Real and Abstract Analysis

Real and Abstract Analysis
A modern treatment of the theory

of functions of a real variable
by
Edwin Hewitt
Professor of Mathematics
The University of Washington
and
Karl Stromberg
Associate Professor of Mathematics
The University of Oregon
With 8 Figures
Springer-Verlag Berlin Heidelberg GmbH

ISBN 978-3-540-78018-2 ISBN 978-3-642-88047-6 (eBook)
DOI 10.1007/978-3-642-88047-6
All rights, especially that of translation into foreign languages, reserved.

It is also forbidden to reproduce this book, either whole or in part, by photomechanical means
(photostat, microfilm and/or microcard) or by other procedure without written permission from
Springer -Verlag.
© by Springer-Verlag Berlin Heidelberg 1965

Ursprilnglich erschienen bei Springer -Verlag Berlin· Heidelberg 1965
Softcover reprint of Ihe hardcover I sl edition 1965
Library of Congress Catalog Card Number 65 - 26609
Title No. 1320

Table of Contents
Preface . . . . . . . . . . . . . . v
Chapter One: Set Theory and Algebra.
Section 1. The algebra of sets. . . 1
Section 2. Relations and functions . 7
Section 3. The axiom of choice and some equivalents 12
Section 4. Cardinal numbers and ordinal numbers . 19
Section 5. Construction of the real and complex number fields 32
Chapter Two: Topology and Continuous Functions 53

Section 6. Topological preliminaries . . . 53
Section 7. Spaces of continuous functions. 81
Chapter Three: The Lebesgue Integral . . . 104

Section 8. The Riemann-Stieltjes integral. 105
Section 9. Extending certain functionals 114
Section 10. Measures and measurable sets . 125
Section 11. Measurable functions. . . . . 148
Section 12. The abstract Lebesgue integral. 164
Chapter Four: Function Spaces and Banach Spaces. 188

Section 13. The spaces ~p(l ~ P < (0) . . . . . 188
Section 14. Abstract Banach spaces. . . . . . . 209
Section 15. The conjugate space of ~p(l < p < (0) 222
Section 16. Abstract Hilbert spaces. 234
Chapter Five: Differentiation . . . . 256

Section 17. Differentiable and nondifferentiable functions 256
Section 18. Absolutely continuous functions . . . . . . 272
Section 19. Complex measures and the LEBESGUE-RADON-NIKODYM theorem 304
Section 20. Applications of the LEBESGUE-RADON-NIKODYM theorem 341
Chapter Six: Integration on Product Spaces . . . . . . 377

Section 21. The product of two measure spaces . . . . 377
Section 22. Products of infinitely many measure spaces 429
Index of Symbols . . . . . 460
Index of Authors and Terms 462

This book is dedicated to
MARSHALL H. STONE
whose precept and example have
taught us both.
Preface
This book is first of all designed as a text for the course usually called
"theory of functions of a real variable". This course is at present cus-
tomarily offered as a first or second year graduate course in United
States universities, although there are signs that this sort of analysis
will soon penetrate upper division undergraduate curricula. We have
included every topic that we think essential for the training of analysts,
and we have also gone down a number of interesting bypaths. We hope
too that the book will be useful as a reference for mature mathematicians
and other scientific workers. Hence we have presented very general and
complete versions of a number of important theorems and constructions.
Since these sophisticated versions may be difficult for the beginner, we
have given elementary avatars of all important theorems, with appro-
priate suggestions for skipping. We have given complete definitions, ex-
planations, and proofs throughout, so that the book should be usable
for individual study as well as for a course text.
Prerequisites for reading the book are the following. The reader is
assumed to know elementary analysis as the subject is set forth, for
example, in TOM M. ApOSTOL'S Mathematical Analysis [Addison-Wesley
Publ. Co., Reading, Mass., 1957], or WALTER RUDIN'S Principles of Mathe-
matical Analysis [2 nd Ed., McGraw-Hill Book Co., New York, 1964].
There are no other prerequisites for reading the book: we define practi-
cally everything else that we use. Some prior acquaintance with abstract
algebra may be helpful. The text A Survey of Modern Algebra, by GARRETT
BIRKHOFF and SAUNDERS MAC LANE [3 rd Ed., MacMillan Co., New York,
1965] contains far more than the reader of this book needs from the
field of algebra.
Modern analysis draws on at least five disciplines. First, to explore
measure theory, and even the structure of the real number system, one
must use powerful machinery from the abstract theory of sets. Second,
as hinted above, algebraic ideas and techniques are illuminating and
sometimes essential in studying problems in analysis. Third, set-theoretic
topology is needed in constructing and studying measures. Fourth, the
theory of topological linear spaces ["functional analysis"] can often be
applied to obtain fundamental results in analysis, with surprisingly little
effort. Finally, analysis really is analysis. We think that handling ine-
qualities, computing with actual functions, and obtaining actual num-
VI Preface
bers, is indispensable to the training of every mathematician. All five of

these subjects thus find a place in our book. To make the book useful
to probabilists, statisticians, physicists, chemists, and engineers, we have
included many "applied" topics: Hermite functions; Fourier series and
integrals, including PLANCHEREL'S theorem and pointwise summability;
the strong law of large numbers; a thorough discussion of complex-
valued measures on the line. Such applications of the abstract theory
are also vital to the pure mathematician who wants to know where his
subject came from and also where it may be going.
With only a few exceptions, everything in the book has been taught
by at least one of us at least once in our real variables courses, at the
Universities of Oregon and Washington. As it stands, however, the book
is undoubtedly too long to be covered in toto in a one-year course. We
offer the following road map for the instructor or individual reader who
wants to get to the center of the subject without pursuing byways, even
interesting ones.
Chapter One. Sections 1 and 2 should be read to establish our
notation. Sections 3, 4, and 5 can be omitted or assigned as outside
reading. What is essential is that the reader should have facility in the
use of cardinal numbers, well ordering, and the real and complex number
fields.
Chapter Two. Section 6 is of course important, but a lecturer should
not succumb to the temptation of spending too much time over it. Many
students using this text will have already learned, or will be in the
process of learning, the elements of topology elsewhere. Readers who
are genuinely pressed for time may omit § 6 and throughout the rest of
the book replace "locally compact Hausdorff space" by "real line" , and
"compact Hausdorff space" by "closed bounded subset of the real line".
We do not recommend this, but it should at least shorten the reading.
We urge everyone to cover § 7 in detail, except possibly for the exercises.
Chapter Three. This chapter is the heart of the book and must be
studied carefully. Few, if any, omissions appear possible. Chapter Three
is essential for all that follows, barring § 14 and most of § 16.
After Chapter Three has been completed, several options are open.
One can go directly to § 21 for a study of product measures and FUBINI'S
theorem. [The applications of FUBINI'S theorem in (21.32) et seq. require
parts of §§ 13-18, however.] Also §§ 17-18 can be studied immediately
after Chapter Three. Finally, of course, one can read §§ 13-22 in order.
Chapter Four. Section 13 should be studied by all readers. Subheads
(13.40)-(13.51) are not used in the sequel, and can be omitted if neces-
sary. Section 14 can also be omitted. [While it is called upon later in
the text, it is not essential for our main theorems.] We believe never-
theless that § 14 is valuable for its own sake as a basic part of functional
Preface VII
analysis. Section 15, which is an exercise in classical analysis, should be

read by everyone who can possibly find the time. We use Theorem
(15.11) in our proof of the LEBESGUE-RADON-NIKODYM theorem [§ 19],
but as the reader will see, one can get by with much less. Readers who
skip § 15 must read § 16 in order to understand § 19.
Chapter Five. Sections 17 and 18 should be studied in detail. They
are parts of classical analysis that every student should learn. Of § 19,
only subheads (19.1)-(19.24) and (19.35)-(19.44) are really essential. Of
§ 20, (20.1)-(20.8) should be studied by all readers. The remainder of
§ 20, while interesting, is peripheral. Note, however, that subheads
(20.55)-(20.59) are needed in the refined study of infinite product
measures presented in § 22.
Chapter Six. Everyone should read (21.1)-(21.27) at the very least.
We hope that most readers will find time to read our presentation of
PLANCHEREL'S theorem (21.31)-(21.53) and of the HARDy-LITTLEWOOD
maximal theorems (21.74)-(21.83). Section 22 is optional. It is essential
for all students of probability and in our opinion, its results are extremely
elegant. However, it can be sacrificed if necessary.
Occasionally we use phrases like "obvious on a little thought", or
"a moment's reflection shows ... ". Such phrases mean really that the
proof is not hard but is clumsy to write out, and we think that more
writing would only confuse the matter. We offer a very large number of
exercises, ranging in difficulty from trivial to all but impossible. The
harder exercises are supplied with hints. Heroic readers may of course
ignore the hints, although we think that every reader will be grateful
for some of them. Diligent work on a fairly large number of exercises is
vital for a genuine mastery of the book: exercises are to a mathematician
what CZERNY is to a pianist.
We owe a great debt to many friends. Prof. KENNETH A. Ross has
read the entire manuscript, pruned many a prolix proof, and uncovered
myriad mistakes. Mr. LEE W. ERLEBACH has read most of the text and
has given us useful suggestions from the student's point of view. Prof.
KEITH L. PHILLIPS compiled the class notes that are the skeleton of the
book, has generously assisted in preparing the typescript for the printer,
and has written the present version of (21.74)-(21.83). Valuable con-
versations and suggestions have been offered by Professors ROBERT M.
BLUMENTHAL, IRVING GLICKSBERG, WILLIAM H. SILLS, DONALD R.TRUAX,
BERTRAM YOOD, and HERBERT S. ZUCKERMAN. Miss BERTHA THOMPSON
has checked the references. The Computing Center of the University
of Oregon and in particular Mr. JAMES H. BJERRING have generously
aided in preparing the index. We are indebted to the several hundred
students who have attended our courses on this subject and who have
suffered, not always in silence, through awkward presentations. We
VIII Preface
are deeply grateful to Mrs. SHANTI THAYIL, who has typed the entire
manuscript with real artistry.
Our thanks are also due to the UniversitieE. of Oregon and Washington
for exemption from other duties and for financial assistance in the pre-
paration of the manuscript. It is a pleasure to acknowledge the great
help given us by Springer-Verlag, in their rapid and meticulous publica-
tion of the work.
Seattle, Washington EDWIN HEWITT
Eugene, Oregon KARL R. STROMBERG
July 1965
CHAPTER ONE
Set Theory and Algebra
From the logician's point of view, mathematics is the theory of sets
and its consequence,>. For the analyst, sets and concepts immediately
definable from sets are essential tools, and manipulation of sets is an
operation he must carry out continually. Accordingly we begin with two
sections on sets and functions, containing few proofs, and intended largely
to fix notation and terminology and to form a review for the reader in
need of one. Sections 3 and 4, on the axiom of choice and infinite arith-
metic, are more serious: they contain detailed proofs and are recommended
for close study by readers unfamiliar with their contents.
Plainly one cannot study real- and complex-valued functions seriously
without knowing what the real and complex number fields are. Therefore,
in § 5, we give a short but complete construction of these objects. This
section may be read, recalled from previous work, or taken on faith.
This text is not rigorous in the sense of proceeding from the axioms
of set theory. We believe in sets, and we believe in the rational numbers.
Beyond that, we have tried to prove all we say.
§ 1. The algebra of sets

(1.1) The concept of a set. As remarked above, we take the notion of set
as being already known. Roughly speaking, a set [collection, assemblage,
aggregate, class, family] is any identifiable collection of objects of any
sort. We identify a set by stating what its members [elements, points]
are. The theory of sets has been described axiomatically in terms of the
notion "member of". To build the complete theory of sets from these
axioms is a long, difficult process, and it is remote from classical analysis,
which is the main subject of the present text. Therefore we shall make
no effort to be rigorous in dealmg with the concept of sets, but will appeal
throughout to intuition and elementary logic. Rigorous treatments of the
theory of sets can be found in Naive Set Theory by P. HALMos [Princeton,
N. J: D. Van Nostrand Co. 1960] and in Axiomatic Set Theory by
P. SUPPES [Princeton, N. J: D. Van Nostrand Co. 1960].
(1.2) Notation. We will usually adhere to the following notational
conventions. Elements of sets will be denoted by small letters: a, b, c,
... , x, y, z; IX, {J, y, ... Sets will be denoted by capital Roman letters:
A, B, C, ... Families of sets will be denoted by capital script letters:
d, !!8, re, ... Occasionally we need to consider collections of families of
sets. These entities will be denoted by capital Cyrillic letters: )1(, ~L ...
Hewitt/Stromberg, Real and abstract analysis 1
2 Chapter 1. Set theory and algebra
A set is often defined by some property of its elements. We will write

{x: P(x)} [where P(x) is some proposition about x] to denote the set of
all x such that P(x) is true. We have done nothing here to sharpen the
definition of a set, since "property" and "set" are from one point of view
synonymous.
If the object x is an element of the set A, we will write x EA; while
x ~ A will mean that x is not in A.
We write 0 for the void [empty, vacuous] set; it has no members at
all. Thus 0 = {x: x is a real number and x2 < O} = {x: x is a unicorn in
the Bronx Zoo}, and so on.
For any object x, {x} will denote the set whose only member is x.
Similarly, {Xl> x 2, ••• , x ...} will denote the set whose members are precisely
Xl' x 2 , ••• , x....
Throughout this text we will adhere to the following notations:
N will denote the set {I, 2, 3, ... } of all positive integers; Z will denote
the set of all integers; Q will denote the set of all rational numbers;
R will denote the set of all real numbers; and K will denote the set of all
complex numbers. We assume a knowledge on the part of the reader of the
sets N, Z, and Q. The sets R and K are constructed in § 5.
(1.3) Definitions. Let A and B be sets such that for all x, x EA
implies x E B. Then A is called a subset 01 B and we write A C B or
B::J A. If A C Band B C A, then we write A = B; A =1= B denies
A = B. If A C B and A =1= B, we say that A is a proper subset 01 B and
we write A <;j; B. We note that under this idea of equality of sets, the void
set is unique, for if 0 1 and O 2 are any two void sets we have 0 1 C O 2
and 0 2 C 0 1 ,
(1.4) Definitions. If A and B are sets, then we define A U B as the set
{x: x EA or x EB}, and we call A U B the union 01 A and B. Let d be a
family of sets; then we define U d = {x: x EA for some A Ed}. Similarly
if {A'}'EI is a family of sets indexed by iota, we write .EI U A. = {x: x EA,
for some LEI}. If I = N, the positive integers, U A ... will usually be
00 00
"EN
written as U A .... Other notations, such as U A ... , are self-explanatory.
n=l n=-oo
For given sets A and B, we define A n B as the set {x: x EA and
x E B}, and we call A n B the intersection 01 A and B. If d is any family
of sets, we define n d = {x: x EA for all A Ed}; if {A'}'EI is a family of
sets indexed by iota, then we write n A, = {x: x EA, for all LEI}. The
00
'EI
notation n A ... [and similar notations] have obvious meanings .
.. =1
Example. If A ... = {x: x is a real number, Ixl < ~}, n = 1,2, 3,
0 A ... =
00
then .. 1 {O}.
§ 1. The algebra of sets 3
For a set A , the family of all subsets of A is a well-defined family of

sets which is known as the power set 0/ A and is denoted by .9'(A). For
example, if A = {I, 2}, then .9'(A) = {0, {I}, {2}, {I, 2}}.
(1.5) Theorem. Let A, B, C be any sets. Then we have:
(i) AU B = B U A; (i') A n B = B n A;
(ii) A U A = A; (ii') A n A = A;
(iii) AU 0 = A; (iii') An 0 = 0;
(iv) AU (B U C) (iv') A n (B n C)
= (A U B) U C; = (A n B) n C;
(v) A cA U B; (v') An Be A;
(vi) A c B if and only if (vi') A c B if and only if
AU B = B; An B=A.
The proof of this theorem is very simple and is left to the reader.
(1.6) Theorem.
(i) A n (B U C) = (A n B) U (A n C);
(ii) A U (B n C) = (A U B) n (A U C) .
Proof. These and similar identities may be verified schematically; the
verification of (i) follows:
B C 8
An (D UC) (A na) u (A nt:)

Fig. 1
A similar schematic procedure could be applied to (ii). However, we may

use (i) and the previous laws as follows:
(A U B) n (A U C) = «(A U B) n A) U «(A U B) n C)
= (A n A) U (B n A) U (A n C) U (B n C) = A U (B n C)
U (B n A) U (A n C) = A U (B n C);
the last equality holds because B n A C A and A n C cA. 0 1
(1.7) Definition. If A n B = 0, then A and B are said to be disjoint.
If d is a family of sets such that each pair of distinct members of dare
disjoint, then d is said to be pairwise disjoint. Thus an indexed family
{A'}'Ei is pairwise disjoint if A. n A7J= 0 whenever t =l= 'f}.
(1.8) Definition. In most of our ensuing discussions the sets in question
will be subsets of some fixed "universal" set X. Thus if A C X, we define
1 The symbol 0 will be used throughout the text to indicate the end of a proof.
1*
the complement of A [relative to XJ to be the set {x : x EX, x ~ A}. This

set is denoted by the symbol A'. If there is any possible ambiguity as to
which set is the universal set, we will write X n A' for A'. Other common
notations for what we call A' are X - A, X \ A, X ~ A, C A, and A c;
we will use A' exclusively.
(1.9) Theorem [DE MORGA:'I'S laws].
(i) (A U B)' = A' n B';
(ii) (A n B)' = A'U B';
(iii) (U A,)' = n A;;
' EI ' EI
(iv) ( n A,),= t UE l
t El
A;.
The proofs of these identities are easy and are left to the reader.
(1.10) Definition. For sets A and B, the symmetric difference of A and
B is the set {x: x EA or x EB and x ~ A n B}, and we write A ,6, B for
this set. Note that A ,6, B is the set consisting of those points which are
in exactly one of A and B, and that it may also be defined by A ,6, B
= (A n B') U (A' n B). The symmetric difference
is sketched in Fig. 2
(1.11) Definition. Let X be a set and let !Ji
8 be a nonvoid family of subsets of X such that
A
(i) A, B EBf implies A U B E &£;
(ii) A, B EBf implies A n B' E!Ji .
A/J.8 Then PA is called a ring of sets. A ring of sets
Fig. 2 closed under complementation [i. e. A Efj£ implies
A' E!JiJ is called an algebra of sets.
(1.12) Remarks. A ring of sets is closed under the formation of finite
intersections; for, if A, B EBf, then (1.11.ii) applied twice shows that
An B = A n (A n B')'E!Ji. By (1.11.i) and (l.1l.ii), we have A,6,B
= (A U B) n (A n B)' EPA. Note also that 0 E!Ji since Bf is nonvoid.
Also !Ji is an algebra if and only if X EBf. There are rings of sets which
are not algebras of sets; e.g., the family of all finite subsets of N is a ring
of sets but not an algebra of sets.
(1.13) Definition. A a-ring [a-algebra J of sets is a ring [algebra J of
00
sets Bf such that if {An: n EN} CBf, then U An EBf.

n=1
Much of measure theory deals with families of sets which form a-rings
or a-algebras. There are a-rings which are not a-algebras, e.g., the family
of all countable subsets of an uncountable set. [For the definitions of
countable and uncountable, see § 4.J
(1.14) Remarks.1 There are many axiomatic treatments of rings and
algebras of sets, and in fact some very curious entities can be interpreted
1 This subhead is included only for its cultural interest and may be omitted by
anyone who is in a hurry.
as rings or algebras of sets [see (1.25)]. Let B be any set. Suppose that
to each a EB there is assigned a unique element a* EB and that to each
pair of elements a, b EB there is assigned a unique element a V b EB
such that these operations satisfy
(i) a V b = b Va,
(ii) a V (b V c) = (a V b) V c,
(iii) (a*V b*)*V (a*V b)*= a.
Sets B with operations V and * [or similar operations J and satisfying
axioms equivalent to (i) -(iii) were studied by many writers in the period
1890-1930. They bear the generic name Boolean algebras, after the
English mathematician GEORGE BOOLE [1815-1864]. The axioms
(i) -(iii) were given by the U. S. mathematician E. V. HUNTINGTON
[1874-1952J [Trans. Amer. Math. Soc. 5, 288-309 (1904)].
The reader will observe that if a, b are interpreted as sets and V
and * as union and complementation, then (i) -(iii) are simple identities.
Other operations can be defined in a Boolean algebra, e.g., /\ [the analogue
of n for setsJ, which is defined by a /\ b = (a* V b*)*. A great deal of
effort has been devoted to investigating abstract Boolean algebras. In the
1930's, the contemporary U. S. mathematician M. H. STONE showed that
any Boolean algebra can be interpreted as an algebra of sets in the follow-
ing very precise way [Trans. Amer. Math. Soc. 40, 37 -111 (1936)]. Given
any Boolean algebra B, there is a set X, an algebra ~ of subsets of X,
and a one-to-one mapping -r of B onto ~ such that -r(a*) = (-r(a))'
[* becomes 'J and -r(a V b) = -r(a) U -r(b) [V becomes U]. Thus from the
point of view of studying the operations in a Boolean algebra, one may
as well study only algebras of sets.
STONE'S treatment of the representation of Boolean algebras was
based on a slightly different entity, namely, a Boolean ring. A Boolean
ring is any ring 5 such that x 2 = x for each xES. [For the definition of
ring, see (5.3).J
STONE showed that Boolean algebras and Boolean rings having a mul-
tiplicative unit can be identified, and then based his treatment on
Boolean rings. More precisely: for every Boolean ring 5, there is a ring
of sets ~ and a one-to-one mapping -r of 5 onto ~ such that
-r(a + b) = -r(a) £:" -r(b)
and
-r(ab) = -r(a) n -r(b) .
That is, addition in a Boolean ring corresponds to the symmetric differ-
ence, and multiplication to intersection.
Proofs of the above results and a lengthy treatment of Boolean al-
gebras and rings and of algebras and rings of sets can be found in
G. BIRKHOFF, Lattice Theory [Amer. Math. Soc. Colloquium Publications,
Vol. XXV, 2nd edition; Amer. Math. Soc., New York, N. Y., 1948].
(1.15) Exercise. Simplify as much as possible:

(a) (A U (B n (C U W')))' ;
(b) «(X' U Y) n (X U Y'))' ;
(c) (A n B n C) U (A' n B n C) U (A n B' n C) U (A n B n C')
U (A n B' n C') U (A' n B n C') U (A' n B' n C) .
(1.16) Exercise [PORETSKY]. Given two sets X and Y, prove that
X = 0 if and only if Y = X 6. Y.
(1.17) Exercise. Describe in words the sets ngltOnA /<) andnO l tgnA/<)
where {AI> A 2 , ••• , A/<, ... } is any family of sets indexed by N. Also
prove that the first set is a subset of the second.
(1.18) Exercise. Prove:
(a) A 6. (B 6. C) = (A 6. B) 6. C ;
(b) A n (B 6. C) = (A n B) 6. (A n C) ;
(c) A 6. A = 0;
(d) 06. A = A .
(1.19) Exercise. Let {A,},EI and {B,},EI be nonvoid families of sets.
Prove that
(i) (U, A,) 6. (U, B,) C U , (A,6. B,) .
Prove by an example that the inclusion may be proper. Can you assert
anything about (i) if the U's are changed to n's?
(1.20) Exercise. For any sets A, B, and C, prove that
A 6. B C (A 6. C) U (B 6. C) ,
and show by an example that the inclusion may be proper.
(1.21) Exercise. Let {Mn}:=1 and {Nn}:=l be families of sets such
that the sets N n are pairwise disjoint. Define Ql = Ml and Qn =
n
Mnn (M1 U···UM.. _1 )'forn=2,3, ... Prove that N n6. QnC U (N/<6.M/<)
"=1
(n = 1, 2, ... ).
(1.22) Exercise. Consider an alphabet with a finite number of letters,
say a, where a > 1. A word in this alphabet is a finite sequence of letter~,
not necessarily distinct. Two words are equal if and only if they have the
same number of letters and if the letters are the same and in the same order.
Consider all words of length l, where l > 1. How many words of length l
have at least two repetitions of a fixed letter? How many have three such
repetitions? In how many words of length l do there occur two specified
distinct letters?
(1.23) Exercise.
(a) Let A be a finite set, and let v(A) denote the number of elements
of A : thus v (A) is a nonnegative integer. Prove that
v(A U B) = v(A) + v(B) - v(A n B) .
(b) Generalize this identity to v(A U B U C) and to v(A U B U CUD).

(c) A university registrar reported that the total enrollment in his
university was 10,000 students. Of these, he stated, 2521 were married,
6471 were men, 3115 were over 21 years of age, 1915 were married men,
1873 were married persons over 21 years of age, and 1302 were married
men over 21 years of age. Could this have been the case?
(d) Help the registrar. For a student body of 10,000 members, find
positive integers for the categories listed in (c) that are consistent with
the identity you found in (b).
(1.24) Exercise. Prove that in any Boolean ring we have the iden-
tities
(a) x + x = 0;
(b) xy = yx.
(1.25) Exercise. (a) Let B be the set of all positive integers that
divide 30. For x, y EB, let x V y be the least common multiple of x andy,
and let x* = 30.
x
Prove that B is a Boolean algebra. Find an algebra of
sets that represents B as in (1.14).
(b) Generalize (a), replacing 30 by any square-free positive integer.
(c) Generalize (b) by considering the set B of all square-free positive
integers, defining x V y as the least common multiple of x and y, x 1\ y
as the greatest common divisor of x and y, and x t::. y as XVI\Y • Show
x Y
that B can be represented as a certain ring of sets but not as an algebra
of sets.
§ 2. Relations and functions
In this section we take up the concepts of relation and function,
familiar in several forms from elementary analysis. We adopt the currently
popular point of view that relations and functions are indistinguishable
from their graphs, i. e., they are sets of ordered pairs. As in the case of
sets, we content ourselves with a highly informal discussion of the subject.
(2.1) Definition. Let X and Y be sets. The Cartesian product 01 X
and Y is the set X x Y of all ordered pairs (x, y) such that x EX and
yEY.
We write (x, y) = (u, v) if and only if x = u and y = v. Thus (1, 2)
9= (2, 1) while {I, 2} = {2, I}.
(2.2) Definition. A relation is any set of ordered pairs. Thus a relation
is any set which is a subset of the Cartesian product of two sets. Observe
that 0 is a relation.
(2.3) Definitions. Let I be any relation. We define the domain 01 I to
be the set doml = {x : (x, y) EI for some y} and we define the range 01 I
to be the set mgl = {y: (x, y) EI for some x}. The symbol 1-1 denotes the
inverse of I: 1-1 = {(y, x) : (x, y) EI}.
(2.4) Definition. Let / and g be relations. We define the composition

[product, iterate are also used] 0/ / and g to be the relation go/ = {(x, z) :
for some y, (x, y) E/ and (y, z) Eg}.
The composition of / and g may be void. In fact., go/ =F 0 if and
only if (mg/) n (domg) =F 0.
(2.5) Definition. Let / and g be relations such that / c g. Then we say
that g is an extension 0/ / and that / is a restriction 0/ g.
We now discuss some special kinds of relations that are needed in the
sequel. Wherever convenient, we will use the conventional notation x/y
to mean that (x, y) E/.
(2.6) Definition. Let X be a set. An equivalence relation on X is any
relation ~ c X x X such that, for all x, y, z in X we have:
(i) x ~ x [reflexive];
(ii) x ~ y implies y ~ x [symmetric];
(iii) x ~ y and y ~ z imply x ~ z [transitive].
(2.7) Definitions. Let P be a set. A partial ordering on P is any
relation ~ c P x P satisfying
(i) x ~ x [reflexive];
(ii) x ~ y and y ~ x imply x = y [antisymmetric];
(iii) x ~ y and y ~ z imply x ~ z [transitive].
If ~ also satisfies
(iv) x, yEP implies x :;;;; y or y :;;;; x [trichotomy],
then ~ is called a linear [also called simple, complete, or total] ordering on P.
If x ~ y and x =F y, we write x < y. The expression x ~ y means y ;s x
and x > y meansy < x.
If ;::::: is a linear ordering such that
(v) 0 =F A c P implies there exists an element a EA such that
a ;::::: x for each x EA [a is the smallest element 0/ A],
then;::::: is called a well ordering on P.
A partially ordered set is an ordered pair (P, ;:::::) where P is a set and :;;;;
is a partial ordering on P. If :;;;; is a linear ordering, (P, :;;;;) is called a
linearly [simply, completely, totally] ordered set. If ;::::: is a well ordering,
then (P, ;:::::) is called a well-ordered set.
Let P be a linearly ordered set. For x, YEP, we define max {x, y} = y
if x;::::: y, and max{x,y} = x ify;::::: x. For a finite subset {Xl> x 2 , . • . , x n }
of P [not all x;'s necessarily distinct], we define max {Xl> x 2 , . . . , x n } as
max{xn' max {Xl> x 2 , ••• , xn- 1 }}. The expressions min {x, y} and
min {Xl' x 2, ••• , X n } are defined analogously.
(2.8) Examples. (a) Let § be any family of sets. Then set inclusion C
is a partial ordering on § and (§, c) is a partially ordered set. For short
we say that § is partially ordered by C. The reader should note that,
depending on §, this relation may fail to be a linear ordering; for example,
take § = &P({O, I}).
§ 2. Relations and functions 9
(b) Let P be the set of all nonnegative rational numbers:

P = {x: x EQ, x ~ O},
and let ~ be the usual ordering on P. Then ~ is a linear ordering on P
and P has a smallest element 0, but P, with this ordering, is not a well-
ordered set, since there are non void subsets of P containing no smallest
element. For example, let A = {x E P: x =1= O}. Then ; E A whenever
x E A, so A contains no smallest element.
(c) The set N of all positive integers with its usual ordering is a linearly
ordered set. It is also a well-ordered set. This last assertion is equivalent
to PEANO'S axiom of mathematical induction.
(2.9) Definition. Let I be a relation and let A be a set. We define the
image 01 A under I to be the set
I (A) = {y : (x, y) E I for some x E A} .
Observe that I(A) =1= 0 if and only if A n doml =1= 0. The inverse image
of A under I is the set 1-1 (A).
(2.10) Definition. A relation I is said to be single-valued if (x, y) E I
and (x, z) E I imply y = z. If I and 1-1 are both single-valued, then I is
called a one-to-one relation. The definitions of many-to-one, one-to-many,
and many-to-many relations are analogous.
Single-valued relations play such an important role in analysis that
we make the following definition.
(2.11) Definition. A single-valued relation is called a lunction
[mapping, translormation, operation, correspondence, application].
(2.12) Examples. The sine function, {(x, sinx) : x E R} is many-to-one.
The inverse of this function, {(sinx, x) : x E R}, is a one-to-many relation.
The relation {(x,y): x,y ER, x 2 + y2= I} is a many-to-many relation.
The function {(x, tanx) : x E R, - ; < x < ;} is a one-to-one function.
(2.13) Definition. Let I be a function and let X and Y denote the
domain and range of I, respectively. For x E X, let I (x) denote that unique
element of Y such that (x, I(x) EI. The element I (x) is called the value
011 at x or the image 01 x under I.
Note that in order to specify a function completely, it is sufficient to
specify the domain of the function and the value of the function at each
point of its domain.
(2.14) Remark. Referring to (2.9), we observe that if I is a function
and A is a set, then I(A) = {/(x) : x EA n dom/} and 1-1 (A) = {x: x Edomf,
f (x) E A}. The reader should verify these statements.
(2.15) Theorem. Let X and Y be sets and let f C X x Y be a relation.
Suppose that {A.},EJ is a family 01 s1tbsets 01 X and that {B.},EJ is a lamily
01 subsets 01 Y. For A e X we write A' lor the complement 01 A relative

to X and lor BeY we write B' lor the complement 01 B relative to Y. Then
(i) I(~A,) = '~If(A,);
(ii) f(n A,) e n f(A,).
'El 'El
The followings results are true il f is a function, but may lail for arbitrary
relations:
(iii) f-I(n B) = n f-I(B)'
'EI' 'EI ' ,
(iv) 1-1 (B') = (t-I(B)';
(v) l(t-I(B) n A) = B n I(A).
The proof of this theorem is left to the reader.
(2.16) Remark. From Theorem (2.15) it follows that the domain and
range of a one-to-one function cannot be distinguished from each other
by any purely set-theoretic properties. If X and Yare sets for which
there is a one-to-one function I with domain X and range Y, then for any
subset A of X we have I(A/) = I(A)/. For any family {A,LEr of subsets
of X, we have I( U A,) = U I(A,) and f( n A,) = n I(A,). Similar
'EI 'EI 'EI 'EI
statements hold for subsets of Y and 1-1. Thus, all Boolean operations
(U, n, 6" ') are preserved under I and 1-1.
(2.17) Definition. Let I be a function such that doml = X and
mgl e Y. Then I is said to be a function Irom [onJ X into [toJ Y and we
write f: X --* Y. If mgf = Y, we say that I is onto Y.
(2.18) Definition. A sequence is a function having N, the set of all
positive integers, as its domain. If x is a sequence, we will frequently
write Xn instead of x(n) for the value of x at n. The value Xn is called the
nth term of the sequence. The sequence x whose nth term is Xn will be
denoted by (xn):= I or simply (xn). A sequence (xn) is said to be in X if
Xn EX for each n EN; we abuse our notation to write (xn) e X.
The following theorem will be used several times in the sequel.
(2.19) Theorem. Let 5' be any family 01 functions such that I, g E5'
implies either f e g or gel, i. e., 5' is linearly ordered relative to C. Let
h = U5'. Then:
(i) h is a lunction;
(ii) domh = U{dom/: I E5'};
(iii) x Edomh implies h (x) = I (x) lor each IE 5' such that x Edom/;
(iv) mgh = U{mg/: I E5'}.
Proof. (i) Obviously h is a relation since it is a union of sets of ordered
pairs. We need only show that h is single-valued. Let (x, y) Ehand
(x, z) Eh. Then there exist I and g in 5' such that (x, y) Ef and (x, z) Eg.
We know that leg or gel; say feg. Then (x,y) Eg and (x,z) Eg.
Since g is a function we have y = z. Thus h is a function.
§ 2. Relations and functions 11
The equality (ii) is true because the following statements are pairwise
equivalent: x Edomh; (x, y) Eh for some y; (x, y) Ef for some I Etr;
x Edoml for some I Etr.
Let x Edomh n doml = doml where IE tr. Then (x, I (x)) EIe h
and h is single-valued so hex) = I(x). This proves (iii).
The equality (iv) follows from the previous conclusions and (2.15.i)
since rngh = h (domh) = h (U{dom/: IE tr}) = U {h (dom/): IE tr}
= U{/(dom/): I Etr} = U{rng/: I Etr}. 0
(2.20) Definition. Let X be any set and E any subset of X. The
function ~E with domain X and range contained in {O, I} such that
~E (x) =
{
°lifXEE'
if x EX n E' ,
is called the characteristic lunction 01 E. It will always be clear from the

context what the domain of ~E is. Characteristic functions are very use-
ful in analysis, and will be encountered frequently throughout this text.
One particular characteristic function is used so much that it has a
special symbol. The diagonal Dol X x X is defined as D = {(x, x) : x EX}.
The value of the characteristic function of D at (x, y) is written 15"'11 and
is called KRONECKER'S !5-symbol. Thus 15"'11 = 1 if x = y and 15"'11 = if
x =1= y; here x and yare arbitrary points in X.
°
(2.21) Exercise. Prove that I 0 (g 0 h) = (log) 0 h for all relations I,
g, andh.
(2.22) Exercise. Show that the equality f(f-l(B) n A) = B n I(A)
fails for every relation I that is not a function.
(2.23) Exercise. For (a, b) and (c, d) in N x N, define (a, b) ~ (c, d)
if either: a < c, or a = c and b ~ d. Prove that, with this relation, N x N
is a well-ordered set.
(2.24) Exercise. Let n be a positive integer and let p .. = {k EN:
kis a divisor ofn}. For a, b EP.. define a;t{. b to mean that a is a divisor of b,
i.e., alb.
(a) Prove that p .. , with ;t{., is a partially ordered set.
(b) Find necessary and sufficient conditions on n that p .. be a linearly
ordered set.
(2.25) Exercise. Let X be a set with a binary operation p defined on
it, i. e., p is a function from X x X into X. Write p (x, y) = xy. Suppose
that this operation satisfies
(i) x(yz) = (xy)z,
(ii) xy = yx,
(iii) xx = x,
for all x, y, z in X. Define ~ on X by x ~ y if and only if xy = y. Prove
that: (a) X is a partially ordered set; (b) each pair of elements of X has
a least upper bound, i.e., if x, y EX, then there exists z EX such that
x ~ z, y ~ z, and if x ~ w, y ~ w, then z ~ w.
(2.26) Exercise. Let 1be a function from X to Y. Suppose that there
is a function g from Y to X such that log (y) = y for all y E Y and
g 01 (x) = x for all x EX. Prove that I is a one-to-one function from X
onto Y and that g = 1-1.
§ 3. The axiom of choice and some equivalents

In the study of algebra, analysis, and topology one frequently encoun-
ters situations in which the tools of elementary set theory [as they have
been informally presented in §§ 1 and 2] are too weak to permit construc-
tions, proofs, or even definitions that one may need. In the early 1900's
the German mathematician ERNST ZERMELO propounded an innocent-
appearing but actually very strong axiom, called the axiom of choice
[Auswahlpostulat], which has many important consequences, and which
has also excited vigorous controversy. In this section we take up the
axiom of choice, establish the equivalence of four other assertions with it,
and point out two important applications. Other applications of the
axiom of choice will appear throughout the book.
(3.1) Definition. Let {A,},u be any family of sets. The Cartesian
product of this family, written X A" is the set of all functions x having
'EI
domain I such that x, = x (t) E A, for each t EI. Each such function x is
called a choice lunction for the family {A,},u. For x E X A, and t E I,
'EI
the value x, EA, is known as the ttlt coordinate 01 x.
One may ask if there are any choice functions for a given family of
sets. Of course if I = 0, then the void function 0 is a choice function for
any family indexed by I. If I =1= 0 and A, = 0 for some t EI, then
X A, = 0. These two special cases are of little interest. In general the
'El
question cannot be answered on the basis of the usual axioms of set
theory. We will use the following axiom.
(3.2) Axiom of choice. The Cartesian product of any non void family
of nonvoid sets is a nonvoid set, i. e.,if{A,},EI is a family of sets such that
I =1= 0 and A, =1= 0 for each £ EI, then there exists at least one choice
function for the family {A,},u.
P. J. COHEN has recently proved that this axiom is independent of
the other axioms of set theory [Proc. Nat. Acad. Sci. U.S.A. 50, 1143 -
1148 (1963); 51, 105-110 (1964)J.
(3.3) Definition. Let A and I be sets. We define A I to be the Cartesian
product X A" where A, = A for each £ EI. Thus AI is the set of all
'EI
functions 1 such that doml = I and mgl cA. If, for some n EN, I is
§ 3. The axiom of choice and some equivalents 13
the set {I, ... , n}, then we write A I = An. Some authors write
A oo for AN.
A typical member of An is, to be sure, a function and as such is a set
of n ordered pairs. We follow conventional notation, however, and list the
values of such a function as an ordered n-tuple. Thus An= {(a v ... , an)
: ak E A for k = 1, ... , n}. Similarly AN = {(a v a 2 , • • • ) : ak E A for
kEN}. The set Rn is called Euclidean n-space and Kn is called unitary
n-space.
°
(3.4) Example. Let A = {O, I}. Then AN is the set of all sequences
a = (a v a 2 , ••• , an' ... ) where each an is or 1. In many ways this set
resembles CANTOR'S ternary set P = [0, 1] n (] !, ! [ u ] ~, ~[
U] ~, : [ u .. .y [see (6.62) infra]. The mapping cp defined by
cp (a) = 2 t ;: is a one-to-one mapping from
n=!
AN onto P. Anticipating
future developments, we remark that AN can be made a metric space by
introducing the metric e, where e(a, b) =~ if tlt = bI , a2 = b2 ,
°
••• ,
an - I = bn - I , and an =!= bn ; and e(a, b) = if a = b. Under this metric on

AN, cp and cp-I are both continuous. The set AN becomes an Abelian
group under the operation + defined by (a + b)n = an + bn (mod 2) for
n EN. [There are many other ways to make AN into an Abelian group.]
(3.5) Definitions. Let (P, ~) be any partially ordered set and let
A C P. An element u E P is called an upper bound for A if x ~ u for each
x EA. An element mE P is called a maximal element of P if x E P and
m ~ x implies m = x. Similarly we define lower bound and minimal
element,1 A chain in P is any subset C of P such that C is linearly ordered
under the given order relation ~ on P.
This terminology of partially ordered sets will often be applied to an
arbitrary family of sets. When this is done, it should be understood that
the family is being regarded as a partially ordered set under the relation
C of set inclusion. Thus a maximal member of .PI is a set M E.PI such that
M is a proper subset of no other member of .PI and a chain of sets is a
family flB of sets such that A c B or B C A whenever A, BE flB.
(3.6) Definition. Lei $ be a family of sets. Then $ is said to be a
family of finite character if for each set A we have A E§ if and only if
each finite 2 subset of A is in $.
We shall need the following technical fact.
1 We agree that every element of P is both an upper bound and a lower bound
for the void set 0 ; but naturally 0 contains neither a maximal nor a minimal element.
2 A set F is said to be finite if either F = 0 or there exist n EN and a one-to-one
function from {I, 2, ... , n} onto F. See (4.12).
14 Chapter I. Set theory and algebra
(3.7) Lemma. Let ~ be a family of finite character and let 1B be a chain

in~. Then U1B E~.
Proof. It suffices to show that each finite subset of U 1B is in ~. Let
F = {Xl> ... ' xn} C U1B. Then there exist sets Bl> ... ' Bn in 1B such that
xi E Bi (j = 1, ... , n). Since 1B is a chain there is a jo E{I, ... , n} such
that B; C B;. for eachj = 1, ... , n. ThenF C Bi.E~. But ~ is of finite
character, and so F E~. 0
There are many problems in set theory, algebra, and analysis to which
the axiom of choice in the form (3.2) is not immediately applicable, but
to which one or another equivalent axiom is applicable at once. We next
list four such statements. The names "lemma" and "theorem" are at-
tached to them only for historical reasons, as they are all equivalent to
Axiom (3.2).
(3.8) TUKEY's Lemma. Every nonvoid family of finite character has
a maximal member.
(3.9) HausdorffMaximality Principle. Every nonvoid partially ordered
set contains a maximal chain.
(3.10) ZORN'S Lemma. Every nonvoid partially ordered set in which
each chain has an upper bound has a maximal element.
(3.11) Well-ordering Theorem [ZERl\iELO]. Every set can be well
ordered; i.e., if S is a set, then there exists some well-ordering ~ on S.
(3.12) Theorem. The following five propositions are pairwise equiv-
alent:
(i) The axiom of choice;
(ii) TUKEY's lemma;
(iii) The Hausdorff maximality principle;
(iv) ZORN'S lemma;
(v) The well-ordering theorem.
Proof. We will prove this theorem by showing successively that (i)
implies (ii), (ii) implies (iii), (iii) implies (iv), (iv) implies (v), and finally
that (v) implies (i). The most difficult of these five proofs is the first.
Suppose that (i) is true and assume that (ii) is false. Then there exists
a nonvoid family ~ of finite character having no maximal member. For
each F E~,let d'F={E E~:F C,f E}. Then {d'F:F E~} is a nonvoid
family of nonvoid sets, so by (i) there is a function f defined on ~ such
thatf(F) Ed'F for eachF E~. Thus we haveF C,f f(F) E~foreachFE~.
A subfamily J of ~ will be called f-inductive if it has the following
three properties:
(1) 0 EJ;
(2) A EJ implies f(A) EJ;
(3) 1B a chain cJ implies U1B EJ.
Since ~ is nonvoid, since 0 is finite, and since (3.7) holds, the family ~
is f-inductive. Let -Fo= n {J: J is f-inductive} = {A E ~: A EJ for
every I-inductive family.J'}. It is easy to see that .J'o is I-inductive. Thus

.J'o is the smallest I-inductive family, so any I-inductive family contained
in.J'o must be .J'o. We will make heavy use of this fact in proving that.J'o
is a chain.
Let.Yl' = {A E.J'o: B E.J'o and Bs;: A imply I (B) c A}. We assert that
if A E.Yl' and C E.J'o, then either C C A or I (A) c C. To prove this asser-
tion, let A E.Yl' and define ~.A = {C E.J'o: C C A or I (A) c C}. It suffices
to show that ~.A is I-inductive. Since 0 E.J'o and 0 C A, (1) is satisfied.
s;:
Let C E ~.A. Then we have either C A, C = A, or I(A) C C. If C A, s;:
then I(C) C A because A E.Yl'. If C = A, then I(A) C I(C). If I(A) C C,
then I(A) C I(C) because C C I(C). Thus in every case I(C) E~.A and
(2) is satisfied. Next let fJI be a chain in ~.A. Then either C C A for each
C E fJI, in which case U fJI C A, or there exists aCE fJI such that
I(A) C C C UfJI. Thus UfJI E~.A and (3) is satisfied. We conclude that
~.A is I-inductive and so ~.A = .J'o·
We next assert that .Yl' = .J'o. We prove this by showing that .Yl'is
I-inductive. Since 0 has no proper subset, .Yl' satisfies (1) vacuously.
Next let A E.Yl' and B E.J'o be such that B ~ leA). Since B E.J'o = ~.A'
we have B C A [the inclusion I (A) C B being impossible]. If B A, thes;:
definition of .Yl' yields I(B) C A C I(A). If B = A, then I(B) C I(A).
In either case, the inclusion I(B) C I (A) obtains, so I(A) E.Yl' and (2) holds
for.Yl'. Next, let fJI be a chain in.Yl' and let B E.J'o have the property that
B ~ UfJI. Since B E.J'o =~.A for each A EfJI, we have either B C A
for some A E fJI or I(A) C B for every A E fJI. If the latter alternative
were true, we would have
B ~ U fJI C U {I (A) : A E fJI} C B ,
which is impossible. Thus there is some A E fJI such that B C A. If B

then, since A E.Yl', we have I(B) C A C UfJI. If B = A, then BE.Yl' and
A,s:
U fJI E.J'o = ~B. This implies that I(B) C U fJI [U fJlc B being impossible].
Thus in either case, we have I (B) C U fJI and so U fJI E.Yl'. This proves
that .Yl'satisfies (3). Therefore .Yl' is I-inductive and .Yl' = .J'o.
We conclude from the above arguments that if A E.J'o =.Yl' and
B E.J'o = ~.A' then either B C A or A C I(A) C B. Accordingly.J'o is a
chain. Let M = U.J'o' Since .J'o is I-inductive, (3) implies that M E.J'o.
Applying (2), we have U.J'o = M ~ I(M) E.J'o. This contradiction
establishes the fact that (i) implies (ii).
We next show that (ii) implies (iii). Let (P, ~) be any nonvoid par-
tially ordered set. We want to show that P contains a maximal chain.
This follows at once from TUKEY's lemma since the family rc of all chains
in P is a nonvoid family of finite character [0 E rc and {x} E rc for each
x EP].
To show that (iii) implies (iv), lei (P, ~) be any nonvoid partially
ordered set in which each chain has an upper bound. By (iii) there is a
maximal chain Me P. Let m be an upper bound for M. Then m is a
maximal element of P, for if there is an x E P such that m ~ x and
m =1= x, then M U {x} is a chain which properly includes M, contradicting
the maximality of M.
To prove that (iv) implies (v), let 5 be any nonvoid set and let :!l'
denote the family of all well-ordered sets (W, ~) such that We S. For
example, ({x}, {(x, x)}) E:!l' for each xES. We next introduce an ordering
on :!l' by defining (WI' ~I) ~ (W2' ~ 2) to mean that either WI = W 2 and
~I = ~2 or there exists a EW 2 such that WI = {x EW 2 : x ~2 a, x =1= a}
and ~2 agrees with ~I on WI> i.e., ~I C ~2' We say that (W2' ~2) is a
continuation 0/ (WI' ~I)' The reader should see without difficulty that
~ is a partial ordering on :!l'.
Let us show that ZORN'S lemma can be applied to the partially ordered
set (:!l', ~). Let Cfi = {(W" ~J}tEI be any nonvoid chain (relative to ~)
in :!l'. Set W = U W, and ~ = U ~ t [recall that each ~ t is a set of
tEl tEl
ordered pairs]. We leave it to the reader to prove that ~ is a linear
ordering on W. Let A be a nonvoid subset of W. There exists lEI such
that A n W, =1= 0. Since (w" ~ t) is a well-ordered set, there is an element
a EA n W, such that a ~ t X for each x EAnn;. Assume that there is an
element b EA such that b ~ a. Then b EW, and b ~ t a, so b = a. Thus A
has a smallest element a in (W, ~). We conclude that (W, ~) E:!l' and
is an upper bound for Cfi.
By ZORN'S lemma, :!l' has a maximal element (Wo' ~o). If Wo = 5,
then ~ 0 is a well-ordering for 5 and we are through. Assume that Wo =1= S.
Let zEsnwh. Define ~=~oU{(x,z):xEWoU{z}} on WoU{z},
i.e., we place z after everything in WOo Then (Wo U {z}, ~) E:!l'. This
contradicts the maximality of (Wo' ~o), and so we have proved that
Wo= S.
It remains only to show that (v) implies (i). Let {At}tEI be any non-
void family of nonvoid sets. Let 5 = U At. Let ~ be a well-ordering
tEl
for S. For each lEI, let / (l) be the smallest member of At relative to the
well-ordering ~. Then / is a choice function for the family {At}tEI' 0
It is frequently useful to make definitions or carry out constructions
by well ordering a certain set Wand making the definition or construction
at a EW depend upon what has been defined or done at all of the prede-
cessors of a in the well-ordering. The general form of this process is
described in (3.13) and (3.14) below.
(3.13) Definition. Let (W, ~) be a well-ordered set and let a EW.
The set I(a) = {x EW: x ~ a, x =1= a} is called the initial segment 0/ W
determined by a.
(3.14) Theorem [Principle of Transfinite Induction]. Let (W, ~) be a

well-ordered set and let A C W be such that a EA whenever I(a) cA. Then
A=W.
Proof. Assume that W n A' =l= 0 and let a be the smallest member of
W n A'. Then we have I(a) C A, so a EA. But a EW n A'. 0
The axiom of choice does not perhaps playa central role in analysis,
but when it is needed, it is needed most urgently. We shall encounter
several such situations in our subsequent study of measure theory and
linear functionals. To give an immediate and important application of the
axiom of choice, we will prove from TUKEY's lemma that every vector
space contains a basis. Exact definitions follow.
(3.15) Definition. A vector space [linear space] is an ordered triple
(X, " F) where X is an additive Abelian group 1, F is a field, and· is a
function from F x X into X, whose value at (IX, x) is denoted lXX, such
that for IX, (3 EF and x, y EX we have
(i) IX (x + y) = IXX + lXy;
(ii) (IX + (3)x = IXX + (3x;
(iii) 1X({3X) = (IX (3)x;
(iv) Ix = x, where 1 is the multiplicative identity of F.
The members of X are called vectors and the members of F are called
scalars. The operation' is called scalar multiplication. For short we say
that X is a vector space over the field F.
(3.16) Remarks. In a vector space we have Ox = IXO = 0 because
Ox = (0 + O)x = Ox + Ox and IXO = IX (0 + 0) = IXO + IXO. Also IX =l= 0 and
x =l= 0 imply IXX =l= 0, since otherwise we would have x = Ix = (lX-llX)X
= IX-I (IXX) = 1X-10 = O.
(3.17) Examples. (a) Let F be any field, let n EN, and let X = F".
For :Xl = (xl> ... , x,,) and y = (Yl' ... , y,,) in X and IX EF define :Xl + y
= (Xl + Yl' ... , x" + y,,) and 1X:Xl = (IXXl , ••• , IXX,,). Then X is a vector space
over F.
(b) Let F be any field, let A be any nonvoid set, and let X = FA.
For /, g EX and IX EF define (f + g) (x) = / (x) + g (x) and (IX/) (x) = IX/ (x)
for all x EA. Then X is a vector space over F. Note that (a) is the special
case of (b) in which A = {I, ... , n}.
(c) Let X = R with its usual addition and let F = Q. For x ERand
IX EQ let IXX be the usual product in R. Then R is a vector space over Q.
(3.18) Definition. Let X be a vector space over F. A subset A of X is
said to be linearly independent [over F] if for every finite subset
{Xl' x 2 , •• • , x,,} of distinct elements of A and every sequence (lXI' 1X2, ••• , IX,,)
n
of elements of F, the equality}; IX,.X,. = 0 implies the equalities
k=1
1 The reader will find a discussion of groups, rings, and fields in § 5.

= DC2 = ... = DC" = 0.1 A nonvoid linearly independent set B such that
DC1
B ~ E C X implies that E is not linearly independent is called a Hamel
basis [or merely basis] for X over F. Thus a Hamel basis is a maximal
linearly independent set.
(3.19) Theorem. Every vector space with at least two elements contains
a Hamel basis.
Proof. Let X be a vector space with at least two elements. Let x =l= 0
in X. Then (3.16) shows that {x} is a linearly independent set. Thus the
family:F of all linearly independent subsets of X is non void. The definition
of linear independence shows at once that:F is of finite character. TUKEY's
lemma proves that :F contains a maximal member, i. e., X contains a
basis. D
(3.20) Theorem. Let X be a vector space over a field F and let B be a
Hamel basis lor X over F. Then lor each x EX there exists a unique lunction
DC Irom B into F such that DC (b) = 0 except lor finitely many b E B and
x = L DC (b) b, i. e., x can be expressed in just one way as a finite linear
bEB
combination 01 members 01 B.
Proof. Let x EX. If x EB, define DC(X) = 1 and lX(b) = 0 for bE B,
b =l= x. Then LIX(b)b = Ix = x. Suppose x ~ B. Then B U {x} is not
bEB
linearly independent, so there is a finite set {x, Xl' x 2, ... , x,,} C B U {x}
and a finite sequence ((3, (31) ... , (3,,) cF not all 0, such that (3x + (31Xl
+ ... + (3" x" = O. Since B is independent, we see at once that (3 =l= O.
Therefore x = - (3-1(31Xl - ••• - (3-1(3"X". Now define IX(X;) = - (3-1(3;
(j = I, ... , n) and IX (b) = Oforb EB n {Xl' ... ' x,,}'. Then x = LIX(b)b.
This proves the existence statement. bE B
To prove uniqueness, suppose that LIXI(b)b = LIX2(b)b. Then
bEB bEB
L (IXI (b) - 1X2(b))b = 0, and this is a finite linear combination of elements
bEB
of B. By independence, IXI (b) - 1X2 (b) = 0 for each b EB and therefore
the two functions IXI and 1X2 are the same. D
(3.21) Exercise. Given a nonvoid set A and a field F, let ~ be the
subset of FA consisting ofthose functions I for which the set {a EA : I (a) =l= O}
is finite. Let the linear operations in ~ be as in (3.17.b). Prove that ~ is
a vector space over F. Prove that every vector space is isomorphic qua
vector space with some vector space ~. 2
(3.22) Exercise. Prove that if P is a set and ;;;;; is a partial ordering
on P, then there exists a linear ordering ;::;; 0 on P such that ;;;;; C ;;;;; o.
1Note that [25 is linearly independent.
S Let Xl and XI be linear spaces over F. An isomorphism T of Xl onto XI is a
one-to-one mapping of Xl onto Xs such that T(X + Y) = T(X) + T(y) and
T (ocx) = OCT (x) for all x, y E Xl and all oc EF. Isomorphic linear spaces cannot be
told apart by any linear space properties.
§ 4. Cardinal numbers and ordinal numbers 19
(3.23) Exercise. Let (L, ~) be a linearly ordered set. Prove that there
exists a set We L such that ~ well orders Wand such that for each
x EL there is ay EW for which x ~ y.
(3.24) Exercise. Let G be a group and let H be an Abelian subgroup
of G. Prove that there exists a maximal Abelian subgroup] of G such
that He]; i. e., ] is Abelian, but no subgroup J* such that] ]* is s;:
Abelian.
(3.25) Exercise. Prove that the following assertion is equivalent to
the axiom of choice: If A and Bare non void sets and I is a function from
A onto B, then there exists a function g from B into A such that
g (y) E1-1 (y) for each y EB.
(3.26) Exercise. Let X be a vector space over a field F. Let A be a
nonvoid linearly independent subset of X and let 5 be a subset of X
such that each element of X is a finite linear combination of elements
of S. [The set 5 is said to span X.] Suppose that A c S. Prove that X
has a Hamel basis B such that A C Be S.
§ 4. Cardinal numbers and ordinal numbers

As noted in (2.16), two sets which can be placed in a one-to-one
correspondence cannot be told apart by any purely set-theoretic prop-
erties although, of course, they may be quite different entities. This
observation leads us to the following definition.
(4.1) Definition. With every set A we associate a symbol, called the
cardinal number 01 A, such that two sets A and B have the same symbol
attached to them if and only if there exists a one-to-one function I with
doml = A and mgl = B. We will write A '" B to mean that such a
one-to-one function exists. If A '" B, we say that A and B are equivalent
[equipollent, equipotent, have the same cardinality, have the same power].
We write A to denote the cardinal number of A. Thus A = 11 if and only
if A '" B.
(4.2) Examples. Some sets are so commonly encountered that we name
their cardinal numbers by special symbols. Thus 0 = 0, {I, 2, ... , n} = n
for each n EN, N = Ko [read "aleph nought"], and R = c [for con-
tinuum].
(4.3) Remark. The reader will easily verify, by considering the iden-
tity, inverse, and composite functions, that set equivalence, as defined
in (4.1), is reflexive, symmetric, and transitive. This fact makes Definition
(4.1) reasonable and also extremely useful.
(4.4) Remark. Our definition of cardinal number is somewhat vague
since, among other things, it is not made clear what these "symbols" are
to be. Some such vagueness is inevitable because of our intuitive approach
to set theory. However our definition is adequate for our purposes. In
2*
one version of axiomatic set theory, the cardinal number of a set is taken
to be a very specific well-ordered set, viz., the smallest ordinal number
that is equivalent to the given set.
We next define an order relation for cardinal numbers.
(4.5) Definition. Let u and 1) be cardinal numbers and let U and V
be sets such that U = u and V = 1). We write u ~ 1) or 1) ~ U to mean
that U is equivalent to some subset of V. One sees by considering com-
posite functions that this definition is unambiguous. We write u < 1) or
1) > U to mean that u ~ 1) and u =1= 1).
(4.6) Theorem. Let u, 1), and w be cardinal numbers. Then:

(i) u ~ u;
(ii) u ~ 1) and 1) ~ w imply u ~ w.
Proof. Exercise.
(4.7) Theorem [SCHRODER-BERNSTEIN]' II u and 1) are cardinal
numbers such that U ~ 1) and 1) ~ u, then u = 1).
Proof. Let U and V be sets such that U = u and V = 1). By hypoth-
esis there exist one-to-one functions I and g such that doml = U,
rngl C V, domg = V, and rngg cU. Define a function q; on &J(U)
into &J (U) by the following rule:
q; (E) = U n [g(V n (I (E))')]' . (1)
It is easy to see that
E C FeU implies q; (E) C q; (F) . (2)
Define !') = {E E &J(U): E C q;(E)}. Notice that 0 E!'). Next let
D = U !'). Since E C D for each E E !'), (2) implies that E C q; (E) C q; (D)
for each E E!'). Therefore D C q; (D). Applying (2) again, we have
q;(D) C q;(q;(D) so q;(D) E!'). Thus we have the reversed inclusion
D = U !') :> q; (D), so that q; (D) = D. According to (I), this means that
D= un [g(V n (I (D))')]' .
Thus un D' = g(V n (I(D))'). It follows that the function h defined
on Uby
h (x) = { I (x) for xED,
g-l (x) for x E U n D' ,
is one-to-one and onto V. 0
The proof of the Schroder-Bernstein theorem does not require the
axiom of choice. Also it does not tell us all that we would like to know
about comparing cardinal numbers: it merely asserts that u < 1) and
1) < U cannot occur. To prove that all pairs of cardinals are actually
comparable, as we do in (4.8), the axiom of choice is needed.
(4.8) Theorem. Let u and 1) be cardinal numbers. Then either u ~ 1)
or l) ~ u.
Proof. Let U and V be sets such that U = u and V = 1). Let 5'
denote the family of all one-to-one functions I such that doml C U
and rngl C V. It is easily seen that 5' is a family of finite character so,
by TUKEY's lemma (3.8), 5' contains a maximal member h. We assert
that either domh = U or rngh = V. Assume that this is false. Then
there exist x E un (domh), andy E V n (rngh)'. But then h U {(x,y)}E5',
contradicting the maximality of h. Thus our assertion is true. If domh= U,
then h shows that u ;:::;; 1). If rngh = V, then h- 1 shows that 1) ;:::;; U. 0
(4.9) Theorem. The ordering;:::;; lor cardinal numbers makes any set
01 cardinal numbers a linearly ordered set.
Proof. This theorem is just a summary of Theorems (4.6), (4.7),
and (4.8). 0
Our next theorem shows that there is no largest cardinal number.
(4.10) Theorem [CANTOR]' Let U be any set. Then (J < [ljJ(U).
Proof. We suppose that U =1= 0, since [ljJ(0) = 1> 0 = '0. Let
U = (J and 1) = [ljJ(U). The function I defined on U by I (x) = {x} E [ljJ(U)
is one-to-one, so U ;:::;; 1). Assume that u = 1). Then there exists a [one-to-
oneJ function h such that domh = U and rngh = [ljJ(U). Define
5 = {x E U: x ~ h(x)} .
Since 5 c U [perhaps 5 = 0J, we have 5 E [ljJ(U). Thus, because h

is onto [ljJ(U), there exists an element a E U such that h(a) = 5. There
are only two alternatives: either a E 5 or a ~ 5. If a E 5, then, by the
definition of 5, we have a ~ h(a) = 5. Therefore a ~ 5. But 5 is the set
h(a), so a f h(a), which implies that a E 5. This contradiction shows
that U =1= 1), and so we have proved that U < 1). 0
(4.11) Remark. Intuitive set theory suffers from the presence of
several well-known paradoxes. These known paradoxes are avoided in
axiomatic set theory by the elimination of "sets" that are "too large".
For example, let C be the "set" of all cardinal numbers. For each a E C,
let Aa be a set such that Aa = a. Define B = U {Aa : a E C}. Let b = B.
Since Aa C B we have a ;:::;; b for every cardinal number a. This conclu-
sion is incompatible with Theorem (4.lO). The trouble is that the "set"
C is "too large". It is indeed very large. We shall have no occasion in
this book to consider such large sets.
(4.12) Definition. A set 5 is said to be finite if either 5 = 0 or
g = n = {I, 2, ... , n} for some n EN. Any set that is not finite is said
to be infinite.
Definitions of "finite" and "infinite" that make no mention of the
natural numbers have been given by TARSKI and DEDEKIND. We
state them in the form of a theorem.
(4.13) Theorem. Let S be a set. Then:

(i) [TARSKI] the set S is finite il and only il each nonvoid lamily 01
subsets 01 5 has a minimal member;
(ii) [DEDEKIND] the set S is infinite il and only il S is equivalent to
some proper subset 01 itsell.
Proof. Exercise. [Use (4.15).]
(4.14) Definition. A set 5 is called countable if either S is finite or
S = N = No. Any set that is not countable is uncountable. A set 5 is
countably infinite [or denumerable] if S = No. If 5 is countably infinite and
I is a one-to-one function from N onto 5, then the sequence (xn) where
Xn = I(n) is called an enumeration of 5. Note that Xn =l= Xm if n =l= m.
(4.15) Theorem. Every infinite set has a countably infinite subset.
Proof. Let A be any infinite set. We show by induction that for each
n EN there exists a set An C A such that An = n. Indeed A =l= 0 so
there exists an Al C A. If An C A and An = n, then, since A is infinite,
there exists an element x EA n A~. Letting An+! = An U {x}, we have
An+! C A and An+! = n + 1.
Next let {An}nEN be any family of subsets of A as described above.
[Notice the use of the axiom of choice in selecting this family.] For each
n EN, define
Bn = A 2n n ( k=O
n-l
U A2k
)'
Then the family {Bn}nEN is a pairwise disjoint family of subsets of A,

and for each n E N we have
n-1 n-1
11n ~ A 2n - 1: A2k = 2n - 1: 2k = 2n - (2n - 1) = 1,
k=O k=O
so each Bn is nonvoid. Apply the axiom of choice to {Bn}nE N to get a

choice function I. Then I is a one-to-one mapping of N into A, so mgl
is a countably infinite subset of A. 0
(4.16) Corollary. II <l is any infinite cardinal number, i.e., the cardinal
number 01 an infinite set, then No ~ <l.
(4.17) Theorem. Any subset 01 a countable set is countable.
Proof. Let A be any countable set and let B cA. If B is finite, there
is nothing to prove. Thus suppose that B is infinite. Then A is countably
infinite. Let (an) be an enumeration of A. Define a one-to-one function I
from N onto B recursively as follows:
1(1) = an, where n i is the smallest n EN such that an E B;
t (k + 1) = anH1 where nk+I is the smallest n E N such that an E
B n {~, a2 , ••• , ani}'. 0
(4.18) Theorem. The Cartesian product NxN is a countable set.
Proof. We must show that N,.., NxN. One way to do this is to
define the mapping I from N x N onto N by I(m, n) = 2m-I(2n - 1).
Since each positive integer is a power of 2 [possibly the ot" power] times
an odd integer, I is onto N. We see that I is one-to-one, for otherwise
there would be an integer which is both even and odd. 0
(4.19) Lemma. II A is any nonvoid countable set, then there exists a
mapping Irom N onto A.
Proof. Since A is countable, there exists a one-to-one mapping g
from A into N. Let a EA. Define I on N by
n _ { g-l (n) for n E rngg ,
I( ) - a for n ~ rngg. 0
(4.20) Lemma. II A and B are two nonvoid sets and il there is a
mapping Ilrom A onto B, then A ~ B.
Proof. Let g be a choice function for the family {I-I (b)hEB. Then g
is a one-to-one mapping from B into A. 0
(4.21) Theorem. The union 01 any countable lamily 01 countable sets
is a countable set, i. e., il {AihEI is a lamily 01 sets such that I is countable
and each Ai is countable, then A = U Ai is countable.
iEI
Proof. Let {AihEI be as in the theorem. We obviously may, and do,
suppose that I and each Ai are nonvoid. Apply Lemma (4.19) to obtain
mappings/iandgsuch that doml i = domg = N, rngg = I, and rngli= Ai
for all i E I. Now define h on Nx N by hem, n) = fg(m) (n). Then h
is onto A. It follows from (4.20) and (4.18) that
A;;:;; NxN= Ko.

By (4.16), A is countable. 0
(4.22) Corollary. Each 01 the following sets is countable:
(i) Z, the set of all integers;
(ii) Q, the set of all rational numbers.
Proof. We have
Z = N U {o} U {-n: n EN}
and
Q=U {~:mEZ}.
n=1 n
0
We next introduce arithmetical operations for cardinal numbers.

We will show that the arithmetic of infinite cardinals is quite simple.
(4.23) Definition. Let a and b be cardinal numbers and let A and B
be sets for which A = a and B = b. If A n B = 0, we define a + b
= AU B. We define ab = A==;zjJ and a" = (AB).
It is easy to show that these are unambiguous definitions. Also we
hasten to point out that a + b is always defined since it is always possible
to find appropriate sets A and B that are disjoint. In fact if A n B *
0,
then define Ao = {(a, 0) : a E A} and Bo = {(b, I) : bE B} to obtain

A,.., A o' B,.., B o, and Ao n Bo = 0.
(4.24) Theorem. Let u, 1', and W be any three cardinal numbers. Then:
(i) u + (1' + w) = (u + 1') + w;
(ii) u + l' = l' + u;
(iii) u(l' + w) = Ul' + uw;
(iv) u(l'w) = (Ul')w;
(v) Ul' = l'U;
(vi) u"utll = uO+ tIl ;
(vii) Utll l'tIl = (Ul')tIl;
(viii) (u")tIl = u"tIl;
lix) u ~ l' implies u + W ~ l' + w;
(x) U ~ l' implies uw ~ l'W;
(xi) u ~ l' implies utll ~ l'tIl;
(xii) u ~ l' implies WU ~ w" .
Proof. All twelve of these conclusions are proved by defining appro-
priate one-to-one mappings. As a sample we prove (viii) while the
remaining eleven are left as exercises.
r
Let U, V, and W be sets such that fj = u, = 1', and W = w. We
must show that (UV)w,.., UVxw. To do this we define a mapping f{J
on (UV) W by the rule
f{J(/) = g E uvxw
where
g(y, z) = (t(z)) (y) E U for (y, z) E Vx W .
Now f{J is onto UVXW since if g E Uvxw, we define I to be that function
on W whose value at z E W is that function on V which assigns to each
y E V the value g(y, z) E U. Then f{J(/) = g. To see that f{J is one-to-one,
suppose that 11 =1= 12 in (UV)w. Then there is a zoE W such that 11(zo)
=1= 12 (zo)· Since these two functions on V are different, there must be a
YoE V such that 11 (zo)(Yo) =1= 12 (zo) (Yo)· Thus [IP (/1)] (yo,zo) =1= [IP (/2)] (Yo, zo)
so IP (/1) and IP (/2) are different functions. 0
(4.25) Theorem. II a is any cardinal number, then a < 2<1. __
Proof. Let A be a set such that A = a. We know (4.10) that a< .9'(A)
and that 2<1 = ({O, I}A). It suffices to show that .9'(A) ,.., {O, I}A. Define
IP on .9'(A) by
f{J(E)=gEE{O,I}A for EcA,
where as in (2.20) {
1 for x E E ,
gE (x) = 0 for x E A n E'. 0
We next consider the cardinal number c = R. The reader is invited
to look ahead to § 5 for a detailed construction of R and for the relevant
properties of R that we use here.
(4.26) Theorem. Let

]0, 1[ = {x ER: °< x < I},
[0, 1[ = {x E R: ° < I},
° x I} .
~ x
[0, 1] = {x E R: ~ ~
=
Then ]0, 1[ = [0, 1[ = [0, 1] = c.
1-2x
Proof. The function I defined by I(x) = x(l-x) is a one-to-one
mapping of ]0, 1[ onto R. Therefore JO, 1 [ = R = c. The rest follows
from the inequalities c = ]0, 1[ ~ [0, 1[ ~ [0, 1] ~ 11 = c and the
SchrOder-Bernstein theorem. D
(4.27) Theorem. 2lto = c.
Proof. Let A = {O, I}N. Definition (4.23) shows that X = 21to. By
(4.26), [0, 1[ = c. Define I on A by I(q;) = £ !Pi:) .Then [see (5.40)]
1'1=1
I is a one-to-one mapping of A into [0, 1[, and so 2lto ~ c. For each
x E [0, 1 [ there is a unique representation of x in the form
00
'" x"
x="",,-
1'1=1 2"
° °
where each x,.. is or 1 and x,.. = for infinitely many n EN: see (5.40).
Define g on [0, 1 [ into A by g (x) = q; where q; (n) = x,.. for each n EN.
Then g is a one-to-one mapping, so that C ~ 21to. Now apply the Schroder-
Bernstein theorem. D
We next point out a few curious arithmetical properties of infinite
cardinal numbers. First we need a lemma.
(4.28) Lemma. II D is any infinite set and F is any finite set such that
D n F = 0, then 15 = D U F.
Proof. Let F = {Yv Ys, ... , y,..} where Yi =1= Yi for i =1= j and let
C = {Xi:j EN} be a countably infinite subset of D where Xi =1= xi for
i =1= j (4.15). Define I from D onto D U F by
I(x) = !
Yi for x = Xi' 1 ~ j ~ n ,
xi_,.. for x = xi,j > n,
x for xED n c/.
Then I is one-to-one. D
(4.29) Theorem. Let a be any infinite cardinal number. Then a + a = a.
Proof. Let A be any set such that A = a. Let B = A x {O, I}.
Then B = {(a, 0): a E A} U {(a, 1): a E A} so, by Definition (4.23),
we have 11 = a + a. Let 5' denote the set of all one-to-one functions I
such that doml C A and rngl = (dom/) x {O, I}. Since A is infinite,
there exists a count ably infinite set C such that C C A (4.15). In view
of (4.21), we see that Cx {O, I} is also count ably infinite. Hence there
is a one-to-one function I with doml = C and mgl = Cx {O, I}. This

proves that 5' =l= 0. Partially order 5' by C. According to the Haus-
dorff Maximality Principle (3.9), 5' contains a maximal chain lr. Let
g = U lr. It is easily checked that g E5'. Let D = domg. The existence
of the function g shows that jj = jj + D. Thus, to complete the proof,
it suffices to show that Jj = a. Let E = AnD'. If E is finite, our Lemma
(4.28) shows that jj = DUE = a. If E is infinite, let G be a countably
infinite subset of E. Let I be any one-to-one mapping of G onto G x {O, I}.
s:
Then h = lUg E5' and g h. This contradicts the maximality of lr.
Therefore E is finite and jj = a. 0
(4.30) Corollary. II a is any infinite cardinal number and b is any
cardinal number such that b ~ a, then a + b = a.
Proof. Since b ~ a, we have a ~ a + b ~ a + a = a. 0
(4.31) Theorem. II a is any infinite cardinal number, then a 2 = aa = a.
Proof. Let A be any set such that ;r = a. Let 5' denote the set of all
one-to-one functions I such that doml C A and mgl = (dom/) x (dom/).
Since A contains a count ably infinite subset (4.15) and since No No = No
(4.18), we see that 5' =l= 0. As in (4.29), we use the Hausdorff Maximality
Principle to prove that 5' contains a maximal member g. Let D = domg.
Then the existence of g shows that D '"" D x D. To finish the proof we
need only show that jj = a. Let E = AnD' and let b = D. If E ~ b,
then (4.30) shows that b = b + II = i5TJE = A = a. The only other
possibility is that b < E (4.8). Assume that this is the case. Then there
is a set G C E such that iJ = b. Since D '"" D x D, we know that b2 = b.
===
Thus D x G = G x D = G x G = b. We appeal to (4.29) to see that
b = b + b + b. It follows that (D x G) U (G x D) U (G x G) = b = G.
Consequently there exists a one-to-one function I from G onto (D x G)
U (G x D) U (G x G). Define h= lUg. Then his a one-to-one correspond-
ence between DUG and (D U G) x (D U G). Thus we have h E5'.
s:
Since g h, we have contradicted the maximality of g. Consequently
E ~ band b = a. The accompanying figure may be helpful.
D ~ DxD DxG
G ~ GxD GxG o
°
(4.32) Corollary. II a is an infinite cardinal number and b is a cardinal
number such that < b ~ a, then a b = a.
Proof. We have a ~ ab ~ aa = a. 0
(4.33) Exercise. Prove that our ordering and our arithmetical opera-
tions for cardinal numbers agree on the set N with the usual ordering
and arithmetical operations for positive integers.
(4.34) Exercise. Let a be any cardinal number such that 2 ~ a ~ c.

Prove that alto = c and that a C = 2c.
(4.35) Exercise. Let A be any infinite set and let $' be the family
of all finite subsets of A. Prove that ffj = A.
(4.36) Exercise. Let A be any infinite set such that A ~ c, and let
~ denote the family of all countable subsets of A. Prove that ~ = c.
(4.37) Exercise. Let W be a set and suppose that q; c W x W is a
relation such thai (W, q;) and (W, q;-l) are both well-ordered sets.
Prove that W is finite.
(4.38) Exercise. Without appealing to the axiom of choice or its
equivalents prove that c2 = c. [Hint. Use decimal representations of
real numbers.]
(4.39) Exercise [KONIG]. Let I be a nonvoid set and let {A,},E[ and
{B,}.u be families of sets such that X. < B, for each tEl. Let A = U A,
'EI
and B = X B .. Prove that A < B. [For each tEl, let n, denote the pro-
,EI
jection mapping of B onto B" i.e., n, (b) = b, for each b E B. Let 1be any
mapping from A into B. Then n, 0 I(A,) =l= B, (4.20), so that there exists
c, E B, n [n, 0 t(A,)], for each tEl. It follows that the element c E B
having c, as tth coordinate is not in rng f. Thus there is no mapping of A
onto B.]
We next present a brief introduction to the theory of ordinal num-
bers. The chief distinction between cardinal numbers and ordinal numbers
is that each set has a cardinal number while only well-ordered sets have
ordinal numbers. There may be many essentially different ways to well
order a given set. Each of these ways has its own ordinal number even
though the set has only one cardinal number. For example, we can well
order N in the following two ways:
1<2<3<,,, ,
2<3<4<",<1.
The first of these is the usual ordering and the second of these is the
same except that 1 has been removed from the beginning and placed
at the end. These well orderings are different since the second has a last
element while the first has no last element. We need some precise def-
initions.
(4.40) Definition. Let A and B be linearly ordered sets. An order
isomorphism Irom A onto B is a one-to-one function 1 from A onto B
such that x ~ y in A implies 1(x) ~ t(y) in B. We write A ~ B to mean
that such an order isomorphism exists. It is easy to see that the relation ~
is reflexive, symmetric, and transitive. With every linearly ordered set A
we associate a symbol, called the order type 01 A, such that two linearly
ordered sets A and B have the same symbol attached to them if and only
if A ~ B. If A ~ B, we say that A and B are order isomorphic or have

the same order type. We write ordA to denote the order type of A. If,
in particular, A is well ordered, we call ordA an ordinal number.
(4.41) Examples. Let {I, 2, ... , n} (n E N), N, and Q have their
usual orderings. We write ord 0 = 0, ord{I, 2, ... , n} = n, ordQ = "I,
and ordN = w. Thus 0, n, and ware ordinal numbers, but "I is not an
ordinal number, since Q is far from being well ordered.
(4.42) Definition. Let A be a linearly ordered set and let x EA. The
initial segment 01 A determined by x is the set Ax = {y E A: y < x}.
If ex and p are ordinal numbers and A and B are well-ordered sets such
that ordA = ex and ordB = p, we write ex < p to mean that there is an
x E B such that A ~ Bx. We write ex ~ p to mean that either ex < p
or ex = p. It is easy to verify that this definition of ordering for ordinal
numbers does not depend on the particular sets A and B that are used,
but only upon their order types.
Let us investigate the properties of this ordering.
(4.43) Theorem. I I A is a well-ordered set and I is an order isomorphism
Irom A into A, then x ~ I(x) lor each x EA.
Proof. Assume that there is an x in A such that I (x) < x and let a
be the smallest such x. Then I (a) < a so l(t (a)) < I (a). But this contra-
dicts the minimality of a. 0
(4.44) Theorem. Let A and B be well-ordered sets. Then:
(i) A is order isomorphic to no initial segment 01 A;
(ii) Ax ~ A:v lor some x, yEA implies x = y;
(iii) il A ~ B, there exists exactly one order isomorphism Irom A onto B.
Proof. Assume that there are an x E A and an order isomorphism I
from A onto Ax. Then by (4.43), we have x ~ I(x). But I (x) E Ax, so
that I (x) < x. This contradiction proves (i).
Now suppose that Ax~A:v for some x,yEA. Assume that x=l=y.
We may suppose that x < y. Then Ax is an initial segment of the well-
ordered set A:v' As shown in (i), this is impossible, and so (ii) is established.
Suppose that I and g are order isomorphisms from A onto B. Then
h = I-log is an order isomorphism of A onto A. Applying (4.43), we
have x ~ hex) for each x E A, i. e., I(x) ~ g(x) for each x EA. Inter-
changing the roles of I and g in this argument, we obtain g (x) ~ I (x)
for each x EA. It follows that I (x) = g(x) for each x E A, i. e., 1= g. 0
(4.45) Theorem. Let ex and p be ordinal numbers. Then exactly one 01
the lollowing three alternatives obtains: ex < p, ex = p, p < ex.
Proof. Theorem (4.44) shows that at most one of these alternatives
can prevail. We now show that at least one of them must.
Let A and B be well-ordered sets such that ordA = ex and ordB = p.
Let 5' denote the family of all mappings I such that I is an order isomor-
phism from either an initial segment of A or A itself onto either an
initial segment of B or B itself [we may obviously suppose that

A =l= 0 =l= B]. If a is the least element of A and b is the least element
of B, then {(a, b)} E ff, and so ff =l= 0. By the Hausdorff Maximality
Principle, there is a maximal chain cr C ff. [Actually cr= ff, but we
do not need ihis fact.] Let h = U cr. It is easy to check that h belongs to
ff. If domh and rngh are initial segments Ax and By of A and B, re-
spectively, then h U {(x, y)} can be adjoined to cr, and this violates the
maximality of cr. Thus we have either domh = A or rngh = B. If
domh = A, then either rngh = B [i. e., (X = fJ], or rngh is an initial
segment of B [i. e., (X < f3]. If domh =l= A, then domh is an initial segment
of A and rngh = B, and so the existence of h- 1 shows in this case that
{3 < (x. 0
(4.46) Corollary. With the ordering defined in (4.42), any set of ordinal
numbers is linearly ordered.
(4.47) Theorem. Let (X be any ordinal number > 0 and let Prz denote
the set of all ordinal numbers < (x. Then Prz , with the order relation of
(4.42), is a well-ordered set and ordPrz = (x.
Proof. Let {3 E Prz and let A and B be well-ordered sets such that
ordA = (X and ordB = {3. Since {3 < (x, there is an x E A such that
Ax F::l B. In view of (4.44.ii), this x is uniquely determined by {3. Thus
define g; ({3) = x. The reader should have no difficulty in verifying that
this defines an order isomorphism g; from Prz onto A. Thus Prz is well
ordered and ordPrz = ordA = (x. 0
(4.48) Theorem. Let tl be a cardinal number. Then there exists an
ordinal number (X such that Prz = tl.
Proof. Let A be any set such that A = tl. According to the Well-
ordering Theorem (3.11), there is a well ordering on A, making A a well-
ordered set. Let (X = ordA. Then (4.47) A F::l Prz. Consequently A ,., Prz
and Prz = .A = tl. 0
(4.49) Theorem. There is a smallest ordinal number D such that P{J
is uncountable. The set P{J has the following properties:
(i) P{J is well ordered;
(ii) (X E P{J implies Prz is countable;
(iii) P{J is uncountable;
(iv) C C P{J and C countable imply there is a {3 E P{J such that (X ;:;;; {3
for each (X E C.
Proof. Choose an ordinal number y such that p" = c. If each member
of P" has only countably many predecessors, set D = y. Otherwise some
members of P" have uncountably many predecessors and we let D be
the smallest of these (4.47). Conclusion (i) follows at once from (4.47).
Conclusions (ii) and (iii) follow from the definition of D. Suppose that C
is a countable subset of P{J' Let D= U {Prz: (XE C}. ThenD is a count-
able union of countable sets so D is countable (4.21). Let {3 E P{J n D'.
Clearly ex ~ {J for each ex E C for otherwise (J E D. This proves (iv). 0

(4.50) Remark. The cardinal number of the set P n is denoted N1 .
The continuum hypothesis is the assertion that Nl = c. This is equivalent
to the assertion that each infinite subset of R is either countable or of
cardinal number c. Many interesting and important theorems have
been proved with the aid of this hypothesis. It was recently shown by
PAUL J. COHEN [lac. cit. (3.2)] that the continuum hypothesis is inde-
pendent of the Zermelo-Fraenkel axioms of set theory.
(4.51) Exercise. Prove that every nonvoid set of cardinal numbers
has a smallest member.
(4.52) Exercise. Let A be an infinite linearly ordered set. Suppose
that no infinite subset of A has a largest element. Prove that ordA = co.
[Recall that co = ordN.]
(4.53) Exercise. Let A be a linearly ordered set such that:
(i) A is count ably infinite;
(ii) A has no first or last element;
(iii) x, yEA, x < y imply that there is a z E A such that x < z < y.
Prove that ordA = 'YJ. [Recall that 'YJ = ordQ.]
(4.54) Exercise. Prove that there are uncountably many different
ways to well order the set N such that no two of these different well-
ordered sets are order isomorphic.
(4.55) Exercise. Let A be any infinite set. Prove that A can be well
ordered in such a way that it has no last element. Also show that there
is a well-ordering of A in which there is a last element.
(4.56) Exercise. A permutation of a set A is any one-to-one mapping
of A onto A. Let A be a set such that .A > 1.
(a) Prove that there exists a permutation I of A such that t (x) =1= x
for all x EA.
(b) Show that if ..4 is an even integer or is infinite, then the permuta-
tion I in (a) can be chosen so that 10 I (x) = x for all x EA. What happens
if ..4 is an odd integer?
(c) Show that the permutation I in (a) can always be chosen so that
lot 0 tot 0 10 t(x) = x for all x EA.
(4.57) Exercise. Let B be a set, let b = 11, and let
t
b! = {I : is a permutation of B}.
Prove that if B is infinite, then b! = 21>.
We now prove a theorem which allows us to define the algebraic
dimension of any vector space.
(4.58) Theorem. Let X be a vector space over a field F and let A and B
be any two Hamel bases lor X over F. Then1' = B.
Proof. We will first use ZORN'S lemma to produce a one-to-one
function from A into B. To this end let 3 denote the set of all one-to-one
functions! such that:

(1) dom! C A;
(2) rng! C B;
(3) (rng!) U [A n (dom!)'] is linearly independent over F.
The fact that A is linearly independent shows that the empty function 0
is an element of 8. Thus 8 =1= 0. Partially order 8 by inclusion. To show
that ZORN'S lemma applies to 8, let ~ be any nonvoid chain contained
in 8 and let g = U ~. An application of (2.19) shows that g is a function
and that (1) and (2) hold for the functiong. One easily sees that gis one-to-
one. We have
(rngg) U [A n (domg)'] = (/~<l:rng/) U [A n (/~<l:dom/),]. (4)
Now let F be any finite subset of the set in (4). Since {rng/: I E ~}
is a chain under inclusion, there is a function 10 E ~ such that
Fe (rng/o) U [A n (/~<l:dom/),]
C (rng/o) U [A n (dom/o)'].
Therefore F is linearly independent, and so g satisfies condition (3).
Thus g is in 8, and g is an upper bound for~. By ZORN'S lemma, 8 has a
maximal member, say h.
We assert that domh = A. Assume that domh =1= A and let ao E
A n (domh),. According to (3), a o is not a linear combination of elements
of rngh. Since a o is a linear combination of elements of B, it follows that
rngh =1= B. Let bo be any element of B n (rngh)'. If the set
{b o} U (rngh) U [A n (domh)']
is linearly independent, then, as is easily seen, the function hU{(ao,bo)}
is in 8, contrary to the maximality of h. We infer that bo is a linear com-
bination of elements of the set (rngh) U [A n tdomh)']; we write
n
bo = 1: IX"X" •
k~l
Since B is linearly independent, bo is not a linear combination of elements

of rngh. Hence there exists a k such that x" E A n (domh), and IX" =1= O.
Thus bo is not a linear combination of elements of the linearly independ-
ent set (rngh) U [A n ({x,,} U domh)'] and therefore the function
h U {(x"' bon is an element of 8. This contradicts the maximality of h.
Consequently domh = A and A ~ B.
Interchanging the roles of A and B in the above argument, we see
also that 13 ~ A. The proof is completed by invoking the Schroder-
Bernstein theorem (4.7). 0
(4.59) Definition. Let X be a vector space over a field F. We define
the algebraic [linear] dimension 01 X to be 0 if X = {O} and to be the car-
dinal number of an arbitrary Hamel basis for X over F if X =1= {O}.
(4.60) Exercise. Let X be a vector space over a field F and let B

be a Hamel basis for this space. Prove that:
(a) X = max{B, F} if B is infinite;
(b) X = FB if B is finite.
(4.61) Exercise. Without using the continuum hypothesis, find the
algebraic dimension of the vector space R over the field Q.
(4.62) Exercise [proposed by M. HEWITT]. Let A be a nonvoid set.
Suppose that there is a family !/ of subsets of A with the following
properties:
(i) B = 3 for all B E !/;
(ii) U!/ = A;
(iii) BI n B2 = 1 for distinct Bl> B2 E !/;
(iv) if x, yEA and x =l= y, then there is exactly one BE !/ contain-
ing {x,y}.
Prove that such an !/ exists if and only if ;r = 3, ;r = 7, or ;r ;;?; leo.
§ 5. Construction of the real and complex number fields

We give in this section a short and reasonably sophisticated construc-
tion of the real and complex numbers, assuming the rational numbers
as known. It seems appropriate to do this, since completeness of the real
number field is the rock on which elementary analysis rests. Also there
is a strong interplay between algebra and contemporary analysis, which
demands the use of the ideas and methods of algebra in analysis. We
begin with a few facts about groups and other algebraic structures.
(5.1) Definition. A set G together with a binary operation (x, y) -'>- xy
mapping G x G into G is called a group provided that:
(i) x(yz) = (xy)z for all x, y, z E G [associative law];
(ii) there is an element e E G such that ex = x for all x E G [e is a
left identity] ;
(iii) for all e as in (ii) and all a EG there exists a-I EG such that
a-1a = e [a- l is a left inverse for a].
If also we have
(iv) ab = ba for all a, bEG,
then G is called an Abelian group [after the Norwegian mathematician
N. H. ABEL (1802-1829)].
(5.2) Remarks. (a) Every left inverse is a right inverse. In fact, for
any e as in (ii) we have
(a-Ia) a-I = ea- I = a-I.
Now let b be a left inverse of a-I, i. e., ba- I = e. Then
b(a-Iaa- l ) = ba- l = e,
(ba- I) (aa- I ) = e,
e(aa- I) = e,
§ 5. Construction of the real and complex number fields 33
and so by (ii)
aa- 1 = e.
Note that the last equality also implies that a is a left inverse of a-I.
(b) For any e as in (ii) and any a E G, we have
ae = a(a- 1 a) = (aa- 1 ) a = ea = a,
i. e., e is also a right identity. If e1 and e2 satisfy (ii), then they are both
right identities also, and so
e1 e2 = e2 [e 1 is a left identity] ,
e1 e2 = e1 [e 2 is a right identity] ,
so that e1 = e2 , i. e., there is a unique left and right identity in G.
(c) Similarly one sees that a-I is unique.
(d) For Abelian groups, we often use additive notation [+ denotes
the binary operation]; in this case we denote the identity by 0, the in-
verse of a by -a, and a + (-b) by a-b.
(5.3) Definition. Consider a set A with two binary operations + and·
[called addition and multiplication, respectively], which is an Abelian
group under +, with identity 0, and in which the equalities
a· (b + c) = (a· b) + (a· c) [left distributive law] ,
(a + b) . c = (a . c) + (b . c) [right distributive law] ,
and
(a· b) . c = a· (b· c) [associative law for multiplication]
hold for all a, b, c EA. Then A is called a ring. If a· b = b· a for all a, b
in a ring A, A is called commutative. An element 1 E A such that 1 . a
= a . 1 = a for all a E A is called a [two-sided] unit for A. A non void
subset I of a ring A is called a left [right] ideal if a - bEl for all a, bEl
andx·aEI [a·xEI] forallaElandxEA. A subset IofA that is a left
and a right ideal is called a two-sided ideal.
(5.4) Remarks. (a) The notation in the statements of the distributive
laws is correct but clumsy. From now on we will follow the universal
algebraic convention that ab means a· b and that ab + cd means
(a· b) + (c· d).
(b) For all x in a ring A, we have xx = x(x + 0) = xx + xO, so
that xO = 0. Similarly Ox = 0.
(c) Evidently two groups or rings can be distinct objects and still
be indistinguishable as groups or rings. Formally, we say that rings A
and A are isomorphic if there is a one-to-one mapping -r carrying A
I
onto A' such that -r(a + b) = -rea) + -reb) and -r(ab) = -rea) db) for all
a, bE A. The mapping -r is called an isomorphism or an isomorphic
mapping. An analogous definition is made for groups. An isomorphism
of a ring or group onto itself is called an automorphism.
We now define an important special type of ring.

(5.5) Definition. A ring F such that F n {O}' is an Abelian group
under multiplication is called a field.
(5.6) Remarks.
(a) Since a group contains an element, our definition of a field shows
°
°
that 1 =l= and that a field contains at least two elements.
(b) The identities (SA.b) show that has to be excluded from F
in order to obtain a group under multiplication.
(c) The simplest field is {O, I}, with operations addition and multi-
plication modulo 2. The addition and multiplication tables for this field
are
+ 0 I .1 0
0 0 I 010 0
1 1 0 1 I0
(d) If P is a prime, the set {O, 1,2, ... , p - I} is a field under addi-
tion and multiplication modulo p. All necessary verifications are easy;
we will give the least easy one, namely that each nonzero element has
a multiplicative inverse. If a E{I, 2, ... , p - I}, then we must show
that there exists x E{I, 2, ... , p - I} such that
ax == 1 (mod P).
Since p is prime, the greatest common divisor of a and p is 1. It follows
that there are integers x and y, x =l= 0, such that
1 = ax + py.
In particular, there are integers x' and y' such that
1 = ax' + py' , 1 ~ x' < p;
hence ax' == 1 (mod P) and x' E {I, 2, ... , p - I}.
(e) If F is any field, then the elements 0, 1, 1 + 1, ... , n 1, ...
°
[where n is any positive integer and n 1 has an obvious meaning] are all
members of F. If n 1 = for some positive integer n, then the smallest
°
positive integer p such that p 1 = is obviously a prime. In this case F
is said to have characteristic p; otherwise F is said to have characteristic 0.
Fields of characteristic p are of no interest at present in elementary
analysis. If F has characteristic 0, then F has a sub field [the definition
of a subfield is obvious] which is isomorphic to the rational number field.
We will always denote the rational number field by the symbol Q.
To see that F contains a [unique] isomorph of Q, consider a mapping
i of part of F onto Q, and see what properties it must have in order to
be an isomorphism. For notational convenience, we ignore the distinc-
tion between the zeros of F and of Q and between the units of F and of Q.
It is clear that 1'(0) = 0 and 1'(1) = 1. A trivial induction shows that
l' (n 1) = n for all n E N. If l' is an isomorphism, we must also have
1
1'«(n 1)-1) = nand 1'(- (n 1)) = - n for all n EN. It follows that we must
have
1'«(m 1) (n 1)-1) = :
and
1'((-(ml)) (nl)-I) = - :
for all m, n EN. It is a routine matter to show that the mapping l' so
constructed is an isomorphism of a subfield of F onto the field Q.
(f) Let G be any group. For A, E C G, we define
A E = {xy: x E A, Y E E} ,
A-I = {X-l: x E A} ,
A2 = AA .
In additive notation, these sets are written as A + E, -A, and 2A.

The set An [nA in additive notation], where n is a positive integer, has
the obvious definition.
(g) For a nonzero element b of a field, we frequently write the multi-
plicative inverse b- 1 as ! ' and for an element a of the field, we frequently
write ab- 1 as : .
(5.7) Definition. A field F is said to be ordered if there is a subset P
of F such that:
(i) P n (- P) = 0 ;
(ii) PU{O}U(-P)=F;
(iii) a, bE P imply a + bE P and ab E P .
If one thinks of F as the rational or real numbers, then P is just the set
of positive rational or real numbers. Accordingly, in the general case
elements of P are called positive; elements of - P are called negative.
Since 0 = -0,0 cannot be an element of P.
(5.8) Theorem. Let F be an ordered field, and let P be as in (5.7).
It a EF and a =+= 0, then a2 E P. In particular, 1 E P. It a, bE F, ab E P,
and a E P, then also bE P.
Proof. If a E P, then a2 E P by (5.7.iii). If a ~ P and a =+= 0, then
aE -P by (5.7.ii), i. e., -aE P. Again by (5.7.iii), we have (-a)2E P.
In any ring, the identity (-a) (-b) = ab holds, and so a2 = (-a)2 E P.
Since 12 = 1, the second assertion holds. To prove the third assertion,
assume that b ~ P. If b = 0, then ab = 0 ~ P, which contradicts the
3*
hypothesis. If bE - P, then - bE P and

a(-b) = - (ab) E P;
therefore we have ab E - P, which is a contradiction. Hence bE P. D
(5.9) Theorem. Every ordered field F contains an isomorph of Q,
and the isomorphism can be taken as order-preserving.
Proof. Since 1 E P, (5.7.iii) implies that n 1 E P for all n EN. This
implies that F has characteristic 0, and so the isomorphism l' of (5.6.e)
can be constructed. It is easy to check that l' preserves order. D
(5.10) Definition. Let F be an ordered field. We write a < band b > a
if b - a E P; the expressions a ~ band b ~ a have obvious meanings.
(5.11) Theorem. Let F be an ordered field. For all a, bE F, we have
a b, and only one of these relations holds.
Proof. The proof is immediate from the definition of P and the fact
that
b- a = °
if and only if b = a ,
b - a E P if and only if b > a ,
a - b E P if and only if a > b. D
Many elementary facts about inequalities are consequences of the

axioms of order (5.7). We now list a few of them.
(5.12) Theorem. If F is an ordered field, if a, b, e, dE F, and a [e < 0], then ae < be [ae > be].
Proof. If a < b, then b - a E P, and therefore if e> 0, we have
c(b - a) E P and eb > ca. If e < 0, then - e E P, and so (-c) (b - a) E P;
hence ae - be E P, i. e., ae > be. D
(5.14) Theorem. In an ordered field F, the inequalities 0< a < b
imply that °< f < ~ .
Proof. Since b f = 1 E P and bE P, (5.8) implies that f E P.
Hence !, ~, and b - a are in P, and it follows that
1 1 1 1
-;; - T = -;; T (b - a) E P,
so that
1 1
O<T<-;;. D
(5.15) Definition. For an element a of an ordered field F, we define
lal =
{ -aaifa~O,
if a < 0.
(5.16) Theorem. For all elements a, b in an ordered field F, we have:

(i) lal = I-aj;
(ii) labl = lallbl;
(iii) la + bl ~ lal + Ibl;
(iv) I laj - jbj j ~ ja - bj .
Proof. Statements (i) and (ii) are obvious consequences of the def-
inition of jaj, and (iv) follows from (iii). We will prove (iii). If a ~ 0,
we have jaj = a; if a < 0, then we have lal = - a > 0. Hence we always
have
a ~ lal,
b~ Ibl,
and so (5.10) implies that
a + b ~ lal + Ibl .
Since -a ~ j-al = laj, we also have
- (a + b) ~ lal + Ibl '
and (iii) follows from these two inequalities. 0
(5.17) Definition. An ordered field F is said to be Archimedean
ordered if for all a E F and all b E P there exists a positive integer n
such that nb > a. In intuitive language, this definition means that no
matter how large a is and how small b is, successive repetitions of b will
eventually exceed a. There are ordered fields which are not Archimedean
ordered; see (5.39).
(5.18) Theorem. Let F be an Archimedean ordered field, and let a, b EF
be such that a < b. Then there exists ~
n
EF, where m and n are integers,
such that a < ~
n
< b. l
Proof. Since b - a > 0, we have (b - a)-I> 0; and so, since F is
Archimedean ordered and 1 > 0, there exists an integer n such that
Using (5.14), we have

n 1 > (b - a)-I> °.
0«n1)-I<b-a;
or, using an obvious notation,
1
-l<b-a.
n
Let n be any integer satisfying the last inequality, and let 5 be the set
5 = {k : k is an integer and k . ~ > a} .

1 The expression nm really means
ml
---;;y- E F; in view of (5.3. e), it does no harm
to suppose that F ~ Q.
Since ~n
°
> and F is Archimedean ordered, we have S =1= 0. Also,
again by the Archimedean order property of F, there is a positive integer p
such that
1
p-;;>-a,
- (p !) < a,
1
(-p)-;;<a.
It follows that -p, - (p + 1), - (p + 2), ... lie outside of S. Hence S
is a non void set of integers which is bounded below, and so it contains a
least element m. Since mE S, we have a < ~
n
.
m-l
Since m - 1 4 S, we have-n- ~ a; and so
m m-l 1 1
-=--+-:0;:
n n n - a+-<a+
n (b-a) =b·'
i. e.,
a<~<b.
n
0
(5.19) Definition. Let F be an ordered field. A sequence (an)

[= (aI' a2 , ••• , an, ... )] of elements of F is called bounded if there is an
element bE F such that lanl ~ b for each positive integer n. A sequence
(an) is called Cauchy if for every e EF such that e > 0, there is a positive
integer N (e) such that lap - aal < e for all p, q ~ N (e). [The French
mathematician AUGUSTIN LOUIS CAUCHY (1789-1857) first considered
this class of sequences, for the case in which F is the real number field.]
A sequence (an) is called null if for every e EF such that e > 0, there is
a positive integer N (e) such that lapl < e for all p ~ N (e). The families of
sequences satisfying these conditions will be denoted by SB, <t, and m,
respectively.
(5.20) Theorem. The inclusions mc <t c SB obtain.
Proof. If (an) E <t, thenp, q ~ N(I) implies lap - aal < 1. In particular,
laN(:DH - aN(l) I < 1 for k = 0, 1, 2, ....
Let b = max {Iall, la 2 1, ... , laN(I)I, laN(I) I + I} [see (2.7) for the defini-
tion of max]; then lapl ~ b for p = 1,2, ... , and so (an) ESB.
m,
If (an) E then for any given positive e EF, we have
1 1
lap - aql ~ lapl + laql <""2 e + ""2 e = e
provided that p, q ~ N( ~ e). It follows that (an) E<t, and so me <t. 0
(5.21) Theorem. For (an), (bn) E <t, let (an) + (bn) = (an + bn) and
(an) (bn) = (anbn ). With these definitions of sum and product, <t is a com-
mutative ring with unit, and mis an ideal in <t such that mS; <t.
Proof. Let us first show that sums and products of Cauchy sequences
are again Cauchy sequences. Let (an), (bn) E <r be given. For a positive
e E F, let N (e) and M (e) be the positive integers associated with the
Cauchy sequences (an) and (b n), respectively. Ifp, q;;;' max{N(~), M( ~)},
I I
then we have lap + bp - (aq + bq)1 ;;:;; lap - aql + Ibp- bql < Z-e+ Z-e=e.
It follows that (an) + (bn) is a Cauchy sequence.
Since <r c m,
there are positive elements c and d of F such that:
lapl ;;:;; c for p = 1, 2, 3, ... ;
Ibpl ;;:;; d for p = 1, 2, 3, .. .
If p, q;;;' max {N( 2~)' M( 2ec )}, then we have

lapbp - aqbql = lapbp - apbq + apbq - aqbql
ce de
~ lapllbp - bql + Ibqllap - aql < 2C + 2d = e .
It follows that (an) (b n) is a Cauchy sequence.
It is now obvious that <r is an additive Abelian group in which (0, 0, ... )
is the zero. Also multiplication in <r is clearly commutative and <r has
the multiplicative unit (1, 1, ...). The distributive law for <r follows
immediately from the distributive law for F. We have thus shown that <r
is a commutative ring with unit.
We will now show that m is a proper ideal of <r. If (an), (bn) Em,
then it is clear that (an) - (bn) E m; thus mis an additive subgroup of <r.
If (an) Em and (bn) E<r, then we must show that (anb n) Em. Let c be a
positive element of F such that
n = 1,2,3, ....
For a given positive e EF, there is a positive integer N ( :) such that
p ;;;. N ( :) implies that
Hence p ;;;. N ( :) implies that
lapbpl;;:;; laplc< : c=e,

and so (anbn) Em. Thus we have shown that m is an ideal. Since
(1, 1, 1, ... ) E <r n m', mis a proper ideal. D
Note that <r is not a field; e. g., (0, 1,0,0, ... ) has no multiplica-
tive inverse in <r.
(5.22) Theorem. Let ~/m denote the set whose elements are the sets
(an) + m [called cosets of m], where (an) E <r. Addition and multiplication
in ([.jm are defined by

«(an) + m) + «(bn) + Q,) = (an) + (b n) + m = (an + bn) + m ,
and
«(an) + m) ((bn) + m) = (an) (bn) + m = (anb n) + m.
These definitions are unambiguous, and with addition and multiplication
so defined, ([.jm is a field.
Proof. Since mis an additive subgroup of ([., any two coselS (an) + m,
(b n) + m are either disjoint or identical. Two sequences (an), (a~) E ([.
belong to the same coset if and only if (a~) = (an) + (c n), for some (cn) Em.
If (c n ), (dn ) E m, then
(((an) + (cn) + m) + (((b n) + (dn ) + m)
= (an) + (b n) + (c n) + (d n) + m= (an) + (b n) + m,
and
(((an) + (cn) + m) (((b n) + (dn) + m)
= (an) (b n) + (an) (dn) + (cn) (b n) + (cn) (dn ) +m
= (an) (bn) + m.
Thus addition and multiplication of cosets are defined unambiguously.
It is a simple matter to verify that ([.jm is a commutative ring with unit.
For example, (1, 1, 1, ... ) + mis the unit of ([.jm.
It remains to show that every nonzero element of ([.jm has a multi-
plicative inverse. Let (an) + m=1= m; i. e., let (an) E ([. n m'. We are
required to find (xn) E ([. such that
(anxn) + m= (l(n») + m;
i. e., such that
(anXn - l(n») Em.
Since (an) is not a null sequence, there is a positive e EF such that for
every positive integer r there is some integer s> r for which lasl ~ e.
For all p, q ~ N( ~ e) we have
1
lap - aql < 2 e .
Let s > N( ~ e) be such that lasl ~ e; then for arbitrary p ~ N( ~ e)

we have
1
e ~ lasl = las - ap + apl ~ las - apl + lapl < 2 e + lapl .
Hence
lapl > ~ e if p ~ N ( ~ e) .
We now define (xn ). Write N ( ~ e) as m, and let

and
1
xp=-
a p
if P~ m.
We have
and hence
(x"a" - 1(,,») = (al - 1, a 2 - 1, ... , am - l - 1,0,0, ... ) .
Thus it is obvious that (x"a" - 1(,,») E m. To complete the proof, we need
only show that (x,,) E <r. If p, q ~ N ( ~ e), then we have
Ix p - xql = I:p - ~q I= I:p 1·1 ~q I lap - aql < +. ~ lap - aql .
For arbitrary positive dE F, it is now obvious that
implies that
Hence (x,,) E <r. 0

(5.23) Notation. The field <rIm will be written as F. Throughout
(5.23-5.30), elements (a,,) + mof <rIm will be denoted by small Greek
letters: oc, /3, .... If a EF, then the element (a(,,») + of F will be m
written as a; it is the coset of m
containing the constant sequence all
of whose terms are a.
(5.24) Theorem. In F, let P = {oc EF: oc =1= 0 and there exists (a,,) E oc
such that a" > 0 tor n = 1,2,3, ... }. With this set P, F is an ordered
field in the sense ot (5.7). The mapping 1': l' (a) = a is an order-preserving
algebraic isomorphism ot F into F.
The proof of this theorem is left to the reader.
(5.25) Definition. Given a sequence (a,,) in an ordered field F and
bE F, we say that the limit ot (a,,) is b, and we write
lim a" = b or a" -+ b ,
......... 00
if for every positive e in F there exists a positive integer L (e) such that
la" - bl < e for all n ~ L (e). An ordered field is said to be complete if
every Cauchy sequence in F has a limit in F.
(5.26) Lemma. A sequence with a limit is a Cauchy sequence. It (a,,)
is a Cauchy sequence and (a"k)%"= 1 (1 ~ n l < n 2 < ... < nl< < ... ) is a
subsequence with limit b, then (a,,) has limit b.
Proof. The first assertion is trivial. To prove the second, choose
e> 0 in F. If L( ~ e) ~ k,
then we have la"k - bl < ~ e. Since (a,,)
is a Cauchy sequence, we have
1
lap - aql <Z-e
if p, q ~ N( ~ e). Choose any fixed k such that k ~ L (~ e) and

nk ~ N (~ e). Then for q ~ N (~ e), we have
laq - bl ~ laq - ankl + lank - bl < e. 0
(5.27) Lemma. For oc > 0, oc EF, there exists e EF such that 0 < e < oc.
It F is Archimedean ordered, then F is also Archimedean ordered.
Proof. Since oc > 0, there exists (an) E oc such that an> 0 for
n = 1, 2, 3, . .. and (an) ~ m. Hence there is a d EF such that
as = lasl ~ d for arbitrarily large s. We have
1
lap - aql <zd
if p, q ~ N(~ d). Choose an s as above such that s ~ NUd);

then we have
1
d ~ as = as - ap + ap ~ las - apl + lapl < Z d + ap;
i. e., ~ d < ap whenever p ~ N (~ d). It follows that (an - ~ d(n») +m
-}- -
= oc - --Xd ~ o. We have
-}- -}--
oc---d>
3 oc--d'2.
2- 0 ,
and so
-}-
oc > -ad.
Hence e = ~ d satisfies the first assertion of the theorem.
Suppose that F is Archimedean ordered, and let oc, {3 EF be such that
O<oc~{3.
By the first assertion of the theorem, there is a positive e EF such that

e< oc. Since every Cauchy sequence is bounded, there is a d EF such
that {3 < d. If m is any positive integer such that me > d, then we have
moc> me > d> {3;
hence moc > {3, and so F is Archimedean ordered. 0
(5.28) Lemma. Let oc EF and (an) E oc. Then we have
Proof. Choose anye > 0 in F and any e > 0 in F such that 0 < e < e;
this is possible by (5.27). We have
lap - aql < e if p, q ~ N (e) .
Now fix P 6 N(e). For n 6 N(e) we have

a/> - an < e,
an - a/> < e.
It follows that
a/> - IX;:;;; e < 8,
IX - a/> ;:;;; e < 8 ;
i. e.,
!IX-a/>!<8 if p6N(e). 0
We now state and prove our main result about F.
(5.29) Theorem. The field F is complete.
Proof. Let (IX/» be any Cauchy sequence in F. If (IX/» is ultimately
constant, there is nothing to prove; if not, there exists a subsequence
(1X/>/)i..1 such that IX/>l =1= IX/>l+1 for 1 = 1,2, .... By (5.26), it suffices to
prove that (IX/>/)i_ 1has a limit. Hence we suppose with no loss of general-
ity that (IX/»P'_I is such that IX/> =1= IX/>+1 (P = 1, 2, 3, ... ). Write
o < !IX/> - IX/>+1! = ft/>.
For all 8 > 0 in 'P, there exists N(8) such that
!IX/> - IXIl! < 8 if p, q 6 N(8);
in particular,
ft/> < 8 if P 6 N (8) .
Using (5.28), we choose a/> EF such that !a/> - IX/>! < ft/> (P = 1,2,3, ... ).
Now choose any e > 0 in F. For p, q 6 N (~ we have e),
!a/> - all! ;:;;; !a/> - IX/>! + !IX/> - IXIl! + !IXIl - all!
1_ 1_ 1_ 1 __
< ft/> + 3" e + ftll < 3" e + 3" e + 3" e = e .
Since the mapping or of (5.24) is an order-preserving isomorphism, it
follows that !a/> - all! < e if p, q 6 N( ~ e), i. e., (a/» E~. Define p
as (a/» + m.
We claim that lim IX/> =
P-+oo
p. To prove this, choose any positive 8 in 'P.
For p 6 N (~ 8), we have !a/> - IX/>! < ft/> < ~ 8. Also (5.28) shows
that there is a positive integer M (~ 8) such that !a/> - P! < ~ 8for
p6 M (~ 8). Hence
!IX/> - P! ;:;;; !IX/> - a/>! + la/> - P!
1 1
<2'8+2'8=8
if p6 max {N (~ 8), M(~ 8)}. 0

(5.30) Theorem. For any ordered field F, F is isomorphic with F;

every Cauchy sequence 01 F differs Irom a constant sequence by a null
sequence.
Proof. For a Cauchy sequence (lXp) of elements of F, let f3 = lim IXp
p-..oo
(5.29). Then (lXp - f3(p») is a null sequence, and the theorem follows. D
(5.31) Theorem. In any Archimedean ordered field, the sequence
(2- P )P'= 1 is null.
Proof. We have 2 P =k~ (~) > p, and so 2- P < ; . Since (; );=1

is null, (2- P) is also null. D
(5.32) Definition. Let F be an ordered field and 0 A C F. An s;:
element bE F is said to be an upper [lower] bound lor A if x ~ b [x;;;:; b]
for all x EA. An upper bound b is called the least upper bound or supremum
of A, and we write b = supA, if b is less than all other upper bounds for A.
The greatest lower bound or infimum 01 A, written infA, is defined analo-
gouslyl. The notations l.u.b.A and g.l.b.A. are sometimes used for what
we call supA and infA.
(5.33) Theorem. Let F be a complete Archimedean ordered field,
and let A be a nonvoid subset 01 F that is bounded above [below J. Then
supA [infA] exists.
Proof. Let b be any upper bound for A, and let a E A. There exist
positive integers M and -m such that M> band -m> -a, i.e.,
m < a ~ b < M. For each positive integer p, let
Sp = {k : k is an integer and ;p is an upper bound for A} .

If k ~ 2P m, then k is not in Sp. Thus Sp is bounded below. Since we have
2 P ME Sp, Sp is nonvoid. It follows that Sp has a least element, say kp.
We define a p = ~: (P = 1, 2, 3, ... ). By the definition of k p , :P':l = ~:
is an upper bound for A and 2:+1
2k - 2
=
k -1
~ is not. Therefore we have
either
so that
2kp 2k p- 1 1
ap+ l = 2 p+1 = ap or ap+ l = 2 p+1 = ap - 2 p+1 '
and hence
(p = 1,2,3, ... ).
1 Clearly we can also define suprema and infima in arbitrary partially ordered
sets.
If q > P ~ 1, then
o~ ap - all = (ap - aH1 ) + (a HI - aH2 ) + ... + (a ll - 1- all)
1
~ 29+1 + 29+1 + ... + 20
2
1
= 1(
29+1 1 + "2 + ... +
1 1)
2.-9-1
= 21'1+1 ( 2 - 2.-1'-1
1) 1
< 2P .
1
We thus have lap - alll = ap - all < 2P whenever q > p ~ 1. From (5.31)
we infer that (a p ) is a Cauchy sequence, and so lim ap exists; call it c.
p...... oo
It is plain that ap ~ c.
We claim that supA = c. To prove it, assume first that c is not an
upper bound for A. Then there is an x E A such that x > c, and hence
there is a positive integer p such that ap - c = lap - cl < x - c; i. e.,
ap < x. Since ap is an upper bound for A, the last inequality cannot
obtain. Therefore c is an upper bound for A. Assume next that there
exists an upper bound c' for A such that c' < c, and choose a positive
integer p such that ;1' < c - c'. We then have ap - ;1' ~ c- ;1'
> c + c' - c = c', and so ap - ;1' is an upper bound for A. However,
ap -2P . b Y d efin"Ihon -2-1'

1 IS -, and ~
kl'- 1
IS not an upper b ound for A .
kl'- 1 .
It follows that c = supA.

A similar proof can be given that infA exists if A is bounded below;
or it can be shown that
infA = - sup(-A). 0
(5.34) Theorem. Any two complete Archimedean ordered fields Fl
and F 2' with sets 01 positive elements PI and P 2' respectively, are algebraically
and order isomorphic, i. e., there exists a one-to-one mapping -r 01 F1 onto F2
such that
-r(x + y) = -rex) + -r(y) ,
-r(xy) = -rex) -r(y) ,
-rex) E P 2 it and only it x E PI'
Proof. Let 11 and 12 be the units of F1 and F 2 and 01 and O2 the zeros.
The mapping -r [d. (5.6.e)] is first defined on the rational elements of F 1 ;
thus:
-r(11) = 12 ;
-r(01) = O2 ;
-r(m11) = m12' where m is an integer;
-r (~
n 11) = ~
n 12' where n is a nonzero integer,'
-r (: 11) = : 12 ,
If X EFl and X is not of the form ~

n 11' then we define
T(X) = sup {: 1.: : 11 < X} .

It is left to the reader to prove that T has the desired properties. 0
(5.35) Definition. The real number field is any complete Archimedean
ordered field; e. g., Q. We will always denote this field by R.
(5.36) Exercise. Let F be any ordered field. For a, bE F, prove that
1
max {a, b} = 2' (Ia - bl + a + b) ,
min {a, b} = ! (-Ia - bl + a + b) .

(5.37) Exercise. Let F be any ordered field, and let a, b, c be any
elements of F. Define
mode{a, b, c} as min{max{a, b}, max{b, c}, max{a, c}} .
Describe the mode in words, and write it in terms of absolute values and
the field operations.
(5.38) Exercise. Let F be any ordered field. A subset D of F is called
a Dedekind cut in F if:
(i) 0 S D SF;
(ii) the relations xED and y < X imply y ED.
(a) Let D be a Dedekind cut in R. Prove that D = {x ER: x < a} for
some a ER or D = {x ER : x ~ a} for some a ER.
(b) If F is an ordered field not order isomorphic to R, prove that F
contains a Dedekind cut that is of neither of these two forms.
(c) Using (a) for the field R, prove that every positive real number
has a unique positive kt11. root (k = 2, 3, 4, ... ).
(5.39) Exercise. Consider the field of all rational functions with
coefficients in Q in a single indeterminate t, and denote this field by the
.
symbol Q (t). Thus a genenc nonzero element of Q (t) has the form
AOO
B (t) ,
,. m
where A (t) = L aktk and B (t) = L biti. The numbers ak and bi are in Q,
k=O ;=0
and an =1= 0 and bm =1= O. Addition and multiplication are defined as usual.
We order Q(t) by the rule that ~~:~ is in P if and only if anbm is a
positive rational number. Prove that Q (t) is an ordered field and that the
order is non-Archimedean. Prove also that every non-Archimedean ordered
field contains a subfield algebraically and order isomorphic with Q (t).
Find the completion of Q (t).
(5.40) Exercise. Let (an):'= 1 be any sequence of integers all greater
than 1. Prove that every real number x such that 0 ~ x < 1 has an
expansion of the form

00
~ x~
~ a1 a I "'a ~ '
k=1
where each Xk E{O, 1, ... , ak - I}. Find a necessary and sufficient

condition for two distinct expansions to be the same real number.
We next construct the complex number field, a much simpler process
than our construction of R.
(5.41) Theorem. Consider the ring R [t] of all polynomials in the
indeterminate t with coefficients in R, and with addition and multiplica-
tion defined as usual. Let I = {(t2 + 1) P (t) : p (t) E R [tn. Then I is an
ideal in R [t]. Let R [t]/J be the set of cosets p (t) + I. Addition and multi-
plication in R [t]/J are defined by
(P(t) + J) + (q(t) + J) = (P(t) + q(t)) + I
and
(P(t) + J) (q(t) + J) = (P(t) q(t)) + I .
These definitions are unambiguous, and with aclclition and multiplication
so defined, R [t]/J is a field.
Proof. It is obvious that I is an ideal in R [t]. Exactly as in the proof
of (5.22), we see that the definitions of addition and multiplication in
R [t]/J are unambiguous, and that R [t]/J is a commutative ring with
zero I and unit 1 + I.
To describe R [t]/J more closely, suppose that (a + bt) + I
= (a' + b't) + I. Then [(a + bt) - (a' + b't)] EI, and so (a - a') + (b-b')t
= (t s + I)P(t), for some P(t)ER[t]. Comparing the degrees of these
two polynomials, we see that P(t) = 0 and that a = a', b = b'. In other
words, each element of the set {(a + bt) + I : (a, b) E R x R} is a
distinct element of R [t]/J. It is an elementary algebraic fact, whose
proof we omit, that every p (t) E R [t] can be written
P(t) = (t S + 1) q(t) + r(t),
where q(t) E R [t] and ret) = a + bt. Thus the coset p (t) + I is equal to
(t 2 + 1) q(t) + (a + bt) + I = (a + bt) + I. This proves that R [t]/J
= {a + bt + I: (a, b) E R x R}, where distinct pairs (a, b) yield distinct
elements of R [t]lI.
Routine computations show that
((a + bt) + J) + (a' + b't) + J) = ((a + a') + (b + b') t) + I
and that
(a + bt) + J)(a' + b't) + J) = ((aa' - bb') + (ab' + a'b)t) + I.
If (a + bt) + I =1= I, then a =1= 0 or b =1= O. Since R is an ordered field, we
have a 2 + b2 > 0, and so a' : b' exists in R. It is clear that
[(a + bt) + JJ [C.: b' - a': b' t) + J] = 1 + J.

This shows that every nonzero element of R [tJ/J has a multiplicative
inverse, and so R [tJ/J is a field. 0
(5.42) Definitions. The field R [tJ/J is called the complex number field
or the field of complex numbers and is denoted by the symbol K. We
write the coset (a + bt) + J as a + bi; a + bi is called a complex number.
The number a is called the real part of a + bi and is written Re(a + bi).
The number b is called the imaginary part 01 a + bi and is written
1m (a + bi). The symbols z = x + iy, w = u + iv, C1 + iT, (X + pi,
etc., will be used to denote complex numbers. The complex number
a + Oi will be written as a alone and 0 + bi as bi alone. For z =
1
X+ iy EK, the absolute value 01 z is defined as (x2 + y2)"i [the nonnegative
square root!J and is written Izl. The complex conjugate 01 z [or simply
conjugateJ is defined as x - iy and is written 'i.
(5.43) Theorem. The field K cannot be ordered.
Proof. Assuming the existence in K of a subset P as in (5.7), we have
i E P or -i E P. If i E P, then i2 = -1 E P, which contradicts (5.8).
If - i E P, then (_i)2 = -1 E P, also a contradiction. 0
(5.44) Theorem. For all z, Zv Z2 E K we have:
(i) z= z;
(ii) Zl + Z2 = 'i1 + 'i2;
(iii) Zl Z2 = Z1 Zs •
Proof. Routine calculation.
(5.45) Remark.1 The foregoing theorem shows that conjugation is an
automorphism of K. The field R has no automorphisms save the identity.
In fact let cp be a function with domain R, range contained in R,
cp (R) =!= {O}, and such that cp (x + y) = cp (x) + cp (y), cp (xy) = cp (x) cp (y).
It is easy to show that cp (1) = 1, cp (0) = 0, and in general that cp (r) = r
°
for all r E Q. If x =!= and cp (x) = 0, then
l=cp(I)=cp(x :)=cp(x) cp(:)=o.

Hence cp (x) =!= °if x =1= 0. If a < b, then
cp (b) - cp (a) = cp (b - a) = cp (((b - a)!)S) = (cp((b - a)t))2 > 0.
Hence cp (a) < cp (b) if a < b. For an arbitrary real number x, choose
r1 , r 2 E Q such that r1 < x < rs, Then
r1 = cp(r1) < cp(x) < cp(r2) = r2 .
1 Subheads (5.45) and (5.46) are included only for cultural interest and are not
referred to in the sequel.
Since r2 - r i can be made arbitrarily small, it follows that IP (x) = x.

(5.46) The functional equation IP (x + y) = IP (x) + IP (y) has 2C dis-
continuous solutions on R. In fact, regard R as a vector space over Q
(3.17.c) and let E be a Hamel basis for Rover Q (3.19). For each x E R
let IX", denote that unique function from E into Q as in (3.20) such that
x = E IX", (b) b. Now for each t ERB define 1P,: R -+ R by the rule
bEB
IPI (x) = E IX", (b) t (b) .
bEB
The reader can easily verify that each such IPI satisfies the desired
functional equation and that IP/(rx) = r IPI (x) for r E Q, x E R. Thus
IPI(r) = rlPl(l) (r E Q), so that if IPI is continuous, then IPI(X) = xlPl(l) for
all x E R. Since 1P1(1) has just c possible values we see that there are just C
continuous IP/s. But RB = CC = 2c [see (4.34)J and t =l= gin RB implies
IPI =l= lPg, so there exist 2C discontinuous IP/s. The preceding paragraph
shows that the additional requirement that IP (xy) = IP (x) IP (y) forces IP
to be continuous.
To illustrate the bizarre nature of some of these additive functions,
define 1p (x) = E IX", (b) for x E R, i. e., 1p = IPI where t (b) = 1 for each
bEB
bEE. Now consider bi =l= b2 in E, C < din R, and r E Q.
Next choose sEQ such that c < rb i + s (bi - b2) < d. Let
u= rb i + s(b i - b2) = (r+ s) bi - sb 2.
Then c< u< d and 1p(u) = (r + s) -s=r. Therefore c< din R implies
that 1p ({x: c < x < d}) = Q. This function is wildly discontinuous.
The field K has 2C automorphisms. This fact depends on the fact that
K is algebraically closed. Only the identity z -+ z and conjugation
z -+ % are continuous in the usual topology on K (6.17).
(5.47) Theorem. For z, w E K we have Izwl = Izl • Iwl, Izl = 1%1,
Izl2 = zz, Z + z = 2 Re(z), and z - z = 2i 1m (z).
Proof. Computation.
(5.48) Lemma. Let z = x + yi be a complex number. Then IRe(z)1 ~ Izl,
and Re (z) = Izl it and only it x ~ 0 and y = O. Also 11m (z) I ~ Izl, and
Im(z) = Izl it and only it x = 0 and y ~ O.
Proof. The following relations are evident:
1 1
-Izl = - (x 2 + y2)"2 ~ - (X2)"2 = -Ixl ~ x
1 1
~ Ixl = (X2)"2 ~ (x 2 + y2)"2 = Izl .
1
Clearly x = (x 2 + y2)"2 if and only if y = 0 and x ~ O. The proof for
1m (z) is the same. 0
(5.49) Theorem. For z, wE K, we have Iz + wi ~ Izl + Iwl, and
equality holds it and only it IXZ = {3w, where IX and {3 are nonnegative real
numbers not both zero.
Proof. Applying (5.47) and (5.48), we write

Iz + wl 2 = (z + w) (z + w) = zz + ww + zw + zw
= Izl2 + Iwl2 + 2 Re (zw)
~ IZl2 + Iwl2 + 21zwl
= Izl2 + Iwl 2 + 21zllwl
= (Izl + IWI)2.
This shows that Iz + wi ~ Izl + Iwl. Equality holds if and only if
Re(zw) = Izwl, and so by (5.48) if and only if zw is a nonnegative real
number. If z = 0, take oc = 1 and {3 = O. If w = 0, take oc = 0 and {3 = 1
If z =1= 0 and w =1= 0 and zw is a positive real number {3, then zlwl 2 = zww
= {3w, and we can take oc = Iw12> O. 0
(5.50) Geometric interpretation. As the reader will already know,
the field K can be very usefully regarded as the Euclidean plane R x R,
in which the point (a, b) corresponds to the complex number a + bi.
Thus the Euclidean distance between (a, b) and (0,0) is the absolute
value of a + bi. Conjugation is simply reflection in the X-axis.
(5.51) Definition. Let z = x + iy be a complex number different
from O. Then arg (z) is the set of all real numbers 0 such that
cos (0) = 1;1 and sin (0) = I~I •
Any element 0 of arg(z) such that -n< 0 ~ n will be denoted by
Arg(z). We define arg(O) = R and do not define Arg(O).
(5.52) Theorem. For every nonzero complex number z, arg(z) is a
countably infinite set, and Arg(z) contains exactly one real number. It
o E arg(z), then arg(z) = {O + 2nn: n EZ}.
Proof. We only sketch the proof; details may be found, for example,
in SAKS and ZYGMUND, Analytic Functions, pp.62-64 [Monografie
Matematyczne, Warszawa, Vol. 28 (1952)]. The real-valued functions
•
SIll (x) = ..~

00 (-I)" x2n+1
(2n +
1) I and cos (x) = ..fn 00 (-I)" x2n
(2n)!
are defined, continuous, and in fact infinitely differentiable for all x E R.

In particular, sin (0) = 0 and cos (0) = 1. The number; is defined as
the least positive zero of cos. One then proves that for every pair (c, d)
of real numbers such that c2 + d2 = 1, there is a unique real number 0
such that - n < 0 ~ n, cos (0) = c, and sin (0) = d. This number is
Arg(z). One also shows that cos (0) = cos (0 + 2nn) and sin (0)
= sin (0 + 2nn) for all 0 E Rand nEZ, and that 2n is the smallest
period of cos and of sin. These facts imply the last two statements of
the present theorem. 0
(5.53) Exercise. For a nonzero complex number z, prove that
arg ( : ) = - arg (z). If z is not a negative real number, prove that
Arg (-H = -Arg(z) . If z is a negative real number, prove that Arg(z)

=Arg(+) = n.
(5.54) Exercise. For z and w nonzero complex numbers, prove that
arg (zw) = arg (z) + arg (w). For every positive integer k, arg (Zk)
= k . arg (z) .
(5.55) Recapitulation.
In the accompanying Figure 3 we illustrate addition and multiplica-
tion of complex numbers. Addition is componentwise; graphically, one
applies the parallelogram law for addition of vectors. Multiplication is
a little more complicated. We have
Arg(zw) = Arg(z) + Arg(w) modulo 2n, and Izwl = Izl ' lwl .
iY
------- -- --- -- --- / /

_-tZ-fW
/
/
/
Fig. 3
In Figure 4, we illustrate the only conditions under which IZI + z21

= IZII + IZ21 (5.49) and also the position of z relative to z.
(5.56) Exponential notation. A sequence (zn) of complex numbers
converges to a limit z if lim Iz - znl = O. [We shall have more to say on
,,-+00
this subject in § 6.] For the moment, we use it to define the exponential
function exp by
00 n k n
exp(z) = I; -;- = lim I; -;- .

,,=0 n k-+oo ,,=0 n .
Just as with real power series, one proves that exp(z) exists [i. e., the
limit exists] for all z EK. The identity
exp (z + w) = exp (z) exp (w)
4*
holds and is proved by multiplying out ( .. ~ ~~ ) (m~o :~) and taking

the limit as k --* 00. It is easy to show that
exp(iO) = cos (0) + i sin (0)
for all 0 E R and so Jexp (i 0) J = 1. Every nonzero complex number z
can thus be written as
z = JzJ (i) = JzJ exp(iO) = Jzl (cos (0) + i sin (0)) .

Here 0 is any number in arg(z).
iY
\
\
\
I
/
I
I
The function z --* i '
defined implicitly above, is used frequently.
It is called the signum and is defined formally by
sgn (z) = (
-IZI if z EK
Z
n {O}',
o if z = O.
(5.57) Exercise. Use Hamel bases to prove that the additive groups R
and K are isomorphic.
(5.58) Exercise. Define addition in K as usual and define multipli-
cation "coordinatewise":
(x + iy) (u + iv) = xu + iyv .
Prove that, with these operations, K is a commutative ring with unit.
Prove also that K is not a field.
CHAPTER TWO
Topology and Continuous Functions

The main goal of this text is to give a complete presentation of
integration and differentiation. Plainly a detailed study of set-theoretic
topology would be out of place here. Similarly, a detailed treatment of
continuous functions is outside our purview. Nevertheless, topology and
continuity can be ignored in no study of integration and differentiation
having a serious claim to completeness.
First of all, there is an intimate connection between measure theory
[which is almost coextensive with the theory of integration] and the
topological notion of compactness. Second, many important facts in the
theories of integrals and derivatives rest in the end on properties of
continuous functions. Third, purely topological notions play a vital
part both in constructing the objects studied in abstract analysis and
in carrying out proofs. Fourth, a great many proofs are just as simple
for arbitrary topological spaces as they are for the real line.
Therefore, in asking the reader to consider constructions involving
topological spaces far more general than the line, we ask for a not in-
considerable preliminary effort, as the length of § 6 will show. In return,
we promise a much more thorough presentation of contemporary analysis.
Section 6 is a self-contained if rather terse treatment of those parts
of set-theoretic topology that have proved important for analysis. With
some reluctance we have omitted the topics of paracompactness and
compactifications of completely regular spaces. But a line had to be
drawn somewhere. In § 7, we embark on a study of continuous functions
and of functions closely related to continuous functions. We are partic-
ularly concerned with spaces of such functions and properties that they
may have. The section culminates with the STONE-WEIERSTRASS theorem,
surely an indispensable tool for every analyst.
§ 6. Topological preliminaries
Set-theoretic topology is the study of abstract forms of the notions
of nearness, limit point, and convergence. Consider as a special but
extremely important case the line R. For x, y ER, we can define the
distance between x and y as the absolute value of x - y:
e(x,y) = Ix - yl·
54 Chapter II. Topology and continuous functions
We can also associate with a point x ER all of the points y E R such that
e (x, y) is less than a specified positive number ex, i. e., the open interval
]x - ex, x + ex[. This gives us a systematic notion of nearness in R
based on the distance-function e. Even a more general idea is needed.
A subset A of R is called open if it contains all points "sufficiently close"
to each of its members. The abstract and axiomatized notion of open
set is one common and convenient way to approach the study of set-
theoretic topology.
To make all of this precise, we begin with some definitions. We
first extend the real number system, and describe certain important
subsets of the extended real numbers.
(6.1) Definition. (a) Let 00 and - 00 be two distinct objects, neither
of which is a real number!. The set R# = R U {-oo} U {oo} is known
as the set of extended real numbers. We make R# a linearly ordered set
by taking the usual ordering in R and defining - 00 < 00 and - 00< x< 00
for each x E R. For a and b in R# such that a < b, the sets
]a,b[={xER#:a< x< b},
[a, b] = {x E R#: a ~ x ~ b},
[a, b[ = {x E R#: a ~ x < b},
and
]a, b] = {x E R#: a < x ~ b}
are called intervals, with endpoints a and b. The interval ]a, b[ is an
open interval, [a, b] is a closed interval, and [a, b[ and ]a, b] are hal/-
open intervals. If a and b are in R, these intervals are said to be bounded.
Otherwise they are said to be unbounded. Note that R itself is the open
interval ]- 00, oo[ and that all intervals under our definition have
cardinal number c.
(b) For future use, we define sums and products in R#, with a few
restrictions, by the following rules. For x, y ERe R#, x + y and xy
are defined as usual. For x ER, we define:
oo+x=x+oo=oo;
(-00) + x = x + (-00) = x -00 = -00;
we also define
00+00=00;
(-00) + (-00) = -00 - 00 = -00;
-(00)=-00 and -(--00)=00;
the expressions 00 + (- 00) and (- (0) + 00 are not defined. For x E R
1 Many writers use the symbol + 00 for what we write as 00. The + sign is a
mere nuisance and so we omit it.
§ 6. Topological preliminaries 55
and X > 0, we define:

OO'x=x'oo=oo;
(-00)' x = X· (-00) = -00.
We define
and
00'00=00.
For x ERand x< 0, we define
OO'x=x'oo=-oo;
(-00)' x = X· (-00) = 00;
the expressions 00' (- 00), (- 00) . 00, and (- 00) . (- (0) are not defined.
Open subsets of R are defined in terms of open intervals, as follows.
(6.2) Definition. A set U C R is said to be open if for each x E U
there is a positive real number e such that ]x - e, x + e[ C U.
Thus we may say informally that a subset U of R is open if for each
of its points x, it contains all points y such that y is sufficiently close to x.
[The "sufficiently close" depends of course on x.] Plainly every open
interval is an open set. Some important properties of open subsets of
R are listed in the following theorem.
(6.3) Theorem. Let () denote the family of aU open subsets of R. Then:
(i) 0 E () and R E ();
(ii) if o/t is a subfamily of (), then U o/t E () l;
(iii) if {Ul , Us' ... , Un} C (), then Ul n U2 n ... n Un E ().
The reader can supply the proof of this very simple result.
The properties of open sets in R given in Theorem (6.3) form the
basis for the concept of a topological space, which we now define.
(6.4) Definition. Let X be a set and () a family of subsets of X with
the following properties:
(i) 0 E () and X E ();
(ii) if o/t is a subfamily of (), then U o/t E ();
(iii) if {Uv Us, ... , Un} C (), then Ul n Us n ... n Un E ().
[That is, () is closed under the formation of arbitrary unions and of
finite intersections.] Then () is called a topology for X and the pair (X, ())
is called a topological space. When confusion appears impossible, we will
call X itself a topological space. The members of () are called open
sets in X.
Definition (6.4) by itself is rather barren. No great number of excit-
ing theorems can be proved about arbitrary topological spaces. However,
certain entities definable in terms of open sets are of considerable interest
in showing the connections among various topological concepts and in
1 Recall that the union of a void family is 0.
showing the simple nature of such ideas as continuity. Also by the ad-
dition of only two more axioms, we obtain a class of topological spaces
in which many of the processes of analysis can be carried out with great
profit. We first give a few examples of topological spaces.
(6.5) Examples. (a) Naturally, the real line R with the topology
described in (6.2) and (6.3) is a topological space. This topology for R
is known as the usual topology for R. We will always suppose that R
is equipped with its usual topology unless the contrary is specifically
stated.
(b) Consider R* and the family l!J* of all sets having any of the
followingfourforms: U, U U ]t, 00], U U [-00, s[, U U [-00, s[U]t,oo],
where U is an open subset of Rand s, t E R. Then (R*, l!J*) is a topo-
logical space; l!J* is called the usual topology for R*.
(c) Let X be any set. The pair (X, .9'(X)) is a topological space,
obviously. The family .9'(X) is called the discrete topology for X. A set X
with the discrete topology will frequently be denoted by XII'
(d) Let X be any set and let l!J = {X, 0}. Then l!J is called the in-
discrete [or concrete] topology for X. This topology is of little interest to us.
We proceed to the definition of various topological concepts from
our basic notion of open sets.
(6.6) Definitions. Let X be a topological space. A neighborhood of a
point x EX is any open subset U of X such that x E U. The space X
is known as a Hausdorff space if each pair of distinct points of X have
disjoint neighborhoods. A set A C X is said to be closed if X n A' is
open. For x E X and A C X, we say that x is a limit point of A if
(U n {x}') n A =1= 0 for each neighborhood U of x. For A C X, the
closure of A is the set A - = n {F : F is closed, A C F c X}; the interior
of A is the set A °= U {U: U is open, U C A}; and the boundary of A
is the set oA = A- n (A')-.
(6.7) Theorem. Let X be a topological space.
(i) The union of any finite collection of closed subsets of X is a
closed set.
(ii) The intersection of any nonvoid family of closed subsets of X is a
closed set.
(iii) The closure A-of a subset A of X is the smallest closed set con-
taining A, and A is closed if and only if A = A-.
(iv) The interior A ° of A is the largest open set contained in A,
and A is open if and only if A = A 0.
(v) A subset A of X is closed if and only if it contains all of its limit
points.
For subsets A and B of X, we have:
(vi) AO = A'-';
(vii) oA=A-nAo,;
(viii) (A U B)- = A- U B-;

(ix) (A n B)O = AO n BO.
For an arbitrary family {A.} of subsets of X, we have:
(x) U A~ C (U A,)-;
(xi) n A~ ::J ( n A,)O .
Finally,
(xii) 121 and X are closed.
Proof. Assertions (i) and (ii) follow at once from de MORGAN'S laws
(1.9.iii) and (1.9.iv) applied to axioms (SA.iii) and (SA.ii) for open sets.
Since A- is the intersection of all closed supersets of A, assertion (ii)
proves (iii). Assertion (iv) is all but obvious.
We next prove (v). Suppose first that A is closed. Then A' is a neigh-
borhood of each point in A I and A I n A = 121, so that no point of A I
is a limit point of A, i.e., A contains all of its limit points. Conversely,
if no point of A' is a limit point of A, then for each x E A', there is a
neighborhood Ux of x such that Ux n A = 121, and therefore
A' = U {Ux : x EA'} is open, i.e., A is closed.
To prove (vi), we compute as follows:
A'-' = [n{F: F is closed andF::J A'}]'
= U{F': F is closed andF::J A'}
= U{F': F' is open andF' C A}
=AO.
Assertion (vii) is immediate from (vi) and the definition of oA.
To prove (viii), notice that (A U B)- is a closed set containing both
A and B, so it must contain both A- and B-. Thus we have
(A U B)-::J A- U B-.
But A- U B- is a closed set containing A U B, so that
(A U B)-CA-U B-
and hence
(A U B)-= A-U B-.
To prove (ix), we write
(A n B)O = (A n B)'-'= (A' U B')-'
= (A'- U B',' = (A'-' n B'-')
= AO n BO.
Assertion (x) follows from the inclusions A,. C U A, and A::- C (U A,)-,
both of which are obvious for all indices '0. Assertion (xi) is obvious
from (x), and (xii) from (S.4.i) and the definition of a closed set. 0
(6.8) Definition. A topological space X is said to be connected if
121 and X are the only subsets of X that are both open and closed.
(6.9) Theorem. The space R with its usual topology is connected.

Proof. Let A be a nonvoid subset of R which is both open and closed.
Assume that A =1= R and let c ERn A'. Since A =1= 0, we have either
A n]-oo,c[=I= 0 or A n]c,oo[=I= 0. Suppose that B=A n]-oo,c[=I= 0
and let a be the supremum of this set (5.33). It is clear that a ~ c. If
e> 0, then a - e is not an upper bound for B, and so there is some x E B
such that a - e < x ~ a. This proves that every neighborhood of a meets
A so, since A is closed, a is in A. Since A is open, there is a (j > 0 such
that]a - (j, a + (j[ cA. Choose any bE R such that a a contradicts the choice of a. A similar con-
tradiction is obtained if A n ]c, 00 [=1= 0. We are thus forced to the
conclusion that A = R. 0
It is often convenient to define a topology not by specifying all
of the open sets but only some of them.
(6.10) Definition. Let (X, l!J) be a topological space. A family fJI C l!J
is called a base tor the topology l!J if for each U E l!J there exists some sub-
family de fJI such that U = Ud. That is, every open set is a union
of sets in fJI. A subfamily [/ of l!J is called a subbase tor the topology l!J
if the family of all finite intersections of sets in [/ is a base for the
topology l!J.
(6.11) Theorem. Let X be a set and let fJI C .9 (X). Define
l!J = {U d: d C fJI}. Then (X, l!J) is a topological space, and fJI is a base
tor l!J, it and only it
(i) UfJI = X
and
(ii) U, V EfJI and x E U n V imply that there exists W EfJI such that
xEWC un V.
Proof. Suppose that l!J is a topology for X. Then X E l!J, so there exists
de fJI such that X = Ude U fJI c X. That is, (i) is true. Next let
U, V be sets in fJI and let x E U n v. Then U n V is in l!J, so there is
some "If'" C fJI such that un V = U"If'". Thus we have x EWe un V
for some WE "If'". This proves (ii).
Conversely, suppose that (i) and (ii) hold. We must show that l!J
is a topology. Let {U.}IEI be any subfamily of l!J. Then, by the defini-
tion of l!J, for each t there exists d. C fJI such that U. = Ud.. [Here
we use the axiom of choice to choose just one d. for each tEl.] Let
d = lEI U d.. It is clear that de fJI and that Ud = lEI Uu.; thus l!J
is closed under the formation of arbitrary unions. Next let U, V be
in l!J. Then there exist subfamilies {U.}IEI and {~}fJEH of fJI such that
U = U U. and V = U ~. Thus for each x E un V, there exist tEl
lEI fJEH
and 1] E H such that x E U. n ~ and therefore, by (ii), there is a J¥s
n n
in 91 such that x Elfs c U. ~ c u v. Let .!II = {lfs: x E U V}. n
Then.!ll C 91 [.!IImay be voidl] and un V = U.!II E~. Thus ~ is closed
under the formation of finite intersections. According to (i) X is in ~,
and, since {21 C 91, we have {21 = U {21 E~. This proves that ~ is a topol-
ogy for X. Clearly 91 is a base for~. 0
The function (x, y) - Ix - yl defined on R x R is an obvious dis-
tance-function. An important although special class of topological spaces
are those in which the topology can be defined from a reasonable dis-
tance-function. The axiomatic definition follows.
(6.12) Definition. Let X be a set and let e be a function from X x X
into R such that for all x, y, z E X we have:
(i) e(x, y) ~ 0;
(ii) e(x,y) = 0 if and only if x = y;
(iii) e(x, y) = e(y, x) ;
(iv) e(x, z) ~ e(x,y) + e(y, z) [the triangle inequality].
Then e is called a metric [or distance-function] for X; e(x, y) is called
the distance from x to y, and the pair (X, e) is called a metric space.
When no confusion seems possible, we will refer to X as a metric space.
(6.13) Examples. (a) Let n be a positive integer, let X = RR or KR,
and let p be a real number such that p ~ 1. For re = (Xl' ... , xR ) and
y = (Yl' ... , YR) in X, define
I
ep(re, y) = C~ IXi - YiI P) P .
Properties (6. 12.i) - (6. 12.iii) are obvious for ep. The triangle ineqUality
(6.12.iv) is a special case of MINKOWSKI'S inequality, which we will
prove in (13.7) infra.
The metric e2 is known as the Euclidean metric on R!' or KR.
(b) For re, y E Rfl or KR define
e(re, y) = max{lxi - Yil : 1 ~ j ~ n}. It is easy to verify that (KR, e)
and (RR, e) are metric spaces.
(c) Let X be any set. For x, Y EX define e(x, y) = ~SY [~ is KRON-
ECKER'S ~-symbol as in (2.20)]. Plainly e is a metric. It is known as the
discrete metric for X.
(d) Consider the set NNo, which we realize in concrete form as the
set of all sequences (a,,)A= I of positive integers. For a = (a,,) and b= (b,,)
in NNo, define:
e(a, b) = 0 if a = b;
e(a, b ) =-;-I l'f

and aR =1= bR'
Then (NNo, e) is a metric space.
(e) Let D = {z E K: Izl ~ I} be the closed unit disk in the complex

plane. For z, wED define
_ {Iz - wi if arg(z) = arg(w) or one of z and w is zero,
e(z, w) - Izl + Iwl otherwise.
Then (D, e) is a metric space. This space is called the "French railroad
space" or the "Washington D. C. space". A picture should be sketched
to appreciate the reasons for these names. Actually this rather artificial-
looking space is [essentially] a certain closed subset of the closed unit
ball in a Hilbert space of dimension c. See (16.54) inlra.
(6.14) Definition. Let (X, e) be any metric space. For e > 0 and
x E X, let
B.(x) = {y EX: e(x,y) < e}.
This set is called the e-neighborhood 01 x or the open ball 01 radius e
centered at x.
(6.15) Theorem. Let (X, e) be a metric space. Let
fJlQ = {B.(x) : e > 0, x EX}.
Then fJlQ is a base lor a topology (!)Q lor X. We call (!)Q the topology generated
bye. The members 01 (!)Q are called e-open sets.
Proof. We need only show that fJlQ satisfies (6.ll.i) and (6. ll.ii).
Property (6.lLi) is obvious. Let B. (x) and B" (y) be in fJlQ and let
z E B.(x) n B,,(y). Then we have e(x, z) < e and e(y, z) < 6. Define
y = min{e - e(x, z), 6 - e(y, z)}.
Thus Y is positive, and for u E By (z), we have e (x, u) ~ e (x, z) + e (z, u)
< (e - y) + y = e and e(y, u) ~ e(y, z) + e(z, u) < (6 - y) + y = 6.
This proves that By(z) C B.(x) n B,,(y) and so (6.ll.ii) is satisfied. 0
(6.16) Remark. Restated slightly, (6.15) says that a set U C X is
e-open if and only if for each x E U there is an e > 0 such that y E U
whenever e (x, y) < e. When we make statements of a topological nature
about a metric space X, we will always mean the topology generated by
the given metric, unless we make some explicit statement to the contrary.
(6.17) Exercise. Let n be a positive integer and let X denote either
Rn or Kn. Prove that all of the metrics defined in (6.l3.a) and (6.13. b)
for X generate exactly the same topology for X, i. e., any two of those
metrics yield the same open sets. This topology is known as the usual
topology lor Rn [Kn].
Every subset of a topological space can be made into a topological
space in a natural way.
(6.18) Definition. Let (X, (!)) be a topological space and let S be a
subset of X. The relative topology on S induced by (!) is the family
{U n S: U E (!)} and the set S with this topology is called a subspace 01 X.
Thus a set V c 5 is relatively open if and only if V = U n 5 for some set

U that is open in X.
(6.19) Examples. (a) Let X = R [with its usual topology] and let
5 = [0, 1]. Then the set] ~ , 1] is open relative to [0, 1] since] ~ , 1]
=] ! ,2[ n [0, 1] and] ! ,2[ is open in R. However] ! ' 1] is obviously
not open in R.
(b) Let X = Rand 5 = Q. Then [V2, V3] n Q is open relative to Q
since [jl2, V3] n Q = ]V2, ]13 [ n Q.
(c) Consider the set L = {(x, x) : x E R} C {(x, y) : x E R, y E R} = R2.
With the usual topology for R2, L has the usual topology for R.
(d) Consider the set C= {( ~ cos (x), ~ sin (x) ) : xE R, 0< x< oo} C R2.
The relative topology of C in R2 [which has its usual topology] is the
usual topology of ]0,00[. [Identify C with ]0, oo[ in the natural way.]
We pass on to some additional important notions.
(6.20) Definition. A subset D of a topological space X is said to be
dense in X if D- = X. A space X is said to be separable if X contains a
countable dense subset. A space X is said to have a countable base if there
is a base for the topology of X which is a countable family.
(6.21) Example. The space Rn with its usual topology is separable
since the set D = {(xv ... , xn ) : xi E Q, 1 ;;;; j ;;;; n} is countable and dense.
However Ref is not separable since each countable subset [like all sub-
sets] of Ref is closed and R is uncountable. Also the French railroad space
(6.13.e) is not separable.
(6.22) Theorem. Any space with a countable base is separable.
Proof. Let X be a space with a countable base /14. For each nonvoid
BE /14 let XB E B. Then the set D = {XB: BE /14} is countable and
dense. 0
(6.23) Theorem. Any separable metric space has a countable base.
Proof. Let X be a metric space containing a countable dense subset D.
Let /14 = {Br(x) : xED, r E Q, r > O}. Then /14 is countable. To see that /14
is a base, let U be open and let z E U. Then there exists B > such that °
Be(z) C U. Since D is dense in X, there is an x E B./3 (z) n D. Now choose
a rational number r such that ! B < r< ! B. Then if y E Br (x), we have
e(y, z) ;;;; e(x, y) + e(x, z) < r + -31 B < B ,
so that Br(x) C B.(z) C U. Also

1
e(x, z) < "3 B < r ,
so that z E Br(x). Thus U is a union of members of /14. 0
(6.24) Definition. A sequence (Xn}:=l in a topological space X is

said to converge to an element x E X, or to have limit x, if for each neigh-
borhood U of x there exists a positive integer no such that Xn E U when-
ever n ~ no. We write lim Xn = x and also Xn -+ x if (Xn}:=l converges
to X. n~oo
(6.25) Theorem. A subset A of a metric space X is closed if and only

if whenever (xn ) is a sequence with values in A and (xn ) has limit x in X,
we have xE A.
Proof. Suppose that A is closed and let (xn ) be a sequence with values
in A for which a limit x in X exists. If x were in A', then A' would be a
neighborhood of x, and so all but a finite number of the values Xn would
lie in AI - a contradiction.
Conversely, suppose that A is not closed. Then by (6.7.v), A has a
limit point x such that x ~ A. For each n EN, choose Xn E A n B.!(x).
Then (xn ) C A, Xn -+ x, and x ~ A. D n
(6.26) Theorem. Let X be a Hausdorff space. Suppose that A ex

and that x is a limit point of A. Then each neighborhood of x contains
infinitely many points of A.
Proof. Exercise.
(6.27) Theorem. Every metric space is a Hausdorff space.
Proof. Exercise.
One of the most important concepts in topology is compactness.
There are several versions of this concept, which we next discuss.
(6.28) Definition. If (xn ) is a sequence and {nl < n 2 < ... < n" < ...}
is an infinite set of positive integers, then the sequence (xnk), defined
by k -+ x nk for kEN, is said to be a subsequence of (xn ).
(6.29) Definition. A topological space X is said to be sequentially
compact if every sequence in X admits a subsequence converging to some
point of X.
(6.30) Definition. A topological space X is said to be Frechet compact
[or to have the Bolzano-Weierstrass property] if every infinite subset of X
admits a limit point in X.
Sequential compactness and Fnkhet compactness are useful enough,
but the most useful notion of this sort is compactness alone, which we
now define.
(6.31) Definition. Let X be a topological space. A cover of X is any
family d of subsets of X such that Ud = X. A cover in which each
member is an open set is called an open cover. A subfamily of a cover
which is also a cover is called a subcover.
(6.32) Definition. A topological space X is said to be compact if each
open cover of X admits a finite subcover.
(6.33) Definition. A family of sets is said to have the finite inter-
section property if each finite subfamily has nonvoid intersection.
(6.34) Theorem. A topological space X is compact if and only if each

family of closed subsets of X having the finite intersection property has
nonvoid intersection.
Proof. This is nothing but an application of de MORGAN'S laws
(1.9). In fact, OIl is an open cover of X if and only if §'= {U' : U E OIl}
is a family of closed sets with void intersection. Thus every open cover
has a finite subcover if and only if every family of closed sets having void
intersection has a finite subfamily with void intersection. 0
(6.35) Theorem. Every compact topological space is Frechet compact.
Proof. Let X be a compact space. Assume that X has an infinite
subset A with no limit points in X. Then A is a closed set (6.7.v). Moreover
each a E A has a neighborhood Ua containing no point of A n {a}'.
Then {Ua : a EA} U {A'} is an open cover of X with no finite subcover.
This contradiction completes the proof. 0
(6.36) Theorem. Every sequentially compact metric space is separable.
Proof. Let X be a sequentially compact metric space. TUKEY'S
lemma (3.8) shows that for each positive integer n there is a maximal
subset A" of X having the property that e (x, y) ~ ~ for each pair
of distinct points x, yEA". Each A" is a finite set since otherwise, for
some n, A" would have an infinite sequence of distinct points with no
00
convergent subsequence. Thus the set A = ,,=1 U A" is countable. We assert

that A is dense in X. If this is not the case, then there exists an xE X n A - I .
Since A-I is open, there is an B > 0 such that Ba(x) C A-I. Choose n EN
such that n < B. Then we have e(x, y) ~ B > n for each yEA", and
1 1
the existence of the set A" U {x} contradicts the maximality of A".
It follows that A- = X. 0
(6.37) Theorem. Let X be a metric space. Then the following three
assertions are pairwise equivalent:
(i) X is compact;
(ii) X is Frechet compact;
(iii) X is sequentially compact.
Proof. The fact that (i) implies (ii) follows from (6.35). Suppose that
(ii) holds and let (x,,) be a sequence with values in X. If (x,,) has only
finitely many distinct terms, it is clear that there exists an infinite set
{n,,: kEN} C N such that nl < ns< ... and X"j; = x", for each kEN.
In this case the subsequence (X"k) converges to x",. Therefore we suppose
that (x,,) has infinitely many distinct values. Then the set {x,,: n EN}
has a limit point x EX. Let x", = Xl. Suppose that x"" ... , x"l: have
been chosen. Since each neighborhood of x contains infinitely many
distinct x,,'s, we choose x,," + 1 EB_l_(X) such that nHl > ni (1;;;; j;;;; k).
IHI
Then the subsequence (xnl:) converges to x. Thus (ii) implies (iii). Next
suppose that (iii) holds. According to (6.36) X is separable so, by (6.23),
X has a countable base &I. Now let !fI be any open cover of X. Let
.91 = {B E &I: Be U for some U E !fI}. For each BE.9I, choose UB E !fI
such that B C UB and let "f/' = {UB : BE .91}. Clearly "f/' is a countable
family. If x E X, then x E U for some U E !fI, and since &I is a base, there
is aBE &I such that x E B c U. Then BE.9I and x E B CUB' We
conclude that "f/' is a countable subcover of !fl. Enumerate "f/' in a se-
"
quence "f/' = (v,.). For each kEN let w,. = U v,.. To prove (i), we
n=1
need only show that w,. = X for some kEN. Assume that this is false.
For each k choose x" E X n W~. Then (x,,) has a subsequence (x"J)
converging to some x EX. Since "f/' is a cover there exists a ko EN such
that x E lIk. C w,. •. Thus w,.. is a neighborhood of x which contains x"
for only finitely many k. This contradiction establishes the fact that (iii)
implies (i). D
(6.38) Theorem. Let X be a Hausdorff space and let A be a subspace
of X that is compact in its relative topology. Then A is a closed subset
of X.
Proof. We will show that A' is open. Let zEA'. For each x E A
choose disjoint open sets Uz and Vz such that x E Uz , z E Yx. Then
{Uz n A : x E A} is an open cover of A, in its relative topology, so there
n n
exists a finite set {Xl' ... , x n } C A such that A C.U UZJ' Let V =.n VZj '
1=1 1=1
Then V is a neighborhood of z and V n A = 0, i.e., V CA'. D
(6.39) Theorem. Let X be a compact space and let A be a closed subset
of X. Then A is a compact subspace of X.
Proof. Let ofF be any family of closed [in the relative topology]
subsets of A having the finite intersection property. Then each member
of ofF is closed in X, so (6.34) implies that n ofF =1= 0. Thus A is compact,
by (6.34). D
We next present a striking characterization of compactness which
shows that we may restrict our attention to very special open covers in
proving that a space is compact.
(6.40) Theorem [ALEXANDER]. Let X be a topological space and let f/
be any subbase for the topology of X [see (6.10)]. Then the/following two
assertions are equivalent.
(i) The space X is compact.
(ii) Every cover of X by a subfamily of f/ admits a finite subcover.
Proof. Obviously (i) implies (ii). To prove the converse, assume that
(ii) holds and (i) fails. Consider the family )I{ of all open covers of X
without finite subcovers. The family )I{ is partially ordered by inclusion,
and plainly the union of a nonvoid chain in )I{ is a cover in)l{. ZORN'S
Lemma (3.10) implies that )K contains a maximal cover "Y. That is,
"Y is an open cover of X, "Y has no finite subcover, and if U is any open
set not in "Y, then "Y U {U} admits a finite subcover. Let "If" = "Y n f/.
Then no finite subfamily of "If" covers X, and so (ii) implies that "If" is
not a cover of X. Let x be a point in X n (U"If")" and select a set V
in the cover "Y that contains x. Since f/ is a subbase, there are sets
.
51' ... ,5" in f/ such that x E,n 5 i c V. Since x ~ (U "If"), no 5 i is in
1=1
"Y. Since "Y is maximal, there exists for eachj a set Ai which is the union
of a finite number of sets in "Y such that 5 i U Ai = X. Hence
.U
V U 1=1 Ai::::> (/1 5 i ) U (.U Ai)
1=1 1=1
= X,
and therefore X is a union of finitely many sets from "Y. This contradicts
our choice of "Y. 0
Another important class of topological spaces are those obtained
by taking the Cartesian product of a given family of topological spaces.
We need a definition.
(6.41) Definition. Let {X,},E! be a nonvoid family of topological
spaces and let X = X X, [see (3.l)J. For each, E I, define n, on X
lEI
by n , (~) = x,, The function n , is known as the projection of X onto X,.
We define the product topology on the set X by using as a subbase the
family of all sets of the form n;-1 (U;), where , runs through I and U;
runs through the open sets of X,. Thus a base for the product topology
is the family of all finite intersections of inverse projections of open sets.
A base for the product topology is the family of all sets of the form
XU;, where U; is open in X, for each, E I and U; = X, for all but a finite
'EI
number of the ,'so Whenever we discuss the Cartesian product of a family
of topological spaces, it is to be understood that the product is endowed
with the product topology unless the contrary is specified.
(6.42) Exercise. Let I = {I, 2, ... , n} for some n E N, and for each
,E I let X, = R [or KJ with its usual topology. Clearly X = 'EI X X, = R"
[or K"J. Prove that the product topology on X is the usual topology on X.
(6.43) TIHONOV'S Theorem 1. Let {X,},E! be a nonvoid family of
compact topological spaces. Then the Cartesian product X of these spaces
is compact [in the product topology J.
Proof. According to ALEXANDER'S theorem (6.40) it suffices to con-
sider open covers of X by subbasic open sets as described in (6.41).
Let CfI be any cover of X by subbasic open sets. For each, EI, let CfI,
denote the family of all open sets U C X, such that n;-I(U) E CfI. We
1 This theorem was proved by A. TIHONOV for the case in which each X, is the
closed unit interval [0,1] [Math. Annalen 102, 544-561 (1930)]. The general case
was first proved by E. CECH [Ann. of Math. (2) 38, 823-844 (1937)].
assert that U <¥t. = XI for some tEl. If this were not the case, there would
be a point ~ E X such that for every tEl, n l (~) = XI E XI n (U <¥t.)';
hence ~ ~ n;-l (U) for all n;-l (U) E tfI. That is, tfI would not be a cover of X.
Hence we can [and do] choose an 1] E I such that U tfl7/ = X7/' Since X7/
is compact, there is a finite family {Ut> ... , Un} C tfl7/ such that
X7/ = U1 U U2 U ... U Un. Plainly {nq-1(Ui):j = 1, ... , n} is a finite
subcover of tfI for X. 0
We next characterize the compact subspaces of R.n and Kn.
(6.44) Theorem [HEINE-BoREL-BoLZANO-WEIERSTRASS]. Let n EN
and let A eRn [or Kn]. Then A is compact [in the relativized usual
topology] it and only it A is closed and bounded 1.
Proof. The mapping X + iy -+ (x, y) of K onto R2 preserves distance:
1
I(x + iy)- (u + iv)l = «(x - U)2 + (y - V)2)2" = e«(x, y), (u, v)). Thus
Kn and R2n are indistinguishable as topological spaces. We therefore
restrict our attention to the case in which A eRn.
We first take the case that n = 1 and A = [a, b], a bounded closed
interval in R. A subbase for the topology of [a, b] is the family [/ of
all intervals of the form [a, d[ or ]c, b] where c, dE [a, b]. Let tfI be any
cover of [a, b] by sets in [/. Since b is covered by tfI, there is a set of the
form ]c, b] in tfI. Let Co = inf{c: ]c, b] E tfI}. Since Co is covered by tfI,
there is an interval [a, d1 [ E tfI such that co< dl . By the definition of
infimum, there is an interval ]ct> b] EtfI such that Cl < dl • Thus {[a, ~[,
]ct> b]} C tfI and [a, b] = [a, ~[U] cl , b]. It follows from this and ALEX-
ANDER'S theorem (6.40) that [a, b] is compact.
.
Now let n be arbitrary and suppose that A is closed and bounded .
Since A is bounded, there exists a cube C =.X [ai' bi] such that A c C.
1=1
The preceding paragraph and TIHONOV'S theorem (6.43) show that C
is compact. Using the fact that A is closed and citing (6.39), we see that A
is compact.
Conversely, suppose that A is compact. By (6.38), A is closed.
00
I t is clear that A cUB ~ (0) = Rn, where B ~ (0) is the open ball of radius
"=1
k centered at 0 = (0, 0, ... , 0) in Rn. Since A is compact, there exists
ko EN such that A c B",(O), i.e., A is bounded. 0
(6.45) Exercise. Prove the following.
(a) Any compact subset of a metric space is bounded.
(b) Theorem (6.44) is not true for arbitrary metric spaces.
(c) Every bounded sequence in Rn [or Kn] admits a convergent sub-
sequence.
1 A subset A of a metric space X is said to be bounded if there exist P E X and
(J E R such that e(P. x) ~ {J for all x E A.
We next take up the study of completeness for metric spaces.

(6.46) Definition. A sequence (X.,) in a metric space X is said to be
a Cauchy sequence if for each B > 0 there exists no EN such that e (xm' X.,) < B
whenever m, n ~ no. A metric space X is said to be complete if each Cauchy
sequence in X converges to a point of X.
(6.47) Example. The real line R is complete (5.25), (5.35). Also it is
easy to see that a subset of a complete metric space is complete if and only
if it is closed.
(6.48) Theorem. Any compact metric space is complete.
Proof. Let X be a compact metric space and let (X.,) be a Cauchy
sequence in X. Then (X.,) has a subsequence (X"k) converging to some
X EX. Let B > 0 be given. Choose no, ko EN such that m, n ~ no implies
e (xm' X.,) < B/2 and k ~ ko implies e (X"k' x) < B/2. Choose kl ~ ko such
that n", ~ no' Then n ~ n", implies e (x." x) ~ e (x." X.,,,) + e (X.,,,,, x)
< B/2 + B/2 = B. Thus lim X., = x, and so X is complete. 0
.. -..00
(6.49) Theorem. Any Cauchy sequence in a metric space is bounded.

Proof. Let (xn ) be a Cauchy sequence in a metric space X. Choose
no EN such that n ~ no implies e(x.., x ..,) < 1. Let (X = max{l, e(Xl' X".),
... , e(X.,._l' X ..,)}. Then e(x." x ..,) ~ (X for each n EN. 0
(6.50) Theorem. Let n E N. Then R" and K" are complete in the Eucli-
dean metric.
Proof. Let X = R" or K" and let (re k ) be a Cauchy sequence in X.
Since (re k ) is bounded (6.49), there exists a real number {3 such that
e(0, :Ilk) ~ {3 for each kEN. Then (re k ) is a Cauchy sequence in the com-
pact metric space (Bp(O))- (6.44), so (re k ) converges (6.48). 0
(6.51) Definition. Let A be a nonvoid bounded set in a metric space X.
The diameter of A is the number
diam(A) = sup{e(x,y): x,y E A}.
(6.52) Theorem [CANTOR]. Let X be a metric space. Then X is complete
if and only if whenever (An) is a decreasing sequence of nonvoid closed
subsets of X, i.e. AI::> A 2 ::> .. " such that lim diam(An) = 0, we have
00 n~oo
n A., =
.. =1
{x} for some X EX .
Proof. Suppose that (An) is a decreasing sequence of nonvoid closed
subsets of X such that diam (A.,) -+ O. For each n EN let X., E A.,. Then
m ~ n implies that e(xm' X.,) ~ diam (A.,) -+ 0 so (X.,) is a Cauchy se-
quence. Let X = lim X.,. For each m, X., E Am for all large n, and Am
.. -..00
00 00
is closed, so XEAm. Thus XE..n

=1
A.,. If X' E..n
=1
A." then e (x, x') ~ diam (An)
00
for every n. Therefore e(x, x') = O. Hence ..n

=1
A., = {x} .
5*
Conversely, suppose that X has the decreasing closed sets property.

Let (xn) be a Cauchy sequence in X. For each nEN, let An = {xm: m~ n}-.
Then (An) is a decreasing sequence of closed sets and, since (xn) is a
00
Cauchy sequence, diam (An) -)- 0. Let nC]1 An = {x}. If e> 0, then there
is an no E N such that diam (An,) < e. But x E An., so n ~ no implies
that 12 (xn' x) < e. 0
(6.53) Definition. Let X be a topological space. A set A C X is
said to be nowhere dense if A - 0 = 0. A set F C X is said to be of first
category if F is a countable union of nowhere dense sets. All other sub-
sets of X are said to be of second category.
(6.54) Baire Category Theorem. Let X be a complete metric space.
Suppose that A C X and that A is oj first category in X. Then X n A'
is dense in X. Thus X is oj second category [as a subset oj itself].
00
Proof. Let A = U An' where each An is nowhere dense in X. We

n=1
suppose that each An is closed [at worst this makes X n A' smaller].
Let V be any non void open subset of X. We will show that V n A' =1= 0.
Choose a nonvoid open set U1 C V such that diam (U1 ) < 1. For example
we may take U1 to be an open ball of radius < ~ . Then U1 is not a
subset of Av so U1 n A~ is a nonvoid open set. Let U 2 be a nonvoid open
set such that U;; CUI n A~ and diam (U2) < ~. Suppose that UV "" Un
have been chosen such that U +l is a non void open set, Ui+ 1 C U n Ai,
j j
and diam (Ui+ I) < j ~ 1 for 1 ~ j ~ n - 1. Then Un n A~ =1= 0, so

there exists a nonvoid open set Un+l such that Un+IC Un n A~ and
diam(Un+I)< n ~ l ' We thus obtain a decreasing sequence (Un) of
non void closed sets such that diam (U;;-) -)- 0. Since X is complete, there
00
n U;;- = {x}. Then x En=1

exists an x EX such that n=1 n Un+ 1 C U1 n n=1
n A~
C V n CQI An)' = V n A'. Since V was arbitrary, it follows that A'
is dense in X. 0
The Baire category theorem has many interesting and important
applications throughout analysis, as we shall see several times in the
sequel. For the moment, we content ourselves with an unimportant
though interesting application.
(6.55) Definition. Let X be a topological space and let A C X.
The set A is called a G~ set if A is a countable intersection of open sets,
and A is called an Fa set if it is a countable union of closed sets.
(6.56) Theorem. The set Q oj rational numbers is not a Gd set in R.
n u.. ,where each U.. is open in R. Then each

00
Proof. Assume that Q =

.. =1
U~ is nowhere dense since it is closed and contains no rational numbers.
00
Let Q = (X.. ):'=1 be an enumeration of Q (4.22). Then R = U (U~ U {x..}).

11=1
But U~ U {x.. } is nowhere dense for each n E Nand R is a complete
metric space. This contradicts (6.54). 0
We next examine the structure of open subsets and closed subsets of R.
(6.57) Definition. Let A be a nonvoid subset of R#. If A has no upper
[lower] bound in R, we say that the supremum [infimum] of A is 00 [ - 00]
and write supA = 00 [infA = - 00 J.
(6.58) Remark. In view of (5.33) and (6.57), every nonvoid subset of
R has both a supremum and an infimum in R#.
(6.59) Theorem. Let U be a nonvoid open subset of R. Then there
exists one and only one pairwise disjoint family f of open intervals of R
such that U = U f. The family f is countable and the members of fare
called component intervals of U. For each IE f, the endpoints of I are
not in U.
Proof. Let x E U and define a" = inf {t : ] t, x] C U} and
b" = sup{t: [x, t[ C U}. Since U is open, it is clear that a" and b" exist
in R#. We first assert that ]a", b,,[ C U and begin by proving that
]a",x]CU. If a"ER, let x .. =a,,+ ~
and if a,,=-oo, let x .. =-n.
In either case a" = inf{x.. : n EN}. By the definition of a" it follows that
for each sufficiently large n E N there exists a real number t.. such that
00 00
a" ~ t.. < x .. and ]t.. , x] cU. Then ]a", x] = U ]x.. , x] C fI.=1S
fI.=n.
U ]t.. , x] cU.
o
Likewise, we have [x, bA C U, and hence ]a", bA C U.
We next show that a" ~ U, b" ~ U. Assume that b" E U. Since U is
open, there is a 15 > 0 such that ]b" - 15, b" + !5[ C U. But then [x, b" +!5[
= [x, bA U [b", b" + !5[ C U and b" + 15 > b". This contradicts the def-
inition of b". Thus b" ~ U. Likewise a" ~ U.
Let J = {Ja", b,,[: x E U}. Since x E U implies x E ]a", b"L we have
U = UJ. We next show thatJ is a pairwise disjoint family. Let x,y E U
and suppose that there exists u E ]a", bA n ]a:v, b:v[. If a" < a:v < u,
then a:v E U and if a:v < a" < u, then a" E U. But neither a" nor a:v is
in U. Therefore a" = a:v. Likewise b" = b:v. Accordingly any two inter-
vals in J are either disjoint or identical, i.e., J is pairwise disjoint.
For each IE J there is a rational number r[ E I. Since J is pair-
wise disjoint, the mapping f: I ~ r[ of f into Q is one-to-one, and so
:7 ~ Q= Ro. Thus J is countable.
It remains only to prove that J is unique. Thus suppose that U = U"
where" is a pairwise disjoint family of open intervals. Let ]a, b[ E ".
Assume that a E U. Then there exists an interval ]c, d[ E " such that
a E Jc, de· Thus Ja, b[ =l= Jc, dL but Ja, b[ n Jc, d[ = Ja, min{b, d}[ =l= 0.
This contradiction shows that a ~ U. Likewise b ~ U. Let x E Ja, be.
Then Ja, xJ C U and [x, b[ C U so Ja, b[ C Ja x , bx [ C U. Since a ~ U
and b ~ U, we have Ja, b[ = Ja x , bA Ef. Therefore f C f. If there
exists Ja x , bA Ef nf', then x E U while x ~ Uf = U, a contradiction.
Therefore f = f. 0
(6.60) Remark. The simple structure of open sets in R has no analogue
in Euclidean spaces of dimension> 1. For example, in the plane R2
open disks play the role that open intervals play on the line as the
building blocks for open sets, i. e., the base for the topology. But it is
plain that the open square {(x,y): 0< x< 1,0< y< I} is not a union
of disjoint open disks, for if it were, the diagonal {(x, x): 0< x< I}
would be a union of [more than oneJ disjoint open intervals, contrary to
the uniqueness statement of (6.59).
Neither do the closed subsets of R have such a simple structure as
the open ones do. The next few paragraphs show this rather complicated
structure. We begin with a definition.
(6.61) Definition. Let X be a topological space and let A eX. A point
a E A is called an isolated point of A if it is not a limit point of A, i. e.,
if there exists a neighborhood U of a such that UnA = {a}. The
set A is said to be perfect if it is closed and has no isolated points, i. e.,
if A is equal to the set of its own limit points.
We will now construct a large class of nowhere dense perfect subsets
of [0, IJ.
(6.62) Definition. Remove any open interval 11 ,1 of length < 1
from the center of [0, IJ. This leaves two disjoint closed intervals
11,1 and 11,2 each having length < ~. This completes the first stage of
our construction. If the nth step of the construction has been completed,
leaving 2n disjoint closed intervals I n,l' In, 2' ••. , In, 2" [numbered
from left to rightJ, each of length < ;n ,
we perform the (n + 1)8t step
by removing any open interval I n +1,k from the center of In,k such that
the length of I n +1,k is less than the length of In,k (1 ~ k;:::;; 2n). This
leaves 2 n +l closed intervals In+1,l"'" In +1, 2n+ 1 each of length < 2,,1+1'
2"-1 2" 00
Let ~ = k=1
U In ' k and Pn = k=1 n Pn = [0, IJ
U In' k (n EN). Let P = n=1
n CQI Vn )'. Any set P constructed in the above manner is known as a

Cantor-like set. In the case that 11 ,1 = ] ~, ~ [ and the length of In+1,k
1
is exactly 3 of the length of In,k for all k, n E N, 1 ~ k ~ 2n, the
resulting set P is known as the Cantor ternary set [or simply the Cantor
setJ. In this latter case J1,l = [0, !]. J1,2 = [~, 1]. 1 2 ,1 = ]!, : [,
12,2 = ] ~, : [, J2,l = [0, ~-]. etc.
(6.63) Theorem. Let P be any Cantor-like set. Then P is compact,
nowhere dense in R, and perfect.
Proof. We use the notation of (6.62). Obviously each Pn is closed,
so that P is closed and bounded and hence compact (6.44). Since no
Pn contains an interval of length ~ ;n
and P C Pn for each n E N, it
follows that P contains no interval. Thus P- o = po = 0; that is,
P is nowhere dense in R. Next let x E P. For each n E N we have x E Pm
so that there exists kn such that x E J ..,k". Thus, given B> 0, there is an
n EN such that ;n < B, and therefore the endpoints of J ..,k" are both in
Jx - B, X + B[. But these endpoints are in P. Hence x is a limit point
of P. We conclude that P is perfect. 0
(6.64) Theorem. Let P be the Cantor ternary set. Then P = { £~ :

.. =1 3"
xn E {O, 2} for each n EN,
}
and therefore P = c.
Proof. Each number x E [0, IJ has a ternary [base threeJ expansion
in the form x = £ ;: ,where each
.. =1
Xn is 0, I, or 2. This expansion is unique
a
except for the case that x = 3m for some a, mEN where 0 < a < 3m
and 3 does not divide a. In this case x has a finite expansion of the form
x = ~ + ... + ;: where Xm = 1 [if a == 1 (mod3) J or xm = 2 [if a == 2
(mod3)J. If xm = 2 we use this finite expansion for x, but if Xm = I, we
. ~ WI'
pre f er t h e expanSIOn x = ""3 + ... + 3m-l + 3m + k.J a;;-. e eave It
Xl Xm _ l 0 2
,,=m+1
to the reader to verify these assertions [ef. (5.40)J. Thus we have assigned
a unique ternary expansion to each x E [0, 1]. One sees by induction that
Pn = {x: 0 ~x~ I, {Xl' ... , xn } C {O, 2}}. For example PI = [0, !] u
u [: ' 1] and P 2 = [0, !] u [:' !] u [~, ~] u [: ' 1] [we write
1
3"=
002] . Thus xE P = 0 Pnif and only ifxn E{0,2}foreachnEN.
I an 00
,,=2 .. -I
00
Clearly the mapping I ;: -+ (xn ) is a one-to-one correspondence be-

,,-I
tween P and {O, 2}N. Therefore P = 2 1t• = C. 0
In view of the following theorem, it is no accident that the Cantor
set has cardinal number c.
(6.65) Theorem. Let X be a complete metric space and let A be a

nonvoid perfect subset of X. Then A ~ c.
Proof. We will construct a one-to-one mapping of {O, l}N into A.
Since A is nonvoid, it has a limit point and therefore A is infinite (6.26).
Let X o =l= Xl in A. Let el = min { ~, ~ e(Xl' x 2)} and define A (0)
= {x E A : e(x o, x) ;;;; el} and A(I) = {x E A : e(xv x) ;;;; ell. Then A(O)
and A(I) are disjoint infinite closed sets each of diameter;;;; 1. Suppose
that n is a positive integer and for each n-tuple (~, ... , an) E {O, I}n
we have an infinite closed subset A (aI' ... , an) of A having diameter
;;;; ~ and such that no two of these sets have a common point. For
(av ... ,an) E {o,l}n, choose x(~, ... ,an,O)=l=x(al, ... ,an,l) in
A(~, ... , an) and let en +1 = min{2(n 1+ 1)' ~ e(x(al ,···, an, 0),
x(al , ... , an, 1))}. Define A(al , ... , an, j) = {x E A(al , ... , an):
e(x(~, ... , an, j), x) ;;;; en+1} (j = 0, I). Then {A(a l , ... , an+!) :
(aI' ... , an +1) E {O, I}n+l} is a pairwise disjoint family of closed in-
finite sets each having diameter;;;; n ~ 1 . Thus for each a = (an) E{O, I}N
we have a decreasing sequence (A(a l , .. . , an))':=l of infinite closed
subsets of A with diameters tending to 0. Hence by CANTOR'S theorem
00
(6.52), there exists a point x(a)EA such that n A(al , .. . , an) =

n=l
{x (a)}.
Suppose a =l= bin{O, I}N. Then, for some no, an, =l= bn,sox(a)EA (~, ... ,an,)
while x(b) ~ A (aI' ... , an) and therefore x(a) =l= x(b). It follows that
the mapping a ~ x (a) is o~e-to-one. Thus A ~ {O, I}N = C. 0
We next present a structure theorem for closed sets.
(6.66) Theorem [CANTOR-BENDIXSON]. Let X be a topological space
with a countable base 81 for its topology and let A be any closed subset of X.
Then X contains a perfect subset P and a countable subset C such that
A = PU C.
Proof. A point x E X will be called a condensation point of A if
UnA is uncountable for each neighborhood U of x. Let P={xEX:x is a
condensation point of A} and let C = A n P'. Since each condensation
point is a limit point, it follows that PeA. Clearly A = PUC. Since
no point of C is a condensation point of A, each x E C has a neighbor-
hood V % E 81 such that A n V % is countable. But 81 is countable so
C C U {A n V%: x E C}, and C is countable.
Next let x E P and let U be a neighborhood of x. Then UnA is
uncountable and un C is countable, so un P = (U n A) n (U n C)'
is uncountable, and hence x is a limit point of P. Thus P has no isolated
points. To show that P is closed, let x E P'. Then x has a neighborhood V
such that V n A is countable. If there is ayE V n P, then V is a neigh-
borhood of y and y is a condensation point of A, so V n A is uncountable.

It follows that V n p = 0, so that x is not a limit point of P. Therefore
P contains all of its limit points, i. e., P is closed. We conclude that
P is perfect. 0
(6.67) Remark. In view of (6.21) and (6.23), every Euclidean space
satisfies the hypothesis of (6.66).
We now make a brief study of continuity.
(6.68) Definition. Let X and Y be topological spaces and let I be a
function from X into Y. Then I is said to be continuous at a point x E X
if for each neighborhood V of I (x) there exists a neighborhood U of x
such that I(U) c V. The function I is said to be continuous on X if I
is continuous at each point of X.
(6.69) Theorem. Let X, Y, and I be as in (6.68). Then I is continuous
on X il and only il 1-1 (V) is open in X whenever V is open in Y.
Proof. Suppose that I is continuous on X and let V be open in Y.
We must show that 1-1 (V) is open in X. For x E1-1 (V), we know that I
is continuous at x, so there exists a neighborhood Ux of x such that
l(Ux) C V, i.e., Ux C 1-1 (V). It follows that l-l(V) = U{Ux : x E1-1 (V)}
which is a union of open sets, so that 1-1 (V) is open.
Conversely, suppose that 1-1 (V) is open in X whenever V is open in Y.
Let x E X and let V be a neighborhood of I (x). Then 1-1 (V) is a neigh-
borhood of x and 1(t-l(V)) C V. Thus I is continuous at x. Since x is
arbitrary, I is continuous on X. 0
(6.70) Theorem. Let X, Y, and I be as in (6.68). Suppose that [/ is
a subbase lor the topology 01 Y and that 1-1 (5) is open in X lor every 5 E [/.
Then I is continuous on X.
to
Proof. Let P-I be the family of all sets of the form B = .n
,~1
51' where
{51' ... , 5 n } is a finite subfamily of [/. Then P-I is a base for the topology
to
.n 1- 1(51), being a finite intersection
of Y (6.10), and the set 1-1 (B) = ,~l
of open sets, is open for every BE P-I. Next, let V be open in Y. Then
V = U'EI
B, for some family {B,},E! C P-I. Therefore 1-1 (V) = 1-1 (U
LEI
B,)
= U l-l(B,) which, being a union of open sets, is open in X. 0
'EI
(6.71) Theorem. Let X, Y, and I be as in (6.68). Suppose that X is a
metric space and x EX. Then I is continuous at x il and only il I (xn) --+ I (x)
whenever (xn) is a sequence in X such that Xn --+ x.
Proof. Suppose that I (xn) -> I (x) whenever Xn --+ x and assume that I
is not continuous at x. Then there is a neighborhood V of I (x) such that
I(U) C V for no neighborhood U of x. For each nEN, choose xn E B.l (x)
.
such that I(xn) ~ V. Then xn --+ x but I (xn)-f--'>- I(x). This contradiction
shows that I is continuous at x.
Conversely, suppose that f is continuous at x and let (xn) be any se-

quence in X such that Xn -+ x. Let V be any neighborhood of f (x). Then
there is a neighborhood U of x such that feU) C V. Since Xn -+ x, there
exists no EN such that n ~ no implies Xn E U. Then n ~ no implies
f(xn) E feU) C V. Thus f(xn) -+ f(x). 0
(6.72) Theorem. Let X, Y, and f be as in (6.68). Suppose that X is
compact and that f is continuous on X. Then f (X) is a compact subspace of Y.
Proof. Let "f/' be any open cover of f(X). Then {t-I(V): V E"f/'}
is an open cover of X, so there exist VI"'" Vn E "f/' such that
X = k91 f-I(V,,) = f- I C91 V,,) . It follows that f(X) C k91 V k. 0
(6.73) Coronary. Let X be a compact space and let f be a continuous
real-valued function on X. Then f is bounded [i.e., f(X) is a bounded set]
and there exist points a and b in X such that f(a) = sup{f(x) : x EX},
feb) = inf{f(x) : x EX}.
Proof. According to (6.72), f(X) is a compact subspace of R. Thus
f(X) is closed and bounded (6.44). Let IX = supf(X) and {3 = inff(X).
Since f (X) is bounded, we have IX, {3 E R. Since f (X) is closed, we have
IX, {3 E f(X). Choose a E f-I({IX}), bE f- I ({{3}). 0
(6.74) Theorem. Let A, B, and C be topological spaces. Let f be a
function from A into B and let g be a function from B into C. Let x E A and
suppose that f is continuous at x and g is continuous at f (x). Then g 0 f
is continuous at x.
Proof. Let W be any neighborhood of g 0 f(x) = g(f(x)). Then there
is a neighborhood V of f(x) such that g(V) C W. Since f is continuous
at x, there is a neighborhood U of x such that feU) C v. Thus we have
found a neighborhood U of x such thatg 0 feU) = g(f(U)) C g(V) C w. 0
(6.75) Corollary. Let A, B, C, f, and g be as in (6.74). Suppose that f
is continuous on A and g is continuous on B. Then g 0 f is continuous on A.
(6.76) Theorem. Let X and Y be topological spaces and let f be a
continuous function from X into Y. Let S C X. Then the function f [with
its domain restricted to S] is a continuous function from S [with its relative
topology] into Y.
Proof. Let xES and let V be a neighborhood of f(x). Then there is a
neighborhood U [open in X] of x such that feU) C V. But then un S
is a neighborhood of x in the relative topology on Sand feU n S)
Cf(U) C v. 0
We next discuss locally compact spaces. These spaces are of great
importance in our treatment of measure theory.
(6.77) Definition. A topological space X is said to be locally compact
if each point x EX has a neighborhood U such that U- is compact.
(6.78) Theorem. Let X be a locally compact Hausdorff space. Let

x E X and let U be a neighborhood 01 x. Then there exists a neighborhood V
01 x such that V- is compact and v-c U.
Proof. Let W be any neighborhood of x such that W- is compact. Let
G = U n w. Then G is a neighborhood of x; since c;- is a closed subset
of W-, it follows from (6.39) that G- is compact. We have G C U, but we
do not know that c;- cU. Recall (6.7.vii) thatoG = G- n GO' = G-n G'.
Thus CJG is compact (6.39). If oG = 0, we may take V = G. Thus sup-
pose CJG =1= 0. For each y E CJG, choose neighborhoods v;, and H~ of x
and y respectively such thatV~ n H~ = 0. Wemaysupposethat~C G,
for otherwise intersect it with G. Then {H:v : y EoG} is an open cover of
oG, and by compactness there exist Yl> ... , Y.. EoG such that CJG C
H:v. U'" U H:Vll = H. Let V =~. n··· n ~". Then V is a neighborhood
of x and V n H = 0. Clearly V C G, so V-c c;- and V- is compact.
Moreover V C H' and H' is closed so V- C H'. Thus V- c c;- n H'
c c;- n (oG)' = G. 0
(6.79) Theorem. Let X be a locally compact Hausdorff space and let
A be a compact subspace 01 X. Suppose that U is an open subset 01 X such
that A c U. Then there exists an open V C X such that A C V C v-c U
and V- is compact.
Proof. Apply (6.78) to each x EA. Thus for each x EA, there exists a
neighborhood Yx of x such that ~- is compact and Vx- C U. The family
{Yx : x E A} is an open cover of A, so there exist Xl' . . . , x.. E A such that
11 11
A C"'::!I Yx.t= V. Then V-="'::!IY;;C U (6.7.viii) und V-, being a finite

union of compact sets, is plainly compact. 0
The following locally compact version of URYSOHN'S lemma will
be adequate for our purposes.
(6.80) Theorem [URYSOHN]. Let X be a locally compact Hausdorff
space, let A be a compact subspace 01 X, and let U be an open set such that
A C U. Then there exists a continuous lunction I Irom X into [0, 1] such
°
that I (x) = 1 lor all x E A and I (x) = lor all x E U'.
Proof; Let Do = {O, I} and for each n E N define D" = {;.. : a EN, a is
odd, 0< a < 2"}'Let D = "goD". Thus D is the set of all dyadic rational
numbers in [0, 1]. We shall define by induction on n a chain {UthED of
subsets of X. First let U1 = A and Uo = U. For n = 1 we have D1 = {!}
and we apply (6.79) to obtain an open set UJ:. such that U1- C UJ:. c Ui: c Uo.
9 2 ~
Next let n 6 2 and suppose that open sets Ut have been defined for all
_1 _1 a
tE "=1
U D" SO that s< tin U D" implies u,-C U8 • For t =
"=0
-2" ED.., we set
a-I a+ I .
t' = ~ and t" = ~ and notIce that lIe, and lIe" are already defined
[a - 1 and a + 1 are evenJ. We again use (6.79) to obtain an open set
Ut such that lIe;;C Ut C lIe- C lIe,. Thus we obtain the desired family
{Ut}tED, and we have lIe- C Us whenever s < tin D.
Now define I on X by f(x) = 0 for x E U' and I (x) = sup{t ED: x E Ut }
for x E U. Clearly I (x) = 1 for all x E A = U1 • It remains to show that f
is continuous. To this end, let 0 ~ ~< 1 and 0 < (J ~ 1. Clearly I (x) > ~
if and only if x E Ut for some t > ~ and therefore l-l(J~, 1]) =
U {Ut : tED, t > ~}, which is open. In like manner I(x) ~ {J if and only
if x E Us for every s< (J. Therefore l-l([{J, 1]) = n {Us: sED, s< {J}
= n {lIe-: tED, t < {J}, which is closed. Taking complements we see
that 1-1 ([0, (J[) is open. These facts together with (6.70) show that I
is continuous. 0
We now take up the notions of limit superior and limit inferior for
sequences of real numbers.
(6.81) Definition. A nondecreasing [nonincreasing] sequence in R# is a
sequence (xn) C R# such that m ~ n implies Xm ~ Xn [xm ~ xnJ. A se-
quence (xn) C R# is said to have limit 00 [ - 00] if to each ~ E R there
corresponds an n" E N such that n ~ n" implies x" ~ ~ [xn ~ ~], and
we write lim Xn = 00 [lim Xn = - 00] or Xn ~ 00 [xn ~ - 00 J. A sequence
n---+oo n---+oo
that is either nondecreasing or nonincreasing is called monotone.
(6.82) Theorem. Every monotone sequence in R# has a limit in R#.
Proof. Let (xn) be nondecreasing and let x = sup{xn: n EN}. Then
lim Xn = X. 0
"->-00
(6.83) Definition. Let (xn) be any sequence in R#. We define the
limit superior 01 (xn) to be the extended real number
lim Xn = inf (sup xn)
"->-00 kEN ,,~k
and the limit inlerior of (xn) to be the extended real number

lim x" = sup( inf x n ).
n->-oo kEN n~k
Obviously the sequences (sup xn).~'_l and (inf Xn)k'=l are monotone
n~k - n~k
sequences, so that lim Xn and lim Xn are just their respective limits.
n---+oo 110---+00
The alternative notations lim sup Xn = lim Xn and lim inf Xn = lim Xn
n->-oo
are often used.
(6.84) Theorem. Let (xn) be a sequence in R# and let L = {x E R#: x
is the limit 01 some subsequence 01 (xn)}. Then lim Xn and lim Xn are in
n-::;OO 11---+00
Land lim Xn = inf L, lim Xn = sup L.

Proof. We prove only the assertions about the limit superior, the
others being obvious duals. Let x = lim x"' and for each kEN, let YII
....... 00
= sup{x,,: n ;;;; k}. Then x = inf{YII: kEN}.

Case I: x = 00. Then Yk = 00 for each kEN, so that for each mEN
there are infinitely many n E N such that x" > m. Choose nl so that
x", > 1. When nl , . . . , nm have been chosen, choose nm+l > nm such
that x"m+1 > m + 1. Then (x"m):'= 1 is a subsequence of (x,,) and
lim x" = 00. Thus x = 00 ELand clearly x = 00 = supL.
m~oo m
Case II: x E R. We have x = inf{Yk : kEN}. Thus for each p > x
there is a Yk < Pand therefore x" > p for only finitely many n [n < kJ.
This proves that there is no element of L greater than x. On the other hand,
Yk ;;;; x for all k so for each mEN there exist arbitrarily large n's such
that x,,>x-....!....
m We conclude that {nEN:x-....!...<
m x,,< x +....!...}
m is
an infinite set for each mEN. Consequently, as in Case I, we can choose
a subsequence (xnm ) of (x,,) such that lim x" = x. Therefore x E L.
m-+oo m
Case III: x = - 00. The argument given in Case II proves that there
is no element of L greater than x. But for each mEN there is a Yk such
that YII < - m. Thus x" ~ - m for all but finitely many n E N; and so
lim x" = - 00 = x. 0
........ 00
(6.85) Exercise. Let (X, e) be a metric space. Prove that:

(a) there exists a complete metric space (X, e) and a function I from
X into X such that I(X) is dense in X and e(t(x), I(y» = e(x,y) for all
x, Y E X [(X, e) is called the completion 01 (X, e)];
(b) (X, e) is unique in the sense that if (Y, a) is a complete metric
space and g is a function from X into Y such that g(X) is dense in Y
and a(g(x), g(y) = e(x,y) for all x,yE X, then there is a function h
from X onto Y such that a(h(IX), h(P) = e(IX, p) for all IX, P EX.
[Functions such as I, g, and h which preserve distance are called isometries. J
[Hints. Let ~ be the set of all Cauchy sequences in X. Define (x,,) '" (y,,)
if e (x"' y,,) ~ O. Let X be the set of equivalence classes. Define e(IX, P)
= lim e (x"' y,,) where (x,,) E IX, (y,,) E p. [cf. the completion of an ordered
....... 00
field in § 5J.J
(6.86) Exercise. Let (X, e) be a metric space. For each nonvoid
subset A of X and each x E X, define
e(x, A) = inf{e(x, a) : a E A}.
The number e(x, A) is called the distance Irom x to A. Prove each of the
following statements.
(a) If 10 =F A C X, then A- = {x EX: e(x, A) = O}.
(b) If 10 =F A C X and x, y EX, then le(x, A) - e(y, A)I ~ e(x, y).
Thus the function I defined on X by I(x) = e(x, A) is continuous on X.

(c) If A and B are two nonvoid disjoint closed subsets of X, then the
function h defined on X by
h(x) _ (1(%, B)
- (1(%, A) + (1(%, B)
is continuous on X. Also h(A) = {I} and h(B) = {o}. Notice that this
gives a simple proof of (6.80) in the case that X is a metric space.
(6.87) Exercise. Let (X, e) be a metric space and let A and B be
nonvoid subsets of X. Define the distance Irom A to B to be the number
e(A, B) = inf{e(a, b): a E A, bE B}.
Prove the following assertions.
(a) If A is compact, then there exists a point a E A such that e(a, B)
=e(A,B).
(b) If A and B are both compact, then there exist points a EA and
bE B such that e(a, b) = e(A, B).
(c) If A is compact and B is closed, then e(A, B) = °
if and only
ifAnB=!=0.
(d) If X is a noncompact metric space with no isolated points, then
X contains nonvoid, closed, disjoint sets A and B such that e(A, B) = 0.
(6.88) Exercise. Let X be a nonvoid complete metric space. Suppose
that I is a function from X into X such that for some constant c E JO, 1 [
we have
e(t(x), I(y) ~ ce(x,y)
for all x, y EX. Prove that there exists a unique point u E X such that
I(u) = u. [Let xE X and consider the sequence x, I(x), 1(t(x), ... J.
This result is known as BANACH'S fixed-point theorem. It implies several
existence theorems in the theory of differential and integral equations.
(6.89) Exercise. Prove that the closed interval [O,IJ cannot be ex-
pressed as the union of a pairwise disjoint family of closed [nondegenerate J
intervals each of length less than 1.
(6.90) Exercise. Suppose that: X is a topological space; Y is a metric
space; and I is a function from X into Y. For each x E X, define
w (x) = inf{diam(t(U): U is a neighborhood of x}.
The function w is called the oscillation lunction lor I. Prove the following
statements.
(a) The function I is continuous at x if and only if w (x) = 0.
(b) For each real number IX, the set {x EX: w (x) < IX} is open in X.
(c) The set {x EX: I is continuous at x} is a Gd set.
(d) There is no real-valued function I defined on R such that
{x E R : I is continuous at x} = Q.
(e) There exists a real-valued function I on R such that {x ER :
I is discontinuous at x} = Q.
(6.91) Exercise. Prove that every locally compact Hausdorff space

is of second category [as a subset of itself]. [Mimic the proof of the Baire
category theorem (6.54) by constructing an appropriate decreasing se-
quence of compact sets.]
(6.92) Exercise. Let X be a topological space and let Y be a metric
space. Suppose that I and US::= 1 are functions from X into Y such
that each In is continuous and lim In (x) = I (x) for each x E X. Let
n->oo
00 OO[OO{
A = kOI m':2 1 n£1m xE X: e(tm(x), In (x}) ;;;; k
l}]O
Prove that:
(a) I is continuous at each point of A;
(b) X n A' is of first category in X; I
(c) if X is of second category [in itself], then {x EX: I is continuous
at x} is dense in X;
(d) l-l(V} is an Fa set for each open set VcY [prove that I-I (V)
= kgl mgl nOm {x EX: e(tn(x), V') ~ !}] ;
(e) the function ~Q is the pointwise limit of no sequence of contin-
uous real-valued functions on R.2
(f) Prove that ~Q(x) = lim [lim {cos (m!nx)}2n] for all x ER.
m~oo n-+oo
(g) Prove that sgn (x) = lim ~ arctan (nx) for all x E R.
n-+oo n
(h) Prove that 1 - ~Q(x) = lim sgn{sin2(m!nx)} for all x E R.
m->oo
(6.93) Exercise. Let l'oo(N) denote the set of all bounded sequences
x = (xn) of real numbers. For x, y E l'oo(N) , define d (x, y) = sup {Ixn - Ynl :
n EN}. Prove the following.
(a) The function d is a metric for l'oo(N}.
(b) The metric space l'oo (N) is not separable.
(c) If (X, e) is any separable metric space, then there exists an isom-
etry I from X into l'oo(N), i.e., d(t(x), I(y) = e(x,y} for all x,yE X.
[Let (Pn)':=l be dense in X and define I(x) = (e(x, Pn) - e(Pn' Pl)':=l]'
(6.94) Exercise. Prove that if X is a compact metric space and I is an
isometry from X into X, then I is onto X.
(6.95) Exercise. Let X be a locally compact Hausdorff space and let
D be a dense subset of X such that D is locally compact in its relative
topology. Prove that D is open in X.
(6.96) Exercise. Let X be a linearly ordered set. The order topology
lor X is the topology on X obtained by taking as a subbase the family
00
1 Show first that if (Em);;'=l is any sequence of subsets of X, then ~ E-;;
00
C ( n Em U
) (00U ~n
'\ .
E;"; m-l
.. =1 m-l m
a Recall that ;Q is the characteristic function of Q (2.20).
of all sets of the form {x EX: c < x} and {x EX: x < d} for c, dE X.
Prove that X, with its order topology, is compact if and only if every
nonvoid subset of X has both a supremum and an infimum in X. [Use
(6.40) as in the proof of (6.44).]
(6.97) Exercise. (a) Use (6.96) to prove that a well-ordered set is
compact [in its order topology] if and only if it contains a greatest
element.
(b) Use (6.96) to show that R* with the usual topology (6.5.b) is
compact.
(6.98) Exercise. Prove that the set P D of all countable ordinal
numbers [see (4.49)] with its order topology (6.96) is sequentially compact
but not compact.
(6.99) Exercise. Let P be CANTOR'S ternary set and let X = {O, l}N
have the product topology, where {O, I} has the discrete topology. For
re = (xv x 2, ••• ) E X, define 11' (re) = 1;
,,=1
23:" . According to (6.64), 11' is
a one-to-one mapping from X onto P. Prove that both 11' and 11'-1
are continuous.
(6.100) Exercise. Prove that if Y is a compact metric space and P is
CANTOR'S ternary set, then there exists a continuous function I from P
onto Y. [Let {v,,}:=1 be a countable base for the topology of Y. For each
nE N, set A",o= v,;- and An,l = Y n V~. For a point x =
00
1;
,,=1
2;" in P,
the set "~1 An,x" is either void or contains just one point. Let
B={XEP: "OI A ".x,,=I= 0} andforeachxEB,letg(x)E ,,61 An, x,,' Prove

that g is continuous from B onto Y. Show that B is closed in P and that
there exists a continuous function h from Ponto B. Finally set I = go h.]
(6.101) Exercise [BANACH]' Let I be a continuous real-valued func-
tion on [a, b] cR.
(a) For each positive integern, letFn = {x: x E [a, b], and/(x') = I (x)
for some x' 6 x + !}.
Prove that F n is a closed set.
00
(b) Let E = [a, b] n "~1 F~. Prove that I is one-to-one on E and that
I(E)=/([a, b]). In fact, each xEE is equal to sup{Y:YE [a,b], I(y)
= I(x)}. Note that E is a Ga set.
(6.102) Exercise. Let I be a real-valued function with domain R
having a relative minimum at each point of R, i. e., for each a E R, there
is a number ~(a) > 0 such that I(t) 6 I(a) if It - al < ~(a).
(a) Prove that I (R) is a countable set.
(b) Find a function as above that is unbounded and also monotone
on no interval containing O.
§ 7. Spaces of continuous functions 81
(6.103) Exercise. Consider a function I with domain R and range

contained in R such that I 0 I = I. Describe I completely. If I is continuous,
what more can you say? [Recall that Rand I(R) are connected (6.9) and
so I(R) is an interval.] If I is differentiable, what more can you say?
(a) A continuous image of a connected space is connected.
(b) A Cartesian product X X, is connected if and only if every X,
lEI
is connected.
§ 7. Spaces of continuous functions

Functions - both real- and complex-valued - are a major object
of study in this text. Given a set X and a set ff of functions defined on X,
we are frequently interested not only in individual functions I in ff,
but also in ff as an entity, or space, in its own right. Often ff admits a
natural topology [or several natural topologies] of interest by themselves
and also for proving facts about ff. Often too ff is a vector space over K
or R, and vector space notions can be most helpful in studying analytic
questions regarding ff. In the present section we take up a simple class
of function spaces - spaces of continuous functions - and a simple
topology for these spaces. Many other function spaces will be studied
in the sequel.
We begin with a few definitions and some notation.
(7.1) Definition. Let X be any nonvoid set [no topology as yet], and
consider the set KX of all complex-valued functions defined on X.
For I, g E KX, let I + g be the function in KX defined by
(i) (f + g) (x) = I (x) + g (x) for all x EX;
let I g be defined by
(ii) (Ig) (x) = I (x) g(x) for all x EX;
for IE KX and ac EK, let acl be defined by
(iii) (ac/) (x) = ac(f(x) for all x EX.
For IE KX, let III be the function such that
(iv) III (x) = II (x) I for all x EX
and I the function such that
(v) I(x) = I (x) for all x EX.
That is, sums, products, scalar multiples, absolute values, and complex
conjugates of functions on X are defined pointwise. The set RX of all
real-valued functions on X can be considered in an obvious way as a
subset of KX, and so definitions (i), (ii), (iii) [for real ac], (iv) and (v)
[f = I if and only if IE RX] hold for RX as well as KX. In addition, RX
admits a natural partial order. For I, gE RX, we write I ~ g [or g ;?; fJ if
(vi) I(x) ~ g(x) for all x EX.
We define max{/, g} and min{/, g} by
(vii) max{f, g}(x) = max{f(x), g(x)} for all x EX

and
(viii) min {f, g} (x) = min {f (x), g (x)} for all x E X.
For some purposes, we also need extended real-valued functions on X.
For IP C (R#)X, we define sup{f: f E IP} by
(ix) sup{f: f E 1P}(x) = sup{f(x) : f E IP},
which can be any element of R#, and inf{f: fE IP} by
(x) inf{f: f E IP} (x) = inf{f(x) : f E IP} .
Thus all of our operations on and relations between functions are de-
fined pointwise.
Finally, for a subset ff of KX, we define W by
(xi) ffr = {f E ff : f (x) E R for all x E X} = ff n RX
and ff+ by
(xii) ff+ = {f E ffr : f (x) ~ 0 for all x E X} .
The set ff+ also can be defined for ff C (R#)x.
(7.2) Remarks. (a) For IX E K, the function "p in KX such that
"p (x) = IX for all x E X is called the constant function with value IX or the
function identically IX. This function is a quite different entity from the
number IX. It would be unwieldy to use a distinct symbol [e.g. C""x] for
this function whenever we need to write it. We will therefore write the
function identically IX simply as IX, trusting to the reader's good sense to
avoid confusion.
(b) It is easy to check that KX is a vector space over K and that
RX is a vector space over R. Also these spaces are commutative rings,
with [multiplicative] unit the constant function 1. It is further obvious
that
(i) lX(fg) = (lXf)g = f(lXg)
for all functions f, g and scalars IX. That is, KX and RX are algebras over K
and R, respectively. [A vector space over a field F that is also a ring in
which (i) holds is called an algebra over F.]
(c) It is also clear that the relation ~ in RX satisfies (2.7.i)-(2.7.iii),
i. e., ~ is a genuine partial ordering. If X> 1, then ~ is not a linear
order. It is also easy to see that (RX, ~) is a lattice: for f, g E RX, there is
a unique hE RX such that h ~ f, h ~ g, and h ~ h' if h' ~ f and h' ~ g;
that is, h is the smallest majorant of f and g. Similarly there is a largest
minorant k of f and g. It is obvious that h = max{t, g} and that
k = min{f, g}.
(d) The partially ordered set RX enjoys a much stronger property
than (c). Let ff be any nonvoid subset of RX bounded above by a func-
tion cp E RX, i.e., f ~ cp for all f E ff. Then ff admits a smallest majorant.
Its value at x E X is of course sup{f(x) : f E ff}. Similar statements hold
for sets ff C RX that admit minorants.
For infinite sets X, the algebras KX and RX are too large to be of

much use in analysis, although their algebraic structure is of great in-
terest to specialists. By a first restriction we obtain a metrizable space.
(7.3) Definition. Let X be a nonvoid set. Let SS (X) denote the set
of all functions IE KX such that
(i) sup{11 (x) I : x EX}
is finite. Such functions are said to be bounded. The number (i), written
as 11/11u> is called the unilorm norm 01 f.
(7.4) Theorem. Let X be a nonvoid set, and consider I, g ESS (X)
and oc EK. Then the lollowing relations hold:
(i) 11011" = 0, 11/11u> 0 il lof 0;
(ii) Iloc/ll" = loci 11/11,,;
(iii) III + gil" ~ 11/11" + Ilgll,,;
(iv) Il/gll" ~ 11/11" Ilgll" .
Similar assertions hold lor I, g ESST (X) and oc ER.
Proof. Simple exercise.
The linear space SS (X) with its norm I II" is an important example of
a class of analytico-algebraic objects which we shall encounter repeatedly.
(7.5) Definition. Let E be a linear space over K [or RJ. Suppose that
there is a function x -;.llxll with domain E and range contained in R
such that:
(i) 11011 = 0 and Ilxll > 0 if x of 0;
(ii) Ilocxll = loclllxli for all x E E and oc E K [or R];
(iii) Ilx + yll ~ Ilxll + Ilyll for all x, y E E.
The pair (E, I III is called a complex [or real] normed linear space, and
I I is called a norm. 1
If E is a normed linear space and also an algebra over K [or R], and if
(iv) IlxY11 ~ Ilxll Ilyll for all x, y E E,
then E is called a complex [or real] normed algebra. If a normed algebra
has a multiplicative unit u, then we will postulate that
(v) Ilull = 1. 2
(7.6) Theorem. Let E be a complex or real normed linear space. Let e
be the lunction on E x E defined by
(i) e (x, y) = Ilx - yll .
Then e is a metric on E.
Proof. Trivial.
1 As usual, where confusion seems unlikely we will call E itself a normed linear
space.
2 Since Ilxll = Iluxll ~ Ilullllxll, we have Ilull ~ 1 without (v). Also, a normed
algebra with unit can be renormed so that the unit has norm 1 and nothing essential
is changed. See Exercise (7.42) infra.
6*
(7.7) Definition. A complex [realJ normed linear space that is com-

plete in the metric Ilx - yll is called a complex [realJ Banach space.
A complex [realJ Banach space that is also a normed algebra is called
a Banach algebra.
Banach spaces are very important in contemporary analysis; many
basic theorems can be couched in abstract terms as assertions about
Banach spaces of one kind or another. We will give many examples
throughout the text of this technique [see in particular § 14]. We turn
next to the principal object of study in the present section, and one of
the important objects of study in the entire text.
(7.8) Definition. Let X be a nonvoid topological space. Let G:(X)
denote the set of all functions in !OS (X) that are continuous complex-
valued functions on X.
(7.9) Theorem. With the algebraic operations (7.1.i)-(7.l.iii) and the
norm 1111" 01 (7.3), G:(X) is a commutative complex Banach algebra with
unit.
Proof. The only non obvious point is the completeness of G: (X) in
the uniform metric. Let (fnr:~l be a sequence of functions in G:(X)
such that
lim Illn - Imll" = 0 .
m,n~oo
(1)
That is,
lim [sup {lin (x) -
1n,n---70OO
1m (x) I : x EX}] = O. (2)
For every fixed x EX, (2) implies that

lim
m, n--+oo
lin (x) - 1m (x)1 = 0,
and so (In(X))':~l is a Cauchy sequence in K. Since K is complete [use

(6.50) with n = 1J, the sequence (In (x)) has a limit in K, which we denote
by I(x). The mapping x --+ I(x) is thus an element of KX. We claim that
lEG: (X) and that
lim III - Inll" = 0 .
n->oo
(3)
Actually, it is easiest to prove (3) first. Let 8 be an arbitrary positive

real number and let the integer p [depending only on 8J be so large that
Illn - Imll" < : (4)
for all m, n ~ p. Now consider a fixed but arbitrary x E X, and choose m

[depending on both x and 8J so large that m ~ p and also
8
11m (x) - l(x)1 < 3 . (5)
Combining (4) and (5), we see that
Iln(x) - I(X)I ~ I/m(x) - In (x) I + 11m (X) - I(x)l

<~+~
3 3
2
="3 8 (6)
for all n ~ p. The integer p is independent of x, and as x is arbitrary in (6),
we may take the supremum in (6) to write
IIln - III .. = sup {lin (x) - I (x) I : x E X}
2
~"38
<8
if n ~ p. That is, (3) holds.
It remains to prove that/E (t(X). Choose n so that III - Inll .. < 1. Then
it is evident that II (x) I < lin (x) I + 1 for all x E X, and so 11/11 .. exists and
does not exceed IIlnll .. + 1. To prove that I is continuous, let x be any
point in X, let 8 be a positive number, and let n be so large that I In - III ..
< ; . Let U be a neighborhood of x such that
e
Iln(Y)-ln(x)I<"3 forall yEU.
For y E U, we thus see that
I/(y) - l(x)1 ~ I/(y) - In(y)1 + Iln(Y) - In (x) I + lin (x) - l(x)1
~ III - Inll .. + Iln(Y) - In (x) I + IIln - III ..
= 8.
This is just the defining property (6.68) of continuity of I at x. 0

(7.10) Remarks. Some topological spaces X admit no nonconstant
continuous real- or complex-valued functions. A simple but trivial example
is (X, lTJ) where X is infinite and lTJ consists of 0 plus all subsets of X
with finite complements. Such spaces are of no interest for the present
text. However, if X > 1 and X is a locally compact Hausdorff space, then
(t (X) contains an abundance of nonconstant functions, as Theorem
(6.80) shows. At the present time, locally compact Hausdorff spaces
seem to be an ideal vehicle for integration theory, as well as for a number
of classical theorems of analysis. Much of the present section will ac-
cordingly be devoted to these spaces.
(7.11) Exercise. Many, although not all, noncompact topological
spaces admit unbounded continuous complex-valued functions. We
will make no detailed study of the space of all continuous functions on
a topological space, but in this exercise the reader is invited to consider
some of the possibilities, and to prove the following assertions.
(a) Every noncompact metric space admits an unbounded con-

tinuous real-valued function.
(b) Let P Q be the well-ordered set defined in (4.49). Let fJI be the
family of all sets ofthe form {o} or {y EP Q : oc < y ;;;; {J} where oc < {J < Q.
Then fJI is a base for the order topology of P Q [d. (6.96) and (6.98) J.
The topological space P Q is a locally compact Hausdorff space; in fact
every subspace {y E P Q : oc ;;;; y ;;;; {J} is compact. Also P Q is non-
compact, although it is both sequentially compact and Fn§chet compact.
[Again see (6.96) and (6.98).J
(c) Every continuous complex-valued function I on P Q is ultimately
constant: there are an ordinal oc E P Q and a complex number t such
that I(y) = t for all y ~ oc. Consequently every continuous complex-
valued function on P Q is bounded.
(d) Let X be a topological space, and let era (X) denote the set of all
continuous complex-valued functions on X, bounded or unbounded.
If era (X) contains unbounded functions, we cannot impose the uniform
norm 1111 .. on <ra(X). However, an analogue of (7.9) does hold. If (fnr:=l
is a sequence of functions in <ra(X) for which all differences In - 1m
are bounded, and if
lim Illn - Imll .. = 0,
m,n~oo
then there is an IE era (X) such that all I - In are bounded and
lim III - Inll .. =
n ...... oo
°.
We return to a study of function spaces on locally compact Haus-
dorff spaces.
(7.12) Definition. Let X be a nonvoid locally compact Hausdorff
°
space. Let <roo (X) be the subset of <r (X) consisting of all IE <r (X) such
that for some compact subset F of X [depending on fJ, I (x) = for all
x EF' n X. Let <ro (X) be the subset of <r (X) consisting of all IE <r (X)
such that for every positive number e, there is a compact subset F of X
[depending on I and eJ such that II (x) I < e for all x EF' n X. Functions
in <roo (X) are said to vanish in a neighborhood 01 infinity and functions in
<ro (X) to vanish at infinity: both phrases are loose but expressive.
(a) The inclusions <roo (X) C <ro(X) C <reX) obtain. If X is non-
compact, then <roo(X) ~ <reX) (6.80). For the space P Q of (7.11.b),
we have <roo (X) = <ro(X). If X is compact, then <roo (X) = <ro(X) = <r(X).
(b) Let X be an arbitrary locally compact Hausdorff space, and
consider <r (X) and SB (X) as metric spaces under the uniform metric
III - gil ... The space <ro(X) is closed in SB(X) [and hence also in <r(X)J:
if InE<ro(X), IESB(X), and limllin-/II .. =O, then IE <ro(X). Also
n ...... oo
<ro(X) is complete as a metric space. [Recall (6.47).J The space <roo (X)
is not in general closed: its closure is <ro(X). [A proof of the last state-
ment is indicated in (7.41) below.]
Several concepts closely related to, but not identical with, continuity
are needed in our treatment of integration theory. We now take them up.
(7.14) Definition. Let X and Y be metric spaces with metrics e and a
respectively. A mapping q; with domain X and range contained in Y
is said to be uniformly continuous if for every e > 0 there is a c5 > 0
such that e(x, x') < c5 implies that a(q;(x), q;(x')) < e.
(7.15) Remarks. (a) It is easy to see from (6.71) that a uniformly
continuous mapping is indeed continuous.
(b) There is a notion of uniform continuity more general than that
in (7.14), based on what are called uniform spaces. We do not need this
concept and hence omit it.
(7.16) Exercise. (a) Consider the function exp (5.56) defined on K.
Prove that exp is continuous but not uniformly continuous.
(b) Let X be a nonvoid set and let e be the discrete metric on X
(6.13.c). Prove that every mapping of (X, e) into a metric space is
uniformly continuous.
(c) Find reasonable necessary and sufficient conditions for a metric
space to admit a continuous real-valued function that is not uniformly
continuous.
(d) Let (fn)':=l be a sequence of complex-valued uniformly con-
tinuous functions on a metric space X such that all differences fn - fm
are bounded and
lim Ilfn - fmll .. = 0 .
m,n~().,)
Prove that the limit function f (7.11.d) is uniformly continuous.

(7.17) Theorem. Let X and Y be as in (7.14) and let q; be a continuous
mapping of X into Y with the following property. For every e> 0, there
is a compact subset A. of X such that a(q;(x), q;(x')) < e for all x, x'
in A~ n X. Then q; is uniformly continuous.
Proof. Let e be an arbitrary positive number, and let A. be as in the
statement of the theorem: if x, x' are in A~ n X, then
a(q;(x), q;(x') < e. (1)
Now look at an arbitrary point y E A B • Since q; is continuous, there is
a positive number 'fly [depending on y] such that
z E B 2 '1Y(Y) implies a(q;(z), q;(y)) < ~ . (2)
[Notation is as in (6.14).] Now consider the family of sets {B'1Y(Y):
yEA.}. This is an open covering of A., and so by (6.32) and (6.18)
there is a finite subfamily {B'1Y, (Yl)' ... , B'1Ym (Ym)} that covers A •. Let
6 be the number min {'fly" 'fly., .•• , 'fIYm}' We claim that this c5 will
satisfy (7.14) for the preassigned 8. If x and x' are in A; n X, than (1)
may be applied. If at least one of x and x' is in A., we may suppose that
x E A •. Then x is in some B'lYk(y,,). If e(x, x') < 15, then we have
e (y", x') ~ e (y", x) + e (x, x')
< 'YJYk + 15
~ 2'YJYk'
That is, x' is in B2'lYk(y,,) and so (2) shows that a(<p(x'), <p(y,,)) < ~.
Since x is also in B2'lYk (y,,), (2) implies that a(<p (x), <p (y,,)) < ~ ,and hence
a(<p(x), <p(x')) ~ a(<p(x), <p(y,,)) + a(<p(y,,), <p(x'))

<~+~
2 2
=8.
Thus <p is uniformly continuous. 0

(7.18) Corollary. If X is a locally compact metric space, then all
functions in <ro (X) are uniformly continuous. If X is a compact metric
space, then all functions in <reX) are uniformly continuous.
Proof. See (7.17). 0
(7.19) Our next notion is restricted to real-valued functions f on a
topological space X. The definition of continuity for f at Xo EX may be
[somewhat artificially] broken up into two parts: for every 8> 0, there
is a neighborhood U of Xo such that
(i) f(x) > f(x o) - 8 for all x E U,
and
(ii) f(x) < f(x o) + 8 for all x E U.
Taken separately, (i) and (ii) define useful classes of functions. It is
advisable also to consider extended real-valued functions.
(7.20) Definition. Let X be a topological space and f an extended
real-valued function defined on X that does not assume the value - 00,
that is, f (x) is real or 00 for all x E X. The function f is said to be lower
semicontinuous at Xo E X if the following conditions hold. If f(x o) < 00,
then for every 8> 0, there is a neighborhood U of Xo such that f(x)
> f(x o) - 8 for all x E u. If f(x o) = 00, then for every positive number IX
there is a neighborhood U of Xo such that f (x) > IX for all x E U. The
function f is called lower semicontinuous if it is lower semicontinuous at
every point of X. The set of all lower semicontinuous functions on X
is denoted by the symbol:m (X). A function on X with values in [- 00, oo[
is called upper semicontinuous if analogous conditions f (x) < f (x o)
+ 8 [f (xo) finite] or f (x) < - IX [f (xo) = - 00] hold near xo' The set of all
upper semicontinuous functions on X is denoted by m(X).

(a) Nonconstant semicontinuous functions exist on every topological
space X having an open set U such that 0 ~ US;: X. The characteristic
function ~u is lower semicontinuous and ~u' is upper semicontinuous.
(b) A function I is lower semicontinuous if and only if - I is upper
semicontinuous. [Recall that - (00) = -00 and that - (-00) = 00 (6.1).J
Thus for any fact about lower semicontinuous functions there is a dual
fact about upper semicontinuous functions. [We will ordinarily state ex-
plicitly only assertions about lower semicontinuous functions.J
(c) The definition of lower semicontinuity at Xo can be recast: for
every real number IX< I(x o), there is a neighborhood U of Xo such that
I(x) > IX for all x E U. This deals with the cases I (xo) < 00 and I (xo) = 00
simultaneously.
(d) A function I from X into J- 00, 00] is lower semicontinuous if and
only if 1-1 (]t, 00]) is open in X for every t ER. [This characterization is
frequently useful.]
(7.22) Theorem. Let X be a topological space.
(i) II IE ml (X) and IX is a nonnegative real number, then IXI E ml (X).
(ii) II I, g E ml(X), then min {I, g} E ml(X).
(iii) II ~ is a nonvoid subset 01 ml(X), then sup{/: I E ~} is a lunction
in ml(X).
(iv) II I, g E ml (X), then I + g E ml (X).
(v) Suppose that X is a locally compact Hausdorff space. Then il
IE ml+(X), we have I = sup{qJ: qJ E <rot (X), qJ ~ I}.
Proof. Assertions (i), (ii), and (iv) are all but obvious, and we leave
their proofto the reader. To prove (iii), write g for the function sup {I :IE~}.
Consider a fixed but arbitrary point Xo EX. For every real number
IX < g (x o), the definition of supremum (5.32) shows that there is an I E ~
such that IX IX for all x E U
[this holds for I (xo) < 00 as well as for I (xo) = 00]. For all x E U, we clearly
have
g(x) ~ I(x) > IX.
Thus g satisfies (7.21.c) and so (iii) is proved.
To prove (v), we use URYSOHN'S theorem (6.80). If 1=0, there is
nothing to prove. If I (xo) > 0 for some Xo EX, consider any real number
IX such that 0< IX< I(x o). There is a neighborhood U of Xo such that
I (x) > IX for all x E U. By (6.80) there is a function qJ E <rot (X) such that
qJ (X) C [0, IX], qJ (xo) = IX, and qJ (U') C {O}. Plainly we have qJ ~ I,
and as IX can be arbitrarily close to I (x o) [f (xo) < 00] or arbitrarily large
[f(x o) = 00], we infer that f(x o) = sup{~(xo): ~ E (S:do(X), ~ ~ f}. Since

Xo is arbitrary, (v) is established. 0
(7.23) Exercise. (a) State and prove the analogue of (7.22) for upper
semicontinuous functions.
(b) Let X be a topological space admitting a nonclosed open set.
Prove that ~nr(x), the real-valued lower semicontinuous functions on X,
do not form a linear space. If every open subset of X is closed, prove that
I~nr (X) is a linear space.
(c) Prove that uniform limits of sequences in ~nr(X) are in mr(x).
(7.24) We now take up a famous and vitally important approxima-
tion theorem. The German mathematician KARL WEIERSTRASS [1815-
1897] published in 1885 a proof that polynomials with real coefficients are
dense in the space (S:r([o, 1]) in the topology induced by the uniform
metric. We will prove a far-reaching generalization of this theorem due to
the contemporary U.S. mathematician M. H. STONE. Every proof of the
STONE-WEIERSTRASS theorem requires some "hard" analysis. Our proof
uses only a little such analysis, which is presented in the next theorem.
We prove somewhat more than we need.
(7.25) Theorem. For any real number (~) = 1 and
IX, let
1X(1X-1)(1X-2)··· (IX-n + 1)
( IX) =
n n!
for n = 1, 2, .... The infinite series
(i)
converges for all x E ]- 1, 1[. For IX > 0, the series
(ii)
converges, and the series (i) converges uniformly and absolutely in [-1, 1].
Finally, we have
(iii)
for x E J-1, 1[ for all real IX. For IX> 0, (iii) holds for all x E [-1, 1J.
Proof. To prove (i), we use the ratio test. If IX is a nonnegative integer,
then all but finitely many of the numbers (:) are 0, and so (i) trivially
converges. Otherwise, for Ixl < 1 and n = 0, 1, 2, ... , we have
IX )xn+l
( n+l
Hence the series (i) converges absolutely for Ixl < 1. Next we prove (ii).
The case in which a is a nonnegative integer is again trivial. For a
not a nonnegative integer, let an = 1(:) I· Then we have
an+! _ la-nl _ n-a
---a:: - n +1 - -n-+T '
the last equality holding for n ~ [aJ + 1 [[aJ is the largest integer not
exceeding a]. Hence for n ~ [aJ + 1, we have (n + 1) an +1 = nan - aan,
and so
(1)
Thus for n ~ [aJ + 1, (nan) is a decreasing sequence, and so has a
00
limit; let lim nan = y ~ O. Now consider the series I.: (nan -(n+ 1) an +1).
n-+oo n=O
The pth partial sum of this series is -(p + l)ap+I' which converges to
00
- y. Hence the series I.: (nan - (n + 1) an+1) converges, and since by (1)
n~O
00
for sufficiently large n, the series l: an converges; i. e., the series (ii)
n~O
converges. Since 1(:) xnl ~ 1(:)1 for Ixl ~ 1, the series (i) converges
absolutely and uniformly in [-1, IJ and so defines a continuous func-
tion on [- 1, IJ.
We now prove (iii). For x E J-l, 1 [ and a E R, let fa.(x) = f (:) xn.
n~O
A power series may be differentiated term by term in its open interval

of convergence, and so we have
f~(x) =
n~l
i; (n+ 1) (n +a 1)xn.
fn(a)xn-l=
n n~O
Using the identity (n + 1) (n: 1) a (a; 1), =
we see that
t~ (x) = an~
00 (a -1)
n xn = at a.-I (x) . (2)
Next we have
(l+x)ta._I(x)=(I+x)I.:
00 (a-I)
n xn
n~O
= 1 + n~ [(a 00 -1) + (a -I)] Xn = nfo (a) Xn = ta.(x) ,

n n-I
00
n
i.e.,
(1 + x) Icc-l (x) = Icc (X) • (3)
(1 + x)f~(x) - IXfcc(x) = 0
for all x E ]- 1, 1 [. Also we have
ddx [fcc (x) (1 + x)-CC] = (1 + x)-CC-l((1 + x) I~(x) - IXlcc(x)) = O.
Hence J~x;)(J. is a constant function in ]- 1, 1 [. Setting x = 0, we deter-

mine that the constant value of this function is 1; i.e., fcc(x) = (1 + X)IX
for Ixl < 1. If IX > 0, then the series also converges for x = ± 1, and hence
the identity (iii) also holds for these values of x. [Two continuous complex-
valued functions that are equal on a dense subset of their common
domain are equal everywhere.] 0
(7.26) Remarks. (a) In proving the important Theorem (7.27), we
need only the special case of (7.25.iii) in which IX = ~ . That is, we need
the identity
(i)
and the fact that the series converges absolutely and uniformly for
Ixl ~ 1.
(b) The STONE-WEIERSTRASS theorem depends ultimately upon order
properties of the space er' (X) for a compact Hausdorff space X. It is
evident that max{f, g} and min{f, g} are continuous if I, g are in er'(X),
and this simple fact will be very useful. The following theorem connects
polynomials with maxima and minima.
(7.27) Theorem. Let X be any nonvoid set [no topology] and let 1jl
and cp be functions in ~' (X). For every e > 0, there is a polynomial
P = L:m L:" lXi k 1jli cpk with real coefficients such that
i=O "=0
(i) Ilmax{1jl, cp} - Pllu < e.
A like assertion holds lor min{1jl, cp}.
Proof. For every real number t, the identity
1
It I = (1 + t2 - 1)2
1
is obvious. For It2 - 11 ~ 1, that is, for It I ~ 22, we have
(1 + t2 - l)i = i; (+)
,,=0n
(t2-1)",
and the series converges absolutely and uniformly for all t such that
1
It I ~ 22 (7.26). For every positive integer p and for all real t such that
1
It I ~ 22 , we thus have
(I)
and the right side of (I) is arbitrarily small for p sufficiently large. Now
consider any nonzero function 1.p in Q3' (X), and for brevity write the
number 111.p11 .. as fJ.
For x EX and PEN, we apply (I) to write
il1.p(X)I- ntfJ(:)[~~~) -Iri

=fJ\fJ- l l1.p(X)I- i(:)[V'~~) -Irl
s- £ \(+)\.
fJ n=p+l n
It follows that for 1.p E Q3' (X) and e > 0, there is a real polynomial Q
in the functions 1 and 1.p2 [i. e., a linear combination with real coefficients
of 1, 1.p2, 1.p4, ... J, the coefficients of which depend solely upon and e, fJ
for which
1I11.p1- QII .. < 2e. (2)
'*'
If rp = 1.p, the relation (i) is trivial. If rp 1.p, then 1.p - rp is not the
zero function, and we may apply (2) with 1.p replaced by 1.p - rp:
11I1.p- rpl- Q(1.p- rp)II .. < 2e.
As noted in (5.36), the identity
1
max{1.p, rp} = 2 (11.p- rpl + (1.p + rp))
obtains, and so we have
Ilmax{1.p, rp} - ~ (Q(1.p- rp) + 1.p + rp)II ..< e.
Setting P(1.p, rp) = ~ (Q(1.p- rp) + 1.p + rp), we obtain (i). To prove (i)
with "max" replaced by "min", note that min{1.p, rp} = -max{- 1.p, -rp}. 0
(7.28) Definitions. Let X be a set and e a family of functions on X
with values in a set Y. Suppose that for all x, y E X such that x y '*'
'*'
there is an 1E e such that 1(x) I(Y). Then we say that e is a separating
lamily 01 lunctions on X. Next suppose that e is a family of real-valued
functions on X. A real polynomial in functions Irom e is any finite
sum of functions a./~'/;··· 'I?I, where the coefficient a. is a real number
and the exponents nj are positive integers. Equivalently, a real poly-
nomial in functions from e is an element of the smallest subalgebra of

RX that contains e. Complex polynomials are defined similarly in KX.
We now prove one version of the STONE-WEIERSTRASS theorem.
(7.29) Theorem. Let X be a nonvoid compact Hausdorff space, and let
e be a subset 01 <r' (X) such that:
(i) e is a separating lamily;
e
(ii) contains the lunction I;
(iii) lEe and oc E R imply ocl E e;
(iv) I, gEe implies I + gEe;
(v) I, gEe implies max {I, g} E e.
Then e is dense in <r' (X) in the topology induced by the unilorm metric. 1
Proof. Consider 10 E <r'(X). If 10 is constant, then the approximation
is trivial. If not, we have
c = inf{/o(x) : x EX} < d = sup{/o(x): x EX}.
Let 1= d 2 c (fo-d) + I, so that I(X)c [-I, I] and inf/=-I,
supl = 1. It obviously suffices to prove that I lies in the closure of e.
Consider the nonvoid compact sets E = {x EX: I(x) ~- f} and
F = {x EX: I (x) ~ ~}. For every x E E andy E F there exists gx,yE e
such that gx,y (x) =!= gx,y (y). Define
hx,y= 3(gx.,(Y) ~gx.,(X)) (gx,y - gx,y(Y) +~ .

We have hx, y (x) = - ~ and hx,y(y) = ~ . Since e is a linear space, it is
clear that hx,y E e. Since hx,y is continuous, there exists for each x E E
andYEF a neighborhood Ux of x such that hx,y(w) < - ~ for all wE Ux'
Since E is compact and U Ux::J E, there are points xl> x 2, ••• , Xm E E
xEE
such that
Ux, U Ux, U ... U UXm ::J E .
Let ({Jy = min{hx"y, hx"y, ... , hXm'y}, Since min{a, b} = -max{-a, -b},
hypothesis (v) shows that ({Jy E e. It is clear that ({Jy (y) = ~ and that
({Jy (x) < - ~ for all x E E. Note that the function ({Jy is defined for every
fixedy E F.
We repeat the above technique to find points Yl> Y2' ... , Yn EF
and functions ({JYl' ({Jy" ... , ({Jy" E e such that ({JYj(x) < - ~ for all
x E E and such that for each x EF, some ({JYj (x) is greater than !.
1 Note that our hypotheses are slightly redundant: if X admits a separating
family of continuous real-valued functions, then X has to be a Hausdorff space.
Hence the function 1p = max{ If"" If"" ... , If"J is in e and satisfies the
following inequalities: 1p (x) < - ! for all x EE and 1p (x) > ! for all
x EF. Now define WI by
wI=min{max{1p,- ~}, ~}.
It is clear that wIEe, WI (E) ={- !}, wI(F) ={!}, and WI(X) C
[- ~ , ~] . The definitions of E and F show that
2
Ilf - WIll .. = 3'
The function ~ (f - WI) is in ~r (X) and has minimum - 1 and maxi-
mum 1. The method used to construct WI can again be used to approxi-
mate ~ (f - WI)' Thus there exists W 2 E e such that
II ~ (f - WI) - w2 11u = ~ .
Multiplying by ~ , we have
Our scheme is now clear. In the next step we approximate
by a suitable function W3 in e, obtaining the equality
In general, if n is any positive integer, there are functions WI' ... , Wn

such that
where each Wj is in e. Since e is a linear space and lim (~)n = 0,

n--+oo 3
the proof is complete. D
The standard version of the approximation theorem is simple to
prove from (7.27) and (7.29).
(7.30) STONE-WEIERSTRASS Theorem. Let X be a nonvoid compact
e
Hausdorff space and a separating family of functions in ~r(x) containing
the function 1. Then polynomials with real coefficients in functions from e
are a dense subalgebra of ~r (X) in the topology induced by the uniform
metric.
Proof. Let <,p be the set of all polynomials with real coefficients in
functions from e, and let - be the closure operation in ~'(X). Clearly<,p
is a sub algebra of <rr (X). Suppose that I and g are in $-, and that
lim III - Inll" = 0 ,
n~oo
lim Ilg - gnll" = 0 ,
n~oo
where In and gn are in $. Theorem (7.27) proves that max {tn' gn} is in $-.
It is easy to see from (5.36) that
Ilmax{tn' gn} - max{t, g}ll" ~ Il/n - III" + Ilgn - gil" ,
and so max{t, g} is in ('ir)- = $-. Thus $- satisfies all the hypotheses
imposed on 6 in (7.29), and (7.29) therefore implies that ($-)- = $-
is <rr(x). 0
(7.31) Corollary [WEIERSTRASS]. Let X be a compact subset 01 R
and let I E<rr (X). Then there is a real polynomial P = P (x) such that
III - PII" is arbitrarily small.
Proof. The corollary is simply (7.30) with 6 = {t, I} where t (x) = x
for all x EX. 0
(7.32) Remarks. The hypothesis in (7.29) and (7.30) that 6 be
a separating family of functions is obviously necessary. Let X be a
compact Hausdorff space containing at least two points. Suppose that
5' is a nonvoid subset of <rr (X) such that for some distinct a, b EX, we
have I (a) = I (b) for all I E5'. Then P (a) = P (b) for every polynomial P
in functions from 5'. It follows that polynomials in functions from 5'
cannot be dense in <rr(x). In fact, no function having different values
at a and b can be arbitrarily uniformly approximated by polynomials
in the functions of 5', and Theorem (6.80) shows that there is a CfJ E<rr (X)
such that CfJ (a) = 1 and CfJ (b) = O. The beauty of the STONE-WEIERSTRASS
theorem lies in the fact that its hypothesis, trivially necessary, is also
sulficient. We will use the STONE-WEIERSTRASS theorem very frequently.
It is an essential tool for the analyst.
(7.33) Examples. (a) Let X be CANTOR'S ternary set (6.62), which we
write as the set of all numbers 2 1; ~:, where eachYk is 0 or 1. For n EN,
k~l
let CfJn be the function on X such that
CfJn (2 i
k~l
~:) = (_1)Yn.
n
I
For n l < n 2 < ... < n z, let CfJnl,n" ••• ,nz = CfJnr
i~l
Then each CfJn is continuous on X, the set {CfJl> CfJ2' ••• , CfJn' ••• } is
a separating family on X, and CfJ; = 1. Hence every function in <rr (X)
is arbitrarily uniformly approximable by a linear combination of the
functions 1 and CfJnl'n" •.. ,nz.
(b) The function exp defined as in (5.56) is real-valued on Rand
satisfies, as every schoolboy should know, the inequality exp (Xl) < exp (x 2 )
if Xl < x 2. Hence the one-element family {exp} is a separating family

on R, and the family of polynomials in exp and 1 is dense in <r' (X)
for every compact subset X of R. These polynomials are precisely all
n
functions P of the form P(x)=J;rx"exp(n"x), where the rx,,'s are
k=O
real numbers, the n,,'s are nonnegative integers, and n = 1, 2, 3 ....
(c) Polynomials in the cosine function and 1 are dense in <rr([o, nJ).
For every n EN, cosn(x) can be written as a linear combination of terms
of the form cos(kx) and 1 [this fact can be proved by induction on nJ.
Hence the family of all linear combinations of the functions 1, cos (x),
cos (2x), ... is dense in <rr([o, nJ).
(d) Consider <rr([o, IJ) and the smallest subset 6 of <rr([o, IJ) con-
taining 1, the function l [l(X) = xJ, and satisfying (7.29.iii)-(7.29.v).
It is not hard to see that 6 is exactly the linear lattice consisting of all
piecewise linear, continuous, real-valued functions f on [0, 1]. That is,
there are a finite number of subintervals [0, x1J, [Xl' X 2J, ... , [Xn - 1, IJ
°
such that < Xl < ... < x n _ 1 < 1 and such that f is linear on each
[Xk-1' Xk]. Theorem (7.29) shows that 6 is dense in <rr([o, IJ).
One would expect a complex version of the STONE-WEIERSTRASS theo-
rem, and indeed there is one. However, in the complex case some ad-
ditional hypothesis is required.
(7.34) Theorem. Let X be a nonvoid compact Hausdorff space. Let 6
be a separating family of functions in <r (X) containing the function 1 and
such that 1E6 whenever f E6. Then polynomials with complex coefficients
in functions from 6 are dense in <r (X) in the topology induced by the uni-
form metric.
Proof. Let g be in <r (X). We will show that the real-valued continuous
functions Reg and Img can be approximated by polynomials in functions
belonging to 6. For f E 6, we have Ref = 1_; 1 and Imf = f 2i 1 ,
and so Ref and Imf are polynomials in functions from 6. The family of
functions 6 0 , consisting of all Ref and Imf for f E 6, is a separating
family on X, for if x, y EX and f (x) =1= f (y), then either Ref (x) =1= Ref (y)
or Imf(x) =1= Imf(y). Also 6 0 contains the function 1. Theorem (7.30)
shows that for every 6> 0, there are polynomials P and Q in functions
from 6 0 such that
1
liP - Regll" < 2 6 ,
1
IIQ - Imgll .. < 2 6 ,
and so
IIP+iQ-gllu<6. D
(7.35) Examples. (a) Let X = {z EK: JzJ ~ I}, and let e = {t}
where t (z) = z. Unifonn limits of polynomials in t and 1 are analytic
on the open unit disk {z E K: JzJ < I} and continuous on the closed unit
disk {z EK : JzJ ~ I}. There are certainly nonanalytic continuous func-
tions on the unit disk, and so polynomials in functions from e are not
dense in (i (X). Thus some additional hypothesis is necessary in the com-
plex STONE-WEIERSTRASS theorem. Conditions onX under which the theo-
rem holds for (i (X) with no additional hypothesis have been found by
W. RUDIN [Proc. Amer. Math. Soc. 7, 825-830 (1956)].
(b) Let T = {z EK: JzJ = I}, and consider the function t on T
given by t(z)=z. For any zET, we have z=exp(ix) = cos (x) +i sin (x),
for exactly one x E [0, 2n[. Hence t(exp(ix) = exp(ix), and we see
that {t} is a separating family for T. We have t(z) = t(exp (ix) = cos (x)
- isin(x) = cos (-x) + isin(-x) = exp(-ix). Recall that exp(ix)fI
= exp(inx), n = 1,2, .... The family {t, t, I} satisfies the conditions
of (7.34). Now let 1 E (i(T) and let e > O. By the above remarks there is
a function
.
exp (ix) -+ 1: cx" exp (ikx)
k=- ..
such that
IkA . cx"exp(ikx) - 1(exp (iX»1 < e

for all x E [0, 2n[ [actually for all x E R]. This result is important in the
theory of trigonometric series [see (16.34)], and the proof we have given
is the shortest one that we know of.
(7.36) Exercise. (a) Let X be any noncompact subset of R. Find a
separating family e in (ir (X) such that polynomials in the family e U {I}
are not dense in (ir (X). Be careful to consider all possible cases. How
small can you make e?
(b) Let A be a nonvoid, finite, discrete space. Prove a sharpened
fonn of the STONE-WEIERSTRASS theorem for this set, and do it without
recourse to (7.30).
(7.37) Exercise. There are several versions of the STONE-WEIERSTRASS
theorem besides (7.29) and (7.30), which are listed below. Prove them,
using (7.30).
(a) For a compact Hausdorff space X, a closed subalgebra ~ of
(i' (X) that separates points is either (i' (X) or {I E (ir (X) : 1(xo) = O}
for a fixed point Xo EX. A subalgebra ~ of (i' (X) that separates points
and vanishes identically at no point of X is dense in (i'(X). [This state-
ment differs from (7.30) in that ~ need not contain the function 1.]
(b) For a locally compact Hausdorff space X, a closed subalgebra
~ of (i~ (X) that separates points is either (i~ (X) or {I E(i~ (X):I (xo) = O}
for a fixed point Xo EX. A separating subalgebra ~ of (t~ (X) vanishing

identically at no point of X is dense in (t~(X).
(c) State and prove complex versions of (a) and (b).
(7.38) Exercise. Theorems (7.29) and (7.30) are incomparable, in the
sense that subalgebras of (tr (X) need not be sublattices and sublattices
need not be subalgebras.
(a) Let P and Q be real polynomials on [0, 1]. Prove that max{P, Q}
is a polynomial if and only if P ~ Q or Q ~ P. Prove, however, that if
Pl' Ql' Pa, and Qa are polynomials such that Pl ~ Pa, Pl ~ Q2' Ql ~ Ps,
and Ql ~ Qa, then there is a polynomial l/J such that max{Pl , Ql}
~ l/J ~ min{Pa, Qa}.
(b) The sublattice e of (tr([o, 1]) defined in (7.33.d) is not a sub-
algebra of (tr([o, 1]). Prove that if I and 12 are both in e, then I is a
constant. Find conditions on I and g in e necessary and sufficient for
Ig to be in e.
(7.39) Exercise. Let X be a nonvoid compact subset of Rn, where
n EN. For re = (Xl' .•. , xn) E X and (kl>"" kn) a finite sequence of
nonnegative integers, consider the function
re -+ X~l x~• ... x!n .
Prove that linear combinations of these functions are dense in (tr(x).
State and prove an analogue for X C Kn and (t(X).
We close this section with an important application of the STONE-
WEIERSTRASS theorem.
(7.40) TIETZE'S Extension Theorem. Let X be a locally compact
Hausdorff space, let Y be a nonvoid compact subspace 01 X, and let U
be an open set such that Y cUe X. Then every lunction in (t(Y) can be
extended to a lunction in (too (X) that vanishes on X n U'.
Proof. In view of (6.79), we lose no generality in supposing that r;-
is compact. It obviously suffices to show that every function I in
°
(tr (Y) admits an extension It in (tr (X) such that It (x) = for all x Exn U'.
Let e be the set of all functions I in (tr (Y) admitting such extensions It.
It is trivial that e is a subalgebra of (tr(y). Let us show that e separates
points of Y. For a =!= b and a, bEY, there is a neighborhood W of a
such that b ~ W. By (6.79), there is a neighborhood V of a such that
V- C U n W and V- is compact. By (6.80) there is a continuous func-
tion g; from X into [0, 1] such that g; (v-) = {I} and g;«U n W)') = to}.
Thus g; (a) = 1, g; (b) = 0, g; (X n U') c to}, and the restriction of g;
to the domain Y is plainly in e. That is, e separates points of Y. Theorem
(6.80) shows too that all constant functions on Y are in e.
Next consider any lEe, and write IX = max{/(y): y E Y},
P= min{/(Y) : y E Y}. It is obvious that 11111" = max{llXl, IPI}. Let It
be a continuous real-valued extension of lover X such that
7·
It (X nU') C {O}. Let oct = max{oc, O} and {Jt = min{{J, O}. Then define
the function q; on X as
q; = min{max{/t, {Jt}, oct}.
It is clear that q; is in err (X), and it is easy to see that: q; (x) = I (x) for
xEY; oct=max{q;(x):xEX}; {Jt=min{q;(x):xEX}; q;(x) =0 for
x EX nu'. Thus q; is an extension of I of the sort we want for which
11q;llu = Il/ilu· That is, if lEe, then I admits an extension It such that
!Wllu = 11/11u·
N ow let g be a function in e~:r (Y) that is the uniform limit on Y
of a sequence (fn)":;:=l of functions in e. Choose a subsequence (fnk)":=l
such that Il/nk - InkJu < 2- k and write gk = Ink - Ink+! (k = 1,2,3, ... ).
Then we have
where the infinite series converges uniformly on Y, and Ilgkllu < 2- k.

Now let gt be a continuous extension of gk over X such that gt vanishes
00
on U' and Ilgtllu = Ilgkllu < 2- k • The infinite series 1: gk converges uni-
k=!
formly on X to a function in crr (X) that vanishes on U' . Therefore g - Inl
is in e and hence g itself is in e.
We have therefore shown that e satisfies the hypotheses of (7.30)
and is uniformly closed in crr (Y). Therefore e is all of crr (Y). 0
(7.41) Exercise. Let X be a locally compact Hausdorff space. Every
function I in cro (X) can be arbitrarily uniformly approximated by
functions in croo (X). [Hint: consider a compact set outside of which I is
small and use (7.40).J
(7.42) Exercise. Let E be a real or complex normed algebra with
multiplicative unit u. For x E E, let
Illxlll = sup{llyxll:y EE, IIYII ~ I}.
Prove the following:
(a) Illulll = 1;
1
(b) l1Ulf Ilxll ~ Illxlll ~ Ilxll for all x E E;
(c) the algebra E with the norm III III is a normed algebra in the sense
of (7.5);
(d) if E is complete in the metric e(x, y) = Ilx - yll, it is also complete
in the metric Illx - ylll·
(7.43) Exercise. Let X be a nonvoid compact Hausdorff space and
let .s: be a lattice of real-valued continuous functions on X, i. e., I, g E .s:
implies max{j, g} E.s: and min {I, g} E.s:. Suppose that q; is a real-valued
continuous function on X having the property that for each B > 0
and each pair of points x, y E X there exists a function IE.s: such that
l<p(x) - I(X)I < e and l<p(y) - t(y)1 < e. Prove that for each e >
there exists a function h E~ such that 11<p - hll u < e.
°
(7.44) Exercise [ROBERT I. JEWETTJ. Let I denote the interval
[0, 1J. A family fr of functions is said to have property V if
(i) fr C IX for some set X,
(ii) I Efr implies (1 - t) Efr,
and
(iii) I, g Efr implies Ig Efr·
Supply IX with the topology of uniform convergence [the topology
induced by the uniform metricJ. Prove assertions (a)-(k). [This exercise
is rather difficult.J
(a) If X is a set and Q( C IX, then Q( is contained in a smallest sub-
family [closed subfamilyJ of!X having property V. [Intersect a collection
of families. J
Suppose that X is a topological space, and let ~ (X) = IX n ~(X).
For n EN, let '.Pn be the smallest subfamily of ~ (In) that has property
V and contains the n projections
(Xl' ... , x n ) -J> X,. (k = 1, ... , n) .
(b) If fr has property V, if {Iv' .. , In} C fr, and if p E'.Pn' then the
function I defined on X by
I(x) = P(tI(X), ... , In (x))
is in fro [Consider the set of all P E~ (In) for which the conclusion holds.J
(c) For every e> 0, there exist functions P and q in '.PI such that
P < e and q > 1 - e on I. [Consider P(x) = xm(1 - x)m.J
° °
(d) If e > and < a 1 - e on [0, aJ,
P< e on [b, IJ .
[Take P (x) = (1 - xm)n for appropriate m, n EN.J
(e) If {aI' ... , an} U {b v ... , bn} C I, then
I
ii b,. - fI a,.\;:;;; i: Ib,. - a,. I .
1<=1 1<=1 1<=1
[Use induction on n.J

(f) If (a, b) E12 = I x I and e and <5 are positive real numbers, then
there is a function P E'.P2 such that
P(x, y) > 1 - e if (x - a)2 + (y - b)2 ;:;;; <5 2
and
P(X,y) < e if (x - a)2 + (y - b)2 ~ (4<5)2
for (x, y) E12.
[Take P(x, y) = (1 - PI (X))P2(X) (1 - P3(Y))P,(Y), where the P/s are

obtained by using (d).]
(g) Let A and B be disjoint compact subsets of ]2, and consider
arbitrary 8 > 0 and P E '.l32' Then there exists q E $2 such that
q 6; P on ]2,
q>I-8 on A,
and
q<P+8 on B.
[Let 4<5 = dist(A, B). Use compactness and (f) to piece together a
qo E$. such that qo is near 1 on B and near 0 on A. Then let q =
1 - (1 - P)qo.]
(h) Define ({J and "P on ]2 by ({J(x,y) = max{x,y} and "P(x,y)
=min{x,y}. For each 8>0 there exist functions r,qE$2 such that
II({J-rll,,< 8
and
II"P - qllu < 8.
[Let 8<5 = 8 < 1. Let C = {(x, y) E]2 : <5 ;;;;; "P (x, y) ;;;;; 1 - <5} and choose
mEN such that (xy)m < <5 on C. Let P (x, y) = 1 - (xy)m. Then P E $2
and 1- <5 <P < 1 onC. Fork 6; OletA,,= {(x,y) EC: P"(x,y) ;;;;; "P(x,y)}
and B" = {(x, y) E C : P" (x, y) 6; "P (x, y)}. Choose n EN such that B" = J25.
Note that A"_l n B" = J25 for all k. For k = 1, ... , n, use (g) to find
d d
q" E$2 such that q" 6; P on ]2, q" > 1 - n on A"_l' and q" < P + n
on B". Put q' = qlq2 .•. q". Prove that 0 ;;;;; p" - "P < <5 on B" n B~+l
and use (e) to show that IP" - q'l < 3<5 on C (k = 1, ... , n - 1). Thus
conclude that I"P - q'l < 4<5 on C. Use (g) again to find q" E$B such that
q" 6; q' on ]B, q" > 1 - <5 if "P 6; 1 - <5, and q" < q' + <5 if "P ;;;;; 1 - 2<5.
Notice that I"P - q"l < 6<5 if "P 6; <5. Use a slight variant of (g) to find
q E$B such that q ;;;;; q" on ]B, q < <5 if "P ~ <5, and q > q" - <5 if "P 6;2<5.
Then Iq - "PI < 8<5 = 8 on ]2.J
(i) Let X be a nonvoid set and let 5' be a closed subfamily of IX
having property V. Then 5' is a lattice. [Use (b) and (h).]
(j) Let X be a compact Hausdorff space and let 5' be a closed separat-
ing subset of !$) (X) having property V. Let S = {x EX: 1(x) E {O, I}
for all 1 E5'}. Then 5' = {I: 1 E!$) (X), 1(5) C {O, I}}. [Use (7.43), (i), (b),
and (d).J
(k) If n E N and 5' is a closed subfamily of !$) (],,) which has property
V and contains the n projections as well as some function which is never 0
or 1, then 5' = !$) (]").
(7.45) Exercise. The algebra 01 quaternions l H is defined as the
4-dimensional vector space over R having a basis, traditionally written
1 Discovered by the Irish mathematician SIR WILLIAM ROWAN HAMILTON
(1805-1865).
as {I, i,j, k}, with the following multiplication table:

i
1 j k
- - I-- -
1 1 i j k
- - - - -
i i -1 k -j
-
j j - k -1 i
k k j - i -1
Products (a I + bi + cj + dk) (a'l + b'i + c'j + d' k) [real coefficients!]

are defined by supposing that H is an associative algebra over R. Prove
the following.
(a) For every al + bi + cj + dk EH,
(al + bi + cj + dk) (al - bi - cj - dk) = (a 2 + b2 + c2 + d2) 1.
(b) The set Hn{OI+Oi+Oj+Ok}, is a non-Abelian group
under multiplication.
(c) For x = al + bi + cj + dk EH, we have x - ixi -jxj-kxk=4a 1.
1
(d) For x = al + bi + cj + dk EH, let Ilxll = (a 2 + b2 + c2 + d2)'2.
Prove that I I is a norm on H (7.5) and that IlxY11 = Ilxll IIYII for all
x,Y EH.
(e) Let X be a topological space and <reX, H) the set of all continuous
mappings t of X into H [make H into a metric and hence topological
space via the norm] for which Iltll .. = sup{llt(t)11 : t EX} is finite. Show
that <rex, H) is a noncommutative Banach algebra over R with a multi-
plicative unit, all operations being pointwise.
(f) [J. C. HOLLADAY]' Let X be a compact Hausdorff space and let Ql
be any subalgebra of <reX, H) [closed in particular under multiplication
by all constant functions] that separates points and contains 1. Prove
that Ql is dense in <reX, H). [Use (c) and (7.30), regarding <rex, R) as a
subring of <reX, H).]
CHAPTER THREE
The Lebesgue Integral

Integration from one point of view is an averaging process for func-
tions, and it is in this spirit that we will introduce and discuss integration.
In applying an averaging process to a class 5' of real- or complex-valued
functions, a number I (f) is assigned to each I E5'. If I (f) is to be an average,
then it should certainly satisfy the conditions
I(f + g) = 1(1) + I (g) ,
I (rxl) = rxI (I)
for I, g E5' and rx ER. A less essential but often desirable property for I
is that I (I) ~ 0 if I ~ O. In some cases these three properties suffice to
identify the averaging process completely.
Let us mention a few such averaging processes. Suppose for example
that 5' is all real-valued functions on the finite set {I, 2, 3, ... , n},
i.e., 5' = Rn, and define
I if j = k,
{
ej(k)= 0 if j+k;
that is, ej (k) = bjll. For each I E5', we have
n
1=1:IU)ei·
j=!
For any "integral" I we must have

n
1(1) = 1: IU) I (e i ) .
j=!
In fact if we choose the numbers I (el ), I (e 2) , ••• , I (en) arbitrarily,

then an integral I satisfying the first two conditions is determined by the
above sum for all I E5'. Hence in this case the integral is merely a finite
sum in which certain weights have been assigned to the points in the
domain of the function. To satisfy the third property, we need only
require that I (e i ) ~ o.
!
If 5' = <r([O, 1J) and 1(1) = f I (x) dx [the Riemann integralJ, then I
o
is an averaging process for 5'. A simple-minded average defined for any
class of functions defined at x = ~ is given by I (I) = I (~ ). Both of
these averaging processes satisfy the third property: I (I) ~ 0 if I ~ o.
§ 8. The Riemann-Stieltjes integral 105
Let (Xn ):=l be an enumeration of a countably infinite set C, let (ce n )

co
be a sequence of complex numbers such that 1: !cen ! < 00, and let 5'
n=l
be the set of all bounded complex-valued functions defined on C. Then I
co
defined on 5' by I (f) = 1: cenf (xn) is an average. It satisfies the third
n=l
property if and only if cen ~ 0 for all n.
As a final example, let 'J be all complex-valued functions t on the
set C of our last example such that}; It (Xn ) 12 < 00. For fixed g E'J
co
let Ig(f) =}; f(xn) g(xn). This series converges absolutely by CAUCHY'S
inequality (13.13), i. e.,
IIgU)1 ~ Cf If(Xn)!2)} Cf Ig(Xn)12)~.

In this chapter we will first discuss averages of continuous functions
and then extend the process to wider classes of functions. In particular,
we will extend the Riemann integral qua averaging process to obtain
the Lebesgue integral.
We begin with a rapid review of the Riemann-Stieltjes integral,
which is a classical device for averaging continuous functions on inter-
vals [a, bJ.
§ 8. The Riemann-Stieltjes integral

(8.1) Definition. Let A be a subset of R. An extended real-valued
function IX defined on A is said to be nondecreasing on A if IX (x) ~ IX (y)
whenever x < y in A. In case IX (x) < IX (y) whenever x < y in A, we say
that IX is strictly increasing on A. The terms nonincreasing and strictly
decreasing are defined analogously. A function is said to be monotone or
monotonic if it is either nondecreasing or nonincreasing 1 .
(8.2) Definition. Let [a, bJ be any closed interval in R. A subdivision
of [a, bJ is any finite [ordered J set LI C [a, bJ of the form
LI = {a = to < t1 < ... < tn = b} .
The family of all subdivisions of [a, bJ is denoted by ~([a, bJ).

(8.3) Definition. Let [a, b] be a closed interval in R, let IX be a real-
valued non decreasing function on [a, b], and let be a bounded real- t
valued function on [a, bJ. For each LI = {a = to < t1 < ... < tn = b} E
1 This definition plainly generalizes that given in (6.81) for sequences: simply
take A = N in (8.1) to obtain (6.81).

106 Chapter III. The Lebesgue integral
9J([a, b]), define

L(/, ac, Lt) = 1:" inf{/(x): x E [ti - I, tiJ}' (ac(ti) - ac(tS_I)) '
i=1
and
U (I, ac, Lt) = 1:" sup {I (x) : x E [ti -I' tiJ} . (ac (tl ) - ac (ts -I)) .
i=1
These numbers are known as lower Darboux sums and upper Darboux
sums, respectively. Suppose that for every e > 0 there exists a Lt E9J([a,b])
[depending on I, ac, and e] such that
U(I, ac, Lt) - L(f, ac, Lt) < e .
Then I is said to be Riemann-Stieltjes integrable with respect to ac on [a, b].
(8.4) Lemma. Let [a, b], I, and ac be as in (8.3). Suppose that Lt,
Lt* E9J([a, b]) and that Lt c Lt*. Then we have
(i) L(/,ac,Lt) ~ L(I,ac,Lt*) ~ U(I,ac,Lt*) ~ U(/,ac,Lt).
Proof. The middle inequality is obvious. Let us show that
L(/, ac, Lt)
~ L(I, ac, Lt*). Suppose first that Lt* contains just one more
point than Lt. Thus if Lt = {a = to < ~ < ... < tn = b}, then
Lt* = {a = to < ~ < ... < tk _ 1 < u < til, < ... < tn = b}
for some k. We have
L (I, ac, Lt*) - L (I, ac, Lt) = inf{/(x) : x E[t k_l , uJ}· (ac(u) - ac(tk_I ))
+ inf{/(x): x E[u, tkJ}' (ac(tk) - ac(u))
- inf{/(x): x E[t k_l , tkJ}' (ac(tk) - ac(tk_l ))
= (inf{/(x):x E[t k_l , uJ} - inf{/(x):x E[tk-l>tkJ}) (ac(u) - ac(tk_l ))
+ (inf{1 (x):x E[u, tkJ} - inf{1 (x):x E[tk_l , tkJ}) (ac(tk) - ac(u))
~o.
Hence L(I, ac, Lt*) ~ L(I, ac, Lt). The proof by induction on the number
of points in Lt* and not in Lt is now clear. Similarly, suprema decrease
on subintervals of an interval, and so the last inequality in (i) also
holds. 0
(8.S) Lemma. II Lt, Lt* are in 9J([a, b]), then L(I, ac, Lt) ~ U(/, ac, Lt*).
Proof. By (8.4) we have L (I, ac, Lt) ~ L (I, ac, Lt U Lt *) ~ U (I, ac, Lt*). 0
(8.6) Theorem. Let I be Riemann-Stieltjes integrable with respect to ac
on [a, b]. Then there exists a unique real number I' such that
L (I, ac,Lt) ~ I' ~ U (I, ac, Lt)
lor every Lt E9J ([a, b]). In fact,
I' = sup{L (I, ac, Lt) : Lt E9J ([a, b])} = inf{U (I, ac, Lt) : Lt E9J([a, bJ)}.
This number I' is called the Riemann-Stieltjes integral 01 I with respect
b
to acover [a, b]. Historically it is denoted by J I (x) dac(x), but in this text we
..
write it as sa.(I; [a, b]). In case a(x) = x for each x E [a, b], we call y
the Riemann integral 01 lover [a, bJ and denote it by 5 (I; [a, bJ).
Proof. Let y = sup{L(f, a, .1) : .1 EE&([a, bJ)} and b = inf{U(I, a, .1):
.1 EE&([a, bJ)}. It follows from (8.5) that y and b are real numbers
and that y ~ <5. We need only show that y = <5. Assume that y < <5.
Since <5 - y > 0, the definition of integrability (8.3) shows that there is a
.1 EE&([a, bJ) such that U (I, a, .1) - L (j, a, .1) < <5 - y. Then
b ~ U (I, a, .1) = (U (f, a, .1) - L (I, a, .1)) + L (f, a, .1) < (<5 - y) +y= <5,
a clear contradiction. Thus y = <5. D

(8.7) Theorem. Let [a, bJ, 1 and a be as in (8.3). If f is continuous
on [a, bJ, then 1 is Riemann-stieltjes integrable with respect to a over
[a, bJ.
°
Proof. Let s > be given. Since [a, b] is compact (6.44), the function f
is uniformly continuous on [a, b] (7.18). Thus there exists a <5 > 0 such
that If (x) - f(y)1 < ec(b) _ :(a) + 1 whenever x,yE[a, bJ and Ix- yl <<5.
Choose .1 = {a = to < tl < ... < tn = b} such that t; - t;_1 < <5 for
j = 1, ... , n. Select x;,Y; E [t;_I, t;J such that
I (x;) = sup{f(x): x E [ti _ l , ti]}
and
fey;) = inf{f(x) : x E [ti _ l , tiJ}
for j = 1, ... , n. [Such selections are possible by (6.73).J Then we have
IXi - Yil ~ t; - ti _ 1 < <5, so that
o ~ I (x;) - f(Yi) < ec(b) _ : (a) +1 .
Thus we also have
n
U(I, a, .1) - L(I, a, .1) = 1: (I(x;) - f(Yi)) (a (ti) - a(ti _ l ))
;=1
n e
< iJd.. ec(b) - ec(a) + 1 (a (t;) - a (t;_I))
= ec(b)-:(a) +1 (a(b)-a(a))<s. D
(8.8) Theorem. II 11 and 12 are both Riemann-stieltjes integrable with

respect to a over [a, b], then so is 11 + 12 and sa.(fl + 12; [a, b])
= sa.(ll; [a, b]) + Sa. (12 ; [a, b]).
Proof. Let s > 0 be given. Choose .1; such that U (Ii, a, L1i) -
L (I;, a, .1 i) < ~ for j = 1,2, and let .1 = .11 U .1 2 ' Using (8.4) and
the simple inequalities inf (11 + 12) ~ infll + inff2 and sup (11 + 12) ~
supll + sup/2 [valid for any bounded real-valued functions on any
nonvoid set], we have

L(l1> a, ,,11) + L(l2' a, ,,12)
L(l1' a, ,,1) + L(l2' a, ,,1)
~
~ L (II + 12' a, LI)
~ U (/1 + 12, a, LI)
~ U (/1> a, LI) + U (/2' a, LI)
~ U(/1> a, Ll 1) + U(/2' a, Ll 2)
< L (/1' a, Ll 1) + L (/2' a, Ll 2) + e .
It follows that U (/1 + 12' a, LI) - L (/1 + 12' a, ,,1) < e and hence 11 + 12
is integrable.
Next let S",(/i; [a, b]) = Yi and let r j be such that
IL (/j, a, r j) - Yjl < ~- , (1)
and
(2)
r= r 1 U r 2 , we have
°
It follows that ~ U (lj' a, r j ) - L (/j' a, r j ) < ~ for j = 1, 2. Setting
L (/1' a, r 1) + L (12' a, 1'2) ~ L (/1' a, r) + L (/2' a, T)

~ L(/l + 12, a, T)
;5U (/1 + 12, a, T)
~ U (/1> a, T) + U (12' a, r)
r
~ U (/1> a, 1 ) + U (/2' a, 2) r
28
< L (/1' a, 1'1) + L (/2' a, 1'2) + 3 '
from which we see that
28
L (II + 12' a, T) - L (11' a, r 1) - L (f2' a, r 2) < 3 (3)
and
28
U (/1' a, r 1) + U (12' a, r 2) - U (II + 12, a, r) < 3 . (4)
From (1) and (3) we infer that
IL(fl + 12' a, T) - (Y1 + Y2)1 < e,
and from (2) and (4) that
IU(ll + 12' a, T) - (Y1 + Y2)1 < e. 0
(8.9) Theorem. II I is Riemann-Stieltjes integrable with respect to a
over [a, b] and il c ER, then so also is cl and S",(cl; [a, b]) = cS",(I; [a, b]).
Proof. Exercise.
(8.10) Theorem. Let [a, b], I, and a be as in (8.3). II f is Riemann-
Stieltjes integrable with respect to a over [a, b] and f ~ 0, then S'" (f; [a, b]) is
nonnegative.
Proof. Trivial.
(8.11) Theorem. Let I be Riemann-5tieltjes integrable with respect to ex

over [a, b] and let a < c < b. Then I is Riemann-5tieltjes integrable with
respect to IX over both [a, c] and [c, b] and 5,,(1; [a, b]) = 5,,(1; [a, c])
+ 5,,(1; [c, b]).
Proof. Let e > 0 be given. Choose LI E~([a, b]) such that
U (I, IX, LI) - L (I, IX, LI) < e.
In view of (8.4) we may, and do, suppose that cELl; say
LI = {a = to < ... < tm = c < tm + 1 < ... < tn = b} .
Let
Lli = {a = to < ... < tm = c} E~([a, c])
and
Then
(U(I, IX, Ll I ) - L(I, IX, Ll I ) + (U(I, IX, Ll 2) - L(I, IX, Ll 2)
= U(I, IX, LI) - L(I, ex, LI) < e.
It follows that I is integrable over both [a, c] and [c, b]. Let 5,,(I;[a,c])
= YI and 5,,(j; [c, b]) = Y2. Clearly 0 ~ U(f, IX, Ll I ) - YI < e and
o ~ U (I, IX, Ll 2) - Y2 < e. Adding these two inequalities, we get
o ~ U(j, IX, LI) - (YI + Y2) < 2e. Since a similar statement is true
for L(I, IX, LI), we conclude that 5,,(1; [a, b]) = YI + Y2. 0
(8.12) Theorem. Let IX be a real-valued nondecreasing lunction de-
fined on R. For I E<r~o(R), define 5,,(/) = 5,,(/; [a, b]) where [a, bJ is any
closed interval in R such that I vanishes outside 01 [a, b]. Then 5,,(/) is
unambiguously defined lor each I E<r~o([a, b]). Moreover the lunction 5"
has the lalla wing properties:
(i) 5,,(j + g) = 5,,(/) + 5,,(g);
(ii) 5,,(c/) = c5,,(I);
(iii) 5,,(1) ~ 0 il IE <rto(R).
Proof. Let I E<r~o(R) and suppose that 1=0 outside of both [a, b]
and [a', b/]. Let a" = min{a, a /} and b" = max{b, b/}. By (8.11), we have
5,,(/; [a", b"]) = 5,,(/; [a", a]) + 5,,(/; [a, b]) + 5,,(1; [b, b"])
= 0 + 5,,(1; [a, b]) + 0
= 5,,(/; [a, b]) .
Similarly we see that 5,,(/; [a", b"J) = 5,,(f; [a', b/]). Therefore
5,,(1; [a, bJ) = 5,,(j; [a', b/]) and so 5,,(/) is well defined. The remaining
assertions of the theorem are now trivial consequences of previous
results. 0
(8.13) Theorem. Let IX be as in (8.12). For IE <roo (R), define 5,,(/)
as 5,,(Re/) + i5,,(Imf). Then lor I, g E<roo(R) and c EK, we have
(i) 5,,(j + g) = 5,,-(j) + 5,,-(g) ,
(ii) 5,,(cl) = c5,,(I) ,

(iii) 5" (I) ~ 0 itt ECito (R).
This simple computation is left to the reader.
(8.14) Definition. For any function I defined on R and any t ER,
let It be defined on R by It (x) = I(x + t). The function It is called the
translate 01 I by t. Let 1* be defined on R by 1* (x) = 1(- x). The function 1*
is called the reflection 01 I.
(8.15) Theorem. Let 5 be the Riemann integral on Cioo(R), i.e., 5=5"
where a: (x) = x lor all x ER. Then 5 has the lalla wing properties,'
(i) 5(1 + g) = 5(1) + 5 (g);
(ii) 5 (c I) = c5 (I) lor all c EK;
(iii) 5 (I) > 0 il I ECito (R) and I =f= 0;
(iv) 5 (It) = 5 (I) lor all t ER;
(v) 5(1*)=5(1).
Proof. Conclusions (i) and (ii) are contained in (8.13). To prove (iii),
let I ECicio (R) where I is not the zero function. Then there exists u ER
such that I(u) > O. Since I is continuous at u, there is some neighborhood
1
Ju - <5, u + <5[ of u such that I(x) > 2 I(u) whenever u - <5 < x < u+ <5.
Suppose that I vanishes off of [a, b], where a O.
The proofs of (iv) and (v) are left to the reader. D
(8.16) Remark. The function 5 is [except for a positive multiple]
the only complex-valued function on Cioo(R) satisfying (8.15.i)-(8.15.iv),
i. e., if 5' is another complex-valued function on Cioo (R) satisfying (8.15.i)
to (8. 15.iv) , then 5' = yS for some positive real number y. We will
prove this fact inlra [see (19.50)] when more analytical machinery has
been developed.
Theorem (8.13) shows that for each nondecreasing a: defined on R,
the Riemann-Stieltjes integralS" is an averaging process on Cioo(R).
We shall see later (19.50) that every averaging process on Cioo(R) is of
the form 5" for some a:. Our next theorem gives a sufficient condition
that two different a:'s give rise to the same averaging process on Cioo(R).
(8.17) Theorem. Let a: and {3 be two real-valued nondecreasing lunc-
tions defined on R and suppose that there exists acE R such that the set
D = {x ER: a:(x) = (3(x) + c} is dense in R. Then 5,,(1) = 5{J(I) lor every I
in Cioo (R) .
Proof. In view of (8.13) it obviously suffices to consider real-valued
f's. Thus let I ECi~o(R). Choose a 0 be given. Since I is uniformly continuous on R, there is a <5 > 0
§ 8. The Riemann-Stie1tjes integral 111
such that II (x) -/(y)1 < <x(b)-IX(a) + 1 whenever Ix- yl <C}. BecauseD
6
is dense in R, there exists a subdivision L1 = {a=to<···<tn=b} C D

such that ti - ti -1 < C} for j = 1, 2, ... , n. Then, as in the proof of (8.7),
it follows that
U(/, IX, L1) - L(/, IX, L1) < e.
Since LI cD, we have U(I, (J,L1) = U(f,IX,L1) and L(I,{J,L1)=L(f,IX,L1).
Therefore 15,8(1) - 5",(/)1 < e [see (8.6)]. Since e is arbitrary, the theorem
is proved. 0
We next examine a few interesting properties of nondecreasing func-
tions. First, a definition is in order.
(8.18) Definition. Let I be a complex-valued function defined on an
open interval ]a, b[ of the real line. We say that the right-hand limit 01
I at a is I(a+) [or that the lelt-hand limit 011 at b is I(b-)], and we write
lim I(x) I(a+) [or lim I (x) = I (b-)] ,
=
x~a xt b
if there exists a complex number I(a+) [or I(b-)] such that for each
e > 0 there is a C} > 0 such that
I/(x) -1(a+)1 < e [1I(x) -I(b-)I < e]
whenever x E]a, b[ and x - a < C} [or b - x < C}].
Next suppose that 1 is defined on [a, b]. We say that 1 is right con-
tinuous at a [left continuous at b] if lim 1(x) = 1(a) [lim f(x) = I(b)].
Xta xt b
(8.19) Theorem. Let be a real-valued nondecreasing function defined
IX
on R. Then IX has finite right- and left-hand limits at all points of R, and IX
is continuous except at a countable set of points of R.
Proof. Let x ER and set IX(X+) = inf{lX(t): x < t}. This infimum
exists in R since oc(x) ~ oc(t) whenever x < t. For e > 0, oc(x+) + e
is not a lower bound for {1X(t): x < t}, so there exists a lJ > 0 such that
oc(x + lJ) < oc(x+) + e. It follows that x < t < x + lJ implies oc(x+) ~
oc(t) ~ oc(x+) + e, i.e., lim oc(t) = oc(x+). Similarly we prove that
t~x
lim lX(t) = sup{oc(t): t < x} = IX(X-).

tt x
Let D = {x ER : oc is discontinuous at x}. Plainly xED if and only
if IX (x-) < IX (x+ ). Also x < y in D implies that oc (x+) ~ oc (y-). Thus
the family Jf = {JIX(X-), oc(x+)[: xED} is a pairwise disjoint family
of nonvoid open intervals of R. By (6.59), Jf is countable, and so D is
countable as well. 0
(8.20) Remarks. (a) Let oc be a real-valued nondecreasing function
on R. Define {J by the rule: (J (x) = IX (x-) - IX (0-) for x ER. Then {J
is nondecreasing, {J (0) = 0, and {J is left continuous at each point of R.
Moreover, since IX is continuous except on a countable set, we have
/3 (x) = a: (x) - a: (0-) on a dense subset of R. It follows from (8.17)

that 5{J(I) = 5,,(1) for allf E<roo (R). Because of these facts, we lose nothing,
in constructing Riemann-Stieltjes integrals on <roo (R), by normalizing
our a:'s to be left continuous and to vanish at o. It must be pointed out
however that this normalization may affect the value of the integral of a
discontinuous function. For example, let a: (x) = 0 for x < 0 and a: (x) = 1
for x ~ 0; /3 (x) = 0 for x ~ 0 and /3(x) = 1 for x> 0; and take 1=/3.
Integrating over [-1, 1], we have L (/3, IX, LI) = L (/3, /3, LI) = 0 and
U(/3, /3, LI) = 1 for all LI in .@([-1, 1]) while U(/3, IX, LI) = 0 if
o ELI E.@([-I, 1]). Therefore 5,,(/3, [-1, 1]) is zero and 5{J(/3, [-1, 1])
does not exist.
(b) The foregoing example of a non Stieltjes integrable function is
undramatic, to say the least. A bit more interesting is the function
on [0, 1] such that f(x) = 0 if x is rational and I (x) = 1 if x is irrational.
The Riemann-Stieltjes integral 5,,(1; [0, 1]) exists only if IX(I) =IX(O), since
all lower Darboux sums are 0 and all upper Darboux sums are IX (1) - IX (0).
A complete description of Riemann integrable functions appears in
(12.51) inlra.
(8.21) Exercise. Prove that if I is a bounded real-valued function
on [a, b] having only a finite number of discontinuities and if IX is a
real-valued nondecreasing function on [a, b] having no discontinuities
in common with I, then I is Riemann-Stieltjes integrable with respect
to IX over [a, b].
(8.22) Exercise. Let IX be a real-valued nondecreasing function on
[a, b] and let (In) be a sequence of bounded real-valued functions on
[a, b], each of which is Riemann-Stieltjes integrable with respect to IX
over [a, bJ. Suppose that lim II/-Inll" = 0, where I is in Q3r([a, b]).
n--->-oo
Prove that I is Riemann-Stieltjes integrable with respect to IX over
[a, b] and that
lim 5,,(ln; [a, b]) = 5,,(1; [a, b]) .
n--->-oo
(8.23) Exercise. By way of contrast with (8.22), find a sequence

(In) where each In E <roo (R), such that In -+ 0 uniformly on R, but
lim 5 (In) =to 0 = 5(0).
n--->-oo
(8.24) Exercise. Let g be defined on R by g(x) = x - [x], where [x] is
the largest integer not exceeding x. Using (8.21) and (8.22), prove that the
function I defined by I(x) = £ g(;:) is Riemann integrable over [0,1].
n=l
Also evaluate 5(1; [0, 1]).
(8.25) Exercise. Let IX and /3 be continuous, real-valued, nondecreasing
functions on R such that IX (0) = /3 (0) = 0 and IX =to /3. Find a function
IE <rto(R) such that 5,,(1) =to 5{J(I).
(8.26) Exercise. Let (Xn ):=l be an enumeration of Q. Define IX on R

by the rule : IX(X) = 1: ; .. , where the summation is extended over all n
such that Xn < x. Prove that:
(a) IX is strictly increasing;
(b) IX is left continuous;
(c) IX is discontinuous at each rational number;
(d) IX is continuous at each irrational number;
(e) lim IX(X) = 1;
x _co
(f) lim IX (x) = 0.
x-+-oo
[This example gives some indication of how bad a monotone function
can be.]
(8.27) Exercise. Let (xn) and IX be as in (8.26). Prove that Srx(f)
= £ t~nn) for every f E<roo (R) . [First consider f E<r~o(R),
,,= 1
show that
every Lis;;;;; this sum and that every U is ~ it.]
(8.28) Exercise. Let P denote the Cantor ternary set. Notation is
as in (6.62) and (6.64) . Define a function tp on [0, IJ as follows :
tp(O) = 0;
2k-l
tp(x) = 2n for x Eln,,, (n = 1,2, .. . ; k = 1,2, ... , 2n - 1 );
tp(x) = sup{tp(t) : t EP', t < x} for x EP n {oy.
The function tp is called LEBESGUE'S singular function 1. Prove that :
(a) tp(x) is defined for all x E [0, 1];
(b) tp is nondecreasing ;
(c) tp is continuous on [0, 1] ;
(d) rngtp = [0, 1];
(e) if x E P and y
~ 2xn h
x = ~ s;;-' were eac
h .L
7
xn =
co
°or 1, then tp(x)
.. = 1 6
.!.
'I
'\' Xn 5
= ~ z,;-' "6
.. =1 I
The accompanying Z
J
Fig. 5, showing part ofthe 6
graph of tp may be help- ~
fu!. [Use the facttha t the ~
dyadic rationals in [0, 1]
178 I l.IUIlZ '-(l1l1 X
[on the y-axis J form a iboT Jl'7l'78 9l'7&l
dense subset of [0, 1].] Fig. 5
1 The French mathematician HENRI LEBESGUE (1875-1941) was a principal

founder of the theories of measure, integration, and differentiation treated in this
hook. His name occurs frequently in the sequel.
(8.29) Exercise. Let 1jJ be as in (8.28) and let t(x) = x for all x E[0,1].
Prove that Sop(t; [0, IJ) = ~. [Consider the sequence of subdivisions
(L1n) where L1n consists of the endpoints of the closed intervals Jn, 11
making up P n .] For the computation of S",(t1l; [0, 1]) for k = 2, 3, ... ,
see (22.44) and (22.45) infra.
§ 9. Extending certain functionals

In § 8 we have presented Riemann-Stieltjes integrals as averaging
processes on <roo(R). We wish to extend these processes to much wider
classes of functions. To be sure, there exist many discontinuous functions
on R which are Riemann-Stieltjes integrable, but these integrals suffer
from serious defects. For example, very simple functions are noninte-
grable [see (8.20) J. Our method of extension is a very general one. It
consists of starting with an averaging process on a space of continuous
functions, extending it to nonnegative lower semicontinuous func-
tions, extending a second time to all nonnegative, extended real-valued
functions, and finally realizing our process as an integral [the Lebesgue
integral] on a large class of functions. This procedure is often called the
Daniell approach to integration theory, after the British mathematician
P. J. DANIELL (1889-1946).
(9.1) Definition. Let X be a nonvoid locally compact Hausdorff
space. A complex-valued function I defined on <roo (X) is called a non-
negative linear functional [sometimes Radon measure] if for all f, g E<roo (X)
and oc EK we have
(i) I (I + g) = I (I) + I (g) [additive],
°
(ii) I(ocf) = ocI(f) [homogeneous],
(iii) I (I) ;?; if f E<rto (X) [nonnegative J.
(9.2) Examples. (a) Let X = R and I = Sa. for any real-valued non-
decreasing oc on R.
00 00
(b) LetX=R2 and let I (I) = J J f(x,y)dxdyforallfE<roo(R2).

-00 -co
[This is the 2-dimensional Riemann integral, familiar to the reader from
elementary analysis. We will discuss such "multiple" integrals in detail
in § 21.]
(c) Let X be any nonvoid locally compact Hausdorff space and let
a be a fixed point of X. The functional Ea defined by Ea (I) = f (a) is
plainly a nonnegative linear functional on <roo(X). It is called an evalua-
tion functional. In addition, Ea satisfies the identity
Ea (f g) = Ea (I) Ea (g)
for all f, g E<roo (X). Such linear functionals are called multiplicative.
§ 9. Extending certain functionals 115
They are of great importance in the theory of Banach algebras, and we

shall return to them later [see (20.52) and (21.65) inlra].
(d) Let X = R and let w be any nonnegative real-valued continuous
function on R [perhaps unbounded]. Then I{I) = S{lw) [S being the
Riemann integralJ defines a nonnegative linear functional I on <roo(R).
Throughout the rest of § 9, except where the contrary is specified,
X will denote a fixed but arbitrary nonvoid locally compact Hausdorff
space, and I will denote a fixed but arbitrary nonnegative linear func-
tionalon <roo (X). We will write <roo instead of <roo (X), for brevity's sake.
(9.3) Theorem.
(i) I I I E<r~o, then I (I) is a real number.
(ii) II I ~ g in <rlio, then I (I) ~ I (g).
Proof. For IE <rlio, write r = max{/, O} and r= -min{/, O}.
It is clear that I = r - r, r ; :; 0, and r ; :;
0 Thus I (I) = IW) - I(n,
so I (I) is the difference of two nonnegative real numbers. This proves (i).
If I, g E<r6o and I ~ g, then g - I ;;:;; 0 and so I (g) - I (I) = I (g - I) ;;:;; o.
This proves (ii). 0
(9.4) Theorem. Let I E<roo. Then II (I) I ~ I (lfI)·
Proof. Write I{I) = e exp(iO) where 0 ~ e < 00 and - n < 0 ~ n.
Let exp(-iO)/=gl+ig2 where gl>g2E<rlio. Then e=exp(-iO)I{I)
= I (exp (-iO)/) = I(gl + ig2) = I (gl) + iI(g2)' But e is real and I (gl)
and I (g2) are real. Therefore I (g2) = 0 and e = I (gl)' Clearly
gl ~ lexp(-iO)/1 = III· Thus 11(1)1 = e = I(gl) ~ 1(111). 0
(9.5) Theorem. Let A be any compact subset 01 X. Then there exists
a real number {J [depending only on A] such that II {Ill ~ {JII/II" lor all
I E<roo such that I(A') C {O}.
Proof. According to (6.79) there is an open set U:J A such-that U-
is compact. By (6.80) there is a continuous function w from X into [0, 1J
such that w(x) = 1 for all x EA and w(x) = 0 for all x EX n U'. Let
t E<roo be such that 1=0 on A'. Then I(x) = I (x) w (x) for all x EX.
It follows that III ~ 1I/1I"w and hence II (I) I ~ I (III) ~ I (1I/1I"w) = 11t1l"I (w).
Set {J = I(w). 0
Our linear functional I is given to us a priori as only finitely additive:
1(/ + g) = 1(1) + I(g). For many purposes, it is useful to have countable
additivity: IC~ In) = n~ I{ln)· This equality is seldom true for all
convergent sequences of functions on X, and to prove it for some se-
quences of functions, we must extend the domain of definition of I.
In (9.6)-(9.18), we carry out this extension and establish properties
of the extended functional that lead up to countable additivity.
(9.6) Theorem. Let !5) be a nonvoid subset 01 <rto such that lor all
f1' f. E!5) there exists f3 E!5) such that 13 ~ min{h, M [we say that !5) is
8*
directed downward). Suppose also that inf{/(x) : IE!S>} = °

lor every x EX.
Then we have
(i) inf{I (I) : I E!S>} = 0,
°°
and lor every e > there is an I. E!S> such that 11/.11" < e.
Proof. Let e> be fixed. Choose 10 E!S> and define Ao = {x EX:
lo(x) > O}-. Since 10 E <roo' Ao is compact. For each I E!S>, let
At = {x EAo: I(x) ~ e}. Then {At: I E!S>} is a family of closed subsets
of the compact space Ao and n {At: IE!S>} = 0. It follows (6.34)
that this family does not have the finite intersection property, i. e.,
there is a finite subset {/I>"" In} of !S> such that At. nAt, n ...
n Atn = 0. Since !S> is directed downward, there is a function I. E!S>
such that I. ~ min{/o, II> ... , In}; plainly 11/.11" < e.
To prove (i), apply (9.5) to find a real number {J [depending only
on Ao and not on eJ such that 11(1)1 ~ {JII/II" for all I E<roo that vanish
on A~. It is clear that I. = °
on A~. Thus
Since e is arbitrary, (i) is established. 0
°
~ I (I.) ~ {JII/.II .. < {J e.
(9.7) Remarks. The reader will note that compactness is used in an

essential way in the proof of Theorem (9.6). This is no accident. Our
function space <roo was chosen so that compactness could be applied.
Many other plausible choices of function spaces and positive linear
functionals fail to produce countably additive extensions. Theorem (9.6)
is the key to obtaining countable additivity of our extended functional,
and so countable additivity depends in the end upon compactness.
Our first extension of I is to the class !m+ of all lower semicontinuous
functions on X with range contained in [0,00]. [See (7.20)-(7.22).]
(9.8) Definition. For each g E!m+, define
l(g) = sup {I (f) : I E<rto, I ~ g}.
Notice that I (g) may be 00. For example, if X = R and 1= 5,

then the function 1 is in !m+ and 5 (1) = 00. For arbitrary X and I =l= 0,
the function 00 is in !m+, and 1(00) = 00.
(9.9) Exercise. (a) Make a careful computation of 5 (eu)' where U
is any open subset of Rand 5 is the Riemann integral. [Hint. Use
(6.59).] The number S(eu) is called the Lebesgue measure 01 U. It will
be discussed in detail below.
(b) For the evaluation functional Ea (9.2.c) and for g E!m+, compute
Ea(g)·
(9.10) Theorem.
(i) II IE <rto, then I (I) = I (I).
(ii) II gl' g2 E!m+ and & ~ g2' then I (gl) ~ I (g2)'
(iii) II g E!m+ and IX ~ 0, then I (lXg) = IXI (g).
These assertions are all obvious upon a moment's reflection.

The next theorem is a counterpart of (9.6).
(9.11) Theorem. Let ~ be a nonvoid subset of m1+ such that for all
gl> g2 E~ there is a ga E~ such that ga ;:;;:; max{gl> g2} [we say that ~ is
directed upward]. Then
(i) J(sup~) = sup{J(g); g E~}.
Proof. Let go = sup{g; g E~}, i.e., go(x) = sup{g(x) ; g E~} for all
x EX. Then (7.22.iii) shows that go E m1+. There are two cases to consider.
Case I; {go} U ~ c <rto. Then the family {go - g ; g E~} satisfies
the hypotheses of (9.6), for if ga ;:;;:; max{gl> g2}' then
go - g3 ;;;;; min {go - gl> go - g2}'
and
inf{go(x) - g(x) ; g E~} = go (x) - sup{g(x) ; g E~} = go (x) - go (x) = 0
for every x EX. Then (9.6) implies that
0= inf{I (go-g) ;gE~} = inf{I (go) -I (g) ;gE~} = I (go) -sup{I (g) ;gE~}.
But I and J agree on <rto, and so J (go) = sup {J (g) ; g E~}.
Case II; {go} U ~ C m1+, i. e., the general case. Since g ;;;;; go for all
g E~, we have sup{J (g) ; g E~} ;;;;; J (go)' To prove the reversed inequality,
we introduce the family Ci = {I E<rto; I;;;;; g for some g E~}. Making
use of (7.22.v), we have
go = sup{g; g E~} = sup {sup {I E<rto; I ;;;;; g} ; g E~} = sup Ci .
Let fJ be any real number less than J (go)' By the definition of J,
there exists a cp E <rto such that cp;;;;; go and I(cp) > fJ. We have
cp = min {cp, go} = min{ cp, supCi{ = sup {min {cp, I}; I ECi} [the reader
should check this sleight of hand carefully J. The result for Case I now
applies with the family {min{cp, I}; I ECi} taking the role of ~ and cp
the role of go' Thus
fJ < I (cp) = sup {I (min {cp, f)} ; I ECi} ;;;;; sup {I (I) ; I ECi}
= sup{J(g); g E~}.
Since fJ is arbitrary, we have proved that
J(go) ;;;;; sup{J(g) ; g E~}. 0
(9.12) Corollary. II gl ;;;;; g2 ;;;;; •.. ;;;;; gn ;;;;; .. " where gn Em1+ for
n = 1,2, ... , then
The proof is immediate.

(9.13) Corollary. For gvg2 E!m+, we have l(gl+g2) =l(gl) +l(gs)·

Proof. We have gl + g2 = sup{/1 + 12: Ii E<rto and Ii ~ gi' j = 1, 2},
and therefore
I (gl + g2) = sup {I (II) + 1(12) : Ii E <rto, Ii ~ gil
= sup {I (II) : II E<rto, II ~ gl} + sup {I (/2) : 12 E<rto,/g ~ g2}
= l(gl) + l(g2)' 0
(9.14) Corollary. For any nonvoid <S) C !m+, we have
1(I: g)= I: l(g).1
gE~ gE~
Proof. For finite <S), (9.13) and an easy induction yield the result.
For infinite <S), apply (9.13) and (9.11). 0
We now make our second, and last, extension of I.
(9.15) Definition. Let 'J+ be the set of all functions defined on X
with values in [0, ooJ. For h E'J+, let
l(h) = inf{l (g) : g E!m+, g ~ h}.
(9.16) Theorem.
(i) For g E!m+, we have I (g). I (g) =
(ii) I I hi ~ h2' hi E'J+, then we have I (hi) ~ I (h2)'
(iii) For h E'J+ and 0 ~ IX < 00, we have l(ock) = IXI(h).
(iv) For hi' h2 E'J+, we have l(h l + h2) ~ l(hl ) + l(h2).
The proof is left to the reader.
Our next theorem is a generalization of B. LEVI'S monotone conver-
gence theorem.
(9.17) Theorem [Generalized B. LEVI theorem). II hn E'J+ lor
n = 1, 2, 3, ... , and hi ~ h2 ~ ... ~ hn ~ " ' , then l( lim hn )
....... 00
= lim l(hn ).
"--->00
Proof. Let h = lim h.. ; we obviously have

"--->00
l(hl ) ~ l(h2) ~ ... ~ lim l(h..) ~ l(h).

n->oo
We must now prove that lim I (h..) ~ I (h); clearly we may suppose that
,,--->00
lim 1 (hn ) < 00. Choose e > O. For each positive integer n, select g.. E!m+
;n < l(h..). We wish to apply (9.12). To
....... 00
such that g.. ~ h.. and l(g..) -

do this, we must replace the functions g.. by a nondecreasing sequence.
Thus define g~ = max{gv g2' ... , g.. } for n = 1,2, ... ; we have g~ E!m+
1 The function 1: g is defined by ( 1: g) (x) = sup {gl (x) + g. (x) + ... +
gE~ gE!{)
gn (x) : {gl' g., ... , gn} C ~}; similarly, we have 1: 1(g) = sup {l (gl) + 1 (g.) +
... + l(gn): {gl' g., ... , gn} C ~}. gE!{)
for n = 1, 2, .... The reader can easily check the identity

g~ +1 + min {g~, gn +l} = g~ + gn +I-
All of the functions above are in ml+, and so we have
l(g~d + l(min{g~, gn+l}) = l(g~) + l(gn+l)'
From the inequalities gn+l ~ hn+l ~ hn and g~ = max{gl> g2' ... , gn}
~ max{hl' h2, ... , hn}=hn' it follows that min{g~, gn+l} ~ hn. Hence
-l(min{g~,gn+l}) ~ -I(hn), and we have
I (g~+l) = I (g~) + I (gn+l) - I (min{g~, gn+l})

~ I (g~) + I (gn+l) -1 (h n )
-, e
< I (g,,) + I- (hn+l) + 2 n +1 -
=
I (hn) .
Summing this inequality over n = 1,2, ... , p, we have
p p p p p
}; l(g~+l) < }; l(g~) +}; 1(hn+l) -}; 1(hn) + ~ };2-n,
,,=1 ,,=1 n=1 ,,=1 n=1
and so
l(gp+l) < l(g;) + I(hp+l) - I(h1) +;
= I(hp+l) + (l(gl) - I(h1)) + ~
< l(hp +1) + e .
Thus the inequality I (gP+l) < I (hp+l) + e is valid for p = 1,2,3, ....
The same inequality obtains with p = 0, and so we have
l(g~)<l(hn)+e (1)
for n = 1, 2, 3, . . .. The sequence (g~) is nondecreasing and g~ E ml+
for n = 1, 2, 3, ... ; (9.12) thus implies that
lim I (g~) = I ( lim g~) = I (supg~) .
n---+ooo n-+oo
Since g~ ~ gn ~ h.. for n = 1,2,3, ... , we have supg~ ~ suphn = h,

and so I (sup g~) ~ I(h). Using (1), we now find that lim I(hn) ~
n-+oo
lim I (g~) - e ~ I (h) - e. The inequality lim I (hn) ~ I (h) follows, since
n-+oo n-+oo
e is arbitrary. 0 1
(9.18) Corollary. Let (hn)~= 1 be any sequence of functions in iY+;
then I (,,~ hn) ~,,~ I(hn).

1 Theorem (9.17) is of course (9.12) with I replaced by 1. It is important to note
that (9.13) and (9.11) fail in general if I is replaced by i: see (10041). The truth of
(9.17) is by itself fairly remarkable.
Proof. Write "Pn = h,. + h2 + ... + hn; then
I ( £ hn) = 1 ( lim "Pn) = lim I ("Pn) = lim I ( £ h,.)

,,=1 n~oo "'-'00 n~oo k=l
.. 00
~ lim }; I(h,.) = }; I(hn). 0

k=1
"-->-00 ..=1
We now define the measure [with respect to I] of a subset of X.
(9.19) Definition. For A C X let t(A) = I(~.A)' We call t(A) the
[outer] measure of A. The function t, defined on &,(X), is known as the
[outer] measure induced by I. In case X = R and 1= S., for some real-
valued nondecreasing function (X on R, we write A.,(A) = S"(~.A) and
call A., the Lebesgue-Stieltjes [outer] measure on R induced by (x. If X = R
and 1= S [the Riemann integral], then we write A(A) = S(~.A) and
call ALebesgue [outer] measure on R. For arbitrary X and I = Ea (9.2.c),
we write Ba(A) = Ea(~.A) and call Ba the unit point mass [or Dirac measure]
concentrated at a.
(9.20) Exercise. For arbitrary X, a EX, and A C X, prove that
Ba(A) = 1 if a EA and Ba(A) = 0 if a ~A. That is, Ba(A) = ~.A(a).
We propose now to investigate properties of the measure t and to
construct an integral [which is the classical Lebesgue integral for t = A]
from it. This integral turns out to be the functional I whenever the integral
exists. The program is somewhat long; it will be completed in Theorem
(12.35) infra. We begin by pointing out some properties of the set func-
tion t.
(9.21) Theorem. The set function t has the following properties:
(i) 0 ~ t(A) ~ 00 for all A eX;
(ii) t(A) ~ t(B) if A c B c X;
(iii) t(0) = 0;
(iv) if (An)':=1 isanysequenceofsubsetsofX,thent(,.QI An) ~ ..~ t(An)
[countably subadditive].
Proof. Assertions (i)-(iii) are trivial consequences of the definition
of t.To prove (iv), we write
00
here we have used the inequality ~ UA ~ }; ~A .. and (9.18). 0

"-1" .. =1
(9.22) Theorem. Let {Uo: () E8} be any pairwise disjoint family of
open subsets of X. Then t( U Uo) = }; t(Uo).
OE@ OE@
Proof. IfU= OE@

U Uo, it is clear that ~u~Ecm+foreachOEeandthat
v
~u= 1: ~uo' Applying (9.14), we have t(U) = l(~u) = 1: l(~uo)

OE@ OE@
= 1: t(Uo). 0
OE@
(9.23) Corollary. II (Jan, bn[):=l is any pairwise disjoint sequence 01
open intervals 01 R, then A CQl Jan, bn[) = n~ (bn - an)·
Proof. This is an immediate consequence of (9.22) and the fact
[which is a trivial exercise] that A(]a, b[) = b - a. [Note that a or b
may be infinite.] See also Exercise (9.9.a). 0
(9.24) Theorem. For every A C X we have
t(A) = inf{t(U) : U is open, A C U} .
Proof. If t(A) = 00, then the result is trivial since A C X, X is open,
e:
and 00 = t (A) ~ t (X) = 00. Thus suppose that t (A) < 00 and let e > 0
be arbitrary. Choose a real number ~ such that 0 < ~ < 1 + t(A) < 1.
Next select gEcm+ such that g~ ~A and l(g)-~<I(~A)=t(A).
Let U = {x EX: g(x) > 1 - ~}. Clearly U is open and A C U. For
1 1
x E U, we have 1-<5 g(x) > 1, and so 1-<5 g;;?; ~u. Thus
- -( 1 ) 1 - 1
t(U)=I(~u)~I 1-<5 g = 1-<5 I (g) < 1_<5(t(A)+~)<t(A)+e. 0
(9.25) Remark. It follows from (9.24), (6.59), and (9.23) that for
ACRwehaVeA(A)=inf{1: (bn-an): nQl ]an,bn[::::>A and {Jan,bn[}:=l
are pairwise diSjOint}. This is how A was originally defined by LEBESGUE

himself, in 1902. LEBESGUE'S fundamental idea was to consider countable
coverings of A by open intervals. Earlier attempts at defining a suitable
notion of measure for subsets of R were similar to LEBESGUE'S, but in
each case only finite coverings of the set in question were considered.
For example, the content 01 A was defined by C. JORDAN to be the number
inf {1 (bn - an) : A C n~l Jan, bn[, P= 1, 2, .. -}- The Jordan content
[Inhalt in German] is still studied by some mathematicians 1, but it
has proved to be unsatisfactory for the purposes of modern analysis.
(9.26) Theorem. Let U be any open set in X; then we have t (U)
= sup{t(F) : F is compact and F C U} = sup{t(V) : V is open in X, V-
is compact, and V- C U}.
1 See for example K. MAYRHOFER, Inhale und Map, Springer-Verlag, Wien, 1952.
Proof. Take any real number fJ such that fJ < £(U). Since U is open,
we have £(U) = l(~u) = sup {I (I) : I E~rio, I ~ ~u}. Thus we can choose
I E~rio such that fJ a}. It is clear
that lim ~F" (x) = lim ~w (x) = ~w (x) for each x EX, and it is clear
n-+ 00 n-+ 00 n
too that the sequences are nondecreasing. Applying (9.17), we have
fJ < 1(1) ~ l(~w) = lim I(~F.. ) = lim l(~w) = lim l(Fn) = lim £ (Tv..) •
n~ 00 K-+ 00 n n-+ 00 11-+ 00
The theorem follows from these inequalities. 0

(9.27) Theorem. II A C X and A- is compact, then 0 ~ £(A) < 00.
Proof. According to (6.79), there is an open set U such that U-
is compact and A - C U. Apply (6.80) to find a continuous I: X -+ [0, 1]
such that I(x) = 1 for all x EA- and I(x) = 0 for all x EU'. Then
~A ~ I E~rio, and so £(A) = 1(~A) ~ 1(1) < 00. 0
(9.28) Theorem. There exists a unique set E C X having the lollowing
properties:
(i) E is closed in X;
(ii) l (E n U) > 0 il E n U =F 0 and U is open in X;
(iii) £(X n E') = o.
The set E is called the support [or carrier, or spectrum] of £.
Proof. Let OIl = {U: U is open in X, t{U) = O} and let V = U OIl,
E = V'. Since ~v ~ I: ~u and ~u E2n+ for each U EOIl, (9.14) yields
UEo/i
l(E') = £(V) = l(~v) ~ 1(I: ~u) = I: l(~u) = I: £(U) = O.

UEo/i UEo/i UEo/i
Thus (i) and (iii) are established. To prove (ii), let W be any open subset
of X such that En W =F 0. Then W ~ OIl and V n W EOlI. Thus
0< £(W) ~ l(E n W) + l(V n W) = £(E n W) .
Thus E satisfies (i), (ii), and (iii).
To prove that E is unique, assume that both E1 and E2 satisfy (i),
(ii) , and (iii) and E1 =F E 2. At least one of E; n E2 and E1 n E~ is non-
void; say E; n E2 =F 0. Since E; is open, (ii) implies that l (E; n E 2) > O.
But E; n E2 C E; and so 0 < £(E; n E 2) ~ £(E;). According to (iii),
we have £(ED = O. This is a contradiction. 0
(9.29) Definition. A subset A of X for which £(A) = 0 is called an
£-null set. If £(B n F) = 0 for every compact set Fe X, then B is called
a locally £-null set. A property which holds for all x EX except for those x
in some £-null set is said to hold £-almost everywhere [abbreviated £-a.e.].
If a property holds for all x E X except for those x in some locally
£-null set, then the property is said to hold locally £-almost everywhere
[locally t-a.e.]. A complex or extended real-valued function f on X

such that f (x) = 0 l-a. e. [locally t-a. e.] is called an t-null function
[locally l-null function]. Where no confusion can result, we will drop
°
the prefix "t-".
(9.30) Theore~. For h Etr+, we have I(h) = if and only if h is an
t-null function. If I(h) < 00, then h is finite t-a.e.
Proof. Let A = {x EX: h (x) > a}. The functions nh, n = 1, 2, ... ,
are all in tr+, and it is obvious that lim nh ~ gAo Thus if l(h) = 0,
n->-oo
(9.17) shows that t(A)=i(gA);:;;; i(lim nh)= lim l(nh) = lim nl(h) = O.
n-+ 00 fl.-+- co "-+ 00
If h is an l-null function, then teA) = 0; using the inequality h ;:;;; lim n;A'
"_00
we have i(h) ;:;;; lim i(ngA) = lim nl(gA) = lim nt(A) = O.
n~ 00 n---7OO K-+ 00
Next suppose that l(h) < 00, and let B = {x EX: hex) = oo}. For
all e> 0 we have gB ~ eh, and so t(B) = l(gB) ~ l(eh) = el(h). Since
i(h) is finite, we infer that t(B) = O. 0
(9.31) Corollary. Let (h,,)::'=1 be a sequence of functions in tr+ and
suppose that lim i(h,,) = O. Then there is a subsequence (h,,~)k°=1 such that
"_00
00
); h".(x) < 00 l-a.e. on X and in particular, lim h".(x) = 0 t-a.e.
k=1 k_oo
00
Proof. We first select a subsequence (h"k) of (h,,) such that}; 1(h",,) < 00.
k=1
Using (9.18), we see that I ( J'I h"l:) < 00, and it follows from (9.30)
00
that }; h"" (x) < 00 t-a. e. 0

k=1
The next theorem is a technicality, but a very useful one for later
purposes.
(9.32) Theorem. Let U be an open subset of X. Then
t(T) = t(T n U) + t(T n U')
for every set TeX.
Proof. Let TeX. It is an immediate consequence of (9.21) that
t (T) ~ t (T n U) + t (T n U'). The reversed inequality is obvious if
leT) = 00. Thus suppose that leT) < 00, and let e> 0 be arbitrary. By
(9.24), there is an open set V::::> T such that t(V) < leT) + : . Use (9.24)
again to choose an open set H ::::> V n u' such that l (H) < l (V n U') + : .
Applying (9.26), choose an open set W such that W- C V n U and
l(W) + : > l(V n U). Let Wa = V n H n (W -)'; then Wand Wo are
disjoint open sets. Since V n u' is a subset of each of the sets V, H,
and (W)', it follows that V n u' cWo C H, and so

o ~ t (Wo) - t (V n U') ~ t (H) - t (V n U') < 4B .
Therefore
\t(W) + £(Wo) - (£(V n U) + t(V n U'))\ ~ \£(Wo) - t(V n U')\
+ \t(W) - n U)\ < : + : = ; .
t (V
Combining this with the fact that W U Wo c V and using (9.22), we have
B B B
t(T) + e> t(V) +"2~ t(W U Wo) +"2= t(W) + t(Wo) +"2
> t(V n U) + t(V n U') ~ t(T n U) + t(T n U') .
Since e is arbitrary, we conclude that
£(T) ~ t(T n U) + t(T n U'). 0
(9.33) Exercise. Prove that:
(a) if a < b in gil:, then A(]a, b[) = b - a; and
(b) if a< bin R, then A(]a, b[) = A([a, b[) = A(]a, b])
=A([a,b])=b-a.
(9.34) Exercise. Let A be a countable subset of R. Prove that
A(A) = O.
(9.35) Exercise. Let P be the Cantor ternary set. Prove that A(P) = O.
(9.36) Exercise. Construct a nowhere dense perfect subset F of
[0, 1] such that A(F) = 0/:, where 0/: is any real number, 0 ~ 0/: < 1.
(9.37) Exercise. Let F be a nonvoid perfect subset of R. Prove that
F contains a nonvoid perfect subset of Lebesgue measure zero.
(9.38) Exercise. Let (an ):'=1 be a sequence of positive real numbers
00
such that.E an = 1. Prove that there exists a pairwise disjoint sequence

.. =1
00
(110 ):'=1 of open intervals such that U In C [0, 1], A(In) = an for each
10=1
n EN, and [0, 1] n CQl
In)' is nowhere dense and perfect in R. [See
(8.26).]
(9.39) Exercise [FATOU'S lemma]. Let X and I again be arbitrary.
Suppose that (hn ) is a sequence of functions in tr+. Prove that
I( lim hn ) ~ lim
n~oo ft.---ioOO
1(hn ) .
Also find a sequence (hn ) C <rto for which the strict inequality holds,
where X = R and I = S, the Riemann integral.
(9.40) Exercise. Let X be a locally compact Hausdorff space. Let X*
be a nonvoid closed subset of X [with the relative topology]. Let 1*
§ 10. Measures and measurable sets 125
be a nonnegative linear functional on croo (X*) and let t* be the set

function defined on the subsets of X* constructed as in (9.19) from the
functional 1*.
(a) For a function f on X, let f* be f with its domain restricted to X*.
Prove that if f E croo(X), then f* E <roo (X*).
(b) Let g E croo(X*). Prove that there is a function f E croo(X) such
that g = f*. [Use TIETZE'S extension theorem (7.40).]
(c) For f Ecroo(X), let 1(1) = 1*(1*). Prove that 1 is a nonnegative
linear functional on croo(X).
(d) Let t be the set function obtained from 1 as in (9.19). Prove that
t(X*') = 0 and t(A) = teA n X*) = t*(A n X*) for every A eX.
(9.41) Exercise. Let X be the product space Rd x R, where the first
factor is the real line with the discrete topology and the second factor
is the real line with its usual topology. For x, a, b ER with a < b, let
U(x, a, b) = {(x,y) EX:a<y<b}={x} xJa,b[.
(a) Show that {U (x, a, b) : x, a, b ER and a < b} is a base for the
product topology on X (6.41).
(b) Prove that X with this topology is a locally compact Hausdorff
space.
For any function f defined on X and any x ER, let f[x] be the function
defined on R by the rule /rx](Y) = t(x,y).
(c) Prove that if f E croo(X), then lex] is identically zero except for
finitely many x ER.
Define 1 on croo(X) by the rule 1(1) =}; SU[x)), where S is the or-
xER
dinary Riemann integral.
(d) Prove that 1 is a nonnegative linear functional on croo(X).
Let £ be the measure obtained from 1 as in (9.19).
(e) Prove that the set A = {(x, 0) : x E R} is locally t-null but is not
t-null.
(9.42) Exercise. Prove that if X is a countable union of compact
sets [such a space is said to be a-compact], then every locally t-null
set is t-null.
§ 10. Measures and measurable sets

(10.1) Introduction. Section 9 was devoted to the construction of the
functional I and the set function t, defined on [ljJ (X). Our ultimate aim
[to be realized in § 12] is to find a reasonably large class of functions
on X for which the equality
(i)
holds. [We have already proved (i) for functions in ml+.] The avenue
we choose toward this goal is through abstract measures and integrals
defined in terms of these abstract measures. The problem of finding
a largest family of functions on X for which (i) holds is unsolved and ap-
parently very difficult. Our approach to the problem is not the only
one possible, but it has the advantages of simplicity and also of introduc-
ing abstract integrals, which every reader should know about anyway.
The present section is mainly concerned with properties of the set
function , and in particular with its good behavior on a certain well-
defined family of subsets of X. The properties of £ that we need in defining
this family are set down in Theorem (9.21). It turns out that set functions
enjoying properties (9.2l.i)-(9.2l.iv) can be studied in abstracto, with
no reference to a topological space or to positive functionals. We make
a formal definition, as follows.
(10.2) Definition. Let X be a set [no topology]. A function /-' de-
fined on 9'(X) is called a [CARATHJ£ODORyl] outer measure if the follow-
ing relations hold:
(i) 0 ~ /-,(A) ~ 00 for all A ex;
(ii) /-,(0) = 0;
(iii) /-,(A) ~ /-,(B) if A C Be X;
(iv) /-' e.gl Aft) ~,,~ /-' (Aft) for all sequences (Aft):'=1 of subsets of X.
Outer measures are not by themselves of great use. Far more important
for integration theory are measures, which we next define.
(10.3) Definition. Let X be a set and .91 an algebra of subsets of X.
A set function /-' defined only on .91 is called a finitely additive measure if:
(i) 0 ~ /-,{A) ~ 00 for all A E.Iif;
(ii) /-,(0) = 0;
(iii) /-,(A U B) = /-,(A) + /-,(B) if A, B Ed and A n B = 0 2 •
A finitely additive measure /-' such that
(iv) /-' e.Ql Aft) =,,~ /-' (Aft) for all pairwise disjoint sequences (Aft):'_1
00
such that Aft Ed and U Aft Ed is called a countably additive measure

11=1
or simply a measure. If.91 is a a-algebra of subsets of X and /-' is a [count-
ably additive] measure on .91, then the triple (X, .9I,/-,) is called a
measure space. Let (X,d, /-,) be a measure space. If /-' (X) < 00, then
1 CONSTANTIN CARATHEODORY (1873-1950), the inventor of outer measures,
was a distinguished German mathematician [of Greek descent], who made many vital
contributions to modem analysis.
I Finitely additive measures as such are only of peripheral interest for this
text. We include their definition largely for the sake of completeness, but also with
an eye to certain applications in § 20.
(X, .JiI, f-t) is called a finite measure space and f-t a finite measure. If X
is the union of a countable family of sets in.Jil each having finite meas-
ure, then (X,.JiI, f-t) and f-t are called (I-finite. If f-t = 0 or f-t assumes only
the values 0 and 00, then (X, .JiI, f-t) and f-t are called degenerate.
Outer measures offer a convenient method of obtaining measures,
if we suitably restrict the family of subsets on which the outer measure
is defined. Before entering on the technicalities of this construction,
we give a few examples.
(10.4) Examples. (a) Let X be any set. For A C X, let
,,(A) =
{ .A if
00 if
A is finite ,
A is infinite.
Then " is a measure and also an outer measure on .9' (X). [It is easy to
see that a measure on .9' (X) is also an outer measure.] The measure "
is usually called counting measure.
(b) Let X be any set. For A C X, let
f-t(A) =
{o00 ifA=.0,
if A =F .0 .
Then f-t is a [plainly degenerate] measure and also outer measure on .9' (X).
(c) The set function identically zero is a degenerate measure on
.9'(X), for any set X.
(d) The most important outer measures for our purposes are the
set functions t constructed as in § 9 for all subsets of a locally compact
Hausdorff space X. See Theorem (9.21) for the proof that t is an outer
measure. The reader should be aware, however, that not all measures
used in analysis are derived from set functions t.
We now begin our construction of measures.
(10.5) Definition [CARATHEODORY]. Let X be a set [no topology]
and f-t an outer measure on .9' (X). A subset A of X is said to be f-t-measur-
able if
f-t(T) = f-t(T n A) + f-t(T n A')
for all TeX. Let ~ denote the family of all f-t-measurable subsets of
X and';y': the family of all subsets A of X such that f-t (A) = O.
(10.6) Remarks. (a) In view of (1O.2.iv) and (1O.2.ii), a subset A
of X is p-measurable provided that f-t(T) ~ f-t(T n A) + f-t(T n A')
for all Tc X.
(b) Definition (10.5) has a somewhat artificial air. It singles out the
subsets A of X for which A splits aU subsets T of X into two pieces on
which f-t adds. How CARATHEODORY came to think of this definition seems
mysterious, since it is not in the least intuitive. CARATHEODORY'S def-
inition has many useful implications. It gives us a (I-algebra, although
not necessarily the largest possible a-algebra, on which p, is a countably

additive measure. [In (10.40), we point out conditions under which
~ is the largest a-algebra on which p, is count ably additive.]
(c) If X is a locally compact Hausdorff space and, is an outer measure
as in § 9, then the family.A, of ,-measurable sets contains all open sets.
This fact, proved in (9.32), has important consequences as we shall see.
(d) Let A be Lebesgue outer measure on R. The family .AJ. of A-
measurable sets is often called the family of Lebesgue measurable sets.
We will use this phrase when convenient.
We proceed to develop the properties of p,-measurable sets. Through-
out (10.7)-(10.11), X is an arbitrary set and p, is an arbitrary outer
measure on &(X).
(10.7) Theorem. Every subset A of X such that p, (A) = 0 is p,-measur-
able, and p, (T) = P, (T n A') for all TeX.
Proof. Let TbeanysubsetofX. Thenp,(T n A) = 0, since TnA cA.
Also, we have p,(T) ~ p,(T n A) + p,(T n A') = p,(T n A') ~ p,(T).
It follows that p, (T) = P, (T n A) + p, (T n A') = p, (T n A'); the first
equality shows that A is measurable. 0
(10.8) Theorem. If A is p,-measurable, then A' is also.
Proof. Trivial.
(10.9) Theorem. Let (A n):'=l be a sequence of pairwise disjoint p,-
measurable subsets of X. Then
(i) p,(T) =n~ p,(T nAn) + p, (T n CQ! AS) for all TeX.
00
Proof. By countable subadditivity, we have p,(T) ~ 1: p,(T nAn)

n=!
+ p,( Tn CQ!An)'}. If p,(T) = 00, (i) follows immediately. Hence we

may suppose that p, (T) < 00. We first prove
p, (T) = J: P, (T nAn) + p, (T n C~!An)'} for all PEN. (1)
We prove this by induction on p. For p = 1, (1) becomes p, (T)

= p,(T n AI) + p,(T n A;). This is true for all T C X since Al is p,-
measurable. Suppose that (1) is true for a positive integer p and all
subsets T of X. Since Ap+l is p,-measurable, we have
p,(T) = p,(T n Ap+l) + p,(T n Ap+l)
= p,(T nAp+l)+ f p,(TnAp+l nAn) + p, (T nAp
n=!
+1 n C~!An)'} .
(2)
[We apply the inductive hypothesis to the set Tn Ap+l.] Since
An C Ap+1 for n =!= p + 1, (2) can be written as
n Ap+I) +n~ p,(T nAn) + p,( Tn Ap+I n C.~lAnn

l:
p,(T} = p,(T
= p,(T nAn} + p, (TnC~: An)'),

and this is (1) for p + 1.
The sequence of numbers (p,( T n C.~lAn)') ) ~=1 is a nonincreasing
sequence bounded below by the number p, (T n CQ 1An)'). It thus has

a limit, which is greater than or equal to p, (Tn CQ AS) .Taking limits
I
in the equality (1), we obtain
p,(T) = p~~ tl p,(T nAn) + p~m(X'# (T n C~IAn)')

~ nE p,(T nAn) + p, (T n CQIAn)') .
Since the reversed inequality has already been established, this completes
the proof. 0
(10.10) Theorem. II A and Bare p,-measurable, then A n B' is p,-
measurable.
Proof. It suffices to prove that if E cAn B' and F C (A n
B ' l',
then p, (E U F) = p,(E) + P, (F). Since F = (F n B) U (F n B') and B
is p,-measurable, we have
p, (E) + p, (F) = p, (E) + p,«(F n B) U (F n B'))
= p, (E) + P, (F n B) + p, (F n B') .
Now since E C A, F n B' C A', and A is p,-measurable, we have
p,(E) + p,(F n B') + p,(F n B) = p,(E U (F n B')) + p,(F n B) .
Again E U (F n B') C nBc B so that
B' and F
p,(E U (F n B')) + p,(F n B) = p,(E U (F n B') U (F n B)) = p,(E U F).
Combining these equalities, we have p,(E) + p,(F) = p,(E U F). 0
(10.11) Theorem. The lamily ~ 01 p,-measurable sets is a a-algebra
01 subsets 01 X, and p, is a countably additive measure on the a-algebra ~.
Proof. Let (An}:'=1 be a sequence of p,-measurable subsets of X.
00
Then
n=l
U An = Al U (Aan A;) U (A3 n A~ n A;) U ... U (An n A~_l
n ... n AD U .. '. By (10.10), each set of the form Bn = (An n A~_l
n ... n AD is p,-measurable. Furthermore, the sets Bn are pairwise
disjoint.
Let TeX. By (1O.9.i) and countable subadditivity, we have
p(T) =n~ p(T n Bn) + P(T n CQIBn)'} ~ p (T n CQIBn))

+ p (T n CQIBn)'} .
By countable subadditivity, we have
p(T) ~ P (T n CQIBn)) + P (T n CQ1Bn)'} ,

and so
p(T) = P (T n CQIBn)) + P (T n CQIBn)') .
00 00 00
This implies that ..~1 Bn is p-measurable. Thus .. ~1 An = "~I Bn is p-measur-

able. This fact and (10.8) imply that the family of p-measurable sets is
a a-algebra (1.13).
00
Upon setting T =
k=1 U Bli in (1O.9.i), we obtain
00 00
p(Bn) + p(0) =
= ~
..=1 . =1 p(Bn) .
~
Thus P is count ably additive on the a-algebra of all p-measurable subsets

of X. 0
Having proved in (10.11) that nontrivial measure spaces exist,
we digress to prove some useful facts about arbitrary measure spaces.
(10.12) Theorem. Let (X, d, p) be a measure space. II A, BEd
and A C B, then p(A) ~ p(B).
Proof. By (lO.3.iii) we have
p(B) = p(A) + p(B n A') ,
and by (lO.3.i) p(B n A') ~ o. 0
(10.13) Theorem. Let (X, d, p) be a measure space. Let (An):'=1
be a sequence 01 sets in d such that Al C A2 C ... C An C .• '. Then
p(
n=1UAn) = lim P (An) .
,,-..00
00 00
Proof. Write Ao =
Then clearly U An = U (An
0.
. =1 ..=1 n A~ _ I). By
(1O.3.iv), countable additivity, we have
00
p ( "~IAn
)
= .. f 00
p(An n A~_I) = l~ ~ p(An .p n A~_I)

= lim p (
p-..oo
G
=1 (An n A~_I)) = p-..oo
..
lim p(Ag). 0
(10.14) Remark. A result strictly analogous to Theorem (10.13) for

intersections of measurable sets cannot be proved. To see this let X = R,
I-' = A, and d =.AA, the Lebesgue measurable sets. Let An = [n,oo[.
Then A(An) = 00 for n = 1,2,3, ... , so that lim A(An) = 00. On the
n ...... co
other hand, A COl An} = A(0) = O. However, we do obtain the following

result.
(10.15) Theorem. Let (X, d, 1-') be a measure space. II (An):=1
is a sequence 01 sets in d such that I-' (AI) < 00 and 2 An::J •. " Al::J A::J ... ::J
then
00
In particular, il n An =
n=1
0, then lim I-'(An)
11"""",+00
= O.
Proof. The sequence (AI n A~):= I is nondecreasing, and all Al n A~

are in d. Applying (10.13), we get
I-' (AI) - lim I-' (An) = lim (p (AI) - I-' (An)) = lim I-' (AI
n---+ 00 n-+ 00 IS-+- 00
n A~)
= I-' CQI (AlnA~)) = I-' (AI n CgIA~))

= I-' (AlnCBIAn)'} = I-'(Al) -I-' CBIAn) .
Subtracting I-'(Al} [which we may do since I-'(A l } < 00], we have
n~col-'(An} = I-' CBIAn). 0
(10.16) Theorem. Let (An):=1 be any sequence 01 sets in d. Then
(i) I-' tgl COlo An)) ~ l~col-'(A,,} .

We also have
(ii) I-' tOI e.gk An)) ~ ~col-' (A,,) provided that I-' tQI A,,) < 00 •
co
n An C n=2
n An C ... cn=k
n An C •... Theorem
00 00
Proof. It is clear that

n=l
(1O.13) implies that I-' CQl CBkAn)) =,,~~I-' CB"An). We also have
I-'(A,,} ~ I-' CBkAn) for all k. This implies that ~ool-' CB . . An) ~
lim I-'(A,,), from which (i) follows. The inequality (ii) is proved in like
" ...... 00
manner. 0
(10.17) Corollary. Hypotheses are as in (10.16). Suppose also that
kO\ CQkAn) kQ\COkAn)

= = Band p < 00. Then l~oop(Ak)
CQ\A,,)
exists and is equal to p (B).
Proof. The assertion follows from the inequalities lim p (A k) ;;;;;
k-+oo
p (B) ;;;;; lim p
k ..... 00
(Ak)' 0
(10.18) Note. In (10.3), as we trust the reader has noted, count ably
additive measures are defined on algebras d that need not be a-algebras.
This is a technicality, but it is occasionally quite useful. There is a
technique for extending a countably additive measure to a a-algebra
containing d. This is not an essential point in our development of meas-
ure theory, and we relegate it to an exercise (10.36). Note however
that the term measure space is reserved for a triple (X, d, p) in which d
is a a-algebra and p is countably additive on d.
We return to locally compact Hausdorff spaces and outer measures l
as in § 9.
(10.19) Definition. For an arbitrary set X and an arbitrary familyC
of subsets of X, let Y (C) denote the intersection of all a-algebras of sub-
sets of X that contain C. Clearly Y (C) is a a-algebra. Thus Y (C)
is the smallest a-algebra of subsets of X containing C. If X is a topolog-
ical space, let f-I (X) be the smallest a-algebra of subsets of X that con-
tains every open set, i.e., f-I(X) = Y((!)) where (!) is the family of all
open sets. The members of f-I (X) are called the Borel sets of X.
(10.20) Theorem. Let X be a locally compact Hausdorff space and let l
be as in (9.19). Then f-I(X) C vII" i.e., every Borel set is l-measurable,
and (X, vII" l) is a measure space.
Proof. Theorem (9.32) is just the statement that all open subsets
of X are l-measurable [although this phrase is not used in (9.32)].
Thus vii, contains the family (!) of all open subsets of X, and so by (10.11),
vii, :::> 9((!)) = f-I(X). 0
(10.21) Remarks. (a) For many choices of X and l, there are l-
measurable sets that are not Borel sets. For example, let X = R, l = A,
and let P be Cantor's ternary set. Since A(P) = 0 (9.35), we have
A(A) = 0 for all A C P, and so_all ~bsets of P are A-measurable. Thu~
&(P)cvll;.C&(R). Since F=R=c, we have 2c =&(P);;;;; vii;.
;;;;; & (R) = 2c, and so 1;. = 2C• There are exactly c open subsets of R.
This follows from the fact that each open subset of R is a union of open
intervals with rational endpoints. It is therefore a corollary to the next
theorem that f-I (R) = c. This crude cardinal number argument shows
that there are 2C A-measurable sets that are not Borel sets, but it gives
no indication of how to construct such sets.
(b) It is possible actually to construct a very large class of A.-measur-

able sets, the so-called analytic sets, which includes all Borel sets as well
as c other sets. The interested reader is referred to the discussions of
analytic sets in S. SAKS, Theory of the Integral [2nd Edition, Monografie
Matematyczne, Warszawa-Lw6w, 1937J, Chapter III, and in W. SIER-
P[NSKI, General Topology [2nd Edition, Univ. of Toronto Press, Toronto,
1956J, Chapter VII.
(10.22) Exercise. (a) Let X, a, Ea, and ea be as in (9.19). Prove
that ~4 = /y (X).
(b) Let X be a locally compact Hausdorff space, let {Xn }:=1 be a
countable [possibly finiteJ subset of X, and let (IXn ):=1 be a sequence
00 00
of positive numbers such that I: IXn < 00. Let I(rp) = I: IXnrp(xn)
for all
n=1 n=1
rp E<roo (X). Prove that: I is a nonnegative linear functional on <roo (X) ; the
00
corresponding measure t is I: IXnex,,; 1. = /y(X).

n=1
00
I: IXn = 00. Find extra conditions

(c) Extend (b) to the case in which
n=1
on {Xn }:=1 necessary and sufficient for I to be finite for all rp E<roo (X),
and prove the second and third assertions of part (b) for the new I
and t.
The proof of the next theorem gives a method of "constructing"
the a-algebra generated by a given family of sets.
(10.23) Theorem 1. Let X be a set, let tff be a family of subsets of X
such that 0 Etff, and let f/ = f/ (C) be the smallest a-algebra of subsets
of X containing C. If j = e and e ~ 2, then !) ~ etc•.
Proof. For each nonvoid family .F of subsets of X, let .F* be the
00
family of all sets of the form U An' where, for each n EN, either An
n=1
or A~ is an element of .F. Let Q denote the smallest uncountable ordinal
number (4.49). We use transfinite induction to define a family Cf% for
each ordinal number IX < Q. Define Co as the family C. Suppose that
o < IX < Q and that C{J has been defined for each {3 such that 0 ;;;; {3 < IX.
U C{J)*, and write d for the family 0:[,f%<!J
Define Cf% as (0:['{J<f% U Cf%' We
assert that d = f/.
It is clear that tffo = tff c f/. Suppose that C{J C f/ for every {3 < IX
00
and let U An E tffIX' For each n EN, either An or A~ is an element of

n=1
00
U tff{J C f/ so that An' A~Ef/. Thus U An is in f/; i.e., Cf% is contained

{J<f% n=1
1 This theorem is not needed in the sequel and may be omitted by any reader
who is pressed for time; similarly for (10.24) and (10.25).
in .!7. Since .91= U C~, we have proved that .91 C .!7. It is trivial
0;;:;", <.Q
that C C .91, and so to complete the proof we need only to show that
.91 is a a-algebra. Since 0 ECo' we have
X= (0 / U 0 U 0 U ... ) EClCd.
Now let A E d. There is an IX < D such that A EC",; hence A'
= (A U A U ... ) Eiff: C Cp for every {3 > IX, and therefore A 'Ed. Next
I I
let (An)::"= I be a sequence of elements of d. For each n EN, there is an

IXn < D such that An Eiff"n' Apply (4.49.iv) to find a {3 < D such that
IXn < {3 for each n EN. Then
nQI An ECQI Crxnr C Cp C d.

Therefore .91 is a a-algebra and .91 = .!7.
By hypothesis, we have Co = e ~ 2. Considering the ways in which
00
the sets U An ECl can be formed [at most 2 e choices for each AnJ,
n=1
we see that Cl ~ (2e)N. = eN. [(4.32), (4.34), (4.24.vii), and (4.24.xi)J.
Now suppose that C{J ~ eN. for all {3 such that 1 ~ {3 < IX, where
1 < IX < D. Then {J<",
U C{J ~ eN. Ko = eN. (4.32), and so, arguing as above,
i. ~ (eN.)N. = eN. [(4.24.viii) and (4.31)J. It follows by transfinite in-
duction that C" ~ eN. for every IX such that 0 ~ IX < D. Therefore
y=;;= a.<!}
U C.t X:::::eN''"l=e N•
- "
[(4.50) and (4.32)J. 0

(10.24) Exercise. Notation is as in (10.16).
(a) Find .!7(C) if C = {0}.
(b) Find .!7 (C) if C = {U1 , U2 , ••• , Urn}, where the Uj are nonvoid
and pairwise disjoint and Ul U ... U Urn = X. What is .9"(C)?
(c) Find .9"(C) if C is an arbitrary finite family of subsets of X.
What is .9" (Iff) ?
(d) Find .!7 (C) if C is the family of all finite subsets of X [X can
be finite or infinite J. What is .9" (C) ?
(10.25) Corollary. If X is a topological space with a countable base
ct for its topology, then ~(X) ~ c.
Proof. The definition of base (6.10) shows that every open subset of X
is in the family ctl [notation is borrowed with evident changes from
(10.23)]. This proves that .!7(ct) = ~(X); applying (10.23) and (4.34),
we obtain
=
~ (X) ~ (Ko)N. = C. 0
The problem of finding non ,a-measurable sets for a given outer meas-
ure ,a can be terribly complicated. Even for outer measures constructed
as in § 9, no general facts or methods are known. For Lebesgue outer

measure A on PJl (R), however, it is reasonably simple to find non A-
measurable sets. We begin with a simple fact about A.
(10.26) Definition. For subsets A and B of R, let A + B = {x + y:
xEA,yEB},A-B={x-y:xEA,yEB}, and -A={-x:xEA}.
For x ER, the set {x} + A will be written x + A, and is called the trans-
late of A by x. The sets A B, A-I [if 0 ~ A], A B-1, and xA are defined
analogously.
(10.27) Lemma. For all x ER and A C R, the equalities A(X + A)
= A(A) = A(-A) obtain.
Proof. By (9.24) and (9.23), we have
A(A) = inf {A(U) : U is open in R and A C U}
= inf { £ (bn -
n=!
an) : A C UJan, bn[} .
n=!
Since the inclusions 00
00
x +Ac U Jan
n=!
+ x, bn + x[ , (1)
00
-A c U ]-b nJ -an [
n=l
are mutually equivalent, and since the three unions of intervals in (1)
have the same Lebesgue measure, the lemma is proved. 0
(10.28) Theorem. Let T be a A-measurable subset of R such that
A(T) > O. Then T contains a subset E that is not A-measurable. In fact,
E can be chosen to have the following property. If .s1 is a a-algebra of sub-
sets of R such that .4). C .s1 and x + A E.s1 whenever A E.s1 and x E R
[for example d = .4).], and it fl is a countably additive measure on d
such that fl(A) = A(A) for all A E.4). and fl(x + A) = fleA) for all
A E.s1 and x E R, then E ~ .s1.
Proof. Since T = U (T
n=!
n [-n, n]), (9.2l.iv) shows that
00
0< A(T) ~ 2: A(T n [-n, n]) .

n=l
Hence there is aPE N such that

O<A(Tn [-P,P]) ~ A([-P,P])=2P<00,
and we lose nothing in supposing that 0 < A(T) < 00 and that T C [ - p, P].
Since countable sets have A-measure 0 (9.34), we must have
T > ~O.l
1 By (10.30) intra, T contains a compact set F of gositive ).-mea~ure. By (9.34),
F is uncountable, and (6.65) and (6.66) imply that F = C. Hence T = c.

Now let D be any countably infinite subset of T (4.15) and let H be the
smallest additive subgroup of R that contains D. That is, H consists
m
of all finite sums}; nkdk' where the nk's are integers and the dk's are
k=l
in D; from this it is clear that H is countable.
Consider the cosets {t + H: t ER}. Since H is a subgroup of R,
these cosets are pairwise disjoint; i. e., for any t1 and t2, the cosets t1 + H
and t2 + H are disjoint or identical. Let {ty: I' Er} be chosen in R so
r,
that the sets ty + H; I' E are all of the distinct cosets of H, i. e., 1'1 =1= 1'2
implies (tYl + H) n (ty, + H) = 0, and for each t ER, there is a I' Er
such that t+H=ty+H. Let To={YEr:(ty+H) n T=I=0}. For
each I' ETo, choose just one element xl' E(ty + H) n T, and let
E = {xl' : I' ETo}. [Since T is contained in U {ty + H: I' ETo} and
ty + H = No, we must have
ro= T> No.J
In finding E we have twice made an uncountable number of arbitrary
choices, and to do this we must invoke the axiom of choice 1.
To prove that E possesses the pathological properties ascribed to it,
define the set J to be H n (T - T). Since D - D c J c H, it is obvious
that J = No· We claim that (Y1 + E) n (Y2 + E) = 0 for distinct
Y1> Y2 EJ. If not, then there are distinct x1> X 2 EE such that Y1 + Xl
= Y2 + x 2· Since Y1> Y2 EH, this implies
x2 = Xl + (Y1 - Y2) EXl + H,
a contradiction to the definition of E, as Xl and X 2 lie in disjoint cosets
of H. Hence the family {y + E : Y EJ} is pairwise disjoint. Now assume
that E is in the a-algebra d. If fl (E) = 0, our hypotheses on fl imply
that
fl (] + E) = fl (U (y + E) = }; fl (y + E)
yEJ yEJ
=};fl(E)=O. (1)
yEJ
If fl (E) > 0, a similar reckoning gives
fl (] + E) = 00 • (2)
Both (1) and (2) are impossible. To see this, we first prove that
TcJ+E. (3)
In fact, if vET, then v Ety +H = xl' + H for some y ETo, and so
1 Every example of a non A-measurable set has been constructed by using the
axiom of choice. A recent announcement by R. SOLOVAY [Notices Amer. Math.
Soc. 12, 217 (1965)] indicates that without the axiom of choice, non A-measurable
sets cannot be obtained at all.
v = x" + h for some h EH. Thus

h = v - x" E(T - T) n H = J,
which proves (3). If (1) holds, then (3) implies that
p,(T) ~ p,U + E) = O.
Since it(T) is positive and p,(T) = it(T), this is a contradiction, and (1)
cannot hold. It is also obvious that
J +E = (H n (T - T)) +E c (T - T) +T
c [-3P, 3P],
so that
p,U + E) ~ p,([-3P, 3P]) = it([-3P, 3P]) = 6p < 00.
Thus (2) is impossible, and the assumption that E Ed must be re-

jected. 0
(10.29) Remarks. (a) There exists a finitely additive measure p, on
f!jJ (R) such that p, (A) = it (A) for all A E..A;. and p, (x + A) = P, (A)
for x ER and A cR. This was first proved by S. BANACH [Fund. Math.
4, 7-33 (1923)]. The construction is sketched in (20.40) inlra. A far-
reaching generalization of BANACH'S result appears in HEWITT and
Ross, Abstract Harmonic Analysis I [Springer-Verlag, Heidelberg, 1963],
pp. 242-245, to which interested readers are referred.
(b) Count ably additive extensions p, of Lebesgue measure to very
large O'-algebras ..-II of subsets of R have been found, retaining the prop-
erty that p, (x + A) = P, (A). One can make 2( new sets p,-measurable,
and in fact there is a family fl) C ..-II such that ~ = 2( and p, (DID.. D 2) = 1
for distinct DI , D2 Efl). Such extensions are implicit in a construction
given by KAKUTANI and OXTOBY [Ann. of Math. (2) 52,580-590 (1950)].
They are given explicitly in a construction by HEWITT and Ross [Math.
Annalen, to appear].
(c) For an interpretation of (b) in terms of a certain metric space,
see (10.45) and (10.47) below.
We return to our outer measures t on locally compact Hausdorff
spaces, proving some useful facts about t-measurable sets.
(10.30) Theorem. Let X be a locally compact Hausdorff space and let t
00
be as in § 9. Let A be an t-measurable subset 01 X such that A c U Bn

..=1
lor some sequence (Bn)::'=l 01 sets such that t(Bn) < 00 lor all n. Then
teA) = sup{t(F) : F is compact, Fe A} .
Proof. (I) Suppose first that t(A) < 00. Let e be any positive number.
By (9.24) there is an open set V such that A c V and t(V) < t(A) + ! e.
Since t(V) = t(A) + t(V n A'), we have t(V n A') < ! e. Using (9.26),
select a compact subset E of V such that t (V n E') < ! B. Using (9.24)

again, choose an open set W such that V n A' eWe V and t (W) < ! B.
The set F = En W' is compact. It is clear that Fe A, for

E n W' C E n (V' U A) c V n (V' U A) = A.
We have
t (A n F') = teA n (E' U W)) ~ t (A n E') + t (A n W)
I I
~ t(V n E') + t(W) < 2 B + 2 B = B.
Using the t-measurability of A, we see that t(F) = t(A) - teA n F')

> t(A) - B. Since B is arbitrary, the theorem follows for t(A) < 00.
(II) Suppose that t(A) = 00. In view of (9.24), we may suppose that
the sets Bn in our hypothesis are t-measurable [in fact, open]. Write
An = A n (BI U .•• U Bn) for n EN and Ao = 0. Then An is t-measur-
00
able, t(An) < 00, An C An+l' and n=!

U An = A.I
By (10.11), (X,.4" t) is a measure space; hence (10.13) implies
that
00 = t(A) = lim t(An) . (1)
n-->-oo
Using part (I), choose for each n EN a compact set Fn such that Fn C An
and t(Fn) ~ ! t(An). It is plain from (I) that
lim t(Fn) = lim t(An) = 00 = t(A). 0
n-+-oo n~oo
(10.31) Theorem. Let X be a locally compact Hausdorff space and

let t be as in § 9. For A C X, the following statements are equivalent:
(i) A is t-measurable;
(ii) t(U) ~ t(U n A) + t(U n A') for all open sets U such that
t(U) < 00;
(iii) A n U is t-measurable for each open set U such that t(U) < 00;
(iv) A n F is t-measurable for every compact set F.
Proof. Theorem (10.20) shows that each compact set F is t-measur-
able, since it is closed; and so (i) implies (iv).
Suppose that (iv) holds and let U be an open set such that t(U) < 00.
Theorem (10.30) shows that for each n EN there is a compact set
1 00
Fn C U such that t(Fn) > t{U). - -.
n Let F = n=!
U Fn. Then we have:
F C U; F is t-measurable; and t(F) ~ t(Fn) > t(U) - ~ for each n EN.
1 A set that is the union of a countable family of sets of finite measure is called
a-ftnite [ef. (10.3)]. Recall also: a set that is the union of a countable number of
compact sets is called a-compact.
It follows that t(F) = t(U) and t(U n F') = O. Hence

A nU= A n [F U (U n F')] = (A n F) U (A nun F')
= CQI (A n Fn)) U (A nun F') ,

and so, since t (A nun F') = 0, A n U is a countable union of t-
measurable sets. Therefore (iv) implies (iii).
Next suppose that (iii) holds and let U be open of finite t-measure.
Then U and A n U are both t-measurable, and so UnA' = un (U n A)'
is t-measurable. Thus
t(U) = t((U n A) U (U n A')) = t(U n A) + t(U n A') ,
i. e., (iii) implies (ii).
Suppose that (ii) holds, and let T be an arbitrary subset of X. In
establishing (10.5), we may suppose that t(T) < 00, since otherwise
t(T) = 00 ;;;; t(T n A) + t(T n A'). For a given B > 0, choose an open
set U such that T c U and t(U) < t(T) + B. Then (ii) implies that
t(T) + B > t(U) ~ t(U n A) + t(U n A') ~ t(T n A) + t(T n A') .
Since B is arbitrary, we are through. 0
(10.32) Corollary. II A is locaUy t-nuU, then A is t-measurable and
t(A) = 0 or 00.
Proof. For each compact set F we have t(F n A) = 0 (9.29), and so
F n A is t-measurable (10.7). It follows from (10.31) that A is t-measur-
able. Suppose that t(A) < 00. Applying (10.30), we have t{A) = sup{t(F) :F
is compact and F C A} = O. 0
(10.33) Remark. Since for some choices of X and t there exist locally
t-null sets which are not t-null [see (9.4 l.e)] , (10.32) shows that (10.30)
cannot in general be strengthened to admit all t-measurable sets. How-
ever if X is a countable union of compact sets [e.g. X = R"], then
every t-measurable subset of X satisfies the hypothesis of (10.30).
(10.34) Theorem. Let X be a locaUy compact Hausdorff space and let t
be as in §9. For every a-finite, t-measurable subset A 01 X, there are subsets B
and C 01 X such that B is a-compact, C is a Borel set, the inclusions Be A c C
obtain, and t(C n B') = O.
Proof. (I) Suppose that t(A) < 00. For each n EN, there is a compact
1 00
set Fn C A such that t(Fn) > t(A) - -.
n
Let B = n=1
U Fn. For each n
we have t(Fn) ~ t(B) ~ t(A), and so t(B) = t(A). Next, for each
1
n EN, select an open set Un::> A such that t(Un) < t(A) + n' Let
00
C = n Un; then C is a Gd set, and hence clearly a Borel set. It is clear

n=1
that t(C) = t{A). Using the t-measurability of A, we have t{C n A')
= t{C) - t{A) = 0, t{A n B') = t{A) - t{B) = 0, and so also

t{C n B') = t{C n A') + t{A n B') = o.
00
(II) If t{A) = 00, then, as in the proof of (10.30), write A = ,,=1

U An
where each An is measurable and has finite measure. By case (I), there
are a-compact sets B .. and Gd sets Cn , n = 1,2,3, ... , such that
00
Bn C An C Cn and t{Cn n B~) = O. Let B = ,,=1

U Bn; B is clearly a-
compact. We have
A n B' = CQIA. ) n Cgl BRr = CQI An) n COl m) c"gl (An n B~) ,
and so
00
U Cn ; C is clearly a Borel set. The argument given above

Now let C = ,,=1
to prove that t{A n B') = 0 can be used to prove that t{C n A') = 0,
and as in part (I) it follows that t(C n B') = o. 0
The functional I of § 9 satisfies the inequality I (f + g) ;;£; I (I) + I (g)
for alII, g E5'+ (9.18). It is possible to exhibit functionals I and functions
I and g such that strict inequality holds [see (10.41)J; i.e., I is not in
general additive on 5'+. However I is additive on special classes of func-
tions, and we now exhibit one such class.
(10.35) Theorem. Let A and B be disjoint t-measurable sets and let
IX and (J be nonnegative real numbers. Then we have
(i) I(IX~A + (J~B) = IXI{~A) + (JI(~B)·
Proof. By the subadditivity of I, it obviously suffices to prove that
I (IX~A + (J ~B) ;;?; IXI (~A) + {JI (~B) . (1)
The inequality (1) is easy to verify if IX = 0 or (J = 0, or if t(A) = 0 or
t(A) = 00, or if t(B) = 0 or t(B) = 00. We leave these verifications to
the reader, and prove (1) under the hypothesis that IX{J> 0,0 < t(A) < 00,
and 0 < t(B) < 00.
(I) Suppose that A and B are compact. By (6.80), there is a contin-
uous real-valued function qJ on X such that qJ(A) = to} and qJ{B) = {I}.
The sets {x EX: qJ (x) < !}
and {x EX: qJ (x) > ~} are open disjoint
sets containing A and B, respectively. Since A and B have finite measure,
they are contained in open sets having finite measure. Taking the
intersections of these open sets with those defined by qJ, we obtain open
sets Uo and Vo such that Uo :::> A and Vo:::> B, Uo n v;, = 0,
0< t(Uo) < 00, and 0 < t(Vo) < 00. We have
I(IX~A + (JEB) ;;£; IXI(~A) + (JI{~B) = IXt{A) + (Jt{B) < 00 •
Now choose 8> O. There is a function I Em1+ such that I ~ ft~.A + (J~B
and
1(/) -
-
a8
1 -
< I(ft~.A + (J~B) .
Choose 15 > 0 such that
0<!5<min{3'~Uo)' 3'~Vo)
,ft,{J}.
For all x EA, we have I (x) ~ IX. By the lower semicontinuity of I, there
is an open set U such that A cUe Uo and I (x) > ft - 15 for all x E U.
Similarly, there is an open set V such that B eVe Vo and I (x) > {J - 15
for all x E V. Thus we have I ~ (IX - 15) ~u + ({J - 15) ~v; therefore
I (I) ~ I «(IX-!5) ~u+ ({J -15) ~v)= (ft-!5) I (~u) + ({J -15) I (~v)
= (ft- !5)t(U) + ({J- !5)t(V) ~ ftt(A) + (Jt(B) - !5(t(U) + t(V))
2
~ IXt(A) + (Jt(B) - !5(t(Uo) + t(Vo)) > ftt(A) + (Jt(B) - a8 .
Summarizing, we have shown that
-
ftI (~.A) + {JI- (~B) - 8 < I- (I) - a1 8 < I- (IX~.A + (J ~B) .
Since 8 is arbitrary, it follows that (1) holds if A and B are compact.
(II) We suppose now that A and B are arbitrary t-measurable sets
of finite positive measure. Choose 8> O. Using (10.30), choose compact
1
sets E and F such that E C A and FeB, IXt(E) > IXt(A) - 2 8, and
(Jt(F) > (Jt(B) - ~ 8. Using part (I) for the compact sets E and F,
we have
I(IX~.A + (J~B) ~ l(ft~E + (J~F) = ftt(E) + (Jt(F) > ftt(A) + (Jt(B) - 8
= IXI(~.AJ + fJI(~B) - 8. 0
We close this section with a large collection of exercises. A few of
these [for example (1O.37)J are actually needed for subsequent theorems
in the main text. The rest of the exercises illustrate and extend the theory
in various directions, and we trust that all serious readers will work
through most of them.
(10.36) Exercise. Let X be an arbitrary set and d an algebra of
subsets of X [d need not be a a-algebraJ. Let fA be a countably additive
measure on d in the sense of (10.3). Define a set function p, on 9fJ(X)
as follows: for T C X, let
P,(T) = inf{..g fA (An) : T C nQIAn and AI' A!!, ... , An"" Ed} .
(a) Prove that p, is an outer measure on 9fJ(X).
(b) Prove that p, is equal to fA on the algebra d.
(c) Prove that all elements of d are measurable with respect to p,.
(d) Prove that p can be extended to a count ably additive measure

defined on a a-algebra of subsets of X that contains d. [The fact that
this can be done is called E. HOPF'S extension theorem.]
(10.37) Exercise. Let X and d be as in (to.36). Let y be a set func-
tion on d satisfying the following conditions:
(i) 0 ~ Y (A) ~ 00 for all A Ed;
(ii) y(A U B) = y(A) + y(B) forA, B EdandA n B= 0;
00
(iii) if Av A 2, ... Ed, if Al ::::> A2 ::::> ••• ::::> An ::::> •• " and n An =
.. =1
0,
then lim y (An) = O•
........ 00
Define y just as in (to.36). Prove that (a), (b), (c), and (d) of (10.36)
hold for the set functions y and y. [This is another version of E. HOPF'S
extension theorem.]
(10.38) Exercise. Let (X, d, p) be a measure space. Prove that p
can be extended to a measure p, on a a-algebra d such that every subset
of every set of p-measure 0 is p,-measurable and has p,-measure O.
The following exercise will be needed in the sequel to prove two im-
portant theorems of the main text [(20.56) and (20.57)], and so we
spell out the proof in some detail.
• (10.39) Exercise. Let X, d, p, and p, be as in (to.36). Let .9 = .9 (d)
be the smallest a-algebra of subsets of X that contains the algebra d.
Suppose that (X,.9,,,) is a measure space such that ,,(A) = peA)
for all A Ed. Prove the following.
(a) If BE.9, then P,(B) ~ ,,(B). [Hint. Let .s(. denote the family
00
of all countable unions of sets in d. If A = U An E.s(. where

.,=1
00
{An}:'=1 C d, then A = Al U nl!,2 (An n A~ n A~ n ... n A~_l) is a

disjoint union of sets in d, and so " (A) = P,(A). Thus
P,(B) = inf{P,(A) : Be A E.s(.} = inf{,,(A) : B CAE.s(} ~ ,,(B).]
(b) If FE.9 and P,(F) < 00, then ,,(F) = P,(F). [Hint. For e > 0,
choose A E.s(. such that Fe A and P,(A) < p,(F) + e. Then use (a)
to show that
P,(F) ~ P,(A) = " (A) = " (F) + " (A nF') ~ " (F) + p, (A n F') < " (F) + e.]
(c) If there is a sequence (Fn):'=l C d such that p(Fn) < 00 for all
00
n and X U F , then ,,(E) = P, (E) for all E E .9, i. e., the extension
n=1 n
=
of p to .9 is unique. [Hint. We may suppose that Fn n Fm = 0 for

n =1= m. Then by (b) we have
00 00
,,(E) = E ,,(E n Fn) = E P,(E n Fn} = P,(E}.]

.. ~1 n=1
(d) If the a-finiteness hypothesis in (c) fails, then fl may have more
than one extension to .:7. [Hint. Let X = [0, 1[ and let d be the algebra
°
of all finite unions of intervals of the form [a, b[ C [0, 1 [. Define fl
on d by fl (0) = and fl (A) = 00 if A =l= 0. Show that there are exactly
2c countably additive measures on the Borel sets of [0, 1[ that agree
with fl on d.]
(10.40) Exercise. An outer measure fl on gJ (X) is said to be regular
if for each E C X, there exists a fl-measurable set A C X such that
E C A and fleA) = fleE).
(a) Prove that any outer measure obtained from a measure on an
algebra of sets as in (10.36) is a regular outer measure.
(b) Prove that if X is a locally compact Hausdorff space and l is
as in § 9, then l is a regular outer measure.
(c) Let X = {O, I}. Construct an irregular outer measure on gJ(X).
(d) Let fl be a regular outer measure on gJ(X), let E C X, and let d
be the smallest a-algebra containing {E} U vII,.. Prove that if fl is finitely
additive on d, then E Evil,..
(e) Notice that, since A is a regular outer measure on R, none of the
extensions of A [as a measure on ~] mentioned in (1O.29.b) can agree
with the outer measure A except at sets in ~.
(10.41) Exercise. Let A be a subset of [0, 1] that is not A-measurable,
and let B = [0, 1] n A'. Prove that S (~A + ~B) < S (~A) + S (~B), where 5
is the Riemann integral.
(10.42) Exercise. Let X and Y be topological spaces. Prove the
following.
(a) If t is a continuous function from X into Y and if B E~(Y),
then 1-1 (B) E~ (X). [Consider the family of all sets B for which the asser-
tion holds.]
(b) If A E~(X) and B E~(Y), then A x B E~(X x Y). [Recall
the definition of the product topology (6,41) and use (a).]
(c) Generalize (b) to products of countably many topological spaces.
(10.43) Exercise [H. STEINHAUS]. Let T be a A-measurable set in R
such that A(T) > 0. Prove that the set T - T contains an interval
[-IX, IX] (IX> 0). The following steps may be useful.
(a) If U and V are open in R and have finite A-measure, the
function x -+ A«(X + U) n V) is continuous on R. [Begin with intervals
and use (6.59) for general U and V.]
(b) If A and B are A-measurable of finite A-measure, then
x -+ A«(X + A) n B) is continuous. [For U:::> A and V:::> B, prove first
that
JA«(X + U) n V) - A«(X + A) n B)J ;;i A(U n A') + A(V n B').]
(c) The set T - T contains an interval [- oc, oc]. [The function

x ~ A«(X + T) n T) is positive at 0, and if (x + T) n T =1= 0, then
x ET - T.]
(10.44) Exercise. Generalize (10.43) to A + B, where A, B are in
..A}, and both have positive measure.
(10.45) Exercise. Let (X, vii, p,) be a measure space. Define cp on
[0, 00] by letting cp (t) = 1 - exp (- t) if 0 ~ t < 00 and putting cp (00) = 1.
For A, BEvil, define
e(A, B) = cp(p,(A 6. B».
(a) Identifying sets A and B for which p, (A 6. B) = 0, prove that
(vii, e) is a complete metric space.
(b) Show that the mappings from vii x vii to vii with values AU B,
A 6. B, and A n B at (A, B) are continuous. Show also that A ~ A'
is continuous from vii to vii.
(10.46) Exercise. Prove that the metric space (..,(I, e) defined in
(10.45) is not compact in the case that X = [0, I], ..,(I = £W([O, I]), and
p, = A.
(10.47) Exercise. (a) Let X be a locally compact Hausdorff space
and let £ be a measure on X as in § 9. Prove that if there exists a countable
base for the topology of X, then the metric space (vA;, e) defined as in
(10.45) is separable.
(b) Note that the metric space (..A)" e) has a countable dense subset,
where ..A}, is the a-algebra of Lebesgue measurable subsets of R and A
is the measure used to define e, as in (10.45). Find the smallest cardinal
number of a dense subset of (vii, e) for the invariant extension I' of
Lebesgue measure described in (10.29.b).
(10.48) Exercise. Let X be a metric space with metric (! and let p,
be any outer measure on {JjJ (X) such that if A, B C X, A =1= 0, B =1= 0,
and e(A, B) > 0, then p,(A U B) = p,(A) + p,(B). Such outer measures
are called metric outer measures. Let U be an open proper subset of X
and let A be a nonvoid subset of U. For each n EN define An = {x EA :
(! (x, U') ~ !}. Prove that:
(a) lim p'(An) = p,(A)
n ......oo
[consider the sets D 2n = A 2n n A~n-l and D2n +1 = A 2n +1 n A~n];
(b) U is p,-measurable;
(c) £W (X) C ~.
(10.49) Exercise: Construction of a class of outer measures. Let X
be a separable metric space, let (f) be the family of all open sets in X,
and let p be a positive real number. For each e > 0 let (f). = {U E(f):
diamU ~ e} U {0}. For each E C X, define
pp,.(E) = infL~1 (diam Un)P : Un E(!IB' E C nQI Un}'

where we define diam 0 = O.
(a) Prove that pp,.(E) is nondecreasing as e decreases.
Define pp(E) = ~i~ pp,.(E) for each E C X.
(b) Prove that pp is a metric outer measure on X.
(c) Prove that if pp(E) < 00 and q > p, then pq(E) = O.
The set function pp is called the Hausdorff p-dimensional [outer]
measure on X. For E C X we define the Hausdorff dimension of E to be
the number sup{P ER: p > 0, pp(E) = oo}, where we let sup 0 = O.
(10.50) Exercise. Let R have its usual metric. We consider Haus-
dorff measures pp on R [see (10.49)]. For E C R, let dimE be the Haus-
dorff dimension of E. Prove that
(a) Pt = A;
(b) dimU = 1 for all nonvoid open sets U C R;
(c) dimE = 0 implies A(E) = 0;
(d) if P is CANTOR'S ternary set, then dimP = ~:!~ [consider the
sets Pn of (6.62)];
(e) there is an uncountable subset E of R such that dimE = O.
(10.51) Exercise: Another class of outer measures. Let (X, e) be a
metric space. For a nonvoid set E C X and t> 0, define n(E, t) as
follows:
n(E, t) = 1 if e(x, x') ~ t for all x, x' EE;
n (E, t) = sup{F : FeE, F is finite, e (x, x') > t for distinct x, x' EF}
if this supremum is finite;
n(E, t) = 00 in all other cases.
Define n(0, t) as zero. Let qJ be a real-valued, positive, strictly de-
creasing function defined on ]0, 1] such that lim qJ (t) = 00. For all E C X,
define I.j,O
· n (E, t)
extl/> (E) =
11m --(t)- .
ItO f{J
For all E c X, define
1I1/>(E) = inf t~ extl/> (A k )} ,
where the infimum is taken over all countable, pairwise disjoint families
00
of sets {AkH"=1 such that k~l Ak = E.

(a) Prove that "1/> is a metric outer measure as defined in (10.48).
[Hint. The only nontriviality is showing that "1/> (A U B) = 1I1/>(A) + "1/> (B)
if e(A, B) > O. This follows from the equality n(A U B, t) = n(A, t) +
+.
n(B, t), which is valid for t < e(A, B).]
(b) Compute "'1' for X = R with the usual metric for Rand tp (t) =
(c) Compare the outer measures "'1' [with suitable tp!] with Haus-
dorff p-dimensional measures.
(d) Prove that lI!p(E) = 0 if E is countable.
(e) Prove that "'1'(E) = lI.p(F) if there is an isometry of E onto F.
[An isometry is a mapping 1p of one metric space onto another such that
e(x,y) = e'(1p(x), 1p(y)), e and e' being the metrics on the two spaces.]
(10.52) Exercise. Let DC be any real-valued nondecreasing function
on R and let A" be the Lebesgue-Stieltjes measure on R induced by the
Riemann-Stieltjes integral as in § 9. Prove that A" ({x}) = 0 for x ER
if and only if DC is continuous at x.
(10.53) Exercise. Let DC and A" be as in (10.52) and suppose that DC
is continuous. Prove the following assertions.
(a) For each e > 0 there exists a nowhere dense perfect set A C [0,1]
such that A,,(A) > A,,([O, 1]) - e.
(b) There exists an Fa set Be [0, 1] such that B is of first category
and A,,(B) = A,,([O, 1]).
(c) There exists a Gd set of second category contained in [0, 1]
having A,,-measure zero.
(10.54) Exercise. In this exercise, we first sketch the construction of
a subset B of R measurable for no measure A" with continuous DC.
(a) Prove that every uncountable closed subset F of R has cardinal
number c. [Use (6.65) and (6.66).]
(b) [F. BERNSTEIN]. Prove that there is a subset B of R such that
B n F =1= {1) and B' n F =1= {1) for every uncountable closed subset F
of R. [Hints. There are just c open subsets of R and hence just c uncount-
able closed subsets. Let We be the smallest ordinal number with correspond-
ing cardinal number c [use (4.47) to show that We exists]. Let {I;: 1j< we}
be a well ordering of all uncountable closed subsets of R. Define B by
transfinite recursion and the axiom of choice, as follows. Let Xo and Yo
be any two distinct points in Fo. Suppose that Xy and yy have been
defined for all 'Y < 1j, where 1j < We' The set A'1= {Xy: 'Y < rJ} U {yy: 'Y < rJ}
has cardinal number < c, because We is the smallest ordinal number of
cardinal c. Hence the set I; n A~ has cardinal number c. Let x'1 and Y'1
be any two distinct points in the set I; n A~. Finally let B={x'1: rJ < we}'
It is clear that B n I; =1= {1) and that B' n I; =1= {1) for allrJ < we.]
(c) Prove that B is non A,,-measurable if DC is continuous and A" =1= O.
[Hints. Assume that B is A,,-measurable. Then by (10.30), we have
A,,(B) = sup{A,,(F) : F is compact, FeB}. The only compact subsets
of B are countable, and since A,,({X}) = 0 for all x ER, it follows that
A/Z (B) = O. Similarly A/Z (B') = 0, and so if B is A/Z-measurable, A/Z is the

zero measure.]
Throughout the remainder of this exercise, assume the continuum
hypothesis, i.e., at1 = c [see (4.49) and (4.50)].
Cd) Prove that there exists an indexing {C'1: 0 ~ 'YJ < Q} of the family
of all nowhere dense closed subsets of [0, 1] by the set PD of all countable
ordinal numbers.
Define [by transfinite recursion and the axiom of choice] a set
5 = {x'1: 0 ~ 'YJ < Q} as follows: let Xo E [0, 1] n C~ and
x'1 E [0,1]n (o~'1 (Co U {XO}))/.
(e) Prove that S = c and that 5 n C'1 is countable for all 'YJ E PD.
(f) Prove that A/Z (5) = 0 for all Lebesgue-Stieltjes measures A/Z such
that eX is continuous.
(10.55) Remark. The set 5 defined above is not a Borel set in R.
In fact, it is known that each uncountable Borel set in a complete separa-
ble metric space contains a nonvoid perfect set [see W. SIERPINSKI,
loco cit. (10.21.b), p. 228].
(10.56) Exercise. (a) Let (X, .91, "') be a measure space such that
o < '" (X) < 00 and", assumes only a finite number of distinct positive
values. Prove that X = El U ... U En U F, where the summands are
d-measurable and pairwise disjoint and have the following properties.
There is a nondecreasing sequence (eX"):=1 of positive numbers such that
if A Ed and ACE", then ",(A) = 0 or ",(A) = eX,,; ",(E,,) = eX,,; and
'" (F) = o. [Hint. Let eXl be the least positive value assumed by '" and
let El be any set in .91 ·for which ",(E1) = eXl. Consider Ei and proceed
by induction.]
(b) Let X be an uncountable set, .91 the family {A : A C X, A is
countable or A' is countable}, and", on .91 defined by ",(A) = 1 if A'
is countable and ",(A) = 0 if A is countable. Prove that (X, .91, "') is
a measure space. Use this example to show that the decomposition
described in (a) need not be unique.
(c) Let (X, .91, "') be a measure space such that 0 < "'(X) < 00 and '"
assumes infinitely many distinct values. Show that there is an infinite
pairwise disjoint family {An}~l cd such that 0 < ",(An) < 00 for all n.
(d) Let (X, .91, "') be a measure space such that every A Ed such
that ",(A) = 00 contains a set B Ed such that 0 < ",(B) < 00. Then
every such A contains a set C Ed such that", (C) = 00 and C is the union
of a countable number of sets of finite measure. [Hints. Let eX = sup{", (B) :
BEd, B c A, ",(B) < oo}. Let (Bn):=l be a nondecreasing sequence of
sets in .91 such that ",(Bn) < 00, Bn C A, and lim ",(Bn) = eX. Let
"--'00
00
C = U Bn. The assumption", (C) < 00 leads at once to a contradiction.]
.. =1
10*
(10.57) Exercise. Let X be a set and let .fL7 be a family of subsets

of X such that A U B E.fL7 and A n B E.fL7 if A, B E.fL7. Suppose also
that X E.fL7 and 0 E.fL7. Such a family .fL7 is called a lattice of sets [with
unit and zero J. Let!!) be the family of all proper differences of sets in .fL7,
i.e.,!!) = {B n A': A, B E.fL7, A C B}. Finally, let Olt be the family of
all finite disjoint unions of sets in !!). Prove the following.
(a) If Dv D2 E!!), then DI n D2 E!!). [If D j = B j n Ai (j = 1,2),
then DI n D2 = (BI n B 2) n «(AI n B 2) U (A2 n BI))'·J
(b) If Uv U 2 EOlt, then UI n U 2 EOlt.
(c) If U EOlt, then U' EOlt. [Use induction on the number n, where
U = DI U D2 U ... U Dn·J
(d) The family Olt is the smallest algebra of subsets of X containing .fL7.
Let d be a a-algebra of subsets of X such that .fL7 C d and let ft
and v be two measures defined on d.
(e) If ft(A)=v(A) for all A E.fL7 and if there exists a sequence
00
(An) c.fL7 such that n~l

U An = X and ft (An) < 00 for all n EN, then
ft(E) = veE) for all E E Y(.fL7). [First show that ft and v agree on Olt
and then use (1O.39).J
(10.58) Exercise. Use (10.57) to prove the following.
(a) If X is a topological space and ft and v are two finite measures
defined on .@(X) that agree on (1) the family of all open sets, or (2) the
family of all closed sets, or (3) the family of all compact sets [in the case
that X is a-compactJ, then ft and v agree on '@(X).
(b) If X is a metric space and ft is a finite measure defined on '@(X),
then ft(E) = inf{ft(U) : U is open, E C U} for all E E'@(X). [Define
v(E) = inf{ft(U): U is open, E C U} for all E C X. Show that v is a
metric outer measure (10.48) and use (a).J
§ 11. Measurable functions

(11.1) Introduction. As was pointed out in (IOAO.d), the outer meas-
ures £ constructed in § 9 from nonnegative linear functionals I need not
be even finitely additive on all sets. However, we have learned that they
are in fact countably additive on their a-algebras of measurable sets.
In like manner, we cannot expect that the extensions I should be finitely
additive on all nonnegative functions [see (IOAl)J. In this section we
construct a large class of functions on which the functionals I are countably
additive. [The countable additivity will be proved in § 12.J These so-
called measurable functions bear a relationship to the family of all func-
tions which is analogous in many ways to the relation between measurable
sets and the family of all sets.
§ 11. Measurable functions 149
Throughout this section, X will denote an arbitrary set and d will

denote an arbitrary a-algebra of subsets of X. The ordered pair (X, d)
is called a measurable space 1.
(11.2) Definition. Let I be an extended real-valued function defined
on X. Suppose that 1-1 (Ja, 00]) Ed for every a E R, i. e., {x EX:
a < I (x) ;;;;; oo} Ed for all real numbers a. Then I is said to be an d-
measurable lunction. [The reader should notice that this definition closely
resembles the definition of lower semicontinuity (7.21.d).] If X is a
topological space and d is the a-algebra fl (X) of Borel sets, then any
fl (X)-measurable function is said to be Borel measurable. If X = R
and d =..A)., then an ..A}.-measurable function is called a Lebesgue
measurable lunction. [Notice that the definition of measurable function
depends in no way upon any measure, but only upon a particular
a-algebra. ]
(11.3) Theorem. Let D be any dense subset 01 R [that is, D- = R].
The lollowing conditions on an extended real-valued lunction I with domain
X are equivalent:
(i) I is d-measurable;
(ii) 1-1 (Ja, 00]) Ed lor all a ED;
(iii) 1-1([a, 00]) Ed lor all a ED;
(iv) 1- 1([-00, aD Ed lor all a ED;
(v) 1- 1([-00, a]) Ed lor aU a ED.
Proof. It is trivial that (i) implies (ii). To see that (ii) implies (iii),
let a ED and let (an) be a strictly increasing sequence in D such that
00
an -+ a. Then we have 1- 1([a, 00]) = n 1-1 (Jan, 00]). To see that (iii)
n=1
implies (iv), observe that 1- 1([-00, aD = (t-1([a, 00]))'. The proof that
(iv} implies (v) is similar to the proof that (ii) implies (iii). It remains
only to show that (v) implies (i). For a E R, choose a strictly decreasing
sequence (bn ) in D such that bn -+ a. Then
1- 1(Ja, 00]) = C61 1-1([-00, bn]))' D
(11.4) Theorem. Let I be an extended real-valued lunction having

domain X. Then I is d-measurable il and only il
(i) 1-1({- oo}) and 1- 1({ oo}) are both in d,
and
(ii) 1-1 (B) Ed lor every B Efl (R).
Proof. Since ]a,oo[Efl(R) and ]a,oo]=]a,oo[U{oo} for every
a ER, it is clear that (i) and (ii) imply that I is d-measurable.
1 A purist might cavil at the term "measurable space", as there is absolutely no
guarantee that a nontrivial measure exists on d. We use the term faute de mieux.
Conversely, suppose that I is d-measurable. Then

00 00
I-l({-oo}) = n 1-1 ([-00,-nJ)Ed

n=!
and 1-1 ({oo})= n 1-1 (In,ooJ)Ed.
n=!
Thus (i) obtains. To prove (ii) , let 09 = {5 c R: 1-1 (5) Ed}. We will
show that 09 is a a-algebra of subsets of R. Clearly 0 E Y. If (5 n ) is
any sequence in 09, then 1-1 C~! 5 n ) = nQ! 1-1 (5 n ) Ed; thus countable
unions of sets in 09 are again in Y. If 5 E 09, then
l-l(R n 5') = [I-I (5) U 1-1 ({- oo}) U 1-1 ({ oo})]' Ed;
thus 09 is closed under complementation. It follows that 09 is a a-algebra
of subsets of R. We next show that 09 contains every open subset of R.
Indeed, since I is d -measurable, R E09, and (i) is true, we have 1-1 (Ja, 00 J)
Ed and 1- 1([-00, b[) Ed whenever - 00 ~ a < 00 and - 00 < b ~ 00.
Thus if Ja, b[ is any open interval of R, we have
1-1 (Ja, b[) = I-l([-oo,b[) n 1-1 (Ja, ooJ) Ed
and so Ja, b[ E Y. It follows that all open subsets of R are in Y. Thus
09 is a a-algebra containing PA (R) and (ii) obtains. 0
(11.5) Corollary. Suppose that X is a topological space and that
d::::> ~(X). Then all real-valued continuous lunctions and all extended
real-valued lower [upperJ semicontinuous lunctions defined on X are d-
measurable.
(11.6) Remark. It is clear that if I is a real-valued Lebesgue measur-
able function on R, then 1-1 (B) is a Lebesgue measurable set whenever B
is a Borel set. It is worth noting that even for certain real-valued con-
tinuous functions I on R there exist Lebesgue measurable sets A such that
t- 1 (A) is not Lebesgue measurable. We sketch the construction of such
a set. Let P be a nowhere dense perfect subset of [0, IJ such that
°
infP = 0, supP = 1, and A(P) > (10.53.a). Let C denote CANTOR'S
ternary set. Let f and f denote the families of component open sub-
intervals of [0, IJ that are complementary to P and C, respectively.
Linearly order both f and f in the obvious way [II < 12 if II lies to
the left of 12 ]. Then f and f are both of order type 'YJ (4.53), and so
there exists an order-isomorphism q; from f onto f. [If P is a Cantor-
like set constructed as in (6.62), then q; can be defined explicitly by
associating complementary intervals having like subscripts.J Define
a function I from [0, IJ onto [0, IJ as follows: 1(0) = 0; for I Ef, define I
linearly from I onto q; (I) by mapping the lower [upperJ endpoint of I
to the lower [upperJ endpoint of q; (I) and joining with a line segment;
and for x E P n {O}', define I(x) = sup{/(t): t < x, t E Uf = [0, IJ n PI}.
Then I is a continuous one-to-one [strictly increasing!J function from
[0, IJ onto [0, IJ and I(P) = C. Let 5 be a non Lebesgue measurable
subset of P (10.28) and let A = 1(5). Then we have Ace, so that

A(A) = 0 and A E.,A),. However l-l(A) = 5 ~.,A),. [Note that the above
construction can be used to show that any two nowhere dense compact
perfect subsets of R are homeomorphic.]
We conclude from (11.4) that the set A in the above example is not
a Borel set. This is yet another proof that there exist Lebesgue measur-
able sets that are not Borel sets [ef. (1O.21.a)].
Let I, A, and 5 be as above and let g = ~.A' It is clear that g 0 I = ~s
(on [0, 1]). Notice that I and g are both Lebesgue measurable and that
g 0 I is not. Thus the composition of two measurable functions need
not be measurable. We do however have the following theorem.
(11.7) Theorem. Let rp be any extended real-valued lunction defined
on R# such that rp-l([a, 00]) n R is a Borel set lor all real a, i.e., rp is
f!J(R#)-measurable. Let I be .!ii-measurable. Then rp 0 I is d-measurable.
Proof. We have
(rp 0/)-1 ([a, 00]) = I-l(rp-l ([a, 00]))
= I-l(rp-l([a, 00]) n R) U A+ U A_)
= 1-1 (rp-l ([a, 00]) n R) U 1-1 (A+) U 1-1 (A_) ,
where A+={oo}nrp-l([a,oo]) and A_={-oo}nrp-l([a,oo]). Since
rp-l([a,oo]) n R is a Borel set by hypothesis, 1-1 (rp-l ([a, 00]) n R)
is in d, by (11.4). It is easy to see that 1-1 (A+) and 1-1 (A_) are also
in.!il; therefore (rp 0 1)-1 ([a, 00]) E.!iI, and so rp 0 I is d-measurable. 0
(11.8) Theorem. II I is .!ii-measurable, then the lollowing assertions
hold.
(i) The lunction I + IX is .!ii-measurable lor all real IX.
(ii) The lunction IXI is .!ii-measurable lor all real IX.
!
(iii) Let
II (x) IP il I (x) is finite,
h (x) = p_ il I (x) = - 00 ,
p+ il I(x) = 00 ,
where p_ and p+ are arbitrary extended real numbers and p is any positive
real number. Then h is .!ii-measurable.
!
(iv) Let
r
[f (x) il I (x) is finite,
hex) = p_ il I(x) = -00,
p+ il I (x) = 00 ,
where m is a positive integer and p_ and p+ are arbitrary extended real
(v) Let h = +
numbers. Then h is .!ii-measurable.
where I is finite and not zero, and let h assume constant
but arbitrary values p+, p_, and Po on the sets {x EX: I (x) = oo},
{xEX:/(x)=-oo}, and {xEX:/(x)=O}, respectively. Then h is

d -measurable.
Proof. In each case, we define a suitable function ep such that the
function in question is equal to ep 0 I and apply (11.7). For (i), let
To prove (ii), let

ep (t) = r+
±oo
IX
if t ER,
if t = ±oo.
~ 1::00
if t = - 0 0 ,
W(t} if t ER,
if t = 00,
if IX> 0, and
if t=-oo,
p(t}~l~oo if t ER,
if t = 00,
if < 0; if IX = 0, the assertion is trivial.
IX
For (iii), let ep(±oo) = {3± and ep(t) = ItIP for real t. Since ep is con-
tinuous on R, it is clear that ep-l ([a, 00 J) n R is a Borel set for all real a.
The proof of (iv) is similar, with ep (t) = tm for realt and ep (± 00) = {3±.
To prove (v), let ep (t) = t1 for t =l= 0,00, - 00, and let ep (0) = {3o,
ep(±oo)={3±. D
(11.9) Lemma. Let I and g be d-measurable. Then the sets
(i) {x EX: I (x) > g(x)} ,
(ii) {x EX: I (x) ~ g(x)} ,
and
(iii) {x EX: I (x) = g(x)}
are in d.
Proof. We have
{xEX: I (x) >g(x)} = u~Q({x EX: I (x) > u} n {x EX: g(x) < u}),
and from this identity the measurability of (i) follows. The set (ii) is the
complement of the set (i) with the roles of I and g interchanged, and so
it too is measurable. The set (iii) is the intersection of two measurable
sets of type (ii) , and so is measurable. D
(11.10) Theorem. Let I, g be d-measurable. Let h (x) = I (x) + g (x)
lor all x EX such that I (x) + g (x) is defined and let h have any fixed value {3
[an extended real numberJ elsewhere. Then h is d-measurable.
Proof. For any real number a, we have
h- 1 (Ja, ooJ) = {x EX: I (x) + g(x) > a} U AfJ
= {x EX: I (x) > a - g (x)} U A fJ '
!
where
({X: I (X) = oo} n {X: g(X) = - oo})
AII= u ({x: I (X) =-oo}n{X:g(X) =oo}) if a<{J,
o if a~{J.
The set {x EX: I (x) > a - g (x)} is in .91 by (11.9); therefore, since All
is also in .91, the set h- 1 (]a, 00]) is in d. 0
(11.11) Theorem. Let I and g be d-measurable lunctions. Let h be
defined on X by
_ {/(X) g(x) il x ~ A
h (x) - {J il x EA ,
where {J is an arbitrary extended real number, and A = {x EX: I (x) = 00
and g(x) =-oo}U{xEX:/(x) =-00 and g(x)=oo}. Then h is an
d-measurable lunction.
Proof. Consider a ER such that a > 0, and let
A ={A if a<{J,
II 0 if a~{J.
We have h- 1 (]a, 00]) = {x EX: h(x) > a} = All U {x EX: I(x) = 00 and
g(x) > O} U {x EX: I(x) > 0 and g(x) = oo} U {x EX: I(x) < 0 and
g(x) = -oo} U {x EX: I(x) = -00 and g(x) < O} U {x EX: I(x) and
g(x) are finite and ~ [(f(x) + g(x))11 - (f(x) - g(X)2] > a}. Applying
(11.4), (11.8) and (11.10), we see thath- 1 (]a, 00]) Ed. Similar expressions
hold for a < 0 and a = 0, and it follows that h is d-measurable. 0
We next study limits of sequences of measurable functions.
(11.12) Theorem. Let (In) be a sequence old-measurable lunctions
defined on X. Then the lour lunctions inf In, sup In' lim In' and lim In
n n "---+-00 11---+-00
[defined pointwise as in (7.1)] are alld-measurable.

Proof. It follows immediately from the identity {x EX: sup In (x) > a}
00
.
= U {x EX: In (x) > a} that sup In is d-measurable. The d-measur-
.. =1 ..
.
[recall that - ( 00) = - 00 and -
.
ability of inf In follows at once from the identity inf In (x) = -
..
sup(-In (x))
(- 00) = 00]' The rest follows from the
first two results and the identities
lim In (x) = sup ( inf In (x))
"->-00 k .. Oi:;k
and
lim I.. (x) = inf (sup I,,(x)). 0
"-+00 k .. Oi:;k
(11.13) Corollary. Let 11' ... , 1m be d-measurable. Then the lunctions

max{/l> ... ,lm} and min{/l> ... ,lm}
[defined pointwise} are d -measurable.
Proof. Define In = 1m for all n > m and apply (11.12). 0
(11.14) Corollary. II (In) is a sequence 01 d-measurable lunctions
defined on X and lim In (x) exists in R* lor all x E X, then lim In is d-
n---+oo n-+oo
measurable.
Proof. Since lim In = lim In = lim In' (11.12) applies. 0
n-+oo fl.-+-
00 ft.-+- 00
We now consider the concept of measurability for complex-valued

[finite] functions.
(11.15) Definition. A complex-valued function I defined on X is
said to be d-measurable if both Rei and Iml are d-measurable.
(11.16) Theorem. Let I be a complex-valued lunction defined on X.
Then the lollowing statements are equivalent:
(i) I is d-measurable;
(ii) 1-1 (U) Ed lor each open U C K;
(iii) 1-1 (B) Ed lor every B E fJ4 (K) .
Proof. Let 11 = Rei and 12 = Imf. Then 1= 11 + i/2. First suppose
that (i) is true and let V = {s + it E K: a < s < b, c < t < d} where
{a, b, c, d} C Q. Then 1-1 (V) = lil(]a, b[) n l;l(]c, d[) Ed. Next, let U
be any open subset of K. There exists a sequence (Vn) of rational rectangles
00
of the form V above such that U = U

,,=1
Yn. It follows that l-l(U)
00
= ..U=1 1-1 (Vn) Ed. Thus (i) implies (ii) .

Now suppose that (ii) is true and set [I' = {5 C K: /-1(5) Ed}.
As in the proof of (11.4), we see that [I' is a a-algebra of subsets of K.
Also [I' contains all open subsets of K, and so fJ4 (K) C [1'. Thus (iii)
follows and therefore (ii) implies (iii).
Finally, suppose that (iii) is true. For aER, let Al ={s+itEK:s>a}
and A 2= {s + itEK:t > a}. Then li1 (]a, 00]) = lil(]a,oo[) = 1-1 (Ai) Ed
because Ai E fJ4(K) (j = 1,2). Thus 11 and 12 are d-measurable, and so (i)
is true. 0
(11.17) Theorem. Let I and g be complex-valued, d-measurable lunc-
tions on X, let IX EK, let mEN, and let p be a positive real number. Then
+,
all 01 the lollowing lunctions are d-measurable on X: I + IX,' IX/,' ItIP;
1m; il I(x) =!= 0 lor all x EX,' I + g,' Ig.
Proof. These results all follow immediately from Definition (11.15)
by applying (11.8), (11.10), and (11.11). 0
(11.18) Theorem. Let (/..) be a sequence 01 d-measurable complex-

valued lunctions on X and suppose that lim I.. (x) = I (x) EK lor each x EX.
n-+oo
Then I is d-measurable.
Proof. Apply (11.15) and (11.14). 0
(11.19) Remark. Theorems (l1.1S) and (11.14) both require that the
sequence in question converge for every x EX. However, a large portion
of our work will deal with the case in which there is some specific meas-
ure ,." defined on d, the functions in question are defined only ,.,,-almost
everywhere, and the convergence of sequences is only ,.,,-a.e 1 • Thus
we would like a corresponding theorem for this case. Such a theorem will
require some additional hypothesis about ,.", for consider the case that
X = R, d = flI(R),,.,, = A, P = CANTOR'S ternary set, A C P, A ~ flI(R),
I = ~A' and I.. = 0 for all n EN. Then each I.. is flI(R)-measurable and I
is not flI (R)-measurable, but I.. (x) -+ I (x) for all x ERn P', i. e., I.. -+ I
A-a. e. To avoid such irritating situations, it is enough to consider complete
measures, defined as follows.
(11.20) Definition. Suppose that,." is a measure defined on d and
that BEd whenever A Ed, ,.,,(A) = 0, and Be A, i.e., all subsets of
sets of measure zero are measurable. Then ,." is said to be a complete
measure and (X, d,,.,,) is called a complete measure space.
Theorem (10.7) implies that if ,." is an outer measure on X, then
(X, -L,.,,.,,) is a complete measure space. We gain much and lose little
[as the next theorem shows] by restricting our attention to complete
measure spaces.
(11.21) Theorem. Let (X, d,,.,,) be any measure space. Define
sf = {E U A: E Ed, A C B lor some B Ed such that ,.,,(B) = O}
and define p, on d by the rule p, (E U A) = ,." (E). Then d is a a-algebra,
p, is a complete measure on d, and (X, d, p,) is a complete measure space.
This measure space is called the completion of (X, d, ,.,,) and p, is called
the completion of ,.".
Proof. Exercise.
(11.22) Definition. If E Ed and ~ = {F Ed: FeE}, then dE
is plainly a a-algebra of subsets of E and (E, dE) is a measurable space.
A function defined on E will be called d -measurable if it is drmeasur-
able.
(11.23) Theorem. Let (X, d,,.,,) be a complete measure space and let I
be an d-measurable lunction defined ,.,,-a.e. on X. Suppose that g is a
lunction defined ,.,,-a. e. on X such that I = g ,.,,-a. e. Then g is d -measurable.
1 The term "p-almost everywhere" and its abbreviation "p-a.e." were defined
in (9.29) for the case in which p is a measure , on a locally compact Hausdorfi
space. The extension to arbitrary measure spaces (X, ~,p) is immediate.
Proof. Let A = {x E (dom/) n (domg) : I (x) = g(x)}. Then #(A') = 0

and all subsets of A' are in d. We suppose that I and g are extended
real-valued, the complex case being similar. For a E R, we have
g-l(]a, 00]) = (g-l(Ja, 00]) n A) U (g-l(Ja, 00]) n A')
= (1-1 (Ja, 00]) n A) U (g-1 (Ja, 00]) n A') Ed. 0
(11.24) Theorem. Let (X, J~, #) be a complete measure space and let
(I,,) be a sequence 01 d-measurable lunctions each 01 which is defined
#-a. e. on X. Suppose that I is defined #-a. e. on X and that lim I" (x) = I (x)
,,->-00
#-a.e. on X. Then I is d-measurable.
Proof. Define A as the set
(dom/) n COl dom/,,) n {x EX: I" (x) -+ I (x)} .

It is obvious that A Ed and that # (A ') = O. For each n EN, define
() _{/,,(X) if xEA,
g" x - 0 if x EA',
and define
if x EA,
g(x) = {~(X)
if X EA'.
Theorem (11.23) implies that g" is d-measurable for all n EN. Clearly
g,,(x)-+g(x) for all xEX. Applying (11.14) or (11.18), we see that g
is d-measurable. Again by (11.23), I is d-measurable. 0
Mathematical analysis is heavily concerned with convergence of
sequences and series of functions. Indeed, a main goal of analysis is the
approximation of complicated functions by means of simple functions.
[The terms "approximation", "complicated", and "simple" have dif-
ferent meanings in different situations.] Up to this point we have met
in this book two kinds of convergence: pointwise [almost everywhere]
and uniform. We now introduce a third kind of convergence and prove
some theorems that show a number of relationships among these three
kinds of convergence.
(11.25) Definition. Let (X, d, #) be a measure space and let I and
(1,,)':=1 be d-measurable functions on X. They may be either extended
real- or complex-valued. Suppose that for every «5 > 0, we have
lim #({x EX: I/(x) -I,,{x)l ~ «5}) =
,,->-00
o.
Then (I,,) is said to converge in measure [or in probability] to I. We write:
I" -+ I in measure.
(11.26) Theorem [F. RIEsz]. Let (X, d, #) be a measure space and let I
and (I,,) be d-measurable lunctions such that I" -+ I in measure. Then there
exists a subsequence (I..~) such that I..~ -+ I #-a.e.
1
Proof. Choose nl EN such that ",({x EX: I/(x) - In, (x)l ~ I}) < 2'
Suppose that nl , n a, •.• , n,. have been chosen. Then choose nk+1 such
that nk+l> n,. and
'" ({XEX: I/(x) - Ink+! (x) I ~ k : I}) < 2kI+l •
Let A;=kQ;{x:I/{X)-ln,t(X)I;;;;; !}for each jEN. Clearly we have

00 00 1
Al ::> Aa ::> .• '. Next let B = .n Ai'
1=1
Since '" (AI) < .E 2k <
k=1
00, it
follows from (10.15) that
",(B) = .lim ",{Ai) .lim £ 2\ = ~im 2~1 = 0,
~ 1->-00
1->-00 k=; 1->-00
00
that is, ",(B) = O. Next, let xE B' =.U Ai. Then there is a jx such that
1=1
kO;.,{y EX: I/(y) - In.t{y)i < !} .

+
x EAi., =
Given e > 0, choose ko such that ko ;;;;; jx and

o
~ e. Then k;;;;; ko
implies that lI(x) - In,t(x)i < ! ~ e. This proves that In,t(x) -+ I (x)
for all x EB'. 0
(11.27) Note. There exist sequences of functions that converge in
measure and do not converge a.e. For example, let X = [0, IJ, d =..AA,
'" = A, and, for each n EN, define
1.. = g[-.L ;+1] where n=2"+j,0~j<2".

2k' 2k
Thus 11 = ~[O, I), 12 = g[0, 2.],

2
... ,/10 = g[2.4' ~],
8
... ; It is clear that
1
A({X: II.. (x) I ;;;;; t5}) as n -+ 00 for every t5 > O. Thus I.. -+ 0
~ 2k -+ 0
in measure. On the other hand, if x E [0, IJ, the sequence ({.. (x)) contains
an infinite number of O's and an infinite number of I's. Thus the se-
quence of functions (In) converges nowhere on [0, IJ.
(11.28) Exercise. Find a subsequence of the sequence (I..) of (11.27)
that converges to zero A-a.e. on [0, IJ. Can you find a subsequence
(Ink) of (/..) that converges to zero everywhere on [0, IJ?
(11.29) Note. We have already seen several instances in which
finiteness or a-finiteness of a measure space is an essential hypothesis:
d. for example (10.15), (10.30), (10.34), and (10.58). Our next two
theorems are stated for finite measure spaces.
(11.30) Lemma. Let (X, d, "') be a finite measure space and let I
and (I..) be d-measurable lunctions that are defined and finite ",-a.e. on X.
Suppose that I.. -+ I ",-a. e. on X. Then lor each pair 01 positive real numbers
15 and e, there exist a set lEd and an integer no E N such that p (]/) < e
and II (x) - I" (x)l < 15 lor aU x EI and n ~ no.
Proof. Let E = {x EX: I (x) is finite, I" (x) is finite for all n EN,
I" (x) --+-1 (x)}. By hypothesis, p (E') = O. For each mEN, let Em = {x EE:
ex)
I/(x) - 1,,(x)1 < 15 for alln ~ m}. WehaveE1 C E2 C," and m=!
U Em=E.
00
Therefore E; :> E~ :> ... and n E:" = E'. Since p(E;) ~ p(X) < 00, it
m=!
follows that lim p (E:") = p (E') = O. Thus choose no EN such that
~oo
p (E:") < e and set I = E.... D

(11.31) Theorem [LEBESGUE]. Let (X, d, p), I, and (/,,) be as in
(11.30). Then I" --+- I in measure.
Proof. Choose arbitrary positive numbers 15 and e. For each n EN,
let S,,(t5) = {x EX: I/(x) - 1,,(x)1 ~ t5}. By (11.30), there exist I Ed
and no EN such that p (]') < e and II (x) - I" (x) I < 15 for all x EI
and for all n ~ no' Thus n ~ no implies S,,(t5) C J'. Therefore n ~ no
implies p(S,,(t5)) ~ p(]') < e. Since e is arbitrary, it follows that
lim p(S" (15)) = 0, i.e., I" --+- I in measure. D
...... 00
(11.32) Theorem [EGOROV]. Let (X, d, p), I, and (I,,) be as in (11.30).

Then lor each e > 0 there exists a set A Ed such that p (A ') < e and
I" --+- I unilormly on A.
Proof. Choose a positive number e. By (11.30), for each mEN there
exist 1m Ed and nm EN such that p (]:,.) < 2: and II (x) - I" (x) I < ~
ex) ex)
forallxE1mandalln~nm.DefineAbyA= n 1m. ThenA ' =

m=!
U
m=!
I:",
and so
00
p(A') ~ ~ p(]:") <~

m=! m=!
ex)
2: = e.
Also n ~ nm implies that
sup lI(x) - 1,,(x)1 ~ sup I/(x) - 1,,(x)1
xEA xE./...
:5:~
- m
for every mEN. Thus I" --+- I uniformly on A. D
(11.33) Caution. Theorems (11.31) and (11.32) depend heavily on
the hypothesis that p(X) < 00. For example, suppose that X = R,
d = u8J., P = A, I" = ~[".n+!l' and 1=0. Then I,,(x) --+- I (x) = 0 for all
x ER. But A({X E R: I/(x) - 1,,(x)1 ~ I}) = A([n, n + 1]) = 1-++0, and
so I" -++ I in measure. Also, if A Eu8J. and A(A/) < 1, then for each nEN
there exists x" E An en, n + 1], and so II (x,,) - I,,(x,,) I = 1, i.e., I" -++ I
uniformly on A.
The remainder of this section is devoted to an investigation of the

structure of measurable functions. As usual, (X, d) is an arbitrary
measurable space.
(11.34) Definition. A simple lunction s on X is a function that
assumes only a finite number of values. If mg s = {1X1' ••• , 1Xn} and
,.
A~ = {x EX: s(x) = IX~} (1 ~ k ~ n), then it is obvious that s = ~IX~EA'"
11=1
(11.35) Theorem. Let I be any complex- or extended real-valued
d-measurable lunction defined on X. Then there exists a sequence (s,,)
01 finite-valued, d-measurable, simple lunctions defined on X such that
IS11 ~ IS21 ~ ••• ~ Is,,1 ~ ••• and s" (x) -'>- I (x) lor each x EX. II I is
bounded, then the lunctions s.. can be chosen so that the convergence is uni-
lorm. II I ~ 0, the sequence (s,,) can be chosen so that 0 ~ Sl ~ S2 ~ ••• ~ I.
Proof. First consider the case I ~ O. For each n EN and for
1 ~ k ~ n' 2", let
A",~={xEX:k 2,,1 ~/(x)<:,,},
and
B" = {x EX: I (x) ~ n}.
Define
",2"
s,,(x) = ~
11=1
k;l EA".,,(X) + nEB,,(x) .
It is clear that all of the sets A", ~ and B" are in d, and so each s" is
an d-measurable simple function. It is also easy to see that
1
o ~ Sl ~ S2 ~ ••• ~ I, Is,,1 ~ n, and I/(x) - s.. (x)1 < 2ft
for all x E B~. It follows that lim s" (x)
n-o-oo
= I (x) for every x EX. Moreover,
if there exists {J ER such that III ~ {J, then sup II (x) - s" (x) I ~ ; ..
"EX
for all n ~ {J; therefore s" -'>- I uniformly on X if I is bounded.
Now consider the general case. If I is extended real-valued, define
r = max{/, O}andr = -min{t, O}. Thenr ~ O,r
~ 0, and I =1+ -r.
If I is complex-valued, we may write
1= /1 - 12 + i(/s - I,)
where each Ii ~ O. In either of these cases we apply the results of the
preceding paragraph to the nonnegative extended real-valued functions
making up I. We leave the details to the reader. 0
(11.36) Theorem [N. N. LUZIN]. Let X be a locally compact Haus-
dorff space and let, and .A; be as in §§ 9, to. Suppose that E E.A;, t (E) < 00,
and that I is a complex-valued .A;-measurable lunction on X such that
I (x) = 0 lor all x EE'. Then lor each e > 0 there exists a lunction g E(ioo (X)
such that t({x EX: I (x) =l= g(x)}) < B. Moreover il 11/11 .. < 00, then g can
be selected so that Ilgll .. ~ 11/11 ...
Proof. Let B > 0 be given. Use (9.24) to select an open set U such that
E e U and t(U) < teE) + : . The set U is fixed throughout the proof.
(I) Suppose that I = ~.A' Then AcE and A is t-measurable. Since
teA) < 00, we may apply (10.30) to find a compact set F such that
Fe A and teA n F') < ; . Next use (9.24) and (6.79) to produce an
open set V with compact closure such that Fe Ve U andt(VnF') < ~ .
Use (6.80) to obtain a continuous function g from X into [0, 1] such that
g(x) = 1 for all x EF and g (x) = 0 for all x E V'. Then we have: gE<roo(X),
g = 0 on U' , and {x EX: I (x) =l= g(x)} e (V n F') U (A n F'). It follows
that
t({x EX: I (x) =l= g(x)}) < ; + ; = B,
and so the proof is complete if I = ~.A'

"
(II) Consider next the case that I is a simple function, say I = }; a.,. ~.A"
"=1
where each Ak is t-measurable. We may [and do] suppose that Ak e E
for each k. Next apply (I) to find functions gk E <roo (X) such that gk = 0
on U' and t({x EX: ~.Al: (x) =l= gk (x)}) < = for 1 ~ k ~ n. Define
"
g=};a.kgk' Then we have: gE<roo(X); g=O on U' ; and {xEX:
"=1
I (x) =l= g (x)} e kldl" {x EX: ~.A" (x) =l= gk (x)}. Thus t({x EX: I (x) =l= g (x)})
" 8
<};-=B.
"=1 n
(III) Now consider the general case. Apply (11.35) to obtain a se-
quence (S,,):=I of ..4f.-measurable, complex-valued, simple functions such
that s" = 0 on E' for each n EN and s" (x) -+ I (x) for all x EX. For each
n EN, apply (II) to obtain a function g" E <roo (X) such that g" = 0 on
co
U' and if A" = {x EX: s" (x) =l= g" (x)}, then t (A,,) < ; ... Let A = "ld A". l
It is clear that A e U and that
Clearly s" (x) = g" (x) for all x EA I and all n EN. Thus g" (x) -+ I (x)
for all x EA I. Since t (U n A') < 00, EGOROV'S theorem (11.32) shows
that there is an t-measurable set B e UnA I such that t (U n A 'n B') < :
and g" -+ I uniformly on B. Next use (10.30) to select a compact set
FeB such that t (B n F') <

: . Plainly gn ~ I uniformly on F U U'
[recall that gn = 1=0 on U']. Since each gn is continuous, it follows
from (7.9) that I is continuous on F U U' [in the relative topologyJ.
Applying the TIETZE extension theorem (7.40), for the compact set F
and its open superset U, we find a function g E<roo (X) such that I (x) = g (x)
for all x EF U U'. We have
{x EX: I (x) ,*,g(x)} C U nF' = AU (U n A' n B') U (B nF') ,
and so
t({x EX: I (x) '*' g(x)}) ~ t(A) + t(U n A' n B') + t{B n F')
e e e
<4+4+4<8.
Finally, suppose that 11/11" < 00 and Ilgll" > 11/11". This is certainly
possible. In this case we tamper a little with g to obtain the desired con-
clusion. Let S = {z EK : Izl ~ II/II,,} and T = {z EK: Izl ~ Ilgll,,}. De-
fine a mapping rp from Tonto S as follows:
z if zES
{
rp (z) = Z
~. 11/11'f
u 1 zETn s' .
It is easy to see that rp is continuous and that Irp (z)l = 11/11" for z E T n S'.
Now let h = rp 0 g. Then h is continuous and h(x) = 0 whenever g(x) = 0;
thus hE <roo (X). Also it is evident that Ilhll" = 11/11" and {x EX: I (x)
'*'h(x)} C {x EX: I (x) g(x)}. D '*'
(11.37) Exercise: Measures on measurable subsets. Let (X, ~,p,) be
a measure space and E a set in d. Let ~ be as in (11.22), and let P,E
be the restriction of p, to dE' Prove that (E, dE, P,E) is a measure space.
(11.38) Exercise: Measures on image sets. (a) Let (X, d, p,) be a
measure space and l' a mapping of X onto a set Y. Let P4 be the family
of all subsets B of Y such that 1'-1 (B) E d. For B E P4, let
lI(B) = p,(1'- 1 (B). Prove that (Y, P4, 11) is a measure space.
(b) State and prove an analogue of (a) for outer measures.
(11.39) Exercise: Measures on nonmeasurable subsets. (a) Let (X, d,p,)
be a measure space and let Y be a subset of X such that if BeY'
and BEd, then p, (B) = O. Let d t be the family of all sets Y n M for
M Ed. For Mt Edt, define p,t(Mt) as p,(M) for an arbitrary M Ed
such that Y n M = Mt. Prove that p,t is well defined, that .r;/t is a
a-algebra of subsets of Y, and that (Y, d t , p,t) is a measure space.
(b) Consider the measure space (R,.A)" A). Find a subset Y of R
that is not A-measurable and which satisfies the condition of part (a).
[Hint. Use a set B as constructed in (1O.54.b).J
(c) Prove that [0, IJ contains a subset D admitting a measure p,
on P4 (D) such that p, (D) = 1, p, (F) = 0 for all compact sets FeD,
and Jl (Y) = inf {Jl (U) : U is open in the topology of D, U ~ Y} for all
Y Ef!l(D). That is, CD, f!l(D), Jl) is irregular, although "outer" regular.
n
[Hints. Construct a subset D of [0, 1] such that D F =1= 0 =1= D' F n
for every uncountable closed set Fe [0, 1]: see (1O.54.b). Then D is
°
non A-measurable, and so < A(D) ~ 1. Let X be a G6 set such that
Dc X c [0, 1] and A(X) = A(D). Then every A-measurable subset
of X n D' has A-measure 0, and we may take Jl to be ().(~) Af as in
part (a). All of the claims made for Jl are easy to verify. Compare this
result with (10.58).]
(11.40) Exercise: Extending a measure. Let (X, .91, Jl) be a measure
space. Let Y be a subfamily of f!P (X) such that:
(i) P E Y n .91 implies Jl(P) = 0;
00
(ii) PI> P 2 , P a, .•• E Y imply U P n E Y.

n=l
Let .91* be the family of all subsets A * of X such that the symmetric
difference A * L, A is in Y for some A Ed. For such a set A *, let
Jl*(A*) = Jl(A).
(a) Prove that .91* is a a-algebra of subsets of X.
(b) Prove that Jl* is well defined on .91* and that (X, .91*, Jl*) is a
measure space.
(c) In what sense is (X, .91*, Jl*) an extension of (X, .91, Jl) ?
(d) Find a simple hypothesis on Y necessary and sufficient for
(X, .91*, Jl*) to be a complete measure space. Prove your assertion!
(11.41) Exercise. Let (X, .91, Jl) be a finite measure space. Let
(In)~l be a sequence of extended real-valued, d-measurable functions
on X. Suppose that lim In (x) = I (x) for Jl-almost all x EX, where I
n~oo
is extended real-valued and .91-measurable. Define arctan ( 00) = ~

and arctan (- 00) = - ~, and consider the functions arctan 0 In and
arctan 0 I. Now formulate and prove an analogue of EGOROV'S theorem
(11.32) on uniform convergence of In to I except on sets of arbitrarily
small measure.
(11.42) Exercise. Let (X, .91, Jl) be a a-finite measure space and let I
and (In) be d-measurable complex-valued functions that are defined
Jl-a.e. on X. Suppose that In -+ I Jl-a.e. on X. Prove that there exists a
00
set H Ed and a family {E k }k=l c d such that X = H U k~l Ek ,

Jl (H) = 0, and In -+ I uniformly on each Ek • [Use EGOROV'S theorem
(11.32).]
(11.43) Exercise. (a) Find a sequence (In) C <r'([O, 1]) and a real-
valued function I such that In (x) -+ I (x) for all x E [0, 1] but In -+ I
uniformly on no subinterval of [0, 1]. [Make I discontinuous on a

dense set.]
(b) Use the sequence constructed in (a) to show that the conclusion
of (11.42) fails for the measure space ([0,1], .9'([0,1]),,,) where ".is
counting measure on [0, 1] (lOA.a). [Show that the E,,'s can be taken
closed and apply BAIRE'S category theorem (6.54).]
(c) Show that there is a sequence (In) C <r'([O, 1]) such that In (x) -+ 0
for every x E [0, 1] but In -+ 0 uniformly on no subinterval of [0, 1].
[For each n EN, let Fn be the set of all numbers in ]0, 1] having the form
:m for an integer k and an integer mE {O, 1, ... , n}. Let In be zero on Fn.
For :m EFn, where k is odd, let In( :m - 2Ll) = 21m. Let In be linear
in all subintervals of [0, 1] where it is not yet defined.]
(11.44) Exercise. Let X be a locally compact Hausdorff space and
let £ be a measure on X as in § 9. Suppose that I is a complex-valued
..A;-measurable function on X such that {x EX: I (x) =!= O} is a-finite
with respect to t. Prove that there exists a Borel measurable function
g on X such that Igl ~ III and £({x EX: I(x) =!= g(x)}) = O. [Use (11.35)
and (10.34).]
(11.45) Exercise. Let (X,.9/, 1-') be a finite measure space. Suppose
that I and (In)':=1 are .9/-measurable complex-valued functions on X.
Prove that In -+ I in measure if and only if each subsequence of (In)
admits a subsubsequence that converges to I I-'-a. e.
(11.46) Exercise. Let X be a topological space. A family (f of com-
plex-valued functions on X is said to be closed under pointwise limits if
I E(f whenever I is a complex-valued function on X and, for some
sequence (In) C (f, I (x) = lim In (x) for all x EX. The family ~ (X) of
11->00
all Baire lunctions on X is defined to be the intersection of all families (f

of complex-valued functions on X such that (f contains all complex-
valued continuous functions on X and (f is closed under pointwise
limits. Notice that KX is such a class (f.
(a) Prove that every Baire function is Borel measurable.
Let ~o be the set of all complex-valued continuous functions on X.
If a is an ordinal number such that 0 < a < D [see (4049)], define ~«
to be the family of all functions I such that I is the pointwise limit of some
sequence (In) C U {~I1: {3 is an ordinal number, {3 < a}. The functions
in ~ are known as the Baire lunctions 01 type a.
(b) Prove that ~(X) = U ~,,[compare the proof of (10.23)].
«<D
(c) Prove that if I and g are Baire functions, then so are 1+ g, Ig,
and III; if I and g are real-valued Baire functions, then so are max{/, g}
and min{t, g} [use (b) and transfinite induction].
11*
Let 8ito(X) be the smallest a-algebra of subsets of X that contains all

sets of the form {x EX: I (x) = O}, where I is a continuous complex-valued
function on X. The sets in 8ito(X) are called the Baire sets 01 X.
(d) Prove that I E ~(X) if and only if I is a complex-valued function
on X and I is 8ito(X)-measurable. [For the "if" statement, first show that
{E eX: ~E E ~(X)} = 8ito(X), and then use (11.35). For the "only if"
statement use (b) and transfinite induction.J
(c) Prove that 8ito(X) = 8it(X) if X is a metric space [use (6.86)J.
§ 12. The abstract Lebesgue integral

This is perhaps the most important single section in the entire
book. In it we construct, and study the remarkable properties of, the
Lebesgue integral on an arbitrary measure space. It turns out that this
r
integral is equal to the functional for all nonnegative measurable func-
tions when the measure space is (X, vIt., l). Throughout the present
section, (X, d, fl) denotes an arbitrary measure space, except where
e
further restrictions are explicitly stated. The symbol denotes all simple,
d-measurable functions on X that are complex- or extended real-valued;
e+ is as usual the set of all nonnegative functions in e.
(12.1) Definition. A measurable dissection 01 X is any finite, pairwise
n
disjoint family {AI> A 2 , ••• , U Ak = X.
An} C d such that k~l
(12.2) Definition. Let I be any function from X into [0, 00]' Define
L (I) = sup t~ inf{/ex) : x E A k} fl (Ak) : {Al' ... , An}

is a measurable dissection of X} .
Since one or more Ak's may be 0, we must define inf 0: as a matter of
convenience we set inf 0 = 0.
For an extended real-valued function I we define [just as in (11.35)J
1+ = max{/, O} and r = -min{/, O}.
Notice that j+ ~ 0, r~ 0, and I = j+ - r.
We define L (f) = LW)-L(n
provided that at least one of the numbers L W) and L (n is finite. If
LW) = L(n = 00, then we do not define L(f). The number L(I) [when
definedJ is called the Lebesgue integral [or simply the integralJ 01 f.
(12.3) Examples.
(a) If <X E R# and I (x) = <X for all x E X, then L (f) = <Xfl (X).
(b) If I (x) = 00 for x E E and fleE) > 0, then L (I) = 00 if it is defined.
(c) If X = [0, IJ, fl = A, and I is nonnegative and Riemann integrable
on [0, IJ, then L(f) ~ S(/; [0, IJ). This inequality is all but trivial: each
lower Darboux sum for I is less than or equal to one of the numbers of
§ 12. The abstract Lebesgue integral 165
which L (I) is the supremum. [Actually the equality L (I) = S (I; [0, 1])
holds: see (12.51.£) inlra.]
n
(12.4) Theorem. Let IE 16+, say I= };a.k~Ek' where the Ek'S are
k=1
n
pairwise disjoint and in .!if. Then L (I) exists and L (I) = } ; a.k '" (E k) .
k=1
Proof. We may suppose that the pairwise disjoint family {E k }%=1
n
covers X. Since inf{/(x): x EE k} = a.k, we have L(I) ~ };a.k",(Ek).
k=1
Now let {Bi}~1 be an arbitrary measurable dissection of X. Then
m m n
}; inf{/(x) : x EB i } ",(Bi ) = }; };inf{/(x): x EB i } ",(E k n B i )
;=1 i=1 k=1
m n
~}; };inf{/(x):xEEknBi}",(EknBi )
i=1 k=1
m n
= } ; } ; a.k ",(E k n B i)
i=1 k=1
n
= }; a." '" (E k ) •
k=1
Thus we obtain L(I) ~

"
};a.k",(E k) and hence L(/) =
"
};a."",(E k). 0
k=1 k=1
(12.5) Theorem. Let I and g be any nonnegative functions on X.
II I (x)~ g(x) lor all x EX, then L (I) ~ L (g).
Proof. Trivial.
(12.6) Theorem. Let I be a nonnegative measurable function. II
",({x EX: t (x) > O}) is positive, then L (I) is positive.
Proof. We will find a set A E.!if and a positive number a. such that
'" (A) > 0 and I (x) ~ a. for all x EA. It will then follow that
L(I) ~ inf{t(x): x EA} ",(A) + inf{t(x): x EA'} ",(A')

~a.",(A»O.
For each positive integer n, let An = {x EX: I (x) ~ :}. We have

00
A1 CA 2 C"·CA n C .. · and ..U

=1
An={xEX:/(x»O}. By (10.13),
we have ~~ ",(An) = "'("~IAn) > O. Hence there is a positive integer

no for which ",(An)o > 0 and I (x) ~ ~on
no
An. 0 0
(12.7) Theorem. Let I and g be in 16+. Then we have L (f + g)

= L(f) + L(g).
m m"
1=1
ft
Proof. Write/= I: ex;~AJandg= k=1

;=1
I: Pk~Bk' where.U Ai= k=1
U Bk=X'
... "
Then 1+ g = I: I: (exi + Pk) ~(AJnBl:)' Thus by (12.4) we have
;=lk=1
... "
L (I + g) = I: I: (ex; + Pk) ft (Ai n B lI )
;=1 k=1
m ,.. II m
= I: I: ex; ft (Ai n Bk) + I: I: Pk ft (Ai n B lI )
~1~1 ~1~1
= ;~ ex; ft (Ai n tQ1 B k)) \~ Pk ft (Bk n C91 Ai))

m "
= I: ex; ft (A;) + I: Pk ft (Bk)
;=1 k=1
= L (I) + L (g). 0
(12.8) Theorem. II I;;;; 0 and t ER, then we have L(tl) = tL(I).
The proof is easy and is omitted.
Our immediate aim is to establish the extremely important identity
for all sequences (I,,) of nonnegative d"-measurable functions. We

begin with a lemma.
(12.9) Lemma. Let I be any extended real-valued lunction on X and
suppose that E = {x EX: I(x) =1= O} is an d"-measurable set. Let d"E
and ftE be as in (11.22) and (11.37). Then, il L (I) exists, we haveL (I) =LE(f),
where LE is the integral lor the measure space (E, dE' ftE)'
Proof. First suppose that I ;;;; 0 and let y be any real number such
that y < L (I). There exists a measurable dissection {AI' ... , A ...} of X
satisfying the inequality
m
y< I: inf{/(x): x EA k} ft(AlI)'
k=1
Using the fact that E is d"-measurable, we have

... m
y< I: inf{/(x): x EA k} ft{Ak n E) + I: inf{/(x): x EAll} peAk n E')
k=1 k=1
m
~ I: inf{/(x): x EAk n E} p(AlI n E)
k=1
~ LE(I) .
[Note that Ak n E' =1= 0 implies inf{/(x) : x EA",} = O.J It follows that
L(I) ~ LE(I)·
Next suppose that I' < LE(f), and let {B k }Z'=l be any measurable
dissection of E such that
m
I' < J; inf{t(x): x EB k } p.(B k ) •
k=l
We have
m
I' < J; inf{t(x): x EB k } p.(B k ) + o· p.(E')
~ L(t) ,
k=l
and so LE (I) ~ L (I). The assertion for arbitrary functions follows im-
mediately. 0
The following result, which looks harmless enough, is the key to
the proof that L is countably additive.
(12.10) Theorem. Let (g,,):=l be any nondecreasing sequence in e+.
5uppose that h E e+ and that lim g" ~ h. Then we have lim L (g,,) ~ L (h).
~oo n~oo
Proof. The theorem is trivial if p. (X) = 0, and so we suppose throughout

that 0 < p.(X) ~ 00. Let h = ')I1~El + 1'2~EI + ... + I'm~Em' where the
m
ETc's are pairwise disjoint, X = k~l ETc, and 0 ~ ')11 < ')12 < ... < ')1m ~ 00.
Suppose that 1'1 = 0; then by (12.9) we have L (h) = LE' (h). Supposing
that the theorem is established for the case 1'1 > 0 and ietting E~ take
the r6le of X, we have
L (h) = L E , (h) ~ lim L E , (g,,) ~ lim L (g,,) .
I n~oo 1 n~oo
It thus suffices to prove the theorem under the assumption that 1'1> o.
Case (I): p. (X) and I'm are finite. For any () > 0, choose e > 0 satisfy-
ing the inequality
e ~ min {2.u~X) , 1'1} .
For every positive integer n, let 5" = {x EX: g,,(x) > h(x) - e}. Since
00
lim g.. ~ h, we have X = U 5... The sequence (g,,) is nondecreasing,
~oo "=1
and so the sequence 51> 52' ... is nondecreasing. From these facts and
the countable additivity of p., we find (10.13) that
lim p. (5..) = p. (X)
,,-+00
and that
lim p.(5~) = 0. 1
n-+OO
We also have L(g..) ~ L(g,.~s,.) ~ L«(h - e)~s.J = L(h~s,J - eL(~HJ

The relations h = h~s.. + h~s~ ~ h~s" + ')Im~s~ imply that L (h) ~
1 Countable additivity of.u is used in Case (1) only to establish this relation.
However, countable additivity is essential: the theorem fails for p.'s that are finitely
but not countably additive.
L(h~s,,) + I'ml-'(S~), Hence

L (gn) ?;, L (h) - I'm I-' (S~) - e I-' (Sn)
?;, L (h) - I'm I-' (S~) - e I-' (X)
?;, L (h) - I'ml-' (S~) - ~ b.

If n is so large that I'm I-' (S~) < ;, then we have L (gn) > L (h) - b.
The inequality lim L (gn) ?;, L (h) follows, as b is arbitrary.
n-->-oo
Case (II): I'm is finite and I-' (X) = 00. We plainly have L (h)
?;, I'll-' (X) = 00. Let e be any number such that 0 < e < 1'1> and define
Sn for n EN just as in Case (I). For x E Sn' we have gn(x) > h(x) - e
?;, 1'1 - e. Therefore the relations L(gn) ?;, L(gn~S,.) ?;, L«(I'l - e)~s,,)
= (1'1 - e) I-' (Sn) obtain, and (10.13) implies that
lim L(gn)?;, (1'1 - e) lim I-'(Sn) = (1'1 - e) (00) = 00 =L(h).

n---+oo n-+oo
Case '(III): I-' (Em) is positive and I'm = 00. Here we have L (h)
?;, I'm I-' (Em) = 00. Choose any real number I' > I'm-I, and let
h" = I'l~El + ... + I'm-I~Em-l + I'~Em' By Cases (I) and (II), we have
lim L (gn) ?;, L (h,,) ?;, 1'1-' (Em). Since I' can be arbitrarily large, it follows
n-->-oo
that lim L (gn) = 00 = L (h).
n-->-oo
m-I
Case (IV): I'm = 00 and I-'(Em) = O. Here we have L (h) = L: I'jl-' (E j).
i=1
Let B = E1 U E2 U ... U Em_I' Then
m-I
gn ?;, gn~B and lim gn~B ?;, h~B =
n-)-oo
L:
1=1
I'j~Ej'
Since I'm-I < 00, Case (I) or Case (II) applies to (gn~B) and h~B' so that
lim L(gn) ?;, lim L(gn~B) ?;, L(h~B) = L(h). D
n-+oo n-?OO
(12.11) Theorem. Let (gn) be a nondecreasing sequence oj junctions in

e+. Then we have limL(gn) = L (lim gn).
~oo ~oo
Proof. Let lim gn = cp, and let I' be any real number such that
n->oo
I' < L (cp). There exists a measurable dissection {AI' A 2 , ••• , Am} of X
such that
I' <E. inf{cp(x) : x E A k} I-'(Ak) = L C~ O!k~Ak) ~ !~n;, L(gn) ,

where O!k = inf{cp(x): x E A k}. Here we have used (12.10). Since I' is
arbitrary, we infer that L (cp) ~ lim L (gn). The reverse inequality is
n->-oo
immediate. D
(12.12) Theorem. Let I and g be nonnegative d-measurable lunctions.

Then
L(f + g) = L(I) + L(g) .
Proof. Let (In) and (gn) be nondecreasing sequences of nonnegative
simple functions with limits I and g respectively (11.35). The sequence
(In + gn) increases to 1+ g, and so by (12.11) and the additivity of Lon
nonnegative simple functions (12.7), we have
L (I) + L (g) = lim L (In)
~oo
+ lim L (gn) = n--+ooo
~oo
lim L (In + gn) = L (I + g). 0
(12.13) Theorem. Let I be a nonnegative d-measurable lunction on X.

Then L (I) = 0 il and only il I = 0 p-a. e.
Proof. Let E = {x EX: I (x) > o}. If L (I) = 0, then it follows from
(12.6) that p (E) = 0, i. e., 1= 0 a. e. Conversely, suppose that p (E) = O.
Then
o ~ L (I) ~ L ( 00 • ~E) = 00 • P (E) = O. 0
(12.14) Theorem. Let I and g be d-measurable, extended real-valued
lunctions on X such that I = g p-a. e. and L (I) is defined. Then L (g) is
defined and L (g) = L (I).
Proof. Let E = {x EX: I(x) =1= g(x)}. By hypothesis p (E) = o.
Case (I): I ~ 0, g ~ O. Apply (12.12) and (12.13) to obtain
L(I) = L (I~E) + L(I~E') = L (I~E') = L(g~E') = L(g~E) + L(g~E') = L(g).
Case (II): general case. For x E X such that I(x) = g(x), we have
I+(x) = max{/(x), O} = max{g(x), O} = g+(x) andg-(x) = -min{g(x),O}
= -min{/(x)' O} = rex).
Therefore
{x EX: r (x) =1= g+ (x)} U {x EX: r (x) =1= g- (x)} C E ,
and so 1+ = g+ a. e. and r g- a. e. Applying Case (I) twice, we conclude
=
that L (I) = L (1+) - L (1-) = L (g+) - L (g-) = L (g). 0
(12.15) Theorem. Let I be an extended real-valued, d-measurable
lunction defined on X and suppose that L (I) is defined and finite. Then
p({x EX: I(x) = ± oo}) = 0, i.e., I is finite p-a.e.
Proof. Let A={xEX:/(x)=oo} and B={xEX:/(x)=-oo}.
By the definition of L we have
00 • p(A) + inf{/+(x) : x E A'} p(A') ~ L(/+) < 00
and
00 • pCB) + inf{/-(x) : x E B'} pCB') ~ L(n < 00.
It follows that p (A) p (B) = o. 0

=
(12.16) Remarks. Let I be an extended real-valued, d-measurable
function defined on X, let E be any set in d, and let a: be any extended
real number. Let 11 be the function on X such that

IX if xEE,
(i) 11 (x) =
{
I (x) if x EE'.
It is obvious from (11.2) that 11 is d-measurable. If p. (E) = 0 and L (I)

is defined, then (12.14) shows that L(/1) is defined and that L(/1) = L(I).
If L (I) is finite, we use (12.15) and the value IX = 0 [say] in (i) to replace I
by a finite-valued function 11 equal to I a.e. and having the same integral
as I. Thus we lose nothing in dealing with d -measurable functions having
finite integrals if we suppose that these functions are finite-valued. It is
also convenient at times to consider functions defined only almost
everywhere. The definition follows.
(12.17) Definition. Let E be a set in d such that p. (E') = O. Let
dE be as in (11.37), and let Ibe an dE-measurable, extended real-valued
function defined on E. Let 10 be any extended real-valued, d-measurable
function on X such that 10 (x) = I (x) for x EE [e.g., 10 (x) = 0 for x EE'].
Let L (I) = L (10) if L (fo) is defined, and leave L (I) undefined if L (10)
is undefined. [It is immediate from (12.14) that L (I) is uniquely deter-
mined by the definition just given.] We shall frequently in the sequel
encounter functions that are defined only on sets E as above and are
dE-measurable. To avoid tedious repetition, we shall call such functions
d'-measurable, although this is not really correct, and we will whenever
convenient think of these functions as being extended over all of X so
as to be d -measurable.
We now introduce a very important space of functions.
(12.18) Definition. We define ~~ (X, d, p.) as the set of all d-
measurable real-valued functions 1 defined p.-a. e. on X such that L (I)
exists and is finite. Where confusion seems impossible, we will write
~r for ~~ (X, d, p.).
The functional L is ordinarily written in integral notation:
J 1(x) dp. (x) = J 1(t) dp. (t) = J 1dp. = J 1dp. .
L (I) =
x x x
We will adopt this notation in dealing with d-measurable functions.
b b
In case X = [a, b] and p. = A, we write J I (x) dx, J I(t) dt, etc., for
II II
00 00 b
J I dA. The notations J f(x) dx, J I(x) dx, and J I(x) dx are self-
~~ -00.. -00
explanatory.
(12.19) Theorem. Let I E ~~, and let 1=11 - 12' where 11 ~ 0, I. ~ 0
and II' I. E ~~. Then
J I dp. = J 11 dp. - fl. dp. .
x x x
Proof. By definition, we have J I d p = J 1+ dp - J r dp. Since

11 - 12 = r - r, we have
x
r
x x
11 + r = + 12' From this equality and
(12.12) we infer that
x x x
r
J 11 dp + J dp = J 1+ dp + J 12 dp. 0
x
(12.20) Theorem. For I, g E ~~ and IX, fJ E R, we have
J (IXI + fJg) IX J I dp + fJ J g dp .
dp =
x x x
That is, the mapping I--+- J I dp is a linear lunctional on ~~.
x
(12.21) Theorem [LEBESGUE]. Let (In) be a sequence 01 nonnegative,
extended real-valued, d-measurable lunctions on X. Then
f(~ Ii) dp = ;gj Ijdp.

x
Proof. For every positive integer m, we have
1: Ii ~ 1: Ii;
00 '"
;=1 ;=1
therefore
f(£li)d PG f(#/i)dP=;~llidP'
x x
and consequently
x
f ( i Ii) dp il
~ j Ii dp .
For every positive integer n, let (S~I<»)k=1 be a nondecreasing sequence

of functions in e+ with limit In. For kEN, let gk = sil<) + 41<) + ... + s~").
The sequence (gk)k=1 is obviously nondecreasing. If m ~ k, then we have
00
sil<) + s~I<) + ... + s~) ~ gk ~ 11 + ... + It ~ 1: Ii •

;=1
Taking the limit with respect to k, we find that
00
11 + 12 + ... + 1m ~ lim gt ~
/1-+00
1: Ii
;=1
for each m. Taking the limit with respect to m, we obtain
Now (12.11) implies that
f( f i= I
Ii) dp = lim J gk dp;;;;
k--->oo X
lim
k--->oo X
J (11 + 12 + ... + Ik) dp
X
k 00
= lim I: J Ii dp = I: J Ii dp. 0
k--->ooi=IX i=IX
(12.22) B. LEVI'S Theorem. Let (lk)k=1 be a nondecreasing sequence 01
extended real-valued, d-measurable lunctions on X such that J fk dp < 00
x
lor some k. Then
Proof. We may suppose with no loss of generality that J I~ dp < 00,

x
and in view of (12.16) that no Ik assumes the value - 00. If any J Ik dp
x
is equal to 00, the result is trivial. Otherwise, for kEN, we define
gk
(x) = {/Hdotherwise.
00
Ik(x)
X) - if Ik(x) < 00,
Then we have ~~ In = ~~ (11 + ~I gk) = 11 + k~ gk' and so by

(12.21) and (12.19),
00
J(lim In) dp = J 11 dp + I: (J IHI dp - J Ik dp) = lim J In dp. 0

x ~oo X k=1 X X ~oo X
(1~.23) FATOU'S Lemma. Let (In)':=1 be a sequence 01 nonnegative,

extended real-valued, d-measurable lunctions on X. Then
J lim
X tt-+oo
In dp;;;; lim
~oo X
J In dp .
Proof. For every positive integer k, let gk = inf{/k' IHI' ...}. Plainly
gk is d'-measurable, (gk) is nondecreasing, and gk ;;;; Ik' The hypotheses
of (12.22) obtain for (gk)k=l, and so we have
J lim In dp = J lim gk dp = lim J gk dp;;;; lim J In dp. 0
X "--"00 X k--->oo k--->oo X ~oo X
(12.24) LEBESGUE'S Dominated Convergence Theorem. Let (In)':=1 be a

sequence 01 extended real-valued, d-measurable lunctions each defined a.e.
on X, and suppose that there is a lunction s E ~t such that lor each n,
the inequality lin (x) I ;;;; s (x) holds a. e. on X. Then
(i) J lim In dp;;;; lim J In dp
X ~oo 1>-+00 X
and
(ii)
II lim In(x) exists lor ft-almost all x E X, then lim fin dft exists and
~oo fl.---+OO X
(iii) f lim In dft = lim fin dft .

X n-)ooo ff.-)o-OO X
Proof. It is obvious that all In are in ~'i. Hence all In' and s, are finite
a.e. on X. Let A = {x EX: In (x) is ± 00 or lin (x) I > s(x) for some n EN},
let B = {x EX: In (x) is undefined for some n EN}, and let C = {x EX:
s(x) is infinite or is undefined}. Let In (x) = s(x) = 0 on AU B U C.
Since ft (A U B U C) = 0, (12.14) shows that none of the integrals appear-
ing in the statement of the theorem has been changed by this definition.
Furthermore we have lin (x) I ~ s (x) < 00 for all n EN and all x EX.
The sequence (s + Inr:~l consists of nonnegative functions. Applying
FATou's lemma (12.23), we find
f s dft +f lim In dft = f[lim (s + In)] dft

X X ..-..00 X 10-+00
~ lim f (s + In) dft = f s dft + lim f I dft·

10-+00 X X n-->oo X
Thus (i) holds. [The reader will note that the function lim In occurring
n-->oo
in (i) is defined only a. e., but is equal a. e. to the function lim In with
n-->oo
the In defined everywhere as in the proof. ] The inequality (ii) is proved
in like manner, starting with the sequence (s - Inr:~l and using the
equality lim IXn = - lim (- IXn).
~oo ~oo
Finally, if lim In exists a. e. on X, (i) and (ii) imply that

10-+00
lim fin dft ~ f lim In dft ~ lim f In dft .

~oo X X 11-700 ~oo X
Hence lim f In dft exists and (iii) holds. 0

10-+00 x
(12.25) Note. The presence of the dominating function s in the above

theorem is of the utmost importance. If no such function exists, the
conclusion may fail. For example, let X = R, ft = A, and In = n~] 0,-1].
n
Then fin dA = n' ~ = 1 for all n EN while lim In(x) = 0 for all
R n n-+oo
X E R. That is, lim fin dA = 1 =1= 0 = f lim In dA.

n-+oo R R 10-+00
We next extend our integral to complex-valued functions.

(12.26) Definition. Let ~1 (X, .5#, ft) [written for brevity as ~lJ
denote the set of all complex-valued functions I such that I is defined
ft-a.e. on X, Rei E ~'i, and Iml E ~r· For I E ~1> we define f I dft
x
= f Rei dft + i f Iml dft. Functions in ~1 are sometimes called inte-
x x
grable or summable.
(12.27) Theorem. Let I, g E ~1 and IX, fJ E K. Then IXI + fJg E ~1

and I (IXI + fJg) dp. = IX I I dp. + fJ I g dp., i. e., ~1 is a complex linear
x x x
space and I ... dp. is a linear lunctional on ~1'
x
This theorem follows at once by considering real and imaginary parts
and applying previous results.
(12.28) Theorem. Let I be a complex-valued sd-measurable lunction
on X. Then
(i) I E ~l il and only il It I E ~l'
and
(ii) il I E ~l' then I I I dp.l ~ I III dp..
x x
Proof. Conclusion (i) follows directly from the inequalities
III ~ IRel1 + IIml1 ~ 2 Ifl

and the fact that Igl = g+ + g- for real-valued g's. To prove (ii) , repeat
the argument of (9.4). 0
(12.29) Note. To find necessary and sufficient conditions for equality
in the inequality I I I dp.l ~ I It I dp., consider h E ~l' We then ask when
x x
the equality
II h dp.1 = I Ihl dp.
holds. It clearly suffices to have h = exp (i IX) Ihl where IX is any real
number. We now show that this condition is also necessary. Suppose
then that I h dp. = exp(ifJ) If h dp.1 for a real number fJ, and define
cP = exp (-i fJ) h = CPl + icp2'
where CPl and CP2 are real-valued functions. We have
I cP dp. = exp(-ifJ) I h dp. = exp(-i fJ) exp(ifJ) II h dp.1 = I Ihl dp..
Hence
1
J cP dp = f CPl dp + i J CP2 dp = f [cpt + cp~J2 dp. ,

and therefore
1
J cP dp = I CP1 dp. ~ J Icp11 dp. ~ I [cpt+ cp~J2 dp. = I cP dp. .

Hence we have CP2 = 0 a.e. and so cP = CPl a.e. Since I cP dp. = I Icpl dp.,
we have cP ~ 0 a.e. Thus the equality cP = exp(-ifJ)h ~ 0 holds a.e.
and from this we conclude that h = exp(ifJ)CP1 = exp(ifJ) Ihl a.e.
(12.30) LEBESGUE'S Dominated Convergence Theorem [complex form].
Let (In) be a sequence in ~l such that n_oo
lim In (x) exists p.-a. e. on X. Suppose
that there exists a lunction s E ~i such that ifni ~ s p.-a.e. lor each n EN.
Then lim In E ~l and

..-00
lim J In dl-' = J 11-+-00

lim In dl-' .
n-?oo X X
Proof. Let I (x) = lim In (x) whenever this limit exists. Clearly I
..-..00
is defined a.e. on X and is d-measurable. Also I/(x) - In (x) I ~ I/(x)1

+ lin (x) I ~ 2s(x) for all n EN and lim I/(x) - In (x) I = 0 a.e. Thus, ,,-+00
by (12.24) and (12.28.ii), we have

IXJ I dl-' - J In dl-'I ~ J II - Inl dl-' ~ JOdI-'
X X X
= O. 0
(12.31) Definition. Let I be any function for which J I dl-' is defined.

X
For each E Ed we define
J I dl-' = J eE I dl-' .
E X
J I dl-' = J I dl-'E ,
E E
where PE is the measure P restricted to the a-algebra dE (11.37).
(12.32) Corollary. Let I be in ~l> let (An):=l be a pairwise disjoint
00
sequence in d, and write A = U An. Then

11=1
00
J I dl-' =}; J I dp .
A 11=1 A ..
Proof. Define gn = I eA. + ... + I eA..· Then

Ignl ~ III E ~l
and
lim gn(x) = I (XHA (x) a.e .
........ 00
From (12.30) we have

J I dp = J leA dl-' = J lim gn dp = lim J gn dl-'
A X
. X ........ 00 ........ 00 X
= lim }; J I dl-'
..-..00 k=1 A)
00
=}; J Idl-" 0
k=1 Ai
(12.33) Corollary. Let (In) be a sequence 01 complex-valued d-measurable

00 00
lunctions on X such that}; 11101 E~l [or, equivalently, }; J 1/101 dl-' < 00].
£1 In is in ~l and f 1;1 In dp i f In dl-"

11=1 .. =1 X
Then =
X X
Proof. Exercise.
LEBESGUE'S theorem on dominated convergence (12.30), and its

cousins (12.21)-(12.24), are used very frequently in analysis. It is
not too much to say that Fourier analysis, for example, depends upon
(12.30). We shall take up some of these applications in the sequel;
for the moment we content ourselves with a simple though nonobvious
corollary of (12.22).
(12.34) Theorem. Let I E ~1 (X, .sf, pl. For every e > 0 there exists
a t5 > 0 depending only on e and I such that lor all E E.sf satislying
p (E) < t5, we have
I III dp < e.E
Proof. For n = 1, 2, .. " let
1pn (x) = {nil (x) I if II (x) I ~ n,

otherwise.
Then (1pn) is a nondecreasing sequence of .sf-measurable functions and
lim 1pn = III. By (12.22), we have
n-+oo
lim I
n-+oo X
1pn dp = I lim 1pn dp
X n-+oo
= I
X
III dp .
Select n so that j (III - 1pn) dp < ~ e. Setting t5 = 2en and choosing

any E E.sf such that p (E) < t5, we have
1
I 1pn dp ~ In dp = np(E) < 2 e.
E E
It follows that
[I I dp[ ~ I III dp = I (If I - 1pn) dp +I 1pn dp
E E E E
1 1 1
< I (If I - 1pn) dp +2 e < 2 e +2 e= e
x
for all E E.sf such that p(E) < t5. 0
We now return to the functionals I, 1, and 10f § 9. We wish to show
that I is actually an integral.
(12.35) Theorem. Let X be a locally compact Hausdorff space, let I
be a nonnegative linear lunctional on <roo (X), I as in (9.15), and t as in
(9.19). Then (X, JI" t) is a measure space (10.20); and lor every non-
negative JI,-measurable lunction I on X, we have
(i) 1(1)= Iidt.
x
Proof. Let (sn) be the sequence of simple functions defined in terms
of I as in (11.35):
where A". k = {x EX: k 2" 1 ~ I (x) < : .. } and B" = {x EX: I (x) ~ n}.
By (10.35) and (12.4), we have
l(s,,) = J s"dt for all n EN.

x
By (9.17), we have
lim l(s,,) = 1(1) .
n->-oo
By (12.11), we have
lim
n .... oo
J s"dt = xJ I dt.
x
Combining these equalities, we have (i). D
Theorem (12.35) is a generalized form of one of the most famous and
most important theorems of modern analysis; we now state it.
(12.36) F. RIEsz's Representation Theorem. Let X be a locally com-
pact Hausdorff space and let 1 be a nonnegative linear lunctional on <roo (X).
Then there is a measure space (X, vH;, t), where vH; contains all Borel sets,
such that
(i) 1(1) = J I (x) dt(x)
x
lor all IE <roo (X).
Proof. This is a special case of (12.35), since I (I) = 1 (I) for all non-
negative I in <roo (X). and 1 and J ... dt are linear functionals. D
x
(12.37) Remark. The importance of RIEsz's representation theorem
lies in the countable additivity of the integral, as described in (12.21)
to (12.24) and (12.30). Frequently we encounter functionals 1 on <roo (X)
that are nonnegative and linear. RIEsz's theorem shows that we can
write 1 as a countably additive integral; and from this useful conse-
quences often ensue.
(12.38) Remark. In (12.36) there is no statement that the measure,
corresponding to a given functional 1 is unique. In fact, in some cases
there are distinct measures t and 'f} defined on f!I (X) such that
J I d t = J I d'f} for all I E<roo (X) [see (12.58) inlra]. However, this
x x
phenomenon does not occur if we restrict our attention to regular
measures.
(12.39) Definition. Let X be a locally compact Hausdorff space and
let p be a measure defined on a (i-algebra.91 of subsets of X such that.91
contains f!I (X), the Borel sets of X. Then p is called a regular measure if:
(i) p (F) < 00 for all compact sets F ex;
(ii) peA) = inf{p(U): U is open in X, A c U} for all A E.9I;
(iii) P (U) = sup{p (F) : F is compact, FeU} for all open sets
UCX.
Hewitt/Stromberg. Real and abstract analysis 12
It follows from (10.20), (9.27), (9.24), and (9.26) that every measure t
defined as in § 9 is a regular measure on Jt.. [Cf. the different but
related definition of regular outer measure in (10.40) supra.]
(12.40) Theorem. Let fl be a regular measure defined on a a-algebra d
01 subsets 01 a locally compact Hausdorff space X. Then
fl(A) = sup{fl(F) : F is compact, Fe A}
lor every A E d that is a-finite with respect to fl.
Proof. Repeat verbatim the proof of (10.30) with t replaced by fl. 0
let fl and v be regular measures defined on a-algebras Jt,. and Jt" respectively.
Suppose that f I dfl = f I dv lor all IE <rto(X). Then fl(E) = v (E)
x x
lor all E EJt,. n Jt".
Proof. Let F be any nonvoid compact subset of X. Use (12.39.ii)
to find sequences (Un) and (v,,) of open sets containing F such that
fl(UI) < 00, v(v;.) < 00, fl(Un) -7 fl(F), and v (v,,) -7 v(F). For each n EN,
n
set Wn = n
k~1
(Uk n Vk). Then each Wn is open, WI:::> W 2 :::> ... :::> F,
V(Wn) -7 v(F), and fl(Wn) -7 fl(F). For each n EN, use (6.80) to obtain
a function In E <rto(X) such that In (X) C [0, 1], In (F) = {I}, and In(W~)
C {o}. Next let gn = min {II' ... , In}. It is clear that gn E <rto (X), that
gl ~g2 ~"', and that !~gn(x) = ~F(X) for all x E COl Wn nF')',i.e.,
f gl dfl ~ fl(WI) <
fl-a.e. and v-a.e. Since 00 and f gl dv ~ v(WI) < 00,
x x
Theorem (12.24) applies to yield
Thus fl (F) = v (F) for all compact sets F C X.

For open sets U C X, we have fl(U) = sup{fl(F): F is compact,
FeU} = sup {v (F) : F is compact, FeU} = v(U). For arbitrary sets
E EJt,. n Jt", we have fl (E) = inf{fl (U) : U is open, E C U} = inf{v (U):
U is open, E C U} = v(E). 0
let fl be a regular measure defined on a a-algebra d 01 subsets 01 X such that
(X, d, fl) is a complete measure space. Suppose that E Ed il and only
il E n FEd lor every compact set F eX. Define I on <roo (X) by I (f)
= f I dfl and let t be the measure constructed Irom I as in § 9. Then d = --4.
x
and fl (E) = t (E) lor aU E E--4..
Proof. Let E be in .91 and suppose that ft (E) < 00. Since ft is regular,
there exist sequences of sets (Fn) and (Un) such that Fn C E C Un'
each Fn is compact, each Un is open, ft(Fn) ->- ft(E), and ft(Un) ->-ft(E).
00 00
Let A = U Fn and B =
n=!
n Un" Then A and B are Borel sets, AcE CB,
n=!
ft(A) = ft(E) = ft(B), and ft(B n A') = O. It follows from (12.35)
and (12.41) that leA) = l(B) and l(B n A') = O. Therefore, since l
is complete, E = AU [E n (B n A')] E 1.. This argument proves:
E Ed and ft(E) < 00 imply E E1.. (1)
Repeating the above argument with the roles of ft and l reversed,

we have
E E1. and leE) < 00 imply E Ed . (2)
Next consider any E in .91 and any compact set Fe X. By hypothesis

we have E n FEd, and of course ft (E n F) < 00. Therefore E n F
is in 1. for all compact sets F. Applying (10.31.iv), we infer that E E 1.,
and so we have proved that .91 c 1.. A very similar argument proves
the reversed inclusion, and so we have .91 = 1.. Finally, (12.41) shows
that ft (E) = l (E) for all E E1.. D
(12.43) Remark. The hypothesis that E E .91 if and only if E n FEd
for all compact sets Fe X is essential to prove (12.42). For example,
let X = Rd X R, define I on <roo (X) as in (9.41), and let l be the measure
induced by I. Let v be the restriction of l to PA (X) and let (X, .91, ft)
be the completion of the measure space (X, PA(X), v) [see (11.21)]. Then
I (I) = f t d,J. for all t E <roo (X), but we also have .91 ~ 1.. We construct
x
a set A E1. n .91' as follows. Let rp be a one-to-one mapping of R onto
PA(R) (10.25). Define A = U {(x,y):y E rp(x)}. The set A is in 1.
xER
because A n F E PA (X) c 1. for all compact F C X. Assuming that
A Ed, we have A = B U C where C meets only countably many lines
v;. = {(x, y) : y E R} and B E PA (X). Then there exists an ordinal number
IX < Q such that B E <ff<Y, [where tffo is the family of all open subsets of X
and succeeding tff<Y,'s are defined as in the proof of (10.23)]. Thus
B n Vx E tff<Y,+2 for all vertical lines v;.. It follows that rp (x) E~+2
[where ~ is the family of all open subsets of R] for all x ER except possibly
those countably many x's for which C n v;. =!= 0. But this contradicts
the known fact that PA(R) n.fFp =!= 0 for all f3 < Q [see K. KURATOWSKI,
C. R. Paris, vol. 176 (1923), 229; also see W. SIERPINSKI, Fund. Math.,
vol. 6 (1924), 39J. Thus A ~ d.
Next we note an important property of integrals with respect to
Lebesgue measure.
(12.44) Theorem. Let I be a Lebesgue measurable lunction R, let t

be a real number, and let It be the translate 01 I by t (S.14). Then
(i) f Idx) dx = f I(-x) dx = f I (x) dx
R R R
whenever any 01 these integrals is defined.

Proof. For I ~ 0, the translation and inversion invariance of the
Riemann integral makes it clear that
(1)
[Use (S.15.iv), (8. 15.v), (9.S) , and (9.15).J Now call on (12.35); and refer
to (12.2) and (12.26) for conditions under which the integrals in (i)
are defined. 0
(12.45) Continuous images of measures. This construction requires
some preliminary explanation, although the basic idea is simple enough.
Let X and Y be locally compact Hausdorff spaces, and let cp be a con-
tinuous mapping of X onto Y. Suppose that we are given a measure fl
on X in the sense of § 9, and make the hypothesis that
(i) cp-l (F) is compact in X for every compact subset F of Y
or that
(ii) fl (X) is finite.
Consider an arbitrary function IE <roo (Y). It is easy to see that the com-
posite function 10 cp is in ,~.\(X,~, fl), since it is in <roo(X) if (i) holds
and is a bounded continuous function in any case and so is in -S:l (X, ~,fl)
if fl (X) is finite. Therefore the mapping
(iii) 1-+ flo cp (x) dfl (x)
x
is a nonnegative linear functional on <roo (Y). Accordingly there is a meas-
ure 'II in the sense of § 9 on Y such that
(iv) flocp(x)dfl(x) = fl(y)dv(y)
x y
for all IE <roo (Y). The measure'll is called the image 01 the measure fl
under the continuous mapping cpo
(12.46) Theorem. Notation is as in (12.45). For all a-finite 'II-measur-
able subsets B 01 Y, we have
(i) v(B) = fl(cp-l(B)) = f ~B 0 cp(x) dfl(x).
x
For every lunction IE -S:l (Y, JI", 'II), 10 cp is in -S:l (X, ~, fl) and we have
(ii) f I (y) dv (y) = flo cp (x) dfl (x).
y x
Proof. We proceed in steps. Suppose first that U is a nonvoid open
subset of Y. Let ~ be the set of all functions I in <rto (Y) such that
I ;£; ~u· URYSOHN'S theorem (6.S0) implies that sup {I: I E ~} = ~u.
and it is obvious that ~ is directed upward in the sense of (9.11). Taking

note of (12.35) and applying (9.11) twice, we find that
v(U) =
y
J ~u(y) dv(y) = sup {Jy I(y) dv(y) : 1 E~}
= sup { J 10 IP(x) df-l(x): 1E~}
x
= J sup{1 0 IP(x): 1 E~} df-l(x) = J ~u 0 IP(x) df-l(x)
x x
= f-l(IP-1(U)). (1)
Next suppose that B is an arbitrary subset of Y. Theorem (9.24)
and (1) show that
v(B) = inf{v(U) : U is open in Y and U::) B}
= inf{f-l(IP- 1 (U)) : U is open in Y and U::) B}
~ f-l(IP-1(B)) . (2)
In particular, if v(B) = 0, then (i) holds for B. If F is a compact subset
of Y, then F is contained in an open set U such that U- is compact
(6.79). Thus v(U) is finite (9.27), and so
v(U) - v(U n F') = v(F) . (3)
Apply (1) to (3) and note that un F' is open:
v(F) = f-l(IP-1(U)) - f-l(IP-1(U n F'))
= f-l(IP-1(U) n (IP-1(U n F'))')

= f-l(IP-1(F)). (4)
Every a-finite v-measurable set B can be written as
B=CQ1Fn)UP, (5)
where Fl C F2 C ... C Fn C . ", each Fn is compact, P is disjoint
00
from U F n , and v(P) = 0. This follows readily from (10.34), and we

n=l
omit the details. Applying (4) to (5) and using (10.13), we obtain
v(B) = CQl Fn) ~ v (Fn) ~ f-l(IP-1(Fn))

v = =
= f-l CQI IP- (Fn)) f-l (IP- CQI Fn)) ~ f-l(IP-

1 = 1 1 (B)) . (6)
Now (i) follows from (6) and (2).
Let us show that IP- 1 (B) E Jt,. for all B E vIt". If B is a-finite, use (5)
to write
00
IP-1(B) = n~1 IP-1(Fn) U IP-1(P);
each IP-1(Fn) is closed, because IP is continuous, and f-l(IP-1(P)) = °

because of (2). If B is not a-finite, let E be any compact subset of X.

We have
En q;-l(B) = En q;-l(q;(E) n B)
since q; is single-valued. The set q; (E) n B is plainly v-measurable and
finite for v, so that q;-l (q; (E) n B) is ,u-measurable, as was just proved.
Thus E n q;-l(B) is in~, and (10.31) implies that q;-l(B) is in ~.
It is now obvious that for every v-measurable complex function t
on Y, the function 10 q; on X is ,u-measurable. Using (i), it is easy to
establish (ii). Consider 1 E £1 (Y, JI", v); we may suppose that 1 ~ O.
By (11.35) there is a non decreasing sequence (Sn):;"~ 1 of simple, v-measur-
able functions such that lim sn(Y) = I(Y) everywhere on Y. Plainly
n-+oo
f
y
sn(Y) dv(y) ~ f
y
I(Y) dv(y) < 00 • (7)
Write Sn = I'" (Xk~Bk' where 0 < (Xl < ... < (Xm' Then V(Bk) IS finite
k~l
for all k, as (7) proves, and (12.4) and (i) imply that
m m
f sn(Y) dv(y) = I (X"v(B,J = I (X" ,u(q;-l(B,,))
Y k~l k~J
=JC~ (Xk~'rl(Bk)(X)) d,u(x)

f Sn 0 q; (x) d,u (x) .
= (8)
x
Take the limit as n -+ 00 of both ends of (8), and cite B. LEVI'S theorem
(12.22). This proves (ii). 0
We close this section with a large collection of exercise. Of these,
(12.48), (12.51), (12.54), and (12.63) are important either for later appli-
cations or for understanding the theory expounded up to this point.
We hope that all readers will work through at least these exercises, and
that most readers will work through all of them.
(12.47) Exercise. Let (X, d, ,u) be any finite measure space and let B'
be the set of all complex-valued d-measurable functions on X. For I, g EB'
define
e (f, g) = f
I/-gi
1 + II _ gl d,u.
x
Prove the following assertions.
(a) fl(l, g) = 0 if and only if 1= g a.e.
(b) fl (I, g) = fl (g, fl·
(c) e(f, h} ~ e(f, g} + e(g, h).
(d) If (In)':~l satisfies lim e(In' fm) = 0, then there exists a complex-
m,H-)-OO
valued measurable function g such that lim fl (In' g) = O.

n->oo
That is, identifying functions equal a. e., we have defined a complete

metric space iJ.
(e) For IE iJ and (In) C iJ we have (! (tn' I) -+ 0 if and only if In -+ I
in measure. For this reason we call (! the metric 01 convergence in measure.
(12.48) Exercise. Let (X, .91, fl) be any measure space and let I
be a nonnegative, real-valued, bounded, d-measurable function on X.
Let IX = inf{/(x) : x E X} and p = sup{/(x) : x E X}. For n E Nand
j = 1,2, ... , n - 1, let
A; = {x EX: IX + (j - 1) n(f3 - oc) ~ I(x) < IX + j(f3 :: oc) }
and let
An = {x EX: IX + (n - l)n(f3 - oc) ~ I(x) ~ p} .
The Lebesgue sums lor I are defined as the numbers
5" = i~ (IX + (j-l)n(f3- OC)) fl(A;).
Prove that lim

n~oo x
J I dfl·
5n =
Next suppose that fl{X) < 00. Let I be any bounded, real-valued,
d-measurable function on X. Define Sn as above. Prove that lim Sn = JI dfl.
n--+oo X
(12.49) Exercise. Let (X, .91, fl) be any finite measure space and let
(tn):~1 and I be complex-valued, d-measurable functions on X such that
In -+ I a. e. Suppose that there exists PER such that I/nl ~ Pa. e. for all
n EN. Use EGOROV'S theorem, not (12.21)-(12.24) or (12.30), to prove
that
lim J In dfl = f I dfl .
n~oo X X
(12.50) Exercise. Let (X, .91, fl) be a a-finite measure space. Let g
be an d-measurable function on X such that I g E ~1 for every I E~1. Prove
that there is a number IX E R such that fl({x EX: Ig(x)l > IX}) = o.
Give an example to show that this conclusion may fail if the hypothesis
of a-finiteness is dropped.
(12.51) Exercise. Let - 00 < a 0 and x E [a, b], define
m.,(x) = inf{/(t) : t E [a, b] n ]x - 15, x + !5[}
and
M,,(x) = sup{/(t) : t E [a, b] n ]x - 15, x + !5[} ,
and define
m(x) =limm6(x), M(x) = lim Mb(X) .
6-}0 "to
(a) Prove that I is continuous at x if and only if m(x) = M(x).
Let (LI;)~l be a sequence of subdivisions of [a, b], say

Lli = {a = xW) < xV) < ... < x~} = b},
such that
~ max{x~) - X~~l: 1 ~ k ~ ni} = O.
1--+ 00
Let
m~) = inf{t(t) : X~~l ~ t ~ x~)}, M~) = sup{/(t) : X~~l ~ t ~ x~)},
Hi (") Hi (')
cP; = 1: ml
k=1
~].,m .,m[,
1:-1> k
and "Pi = 1: M I
k=1
~].,m .,11)[.
k-l' l:
Prove that:
(b) if x E [a, b] and x is distinct from all x~), then ~im CPi(x) = m(x)
1--+00
and Fm "Pi (x) = M (x) ;
1--+00
(c) m and M are Lebesgue measurable on [a, b];
(d) if L (I, Llj) and U (I, Ll i ) are the lower and upper Darboux sums
respectively [defined with IX (x) = x] for the function I corresponding to the
subdivision LI j, then
b
~im
1-'J-00
L(I, Ll i ) = f
a
m(x) dx
and
b
lim U(/, Ll i )
1~OO
=
a
f M (x) dx;
(e) I is Riemann integrable on [a, b] if and only if I is continuous

a.e., i.e., A({x E [a, b] : I is discontinuous at x}) = 0;
(f) if I is Riemann integrable on [a, b], then I E ~1 ([a, b],..A)., A) and
b
5(1; [a, b]) = f I (x) dx,
a
where 5 denotes the Riemann integral (8.6). [The reader should note
that the foregoing applies only to bounded functions on finite intervals.]
(12.52) Exercise. Find a real-valued function I on [0, 1] such that I
is continuous on ]0, 1], lim f I dA is finite, and the Lebesgue integral
d.j.O [d, I]
of lover [0, 1] is not defined.
(12.53) Exercise. Find a bounded, real-valued, Lebesgue measur-
able function Ion [0, 1] such that f 1/- gl dA > for every Riemann
[0, I]
°
integrable function g on [0, 1].
(12.54) Exercise. Let - 00 < a < b < 00 and let I be a Lebesgue
.,
measurable function on [a, b] such that f
a
I(t) dt = °for every x E [a,b].
Prove that I = 0 A-a. e. [This exercise is needed for the proof of Theorem
b
(16.34) inlra. Hints are as follows. Since J I (t) dt exists and is finite,
a
III is in ~l([a, bJ, vA)., A). It is evident from (6.59), (12.32), and our hy-
pothesis that J I dA = 0 for all open subsets U of [a, b]. Use (9.24) to infer
u
that J I dA = 0 for all A-measurable sets A C [a, b]. From this the identity
A
1= 0 A-a.e. is immediate.J
(12.55) Exercise. Let X be a locally compact Hausdorff space such
that every open subset of X is a-compact. Suppose that I-' is a measure
defined on 1!4 (X) such that I-' (F) < 00 for all compact sets F C X.
Prove that I-' is a regular measure. [First consider the case that X is
compact. Consider the family &f = {E E I!4(X) : I-'(E) = inf{I-'(U) : U
is open, U:::> E} and I-'(E) = sup {I-' (F) : F is compact, Fe En.J
(12.56) Exercise. Let I-' be a measure defined on I!4(R) such that
1-'([0, IJ) = 1 and I-'(E + x) = I-'(E) for every E E I!4(R) and x E R.
Prove that I-'(E) = A(E) for all E EI!4(R). [First prove that I-'({x}) = 0
for all x ER. Next show that I-'(Ja, b[) = b-a for all a < bin R. Use
(12.55)J.
(12.57) Exercise. Let (X, .91, 1-') be an arbitrary measure space and
let I and (In)':=1 be complex-valued measurable functions on X. Suppose
that In --+ I in measure and that there exists a function g E ~~ such that
1/.. 1 ~ g a.e. for all n EN. Prove that lim J II - Inl dl-' = o. [Assume
n~oox
that lim J II - Inl dl-' = > 0 and choose a subsequence (Ink) such that
n~oox
Q(,
lim J II - Ink I dl-' = Then use (11.26)J.

Q(,.
x
k-HXl
(12.58) Exercise. Let 1 be the nonnegative linear functional on

(S.OO(Rd x R) defined in (9.41). Define 'fJ on I!4(Rd x R) by the rule
'fJ(E) = E A({Y: (x,y) E E}). Prove that:
xER
(a) 'fJ is a measure on 1!4 (Rd x R) ;
(b) 1(1)= J Id'fJforall/E{S.oo(RdxR);
RdXR
(c) 'fJ is not a regular measure.
let I-' be a regular measure defined on a a-algebra d of subsets of X
[of coursed:::> I!4(X)J such that I-'({x}) = 0 for all x EX. Suppose that
B is in .91 and that
I-' (B) = sup {I-' (F) : F is compact, FeB} < 00 •
Prove that for each Q(, E [0, I-'(B)J there exists a a-compact set A C B
such that I-' (A) = Q(,.
(12.60) Exercise. Let (X, .91, 1-') be a measure space as described in
(10.56.a). Let I be an d-measurable function on X. Prove that I is con-
stant on each E" except for a set of ,u-measure zero, and accordingly
write the integral J I d,u as a certain finite sum.
x
(12.61) Exercise. Let (X, .91, ,u) be a measure space that is degenerate
in the sense of (10.3): ,u (A) = 0 or ,u (A) = 00 for all A Ed. Show that
every function I E ~l (X, d,,u) vanishes except on a set of ,u-measure
zero and that Jill d,u = O. [This unpleasant property justifies the term
x
"degenerate" for the measure spaces under consideration.]
(12.62) Exercise. Let X be a locally compact, u-compact Hausdorff
space and let ,u be a regular measure defined on au-algebra .91 of subsets
of X [.91:::> gj(X)J. Let I be an d-measurable function on X such that
I(X) C [0, ooJ and such that I~F E ~l(X, d,,u) for all compact subsetsF
of X. [Such an I is called locally ,u-integrable.J Define the set-function"
ondby
,,(A) = J I (x) d,u(x).
A
Prove that" is a regular measure on d.

(12.63) Exercise: Integrals on the completion of a measure space.
Let (X, .91, ,u) be a measure space and (X, d, P) its completion (11.21).
(a) Let i be a complex- or extended real-valued d-measurable
function defined on X. Prove that there is an d-measurable function I
such that I = i p-almost everywhere on X. [Hints. Suppose that i is
extended real-valued. Use (11.35) to find a sequence (S,.) of real-valued,
d -measurable, simple functions such that lim s,. = i everywhere. Each
I: IX","~.A".II with A ... " Ed and the A ... ,,·s pairwise dis-
....... 00
S.. has the form

"
joint. EachA ... "is contained in a set B... " Edsuch thatp(B... "nA~.k) =0.
Let s.. = I: 1X".k~B•• 1I and define I as lim s... J
A n~oo
(b) Let ibe a function in ~l (X, d, P). Prove that there is a function I
in ~l (X, .91, ,u) such that
J 1/- iI dp = 0
x
and
fld,u= fidp.
x x
[It suffices to consider nonnegative functions l By adding sets of ,u-
measure 0, it is easy to make the simple functions s.. of part (a) into a
nondecreasing sequence. By (12.22) we then find
x
J i dp = .......
lim J s.. dp = lim f s.. d,u = J I d,u.
00 x .......00 x x
The other equality is obvious.]
(c) Let X be a locally compact Hausdorff space and (X,.,(, t)
as in §§ 9 and 10. Suppose that X is u-finite with respect to t. Let gj (X)
denote the Borel sets of X. Prove that (X,.A" t) is the completion of

(X, ,qj (X) , t). What does this tell you about Borel measurable functions
and arbitrary .A,-measurable functions? [Use part (a).]
(d) Drop the hypothesis in part (c) that X be a-finite. Let I be an
arbitrary function in ~1 (X,.A" t). Prove that there is a Borel measur-
able function 11 on X such that 11 = I t-almost everywhere and such that
1/11 ~ III· [Consider a subset Y of X a-finite with respect to t such that I
vanishes on Y' [Y can be chosen to be open if you like], and then argue
as in part (c).]
(12.64) Exercise. Let (X, SiI, p,) be a measure space and let I be a
complex-valued SiI-measurable function on X. Prove that I E ~1 (X, SiI, p,)
if and only if there exists a sequence (s.. ) of simple functions such that
(s..) C ~l> S.. --+- I in measure, and
lim
mJfI,~oo
J Is", - s..1dp, =
X
O.
In this case we have
x
J I dp, = lim J s.. dp, .
tI->-co X
[If one first defines J s dp, for complex-valued, d-measurable, simple

x
functions, then the above facts can be used to define ~1 and the integral
on ~1' This approach is useful when dealing with functions with values
in a Banach space. It does not depend directly on the ordering of the
real numbers as our definition of the integral does.]
CHAPTER FOUR
Function Spaces and Banach Spaces

The theory of integration developed in Chapter Three enables us
to define certain spaces of functions that have remarkable properties
and are of enormous importance in analysis as well as in its applications.
We have already, in § 7, considered spaces whose points are functions.
In § 7, we considered only the uniform norm 1111" [see (7.3)J to define
the distance between two functions. The present chapter is concerned
with norms that are defined in one way or another from integrals. The
most important such norms are defined and studied in § 13. These
special norms lead us very naturally to study abstract Banach spaces, to
which § 14 is devoted. While we are not concerned with Banach spaces
per se, it is an inescapable fact that many results can be proved as easily
for all Banach spaces [perhaps with some additional propertyJ as for
the special Banach spaces defined in §§ 7 and 13. Our desires both
for economy of effort and for clarity of exposition dictate that we treat
these results in general Banach spaces. In § 15, we give a strictly com-
putational construction of the conjugate spaces of the function spaces
~P (1 < p < (0). We have chosen this construction because of its elemen-
tary nature and also because we think that manipulation of inequalities
is something that every student of analysis should learn. In § 16, we
consider Hilbert spaces, which are ~2 spaces looked at abstractly, and
also give some concrete examples and illustrations.
All of the sections of this chapter are important, and the reader is
advised to study them all.
§ 13. The spaces 2p (1 ~ P < CD)

As usual, we begin with a definition.
(13.1) Definition. Let p be a positive real number, and let (X, .91, 1')
be an arbitrary measure space. Let f be a complex-valued .9I-measurable
function defined {t-a. e. on X such that ItiP E~~. We then say that
f E~p(X,.9I, 1'), and we define the symboillfilp by
Ilfllp = [fx ItiP d{t]i.

§ 13. The spaces 2p(1 ~ P< (0) 189
Where no confusion seems possible, we will write ~p for ~p(X,.9I, f.l).1

For p ~ 1 and I E~, we call 1I/IIp the norm 01 I or the ~p-norm 01 I. a
The symbols LP, Lp, and ~p are employed by some writers to denote ~p.
For 1 ~ P< 00, the function 1--* 1I/IIp on ~p satisfies all the axioms
for a norm set down in (7.5), except for the positivity requirement:
1I/IIp > 0 if I =l= O. [If .91 contains a nonvoid set E such that f.l (E) = 0,
then ~E =l= 0 but II~Ellp = O.J The only nontrivial fact is the triangle
inequality (7.5.iii), which for ~p is
III + gllp ~ 1I/IIp + Ilgllp .
We shall first prove this inequality, paying attention to possible equalities
and also obtaining some other inequalities useful in the sequel.
(13.2) Theorem [YOUNG'S inequality]. Let q; be a continuous,
real-valued, strictly increasing lunction defined on [0, 00 [ such that
lim q; (u) = 00 and q; (0) = O. Let 1jJ = q;-1. For aU x E [0, oo[ define
u ..... oo
x
(j) (x) = J q; (u) du
o
and
x
'l'(x) = J 1jJ(v) dv.
o
Then a, b E [0, 00 [ imply
ab ~ (j) (a) + '1' (b)
and equality obtains il and only il b = q; (a).
c
Proof. A formal proof can be given using the fact that J q; (u) du
~ 0
+J 1jJ(v) dv = cq;(c) for all c ~ O. However, interpreting the integrals
o
as areas, we render the result obvious by the accompanying Fig. 6. 0
y y
1 For p = I, the present definition is consistent with our earlier definition of

21 given in (12.26), in view of the assertion (12.28.i).
z For 0 < p < I and all but a few measure spaces, the function f --* Ilfllp
on 2p is not a norm in the sense of (7.5). See (13.25.c) for a discussion.
190 Chapter IV. Function spaces and Banach spaces
For any positive real number p such that p =l= 1, define p' = P~l
[thus ; + ;, = 1] .
(13.3) Corollary. For p > 1 and a and b any nonnegative real numbers,
we have
(i)
Equality holds in (i) il and only il aP = bP'.

Proof. For u E[0, 00 [, define cp (u) = uP -1; cp is continuous and strictly
increasing, lim cp (u) = 00, and cp (0) = 0. The inverse tp of cp is given by
"->-00
1 a a p
tp(v)=v P- 1• We have (/>(a)=Jcp(u)du=JuP-ldu=~ and lJI(b)

o 0 P
b b _1_ bP '
= J tp(v) dv = J v P- 1 dv=y [the Lebesgue integral and the Riemann in-
o 0
tegral agree on Riemann integrable functions (I2.51.£)J. The corollary
follows at once from (13.2). D
(13.4) Theorem [HOLDER'S inequality for p > 1]. Let I E5:.p and
g E5:. P" where p > 1. Then Ig E5:.1 ' and we have
(i) [J Ig d,u[ ~ J [/g[ d,u
and
(ii) J [/g[ d,u ~ [[/[[p [[grip';
and so also
(iii) [J Ig d,u[ ~ [[/[[p [[grip' .
Proof. We first prove (ii). [Note that (ii) and (I2.28.ii) imply (i).J
If I or g is zero ,u-a.e., then (ii) is trivial. Otherwise, using (13.3), we have
If(u)1 Ig(u)1 1 If(uW 1 Ig(u)i"'
~ lfijf;: ~ p lfm + 7 Ilgll~:
for all u in X such that I (u) and g (u) are defined, i. e., for ,u-almost all u.
Thus we have
Ilfllp ~Igll,,' f [Ig[ d,u ~ p II~II~ f [tiP d,u + P' II~II~: f [g[P' d,u = ~ + ;, = 1,
and this proves (ii). The inequality (iii) is immediate. D
For p = p' = 2, the inequality (ii) is called CAUCHY'S inequality,
or SCHWARZ'S inequality, or BUNYAKOVSKiJ's inequality; sometimes the
three names are listed together.
(13.5) Conditions for equality in (13.4). To get equality in (I3.4.ii),
it is clearly necessary and sufficient that we have
II (u) I Ig(u)1 = ~ I/(u)lp +~ Ig(u)I'"
~/II.. IIglil" P 11/11: p' IIgll::
§ 13. The spaces ~p(1 ~ P< 00) 191
for almost all uEX. By (13.3), this happens if and only if 111;:1; = I\!II~;
almost everywhere. Thus equality obtains in (13.4.ii) if and only if there
are nonnegative real numbers A and B, not both zero, such that
A ItIP = B IgIP'
almost everywhere.
The reader can easily formulate from this and (12.29) a necessary
and sufficient condition that equality hold in (13.4.iii).
(13.6) Theorem [HOLDER'S inequality for 0 <p < 1]. Let 0 < p < 1
and let I and g be lunctions in ~t and ~t" respectively. Then we have
1 1
(i) fig dp. ~ (J IP dp.)p (J gP' dp.)P'"
unless f gP' dp. = 0 [note that P' < OJ.
Proof. In the case we are concerned with, we have 0 < f gP' dp. < 00,
and since P' < 0, this implies that g (x) > 0 for almost all x EX. Let
1 1 1
1 -- - - , ,
q = -, and define q; = g q and 1j! = g q I q • It is easy to see that q;q = gP •
P
and so q; E~q" If fig dp. = 00, then (i) holds trivially. Otherwise Ig
is in ~l' and so 1j! is in ~(l' Applying (13.4) with p replaced by q. we have
1 1 1
f IP dp. = f q;1j! dp. ~ (J 1j!q dp.F (J q;q'dp.)7 = (f Ig dp.)P (f gP' dp.)7.
It follows immediately that
1 1
fig dp. ~ (J IP dp.)p (J gP'dp.f7P •
and since - q~p = ;, • the theorem follows. 0
(13.7) Theorem [MINKOWSKI'S inequality]. For 1 ~ P< 00 and
I, g E~p. we have
(i) III + gllp ~ 1I/IIp + Ilgllp .
Proof. Suppose first that p > 1. We have
II + glP ~ (It I + IgJ)P ~ [2 max {III. Igl}JP
= 2Pmax{lfIP. IgIP} ~ 2P(lfIP + IgIP) .
This crude estimate shows that II + glP E~l' i.e .• 1+ g E~p, Thus (13.4)
implies that
11/+ gll~ = f II + glP dp. ~ f II + glP-l III dp. + f II + glP-1lgl df'
1 1 1 1
~ (J IflPdp.)P (J II + gl(P-l)P' dp.)P'" + (J IgIPdp.)p (J 1/+ gl(P-l)P' dp.)P'"
.P..
= (llllIp + IIgllp) 11/+ gill' •
The inequality
p_..P...
III + gllp P' ~ 1I/IIp + Ilgllp
thus holds. Observing that p - :' = 1, we obtain MINKOWSKI'S inequality
for p > 1. Since f II + gl dfl ~ fill dfl + f Igl dfl, the inequality is
trivial for p = 1. 0
We now give conditions for equality in MINKOWSKI'S inequality.
(13.8) Theorem. For p = 1, we obtain equality in (13.7.i) il and only il
there is a positive measurable lunction e such that
I(x) e(x) = g(x)

almost everywhere on the set {x: I(x) g(x) =1= o}. Equality obtains lor
1 O.
Proof. Exercise.
(13.9) Theorem. For 0 < p < 1 and I, g E~t, we have
(i) III + gllp ~ 1I/IIp + Ilgllp .
Proof. The estimate given in (13.7) for 11+ glP shows again that
1+ g E~p. To prove (i), use (13.6.i) and the argument of (13.7). 0
We next describe the exact sense in which ~p is a normed linear
space (P ~ 1).
(13.10) Theorem. For 1 ~ P < 00, ~p is a normec linear space over K,
where we agree that I = g means I (x) = g (x) lor fl-almost all x EX. [Alter-
natively, let m = {I E~p : I (x) = 0 a. e. on X}; then m
is a closed linear
subspace 01 ~p. What we call ~p upon identilying lunctions that are equal
a. e. is really ~p;m.]
Proof. It is trivial that I a/lip = lalll/lip. All other necessary verifica-
tions have been made. D
The following theorem is of vital importance in many applications
of integration theory. A very special case, the RIEsz-FISCHER theorem,
was regarded as sensational when it was first enunciated in 1906. Now,
as we will see, the general theorem is not hard to prove.
(13.11) Theorem. For 1 ~ P < 00, ~p is a complex Banach space,
i. e., in the metric e(I, g) = III - gllp, ~p is a complete metric space.
Proof. Let (In)':~1 be a Cauchy sequence in ~p, i.e., (In) has the prop-
erty that lim Illn - Imllp = O. The sequence of numbers (tn(X)):~1
ffl,n--+oo
may converge at no point x EX [the sequence (In) constructed in (11.27)
serves as an example of this phenomenon.] However, we can find a
subsequence of (In) that does converge fl-almost everywhere. In fact,
choose (lnk)k~1 as any subsequence of (In) such that n 1 < n 2< ...
§ 13. The spaces ~I>(I ~ P< 00) 193
00
< n" < ... and I; 11I"A:+1 - I",tllp = at: < 00. This is possible: e.g., we can
k=1
select increasing n,,'s such that 111m - l"A:llp < 2-" for all m ~ n". Now
define
g" = 1/... 1+ 1/...- 1...1+ ... + 1/"A:+1- I..,tl, for k = 1,2,3, ....
It is clear that
r
Ilgtlll = Ilg"lI~ = (111/...1 + 1/..,- 1...1 + ... + I/n,t+1- In,tlllp)P
~ (IItn.llp +;~ 11/ +1- In;llp
1IJ ~ (1I/n.llp + at:)P < 00 •
Let g = lim g". By B. LEVI'S theorem (12.22), and the above estimate,
k~oo
we have
J gP dp. = J lim k~oo
g~ dp. = lim
k ..... oo
J g~ dp. < 00 •
Hence g is in ~p; i.e.,
f [1/...1+ ;~ 1/"J+1- 1..;1] P dp. < 00 •
The nonnegative integrand above must be finite p.-a.e., and so the series
00
I; 1/";+1 (x) - I..; (x) I converges p.-a.e. Obviously the series

j=1
00
In. (x) + I; (1";+1 (x) - In; (x))

;=1
also converges p.-a. e. The ktl' partial sum of this series is I"l:+1 (x), and
so the sequence (lnA:(X))'k=1 converges to a complex number I(x) for all
x EA, where A Ed and p.(A') = O. Define I(x) as 0 for all x E A'. It
is easy to see that I is d-measurable, and obviously I is complex-valued
onX.
We will show that I is the limit in S!.p of the sequence (In), and this
will of course prove that S!.p is complete in the metric induced by the ~p
norm. Given e > 0, let l be so large that
III. - Itllp < e for s, t ~ n, .
Then for k ~ land m > n l , we have
111m - InA:IlP < e.
By FATOU'S lemma (12.23), we have
J II - Iml Pdp. = J klim
.....
I/"A: - Iml Pdp. oo
Thus for each m > n l , the function I - 1m is in ~p, and so I = I - 1m + 1m

is in ~p; and
lim
,,-+00
III - Inllp = O. 0
(13.12) Remark. The function spaces ~p [real-valued d-measurable

1
functions defined ft-a.e. on X such that 1I/IIp = (f IN dft)P < 00] are
real normed linear spaces for 1 ~ P < 00, and they too are complete.
The proofs are very like the proofs for the complex spaces ~p.
(13.13) Example. Let D be any nonvoid set and consider all complex-
valued functions I on D such that 1: I/(xW < 00, where 0 < P< 00.
xED
[Recall that 1: II (x) IP = sup {1: II (x) IP: F is a finite subset of D}.]
xED xEF
If d is all subsets of D and I' the counting measure defined in (lOA.a),
then these functions are the elements of ~p(D, d, 1'). Custom dictates
that this space be designated by lp (D), and if D = N, simply by lp.
If 1 ~ P< 00, then lp (D) is a complete metric space in which the metric
is obtained from the norm
1
1I/IIp = (1: II (xW)P .

xED
The HOLDER and MINKOWSKI inequalities take the forms

1 1
1: I/(x)g(x)l ~ (1: 1(f(xW)p (1: Ig(xW,}fT
xED xED xED
and
1 1 1
(1: I/(x) + g(x)IP)p ~ (1: II (x)IP)P + (1: Ig(x)IP)p,
xED xED xED
respectively. If D is finite, say D = {1, 2, ... , n}, then the foregoing

produces the lp norm and its corresponding metric on Kn and Rn. The
distance between two points (xl>
y x 2 , ••• , x n) and (Yl> Y2' ... , Yn) is
1
( ,f Ix; - Y;IP)P.
1=1
For p= 2, we
obtain the classical Euclidean
metric. The topologies induced by
the lp metrics on Kn and Rn are
all the same [d. (6.17)].
The first quadrant of the unit
balls in R2 for various values of p
are sketched in Fig. 7.
o (13.14) Examples. The spaces
~p([O, 1]) and ~p(R), where 0< p
Fi&. 7. < 00 [it is understood that I' = A.
§ 13. The spaces ~p(1 ~ P < 00) 195
and d = ..-HAJ, are very important function spaces in both pure and ap-
plied analysis.
(13.15) Example. For p = 2, we obtain the famous function space
~2(X, d, fl). In this case p = P' = 2, and so HOLDER'S inequality takes
the form
f [fg[ dfl ~ [[f[[2 [[g[[2
for f, g E~2· Consider the mapping that takes ~2 x ~2 into K by the rule
(I, g) --+ f f g dfl = (f, g) ,
where the equality defines (f, g). This mapping has the following prop-
erties:
(fl + f2' g) = (fl' g) + (f2' g);
(rxf, g) = rx(f, g) for rx EK;
(f, g) = (g, f);
(f, f) > 0 for f"4= 0 .
We infer from these identities [or directly] that
(f, gl + g2) = (f, gl) + (f, g2) ,
(f, rxg) = Ii (f, g) ,
and
(0, f) = (f, 0) = 0 .
The spaces ~2 can be described in abstract terms, as follows.
(13.16) Definition. Let H be a linear space over K having an inner
product
(x, y) --+ (x, y) EK
mapping H x H into K such that
(x + y, z) = (x, z) + (y, z) ,
(rxx, y) = rx(x, y) for rx EK ,
(x,y) = (y, x),
(x, x) > 0 if x "4= O.
[The other properties of ( , ) listed for ~2 in (13.15) can be proved for H
from the above relations.] Then H is called an inner product space or
a pre-Hilbert space. For x EH, define
1
[[xli = (x, x)-Z.
The inequalities
[(x,y)[ ~ [[xli . [[y[[
and
I[x + y[[ ~ [lx[1 + [[yl[
13*
can be proved. Thus H is a normed linear space. If H is complete rela-

tive to this norm, then H is called a Hilbert space. There is a very
extensive theory of Hilbert spaces. We will take up the rudiments of
this theory in § 16 infra. One of the most striking facts of this theory
is that every Hilbert space is identifiable qua Hilbert space with some
12 (D). Thus in particular every ~2 space can be identified with some space
12(D). We will deal with this identification problem in (16.29).
We now return to the spaces ~p, establishing a few more simple
facts.
(13.17) Theorem. If f-t (X) < 00 and 0 1. For any I E~q, we have
l i P q-p
J ItIP df-t ~ (J ItiP' df-tY (J I" df-t)""'" = (J Wdf-tF (f-t (X))-q .
It follows that I E~p, and that
q-p 1 1
Ilfllp ~ Ilfllq (f-t (X))pq = 1I/IIq(f-t(X))p-q. 0
(13.18) Theorem. II 0 < p < q < 00, then lp (D) C lq (D); and the
inclusion is proper il D is infinite.
Proof. Suppose that I Elp (D); then we have
I: I/(x)lq = I: If(x)iP I/(x)l'-P ~ Aq-p I: If(x)iP ,
xED
where A is a constant such that II (x) I < A for all xED. The reader
should find it easy to construct an example illustrating that the inclu-
sion is proper if D is infinite. 0
(13.19) Theorem. II I E~p n ~q, where 0< P< q< 00, and if
p < r < q, then I E~,. Also, the function q; defined by
q; (r) = log (1Ifll~)
on [p, q] is convex, i. e., 0 < IX < 1 implies
+ (1 - IX)q) ~ IXq;(P) + (1 - IX) q;(q) .
q;(IXP
Proof. Let r = IXP + (1 - IX) q, 0 < IX < 1. Using HOLDER'S inequality
with ~ [note that (~)' = l~,J, we have
JI/I'df-t = J I/lexP+(l-ex)q df-t ~ (J ItlexP.~ df-tr (J 1/1(l-ex)q 1~" df-tY-"

= (J IN df-t)ex (J I/lq df-t)1-ex .
§ 13. The spaces ~1>(1 ~ P< 00) 197
[Notethatthefunctions 1/1"Pand 1/1(l-,,)qarein~~ and~(H' respectively.]

Hence we have I E~",p+(t-",) q' and
11/11~tta=~~g ~ (1I/IIt)" (11/11~}1-" .
Taking the logarithm of both sides of this inequality, we have
rp(~P + (1 - ~) q) ~ ~rp(P} + (1 - ~) rp(q} ,
i. e., rp is convex. 0
(13.20) Theorem. Consider any ~p, 1 ~ P < 00. For every I E~
and every B > 0, there exists a simple lunction a E~P such that lal ~ III
and Iia -/llp < B. In particular, e n ~P is dense in ~p.
Proof. Note first that e n ~P C ~p. Suppose next that I ~ 0, I E ~p.
According to (11.35) there exists a nondecreasing sequence (sn) of non-
negative simple functions such that Sn (x) --+ I (x) f-t-a. e. For each n EN
we have
1/- SniP ~ II + SniP ~ 12fiP = 2PIP E~1·
It follows from LEBESGUE'S theorem on dominated convergence (12.24)
that
and so we can choose sn Ee n ~P so that sn ~ I and Iisn -/llp is arbitrar-

ily small. For an arbitrary I E~p, write I = 11 -/2 + i(f3 - 14}, where
Ii E~t and Id2 = 13/4 = O. For B > 0, choose ai E ~t n e such that
ai ~ Ii and Ilai -1;llp < : (j E{I, 2, 3, 4}}. Define a as a 1 - a 2+ i (a 3- a4);
obviously a is in e n ~p. Also we have
4
11/- allp ~ ~ Illi - a;llp < B
i=i
and
4 4
lal 2 = (a1 - a 2}2 + (a3 - a4}2 = ~ a'f ~ ~ ff = 1/12. 0
i=i i=i
(13.21) Theorem. Let X be a locally compact Hausdorff space, let
t be any measure on X as in § 9, and let vii. be the a-algebra 01 t-measur-
able sets. Then <roo(X) C ~p(X, vII., t} and <roo(X} is dense in ~p(X, vII., t}
lor 1 ~ P< 00. That is, il I E~P (X, vII., t) and B > 0, then there exists
a lunction rp E<roo (X) such that 11/- rpllp < B. Moreover, ill is bounded,
then rp can be chosen so that Ilrpll .. ~ 11/11 ...
Proof. If rp E<roo (X), then there exists a compact set F [say the sup-
port of rp] such that IrplP ~ Ilrpll~~F. Since t (F) < 00, itfollows that rp E~p.
Let I E~P (X, vII., t) and let B > 0 be given. Apply (13.20) to obtain a
function a Ee n ~P such that lal ~ III and 11/- allp < ; . Since a
takes on only finitely many complex values, it is clear that Iiall .. < 00.
If a = 0, then a E <roo (X) and the proof is complete. Thus suppose that
Iiall .. = M > O.
Let E = {x EX: a (x) =1= O}. Since a E en S!,p, we have t(E) < 00.
We apply LUZIN'S theorem (11.36) to obtain a function f/J E <roo such
that Ilf/JII .. ~ M and if A = {x EX: f/J(x) =1= a (x)}, then t(A) < (4~ y.
We therefore have
1
III - f/Jllp ~ III - allp + Ila- f/Jllp < ; + (/ la - f/JI P dt) P

1 1
=; + (!la-f/JIPdt)P~; + (!(2M)Pdt)P
=; +2M·t(A)P< e.
1
0
Definition. A step lunction on R is any function of the form

.. (13.22)
1: OCk ~Ik where OC1 , ••• , OCn are complex numbers and each I k is a bounded
k=1
interval [open, closed, or half-open].
(13.23) Theorem. Let t be any measure on R as in § 9 and let 1 ~ P
< 00. Then the step lunctions on R lorm a dense subset 01 s!'p (R, JI., t).
Proof. Clearly each step function is bounded and vanishes off of a
compact set, and so each step function is in s!'p.
r
Let I E s!'p and e> 0 be given. Use (13.20) to find a simple function aEs!'p
m
suchthatllt-allp< ; ,saya= 1: .Bj~Bj,whereBjEJl.forallj.SinceaEs!'p,
1=1
we may suppose thaI< (B;) < 00 10' all j. Let b ~ ; (2 (I +:~ IP,lj
and fix j in {I, 2, ... , m}. Use (9.24) to obtain an open set U j such that
00
00
where the Ij,k's are pairwise disjoint open intervals. We have 1: t(Ij,k)
k=1
00
= t(Uj) < 00, and so we may choose ko such that 1: t(Ij,k) < ~.
k=k.+1
k.
Let V; = k':!l Ij,k and TV; = Uj n V;'. We see at once that £(TV;) < ~,
V; L. B j C ([1; n Bj) U TV;, and £(V; L. B;) ~ t ([1; n Bj) + £(l1j) < 2~.
Therefore
1
IlgB;- gvjllp= IIgv;6dlp< (2r5)P.
m
Next set s = 1: .BigVj' Since each V; is the union of a finite number of
;=1
§ 13. The spaces .s!p{l ~ P < 00) 199
intervals, s is a step function. Also
Iia - slip = II~ Pi(~Bi- ~Vi)llp

~ £IPil'"~Bi-~vilp< (2!S)*(~IPil)<;'
and so
111- slip ~ 11/- allp + lIa - sllp< e. 0
The norm in the spaces ~p (R, ..A)., A) has an interesting and useful
continuity property, which we now establish. Recall the definition (8.14)
of the translate I" of a function I on R by a real number h.
(13.24) Theorem. Let X = R and let A be as usual Lebesgue measure.
Let p be a real number such that 1 ~ P < 00, and let I be any lunction in
~p(R,..A}., A). Then
(i) lim III" -flip = lim III" -/lip = 0 .

"to "to
Proof. Let e be an arbitrary positive number. Applying (13.21),
choose cp E<roo (R) such that IIcp -/lip < ; . Then cp is uniformly contin-
uous (7.18), and also there is a positive real number at: such that cp(t) = 0
if ItI ;::;; at:. Choose !S > 0 so that !S < 1 and
Y
1
Icp(t + h) - cp(t)1 < ; (2{CX ~ 1) for Ihl <!S.

Then Ihl < !S implies that
«+1
jlcp(t+h)-cp(t)iPdt= j Icp(t+h)-cp(t)iPdt < (;r,

R -(<<+1)
which is to say that

IIcp,,- cpllp< ; .
It is clear from (12.44) that
IIcp" -/"lIp = IIcp -/lip for all hER,

and so we have
III" -/lip ~ III" - cp"lIp + IIcp" - cpllp + IIcp -flip < e
if Ihl <!S. 0
(13.25) Exercise. Let p be a real number such that 0 < p < 1.
Let (X, d, "') be any measure space, and let I, g be functions in
~p(X, d, ",).
(a) Prove that
~-l
(i) III + gllp ~ 2P ""lIp + IIgllp) .
[Hints. For 0 ~ t< 00, prove that

1 + tP ~ (1 + t)P .
This implies that
III + gll~ ~ II/II~ + Ilgll~ .
! !
Next look at the function 1p(t) = (1 + tp ) (1 + tfp. The function 1p
has exactly one minimum in [0, 00 [, at t = 1. Computing this minimum,
one finds that (i) can be established. J
(b) Prove that e(I, g) = III - gll~ is a metric on 5!.p under which 5!.p
is a complete metric space. [Hint. Imitate the proof of (13.11).J
(c) Suppose that X contains two disjoint d-measurable sets A
and B each of finite positive ,a-measure. Prove that I lip is not a norm,
i.e., that there are functions I, g E5!.p for which III + gllp > 1I/IIp + Ilgllp.
[Write q for the number ~ . Note that (1 + W> 1 + fl for all positive
real numbers t. Let I = (X~A and g = fJ~B' where IX and fJ are positive
real numbers. Now computing the 5!.p-norms of I, g, and 1+ g, and
using the inequality just noted, one can choose (X and fJ so as to solve
the present problem.J
(d) In this part, let p be any positive number. Suppose that no two
sets in d of finite positive ,a-measure are disjoint. Determine completely
the structure of 5!.p(X, d, ,a). From this determination, show that
5!.p (X, d, ,a) is a trivial normed linear space for all p > O. [Hints. Sup-
pose that there are no sets at all of finite positive ,a-measure. Then every
5!.p reduces to {O}. Suppose next that there is some set of finite positive
,a-measure. Then, under our hypothesis, every 5!.p reduces to K alone. J
(13.26) Exercise: Generalized HOLDER'S inequality. Let IXv IX2' •. , IXn
n
be positive real numbers such that I; (Xj = 1. For Iv 12' ... , In in 5!.i,
i=!
we have
I~' I~' . . . I~n E5!.t '
and
x
f
(t~1/~' ... r::) d,a ~ II/lll~' 11/211~' ... Il/nll~" .
(13.27) Exercise. Write out carefully and prove the conditions for
equality in MINKOWSKI'S inequality for p = 1 and also for 1 0 and t({x}) = 0 for all x EX.
Let p be any positive number.
(a) Find a function I E5!.p such that I ~ 5!.PH for all <5 > O.
2""
[Hint. Use (12.59) and the fact that I; -;2 =
00
00 for all <5 > 0.]
n=!
§ 13. The spaces ~p(1 ~ P< 00) 201
(b) Find a function I on X such that I E~p_" for all r5 > 0 and I ~ ~p.
00
[Recall that 1.: n-rx converges if a > 1 and diverges if 0 < a ~ 1.]
n~l
(c) Find a nonnegative real-valued function that is in no ~p.
(13.29) Exercise. Consider the set [0, 00 [ and Lebesgue measure A on

it. For every p > 0, find a function I on [0, 00 [ such that I E~p and I ~ ~q
if P * q. [Hint. Consider the function g such that g (x) = x (1 + I~og (x) \)' .]
(13.30) Exercise. Let (X, .91, p) be a finite measure space and let I
be any bounded measurable function on X. Prove that
lim
p-+oo
1I/IIp = inf{a ER: a> 0, p({x EX: I/(x)1 > a}) = o}.
(13.31) Exercise. Let p be a real number such that 1 ~ P< 00, and
let I be a function in ~p (R) such that I is uniformly continuous. Prove
that I E~o (R). Show by examples that each ~p (R) contains an unbounded
continuous function.
(13.32) Exercise. Let (X,d, p) be a measure space such that
p(X) = 1 and let I be a function in ~t(X,.9I, p). Define log (0) as - 00.
(a) Prove that
(i) flog I (x) dp (x) ~ log ( f I (x) dp (x) .
x x
[Hint. Check the inequality log (t) ~ t - 1 for 0 ~ t < 00. Replace t
by 11:111 I(x) and integrate.]
(b) Prove that equality holds in (i) if and only if I is a constant func-
tion a. e. [Hint. Check that log (t) < t - 1 if t =F 1. J
(c) Prove that
(ii) lim IIfll, = exp [f logf(x) dp(x)].
'to x
[Hints. Show that W - I)ly decreases to logl as y 1- 0, and apply the
dominated convergence theorem to prove that
lim y-l( flrdp - 1) = flog I dp.
'to x x
Using (i), show that
! [fx rdp - 1] ~ +logxf tr dp ~ +xflog W) dp x

f log I dp .]=
(13.33) Exercise. Let (X, .91, p) be a measure space and let p be any
positive real number. Prove that if I and (fn)':~l are in ~p (X, .91, p) and
III - Inllp --+ 0, then In --+ I in measure. Find a sequence (fn)':'~l C
~p([O, IJ, JI}., A) such that In --+ 0 in measure but II/nllr+-+ O.
(13.34) Exercise: Convex functions. Let I be an interval of R.

A real-valued function (/J defined on I is said to be convex if whenever
a< b in I and 0 ~ t ~ 1 we have
(/J(ta + (1- t) b) ~ t(/J(a) + (1- t) (/J(b),
i.e., on the interval [a, b] the graph of (/J is never above the chord [line
segment] joining the points (a, (/J(a)) and (b, (/J(b)). Let (/J be a convex
function.
(a) Prove that if tl , . . . , tn are positive real numbers and
{Xl> ... , xn} C I, then
(i) (/J ( tl Xl + ... + t"X,,) :5: t1/fJ (Xl) + ... + tft /fJ (Xft) •
; + ... + t" - tl + ... + t"
[Use induction.]
(b) Prove that (/J is continuous on the interior JO of I and show by
an example that (/J may be discontinuous at the endpoints of I.
(c) Prove that if c is in JO, then there exists a real number ex such
that (/J(u) ;;:::; ex(u - c) + (/J(c) for all u EI, i.e., the line through (c, (/J(c))
having slope ex is always below or on the graph of (/J.
(d) Prove the following generalization of inequality (i). Let (X, .91, p)
be a finite measure space. If I E~~ (X, .91, p), if I(X) C I, and if (/J 0 I
E~~ (X, .91, p), then
(ii)
x
(/J L(~)
I dp] ~ P(~}
x
f((/J 0 I) dp. f
Inequalities (i) and (ii) are known as JENSEN'S inequalities. [Hints.
Let c = p (~) f I dp. Show that eEl. For the case that c EJO, let ex
x
be as in (c). Then
(/J o I (x) ;;:::; ex(t(x) - c) + (/J(c) for all x EX.
Integrate both sides of this inequality. The other case is straightforward.]
(13.35) Exercise. Let f(J be a real-valued nondecreasing function
defined on an interval [a, b[ C R. For a ~ x< b, let (/J(x) = J" f(J(u) duo
a
Prove that (/J is convex on [a, be.
(13.36) Exercise. Let (/J and lJI be as in YOUNG'S inequality (13.2).
Let (X, .91, p) be a a-finite measure space and let ~~ be the set of all
complex-valued d-measurable functions on X such that (/J 0 ItI E
~t(X,d,p).
(a) If (/J increases too rapidly we may have I E~~ and 21 ~ ~~. Give
1
such an example. [Try f(J(u) = exp(u) - 1 and I(t) = log(t-"2).]
§ 13. The spaces ~I>(I ~ P< 00) 203
Let ~(/) be the set of all complex-valued d-measurable functions I

on X such that
II/II(/) = sup{ Il/glll: g E ~~,

x
J':P 0 Igl dft ~ 1}< 00 •
Prove that:
(b) ~~ c ~(/) [use (13.2)J;
(c) ~(/) is a complex linear space;
(d) I II(/) is a norm on ~(/), where functions equal a.e. are identified;
(e) with the norm 1111(/), ~(/) is a Banach space. [First suppose that
ft(X) < 00. Prove that if Il/n - Imll(/) -+ 0, then Il/n - Imlll -+ O.J
The spaces ~(/) are called BIRNBAUM-ORLICZ spaces. For further infor-
mation about these spaces, the reader should consult A. C. ZAANEN,
Linear Analysis, Vol. I [New York: Interscience Publishers, 1953J.
(13.37) Exercise. Define if> on [0, 00 [ by if> (t) = 0 if 0 ~ t ~ 1 and
if> (t) = t ·logt if t ~ 1. The Birnbaum-Orlicz space ~(/) (X, d, ft) [see
(13.36)J is often denoted ~ log+~. Prove that for the measure space
([0, IJ,~, A) we have
~p C ~ log+ ~ C ~1
for all p > 1. The space ~ log+ ~ arises quite naturally in Fourier analy-
sis. See for example A. ZYGMUND, Trigonometric Series, 2nd Edition.
[2 Vols. Cambridge: Cambridge University Press, 1959J, and also Theorem
(21.80) inlra.
(13.38) Exercise: VITALI'S convergence theorem. Let (X, d, ft) be a
measure space and let 1 ~ P < 00. Let (fS:=l be a sequence in ~p(X,d, ft)
and let I be an d-measurable function such that I is finite ft-a.e. and
In -+ I ft-a.e. Then I E~p(X, d, ft) and III - Inllp -+ 0 if and only if:
(i) for each e > 0, there exists a set A. Ed such that ft (As) < 00
and J I/nl P d,u< e for all n EN;
A~
and
(ii) lim
p(E)-;'()
f I/nl P d,u = 0
E
uniformly in n, i. e., for each e > 0 there is a t5 > 0 such that E Ed
and ,u (E) < t5 imply J I/nl P dft < e for all n EN.
E
Prove this theorem. [To prove the necessity of (i), let e > 0 be given,
choose no EN such that 111- Inllp < e for all n ~ no, choose B., C. Ed
of finite measure such that J IfIP dft < e and J Ifnl P dft < e for n = 1, ... , no.
B~ c~
Then put A. = B. U C•. The necessity of (ii) is proved similarly by using
(12.34). Next suppose that (i) and (ii) hold. Use (i), FATou's lemma and
MINKOWSKI'S inequality to reduce the problem to the case that ft (X) < 00.
For e> 0, let (J be as in (ii). Use EGOROV'S theorem to find BEd

such that p,(B) < (J and In --+ I uniformly on B'. Use FATOU'S lemma to
prove that 1 IfIP dp, < e. Then use MINKOWSKI'S inequality to show
B
that 1 II - Inl Pdp, < 3Pe for all large n. Thus conclude that 1= (f - In)
X
+ In E: ~P and III - Inllp --+ o.
VITALI'S convergence theorem has considerable theoretical impor-
tance and can also be frequently applied to prove other useful theorems.
The next exercise is also useful for applications [see for example (20.58)
inlraJ and so we provide plentiful hints for its proof.
(13.39) Exercise. Let (X, .91, p,), p, (In), and I be as in (13.38).
Suppose that In --+ I p,-a. e. For each (n, k) EN x N let Bn,k = {x EX:
lin (x)IP ~ k}.
(a) Suppose that condition (13.38.i) holds [as it does, for example,
if p, (X) < (Xl J. Prove that the following four assertions are equivalent:
(i) I E~P and III - Inllp --+ 0;
(ii) if(E k)%"=lcd,E1 ::JE 2 ::J···,and n
k= 1
E k =0,thenlim
k-+oo Ek
Il/nlPdp,
= 0 uniformly in n;
(iii) lim 1 I/nl P dp, = 0 uniformly in nl;
k-+oo Bn.k
(iv) condition (13.38.ii) holds.

[Hints. Assertions (i) and (iv) are equivalent by (13.38). To show
that (i) implies (ii), consider e> 0 and no EN such that Il/n - Illp< e
for all n ~ no. Then for n ~ no, we have
1 1 1
(/I/nIPdp,)P~ (I lfiPdp,)P + (/I/n-fiPdp,)p
Ek Ek Ek
1
< (I lfiP dp,)P + e; (I)

Ek
now apply dominated convergence to (lfiP ~Ek)~l to show that (1) is

less than 2e for k ~ ko and all n ~ no. If n E{I, ... , no}, then
1 I/nl Pdp, ~ 1 max{I/IIP, ... , I/nol P} dp, ,
Ek Ek
and dominated convergence implies (ii).
00
Next suppose that (ii) holds, and write E k = n~k Bn,k. Plainly
E 1 ::JE2 ::J···, and lim

n-+oo
lin (x)1 = (Xl on nE
k= 1
k . Hence p,( nEk) is 0;
k= 1
write Fk = Ek n tOlEk)'. Use (ii) to choose a ko such that for k ~ ko

1 A sequence (If.. I'')~l satisfying (iii) is said to be uniformly integrable.
§ 13. The spaces ~~(l ~ P < ee) 205
and for all n,
For n ~ k o, we have Bn, ko C E k , and so for n ;:?; ko and k ;:?; k o, it follows

that
J Itnl Pdp ~ J Itnl Pdp ~ J Itnl Pdp = J Itnl Pdp < e.
B ... k B ... ko Eko Fko
For n E{I, ... , ko - I}, we have Itnl P~ max{itIIP, ... , Itko-II P} = g,

k,-I
and so J Itnl Pdp ~ Bj;J g dp,
Bn.k
where Bk = U Bn,k = {xEX :g(x) ~ k}.
,,=1
Thus dominated convergence applies, and so (iii) holds if (ii) holds.
Finally, suppose that (iii) holds. Choose ko so large that if k;:?; ko and
n EN, we have
J Itnl Pdp < eP.
Bn.1:
If E E.s;1 and p (E) < kol eP, then

1 1
(J Itnl Pdp)P ~ ( J ItnlPdp)p +
E EnB... ko
Hence (iv) holds if (iii) holds.]

(b) Prove that condition (ii) of part (a) implies conditions (13.38.i)
and (13.38.ii).
(13.40) Discussion. We conclude this section with a study of a
concept of convergence in ~ spaces different from norm convergence.
Thus far we have considered four important concepts of convergence
for sequence of functions: uniform [unif.]; pointwise almost everywhere
[a.e.]; in measure [meas.]; and in the ~p norm [mean-pJ. We have also
expended considerable effort in examining the relationships among these
four types of convergence. Let us summarize our main results. It is
trivial that [unif.] implies [a. e.], in fact "everywhere". Obvious examples
show that the converse fails. Nevertheless, it is easy to infer from (9.6)
that if X is a compact Hausdorff space and (In) is a monotone sequence or
directed family in Irr (X) that converges pointwise to a function t EIrr (X),
then tn -+ t uniformly. [This fact is called DlNI's theorem.] Our most
useful result in this direction is EGOROV'S theorem (11.32). The relation-
ship between [a. e.] and [meas.] was thoroughly examined in (11.26),
(11.27), (11.31), and (11.33). RIEsz's theorem (11.26) is often valuable in
weakening an hypothesis of [a. e.] to [meas. ] [see (13.45) intra and (12.57)].
We have a number of theorems on interchanging the order of limit and
integral, viz. (12.21)-(12.24), (12.30), (13.38), and (13.39). These theorems
can all be regarded as relating [a.e.] to [mean-Pl The relation between
[meas.] and [mean-p] is set down in (13.33). Plainly [unif.] is much
stronger than either [meas.] or [mean-p] on finite measure spaces, but

for infinite measure spaces there is no implication running either way.
We now introduce a fifth kind of convergence for functions in S!.p
spaces, and will study its relations with the notions studied previously.
(13.41) Definition. Let (X, .91, f-l) be a measure space, let 1 ~ P< 00,
and let I and (In)'::=1 be functions in S!.p(X, .91, f-l). If P > 1, then (fn)
is said to converge to I weakly [in S!.p] if
lim J In g d f-l = Jig d f-l
11->00 X X
for every g ES!.P" If P = 1, then (In) is said to converge to I weakly [in S!.1]
if
lim J Ing d f-l = Jig d f-l
n--->oox X
for every bounded d-measurable function g on X.

(13.42) Theorem. Notation is as in (13.41). II 111- Inllp --+ 0, then
In --+ I weakly.
Proof. This follows at once from HOLDER'S inequality. 0
(13.43) Examples. We now give a few examples to show that there
is no connection at all between weak convergence and the four kinds of
convergence discussed in (13.40) [except for (13.42) of course] unless
further hypotheses are imposed on either the sequence or the measure
space. In all of these examples we use Lebesgue measure A.
(a) For each n EN, define In on [0,2n] by In (x) = cos (nx). Then
(In) C S!.p([O, 2n]) for eachp ~ 1. The Riemann-Lebesgue lemma [which
°
we prove in (16.35) inlra] shows that In --+ weakly for all p ~ 1. Since
2"
J 1'; dA =
o
n for all n EN, (12.24) and (12.57) show that In --+ ° for none
of the other four kinds of convergence.
(b) Take In = n~[o,~]. Then (In) C S!.p([O, 1]) for all p ~ 1, In --+ 0
a.e. and in measure, but In ++ ° weakly [take g = ~[o,11].
(c) Let In = n ~[l,eXp(n)]'

1
Then (In) C S!.p (R) for all p~ 1. Write
g(x) = x
1
for x ~ 1 and g(x) = 0 for x< 1. Then g ES!.p,(R) for all
f IgdA = ~ f d: = 1
exp(n}
P> 1, g is bounded, and for all n EN.
° R
Thus In --+ uniformly on R but In ++ weakly in S!.p(R). ° 1
For finite measure spaces, we know that S!.p C S!.1 if P ~ 1 (13.27),

and so uniform convergence implies S!.p-weak convergence for finite
measure spaces.
In spite of the negative results just exhibited, we do have some pos-
itive results if our sequences (In) satisfy certain side conditions.
§ 13. The spaces ~p(l ~ P< 00) 207
(13.44) Theorem. Notation is as in (13.41). Suppose that 1 < p < 00

and that (1Ifnllp):'=l is a bounded sequence 01 numbers. II In ~ I p.-a.e .•
then In ~ I weakly in ~p.
Proof. Choose (X ER such that IItnllp ~ (X for all n EN. By FATou's
lemma (12.23), we have
II/II~ =1 ItiP dp. = x1 .......
x
lim I/nlP dp.
""
(1)
~ lim 1 I/nl dp. ~ (XP •

P
.......""x
Let e > 0 and g E~p. be given. Use (12.34) to obtain 15 > 0 such that
for all E Ed for which p. (E) < 15, we have
1
2(X(J IglP' dp.)P' < ; . (2)
E
Next select A Ed such that p.(A) < 00 and

1
2(X( 1 IgIP' dp.)P' < 36 • (3)

A'
Apply EGOROV'S theorem (11.32) to obtain BEd such that B C A,

P. (A n B') < 15, and In ~ I uniformly on B. Finally. choose no EN
such that n ~ no implies
~ 6
If(x) - In (x) I (p. (B))P Ilgllp' < 3'
for all x EB. Then n ~ no implies
1 6
( 1 II - Inl Pdp.)p Ilgllp, < 3' . (4)
B
Thus, combining (1). (2) [with E = A n B']. (3). and (4) and using
HOLDER'S and MINKOWSKI'S inequalities, we have
IX1 I gdp. - X1 In gdp.1 ~ 1 II - Inllgl dp.

X
= 1 II - Inllgl dp. + 1 II - 1.. llgl dp.
AnB' A'
1
+ 1 II - 1.. llgl dp. ~ III - 1.. llp ( 1 IglP' dp.)P'
B AnB'
1 1
+ 111- 1.. llp(/lgIP' dp.)P' +(1 1/-/.. IPdp.)Pllgllp'

A' B
<~+~+~=e
333
for all n ~ no. 0
(13.45) Corollary. The hypothesis I.. ~ I p.-a.e. in (13.44) can be re-
placed by the hypothesis that In ~ I in measure.
Proof. Assume that In does not converge weakly to I. Choose g Es!'p'

such that
Use (6.84) to find integers n l < n 2 < ... such that
lim j f
k...... oo X
(I - Ink) g dflj = rL • (1)
Next use (11.26) to find a subsequence (lnk)';=! of (fnJk=! such that

lim Ink. = I fl-a.e. It follows from (13.44) that
1-+00 1
lim j J (I - Ink) g dflj = 0.

1...... 00 X '
But this equality is incompatible with (1). 0

(13.46) Remark. Example (13.43.b) shows that neither (13.44) nor
(13.45) is true for the case p = 1. However, if we replace the hypothesis
that (1I/nlll) be a bounded sequence by the hypothesis that Illnlll ~ 11/111>
we get a much stronger conclusion.
(13.47) Theorem. Notation is as in (13.41). Suppose that p = 1,
that In ~ I fl-a. e., and that Illnlll ~ 11/111' Then
(i) J Ilnl dfl ~ Jill dfl lor all E Ed,
E E
(ii) III - Inlll ~ 0,
and
(iii) In ~ I weakly in S!,l'
Proof. Let E Ed. Then FATOU'S lemma (12.23) shows that
lim
n-+oo E
f Ilnl dfl ~ EJill dfl = X
Jill dfl - Jill dfl
E'
Hence lim J
n.....oo E
Ilnl dfl exists and (i) holds.
To prove (ii) , let B > 0 be given. Select A Ed such that fl (A) < (X)
and fill dfl < ; . Use (12.34) to obtain a 0> 0 such that if E Ed and
A'
fl (E) < 0, then fill dfl < ; . Next apply EGORov'S theorem (11.32)
E
to find BEd such that Be A, fl(A n B') < 0, and In ~ I uniformly
on B. Choose no EN such that
[sup I/(x) - In(x)I]' fl(B) < ;
xEB
§ 14. Abstract Banach spaces 209
for all n ~ no. Now apply (i) to get
~ jl/-Inldp.= ~~[jl/-Inldp.+ J I/-Inldp.+ JI/-lnldp.]

x ~ AnW B
~ ~~ [f III dp.+ f Ilnl dp. + fill dp.+ f I/nl dp.] + ;

A' A' AnB' AnB'
= 2 fill dp. + 2 fill dp. + ; < 258 + 258 + ; = 8 •

A' AnB'
Since 8 is arbitrary, we have proved (ii). Conclusion (iii) follows at once

from (ii). 0
(13.48) Note. Theorem (13.42) admits a partial converse involving
no hypothesis about pointwise convergence or convergence in measure.
This converse is easy to prove once certain inequalities are established,
and we postpone it to (15.17) inlra.
(13.49) Exercise. Notation is as in (13.41). Suppose that II/nllp -+ II/lIp.
Prove the following.
(a) If In -+ I p.-a.e., then III - Inllp -+ o.
(b) If In -+ I in measure, then III - Inllp -+ o. [Recall that 1 ~ P < 00.]
(13.50) Exercise. Notation is as in (13.41). Suppose that In -+ I
p.-a. e. and suppose that there is a function g E~t such that Ilnl P ~ g
for all n EN. Prove that for every 8 > 0 there is a set B E.JiI such that
p.(B') < 8 and In -+ I uniformly on B. [Take a hard look at the proof
of (11.30) and then proceed as in (11.32).]
(13.51) Exercise. Let (X, .JiI, p.) be a measure space such that {x} E.JiI
and p. ({x}) > 0 for all x E X. Let I, In E~1 (X, .JiI, p.) for n = 1,2, ....
(a) If In -+ I weakly in ~l> then In (x) -+ I (x) for all x EX.
(b) The converse of (i) is false except when X is a finite set.
(c) If In -+ I weakly, then III - Inll1 -+ O. [Write 10 = I and note that
00
U {x EX: In (x) =l= O} is a countable set. Use (a) and (13.47).]

H=O
(d) For p > 1, find a sequence (/n) C lp such that In -+ 0 weakly but
II/nllp-f-+ o. [See (13.13).]
§ 14. Abstract Banach spaces

We have already defined Banach spaces (7.7) and have met several
specific examples: <t(X) and <to(X) in § 7 and ~p(X,.JiI, p.) in § 13.
In the present section we give a short introduction to the abstract
theory of Banach spaces and prove some important theorems about
these spaces. For a thorough treatment of the subject, the reader is
invited to consult the treatise Linear Operators Part I by NELSON
DUNFORD and JACOB T. SCHWARTZ [New York: Interscience Publishers,

1958]. Throughout this section F will denote either the field R or the
field K.
(14.1) Definition. Let A and B be linear spaces over F. A function T
from A into B is called a linear transformation [or linear operator] if
(i) T(x + y) = T(x) + T(y)
and
(ii) T (IXX) = IX T (x)
for all x, yEA and IX EF. If A and Bare normed linear spaces, a linear
transformation T from A into B is said to be bounded if there exists a
nonnegative real number M such that
(iii) I T (x) I ;;;;; M Ilxll for all x EA
[i.e., T is bounded on the unit sphere of A]. In this case we define the norm
of T to be the infimum of the set of all M's that satisfy (iii), and we write
IITII for the norm of T. This norm is called the operator norm.
(14.2) Theorem. Let A and B be normed linear spaces and let T be
a bounded linear transformation from A into B. Then
IITII = sup { I ~;Ii)11 : x EA, x + o}

= sup{IIT(x)ll: x EA, Ilxll = I}
= sup{IIT(x)ll: x EA, Ilxll ;;;;; I},
and
IIT(x)ll;;;;; IITII'llxll forall x EA.
Proof. Exercise.
(14.3) Theorem. Let A and B be normed linear spaces and let T be
a linear transformation from A into B. Then the following three statements
are mutually equivalent:
(i) T is bounded;
(ii) T is uniformly continuous on A;
(iii) T is continuous at some point of A.
The continuity statements are, as always in this section, understood to be
relative to the metric topologies induced on A and B by their respective norms.
Proof. If (i) holds, we have
IIT(x) - T(y)11 = IIT(x - y)11 ;;;;; IITII '11x - yll
for all x, YEA, and so (ii) follows. Trivially (ii) implies (iii). Next sup-
pose that (iii) holds, say T is continuous at Xo EA. Then there exists a
~ > 0 such that I T(x) - T(xo)11 ;;;;; 1 whenever Ilx - xoll ;;;;; ~. Therefore
Ilxll ;;;;; 1 implies I (bx + x o) - xoll ;;;;; b, and so
1 1
II T xii = b I T (~x + x o) - T (xo) II ;;;;; b .
Thus T is bounded and IITII ;;;;; ~. D
(14.4) Theorem. Let A and B be normed linear spaces over F and let
S8 (A, B) denote the set of all bounded linear transformations from A into
B. Then, with pointwise linear operations and the operator norm, ~ (A, B)
is a normed linear space. Moreover ~(A, B) is a Banach space if B is a
Banach space.
Proof. We prove only that ~(A, B) is complete if B is complete;
the other verifications are routine and we omit them. Suppose that B
is complete and let (Tn) be a Cauchy sequence in S8(A, B). For x EA
we have IITn(x) - Tm(x)11 ~ IITn - Tmll'llxll, and so (Tn(x)) is a Cauchy
sequence in B. Thus for each x EA, there is a vector T(x) EB such that
IITn(x) - T(x)ll-+ O. This defines a mapping T from A into B. For
x,y EA, we have IIT(x + y) - [T(x) + T(y)]11 ~ I T (x+y) - Tn (x+y) I
+ IITn(x) - T(x)1I + IITn(y) - T(y)lI-+ 0, and thus T(x + y) = T(x)
+ T(y). We prove similarly that T(a:x) = a:T(x) for all x EA and a: EF.
Thus T is linear.
Since (Tn) is a Cauchy sequence, there is a positive constant fJ such
that IITnll ~ fJ for all n EN. For Ilxll ~ 1, we have I T (x) I ~ I T (x) -
Tn(x)1I + I Tn (x) I ~ 1 + fJ [for all large n], and hence T is bounded,
i.e., T EQ3(A, B). It remains only to show that liT - Tnll-+ O. Let
e > 0 be given. Choose an integer p so large that m, n ;;?; p implies
e
II Tm - Tnll < 2' Next let x EA be such that IIxll ~ 1 and choose mx EN
such that mx ;;?; p and I T (x) - Tmz (x) I < ; . Then n ;;?; p implies that
I T(x) - Tn (x) I ~ I T(x) - Tm",(x)1I + I Tm", - Tnll < ; + ; = e.

It follows that n ~ p implies
liT - Tnll = sup{IIT(x) - Tn (x) II : IIxll ~ I} ~ e. 0
(14.5) Remark. The reader should notice the similarity between
the above proof and the proof of (7.9).
(14.6) Definition. Let E be a linear space over F. A linear functional
on E is a linear transformation from E into F [where F is regarded as a
one-dimensional linear space over F]. If E is a normed linear space
[and the absolute value is used as a norm on F], let E* denote the space
of all bounded linear functionals on E, i.e., E* = ~(E, F). Since F is
complete, it follows from (14.4) that E* is a Banach space. The space E*
is called the conjugate [adjoint, dual] space of E. The conjugate space E**
of the space E* is called the second conjugate space of E, etc.
(14.7) Discussion. Let E be a normed linear space. There is a so-
called natural mapping of E into E** defined as follows. For x EE,
define x on E* by the rule x(f) = f(x). Simple computations show that
each x is a linear functional on E* and that the mapping x -+ x is a linear
14·
transfounation. Also
sup{lx(f)i: f EE*, 11111 ~ I} sup{l/(x)l: f EE*, 11111 ~ I}
=
~ sup{ll/llllxll : f
EE*, 11111 ~ I} ~ Ilxll
for each x EE. Thus the mapping x -+ x is a bounded linear transfouna-
tion from E into E** of noun ~ 1. Several questions arise. (1) Is this
mapping one-to-one? (2) Does it preserve nouns? (3) Is it onto E**?
(4) Indeed, are there any nonzero elements in E* ? In general none of these
questions have obvious answers; however we are able to answer (1),
(2), and (4) with the aid of the Hahn-Banach theorem, which is next
on our program. Question (3) will be answered in the exercises.
(14.8) Definition. Let E be a real linear space. A real-valued function
p defined on E is said to be a sublinear functional if
(i) P(x + y) ~ P(x) + P(y)
and
(ii) p (OI:x) = OI:p (x)
for all x, y EE and all positive real numbers 01:. Notice that a noun is
a sublinear functional.
(14.9) Hahn-Banach Theorem. Let E be a real linear space and let M
be a linear subspace 01 E. Suppose that p is a sublinear lunctional de-
fined on E and that I is a linear functional defined on M such that I (x)
~ p (x) lor every x EM. Then there exists a linear lunctional g defined on
E such that g is an extension 01 f [i.e., f c gJ and g(x) ~ P(x) lor every
xEE.
Proof. Let 8 be the set of all real functions h such that domh is a linear
subspace of E, h is linear, f c h, and h(x) ~ P(x) for all x Edomh.
Notice that I E8. Partially order 8 by c [recall that a function is a set
of ordered pairs]. Let ~ be any chain in 8 and let h = U ~. Then, with
the help of (2.19), we see that h E8. Applying ZORN'S lemma (3.10),
we see also that 8 has a maximal member, say g. To complete the proof
we need only show that domg = E. Assume that this is false, and let y
be any element in En (domg),. Let G = domg and define H = {x+ OI:y:
x E G, 01: E R}. Clearly H is a linear subspace of E and G H. Let c S
be a fixed, but arbitrary, real number and define h on H by
h(x + OI:y) = g(x) + OI:C.
Then h is well-defined since if Xl + OI:lY = X 2 + 0I:2Y, where xl> X 2 EG
and 01:1,01:2 ER, then (01:1 - 01:2) y = x 2 - Xl EG so that 01:1 = 01:2 and Xl =x2•
S
Clearly h is a linear functional and g h. If we can select c in such
a way that h (x) ~ P(x) for all x EH, then we will have h E8, which
contradicts the maximality of g and will complete our proof. The re-
mainder of our proof is therefore devoted to showing that c can be so
selected.
Our requirement is that g(x) + exc = h(x + exy) ~ P(x + exy) for
all x E G, ex ER. By the linearity of g and the sublinearity of p, this is
equivalent to the two requirements
g ( :) + c ~ p (: + y) for x EG and ex > 0 ,

and
g ( :) + c ;:?; - p (- : - y) for x EG and ex < 0 .
Therefore it is sufficient to have
g(u) - P(u - y) ~ c ~ -g(v) + P(v + y)
for all u, v EG. But we do have
g(u) + g(v) = g(u + v) ~ P(u + v) ~ P(u - y) + P(v + y)
for all u, v EG. Write
a = sup{g(u) - P(u - y): u EG}
and
b = inf{-g(v) + P(v + y) : v EG}.
It is clear that a ~ b. Taking c to be any real number such that a ~ c ~ b,
we complete our construction. 0
(14.10) Remark. The crux of the Hahn-Banach theorem is that the
extended functional is still majorized by p. If this requirement were
not made we could obtain an extension of I simply by taking any Hamel
basis for M, enlarging it to a Hamel basis for E, defining g arbitrarily on
the new basis vectors, and defining g to be linear on E.
(14.11) Corollary. Let E be a real normed linear space and let M be
a linear subspace 01 E. I I I EM*, then there exists g EE* such that leg
and Ilgll = 11111·
Proof. Define p on E by P (x) = 11/11 • IIxli. Then p is a sublinear func-
tional on E and we have I (x) ~ II (x) I ~ p (x) for all x EM. Apply
(14.9) to obtain a linear functionalg on E such that leg and g(x) ~ p (x)
for all x EE. Clearly g EE* and IIgll ~ 11/11. But we also have
IIgll = sup{lg(x)! : x EE, IIxll ~ I}
EM, IIxll ~ I}
;:?; sup{lg (x) I : x
= sup{l/(x)1 : x EM, IIxll ~ I} = 11/11 .
Thus IIgil = 11/11. 0
(14.12) Theorem [BOHNENBLUST-SOBCZYK-SUHOMLINOV]. Let E be
a complex normed linear space and let M be a linear subspace 01 E. II
I EM*, then there exists g EE* such that leg and IIgil = 11/11.
Proof. For each x EM, write I (x) = 11 (x) + i 12 (x) where 11 and 12
are real-valued. An easy computation shows that 11 and 12 are real
linear functionals on M, i. e., Ii (x + y) = Ii (x) + Ii (y) and Ii (exx) = exli (x)
for oe E R. It is also obvious that I/i(x)1 ~ I/(xli ~ IIIII . Ilxll, and so 11

and 12 are bounded and Illill ~ 11/11. Now, regarding E and M as real
linear spaces [simply ignore multiplication by all but real scalars],
we apply (14.11) to obtain a bounded real linear functional gl on E
such that 11 C gl and Ilglll = 11M. Next define g on E by the rule
g (x) = gl (x) - ig1(ix) .
It is easy to see that g is a complex linear functional, e.g., ig (x) = ig1 (x)
+ gl(ix) = gt(ix) - ig1(i(ix)) = g(ix). To see that leg, notice that
for x EM we have
gl(ix) + i/2(ix) = 11 (ix) + i/2(ix) = I(ix) = i/(x)
= -/2(x) + ill (x) = -/2 (x) + ig1(x),
so that gl (ix) = - 12 (x) and therefore g (x) = gl (x) - ig1(ix) = 11 (x)
+ i/2(x) = I(x). We need only show that g is bounded and that Ilgll = 11/11.
Let x E E be arbitrary and write g(x) = r exp(iO) where r ~ 0 and 0 E R.
Then we have
Ig(x)1 = r = exp(-iO) g(x) = g(exp(-iO)x) = gl(exp(-iO)x)
~ IIg111 . Ilxll = 11/111 . Ilxll ~ IIIII . Ilxll .
This proves that g is bounded and that Ilgll ~ 11/11. As in (14.11), it is
obvious that IIIII ~ Ilgll· Therefore Ilgll = 11/11· 0
(14.13) Corollary. Let E be a normed linear space and let S be a
linear subspace 01 E. Suppose that z E E and dist (z, S) = d > O. Then
there exists g E E* such that g(S) = {O}, g(z) = d, and Ilgll = 1. In partic-
ular, il S = {O}, then we have g (z) = Ilzll.
Proof. Let M = {x + oez: xES, oe EF}. Then M is a linear sub-
space of E. Define I on M by I(x + oez) = oed. Clearly I is a well-defined
linear functional on M such that I(S) = {O} and I(z) = d. Also I1I1I
= sup { 11% I:d~zll : x + oez EM, Ilx + oezll =!= O} = sup {II-/+ zll : YES}
= ~ = 1. Apply (14.11) if F = R or (14.12) if F = K to obtain the re-
quired functional g E E*. 0
We now return to the mapping x -c>- X discussed in (14.7).
(14.14) Theorem. Let E be a normed linear space and let :n; be the
natural mapping 01 E into E**: :n; (x) (f) = I (x). Then:n; is a norm-preserv-
ing linear translormation Irom E into E**. Consequently :n; is one-to-one.
Proof. We have already observed in (14.7) that :n; is a bounded linear
transformation from E into E** and that 11:n;11 ~ 1. Let x be any non-
zero element of E. According to (14.13), there is an element g E E*
such that Ilgll = 1 and g(x) = Ilxll. Thus
Ilxll =g(x) ~ sup{11 (x) I :/EE*, IIIII = 1}= 11:n;(x)11 ~ Ilxll,

that is,
lin (x) I = Ilxll .
Clearly lin (0) I = 0 = 11011. We have thus proved that n preserves nonns.
Consequently x =l= y in E implies that Iln(x) - n(y) I = Iln(x - y)11
= Ilx - yll =l= 0, and so n(x) =l= n(y). 0
(14.15) Remark. In view of (14.14), a nonned linear space E is
indistinguishable qua nonned linear space from the subspace n(E)
of E**. The mapping n need not be onto E** [see (14.26)]. In case
n(E) = E**, the space E is called reflexive. Since E** is complete and n
is an isometry, every reflexive nonned linear space is a Banach space.
In § 15 we will show that every 2.p space (1 < p < 00) is reflexive.
We next present three theorems which, together with the Hahn-
Banach theorem, are often regarded as the cornerstones of functional
analysis. These are the open-mapping theorem, the closed-graph theorem,
and the uniform boundedness principle. Several applications of these
theorems will be given in the corollaries and the exercises. Unlike the
Hahn-Banach theorem, these three theorems require completeness.
(14.16) Open mapping theorem [BANACH]. Let A and B be Banach
spaces and let T be a bounded linear transformation from A onto B.
Then T(U) is open in B for each open subset U of A.
Proof. For each e > 0, define A. = {x EA : Ilxll < e} and B. = {y EB:
Ilyll < e}. Let e> 0 be given. We will show that there exists a 15 > 0
such that T (A.) ::::> BiJ. For each n EN, let en = ; ... It is clear that if n
is fixed, then
00
A =.U jA.,.
1=1
[we definejA.,. as in (5.6.£)], and so we also have
00
B= T(A) =.U T(jA.).

1=1 ..
Since B is complete, the Baire category theorem (6.54) implies that not
every T(jA ...), j = 1,2, ... , is nowhere dense. Thus there is a jn EN
such that [T(jnA.,,)r has nonvoid interior. But
[T(A .../2 )]- = 2~.. [T(jn A .,.)]-,
and (X W is open in B if (X =l= 0 and W is open in B. Thus there exists a
v..
nonvoid open set C [T(A.nI 2)]-' It follows that
[T(A.,.)]-::::> [T(A.nI2) - T (A",/2)]-::::> [T(A",/2)]-
- [T(A",/2)]- ::::> v.. - v...l
1 For subsets C and D of B, we write C-D={x-y:xEC, yED}: see
(5.6.£).
Since 0 E~ - ~ and ~ - ~ = U {~- x: x E~} is open in B, there

exists a <5n > 0 such that
(I)
We may suppose that <5n <~

n
for every n EN. We will now show that
B d, C T(A.).
To this end, let y be any element of B d,. We must find an x EA.
such that T(x) = y. By(I), there exists Xl EA., such that Ily -T(xl)11 < ~2'
so that y - T(x l ) EB d,. In view of (I) there exists X 2 EA •• such that
Ily - T(XI) - T(x 2} II < ~3' Continuing by finite induction, we find a
sequence (xn }:,= 1 such that for each n EN, Xn is in A ... and
(2)
Let Zn = Xl + X 2 + ... + xn • For m < n, we have
which has limit 0 as m --+ 00. Thus (zn) is a Cauchy sequence in A ; since A
is complete, there is an x EA such that Ilx - znll --+ O. It is clear that
Ilxll = lim Ilznll

n-+oo
~ £ Ilx,,11 < £ ;k = e,
k=1 k=1
and so x is in A •. It follows from (2) that IIY - T (zn) II --+ O. Since T is

continuous, we have liT (zn) - T (x) II --+ 0, and so y = T (x).
Finally, let U be any nonvoid open subset of A, and let y be any
element of T(U). Then there exists x EU such that T(x} = y. Since U
is open, there is an e> 0 such that x + A. C U. Applying our previous
result, we find that there is a ~ > 0 for which Bd C T(A.}. Therefore
y + Bd C T(x) + T(A.) = T(x + A.} C T(U). Thus Y is an interior
point of T(U), and T(U) is open. D
(14.17) Corollary. If A and B are Banach spaces and T is a one-to-one
continuous linear transformation from A onto B, then T-I is continuous.
Proof. If U is open in A = mgT-I, then (T-I)-l (U) = T(U) is
open in B = dom T-I. D
(14.18) Corollary. Let E be a linear space over F and suppose that
IIII and 1111' are two Banach space norms for E. Then the metric topologies
induced on E by I I and I II' are identical if and only if there exists a
positive constant <X such that
<xllxll ~ Ilxll'
for all x EE.
Proof. Consider the identity mapping on E as a linear transformation

from the Banach space (E, 1111) onto the Banach space (E, 1111'). We leave
the details as an exercise. 0
(14.19) Lemma. Let A and B be normed linear spaces. Then A x B,
with coordinatewise linear operations and the norm
II (x,y)1I = I x ll + lIyll
is a normed linear space. Moreover A x B is complete if and only if both
A and B are complete.
Proof. Exercise.
(14.20) Definition. Let A and B be normed linear spaces. A linear
transformation T: A -+ B is said to have a closed graph if whenever
Xn -+ x in A and T(xn ) -+ yin B we have T(x) = y, i.e., T, as a set of
ordered pairs, is a closed set in A x B.
It is trivial that if T is continuous, then T has a closed graph. The
converse is not always true. However the converse is true if A and B
are Banach spaces.
(14.21) Closed graph theorem. Let A and B be Banach spaces and let T
be a linear transformation from A into B such that T has a closed graph.
Then T is continuous.
Proof. Let G = {(x, T(x): x EA} be the graph of T [actually Gis T).
Then G is a closed linear subspace of the Banach space A x B, and
so G is a Banach space. Let PI and P 2 be the projections of G into A
and B respectively, i.e., PI(x, T(x) = x and P 2 (x, T(x) = T(x) for
all x EA. We have
I PI (x, T(x)11 = IIxll ~ lI(x, T(x)11
and
IIP 2 (x, T(x)11 = I T (x) I ~ I (x, T(x)11
and therefore PI and P 2 are continuous linear transformations. Since T
is single-valued and dom T = A, PI is one-to-one and onto A. It follows
from (14.17) that Pi l is continuous. Clearly T = P 2 0 Pi l , and so T
is continuous. 0
(14.22) Lemma. Let B be a Banach space and let I be a nonvoid set.
r
Let denote the set of all functions I' from I into B such that sup {II I' (t) I :
r,
tEl} < 00 and let I I'll denote this supremum. Then with pointwise linear
operations and the above norm, is a Banach space.
The proof is almost the same as that given in (7.9) and we therefore
omit it.
(14.23) Theorem: Uniform boundedness principle. Let A and B be
Banach spaces and let {T.: tEl} be a nonvoid family of bounded linear
transformations from A into B such that
sup{IIT.(x)1I : tEl} < 00
for every x EA. Then

sup{IIT.11 : tEl} < 00 •
Proof. Let rbe as in (14.22). Define a mapping 5: A -+ rby

5 (x)(t) = T.(x) for x EA, t EI.
Our hypothesis that the family {T.: tEl} is pointwise bounded shows
r
that 5 (x) E for each x EA. Clearly 5 is linear. We show that 5 is con-
tinuous by using the closed graph theorem. Thus suppose that Xn -+ x
in A and that 5 (xn) -+ yin r.
For each tEl, we have
Ily(t) - 5 (x)(t) I ~ Ily(t) - 5 (xn}{t) I + 115 (xn){t) - 5 (x)(t) I
~ Ily - 5 (xn) I + 11T.(xn ) - T. (x) I .
The last expression has limit 0 as n -+ 00 because T. is continuous. There-
fore y (t) = 5 (x)(t) for all tEl, and so y = 5 (x). This proves that the
graph of 5 is closed, and hence 5 is continuous, i.e., 11511 = sup{115(x)ll:
Ilxll ~ I} < 00. Since
IIT.II = sup{llT. (x) I : Ilxll ~ I}
= sup{115 (x)(t) I : Ilxll ~ I}
~ sup{115 (x) I : Ilxll ~ I} = 11511
for every tEl, we conclude that
sup{IIT.11 : tEl} ~ 11511 < 00. 0
(14.24) Corollary [Banach-Steinhaus theorem]. Let A and B be
Banach spaces and let C1;,):'= 1 be a pointwise convergent sequence of bounded
linear transformations from A into B. Then the mapping T: A -+ B
defined by
T(x) = lim Tn(x)
~co
is a bounded linear transformation.

Proof. It is obvious that T is linear. It is also clear that for each
x EA we have
sup{llT.. (x) I : n EN} < 00 .
It follows from (14.23) that there exists a positive constant M such that
IIT..II ~ M for all n EN. Thus x E A implies
I T(x)11 = n--+co
lim liT.. (x) II ~ Mllxll '
and so T is bounded and II Til ~ M. 0

(14.25) Exercise. Let D be a nonvoid set and let co(D) denote the
set of all complex-valued functions f defined on D such that for each
e > 0 the set {x ED: If(x)1 s e} is finite. Thus co(D) = <ro(D) where D
is equipped with the discrete topology [see (7.12) J. Define linear operations
pointwise on co(D), and for IE Co (D) define 11/11 .. = sup{l/(x)/ : xED}.

(a) Co (D) is a Banach space.
(b) co(D) is separable if and only if D is countable.
(c) If qJ is a bounded linear functional on co(D), then there exists
a function g EII (D) [see (13.13)] such that
qJ(I) =}; I (x) g(x)
xED
for all I ECo (D).
(d) The mapping qJ -+ g described in (c) is a norm-preserving iso-
morphism from Co (D) * onto II (D).
(14.26) Exercise. Let D be a nonvoid set and let loo (D) denote the
set of all bounded complex-valued functions defined on D. [The space
loo (D) is denoted m (D) by some writers.] Define linear operations point-
wise in loo(D), and for I Eloo(D) define
11/11 .. = sup{l/(x)1 : xED}.
Notice that loo (D) = <reD) where D has the discrete topology. Notation
is as in (14.25). Prove the following.
(a) If qJ Ell (D)*, then there exists a gin loo (D) such that
qJ(I) = 1.: I (x) g(x)
"ED
for all I Ell (D).
(b) The mapping qJ -+ g described in (a) is a norm-preserving iso-
morphism from ~(D)* onto loo(D).
(c) If D is infinite, then co(D) is not reflexive. [Compute the natural
mapping of Co (D) into loo(D) explicitly.]
The density character of a topological space X is the smallest cardinal
number such that there exists a dense subset of X having that cardinal
number.
(d) If D is infinite and jj = b, then both Co (D) and II (D) have density
character b, but loo(D) has density character 2b.
(e) If D is infinite, then there exists an element qJ E loo (D)* such that
qJ (I) = 0 for all I E Co (D) and II qJll = 1. [Use an extension theorem.]
(f) If II (D) is reflexive, then D is finite. [Use (b) and (e). For an
explicit computation of loo (D)*, see (20.27) -(20.35) inlra.]
(14.27) Exercise. Let E be a real linear space and let PeE be such
that
(i) x, yEP and (x, {J ~ 0 imply (Xx + {Jy E P;
(ii) x E P and - x E P imply x = o.
Then P is called a convex cone. For x, y E E, define x ~ y to mean that
y - x E P. Prove that ~ is a partial ordering on E. Let S be a linear
subspace of E such that for all x EE, (x + S) n P =1= 0 if and only if
( - X + 5) n P =l= 0. Suppose that I is a linear functional on 5 such that

xES and x ~ 0 imply I (x) ~ o. Prove that I can be extended to a
linear functional g on E such that x E E and x ~ 0 imply g(x) ~ o.
[Use the Hahn-Banach theorem, defining
P(x) = inf{/(Y):Y ES,y ~ x}
for all x in the linear span of 5 U P. Alternatively, give a direct proof
using ZORN'S lemma.] This result is known as KREIN'S extension theorem
lor nonnegative linear lunctionals.
(14.28) Exercise. Prove that there exists no sequence (Cn):=l of
00
complex numbers such that an infinite series I: an of complex numbers

n=l
converges absolutely if and only if (cnan):=l is a bounded sequence.
[Assume such a sequence exists with Cn =1= 0 for all n. Consider the
mapping T:loo(N)--+ll(N) given by T(f)(n) = fIn) and use the open
en
mapping theorem.]
(14.29) Exercise. Let E be a normed linear space and let D be a
nonvoid subset of E such that
sup{11 (x) I : xED} < 00
for each IE E*. Prove that sup{llxll : xED} < 00. [Consider n(D) CE**.]
(14.30) Exercise. Let E be a Banach space and let A and B be
closed linear subspaces of E such that A n B = {o}. Prove the following.
(a) If A + B is closed in E, then the mapping x + Y --+ x for x E A,
Y E B is a continuous linear mapping of A + B onto A.
(b) For x E A, Y E B let Ilx + yll' = Ilxll + Ilyll· Then 1111' is a complete
norm for A + B.
(c) The set A + B is closed in E if and only if IIII and 1111' induce the
same topology on A + B.
(14.31) Exercise. (a) Let E be a normed linear space and let M be
a closed linear subspace of E. Suppose that z EE n M'. Let 5 = {x+az:
x EM, a EF}. Thus 5 is the smallest linear subspace of E containing
M and z. Prove that 5 is closed in E. [Define I on 5 by I(x + az) = a.
Show that IE 5* and IIIII ;;:;:; dist(~,M) . Then use the fact that F is
complete.]
(b) Prove that every finite-dimensional linear subspace of E is closed
in E. [Use (a) and induction].
(14.32) Exercise. Prove that there exists no Banach space of al-
gebraic dimension No. [Use (14.31) and the Baire category theorem.]
For an arbitrary nonzero cardinal number m [finite or infinite], construct
a normed linear space of algebraic dimension m.
(14.33) Exercise. Let A and B be Banach spaces and let T be a

linear transformation from A into B such that goT E A * for every
g EB*. Prove that T is continuous.
(14.34) Exercise. Let A and B be normed linear spaces and let T
be a bounded linear transformation from A into B. For g EB*, define
T* (g) = goT. [The transformation T* is frequently called the adjoint
01 T.] Prove that:
(a) T* is a bounded linear transformation from B* into A *;
(b) IIT*II = II Til ;
(c) T* is one-to-one if and only if T(A) is dense in B;
(d) T is one-to-one if and only if T* (B*) separates points of A.
(14.35) Exercise. Let A be a normed linear space.
(a) Prove that there exists a Banach space B and a norm-preserving
linear transformation T: A -+ B such that T(A) is dense in B. [Consider
the natural mapping of A into A **.]
(b) Prove that if B1 and B2 are any two Banach spaces having the
property ascribed to B in part (a), then there exists a norm-preserving
linear transformation from B1 onto B 2.
(14.36) Exercise. Let I be a nonvoid set, and for each tEl, let E,
be a normed linear space over F. Let p be a real number ~ 1. Let E
be the set of all x = (XI) E X EI such that I; Ilxili P < 00, and for all
.EI lEI
1
X EE, let Ilxll = [I; Ilx.IIP)p. With linear operations (x + y)(t) = x(t)
lEI
+ Y (t) and (exx) (t) = ex (x (t)) and the norm just defined, E is a normed
linear space. The space E is a Banach space if and only if each E. is.
Prove the preceding two assertions.
(14.37) Exercise. Let E be a finite-dimensional linear space and let
11111 and 11112 be norms on E. Prove that there are positive numbers ex
and fJ such that
exllxl11 ~ IIxl1 2 ~ fJllxl11
for all x EE. Thus all pairs of norms on E are "equivalent", and all
norms make E a Banach space.
(14.38) Exercise. Let E be a normed linear space and let M be a
closed linear subspace of E. Consider the quotient space EjM = {x + M:
x EE}, where linear operations are defined by
(x + M) + (y + M) = (x + y) + M
and
ex (x + M) = (exx) +M .
Define the quotient norm on EjM by the rule
(i) Ilx + Mil = inf{llx + mil: m EM}.

(a) Formula (i) defines a norm on ElM.
(b) If E is a Banach space, then so is ElM. [For a given Cauchy
sequence in ElM, choose a subsequence (xn + M):'=l such that
Ilx/< - Xn + Mil < 2- n for k ~ n. Then choose Zn EM such that Ilxn+1 - Xn
+ znll < 2- n for each n. Let Yn = Xn + zn-l + Zn-2 + ... + Zl' Prove that
(Yn) is a Cauchy sequence in E, letYn -* Y, and prove that Xn + M -* y+M
in ElM.]
(c) The natural mapping q; of E onto ElM: q;(x) = x + M, is a
bounded linear transformation with I q;11 ~ 1, and q; sends open sets
onto open sets.
(d) If E is a Banach space, then q; maps the open unit ball of E onto
the open unit ball of ElM.
(e) If M and ElM are both complete, then so is E.
(14.39) Exercise. Let E be a separable Banach space, let (Xn );;:l
be a countable dense subset of E, and let B = {I EE*: IIIII ~ I}. For
I, g EB, define
Prove that
(a) e is a metric for B,
and
(b) with the metric e, B is a compact metric space.
Let P be CANTOR'S ternary set.
(c) Prove that there exists a norm-preserving linear transformation
T from E onto a closed subspace of <r(P), where <r(P) has the uniform
norm. [Use (b) and (6.100) to obtain a continuous mapping q; from P
onto B. Define T(x)(t) = q;(t)(x), for x EE and t E P.]
(d) Prove (c) with P replaced by [0, IJ.
(14.40) Exercise. Let E be a normed linear space and let S be a
linear subspace of E that is dense in E. Suppose that I is a bounded linear
functional on S. Without recourse to the Hahn-Banach theorem, prove
that there exists a unique g EE* such that g(x) = I (x) for all xES.
Prove also that Ilgll = 11/11·
§ 15. The conjugate space of 2p (1 < P < 00)

In this section we construct the conjugate spaces of an important,
if special, class of Banach spaces. Throughout this section (X,.JiI, ft)
denotes a fixed but arbitrary measure space and p a fixed but arbitrary
real number such that 1 (1 < P< 00) 223
(15.1) Theorem. Let g ES!p' and define Lg on S!p by the rule

Lg (I) = J fg dp. for all I ES!p .
x
Then Lg ES!t and I L,II = Ilgllp'·
Proof. By HOLDER'S inequality (13.4) we have
Thus Lg is a bounded linear functional on S!p [the linearity of Lg is

obvious], and
IILgl1
Ilgllp'· ~
In fact, equality holds here, i.e., IILgl1 = Ilgllp'. To see this, let I be
IgIP' -1 sgn (g); then we have ItiP = IgI P'. Thus I is in S!p, and
P'
1I/IIp = Ilgll: = Ilgll~:-l .
Hence the equalities
Lg (I) = Jig dp. = J IgIP' -1 sgn (g) gdp. = J IgI P'dp.
= Ilgllf = IIgll~:-l Ilgllp' = 1I/IIp Ilgllp'
hold, and so also IILgl1 ~ Ilgllp'. Hence we have IILgl1 = Ilgllp,.1 0
(15.2) Remark. Our goal in this section is to prove that every bounded
linear functional on S!p has the form Lg for some g ES!p'. It will then follow
that S!; and S!p' are indistinguishable as Banach spaces, even though,
naturally, they consist of quite different entities. Our proof is elementary,
making use of no sophisticated facts. Only techniques of the calculus
are needed.
(15.3) Lemma. Suppose that p ~ 2. Then the inequality
(i) C~ y+ C; y~ ~ (1 +
x x x P)
obtains lor aU x E [0, 1].
Proof. Define F(x)=(11xy+(I;xy-! (l+xP). We must
show that F(x) ~ 0 for 0 ~ x ~ 1. Since F(O) = 2- 1 (2- P+2- 1) and
p ~ 2, we have F(O) ~ O. For 0 < x ~ I, it is convenient to consider
the function (/) defined by
(1)
Thus
(/) (x) = [(! + 1 + (! - 1 r r- 2 P- 1 (:/> + 1)] ;
clearly (/)(1) = o. Let us prove that (/)' (x) ~ 0 for 0 < x < 1. This deriv-
1 We could just as well have used gas g in the definition of the functional L •.
For the case p = 2, g is more natural, as we will see in § 16, and so we keep it here.
ative is
(/)' (x) = - X;+1 [(1 + X)I>-1 + (1 - X)I>-1 - 21>-1] • (2)
Write IX = P- 1 [note that IX ~ 1], and consider the function lJI de-
fined by
We have
lJI'(x) = IX(1 + X)ot-l_ IX(I- X)ot-l ~ 0 for 0 < x < 1.
Thus lJI is a nondecreasing function on [0, 1], and since lJI(I) = 0, the
mean value theorem implies that lJI (x) ~ 0 for 0 ~ x ~ 1. Going
back to (2), we infer that (/)' (x) ~ 0 for 0 < x < 1, and since (/)(1) = 0,
(/)(x) is nonpositive for 0 < x < 1. The definition (1) shows that F(x) is
also nonpositive for 0 < x < 1. 0
(15.4) Lemma. Let z and w be complex numbers, and suppose that
p ~ 2. Then we have
(i) z+w II> +-2-
I-2-
Iz-wll> ~2+-2-·
\zIP Iwlt>
Proof. If w = 0, the inequality becomes 21:~1 ~ I~t> , which holds

since p - 1 ~ 1. Thus we may suppose that Izl ~ Iwl > o. The inequality
(i) is equivalent to
(1)
which we will now prove. The inequality (1) can be written in the form
r
11 + r~Xp(ilJ) + 11 -r~p(ilJ) II> ~ ~ (1 + rl» (2)
where 0 < r ~ 1 and 0 ~ 0 < 2n. If 0 = 0, the inequality (2) is just

(I5.3.i). The proof will be complete if we show that the left side of (2)
is a maximum when 0 = 0, for fixed r. Clearly we may consider only 0
such that 0 ~ 0 ~ ~ . We must show that the function g defined by
g(O) = 11 + r exp(iO)!f> + 11 - r exp(iO)!f>
has a maximum on [0, ~] at 0 = o. We have
I> I>
g(O) = [1 + r2+2rcos{O)]2 + [1 + r2 - 2rcos(O)]2
and so
I>
g' (0) = ~ (1 + r2 + 2r cos (0))2- 1(- 2rsin (0))
I>
+~ (1 + r2 - 2r cos (0))2- 1(2r sin (0))
I> I>
= - prsin(O)[(I +r2 +2rcos(O))2- 1 - (1 +r2-2r cOS(O))2-1]. (3)
§ 15. The conjugate space of .£1>(1 <p< 00) 225
Since p ~ 2, it is clearfrom (3) that g' (0) ~ °for 0 E[0, ~]. Therefore
the function g is nonincreasing in [0, ~l i.e., g assumes its maximum
value at 0. 0
(15.5) CLARKSON'S inequality for p ~ 2. For p ~ 2 and I, g E2.p,

we have
(i) I f ~ gil: + /I f ~ II: ~ ~ II/II~ + ~ Ilgll~·

g
Proof. We may suppose that I and g assume complex values and are
defined p,-a.e. [(12.18) and (12.26)]. Then for all x EX such that I(x)
and g (x) are defined, (15.4.i) implies that
f(x) + g(x) IP I f(x) -g(x) IP < If(xW Ig(x)IP

(1)
I 2 + 2 = 2 +2'
Integrating both sides of (1) over X, we obtain (i). 0
There is an analogue of (15.5.i) for 1 < p < 2, which we establish
next. The inequality and its proof, for some reason, are more com-
plicated than for p ~ 2.
(15.6) Lemma. Suppose that 1 < P ~ 2. Then the inequality

I
(i) (1 + x)P' + (1 - x)P' ~ 2(1 + XP)P-I
obtains lor all x E [0, IJ.
Proof. The result is trivial if p = 2. Thus we suppose that 1 < p < 2.
For x = ° and for x = 1, (i) becomes an equality. As u runs from ° to 1,
the function ~ ~: decreases [strictly] from 1 to 0. Hence our desired
inequality (i) is equivalent to
(1+I+u + (1 _ ~)P' S 2(1

+ (II+u
I
I - u )P' - u )P)P=I ( 1)
I+u -
for °< u < 1. Multiplying both sides of (1) by (1 + u)P', we obtain

I
2P' (1 + uP,) ~ 2 [(1 + u)P + (1 - u)P] P-I . (2)
Raising both sides of (2) to the (p - 1)8t power, we get
(1 + UP')P-I ~ ~ [(1 + u)P + (1 - u)P] , (3)
for °< u < 1. It is clear that the steps going from (i) to (3) are reversible,
so that we need only to prove (3). Expanding in power series, we have
~ [(1 + u)P + (1 - u)P] - (1 + UP')P-l

= ~2 [£ (P)Uk+
k~O k
£
(P)(_I)k Uk] _
k~O k k~O
(P - I)UP'k
k
£
=k~[(:k)U2k_(P ~ I)UP'k]
= £ [(
k~l
P )U2k
2k
_ (P - 1 )UP'(2k-l) _ (P -
2k - 1 2k
I) uP' 2k] (4)
As shown in (7.25), the last line of (4) converges absolutely and uni-
formly for U E [0, I]. We will show that each term [... J in this series
is nonnegative. Plainly this will prove (3). The kth term is
P (P - I) (P - 2) ... (p - (2k - I)) 2k _ (P-I) (P - 2) ... (p - (2k - I)) P'(2k-l)

(2k)! U (2k _ I)! U
_ (P - I) (P - 2) ... (P - 2k) UP'2k

(2k) !
= P(P-I) (2-P) .. , (2k-I-P) 2k_ (P-I) (2-P) (3-P) .. ·(2k-I-P) P'(2k-l)

(2k)! U (2k _ I)! U
(P-I)(2-P)···(2k-P) P'2k
+ (2k)! U
_ U2k (2-P) (3-P)'" (2k-P)
- (2k-I)!
X [ P (P-I) - (P-I) UP'(2k-l) -2k + (P-I) UP'2k-2k]

(2k) (2k - P) (2k -p) (2k) .
The first factor here is obviously positive. Rewrite the expression in

brackets as
2k-P 2k ]
1 1 1 P-l 1 p-l
[
2k-p - 2k - 2k-p U + 2k U
p-I p-I p-I p-I
[ 1_u2:~i _ l-u R] (5)

- 2k-p 2k'
p-I p-I
An elementary argument [which the reader should carry out J shows

that for any U > °the function with values 1 ~ ut , °< t < 00, is de-
. 0 f t. S'mce
creasing as a functIOn 2k - P
~ < p 2k . f0 11ows t 1lat ().
_ 1 ,It 5 IS
positive. D
§ 15. The conjugate space of ~p(1 <p< (0) 227
(15.7) Theorem. Let z and w be complex numbers, and suppose that

1< P ~ 2. Then we have
I
(i) Iz + wlP' + Iz - wlP' ~ 2(/z1P + IwIP)P-I .
°
Proof. If z = or w = 0, (i) is obvious. Otherwise, we may suppose
°
that < Izl ~ Iwl. The desired inequality is thus equivalent to the in-
equality
t' t'
I
11 + ~ + 1-1 + ~ ~ 2 (I ~ IP+ 1)t>=I. ( 1)

Write (1) in the form
I
11 + r exp(ieW' + 1-1 + r exp(ieW' ;£; 2(rP+ 1) P-I , (2)
where ~ = r exp(ie), 0 < r ~ 1, and 0 ~ e < 2n. For e= 0, the in-

equality (2) is just (15.6.i). Just as in the proof of (15.4.2), one shows
that the expression on the left in (2) attains its maximum on [0, ;] at
e= 0. Thus (2) holds for all e. 0
(15.8) CLARKSON'S inequality for 1 < p < 2. For functions f and g
in 5!p, the inequality
I
(i) I f ~ gil:' + I f ; g II:' ~ [~ IIfll~ + ~ IIgll~] p-I

holds.
Proof. By MINKOWSKI'S inequality for 0 < P< 1 (13.9), we have
The left side of (1) is the left side of (i), since IIIW' lip-I = Ilhll~' for
any h E 5!p. The right side is
which by (15.7) is less than or equal to
[f 2P-l (I ~ IP + I; n I
dfl]t>=I = [~ IIfll~ + ~ IlglI~]t>=I.

I
Throughout (15.9)-(15.11), p is fixed and greater than 1, 5!p denotes

an arbitrary 5!p (X, .511, fl), and L is an arbitrary bounded linear functional
on 5!p different from O.
(15.9) Theorem. There is a function CPo E 5!p such that II CPo lip = 1 and
L (CPo) = II L II, that is, L assumes a maximum absolute value on the unit
ball of 5!p.
15*
Proof. The definition (14.1) of IILII shows that there is a sequence

(!P~)~l in S!p such that 11!p~llp = 1, IL(!p~)1 > ~
2
liLli, and n-+oo
lim IL(!p~)1
= I L II· Let !Pn = sgn [L (!p~) ] !p~. Then we obviously have:
L(!Pn) = IL(!p~)1 > ~ IILII > 0; (1)
I !Pnllp = 1; (2)
lim L(!Pn)
n-+oo
= IILII . (3)
We will show that (!Pn) is a Cauchy sequence in 5!-p. In the contrary case,
there are a positive number (X and subsequences (!pnk)'k~l and (!pmk)k~l
such that I !pnk - !pmkllp > (X for k = 1,2, .... For p ;;?; 2, we use CLARK-
SON'S inequality (15.5) to write
(4)
For 1 < p < 2, we use CLARKSON'S inequality (15.8) to write
[~ I !pmkll~ + ~ I !pnJ~p-l
1
I q?mk ; q?nk II:' + I q?mk~ q?nkII:' ~ = 1. (5)

For p ;;?; 2, the inequality (4) implies that
(6)
and for 1 1, a number {3 EJO, 1 [ that
is independent of k and such that
(8)
for k = 1,2, .... Consider the sequence of functions (gk)k~l defined by

q?mk + q?nk (9)
gk = IIq?mk + q?nkllp
No denominator in (9) is zero, for otherwise we would have !pnk = - !pmk'
and hence the equality L (!pnk) = - L (!pm k) would hold, contradicting (1).
For k = 1, 2, ... , (8) and (9) show that
1 [1 1 ]
L (gk) = I q?mk ; q?uk lip 2 L (!pm k) + 2 L (!pnk)
> 1 -1 f3 [ 21 L (!pmk) + 21 L (!pnk) ] . (10)

§ 15. The conjugate space of ~p(1 <p< (0) 229
By (3) we have lim L(gJmJ = lim L(gJnk ) =

k-:;.oo k-+oo
IILII. Thus (10) implies that
lim L (gk)
k--;.oo
~ 1 ~ f3 IILII .
Since Ilgkllp = 1, this is an evident contradiction. Therefore (gJn) is a
Cauchy sequence in S!p and so has a limit gJo in S!p (13.11). It is clear
from (3) that lim L (gJn) = L (gJo) = IILII. 0
n--;.oo
(15.10) Lemma!. Let E be a complex normed linear space, and let L

be a nonzero bounded linear functional on E for which there exists gEE
satisfying the conditions Ilgll = 1 and L (g) = IILII. Consider the function
(i) t -7 Ilg + t fll = 1pf (t)
defined on R, where f is any element of E. If 1pf and 1p-if are differentiable
at t = 0, then we have
(ii) II~II L (f) = 1p; (0) + i1p'---if (0).
Proof. Suppose without loss of generality that IILII = 1. For any
complex number z, we have
L(g + z(f - L(f) g») = L(g) + z(L(f) - L(f) L(g) = L(g) = 1.
Since IL(h)1 ~ Ilhll for all h EE, it follows that
Ilg + z(f - L (f) g)11 ~ 1 for all z EK . (1)
For each t ER different from .Z(~)' write g + tf in the form g + t f
= (1 + tL (f) [g + t 1 + ~L (f) (f - L (f) g) J. The norm of the expression

in brackets is greater than or equal to 1 for all t, and for t = 0 it is equal
to 1. Hence
Ilg+ tfll-llgll ~ 11 + tL(f)I-l
1
= [(1 + tRe(L(f))2+(tIm(L(f)))2]2 -1
~ 1 + t Re(L (f) - 1 = t Re(L (f) ,
which implies that
Ilg + tf~1 - Ilgll ~ Re(L (f) if t > 0, (2)
and
Ilg + tfJI-llgll ~ Re(L(f) if t< o. (3)
Both (2) and (3) follow trivially from (1) if L (f) = o. It follows that
1p; (0) = Re(L (f) for all fEE. (4)
1 This lemma is due to E. J. MCSHANE [Proc. Amer. Math. Soc. 1 (1950), p. 402].
Applying (4) to the function -ii, we obtain

tp~i/(O) = Re(L(-i/)) = Im(L(f)); (5)
and (4) and (5) imply (ii). 0
(15.11) Theorem [F. RIEsz]. Let L be a bounded linear lunctional
on s!'p (1 < P < (0). There is a lunction h ES!,p' such that L(f) = f Iii df-'
x
lor all I Es!'p.
Proof. The result is trivial for L = 0, so we suppose that L =l= O.
Using (15.9), select a function g Es!'p such that L (g) = IILII and Ilgllp = 1.
We want to apply (15.10), and to do this we must show that the function
t-+ Iitl + gllp = tpf(t)
is differentiable at t = 0 for every I Es!'p. Let w (t) = tpf (t) = f It 1+ glP df-'.
x
Writing I = 11 + i/2 and g = gi + ig2, we have
P
It I + glP = [(tIl + gI)2 + (t12 + g2)2]2,
and so almost everywhere on X we have
d
dt It I + glP = PIt I + gIP-2[(tli + gI) 11 + (t12 + g2) 12] (1)
for all t. [If 1 < P < 2 and the points x EX and t ER are such that
tl(x) + g(x) = 0, then the first factor in the above expression for
:t It I + glP is undefined, and the second factor is zero. In this case,
as the reader can check, the derivative is actually zero.] For every
t =l= 0, we have
OJ(t) - OJ(O) =
t t f-'.
f Itt +glp -Iglp d
(2)
x
Using the mean value theorem and (1) to rewrite the integrand in (2),
we have
OJ(I) ~ OJ(O) = f
P It'l + g1P-2 [(t'/I + gI) 11 + (t'/2 + g2) 12] df-' (3)
x
where 0 < WI < It I and t' is a lunction 01 x EX. [If 1 < P < 2 and t'/(x)
+ g(x) = 0, then the integrand is zero.] Since (t'I, + g,) ~ It'l + gl
and I, ~ III, the absolute value of the integrand in (3) is less than or
equal to 2p It' 1+ glP-l III. If It I ~ 1, then we have 2p It' 1+ glP-l III
~ 2P(1/1 + Igl)P-1I/I· The functions III and Igi are both in S!,p, and so
([II + Igl)P-l is in S!,P" and (III + Igl)P-1I/I is in S!,l' by HOLDER'S inequality
(13.4). Thus for all It I ~ 1, the integrand in (3) is less than or equal to
the fixed function 2p ([II + Igl)P-1I/I, which is in S!,l' LEBESGUE'S theorem
§ 15. The conjugate space of 21>(1 <p< 00) 231
on dominated convergence (12.24) implies that
~~ f
Ig + ttl; -Igll> dp = f
P /g/P-2[gdl + g2f2] dp. (4)
x x
[If 1 < p < 2 and g(x) = 0, then the integrand in (4) and in the following
integrals is zero.] Combining (2) and (4), we see that w'CO) exists and
that
w' (0) = f P /g/P-2 [glfl + g2f2] dp . (5)
x
Consequently "PI (0) also exists. Using (5), we write
(!
1
"PI (0) = ~ /g/Pdp ) p-I . w' (0) = ~ IIgll}-P w' (0)
= f
/g/P-2 [gdl + g2f2] dp . (6)
x
Lemma (15.10) and (6) imply that
L (I) = IILII ("PI (0) + i"P'--i/(O))
= IILII f /g/P-2((gdl+g2f2)+i(glf2-g2fl))dp=II L II' f /g/P-2gjdp.
x x
The theorem follows when we set h = IILII . /g/P-I sgn(g); i.e.,
f fit dp. 0
L (I) =
x
(15.12) Theorem. Let (X, .91, p) be an arbitrary measure space and let
p be a real number such that 1 < p < 00. Then the mapping T defined by
T(g) = Lg
[see (15.1)] is a norm-preserving linear transformation from !i!.p' onto !i!.t-
Thus, as Banach spaces, !i!.p' and !i!.; are isomorphic.
Proof. The fact that T is a norm-preserving mapping from !i!.p'
into !i!.: is (15.1). It follows from (15.11) that T is onto !i!.t- It is trivial
that T is linear. Since T is linear and norm-preserving, T is one-to-one. 0
(15.13) Exercise [J. A. CLARKSON]' Let (Y, .91, p) be a measure
space such that .91 contains two disjoint sets of finite positive measure.
There is a [unique] least positive number c such that
~ < lit + gil; + lit - gil; <
c = 2 (11tll; + Ilgll;) = c
for all f, g E !i!.p (1 < p < (0) such that IIfllp and IIgllp are not both zero.
12-PI 1
Prove that c exists and that c = 2 P . Also, the constants c and -
c
are attained.
(15.14) Exercise. (a) Let (X, d, ft) be a measure space, let p be a

real number such that p > 1, and let I be an d-measurable function on X
such that:
(i) {x EX: I (x) =l= O} is the union of a countable number of sets in
d having finite measure;
(ii) I g E .£1 (X, d, ft) for all g E .£p (X, d, ft)·
Then I is in .£p,(X, d, ft). [Hints. Construct a sequence of functions
(fn)':~l such that (1Inl)::'~l is nondecreasing, Ilnl-* III everywhere, and
each In vanishes except on a set of finite measure. Then use (12.22),
(15.1), and (14.23) to infer that I E .£p,.J
(b) [E. B. LEACH]' Let (X, d, ft) be a measure space and suppose
that every set A in d such that ft (A) = (Xl contains a set BEd for
which 0 < ft (B) < 00. Let I be any d-measurable function on X for
which (ii) above holds. Then I is in .£p' (X, d, ft). [Hints. Let An
= {x EX: II (x) I;::;; :}. If ft (Am) = 00, use (1O.56.d) to find a subset C of
Am such that C Ed, ft (C) = 00, and C is a-finite. Then no
satisfies (i)
above and also (ii), since I satisfies (ii). It follows that no
E .£p', a contra-
diction. Hence I satisfies (i), and (a) applies. J
(15.15) Exercise. (a) Let (X, d, ft) be the measure space described
in (1O.56.b). Show that the conclusion of (15.13) fails for this measure
space for each p such that 1 < p < (Xl.
(b) Let (X, d, ft) be a measure space for which there is a set D Ed
such that ft (D) = 00 and no d-measurable subset of D has finite posi-
tive measure. Prove that there is an d-measurable, nonnegative, real-
valued function I on X such that (15. 14.ii) holds and I is in no .£r
(0 < r < (Xl).
(15.16) Exercise. Let E be a [real or complexJ normed linear space
such thatfor all s > 0 and x, y E E such that Ilxll = IIYII = 1 and Ilx - yll > s,
the inequality
(i) II ~ (x + y)11 ~ (1 - b)
obtains, where 15 = 15 (s) is independent of x and y and 0 < 15 < 1. Such

spaces are called unilormly convex [by some writers unilormly rotund].
(a) Let E be a uniformly convex Banach space and L a bounded
linear functional on E. Prove that there is an x E E such that Ilxll = 1
and L(x) = IILII. [Imitate the proof of (15.9), noting that (15.9.8) is
simply the assertion that .£p is uniformly convex. J
(b) Prove that a uniformly convex Banach space is reflexive. [Use
MCSHANE'S lemma (15.1O).J
(c) Let E be a uniformly convex normed linear space, let 5 be a
proper linear subspace of E that is complete in the norm on E1, and let x
1 For example, 5 can be any closed subspace of E if E is a Banach space, or
any finite-dimensional subspace of an arbitrary E [see (14.31.b) and (14.37)J.

§ 15. The conjugate space of ~I>(I <p< (0) 233
be any element of E n 5'. Show that there is one and only one element
Yo E5 such that
Ilyo - xii = inf{lly - xii: Y E5} .
(15.17) Exercise: Weak convergence and the RAnoN-lb:Esz theorem.
In (13.41), we defined weak convergence for sequences of functions in ~p
and thereafter studied relations between weak convergence and other
sorts of convergence. Weak convergence can be defined for sequences
(xn) in any normed linear space E, as follows. The sequence (xn) converges
-+
weakly to x EE if I (xn) I (x) for all I in the conjugate space E* of E.
Theorem (15.11) shows that the present definition is consistent with
the definition offered in (13.41) for E = ~p (1 
see (20.19) inlra.
(a) Let E be a normed linear space [over R or K] with the property
that if (xn) is a sequence in E, x EE, Ilxnll = 1, and Ilxll = 1, then the
relation
IIX
n
:xll-+ 1
implies
Ilxn - xii -+ 0 .
[Such spaces are called locally unilormly convex.]. Let (Yn) be a sequence
in E and y an element of E such that:
(i) IIYnl1-+
Jlyll ;
-+
(ii) Yn Y weakly.
Prove that llYn - yll -+
o. [If Jlyll = 0, the assertion is trivial. Otherwise,
write Xn = IIYnll-1Yn and x = IIYII- 1y. Then Xn x weakly and Ilxnll-+
= Ilxll = 1. Also if Ilxn - xii -+
0, it follows that llYn - yll O. Hence -+
we need only to show that Ilx n- -+
xii O. Assume the contrary. Then by
local uniform convexity, there is a subsequence (Xnk)k~l such that
II ~ (Xnk + x)11 < DC < 1 (k = 1,2,3, ... ) . (1)

By (14.13), there is an I EE* such that IIIII = 1 and I (x) = 1. For this I,
(1) implies that
so Xn does not converge weakly to x.]

(b) [RADON-RIESZ theorem]. Let p be a real number such that
1 < P < 00 and let (X,d, /1) be a measure space. Write ~p for ~p(X,d,/1).
Let (In) be a sequence of functions in ~p and I a function in ~p such that
-+ -+ -+
In I weakly and Il/nllp 11/11p· Prove that Il/n - Illp o. [Use CLARKSON'S
inequalities (15.5) and (15.8) to show that ~p is locally uniformly convex.
Then apply part (a)1.]
1 This short proof of the RADON-RIESZ theorem, and part (a), were kindly sug-
gested to us by Professor IRVING GLlCKSBERG.
(15.18) Exercise. Define I and In' n = 1,2, ... , on [0,2nJ by the

rules: I(x) = 1, In (x) = 1 + sin (nx). Notice that I, InE~I([O, 2nJ,~, A)
for all n EN. Prove that:
(a) In --* I weakly in ~I [use (16.35)J;
(b) IIIIII = Il/nlll = 2n for all n EN;
(c) In --f-'>- I in measure;
(d) Il/n - IIII --f-'>- 0;
(e) In--f-'>-/a.e.;
(f) ~I ([0, 2nJ, ~, A) is not locally uniformly convex.
§ 16. Abstract Hilbert spaces

(16.1) Inner product spaces. Recall (13.16) that an inner product
space is a linear space Hover K together with a mapping (x, y) --* <x, y)
of H x H into K such that:
<x + y, z) = <x, z) + <y, z);
<ccx,y) = cc<x,y);
<y, x) = <x,y);
<x,x»O if x=!=O;
[for all x, y, z EHand cc EKJ. The above relations imply trivially that:
<O,y) = <y, 0) = 0;
<x, ccy) = oc<x,y);
<x,y + z) = <x,y) + <x, z);
for all x, y, z EHand cc EK.
An inner product space over R is defined similarly; in this case we
have <y, x) = <x, y). For many reasons complex inner product spaces
are more useful in analysis than are real inner product spaces.
(16.2) Theorem [Inequality of CAUCHY-BuNYAKOVSKii-SCHWARZ].
For x, y EH we have
(i) l<x,y)12;;:; <x, x) <y,y).
Equality obtains in (i) il and only il x and yare linearly dependent.
Proof. If y = 0, equality obtains and we have Ox = ly. Thus suppose
°
that y =!= and let y E K. Then we have
0;;:; <x - yy, x - yy) = <x, x) - y<y, x) - y(x,y) + IYl2 <y,y). (1)
Setting y = «x,y,yy» in (1), we obtain
°s: <x x) _
-,
(x, y) (y, x) _ (X:'Y) (x, y)
(y,y) (y,y)
+ I(x, y)I' (y, y)
(y, y)2
_ <x,x) -
-
I(x,y~
(y,y) .
§ 16. Abstract Hilbert spaces 235
It is clear that strict inequality holds throughout this computation

unless x - yy = O.
If IXX = py, where IIXI + IPI =l= 0, it is easy to see that the two sides
of (i) are equal. 0
1
(16.3) Theorem. Let IIxil =
(x, X)2. With this definition 01 norm,
an inner product space H is a normed linear space.
Proof. All of the verifications that I I is a norm on H are trivial ex-
cept for the triangle inequality. Evidently
IIx + Yll2 = (x + y, x + y) = (x, x) + (x,y) + (y, x) + (y,y)
= I xII 2+ 2 Re(x,y) + IIyll2.
Applying (16.2), we have
2 Re(x,y) ~ 21(x,y)1 ~ 211xll 'IIYII .
Therefore IIx + Yll2 ~ (IIxil + IIYII)2, and so IIx + yll ~ IIxil + IIYII.
0
(16.4) Exercise. Find a necessary and sufficient condition that
IIx + yll = IIxll + IIYII·
(16.5) Exercise: The polar identity. Prove that if H is a complex
inner product space, then
4(x,y)= Ilx+yll2- IIx-yll2+illx+iyll2-ilix-iyll2
for all x, y EH.
(16.6) Exercise. Let E be a complex normed linear space. Prove that
there exists an inner product on E that induces the given norm as in
(16.3) if and only if the given norm satisfies the parallelogram law
(i) IIx + Yll2 + IIx - Yll2 = 2(IIxll 2+ IIYll2) for all x, y EE .
[If (i) holds, define (x, y) by Re (x, y) = ~ (lix + Yll2 - I xII 2- 1IY1I2)
and 1m (x,y) = ~ (lix + iYll2 - IIxll 2- IIYII2). Then (16.1) is easy to
check. The reader should draw a diagram to show that (i) is really a
well-known elementary fact about sides and diagonals of a parallelo-
gram. This exercise shows that inner product spaces are the normed linear
spaces in which all two-dimensional subspaces "look like" Euclidean
spaces.]
(16.7) Definition. An inner product space which, with the norm
defined in (16.3), is a Banach space is called a Hilbert space. An in-
complete inner product space is sometimes called a pre-Hilbert space.
(16.8) Examples. (a) If (X, d, ft) is any measure space, then
~2(X, d, ft) is a Hilbert space [(13.11) and (13.15)].
(b) The space <roo(R) with
00
(I, g) = J 1ft) g(t) dt

-00
is an incomplete inner product space. To see this, approximate ';[0,1)

by the sequence (gn)~l of continuous functions defined as follows:
gn (x) = 0 for - 00 <x ~ - n-1 and for 1 + n1 ~ x < 00; gn (x) = 1
for 0 ~ x ~ 1; and gn is linear on [- ~ , 0] and [1, 1 + ~]. The
sequence (gn) converges to ';[0,1) in ~2' and so is a Cauchy sequence.
However, it is clear that (gn) converges to no function in <roo(R).
(c) The sequence space 12(N), consisting of all complex sequences
00
x = (xn) such that 1: Ixn 2 <l 00, is a Hilbert space in which «(xn), (Yn)
n~l
00
= 1: xnYn' [As we have observed before (13.13), l2(N) is actually the

n~l
space ~2 (X, d, fl) in which X is the set N, d is all subsets of X, and fl

is counting measure. ]
(d) The space consisting of all sequences x = (xn ) such that Xn is
ultimately zero is an incomplete inner product space [the inner prod-
uct is that induced by 12J. For example, the sequence (x(m»):~l in
which x(m) = (1, ~, ... , ~ ,0,0, ... ) converges in l2' but its limit has
no zero terms. The reader should note the analogy between this space
and <roo (R); indeed if the positive integers N are given the discrete
topology, then the space being considered is <roo(N).
We now take up a notion that lies ready to hand in any inner prod-
uct space, and which we will use later to classify all I-Iilbert spaces.
(16.9) Definition. Let H be an inner product space. If (x,y) = 0
for elements x and Y of H, then x and yare said to be orthogonal to each
other, and we write x ..l y. If E is a subset of H such that (x, y) = 0
for all x, y E E such that x =1= y, then E is said to be orthogonal. If in
addition Ilxll = 1 for all x EE, then E is called orthonormal. If E and F
are subsets of H such that x ..l y for all x EE and y EF, then we say
that E and F are orthogonal, and we write E ..l F.
The sets 0 and {x} are orthogonal for all x EH. The vector 0 is
orthogonal to every vector in H.
(16.10) Theorem. If {Z1' ... , zn} is an orthogonal set in H, then
IIz1 + ... + znl1 2 = IIz1112 + ... + I znl1 2 .
Proof. We have
n
= 1: IIZjI12. 0
i~l
(16.11) Definition. Let E be a nonned linear space and let (Xn ):'=1
00
be a sequence of elements of E. We say that the series E Xn converges,

n=!
and we write 1: Xn
n=!
= x, if there exists an xEE such that lim
P-+oo
[[x - !; Xn[[
n=!
=0.
(16.12) Theorem. Let H be a Hilbert space and let {Zn}:'=l be an
Zn -1 Zm tor n =1= m. E Zn converges it and
00
orthogonal set in H, i.e., Then

n=!
00 00 00
only it E I znl1 2< 00. It E Zn = z, then IIzI12 = E I znl1 2.

n=! n=! n=!
Proof. For n > m, it is plain that
[\k#1 Zk - il Zk\\2= Ilk=EI Zk I1 2= k=Elllzkl12 = il11zkl12 - k~ Il zkl1 2.

Thus (~! Zk) ~1 is a Cauchy sequence in H if and only if (ilIIZkI12) ~=!
is a Cauchy sequence in R. This proves our first assertion.
00 n
Now suppose that I: Zk = z. Writing sn = I: Zk' we have liz - snll -+ 0
k=! k=1
n
and IIsnl12 = I: Il zkl1 2. Thus (11zll + Ilsnll):'=1 is a bounded sequence of
k=1
numbers. Also we have Illzll - Ilsnlll £ liz - snll-+ 0, and so
IIIZI12 - k~ I Zkl1 21= IIIzI12 - I snl1 21
= (11zll + Ilsnll) Illzll - Ilsnlll-+ 0,
i. e.,
n
lim I: I zkl1 2= Ilz112. 0
11-+00
k=1
(16.13) Theorem. Let H be an inner product space and let E be an
orthogonal subset ot H not containing o. Then E is linearly independent.
Proof. Suppose that {Xl> ... , xn} C E, that 1X1 , ••• , IXn are scalars,
n
and that I: IXkXk = o. Then for each j E{I, ... , n} we have 0 = (0, Xi)
k=1
= <il IXkXk, Xi) = t!lXk(Xk , Xj) = IXj Ilxj l1 2. But Xj =1= 0, and so IIxjl12 =1= O.
Therefore alllX/s are zero. 0
(16.14) Definition. Let E be an arbitrary orthogonal set in an inner
product space H. For X EHand Z EE, we define the Fourier coefficient
ot X with respect to Z to be the number (x, z).
In the real Hilbert space R3 [3-dimensional Euclidean space], the

vectors (1,0,0), (0, 1,0), and (0,0, 1) form an orthonormal set, and the
Fourier coefficient of a vector x with respect to each of these unit vectors
is simply the length of the perpendicular projection of x in the direction
determined by that unit vector.
The following example motivates the use of the term "Fourier
coefficient" in (16.14).
(16.15) Example. Consider the space ~2([-:n,:n],~,-!- A). In
this space, the inner product is n
,.
(I, g) = 2~ JI(t) g(t) dt
-,.
for I, g E~2' For nEZ, the functions Xn: t -+ exp (int) are defined and
continuous on R, and so are in ~2([-:n, :n]). We will show that these
functions form an orthonormal set. For each nEZ, we have IIXnl1 2
,. ,.
= 21n JIXn(t)12 dt n J1 dt If m =t= n in Z, we have (Xm' Xn)
= 21 = 1.
,.
-n
,. -n
,.
= 2~ JXm (t) Xn (t) dt 2~ Jexp (i (m - n) t) dt = 2~ Jcos«m- n) t) dt
=
-,. -,. -n
n
+ J sin«m - n)t) dt O. [Use elementary calculus to evaluate
2in =
-n
these integrals.] Hence {Xn}:=-oo is an orthonormal set. For IE~l([ -:n,:n]),
the classical nth Fourier coefficient of 1is the number
JI(t) exp(-int) dt
n
I(n) = 2~ = (I, Xn)·

-n
For an inner product space H and an orthonormal set {zv ... , zn} C H,
one often wants to know how closely an element xEH can be approximat-
ed in the metric of H by a linear combination of the z,:s. This question
is completely answered by the following theorem.
(16.16) Theorem. Let H be an inner product space, let {zv . .. , zn}
be an orthonormal set in H, and let x be any element 01 H. Then the lunction 1
defined on Kn by
(i) 1«(Xv .•. , (Xn) = Ilx - g(X"z,,11

attains an absolute minimum value at one and only one point 01 Kn, viz.,
(X" = (x, z,,) (k = I, ... , n). Furthermore, the inequality
n
(ii) }; I(x, Z,,)12 ~ IIxl1 2
"=1
holds.
Proof. We have
.. \\2 .. ..
\\ x - k-?: (Xk Z,. = I xl1 2 - k~ (Xk(X, Zk) - k~ (Xk(Z,., x) + k~.. l(Xkl 2
n n
= IIxl1 2 + I; [l<x, Z,.)12 - OC,. (x, Z,.) - (Xk (x, Z,.) + l(XkI 2] - I; I(x, Z,.)12
k=l
. . k=l
= IIxl1 2 + I; I(x, Zk) - (Xk1 2 - I; I(x, Zk)12.

k=l k=l
Hence t ((Xv' .. , (Xn) is a minimum if and only if (Xk = (x, Zk)' k = 1, ... , n;
and in this case we see that
n
o ~ IIxl1 2 - I; I(x, Zk)l2. 0
k=l
(16.17) Theorem [BESSEL'S inequality]. Let E be a nonvoid ortho-
normal set in an inner product space H, and let x EH. Then we have
(i) I; I(x, z)12 ~ Ilx11 2 ,
zEE
and so {z EE : (x, z) =l= O} is countable.
Proof. The inequality (16.16.ii) implies that for each nonvoid finite
set FeE we have
I; I(x, z)12 ~ IIxl1 2 •
zEF
Therefore
I; I(x, z)12 = sup{ I; I(x, z)12: F =l= 0, F finite, FeE} ~ Ilx11 2 • 0
'EE zEF
(16.18) Theorem. Let {Zk}k'=l be an orthonormal set in a Hilbert space H.
00
For every x EH, the vector Y = I; (x, Z,,)Zk exists in H and x - Y is orthog-
k=l
onal to every z,..
Proof. The existence ofy follows from (16.12) and BESSEL'S inequality.
Let mEN. We must show that (x - y, zm) = O. For each n EN let
n
Yn = I; (x, Zk) Zk' Then for all n ~ m, we have
k=l
I(x - y, zm)1 ~ I(x - Yn' zm)1 + I(Yn - Y, zm)1
~ \(x, zm) - k~ (x, Z,.) (Z,., Zm)\ + llYn - Yllllzmll

= 0 + llYn - yll .
We have used the fact that {Zk} is orthonormal. Since llYn - YII-+ 0,
it follows that (x - Y, zm) = O. 0
We shall now investigate the problem of writing an arbitrary element
of an inner product space as a limit of linear combinations of elements of
an orthonormal set. We first make a definition.
(16.19) Definition. An orthonormal set E in an inner product space H

is said to be complete if the only vector orthogonal to all elements of E is O.
(16.20) Theorem. Every inner product space that is not {O} contains a
complete orthonormal set. Intact, every orthonormal subset of H is contained
in a complete orthonormal set.
Proof. We use TUKEY's lemma (3.8). Let A be any nonvoid ortho-
normal subset of H; for example, A = {llxll-1x} where x =1= 0, x EH.
Next let ff = {B: Be H, A U B is an orthonormal set}. To test ortho-
normality one tests only two vectors at a time, and so it is clear that ff
is of finite character. Also A E ff, so that ff is nonvoid. By TUKEY's
lemma, ff has a maximal member E. It is obvious that E::J A. We
assert that E is complete. Assume that Y =1= 0 and that y.l E. Set
z = 1IY11-1 y. Then E U {z} E ff and E S; E U {z}. This contradicts the
maximality of E. Thus Y ..1 E implies Y = O. 0
The above proof is not constructive; it gives no clue as to how to
construct a complete orthonormal set in any given inner product space.
There are methods of actually constructing complete orthonormal sets
provided that they are not too large. We now take up this construction.
(16.21) Lemma. Let 5 be a dense subset of an inner product space H.
It x ..1 5, then x = O.
Proof. Choose a sequence (Yn) in 5 such that Ilx - Ynll ~ O. Then
IIxl1 2 = <x, x) - <x,Yn) = <x, x - Yn);;;;; Ilxll . Ilx - Ynl! ~ O.
Therefore Ilxll = o. 0
(16.22) The Gram-Schmidt orthonormalization process. Let H be an
inner product space and let {Y1' ... , Yn' ... } be a finite or count ably
infinite linearly independent subset of H. Let Z1 = Y1 and set
U1 = Il z111- 1Z1 ,
and
Z2 = Y2 - <Y2' u1) u1 .
The vector Z2 is not zero, for Y2 is not a multiple of Y1' Define
U2 = Il z211- 1z2'
We have
<u2' u 1) = Ilz211-1<z2' u 1) = O.
Thus the set {u 1 ' u 2} is orthonormal and it spans the same 2-dimen-
sional subspace as {Y1' Y2}'
We will define inductively an orthonormal set {Ul>"" Un' ... }
such that for each positive integer k, the set {u 1 ' ••• , Uk} spans the same
subspace as {Y1' ... , Yk}' Thus suppose that {Ul> ... , un} has been con-
structed and that span {U1' . . . , Uk} = span {Yl> ... , Yk} for k = 1, ... , n.
If {Yl> Y2' ... , Yn} is aU of the y's, stop the construction. If Yn+1 exists,
let
.
zn+1 = Yn+1 - E (Yn+1' UII) ull .
11=1
Now zn+1 is not the zero vector, because Yn+1 ~span{Yl'" .,Yn}
= span{ul , ... , un}. Next define
Un+l = IIZn+lll-l zn+l'

For 1 ~j ~ n, we have
1
(un+1' u;) = II Zn+1II- (Yn+1 - 1;1 (Yn+1' UII) ulI' U;)
= IIZ"+111- 1 (Yn+1' u;) - (Yn+1' u;) • (u;, u;») = 0 .
Thus the set {Ul> ... , un+1} is orthonormal; it remains to show that
it spans the same subspace as {Yl> ... , Yn+1}' It is obvious from the def-
initions of zn+1 and Un +1 thatY"+1is a linear combination oful , ... , un+!'
Therefore
{Yn+!} U (span{Yl' ... , Yn}) C span{ul> ... , Un+l} •
and so
span{yl> ... , Yn+!} C span{ul> ... , un+l} .
Similarly un+1 is a linear combination of the vectors Yn+1 and Ul> ...• Un'
By the inductive hypothesis, it follows that un+! Espan{Yl> ... , Yn+l},
and therefore span{ul>"" Un+l} C span{Yl>" "Yn+l}' Thus these two
subspaces are the same. [Alternatively, we could have reversed the
first inclusion by a dimensionality argument.]
The Gram-Schmidt process yields an orthonormal set which is es-
sentially unique. More precisely, if {Ul> ... , un} is to span the same
subspace as {Yl> ...• y,,} for each n, then we must have u l = II~II Yl>
where Irl = 1. Hence the choice of u l is unique up to a multiplicative
constant of absolute value 1. Having defined {Ul> ...• un}. we must take
un+l Espan {Ul' ... , un.Yn+1}· Thus for some complex numbers Pl' ... , Pn'
ex, we have
The number ex cannot be zero since, if it were, then Un +1 Espan{Ul>'" ,un}

= span{Yl"" ,Yn},fromwhichit would follow thatYn+lEspan{u l , ..• ,Un+l}
= span{Yl> ... , Yn}. This would contradict the linear independence of
the Yll's. Hence

and so also
and
lJ j = - (Yn+l' Uj) (1;;;' j ;;;, n) .
Thus we have
just as we defined zn+l' The number (X is now determined from this last
equality by taking norms and noting that IIUn+ll1 = 1. Clearly (x, and
hence U n + v is unique up to a multiplicative factor of absolute value 1.
(16.23) Theorem. Let H be an inner product space, not {O}, that con-
tains a countable dense subset D. Then H contains a countable complete
orthonormal set that is obtained from D by the Gram-Schmidt process.
Proof. Suppose, as we may, that 0 ~ D, and enumerate D as (Xn):'l'
Define Yl = x n , where n 1 = 1. Suppose that Yl = x n " ..• , y,. = xnkhave
been defined and are linearly independent, and that n 1 < n 2 < ... < n,..
If there is no j > n,. such that {yv ... , Y,., Xj} is linearly independent,
stop the process. Otherwise let n"+l be the smallest such j and define
Y"+l = xnk+1' We have thus defined a finite or countably infinite linearly
independent set {Yl' Y2' ... } C D. Let 5 be the smallest linear subspace
of H containing {Yv Y2' ... }. It is clear that DeS, since if Xj ED, then
Xj is a linear combination of Yl' ... , y", where k is chosen so that
n" ;;;, j < n1<+ 1 [or else y" is the last Y selected and n,. ;;;, j]. Hence 5 is dense
in H. Let {u v us' ... } be the orthonormal set obtained from the set
{Yl' Y2' ... } by the Gram-Schmidt process. We will show that {uv us' ...}
is complete. Suppose that x E H and (x, un) = 0 for all n. Then
(x, g /XnUn) = 0 for all finite linear combinations of the un's, and so
(x, y) = 0 for all yES. It follows from (16.21) that x = o. 0
(16.24) Corollary. Let n EN. T.hen an inner product space H is in-
distinguishable [as an inner product space] from Kn if and only if the
algebraic dimension of His n. [In Kn we define «(Zl' ... , zn), (wv ... , wn)
n
= 1: ZjWd
i=1
Proof. It is clear that if H has dimension n, then the process of
choosing the y's in the previous proof stops with {Yv ... , Yn}. Thus we
obtain a complete orthonormal set {ull •• '1 un} C H. The mapping of
H onto Kn given by
n
1: (Xi Ui -+ ((Xl' • • ., (Xn)
i=1
preserves all inner product space structure, i. e., it is one-to-one, onto,

linear, and preserves inner products [hence norms as well]. The converse
is trivial. 0
(16.25) Example. By way of illustration, we work out a certain classi-
cal orthonormal set. For each integer n ~ 0, define In on R by
In(x) = xn exp [_ ~2] .

Since £
k~1
k2(:2) < 00 for all n ~ 0, each In is evidently in the Hilbert
exp
space S!.2(R, Jt)., A). Since each polynomial has only finitely many roots,
the set {In}:~o is linearly independent over K. For each integer n ~ 0,
define
where the superscript (n) denotes the nth derivative of the function
x -+ exp [- x 2 J. The functions (Hn):~o are clearly all polynomials. They
are called the Hermite polynomials. The first three Hermite pol ynomials are
Ho(x) = 1,
HI (x) = 2x,
H 2 (x)=4x 2 -2.
One can go on computing them as long as patience will permit. Next let
these functions are called the Hermite lunctions. They are all in
S!.2(R, Jt)., A), and, as we will now show, they are an orthogonal set.
First we have
cp~ (x) = (_ l)n {(x 2 + 1) exp [ ~2] exp(n) [- X2]
+ 2xexp[ ~2] exp(n+l) [_ X2] + exp[ ~2] exp(n+2) [-X 2]}. (1)
Using LEIBNIZ'S rule for finding the derivatives of a product, we have

exp(n+2) [_X2] = {-2xexp[-x2J}<n+l)
1;1 (n +k 1) (_ 2x) (/') exp(n+l-k) [_ X2]

k~O
= (-2x) exp(n+l) [_X2] + (n + 1) (-2) exp(n) [_X2] .
16*
Substituting this expression in (1), we obtain
rp:: (x) = + 1) exp(n) [- X2] + 2 x exp(n+1) [- X2]

(-l)n exp [ ~2] {(X2
+ (-2x exp(n+1) [_X2] - 2(n + 1) exp(n) [_X2J)}
= (-1)n exp [~2] exp(n) [_ X2] (x2- 2n - 1) = (X2- 2n -1)rpn (x) .
Thus every rpn satisfies the differential equation
rp::(x) = (x 2 - 2n -1)rpn(x).
Hence for every pair of nonnegative integers m and n, we have
rp:: rpm- rpn rp::' = (x2-2n-1) rpnrpm- (x 2-2m-l)rpnrpm=2(m-n) rpnrpm·
If m =1= n, we have
00 00
-00 -00
[This computation requires LEBESGUE'S theorem on dominated conver-

gence.] Thus the set {rpn}:~o is orthogonal.
00
To normalize therpn's, we now compute J rp~ dx. We begin by establish-

-00
ing the equality

H~ = 2nHn_1 (n = 1,2,3, ... ) . (2)
We have
H~(x) = (-1)n{2x exp[x2] exp(n) [_X2] + exp[x2 ] exp(n+1) [_X2J}.
Computing as before, we find that
exp(n+1) [_X2] = -2x exp(n) [_X2] - 2n exp(n-l) [_X2] ,
and therefore
H~(x) = (-1)n {2x exp [X2] exp(n) [_X2]
+ exp[x2] (-2x exp(n) [_X2] - 2n exp(n-l) [_X2J)}
= (_1)n-l 2n exp [X2] exp(n-l) [_X2] = 2nHn_1 •
This establishes (2). To evaluate our integrals, we first observe that

00 00 1
J rp~(x) dx = J exp [_X2] dx = 1(;2,
-00 -00
as is well known. [See (21.60) infra.J Next, we have

00 00
J cp~(x) dx = J exp [_X2] H;(x) dx

-00 -00
00
= J exp [_X2] Hn(x) exp [X2] (-1)n exp(n) [_X2] dx

-00
00
-00
= lim {(-1)nHn(X)
A~oo
exp(n-1)[-x2]1~.A
+ (_1)n-1 /H~(x) exp(n-1) [-x2J dX}
-A
00
-00
00
= 2n
-00
J CP;-l (x) dx .
This establishes the recursive formula
00 00
J cP; (x) dx = 2n J CP;-1 (x) dx,

-00 -00
and it follows that

1
J cp;(x) dx =
00
n"22nn!
-00
for n = 0, 1,2, . . .. Hence the functions {1J!n}:=o given by 1J!n(x)

1 1
= (n+ 2n n!t"2 CPn{x) = {n+ 2 n n!t"2 (-I)n exp [~2] exp(n) [_X2] form
an orthonormal set. The functions {1J!n}:=o are obtained from
{xn exp [- ~2 ]}::o
by the Gram-Schmidt process. This follows readily
from the fact that Hn has degree n for all n, and from the essential unique-
ness pointed out in (16.22).
(16.26) Theorem. Let H be a Hilbert space. The following properties
01 an orthonormal subset E at H are equivalent.
(i) The set E is complete.
(ii) For each x EH, we have x = I.; <x, z)z [Fourier seriesJ.I
zEE
(iii) For all x EH, we have II xII 2 = I.; I<x, z)12 [PARSEVAL'S identity].
zEE
1 The equality (ii) means that the right side has only a countable number of
nonzero terms and that for every enumeration of these terms the resulting series
converges to x as in (16.11). The equalities (iii) and (iv) have analogous meanings.
(iv) For all x,y EH, we have (x,y) = I; (x, z) (y, z) [PARSEVAL'S
lEE
identity].
(v) The smallest subspace 01 H containing E is dense in H.
Proof. Suppose that (i) holds and let x EH. According to BESSEL'S
inequality (16.17), there are only countably many z EE such that
(x, z) =l= 0; enumerate these as (z..). By (16.18), the vector y = I; (x,z.. )z..
= I; (x, z)z exists and x - y is orthogonal to E. Since E is complete,
.
JEE
it follows that x - y = 0; and so (ii) holds.
To show that (ii) implies (iv), let x, y EH be given and let (z,,) be
an enumeration of all z EE such that (x, z) =l= 0 or (y, z) =l= o. Let
.
x.. = I; (x, z,,) Z"
"=1
and
.
y .. = I; (y, z,,) z".
"=1
Then we have
/(X,y) -,,~ (x, z,,) (y, Z,,)! = I(x,y) - (x.. ,y.. )1
~ I(x,y) - (x",y)1 + I(x.. ,y) - (x.. ,y.. )1

~ Ilx - x.. II . Ilyll + II x.. II . Ily.. - yll -40
because Ilx - x.. II -40, Ily.. - yll --+ 0, and II x.. II --+ Ilxll. Thus (x, y)
00
= I; (x, z,,) (y, z,,) = I; (x, z) (y, z); hence (iv) is established.
"=1 JEE
It is obvious that (iv) implies (iii). if (iii) holds and (x, z) = 0 for all
z EE, then it is plain that Ilxll = o. Thus (iii) implies (i). This completes
the proof that (i), (ii), (iii), and (iv) are pairwise equivalent.
Plainly (ii) implies (v). Finally we show that (v) implies (i). Suppose
that x EH and (x, z) = 0 for all z EE. Then it is clear that (x, y) = 0
for ally in the linear span ofE. It follows from (v) and (16.21) that X= o. 0
(16.27) Theorem. Any two complete orthonormal sets in a Hilbert
space H have the same cardinal number.
Proof. Ignoring a trivial case, we suppose that H =l= {O}. Let A and B
be any two complete orthonormal sets in H. If A is finite, it follows from
(16.26.ii) and (16.13) that A is a Hamel basis for Hover K. Since B
is linearly independent, (3.26) shows that B is contained in a Hamel basis
C, so that 11 ~ {j, and {j = .1 by (4.58). Thus B is also finite, and is
also a Hamel basis. Another reference to (4.58) shows that 11 = .1.
The case in which A and B are infinite remains to be treated. For
each a E A, let Ba = {b EB: (a, b) =l= O}. Then Ba is countable for all
a EA. For any bE B we have 1 = Ilblls = 1: I(b, a)12 (16.26.iii), and so

aEA
there is some a EA such that (a, b) =1= 0, i.e., bE Ba. Thus B = U Ba.
aEA
It follows that B ~ itoA' = A'. Interchanging the r6les of A and B
in this argument, we see also that A' ~ B. It follows from the SchrOder-
Bernstein theorem that A' = B. 0
(16.28) Definition. Let H be a Hilbert space. If H =1= {O}, we define
the orthogonal dimension of H to be the [unique!] cardinal number of
any complete orthonormal set in H. If H = {O}, we say that H has orthog-
onal dimension zero.
(16.29) Theorem. Let H be a nonzero Hilbert space. Then there
exists a set D and a linear transformation T from H onto l2 (D) that preserves
inner products [hence norms as well]. Also 15 is the orthogonal dimension of H.
Proof. Let D be an arbitrary complete orthonormal set in H; then
!J is the orthogonal dimension of H. For x EH, let T (x) be that function
on D such that
[T(x)](z) = (x, z)
for all z ED. Then Tmaps H into l2(D), for 1: I(x, Z)llI < 00 by BESSEL'S
rED
inequality. Also, for x, y EH, we have
[T(x + y)](z) + y, z) = (x, z) + (y, z) = [T(x)](z) + [T(y)](z)
= (x
for all zED; that is, T (x + y) = T (x) + T (y). Similarly T (exx) = ex T (x)
for all ex EK. Thus T is linear. Using PARSEVAL'S identity (16.26.iv),
we have
(T(x), T(y) = 1: [T(x)(z)] [T(y)(z)] = 1: (x, z) (y, z) = (x,y) ,
rED lED
and so T preserves inner products. Finally, we show that T is onto
12 (D). If / E12 (D), then E 1/(Z)j2< 00, and it follows from (16.12) that
rED
x= E f (z) z is in H. Let (Z,.) be an enumeration of all zED such that
:ED
fez) =1= 0 or (x, z) =1= o. For a fixed p and any m ~ p, we have
I(x, zp) -/(zp)1 = I(X, zp) - n~ f(z,.) . (Z,., Zp)1
~ Ilx - g f(z,.) Z,.II· Ilzpll-+ 0 as m -+ 00.
Therefore fez) = (x, z) = T(x)(z) for all zED; hence f = T(x). 0
(16.30) Remark. It follows from (16.29) that every .s!2(X,.s;I, p)
is completely determined qua Hilbert space by a single cardinal number,
and in fact is indistinguishable from a certain l2(D). No characterization
of this kind for .s!p(X,.s;I, p), P =1= 2, is known to the authors. We also
see that there exists just one [complex] Hilbert space for each cardinal
number, i. e., a Hilbert space is completely determined by its orthogonal

dimension.
(16.31) Theorem. Let H be a Hilbert space and let I be a bounded
linear lunctional on H. Then there exists a unique y EH such that
I(x) = (x, y)
lor every x EH. Moreover
11I11 = IIYII .
Proof. Since H may be identified with an l2 (D), the theorem follows
from (15.12). 0
(16.32) Theorem. The lunctions X.. : t ~ exp(int) (n EZ) lorm a
complete orthonormal set in ~2 ([-n, n],.AA, 2~ A).
Proof. We have already shown (16.15) that {x..}..ez is an orthonormal
set in ~2' We now prove that it is complete. Let T = {z EK: Izl = I}
= {exp(it) : -n ~ t ~ n}. For every integer n, the function
exp (it) ~ exp (int)

is continuous on T. Let <I' denote the set of all functions of the form
exp (it) ~
"
}; IX" exp (ikt)
k=-"
where IX" EK. These functions are called, for an obvious reason, trig-
onometric polynomials. We have proved in (7.35.b) that <I' is uniformly
dense in <reT). It is plain that this implies that any function IE<r([-n, n])
such that I(-n) = I(n) can be uniformly approximated [arbitrarily
closely] by functions in <I' [regarded as functions of t E [- n, n]]. Now
let rp E~2([-n, n]) and let e be a positive real number. Use (13.21)
to select a function gE<r([-n,n]) such that Ilrp-gI12~;' Next,
by changing the values of g on an interval [n - ~, n] for an appropriate
~ > 0 [if necessaryJ. it is easy to find a function I E<r ([- n, n]) such that
1(- n) = I (n) and Ilg - 1112 < ;. Finally, choose a function P E<I'
such that III - PII" < ; . Then
Ilrp - PI12 ~ Ilrp - gl12 + Ilg - 1112 + III - PI12
I
< ; e+~ }t-PI'dl.)'

,; (2~ 1(;)'01)' +;
1
= e.
Thus ~ is dense in ~2([-~' ~]). By (16.26.v), {X"},,EZ is complete. 0

(16.33) Definition. For IE~I([-~,~],~,A), let
n
I(n) = 2~ f I(t) exp(-int) dt (n EZ).

-n
The number I(n) is called the nth Fourier coelficient 011. The function I
on Z is called the Fourier translorm 011.
(16.34) Theorem. Let I E~l([-~' ~]). II the identity I = 0 holds, then
1= 0 in ~l'
Proof. Let ~ be the set of all trigonometric polynomials on T, defined
as in the proof of (16.32). If 1= 0, then it is obvious that
n
J I(t) pet) dt = 0
-n
for allp E~. Let x be a fixed number in ]-~, ~]. It is easy to construct a
nondecreasing sequence (g,,)::"=l of nonnegative continuous functions
such that:
g,,(-~) = g,,(~) = 0;
lim g,,(t) = ~l-n,xdt) for all t E [-~,~] .

fl-+OO
By (7.35.b), there exists for each n EN a polynomial p" E~ such that

1
Ilg" - All ... < n· Then we have
I(t) ~l-n,%[(t) = lim I(t) p,,(t)
fl-+OO
for all t E[-~,~] ,
and plainly
LEBESGUE'S dominated convergence theorem (12.30) shows that

% n n
J I(t) dt = -nJ l(tHl-n,x[(t) dt =
-n
lim
~OO_n
J I(t) P,,(t) dt = O.
From (12.54) we infer that 1=0 a.e. on [-~, ~]. 0
(16.35) Riemann-Lebesgue Lemma. Let I E ~l ([-~, ~]). Then
lim I(n) = o.
1"1-+00
Proof. Let 8>0 be given. Use (13.21) to choose gE(t([-~,~])
such that 11/- gill < ; . Clearly g E~2([-~' ~]). By BESSEL'S inequality
00
we have E Ig(n)12 ~ IIgll~. Thus there exists a positive integer p

11=-00
such that Ig(n)1 < ~ whenever Inl ;s p. Hence Inl ;s p implies

I/{n) I ~ II (n) - g (n) I + Ig (n)l
n
< 2~ f (t(x) - g(x) exp(-inx) dx +;

-n
,.
~ 21n f If (x) - g(x)l dx +;
-n
e e
<2+2=e. D
(16.36) Remark. There is an analogous theory of Fourier transforms

for the space ~1 (R,~, ),) = ~1 (R). In this case, one defines
I(y) = (2n)-+ f f(x) exp(-iyx) dx

R
for each y ER. Once again the equality 1=0 implies f = 0 a.e. Also,
lim I (y) = O. These transforms are important in several parts of analysis.
Iyl-+oo
Using them, it is possible to prove that the Hermite functions are a
complete orthonormal set in ~2 (R); for a short proof, see (21.64.b)
infra.
(16.37) Theorem [PARSEVAL'S identity]. For f E~2([-n, n]), the
equality
(i) 2~ f " If(t)1 dt ;: £

2
"=-00
I/(n)12
-"
holds.
Proof. This is an immediate consequence of (16.26), (16.32), and the
definition (16.33) of I(n). D
(16.38) Remark. There is a generalization of (16.37) for ~p([ - n, n]),
1 < p < 2. In this case, the inequality
I ~
C~~II(n)IP)" < (;.jlf(~l'dt)'

holds, unless f(x) = cx exp(imx) for some cx EK and m EZ. This is a
nontrivial fact, and its proof is fairly sophisticated. See for example
EDWIN HEWITT and 1. 1. HIRSCHMAN JR., Amer. J. Math. 76, 839-852
(1954).
(16.39) Riesz-Fischer theorem. Let (cxn)::'=_oo be in l2(Z), i.e.,
00
:E Icxn I2 <oo. Then there is a function fE~2([-n,n]) such that

n=-oo
I (n) = CX n , tor all n EZ.
Proof. This is a trivial consequence of the completeness of ~2 (13.11).

Let
"
I,,(t) = }; ocjexp(ijt) (k = 1,2, ... ).
;=-"
-"-I 1
Then the equality III" - I,II~ = }; locjl2 + }; locjl2 shows that
;=-1 ;=k+1
(I,,) is a Cauchy sequence. Let I be the limit in ~2 of (I,,). We clearly
have/(n) = lim /,,(n) = <Ln. 0
BOO
(16.40) Theorem. Let H be a Hilbert space and let ::B (H) denote the
space 01 all bounded linear operators [translormations] Irom H into H.
Then ::B (H) is a Banach algebra with unit, where we take composition
lor multiplication. Moreover, lor each T E::B (H) there exists a unique
T* E<:13 (H) such that
(T(x),y) = (x, T*(y)
lor all x, y E H. Also T** = T and II T* II = II Til. The operator T* is
called the adjoint of T.
Proof. We have proved in (14.4) that ::B (H) is a Banach space. For
Tv T2 E <:13 (H) and x E H, we have
II Tl T2(X) II ~ IITl ll'IIT2 (x)11 ~ I Tlil '11T2 11 ·llxll·
Therefore Tl T 2 E ::B (H) and
II Tl T211 ~ IITl ll'IIT2 11 .
It is easy to see that <:13 (H) is an algebra over K. Thus <:13 (H) is a Banach
algebra.
Let T E::B(H). For eachy E H, define lyon H by
Iy(x) = (T(x), y) .
Then we have
I/y(x) I = I(T(x),y)1 ~ II Til '1Ixll'llyll
for all x E H. Obviously Iy is linear. Thus Iy is a bounded linear functional
on Hand Il/yll ~ II Til . IIYII. It follows from (16.31) that there exists a
unique element T*(y) EH such that (T(x),y) = Iy(x) = (x, T*(y)
for all x E Hand
II T* (y) II = Il/yll ~ II Til' IIYII . (1)
This defines T* on H. A simple computation shows that T* is linear,
and (1) implies that T* is bounded and
II T*II ~ II TIl . (2)
Applying the above results to T*, we have (T*(x), y) = (y, T*(x)
= (T(y), x) = (x, T(y) for all x, y E H. Since the adjoint is unique,
this implies that T = (T*)* = T**. We next apply the inequality (2)
with T* taking the role of T to obtain I Til = I T**II ~ I T*II. Thus

I Til = IIT*II· D
(16.41) Theorem. Let Hand "B (H) be as in (16.40). Then the mapping
T ~ T* of Q3 (H) into "B (H) has the following properties:
(i) (Tl + T2)*= TT+ n;
(ii) (oc T)* = fi. T* for oc EK;
(iii) (TIT2)*=T~TT;
(iv) T** = T;
(v) I T* Til = I T112.
Proof. Equality (iv) was established in (16.40). Equalities (i)-(iii)
follow from the uniqueness of the adjoint and the following computa-
tions:
«(Tl + T 2 )(x),y) = <Tdx),y) + <T2 (x),y)
= <x, Tt(y) + <x, n(y)
= <x, (TT + Tr}{y);
«ocT)(x),y) = Ct<T(x),y) = Ct<x, T*(y)

= <x, fi.T*(y);
«(Tl T 2 )(x) , y) = <T1 (T 2 (x)),y) = <T2 (x), Tt(y) = <x, TrTT(y).
We next prove (v). First we have I T* Til ~ I T*II . I Til = I T112. Next,
for all x EH we have
I T(x)i12 = <T(x), T(x) = <x, T* T(x) ~ Ilxil'll T*T (x) I ~ Ilx11 2'11 T* Til;
1 1
hence liT (x) I ~ IIT* TII2"· Ilxll· Therefore I Til ~ I T* TII"2, and so
IITI12 ~ IIT*TII. D
(16.42) Exercise. Let H be a Hilbert space and let M be a closed
linear subspace of H. Prove that there exists a closed linear subspace
MJ. of H such that M.l.. MJ. and H = MEBMJ., i.e., M + MJ. = H
and M n M J. = {o}. [Hint. Consider a complete orthonormal set in M;
extend it to a complete orthonormal set in H; and let M J. be the closed
linear subspace spanned by the added orthonormal vectors.]
(16.43) Exercise. Let H be an inner product space and let A be a
nonvoid subset of H that is complete in the norm metric of H, and also
has the property that ~ (x + y) EA if x, YEA. Prove the following
statements.
(a) The set A is closed in H.
(b) For Xl> ••• , Xn EA and positive real numbers Ctl> ••• , Ctn such that
" n
1: Ct" = 1, the element 1: Ct"X" is in A. [This is the property of convexity.]
"=1 k=1
(c) Every finite-dimensional linear subspace of H is complete in its
norm metric, i. e., it is a Hilbert space.
(16.44) Exercise. Let H and A be as in (16.43) and let z be any

element of H. Prove that there is a unique element Yo EA such that
(i) Ilyo - zll = inf{lly - zll :y EA}.
[The right side of (i) is the distance Irom {z} to A as defined in (6.87)].
[Hints. Cf. (IS. 16.c) , noting that H is uniformly convex; or apply the
parallelogram law directly. Choose a sequence (Yn) of elements in A
such that lim llYn - zll = inf{lly - zll:Y EA}. Apply uniform convexity
1>-+00
to Yn - z and Ym + z to infer that (Yn) is a Cauchy sequence and so has a

limit Yo in A. Uniqueness is proved similarly.]
(16.45) Exercise. Let H be an inner product space and let {Xn}:'=l
be a set of vectors in H. Let Sn be the linear space spanned by {xv . .. , xn}.
Suppose that Xn is the element of Sn nearest to xn+l for n = 1,2,3, ....
[This element exists by (16.43.c) and (16.44).] The set {Xn}:'=l is called
a martingale in the wide sense. Write Yl = Xl and Yn = Xn - Xn -1 for n > 1.
(a) The vectors Yn are pairwise orthogonal and Xn = Yl + ... + Yn'
(b) The inequalities Ilxlll ~ IIx211 ~ ... ~ Ilxnll ~ ... hold.
(c) If {Zn}~l is an orthogonal set, then {Zl + ... + Zn}:'=l is a martin-
gale in the wide sense.
(16.46) Exercise. Let H be a Hilbert space and let TE!:8 (H).
Prove:
(a) if (T(x), y) = 0 for all x, Y EH, then T = 0;
(b) if (T(x), x) = 0 for all x EH, then T = 0;
(c) T* T = TT* if and only if II Txll = II T*xll for all x EH.
(16.47) Exercise. Let H be a Hilbert space and let M be a closed
linear subspace of H. As in (16.42), we have H = M EB Ml.. Prove that:
(a) for each x EH, there exists a unique (y, z) EMx Ml. such that
x = Y + z; define P(x) = Y;
(b)PE~(H);
(c) p2= P;
(d) P= P*;
(e) P(H) = M;
(f) p-l (0) = Ml..
The operator P is known as the projection 01 H onto M.
Prove that if TE!:8(H) satisfies P = T and T = T*, then T is
the projection of H onto some closed subspace of H.
(16.48) Exercise: Construction of a particular projection operator.
(a) Let (X, d, p,) be a measure space such that 0 < p, (X) < 00; write
~2 for ~2(X, d, p,). Let ff be a dense linear subspace of ~2 closed under
the formation of complex conjugates and containing only bounded
functions. Let 9n be a closed linear subspace of ~2 such that 1E9n
and g EiY imply g I E921 [i. e., 921 is invariant under multiplication by
lunctions in iYJ. Let P be the projection of ~2 onto 921 as defined in (16.47).
Prove that peep) = P(I)· ep for all ep E~2 and that P(I) = ~E for some
set E Ed. Thus 921 is exactly the set of all functions in ~2 that vanish
on E'. Note that every such set is plainly a closed subspace of ~2 invariant
under multiplication by functions in iY. [Hints. Consider any hE 921..1.,
IE 921, and g EiY. We have
J I g h dfl = JUg) ii dfl = 0
x x
and so 921..1. too is invariant under multiplication by functions in iY.
Since 1- P(I) E921..1., it follows that P(g(1 - P(I))) = 0 and so
peg) = P(gP(I)) = gP(I) for all g EiY. As iY is dense in ~2' we infer
that P (ep) = ep P (1) for all ep E~2. Since P (I) = P (1)2, we haveP (I) = ~E.]
(b) Let (X, d, fl) be a a-finite measure space, and let 921 and iY
be subspaces of ~2 (X, d, fl) just as in part (a). Prove that the projection
operator P of ~2(X, d, fl) onto 921 has the form P(f) = ~EI for some
00
E Ed.! [Write X = }:11 F n, where the Fn's are pairwise disjoint sets in d
of finite measure. Apply (a) to each subspace ~Fn ~2(X, d, fl) and add.]
(16.49) Exercise. Let H be the real Hilbert space l~ = l~(N). Define
C= {I EH : [I (n) [ ~ : for all n EN}. Prove that C is a compact, no-
where dense subset of H. The metric space C is known as the Hilbert
parallelotope.
(16.50) Exercise. Let x, y, z be elements of an inner product space H
such that [[x[[ = llY[[ = [[z[[. Prove that
(i) [<x, x)<z, y) - <x, y) <z, X)[2 ~ [<x, X)2 - [<y, X)[2][ <x, X)2_[<Z, X)[2].
[Hint. You may obviously suppose that [[x[[ = llY[[ = [[z[[ = 1. Use the
Gram-Schmidt process to replace H by K3, X by (1,0,0), Y by (ex:, fl, 0),
and z by (y, b, s), where [ex:[2 + [fl[2 = [y[2 + [b[2 + [S[2 = 1. Then (i)
becomes nearly triviaL]
(16.51) Exercise. Let x, y, z and H be as in (16.50). Prove that
(i) <x, X)2([<y, Z)[2+ [<y, X)[2+ [<z, X)[2)
~ <X,X)4 + <x,x) <z,y) <y,x)<x,z) + <x,x)<y,z)<x,y)<z,x).
(16.52) Exercise. Prove that a Hilbert space H is separable if and
only if the orthogonal dimension of H is ~ toto [see (16.29)J.
(16.53) Exercise. Let (X, d, fl) be a measure space such that
0< fleX) < 00 and let (d, e) be the metric space defined in (10.45).
Let m be the smallest cardinal number of a dense subset of (d, e).
1 For a yet more general result, see (19.76).
Let b be the orthogonal dimension of the Hilbert space ~2(X,.9I, ft).

Prove the following assertions. If b is finite, then m = 2b. If b is infinite,
then m = b. [Hints. Consider first the case in which ft assumes only
finitely many values, and use (lO.56.a) and (12.60) to prove that m = 2b.
If ft assumes infinitely many values, then (lO.56.c) shows that b is in-
finite. In this case, the inequalities m ;:;;; band b ;:;;; m are proved by
simple arguments.]
(16.54) Exercise. Let H be the real Hilbert space lHR). For each
t ER let U t be that element of H defined on R by Ut (s) = <5 t • s [KRON-
ECKER'S delta]. Define X = {t EH: t = IXUt for some t ER, 0;:;;; IX;:;;; I}.
Then X, with the metric of H, is a metric space. Prove that there exists a
one-to-one mapping cp of X onto the French railroad space (D, e) [see
(6.13.e)] such that both cp and cp-l are uniformly continuous.
(16.55) Exercise. Let H be a Hilbert space and let T EQ3 (H). Write
IX(T) = sup{I(Tx, x)1 : x EH, Ilxll = I}.
(a) Prove that IX(T) = II Til if T = T*. [Use the identity
411Tx113= (T(IITxllx+ Tx), IITxllx+ Tx) - (T(IITxllx- Tx), IITxllx- Tx)
and the parallelogram law.]
(b) Find a T for which IX(T) =l= I Til.
(16.56) Exercise. Use (16.42) to give a proof of (16.31) that does not
invoke the results of § 15. [For t EH*, let M = t-I({O}). If M = H,
let y = O. If M =1= H, let z be a nonzero element of M 1. and set y =
t(z) Il zll-2 z. Note that (x- ji:; z) EM for all x EH.]
(16.57) Exercise. Let H be a Hilbert space and let T Ec:a (H).
Prove that [T-l({O})] 1- [T*(H)] and H = [T-l({O})] + [T*(H)].
[See (16.42). Show that T*(H)1. = T-l({O}).]
(16.58) Exercise. Let H be a Hilbert space and let T Ec:a (H) be
such that T = T*. Suppose that there exists a positive constant c
such that I T (x) I ~ c Ilxll for all x EH. Prove that
(a) T is one-to-one,
(b) T(H) = H,
and
(c) T-IEQ3(H).
[Use (16.57) and (14.17).]
CHAPTER FIVE
Differentiation
This chapter contains first a brief but reasonably complete treatment
of the theory of differentiation for complex-valued functions defined
on intervals of the line. Section 17 is severely classical, containing
examples and LEBESGUE'S famous theorem on differentiation of functions
of finite variation. In § 18, we explore the conditions under which the
classical equality
b
t(b) - t(a) = J f'(t) dt
a
is valid. This exploration leads to interesting and perhaps unexpected

measure-theoretic ideas, which have little to do with differentiation and
which have applications in extraordinarily diverse fields. The main
result in this direction is the LEBESGUE-RADON-NIKODYM theorem, which
we examine thoroughly in § 19 and apply to the decomposition of
measures on R. In § 20, we present several other applications of the
LEBESGUE-RADON-NIKODYM theorem to problems in abstract analysis.
Sections 17 and 18 are important, and should be studied by all readers.
The same is true of § 19, up to and including (19.24). The remainder of
§ 19 may be omitted by readers pressed for time. Of § 20, (20.1)-(20.5)
and (20.41)-(20.52) are topics important for every student. The re-
mainder of § 20 is in our opinion interesting but less vital, and it too
may be omitted by readers pressed for time.
§ 17. Differentiable and nondifferentiable functions

This section deals solely with functions defined on intervals of R.
While reasonably elementary, it is an indispensable introduction to the
more sophisticated matters considered in §§ 18 and 19. Throughout this
section, "almost everywhere" means "A-almost everywhere" and
"measurable" means "vItA-measurable". As usual, we begin with some
definitions.
(17.1) Definition. Let a ERand <5 > O. If g; is a real-valued function
defined on ]a, a + <5[, define
lim g;(h) = sup{inf{g;(h): a < h < t}: a < t ~ a + <5}
Ma
and
§ 17. Differentiable and nondifferentiable functions 257
These two extended real numbers are called the lower right limit and the
upper right limit 01 rp at a respectively. If rp is a real-valued function
defined on ]a - ~, a[, define the lower lelt limit and the upper lelt limit 01 rp
at a to be the extended real numbers
lim rp(h) = sup{inf{rp(h): t < h < a}: a - ~ ~ t < a}
"t"
and
lim rp(h) = inf{sup{rp(h): t < h < a}: a - ~ ~ t < a}
"t"
respectively.
(17.2) Definition. Let a ER and ~ > o. If I is a real-valued function
defined on [a, a + ~ [, define
D I(a) = lim I(a + h) - I(a)
+ "to h
and
D+/(a) = lim I(a + h) - I(a) •
lito h
If I is a real-valued function defined on ]a - ~, a], define
D_/(a) = lim I(a + h) - I(a)
"to h
and
D-/(a) = lim I(a + h) - I(a) •
"to h
These four extended real numbers are known as the Dini derivates 01 I
at a; D+/(a) is the lower right derivate, D+/(a) is the upper right derivate,
D_/(a) is the lower lelt derivate, and D-/(a) is the upper lelt derivate.
(17.3) Remarks. The inequalities
(D+f)(a) ~ (D+/)(a)
and
(D_/)(a) ~ (D- I)(a)
obviously hold. Also it is easy to see that (D+t)(a) [(D+t)(a)] is the
largest [smallest] limit of a sequence (' (a + h~~ - I (a) ) , where h" > 0
and lim h" = o. Similar statements hold for (D- t)(a) and (D_t)(a).
n--->-oo
(17.4) Definition. If (D+f)(a) = (D+t)(a), then I is said to have a

right derivative at a, and we write I~ (a) for the common value (D+f)(a)
= (D+f)(a). The lelt derivative 01 I at a is defined analogously, and is
written t'-- (a). If I~ (a) and 1'- (a) exist and are equal, then I is said to be
differentiable at a, or to have a derivative at a, and we write I' (a) for the
common value I~ (a) = I~ (a). The number f' (a) is called the derivative 01 I
at a. Notice that our definition does not exclude 00 or - 00 as a value for
1
f' (a). For example, if I (x) = x3" (x E R), then f' (0) exists and f' (0) = 00.

258 Chapter V: Differentiation
(17.5) Definition. If I is a complex-valued function defined on

[a, a + <5[, we say that I is right differentiable at a or has a right derivative
at a if
lim I(a + h) - I(a)
"to h
exists and is a complex number. Left derivatives and two-sided deriv-
atives [which are simply called derivatives] are defined similarly. It is
obvious that I has a derivative at a if and only if Rei and Iml have
finite derivatives at a. Then
(Rei), (a) + i (1m I)' (a) .
f' (a) =
(17.6) Remarks. (a) There is a slight solecism in Definitions (17.4)

and (17.5), since a real-valued function is certainly complex-valued,
and ± 00 are admitted as values of f' if I is real-valued. The point should
cause no trouble, however, and we will not bother with the special
terminology that would be needed to remove the difficulty.
(b) If 1'+ (a) exists and is finite, then lim I(a + h) = I(a). Similarly,
litO
if 1'- (a) exists and is finite, then lim I(a
litO
+ h) = I(a). Hence if I has a
finite derivative at a, then it is continuous at a. [Is this true if f' (a) = 00
or - oo?] The following two theorems show that the converse of this
statement fails in a striking way.
(17.7) Theorem. Let a be an odd positive integer and b a real number
such that 0 1 + 32n . Let I be the lunction
defined on R by
00
I(x) = 1: bl< cos [aknx] .

k=O
Then I is continuous and bounded on R, and I has a finite derivative at no
point l .
Proof. For each k we have Ib k cos [aknx]1 ~ bk for all x ER. Thus
the series defining I converges absolutely and II (x) I ~ i: bl<
k=O
= 1 ~b
1 The investigation of the relationship between continuity and differentiability
has a long history. The function I defined here was constructed by WEIERSTRASS
(ca. 1875). It was minutely examined by HARDY in 1916 [Trans. Amer. Math. Soc.
17,301-325 (1916)]. Among other things, HARDY was able to show that I has the
stated properties if ab ~ 1. In the same paper, he showed that the continuous
function
00 sin[n2nx]
g(x) = E n2
11=1
is nowhere differentiable; this is considerably more difficult to prove than the
corresponding statement for f. RIEMANN had conjectured many years earlier that
g is nowhere differentiable.
§ 17. Differentiable and nondifferentiab1e functions 259
.. -1
for all x ER. Also, if f",(x) = 1: bk cos [aknx], we have
k=O
as n ~ 00. It follows as in (7.9) that f is continuous on R.

We will show that f is differentiable at no point of R. Let x ER be
fixed. For n EN and h > 0, write
f(x + k) - f(x) = ,,£1 bk cos [a~n(x + k)] - cos [aAnx]
k k=O k
~ bk cos [akn(x + k)] - cos [aknx] _ S R
+"""
k= ..
k - ",+ ",.
Using the mean value theorem, we write
cos[a~n(x + k)]k - cos [aknx] k ' [k (
=-ansmanx+,
h')]
(1)
°
where < h' < h. The absolute value of the right side of (1) is less than
or equal to a"n, and so
n-l a"b" _ 1 na"b"
IS", I ~ 1: akbkn = n ab _ 1 < ab _ 1 • (2)
k=O
Now write
a"'x = IX", + f3", ,
where IX." is an integer, and - ~ ~ f3", < ~ . Let

hn -_ 1-
an
{J"
•
·
Smce 3
2"" ~ 1-
f1", > 2""1 ,we h ave 3 1 - (J"
2a" ~ -a-"- > 1
2a"
d
an so
2a" 1
-~-<2a'"
3 - k" •
We now estimate IR",I. For k ~ n, consider

akn(x + h",) = ak-"'a"'n(x + h",) = ak-"n(a"'x + 1 - f3,,) = ak-"n(l + IX,,} •
Since a is odd, the equalities

cos [akn(x + h",)] = cos [a k -"n(1 + IX,,)] = (_1)1+«"
hold. We also have
- cos [nakx] = -cos [nak-"a"x] = -cos [na k -" (IX", + f3,,)]

= -cos [nak-"IX.,,] cos [na k -", f3,,] = (- W+IX" cos [na k -"'f3,,].
17*
Upon setting h = hn' we find that IR.. I becomes

IRnl=! £ bk (-1)1+0:,. + (-1)1+o:,.COS[3tak-nPnJ!
k=n hn
since cosn (In ~ 0. Combining this estimate for IRnl with (2), we obtain
I f(x + h;~ - f(x) I ~ IRnl- 15.. 1> n

2a;b - ::~; = (ab)n [~ - ab~1]'
Since ab > 1 + ~ n, [~ - ab~ 1] is a positive constant, and it follows
that
lim If (x + hn) - f (x) I = 00 •
n-+oo hn
Since lim h.. = 0, it is clear that at least one right derivate of

n-+oo
1 at x
is infinite. D
Our next theorem gives an indirect proof that continuous nowhere
differentiable functions exist. It shows actually that in a certain sense
most continuous functions are nowhere differentiable.
(17.8) Theorem. Consider the real Banach space <rr = <rr([o, 1J).
Let ~ = {f E<rr: D+f(x) and D+f(x) are both finite for some x E [0, 1[}.
Then ~ is 01 first category in the complete metric space <rr. Thus the
set of all continuous functions on [0, 1J which have at least one infinite right
derivate at every point of [0, 1[ is dense in <rr([o, IJ).
Proof. For each integer n> 1, let ~n = {f E<rr: there exists an
XE[O,l-~] such that If(x+hl-f(X)I~nforallhE ]O,~]}. We
00 co
first show that ~ = U ~n' Obviously U ~n C ~. Let 1E~. Then there
n=2
exists an x E [0, 1[and a constant IX > °such that for some (), °<() < 1- x,
n=2
the inequality
If(x + h)h - f(x) I <IX
holds for every h EJO, ()[. Select an integer n for which n > max {~ , IX}.
00
It is plain that f E~n' U ~n'

and so ~ C n=2
Next we show that each ~n is closed in <rr. To this end fix n, let
f E~, and choose a sequence (f,J'k=l of functions in ~n such that
Ilf - f"ll" -+ 0. For each k, choose x" E [0, 1 - ~] to correspond to f"
as in the definition of ~n' Since [0, 1 - !] is compact, the sequence
(x,,) has a convergent subsequence whose corresponding subsequence of
§ 17. Difierentiable and nondifferentiable functions 261
(f,,) converges in norm to I; we denote these subsequences again by (x,,)

and (f,,) respectively. Let x = lim x,,; clearly x E
HOO
[0, 1-~].
n
Now fix
hE ]0, ~] and let e> °be arbitrary. Choose k so large that

he he he
III - "'II" < 4 ' I/(x,,) - l(x)1 < 4 ' and II (x" + h) - I(x + h)1 < 4 ·
Then I/(x + h) - l(x)1 ~ I/(x + h) - I (x" + h)l + II (x" + h) - 1,,(x,,+h)1
he he
+ II" (x" + h) - 1,,(x,,)1 + 1/,,(x,,) - I (x,,) I + I/(x,,) - I(x)l < 4 + 4
+ nh + h4e + h4e = h(n + e). Since e is arbitrary, it follows that
If(x + hl- f(x) I~ n
for all hE ]0, ~] ;hence I E 'fn . It follows that 'fn is closed.

We now show that every 'fn has void interior. Assume the contrary,
i.e., that there exist an n, an IE 'fn' and an e> such that ~8(f)
={cp Eerr : III - cpll" < e}c'fn • By WEIERSTRASS'S approximation theorem
°
(7.31), there is a polynomial P such that liP - III" < e. Let ~ = e -liP - III".
Then we have $,,(P) C $.(1) C 'fn . Next we construct a function g E err
such that: Ilgll" < ~,g,+ (x) exists, and Ig,+ (x) I > n + IIP'II" for all x E [0, 1 [.
[Such g's clearly exist; e.g., let g be a nonnegative "sawtooth" function
on [0, 1] with maximum ~ and slopes greater in absolute value than
the constant n + IIP'II".] Then g + P E $,,(P), and we also have
I(g + P)'+I = Ig,+ + P'I ~ Ig'+I-1P'1 ~ Ig'+l- IIP'II" > n
at all points of [0, 1[. Thus g + P ~ 'fn . This contradiction proves that
'f~ = 0 for all n.
We conclude that each set 'fn is nowhere dense in err. Therefore
00
!$) = U 'fn is of first category in err. Since err is a complete metric space,
n=2
it follows from the Baire category theorem (6.54) that err n !$)' is dense
in err • l 0
The technique used in the proof of (17.7) is important. Many existence
proofs throughout analysis and set-theoretic topology are carried out
in just this way.
We next examine the extent to which a function can have different
right and left derivatives.
1 Many writers have made constructions of this sort. Our construction is taken
from S. BANACH, Studia Math. 3, 174-179 (1931). See also K. KURATOWSKI.
Topologie I. Deuxieme Edition. Monografie Matematyczne. Tom XX. Warszawa-
Wroclaw. 1948. pp. 326-328.
(17.9) Theorem. Let ]a, b[ be any open interval 01 R and let I be an

arbitrary real-valued lunction defined on ]a, be. Then there exist only count-
ably many points x E]a, b[ such that 1'- (x) and l't (x) both exist [they
may be infinite} and are not equal.
Proof. Let A = {xE ]a, b[: 1'- (x) exists, l't (x) exists, l't (x) < 1'- (x)}
and let B = {xE ]a, b[: 1'- (x) exists, l't (x) exists, l't (x) > 1'- (x)}. For
each x EA, choose a rational number rx such that l't (x) < rx < 1'- (x).
Next choose rational numbers Sx and tx such that a < Sx < x < tx < b,
fry) -f(x) >rx if sx<y<x, (1)
y-x
and
fry) -f(x) <rx if x<y<t x ' (2)
y-x
Combining (1) and (2), we have
I(y) - I(x) < rx(Y - x) (3)
whenever y =l= x and Sx < Y < tx' Thus we obtain a function cp from A
into the countable set Q3, defined by <p(x) = (rx' sx, tx). We will prove
that A is countable by showing that cp is one-to-one. Assume that there
are distinct x and y in A such that cp (x) = cp (y). Then ]s:v' t:v[ = ]sx' tx[,
and x and yare both in this interval. It follows from (3) that
I(y) - I(x) < rx(Y - x)
and
I(x) - I(y) < r:v(x - y) .
Since rx = r:v, adding these two inequalities yields 0 < O. This is a con-
tradiction, so that cp is one-to-one and A is countable. Similar reason-
ing proves that B is countable. 0
Our next goal is to prove H. LEBESGUE'S famous theorem that a
monotone function has a finite derivative almost everywhere. The main
tool used in the proof is a remarkable theorem of VITALI, which we
present next. VITALI'S theorem has numerous applications in classical
analysis, particularly in the theory of differentiation.
(17.10) Definition. Let E C R. A family l ' of closed intervals of R,
each having positive length, is called a Vitali cover 01 E if for each x EE
and each e > 0 there exists an interval I El ' such that x EI and A(I) < e,
i.e., each point of E is in arbitrarily short intervals of 1'.
(17.11) VITALI'S covering theorem. Let E be an arbitrary subset 01 R
and let l ' be any [nonvoid} Vitali cover 01 E. Then there exists a pairwise
disjoint countable lamily {In} C l ' such that
ACE n (IJ In)') = 0.
Moreover, il A(E) < 00, then lor each e > 0 there exists a pairwise disjoint
finite family {II> ... , Ip} C "Y such that
A (E n C~1 In)') < e.

ProoF. Case I: A(E) < 00. Choose an open set V such that E C V
and A(V) < 00. Let 1';; = {I E"Y: I C V}. Plainly 1';; is a Vitali cover of E.
Let II E1';;. If E C II> the construction is complete. Otherwise we con-
tinue by induction as follows. Suppose that II> 12 , ••• , In have been
n
selected and are pairwise disjoint. If E C U III' the construction is
k=1
complete. Otherwise, write
n
An = k=1
U III, Un = V n A~ .
Clearly An is closed, Un is open, and Un n E =1= 0. Let
15n = sup{A(I): I Ej/~, I C Un}. (1)
1
Choose In +1 E1';; such that In +1 C Un and A(In +1) > 215n- If our
process does not stop after a finite number of steps [in which case there
is nothing left to prove], then it yields an infinite sequence (In)~1 of
00
pairwise disjoint members of 1';;. Let A = U In. We must show that

n=1
A(E n A') = 0. For each n, let In be the closed interval having the same
midpoint as In and such that
A(]n) = 5A(In) .
We have
A CQl In) ~ EI A(]n) = 5 E}(In)
= 5A(A) ~ 5A(V) < 00. (2)
Theorem (10.15) shows that
lim
p-+oo
A(U
n=p
In)=O.
Thus, to prove that A(E
00
n A') = 0, it suffices to prove that En A'
C U In for every
n=p
PEN. Fix PEN and let x EEn A'. Then we have
x EE n Ap C Up, and so there exists an I E1';; such that x EI CUp.
It is evident that 15n < 2A(In+1)' and (2) shows that A(In) --+ as n --+ 00.
Hence there is an integer n such that 15n < A{I). Thus, by (1), there
°
exists an integer n such that I ¢ Un; let q be the smallest such integer.
It is obvious that P < q. We infer that
I n Aq =1= 0 and I n A q_1 = 0 .
1 We give the ingenious proof of this theorem due to S. BANACH [Fund. Math.
6, 170-188 (1924)J.
It follows that
(3)
and, since Ie Uq _ v we have
1.(1) ~ <5 q_1 < 21. (Iq) . (4)
Since A.(}q} = 51. (Iq) , (3) and (4) show that
n A 'eU
00 00
so that x E U
n=p
In. Hence we have E
n=p
In' which implies that
A.(E n A'} = o.
Now let e > 0 be given and choose an integer p so large that
00
}; A.(In) < e .
n=p+l
Then
En Ap C (E n A') U (
n=p+l
U In),
and so
A(E U In) <
n Ap} ~ 0 + A ( n=p+l e.
Thus the proof is finished if A(E) < 00.
Case II: A(E) = 00. For each nEZ, let En = E n ]n, n + I[ and let
n.
i'; = {I Er : I C ]n, n + 1 Clearly i'; is a Vitali cover of En. Apply
Case I to find a countable pairwise disjoint family J" C i'; such that
ACEn n (U J,,),) = 0 for each n EZ. Let J
00
= U J". Then J is a countable

n=-oo
pairwise disjoint subfamily of rand
En (U J)' cZ U [n=l2oo(En n (UJ,,)/)] .

We see that
00
ACE n (U J)') ~ A(Z} + }; 0= o. 0

n=-oo
(17.12) Theorem [LEBESGUE]. Let [a, b] be a closed interval in R

and let f be a real-valued monotone function on [a, bJ. Then f has a finite
derivative almost everywhere on [a, bJ.
Proof. We suppose that f is non decreasing [otherwise consider - tJ.
Let E = {x: a ~ x < b, D+f(x} < D+f(x)}. We will first show that
A(E) = o. For every pair of positive rational numbers u and v such that
u < v, let
Eu,v = {x EE: D+f(x} < u < v < D+f(x)} .
Clearly E = U {Eu,v: u, v EQ, 0 < u < v}. Since this is a countable union,
it suffices to show that A(Eu, v) = 0 for all 0 O. Let 8 be such that

IX (v - u)
0< 8 ElL." such that A(U) < 0: + 8. For each x EE".",
there exist arbitrarily small positive numbers h such that [x, x + h] C U
n [a, b] and
f(x+h)-f(x)<uh. (1)
The family "f/ of all such closed intervals is a Vitali cover of E u ,,,, and so,
by (17.11), there exists a finite, pairwise disjoint subfamily {[Xi' Xi+ hi J}i"=l
of "f/ such that
...
Let V = .U ]Xi' Xi
.=1
+ hie. Then we have
A(Eu•v n V') < 8. (2)
The inclusion V C U implies that
...
1: hi = A(V) ~ A(U) < 0: + 8,
i=1
and so (1) yields the inequalities

m m
1: (t(Xi + hi) - f(x i )) vk . (4)
The family of all such closed intervals is a Vitali cover of Eu,v n V,
and so there is a finite, pairwise disjoint family {[Yj' Yj + kj J}j'=l of such
intervals with the property that
A (Eu, " n V n C91[Y;'Y; + k j ])') < 8.
This inequality together with (2) implies that
0: = A(ElL,,,) ~ A(Eu•v n V') + A(Eu,,, n V) < 8+ (8 + i kj ) . (5)

Next, using (4) and (5), we have
n n
v (0: - 28) < v 1: k; < 1: (t(YJ + k;) - f(y,)). (6)
i=1 ;=1
n m
Since.U
1=1
[Yj' Y; + k,] c .U [Xi' Xi
.=1
+ hi] and f is nondecreasing, we also
have
n m
L (t(y; + k;) - 1(y;)) ~ L (f(Xi + hi) - 1(Xi)) . (7)
i=1 i=1
Combining (6), (7), and (3) gives

v(a - 2e) < u(a + e) ,
which contradicts our choice of e. Thus A(E) = 0, and so (x) exists I:
a.e. on [a, bJ. Similarly ( (x) exists a.e. on [a, bJ. Now apply (17.9)
to see that f' (x) exists a. e. on [a, bJ.
It remains only to show that the set F of points x in Ja, b[ for which
f' (x) = 00 has measure zero. Let P be an arbitrary positive number.
For each x EF, there exist arbitrarily small positive numbers h such that
[x, x + hJ C Ja, b[ and
I(x + h) - I(x) > ph. (8)
By VITALI'S theorem (17.11), there exists a countable pairwise disjoint
family {[xn' Xn + hnJ} of these intervals such that
A(F n (';I [xn' Xn + hnJ)') = 0.
From this fact and (8) we obtain
PA(F) ~ PI; hn < I; (t(xn + hn) - 1(xn)) ~ I(b) - I(a) .
n n
Thus
PA(F) </(b)-/(a) forall PER,
which implies that A(F) = 0. 0
(17.13) Question. Suppose that A(A) = 0, A C [a, bJ. Is it possible
to find a monotone function 1 on [a, bJ such that f' exists exactly on
A' n Ja, be? The complete answer seems to be unknown.
(17.14) Definition. Let 1 be a complex-valued function defined on
[a, bJ cR. Define
V;I = sup {i;

k=1
I!(x,,) - l(x"-l)1 : a = Xo < Xl < ... < Xn = b} .
The extended real number V;
1 is called the total variation allover
[a, bJ. If V;I < 00, then I is said to be of finite variation [or bounded
variation J over [a, bJ.
(17.15) Remarks. (a) The function I has finite variation if and only
if the functions Rei and Iml have finite variation.
(b) The equality Yab I + VN = V; I holds for a < b < c.
V:
(c) The function x -7 I is nondecreasing.
(17.16) Theorem (Jordan decomposition theorem]. A real-valued
lunction 01 finite variation is the difference 01 two nondecreasing lunctions.
Proof. Write I(x) = V:I - (V:I - I(x)), where we define "Vaal = O.

V:
Evidently the function x -? I is nondecreasing. The function
x -? "Vax I - I (x)
is also nondecreasing, for if x' > x, then
V;' 1- I (x') - (V: 1- I (x))= V:' 1- (I(x') - I(x)) ;;;; O. 0
(17.17) Theorem [LEBESGUE]. A complex-valued lunction 01 finite
variation has a finite derivative a. e.
Proof. This is an immediate consequence of (17.16), (17.15.a), (17.12),
and (17.5). 0
(17.18) Theorem [FUBINIP. Let (Inr:=1 be a sequence olnondecreasing
[or nonincreasing] real-valued lunctions on an interval [a, b] such that
00
}; In (x) = s (x) exists and is finite in [a, bJ. Then
..=1
00
(i) s' (x) = }; I~(x)

.. =1
a. e. in ]a, b[.
Proof. There is no harm in supposing [and we do] that all In are
nondecreasing. Also, by considering the functions In - In(a), we may
00
suppose that In ;;;; O. Thus s = }; In is nonnegative and nondecreasing.

.. =1
The derivative s' (x) exists and is finite for almost all x E]a, b[, as (17.12)
shows.
Consider then the partial sums Sn = 11 + 12 + ... + I... and the
remainders r n = S - Sn. Each Ii has a finite derivative a. e.; hence there
is a set A C ]a, b[ such that A(A' n ]a, b[) = 0,
s~(x) = I~ (x) + I~ (x) + ... + I~(x) < 00
for all x EA and all n, and s' (x) exists and is finite for x EA. For any
x E]a, b[ and every h> 0 such that x + h E]a, b[, it follows from the
equality
Six + h) - six) sn(x + h) - sn(x) rn(x + h) - r,,(x)
h = h + h
that
S,,(x + h) - sn(x) S s(x + h) - s(x) .
h - h '
and this inequality implies that s~ (x) ;;;;; s' (x) for all x EA. The inequality
s~ (x) ;;;;; S~+1 (x) is clear, and so we have
s~ (x) ~ s~+1 (x) ~ s' (x)
1 This is not the theorem ordinarily called "FUBINI'S theorem", which deals
with product measures and integrals and will be taken up in Chapter Six.
for x EA and n = 1,2, .... Hence

1: Ii (x)
00
lim s~ (x) =
11-->00 i= 1
exists a. e., and it remains to show that lim s~ (x) = s' (x) a. e. Since the
,,~oo
sequence (s~(X):=l is nondecreasing for each x EA, it suffices to show

that (s~) admits a subsequence converging a.e. to s'. To this end, let
nl> n 2, ... , nk, . .. be an increasing sequence of integers such that
00
1:
[s(b) - s"k(b)J < 00.
k=l
For each nk and for every x EJa, be, we have
o~ s(x) - S"k(X) ~ s(b) - s"k(b) .
The terms on the left side of this inequality are bounded by the terms of
1: [s (x)
00
a convergent series of nonnegative terms. Hence - snk (x) J

k=l
converges. The terms of this series are monotone functions that have
finite derivatives a. e. Therefore the argument used above to prove that
1: Ii (x) 1: [s' (x) -
00 00
converges a. e. also proves that S~k (x)J converges

i=l k=l
a. e.; and of course it follows that lim S~k (x) = s' (x) a. e. 0
k~oo
We close this section with a long collection of exercises. A number are

merely illustrative examples; several are minor theorems with sketched
proofs [(17.24), (17.25), (17.26), (17.27), (17.31), (17.36), (17.37)J; and
(17.33) and (17.34) are needed for later theorems of the main text. The
reader should bear these facts in mind when doing the exercises.
(17.19) Exercise. Let 1p be LEBESGUE'S singular function, defined
in (8.28). Compute all of the derivates of 1p at each point of [0, IJ.
(17.20) Exercise. Define the function f{J on R by
f{J(x) = Ix
1- x if
f{J(x + k) = f{J(x) for all k EZ .

Let
1: 2- n f{J(2 n x) .
00
I(x) =
Prove that I is continuous on R. Compute all four derivates of I at each

dyadic rational point. Prove that I fails to have a finite derivative at
every point not a dyadic rational.
(17.21) Exercise. Let {an}~l be a set of distinct points in the inter-
val [a, bJ. Let (Un)~l and (V n);:"=l be sequences of real numbers such
00 00
that 1: Iunl < 00 and 1: Ivnl < 00. Define

.. =1 .. =1
1
0 if x<an ,
In (x) = Un if X = an ,
Vn if x> an.
00
Prove that s (x) = 1: In (x) has a finite derivative a. e., and that s' (x) = 0
.. =1
a. e. [Hint. The function s has finite variation; find each V; by using
the numbers Iunl and Ivnl. Then apply (17.18).]
(17.22) Exercise. Find a real-valued strictly increasing function I
on R such that I' (x) = 0 a. e.
(17.23) Exercise. Let I E~r([a, b]). Suppose that there exist real
constants ex < f3 such that
for all x E [a, be. Prove that

h ex ~ I (x + h) - I (x) ~ h f3
if a ~ x<x + h ~ b. [Hint. Assuming that
I(xo + ho) -I(xo) < yho < exho
and writing hI = sup{h: 0 < h < ho, I(xo + h) -I(xo) ~ yh}, show that
D+/(xo + hI) < ex.]
(17.24) Exercise. Let I be a function in ~r([a, b]) and let c be a
number in ]a, be. Suppose that D+ I (c) is finite and that D+ I is contin-
uous at c. Prove that I'(c) exists. [Use (17.23).]
(17.25) Exercise. Let I be a real-valued nondecreasing function on
[a, bJ. Suppose that U ~ 0 and E C [a, b] are such that for each x E E,
there exists some derivate of I at x which does not exceed u. Prove that
J..(I(E)) ~ uJ..(E). [Consider an appropriate Vitali cover of I(A), where
A = {x E E: I (x) =1= I(y) for all y E [a, b] n {x}'}. Notice that I(E n A')
is countable.]
(17.26) Exercise. Let I be as in (17.25). Suppose that v ~ 0 and
Fe [a, b] are such that for each x EF, some derivate of I at x is greater
than or equal to v. Prove that J..(I(F)) ~ vJ..(F). [Consider an appro-
priate Vitali cover of B, where B = {x EF: I is continuous at x}. Notice
that F n B' is countable.]
(17.27) Exercise. Let I be a real-valued function defined on [a, b].
Suppose that c ~ 0 and E C [a, b] are such that I' (x) exists and II' (x) I ~ c
for all xEE. Prove that J..(I(E)) ~ cJ..(E). [Consider a Vitali cover of
I (E) by intervals [I (x), I(x + h)] such that I([x, x + h]) C [I(x),
I(x + h)].]
(17.28) Exercise. Let E be a subset of R that is the union of a family

of quite arbitrary intervals, each being open, closed, or half open and
half closed. Prove that E is Lebesgue measurable. [Use VITALI'S theorem.J
(17.29) Exercise. Let a; and (3 be positive real numbers. Define I on
[0, IJ by I(x) = x'" sin (rll) (0 < x ~ 1), 1(0) = 0. Prove that I is of
finite variation on [0, IJ if and only if a; > (3.
(17.30) Exercise. Prove or disprove the following statement. If I
is a function in <rr([o, 1]), then there exist a, b ERsuchthatO ~ a <b~ 1
and I is of finite variation on [a, bJ.
(17.31) Exercise. A function I defined on an interval I of R is said
°
to satisfy a Lipschitz condition 01 order a; > if there exists a constant
°
M ~ such that
I/(x) - l(y)1 ~ Mix - YI'"
for all x, y E I. We write I E~ip",(I). Prove the following.
(a) If a; > 1 and I E ~ip",(I), then I is a constant.
(b) If 0< a; < 1, then there exists a function I E~ip",([O, IJ) such
that I has infinite variation over [0, IJ.
(c) There exists a continuous function of finite variation on [0, IJ
which satisfies no Lipschitz condition.
(d) If I E<rr( [a, bJ), then I E~ipi ([a, bJ) if and only if D+ I is bounded
on [a, be. [Hint. Use (17.23).J
(17.32) Exercise. Let I be a complex-valued function of finite varia-
tion on [a, bJ. Suppose that I is continuous at c E [a, bJ. Prove that the
function g : x --0> v,.x I, where g (a) = 0, is continuous at c.
(17.33) Exercise. Let I be a function in <rr([a, bJ). For each sub-
division LI = {a = Xo < Xl < ... < Xn = b} of [a, bJ, define
ILl I = max{xk - X k - l : 1 ~ k ~ n} and
Wk = max{/(x) : Xk-l ~ x ~ xk } - min{/(x) : X k - l ~ x ~ xk }
for k = 1, ... , n. Prove that
n
V;I = lim I: Wk'
i.dl-?O k=!
(17.34) Exercise. Let I be a function in (P([a, bJ). For each y ER,
let A:v = {x E [a, bJ : f(x) = y}. Define v on R by
A:v if A:v is finite,
{
v (y) = 00 if A:v is infinite.
Prove that the function v is Lebesgue measurable and that
J v (y) dy = V; I .
R
[Hint. Use (17.33). Let Lli C Ll2 C ... be a sequence of subdivisions of

[a, bJ such that ILlnl--o> 0, say LIn = {a = x&n) < ... < x~ = b}. iFor
§ 17. Differentiable and non differentiable functions 271
each n, define
.....
Vn = I: ~Bn.k '
k=l
where Bn,k = 1([X~~l' x~")]). Prove that vn(y) -+ v(y) for ahnost all y ER
and apply B. LEVI'S monotone convergence theorem.] The function v is
known as the Banach indicatrix 01 I.
(17.35) Exercise. For an interval [a, b] c R, let ~ ([a, b]) denote the
set of all complex-valued functions I on [a, b] such that V; 1< 00 and
I(a) = O. For I E~ ([a, b]), define II1II = Yab I. Prove (a)-(c) and answer
(d) and (e).
(a) With pointwise operations ~ ([a, b]) is a complex linear algebra.
(b) For I E~ ([a, b]), the inequality 11/11 .. ~ IIIII holds.
(c) With the above norm, ~ ([a, b]) is a Banach space.
(d) Is it true that Il/gll ~ II1II . Ilgll for all I, g E~ ([a, b]) ?
(e) Find the least cardinal number of a dense subset of ~ ([a, b])
in the topology defined by the variation norm.
(17.36) Exercise. Let x be a real number and let I be a real-valued
function defined in a neighborhood of x. The upper [lower] first and
second symmetric derivatives of I at x are defined to be the limits superior
[inferior] of the expressions
f(x + h) - f(x- h)
(1)
2h
and
I(x + h) + f(x - h) - 2f(x)
(2)
hI
as h t 0, respectively. These derivatives are denoted by D11 (x) and
IJ21(x) Wd (x) and lI21 (x)] respectively. If Dd (x) = lI11 (x) [IJ 21(x)
= lI 2 /(x)], we call this common value the first [second] symmetric deriv-
ative 01 I at x and denote it by Dd(x) [D 2 /(x)]. Prove the following.
(a) If f' (x) exists, then so does D11 (x) and they are equal.
(b) The converse of (a) is false.
(c) If f' exists and is finite in a neighborhood of x and I" (x) is finite,
then D 2 /(x) exists and is equal to f" (x). [Use the mean value theorem
on (2) as a function of h.]
(d) D 2 /(x) may exist even when I is continuous only at x.
(17.37) Exercise: More on convex functions. Let I be an open interval
in R and let I be a convex function [see (13.34)] defined on I.
(a) Prove that 1'+ (x) and 1'- (x) exist and are finite for all x EI,
also that 1'+ and 1'- are nondecreasing functions and 1'- ~ 1'+ on I.
Thus f' exists and is finite a. e. on I. [Hints. For x < y < Z, we have
z-y y-x z-y y-x

y=--x+
z-x
- - z and so fey)
z-x
~ --f(x)
z-x
+ --fez).
z-x
Hence
t(y) - t(x) ~ t(z) - t(x) ~ t(z) - t(y) (1)
y-x - z-x z-y
From (1) our assertions follow easily.]
(b) Prove that f is in ~if.\([a, bJ) for all closed bounded subintervals
[a, b] of I. [For a < x < y < b, (1) implies that
t(x) - t(a) < t(y) - t(x) < t(b) - t(y)
x-a - y-x = b-y
and from this and (a) it is easy to see that
It(~ := ~(x) I ~ max{lf~ (a)l, It~ (b)I}.J

(c) Let g be a real-valued function on an open interval I in R.
Prove that g is convex if and only if g is continuous and D 2g ~ 0 on
I [see (17.36)]. [First suppose that D2 g > 0 on I, assume that g is not
convex on I, and find a point x such that D2g(X) ~ o. Next consider the
functions gn (x) = g (x) +~ X2, n = 1, 2, .... J
§ 18. Absolutely continuous functions

In this section, we identify the class of functions F of the form
x
F(x)=Jf(t)dt for fE~l([a,b]). We also identify the functions on
a
intervals of R that are integrals of their derivatives. This study leads
directly to some classical facts in the theory of Fourier series, which we
also take up. As in § 17, "almost everywhere" means "A-almost every-
where", and "measurable" means "Jt).-measurable". We begin with
some simple theorems.
(18.1) Theorem. Let f E~1 ([a, b]) and define F on [a, b] by
x
F(x) = J f(t) dt.
a
[The function F is called the indefinite integral of f.] Then F is uniformly

b
continuous and has finite variation, and V: F = J It(t)1 dt. A similar
a
x
assertion holds for f E~l(R) and F(x) = J f(t) dt. [If q; is a complex-
-00
valued function on R, then we define V~oo(q;) = lim V!Aq;; V~oo(q;)

A--..oo
and v,.oo (q;) are defined similarly.]
§ 18. Absolutely continuous functions 273
Proof. For x' > x, the equality IF(x') - F(x)1 = Iff(t) dtl holds;
therefore it is clear from (12.34) that F is uniformly continuous. If
a = Xo < Xl < ... < X" = b, then we have
i,IF(Xk) - F(Xk_l)1 = kg I.li f(t) dtl ~ k~ ./: If(t)1 dt = ! If(t)1 dt.

b
Hence the inequality V': F ~ J If(t)1 dt holds, and so F has finite
variation. "
To prove the reversed inequality, first recall that step functions
.
a= 1: rx-k ~L"i-l''''k[ (a = Xo < Xl < ... < X" = b) (1)
k=l
are dense in ~l([a. b]) (13.23). Consider the function sgnf; for every
positive integer m, select a step function am of the form (1) such that
Ilam - sgnflll < !. (2)

Since Isgnf (x) I = 1 or 0 for every x, it is easy to see that the inequality (2)
is only improved by replacing every rx-k such that Irx-kl > 1 by the number
rx-klrx-kl- l . There is thus no harm in supposing that lam (x) I ~ 1 for all
X E [a, b] and mEN. Clearly am -+ sgnf in measure, and so by (11.26)
there is a subsequence (am) of (am) such that
~im amj(t) = sgnf(t) a.e. in [a, b] .
1-->00
We then infer from LEBESGUE'S theorem on dominated convergence

(12.30) that
b b b
J lI(t) I dt = aJ f(t) sgnf(t) dt = ,~im
a
J f(t) amj(t) dt.
.... a 00
(3)
Since amj has the form (1), the absolute value of the last integral in (3)
has the form
..
~ 1: Irx-kl . IF(Xk) - F(Xk_l)1
.
k=l
~ 1: IF(Xk) - F(Xk_l)1 ~ V: F . (4)

k=l

b
J If(t)1 dt ;£; V:F. 0
" 18
Hewitt/Stromberg, Real and abstract analysis
The foregoing theorem shows that indefinite integrals are continuous

and have finite variation, We wish now to show that the derivative of
an indefinite integral is the integrated function [a, e,!]. To prove this,
we need a preliminary, which is of some interest in its own right,
(18.2) Theorem. Let A be an arbitrary subset 01 R, Then
(1') Iim A(An]x,x+k[) =
I'1m --'---"-:----=-
A(An]x-h.x[)
ktO k "to h
= lim A(A n]x - h. x + k[) = 1
".ktO h+k
lor almost all x EA, II A is A-measurable, the limits in (i) are equal to zero
lor almost all x EA',1
Proof. With no harm done, we can [and do] suppose that A is bounded.
There are bounded open sets Un' n = 1, 2, ' . " such that
U1 :> U2 :> •.• :> Un:> •.• :> A
and A(Un) - 2- n < A(A), Let a = inf Uv and consider the functions
CPn(x) = A(Un n ]a, x[)
and
cp(x) = A(A n ]a, x[) .
For x EUn and sufficiently small positive h, it is clear that
1P.. (x + h) - 1P.. (x) = 1P.. (x) - 1P.. (x - h) = l'
h h '
hence cp~ (x) exists for all x E Un and
cp~(x) = 1.
We want to apply FUBINI'S theorem (17.18) to the sum
(911 - 91) + (912 - 91) + ... + (CPn - 91) + ... ;
we first show that each cpn - 91 is monotone. For x' > x, we have
CPn (x') - 91 (x') - (CPn (x) - 91 (x)
= A(Un n [x, x'[) - A(A n ]a, x'[) + A(A n ]a, x[)
~ A(Un n [x, x'[) - A(A n [x, x'[) ~ 0
because
A(A n ]a, x'[) ~ A(A n ]a, x[) + A(A n [x, x'[)
and
A n [x, x'[ C Un n [x, x'[;
thus CPn - 91 is monotone. Now let b = sup U1; then
CPn(b) - cp(b) = A(Un) - A(A) < 2- n,
1 Points x for which the relations (i) hold are called points 0/ density of A.
§ IS. Absolutely continuous functions 275
and so for a ~ x ~ b we have

00 00 00
~ (tp,,(x) - tp(x)) ~ ~ (tp,,(b) - tp(b)) ~ ~ 2-" < 00 •

.. =1 .. =1 .. =1
Let 00
s(x) = ~ (tp,,(x) - tp(x)) .

.. =1
By (17.18) and (17.12), the relations

00
s' (x) = ~ (tp~ (x) - tp' (x)) < 00

.. =1
hold for almost all x in ]a, b[, and so also we have

lim tp~ (x) = tp' (x)
........ 00
00
a. e. in ]a, b[. Thus tp' (x) = 1 on ..n

=1
U" except on a set of A-measure zero,
and this implies the first assertion of the theorem. 1
If A is A-measurable, then
1= }.(An]x-h,x+kD + }.(A'n]x-h,x+kD
h+k h+k
= V' A (x) + V'A' (x)
for all h, k. As hand k go to 0, V'A' (x) goes to 1 for almost all x EA'
[apply the first part of the theorem to the set A']. Hence V'A (x) goes to
zero a.e. on A'. 0
(18.3) Theorem. Let I E·~.\([a, b]), and let F be as in (18.1). Then the
equality
(i) F' (x) = I (x)
holds lor almost all x E]a, bE.
Proof. If 1= EA, where A is a measurable subset of ]a, bE, then
F(x) = A(]a, x[ n A); and (18.2) shows that F' (x) = EA (x) a.e. in
n
]a, bE. Next, let s = ~ IX" EAk be a nonnegative simple measurable func-
k=1
tion, so that % n %
5(x)=Js(t)dt= };IX"JEAk(t)dt.
a "=1 a
Theorem (18.2) implies that
5'(x)=s(x) a.e.in]a,b[. (1)
For a nonnegative function I in ~l' let (S,,):=1 be a nondecreasing se-
quence of simple measurable functions such that lim s" (x) = I (x) for all
........ 00
1 We have actually proved a little more than claimed in the theorem. We have
n U.. ;
00
9" (x) = 1 a.e. on the set if A is nonmeasurable, the nonmeasurable set
(n
..-1
.. -1
U.. ) n A' does not have measure O.
lS*
x
X E [a, b] (11.35). Write S .. (x) = J s.. (t) dt for n EN; B. LEVI'S theorem
II
(12.22) shows that

x x
F(x) = J f(t) dt = n-+oo
a
lim J sn(t) dt = lim S.. (x)
a n-+oo
00
= SI (X) +E [Sn+1 (X) - S.. (x)] (2)

,,=1
for all X E [a, b]. Each function S"+1 - Sn is the integral of a nonnegative
function and so is non decreasing. FUBINI'S theorem (17.18) applied to
(2) gives us the equalities
00
F'(x) = S~(x) +E [S~+1(x) - S~(x)]

,,=1
= lim S~(x) a.e. in ]a, b[, (3)
"-+00
and (1) gives us

lim S~ (x) = lim s.. (x) a. e. in ]a, b[ • (4)
tJ-+oo n-+oo
Combining (3) and (4), we obtain (i).

Finally, if I is an arbitrary function in ~1 ([a, b]), write
f = (11 - 12) + i(la - 14)
where Ii E~t and apply (i) for nonnegative functions. 0
Theorem (18.3) can be sharpened considerably, as the next two
assertions show.
(18.4) Lemma [LEBESGUE]. Let I be a function in ~([a, b]). Then
there is a set E C ]a, b[ such that A(E' n [a, b]) = 0 and
x+" x
(i) lim J If (t)
IXI dt =
- lim J II (t) - IXI dt = II (x) - IXI
lito " h.j.O x-"
lor all IX EK and all x EE.
Proof. Let {P"}:'=1 be any countable dense subset of K. The functions
g.. defined by
g.. (t) = I/(t) - P..I (n EN)
are in ~1([a, b]). By (18.3), there are sets E.. C ]a, b[ such that
A(E~ n [a, b]) = 0 and
x+" "
lim hl j g.. (t) dt = lim hl jg.. (t) dt = g,,(x)
"to x
"to x-h
n E .. ; clearly A(E' n [a, b]) = o.

00
for all xEEn . Let E be the intersection ,,=1

IX EK, IP.. - IXI < 3. Then we have
Ii
For e> 0 and select an n such that
11/(t) - 1X1-I/(t) - P.. II ~ IP.. - IXI <; for all t Era, b] .

It follows that
s+h %+11 s+h
!f I/(t) - (XI dt - !f I/(t) - fJnl dt ~ !f ; dt = ; ,
and in turn that
s+h
!f %
I/(t) - 0:1 dt - I/(x) - 0:1
%+h s+h
~ !f I/(t) - 0:1 dt - !f I/(t) - fJnl dt
if x EE and 0 < h < ho' where ho depends on e and n. But n depends

only on e and 0:. Thus we conclude that
%+11
lim~f
11+0 h
I/(t) - 0:1 dt = I/(x) - 0:1
s
for all x EE and 0: EK. A similar argument shows that
lim~
litO h
f "I/(t) - 0:1 dt = I/(x) - 0:1
x-II
for all x EE and 0: EK. 0

(18.5) Theorem [LEBESGUE]. Let IE~l([a, b]). Then we have
II
(i) lim
lito
hI f I/(x + t) + I(x - t) - 2/(x)1 dt =0
o
lor almost all x E]a, b[.
Proof. For fixed x E ]a, bL write
II
! of I/(x + t) + I(x - t) - 2/(x)1 dt
%+11 %
~ !f I/(t) - l(x)1 dt +! f I/(t) - I(x)l dt.

%-11
Applying (18.4) with 0: = I(x), we see that (i) holds for almost all
x E]a, b[. 0
(18.6) Definition. Suppose that I E~l ([a, bJ) and that x EJa,b[.
Then x is called a Lebesgue point lor I if (18.5.i) holds. The set of all
Lebesgue points for I is called the Lebesgue set lor I.
It is obvious that every point of Ja, b[ at which I is continuous is
a Lebesgue point for I. However I need not be continuous anywhere,
and yet almost every point is a Lebesgue point. The Lebesgue set plays
an important role in the theory of Fourier series and integrals, as we
shall see in (18.29) and (21.43).
We now inquire into the "reverse" [not really a converseJ of Theorem
(18.3). Given a continuous function cp that is differentiable a.e., is it
true that
x
cp(x) - cp(a) = J cp'(t) dt?
a
That is, does the fundamental theorem of the calculus hold with the
Riemann integral replaced by the Lebesgue integral and differentiability
replaced by differentiability a. e.? The following examples answer this
question with an emphatic "no".
(18.7) Example. As in (8.28), let 1jJ be LEBESGUE'S singular function.
Then 1jJ is continuous on [0, 1J and it is clear that 1jJ' (x) = for all x °
in [0, IJ that are not in P [CANTOR'S ternary set]. Thus 1jJ' (x) = a.e. °
It follows that
1jJ(I) -1jJ(0) = 1 =!= ° =
o
1
J 1jJ'(x) dx.
The next example seems not as geometrically obvious as (18.7),
but it is much more dramatic.
(18.8) Example [Adapted from RIEsz-NAGY]. We will exhibit a
real-valued function F on [0, IJ such that F(O) = 0, F(I) = 1, F is con-
tinuous and strictly increasing, and F' (x) = °
a.e. We define continuous
functions Fn, n = 0, 1, 2, ... , inductively as follows. First, let (tn);:'=l
be any sequence of numbers in JO, 1 [. Let Fo (x) = x, and define Pi (0) = 0,
Pi(I) = 1, and Pi (21) = I-t
-2-
1
• °+ l+t
-2-
1
,1; and define FI to be
linear on [0,
~] and [~, 1]. Suppose that Fo' FI> ... , Fn have been
defined. Then we define:
Fn+I(;n)=Fn(;n) for k=0,1, ... ,2n ;
F.
n+l
+ 1)
( 2k2n+1 = 1 - tn+1 F. (~)
2 n 2n + 1 +2tn+1 F.n (~)
2n
for k=O, 1, . .. ,2n-1;

and define Fn+1 to be linear in the intervals [ 2:+1' k2~+11] for k = 0,
1, ... , 2n +1_ 1. The functions F n are plainly continuous. They are also
strictly increasing. Indeed, for 0 < t < 1 and any oc, Psuch that oc < P,
the inequalities
R
{'- ( -12 t OC + -12+-t{R)
- - ' -_ -12 t (R
- - {,-OC) > °
and
I-t l+t l+t
- 2 - OC +-2- P - oc= -2-(P- OC) >0
hold. These inequalities show that if F n is strictly increasing, then
for k = 0, 1, ... , 2n - 1. The piecewise linearity of Fn+1 proves that

it too is strictly increasing. Also, if oc < P, we have
I-t l+t cx+{3 t
-2-oc+-2-P - -2-= 2 (P-oc»O,
and from this inequality it follows that
(2k+l)
Fn~ <Fn+l ~
(2k+l)
for k = 0, 1, ... , 2n- 1. Hence, again by linearity, the inequality
Fn (x) ~ Fn+ 1 (x)
holds for all x E [0, 1]. Thus the sequence (Fn (X)):=1 converges for all
x E [0, IJ; let
F(x) = lim Fn(x) .
n-->oo
It is clear that F is non decreasing. Actually it is strictly increasing. For,

if x < x' and k and n are such that x < :n < x', then we have
F(x) ~ F Un) = Fn (:n) < Fn(x') ~ F(x').
We next consider any sequence of pairs of numbers (oc n, Pn):=o satis-
fying the following conditions:
OCn ~ OCn+l and Pn+1 ~ Pn (n = 0,1,2, ... ); (1)
(2)
where k n E{O, 1, ... , 2n- I} (n = 0, 1,2, ... ). Thus we have OCn = OCn+1
1 1
and Pn+1 = Pn - 2n +l ' or Pn+1 = Pn and OCn+1 = OCn + 2n +l • In the first
case, we go to the left in proceeding from (oc n, Pn) to (OCn+I' Pn+l);
in the second, we go to the right.
Let k be a fixed nonnegative integer. Suppose that we go to the

right in going from (a", PI!) to (aHI' PHI). Then we have
r -;0+1
F(PHI) - F(aHI) = FHI (PHI) - FHI (aHI)
= F" (P,,) - F" (a,,) + 1 +;k+l F" (P,,)}
= 1 -;0+1 (F(P,,) _ F(a,,)) . (3)
If we go to the left in going from (a", P,,) to (aHI' PHI)' then a like
computation shows that
F(PHI) - F(aH1) = 1 +2t 0+ 1 (F(P,,) - F(a,,)) . (4)
"g e+tt
A simple induction shows that
F(P,,) - F(a,,) = k
) ,
where the B,,'S are lor -1. Thus
IF(P,,} - F (a,,) I ~ /J" (1- 2+ tk)

- ,
and if all t" are less than a number less than 1, F is obviously continuous.
[We will not bother with exploring necessary conditions for the conti-
nuity of F.]
We now look at the derivates of F. Consider any dyadic rational
;" such that 0 ~ ;" < 1. Define a sequence (a,., P,,) satisfying (1) and (2)
for which ap = ap+l = ... = ;" : we care not what al> PI'··· ,ap-I,pp-I
are. Then we must have
I 1
PI>+s = 2P + 2"+1 for s = 0, 1, 2, ....
Applying (4), we see that
F{PH') -F(;,,)
/12 (1 + tp+;) (F(Pp) -
s 1
I = 2P+ 8 F(ap))
PH. - 2P
If the series I; tIS diverges, it follows that D+F ( ;" ) =

,,=1
00 [log (1 + t) ;::;; ~
for 0 ~t~ 1]. Similarly we have D _ F ( ;" ) = 0 if 0 < ;" ~ 1 and
00 __
II (1 -
,,=1
tIS) = O. For these two results it is sufficient that lim tIS be
~oo
positive. Consider next a point x = f; ;: , where x" = 0 or x" = 1, and

"=1
each value is assumed for infinitely many k's, i. e., x is not a dyadic
rational. For each n, there is a unique l such that ;" < x < 1 ~ I ; let
1 1 +1 .
cx" = 2" and {J" = ~. In fact these numbers are gIVen by
W e WI'11 compute F(fJ,,} - F(oc,,}

1 •
If x" =
0, t h en cx" = 1X,,-1 an d
2n
{J" + 2n
1 IX,.-l + fJ"-l
= IX,,_I = 2
If x" = 1, then {J" = {J,,-I, and
+ fJ"-l
IX" = IX,,_I + 2" =
1 IX,.-l
2
In the first case we have
F (fJ,,) - F (oc,,)
~ =
- t" F.
2" { -1 2 ( ) 1 + t" F. ({J ) F. (
- ,,-1 cx,,-I + - 2 - ,,-1 ,,-1 - ,,-1 cx,,-I
)}
2"
= 2" { 1 ~ t" • (F,,-I ((J,,-I) - F,,_ dlX,,_I»} .
In the second case the factor 1 ~ t" is replaced by 1 ~ t" in the preceding
line, and so we obtain
F(fJ,,} - F(oc,,} =
1
2" (1 + (-W"t,,)
2
• (F. ({J ) _ F. (
,,-1 ,,-1 ,,-1 IX,,_I
)
2"
II
='" = lI(l + (-Wkt,,) (5)

"=1
for all n. We know that the function F has a finite derivative a.e., and
hence the limit of the product in (5) exists, is finite, and is equal to F' (x)
,,+1
II (1 + (-I}%kt k )
for almost all x. The ratio k~: is 1 ± t"+l> and hence it con-
II (1 + (- Wkt k )
k~1
verges to 1 if and only if lim tIl = O. Thus if lim t" > 0, the product
~oo ~oa
II" (1 + (_l)Xkt,,) cannot converge to a positive finite number, and so

"=1
F' (x) = 0 for almost all x.
We summarize. Given a sequence (tn):'=1 with values in ]0, 1[, we have
constructed a real-valued function F on [0, 1] having the following
properties:
(i) F(O) = 0, F(l) = 1, F is strictly increasing;
(ii) if lim tn < 1, then F is continuous;

" ..... 00
(iii) if lim tn > 0 and x is a dyadic rational in ]0, 1 [, then D+ F (x) = 00

" ..... 00
and D_F(x) = 0;
(iv) if lim tn > 0, then F' (x) = 0 for almost all x E]0, 1 [ .
....... 00
Thus if 0 < lim tn < 1, then F is continuous, strictly increasing, and

....... 00
F' = 0 a. e. The reader should sketch the first few approximants F n to F

for a special choice of (tn), say tn = ! for all n, to see what is going on.
(18.9) Note. The construction in (18.8) proves also a curious measure-
theoretic fact. If (tn):'=l is a sequence of numbers in ]0, 1 [ not having
limit 0, then
00
lI(1 + (-I)"kt k) = 0
k=l
x
E 2= .
00
for almost all numbers x =
k=l
We now identify the class of functions that are indefinite integrals of
functions in ~l.
(18.10) Definition. Let I be a complex-valued function defined on a
subinterval J of R.l Suppose that for every e > 0, there is a () > 0 such
that
" k) - I(Ck)1 < e
(i) EI/(d
k=l
for every finite, pairwise disjoint, family {JCk, dk[}k=l of open subinter-
vals of J for which
"
E (dk - ck ) < () .
(ii)
k=l
Then I is said to be absolutely continuous on J.
(18.11) Examples. (a) Theorem (12.34) shows that the indefinite
integral of a function in ~l([a, b]) is absolutely continuous. Our next
project will be to prove that every absolutely continuous function is an
indefinite integral.
(b) LEBESGUE'S singular function 'IjJ is not absolutely continuous.
00
We can enclose CANTOR'S ternary set P in a union U ]ak, bk[ of pairwise

k=l 00
disjoint open intervals such that E (b k - ak) is arbitrarily small. Extend

k=l
'IjJ so that 'IjJ(x) = 0 for x < 0 and 'IjJ(x) = 1 for x> 1. Then it is easy to
1 Recall that by (6.1) J can be open, closed, or half-open, and that J can be
bounded or unbounded.
1
z
00 n
see that 1: C"p(b,,) - tp(a,,)) = 1, and so 1: (tp(b,,) - tp(a,,)) ;;:;; for
k=1 k=1
n
sufficiently large n, while 1: (b" - a,,) is arbitrarily small.
k=1
(c) None of the functions F of (18.8) is absolutely continuous. This is
most easily seen from Theorem (18.15) inlra.
We first set down some elementary properties of absolutely continuous
functions.
(18.12) Theorem. Any complex-valued absolutely continuous lunction I
defined on [a, b] has finite variation on [a, bJ.
Proof. Let () > 0 satisfy the conditions of Definition (18.10) for e = 1.
Let n be any integer such that n > b ~ a ,andsubdivide[a,b]bypoints
b-a
a= Xo < Xl < ... < Xn = b such that x" - X"_l = -n- < () for k = 1, 2,
... , n. From our choice of () it follows that V~~.t;;:;; 1 for all k. Thus
n
V:I = 1: V~~.t ;;:;; n. 0
k=1
(18.13) Theorem. Any absolutely continuous lunction I on [a, b]
is continuous, and can be written as
(i) 1= 11 - 12 + i(f3 - 14) ,
where the I; are real, nondecreasing, and absolutely continuous on [a, b].
Proof. If I is absolutely continuous, then the continuity of I and the
absolute continuity of Iml and Rei are obvious. For a real-valued,
absolutely continuous function g, write gl (x) = V; g. Then g = gl - (gl - g),
and the proof will be complete upon showing that gl is absolutely con-
tinuous. [Note that gl and gl - g are nondecreasing (17.16).] For an
n
arbitrary e > 0, let () > 0 be so small that 1: Ig(d,,) - g(c,,) I < ;
k=1
whenever the pairwise disjoint intervals ]c", d,,[ are such that
.
1: (d" - c,,) < () . (1)
k=1
Let {Jc", d,,[}Z=1 be a fixed system of pairwise disjoint intervals satis-
fying (1). Since g has finite variation, there is for each k E{I, 2, ... , n}
a subdivision c" = a~k) < a\k) < ... < al:> = d" such that
Hence we have
.. n n 1~-1
1:lgl(d,,)-gdc,,)I= 1:V~~g< 1: 1: Ig(aWl)-g(aY'»)I+ ~

k=1 k=1 k=1 ;=0
e e
<z+-z-= e;
and so gl is absolutely continuous. 0
(18.14) Theorem. II I is a real-valued, nondecreasing lunction on

[a, b], then f' is Lebesgue measurable and
b
(i) J f'(x) dx ~ I(b) - I(a) .
..
II g is a complex-valued lunction 01 finite variation on [a, b], then
g' E~1 ([a, b]).
Proof. For x > b, let I (x) = I(b). Let
f (x + ~) - f (x)
I.. (x) = 1
n
for n = 1,2, 3, ... and a ~ x ~ b. Then (I..) is a sequence of nonnegative
measurable functions and
lim I.. (x) = f' (x)
........ 00
for almost all x E]a, b[ [see (17.12)]; thus f' is measurable. By FATOU'S
lemma (12.23) and (12.44) we have
J lim I.. (x) dx ~ lim J I.. (x) dx

b b b
J
....
f' (x) dx =
.. ft-io.OO n-+oo
!~~n J [/(x+ ~)-/(x)]dx

b
=
..
~ ~ [ntf (X)4X-n7~(X) dX]
'" ~ [ntf(b) dx - nJ'-f(4) 4%]
= I(b) - I (a) .
This proves our first assertion. The second assertion plainly follows from
the first and the fact that g can be expressed as a linear combination of
four nondecreasing functions. 0
(18.15) Theorem. Let I be an absolutely continuous complex-valued
lunction on [a, b] and suppose that f'(x) = 0 a.e. in ]a, be. Then I is a
constant.
Proof. We lose no generality by supposing that I is real-valued, for
otherwise we examine Rei and Iml separately. We will show that
I(c) = I(a) for all c E]a, b]. Thus let c E]a, b] and e > 0 be arbitrary.
Select a number ~ > 0 corresponding to the given e for which the condi-
tion in the definition of absolute continuity (18.10) is satisfied. Let
E = {x E ]a, c[: f' (x) = o}. Clearly A(E) = c - a. For each x EE there
exist arbitrarily small h > 0 such that [x, x + h] C ]a, c[ and
ell
I/(x+ h) - l(x)1 < c _ a . (1)
The family of all such intervals [x, x + hJ is a Vitali cover of E, and so by
VITALI'S theorem (17.11) there exists a finite pairwise disjoint family
{[Xk' Xk + hkJ}i:~1 of these intervals such that
A(E n (k~l [Xk,Xk+ hk]r) < <5.

Then
n
A(]a, cD = A(E) < 15 + I: hk · (2)
k~l
We may [and do] suppose that Xl < X 2 < ... < x n . It follows from (2)
that the sum of the lengths of the open intervals
]a, xI [, ]XI + hI' x2 [, ••• , ]Xn + hn' c[
n
complementary to k~l U [Xk' Xk + hk] is less than 15, and so, in view of our
choice of <5, we have
n-l
I/(a) - I (Xl) I + I: I/(Xk + hk) - I (Xk+l) I + I/(xn + hn) - l(c)1 < e. (3)
k~l
The inequalities (1) and (3) combine to yield

n-l
I/(a) - l(c)1 ~ I/(a) - I (Xl) I + I: I/(Xk + hk) - I (x k+1) I
k~l
n
+ I/(xn + hn) - l(c)1 + I: II ex,. + hk) - I (xk)l
k=l
Since e is arbitrary, it follows that I (c) = I(a). D

(18.16) Theorem [Fundamental theorem of the integral calculus for
Lebesgue integrals]. Let I be a complex-valued, absolutely continuous
lunction on [a, b]. Then f' E~l ([a, bJ) and
x
(i) I (x) = I (a) + f f' (t) dt
a
lor every x E [a, b].
Proof. From (18.12) and (18.14) it follows that t' E~l' Let g(x)
x
= f t' (t) dt. Then g is absolutely continuous and, by (18.3), g' (x) = t' (x)
a
a. e. Thus the function h = I - g is absolutely continuous and h' (x)
= t' (x) - g' (x) = 0 a. e. It follows from (18.15) that h is a constant.
286 Chapter V; Differentiation
Therefore
x x
I (x) = hex) + g(x) = heal + J f' (t) dt = I (a) + J f' (t) dt
a a
for all x E [a, b]. 0
(18.17) Theorem. A lunction I on [a, b] has the lorm
x
I(x) = I (a) + J rp(t) dt
a
lor some fjJ E5~\([a, b]) il and only il I is absolutely continuOZls on [a, b].
In this case we have fjJ (x) = f' (x) a. e. on ]a, b[.
Proof. This is just a summary of (18.3), (18.11.a), and (18.16). 0
Indefinite integrals on R can be characterized in much the same way.
(18.18) Theorem. A lunction I on R has the lorm
x
(i) I(x) = J rp(t) dt
-00
lor some fjJ E~1 (R) il and only il I is absolutely continuous on [- A, A] lor
all A > 0, V~ool is finite, and lim I(x) = 0.
x-+-oo
Proof. Suppose that I has the form (i). Then I is absolutely con-
tinuous on [-A,A] by (12.34), and by (18.1) we have
V~ool = J \fjJ(t) \ dt < 00.
R
The dominated convergence theorem (12.30) implies that lim I (x) = 0.

x-+--oo
Conversely, if I is absolutely continuous on [-A, A] for all A> 0,
(18.17) shows that
x
I(x) = I(-A) + J f' (t) dt
-A
for all positive real A and x such that x > - A. Taking the limit as
A -+ 00, we have
x
lim J f' (t) dt.
I(x) = (1)
A-->oo -A
Applying (12.22), (18.17), and (18.1), we have
00 n
J 1f'(t)ldt= lim J 1f'(t)ldt
-00 n~oo -n
lim V!:nl ;2; V~ool < 00 •

=
n-->oo
Thus f' is in ~1 (R), and the dominated convergence theorem applied to
..
the right side of (1) shows that I (x) = J f'(t) dt. 0
-00
The formula for integration by parts holds for absolutely continuous

functions and Lebesgue integrals.
(18.19) Theorem. Let I, g be lunctions in ~l([a, b]), let
F(x) = rx + J" I(t) dt,

a
and let
G(x) = {J + J"g(t) dt.
a
Then
b b
(i) J G(t) I(t) dt + J g(t) F(t) dt = F(b) G(b) - F(a) G(a) .
a a
Proof. The inequality

IF(v)G(v) -F(u)G(u)/ ~ IIFlluIG(v) - G(u)/ + IIGllulF(v) -F(u)/
shows that FG is absolutely continuous. Hence FG is differentiable a.e.,
and
(FG)' = FG' + F'G ,
as an elementary calculation shows. By (18.3) we have G' = g a.e. and
F' = I a.e.; thus (i) follows from (18.16). D
(18.20) Corollary. Let I and g be absolutely continuous lunctions on
[a, b]. Then
b b
(i) J I(t) g' (t) dt + J t' (t) g(t) dt = I(b) g(b) - I (a) g(a) .
a a
Proof. This is just (18.19) rewritten with the aid of (18.17). D
(18.21) Corollary. Let I and g be lunctions on R satislying the condi-
tions 01 (18.18). Then
(i) J I(t) g' (t) dt + J t' (t) g(t) dt = J t' (t) dt· J g' (t) dt
R R R R
= lim I (x) . lim g(x) .
x~oo ~oo
Proof. Take limits as a ~ - 00 and b ~ 00 in (18.20.i). D

(18.22) Note. Another famous integral formula is the elementary
formula for integration by substitution:
(J b
(i) J l(y) dy = J 1(f{J(x) f{J'(x) dx,
'" a
where I is Riemann integrable and f{J is a function with positive contin-
uous derivative on [f{J-l(rx), f{J-l(,B)] = [a, b]. A much more general
formula is in fact true, which subsumes (i) as a very special case. The
proof seems most easily carried out by using the LEBESGUE-RADON-
NIKODYM theorem (19.24), and we postpone it to (20.4). We continue
here with some technical facts about absolutely continuous functions
needed in (20.4) and (20.5).
(18.23) Theorem. Let cp be a complex-valued, absolutely continuous

function on [a, bJ. For [x, y] C [a, b], let
O1<p(x,y) = sup{lcp(u) - cp(v)l: u, v E [x,yJ}.
Then for every [', > 0, there is a 15 > 0 such that
n
(i) l: O1<p (Ck' dk ) < [',
k~1
for every finite, pairwise disjoint, family {Jc k, dk[}k~1 of open subintervals
of [a, b] for which
n
(ii) E (d k - ck ) < 15 .1
k~1
Proof. The function cp being continuous and the interval [c k , dk ] being

compact, it is easy to see that [c k, dk] contains points Uk and Vk such that
Uk < V k and Icp(u,,) - cp(v,,)1 = O1<p(c", d,,). Since
n n
E (v" - Uk) ~ l: (d k - Ck) < 15 ,
"~I k~1
we obtain (i) at once from (18.1O.i). 0

For Theorem (20.4), we will need another definition.
(18.24) Definition. Let g be a function with domain [a, b] C Rand
range [IX, fJ] cR. If A(E) = 0 implies A(g(E)) = 0 for all E C [a, b],
then g is said to be an N-function or to satisfy the condition N.2
(18.25) Theorem. Let cp be a continuous function of finite variation
with domain [a, b] C R and range [IX, fJ] CR. Then cp is an N-function if
and only if cp is absolutely continuous. 3
Proof. Suppose that cp is absolutely continuous and that A(E) = O.
Let [', be an arbitrary positive number and let 15 be as in (18.23). Since
).,(cp({a, b})) is trivially zero, we may suppose that E C ]a, be. We choose
a family {JCk' dk[}k~1 of pairwise disjoint open subintervals of ]a, b[ such
00
that E C U ]ek' dk[ and

k~1
l: (d k -
k~1
ek) < 15. By (18.23), we have
n
l: O1<p (ek' dk) < [', for all n
k~1
and so
00
l: O1<p(c k, dk) ~ [', . ( 1)

k~1
1 Thus we can replace the condition (18.10.i) by the apparently stronger condi-
tion (18.23.i) in the definition of absolute continuity.

2 This terminology, and the concept itself, are due to N. N. LUZIN (1915);
he thought of the property as a "null condition".

3 This theorem is due to BANACH [Fund. Math. 7, 225-236 (1925)]; we give
his original proof.
Plainly we have
epeE) c ep (kgl Jc", d,,[) kgl ep(Jc", d,,[) .
= (2)
It is also evident that
A(ep(JC", d,,[)) = A(ep([C", d"J)) = w",(c", d,,) .

Hence (2) and (1) imply that A( ep (E)) ;;;: 8. Since 8 is arbitrary, A( ep (E)) = 0
and ep is an N-function.
The converse is less obvious. Suppose that ep is an N-function and
assume that ep is not absolutely continuous. By (18.23), there is a positive
number 8 0 such that we can find a sequence
{ J C(l)
1 ,
del)
1
[, ••• ,
JC(l)
11 ,
d(1)
11
[} = f!) l'
{ J C(2) d(2) [ JC(2) d(2) [} = 17),

1 , 1 , •.. , Is' I" :;zJ2J
{ JC(n)
II
d(n)[
I
J
(n) d(n)[} _ 17),
, . . . ,GIn' In -;;zJnJ
with the following properties. First, the intervals comprising each f!)n
are pairwise disjoint. Second, the inequalities
In
'"' W (c(n) den») > ~ (3)
£..oJ '" k ' k = "0
k=1
hold for all n. Third,
(4)
n=1 k=l
For each n, and for y E [IX, fJJ, let Nn(Y) be the number of intervals
Mnl, d~n)[ in f!)n that have nonvoid intersection with ep-l({y}). As the
intervals Jc~n), 4n )[ are pairwise disjoint, it is evident that
(5)
here v is the Banach indicatrix of ep, defined in (17.34). It is obvious that
In
Nn = 1: ;"'(]C~n), dkn )[).
k=1
Since ep ([4n), din)J) is a closed interval [whose measure is w'" (c~n), 4n »)]
and since (ep (Jc\:'\ d~n) [))' n ep ([cin ), 4n )J) contains at most two points,
N n is Borel measurable [actually the pointwise limit of a sequence of
continuous functions] and
(6)
Let A be the set {y : y E[IX, fJ], lim N n (y)

o}. LetA l = {y EA :v(y)= oo}.
=1=
~oo
Since ep has finite variation by hypothesis, v is in ~1 ([IX, fJ]) and so

290 Chapter V. Differentiation
l(AI) = o. Consider any point Yo EA n A;. There exists a sequence

(Xnj )j':l (nl < n 2 < ... ) of points in [a, b] such that
Inj
x nj E k'::!l Mnj), 4 nj )[ and ep(xnj ) = Yo

(j = 1,2, ... ). Since Yo ~ AI> only a finite number of the points x nj are
distinct, and so there is an Xo E [a, b] in infinitely many of the sets
z,.
U ]c(n)
k~l k' k
din) [ such that m(x )
=
rOO
y . Write
E = i Ogi t~l
l Mn), dkn )[) •
It is obvious that Xo EE. From (4) and (10.15) we infer that l (E) = o.
Since ep is an N-function, we have l(ep(E) = O. We have proved above
that ep(E) :J A n A;, so that
l(A) = l(A n A;) = O. (7)
The definition of A and (7) show that lim N n (y) = 0 for almost all y in
1<->00
[ae, [3J.AsNn ~ vandv E~d[ae, [3]),weinferfromLEBEsGuE'sdominated

convergence theorem (12.24) that
lim
1<->00 R
J Nn(y)dy=O. (8)
Since (6) and (8) contradict each other, the proof is complete. 0
The hypothesis in (18.25) that ep have finite variation is essential,
as the following example shows.
(18.26) Example. Consider any interval [a, b] and a perfect nowhere
dense subset F of [a, b] that contains both a and b. The measure l(F)
may be zero or positive. Write the open set [a, b] n F' as U Jan, bn [, where
00
n~l
the intervals Jan, bn [ are pairwise disjoint and are enumerated in an

arbitrary order. Let Cn = ~ (an + bn) and let (tn):~l be a sequence of
positive numbers with limit zero. Define a function g on [a, b] as follows:
g(X) = 0 for all x EF;
g(cn ) = tn (n = 1,2, ... ) ;
g is linear in [an, cn] and in [cn' bn]
(n = 1,2, ... ) .
It is easy to see that g is continuous. Also it is easy to see that
V; g = 2 1: tk •
00
k~l
We leave both proofs to the reader. To see that g is an
N-function, consider any set E C [a, b] such that l(E) = O. Using
(2.15.i), we have
00
gee) = g(E n F) U "'::!1 geE n ]a", b,,[) .

Since g is linear on [ak, Ck] and on [Ck' bk], it is plain that
and so
+ E A(g(E n ]ak' bk[)) =
00
A(g(E)) ~ A({O}) O.
k=1
E tk =
00
If 00, then g certainly fails to be absolutely continuous, since it
k=1
has infinite variation.
(18.27) Discussion. We close this section by giving a famous applica-
tion of (18.5) [which in fact led LEBESGUE to the definition (18.6) of the
Lebesgue set]. Consider a function I E ~1([-:n;, :n;]) and its Fourier co-
efficients I(n) (16.33). The uniqueness theorem (16.34) tells us that I is
determined [as an element of ~l([-:n;, :n;]), of course] by the function I
defined on Z. This theorem leaves untouched the problem of reconstruct-
ing I from f. This problem is important not only for its own sake but also
for applications to physics, chemistry, and engineering, since many data
obtained by spectroscopy, X-ray analysis, and the like, are nothing other
than Fourier coefficients of functions which one wishes to determine.
The simplest way to try to recapture I from I is by means of the Fourier
series 01 I, the partial sums of which are defined as
n
(i) snl(x) = E I(k) exp(ikx) (n = 0, 1,2, ... ) .
k=-n
In order to rewrite (i) and other expressions to be defined shortly, we
define I on the entire line R by periodicity: I(x + 2k:n;) = I(x), for
x E [-:n;,:n;[ and k EZ [the number I (:n;) has no importance for
I E~l([-:n;, :n;])].
We then have:
n
(ii) s,d(x) = i;
k=-n
2~ f I(t) exp(-ikt) dt· exp(ikx)
-n
= 21:n; j I(t) L.J:n exp(ik(x - t))] dt

-n
= 21:n; j I (x - t) L.J:n exp (ikt)] dt.

-n
[The reader should check the last equality in (ii).] It is elementary to
show that
sin((n+{)t) 1.f
n --'-'-"-;-1--:-'--'-- exp (i t) =1= 1 ,
(iii) E exp(ikt) = sin(2"t)
k=-n
2n+ 1 if exp(it) = 1.
19·
The function defined by (iii) is called the Dirichlet kernel and is denoted
by Dn(t). Thus we may write
n
(iv) snl (x) = 21n f I (x - t) Dn (t) dt .

-n
For many functions, the sequence snl does in fact converge to 1.1 For
others it does not. To reconstruct I from I, we follow FEJER 2 in taking
the arithmetic means of the partial sums (i). Accordingly we define
1
(v) Gnl(x) = ~ [sol (x) + sl/(x) + ... + snl(x)]
= i (1 - I~ 1) 1(k) exp (ikx) .
k=-n
n
Using (iv), we write
nf I(x-t) [n ~ 1 (D o(t)+D (t)+"'+Dn(t))] dt.

"
(vi) Gnl(x) = 21 1
-"
The expression [... J in (vi) is called the Fejer kernel; it is denoted by
Kn (t); and one easily proves that
1 [sin(f(n+l)t)]2 1
1
if sinh-t) =l= 0 ,
(vii) Kn(t)= n+l sin(ft)
n +1 if sin (ft) = O.
The reader can easily verify the following:
(viii) Kn (- t) = Kn (t) ;
(ix) 0 ~ Kn (t) ~ n + 1;
n
(x) 21n
-n
f Kn(t) dt = 1;
since sin (8) > ! 8 for 0 < 8 < ~ ,

n2
(xi) Kn (t) ~ (n + 1) t2 for 0 < It I ~ n .
It follows trivially from (xi) that
n
(xii) lim f Kn (t) dt = 0 for b EJO, n[ .
n-->-OO "
Our first inversion theorem is elementary.

1 For all details of this fact, and indeed the whole theory of trigonometric
series, the best guide is undoubtedly the classical work of ZYGMUND, Trigonometric
Series [2 vots., Cambridge University Press, 1959].
2 LEOPOLD FEJER (1880-1959) was a distinguished Hungarian mathematician.
(18.28) Theorem. Let p be a real number such that 1 ;:;;;; p < 00, and
let I be a lunction in ~p([-~, ~]). Then
(i) lim III - O"nilip = 0.
n->-oo
Proof. We choose an auxiliary function g, which for p> 1 is an

arbitrary function in ~p,([-~, ~]) such that IIgllp';:;;;; 1, and which for
p = 1 is the function identically 1. We then write
n
21n J (f(x) - O"nl(x)) g(x) dx

-n
n n n
1
= 21n J (/(x) 2n J Kn(t) dt - 21n J I(x-t) Kn(t) dt) g(x) dx
-n -n -n
n n
;:;;;; 4~1 J J II (x) - I(x - t)IIg(x)l Kn(t) dt dx. 1 (1)

-.11 -:11
We anticipate FUBINI'S theorem (21.13) [which of course is proved

without recourse to the present theorem] to reverse the order of inte-
gration in the last expression of (1). This produces
n n
4~1 J JI/(x)-/(x-t)IIg(x)ldxKn(t)dt. (2)

-,11: -:rr
Now use HOLDER'S inequality (13.4.ii) on the inner integral in (2):

n
21n J I/(x) - I(x - t)i Ig (x) I dx;:;;;; III - I-tllp' IIgll p'
-n
(3)
Going back to (1), we therefore have
;:; ; "' - '-tllp·
n n
21n J (f(x) - O"nl(x)) g(x) dx ;:;;;; 21n J Kn(t) III - I_tllpdt

-n -n
=_1
2n
J+- J 1
2n
Itl;:;;~ Itl>"
;:;;;; sup{II1 - I-tllp: It I ;:; ; 15} 21n J Kn(t) dt
Itl;:;;;"
+ 2","p 2~ J Kn(t) dt. (4)
Itl>"
By (13.24), the supremum in (4) is arbitrarily small if 15 is sufficiently
small. By (18.27.xii), the limit of the last expression as n -+ 00 is zero, no
1 We will prove in § 21 that this iterated integral is well defined.
matter how small 6 may be. That is, the first expression in (4) is arbitrar-
ily small for n sufficiently large. This implies by (15.1) that
lim III - anillp = 0 (p > 1) .

........ 00
For p = 1, use (3) and repeat the argument with obvious changes. 0 1
Our point in going through (18.28) was to lead up to the much
subtler fact that ani converges to I not merely "in the mean" [i.e., in the
~p norm] but also pointwise almost everywhere.
(18.29) Theorem [LEBESGUE]. Let I be a lunction in ~1([-n, nJ).
Then il x is in the Lebesgue set 011, we have
(i) lim ani (x) = I (x) .
n->-oo
Proof. For brevity we write I(x + t) + I(x - t) - 2/(x) as q;(x, t).

We also define
,
~(x, t) = J Iq;(x, u)1 du
o
+~(x,
and we write the number ~ (x, n) as a. Theorem (18.5) shows that
t) ~ 0 as t ~ 0 [x is in the Lebesgue set of II]. Consider any e > 0
and choose IX>O so that j+~(X, t)j < e for ItI ~ IX. Next use (18.27.xi)
to choose an integer no > ~ so that
n
ex
~ no and IX ~ t ~ n imply IKn(t) I < a: I . (1)
Note that
n ~ no implies n~(x, !) < e. (2)
n
ani (x) -I (x) = 21n; f q; (x, t) Kn (t) dt

o
and hence
n
2n lanl(x) -/(x)l ~ f Iq;(x, t)l Kn(t) dt
o
l/n "
= f 1q;(x,t)1 Kn(t) dt+ f 1q;(x,t)1 Kn(t) dt
o l/n
n
+ f Iq;(x, t)1 Kn(t) dt = 51 + 52 + 53' (3)
"
1 For p > I, we even have lit - s"fllp ~ O. This is much harder to prove, and
is typical of the intriguing and delicate results obtained in the theory of Fourier
series. See ZYGMUND, loco cit., Chapter VII, Theorem (6.4).
Now suppose that n ~ no' For 53' (I) implies that

n
53~jlcp(x,t)1 a~1 dt~ a~1 a<e. (4)

"
By (18.27.ix) and (2), we have
lin
51 ~ j Icp(x, t)1 (n + I)dt = (n + I)q> (!) ~ 2nq> (!) < 2e. (5)
o
Using (18.27.xi), we have
" "
52 = j Icp(x,t)1 Kn(t) dt~ j Icp(x, t)1 n: I fdt.
lin lin
We now apply (18.19), with

g(t) =-2r3 andG(t) =n2 + Jg(u)du=t- 2
lin
This yields
" "
jlcp(x,t)lt- 2 dt= q>(X,ex)ex- 2 - q>(x, !)n 2 +2jq>(x,t)r3dt.
~ ~
Hence
:n;2
52~ n+1 q>(x,ex)~
I
+
2:n;2
n+1
j "q>(x,t)r dt 3
lin
e 2:n;2 j "
:n;2
:s;: - - -
-n+lex
+-
n+1
- q>(X' t) r 3dt
lin
<n2 e+54 •
Finally we have
"
5 :s;: ~j er2 dt = 2:n;2e [n-~] < 2:n;2e • n < 2n2e
4 - n+1 n+1 ex n+1
lin
and hence
(6)
The relations (3), (4), (5), and (6) show that n ~ no implies
2nlQ'nl(x) - l(x)1 < 3e + 3n2 e < 33e. 0
(18.30) Exercise. Let E be a subset of R [not a priori measurable]
such that for some positive real number b, the inequality .A. (E n I) ~ b.A. (I)
holds for all intervals Ie R. Prove that E' is a set of measure zero.
[Use (18.2).]
(18.31) Exercise. Find a real-valued, absolutely continuous function

on [0, IJ that is monotone on no interval. [Construct a measurable set
A C [0, IJ such that A(A n J) > 0 and A(A' n J) > 0 for every interval
J C [0, 1]. This can be done by using Cantor-like sets (6.62). Now inte-
grate ~A - ~A"]
(18.32) Exercise. Consider a positive real number IX, and define a
function I,. on [0, IJ by
1,.(0) = 0,
I,.(x) = XIX cos (X-l) for 0 < x;;;;; 1.
[Define XIX as exp(1X log (x) with log(x) real.J
(a) For what IX'S does I,. have finite variation?
(b) For what IX'S is I,. absolutely continuous?
(c) For what IX'S does I,. have a finite derivative in JO, 1[ and a finite
right derivative at O?
(18.33) Exercise. (a) Let I be a continuous complex-valued function
of finite variation on [a, bJ and suppose that I is absolutely continuous
on [a, cJ for all c such that a < c < b. Prove that I is absolutely contin-
uous on [a, bJ.
(b) Show that part (a) fails for some continuous functions on [a, bJ
having infinite variation on [a, b].
(18.34) Exercise. Prove the following. A complex-valued function
on [a, bJ is in ~ipl([a, bJ) [see the definition in (17.31)J if and only if for
every e > 0 there is a ~ > 0 such that for all sequences ([a", b"m=1 of
subintervals of [a, bJ for which
holds, the inequality

.
E I/(b,,) - I (a,,) I < e
"=1
holds. That is, I is "absolutely continuous with overlap permitted".
(18.35) Exercise. Let t be a nondecreasing function in (tr([a, bJ) and
let E = {x : a ;;;;; x < b, D+ t (x) = co}. Prove that t is absolutely continuous
if and only if A(f(E» = O. [Use (18.25) and (17.25).J
(18.36) Exercise. Let I be a complex-valued function on [a, b].
Prove that I is absolutely continuous on [a, bJ if and only if there exists
a sequence (1.. )':'=1 of functions in ~ipl([a, b]) such that V;(f - I..) ~ O.
"
[If / is absolutely continuous, let /.. (x) =
,.J g.. (t) dt, where gn (t) = f' (t) if
If' (t)J ;;;;; nand gn (t) = 0 otherwise.]
(18.37) Exercise. Let g be absolutely continuous on [a, bJ, let

[01:, PJ = rngg, and let t be absolutely continuous on [01:, PJ.
(a) Prove that tog is absolutely continuous on [a, bJ if and only if
tog is of finite variation on [a, bJ. [Use (18.25).J
(b) Give an example to show that tog need not be absolutely contin-
uous on [a, bJ.
(18.38) Exercise. Let t be a complex-valued function defined on a
closed interval [01:, PJ. Prove that t E~ipl ([01:, PJ) if and only if log is
absolutely continuous on [a, bJ for all closed intervals [a, bJ and abso-
lutely continuous functions g with domg = [a, bJ and rngg C [01:, PJ.
(18.39) Exercise. (a) Find a strictly increasing function IE <rr([o, IJ)
for which there is a subset A of [0, IJ such thatA(A) = and A(f(A)= 1. °
[Use (18.8).J
(b) Show that a function in <rr([a, bJ) maps every measurable set
onto a measurable set if and only if I is an N-function. [Hints. If t is an
N-function and A is a measurable set, write A as a a-compact set plus a
set of measure zero. If t is not an N-function, apply (10.28).J
(18.40) Exercise. By a particular choice of the numbers tn in (18.26),
functions with quite extraordinary properties can be found.
(a) [RUZIEWICZ-SAKSJ Let F be as in (18.26). The structure of F
makes it clear that the family of intervals {[al> blJ, ... , [an, bnJ} are
pairwise disjoint for all n: let en be the maximum of the lengths of the
open subintervals of [a, bJ that form the set [a, bJ n [a k, bkJ)'. It is t91
obvious that lim (b n - an) = 0. Since F is nowhere dense, a simple
en =
n~oo
argument proves that lim 0. In the definition of gin (18.26), let

n~oo
tk = ~ (b k - ak) + ek'
Then g has a finite derivative at no point of F.
Thus if A(F) > 0, we have an N-function that fails on a set of positive
measure to have a derivative. [Hints. Show first that if g is differentiable
at x EF, then g' (x) = 0. Show next that a derivative of g is nonzero
if x = ak or x = bk • For other points x EF and for each index n, there is
an interval Jaj(n), bj(n) [ among the intervals Jal> blL ... , Jan, bn[ at mini-
mum distance from x [there are obviously at most two such intervalsJ.
A moment's thought shows that lim j (n) = 00, that
n.-,.oo
and that lim Cj(n) = x. Hence we have

n ..... oo
which shows that either D+ g (x) ~ 1 or D _g (x) ~ - 1. Thus g is differen-

tiable nowhere on F.]
(b) Suppose that A(F) > O. Prove that kf00 (h~= 00.
[00
If ~ (h~ were
finite, then g would have finite variation and would be differentiable al-
most everywhere.]
(c) Consider CANTOR'S ternary set Pc [0, 1], with the notation of
(6.62). Write its complementary intervals in the order
1 1,1> 12,1> 12,2' 1 3 ,1' 1 3,2' 1 3,3' 1 3", •• • ,1",1' .. . ,1",2....1 , ••••
00
Compute the sum I: e", where e" is defined as in (a). Why does this not
k=1
conflict with (b) ?
(18.41) Exercise. This exercise is somewhat demanding, but its
part (d) is so elegant a result that we hope all readers will work through
it. The closed interval [a, b] is fixed throughout.
(a) Let A be a subset of [a, b] of measure O. There is an absolutely
continuous non decreasing function 1jJ on [a, b] such that 1jJ' (x) = 00 for
all x EA. [Hints. Let (U,,):'=1 be a decreasing sequence of open supersets
..
of A such that A(Un) < 2-". Let ({J" = I: ~Uk' and ({J = lim ({J". Verify
k=1 ....... 00
b %
that ({J E~i([a, b]) and that J ((J(t) dt < 1. Let 1jJ(x) = J ((J(t) dt, and
a a
%
1jJ,,(x) = J ((J,,(t) dt. If x EA and [x, x + h] C Un' then

a
,,* + h) -",,(x) >~f

h = h
%+h
({J"
(t) dt =
n.
"
Thus 1jJ~ (x) = 00; similarly for 1jJ'- (x) .]
(b) [A. ZYGMUND]. For I E(t"([a, b]), define
E = {x: a ~ x < b, D+/(x) ~ O}.
Suppose that t (E) contains no interval. Then I is nondecreasing. [Hints.
If I(c) > I(d), where a ~ c < d ~ b, choose any Yo E]t(d), I(c)[ and
define Xo as sup {x : c ~ x < d, I (x) ~ Yo}. Show that t (xo) = Yo, and hence
that D+/(xo) ~ O. This implies that I(E) ::::> ]/(d), l(c)L a contradiction
to our hypothesis.]
(c) Consider any IE (t"([a, b]). Suppose that D+I is nonnegative
almost everywhere on [a, b[. Suppose also that the set B, defined by
B = {x: x E[a, bL D+/(x) = - oo}, is countable. Then I is nondecreasing.
[Hints. Let A = {x: x E]a, b[, D+/(x) or D+/(x) is negative and finite}.
Let1jJbe as in part (a) for the set A, let a (x) = 1jJ(x) +x, andletg= 1+ ea,
where e is a positive number. Show that D+g(x) = 00 for x EA and that

D+ g (x) is positive for x EA' n B'. Hence the set E = {x : D+ g (x) ~ o}
is contained in the countable set Band g (E) can accordingly contain no
interval. By part (b), g is nondecreasing. As e is arbitrary, I too must be
nondecreasing. ]
(d) [Main result]. Let I be a function in <r([a, bJ). Suppose that I' (x)
exists and is finite for all but a countable set of x in Ja, b[ and that
I' E~l ([a, bJ). Then
"
l(x)-/(a)=JI'(t)dt for a~x~b,
a
and in particular I is absolutely continuous. [A sketch of the proof
follows. Consider IE <rr([a, bJ), the complex case being a trivial extension
of this. For each positive integer n, define
gn = max{l', -n} and In(x) = J gn(t) dt.
"
a
Apply (12.24) to prove that
lim In(x) = J" I' (t) dt.
"-+00 a
Next show that
D+(/n - I)(x) = I~(x) -I'(x) = gn(x) -I'(x) ~ °
for almost all x EJa, b[. Since
"+,,
I,,(x+ hl - I.. (xl :?:-.!..J(-n) dt =-n
h - h '
"
one sees that D+ (In - I) (x) is greater than - 00 except on the countable
set where I' is nonfinite or does not exist. By part (c), In - I is non-
decreasing. Thus
In (x) - I (x) ~ In(a) - I (a) = -/(a) ,
and so
J"I'(t) dt = lim In(x) ~ I(x) - I (a) .
a. ,,-+00
Replace I by - I to reverse the last inequality.J

(18.42) Exercise. Let I be defined on [0, IJ by I(x) = xlsin (X-I) for
° < x ~ 1 and 1(0) = 0. Prove that I has a finite derivative at all points
of JO, 1[ and I~ (0) = 0, but I' ~ ~l ([0, IJ). Is I absolutely continuous on
[0, IJ? Is I of finite variation on [0, IJ? [Compare this with (18.4l.d).
This example raises the following question. If I is a continuous function
on [a, bJ such that I' is finite everywhere on Ja, b[ but is not in ~l ([a, bJ).
then how can I be reconstructed from I'? This problem is solved by
making use of the Denjoy integral, which was invented for just this pur-
pose. The reader who wishes to learn the details about this integral and
other integrals more general than that of LEBESGUE is referred to S. SAKS,

Theory of the Integral, 2 nd Ed., Monografie Matematyczne, Warszawa-
I.wow, 1937].
(18.43) Exercise: Integral representation of convex functions. Prove
the following. Let I be an open interval in R and I a real-valued function
on I. The function I is convex if and only if there are a nondecreasing
function q; on I and a point c EI such that
I (x) = j
/(C) +
I(c)-fq;(t)dt
j q;(t) dt
••
for
for
x
x<c.
~ c,
"
[The "if" part is a tiny extension of (13.35). For the "only if" part,
combine (17.37.b), (18.16), and (17.37.a).]
(18.44) Exercise. Let I be a complex-valued function of period 2:rt
defined on R such that I E~1 ([ -:rt, :rtJ). Suppose that I is continuous at
every point of [a, b]. Prove that (an l)':=1 converges to I uniformly on
[a, b]. [Use uniform continuity and the Fejer kernel as in (18.29).]
(18.45) Exercise. Use the uniform boundedness principle to prove
that there exists a real-valued continuous function of period 2:rt defined
on R whose Fourier series diverges at O. [Let
'-13 = {t E<rr([-:rt, :rt]): I(-:rt) = I(:rt)}.
Then '-13, with the uniform norm, is a real Banach space. For each n EN,
n
define T..: '-13 ~ R by T..I = snl(O) = 2~

-n
JI(t) Dn(t) dt, where the nota-
n
tion is as in (18.27). Prove that T..E'-13* and that IIT..II= 21n JID.,(t) Idt
-n
for each n. Next show that lim IIT..II = 00. Finally conclude from (14.23)
...... 00
that there exists an I E'-13 for which lim ISnl(O)i = 00.]

....... 00
(18.46) Exercise. (a) Let X be a complex-valued Lebesgue measurable

function defined on R such that:
(i) X is bounded;
(ii) X is not identically zero;
and
(iii) X(x+Y)=X(x)X(Y) for all x,yER.
Prove that there exists a real number IX such that X(x) = exp(ilXx)
for all x ER. [Hints. Use (ii) and (iii) to prove that X vanishes nowhere.
Noting (i), define I on R by the rule I (x) = f"X(t) d t. Choose a ER such
o
that I (a) =t= 0, prove that X(x) I (a) = I (x + a) - I (x) for all x ER, and
conclude that X is absolutely continuous. These facts imply that I. and

therefore X. has a continuous derivative on R. Differentiate (iii) with
respect to y and then set y = 0 to obtain X' (x) = X (x) X' (0). Next set
X' (0) = i IX, where IX EK.
Verify that :x
[X (x) exp (- ilXx)] = 0 for all
x ER. Conclude that there exists a {J EK such that X(x) = {J exp (iiXX)
and show that X (0) = 1 so that (J = 1. Finally, use (i) to prove that IX ER.
Functions X satisfying (i) - (iii) are called characters 01 R.]
(b) Let "P be a real-valued Lebesgue measurable function on R that
satisfies (ii) and (iii) above [naturally with X replaced by "PJ. Prove that
"P(x)=exp({Jx) for a real number (J. [It is elementary to show that
"P(x) > 0 for all x. Use (10.43) to show that "P is continuous at 0 and
use (iii) to show that "P is continuous everywhere. Now integrate as in
part (a).]
(c) Let w be a complex-valued Lebesgue measurable function on R
that satisfies (ii) and (iii). Prove that w (x) = exp (yx) for some y EK.
[Use part (a) on w Iwl-1 and part (b) on Iwl.]
(d) Construct examples to show parts (a) and (b) fail if the hypothesis
of measurability is dropped. [Use a Hamel basis for Rover Q as in (5.46).
and note that a discontinuous X as in part (a) cannot be Lebesgue meas-
urable; similarly for "P's as in part (b).]
(18.47) Abel summability of Fourier series. Theorem (18.29) has
an analogue for another classical summability method, and in fact the
exceptional set for this method may be much smaller than the complement
of the Lebesgue set. We sketch the construction and proof, leaving many
details to the reader as exercises. All notation not explained here is as in
(18.27). For I E~l([-n. n]) and for 0 < r < 1, let
(i) IXrl (x) = £
k=-oo
rlkll (k) exp (ikx) .
[Compare this with the definition of unl in (18.27.v).] The function IXrl
1:
00
is called the rtl• Abel sum 01 the series I(k) exp (ikx). Using the uniform
k=-oo
convergence of the series in (i). show that
n
(ii) IXrf(X) = 2~
-n
f I(x - t) L=~oo rlkl eXP(ikt)] dt.
The expression [... ] in (ii) is called the Poisson kernel, and is denoted
per, t). A simple computation shows that
''') P()
(11l r, t = 1 + 1'21_- 1'2
2rcos(t) .
Also easy to verify are the relations

(iv) Per. t) = per, -t) ,
(vi) 2~ f P(r, t) dt = 1,
-n
.., (1 - r2) 2r sin(t)
(vn) P (r, t) = - (1 + r2 _ 2r cos (t))2
[differentiation with respect to t].

x
Consider the functionF (x) = 21;n; f ! (t) dt, and consider any xE]-n, n[
-n
at which the symmetric derivative
lim F(x+h) -F(x-h) = D F(x)
hj-O 2h 1
exists and is finite [see (17.36)]. We wish to prove that

(viii) lim IXr!(X) = D1F(x) .
,t 1
In view of (17.36.a) and (18.3), this will prove that lim IXr! (x) = ! (x) a. e.
Ttl
on [- n, n ] [note that the set where this occurs contains, perhaps
properly, the Lebesgue set for! (18.5)]. By adding a constant to! [which
disturbs nothing] we may suppose that F(n) = o. Applying (18.19) to
(ii), we see that
n
(ix) IX,! (x) = 21;n; f F (t) P' (r, x - t) dt

-n
n
= 21;n; fF(x-t)P'(r,t)dt
-n
n
= - 2~ f F(x+ t) P'(r, t) dt
-n
= _1_
2;n;
f n
F (x + t) - F (x - t)
2 sint
M( )d
r, t t ,
-t<
where the kernel M (r, t) is defined by

M (1 - r2) 2r sin 2 (t)
(x) (r, t) = (1 + r2 _ 2r cos (t))" - sin (t) P' (r, t) •
The equality
P' (r, t) = 1:
k=-oo
rlkl (ik) exp (ikt)
holds because the infinite series converges uniformly in t. Therefore
- sin (t) P' (r, t) =~

k=-oo
t rlkl [- k exp(i (k + 1) t) + k exp(i (k - 1) t)] ,
so that
n
(xi) 21n f M(r, t) dt

-n
= r.
Since 1 + r2 - 2r cos(t) = 11 - r exp(it)12, it is easy to see from (x) that

for every c5 E JO, n[. the equality
(xii) lim [max{M (r, t) : c5 ~ It I ~ n}] = 0
,+1
holds. Finally, for It I sufficiently small, I
F (x + t)2 sint
- F (x - t)
- DIF (x)
Iis
arbitrarily small. Combining this with (xii), (xi), and (ix), we obtain (viii)
forthwith.
(18.48) Exercise: More on N-Functions. Let [a, bJ be a compact interval
in R and let I be a real-valued function defined on Ja, be.
(a) Suppose that E C Ja, b[ and {3 ~ 0 are such that D+/(x) ~ {3 and
D_/(x) ~ - {3 for every x E E. Prove that A(I(E) ~ {3ACE). [Hints. For
8> 0 and n EN, define En ={x EE: I(t) -/(x) < ({3 + 8) It - xl for all
t E Ja, b[ for which It - xl < ~}. Then EI C E2 C ... and nQl En = E.

For each n, let {In.k}k=l be a cover of En by intervals of length < ~
00
such that }; A(In. k) < A(En) + 8. Then

k=1
00 00
A(I (En) ~ }; A(I (En n In. k) < ({3 + 8) }; A(In, k) < ({3 + 8) (A (E) + 8)
k=l k=l
for all n EN. Let n go to 00 and use (9.17).J

(b) Suppose that I has a finite derivative at all but count ably many
points of [a, bJ. Prove that I is an N-function. [Hint. Consider the sets
An = {x E Ja, b[: I' (x) exists and II' (x) I ~ n} and use part (a).J
(c) Suppose that B is a Lebesgue measurable subset of Ja, b[ and that
I has a finite derivative at each point of B. Prove that I and I' are both
Lebesgue measurable on B and that A(I (B) ~ J II' (x) I dx. [Hints. For
B
8> 0 and n EN, let Bn = {x E B: (n - 1)8 ~ I!,(x) I < n8}. Applying
part (a), we have
00 00
A(I(B» ~ }; A(I(Bn) ~ }; nd(Bn)

n=l n=l
00
~ }; [J II'(x)1 dx + d(Bn)]
n=l B..
= J I!,(x) I dx + deB) .J
B
(d) Use part (c) [not (18.25)J to prove that if I is a continuous N-

function of finite variation on [a, bJ, then I is absolutely continuous on
[a, b]. [Hints. For [c, d] C [a, b], let B = {x E ]c, d[: I' (x) exists and is
finite}, and let A = [c, d] n B'. Then It(d) - t(c)1 ~ A(t([C, dJ)) = A(t(B))
d
+ A(f(A)) = A(t(B)) ~ J If'(x)ldx. Recall that I' E~l([a, b]).p
c
§ 19. Complex measures and the LEBESGUE-RADON-NIKOD"fM

theorem
In this section we make a further study of the measure-theoretic
significance of absolutely continuous functions and functions of finite
variation. We begin by examining abstract analogues of some of the clas-
sical notions of §§ 17 and 18. We will then use our abstract results to
obtain further information about the classical case.
The most useful generalization of the notion of indefinite integral
seems to be the following. Let (X, d, f-t) be an arbitrary measure space
and let t be any function in ~1 (X, d, f-t). Define '/I on d by
'/I(E) = J tdp.
E
for E Ed. Clearly '/I is complex-valued, '/1(0) = 0, and '/I is count ably
additive (12.32). Thus '/I enjoys two essential properties of a measure.
Since '/I can assume arbitrary complex values, it is not always a measure
in the sense of (10.3). This leads us to define and study signed measures
and complex measures.
(19.1) Definition. Let (X, d) be an arbitrary measurable space. An
extended real-valued function '/I defined on d is called a signed measure if
(i) '/1(0)=0
and
(ii) '/I ("Ql E..) = ~1 '/I (E ..)
for all pairwise disjoint sequences (E..):'=1 of elements of d. A complex-

valued function '/I defined on d that satisfies (i) and (ii) is called a complex
measure.
(19.2) Note. It is implicit in the above definition that the infinite
series appearing in (19.1.ii) must always be meaningful and must converge
[or definitely diverge] to the value on the left side of the equality. In
particular, a signed measure '/I can assume at most one of the values 00
and - 00. For, if '/I(E) = 00 and '/I(F) = - 00, then the right side of the
equality
'/I (E U F) = '/I (E n
F') + '/I (E F) + '/I (E' F) n n
is undefined since it contains both 00 and - 00 among its three terms
[see (6.l.b)].
1 Note that we have here a short proof of (18.25).
§ 19. Complex measures and the LEBESGUE-RADON-NIKODYM theorem 305
Our first goal is to show that just as a function of finite variation can
be expressed as a linear combination of four monotone functions, so a
complex measure can be expressed as a linear combination of four
measures. It is obvious that any complex measure v can be expressed
uniquely in the form v = VI + iV2 where VI and V2 are real-valued signed
measures; simply let VI (E) = Re V(E) and V2(E) = 1m V(E) for all E Ed.
We therefore take up signed measures first.
(19.3) Theorem. Let V be a signed measure on a measurable space
(X, d). We have:
(i) if E, FEd, Iv(E)1 < 00, and FeE, then Iv(F)1 < 00;
(ii) if An Ed (n = 1, 2, 3, ... ) and Al C A2 C ... CAn C .. "
then
lim V(An)
n--)o-oo
= U An);
V (n=l
(iii) if An Ed (n = 1, 2, 3, ... ), if Al:J A 2 :J···:J An:J"', and
Iv (AI) I < 00, then
Proof. To prove (i), observe that v(E) = v (F) + v(E n P). In order
that V (E) be finite, it is necessary and sufficient that both summands on
the right side be finite. Conclusions (ii) and (iii) are proved by repeating
verbatim the proofs of (10.13) and (10.15), respectively. In the proof of
(iii) we use (i) to write V(AI n A~) = V(Al) - v(An). 0
(19.4) Definition. Let v be a signed measure on a measurable space
(X, d). A set P Ed is called a nonnegative set for v if v (P n E) ;:;:; 0 for
all E Ed. A set M Ed is called a nonpositive set for v if v (M n E) ~ 0
for all E Ed. Note that 0 is both a nonnegative set and a nonpositive
set for v. If P is a nonnegative set for v and P' [the complement of P in X]
is a nonpositive set for v, then the ordered pair (P, P') is called a Hahn
decomposition of X for v.
(19.5) Lemma. Let v be a signed measure on (X, d) and suppose that E
is a set in d such that 0 < v (E) < 00. Then there exists a set 5 Ed such
that 5 C E, 5 is a nonnegative set for v, and v(5) > O.
Proof. It follows from (19.3.i) that Iv(F)1 < 00 for allF Ed such that
FeE. Assume that no set 5 of the required sort exists. In particular,
E is not a nonnegative set for v. Let n l be the smallest positive integer for
which there exists a set F; E d such that FJ. C E and v (FI) < - ~.
n1
Then we have
v(E n F I ') = v(E) - v(FrJ > v(E) > 0,
and so, by our assumption, E nFl' is not a nonnegative set for v. As
before, let n2 be the smallest positive integer for which there exists a set
F2 Ed such that F2 c En F,' and V (F2) < -~.

n 2 Then
v(E n (F, U F2)') = v(E) - v(FI) - V (F2) > 0,
and so E n (F, U F2)' is not a nonnegative set for v. Continuing this
process, we obtain a sequence (n")k~1 of minimal positive integers and a
corresponding pairwise disjoint sequence (F")k~1 of sets in d such that
1 00
v (F,,) < - -n for each kEN. Let F = U F". Then we have
k~1
k
1
+ 1) - > 0 .
00 00
00 > v(E n F') = v(E) - v (F) = v (E) -1) v (F,,) > v (E)
k~1 k~1 n k
Thus 1; _1_
k~1 nk
< 00 and E n F' is not a nonnegative set for v. Choose
A Ed such that AcE n F' and v(A) < 0, and then choose k so large
1
that v(A) < - -nk and n" > 2. Now
AUF" C E n (;9: F;)'

and
This contradicts the minimality of n". We conclude that a set 5 of the

required sort exists. D
(19.6) Hahn Decomposition Theorem. Let v be a signed measure on a
measurable space (X, d). There exists a Hahn decomposition ot X tor v.
Moreover, this decomposition is unique in the sense that it (PI' and pn
(P2, P;) are any two such decompositions, then V (PI n E) = v(P2 n E) and
v(PI' n E) = v(P; n E) tor every E Ed.
Proof. Since v takes on at most one of the values 00 and - 00, we sup-
pose that v (E) < 00 for all E Ed. In the other case, simply interchange
the roles of positive and negative by considering the signed measure - v.
Let IX = sup {v (A) : A is a nonnegative set for v}. Choose a sequence
(A")k~1 of nonnegative sets for v such that lim v (A,,) = IX. Define
k-+oo
00 n
P = U A" and Pn
k~1
= U A" (n = 1,2, ... ). It follows by induction on n
k~1
that each Pn is a nonnegative set for v and that v (Pn) ~ v (An). Applying
(19.3.ii), we have v (P n E) = lim v (pn n E) ~ 0 for all E Ed. Thus P
n-+oo
is a nonnegative set for v and v(P) = IX. [Note that IX < 00.]
We next show that P' is a nonpositive set for v. Assuming the contrary,
choose a set E Ed such that E C P' and v (E) > O. Since v does not
assume the value 00, we have v (E) < 00. Apply (19.5) to obtain a set
5 C E such that 5 Ed, 5 is a nonnegative set for v, and v(5) > o. Then
5 U P is a nonnegative set for v and v(S U P) = v(5) + IX> a; this
violates the definition of IX. Therefore P' is a nonpositive set for v.
To prove our uniqueness assertion, suppose that (PI' pn and (P2 , P;)
are two Hahn decompositions of X for'll, and select E Ed. Since
E n PI n P; is a subset of both PI and P; , we have'll (E n PI n pn = 0;
similarly v(E n PI' n P2 ) = o. Thus we have v(E n Pd = 'liCE n (PI U P2 ))
= v(E np2) and v(E npn = 'liCE n (P{ UP;)) = v(E np;). 0
(19.7) Definition. Let 'I' be a signed measure on (X, d) and let (P, P')
be a Hahn decomposition of X for'll. Define '1'+, '1'-, and 1'1'1 on d by:
v+(E) = v(E n P) ;
v-(E) = -v(E n P');
and
1'1'1 (E) = v+(E) + v-(E)
for all E Ed. The set functions'll+' '1'-, and 1'1'1 are called the positive
variation 01 '1', the negative variation 01 '1', and the total variation 01 '1',
respectively.
(19.8) Theorem. Notation is as in (19.7). The set lunctions '1'+, '1'-,
and 1'1'1 are well-defined measures on (X, d). Also we have
(i) v(E) = '1'+ (E) - v-(E) lor all E Ed .1
The proof is very simple, and we omit it.
(19.9) Example. Let (X, d, f-l) be a measure space and let I be an
d-measurable, extended real-valued function on X for which J I df-l is
x
defined. Define 'I' on d by
'I' (E) = J I df-l .
E
Then we have
and
1'1'1 (E) = Jill df-l
E
for all E Ed. If P = {x EX: t (x) > O}, then (P, P') is a Hahn decom-
position of X for'll. Notice that the nonequality 1'1'1 (E) =1= 1'1' (E) I must
occur for some sets E Ed if '1'+ and '1'- are nondegenerate.
tf
(19.10) Theorem. Let 'I' be a signed measure on (X, d). Then
(i) 1'1'1 (E) = sup 1'1' (E k) I: {El> ... , En} is a measurable dissection 01 E}
lor every E Ed.
1 The expression v = v+ - v-is known as the]ordan decomposition of v. in
analogy with (17.16).
20·
Proof. Let E be any fixed set in d, and let (3 denote the right side
of (i). Then we have
n n
£ Iv(E,,)1 = £ Iv+(E,,) - v-(Ek)i
"=1 k=1
n
;2; £ ('11+ (E,,) + v-(E k )
k=1
n
= £
k=1
1'111 (E,,) = 1'111 (E)
for every measurable dissection {EI' ... , En} of E; hence {3 ;2; 1'111 (E).
Consider the dissection {E n P, En PI} where (P, PI) is a Hahn de-
composition of X for v. We get
(3 ~ Iv(E np)1 + Iv(E n PI)I = v+(E) + v-(E) = 1'111 (E) . 0
In view of (19.10), we make the following definition with no risk of
inconsistency.
(19.11) Definition. Let v be a complex measure on (X, d). The total
variation of v is the function 1'111 defined on d by the formula (19.1O.i).1
(19.12) Theorem. Notation is as in (19.11). The set-function 1'111 is a
measure on (X, d).
Proof. It is obvious that 1'111 (0) = O. Thus we need only show that 1'111
is count ably additive. Let (Ai)~1 be a pairwise disjoint sequence of sets
00
in d and let A = .U Ai. Let {3 be an arbitrary real number such that

1=1
{3 < 1'111 (A). Choose a measurable dissection {E}I ... , En} of A such that
n
{3 < £ Iv(E,,)I· Then we have
k=1
n 00
;2; £ £ Iv(E" n Ai)1

k=1 ;=1
00 n
= £ £ Iv(E" nAi)1
;=1 k=1
00
;2; £ 1'111 (Ai) .

;=1
Since {3 is arbitrary, it follows that

00
1'111 (A) ~ £ 1'111 (Ai) . (1)

;=1
If 1'111 (A) = 00 [which, as we shall see in (19.13.v), is impossible], then the

1 Notice the similarity of this definition with our definition of V!/ in (17.14).
reverse of this inequality is obvious. Thus suppose that 1'111 (A) < 00. For
any jo EN and any measurable dissection {B1' ... , Bm} of Ai., we have
and so 1'111 (Ai) < 00 for all j EN. Let e > 0 be arbitrary. For each j,
choose a measurable dissection {Ei'l' ... , E i ,lIj} of Ai such that
~ e
1) Iv(Ei,k)1 > 1'111 (Ai) -V·
k=1
Then for all mEN, we have
i~ 1'111 (Aj) <i~ (-;j + k~ Iv(Ei'k)l)

< e + I'll C=Q+1 Aj)1 + i~ k~llv(Ei.k)1
~ e + 1'111 (A) .
Since e is arbitrary, we have
m
1) 1'111 (Ai) ~ 1'111 (A)
i=1
for every mEN, and so
00
1) 1'111 (Ai) ~ 1'111 (A) . (2)

i=1
Combining (1) and (2), we see that 1'111 is countably additive. 0
(19.13) Theorem. Let 'II be a complex measure on (X, d) and let 'Ill
and '112 be the real and imaginary parts of'll, respectively. Then:
(i) 'Ill and '112 are signed measures on (X, d);
and for all E Ed we have
(ii) v(E) = 'lit (E) - 'Ill (E) + ivt(E) - iV2(E);1
(iii) 1'111 (E) ~ 'lit (E) + 'Ill (E) + 'lit (E) + '112 (E);
(iv) sup{lv(A)I: A Ed, ACE} ~ 1'111 (E)
~ 4· sup{lv(A)1 : A Ed, ACE},·
(v) I'll (E) I ~ 1'111 (X) < 00;
and
(vi) 'lit (E) $; 1'111 (E) and vi (E) ~ 1'111 (E) for j = 1, 2 .
Thus 'lit, 'Ill' 'lit, '112, and 1'111 are finite measures on (X, d).
Proof. Parts (i) and (ii) are obvious. Let (PI> P;) and (P 2, P~) be Hahn
decompositions of X for 'Ill and '112 respectively. For any measurable
1 We call this the Jordan decomposition of'll, again in analogy with (17.16).
dissection {El' ... , En} of E, we have

11 11
1: Iv(Ek)i ~ 1: (vi (Ek) + vl'(E k) + vt(E k) + v2'(E k))

k=l k=l
= vi (E) + vI' (E) + vt (E) + v2' (E) .

Take the supremum over all {Ev ... , En} to obtain (iii). Next let
IX = sup{lv(A)1 : A Ed, ACE}. Applying (iii) and (19.7), we have
Ivl (E) ~ VI (E n PI) + IVI (E n pal + vs(E n Ps) + IVs(E n P~)i

~ Iv(E n PI) I + Iv(E n P;)! + Iv(E n Ps)1 + Iv(E n P~)!
~ 41X. (1)
If A Ed and AcE, then {A, En A'} is a measurable dissection of E,
and so Iv(A)1 ~ Iv(A)1 + Iv(E n A')! ~ Ivl (E). Take the supremum over
all such A's to obtain
IX ~ Ivl (E) . (2)
Combine (1) and (2) to get (iv).
Conclusion (v) follows from (iv) and (iii) [recall that v is complex-
valued and that 00 is not a complex number].
Finally, (vi) follows from (iv) because
vt (E) = Vi (E n Pi) ~ Ivl (E)
and
vi (E) = IVi (E n Pi)! ~ Ivl (E). 0
The equality (19. 13.ii) suggests our next definition, which also is useful
in later arguments and constructions.
(19.14) Definition. Let (X, d) be a measurable space, let 1-'1 and 1-'2
be complex measures on (X, d), and let 1X1 and IXs be complex numbers.
The set-function 1X11-'1 + IXSl-'2 on .91 is defined by
( 1X11-'1 + IXs 1-'2) (E) = 1X1 1-'1 (E) + 1X2 1-'2 (E)
for all E Ed. If I-' is a signed measure on (X, d) and IX ER, then IXI-' is
the set-function on .91 such that
(IXI-')(E) = IX(P(E)) .
If I-' and v are signed measures on (X, d) and if the simultaneous equal-
ities I-' (E) = 00, v (E) = - 00 or I-' (E) = - 00, v (E) = 00 hold for no
E Ed, then we define I-' + v by
(I-' + v) (E) = I-' (E) + v (E) (E Ed) .
In this case, I-' + v is said to be defined. If either of the interdicted pair of
eqUalities holds for some E E.91, then I-' + v is undefined.
(19.15) Notes. (a) Notation is as in (19.14). It is all but obvious that

IXtPI+ 1X2P2 is a complex measure on (X, d), that IXp is a signed measure,
and that P + 'II, if defined, is a signed measure.
(b) If PI and P2 are measures on (X, d) and 1Xl> 1X2 are in [0, 00[' then
1X1PI+ 1X2 P2 is a measure on (X, d).
(c) For notational convenience, we will usually write the Jordan
decomposition (I9.13.ii) of a complex measure'll as
We now consider integration with respect to complex measures, be-

ginning with a useful if rather obvious fact.
(19.16) Theorem. Let 'II be a complex measure on (X, d) and let
4
'II = } ; IX" 'II" be its Jordan decomposition. A complex-valued lunction I
"=1
defined I'III-a. e. on X is in ~l (X, d, 1'111) il and only il it is in ~l (X, d, 'II,,)
lor each k E{I, 2, 3, 4}.
Proof. The theorem is true if I is an d-measurable, nonnegative,
m
simple function [say I = }; P; ~EJ]' because
;=1
m
x
J Id 1'111 = ;=1
1: Pi 1'111 (EI )
~ £P; C~ 'II,,(E,))
4
= }; J Idv" , (I)
k=IX
as (19.13.iii) shows. Furthermore,

m
J I d'll" = 1: P;'II" (E;)
x ;=1
m
~ }; P; 1'111 (E;) = J Idl'lll ' (2)
;=1 x
by (19.13.vi). In the general case, let (Sn):"1 be a nondecreasing sequence
of d-measurable, nonnegative, simple functions which converges a.e.
to III. [Note that 1'111 (A) = 0 if and only if 'II,,(A) = 0 for all k.] As usual,
we apply B. LEVI'S theorem (12.22) to (1) and to (2) with Sn in place of I
to obtain
4
Jill dl'lll ~ }; Jill d'll"
x k=IX
and
fill d"" ~ fill d 1"1
x x
for all k. D
We may now make the following definition.
(19.17) Definition. Notation is as in (19.16). For I E~l(X, .91, 1"/),
define
4
f I d" = 1) rx" f I d"".

X "=1 X
(19.18) Theorem. Let " be a complex measure on (X, d). II

I, g E~1 (X, .91, 1,,1) and rx EK, then
(i) f(/+g)d,,= fl d ,,+ fgd"
x x x
and
f rxl d" = rx f I d" .
(ii)
x x
Thus f· .. d" is a linear functional on ~1 (X, .91, I"/).
x
This is proved by an obvious computation, which we leave to the
reader.
We next define absolute continuity lor measures. As we shall prove in
(19.53), Borel measures on R absolutely continuous with respect to A. are
just the Lebesgue-Stieltjes measures induced by absolutely continuous
nondecreasing functions.
(19.19) Definition. Let (X, d) be a measurable space and let I-' and"
be signed or complex measures on (X, d). We say that" is absolutely
continuous with respect to 1-', and we write ,,~ 1-', if 11-'1 (E) = 0 implies
,,(E) = 0 for all E Ed.
(19.20) Theorem. Let I-' and" be complex or signed measures on (X, d)
4
and let 1) rx"",, be the Jordan decomposition 01 ". Then the lollowing are
k=1
equivalent:
(i) "~I-';
(ii) "" ~ I-' for k E{I, 2, 3, 4};
(iii) 1"1 ~ I-' .
Proof. We consider only the complex case. Suppose that E Ed and
that 11-'1 (E) = O. Suppose that (i) holds. Since l.ttl (F) = 0 for all subsets
FofE such thatF Ed, we alsohave" (F) = 0 for all suchF. From (19.10.i),
we infer that 1"1 (E) = O. Thus (i) implies (iii). If (iii) holds, then 1"1 (E) = 0
and so (19.13.vi) shows that "" (E) = 0 for all k. That is, (iii) implies (ii).
4
Finally, it is obvious that (ii) implies (i), since" = 1) rx"",,. D
k=1
As noted in the introductory remarks to this section, if (X, .91, fl) is a

measure space and I E~1 (X, .91, fl), then the function v defined on .91 by
veAl = J I dfl (1)
A
is a complex measure on (X, d). It is clear that v ~ fl. The LEBESGUE-

RADON-NIKODYM theorem asserts that if v ~ fl and certain other condi-
tions are met, then v has the form (1). We will present several avatars
of this theorem. First, a technicality.
°
(19.21) Lemma. Let fl and v be measures on (X, d) such that
veE) ~ fleE) lor all E Ed. II P > and I E~p(X, .91, fl), it lollows that
I E~p(X, .91, v) and J IfiPdv ~ J IfiPdfl·
x x
Proof. If I E~p (X, .91, fl), then there is a sequence (O'n)~l of .91-
measurable, nonnegative, simple functions increasing to IfiP, and so
It is clear that J O'ndv ~ J O'ndfl, and it follows that

x x
The next lemma, which may at first glance appear rather strange, is
actually the crucial step in our proof of the LEBESGUE-RADON-NIKODYM
theorem.
(19.22) Lemma. Let fl and v be finite measures on (X, d) such that
v ~ fl. Then there is an d-measurable lunction g on X such that g (X) C [0, 1 [
and
(i) JI(I-g)dv= JIgdfl
x x
lor all I E~2(X, .91, fl + v).
Proof. For I E ~2(X, .91, fl + v), define
L(f) = J I dv. (1)
x
Since fl and v, and with them fl + v, are finite, I is also in ~1 (X, .91, fl + v);
and so by (19.21), I is in ~l(X, .91, v). Thus L is defined and finite on
~2(X, .91, fl + v). It is clear that L(rxl + (Jg) = rxL(f) + (JL(g) for all
rx, (J EK and I, g E~2 (X, .91, fl + v). Inequality (13.4.iii) shows that
1 1
IL(f)1 = I Jldvl ~(J IfI2dv)"2(v(X)"2
x X
1 1 1
~(J IfI2d(fl + v))2(v(X))"2 = II/Mv(X)"2.
X
[Here 11/112 denotes the norm in ~2(X, .91, I' + v).] Thus L is a bounded
linear functional on ~2(X, .91, I' + v), and so by (15.11), there is a func-
tion h E~2(X, .91, I' + v) such that
L (I) = fiJi d(1' + v) .1 (2)
X
Actually h is real-valued and nonnegative (fJ, + v)-a.e., as we now show.

For any 1E~2 (X, .91, I' + v), we can write
L(f) = f 1Reh d(1' + v) - if 1Imh d(1' + v) .
x x
Assume that Imh fails to vanish (I' + v)-a. e.; say the set
A = {x: Imh(x) > O}
satisfies (I' + v)(A) > 0. Then L (';A) = f Reh d(fJ, + v) - i f Imh d(1' + v)
A A
is not real. By (1), L is obviously real-valued on real-valued functions.
This is a contradiction. Similarly, if h were negative on a set B such that
(I' + v)(B) > 0, we would have L (';B) < 0, which again contradicts the
definition of L. Thus h is real and nonnegative (I' + v)-a. e.; and we
may suppose it to be so everywhere. The definition (1) of L and the rep-
resentation (2) show that
fl(l-h)dv= flhdl' (3)
x x
for all 1 E~2(X, .91, I' + v).
Next, let
E = {x EX: hex) ~ I} .
Since ';E is in ~2 (X, .91, I' + v), we may apply (3) with 1= ';E to obtain
° ~ I'(E) = f ';Edl'
x x
= f ';E(l - h)dv ~ 0.
x
~ f ';Eh dl'
°
Thus we have I' (E) = 0, and so v (E) = also. [This is our only use of the
hypothesis v ~I'.] Let g = h';E" Then g(X) C [0, 1[ and g = h almost
1 Theorem (15.11) is of course much more than we need to produce the represen-
tation (2): only the case p = 2 is needed. As is the case with many problems in-
volving 2p spaces, the case p = 2 is much the simplest, and there is in fact a proof
of (15.11) forp = 2 couched in terms of abstract Hilbert spaces. This proof is sketched
in (16.56). Using (16.56), we could then prove (19.22) and so also (19.24) and (19.27)
without recourse to (15.11). It would be then possible to prove (15.11) for all
2p (X, d, p,) such that (X, d, p,) satisfies the hypotheses of (19.27). We prefer the
proof given in (15.11), partly because it is completely general and partly because it is
constructive and classical in spirit. In § 20 we construct the conjugate space of
21 (X, d, p,), a process that apparently requires (19.27). Then the general case of
(15.11) could be proved from (19.24).
everywhere with respect to both ft and'll. Thus (3) shows that

fl(1-g)d'll= flgdft
x x
for every I E~2 (X, .91, ft + 'II). 0
(19.23) Theorem [LEBESGUE-RADON-NIKODYM]. Let (X, d) be a
measurable space and let ft and'll be finite measures on (X, d) such that
'II ~ ft. Then there exists a Iztnction 10 E~t (X, .91, ft) such that
(i) f I d'll = f Ito dft
x x
lor all nonnegative, extended real-valued, d-measurable lunctions I on X.
For I E~l (X, .91, 'II), the lunction Ito is in ~l (X, .91, ft) and (i) holds. In
particular, we have
(ii) 'II(A) = flo dft
A
lor all A Ed.
Proof. First consider any bounded, nonnegative, d-measurable
function I. Let g be the function of (19.22.i). Since both g and I are bounded
and ft + 'II is a finite measure, the function (1 + g + ... + gn-l) I is in
~2 (X, .91, ft + 'II) for every positive integer n; and by (19.22) the equality
f (1 + g + g2 + ... + gn-l) 1(1 - g) d'll

X
= f (1 + g + g2 + ... + gn-l) Ig dft

x
holds. Since 0 ~ g (x) < 1 for all x EX, this equality can be written
x
J(1 - gn) I d'll =
x
J
1 ~ g (1 - gn) I dft . (1)
The sequence of functions (1 - gn)1 increases to I as n goes to infinity.

Using (12.22) to pass to the limit in both sides of (1), we have
x
J
Id'll=
x
J
1 ~g Idft· (2)
Putting I= 1 in (2), we see that the function 1 ~ g is in ~t (X, .91, ft) ;

define 10 as the function 1 ~ g •
If I is an unbounded. nonnegative. d-measurable function, then we
may write I = lim 1m' where 1m = min{/. m}, and apply (12.22) to (2)
m ..... oo
to obtain (i). The other assertions of the theorem are now clear. 0
(19.24) LEBESGUE-RADON-NIKODYM Theorem. Let ft and'll be a-finite
measures on (X, d) such that 'II ~ ft. Then there exists a nonnegative,
finite-valued, d-measurable lunction 10 on X such that
(i) f I d'll = f Ito dft
x x
lor all nonnegative, extended real-valued, d-measurable lunctions I on X.

For I E~1 (X, d, 'II), the lunction Ito is in ~1 (X, d, p), and (i) holds. In
particular, we have
(ii) 'II (A) = !Iodp
A
lor all A Ed. Moreover, 10 is unique in the sense that il go is any nonnegative,
extended real-valued, d-measurable lunction lor which (ii) holds, then
go = 10 p-a.e.
Proof. Let {A ..}:=l and {B ..}:=l be pairwise disjoint families of d-
measurable sets, each with union X, such that p (A ..) < 00 and'll (B..) < 00
for all n. The family f'(f = {Am n B ..}:, .. =l is pairwise disjoint, and its
union is X. Also, each member of this family has finite'll and p measure.
Let (E.. ):=l be any arrangement of f'(f into a sequence of sets. For each n,
define P.. and'll.. on d by Pn(A) = p(A n En) and 'IIn(A) = 'II (A n En).
Then Pn and 'lin are finite measures on (X, d) and 'lin ~ P.. for each n;
and so (19.23) applies. Thus, for each n, we obtain a nonnegative, finite-
valued, d-measurable function In on X such that
(1)
for all nonnegative d-measurable functions Ion X. Let 10 be the func-

tion on X which is equal to In on E .. for all n EN. It is easy to see that
10 is nonnegative, d-measurable, and finite-valued. Also, by (12.21), if I
is a nonnegative, d-measurable function on X, then
This proves (i) for nonnegative, d-measurable I; (i) for I E~1 (X, d, 'II)
and (ii) follow at once.
To prove the uniqueness of 10' let go be d-measurable and satisfy (ii).
Assume that there exists a set E Ed such that p (E) > 0 and 10 (x) > go (x)
for all x EE. For some n, we have p(E n En) > 0; if A = En En' then
we have 'II (A) < 00,0 < p(A) and 10 - go> 0 on A. Applying (12.6) and
(ii), we obtain
0< !(/o - go) dp = 'II (A) - 'II (A) = O.
A
This contradiction shows that 10 ~ go p-a.e.; similarly 10 ~ go p-a.e. 0

The most general form of the LEBESGUE-RADON-NIKODYM theorem
of any conceivable use deals with an arbitrary 'II ~ P and a p that can be
decomposed in such a way that (19.24) can be applied to each piece. The
definition is as follows.
(19.25) Definition. Let (X, d, fl) be a measure space. Suppose that

there is a subfamily .fF of d with the following properties:
(i) 0 ~ fl(F) < 00 for all F E.fF;
(ii) the sets in .fF are pairwise disjoint and U .fF = X;
(iii) if E Ed and fl (E) < 00, then fl (E) = }; fl (E n F) ;1
FE:!'
(iv) if 5 C X and 5 n FEd for all F E.fF, then 5 Ed.
Then (X, d, fl) and fl itself are said to be decomposable and .fF is called
a decomposition of (X, d, fl).
Our general LEBESGUE-RADoN-NIKODYM theorem holds for a decom-
posable fl and an arbitrary v such that v ~ fl. We need the following
technical lemma.
(19.26) Lemma. Let (X, d) be a measurable space, and let fl and v be
measures on (X, d) such that fl (X) < 00 and v ~ fl. Then there exists
a set E Ed such that:
(i) for all A Ed such that ACE, v(A) = 0 or v(A) = 00;
(ii) for all A Ed such that ACE, fl(A) = 0 if v(A) = 0;
(iii) v is a-finite on E'.
Proof. With an eye to proving (i), consider the family £0 = {B Ed:
C C Band C Ed imply that v(C) = 0 or v(C) = oo}. Note that 0 E£0.
Define ex by
ex = sup{fl (B) : B E£0} ;
it is obvious that ex ~ fl (X) < 00. There is a non decreasing sequence of
00
sets (Bn):=1 in £0 such that lim fl (Bn) = ex; let D = U Bn· Since fl is
n~oo n=l
count ably additive, it is clear that fl (D) = ex (10.13). The set D is in £0,
for if C Ed and C CD, then
v(C) = v(C n B l ) + !; v(C n (Bn n B~_l» .
n=2
Since each set C n (Bn n B~_l) is in d and has v measure 0 or 00, the
same is true of C.
Now consider the set D'. We will show that for every set FeD' such
that FEd and v (F) > 0, there exists a set FI in d such that Fj c F and
(1)
If v(F) < 00, then (1) is trivial. Thus suppose that v(F) = 00 and assume
that v(G) = 0 or v(G) = 00 for every subset G of F such that G Ed.
Under this assumption, it is clear that F U D E£0. Since v ~ fl and
1 This possibly uncountable sum is defined as the supremum of the sums
E f.l (E n F), where p) runs through all finite subfamilies of ~.
FEfId
'/I(F) > 0, wehave,u(F) > 0. Butwealsohave,u(FU D) = ,u(F) + ,u(D) > iX,

and this is a contradiction since F U D EP). The existence of a set Fl
satisfying (1) follows.
We will next show that '/I is a-finite on D'. To this end, let
.f7 = {F Ed : FeD' and '/I is a-finite on F} .
There is a nondecreasing sequence (Fn)~=1 in .f7 such that lim ,u(Fn)
....... 00
00
= sup{,u (F) : F E.f7} = (3 ; let F = U Fn. Since F is a countable union

n=1
of sets on which '/I is a-finite, '/I is also a-finite on F; thus F E .f7. Also, the
equality ,u(F) = (3 follows from (10.13). We claim that '/I(F' n D') = 0.
If not, then by the preceding paragraph, there exists a set H Ed such
°
that H C D' n F' and < '/I (H) < 00; hence F U H E.f7 and ,u (H) > 0.
However we have
,u(F U H) = ,u(H) + ,u(F) > ,u(F) = (3 ~ ,u(F U H) .
This contradiction shows that '/I(F' n D') = 0, and so '/I is a-finite on D'.
Finally, we define the promised set E. Let
(g={BEd:BCD and '/I(B)=O}.
There is a nondecreasing sequence of sets (Bn)~=1 in (g such that
00
lim ,u (Bn) = sup{,u (B) : B E(g} = y. Let G = U Bn, and let E = D n G'.
n~oo n=l
Since '/I (G) = 0, '/I is a-finite on E' = D' U G; i. e., (iii) is satisfied. The
assertion (i) is clear since E cD. To prove (ii), assume that there is a set
BeE such that BEd, '/I (B) = 0, and ,u (B) > 0. Then we must have
G U B E(g. But this is impossible since
,u(G UB) = ,u(G) + ,u(B) > ,u(G) = y ~ ,u(G U B);
and so (ii) is proved. 0
We can now prove our final version of the LEBESGUE-RADoN-
NIKODYM theorem.
(19.27) LEBESGUE-RADON-NIKODYM Theorem. Let (X, d,,u) be
decomposable with decomposition .f7, and let '/I be any measure on (X, d)
such that '/I ~,u. There exists a nonnegative, extended real-valued, .91-
measurable lunction loon X [which can be chosen finite-valued on each
FE.f7 where '/I is a-finite] with the lollowing properties:
(i) '/I(A) = flo d,u
A
lor all A Ed that are a-finite with respect to ,u;
(ii) f I d'/l = f Ito d,u
x x
lor all nonnegative, extended real-valued, d-measurable lunctions I on X
such that {x EX: I (x) > O} is a-finite with respect to ,u;
(iii) il I E~1 (X, .91,,,) and {x EX: I (x) =1= O} is a-finite with respect
to ft, then Ito E~1 (X, .91, ft) and J I d" = J Ito dft·
x x
Also 10 is unique, in the sense that il go is any nonnegative, extended real-
valtted, .9I-measurable lunction on X for which
(iv) ,,(A) = J go dft
A
lor all A Ed such that ft(A) < 00, then 10~E and gO~E are equal ft-a.e. lor
all E Ed that are a-finite with respect to ft.
Proof. For eachF E ff, the restriction of" toF is absolutely continuous
with respect to the restriction of ft to F, and so by (19.26) there are sets
DF and EF in .91 such that: DF n EF = 0 ; DF U EF = F ; (19.26.i) and
(19.26.ii) hold for E F ; and" is a-finite on DF . If" is a-finite on F, then
we take DF = F. Since" is a-finite on Dp and ft is finite on DF, we can
apply (19.24) to assert that there is a nonnegative, finite-valued, .91-
measurable function f&F) defined on DF such that the conclusions of
(19.24) hold for ft and" restricted to DF . Now let fo be the function on X
such that for all F Eff,
x = { fbF) (x) if x EDF ,
foe) 00 ifxEEF • (1 )
Plainly to is finite-valued if each DF is F. It is easy to see that fo is

.9I-measurable; we leave this to the reader. Let us also write D for
U DF and E for U EF.
FE.'F FE.'F
To show that fo has all of the properties ascribed to it, we first con-
sider a set A Ed for which ft (A) < 00. Condition (19.25.iii) ensures that
ft(A) = }; ft(A nF) , (2)
FEF,
the family ffo being a countable subfamily of ff. Condition (19.25.iii)
also implies that
,u(A n (U{F: F Eff n ~'})) = 0,
and since "~ft' we have
,,(A n (U{F:FEffn~'}))=O,
so that
,,(A) = }; peA n F) . (3)
FE~
It is clear from (1) that
" (A n F) = peA n DF) + " (A n EF)
= J fodft+p(AnE F)· (4)
AnDp
By (19.26.i), the value" (A n EF ) is either 0 or 00; by (19.26.ii) and ab-

solute continuity, peA n EF) is zero if and only if ft(A n EF) is zero.
Therefore
Combining (3), (4), and (5), and harking back to (12.21), we find that
v(A)=,Ev(AnF)=,E flodf-t=flodf-t. (6)
FE.'F, FE.'F, AnF A
Assertion (i) is now obvious, since both ends of (6) are countably additive.
The equality (ii) is proved by considering characteristic functions,
then simple functions, and finally passing to the limit using (11.35) and
(12.22). Equality (iii) follows upon writing I as a linear combination of
functions in ~t (X, .91, v).
It remains to prove the uniqueness of 10. Let go be as in liv). Then for
every FE.'F and every .9I-measurable subset A of DF , we have
v(A) = flodf-t= fgodf-t,
A A
and so by (19.24), 10 (x) = go (x) for f-t-almost all x EDF • Now assume that
there exists a set A Ed such that A C EF , f-t (A) > 0, and
go(x) < 00 = 10 (x) for all x EA.
From (19.26.ii), we see that v(A) > O. For each n EN, write
An = {x EA: go (x) < n}.
00
Then (A n):'=1 is a nondecreasing sequence and ,,=1
U An = A. Applying
(10.13), we infer that
0< v(A) = lim v (An) ;
"->-00
hence there exists an n such that v(An) > 0; since An C EF , we have

v (An) = 00. A glance at (iv) reveals that
00 = v(An) = f go df-t ~ nf-t(An) < 00.
A"
This contradiction proves that go = 10 f-t-a.e. on F. 0
(19.28) Corollary. Let (X, .91, f-t) be a a-finite measure space and let v
be any measure on (X,d) such that v ~ f-t. Then (19.27.i), (19.27.ii),
(19.27.iii) hold lor all A Ed, all nonnegative, extended real-valued, .9I-meas-
urable I, and all I E~l (X, .91, v), respectively.
Proof. Since X is the union of a countable family of sets of finite
f-t-measure, (X,d, f-t) is plainly decomposable; and the restrictions im-
posed in (19.27.i), (19.27.ii), and (19.27.iii) are no restrictions at all in
the present case. 0
(19.29) We next consider the LEBESGUE-RADON-NIKODYM theorem
for measure spaces (X,.A" £), where X is a locally compact Hausdorff
space and £ is a measure as in § 9. It turns out that every such measure
space (X, -4., t) is decomposable in the sense of (19.25), and so we will be

able to apply (19.27). The discussion is somewhat technical, unavoidably
so in our opinion.
(19.30) Theorem. Let X be a locally compact Hausdorff space, and let
(X, -4., t) be a measure space constructed as in §§ 9 and 10. There exists a
family ~ of subsets 0/ X with the following properties:
(i) the sets in ~ are compact and have [finite!} positive measure;
(ii) the sets in ~ are pairwise disjoint;
(iii) if F E~, U is open, and un F =1= 0, then t(U n F) > 0;
(iv) if E E -4. and t (E) < 00, then E n F is nonvoid for only a countable
number of sets F E~;
(v) the set D = X n (U~)' is t-measurable and is locally t-null;
(vi) if Y is a subset of X such that Y n F E -4. for all F E~, then
Y E-4..
Proof. Let )K be the collection of all families F of subsets of X
enjoying the following properties:
(1) the sets in F are compact and have positive t-measure;
(2) the sets in F are pairwise disjoint;
(3) if F E F and U is an open subset of X for which U n F =1= 0,
then t(U n F) =1= O.
Clearly)K is nonvoid, since the void family satisfies (1) -(3) vacuously.
It is also clear that )K is a partially ordered set under inclusion: for
~, ~ E)K we have ~ C ~ or we do not. If )Ko C )K and )Ko is linearly
ordered by inclusion, then it is clear that U {F: F E )Ko} E)K. Thus
ZORN'S lemma implies that )K contains a maximal family, which we
call ~.
Let us show that ~ satisfies all of our conditions. Condition (i)
holds because of (1) and the fact that compact sets have finite t-measure
(9.27). Condition (ii) is just (2), and (iii) is the same as (3). To verify (iv),
consider any E E -4. such that t (E) < 00, and select an open set U such
that E C U and t(U) < 00 [see (9.24)]. Assume that En F is nonvoid for
an uncountable number of sets F E~. The same is then true of U n F,
and property (iii) shows that t(U n F) > 0 for an uncountable number
of F E~. Hence t(U) = 00, and this contradiction proves (iv).
We next prove (v). Let U be any open set such that t(U) < 00 and
let ~ be the [countable] subfamily of ~ consisting of all F such that
t{U n F) > O. Then (iii) shows that ~ = {F E~: un F =1= 0}. It is
plain that
t(U) = t(U n (U~)) + t(U n (U~)')
since U ~ is a-compact and hence t-measurable; hence also
t(U) = t(U n (U ~)) + t(U n (U ~)') ,
and (10.31) implies that U ~ and (U ~)' = Dare t-measurable. Up to

this point in the proof, we have not needed the maximality of ~. To
prove that D is locally t-null, we need this property. If D is not locally
t-null, then by definition there is a compact set C such that t (C n D) > 0,
and from (10.30) and the fact that D is t-measurable, we infer the existence
of a compact set H such that H c C n D and t (H) > O. Consider the
family Olt of all open sets U such that t (U n H) = O. Then t(H n (U Olt)) = 0,
for otherwise H n (U Olt) would contain a compact set E of positive
t-measure (10.30), and E would be covered by a finite number of sets
U n H, each of zero t-measure. The set H n (U Olt)' is compact and
contained in D, and
t(H n (U Olt)') = t(H) - t(H n (U Olt)) = t(H) > O.
Also if V is open and V n H n (U Olt)' =l= 0, then V ~ Olt, so that
t(V n H n (U Olt)') = tty n H) - t(V n H n (U Olt)) = t(V n H) > 0 .
Therefore we can adjoin H n (U Olt)' to ~ and still preserve properties
(1) - (3). This contradicts the maximality of ~.
It remains only to prove (vi), to do which we appeal to (10.31). Let U
be an open set such that t(U) < 00, let..q: = {F E~: un F =l= 0}, and
let Y be as in (vi). Then we write
U nY= (U n Y n D) U (U n Y n U~)
= (U n Y n D) U F~, (U n Y n F) .
The set U n Y n D is t-measurable because it is locally t-null (10.32)
[note that every subset of a locally t-null set is locally t-null]. The set
U (U n Y n F) is a countable union of t-measurable sets and so is
FE.'F,
t-measurable. Hence un Y is t-measurable, and (10.31) shows that Y
is t-measurable. 0
(19.31) Corollary. Let (X, vK" t) be as in (19.30). The measure space
(X, vK" t) is decomposable in the sense of (19.25).
Proof. Let ~ and D be as in (19.30) and let ~= ~ U {{x}: x ED}.
It is clear from (19.30.ii) and (19.30.v) that the sets in ~ are pairwise dis-
joint and that U~ = X; i.e., (19.25.ii) holds for~. Since each set in ~
is compact, (19.25.i) holds. Suppose that E EvK, and that t(E) < 00.
Since D is locally t-null, we have t(E n D) = 0, and so t(E n {x}) = 0
for all xED. Thus it follows from (19.30.iv) and the countable additivity
of t that
£ teE n F) = £ teE n {x}) + 1: t(E n F)
FE.'JI' "ED FEF,
= teE n D) + teE n (U~)) = t(E);
hence (19.25.iii) is satisfied by ~. Condition (19.25.iv) follows at once
from (19.30.vi). 0
§ 19. Complex measures and the LEBESGUE-RADON-NIKOnYM theorem 323
We now present our version of the LEBESGUE-RADON-NIKODYM

theorem for locally compact Hausdorff spaces.
(19.32) Theorem. Let X be a locally compact Hausdorff space and t a
measure on X as constructed in §§ 9 and 10. Let 11 be any measure whatever
on (X, J,) such that 11 ~ t. Then all 01 the conclusions 01 (19.27) hold
with I' replaced by t and d by J,.
Proof. By (19.31), (X, J" t) is decomposable. Now we have only
to apply (19.27). 0
(19.33) Remark. If X is a locally compact Hausdorff space and if
t and 1J are any two outer measures on X constructed from nonnegative
linear functionals as in § 9 such that t (E) = 0 implies 1J (E) = 0, then
J, C ~. To see this, let A EJ, and let F be compact. According to
(10.34), we have A n F = B U E where BE f11I(X) and teE) = O. Since
f11I (X) C ~ and 7J (E) = 0, we have A n F E~. It follows from (10.31)
that A E~. Thus (19.27) holds for t and 1J on (X, J,).
(19.34) Note. Our labors throughout (19.25) to (19.33) would be
unnecessary if all measures were a-finite and all locally compact Hausdorff
spaces a-compact. One may well ask if the generality obtained in (19.27)
and (19.33) is worth the effort. Many mathematicians believe it is not.
But we feel it our duty to show the reader the most general theorems that
we can reasonably produce and that he might reasonably need. Examples
showing the failure of other plausible versions of the LEBESGUE-RADON-
NIKODYM theorem appear in Exercise (19.71).
(19.35). It is easy to extend the LEBESGUE-RADON-NIKODYM theorem
to the cases that I' and 11 are signed measures or complex measures, by
making use of Hahn and Jordan decompositions. We restrict our attention
to just one important extension of this sort.
(19.36) Theorem. Let (X, d, 1') be a a-finite measure space and let 11
be a complex measure on (X, d) such that 11 ~ 1" Then there exists a unique
10 E~1 (X, d, 1') such that
(i) J I d'P = J 110 dl'
x x
lor alii E ~ (X, d, 1111) and
(ii) 'P (A) = J 10 dl'
A
lor all A Ed. Moreover,
(iii) 1111 (A) = J 1/01 dl'
A
lor all A Ed, and in particular
(iv) I'PI (X) = J 1/01 dl' = 11/0111'
X
Proof. Let 'P = 1:" eX"'P,, be the Jordan decomposition of 'P [(19.13.ii)
,\=1
and (19.15.c)]. According to (19.13) and (19.20), "1' lIa, 'P3' 'P" and I'PI are
21*
finite measures on (X, d), and each of them is absolutely continuous

with respect to ft. By (19.24), there exist nonnegative, finite-valued,
d-measurable functions Ik on X such that I E ~l(X, .91, Vk) implies
J I dVk = J Ilk df' (1)
x x
4
(k E {I, 2, 3, 4}). Let 10 = }; rxklk' where rxl = 1, rx 2 = -1, rxa = i,
k=l
and rx4 = -i. Of course 10 is in ~l (X, .91, ft) since each Ik is in ~l (X, .91, ft)
[set I = 1 in (I)J. Then if I E~l (X, .91, Iv!), (19.16) shows that IE~l (X,d, Vk)
for all k, and so (1) and (19.17) yield
4 4
J I dv = }; rxk J I dVk =
rxk J Ilk dft = J 110 df'
};
X k=l X k=l X X
[here we have used the fact that Ilk E ~l (X, .91, ft) for all k, which is
evident from (1)]. This proves (i). The identity (ii) follows from (i) upon
taking I = ~.A' since Ivi is a finite measure.
To prove the uniqueness of 10 in ~l (X, .91, ft), suppose that
ho E ~l (X, .91, ft) and that v (A) = J ho dft for all A Ed. Let v = Cf + iT,
A
where Cf and T are real-valued. Then
J Reho dft = Cf(A) = J Re/odft
A A
for all A Ed. It follows, as in the uniqueness proof of (19.24), that

Re ho = Re loft-a. e.; similarly 1m ho = 1m loft-a. e. Thus ho and 10 are the
same element of ~l (X, .91, ft).
Finally we prove (iii). Let A Ed be fixed. For an arbitrary measur-
able dissection {AI> ... , An} of A we have
n n n
}; Iv(A;)1 = }; I J 10 dftl~ }; J 1/01 dft = J 1/01 dft·
i=l i=l Aj i=l Aj A
Taking the supremum over all such dissections, we obtain

Ivl (A) ~ J 1/01 dft .
A
(2)
To prove the reversed inequality, use (11.35) to choose a sequence

(Cfm):'=l of d-measurable simple functions such that Cfm--'>- ~A sgnfo ft-a.e.
and ICfml ~ I~.A sgnfol ~ 1. Then
I/oCfml ~ 1/01 E~l (X, .91, ft)
for all m, and so LEBESGUE'S dominated convergence theorem (12.30)
implies that
J 1101 dft = J 10~A sgnfo dft
A X
= lim
'n~oo
J 10Cfm dft .
X
(3)
II
Each (Jm has the form 1: Pi~AJ' where {AI> ... , An} is a measurable dis-
;=1
section of A and IPil ~ 1 for all j. Therefore
Ij 10(Jm dftl = I;~ Pi i 10 dftl ~ ;~ [1 10 d ft [

n
= 1: Iv(Ai)1 ~ Ivl (A). (4)
;=1
J 1/01 dft ~ Ivl (A) . (5)
A
Now (2) and (5) together imply (iii). Setting A = X, we get (iv). 0
(19.37) Note. The reader should find it illuminating to compare the
statement and the proof of (19.36.iii) with the corresponding result about
absolutely continuous lunctions (18.1).
(19.38) Corollary. Let v be a complex measure on (X, d). Then there
exists an d-measurable lunction 10 on X such that:
(i) Ito I = 1;
(ii) v(A) = J lodlvl loraU A Ed;
A
and
(iii) J I dv = J!fo d Ivl lor aU IE 5:.1 (X, d, Ivl) .
x x
Also,
(iv) I J I dvl ~ Jill d Ivl lor aU I E5:.1 (X, d, Ivl) .
x x
Proof. Obviously we have v ~ Ivl. Define 10 as in (19.36); then (ii) and
(iii) are immediate. LetA = {x EX: 110 (x) I < l}andB = {x EX: 110 (x) I > I}.
Apply (19.36.iii) to get
J (1 - 1/01) dlvl = 0 = J (1/01- 1) dlvi .
A B
From (12.6), we see that

Ivl (A) = 0 = Ivl (B) .
Thus, with no harm done, we redefine 10 on AU B so that (i) holds. For
I E5:.1 (X, d, Ivl) we have
IxJ Idvl = IxJ !fodlvll ~ xJ l!foldlvl
=
x
Jill dlvl '
and so (iv) holds. 0
We next consider a relationship between pairs of measures that is the
antithesis of absolute continuity.
(19.39) Definition. Let p. and v be measures, signed measures, or

complex measures on (X, d). We say that p. and v are [mutually] singular,
and we write p. 1- v, if there exists a set BEd such that 1p.1 (B) = 0 and
1111 (B') = O. We also say that p. [11] is singular with respect to 11 [p.].
(19.40) Theorem. Let p., v, and (J be complex or signed measures on
(X, d) such that v + (J is defined, and let tX be in K. Then:
(i) v ~ p. and (J ~ p. imply tXV ~ P. and (v + (J) ~ p.,'
and
(ii) v 1- p. and (J 1- p. imply tXlI1- p. and (v + (J) 1- p..
Proof. Suppose that v ~ p. and (J ~ p., and let E Ed be such that
1p.1 (E) = O. Then (tXv)(E) = tXV (E) = tX· 0 = 0 and (11 + (J)(E) = 11 (E) + (J (E)
= O. Thus (i) holds.
Next suppose that 111- p. and (J 1- p.. Choose A and B in d such that
1p.1 (A) = 1p.1 (B) = 0, 1111 (A') = 0, and I(JI (B') = O. Let C = AU B. It is
clear that
111 + (JI (C') ~ (1111 + 1(Ji) (C') ~ 1111 (A') + I(JI (B') = 0
and that 1p.1 (C) ~ 1p.1 (A) + 1p.1 (B) = O. Thus (v + (J) and p. are singular.
It is obvious that tXlI 1- p.. 0
(19.41) Theorem. Let p. and 11 be complex or signed measures on (X, d).
Then the following are equivalent:
(i) p. 1- v,'
(ii) 1p.11- 1111;
and
4 4
(iii) p." 1- Vi for all j, k E {I, 2, 3, 4}, where }; tX"p." and }; tXilli are the
k=1 ;=1
Jordan decomposition of p. and v, respectively.
Proof. This follows easily from (19.40) and (19.13). We omit the
details. 0
Our next theorem shows that if a (J-finite measure p. is given on a
measurable space, then all (J-finite measures on that space can be analyzed
by considering only those that are absolutely continuous or singular
with respect to p..
(19.42) LEBESGUE Decomposition Theorem. Let (X, d, p.) be a (J-finite
measure space and let v be a complex measure or a (J-finite signed measure on
(X, d). Then we have
(i) v = 111 + Vs '
where 111 ~ P. and 1121- p.. Moreover the decomposition (i) is unique,' indeed,
if the decomposition v = ii1 + iis has the same properties, then ii1 = v1 and
ii2 = V2'
Proof. In view of the Jordan decomposition of 11 and because of
(19.20), (19.40), and (19.41), it suffices to consider the case in which v
is a measure. Thus suppose that" is a a-finite measure on (X, .sat). The

measure " is absolutely continuous with respect to the measure p. + ",
and both are a-finite. Hence, by (19.24), there exists a nonnegative, real-
valued, .sat-measurable function 10 on X such that
Jld,,= Jllod(P+,,) (1)
x x
for all nonnegative .sat-measurable functions I. We claim that 10 ~ 1 a.e.
with respect to p. + ". To see this, let E = {x: 10 (x) > I} and assume that
(p. + ,,)(E) > O. We can write
E= U{x: lo(x) ~ 1 +~},

ft=l n
and from this equality it is clear that there exists a number ac > 1 such
that (p. + ,,)(F) > 0, where
F={x:/o(x) ~ ac> I}.
Since p. +" is a-finite, there is a set A E.sat such that A c F and
0< (p. + ,,)(A) < 00. Putting I = ~.A in (1), we obtain
,,(A) = J 10 dp, + J 10 d" ~ acp,(A) + ac,,(A),
A A
and so
(1 - ac)" (A) ~ acp, (A) .
Since ac> 1, the inequality " (A) > 0 implies that acp.(A) < 0; it follows
that " (A) = O. Hence we have 0 ~ acp,(A), and so p,(A) = 0 also. Thus
the equality (p, + ,,) (A) = P, (A) + ,,(A) = 0 holds, and this is a contra-
diction. If we write 11 = min{/o, I}, then 0 ~ 11 ~ 1 and
J I d" = Jill d(p. + ,,) . (2)
x x
Now consider the set
B= {x EX: 11 (x) = I} .
For a set C E.sat such that C C B, p,(C) < 00, and ,,(C) < 00, we put
I= ~c in (2) to find that
,,(C) = J 11 dp. + J 11 d" = p.(C) + ,,(C) ,

c c
so that p.(C) = O. But P, is a-finite on B, and it follows that p,(B) = O.
Defining "2 on d by
"2(A) = ,,(A n B) ,
"2 "2
we obtain (B') = o. Thus and p. are mutually singular. [Obviously".
is a measure on (X, .sat).] Writing
"1(A) = " (A n B')
for A E.sat, we obviously have" = "1 + "2'
We must show that '1'1 is absolutely continuous with respect to It.

To do this, first consider any C Ed such that /t(C) = 0 and v(C) < 00.
We have
and so
J (1 - 11) dv = 0 . (3)
enB'
The function 1 - 11 is positive on B', and so the equality (3) implies that
'1'1 (C) = v(C n B') = 0, as desired. For an arbitrary C in d such that
00
/t (C) = 0, write C = U Cn where the Cn's are pairwise disjoint sets in d

n=!
and V (C n) < 00 for all n. The case just considered applies to each Cn' and
so '1'1 (C n) = 0 for n = 1, 2, ... ; and of course it follows by countable
additivity that '1'1 (C) = o.
Finally, we prove the uniqueness of the decomposition. Suppose that
V = '1'1 + '1'2 = VI + V2, where VI and VI are absolutely continuous with
respect to fl and '1'2 and v2 are mutually singular with respect to fl. Let B
and 13 be sets in d such that fl(B) = fl(13) = 0 and '1'2 (B') = V2(13') = O.
For a set C in d such that C C BUB we have /t(C) = 0, and so by the
absolute continuity of '1'1 and VI the equality v(C) = v2(C) = V2(C) holds.
If C c B' n 13' and C Ed, then the equality '1'2 (C) = v2(C) = 0 holds.
Hence for an arbitrary set A Ed we have
v2(A) = v2(A n (B U B)) + v2(A n (B' n 13'))
= v2(A n (B U 13)) + v2(A n (B' n 13')) = v2(A) .
Since '1'2 = V2 and every measure in sight is a-finite, the equality '1'1 = VI
also holds. 0 1
(19.43) Definition. The [essentially uniqueJ function 10 appearing
in the LEBESGUE-RADON-NIKODYM theorem [(19.23), (19.24), (19.27),
(19.36)J is often called the LEBESGUE-RADON-NIKODYM derivative 01 V
with respect to It, and the notation ~: is used to denote 10' Also this
relationship among fl' v, and 10 is sometimes denoted by the formulae
dv=/od/t and v=/o/t.
(19.44) Theorem [Chain Rule]. Let /to, /tl> and /t2 be a-finite measures
on (X, d) such that
Then
(i) /t2 ~ /to
1 There is a proof of (19.42) in S. SAKS, loco cit. (18.42), which does not use the
LEBESGUE-RADON-NIKODYM theorem.
and
(1'1') --=--.--
dps dpB dP1 p,-a e
dpo dp1 dpo 0 ••
Proof. Assertion (i) is trivial. Let 10 = dd P1 and 11 = dd P B • Assertion (ii)

Po P1
follows from the equalities
f I dp,2= xf ttl dp,l = xf 111/0 dp,o·
x
0
We turn now to a detailed study of the relationship between absolute

continuity for functions and for measures. We first show that the map-
pings
described in §§ 8 and 9 establish one-to-one correspondences between the

set of all normalized nondecreasing functions ac on R (8.20), the set of all
nonnegative linear functionals on <roo(R), and the set of all [regular]
Borel measures on R.I
(19.45) Theorem. Let t be any regular Borel measure on R. Define ac
on R by the rule
(i) ac(t) = 0 j
t([O,t[) il t> 0,
il t = 0,
-t([t,OD if t < o.
Then ac is a nondecreasing, real-valued, left-continuous function on R. Also,
we have lim ac(t) > - 00 if and only if t(]- 00, O[) < 00 and lim ac(t) < 00
~-oo ~oo
if and only if t([O, ooD < 00.

Proof. If 0 < tl < t2 , then
ac(t2) - ac(tI) = t([O, t2[) - t([O, tID = t([tv t2D ;?; 0,
and so ac(t2) ;?; ac(tI)' Iftl < 0;:::;; t2, then the inequalities ac(tI);:::;; 0;:;;; ac(t2)
hold, and so again ac(fI) ~ ac(t2)' If tl < t2 < 0, then
ac(t2) - ac(tl) = -t([t2, OD + t([tl' O[) = t([tl' t2[) ;?; O.
Thus in all cases the relation t2 > ~ implies that ac (t2) ;?; ac (t I) .
To show that ac is left continuous, consider first t > O. Let (8,,):'=1 be
any decreasing sequence of positive real number converging to 0 and such
00
that 81 < t. We have [0, t[ = ,,=1

U [0, t - 8,,[, and so
ac(t) = t([O, t]) = lim t([O, t - 8"D = lim ac(t - 8,,) .

fI--+oOO ~oo
1 A Borel measure is of course a measure defined on the a-algebra of Borel sets.

In (19.45), we use the symbol "," to denote a regular (12.39) Borel measure, although
,in §§ 9 and 10 was defined as an outer measure on all subsets with a a-algebra .,If.
of measurable sets. The distinction is wiped out by Theorem (19.48). It is worth
while as well to note that if P is a Borel measure on R and P (F) < ()() for all compact
sets FeR, then p is automatically regular (12.55).
Since n[- 2. ,°[

n= 1 n
= 0 , lim ex (- 2.)
we have n-+-oo n
= lim
"-+00
£ ([- 2.,
n
°[) = £( 0 )
= 0 ; and so ex is left continuous at 0. Finally, if t < 0, then the equalities
ex(t- !)= - £([t- ! , O[)=-[£([t- !,t[)+£([t,O[)]
show that ex(t)-ex(t- !)=£([t- !,t[). Since .. 0[t- !,t[=0, it

1
follows that ~~ £ ([t - ! ' t[) = 0, and so ex is left continuous at t.

By (10.13), the equalities lim ex(t)
'-'00
= £([0, ooD and - lim ex(t)
t-+- 00
= £ (]- 00, 0]) hold, and these equalities prove the last assertion of the
theorem. 0
(19.46) Remarks. The function ex of (19.45) may fail to be right
continuous. For example, if £ is the point mass defined by
£(A)=
{
°I ifif °O~A
EA ,
,
it is easy to see that the corresponding ex is not right continuous at 0. The
choice of the definition of ex was in this respect arbitrary ; it could as well
°
have been defined so as to be right continuous and nondecreasing. Also,
the choice that ex (0) = is an arbitrary normalization. For finite measures
£ [i. e., £(R) < 00], it is often more convenient to normalize ex so that
lim ex(t) = 0. As in (8.20), we shall use the term normalized nondecreasing
I~-(X)
lunction to mean a nondecreasing function ex on R that is left continuous

and satisfies ex (0) = 0. Theorem (19.45) therefore defines a mapping £ -* ex
of the set of all regular Borel measures on R into the set of all normalized
nondecreasing functions. We next show that this mapping is onto.
(19.47) Theorem. Let ex be a normalized nondecreasing lunction on R,
let 5« be the Riemann-Stieltjes integral corresponding to ex as in § 8, and let
A« be the Lebesgue-Stieltjes measure on R constructed Irom 5« in §9 see [in
particular (9.19)). II {J is the normalized nondecreasing lunction constructed
Irom A« in (19.45) , then (J = ex. Thus a < bin R implies
(i) A«([a, b[) = ex(b) - ex(a) .
Proof. The equalities (J (0)
= °
ex (0) = are trivial. Take t >
in R. Then (J(t) = A«([O, tD, and
°
we want to show that this num-
ber is ex (t) . Consider any decreas-
ing sequence (e..):=1 of positive
numbers such that e.. -* and such
that 0< e.. < e1 < t. Let I.. be the
°
function whose graph is pictured
Fig. 8 in Figure 8.
The functions (In)':=1 converge to ~[O,t[ everywhere, and

In ~ ~[-",t[ E~1 (R, ..A'Aot' Atz) for n = I, 2, ....
By (12.24) we have lim J IndAtz = J lim IndAtz = Atz([O, t[) = f3(t). By
11-+00 R R "-+00
(12.36) the equality

J IndAtz =
R
Stz(ln)
_co
holds for n = 1, 2, . We complete the proof by showing that
lim Stz(ln) = ex (t). If L1n is a subdivision of [-e1> t] such that
{-e n,O,t-en}CL1,
then
Stz (In) ~ U (In' ex, L1n)
~ (ex(t) - ex(t - en)) + (ex(t - en) - ex (0)) + (ex(O) - ex(-en ))
= ex(t) - ex(- en) -+- ex(t) - ex(O) = ex(t)
and
Stz (/n) ~ L (In' ex, L1n) ~ ex (t - en) - ex (0) -+- ex (t -) = ex (t) .
[Recall that ex is left continuous.] Thus
ex (t) = lim Stz (In) = lim
n~oo ~OOR
J IndAtz = f3 (t) .
A similar argument shows that f3(t) = ex(t) for t < O. Relation (i) now
follows from the definition of f3. 0
(19.48) Theorem. The mapping
(i) t -+- ex
defined in (19.45) is a one-to-one mapping 01 the set 01 aU regular Borel
measures on R onto the set 01 aU nondecreasing, real-valued, lelt continuous
lunctions ex on R such that ex (0) = O. The inverse 01 this mapping is the
mapping
(ii) ex -+- Atz.
Thus every regular Borel measure on R is a Lebesgue-StieUjes measure.
Proof. Let ex be given. Theorem (19.47) shows that ex is the image of
Atz in (i) and so this mapping is onto. Suppose that ex is also the image of
the regular Borel measure t. Now (19.47.i) and (19.45.i) show that
Atz([a, b[) = ex(b) - ex(a) = t([a, b[) (1)
for all a < bin R. Each open subset U of R can be expressed as a countable
disjoint union of sets of the form [a, b[, e.g., ]0, 1[= ,,=1 n +1 1 ' ~[,
n
and U[
so (1) implies thatAtz(U)=t(U) for all open UCR. Since Atz and t are both
regular, this implies that Atz(E) = t(E) for all E EfJI(R). Thus A.. = t and
the mapping (i) is one-to-one. The rest is clear. 0
(19.49) Remark. There is a different proof of (19.48) which does not

use the regularity of l except for the requirement that l([a, b[) < 00 for
all a < bin R, this being needed to define IX. [Of course (12.35) implies that
such an l is regular.] This shows again that all Borel measures on R
satisfying t([a, b[) < 00 are Lebesgue-Stieltjes measures and hence are
regular. The alternate proof runs as follows. Use (19.48.1) to show that t
and A.o: agree on the algebra of all finite disjoint unions of intervals of any
of the forms [a, bL ]- 00, be, or [a, 00[. Note that A.o: and l are both a-
finite on this algebra, and finally use the uniqueness part of HOPF'S
extension theorem (10.39) to infer that A.o: and t agree on the a-algebra
generated by this algebra, namely fJI (R).
(19.50) Theorem. Riemann-5tieltjes integrals So: are the only nonnega-
tive linear lunctionals on <roo(R).
Proof. Let I be any nonnegative linear functional on <roo(R) and let t
be the measure constructed from I as in §§ 9 and 10. Use (19.48) to find an
IX such that l = A.o: on fJI(R). Finally infer from RIEsz's representation
theorem (12.36) that
I (f) = f I dl = f I dAo: = 50: (f)
R R
for all I E<roo(R). 0 1
(19.51) Notation. To simplify our statements, we will write
t -+ IX
to mean that l is a regular Borel measure on R and that IX is the nor-

malized nondecreasing function on R obtained from tin (19.45). Of course
(19.48) shows that l = A.o:.
(19.52) Theorem. Let t -+ IX. Then IX is continuous il and only il
l({X}) = 0 for all x E R. In fact, t{{x}) = lim IX (x + h) - IX (x) for all x E R.
htO
Proof. For all x ER, we have
l({X}) = leO! [x, x + ~ [) = ~n;, t([x, x + ~ [)

= lim IX (x +~)
~oo n
- IX(X) = lim IX (x
htO
+h) - IX(X) ,
as (19.47.i) shows. 0
(19.53) Theorem. The lunction IX is absolutely continuous (18.10) in
every interval [- p, P], PEN, il and only il the corresponding measure t is
absolutely continuous with respect to Lebesgue measure k 2
1 Theorems (19.50) and (19.48), together with (12.56), prove that the Riemann
integral is the only invariant nonnegative linear functional on ~oo(R), up to a positive
constant. See (8.16) supra.
2 The a-algebra Jt). of course contains ~ (R). In applying the definition of
absolute continuity of measures we here consider the measure spaces (R, ~ (R), £)
and (R. ~ (R). J.).
Proof. Suppose first that oc is absolutely continuous on [- p, P] for

all PEN. Let A be any subset of R such that A(A) = 0, and let E be a
positive number. For every PEN, there exists tJ(P) >0 such that
.
J; (oc(b k) - oc(ak)) < e whenever {]ak' bk[}k=l is a pairwise disjoint
k=l
.. ..
family of intervals such that U ]ak' bk[C [- P,P] and}; (b k - ak) <tJ(P).
k=1 k=1
Since A(A n [- P,P]) = 0, there are pairwise disjoint intervals {]a k, bk[}k'::1
<Xl <Xl
such that A n ]- p, P[ C
U ]a k, bk[ C [- P, P] and}; (b k- ak) < tJ (P).
k=1 k=1
By the choice of tJ(P), we have
00
J; (oc(b k) - oc(ak)) ;:;;;; e.

k=1
In view of (19.47) and the continuity of oc, we also have oc(b k) - oc(ak)
= t(]ak' bk[) for all k EN, and so
e ~ k~ (oc(bk) - oc(ak)) = %1 t(]ak, bk[ = t (kQI ]ak' bk[) ,
which implies that teA n ]-p.P[);:;;;; e. As e is arbitrary. the equalities

teA) = lim teA
p-..oo
n ]-P.P[) = 0
follow.
Next suppose that t is absolutely continuous with respect to A
[on Borel sets]. and consider any interval [-P,P] where P EN. By
(19.24), there is a nonnegative Borel measurable function 10 such that
teA) = J 10 dA
A
forallBorelsetsACR. In particular J 10d)"=t([-p.P]) <00. and

[-P.P]
so 10 E ~l([-P, P]. gj(R), )"). Let e> 0 be given. By (12.34) there is a
tJ> 0 such that J 10 dJ.. < e for all Borel sets A C [-P. P] such that
A
J..(A) < tJ. Thus let {Jak' bk[}k=1 be any pairwise disjoint family of open
intervals with union A C [- P. P] for which the inequality A. (A) < tJ
holds. Then we have
J 10 dA. < e.
A
and the integral is obviously equal to

.. .
L t([a k, bk[) = J; (oc(b k) - oc(a k))·
k=1 k=l
That is, oc is absolutely continuous on [- p. P]. 0

We close this section by obtaining a classical decomposition of

measures £, and with it a corresponding decomposition for functions IX.
We begin by "skimming off" the discontinuous part of £.
(19.54) Theorem. For x ER, let e% be the set function such that
e% (A) = ';..4. (x) for all A CR. Let {Xk}k'=l be any countable subset of R,
and let cP be a mapping of {Xk} into ]0, oo[suchthatE{cp(xk): IXkl ~ n}< 00
00
for n = 1,2,3, .... Let t(A) = }; cp(xk)e%k(A) for all A cR. Then t
k=l
[on f1I (R) J is the unique regular Borel measure obtained from the nonnegative
linear functional
00
I (I) = }; cP (Xk) f (Xk)
k=l
on (too(R).
(19.55) Definition. Measures t defined as in (19.54) are called purely
discontinuous. A measure t such that t ({x}) = 0 for all x ER is called
continuous. 1
(19.56) Theorem. If t is purely discontinuous, then the corresponding IX
has derivative 0 a. e. 2
Proof. Let a.. = ';).. ,00[. For t > 0, we have
lX(t) = £([0, t[) = E{£({u}) : 0 ~ u < t}
= }; t ({Xk}) a%k (t) .
%k~O
By (17.18), the equality 1X'(t) = } ; £ ({Xk}) a~k(t) holds a.e. Since

%k~O
a~k (t) = 0 for t =!= Xk, we have IX' (t) = 0 a. e. for t > o. For t < 0, consider
the function Tu = ';)-oo,u]. Then lX(t) = -}; t({Xk})£%k(t), and so IX' (t) = 0
%1:<0
a. e. for t ~ O. 0
(19.57) Theorem. Let £ be any regular Borel measure on R. Then £
can be expressed in exactly one way in the form
(i)
where £" is a purely discontinuous measure as in (19.54) and tc is a continuous
regular Borel measure.
Proof. Let D = {x ER : t({x}) > O}. Since t([ -n, n]) is finite for each
n EN, D must be countable. If D is void, let t" = 0 and tc = t. Otherwise
let D = {Xl> x2, ••. } be an enumeration of D, where xi =1= X k ifj =1= k,and
1 Theorem (19.54) and Definition (19.55) have obvious generalizations to

(X, .Ii" t) for an arbitrary locally compact Hausdorff space X. See (9.20) and
(10.22).
I Wherever "almost everywhere", "a. e.", etc. is written with no further
qualification, we mean with respect to Lebesgue measure.

define
It is obvious that tel is a purely discontinuous measure as in (19.54) and

that
tel(E) = I: t({X}) ~ t(E) (1)
xEE
for every Borel set E. It is immediate from (1) that

I: I (Xl<) t({Xl<}) = f I dtel ~ f I dt (2)
k R R
for every nonnegative Borel measurable function I on R. Define I. on

ctoo(R) by
(3)
Since t (F) is finite for every compact subset F of R, both integrals on the
right side of (3) are finite. Thus (2) shows that I. is a nonnegative linear
functional on ctoo(R). Let t. be the regular Borel measure on R constructed
from I. as in § 9. Theorem (12.36) shows that
f I dt = f I dt. + f I dtel = f I d(tc + t4)
R R R R
for all I Ectoo(R). The equality (i) follows from (19.54) and (12.41). For
all x ER, it is clear that tc({x}) = t({x}) - tel ({x}) = 0, and a moment's
thought shows that the decomposition (i) of t is unique. 0
(19.58) Remark. Let t be a regular Borel measure on R and let t. and
be as in (19.57). Let the correspondences t -'>- a, t. -'>- a., and td -'>- ad
tel
be as in (19.51). It is easy to see from (19.45) that
a = ac + ael'
The function a c is continuous (19.52). The function ael is called a saltus
lunction; it has derivative zero a. e. (19.56) and a jump [or saltus] equal
to a(x+) - a(x) = t{{x}) at each of its [countably many] discontinuities.
Next we turn to the question of singularity as regards measures and
their corresponding functions a. First we need a lemma.
(19.59) Lemma. Let ft and v be a-finite measures on a measurable space
(X, d). Then we have v .1 ft il and only il there exists no a-finite measure ji
on (X, d) such that ji =i= 0, ji ~ v, and ji ft. <
°
Proof. If v .1 ft, then v = + v is the unique Lebesgue decomposition
of v given in (19.42). Suppose that ji is a a-finite measure on (X, d) such
<
that ji ~ v and ji ft. Then there is a a-finite measure n on (X, d) such
that v = ji +n [define n (E) = lim (v (E nAn) -
11->'00
ji (E n An)), where
Al e A2 e···, v(An) < 00, and X = nQ1An]. Thus

')I = v+ :7t1 + :7t2 , (1)
where :7t1 ~ fl, and :7t2 1.. fl,. Since v + :7t1 ~ fl" (1) is a Lebesgue decompo-
sition of ')I. Therefore v = O.
Suppose that ')I and fl, are not mutually singular. Use (19.42) to write
')I = ')11 + ')12' where ')11 ~ fl, and ')12 1.. fl,. Then ')11 =1= 0 and ')11 ~ ')I; hence ')11 is
the desired v. 0
(19.60) Theorem. Let l be any regular Borel measure on R and let oc
be the normalized nondecreasing function associated with l as in (19.45).
Then l and A are mutually singular if and only if 0c' vanishes almost every-
where.
Proof. Suppose first that l and Aare not mutually singular. By (19.42),
we have l = II + l2' where II =1= 0, II ~ A, and l2 1.. A. Then II is a regular
measure, as (19.49) shows. 1
Next suppose that II -> OCl> i.e., II = A"". Then OC1 is absolutely continuous
on every interval [-P, PJ (19.53), and (19.45) and (18.16) imply that
il ([0, x[) = oc1 (x) = J" oc~ (t) dt for x> 0

o
and
o
ll([X, O[) = -oc1 (x) = J oc~(t) dt for x < O.
Since II =1= 0 and II is regular, oc~ is positive on a Borel set E such that
A(E) > o. If l2 -'>- oc2 , then oc = OC1 + oc2, and so oc' = oci + oc~ ~ oc~ a. e. ;
hence oc' cannot vanish a. e. We have shown that 0c' = 0 a. e. implies i 1.. A.
To prove the converse, suppose that 0c' > 0 on a Borel set A of positive
Lebesgue measure. By (18.14), 0c' is Lebesgue measurable on R and of
course oc' ~ 0 a. e. Define a on !!J (R) by
a(E) = J oc' (t) dt.
E
It is obvious that
(1)
and that
a(A»O. (2)
1 Here is an alternate proof. If E is a bounded Borel set, choose a decreasing
sequence (U")~=l of bounded open sets such that Ee n Uti =
n=l
00
B, and A(B n E')
= O. Then 11(B n E') = 0, 1,(U,) ~ I(U,) < 00, and 1,(E) = 1,(B) = lim 1,(Un ) •
...... 00
Unbounded Borel sets of finite I,-measure are now easily dealt with. If U is
open in R, choose an increasing sequence (Fn)~=l of compact sets such that
n A') = O. n A') = 0
00
A = U Fn e U and A(U Then 1,(U and 11(U) = 1,(A)
n=l
= lim 1,(Pn). Thus 1, is regular .
...... 00
From (18.14) we have

b
O'([a, b[) = J <x'(t) de ~ <x(b) - <x(a) = t([a, b[) (3)
II
for all a < bin R. Like tl in the preceding paragraph, 0' is a regular Borel
measure on R. It follows from (3), as in the proof of (19.48), that
O'(E) ~ t(E) (4)
for all E EffI(R). Combining (1), (2), (4), and (19.59), we see that t..L A
cannot obtain. Thus t ..1. A implies <x' = 0 a. e. 0
We now present our main decomposition theorem for regular Borel
measures on R and their corresponding nondecreasing functions.
(19.61) Theorem. Let t be any regular Borel measure on R. Then t can
be expressed in just one way in the form
(i) t = ta + t8 + td'
where ta' t8, and td are regular Borel measures on R, ta ~ A, ta ..L A, td is
purely discontinuous, and ta is continuous. If <X, <xa, <X., and <Xd are the
corresponding nondecreasing functions [see (19.45)], then:
(ii) <X = <Xa + <X, + <Xd;
<Xa is absolutely continuous on every compact interval; <X, is continuous;
<X~ = 0 a. e.; and <Xd is a saltus function. Furthermore, we have
(iii) ta (E) = J <X' (t) dt

E
tor all Borel sets E, and
(iv) <Xa(x) = (i- J <x'(t) dt

0
for x
<X' (t) dt for x < 0 .

~ O.
"
Proof. We proved in (19.57) that t = 'c + 'd' where 'c is regular and
continuous and 'd
is purely discontinuous. We also proved that this
decomposition is unique. In (19.58) we produced the decomposition
<X = <Xc + <Xd' where <Xc is continuous and <Xd is a saltus function. In (19.56)
we showed that <X~ = 0 a. e. ; hence <X' = <X~ a. e. Now define by formula 'a
'a
(iii). Then is the regular measure 0' that appears in the proof of (19.60).
It is clear that 'a ~ A. Let 'a~ <Xa. Then <Xa is absolutely continuous on
every compact interval (19.53) and 'a(E) = J <x~(t)dt for all E EffI(R).
E
Applying (19.45) and (18.14), we have
b
<Xa(b) - <Xa(a) = 'a([a, b[) = J <x~(t)dt ~ <Xc(b) - <Xc (a) (1)
II
whenever a < b in R. Let <X, be the function <Xc - <Xa. Then <X, is continuous
and, by (1), IX. is nondecreasing. Next let t8 -+ IX" i.e., t, = All.' Then t, is
continuous (19.52) and regular and
(2)
Formula (iv) follows at once from (iii) and the definition of lXa' From this
fact and (18.3) we get IX~ = IX' = IX~ a. e.; hence a; = 0 a. e., and so t,.l A
(19.60). Thus (2) gives the unique Lebesgue decomposition of tc with
respect to A. 0
As usual, we close the section with a set of exercises. None of them is
essential for subsequent work, but they all illustrate the theory in one
way or another, and we recommend their study to the reader.
(19.62) Exercise. Let" be a signed measure on (X, d). Prove that
,,+(E) = sup{,,(F) : FEd, FeE}
and
,,-(E) = -inf{"(F) : FEd, FeE}
for all E Ed.
(19.63) Exercise. Let X be a nonvoid set and let d be an algebra of
subsets of X. Suppose that" is a real-valued set function defined on d
such that "(0) = 0 and " (A U B) = ,,(A) + " (B) if A and B are disjoint
in d. Let ,,+
and ,,- be defined as in (19.62).
(a) Prove that ,,+
and ,,- are finitely additive measures on (X, d).
(b) Suppose that sup{I,,(A)1 : A Ed} < 00. Prove that" =,,+ - ,,-.
[This is the Jordan decomposition of a finitely additive signed measure.]
(19.64) Exercise. Let X = [- 1, 1[ and let d be the algebra of all
finite disjoint unions of intervals of the form [a, b[ C [-1, 1[. Let
f (x) = x1 for x =l= 0 and f (0) = O. Define" on d by
"C.~l[a", bloC) = ,,~f(b,,) - f(a,,).

Prove that" is well defined and satisfies the hypothesis of (19.63), but
" =l= ,,+ - ,,-. Is there a Hahn decomposition of X for,,?
(19.65) Exercise. Let X = [0, 1] and d = &1([0, 1]). Define" on d
by ,,(E) = A(E) + iA(E n [0, ;]).
(a) Compute 1"1 in terms of A.
(b) Show that, in this case, strict inequality holds in (19.13.iii).
(c) Find a Borel measurable function g on [0, 1] such that Igl = 1
and ,,(E) = f gd 1"1 for all E Ed.
E
(19.66) Exercise. Let (X, d) be a measurable space. Prove that the
set of all complex measures on (X, d) with setwise linear operations and
11"11 = 1"1 (X) is a complex Banach space.
(19.67) Exercise. Let (X, d) be a measurable space and It and 11

complex or signed measures on d. Then 11 is called #-continuous if
lim 11 (E) = 0 [that is, for every e > 0 there is a <5 > 0 such that
Ipj(Ej-+-O
1#1 (E) < <5 implies 111(E)1 < eJ. If 11 is finite, prove that 11 is #-continuous
if and only if it is absolutely continuous with respect to #.
(19.68) Exercise. Let (X, vIt, #) be a measure space and (11n):=1 a
sequence of finite measures on vIt that are absolutely continuous with
respect to #. Suppose that lim 11n (E) = 11 (E) exists and is finite for all
..-..00
E E vIt.
(a) Prove that the 11n'S are uniformly absolutely continuous with
respect to #; i.e., lim 11n(E) = 0 uniformly in n.
p(E)->O
(b) Prove that 11 is a measure.
(c) Do (a) under the modified hypothesis that # is a complex or
signed measure on vIt and that the 11n'S are complex measures. Prove
that, in this case, 11 is a complex measure.
[Hints. Consider the metric space (vIt, e) defined in (10.45). Show that
each 11n is well defined and continuous on this space. For given e > 0, the
families
JI.."n= {E Evil: l11n(E) - 11m (E)I ~ ;}, m, n = 1,2,3, ... ,
and
vi(, -
p-
n
m,n~p
Jt..m.n. p= 1,2•... ,
are thus closed. Apply the Baire category theorem to obtain an ~
having an interior point A. For a set B E vIt, write
11n(B) = 11q(B) + [11.,(B) - 11q(B)]
and use the identity 11" (B) = 11" (A U B) - 11" (A n B'), k = 1, 2, 3, ... ,
to estimate 11.. (B). Use (a) to prove (b). To do (c), use # to define a metric
space analogous to (vIt, e) and proceed as in (a) and (b).]
(19.69) Exercise. Let X be a nonvoid set and vIt a a-algebra of subsets
of X. Suppose that (11..):=1 is a sequence of nonzero complex measures on
vIt such that lim 11n (E) = 11 (E) exists and is finite for all E E vIt. Prove
..-..00
that 11 is a complex measure on vIt. [Hint. Let # (E) = n~ I;:I(~) 2-",
show that each 11n is absolutely continuous with respect to #' and apply
(19.68).]
(19.70) Exercise. Let {#,,}..EI be a family of measures on a a-algebra
d. Prove that the set function # given by
#(A) = 1.: #,,(A)
"EI
is a measure on d.
22*
(19.71) Exercise: Examples relating to the LEBESGUE-RAnON-NmonDl

Theorem.
(a) Let (X, .91, p,) be an arbitrary measure space and let 'II on .91 be
defined by: 'JI(A) = 0 if p,(A) = 0 and 'II (A) = 00 if p,(A) > O. Prove that
(X, .91, 'II) is a measure space and that 'II ~ p,. Find a function 10 for which
the conclusions of (19.24) are valid.
(b) Let X be a locally compact Hausdorff space admitting a nonzero
measure, as in § 9 such that ,({x}) = 0 for all x EX. Let p, be counting
measure (lOA.a), with its domain of definition restricted to 1.. Show
that '~p, and that there is no function 10 on X for which (19.24.ii)
obtains. Find an 10 for which (19.24.ii) holds for all A E1. that are
a-finite with respect to p,.
(c) Let X be a locally compact Hausdorff space admitting a measure,
as in § 9 such that X is not a-finite with respect to t and such that
t({x}) = 0 for all x EX. Prove that there is a subset of X that is locally
t-null but not t-null. [Use (19.30), and choose a set 5 containing just one
point in each F E~.]
(d) Let X = R2. Impose first on X the topology Rd x R [refer to
(9041) for a description of this topology]. Let I be any nonnegative linear
functional on crOO(Rd x R) such that t ({a} x R) > 0 for all a ER and t
vanishes for points. [It is evident that many such 1's exist.] Next repeat
this construction with the topology of R x RtJ on X; construct a non-
negative linear functional J on croo(R x Rd) with corresponding measure
'f} for which 'f} (R x {b}) > 0 for all bE Rand 'f) vanishes for points.
Consider the measure space (X, 1. n~, ,+ 'f}). Observe that
t ~ t + 'f}. Prove that (19.27.i) holds for no function 10 on X, even for all
sets finite with respect to t + 'f}.
(e) Generalize part (d) to products X x Y, where X and Yare suit-
able locally compact Hausdorff spaces.
(19.72) Exercise. State and prove an analogue of FUBINI'S theorem
(17.18) on term-by-term differentiation, recast in terms of an infinite
series of measures on a measurable space.
(19.73) Exercise. Suppose that p, and 'II are a-finite measures on
dv dp. / dv
(X,d) such that p, ~ 'II and 'II ~p,. Prove that dp. =1= 0 a. e. and d; = 1 dp.
a. e. [Note that p, und 'II have exactly the same sets of zero measure.]
(19.74) Exercise. Let [a, b] be a closed interval in R and let 1 be a
function of finite variation on [a, b]. Prove that
(a) 1= g + h, where g is absolutely continuous on [a, b] and h' = 0
a.e. on [a, b], and
(b) the decomposition in (a) is unique except for additive constants.
(19.75) Exercise. There exist a-finite measures on (fj(R) that are not
Lebesgue-Stieltjes measures. Consider the following pathological example.
§ 20. Applications of the LEBESGUE-RADON-NIKODYM theorem 341
Let (rn)~l be an enumeration of Qand let g be the function such that

1
g(x)=x- 2 if O<x<l,
g (x) = 0 otherwise.
Define f on R by the rule
00
f(x) = }; 2-ng(x + rn) .

n=l
(a) Prove that t E~1 (R).
Define fl on ~ by
fl(E) = J f2dA.
E
Note that fl ~ A.
(b) Prove that fl([a, bJ) = 00 for all a < b in R.
(c) Prove that fl is O'-finite.
00
(d) Find a sequence (Fn)':=l of Borel subsets of R such that U Fn = R
n=l
and fl(Fn) < 00 for all n EN. Can the Fn's be chosen so as to be compact?
(e) Why is this example not a counterexample to (19.48) ?
(19.76) Exercise. Extend the result of (16.48.b) to subspaces of
~2 (X, d, fl), where (X, d, fl) is an arbitrary decomposable measure
space. That is, let !m and ff be as in (16.48.a). Then there is a set E Ed
such that projection onto !m is just multiplication by ~E' and so !m is the
set of all t in ~2(X, d, fl) that vanish on E'.
(19.77) Exercise: Lebesgue decomposition for decomposable measures.
Let (X, d) be a measurable space and let fl and v be decomposable
measures on (X, d).
(a) Prove that fl + v is a decomposable measure.
Modify the definition (19.39) of singularity as follows. Say that fl and v
are mutually singular if there is a set BEd such that fl (B n E) = 0 for
all E Ed that are O'-finite with respect to fl and 'II(B' n F) = 0 for all
FEd that are O'-finite with respect to v.
(b) State and prove an analogue of (19.24) for arbitrary decomposable
measures fl and v. Show that your result is a true generalization of (19.42).
§ 20. Applications of the LEBESGUE-RADON-NIKODYM theorem

(20.1) Introduction. The LEBESGUE-RADON-NIKODYM theorem has a
large number of applications. It is very useful in establishing some weIl-
known properties of integrals, in computing the conjugate spaces of
various classical Banach spaces, in elucidating certain concepts of
probability theory, and in studying product measures. We cannot give
all known applications of the LEBESGUE-RADON-NIKODYM theorem,
or even any large part of them, and keep the text of reasonable size.
We choose four famous applications, which are either themselves impor-

tant theorems in analysis or needed to establish such facts.
(20.2) Integration by substitution. We will applytheLEBEsGUE-RADON-
NIKODYM theorem to prove a very general theorem on integration by
substitution or by "change of variable". We wish to use the construction
of continuous images of measures given in (12.45) and (12.46), and also
(19.24). Hence we consider locally compact Hausdorff spaces X and Y
and a continuous mapping ffJ of X onto Y. It is convenient to suppose
that Y is a-compact. [We could avoid this hypothesis, but at the cost of
tedious complications.] Let p, and ebe measures on X and Y, respectively,
in the sense of § 9. Suppose as in (12.45) that ffJ-l(F) is compact in X for
F compact in Y, or that p, (X) < 00. Then define the measure'll on Y just
as in (12.45). Since Y is a-compact by hypothesis, 'II must be a-finite
(9.27). Theorem (12.46) implies that
(i) 'II(B) = p,(ffJ-l(B))
for all 'II-measurable subsets B of Y. With these preliminaries, we can
state and prove our general theorem on integration by substitution.
(20.3) Theorem. AU notation is as in (20.2). Suppose that
(i) e(ffJ(E)) = 0 lor aU E C X such that p,(E) = O.
Then there is a nonnegative, real-valued, Borel measurable lunction w on X
with the lollowing properties. For every I E~1 (Y, ~, e), the lunction
(10 ffJ)w is in ~1(X,~, p,), and
(ii) J I (y) de (y) = JI 0 ffJ (x) w (x) dp, (x) .
y x
Furthermore, w can be taken to have the lorm 11 0 ffJ lor a Borel measurable
lunction lion Y.
Proof. Let 'II be as in (20.2). Suppose that BeY and that 'II(B) = o.
Then as noted in (20.2.i), we have 0 = 'II(B) = p,(ffJ-l(B)), and so our
hypothesis (i) implies that e (B) = e( ffJ( ffJ-1 (B))) = O. That is, e < 'II. We
apply Theorem (19.24) to the measurable space (Y,9i(Y)) and the
measures'll and e. Thus there exists a nonnegative, real-valued, Borel
measurable function lion Y such that
y
J g(y) de(y) = yJg(y) 11(y) d'll(Y) (1)
for all Borel measurable functions g on Y for which the left side of (1) is
defined. Also, if g is in ~1(Y' 9i(Y), e), then gft is in ~l(Y' 9i(Y), 'II).
We now appeal to (12.46.ii) to write
y
J g(y) 11(y) d'll(Y) = xJ go ffJ(x) 11 0 ffJ(X) dp,(x) (2)
for allg E~1(Y' 9i(Y), e). Since ffJ is continuous and ft is Borel measurable,
it follows that 110 rp is Borel measurable [see (1O.42.a)J. Combining (1)

and (2) and writing w = 11 0 rp, we obtain
J g(y) de(y) = J go rp(x) w(x) df-t(x) (3)
y x
for all g E~l(Y' 8l(Y), e)·
Suppose that BeY and that e (B) = o. Let (Un) be a decreasing
sequence of open subsets of Y such that Un::) B, e (U1 ) < 00, and
00
lim e (Un) = O. Write A =

........ 00
n Un.
,,=1
Put g = ~A in (3) and use (10.15) to
obtain
0= e(A) = J ~A 0 tp(x) w(x) df-t(x) .
x
Thus (~A 0 tp) w vanishes f-t-a. e. on X, and hence (~B 0 tp) w also vanishes
f-t-a. e. on X. If h is any ~-measurable function on Y that vanishes
except on B, we clearly have (h 0 tp) w = (h 0 tp) (~B 0 tp) w, and so
(h 0 tp)w is ~-measurable and vanishes f-t-a.e. on X. Finally, consider
an arbitrary I E~1 (Y,~, e). By (12.63), we can write
I=g+h,
where g is Borel measurable and h vanishes e-a. e. on Y. Applying the last
observations, we see that (f 0 tp) w = (g 0 tp) w + (h 0 tp) w, and so (f 0 tp) w
is the sum of a Borel measurable function and a function that vanishes
f-t-a. e. on X. Hence (f 0 tp) w is ~-measurable. Furthermore, (3) yields
J I de = yJ g de = xJ (g 0
y
tp) w df-t =
x
J (f 0 tp) w dll· 0
A classical case is that in which X and Yare closed intervals, as

follows.
(20.4) Theorem. Let [a, bJbe an interval in R and let tp be a nonconstant,
real-valued, continuo~ts N-lunction 1 defined on [a, bJ. Let [IX, PJ be the
image interval tp([a, bJ). There is a nonnegative, real-valued, Borel measur-
able lunction w on [a, b] such that lor all I E~1([1X, (n~, A.), (f 0 tp)w
is in ~1 ([a, b], ~, A.) and
fJ b
(i) J I(y) dy = flo tp(x) w(x) dx.
a
"
Furthermore, w can be taken to have the lorm 11 0 tp for a certain Borel
measurable function lIon [IX, PJ.
Proof. In (20.3), we take X = [a, bJ and Y = [IX, PJ. For f-t we take
Lebesgue measure Aon [a, bJ, and for e, Lebesgue measure Aon [IX, PJ.
Definition (18.24) ensures that our mapping function tp satisfies (20.3.i).
Thus (20.3) applies. 0
1 For the definition, see (18.24).
The function w in (20.4) can be rather complicated to compute [see

(20.6)-(20.8)]. For monotone cp, however, it is simply Icp'l, as we now
show. [This is the classical form of the theorem on integration by sub-
stitution. ]
(20.5) Corollary. Let cp be a monotone continuous N-function with
domain [a, bJ and range [Ct, fJ] (Ct < fJ). Then cp is absolutely continuous,
and the lunction w 01 (20.4) is equal to Icp'l A-almost everywhere on [a, b].
Thus lor I E~1 ([Ct, fJ]), we have (f 0 cp) Icp'l E~1 ([a, b]), and
P b
(i) f I(y) dy = flo cp(x) Icp' (x) I dx.
" II
Proof. Theorem (18.25) shows that cp is absolutely continuous, since
as a monotone function cp has finite variation. With no loss of generality,
we suppose that cp is nondecreasing. Applying (20.4.i) to the function 1 on
[Ct, fJJ, we obtain
P b
cp(b) - cp(a) = f dy = f w(t) dt. (1)
" a
For every x E[a, b], the inclusion cp-l( cp ([a, x]) :::> [a, x] holds, and hence
by (20.4.i) applied to ~'1'{[II,XJ) we have
p
cp(x) - cp(a) = f ~'1'{[a,xJ)(y) dy
"b
= f ~'1'{[a,xJ) 0 cp (t) w (t) dt
a
b
= f ~'1'-l('1'{[a,xl)(t)w(t)dt
II
x
~ fw(t)dt. (2)
a
Similarly
b
cp(b) - cp(x) ~ f w(t)dt. (3)
x
Inspection of (1), (2), and (3) shows that equality must hold in both (2)
and (3). Since w E~l([a, bJ) by (20.4) and
x
cp(x) = f w(t) dt + cp(a) ,
II
(18.3) shows that cp' (x) = w (x) a. e. 0

The following exercises illustrate the complications attendant on
choosing functions w as in (20.4) for nonmonotone mapping functions cp.
(20.6) Exercise. Let cp be the function on [0, 1J such that cp (t)
= min{t, 1 - t}. Note that cp is absolutely continuous and hence an
N-function (18.25). Consider cp as a mapping of [0, 1] onto [0, ~]. to

which (2004) may be applied.
(a) Let v be a function in ~1 ([0, 1], ..A)., A) with the property that
4 1-.
(i) J v dA + J v dA = d- e
• 1-4
for all [e, d] C[0, ~]. Prove that

1
2" 1
(ii) J I(y) dy = J (t 0 cp(x)) v(x)dx
o 0
for all I E~1 ([0,

~]...A)., A).
(b) If v is any function in ~1 ([0, 1], ..A)., A) for which (ii) holds, prove
that (i) holds for v.
(c) Infer that the function w in Theorem (2004) is never unique if cp
fails to be monotone.
(d) Is the function Icp'l of (20.5) the only function [a.e. of course] for
which (20.5.i) holds?
Extra complications appear with more complicated mapping functions
cp, as the following exercise shows.
(20.7) Exercise. Let "I' be the function of period 1 on R such that
1 1
'P (t) = t for 0 ~ t ~"2 and "I' (t) = 1 - t for "2 ~ t ~ 1. Define the
function cp on [0, 1] by the following rules:
for 2-1 ~ t ~ 1, cp (t) = 2-1 "1' (2t) ;
for 2-2~ t ~ 2-1 , cp(t) = 2-3 1p(23 t);
for 2-3~ t ~ 2-2, cp(t) = 2-5 1p(25 t);
cp(O) = O.
[Draw a sketch of the graph of cp.]
(a) Prove the following assertions. The variation Vol cp is equal to 1.
The function cp is absolutely continuous; in fact, cp is in ~ipd[O, 1]). The
derivative cp' assumes only the values ± 1.
(b) Consider cp as a mapping of [0, 1] onto [0, !]
to which Theorem
(2004) may be applied. Find all of the ..A).-measurable functions w on
[0, 1] for which (2004.i) holds.
(20.8) Exercise. Let g be the function constructed in (18040.a) supra.
It is an N-function mapping [0, 1] onto [0, {3], where
{J = max {~ (b ll - all) + ell}~=1 .

Thus Theorem (20.4) can be applied, with g = cp. Since g' does not exist
on F, no function w for which (20.4.i) holds can possibly be the derivative
of g. Find all ..A).-measurable functions on [0, 1] for which (20.4.i) holds
for cp = g.
(20.9) Note. There are obvious questions concerning classical trans-
formations of integrals other than those treated in (20.4) and (20.5). For
example, for a continuous mapping cp of Rn into Rn, the function w of
(20.3) is in certain cases the absolute value of the Jacobian of the trans-
formation cp. Lack of space and time compels us to omit this interesting
subject, although its main outlines are clear enough from (20.3) and the
n-dimensional Lebesgue integral to be defined in § 21.
In § 15 we computed the conjugate space for each of the Banach spaces
£p(X, .91, p,) for 1 < P < 00 and arbitrary measure spaces (X, .91, p,), but
omitted all mention of this problem for spaces £1 (X, .91, p,). We now
address ourselves to this computation. The results and the techniques
employed here are quite different from those of § 15: our main tool is the
LEBESGUE-RADON-NIKODYM theorem.
(20.10) Remarks. Let (X, .91, p,) be any measure space. Let g be a
bounded d-measurable function on X and let Lg be defined on
£1 (X, .91, p,) by the rule
(i) Lg (f) = J f g dp, .
x
Obviously Lg is linear on £1 and
ILg (f) I ;;:; Ilgll .. Ilflll
for all f E £1' Hence Lg is a bounded linear functional on £1' If the function
g is tampered with in any way whatever on a set of measure zero to obtain
a new d-measurable function h, then it is clear that Lh = L g • Thus g
need not be bounded to give rise to an element of £1. It turns out in many
important cases that all elements of £1 indeed have the form L g , where g
is "essentially" bounded. We begin by making this idea precise.
(20.11) Definition. Let (X, .91, p,) be a measure space. A set A E .91 is
said to be locally p,-null if p, (A n E) = 0 for all E E .91 such that p, (E) < 00
[d. (9.29)]. An d-measurable function g on X is said to be essentially
bounded if for some real number a;;;: 0 the set {x EX: Ig(x)1 > a} is
locally p,-null. Let Ilglloo be the infimum of the set of all such numbers a.
The number Ilglloo is called the essential supremum of Igl. This number is
sometimes denoted by the symbols "ess sup Igl" or "vrai max Igl".
(20.12) Theorem. If (X, .91, p,) is a a-finite measure space and A Ed,
then A is locally ",,-mtll if and only if "" (A) = O.
Proof. Trivial.
(20.13) Theorem. Let (X, .91, p,) be any measure space. Suppose that
t E£1 (X, .91, p,) and that g is essentially bounded. Then:
(i) the set {x EX: If (x) I > O} is a-finite;
(ii) Ig E~1;
and
(iii) I Jig d",1 ~ Ilgll eo 11/111'
x
Proof. Let E = {x EX: If(x) I > O} and En = {x EX: I/(x)1 ;;;; ~} for
eo
U En and", (En) <
n = 1, 2, . . . . Obviously E = 11=1 for all n. Thus (i) 00
holds. Let a be any nonnegative real number such that the set
A = {x EX: Ig(x)J > a}
is locally ",-null. By (12.22), we have
J I/gld", = J Ilgld", + lim J I/gld", +EnA'
J Itgld",
x E' ....... eo E"nA
~ 0 + 0 + aEnA'
Jill d",
~ a Jill d", = a 11/111'

x
Taking the infimum over all such a, we obtain (ii) and (iii). 0
(20.14) Theorem. Let (X,d, "') be any measure space. Let g be a
complex-valued, .9I-measurable lunction on X.
(i) II g is essentially bounded, then the set {x EX: Ig (x) I > Ilglleo} is
locally ",-null, i. e., the infimum in Definition (20.11) is attained.
(ii) The lunction g is essentially bounded il and only il there is a
bounded .9I-measurable lunction q; on X such that the set {x EX: g (x) =!= q; (x)}
is locally ",-null.
(iii) II g is essentially bounded, then Ilgll eo = in/{11 q;llu: q; is as in (iin,
and this infimum is attained.
Let ~eo (X, .91, "') denote the set 01 all essentially bounded .9I-measurable
lunctions on X, two lunctions being identified il they differ only on a locally
",-null set. Under pointwise linear operations and the norm I Ileo' ~eo is a
complex Banach space.
Proof. Assertion (i) follows from the evident fact that any countable
union of locally ",-null sets is locally ",-null. Assertion (ii) is obvious.
To prove (iii), assume that there is a function q; as in (ii) such that
~ q;llu < Ilglloo' Then {x EX: Ig (x) I > I q;llu} is not locally ",-null and
g (x) =!= q; (x) at all points of this set. This contradicts the choice of q;,
and so 11q;llu;;;; Ilglleo for all q; as in (ii). Let B = {x EX: Ig(x)1 ~ Ilglleo}'
Then g~B is bounded, .9I-measurable, and equals g except on a locally
",-null subset of B'. Also, Ilg~Bllu = Ilglleo [if B = 0, then Ilglleo = 0]. This
proves (iii).
To prove that ~eo is a Banach space, first notice that (ii) shows that
each equivalence class in ~eo contains a bounded function. Thus we may
regard ~eo [as a linear space] as being the quotient space r:B/m, where liB
is the linear space of all bounded, complex-valued, d-measurable

functions on X and mis the linear subspace of all of those functions in sa
that vanish except on a locally ,u-null set. It is clear that sa, with the
m
uniform norm, is a Banach space and that is closed in sa. Assertion (iii)
shows that" II"" is the quotient norm on sajm [see (14.38)]. Thus (14.38.b)
shows that ~"" is a Banach space. 0
The reader will recall from (15.14.b) that, for certain measure spaces,
if p > 1 and g is a measurable function such that I g E~l for all f E~p,
then g E~p'. The proof of this nontrivial fact suggested in (15.14) requires
the uniform boundedness principle. The corresponding result for p = 1
and p' = 00 is considerably easier to prove, as we shall now see.
(20.15) Theorem. Let (X, d,,u) be a measure space and let g be a
complex-valued, d -measurable lunction defined on X such that J IIgl d,u < 00
lor all I E~l(X, d, ,u). Then g is in ~",,(X, d, ,u). x
Proof. Assume that g ~ ~oo. Then there exist sequences of positive real
numbers (IX,.) and of d-measurable sets (A,.) such that:
"" 1
1X1 < 1X2<···<1Xn<··· and };-<oo· (1)
IX '
.. =1 "
o <,u(A,.) <00; (2)
and
IX,. < Ig (x) I ~ IX,.+! for xEA,.. (3)
To see this, let ({J,.) be any sequence satisfying the conditions in (1). The
assumption that g ~ ~oo implies that for each n the set
{x EX: Ig (x) I > "" {x EX: {J,. < Ig(x)i ;:;;; {J"+i}
{J,.} = i~1
is not locally ,u-null, and so for each n there exists a j such that
{x EX: {J,. < Ig (x) I ;:;;; {J,.+i} has a subset of finite positive ,u-measure.
Define the subsequence (1Xn) of ({J,.) by letting IXI = {Jl' and, having
defined 1Xn, letting 1Xn+1 be the smallest value of {J such that
{x EX: IXn < Ig(x)i ;:;;; 1Xn+1}
has a subset A,. of finite positive ,u-measure. This establishes (1) -(3).
Let
"" 1
I = ..~ IXnI'(A,,) ~.A".
1
J Id,u =
CD
Then we have }; - < 00, and so I E~. But we also have
x .. =1 IXn
jllgld,u = ..~
"" 1
IXnI'(A,,) 1lgld,u
1
J IXnd,u =
CD
~ }; (A) 00. 0
.. =1 IXnI' .. A"
We now return to our functionals Lg on ~l'

(20.16) Theorem. Let (X, d, f-t) be any measure space and let g be an
element 01 ~oo (X, d, f-t). Define Lg on ~l (X, d, f-t) by the rule
Lg(l) = J Igdf-t.
x
Then Lg is a bounded linear lunctional on ~l(X, d, f-t) and IILgl1 = Ilglloo'
Proof. Clearly Lg is linear; (20.13) shows that
for all I E~l' Thus Lg is a bounded linear functional on ~l and

(1)
It remains only to prove the reverse of this inequality. If Ilglloo = 0, this is
obvious. Thus suppose that Ilgll > 0 and let e be an arbitrary real number
00
such that 0< e < Ilglloo' Then the set {x EX: Ig(x)1 > Ilglloo - e} is not
locally f-t-null, and so it has a subset E Ed such that 0 < f-t (E) < 00.
Define
1
1= p,(E) ;E sgn(g) .
It is obvious that I E~l and that 11/111 ~ 1. Also, we have
_ 1
Lg (I) = jig df-t = fL (E) j Igl df-t
~ Ilglloo - e . (2)
It follows from (1), (2), and the definition of IILgl1 that IILgl1 = Ilglloo' 0
It is tempting to conjecture that every bounded linear functional on
an ~l space has the form Lg for some g E~oo' This, however, is not the
case for all measure spaces, as the following example shows.
(20.17) Example. Let 1= [0, 1] and let X be the unit square I x I.
Let d be the a-algebra of Borel sets in X [usual topology] and define v
and f-t on d by
v(E) = L: A({y EI: (x, y) EE})
xEI
and
f-t(E) = v(E) + L: A({x EI: (x,y) EE}).
rEI
[This construction is very like the construction in (I9.71.d).] Define Lon
~l (X,d, f-t) by
L(I) = J Idv.
x
Since v ~ f-t, it is plain that L E ~T (X, d, f-t) and that IILII ~ 1. Assume
that there exists a function g E ~oo (X, d, f-t) such that L = Lg. For each
fixed y E I, set H)' = {(x, y) : x E I} and 1= ;Hysgn(g). Then I is in
~1 (X, d, p) and we have

1
f Ig(x,y)ldx = fig dp = L,(f)
o x
= L(f) = f Id'jl
x
=0.
Since y EI is arbitrary, it follows that for each y EI we have g (x, y) = 0
for A-almost all x. Anticipating FUBINI'S theorem (21.12) [which of
course depends in no way upon this example], it follows that there is a
1
point Xu EI such that f Ig(xo,y)1 dy = O. [In fact, A.-almost all x have
o
this property.] Now let V = {(xu, y) : y EI} and let I = ey. Then I is in
~dX, d, p), and we have
1 = 'jI(V) = f I d'jl = L(f) = L,(f) = fig dp

x x
1
=fgdp~ fg(xu,y)dy=O.
v 0
This contradiction shows that no such g exists.

This example notwithstanding, it is true that for decomposable
measure spaces (19.25), every L E~r is an L, for some g E~oo' To prove
this fact, we need the following lemma.
(20.18) Lemma. Let (X, d, p) be a decomposable measure space
and let ofF be a decomposition 01 (X, d, p) [see (19.25)). Suppose that
I E~1 (X, d, p). Then there exists a countable sublamily ~ 01 ofF such that
1= 0 p-a.e. on X n (U~)'. Moreover il (Fn):=1 is any enumeration 01 <"C,
then we have
(i) lim
p-+oo
III - t ieF.11
.. =1 1
= 0
and
00
(ii) f I dp = I f Idp .
x ..=IF..
Proof. For n EN, write
En={xEX:I/(x)I>!} and E={xEX:/(x)=t=O}.

00
ThenE = U En and p{En) < 00 for all n EN. Let~n={FEofF:p(FnEn)

.. =1
> O} and let An= En n (U~n)" Since p(En) = I p(F n En) (19.25.iii),
F~
co
it follows that <"Cn is a countable family and p (An) = O. Let <"C = ..U
=1
<"Cn
co
and A = U An. Then ~ is countable, p(A) = 0, and I vanishes except
..=1
on A U (U~). Now let (Fn):'=1 be as above. It is clear that
~ n~ I ~F. = 1ft-a. e. and II - 1;1 I ~F.I ~ III E ~l for all PEN. Thus
(i) and (ii) follow from LEBESGUE'S theorem on dominated convergence. 0
(20.19) Theorem. Let (X,.9I, ft) be a decomposable measure space
(19.25). Suppose that L is a bounded linear lunctional on ~l (X, .91, ft).
Then there exists a lunction g E~<XI (X, .91, ft) such that L = L, as in
(20.16).
Proof. (I) Suppose that ft (X) < 00. For E E .91, define
(1)
We claim that'll is a complex measure on (X, .91) such that 'II ~ ft. Let
<XI
{En}:"'1 be a pairwise disjoint family of sets in d, let E = U En' and let

n=1
<XI <XI
Fp= U En for each PEN. We haveF1 :::>F2 :::>··· and n Fp= 0; it

n=p+l P=1
follows from (10.15) and the hypothesis ft (X) < 00 that
I'II(E) - n~ 'II (En) = I IL(~E) - LC~ ~E.)I = IL(~Fp)1

~ IILII • II~Fplll = IILII • ft(Fp) -+ 0
as p -+ 00. Therefore
.. =1
and so 'II is count ably additive. If ft (E) = 0, then ~E = 0 in ~l' and

therefore'll (E) = L (~E) = L (0) = 0; hence 'II ~ ft.
We infer from (19.36) that there exists a uniquefunctiongE~l (X,.9I,ft)
such that
'II(E) = J gdft (2)
E
and
1'111 (E) = J Igl dft (3)
E
for all E E.9I. We assert that Igl ~ IILII ft-a.e. To see this, let
A = {x EX: Ig(x)1 > IILII}
and assume that ft (A) > O. From (3) and (12.6) we have
1'111 (A) =
A
J Igl dft > AJ IILII dft = ilL I ft(A) .
Thus there exists a measurable dissection {Al' ... , An} of A such that
n
E I'll (Ai)! > IILllft(A).
;=1
Invoking (1), we obtain

II II II
IILII ft(A) <}; 111 (Ai)l =}; IL(~AJ)I ~ }; IILII '11~AJlll

;=1 ;=1 ;=1
II
= IILII . }; ft (Ai) = IILII ft (A) .

;=1
This contradiction shows that Igl ~ IILII ft-a.e., and so we suppose with
no harm done that Ig(x)1 ~ IILII for all x EX.
It obviously follows from (1) and (2) that
J sg af'
L(s) = (4)
x
for all complex-valued, d-measurable, simple functions s defined on X.
Now let I E~l(X,d,ft), and use (11.35) to select a sequence (Sn):'=1 of
d-measurable simple functions such that ISll ~ IS21 ~ ••• ~ III and
sn(x) ~ I (x) for all x EX. Plainly ISngl ~ IILII '1/1 E ~l and II - snl
~ 21/1 E ~l for all n EN; hence (12.30), (4), and the continuity of Limply
that
This proves the theorem for the case that ft (X) < 00.
(II) We now consider the general case. Let ~ be a decomposition of

(X, d, ft). For each F E~, define ftF on (X, d) by ftF(E) = ft(E n F)
and define LF on ~l(X, d, ftF) by LF(/) = L(/~F)' Plainly (X, d, ftF)
is a finite measure space and LF is a bounded linear functional on
~ (X, d, ftF) such that IILFII ~ IILII. Apply (1): for each F E~, there is
a function gF E ~oo(X, d, ftF) such that IgF (x) I ~ IILFII ~ IILII for all
x EX and
(5)
for all I E ~l(X, d, ftF)' Since ~dX, d, ft) C ~(X, d, ftF), (5) plainly
holds for all I E ~l (X, d, ft).
Next consider the functions gF' All of them are defined everywhere
on X, but the values of gF on F' are of no consequence in determining L F.
In any event, we can [and do] define a function g on all of X by
g(x) = gF(X) if x EF E~.

It is clear that g E ~oo (X, d. ft) [the d-measurability of g follows from
(19.25.iv)] and that Ilglloo~ Ilgllu ~ IILII·
Let I E ~l (X. d. ft) and let {Fn }:'=1 be the countable subfamily of ~
described in (20.18). The continuity of L. (20.18), the boundedness of g.
and (12.30) show that
L(f) = ~~ LC~ /I;F.. ) ~~ = tl

.. L(fI;FJ
1.:I f f1.:P f!;Fngdfl

P
= lim Ig dfl = lim
P-->oo n~ p-->oo n~ 1
FlO X
=jlgdfl' D
x
(20.20) Theorem. Let (X, d, fl) be a decomposable measure space
(19.25). Then the mapping T defined by
T(g) = Lg
[see (20. 16)} is a norm-preserving linear mapping 01 ~oo onto the conjugate
space ~f. Thus, as Banach spaces, ~oo and ~f are isomorphic.
Proof. The fact that T is a norm-preserving mapping from ~oo into
~f is (20.16). It follows from (20.19) that T is onto ~f. It is trivial that
T is linear. Since T is both linear and norm-preserving, it is one-to-one. D
(20.21) Note. As we have shown in (20.17), the conclusion in (20.20)
fails for some nondecomposable measure spaces. However J. SCHWARTZ
has found a representation of ~i" (X, d, fl) for arbitrary (X, d, fl)
[Proc. Amer. Math. Soc. 2 (1951), 270-275J, to which the interested
reader is referred.
let t be an outer measure on f!J (X) as in § 9. Prove that the definitions of
local t-nullity given in (9.29) and in (20.11) are equivalent.
(20.23) Exercise. Let (X, d, fl) be a degenerate measure space such
that fl (X) = 00 [see (to.3) for the definition]. Is this measure space
decomposable? Find ~l> ~f, and ~oo explicitly for this measure space.
(20.24) Exercise. Let (X, d, fl) be any measure space and let
I E~l (X, d, fl)· Define L on ~oo (X, d, fl) by
L(g) = j gfdfl.
x
Prove that L E~! and that IILII = 11/111'
(20.25) Exercise. Prove that ~1([0, IJ) [with Lebesgue measureJ is
not reflexive by showing that not every L E~! ([0, 1J) has the form
described in (20.24). [Hint. Use the Hahn-Banach theorem to produce
an L =!= 0 such that L (g) = 0 for all g E~oo for which g is essentially
continuous, i.e., Ilg - hll oo = 0 for some h E<r([0, IJ).J
(20.26) Exercise. (a) Prove that ~oo([O, IJ) is not separable.
(b) Find necessary and sufficient conditions on a measure space that
its ~oo space be separable. [Do not forget (20.23).J
Having found the conjugate space of ~P (X, d, fl) for 1 < p < 00 and
any measure space (X, d, fl), and of ~l (X, d, fl) for a large class of
measure spaces, we find it natural to ask about the conjugate space of

~oo (X, d, p,). It turns out that each functional in ~~ can be represented
as an integral with respect to a certain bounded, complex-valued,
finitely additive measure on d. We now sketch this representation. First
we need some definitions.
(20.27) Definition. Let (X, d, p,) be a measure space and let
F (X, d, p,) denote the set of all complex-valued functions T defined on d
such that
(i) sup{IT(A)I:A Ed}<oo;
(ii) T(A U B) = T(A) + T(B) if A, B Ed and A nB = 0;
and
(iii) T (A) = 0 if A E d and A is locally p,-null.
For such a Twe define ITI on d [just as in (19.10) and (19.11)J by the rule
ITI (A) = sup{t; IT (A;) I : {AI> ... , An} is a measurable dissection of A}.
ITI EF(X, d, p,) [ef. (19.12)J. We define the norm

It is easy to show that
of Tto be the number IITII = ITI (X). One easily verifies that F(X, d, p,)
is a complex normed linear space with this norm and with setwise linear
operations.
It is an interesting fact that integrals can be defined for d-measur-
able functions on X with respect to finitely additive measures T as defined
in (20.27). [Of course there is no analogue of the limit theorems (12.22)
and (12.24).J We now outline the construction of integrals J ... dT.
m
(20.28) Lemma. Notation is as in (20.27). If f = 1: IX;~AI and
.. ;=1
g = 1: {l/<~BJ; are complex-valued simple functions on X, where {AI' . .. ,Am}
k=l
and {BI' ... , Bn} are measurable dissections of X, then for any T EF (X, d,p,)
we have
(i) !i~ IX;T(A;) - i1 {l/<T(B/<)! ~ IITII . IIf - gil,,·

Proof. We write
jig IX;T(A;) - k~ {lkT(B/<)! = Ii i1

m ..
(IX; - {lk) T(A; n B/<)!
~ 1: 1: \IX; - {l/<\' /T(A; n Bk)\
i=l k=l
m ..
~ 1: 1: \It - gll,,/T(A; n B k )\
i=l k=l
~ IIf - gil" . IITII;
the next to last inequality holds because either Ai n Bk = 0, in which

case the summand is zero, or there is an x EAi n B k , in which case
lexi - I1kl = I/(x) - g(x)1 ~ III - gil .. · 0
(20.29) Definition. Let (X, d, p.) be a measure space and let
m
• EF(X, d, p.). If s = I: exi~Aj is an d-measurable simple function on
;=1
X, we define
m
(i) J s d. = I: exi.(A i) .
;=1
x
If I is a bounded, complex-valued d-measurable function on X, use
(11.35) to obtain a sequence (Sn):'=1 of d-measurable simple functions
on X such that 111- snll .. ~ o. In view of Lemma (20.28), the sequence
(J sn dT ):'=1 is a Cauchy sequence of complex numbers, and we define
x
(ii) J I dT = lim J sn dT .
x """'00 x
It is easy to see that this definition does not depend upon the particular
sequence (sn) [provided, of course, that it converges uniformly to f], and
so the integral is well defined, and definitions (i) and (ii) are consistent.
It is also easy to see that this definition agrees with Definition (19.17) in
the case that. is a complex measure.
(20.30) Theorem. Let (X, d, p.) and. be as in (20.29). Let I and g be
bounded d-measurable lunctions on X and let ex EK. Then
(i) J exl d. = ex J I d. ,
x x
(ii) J (I + g) dT = J I dT + J g dT ,
X X X
and
(iii) I J I d.1 ~ Jill d 11'1 .
x
x
Proof. Exercise.
(20.31) Theorem. Let (X, d, p.) and. be as in (20.29). Let h be a
bounded d -measurable lunction on X such that the set A = {x EX: h (x) =l= O}
is locally p.-null. Then J h d l' = o.
x
Proof. Using (20.30) and (20.27.iii), we find
I J hd.1 ~ J Ihl dl.1 = J ~Alhl dlTI + J ~A'lhl dl.1
x x x x
~ Ilhll .. I.1 (A) + 0 = O. 0
(20.32) Definition. Let (X, d, p.) and • be as in (20.29) and let
g E~oo(X, d,p.). Let Ibe a bounded function in ~oosuch that 11/-glloo=O.
Define
J gdT = J I d•.
x x
23*
In view of (20.31), this definition does not depend on the particular

bounded function f that is drawn from the ~oo-c1ass determined by g, and
so the definition is unambiguous.
(20.33) Theorem. Let (X, d, fl) be any measure space and let
'i EF (X, d, fl). Define LT on ~oo (X, d, fl) by the rule
(i) LT(g) = f g d'i.
x
Then LT is a bounded linear functional on ~oo and
(ii) [[LT[[ = [['ill .
Proof. For g E~oo' choose a bounded f E~oo such that [[f - g[[oo = 0
and [[f[[ .. = [[g[[oo (20.14). By (20.32), we have
[LT (g) [ = [f g d'i[ = [ f f d'i[ ~ f [f[ d ['it ~ [[f[[ .. [['ill
x x X
= [[g[[oo[['i[[ .
Thus LT E~!, and
[[LT[[ ~ [['ill . (1)
Let e > 0 be given and select a measurable dissection {AI> ... , An} of X
such that
n
L ['i(Aj)[ > [['ill - e.
i~l
n
For each j, let l;.j = sgn('i(Aj)) and set g = L l;.j~Ar It is plain that
i~l
g E~oo' that [[g[[oo= [[g[[" ~ 1, and that
[LT (g) [ = [j g d'i[ = \i~ l;.j'i(A j)! = i~ ['i(Aj)[ > [['ill - e.

Since e is arbitrary, this shows that
[[L T [[ ;;;; [['ill . (2)
Now combine (1) and (2) to get (ii). 0
The converse of (20.33) holds: every bounded linear functional on ~oo
is the integral with respect to a finitely additive measure. Plainly we get
this measure by looking at characteristic functions of sets. The details
follow.
(20.34) Theorem. Let (X, d, fl) be any measure space and let L be a
bounded linear functional on ~oo (X, d, fl). Then there exists a'i EF (X,d,fl)
such that L = LT as in (20.33).
Proof. For each A Ed, let
(1)
We have
sup{['i(A)[ : A Ed} ~ sup{[L (g)[ : g E~oo' [[g[[oo ~ I} = [[L[[ .
Thus (20.27.i) holds for -r. Also, if A and B are disjoint sets in .91, then
~A + ~B = ~AUB' and so
-r(A U B) = L(~AUB) = L(~A + ~B) = L(~A) + L(~B) = -r(A) + -r(B);
hence (20.27.ii) holds. Next, if A Ed and A is locally p-null, then ~A =0
in ~ 00 and therefore
-r(A) = L(~A) = L(O) = O.
Thus 120.27.iii) holds, and so -r EF (X, .91, p). Let g be any function in ~oo.
Choose a bounded t E ~oo such that Ilg - 11100= 0 and 11/11 .. = Ilglloo' Choose
a sequence (sn) of d-measurable simple functions such that III - snllu -+ O.
It is clear from (1), the linearity of L, and (20.29.i) that
L(sn)=fsnd-r (2)
x
for each n EN. Since 111- snll 00 -+ 0 and L is continuous on ~oo' (2) and
(20.29.ii) imply that
L(g) = L(f) = lim L(sn) = lim f Sn d-r = f I d-r
n->-OO n->-OO X X
= f g d -r = Lr (g). 0
x
(20.35) Theorem. Let (X, .91, p) be an arbitrary measure space. Then
the mapping T defined by
T(-r) = Lr
[see (20.33)] is a norm-preserving linear mapping 01 F(X, .91, p) onto the
conjugate space ~!. Thus F is a Banach space and, as Banach spaces,
F and ~! are isomorphic.
Proof. The fact that T is a norm-preserving mapping from F into~!
is (20.33). It is trivial that T is linear. Theorem (20.34) shows that Tis
onto ~!. Being both linear and norm-preserving, T is one-to-one and
preserves Cauchy sequences. Since ~t, is complete, so is F. 0
(20.36) Remark. Let X be an arbitrary nonvoid set. As in (7.3), let
~ (X) denote the space of all bounded, complex-valued functions on X.
This space has several other names: if X is regarded as a topological space
with the discrete topology, then ~ (X) = <r (X) (7.8); if P is the counting
measure defined on gJ(X) (lOA.a), then ~(X) = ~oo(X, gJ(X), p); and
in (14.26) this space was denoted by 100 (X) [the conjugate space of 11 (X)].
In all cases the norm used on ~ (X) has been the uniform norm. Thus
Theorem (20.35) shows that the conjugate space ~ (X)* of the Banach
space ~ (X) is isometrically isomorphic to the space of all bounded,
complex-valued, finitely additive measures defined on gJ(X).
(20.37) Exercise. Let X be a nonvoid set. Suppose that -r is a finitely
additive measure defined on gJ (X) such that -r (X) = 1 and -r (A) = 0 or 1
for all A eX. Let 0/1 = {A : A eX, -r(A) = I}.
(a) Prove that 1ft has the following properties:

(i) 0 ~ 1ft;
(ii) if A Elft and A c B C X, then B Eo/!;
(iii) if A, B EIft, then An B EIft;
(iv) if A C X, then A E1ft or A' E1ft.
Any family 1ft that satisfies (i) -(iii) is called a filter in X. A filter in X
satisfying (iv) is called an ultrafilter in X.
(b) Prove that if ny is any ultrafilter in X and if a is defined on
9(X) by
I if A Eny,
{
a(A) = 0 if A ~ny,
then a is a finitely additive measure on (X, 9 (X)).
Thus we have set up a one-to-one correspondence between ultra-
filters and finitely additive zero-one measures.
(c) Let (X, d, p,) be any measure space and let or be a finitely additive
measure in F(X, d, p,). Prove that
Jig dor = J I dor' J g dor
x x x
for alII, g E~oo if and only if or(A) = 0 or 1 for all A Ed.
(20.38) Exercise. Let X be a non void set. A filter 1ft in X is said to be
Iree if no/! = 0. All other filters are said to be fixed. Prove the following.
(a) If 1ft is a fixed ultrafilter in X, then there is a point P EX such that
1ft = {A E9 (X) : PEA}.
(b) If X is finite, then every ultrafilter in X is fixed.
(c) If X is infinite, then there exists a free ultrafilter in X. [Let
F = {A E9(X) : A' is finite}. Prove that F is a filter. Use ZORN'S
lemma to show that there is a maximal filter ny containing F. Prove that
ny is a free ultrafilter.]
(d) If ny is a free ultrafilter in N and a is as in (20.37), then a (F) = 0
for all finite sets FeN, and so a is not count ably additive.
(e) If a is as in (d), then for alII Elr., we have
lim I(n) ~ J Ida ~ fiiii I(n).
n-+oo N n-+oo
[First consider the case that lim I(n) = O. In general, find functions
11->-00
[sequences!] g and h such that g ~ I ~ h, lim I(n) = lim g(n) , and

~oo 11-+00
lIm I(n) = lim h(n).]

"""""00 11-+00
(20.39) Exercise. Prove that there exists a nonnegative, real-valued,

finitely additive measure or defined on 9([0,1]) such that or(A) = A(A)
for all A E.A). for which A C [0, I]. [Hints. Define L on {I : I is a bounded,
real-valued, Lebesgue measurable function on [0, In by L (f) =
o
1
J I (x) dx.
Use KREIN'S extension theorem (14.27) to extend L to a nonnegative

linear functional on the space mr([O, IJ) of all bounded real-valued func-
tions on [0, 1].]
(20.40) Exercise. (a) Prove that there exists a linear functional M
m
on m1' (R) such that for all I E r (R) and all t ER, we have
(i) inf{/(x): x ER} ~ M(f) ~ sup{/(x): x ER}
and
(ii)
[Hints. Let ~ be the linear subspace of m1' (R) consisting of all finite sums
of functions of the form It - I for IE mr(R) and t ER. For
..
h = 1: [(fk)tk - IkJE ~ ,
k=I
prove that inf{h (x) : x ER} ~ 0. Assuming that this is false, choose
°
e > such that h (x) ~ e for all x E R. Let p be an arbitrary positive
integer, and let (/J denote the set of all functions cp from {I, 2, ... , n} into
{I, 2, ... , P}. Clearly ($ = pIt. For each cp ((/J, let x (cp) = cp (1) tl + cp (2) t2
+ ... + cp(n)t". Show that
1: [Mx(cp) + tk ) - Mx(cp)] ~ 2P"-11IM"
'PE<l>
for each k E{I, 2, ... , n}, and then conclude that

..
pn e ~ 1: hex (cp» ~ 2P"-11: 11M ...
q;E<l> k=I
This contradiction shows that no h E ~ has a positive lower bound. Next

use (14.13) to obtain ME Q31'(R) * such that M (1) = 1, IIMII = 1, and
M (h) = 0 for all hE fl. Now (i) and (ii) follow easily.]
(b) Prove that there exists a [nonnegative, extended real-valued,
finitely additive measure ft defined on fJJ (R) such that
(iii) ft(A) = A(A) for all A E~
and
(iv) ft(A + t) = ft(A) for all A C R and all t ER.
[Hints. Let T be as in (20.39). Define 11 on fJJ (R) by
00
1I(A)= 1: T((An[n,n+l[)-n).
1$=-00
Show that 11 is finitely additive and that 11 (A) = A(A) for all A E~.
For A C R, define IA on R by the rule IA(t) = 1I(A + t). Let M be as in
part (a) and define ft on fJJ(R) by the rule
ft(A) = lim M(min{IA' 1~}).
~oo
1 As usual, I. denotes the translate of I by t : I. (x) = I (x + t) .

It is easy to prove (iii). To prove that I-' is additive, use the inequalities
min {fA + fB' n} ~ min {fA, n} + min {fB' n} ~ min {fA + fB' 2n}. To prove
(iv), use the equality !cHt) (x) = fA (x + t).]
Our third application of the LEBESGUE-RADON-NIKODYM theorem is
to the study of yet another conjugate space. In (12.36), we saw that if X
is a locally compact Hausdorff space, then every nonnegative linear
functional I on <E:oo(X) has the form I (I) = J f dt for some regular Borel
x
measure ton X. This fact is useful in identifying the conjugate space of the
complex Banach space <E:o(X).
(20.41) Definition. Let X be a locally compact Hausdorff space.
A complex measure I-' defined on the a-algebra !Ji (X) of Borel sets of X
is said to be a complex regular Borel measure on X if for each E E!Ji (X)
and each e > 0 there exist a compact set F and an open set U such that
FeE c U and II-' (A)I < e for all A E!Ji(X) such that A c F'. Let un
M (X) denote the set of all complex regular Borel measures on X.
(20.42) Note. If I-' EM (X) and I-' ~ 0, then, since all complex measures
are bounded (19.13.v), I-' is a finite measure. Theorem (12.40) thus shows
that the definitions of regularity of I-' given in (20.41) and in (12.39) are
equivalent.
1-', 'II EM (X), and let IX (K. Then, operations being defined setwise, we have
(i) IXI-' EM(X)
and
(ii) I-' + 'II EM(X).
Thus M(X), with the norm defined by 1/1-'1/ = 11-'1 (X), is a complex normed
linear space.
Proof. Exercise.
(20.44) Theorem. Let X be a locally compact Hausdorff space, let I-' be
4
a complex measure defined on !Ji (X), and let I lXi I-'i be the Jordan decom-
i=l
position of I-' as in (19.15.c). Then the following three statements are pairwise
equivalent:
(i) I-' EM(X);
(ii) II-' I EM(X);
(iii) I-'i EM(X) for j = 1,2,3,4.
Proof. We make heavy use of Theorem (19.13). The fact that (i) implies
(ii) follows from the inequality
II-'I(U n F') ~ 4· sup {II-' (A)I : A E!Ji(X), A c un F'}.
The inequality l-'i(U n F') ~ II-'I(u n F') (j = 1, 2, 3, 4) shows that
4
(ii) implies (iii). From the inequality 1.u(A)1 ~ E .ui(A), we infer that
;=1
(iii) implies (i). 0
.u EM (X), and define L,. on (£:o (X) by the rule
(i) J fd.u.
L,.(I) =
x
Then L,. is a bounded linear functional on the Banach space {£:o (X) and
(ii) IIL,.II = 11.u11 .
Proof. Each f E (£:o is a bounded continuous function and so, since l.u I
is a finite measure (19.13.v), f is in ~l(X, &6'(X), l.ul). Thus L,. is a linear
functional on {£:o' Also (19.38.iv) shows that if f E {£:o, then
Thus L,. E (£:: and

(1)
To prove the reverse of inequality (1), use (19.38) to obtain a complex-
valued Borel measurable function fo on X such that Ifo (x) I = 1 for all
x EX and
(2)
for all f E ~l(X, &6'(X), l.ul). Let e > 0 be arbitrary. By (13.21), there
exists a function g E (£:oo such that Ilgllu ~ Illoliu = 1 and
x
J 110 - g I dl.ul < e . (3)
Now (2) and (3) imply that

11.u11 = 1.uI(X) = J 'ofodl.ul = xJ ' od.u < IxJ g d.ul + e =
x
IL,. (g) I + e. (4)
Since e is arbitrary and Ilgllu ~ 1, (4) implies that

11.u11 ~ IIL,.II . (5)
Combine (1) and (5) to complete the proof. 0
We will next show that every element of (£:0(X)* has the form L,. for
some.u EM(X). First we need a lemma.
(20.46) Lemma. Let X be a locally compact Hausdorff space and let
L E (£:o(X)*. Define I on {£:ri by the rule
(i) I (I) = sup{IL (g)J : g E {£:o, Igl ~ I} lor alii E {£:ri·
Then I can be extended to a nonnegative linear functional in cr:
such that
11111 = IILII·
Proof. Obviously if I E <rci. then

1(1) ~ sup{!!L!!·!!gll .. :g E<ro. jgj ~ I} = jjLjj·jj/jj .. < 00. (1)
Thus I is real-valued on <rt. It is also clear that
I (rx/) = rxl (I) (2)
for IE <rt and rx~ O. Let us show that! is additive on <rt. Let Iv 12 be in <rt.
If e> Oisgiven. choosegvg2 E<rosuch that Igil ~ Ii and IL(gi)1 > I (Ii) - ;
(j = 1.2). Now write L(gi) = PiIL(gi)l. where Pi EK and IPil = 1
(j = 1.2). and let P = P2 Pl' Then IPI = 1 and we have
/(/1) + 1(12) < IL(gl)1 + jL(g2)1 + e = P1 L (gl) + P2 L (g2) + e
= IPL(gl) + L(g2)1 + e = IL(Pg1 + g2)1 + e ~ /(/1 + 12) + e
[note that IPg1 + g21 ~ Ig]1 + Ig21 ~ 11 + 12J· Thus
/(11) + 1(12) ~ /(11 + 12)' (3)
To prove the reversed inequality. choose g E <ro such that Igl ~ 11 + 12
and IL(g)1 ~ /(/1 + 12) - e. Let hI = min{t1' Ig!} and h2 = Igl- hI' It is
plain that hv h2 E <rt. hI ~ II. h2 ~ 12' and hI + h2 = Igl· Let gi = hisgng
(j = 1.2). Then we have: gi E <ro; Igil = hi ~ Ii; andg1 + g2 = g. Therefore
/(/1 + 12) - e ~ IL(g)1 = IL(gl + g2)1 ~ IL(&)I + IL(g2)1
~ /(11) + 1(/2);
together with (3). this proves that
/(/1 + 12) = /(11) + 1(/2) . (4)
We now extend I to <ro in two steps. First. if I E<r~. write 1= 1+ - r.
where as usual r
= max{/. O} and r
= - min{/. O}. and define
1(/) = I W) - I (n .
Notice that if gl and g2 are any two functions in <rt such that I = gl - g2'
then 1+ + g2 = gl + r.
and so (4) implies that 1(/) = I (gl) - I (g2)'
A simple computation shows that I is real linear on <r~. Next. for I E <roo
define
1(/) = I (Rei) + il (lm/) .
Another obvious computation shows that I is complex linear on <roo
It is now clear [using (i)J that I is a nonnegative linear functional
on <roo Thus. just as in (9.4). for all I E <ro we have
II (I) I ~ I (lfI) .
This fact together with (1) yields
11111 = sup{ll (I) I : I E <roo II/I! .. ~ I} ~ sup{I(I/D : I E <roo 11/11 .. ~ I} ~ IILII·
To prove the reversed inequality, let e> 0 be given and then select
g E<io such that Ilgllu ~ 1 and IL (g) I > IILII - e. From (i) we have
IILII - e < IL(g)1 ~ I(lgi) ~ 11 1 11.
Hence 11111 = IILII. 0
(20.47) Theorem. Let X be a locally compact H ausdortJ space and let
L E<io (X) *. Then there exists a complex measure p, EM (X) such that
(i) L (f) = J I dp, = Lp(f)
x
lor all I E<io (X) .
Proof. Let I be constructed from L as in (20.46). By F. RIEsz's
representation theorem (12.36), there exists a measure £ on X as in § 9
such that
I(f)=Jld£
X
for all I E<ioo (X). It follows that
£(X) = 1(1) = sup {I (f) : I E<ito, I ~ I}
~ sup{II(f)I: I E<io' 1I/IIu ~ I}
= 11111 = IILII·
Thus t is a finite measure. We know (13.21) that <ioo is a dense linear
subspace of ~l (X, J(, t). It follows from (20.46.i) that, relative to the ~l
norm, L is a bounded linear functional of norm ~ 1 on this dense sub-
space: in fact,
IL(f)1 ~ 1(1/1) = J I/ldt = 11/111
x
for all IE <ioo . Thus, by (14.40) [or the Hahn-Banach theorem], L can be
extended to a bounded linear functional L' on ~l (X, J(, t) such that
IIL'II ~ 1. Now apply (20.19) to obtain a function g E~oo (X, J(, t) such
that Ilglloo= IIL'II ~ 1 and
L'(f)=JIgdt
x
for all I E~l (X, J(, t). Next define p, on J( by the rule
p, (E) = J g d t.
E
Then p, is a complex measure on (X, J(), p, is absolutely continuous with

respect to t, and g is the unique LEBESGUE-RADON-NIKODYM derivative
of p, with respect to t as in (19.36). [Actually 1p,1 is t, but we do not need
this fact here.] Now (19.36) implies that if IE <io(X) C ~l (X, J(, Ip,!),
then
J I dp, = Jig d t = L' (I) = L (I) ;
x x
the last equality follows from the fact that Land L' are both continuous
in the ~1 (X, vII., t)-norm on ~o and agree on the ~l-dense subspace ~oo.
All that remains is to show that p, is regular. Let E Evii. and E> 0 be
given. Since t is regular and finite, there exist a compact set F and an open
set U such that FeE c U and t(U n F') < E. Thus if A Evii. and
A C U n F', then
1p,(A)I = If gdtl ~ teA) ~ t(U n F/) < E.
A
Hence p, EM (X) and the proof is complete. 0

(20.48) Riesz Representation Theorem. Let X be a locally compact
H ausdorD space. Then the mapping T defined by
T(p,) =L,.
[see (20.45)} is a norm-preserving linear mapping 01 M(X) onto ~o(X)*.
Thus M(X) is a Banach space, and M(X) and ~o(X)* are isomorphic as
Banach spaces.
Proof. The fact that T is a norm-preserving mapping from M into ~
is (20.45). It follows from (20.47) that T is onto ~. It is trivial that T
is linear. Since T is both linear and norm-preserving, T is one-to-one and
preserves Cauchy sequences. Thus, since ~~ is complete, so is M. 0
let p, EM(X). Let I be the nonnegative linear functional on <ro(X)
constructed from L,. in (20.46). Prove that if t is the measure correspond-
ing to I as in (12.36), then t = 1p,1. [Show that f 1dlp,1 = sup{lf g dp,l:
x x
g E<ro' Igl ~ I} for all 1E~t.]
(20.50) Exercise. Let X be a locally compact Hausdorff space for
which there exists a p, EM(X) such that p,({x}) = 0 for all x EX and
1p,1 (X) > O. Prove that the Banach space ~o(X) is not reflexive. [Define
fP on M(X) by fP(v) = L; v ({x}).]
xEX
(20.51) Exercise. (a) Let X be a locally compact Hausdorff space and
suppose that 7: is a [nonnegative] finitely additive measure defined on
~ (X) such that 7: is regular; i. e., 7: satisfies (i), (ii), and (iii) of Definition
(12.39). Prove that 7: is count ably additive. [Use (12.36) to obtain a
regular Borel measure t on X such that f 1d7: = f 1d t for all 1E<roo.
x x
Then show that 7:(E) = teE) for every Borel set E.]
(b) Prove that any bounded, complex-valued, finitely additive
measure 7: defined on ~(X) which is regular in the sense of (20.41) is
countablyadditive, i.e., 7: EM(X).
In parts (c) and (d), let t be a a-finite regular Borel measure on X.
(c) If 'JI is a complex measure defined on ~(X) such that 'JI ~ t, then
'JI EM(X).
(d) If '/I EM(X) and '/I = '/11 + '/12 is the Lebesgue decomposition of '/I
with respect to t, then '/Ii E M(X) (j = 1,2).
(20.52) Exercise. More on the structure of 'ioo(X) and'io(X). (a) Let
X be a non void locally compact Hausdorff space, and let M be a multi-
plicative linear lunctional on <ro(X). By this we mean that M satisfies
(9.l.i) and (9.l.ii), that M =1= 0, and that M (I g) = M (I) M (g) for all
I, g E<ro(X). Prove that there is a point a EX such that M (I) = Ea(l) = I (a)
for all IE <ro(X). [Hints. It is convenient to prove first that M is con-
tinuous. In fact, we have 1M (1)1 ~ 1I/IIu for all I E <ro. To see this, assume that
00
M(I) = IX IIXI> 11/11u. Let g = - L IX-kfk. The series converges
where
k=l
uniformly and so g E<ro. Check that IX-I I + g - IX-I I g = O. Then one has
0= M(O) = IX-1M(f) + M(g) - IX-IM(I)M(g) = 1 + M(g) - M(g) = 1.
Now knowing that M is bounded, we apply (20.47) to write
M(f) = J Idl-' = J Igdt
x x
where I-'EM(X), tEM+(X), gE~oo(X,&#(X),t), and Ilglloo~ 1. If
A, B E &# (X) and t (A) and t (B) are positive, then A n B =1= 0. This
follows readily from the regularity of t and the identity M (lIi2)
=M(fl)M(f2)' Now it is easy to see that t assumes only one positive
value, since teA) = t(A n B') + t{A n B) and t(B) = teA' n B) +
teA n B). The family {F eX: F is compact, t(F) > O} is nonvoid and
has the finite intersection property and hence has non void total inter-
section, say Fa. It is not hard to see that t (Fa) > 0 and that Fo has no
proper nonvoid subsets, i.e., Fa = {a} for some a EX. It follows immediately
that M(f) = I(a).]
(b) Prove that every multiplicative linear functional M on <roo (X)
has the form Ea for some a EX. [Hints. Choose 10 E <roo such that
M (to) = 1. Let U = {x EX: lo(x) =1= O}. Let 5' = {I E<roo (X) : I(U') c to}}.
Then 5' can be identified in an obvious way with <r0(U) [note that U is
locally compact]. The functional M is multiplicative and nonzero on 5'
and so part (a) shows that M(f) = I (a) for all I E5', a being a point in U.
Given g E <roo (X), the function glo is in 5' and M (glo) = M (g). Since
lo(a) = 1, the proof is complete.]
(c) Generalize part (a) as follows. Let (X, .5#, 1-') be a finite measure
space such that I-' (X) > O. Prove that Jig d I-' = J I d I-' J g d I-' for all
I E~l (X, .5#, 1-') if and only if I-' assumes only the values 0 and 1. [Compare
this with (20 37.c).]
(d) Part (a) can be interpreted in the following way. A lelt ideal in an
algebra A is a linear subspace I that is closed under left multiplication
by arbitrary elements of A: x E I and yEA imply yx EI. Right ideals
are defined similarly. A set I that is both a left and a right ideal is called
a two-sided ideal. In commutative algebras, the distinction between right
and left ideals disappears, of course, and we use the term "ideal". An
ideal is called maximal if the only ideal containing it properly is the entire
algebra. In the commutative Banach algebra <ro(X), every set
{f E <ro (X) : I (a) = O} is a closed maximal ideal. Furthermore, every
maximal ideal [not assumed a priori to be closed] in <ro(X) has this form.
[All assertions are simple to verify except the last, for which the following
hints are offered. If S is an ideal in <ro (X) and if there is a point a E X such
that I (a) = 0 for all f E S, then S can be maximal only if S = {I E<ro (X):
I (b) = O} for some b EX. If S is an ideal such that for all a E X, I (a) =f= 0
for some IE S, then a simple compactness argument, and the fact that
11 E S if IE S, show that S ~ <roo (X). Consider the algebra = <ro(X)jS. e
It is a linear space over K and also an algebra with no proper ideals at all.
This implies that e is a field or is K as a linear space with all products
equal to o. Let -r denote the canonical mapping of <ro (X) onto e:
-r (I) = I + S for all I E <ro (X). If e is a field, let h be a function in <ro (X)
such that -r(h) is the multiplicative unit of e. Since S contains <roo (X) ,
00
there is an element q; E S such that Ilh - q;11 .. < 1. Let 1p = - 1: (h- q;)k.
k=1
Then as in part (a), one can show that
h = q; - 1p q; + 1ph - 1p .
Clearly q; - 1pq; is in S, and 1ph - 1p is in S because
-r(1ph - 1p) = -r(1p) -r(h) - -r(h) = -r(h) - -r(h) = 0.
Hence h is in S, and so -r(h) = 0, a contradiction. If e is not a field, then
e
all products in are 0, and in particular, if IE <rt, then 1= f2 . l2- is in S.
1 1
From this it follows at once that S = <ro.]

(f) Part (e) admits the following extension. Let Bbe any closed proper
ideal in <ro(X). Then there is a nonvoid closed subset F of X such that
B = {I E <ro(X) : I (F) = {On. [Hint. Use (6.80).]
(20.53) Discussion. Our last application of the LEBESGUE-RADoN-
NIKODYM theorem deals with sequences of a-algebras on a fixed set and
with measures on them. We first generalize the LEBESGUE-RADoN-
NIKODYM derivative defined in (19.43). Let X be a set, d a a-algebra of
subsets of X, f-l a a-finite measure on d, and 'YJ a a-finite signed measure
on d. Let 'YJ = 'YJa + 'YJ. be the [unique] Lebesgue decomposition of 'YJ with
respect to f-l [see (19.42)]. Let I be an extended real-valued, d-measurable
function on X. Then 1 is said to be a derivative 01 'YJ with respect to f-l if the
following conditions hold:
(i) 'YJa(A) = f I df-l for all A Ed;
A
there exist sets B, P Ed such that
(ii) f-l (B) = ° and l'YJsl (B') = °;
(iii) (P, PI) is a Hahn decomposition for 'fJ.;

(iv) I = 00 on B n P and I = - 00 on B n P'.
It is obvious that this definition agrees with (19.43) in the case that
'fJ 4t,. f-t ['fJ. = 0]. In this case we merely take B = P = 0.
To see that such derivatives exist, let B be any set in d such that (ii)
holds ['fJ• ..L f-t], let (P, P') be any Hahn decomposition for 'fJ. (19.6), and
let 10 be any LEBESGUE-RADoN-NIKODYM derivative of 'fJa with respect to
l
f-t [apply (19.24) to 'fJ! and 'fJ; to obtain It and I~]. Now define I by
fo (x) for x EB',
f(x) = ~ 00 for x EB n P,
for x EB n P'.
Since I = fo f-t-a.e., it is clear that f is a derivative.
The following lemma characterizing derivatives will be very useful.
It has the advantage that it makes no mention of Lebesgue or Hahn
decompositions.
(20.54) Lemma. Let g be an extended real-valued, d-measurable func-
tion on X. For each real number cx, let Gar. = {x EX: g (x) ~ cx} and
Lar. = {x EX: g (x) ~ cx}. The function g is a derivative of fJ with respect to f-t
il and only if the following conditions obtain:
(i) for every real number cx and every A Ed, we have
fJ(Gar. n A) ~ cxf-t(Gar. n A);
(ii) for every real number cx and every A Ed, we have
fJ(Lar. n A) ~ cxf-t(L« n A) .
Proof. Suppose that g is a derivative of fJ with respect to f-t and
let Band P be as in (20.53) [with I replaced by g of course]. Then for
all A Ed and cx ER we have
'fJ(Gar. n A) = 'fJa(Gar. n A) + 'fJ,(Gar. n A)
= J g df-t + 'fJ,(Gar. nAn B)
G",nA
~ CXf-t(Gar. n A) + 'fJs(Gar. nAn B) . (1)

Since g ~ cx> - 00 on Gar., it follows from (20.53.iv) that Gar. n A nB c P,
and so 'fJ.(Gar. nAn B) ~ 0 (19.4). This fact and (1) imply (i). A similar
argument, which we leave to the reader, establishes (ii).
Conversely, suppose that (i) and (ii) hold for the function g. Let I
be any derivative of fJ with respect to f-t as defined in (20.53), and let B
and P be as in (20.53). We first show that g = f f-t-a.e. If this is false,
then a moment's reflection reveals that either:
(a) there exist real numbers cx and f3 and a set FEd such that
0< f-t(F) < 00, IfJ.I(F) = 0, and /(x) ~ cx < f3 ~ g(x) for all x EF;
or
(b) there exist real numbers IX and (3 and a set FEd such that
0< fl(F) < 00, I'I].I(F) = 0, and g(x) ~ IX < (3 ~ t(x) for all x EF.
Assume that (a) holds. By (i) we have
'I] (F) = 'I](G{J nF) ~ (3fl(G{J nF) = (3fl(F)· (2)
We also have
'I] (F) = 'l]a(F) = J t dfl ~ IXfl(F) . (3)
F
An obvious contradiction ensues from (2) and (3), and so (a) must fail.
Likewise (b) fails. Thus g = t fl-a.e., and so (20.53.i) holds for g.
Now make the following definitions: L={xEX:g(x)=-oo};
G = {x EX:g(x) = oo};Bo= B n (L U G); and Po = B n G. We wish to
show that (20.53.ii)-(20.53.iv) hold for B o, Po, and g. Since Bo C B,
it is obvious that fl (Bo) = O. It is also obvious that (20.53.iv) holds. Thus
it suffices to show that 1'1].1 (B~) = 0 and that (Po' Po') is a Hahn de-
composition for '1] •.
Since fl (B) = 0 and 'l]a < fl' we have l'I]al (B) = O. Thus condition (ii)
implies that for all A Ed and IX ER, we have
'I].(L cx nAn B) = 'I](Lcx nAn B) ~ IXfl(L cx nAn B) = O. (4)
It is obvious that
00
L 1 C L Z C··· and U Ln
n=! = G" ,
hence (19.3.ii) and (4) imply that
'fJ.(G' nAn B) = lim 'I].(L n
n-->oo
nAn B) ~ 0
for all A Ed. We conclude that G' nB is a nonpositive set for '1] •.
00
Likewise we use (i) and the facts G_1 C G_ z C ... and U G_n = L'
n=!
to prove that L' n B is a nonnegative set for '1] •. Thus the set L' n G' n B
is both nonpositive and nonnegative for '1]., and so I'I].I(L' n G' n B) =0.
Now B~ = B' U [(L' n G') n BJ and l'I]sl(B') = 0; hence I'I].I(B~) = O.
It remains only to prove that (Po, Po') is a Hahn decomposition for '1] •.
00 00
It is plain that G = n Gn
n=!
and that L =
n=!
n L_n . Thus we may argue as
above to prove that
'I].(G nAn B) ~ 0 and 'fJ.(L nAn B) ~ 0
for all A Ed [this time we use (19.3.iii) and the fact that '1]. is a-finite];
we leave the details to the reader. Therefore G n B is a nonnegative set
for 'fJ. and L n B is a nonpositive set for '1]8' Since Po is equal to B n G,
Po is a nonnegative set for 'fJ •. Since
P; = (B n L' n G') U(L n B) UB'
and since 11']81(B') = 0 and 11']81(B n L' n G') = 0, we see that is a P;

nonpositive set for 1']8' This completes our verification that g is a derivative
of 1'] with respect to f-l if (i) and (ii) hold. D
(20.55) More discussion. The following notation and hypotheses are
fixed throughout (20.55)-20.60). Let X be a set and vIt a a-algebra of
subsets of X. Let f-l be a measure on vIt and 1'] a signed measure on vIt.
Let (J(,.):~l be a sequence of a-algebras of subsets of X such that
00
U J(,. C vIt. We suppose that f-l and 11']1 are a-finite on each of the a-alge-
k~l
bra J(,..
For each n EN, consider the measure spaces (X, J(,., f-l) and (X, J(,.,1']);
that is, we restrict the domains of f-l and 1'] to vltn. We write f-ln for f-l
on J(,. and 1']n for 1'] on vltn- Let fn be a derivative of 1']n with respect to f-ln'
We are concerned with pointwise limits of the sequence of functions
(fn) under two hypotheses concerning (J(,.). Our first theorem is the
following.
(20.56) Theorem. Suppose that
(i) vitI C vlt2 C vita C .. "
00
and write vltw for the smallest a-algebra containing U vltn· Let f-lw and 1']w
n~l
be f-l and 1'], respectively, restricted to vltw' Then both of the functions
(ii) I = lim In and f = lim fn
- n---+oo ~ 00
are derivatives of 1']w with respect to f-lw' Thus: lim fn (x) exists for f-l-almost
,.....,.00
all x EX; f (lim In) df-l =

A ,.....,.00
1']a(A) for all A Evltw,' and the f-lw-singular
part of 1']w is confined to the set {x EX: lim In (x) = ± oo}, as set forth
""""00
in (20.53).
Proof. Let 0( be a real number. In this proof, write L" = {x EX:
f(x) ~ O(} and G,,={xEx:f(x) ~ O(}. In view of (20.54),itsufficesto
prove that
1'](L" n A) ~ O(f-l(L" n A) (1)
and
1'](G" n A) ~ O(f-l(G" n A) (2)
for all A Evltw' [To obtain (20.54.i) and (20.54.ii) from (1) and (2), use
the obvious inclusions
{xEx:f(x) ~ O(}:::>{xEX:t(x) ~ O(}
and
{x EX: I (x) ~ O(}:::> {x EX: f(x) ~ o(}.
and replace A in (1) and (2) by its intersection with the smaller set.]
Let (O(n):~l be a strictly decreasing sequence of real numbers with limit 0(.
For n EN, let

Hn = {x EX: inf{!n+1 (x), In+2(x), ...} < a.a},
lIn,1 = {x EX: In+1(x) <~},
and
Hn,p = {x EX: min {In+1 (x), ... , In+l>-dx)} ~ OCn, In+l>(x) < ocn},
(P ~ 2) .
It is clear that Hn,p E .A(.+I>' that {Hn,P}P=1 is a pairwise disjoint family,
n H...
00 00
that Hn = U Hn,p, and that L .. = Let A be any set in the algebra

1>=1 .. =1
00 00
n o .A(. for some no. The set Hn ' p n A is in .A(.+I>

U .A(., so that A E,.,=fl.
n=1
for n ~ no and p ~ 1. We assemble all of these facts and (20.54) to write
00 00
'fJ(Hn n A) = }; 'fJ(Hn,p n A) =}; 'fJn+l>(Hn,p n A)

1>=1 1>=1
n A) = n A) ,
00
~ }; ocnPn+l>(H.. ,p ocnp(Hn (3)

1>=1
the relations (3) holding for all n ~ no. If i'fJi(A) < 00 and peA) < 00,
then we take the limit as n -+ 00 in (3) to write

'fJ(L .. n A) = lim 'fJ(Hn n A) ~ lim ocnp(Hn n A) = ocp(L .. n A). (4)
n-+co fS-+OD
00
Since i'fJi and p are a-finite on 1}> (4) holds for all A E U.A(. .
.. =1
Let {Fk}k'=1 be a pairwise disjoint family of sets in 11 such that
00
X = U Fk and p and i'fJi are finite on each Fk . Let 11k be the set function on
k=1
00
U .A(. defined by
.. =1
lIk(A) = OCp(Fk n L .. n A) - 'fJ(Fk n L .. n A) .
A routine computation and (4) show that 11k is a countably additive,
00
nonnegative, finite-valued measure on the algebra U.A(., in the sense of
.. =1
00 00
(10.3). Let 11 be the set function}; "k' also defined only on U.A(.. It is
k=1 ..=1
easy to see that 11 is a nonnegative, countably additive, a-finite measure
00
on U.A(.. By (1O.39.c), 11 admits a unique [count ably additive!] exten-
#=1
sion over the a-algebra 1w. All this implies that (4) holds not only for
00
A E U.A(. but for all A E1w. Thus condition (1) is established.
n=1
The proof of (2) is very like the proof of (1), and we leave it to the
reader. 0
The following special result is frequently useful.
(20.57) Corollary. Suppose that rJn ~ fln lor all n EN and that the
equality
(i) lim J In dfl = J I., dfl
1\-700 E E
00
holds for all E in the algebra U Jtk. Then rJ., is absolutely continuous with
k=1
respect to fl.,.
Proof. For each fixed kEN and every E E Jtk , we have
rJ (E) = rJk (E) = (1]k)a (E) = J Ik dflk = J Ik dfl . (1)
E E
For n > k, E is also in Jt.., and so (1) can be extended to
1] (E) = rJn (E) = (rJn)a (E) = J In dfl . (2)
E
Take the limit as n -+ 00 in (2) and apply (i). This yields

rJ(E) = lim
1\-700
J Indfl = EJ I.,dfl·
E
(3)
Since I., is a derivative of rJ., with respect to fl." we have
and from (3) we infer that

rJ (E) = rJ., (E) = (rJ.,)a(E) (4)
for all E E Jtk. Since k is arbitrary, we have proved that rJ., and (rJ.,)a
00
agree on U Jtk • By (10.39.c), rJ., is equal to (rJ.,)a on the entire a-algebra

k=1
Jt.,. 0
(20.58) Remark. Condition (20.57.i) is of course just the condition
that J(lim In) dfl = lim J In dfl. If all of the functions Ifni are
E ~oo _OOE
bounded by a fixed function in ~t (X, Jt, fl), then LEBESGUE'S dominated
convergence theorem (12.24) guarantees (20.57.i). If lim Illn - Imlll = 0
m,n---+oo
and fl (X) < 00, then (13.39) implies that (20.57.i) holds. Other conditions
under which (20.57.i) holds are set down in (13.39).
We have a second limit theorem, like (20.56), but dealing with
descending instead of ascending sequences of a-algebras.
(20.59) Theorem. Suppose that
(i) vI4::) Jt2 ::) • • • ::) Jt.. ::) . . .,
and write ~ for the a-algebra ..n
00
=1
Jt... Let flo and rJo be fl and rJ, respectively,
restricted to the a-algebra Jto. Suppose that flo and IrJol are a-finite. Let f
and 1 be defined as in (20.56.ii). Then both t and 1 are derivatives 01 rJ~
with respect to flo.
24*
t 1
Proof. We note first that and are ~-measurable. To see this, write
hn = k->-oo
lim [min {!n' !n+v ... , !n+k}].
It is plain that hn is ~-measurable, and also that lim hn =
n->-oo
t [d. (6.83)].
-
By (i), all of the functions hm' hm+1' ... are Jim-measurable, and since
t=
-
lim hn +m , we infer that t is Jim-measurable for all nt, i.e., t is Jlo-
~oo - -
measurable. The proof for 1is similar.

We borrow the notation La. and Ga. from the proof of (20.56). It is
clear that La. and Ga. are in~. As in the proof of (20.56), it is sufficient
to prove (20.56.1) and (20.56.2) for all A EJlo• For each real number 0(,
let
Ma. = {x EX: inf{!l(x), !2(X), ... } < a:}
and
H" = {x (X: SUp{/l (x), !2(X), ... } > O(}.
The sets Ma. and Ha. are in ~, but need not be in ~. Suppose that the
inequalities
r;(Ma. n A) ~ a:fl(Ma. n A) (1)
and
r;(Ha. n A) ~ a:fl(Ha. n A) (2)
obtain for all A EJlo. For all e > 0, the inclusions La. C Ma.+e and
Ga. C Ha._. are evident, and so (1) implies that
r;(La. n A) = r;(Ma.+e n La. n A) ~ (a: + e) fl(Ma.+. n L(Z n A)
= (a: + e) fl(La. n A). (3)
Since 1r;1 and fl are a-finite on JlI> a simple argument and (3) imply that
r;(La. n A) ~ a:fl(La. n A) ,
i.e., (20.56.1) holds if (1) holds. Similarly we have
r;(Ga. n A) ~ (a: - e) fl(Ga. n A)
if (2) holds, and (20.56.2) follows from this. Therefore to prove the
present theorem we need only to verify (1) and (2).
To prove (1), we write for each n EN
In = {x EX: min{tl(x), ... , tn(x)} < a:},
In,p = {x EX: tp(x) < a:, !P+1(x) ~ a:, ... , In (x) ~ a:}
(1 ~ P < n),
and
In,n = {x EX: In (x) < a:} .
11
Then In,P is in JIp, {]n,p}p=1 is a pairwise disjoint family, and p'd,dn,p = In.
From the definition of Ip and (20.54), we infer that

n
'YJUn n A) =}; 'YJUn,p n A)
.
P=1
=}; 'YJ({x (X: h(x) ~ IX} n 1n,p n A)
..
P=1
~ }; IXp,({xEX:/p(x) ~ IX} n1n,p n A)
P=1
= IXP,Un n A) . (4)
00
U 1n = MIX. Formula (4) and count-

It is clear that 11 C 12 C ... and that ,,=1
able additivity imply that
'YJ(MIX n A) ~ IXp,(MIX n A) ,
i.e., (1) is verified. The inequality (2) is proved in like manner; we leave
it to the reader. 0
(20.60) Remarks. The limit theorems (20.56) and (20.59) are versions
of what probabilists call martingale theorems. Our treatment is taken
from a paper of SPARRE ANDERSEN and JESSEN [Danske Vid. Selsk.
Mat.-Fys. Medd. 25 (1948), Nr. 5J. The interested reader may also consult
J. L. DOOB'S treatise Stochastic Processes [John Wiley and Sons, New
York, 1953J, Ch. VII. We have included these limit theorems primarily
for their applications to infinite products of measure spaces [see § 22J.
Several interesting and unexpected results follow immediately from them,
however, and we shall now point out a few of these in the form of
exercises with copious hints.
(20.61) Exercise: Differentiation on a net. Let (X, d, p,) be a a-fi-
nite measure space. Consider a sequence (,#,;):'=1 of subfamilies of d
having the following properties.
(i) The sets in each .#,; are pairwise disjoint and U.#,; = X.
(ii) If E E.#,;, then 0 < p, (E) < 00.
(iii) For each E E.#,;, E = U{F: F ('#';+1' FeE}.
Such a sequence (,#,;):'=1 is called a net. Let ~ be the a-algebra
generated by.#,;, i.e., the family of all unions of sets in.#,;.
(a) Let 'YJ be a a-finite signed measure on d. For each n EN, let
In -- ~
""
1/{E) I: .
IL{E) 'iiE
EE.¥" t'"
Prove that lim

~oo
In (x) exists for p,-almost all x EX and in fact that lim
11-+00
In
is a derivative of 'YJ with respect to p, for the smallest a-algebra containing
all.#,;. [From (ii) we see that each.#,; is countable. Also if E E~
and p, (E) = 0, then E = 0. The relation I'YJn I <{ p,n is therefore trivially
satisfied, and the function In is plainly the LEBESGUE-RADON-NIKODYM
derivative of 'YJn with respect to I'n' Now apply (20.56) and the definition
of derivative in (20.53).J
(b) Let cp be a Lebesgue measurable function on R such that
a
f Icpl dA < 00 for all a> O. For every x ER, let ](n, x) be the interval
-a
[k . 2- n , (k + 1) . 2-n [ that contains x (n EN, k EZ). Prove that
lim 2n J cp d'A = cp(x) for A-almost all x ER. [Hint. Apply part (a)
........ 00 ](n,x)
with X = R, d = ~, ~ = {[k2- n, (k + I)2- n [}k'=_00, 'YJ (A) = J cp dJ....J
A
Compare this result with (18.3).
(c) Let 'YJ be a a-finite signed measure or a complex measure on
PA (R) such that I'YJI and 'A are mutually singular. Prove that
lim 2n 'YJ(](n, x)) = 0 for 'A-almost all x ER. For 1] a signed measure,
........00
prove that lim 2n 'YJ(](n, x))
........ 00
= ±oo on a set E such that I'YJI (E') = O.
[Apply part (b), (20.56), and (20.53.iv).J Compare this with the behavior
of (X's described in (19.60).
(20.62) Exercise: Densities. Notation is as in (20.61). Let E be any
set in d. Let 'YJ be the measure on d such that
I'(A) = I'(E n A) forall A Ed.
Plainly 'YJn ~ I'n; let Wn be a LEBESGUE-RADoN-NIKODYM derivative of
'YJn with respect to I'n [here n E {I, 2, ... , w}]. The function 'Ze'n is called
a density 0/ E with respect to JI...
(a) Prove that lim Wn (x) = Wro (x) for I'-almost all x EX. [This is
....... 00
a direct application of (20.56).J
(b) Suppose that each Wn is a constant real function I'-almost every-
where, in addition to being JI..-measurable. Suppose also that there is a
set D Evitro such that I' (D 6 E) = O. Then I' (E) = 0 or I' (E') = O.
[Plainly lim Wn is a constant I'-almost everywhere. A simple argument
........ 00
shows that Wro is equal to ~D I'-almost everywhere. Thus ~D is a constant
I'-a.e. J
(c) Consider the space X = [0, I [, and for each n EN, let
~ = {[k 2- n , (k + I)2-n[: k E{O, 1, ... , 2 n - In.
Let JI.. be the [obviously finite J algebra of sets generated by ~. Let
vIt = ~([O, I[). Let P be the Cantor-like set in [0, I[ obtained by re-
moving middle one-quarters at each step (6.62). Determine 'A(P). Let I'
be A and let 'YJ(A) = 'A(A n Pl. Compute Wn as defined above and find
lim W n •
>1---+00
(20.63) Exercise: Application of (20.59). Suppose in (20.59) that

I' (X) = 1 and that 1'0 assumes only the values 0 and 1 on the a-algebra vito.
(a) Prove that lim In is a constant ,u-almost everywhere. [Hints .

...-..00
Consider the Lebesgue decomposition 'YJo = 'YJOa + 'YJ08' Let Bo E ~ be
°
a set such that ,uo(Bo) = and 1'YJ081 (B~) = 0. Let g be a LEBESGUE-
RADON-NIKODYM derivative of 'YJOa with respect to ,uo' Then for every
E E .Ao, we have
'YJoa(E) = f g d,uo = f g d,uo·
°
E EnB,'
Since ,uo assumes only the values and 1, it follows from (12.60) that
there is a number IX for which ,uo({x EX: g(x) = IX}) = 1 and so lim In = IX
...-..00
,uo-almost everywhere on B~ and so ,uo-almost everywhere on X.]
(b) Prove that the IX of part (a) is equal to f 11 (x) d,ul (x) = 'YJla (X),
x
where 'YJla is the ,ul-absolutely continuous part of 'YJl'
(20.64) Exercise. Let X = [0, 1[. For each n EN, let ..II.. = {A eX:
2"-\
there is an .A).-measurable set Be [0,2- n [ such that A = U (B + k2-n)}.
k=O
Note that ..II.. is a a-algebra and that .AI ~ .A2 ~ ... ~ ..II.. ~ ....
Prove that if A E.Ao = ""
n..ll..,
,,=\
then A(A) = °
or A(A) = 1. [Hints. Apply
(20.62.b) to the set A, and use the increasing algebras of sets described in
(20.62.c). It is evident that each Wn is a constant, and so A(A) = or °
A([O, 1[ n A') = O.J
(20.65) Exercise [JESSENJ. Let I be any function in ~d[O, 1[,.A)., A)
and extend lover R by the definition I (x + k) = t (x) for all k (Z and
x E [0, 1[. Prove that
(i) ~~ 2-n [2r/ +

(X k2- n)] = i I(t) dt
for A-almost all x E R. [Hints. Let ..II.. and .Ao be as in (20.64). Let In
be the function on the left side of (i). On the domain [0, 1 [, In is plainly
.An-measurable, and is the LEBESGUE-RADON-NIKODYM derivative of the
measure 'YJn' where 'YJ(A) = f IdA for A E.Al . By (20.59), In converges
A
A-almost everywhere to the LEBESGUE-RADON-NIKODYM derivative 10
\ \
of 'YJo with respect to A.a. Since 'YJo([O, 1[) = flo dAo = f t dA, and since
o 0
1
(20.64) and (20.63) hold, we have lim In (x) = f I dA A-a.e. on [0, 1[;
""-"00 0
and by periodicity, A-a.e. on R.]
(20.66) Exercise: A martingale theorem. Let X,.A,..II..,.AU) and ,u
be as in (20.56) and suppose that ,u(X) < 00. Suppose that (fn)~\ is a
sequence of functions on X each with values in [0, oo[ such that:
(i) In is .An-measurable for all n EN;
(ii) J In df-t = J In+1 df-t for all A E.,A;. and all n EN;
A A
(iii) J In df-t = 1 for all n EN.
x
00
U .,A;.
(a) Prove that there is a finitely additive measure 'YJ on ,,=1
such that
(iv) 'YJ (A) = J In df-t for all A E.,A;. and all n EN.
A
(b) Prove by giving an example that n need not be countably ad-
00
U .,A;..
ditive on ,,=1
(c) Prove that lim In exists and is finite f-t-almost everywhere on X.
"-.00 00
[Hints. Consider the set Q of all finitely additive measures w on "~1 .,A;.
that assume the values 0 and 1 and no other values and vanish
for f-t-null sets. Make Q into a topological space by neighborhoods
00
LI.A = {w EQ: w(A) = I}, for A E,,=1 U vIIn . Then Q is a compact Haus-
dorff space, and f-t and 'YJ can be transferred to Q. Apply (20.56) appro-
priately and go back to X.]
CHAPTER SIX
Integration on Product Spaces
§ 21. The product of two measure spaces

(21.1) Remarks. Suppose that (X, JI, f-') and (Y,.AI", 'JI) are two
measure spaces. We wish to define a product measure space
(Xx Y, Jlx.Al", f-' x 'JI),
where JI x.Al" is an appropriate a-algebra of subsets of X x Y and f-' x 'JI

is a measure on JI x .AI" for which
f-' x 'JI(AxB) = f-'(A) o'JI(B)
whenever A EJI and B E.AI". That is, we wish to generalize the usual geo-
metric notion of the area of a rectangle. We also wish it to be true that
f t df-' x 'JI = f f t d'JI df-' = f f t df-' d'JI , (1)
XXY XY YX
for a reasonably large class of functions t on X x Y. Thus we want a

generalization of the classical formula
b d d b
f t(x,y)dS=f ft(x,y)dydx=f ft(x,y)dxdy,
[a,blX[c,dl a c e a
which, as we know from elementary analysis, is valid for all functions

t E<r([a, bJ x [c, dJ).
In the case that X and Y are locally compact Hausdorff spaces
and f-' and 'JI are measures constructed as in § 9 from nonnegative linear
functionals 1 and I on <roo (X) and <roo(Y) respectively, this program
can be carried out by first constructing a nonnegative linear functional
1 x I on <roo (X x Y) and then letting f-' x 'JI be the outer measure
induced on X x Y by 1 xl just as in § 9. A brief outline of this con-
struction follows. For f E<roo (X x Y) and y EY, the function t[Yl:
x~ t(x, y) is in <roo (X). Thu'> we define a function on Y by
y~ 1 (flYl) .
To indicate that we "integrate" with respect to x, we denote this function
by 1%(1). Similarly we obtain a function J:v(t) on X. Next choose open
sets U C X and V C Y such that U- and V- are compact and U x V
:::> {(x, y) EX x Y : t (x, y) =1= O}-. Use (9.5) to find positive constants ac
378 Chapter VI. Integration on Product Spaces
and (J for which

II(q;)I;;;; IX IIq;II .. (2)
for all q; E<roo (X) such that q;(U') C {O} and
IJ('I')I;;;; {J 11'1'11 .. (3)
for all 'I' E<roo (Y) such that 'I' (V') C {O}. Consider any e > 0 and use
the STONE-WEIERSTRASS theorem (7.30) to find a function g on XxY
of the form
..
g (x, y) = }; q;; (x) '1'; (y) ,
;=1
where the q;'s and 'I"s satisfy the conditions of the preceding sentence,
such that 11/- gil .. < e. Since IrYl - grYl is in <roo (X) and vanishes on U'
for every y E Y, we deduce from (2) that
..
IIx (I) (y) - }; I (q;;) '1'; (y) I = II (lrYl) - I (grYl) I
;=1
;;;; IX· sup{l/(x,y) - g(x,y)l: x EX}
~ IX e 11/- gil .. < IX(4)
for all y E Y. Thus Ix (I) is the uniform limit of a sequence of functions
in <roo (Y) all of which vanish on V'; hence Ix (I) E <roo (Y) and Ix (I)
vanishes on V'. Combining this fact with (4), we infer from (3) that
. .
IJeIx(l)) - }; I (q;;) J ('1';) I = IJ(Ix(l) - }; I (q;;) '1';)1 ;;;; (J IX e.
;=1 ;=1
Similarly we obtain
.
II(]y(l)) - }; I (q;;) J('I';)I ;;;; (J IX e.
;=1
We conclude that
J(Ix(l)) = I(]y(l))
and we denote this common value by I xJ (I). This defines a nonnegative
linear functional IxJ on <roo (X x Y). Let p. x v denote the outer
measure on Xx Y constructed from IxJ as in § 9. Using (12.35), it is
easy to show that (1) holds for all I E<roo (X x Y). It is also true that (1)
holds for all I E~1 (X X Y, Jt,..x., P. x v). For a detailed treatment of
this approach to our problem the reader should consult HEWITT and
Ross, Abstract Harmonic Analysis I [Springer-Verlag, Heidelberg, 1963],
pp. 150-157.
While the above approach has aesthetic appeal, as well as more
generality in that a-finiteness of the measure spaces need not be sup-
posed, it produces product measures only for measures constructed as
in § 9 on locally compact Hausdorff spaces. We prefer a construction
that can be carried out for any two a-finite measure spaces. These
§ 21. The product of two measure spaces 379
measure spaces include most of those that arise in classical analysis. The
subject exhibits several technicalities, which can be an annoyance or a
source of fascination: depending upon one's point of view. We proceed to
some exact definitions.
(21.2) Definitions. Let (X, vii) and (Y,.AI') be any two measurable
spaces. For A Evii and BEvY, the set A x B c X x Y is called a
measurable rectangle. The smallest a-algebra of subsets of X x Y contain-
ing all of the measurable rectangles is denoted by vii x .AI' 1 and is
called the product a-algebra. For E C X x Y and x EX, let
Ex = {y EY : (x, y) EE} ;
similarly, for y EY let
£3' = {x EX: (x, y) EE} .
These sets are called X- and Y-sections 01 E, respectively. For a function I
on X x Y and a fixed x EX, let /rx] be the function defined on Y by
/rx](Y) = I (x, y) ;
similarly, for each y EY define Iry] on X by
I[Y] (x) = I (x, y) .
These functions are called X- and Y-sections all, respectively.
(21.3) Theorem. Let (X, vii) and (Y,.AI') be measurable spaces. Then
the lamily .91 01 all finite, pairwise disjoint unions 01 measurable rectangles
is an algebra 01 subsets 01 X x Y.
Proof. First we note that if Ax B and C x D are two measurable
rectangles, then
(AxB) n (CxD) = (A n C) x (Bn D),
m
as the reader can easily verify. Thus if E =.U (Ai x B i ) and F =
<=1
.U" (C j x D j ) are in .91, where these are pairwise disjoint unions of

/=1
measurable rectangles, then
m ..
En F = .':11;':11 [(Ai n C j) x (Bi n D j )] ,
and so E n FEd. That is, .91 is closed under the formation of finite
intersections.
If Ax B is a measurable rectangle, one easily checks that
(Ax B)' = (A'xB) U (XxB') ,
which is the union of two disjoint measurable rectangles. For
1 Note that Jt x.IV is not the Cartesian product of Jt and .IV. Convention de-
mands that this notation be used; we trust it will cause no confusion.
m m
E = .U (Ai x B i ) Ed, we have E' = .n (Ai x B i )', which is a finite
,=1 ,=1
intersection of sets in .91, and so E' Ed. 0
(21.4) Theorem. Let (X, JI) and (Y,.AI) be measurable spaces and
let E EJlx.Al. Then
(i) Ex E.AI for all x EX
and
(ii) EY EJI for all y E Y.
Proof. To prove (i), let Y = {E E Jlx.Al: Ex E .AI for all x EX}.
For any measurable rectangle A x B and any x EX, we have
BifXEA'
{
(A x B)x = 0 if x ~ A .
Thus Y contains all measurable rectangles. To complete the proof of (i),

we need only show that Y is a a-algebra. If (En):=1 C Y, then it is easy
to see that CQI Ent = nQI (En)x E.AI for all x EX; hence Y is closed
under the formation of countable unions. For E EY, we have (E')x
= (E xl' E.AI for all x EX, and so Y is also closed under complementation.
This proves (i). The proof of (ii) is similar. 0
(21.5) Theorem. Let (X, JI) and (Y,.AI) be measurable spaces, and
let f be an extended real- or complex-valued JI x %-measurable function
onXxY. Then
(i) hXJ is %-measurable for all x EX,
and
(ii) frYJ is JI-measurable for all y E Y.
Proof. Suppose that f = ~E for some E EJI x.Al'. Since the state-
ments hXJ (y) = 1; (x, y) EE; y EEx; ~E. (y) = 1 are mutually equivalent,
it follows that hXJ = ~Ex for all x EX. Thus (i) follows from (21.4.i) in the
case that t = ~E' It is now plain that (i) holds if t is a simple function.
The general case follows from the above by using (11.35), (11.14), and
(11.18). A similar argument proves (ii). 0
We will need the following purely set-theoretic fact.
(21.6) Theorem. Let T be any set and let .91 be an algebra of subsets of T.
Then the a-algebra Y (d) generated by .91 is the smallest family ~ of
subsets of T that contains .91 and satisfies the following two conditions.
00
(i) If En E ~ and En C En+! for n = 1,2, ... , then U En

n=1
E ~.
(ii) If Fn E ~ and Fn ~ Fn+l for n = 1, 2, ... , then nOI Fn E ~.
00
Thus, in particular, if the algebra .91 satisfies (i) and (ii), then .91 is a
a-algebra!.
1 Families that satisfy (i) and (ii) are called monotone families.
Proof. Note first that the family ~ exists, for [!P(T) is a monotone
family and the intersection of all monotone families containing d is
again a monotone family containing d; this intersection is ~.
Since any a-algebra is a monotone family and g (d) :J d, it follows
that g (d) :J~:Jd. To finish the proof it therefore suffices to prove
that ~ is a a-algebra. For a monotone sequence (Hn):=I, we write limHn
00 00
for the set U Hn or n Hn according as (Hn) is an increasing or a

.. =1 .. =1
decreasing sequence. If E E ~, write
~= {F E ~ : F n E' E ~, E n F' E ~, E U F E ~} .
Note that F E ~ if and only if E E~. It is also clear that if (F..):=1
is a monotone sequence in ~, then
(limFn) n E' = lim(F.. n E') E ~ ,
En (limFn)' = E n (limF~) = lim(E n F~) E ~ ,
and
E U (limF,.) = lim(E U F,.) E ~ ,
since ~ is a monotone family. Thus ~ is a monotone family for every
EE~·
If E, FEd, then, since d is an algebra, E belongs to ~ and
F belongs to ~. It follows that d C ~ for all E Ed. Since ~ is
the smallest monotone family containing d, we therefore have ~ C ~
for all E Ed. Thus for any F E ~ and E Ed we have F E~; hence
E E~. This shows that d C ~ for all F E~. Since each ~ is a
monotone family, it follows that ~ C ~ for all F E~. The definition
of ~ now shows that ~ is an algebra.
To show that ~ is a a-algebra, let (Fn)~1 c~. For each n EN,
write Gn = Fl U F2 U ... U Fn. Then G1 C G2 C··· and, since ~ is an
algebra, (G"):=1 C~. Thus
00 00
U F,.
n=1
= U Gn
n=1
= lim Gn E ~. 0
(21.7) Corollary. Let (X,.A) and (Y, %) be measurable spaces.

Then .A x % is the smallest monotone family of subsets of X x Y that
contains all finite disjoint unions of measurable rectangles.
Proof. This follows at once from (21.3) and (21.6). 0
(21.8) Theorem. Let (X,.A, p,) and (Y, %, 'II) be a-finite measure
spaces and let E E.A x %. Then the following assertions obtain:
(i) the function x-+ 'II (Ex) on X is .A-measurable;
(ii) the function y-+ p,(EY) on Y is %-measurable;
(iii) J'II(Ex) dp,(x) = J p,(EY) d'll(y) .
x y
Proof. Let F be the family of all sets E E.Lx.Al" such that (i), (ii),
and (iii) hold for E. We will show that F is a monotone family that
contains all finite disjoint unions of measurable rectangles. Then (21.7)
will show that F = .Lx.AI'.
Suppose that E = Ax B is a measurable rectangle. Since E~= B
or 0 according as x EA or x EA', and E'Y = A or 0 according as y E B or
y EB', we have '/I(E~) = v(B) ~.A (x) and ",(E'Y) = '" (A) ~B(Y); hence (i)
and (ii) hold for this E. We may also write
J'/I(E~) d",(x) = J'/I(B) ~.Ad", = '/I (B) • ",(A)
x x
=
y
J",(A) ~B d'/l = yJ ",(EY) d'/l(y) .
Thus (iii) holds for this E, and so E EF. Thus F contains all measurable
rectangles.
Let {E v "" E1>} be a finite, pairwise disjoint subfamily of F. Since
(
..~l
1>
..) x = 1> (1)))' 1>
"~I(EfI)~ and "~I Efl = ,,~,(EfI)'Y for all x EX andy EY, it is
clear that (..~lfl) E F. Thus, F contains all finite disjoint unions of
measurable rectangles.
00
Now let (EfI):=, be an increasing sequence in F and let E = U Efl
.. =1
= lim Efi' For all x EX, we have
as (10.13) shows, and so (11.14) implies that (i) holds for E. In like
manner, (ii) holds for E. Applying B. LEVI'S theorem (12.22), we have
J '/I(E~) d",(x) = lim J '/I «(EfI)~) d",(x)
x "-+cox
= J ",(E'Y) d'/l(y) .
y
Therefore E is in F, and so F is closed under the formation of unions

of increasing sequences.
It remains only to show that F is closed under the formation of
intersections of decreasing sequences. To do this, we must use our
O'-finiteness hypothesis, since we shall call upon (10.15) and (12.24), each
of which contains a finiteness hypothesis. Let (FfI):=1 be a decreasing
sequence in F such that ~ c A x B for some measurable rectangle
00
AxB for which peA) < and '/I (B) < 00. Write F = ..C1IFfI' For each
00
x EX, we have (~)~ c B, and so '/I «(~)~) < 00. It follows from (10.15)
that
for every x EX. Thus the function x -+ ,,(F%) is the pointwise limit of a
sequence of .A-measurable functions, and so it is .A-measurable (11.14);
hence (i) holds for F. Likewise (ii) holds for F. Since
J,,((~)%)dft(x);:;;; J,,(BHAdft< 00
x x
and
J ft ((~),,) d,,(y) ;:;;; J ft(A) ~Bd" < 00,
y y
LEBESGUE'S dominated convergence theorem (12.24) implies that

J ,,(F%) d ft (x) = lim J" ((Fn) %) d ft (x)
x ,,-+00 X
= Jft(P) d,,(y) ,
y
and therefore (iii) holds for F.

Use the a-finiteness hypothesis to choose increasing sequences
(A")k=1 c.A and (B")k=1 c.#" such that ft(A,,) < 00 and ,,(B,,) < 00
00 00
U A" and Y = U B". Let tff ={E E.Ax.#":
for all k, and such that X = k=1 k=1
En (A" x B,,) E F for all kEN}. Since the family d of all finite
disjoint unions of measurable rectangles is an algebra (21.3) and d C F,
we have de tff. If (En):'=1 is an increasing sequence in tff, then
(..QI En) n (A"xB,,) = ..QI (En n (A" XB,,)) E F

because F is closed under the formation of unions of increasing se-
quences. Thus tff is closed under limits of increasing sequences. If (En):'=1
is a decreasing sequence in tff, then
("~I En) n (A" x B,,) = "~I (En n (A" X B,,)) E F,

as proved in the preceding paragraph. Thus tff is also closed under the
formation of intersections of decreasing sequences. By (21.7) we see that
tff = .Ax.#". Now let (F,.):'=I be any decreasing sequence in F and
00
let F = ..n
=1
Fn. Since FE tff, we have F n (A" x B,,) EF for all kEN .
From the fact that F is closed under the formation of unions of in-
creasing sequences, it follows that
00
F = k~1 (F n (A" x B,,)) EF.

This completes the proof that F is a monotone family. 0
We now define a set function on .Ax% that turns out to be the

desired product measure.
(21.9) Definition. Let (X,.A, p) and (Y, %, v) be a-finite measure
spaces. For E E.,I( x %, define
p x v (E) = Jv (Ex) dp(x) .
x
According to (21.8.iii), we also have
p x v (E) = J p(EY) dv(y) .
y
(21.10) Theorem. Notation is as in (21.9). The set function p x v is a

[countably additive] a-finite measure on .Ax%.
Proof. Let {E..}:'=1 be a pairwise disjoint countable family of sets in
vIIx %. Then (12.21) implies that
00
= ~ J v ([E.. ]x) df.t (x)

.. =1 X
00
= ~ f.txv(E ..) .
.. =1
Thus f.t x v is countably additive. Plainly f.t x v ;;S 0 and f.t x '11(0) = 0;
hence p x v is a measure on vIIx %. To show that p x v is a-finite, let
00
(A k)k=1 and (B k)k=1 be as in the proof of (21.8). ThenXxY = k'::.l (Ak xB k)

and f.t x V(Ak x B k) = f.t(Ak) . V(Bk) < 00 for all kEN. 0
(21.11) Note. A simple computation made in the proof of (21.8)
shows that p x v (A x B) = p (A) . v (B) for all measurable rectangles
AxE. [Recall that 0'00 = 0.] If w is any measure on vIIx% such
that w(AxB) = f.t(A)·v(B) for all measurable rectangles AxB, then
p x v (E) = w (E) for all E in the algebra d of all finite disjoint unions
of measurable rectangles. Since f.t x v is a-finite and f/ (d) = vIIx %,
it follows from the uniqueness part of HOPF'S extension theorem (10.39. c)
that f.t x v (E) = w (E) for all E EvIIx %. Therefore the product
measure p x v is uniquely determined by the requirements that it be a
measure on vIIx % and that f.t x v(Ax B) = p(A)· v (B).
We can now prove two versions of FUBINI'S theorem for integrals on
product spaces.
(21.12) Theorem [FUBINI]. Let (X, vii, f.t) and (Y, %, 'JI) be a-finite
measure spaces and let (X x Y, vii x %, f.t x 'JI) be the product meaS$lre
space constructed above. II I is a nonnegative, extended real-valued, -A'x.;V-

measurable lunction on X x Y, then:
(i) the lunction x -+ I(x, y) is -A'-measurable lor each y E Y;
(ii) the lunction y -+ I (x, y) is ';v-measurable lor each x EX;
(iii) the lunction y -+ f I (x, y) d P, (x) is .;V-measurable;
x
(iv) the lunction x -+ f I (x, y) dv(y) is -A'-measurable;
Y
and
(v) the equalities
f I(x,y) dp, x v(x,y) = f f I(x,y) dp,(x) dv(y)

XXY YX
= f f I(x,y) dv(y) dp,(x) 1
XY
hold.
Proof. Conclusions (i) and (ii) are just (21.5.i) and (21.5.ii): we have
written them again only for the sake of completeness. To deal with
(iii)-(v), suppose first that I = ~E for some E E -A'x';v. It is clear that
f ~E (x, y) dp, (x) = f ~EY (x) dp, (x) = P, (P)

x x
for every y EY and that
f~E(X,y)dv(y) = f~E",(y)dv(y) = v(E",)

Y Y
for every x EX. Thus (iii), (iv), and (v) for ~E follow at once from (21.8)
and the definition (21.9) of p, x 'II. For a simple -A'x ';v-measurable
function I, assertions (iii)-(v) are clear from the linearity of all of our
integrals.
Finally, let I be any nonnegative, extended real-valued -A'x';v-
measurable function on X x Y. Let (0'.,):=1 be an increasing sequence of
nonnegative, real-valued, -A'x .;V-measurable simple functions on
XxY such that O'.,(x,y)-+I(x,y) for all (x,y) EXxY (11.35). For
every y EY, (12.22) shows that
f I (x, y) dp, (x) = lim f 0'., (x, y) dp, (x) ;

x n-+oox
hence the function in (iii) is the pointwise limit of a sequence of .;V-

measurable functions, and so (iii) follows from (11.14). Likewise (iv)
1 In interated integrals we treat the symbols f··· d like parentheses. For
example, J J f(x,y) dp(x) dv(y) = f[ fI(x, y) dp(x)] dv(y).

YX Y x

holds. To prove (v), use (12.22) once more to write

J f (x, y) dft x v (x, y) = lim J an (x, y) dft x v (x, y)
xxY n~ooXXY
= lim J Jan(x,y)dft(x)dv(y)
~OOY x
= J [lim J an (x, y) dft (x)] dv(y)
Y ~oox
= J J f(x,y)dft(x)dv(y).
YX
A like computation proves the equality

J f(x,y)dft x v(x,y) = J J f(x,y)dv(y)dft(x). 0
XXY x Y
The following version of Fubini's theorem is particularly useful.

(21.13) FUBINI'S Theorem. Let (X, vii, ft) and (Y,.AI, v) be a-finite
measure spaces and let (X x Y, vIIx.Al, ft x v) be the product measure
space constructed above. Let f be a complex-valued vii x .AI-measurable
function on X x Y and suppose that at least one of the three integrals
J If(x,y)ldft x v(x,y) ,
XXY
J J It (x, y)1 dft(x) dv(y) ,

YX
and
J J If(x,y)ldv(y)dft(x)
XY
is finite l • Then:
(i) the function x -+ f (x, y) is in ~l (X, vii, fl) for v-almost all y E Y;
(ii) the function y -+ f (x, y) is in ~l (Y, .AI, v) tor ft-almost all x EX;
(iii) the function y -+ J f (x, y) dft (x) is in ~dY,.AI, v) 2;
x
(iv) the function x -+ J f (x, y) d v (y) is in ~l (X, vii, ft);
Y
and
(v) the equalities
J f(x,y)dft x v(x,y) = J J f(x, y) dft(x) dv(y)
XXY Y x
= J J f(x, y)dv(y) dft (x)

XY
obtain.
Proof. Our hypothesis and (21.12) show that
J Itldft x v = J J Ifldft dv = J J Ifldvdft < 00 • (1)
XXY YX XY
1 If one of these integrals is finite, then by (21.12.v), all are finite [and equal].
It is to be understood that this function is defined only for those y E Y such
2
that x -+ t (x, y) is in ~l (X, JI, p). A similar remark applies to assertion (iv).
Thus /ES!.l(XXY, Jtx,Al', ft XV). Write /=/I-/2+i(/3-/4)'

where Ii E S!.i (X x Y, Jtx,Al', ft x v) and Ii ~ III, for j = 1,2,3,4.
The functions described in (i) and (ii) are measurable by (21.5). From (1)
we have
J Ili(x,y)ldft(x) ~ J I/(x,y)ldft(x) < 00
x x
for v-almost all y, and so /fY] E S!.l (X, Jt, ft) for v-almost all y (j = 1, 2,
3,4). Thus (i) holds. Assertion (ii) is proved in like manner. Because of
(i), the function in (iii) is defined for v-almost all y EY, and its ,AI'-meas-
urability follows upon applying (21.12.iii) to each Ii and taking linear
combinations. Thus for v-almost all y, we have
I J I(x, y) dft (x) I ~ J II (x, y)1 dft(x)
,
x x
and so (iii) follows upon applying (1). The proof of (iv) is similar. Finally,
apply (12.12.v) to each Ii and then take linear combinations to obtain (v).
This last step is legitimate since the integrals in question are linear on
the various S!.l-spaces and (i)-(iv) hold. 0
(21.14) Remarks. (a) The product a-algebra Jtx,Al' may be
"quite small" even when Jt and,Al' are "very large". In fact, if X is a
Hausdorff space such that X> C, then 9'(X) x 9'(X) does not contain
all closed subsets of X x X [see (21.20.c)].
(b) The measure spaces (XxY, Jtx,Al', ft x v) are seldom complete
(11.20) even when (X, Jt, ft) and (Y,.AI; v) are both complete [see (21.21)J.
As in (11.21), let (X x Y, Jtx ,AI', ft x v) denote the completion
of the measure space (Xx Y, Jtx,Al', ft x v). The following lemma
allows us to extend FUBINI'S theorem to this completed product space.
(21.15) Lemma. Let (X, Jt, ft) and (Y,,AI', v) be two complete,
a-finite measure spaces. Let E E Jtx,Al' be such that ft x v (E) = 0,
and suppose that FeE. Then
(i) ft (FY) = 0 v-a. e.
and
(ii) v (Fx) = 0 ft-a. e.
Proof. We content ourselves with proving (ii). By (21.12), we have
0= ft x v(E) = J J eEdvdft = J v (Exl dft (x) .
x y x
Since v(Ex) ~ 0 for all x E X, it follows that v(Ex) = 0 ft-a.e. Plainly
Fx C Ex for all x, and so Fx E,AI' and v (Fx) = 0 for ft-almost all x. 0
(21.16) Theorem [FuBINIJ. Let (X, Jt, ft) and (Y,,AI', v) be com-
plete, a-linite measure spaces and let I be a nonnegative, extended real-
valued Jtx ,AI'-measurable lunction on X x Y. Then:
(i) the lunction x -+ I (x, y) is Jt-measurable lor v-almost all y E Y;
Iii) the lunction y -+ I (x, y) is ,AI'-measurable lor ft-almost all x EX;
25*
(iii) the junction y-+ J j(x,y)dfl(x) is .AI'-measurable;

x
(iv) the junction x-+ J j(x,y)dv(y) is J/-measurable;
y
and
(v) the equalities
J j(x,y)dfl x v(x,y) = J J j(x,y)dfl(x)dv(y)
xxv YX
= J J j(x,y)dv(y)dfl(x)
xy
hold.
Proof. Let HE J/x.At. According to (11.21), H has the form G U F,
where G E J/x.Al' and FeE for some E E J/x.Al' such that fl x veE)
= O. For each x EX, we have Hx = Gx U Fx, and so it follows from
(21.15) and (21.4.i) that Hx E.AI' for fl-almost all x E X. This proves (i)
for the case that j = ~H Wi) is similar].
The preceding paragraph shows too that v(Hx) = v(G x) for fl-almost
all x EX. Since
J ~H(X, y)dv(y) = v(Hx) ,
y
(21.8.i) shows that the function
x -+ J ~H(X, y) dv(y)
y
is equal fl-a.e. to the J/-measurable function x -+ v(G x ). This proves

(iv) for j = ~H [(iii) is similar].
To prove (v) for ~H' we note that
J ~Hdfl x v = fl x v (H) = fl x v(G) = J J ~Gdfldv = J J ~Hdfldv.
xxv yx yx
The second equality in (v) is similar. The remainder of the proof is like
that of (21.12). 0
(21.17) Note. Fubini's theorem (21.13) is also valid for J/x.Al'-
measurable functions. This fact can be deduced from (21.16) in just the
same way that (21.13) was deduced from (21.12). It seems unnecessary
to repeat the details.
A great many arguments in the theory of integration depend upon
regularity of the measure or measures under consideration. Measures t
as constructed in § 9 are automatically regular [ef. (9.24) and (10.30)].
Regularity is far from obvious, however, for products of such measures.
In the next theorem, we prove a little more: namely, the completion of a
product measure is regular if the factors are regular [and a-finite]. The
proof is rather long, in fact tedious, but it presents no technical diffi-
culties.
(21.18) Theorem. Let X and Y be locally compact Hausdorff spaces,

and let (X,~, p) and (Y, JI", v) be a-finite measure spaces as in §§ 9
and 10. Then the completion p x v of p X v is regular on ~ x JI" in the
sense that if E E~ x JI", then
(i) p x v (E) = inf{p x v(U):E C U, U E..A,. x JI", U is open}
and
(ii) p x v (E) = sup{p x v (F) :E:::> F, FE..A,. x JI", F is compact}.
Proof. Let Bt be the family of all sets E E..A,. x JI" for which (i)
and (ii) hold. We will prove that Bt = ..A,. x JI". Suppose that (En ):'=\ C Bt
00
and write E = n~\ En. If P x v (E... ) = 00 for some no, then (i) is trivial for
E and, by (ii) , En. has compact subsets of arbitrarily large [finite!]
measure; hence E is in Bt. Thus suppose that p x v (En) < 00 for all
n EN. For arbitrary e > 0 and for each n EN, choose a compact set
Fn E..A,. x JI" and an open set Un E..A,. x JI" such that Fn C En C Un
e
and p X v(Un n Fn} < ~. Let U = n~\ Un and F
I 00 00
= n~\ Fn. Then we
have: FeE C U; U, FE ..A,. x.A,.; and also
00
p X v(U n F') ~ 1: p X v(Un n F')

n=\
~ 1; p X
n=1
V (Un n F~) < l' ;..
n=\
= e.
It follows that
p x v(U) = P xv(E) +Px v(U n E') ~ p x veE) +e (1)
and
p X v (E) = p x v (F) + P x v (E n F') ~ P X v (F) + e . (2)
Since U is open, (1) implies that (i) holds for E. Applying (2) and (10.13),
we obtain
p x v (E) ~ lim p x v (1\ U·· . UFp) + e,
p.-..oo
and since 1\ U ... U Fp is compact for all p, (ii) holds for E. Hence Bt
is closed under the formation of countable unions.
Next, let A x B be a measurable rectangle. Use our a-finiteness
hypothesis to select ascending sequences (An)~\ c..A,. and (Bn):'=\ C
00 00
.A,. such that X = n=\ U B n , and p(An) <
U An' Y = n=\ 00 and v(Bn) < 00
for all n EN. Then

00
A x B = n=\
U [(A n An) X (B n Bn)] ,
and so if we can show that Bt contains each measurable rectangle having
sides of finite measure, it will follow from the preceding paragraph that
A x BE Bt. Thus we suppose that peA) < 00 and v(B) < 00. Use the
regularity of {' and v to select ascending sequences (C n ):'1 and (D n )::'=1

of compact sets and descending sequences (Un )::'=1 and (v,,)::'=1 of open
sets such that:
Cn cA c UnCX;
Dn C B c v" c Y;
1
{'(Un) - {'(Cn) < n;
and
Then for each n EN, Un X Vn is open, Cn x Dn is compact, and Cn x Dn

C A xB C Un X Vno Also we have
{' x v(A x B) = {'(A) . v(B)

= lim [{' (Un) . V (Vn)] = lim {' x v (Un X Vn)
n~oo n-+oo
and
{' x v(A x B) = {'(A) . v(B)
= lim [{'(Cn) . v (Dn)J = lim {' x v(Cn x Dn) .
n---+oo n-+oo
Thus A x B E~, and so ~ contains all measurable rectangles.

To finish the proof that ~ = ~ x 1., we need only to show that
~ is closed under complementation. Let (An X B n )::'=1 be as above.
Define ~n = {E t:.~:E C An X Bn}. Clearly ~n is closed under the
formation of countable unions. Let E be in ~n and let B > 0 be given.
Since E E~, there exist a compact set F and an open set U [both in
~ x 1.J such that F C E C U and {' x v(U n F'} < B. Since
An X Bn E~, there exist a compact set J and an open set W [both in
~ x 1.J such that J C An X Bn C Wand {' x v(W n]'} < B. Now
W n F' is open and J n u' is compact. Furthermore it is clear that
and that
{' x v(W n F') - {' X v(] nU') = {' X v((Wn F') n (] nu')')
~ {' X v(W n F' n]') + {' x v(W n F' n U)
< B +B= 2e.
This proves that (An X Bn) n E' is in ~n' We conclude that ~n is a
a-algebra of subsets of An X Bn.
Now let tff = {E EJI", x Jt" : E n (An X Bn) E~n for every n EN}.
It is easy to see from our previous results that tff is a a-algebra of subsets
of X x Y that contains every measurable rectangle. Thus tff = ~ x Jt".
Then for any E E~ we have E' Etff; hence E' n (An X Bn) is in ~ for
<XI
U [E' n (A .. x B ..)] EfIA. Therefore fIA is closed under

all n, and so E' = n=!
complementation. Altogether we have proved that fIA = .A,.. x vII".
Finally, let H be any set in .A,.. x vII". Then H = G U F where
G E.A,.. x vii" and FeE for some E E .A,.. x vii" for which ft x v (E) = o.
If ft x v (G) = 00, it is clear that (i) and (ii) hold for H; thus suppose
that ft x v(G) < 00. Given e > 0, choose open sets U and V and a
compact set C [all in .A,.. x vII,,] such that C C G c U, E c V,
ft x v(U n C') < e, and ft x v(V) < e. Then we have
CCHCUUV
and
ft x v(U U V) - ft x v(C) ~ ft x v(V) + ft x v(U n C') < 2e.
It follows that
ft x v(U U V) < ft x v (H) + 2e
and
ft x v(H) < ft x v(C) + 2e ,
and so again (i) and (ii) hold for H. D
In the following exercises (21.19) and (21.20), the reader may get
some idea of the size of a-algebras .Ax,Ai: actually they are rather
small.
(21.19) Exercise. Let X and Y be topological spaces, each having
a countable base for its topology. Prove that fJI (X) x fJI (Y) = fJI (X x Y).
[Use (6.41) and (l0.42).]
(21.20) Exercise. (a) Let T be a set and let 8 be a family of subsets
of T. Prove that the a-algebra f/' (8) of subsets of T generated by 8
consists precisely of those sets F C T such that F E f/' (~) for some
countable family ~ C tff.
(b) Use (a) to prove that if X and Yare topological spaces and if
FE fJI(X) x fJI(Y), then there exists a countable family ~ of sets of the
form U x V, where U is open in X and V is open in Y, such that
F (. f/'{~).
(c) Use (a) to prove that if X is a set and if D = {(x, x):x E X} is the
diagonal in X x X, then D E.9 (X) x .9 (X) if and only if X ~ c.
[If X ~ c, suppose that Xc R and use (21.19). If D E.9 (X) x .9 (X),
choose a countable family ~ C .9 (X) such that DE f/'({A x B: A,
B E ~}) and then show that the mapping x -+ {C E~ : x EC} from X into
.9(~) is one-to-one.]
(d) Combine (21.19) and parts (a), (b), and (c) to find all Hausdorff
spaces X such that fJI (X) x fJI (X) contains all closed subsets of X x X.
Product measures are as a rule incomplete. This is brought out in the
following exercise.
(21.21) Exercise. (a) Let (X, Jt, p,) and (Y,.AI, 11) be a-finite
measure spaces. Suppose that there exists a set A C X such that A ~ Jt
and suppose that there exists a nonvoid set BE.AI such that lI(B) = o.
Prove that (X x Y, Jt x.Al, p, X 11) is incomplete.
(b) Prove that (R2, ~ x~, Ax A) is an incomplete measure
space.
(21.22) Exercise. For each of the functions I on R x R defined by
1 1 1 1
the following formulas, compute f f I (x, y) dxdy, f f I (x, y) dydx,
o 0 0 0
1 1 1 1
ff I/(x,y)1 dxdy, and ff It(x,y)1 dydx:
o 0 0 0
x2_y2
(a) I(x, y) = (x2 + y2)2 ;
( x-1 1)3 for 0 < y < Ix - ~ I '

(b) I(x, y) = ( 2
o otherwise;
x-y
(c) I(x, y) = Tx.--:ty2)3/2 ;
1
(d) I(x, y) = (1- xy)P' where p > O.
Compare your findings with (21.13).
(21.23) Exercise. Let (X, Jt, p,) be a a-finite measure space. Let
I be a nonnegative, extended real-valued, Jt-measurable function de-
fined on X. Define
V* I = {(x, t) EX x R : 0 ~ t ~ I (x)}
and
V*I = {(x, t) EX x R : 0 ~ t < I (x)} .
Prove that V* I and V*I are in the a-algebra Jt x f!J (R) and that
(i) p, x A(V*/) = p, x A(V*f) = f Idp,.
x
[The equality (i) asserts that the integral of I is the "area under the
curve y = I(x)".]
(21.24) Exercise. Let p, be counting measure on [0, I] and let
D = {(x, x) : x E [0, IJ}. Write X = Y = [0, I]. Prove that
ff ~Ddp,dA '*= f f ~DdAdp, .
YX x y
Why does this not contradict (21.12) ?

(21.25) Exercise. Let (X,~, p,) and (Y, Jt", 11) be measure spaces
as in (21.18). Prove the following.
(a) Every compact Gd subset of X x Y is in ~ x Jt".
(b) Every function in ~oo(X x Y) is ~ x Jt,,-measurable.
(21.26) Exercise. (a) Let I = [0,1]. Suppose that E C I x I is

such that A. (Ex) = A.(I n (E:I')') = 0 for all x, y EI. Prove that E is not
.A}. x .A}.-measurable.
(b) Prove that sets E as described in part (a) exist. [Let LJ be the
least ordinal number such that ~ = c [see (4.47) and (4.48)]. Let
oc -+ XIZ be any one-to-one mapping of PLI onto I. Define E = {(XIZ' x{J):
{J < oc < LJ}. Then 1\ < c and I n (E:I')' < C for all x, y EI. If we accept
the continuum hypothesis (4.50), then all of these sets are countable.
In any case, it follows from (to.30), (6.66), and (6.65) that if these sets
are measurable, then they have measure zero.]
(21.27) Exercise. Prove that there exists a subset S of I x I
[I = [0, 1]] such that Bx ;:; :; 1 and B:I' ;:;;:; 1 for allx, y EI, but S ~.A}. x .A}..
[Supply the many missing details in the following outline. Let §' be the
family of all compact sets F C I x I such that A. x A. (F) > O. Let LJ be
as in (21.26) and let oc -+ FIZ be a one-to-one mapping of P"j onto §'.
Choose (xo, Yo) E Fo· If {J < LJ and S{J = {(XIZ' YIZ) : oc < {J} have been
chosen so that no vertical or horizontal section of S{J has more than one
point, let B{J = {x EI: A.«(Fp)x) > O}. Since A.(B{J) > 0, there exists
x{J EB{J n {XIZ : oc < {J}'. Now since A. «(Fp)x{J) > 0, there exists Y{J E (Fp)x{J
n {y1Z: oc < {Jy. Let S = {(XIZ' YIZ) : oc < LJ}. Clearly S n FIZ =1= 0 for every
oc, and so, since any measurable set of positive measure contains some
FIZ , S cannot be measurable.]
(21.28) Exercise. Recall our discussion of ultrafilters and their
corresponding finitely additive measures given in (20.37). Let 0/1 and "I'"
be ultrafilters of subsets of R such that 0/1::::> {[a, 00 [ : a ER} and
"I'"::::>{J-oo,b]:bE:R}. For all ECR, define p(E) = 1 if EEo/I,
p(E) = 0 if E ~ 0/1, v(E) = 1 if E ("I'", and v (E) = 0 if E ~ "1'". Then p and
v are finitely additive on (!J' (R). Let t be any bounded real-valued function
on R such that lim t (x) = oc and lim t (x) = {J, where oc and {J are any
%-+00 %~-OO
given real numbers. Prove that:

ta) f t(x + y) dp (x) = oc for every y ER;
R
(b) f t(x + y)dv(y) = {J for every x ER.
R
(c) f f t(x + y)dp(x)dv(y) = oc;
RR
and
(d) f f t(x + y)dv(y)dp(x) = {J.
RR
Thus FUBINI'S theorem may fail completely for very simple functions
in the absence of countable additivity.
The following theorem describes the behavior of absolute continuity

and singularity under the formation of product measures.
(21.29) Theorem. Let (X,.A) and (Y,.AI) be measurable spaces.
Let I' and p,t be a-linite measures on (X,..A) and let v and vt be
a-linite measures on (Y, .AI). II p,t ~ I' and vt ~ v, then we have
p,t x '/It ~ I' x v and
(i) d (,ut x vt) (x ) = d,ut (x) • dvt ( )
d(,uxv) ,y d,u dv y
lor all (x, y) E X x Y. II p,t 1- I' or '/It 1- v, then we have 1'+ xvt 1- I' xv.
Writing subscript a's and s's lor the Lebesgue decomposition 01 p,t with
respect to 1', etc., we thus have
(ii) (p,t x vt)G = p,! x,,!
and
(iii) (p,t xv t ), = (p,! x vn + (p,! x v!) + (p,! xv!).
Proof. Suppose that I' t ~ I' and vt ~ v, and let E E..A x.AI be
such that I' x v (E) = o. By (21.9), we have
0= I' x v (E) = f v (E,J dp, (x) . (1)
x
By (12.6), the set A = {x EX:v(Ex) > O} has p,-measure o. By hypoth-
esis, we have p,t(A) = 0 and also vt(Ex) = 0 if x EA'. Therefore
p,t x vt(E) = f vt(Ex) dp,t (x) = f vt(Ex)dp,t (x) +f vt(Ex) dp,t (x)
x A A'
= 0 + f 0 dp,t (x) = 0.
A'
This shows that p,t x vt ~ I' x v. To prove (i), write d:: as 10 and ~vvt
as go [for brevity's sake]. Applying (21.13) to an arbitrary I belonging to
.~\(X x Y,..A x.Al, 1'+ x vt ), and taking note of (19.24), we obtain
f Idp,+ x vt = ff I(x,y)dvt(y)dp,t(x)
XXY x Y
= ff I (x, y)go(Y) dv(y) dp,t(x)

XY
= ff I(x,y)go(y) dv (y) 10 (x) dp,(x)
XY
= f f l(x,y)/o(x)go(y)dv(y)dp,(x). (1)
XY
The function (x,y) -+ l(x,y)/o(x)go(y) on Xx Y is plainly ..Ax.Al-
measurable, and so we can apply (21.13) to the last integral in (1),
finding that
f f I (x, Y)/o(x)go(y)dv(y)dp, (x)
XY
= f l(x,Y)/o(x)go(y)dp,xv(x,y). (2)
XXY
Now combine (1) and (2) and let I= ~A' where A Evi( x % and
p.t x vt(A) < 00. This gives
p.txvt(A) = J
XXy
~A(x,y)/o(x)go(y)dp. x v(x,y). (3)
Since p.t x vt is a"-finite, (3) holds for all A E vi( x%, and the unique-
ness provision in (19.24) proves (i).
N ow suppose that p.t 1- 1'. Let B be a set in vi( such that I' (B) = 0
and p.t (B') = O. Then, as noted in (21.11), we have
p.xv(BxY) =p.(B)v(Y) =0
and
p.t x vt«(B x Y)') = p.tx 'lit (B'xY) = p.t(B') vt(y) = O.
This implies that p.tx 'lit 1- P. x v. For arbitrary 1', p.t, v, and vt , we have
p.txvt = (p.! + p.!) x ('II! + 'II!)
=Wx~+Wx~+Wx~+Wx~,
from which (ii) and (iii) follow easily. 0
FUBINI'S theorem and LEBESGUE'S dominated convergence theorem
are cornerstones of analysis. The theory of Fourier transforms, as well as
many other theories, depends in the last analysis on these two theorems.
We devote the remainder of the present section to a number of applica-
tions of these theorems. Our first result is a simple lemma that is useful
in establishing the p. xv-measurability of functions on spaces X x Y.
(21.30) Lemma. Let cp be a real-valued Borel measurable lunction
defined on R2 such that il MeR and A(M) = 0, then cp-I(M) E ~ x ~
and A x A (cp-I(M)) = O. Then 10 cp is ~ x ~-measurable lor every
complex-valued Lebesgue measurable lunction I defined A-a. e. on R.
Proof. First suppose that I = ~A' where A (~. By (10.34), we have
A = BUM where B EfJI (R) and A(M) = O. Then we may write ~A °cp
= ~Al' where
Since
and
cp-l (M) E ~ x ~ ,
the lemma follows for I = ~A' Since (I + g) 0 p = lop + gop, the
lemma also holds if I is a simple function.
Finally let I be a complex-valued, Lebesgue measurable function
defined on R n F', where FeR and A(F) = O. Use (11.35) to obtain a
sequence (sn) of complex-valued, Lebesgue measurable, simple functions
on R such that sn(x) -+ I(x) for all x ERn F'. Then SnO P is ~ x~-
measurable for all n EN and Sn 0 cp ~ I 0 cp except on cp-l (F). Since

A x A (cp-l(F)) = 0, it follows that I 0 cp is ~ x ~-measurable
(11.24). 0
(21.31) Theorem. Let I and g be in ~l (R). Then lor almost all x ER
the lunction y ~ I (x - y) g(y) is in ~l (R). For all such x define
00
(i) 1* g(x) = J I(x - y) g(y) dy .

-00
Then I * g E~l (R) and III * gill ~ 11/111 . Ilglll' [The lunction 1* g is
called the convolution of and g.] t
Proof. Suppose for the moment that the function (x,y) ~/(x-y)g(y)
is ~ x ~-measurable on R2. We apply (21.12) to write
00 00 00 00
J J If(x-y)g(y)ldxdy=Jlg(y)I'JI/(x-y)ldxdy
-00 -00 -00 -00
00
= J Ig(y)1 .
-00
11/111 dy = 11/111' Ilglll
< 00.
Thus the hypothesis of FUBINI'S theorem (21.13) is satisfied, and so

y ~ I (x - y) g(y) is in ~l (R) for almost all x ER, f * g E~l (R), and
00 00
III * gill = J
-00
IJ
-00
I(x - y) g(y) dYI dx
00 00
~ J J I/(x-y)g(y)ldydx
-00 -00
00 00
= J J If(x-y)g(y)ldxdy
-00 -00
= 11/111 . Ilglll < 00 •
We now proceed to prove that the function (x,y) ~ I(x - y) g(y)

is ~ x ~-measurable. The function (x, y) ~ g (y) is ~ x ~
measurable since for all B Epjj (K), we have {(x,y) ER2 : g (y) EB}
= R X [g-l (B)] E ~ x~. Thus we have only to show that the function
(x, y) ~ I (x - y) is ~ x ~-measurable. Let cp (x, y) = x - y for
(x, y) ER2. Then cp, being continuous, is Borel measurable, and so the
desired result will follow from (21.30) once we show that AX A(cp-I(M)) = 0
whenever A(M) = O.
Let M be such a subset of R. Then
00
cp-l (M) = {(x, y) ER2 : (x - y) EM} = U Pn

n~l
,
where Pn = {(x, y) ER2 : (x - y) EM, Iyl ~ n}. We complete the proof

by showing that Pn E ~ x ~ and A x A(Pn) = 0 for all n EN. Fix
n EN and choose a decreasing sequence (Uk)k~l of open subsets of R
00
such that Men

k=l
U~ and o. Let Bk={(x,y) ER2 : (x - y) E U~,
A(U~) ~
00
Iyl ~ n}. We see at once that (B~) C 81 (R2) and Pn C n B~. Use (10.15),
k=l
(21.12), and (12.44) to write
n
A x A (k=l B~) = lim A. x
k ..... oo
A.(B~)
00 00
= lim
k-+oo
J J ~Bk(X, y) dx dy
-00 -00
.. 00
= lim
k.-..oo _
J J ~Uk(X -
-00
y) dxdy
.. 00
= lim
k.-..oo _ J J ~Uk(X) dx dy
-00
= lim 2n A(U~)
k.-..oo
= o.
00 _ __
Since Pn C k01B~, we have proved that A x A(Pn) = O. 0

It is possible to convolve some pairs of functions not both of which are
in ~I (R). This is brought out in the following two theorems and in
Exercise (21.56).
(21.32) Theorem. Suppose that 1 (R). Then for almost every x ER, the functions y ~ I(x-y) g(y)
and y ~ I(y) g (x - y) are in ~I(R). For aU such x, we write
1* g(x) = J I(x - y) g(y) dy
R
and
g * I (x) = J g(x - y) fey) dy.
R
Then 1* g = g * I a. e., 1* g E~t>(R), and III * gilt> ~ 111111' IIgllp·

Proof. As always, let p' = p~ 1 ' and let h E~P' (R). As in (21.31), we
see that the functions (x, y) ~ I (x - y) and (x, y) ~ g (x - y) are
-AA x -AA-measurable. Applying (12.44), (21.12), and HOLDER'S m-
equality (13.4), we have
JJ I/(x - y) g(y) h(x)ldy dx = J Ih(x)! J /lCx - y) g(y)1 dy dx
RR R R
= J Ih(x)! J /I(t) g(x - t)! dt dx
R R
= J I/(t)1 J Ig(x - t) h(x)1 dx dt

R R
~ J II(t)I' IIg-tllp' II hllp' dt

R
= J II (t)! • IIgllt> • IIhllt>, dt

R
= 11//11 • IIg/lt> • IIhllt>, < 00 • (1)

[As in (8.14) and (12.44), g-t denotes the translate of g by -t.] Since h
can be taken to never vanish, e. g. h(x) = exp(-x2), (1) implies that the
integrals f I/(x - y) g(y)1 dy and f I/(t) g(x - t)1 dt are both finite for
R R
almost all x ER. This proves our first assertion.
It also follows from (1) and (21.13) that the mapping
h ->- f h (x) 1* g (x) dx
R
is a bounded linear functional on 5!,p' (R) with norm not exceeding

11/111 . Ilgllp· Theorem (15.12) shows that there exists a function cp E5!,p (R)
such that
f h(x) cp(x) dx = f h(x) t * g(x) dx (2)
R R
for all It E5!,p' (R) and such that

Ilcpllp ~ 11/111 . Ilgllp·
From (2) and a standard argument, we infer that cp = I * g a. e., and
also that t * g E5!,p(R) and III * gllp ~ 11/111 Ilgllp. Finally we have
1* g(x) = f I (x - y) g(y) dy = f I(t) g (x - t) dt = g * I (x)
R R
for almost all x, i. e., for all x such that these integrands are in 5!,1 (R). 0
(21.33) Theorem. Let 1 ~ P < 00, I E5!,p (R), g E5!,p' (R) [where
P' = p P 1 it P > 1 and I' = 00]' Define t * g on R by
1* g(x) = f I(x - y) g(y) dy.
R
Then 1* g is unilormly continuous on R and III * gllu ~ 11/11p' Ilgllp" II

P> 1, then I * g E<ro(R).
Proof. For P > 1 use (13.4), and for p = 1 use (20.16), to infer that
I * g (x) exists for all x ER and that
III * gllu ~ 1I/IIp Ilgll p' .
Now consider any B > O. By (13.24), there is a b > 0 such that x, z ER
and Ix - zl 1. Given B> 0, choose a compact interval
F = [-a, a] C R such that
f IIIP dA < BP and f IgIP' dA < BP'.
F' F'
[The existence of F follows at once from (12.22) or (12.24).] Then if

x ER and Ixl > 2a, we have [x - a, x + a] C F' and hence
II *g(x)1 ;:;; I J I(x - y) g(y) dyl + I J I(x - y) g(y) dyl
F r
1 I
;:;; (J II (x - y)ll'dy)p IIgll p' + IItllp' (J Ig(Y)lp'dy)7
F r
1
;:;; (l/1I/(Y)/f' dY)P, IIgll p' + 111111' • e

1
;:;; (J I/(y)/f'dy)p 'lIgll p'+ IItllp' e
F'
;:;; (11/111'+ IIgli p') e.

Thus I * g vanishes at infinity, and so I * g 8 cro(R). 0
(21.34) Theorem. With convolution as multiplication, ~l (R) is a
complex commutative Banach algebra.
Proof. We leave it to the reader to make the necessary computations
to show that 1* (g * h) = (f * g) * h, 1* (g + h) = (f * g) + (I * h), and
a. (f * g) = (a./) * g = 1* (a.g) for all I, g, h E~dR) and a. (. K. We saw
in (21.32) that convolution is commutative and we saw in (21.31) that
III * gill;:;; 11/111 • IIglll' Since ~l (R) is a complex Banach space (13.11), all
of the requirements of Definition (7.7) are fulfilled. 0
The algebra ~l (R) is often called the group algebra 01 R.
(21.35) Theorem. The algebra ~l (R) has no multiplicative unit.
Proof. Assume that ~l (R) has a multiplicative unit u, i. e., u E~l (R)
and u * I = I a. e. for all I E~l (R). By (12.34), there exists a real number
2~
lJ > 0 such that J lu (t)1 dt < 1. Let 1= ;[-.".,1. Then I E ~ (R), and so
-2~
for almost all x ER we have

I(x) = u * I (x) = J u(x - y) I(y) dy
R
d x+d
= Ju(x-y)dy = J ult)dt.
-4 x-d
Since A ([-lJ, 6]) > 0, there must be an x E[-6, 6] such that
z+d
1 = I (x) = J u(t)dt.
x-d
Since [x - 6, x + 6] C [-26,26], our choice of 6 implies that
1= Iz1a u(t) dtI~ zE lu(t)l dt ~ -fa lu(t)l dt < 1.

z+d x+d 2d
This contradiction proves the theorem. 0

Even though ~l (R) has no unit, it doe,> have "approximate units",

which for many purposes serve just as well. We give a precise definition.
(21.36) Definition. A sequence (Un):~1 C ~1 (R) is called an approximate
unit [or a positive kernel] if:
(i) Un ~ 0 for all n;
(ii) I Unlll = 1 for all n;
(iii) for each neighborhood V of 0 we have
lim Jun(t)dt=O.
n---+oo v'
It is obvious that approximate units exist; e. g., take Un = ; e[_~ ~J'
n'n
The next theorem justifies our terminology.
(21.37) Theorem. Let (un) be an approximate unit in ~l (R). If
1 ~ P < 00, then lim Ilf * un-/llp= 0 for all f E~p(R).
1>--+00
Proof. Let I E~p(R) and let B > 0 be given. Apply (13.24) to obtain
a neighborhood V of 0 in R such that 21If-:v - flip < B whenever y EV.
Next use (21.36.iii) to choose an no EN such that 411fllp J Un (y) dy < B for
v'
all n ~ no' Now fix an arbitrary n ~ no' Then (21.31) or (21.32) shows
that (f * Un - f) E~p (R), and so for any h E~p' (R) [recall that l' = 00],
FUBINI'S theorem (21.13) and HOLDER'S inequality (13.4) show that
If (f * Un (x) -/(x)) hex) dxl = IJ J(f(x- y) un(y) - f(x) un(y)) dy hex) dxl
R RR
~ J IUn (y) I J If (x - y) - f (x) Ilh (x) I dx dy
R R
<
Ilhllp"
B
Thus the bounded linear functional h -'>- J (f * Un - f) hd)" on ~P' (R) has
R
norm not exceeding B, and so it follows from (15.1) that Ilf*un - fllp~ B
if p > 1. For the case p = 1, we simply take h = 1 in the above com-
putation. 0
(21.38) Remarks. (a) We now take up the Fourier transform for various
classes of functions on R. This transform is of great importance in
applications of analysis, and it is also very useful in describing the
structure of the Banach algebra ~l (R). There are close similarities, as
well as some important differences, between the theories of Fourier
series and Fourier transforms. We will point these out as they come up
in our exposition.
(b) Recall our definition (16.36) of the Fourier transform f of a

function I E~l (R) : for all y ER,
1
(i) f(y) = (2n)-2 J I (x) exp(-ixy) dx.
R
1
The factor (2n) -2 in (i) is placed there as a matter of convenience. The
reader will note the normalization used in the definition of Fourier
coefficients f(n} (16.33): we divided all integrals by 2n. This was done
to render the set {exp (inx)}nEZ orthonormal over [-n, n], and had
useful by-products in (16.37), (18.28), and (18.29). All of these theorems
n
would be slightly more complicated to state had we used J ... d it
-n
n
instead of (2n}-1 J ... dit. The situation is similar in the case of Fourier
-n
transforms. There are good reasons for "normalizing" our integrals with
1
the factor (2n) -2, and we will point them out at the appropriate places.
(c) It is in fact convenient to replace all integrals J ... dit by
R
1
(2n}-2 J ... dit. Let us agree to do this throughout (21.38)-(21.66).
R
Let us also agree that in (21.38}-(21.66),
1
[(2 n) -2jlfIP dit]

1 -
111111' = I'
for 1 ~ P< 00. With this reinterpretation, we have

1
I*g(x} = (2n)-2 J I(x - y) g(y) dy;
R
and the inequalities III * gill ~ IIlIlIllgllI from (21.31) and 11/*gllp ~ 1I/IIdgilp
from (21.32) evidently remain valid.
Our first theorem is simple enough.
(21.39) Riemann-Lebesgue Lemma. II I is in ~l (R), then fisin G:o(R) I.
Proof. For nonzero y ER, we have
f(y) = (2 n) -2JI (x) exp(-ixy) dx

1
(-1) exp(-ni) (2n) -2 JI (x) exp(-ixy) dx

1
=
R
(-1) (2n) -2 JI (x) exp{ -i(x + ;) y) dx

1
=
R
JI{x- ;) exp(-ixy) dx.

1
= (-1) (2n)-2
R
1 Compare this fact with (16.35).
Thus
1 1
21/(y)1 = 1(2n)-2" ft (X) exp(-ixy) dx - (2n)-2" f t(x- ;)exp(-ixy)dxl
R R
1
~ (2n) -2" f It (x) - t (x - ; )1· lexp(-ixy)1 dx
R
;)1 dx.
1
= (2n)-2" f It(X) - t (x -
R
A look at (13.24) shows that I/(y)1 is arbitrarily small if Iyl is sufficiently

large.
It remains to show that I is continuous. Given B > 0, choose a
compact interval I = [-a, a] (a> 0) such that
4 J It (x) I dx 0 such that
a
2ac5 f It(x)1 dx < B. (2)

-a
Since Isin (u) I ~ lui for all u E R, it follows from (1) and (2) that if y, t E R
and It I < c5, then
1
(2n)2"I/(y+t) - l(y)1 = I jl(x) exp(-iyx) (exp(-itx) -1) dxl
R
~ f It(x)l·lexp(-itx) - 11 dx
R
= f It(x)l . 21sin (t;)1 dx

R
a
~ 2 f It(x)1 dx + f It(x)1 . Itxl dx

[' -a
a
< ~ + ac5 f It(x)1 dx

-a
< B.
Thus I is [uniformly] continuous on R. 0

(21.40) Remarks. It is plain that the Fourier transform I- I, which
maps ~l(R) into <ro(R), is linear. It is also bounded, since
1
11/11" = sup 1(2n)-2" ft(x) exp(-iyx) dxl
yER R
1
~ (2n)-2" jlt(x)1 dx = 111111.
R
Our next theorem shows that this transformation also preserves products
[convolution in ~l turns into pointwise multiplication in ~oJ. The
Fourier transform is also one-to-one (21.47) and its range is uniformly
dense in ~o (21.62.b). There seems to be no simple way to describe
intrinsically the functions in ~o that have the form I for some I E~l'
(21.41) Theorem. Let I, g be lunctions in ~l (R). Then
............
(i) I * g = I- g.
Proof. For all y ER we have
1
l--;g(y) = (2n)-2j I*g(x) exp(-iyx) dx
R
1 1
= (2n)-2j (2n)-2j I(x-t)g(t) dtexp(-iyx) dx
R R
= 2~ j g(t) j I(x - t) exp(-iyx) dx dt

R R
= 2~ j g (t) j I (u) exp( -i (t + u) y) du dt

R R
= 2~ j g(t) exp (-iyt) j I(u) exp(-iyu) du dt

R R
=f(y)·g(y);
we have made free use of FUBINI'S theorem and of (12.44). 0

We next take up the problem of reconstructing a function from its
Fourier transform. The analogous problem for Fourier series was treated
in (18.29) and (18.47) supra. The following lemma points up the close
connection between this problem and that of approximating a function
by convolving it with an approximate unit.
A
(21.42) Lemma. Let I, k be lunctions in ~l (R), write u = k, and

suppose that u (t) = u (- t) lor all t ER. Then
1
(i) (2n)-2 jl(y) k(y) exp(ixy) dy = I * u{x)
R
1
= (2n)-2 jl(x - s) u{s) ds
R
lor all x ER.
26*
Proof. Let x E R. Since the function (s, t) -+ I(s) k(t) is in ~l (R x R),

FUBINI'S theorem can be applied in the following computation:
1
(2n)-"2 Iffy) key) exp(ixy) dy
R
= 2~ I II(t) exp(-iyt)dt key) exp(ixy) dy
R R
= 2~ II(t) I keY) exp(-i(t - x)y)dydt

R R
1
= (2n)-"2II(t) k(t-x)dt
R
1
= (2n)-"2II(t) u(x - t) dt
R
(21.43) Pointwise summability Theorem. Let (kn):=l be a sequence 0/

A
lunctions in ~1 (R) and write Un = k n. Suppose that lor each n EN we have

1
UnE~l(R), (2n)-"2 lun{t)dt= 1, and un{-t)=un{t) lor all tER.
R
Furthermore, suppose that there is a lunction u such that:
(i) u E~i([O, ooD;
(ii) u is nonincreasing and absolutely continuous on [0,00[;
(iii) lun{t)i ~ nu{nt) lor aU t ~ 0 and aU n EN.
Then il I is a lunction in ~1 (R) and x is a Lebesgue point lor I, we have
1
(iv) !!!:!,(2n)-"2II(y) kn(y) exp(ixy) dy = I(x).
R
In particular, (iv) holds il I is continuous at x.
Proof. Let x be a fixed Lebesgue point of I. In view of Lemma (21.42),
it suffices to prove that lim I * un (x) = I(x). We have
n~oo
1 1
1* un (x) - I (x) = (2n)-"2I I(x - t)un(t) dt - (2n)-"2II(x) un(t) dt
R R
1 00 _.!.. 0
= (2nf"2I[t(x-t)-/(x)]u n (t)dt+(2n) 2 1[/(x-t)-/(x)]un(t) dt
o -00
1 00
= (2n)-"2 l[t(x+ t) + I(x-t) - 21 (x)] un(t) dt . (1)

o
As in (18.29), we write ep(t) = I(x + t) + I(x - t) - 2/(x). Since lu.. (t)1

~ nu (nt) for all t ~ 0 and all n EN, (1) yields
1 00
II * u.. (x) - l(x)1 ~ (211;)-2:jlep(t)1 nu(nt) dt. (2)

o
Thus we need only to show that the right side of (2) has limit zero as
n -+ 00. To this end, let e > 0 be given and write
00
c = 6e + 6u(0) + 3 J u(y) dy + 611/111 + 61/(x)l + 1.

o
For h > 0, set
II
fP(h) = J lep(t)1 dt.
o
Since x is a Lebesgue point (18.6) for I, there exists a number a; E]0, 1]
such that
1
TfP(h) < ce for hE ]0, a;]. (3)
Since u is nonincreasing and in ~t([O, oo[), it follows at once from domi-

nated convergence (12.24) that there exists an integer no such that if
n > no, then
00
j u(y) dy < : ' (4)

and also "'"
"'"
j
00
2 e (5)
~--; u(y)dy<c'
"'"
Now let n be any integer> nl = max {~, no}. We have

00 1/" '"
jlep(t)1 nu(nt) dt = jlep(t)1 nu(nt) dt + jlep(t)1 nu(nt) dt
o 0 1/"
00
+jlep(t)1 nu(nt) dt
'"
(6)
We apply (3) to II and find that

lin
II ~ j Irp(t)1 nu(O) dt = n~ (:) u(O) < u(O) : < ; . (7)

o
To majorize 12 , we integrate by parts (18.19), use (3) and (5), and inte-
grate by parts again. This yields the following estimates:
0:
12 = j Irp(t)1 nu(nt) dt
lin
0:
= ~(rx.) nu(nrx.) - ~(:)nu(l) - j ~(t)n2u'(nt)dt

lin
0:
~ :~(rx.)ntf(nrx.)+n~(:)u(O)- j~(t)n2u'(nt)dt
lin
j
0:
e'
< C. + Ce u (0) - e
t c n 2u' (nt) dt 1
lin
j
0:
e2 e e [ nrx.u(nrx.)-u(l) ] +-c
=c.+c-u(O)-c en u(nt)dt
lin
j
no:
< ce' e
+cu(O) + c + cu(O) + c
e' e e
u(y) dy
!
I
< : (2e+ 2.(0)+ d

u(Y) Y) < ; . (8)
Next we use obvious estimates and (4) to write

00
Ia = jlrp(t)1 nu(nt) dt
0:
00 00
~ jCI/(x + t)1 + It (x - t)l) nu(nt)dt +21/(x)lnju(nt) dt

0: 0:
00
~ nu(nrx.) . 211/111 + 21t(x)lju(y) dy

no:
~ : (2111111 + 21/(x)I)< ~ . (9)
Thus, combining (2), (6), (7), (8), and (9), we infer that
It * un (x) - l(x)1 < e

1 Recall that u' (nt) ~ O.
if n > n i . Since every point of continuity of I is a Lebesgue point of I,

our last assertion also holds. 0
(21.44) Notes. The reader may well have noticed the similarity
between the proofs of (21.43) and (18.29) [(18.47) is different]. It would
be no trouble to generalize (18.29) to a class of kernels (Un):=l of period
2:n; satisfying hypotheses like those imposed on (Un):=l in (21.43). The
only essential difference in the arguments of (21.43) and (18.29) is that we
need FUBINI'S theorem to interchange the order of integration in the
former, while only sums and integrals are involved in the latter. [Of
course equalities 1: J = J1: are special cases of FUBINI'S theorem.]
There are many sequences (kn):=l that satisfy the hypotheses of (21.43).
We will now give three classical examples of such sequences. The reader
should check in each case that the hypotheses of (21.43) hold.
(21.45) Examples. (a) Let kn (y) = exp (- I~I). For each ex > 0, we have
o 00
!exp(- exlyl)exp(-ity) dy = !exp((ex-it)y)dy + !exp((-ex-it)y) dy

R -00 0
= lim [_1_. _ _1_. exp(-(ex-it)A) +_1_. _ _1_. exp(-(ex+it)A)]

A.... oo ex-zt ex-zt ex+tt ex+,t
_ 2ex I
- ex2 +t2 •
Hence 1
Un (t) = kn (t) = (!) 2 --=1-:+-:72t:::-2
For the function U of (21.43), we may evidently take

1
2)2 1
u(t) = (n 1+t2'
This sequence (Un):=l is known as ABEL'S kernel.
(b) Let kn (y) = (1- I~I) E[-n,nl (y). For every ex> 0, integration
by parts yields
!
« «
(1-1~I)exp(-ity)dY=2! (1- ~)coS(ty)dy

-« 0
= 2! «
sin (ty)
ext
dy
o
2 (l--cos (ext))
- ext2
_ [Sin(t ext)]2
- ex text •
1 This is one of the few Fourier transforms that is computable by inspection.
Hence
" --.!. [Sin(tnt)]2
U.,(t) = k.,(t) = (2n) 2 n tnt .
In this case, we may take u (t) = 1 ~t • . This sequence (u.,):= 1 is known as
FEJER'S kernel. See also (21.55) inlra.
(c) Let k., (y) = exp (- ;;.). It is shown in (21.60) inlra that
u.,(t)=k.,(t)=nexp(- n~t.).
Here we take u = u1 • This sequence is called GAUSS'S kernel.
(21.46) Notes. (a) Theorem (21.43) and Examples (21.45) show why
1
the factor (2n)-"2 is used in the integral defining f. With this factor, we use
the same integral for integrating Fourier transforms that we use for inte-
grating the original functions. This is convenient, and it becomes useful
in some later developments (21.53).
(b) All of the kernels listed in (21.45) can plainly be taken to depend
upon an arbitrary positive real number a instead of a positive integer n.
The equality (21.43.iv) holds as a -+ 00 for all three kernels.
(21.47) Corollary [Uniqueness Theorem]. I I I and g are in ~l (R) and
1= g, then I = g a.e. Thus the mapping I -+ I is one-to-one.
Proof. Let h = I-g. Then h = 0, and so (21.43) and (21.45.a) imply
that
h(x) = lim
n .....oo
fO. kn(y) exp(ixy) dy = 0
R
for all Lebesgue points of h, i.e., for almost all x ER (18.5). 0

(21.48) Remarks. Our next theorem shows that sometimes a thor-
oughly simple-minded device will recapture a function I E~l (R) from
its Fourier transform f. An analogue for Fourier transforms of the partial
sum snl of a Fourier series is evidently
1 A
(i) (2n)-"2 ff(y) exp(ixy) dy ,
-A
and the limit of this expression as A -+ 00, when it exists, is an analogue
of the sum !:
n=-oo
f (n) exp (inx) of a Fourier series. Neither of these limits
need exist, as is well known. In case f is in ~l (R), however, the limit of (i)
as A -+ 00 plainly exists, and remarkably it is I(x) a.e., as we will now
show.
(21.49) Fourier Inversion Theorem. Let I be a lunction tn ~1 (R).

II I is also in ~1 (R), then
1
(i) (2n)-2 fl(y) exp(ixy) dy = I(x)
R
lor every Lebesgue point x 01 I. Hence I is equal a. e. to a lunction in

<ro(R) n ~l(R). II I is continuous, then (i) holds everywhere.
Proof. Suppose that I E~1 (R) and let x be a Lebesgue point of I.
According to (21.43) and (21.45.a), we have
1
I(x) = ~~ (2n)-2 f I(y) exp (- I~I) exp(ixy) dy. (1)
R
Moreover
I/(y)exp(-I~I)exp(ixY)1 ~ I/(y) I (2)
for all n EN and all y ER, and
lim exp (-
~oo
M)
n
= 1 (3)
for ally ER. Since III E~1 (R), it follows from (1), (2), (3), and LEBESGUE'S
dominated convergence theorem (12.30) that (i) holds. Theorem (21.39)
shows that the left side of (i) is a function in <ro (R); and so the second
assertion holds. Two continuous functions that are equal a.e. are the same
function, and so the last statement holds. D
We propose now to define the Fourier transform for all functions in
~2(R). We need two lemmas.
(21.50) Lemma. Suppose that I E~1 (R) n ~oo (R) and that I is real-
valued and nonnegative. Then I is in ~1 (R) and so the conclusions 01
(21.49) hold lor f.
1
Proof. Consider ABEL'S kernel un(t) = (!)2" (1 +nn.t.) . We have

1
II * un (x) I = (2n) -2"1 f I(x - t) Un(t) dtl = ~ 1f I(x - t) 1 +nn't dtl 2

R R
=~lfl(x-~)-1
1£ n 1 + dsl S2
R
::;; 11/1100 f~ = IIIII

- 1£ 1 + S2 00
R
for all x ERand n EN. We set x = 0 in (21.42.i) and use (21.45.a) to infer
that
1
(2n) -2"f I(y) exp (- I~I) dy = I * un (0) ~ 11/1100 < 00 (1)
R
for all n EN. Since I is nonnegative, we may apply B. LEVI'S theorem

(12.22) to the left side of (1) to obtain
1
(2:rr;)-2 jl(y) dy ~ 11/1100 < 00.
R
Thus I is in ~dR). The rest follows from (21.49). 0
(21.51) Lemma. Let I E~1 (R). Define , on R by '(x) = 1(- x). Then
f(y) = fTY) lor aU y ER.
Proof. We have
1
f(y) = (2:rr;)-2 jf(x) exp(-iyx) dx
R
1
= (2:rr;)-2 jl(-x) exp(iyx) dx
R
1 ".-------
= (2:rr;)-2 jl(t) exp(-iyt) dt
R
=TfY). 0
A first step in defining I for I E~2 (R) follows.
(21.52) Theorem. Let I E~1 (R) n ~2 (R). Then I E~2 (R) and
1 1
(i) (2:rr;) -2 jl/(y)l2dy = (2:rr;) -2 jlt(x)l2 dx. 1
R R
Proof. Let' be as in (21.51) and let g = 1*'. Since I, f E~(R),
we have g E~ (R) (21.31); and so (21.41) and (21.51) imply that
g= II = 1/12 ~ O.
Since I, f E~2 (R), (21.33) shows that g E~o (R). Thus g E~l (R) n ~oo (R),
and so (21.50) shows that 1/12 = g E~l (R) and that the inversion formula
(21.49.i) holds for g everywhere, since g is continuous. Thus
1 1
(2:rr;) -2 jl(x+ y) I(y) dy = (2:rr;) -2 jl(x - y) I(-y) dy
R R
1
= I * f(x) = g(x) = (2:rr;) -2 jg(y) exp (ixy) dy
R
1
= (2:rr;)-2 jl/(y)l2 exp (ixy) dy (1)
R
for all x ER. Setting x = 0 in (1), we have (i). 0
1
1 The (2n)-"i in (i) is of no consequence, but the equality would not hold as
1
written without the factor (2n)-' in (21.38.i).
(21.53) PLANCHEREL'S Theorem. There exists a unique bounded linear

translormation T Irom ~2 (R) into ~2 (R) such that T I = f lor all I in
~l(R) n ~2(R)1. Moreover:
(i) IITII12 = 11/112 lor all I E ~2(R);

(ii) (TI, Tg) = (I, g) lor all I, g E~2(R);
(iii) T carries ~2(R) onto ~2(R).
Proof. Define T on ~1 n ~2 by
TI=/.
Since <roo C ~1 n ~2' it follows from (13.21) that ~1 n ~2 is dense in ~2'
For I E~2' let (In) be any .,equence in ~1 n ~2 such that 111- Inl12 -+ O.
Then (In) is a Cauchy sequence in ~2' and so (21.52) implies that
Thus (Tin) is a Cauchy sequence in ~2 and, since ~2 is complete (13.11),

there is a unique function TIE ~2 such that
It is easy to see that T I is independent of the particular sequence

(In) that is used. Thus T is defined from ~2 into ~2' It is also easy to see
that T is linear. Again (21.52) implies that
I Tills = lim
H-+OO
I Tlnlls = lim IIInl12 =
n~oo
lillis,
where (In) is as above, and so (i) holds. Since T preserves norms, it is
bounded and one-to-one. Conclusion (ii) follows from (i) and the polar
identity (16.5), which shows that the inner product is determined by the
norm. The uniqueness statement follows from the fact that if two con-
tinuous mappings [into a Hausdorff space] agree on a dense subset of
their common domain, then they agree throughout that domain.
It remains only to show that T carries ~s onto all of ~2' Since T
preserves norms and ~s is complete, the range of T is closed in ~s. Thus
it will suffice to show that the range of T is dense in ~2'
Let us compute the adjoint T* of the operator T (16.40). For
I, rp E~1 n ~2' write
f
1
g(x) = (211:)-2 rp(y) exp(ixy) dy. 2 (1)
R
1 The function Tf is called the Fouriet' transform of f. Some writers call it the
Planckerel transform, but the term "Fourier transform" is more common, and we
will retain it.
a That is. g is the inverse Fourier transform of rp. as defined in (21.49.i).
Then we have
1
(I, T*qJ) = (TI, qJ) = (2nr2 f /(y) qJ(y) dy
R
= 211& f f I (X) exp(-iyx) dx qJ(y) dy

R R
= 211& fl(x)f qJ(y) exp(ixy) dydx

R R
1
= (2n)-2 f f(x) g(x) dx = (I, g) , (2)
R
the change of order being justified by FUBINI'S theorem [the integrand

is in ~dR2) because I, qJ E~dR)]. Since ~l n ~2 is dense in ~2 and the
mapping I -4 (I, h) is continuous on ~2 for all h E ~2' (1) and (2) imply
that
1
[T*qJ](x) = (2n)-2 f qJ(y) exp(ixy) dy = pC-x) (3)
R
for all qJ E~l n ~2 and all x ER. Let qJ, tp be in ~l n ~2 and write 1= T*qJ,
g = T*tp. Then (21.41) and (3) imply that
M(x) = p(x) 1jJ(x) = I(-x) g(-x) (4)
for all x ER. Since I, g are in ~2' (13.4) shows that I g and so also A
are in ~l' In addition qJ * tp is continuous (21.33). Thus we can apply
(21.49) to invert M: for ally ER, we have
A
1
qJ * tp(y) = (2n)-2 I (x) exp(iyx) dx
R
1
= (2n) -2II(-x) g(-x) exp(iyx) dx
R
1
= (2n)-2/(lg)(x) exp(-iyx) dx
R
/'.
=Ig(y). (5)
Also, (4) and (21.39) show that I g E<ro' and so
IgE~ln<rOC~ln~2·
[We omit the easy verification of the above inclusion.] Hence (5) shows
that qJ * tp is in the range of T, for all qJ, tp E~l n ~2'
Now let (tpn) be an approximate unit (21.36) such that (tpn) C ~ n ~Il
and let qJ E~ n ~2' The preceding paragraph shows that qJ * tpn Erng T
§ 2l. The product of two measure spaces 413
for all n EN, and (21.37) shows that

lim 11-00
hence, rng T being closed in ~2' we have <p Erng T. Thus ~1 n ~2 C rng T,
and so rng T is dense in ~2' 0
(21.54) Remarks. (a) The proof of (21.53) shows that if <p, VJ E~1 n ~2'
then there exists a function h E~1 n <ro such that Ii = <p * VJ. In fact,
h = I g where I = T* <p and g = T* '!jJ.
(b) Theorem (21.53) is the analogue for the line of the RIEsz-FISCHER
theorem (16.39). It is of course much harder to prove than (16.39). This
is accounted for by the fact that ~d[-n, n]) ::> ~2([-n, n]), while ~2(R)
neither contains nor is contained in ~dR).
(21.55) Example. PLANCHEREL'S theorem can be used to evaluate
certain integrals. We illustrate by integrating FEJER'S kernel (21.45.b).
For oc > 0, it is obvious that
1 1
II~[-cx,cxlll~ = (2n) -2 j ~[-cx,cxl(x) dx = (!)2 oc.
R
By (21.52.i), we have
1
II~[-cx,Gtlll~= IIT~[_Gt,Gtlll~=(!)2 oc (1)

as well. Plainly too
1 Gt
~[-Gt.Gtl(Y) = (2n)-2 jexp(-iyx) dx
-Gt
= (2n)- t exp (-iYIX)-:-exp (iylX)
-~y
_ (2)2
- -;; -y-.
1
sin(IXY)
(2)
r
Combining (1) and (2), we obtain
j [ Sin;IXY) dy = noc (3)
R
for all positive real numbers oc. For FEJER'S kernel, then, we have
(2n) - t j(2n) - t n [Si~~tnt) rdt = 1, (4)

R
for all n EN.
Before presenting our next application of FUBINI'S theorem, we give a
number of exercises illustrating and extending the notions of convolution
and Fourier transform.
(21.56) Exercise [W. H. YOUNG]. p, q, and T be real numbers

Let
1 1 1
such that p> 1, q> 1, and also -p + - - 1 = - r > O. Suppose that
q
I E~i>(R) and g E~q(R). Prove that the convolution 1* g is in ~,(R)
and that III * gil, ~ 11/111' . Ilgll q• [Hints. Let a, b, and c be real numbers
111 111 111
such that a=T'p -=- a
+- b'
and-=-+-.
q a cab
Note that-+-+- c
= 1, and use the generalized HOLDER inequality (13.26) on the product
I' q (1 1) (1 1)
(i/(x - y)la Ig (y) Iii) (i/(x - y)li> "i-a) (ig(Y)l q -q-a ).]
(21.57) Exercise. (a) For IE~l(R) and aER, let la(x)=/(x+a).

Prove that (J;)(y) = I(y) exp(iay) for ally ER.
(b) Prove that if I E~l(R), a ER, and g(x) = exp(-iax), then
"....
Ig(y) = (/)a(Y) for ally ER.
(c) Let I, g E~l(R). Prove that J I (x) g(x) dx = J I(x) g(x) dx.
R R
(d) Let I be in ~l (R). Find a necessary and sufficient condition on I
for I to be real-valued; similarly for I to be even.
(21.58) Exercise. (a) Find two functions I, g E~ (R), neither of
which vanishes anywhere, such that 1* g = O. [Hint. Use (21.57.b),
(21.45.b), (21.47), and (21.41).]
(b) Suppose that I E~l (R) and that I * I = I a. e. Prove that 1= 0
a. e.
(c) Suppose that I E~l (R) and that I * I = 0 a. e. Prove that 1= 0
a. e.
(21.59) Exercise. (a) Let I E~l (R), write g(x) = -ixl(x) for all
x ER, and suppose that g E~l (R). Prove that I has a finite derivative at
every point of R and that I' = g. [Hint. Prove that exp (i~h) - 11 ~ Ixl I
and use (12.30).]
(b) Suppose that I E~l (R), that I is absolutely continuous on R,
and that I' E~ (R). Prove that ?(y) = i Y I (y) for all y ER. [HintS.
b
Write I(b) -/(a) = J I' (x) dx and apply (12.30) to prove that lim I(b)
a HOO
'" 1 b
= lim I(a) =0. Then write I'(y) = lim (2n)-2
/1-+-00 HOO
J I'(x)
-b
exp (-iyx) dx
and integrate by parts.]

(21.60) Exercise. Define f on R by f (x) = exp (- ~2). Use (21.59.a)
to prove that f = f. [Hints. Write 1= cpo By (21.59.a) we have
cp' (y) = i(2n) -2"

1
f (-x) exp (- ~2) exp(-iyx) dx.
R
Integrate by parts to obtain cp' (y) = - y cp (y) for all y ER. Conclude
that d~ [cp (y) exp ( ~2 ) ] = 0 for all y ER. Show that cp (0) = 1 by
noticing that
= fob
f expO
0
O]
(- ~) r dr dO 2n. =
(21.61) Exercise. (a) Let cp E~1 (R). Suppose that cp is twice dif-
ferentiable on R, that cp', cp" are in ~1 (R), and that cp and cp' are absolutely
continuous on R. Prove that there exists a function f E~1 (R) such that
f = cpo [Use (21.59.b) to show that ?(y) = _y2 ¢(y) for ally ER. Thus
conclude that ¢ E ~l(R), and then use the inversion theorem (21.49).]
(b) Prove that cp satisfies the hypothesis of (a) if cp, cp', and cp" are
all in ~l(R) n cro(R).
(21.62) Exercise. (a) Let F be a compact subset of R and let U be an
open subset of R such that FeU. Prove that there exists a function
f E~1 (R) such that /(y) = 1 for ally EF and /(y) = 0 for ally ERn U ' .
[Use (21.61) or (21.54.a).]
(b) Let Ql(R) = {/: f E~1 (R)}. Prove that Ql(R) is dense in cro(R)
in the topology induced by the uniform norm. [Use part (a) and the
STONE-WEIERSTRASS theorem.]
r
(21.63) Exercise. (a) For a ~ 0, prove that
f [ Sin;a y) dy = 2~n .
R
[Let t (x) = (1 - ':');[ -a, a] , compute I, and apply PLANCHEREL'S

theorem. Cf. (21.55).]
(b) Compute
J
R
r
[Sin;a y ) dy.
[Use (21.53.ii), (a), and (21.55).]
J [sin;a r
(c) Evaluate
y) dy
R
for a ERand n E{5, 6, 7, .. .}.

(21.64) Exercise. Some rudimentary facts about analytic functions
are needed in this exercise l •
(a) Let t and g be functions in ~l (R) such that t (x) = g (x) = 0
for all x < O. Suppose that t * g = 0 a. e. Prove that t = 0 a. e. or g = 0
a. e. [Hints. Consider a complex number z = s + it with t ~ O. The
Fourier transform I can be extended to
1 00
I(z) = (2n)-2 Jexp(-izx) t(x) dx.

o
Show that I is an analytic function in {z: Imz < O} and a continuous
function in {z: Imz ~ O}. Show that
].g(z) = I(z) g(z) for Imz ~ O.

Thus the analytic function I g vanishes identically in {z: Imz < O},
and this implies that 1= 0 or g = 0, in {z: Imz < O}. If 1= 0, then
I (s) = 0 as well for all s ER, and the uniqueness theorem (21.47) shows
that t = 0 a. e.]
(b) Prove that the Hermite functions (16.25) are a complete ortho-
normal set in ~2(R). [Hints. Let t be any element of ~2(R). For all
z EK, let
1
F(z) = (2n) -2 J exp (- izx - ~ ) t(x) dx.
R
Show that F (z) is defined for all z EK, and that F is analytic in the
entire z-plane. Show too that the nth derivative F(n) of F is given by
1
F(n) (z) = (2n) -2 (-i)n J xnexp (- izx - ~2_) t(x) dx) .
R
1 No knowledge of analytic functions is presupposed elsewhere in the text,

although most readers will surely know the fundamentals of the subject. In any
case, (21.64) is used nowhere else in the book.
If I is orthogonal to all of the Hermite functions, then we have
f xnexp (- ~2) I (x) dx = 0

1
F(n) (0) = (2n) -2 (-i)n
R
for all n E{O, 1,2, ... }, and so F itself must vanish identically. For real
z, this yields
f exp(-isx) exp(- ~I) I (x) dx

1
0= F(s + i 0) = (2n)-2
R
and so by (21.47), exp (- X;)

I (x) = 0 a. e. Therefore 1=0 a. e.]
(21.65) Exercise: More on the structure of 2 1 (R). In this exercise, we
point out some algebraic properties of the Banach algebra ~1 (R) anal-
ogous to those obtained in (20.52) for <ro(X).
(a) Let 6 be a closed ideal in ~1 (R) [for the definition, see (20.52)].
Prove that I E6 and a ER imply I a (. 6. [Hints. For g, k E~1 (R) and
a ER, check that
(ga) * k = g * (ka) = (g * k)a .
Now if (u(n») is an approximate unit in ~1 (R) (21.36), the relations
u(n) * (fa) -+ la' (u~"») * IE 6
prove that la E 6.]
(b) Let M be a multiplicative linear functional on ~ (R), i. e.,
a nonzero linear functional on ~1 (R) for which M (I * g) = M (I) M (g)
for all I, g E~1 (R). Prove that 1M (1)1 ~ 11/111 for alii E~1 (R). [Repeat the
proof sketched in (20.52) for <ro{X).]
(c) Let M be as in part (b). Prove that {I E~1 (R) : M (I) = O} = 8M
is a closed maximal ideal in ~1 (R).
(d) Let M be as in part (b), let x be any real number, and let I be
in ~l(R) n 8;"'. Prove that the number X(x) = M (lx)/M (I) is independent
of I, that the function x -+ X (x) is a continuous function on R such that
X (x + y) = X (x) X(y) for all x, y ER, and that Ixi = 1. [Hints. If I, g
E~1 (R) and x ER, we have (Ix) * g = I * (gx) and M (Ix) M (g)
= M (I) M (gx). Thus
M(f~) = M(g~)
M(t) M(g)'
Now consider x, y ER. If M(I) =1= 0, then M (Ix) =1= 0, as parts (c) and (a)
show. Hence:
M(fH.) M(tH') M(f~) () ()
X (x + y) = M(t) = M(f~) M(f) = XY Xx .
Choose I E~1 (R) such that M (I) = 1. Then
Ix(x + y) - x (x)! = IM(lx+3') - M(lx)1 ~ Il/x+3' -Ixill
= 11/3' -lilt '
and by (13.24) we have lim Ix(x + y) - x(x)l = O. Thus X is continuous.

"......0
Also Ix(x)l = IM(tx)l ~ Il/xlll = 11/111' so that X is bounded. This implies

that Ixl = I.]
(e) Let M and X be as in parts (b) and (d). Then we have
1
M(f) = (2n) -"2 f x (x) I (x) dx
R
for all I E~l(R). [Hint'>. By (20.19), there is an h E~oo(R) such that

f th dA
1
M (I) = (2n) -"2 for all t E~l (R). Take g E~l (R) such that
R
M (g) = 1. Then for all y ER,
1
X(y) = M(g-:v) = (2n)-2" fg(x - y) h(x) dx.
R
For an arbitrary I E~ (R), we thus have

1
(2n) -"2 ft(y) X(y) dy = (2n)-lf fg(x - y) h(x) dx t(y) dy
R RR
= (2n)-lf fg(x - y) t(y) dy h(x) dx

RR
1
= (2n) -"2 fg * t(x) h(x) dx
R
=M(g*/)
= M(I).]
(f) Every multiplicative linear functional M on the Banach algebra
~l (R) has the form
1
M(/) = (2n)-"2 ft(x) exp(-iyx) dx
R
for some fixed y ER. That is, the Fourier transform f of I describes the
values at I of all multiplicative linear functionals. [Use part (e) and
(18.46.a).]
(21.66) Note. The closed ideals of <£o(X) are identified in (20.52.f).
For ~ (R) the closed ideals are far more complicated, and no complete
description of them has as yet been found. An obvious class of such
ideals is the following. For a closed set FeR, let SF = {I E~ (R) :
f (y) = 0 for all y EF}. It is known that there are closed ideals in ~ (R)
not of this form, but little more can be said at present. For a discussion,
see W. RUDIN, Fourier Analysis on Groups [Interscience Publishers,

New York, 1962], Chapter 7.
We now turn to a quite different application of FUBlNI'S theorem,
obtaining a formula for integration by parts under very general circum-
stances.
(21.67) Theorem [Integration by parts for Lebesgue-Stieltjes integrals].
Let IX and f3 be any two real-valued nondecreasing functions on R and let
A", and Ap be their corresponding Lebesgue-5tieltjes measures (9.19). Then
a < b in R implies
(i) A",([a, b[) = lX(b-) - lX(a-);
(ii) A,,({b}) = lX(b+) - lX(b-);
(iii) A,,([a, b]) = lX(b+) - lX(a-);
(iv) J f3(x+) dA" (x) + J IX(X-) dAp(X)
[a,b) la,b)
= lX(b+) f3(b+) - lX(a-) f3(a-);

and
(v) f P(x+)-; P(x-) dA,,(X) +f OI:(x+) ~ OI:(x-) dAp(X)
[a,b) [a,b)
=1X(b+) f3(b+) - lX(a-) f3(a-) .
Proof. For x E R, define 1X0(x) = IX(X-) - IX(O-). Then 1X0 is a left

continuous nondecreasing function on R such that 1X0(0) = 0; i. e.,
1X0 is normalized in the sense of (19.46). Clearly 1X0 = IX - IX(O-) except
possibly at the [countably many] discontinuities of IX. Thus (8.17)
implies that the Riemann-Stieltjes integrals 5", and 5",. agree on <roo(R);
hence (9.19) shows that A" = A",. It follows from (19.47) that A", ([a, b[)
= A". ([a, b[) = ao(b) - 1X0(a) = lX(b-) - lX(a-); hence (i) holds. From
(19.52) we have
A,,({b}) = A",. ({b}) = lim- 1X0(b + h) - 1X0(b)

= lim 1X(b + h) -) - lX(b-) = lX(b+) - lX(b-);
io-j.O
hence tii) holds. We obtain (iii) by adding equalities (i) and (ii).
To prove (iv) , let E = {(x, y) E [a, b] x [a, b] : y ~ x}. Since E is
compact, we have E EfJ(R2) = fJ(R) xfJ(R). Applying (2 1.8.iii) , we
obtain
J Ap(Ex) dA",(X) = J A",(E~) dAp{Y) . (I)
[a,b) [a,b)
Since Ex = [a, x] and E~ = [y, b] for all x,y E [a, b], (iii) applied to (1)
27*
[(iii) obviously holds for 13 as well as ac] yields

J 13 (x+) dA",(X) - 13 (a-)[ac(b+) - ac(a-)]
[a,b]
= J [13 (x+) - 13 (a-)] dA",(X)
[a,b]
= J [ac(b+) - ac(y-)] dAp(Y)
[a,b]
= ac(b+) [f3(b+) - f3(a-)] - J ac(y-) dAp(y) . (2)
[a,b]
Now replace y by x in (2) and rearrange the terms to obtain (iv). To get
(v), simply interchange the roles of ac and 13 in (iv), add this new equality
to (iv), and then divide by 2. 0
(21.68) Remarks. A theorem more general than (21.67) can be for-
mulated by allowing ac and 13 to be any functions that are of finite
variation on each bounded interval of R and considering the correspond-
ing signed or complex Lebesgue-Stieltjes measures. All that is involved
is to reduce the more general case to that of (21.67) by invoking the
Jordan decomposition theorem 117.16). We leave this to the reader. In
case ac and 13 are both absolutely continuous functions, i. e., A", ~ A and
Ap ~ A, (21.67) reduces to (18.19).
(21.69) Theorem [First mean value theorem for integrals]. Let p, be any
finite [nonnegative, countably additive} Borel measure on [a, b] and let
IE (,t'"([a, b]). Then there exists a real number ~ such that a < ~ < band
J I(x) dp,(x) = 1m p,([a, b]) .
[a,b]
Proof. This follows immediately from the fact that I([a, b]) is a
closed interval [perhaps a single point] and from the obvious in-
equalities
min{/(x): x E [a, b]} ~ I-'([!,b]) f I (x) dp,(x)
[a,b]
~ max{/(x) : x E [a, bJ}. 0

(21.70) Theorem [Second mean value theorem for integrals]. Let ac and 13
be real-valued nondecreasing lunctions on [a, b]. Suppose that 13 is continuous
and let Ap be the Lebesgue-StieUjes measure corresponding to 13 (9.19). Then
there exists a ~ E]a, b[ such that
(i) J ac(x) dAp(X) = ac(a)[f3m - f3(a)] + ac(b)[f3(b) - 13m]·
[a,b]
Proof. Let ac (x) = ac (a) and f3(x) = 13 (a) for all x < a, and let ac (x)
= ac (b) and 13 (x) = 13 (b) for all x > b. Let A", and Ap be the Lebesgue-
Stieltjes measures corresponding to ac and 13. Then ac(a-) = ac(a),
ac(b+) = ac(b), and f3(x+) = f3(x-) = 13 (x) for all x ER, and so (21.67.v)
becomes
f {3(x) dA.",(x) + f IX{X+) ~ IX{X-) d)'{J(x) = ct(b) {3(b) - ct(a) {3(a) (1)
[a,b) [a,b)
By (21.69), there exists a ~ E]a, b[ such that

f {3(x) dA.",(x) = {3(~)A.",([a, b]) = {3(~)[ct(b) - ct(a)] . (2)
°for all x (19.52), and so

[a,b)
Since {3 is continuous, we have A.{J({x}) =
f IX{X+) ~ IX{X-) dA.{J(x) = f ct(x) dA.{J(x) (3)

[a,b) [a,b)
because the two integrands differ only on a countable set.

Applying (2) and (3) to (1), we obtain
{3(~)[ct(b) - ct(a)] + J ct(x) dA.{J(x) = ct(b) {3(b) - ct(a) {3(a) . (4)
[a,b)
Plainly (i) follows at once from (4). 0

Our next application of FUBINI'S theorem is of interest in its own
right and also is needed for yet another application that we have in mind
[Theorems (21.76) and (21.80) infraJ.
(21.71) Theorem. Let (X, vii, fl) be a a-finite measure space, let f
be a nonnegative, real-valued, vii-measurable function on X, and let E
be any set in vii. Let fIJ be a real-valued nondecreasing function with do-
main [0, 00 [ that is absolutely continuous on every internal [0, a] for
a> 0. Suppose also that fIJ{O) = 0. For t ~ 0, let Gt = {x EX: f(x) > t}.
Then
J fl (E n Gt )
00
(i) J fIJ 0 f (x) dfl (x) = fIJ' (t) dt .

E 0
Proof. Using (18.16), we see that

f fIJ 0 f (x) dfl (x) = f ~E (x) (fIJ 0 f) (x) dfl (x)
E X
f(x)
= J ~E(X) J fIJ' (t) dt dfl (x)
x 0
00
= f ~E (x) f ~lo'/(x)[(t) fIJ'(t) dt dfl(x) . (1)

x 0
The function (x, t) -+ ~[O,f(x)[ (t) on X x [0, 00 [ is the characteristic function

of the set {(x, t): f(x) > t}. This set is V*f, defined as in (21.23). Hence
it is vIIx vH).-measurable, and so the function
(x, t) -+ ~E (x) ~[O,f(x)[ (t) fIJ' (t)
on X x [0, 00 [ is nonnegative and vIIx vH).-measurable, so that we may
apply (21.12) to the last integral in (1). This yields

00 00
J ~E(X) J ~[O,f(x)[(t) q/ (t) dt dp, (X) = JJ ~E(X) ~[O,f(x)[(t) rp'(t) dt dp,(x)

x 0 x 0
00
= J J ~E (x) ~[O.t(x)[ (t) dp, (x) rp' (t) d t

ox
00
=JJ ~E (x) ~G. (x) dp, (x) rp' (t) dt

ox
00
=Jp,(EnGt)rp'(t)dt.D
o
(21.72) Corollary. For (X, Jt, p,), E, I, and Gt as in (21.71) and lor
p > 0, we have
(i) J IPdp, = jptP-1p,(E n Gt) dt.
E 0
Proof. Set rp(t) = tP in (21.71.i). 0
(21.73) Note. For p = 1, the equality (21.72.i) serves as a definition
of J I dp, if the Lebesgue integral on [0, is known. This technique (Xl [
E
was used by J. RADON [Theorie und Anwendungen der absolut additiven
M engenlunktionen, Sitzungsberichte Akad. Wissenschaften Wien 122,
1295-1438 (1913)J to define Lebesgue-Stieltjes integrals. Note also that
(21.72.i) holds for I E~p(X, Jt, p,) even if X is not a-finite, as the func-
tion I vanishes outside of a certain a-finite Jt-measurable set.
Our final application of FUBINI'S theorem is to the proof of a famous
theorem due to the English mathematicians G. H. HARDY (1877-1947)
and J. E. LITTLEWOOD (1885- ).
(21.74) Notation and Definitions. We will adhere to the following
notation and definitions throughout (21.74)-(21.83). First, I is a non-
negative, extended real-valued, Lebesgue measurable function on R
such that J I dA< for all compact sets F. Define functions 1,1(,), tA(l),
(Xl
F
and 1,1 by the following rules:
I""(x) ~ ,up l.~x ild)':. Elx, ml};

1""(.) ~ ,up 1x~. i I d).:. E]- 00, Xl} ;
IA(x) = max{IA(')(x), IA(l)(X)} .
For each t > 0, let
Gt = {x: I (x) > t};
Mlil = {x: IA(i)(X) > t} (j = t, r);
and
Mt = {x: 1,1 (x) > t} .
(21.75) Lemma. The equality

(i) A.(Mli» = -} f t dA. (j = r, l)
Mlj)
and the inequality
(ii) A.(Mt) ~ ~ f IdA.
M,
hold lor every t > O.
Proof. We prove (i) only forj = r, the case j = lbeing almost the same.
S
It is easy to see that the set Ml'> is open, since the function s ~ -1-fl
s-x
dA.
x
is continuous on Jx, 00[. Let {]Pk' Yk[}k'=l be the unique family of pair-
wise disjoint intervals such that
00
Ml'> = kl},l JPk, Yk[

(6.59). Consider an interval JPk' Yk[ [which may of course be unboundedJ.
For each x EJPk' YkL the open set
S
Nx = {s: J IdA. > t(s - x), s EJx, YkJ n R}

x
is nonvoid. This is trivial if Yk = 00. If Yk < 00 and Nx is void for some

w
x EJPk' Yk[, there must be a w > Yk such that J I dA. > t(w - x). We have
x
W W l'k
J I dA. = J I dA. - J I dA. > t(w - x) - t(Yk - x) = t(w - Yk) .
l'k x x
This inequality implies that Yk EMl'>, a contradiction. Let sx = sup Nx.
Sz
We will prove that s" = Yk' If Sx < Yk' then the equality J I dA = t(sx-x)
" set
holds; an obvious continuity argument proves this. The N.z is non-
:Y
void, so there is a real number y E Js", YkJ such that J IdA> t(y-s,,).
Sz
:Y
It follows that J I dA > t(y - x), a contradiction since y > s". Hence for
x
all x E JPk' YkL we have Sx = Yk' and so the inequality
l'k
J ldA.~ t(Yk- X )
x
holds. Letting x ~ Pk' we obtain

l'k
J I dA. ~ t (Yk - Pk) .
Pk
If Pk = - 00 or Yk = 00, the equality (i) follows. If JPk' Yk[ is bounded,
we have
since (3" is not in M~'). Hence in all cases we have

Yk
f fd).= t(y" - (3,,).
fJk
The equality (i) follows.
To prove (ii), note that Mt = M\') U MIl). Hence we have
).(Mt ) ~ ).(M\'») + )'(MV»)
=+[ ! fd)'+ ! Id)']
Mi') Mil)
~ ~ 1f d)'.
M,
0
(21.76) HARDy-LITTLEWOOD Maximal Theorem for 2p (p > 1). Let p

be a real number> 1. Notation is as in (21.74). Then
1 1
(i) [I WI (j»)Pd)']p ~ p~ 1 [IIPd)']p (j=r,l)

R R
and
/!
1 1
(ii) [I (fd)P d).F ~ 1 [liP d).F .

R R
Proof. We use (21.72), (21.75.i), FUBINI'S theorem (21.12), and
HOLDER'S inequality (13.4) [in that order] to calculate as follows:
=1 ptP- )'(M\i») dt
00
IW(i) (x))P dx 1
R 0
00
= p!tP- 2!f(x) dx dt
o Mil)
! ! ~Mii) (x) I (x) tP-2 dt dx

00
= p
R 0
00
= pi! ~lO.fd(i)(x)[ (t) tP-2 I (x) dt dx

R 0
= 1
P I (x)
[fd(i) (X)]P-l
p_ 1 dx
R
1 1
~ P~ 1 [I l(x)PdxF [I (ld(i) (x))P'(P-I)dxP'

R R
1 1
= P~ 1 [I f(x)Pdx]P [I (ld(i) (x))Pdxp>' .

R R
Since 1- ;, = ~ , the inequality (i) follows if IJ(j) E s!,p. To check this,

use (21.79.i) [which depends only upon (21.75)J to write
oc
fUJ(j) (x)P dx ~ l~k f tP - 2fl(x)dxdt.

R 0 ~
Then argue as above to obtain
fW(i) (x)P dx ~ (p!l~l(~P_k) f(t(x)p dx.

R R
The use of FUBINI'S theorem is justified because the inverse image of
°
the set Ja, ooJ under the map (x, t) -+ ~Mlj)(x) is 0 if a ~ 1, is R x [O,oo[
if a < 0, and is {(x, t) : /,1 (i) (x) > t} if ~ a < 1. Each of these sets is
product measurable. A like calculation, based on (21.75.ii), proves (ii). D
The preceding theorem is ordinarily stated with R replaced by an
interval and with smaller functions /,1('), /,1(1), and /,1. This case is con-
tained in the following corollary.
(21.77) Corollary. Let / be as in (21.74), and suppose lurther that E
is a Lebesgue measurable set such that /(E') = {O}. For p> 1, we have
(i) f W(i)(x)Pdx ~ (p pIt f /(x)Pdx
E E
lor j = r, 1; and
(ii) f
E
t
1,1 (x)P dx ~ (p 2p 1 f l(x)P dx.
E
Proof. Since f g d)' ~ f g d)' for nonnegative g, the result is an im-
E R
mediate consequence of (21.76). D
(21.78) Remark. If the set E of (21.77) is contained in an interval
!
[(X, {3], then it is clear that
1"'" (x) ~ sup {t I x f dl : x <t " pI

if (X < x ~ p. This is the customary definition of IJ(r) in the case of an
interval; similarly with /,1(1) and 1,1.
There is also a version of the HARDy-LITTLEWOOD maximal theorem
for functions in S!,l. As frequently happens [for some mysterious reasonJ,
less is true in the S!,l case than in the s!'p case for p > 1. We need a lemma,
as follows.
(21.79) Lemma. Notation is as in (21.74). For all k EJO, 1[ and all
t> 0, we have
(i) ). (M\i») ~ (1 ~k) t f I d). (j = r, 1) ,
GkI
and
(ii) A(M t );;:;; (1 2k)t IldA.
Gk.
Proof. Define a function g on R by
x = {I (x) if I(x»kt,
g () 0 otherwise .
We have
IA(r)(x)=supj_l_IYgdA+_l-
y-x y-x
IldA:Y>X)
x ]X,y[nGk'
;;:;; gA(r) (x) + kt .

Write N" = {x: gA(r) (x) > u} (u> 0). It is plain that
M~r) C N(l-I.)t .
Applying (21.7S.i) to the function g, we obtain
A(Ml'»);;:;; (1 ~k) t I gdA.

N(l-k).
(1)
Since g = 0 on {x: I (x) ;;:;; kt}, the integral on the right side of (1) is
J IdA;
N(l_k).nG~
it follows that
A(M~'»);;:;; (l~k)t I IdA.
G~
Thus (i) is established for j = r; it is clear how to obtain (i) for j = l

and also how to prove (ii). 0
(21.80) HARDy-LITTLEWOOD Maximal Theorem for S!1. Let I be a lunction
as in (21.74) and let E be any set in .Ai. For each k such that 0 < k < 1,
we have
(i) I IAU) dA ;;:;;
E
! A(E) + 11k I I (x) log+/(x) dx
R
1 (j = r, l),
(ii) I IAdA;;:;;
E
! A(E) + 1 2 k I f(x) log+/(x) dx.
R
For 0 < P< 1, we have
(iii) I (fA(i»)Pd)';;:;; ~(~l~P (1IdAY (j=r,l)
E R
and
(iv) I (fA)P dA ;;:;; 2P ~(~; (I I dAy.
E R
----
1 Recall that log+t = max{logt, O} (0 < t < (0): see (13.37).
Proof. To prove (i), we use (21.72) and (21.79) to compute as follows:
J IA(j) dJ. =
E
r
0
).({y : $E IA(J) (y) > t}) dt
J ).(Mlj) n E) dt
00
=
o
=j +j ~ ~
o
11k
11k
00
)'(E) + 11k j
11k
00
+ j I (X) dx dt
GlcI
=-k-+
A(E) 1
l-k
JOO 1
t j $GlcI(X) I (X) dxdt
11k R
= -k-
A(E) + l-k
1 j I (X) { JOO$GlcI(X) tI}
dt dx. (1)
R 11k
The integral {... } in the last line of (1) is equal to

f(xllk
j +dt = log/(x)
11k
if I (x) > 1 and is 0 if I (x) ~ 1 [use elementary calculus or, if you like,
(20.5)J. Thus the last line of (1) is equal to
A(E)
-k- + 1- k
1 j +
I(x) log I (x) dx . (2)
R
From (1) and (2), (i) is immediate. To prove (ii) , use (21.79.ii) instead
of (21.79.i) in (1).
To prove (iii), a slightly different argument is required. Let ex be
any positive real number. We may suppose that), (E) > OandJ Id)' < 00.
R
Then, using (21.72), we write
00
jUA(il)P d). = PjtP- 1 )'(Mlj) n E) dt

E 0
"Ik
=pj +p j
00
o a.lk
00
~)'(E) ;: + l~k jtP- 2 jl(x)dxdt

a.lk GlcI
= :: ). (E) + 1 ~k j I (x) { j'tP-2 $GlcI (x) dt} dx . (3)

R a./k
Corollary (20.S) shows that
Ur-1f SP-2 ;G, (x) ds ,

00 00
f tP-2 ;Gk. (x) dt =

IXjk IX
and it is easy to verify that
fooSP-2 ;G,(x) ds = jP ~ 1 «(t(X))P-l - o;p-l) if t(x) > 0;,

IX 0 if t(x)~o;.
Since p is less than 1, the last line of (3) is therefore equal to
:: A(E) + 1~P hP-l (~ _ h) f t (x) max{O, o;P-l - t (X)P-l} dx (4)

R
and in turn (4) does not exceed
:P A(E) o;P + CI-P) h~ 1(1- h) f fdA) o;p-l. (S)

R
Regarding (S) as a function of 0;, we see that it has exactly one minimum
value, attained at 0; = 1 ~ h (A (E))-l f /dA. The value of (S) for this 0; is
R
1
(1- h)P
A(E)l-P
--r=p-
(f fdA,)P
R
so that for each k such that 0 < k < 1 we have
f (tA(i))PdA~ 1 A(E)l-P
(1-k)P I-p
(f/dA)P.
E R
Letting k --+ 0, we obtain (iii). Obvious modifications in the above proof,
using (21.79.ii), yield (iv). 0
(21.81) Exercise. Let
/(xl j
~ ;(IO~X)' if
otherwise.
xE ]0, ~ [,
Prove that / ES!,l (R), but that tA(l) ;]O,-H ~ S!,l (R). [Hint. Show that
1
/ ,1(1) (x) ~ x jlogxj f or x E] 0'"2
1[ ]
.
(21.82) Exercise [T. M. FLETT]. For a function / ES!,t (R), p > 1, let
Af(P) denote the number such that II/A(I)llp = Af(P) II/lip. Define a se-
quence in s!'p (R) by
x(n-'-1)p- if x E]O, 1[, 1
{
/n(x) = 0 otherwise.
§ 22. Products of infinitely many measure spaces 429
Prove that lim Atn(p) =

~oo
_L,
p-l
thus proving that the constant -p-
p-l
in (21.76.i) is the best possible.
(21.83) Exercise [K. L. PHILLIPS]. Let f be as in (21.74). Prove that
fAt x) ~ sup ( ~
u , /f d.l : - = < t ,; x ,; u < =. t '" U) ,
§ 22. Products of infinitely many measure spaces

(22.1) Introductory Remarks. If one flips a coin n times, the
number of possible outcomes is 2n , the number of n-tuples having and 1
as entries; or, it is the number of functions from {I, 2, 3, ... , n} into
°
{O, I}. It is intuitively obvious that any two of these functions are
equally likely to occur if the coin is unbiased. Thus it is just as likely that
all n flips will yield heads as that they will alternate from heads to tails
to heads .... It seems also intuitively clear that "in the long run" [as n
goes to 00] it is probable that the number of heads obtained is about one
half the total number of tosses. We can interpret the expression "is
probable that" to mean that for some appropriately chosen measure
fl on {O, l}N [= all sequences t = (tv t 2 , ••• , tn> ... ), where tj is or 1
for allj], we have
°
Perhaps the most convenient way to study probability measures such

as the one indicated here is to consider infinite products of measure spaces,
and this fact is a sufficient reason for including the present section.
Furthermore, many important and useful constructions in measure
theory and its applications throughout analysis depend upon measures
on infinite products of measure spaces. The subject is too important to
be ignored.
The ideas used in the study of infinite product measures are not at
all recondite, but the notation is complicated and, it may be, a little
forbidding. The reader should be sure to keep the notation of (22.2)
firmly in mind throughout.
(22.2) Definitions and Notation. Throughout this section (T",~, fly)
will den~te a measure space for each y contained in an index set r,
where r ~ No and fly(T,,) = 1 for every y Er. Let T = X T,,; we
yEr
will often write t = (ty) for elements of T, where ty denotes the value
of t at y. The symbol Q, with or without a subscript, will be reserved for
finite subsets of r. The complement Q' will always be with respect to
r: Q' = r n Q'. For any setLl such that 0 =l= LI c r, we let ~ = X T"
yELl
[thus in particular Tr = T]. Note that ~ is not a subset of T if LI S; r.
r
For Q = {I'l> 1'2' ••. , I'm} C [the I'/s are distinct], let GD be the family
of all sets
A", x A". x ... X A"m (A"J EJ{"J) .
These are the analogues of measurable rectangles for the product of
two spaces. For an arbitrary subset LI of r,
let ~ be the smallest
algebra [not a-algebra] of subsets of ~ that contains all sets AD x TLI nD',
where Q runs through all finite subsets of LI and AD through all sets in
GD.l Let ~ be the a-algebra !/ (~). We write ,AI' for %r and vii for
vIIr. Note that for finite Q, the a-algebra vIID is just !/(GD).
(22.3) Discussion. Our aim is to construct a [countably additive]
measure p on vii such that
(i) p(T} = 1
and
IL((A
(ii) r "l
X ' " x A } x 1',
".. {"I.",,·· .•"..},)
= p".(A".) P,," (A,,") ••• p""(A,,.. )
if A"J EJ{"J (j = 1,2, ... , n). We will then prove two analogues of
FUBINI'S theorem for this product measure. We begin with a technical
lemma.
(22.4) Lemma. Let {LlQ}eEP be a pairwise disjoint family of nonvoid
r
subsets of such that U LlQ =
eEP
r.
Then the mapping
(i) (ty) -+ ((t,,)yELle)QEP = <P(t)
is a one-to-one mapping of T onto the product space
XTLI = T+.
I!EP Q
The mapping <P carries vii onto the smallest a-algebra vIIt of subsets of Tt
that contains all sets
XA LI ,
eEP e
where ALI EvilLI for all e EP and only a finite number of the sets ALI are
e Q Q
dilferent from TLI Q.
Proof. The first assertion of the lemma is obvious on a little re-
flection: the mapping <P merely "regroups" the terms t" in (t,,}yEr' To
prove the second assertion, consider first a set
A=XA"CT, (1)
yEr
1 We commit a slight inaccuracy in writing AD X TA nD' as a subset of T...
although the intended meaning is clear enough. See (22.4) for a justification.
§ 22 Products of infinitely many measure spaces 431
where Ay E~ for all Y and the set Q = {y Er : Ay =1= :q is finite.

Plainly W(A) is the set
B=W(A)=X(XAy). (2)
~EP YEAQ
Since Q is finite, all but a finite number of the sets X Ay are equal to
YEA(/
TA , and it is plain that each X Ay is in ~ C vilA . Therefore W(A)
Q YEA~ II Q
is in vIIt. Obviously vii is the smallest a-algebra of subsets of T contain-

ing all sets A of the form (1). Since W is one-to-one, it preserves all
Boolean operations, and therefore W(vII) C vIIt. It is easy to see that
the family vIIt is the smallest a-algebra of subsets of Tt containing all
sets B ofthe form (2). Since W-l(B) has the form A, we have W-l(vllt)
C vii, and so vIIt C W(vII). 0
Our first step in constructing the measure /-t on T is to show that /-t
is uniquely determined for finite products by the requirements in (22.3).
(22.5) Lemma. Let Q = {Yl> Y2' ... , Ym} be any finite nonvoid
subset of r.
There is a unique measure /-tQ on vIIQ such that
(i) /-tQ C~1 Ay;) =

m
/J /-ty; (AYj)
for all X A1'; EtlQ.

i=1
Proof. (I) Suppose first that Q = {Yl> Y2}' Let /-tQ be the product
measure /-ty. x /-ty. on vIIQ = vIIy• X vIIy• defined in (21.9). The uniqueness
of /-tQ follows from (21.11).
(II) We complete the proof by induction on the number of ele-
ments in Q. Suppose that the result holds for all subsets of having r
n elements, and that Q = {Yl> Y2' ... , Yn> Yn+l}' Let /-t' be the unique
measure on ~y.,y"""yn} satisfying the inductive hypothesis, and let /-t
be the unique measure on ~y"".,yn} X vIIyn+. such that /-t (B x Ay,HJ
= /-t' (B) . /-ty"+1 (AYnJ for all B E~Yl,,,,,yn} and AYn+. EvIIYn+1' The
existence and uniqueness of /-t are guaranteed by part (I). Lemma (22.4)
shows that ~Y"Y.' ...'Yn} X vIIYn+. can be identified with vIIQ. We can
therefore define the measure /-t on vIIQ. Doing this, we have
/-t(Ay. x Ay. X· •• X AYn x Ayn+.)
= /-t((Ay. X AYl x ... x Ay.. ) X AYnJ
= /-t' (Ay. x ... x Ay.. ) . /-tYR+' (AYnJ
= /-t1'. (A1',) . /-t1'1 (AYI) ••• /-ty" (AyJ • /-t1'''+1 (A1'"J
for sets Ay; in ~r Take the measure /-t just constructed to be /-tQ in the
statement of the theorem. To prove that /-tQ is unique, observe that any
measure p, on 10 satisfying (i) defines, in an obvious way, a measure

on .A{Y"ooo,Yn}o In view of the uniqueness of 1-", this measure is equal to
1-". It follows that p, = 1-'0 by the uniqueness of 1-'0, for the case
D = {Y1' Y2}' 0
Our next result is also a useful technicality.
(22.6) Lemma. Ij D1 n D2 = 0 and Bo; E1o;(j = 1,2), then
(i) I-'0IU02(Bol x B 02 ) = I-'01(B01) • 1-'02 (B 02 )'
Proof. Use Lemma (22.4) to identify 101U02 with 101 x 1 02 ,
If Bo; is in tB'0; (j = 1, 2), then BOI x B02 is in tB'0IU 02' and it is clear that
I-'0IU02(Bol x B 02 ) = I-'01(Bo1) • 1-'02(B 02 ) :
both sides of this equality are the same products II I-'y;(Ay;). Thus
I-'0IU02 and 1-'01 x 1-'02 are measures on 101U02 that satisfy (22.5.i)
for D = D1 U D2. By (22.5), they are the same measure. 0
We now consider the full infinite product T = X Ty, first proving
a preliminary fact. yEr
(22.7) Theorem. There is a unique finitely additive measure I-' on

the algebra oj sets,Ai such that
(i) I-' (Ao x To,) = 1-'0 (Ao)
jor all D and all Ao E1 0 ,
Proof. Let the expression (i) define 1-" We first show that this defini-
tion is unambiguous. Thus suppose that
AOl x To;. = A02 x T02 .1 (1)
Write Ds = D1 U Dl!' Then (1) can be rewritten as
AOl x T03noi x Tos = A02 x T03no; x TOa . (2)
The sets AOl x TOanoi and A02 x T03noil are in 1 0a , and (2) shows that
they are equal. Lemma (22.6) shows that
and
I-'oa(A02 x TOanoia) = I-'02(A02 ) • 1 = I-'02(A02 ) .
Thus I-' is well defined by (i).
For every set A E,Ai, it is easy to see that there exists an D and a set
AD E.A"iJ such that
A = Ao x To' ; (3)
we omit the proof of this simple fact. If A1 and All are disjoint sets in,Ai,
write
(j = 1,2)
1 Again we commit a slight solecism in writing AOI x T 0i = A O2 X T 02' as
the elements of these two sets are really different entities. Again Lemma (22.4)
saves the day.
as in (3); next write Da = D1 U D2 ; and then write

Ai = Ag! x TDs (j = 1, 2) ; (4)
it is clear that this can be done and that Ag! E ADa' It is also clear that
Ag; n A~; = 0, and so, using (i) and the additivity of f'Da (22.5), we
have
f' (AI U A2) = f'Ds (Ag; U A~m
= f'Da(A~m + f'Da(A~;)
= f' (AI) + f' (A2) .
That is, f' is finitely additive on .AI. 0

It is tempting to use the technique of (22.7) to try to prove that f'
is actually count ably additive, for each f'D is countably additive. This
approach must fail, since we cannot necessarily obtain a finite subset D
r
of such that each of the countably many sets in question is a subset of
TD • Furthermore, one cannot apply (10.36), since there may exist
00
pairwise disjoint families {Cn}:=1 c.AI such that n~ICn E.AI, but no
..AD contains all of the sets AD.. E ..AD" such that Cn = AD" x TD;.. For
example, let r = N, T = [0, I]N, and Bn = {t ET: 0 <t~ <! for
k = 1, 2, ... , n}. Then write
C1 = B~, Cn = B~ n (B1 n ... n B n - 1) for n> 1.
00
It is clear that n~ICn = T, that Cn = A.. x 1{1, ... ,nY for A.. E ..4{1, ... ,n},
and that Cn does not have the form Dn- 1 x 1{1, ... ,n-1Y for any D n- 1
E..4{1,... ,n-l}' We proceed to the theorem itself. The proof is perhaps
complicated, but its basic idea is simple enough.
(22.8) Theorem. The finitely additive measure f' on.AI admits a
unique extension over ..A that is countably additive.
Proof. The uniqueness of f"s extension over ..A [if it exists at all]
is proved by an obvious application of (21.6). To prove that f' has some
count ably additive extension over ..A, we need only show that
lim f' (Fn) = 0 (1)
..-..00
for every sequence (Fn)::"=1 such that F.. E.AI, 1;;. :::> F2 :::> ••• :::> Fn :::> ••• ,
00
and n Fn =
n=1
0. This follows from (10.37).1
1 The reader will recall that (10.37) is an exercise, for which he must supply
the proof. We believe that all readers who have worked through the text to this
point will now be able easily to prove (10.37), if they omitted to do so on first
reading § 10.
Hewitt/StlOmberg, Real and aLstract analysis 28
In this paragraph, we make some reductions in order to simplify

subsequent notation. Each Fn has the form A.Qn x T.Q', where A.Qn E.AD.n ft
00
Let L1 = n':!l [In' By (22.7), there is a finitely additive measure ft.!1 on.;¥,1
such that ft.!1 (A.Q x T.!1n.Q') = ft.Q(A.Q) for all [I c L1 and A.Q E .A.Q. For
each n, let F~ = A.Q.. x T.!1n.Q~ C JA. It is then clear that:
each F~ belongs to .;¥,1;
ft.!1 (F~) = ft (F..) ;
F1 :::> F1 :::> ••• :::> F~ :::> ••• ;
n F.!1n =
00
and n=l
10.
It clearly suffices to prove that lim ft.!1 (F~) = O. In other words,

r
~oo
we lose no generality in supposing that is countably infinite. It is now

just a notational matter to suppose that = N = {I, 2, ... }. Let r
kn = max [In' We may suppose with no loss of generality that
[In = {I, 2, ... , kn} and that kl < k2 < ka < .. '. Define the sequence of
sets (Em):;';'=l by the following rule:
T if I ~ m<kl'
E ={
m Fn if kn ~ m < kn+1'
n Em = n Fn =
00 00
Then we have 50 and lim ft(Em) = lim ft(Fn}. We

m=l H=l ~oo ~oo
must show that lim ft (Em) = O. Let em = {I, 2, ... , m} for each m.
tn--+oo
Note that each Em has the form Am X Te;.., where Am E.A-@",.
As noted in the proof of (22.5), we have fte"+1 = fte .. x ftm+l for all m.
It is also easy to see that a set in .A-@"'+1 is .A'e.. x .Am+l-measurable.
Now apply (22.7) and (21.12.v) to write
ft (Em) = fte .. (Am) = J ~A .. (t) dftem (t)
Te",
where t* denotes a generic element of Tem_,. By (21.I2.iii), the inner

integral in (2) is .Aem_,-measurable, and so we can apply (22.5) and
(21.12.v) again. Doing this m - I times, we arrive at the equality
Assume that ft (Em) does not go to zero. For S1 ET1, let

!t.m (S1) = J ... J ~A", (s1> t2, ••. , tm) dftm (tm) .•. dft2 (t2)
T. Tm
for m = 1, 2, 3, .... That is, we carry out all but the outermost integration
in (3), and leave the first variable unintegrated. It is plain that
# (Em) = J II,m (tl) d#l (tl)
T,
and that II,.,. (I;.) C [0, 1J. It is not the case that
lim II,m(tl ) = 0 everywhere on T I ; (4)
m--+oo
if (4) held, then LEBESGUE'S dominated convergence theorem (12.24)

would imply that lim # (E.,.) = O. Hence there is a point al E Tl such
m--+oo
that II,.,. (a l ) does not go to zero as m ~ 00. Next define 12,.,. by
12,.,. (S2) = J ... J ~Am (at> S2' ta, ... , t.,.) d#m (t.,.) •.. d#s (ta) .
T. Tm
If it were the case that lim la,.,. (S2) = 0 for all Ss ET s' then by (12.24)
m--+oo
we would also have lim Jla,.,.(ta)d#s(t2) = 0; i.e., we would have
tn--+oo T"
lim II,.,. (a l ) = O. Hence there is a point as ETs such that 12,m (as) does
m--+oo
not go to zero as m ~ 00.
In this way we construct a sequence of points (at> a2, ... , an' ... ) = a
in T with the following property. For every n EN, the sequence of
numbers
defined for m > n, does not have limit zero as m ~ 00. Now the integrand
in (5) cannot vanish identically for all large m, and so
(at> aa, ... , an, Sn+t> ... , sm) EA.,. (6)
for appropriately chosen Sj E T;, and for arbitrarily large m. Since
Em = Am X T@;,., we can choose a point s(m,n) E T@~ such that
(ai' aa, ... , an) x s(m,n) EEm .
Since E.,. C En' we also have
(ai' aa, ... , an) x s(m,n) EEn . (7)
Since En =
.
An X T@', (7) implies that
(at> as, ... , an) x T@, C En. II
In particular, a EEn. Since n is any positive integer, we infer that
a E..n
00
=1
En .
n En =
00
This contradicts the equality 0 and completes the proof. 0
.. =1
28*
Thus we have a count ably additive measure fl on a a-algebra of

subsets of T that behaves like a product measure. A first and very
simple application is to a classical problem in probability.
r
(22.9) Example. Let = {I, 2, 3, ... } and let Tn = {O, I} for each
r.
n E Then T = {t: t = (tn), tn = 1 or tn = O}. Define the measure fln
on each Tn by setting fln({O}) = fln({I}) = ~, and let ~ be all four
r
subsets of {O, I}. For a finite subset {kl' ... , k n } of [all k's distinct] and
any sequence (a k" ak" ... , akJ of O's and I's, write E (ak" ak" ... , akJ for
the set {t ET: t k, = ak" t k, = ak" ... , tkn = akn }' The definition of fl as
given in (22.3.ii) and (22.7) shows at once that
fleE (ak" all" ... , akJ) = 2-n .

1 1
For n EN, define In on Tby In(t) = tn - 2' and let hn = n (11 + ... + In)·
We will show that
(I )
1
It IS clear that Ii = ;E(I;) - 2' that
1 1 1
Ii Ik = ;E(Ij.1k) - 2 ;E(lj) - 2 ;E(lk) + 4" (j oF k),
1
and that IT = 4- .
Therefore
This proves (I). From (I) and (13.33), we infer that hn -+ 0 in measure.
Thus for every e > 0, the equality
~~ fl ({t ET: I_/'_~~~ -~-± ~ - ~ I> e}) = 0 (2)

obtains.
If we let 0 and 1 correspond to obtaining heads or tails, respectively,
upon flipping a coin, then I, + ...
n
+ In is the proportion of tails obtained in
n flips. The equality (2) then asserts that the probability that the propor-
tion of tails is farther than e from ~ decreases to 0 as n [the number of
flips] goes to infinity, for every e > O. If the coin is unbiased [fln({O})
= fln({I}) = ~] this is what we would expect.
The equality (2) is one form of the weak law ot large numbers. [See
also (22.32.b) intra.]
(22.10) Exercise. Consider a generalization of (22.9), as follows.
Let rbe arbitrary [but infinite of course]. For each y, let Ay be any set
in~and let Iy be the function on Tsuch that Iy(t) = ~A y (ty) - fly(A y}.
For Q = {Y1' Y2' ... , Yn} C r, let w(Q) be a positive number and let
hn = w(Q} I: Iy·
yEn
(a) Show that J Ihnl sdfl = W(Q}2 I: [fly (Ay) - (fly(Ay})2].

T yEn
(b) Generalize the notion of convergence in measure: hn ~ 0 in
measure if for every c5 > 0 and every e > 0 there is an Q o c such that r
fl({t ET: Ihn(t)! ~ c5}) < e
for all Q ::> Q o. Find reasonable conditions on w (Q) for hn to converge

to 0 in measure. What simple form can you give w(Q) if all fly(Ay} are
equal? What happens if fly(Ay} = O? If fly (Ay) = I?
There are several quite distinct analogues of FUBINI'S theorem for
infinite products, all of which coalesce trivially for finite products. These
distinct versions arise because of the various different ways in which we
can approximate J I dfl and I by integrals over finite numbers of co-
T
ordinates. Our first Fubini-esque theorem deals with ~p-convergence
and is quite general. We need three lemmas.
(22.11) Lemma. Let (T;, ~, fl;) (j = 1,2) be measure spaces such that
fl;(T;) = 1; let (T, vii, fl) = (T1 X T 2, viiI X v112, fl1 x fl2); and let p be a
real number ~ 1. For IE fi!.p(T, vii, fl), let SI be the lunction on T such that
lor all S2 ETa· Then 51 is in ~p(T, vii, fl) and IISllip ~ II/IIp, so that 5 is a
norm nonincreasing linear translormation 01 ~p(T, vii, fl} into itsell.
Proof. Since fl(T} = 1, we have ~p(T, vii, fl} c ~l(T, vii, fl}. Thus
I is in ~1 (T, vii, fl), and it follows from (21.13.iv) that the function
Sl ~ J I (Sl' ts) dfls (ta)

T,
is in ~l (7;.,~, fl1)' In particular this function is ~-measurable, and

since the function
does not depend on S2' it is plainly vii-measurable. Using (12.28.ii),

(13.17), and (21.12), we obtain
liS fll~ = J I J f (SI' t2) dfl2 (t2) IP d (fll x fl2) (s1> S2)
T,XT, T,
Hence Sf is in S!p(T, vii, fl) and IISfilp ~ Ilfllp. 0

(22.12) Lemma. Let r be an arbitrary infinite index set, and suppose that
o ~ L1 ~ r.
Let fl,j and fl,j' be the measures on the a-algebras ~ and vII,j'
constructed as in (22.7) and (22.8). Identifying vii and ~ x vII,j' [the
mapping cP of (22.4) allows us to do this j, we have fl = fl,j x fl,j'.
Proof. The measures fl and fl,j x fl,j' agree on sets of the form
AD x TD " where Q is a finite subset of r, and so by the uniqueness of fl
[(22.7) and (22.8)J they agree throughout vii. 0
The next lemma is a necessary technicality.
(22.13) Lemma. For every BEvil and every c> 0, there is a set
A E.AI" such that II~A - ~Blll = fl(A 6. B) < c.
Proof. Define the family g> as {B Evil: for all c > there exists
A E:.AI" such that fl (A 6. B) < c}. It is trivial that g> ::> .AI"; to prove
°
that g> = vii, it suffices to prove that g> is a a-algebra. We do this by
appealing to (21.6). Let (Bn) be a monotone sequence in g> [either
increasing or decreasingJ and write B = lim Bn- Given c > 0, use (10.13)
or (10.15) to select mEN such that fl(B 6. Bm) < ~ . Since BmEf!/J,
there exists a set Am E.AI" such that fl (Am 6. Bm) < ~ . Then we have
Am 6. B = (Am n B') U (B n A:")

C (Am n B:") U (B 6. Bm) U (Bm n A:")
= (Am 6. Bm) U (B 6. B m),
and so fl (Am 6. B) < ~ + ~ = c; hence B E{J/J. Thus g> is a monotone

family. Since g> contains the algebra.Al", (21.6) implies that g> contains
.9'(.AI"), and so we have

vIt ::::> f!J ::::> .9'(%) = vIt. 1 0
We can now state and prove a mean convergence version of FUBINI'S
theorem, due to B. JESSEN.
(22.14) Theorem. Let r be an arbitrary infinite index set. For every
finite subset D 01 r,
regard (T, vIt, fl) as (T.o x T.o', vIt.o X vIt.o',
fl.o x fl.o') [making use 01 (22.12)). For 1 ~ P < 00 and f E~p(T, vIt, fl),
define 1.0' on T by
fD' (tD' tD') = f I (u.o, tD') dflD (u.o) .
T.o
That is, ID' is a lunction 01 the lorm SI as in (22.11). Then 1.0' is in
~p(T, vIt, fl) and
(i) lim [[I D' - f I dfl[[p = 0. 2
D T
Also, let
Then 1.0 is in ~p(T, vIt, fl) and

(ii) lim [[ID - f[[p =
D
o.
Proof. (I) We first consider functions I of a very special kind: suppose
that I = ;AnXTn" where AD E vltD . For D::::> Do, we have
~.ro ~.ro 0 0
As a function on T.o, ;AD.XTDnD;XT.o' is merely the characteristic func-

tion of the set A.o. x T.onD;' Thus the integrals in (1) are equal to
flD (AD. x TDnD;l = fl.o. (AD.) = fl (AD. x TD;) = f I dfl; i. e.,
T
Thus (i) is established for our special function I. To establish (ii) for I,
again let D ::::> Do and observe that the integrand in
1.0 (tD' t.o') = Tn'

f ;AD XT.on.o,XT!J' (t.o, UD') dflD'(UD')
0 0
1 Observe that (22.13) holds for any finite measure space (T. vii. t-t) and any
algebra,Al" C vii such that [/ (,AI") = vii.
2 By this limit we mean that for every e > 0, there is an Do C r
such that if
D::::> Do, then IltD' - f f dt-tllp < e. The limit in (ii) has a similar definition.
T
is equal to 1 for all UD' if tDID; EAD.; therefore in this case 1.0 = 1=1.
If tDID. ~ AD., the integrand vanishes and so 1.0 (tD' t D.) = O. Thus we have
and (ii) is established.

ID(tD, tD·) . .
= ~ADXTD·(tD' t D·) ,
(II) To establish (i) and (ii) for all I E~p(T, vii, p), let 5 be the
subset of ~p for which (i) and (ii) are true. We prove first that 5 is a
closed linear subspace of ~p. It is obvious that 5 is a linear subspace, and
so we have to show that it is closed. Suppose that lim II/(n) - Ilip = 0,
~oo
where I(n) E5 for n = 1, 2, 3, .... For every 8 > 0 there is a set Dn such
that "/~) - J I(n) dpllp < 8 (n = 1, 2, 3, ... ), and such that the same
n T
inequality obtains with Dn replaced by any larger finite set. By (22.11),
the inequality II/(n) - Ilip < () implies the inequalities IIf~) - ID·llp < ()
and II/~) - ID·llp < () (n = 1,2,3, ... ) for any () > 0 and any"D. Choose n
so large that 11/(n) - Ilip < 8, and let D ::::> Dn. Then we have
11/.0' - J I dpllp ~ 11/.0' - IW:)IIp + 1I/W:) - J I(n) dpllp

T T
+ I J I(n) dp - J I dpllp
T T
< 8 + 8 + I J I(n) dp - J I dpi
T T
~ 28 + J lI(n) -II dp
T
~ 28 + II/(n) - Ilip < 38 .
Here we have used (13.17) and the fact that IIC11p = ICI for any con-
stant C. Thus the inclusion D::::> Dn implies that 11/.0'- J I dpll p < 38, and
T
as 8 is arbitrary, (i) follows for the function I. The relation (ii) for I is
proved in like manner, and so 5 is closed. By step (I), 5 contains all
functions of the form ~A for A E .AI', and since 5 is closed, Lemma (22.13)
1
and the trivial identity II~Ellp = p(E)P prove that 5 contains all ~B for
BEvil. Thus 5 contains all vii-measurable simple functions, and as these
are dense in ~p (13.20), the proof is complete. 0
(22.15) Note. Theorem (22.14), which is of course two theorems,
tells us all we could hope for about mean convergence of integrals over
partial products of X I'y, either to the integral (22. 14.i) or to the
yEr
integrand (22. 14.ii). For finite products, (22.14.i) becomes trivial, and
(22.14.ii) becomes meaningful [and immediately trivial] only if we agree
that integration for a void set of coordinates does nothing at all.
For rcount ably infinite, the mean convergence of (22.14) can be

replaced by pointwise convergence ,u-almost everywhere. These results
follow readily from (20.56) and (20.59), as we shall now show.
(22.16) Notation. Throughout (22.16) - (22.23), the following notation
will be used. The set r will be {I, 2, 3, ... }, the set Q n will be {I, 2, 3,
... , n} for n EN, and I will be an arbitrary function in ~1(T,..II, ,u).
The function In will be the function If)" of (22.14), i. e.,
(i) In (t) = In (tv' .. , tn, tn+v ... )
= J I(tv ... , tn> Un+!' ... ) d,un' (Un+!, ... ) .
Tn'
. ..
The function I~ will be the function If/ of (22.14), i. e.,

"
(ii) I~ (t) = I~ (t1' ... , tn' tn+!' ... )
= J l(u1, ... , Un' tn+!' ... ) d,un,,(u 1, ... , Un) .
Tn"
(22.17) Theorem [JESSEN]. The relation
(i) lim In (t)
n--->oo
= I (t)
holds lor ,u-almost all t E T.
Proof. We wish to apply the limit theorem (20.56). To do this, con-
sider the a-algebra ..lin.. of subsets of 1;, x ... x Tn as defined in (22.2),
and let ..II(n) be the family of all subsets of T having the form An x Tn' ,
"
where Ann E ..lin". It is evident that ..11(1) C ..11(2) C ... C ..II(n) C .. "
.
that each ..II(n) is a a-algebra, and that ..II is the smallest a-algebra
00
containing U ..II(n). Thus the hypotheses of (20.56) are satisfied,
n~l
where the ..It.. of (20.56) is our present ..II(n). The a-algebra ..1(0) is our
present ..II. As the measure ,u of (20.56) we take our product measure ,u,
and we define the measure 'YJ of (20.56) by
'YJ(A) = J Id,u
A
for all A E..II. [To satisfy the hypothesis that 'YJ be a signed measure,
we must consider first I E~~ (T, ..II, ,u). The complex case obviously
follows at once.J It is trivial that i'YJi < ,u.
Now look at the definition (22.16.i) of In' and use Lemma (22.12)
and (21. 12.iv). These assertions show that In is ..II(n}-measurable. We
must show that In is a LEBESGUE-RADON-NIKODYM derivative of 'YJ(n)
with respect to ,u(n) ['YJ(n) and ,u(n) are restrictions to ..II(n}, as in (20.55)J,
i. e., we must show that
'YJ(A) = J Itl d,u (1)
A
for all A E j(n). Write A = An" x Tn~, where An" E ..II!}.. , Applying
(22.12) and (21.13), we write
l7(A) = J ~A(t) I(t) dp,(t)

T
Since ~An XTn'

~"n .10"'"
(tn, tn')
ft"
= ~An (tn ), the last integral in (2) is equal to
;,.,,"
J ~An,.(tn,,) J I(tnn• tnJ dp,n~(tnJ dp,n,.(tn,,)

Tn,. Tn~
= J ~An (tn..} In (tn,,) dp,n,. (tnn ) . (3)

Tn.. ,.
A similar but simpler computation using (22.12) and (21.13) shows that
the right side of (3) is in fact
J ~A (t) In (t) dp, (t) .
T
Thus (1) holds. Since I is plainly the LEBESGUE-RADON-NIKODYM deriv-

ative of 17 with respect to p, on the a-algebra vIt, we apply (20.56) to
infer that lim In(t) = I(t) p,-a. e. on T. This is (i). 0
n---+oo
(22.18) Exercise. Suppose that 17 is a a-finite measure on T such that

17 ~ p,. Express :: as an iterated limit of functions each of the form
(22. 16.i). [Hints. By (19.24), there is a nonnegative, real-valued, vIt-
measurable function I on T such that J I dp, = 17 (A) for all A EvIt. Let
A
I(k) min {I, k} for all kEN. Define I~") as in (22.16.i) for
= 1("). Then
(22.17) shows that lim I~") = 1(10) p,-a. e. and so
n---+oo
lim [lim Ir:)] = I p,-a. e.]

HOC ~oo
(22.19) Exercise. Let T and P,n be as in (22.9).

(a) Let L be an infinite subset of N, let a be a fixed element of T,
and let B = {t E T: tn = an for all n EL}. Prove that p,(B) = O.
(b) In the notation of (22.9), we have
E(01) Os, ... , 02n-l) = {t ET: tl = ts = ... = t 2n- 1 = O}.
Write Sn for this set. Prove that
lim 2n ~s.. (t)
~oo
= 0
for p,-almost all t ET. [Hints. Let 17 be the product measure on T defined
from measures l7n on each {O, l}n as follows. If n is even, then l7n({O})
= 'l}n ({1}) = ~ . If n is odd, then 'l}n ({O}) = 1 and 'l}n ({1}) = O. Show
that 'l} and '" are mutually singular. Let J{(n) be defined as in (22.16).
Show that 'l} on J{(n) is absolutely continuous with respect to '" on J{(n)
and that 2n ~s" is its LEBESGUE-RADON-NIKODYM derivative. Now
apply (20.56).]
(c) Find a set D of 'l}-measure 1 and ",-measure 0 such that _lim
00
2n~s (t)
..
= 00 'l}-almost everywhere on D. Note that this is consistent with
(20.53.iv).
We now present an important corollary of Theorem (22.17).
(22.20) Definition. Let t = (tn ):'=1 and u = (Un ):'=1 be points in T.
We say that t und u are ultimately equal if there is some no EN such that
tn = Un for all n ~ no.
(22.21) Theorem: The Zero-One Law. Let U be a set in J{ such that lor
all t ET, t is in U il and only il all points u ultimately equal to t are also
in U. Then ",(U) is either 0 or 1.
Proof. Consider the function I = ~u and form the functions In as
in (22. 16.i). For all n and all (tv' .. , tn) and (t;, ... , t~) in TD", the points
(tv' .. , tn, un+!' un+2 , ••• ) and (t;, ... , t~, Un+!' un+2 , ••• ) are ultimately
equal, and so ~u has the same value at these two points. The definition
(22.16.i) shows that In(t) is actually a constant on all of T, and so ~u(t),
which by (22.17.i) is the limit ",-a.e. of In (t), must be a constant ",-a.e.
As ~u assumes only the values 0 and 1, we have", (U) = 0 or '" (U) = 1. 0
With Theorems (22.21) and (20.59), we can prove another pointwise
limit theorem.
(22.22) Theorem [JESSEN]' Notation is as in (22.16). The relation
_00
(i) lim t~ (t) = Tf Id",
holds tor ",-almost all t E T.
Proof. We wish to apply the limit theorem (20.59). To do this,
consider the a-algebra J{D~ of subsets of TD~' and now let J{(n) be the
family of all subsets of T having the form TDn x AD'n AD' EJ{D'. It
for i l"
is evident that J{::> J{(1) ::> J{(2) ::> •.• ::> J{(n) ::> • •• and that each
n
00
J{(n) is a a-algebra of subsets of T. A set U in .-A;, = J{(n) clearly
.. =1
satisfies the hypotheses of (22.21) and so has ",-measure 0 or 1. Define the
measure 'l} on J{ by
'l}(A) = f Id",.
A
Modifying in an obvious way the computation used in proving (22.17),

we see that I~ is J{(n)-measurable and that
'l} (A) = f I~ (t) d",(t)
A
Thus I~ is a LEBESGUE-RADON-NIKODYM derivative of 'YJ with respect to

I-' for JI(n), and (20.59) implies that lim I~(t) = lo(t) exists and is a
_00
LEBESGUE-RADON-NIKODYM derivative of 'YJ with respect to I-' on-Au.

Since 10 is -Au-measurable and I-' assumes only the values 0 and 1 on
-Au, it is easy to see that there is a number a; such that lo(t) = a; for all
t in a set of I-'-measure 1. Thus we have
'YJ(T) = J Idl-' = J lodl-' = a;,
T T
and so (i) is proved. 0
(22.23) Exercise. Let I be an JI-measurable complex-valued func-
tion such that
(i) I (UI> u 2, ••• , Un' tn+1' tn+2 , ••• ) = I (VI> V2, ••• , Vn, tn+1' tn+2, ••• )
for all positive integers n and all choices of UI> U 2, ••• , Un and vI> V 2, ••• , Vn •
Prove that I is a constant I-'-a.e. [Hint. Use the argument of (22.21).]
(22.24) Remarks. Theorems (22.14), (22.17), and (22.22) assume a
particularly simple form for functions that are products of functions
depending on a single coordinate. Making no effort to be exhaustive,
we list a few examples.
(a) Let rbe an arbitrary infinite index set, let Q be a nonvoid finite
subset of r, and let Iy be a function in ~1 (7;, ~,p,y) for each y EQ.
Let g be the function t ~ II Iy(ty) on T. Then we have
yEO
(i) J gdl-' = II J Iydp,y .

T yEO Ty
This follows immediately from (22.14) and (21.13).
r
(b) Let = N = {I, 2, 3, ... }, let In be a function in ~1 (7;., .A(.,I-'n)
P 00
for each n EN, and suppose that lim II In (tn) = II In (tn) exists and is
/>-+00 .. =1 .. =1
00
finite for I-'-almost all (tn) ET. The function g defined by t ~ II In (tn) is
.. =1
certainly JI-measurable. Suppose that g E~1 (T, JI, 1-'). Then we have
P
(ii)
T
J gdl-' = lim II
/>-+00 .. =1 Tn
J Indl-'n .
This too follows at once from (22.14). By applying (12.22), the reader
can easily extend (ii) to the case in which In E~t (7;., .A(., I-'n) and
In ~ 1, with no assumption on g.
(c) A special case of (ii) is the equality
co
(iii) I-'C X An) = II I-'n (An) ,
nEN ..=1
which holds for all sequences (An):=1 of sets such that An E .A(. for all n.
(d) Nothing like (iii) holds for uncountable products even if we

consider th~ completed measure space (T,.ii, P). For example, sup-
pose that F> Ko and Ty = [0, IJ for all y E F. For each y, let Ey be
A-measurable, Ey s:
[0, IJ, and A(Ey) = 1. Then XEy is not measurable
yEr
in the product space in which each coordinate has Lebesgue measure.
For a proof of this rather delicate fact, see HEWITT and Ross, Abstract
Harmonic Analysis I [Springer-Verlag Heidelberg 1963J, p. 228. Also,
for general (Ty, vii,,, It"I) , if r> Ko, if A" EJIy, and It"I (A,,) < 1 for un-
count ably many y's, then pC XAy) = 0. This is very simple to show,
yEr
and we omit the argument.
We continue with a fact (22.26) related to but not dependent on
(22.21), for which an elementary lemma is needed. Lemma (22.26) is of
independent interest and is also needed in the proof of (22.31).
(22.25) Lemma. Suppose that IX" E [0, 1[ for all kEN. Then if
1.: IX" = 1.:
00 00 00
00, we have II (1 - IX,,) = O. If (x" < 00, then we have

k=1 k=1 k=1
00
II(I- IX,,) > O.

k=1
n
Proof. It is obvious that lim II (1 - (X,,) exists. For k = 1, 2, ... ,
11->00 11=1
1
we have (1 - (X,,) (1 + (X,,) = 1 - (X~ ~ 1, so that 1 + (XII ~ 1_ cxl: . Thus
n n
1.: (X" ~ II (1 + (Xk) ~ -n---
k=1 k=1 II (1 - CXk)
k=1
n 00
from which it follows that lim

n~ook=1
II (1 - IXII) = 0 if 1.: IX" =
k=1
00. Suppose
1
1.: IX" < (Xk
00
conversely that 00. Then for k such that ~ 2' we have
k=1
1
1 - IXII ~ 1 + 2 Q(k
and we also have 1 + 2 (X" ~ exp(2(XII) for all k. Hence
k!! (1 - (X,,) ~ exp (- 2 (f (X,,)) ~ exp (- 2 (f (X,,))

00
for a certain m and all n ~ m. This obviously implies that II (1 - (XII)

> O. 0 k=1
(22.26) BOREL-CANTELLI Lemma. Let F = N = {I, 2, 3, ... }. For each

n EN, let E(n) be a set in J(., and write En for the set E(n) x 1{n r. Let
Proof. The set F plainly satisfies the hypotheses of (22.21), and so

fl (F) must be 0 or 1. The present lemma tells us which. It is clear that
fl(F) ~ fl CQn E,,) ~ kf,fl(E,,)

for all n; therefore fl (F) = 0 if n~ fln (E(n») < 00. [A like result holds for
)~ll (kQn Ak) where the A,,'s are measurable sets in any measure space.]
To prove the second statement, we need our special sets En. Observing
that F' = nQI U5n E;) and that E; = (E(k)') x 1{kY, we have
fl (F') = lim fl (
n---+oo
0 E;) = lim fl (C X(E("»,) x
k-n n---+oo k=n
TI x ... x Tn-I)
00 00
= lim II fl,,((E(k»),) = lim II (1 - fl" (E(k»)) .

n-;.oo k=n ~oo k=n
If £ flk (E(k»)
k=1
= 00, then it follows from (22.25) that li (1 -
k=n
fl" (E(k»))
= 0 for n = 1,2,3, ... , and hence that fl (F') = 0; thus fl (F) = 1. 0
Theorem (22.22) has many applications. As an example, we will use
it to prove a famous limit theorem called the strong law large numbers. at
We first present two elementary lemmas.
(22.27) Lemma. Let (fJn);:'=1 be any sequence at complex numbers
having a limit; say lim fJn = fJ. Then we also have
1>-->00
lim
n->oon
2- (fJI + fJ2 + ... + fJn) = fJ .
Proof. Given B> 0, choose no such that k > no implies that IfJ" - fJl
s
< 2. If n > no, then
I n.
Thus for n so large that - E IP~ - PI < 26 , we have
nk=l
I~ k~ P~ - pi < E. 0
(22.28) Lemma. Let (oc~) be a sequence of complex numbers such that

00 1
1: II OC~ converges. Then we have
k=l
Proof. Let So = °and Sn = 1: k k=l

n 1
OC~ for n ~ I, and let tn =
OCt + OC2 + ... + ocn(n ~ I). With this notation we have
oc~ = k (s~ - S~_l) ,

and therefore
n+l n n+l
tn+l = 1: k (s~ - S~_l) = (n + I) sn+l + 1: ks~ - 1: kS~_l
k=l k=l k=2
n n n
= (n + I)sn+l + 1: ks~ - E (k + I)s~ = (n + I)sn+l - 1:s~;
k=l k=l k=l
i.e.,
n
t''l+1 = (n + I) sn+l - 1:s~
k=l
for n = I, 2, 3, .... Thus we have
n~ 1tn+! = sn+! - (n : 1) ~ C~ Sk) .

The partial sum sn+! goes to y E K as n goes to infinity, ~ (
n
£ s~) goes
k=l
to y by (22.27), and lim ~1 = I. Thus we have lim _1-1 tn+! = 0,

n--+oo n + n->-oo n +
as we wished to show. 0
(22.29) Theorem: The Strong Law of Large Numbers. Let r= N. For
each kEN, let g~ be a function in ~2 (T~, Jt~, p.~) such that J g~ dp.~ = 0,
Tk
and suppose that
.x>
(i) 1: k-2 Ilg~ll~ < 00.

k=l
Let f~ be the function on T such that f~(t) = g~(t~). Then we have
(ii) lim
n--+oo
[~f
n k=l
f~(t)] = 0 for almost all t ET.
Proof. We first prove that the series of functions £ ~ converges in

k=1 k
the space ~2 (T,"/{, f-t). If m < n, then we have
£.. kf~ - k=1

II k=1 £m kf~ 1122 = f I £..
k=m+l
kf~ 12df-t = £..
i,k=m+1 J
1 f iJtk-df-t
---:y;
T T
=
H=k
i
.£ j~ fg; df-t; f tk df-tk + k=m+l ;2 flgkl2 df-tk .
7j Tk Tk
We have used (22.24.i) to write the last equality. The first term of the
last expression above is zero by hypothesis, and the second term goes
to zero as m and n go to 00, by (i). Thus we have
so that the partial sums of £

k=l
~ form a Cauchy sequence in ~2 (T,..A; f-t).
Let h be the ~2 limit of i ~ ; then h is also in ~. Theorem (13.17)

k=l
shows that h is also the ~1 sum f ~ .We claim that J hdf-t = O. Write
k=l T
If hdf-tl =
T
IJ: f! tk df-t +f (h - g ! tk) df-tl
T T
~ If (It + ~ t2+ ... + ! tn) df-tl +f !

Ih - kfl tkl df-t.
T T
The first term on the right is zero and the second term goes to zero as
n goes to 00. It follows that J hdf-t = O. We now use Theorem (22.22)
l'
to write
(1)
for almost all t E T. [Recall that Qn= {I, 2, ... , n} and note that the
expression J hdf-tD" is a function of t which is independent of the first n
TD..
coordinates.] We have
The first integral on the right is zero, and the integrand in the second is
independent of the first n coordinates. [Note that

k~n+1
f ~ j" converges in
the ~1 metric to a function in ~l'] The integral
f( 1; ~/,,) dftQn
+I,,;
k~n+1
TQn
is plainly equal to the function f

k~n+1
thus (1) and (2) show that
lim
n-+oo k~n+
£ 1
~ I" (t) = 0 for almost all t ET. That is, the series £
k~1
~ I" (t)
converges a. e. in T. By (22.28), we have
lim ~(fl(t)
n--:;..oon
+ ... + In (t)) = 0
a. e. in T. 0
(22.30) Example. Let 7;" ftn and In be just as in (22.9). We have
11f"J[~ = ~ ,so that the hypotheses of (22.29) are satisfied. Hence we have
lim 11 + ... + In = lim (11 - f) + (t2 - f) + ... + (tn - f) = 0

~oo n n---+oo n
a. e.; i. e.,
lim (tt + t2 + ... + tn _~) = 0
~oo n 2
a. e. Interpreting the occurrence of obtaining heads or tails upon flipping
an unbiased coin as a 0 or 1, respectively, this result says that the portion
of heads [or tails] obtained in n flips goes to ~ as n goes to infinity for
almost all sequences of flips. This result is far stronger than that obtained
in (22.9).
The following is another version of the strong law of large numbers.
(22.31) Theorem. Let r = N. For each kEN, let g" be a junction in
~1 (T", vii", ft,,). Write g,,= If,,+ i1p", where If" and 1p" are real-valued,
and suppose that
(i) the numbers ft"({t,, ET" : If" (t,,) > ce}) and ft"({t,, ET" : 1p,,(t,,) > ce})
are independent 01 k lor every real number ce.
Let I" be the lunction on T such that I,,(t) = g,,(t,,). Then we have
(ii) ~~ [~ /;I/,,(t)] III dft lor almost all t ET.

=
Proof. A moment's reflection shows that there is no harm in sup-

posing that each g" is real, and we do this. For each k, define
I t = {/,,(t) if II" (t) I ;;;; k,
Ik ( ) 0 otherwise.
We will show first that for almost all t E T, there is a positive integer mo
[depending on t] such that
Im(t) = I'",(t) if m ~ mo. (1)
Under the hypothesis (i), it is trivial to show that the numbers
f-lk({tk E T k : Igk(tk)1 > IX}) are independent of k for every IX ER. Using
this fact, we have
00 00
E f-l({t ET: Ik(t) =1= I~(t)}) = E f-l({t ET: IIk(t)1 > k})
k=1 k=l
00
= E f-lk({tk E Tk : Igk(tk)1 > k})

k=1
= 1;1 L~kf-ll({tl E~: n < Igl(tl)1 ~ n+ I})]
00
= Ekf-ll({tlE~:k< Igl(t1)i ~ k+ I})

k=1
(2)
Next write E(k)= {tkETk:lgk(tk)i>k} and E k=E(II)xl{kY. By (2),

00
the series E f-lk (E(II») converges, and so the BOREL-CANTELLI lemma

k=1
(22.26) implies that f-l CQ 1 (kOn E~)) = 1. Thus almost all t E T have
00
the property that tEn
k=n
E~ for some n. This is exactly the assertion (1).
From (1) it is immediate that

1 n 1 n ]
lim [ -Elk(t)--E/~(t) =0 (3)
...-.."" n k=1 n k=1
for almost all t ET.
We wish to apply (22.29) to the functions I;' - J I;' df-l. To establish
(22.29.i), it suffices to show that T
00
E k- 2 III;' - J I~ df-lll~ < 00 • (4)

k=1 T
We first write
J I~ df-lll~ =
E k- 2 [J 1;'2 df-l - (J I;' df-l)2]
"" 00
E k- 2 III;' -
k=1 T 11=1 T T
"" 2
~ E k- J 1;'2 df-l . (5)
11=1 T
Now for each kEN, define the function hk on ~ by

h (t) = {gl (~) if Igl (tl) 1 ~ k ,
k 1 0 otherwise.
From (i), from the primeval definition (12.2) of the integral, and from
(12.21), it is clear that
~OOk_2/ 1~2dll- = f (tl k-2h~) dll- 1 ' (6)

T,
00
We will show that the function w = E k-2h~ is in ~l (7;.). Consider
k=1
any point ~ E7;. such that Igl (tl) I > O. There is a [unique] positive
integer p such that p - 1 < Igdtl)i ~ p. We have h1(t1) = h2 (t1) = ...
= hP- 1 (tl) = 0 and hp (tl) = hp+1 (tl) = ... = gl (~), so that
00 00 00
E k-2 h~ (tl) = E k- 2g~ (tl) ~ Igl (tl) IE k- 2 P . (7)

k=1 k=p k=p
For every positive integer p, the relations
£ k- < f
00
2 dx =~
k=fJ+I 0 (x+ P)2 P
are obvious, and so we have
f
k=p
k- 2 P < ~ +1~ 2. (8)

w~ 21g11.
Since gl E ~l (Tl) by hypothesis, we can retrace our steps (6) and (5) to
see that (4) does hold.
Thus the hypotheses of (22.29) are satisfied for the functions
I~ - JI~ dll-; the conclusion (22.29.ii) assumes here the form
T
lim
....... 00
[~n k=1;1 f~(t) - ~n k=E1 Tff~ dll-] = 0
for almost all t ET. In view of (3), the present proof will be completed
by showing that
(9)
With h" as defined above, we again have
T
J I~ dll- = T,J h" dll-1 ' (10)
and (12.24) implies that

lim
HOO T,
J h" dll-l = T,J gl dll-1 = TJ 11 dll- . (11)
The equality (9) follows from (10), (11), and (22.27). 0

29*
(22.32) Exercise. (a) Prove the following analogue of (22.29). Nota-

tion is as in (22.29). Replace the hypothesis (22.29.i) by
00
(i) E IX,. IIg,.ll~ < 00,

k=1
E IX" <
00
where IX,. > 0 and 00. Then the infinite series

"=1
E IX,. I,,(t)
00
(ii)
k=1
converges for almost all t ET.
(b) Prove the following analogue of (22.29), which is known as
the weak lawai large numbers. Again notation is as in (22.29). Replace
(22.29.i) by the hypothesis
(iii) lim [-\-

H->OO n k=1
i II/"II~] = 0.
Then for every e > 0, the equality
~~fl ({tET: \!ll,,(t)\ > en = 0
holds. That is, the sequence of functions ( n"f

1 n )00
I,. n=1 converges to zero
+1'
in measure.
(22.33) Exercise. Let T..= {O, 1,2, ... , r - I} and let fl .. (A) =
(n = 1,2, ... ). For a fixed l E{O, 1, ... , r - I}, define
lift.. =l,
{
g.. (t) = 0 if t.. =1= l
for all t in the product space T.
Prove that
~fl ({tET: I!,,~ g,,(t) -+1> en = 0

for all e> O.
(22.34) Exercise. For x E]0,1[, a fixed integer r> 1, and
l E{0,1, ... , r - I}, let b" (x) be the number of l's among the numbers
Xl' •.• , X", where
.
00
X =
~
'\' r-" x '
n=1
x.. E{0,1, ... , r - I}, and x.. =1= 0 for infinitely many n's. Prove that
lim ~ b" (x) = ~ for [Lebesgue] almost all x E]0,1[. [This follows
k--->oo k r
from (22.31).]
(22.35) We now present an application of the limit theorem (20.56)
somewhat different from those given above. As in (22.16) and (22.17),
let F={1,2,3, ... }, let Qn={1,2" ... ,n}, and let vIt(n) be the
a-algebra of all sets Aa x Ta, for Aa Evita . Now consider measures II.n
n " r-
and 'YIn on Cr,.,.A;,) such that P,n Cr,.) = 'YIn Cr,.) = 1, and let p, und 'YI be
1& "
the product measures formed from the measures P,n and 'YIn' respectively.
Our first result deals with the case that 'YIn ~ P,n for all n, and establishes
a remarkable fact about 'YI and p,.
(22.36) Theorem [So KAKUTANI]. Notation is as in (22.35). Suppose
that 'YIn ~ P,n lor all n. Then we have either
(i) 'YI ~ P,
or
(ii) 'YI 1. p,.
Let In be a lunction in ~i Cr.., .A;" P,n) such that
(iii) J In dp,n= 'YIn(En)
E"
lor all En E.A;" i. e., let In be a LEBESGUE-RADON-NIKODYM derivative
dd'YJ
n in the sense 01 (19.43). Then (i) holds il and only it
/-In
00 1
(iv) II( J I! dp,,.) > 0,
k=l Tk
and (ii) holds il and only il
00 1
(v) II (J I! dp,k) = O.
k=l Tk
Proof. We observe first of all that

1 1 1 1 1
0< J I! dp,,. ~ [J I,. dp,,.]2 [J 12 dp,,.]2= 'YI,. (T,.)2 p,,.(T,.f2 = 1,
Tk Tk Tk
as (13.4) shows. Hence the infinite product in (iv) and (v) is a number
in [0,1]. For each n EN, consider the finite product Ta.. and the product
measures P,1 x . .. X p,n and 'YIl x ... x 'YIn on vita... It follows from
(21.29) by induction on n that
'YIl x . .. x 'YIn ~ P,1 X •• • X P,n
and that the function
(1)
is a LEBESGUE-RADON-NIKODYM derivative of 'YIl X· •• x 'YIn with
respect to fl-I x ... X p,n' Now consider the a-algebra vIt(n) of all sets
Aa11 x Ta, where Aa1E1
tI
vita" . Let I(n) be the function on T such that
and let p,(n) and 'YI(n) be the measures p, and 'YI, respectively, restricted
to the O'-algebra ,L(n). It is clear from (1) and (22.24.i) that
II(n) df-t =
T
l1 Tk
II,. df-t,. = kiJ 'fJ,.(T,.) = 1 (2)
and that for 1 ~ m < n,

I(n)(t) I(m)(t) = mtl) •.. fm (tm) 1m+! (tm+!) ..• In (tn) . (3)
It is evident from (21.29) that 'fJ(n) ~ f-t(n) and that I(n) is a LEBESGUE-
RADON-NIKODYM derivative of 'fJ(n) with respect to f-t(n). We therefore
cite (20.56) to assert that
lim I(n) (t) = I (t) (4)
........ 00
exists for f-t-almost all t ET and is a derivative of 'fJ with respect to f-t
in the sense of (20.53).
Suppose that (v) holds. Then we apply (22.24.i), (12.23), and (4) to
write
(5)
From (5) it follows that 1= 0 f-t-a.e., and so (20.53) implies that 'fJ .1 "',
since the ",-absolutely continuous part of 'fJ is obtained by integrating I.
Regardless of the value of /100/ Ii1df-t,., we can compute as follows.
Tk
For m < n, we use (13.4) and (2) to write
I I/(m) - I(n) I df-t I I(/(m») 2" + (I(fI») 2"1 I(/(m») 2" -

=
1 1 1 1
(I(n»)2"1 df-t
T T
1 1
~ [/1(/(m»)~ + (I(n»)~12df-tY [/I(I(m»)~ _ (/(n»)~12df-t]2"

T T
2[1 +l(f(m»)~' (f(n»)~ df-t]2" [1 - l(f(m»)~ (I(n»)~ df-tY.

1 1
= (6)
T T
Now taking note of (3) and (22.24.i), we write
l(f(m)r~ (f(n»)i df-t

T = III d"'l X ••• X 11m df-tm X 11!+1 d"'m+! X ••• X II!df-tn
T, T.. Tm+, T..
=
k=m+l
Ii Tk
II: df-t,. . (7)
Combining (6) and (7), we obtain
/1: dft k) 2]
1
/1/(n) - I(m) 1 dft ~ 2 [1- ()j 2. (8)

T k-m+1 n
If (iv) holds, then it is clear that
lim n
m, ....... OOk=m+1
(/I:dftk) = 1.
Tk
Hence (8) shows that (l<n»):=1 is a Cauchy sequence in ~l(T, JI, ft).
We now appeal to (20.58) and (20.57) to infer that 'YJ ~ ft. 0
(22.37) Remarks. Notation is as in (22.35). If not all 'YJn are absolutely
continuous with respect to ft", then 'YJ cannot be absolutely continuous
with respect to ft, but it may still have a large absolutely continuous
part. Suppose that for some lEN, we have
(i) 'YJ! = lX!e! + (1 - IX!) a! •
°
where e! and a! are measures on (~, JI!) such that e!(~) = a!(~) = 1.
e! ~ ft!, a! ..L ft!, and ~ IX! < 1. If IX! = 0, i.e., if 'YJ! ..L ft!, then (21.29)
shows that 'YJ ..L ft. Otherwise, let 'YJ' be the product of all 'YJk for k =1= l,
on the space 1{W. It is easy to see that 'YJ = IX! (e! x 'YJ') + (1 - IX!) (a! x 'YJ').
As (a! x 'YJ') ..L ft. 'YJ cannot be absolutely continuous with respect to ft,
but it is possible for e! x 'YJ' to be singular with respect to ft, absolutely
continuous with respect to ft, or to be "mixed". A precise description
of'YJ in terms of the decomposition (i) for all lEN could be given: we leave
the details to any interested readers.
(22.38) Exercise. Notation is as in (22.35). Let T., = {O, I} for all
n EN, and let /Z be a sequence (a.,):=1 with values in ]0, 1[. Let ft ...
be the measure on (T, JI) that is the product of the measures ft" on
{O, I} such that ft" ({O}) = IX", ft" ({I}) = 1 - IX". Suppose that /Z and p
are any two such sequences.
(a) Prove that exactly one of the two following assertions holds:
(i) ft ... ~ ftp and ftp ~ ft ... ;
or
(ii) ft ... ..L ftp·
Prove also that (i) holds if and only if
00 1 1 1 1
(iii) 1: (1- IX!{J: - (1 - IXn)9(I- (J,,)9) < 00,
n=1
and that (ii) holds if and only if
(iv) the series in (iii) diverges.
[Hints. Apply (22.36) to the measures ft ... and ftp. The factor measures
are evidently absolutely continuous with respect to each other, and the
integral over T., of I! is IX! {J! + (1 - a.,)9(1 - {J,,)9. Now apply (22.29).
1 1 1 1 1
It makes no difference which of ft ... and ftp is taken as ft in (22.35).]

(b) Suppose that for some ~ > the inequalities ~ ~ IX.. ~ 1 - ~°

and ~ ~ fJ.. ~ 1 - ~ hold for all n EN. Prove that (iii) holds if and only if
E
00
(v) (IX.. - fJ..)2 < 00.

n=!
[Hint. Use the identity
I I I 1
(vi) 1 - 1X2 fJ2 - (1- 1X)2 (1 - fJ)2
1 1 I 1 1
= 2 [(1X2 - fJ2)2 + «(1 - 1X)2 - (1 - fJ)2)2]
and the mean value theorem of the differential calculus.J
(c) Suppose that fJ.. is constant: fJ.. = fJ for some fJ EJO, 1[. Show
that (iii) holds if and only if (v) holds. [Hint. If lim IX.. = fJ, then part (b)
........ 00
can be applied. Otherwise (vi) shows that the terms of the series in (iii)
do not have limit O.J
(22.39) Exercise. Prove that there is a set 8 of measures on (R, gj (R))
such that: 0" (R) = 1 for all 0" E8; each 0" E8 is regular; each 0" E8
is continuous; each 0" has support the interval [0, 1]; 0" 1.. 0"' for distinct 0"
and 0"' in 8; and § = c. [Hints. Let T be as in (22.38), and let ffJ (t)
E 2-k tk
00
= for t ET. The mapping ffJ carries Tonto [0, 1]. For every
k=!
number I' E]0, 1[, let fly be the measure on T constructed from the con-
stant sequence (1', 1', 1', ...) as in (22.38). For I' =!= 1", fly and fly' are
obviously mutually singular. For I' E]0, 1[ let ay be the measure on [0, 1J
constructed as in (12.45) and (12.46) from the measure fly on T and the
continuous mapping ffJ. It is simple to verify that {ay : I' E ]0, 1 [} can be
taken for the set 8 of measures.]
(22.40) Exercise. A set 8 1 of measures on gj (R) with all of the pro-
perties of 8 [see (22.39)] except that supports be [0, 1] can be constructed
without recourse to KAKUTANI'S theorem (22.36). Fill in the details of
the following argument. Let T = {O, I}N and consider the measure fll
2
on (T, .-'I) as in (22.39). For each u = (u}> u 2, •••• 1t.. , ••• ) ET, Let ffJu be
the mapping of T into [0, 1] given by
For A Egj(R), let au(A) = fll(ffJ;;l(A

2
n ffJu(T)).
Then the set 8 1 = {au: u E T} of measures has all of the asserted
properties.
(22.41) Exercise. Consider the measure al' constructed as in (22.39).
2
Prove that al is Lebesgue measure on [0, IJ.
2
(22.42) Exercise. (a) Alter the construction of (22.39) in the following

way. Let T and Ily be as in (22.38) and (22.39), but define the mapping
00
q; of T into R by q; (t) = 2 E 3-"t". For I' E]0, 1[, let Ty be the image of
k=l
Ily under q; as in (12.45) and (12.46). Prove that the support of each Ty
is the Cantor ternary set. Prove too that T.! is the Lebesgue-Stieltjes
2
measure that corresponds to LEBESGUE'S singular function (8.28).
(b) Prove that f x dTy(X) = 1 - y. [Hint. The following steps are
[0,1]
easy to check:
f
~,I]
x dTy(X) =
T
f (2 g 3-"t,,) dlly (t) = 211 f t" dlly(t) 3-"
T
00
= 2(1- 1') E 3-" = 1 - 1'.]

k=1
(c) Prove that
f X2 dTy(X) = 2-1 (1 - 1') + 2-1 (1 - 1')2.
[0,1]
[Hints. The computation can be made as follows:
f
[0,1]
x 2dTy(X) = 4
T
f C~3-"t"rdlly(t)
= 4 [g f t"dlly(t) + 21~ C£1 f t"t, dlly(t))]
3-210 3-"-'
T T
= 4 [(1 - 1') k~ 3-210 + 2 (1 - 1')2 I~ 3-111- 1 (J:3-11)]

1 1
="2 (1 - 1') +"2 (1 - 1')11. ]
(d) Prove that
f XSdTy(X) = -2 6
8 [(1- 1') + 1: (1- 1')2 + ! (1- 1')3] .
[0,1]
(22.43) Exercise. Notation is as in (22.42). (a) Let ex be any complex

number. Prove that
00
f exp(exx) dTy(X) = lI[Y + (1 - 1') exp(3-"2ex)].

[0,1] k=1
(b) Prove that
00
f exp(2ni 3Px) dTy(X)

(i) = lI[y + (1 - 1') exp(4ni 3-")]
[0,1] k=1
for all PEN.
(c) Prove that the value of (i) is a positive real number for I' = !.
What can you say for other values of I' in [0, I]?
(22.44) Exercise. (a) Generalize the result of (22.42) in the following

way. Let a be any number in the interval ]- 1, 1 [. Define a map f{J
00
of T into R by cp(t) = }; aktk' Let Wy be the image of /-ty under f{J as in

k~1
(12.45). Thus for every continuous t on R we have

J t dwy = J to f{J d/-ty.
R T
Let 11n = J x" dwy(x) (n = 0,1,2, ... ). Prove that 110 = 1 and
R
(1.) 11 n = a1 _
n
(1 - y) ( ;,
an kJ
(n)j 11 n-)
.) .
1
1~1 ;
[Hints. For n EN, we have
(cp(t))n = an (tl + kf2 ak-ltkf
= an [(~ ak-lik)n +£ (~) i l (~ak-lik)n-i] ,

k-2 1~1 J k-2
since t~ = t i . Integrating over T, we find
11n = an [f (£/k-likf d/-ty(t) + 1~ G) f tl (£2 ak-likf-i d/-ty(t)]

T T
From this, (i) is immediate.]

1
(b) For the case a = y = 2' (22.41) shows that WI = 0"1 = A on
"2 i
[0, IJ. In this case, compute 11n directly and then verify formula (i).
(c) Let Ty be as in (22.42) and write 11~ = J xn dTy(X) (n=O, 1,2, ... ).
[0,1]
Prove that 11~ = 1 and
(ii) 11~= ;n ~ £ (~)2i 11~-i (n = 1,2, ... ).
1~1 J
(22.45) Exercise [H. S. ZUCKERMAN]. Notation IS as III (22.44).
·
F or b revlty, . b f an (1 - y)
wnte n or I-a"
(a) Prove that for n ~ 1,
(i) 11n = bn {I + y/ ( _')' (" _.)~.!

k~ 1 n t 1 • t1 22 ,
.. ("
t k- 1
_.) , . ,"
tk •tk•
bi,bi, .• 'bik} ,
where the sum };' is taken over all k-tuples (iv i 2 , ••• , i k ) of positive
integers such that n > i l > i2 > ... > i k > 0. [Hints. Rewrite (22.44.i)
1 This recursion formula was kindly suggested to us by Professor R. M. BLUMEN-
THAL.
in the form
note that
n!
(n - ill! (il - i.l! (i. - isl! ... (i h - l - ihl! i k ! =
(n)il (il)
i.'"
(ik-l)
ik '
and use induction on n.]

(b) Prove that for n ~ 1,
.. n" n!
1: 1: 11·12···
(11) LIn =
k=l
.,., . , bit+i'+"'+h bi'+i'+"'-I-h bi'+i'+"'+h ••• bik ,
')k'
the sum 1:" being taken over all k-tuples (jI' j2' ... , jk) of positive integers
such thatjI + j2 + . , . + jk = n. [Hint. Rewrite (i).]
Index of Symbols
dE 155 L [Lebesgue integral] 164
arg, Arg 50 lp, lp(D) 194
19 loo (D) 219
30 ~1 173
301 ~~ 170
~p 188
f!I(X) [Borel sets] 132
~oo 347
~(X) 83 ~4> 203
~(H) 251 ~ log+ ~ 203
B.(x) [E-neighborhood] 60 L (f, IX, .1) [Darboux sum] 106
~(A,B) 211
lim, lim 76, 256
c 19 ~iPa; 270
Co (D) 218 A [Lebesgue measure] 120
(£(X) 84 Aa; 120
(£0 (X), (£00 (X) 86
430
Dv D2 271 [,-measurable sets] 128
D" [Dirichlet kernel] 292 [Lebesgue measurable sets]
D+, D+, D-, D_ 257 128
.@([a, b]) 105 viifJ [p,-measurable sets] 127
diam 67 m(D) 219
domf [domain] 7 921 (X) 88
!5xy [KRONECKER'S delta] 11
M(X) 360
t,. [symmetric difference] 4 max {x, y}, min {x, y} 8
E" [E,,(f) = f(a)] 114 max{t, g}, min{f, g} 82
tin 430 161
exp [exponential] 51
Ea [Ea(A) = ~A (a)] 120 N [positive integers] 2
1)", 1)8 366 .t¥,j,% 430
Fa 68 ordA 28
F(X, d, p,) 354 w 28
Ga 68
[J 29
[J [Cr] 429
I [nonnegative linear function-
al on (£00] 114
I 116
P' [ = _P-]
P-l
190
f 118 9i' (X) [all subsets of X] 3

1m [imaginary part] 48
inf 44,69,82 Q [rational numbers] 2
[,(A) = 1(~A)] 120

R [real numbers] 2
337
R" 13
334
R* [extended real numbers] 54
K [complex numbers] 2 Rd [discrete reals] 56
K" 13 Re [real part] 48
K,. [Fejer kernel] 292 rngf [range] 7
Index of Symbols 461
10 [simple functions] 164 A [cardinal number] 19

9' (If) 132
~f(n)
81
S (I) [Riemann integral] 11 0 401
S (I; [a, b]) 107 238
Sa. (I) 109 1 [1(x) = I(-x)] 410
Sa. (I; [a, b]) 107
sgn [signum] 51 Ml. 252
s,,1 291 T* [adjoint operator] 221, 251
ani 292 E*, E** [conjugate spaces] 211
sup 44,69,82
1+ [= max {I, O}] 164
1- [= - min{/, O}] 164
TLl, T 430
1* [/*(x)=/(-x)] 110
U(I, ac, LI) [Darboux sum] 106 /" f+, 1'- 257
1['] 379
ILl, ILl(r), ILl(I) 422
v,.& [total variation] 266
5" [real-valued functions in 5']
82
5'+ [nonnegative real functions in
z [integers] 2 5'] 82
E' 379
~E [characteristic function] 11 A0 [interior] 56
o [void set] 2 A' [complement of A] 4
A- [closure] 56
A" 13
gol 8
Al 12
1* g 396----398,414
v+,v- 307
A"", B 19
A ~ B 27
I~ g 81
It [/e(x) = I(x + t)] 110
I[x] [/[x] (y) = I (x, y)] 125
v <{ P, 312
Ex 379
v 1. P, 326
x 1. y, E 1. F 236
x +A [= {x + y : yEA}] 135
(X, .JII) 149
(X, .JII, p,) 126
A +B [= {x + y: x E A, Y E B}] 135
A-B [={x-y:xEA,yEB}] 135 -A [= {-x:xEA}] 135
Jlx.tV 379 I(a+), I(b-) 111
XxV [Cartesian product] 7 (:) [binomial coefficient] 90
p,xv 384 00, - 00 54
(X,.) [sequence] 10
XA, [Cartesian product] 12
Ilxll 83
'EI
Ilxll [= (x,x)t] 195,235 dv = 10 dp" v = lop, 328
11/11.. 83 :; 328
1I/IIp 188 f I (x) dp,(x) [= Jtdp,= Jtdp,] 170,
IIglloo 346 x x
1I/II!Il 203 312,355
Ilx+MII [norm on ElM] 221 b
fl(x)dx [= f IdA] 170
II Til 210 a [a,b]
11p,11 360 f Idp, [= f ~Eldp,] 175
IITII 354 E X
Ivl [total variation] 307, 308 ~ 10
(x,y) 195, 234 [a, b], ]a, be, etc. 54
(I, g) 195 o [end of proof] 3
Index of authors and terms
Abel summability of Fourier series 301 Baire functions 163
ABEL, N. H. 32 - - of type IX 163
Abelian group 32 Baire sets 164
ABEL'S kernel 407 ball, open 60
absolute continuity of infinite product Banach-Steinhaus theorem 218
measures 453 Banach algebra 84
- - of the integral 176 Banach indicatrix of a function 271
absolute value 48 Banach space 84
absolutely continuous functions 282 -, conjugate of a 211
- - -, composition of 297 -, locally uniformly convex 233
absolutely continuous measure 312 -, products of 217, 221
- - -, derivative of an 328 -, quotient spaces of 221
additive functions on R 49 -, reflexive 215
adjoint of a linear operator 221, 251 -, uniformly convex 232
adjoint space 211 -, uniformly rotund 232
ALEXANDER'S subbase theorem 64 BANACH, S. 80, 137, 215, 261, 263, 288
algebra of operators 251 BANACH'S fixed-point theorem 78
algebra of R, group 399 base for a topology 58
algebra of sets 4 basis for a vector space 18
algebra over a field 82 -, Hamel 18
algebra, Banach 84 -, linear 18
-,linear 82 BERNSTEIN, F. 146
-, normed 83 BESSEL'S inequality 239
algebraic dimension of a vector space 31 binomial coefficients 90
algebras, Boolean 5 binomial series 90
almost everywhere 122, 155 BIRKHOFF, G. 5
- -, locally 122 Birnbaum-Orlicz spaces 203
almost uniform convergence 158 BOHNENBLUST, H. F. 213
analytic sets 133 BOLZANO 66
approximate unit 400 Bolzano-Weierstrass property 62
approximation by simple functions 159 BOOLE, G. 5
approximation theorem, WEIERSTRASS Boolean algebra 5
96 Boolean ring 5
Archimedean ordered field 37 BOREL, E. 66
- - -, complete 44 BOREL-CANTELLI lemma 445
- - -, non-46 Borel measurable function 149, 187
areas, integrals as 392 Borel measures 329
arg 50 - -, complex regular 360
arithmetic means for a Fourier series 292 - - on R, regular 329
arithmetic, cardinal number 23 Borel sets 132
automorphism 33 - - in product spaces 391
automorphisms of the real field 48 bound in an ordered field, greatest
axiom of choice 12 lower 44
- - -, least upper 44
SBr(R). invariant mean on 359 bound, lower 13
Baire category theorem 68, 79 -, upper 13
Index of authors and terms 463
boundary 56 COHEN, P. J. 12,30

bounded functions 83 commutative ring 33
- -, essentially 346 compact space 62
bounded interval 54 -, Frechet 62
bounded linear functionals on Hilbert -, locally 74
spaces 248, 255 -, sequentially 62
bounded linear transformation 210 -,a-125
bounded sequence 38 compact support, continuous functions
bounded set in a metric space 66 with 86
bounded variation, function of 266 complement of a set 4
BUNYAKOVSKII'S inequality 190 - of a subspace, orthogonal 252
complete Archimedean ordered field 44
G:o (X), conjugate space of 364 measure 155
- , ideals in 365 measure space 155
Cantor-Bendixson theorem 72 metric space 67
Cantor ternary set 13, 70 orthonormal set 240, 242
Cantor-like set 70 completeness of £p 192
CANTOR, G. 21, 67 - of the Hermite functions 416
CARATHEODORY, C. 126, 127 - of the trigonometric functions 248
cardinal numbers 19 completion of a measure space 155
- , arithmetic of 23 - - - -, integrals on 186
- -, no largest 21 of a metric space 77
- -, order relation for 20 of a normed linear space 221
cardinality 19 of an ordered field 38
Cartesian product of a family of sets 12 complex conjugate 48
- - of topological spaces 65 complex field, geometric interpreta-
- - of two sets 7 tion of the 50
category theorem, Baire 68, 79 complex measure 304
category, first 68 -, integral relative to a 312
-, second 68 - -, Jordan decomposition of a 309
CAUCHY-BuNYAKOVSKd-sCHWARZ in- - -, total variation of a 308
equality 234 complex measures, LEBESGUE-RADON-
Cauchy sequence 38, 67 NIKODYM theorem for 323
CAUCHY'S inequality 190 complex number field 48
CECH, E. 65 complex number, imaginary part of a
Cesaro summability for Fourier series 48
293 - -, real part of a 48
chain rule 328 complex plane 50
change of variable in integrals 342 component intervals 69
characteristic function of a set 11 composition of absolutely continuous
characteristic of a field 34 functions 297
characters of R 301 composition of relations 8
choice function 12 condition N 288
choice, axiom of 12 cone, convex 219
CLARKSON, J. A. 231 conjugate of a Banach space 211
CLARKSON'S inequality for 2 ~ P < 00 conjugate space 211
225 of G: o (X) 364
- - for 1 < p < 2 227 of £00 357
closed graph theorem 217 of £p 230
closed interval 54 of £1 353
closure of a set 56 -, second 211
coefficients, binomial 90 conjugate, complex 48
-, Fourier 237, 249 connected space 57
464 Index of authors and terms
constant function 82 DE MORGAN'S laws 4

content, Jordan 121 decomposable measures 317
continuity at a point 73 -, LEBESGUE-RADON-NIKODYM
continuity of translation in ~p 199 theorem for 318
continuous function 73 -, Lebesgue decomposition for 341
- -, uniformly 87 decomposition of a complex measure,
continuous functions, absolutely 282 Jordan 309
-, pointwise limits of 79 of a finitely additive signed measure,
that vanish at infinity 86 Jordan 338
- that vanish in a neighborhood of of a measure space 317
infinity 86 of a signed measure, Jordan 307
-, uniformly 87 decomposition theorem, Hahn 306
- with compact support 86 - -, Jordan 266
continuous images of measures 180 - -, Lebesgue 326
continuous measures 334 decreasing function, strictly 105
continuous nowhere differentiable func- DEDEKIND, R. 22
tions 258 Dedekind cut in an ordered field 46
continuous, left III degenerate measure 127, 186
continuous, right III delta, Kronecker 11
continuum 19 Denjoy integral 299
continuum hypothesis 30 dense subset of a topological space 61
contraction mapping 78 dense subsets of ~p 197
density character of a topological space
convergence in measure 156
219
in measure, metric of 182
density of a set with respect to a G-
- in probability 156
algebra 374
- weak, in ~p 206
denumerable set 22
convergence theorem, LEBESGUE'S do-
derivates, Dini 257
minated 172, 174, 185
derivative 257
- -, monotone 172
of a signed measure 366
- -, VITALI'S 203
of an absolutely continuous measure
convergence, almost uniform 158 328
-, weak 233 of an indefinite integral 275
convergent sequence 62 -, LEBESGUE-RADON-NIKODYM 328
convex cone 219 -, left 257
convex functions 202, 271 -, right 257
- -, integral representation of 300 -, symmetric 271
convex set 252 diameter of a set 67
convolution of functions 396 differentiability of monotone functions
coordinate 12 264
coordinate space, projection onto a 65 differentiable function 257
countable set 22 differentiation on a net 373
countably additive measure 126 differentiation theorem, LEBESGUE'S 264
countably infinite set 22 differentiation, term by term 267
counting measure 127 dimension of a Hilbert space, orthogonal
cover 62 247
-, open 62 dimension of a vector space, algebraic 31
-, Vitali 262 - - - - - , linear 31
covering theorem, VITALI'S 262 dimension of ~2' orthogonal 255
cut in an ordered field, Dedekind 46 dimension, Hausdorff 145
Dini derivates 257
DANIELL, P. J. 114 DINI'S theorem 205
Darboux sums 106 Dirac measure 120
Dirichlet kernel 292 field 34

discontinuities of a function, set of 78 - of complex numbers 48
discontinuous measures, purely 334 -, Archimedean ordered 37
discrete metric 59 -, automorphisms of the real 48
discrete topology 56 -, characteristic of a 34
disjoint sets 3 -, complete Archimedean ordered 44
dissection, measurable 164 -, completion of an ordered 38
distance between two sets 78 -, non-Archimedean ordered 46
distance from a point to a set 77 -, ordered 35
distance function 59 -, real number 46
divergent Fourier series 300 filter 358
domain of a relation 7 -, fixed 358
dominated convergence theorem, LE- -, free 358
BESGUE'S 172, 174, 185 finite character, family of 13
DooB, J. L. 373 finite intersection property 62
dual space 211 finite measure 127
DUNFORD, N. 210 finite measure space 127
finite set 21
finite variation, function of 266
EGOROV'S theorem 158
finitely additive measures 126, 354
endpoints 54
- - -, regular 364
enumeration of a set 22
- - -, integral for 355
e-neighborhood 60
finitely additive signed measure, Jordan
equivalence relation 8
decomposition of 338
equivalent sets 19
first category 68
essential supremum 346
first mean value theorem for integrals
essentially bounded functions 346
420
Euclidean metric 59
fixed filter 358
Euclidean n-space 13
fixed-point theorem, BANACH'S 78
evaluation functional 114
FLETT, T. M. 428
exp 51
Fourier coefficients 237, 249
expansions of real numbers 46, 47
Fourier integrals, summability of 404
exponential function 51
Fourier series 245, 291
extended real numbers 54
-, Abel summability of 301
extending a measure 162
-, arithmetic means for a 292
extension of a relation 8
-, Cesaro summability for 293
of Lebesgue measure, invariant 359
-, divergent 300
of the Riemann integral, Lebesgue
-, uniform summability of 300
integral as an 184
Fourier transform 249, 401
theorem, Hahn-Banach 212
theorem, HOPF'S 142 - on.s:!2 411
theorem, KREIN'S 220 - -, inversion of 409
- -, uniqueness theorem for the 408
theorem, TIETZE'S 99
Frechet compact space 62
extensions of Lebesgue measure 137,359
free filter 358
- of measures 141, 148
FUBINI, G. 267
FUB~NI'S theorem 384, 386
Fa set 68 function 9
family of finite character 13 -, absolutely continuous 282
family, monotone 380 -, additive on R 49
FATOU'S lemma 124, 172 -, Baire 163
FEJ:ER, L. 292 -, Banach indicatrix of a 271
Fejer-Lebesgue theorem 294 -, Borel measurable 149,187
Fejer's kernel 292, 408 -, bounded 83
function, choice 12 functionals on ~1 (R), multiplicative

-, constant 82 linear 417
-, continuous 73 functions, convolution of 396
-, -, that vanishes at infinity 86 -, differentiability of monotone 264
-, -, that vanishes in a neighborhood -, Hermite 243 .
of infinity 86 -, integral representation of convex 300
-, -, with compact support 86 -, lattice of 82
-, convex 202, 271 -, N- 288, 303
-, differentiable 257 -, normalized Hermite 245
-, distance 59 -, pointwise limits of continuous 79
-, exponential 51 -, separating family of 93
-, indefinite integral of a 272 -, singular 278
-, integrable 173 -, uniformly integrable sequences of
-, Lebesgue measurable 149 204
-, Lebesgue points for a 278 fundamental theorem of the integral
-, Lebesgue set for a 278 calculus for Lebesgue integrals 285
-, LEBESGUE'S singular 113 G,,-set 68
-, locally integrable 186 GAUSS'S kernel 408
-, locally null 123 generalization of monotone convergence
-, lower semicontinuous 88 theorem 118
-, measurable 149, 154 generalized B. LEVI theorem 118
-, monotone 105 generalized HOLDER'S inequality 200
-, nondecreasing 105 generated by a family, a-algebra 132
-, nonincreasing 105 geometric interpretation of the complex
-, normalized nondecreasing 112, 330 field 50
-, null 123 Gram-Schmidt orthonormalization
- of bounded variation 266 process 240
- of finite variation 266 greatest lower bound in an ordered
-, onto 10 field 44
-, oscillation 78 group 32
-, reflection of a 110 group algebra of R 399
-, Riemann-Stieltjes integrable 106
Hahn-Banach theorem 212
-, section of a 379 Hahn decomposition 305
-, set of discontinuities of a 78 Hahn decomposition theorem 306
-, simple 159 half-open intervals 54
-, step 198 HALMOS, P. 1
-, strictly decreasing 105 Hamel basis 18
-, strictly increasing 105 HAMILTON, SIR W. R. 102
-, summable 173 HARDy-LITTLEWOOD maximal theorem
-, symmetric derivatives of a 271 424,425,426
-, total variation of 266, 270, 272 HARDY, G. H. 258, 422
-, translate of a 110 Hausdorff dimension 145
-, uniformly continuous 87 Hausdorff maximality principle 14
-, upper semicontinuous 88 Hausdorff measure 145
-, value of a 9 Hausdorff space 56
functional, evaluation 114 HEINE-BoREL-BOLZANO-WEIERSTRASS
-, linear 211 theorem 66
-, multiplicative linear 114, 365 HEINE-BOREL theorem 66
-, nonnegative linear 114, 220 Hermite functions 243
-, sublinear 212 - -, completeness of the 416
functionals on Hilbert spaces, bounded - -, normalized 245
linear 248, 255 Hermite polynomials 243
HEWITT, E. 137, 250, 378, 445 infinite products of measure spaces 429
HEWITT, M. 32 infinite set 21
Hilbert parallelotope 254 - -, countably 22
Hilbert space 196, 235 initial segment 28
-, bounded linear functionals on a - - of a well ordered set 16
248,255 inner product space 195, 234
-, orthogonal dimension of a 247 integers 2
-, pre- 195 - , positive 2
-, projections in a 253 integrability, Riemann 105, 183
HIRSCHMAN JR., I. I. 250 integrable function 173
HOLDER'S inequality for 1 < p < 00 190 - -, locally 186
- - for 0 < p < 1 191 - -, Riemann-Stieltjes 106
- -, generalized 200 integral as a set function 175
HOLLADAY, J. C. 103 integral of a function, indefinite 272
HOPF's extension theorem 142 - relative to a complex measure 312
HUNTINGTON, E. V. 5 integral representation of convex func-
tions 300
ideals in a ring 33 integral, absolute continuity of the 176
- in (£0 (X) 365 -, Denjoy 299
- in £l(R) 417 -, derivative of an indefinite 275
identity, PARSEVAL'S 245,246,250 -, iterated 385
-, polar 235 -, Lebesgue 164
image sets, measures on 161 -, Lebesgue integral as an extension
images of measures, continuous 180 of the Riemann 184
imaginary part of a complex number 48 -, linearity of the 169, 171, 174
increasing function, strictly 105 -, Riemann 107
indefinite integral of a function 272 -, Riemann-Stieltjes 106
- -, derivative of an 275 integrals as areas 392
independence, linear 17 for finitely additive measures 355
indicatrix of a function, Banach 271 on infinite product spaces, iterated
indiscrete topology 56 439
induction, transfinite 17 on the completion of a measure space
inequalities, CLARKSON'S 225, 227 186
-, HOLDER'S 190, 191 -, change of variable in 342
-, JENSEN'S 202 -, first mean value theorem for 420
inequality, BESSEL'S 239 -, fundamental theorem of the integral
-, BUNYAKOVSKII'S 190 calculus for Lebesgue 285
-, CAUCHY-BuNYAKOVSKrI-SCHWARZ -, integration by parts for Lebesgue-
234 Stieltjes 419
-, CAUCHY'S 190 -, second mean value theorem for 420
-, generalized HOLDER'S 200 -, summability of Fourier 404
-, MINKOWSKI'S 191 integration by parts 287
-, SCHWARZ'S 190, 234 - - for Lebesgue-Stieltjes inte-
-, YOUNG'S 189 grals 419
infimum in an ordered field 44 by substitution 342
- inR#69 integration, term by term 171, 175
infinite product measures 432 interior of a set 56
-, absolute continuity of 453 intersections of sets 2
-, singular measures induced by interval, bounded 54
456 -, closed 54
-, singularity of 453 -, open 54
infinite product spaces, iterated integrals -, subdivision of an 105
on 439 -, unbounded 54
Hewitt/Stromberg, Real and abstract analysis 30·
intervals, component 69 Lebesgue-Stieltjes integrals, integration

-, half-open 54 by parts for 419
invariant extension of Lebesgue meas- - - measures 330
ure 359 - - outer measures 120
invariant mean on Q3r (R) 359 Lebesgue decomposition for decompos-
inverse of a relation 7 able measures 341
inversion of Fourier transforms 409 decomposition theorem 326
irregular measure 185 definition of measure 121
isolated point 70 differentiation theorem 264
isometry 77 dominated convergence theorem 172,
isomorphic rings 33 174, 185
isomorphism 33 integral 164
-, order 27 - as an extension of the Riemann
iterated integral 385 integral 184
iterated integrals on infinite product -, linearity of the 169
spaces 439 -, fundamental theorem of the in-
JENSEN'S inequalities 202 tegral calculus for 285
JESSEN, B. 373, 375, 439, 441, 443 Lebesgue measurable function 149
JEWETT, R. 1. 101 - - sets 128
Jordan content 121 Lebesgue measure, extensions of 137
Jordan decomposition of a complex - -, invariant extension of 359
measure 309 - -, uniqueness of 185
of a finitely additive signed meas- Lebesgue outer measure 120
ure 338 points for a function 278
of a signed measure 307 set for a function 278
Jordan decomposition theorem 266 singular function 113
JORDAN, C. 121 sums 183
LEBESGUE, HENRI 113, 121, 158, 171,
KAKUTANI, S. 137, 453 264, 267, 276, 277, 294
kernel, ABEL'S 407 left continuous 111
-, DIRICHLET'S 292 left derivative 257
-, FEJER'S 292, 408 left hand limit 111
-, GAUSS'S 408 LEVI theorem, generalized 118
-, POISSON'S 301 LEVI, B. 172
-, positive 400 LEVI'S theorem 172
KONIG'S theorem 27 limit inferior 76
KREIN'S extension theorem 220 - of a sequence 62
KRONECKER'S delta 11 - point 56
lattice of functions 82 -, left hand 111
- of sets 148 -, right hand 111
law of large numbers, strong 447, 449 limits of continuous functions, pointwise
- - - -, weak 452 79
law, zero-one 443 - of measures, setwise 339
LEACH, E. B. 232 linear algebra 82
least upper bound in an ordered field 44 linear basis 18
LEBESGUE-RADON-NIKODYM derivative linear dimension of a vector space 31
328 linear functional 211
theorem 315 -, multiplicative 114, 365
for complex measures 323 -, nonnegative 114, 220
for decomposable meas- -, outer measure induced by a
ures 318 nonnegative 120
for regular measures 323 linear functionals on Hilbert spaces,
-, examples concerning 340 bounded 248, 255
linear functionals on £1 (R), multipli- maximality principle, Hausdorff 14

cative 417 MAYRHOFER, K. 121
linear independence 17 MCSHANE, E. J. 229
linear operator 210 mean value theorem for integrals, first
- -, adjoint of a 221 420
linear ordering 8 - - - - -, second 420
linear space 17 measurable dissection 164
- -, completion of a normed 221 function 149, 154
- -, normed 83 function, Borel 149, 187
linear span 19 function, Lebesgue 149
linear transformation 210 rectangle 379
- -, bounded 210 sections, nonmeasurable sets with 393
linearity of the integral 169, 171, 174 set 127
linearly ordered set 8 measurable sets 138
Lipschitz condition of order CI: 270 - -, Lebesgue 128
LITTLEWOOD, J. E. 422 - -, mappings of 150, 269, 288, 297
£00 spaces 347 measurable space 149
-, conjugate space of 357 measurable subsets, measures on 161
£1 spaces 173 measure 126
£1 (R), ideals in 417 measure space 126
£dR), multiplicative linear functionals -, complete 155
on 417 -, completion of a 155
£" norm 188 -, decomposable 317
- spaces 173 -, decomposition of a 317
-, completeness of 192 -, finite 127
-, conjugate space of 230 -, integrals on the completion of a
-, continuity of translation in 199 186
-, dense subsets of 197 -, nondecomposable 349
-, weak convergence in 206 -, a-finite 127
£1> conjugate space of 353 measure spaces, infinite products of 429
£., Fourier transform on 411 measure, Borel 329
-, orthogonal dimension of 255 -, complete 155
locally almost everywhere 122 -, complex 304
locally compact space 74 -, complex regular Borel 360
locally integrable function 186 -, convergence in 156
locally null function 123 -, countably additive 126
locally null set 122, 346 -, counting 127
locally uniformly convex Banach spaces -, decomposable 317
233 -, degenerate 127, 186
lower bound 13 -, derivative of a signed 366
lower semicontinuous function 88 -, derivative of an absolutely continu-
LUZIN, N. N. 159, 288 ous 328
LUZIN's theorem 159 -, Dirac 120
-, extending a 162
mapping 9 -, extensions of Lebesgue 137
mapping, contraction 78 -, finite 127
mappings of measurable sets 150, 269, -, finitely additive 126
288,297 -, Hausdorff 145
martingale in the wide sense 253 -, integral relative to a complex 312
martingale theorems 369, 371, 375 -, invariant extension of Lebesgue 359
maximal element 13 -, irregular 185
maximal theorem, HARDy-LITTLEWOOD -, Jordan decomposition of a complex
424, 425, 426 309
measure, Jordan decomposition of a monotone convergence theorem 172

finitely additive signed 338 - - -, generalization of 118
-, Jordan decomposition of a signed 307 monotone family 380
-, LEBESGUE'S definition of 121 monotone function 105
-, metric of convergence in 182 - -, differentiability of 264
-, negative variation of a signed 307 monotone sequence in R=iI= 76
- nonnegative set for a signed 305 multiplication, scalar 17
-, nonpositive set for a signed 305 multiplicative linear functionals 114,365
-, outer 126 - - - on 2 1 (R) 417
-, positive variation of a signed 307
-, Radon 114 N-functions 288, 303
-, regular 177, 185 n-space, Euclidean 13
-, regular finitely additive 364 -, unitary 13
-, regular outer 143 negative variation of a signed measure
-, a-finite 127 307
-, signed 304 neighborhood 56, 60
-, support of a 122 net, differentiation on a 373
-, total variation of a complex 308 non-Archimedean ordered field 46
-, total variation of a signed 307 nondecomposable measure space 349
-, uniqueness of Lebesgue 185 non decreasing function 105
measures on image sets 161 - -, normalized 112, 330
on measurable subsets 161 nonincreasing function 105
- on nonmeasurable subsets 161 nonmeasurable sets 135, 146
- on R, regular Borel 329 - - with measurable sections 393
-, absolutely continuous 312 - -, measures on 161
-, continuous 334 nonnegative linear functional 114, 220
-, continuous images of 180 - - -, outer measure induced by a
-, extensions of 141, 148 120
-, finitely additive 354 nonnegative set for a signed measure 305
-, infinite product 432 nonpositive set for a signed measure 305
-, integrals for finitely additive 355 norm 83
-, Lebesgue-Stieltjes 330 -, 2/> 188
-, Lebesgue decomposition for decom- -,2 00 346
posable 341 -, operator 210
-, metric outer 144 -, total variation 271
-, product of two 384 -, uniform 83
-, purely discontinuous 334 normal numbers 452
-, regularity of product 389 normalized Hermite functions 245
-, setwise limits of 339 - nondecreasing function 112, 330
-, singular 326 normed algebra 83
-, zero-one 358 normed linear space 83
metric 59 - - -, completion of a 221
-, discrete 59 nowhere dense set 68
-, Euclidean 59 nowhere differentiable functions, con-
- of convergence in measure 182 tinuous258
- outer measures 144 null function 123
metric space 59 function, locally 123
-, bounded set in a 66 sequence 38
- -, complete 67 set 122
- -, completion of a 77 set, locally 122, 346
metric topology 60 number, cardinal 19
minimal element 13 -, ordinal 28
MINKOWSKI'S inequality 191 numbers, complex 2
numbers, expansion of real 46, 47 oscillation function 78

-, extended real 54 outer measure 126
-, field of complex 48 - induced by a nonnegative linear
-, normal 452 functional 120
-, order relation for cardinal 20 -, Lebesgue 120
-, rational 2 -, Lebesgue-Stieltjes 120
-, real 2 -, metric 144
-, strong law of large 447, 449 -, regular 143
-, weak law of large 452 OXTOBY, J. C. 137
one-to-one relation 9 parallelogram law 235

onto function 10 parallelotope, Hilbert 254
open ball 60 PARSEVAL'S identity 245, 246, 250
- cover 62 partial ordering 8
- interval 54 partially ordered set 8
open mapping theorem 215 PEANO, G. 9
open set 55 perfect set 70
- - in R 55, 69 perfect sets, uncountability of 72
open, relatively 61 permutation 30
operator norm 210 PHILLIPS, K. L. 429
operator, linear 210 Plancherel transform 411
-, adjoint of an 221, 251 PLANCHEREL'S theorem 411
operators, algebra of 251 point mass, unit 120
order isomorphism 27 point, continuity at a 73
order relation for cardinal numbers 20 -, isolated 70
order topology 79 -, Lebesgue 278
order type 27 -, limit 56
ordered field 35 - of density of a set 274
-, Archimedean 37 pointwise limits of continuous functions
- , complete Archimedean 44 79
-, completion of an 38 pointwise operations and relations 82
-, Dedekind cut in an 46 POISSON'S kernel 301
-, greatest lower bound in an 44 polar identity 235
-, infimum in an 44 polynomials, Hermite 243
-, least upper bound in an 44 -, trigonometric 248
-, non-Archimedean 46 PORETSKY 6
-, supremum in an 44 positive kernel 400
ordered set, linearly 8 positive variation of a signed measure
- -, partially 8 307
- -, well 8 power set 3
ordering, linear 8 pre-Hilbert space 195
-, partial 8 product measures 384
-, well 8 -, absolute continuity of infinite
ordinal number 28 453
Orlicz spaces 203 -, infinite 432
orthogonal complement 252 -, regularity of 389
dimension of a Hilbert space 247 -, singular measures induced by
- dimension of ~2 255 infinite 456
- vectors 236 -, singularity of infinite 453
orthonormal set 236 product of a family of sets, Cartesian 12
- -, complete 240, 242 product of topological spaces, Cartesian
orthonormalization process, Gram- 65
Schmidt 240 product of two sets, Cartesian 7
product a-algebra 379 Riemann-Lebesgue lemma 249, 401

product topology 65 Riemann-Stieltjes integrable function
products of Banach spaces 217,221 106
products of measure spaces, finite 384 Riemann-Stieltjes integral 106
- - - -, infinite 429 Riemann integrability 183
projection onto a coordinate space 65 Riemann integral 107
- in a Hilbert space 253 - -, Lebesgue integral as an extension
proper subset of a set 2 of the 184
purely discontinuous measure 334 Riesz-Fischer theorem 250
Pythagorean theorem 236 Riesz representation theorem 177, 364
RIEsz, F. 156, 230
quaternions 102 right continuous 111
quotient spaces of Banach spaces 221 right derivative 257
right hand limit 111
RADON-NIKODYM theorem [see LE- ring 33
BESGUE-RADON-NIKODYM Theorem] - of sets 4
Radon measure 114 - of sets, a- 4
RADON-RIESZ theorem 233 ring, Boolean 5
RADON, J. 422 -, commutative 33
range of a relation 7 Ross, K. A. 137, 378, 445
real number field 46 RUDIN, W. 98, 419
- - -, automorphisms of the 48 RUZIEWICZ. S. 297
real numbers 2 R, additive functions on 49
real numbers, expansions of 46, 47 -, characters of 301
- -, extended 54 -, group algebra of 399
real part of a complex number 48 -, open set in 55
rectangle, measurable 379 -, regular Borel measures on 329
reflection of a function 110 -, structure of open sets in 69
reflexive Banach space 215 -, usual topology for 56
regular Borel measure, complex 360
regular Borel measures on R 329
regular finitely additive measure 364 SAKS, S. 133, 297, 300, 328
regular measure 177, 185 scalar multiplication 17
- -, LEBESGUE-RADON-NIKODYM Schroder-Bernstein theorem 20
theorem for 323 SCHWARTZ, J. 210, 353
regular outer measure 143 SCHWARZ'S inequality 190, 234
regularity of product measures 389 second category 68
relation 7 second conjugate space 211
-, domain of a 7 second mean value theorem for integrals
-, equivalence 8 420
-, extension of a 8 section of a function 379
-, inverse of a 7 - of a set 379
-, one-to-one 9 segment [see initial segment]
-, range of a 7 semicontinuous function, lower 88
-, restriction of a 8 - -, upper 88
- , single valued 9 separable space 61
relations, composition of 8 separating family of functions 93
-, pointwise operations and 82 sequence 10
relative topology 60 -, bounded 38
relatively open 61 -, Cauchy 38, 67
representation theorem, F. RIEsz's 177, -, convergent 62
364 -, limit of a 62
restriction of a relation 8 -, monotone 76
sequence, null 38 sets, distance between two 78

-, term of a 10 -, equivalent 19
sequences of functions, uniformly inte- -, intersections of 2
grable 204 -, lattice of 148
sequences, ultimately equal 443 -, Lebesgue measurable 128
sequentially compact space 62 -, mappings of measurable ISO, 269,
series, Abel summability of Fourier 301 288, 297
-, arithmetic means for a Fourier 292 -, measurable 138
-, binomial 90 -, nonmeasurable 135, 146
-, Cesaro summability for Fourier 293 -, ring of 4
-, divergent Fourier 300 -, a-algebra of 4
-, Fourier 245, 291 -, a-ring of 4
-, uniform summability of Fourier 300 -, symmetric difference of two 4
set 1 -, union of 2
-, Cantor 13, 70 setwise limits of measures 339
-, Cantor-like 70 sgn 52
set, complement of a 4 SIERPINSKI, W. 133, 147
-, complete orthonormal 240, 242 a-algebra 4
-, convex 252 - generated by a family 132
-, countable 22 -, product 379
-, countably infinite 22 a-compact set 138
-, denumerable 22 - - space 125
-, diameter of a 67 a-finite measure 127
-, distance from a point to a 77 - measure space 127
-, enumeration of a 22 - set 138
-,Fu 68 a-ring 4
-, finite 21 signed measure 304
-, Cd 68 -, derivative of a 366
-, infinite 21 -, Jordan decomposition of a 307
-, linearly ordered 8 -, Jordan decomposition of a fi-
-, locally null 122, 346 nitely additive 338
-, measurable 127 -, negative variation of a 307
-, nowhere dense 68 -, nonnegative set for a 305
-, null 122 -, non positive set for a 305
-, open 55 -, positive variation of a 307
-, orthonormal 236 -, total variation of a 307
-, partially ordered 8 signum 52
-, perfect 70 simple functions 159
-, power 3 - -, approximation by 159
-, section of a 379 single valued relation 9
-, a-compact 138 singleton set 2
-, a-finite 138 singular function 278
-, singleton 2 singular function, LEBESGUE'S 113
-, subset of a 2 singular measures 326
-, uncountable 22 - - induced by infinite product
-, void 2 measures 456
-, well-ordered 8 singularity of infinite product measures
sets, algebra of 4 453
-, analytic 133 SOBCZYK, A. 213
-, Baire 164 space, adjoint 211
-, Borel 132 -, Banach 84
-, Cartesian product of 7, 12 -, compact 62
-, disjoint 3 -, complete metric 67
space, conjugate 211 subset of a set 2

-, conjugate of a Banach 211 - - - -, proper 2
-, connected 57 subspace of a topological space 60
-, dual 211 subspace, orthogonal complement of a
-, Euclidean 13 252
-, Fn!chet compact 62 substitution, integration by 342
-, Hausdorff 56 SUHOMLINOV 213
-, Hilbert 196, 235 summability of Fourier integrals 404
-, inner product 195, 234 summability of Fourier series, Abel 301
-, linear 17 - - - -, Cesaro 293
-, locally compact 74 - - - -, uniform 300
-, measurable 149 summable function 173
-, measure 126 sums, Darboux 106
-, metric 59 -, Lebesgue 183
-, normed linear 83 SUPPES, P. 1
-, pre-Hilbert 195 support of a measure 122
-, reflexive Banach 215 supremum in an ordered field 44
-, second conjugate 211 - inR# 69
-, separable 61 -, essential 346
-, sequentially compact 62 symmetric derivative of a function 271
-, a-compact 125 symmetric difference of two sets 4
-, a-finite measure 127
-, topological 55
TARSKI, A. 22
-, unitary 13
term by term differentiation 267
-, vector 17, 31
- - - integration 171, 175
spaces, Birnbaum-Orlicz 203
term of a sequence 10
-, Cartesian product of topological 65
ternary set, Cantor 13, 70
-,£",,347 theorem, ALEXANDER'S subbase 64
-, £1 173 -, B. LEVI'S 172
-, £1> 173 -, Baire category 68, 79
-, uniform 87 -, Banach-Steinhaus 218
span 19 -, BANACH'S fixed-point 78
-, linear 19 -, Cantor-Bendixson 72
Steinhaus theorem 143 -, closed graph 217
STEINHAUS, H. 143 -, DINI'S 205
SPARRE ANDERSEN, E. 373 -, EGOROV'S 158
step function 198 -, Fejer-Lebesgue 294
Stieltjes [see Riemann-Stieltjes and -, FUBINI'S 384, 386
Lebesgue-Stieltj es] -, generalized B. LEVI 118
STONE-WEIERSTRASS theorem 94, 95 -, Hahn-Banach 212
- - -, complex version of 97 -, Hahn decomposition 306
- - -, other versions of 98 -, HARDy-LITTLEWOOD maximal 424,
STONE, M. H. 5, 90 425, 426
strictly decreasing function 105 -, HEINE-BOREL 66
strictly increasing function 105 -, HOPF's extension 142
strong law of large numbers 447, 449 -, Jordan decomposition 266
subbase for a topology 58 -, KONIG'S 27
subbase theorem, ALEXANDER'S 64 -, KREIN'S extension 220
subcover 62 -, LEBESGUE-RADON-NIKODYM 315,
subdivision of an interval 105 318,323
sublinear functional 212 -, LEBESGUE'S decomposition 326
subsequence 62 -, LEBESGUE'S differentiation 264
theorem, LEBESGUE'S dominated con- ultimately equal sequences 443

vergence 172, 174, 185 ultrafilter 358
-, LUZIN'S 159 unbounded interval 54
-, monotone convergence 172 uncountable set 22
-, open mapping 215 uniform boundedness principle 217
-, PLANCHEREL'S 411 uniform norm 83
-, RADON-RIESZ 233 uniform spaces 87
-, Riesz-Fischer 250 uniform summability of Fourier series
-, Riesz representation 177,364 300
-, Schroder-Bernstein 20 uniformly continuous function 87
-, Steinhaus 143 uniformly convex Banach spaces 232
-, STONE-WEIERSTRASS 94, 95, 97, 98 uniformly integrable sequences of func-
-, TIHONOV'S 65 tions 204
-, TIETZE'S extension 99 uniformly rotund Banach spaces 232
-, VITALI'S convergence 203 unions of sets 2
-, VITALI'S covering 262 uniqueness of Lebesgue measure 185
-, WEIERSTRASS approximation 96 uniqueness theorem for Fourier trans-
-, well-ordering 14 forms 408
TIHONOV, A. 65 unit of a ring 33
TIHONOV'S theorem 65 unit point mass 120
TIETZE'S extension theorem 99 unit, approximate 400
topological space 55 unitary n-space 13
-, dense subset of a 61 upper bound 13
- -, density character of a 219 upper semicontinuous function 88
- -, subspace of a 60 URYSOHN, P. 75
topological spaces, Cartesian product of URYSOHN'S lemma 75
65 usual topology for R 56
topology 55 R# 56
topology for R#, usual 56 - - - R" or K" 60
- - R" or K", usual 60
- - R, usual 56 value of a fUllction 9
topology, base for a 58 variation [see total variation]
-, discrete 56 variation of a signed measure, negative
-, indiscrete 56 307
-, metric 60 - - - - -, positive 307
-, order 79 - - - - -, total 307
-, product 65 vector space 17, 31
-, relative 60 -, algebraic dimension of a 31
-, subbase for a 58 - -, basis for a 18
total variation norm 271 - -, linear dimension of a 31
total variation of a complex measure 308 Vitali cover 262
- - - - function 266, 270, 272 VITALI'S convergence theorem 203
- - - - signed measure 307 VITALI'S covering theorem 262
transfinite induction 17 void set 2
transform, Fourier 249, 401
-, Plancherel 411 weak convergence 233
transformation 9 weak convergence in ~p 206
-, bounded linear 210 weak law of large numbers 452
-, linear 210 WEIERSTRASS, K. 66, 90, 258
translate of a function 110 WEIERSTRASS approximation theorem
translation in ~P' continuity of 199 96
trigonometric polynomials 248 well-ordered set 8
TUKEY'S lemma 14 - - -, initial segment of a 16
well-ordering theorem 14 ZERMELO, E. 12, 14

zero-one law 443
YOUNG, W. H. 414 zero-one measures 358
YOUNG'S inequality 189 ZORN'S lemma 14
ZUCKERMAN, H. S. 458
ZAANEN, A. C. 203 ZVGMUND, A. 203, 292, 298

Real and Abstract Analysis - A Modern Treatment of The Theory of Functions of A Real Variable (PDFDrive)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Real and Abstract Analysis - A Modern Treatment of The Theory of Functions of A Real Variable (PDFDrive)

Uploaded by

Copyright:

Available Formats

A modern treatment of the theory

Springer-Verlag Berlin Heidelberg GmbH

All rights, especially that of translation into foreign languages, reserved.

© by Springer-Verlag Berlin Heidelberg 1965

Title No. 1320

Chapter Two: Topology and Continuous Functions 53

Chapter Three: The Lebesgue Integral . . . 104

Chapter Four: Function Spaces and Banach Spaces. 188

Chapter Five: Differentiation . . . . 256

Chapter Six: Integration on Product Spaces . . . . . . 377

Index of Symbols . . . . . 460

Index of Authors and Terms 462

bers, is indispensable to the training of every mathematician. All five of

analysis. Section 15, which is an exercise in classical analysis, should be

Seattle, Washington EDWIN HEWITT

Eugene, Oregon KARL R. STROMBERG

§ 1. The algebra of sets

A set is often defined by some property of its elements. We will write

Example. If A ... = {x: x is a real number, Ixl < ~}, n = 1,2, 3,

For a set A , the family of all subsets of A is a well-defined family of

An (D UC) (A na) u (A nt:)

A similar schematic procedure could be applied to (ii). However, we may

the complement of A [relative to XJ to be the set {x : x EX, x ~ A}. This

sets Bf such that if {An: n EN} CBf, then U An EBf.

(1.15) Exercise. Simplify as much as possible:

(b) Generalize this identity to v(A U B U C) and to v(A U B U CUD).

(2.4) Definition. Let / and g be relations. We define the composition

(b) Let P be the set of all nonnegative rational numbers:

01 subsets 01 Y. For A e X we write A' lor the complement 01 A relative

is called the characteristic lunction 01 E. It will always be clear from the

§ 3. The axiom of choice and some equivalents

an - I = bn - I , and an =!= bn ; and e(a, b) = if a = b. Under this metric on

(3.7) Lemma. Let ~ be a family of finite character and let 1B be a chain

every I-inductive family.J'}. It is easy to see that .J'o is I-inductive. Thus

which is impossible. Thus there is some A E fJI such that B C A. If B

(3.14) Theorem [Principle of Transfinite Induction]. Let (W, ~) be a

1 The reader will find a discussion of groups, rings, and fields in § 5.

§ 4. Cardinal numbers and ordinal numbers

(4.6) Theorem. Let u, 1), and w be cardinal numbers. Then:

Since 5 c U [perhaps 5 = 0J, we have 5 E [ljJ(U). Thus, because h

(4.13) Theorem. Let S be a set. Then:

Then the family {Bn}nEN is a pairwise disjoint family of subsets of A,

so each Bn is nonvoid. Apply the axiom of choice to {Bn}nE N to get a

A;;:;; NxN= Ko.

We next introduce arithmetical operations for cardinal numbers.

then define Ao = {(a, 0) : a E A} and Bo = {(b, I) : bE B} to obtain

(4.26) Theorem. Let

is a one-to-one function I with doml = C and mgl = Cx {O, I}. This

(4.34) Exercise. Let a be any cardinal number such that 2 ~ a ~ c.

if A ~ B. If A ~ B, we say that A and B are order isomorphic or have

initial segment of B or B itself [we may obviously suppose that

Clearly ex ~ {J for each ex E C for otherwise (J E D. This proves (iv). 0

functions! such that:

Since B is linearly independent, bo is not a linear combination of elements

(4.60) Exercise. Let X be a vector space over a field F and let B

§ 5. Construction of the real and complex number fields

We now define an important special type of ring.

In additive notation, these sets are written as A + E, -A, and 2A.

hypothesis. If bE - P, then - bE P and

Many elementary facts about inequalities are consequences of the

(5.16) Theorem. For all elements a, b in an ordered field F, we have:

Using (5.14), we have