P. 1
H.L. Royden - Real Analysis

H.L. Royden - Real Analysis

|Views: 1,915|Likes:
Published by boraakdeniz

More info:

Published by: boraakdeniz on Apr 03, 2011
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

05/07/2013

pdf

text

original

H. L. Royden

STANFORD

UNIVERSITY

Real Analysis
SECOND EDITION

The Macmillan Company Collier-Macmillan Limited, London

Preface

This book is the outgrowth of a course at Stanford entitled "Theory of Functions of a Real Variable," which I have given from time to time during the last ten years. This course was designed for first-year graduate students in mathematics and statistics. It presupposes a general background in undergraduate mathematics and specific acquaintance with the material in an undergraduate course on the fundamental concepts of analysis. I have attempted to cover the basic material that every graduate student should know in the classical theory of functions of a real variable and in measure and integration theory, as well as some of the more important and elementary topics in general topology and normed linear space theory. The treatment of material given' here is quite standard in graduate courses of this sort, although Lebesgue measure and Lebesgue integration are treated in this book before the general theory of measure and integration. I have found this a happy pedagogical practice, since the student first becomes familiar with an important concrete case and then sees that much of what he has learned can be applied in very general situations. The preparation of a revised edition has given me the opportunity to expand points that students have found difficult and to improve a number of awkward proofs. Except for the addition of problems, little change has been made in the first seven chapters. The remaining chapters in Part Two have been enlarged by the addition of sections on products of topological spaces, the Stone-Cech compactification, and topological vector spaces. Part Three, on the other hand, has been considerably revised. The approach to the abstract Lebesgue integral in Chapter 11 has been altered slightly to make the development parallel to the one used in Chapter 13. The proof of the Radon-Nikodym theorem has been simplified, and a section on LP spaces has been added. Chapter 12 contains new material on inner
vii

viii

Preface

measure and its role in extension and uniqueness theorems. Chapter 13 on the Daniell integral has been rearranged to correct several erroneous proofs that made tacit use of things not proved until later. A new chapter on measure theory in locally compact spaces has been added. There is considerable independence among chapters, and the chart on page 4 gives the essential dependencies. The instructor thus has considerable freedom in arranging the material here into a course according to his taste. Sections that are peripheral to the principal line of argument have been starred (*). The Prologue to the Student lists some of the notations and conventions and makes some suggestions. The material in this book belongs to the common culture of mathematics and reflects the craftsmanship of many mathematicians. My treatment of it is particularly indebted to the published works of Constantine Caratheodory, Paul Halmos, and Stanislaw Saks and to the lectures and conversations of Andrew Gleason, John Herriot, and Lynn Loomis. Chapter 15 is the result of intensive discussion with John Lamperti. I also wish to acknowledge my indebtedness for helpful suggestions and criticism from numerous students and colleagues. Of the former I should like to mention in particular Peter Loeb, who read the manuscript of the original edition and whose helpful suggestions improved the clarity of a number of arguments, and Charles Stanton, who read the manuscript for this revision, correcting a number of fallacious statements and problems. Among my colleagues, particular thanks are due Paul Berg, who pointed out Littlewood's "three principles" to me, to Herman Rubin, who provided counterexamples to many of the theorems the first time I taught the course, and to John Kelley, who read the manuscript, giving helpful advice and making me omit my polemical remarks. (A few have reappeared as footnotes, however.) Finally, my thanks go to Margaret Cline for her patience and skill in transforming illegible copy into a finished typescript for the original edition, to William Glassmire for reading proofs of the revised edition, to Valerie Yuchartz for typing the material for this edition, and to the editors at Macmillan for their forbearance and encouragement during the years in which this book was written.
H.L.R.

Stanford, California

Contents

Prologue to the Student

1

1 Set Theory
1 Introduction, 5

5

2 3 4 5 6 7 8 9

Functions, 8 Unions, intersections, and complements, 11 Algebras of sets, 16 The axiom of choice and infinite direct products, 18 Countable sets, 19 Relations and equivalences, 22 Partial orderings and the maximal principle, 23 Well ordering and the countable ordinals, 24

Part One

0

Theory of Functions of a Real Variable

2 The Real Number System
1 Axioms for the real numbers, 29 2 The natural and rational numbers as subsets of R, 32 3 The extended real numbers, 34 4 Sequences of real numbers, 35 5 Open and closed sets oj real numbers, 38 6 Continuous junctions, 44 7 Borel sets, 50 ix

29

x

Contents

3

Lebesgue Measure
I Introduction, 52 2 Outer measure, 54 3 Measurable sets and Lebesgue measure, 56 *4 A nonmeasurable set, 63 5 Measurable functions, 65 6 Littlewood's three principles, 71

52

4

The Lebesgue Integral 1 The Riemann integral, 73
2 The Lebesgue integral of a bounded function measure, 75 3 The integral of a nonnegative function, 82 4 The general Lebesgue integral, 86 *5 Convergence in measure, 91 over a set of finite

73

5

Differentiation 2 3 4 *5

and Integration
of monotone junctions, 94

94

1 Differentiation

Functions of bounded variation, 98 Differentiation of an integral, 101 Absolute continuity, 104 Convex junctions, 108

6 The Classical Banach Spaces
I The LP spaces, III 2 The Holder and Minkowski inequalities, 112

III

3 Convergence and completeness, 115 4 Bounded linear [unctionals on the LP spaces, 119

Part Two

0

Abstract Spaces
127
127 131

7

Metric Spaces
1 Introduction,

2 Open and closed sets, 129
3 Continuous/unctions and homeomorphisms, 4 Convergence and completeness, 133 5 Uniform continuity and uniformity, 135 6 Subspaces, 137 7 Baire category, 139

'"Indicates sections peripheral to the principal line of argument.

Contents

xi

8 Topological Spaces
1 Fundamental notions, 142

142

Bases and countability, 145 The separation axioms and continuous real-valued/unctions, 147 Product spaces, ISO Connectedness, 150 Absolute g~'s, 154 *7 Nets, 155

2 3 4 5 *6

9 Compact Spaces
1 Basic properties, 157

157

2 Countable compactness and the Bolzano- Weierstrass property, 159 3 Compact metric spaces, 163 4 Products 0/ compact spaces, 165 5 Locally compact spaces, 168 *6 The Stone-Cech compactification, 170 7 The Stone-Weierstrass theorem, 171 *8 The Ascoli theorem, 177

10 Banach Spaces
1 Introduction, 181 2 Linear operators, 184 3 Linear functionals and the Hahn-Banach theorem, 186 4 The closed graph theorem, 193 *5 Topological vector spaces, 197 *6 Weak topologies, 200 *7 Convexity, 203 8 Hilbert space, 210

181

Part Three

0

General Measure and Integration Theory
217

11 Measure and Integration
1 Measure spaces, 217 2 Measurable functions, 223 3 Integration, 225 *4 General convergence theorems, 231 5 Signed measures, 232 6 The Radon-Nikodym theorem, 238 7 The LP spaces, 243

on ceX). 288 286 3 Uniqueness. 261 4 Product measures. 283 *7 Caratheodory 13 The Daniell Integral 1 Introduction. 253 *3 The Lebesgue-Stieltjes 250 250 integral.xii Contents 12 Measure and Outer Measure 1 Outer measure and measurability. 313 15 Mappings of Measure Spaces 1 Point mappings and set mappings. 295 14 Measure and Topology 1 Baire sets and Borel sets. 2 The extension theorem. 317 2 Measure algebras. 294 4 Measurability and measure. 319 3 Borel equivalences. 328 5 The isometries oj LP. 274 *6 Extension by sets of measure zero. 308 301 304 of a measure. 2 The extension 286 theorem. 281 outer measure. 264 *5 Inner measure. 301 2 Positive linear functionals 3 Bounded linear [unctionals *4 The Borel extension and Baire measures. 324 4 Set mappings and point mappings on complete metric spaces. 331 317 Epilogue Bibliography Index of Symbols Subject Index 335 337 339 341 .

Real Analysis .

It is assumed that the reader already has some acquaintance with the principal theorems on continuous functions of a real variable and with Riemann integration. Since the remainder of the book is full of applications of set theory. The second is devoted to general topology and to the theory of general Banach spaces. The theory of sets underlies all of the material in this book. and I have sketched in Chapter 1 some of the basic facts from set theory. The material in Chapter 2 is. These include the classical theory of functions of a real variable itself. Prerequisites.Prologue to the Student This book covers a portion of the material that every graduate student in mathematics must know. The first part contains the classical theory of functions. however. and the basic notions of linear vector spaces are used in Chapter 10. and Chapter 2 provides (formally) all of the basic theorems required. pointset topology. presented in a rather brief fashion and is intended for review and as introduction to the succeeding chapters. the student should become adept at set-theoretic arguments as he 1 . and the theory of normed linear spaces. The definitions and elementary properties of groups and rings are used in some of the peripheral sections. We also presuppose some acquaintance with the elements of modern algebra as taught in the usual undergraduate course. by which I mean those parts of modern mathematics which have their roots in the classical theory of functions of a real variable. For want of a better name the material here is denoted by real analysis. and the third to abstract treatments of measure and integration. This book is accordingly divided into three parts. The reader to whom this material is not already familiar may find it difficult to follow the presentation here. including the classical Banach spaces. measure and integration. No formal use of this knowledge is made here.

It is convenient to use some abbreviations for logical expressions. 'A only if B'. I recommend that he first read Chapter 1 lightly and then refer back to it as needed.B))... then B'. Since saying that there is an x such that A (x) means that it is not the case that for every x we have ---. p. (x) A(x) {:::}---. means 'not' or 'it is not the case that'. and '(3x)' to mean 'there is an x' or 'for some x'. 'V' means 'or' so that 'A V B' means 'A or B (or both)'. and 'A is necessary and sufficient for B'. -. Thus r+r -r-rr {(x) (3y) (x < y)} {:::}r-r-r (x) -r-rr (y) -. or 'B is necessary for A'.-(:Ix) --.. (3y)(x)(x < y) says that there is a y which is larger than every x.A' means 'it is not the case that A'.(A & (-. (x < y) {:::}(:Ix) (y) -. Logical notation. It has a number of synonyms in English.(x) A(x). We use' &' to mean 'and' so that 'A & B' means 'A and B'...(x). The books by Halmos [9]1 and Suppes [23] contain a more thorough treatment of set theory and can be profitably read by the student while he reads this book. I Numbers in brackets refer to the Bibliography. In addition to the preceding symbols we use two further abbreviations: '(x)' to mean 'for all x' or 'for every x'. 'A is equivalent to B'. Similarly..2 Prologue to the Student progresses through the book. Note that these two statements are different: As applied to real numbers. so that '-. 'A iff B'. We also use the notation 'A {:::} to mean '(A =} B) & (B =} A)'. we see that (3x) A (x) A {:::}__.. . the first is true and the second is false. 337. 'A is sufficient for B'. (x < y) {:::}(:Ix) (y) (y ~ x). The statement 'A =} B' is equivalent to each of the statements '(-. A(x). Similarly. Thus the statement (x)(3y)(x < y) says that for every x there is a y which is larger than x.A) V B' and '-. Another important notion is the one that we express by the symbol '=}'. Gleason [7] also contains useful material on set theory and has an excellent discussion of the role of set theory in characterizing abstract mathematical structure. English synonyms B' for 'A {:::}B' are 'A if and only if B'. so that the statement 'A =} B' can be expressed by saying 'if A. The reader will also find there a good discussion and development of the real number system and its properties. This rule is often convenient when we wish to express the negation of a complex statement. 'A implies B'.

All students are enjoined in the strongest possible terms to eschew proofs by contradiction! There are two reasons for this prohibition: First. then B' or in symbols 'A => B'. We will sometimes modify the standard logical notation slightly and write (€ > 0) (.. proposition. then so is the other.) greater than o such that (.. the contradiction on the final page arising from an erroneous deduction on an earlier page. etc. ) to mean 'for every E greater than 0 ( )'... A). and 'there is an x in the set A such that (. whereas both the direct proof and the proof by contraposition construct a chain of argument connecting A with B. 'there is a . ) would be written in standard notation (E) {(€ > 0) => ( . deduce various consequences from it.. Statements and their proofs. A lemma is usually used only for proving propositions and theorems in the same section.. ).. A theorem is a statement of such importance that it should be remembered since it will be used frequently. A similar convention is followed with respect to prob-r-r r.. The principal statements in this book are numbered consecutively in each chapter and are variously labeled lemma. . rather than from the incompatibility of A with -r-r B. and end with B. A proposition is a statement of some interest in its own right but which has less frequent application. The contrapositive of A => B is the statement (~B) => (---.£ x).. (30 > 0) ( ). It is sometimes easier to prove a theorem by contraposition. It is readily seen that a statement and its contrapositive are equivalent. that is. References to statements in the same chapter are made by giving the statement number. A third method of proof is proof by contradiction or reductio ad absurdum: We begin with A and -r-r B and derive a contradiction. and (:lx e A) ( .. )}. For example (E > 0) (. meaning Proposition 21 of Chapter 3.21.. Most of the principal statements (theorems. The direct method of proving a theorem of the form 'A => B' is to start with A. if one is true. This modification shortens our expressions.. Second. y. For a thorough discussion of the formal use of logical symbolism the student should refer to Suppes [22). References to statements in another chapter take the form Proposition 3. that is. such a proof gives little insight into the connection between A and B.. theorem.Prologue to the Student 3 where we have used properties of the real numbers to infer that (x < y) ¢=} (y .. or corollary.) in mathematics have the standard form 'if A. even when correct. propositions. by starting with ---. such proofs are very often fallacious.B and deriving r-rr A. as Theorem 17.

I have tried to restrict the essential use of inter chapter references to named theorems such as 'the Lebesgue Convergence Theorem'. one. which has the meaning of 'this completes the proof. proving B =} A. the 'only B'. etc. Interdependence of the chapters..4 Prologue to the Student lems.' If a theorem has the form 'A {::::} the proof is usually divided into two parts. proposition. if' part. The dependence of one chapter on the various preceding ones is indicated by the following chart (except for a few peripheral references). 9 Compact Spaces 6 Clossi co l Banach Spaces lab Integral Topological Spaces Vector . in this book begins with the word 'proof' and ends with the symbol 'I'. Chapters 14 and 15 depend on most of the preceding chapters. the references to numbered statements are mostly auxiliary references which the student should find it unnecessary to consult. The proof of a theorem. proving A => B. the 'if' part. the other. Chapter lOa indicates Sections 1 through 4 and 8 of Chapter ]0. Chapter 10b denotes Sections 5 through 7 of that chapter.

Our purpose is descriptive and the arguments given are directed toward plausibility and (hopefully) understanding rather than toward rigorous proof in some fixed basis for set theory. and others. but unfortunately this set theory is not quite as simple and natural as Frege and Russell supposed. or sets (col1ections) whose elements are subsets of X. for it was soon discovered that a free and uncritical use of set theory leads to contradictions and that set theory had to have a careful development with various devices used to exclude the contradictions. Frege. In this book we shall avoid these contradictions by having some set or space X fixed for a given discussion and considering only sets whose elements are elements of X. It is in fact possible to base most of mathematics on set theory. and so forth. Russell. The descriptions and notations given here are for . In the first few chapters X will usually be the set of real numbers. Roughly speaking. and it appeared that all mathematics could be based on set theory alone." such as trying to speak of a set which contains everything. the contradictions appear when one uses sets that are "too big.1 Set Theory 1 Introduction One of the most important tools in modern mathematics is the theory of sets. The study of sets and their use in the foundations of mathematics was begun just before the turn of the century by Cantor. or sets (families) whose elements are collections of subsets of X. In the present chapter we shall describe some of the notions from set theory which will be useful later.

and where the set X is understood we sometimes write! A = [x: P(x)}. that is. p. which can be defined (as in Halmos) in terms of the more primitive notions of set theory. It is perhaps unfortunate that the English phrase "contained in" is often used to represent both the notions E and C. 1 the most part consistent with the set theory described by Halmos in his book Naive Set Theory [9J. Since a set is determined by its elements. We also shall take for granted the principle of mathematical induction and the well-ordering principle. We express this latter notion by E. then A = B. although we assume as known. if two sets A and B have the property that x E A if and only if x E B. Otherwise we are confronted with the Russell paradox (cf. 6). several notions such as the natural numbers. Suppes [23]. General Topology [14]. I recommend Suppes' book Axiomatic Set Theory [23] or the appendix to Kelley's book. functions. Thus we always have A c A. the rational numbers. Suppose that each x in a set A is in the set B. that x is not an element of A.6 Set Theory [eh. The natural numbers (positive integers) play such an important role in this book that we introduce the special symbol N for the set of natural numbers. . as in the definition: The set A is the set of all elements x in X which have the property P. The principle of mathematical induction states that if Pen) is a proposition defined for each n in N. that is. The well-ordering principle asserts that each nonempty subset of N has a smallest element. We abbreviate this by writing A = [x E X: P(x)}. then {P(l) & [Pen) =? Pen + l)J} :=} (n)P(n). and if A c Band B c A then B = A. and so forth. one of the commonest ways of determining a set is by specifying its elements. and write 'x E A' for the statement 'x is an element (or member) of A'. 1 The presence (explicit or implicit) of the qualifying set X is essential. x E A =? X E B. We write 'x e A' to mean 'not (x E A)" that is. then we say that A is a subset of B or that A is contained in Band write A c B. For an axiomatic treatment. The basic notions of set theory are those of set and the idea of membership in a set. or primitive. A set is completely determined by its members. but we shall use it only in the latter context.

where we distinguish between the first element x and the second element y.Sec. . the set {x. w). we always have x 8 {x}. y} an unordered pair. For example. In {x. Halmos). and so 0 cA. and z. 3. The set {x} is called a unit set. then X X X is the set of ordered pairs of real numbers and. y. z. then x is a green-eyedlion. Since a set is determined by its elements. x}. second. One should distinguish carefully between x and {x}. y. product X X Yto be the set {(x. Similarly. z are elements of X. y. we shall not do so here. z} to be the set whose elements are x. It is often useful to consider also the ordered pair (x. Thus (x. X3 for X X X X X. while we seldom have x 8 x. quadruplets (x. that is. and so forth. x) if x ~ y. y. in terms of unordered pairs (cf. the set {x. third. where we distinguish between the first. If X is the set of real numbers. y) ~ (y. but it turns out to be convenient to consider also a set which has no members. is equivalent to the set of points in the plane. {x. y. 1. If x. y e Y. and z e Z. . y). or direct. Show that if x 80.. as we know from analytic geometry. 1] Introduction 7 We usually think of a set as having some members. y. We sometimes write X2 for X X X. y)} of all ordered pairs whose first element belongs to X and whose second element belongs to Y. y} = {y. Problems Show that {x: x ~ x} = 0. Show that in general the sets X X (Y X Z) and (X X Y) X Z are different but that there is a natural correspondence between each of them and X X Y X Z. . and we call it the empty set and denote it by 0. Thus the empty set is a subset of every set. y} to be the set whose elements are exactly x and y. we define the Cartesian. X X Y X Z is the set {(x. elements. b) if and only if x = a and y = b.. Similarly. Although it is possible to define ordered pairs. and so forth. triplets. 2. so that (x. or the singleton of x. we define the set {x} to be the set whose only element is x. y) = (a. If A is any set. y} there is no preference given to x over y. then each member of 0 (there are none) is a member of A. and so forth. we shall consider ordered triplets (x. For this reason we call {x. z). If X and Yare two sets. z)} of all ordered triples such that x e X. and so forth. there is only one such set.

] 5. 1 4. is called the range of f The range of a function f will in general be smaller than Y. 2 The word 'mapping' is often used as a synonym for 'function'. Show that the well-ordering principle implies the principle of mathematical induction. . we must have axioms describing the properties of functions. [Given a set S of positive integers. Thus the range offisf[X]. many people like to define a function to be its graph. The set of values taken by f. for example. let Pen) be the proposition "If n e S. such as (f = g) ¢=} (x)[J(x) = g(x)]. then S has a least element". A subset G of X X Y is the graph of a function on X if and only if for each x E X there is a unique pair in G whose first element is x. 2 onto Y if and only if Y = f[XJ. (Another terminology for this case: f is surjective. We denote this image by f[A]. in definmg the graph of the identity function i defined by i(x) = x for all x. The collection G of pairs of the form \x.8 Set Theory [Ch.) If A is a subset of X. [Consider the set {n e N: Pen) false}. We shall not use it here. Use mathematical induction to establish the well-orderingprinciple. the set {y e Y: (3x)[y = fex)]). then we say thatfis a function onto Y. we define the image under f of A as the set of elements in Y such that y = f(x) for some x in A. Thus f[A] = {y e Y: (3x) [x e A andfis and y = fex)]}. The equivalence between functions and their graphs is valrd only for functions from a given set X to a set Y. The set X is called the domain (or domain of definition) of f. There are difficulties.] 2 Functions By a function f from (or on) a set X to (or into) a set Y we mean a rule which assigns to each x in X a unique elementf(x) in Y. In a formal treatment which takes function as a primitive notron. as well as axioms enablmg us to construct functions.f(x) in X X Y is called the graph of the function f. Since a function is determined by its graph. It is irrelevant for our purposes whether we do this or consider the notion of function as primitive. We often express the fact that f is a function on X into Y by writing f:X~ Y. If the range of'j is Y. that is.

Fortunately for our notation. Similarly. we speak of a sequence from (or in) X or of a sequence of elements of X.) In this case there is a function g: Y --> X such that for all x and y we have g(f(x)) = x and f(g(y)) = y. we can define a new function g: A --> Y by defining g(x) = f(x) for x E A. an infinite sequence is a function whose domain is the set N of natural numbers. We use the term 'sequence' to mean a finite or infinite sequence. or n-tuple. f-I[B] It should be noted that = {x E X: f(x) E B}. 2] Functions 9 More important than the notion of image of a set under / is the notion of an inverse image. We shall often denote ordered n-tuples by <Xi!~=1 0 . and to call this value the ith element of the sequence. It should be noted that.) Similarly. f is onto Y if and only if the inverse image of each nonempty subset of Y is nonempty. n}. and the inverse images under g are different from those under f Having mentioned the possibility of defining functions by means of ordered pairs. It is customary to depart somewhat from the usual functional notation when dealing with sequences and denote the value of the function at i by x. If B is a subset of Y. This new function g is called the restriction of f to A and is sometimes written / I A. The function g is called the inverse of / and is sometimes denoted by f-l. 2}. Functions which are one-to-one from X onto Yare often called one-to-one correspondences between X and Y. that is. the set {i r N: i ::. conversely. is a function whose domain is the first n natural numbers. or injective) if/(xI) = f(X2) only when Xl = X2. we define a new function h: X --> Z by setting hex) = g(f(x)). ordered pairs may be defined in terms of the notion of a function: An ordered pair is a function whose domain is the set {I.Sec. it is only fair to point out that. these are the same set. The function h is called the composition of g with / and denoted by g f Iff: X --> Y and A is a subset of X. A functionj: X --> Yis called one-to-one (or univalent. if we denote g by /-I. a finite sequence. Ifj: X --> Y and g: Y --> Z. (They are also called bijective. (We call such a set a segment of N. then/-I[E] can be considered to be the inverse image of E under f or the image of E under /-1. If the range of a sequence is in a set X. we define the inverse imagef-I[B] of B to be the set of those x in X for whichf(x) is in B. In many cases it is important to distinguish carefully between the functions g and f They have different ranges. that is.

) will be denoted by {x}. Thus the range of an ordered n-tuple (Xi?:=1 is the unordered n-tuple {Xi}:= l' This notation is reasonably consistent with our earlier notation concerning ordered and unordered pairs. A set A is called countable if it is the range of some sequence and finite if it is the range of some finite sequence.. = a an d X.n) = Xt. We say that a mapping g of N into N is monotone if (i > j) =:} (g(i) > g(j». we have x. The existence of such a sequence is intuitively clear: We define = a. triplets. .x.m. One of the most useful ways of defining an infinite sequence is given by the following principle: Principle of Recursive Definition: Let f be a function from a set X to itself.10 Set Theory [Ch. and let a be an element of X.. such that XI (n) (n) x.. and so forth.). An important notion connected with that of a sequence is the notion of a subsequence. A more formal proof can be given as follows: We first prove by induction on n that for each natural number n there is a unique finite sequence Xl Xl (n) . . and so on.n) for 1 ::. a function whose domain is N). . Then there is a unique infinite sequence (x. We shall return to this notion again in Section 6. but our definition includes the finite sets among the countable sets. i < n. (n) + 1 = f( X.) satisfies the requirements of our principle..m) for i ::. n ::. From the uniqueness. and we see that the sequence <x.X2 (n) .. it follows that x~n) = x. Thus if we define Xn to be x~n). X2 = f(a). A set which is not finite is called infinite. . 1 and infinite sequences by (Xi?~l' When no misunderstanding is likely to arise we often write simply (Xi? The range of the sequence (x.) We use the term 'countably infinite' for infinite countable sets. we say that h is an infinite sub- . X3 = f(f(a». from X such that Xl = a and Xi+l = f(xi)for each i. A slight extension of this principle is the following: For each natural number n letfn be a function on xn to X and let a 8 X. Iff is an infinite sequence (that is. (Many authors restrict the use of the word 'countable' to sets which are infinite and countable. Then there is a unique sequence (Xi? from X such that Xl = a and xt+1 = hex!> .

that is. and Complements 11 sequence of f if there is a monotone mapping g of N into N such that h = fog. Show that f is onto if there is a mapping g: Y ---7 X such that / g is the identity map on Y.) 0 8. 3] Unions. Problems 6. such that g(f(x» = x for all x 8 X. Let f: X ---7 Y be a mapping of X into Y. We have (A n B) n C = A n (B n C) and write this set as A n B n C. and Complements Let us fix a given set X and consider the set cp(X) consisting of all subsets of X. Use mathematical induction to prove the generalized principle of recursive definition stated in the text. we define their intersection A n B to be the set of all elements which belong to both A and B. It is the set of all elements which belong to each of the sets A.Sec. 3 Unions. Also A nBc A and A n B = A RAe B. Thus A n B = {x: x e A & x e B}. A n B = B n A. B. then we usually denote fog by (fg). Intersections. Let f: X ~ Y be a mapping of a nonempty space X into Y. If A and B are subsets of X. Show that/is one-to-one if and only if there is a mapping g: Y ~ X such that go/ is the identity map on X. We define the union A u B of two sets A and B to be the set of elements that are in either A or B. We note that the definition is symmetric in A and B. 7. Thus AuB= {x:xeAVxeB}. AuB=BuA A u (B u C) = (A u B) u C = A u B u C AcAuB A = A UBRBCA. and C. If we write f as (J. such that f(g(y») = y for all y 8 Y.) and g as (gi). There are certain set-theoretic operations that we can perform on subsets of X. We have . that is. (For the converse see Problem 20. that is. Intersections.

We sometimes write ".(A u B) = ". to unions and intersections Two special laws relating complements are De Morgan's laws: ".. AuX= If A is a subset of X. X.. or relative complement of A in B. as the set of elements in B which are not in A. We have B . Thus A of A (relative A = {x eX: x e A}. Thus B". The empty set 0 and the space X play special roles: Au which are 0 = A... If A and B are two subsets of X we define the difference B '" A.. An0 = 0 A n X = A. B for the symmetric difference of two sets defined by A £::. we define the complement to X) as the set of elements not in A. B = (A '" B) u (B '" A). = B n A. We have X=0 A nA = A= A. The symmetric difference of two sets consists of all those points which belong to one or the other of the two sets but not to both.. . A We shall also use the notation A £::.(A () B) = A nB A u B. 1 We also have relations between unions and intersections called distributive laws: A n (B u C) = (A n B) u (A () C) A u (B n C) = (A u B) n (A u C). A = {x: x e B & x e A}. A.A instead of 0=X. A SZ5 cB ¢=:} B c A.12 Set Theory [Ch. A u A = X..

Similarly. It follows from our definition that the union of an empty collection of sets is empty and that the intersection of the empty collection of sets is X. A collection e of sets is said to be a disjoint collection of sets or a Aee nA = Aee n n {x e X: (A)(A e e =} x e A)}. for the union of the •=1 U A. we write range of the sequence.)} . a mapping of N(or a segment of N) into cp(X). we can give a definition of intersection for an arbitrary collection e of sets: The intersection of the collection e is the set of those elements of X which belong to each member of e.. We denote this intersection by A or {A: A e e}.. Thus If the intersection of two sets is empty we say the sets are disjoint.. By a sequence of subsets of X we mean a sequence from cp(X). Intersections. we define the union of an arbitrary collection of sets by zl s e UA = {x e X: (~A)(A e e&x e A)}. Thus U . The process of taking unions (or intersections) of two sets can be extended by repetition to give unions (or intersections) of any finite collection of sets. and Complements 13 collection of pairwise disjoint sets if any two sets in e are disjoint.) is an infinite sequence of subsets of X.. = 00 {x: (3i)(x eA. However.[U A] Aee Ace = Aee nA U A.=1 00 A. De Morgan's laws hold for arbitrary unions and intersections: . . 3] Unions. If (A. '"[n AJ = We also have the distributive laws: B n[ B U[ Aee Aee U A] = U Aee Aee (B n A) Aee n AJ = n (B u A).Sec. that is.

)} and "eA n A" = {x E X: (X)(x E A =} XE A. for a collection {BJ of subsets of Y. 1 Similarly.J is a collection of subsets of X.. I-I [y B"J = [Q BA] = 1-1 . we write n B.} is indexed by A.: X r A}. We say that {x".. we usually write Xi. In keeping with the notation for sequences.14 Set Theory [Ch.)}. then but we can conclude only that For inverse images we have.." e A & x e A.. for the intersection n B.. We define the union and intersection of an indexed set to be the union and intersection of the range of the function defining the indexed set. If I maps X into Yand {Ai. then the notion of an indexed set coincides with the notion of a sequence. n n i=1 of the range of the sequence. so that = s.. If A is the set of natural numbers. n u. This notation for sequences of sets is so convenient that we often generalize it to arbitrary collections of sets by using the notion of an indexed collection: An indexed subset of X (or collection of subsets of X) is a function on an index set A to X (or the set of subsets of X). In the case when A is the set N of natural numbers we have teN n At = t=1 n At) and similarly for unions. if (B~)~=1 is a finite sequence of subsets of X. instead of xC. n B2 n .). Thus U A" "cA = {x e X: (3X)(. and denote the indexed set itself by {XJ or {Xl.

". BeY. a. B) n E 14. Show f[UAx] b. c nf[AxJ. C) (A 6. B c A. (B n E). = = Uf-I[Bx]. = a. n [U A] = Aee Aee U (B n A). Show that if C't and CBare two collections of sets. B) 6. Problems 9. Show that: a. A /: B c. Show f[nAx] = = U{AnB: Uf[Ax]. and Complements 15 and for BeY. B d. Show that: ¢::} ¢::} A UB = B. Intersections. a. A and A 6. (B 6. 15. f[f-I[B]J CB then . then [U{A:ABC't}]n[U{B:BBCB}] 16. = X ¢::} A = B. c. 3] Unions. fI[E] = 18. Prove the distributive laws. c. X = = e. Show that A C B 12. Give an example where 17. f-I[UBx] b. Show that B (A n E) 6. and for A Also c X and BeY.B)BaxCB}. f-I[nBx] nf-I[Bx]. B b. (A 6. Show that A C B ¢::} A n B = A 10. Show that iff maps X into Y and A C X. (A. = 0 ¢::} A = B. Prove De Morgan's laws (for arbitrary unions and intersections). A 6. = B b. A.Sec. A f::. A 6.. 13. 11.f-I[B] for BeY. 0= A and A b. C.

. Similarly. Give examples to show that we need not have equality. it follows that if ill is an algebra containing e. we have A u B in {CB: ill e s}. Proof: Let . m and 2. then CB ::::> a.. U An is again in a.y contains e. From the definition of a. n An is in a. then for each CB e . we see that if At...An are sets in a then At U A2 U .y we have A B CB and B e CB.Y.. a is an algebra. By taking unions two at a time. then 4 Algebras of Sets A collection a of subsets of X is called an algebra of sets or a Boolean algebra if (i) A u B is in a whenever A and B are. there is a smallest algebra a which contains e. If a collection a of subsets of X satisfies (ii) and (iii). Show that iff maps X onto Y and BeY. Proposition: Let a be t=l U 00 B. Al n A2 n . . 1 and b. . Then e is a subcollection of a. then A e a. there is an algebra a containing e and such that if CB is any algebra containing e. since each CB In .. = 7=1 U 00 A. The first IS the following: 1. For if A and B are in a. We shall find several propositions concerning algebras of sets useful. we see that if A e a. Since ill is an algebra.' that is.y be the family of all algebras (of subsets of X) which contain e. I n n an algebra of subsets and <A~)a sequence of sets in a. It follows from De Morgan's laws that (iii) A n B is in a whenever A and B are. Proposition: Given any collection e of subsets of X. Since this is true for every CB e . Similarly. Moreover. Then there is a sequence <B~) of sets in a such that B.. then by De Morgan's laws it also satisfies 0) and is therefore a Boolean algebra. A u B belongs to CB. Let a = {CB: CB £ s}.. n Bm = 0 for n rf:. c. and (ii) A is in a whenever A is. then CB contains a.16 Set Theory [Ch.

. n Am n . Let n 8 1 be the smallest value of i such that x x 8 A.) to be an infinite sequence. and suppose m < n.Sec. and so n=l U e. 4) Algebras of Sets 17 Proof: Since the proposition is trivial when <A. = An = f" [AI U A2 U U An-d An n . Let B. From De Morgan's laws it follows that the intersection of a countable collection of sets in ct is again in ct. C An. Then Bm C Am. we assume <A. 8 U At. That is.. C At> we have U Let x 00 co B. and Bm be two such sets. I An algebra ct of sets is called a e-algebra.) is a sequence of sets.. cU co A •. then U A.=1 must again be in ct. C Am n s. Set B1 = AI.L n A2 n nAn-I. Then x 8 Bn. or a Borel field. if <A. and we have n=l U 00 U n=I OC! An. :l U B. = OC! co n=l 00 An.. Since B. Since the complements and intersections of sets in ct are in e.. Thus n=l U s. 8 ct. We also have B.) is finite. t= 00 Then x must belong to at least one of the A:s. we have each B. and so s. and for each natural number n > 1 define s. if every union of a countable collection of sets in ct is again in ct. A slight modification of the proof of Proposition 1 gives us the following proposition: . = (Am n Am) n ." = 0' n'" = 0'. n s. = Am n An n ..

Prove Proposition 3. ). but we shall be rather informal about its use.18 Set Theory [Ch.. If A = {I. we may think of the axiom of choice as asserting the possibility of selecting a "parliament" consisting of one member from each of the sets in e. that is... of course. but we need the axiom of choice in case the collection e is infinite. . The axiom is the following: e be any collection of nonempty sets. there is a smallest a-algebra which contains e.. If the sets in e are disjoint. the Xth coordinate of z." we call x). The function F is called a choice function. Let e = {X)J be a collection of sets indexed by an index set A. Problem 19.. Then there is a function F defined on e which assigns to each set A e e an Axiom of Choice: Let element F(A) in A. e X). Many mathematicians like to be very explicit about the use of the axiom of choice and its consequences. there is a a-algebra a containing e such that if <E is any a-algebra containing e then a C <E.} indexed by A and having the property that x).. 2}. 1 3. It is somewhat less elementary than the other axioms used in axiomatic set theory and is known to be independent of them. We define the direct product to be the collection of all sets {x). and its existence may be thought of as the result of choosing for each of the sets A in e an element in A. Proposition: Given any collection e of subsets of X. 5 The Axiom of Choice and Infinite Direct Products An important axiom in set theory is the so-called axiom of choice. no difficulty in doing this if there are only a finite number of sets in e. There is. we have our earlier definition of the direct product Xl X X2 of the two sets Xl and X2• If z = {xJ is an element of X X).

For this reason Bertrand Russell axiom. the finite set {Xl. 6] Countable Sets 19 If one of the Xx is empty. In fact. we have a finite set. The set N of natural numbers is an example of a count ably infinite set. It is convenient. A set is called countable (or denumerable) if it is either empty or the range of a sequence. prefers to call the axiom of choice the multiplicative Problem g: Y --+ X 20. Let j: X --+ Y be a mapping onto Y. for i > n..xn} is the range of the infinite sequence defined by setting X~ = x.Sec. we had better come to terms with the empty set. For example. It is usual in mathematics to give a slightly different but equivalent definition based on the notion of a one-to-one correspondence. [Apply the axiom of choice to the collection {A: (3y e Y) with A = j-l[{y} ]}. then X X>. It follows at once from this definition that the image of any countable set is countable. . is also x empty. that is. but the range of an infinite sequence may also be finite. If it is the range of a finite sequence.. and similarly for finite sets. Thus a set is countably infinite if it is the range of some infinite sequence but not the range of any finite sequence.] 6 Countable Sets In Section 2 we defined a set to be countable if it was the range of some sequence. Then there is a mapping such tr1tj 0 g is the identity map on Y. however. Before proceeding further.. . The axiom of If none of the Xx choice is equivalent to the converse statement: are empty. that the range of any function with a countable domain is itself countable. The empty set is not the range of any sequence (unless we admit sequences with zero terms). We . then X Xx x is not empty. every nonempty finite set is the range of an infinite sequence. Hence: Definition: A set is called finite if it is either empty or the range of a finite sequence. to define finite and countable sets so that the empty set is both finite and countable.

Thus to show that our definition is equivalent to the customary one. To do this we define a function cp from N into N by recursion as follows: Let cp(l) = 1. if x. ~ x.. any set which can be put in one-to-one correspondence with N is countably infinite. Let f be the function on N which assigns to the natural number n the finite sequence (Xl. If A is empty. ) be the sequence of prime numbers. Since the set N of natural numbers is countable but not finite. Define a new sequence (Yn) by setting Yn = x. If A is not empty. 3. Proposition: Every subset of a countable set is countable.Pk. Thus we have shown that a set is countably infinite if and only if it can be put into one-to-one correspondence with N. where x.• . . 1 first note that any set that can be put in one-to-one correspondence with a finite set is finite and that any set that can be put in one-to-one correspondence with a countable set must be countable.. choose x in A. I 3 Except 1. Then A is the range of (Yn) and is therefore countable.20 Set Theory [Ch.. for all i ::. we agree to wnte 1 = 2°. The correspondence n ~ x'I'(n) is a one-to-one correspondence between Nand E.. cp(n). such an m always exists. and let A be a subset of E. it can be put into one-to-one correspondence with a subset of the set N of natural numbers.} be a countable set. we must show that. E No = N u {a} and Xk > O. if an infinite set E is the range of a sequence (xn). and by the well-ordering principle for N there is always a least such m. Xk) from No. Since E is infinite. . We are now in a position to prove some simple propositions about countable sets: 4. Proof: Let E = {x. and define cp(n + 1) to be the smallest value of m such that x. E A and Yn = X if Xn t A. It is customary to use this property to define the notion of countably infinite. Then the set of all finite sequences from A is also countable. . 11. Then each n in N has a unique factorization of the form" n = 2X1 3X2 ••• PI/-. Thus it suffices to prove that the set S of all finite sequences of natural numbers is countable. I 5. •. . 5. Let (2. then E can be put in one-to-one correspondence with N.. Proposition: Let A be a countable set. Proof: Since A is countable. Then S is a subset of the range off Hence S is countable by Proposition 4. A is countable by definition. 7.

22. Then there is a set E C X which is not in the range of f. 3) --7 ° -p/q is a function whose range is the set of rational numbers and whose domain is a subset of the set of finite sequences from N. Thus e is the range of an infinite sequence (An):=l' and each An is the range of an infinite sequence (Xnm):=l' But the mapping of (n. [Hint: The mappmg (p. = 1 . Proposition: The union of a countable collection of countable sets is countable. I Proof: Let e be a countable Problems 21. and for each v 8 N we have (bn> ~ (avn). Show that the set E of infinite sequences from {a.] 23. [Hint: Let f be a function from N to E. . Then (bn) is again a sequence from {a. Letfbe a function from a set X to the collection (fl(X) of subsets of X. m) to Xnm is a mapping of the set of ordered pairs of natural numbers onto the union of e. If all the sets in e are empty. the union of the collection e must also be countable.] 24. I}.) 25. 1) --7 p/q (p. Then f(v) is a sequence (am):=l' Let b. collection of countable sets. and since the empty set contributes nothing to the union of e. I} is not countable. we can assume that the sets in e are nonempty. Thus we may as well assume that e contains nonempty sets. the union is empty and thus countable.Sec. Show that every subset of a finite set is finite. q. Use the axiom of choice and the generalized principle of recursive definition to show that each infinite set X contains a countably infinite subset. 6] Countable Sets 21 6. Since the set of pairs of natural numbers is countable. (Let E= {x:xtf(x)}. q.apv. 7. This method of proof is known as the Cantor diagonal process. Prove Proposition 6 by using Propositions 4 and 5. Proposition: The set of all rational numbers is countable. 2) --7 (1. 1.

Thus any element of X which is equivalent to an element of Ex is itself an element of Ex.y. whence z == x." A relation R is said to be transitive on a set X if x R y and y R z imply x R z for all x.x. and c are no longer relations. we define the graph of R to be the set {(x. either x stands in the relation R to y (written x R y) or x does not stand in the relation R to y. x e y. For this reason I prefer some treatment such as the one III Kelley [14]. 1 7 Relations and Equivalences Two given entities x and y may be "related" to each other in many ways. In general we say that R denotes a relation if. y. and so no equivalence class is empty. let Ex be the set of elements equivalent to x. each relation on a set X is uniquely determined by its graph. Thus = and < are transitive relations on the set of real numbers. and conversely each subset of X X X is the graph of some relation on X. where relations are not necessarily sets of ordered pairs. or for numbers x < y. A relation R is said to be a relation on a set X if x R y implies x e X and y e X. for any two elements x and y of X. Note that x e Ex. or Halmos [9]. are either identical (if x . e. Thus any two elements of Ex are equivalent.22 Set Theory [Ch. then y == x and z x. and by symmetry and transitivity we have z . If R is a relation on a set X. the sets Ex and E. Consequently.y) or disjoint (if x ¢ y). Suppose that == is an equivalence relation on a set X. p. and z in X. reflexive. It is said to be reflexive on X if for all x e X we have x R x. If y and z are both in Ex. x C y. 260. that there IS a difficulty III this approach III that =. given x and y. as in x = y. The sets in the collection {Ex: x eX} are called equivalence sets or classes of X under -. In many formalized treatments of set theory a relation is in general defined simply as a set of ordered pairs. Thus X is the disjoint union of the equivalence classes under . The collection of equivalence classes under an equivalence == is called the quotient of X with respect to ==. then z y and y .x}. p. p 26. that is Ex = {y: y . It should be pointed out. Thus we may if we like identify a relation on X with its graph and define a relation to be a subset of X X X. A relation which is transitive. however. Since we consider two relations Rand S to be the same if (x R y) ¢:::} (x S y). and so z e Ex. For a given x e X. and is sometimes denoted 4 Cf Suppes [23]. y): x R y}. and symmetric on X is said to be an equivalence relation on X or simply an equivalence on X. . A relation R is said to be symmetric on X if x R y implies y R x for all x and y in X. 57. If y e Ex and z == y.

8 Partial Orderings Maximal Principle and the A relation R is said to be antisymmetric on a set X if x R y and y R x imply x = y for all x and y in X. A relation -< is said to be a partial ordering of a set X (or to partially order X) if it is transitive and anti symmetric on X. choose x 8 E. Thus ~ linearly orders the set of real numbers. The induced operation then makes the quotient space into a group. 8] Partial Orderings and the Maximal Principle 23 by X/=. Thus ~ is a partial ordering on the real numbers and c is a partial ordering on cp(X). 27. For further details the reader may consult Birkhoff and MacLane [2]. Since == is an equivalence. y 8 F and define E + F to be E(x+y). Let X be an Abelian group under +. A partial ordering -< on a set X is said to be a linear ordering (or simple ordering) of X if for any two elements x and y of X we have either x -< y or y -< x. we often say that a precedes b or that b follows a. I 45ff. We say that an equivalence relation '= is compatible with a binary operation if x '= x' and y '= y' imply that (x y) '= (x' y'). It should . and x ~ a. In this case + defines an operation on the quotient Q = XI == as follows: If E and F belong to Q. we say that an element a e E is the first element in E or the smallest element in E if. Sometimes we say that a is less than b or b greater than a. Then == is compatible with + if and only if x '= x' implies x + y '= x' + y. while c is not a linear ordering on cp( X). The mapping x ~ Ex is called the natural mapping of X onto X/~. and similarly for maximal elements. we have a -< x. Similarly for last (or largest) elements. An element a 8 E is called a minimal element of E if there is no x e E with x =P a and x -< a. Prove that F + G as defined above depends only on F and G. + + + + Problems 26. If -< is a partial order on X and if a -< b. E F is seen to depend only on E and F and not on the choice of x and y.Sec. pp. A binary operation on a set X is a mapping from X X X to X. If E C X. whenever x e E.

then that element is a minimal element. pp. If we have x -< x for all x. Give an example of a partially ordered set which has a unique minimal element but no smallest element. a subset S of X which is linearly ordered by -< and has the property that. 1 be observed that if a set has a smallest element. Problems 28. 31-36. or Kelley [14]. if we take X = Nand < to mean less than. If < is any partial order we use ::. .31-36). we call -< a reflexive partial order." The following principle clearly implies the axiom of choice and can be shown equivalent to it (cf. that is. if SeT c X and T is linearly ordered by -<. y 29. but in general it is possible to have minimal elements which are not least elements.. then -< is called a strict partial order. I .. Then there is a unique strict partial order < and a unique reflexivepartial order ::. Chapter 8. the set R of all real numbers is not well ordered by the relation "less than. On the other hand. y) with x ~ y.. For a proof of this equivalence and a discussion of related principles see Suppes [23].on X such that for x ~ y we have x -< y ¢=} x < y ¢=} x ::. If we never have x -< x. Thus. pp. 9 Well Ordering and the Countable Ordinals A strict linear ordering < on a set X is called a well ordering for X or is said to well order X if every nonempty subset of X contains a first element.for the associated reflexive partial order. a minimal element is a least element.24 Set Theory [Ch. The following principle is equivalent to the axiom of choice and is often more convenient to apply. . Thus < is a strict partial order for the real numbers and < is a reflexive partial order. or Kelley [14]. then S = T. Then there is a maximal linearly ordered subset S of X. then N is well ordered by <. If < is a linear ordering. Our definition of partial order makes no assertion about the possibility or necessity of x -< x. Let -< be a partial order on X. Hausdorff Maximal Principle: Let < be a partial ordering on a set X. Suppes [23]. To any partial order -< there is associated a unique strict partial order and a unique reflexive partial order that agree with -< for all ~x. Chapter 8.

as an ordered set. Proposition: There is an uncountable set X which is well ordered by a relation < in such a way that: i. 31. 8. If < is a partial order on X with the property that every nonempty subset of X has a least element. set. If Y does not have a last element. then the set {y e X: y < x} is countable.b* for each upper bound b*. replace Y by Y u {a}. then < is a linear ordering and consequently a well ordering. and extend the order < by setting y < a for all y e Y. take an element a e Y. b. It can be shown to be unique in the sense that. The set of y in Y for which the set {x e Y: x < y} is uncountable is a nonempty set. Proof: Let Y be any uncountable Problems 30. say the one given by Problem 23. i. (An element b is an upper bound for E if x ::. Let Y be the set of ordinals less than the first uncountable ordinal. then there is a one-to-one order preserving correspondence between X and Y. then {x: x < w} is the set of finite ordinals and is equivalent. if Y is any other well-ordered set with the same properties.. Then X is the required set. it is a least upper bound if b b ::.) . By the well-ordering principle. ii. since it contains the last element of Y. Show that any subset of a well-ordered set is well ordered. There is a last element Q in X. Show that every countable subset E of Y has an upper bound in Yand hence a least upper bound.. The elements x < Q are called countable ordinals. 9] Well Ordering and the Countable Ordinals 2S Well Ordering Principle: Every set X can be well ordered. If x e X and x ~ Q. The last element Q in X is called the first uncountable ordinal and X is called the set of ordinals less than or equal to the first uncountable ordinal. for each x e E. and let X = {x e Y: x < Q or x = Q}. If w is the first nonfinite ordinal. there is a relation < which well orders X. a. I The well-ordered set X given in the proposition will be very useful for constructing examples.' that is. Let Q be the smallest element in this set.e. there is a well ordering < for Y.Sec.. If {y: y < x} is finite we call x a finite ordinal. Y = {x e X: x < Q}. This new Y has a last element and is well ordered by <. to the set N of positive integers.

33. Show that there is a successor-preserving map / of the union 8 of this collection into Yand that either S = X or /[8] = Y. A subset 8 of a well-ordered set X is caIIed a segment if 8 = {x e X: x < y} for some y e X or if S = X. .26 Set Theory [Ch. Show that the well-ordered set X in Proposition 8 is unique up to isomorphism. If X and Yare well-ordered sets. A function / from X to Y is called successor preserving if for each x e X the element/ex) is the first element of Y not inf[ {z: Z < x}]. Show that there is at most one successor-preserving map of X into Y. then the restriction e. 1 32. Show that the range of a successor-preserving map is a segment. d. either there is a successor-preserving map of X onto a segment of Y or a successor-preserving map of Y onto a segment of X. Show that a union of segments is again a segment. c. that is. Show that if / is successor preserving. [Hint: Consider the collection of all those segments of X on which there is a successor-preserving map into Y. Let X and Y be two well-ordered sets. If/is a successor-preserving map of Xinto of/ to a segment is also successor preserving. then f is a one-to-one order-preserving map and/-l is also order preserving. Y. then there is a successorpreserving map from one of them onto a segment of the other. a.] f. b.

Part One 0 Theory of Functions of a Real Variable .

Such a program gives an elegant construction of the real numbers out of more primitive concepts and set theory. (x + y) + z = x + (y + z). The first group describes the algebraic properties and the second the order properties. A. and in fact these axioms completely characterize the real numbers. We shall not concern ourselves here with the construction of the real numbers but will think of them as already given. the set P of positive real numbers. All the properties we need are consequences of these axioms. which we list in three groups. y. the rational numbers in turn being defined in terms of the natural numbers. The present chapter is devoted to a review and systematization of those results which will be useful later. x A2. + y = y + x. and z we have: AI. and the functions '+' and '. 29 .' on R X R to R and assume that these satisfy the following axioms. We thus assume as given the set R of real numbers. The third comprises the least upper bound axiom. and list a set of axioms for them. The Field Axioms: For all real numbers x.2 The Real Number System 1 Axioms for the Real Numbers We assume that the reader has a familiarity with the set R of real numbers and those basic properties of real numbers which are usually treated in an undergraduate course in analysis. One approach to the subject of real numbers is to define them as Dedekind cuts of rational numbers.

£. xy = yx.x e P. (xy)z = x (yz). We write 'x ::. that is. (x £. E R. (x = 0) or (x P) or (-x E P). A 7. A6. It follows from Al that the 0 in A3 is unique. and A8. xy P. R. any system satisfying Al through A9. .y' for 'x < y or x = y'. I = x for all x £. A8. The rational numbers give another example of an ordered field. We shall use the various consequences of these axioms without explicit mention. R there is a w R such that x +w = O. ::11E R such that I ~ 0 and x . Thus the real numbers are an ordered field. A9. we can perform all the operations of elementary algebra.). The win A4 is unique and denoted by '-x'. When we do this our second group of axioms takes the following form: + B. For each x AS. 2 R such that x E +0 = x for all x £. We define subtraction x . a fact which we have assumed in the formulation of A4. The second class of properties possessed by the real numbers have to do with the fact the real numbers are ordered. x(y + z) = xy + xz.30 A3. A4. but it is somewhat more convenient to use the notion of a positive real number as the primitive one. For each x in R different from 0 there IS w E R such that xw = 1. In terms of < Axiom BI is equivalent to (x < y &z < w) ==? X + z < y + w. ==? X ==? ==? ==? + YEP. E P) R) E -x t P. In an ordered field we define the notion x < y to mean y . (x. Any set which satisfies these axioms is cal1ed a field (under and . YEP) B3. including the solution of simultaneous linear equations. (x B4. (x. A 7. The w in A8 can be shown to be unique and is denoted by 'x-Po If we have a field. The I in A 7 is unique. Any system satisfying the axioms of groups A and B is called an ordered field. We could axiomatize the notion of a less than b. YEP) B2. Axioms of Order: The subset P oj positioe real numbers satisfies the following: Bl.y as x + (-y). 30 E The Real Number System [Ch.

b for each upper bound b of S. while B4 states that of any two different numbers one must be the larger. Then either L has a greatest element or U has a least element. In contrast to our cavalier policy about the consequences of the first two axiom groups we shall be explicit about the use of this last axiom. C. Since Axiom BI implies that the relation < is transitive. 1] AXioms for the Real Numbers 31 and B2 is equivalent to (0 <x <y &0 <z < w) =} xz < yw. b. Except for the discussion in the beginning of the next Section we take all the consequences of these two axiom groups for granted and use them without explicit mention. We shall often denote the least upper bound of S by sup S or by sup x and occasionally by sup {x: x s S}. and it is this axiom which distinguishes the real numbers from other ordered fields. we see that the real numbers are linearly ordered by <. For further properties of ordered fields the reader should see Birkhoff and MacLane [2]. We sometimes express this by writing S ::. Proposition: Let Land U be nonempty subsets of R with R = L u U and such that for each I in L and each u in U we have I < u. let us introduce some terminology: If S is a set of real numbers. Note that inf x = -sup -x. and it follows from Axiom C that every set of real numbers with a lower bound has a greatest lower bound. We denote the greatest lower bound of a set S by inf S or by inf x. The third axiom group consists of a single axiom. Completeness Axiom: Every nonempty set S of real numbers which has an upper bound has a least upper bound. We can define lower xeS bounds and greatest lower bounds in a similar fashion. As a consequence of Axiom C we have the following proposition: 1.Sec. we say that b is an upper bound for S if for each xeS we have x ::. Axiom B3 asserts that a number cannot be both greater than and less than another. Our final axiom for real numbers simply guarantees its existence for sets with an upper bound. A number c is called a least upper bound for S ifit is an upper bound for S and if c ::. b. Clearly the least upper bound of a set S is unique if it exists. xeS xeS xeS . Before stating this final axiom.

We often denote max (x. 2 The Natural and Rational Numbers as Subsets of R We have adopted the procedure of taking the natural numbers for granted and of using them as counting numbers. c. IxYI = IxIIYI· b. 4. Show that a.) We shall show that the ++ + + . By the principle of recursive definition there is a function i{J from the natural numbers to the real numbers defined by i{J(1) = 1 and i{J(n 1) = i{J(n) 1. (-x) /\ (-y) -(x V y). then Ixl ~ y. z(x V y) = (zx) V (zy) if z o. c. The Real Number System [Ch. Ixl = x V (-x). We define Ixl to be x if x 2:: 0 and to be . b. y) to be x if x 2:: y and y if y 2:: x. Yet we all regard such a number as 3 not only as a natural number but also as a real number. < O. d.x if x a. if -y ~ x ~ y. y) to be the smaller of x and y. = (x + z) V (y + z).yl). If x and yare two real numbers. x /\ y d. In fact. and one is tempted to say that we define the real number 3 as 1 1 1. Prove Proposition 1 using Axiom C. 3. x V y lex + y + e. ?: e. (Here 1 denotes a real number on the right side and a natural number on the left. x V y +x +z Vy = = x + y. (x /\ y) /\ z = x /\ (y /\ z). and that in a "similar" fashion we can define real numbers corresponding to any natural number.32 Problems 1. y) by 'x V y'. Similarly. we define min ex. Show that [x + yl ~ Ixl = + Iyl· [x . Use Axiom C to show that every nonempty set of real numbers with a lower bound has a greatest lower bound. we define max (x. 5. and denote it by 'x /\ y'. 2 2. Show that 1 e P. Actually we may use the tools we have at hand to do this in a more precise fashion. we have used the symbol I not only to denote the first natural number but also the special real number given by Axiom A7.

then there is a rational r with x < r < y. Proof: Let S be the set of integers k such that k . and taking quotients of integers gives us the rationals. I 4. 2] The Natural and Rational Numbers as Subsets of R 33 mapping <pis a one-to-one mapping of N into R. It was first used by Eudoxus. y . and the rational Ifwe make use of Axiom C we can prove several further facts about the integers and rational numbers as subsets of the reals. and so Ck + 1) Z S. and we see that the mapping tp is one-to-one. Let p and q be two different natural numbers. and we shall show that <p(p) < <p(q) by induction on n. products.t. Proposition: the natural numbers. By taking differences of natural numbers. Thus we have shown the following: + + + 2. . x. and the relation <.s. Since k + 1 is an integer not in S. it has a least upper bound y by Axiom C. we obtain the integers as a subset ofR. But k + 1 > y + t > y.t cannot be an upper bound for S. One of the most important is the following theorem. Thus <pgives a one-to-one correspondence between the natural numbers and a subset of R.Sec. < y. and so <pCP + n) > <pCp) implies <pCP + n + 1) > <pCp)· Thus by induction <pCP + n) > <p(p). to) the integers. which for historical reasons 1 is called the Axiom of Archimedes: 3. Since S has the upper bound x. For n = 1 we have q =p 1 and <p(q) = <p(p) 1 > <p(p). the same results hold for any ordered field. Since y is the least upper bound for S. we shan consider the set N of natural numbers to be a subset ofR. we should distinguish between the natural number n and its image <p(n) under sp. ifx 1 that is. Since Axiom C was not used in this discussion. For general n we have <pCP+ n + 1) = <p(p + n) + 1 > <p(p + n). We can also prove by induction that <p(p + q) = <p(p) + <p(q) and <p(pq) = <p(p)<p(q). say p < q. Strictly speaking. there is an < n. we must have k + 1 greater than x by the definition of S. Then q = p n. Every ordered field contains (sets isomorphic numbers. and so there is a k e S such that k > y . and <p preserves sums. Axiom of Archimedes: integer n such that x Given any real number x. Corollary: Between any two real numbers is a rational. but we shall not make the distinction here.

A function whose values are in the set of extended real numbers is called an extended real-valued function. Then (p . We define x + 00 = 00. and so has a least element p. we define sup S to be the least upper bound of S.00 is left undefined. This enlarged set is called the set of extended real numbers. 00' (±oo) = ±oo. The operation 00 . but we shall adopt the arbitrary convention that O· 00 = O.x) < (plq) .34 The Real Number System [Ch. By the Axiom of Archimedes there is an integer q > (y . and there is a rational r with n x<r<n y. The set of integers n such that y ~ (n/q) is a nonempty (by the Axiom of Archimedes) set of positive integers. for each real number x.l)/q lies between x and y.l)/q. One use of extended real numbers is in the expression "sup S".n is a rational between x and y.(y . we can find an integer n such that n > .(l/q) = (p . -00-00=-00 -00' (±oo) = =FOO. If x < 0.x. and x = y .00 < x < 00.00. We extend the definition of < to the extended real numbers by postulating .l)lq < y ~ (plq).x. we write sup S = 00. 2 Proof: Let us first suppose 0 ~ x.X)-l. 00 x. Then (l/q) < y . and if we define sup 0 to be . If S is a nonempty set of real numbers with an upper bound. Thus r = (p . and set 00+00=00. Then sup S is defined for all nonempty sets S. and r . Then n x > 0.00.00 00 = . . Similar conventions are adopted with respect to inf S. I + + + 3 The Extended Real Numbers It is often convenient to extend the system of real numbers by the addition of two elements + 00 and .00 X'oo-oo ifx>O if x > 0 x· - for all real numbers x. then in all cases sup E is the smallest extended real number which is greater than or equal to each element of E. If S has no upper bound.

1 there is an N such that for all n 2: N we have x.. In those cases in which it is important to distinguish between the two concepts of a limit we shall try to be explicit by the use of such phrases as 'converges to a real number' or 'converges in the set of extended real numbers. Thus I is a cluster point of (xn) if. and we denote this limit by lim x. 3n ~ N such that \xn . it depends on whether or not we mean a limit which is a real number or an extended real number. Xn+b we write Xn i I. 4] Sequences of Real Numbers 35 4 Sequences of Real Numbers By a sequence" (xn) of real numbers we mean a function which maps each natural number n into the real number Xno We say that a real number' is a limit of the sequence (xv) if for each positive ~there is an N such that for all n 2: N we have IXn . A sequence is called convergent if it has a limit. If. This definition is ambiguous. > D. there is an N such that for all n 2: N and all In 2: N we have IXn . It is easily verified that a sequence can have at most one limit. In this case we say that I is a cluster point of the sequence (xn). (xn> is monotone. In the case of a real number we can paraphrase the definition of limit as follows: I is the limit of (xn) if. when it exists.Sec. we often write Xn -7 I. given . given € > 0. a1lbut a finite number of terms of the sequence (xn) are within € of I. . Xn ::. given € > 0.xml < E. which requires a limit to be a real number. and given N. In most of analysis it is more usual to use the restricted definition for convergence. In symbols I = lim Xn if (E > 0)(3N)(n 2: N)(lxn - II < E). that is. A weaker requirement is to have infinitely many terms of the sequence within € of I. given E > 0. The Cauchy Criterion states that a sequence of real numbers converges if and only if it is a Cauchy sequence (see Problem 10).IJ < ~. We extend this notion of limit of a sequence to include the value 00 as follows: lim Xn = 00 if given . Since we shall be mostly interested in infinite sequences in the remainder of this book. we drop the adjective "infinite" and assume all sequences infinite unless otherwise specified. We extend this definition to the case I = 00 by saying that OCI is a cluster point of (xn> if.1 2 These are infinite sequences in the terminology of Chapter 1. but we shall find it convenient in the next few chapters to allow ± 00 as limits in good standing.'I < E.' If I = lim Xn. in addition. A sequence (xn) of real numbers is called a Cauchy sequence if.

Xn = -lim Xm and lim Xn :::. there is a k ~ n such that Xk > Ll. we have lim x. the sequence (-l)n has + 1 and -1 as cluster points but we define its limit superior by lim Xn = inf sup n kc.n xi. lim (x. b. The extended real number r/J is the limit superior of (xn) if and only if. :::. Show that lim Xn and lim x. then I is a cluster is not usually true. 8. The symbols lim and lim sup are both used for the limit superior.:::: N applies to .co is the limit superior of (xn) if and only if . and given n.e..36 The Real Number System ECho 2 and given N.. = sup inf n kc.co . given Ll and n. Show that a sequence can have at most one limit. 7. 3n . = has no limit. If (xn) and (Yn) are two sequences. Thus. Is this statement true if we omit the word 'extended'? .lim Xn + lim Yn- Problems 6.lim + lim Yn (x. + Yn) :::. a. point.: The sequence (xn) converges to an extended real number I if and only if 1 = lim Xn = lim Xn. If (xn! is a sequence.co = lim Xn• We define the limit inferior by lim x. The extended real number . Show that a sequence (xn> is convergent if and only if there is exactly one extended real number which is a cluster point of the sequence. For example. Show that every bounded infinite sequence has a subsequence which converges to a real number. but the converse (xn) defined by x.n xi: We have lim .::::n such that Xk > I .::::Ll. 3n such that Xk < 1 + e for all k . Show that I is a cluster point of (xn) if and only if there is a subsequence (Xn)':=l which converges to I. A real number I is the limit superior of the sequence (xn) if and only if (i) given € > 0. An obvious modification if a sequence has a limit I. e > 0. 3k . + lim Yn :::. are the largest and smallest cluster points of the sequence (xn). 9.:::: and (ii) given n. + Yn) :::. lim x. lim x. such that Xn .

lim Xn 00 + lim Yn. 4] Sequences of Real Numbers 37 10.. and (ii) given to > 0 and n. then the original sequence converges to 1. lim Xn and that lim Xn only if I = lim Xn. Show that if a Cauchy sequence has a subsequence which converges to I.'). 14. provided the right and left sides are not of the form 16. Establish the Cauchy Criterion: There is a real number 1 to which the sequence (Xnl converges if and only if (xnl is a Cauchy sequence. ~ 0.oo = Xl +L 00 (Xv+1 - x. (lim xn)(lim Yn). 00. Show that each Cauchy sequence is bounded. v=1 L x. d. has n 18. Show that lim Xn :::.6. + Yn) :::. = I if and lim x. Show that x and only if x = lim x« if n--. Let (xnl be a sequence of real numbers. Show that lim Xn Xk > = 00 if and only if given Lland n. We shall say that a sequence (or series) (xv I is summable to the real number s or has a sum s if the sequence (sn) defined by Sn = S as a limit. b.. Show that the series (XvI has a sum if 19. + lim Yn :::. Show that the real number 1 is the limit superior of the sequence (xn) if and only if (i) given E > 0. lim (x.::. Show that x = lim Xn if and only if every subsequence of (xn) has in turn a subsequence which converges to x. c. ~ ° and Yn ~ 0. 11. 12. 17. In this case we write s = there is always an extended real number s such that v=1 L 00 xv' Show that. if each x. a. Show that a sequence (xn) which converges to a real number 1 is a Cauchy sequence. 15. + 13. Prove that if x. 3k ~ n with = . Prove that lim x. 3k ~ n such that Xk > I-E. v=l . 3n such that Xk < 1 E for all k ~ n. then lim (xnYn) . provided the product on the right is not of the form 0· 00.Sec.

. b. bJ is defined to be {x: a < x ::.x ::. in which case there are exactly two such sequences. Sometimes we write (. this sequence is called the decimal expansion of X.38 The Real Number System ECho 2 20. 00) = {x: a < x} and (-!JJ.4. A generalization of the notion of an open interval is given by that of an open set: Definition: A set 0 of real numbers is called open if for each x e 0 there is a o > 0 such that each y with Ix . Show that R is uncountable. We define the open interval (a. If p = 10. a. Let E be a set of positive real numbers. (0) for the set of all real numbers. Show that xeE Lx < 00 00 only if E is countable. Show that if E is countable and (Xn) is a one-to-one mapping of N onto E. Another proof 5 Open and Closed Sets of Real Numbers The simplest sets of real numbers are the intervals. sup Fe'J SF. 0 < x < l.. 22. Show that. the series converges to a real number x with 0 ::. will be given by Corollary 3.. the ternary expansion.yl < 0 belongs to O. x ~ 1. and for p = 3. Show that there is a sequence (an> of integers with 0 ~ an < p such that and that this sequence is unique except when x is of the form q/pn. For closed intervals we take a and b finite. and [a. For p = 2 it is called the binary expansion. b) = {x: a ~ x < b}. bJ to be the set {x: a ::. The half-open interval (a. then xeE Lx = n=l L Xn- 21. Let p be an integer greater than 1. if (an) is any sequence of integers with 0 ::.23.] [Use Problem 1. b) = {x: x < b}. We define SF Lx xeE to be where if is the collection of finite subsets of E and is the (finite) sum of the elements of F. b}.00. We always take a < b but we consider also the infinite intervals (a. b) to be the set {x: a < x < b}. b} . an < p. conversely. We define the closed interval [a. and x a real number.

however.yl < 02 belong to O2. and {a} is not n~1 set. Leta = inf {z: (z. form of the converse of this is also true: n 00 8. we have by the definition of b a number y > w such that (x. and so w e O.e. Then a < x < b. union of open intervals is an open set by Proposition 7. Corollary: The intersection of any finite collection of open sets is open. Take. and x E U.yl < 0. 7. On to be an open Every A strong the open interval (-~. to 01 n O2. and both the empty set )25 and the set R of real numbers are open.yl < 01 belong to 01• Similarly. b) is an open interval containing x. I It follows from Proposition 5 that the intersection of any finite collection of open sets is open. for example.. Proof: Let U be the union of the collection e. there is a 01 > 0 such that all y with [x . I 6. Then On = {O}. then y belongs to both 01 and O2. Proof: Since 0 is open. y) CO. Since 0 is open. there is an E > 0 such that all y with [x . Since x E 01 and 01 is open. Then there is an 0 E e with x EO. for each x E 0 there is a y > x such that (x. there is a O2 > 0 such that all y with [x . i. Now Ix c 0. Proposition: The intersection 01 n O2 of two open sets 01 and O2 is open.yl < E belong to 0 and hence to U. that the intersection of any collection of open sets is open. ~). y) cO. Proposition: Every open set of real numbers is the union of a countable collection of disjoint open intervals. since o c U. Letb = sup {y: (x. 5. x) cO}. Proof: Let x E 01 n O2. and Ix = (a. . for if w E Is. It is not true. The open intervals are examples of open sets. say x < w < b. Then 0 > 0. y) cO}.Sec. Take 0 to be the smaller of 01 and O2. Thus U is open. 5] Open and Closed Sets of Real Numbers Another way of phrasing this definition is to say that a set 0 is open if for every x in 0 there is an open interval I such that x e I c O. and if [x . We establish some properties of open sets. Proposition: The union of any collection e of open sets is open.

Since a does not belong to 0 and hence not to (c. b) = (c. we have 0 = Let (a.40 The Real Number System [Ch. X B O. b = d. and (a. CIJ Proof: Let U = U{O: 0 Be}. But each open interval contains a rational number by the corollary to the Axiom of Archimedes. is contained in 0. b) and we have c :s: a.yl < o. then for some E > 0 we have (b . It follows from Corollary 4 that we can find an open interval Jx with rational endpoints such that x B Jx C Ix. Proposition (Lindelof): Let e be a collection of open sets of real numbers. Since c does not belong to 0. b t 0. we have a ::. Then we must have c < b and a < d. whence (x.c. Since the collection of all open intervals with rational endpoints is countable. b E) C 0. and U = U Jx• For each o interval in {/J choose a set 0 in countable subcollection e which IgU contains it. a t o. each open interval contains a different rational number. and the collection can be put in one-to-one correspondence with a subset of the rationals. d). Since we have a collection of disjoint open intervals. Then there is a countable subcollection {oj of e such that Oge U 0= t=l U o. I + + uz. Each . and U = U 00 01• I We shall also study the notion of a closed set which generalizes the notion of a closed interval. d) be two intervals in this collection with a point in common. d). 2 Moreover. for if b B 0. Then there is an B e with x B o. We begin by defining a point of closure: Definition: A real number x is called a point of closure of a set E if/or every 0 > 0 there is a yin E such that [x . Thus a = c. Similarly. Thus 0 is the union of the disjoint collection {Ix} of open intervals. there is an open interval I« such that x B Ix C O. Consider the collection of open intervals {Ix}. Thus two different intervals in the collection {Ix} must be disjoint. Since 0 is open. is countable. b E) C 0.E. This gives a t=l {OJ:'l of e. and each I. and x E U. Thus it is a countable collection.. the collection {Jx}. contradicting the defirntion of b. Similarly. b) and (c. This is equivalent to saying that x is a point of closure of E if every open interval containing x also contains a point of E. and it remains only to show that this collection is countable. Since each x in 0 is contained in Ix. 9. x B U. it does not belong to (a.

Thus [x .zl of closure of x be a point of closure of E. Proposition: The intersection of any collection e of closed sets is closed.yi < 02. Thus :Ii u B c (A u B). Then there is a 01 > such that there is no y £ A with [x . and we have (A u B) c A u B. 10.z] < 0. Since A C A u B.Sec. Also (A u B) = A u B. Since we always have Fe F. Proposition: For any set E the set E is closed. 5] Open and Closed Sets of Real Numbers 41 point of E is trivially a point of closure of E. (2)' there is no y £ A u B with [x . We denote the set of points of closure of E by E. Thus E c E. Since such a y belongs to each F £ e. and there is a 02 > such that there is no y £ B with [x . Proposition: The union closed. there is a z e E < 0/2. Similarly. that is. there e E with Ix . It is customary to use the letter F to denote closed sets (French. if F contains all of its points of closure. B c (A u B). Then. I FI U F2 12. Proof: By Proposition of two closed sets FI and F2 is 10 we have 13. given o > 0. Suppose that x it A u B. DO) are closed. Then. then :Ii C B. there is a y e {F: F £ e} such that Ix . and we see that x is a point E. Thus if 0 = min (01.yl < o. 11. The empty set )25 and the set R of all real numbers are closed. we see that x is a point of closure of each n n . E = E. a set F is closed if F C F. we have A C (A u B). fermi). b] and [a. I °° Definition: A set F is called closed ifF = F. x it (A u B).YI < 8/2. Proposition: If A c B. Consequently.yl < 01. Proof: Let x be a point of closure of [F: F £ e) .yl < O. given 0 > 0. Proof: The first part follows immediately from the definition of points of closure.' that is. The closed intervals [a. Proof: Let is a point y with Iy . Since y e E.

there is an E > 0 such that the interval (c . If e contains only a finite number of sets. With this terminology we state the following theorem: It 15. we call e a finite covering. Since c e [a.00 < a < b < 00. Thus F is open. This terminology is inconsistent: In 'open covering' the adjective 'open' refers to the sets in the covering. I We say that a collection e of sets covers a set F if FeU {O: 0 e e} . where .yl < 0. Thus the term 'open covering' is an abuse of language and should properly be 'covering by open sets' . Since each F is closed. Then each open covering of F has a finite subcooering. b]. C E) + . in 'finite covering' the adjective 'finite' refers to the collection and does not imply that the sets in the collection are finite sets.. If x e 0.42 The Real Number System [Ch. then yeO. since x is not a point of closure of F. then y e F.E. Let E be the set of numbers x ::. let F be closed and x e F. x] can be covered by a finite num ber of the sets of e. b]. there is an 0 e e which contains c. the former terminology is well established in mathematics. Hence xe {F: Fee}. Proposition: The complement of an open set is closed and the complement of a closed set is open. 2 Fee. That is.. Since 0 is open.yl < o. there is a 0 > 0 such that if [x . there is a 0 > 0 such that there is no y e F with [x .b with the property that the interval [a.. since there is no yeO with Ix . Proof: Let 0 be open. Proof: Let us first consider the case that F is the closed interval [a.yl < 0. we have x e F for each F in e. Thus 0 contains all of its points of closure and is therefore closed. then there is a finite collection {Ob . Theorem (Heine-Borel): Let F be a closed and bounded set of real numbers. . The set E is bounded by b and so has a least upper bound c. On the other hand.yl < o. Then. Unfortunately. If each 0 e e is open. On} of sets in e such that n FeU i=l 0. In this case the collection e is called a covering of F. Hence. if [x . I n 14. Hence x cannot be a point of closure of 0. we call e an open covering of F. if e is a collection of open sets such that FeU {O: 0 e e}.

A number x is called an accumulation point of a set E if it is a point of closure of E ~ {x}.. there is a finite collection {Ob"" Ok} of sets in e which covers [a. What are the sets of real numbers that are both open and closed? 25. Is the set of rational numbers open or closed? 24. If the subcollection contains F. b] can be covered by a finite number of sets from e. F}. Show that x is a point of closure of Eif and only if there is a sequence 27. F is open. On. If this finite subcollection does not contain ft.. Proposition: Let e be a collection of closed sets (of real numbers) with the property that every finite subcollection of e has a nonempty intersection. By hypothesis FeU {O: 0 c e}.e is not an upper bound for E. . it is a subcollection of e and the conclusion of our theorem holds. Since F is bounded it is contained in some closed bounded interval [a. 5] Open and Closed Sets of Real Numbers 43 is contained in O. b]. Find two sets A and B such that A n B (Yn) with Yn e E and x = lim Yn' = 0 and A n 13~ 0. and e* is a collection of open sets. c + e).Sec. e* = e u {F}.. Since no point of F is contained in F. Then Fe F u 01 U U On.e. we have Fe 01 u U Om and the collection {Ol' . Since F is closed. c + e) would be in E if it were less than or equal to b. On} is a finite subcollection of e which covers F. 26. Since no point of [c. By our previous case there is a finite subcollection of e* which covers [a. Show that the set E' of accumulation points of E is a closed set.. Consequently. and suppose that one of the sets in e is bounded. Now c . and so there must be an x c E with x > c . we must have c = band b e E. Then Fef2 nF ~ 0. c + e) except c can belong to E. Thus each point of [c. .. . I 16. O} covers [a. x]. proving our special case. Since x c E. denote it by {Ol'" . Problems 23. b] and hence F. b]. the finite collection {Ol' . Oi. Let e* be the collection obtained by adding F to e: that is. Thus e* is an open covering ofR and therefore of [a. Now let F be any closed and bounded set and e an open covering of F. and so R = F u Fe F u U {O: 0 c e} = U {O: 0 c e*}. Thus [a.

A set is called isolated if EnE' of real numbers is countable. 33. 32. x + 0) is contained in A. 35. AO = AO. and so on. The functionfis said to be continuous on a subset A of E if it is continuous at each point of A. 38. is bounded.yl < 0 we have I/(x) . Problem 21) (an> for which an is never 1. Show that every isolated set 30. A set D is called dense in R if D = R. Show that E = E u E'. 1] which have ternary expansion (cf. Give an example to show that this conclusion may be false if we do not require one of the sets to be bounded. 2 28.(A). = 29. we put x in the Cantor set if one of the expansions has no term equal to 1. If we merely say thatfis continuous. Derive Proposition De Morgan's laws. 1]. 6 Continuous Functions Letfbe a real-valued function whose domain of definition is a set E of real numbers. A point x is called an interior point of a set A if there is a 0 > 0 such that the interval (x . ~) from [0. 1]. Show that the set of accumulation points of the Cantor set is the Cantor set itself. usmg = ". then ~=1 n F. then removing the middle thirds (§-. Let <Fn> be a sequence of nonempty closed sets of real numbers with Fn+ 1 C Fn.44 The Real Number System [Ch. The set of interior points of A is denoted by AO. A is open if and only if A b. we mean thatfis continuous on its domain. there is a 0 > 0 such that for all yin E with [x . Show that if one of the sets F. 16 from the Heine-Borel Theorem 34. 13. Show that a. and that C is obtained by first removing the middle third Ct. 37.0. Show that the Cantor set can be put into a one-to-one correspondence with the interval [0. and 14. (If x has two ternary expansions.~) of the remaining intervals. 36. and 14. given E > 0. 31. We say that f is continuous at the point x in E if. 0.) Show that C is a closed set. . Show that the set of rational numbers is dense in R.f(y) I < E. Prove Propositions 5 and 7 using Propositions 12. Prove Propositions 12 and 13 using Propositions 5.~) and (~. 7. ~ 00 95. The Cantor ternary set C consists of all those real numbers in [0.

.. and hence If(y) I < 1 If(xk)I sM. we have 1 f(y) 1 S 1 f(x) 1 + 1 and sofis bounded in t. Consequently.f(x) 1 < 1 for y e I. and our goal is to show that there is an Xl e F such that f(Xl) = m. given e > 0. that is.• + + + 18. f(x) s f(X2) for all x in F. Since f is continuous on F.. f(x) e) is an open set. and let x be an arbitrary real number. The collection {Ix: x e F} is a collection of open intervals which covers F. . for each x e Fthere is an open interval Z. there is an X2 at whichfassumes its minimum . 6] Continuous Functions 45 17.e. containing x such that If(y) . But this is impossible. . Suppose now that 1 continuous and that 0 is an open set. But this implies that if Iy . we can find a finite number of these intervals {Ixi' . Hence f is continuous at x. This shows that f is bounded (by M) on F. x 0) Cf-l[I]. Then each y in F belongs to some interval IXk in the finite subcollection.xl < 0. Let M = 1 max [if(xl)l. there must be some 0 > such that (x . since tea m) < m. Then. and so its inverse image f-l[I] must be open. m is finite. then fey) e (f(x) . Then f is bounded on F and assumes its maximum and minimum on F. Ix) which covers F. and by continuity there is an open interval Ix containing x such that fey) < Hf(x) m) for all y e Ix n F. Since + + f xeF is bounded. n F.Ix) which cover F. Since x was arbitrary. n F. Suppose not.. Then f(x) < m for each X e F. Since x e l-l[I]. . Let is x be a point off-1[0]. .00.0.Sec.f(y) 1 < e. there are points Xl and X2 in F such thatf(xI) . Similarly. Thenf(x) e 0. l. that is. let m = sup f(x). To see that f assumes its maximum on F. the interval I = (l(x) .Borel Theorem. Then each y e F belongs to one such interval IXk and f(y) < ![ f(Xk) m] S !(a + m). Proposition: Let f be a real-valued function defined and continuous on a closed and bounded set F. 00). Set a = max [f(Xl). [(x) e).. Thus for y e I.If(xn)l].. . . Thus Ha + m) is a bound for Ion F.f(xn)]. and by the Heine-Borel Theorem there is a finite subcollection {IxI' . Again using the Heine. Proof: We shall first show that f is bounded on F. Proof: Suppose 1-1[0] is open for each open set 0.:::.f(x) . there must be an Xl such thatf(xl) = m. (is continuous. and there is an e > 0 such that + + ° + . Proposition: Let f be a real-valued function defined on ( .e... Then f is continuous if and only if for each open set 0 of real numbersf-l[O] is an open set.

Hence ex . . < +0 ~ ox.. for every y e (x .~ox.x2) < ~ox.yl < ax implies Ifex) . Given two points y and z in Fsuch that Iy .f(y») < ~f. .f(x) + f) C O.f.fey») < f. Ox) which covers F. Then {Ox: x e F} is an open covering of F By the Heine-Borel Theorem there is a finite subcollection [OXI' . there is an N such that for all n ~ N we have )f(x) . Definition: A sequence <fn? of functions defined on a set E is said to converge uniformly on E if.0. x + ~ox). I Definition: A sequence <fn> of functions defined on a set E is said to converge pointwise on E to a function f if for every x in E we have f(x) = limfn(x).. 2 (f(x) . Proposition: If a real-calued Given f > and x in F. Proof: 1 ' ••• . x + 0) C f-1[0].x21 ~ [z . Consequently. I ° Definition: A real-valued function f defined on a set E is said to be uniformly continuous (on E) if. Let 0 = ~ min (ox ox. Let Ox be the interval (x .. there is a 0 > such that Ifex) . x + 0). it is uniformly continuous on F. Thus. there is a 0 > such that for all x and y in E with \x . and so f-1[0) is open. that is.46 The Real Number System [Ch..f.f(y)/ < f for [x . the point y must belong to some Ox" and hence there is an i such that Iy . given f > 0. we havef(y) e (f(x) . f(x )i i f and < -. there is a Ox > such that [x . ° 19.yl Hence If(y) + Iy - -x 2 2) < ~ox.yl < o. given f > 0.fnex) \ < f. 2 whence If(z) . ° ° [z .zl < 0.0. showing that f is uniformly continuous on F.f(x») < E. if. there is an N such that for all x 8 E and all n ~ N we have Ifnex) . f(x) + f) C O. Sincefis continuous at x.y\ < 0 we have If(x) . given x r E and f > 0. function f is defined and continuous on a closed and bounded set F of real numbers.) Then 0 is positive.fey») < f.

b] with )f(x) . 45. 61 Contmuous Functions 47 Problems 39. fey) whenever x:S. y. Let Un) be a sequence of continuous functions defined on a set E. Show that if I and g are continuous so are Iv g and j" A g. A continuous function sp on [a. Prove that if Un) converges uniformly to Ion E then I is continuous on E. 1) with the ternary expansion (an> (cf. b] is called polygonal (or piecewise linear) if there is. Let / be a continuous function on the interval [a. Show that if I and g are continuous functions. If/is continuous. It is called monotone (or strictly monotone) if I or -lis monotone (or strictly monotone) increasing.. Let N = co if none of the an are 1.r.Sec. d. < x. b].. A real-valued function I is said to be monotone increasing if I(x) :s.co . b. In this case we also have f(g(y» = y for each y between lea) and f'tb).+ 1]' Let f be an arbitrary continuous function on [a. A function I which has a continuous inverse is called a homeomorphism (between its domain and its range). [Hint: Take g to be linear in each of the intervals of which F is composed. then so is/o g. LetI V g be the function defined by (f V g)(x) = f(x) V g(x). q At what points is f continuous? /+ 43. a. Prove that/is continuous if and only if for each open set 0 there is an open set U such that /-1[0] = En U. Show that iff and g are continuous. then so is If I· 44. Let I be that function defined by setting if x irrational J(x) ={ ~ 1 p smq if x = f!_ in lowest terms. and define / A g similarly. Then there is a continuous function g such that g(f(x» = x for all x e [a.p(x») < f for all x B [a. b] and f a positive number. and otherwise let N . = b such that r. Let x be a real number in [0. 00) such that/ex) = g(x) for each x e F. then the functions g andlg are continuous. b].p on [a. a subdivision a = Xo < x .] 40. c. b] if and only iff is strictly monotone. 46. Let F be a closed set of real numbers and/ a real-valued function which is defined and continuous on F. and strictly monotone increasing if f(x) < fry) whenever x < y. Problem 21). Letfbe a real-valued function with domain E.p is linear on each interval (x" x. Show that there is a polygonal function r. Show that there is a function g defined and continuous on (. 42. < . 41.

Show that f is constant on each interval contained in the complement of the Cantor ternary set (Problem 36). then there is a sequence (xn> with = Y such that y = lim x. monotone function on the interval [0. d. f. c. Similarly.00 and fey) limf(x). given € > 0. lim f(x) ::. f is called upper semicontinuous at y if f(y) . 1]. 47.yl < 0 we havef(x) S A + E.48 be the smallest value of n such that an bN = 1. 48. a. .€. limf(x) ::::: if and only if.i of a real variable.F Y and y = lim x. Semicontinuous functions.>y lim f(x) = inf = o>OO<lx-yl<o sup fex) f(x) lim f(x) x-'>y+ inf 0>0 sup O<x-y<o with similar definitions for lim. there is a a > Osuch x such that x->y 0 < Ix . b.yl < a andf(x) 2:: A . lim fex) x->y that for all x with < [x .. 2 1.F + 00 x->y lim f(x) if and only if I = s . = !an for n < Nand is independent of the ternary expansion of x (if x has two expansions) and that the function f defined by setting is a continuous. 1]. Show that = The Real Number System [Ch.F Y such that Xn . then limf(xn) e. and A = limf(xn). If limf(x) x->y = A and {xn> is a sequence with ::. lim f(x) with equality (for limf. and thatfmaps the Cantor ternary set onto the interval [0..F x-'>y x-'>y x->y ± (0) if and only if limj'(x) exists.F y = lim xn. Let b. Let f be a real (or extended real) valued function defined for all x in an interval containing y. If limf(x) x->y A.. given E > 0 and 0 > 0. An extended real-valued function f is called lower semicontinuous at the point y if f(y) . (This function is called the Cantor ternary function. Xn . For a real number l we have I limf(xn) x->y for every sequence (xn> with x. Limit superior of a function We define x-. there is an A ° sA if and only if.F . = A.

b]. f(x) € for all x with Ix .) is less than or equal to the smaller of the two values assumed in (x. 1»O/x-y/<1l . bJ. The function f is upper semicontinuous if and only if the function . h. that is. Show that if V g andf g. 49. Let feY) be finite.yl < o. ~o > 0 such that f(y) . b] such that for each x e [a. in (g) can be replaced by con- I.. A functionf defined on an interval [a.+l)' Show that a step function 'P is lower semi continuous iff 'P(x. b] such thatf(y) ::. b) if and only if the set {x: f(x) > X} is open for each real number A.:s. aref d. b] we have f(x) = lim 'Pnex). We define the lower envelope g off to be the function g defined by g(y) = sup 0<0 Ix-yl<o inf f(x). Prove that a function f which is defined and lower sernicontinuous on a closed interval [a. that there is a y 8 [a. 6] Continuous Functions 49 on and f(y) 2:: Jimf(x)._ f(x) for all x 8 [a. Let f be a real-valued function defined on [a.Sec. Let Un) be a sequence of lower semicontinuous functions.. Prove that f is lower semicontinuous at y if and only if given e > 0. g. A function f is continuous (at a point or in an interval) if and only if it is both upper and lower semicontinuous (at the point or in the interval). x--.y We say that f is lower (upper) semicontinuous an interval if it is lower (upper) sernicontinuous at each point of the interval. b J is called a step function if there is a partition a = Xo < Xl < . and the upper envelope h by h(y) = inf sup f(x). Show that a real-valued function f is lower semicontinuous on (a. + b._ b x. b] is bounded from below and assumes its minimum on [a. c. so e. f. bJ.) and (X1' Xi+l). a. Show that the function f defined by f(x) = supfn(x) is also lower semin continuous.f is lower semicontinuous. + f and g are lower semicontinuous functions. A real-valued function 'P defined on an interval [a.. Show that the step functions tinuous functions. b] is lower semicontinuous if and only if there is a monotone increasing sequence ('Pn) of lower semicontinuous step functions on [a. Upper and lower envelopes of a function. < Xn = b such that for each i the function 'P assumes only one value in the interval (x" X.

This leads us to the following definition: Definition: The collection ill of Borel sets is the smallest a-algebra which contains all of the open sets. and g(x) = f(x) if and only iff is lower semicontinuous at x. of course. If f is bounded. Thus the complement of an 5'rr is a g~. Such a smallest e-algebra exists by Proposition 1. A countable union of sets in 5'rr is again in 5' a» Since (a. For example. b]. the set of rational numbers is the union of a countable collection of closed sets each of which contains exactly one number.and conversely. while g(x) = hex) if and only iff is continuous at x. 2 a. We could also consider sets of type 5'rr~' which are the intersections of countable . the union of a countable collection of closed sets need not be closed. The 5'rr and g~ are relatively simple types of Borel sets.3. every closed set. 0 for durchschnitt). and hence each open set is an 5'rr. We say that a set is a g~ if it is the intersection of a countable collection of open sets (g for open. b]. each open interval is an 5'rr. n b- - n IJ . A set which is a countable union of closed sets is called an 5' a (5' for closed. b. b]. Thus every countable set is an 5'rr. while h is upper semicontinuous. b) = U 00 [ a n=l I + -. then lI'(x) ~ g(x) for all x e [a. If II' is any lower semicontinuousfunction such that 'P(x) :$ f(x) for all x e [a. hex). the function g is lower semicontinuous. Thus if we are interested in c-algebras of sets which contain all of the closed sets we must consider more general types of sets than the open and closed sets. For each x e [a. It is also the smallest e-algebra which contains all closed sets and the smallest e-algebra which contains the open intervals.50 The Real Number System [Ch. 7 Borel Sets Although the intersection of any collection of closed sets is closed and the union of any finite collection of closed sets is closed. c. (T for sum). g(x) :$ f(x) ::. as is.

Letfbe a lower semicontinuous function defined for all real numbers.Sec. but we shall need only the properties that follow directly from the fact that they form the smallest e-algebra containing the open and closed sets. we can construct the classes Soer. {x: f(x) :::. Problems 50. {x: f(x) < a}. ~O"oO". etc. What can you say about the sets {x:f(x) > a}. Further theory of Borel sets can be found in Kuratowski [16]. not every Borel set belongs to one of these classes.a}. Prove that the set of points at whichf is continuous is a So. 52. and {x: f(x) = a}? 51. Thus the classes in the two sequences are all classes of Borel sets. Let Un) be a sequence of continuous functions defined on R. Similarly. However. Show that the set C of points where this sequence con verges is an ~ 0"0' . Letfbe a real-valued function defined for all real numbers. 7] Borel Sets 51 collections of sets each of which is an ~O". {x:f(x) ~ a}.

that is. m is translation invariant. to be the difference of the endpoints of the interval. iii. we would like to construct a set function m which assigns to each set E in some collection .m = C9(R).3 Lebesgue Measure 1 Introduction The length 1(1) of an interval I is defined. we should like m to have the following properties: i. iv. if E is a set for which m is defined and if E + y is the set {x + y: x e E}. We should like to extend the notion of length to more complicated sets than intervals. Since the class of open sets is still too restricted for our purposes. ii. that is. a function which associates an extended real number to each set in some collection of sets. we could define the "length" of an open set to be the sum of the lengths of the open intervals of which it is composed. for an interval I. if (En) is a sequence of disjoint sets (for which m is defined). m(U En) = 'LmEn. In the case of length the domain is the collection of all intervals. that is. as usual. . then m(E + y) 52 = mE. Length is an example of a set junction. mE is defined for each set E of real numbers.m of sets of real numbers a nonnegative extended real number mE called the measure of E. . Ideally. obtained by replacing each point x in E by the point x + y. For instance. mI = 1(1).

1 2 If we assume the continuum hypothesis (that every noncountable set of real numbers can be put in one-to-one correspondence with the set of all real numbers). then m 0 = O. . Let m be a countably additive measure defined for all sets in a o-algebra 1. as we shall see in Section 4. 3. which is satisfied by the outer measure we construct in the next section (see Problem 2). This property is called monotonicity. [Hint: Use Proposition 1." We shall want mE to be defined for as many sets as possible and will find it convenient to require the family ~ of sets for which m is defined to be a zr-algebra. Then m(U En} ~ L_mEn. it is also possible to replace property (iii) of countable additivity by the weaker property of finite additivity: for each finite sequence <En) of disjoint sets we have m(U En) = 'E.2.] This property of a measure is called countable subadditicity. Weakening property (I) is not the only approach. This measure is called the counting measure. 1] Introduction 53 Unfortunately.21). and it is not known whether there is a set function satisfying the first three properties. 2.m: with A C B. 4. and it is most useful to retain the last three properties and to weaken the first condition so that mE need not be defined for all sets E of real numbers. Let nE be 00 for an infinite set E and be equal to the number of elements in E for a finite set. one of these properties must be weakened. then such a measure is impossible. If there is a set A in . it is impossible to construct a set function having all four of these properties. If A and B are two sets in . Show that n is a countably additive set function which is translation invariant and defined for all sets of real numbers. Thus we shall say that m is a countably additive measure ifit is a nonnegative extended real-valued function whose domain of definition is a o-algebra ~ of sets (of real numbers) and we have m(U En) = L_mEn for each sequence (En) of disjoint sets in ~. Let <En) be any sequence of sets in ~.Sec.mEn (see Problem 10. Our goal in the next two sections will be the construction of a countably additive measure which is translation invariant and has the property that mI = leI) for each interval I.m: such that mA < 00. 1 Consequently. then mA ~ mB. Problems ~. Another possible alternative to property (iii) is countable subadditivity.

2 Outer Measure For each set A of real numbers consider the countable collections {In} of open intervals which cover A. and for each such collection consider the sum of the lengths of the intervals in the collection. But this is equivalent to showing that if {In} IS any countable collection of open intervals covering [a. there must be an 3 In order to distinguish this outer measure from the more general outer measures to be considered in Chapter 12. lea ..a.). and since b.'s which contains a. b] ::. after Henri Lebesgue. Also each set consisting of a single point has outer measure zero..a.€. Let this be the interval Cal. b] :2: b . b] :s. b]. b] contains a finite subcollection which also covers [a. . b]. bJ for each positive €. b . m*B. then b. b]. bl)' We have al < a < h. By the Heine-Borel theorem. then + (1) L '(In) :2: b .a + 2€. Since we consider no other outer measure in this chapter. b] :s. we have m*[a. that is. we must have m*[a. b. It follows immediately from the definition of m* that m*0 = 0 and that if A C B. collections for which A c U In. this sum is uniquely defined independently of the order of the terms. Proposition: The outer measure of an interval is its length. Since the lengths are positive numbers. Since the open interval (a . say [a. Since m*[a. b . If b i :s.a.. it suffices to prove the inequality (1) for finite collections {In} which cover [a.€. b].a + 2€ for each positive €. b €) contains [a. then m*A ::. b].. Since a is contained in U 1m there must be one of the I. In an abbreviated notation m* A = ACUln inf L 1(1. and since the sum of the lengths of the finite subcollection is no greater than the sum of the lengths of the original collection. Proof: We begin with the case in which we have a closed finite interval. we call this outer measure Lebesgue outer measure. we refer to m* simply as outer measure. b + €) = b . any collection of open intervals covering [a. Thus we have only to show that m*[a. We establish two propositions concerning outer measure: 1. e (aI' bl). We define the outer measure" m*A of A to be the infimum of all such sums. e [a.

< bi_1 < i. if ale < b < be.ble-l) ..:::: l(ai. then. our process must terminate with some interval (ab ble)' But it terminates only if b e (ak' bi). Thus L f(ln) . the inequality holds trivially.i and L l(In. (a2 - bl) . < b2.a..ale-I) + ..a. there is a closed interval J c I with l(J) = A. given E > 0. a2 < b.al > since a.al) = ble . given E > 0. there is a closed interval J c I such that I(J) > 1(1) . < a. If m* An is finite. there is a countable collection {In. E > 0. If I is any finite interval. and so m*I = 1(1).(ale . m* Since m*I . that is.a). Then of sets of real Proof: If one of the sets An has infinite outer measure. + (b .i} n. 2] Outer Measure 55 interval (a2' b2) in the collection {In} such that b i e (a2' bs): that is.i) < m* An + 2-n€. Continuing in this fashion we obtain a sequence (aI' bI). Hence leI) Thus for each E < l(l) = m*J ~ m*I ~ m*i = l(l) = I(I). then.E.a. Hence m* I . > b . whence "Ll(ln) > (b . 1(1) E < m*I ~ 1(1).i} i of open intervals such that An C U In. m*I = 00 = 1(J). If I is an infinite interval.:::: J = I(J) = A. (ab ble) from the collection {IJ such that a.bk-2) . ble aI.. But bk > band a..J n t i is countable. b] = b . < bt_l. This shows that m*[a. being the union of a countable number of .Sec. . then given any real number ii.. and so we have bk . bi) L = (bk - - ak) - + (ble-I . .:::: for each A. Since [In} is a finite collection. I A 2.i = • U {In. Now the collection Un. .(ak_l . Proposition: Let {An} be a countable collection numbers.

Prove Proposition 5. . however. there is an open set such that A c 0 and m*O S. m* A = 0. Prove that m* is translation invariant. and let {In} be a finite collection of open intervals covering A. 6. m*(A n E) m*(A n E). Proposition: Given any set A and any e > 0. it is not countably additive. was an arbitrary positive number. o Problems 5. 3. There is aGe 91l such that A c G and m*A = m*G. and we say E IS Lebesgue measurable. 3 Measurable Sets and lebesgue Measure While outer measure has the advantage that it is defined for all sets. Let A be the set of rational numbers between and 1. 5. and covers U An. m* A + E. 8. we see that E is measurable if (and only if) for each A we have m* A 2:: 4 + In the present case m* IS Lebesgue outer measure. n n. if for each set A we Since we always have m* A S.56 Lebesgue Measure [Ch. Thus n. 1] is not countable. It becomes countably additive. 7. Corollary: The set [0. More general nouons of measurable set are considered m Chapters 11 and 12. if we suitably reduce the family of sets on which it is defined.t = Since E L: mr A. then m*(A U B) = m* B. Then '[j(ln) 2:: 1. 4. + E. Corollary: If A is countable. Prove that if m* A = ° 0. Perhaps the best way of doing this is to use the following definition due to Caratheodory: Definition: A set E is said to be measurable" have m* A = m*(A n E) + m*(A n E). 3 countable collections.

we have E measurable whenever E is. m*(A n [E1 u E2]) :::. I E. me of measurable sets is an algebra of 9.m*(A n E1) Thus m*(A n [E1 u Ed) n E2 n E 1)' + m*(A + m*(A n E1 n £2) n E2 n £1) = m*(A nEd + m*(A < m*(A + m*(A n E1) n £1 n £2) n E1) = m*A. Lemma: Let A be any set. disjoint measurable sets.1 sets E~. Then A nEe m* E = O. then E is measurable. + o 6. Corollary: The family sets.. Proof: Let A be any set. Since E2 is measurable. Lemma: If m* E = 0. En a finite sequence of Proof: We prove the lemma by induction on n. so is E1 U E2• Proof: Let A be any set.(E1 u E2) = E1 n E2.Since . Lemma: If E1 and E2 are measurable. . and E1. Since the definition of measurability IS symmetric in E and E.Sec. by the measurability of E1. 7. 3] Measurable Sets and Lebesgue Measure 57 m*(A n E) m*(A n E). Then . Since ". I 8. and we assume it is true if we have n . and so m*(A n E) < + m*(A n E). and so m* A 2': m*(A n E) = m*(A n £) and therefore E is measurable.. we have m*(A n E1) = m*(A n £1 n E2) + m*(A + m*(A n E1 n £2). Also A => A n E. we have and since A n (E1 u E2) = [A n E1] u [A n E2 n Ed. this shows that E1 u E2 is measurable. Clearly and the set R of all real numbers are measurable. It is clearly true for n = 1.

are disjoint sets. n Then F. Hence t=l U E. By Proposition 1. and let F. is By Lemma 9 m*(A n Fn) = i=l L n m*(A n Ei). = measurable. every set with outer measure zero is measurable.1 sets. that is.58 Lebesgue Measure [Ch. Thus m*A ~ L ~=l n m*(A n Ei) + m*(A n if). and F n ~ E.2 such an E must be the union of a sequence <En! of pairwise disjoint measurable sets. Let A be any set. 3 the E. . we have and Hence the measurability of En implies = m*(A n En) +L i=l n-l m*(A n Ei) by our assumption of the lemma for n . Theorem: The collection ~ of measurable sets is a a-algebra. Moreover. Proof: We have already observed that ~ is an algebra of sets. the complement of a measurable set is measurable and the union (and intersection) of a countable collection of measurable sets is measurable. and so we have only to prove that if a set E is the union of a countable collection of measurable sets it is measurable. I 10.

3] Measurable Sets and Lebesgue Measure 59 Since the left side of this inequality is independent of n. If m* A = 00.00. I 11. In particular each open set and each closed set is measurable. we have Thus m*Al + m*A 2 L (m*l~ + m*I::) ~L l(ln) < m* A + ~ E. since A 2 C U I~'. A2 = An (-oo. 00). we have m* Al ~ m*(U I~) < L -et. Proof: Let A be any set. Then we must show m* A I m* A 2 ~ m* A. 00) and I~' = In n (. given E > 0. we have m*A >L i=l 00 m*(A n Ei) + m*(A n E) . and so we must have m* A I m*A2 ~ m*A. Lemma: The interval (a. a]. and. Then I~ and I~' are intervals (or empty) and I(In) = I(I~) + 1(1::) = m*I~ + m*I::.1 + 12. Al = A n (a.a]. .. Since A leU I~. there is a countable collection {In} of open intervals which cover A and for + which Let I~ = In n (a. If m* A < 00.Sec. then. then there is nothing to prove. (0) is measurable. Theorem: Every Borel set is measurable.::: *(A n E) m + m*(A n E) by the countable subadditivity of m*. But E was an arbitrary positive number.

00. i=l 00 n . a] measurable for each a since (. If (Ei) is a finite sequence of disjoint measurable sets. Then i=l U s. Hence each open interval (a. Since (-oo.b - lin). a] = ".00.00. Two important properties of Lebesgue measure are summarized by the following propositions: 13. we define the Lebesgue measure mE to be the outer measure of E. we have (. Then If the sets En are pairwise disjoint. [Note: The theorem also follows immediately from the fact that mr is a o-algebra containing each interval of the form (a. then Proof: The inequality is simply a restatement of the subadditivity of m* given by Proposition 2. since CB is the smallest o-algebra containing the open sets. Let (Ei) be an infinite sequence of pairwise disjoint measurable sets. J I If E is a measurable set.b) = U n=l 00 (-oo. (0) and the fact that CB is the smallest o-algebra containing all such intervals. Thus mr is a o-algebra containing the open sets and must therefore contain the family CB of Borel sets. But each open set is the union of a countable number of open intervals and so must be measurable. Thus m is the set function obtained by restricting the set function m* to the family mr of measurable sets. b) measur- able. we have (-00. (0) is measurable. b) = (. Proposition: Let (Ei) be a sequence of measurable sets.» U e. 3 Proof: Since the collection mr of measurable sets is au-algebra. b) n (a.60 LebesgueMeasure [Ch. (0).(a. then Lemma 9 with A = R implies that and so m is finitely additive.

.mEn) .. Hence m(E1 . = lim mEnn--->co i=l nE 00 and let Fi E1. E i=l But mEl = mE + m(E1 . ) = E t=l L mr.. Ei+1) = mli. E Ei+1).. ). ...Then i=l U r.Sec... ::. and mli.. = mEi+1 + m(Ei ....mEandm(Ei. E = = co Ei. since E c E1 and Ei+l C Ei.mEi+1. .. i+1).. ..mE = i=l L ao (mli.... are pairwise disjoint.... 3] Measurable Sets and Lebesgue Measure 61 and so Since the left side of this inequality is independent of n... and the sets F...mEi+1) lim n.. Ei+1.Thus mli. Let mEl ((1 Ei) t= 1 i...r-}CO i=l L n (mli. Then m Proof: Let E = c Enlor each n.... we have The reverse inequality follows from countable subadditivity..mEl < 00.. a sequence with En+1 be finite. = L co 00 m(Ei .. E) = mEl .. we have m(E1. and we have 14.. Proposition: Let (En! be an infinite decreasing sequence of measurable sets.. ... Since mii.--. that is....mEi+1) = lim (mEl .

we have mE = lim mEn. E is measurable.. Problems 9. Show that the condition mE 1 < 00 is necessary in Proposition 14 by giving a decreasing sequence (En) of measurable sets with 0 = En and mEn = 00 for each n. is finite. e. Then m*(A n t=1 U 00 E. Proposition 5). there is an F in Ifm*E VI. 3 Since mEl < 00.. 15. there is a closed set FeE with m*(E '" F) E. (i) => (ii) 13. Let (Etl be a sequence of disjoint measurable sets and A any set. n 12. then each translate E also measurable. there is a G in S" with E c G.F) = 0. Prove Proposition 15. there is a finite union U of open intervals such that m*(U 6. m*(E . 11. I The following proposition expresses a number of ways in which a measurable set is very nearly a nice set. m(E1 n E2) = mEt mE2. Then the following five statements are equivalent: i.] .). E) < E. the above statements are equivalent to: given E > 0. Use (a) to show that for arbitrary sets E. Show that if E 1 and E2 are measurable. + y of E is 1 + then m(E U E2) + 11.. < E. Proposition: Let E be a given set. Show that for m*E < 00. b.. Show that if E is a measurable set. (i) => (ii) => (iv) => (i).) = t=l L m*(A 00 n E. m*(G '" E) = 0. there is an open set 0 :) E with m*(O '" E) < 0. [Hints: a. IV.62 Lebesgue Measure [Ch. The proof is left to the reader (Problem 13). 10. Ill. ¢::} (vi) (cf. with FeE. given given E E > > 0. Use (b) to show that (i) => (iii) => (v) => (i). v. c.

1). 1) be a measurable set. 1). If x and yare real numbers in [0. Hence E y is measurable and m(E y) = m(El y) + m(E2 y) = mli. The following lemma shows that Lebesgue measure is invariant under translation modulo 1. 1] constructed in the same manner as the Cantor ternary set except that each of the intervals removed at the nth step has length (y}-n with < a < 1. translation modulo 1 by y corresponds to rotation through an angle of 27rY. Let F be a subset of [0.y. Let us denote the sum modulo 1 of x and y by x y. Then for each y e [0. If E is a subset of [0. 1) into numbers in [0. F dense in [0. and so Now E. + + + + 16. Such a set F is called a generalized Cantor set. Also E2 y= E2 + (y .1 if x + y ~ 1.y) and E2 = En [1 . I . + mE2 = mE . 4] A Nonmeasurable Set 63 14. Then F is a closed set. If we consider addition modulo 1 as addition of angles. we define the translate modulo 1 of E to be the set E y = {z: z = x y for some x 8 E}.Sec. and to be x + y . we define the sum modulo 1 of x and y to be x + y. 1) the angle 27rx. and E2 are disjoint measurable sets whose union is E. 1). Show that the Cantor ternary set (Problem 2. a. y is measurable and we have m(El y) = mEl. and so E2 y is measurable and m(E2 y) = mE2• But E y = (El y) u (E2 y) and the sets (El y) and (E2 y) are disjoint measurable sets. + + Proof: Let E. if x + y < 1. ° *4 A Nonmeasurable Set We are going to show the existence of a nonmeasurable set. + y. = En [0. If we assign to each x 8 [0. 1] and mF = 1 . y = E. and so E.I). Lemma: Let E c [0.a.• + + + + + + + + + + + + + + + . since m is translation invariant. 1). then addition modulo 1 corresponds to the addition of angles. Then is a commutative and associative operation taking pairs of numbers in [0. 1) the set E y is measurable and m(E y) = mE. b. Then E.36) has measure zero.

But Pi . 1) is either zero or infinite. Since each P.Ps = rj . But if x differs from an element in P by the rational number r. we must have i = j. 1) = 1. it should be noted that (until the last sentence) we have made no use of properties of Lebesgue measure other than translation invariance and countable additivity. we say that x and yare equivalent and write x '" y. 00 = t=l I:mP. P~ n PJ = >6. that is.64 Lebesgue Measure [Ch. This implies that if i ~ j. depending on whether mP is zero or positive. Let (ri)'::=O be an enumeration of the rational numbers in [0. whence Pi '" P» Since P has only one element from each equivalence class. . = pj + rj with Pi and PJ belonging to P. and consequently P cannot be measurable. is a translation modulo 1 of P. This is an equivalence relation and hence partitions [0. But if this were the case. each Pi will be measurable if P is and will have the same measure. 1) is in some equivalence class and so is equivalent to an element in P. If x . then m[O. 1) into equivalence classes. each real number x in [0. Theorem: If m is a countably additive. 00 and the right side is either zero or infinite. 3 We are now in a position to define a nonmeasurable set. then x E P~. 1). classes such that any two elements of one class differ by a rational number. The nonmeasurability of P with respect to any translation invariant countably additive measure m for which m[O. and define Pi = P rio Then Po = P. translation invariant measure defined on a a-algebra containing the set P.Thus U P~ = [0. is a rational number. But this is impossible since m[O. 1) = i=l I:ml'. 1) is 1 follows by contraposition.. While the above proof that P is not measurable is a proof by contradiction. 1) with r« = 0. + m[O.r. while any two elements of different classes differ by an irrational number. Hence the foregoing argument gives a direct proof of the following theorem: 17. Let x E Pi n PJ" Then x = Pi + r. By the axiom of choice there is a set P which contains exactly one element from each equivalence class. On the other hand. that is. that (Pi) is a pairwise disjoint sequence of sets.y is a rational number.

The measurability of E. . then mE = O. measurable. it is of great importance to know that sets which arise naturally in certain constructions are measurable. Now (i) =} set set set set {x: f(x) {x: f(x) {x: f(x) {x: f(x) is 2': a} is < a} is ::. For each real number a the iii. Similarly. a. For each real number a the These statements measurable. 5 Measurable Functions Since not all sets are measurable. (ii) =} of a sequence of measurable sets is measurable. [Hint: If A C (0.m* E. if A is any set with m* A > 0.] +- 16. Then (E. 2: a and the intersection Similarly. Show that if E is measurable and E C P. 1].) < 1. = mE. ::. m* s: < b. We have (i) =? (iv). = 0. Give an example of a sequence of sets (E'l with E. ::)Et+ (f). implies mE. (i).) is a disjoint sequence of measurable sets and mE.lin}. Proof: Let the domain of f be D. since {x: f(x) ::.] 17. Then the following statements are equivalent: i. Proposition: Let f be an extended real-valued function whose domain is measurable. 5] Measurable Functions 65 Problems 15.. 1). and m* (n E.) < L. = m U E.. Give an example where (E.. let E. Thus L.mE. since {x: f(x) > aJ = n=1 U 00 [x: f(x) + lin}..) is a disjoint sequence of sets and m*(U E.m*E. For each real number a the ii. Show that.. lim m* E. 2': m*A > 0. measurable. since {x: f(x) 2': a} = n=l n OC! {x: f(x) > a .m[O. imply real number a the set {x: f(x) v. For each real number a the iv. while L. = E r. measurable. (iv) =} (i) and (ii) ¢::} (iii).a} is > a} measurable.Sec. then there is a nonmeasurable set E CA. If we start with a function f the most important sets which arise from it are those listed in the following proposition: 18. = A n Pt. For each extended = a} is (ii).a} = D "" {x: f(x) > a} and the difference of two measurable sets is measurable. [Hint: Let E.

and fg are also measurable. the most important sets connected with them are measurable. Proposition: Let c be a constant and I and g two measurable real-valued functions defined on the same domain. Since the rationals are countable. and of course each step function is measurable. and so (ii) and (iv) ===} (v) for a real. If a is a real number. 3 and the union of a sequence of measurable sets is measurable. then I(x) < a . Proof: We shall use condition (iii) of Proposition 18. The following proposition tells us that certain operations performed on measurable functions lead again to measurable functions: 19. this set is measurable and so f +g . Iff is a measurable function and E is a measurable subset of the domain of I.g(x) and by the corollary to the axiom of Archimedes there is a rational number r such that f( x) < r < a- g( x). Similarly. f is said to be if it satisfies (Lebesgue) one of the Thus if we restrict ourselves to measurable functions. and so I + c is measurable when I is. If f(x) + g(x) < a. I ===} (v) for =- 00. cf. a (ii) ===} (v) for a = 00.r}).66 Lebesgue Measure [Ch. It should be noted that a continuous function (with a measurable domain) is measurable.f. Then {x:/(x) +c < a} = {x:f(x) < a . (iv) have (ii) & (iv) ===} (v). {x: f(x) = a} = {x: f(x) ~ a} n {x: f(x) ::. f + g. then the function obtained by restrictingfto E is also measurable. This shows that the first four statements are equivalent. and we Definition: An extended real-calued function measurable if its domain is measurable and first four statements of Proposition 18. a}. Hence [x: I(x) + g(x) < a} = U ({x: r f(x) < r} n (x: g(x) < a .c}. Since {x:f(x) = oo} = n=l n {x:f(x) 00 ~ n}. A similar argument shows cf to be measurable. g . Then the functions f + c.

n n Proof: If h is defined by hex) = sup {fl(X). implies that of h.). inf {fll .f2 . fg is always measurable and 1+ g is measurable if we always take the same value for f + g at points where it is undefined. Since -g = (-I)g is measurable when g is.=1 U 00 n {x: I~(x) > a}. then of the {x: hex) > a} = I. . . Similarly.Sec... Also. 6] Measurable Functions 67 f- is measurable. I We will often want to use Proposition 19 for extended real-valued functionsfand g. f + g is measurable no matter what values we take at the points where it is not defined.00. 20. . The function f2 is measurable.g2J is measurable. Hence the measurability {x: fn(x) > a}. Then the functions sup {/l. then . Unfortunately. measurable. sup fm inffm limfm and Hmfn are all measurable. However.p. A property is said to hold almost everywhere" (abbreviated a. we have g measurable.e..rs.. .e. since {X:f2(X) for a ~ ° > a} = {x:f(x) (X:f2(X) > v~} U {x:f(x) > a} = D < -Va} and if a < 0.. . if g is defined by g(x) = sup fn(x). and so g is measurable.In}. Theorem: Let <fn) be a sequence of measurable functions (with the same domain of definition). . and similarly n k [x: g(x) > a} = U >». Thus in particular we say thatf = g a. if I and g have the same domain 5 French: presque partout (p. See Problem 22.) if the set of points where it fails to hold is a set of measure zero.ln(x)}.. provided these points are a set of measure zero.j" g is not defined at points where it is of the form 00 . A n=l similar argument establishes the corresponding statements for info Since Hmfn = inf sup fb we have Iim j'. where D is the domain of f Thus fg = M(f + g)2 + .

then we may choose thefunctions g and h so that m :S g ~ M and m ~ h :S M. we can find a step function g and a continuous function h such that If - gl < E and If - hi < E except on a set of measure less than E.h(x)\ 2 c} < E. > a} E U {x e E: g(x) > a} E: g(x) ~ a}._. e. By hypothesis mE = 0. we define the characteristic function X A of the set A to be the function given by XA(x) = {0 I if x e A if x t A..68 Lebesgue Measure [Ch.e. and assume that f takes the values ± 00 only on a set of measure zero. m{x: If(x) . since / is a measurable function. b]. then Proof: Let E be the set {x: f(x) Now {x: g(x) > a} = {x:/(x) . Similarly. The last two sets on the right are measurable since they are subsets of E and mE = O. The first set on the right is measurable. x { ~ g(x)}.g(x) I 2 c} < E and m{x: If(x) . and so g is measurable. 22. . 3 and m{x: f(x) ~ g(x)} = O. The proof is left to the reader (cf. i.. Then given E > 0. Problem 23). we say that z. converges to g almost everywhere if there is a set E of measure zero such that fn(x) converges to g(x) for each x not in E. If A is any set. I The following proposition tells us that a measurable function is "almost" a continuous function. Proposition: Let f be a measurable function defined on an interval [a. If in addition m ~ f ~ M. Proposition: g is measurable. If f is a measurable function and f = g a.e. Thus {x: g(x) > a} is measurable for each a. is the following: 21. One consequence of equality a..

Show that XAnB XAUB = XA . . and such thatf assumes each value at most once. b.. that is. a set of real numbers such that every interval contains an element of D. a given nonmeasurable set. Let D be a dense set of real numbers. Thus the existence of a nonmeasurable set implies the existence of a nonmeasurable function. n where A.00 + 00. 19.. . Show that the sum and product of two simple functions are simple. and difference of two simple functions are simple.}. c.. Let f be an extended real-valued function on R such that {x: f(x) > a} is measurable for each a e D. then f + g is measurable if we define f + g to be a whenever it is of the form 00 . product. = {x: cp(x) = a. 20. Then f is measurable if and only if D 1 and D 2 are measurable and the restriction off to D . Let f be a function with measurable domain D. Prove that the product of two measurable extended real-valued functions is measurable. . If cp is simple and has the values al. D2 = {x:f(x) = -oo}. Show that f is measurable iff the function g defined by g(x) = f(x) for x e D and g(x) = 0 for x t D is measurable. Iff and g are measurable extended real-valued functions and a a fixed number. Show thatf is measurable if and only if its restrictions to D and E are measurable. a. A real-valued function cp is called simple if it is measurable and assumes only a finite number of values. (Dl U D2) is measurable. The sum.XA . 22. a. Let D and E be measurable sets and f a function with domain DUE. an then cp = c=L I:a. and let Dl = {x:f(x) = oo}.00 or . Let f be an extended real-valued function with measurable domain D.Sec._. XB = XA + XB XA· - XA • XB X_A = 1- 21. Thenfis measurable. 5] Measurable Functions 69 The function X A is measurable if and only if A is measurable. b. Problems 18. Show that (v) does not imply (iv) in Proposition 18by constructing a functionfsuch that {x: f(x) > O} = E.

Let/be measurable and B a Borel set. [Hint: The class of sets for whichf-I[E] is measurable is a rr-algebra. If m :::. 3 d. Show that there is a measurable set A such that g-I[A] is not measurable. M b. M. If / is Borel measurable. Show that there is a measurable set which is not a Borel set. b]. 2]. Iff is Borel measurable and g is Lebesgue measurable. Problem 2. then we may take SO so that m S SO ::. <p ~ M. / c. ~ M. . e. Given a measurable function I on (a.00. 23. Show thatf maps the Cantor set onto a set F of measure 1. then we may take h so that m ~ h :::. then /-I(E] is a Borel set. Given a step function g on [a. 1] onto [0. Given a simple function <pon [a. Prove Proposition 22 by establishing the following lemmas: a. .] If m ::. Verify that Propositions 18 and 19 and Theorem 20 remain valid if we replace "measurable set" by "Borel set" and "(Lebesgue) measurable" by "Borel measurable. Iff and g are Borel measurable. so is log. Show that if/is a meaurable real-valued function and g a continuous function defined on (. g::. and given E > 0. Borel measurability.00. Show that j" is a homeomorphism of [0. there is a step function g on [a. How much of the preceding problem can be carried out if we replace the class CB Borel sets by an arbitrary o-algebra a of sets? of 28. c. Let/be a measurable function on [a. b] which takes the values ± 00 only on a set of measure zero.70 Lebesgue Measure [Ch. Then/-I[B] is a measurable set. b] such that g(x) = <p(x) except on a set of measure less than E/3.. Let g = 1-1. a. [Hint: Use Proposition 15.. M < 24.. d.] 25. Let /1 be the Cantor ternary function (cf. Compare with Problems 25 and 26. except on a set of measure less than t/3..' M.SO(x)I < E except where I/(x)1 2:: M." Every Borel measurable function is Lebesgue measurable. b]. there is a simple function <p such that i/(x) . Thenj" + g is measurable no matter how it is defined at points where it has the form 00 . there is a continuous function h such that g(x) = hex) except on a set of measure less than E/3. 0 26. Given E > 0 and M.. If m g < M. A function / is said to be Borel measurable if for each a the set {x:f(x) > a} is a Borel set. then we can take g so that m s. b]. then log is Lebesgue measurable.46). b. (0). and define I by I(x) = 11(x) + x. d. Letj" and g be measurable extended real-valued functions which are finite almost everywhere. then g I is measurable. 27. there is an M such that III :::. Give an example of a continuous function g and a measurable function h such that hog is not measurable. and B is a Borel set.

by Proposition 14. 26.:::: N. 2 for some n 2 N}. and a third is given by Problems 4. every convergent sequence of [measurable] functions is nearly uniformly convergent. If one of the principles would be the obvious means to settle the problem if it were 'quite' true. roughly expressible in the following terms: Every (measurable) set is nearly a finite union of intervals. Thus EN = 0. Let f be a measurable real-valued function such that for each x in E we have fn(x) -7 f(x). every [measurable] function is nearly continuous. Ifn(x) . The following proposition gives one version of the third principle." "The extent of knowledge required is nothing like so great as is sometimes supposed. Littlewood says. and the student armed with them should be equal to most occasions when real variable theory is called for. another version by Problem 31.15 and 6. and <fn) a sequence of measurable functions defined on E. 1944. and for a problem that is actually solvable it generally is. lim mEN = O. p. 6] Littlewood's Three Principles 71 6 Littlewood's Three Principles Speaking of the theory of functions of a real variable. A slightly stronger form is given by Egoroff's theorem (Problem 30). Most of the results of [the theory] are fairly intuitive applications of these ideas. There are three principles.14. Then.f(x)1 Proof: Let Gn = {x e E: Ifn(x) .f(x) < E. 23. Proposition: Let E be a measurable set of finite measure.Sec. . Hence given 0 > 0. and for each x E E there must be some EN to which x does not belong. E. One version of the second principle is given by Proposition 22. there is a measurable set ACE with mA < 0 and an integer N such that for all x t A and all n . given E > 0 and 0 > 0." We have already met two of Littlewood's principles: Various forms of the first principle are given by Proposition 15. ':IN so that n 6 Lectures on the Theory of Functions.I(x)i E We have EN+1 C EN. since fn(x) -7 f(x). but you will generally find the weak form adequate. it is natural to ask if the 'nearly' is near enough.:::: E}. and so. Oxford. and set EN = n=N U rLJ Gn = {x e E: Ifn(x) . J. I .

29. b]. and j.72 LebesgueMeasure [Ch. that is. given E > 0 and 0 > 0. Hint: Let P.J 32. lin (x) Problems . there is a continuous function cpon [a. Can you do the same on the interval (. For 2-. 00 in Proposi- tions which converge to a real-valued function f a. 1] such that for each x we have limldx) = 0. and an N such that for all x t A and all n 2: N. be the sets in Section 4. as in the hypothesis of the proposition. oo)? [Hint: Use Egoroff's theorem. t if x e P.-1 ::. . Prove Egoroff's Theorem: If (In) is a sequence of measurable func- < 2. we have fn(x) ~ f(x) for each x. We have the following trivial modification of the last proposition: 24.+lt . If there is a subset B of E with mB = 0 such that fn ~ Ipointwise on E '" B. we say that j". there is a set ACE with mA < 0. we say that the sequence <fn) converges pointwise to f on E.l(x)1 2: € for some n 2: N} < o.(x) ~ {~ 1->0 30. m{x e E: Iln(x) . [Hint: Apply Proposition 24 repeatedly with En = lin and On = 2-nl1·] 31. then.I(x) I < € for all n 2: N}. that is. Give an example to show that we must require mE < tion 23. Then given 0 > 0. ~ fa. on E. construct a family Ut> of measurable real-valued functions on [0.e. Propositions 15 and 22.define It by t x = 2. Proposition: Let E be a measurable set of finite measure. If we write A for this EN' then m A < 0 and A = {x e E: I/nex) . and <fn>a sequence of measurable functions which converge to a real-valued function f a. 3 mEN < 0. and Problem 2. I If. b] such that m{x: I(x) ~ cp(x)} < o. Then.1 otherwise. on E. > o.e.39. Show that Proposition 23 need not be true if the integer variable n is replaced by a real variable t.e. on a measurable set E of finite measure. there is a subset ACE with mA < Y} such that In converges to I uniformly on E '" A.I(x) I < E. but for some 0 > Owe have m*{x:lt(x) > ~.00. given 71 > 0. Prove Lusin's Theorem: Letfbe a measurable real-valued function on an interval la.

. b]. inf ~'-I<X~~. f(x). b] and let a = ~o < ~ < . with the infimum taken over all possible subdivisions Similarly. Then for each subdivision we can define S= and S L t=l n at (~t - ~t_l)Mt = . We then define the upper Riemann integral of f by Rib f(x) dx = inf S of [a. . b]. f(x)._z)mt.4 The Lebesgue Integral 1 The Riemann Integral We recall a few definitions pertaining to the Riemann integral. .=1 L n ~. = sup ~'-I<X~~. m. Let f be a bounded real-valued function defined on the interval [a. < ~n = b 1 be a subdivision the sums of [a. = where M. we define the lower integral Rlb f(x)dx 73 = sups.

Under practically anybody's definition of an integral we have With this in mind we see that for all step functions 1/. f(x) = {1 and o x irrational x rational. 4 The upper integral is always at least as large as the lower integral.a b. which we shall consider later. bJ and some set of constants c. and if the two are equal we say thatfis Riemann integrable and call this common value the Riemann integral off.74 The Lebesgue Integral [Ch. Riemann integrable functions such that In increases monotonically to f What does this imply about changing the order of integration and the limiting process? . which has the form for some subdivision of [a. Similarly. Rib Problem 1.(x) ~ f(x). a. Show that if f(x) dx = sup ib cp(x) dx for all step functions cp(x) S. f(x). Construct a sequence Un> of nonnegative.. We shall denote it by Rib f(x) dx to distinguish it from the Lebesgue integral. then Rl -b f(x)dx = b. By a step function we mean a function".

However. 2] The Lebesgue Integral of a Bounded Function 7S 2 The Lebesgue Integral of a Bounded Function over a Set of Finite Measure The problem in the preceding section shows some of the shortcomings of the Riemann integral. A linear combination cp(x) = i=l L n aiXE/x) is called a simple function if the sets E. . -The function XE defined by xeE xzE is called the characteristic junction of E. we define i=l L atXA. are measurable. and the following lemma is useful: . distinct and nonzero. then where Ai = {x: cp(x) = aJ. This representation for cp is called the canonical representation and it is characterized by the fact that the Ai are disjoint and the a. This representation for cp is not unique. we define the integral of cp by when cp has the canonical representation cp = times abbreviate the expression for this integral to measurable set. n We some- J cpo If E is any It is often convenient to use representations which are not canonical. we would like a function which is 1on a measurable set and zero elsewhere to be integrable and have as its integral the measure of the set.. If cp is a simple function and {at. In particular.an} the set of nonzero values of cp.Sec. .. we note that a function cp is simple if and only if it is measurable and assumes only a finite number of values. If cp vanishes outside a set of finite measure.

1: a. a.) = a I cp + b I if. form a finite disjoint collection of measurable sets. Let Ao and Bo be the sets where tp and if. Then the sets Ek obtained by taking an the intersections Ai n B. To since the integral of a simple function which is greater than or equal to zero a. 4 1.76 The Lebesgue Integral [Ch. then Proof: Let {AJ and {BJ be the sets which occur in the canonical representations of cp and 1/. is a measurable set offinite Proof: The set Aa = {x: cp(x) = a} = U Ei• Hence a mli.e. n EJ = )25 for measure. i=l • n with E.. and we may write and so whence prove the second statement.. is nonnegative by the definition of the integral. Then and. I which vanish out- 2. I . = 2: aimEi by the additivity of m. we note that + bif. I (acp + bif.e. if cp :2: if. Lemma: Let cp = 1: aiXE.mli . are zero. + bbk)XE/c. Then i ~ j. and so f cp(x) dx = = 1: amn. be simple functions side a set of finite measure. follows from acp Lemma 1. = L (aa. Proposition: Let cp and 1/. Suppose each set E..

disjoint. be disjoint unnecessary.Sec.aimEi' and so the restriction of Lemma 1 that the sets E.. Thus L k=-n n mEk = mE. and have union E. E and sup "'~I E cp. The simple functions defined by and M CPn(x) = - n k=-n L n (k - l)XE. the numbers "'21 L. In order that Is'" inf f I/. 2] The Lebesgue Integral of a Bounded Function n 77 It follows from this proposition that if cP = ~ aiXE" . Proposition: Let f be defined and bounded on a measurable set E with mE finite. The answer is given by the following proposition: 3. Proof: Let f be bounded by M and suppose that f is measurable.(x) . Then the sets -n ~ k ~ n.=1 then f cp = IS Let f be a bounded real-valued function and E a measurable set of finite measure. are measurable. inf f f I/. f cp(x) dx E for all simple functions cp and 1/.. it is necessary and sufficient that f be measurable. and ask when these two numbers are equal. By analogy with the Riemann integral we consider for simple functions cp and I/.(x) dx = sup E /2".

. 4 satisfy ~n(X) ~ f(x) ~ fn(x). Thus inf and J r f(x) E dx ~ J r I{.78 The Lebesgue Integral [Ch.sup IE cp(x) dx = 0. Suppose now that if'?! inf f E f(x) dx = sup cpS! f E <p(x) dx. n Since n is arbitrary we have inf IE I{. and <p*(x) ::.n(X) dx E = M n k=-n ± kmEk whence o ::. I{.* (x). - E n k=-n L n M mEk = .mE. inf J I{.20. f(x) ~ I{.(x) dx .(X) dx - sup E J M ~(x) dx :::. and the condition is sufficient. n Then given n there are simple functions <Pn and and such that Then the functions and <p* = sup <Pn are measurable by Theorem 3.

*(x) < 1/. I Definition: If f is a bounded measurable function defined on a measurable set E with mE finite. we write Note that IEf = If' If for fE f XE• The following corollary to Proposition 3 shows that the Lebesgue integral is in fact a generalization Riemann integral. 2] The Lebesgue Integral of a Bounded Function 79 Now the set A = {x: ". Thus ". Proposition: Let f be a bounded function defined on [a. b].n(x) . and this latter set has measure less than vln.*(X) - H. nus.(x) dx for all simple functions l/. we write ( ( Ja f instead of i[a. ~ b f JE We sometimes write the integral as ( f If E = [a.. then it is measurable and If Proof: Since every step function is also a simple function we have R (b f(x) !a dx s sup i: cp(x) dx s: y.* except on a set of measure zero. b]. and so nus = 0. = 0.?J t" inf <p9 ia (b l/. I . and ".(x) < 1/. of the f 4. But each is contained in the set {x: ".. b].* = 1/. we define the (Lebesgue) integral of f over E by fE f(x) dx = inf fE 1/. is Riemann integrable on [a.21. Thus f is measurable by Proposition 3. and the condition is also necessary. Since f is Riemann integrable. and f is measurable by Proposition 3. Since n is arbitrary.b] f If f is a bounded measurable function which vanishes outside a set E of finite measure.*(x) < ¥-*(x)} is the union of the sets AI' ~p = {x: ".l/v}.* = f except on a set of measure zero.(x) dx s R Ja f(x) (b dx.Sec. the inequalities are all equalities.

aV' = a sup <p5. then lh 1/. 4 5. if. then: i. ii. Proposition: If f and g are bounded measurable functions defined on a set E of finite measure. and conversely (if a Hence for a > 0. = a fE f. then iii. is a simple function so is a1/..e.80 The Lebesgue Integral [Ch.! E using Proposition If 1/. then Hence Iff! iv. Hence + + IE f +g S IE (1/.! inf ( af = a inf JE if.. Iff = g a.1 is a simple function not less than f and 1/. 0). Iff s g a.2:. s fIfl· v.?! fE f = af E f. IE a f = <p~! inf fE 3.2) = IE 1/. If A s f(x) s B..z!. then AmE S IEfs BmE. ( af = JE If a < 0. we have .2. If A and B are disjoint measurable sets of finite measure.2 a simple function not less than g.2 is a simple function not less thanf g.e. if.2:. then Proof: If 1/.1 + 1/..1 + IE 1/. :.! fE V' = a inf h 1/. Since the infimum of the right side is f f + f g.

2] The Lebesgue Integral of a Bounded Function :::. This proof also serves to establish (iii).. Proposition (Bonnded Convergence Theorem): Let (/. 6.e.g = 0 a. we have = ({)1 ({)2' IE and statement (i) of the proposition follows. ~ 0 a. If f(x) = lim /. whence (ii)./x) for each x in E. Statement (iv) follows from (iii) and the fact that IE I = mE.e. 2:: this it follows that f- g.. From whence Similarly.: g imply ({)l + ({)2 is a simple func- IE f + g 2:: IE + ({)2) Since now the supremum of the right + IE side is J f + Jg. Hence «({)l f and ({)2 :::. if. ({)l tion not greater thanf + g. Statement (v) follows from (i) and the fact that XAUB = XA +X B• I We next prove a proposition which will be used to prove Theorem 15. and suppose that there is a real number M such that I!n(x)/ ::. The proposition is a special case of Theorem 15. it follows that if 'if. then IE f = lim IE r: .> be a sequence of measurable functions defined on a set E of finite measure.: 81 On the other hand.Sec. To prove (ii) it now suffices to show that Since f . M for all n and all x.

z!: o} . there is an N and a measurable set AcE with mA < f/4M such that for n 2: Nand x EO E . which states that." The conclusion of the proposition would be trivial if (In) converged to I uniformly.. A Thus Hence Problem 2. -b But there is a sequence 'Pn t h. :.49). b] is Riemann integrable if and only if the set of points at whichf is discontinuous has measure zero. 4 Proof: The proof of this proposition furnishes a nice illustration of the use of Littlewood's "three principles. given f > 0. b]. By Proposition 6 we have i ('Pn) of step functions such that b h = lim i b 'Pn 2: R i J.23. Then R function.. then i f = i h.f f E where h is a bounded measurable function such that m{x: hex) is finite. (If'P -b b 2: fis a step i 'P 2: h except at a finite number of points and so b h < R fa J. if (In) converges to I pointwise. we define I = sup h. Use part (a) to prove Lebesgue's Theorem: A bounded functionf on [a.. f E h5. Littlewood's third principle states that.82 The Lebesgue Integral [Ch. Let f be a bounded function on [a. a. Problem 2.I(x) I < f/2m£.. we have Iln(x) . and let h be the upper envelope off (cf. then (In) is "nearly" uniformly convergent to f A precise version of this principle is given by Proposition 3. 3 The Integral of a Nonnegative Function If I is a nonnegative measurable function defined on a measurable set E.) -b b.

e. Then we define the functions hand k by setting + hex) = min (f(x). let l be a bounded measurable function which vanishes outside a set of finite measure and which is not greater than f g.(x) and k(x) :::. then: 1.f(x) + g(x). we have k :::. fE f + g. Hence fE I= fE h + fE k < fEI + fE g. and so 8. g a. while hand k are bounded by the bound for I and vanish where I vanishes. We have hex) :::. fE I = fE f. and we prove only (ii) in detail.Sec. Proposition: If f and g are nonnegative measurable functions. and so fE h + fE Taking suprema. > o.hex). we have hex) + k(x) ::. I(x» and k(x) = I(x) . C C C iii. g(x). On the other hand. Theorem (Fatou's Lemma): If (fn) is a sequence of nonnegative measurable functions and fn(x) ~ f(x) almost everywhere on a set E.(x) f and k(x) :::. If hex) :::. then Proof: Parts (i) and (iii) follow directly from Proposition 5.. fE f + g = fE f + fE g. then . 3] The Integral of a Nonnegative Function 83 7. Iff::. 11. f g(x).

= lim E. Define a function h . Now hn(x) --? hex) for each x in E'. since integrals over sets of measure zero are zero. Then h. Proposition: Let f be a nonnegative function and (Et) a disjoint sequence of measurable sets. we get 9. 4 Proof: Without loss of generality we may assume that the con- vergence is everywhere.n. fn(x)}.. measurable 10. and let/ = lim/no Then Proof: By Theorem 8 we have ff lim Hence < lim f fn. Let h be a bounded measurable function which is not greater than/ and which vanishes outside a set E' of finite measure. Taking the supremum over h. --JE < lim f r fn. Monotone Convergence Theorem: Let <fn) be an increasing sequence of nonnegative measurable/unctions. and so f In ::. and let I= n=l L u. J f. Ifn s If But this implies But for each n we have I« < f.84 The Lebesgue Integral [Ch. 00 Then 11. Thus by Proposition 6 we have JE h r = JE h r. be a sequence of nonnegative functions. Corollary: Let u. is bounded by the bound for h and vanishes outside E'. Then . Let E = U E. by setting hn(x) = min {hex).

Then each In is bounded and In converges to I at each point. + NmA I < 2: + "2 = E .fN) E f. Set fn(x) = f(x) if f(x) :s. we have E/2. and so g is integrable.Sec. and [» < E/2. Proof: The proposition would be trivial if f were bounded. Choose 0 < 2~' If mA < ii i f= (f - IN) + i IN < (f . then g is also integrable on E. the terms on the right must also be finite. I and so the proposition Definition: A nonnegative measurable func tion f is called integrable over the measurable set E if 12. 3] The Integral of a Nonnegative Function 85 Proof: Let u. By the monotone convergence theorem there is an N such that if - biN > lEf 0. Thenj": XE = LU" follows from the preceding corollary. Proposition: Let f and g be two nonnegative measurable functions. Proposition: Let f be a nonnegative function which is integrable over a set E. I 13. nand fnCx) = n otherwise. Iff is integrable over E and g(x) < f(x) on E. Then given E > 0 there is a 0 > 0 such that for every set ACE with mA < 0 we have Lf< E. = f· XE. and Proof: By Proposition 7 Since the left side is finite.

00. with Lemma.] + 1. Then J 1= limJt. Let implies I 4. that is.e. 00) such that f. function.] 8. Let Un) be a sequence of nonnegative measurable converges to I..by measurable. Let I = be a nonnegative 0 a. b. Show that there is an increasing sequence ('Pn) of nonnegative simple functions each of which vanishes outside a set of finite measure such that I = lim 'Pn. Show that the function F F(x) 5. If I is . Show that the Monotone Convergence Theorem need not hold for decreasing sequence of functions.I· integrable function. a. Show that we may have strict inequality in Fatou's [Consider the sequence Un} defined by In (x) = 1 if n S x < n In(x) = 0 otherwise. 4 Problems 3.86 The Lebesgue Integral [Ch. We have 1- = (- f) V O.e. and suppose In S I for each n. 6. Prove the following generalization of Fatou's is a sequence of nonnegative functions. Let Un) be a sequence of nonnegative functions on ( . Show that fI = sup S'P over all simple functions 'P :::. ---> I a. a nonnegative measurable measurable function. then Lemma: If Un> Jlim In S lim fIn. Show that f 1= 0 I be a. 9. 4 The General Lebesgue Integral IV By the positive part 0. 1+ of a function I we mean the function 1+ = I+(x) = max {/(x). 7. so are/+ and 1-. we define the negative part 1. Then for each measur- f able set E we have JEIn ---> IEf. b. and suppose that fIn ---> I < 00. O}.In(x) = I for x ~ n. Let j be a nonnegative defined by is continuous = J~ooI functions which by using Theorem 9. [Let In(x) = 0 if x < n. Similarly.

if f and g are integrable. and I (f + g) = I ir: + g+) .J f. Hence f. The function cf is integrable over E.f ir + g-) = J r + f + .(f- + + + so are j+ + g+ and g-). Proposition: Let f and g be integrable over E. then JEf S. By Proposition 7.e.J r J I. we first note that if fl and h are nonnegative integrable functions with f = fl . + + and so J f = J r: . If A and B are disjoint measurable sets contained in E. then Proof: Part (i) follows directly from the definition of the integral and Proposition 7. Then: 1. g a.J «: g = f f+ fg· . To prove part (ii). 4] The General Lebesgue Integral 87 and jfl =T" +r. and IE cf = The function C IE f f+ g is integrable over E. With these notions in mind we make the following definition: Definition: A measurable function f is said to be integrable over E if f+ and f. In this case we define 14. . Iff IV. IE g. = But.f2' then f+ f2 = ffl.Sec. S.. (f g) = Cf+ g+) . 11.+ g-.are both integrable over E..J /2. and iii.

88 The Lebesgue Integral [Ch. 00 It should be noted that I + g is not defined at points where 15. we get and the theorem follows.00 and g = 00. Hence the integrability and the value of f (I + g) IS independent of the choice of values in these ambiguous cases. since f and g are integrable.In is nonnegative. f is integrable. and so by Theorem 8 Since III :::. the proof does not require quite so much. I 1= and g = . For (iv) we have LUB f = IXAUB ! = ! + flxB IXA = L f+ IB [. whence Similarly. the set of such points must have measure zero. 4 Part (iii) follows from part (ii) and the fact that the integral of a nonnegative integrable function is nonnegative. However. Then Proof: The function g .00 and where 1= . and we have g. considering g + fn. I The above theorem requires that the sequence <In) be dominated by a fixed integrable function g. If we replace the appropriate g's in the above proof . Lebesgue Convergence Theorem: Let g be integrable over E and let (In) be a sequence of measurable functions such that Ifni :s: g on E and for almost all x in E we have f(x) = limfn(x). However.

then Fatou's Lemma.lim fin. 4] The General Lebesgue Integral 89 by gn's. Let (fn) be a sequence of measurable functions such that Ifni:::. to an integrable function g.e. we get the following generalization of the Lebesgue Convergence Theorem: 16. the conclusion of Fatou's Lemma is weaker than that of the others: We can only assert ff :::. be bounded from above and below by fixed integrable functions and then asserts the equality of ff and lim ffn. Fatou's Lemma and the Monotone Convergence Theorem are very close in the sense that each can be derived from the other using only the fact that integration is positive and linear. If then If \ fn) is a sequence of measurable functions which converge a. S« and (fn) converges to fa. to J. the Monotone Convergence Theorem. iff is integrable. The Monotone Convergence Theorem (as generalized in Problem 6) is something of a hybrid: It requires that the z. and the Lebesgue Convergence Theorem all state that under suitable hypotheses we can assert something about ffin terms of ffn. Consequently. Show that Iff is integrable over E.e.e. The Lebesgue Convergence Theorem requires that j'.Sec. a. be bounded from below by zero (or an integrable function) and above by the limit function! itself. Theorem: Let (gn) be a sequence of integrable functions which converge a. Problems 10. Of course. Fatou's Lemma has the weakest hypothesis: We need only have /'1 bounded below by zero (or more generally by an integrable function). then so is If I and Does the integrability of If I imply that off? . but the advantage of Fatou's Lemma and the Monotone Convergence Theorem is that they are applicable even if f is not integrable and are often a good way of showing that f is integrable. this is a special case of the Lebesgue Convergence Theorem.

then . Show that I fn and I I have a meaning and that If S lim I In. on E. 14.e. Show that they are the same.gl < E. such that IE 11-1/.~ 17. ~ I a. 00). if is integrable. (0). Let I be integrable over E. withf integrable. Let h be an integrable function and (In> a sequence of measurable functions with In 2:: -h and limfn = f. Under the same hypothesis there is a continuous function g vanishing outside a finite interval such that IE 11. If cp is a simple function. integral of a function may exist without (in the sense of Lebesgue). (0). a. Establish the Riemann-Lebesgue Theorem: If f is an integrable function on (. show that the improper Lebesgue integral whenever the former 11.fnl ~ 0 if and only if Ilfnl ~ fill· 15. Then given E > 0. Icp. Let Un> be a sequence of integrable functions such thatj'.00. c. Use Problem 15.1 < [Combine part (a) with Proposition 3.00. that on 12. a. Let g be an integrable function on a set E and suppose Un) is a sequence of measurable functions such that Ifn(x)1 < g(x) a.g. 16. The improper Riemann the function being integrable I(x) = (sin x)/x on [0.22.90 The Lebesgue Integral [Ch. Then Ilf . If f Riemann integral is equal to the exists.] integrable over (. [Hint: The theorem is easy iff is a step function.e. we have two definitions for page 75 and that on page 87.] E. e. . there is a simple function cp such that [Apply Problem 4 to the positive and negative parts off. Under the same hypothesis there is a step function 'if. Letfbe J~oof(x) cos nx dx = O. 4 b. Then J f(x)dx = I f(x + t) dx. Then 13.] b..

Let g be a bounded measurable function.f(x + t)][ = O. t) which is defined in the square Q = {(x. . and suppose that for each fixed t the functionfis a measurable function of x.. t) is continuous in t for each x. t) e Q let the partial 19. given n ~ N we have € > is said to 0. where g is an integrable function on [0.. Let f be a function of two variables (x. t :::. This leads us to the following definition: Definition: A sequence (fn) of measurable functions converge to f in measure if. 5] Convergence in Measure 91 b. t): 0 S x :::. Then at d dt}o t at fex.: 1.. Suppose that af is bounded in Q.1.47(f) is useful here.Sec. t-40 [Problem 2. } and which is a measurable l function of x for each fixed value of t. Then lim If(x.I}. *5 Convergence in Measure Suppose that < fn> is a sequence of measurable functions such that flfnl -> O. be a function defined and bounded in the square Q = {(x. t): 0 ::.] Show also that if the function f(x.fn(x) e} < e. [Hint: Iff is continuous and vanishes outside a finite interval. Suppose that limf(x. t) = f(x) and 1-->0 that for all t we have [f(x. 0 :s. t) dx = Jr f(x) dx.] 18. g(x). 0 :S t :::.: x ::. this result follows from the uniform continuity off Apply Problem 15. t) dx is a continuous function of t.. Then !~1:00 [g(x)[f(x) . t)[ :::. Let f derivative af exist. t) dx = }o at t af dx. For each (x. there is an N such that for all I~ m{x: If(x) . 1]. then h(t) = 1[t». What can we say about the sequence (fn)? Perhaps the most important property of such a sequence is that for each positive rJ the measure of the sets [x: I fnex) I > rJ} must tend to zero.

fex)i 11 2:: 2.I(x) ----+ fex) I< p=k U ----+ 00 E we have p. 2-v for 2:: k.11. Problems 20. Deduce Proposition 2.}. however. there is an integer n.' is replaced by 'convergence in measure'. such that for all n 2:: n. and so InJx) I(x). v Then if x t .92 The Lebesgue Integral [Ch.andfn(x) = V P . = {x: Ifn/x) IlnJx) fnvex) 00 . Let E. Hence for any x eA = kQ vQ e. Prove that a sequence Un) of measurable functions converges tofin measure if and only if every subsequence of (fn) has in turn a subsequence which converges in measure to f.e. ] °° to zero in measure for any x in [0. Proposition: Fatou's Lemma and the Monotone and Lebesgue Convergence Theorems remain valid if 'convergence a. have the following proposition: 17. Then m{x: and so In ----+ in measure. 1] but such that (fnex) does not converge can be constructed as follows: Let n = k + 2". fnex) = 1 if x 8 [k2. 1] :5 k < 2".(k + 1)2. 21. 18 from Proposition 17 using Problems 20 and 22. ° lin (x)1 > e} :5 -. the sequence (fnex) has the value 1 for arbitrarily large values of n and so does not converge. although for any x e [0. 4 An example of a sequence (fn) which converges on [0. and set otherwise. then each subsequence (!nk) converges to f in measure. But mA k :5 m LQk EpJ :s p=k L mEp = 2-k+1• Hence mA = 0. 1].• 18. . n 2 Proof: Given we have 11. Proposition: Let (fn) be a sequence of measurable functions which converges in measure to f Then there is a subsequence (fnk) which converges to I almost everywhere. We do. Show that if Un) is a sequence which converges tofin measure.

You're Reading a Free Preview

Download
scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->