(MA4613&4623, MA6213&6223) Introduction To Real Analysis 2e PDF

Introduction to Real Analysis
(*, ) Manfred Stoll

English reprint copyright ®2004 by Pearson Education North Asia Limited and China Machine Press.
Original English Language title: Introduction to Real Analysis, Second Edition by Manfred Stoll
EISBN 0-321-04625-0
Copyright ®2001 by Addison Wesley Longman, Inc.
All right reserved.
Published by arrangement with the original publisher, Pearson Education. Inc. , publishing as Addison Wesley
Longman, Inc.
1 L EI1 Pearson Education (fff ( J` I) X aa° 11 t T_ R Lh ' Wt * jf .

For sale and distribution in the People's Republic of China exclusively ( except
Taiwan, Hong Kong SAR and Macao SAR)
09 Cx) 1r
Ifs 01-2004-2085
'M!FAVF PearsonEducation( * &*L$ SIM)&* Fjjjj#-No 3E *;Ia
® iii J112 (CIP) as

/i} 3(S2lib/(*)Wiilj(Stoll,Nil.
-Itht: V *2AthJW#.t, 2004.7
"ROW ffl# 4it3 t!ktt*ic
ISBN 7-111-14747-2
I .... aI. j1... II.
IV. 0174.1
q3W&*ffi4 CIPft*a*(2004) J059406'9,
22 -Ea WBittffili81100037)
WfIANIA: Y$ 3R
Iii #: I7 194,1:! £C
2004*7)1 Z 1 NJO I11M*

787mm x 1092mm 1/16 35.5 FPW 881 b=iz
2ft : 55.00 5
A.IIji443, IIU k S MX, RX, dl It T7 R1JJ

* * N 45 VL,q i& (010) 68993821, 88379646
Contents
PREFACE x
TO THE STUDENT xv
i. J The Real Number System

1.1 Sets and Operations on Sets 2
1.2 Functions 6
i
1.3 Mathematical Induction 15

1.4 The Least Upper Bound Property 20
1.5 Consequences of the Least upper Bound Property 28
1.6 Binary and Ternary Expansions 30
1.7 Countable and Uncountable Sets 34
Notes 43
Miscellaneous Exercises 44
Supplemental Reading 46
21 Sequences of Real Numbers

2.1 Convergent Sequences 48
47
2.2 Limit Theorems 53
2.3 Monotone Sequences 60
2.4 Subsequences and the Bolzano-Weierstrass Theorem 67
2.5 Limit Superior and Inferior of a Sequence 73
2.6 Cauchy Sequences 80
2.7 Series of Real Numbers 86
Notes 90
vi
Contents vii
3 1 Structure of Point Sets 93

3.1 Open and Closed Sets 93
3.2 Compact Sets 101
3.3 The Cantor Set 107
Notes 110
Q Limits and Continuity 115

4.1 Limit of a Function 116
4.2 Continuous Functions 130
4.3 Uniform Continuity 144
4.4 Monotone Functions and Discontinuities 148
Notes 162
5 j Differentiation 165
5.1 The Derivative 166
5.2 The Mean Value Theorem 176
5.3 L'Hospital's Rule 190
5.4 Newton's Method 197
Notes 203
The Riemann and 207

6 Riemann-Stieltjes Integral
6.1 The Riemann Integral 208
6.2 Properties of the Riemann Integral 223
6.3 Fundamental Theorem of Calculus 23A
6.4 Improper Riemann Integrals 239
6.5 The Riemann-Stieltjes Integral 245
6.6 Numerical Methods 260
6.7 Proof of Lebesgue's Theorem 272
Notes 276
7J Series of Real Numbers

7.1 Convergence Tests 280
279
7.2 The Dirichlet Test 294

viii Contents
7.3 Absolute and Conditional Convergence 299

7.4 Square Summable Sequences 306
Notes 313
$ Sequences and Series of Functions

8.1 Pointwise Convergence and Interchange of Limits 318
317
8.2 Uniform Convergence 323

8.3 Uniform Convergence and Continuity 330
8.4 Uniform Convergence and Integration 337
8.5 Uniform Convergence and Differentiation 339
8.6 The Weierstrass Approximation Theorem 346
8.7 Power Serves Expansions 353
8.8 The Gamma Function 372
Notes 377
j
..
q
r Orthogonal Functions and
Fourier Series
9.1 Orthogonal Functions 380
9.2 Completeness and Parseval's Equality 390
379
9.3 Trigonometric and Fourier Series 394

9.4 Convergence in the Mean of Fourier Series 404
9.5 Pointwise Convergence of Fourier Series 415
Notes 426
Lebesgue Measure and Integration 429

10.1 Introduction to Measure 430
10.2 Measure of Open Sets; Compact Sets 432
10.3 Inner and Outer Measure; Measurable Sets 444
10.4 Properties of Measurable Sets 449
10.5 Measurable Functions 455
10.6 The Lebesgue Integral of a Bounded Function 462
10.7 The General Lebesgue Integral 473
10.8 Square Integrable Functions 484
Notes 491
Contents ix
APPENDIX: Logic and Proofs 495

A.1 Propositions and Connectives 496
A.2 Rules of Inference 500
A.3 Mathematical Proofs 507
A.4 Use of Quantifiers 515
Bibliography 522
Hints and Solutions to Selected Exercises 523
Notation Index 543
Index 545
Preface
The subject of real analysis is one of the fundamental areas of mathematics, and is the
foundation for the study of many advanced topics, not only in mathematics, but also in
engineering and the physical sciences. A thorough understanding of the concepts of real
analysis has also become increasingly important for the study of advanced topics in
economics and the social sciences. Topics such as Fourier series, measure theory, and
integration are fundamental in mathematics and physics as well as engineering, eco-
nomics, and many other areas.
Due to the increased importance of real analysis in many diverse subject areas, the
typical first semester course on this subject has a varied student enrollment in terms of
both ability and motivation. From my own experience, the audience typically includes
mathematics majors, for whom this course represents the only rigorous treatment of
analysis in their collegiate career, and students who plan to pursue graduate study in
mathematics. In addition, there are mathematics education majors who need a strong
background in analysis in preparation for teaching high school calculus. Occasionally,
the enrollment includes graduate students in economics, engineering, physics, and other
areas, who need a thorough treatment of analysis in preparation for additional graduate
study either in mathematics or their own subject area. In an ideal situation, it would be
desirable to offer separate courses for each of these categories of students. Unfortu-
nately, staffing and enrollment usually make such choices impossible.
In the preparation of the text there were several goals I had in mind. The first was
to write a text suitable for a one-year sequence in real analysis at the junior or senior
level, providing a rigorous and comprehensive treatment of the theoretical concepts of
analysis. The topics chosen for inclusion are based on my experience in teaching grad-
uate courses in mathematics, and reflect what I feel are minimal requirements for suc-
cessful graduate study. I get to the least upper bound property as quickly as possible,
and emphasize this important property in the text. For this reason, the algebraic prop-
erties of the rational and real number systems are treated very informally, and the con-
struction of the real number system from the rational numbers is included only as a mis-
cellaneous exercise. I have attempted to keep the proofs as concise as possible, and to
x
Preface xi
let the subject matter progress in a natural manner. Topics or sections that are not specif-
ically required in subsequent chapters are indicated by a footnote.
My second goal was to make the text understandable to the typical student enrolled
in the course, taking into consideration the variations in abilities, background. and mo-
tivation. For this reason, Chapters 1 through 6 have been written with the intent to be
accessible to the average student, while at the same time challenging the more talented
student through the exercises. The basic topological concepts of open, closed, and com-
pact sets, as well as limits of sequences and functions, are introduced for the real num-
bers only. However, the proofs of many of the theorems, especially those involving
topological concepts, are presented in a manner that permits easy extensions to more
abstract settings. These chapters also include a large number of examples and more rou-
tine and computational exercises. Chapters 7 through 10 assume that the students have
achieved some level of expertise in the subject. In these chapters, function spaces are
introduced and studied in greater detail. The theorems, examples, and exercises require
greater sophistication and mathematical maturity for full understanding. From my own
experiences, these are not unrealistic expectations.
The book contains most of the standard topics one would expect to find in an
introductory text on real analysis-limits of sequences, limits of functions, conti-
nuity, differentiation, integration, series, sequences and series of functions, and
power series. These topics are basic to the study of real analysis and are included
in most texts at this level. In addition, I have included a number of topics that are
not always included in comparable texts. For instance, Chapter 6 contains a section
on the Riemann-Stieltjes integral, and a section on numerical methods. Chapter 7
also includes a section on square summable sequences and a brief introduction to
nonmed linear spaces. Both of these concepts appear again in later chapters of the
text.
In Chapter 8, to prove the Weierstrass approximation theorem, I use the method of
approximate identities. This exposes the student to a very important technique in analy-
sis that is used again in the chapter on Fourier series. The study of Fourier series, and
the representation of functions in terms of series of orthogonal functions, has become
increasingly important in many diverse areas. The inclusion of Fourier series in the text
allows the student to gain some exposure to this important subject, without the neces-
sity of taking a full semester course on partial differential equations. In the final chap-
ter I have also included a detailed treatment of Lebesgue measure and the Lebesgue in-
tegral. The approach to measure theory follows the original method of Lebesgue, using
inner and outer measure. This provides an intuitive and leisurely approach to this very
important topic.
The exercises at the end of each section are intended to reinforce the concepts of
the section and to help the students gain experience in developing their own proofs. Al-
though the text contains some routine and computational problems, many of the exer-
cises are designed to make the students think about the basic concepts of analysis, and
to challenge their creativity and logical thinking. Solutions and hints to selected exer-
cises are included at the end of the text. These. problems are marked by an asterisk (*).
At the end of each chapter I have also included a section of notes on the chapter,
miscellaneous exercises, and a supplemental reading list. The notes in many cases pro-
xii Preface
vide historical comments on the development of the subject, or discuss topics not in-
cluded in the chapter. The miscellaneous exercises are intended to extend the subject
matter of the text or to cover topics that, although important, are not covered in the
chapter itself. The supplemental reading list provides references to topics that relate to
the subject under discussion. Some of the references provide historical information;
others provide alternative solutions of results or interesting related problems. Most of
the articles appear ih the American Mathematical Monthly or Mathematics Magazine.
and should be easily accessible for students' reference.
To cover all the chapters in a one-year sequence is perhaps overly ambitious. How-
ever, from my own experience in teaching the course, with a judicious choice of topics
it is possible to cover most of the text in two semesters. A one-semester course should
at a minimum include all or most of the first five chapters, and part or all of Chapter 6
or Chapter 7. The latter chapter can be taught independently of Chapter 6; the only de-
pendence on Chapter 6 is the integral test, and this can be covered without a theoretical
treatment of Riemann integration. The remaining topics should be more than sufficient
for a full second semester. The only formal prerequisite for reading the text is a stan-
dard three- or four-semester sequence in calculus. Even though an occasional talented
student has completed one semester of this course during their sophomore year. some
mathematical maturity is expected, and the average student might be advised to take the
course during their junior or senior year.
Features New to the Second Edition

In content, the second edition remains primarily unchanged from the first. The subject
of real analysis has not changed significantly since publication of the first edition. In
this edition I have incorporated many of the valuable suggestions from reviewers, in-
structors, and students. Some new topics have been included, and the presentation of
others has been revised.
New examples and revised explanations appear throughout this edition of the text.
The second edition also contains additional illustrations and expanded problem sets.
The problem sets in all sections of the first six chapters have been expanded to include
more routine and computational problems. The challenging problems are still there.
With the addition of more routine problems, instructors using this text will have greater
flexibility in the assignment of exercises. The supplemental reading lists have all been
updated to include relevant articles that have appeared since 1996.
Two of the more substantive changes are the inclusion of a proof of Lebesgue's the-
orem in Chapter 6, and the addition of an appendix on logic and proofs. In the first edi-
tion, Lebesgue's theorem was stated in Section 6.1 and then proved in Chapter 10. At
the recommendation of my colleague Anton Schep, I have included a self-contained
proof of Lebesgue's theorem as a separate section in Chapter 6. The proof is based on
notes that he has used to supplement the text. In the proof, as in the statement of the
theorem, the only reference to measure theory is the definition of a set of measure zero.
With this change it is now possible not only to state but also to prove this important the-
orem without first having to develop the theory of Lebesgue measure and integration.
The greatest difficulty facing many students taking a course in real analysis is the
ability to write and to understand proofs. Most have never had a course in mathemati-
Preface xiii
cal logic. For this reason I have included a brief appendix on logic and proofs. The ap-
pendix is not intended to replace a formal course in logic; it is only intended to intro-
duce the rules of logic that students need to know in order to better understand proofs.
These rules are also crucial in helping students develop the ability to write their own
proofs. The various methods of proof are discussed in detail, and examples of each
method are included and analyzed. The appendix also includes a section on the use of
quantifiers. with special emphasis on the proper negation of quantified sentences. The
appendix itself is independent of the text: however, references to it are included
throughout the first several chapters of the text. The appendix can be included as part
of the course, or assigned as independent reading.
LL Acknowledgments
I would like to thank the students at South Carolina who have learned this material from
me, or my colleagues, from preliminary versions of this text. Your criticisms, com-
ments, and suggestions were appreciated. I am also indebted to those colleagues, espe-
cially the late Jeong Yang, who agreed to use the manuscript in their courses.
Special thanks are also due to the reviewers who examined the manuscript for the
first edition and provided constructive criticisms and suggestions for its improvement:
Joel Anderson, Pennsylvania State University; Bogdam Baishanski, Ohio State Univer-
sity; Robert Brown, University of Kansas; Donald Edmondson, University of Texas at
Austin; Kevin Grasse, University of Oklahoma; Harvey Greenwald, California Poly-
technic State University; Adam Helfer, University of Missouri, Columbia: Jan Kucera.
Washington State University; Thomas Reidel, University of Louisville; Joel Robin,
University of Wisconsin, Madison; Stuart Robinson, Cleveland State University; Dan
Shea, University of Wisconsin, Madison; Richard B. Sher, University of North Car-
olina: Thomas Smith, Manhattan College. Your careful reading of the manuscript
helped to turn the preliminary drafts into a polished text.
1 would also like to thank Carolyn Lee-Davis and the staff at Addison Wesley
Longman for their assistance in the preparation of the second edition, and the re-
viewers for this edition for their comments and recommendations: William Barnier,
Sonoma State University; Rene Barrientos, Miami Dade Community College; Denine
Burkett, Lock Haven University; Steve Deckelman, University of Wisconsin; Lyn
Geisler, Randolph-Macon College; Constant Goutziers, State University of New
York; Christopher Heil, Georgia Institute of Technology; William Stout. Salve Regina
University. Special thanks go to my colleague, George McNulty, for his careful read-
ing of the appendix. His constructive criticisms and suggestions were appreciated. I
am also grateful to the readers who informed me of errors in the first edition, and to
the instructors who conveyed to me some of the difficulties encountered while using
the book as a text. Hopefully all of the errors and shortcomings of the first edition
have been corrected.
Finally, I would especially like to thank my wife, Mary Lee, without whose en-
couragement this project might never have been completed.
Manfred Stoll
To the Student
The difference between a course on calculus and a course on real analysis is analogous
to the difference in the approach to the subject prior to the nineteenth century and since
that time. Most of the topics in calculus were developed in the late seventeenth and
eighteenth centuries by such prominent mathematicians as Newton, Leibniz, Bernoulli,
Euler, and many others. Newton and Leibniz developed the differential and integral cal-
culus; their successors extended and applied the theory to many problems in mathe-
matics and the physical sciences. They had phenomenal insight into the problems, and
were extremely proficient and ingenious in deriving complex formulas. What they
lacked, however, were the tools to place the subject on a rigorous mathematical foun-
dation. This did not occur until the nineteenth century with the contributions of Cauchy,
Bolzano, Weierstrass, Cantor, and many others.
In calculus, the emphasis is primarily on developing expertise in computational
techniques and applications. In real analysis, you will be expected to understand the
concepts and to develop the ability to prove results using the definitions and previ-
ous theorems. Understanding the concept of a limit, and proving results about limits,
will be significantly more important than computing limits. To accomplish this, it is
essential that all definitions and statements of theorems be learned precisely. Most of
the proofs of the theorems and solutions of the problems are logical consequences of
the definitions and previous results; some, however, do require ingenuity and cre-
ativity.
The text contains numerous examples and counterexamples to illustrate the partic-
ular topics under discussion. These are included to show why certain hypotheses are re-
quired, and to help develop a more thorough understanding of the subject. It is crucial
that you not only learn what is true, but that you also have sufficient counterexamples
at your disposal. I have included hints and answers to selected exercises at the end of
the text; these are indicated by an asterisk (*). For some of the problems I have provided
complete details; for others I have provided only brief hints, leaving the details to you.
As always, you are encouraged to first attempt the exercises, and to look at the hints or
solutions only after repeated attempts have been unsuccessful.
xc
To the Student xV
At the end of each chapter I have included a supplemental reading list. The journal
articles or books are all related to the topics in the chapter. Some provide historical in-
formation or extensions of the topics to more general settings; others provide alterna-
tive solutions of results in the text. or solutions of interesting related problems. All of
the articles should be accessible in your library. They are included to encourage you to
develop the habit of looking into the mathematical literature.
An excellent source for additional historical information and biographies of fa-
mous mathematicians is the MacTutor History of Mathematics archive at the University
of St. Andrews, Scotland. The URL of their webpage is http://www-history.mcs.st-
andrews.as.uk/
On reading the text you will inevitably encounter topics, formulas, or examples that
may appear too technical and difficult to comprehend. Skip them for the moment; there
will be plenty for you to understand in what follows. Upon later reading the section, you
may be surprised that it is not nearly as difficult as previously imagined. Concepts that
initially appear difficult become clearer once you develop a greater understanding of the
subject. It is important to keep in mind that many of the examples and topics that ap-
pear difficult to you were most likely just as difficult to the mathematicians of the era
in which they first appeared.
The material in the text is self-contained and independent of calculus. I do not use
any results from calculus in the definitions and development of the subject matter. Oc-
casionally, however, in the examples and exercises I do assume knowledge of the ele-
mentary functions and of notation and concepts that should have been encountered else-
where. These concepts will be defined carefully at the appropriate place in the text.
Manfred Stoll
The Real Number System
1.1 Sets and Operations on Sets
1.2 Functions
1.3 Mathematical Induction
1.4 The Least Upper Bound Property
1.5 Consequences of the Least Upper Bound Property
1.6 Binary and Ternary Expansions
1.7 Countable and Uncountable Sets
The key to understanding many of the fundamental concepts of calculus, such as limits,
continuity, and the integral, is the least upper bound property of the real number system
It As we all know, the rational number system contains gaps. For example. there does
not exist a rational number r such that r2 = 2, i.e., f is irrational. The fact that the
rational numbers do contain gaps makes them inadequate for any meaningful discus-
sion of the above concepts.
The standard argument used in proving that the equation r2 = 2 does not have a
solution in the rational numbers goes as follows: Suppose that there exists a rational
number r such that r2 = 2. Write r = m/n where in, it are integers that are not both
even. Thus m' = 2n2. Therefore in2 is even, and hence in itself must be even. But m2,
and hence also 2n2, are both divisible by 4. Therefore n2 is even, and as a consequence
n is also even. This, however, contradicts our assumption that not both in and n are even.
The method of proof used in this example is proof by contradiction; namely, we assume
the negation of the conclusion and arrive at a logical contradiction. A discussion of the
various methods of proof is included in Section A.3 of the Appendix.
The above argument shows that there does not exist a rational number r such that
r2 = 2. This argument was known to Pythagoras (around 500 B.C.), and even the Greek
mathematicians of this era noted that the straight line contains many more points than
the rational numbers. It was not until the nineteenth century, however, when mathe-
maticians became concerned with putting calculus on a firm mathematical footing, that
the development of the real number system was accomplished. The construction of the
real number system is attributed to Richard Dedekind (1831-1916) and Georg Cantor
0
2 Chapter I The Real Number System
(1845-1917), both of whom published their results independently in 1872. Dedekind's

aim was the construction of a number system. with the same completeness as the real
line, using only the basic postulates of the integers and the principles of set theory. In-
stead of constructing the real numbers, we will assume their existence and examine the
least upper bound property. As we will see, this property is the key to many basic facts
about the real numbers that are usually taken for granted in the study of calculus.
In Chapter 1 we will assume a basic understanding of the concept of a set and also
of both the rational and real number systems. In Section 1.4 we will briefly review the
algebraic and order properties of both the rational and real number systems and discuss
the least upper bound property. By example we will show that this property fails for the
rational numbers. In the subsequent two sections we will prove several elementary con-
sequences of the least upper bound property. In Section 1.7 we define the notion of a
countable set and consider some of the basic properties of countable sets. Among the
key results of this section are that the rational numbers are countable, whereas the real
numbers are not.
Sets and Operations on Sets

Sets are constantly encountered in mathematics. One speaks of sets of points, collec-
tions of real numbers, and families of functions. A set is conceived simply as a collec-
tion of definable objects. The words set, collection, and family are all synonymous.
The notation .r e A means that x is an element of the set A; the notation x e A means
that x is not an element of the set A. The set containing no elements is called the empty
set and will be denoted by 0.
A set can be described by listing its elements, usually within braces { }. For
example,
A = {-1,2.5.4}
describes the set consisting of the numbers -1, 2, 4, and 5. More generally, a set A may
be defined as the collection of all elements x in some larger collection satisfying a given
property. Thus the notation
A = {x : P(x)}
defines A to be the set of all objects x having the property P(x). This is usually read as
"A equals the set of all elements x such that P(x)." For example, if x ranges over all real
numbers, the set A defined by
A = {x:1 <x<5}
is the set of all real numbers that lie between 1 and 5. For this example, 3.75 E A.
whereas 5 E A. We will also use the notation A = {x E X : P(x)} to indicate that only
those x that are elements of X are being considered.
Some basic sets that we will encounter throughout the text are the following:
N = the set of natural numbers or positive Integers = {1. 2, 3, ...},
Z = the set of all integers = {. . ., -2, -1, 0, 1, 2... .},
1.1 Sets and Operations on Sets 3
O = the set of rational numbers = { p/q : p, q E Z, q # 0}. and

IlR = the set of real numbers.
In addition, we will occasionally encounter the set 10, 1, 2, 3, .. .} of nonnegative in-
tegers.
Real numbers that are not rational numbers are called irrational numbers. Since
many fractions can represent the same rational number, two rational numbers
r, = p,/q, and r, = p2/q2 are equal if and only if q, P2 = ptg2. Set-theoretically the ra-
tional numbers can be defined as sets of ordered pairs of integers (m, n), n * 0, where
two ordered pairs (p,, q,) and (p2, q2) are said to be equivalent (represent the same ra-
tional number) if ptq, = p2q,. We assume that the reader is familiar with the algebraic
operations of addition and multiplication of rational numbers.
A set A is a subset of a set B or is contained in B, denoted A C B. if every ele-
ment of A is an element of B. The set A is a proper subset of B, denoted A C# B if A
is a subset of B, but there is an element of B that is not an element of A. Two sets A and
B are equal, denoted A = B, if A C B and B C A. By definition, the empty set (6 is
a subset of every set.
Set Operations
There are a number of elementary operations that may be performed on sets. If A and
B are sets, the union of A and B, denoted A U B, is the set of all elements that belong
either to A or to B or to both A and B. Symbolically,
AUB={x:xEA or xEB}.
The intersection of A and B, denoted A fl B, is the set of elements that belong to both
A and B; that is,
Af1B={x:xEA and xEB}.
Two sets A and B are disjoint if A fl B = ¢. The relative complement of A in B, de-
noted B \A, is the set of all elements that are in B but not in A. In set notation,
B\ A= {x : x E B and x q-A}.
If the set A is a subset of some fixed set X. then X \ A is usually referred to as the com-
plement of A and is denoted by A`. These basic set operations are illustrated in
Figure 1.1 with the shaded areas representing A U B, A fl B, and B \ A, respectively.
AUB AnB BMA
Figure 1.1
4 Chapter 1 The Real Number System
There are several elementary set theoretic identities that will be encountered
throughout the text. We state some of these in the following theorem; others are given
in the exercises.
1.1.1 THEOREM If A. B, and C are sets, then

(a) An(BUC)=(AnB)u(Anc),
(b) AU(BnC)=(AUB)n(AUC),
(c) C \ (A U B) = (C \ A) n (C \ B),
(d) c \ (A n B) = (C \ A) U (C\ B).
The identities (a) and (b) are referred to as the distributive laws, whereas (c) and
(d) are De Morgan's laws. If A and B are subsets of a set X, then De Morgan's laws
can also be expressed as
(AUB)` =A`nB`, (A nB)` =AÙB`.
A more general version of both the distributive laws and De Morgan's laws will be
stated in Theorems 1.7.12 and 1.7.13.
Proof. We will provide the proof of (a) to illustrate the method used in proving these
results. The proofs of (b)-(d) are relegated to the exercises (Exercise 7). To prove the
identity in (a) we must prove that
An(BUC)C(AnB)U(AnC) (AnB)U(AnC)CAn(BUC).
and
Suppose xEAn(BUC).Then xEAand xEBUC.Since xEBUC,xEBor
xEC.IfxEB,then xEAnBandtherefore
x xE C x n (A n C). This then proves that
An(BUC) c (AnB)u(Anc).
To complete the proof, it still has to be shown that (A n B) U (A n C) C An
(B U C), thereby proving equality. If x E (A n B) U (An c). then by definition
xEAnBor xER An C,or both. But if xEAnB,then xEE Aand xEB.Since
xEBwealso have xEBUC.Therefore xEAn(BUC).Similarly, ifxEAnC,
then xEAn(BUC). Q
If A is any set, the set of all subsets of A is denoted by 91(A). The set 9'(A) is some-
times referred to as the power set of A. For example, if A = {I, 2}, then
91(A) = {4, {l}, (2), {1, 2}}.
In this example, the set A has 2 elements and 91(A) has 4 or 22 elements, the elements
in this instance being the subsets of A. If we take a set with 3 elements, then by listing
the subsets of A it is easily seen that there are exactly 23 subsets of A (Exercise 8). On
the basis of these two examples we are inclined to conjecture that if A contains n ele-
ments, then @'(A) contains 2" elements.
1.1 Sets and Operations on Sets S
We now prove that this is indeed the case. We form subsets B of A by deciding for
each element of A whether to include it in B or to leave it out. Thus for each element of
A there are exactly two possible choices. Since A has n elements, there are exactly 2"
possible decisions, each decision corresponding to a subset of A.
Finally, if A and B are two sets, the Cartesian product of A and B, denoted A X B.
is defined as the set of all ordered pairs' (a. b), where the first component a is from A
and the second component b is from B, i.e.,
A X B= {(a, b) : a E A, b E B}.
For example, if A = (1, 2} and B = {-1, 2, 4}, then
A X B = ((l, -1), (1, 2), (1, 4), (2, -l),(2,2),(2,4)).
The Cartesian product of R with R is usually denoted by R2 and is referred to as the eu-
clidean plane. If A and B are subsets of R, then A X B is a subset of R2. The case where
A and B are intervals is illustrated in Figure 1.2.
a
A
Figure 1.2 The Cartesian Product A x B
EXERCISES 1.1
1. LetA = {-1,0,1,2},B = {-2,3}, and C = (-2,0,1,5).
a. Find each of the following: (A U B), (B U C), (A n B), (B n C), (A n C). A n (B u C), A \ B. C \ B.
A\ (BUC).
1. A set theoretic definition of ordered pair can be given as follows: (a, b) - ((a), {a, b}}. With this
definition two ordered pairs (a, b) and (c, d) are equal if and only if a = c and b = d (Miscellaneous
Exercise 1).
b. Find each of the following: (A X B), (C x B), (A x B) n (C X B). (A n c) x B.

c. On the basis of your answer in (b), what might you conjecture about (A n c) X B for arbitrary sets A. B. C°
2. Let A={xER:-1<-x:5 5}.B={xE18:0Sx:53),C=(.rElB:2sx:5 4).
*a. Find each of the following: AnB.An7L,BnC,AUB,BUC.
*b. Find each of the following: A X B. A X C, (A X B) U (A X C).
c. Sketch the sets (A X B) U (A X C) and A X (B U C).
d. On the basis of your answer in (c), what might you conjecture about A X (B U C) for arbitrary sets A, B. C.'
3. If A and B are sets, prove that
a. Anm=d., AU46 =A. b. AfA=A. AUA=A.
c. AnB=BnA, AUB=BUA.
4. Prove the following associate laws for the set operations U and n:
*a. An(BnC)= (AnB)nc. b. AU(BUC) _ (AU8)UC.
5. If A C B. prove that
a. AnB=A. b. AUB=B.
6. If A is a subset of X. prove that
a. A U A` = X, b. A n A` = ¢,
c. (A`)` = A.
7. If A. B. and C are sets, prove that
*a. AU(Bnc) _ (AUB)n(AUC), b. C\(AUB) _ (C\A)n(c\B).
c. C\(A n B) = (C\A) U (C\B).
8. *Verify that a set with three elements has eight subsets.
9. If A and B are subsets of a set X. prove that A \ B = A n B.
10. If A and B are any sets, show that A n B and A \ B are disjoint and that
A = (AnB)U(A\B).
11. Prove that A X (B, U B,) = (A X B,) U (A X B,)-
12. Suppose A, C are subsets of X. and B, D are subsets of Y. Prove that
(AxB)n(CxD)=(Anc)x(BnD).
1.21 Functions
We begin this section with the fundamental concept of a function. In many texts, a func-
tion or a mapping f from a set A to a set B is described as a rule that assigns to each
element x E A a unique element y E B. This is generally expressed by writing y = f(x)
to denote the value of the function f at x. The difficulty with this "definition" is that the
terms "rule" and "assigns" are vague and difficult to define. Consequently we will de-
fine "function" strictly in terms of sets, using the notation and concepts introduced in
the preceding section.
The motivation for the following definition is to think of the graph of a function;
namely the set of ordered pairs (x, v) where .v is given by the "rule" that defines the func-
tion.
1.2 Functions 7
1.2.1 DEFINITION Let A and B be any two sets. A function f from A into B is a subset of
A X B with the property that each x E A is the first component of precisely one ordered
pair (x, y) E f ; that is, for every x E A there exists Y E B such that (x, y) E f, and if
(x, y) and (x, y') are elements of f, then y = y'. The set A is called the domain off, de-
noted Dom f The range of f, denoted Range f, is defined by
Range f= (yEB:(x,y)Ef forsomnexEA).
If Range f = B, then the function f is said to be onto B. (See Figure 1.3.)
AxB
1- - - -f- - (x.y) I
Figure 1.3 A Function as a Graph
If f is a function from A to B and (x, y) E f, then the element y is called the value
of the function f at x and we write
y = f (x) or f : x -+ y.
We also use the notation f : A -± B to indicate that f is a function from (or on) A into
(or to) B, or that f maps A to B or is a mapping of A to B. If we think of a function
f : A -> B as mapping an element x E A to an element y = f (x) in B, then this is often
represented by a diagram as in Figure 1.4. If f : A -* Q8, then f is said to be a real-
valued function on A.
A function f from A to B is not just any subset of A X B. The key phrase in Defini-
tion 1.2.1 is that "each x E A is the first component of precisely one ordered pair
(x, y) E f." To better understand the notion of function we consider several examples.
Figure 1.4 A Function as a Mapping
1.2.2 EXAMPLES
(a) Let A = 1-3, -2, -1, 0, 11 and B = Z. Consider the subset f of A X B given by
f = {(-3. 2). (-2. -2). (-1.4). (0. -6), (1.4)}.
Since each x E A belongs to precisely one ordered pair (x. y) E f. f is a function from
A into B with Range f = (-6. -2. 2.4). Figure 1.5 indicates what f does to each el-
ement of A. Even though the element 4 in B is the second component of the two dis-
tinct ordered pairs (- 1, 4) and (1, 4). this does not contradict the definition of function.
-3 2
-2 --2
0'--6
-1 4
1 4
Figure 1.5
(b) Let A and B be as in (a) and consider g defined by

g = ((-3,2).(-2.4).(-2.I).(-l.4),(0,5). (1, I)).
Since (-2.4) and (-2, 1) are two elements of g with the same first component, g is not
a function from A to B.
(c) In this example we let A = B = R. and let h be defined by
h=((x,v)ERXR:y=x'+2).
This function is described by the equation), = x2 + 2. The standard way of expressing
this function is as
h(x)_X''+2, Domh=R.
This specifies both the equation defining the function and the domain of the function.
For this example, Range It = {y E R : y ? 2}.'
2. Since x' z 0 for all x E R. the range of It is a subset of (v E R : ? 2). To obtain equality. we re-
y
quire that for every e > 2 there exists an .r E R such that x' + 2 = v. The existence of such a y will fol-
low as a consequence of Example 1.4.8.
1.2 Functions 9
(d) Let A be any nonempty set and let

i = {(.r, x) : x E Al.
Then i is a function from A onto A whose value at each x E A is x; i.e., i(r) = x. The
function i is called the identity function on A.
(e) Let A and B be two nonempty sets and consider the projection function p from
A X B to A defined by
p = {((a, b), a) : (a, b) E A X B).
In this example Dom p = A X B. Since p : (a, b) -* a, we denote this simply by

p(a, b) = a. For example, if A = 11, 2, 3) and B = { -1, 11, then A X B = {(1, -1).
(1, 1), (2, -1), (2,1), (3, -1), (3, 1)} and p(l, -1) = 1,p(1, 1) = l,p(2, -1) = 2.
etc.
As was indicated in (c), if our function h is given by an equation such as

y = x' + 2, we will simply write h(x) = x'' + 2, Dom h = Fl, to denote the function h.
It should be emphasized, however, that an equation such as h(x) = x2 + 2 by itself does
not define a function; the domain of h must also be specified. Thus
h(x) = x' + 2, Dom h = I8,
and
g(x)=x2+2, Domg={xEIB:-1 :5 x:5 2},

define two different functions.
Image and Inverse Image

1.2.3 DEFINITION Lei f be a function from A into B. If E C A, then f(E), the Image of
E under f, is defined by
f(E) = {f(x) : x E E}.

If H C B, the inverse Image of H, denoted f -'(H), is defined by
f-'(H)=(xEA:f(x)EH).
If H = {y}, we will write f -'(y) instead of f ''({y}). Thus for Y E B.
f"'(y) _ {x E A : f(x) = y}.

It is important to keep in mind that for E C A, f(E) denotes a subset of B, while
for H C B, f-' (H) describes a subset of the domain A. It should be clear that f(A) =
Range f, and that'f is onto B if and only if f(A) = B. To illustrate the notions of image
and inverse image of a set we consider the following examples.
1.2.4 EXAMPLES
(a) As in Example 1.2.2, let A = {-3, -2,-1.0, 1},B = 71, and f : A -* 2 the func-
tion given by
f = ((-3,2),(-2. -2),(-l,4).(0. -6),(1.4)1.
Consider the subset E _ {-1, 0, 1) of A. Then
f(E) {f(-1), f(0). f(l)} {-6,4}.
If H = (0, 1, 2, 3, 4}, then
f-'(H)= {xEA: f(x)EH}= {-3. -1,1}.
Since both f(- 1) and f (l) are equal to 4, f '(4) -1, 1). On the other hand, since
(x, 0) 0 f for any x E A, f -'(0) _ 46.
(b) Consider the function g : 7 -+ Z given by g (x) = x 2, and let E -1, -2,
-3,...). Then
g(E) = {(-n)2 : n E NI} _ {1,4,9,. ..}.
On the other hand,
g-'(g(E)) = 71\ {0}.
For this example, E C# g-'(g(E)).
(c) Let h be the function defined by h(x) = 2x + 3, Dom h = R. If E _ {x E R :
-1 < x s 2}, then
h(E) = {2x+3:-1 5 xs2} _ {yER: l y:5 7).
For the set E we also have
h-'(E) ={xER:2x+3EE}= {x:_2 s x s - 1}

2
For each y E R, x E h-' (y) if and only if 2x + 3 = y, which upon solving for x gives
x = 1(y - 3). Thus for each y E R. h-'(y) _ {2 (y - 3)}. Since h-'(y) # for each
y E R, the function h maps R onto R.
Suppose f is a function from A into B. If A, and A2 are subsets of A with A, C A2,
then from the definition of the image of a set it follows immediately that f(A1) C f(A2).
Likewise, if B, C B2 C B, then f -' (B,) C f -' (B2). The operations of finding the im-
age or inverse image of a set usually also preserve the basic set operations of union, in-
tersection, and complementation. There are two important exceptions, which are pre-
sented in part (b) of the next theorem and in Exercise 9.
1.2.5 THEOREM Let f be a function from A into B. If A, and A2 are subsets of A. then
(a) f(Ai U A2) = f(A,) Uf(A2)
(b) f(A, n A2) C f(A,) nf(A2)
1.2 Functions 11
Proof. To prove (a), let y be an element of f(A, U A,). Then Y = f(x) for some x in
A, U A2. Thus x E A, or x E A,. Suppose x E A,. Then y = f(x) E f(A,). Similarly, if
x E A2, Y E f (A2). Therefore y E f (A,) U f (A2). Thus
f(A, U A2) C f(A,) U f(A,).
Since it is clear that f(A,) and f(A2) are subsets of f(A, U A,), the reverse inclusion
also holds, thereby proving equality.
Since f(A, n A2) is a subset of both f(Al) and f(A2), the relation stated in (b) is
also true.
To. see that equality need not hold in (b), consider the function g(x) = x2
Dom g = 7L, of Example 1.2.4(b). If A, = (0, -1, -2, -3, ...), and A2 = (0. I, 2,
3, . . . } , then f(Al) = f(A2) _ {0, 1, 4, 9, . . .}, but A, n A, = {0}. Thus
f(A, fl A2) = f({0}) = {0} # f(A,) n f(A2) _ (0, 1, 4, 9, ...).
1.2.6 THEOREM Let f be a function from A to B. If B, and B2 are subsets of B, then

(a) f -'(B, U B2) = f -'(B,) U f-'(B2),
(b) f -'(B, n B2) = f -'(B,) n f-'(B2),
(c) f-'(B\B,) = A\f-'(B,).
Proof. To illustrate the method of proof we will prove (a), leaving the proofs of (b)
and (c) to the exercises (Exercise 8). Suppose x is an element of f -'(BI U B2). Then by
the definition of the inverse image of a set, f (x) E B, U B2. Thus f (x) E B, or
f(x) E B2. If f(x) E B,, then x is in f '(81), and thus x E f -' (B,) U f (B,). The
same argument also applies if f(x) E B2. In either case we have f -'(BI U B2)
C f-(B1)Uf-'(B2)
To prove the reverse containment, we first note that both B, and B2 are subsets of
B, U B2. Therefore both f -' (B,) and f - (B2) are subsets of f `(BI U B2), and thus
f -' (B,) U f -' (B2) C f (B, U B2). This proves the reverse containment thereby
giving equality.
Inverse Function
1.2.7 DEFINITION A function f from A into B is said to be one-to-one if whenever
x, * x2, then f(x,) * f(x2).
Alternatively, a function f is one-to-one if whenever (x,, y) and (x2, y) are elements
off then x, = x2. From the definition it follows that f is one-to-one if and only if
f `(y) consists of at most one element of A for every y E B. If f is onto B. then
f -'(y) # rb for every y E B. Thus if f is one-to-one and onto B, then f `(y) consists
of exactly one element x E A and
g={(y,x)EBXA: f(x)=y}
defines a function from B to A. This leads to the following definition.
1.2.8 DEFINITION If f is a one-to-one function from A onto B. let

f -' = {(y, x) E B X A: f(x) = y}.
The function f from B onto A is called the inverse function of f. Furthermore. for
each y E B.
x = f -'(y) if and only if f(x) = y.
There is a subtle point that needs to be clarified. If f is any function from A to B.
then f''(y) (technically f -'({y})) is defined for any y E B as the set of points x in A
such that f(x) = y. However, if f is a one-to-one function of A onto B, then f -'(y) de-
notes the value of the inverse function f -' at y E B. Thus it makes sense to write
f -'(y) = x whenever (y. x) E f '. Also, if f is a one-to-one function of A into B. then
f'' defined by
f -' = {(y, x) : y E Range f and f(x) = y}
is a function from Range f onto A. (See Figure 1.6.)
Figure 1.6 The Inverse Function
1.2.9 EXAMPLES
(a) Let h be the function of Example 1.2.4(c); that is, h(x) = 2x + 3, Dom h = R. The
function h is clearly one-to-one and onto R with
x=h-'(v)=2(y-3). Domh-' - R.
(b) Consider the function f defined by the equation y = 2. If we take for the domain
of f all of R, then f is not a one-to-one function. However, if we let
Domf=A={xER:xa0},
then f becomes a one-to-one mapping of A into A. To see that f is one-to-one, let
x,, x2 E A with x, # x2. Suppose x, < x2. Then x1 < x2; that is, f(x1) # f(x2). Tbero-
fore f is one-to-one. To show that f is onto A. we need to show that for each y E A.
y > 0, there exists a positive real number x such that
x2=y.
1.2 Functions 13
Intuitively we know that such an x exists; namely, the square root of y. However.
a rigorous proof of the existence of such an x will require the least upper bound
property of the real numbers. In Example 1.4.8 we will prove that for each y > 0
there exists a unique positive real number x such that X' = y. The number x is called
the square root of y and is denoted by \. Thus the inverse function of f is
given by
f-'(y) _ Domf ' _ {y E R:y :0}.
Composition of Functions
Suppose f is a function from A to B and g is a function from B to C. If a E A. then f(a)
is an element of B, the domain of g. Consequently we can apply the function g to f(a)
to obtain the element g(f(a)) in C. This process, illustrated in Figure 1.7. gives a new
function h which maps a E A to g(f(a)) in C.
h(a) =g(f(a)) C
Figure 1.7 Composition of g with f
1.2.10 DEFINITION If f is a function from A to B and g is a function from B to C, then the

function g of: A -i C. defined by
gof= {(x.z)EA X C:z=g(f(x))}
is called the composition of g with f
If f is a one-to-one function from A into B, then it can be shown that
(f o f) (x) = x for all x E A and that (f o f ') (y) = y for ally E Range f (Exercise
10). This is illustrated in (b) of the following example.
1.2.11 EXAMPLES
(a) If f (x) = v -+x with Dom f = {x E R : x z - I } and g(x) = x', Dom g = R.
then
(g ofXx) = g(f(x)) = ( )'' = I + x,

Dom(gof)= {x ER:xz -1}.
Even though the equation (g of)(x) = 1 + x is defined for all real numbers x, the do-
main of the composite function g o f is still only the set {x E R : x,2: -1 }. For this ex-
ample, since Range g C Dom f, we can also find f o g; namely,
(f o g)(x) =.f(g(x)) = 1 + x2, Dom f o g = R.

(b) For the function fin (a), the inverse function f is given by
f-'(y)=y2- 1, Dom f'' = Range f = {y E R :Y z 0}.

Thus for x E Dom f,
(f f)(x) = f _' U W) = (f (x))2 - I = (\/j)2 - 1 = X.

and for y > 0,
(fof-')(Y) =f(f+I(Y)) = ( = Y.
R EXERCISES 1.2
1. Let A = {-1, 0, 1, 2) and B = N. Which of the following subsets of A X B is a function from A into B? Explain
your answer.
a. f= {(-l.2).(0.3).(2.5)} b. g = {(-1,2),(0,7).(1,3),(2,7)}
c. h = -1)} d. k = {(x, y) : y = 2r + 3, x E A}
2. *a. Let A = {(x, y) E R X R : x2 + y2 = 11. Is A a function? Explain your answer.
b. Let B = {(x, y) E A : y z 0}. Is B a function? Explain your answer.
3. Let f : N -* N be the function defined by f(n) = 2n - 1. Find f(E) and f -'(E) for each of the following sub-
sets E of N.
*a. {1,2,3,4} b. {1,3,5.7} c. N
4 . Let f= {(x,y):xE R,y =x3 + 1}.
*a. Let A = {x : -1 <- x 5 2}. Find f(A) and f '(A).
b. Show that f is a one-to-one function of R onto R.
*c. Find the inverse function f - 1.
S. Let f, g mapping Z into Z be given by f(x) = x + 3 and g(x) = 2x. Find (f o g)(N) and (g of)(N).
6. For each of the following real-valued functions, find the range of the function f and determine whether f is one-
to-one. If f is one-to-one, find the inverse function and specify the domain of f-'.
*a. f(x)3x-2, Domf=R b. f(x)=5x+4. Domf=R
*c. 1(x) = 1 1
s Dom f= {x E R : O s x < l} d. Ax) = I X X Dom f= {x E R : 0 < x < 1)
e. Ax) _ + Dom f= {x E R : -1 5 x s 11 1. Ax) = sin x, Dom f = {x E R : 0!5 x :s r)

I,
*g. f(x. y) = x. Dom f = R2

7. Let A = {t E R : 0:5 t < 21r} and B = R2. and let f : A -+ B be defined by J(t) = (co. r. sin t).
*a. What is the range of f?
*b. Find f ' of each of the following points: (1, 0), (0, -1), ;-, ), (0, 1).
c. Is the function f one-to-one?
8. Prove parts (b) and (c) of Theorem 1.2.6.
9. Letf:A -+ BandletFCA.
a. Prove that f(A) \f(F) C f(A \ F). b. Give an example for which f(A) \f(F) * f(A \ F).
10. Let f be a one-to-one function from A into B. Show that' o(f f)(x) = x for all x E A and that (f of
for ally E Range f.
11. *Let f : A -* B and g : B -> A be functions satisfying (g o f)(x) = x for all x E A. Show that f is a one-to-one
function. Must f be onto B?
12. If f : A -s B and g : B -+ C are one-to-one functions, show that (g of) f -' o g' on Range (g of).
1.3 Mathematical Induction

Throughout the text we will on occasion need to prove a statement, identity, or in-
equality involving the positive integer n. As an example, consider the following iden-
tity. For each n E N,
r"+'
r+r2+ +r"= r -
Mathematical induction is a very useful tool in establishing that such an identity is valid
for all positive integers n.
1.3.1 THEOREM (Principle of Mathematical Induction) For each n E N, let P(n) be a

statement about the positive integer n. If
(a) P(1) is true, and
(b) P(k + 1) is true whenever P(k) is true,
then P(n) is true for all n E N.
The proof of this theorem depends on the fact that the positive integers are well-
ordered; namely, every nonempty subset of N has a smallest element. This statement is
usually taken as a postulate or axiom for the positive integers; we do so in this text. Since
it will be used on several other occasions, we state it both for completeness and emphasis.
1.3.2 WELL-ORDERING PRINCIPLE Every nonempty subset of N has a smallest element.

The well-ordering principle can be restated as follows: If A C F%, A # d,, then
there exists n E A such that n :5 k for all k E A.
To prove Theorem 1.3.1 we will use the method of proof by contradiction. Most
theorems involve showing that the statement P implies the statement Q; namely, if P is
true, then Q is true. In a proof by contradiction one assumes that P is true and Q is false.
and then shows that these two assumptions lead to a logical contradiction; namely, show
that some statement R is both true and false. Further details on the method of proof by
contradiction are provided in Section A.3 of the Appendix.
Proof of Theorem 1.3.1. Assume that the hypotheses of Theorem 1.3.1 are true,
but that the conclusion is false; that is, there exists a positive integer n such that the
statement P(n) is false. Let
A = {k E N : P(k) is false}.
By our assumption, the set A is nonempty. Thus by the well-ordering principle A has a
smallest element k,,. Since P(1) is true, k > 1. Also, since ko is the smallest element of
A. P(k,, - 1) is true. But then by hypothesis (b), P(k,,) is also true. Thus k, a A. This,
however, is a contradiction. Consequently, P(n) must be true for all n E N. Q
1.3.3 EXAMPLES We now provide two examples to illustrate the method of proof by math-
ematical induction. The first example provides a proof of the identity in the introduc-
tion to the section. An alternative method of proof will be requested in the exercises
(Exercise 7).
(a) To use mathematical induction, we let our statement P(n), n E N, be as follows:
r+ +r"= r - r"+i . r # 1.
I -r
When n = I we have
r(1 -r) _ r-r2
r = (1 - r) I - r , provided r * 1.
Thus the identity is valid for n = 1. Assume P(k) is true for k i' I; i.e.,
r - rA+I
r+ +r I-r r # 1.
We must now show that the statement P(k + 1) is true; that is.
r+ r- r(A+ p+ i
r # 1.
I -r '
But
r+ + rk+i = r + + rk + rk+1,
which by the induction hypothesis

r - rk+i r - rA+i + (1 - r)rk+l
I-r I - r
r - rk+'
1-r
1.3 Mathematical induction 17
Thus the identity is valid for k + 1, and hence by the principle of mathematical induc-
tion for all n E N.
(b) For our second example, we use mathematical induction to prove Bernoulli's in-
equality. If h > -1, then
(I +h)">1 + nh for alln EN.
When n = 1, (1 + h)' = I + h. Thus since equality holds, the inequality is certainly
valid. Assume that the inequality is true when n = k, k -- 1. Then for is = k + 1,
(I + h)'-I = (I + h)'"(1 + h),
which by the induction hypothesis and the fact that (1 + h) > 0
?(1 +kh)(1 +h)= I +(k+ 1)h+kh=
1 + (k + 1)h.
Therefore the inequality holds for n = k + 1, and thus by the principle of mathemati-
cal induction for all n E N.
. Although the statement of Theorem 1.3.1 starts with n = 1, the result is still true if
we start with any integer n,, E Z. The modified principle of mathematical induction is
as follows: If for each n E 77, n ? n,,, P(n) is a statement about the integer n satisfying
(a) is true, and
(b') P(k + 1) is true whenever P(k) is true. k ? no,
then P(n) is true for all n E 77, n ? n,.
The proof of this follows from Theorem 1.3.1 by simply setting
Q(n)=P(n+n,- I), n E N,
which is now a statement about the positive integer n.
Remark. In the principle of mathematical induction, the hypothesis that P(I) be true
is essential. For example, consider the statement P(n):
n+i=n, nEN.
This is clearly false! However, if we assume that P(k) is true, then we also obtain that
P(k + 1) is true. Thus it is absolutely essential that P(n,) be true for at least one fixed
value of n,.
There is a second version of the principle of mathematical induction that is also

quite useful.
1.3.4 THEOREM (Second Principle of Mathematical Induction) For each n E N. let

P(n) be a statement about the positive integer n. If
(a) P(1) is true, and
(b) for k > 1, P(k) is true whenever P(j) is true for all positive integers j < k,
then P(n) is true for all n E N.
18 chapter 1 The Real Number System
Proof. Exercise 3. Q
Mathematical induction is also used in the recursive definition of functions defined
for the positive integers. In this procedure, we give an initial value of the function fat
n = 1. then assuming that f has been defined for all integers k = 1. .... n, the value
off at n + I is given in terms of the values off at k, k 5 n. This is illustrated by the
following examples.
1.3.5 EXAMPLES
(a) Suppose f : N -' N is defined by f(1) = t and f(n + 1) = nf(n). n E N. The
values off for n = 1, 2, 3.4 are given as follows:
3.2.1.
f(1) = 1, f(2) = lf(1) = 1, f(3) = 2f(2) = 2.1, f(4) = 3f(3) =
Thus we conjecture that f(n) (n - 1)!, where 0. is defined to be equal to one, and
for n E N, n! (read n factorial) is defined as
The conjecture is certainly true when n = 1. Thus assume that it is true for n = k,
k 1; that is, f(k) = (k - 1)!. Then for n = k + 1,
f(k + 1) = kf(k),
which by the induction hypothesis
k(k - 1)! = k!
Therefore the identity holds for it = (k + 1), and thus by the principle of mathematical
induction, for all it E N.
(b) For our second example, consider the function f : N - R defined by f (l) = 0,
f(2) = 3, and for n ? 2 by f (n) = (1, +i) f (n - 2). Computing the values of f for
n = 3, 4, 5, and 6, we have
f(3) = 0, f(4) = 5, f(5) = 0. f(6) =
From these values we conjecture that

0, if n is odd,
f(n) = 1
if it is even.
tn
To prove our conjecture we will use the second principle of mathematical induction.
Our conjecture is certainly true for it = 1, 2. Suppose n > 2, and suppose our conjec-
ture holds for all k < n. If n is odd, then so is (n - 2), and thus by the induction
hypothesis f(n - 2) = 0. Therefore f(n) = 0. On the other hand, if n is even, so is
(n - 2). Thus by the induction hypothesis f(n - 2) = ;,?-y. Therefore
n
f(n)
-(n1) n-2) -(n-11 1 = I .
n+l f( n+1)n-1 n+l IN

EXERCISES 1.3
1. Use mathematical induction to prove that each of the following identities are valid for all n E N.
n(n + 1)
a. 1+2+3+ +n= 2 *b. 1 +3+5+ +(2n- 1)=n2
n(n + 1)(2n + 1)
. 12+22+ +n2= d. 13+23+ +n3=[Zn(n+l)]2
6
e. 2+22+23+ +2"=2(2"- 1)
*f For x, y E Il,
x"+1 -y",i =(x-y)(x"+x"-iy+...+y")
g' 1(2)+2(3)+ + n(n+1) n+1

n
2. Use mathematical induction to establish the following inequalities.
*a. 2" > n for all n E N b. 2" > n2 for all n E N. n ? 5
*c. n! > 2" f o r all n E !N, n L - 4 d. 13 + 23 + + n3 < n4 for all n E 1N, n 2
e. 13 + 23 + + n3 < £n4 for all n E !N, n >- 3
3. Prove Theorem 1.3.4.
4. *Let f : N - N be defined by f (l) = 5, f(2) = 13, and for n z 2, f(n) = 2f(n - 2) + f(n - 1). Prove that
f(n)=3.2"+(-1)"foralln E Ni.
5. For each of the following functions f with domain N, determine a formula for f(n) and use mathematical induction
to prove your conclusion.
a. f(1)=2,andforn> 1,f(n)=(n-1)f(n-1)-n+1.
*b. f(l)=1,1(2)=4, andforn>2,f(n)=2f(n-1)-f(n-2)+2.
(n
c. f (l) = 1, and for n > 1, f (n) = 3n 1) f (n - 1).
* d f(1) = I f(2) = 0 and

'
for n > 2 f(n) _ - f(n - 2)
n(n-1)*
An - 2)
C. For a,, a2 E R arbitrary, l et f ( l ) = a,, f(2 ) = a2, and for n > 2 , f (n)
n(n - 1)'
f. For a 1, a2 E R arbitrary, let f (l) = a,, f (2) = a2, and for n > 2,f(n) _ (__4)f(n - 2).
6. Let f: N! -*NI be defined by f(1) = 1, f(2) = 2, and
f(n + 2) =(n + 1) + f(n)).

Use Theorem 1.3.4 to prove that 1 s f(n)s 2 for all n e N.
7. *Prove that
r+r2+ +r"= r - r 1
, r# nEN
without using mathematical induction.
20 Chapter) The Real Number System
8. Use mathematical induction to prove the arithmetic geometric mean inequality. If a,, a,.. . . . a,,, n E N. are
nonnegative real numbers, then
a, + a2 + ' - - +a.
ala, , .
-a. S (
with equality if and only if a, = a, = = a".
1.4 The Least Upper Bound Property

In this section we will consider the concept of the least upper bound of a set and intro-
duce the least upper bound or supremum property of the real numbers R. Prior to in-
troducing these new ideas we briefly review the algebraic and order properties of 0
and R.
Both the rational numbers 0 and the real numbers R are algebraic systems known
as fields. The key fact about a field is that it is a set F with two operations, addition (+)
and multiplication (), that satisfy the following axioms:
1. If a,bE F.then a+bE
2. The operations are commutative; that is, for all a, b E F
a+b=b+a and
3. The operations are associative; that is. for all a, b, c E F
a + (b + c) _ (a + b) + c and a (b c) = (a b) c.
4. There exists an element 0 E F such that a + 0 = a for every a E F.
5. Every a E F has an additive inverse; that is, there exists an element -a in F
such that
a+(-a)=0.
6. There exists an element 1 E F with 1 # 0 such that a 1 = a for all a E F.
7. Every a E F with a * 0 has a multiplicative inverse; that is, there exists an
element a' in F such that
1.
8. The operation of multiplication is distributive over addition; that is, for all
a, b, c E F,
The element 0 is called the zero of F and the element I is called the unit of F. For
a # 0, the element a-' is customarily written as or 1/a. Similarly, we write a - b in-
stead of a + (-b), ab instead of a b, and alb or instead of a b-.
The real numbers R contain a subset P known as the positive real numbers satis-
fying the following:
(01) If a, b E P, then a+ b E P and a -b E P.

(02) If a E R then one and only one of the following hold:
aEP, -aEP, a =0.
Properties (01) and (02) are called the order properties of R. Any field F with a non-
empty subset satisfying (01) and (02) is called an ordered field. For the real numbers
we assume the existence of a positive set P. For the rational numbers 0, the set of pos-
itive rational numbers is given by P n 0, which can be proved to be equal to
{p/q:p,gE7L,4 *O,pgE1i}.
Let a, b be elements of R. If a - b is positive, i.e., a - b E P. then we write
a > b or b < a. In particular, the notation a > 0 (or 0 < a) means that a is a positive
element. Also, a s b (or b ? a) if a b, then a + c > b + c.
(b) If a > b and c > 0, then ac > be.
(c) If a > b and c < 0, then ac < be.
(d) If a * 0, then a2 > 0.
(e) If a > 0, then 1/a > 0; if a < 0, then 1/a < 0.
To illustrate the method of proof, we provide the proof of (b). Suppose a > b; i.e.,
a - b is positive. If c is positive, then by (01), (a - b)c is positive. By the distributive law,
(a - b)c = ac - be.
Therefore ac - be is positive; that is, ac > be. The proofs of the other results are left
as exercises.
Upper Bound of a Set

We now turn our attention to the most important topic of this chapter; namely, the least
upper bound or supremum property of R. In Example 1.4.5(c) we will show that this
property fails for the rational numbers Q. First, however, we define the concept of an
upper bound of a set.
1A.1 DEFINITION A subset E of R is bounded above if there exists,0 E R such that

x 5 /3 for every x E E. Such a (3 is called an upper bound of E.
The concepts bounded below and lower bound are defined similarly. A set E is
bounded if E is bounded both above and below. We now consider several examples to
illustrate these concepts.
1A.2 EXAMPLES
(a) Let A = {0, 12, 3, s, ...} = { l - n = 1, 2, 3, ...} (see Figure l.8). Clearly A
is bounded below by any real number r 0 and above by any real numbers ? 1.
o 1 2 .1 a
2 a 4s
Figure 1.8
(b) N = {1, 2, 3, ...}. This set is bounded below; e.g., I is a lower bound. Our intu-
ition tells us that N is not bounded above. It is obvious that there is no positive integer
n such that j n for all j E N. However, what is not so obvious is that there is no real
number .6 such that j s /3 for all j E N. In fact, given /3 E R, the proof of the exis-
tence of a positive integer n > /3 will require the least upper bound property of R (The-
orem 1.5.1).
(c) B = {r E 0 : r > 0 and r2 < 2}. Again it is clear that 0 is a lower bound for B.
and that B is bounded above; e.g., 2 is an upper bound for B. What is not so obvious,
however, is that B has no maximum. By the maximum or largest element of B we mean
an element a E B such that p a for all p E B. Suppose p E B. Define the rational
number q by
_ 2-p2 2p+2
q-p+(p+2) p+2
With q as defined, a simple computation gives
z_
q2-=
2 (p+2)2
Since p2 < 2, q > p and q2 < 2. Thus B has no largest element. Similarly, the set
has no minimum or smallest element. Intuitively, the largest element of B would satisfy
p2 = 2. However, as was shown in the introduction, there is no rational number p for
which p2 = 2.
Least Upper Bound of a Set

1A.3 DEFINITION Let E be a nonempty subset of R that is bounded above. An element
a E R is called the least upper bound or supremum of E if
(1) a is an upper bound of E, and
(ii) if /3 E R satisfies /3 < a, then 13 is not an upper bound of E.
Condition (ii) is equivalent to a /3 for all upper bounds J3 of E. Also by (ii), the
least upper bound of a set is unique. If the set E has a least upper bound, we write
a = sup E
to denote that a is the supremum or least upper bound of E. The greatest lower bound
or ifmum of a nonempty set E is defined similarly, and if it exists, is denoted by inf E.
There is one important fact about the supremum of a set that will be used repeat-
edly throughout the text. Due to its importance we state it as a theorem.
1 AA THEOREM Let A be a nonempty subset of H that is bounded above. An upper bound

or of A is the supremum of A if and only if for every 13 < a, there exists an element
x E A such that
13 <x5a.
Proof. Suppose a = sup A. If 13 < a, then 13 is not an upper bound of A. Thus there
exists an element x in A such that x > B. On the other hand, since a is an upper bound
ofA,x: a.
Conversely, if a is an upper bound of A satisfying the stated condition, then every
(3 < a is not an upper bound of A. Thus a = sup A. Q
1.4.5 EXAMPLES In the following examples, let's consider again the three sets of the pre-
vious examples.
(a) As in Example 1.4.2(a), let A = {0, Z, 3, ;, ...}. Since 0 is a lower bound of A and
0 E A, infA = 0. We now prove that sup A = 1. Since I - I'-1 < I for all n = 1,
2, ... , 1 is an upper bound. To show that I = sup A we need to show that if 13 E R
with $ < 1, then f3 is not an upper bound of A. Clearly if 0 0, then $ is not an up-
per bound of A. Suppose, as in Figure 1.9, 0 < 13 < 1. Then our intuition tells us that
there exists an integer no such that
1
n1
1 E A, and thus 13 is not an upper bound. Therefore sup A = 1. The existence

n
of such an integer no will follow from Theorem 1.5.1. In this example, inf A E A but
sup A e A.
(b) For the set N, inf N = 1. Since N is not bounded above, N does not have an upper
bound in R.
(c) In this example we prove that the supremum of the set
B={rEQ:r>0 and r2 <2},
if it exists, is not an element of 0. Suppose a = sup B exists and is in 0. Since a is ra-
tional, a2 0 2. Thus a2 < 2 or a2 > 2. But if a 6 B, then since B contains no largest
element, there exists q E B such that q > a. This contradicts the fact that a is an up-
per bound of B. Similarly, if a2 > 2, then there exists a q < a such that q2 > 2. But
then q is an upper bound of R, which is a contradiction of property (ii) of Definition
1.4.3. The least upper bound of B in N is V2 (Section 1.5, Exercise 9), which we know
is not rational.
Least Upper Bound Property of R

The following property, also referred to as the completeness property of R. distin-
guishes the real numbers from the rational numbers and forms the foundation for many
of the results in real analysis.
1A.6 SUPREMUM OR LEAST UPPER BOUND PROPERTY OF N Even nonempn sub-

set of P that is bounded above has a supremum in R.
For our later convenience we restate the supremum property of P as the infimum
property of R.
1.4.7 INFIMUM OR GREATEST LOWER BOUND PROPERTY OF P Ever nonempn

subset of H that is bounded below has an infimum in B.
Although stated here as a property, which we will assume as a basic axiom about
P, the least upper bound property of P is really a theorem due to both Cantor and
Dedekind, both of whom published their results independently in 1872. Dedekind, in
the paper "Stetigkeit and irrationale Zahlen" (Continuity and irrational numbers), used
algebraic techniques now known as the method of Dedekind cuts to construct the real
number system P from the rational numbers O. He proved that the system R contained
a natural subset of positive elements satisfying the order axioms (01) and (02), and fur-
thermore, that P also satisfied the least upper bound property. The books by Burrill and
t by Spooner and Mentzger cited in the Supplemental Readings are devoted to number
systems. Both texts contain Dedekind's construction of R. Cantor, on the other hand,
constructed P from 0 using Cauchy sequences. In the miscellaneous exercises of
Chapter 2 we will provide some of the key steps of this construction.
1.4.8 EXAMPLE In this example we show that for every positive real number y > 0, there
exists a unique positive real number a such that a2 = y; i.e., a = v. The uniqueness
of a was established in Example 1.2.9(b).
We only prove the result for y > 1, leaving the case 0 < y s I to the exercises (Ex-
ercise 6). Let
C={xEH:x>0 and x2<y}.
With y = 2, this set is similar to the set B of Example 1.4.2(c), except that here we con-
sider all positive real numbers x for which x2 < Y. Since y > 1, 1 E C and thus C is
nonempty. Also since y > 1, y2 > y, and thus y is an upper bound of C. Hence by the
least upper bound property, C has a supremum in P. Let a = sup C. We now prove that
a2 = y. To accomplish this we show that the assumptions a2 < Y and a2 > v lead to
contradictions. Thus a2 = y.
Define the real number $3 by

_ y(a + 1)
C2-_
/3=a+(a+r) a+r (1)
Then
y(! - I)(a2 )
P'- - y = (2)
(a + V)2
If a2 < y, then by (1) $3 > a, and by (2) $32 < v. This contradicts that a is an upper
bound for C. On the other hand, if Cr' > y, then by (1),6 < a and by (2), j32 > Y. Thus
if x E R with x ? $3, then x2 > y. Therefore $3 is an upper bound of C. This contradicts
that a is the least upper bound of C. Since S defined by (I) may not be rational, the
same proof will not work for the set B of Example 1.4.2(c). However, using Theorem
1.5.2 of the following section, it is possible to also prove that sup B = 12.
For convenience, we extend the definition of supremum and infimum of a subset E
of N to include the case where E is not necessarily bounded above or below.
1.4.9 DEFINITION If E is a nonempty subset of R, we set

sup E = oo if E is not bounded above, and
inf E _ - oo if E is not bounded below.
For the empty set ¢, every element of i is, an upper bound of 46. For this rea-
son the supremum of the empty set 44 is taken to be -oo. Similarly, inf d = cc.
Also, for the symbols -oo and oo we adopt the convention that -oo < x < cc for
every x E R.
Intervals
Using the order properties of R, we can define certain subsets of R known as intervals.
1.4.10 DEFINITION For a, b E R, an-55 b, the open interval (a, b) is defined as

(a,b)={xE68:a<x<b},
whereas the dosed interval [a, b] is defined as
[a, b] _ {x E IF8 : a s x < b}.
In addition, we also have the half-open (half-closed) intervals
[a,b)={xEJl:a<x<b},
(a,b]={xEl2:a<xsb},
and the infinite intervals
(a,oo)={xER:a <x<oo},
[a,oo)=Ix ER:asx<oo},
with analogous definitions for (-cc, b) and (-oo, b). The intervals (a. 00), (-oo, b)
and (-oo, oo) = R are also referred to as open intervals, whereas the intervals [a, oo)
and (-oo, b) are called closed intervals. (See Figure 1.10.)
(a. b)
a b
[a, b]
a b
(a, b)
Figure 1.10 Types of Intervals
In this definition, when b = a, (a, a) = 0 and [a, a] = {a}. Although the empty
set 0 and the singleton {a} do not fit our intuitive definition of an interval, we will in-
clude them as the degenerate case of open and closed intervals, respectively. It should
be noted that the intervals of the form (a, b), (a, b], [a, b), and [a, b] with a, b E R.
a s b, are all bounded subsets of R.
An alternative way of defining intervals without use of the adjectives open and
closed is as follows.
1.4.11 DEFINITION A subset J of R is an interval if whenever x, y E J with x < y, then

every t satisfying x < t < y is in J.
This definition also allows the possibility that J is empty or a singleton. One can
show that every set J satisfying the definition is one of the intervals defined in Defini-
tion 1.4.10 (Exercise 21).
EXERCISES 1.4
1. Use the axioms for addition and multiplication to prove the following: If a E R, then
a. 0. b. -a.
c. -(-a) = a.
2. Let a, b E It Prove the following.
a. If a # 0, then I/a # 0. b. If ab - 0, then either a = 0 or b = 0.
3. Let a, b, c E R. Prove the following.
a. If a > b, then a + c > b + c. b. If a > b and c < 0, then ac < bc.
c. If a * 0, then a2 > 0. d. If a > 0 then I/a > 0, and if a < 0 then 1/a < 0.
4. *If a, b E IR, prove that ab ](a2 + b2).
5. Find the supremum and infimum of each of the following sets.
*a.A1,ll1,...}=
248 1l 1J ll
nEll}
'):E N b.Bsinn 22 :nEN}
ll J
'c. C {n cosna:nEN} d. D {cos ni :nEN}
"e.E{2nn:nEt`11 f. F {(-lr-!:nGN}
1(
"g. G{(1 nEN) h. H +(-lY')":nEN}
i.l={xER:xz<4} j.J={x2:-2<x<2}
6. If 0 < y < 1, prove that there exists a unique positive real number x such that x2 = y.
7. Prove that there exists a positive real number x such that x3 = 2.
8. Let A be a nonempty subset of R. If a = sup A is finite, show that for each e > 0, there is an a E A such that
a - e < a < a.
9. Let E be a nonempty subset of R that is bounded above, and set U = jig E R is an upper bound of E}. Prove
that sup E = inf U.
10. Let A be a nonempty subset of R and let -A = (-x: x E A). Prove that infA = -sup(-A).
11. Use the least upper bound property of R to prove that every nonempty subset of R that is bounded below has an
infimum.
12. If A and B are nonempty subsets of R with A C B, prove that
inf B s infA s sup A <_ sup B.
13. Suppose that A and B are bounded subsets of R. Prove that A U B is bounded and that
sup(A U B) = sup{sup A, sup B}.
14. For A. B, subsets of R, define
A+B=(a+b:a6A,bEB) and
A = (-1, 2, 4, 7) and B = {-2, -1, 1), find A + B and A B.
'b. If A and B are nonempty and bounded above, prove that
sup(A + B) = sup A + sup B.
c. If A and B are nonempty subsets of the positive real numbers that are bounded above, prove that
sup(A B) = (sup A)(sup B).
d. Give an example of two nonempty bounded sets A and B for which sup(A B) # (sup A)(sup B).
15. Let f, g be real-valued functions defined on a nonempty set X satisfying Range f and Range g are bounded sub-
sets of R. Prove each of the following.
*a. sup{f(x) + g(x) . X E X) s sup{f(x) : x e X} + sup{g(x) : x E X}
b. inf{f(x) : x E X} + inf{g(x) : x E X} s inf{f(x) + g(x) : x r= X}
c. If f(x) : g(x) for all x E X. then sup{f(x) : x E X} s sup{f(x) : x e X}.
16. Let X = Y = [0, 1 ] and let f : X X Y -+ R be defined by f (x, y) = 3x + 2y.
a. For each x E X, find F(x) = sup{f(x, y) y E Y}; then find sup{F(x) : x E X}.
b. For each y E Y, find G(y) = sup{ f(x, y) : x E X}; then find sup(G(y) : y E Y).
c. Find sup{f(x, y) : (x, y) E X X Y}. Compare your answer with your answers in parts (a) and (b).
17. Perform the computations of Exercise 16 with X = [-1, 11. Y = [0, 2], and f(x, v) = 3x - 2y.
18. Let X, Y be nonempty sets, and let f be a nonnegative real-valued function defined on X X Y. For each x E X and
y E Y, define
F(x) = sup{f(x, y) : y E Y}, G(y) = sup{f(x. v) : x E X}.
Prove that
sup{F(x) : x E X} = sup{G(v) : y E Y} = sup{f(x, y): (x, v) E X X Y}.
19. Let X, Y be nonempty sets and let f : X X Y -+ R be a function with bounded range. Let
F(x) = sup{f(x, y) : y E Y} and H(y) = inf{f(x, y) : x E X}. Prove that
sup{H(y) : y E Y} inf{F(x) : x E X}.
20. Let X = Y = [0, I ]. Perform the computations of Exercise 19 for each of the following functions f(x, y).
1, x=y,
a. f(x, v) = 3x + 2r b. f(x, y) =
x*v
21. Let J be a subset of R that has the following property: if x, y E J with x < y, then t E J for every t satisfying
x < t < Y. Prove that J is an interval as defined in Definition 1.4.10.
1.5 1 Consequences of the Least Upper Bound Property

In this section we look at a number of elementary properties of the real numbers, which
in more elementary courses are usually taken for granted. As we will see, however,
these are all actually consequences of the least upper bound property of the real numbers.
1.5.1 THEOREM (Archimedian Property) If x, y E Ll and x > 0, then there exists a posi-
tive integer n such that
nx > Y.
Proof. If y <- 0, then the result is true for all n. Thus assume that y > 0. We will
again use the method of proof by contradiction. Let
A={nx:nEN}.
If the result is false, that is, there does not exist an n E N such that nx > y. then nx y
for all n E N. Thus y is an upper bound for A. Thus since A * 46, A has a least upper
bound in R. Let a = sup A. Since x > 0, a - x < a. Therefore a - x is not an upper
bound and thus there exists an element of A, say mx, such that
a - x<mx.
But then a < (m + 1)x, which contradicts the fact that a is an upper bound of A.
Therefore, there exists a positive integer n such that nx > y.
Remark. One way in which the previous result is often used is as follows: given
e > 0, there exists a positive integer n such that noe > 1. As a consequence,
1
- < E
n
for all integers n, n ? no.

1.5 Consequences of the Least Upper Bound Property 29
1.5.2 THEOREM If x, .v E l and x < v, then there exists r E O such that

x<r<y.
Proof. Assume first that x ? 0. Since y - x > 0, by Theorem 1.5.1 there exists an
integer n > 0 such that
n(y - x) > 1 or ny > I + nx.
Again by Theorem 1.5.1, {k E N : k > nx} is nonempty. Thus by the well-ordering
principle, there exists m E N such that
m - 1 :5 nx<m.
Therefore
nx<m<_I+nx<ny,
or dividing by n,
x< n <y.
If x < 0 and y > 0, then the result is obvious. Finally, if x < y < 0, then by the pre-
ceding calculations there exists r E 0 such that -y < r < -x; i.e., x < -r < y.
The conclusion of Theorem 1.5.2 is often expressed by the statement that the ra-
tional numbers are dense in the real numbers; that is, between any two real numbers
there exists a rational number. A precise definition of "dense" is given in Definition
3.1.12.
Another consequence of the least upper bound property is the following theorem
concerning the existence of nth roots.
1.5.3 THEOREM For every real number x > 0 and every positive integer n, there exists a
unique positive real number y so that y" = x.
The number y is written as lcx or x'1" and is called the nth root of x. The unique-
ness of y is obvious. Since the existence of y will be an immediate consequence of the
intermediate value theorem (Theorem 4.2.11), we omit the details of the proof in the
text. A sketch of the proof of Theorem 1.5.3, using the least upper bound property, is
included in the miscellaneous exercises. It should be emphasized that the proof of The-
orem 4.2.11 also depends on the least upper bound property.
1.5.4 COROLLARY If a, b are positive real numbers, and n is a positive integer, then
(ab)Un = al/nb'l".
Proof. Set a = a'"" and R = P". Then

ab=a"/3"
Thus by uniqueness, af3 = (ab)'t".
0 EXERCISES 1.5
1. *If r and s are positive rational numbers, prove directly (without using the supremum property) that there exists
an n E N such that nr > s.
2. Given any x E R, prove that there exists a unique n E Z such that n - 1 5 x < n.
3. If r * 0 is a rational number and x is an irrational number, prove that r + x and rt are irrational.
4. Prove directly (without using Theorem 1.5.1) that between any two rational numbers there exists a rational
number.
5. If x. y E R with x < y, show that x < ty + (I - t)x < y for all t, 0 < t < 1.
6. *a. Prove that between any two rational numbers there exists an irrational number.
*b. Prove that between any two real numbers there exists an irrational number.
7. If x > 0, show that there exists n E N such that 1/2" < x.
8. Let x, y E R with x < y. If u E P with u > 0. show that there exists a rational number r such that x < ru < Y.
9. Let B = {r E 0 : r > 0 and r2 < 2} and a = sup B. Prove that a2 = 2.
1.6 Binary and Ternary Expansions3

In our standard base 10 number system we use the integers {0, 1.. . . , 9} to represent
real numbers. Base 10, however, is not the only possible base. In base 3 (ternary) we
use only the integers {0, 1, 2}, and in base 2 (binary) we use only {0, I }. In this section
we will show how the least upper bound property may be used to prove the existence
of the expansion of a real number x. 0 < x 5 1, in a given base. For purposes of illus-
tration, and for later use, we will use base 2 and 3, which are commonly referred to as
the binary and ternary expansions, respectively.
In base 10, given the decimal .1021, what we really mean is the real number
-
given by
0 2
-+1
10 107+70+70-
1
However, in base 3, the expansion .1021 represents

1 0 2 1
3+32+3;+34,
which is the ternary expansion of 34/81.
For a given x, 0 < x s 1, the ternary or base 3 expansion of x is defined induc-
tively as follows.
3. This section can be omined on first reading. The ternary expansion of a real number is only required in
Section 3.3 and can be covered at that point.
1.6 Binary and Ternary Expansions 31
1.6.1 DEFINITION Let n, E {0, 1, 2} be the largest integer such that

n
3 < X.
Having chosen n1, ... , nk, let nk+ 1 E {0, 1, 2} be the largest integer such that
n,
+n2+. . .+nk+nk+1 <X.
3 32 3k 3k+'
The expression .n,n2n3 is called the ternary expansion of x.

If we set
E={3 +.+3k:k=1,2, ..},

then E * 0 and E is bounded above by x. As we will shortly see, sup E = x. In terms
of series, which will be covered in detail later, we have
00 n
X = k.
k=1 3
The binary expansion of x, or the expansion of x to any other base, is defined simi-
larly. For the binary expansion, the integer nk at each step is chosen as the largest inte-
ger in {0, 1).
1.6.2 EXAMPLES
(a) We now use Definition 1.6.1 to obtain the ternary expansion of 3. As the first step.
we must choose n, as the largest integer in {0, 1, 2} such that
3 < 3'
This inequality fails for n, = 1, 2. Thus n, = 0. To find n2, we choose the largest inte-
ger n2 E {0, 1, 2) such that
3+32<3'
which is satisfied by n2 = 2. To find n3 we must have
2 1 3.
0+ + n3 <
3 32 33
It is left as an exercise to show that this is satisfied for n3 = 0, 1, 2. Thus we take

n3 = 2. At this stage we conjecture that nk = 2 for all k >_ 2, and that
1
3 = .02222 (base 3).
32 Chapter t The Real Number System
To see that this indeed is the case, we use the fact that for the geometric series.
rk=(11
k=o
r), O<r<1.
Thus
.0222...=
k=, 3k
- _-
2 2
9 kap 3
I k 2
9 1
1
i
1
In (c) we will illustrate how mathematical induction may also be used to prove such a
result. The above ternary expansion is not unique. The number ;also has the finite ex-
pansion
I = .1000.
(base 3).
3
We will discuss this in greater detail at,the end of this section.

(b) The binary expansion of is given by
3 = .010101 (base 2).
This expansion can be obtained using the definition and induction (Exercise 3). Alter-
natively, using the geometric series we have
X
r_I 2
1 I" I I I
. 4
-U 4 = 3.
(c) The ternary expansion of 1 is given by
1 =.1111 . (base 3).

2
We now show in detail how this expansion is derived. We will use the second principle
of mathematical induction to prove that nk = I for all k E K. Since < ; and
i > n, = 1. Thus, the result is true for k = 1. Let k > I and assume that n; = I for
Using the identity

1+
3
+ I
all j < k. By definition nk is the largest integer in {0, 1, 2) such that
3k-1
+nk<1
3` 2
(3)
r - r"+i
r+r2+ +r"=
1-r'
r # 1, (Example 1.3.3(a))
with n = k - 1, we obtain
3+
+ 3kli = ' 2
` 1 - 3k1_ 1
1.6 Binary and Ternary Expansion 33
Substituting into equation (3) and multiplying by 2 gives

2;tk
1- 3k. + < I.
This inequality is true if nk = 0 or I and is false if nk = 2. Since nk is to be chosen as
the largest integer of {0, 1, 2} for which (3) holds, nk = 1. Thus by Theorem 1.3.4.
n; = 1 for all k E N.
1.6.3 THEOREM Let x E U8 with 0 < x <- 1. and with {n} as defined in Definition
1.6.1, let
(n1 nk
3 3'
Then sup E = x.
Proof. Let a = sup E. Since x is an upper bound, a s x. Suppose a < r. Let k be
the smallest positive integer such that
1 <.r - a or a+ .<x.
3k 3k
Since a = sup E,
But
n1
+...+3k+;k=
I
3 +... i
irk n1 nj.._ 1
3k i+ nk + < x.
3 3k
If nk = 0 or 1, this contradicts the choice of nk. If nk = 2. then we have

11 nk_ r + 1
3 + .
. . + 3k_1 < x.
If any n;, 1 s j <_ k - l is 0 or 1, we have a contradiction to the choice of n;. If all the
ni = 2, then we obtain
2
3+3<xsl, 1
which is also a contradiction.

The expansion
x = .nln,n3
is finite or terminating if there exists an integer m E tk such that nk. = 0 for all k > m.
Otherwise, the expansion is infinite or nonterminating. The expansion of a real num-
ber in a given base is not always unique; when x has a finite expansion, it has an infi-
nite expansion as well. For the ternary expansion, when
x=3m, aEN with O<a<3'",

and 3 does not divide a, then x has a finite expansion of the form
x= + +3m, a,"E{1,2}.
or an infinite expansion of the form
3
a, 0 2
x= +- +3m+ k when a,"=1,
3 k=m+i 3*
or
x=3+ + + k, when am = 2.
I k=m +1 3
© EXERCISES 1.6
1. Find the ternary expansion of each of the following.
*a. i b. i3
2. Find the real number determined by each of the following finite or infinite ternary expansions. In the case of an
infinite expansion, use the geometric series as in Example 1.6.2 (c) to determine your answer.
*a. .0022 b. .00222 c..1010
*d..101010 e..001001001 f. .121212
3. *Find the binary expansion of 3. Use induction to prove the result.
4. Find both the finite and infinite binary expansions of 6.
5. If 0 < x s 1, prove that the infinite ternary expansion of x is unique.
6. If x = a/2m. with a E N odd, and 0 < a < 2m, show that x has a finite binary expansion of the form
al ni 0
x=
2
+ + a'",
2'"
a.m = 1, and an infinite expansion x=
2
+ +
2m
+7
k.m+i
2k.
1
1.71 Countable and Uncountable Sets

In discussing sets, we all have an intuitive understanding of what it means for a set to
be finite or infinite, and what it means for two finite sets to be of the same size; that is,
to have the same number of elements. For example, the sets A = {2, 7, 11, 211 and
B = {7, 3, 19, 32} both have the same number of elements; namely, four. We discover
this by counting the number of elements in each of the two sets. Alternatively, we can
reach the same conclusion without counting, by simply pairing up the elements:
7+2
3+7
19 + 11
32 + 21
For infinite sets, the concept of two sets being of the same size or having the same
number of elements is vague. For example, let S denote the squares of the positive in-
tegers; namely,
S = {12, 22, 32, . . J.
Then on one hand, S is a proper subset of the positive integers N, yet as Galileo
(1564-1642) observed, the sets N and S can be placed into a one-to-one correspondence
as follows:
1+12
2+22
3+32
This example caused Galileo, and many subsequent mathematicians, to conclude that
the standard notion of "size of a set" did not apply to infinite sets. Cantor, on the other
hand, realized that the concept of one-to-one correspondence raised many interesting
questions about the theory of infinite sets. In this section we take a closer look at infi-
nite sets and what it means for an infinite set to be countable. We begin by defining the
concept of equivalence of sets.
1.7.1 DEFINITION Two sets A and B are said to be equivalent (or to have the same car-
dinaiity), denoted A -- B, if there exists a one-to-one function of A onto B.
The notion of equivalence of sets satisfies the following:
(I) A -A. (reflexive)
(ii) If A B, then B --A. (symmetric)
(iii) If A - B and B -- C, then A ^- C. (transitive)
1.7.2 DEFINITION For each positive integer n, let N. = {1, 2, ... , n}. As in Section 1.1,
N denotes the set of all positive integers. If A is a set, we say:
(a) A is finite if A N. for some n, or if A = t¢.
(b) A is infinite if A is not finite.
(c) A is countable if A N.
(d) A is uncountable if A is neither finite nor countable.

(e) A is at most countable if A is finite or countable.
The reader might want to ponder Exercise 21 at this point. Countable sets are of-
ten called denumerable or enumerable sets. It should be pointed out that some text-
books, when using the term countable, include the possibility that the set is finite.
1.7.3 EXAMPLES
112 ,
22,32, J. Then the function g(n) = n2 is a one-to-one
(a) As before, let S =
mapping of N onto S. Thus S - N, and S is countable.
(b) For our second example, we show that 7L -- N. To see this, consider the function
f: N -+ Z defined by
n
(n even),
.f(n) (n 2
2 1), (n odd).
Figure 1.11 illustrates what the mapping f does to the first few integers. It is left as an
exercise (Exercise 1) to show that this function is a one-to-one mapping of N onto Z.
Thus 7L -- N, and the set Z is also countable.
Figure 1.11
As another illustration of countable sets consider the following theorem.
1.7.4 THEOREM N x N is countable.

Proof. For this example, it is easier to construct a one-to-one mapping of N X N
onto N. Such a function is given by
f(m, n) = 2-'(2n - 1).
It is left as an exercise (Exercise 3) to show that f as defined is a one-to-one mapping
ofNXNontoN. 0
One of our goals in this section is to show that the set Q of rational numbers is
countable.
Sequences
1.7.5 DEFINITION If A is a set, by a sequence in A we nean a function f from N into A.
For each n E N, let x = f(n). Then x is called the nth term of the sequence f.
For notational convenience, sequences are denoted by or just rather
than the function f Note, however, the distinction between which denotes the
sequence, and {x : n = 1, 2, . . .}, which denotes the range of the sequence. For ex-
ample, {1 - (-1)"} denotes the sequence f where
f(n) = X. = 1 - (-1)".
On the other hand, {x : n = 1, 2, ...} = {0, 2}.
By definition, if A is a countable set, then there exists a one-to-one function f from
N onto A. Thus
A = Rangef = {x,,:n = 1,2,...}.
The sequence f is called an enumeration of the set A; i.e., A = {x : n = 1, 2, .. .}
with x * x,R whenever n * m. This ability to enumerate the elements of a countable
set plays a key role in the proof of the following theorem.
1.7.6 THEOREM Every infinite subset of a countable set is countable.

Proof. Let A be a countable set and let Ix.: n = 1, 2, ...} bean enumeration of A.
Suppose E is an infinite subset of A. Then each x E E is of the form xk for some k E N.
We inductively construct a function f : N -) E as follows: Let n, be the smallest pos-
itive integer such that x.. E E. Such an integer exists by the well-ordering principle.
Having chosen n1..... nk - ,, let nk be the smallest integer greater than nk _ , such that
E E\ Set f(k) = x,,. Since E is infinite, f is defined on N.
If m > k, then nk and thus x,,. Therefore f is one-to-one. The function
f is onto E since if x E E, then x = x, for some j. By construction, nk = j for some k.
and thus f(k) = x.
1.7.7 THEOREM If f maps N onto A. then A is at most countable.

Proof. If A is finite, the result is certainly true. Suppose A is infinite. Since f maps
N onto A, each a E A is of the form f (n) for some n E N. For each a E A, by the well-
ordering principle
f-'({a}) = {n E N J (n) = a}
has a smallest integer, which we denote by n,,. Consider the mapping a -a nd of A into
N. If a * b, then since f is a function, n. * nb. Also, since A is infinite, {n, : a E A)
is an infinite subset of N: Thus the mapping a--+ n, is a one-to-one mapping of A onto
an infinite subset of N. Therefore by-Theorem 1.7.6, A is countable.
Indexed Families of Sets

In Section 1.1 we defined the union and intersection of two sets. We now extend these
definitions to larger collections of sets. Recall that if X is a set, 91(X) denotes the set of
all subsets of X.
1.7.8 DEFINITION Let A and X be nonempty sets. An indexed family of subsets of X with
index set A is a function from A into 9(X).
If f : A -+ 13(A), then for each a E A, we let Ea = f(a). As for sequences, we
denote this function by {Ea}aE,%. If A = N, then {En}nEN is called a sequence of sub-
sets of X. In this instance, we adopt the more conventional notation {En} 1 to denote
{En}nEN
1.7.9 EXAMPLES The following are all examples of indexed families of sets.
(a) The sequence {N.} 1, where N. = {1, 2.... , n}, is a sequence of subsets of N.
(b) For each n E N, set 1n = {x (E R : 0 < x < Then {1n} 1 is a sequence of sub-
sets of N.
(c) For each x, 0 < x < 1, let
Ex= {rE0:0t=- r<x}.
Then {E;},,E(0.1)is an indexed family of subsets of 0. In this example, the open interval
(0, 1) is our index set.
1.7.10 DEFINITION Suppose {Ea}aEA is an indexed family of subsets of X. The union of the
family of sets (Ea}aEA is defined to be
6A E. = {x E X : x r= Ea for some a E A).

The intersection of the family of sets {Ea}aEA is defined as
n E. = {x E X: x E Ea for all a E A}.
If A = N we use the notation

00 00
U E. and n-1
n=1
fl En instead of u E.
nEN
and fl En,
nEN
respectively. Also, if A = NO then

k
nEN, E. is denoted by U En,

n=1
with an analogous definition for the intersection. Occasionally, when the index set A is
fixed in the discussion, we will use the shorthand notation U. E. or !1a E. rather than
aU E. or f A Ea.
1.7.11 EXAMPLES We now consider the union and intersection of the families of sets given
in the previous example.
(a) With N. = {1, 2, ... , n}, we have

00 00
nN,, = {1}, and UN,, = N.
n=1 n-i
Since I is the only element that is in N. for all n, fl,, Ni,, = {1}. For the union, since
N. C Ni for all n, U. Nn C Ni. On the other hand, if n E N, then n E NI,,, and as a
consequence, N C U,, N,,, which proves equality.
(b) As in the previous example, for n E NI, let In = {x E R:0 < x < .1}. We first
show that 00
n in = (Q.
Suppose the conclusion is false. Then there exists x E H such that x E In for all n; i.e.,
0<x< 1,n forallnEN.

This, however, contradicts Theorem 1.5.1, which guarantees the existence of a positive
integer n such that nx > 1. For the union, since In C 11 for all n ? 1,
00
R=1
={xE08:0<x< 1}.
(c) We leave it as an exercise (Exercise 9) to show that if E, is defined as in Example
1.7.9(c); i.e., E, = {r E O : 0 5 r < x}, then
sEn E,={0} and 1}.
As for a finite number of sets, we also have analogues of the distributive laws and
De Morgan's laws for arbitrary unions and intersections.
1.7.12 THEOREM (Distributive Laws) If Ea, a E A. and E are subsets of a set X, then
(a) En( U Ea) = aU (E n

E U (fl Ea) n (E U E.).
*eA
1.7.13 THEOREM (De Morgan's Laws) If {EQ},EA is a family of subsets of X, then
(a)
aE
UA EQ = aEA
fl Ea,
c
(b) *GA
n Ea = aEA
u E:-
The proofs of both of these theorems are left to the exercises. The following theo-
rem is an analogue of Theorems 1.25 and 1.2.6.
1.7.14 THEOREM Lei f be a function from X into Y.

(a) If {E,},EAis a famil of subsets of X, then
f (,U E0) = U f(E )

f(n E,) C n f(E,).
(b) If {B,}OEA is a family of subsets of Y. then
P B,)
(U = U f -'(B.).
f '(n B.) = n f-,(B,.).

GA A
Proof. (a) To illustrate the method of proof, we will prove that f (n. E,) C fl, f (E,).
The proof that f (U, E,) = U f(E,) is left to the exercises.
Suppose y E f(fl, E,.). Then by definition of the image of a set, Y = f (.Y) for some
x E fl, E,. But then x E E, for all a E A, and thus f (x) E f (E.) for all a E A. There-
fore y E fl, f(E,).
(b) Again we will only prove one of the two identities. The other is left to the exercises.
In this instance we will show that f -' (U, B,) = U. f Let B = U, B,. Since
B. C B for all a e A, we have f `(B,) C f `(B) for all a e A. Thus U. f "' (B,)
C f -' ( U,, B.). In the other direction, suppose x E f -'(U,, B,). Then by the definition
of the inverse image of a set, f (x) E U. B,. Thus f (x) E B,,, for some index a,, E A.
But then x E f -' (B,), and hence x E U, f '(B.). This proves equality. U
The Countability of 0
The countability of the set of rational numbers Q will. follow as a corollary of the fol-
lowing theorem.
1.7.15 THEOREM If (E,)'".1 is a sequence of countable sets and

x.
S = n-I
U
then S is countable.
Proof. Since E is countable for each n E N. we can write
E = {x,,,k:k= 1,2,...).
Since E, is an infinite subset of S, the set S itself is infinite. Consider the function
h: N X N - Shy
h(n, k) = x,k.
.The function h, although not necessarily one-to-one, is a mapping of N X N onto S.

Thus since N X N - N, there exists a mapping of N onto S. Hence by Theorem 1.7.7
the set S is countable.
1.7.16 COROLLARY CQ is countable.
Proof. For each m E N, let
Em= {iE}.
Ua_, Eby
Theorem
Then E. is countable, and since 0 = 1.7.15 the set 0 is count-
able.
The UncountabUity of 08
In November 1873, in a letter to Dedekind, Cantor asked whether the set R itself was
countable. A month later he answered his own question by proving that 01 was not count-
able. We now prove, using Cantor's elegant "diagonal" argument, that the closed interval
[0, 11 is uncountable,. and thus 08 itself is uncountable. For the proof we will use the fact
that as in Section 1.6, every x E [0, 11 has a decimal expansion of the form x = n ,n,
with n; E {0, 1, 2, . .. , 9}. As for the binary and ternary expansions, the decimal expan-
sion is not necessarily unique. Certain numbers such as ,'-o have two expansions; namely.
.100 and 10 = .0999

10
This, however, will not be crucial in the proof of the following theorem.
1.7.17 THEOREM The closed interval [0, 11 is uncountable.

Proof. Since there are infinitely many rational numbers in [0, 1 ], the set is not finite.
To prove that it is uncountable, we only need to show that it is not countable. To ac-
complish this, we will prove that every countable subset of [0, 1} is a proper subset of
[0, 1 ]. Thus [0, 1 ] cannot be countable.
Let E = {xn : n = 1, 2, .. .} be a countable subset of [0, 1]. Then each xn has a
decimal expansion
. .
xn = xn.lxn.2xn,3 .
where for each k E N, x,.k E 10, 1, .. .. 9}. We now define a new number
Y = .yly2Y3 ..
as follows: If x, ,, c 5. define yn = 6; if xn ? 6, define yn = 3. Then y E [0, 1]. and
since yn * 0 or 9, y is not one of the real numbers with two decimal expansions. Also.
since for each n E N. yn # xn,n, we have y * xn for any n. Therefore Y E E; i.e., E is a
proper subset of [0, 11.
Another example of an uncountable set is given in the following theorem.
42 Chapter) The Real Number System
1.7.18 THEOREM If A is the set of all sequences whose elements are 0 or 1, then A is
uncountable.
Remark. The set A is the set of all functions f from N into {0, I }. Thus a sequence
f E A if and only if f(n) = O or I for all n.
Proof. As in the previous theorem, we will also prove that every countable subset of
A is a proper subset of A, and thus A cannot be countable.
Let E be a countable subset of A and let {s : n = 1 , 2, .. .} be an enumeration of
the set E. For each n, s. is a sequence of 0's and l's. We construct a new sequence s as
follows: for each k E N, let
s(k) = I - sk(k).
Thus if sk(k) = 0, s(k) = 1, and if sk(k) = 1, s(k) = 0. Thus s e A. Since for all k E N
s(k) # sk(k), we haves* s for any n E N. Therefore s a E; i.e., Ec A. 0
In the previous two theorems we proved that the closed interval [0, 1 ] and the set
of all sequences of 0's and I's are both uncountable sets. One may ask, are these two
sets equivalent? Considering the fact that every real number x E [0,1 ] has a binary ex-
pansion, which is really a sequence of 0's and I's, one would expect that the answer is
yes. This indeed is the case (Miscellaneous Exercise 7).
EXERCISES 1.7
1. a. Prove that the function f of Example 1.7.3(b) is a one-to-one function of N onto Z.
b. Find a one-to-one function of Z onto N.
2. *Let 0 denote the set of odd positive integers. Prove that 0 -- N.
3. Prove that the function f of Theorem 1.7.4 is a one-to-one function of N X N onto N.
4. *a. If a. b e R with a < b, prove that (a, b) - (0, 1).
b. Prove that (0, 1) -- (0, oo).
5. Suppose X. Y, Z are sets. If X - Y and Y- Z, prove that X Z.
6. *a. If A ^- X and B Y, prove that (A X B) (X X Y).
b. If A and B are countable sets, prove that A X B is countable.
7. If X - Y, prove that 9(X) - 9(Y).
& Find U , A. and fl°, A. for each of the following sequences of sets
nEN b.A.={xCE R:-1,<x<1}. nEN
nEN x:5 l-,-,), nEN
e.A ={xER: x51nEN,n; 2 nEN
9. For each x E (0, 1), let Ex = {r E Q : 0 s r < x}. Prove that
n E,={0}
xE(0,I)
and U E,={r E Q:0sr< 1).
Notes 43
12. Let f be a function from X into Y.

a. If {Ea}eA is a family of subsets of X, prove that f(Ua Ea) ='U. f(E*).
b. If {Ba}aEA is a family of subsets of Y, prove that f -' (laBa) = fla f -'(B.).
13. a. If A is a countable subset of an uncountable set X. prove that X \ A is uncountable.
b. Prove that the set of irrational numbers is uncountable.
14. Suppose f is a function from X into Y. If the range of f is uncountable, prove that X is uncountable.
15. a. For each n E N, prove that the collection of all polynomials in x of degree less than or equal to n with rational
coefficients is countable.
b. Prove that the set of all polynomials in x with rational coefficients is countable.
16. Prove that the set of all intervals with rational end points is countable.
17. Let A be a nonempty bounded subset of R and let a = sup A. If a a A, prove that for every e > 0. the interval
(a - e, a) contains infinitely many points of A.
18. *a. Prove that (0, 1) - (0, 1 ]. (This problem is not easy!)
b. Prove that (0, 1) -- [0. 1 ].
19. *A real number a is algebraic if there exists a polynomial p(x) with integer coefficients such that p(a) = 0. Prove
that the set of algebraic numbers is countable.
20. Prove that any infinite set contains a countable subset.
21. Prove that a set is infinite if and only if it is equivalent to a proper subset of itself.
22. *Prove that any function from a set A to the set 9(A) of all subsets of A is not onto 9(A).
23. *Prove that [0, 1] x [0, 1 ] - [0, 1].
NOTES
The most important concept of this chapter is the least up- pend on previous theorems that require the least upper
per bound property of the real numbers. This property bound property in their proofs.
will be fundamental in the development of the underlying The emphasis on the least upper bound property is
theory of calculus. In the present chapter we have already not meant to overshadow the importance of the concepts
seen its application in proving the Archimedian property of countable and uncountable sets. The fact that the ratio-
(Theorem 1.5.1). For the rational number system this nal numbers are countable, and thus can be enumerated,
property can be proved directly. For the real number sys- will be used on several occasions in the construction of
tem it was originally assumed as an axiom by Archimedes examples. In all the examples and exercises. every infinite
(287-212 B.c.). Cantor, however, proved that the Archi- subset of R turns out to be either countable or equivalent
median property was no axiom, but a proposition deriv- to [0, 1 ]. Cantor also made this observation, and it led
able from the least upper bound property. him to ask whether this result was true for every infinite
In subsequent chapters the least upper bound prop- subset of R. Cantor was never able to answer this ques-
erty will occur in proofs of theorems, either directly or in- tion; nor has anyone else. The assertion that every infinite
directly, with regular frequency. It will play a crucial role subset of R is either countable or equivalent to [0, 11 is
in the characterization of the compact subsets of R and in known as the continuum hypothesis. In 1938 Kurt Godel
the study of sequences of real numbers. It will also be re- proved that the continuum hypothesis is consistent with
quired in the proof of the intermediate value theorem for the standard axioms of set theory; that is. Godel showed
continuous functions. One of the corollaries of this theo- that the continuum hypothesis cannot be disproved on the
rem is Theorem 1.5.3 on the existence of nth roots of pos- basis of the standard axioms of set theory. On the other
itive real numbers. Many other results in the text will de- hand, in 1963 Paul Cohen showed that the continuum hy-
pothesis is undecidable on the basis of the current axioms is equivalent to some subset of B. We also write A < 8
of set theory. provided A s B and A # B. C)early, N < [0, 1 ]. Also.
Cantor's creation of the theory of infinite sets was Exercise 22 of the previous section proves that for any set
motivated to a great extent by problems arising in the
A, A < Using equivalence of sets, it follows that if
study of convergence of Fourier series. We will discuss
some of these problems in greater detail in Chapter 9. The A s B and B C, then A s C. This naturally leads to
notion of one-to-one. correspondence allowed Cantor to the following question: If A and 8 are sets such that
define the "power" or what is now called the "cardinality" A c B and B < A. is A = B? The answer is yes, and the
of a set. For a given set A, the cardinal number or power result is known as the Schroder-Bernstein theorem.
of A is denoted by A. Two sets A and B are said to have Cantor did more than just introduce cardinal num-
bers; he also defined addition, multiplication, and expo-
the same cardinality, denoted A = B, provided that A is
nentiation of cardinal numbers. Cantor's original work
equivalent to B. Some commonly encountered sets have
on cardinal numbers and the theory of infinite sets can
special symbols attached to their cardinal number. For ex-
be found in his monograph listed in the Supplemental
ample 0 = 0, N, = n, N = K. and [0. 1) = c. Readings. A more modern treatment. including a proof of
An ordering can be defined on cardinal numbers as the Schrdder-Bemstein theorem, is available in the text by
follows: If A and B are sets, then A B provided that A Halmos.
MISCELLANEOUS EXERCISES
1. Let A and B be nonempty sets. For a E A, b E B. define the ordered pair (a. b) by
(a, b) _ {{a}, {a, b}}..
Prove that two ordered pairs (a, b) and (c, d) are equal if and only if a = c and b = d.
The following-two exercises are detailed and lengthy. The first is a sketch of the proof of Theorem 1.5.3. The second
shows how the least upper bound property may be used to define the exponential function b', b > 1.
2. LetE= {t E R:t> 0 and t' <x).
a. Show that E # ¢ by showing that x/(x + 1) E E.
b. Show that I. + x is an upper bound of E.
Let y = sup E. The remaining parts of the exercise are to show that y' = x. This will be accomplished by show-
ing that y" < x and y" > x lead to contradictions, leaving y" = x. To accomplish this, the following inequality
will prove useful. Suppose 0 < a < b, then
b"-a"=(b-a)(b"-'+ab" +... '+ a`)<n(b-a)b"''.
c. Show that the assumption y" < x contradicts y = sup E as follows: Choose 0 < h x also leads to 'a contradiction of the definition of y as follows: Set
k=yy"_x
Show that if tat y-k,then t44;E.
3. Fix b > 1.
a. Suppose m, n, p, q are integers with n > 0 and q > 0. If m/n = p/q, prove that
(hl" =
-Thus if r is rational, b' is well defined. '
b. If r, s are rational, prove that b"' = b' b'.

c. If x E R, let B(x) = {b': r E 0. t x}. Prove that b' = sup B(r) when r E 0. Thus it now makes sense to de-
fine b' = sup B(x) when x E R.
d. Prove that b"i" = bxb'' for all real numbers x,
The following two exercises provide a detailed development of the field of complex numbers.
4. Definition. A complex number is an ordered pair (a. b) of real numbers. If ;, = (a, b) and w _ (c, d), we write z = w
if and only if a = c and b = d. For complex number z and w we define addition and multiplication as follows:
z+w=(a+ c, b + d)
zw (or - bd, ad + bc).
The set of ordered pairs (a, b) of real numbers with the above operations of addition and multiplication is denoted
by C.
a. Find elements 0 and 1 in C such that 0 + z = z and lz = z for all z E C.
b. Show that if z = (a, b), then -z - (-a,-b) is the additive inverse of z.
c. For z E C with z * 0, find the multiplicative inverse z-'.
d. Prove that the set of complex numbers C with addition and multiplication as defined is a field.
e. Set i = (0, 1). Show that i2 = -1.
f. Show that every complex number z can be written as z = a + bi where a. b E R. The real numbers a and b are
called the real part and the imaginary part of z, respectively. We write a = Re(z) and b = Im(z).
g. Prove that C is not an ordered field.
5. Definition. If z = a + bi E C. then the complex number i = a - bi is called the conjugate of z. The absolute
value of z, denoted IzI, is defined by IzI = a2 +
a. Prove each of the following.
(1) i z+
(ii) zw = z N
(iii) z + i = 2 Re(z), z - z = 2i Im(z)
(iv) zz = 1z12
b. Prove each of the following.
(1) IZI = IzI
(ii) Izwl = IzIIwI
(iii) IRe(z)I IzI, IIm(z)I s IzI
(iv) Iz + w12 = Izl2 + Iwl2 + 2 Re(zw)
(v) Iz - w12 = 1z12 + Iwl2 - 2 Re(zw)
(vi) Iz + wl IZI + Iw'I
The following result, known as the Schroder-Bernstein theorem, is nontrivial, but very- important. It is included as an
exercise to motivate further thought and additional studies. A proof of the result can be found in the text b) Halmos
listed in the Supplemental Reading.
6. Let X and Y be infinite sets. If X is equivalent to a subset of Y. and Y is equivalent to a subset of X. prove that X is
equivalent to Y.
7. As in Theorem 1.7.18, let A denote the set of all sequences of 0's and I's. Use the previous result to prove that
A-[0,1].
46 Chapter i The Real Number System
SUPPLEMENTAL READING
Buck, R. C., "Mathematical induction and recursive defi- GOdel, Kurt, "What is Cantor's continuum problem?"
nition:' Amer. Math. Monthly 70 (1963),128-135. Amer. Math. Monthly 54 (1947), 515-525.
Burrill, Claude W., Foundations of Real Numbers. Halmos, Paul. Naive Set Theory, Springer-Verlag. New
McGraw-Hill, Inc., New York, 1967. York. Heidelberg. Berlin, 1974.
Cantor, Georg, Contributions to the Founding of the The- Richman, F. "Is 0.999... = 0- Math. Mag. 72
ory of Transfrnite Numbers (translated by Philip E. (1999), 396-400.
B. Jourdain), Open Court Publ. Co., Chicago and Shrader-Frechette. M., "Complementary rational num-
London, 1915. bers," Math. Mag. 51(1978), 90-98.
Dauben, Joseph W., Georg Cantor: His Mathematics Spooner, George and Mentzer, Richard, Introduction to
and Philosophy of the Infinite, Princeton University Number Systems, Prentice-Hall, Inc., Englewood
Press, Princeton, N.J., 1979. Cliffs, N.J., 1968.
Sequences of Real Numbers
2.1 Convergent Sequences
2.2 Limit Theorems
2.3 Monotone Sequences
2.4 Subsequences and the Bolzano-Weierstrass Theorem
2.5 Limit Superior and Inferior of a Sequence
2.6 Cauchy Sequences
2.7 Series of Real Numbers
In our study of sequences of real numbers we encounter our first serious introduction to
the limit process. The notion of convergence of a sequence dates back to the early nine-
teenth century and the work of Bolzano (1817) and Cauchy (1821). Some of the con-
cepts and results included in this chapter have undoubtedly been encountered previ-
ously in the study of calculus. Our presentation, however, will be considerably more
rigorous-emphasizing proofs rather than computations.
We begin the chapter by introducing the notion of convergence of a sequence of real
numbers and by proving the standard limit theorems for sequences normally encountered
in calculus. In Section 2.3 we will use the least upper bound property of 1q to prove that
every bounded monotone sequence of real numbers converges in R. The study of subse-
quences and subsequential limits will be the topic of Section 2.4. In this section we also
prove the well-known result of Bolzano and Weierstrass that every bounded sequence of
real numbers has a convergent subsequence. This result will then be used to provide a short
proof of the fact that every Cauchy sequence of real numbers converges. Although the
study of series of real numbers is the main topic of Chapter 7, some knowledge of series
will be required in the construction of certain examples in Chapters 4 and 6. For this rea-
son, we include a brief introduction to series as the last section of this chapter.
Even though our emphasis in this chapter is on sequences of real numbers, in sub-
sequent chapters we will also encounter sequences of functions and convergence of se-
quences in normed linear spaces. A good understanding of sequences of real numbers
will prove very helpful in providing insight into properties of sequences in more gen-
eral settings.
47
48 Chaprer2 Sequences of Real Numbers
2.11 Convergent Sequences

Before we begin our study of sequences we first introduce the absolute value of a real
number.
2.1.1 DEFINITION For a real number x. the absolute value of x, denoted ix . is de-
fined by
V. ifx > 0,
1x1 = -.r, ifx - 0.
For example, 141 = 4 and 1-5I = 5. From the definition, Ixl ? 0 for all x E R.
and 1x1 = 0 if and only if x = 0. This last statement follows from the fact that if x * 0.
then -x * 0 and thus 1x1 > 0. The following theorem, the proof of which is left to the
exercises (Exercise 1), summarizes several well-known properties of absolute value.
2.1.2 THEOREM
(a) I-xI = Ix1 for all x E R.
(b) lxvI = I.xI IyI for all x. Y E R.
(c) IxI = \ for all x E R.
(d) If r > 0. then lxl < r if and only if -r < x < r.
(e) - Ixl s x < Ixl for all x E R.
The following inequality is very important and will be used frequently throughout
the text.
2.1.3 THEOREM (Triangle Inequality) For all.r, y E R, we have

Ix +v1l< lx1+h'I.
Proof. The triangle inequality is easily proved as follows: For x,.%- E R.
0:!:_ (x+v)-=x-+2xv+v-
Ixl' + 21x1 hvi + lyl' = (Ixl + w;)=.
Thus by Theorem 2.1.2(c),
Ix+yl= (x+y)':S (L,I+Iv1)'-1xl+i1.
As a consequence of the triangle inequality, we obtain the following two useful in-
equalities.
2.11A COROLLARY For all x, v, E IL we have

(a) Ix - yl s Ix - zl + j:.-yl,and
(b) IIx I - ly11 Ix - yl.
Proof. We provide the proof of (a), leaving the proof of (b) as an exercise (Exercise
2). If x, y, z E 08, then by the triangle inequality,
Ix - yl = I(x-z)+(z- y)I S Ix - zI + Iz - yI. I]

The following example illustrates how properties of absolute value can be used to
solve inequalities.
2.1.5 EXAMPLE Determine the set of all real numbers x that satisfy the inequality
12x + 41 < 8. By Theorem 2.1.2(d), 12x + 41 < 8 if and only if -8 < 2x + 4 < 8.
or equivalently, -12 < 2x < 4. Thus the given inequality is satisfied by a real number
x if and only if -6 < x < 2.
Geometrically, 1x1 represents the distance from x to the origin 0. More generally,
for x, y E 1i8, the euclidean distance d(x, y) between x and y is defined by
d(x,y)= Ix - yl
For example, d(-1, 3) = I-1 - (3)1 = 1-41 = 4, and d(5, -2) = 15 - (-2)1 = 7.
The distance .d, may be regarded as a function on R X R which satisfies the following
properties: d(x, y) _,- 0, d(x, y) = 0 if and only if x = y, d(x, y) = d(y, x), and
d(x, y) s d(x, z) + d(z, y)
for all x, y, z E R. This last inequality, also referred to as the triangle inequality, fol-
lows from Corollary 2.1.4(a).
Neighborhood of a Point
The study of the convergence of a sequence or the limit of a function requires the no-
tion of one real number being "close to" another. Since the euclidean distance between
two points a and x is given by d(a, x) = j a - xl, saying that x is "close to" a is equiv-
alent to saying that the distance lx - aI between them is "small" A convenient method
for expressing this idea is with the concept of an e-neighborhood of a point. This con-
cept will prove useful not only in the study of the limit of a sequence but also in our
study of the limit of a function and the structure of point sets in R.
2.1.6 DEFINITION Let p E R and let e > 0. The set

N1(p)_{xER:Ix -pl <e}
is called an e-neighborhood of the point p.
Whenever we use the term neighborhood, we will always mean an e-neighborhood
with e > 0. By Theorem 2.1.2(d), for fixed p E R and E > 0,
N,(p)={x:p-E<x<p+e}.
The set N,(p) is the open interval (p - e, p + e) centered at p with radius e. This is il-
lustrated in Figure 2.1.
50 Chapter2 Sequences of Real Numbers
p-E p p+E
Figure 2.1 N.(p)
Convergence of a Sequence
Recall from Definition 1.7.5 that a sequence in R is a function f :N -i R. For each
n E N, p = f(n) is called the nth term of the sequence f, and for convenience, the se-
quence f is denoted by I or simply
2.1.7 DEFINITION A sequence {p.)' t in R is said to converge if there exists a point

p E R such that for every e > 0, there exists a positive integer no such that p,, E Ni(p)
for all n a n,. If this is the case, we say that { p is the limit
of the sequence (p.), and we write
yip" = p or p" _+p.

if { does not converge, then {p,,) is said to diverge.'
In the definition, the statement p E N,(p) for all n at n, is equivalent to
1p - p l < e for all n ? n,.
As a general rule, the integer n,, will depend on the given e. This will be illustrated in
the following examples.
2.1.8 EXAMPLES
(a) For our first example we show that the sequence t converges to 0 in R. The
proof of this is the remark following Theorem 1.5.1; namely, given e > 0, there exists
a positive integer n, such that n,e > 1. Thus for all n _> no,
H_o <e.
Therefore 1im,, = 0. In this example, the integer n, must be chosen so that
n "o
n, > lie.
(b) If p E R, the sequence defined by p = p for all n E N is called the constant
sequence p. Since Jp - pl = 0 for all n E 101, we have lim p = p.
(e) Consider the sequence

1E,0_
2n + 2 We will show nthat
**0
I.
lim 3n+2=3.
1. The negation of the definition of convergence of a sequence is included in Section A.4 of the Appendix.
Since
2n+ l 2 l 1
3n + 2 3 3(3n + 2) < 9n'

given e > 0, choose n E N such that n > 1/9e. Then for all n ? no,
2n+ 1 2
< e.
I 3n + 2 3
Thus the given sequence converges to 3.

(d) The sequence { Ip.+11
- (- diverges in R. To prove this, we first note that for
this sequence, j ph - = 2 for all n. Suppose p -+ p for some p E R. Let
0 < e < 1. Then by the definition of convergence, there exists an integer n, such that
Ip - pj < efor all n?n0.But if n-a no,then
2=1pn-pn+11 _,; 1p0-pI+Ip-p,,+ij< 2e<2.
This, however, is a contradiction. Thus our assumption that the sequence converges is false;
i.e., the sequence diverges.
(e) As our final example we prove that li(Vn+ I - Vn) = 0. We first note that
( n+1-y) (Vn + I +V)

1 ( n+1+\)
I
n+1+Vn-
1
2Vn-
Given e > 0, we want to choose no E NI such that 1/(2\) < e for all n y n,. This
is easily verified to be the case if n, E N is chosen so that n, a 1/4E2. With this choice
of n, we now have I n + 1 - VnI < e for all it at n,.
2.1.9 DEFINITION A sequence { in R is said to be bounded if there exists a positive

constant M such that M for all it E N.
This definition is equivalent to saying that the range {p : it E N} of the sequence
{p,} is a bounded subset of R.
2.1.10 THEOREM
(a) If a sequence in ir'8 converges, then its limit is unique.
(b) Every convergent sequence in R is bounded.
Proof. (a) To prove (a) we use the method of proof by contradiction. Thus we as-
sume that the sequence converges but that its limit is not unique. So suppose the se-
quence {p"} converges to two distinct points p,q E R. Let e = 11p - q1. Since
p" - * p, there exists an integer n, such that I p" - p I < e for all n ? n1. Also, since
p" - q, there exists an integer n2 such that I p. - qI < e for all is ? n2. Thus if n
max{ni, n2}, by the triangle inequality (Corollary 2.1.4(a)),
2
Ip - qI s Ip" - pi + Ipn - qI < 2e=-glp-qI
The inequality Ip - qI <'Ip - qI however gives a contradiction. Thus the limit must
be unique.
(b) Let {p"} be a convergent sequence in 11 that converges top E R. Take e = 1.
For this e, there exists an integer na such that I p" - pI < 1 for all n > no. Thus by
Corollary 2.1.4(b), I p" I no. Let
M = max{Ip,l. .Ip".I. Ipl + 1}.
Then Ip"I s M for all n E N. Therefore the sequence is bounded. O
2.1.11 EXAMPLES
(a) According to the previous theorem, every convergent sequence is bounded. The
converse, however, is false. The sequence (I - (-1)"}"= I is bounded, but by Exam-
ple 2.1.8(d), the sequence does not converge. The sequence is bounded since Ip"I =
II -(-1)"1:5 2forallnEN.
(b) The sequence {n(- I r} is not bounded in R, and thus cannot converge.
EXERCISES 2.1
2. *Prove Corollary 2.1.4(b).
3. Prove that for xi...._x. E R, Ix, + + X.1 Sri I + + Ix" I
4. If x, y E R, show that (x + yl = Ixl + lyl if and only if xy ? 0.
5. If a, b e R, prove that labs s 21(a2 + b2).
6. Determine all x E R that satisfy each of the following inequalities.
"a.13x-21<-I1 b. Ix2-4I<5 'c.lxl+Ix-11<3 d. Ix-li<Ix+l1
7. Determine and sketch the set of ordered pairs (x, y) in R X R that satisfies the following.
a. 4x1=II b.IxI:IyI C.Ixyl!5 2
8. For each of the following sequences, prove, using an e, no argument that the sequence converges to the given limit
p; that is, given e > 0, determine n such that I p" - pI < e for all n ;' no.
13n ++5 _ 3 b2n+5 _ 1 n2+ I
'a. P 2\
6n-3 p 3 c 2n2
2n 7 ' 2 'P
j1-(- rJ,p=1
/1
( 1
M'p=0 f. In 1 +R 1 J},p=2.
l n
9. Show that each of the following sequences diverges in R.
*a {n(1 + (-1)")} b. +*c. ! ll

sin}
d. j .sin
2 e. l(+)nl J
10. If b > 0, prove that lim 1 = 0.

n-- 1 + nb
11. *a. If b > 1, prove that lim b = 0.
b. If 0 <- b < I, prove that nix
lim b" = 0.
12. *Let {a,} be a sequence in Il withnix
lim a" = a. Prove that lim a,
13. Let {a.} be a sequence in R withn-+oc
lim a, = a. Prove that lim
n xq3, = a3.
14. *If an a 0 for all n and lim an = a, prove that lim \ _ V.
15. Prove that if {a,} converges to a, then {la,I} converges to ]a]. Is the converse true?
16. Let {a.} be a sequence in P with lim an = a. If a > 0, prove that there exists n,, E N such that an > 0 for all
nano.
17. Let {a,} be a sequence in P satisfying Ia. - a.. I ? c for some c > 0 and all n E N. Prove that the sequence
{a,} diverges.
2.2 Limit Theorems

In this section we will emphasize some of the important properties of sequences of real
numbers and investigate the limits of several basic sequences that are frequently en-
countered in the study of analysis. Our first result involves algebraic operations on con-
vergent sequences.
2.2.1 THEOREM If {a.} and are convergent sequences of real numbers with
lima.=a
n- oo
and limb,=b,
n- 00
then
lira (a, + b,) = a + b, and
(a) n_00
(b) ltm
.-000
a,b, = ab.
L. = b.
(c) Furthermore, if a * 0, and a. * O for all n, then lint
n--an a
Proof. The proof of (a) is left to the exercises (Exercise 1). To prove (b), we add and
subtract the term a,b to obtain
la,b. - abl = I(a,b, - a,b) + (a,b - ab)l
c la,llb. - bl + Iblla. - al.
Since {an} converges, by Theorem 2.1.10(b), {an} is bounded. Thus there exists a con-
stant M > 0 such that Ia.I <- M for all n. Therefore,
lanbn - abl s Mlbn - bl + IbI Ian - at.
Let e > 0 be given. Since an -+a, there exists a positive integer n, such that
E
In - al <
n 2(lbl + 1)
for all n a n1. Also, since bn -+ b, there exists a positive integer n_ such that
Ibn-bt<2mf
for all n at n2. Thus if n a max{n1, n2},
abl < M(2M

+ IbIC2(IbI+ 1)) < E.
Therefore lira anbn = ab.
To prove (c) it suffices to show thatA00
lira I/an = 1/a. The result (c) then follows
from (b). Since a # 0 and an a, there exists a positive integer n, such that
21
Ian - al < 21aI
for all n a n,. Also, since
IaISIa - and+lal <2ial+lanl
for n a no, we have
Ianl z Ilai
for all n a n,. Therefore,

1 Ia - a.I 2
< Ian - al.
Ian a I Ianl Iai Iaj2
Let e > 0 be given. Since an -+a, we can choose an integer n, a n, so that
Ian - al < E 2-
for all n a n1. Therefore,
a_aI<E
I
for all n >_ n1, and as a consequence,
lim
I
n-+oo a, - a'
--
2.2.2 COROLLARY If {a,,} is a convergent sequence of real numbers with lim an = a, then
for any c E R, n"°`
(a) lim (an + c) = a + c, and
(b) Inn
n-ow
can = ca.
Proof. If we define the sequence {c,} by c, = c for all n E N, then the conclusions
follow by (a) and (b) of the previous theorem.
2.2.3 THEOREM Let {a,,} and {b,,} be sequences of real numbers. If {b,,} is bounded and
lima,, = 0, then
lim anbn = 0.
n-.oo
Proof. Exercise 3.
Remark. Since the sequence {b,} may not converge, Theorem 2.2.1(c) does not ap-
ply. The fact that the sequence {b,,} is bounded is crucial. For example, consider the se-
quences {,1,} and {3n}.
2.2.4 THEOREM Suppose {a,,}, {b,,}, and {c,,} are sequences of real numbers for which
there exists n, E N such that
an<bn5c, for all nEN,n?n
and that lint an = lim c. = L. Then the sequence {b,} converges and
n-roo n-+oo
limb,,=L.
n--.W
Proof. Exercise 4.
This result, commonly called the "squeeze theorem," is very useful in applications.
Quite often to show that a given sequence {a,,} in R converges to a, we will first prove
that
la, - al S Mb,
for some positive constant M and a nonnegative sequence {b,,} with lim b, = 0. (See,
for example, the proof of Theorem 2.2.1(c).) At this point we can uselheorem 2.2.4
to conclude that the sequence {(a,, - a1} converges to zero, or equivalently, that
lim an = a.
now
56 Chapter 2 Sequences of Real Numbers
This can also be proved directly. SinceM-OGlim b" = 0, given e > 0. there exists
n,, E IN such that 0 S b" < e/(M + 1) for all /n ? no. Thus for n ? n,,,
Ia"-aI5Mb"<M`M+ 1)<E.
Therefore lim a" = a.
Some Special Sequences

We next consider some special sequences of real numbers that occur frequently in the
study of analysis. For the proof of Theorem 2.2.6 we require the following result.
2.2.5 THEOREM (Binomial Theorem) For a E l8, n E N,
(1+a)"= k±(k)a&= (0) +(n)a+(2)a2+ +(n)a"

where
n _ n!
k k!(W:: k)!
is the binomial coefficient.
As in Section 1.3, for n E N, n! (read n factorial) is defined by
2-1,
with the usual convention that 0! = 1. Since the binomial theorem is ancillary to our
main topic of discussion, we leave the proof, using mathematical induction, to the ex-
ercises (Exercise 13). An alternative proof using Taylor ser ies will be provided in Sec-
tion 8.7.
2.2.6 THEOREM
(a) If p > 0, then limo 0.
(b) If p > 0, then lim = 1.

(c) lim " n = 1.
lint no
(d) If p > I and a is real, then - = 0.
P
(e) If IPI < 1, then lim p" = 0.
(f) For all p E R, lim nt = 0.
Proof. The proofs of (a) and (b) are left to the exercises (Exercise 5). The proof
of (a) is straightforward and the proof of (b) (for p > 1) is similar to the proof
of (c). For the proof of (c), let x" _ '"In-n - 1. Since x" is positive, by the binomial
theorem
.(nl n(n2- 1)
n = (1 + x")" /f xn =
for all n ? 2. Therefore, xn 5 2/(n - 1) for all n ? 2, and as a consequence,
Thus by (a) and Theorem 2.2.4, lim x" = 0, from which the result follows.
"
(d) Let k be a positive integer so that k > a. Since p > 1, write p = (I + q), with
q > 0. By the binomial theorem, for n > 2k,
P"=0 +q)"> kn qk= k!

qk.
Since k < in, n - k + 1 > in + 1 > in. Therefore,

n(n- 1)' '(n-k+ 1) nk
kt 2kk!'
and as a consequence,
n°
05 c MnF"
k-"
The result now follows by part (a) and Theorem 2.2.4.
(e) Write p as p = ±1/q, where q > 1. Then,
IP"I = IPA" =
4"
which by (d) (with a = 0) converges to 0 as n -- oo.
(f) Fix k E N such that k > dpi. For n > k,
IPr < klk-t) IPI "
n! - n! (k - 1)! ( k }
Since ip(/k < 1, the result follows by (e).
2.2.7 , EXAMPLES We now provide several examples to illustrate the previous theorems.
f 2n + 1
(a) As in Example 2.1.8(c) consider the sequence' } We write
in + 2
2n+1_n(2+'-) 2+-',
3n + 2 3+
1
Since lim = lam 2 = 0, by Corollary 2.2.2(a),
n-"o n n-.oo n
lim2 + 1n I = 2 and 1im3 + ? I = 3.

R"00(
n--"O( n
Therefore by Theorem 2.2.1,
(2+n?)-_1
3+n
oo\2+nJ3+?3
/
)=2-1=2.
1 1
(b) Consider the sequence 12N/n1. We first note that

+7
(-1)" -
05
I2\+7 2\
I 1
Thus by Theorem 2.2.4 and Theorem 2.2.6(a) with p = 2,
On
0.
"-Cc 2 n+
n 3l
(c) For our next example we consider the sequence +n} As in (a), we first want
3'
to factor out the dominant power in both the numerator and the denominator. By
Theorem 2.2.6(d), lim n°/p" = 0 for any a E R and p > 1. This simply states that
p" (p > 1) grows taster than any power of n. Therefore the dominant terms in the nu-
merator and denominator are 2" and 3", respectively. Thus
2"+n3 _ 2"(1 +29') _( 2 (1 +Y)

3n + n2 3n(1 + 3,) (I +
3 y
By Theorem 2.2.1 and Theorem 2.2.6(d),
"' 1+
11+F
3
1.
Finally, since lim ()" = 0 (Theorem 2.2.6(e)), we have

n-4W
3
n-. 2n +n2=0.
3"
(d) As our final example we consider the sequence {n((1 + n)'2 - 1)}. Before we can
evaluate the limit of this sequence we must first simplify the nth term of the sequence.
This is accomplished as follows:
1)=n((l }!)2 - 1)=n((n+1)2- 1)

2n - 1 _ -2n2 - n
(n + 1)2 F n+ l)2
Now we can factor out an n2 from both the numerator and the denominator. This gives
-2 -
X, (1 + i)2.
"
Using the limit theorems we now conclude that lim x" = -2.
EXERCISES 2.2
1. Prove Theorem 2.2.1(a).
2. Let {a"} and {b"} be sequences of real numbers.

a. If {a"} and {a" + b"} both converge, prove that the sequence {b"} converges.
b. Suppose b" * 0 for all n E N. If {b"} and both converge, prove that the sequence {a"} also converges.
5. a. If p > 0, prove that lim

1=
v
0.
n
*b. If p > 0, prove that lim = 1.
6. Find the limit of each of the following sequences.
(3n2+2n+l)°°
5n2 - 2n + 3 "..
( n °°
Sc. 5 n
L{ n2+n-n}.,
{\ 11+" - i)E
g. +a - h. {(2" + 3)""r..,
7. For each of the following sequences, determine whether the given sequence converges or diverges. If the sequence
converges, find its limit; if it diverges, explain why.
I + (-1)"1°0 1
-sin -}
-
(( n "-1 n 2
Sc. 1n2\2n+3)2 } 00 d. "

l 2"+n2J.i
a8n3 +51ao In cos nhr 00
e. fV9W L
2n+3 }
8. -Prove that "lim R cos n = 0.

2
9. Let {x"} be a sequence in R with x" - 0. and x" * 0 for all n. Prove that lim x" sin = 0.
z
10. Let {a"} be a sequence of positive real numbers such that
lim
.-. a" =L.
*a. If L < 1, prove that the sequence {a"} converges and that lim a = 0.
RtiOG
b. If L > 1, prove that the sequence {a"} is unbounded.
c. Give an example of a convergent sequence {a"} of positive real numbers for which L = 1.
d. Give an example of a divergent sequence {a"} of positive real numbers for which L = 1.
11. Use the previous exercise to determine convergence or divergence of each of the following sequences.
*a. {n2a"},0 < a < 1 b. {}.o< I
12. *Suppose Sao

lim (a" - 1)/(a" + 1) = 0. Prove that lim a" = 1.
n--.oo
13. a. For n E I !s k s n, prove that
(k)+\k- I )
N,
(n
k l)*b.
Use mathematical induction to prove the binomial theorem (Theorem 2.2.5).
14. Let {ak}k i be a sequence in R. For each n E N, define
a,+ +a"
s" n
*a. Prove that if li-m ak = a, then lim s" = a.
b. Give an example of a sequence {ak} which diverges, but for which {s") converges.
2.3' Monotone Sequences

In this section we will briefly consider monotone sequences of real numbers. As we
will see, one of the advantages of such sequences is that they will either converge
in R or diverge to oo or -oo.
2.3.1 DEFINITION A sequence {a"},°_ 1 of real numbers is said to be:

(a) monotone Increasing (or nondecreasing) if a" s a,, + t for all n E N;
(b) monotone decreasing (or nonincreasing) if a" ? a"+ 1 for all n E N;
(c) monotone if it is either monotone increasing or monotone decreasing.
A sequence {a"} is strictly increasing if a" < a"+, for all n. Strictly decreasing is
defined similarly.
As a general rule, bounded sequences need not converge; e.g., {I - (-1)"}. For
monotone sequences, however, we have the following convergence result.
2.3.2 THEOREM If {a, is monotone and bounded, then {a"},` , converges.

Proof Suppose (an} is monotone increasing. Set
E=fan: n=1,2,...}.
Then E # ¢ and is bounded above. Let a = sup E. We now show that
lim a" = a.
n-ti00
Let e > 0 be given. Since a - E is not an upper bound of E, there exists a positive in-
teger n,, such that
a - E <an 5a.
This is illustrated in Figure 2.2. Since {an} is monotone increasing,
a - e < an<- a forallnatn0.
Thus an E NE (a) for all n ? no and therefore lim a" = a.
ano-I anu anyN
a-t a
Figure 2.2
Nested Intervals Property

As an application of the previous theorem we prove the following result, usually re-
ferred to as the nested intervals property. The term nested comes from the fact that the
sequence {I"} of intervals satisfies 1" D In+1 for all it E N (see Figure 2.3).
2.3.3 COROLLARY (Nested Intervals Pro)erty) If {In} , is a sequence of closed and
a, a2 an a,.«, b,,., b b2 b,
Figure 2.3
bounded intervals with In D In+, for all n E N, then

00
nI ,,*o.
Proof. Suppose In = [an, bn), a", b" E R, an 5 bn. Since 1n 3 /".," for all m ? 0,
an!5 an+m:s bnt,,,<_b,,,
for all n, m E N. Thus the sequence {a"} is monotone increasing and bounded above
by every b" m E N. Thus by the previous theorem, a = lim an exists with a s b", for
all m E N. Therefore a E In, for all m E N and thus "~'°
00
a E nm-l/,,,,
which proves the result. Q
Remark. Similarly, if b = lim bthen b E In for all n, and thus
[a, b] C 1 1n.
In fact, one can show that equality holds (Exercise 1).
2.3A EXAMPLES
(a) Our first example shows that the conclusion of Corollary 2.3.3 is false if the inter-
vals 1" are not closed. As in Example 1.7.11(b), for each n E N set 1" = (0, ). Then
1,, 3 1n+, for all n, but
00
n 1" = o.
The conclusion of Corollary 2.3.3 may also be false if the intervals 1" are unbounded
(Exercise 2).
(b) Consider the sequence {j?'}. , with 0 0, s" > 0 for all n E N. Thus {sn} is bounded below. Also,
"+'
The last inequality follows since s,, > 0 and 0 < p < 1. Thus the sequence {s"} is
monotone decreasing, bounded below, and hence by Theorem 2.3.2, is convergent. Let
s = lim
n-O oW
s". Then
s= lim
n-"O
sn+, = n-+oo
limp"+' =p lims"=ps.
Therefore s = ps or s(1 - p) = 0. But p # I and thus we must have s = 0.
(c) Let a, = 1 and for n ? I set an+i = 6(2a,, + 5). The first three terms of the se-
quence {a"} are a, = 1, a2 = 6, and a3 = y. Thus we suspect that the sequence {a"} is
monotone increasing and bounded above by 2. Since a, < 2, if we assume that a" <_ 2,
then
a"+,= I(2a,, +5)56(4+9)<2.

Thus by mathematical induction a, s 2 for all n E N. Likewise, since a, < a2, if our
induction hypothesis is an < then the result is true for n = 1. and
an+I = (tan + 5) < 6(2an+I + 5) = an+2.

6
Therefore the sequence {an} is monotone increasing, bounded above, and thus is con-
vergent. Let a = lim an. Then
nr00
a lima..,,
R-KO
6 (tan + 5) = (2a + 5).
aim 6
Solving the equation a = '-6(2a + 5) for a gives a = ;.
(d) Let a, = 1, and for n ? 1, set To investigate the convergence of the
sequence {an}. we will establish by induction that
1 a.<a,, 1<2,
for all n r= N. When n = 1, we have
1=a,<N/2 =a2<2.
Thus the statement is true for n = 1. Assume that it is true for n = k. Then
I S ak+t = V Lak < V 2ak+, = ak+2,
and
ak+2= VL[lk+, < V4=2.

Thus the sequence {an} is monotone increasing, bounded above by 2, and hence by The-
orem 2.3.2 is convergent. It is possible to prove directly that
sup{an:n= 1, 2, ...}=2.
However, if we let a = lim an, then
a = lim
n-"0
an+, = lim
w_00
= \1521
The last equality follows by Exercise 14 of Section 2.1. Therefore a is a solution of

a2 = 2a, which since a a 1, implies a = 2. N
Euler's Number e
2.3.5 EXAMPLE In this example we consider in detail the very important sequence
{tn}= 1, where for each n E N,
tn='1+nJn.
We will show that the sequence {t,} is monotone increasing and bounded above, and thus
has a limit. The standard notation for this limit is e (in honor of Leonhard Euler); i.e.,
11n
e= lim I+
n- n
By the binomial theorem
t"=C1+n)"
I n(n - 1) 1 n(n - 1) 1 1
+n n2+
+ (1)
n+ 1.2 1.2 n n"
For k = 1, . .. , n, the (k + 1)-st term on the right side is

n(n- 1). . .(n-k+ 1) 1
1.2...k nt'
which is equal to
n)(1 n)...(1 -k n' (2)
if we expand in the same way, we obtain n + 2 terms, and for k = 1, 2..

the (k + 1)-st term is
1)...(I
1 2 1 . .k (1 n+ 1)( n+ - n + 1),
which is greater than the corresponding term in (2). Thus t < r,, for all n. From (1)
we also obtain
4 + + 1+
1.2 1.2.3
+- 1 + 1
2 1+1+2+22+
+r---
which by the identity I + r + + r'-' _ (1 - r")/(1 - r), r # 1,
=1+ I <I+ .. !- . = 3 ..
1 -g i -g
Thus {t"} is bounded above by 3, and we can apply Theorem 2.3.2. Since t" s 3 for all
n e N we also have that e s 3. To five decimal places, e = 2.71828 . The num-
ber e is the base of the natural logarithm function which will be defined in Example
6.3.5 as a definite integral.
Infinite Limits
If a monotone increasing sequence {a"} is bounded above, then by Theorem 2.3.2 the
.sequence converges. If the sequence {a'} is not bounded above, then for each posi-
tive real number M there exists n E N such that a" ? M for all n ? n,,. Since the
real number M can be taken to be arbitrarily large, this is usually expressed by say-
ing that the sequence {a"} diverges to oo. We make this concept precise, not only for
monotone sequences, but for any sequence of real numbers, with the following def-
inition.
23 Monotone Sequences 65
2.3.6 DEFINITION Let {a,j be a sequence of real numbers. We say that {an} approaches
infinity, or that {an} diverges to oc, denoted a,, -+ oo, if for every positive real number
M, there exists an integer n,, E N such that
a > M for all n ? n,,.
We will also use the notation lim an = oo to denote that a - oc as n -+ oc. The
n"X
concept of a,, - -oo is defined similarly.
2.3.7 THEOREM If {a"} is monotone increasing and not bounded above. then a,, -> oc as
n--)I oo.
As a consequence of Theorems 2.3.2 and 2.3.7, every monotone increasing se-
quence {an} either converges to a real number (if the sequence is bounded above) or di-
verges to oo. In either case,
lim a,, = sup{an : n E N}.
Remarks. Although the definition of diverging to infinity is included in this section

on monotone sequences, this should not give the impression that Definition 2.3.6 is ap-
plicable only to such sequences. In the following we give an example of a sequence that
diverges to infinity but is not monotone. Also, it is important to remember that when we
say that a sequence converges, we mean that it converges to a real number.
2.3.8 EXAMPLE Consider the sequence {n(2 + (- I )")}. If n is even, then n(2 + (- I Y')
= 3n; if n is odd, then n(2 + (-1)") = n. In either case,
n(2+(-1)")ten,
and thus the sequence diverges to oo. The sequence, however, is clearly not monotone.
EXERCISES 2.3
1. Let 1n = [an, n E N, be closed and bounded intervals satisfying I,, D I,,. i for all n. Prove that
x
n 1n = [a, b],
"=1
where a = sup{an : n E N} and b = inf{b : n e NJ.
2. 'Show by example that the conclusion of Corollary 2.3.3 is false if the intervals 1,, with 1. D 1_ are not
bounded.
3. Show that each of the following sequences are monotone. Find a lower or upper bound if it exists; find the limit if
you can.
{ a+If)(a- J-)1,wherea>
t It
b. 1
d. {sn}, where sn = cos' + cos- 2+ + cos'

1117
2
4. Define the sequence {a,,} as follows: a, ', and a,r, = V + Ct".

a. Show that a" s 2 for all n.
b. Show that the sequence (a") is monotone increasing.
c. Find lim a".
5. *Let a, > 1, and for n E N, define a"_, = 2 - I/a". Show that the sequence {a") is monotone and bounded.
Find n-+x
lim a".
6. Let 0 < a < 1. Set t, = 2, and for n E N, set:,, = 2 - alt,. Show that the sequence {t") is monotone and
bounded. Find lim t". n
7. For each of the following, prove that the sequence (a") converges and find the limit.
a. a"+, = 1(2a" + 5), a, = 2 b. a"+, _ V, a, = 3
*c.a"+ti2a, a,=1 d.a"+,=V2.. + 3, a4
*e. a"_, = 3a" - 2, a, = 4 f. an+, = 3, a, = I
8. Set x, = a. where a > 0 and let x" x" + (1/x"). Determine if the sequence {x"} converges or diverges.
9. Let a > /0. Choose x, > Va. For n = 1, 2, 3, .... define
xn+1 = 21X,+SJ.
\
*a. Show that the sequence {x"} is monotone and bounded.
b. Prove that lim x" = %/a-.
nb.
c. Prove that 0 c x" - NF. s (x.
10. In Exercise 9, let a = 3 and x, = 2. Use part (c) to find x" such that Ix" - \/31 < 10-5.
11. Let A be a nonempty subset of R that is bounded above and let a = sup A. Show that there exists a monotone in-
creasing sequence {a"} in A such that a = I m a". Can the sequence (a"} be chosen to be strictly increasing?
12. Use Example 2.3.5 to find the limit of each of the following sequences.
- n) '
1 j(1
*a. lI + +
d.
*c. 1 jl l
+ 2n/3n}
n
Sn= +2+ 1 1
+n
Show that {s,} is monotone increasing but not bounded above.
14. For each n E N, let
I I
s"= + 72+ +
n
Show that the sequence {s"} is monotone increasing but not bounded above.
15. *For each n E N. let
S"= i2+22+.+n2.
1I 1
Show that the sequence {s"} is monotone increasing and bounded above by 2.
16. Let 0 < b < 1. For each n E N, let s" = I + b + b2 + + b". Prove that the sequence {sue} is monotone in-
creasing and bounded above. Find lim s".
17. Show that each of the following sequences diverges to 00.
n n
18. *Which of the sequences in the previous exercise are monotone? Explain your answer.
19. If a -+oc and (b^} converges in R, prove that (a, + diverges to oc.
20. If a, > 0 for all n E h and lim a^ = 0, prove that I la, -+ oc.
21. Suppose a, > a, > 0. For n ? 2, set a^ _ i = ; (a, + a^ 1). Prove that
a. {au, ;} is monotone decreasing. b. {a2k) is monotone increasing, and
c. {a^} converges.
22. Let {s,} be a bounded sequence of real numbers. For each n E N. let a^ and b^ be defined as follows:
a, = inf{sk : k ? n}, b" = sup{sk : k ? n}.
a. Prove that the sequences {a^} and (b,,) are monotone and bounded.
b. Prove that lim a^ = lim b" if and only if the sequence {s^} converges.
nix nix
23. *In Theorem 2.3.2 we used the supremum property of R to prove that every bounded monotone sequence con-
verges. Prove that the converse is also true; namely, if every bounded monotone sequence in l8 converges. then
every nonempty subset of R that is bounded above has a supremum in R.
24. *Use the nested intervals property to prove that [0, 1 ] is uncountable.
2.4 Subsequences and the

Bolzano-Weierstrass Theorem
In this section we will consider subsequences and subsequential limits of a given se-
quence of real numbers. One of the key results of the section is that every bounded
sequence of real numbers has a convergent subsequence. This result, known as the se-
quential version of the Bolzano-Weierstrass theorem, is one of the fundamental results
of real analysis.
2.4.1 DEFINITION Given a sequence { p^} in OB, consider a sequence {nk}k 1 of positive
integers such that n, < n, < n3 < . Then the sequence { p^,}k , is called a sub-
sequence of the sequence { p"}.
If the sequence {p,,} converges, its limit is called a subsequential limit of the se-
quence { p,}. Specifically, a point p E 18 is a subsequential limit of the sequence { p"}
if there exists a subsequence {p,,,} of {p,,} that converges to p. Also, we say that oo is
a subsequential limit of { p"} if there exists a subsequence { p,, } so that p^, -+oo as
k -+ oo. Similarly for -no.
2.4.2 EXAMPLES
(a) Consider the sequence {(I - (-1)")}. If n is even, then a, = 0, and if n is odd, then
a = 2. Thus 0 and 2 are subsequential limits of the given sequence. That these are the
only two subsequential limits is left to the exercises (Exercise 1).
(b) As our second example, consider the sequence {(-1 )" + }. Both I and - I are
subsequential limits. If n is even, i.e., n = 2k, then
an = au = l +
2k'
which converges to t. On the other hand, if n is odd, i.e., n = 2k + 1, then
1
a,, = a2k+ i = -1 +
2k + 1'
which converges to -1. This shows that -1 and I are subsequential limits. Suppose
{a",} is any subsequence of {a"}. If the sequence Ink) contains an infinite number of
both odd and even integers, then the subsequence {a",} cannot converge. (Why?) On the
other hand, if all but a finite number of the nk are even, then {a",} converges to 1. Sim-
ilarly, if all but a finite number of the nk are odd, then {a",} converges to -1. Thus -1
and I are the only subsequential limits of {a"}.
(c) Consider the sequence {n(1 + If n is even, then n(1 + (-I )") = 2n,
whereas if n is odd, n(1 + (-1)") = 0. Thus 0 and oo are two subsequential limits of
the sequence. The same argument as in (b) proves that these are the only two subse-
quential limits.
Our first result assures us that for convergent sequences, every subsequence also
converges to the same limit.
2.4.3 THEOREM Let (p,,) be a sequence in R. If {p.1 converges to p, then every subse-
quence of { p"} also converges to p.
Proof. Let {p",} be a subsequence of {p"}, and let e > 0 be given. Since p"-+p,
there exists a positive integer n such that 1p. - pl < e for all n ? n0. Since {nk} is
strictly increasing, nk ? no for all k ? n". Therefore,
Ipn, - pI < e
forallk? n0;i.e.,p",-+p.
2.4A EXAMPLES
(a) In our first example we illustrate how Theorem 2.4.3 may be used to compute the
limit of a sequence. Consider the sequence {p"} where 0 < p < 1. By Example
2.3.4(b), the sequence {p"} is monotone decreasing, bounded below, and hence con-
verges. Let
a = lim p".
By Theorem 2.4.3 the subsequence {pen} also converges to a. But p2" = (p")2, and thus
a = lim p2" = lim (p")2 = a2.
Thus a2 = a. Since 0 a < 1, we must have a = 0.

2.4 Subsequences and the Bolzano-Weierstrass'fheorem 69
4 (b) In our second example we show how the previous theorem may be used to pros e
divergence of a given sequence. Consider the sequence { sin n07r}, where 8 is a ra-
tional number with 0 < 8 < 1. Write 0 = alb, with a, b E N and b ? 2. When
n = kb, k E N, then sin n8ir = sin kair = 0. Therefore 0 is a subsequential limit of
the sequence. On the other hand, if n = 2kb + 1, k E N, then
sin n0ar = sin (2kb + l) 6 ar = sin 12kalr + b ir)
= cos(2kaar) sin b ar = sin b ar.
Since 0 < alb < 1, sin bar * 0. Thus sin bar is another distinct subsequential limit
of {sin nBar}. Hence as a consequence of Theorem 2.4.3, the sequence { sin n07r}
diverges. The result is still true if 0 is irrational. The proof however is much more dif-
ficult.
Limit Point of a Set

In order to state the Bolzano-Weierstrass theorem we need to introduce the concept of
a limit point of a set. This concept is fundamental in the study of analysis and will oc-
cur again when we discuss the structure of point sets and limits of functions.
2.4.5 DEFINITION Let E be a subset of R.

(a) A point p E R is a limit point of E if every e-neighborhood N,(p) of p con-
rains a point q E E with q* p.
(b) A point p E E that is not a limit point of E is called an isolated point of E.
Remarks. In the definition of limit point it is not required that p is a point of E.

Also, a point p E E is an isolated point of E if there exists an e > 0 such that
N,(p)flE={p}.
2A.6 EXAMPLES
(a) E _ (a, b), a < b. Every point p, a 0, there exists a point x E (a, b) satisfying
p < x < p + e. These, however, are not the only limit points. Both a and b are also
limit points of E, but they do not belong to E.
(b) E _ {,'-, : n = 1, 2, ...}. Each n is an isolated point of E. If a is chosen so that
0<E 0, choose n E IN so that 1- < e. Such a choice of n is possible by Theorem 1.5.1.

Then
1 EN,(0)flE,
n
and thus 0 is a limit point of E.
(c) Let E =o fl [0, 1 ]. If p (4 [0, 1 ], then p is not a limit point of E. For if p > 1.
then for e = 1(p - 1) we have k(p) fl E = 46. Likewise when p < 0. On the other
hand, every p E [0, 1] is a limit point of E. Let e > 0 be given. Suppose first that
0 s p < 1. Then by Theorem 1.5.2 there exists r E Q such that
p<r<min{p+e.l}.
When p = 1, Theorem 1.5.2 also guarantees the existence of an r E Q fl [0, ii with
p - to < r < p. Thus for every e > 0, N,(p) contains a point r E E with r * p. The
same argument also proves that every point of R is a limit point of Q.
2.4.7 THEOREM Let E be a subset of 08.

(a) If p is a limit point of E, then every neighborhood of p contains infinitely
many points of E.
(b) If p is a limit point of E, then there exists a sequence { p
E N, such that
Proof.2 (a) Suppose there exists a neighborhood Na of p which contains only finitely
many points of E, say q1, ... , q, # p. Let
emin{lp-q1J... >lp-q.11.
Then N,(p) fl E contains at most p. Thus p is not a limit point of E.
(b) We construct the sequence in E as follows: Since p is a limit point of E,
for each positive integer n, by the definition of limit point, there exists p E E with
p # p such that I p - p I < ,1-,. The sequence {p.) clearly satisfies p --.>p. L)
2.4.8 COROLLARY A finite set has no limit points.
2.4.9 EXAMPLE In this example we illustrate part (b) of the previous theorem. Since
is a limit point of Q (Example 2.4.6(c)), by Theorem 2.4.7(b) there exists a sequence
of rational numbers with r -> N/-2. Such a sequence, however, need not be unique.
If r V, then the same is also true for the sequence {r + ,1-,}.
2. The proof of (a) is an example of a proof by contraposition. In this method of proof we assume the nepa-
tion of the conclusion and prove the negation of the hypothesis.
Bolzano-Weierstrass Theorem
We are now ready to state and prove the Bolzano-Weierstrass theorem. This impor-
tant result was originally proved by Bernhard Bolzano (1781-1848) and modified
slightly in the 1860s by Karl Weierstrass (1815-1897). The theorem is one of the
fundamental results of real analysis. Its importance will become evident in subse-
quent sections. Although not immediately obvious from the proof, the least upper
bound property plays a crucial role in the proof of the theorem. The least upper
bound property was used in proving the nested intervals property (2.3.3), and this re-
sult is crucial in the proof.
2A.10 THEOREM (Bolzano-Weierstrass) Every bounded infinite subset of R has a limit

point.
Proof. Let S be a bounded infinite subset of R. Since S is bounded, there exist
a, b E Il with a < b such that S C 11 = [a, b]. Divide 11 into two closed subin-
tervals
b1 fa+b b]
a,a 2 Jl 2 J
each of length (b - a)/2. At least one of these, call it 12, must contain infinitely many
points of S. If not, then S would be the union of two finite sets and thus be a finite set
itself. Repeating this process, we obtain a sequence {1"} of closed and bounded inter-
vals satisfying
(a) [a, b]=11DIZD.. .D1"D...,
(b) length of I" = (b - a)/2"-, and
(c) for each n, in fl s is infinite.
By Corollary 2.3.3, lono- 11, * 0. Let x e fly 1 In. It remains to be shown that x is a
limit point of S. If e > 0 is given, choose n E N so that (b - a)/2"-1 < e. If y E I",
then
ly - xj S length I" < e.
Therefore 1" C N.(x). Since In fl S is infinite, there exists y E S fl N1(x) with y * x.
Since e > 0 was arbitrary, x is a limit point of S.
The conclusion of the Bolzano-Weierstrass theorem may fail if either hypothesis is
removed. By Corollary 2.4.8 a finite set has no limit point. On the other hand, the set N
is an infinite unbounded subset of II with no limit points. The following corollary is of-
ten called the sequential version of the Bolzano-Weierstrass theorem.
2A.11 COROLLARY (Bolzano-Weierstrass) Every bounded sequence in R has a conver-

gent subsequence.
Proof.3 Let {p,,} be a bounded sequence in R, and let S = { p" : it = 1, 2, . . .}. If S

is finite, then there exists a point p E S and a sequence Ink.) with n, < n_ < such
that
Pn,=Pn_=.=P
The subsequence {pn,} obviously converges to p.

If S is infinite, then by the Bolzano-Weierstrass theorem, S has a limit point p E R.
Choose n1 so that I p - pn,I < 1 . Having chosen n1, ... , nkchoose an integer
nk > nk _ , so that
IP - Pn,I < kI
Such an integer nk exists since every neighborhood of p contains infinitely many points
of S. The sequence {p",}k , is a subsequence of (pn) that by construction converges
to p.
An argument similar to the one used in the previous Corollary may be used to prove
the following.
2.4.12 THEOREM Let { pn} be a sequence in R. If p is a limit point of { p" : n E K), then
there exists a subsequence {p.,) of { such that p,,, -+p as k -+ co.
Proof. Exercise It.
As an application of Theorem 2.4.12 we consider the following example.
2.4.13 EXAMPLE Since the set E = a fl [0, 1 ] is countable, there exists an enumeration
{rn : n E NJ of E. The sequence {rn}W , is called an enumeration of the rational numbers
in (0,1 ]. By Example 2.4.6(c), every p E [0, 1 ] is a limit point of {r" : n = 1, 2, ...}.
Thus if p E [0, 1 ], there exists a subsequence {rn } of {r,} such that rn, -+p. The se-
quence {rn} has the property that every p e [0, 1) is a subsequential limit of the se-
quence. This sequence also provides an example of a sequence for which the set of sub-
sequential limits of the sequence is uncountable.
EXERCISES 2.4
1. a. Prove that 0 and 2 .are the only subsequential limits of the sequence (1 - (-1n
b. Prove that 0 and oo are the only subsequenual limits of the sequence {n(1 +
2. a. Construct a sequence {sn} for which the subsequential limits are {-oo, -2. 1}.
b. Construct a sequence {s"} for which the set of subsequential limits of the sequence is countable.
3. The proof of Corollary 2.4.11 is an example of a proof by cases (see Appendix A.3). If S = { p, : n E101}.
then there are two cases; namely. S is finite or S is infinite. For each case we prove the existence of a con-
vergem subsequence of {p"}.
b. s cos
Slln
-
3. Find all the subsequential limits of each of the following sequences.
*a. s sin 2
2
l
c. Sll-I)" +12sin
l -I"
2 1
d.
l
sin?
*e. I--
n
f {(1.5
4. Use Example 2.3.5 to find the limit of each of the following sequences. Justify your answer.
*a.
I \ + 3n1J b. .CI + 2nI
j c_ 1(I + n/nj
5. Suppose p > 1. Use the method of Example 2.4.4 to show that lim = 1.
6. For n E Nl, set p" = n'I".
a. Prove that I < p, , < p" for all n ? 3.
b. Let p = limp". Use the fact that the subsequence {p,,,} of {p,,} also converges to p to conclude that p = 1.
7. Determine the limit points and the isolated points of each of the following sets.
"
((
*a. S1-
n
:nEN! b. -1)"+!:nE!i } 1
ll
c. (0. l) U {2} d. N
e. Q8\Q f. Qfl(0,1)
8. Let A be a nonempty subset of U8 that is bounded above and let a = sup A. If a E A. prove that a is a limit point
of A.
9. Let { p"} be a bounded sequence of real numbers and let p E R be such that every convergent subsequence of
converges to p. Prove that the sequence {p") converges to p.
10. a. Construct a subset of 68 with exactly two limit points.
b. Find an infinite subset of R with no limit points.
c. Construct a countable subset of U8 with countably many limit points.
d. Find a countable subset of R with uncountably many limit points.
12. Prove that every sequence in R has a monotone subsequence.
13. Use the Bolzano-Weierstrass theorem to prove the nested intervals property (Corollary 2.3.3).
14. *Use the nested intervals property to prove the least upper bound property of R.
15. Prove that every uncountable subset of R has a limit point in R.
25 Limit Superior and Inferior of a Sequence

In this section we define the limit superior and limit inferior of a sequence of real num-
bers. These two limit operations are important because unlike the limit of a sequence.
the limit superior and limit inferior of a sequence always exist. The concept of the limit
superior and limit inferior will also be important in our study of both series of real num-
bers and power series.
Let {s"} be a sequence in R. For each k E N, we define ak and bk as follows:

ak = inf{s" : n ? k},
bk = sup{s" : n ? k}.
Recall that for a nonempty subset E of R, sup E is the least upper bound of E if E is
bounded above, and co otherwise.
From the definition, ak s bk for all k. Furthermore, the sequences {ak} and {bk} sat-
isfy the following:
ak < ak.l and bk'- bk+l (3)
for all k. To prove (3), let Ek = {s" : n zz k}. Then Ek+, C E.. Therefore, if bk =
sup Ek, s" s bk for all n ? k. In particular
s" : bk for all n ? k + 1.
Therefore bk+, = sup Ek+, s bk. A similar argument will show that the sequence {ak}
is nondecreasing.
As a consequence of (3), the sequence {ak} is monotone increasing and the se-
quence {bk} is monotone decreasing. Thus by Theorems 2.3.2 and 2.3.7, these two se-
quences always have limits in i8 U {-oo, oo}.
2.5.1 DEFINITION Let {s"} be a sequence in R. The limit superior of {s"}, denoted lim s"
"~"
or lim s", is defined as
ligoo s" k
urn bk A sup{S" : n > k}.
The limit inferior of {s"}, denoted lim s" or lim s", is defined as
"-00
lim s" = lim
k-oo
ak = sup inf{s" : n at k}.
nToo kEN
We now give several examples for which we will compute the limit inferior and
limit superior. As will be evident, these computations are very tedious. An easier
method will be given in Theorem 2.5.7.
2.5.2 EXAMPLES
(a) {1 + (-1. Let s,, = 1 + (-1)". Then s" = 2 if n is even, 0 otherwise. Thus
ak = 0 for all k and bk = 2 for all k. Therefore
lim s" = 2 and lims"=0.
(b) {n(1 + (-,. In this example,
ifnisodd.
s"=n(1 +(-1)")={0
2. if n is even.
Set Ek = {s,,: n = k}. Then

Ek = {0, 2(k + 1), 0, 2(k + 3),. ..} if k is odd
Ek = (2k, 0, 2(k + 2), 0. 2(k + 4),. ..} if k is even.
Therefore ak = inf Ek = 0, and bk = sup Ek = oo. Thus

limsn=0 and ii-m- S, = 00.
n-x
n-+x
(c) {(-1)"+,,,}°,. Setsn= (-1)"+1.Then

-1+n n odd,
1+ n even.
n
The first few terms of the sequence {sn} are illustrated in Figure 2.4.
$5 S3 $3 S( 'i4 52
0 I 75 3
5 3 64 2
Figure 2.4
To compute the limit superior and inferior of the sequence {sn}, we set Ek =
{sn : n ? k}. If k is even, then
Ek1+k,-1+k+1' I+kI . 1.
Therefore, for k even,
bk = sup Ek = 1 + k and ak = inf Ek = - I.
Similarly, fork odd,

1
bk=supEk= 1+ k+ l and ak=infEk= -I.
As a consequence,
1 + k, k even,
a. _ -1 for all k. and b. _
1
1+ k odd.
k 1
Thus
lim and limsn=-1.

or-_+W
Our first theorem provides an (e, n0) characterization of the limit superior. An anal-
ogous characterization for the limit inferior is given in Theorem 2.5.4.
2.5.3 THEOREM Let {sn}' , be a sequence in R.

(a) Suppose lim sn E R. Then /3 = lim s if and only if for all e > 0,
(i) there exists n E N such that s < 0 + e for all n ? n0, and
(ii) given n E N, there exists k E N with k ? n such that sk > /3 - e.
(b) lim s = oo if and only if given M and n E N, there exists k E N with
n such that sk ? M.
(c) n-00
lim sn = -oo if and only if sn -+ -oo as n -4 oo.
Remark. The statement "s,, < (3 + e for all n a means that sn < /3 + e for all
but finitely many n. On the other hand, the statement "given n, there exists k E N with
k ? n such that sk > /3 - e" means that sn > /3 - e for infinitely many indices n.
2.5A THEOREM Let {sn} be a sequence in H.

(a) Suppose urn sn t R. Then a = lim sn if and only if for all E > 0,
n-4ou n+oc
(i) there exists no E N such that s > a - e for all n ? no, and
(ii) given n E N, there exists k E Ni with k ? n such that sk < a + e.
(b) lim sn = -oo if and only if given M and n E N, there exists k E N with
i_-"_Q
k ? n such that sk <_ M.
(c) lim s = oo if and only if s,, - oo as n --> oo.
Proof of Theorem 2.5.3. We will only prove (a). The proofs of (b) and (c) are left
to the exercises (Exercise 5).
(a) Suppose /3 = lim S. = rlim bk where
bk = sup{ s,, : n ? k}.

Let e > 0 be given. Since lim bk = (3, there exists a positive integer n,, such that
bk < /3 + e for all k >t n0. Since
k sn s bk for all n ? k,
s,, </3+e for allna, no.
This proves (1). Suppose n E N is given. Since bk-+13, and {bk} is monotone decreas-
ing, bk ? /3 for all k. In particular, bn ? /3. By the definition of b,,, however, given
e > 0, there exists an integer k z n such that
sk> b,, - e? /3 -e,
which proves (ii).
Conversely, assume that (i) and (ii) hold. Let e > 0 be given. By (i) there exists
n. E N such that s < /3 + e for all n a na. Therefore,
b,,,= sup{s,,:n?n0}:/3+e.
Since the sequence {bn} is monotone decreasing. bn <- 6 + E for all n ? n,,. Thus
Since c > 0 was arbitrary, urn sn S /3.

Suppose $3' = lim sn < (3. Choose e > 0 such that $3' < 6 - 2E. But then there
exists n such that
sn<13'+E<$3-E foralln?n0,
which contradicts (ii). Thus Tim sn = /3. p
To illustrate the previous two theorems, consider the sequence sn = (- I)" + 1/ of
Example 2.5.2(c). For this sequence, lim s" = 1 and lim s,, _ -1. Given e > 0, then
s"_ e. Since the odd terms get close to -1, we can never have the
existence of an integer n such that sn > 1 - E for all n ? no. On the other hand, given
any n E N, there exists an even integer k ? n such that Sk > I - E.
An immediate consequence of the previous two theorems is as follows.
2.5.5 COROLLARY Jim sn = lim s" if and only if Jim s" exists in R U {-oo, oo}.
n-+00 nix
n-M
Proof.. Suppose lim s" = lim sn = a E R. Let e > 0 be given. By (a) of the previ-
ous two theorems, there exist positive integers ni and n2, such that
s" < a + e for all n ? n,, and
sn > a - E for all n ? n,.
Thus if n,, = max{nt, n2},
a - e < sn<a+E
for all n ? no; i.e., n-.oo
lim sn = a. The proofs of the cases a = oo or a = -oo are similar.
If lim sn = a, then it easily follows that both Jim s,, = a and lim s, = a. O
2.5.6 THEOREM Let {a"} and {b"} be bounded sequences in R. Then

lim a,, + Jim bn S lim (an + bn) <_ lim a, + lim b"
n-.oo n-too n--.x w
<- nyOG
lim (an + bn) S ntiR
lim an + lim b,,.
Proof. Exercise 6. U
The following theorem relates the limit superior and inferior of a sequence to the
subsequential limits of the sequence, and is in fact very useful for finding Tim- S. and
lin s of a sequence {sn}.
2.5.7 THEOREM Let {sn}' 1 be a sequence in IR, and let

E = the set ofsubsequential limits of {sn} in R U {-oo. oo}.
n:
Then lim s, and lim s" are in E and
"x-100
(a) lim s" = sup E. and

(b) lim s" = inf E.
"-400
Proof. Let s = lim s". Suppose s E R. To show that s E E, we show the existence of
a subsequence {s"j of {s"} which converges to s. Take e = 1. Let n, be the smallest in-
teger such that
s - I <S". < S + 1.
Such an integer exists by (i) and (ii) of Theorem 2.5.3(a). Suppose the integers
nl < n2 < .. < nt have been chosen. Take e = 1/(k + 1). Let n1 be the smallest
integer greater than nk such that
$ -k+1<S",. <S+k+1
1 1
Again, such an integer exists by (i) and (ii) of Theorem 2.5.3(a). Then (sn j is a subse-
quence of {s"} which clearly converges to s. Therefore s E E. The case s = no is
treated similarly. Ifs = -oo, then by (c) of Theorem 2.5.3, sn -i -oo as n -* no.
Since s E E, s s. Suppose S # no. Then there exists a E E
such that
sass.
Since a E R, we can choose e > 0 such that s + e < a - e. For this e, there exists
n0 E N such that s" < s + e for all n z n0. Hence there can exist only finitely many k
such that
(sk - al < e.
Consequently no subsequence of {s"} can converge to a. This contradiction shows that
sup E = s. The case S = no is treated similarly. Q
2.5.8 EXAMPLES In the following examples we use Theorem 2.5.7 to compute lim s" and
lim s" for each of the given sequences {s"}.
(a) Let s,, = (- I)" + 1. By Example 2.4.2(b), the set of subsequential limits of {sn} is
{-1, I}. Thus by the previous theorem,
lim sn = -1 and lim s" = 1.
(b) Let sn = n(1 + (-1)"). By Example 2.4.2(c) the subsequential limits of {sn} are 0
and no. Therefore,
lim s" = 0 and lim s" = no.
(c) Let s" = sin" If n is even, i.e., n = 2k, then s,t = sin kir = 0. On the other
.
hand, if n is odd, i.e., it = 2k + 1, then s,+, = sin(2k + 1)121 _ (-1)r. Hence

the set of subsequential limits of the sequence {s"} is {-1, 0, 1}. As a conse-
quence,
lim s" = -1 and Urn s" = 1.
EXERCISES 2.5
1. Find the limit inferior and limit superior of each of the following sequences.
*a. } b. j(1 +(-Ir)sin

n + (-1)"n2) *d I - 2(-1)"n
n2+1 1 ll f 3n+2
e. 11+n(I
n
+ cos nir) l f, {[ 1.5 + (- I)"J"}
g. {cos n9w}, 9EO

2. Let {a"} be a sequence in R. If ii-m- la"I = 0. prove that lim a" = 0.
3. *Let Jr.) be an enumeration of the rational numbers in (0, 1). Find lint r" and urn r".
4. Let {s"} be a sequence in R. If s E H satisfies that for every e > 0, there exists no E N such that s" < s + e for
all n ? n prove that liras" s s.
5. a. Prove Theorem 2.5.3(b).
b. Prove Theorem 2.5.3(c).
6. Let {a"} and {b"} be bounded sequences in R.
*a. Prove that lim a" + limb"slim (a" + b") <_ Iim a" + limb,,.
b. Prove that lim (-a") = -Fun a..
c. Given an example to show that equality need not hold in (a).
7. a. If a" and b" are positive for all n, prove that lira (Tim a")(Iim b"), provided the product on the right is
not of the form 0 oo.
b. Need equality hold in (a)?
8. *Let s, = 0. For n E N, n > 1, let s" be defined by
S2m-I 1
S2m = , S2,"+I = 2 + s2,,,.
2
Find Iim s" and lim s".
a='.
9. Let a" > 0 for all n. Prove that Tim " a" lim
a"
10. *Suppose (a"}, {b"} are sequences of nonnegative real numbers with limo b, = b * 0 and lim a" = a. Prove that
lima"b"=ab.
n- =
2.6. Cauchy Sequences

In order to apply the definition to prove that a given sequence 1 1),, } converges, we must
know the limit of the sequence For this reason, theorems that provide sufficient
conditions for convergence, such as Theorem 2.3.2. are particularly useful. The draw-
back to Theorem 2.3.2 is that it applies only to monotone sequences of real numbers.
In this section we consider another criterion that. for sequences in R. is sufficient to en-
sure convergence of the sequence.
2.6.1 DEFINITION A sequence {p,),',-, in R is a Cauchy sequence if for every r: > 0.

there exists a positive integer it,, such that
IP,.-P.,I <E
for all integers n, m ? n,,.
Remark. In Definition 2.6.1, the criterion I p - p,,,l < e for all integers n. m ? it,,
is equivalent to
Inn-4 - P,.I < E
for all n ? it,, and all k E N. Thus if { is a Cauchy sequence in R.
PnI = 0 (4)
for every k E N. The converse, however, is false: namely, if { p,,} is a sequence in 6a that
satisfies lim 0 for every k E N. this does not imply that'the sequence
n-.x
is a Cauchy sequence (Exercise 4). The hypothesis (4) only implies that for each
k E N, given E > 0, there exists a positive integer n such that Ip,,.a - E for all
n at n,,,
2.6.2 THEOREM
(a) Every convergent sequence in R is a Cauchy sequence.
(b) Every Cauchy sequence is bounded.
Proof. (a) Suppose that {p,,} converges to p E R. Let e > 0 be given. Then for the
given e, there exists a positive integer n,, such that
l
In. - PI < e
for all n ? n,,. Thus by the triangle inequality, if it, m a n,,,
Ip.. P1 c!pn PI+IP pn.I<Ie+2E=E.

(b) Take e = 1. By the definition of Cauchy sequence, there exists n E N such
that I p,, - p0I < I for all n, m ? no. Therefore, with m = nby Corollary 2.1.4(b).
I PnI { 1 + IP I for all n ? n,,. Let
M = max{1 + Ipnj, Ipil,. . ., Ipn _,I}.
Then for all n, I pnI :SM. Thus {pn} is bounded. 0
2.6.3 THEOREM If { pn} is a Cauchy sequence in l that has a convergent subsequence,

then the sequence { converges.
Proof. Suppose { p,, } is a convergent subsequence of { pn} with lim pn, = p. Let
e > 0 be given. Since {pn} is Cauchy, there exists an integer ni such that
IPA - PmI <2e foralln,m?n,.

Since p,, -pp, for the given e, there exists an integer k, such that
IPn, - PI <2e forallkaki.

Let no = max{k,, n,}, and choose nk such that k at no. Then nk a n,. Thus if is >_ no, by
the triangle inequality,
IPn - pi < IPn - Pn,I + IPn, - PI < 2e + 2e = e.

Therefore lim p,, = p, which proves the result. 0
2.6.4 THEOREM Every Cauchy sequence of real numbers converges.

Proof. Let {p,,} be a Cauchy sequence in R. By Theorem 2.6.2, the sequence { pn} is
bounded. Thus by Corollary 2.4.11, the sequence {pn} has a convergent subsequence.
The result now follows by Theorem 2.6.3. U
Remark The statement that "every Cauchy sequence in R converges" is often ex-
pressed by saying that R is complete. Since the proof-of Theorem 2.6.4 used the
Bolzano-Weierstrass theorem, the completeness of R ultimately depends on the least
upper bound property of R. Conversely, if we assume completeness of K, then we can
prove that R satisfies the least upper bound property (Exercise 12). For this reason, the
least upper bound or supremum property of R is often called the completeness prop-
erty of R.
2.6.5 EXAMPLES
(a) For our first example we consider the sequence {sn} where for n E N,
1+
2+. .+'2.
.
2 n
For k E N,
(n+l)z+...+(n+k)z
(n -n+ 1) + +(n+k- 1 n+k/]

n n + k
In the above we have used the inequality

1 1
(n+m)2 n+m-1 n+m
valid for all n, m E N. Since the sequence {.1} converges, it is a Cauchy sequence.
Thus given e > 0, there exists n, E Nl such that I1, - kI < e for all n > no and all
"n
k E N. Thus IS. +k - S. I < e for all n ? no and all k E N. Therefore the sequence {s"}
is a Cauchy sequence.
(b) In this example we illustrate how the concept of a Cauchy sequence may be used
to prove convergence of a given sequence. Additional applications will be given in the
exercises. Let a,, a2 be arbitrary real numbers with a1 # a2. For n > 3, define an in-
ductively by
an = I(an-1 + an-2) (S)
Our first goal is to show that the sequence {a"} is Cauchy. We first note that
an+1-an=-2(an-an-1)
As a consequence, for n ? 2,
1 2"
an+1 - an (a2 - a1). (6)
This last statement is most easily verified by induction (Exercise 5). For m ? 1, con-
Ian+'n - anI
m:Ian+k+I
sider Ian+," - a,, 1. By the triangle inequality,
- an+k) S I
7,(an+k+I -
-1
-0
which by (6)
I 1 I m
1
Ia2 - a]I 7, 2n+k-f 2"-21a2 - all 7, 2k.
By Example 1.3.3(a)
rk r-r'* r # 1. (7)
kl 1-r
Thus with r = 2,
M 1
2 - 2)m+I l
= 1 - - < 1.
k=12k 1-Z 2m
Therefore,
1
Ian+m - and :S- 22 Ia2 - all (S)
2
for all n >- 2 and m E N. Let e > 0 be given. Choose no such that lag - a, I/2"-2 < e
for all n ? no. Then by (8),
Ian+m - and < E
for all m E N, n ? nn. This, however, is just another way of stating that
Ian - am < e for all m, n ano.
Therefore the sequence {an) is a Cauchy sequence in R, and thus by Theorem 2.6.4.
a = lima"
n- 00
exists in R.
Can we find the limit a here? If we follow the approach in Example 2.3.4(c), by
taking the limit of both sides of equation (5), we only get a = a. To find the value of a,
let us observe that
an+1 - a, = (an+, - an) + (an - a"-,) + ... + (a2 - a1)
n
Y, (ak+ I - ak),
k=1
then use (6) to get

I k-1
(a2-al)I(-2)
( n
The last equality follows from formula (7). Since an+, and (-3f -+0, upon tak-
ing the limit of both sides we obtain
2
a -a, = (a2 - a,) or a=a,+3(a2-a,).
Contractive Sequences
One of the key properties of the sequence {an} of the previous example was that
Ian+, - and < Ia,, - a.-,I

2
for all n ? 2. This property was used to show that the sequence {a"} was a Cauchy se-
quence and thus converged. Such sequences are commonly referred to as contractive
sequences. We make this precise in the following definition.
2.6.6 DEFINITION A sequence { p"} in R is contractive if there exists a real number

b, 0 < b < 1, such that
IPn+I - P"I t bIP" - Pn
forallnEN,na2.
If {p.) is a contractive sequence, then an argument similar to the one used in the
previous example shows that
IPn+I - P"I `- b"-'IP2 - P1I
for all n >_ 1, and that
IP"+m-P.1
Cb"-'IP2-PII(l+b+. +b'"-I)< 1-bIP,-PII
b"
for all n, m E N. As a consequence, every contractive sequence is a Cauchy sequence.

Therefore, every contractive sequence in R converges to a point in R. We summarize
this in the following theorem.
2.6.7 THEOREM Every contractive sequence in R converges in R. Furthermore. if the se-
I(b)
quence {p.1 is contractive and p = lim p", then
(a) IP - P"I l_b
P-P"::=:':r'
where 0 < b < 1 is the constant in Definition 2.6.6.
Proof. We leave the details of the proof to the exercises (Exercise 7). 0
2.6.8 EXAMPLE Suppose we are given that the polynomial p(x) = x2 - 3x + I has ex-
actly one zero in the open interval (0, 1). If c r= (0, 1) is such that p(c) = 0, then
c = (c2 + 1). We start with cl E (0, 1) arbitrary, and for n a- I we set
i
1 ,
c"+1=3(c;,+I).
Since cl E (0, 1) we have c2 E (0, 1), and by induction, c" E (0, 1) for all n E N. To
prove that the sequence {c"} converges we prove that it is contractive. For n > 2 we
have
Ic"+I-c"I=I3(c+1)-3(c,,-1+1)I
= 3I(Cn - C.-1)(c + Cn-1)I
2
:5 3ICn - Cn-1I
Thus the sequence is contractive with constant b = 3. If c = nlimc

max.
then c
(c2 + 1) or p(c) = 0.
Suppose we begin with c1 = .5 and we wish to determine the value of n such that
Ic - Cl < 10-3. By Theorem 2.6.7 (a) it suffices to determine n such that
bn-t
IC2 - ct I <
(b - 1)
By computation c2 = aZ. Thus since b = 3, we want to determine n such that
4(3)"-i
< 10-1.
3 12 2
Solving the above inequality for n - I gives
(n - 1) >
3 In 10 - In 4
n3 - In 2 = 13.62 (to two decimal places).
Thus the choice of n = 15 is sufficient to ensure accuracy to within 10-3. This estimate
on n however is very crude. In fact, after six steps, part (a) of Theorem 2.6.7 only gives
.Ic - c6I < 3(3)SIc2 - c1I = .032922. The actual approximation however is much bet-
ter than this. By computation we find that c3 = .391204, c4 = .384347, cs = .382574.
and c6 = .382121. By part (b) of Theorem 2.6.7, Ic - c6I 21c6 - csl = 000906,
and this is less than 10-3.
0 EXERCISES 2.6
1. If and are Cauchy sequences in R, prove (without using Theorem 2.6.4) that {a + and are
also Cauchy sequences.
2. For each of the following, determine whether the given sequence is a Cauchy sequence.
+l} b. {(-I)"} c.
(n+(-I}"}
-a. In in ll(( n
adI
. 1(1)'n {.1 + (-l)nn2}
l
1n2+3 2n2 +3
3. For n E 4 let
S. +21+31++n1.
Prove that is a Cauchy sequence.

4. Consider the sequence defined for n = 1, 2, 3, ... by
s=1+2+:..+1
*a. Show that {sn} is not a Cauchy sequence.

b. Even though {sn} is not a Cauchy sequence, show that I'M Isn.k - s,, = 0 for all k E N.
n+x
5. Use mathematical induction to prove identity (6).
6. Let {an} be the sequence of Example 2.6.5.
a. Use mathematical induction to prove thatf
I
aa, , = 22I i (a, +a, ) + (a, + 2a.,)I I
b. Use the result of (a) to find lim an.

8. 'Let a, > 0, and for n ? 2, define an = (2 + a, - Prove that {an} is contractive, and find lint
nixan.
9. Let c, E (0, 1) be arbitrary, and for n E N set cn _ s (c.2 + 2).
a. Show that the sequence {c,} is contractive.
b. Let c = lim en. Show that c is a solution of x' - 5x + 2 = 0.
c. Let c, = Z. Using the result of Theorem 2.6.7(b), determine the value of n such that Ic - 10-3.
10. Consider the polynomial p(x) = x3 + 5x - 1. It can be shown that p(x) has exactly one root in the open interval
(0. 1). Let a, E (0, 1) be arbitrary, and for n ? 1, set a _ ;(I - a,3,).
a. Prove that the sequence {an} is contractive.
b. Show that if a = lim an, then p(a) = 0.
c. Let a, = Z. Using the result of Theorem 2.6.7(b), determine the value of n such that la - an! < 10-'.
11. Let a, # a, be real numbers, and let 0 < b < 1. For n ? 3, set an = ban _, + (l - b)an -,.
a. Show that the sequence (an) is contractive.
*b. Find nom
lima,.
12. Prove that if every Cauchy sequence in R converges, then every nonempty subset of R that is bounded above has a
supremum.
_2.77 Series of Real Numbers

In this section we will give a brief introduction to series of real numbers. Some
knowledge of series, especially series with nonnegative terms, will be required
in Chapter 4. The topic of series in general, including various convergence tests, al-
ternating series, etc., will be treated in much greater detail in Chapter 7. We begin
with some preliminary notation. If {an}, is a sequence in N and if p, q E N with
p q, set
Tlak=ap+ap,,+ +av.
k=p
2.7.1 DEFINITION Let {an}'°_, be a sequence of real numbers. Let {s,,}., be the sequence
obtained from where for each n E N, sn = 7-k=, ak. The sequence is called
an infinite series, or series, and is denoted either as
ak or as a1 + a2 + + an +
k-I
For each n E N, s" is called the nth partial sum of the series and a" is called the nth
term of the series.
The series 7,I ak converges if and only if the sequence (s"} of nth partial sums
converges in R. if im s" = s, then s is called the sum of the series, and we write
00
s = I a,
k=1
If the sequence {s"} diverges, then the series joko. I ak is said to diverge.
2.7.2 EXAMPLES
(a) For (rl < 1, consider the geometric series
00
lrrk
k=1
For n(=- N1,
s"=
k-I
rk=r+r2+ +r
Thus
(1 -r)s"=s"-rs"=r-r"+',
S" _ r - r"+
I-r
Since Iri < 1, by Theorem 2.2.6(e),"-.00
lim r" = 0. Therefore lira s" = r/(1 - r), and
thus '~°°
°O _ r
A rk n!<I.
For Iri 1 the series =1 rk diverges (Exercise 3).

(b) Consider the series Moo I ak, where for each k E N, ak = (k - k
1
1)
Then
n
s" _ j ak
k-I
=(1 - 2)+(2 3)+...+(n

n+ 1}
=1- n+1 1
Thus lim s" = 1 and hence Jk I ak = 1.

(c) Consider Y,(-1 )k. Then

0, if n is even.
sn = I (- 1)k
- _ 1, if n is odd.
k=1
Thus since {sn} diverges, the series diverges.
The Cauchy Criterion

The following criterion, which provides necessary and sufficient conditions for the
convergence of a series, was formulated by Augustin-Louis Cauchy (1789-1857) in
1821.
2.7.3 THEOREM (Cauchy Criterion) The series Y,k I ak converges if and only if given
e > 0, there exists a positive integer n,,. such that
< e
for all in > n 2,- n0.

Proof. Since
k_n+I akI
the result is an immediate consequence of Theorem 2.6.2 and Theorem 2.6.4. O
Remark The previous theorem simply states that the series I ak converges if and
only if the sequence {s,,} of nth partial sums is a Cauchy sequence.
2.7.4 EXAMPLE In this example we show that the series 7" 1 k diverges. We accomplish
this by showing that the sequence (sn) of partial sums is not a Cauchy sequence. Con-
sider
stn-sn=n+t.+.
nEN.
+2n'
There are exactly n terms in the sum on the right, and each term is greater than or equal
to 1/2n. Therefore,
)= 1'
(s2. -sn ? n J
2n 2
The sequence {sn} therefore fails to be a Cauchy sequence and thus the series diverges.
The divergence of this series appears to have been first established by Nicole Oresme
(1323?-1382) using a method of proof similar to that suggested in the solution of Ex-
ercise 13 of Section 2.3.
2.7.5 COROLLARY If ET , ak converges, then Jim ak = 0.

Proof. Since ak = sk - sk_ i, this is an immediate consequence of the Cauchy crite-
rion.
Remark. The condition lim at = 0 is not sufficient for the convergence of I ak. For
example, the series 1 diverges, yet slim k = 0.
k -oc
2.7.6 THEOREM S u p p o s e ak at 0 f o r all k E N. T h e n E , at converges if and only if
is bounded above.
Proof. Since ak a 0 for all k, the sequence is monotone increasing. Thus by The-
orem 2.3.2, the sequence converges if and only if it is bounded above.
EXERCISES 2.7
1. *'Using the inequality
1 I 1 1
k2 k(k - l) k-I k
prove that the series 71 converges.

k-ikz
2. Prove that the series X,, converges.
k., k2+k
3. If Irl _- 1, show that the series 7,k'., rk diverges.
4. Prove that the series k 0

k!
5. *Suppose ak >_ 0 for all k. Prove that if B ak converges, then

k.,
I
converges. (See Exercise 3 of Section 2.6.)
k
converges.
6. If Jk , ak and 1:T=, bk both converge, prove each of the following.

a Yk , cak converges for all c E R.
b. 7,'k., (a,, + bk) converges.
7. If I' , (ak + bk) converges, does this imply that the series
a, +b, +a2+b2+
converges?
8. Suppose bk?ak?0for all kE N.
a. If 10ko. , bk converges, prove that J:e , ak converges.
b. If 7Q1. , ak diverges, prove that 7,x I bk diverges.
9. Consider the series T, k,,, p E R.

k
a. Prove that the series diverges for all p 5 1.

b. Prove that the series converges for all p > 1.
NOTES
This chapter provided our first serious introduction to the consequences of the least upper bound property of R. The
limit process. In subsequent chapters we will encounter least upper bound property was used to prove that every
limits of functions, the derivative, and the integral, all of bounded monotone sequence converges. This result was
which are further examples of the limit process. Of the subsequently used to prove the nested intervals property.
many results proved in this chapter, it is difficult to select which in turn was used to prove the Bolzano-Weierstrass
one or two for special emphasis. They are all important! theorem. By Exercise 23 of Section 2.3 and Exercises 13
Many of them will be encountered again-either directly and 14 of Section 2.4. each of these imply the least upper
or indirectly-throughout the text. bound property of R. Another property of the real num-
Some of the concepts and results of this chapter have bers that is equivalent to the least upper bound property is
certainly been encountered previously; others undoubtedly the completeness property of R; namely, every Cauchy
are new. Two concepts that may not have been previously sequence of real numbers converges. Other consequences
encountered are limit point of a set and the limit superior of the least upper bound property will be encountered in
(inferior) of a sequence of real numbers. The notion of subsequent chapters.
limit point of a set is one of the fundamental concepts of Cauchy sequences were originally studied by Cantor
analysis. We will encounter limit points again when we in the middle of the nineteenth century. He referred to
characterize the closed subsets of R. The notion of limit them as fundamental sequences and used them in his con-
point will also be crucial in the definition of the limit of a struction of the real number system R (see Miscellaneous
function. The results of Theorem 2.4.7, although elemen- Exercises 4-11). The main reason that these sequences are
tary, are very useful. The fact that every limit point of a set attributed to Cauchy, rather than Cantor, is because his
is the limit of a sequence of distinct points in the set will 1821 criterion for convergence of a series (Theorem 2.7.3)
be exploited in several instances in subsequent chapters. is equivalent to the statement that the sequence of partial
The primary importance of the limit superior and in- sums is a Cauchy sequence. The fact that Cauchy was a
ferior of a sequence is that these two limit operations al- more prominent mathematician than Cantor may also have
ways exist in R U {-oo, oo}. As we will see in Chapter been a factor. In later chapters we will encounter examples
7, this will allow us to present the correct statements of of spaces of functions that have defined on them a func-
the root and ratio test for convergence of a series. The tion, called a norm, having properties analogous to those
limit superior will also be required to define the radius of of the absolute value function. For such spaces it will also
convergence of a power series. There will be other in- be possible to define both convergence of a sequence and
stances in the text where these two limit operations will the notion of Cauchy sequence. Many of these spaces will
be encountered. also have the property that they are complete; that is, every
In this chapter we have proved several important Cauchy sequence in the space converges.
The first three exercises involve the concept of an infinite product. Let {ak} be a sequence of
nonzero real numbers. F o r e a c h is = 1, 2, ... , define
pn = flak = a, . a2 . . . an.
k=1
If p = lim pn exists, then p is the infinite product of the sequence {ak}k 1, and we write
TM1
p=1
k-1
ak.
If the limit does not exist, then the infinite product is said to diverge. Some authors require that
p * 0. We will not make this requirement; rather we will specify p # 0 if this hypothesis is re-
quired in a result.
1. Determine whether each of the following infinite products converge. If it converges, find the infinite product.
a. H(- I)k c.frlll-

2. If II' 1 ak = p with p * 0, prove that lim ak = 1.
3. If a ? 0 for all n E N, prove that
(1 + ak) converges if and only if ak converges.
k-1 k-1
To prove the result, establish the following inequality:

a,+ +a,, (1+a,) (1+a°):5 e°'` '°
CONSTRUCTION OF THE REAL NUMBERS
In the following exercises we outline the construction of the real number system from the rational
number system using Cantor's method of Cauchy sequences.
Let Q denote the set of rational numbers. A sequence in 0 is Cauchy if for every r f-= Q, r > 0, there exists
a positive integer n° such that Ia - a,,' < r for all n, m z n°. A sequence in 0 is called a null sequence if
for every r (=- 0, r > 0, there exists a positive integer n° such that Ia I < r for all n z n°. Two Cauchy sequences
and in Q are said to be equivalent, denoted provided (a° - is a null sequence.
4. Let and be Cauchy sequences in Q. Prove the following.
a. {a°) -- b. If {b.), then {b°} -
c. If and then {a°} d. If {b°}, then

e. If {a°) - {cn} and {b°} then {a + {c + and {c,d°}.
Given a Cauchy sequence in Q. let denote the set of all Cauchy sequences in Q equivalent to
The set is called the equivalence class determined by
S. Given two Cauchy sequences and (b,,) in 0, prove that provided and
n [{b°}] = 0 otherwise.
Let 9t denote the set of equivalence classes of Cauchy sequences in Q. We denote the elements of 9t by lower-
case Greek letters a, 6, y, .... Thus if a E 9t, a = [{a°}] for some Cauchy sequence in Q. The sequence
is called a representative of the equivalence class a. Suppose a = and P = [{b,,)]. Define -a, a + j9,
and a - 6 as follows:
-a =
One needs to show that these operations are well defined; that is, independent of the representative of the equiva-
lence class. For example, to prove that -a is well defined, we suppose that and are two representatives
of a; i.e., {a°} -- {b°}. But by 4(d), Therefore, [{-a.)) = [(-b°)]. This shows that -a is well
defined.
6. Prove that the operations + and are well defined on 9t.
For each p E Q, let { p} denote the constant sequence p. and set ap = [{ p}]. Also, we set
0 = [{0}], t, = [{l}].
As we will see, the element 0 will be the zero of 9t and k will be the unit of R. A Cauchy sequence in 0 be-
longs to 0 if and only if b -> 0. Similarly, E i if and only if (a - 1) -+ 0. The following problem provides
us with the multiplicative inverse of a # 0.
7. If a # 0, prove that there exists E a such that a, * 0 for all n E N, and that { is a Cauchy sequence.
Define a-' _
S. Prove that 9t with operations + and is a field.
We now proceed to define an order relation on 9t. A Cauchy sequence in Q is positive if there exists
r E Q, r > 0. and no E N such that a > r for all n >_ no. Let 91 be defined by
91 _ is a positive Cauchy sequence}.
9. Prove that the set 'satisfies the order properties (01) and (02) of Section 1.4.
10. Show that the mapping p -> a,, is a one-to-one mapping of Q into % which satisfies
ap+a, =ap,q
for all p, q E Q. Furthermore, if p > 0, then a,, E '.

11. Prove that every nonempty subset of 9t which is bounded above has a least upper bound in 9t.
The above exercises prove that 9t is an ordered field that satisfies the least upper bound property
One can show that any two complete ordered fields are in fact isomorphic, that is, there exists a
one-to-one mapping of one onto the other that preserves the operations of addition, multiplication,
and the order properties. Thus 9t is isomorphic to the real numbers R.
Aguirre, J. A. F., "A note on Cauchy sequences," Math. subsequences;' Amer. Math. Monthly 95 (1988),
Mag. 68 (1995), 296-297. 44-45.
Bell, H. B., "Proof of a fundamental theorem on se- Staib, J. H. and Demos, M. S., "On the limit points of
quences;' Amer. Math. Monthly 71 (1964), 665-666. the sequence (sin n 1:' Math. Mag. 40 (1967),
Goffman. C., 'Completeness of the real numbers;' Math. 210-213.
Mag. 47 (1974), I-8. Wenner, B. R., 'The uncountability of the reals;' Amer
Newman, Donald J. and Parsons, T. D., "On monotone Math. Monthly 76 (1969), 679-680.
3 Structure of Point Sets
3.1 Open and Closed Sets
3.2 Compact Sets

3.3 The Cantor Set
In this chapter we introduce some of the basic concepts fundamental to the study of limits
and continuity, and study the structure of point sets in R. The branch of mathematics con-
ceded with the study of these topics-not only for the real numbers but also for more gen-
eral sets-is known as topology. Modern point set topology dates back to the early part of
this century; its roots, however, date back to the 1850s and 1860s and the studies of Bolzano,
Cantor, and Weierstrass on sets of real numbers. Many important mathematical concepts de-
pend on the concept of a limit point of a set and the limit process, and one of the primary
goals of topology is to provide an appropriate setting for the study of these concepts.
Although we restrict our study to the topology of the real line, all of the concepts
encountered in this chapter can be defined in the more general setting of metric spaces.
A thorough understanding of these topics on the real line will prove invaluable when
they are encountered again in more abstract settings. On first reading, the concepts in-
troduced in this chapter may seem difficult and challenging. With perseverance, how-
ever, understanding will follow.
3 ,1 Open and Closed Sets

In the previous two chapters we used the terms open and closed in describing intervals
in R. The purpose of this section is to give a precise meaning to the adjectives open and
closed, not only for intervals, but also for arbitrary subsets of R. Before defining what
we mean by an open set, we first define the concept of an interior point of a set.
93
94 Chaprer3 Structure of Point Sets
3.1.1 DEFINITION Let E be a subset of R. A point p E E is called an interior point of E

if there exists an e > 0 such that N,(p) C E. The set of interior points of E is denoted
by Int(E), and is called the interior of E.
Recall that for p E R and e > 0, the e-neighborhood N,(p) of p is defined as
{xER:Ix - pl < e}.
3.1.2 EXAMPLES
(a) Let E = (a, b] with a < b. Every p satisfying a 0,
N,(b) = (b - e, b + e)contains points that are not in E. Any x satisfying
b < x < b + e is not in E. This is illustrated in Figure 3.1. For this example,
Int(E) = (a, b).
NN(P) N,(b)
a P b
Figure 3.1
(b) Let E denote the set of irrational real numbers, i.e., E = R \ Q. If p E E, then by
Theorem 1.5.2, for every e > 0 there exists r E Q fl N,(p). Thus N,(p) always con-
tains a point of R not in E. Therefore no point of E is an interior point of E; i.e.,
Int(E) = 0. Using the fact that between any two real numbers there exists an irrational
number (Exercise 6, Section 1.5), a similar argument also proves that Int(Q) _ ¢.
Open and Closed Sets

Using the notion of an interior point, we now define what we mean by an open set.
3.1.3 DEFINITION
(a) A subset 0 of R is open if every point of 0 is an interior point of 0.
(b) A subset F of P is dosed if F` = R \ F is open.
Remark. From the definition of an interior point it should be clear that a set 0 C R
is open if and only if for every p E 0 there exists an e > 0 (depending on p) so that
N,(p) C 0. In Theorem 3.1.9 we will provide a characterization of closed sets in terms
of limit points.
3.1.4 EXAMPLES
(a) The entire set N is open. For any p E N and e > 0, N,(p) C R. Since R is open,
by definition the empty set 0 is closed. However, the empty set is also open. Since 4)
contains no points at all, Definition 3.1.3(a) is vacuously satisfied. Consequently R is

also closed.
(b) Every e-neighborhood is open. Suppose p E I8 and E > 0. If q E NE(p), then
p - qI < e. Choose S so that 0 < S s e - p - qI. If x E N5(q), then
Ix-PI:5 IP - qI+Ix - qI
< IP - qi + S<_ IP-qi +e - IP - qi = e.
Therefore N6(q) C NE(p) (see Figure 3.2). Thus q is an interior point of N1(p). Since
q E NE(p) was arbitrary, NE(p) is open.
Ne(p)
-N6(q)
p-E p q p+E
Figure 3.2
(c) Let E = (a, b], a < b, be as in Example 3.1.2(a). Since the point b E E is not an
interior point of E. the set E is not open. The complement of E is given by
E`=(-oo,a]U(b,oo).
An argument similar to the one given in Example 3.1.2(a) shows that a is not an inte-
rior point of E`. Thus E` is not open and hence by definition E is not closed. Hence E
is neither open nor closed.
(d) Let F = [a, b], a < b. Then
F` _ (-oo, a) U (b, oo),
and this set is open. This can be proved directly, but also follows as a consequence of
Theorem 3.1.6(a) below. Thus F is closed.
(e) Consider the set 0. Since no point of 0 is an interior point of Q (Example
3.1.2(b)), the set Q is not open. Also, C is not closed.
The use of the adjective open in describing the intervals (a, b), (a, on), (-cc, b)
and (-oo, oo) is justified by the following theorem.
3.1.5 THEOREM Every open interval in R is an open subset of It

Proof. Exercise 1.
3.1.6 THEOREM
(a) For any collection {Oa},,EA of open subsets of R, n o. is open.
&FA
(b) F o r any f i n i t e collection {O,, ... , of open subsets of R. fl O; is open.
96 Chapter 3 Structure of Point Sets
Proof. The proof of (a) is left as an exercise (Exercise 6).

(b) If f l o, = 46, we are done. Otherwise, suppose
pE0=.n 0;.
Then p E 0, for all i = 1, . .. , n. Since 0; is open, there exists an e; > 0 such that
NE,(p) C 0..
Let e = min{e,, ... , e,}. Then e > 0 and NE(p) C 0, for all i. Therefore
NE(p) C 0; i.e., p is an interior point of 0. Since p E 0 was arbitrary, 0 is open. C)
As a consequence of the previous theorem every closed interval [a. b], a, b E R with
a s b, is a closed subset of R. Since
H \ [a, b] = (-oo, a) U (b, oo)
is the union of two open intervals, by the previous theorem P \ [a, b] is an open subset
of R. Thus [a, b] is a closed set.
For closed subsets of R we have the following analogue of the previous result.
3.1.7 THEOREM
(a) For any collection {Fa}aEA of closed subsets of P. n F. is closed.
(b) For any finite collection IF,, ... , of closed subsets of R. U F is closed
j°
Proof. The proofs of (a) and (b) follow from the previous theorem and De Morgan's
laws:
A EA (")c
j=1 j .1
Remark. The fact that the intersection of a finite number of open sets is open is due
to the fact that the minimum of a finite number of positive numbers is positive. This
guarantees the existence of an e > 0 such that the e-neighborhood of p is contained in
the intersection. For an infinite number of open sets, the choice of a positive e may no
longer be possible. This is illustrated by the following two examples.
3.1.8 EXAMPLES We now provide two examples to show that part (b) of Theorem 3.1.6
is, in general, false for a countable collection of open sets. Likewise, part (b) of Theo-
rem 3.1.7 is, in general, also false for an arbitrary union of closed sets (Exercise 6b).
(a) For each n = 1, 2, .. . , let 0 = (-,1-,,,-,). Then each O is open, but
00
n1 O = {0},
which is not open.

(b) Alternatively, if we let G,, = (0, 1 + ), n = 1, 2.... , then again each G. is

open, but
x
n
n-I
G,, = (0. 1 ],
which is neither open nor closed.

The following theorem provides a characterization of the closed subsets of R. Be-
fore stating the theorem we recall the definition of a limit point (Definition 2.4.5). For
E C R, a point p E R is a limit point of E if for every e> 0. (N,(p)1 { p}) n E t- 4,.
3.1.9 THEOREM A subset F of R is closed if and only if F contains all its limit points.
Proof. Suppose F is closed. Then by definition F` is open and thus for every p E F'
there exists e > 0 such that N, (p) C F'; that is, Ne(p) fl F = 0. Consequently. no
point of F` is a limit point of F. Therefore F must contain all its limit points.
Conversely, let F be a subset of R that contains all its limit points. To show F is
closed we must show F'' is open. Let p E V. Since F contains all its limit points. p is
not a limit point of F. Thus there exists an e > 0 such that N,(p) n F = 44. Hence
N,(p) C F' and p is an interior point of V. Since p E F` was arbitrary, F` is open and
therefore F is closed.
Closure of a Set
3.1.10 DEFINITION If E is a subset of R, let E' denote the set of limit points of E. The
closure of E, denoted E, is defined as
E=EUE'.
3.1.11 THEOREM If E is a subset of R, then
(a) E is closed.
(b) E = E if and only if E is closed.
(c) E C F for every closed set F C R such that E C F.
Proof. (a) To show that E is closed, we must show that k is open. Let p E E`. Then
p 44 E.and p is not a limit point of E. Thus there exists an e > 0 such that
N,(p)flE_¢.
We complete the proof by showing that NE(p) fl E' is also empty and thus N,(p) fl E =
Therefore N,(p) C E`; i.e., p is an interior point of E`.
Suppose N,(p) fl E' * 4,. Let q E NE(p) n E', and choose S > 0 such that
Na(q) C N,(p). Since q E E, q is a limit point of E and thus N8(q) fl E * 46. But this
implies that N. (p) n E * tb, which is a contradiction. Therefore N,(p) n E' = 46,
which proves the result.
_ (b) If E = E, then E is closed. Conversely, if E is closed, then E' C E and thus
E. _
(c) If E C F and F is closed, then E' C F. Thus E C F.
98 Chapter3 Structure of Point Sets
3.1.12 DEFINITION A subset D of H is dense in R if D = H.

The rationals 0 are dense in R. By Example 2.4.6(c), every point of H is a limit
point of Q. Hence 0 = H. This explains the comment following Theorem 1.5.2. The
rationals are not only dense; they are also countable. Countable dense subsets play a
very important role in analysis. They allow us to approximate arbitrary elements in a set
by elements chosen from a countable subset of H. Since the rationals are dense in R,
given any p E 11 and e > 0, there exists r E 0 such that Ip - rI < e. Additional ex-
amples of this will occur elsewhere in the text.
Characterization of the Open Subsets of R'

If {In} is any finite or countable collection of open intervals, then by Theorem 3.1.6,
U = U I is an open subset of N. Conversely, every open subset of H can be expressed
as a finite or countable union of open intervals (see Exercise 17). However, a much
stronger result is true. We now prove that every open set can be expressed as a finite or
countable union of pairwise disjoint open intervals. A collection {!n} of subsets of R is
pairwise disjoint if 1, fl Im = 0 whenever n # m.
3.1.13 THEOREM If U is an open subset of R, then there exists a finite or countable col-
lection {1n} ofpairwise disjoint open intervals such that
U=UI..
Proof. Let x E U. Since U is open, there exists an e > 0 such that
(X-E,x+E)C U.
In particular (s, x) and [x, t) are subsets of U for some s < x and some t > x. Define
r, and lx as follows:
r, = sup{t : t > x and [x, t) C U}, and
lx = inf{s : s < x and (s, x] C U}.
Then x < r, < oo and -oo s 1, < x. For each x E U, let 1, = (t r,). Then
(a) 1, C U, and
(b) if x, y E U, then either Ix = I, or 1, fl /,. = 4).
The proofs of (a) and (b) are left as exercises (Exercise 16).
To complete the proof, we let 2 = {I, : x E U}. For each interval I E 9, choose
rt E Q such that rt E I. If !, J E 2 are distinct intervals, then rt * rj. Therefore the
mapping I - rt is a one-to-one mapping of T into O. Thus the collection Y is at most
countable and therefore can be enumerated as {I;}iEA, where A is either a finite subset
of N, or A = N. Clearly
U=iE Ii,
1. This topic can be omitted upon first reading of the text. The structure of open sets will only be required in
Chapter 10 when defining the measure of an open subset of R.
and by (b), if n * j, then 1 f11, = 4). Thus the collection {1%EA is pairwise dis-
joint.
Relatively Open and Closed Sets

One of the reasons for studying topological concepts is to enable us to study properties
of continuous functions. In most instances, the domain of a function is not all of R, but
rather a proper subset of R, as is the case with f(x) = NA for which Dom f = [0, oc).
When discussing a particular function we will always restrict our attention to the do-
main of the function rather than all of R. With this in mind we make the following
definition.
3.1.14 DEFINITION Let X be a subset of R.

(a) A subset U of X is open In (or open relative to) X if for every p E U, there
exists e > 0 such that NN(p) fl x C U.
(b) A subset C of X is closed in (or closed relative to) X if X \ C is open in X.
3.1.15 EXAMPLE Let X = [0, oo) and let U = (0, 1). Then U is not open in R but is open
in X. (Why?)
The following theorem, the proof of which is left as an exercise (Exercise 21), pro-
vides a simple characterization of what it means for a set to be open or closed in X.
3.1.16 THEOREM Let X be a subset of R.

(a) A subset U of X is open in X if and only if U = X fl O for some open subset
OofR.
(b) A subset C of X is closed in X if and only if C = X fl F for some closed
subset F of R.
Connected Sets2
Our final topic in this section involves the notion of a "connected set." The idea of con-
nectedness is just one more of the many mathematical concepts that have their roots in
Cantor's studies on the structure of subsets of R. When we use the term "connected sub-
set' of R, intuitively we are inclined to think of an interval as opposed to sets such as the
positive integers N or (0, 1) U {2}. We make this precise with the following definition.
3.1.17 DEFINITION A subset A of R is connected if there do not exist two disjoint open
sets U and V such that
(a) AlU*4)andAflV*e(,,and
(b) (An u)U(Af1V)=A.
2. this concept, though important and used implicitly in several instances in the text, will not be required
specifically in subsequent chapters except in a few exercises. Thus the topic of connectedne ss can be omitted
upon first reading of the text.
The definition for a connected set differs from most definitions in that it defines
connectedness by negation; i.e., defining what it means for a set not to be connected.
According to the definition, a set A is not connected if there exist disjoint open sets U
and V satisfying both (a) and (b). As an example of a subset of R that is not connected,
consider the set of positive integers N. If we let U = (1, Z) and V = (, co). then U and
V are disjoint open subsets of 68 with
u n Nl l} and v n N = {2, 3, ...}
that also satisfy (U n N) U (V n N) = N. That the interval (a, b) is connected is a con-
sequence of the following theoreth, the proof of which is left to the exercises (Exercise 25).
3.1.18 THEOREM A subset of 118 is connected if and only if it is an interval.
EXERCISES 3.1
2. *Show that every finite subset of I8 is closed.
3. Show that the intervals (-oo, a) and [a, oo) are closed subsets of R.
4. For the following subsets E of R, fill in the chart.
E Int(E) E' Isol. pts. of E E Open? Closed?

(0, 1) U {2}
(a, b)
(a,b]
[a, b]
[o, 1]na
5. a. Let F be a closed subset of R and let { p,} be a sequence in F which converges to p E R. Prove that p E F.
b. Show by example that the conclusion is false if F is not closed.
6. *a. Prove Theorem 3.1.6(a).
b. Give an example of a countable collection of closed subsets of R such that U.=, F. is not closed.
7. Let A, B be subsets of R.
a. If A C B, show that Int(A) C Int(B). b. Show that Int(A n B) = Int(A) n Int(B).
c. Is Int(A U B) = Int(A) U Int(B)?
8. Let E be a subset of R.
*a. Prove that Int(E) is open. b. Prove that E is open if and only if E = Int(E).
c. If G C E and G is open, prove that G C Int(E).
9. Let A. B be subsets of R.
*a. Show that (A U B) = A U B. b. Show that (A fl B) C A fl B.
c. Give an example for which the containment in part (b) is proper.
10. Prove that the set of limit points of a set is closed.
11. Let E C R. A point p E 18 is a boundary point of E if for every e > 0, N,(p) contains both points of E and
points of E. Find the boundary points of each of the following sets.
*a. (a, b) b. E = {,'-, : n E N} c. N d. Q
12. a. Prove that a set E C B8 is open if and only if E does not contain any of its boundary points.
b. Prove that a set E C 18 is closed if and only if E contains all its boundary points.
13. *Prove that the set of irrational numbers is dense in R.
14. *If D is dense in 18, prove that for every p E IB there exists a sequence in D with lim p = p.
15. Let Do = {0, 11, and for each n E N, let D" = {a/2" : a E N, u is odd, 0 < a < 2"}. Let D = U `-0 D". Prose
that D is a countable dense subset of [0, I ].
16. Prove statements (a) and (b) of Theorem 3.1.13.
17. *Prove that there exists a countable collection 9 of open intervals such that if U is an open subset of E8 and
p E U, there exists / E I with p E /" C U.
18. Let X = (0, oo). For each of the following subsets of X determine whether the given set is open in X. closed in X.
or neither.
*a. (0, 1 ] b . (0, 1 ) *c. (0, 1 ] U (2, 3) d. (0, 1 ] U (21 e. (1, : n E NJ
19. For each of the following subsets of 0, determine whether the set is open in 0, closed in Q. both open and
closed in 0, or neither.
a.A={pE0:I<p2<2} b.B=(pE0:2<p2<3) c.N
20. *If U is an open subset of 68, prove that E C U is open in U if and only if E is an open subset of R.
22. Let X be a subset of R. Prove each of the following.
a. If {Ua},EA is any collection of subsets of X with U. open in X for all a E A. then Ua is open in X.
b. If { U,, ... , Ua} is any finite collection of subsets of X with each U; open in X. then f;'= , U, is open in X.
23. For each of the following, use the definition to prove that the given set is not connected.
*a. (0, 1) U {2} b. n= 1.2,...} c. (pE0:p>0andI <p2<3}
24. *a. If A is connected, prove that A is connected.
b. Give an example of a set A such that A is not connected, but A is connected.
25. *Prove Theorem 3.1.18.
Compact Sets
In this section we introduce the concept of a "compact set" and provide a characteriza-
tion of the compact subsets of R. The notion of a compact set is very important in the
study of analysis, and many significant results in the text will depend on the fact that
every closed and bounded interval in F8 is compact. The modern definition of a compact
set given in Definition 3.2.3 dates back to the second half of the nineteenth century and
the studies of Heine and Borel on compact subsets of R.
3.2.1 DEFINITION Let E C It A collection {O,},.EA of open subsets of R is an open

cover of E if
E C QU 0..
An alternative definition is as follows: The collection {O,}REA of open sets is an

open cover of E if for each p E E, there exists an index a E A such that p E 0,.
3.2.2 EXAMPLES
(a) Let E = (0, 1) and 0n = (0, 1 - ), n = 2,3..... Then 2 is an open
cover of E. To see this, suppose x E E. Then since x < 1, there exists an integer n such
that x < 1 - R. Thus x E 0,,, and as a consequence,
all
E C n=2
U 0,,.
which proves the assertion. In fact, since 0 C E for each n. we have E = U. 't2 0n.
(b) Let F = ;0, oo) and for each n E I,1, let U,, _ (-1, n). Then {Un}nENis an open
cover of F.
(c) If E is any nonempty subset of R, then for every e > 0. the collection
{N,,(x): x E E} is an open cover of E.
3.2.3 DEFINITION A subset K o f U8 is compact i f every o p e n cover o f K has a f i n i t e sub-

cover o f K; that is, i f {0,}a,EA is an o p e n cover o f K , then there exists a,,. .. , a E A
such that
R
K C U0,.
Remark To prove (using the definition) that a given set K is compact, we must prove
that f o r m= open cover {O,} of K, there exists a finite subcollection 0,,, ... , 0..
whose union covers K. This is illustrated in (a) of the following examples. It is W suf-
ficient to find an open cover of K for which a finite subcollection also covers K. Such
an open cover always exists. (Why?) On the other hand, to show that a set E is not com-
pact, it suffices to find (construct) an open cover {O,} of E for which the union of no
finite subcollection contains E. This is illustrated in Examples 3.2.4 (b) and (c).
3.2A EXAMPLES
(a) Every finite set is compact. Suppose E = {p,, ... , pJ is a finite subset of N and
{0,},EA is an open cover of E. Then for each j, j = 1, ,.. , n, there exists a; E A such
that pi E 0,; But then {O,}7. is a finite subcollection which covers E.
(b) The open interval (0, 1) is not compact. For the open cover {O"} 2 of (0, 1) in Ex-
ample 3.2.2(a), no finite subcollection can cover (0, 1). Suppose on the contrary that a
finite number, say O . . . , O",, covers (0, 1). Let N = max{n . . . , Pik}. Then
k N
(0.1)000"CUO"= 0,1-N,
which is a contradiction.
(c) The closed set F = [0, oo) is not compact. For the open cover CU.
of F, no finite subcollection can cover F. If there exist a finite number of sets in al1
which cover F, then there exists N E N such that F C (-1, N). (Why?) This, however,
is a contradiction.
Properties of Compact Sets

Before we provide a characterization of the compact subsets of U8, we first prove sev-
eral properties of compact sets that require only the definition of compactness. As a
consequence, all three of the following theorems are true in more general settings; e.g.
in n-dimensional space R", and in general metric spaces. (For a discussion of metric
spaces, see the notes at the end of the chapter.)
3.2.5 THEOREM
(a) Every compact subset of IR is closed and bounded.
(b) Every closed subset of a compact set is compact.
Proof. (a) Let K be a compact subset of R. To show that K is bounded, consider the
open cover {(-k, k)}5EN of 6$, and hence also of K. Since K is compact, there exist
k,, ... , k" E IN such that
KCU
If N = max{k,, ... , k"}, then K is a subset of (-N, N). Thus K is bounded.

To show that K is closed, we need to show that K` is open. Let p E K` be arbitrary.
For each q E K, choose eq > 0 such that
N,,(q)nN.,(p) _ 4,
Any eq satisfying 0 < eq < z p - ql will work. Then {N,(q)}qE,, is an open cover of
K. Since K is compact, there exists q,, ... , q" such that
K C U N. (qj).
j=1 qj
Let e = min{eqj : j = 1, .. . , n}, which is positive. Then

N,(p)f N,gj(g) = 0 for all j = 1.. ... n.
Thus N,(p) fl K = 0; i.e., N,(p) C K. Therefore p is an interior point of K` and thus
K` is open.
(b) Let F be a closed subset of the compact set K and let {4".},,E.a be an open cover
of F. Then
{Oa}aEA U {F`}
is an open cover of K. Since K is compact. a finite number of these will cover K. and
hence also F.
As a consequence of the previous theorem, the open interval (0. 1) is not compact
since it is not closed. Likewise, the set A = {n : n E N} is not compact since again A
is not closed. By Example 2.4.6(b), the point 0 is a limit point of A but 0 0 A. Also, the
closed set [0, oo) is not compact since it is not bounded. In Theorem 3 .2.9 we will prove
the converse of Theorem 3.2.5(a); namely, every closed and bounded subset of l is
compact.
Remark. In proving that the compact set K was closed, compactness allowed us to
select a finite subcover from the constructed open cover of K. Finiteness then assured
that the e as defined was positive. This method of first constructing an open cover pos-
sessing certain properties and then using compactness to assure the existence of a finite
subcover will be used on other occasions in the text.
3.2.6 THEOREM If S is an infinite subset of a compact set K. then S has a limit point in K.
Proof.3 If no point of K is a limit point of S, then for each q E K, there exists a
neighborhood Nq of q so that Nq contains at most one point of S, namely q if q E S.
Since S is infinite, no finite subcollection of {Nq}qEK can cover S, and consequently no
finite subcollection of {Nq}qEK can cover K. This is a contradiction.
Remark. Since every compact subset of R is closed and bounded, the previous the-
orem can also be proved using the Bolzano-Weicrstrass theorem (2.4.10) (see Exercise
6). Moreover, as we will see in Theorem 3.2.9, if K is a subset of R having the prop-
erty that every infinite subset of K has a limit point in K, then K is compact. The prop-
erty that "every infinite subset of K has a limit point in K" is usually referred to as the
Bolzano-Welerstrass property.
Another useful consequence of compactness is the following analogue of the
nested intervals property.
3.2.7 THEOREM If t is a sequence of nonemply compact subsets of R with

K K. + t for all n, then
K= fIK,,
is nonempty and compact.
3. The proof of Theorem 3.2.6 is again a proof by contradiction. In the proof we assume that the hypothesis
"S is an infinite subset of K and K is compact" is true, and that the conclusion "S has a limit point in K" is
false.
Proof. We first show that fl , K # ¢. Let O = K. By Theorem 3.2.5, K. is

closed and thus O is open. Furthermore,
00 00
n K. = 45 if and only if Ul O = R.
Thus if fl. , K. then is an open cover of R, and thus also of K1. But K,
is compact. Therefore there exists n, < < nk such that
k
K, C j_
U 0,,J..
But then K, fl K,,, fl fl K,,, _ ¢. This, however, is a contradiction, since the in-
tersection is equal to K,,,, which by hypothesis is nonempty. Thus K= fl K # 40. By
Theorem 3.1.7, K is closed, and hence by Theorem 3.2.5(b), K is compact. Q
Characterization of the Compact Subsets of R

We now turn to our goal of providing a characterization of the compact subsets of the
real line R. The first of the following two results is attributed to Eduard Heine
(1821-1881) and Emile Borel (1871-1956); the second is due to Bernhard Bolzano
(1781-1848) and Karl Weierstrass (1815-1897). The two theorems rank very high
among the many important advances in the foundations of analysis during the nine-
teenth century. The importance of these results will become evident in later chapters.
As is to be expected, the least upper bound property of u8 will play a crucial role in the
proofs of these theorems.
3.2.8 THEOREM (Heine-Borel) Every closed and bounded interval [a, b) is compact.
Proof. Let °1L = {U,},,, be an open cover of [a, b] and let
E = {r e [a, b] : [a, r] is covered by a finite number of the sets U,}.
The set E is bounded above by b, and since a E U. for some a E A, E is nonempty.
Thus by the least upper bound property, the supremum of E exists in R. Let c = sup E.
Since b is an upper bound of E, c < b.
We first show that c E E; i.e., [a, c] is covered by a finite number of sets in 91.
Since c E [a, b], c E Up for some (3 E A. Since Up is open, there exists e > 0 such
that (c - e, c + e) C Up. Furthermore, since c - e is not an upper bound of E, there
exists r E E such that c - e < r c. But then [a, r] is covered by a finite number,
s a y U.,,, ... , U , , of sets in °U. But then the finite collection (U.,, ... , U,, , Up} cov-
ers [a, c]. Therefore, c E E.
To conclude the proof, we show that c = b. Suppose c < b. If we choose s < b
such that c < s < c 4- e, then the collection {U,,, ... , U,,, Up} also covers [a. s].
Thus s E E, which contradicts that c = sup E. Hence c must equal b. Q
An alternative proof of the Heine-Borel theorem using the nested interval property
(2.3.3) is outlined in the exercises (Exercise 11). The statement of the Heine-Borel theo-
rem was initially due to Heine, a student of Weierstrass, who used the result implicitly in
the 1870s in his studies on continuous functions. The theorem was proved by Borel in
1894 for the case where the open cover was countable. For an arbitrary open cover the re-
sult was finally proved in 1904 by Henri Lebesgue (1875-1941). Using the Heine-Borel
theorem, we now prove the following characterization of the compact subsets of R.
3.2.9 THEOREM (Heine-Borel-Bolzano-Weierstrass) Let K be a subset of R. Then the fol-

lowing are equivalent:
(a) K is closed and bounded.
(b) K is compact.
(c) Every infinite subset of K has a limit point in K.
Proof. (a) (b). Since K is bounded, there exists a positive constant M so that
K C [-M, M]. Since [-M, M] is compact and K is closed, by Theorem 3.2.5(b), K is
compact.
(b) (c). This is Theorem 3.2.6.
(c) = (a). Suppose the set K is not bounded. Then for every n E N there exists
p E K with p * p. for n * in such that Ip.i > n for all n. Then (p,,: n = 1, 2, ...)
is an infinite subset of R with no limit point in R. and hence none in K, which is a con-
tradiction. Thus K is bounded.
Let p be a limit point of K. By definition of limit point, for each n E N there exist
p E K with p * p such that
1n
IN - A <
Then S = {p,,: n = 1, 2, ... } is an infinite subset of K, and p is a limit point of S. To
complete the proof we must show that p is the only limit point of S, and hence by hy-
pothesis must be in K; i.e., K is closed.
Suppose q E QR with q # p. Let e = iI p - 9I, and choose N E N such that
1/N < e. Then for all n a N,
Ip - qI < Ipp - ql+Ipn - pI < Ipp - qI + n<Ipp-ql+e.

Therefore, for all n ? N,
1p.-qI > Ip-qI -e=2Ip-qI.

Thus N1(q) contains at most finitely many p, and as a consequence q cannot be a limit
point of S. Therefore, no q E R with q * p is a limit point of S.
Several additional consequences of compactness are given in Exercises 5 and 6.
Another is the following theorem.
3.2.10 THEOREM Let K be a nonempty compact subset of R. Then every sequence in K has
a convergent subsequence that converges to a point in K.
Proof. Exercise 7.
D EXERCISES 3.2
1. Let A = {;,:n = 1, 2,. . . }.
a. Show that the set A is not compact by constructing an open cover of A that does not have a finite subcoser
of A.
b. Prove directly (using the definition) that K = A U {0} is compact.
2. Suppose is a convergent sequence in R with lim p = p. Prove, using the definition. that the set
A = {p} U {p" n E N} is a compact subset of W.
:
3. Show that (0, 1 ) is not compact by constructing an open cover of (0. 1' that does not have a finite subcover.
4. Suppose A and B are compact subsets of R.
'a. Prove (using only the definition) that A U B is compact. b. Prove that A (1 B is compact.
5. 'Let K be a nonempty compact subset of R. Prove that sup K and inf K exist and are in K.
6. Let K be a compact subset of R. Use the Bolzano-Weierstrass theorem (2.4.10) to prove that every infinite subset
of K has a limit point in K.
8. Prove that a compact set has only a finite number of isolated points.
9. Let K be a compact subset of R and p E R \ K.
a. Prove that there exist points a, h E K such that
Ja - p, = inf{lx - pI : .r E K} and jb - p = sup{jx - pI : x E K}.
b. Are either of the above still true if K is only closed'?
10. a. Prove that the intersection of an arbitrary collection of compact sets is compact.
b. Find a countable collection {K,} . , of compact subsets of R such that U -, K, is not compact.
11. This exercise outlines an alternative proof of the Heine-Borel theorem. Suppose [a. h: is not compact. Then there
exists an open cover 1t = of [a, h] such that no finite subcollection of 1l covers 'a. b'.We now proceed
to show that this leads to a contradiction. Divide [a, b] into two closed subintervals [a. "- * n] and [" t. b'. each
of length (b - a)/2. At least one of these, call it h, cannot be covered by a finite number of the U,. Repeating
this process, obtain a sequence of closed and bounded intervals satisfying
(a) (a, b] D /, D 1, D D/D (b) length of 1 = (b - a)/2", and
(e) for each n, 1" is not covered by a finite number of the U,,.
Now use Corollary 2.3.3 to obtain a contradiction.
3.3 1 The Cantor Set

In this section we will construct a compact subset of [0, 1 ], known as the Cantor set.
that has a number of interesting properties. This set is constructed by induction, the first
two stages of which are illustrated in Figure 3.3.
Let P. = [0, 11. From PO remove the middle third open interval (3, 3). This leaves
two disjoint closed intervals
J,.= [o.''] and J,,, =

( 2I 1
J 1.2
f1rY
9
VF1}
9 3
2
3
ìu}
JZ ,
r
7
9
8
9
zs
Figure 3.3
Set P, = J,,, U J,,2.

From each of J,., and J1,2 remove the middle third open intervals
1 2 7 8
C 32, 32 and 32,
32
of length 1/32 This leaves 22 disjoint closed intervals J2.1' J2.2, J2,3, J24 of length 1/32;
namely,
326 7
L 32' 3 J L ' 3 ,' 32' 3 ,
Set P2 = J2,, U J2.2 U J2,3 U J2.4. In Figure 3.3, the shaded intervals indicate the open
intervals that are removed at each stage of the construction.
We continue this process inductively. At the nth step, each P. is the union of 2" dis-
joint closed intervals each of length 1/3"; i.e.,
r
P. = U J..j,
where for each j, J",; is a closed interval of the form
zj z!+ 1
Jn.j = 3n' 3"
Since each P, is a finite union of closed intervals, P. is closed and bounded, hence com-
pact. Furthermore, since P0 D P, D P2 D , by Theorem 3.2.7,
a
P= t5 P"
is is a nonempty compact subset of (0, 1 ]. The set P is called the Cantor ternary set. We
now consider some properties of the set P.
PROPERTY I P is compact and nonempry.
PROPERTY 2 P contains all the end points of the closed intervals {J",r}, n = 1,
2, . . , k = 1,2,. . ,2".
PROPERTY 3 Every point of P is a limit point of P.

Proof. Let p E P and let e > 0 be given. Choose in E I 1 such that 1/3< e. Since
p E P P E J,".k for some k, 1 <_ k s 2"'. But
xk L L 1
Jm. k = 3,,,
3m
Since the length of = 1/3'" < E, Jm.k C N,(p). Thus both end points of Jm., are in
P fl Nf(p), and at least one of these is distinct from p. 0
PROPERTY 4 The sum of the lengths of the intervals removed is 1.

Proof. At step 1, we removed one interval of length 1/3 . At the second step, we re-
moved two intervals of length 1/32. At the nth step, to obtain P", we removed 2" -' in-
tervals of length 1/3". Thus the sum of the lengths of the intervals removed
'
=3+232+...2"-13"+..
x 2'?- 1x 2
3" 3 3
t
31
O
3
As a consequence of Property 4,
PROPERTY 5 P contains no intervals.

For x E [0, 1 let x = .nln2n3 be the ternary expansion of x. As we indicated
in Section 1.6, this expansion is unique except when
x= 3m, aEN with 0 < a < 3m,
where 3 does not divide a. In this case, x has two expansions: a finite expansion
x=3+ + 3m, a. E {1, 2},
and an infinite expansion. If am = 2, we will use the finite expansion. If am = 1. we will

use the infinite expansion
x=-+ a1
3
++3m
0
co 71
k.m+l
2
3k
With this convention we have
PROPERTY 6 if for each x E [0,1 J, x = .nln2n3 is the ternary expansion of x,

then
x E P if and only if nk E {0, 2}.

Proof. Exercise 2.
As a consequence of Property 6 and Theorem 1.7.18,
PROPERTY 7 P is uncountable.
For each n, the set P. has only a finite number of end points. As a consequence, the
set of points of P that are end points of some open interval removed in the construction
is countable. Since P is uncountable, P contains points other than end points. By Exer-
cise 1(a) of Section 1.6, the ternary expansion of; is
1
= .020202
4
Thus E P. but 1 is not an end point of any closed interval J,",k.
;
Remark. By Property 4, the sum of the lengths of the intervals removed is 1. This
seems to imply that P is in some sense very "small." On the other hand, by Property 7.
P is uncountable, which seems to imply that P is "large:'
EXERCISES 3.3
1. Determine whether ; is in the Cantor set.
2. Prove Property 6 of the Cantor set.
3. Let 0 < a < 1. Construct a closed subset F of [0, 1 ] in a manner similar to the construction of the Cantor set
such that the sum of all the open intervals removed is equal to a.
4. *Prove that the Cantor set P is equivalent to [0, I ].
NOTES
Without a doubt, the most important concept of this chap- set of K has a limit point in K. The converse of this result
ter is compactness. The fact that every open cover of a is also true, not only in R (Theorem 3.2.9), but also in the
compact set has a finite subcover will be crucial in the more general setting of metric spaces. A proof of this im-
study of continuous functions, especially uniform conti- portant result is outlined in the miscellaneous exercises.
nuity. As we will see in many instances, the applications The notion of a metric and a metric space is moti-
of compactness depend on the ability to choose a finite vated by geometrical considerations in R and n-dimen-
subcover from a particular open cover. A good example of sional euclidean space R". Many of the results involving
this is the proof of Theorem 3.2.5. Other instances will the structure of point sets in R depend on the fact that the
occur later in the text. euclidean distance d(x, y) = Ix - yI between the points x
Since compactness is the most important concept, and y in R satisfy the following properties: d(x. y) z 0,
Theorems 3.2.8 and 3.2.9 are the two most important re- d(x, y) = 0 if and only if x = y, d(x, y) = d(y. x), and
sults. In the Heine-Borel theorem we proved that every
d(x, y) s d(x. z) + d(z, y)
closed and bounded interval is compact, whereas in the
Heine-Borel-Bolzano-Weierstrass theorem, we character- for all x, y, z E R. Using these properties as a motivation.
ized the compact subsets of R. In Theorem 3.2.6 we the notion of distance can be extended to sets other than
proved that if K is a compact set, then every infinite sub- P as follows.
DEFINITION Let X be a nonempty set. A real-valued of a metric. It allows one to prove that every e-neighbor-
function d defined on X X X satisfying hood of a point is an open set. Also, combined with the
property that d(p, q) = 0 if and only if p = q, it guaran-
(a) d(x, y) ? O for all x, y E X.
tees that the limit of a convergent sequence in X is unique.
(b) d(x, y) = 0 if and only if x = y, With the exception of Theorems 3.1.13 and 3.1.18, all of
(c) d(x, y) = d(y, x), the results of Section 3.1 are still valid for any metric
space (X, d). In fact, the proofs follow verbatim the proofs
(d) d(x, y) s d(x, z) + d(z, y) for all x, y, z E X,
given in the text. The proofs of Theorems 3.1.13 and
is called a metric or distance function on X. The set X 3.1.18, however, require properties of the real number
with metric d is called a metric space, and is denoted by system and thus cannot be expected to be valid in an ar-
(X, d). bitrary metric space.
As for the real line, if (X, d) is a metric space, a sub-
Inequality (d) is called the triangle inequality. If
set K of X is compact if every open cover of K has a finite
d is a metric on a set X, then as in Definition 2.1.6, if
subcover of K. Again, since the proofs of Theorems 3.2.5.
p E X and e > 0, the e-neighborhood N,(p) of p is de-
3.2.6, and 3.2.7 only require the definition of compactness.
fined as
these results are again valid for an arbitrary metric space. In
N,(p) = {x E X : d(p,x) < e). particular, if K is a compact subset of (X, d) then K is closed
Since the notions of limit point of a set, limit of a se- and bounded. In this setting, a subset E of X is bounded if
quence, and interior point of a set are all defined in terms there exists a positive constant M and p e X such that
of e-neighborhoods, all of these concepts can be defined d(p, x) M for all x E E. The converse, however, is false.
verbatim for an arbitrary metric space. As in R, if (X, d) For example, if we set p(x, y) _ Ix - yI/(1 + Ix - v ).
is a metric space, a subset 0 of X is open in (X, d) if and then p is a metric on R satisfying p(x, y) < I for all
only if every point of 0 is an interior point of O. It is im- x, y E R (see Miscellaneous Exercise 4). In particular.
portant to note, however, that different metrics on a given [0. oo) is a closed and bounded subset of (R, p) that is not
set result in different c-neighborhoods, and thus a set that compact. Thus the Heine-Borel-Bolzano-Weierstrass theo-
is open with respect to one metric may not be open with rem is not valid in an arbitrary metric space. This is to be ex-
respect to a different metric. If X is any nonempty subset pected since the least upper bound property of R was used
of lR, then d defined on X X X by d(x, y) = Ix - yI is a in part of the proof. However, there is one important part of
metric on X. In this setting, the concept of a subset U of this theorem that is still valid; namely, if a subset K of a met-
X being relatively open in X is consistent with U being an ric space (X, d) satisfies the Bolzano-Weierstrass property,
open subset of X. then K is compact. The proof of this very important result is
The triangle inequality (d) is crucial in the definition outlined in Miscellaneous Exercise 7.
The first two exercises involve the geometric and euclidean metric structure of IB". For n > 2,
R" _ {(xt, .. , x") xi E R, i = 1, .. , n). For p = (pt, , p"), q = (qi, . q") in I2" and
c E R. define p + q = (pt + qt, .. , p" + q"), and cp = (cpt..... cp.).
Also. let 0 = (0, ... , 0). For p, q E R", the inner product of p and q, denoted (p, q), is definedas
(p,q)=ptq.+"+p"q"
1. Prove each of the following. For p, q. r E P8",
a. (p, p) a 0 with equality if and only if p = 0.
b. (p, q) _ (q, p).
c. (ap + bq, r) = a (p, r) + b (q, r) for all a, b E

d. 1(p, q)I z ( (9
This last inequality is usually called the Cauchy-Schwarz inequality. As a hint on how to prove part (d), for
A E R, expand (p - Aq, p - Aq) and then choose A appropriately. Note that by part (a). (p - Aq, p - Aq) z 0 for
allAER.
2. For p = (Pi.... ,p") E R', set lIP112 = )P)P
=\p . The quantity IIPII: is called the norm or
the euclidean length of the vector p.
a. Use the result of Exercise 1(d) to prove that Ilp + g112 IIPI2 + I1g112 for all p, q E R".
b. Using the result of part (a), prove that d2(p, q) = lip - q4' is a metric on R".
c. In R2 sketch the 1-neighborhood of 0.
3. For p=(pi.. .,p.,)and in R", set
d, (p,q)_ 1p:-q,1.
a. Prove that d, is a metric on R".
b. In R2 sketch the 1-neighborhood of 0.
c. Suppose that {pk} is a sequence in R", where for each k E N. pk = (pk,,, ... , pk."). Prove that the sequence
{pk} converges top = (p ... , p") if and only if lim pk.; = p; for all i = 1, ... , n.
d Prove that a sequence {pk} in R" converges to p E R' with respect to the metric d, if and only if it converges to
p with respect to the metric d2. Specifically, prove that lim d,(pk, p) = 0 if and only if Alimd2(pk, p) = 0.
4. a. If (X, d) is a metric space, prove that p(x, y) = d(x, v)/(I + d(x, y)) is also a metric on X.
b. Prove that a subset U of X is open in (X, d) if and only if it is open in (X, p).
5. If E is an uncountable subset of R. prove that some point of E is a limit point of E. (Hint: Use Exercise 17 of Sec-
tion 3. 1).
6. Let {D"} be a countable collection of dense open subsets of R. Prove that (1 , O" is dense in R.
The following exercise is designed to prove the converse of Theorem 3.2.6; namely, if K is a subset
of a metric space (X, d) having the properly that every infinite subset of K has a limit point in K.
then K is compact.
7. Let K be a subset of a metric space (X. d) that has the property that every infinite subset of K has a limit point in K.
a. Prove that there exists a countable subset D of K which is dense in K. (Hint: Fix n E N. Let p, E K be arbi-
trary. Choose p2 E K, if possible, such that d(p,, P2) ? Suppose pl.. . . , pi have been chosen. Choose
pj+,, if possible, such that d(pi, pj ) >_ in for all i = 1, ... , j. Use the assumption about K to prove that this
process must terminate after a finite number of steps. Let 91 denote this finite collection of points, and let
D = U "eN P". Prove that D is countable and dense in K.)
b. Let D be as in (a), and let U be an open subset of X such that u n K * 0. Prove that there exists p E D and
n E N such that N,t"(p) C U.
c. Using the result of (b), prove that for every open cover qt. of K, there exists a finite or countable collection
{ U")" C °U, such that K C U Un.
d. Prove that every countable open cover of K has a finite subcover. (Hint: If is a countable open cover of
K, for each n E N. let W. = U, U,. Prove that K C W. for some n E N. Assume that the result is false, and
obtain an infinite subset of K with no limit point in K, which is contradiction.)
Asic, M. D. and Adamovic, D. D., "Limit points of se- Geissinger, Ladner, "Pythagoras and the Cauchy-
quences in metric spaces," Amer. Math. Monthly 77 Schwarz inequality," Amer. Math. Monthly 83
(1970), 613-616. (1976), 40-41.
Corazza, P., "Introduction to metric-preserving func- Kaplansky, Irving, Set Theory and Metric Spaces.
tions:' Amer. Math. Monthly 106 (1999). 309-323. Chelsea Publ. Co., New York, 1977.
Dubeau, Francis, "Cauchy-Bunyakowski-Schwarz in- Kraft, R. L., "A golden Cantor set." Amer. Math.
equality revisited," Amer. Math. Monthly 99 (1990), Monthly 105 (1998), 718-725.
419-421. Labarre, Jr., A. E., "Structure theorem for open sets of
Espelie, M. S. and Joseph, J. I-, "Compact subsets of the real numbers;' Amer. Math. Monthly 72 (1965),
Sorgenfrey line," Math. Mag. 49 (1976), 250-251. 1114.
Fleron, Julian F., "A note on the history of the Cantor set Nathanson, M. B., "Round metric spaces," Amer. Math.
and Cantor function;' Math. Mag. 67 (1994),136-140. Monthly 82 (1975), 738-741.
Limits and Continuity
4.1 Limit of a Function
4.2 Continuous Functions
4.3 Uniform Continuity
4.4 Monotone Functions and Discontinuities
The concept of limit dates back to the late seventeenth century and the work of Isaac
Newton (1642-1727) and Gottfried Leibniz (1646-1716). Both of these mathemati-
cians are given historical credit for inventing the differential and integral calculus. Al-
though the idea of limit occurs in Newton's work Philosophia Naturalis Principia
Mathematica of 1687, he never expressed the concept algebraically; rather, he used the
phrase "ultimate ratios of evanescent quantities" to describe the limit process involved
in computing the derivatives of functions.
The subject of limits lacked mathematical rigor until 1821 when Augustin-Louis
Cauchy (1789-1857) published his Cours d'Analyse in which he offered the following
definition of limit: "If the successive values attributed to the same variable approach
indefinitely a fixed value, such that finally they dijf°er from it by as little as desired, this
latter is called the limit of all the others." Even this statement does not resemble the
modem delta-epsilon version of limit given in Section 4.1. Although Cauchy gave a
strictly verbal definition of limit, he did use epsilons, deltas, and inequalities in his
proofs. For this reason, Cauchy is credited for putting calculus on the rigorous basis we
are familiar with today.
Based on the previous study of calculus, the student should have an intuitive notion
of what it means for a function to be continuous. This most likely compares to how
mathematicians of the eighteenth century perceived a continuous function; namely, one
that can be expressed by a single formula or equation involving a variable x. Mathe-
maticians of this period certainly accepted functions that failed to be continuous at a fi-
nite number of points. However, even they might have difficulty envisaging a function
115
116 Chapter 4 Limits and Continuity
that is continuous at every irrational number and discontinuous at every rational num-
ber in its domain. Such a function is given in Example 4.2.2(g). An example of an in-
creasing function having the same properties will also be given in Section 4 of this
chapter.
4.11 Limit of a Function

The basic idea underlying the concept of the limit of a function f at a point p is to study
the behavior off at points close to, but not equal to, p. We illustrate this with the fol-
lowing simple examples. Suppose that the velocity v (ft /sec) of a falling object is given
as a function v = v(t) of time t. If the object hits the ground in t = 2 seconds, then
v(2) = 0. Thus to find the velocity at the time of impact, we investigate the behavior of
v(t) as t approaches 2, but is not equal to 2. Neglecting air resistance, the function v(t)
is given as follows:
32t, 05t<2,
(t) 0, t ? 2.
Our intuition should convince us that v(t) approaches -64 ft/sec as t approaches 2. and
that this is the velocity upon impact.
As another example, consider the function f(x) = x sin x * 0. Here the function
f is not defined at x = 0. Thus to investigate the behavior off at 0 we need to consider
the values f(x) for x close to, but not equal to 0. Since
I
V(x) I = jx sin x 5 kxi
for all x * 0, our intuition again should tell us that f(x) approaches 0 as x approaches
0. This is indeed the case as we will see in Example 4.1.10(c).
We now make this idea of f(x) approaching a value L as x approaches a point p
precise. To make the definition meaningful, we must require that the point p be a limit
point of the domain of the function f.
4.1.1 DEFINITION Let E be a subset of 68 and f a real-valued function with domain E.

Suppose that p is a limit point of E. The function f has a limit at p if there exists a num-
ber L E Q8 such that given any e > 0, there exists a 8 > O for which
1f(x)-LI <e
for all points x E E satisfying 0 < Ix - pI < S. If this is the case, we write
1 m f(x) = L or f(x) - L as x -> p.
The definition of the limit of a function can also be stated in terms of e and 6 neigh-
borhoods as follows: If E C R, f : E --- R, and p is a limit point of E. then
lim f (x) = L
XP
if and only if given e > 0, there exists a S > 0 such that

f(x) E N,(L) for all x E Efl (Na(p)\{p}).
This is illustrated graphically in Figure 4.1.
L+E
Lt
Figure 4.1 lim f(x) = L

x-p
Remarks
(a) In the definition of limit, the choice of S for a given e may depend not only on e
and the function, but also on the point p. This will be illustrated in Example 4.1.2(g).
(b) If p is not a limit point of E, then for S sufficiently small, there do not exist any
x E E so that 0 < Ix - pI < S. Thus if p is an isolated point of E, the concept of the
limit of a function at p has no meaning.
(c) In the definition of limit, it is not required that p E E, only that p is a limit point of
E. Even if p E E, and f has a limit at p, we may very well have that
lf(X) *f(p).
This will be the case in Example 4.1.2(c).
(d) Let E C R and p a limit point of E. To show that a given function f does tlt have
a limit at p, we must show that for every L E R, there exists an e > 0, such that for
every S > 0, there exists an x E E with 0 < Ix - pI < S, for which
Lf(x)-Llae.
We will illustrate this in Example 4.1.2(e).
4.1.2 EXAMPLES
(a) Let E be a nonempty subset of 1R and let f, g, and h be functions on E defined by

f (x) = c (c E 18), g(x) = x, and h(x) = x2, respectively. If p is a limit point of E. then
lim f(x) = c, lim g (x) = p. lim h(x) = p2.
X-P
These limits are also expressed as limc = c, limx = p, and limx'' = p2.
Even though we may feel that these limits are obvious, theypstill have to be proved.
We illustrate the method of proof by using the definition to prove that lim h(x) = p2.
The proofs of the other two limits are left to the exercises (Exercise 2). Fpor x E E,
Ih(x) - p2I = Ix2 - p2I = Ix - pIIx + pi (IkI + Ipj)Ix - pI
If pl < 1,then kI < IpI + 1. Hence forallxEEwithIx - pl < 1,
Ih(x) - p2I < (2lpl + l)Ix - pI.
This last term will be less than e provided Ix - pl < e/(21 pl + 1). Thus given e > 0,
we choose 8 = min{ 1, E/(2 1 pI + 1)}. With this choice of 8, if x E E with 0 < Ix - pl
< S,we first of all have Ix - pl < 1, and therefore also
Ih(x) - p2I < (2lpl + 1)Ix - pl < (2IPI + 1) (2IPIE+ 1)

= e.
Thus lim x2 = p2.

X" p
(b) For x * 2, let f(x) be defined by

f (X) = X2 -4
x-2
The domain of f is E = (-oo, 2) U (2, oo), and 2 is clearly a limit point of E. We now
show that lim f (x) = 4. For x # 2,
If(x)-4I = Ix-2 -4I = Ix+2-4I = Ix-2I.

Thus given e > 0, the choice S = e works in the definition.
(c) Consider the following variation of (b). Let g be defined on R by
x2 -4
x#2,
g(x) = x - 2
2, x2.
For this example, 2 is a point in the domain of g, and it is still the case that lim g(x) = 4.
However, the limit does not equal g(2) = 2. The graph of g is given in Figure 4.2.
(d) Let E = (-1, 0) U (0, oo). For x e E, let h(x) be defined by
h(x)=
x+1-1
x
g(x), x s 2
Figure 4.2 Graph of g
We claim that l h(x) = Z. This result is obtained as follows: For x # 0,
x+1-1 x+1-ill x+1+1

x x 777-1 + 1/
_ x 1
x( x+1+1) x+I+1
From this last term we now conjecture that h(x) -+ as x --+0. By the above,
h(x) - 12 (
+1 121 - I2( I x +X+
l + 11)
x+1)(1+Vx- -+I) _ -x
2(V, +1+1)2 I -I2(V+
IxI
2( x+1+1)2
For x E Ewe have ( x + 1 + 1)2 > 1, and thus
Ih(x)I < 2I
Given e > 0, let S = e. ThenI for all x e E with 0 < jxI < S,
Ih(x) 21 < ZI < 2 < E,

and thus liim h(x)
(e) Let f be defined on R as follows:
f(x) - 10, x E O,
x le a.
We will show that for this function, lim f(x) fails to exist for every p E R. Fix p E R.
Let L E R and let
e = max( IL - 11, ILI).
Suppose e = IL - 11. By Theorem 1.5.2. for any S > 0, there exists an x E 0 such
that 0 < Ip - xI < S. For such an x,
If(x)-LI=11-LI=e-
lf e = ILI, then by Exercise 6, Section 1.5, for any 8 > 0, there exists an irrational
number x with 0 < Ix - p l < S. Again, for such an x, V (x) - LI = e. Thus with
e as defined, for any 8 > 0, there exists an x with 0 < Ix - pI < S such that
V (x) - LI ? e. Since this works for every L E 6L lim f(x) does not exist.
(f) Let f : 68 -* l be defined by
0' x E Q,
AX) - IX, x Q.
Then lim f (x) = 0. Since L f (x)1 jx i for all x, given e > 0. any S. 0 < 8 e, will
work in the definition of the limit. A modification of the argument given in (e) shows
that for any p 0 0, iim f (x) does not exist. An alternative proof will be provided in
Example 4.1.5(b).
(g) Our final example shows dramatically how the choice of S will generally depend
not only on e, but also on the point p. Let E _ (0, oo) and let f : E - R be defined by
f(x) I
We will prove that for p E. (0, oo),

1
liml = p
-+p x
If x > p/2, then

1-11 Ix-PI 2
x
= <-.
P XP Ix - PI,
Therefore, given e > 0, let S = min{p/2, pee/2}. Then if 0 < Ix - pI < S. x > p/2,
and
x -P < p2Sse.
The S as defined depends on both p and e. This suggests that any S that works for
a given p and a must depend on both p and e. Suppose on the contrary that for a given
e > 0, the choice of 8 is independent of p E (0, oo). Then with e = 1, there exists a
S > 0 such that
z- p1<1
I
for all x, p E (0, oo) with 0 < Ix - p I < S. Since any smaller S must also work, we can
assume that 0 < S < 1. But now if we take p = 18 and x = 8, then 0 < Ix - pI < S,
and thus I f(x) - f(p) I < 1. However,
If(x) -1(P)I (s > 1.

8I S
This contradiction proves that the choice of S must depend on both p and e. 0
Sequential Criterion for Limits

Our first theorem allows us to reduce the question of the existence of the limit of a func-
tion to one concerning the existence of limits of sequences. As we will see, this result
will be very useful in subsequent proofs, and also in showing that a given function does
not have a limit at a point p.
4.1.3 THEOREM Let E be a subset of l1, p a limit point of E, and f a real-valued function
defined on E. Then
lim f (x) = L if and only if lint f (pn) = L
X-+p n-+oc
for every sequence {pn} in E, with pn # p for all n, and lim p = p.
Remark. Since p is a limit point of E, Theorem 2.4.7 guarantees the existence of a

sequence { pn} in E with pn # p for all n E N and pn -- p.
Proof. Suppose xlpiim f (x) = L. Let {pn} be any sequence in E with pn * p for all n
and pn -+ p. Let E > 0 be given. Since x-.p
limf(x) = L, there exists a S > 0 such that
V (x)
(x) - L I < e for all'x E E, 0 < Ix - p I < S. (1)
Since lint pn = p, for the above S, there exists a positive integer no such that
m
0 < Ipn - P I < S for all n ? no.
Thus if n ? n by (1), l f (pn) - LI < e. Therefore, lint f (pn) = L.
R-00
Conversely, suppose f(pn) -4 L for every sequence { pn} in E with pn * p for all
a and pn -+ p. Suppose lim f (x) . # L. Then there exists an e > 0 such that for every
5 > 0, there exists an
x-Px E Ewith 0<Ix-p1 <8andlf(x)-LIae.Foreach
n r= N, take S = ,-,. Then for each n, there exists p,, E E such that
0<IPn-PI <1. and If(Pn)-LIae.

Thus pn -+ p, but {f(pn)) does not converge to L. This contradiction proves the re-
sult. p
The following uniqueness result is an immediate consequence of the previous
theorem.
4.1.4 COROLLARY If f has o limit at p, then it is unique.

Theorem 4.1.3 is often applied to show that a limit does not exist. If one can find
a sequence {p"} with p" -- p, such that {f(p")} does not converge, then lim f(x) does
not exist. Alternatively, if one can find two sequences {p"} and {r"} both converging to
p, but for which
litn .f(p,,) * iim f(rM),
1W -00
then again 1 m f(x) does not exist. We illustrate this with the following two examples.
4.1.5 EXAMPLES
(a) Let E = (0, oo) and f(x) = sin (11x), x E E. We use the previous theorem to show
that
lim sin
x-.0
-
1
does not exist. Let p" = zi + 1 r;. Then
sin(2n + 1)2 = (-1)".
Thus ii f (p,) does not exist, and consequently by Theorem 4.1.3. liim f (x) also does
"7
not exist. The graph of f(x) = sin(1/x) is given in Figure 4.3.
2rr
Figure 4.3 Graph of f(x) = sin(l/x), x > 0

4.1 limit of a Function 123
(b) As in Example 4.1.2(f) let

J0, X E Q,
AX) = lx,
x qE Q.
Suppose p E R, p * 0. Since Q is dense in R, there exists a sequence C Q with
pn * p for all n E N such that pn -> p. Hence, lim 0. On the other hand.
since R \ Q is also dense in R, there exists a sequence
" *00{qn} of irrational numbers with
qn -? p. But then lim
n-+oo
lim q = p. Thus since p * 0, by Theorem 4.1.3,
n- 00
lim f (x) does not exist.
Limit Theorems
4.1.6 THEOREM Suppose E C R, f. g : E -+ P. and p is a limit point of E. If
limf(x) = A and x-4p
lim g(x) = B,
then
(a) xlim [f(x) + g(x)] = A + B,

(b) 1 m f(x)g(x) = AB, and
x
(c) lim = B, provided B * 0.
Proof. For (a) and (b) apply Theorem 4.1.3 and Theorem 2.2.1. We leave the details
to the exercises (Exercise 11).
Proof of (c): By (b) it suffices to show that
1 1
lim g(x)
=B
We first show that since B # 0, g(x) * 0 for all x sufficiently close to p, x * p. Take
e = IBI/2. Then by the definition of limit, there exists a St > 0 such that
11
2
Ig(x) - BI <
for all x E E, 0 < Ix - pI < SI. By Corollary 2.1.4, Ig(x) - BI > Ilg(x)I - IBII.
Thus
forallxEE,0< Ix - pI < Si.

We can now apply Theorem 4.1.3 and the corresponding result for sequences of
Theorem 2.2.1. Let { pn} be any sequence in E with pn -+ p and pn * p for all n. For
the above S1, there exists an no E Rl such that 0 < jp" - pl < S1 for all n ? no. Thus
g(p") * 0 for all n ? n0. Therefore by Theorem 2.2.1(c),
lim
"°° g(p")
1
= -
1
B.
Since this holds for every sequence p" -> p, by Theorem 4.1.3,
lmg(x) = 1
B.
L1
The proofs of the following two theorems are easy consequences of Theorem 4.1.3
and the corresponding theorems for sequences (2.2.3 and 2.2.4). First, however, we give
the following definition.
4.1.7 DEFINITION A real-valued function f defined on a set E is bounded on E if there

exists a constant M such that V(x)l t M for all x E E.
4.1.8 THEOREM Suppose E C R, p is a limit point of E, and f, g are real-valued functions

on E. If g is bounded on E and lim f (x) = 0, then
t-.p
lim f(x)g(x) = 0.
Proof. Exercise 12.
4.1.9 THEOREM Suppose E C R, p is a limit point of E, and f, g, h are functions from E

into R satisfying
g (x) f (x) : h (x) for all x E E.
If lire g (x) =lim h(x) = L, then 1 m f (x) = L.
Proof, Exercise 13.

We now provide several examples to illustrate the previous theorems.
4.1.10 EXAMPLES
(a) By Example 4.1.2(a), lim x = c. Thus, using mathematical induction and Theorem
4.1.6(b), x.+c
lim x" = c" for all n E N. If p(x) is a polynomial !unction of degree n, that is,
.+a,x+ao,
where n is a n o n n e g a t i v e i n t e g e r and a 0 ,
. .. , a" E R with a" * 0, then a repeated ap-
plication of Theorem 4.1.6(a) gives lim p(x) = p(c).
(b) Consider
lim
x'+2x2-2x-4
x-.-2 x2 - 4
4A Limit of a Function 125
By part (a), xlim2 (x' + 2x2 - 2x - 4) = 0 and dint (x2 - 4) = 0. Since the denomi-
nator has limit zero, Theorem 4.1.6(c) does not apply. In this example, however, for
x # -2,
x3+2x2-2x-4 (x+2)(x2-2) x' -2
x2-4 (x+2)(x-2) x-2
Since slim (x - 2) = -4, which is nonzero, we can now apply Theorem 4.1.6(c) to
conclude tat
lim
x3+2x2-2x-4= lim x2-21
x2-4 x--2x-2 2
(c) Let E = R \ {0}, and let f : E -r R be defined by
f (x) = x sin x.
Since I sin (l /x) I <- I for all x E i8, x # 0, and Ii m x = 0, by Theorem 4.1.8
lim x sin 1 = 0.
x-w x
The graph of f(x) = x sin(1/x) is given in Figure 4.4.
-0.2
Figure 4.4 Graph of f(x) = xsin(1/x), x # 0
(d) Let E = (0, oo) and let f be defined on E by

f(t) _ sin t
We now prove that
lim
r-0 -smt= 1.
1
As we will see in the next chapter, this limit will be crucial in computing the derivative
of the sine function. From Figure 4.5, we have
area(LOPQ) < area(sector OPR) < area(DOSR).
Figure 4.5
In terms of t, this gives
2sintcost<2t<2tan:.
Therefore,
sin t < 1
cost <
t cost
Using the fact that lim cos t = 1 (Exercise 5), by Theorem 4.1.9 we obtain
sin t
li m
1
= I.
Limits at Infinity
Up to this point we have only considered limits at points p E N. We now extend the def-
inition to include limits at oo or -oo. The definition of the limit at oo is very similar to
lim f(n) where f : N , R; that is, f is a sequence in R.
4.1.11 DEFINITION Let f be a real-valued function such that Domf tl (a, oo) * ¢ for
every a E P. The function f has a limit at oo if there exists a number L E R such that
given e > 0, there exists a real number M for which
lf(x) - LI < e
for all x E Dom f fl (M, oo). (See Figure 4.6.) If this is the case, we write
lim (x) = L.
X-*00f
Similarly, if Dom f fl (-oo, b) * O for every b E R,

lim f (x) = L
X--00
if and only if given e > 0, there exists a real number M such that
If(x) - LI < e
for all x E Dom f fl (-oo, m).
The hypothesis that Dom f fl (a, oo) * 0 for every a E R is equivalent to saying
that the domain of the function f is not bounded above. If Dom f = N, then the above
definition gives the definition for the limit of a sequence. Readers should convince
themselves that all theorems up to this point involving limits at a point p E R are still
valid if p is replaced by o0 or -oo.
L+E
L
L-E
Figure 4.6 lim f(x) = L

x-t=
4.1.12 EXAMPLES
(a) As our first example, consider the function f(x) = (sin x)/x defined on (0, oo).
Since Isin xl 1,
U(x)I x
for all x E (0, oo). Let e > 0 be given. Then with M = 1/e,
I f (x)I < e for all x > M.
Therefore, 1 (sin x)/x = 0.
(b) For our second example consider f(x) = x sin irx. If we set p" = (n + ;). n E N.
then
f(p") = (n + !)sin(n + ;)ir = (-1)"(n + 2).
Thus the sequence {f(p")}n°_ I is unbounded, and as a consequence, lim x sin zrx does
not exist. x
EXERCISES 4.1
1. Use the definition to establish each of the following limits.
*a. lim (2x - 7) = -3 b6 fim2 (3x + 5) = -1
x _ I
*c. I m d. lim 2x2-3x-4=
l+ x 2 x-+-i
i x3-2x-4 = 5
e..-- I x + 1 - 3 L lim
x2-4 2
2. Use the definition to establish each of the following limits.
a. lim c=C b. lim x = p
x-v x-+v
C. lim x3 = p3 d. lim
x-0 x" = p", n E RI
-v
p>0 x +o x
I
, p>0
v 2Vp
3. For each of the following, determine whether the indicated limit exists in R. Justify your answer!
limWx-
*a x-.° limx2x +- I
b.x-.i I
*C. lim cos- d. lim cos x

X-O
)2 -
e. lim x 14 f. lim (x + lx 1
4. *Define f : (-1, 1) R by
f(x)=x2-x-2
x+l
Determine the limit L of f at -1 and prove, using e and S, that f has limit L at -1.
5. *a Using Figure 4.5, prove that I sin hl -- jhI for all It E R.
b. Using the trigonometric identity I - cos It = 2 sin 2 2, prove that
(i) limo cos h = I.
1 - cosh
00lim h 0.
6. Let E C R, p a limit point of E, and f : E -* R. Suppose there exist a constant M > 0 and L E R such that
[f(x) - LI c Mix - pl for all x E E. Prove that lim
X-P
f(x) = L.
7. Suppose f: E -i R, p is a limit point of E, and limf(x) = L.

x-p
*a. Prove that lint
x-o
If(x)I = ILI
b. If, in addition, f(x) >- 0 for all x E E, prove that lim = VL.
c. Prove that lim (f(x))" = L" for each n E N.
8. Use the limit theorems, examples, and previous exercises to find each of the following limits. State which theo-
rems, examples, or exercises are used in each case.
.a, littt 5x2 + 3x - 2 b. lim

x3 _ x2 + 2
_-'-+ x-1 x-.-t x+1
Ix + 13n
se lint d.
m 2x + 5 xm2 x + 2)
-
x-4 f lim-
.r(
I
Ix+21
'1
1
s
9. 'Suppose f : (a, b) -+ R, p E [a, b], and + m f (x) > 0. Prove that there exists a 8 > 0 such that f (x) > 0 for all
x E (a, b) with 0 < Ix - pi < S.
10. Suppose E C R, p is a limit point of E, and f : E -+ R. Prove that if f has a limit at p, then there exists a posi-
tive constant M and a S > 0, such that [f(x)I s M for all x E E, 0 < Ix - pi < 6.
11. a. Prove Theorem 4.1.6(a).
b. Prove Theorem 4.1.6(b).
12. 'Prove Theorem.4.1.8.
14. Let f, g be real-valued functions defined on E C R and let p be a limit point of E.
a. If limf(x)
xrp
and lim
x-p
(f(x) + g(x)) exist, prove that lim
x-p
g(x) exists.
b. If lim f(x) and lim (f(x)g(x)) exist, does it follow that lim g(x) exists?
x-ap X-P X-P
15. Let E be a nonempty subset of R and let p be a limit point of E. Suppose f is a bounded real-valued function on
E having the property that iiin f(x) does not exist. Prove that there exist sequences (p.) and {q"} in E with
lim p" = lim q" = p such that lim f(p") and lim f(q") exist, but are not equal.
.-CO "-W
16. *Let f be a real-valued function defined on (a, oo) for some a > 0. Define g on by g(t) Prove that
lim f (x) = L if and only if lint g(t) = L.
17. Investigate the limits at oo of each of the following functions defined on (0, oo).
3x2
a. f(x) =
2+ +x ! 1
b. f (x) =
1 + x2
V-47+ 1 d. f(x) - 2x + 3
c. f(x) = X x+l
e. f( x) = AX) = V x - 2x
2NA + 3x
g. f (x) = x cos x h. f(x) = x sin
18. Let f : (a, oo) R be such that lim x f (x) = L where L E R. Prove that lim f (x) = 0.
130 Chapter4 Limits and Continuity
19. Let f : R -> R satisfy f(x + y) = f(x) + f(y) for all.r. v E R. If lim f(.r) exists, prove that
a. lim f(x) = 0, and b. lim f(x) exists for every p E R.
=-v
4.2j Continuous Functions

The notion of continuity dates back to Leonhard Euler (1707-1783). To Euler, a con-
tinuous curve (function) was one that could be expressed by a single formula or equa-
tion of the variable x. If the definition of the curve was made up of several parts, it was
called discontinuous. This definition was sufficient to convey the concept of continuity
in Euler's time as mathematicians were primarily concerned with elementary functions;
namely, functions built up from the trigonometric and exponential functions, and in-
verses of these functions, using algebraic operations and composition.
The more modern version of continuity is credited to Bolzano (1817) and Cauchy
(1821). Both men were motivated to provide a clear and precise definition of continu-
ity in order to prove the intermediate value theorem (Theorem 4.2.11). Cauchy's defin-
ition of continuity was as follows: "The function f (x) will be, between two assigned val-
ues of the variable x, a continuous function of this variable if for each value of x
between these limits, the numerical value [i.e., absolute value] of the difference
f(x + a) - f(x) decreases indefinitely with a'. Even this definition appears strange in
comparison with the more modem definition in use today. Both Bolzano and Cauchy
were concerned with continuity on an interval, rather than continuity at a point.
4.2.1 DEFINITION Let E be a subset of U8 and f a real-valued function with domain E.

The function f is continuous at a point p E E, if for every e > 0, there exists a S > 0
such that
IA(x) -f(P)I < e
for all x E E with Ix - pI < S. The function f is continuous on E if and only if f is
continuous at every point p E E.
This definition can be rephrased as follows: A function f : E - IR is continuous at

p E E if and only if given e > 0, there exists a S > 0 such that
f(x) E NE(f(p)) for all x E N5(p) fl E.
This is illustrated in Figure 4.7.
Remarks
(a) If p E E is a limit point of E, then f is continuous at p if and only if
Lmf(x) =f(p).
X-P
1. Cauchy Course d'Analyse, p. 43.

Figure 4.7
Also, as a consequence of Theorem 4.1.3, f is continuous at p if and only if

f(p)
for every sequence in E with p, -+ p.
(b) If p E E is an isolated point, then every function f on E is continuous at p. This
follows immediately from the fact that for an isolated point p of E, there exists a S > 0
such that N8(p) (1 E = { p}.
We now consider several of the functions given in previous examples, and also
some new examples.
4.2.2 EXAMPLES
(a) Let g be defined as in Example 4.1.2(c), i.e.,

1xZ-4 x-2.
x#2,
g(x)=
2, x=2.
At the point p = 2, lim g(x) = 4 # g(2). Thus g is not continuous at p = 2. However,
if we redefine g at p = 2 so that g(2) = 4, then this function is now continuous at
p=2.
{0. x E O,
(b1 Let f be as defined in Example 4 . 1 . 2(f ): i . e., f (x) =
x x Q
132 Chapter4 Lindta and Continuity
Since
limf(x) = 0 = f(0),
f is continuous at p = 0. On the other hand, since limf(x)
-P fails to exist for every
p * 0, f is discontinuous at every p E Il, p * 0.
(e) The function f defined by
f(x)
- 11, xE0,
0, x VE Q,
of Example 4.1.2(e) is discontinuous at every p E R.

(d) As in Example 4.1.2(g), the function f (x) = 1/x is continuous at every p E (O,x).
Thus, f is continuous on (0, a).
(e) Let f be.defined by
0, x = 0,
f (X)
x sin -x , x:# 0.
By Example 4.1.10(c),
li mf(x) = 0 = f(0).
Thus f is continuous at x = 0.
(f) In this example we show that f (x) = sin x is continuous on R. Let x, y E R. Then
If(s) -.f(x)l = Isiny - sinxl

= 2 cos (y + x) sin 2(y - x) I
2
).
{2lsinI(y-x+
By Exercise 5 of the previous section, I sin h I < Ih 1. Therefore,
If(y) -f(x)I S Iy - xl,
from which it follows that f is continuous on R.
(g) We now consider a function on (0, 1) that is discontinuous at every rational num-
ber in (0, 1) and continuous at every irrational number in (0, 1). For x E (0, 1)
define
0, if x is irrational,
AX) m
= in1, if x is rational with x =
n
in lowest terms.
The graph of f, at least for a few rational numbers, is given in Figure 4.8.
To establish our claim we will show that
lira f(x) = 0
X-P
for every p E (0, 1). As a consequence, since f(p) = 0 for every irrational number
p E (0, 1),f is continuous at every irrational number. Also, since f(p) # 0 when
p r= 0 fl (0, 1), f is discontinuous at every rational number in (0, 1).
8
16
- 1 I 3 I 5 3 7 1 9 S 11 3 13 '7 U
16 8 16 4 16 8 16 2 16 8 16 4 16 8 16
Figure 4.8
Fix p E (0, 1) and let e > 0 be given. To prove that lim f (x) = 0 we need to show
X __*P
that there exists a 8> 0 such that
lf(x)I < E
for all x E Na(p) 11(0, 1), x * p. This is certainly the case for any irrational number x.
On the other hand, if x is rational with x = m/n (in lowest terms), then f (x) = I/n.
Choose n0 E 1\J such that 1/n,, < E. There exist only a finite number of rational num-
bers m/n (in lowest terms) in (0, 1) with denominator less than no. Denote these by
rl, ... , rk, and let
8=min{Ir1-pI:i= 1,.. ,k,r,*p}.
(Note: Since p may be a rational number and thus possibly equal to r, for some
i = 1, ... , k, we take the minimum of fir, - pI) only for those i for which r, # p).
Thus 8 > 0, and if r E 0 fl Na(p) fl (0, 1), r * p, with r = m/n in lowest terms, then
n ? n0. Therefore,
if(r)I < E.
Thus I f (x) I < e for all x E Na(p) fl (0, 1), x # p.

If f and g are real-valued functions defined on a set E, we define the sum f + g,
the difference f - g, the product fg, and the absolute value lfl off on E as follows:
For x E E.
(f + g) (x) =1(x) + 8(x).
(f - 8)(x) = f(x) - 8(x),
(fg) (x) = f(x)g(x),

Ifl(x) = V(X)1-
Furthermore, if g(x) * 0 for all x E E, we define the quotient fig by
\g/(x) g(x)*
More generally, if f and g are real-valued functions defined on a set E. the quotient f/g
can always be defined on E, = {x E E : g(x) * 0}.
As an application of Theorem 4.1.6 we prove that continuity is preserved under the
algebraic operations defined above. The proof that Ifl is continuous whenever f is con-
tinuous is left as an exercise (Exercise 6).
4.2.3 THEOREM If E C R and f, g : E -> 68 are continuous at p E E, then

(a) f + g and f - g are continuous at p, and
(b) f g is continuous at p.
(c) If g(x) * 0 for all x E E, then fig is continuous at p.
Proof. If p is an isolated point of E, then the result is true since every function on E
is continuous at p. If p is a limit point of E, then the conclusions follow from Theorem
4.1.6. Q
Composition of Continuous Functions

In the following theorem we prove that continuity is also preserved under composition
of functions.
4.2.4 THEOREM Let A, B C l18 and let f : A -+ R and g : B - l be functions such

that Range f C B. If f is continuous at p E A and g is continuous at f(p), then
h = g e f is continuous at p.
Proof. Let e > O be given. Since g is continuous at f (p), there exists a S, > 0 such that
lg(y) - g(f(p))I < e for ally E B fl NN,(f(p)). (2)
Since f is continuous at p, for this S,, there exists a S > 0 such that
Lf(x)-f(p)I <S, for all xEAf1N8(p).
Thus, if x E A with Ix - pl < S, by (2)
lh(x) - h(p)I = Ig(f(x)) - g(f(p))I < e.
Therefore h is continuous at p.
4.2.5 EXAMPLES
(a) If p is a polynomial function on R, then by Example 4.1.10(a), lim p(x) = p(c) for
every c E R. Thus every polynomial function p is continuous on i$.`~`
(b) Suppose p and q are polynomials on P and E _ {x E R : q(x) = 01. Then by The-
orem 4.2.3 the rational function r defined on H \ E by
q(x)
r(x)=p(-,
xER\E,
is continuous on H \ E.
(c) By Example 4.2.2(f), f(x) = sin x is continuous on P. Hence if p is a polynomial
function on R, by Theorem 4.2.4, (f o p)(x) = sin(p(x)) is also continuous on R.
Topological Characterization of Continuity

Before considering several consequences of continuity, we provide a strictly topologi-
cal characterization of a continuous function. In more abstract courses this is often
taken as the definition of continuity.
4.2.6 THEOREM Let E be a subset of P and let f be a real-valued function on E. Then f

is continuous on E if and only if f -'(V) is open in E for every open subset V of R.
Proof. Recall (Definition 3.1.14) that a set U C E is open in E if for every p E U
there exists a 8 > 0 such that Na(p) fl E C U.
Suppose f is continuous on E and V is an open subset of R. If f ''(V) we are
done. Suppose p E f -(V). Then f(p) E V. Since V is open, there exists e > 0 such
that N1(f(p)) C V. Since f is continuous at p, there exists a S > 0 such that
f(x) E NE(f(p)) for all x E Na(p) fl E; i.e., Na(p) fl E C f-'(V). Since p E f -t(V)
was arbitrary, f -'(V) is open in E.
Conversely, suppose f -'(V) is open in E for every open subset V of R. Let p E E
and let e > 0 be given. Then by hypothesis f -'(N,(f(p))) is open in E. Thus there ex-
ists a S > 0 such that
E fl Na(p) C f -'(Nf(f(p)));
that is, f(x) E N.(f (p)) for all x E Na(p) fl E. Therefore f is continuous at p. O
4.2.7 EXAMPLES
(a) We illustrate the previous theorem for the function f(x) = \, Dom f = [0, oo).
Suppose first that V is an open interval (a, b) with a < b. Then
1
¢, b < 0,
f-,(V)= [O,b2), as0<b,
(a2, b2), 0 < a.
Clearly 46 and (a2, b2) are open subsets of H and hence also of [0, oo). Although [0, b2)
is not open in R,
[0, b2) = (-b2, b2) fl [0, oo).
Thus by Theorem 3.1.16, [0, b2) is open in [0, oo). If V is an arbitrary open subset of
R, then by Theorem 3.1.13 (or Exercise 17, Section 3.1) V = U 1,,, where is a fi-
nite or countable collection of open intervals. Since f -'(V) = U f (Theorem

1.7.14) and each f is open in [0, oo), f -'(V) is open in [0, oo). Therefore
f(x) _ N/x is continuous on [0, oo).
That f -'(V) is open in 10, x) for every open set V can also be proved as follows:
Let p E f"'(V) be arbitrary. Since f(p) E V and V is open, there exists an open in-
terval I with f(p) E I C V. But then p E f -'(I) C f -(V). By the above, f -'(I) is
open in [0, cc). Thus there exists a 8 > 0 such that N8(p) fl [0, x) C f -'(I) C f - '(V ).
Since p E f -'(V) was arbitrary, f -'(V) is open in (0, oo).
(b) In this example we show that if f : E -> R is continuous on E and V C E is open
in E, then f(V) is not necessarily open in Range f. Consider the function f : R - R
given by
x2, x < 1,
f(x) - {3-2x, x> 1.
Then f is continuous on R and Range f = R (Exercise 10). However, f((-I, 1))
[0, 1), and this set is not open in R. (See Figure 4.9.)
X2 xx> I.
Figure 4.9 Graph of f(x) _
13-2x, >I
Continuity and Compactness

We now consider several consequences of continuity. In our first result we prove that
the continuous image of a compact set is compact. In the proof of the theorem we use
only continuity and the definition of a compact set. An alternative proof using the
Heine-Borel-Bolzano-Weierstrass theorem (Theorem 3.2.9) is suggested in the exer-
cises (Exercise 25).
4.2.8 THEOREM If K is a compact subset of P and if f : K -.> H is continuous on K, then

f (K) is compact.
Proof. Let { V,},EA bean open cover of f(K). Since f is continuous on Kj -'(V.) is
open in K for every a E A. By Theorem 3.1.16, for each a there exists an open subset
U. of l such that
f -'(V,) = K n U,.
We claim that {U,}0EA is an open cover of K. If p E K, then f(p) Ef(K) and thus
f(p) E V. for some a E A. But then p is in j-'(V.) and hence also in U,. Since each
U. is also open, the collection { U«}.eA is an open cover of K. Since K is compact, there
exists a,,. . . , an E A such that
K C ,U U..
I
Therefore,
n n
K =U
j=t
(u.fl K) = Utf
j- -'(V,),
and by Theorem 1.7.14(a),
n
f(K) = Uf(r'(V0,))
Since f (f -'(V.,)) C V,, f (K) C U; _ IV.,. Thus f (K) is compact.
As a corollary of the previous theorem we obtain the following generalization of
the usual maximum-minimum theorem encountered in calculus.
4.2.9 COROLLARY Let K be a compact subset of R and let f : K -+ R be continuous.

Then there exist p, q E K such that
f(q)<_f(x)sf(p) forallxEK.
Proof. Let M = sup{f(x) : x e K}. By the previous theorem f(K) is compact. Thus,
since f(K) is bounded, M < oo. Also, since f(K) is closed, M E f(K). Thus there ex-
ists p E K such that f(p) = M. Similarly for m = inf{f(x) : x E K}.
We now provide examples to show that the result is false if K C 68 is not compact:
that is, not both closed and bounded.
4.2.10 EXAMPLES
(a) Suppose E is a subset of R. If E is unbounded, consider
f(x) = x2
I +x"
Then f is continuous on E, sup { f (x) : x E E} = 1, but f(x) < I for all x e E. To see
that the supremum is 1, we note that since E is unbounded, there exists a sequence {xn}
in E such that x2. --* oo as n -+ oo. But then
lim f(zn) = lim 1.

R-00 n-xl + X 1
Thus if 0 < (3 < 1, there exists an integer n such that f (x.) > p. Hence sup{ f (x) :
x E E} = 1. The graph of f with E = (0, oo) is given in Figure 4.10.
4
1 2 3 4 5 6
E=(0.-)
x2
Figure 4.10 Graph of f(x) = x E [0,oo)
+z2
(b) If E is not closed, let x be a limit point of E that is not in E. Then

1
8(x) _ .
1 + (x - x,)2
is continuous on E, with g(x) < 1 for all x e E. A similar argument shows that
sup{g(x) : x E E} = 1. This is illustrated in Figure 4.11 with E = (0, 1) and xo = 1.
Intermediate Value Theorem

The following theorem is attributed to both Bolzano and Cauchy. Cauchy, however, im-
plicitly assumed the completeness of R in his proof, whereas the proof by Bolzano
(given below) uses the least upper bound property. An alternative proof is outlined in
the miscellaneous exercises.
4.2.11 THEOREM (intermediate Value Theorem) Let f: [a, b) -> R be continuous.

Suppose f(a) < f(b). If y is a number satisfying
f(a) < y < f(b),
then there exists c E (a, b) such that f(c) = y.
0.5
0 E=[0, I) I
Figure 4.11 Graph of g(x) =

I+ (x - 1)2, x E (0, 1)
The statement and conclusion of the intermediate value theorem is illustrated

in Figure 4.12. The theorem simply states that if f is continuous on [a, b] with
f(a) < f(b), and y E 68 satisfies that f(a) < y < f(b), then the graph off crosses the
line y = y at least once.
Proof. Let A = {x E (a, b) : f(x) 5 y}. The set A * 45 since a E A, and A is
bounded above by b. Thus by the least upper bound property, A has a supremum in R.
Let c = sup A. Since b is an upper bound, c s b.
We now show that f(c) = y. Since c = sup A, either c E A or c is a limit point of A.
If c e A, then f(c) < y. If c is a limit point of A, then by Theorem 2.4.7 there exists a se-
quence in A such that x -* c. Since x. E A, f 5 y. Since f is continuous,
f(c) = lim f(x,,) <_ y
Thus in either case, f(c) s y.
Suppose f(c) < y. Then e = 2(y - f(c)) > 0. Since f is continuous at c, for this
e there exists a S > 0 such that
f(c) - e < f(x) < f(c) + e for all x E fl [a, b].
Since f(c) < y, c # b. Thus (c, b] fl N5(c) # 0. But for any x E (c, b] with c < x <
c + S,
1(x) <f(c) + e =f(c) + 2y - 2f(c) = !U(c) + y) < Y.

Figure 4.12 Intermediate Value Theorem
But then x E A and x > c, which contradicts c = sup A. Therefore f (c) = y.

The intermediate value theorem is one of the fundamental theorems of calculus.
Simply stated, the theorem implies that if I is an interval and f : 1 R is continuous.
then f(!) is an interval. Due to the importance of this result we state it as a corollary.
4.2.12 COROLLARY If I C R is an interval and f :I -> R is continuous on 1, then f(1) is

an interval.
Proof. Let s, t e f(I) with s < t, and let a, b E I with a * b be such that f(a) = s
and f(b) = t. Suppose y satisfies s < y < t. If a < b, then since f is continuous on
[a, b], by the intermediate value theorem there exists c E (a, b) such that f(c) = y.
Thus y E f(1). A similar argument also holds if a > b.
There is an alternative way to state the previous corollary using the terminology of
connected sets. If I is a connected subset of R and f : I-+ R is continuous on I, then
f(!) is connected. This result can be proved using only properties of continuous func-
tions and the definition of a connected set (Exercise 28). The corollary now follows
from the fact that a subset of R is connected if and only if it is an interval (Theorem
3.1.18). The proof of Theorem 3.1.18, however, also requires the least upper bound
property of R. Consequently, the supremum property of the real numbers cannot be
avoided in proving Corollary 4.2.12.
The following two corollaries are additional applications of the intermediate value
theorem. Our first result is the proof of Theorem 1.5.3.
4.2.13 COROLLARY For every real number y > 0 and every positive integer n, there exists
a unique positive real number y so that y° = y.
Proof. That y is unique is clear. Let f(x) = x", which by Example 4.2.5(a) is contin-
uous on R. Let a = 0 and b = y + 1. Since (y + 1)" > y, f satisfies the hypothesis of
Theorem 4.2.11. Thus there exists y, 0 < y < y + 1, such that
f(y) = y" = y
4.2.14 COROLLARY If f : [0, 1 ] -* [0, 1 ] is continuous, then there exists y E [0, 1 ] such
that Ay) = y.
Proof. Let g(x) = f(x) - x. Then g(0) = f(0) ? 0 and g(1) = f(1) - I 0. Thus
there exists y E [0, 11 such that g(y) = 0; i.e., f(y) = y. Q
4.2.15 EXAMPLES
(a) In the proof of Theorem 4.2.11 continuity of the function f was required. The fol-
lowing example shows that the converse of Theorem 4.2.11 is false; that is, if a func-
tion f satisfies the intermediate value property on an interval [a, b], this does not im-
ply that f is continuous on [a, b]. Let f be defined on [0, 2] as follows-
f(x) =
1
x
, 0<xs?, or,
- -1, x = 0.
Then f(0) _ -1, f(r) = 1, and for every y, -1 < y < 1, there exists an x E (0,
such that f(x) = y. However, the function f is not continuous at x = 0 (see Figure 4.3).
(b) In this example we show that the conclusion of the intermediate value theorem is
false if the interval [a, b] of real numbers is replaced by an interval of rational numbers.
Let E _ {x E 0 : 0 5 x < 2}, and let f (x) = x2. Then f is continuous on E with
f(O) < 2 < f(2). However, there does not exist r E E such that f(r) = 2. 0
EXERCISES 4.2
1. For each of the following, determine whether the given function is continuous at the indicated point x".
2x2-5x-3 x03, _ Vc-2 x#4,
f(x) h(x)
at x" = 3 x 4 x at x" = 4
o x 6, 3 x= 3, 4 4
1- cos x
`c. g(x) =
x# 0, d. k(x)
- J x2, x 5 2,
at x, =.2
x at x° = 0 1 4 - x, x>2.
0, x = 0,
2. Let f : R -' R be defined by
f(x) _ 8x, when x is rational,
2x2 + 8, when x is irrational.
a. Prove, using e and 8, that f is continuous at 2.
b. Is f continuous at 1? Justify your answer.
3. Let f : R -r R be defined by
f(x) __ x2, x E Q,
x+2, x0Q.
Finds all points (if any) where f is continuous.
4. *Prove (without using Example 4.2.7) that f(x) = V is continuous on [0, oo).
5. Define f : (0, 1 ] - R by
f x+l
x
a. Justify that f is continuous on (0, 1].
b, Can one define f(0) so that f is continuous on [0, 1]?
6. Let E C R, and suppose f : E --+ R is continuous at p e E.
*a. Prove that Ifl is continuous at p. Is the converse true?
b. Set g(x) =, x E E. Prove that g is continuous at p.
7. *Let E C R and suppose f : E-*R is continuous on E. Prove that f" defined by f'(x) _ (f(x))" is continuous on
E for each n E N.
8. a. Prove that f (x) = cos x is continuous on R.
b. If E C R and f : E -+ R is continuous on E, prove that g(x) = cos (fix)) is continuous on E.
9. For each of the following equations, determine the largest subset E of R such that the given equation defines a
continuous function on E. In each case state which theorems or examples are used to show that the function is
continuous on E.
*a. AX) =
x 4x
b. g(x) = sin-
4)5
z
*c. d. k(x) =cos
h(x) x2 + sin x
10. Prove that
x2, x s 1,
f(x) _ { 32x, x> 1,
is continuous on R and that Range f = R.
11. As in Example 4.2.7(a), use Theorem 4.2.6 to prove that each of the following functions is continuous on the
given domain.
1,
a. f(x) = Dom f = (0, oo) b, g(x) = x2, Dom g = R
X
12. Suppose E is a subset of R and f, g : E -+ R are continuous at p E E. Prove that each of the functions defined
below is continuous at p.
*a. max{f, g}(x) = max{f(x), g(x)}, x E E
b. min{f, g}(x) = min{f(x), g(x)}, x E E
c. f'(x) = max{f(x), 0), x E E
13. Suppose f : [a, b] -* R and g : [b, c] - R are continuous on [a, b] and [b. c], respectively. Leth : [a. c] -+R be
defined by
h(x) _ {f
(x a s x s b,
g(x). b < x s c.
Prove that h is continuous on [a, c] if and only if f(b) = g(b).
14. Use the intermediate value theorem to prove that every polynomial of odd degree has at least one real zero.
15. Prove that there exists x E (0, z) such that cos x = x.
16. *Suppose f : [ -1, 1 ] -+ R is continuous and satisfies f(- I) = f(l). Prove that there exists y E [0, 17 such that
f(Y) = f(Y - 1)
17. Suppose f : [0, 1 ] --? R is continuous and satisfies f(0) = f(l). Prove that there exists y E (0, ;] such that
ICY) = f(y + D.
18. *Let E C R and let f : E -a R be continuous. Let F = {x E E : f(x) = 0}. Prove that F is closed in E. Is F neces-
sarily closed in R?
19. Suppose f : (0, 1) -- R is continuous and satisfies f(r) = 0 for each rational number r E (0, 1). Prove that
f(x) = 0 for all x E (0, 1).
20. Let f be a real-valued function on R satisfying f(x + y) f(x) + f(y) for all x, y E R.
a. If f is continuous at some p E R, prove that f is continuous at every point of R.
*b. If f is continuous on R, prove that f(x) = cx for all x E R, where c = f(1).
21. Let E C R and let f be a real-valued function on E that is continuous at p E E. If f(p) > 0, prove that there ex-
ists an a > 0 and a S > 0 such that f(x) 2- a for all x e N$(p) fl E.
22. *Let f : E.- R be continuous at p E E. Prove that there exists a positive constant M and S > 0 such that
lf(x)I sMforallxEEflN6(p).
23. Let f : (0, 1) --> R be defined by
f(x) _ 0, if x is irrational,
- In, if x is rational with x=m/n in lowest terms.
a. Prove that f is unbounded on every open interval I C (0, 1).
b. Use part (a) and the previous exercise to conclude that f is discontinuous at every point of (0, 1).
24. Suppose E is a subset of R and f, g : E -- R are continuous on E. Show that {x E E : f(x) > g(x)} is open in E.
25. *Let K be a compact subset of R and let f : K -a R be continuous on K. Prove that f(K) is compact by showing
that f(K) is closed and bounded.
26. Let E C R and let f be a real-valued function on E. Prove that f is continuous on E if and only iff-'(F) is
closed in E for every closed subset F of R.
27. Let A, B C R and let f : A -- ). R and g : B -a R be functions such that Range f C B.
a. If V C R, prove that (g -f)-'(V) = f-'(g-'(V)).
b. If f and g are continuous on A and B respectively, use Theorem 4.2.6 to prove that g of is continuous on A.
28. Suppose 1 is a connected subset of R and f : I--> R is continuous on I. Prove, using only the properties of continu-
ity and the definition of connected set, that f(1) is connected.
29. *Let K C R be compact and let f be a real-valued function on K. Suppose that for each x E K there exists
e, > 0 such that f is bounded on NE,(x) n K. Prove that f is bounded on K.
30. Let A C R. For p E R, the distance from p to the set A, denoted d(p, A), is defined by
d(p, A) = inf{Ip - xJ : x E A}.
a. Prove that d(p, A) = 0 if and only if p E A.
b. For x, y E R, prove that Id(x, A) - d(y, A) Jx - y1.
e. Prove that the function x -+ d(x, A) is continuous on R.
d. If A, B are disjoint closed subsets of R, prove that
d(x, A)
AX)
(x)
= d(x, A) + d(x, B)
is a continuous function on R satisfying 0 s f(x) 1 for all x E R. and

0, x E A.
f(x) - 11, xEB.
31. Let f be a continuous real-valued function on R satisfying f(0) = I and f(x + v) = f(x)f(v) for all x.y E R.
Prove that f(x) = a` for some a E R. a > 0.
Uniform Continuity
In the previous section we discussed continuity of a function at a point and on a set. By
Definition 4.2.1, a function f : E -i R is continuous on E if for each p E E, given any
e > 0, there exists a S > 0 such that Lf(x) - f(p)I < e for all x E E fl N5(p). In gen-
eral, for a given e > 0, the choice of S that works depends not only on a and the func-
tion f, but also on the point p. This was illustrated in Example 4.1.2(g) for the function
f(x) = 1/x, x E (0, oo). Functions for which a choice of S independent of p is possible
are given a special name.
4.3.1 DEFINITION Let E C P and f : E - R. The function f is uniformly continuous

on E if given e > 0, there exists a S > 0 such that
1f(x) - fCv)I < e

for all x,vEEwith x-v1 <S.
The key point in the definition of uniform continuity is that the choice of S must
depend only on e, the function f and the set E; it has to be independent of any x E E.
To illustrate this, we consider the following examples.
4.3.2 EXAMPLES
(a) If E is a bounded subset of R, then f(x) = .r` is uniformly continuous on E. Since
E is bounded, there exists a positive constant C so that 1x1 C for all x E E. If
x, y E E, then
Lf(x) - f(v)I = 1x2 - y2I = Ix + yI Ix - v1
(Ix1 + yI)Ix - yI 2CIx - v1.
Let e > 0 be given. Take S = e/2C. If x, y E E with Ix - yI < S, then by the above
Lf(x) -f(y)1 s 2CIx - yI < 2C8 =
Therefore, f is uniformly continuous on E. In this example, the choice of S depends
both on e and the set E. In the exercises you will be asked to show that this result is
false if the set E is an unbounded interval.
(b) Let f(x) = sinx. As in Example 4.2.2(f),
Lf(y) -f(x)I < Iy - xI
for all x, y E R. Consequently, f is uniformly continuous on R.
(c) In this example we show that the function f(x) = I /.r, x E (0. oo) is r= uniformly
continuous on (0, oo). Suppose, on the contrary, that f is uniformly continuous on
(0, oo). Then, as in Example 4.1.2(g), if we take e = 1, there exists a S > 0 such that
If(x) - f(Y) I X- y < I

for all x, y E (0,-oo) with Ix - yl < S. Choose no E Rl such that 1/n < S, and for
n E f i set x = 1/n. Then for all n ? no,
X. - X.+1 =I
n-n+1 < n <
but
If(x.) -f(xn+i)I = In - (n + 1)I = 1.

This however contradicts that If(x) - f(y) I < 1 for all x, y E (0. oo) with Ix - yl < S.
Thus f is not uniformly continuous on (0, oo).
The function f(x) = l/x, however, is uniformly continuous on (a, oo) for every
fixed a > 0. Suppose x, y E [a, oo). Then
[f(x) - f(y)I = ly xyx l 5 a21x - yl.

Hence, given e > 0, if we choose S such that 0 < S < a2e, then as a consequence of
the above inequality, if(x) - f(y)I < e for all x, y E [a. oo) with Ix - yi < S.
Lipschitz Functions
Both of the functions in Example 4.3.2(a) and (b), and the function f(x) = 1/x with
Dom f = [a, oo), a > 0, are examples of an extensive class of functions. If E C R, a
function f : E -+ R satisfies a Lipschitz condition on E if there exists a positive con-
stant M such that
[1(x) - f(Y) I < Mix - Y I
for all x, y E E. Functions satisfying the above inequality are usually referred to as Lip-
schitz functions. As we will see in the next chapter, functions for which the derivative
is bounded are Lipschitz functions.. As a consequence of the following theorem, every
Lipschitz function is uniformly continuous. However, not every uniformly continuous
function is a Lipschitz function. For example, the function f(x) = V is uniformly
continuous on (0, oo), but f does not satisfy a Lipschitz condition on [0, oo) (see Ex-
ercise 5).
4.3.3 THEOREM Suppose E C R 'and f : E - R. If there exists a positive constant M

such that
V(x) - f(y)I <- MIx - yI
for all x, y E E, then f is uniformly continuous on E.
Proof. Exercise 1. O
Uniform Continuity Theorem

If the function f does not satisfy a Lipschitz condition on E, then determining whether
f is uniformly continuous on E is much more difficult. The following theorem provides
a sufficient condition on the set E such that every continuous real-valued function on E
is uniformly continuous.
4.3.4 THEOREM If K C R is compact and f : K -+ R is continuous on K, then f is uni-

formly continuous on K.
Proof. Let e > 0 be given. Since f is continuous, for each p E K, there exists a
6P > 0 such that
If(X) - f(p) I < 2 (3)
for all x E K fl N2s,(p)

The collection {Nso(p)},eK is an open cover of K . Since K is compact, a finite num-
ber of these will cover K. Thus there exist a finite number of points P i . . . . . p in K
such that
K C U Ns,(pi).
i=[
Let
S = min{SP, : i = 1, , n).
Then S > 0. Suppose x, y E K with Ix - yi < S. Since x E K, x E N,&(pi) for some
i. Furthermore, since k - yI < S s Sp,
x, y E Nzs,.(Pi)
Thus by the triangle inequality and inequality (3),
LAX) - f(y)I C If(x) - f(pi)I + U(pa) - f(y)I < 2 + 2 = e.
4.3.5 COROLLARY A continuous real-valued function on a closed and bounded interval

[a, b] is uniformly continuous.
The definition of uniform continuity and the proof of Corollary 4.3.5 appeared in
a paper by Eduard Heine in 1872.
4.3.6 EXAMPLE In this example, we show that the properties closed and bounded are both
required in Corollary 4.3.5. The interval [0, oo) is closed, but not bounded. The func-
tion f(x) = x2 is continuous on (0, oo), but not uniformly continuous on [0, co) (Exer-
cise 2). On the other hand, the interval (0, 1) is bounded, but not closed. The function
f(x) = 1/x is continuous on (0, 1), but is not uniformly continuous on (0, 1).
EXERCISES 4.3
2. Show that the following functions are not uniformly continuous on the given domain.
I. c. h(x) = sin x, Dom h = (0, co)
*a. AX) = x2, Dom f = [0, oo) b, g(x) = Dom g = (0, oo)
3. Prove that each of the following functions is uniformly continuous on the indicated set.
*a. AX) 1 + x , x E [0. oo) b. g(x) _ .r'-, x E hi
c. h(x) = x2 xER d. k(x) = cos x, x E R

+
sinx,
e. e(x) = x + 1, x E (0, oo) *1 f(x) = x E (0, I)
4. Show that each of the following functions is a Lipschitz function.
*a. f(x) = , Dom f = [a, oo), a > 0 b. g(x) = x2 + 1, Dom g = (0. oo)
c. h(x) = sin x, Dom h = [a, oo), a > 0 d. p(x) a polynomial, Dom p = [ -a, a], a > 0
5. *a. Show that f(x) _ \ satisfies a Lipschitz condition on [a, oo), a > 0.
b. Prove that V is uniformly continuous on (0, oo).
c. Show that f does not satisfy a Lipschitz condition on (0, oo).
6. Suppose E C R and f, g are Lipschitz functions on E.
a. Prove that f + g is a Lipschitz function on E.
b. If in addition f and g are bounded on E, or the set E is compact, prove that fg is a Lipschitz function on E.
7. Suppose E C R and f. g are uniformly continuous real-valued functions on E.
a. Prove that f + g is uniformly continuous on E.
*b. If, in addition, f and g are bounded, prove that fg is uniformly continuous on E.
c. Is part (b) still true if only one of the two functions is bounded?
8. Suppose E C R and f : E - R is uniformly continuous. If is a Cauchy sequence in E, prove that { f(x )} is a
Cauchy sequence.
9. Let f : (a, b) -+ R be uniformly continuous on (a, b). Use the previous exercise to show that f can be defined at a
and b such that f is continuous on [a, b].
10. Suppose that E is a bounded subset of R and f : E - R is uniformly continuous on E. Prove that f is bounded
on E.
11. Suppose -oo s a < c < b <_ oo, and suppose f : (a, b) - R is continuous on (a. b).
a. If f is uniformly continuous on (a, c) and also on (c, b), prove that f is uniformly continuous on (a, b).
b. Show by example that the conclusion in part (a) may be false if f is not continuous on (a, b).
12. Let a E R. Suppose f is a continuous real-valued function on [a, oo) satisfying lim f(x) = L. where L E R.
Prove that
a, f is bounded on [a, oo), and
b. f is uniformly continuous on (a, oo).
13. Let E C R. A function f : E E is contractive if there exists a constant b, 0 < b < I. such that
If(x) - f(y)I b Ix - yI.
*a. If E is closed, and f : E -* E is contractive, prove that there exists a unique point x E E such that x,,.
(Such a point xo is called a fixed point of f.)
b. Let E _ (0,;]. Show that f(x) = r' is contractive on E. but that I 'does not have a fixed point in E.
14. A function f : Il -> R is periodic if there exists p E R such that f(x + p) = f(x) for all x E R. Prove that a con-
tinuous periodic function on R is bounded and uniformly continuous on R.
44.4 Monotone Functions and Discontinuities

In this section we take a closer look at both limits and continuity for real-valued func-
tions defined on an interval 1 C R. More specifically, however, we will be interested in
classifying the types of discontinuities that such a function may have. We will also in-
vestigate properties of monotone functions defined on an interval I. These functions will
play a crucial role in Chapter 6 on Riemann-Stieltjes integration. First, however, we be-
gin with the right and left limits of a real-valued function defined on a subset E or R.
Right and Left Limits

4.4.1 DEFINITION Let E C I8 and let f be a real-valued function defined on E. Suppose
p is a limit point of E fl (p, oo). The function f has a right limit at p if there exists a
number L E IB such that given any e > 0, there exists a S > 0 for which
Lf (x) - LI < e
for all x E E satisb,ing p < x p
Similarly, if p is a limit point of En( - oo, p), the left limit off at p, if it exists, is de-
noted by f( p- ), and we write
f(p-) = lim.f(x) = limf(x).
x 0, E n (p, p + S) * 4i. If E is an open interval (a, b), -oo < a < b s oo, then
any p satisfying a s p R, then f has a limit at p E lnt(I) if and only if
(a) f (p+) and f (p-) both exist, and
(b) f(p+) =f(p-).
The hypothesis that p E Int(!) guarantees that p is a limit point of both (-oo, p) fl 1
and I fl (p, oo). If p is the left endpoint of the interval 1, then the right limit of f at p
coincides with the limit off at p. The analogous statement is also true if p is the right
endpoint of I.
We also define right and left continuity of a function at a point p as follows.
4A.2 DEFINITION Let E C R and let f be a real-valued function on E. The function f

is right continuous (left continuous) at p E E if for any e > 0, there exists a S > 0
such that
If(x) -f(p)I < e
for all xEEwith p<_x<p+S(p-6<xsp).
Remarks. If p E E is an isolated point of E or is not a limit point of E fl (p, oo), then
there exists a S > 0 such that E fl (p, p + S) = 4,. Thus if f': E -+ R and e > 0 is arbi-
trary, then If (x) - f (p)l < e for all x E. E fl ['p, p + S). Thus every f : E -> R is right
continuous at p. In particular, if E is a closed interval [a, b], then every f : [a, b] -+ R is
right continuous at b. Also, f is left continuous at b if and only if f is continuous at p.
The following theorem, the proof of which is left to the exercises, is an immediate
consequence of the definitions.
4A.3 THEOREM A function f : (a, b) -+R is right continuous at p E (a, b) if and only if
f (p+) exists and equals f (p). Similarly, f is left continuous at p if and only if f (p- )
exists and equals f (p).
Types of Discontinuities
By the previous theorem a function f is continuous at p E (a, b) if and only if
(a) f (p+) and f (p-) both exist, and
(b) f(p+) =f(p-) =f(p).
A real-valued function f defined on an interval I can fail to be continuous at a point
p E 1 (the closure of I) for several reasons. One possibility is that lim f(x) exists but
xZ
either does not equal f (p), or f is not defined at p. Such a function can easily be made
continuous at p by either defining or redefining f at p as follows:
lim f(x).
f(p) = X-P
For this reason, such a discontinuity is called a removable discontinuity. For example,
the function
-4 x #.2,
g(x) x-2'
2, x = 2,
of Example 4.2.2(a) is not continuous at 2 since
1 m g(x) = 4 * g(2).
By redefining g such that g(2) = 4. the resulting function is then continuous at 2. An-
other example is given by f (x) = x sin(] /x), x E (0, oo), which is not defined at 0. If
we define f on [0, oo) by
0, x=0
f(x) i , x > 0,
x sin
x
then by Example 4.2.2(e), f is now continuous at 0.
Another possibility is that f(p+) and f(p-) both exist, but are not equal. This
type of discontinuity is called a jump discontinuity. (See Figure 4.13.)
4AA DEFINITION Let f be a real-valued function defined on an interval I. The function f

has a jump discontinuity at p E Int(1) if f(p+) and f (p-) both exist, but f is not
continuous at p. If p E I is the left (right) endpoint of 1, then f has a jump discontinu-
ity at p if f (p+) (f( p- )) exists, but f is not continuous at p.
Ap+)
AP-)
Figure 4.13 Jump Discontinuity off at p
Jump discontinuities are also referred to as simple discontinuities, or discontinu-

ities of the first kind. All other discontinuities are said to be of second kind. If f( p+ )
and f (p-) both exist, but f is not continuous at p, then either
(a) f(p+) # f(p-), or

(b) f(p+) =f(p-) #f(p)
In case (a) f has a jump discontinuity at p, whereas in case (b) the discontinuity is re-
movable. All discontinuities for which f (p+) or f (p-) does not exist are discontinu-
ities of the second kind.
4.4.5 EXAMPLES
(a) Let f be defined by
x, 0<x-1,
f(x)= 3-x2, x> 1.
The graph off is given in Figure 4.14. If x < 1, then f(x) = x. Therefore,
f(I-)= lint f(x) = limx=
X +F X-1
1 =f(1).
Likewise, the right limit off at 1 is
Figure 4.14
f(1+)=Xlimf(x)=lim(3-x2)=2.
Therefore, f (I -) = f (l) = 1, and f (I +) = 2. Thus f is left continuous at 1, but not
continuous. Since both right and left limits exist at 1, but are not equal, the function f
has a jump discontinuity at 1.
(b) Let [x] denote the greatest Integer function; that is, for each x, [x] = largest
integer n that is less than or equal to x. For example, [2.9] = 2,[3.1) = 3, and
[ 1.5] = -2. The graph of y = [x] is given in Figure 4.15. It is clear that for each n E Z,
lim [x] = n - 1 and lim. [x] = n.
Thus f has a jump discontinuity at each n E Z. Also, since f (n) = [n] = n, f (x) = [x]
is right continuous at each integer. Finally, since f is constant on each interval
(n - 1, n), n E Z, f is continuous at every x E R\7L.
34-
24-
I +
-1 1 2 3
00 -1
Figure 4.15 Graph of [x]
(c) Let f be defined on R by

0, if x<_0,
f(x) = sin if x > 0.
X
Then f(0-) = 0, but f(0+) does not exist. Thus the discontinuity is of second kind.
(d) Consider the function g : R -+ R defined by
g(x) = sin (21rx[x]).
For x E (n, n + 1), n E Z, x[x] = nx, and thus g(x) is continuous on every interval
(n, n + 1), n E Z. On the other hand, for n E 71,
lim sin(2ax[x]) = sin(21rn2) = 0, and
sn+
lim sin(2irx[x]) = sin(2an(n - 1)) = 0.
Since g(n) = sin(21rn2) = 0, g is also continuous at each n E Z. Thus g is a bounded

continuous function on R. The function g, however, is not uniformly continuous on R
(Exercise 7). The graph of g for x E (-4.4) is given in Figure 4.16. 0
Monotone Functions
4.4.6 DEFINITION Let f be a real-valued function defined on an interval I.
(a) f is monotone increasing (increasing, nondecreasing) on I if f(x) < f(y)
for all x, y E I with x < y.
A A
AAIA
AA
I IV TV-11- A V
Figure 4.16 Graph of g(x) = sin(2ax[x]), x E (-4,4)
(b) f is monotone decreasing (decreasing, nonincreasing) on I if f (x) at f (y)

for all x, y E I with x < y.
(c) f is monotone on 1 if f is monotone increasing on I or monotone decreasing
on I.
A function f is strictly increasing on I if f (x) < f (y) for all x, y E I with x < y.
The concept of strictly decreasing is defined similarly. Also, f is strictly monotone
on I if f is strictly increasing on I or strictly decreasing on I. Our main result for mono-
tone functions is as follows.
4A.7 THEOREM Let I C R be an open interval and let f : I -+ H be monotone increasing

on I. Then f (p+) and f (p-) exists for every p E I and
supf(x) inff(x).
=f(p-) :5f(p):5 f(p+) = p<x
x<p
Furthermore, if p,q E I with p < q, then f(p +) 5 f(q-).

Although we stated the theorem for monotone increasing functions, a similar state-
ment is also valid for monotone decreasing functions.
Proof. i+ix p E I. Since f is increasing on 1, {f(x) : x < p, x E I} is bounded above
by f (p). Let
A = sup{ f (x) : x < p, x e I}.
Then A f (p). We now show that
lim f (x) = A.
X- +P
The proof of this is similar to the proof of Theorem 2.3.2. Let e > 0 be given. Since A
is the least upper bound of {f (x) : x < p}, there exists xo < p such that
A-e<f(x,,)5A.
Thus if xo < x < p, A - e < f (x4,) <- f (x) s A. Therefore,
Lf(x) - AI < e
x+p
f(x) = A. Similarly,
f(p) f(p+) = inf{f (x) : p < x, x E l}.
Finally, suppose p < q. Then

f(p+)=inf{f(x):p<x,xE1}sinf{f(x):p<x<q}
sup { f (x) : p < x < q} sup { f (x) : x < q, x E l } = f (q - ).
4.4.8 COROLLARY If f is monotone on an open interval 1, then the set of discontinuities

off is at most countable.
Proof. Let E p E 1: f is discontinuous at p}. Suppose f is monotone increasing
on I. Then
p E E if and only if f (p-) < f(p+).
For each p E E, choose r1, E 0 such that
f(p-) < r, <.f(p+).
If p < q, then f(p+) s f(q-). Therefore, if p. q E E, r, * r,, and thus the function
p -), ro is a one-to-one map of E into Q. Therefore E is equivalent to a subset of 0 and
thus is at most countable.
Construction of Monotone Functions

with Prescribed Discontinuities
We now proceed to show that given any finite or countable subset A of (a, b), there ex-
ists a monotone increasing function f on [a, b] that is discontinuous at each x E A and
continuous on [a, b] \A. We first illustrate how this is accomplished for the case when
A = {a,, ... , a finite subset of (a, b). To facilitate this construction, we define
the unit jump function Ion R as follows.
4.4.9 DEFINITION The unit jump function 1: R -> R is defined by

when x < 0,
1(x) = 10.
1, when x ? 0.
The function I is right continuous at 0 with 1(0+) = 1(0) = 1 and 1(0-) = 0. For
aER,set
0 when x < a,
la(x) = I(x - a) =
1, when x ? a.
Then la has a unit jump at a and is right continuous at a (see Figure 4.17). Suppose
{c,, ... , are positive real numbers. Define f on [a, b] by
f(x) _ ck1(x - ak).
k- l
The reader should verify that the function f is

(a) monotone increasing on [a, b],
(b) continuous on [a, b]\{a,, a2, ... , aj,
Figure 4.17 Graph of 1(x - a)
(c) right continuous at each ak, k = 1, 2, ... , n, and

(d) discontinuous at each ak with f(ak+) - flak-) = ck for all k = 1. 2, ... , n.
That such a function exists f o r any finite subset {a,, ... , of (a, b) is not surprising.
However, that such a function exists for any countable subset A of (a, b) may take some
convincing, especially if one takes A to be dense in [a, b]; e.g., the rational numbers in
(a, b).
4A.10 THEOREM Let a, b e R and let be a countable subset of (a, b). Let
be any sequence of positive real numbers such that 7,0,0. , c converges. Then there ex-
ists a monotone increasing function f on [a, b) such that
(a) f(a) = 0 and f(b) _ c,,,
(b) f is continuous on [a, b]\ {x : n = 1, 2, ...},

(c) for all n, i.e., f is right continuous at all x,,, and
(d) f is discontinuous at each x with
f(xJ
Proof. For each x E [a, b], define
00
AX) _ 7, c 1(x -
=t
Since 0 s c I(x - x E [a, b], we have
00
s (X) = Ckl(X - Xk) S Ck < Ck.
k-t k=t
Thus for each x E [a, b], the sequence {s (x)} of partial sums is monotone increasing
and bounded above, and hence by Theorem 2.7.6 converges. Since
n= 1,2,.. .
for all x, y with x < y, f is monotone increasing on (a, b). Furthermore, since x > a for
all n, 1(a - 0 for all n. Therefore f(a) = 0. Also, since 1(b - I for all n,
f(b) = 7, C.
k-l
This proves (a).

We now prove (b). Fix p E [a, b), p # x for any n. Let E = {x : n E N}. There
are two cases to consider.
(1) Suppose p is not a limit point of E. If this is the case, there exists a S > 0 such
that NB(p)flE= ¢. Then
I(x-xk)=1(p-xk) forallxE(p-6,p+6)
and all k = 1, 2, .... Thus f is constant on (p - 8, p + 8) and hence continuous.
(ii) Suppose p is a limit point of E. Let e > 0 be given. Since the series 7, 1 Ck
converges, by the Cauchy criterion there exists a positive integer N such that
Ck < E.
k-N+1
Choose S such that

0 < S < min{1p - n = 1, 2, ... , N}.
With this choice of 8, if xk e N6(p) fl E, we have k > N. Suppose p < x p, we always have
021(x-xk)-1(p-xk): 1 forallk6N.
Therefore, for p < x < p + S,
00
0 t f(x) - f(P) 5 '0 Ck(I(x - xk) - 1(P - xk)) 5.7, Ck < E.

k-N+l k=N+1
Thus f is right continuous at p. Similarly, f is left continuous at p, and therefore f is

continuous at p.
For the proof of (c), fix an x E E. If x is an isolated point of E, then as above,
there exists a S > 0 such that En (x,,, x + S) = ¢. Therefore, f(y) = for all
y, x < y < x + S. Thus f(x ). Suppose x is a limit point of E. Let e > 0
be given. Again, choose a positive integer N so that
I00
kN+l
Ck < E.
As in (b), there exists a S > 0 such that if xk. E (x,,, x,, + S) fl E, then k > N. Thus
a
0 s f(y) - 7, ck < e for all y E (x,, x + S).
k=N+l
Therefore, f is right continuous at each x,,.

For the proof of (d), suppose y < x,,. Again, if x is an isolated point of E, there ex-
ists a S > 0 such that (x - S, n E = 4. Therefore, for all k * n,
for all y < xn,
I(xn-xn)=1(0)= 1.
Therefore,
f(xn) - f(y) = cn for all y, x - S < y < x,,.

Suppose x is a limit point of E. Given c > 0, choose N such that N+I ck < e. For
this N, choose S > 0 such that if xk E (x - S, fl E then k > N. Then for all
yE[a,b] with <x,,
C. f(xn) - f(y) < C. + 7, ck < C. + e.

k-N+1
Therefore, f(xn) - f(xn-) = Cn O

4A.11 EXAMPLES
( a ) T a k e c = 2', x = 1 = 1/(n + 1), n = 1, 2, ... , and (a, b) = (0, 1). As in
Theorem 4.4.10, let
00
Ax) _ 71
n-1
In this example, the sequence {xn} satisfies 0 < xt < x2 < < I. If 0 < x <
xt = 1, then I(x - 0 for all n. Thus
f(x)=0, xE [0'2).
If x1 x < x2 = 3, then !(x - x1) = 1 and !(x - xk) = 0 for all k =- 2. Therefore,
f(x)=c1=2, X
If x2 5 x < x3 = 4, then I(x - xk) = 1 for k = 1,2 and I(x - xk) = 0 fork a 3.
Therefore,
1 34, J23. 34
AX) +c2= xE ,
2+
and so forth. The graph off is depicted in Figure 4.18.
7
8
0-0
4
.-o
2
S 0
0 I 2 3 4
2 3 4 S
Figure 4.18
(b) Let c" = 2-", and let {x"} be an enumeration of the rationals in (0, 1). Theorem
4.4.10 guarantees the existence of a nondecreasing function on [0. lj, which is discon-
tinuous at each rational number in (0, 1). and continuous at every irrational number in
(0, 1).
(c) If in the proof of Theorem 4.4.10, we take to be a countable subset of R
and choose {c"}"EN (c" > 0) such that Z'n'o- , c" = 1, then we obtain a nondecreasing,
real-valued function f on R satisfying lie f(x) = 0 and lim f(x) = 1. (See Exercise
20.) Such a function is called a distribution function on d. Snch functions arise natu-
rally in probability theory.
Inverse Functions
Suppose f is a strictly increasing real-valued function on an interval I. Let x, y E I with
x 0 y. If x < y, then since f is strictly increasing. f(x) < f(y). Similarly, if x > y then
f(x) > f(y). Thus f(x) # f(y) for any jr, y E I with x # y. Therefore, f is one-to-one,
and consequently has an inverse function f -' defined on f(1). In the following theorem
we prove that if f is continuous on 1, then f -' is also continuous on f(1).
4.4.12 THEOREM Let I C R be an interval and let f : I-+ R be strictly monotone and
continuous on !. Then f -' is strictly monotone and continuous on J = f(1).
Proof. Without loss of generality, we consider the case where f is strictly increasing
on I. Since f is continuous, by Corollary 4.2.12, f(1) = J is an interval. Furthermore,
since f is strictly increasing on I, f is a one-to-one function from I onto J. Hence f
is a one-to-one function from J onto I.
Suppose y,, y2 E J with y, < y2. Then there exist distinct points x,, x2 E I such
that f(x;) = y;, i = 1, 2. Since f is strictly increasing, we have x, < x2. Thus f -'(y,) <
f -' (y2); i.e., f ` is strictly increasing.
It remains to be shown that f ` is continuous on J. We first show that f `' is left

continuous at each y, E J for which (-oo, yo) fl J * 0. This last assumption means
only that y, is not the left endpoint of J. Let xa E I be such that f(xo) = y0. Then
(-oo, x,) n l * ¢. Let e > 0 be given. Without loss of generality, we can assume that
e is sufficiently small so that x, - e E I. Since f is continuous and strictly increasing,
f((xo - e, xo]) = (f(xo - e),f(xo)) = (yo - 5, yo]
where S > 0 is given by S = f(x,) - f(x, - e). Thus since f-' is strictly increasing,
x, - E= f <f-'(y) <f-'V.)=x,
for all y E (y, - 8, y,) (see Figure 4.19). Hence,
1f -'(yo) - f -'(y)l < e for ally E (y0 - s, yo]
Figure 4.19
Therefore, f "' is left continuous at y,. A similar argument also proves that f -' is right
continuous at each y, E J that is not the right endpoint of J. Thus f -' is continuous at
each y, E J.
For a strictly increasing function f on an open interval I, an alternative proof of the
continuity of the inverse function f `' is suggested in Exercise 14.
4.4.13 EXAMPLE The function f(x) = x2 is continuous and strictly increasing on I = [0, oo)
with J = f(I) = [0, oo). Thus the inverse function f `'(y) = N /y is continuous on
(0, oo). As a consequence, the function g(x) = N/x is continuous on (0, oo). Applying
the same argument to f (x) = x" shows that the function g(x) = 'Clx is strictly increas-
ing and continuous on (0, oo). 0
Remark. In the statement of Theorem 4.4.12 we assumed that f was strictly mono-
tone and continuous on the interval 1. The fact that f is either strictly increasing or
strictly decreasing on I implies that f is one-to-one on the interval 1. Conversely, if f is
one-to-one and continuous on an interval 1. then as a consequence of the intermediate
value theorem the function f is strictly monotone on I (Exercise 15). This, however, is
false if either f is not continuous on the interval 1, or if Dom f is not an interval. (See
Exercises 16, 17, and 18.)
EXERCISES 4.4
2. For each of the following functions f defined on R \ {0}, find ,4 m_ f(x) and -Ii m, f(x), provided the limits exist.
Ix[x]. X < 0,
a. f(x) = Ixl *b. f(x) _ [X] X>0
X
C. f(x) = [I - x2] d. f(x) = [x2 - 1]
e. f(x) _ [ X1, *L f(x) = x[ x

3. For each of the functions f in Exercise 2, determine whether f has a removable discontinuity, a jump discontinu-
ity, or a discontinuity of second kind. at x = 0. If f has a removable discontinuity at 0. specify how f(0) should
be defined so that f is continuous at 0.
4. *Investigate continuity of g(x) = (x - 2) [x) at x = 2.
5. Let f(x) = x - [x]. Discuss continuity of f. Sketch the graph of f.
6. *Let f be defined by
f(x) = _ {x x2[x}, x < -2.
+ b, x > -2.
Determine the value of b such that f has a removable discontinuity at x = -2.
7. *Prove that g(x) = sin(2rrx[x]) is not uniformly continuous on R.
8. Prove that the function f of Example 4.4.11(a) is continuous at x = 1.
9. Let E C Fl and let f be a real-valued function on E. Suppose p E R is a limit point of E fl (p, oo). Prove that
lim f(x) = L if and only if 1 m f(p") = L for every sequence {p"} in E with p" > p for all n E N and p" --> p.
X-P
10. Let f be a real-valued function on (a, b].
a. If f is continuous on (a, b} and lim f(x) exists. prove that f is uniformly continuous on (a, b].
b. If f is uniformly continuous on (a, b), prove that lint f(x) exists.
.' a-
11. Let f : [0, 2] - R be defined by
X, 05x 1,
f(x) - 1 + x2, 1 < x 2.
Show that f and f -t are strictly increasing and find f([0, 2]). Are f and f continuous at every point of their
respective domains?
12. *If m E 7L, n E N, prove that f(x) = x'"1" is continuous on (0, oo).
13. a. If f and g are monotone increasing functions on an interval I. prove that f + g is monotone increasing on I.
b. If in addition f and g are positive on 1, prove that fg is monotone increasing on I.
c. Show by example that the conclusion in part (b) may be false if f and g are not both positive on I.
14. Let I C R be an open interval and let f : I -+ R be strictly increasing and continuous on I.
a. If U C I is open, prove that f(U) is open.
b. Use (a) and Theorem 4.2.6 to prove that f'' is continuous on f(I).
13. Let I C R be an interval and let f be a one-to-one continuous real-valued function on 1. Prove that f is strictly
monotone on I.
16. Let f : [0,1 ) -' R be defined by
2x, Osx<2.
AX)
3-2x, ?5xsl.
a. Sketch the graph of f.
b. Show that f is one-to-one on [0, 1 ], but not strictly monotone on [0, 1 ].
c. Show that f([0.. I ]) = [0, 2).
d. Find f-' (y), y E [0, 2]. and show that f-' is not continuous at y, = 1.
17. Let E = [0, 11 U (2, 3), and for x E E set
x2, Q :!:-: x < 1,

f(x) - I 4-x, 25x<3.
a. Sketch the graph of f.
b. Show that f is one-to-one and continuous on E.
e. Show that f(E) _ (0, 2].
d. Find f-' (y), y g [0, 2], and show that f`' is not continuous at y, = 1.
18. Let f : [0, 1] -+ R be defined by
x, x E dQ,
f(x) - S ll-x,
Prove that f is one-to-one, that f([0, 1 ]) _ [0, 1], but that f is not monotone on any interval 1 C [0, 11.
19. Let ! C R be an interval and let f : I -+ R be monotone increasing. For p E Int(I), the jump of f at p, denoted
Jf(p), is defined by Jf(p) = f(p+) - f(p -)
If p is the left endpoint, set Jf( p) = f( p +) - f(p), and if p is the right endpoint, set Jf(p) = f(p) - f(p -).
a. Prove that f is continuous at p E I if and only if Jf(p) = 0.
b. If p e Int(I), prove that Jf(p) = inf{ f(y) - f(x) : x < p < y, x, y E I}.
20. Let be a countable subset of R and a sequence of positive real numbers satisfying E c = 1. Let
f:R -iRbe defined by
f(x) _ C. !(x -
Prove that lint f(x) = 0 and X__

lim f(x) = 1.
NOTES
The limit of a function at a point is one of the fundamen- a unique positive real number y such that y" = x. Even
tal tools of analysis. Not only is it crucial to continuity, though the existence of nth roots can be proved without
but also to many subsequent topics in the text. The limit the intermediate value theorem, any such proof is simply
process will occur over and over again. We will encounter the statement that the function f(x) = x" satisfies the in-
it in the next chapter in the definition of the derivative and termediate value property on [0, a] for every a > 0. Other
again in the chapters on integration, series, etc. applications of the intermediate value theorem will occur
Another very important concept that will be encoun- elsewhere in the text.
tered often in the text is uniform continuity. Uniform con- The proof of the intermediate value theorem de-
tinuity is important in that given e > 0, it guarantees the pended on the fact that the connected subsets of R are the
existence of a S > 0 such that If(x) - f(y) I < e for all intervals (Theorem 3.1.18) and that the continuous image
x, y E Dom f with Ix - yI < 6. In Chapter 6 we will use of a connected set is connected (Exercise 28 of Section
this to prove that every continuous real-valued function 4.2). Assuming these two results, the intermediate value
on [a, b], a, b E R, is Riemann integrable on [a, b]. theorem is an immediate consequence as follows: Sup-
Other applications of uniform continuity will occur in pose f is continuous on [a, b]. Let I = f([a, b]). Then I
many other theorems and in the exercises. is connected and thus must be an interval. Thus if y satis-
One of the most important results of this chapter is fies f(a) < -y < f(b), then y E I, and hence there exists
the intermediate value theorem (Theorem 4.2.11). The in- c E [a, b] such that f(c) = y. That the continuous image
termediate value theorem has already been used in Corol- of a connected set is connected follows from the defini-
lary 4.2.13 to prove the existence of nth roots; namely, for tion. However, the proof that the connected subsets of R
every positive real number x and n E 101, there exists are the intervals requires the least upper bound property.
1. Let f be a continuous real-valued function on [a, b] with f(a) < 0 < f(b). Let c, = 2(a + b). If f(c,) > 0, let
c, = } (a + c,). If f(cl) < 0, let c2 = 1(c, + b). Continue this process inductively to obtain a sequence {c"} in
(a, b) which converges to a point c E (a, b) for which f(c) = 0.
2. Let E C R, p a limit point of E, and f a real-valued function defined on E. The limit superior of f at p, denoted
1 m f(x), is defined by
1 m f(x) = anfo sup{ f(x) : x E (N8( p) \ { p}) fl E}.
Similarly, the limit inferior off at p, denoted lim f(x), is defined by
=FP
!,f f (x) = ssu$ inf{}(x) : x e (N8(p) \ { p}) fl }.
x-P
Prove each of the following:
a. lim f (x) <_ L if and only if given e > 0, there exists a S > 0 such that f(x) < L + e for all x E E,
0<Ix-pI<S.
b. lim f(x) z L if and only if given e > 0 and 8 > 0, there exists x e E with 0 L-e.
c. If I m f(x) = L, then for any sequence {xn} in E with x" # p for all n E N. and limx" = P, lim f(x") 5 L.
d. There exists a sequence {x"} in E with x" # p for all n E N, such that and limx" = P,
lim f(xn) = lim f(x).
xp
3. Let X C R and f a real-valued function on X. For p E X, the oscillation of fat p, denoted w(f; p), is defined as
w(f; p) = inf sup{ If(x) - f(y) I : x, y e Na(p) fl X}.
8>0
Prove each of the following:

a. The function f is continuous at p if and only if w(f: p) = 0.
b. For every s E i8, the set {x E X : w(f; x) < s} is open in X.
c. The set {x E X : f is continuous at x} is the intersection of at most countably many sets that are open in X.
4. Find w(f; x) for the functions f of Example 4.1.2(e) and Example 4.2.2(g).
The following set of exercises involves the Cantor ternary function. Let P denote the Cantor ternary set of Sec-
tion 3.3. For each x E (0, 1], let x = .aia2a3 denote the ternary expansion of x. Define N as follows:
oo, if a # I for all is E fJ,
N
min{n : a = 11, otherwise.
Define b = 1a for is < N, and bN = 1, if N is finite. (Note: b E {0,1} for all n.)
5. If x E (0, 1) has two ternary expansions, show that
"b
T, 2
n-i
is independent of the expansion of x.
f(x)= _12
I
The Cantor ternary function f on [0, 1 ] is defined as follows: f(0) = 0, and if x E (0. 1 ] with ternary expan-
sion x = .aja2a3 .
b
, set
where N and b are defined as above.

6. Prove each of the following:
a. f is monotone increasing on [0, 1].
b. f is constant on each interval in the complement of the Cantor set in [0, 1 ].
c. f is continuous on [0, 1 ].
d. f(P) = [0, 1].
e. Sketch the graph of f.
Bryant, J., Kuzmanovich, J. and Pavlichenkov, A., Grabinger, Judith V., "Who gave you the epsilon?
"Functions with compact preimages of compact Cauchy and the origins of rigorous calculus;' Amer.
sets:' Math. Mag. 70 (1997), 362-364. Math. Monthly 90 (1983), 185-194.
Bumcrot, R. and Sheingorn, M., "Variations on continuity: Martelli, M., Dang, M. and Seph, T.. "Defining chaos;'
Sets of infinite limits,' Math. Mag. 47 (1974), 41-43. Math. Mag. 71 (1999), 112-122.
Cauchy, A. L., Cours d'Analyse, Paris, 1821, in Oeuvres Snipes, Ray F., "Is every continuous function uniformly
compldtes d'Augustin Cauchy, series 2, vol. 3, continuous?" Math. Mag. 57 (1994),169-173.
Gauthier-Villars, Paris, 1899. Straffin, Jr., Philip. D., "Periodic points of continuous
Fleron, Julian F., "A note on the history of the Cantor set functions;' Math. Mag. 51 (1978), 99-105.
and Cantor function;' Math. Mag. 67 (1994), Velleman, D. J., "Characterizing continuity," Amer.
136-140. Math. Monthly 104 (1997), 318-322.
Differentiation
5.1 The Derivative
52 The Mean Value Theorem
5.3 L'Hospital's Rule
5.4 Newton's Method
The development of differential and integral calculus by Isaac Newton (1642-1727)

and Gottfried Wilhelm Leibniz (1646-1716) in the mid-seventeenth century constitutes
one of the great advances in mathematics. In the two years following his degree from
Cambridge in 1664, Newton invented the method of fluxions (derivatives) and fluents
(integrals) to solve problems in physics involving velocity and motion. During the same
period, he also discovered the laws of universal gravitation and made significant con-
tributions to the study of optics. Leibniz, on the other hand, whose contributions came
ten years later, was led to the invention of calculus through the study of tangents to
curves and the problem of area. The first published account of Newton's calculus ap-
peared in his 1687 treatise Philosophia Naturalis Principia Mathematica. Unfortu-
nately, however, much of Newton's work on calculus did not appear until 1737, ten
years after his death, in a work entitled Methodus fuxionum et serierum infinitorum.
Mathematicians prior to the time of Newton and Leibniz knew how to compute tan-
gents to specific curves and velocities in particular situations. They also knew how to
compute areas under elementary curves. What distinguished the work of Newton and
Leibniz from that of their predecessors was that they realized that the problems of find-
ing the tangent to a curve and the area under a curve were inversely related. More im-
portantly, they also developed the notation and a set of techniques (a calculus) to solve
these problems for arbitrary functions, whether algebraic or transcendental. In Newton's
presentation of his infinitesimal calculus, he looked upon y as a flowing quantity, or flu-
ent, of which the quantity v was the fluxion or rate of change. Newton's notation is still
in use in physics and differential geometry, whereas every student of calculus learns the
165
166 Chapter5 Differentiation
d (for difference) and f (for sum) notation of Leibniz to denote differentiation and in-
tegration. Many of the basic rules and formulas of the differential calculus were devel-
oped by these two remarkable mathematicians. In the paper A New Method for Maxima
and Minima, and also for Tangents, which is not Obstructed by Irrational Quantities,
published in 1684, Leibniz gave correct rules for differentiation of sums, products. quo-
tients, powers, and roots. In addition to his many contributions to the subject. Leibniz
also disseminated his results in publications and correspondence with colleagues
throughout Europe.
Newton and Leibniz, with their invention of the calculus, had created a tool of such
novel subtlety that its utility was proved for over 150 years before its limitations forced
mathematicians to clarify its foundations. The rigorous formulation of the derivative did
not occur until 1821 when Cauchy provided a formal definition of limit. This helped to
place the theory on a firm mathematical footing. Cauchy's contributions to the rigorous
development of calculus will be evident in both this and subsequent chapters.
In this chapter we develop the theory of differentiation based on the definition of
Cauchy, with special emphasis on the mean value theorem and consequences thereof.
The first section presents the standard results concerning derivatives of functions ob-
tained by means of algebraic operations and composition. In the examples and exercises
we will derive the derivatives of some of the basic algebraic and trigonometric func-
tions. However, throughout the chapter we will assume that the reader is already famil-
iar with standard techniques of differentiation and some of its applications. Therefore,
we will concentrate on the mathematical concepts of the derivative. emphasizing many
of its more subtle properties.
5.1 The Derivative

In an elementary calculus course, the derivative is usually introduced by considering the
problem of the tangent line to a curve or finding the velocity of an object moving in a
straight line. Suppose y = f(x) is a real-valued function defined on an interval [a, b].
Fix p E [a, b]. For x E [a, b], x # p, the quantity
Q(x) = f(x) - f(p)
x-p
represents the slope of the straight line (secant line) joining the points (p,f(p)) and
(x,f(x)) on the graph off (see Figure 5.1). The function Q(x) is defined for all values
of x E [a, b], x # p. The limit of Q(x) as x approaches p, provided this limit exists, is
defined as the slope of the tangent line to the curve y = f(x) at the point (p,f(p)).
A similar type of limit occurs if we consider the problem of defining the velocity
of a moving object. Suppose that an object is moving in a straight line and that its dis-
tance s from a fixed point P is given as a function of t. namely, s = s(t). If t,, is fixed.
then the average velocity over the time interval from to to t, t # to, is defined as
s(t) -
t-r
Figure 5.1
The limit of this quantity as t approaches to, again provided that the limit exists, is taken
as the definition of the velocity of the object at time ta.
Both of the previous examples involve identical limits; namely,
s(t) - s(t°)
limf (x) - A p) and lira
x-p 1-.1, t - t,
These limits,. if they exist, are called the derivatives of the functions f and s at p and t
respectively. The term derivative comes from the French fonction derivee.
5.1.1 DEFINITION Let I C R be an interval and let f be a real-valued function with do-
main I. For fixed p E 1, the derivative off at p. denoted f'(p), is defined to be
Pp) = limf(x)
x-,P x-p
-f(p)
provided the limit exists. If f(p) is defined at a point p E 1, we say that f is differen-
tiable at p. If the derivative f is defined at every point of a set E C 1, we say that f is
differentiable on E.
If p is an interior point of I, then p + h E I for all h sufficiently small. If we set
x = p + h, h # 0, then the definition of the derivative of f at p can be expressed as
+ hh' - f(P)
f'(P) = lt_mf(P
provided the limit exists. This formulation of the derivative is sometimes easier to use.
In the definition of the derivative we do not exclude the possibility that p is an
endpoint of I. If p E I is the left endpoint of 1, then
f(x) - f(p) = f(p + h) - f(p)

P(P) = Rimp. x-p ti! h
provided, of course, that the limit exists. The analogous formula also holds if p E I is
the right endpoint of L In analogy with the right and left limit of a function, we also de-
fine the right and left derivative of a function.
5.1.2 DEFINITION Let I C I be an interval and let f be a real-valued function with do-
main I. If p E I is such that I fl (p, oo) # 0, then the right derivative off at p, de-
noted f' (p), is defined as
f(P + h) - f(P)
f+(P) = lim
h-4O h
provided the limit exists. Similarly, if p E I satisfies (-oo, p) fl I # 4,, then the left de-
rivative off at p, denoted f'_(p), is given by
f(P + h) - f(P)
f'(P)=hli3t- h
provided the limit exists.
Remarks
(a) If p E Int(/), then f'(p) exists if and only if both f"+(p) and f_(p) exist and are
equal. On the other hand, if p E I is the left (right) endpoint of I, then f'(p) exists if
and only if f+(p) (f_(p)) exists. In this case, f'(p) = f+ (P) (.f' (P))
The reader should note the distinction between f+(p) and f'(p+). The first de-
notes the right derivative off at p, whereas the latter is the right limit of the deriva-
tive; i.e.,
f'(P+) = lim f'(-).
Here, of course, we are assuming that f' is defined for all x E (p, p + 8) for some
8>0.
(b) If f is a differentiable function on an interval 1, we will also occasionally use Leib-
niz's notation
d
f(x), "& , or '*,
dx
to denote the derivative of y = f(x).

(c) If f is differentiable on an interval I, then the derivative f'(x) is itself a function on
I. Therefore, we can consider the existence of the derivative of the function f' at a point
p E I. If the function f' has a derivative at a point p E 1, we refer to this quantity as
the second derivative off at p, which we denote f"(p). Thus
hmf'(P + h) -f'(P)
f"(P) = h
In a similar fashion we can define the third derivative off at p, denoted f'"(p) or f(')(p).
In general, for n E N, ft"t(p) denotes the nth derivative off at p. In order to discuss
the existence of the nth derivative off at p, we require the existence of the (n - 1)st de-
rivative off on an interval containing p.
5.1.3 EXAMPLES
(a) In the exercises (Exercise 2) you will be asked to prove that if f(x) = x", n E Z,
then f'(x) = nx"- ' for all x E IB (x * 0 if n is negative). For the function f(x) = xz,
the result is obtained as follows:
z -
,y9 (x +
fi(x) = h = h) = 2x.
A similar computation shows that f"(x) = 2.

(b) Consider f(x) = N /x, x > 0. We first note that for h * 0,
f(x+h)-f(x) -vrx
h
= ( x+h-V)( x+h+V)
h ( x+h+Vx)
x+h+;7x
Since h x -+h = ix, we have
f'(x)=h+ 1
=
1
x + h + vx 2Vx
(c) Consider f(x) = sin x. From the identity
sin(x + h) = sin x cos h + cos x sin h,
we obtain
sin(x + h) - sin x r cos h-I sin h
h = sin x l h I+ cos x f h
By Example 4.1.10(d) and Exercise 5, Section 4.1,
lmi-
sin h
h
=l and lim
cos h - 1 = 0.
Therefore,
sin(x + h) - sin x
f'(x).
LIn h
cos h - 11 [ sin h
= sin x lim i + cos x l
= cos X. J J
In Exercise 3 you will be asked to prove that
d (cos x) = -sin x.
170 Chapters Differentiation
(d) Let f be defined by

x x ? 0,
.f(x) = IxI =
-x, x<0.
Then
h _ h
.f+(0) = lim o h= 1, and
ti h
ii
01
= lim _h
Thus f (0) and f'_(0) both exist, but are unequal. Therefore f'(0) does not exist.
(e) In this example let g(x) = x312 with Dom g = [0, oo). Then for p = 0,
x3/2
8'(0) = 8+(0) =I-,Q
limX- = Jim
.Ho
V = 0.
Thus g is differentiable at 0 with g'(0) = 0.
(f) Let f be defined by
xsin,
x x#0,
x
f(x) =
For x # 0, f'(x) _ -x cos X + sin I (Exercise 8). When x = 0,

h
f(h) - f0
f'(0) _ h m =him sin h,
which by Example 4.1.5(a) does not exist. Therefore, f'(0) does not exist.
g(x) =
II
(g) Consider the following variation of (f). Let
x2 sin x
0,
, x # 0,
x0.
This example is very important! In Exercise 9 you will be asked to show that g'(x) ex-
ists for all x E R with g'(0) = 0, but that the derivative g' is not continuous at 0.
When Cauchy gave the rigorous definition of the derivative, he assumed that the
given function was continuous on its domain. As a consequence of the following theo-
rem this requirement is not necessary.
5.1.4 THEOREM If 1 C R is an interval and f : I -+ R is differentiable at p E 1, then f is

continuous at p.
Proof. Fort # p,
(f(t` _ P(P))(t
f(t) -f(P) = - P)
Since
limi(t) -f(p)
'-+p t-p
exists and equals f'(p), by Theorem 4.1.6(b),
(f(t) - f(P)1Jlim(r - P) = Pp) 0 = 0.
lim(f(t) - f(p)) = lim
r-+p _P t-p r-+p
Therefore, lm f(t) = f(p) and thus f is continuous at p. In the above, if p is an endpoint,

then the limits are either the right or left limit at p, whichever is appropriate.
Remark In both Examples 5.1.3(d) and (f), the given function is continuous at 0, but
n o t d i ff e r e n t i a b l e a t 0. Given a f i n it e number of points, say p i, ... , pn, it is easy to con-
struct a function f which is continuous but not differentiable at pI, .. . , p,,. For example,
n
f(x)= k-I
I Ix-Pk1
has the desired properties. In 1861, Weierstrass constructed a function f which is con-
tinuous at every point of 11 but nowhere differentiable. When published in 1874, this
example astounded the mathematical community. Prior to this time mathematicians
generally believed that continuous functions were differentiable (except perhaps at a
finite number of points). In Example 8.5.3 we will consider the function of Weierstrass
in detail.
Derivatives of Sums, Products, and Quotients

We now derive the formulas for the derivative of sums, products, and quotients of func-
tions. These rules were discovered by Leibniz in 1675.
5.1.5 THEOREM Suppose f, g are real-valued functions defined on an interval L 1f f and

g are differentiable at x E 1, then f + g,fg, and fig (if g(x) # 0) are differentiable at
x and
(a) (f + 8)'(x) = fl(X) + g'(x),
(b) (J)'(x) = f'(x)g(x) +f(x)g'(x)
(c) ()'(x) = fl(x)8(x) - f2x)g'(x)'
provided g(x) # 0.
(g(x))
Proof. The proof of (a) is left as an exercise (Exercise 4). For the proof of (b), by
adding and subtracting the term f (x + h)g(x), we have for h * 0,
(fg)(x + h) - (fg)(x) + h) - g(x)
= f(x + h)(Ax
+ (f(x + h) - f(x)lAX)
h J
By Theorem 5.1.4, since f is differentiable at r, i a f(x + h) = f(x). Thus since each

of the limits exist, by Theorem 4.1.6,
(.fg)(x + h) - Ug)(x)
(fg)'(x) = lim
ti0 h
/ mig(x+hh-g(x)
=ii mf(x+h)ii
+\h)
- f(x)
+ g(x) him/f(x
l h
= f(x)g'(x) + g(x)f'(x).
To prove (c), we first prove that (1/g)'(x) = -g'(x)/[g(x)]2, provided g(x) * 0.
The result then follows by writing f/g as f (1/g) and applying the product formula (b).
Suppose g(x) * 0. By continuity of g at x, there exists a S > 0 such that g(x + h) * 0
for all h, I h i < S. Thus for h sufficiently small and nonzero,
1 1
g(x + h) g(x) - g(x + h) - g(x)' 1

) g(x)g(x + h)
h h
Again, using the fact that lim g(x + h) = g(x), by Theorem 4.1.6,
1 1
1ll' __ lim
g(x + h) g(x)
C-/
8 (x) h-0 h
(g(x + h) - g(x)1
_ -lim
11-0 \
) lim
n-o g(x)g(x + h)
h
-g'(x)
= g2(x) ,
The Chain Rule

The previous theorem allows us to compute the derivatives of sums, products, and quo-
tients of differentiable functions. The chain rule, on the other hand, allows us to compute
the derivative of a function obtained from the composition of two or more differentiable
functions. Prior to stating and proving the result, we introduce some useful notation.
Suppose f is differentiable at x E 1. For t E 1, t # x, set
Q(t) = f(t) - f(x)
t - x
Then by Definition 5.1.1, Q(t) --+f(x) as tax. If we let u(t) = Q(t) - f'(x). then
u(t) -- 0 as t -+ x. Therefore, if f is differentiable at x, fort * x,
f(t) - f(x) = (t - x)[f'(x) + u(t)] where u(t) -+0 as (1)
By setting u(x) = 0, the above identity is valid for all r E I.
5.1.6 THEOREM (Chain Rule) Suppose f is a real-valued function defined on an inter-

val I and g is a real-valued function defined on some interval J such that Range f C J.
If f is differentiable at x E I and g is differentiable at f(x), then h = g of is differen-
tiable at x and
h'(x) = g'(f(x))f'(x)
Proof. Let y = f(x). Then by equation (1),
f(t) - f(x) _ (t - x)[f'(x) + u(t)I. (2)
g(s) - g(y) = (s - y)[g'(y) + v(s)], (3)
where t E I, s E J, and u(t) -> 0 as t -> x and v(s) -+0 as s -+y. Let s = f (t). Since f
is continuous at x, s --* y as t -> x. By identity (3) and then (2),
h(t) - h(x) = g(f(t)) - g(f(x))
[f(t) - f(x)I [g'(y) + v(s)]
= (t - x)[f'(x) + u(t)][g'(v) + v(f(t))].
Therefore, for t * x,
h(t)- hx_
t-x ) = [f'(x) + u(t)N91(y) + v(f(t))).
Since v(f(t)) and u(t) both have limit 0 as t -+ x,
h(t) - h(x) =
lim r(X) g'(y) = g'(f(x))f'(x). O
t-x
To illustrate the previous theorem we consider the following examples.
5.1.7 EXAMPLES
(a) By Example 5.1.3(c) the function f (x) = sin x is differentiable on R. Hence if
g :1 * R is differentiable on the interval 1, h(x) = If o g)(x) = sin g(x) is differentiable
on I with
h'(x) = f'(g(x)) g'(x) = g'(x)cos g(x).
In particular, if g(x) = 1/x2, Dom g = (0, oo), then by Theorem 5.1.5(c),
- 2x 2
g'(x) _ (x2)2 = x3, x E (0, oo).
Therefore,
d
dx
sm-
.
x2
-=-cos-
2
x2
1 1
2.
(b) By Exercise 2,
dx"=nx"-' forallnEfol.
Thus if f : I -+ R is differentiable on the interval 1, then by the chain rule g(x) = [f(x) ]1.
n E N, is differentiable on I with g'(x) = n[f(x)]"-'f'(x). This formula can also be ob-
tained from Theorem 5.1.5(b) using mathematical induction.
EXERCISES 5.1
1. Use the definition to find the derivative of each of the following functions.
*a. f(x)=x;, x E R b. g(x) = x+2, x> -2
X, x # 0
*c. h(x) = 1 d. k(x) = l
'
x > -2
*e. f (x)
x+1
x* -1 f. g(x) = xZ + 1' x E R
2. *Prove that for all integers n,
dx" = nx"-' (x * 0 if n is negative).

3. *a. Prove that
d (cos x) = -sin x.
b. Find the derivative of tan x = sin x/cos x.

4. Prove Theorem 5.1.5(a).
S. For each of the following, determine whether the given function is differentiable at the indicated point x°. Justify
your answer!
x E 0'
*a. f(x) = xIxl, at X. = 0 b. f(x) = atx° = 0
lo0, , x00,
*c. g(x) = (x - 2)[x], at x° = 2 d. h(x) = x + 2, x E [ -2, oo), at x° _ -2
x sin X, xEQ.
e. f()= x, x00. atx
X. =0
6. Let f(x) = lxI3. Compute f'(x), f"(x), and show that f"'(0) does not exist.
7. Determine where each of the following functions from R to R is differentiable and find the derivative.
*a. 1(x) = x(x]
b.g(x)=Ix-21+Ix+tI
*c. h(x) = I sin x I
{x2[].
0, :::"
8. Use the product rule, quotient rule, and chain rule to find the derivative of each of the following functions.
a. f (z) = x sin x*0 b. f(x) = (cos (sin x) ), n, m E h!
c. f(x)= x+1Y d. f(x)=x°(2+sink), x * 0
9. Let g be defined by
g(x) =
x2 sin1,
x
x0,
a. Prove that g is differentiable at 0 and that g'(0) = 0.

b, Show that g'(x) is not continuous at 0.
10. Let f be defined by
f(x) = fx2 + 2, x s 2,
ax+b, x> 2.
a. For what values of a and b is f continuous at 2?
b. For what values of a and b is f differentiable at 2?
ax + b, x < -l,
f(x) = x3+1. -lsx:5 2,
cx+d, x> 2.
Determine the constants a, b, c, and d such that f is differentiable on R.
12. Assume there exists a function L : (0, oo) -+ R satisfying L'(x) = l/x for all x E (0, oo). Find the derivative of
each of the following functions.
a. f(x) = L(2x + 1), x > -2
b.g(x)=L(x2), x#0
*c. h(x) = [L.(x)]', x > 0
d. k(x) = L(L(x)), x E {x > 0 : L(x) > 0}
13. For b real, let f be defined by
1x° sin 1-, x > 0,

f(x) = x
0, x50.
Prove the following:
a. f is continuous at 0 if and only if b > 0.
b. f is differentiable at 0 if and only if b > I.
c. f is continuous at 0 if and only if b > 2.
mfx +h)-f(xo-h)
14. a. If f is differentiable at x, prove that li =fI M
2h
,+h- fx, - h )
b. If Ii f (x exists, is f differentiable at x0?
15. If f : (a, b) -+ R is differentiable at p E (a, b), prove that f'(p) = lim n[f(p + ,l,) - f(p)]. Show by example that
the existence of the limit of the sequence {n[ f(p + f(p)j) does not imply the existence of f'(p).
16. Leibniz's Rule: Suppose f and g have nth order derivatives on (a, b). Prove that
(nk) f (k)(X)e - kkX).

(fg)t11(x) _
4-0
5.21 The Mean Value Theorem

In this section we will prove the mean value theorem and give several consequences of
this important result. Even though the proof itself is elementary, the theorem is one of
the most useful results of analysis. Its importance is based on the fact that it allows us
to relate the values of a function to values of its derivative. We begin the section with a
discussion of local maxima and minima.
Local Maxima and Minima
5.2.1 DEFINITION Suppose E C R and f is a real-valued function with domain E. The

function f has a local maximum at a point p E E if there exists a 8 > 0 such that
f (x) s f(p)for all x E E fl Na(p)- The function f has an absolute maximum at p E E
iff(x) 5 f(p) for all x E E.
Similarly, f has a local minimum at a point q E E if there exists a 8 > 0 such that
f(x) ? f(q) for all x E E fl %(q) and f has an absolute minimum at q E E
if f (x) ? f (q) for all x E E.
Remark. As a consequence of Corollary 4.2.9. every continuous real-valued function

defined on a compact subset K of I has an absolute maximum and minimum on K.
The function f, illustrated in Figure 5.2, has a local maximum at a, p,, and p, and
a local minimum at and b. The points (p4,f(p4)) and (p,.f(p,)) are absolute
maxima and absolute minima, respectively.
Figure 5.2
The following theorem gives the relationship between local maxima or minima of
a function defined on an interval and the values of its derivative.
5.2.2 THEOREM Let fbe a real-valued function defined on an interval 1, and supposef has
either a local minimum or local maximum at p E Int(l). 1f f is differentiable at p, then
f(P) = 0.
Proof. If f is differentiable at p E Int(1), then f'- (p) and f; (p) both exist and are
equal. Suppose f has a local maximum at p. Then there exists a S > 0 such that
f (t) f (p) for all t E I with I t - p I < S. In particular, if p < t < p + S, t E 1, then
f(t) - f(p) s 0.
t-P
Thus f+(P) 0. Similarly, if p - S < t < p,
f(t) - f(P) 0,
t-p
and therefore f'_(p)' 0. Finally, since f+ (p) = f'_ (p) = f'(p), we have f(p) = 0.
The proof of the case wheref has a local minimum at p is similar. O
As a consequence of the previous theorem we have the following corollary.
5.2.3 COROLLARY Lei f be a continuous real-valued function on [a, b]. 1f f has a local
maximum or minimum at p E (a, b), then either the derivative off at p does not exist,
or f(p) = 0.
Remark. The conclusion of Theorem 5.2.2 is not valid if p E I is an endpoint of the
interval. For example, if f :[a, b] -- R has a local maximum at a, and if f is differen-
tiole at a, then we can only conclude that f'(n) = f+ (a) 5 0. This is illustrated in the
following examples.
5.2.4 EXAMPLES
(a) The function
f(x)= x- lZ , 05x2,
has a local maximum at p = 0 and p = 2, and an absolute minimum at q = ;. By com-
putation, we have f'(0) _ -1, f' (2) = 3, and f (2) = 0. The graph of f is given in
Figure 5.3.
(b) The function f (x) _ x 1, x E [ -1, 11, has an absolute minimum at p = 0. How-
ever, by Example 5.1.3(d) the derivative does not exist at p = 0.
Rolle's Theorem
Prior to stating and proving the mean value theorem, we first state and prove the fol-
lowing theorem credited to Michel Rolle (1652-1719).
178 Chapter 5 Differentiation
24J
21
Il
4
Figure 5.3 Graph of f(x) _ (x - 1)2, 0 5 x 5 2
5.2.5 THEOREM (Rolle's Theorem) Suppose f is a continuous real-valued function on

[a, b] with f(a) = f(b), and that f is differentiable on (a, b). Then there exists
c E (a, b) such that f (c) = 0.
Since the derivative off at c gives the slope of the tangent line at (c, f (c)), a geomet-
ric interpretation of Rolle's theorem is that if f satisfies the hypothesis of the theorem, then
there exists at least one value of c E (a, b) for which the tangent line to the graph off is
horizontal. For the function f depicted in Figure 5.4, there are exactly two such points.
Proof. If f is constant on [a, b], then f'(x) = 0 for all x E [a, b]. Thus, we assume
that f is not constant. Since the closed interval [a, b] is compact, by Corollary 4.2.9,f
has a maximum and a minimum on [a. b]. If f(t) > f(a) for some t, then f has a max-
imum at some c E (a, b). Thus by Theorem 5.2.2, f'(c) = 0. If f(t) < f(a) for some t,
then f has a minimum at some c E (a, b), and thus again f'(c) = 0.
Remarks
(a) Continuity off on [a, b] is required in the proof of Rolle's theorem. The function
Ix, 05x< 1,
f(x)= 0, x=1
is differentiable on (0, 1) and satisfies f(0) = f(1) = 0; yet f' (x) * 0 for all x E (0, 1).
The function f fails to be continuous at 1.
(b) For Rolle's theorem, differentiability of f at a and b is not required. For example,
the function f(x) = - xZ, x e [-2, 2], satisfies the hypothesis of Rolle's theo-
rem, yet the derivative does not exist at -2 and 2. For x E (-2, 2),
fi(x) _
V4 - X2
and the conclusion of Rolle's theorem is satisfied with c = 0.
Figure 5.4 Rolle's Theorem
The Mean Value Theorem

As a consequence of Rolle's theorem we obtain the mean value theorem. This result is
usually attributed to Joseph Lagrange (1736-1813).
5.2.6 THEOREM (Mean Value Theorem) If f : [a, b] -+ R is continuous on [a, b] and

differentiable on (a, b), then there exists c E (a, b) such that
f(b) - f(a) =f(c)(b - a).

Graphically, the mean value theorem states that there exists at least one point
c E (a, b) such that the slope of the tangent line to the graph of the function f is equal
to the slope of the straight line passing through (a, f (a)) and (b, f (b)). For the function
of Figure 5.5, there are two such values of c, namely c, and c2.
Proof. Consider the function g defined on [a. b] by
{f(b)b
g(x) = f(x) - f(a) - - a(a),(x - a).
Then g is continuous on [a, b], differentiable on (a, b), with g(a) = g(b). Thus by
Rolle's theorem there exists c E (a, b) such that g'(c) = 0. But
f(b) - f(a)
g'(x) =f'(x) - b - a
f (b)
for all x E (a, b). Taking x = c gives f(c) =
now follows.
- a(a), from which the conclusion
Figure 5.5 Mean Value Theorem
The mean value theorem is one of the fundamental results of differential calcu-
lus. Its importance lies in the fact that,it enables us to obtain information about a
function f from its derivative f'. In Example 5.2.7 we will illustrate how the mean
value theorem can be used to derive inequalities. Other applications will be given
later in this section and in the exercises. It will also be used in many other instances
in the text.
5.2.7 EXAMPLE In this example we illustrate how the mean value theorem may be used
in proving elementary inequalities. We will use it to prove that
x
s ln(1 + x) < x for all x > - 1,
1 + x
where In x denotes the natural logarithm function on (0, oo). This function is defined
and considered in detail in Example 6.3.5 of the next chapter. There it is proved that the
derivative of In x is 1/x. Let f(x) = ln(1 + x), x E (-1, oo). Then f(0) = 0. If x > 0,
then by the mean value theorem, there exists c e (0, x) such that
In(1 + x) = f(x) - f(0) = f(c)x.
But f'(c) _ (1 + c)-' and (I + x)-' < (1 + c)-t < 1 for all c E (0, x). Therefore,
x
1 +x<f'(c)x<x
x
1 +x`ln(1 +x)<x for all xz0.
Now suppose -1 < x < 0. Then again by the mean value theorem there exists c E (x. 0)
such that
ln(1 + x) = f(x) - f(0) =

1 + c,
But since x < c < 0, 1 < (1 + c)-' < (1 + x)-', and since x is negative
< ln(1 + x) < .r.

1 + x
Hence the desired inequality holds for all x > -1, with equality if and only if x = 0.
U
The following theorem, attributed to Cauchy, is a useful generalization of the mean
value theorem.
5.2.8 THEOREM (Cauchy Mean Value Theorem) If f, g are continuous real-valued func-
tions on [a, b] that are differentiable on (a, b), then there exists c E (a, b) such that
[f(b) - f(a)]g'(c) _ [g(b) - 8(a)] .`'(c)
. Proof. Let
h(x) = [f(b) -f(a)]g(x) - [8(b) - g(a)]f(x).

Then h is continuous on [a, b], differentiable on (a, b) with
h(a) = f(b)g(a) - f(a)g(b) = h(b).
Thus by Rolle's theorem, there exists c E (a, b) such that h'(c) = 0, which gives the
result. U
The geometric interpretation of the Cauchy mean value theorem is very similar to
that of the mean value theorem. If g'(x) * 0 for all'x E (a, b), then g(a) 0 g(b) and the
conclusion of Theorem 5.2.8 can be written as
f(b) - f(a) _ f (c)

g(b) - g(a)
guppose.x = g(t), y = f(t), a s t s b, is a parametric representation of a curve C in
the plane. As t moves along the interval [a, b], the point (x, y) moves along C from the
point P = (g(a),f(a)) to Q = (&)J (b)). The slope of the line joining P to Q is given
by [f(b) - f(a)]/[g(b) - g(a)] (see Figure 5.6). On the other hand, the quantity
f'(t)/g'(t) is. the slope of the. curve C at the point (g(r), f(t)). Thus one meaning of
Theorem:5.2.8 is that there must be a point on the curve C where the slope of the curve
is the same as the slope of the line joining P to Q.
Q=(g(b). f(b))
P=(g(a), f(a))
Figure 5.6
Applications of the Mean Value Theorem

We now give several consequences of the mean value theorem. Additional applications
are also given in the exercises. In the following, I will denote an arbitrary interval in R.
5.2.9 THEOREM Suppose f : 1--+ R is differentiable on the interval L

(a) If f'(x) ? O for all x E I, then f is monotone increasing on 1.
(b) If f (x) > O for all x E I, then f is strictly increasing on I.
(c) If f'(x) 5 O for all x E 1, then f is monotone decreasing on L
(d) If f'(x) < O for all x E 1, then f is strictly decreasing on I.
(e) If f'(x) = O for all x E I, then f is constant on I.
Proof. Suppose x1, x2 E I with x, < x2. By the mean value theorem applied to f on
[x1, x2].
f(x2) -f(xl) =f'(C)(x2 - xl)

for some c E (x1, x2). If f(c) ? 0, then f(x2) ? f(x1). Thus, if f'(x) a 0 for all x E 1,
we have f(x2) ? f(xl) for all xj, x2 E I with xi < x2. Thus f is monotone increasing on
I. The other results follow similarly. O
Remark It needs to be emphasized that if the derivative of a function f is positive at

a point c, then this does not imply that f is increasing on an interval containing c. The
function f of Exercise 18 satisfies f'(0) = 1, but f'(x) assumes both negative and
positive values in every neighborhood of 0. Thus f is not monotone on any interval
containing 0. If f'(c) > 0, the only conclusion that can be reached is that there exists
a 8 > 0 such that f(x) < f(c) for all x E (c - S, c) and f(x) > f(c) for all
x E (c, c + 8) (Exercise 15). This, however, does not mean that f is increasing on
(c - 8, c + 8). However, if f'(c) > 0 and f' is continuous at c, then there exists
a 8 > 0 such that f'(x) > 0 for all x E (c - 8, c + 8). Thus f is increasing on
(c-8,c+ 8).
Theorem 5.2.9 is often used to determine maxima and minima of functions as fol-
lows: Suppose f is a real-valued continuous function on (a, b), and c E (a, b) is such
that f'(c) = 0 or f(c) does not exist. Suppose f is differentiable on (a, c) and (c, b). If
f'(x) < 0 for all x E (a, c) and f'(x) > 0 for all x E (c, b), then by Theorem 5.2.9. f
is decreasing on (a, c) and increasing on (c, b). As a consequence, one concludes that f
has a local minimum at c. This method is usually referred to as the first derivative test
for local maxima or minima. The natural inclination is to think that the converse is also
true; namely, if f has a local minimum at c, then f is decreasing to the left of c and in-
creasing to the right of c. As the following example shows, however, this is false!
5.2.10 EXAMPLE Let f be defined by
f(x) = {x4(2 x ,
0, x = 0.
The function f has an absolute minimum at x = 0; however, f'(x) has both negative
and positive values in every neighborhood of 0. The details are left as an exercise
(Exercise 19). The graph of f'(x) = 4x3(2 + sin 1/x) - x2 cos l/x, x * 0, for x in a
neighborhood of zero is given in Figure 5.7.
Figure 5.7 Graph of f'(x) = 4x3(2 + sin j) - x2 cos , x * 0

The following theorem, besides being useful in computing right or left derivatives
at a point, also states that the derivative (if it exists everywhere on an interval) can only
have discontinuities of the second kind.
5.2.11 THEOREM Suppose f :[a, b) -+ R is continuous on [a, b) and differentiable on (a, b).
If lim f'(x) exists, then f+ (a) exists and
x-+a
f, (a) = 1 m, f'(x).
Xa
Proof. Let L = lim'

x-ra
f'(x), which is assumed to exist. Given e > 0, there exists a
S > 0 such that
If'(x)-LI <e for all x, a < x < a + S.
Suppose 0 < h < 8 is such that a + h < b. Since f is continuous on [a, a + h] and
differentiable on (a, a + h), by the mean value theorem f(a + h) - f(a) = f'(C,,)h for
some Ch E (a, a + h). Therefore,
f(a + h) - f(a) _ LI
= If'(Sh) - LI < e
h
for all h, 0 < h < S. Thus f; (a) = L. Q
5.2.12 EXAMPLES
(a) To illustrate the previous theorem, consider the function
Jx2 + 1, x < 1,
f(x) ° 3-x, x> 1.
For x < 1, f'(x) 2x, which has a left limit of 2 at x = I. Thus by the theorem,
f'_(1) = lim 2x = 2.
Similarly,
f+'(1) = lim (-2x) = -2.
(b) The converse of Theorem 5.2.11 is false. The function
Ix2 sin 1 x # 0,
$(x) = x
0, x = 0,
of Example 5.1.3(g) has the property that g'(0) exists but lidg'(x) does not.
Intermediate Value Theorem for Derivatives

Our second important result of this section, credited to Jean Gaston Darboux (1842-
1917), is the intermediate value theorem for derivatives. The remarkable aspect of this
theorem is that the hypothesis does not require continuity of the derivative. If the deriva-
tive were. continuous, then the result would follow from Theorem 4.2.11 applied to f
5.2.13 THEOREM (Intermediate Value Theorem for Derivatives) Suppose I C R is an

interval and f : I - P is differentiable on I. Then given a, b in I with a < b and a real
number A between [(a) and f'(b), there exists c E (a, b) such that f(c) = A.
Proof, Define g by g(x) = f(x) - Ax. Then g is differentiable on I with g'(x) _
f'(x) - A.
Suppose f'(a) < A < f'(b). Then g'(a) < 0 and g'(b) > 0. As in the remark fol-
lowing Theorem 5.2.9. since g'(a) < 0, there exists an x, > a such that g(.r,) < g(a)-
Also, since g'(b) > 0, there exists an x, < b such that g(x.) < g(b). As a conse-
quence, g has an absolute minimum at some point c E (a, b). But then
g'(c) = f'(c) - A = 0,
i.e., f'(c) = A. Q
The previous theorem is often used in calculus to determine where a function is in-
creasing or decreasing. Suppose it has been determined that the derivative f is zero at
c, and c2 with c, < c2, and that f'(x) * 0 for all x E (c,, c2). Then by the previous the-
orem, it suffices to check the sign of the derivative at a single point in the interval
(cr, c2) to determine whether f is positive or negative on the whole interval (c c,).
Theorem 5.2.9 then allows us to determine whether f is increasing or decreasing on
(cj, c2).
Inverse Function Theorem

We conclude this section with the following version of the inverse function theorem.
5.2.14 THEOREM (Inverse Function Theorem) Suppose ! C R is an interval and f : I -+ R

is differentiable on I with f'(x) * O for all x E I. Then f is one-to-one on I, the inverse
function f` is continuous and differentiable on J = f(!) with
1
(f-')'(f(x)) = f'(x)
for all x E 1.
Proof. Since f'(x) * 0 for all x E 1, by Theorem 5.2.13, f is either positive or neg-
ative on 1. Assume that f'(x) > 0 for all x E I. Then by Theorem 5.2.9,f is strictly in-
creasing on I and by Theorem 4.4.12, f -' is continuous on J = f(I).
It remains to be shown that f -' is differentiable on J. Let y, E J, and let {y.) be
any sequence in J with yn y0, and y 0 y for all n. For each n, there exists x E 1
such that y,,. Since f'' is continuous, x -+x0 = fHence
f- f - `Cy0) xn - x
1' Y. - y0 - - oof(xx) - f(x0)
=f
(xo)
Since this holds for any sequence with y, -r y,,, y * y,,, by Theorem 4.1.3 and the
definition of the derivative,
(f TO = rl(xo)' v
Remark, The hypothesis that f'(x) # 0 for all x E I is crucial. For example, the
function f(x) = x3 is strictly increasing on [ -1, 1 ] with f'(0) = 0. The inverse func.
tion f -'(y) = y'13, however, is not differentiable at y = 0.
5.2.15 EXAMPLES
(a) As an application of the previous theorem, we show that f (x) = x'1", x E (0, oo),
n E N, is differentiable on (0, oo) with
n
for all x E (0, oo). Consider the function g(x) = x", n E N, Dom g = (0, oo). Then
g'(x) = nx"-' and g'(x) > 0 for all x E (0, oo). By the previous theorem, g-' is dif-
ferentiable on J = g((0, oo)) _ (0, oo) with
(g-')'(g(x)) = = 1-1
g'(x)
If we set y = g(x) = x", then x = y"" and
(g-')'(y) = n(y )"-1 = n yI/"-t.
Since f = g' the desired result follows.

(b) As in Example 5.2.7, let L(x) = In x denote the natural logarithm function on
(0, co). Since L'(x) = l/x is strictly positive on (0, oo), the function L is one-to-one,
the inverse function L-' is continuous on Q8 = Range L, and by Theorem 5.2.14,
If we set E = L', then E'(L(x)) = x, or E'(y) = E(y) where y = L(x). The function
E(x), x e 08, is called the natural exponential function on } and is usually denoted
by e', where e is Euler's number of Example 2.3.5. The exponential function E(x) is
considered in greater detail in Example 8.7.20.
(c) In this example we consider the inverse function of g(x) = cos x, x E [0, ir]. Since
g'(x) = -sin x is strictly negative for x E (0, ir), the function g is strictly decreasing
on [0, a] with g([0, a]) _ [- I, 1 ]. Therefore its inverse function g- 1, which we de-
note by Arccos, exists on [ -1, 1 ]. Thus for y E [ -1, 1 ], x = Arccos y if and only if
y = cos x. Finally, since g'(x) 0 for x E (0, ir), by the inverse function theorem,
-1
(8-,)'(g(x)) = g'(x) -sinx = N/l z
'
or since y = cos x,
-l
Arccos y =
dy
The graphs of both cos x, x E [0, ir], and Arccos x, x E [ -1, 13, are given in Figure 5.8.
Figure 5.8 Graphs of cos x, x E [0, ar], and Arccos x, x E [-1, 11
EXERCISES 5.2
1. For each of the following functions, determine the interval(s) where the function is increasing or decreasing, and
find all local maxima and minima.
*a f(x)=x3+Gx-5, xER b.g(x)=4x-x°, xER
c. h(x) = 1+x2x2, x e R d. k(x) _ V- x, xz0
2
e.l(x)=x+7, x#0 f f(x)x-b, a*b,x*b
2. Let f (x) _ (x - a,)2, where a,, a2, ... , a are constants. Find the value of x where f is a minimum.
3. As in Example 5.2.7, use the mean value theorem to establish each of the following inequalities.
a. 1 +xs1+Ix, x> -I b.exZI+x, xER
C., (I +x)a? I + ax, x> -1,a> I
1. For a E N, this inequality was proved by mathematical induction in Example 1.3.3(b). In this exercise,
and in Exercise 4(b), you may assume that for a E R.
d x° = ax°-
4. Prove each of the following inequalities.

a.all. -b""<(a-b)'/", a>b>0,nEE N,na: 2
*b.a°b'-°:S aa+(1 -a)b, a,b>0,0<a < 1
5. Second Derivative Test.: Let f : [a. bJ -* R be differentiable on (a, b). Suppose c E (a, b) is such that f'(c) = 0.
and f"(c) exists.
*a. If f"(c) > 0, prove that f has a local minimum at c.
b. If f"(c) < 0, prove that f has a local maximum at c.
c. Show by examples that no conclusion can be made if f"(c) = 0.
6. Suppose f : (a, b) -> R satisfies If(x) - f (y)I <- Mix - yI" for some a > I and all x, y E (a, b). Prove that f is a
constant function on (a, b).
7. 17ind a polynomial P(x) of degree less than or equal to 2 with P(2) = 0 such that the function
x2, x s 1.
f(x}
P(x), x > 1.
is differentiable at x = 1.
8. Let g be defined by
g(x)- f2sinx+cos2x, xs0,
axe+bx+c, x>0.
Determine the constants a, b, and c such that g'(0) and g"(0) exist.
9. a. Suppose f is differentiable on an interval I. Prove that f is bounded on I if and only if there exists a constant
M such that If(x) - f(y)I s MIx - yI for all x, y E I.
b. Prove that I sin x - sin y I 0.
10. a. Show that tan x > x for x E (0, ).

b. Set
sinx
x
x E (0,;
f(x) -
1, x=0.
Show that f is strictly decreasing on [0, i ].
c. Using the result of pan (b), prove that x < sin x <- x for all x E [0, F].
11. Give an example of a uniformly continuous function on [0, 1 ] that is differentiable on (0, 1) but for which f is
not bounded on (0, 1).
12. ea Suppose f'(x) exists for all x E (a. b). Let c E (a. b). Show that there exists a sequence {x"} in (a, b) with
x" * c and x" -+ c such that f'(x") - +f'(c).
b. Does f'(x,) -+f'(c) for every sequence {x"} with x" -* c?
13. Let f, g : [0, oo) - R be continuous on (0, oo) and differentiable on (0, oo). If f(0) = g(0) and f'(x) a g'(x) for
all x e (0, oo), prove that f(x) z g(x) for all x E (0, oo). .
14. A differentiable function f : [a, b] - R is uniformly differentiable on [a. b] if for every E > 0 there exists a
S > 0 such that
f(t) - f(x) <E
t-x
for all r, x E (a, b] with 0 < It - xl < S. Show that f is uniformly differentiable on (a. b] if and only if f is
continuous on [a, b].
15. *Suppose f : [a, b] -> R with f. (a) > 0. Prove that there exists a S > 0 such that f (x) > f(a) for all
x, a<x<a+S.
16. Prove that the equation x3 - 3x + b = 0 has at most one root in the interval [ -1. 1 ].
M. 'Suppose gas differentiable on (a. b) with I g'(x) I s M for all x E (a. b). Prove that there exists an e > 0 such
that the function f(x) = x + eg(x) is one-to-one on (a, b).
18. Let
I x+2x2 sin X, x#0,

f(x)
0, x=0..
a. Show that f'(0) = 1.-
b. Prove that f'(x) assumes both positive and negative values in every neighborhood of 0.
.
x'(2r.,+ all) X :A 0, ..
f(x) =
0, x=0.
a. Show that f has an absolute minimum at x = 0.
b. Show that f'(x) assumes both negative and positive values in every neighborhood of 0.
20. Let f(x)=x2,g(x)=x3,xE[-1,l].
a. Find c E (-1, 1) such that the conclusion of Theorem 5.2.8 holds.
b. Show that there does not exist any c E (-1, 1) for which
Al) -f(-1) f(c)
g(1) - g(-1) g'(c)'
21. For r E 0 and x > 0, let f(x) = x'. Prove that f'(x) = rx'-'.
22. Suppose L : (0, oo) R is a differentiable function satisfying L'(x) = 1/x with L(1) = 0. Prove each of the fol-
lowing.
a. L(ab) = L(a) + L(b) for all a, b E (0, oo)
b, L(I/b) = -L(b), b > 0
c. L(b') = rL(b), b > 0, r E R
d. L(e) = 1, where e is Euler's number of Example 2.3.5
e. Range L = R
23. Let g(x) - tan x, -f < x < .
a. Show that g is one-to-one on (-f, i) with Range g = R.
'b. Let Arctan x, x E R, denote the inverse function of g. Use Theorem 5.2.14 to prove that
d
An:tan x
+ x2 .
24. a. , Show that f(x) = sin x is one-to-one on [ -i, f] with f([ - z, ]) _ [ - I,.1 J.
z
b. For x E (- I, 1]. let Arcsin x denote the inverse function off. Show that Arcsin x is differentiable on (-1, 1).
and find the derivative of Arcsin x.
25. Let f : (0, oo) -> R be differentiable on (0, oo) and suppose that lim f'(x) = L.
f (x + h f(x)
a. Show that for any h > 0, lim
h
- = L.
x
b. Show that lim f )) = L.
5.3 L'Hospital's Rule

As another application of the mean value theorem, we now prove ('Hospital's rule for
evaluating limits. Although the theorem is named after the Marquis de l'Hospital
(1661-1704), it should be called Bernoulli's rule. The story is that in 1691, l'Hospital
asked Johann Bernoulli (1667-1748) to provide, for a fee, lectures on the new subject
of calculus. L'Hospital subsequently incorporated these lectures into the first calculus
text, L'Analyse des infiniment petis (Analysis of infinitely small quantities), published
in 1696. The initial version (stated without the use of limits) of what is now known as
l'Hospital's rule first appeared in this text.
Infinite Limits
Since l'Hospital's rule allows for infinite limits, we provide the following definitions.
5.3.1 DEFINITION Let f be a real-valued function defined on a subset E of R and let p be a

limit point of E. We say that f tends to oo, or diverges to oo, as x approaches p. denoted
Jim
X-+p
f (x) = 00,
if for every M E R, there exists S > 0 such that

f(x) > M for all x E E with 0 < Ix - pI < S.
Similarly,
lim f(x) _ -00,
if for every M E R, there exists a S > 0 such that
f(x) < M for all x E E with 0 < Ix - pI < S.
For f defined on an appropriate subset E of P. it is also possible to define each of
the following limits:
lint, f (x) =too, lim f (x) =too, Jim f (x) =too, Jim f (x) =too.
X+p X-+p X- 00 XH-x
Since these definitions are similar to Definitions 4.1.11 and 4.4.1 they are left to the ex-
ercises (Exercise 1).
5.3 !Hospital's Rule 191
Remark. Since we now allow the possibility of a function having infinite limits, it
needs to be emphasized that when we say that a function f has a limit at p e I3
(or at ± oo), we mean a finite limit.
L'Hospital's Rule
L'Hospital's rule is useful for evaluating limits of the form
gx(x)
lim
where either (a) lim f(x) = lim g(x) = 0 or (b) f and g tend to ± oo as x -+p. If (a)
holds, then 1 m W 419W) is usually referred to as indeterminate of form 0/0, whereas
in (b) the limit is referred to as indeterminate of form oo/oo. The reason that (a) and (b)
are indeterminate is that previous methods may no longer apply.
In (a), if either lim f (x) or lim g(x) is nonzero, then previous methods discussed in
x-ip
Section 4.1 apply. For example, if both f and g have limits at p and lim g(x) * 0, then
by Theorem 4.1.6(c), p
lim f(x)
f(x)
lim -_
x-*p 8(x) lim g(x)
On the other hand, if lim f (x) = A * 0 and g(x) > 0 with lim g(x) = 0, then as
X-P
x-ip, f(x)/g(x) tends to `Z
oo if A > 0, and to -oo if A < 0 (Exercise 5). However, if
lim f(x) = lim g(x) = 0, then unless the quotient f(x)/g(x) can somehow be simplified,
previous methods may no longer be applicable.
5.3.2 THEOREM (L'Hospital's Rule) Suppose f, g are real-valued differentiable functions

on (a, b), with g'(x) # O for all x E (a, b), where -oo a < b s oo. Suppose
lim f,x) = L, where L E R U {-oo, oo).

x-.a g 'W
if
(a) lim f(x) = lim

x-sa*
g(x) = 0, or
(b) lint g(x) = ±oo,
then
lim f ((x() = L.
x-.a' g(x)
Remark. The analogous result where x b- is obviously also true. A more elemen-
tary version of 1'Hospital's rule, which relies only on the definition of the derivative, is
192 ChaprerS Differentiation
given in Exercise 2. Also, Exercise 7 provides examples of two functions f and g satis-
fying (a) for which lim (f(x)/g(x)) exists but lim (f'(x)/g'(x)) does not exist.
x-a x-a
Proof. (a) Suppose (a) holds. We first prove the case where a is finite. Let be a
sequence in (a, b) with x -+a and x * a for all n. Since we want to apply the gener-
alized mean value theorem to f and g on the interval [a, x,), we need both f and g con-
tinuous at a. This is accomplished by setting
f(a) = g(a) = 0.
Then by hypothesis (a), f and g are continuous at a. Thus by the generalized mean value
theorem, for each n E NI there exists c between a and /x such that
p
[f(xa) - J g(a))f (cn),
or
f(xn) _ f'(Ca)
g(xa) g'(Cn)
Note, since g'(x) # 0 for all x E (a, b), g(x,) # g(a) for all n. As n --I- oc, c -> a'.
Thus by Theorem 4.1.3 and the hypothesis,
'_°° AX-) =
l- f' (x)
m g'(x) -
= L.
Since the above holds for every sequence with x tea', the result follows.
Suppose a = -oo. To handle this case, we make the substitution x = -1/t. Then
as t -> 0+, x -> -oo. Define the functions (p(t) and *(t) on (0, c) for some c > 0 by
ap(t) = f (- ) and *(t) = g(- ! ).
We leave it as an exercise (Exercise 3) to verify that
t-0 +G r)
x lim
- g(X)L,
and that
1m P(t) = lim iy(t) = 0.
Thus by the above,
t
= L.
xli.m- g(x) = 1-'0 00)
(b) Suppose I m, g(x) = oo. The case in which g(x) -+ -oo is treated similarly.
Rather than treating the finite case and infinite case separately, we provide a proof that
works for both.
5.3 CHospital's Rule 193
Suppose first that -oo s L < oo, and /3 E R satisfies /3 > L. Choose r such that
L < r < /3. Since
lim f, (x) < r,

r-.a g,(x)
there exists c, E (a, b) such that
r for all C, a < < c1.

g'(C)
Fix a y, a < y < c1. Since g(x) -* oo as x -+a+, there exists a c2, a < c2 < y, such
that g(x) > g(y) and g(x) > 0 for all x, a < x < c2. Let x E (a, c2) be arbitrary. Then
by the generalized mean value theorem, there exists E (x, y) such that
f(x) -AY) = f'(C) < r. (v)
g(x) - g(Y) g'(C)
Multiplying inequality (4) by (g(x) - g(y))/g(x), which is positive, we obtain
f(x) - f(Y) <
g(x)
r( l-- g(Y)1
g(x)
or
g(x) < + rl l - (5)

g() g(x
for all x, a < x < c2. Now for fixed y, since g(x) -+ oo,
hm. = 0.
xm g-) =
Therefore,
8X)
1'm_IgX)+rl I
L<
--ii =r</3.
Thus there exists c3, a < c3 c2, such that /
- +r(l
for all x, a < x < c3. Thus by inequality (5),
f(x)
< /3 for all x, a < x < c3. (6)
x
AX)
If L = -oo, then for any /3 E Q8, there exists c3 such that inequality (6) holds for
all x, a < x < c3. Thus by definition,
f(x) -oo.
'tea' g(x )
If L is finite, then given e > 0, by taking /3 = L + e, there exists c3 such that

gXx))
x
<L+e for all x, a < x < c3. (7)
Suppose -oo < L <_ oo. Let a E R, a < L be arbitrary. Then an argument sim-
ilar to the above gives the existence of c3 (=- (a, b) such that
f(x) > a
for all x, a <x<c3 .
g(X)
If L = oo, then the above implies that
limf(x)
,/ = oo.
x-+a' g(x)
On the other hand, if L is finite, taking a = L - e gives the existence of c3 such that
f(x)
>L-e forallx , a<x<c3 .
8()
x
Combining this with inequality (7) proves that
f(x)
lim_ = L. Q
x-*a 8(x)
Remarks
(a) The proof of case (a) could have been done similarly to that of (b), treating the case
where a is finite and -oo simultaneously. I chose not to do so since making the substi-
tution x = -I It is a useful technique, reducing problems involving limits at -oo to
right limits at 0. Conversely, limits at 0 can be transformed to limits at too with the
substitution x = 1/t. These new limits are in many instances easier to evaluate than the
original. This is illustrated in Example 5.3.4(c).
(b) In hypothesis (b), we only required that lim g(x) = too. If lira f (x) is finite, then
x-a x-,a'
it immediately follows that
lint-=0,
f (X)
x+a 8(x)
and l'Hospital's rule is not required (Exercise 4). Thus in practice, hypothesis (b) of
1' Hospital's rule is used only if both f and g have infinite limits.
For convenience we stated and proved l'Hospital's rule in terms of right limits.
Since the analogous results for left limits are also true, combining the two results gives
the following corollary.
5.3.3 COROLLARY (L'Hospital's Rule) Suppose f, g are real-valued differentiable functions

on (a, p) U (p, b) with g'(x) * Oforall x E (a, p) U (p, b), where -oo a < b < oo.
Suppose
lim' (x) = L, where L E R U {-oo, oo}.

x-ip g'(x)
5.3 L'Hospital's Rule 195
If
f(x) = lim g(x) = 0, or
(a) limf(x)
(b) lim g(x) 00,
then
f(x)
limg(x) = L.
5.3.4 EXAMPLES
(a) Consider
ln(1+x
lira
x-.0 X
where In is the natural logarithm function on (0, oo). This limit is indeterminate of form
0/0. With f(x) = In(1 + x) and g(x) = x,
f'x 10 1+x=1.
x .O g'(x)_
Thus by 1'Hospital's rule,
ln(1 + x)
A
lira
X
= 1.
Although l'Hospital's rule provides an easy method for evaluating this limit, the re-
sult can also be obtained by using previous techniques. In Example 5.2.7 we proved that
+xsln(1+x)5x
forallx> -1. Thus

1 ln(1+x)
:5
1 +x X
for all x > 0. Thus by Theorem 4.1.9,
11>o
ln(1 + x)
x
- 1.
(b) In this example we consider

1 - cos x
O xz
This is indeterminate of form 0/0. If we apply l'Hospital's rule we obtain

stn x
1
2x
which is again indeterminate of form 0/0. However, applying ('Hospital's rule one more
time gives
cosx
lim
x-O 2
= -
I
Therefore,
lim
1-cosx I
2.
xti0 x2
(e) Consider
e-six
lim
x-.0- X
Since lim a-11x = 0, the above limit is indeterminate of form 0/0. If we apply I'Hospi-
x0'
tal's rule, we obtain
0X e- tix
and this limit is more complicated than the original limit. However, if we let t = 1/x,
then
e-11X t
-,0 X = lim e .
This limit is indeterminate of form oo/oo. By l'Hospital's rule,

t
lim
roc er
-= lim -=0.
,tea e'
1
Therefore, lim a-llx/x = 0.
EXERCISES 5.3
1. Provide definitions for each of the following limits.
a. A-P
limf(x)=oo b. limf(x)=oo
x-4ao
2. *Suppose f g are differentiable on (a, b), xa E (a, b) and g'(xa) * 0. If f (xa) = g(xa) = 0, prove that
f(x) f'(xa)
lim
X _X, g(x) g,(x0)
(Hint: Apply the definition of the derivative.)
3. Let h(x) be defined on (-oo, b). Show that there exists a c > 0 such that rp(t) = h(-;) is defined on (0, c), and
that l-.-aa
lim h(x) = lim op(t).
1H04
4. 'Let f, g be real-valued functions defined on (a, b). If lim f(x) exists in R andx-a'
lim g(x) = oo, prove that
lint
xa' f(x)/g(x) = 0.
5. Suppose f, g are real-valued functions on (a, b) satisfying lim, f(x) = A * 0, lim* g(x) = 0. and g(x) > 0 for all
1
x E (a, b). If A > 0, prove that lim f(x)/g(x) = oc. If A <0, prove that lim. f(x)/g(x) _ -oc.
6. Use I'Hospital's rule and any of the differentiation formulas from calculus to find each of the following limits. In
the following, In x, x > 0, denotes the natural logarithm function.
x5 +2x-3 x`+2x-3
a. lim b. ,Imi
2x3 - x2 - I 2x' - x' - I
I - cos 2A-
c. lim -x In x
d. lira
_.0 sin x
e lim_ox° In x where a > 0 f. lim

,-.o sin x
(In x)° (In x)"
g. lim , where a > 0 h. lim
Ay y , p, q E fl
e` .r
(i. lim
=-o' .r sin .rJ
f( x) fg'(-x(x)
7. Let f(x) = x2 sin(1/x) and g(x) = sin x. Show that lim exists but that lim ) does not exist.
box)
p(X) p(x),
8. Investigate lim
. q(x)
and lim where p and q are polynomials of degree it and in. respectively.
x_'-« q(x)
9. Let f (x) = (sin x)/x for x * 0, and f (0) = I .
a. Show that f'(0) exists, and determine its value.
b. Show that f"(0) exists, and determine its value.
e_
10. Let f be defined on R by, f(x) = for x # 0. and f(0) = 0. Prove that ft"'(0) = 0 for all n = I. 2. .
5.4 Newton's Method2

In this section we consider the iterative method, commonly known as Newton's method.
for finding approximations to the solutions of the equation f (x) = 0. Although the
method is named after Newton, it is actually due to Joseph Raphson (1648-1715). and
in many texts the method is referred to as the Newton-Raphson method. Newton did de-
rive an iterative method for finding the roots of a cubic equation; his method, however.
is not the one used in the procedure named after him.
Suppose f is a continuous function on [a, b] satisfying f(a)f(b) < 0. Then f is of
opposite sign at the end points a and b and thus by the intermediate value theorem (The-
orem 4.2.11), there exists at least one value c E (a, b) for which f (c) = 0. If, in addi-
tion, f is differentiable on (a. b) with f'(x) * 0 for all x E (a, b), then f is either strictly
increasing or decreasing on [a, b]. In this case the value c is unique; that is, there is ex-
actly one point where the graph off crosses the x-axis.
2. The topics of this section are not required in subsequent chapters.

An elementary approach to finding a numerical approximation to the value c is the

method of bisection. For this method, differentiability off is not required. To illustrate
the method, suppose f (as in Figure 5.9) satisfies f(a) < 0 < f(b). Let
c,=2(a+b).
If f(c,) = 0, we are done. If f(c1) * 0, then c belongs to one of the two intervals
(a, c1) or (c1, b), and thus Ic1 - ci < I' (b - a). Suppose (as in Figure 5.9) f(c1) > 0.
Then c E (a, c1), and in this case we set co = a, and
C2 = 2 (CO + CO-
If f(c2) = 0, we are done. If not, then suppose (as in Figure 5.9) f(c2) < 0. Then
c E (c2, c1), and as above, we set
C3 = I (Cl + C2).
I//
c1,C2,...,cn,n?2 C. = lCn-1 + C)
Figure 5.9
In general, suppose have been determined with

for some j = 0, . . , n - 2. If by happenstance f (cn) = 0, then we have obtained the
.
exact value. If f(cn_ 1)f(cn) < 0, then c lies between cn_ 1 and c,, and we define
cnr1 = -2 2 (C. + Cn_1),

On the other hand, if 0, then c lies between c and c,, and in this case.
we define
This gives us a sequence which satisfies
:2I(b-a).
Thus lim c = c and by continuity, f(c) = 0.
Although this method provides a sequence of numbers that converges to the zero off.
it has the disadvantage that the convergence is rather slow. An alternative method, due to
Raphson, uses tangent lines to the curve to find successive points c approximating the zero
off. As we will see, this method will often converge much more rapidly to the solution.
As above, assume that f is differentiable on [a, b] with f (a) f (b) < 0 and
f'(x) # 0 for all x E [a, b]. Let ct be an initial guess to the value c. The line tangent
to the graph off at (c,, f (c,)) has an equation given by
Y =f(ct) +f(ct)(x - CO.
Since f (c,) # 0, the line crosses the x-axis at a point that we denote by c2 (Figure 5.10).
Thus
0 =f(ci) +f'(ct)(c2 - Cl),
that, upon solving for c2, gives
f(CI)
C2 = C, -
f'(CI)
Figure 5.10 Newton's Method

We now replace the point c, by the second estimate c2 to obtain c3, and so forth. In-
ductively, we obtain a sequence (cn) given by the formula
c
----- , n = 1,2... (8)
c"
in which c, is an initial guess to the solution f(c) = 0. As we will see, under suitable
hypothesis, the sequence {c"} will converge very rapidly to a solution of the equation
f(x) = 0. Before we prove the main result, we illustrate the above with an example.
5.4.1 EXAMPLE Let a > 0 and consider the function

f(x)=x2-a.
If a > 1, then fhas exactly one zero on [0, a], namely \. If 0 < a < 1. then the
zero off lies in [0, 1). Let c, be an initial guess to Va. Then by formula (8), for n ? 1,
cl. - a l a
Cn+1 - Cn
2c" - 2(C" + Cn).
This is exactly the sequence of Exercise 9 of Section 2.3, where the reader was ask
to prove that the sequence converges to %/a-. With a = 2, taking c, = 1.4 as an initia
guess yields
c2 = 1.4142857,
c3 = 1.4142135,
which is already correct to at least seven decimal places.
5.4.2 THEOREM Let f be a real-valued function on [a, b] which is twice differentiable on

[a, b]. Suppose that f(a)f(b) < 0 and that there exist constants m and M such that
If'(x)I m > 0 and [f"(x) 1 s M for all x E [a, b]. Then there exists a subinterval I
of [a, b] containing a zero c off such that {for any c, E 1, the sequence {c"} defined by
T(-
J (Cn)
c.), nEN,
Cn+l=cn-f,(c),
is in 1, and n-co
lim c" _. c. Furthermore,
C"+, - cl S ICn - W. (9)
Prior to proving Theorem 5.4.2 we first state and prove the following lemma. The
result is in fact a special case of Taylor's theorem (8.7.16), which will be discussed in
Chapter 8.
5.4.3 LEMMA Suppose f : [a, b] -a R is such that f and f' are continuous on [a, b] and
f"(x) exists for all x E (a, b). Let x" E [a, b). Then for any x E [a, b], there exists a
real number between x, and x such that
f(x) =f(xo) +f'(xe)(x - xo) + 2f"(C)(x - xo)2.

Proof. For x E [a, b], let a E R be determined by

f(x) = f(x,) + CO -
Define g on [a, b] by
g(t) = f(t) - f(x.) - x,) - a(t - x,)2.
If x = x, then the conclusion is true with C = x". Assume that x > x". Then g is con-
tinuous and differentiable on [x x] with g(x") = g(x) = 0. Thus by Rolle's theorem
there exists c E (x0, x) such that g'(c) = 0. But
g'(t) = fi(t) -f'(x,,) - 2a(t - x,).
By hypothesis, g' is continuous on [x c], differentiable on (x c), and satisfies g'(x,) _
g'(c) = 0. Thus by Rolle's theorem again, there exists C E (x c) such that 0. But
g"(t) = f"(t) - 2a.
Therefore, a = if"(C)
Proof of Theorem 5.4.2. Since f(a) f(b) < 0 and f'(x) # 0 for all x E [a. b], f
has exactly one zero c in the interval (a, b).
Let x, E [a. b] be arbitrary. By Lemma 5.4.3 there exists a point C between c and
x, such that
0 = f(c) = f (xo) + f'(xo)(c - x,) + f"(C)(c - x,)2,
or 2
- Ax.) = f' (xo)(c - x0) + 2 f "(3'xc - x,)2. (10)
If x, is defined by
f(x.)
X1 = X. - f,(xo),
then by equation (10).
X, =x,+(c- x,)+ 2 f,"(0 1

f )(c-x,)2.
0
Therefore,
"
IX, - cl = 2 Ic - x,12 s Ic - x,12. (11)
x,
Choose s>0sothat 8<2m/M andI=(c-S,c+S] C [a,b].Ifc"E1,then
lc - c"I < S. If c"+, is defined by formula (8), then by formula (11),
M
c"+,-c1<- 82<S.
Therefore, c"+, E I. Thus if the initial choice c, E I, then c" E I f o r all n = 2, 3, ... .
202 Chaprer5 Differentiation
It remains to be shown that 1 m c = c. If ct E 1, then by induction
ICn+I - CI <
2 - I.
But by our choice of S. (M/2m) S < 1, and as a consequence cn -*c. Q
Remarks
(a) For a given function f satisfying the hypothesis of the theorem, the constants M and
m, and thus S can be determined. To determine the interval I, one can use the method of
bisection to find an approximation xn to c satisfying Ix,, - x - I I < S. If cl is taken to
be xn, then Icl - ci < S. In practice, however, one usually makes a judicious guess for
c1 and proceeds with the computations.
(b) Let en = c - cn be the error in approximating c, and let K = M/2m. Then in-
equality (9) can be expressed as
ien+1i C Kleni2.
Consequently, if I e,,I < 10', then I en + I I < K 10-2. Thus, except for the constant
factor K, the accuracy actually doubles at each step. For this reason. Newton's method
is usually referred to as a second order or quadratic method.
(c) Even though Newton's method is very efficient, there are a number of things that
can go wrong if c1 is poorly chosen. For example, in Figure 5.11, the initial choice of
c1 gives a c2 outside the interval, and the subsequent cn tend to -cc. Such a function is
given by f (x) = x/(x2 + 1). In Figure 5.12, the initial choice of c1 causes the subse-
quent values to oscillate between c1 and c2. A function having this property is given by
g(x) = x - J Taking c1 = 1 gives c2 = -1, c3 = 1, etc. For this reason, the initial
choice of c1 for many functions has to be sufficiently close to c in order to be sure that
the method works.
(cl.ftc1))
(c2. f(c2))
Figure 5.11
Notes 203
Figure 5.12
EXERCISES 5.4
1. *For a > 0, apply Newton's method to f(x) = x3 - a to obtain a sequence that converges to the cube root
of a.
2. Use Newton's method to find approximations to the roots, accurate to six decimal places, of the given functions on
the interval [0, 1 ].
*a. f(x) = x3 - 3x + 1 b. f(x) = 3x3 - 5x2 + I
c. f(x)=x`-3x3+1
3. Use Newton's method to approximate the real zeros of f(x) = x' - 4x - 3 accurate to four decimal places.
4. Show that f(x) = In x - x + 3, x E (0, oo), has two real zeros. Use Newton's method to approximate them accu-
rate to four decimal places.
5. Suppose g : [a, b] -i, R is differentiable on an interval [a, b] and suppose that there exists a constant c, 0 < c < I,
such that Ig'(x)I s c for all x E [a, b]. Prove, that there exists a unique point x, E [a, b] such that g(xo) = x,. (See
Exercise 13, Section 4.3.)
6. Let f : [a, b] -* R be differentiable on [a, b] with f (a) < 0 < f(b). Suppose there exist constants m and M such
that 0 < m <_ f'(x) s M for all x E [a, b]. Let c, E [a, b] be arbitrary, and define
C-1 = f(c,,)
M
Prove that the sequence converges to the unique zero off on [a, b]. (Hint: Consider the function
g(x) = x - f(x)/M.)
NOTES
Without question the most significant result of this chap- was illustrated by the subsequent theorems and exercises.
ter is the mean value theorem. The simplicity of its proof Additional applications will be encountered throughout
disguises the importance and usefulness of the result. The the text.
theorem allows us to obtain information about the func- Although the mean value theorem is attributed to La-
don from its derivative. This has many applications, as grange, his proof, which appeared about 1772, was based
M Chapter 5 Differentiation
on the false assumption that every function could be The Bernoulli brothers, Jakob (1654-1705) and Jo-
expanded in a power series. Cauchy, in his 1823 text Rf- hann (1667-1748), were among the first mathematicians
sumf des Lecons donnees a L'Ecole Royale Polytechnique in Europe to use the new techniques of Newton and Leib-
sur le Calcul Infinitesimal, used the modern definition of niz in the study of curves and related physical problems.
the derivative to provide a proof of the mean value theo- Among these were finding the equations of the catenary
rem. His statement and proof of the theorem, however, and isochrone.` Both brothers also contributed to the
differ from the version detailed here in that he assumed study of differential equations by solving the Bernoulli
continuity of the derivative. What Cauchy actually proved equation: y' + P(x)v = Q(x)y". Through their numerous
was that if f is continuous on [a, b], then the quantity publications and correspondence with other mathemati-
{f(b) - f(a)}/{b - a) lies between the minimum and cians, the Bernoulli brothers helped to establish the utility
maximum values of f on [a, b]. (See Miscellaneous Ex- of the new calculus. The first text on differential calculus
ercise 4.) Then by the intermediate value theorem (Theo- by ('Hospital also contributed significantly to populariz-
rem 4.2.1 1) applied to the continuous function f, there ing the subject.
exists c E (a, b) such that Leonhard Euler (1707-1783), one of the most pro-
lific mathematicians in history, contributed significantly
f(b) - f(a) = f'(c)(b - a). to establishing calculus as an independent science. Even
It is worth noting that our proof of the mean value theo- though the calculus of exponential and logarithmic func-
rem depends ultimately on the completeness property of tions was basically developed by Johann Bernoulli, it was
R (through Rolle's theorem and Corollary 4.2.9). Euler's expositions in the eighteenth century that brought
The mean value theorem can justifiably be called the these topics into the mainstream of mathematics. Much of
fundamental theorem of differential calculus. It allowed what we know today about the exponential, logarithmic.
the development of rigorous proofs of many results that and trigonometric functions is due to Euler. He was also
were previously taken as fact or "proved" from geometric among the first mathematicians to define the concept of a
constructions. Although the modern proof of l'Hospital's function. However, to Euler-as to the other mathemati-
rule uses the mean value theorem, it should be remem- cians of that period-a function was one that had a power
bered that the original version for calculating the limit of series expansion. It is important to note that most mathe-
a quotient where both the numerator and denominator be- maticians of the eighteenth century. including Euler, were
come zero first appeared in 1696, over 70 years before primarily concerned with computations needed for the ap,
Lagrange's proof of the mean value theorem. The original plications of calculus; proofs did not gain prominence un-
version was stated and "proved" in a purely geometric til the nineteenth century. For this reason, numerous re-
manner without reference to limits. For further details, in- sults of that era that were assumed to be true were
cluding a history of calculus, the reader is referred to the subsequently proved to be true only under more restric-
text by Katz listed in the Bibliography. tive conditions.
1. Let f : (a, b) -> R and suppose f" exists at x E (a, b). Prove that
limf(x,+h)+f(x0-h) -2f(x")
f"(z,) hn0 h2
Give an example where this limit exists at x,, but f" does not exist at x,.
3. The catenary problem involves finding the equation of a freely hanging cable, whereas the isochrone
problem involves finding the equation of a curve along which an object would fall with uniform vertical
velocity.
2. Let f be a real-valued differentiable function on R.

a. If there exists a constant b < I such that I f'(x) I < b for all x E Ca, prove that f has a fixed point in R. (See
Exercise 13, Section 4.3.)
b. Show that the function f(x) = x + (1 + e`)-' satisfies I f'(x)l < I for all x E R but that f has no fixed point
in R.
3. A function f is convex (or concave up) on the interval (a. b) if for any x, y E (a, b), and 0 < t < 1,
f(tx + (I -r)y)<-tf(x)+(I -t)f(y).
a. If f is convex on (a. b). prove that f is continuous on (a, b).
b. If f is convex on (a, b), prove that f+(p) and f _(p) exist for every p E (a, b). Show by example that a convex
function on (a, b) need not be differentiable on (a, b).
c. Suppose f"(x) exists for all x E (a, b). Prove that f is convex on (a, b) if and only if f"(x) at 0 for all
x e (a, b).
4. Suppose fis differentiable on [a, b] and that f is continuous on [a, bl.Without using the mean value theorem,
prove that
min{f'(x):x e [a, b]} sf(bb Q(a) - max{f (x) :x E [a, b]}.
5. (T. M. Flett) If f is differentiable on [a, b] and f'(a) = f'(b), prove that there exists ( E (a, b) such that
f(C) - f(a) =
C - a
6. Suppose f: (a, b) R is differentiable at c E (a, b). If (s") and {t"} are sequences in (a, b) with s,, < c < t" and
limoo (t" - s") = 0, prove that
f(s") =
lim
t"-s"
Baxley, J. V. and Hayashi, E. K., "Indeterminate forms f (a) to f'(a+)," Math. Mag. 39 (1966),
of exponential type," Amer. Math. Monthly 85 112-120.
(1978), 484-486. Katznelson, Y. and Stromberg, K., "Everywhere differ-
Cajori, Florian, "Historical note on the Newton-Raphson entiable, nowhere monotone function;' Amer. Math.
method of approximation;' Amer. Math. Monthly 18 Monthly 81(1974). 349-354.
(1911), 29-33. Miller, A. D. and Vyborny, R., "Some remarks on func-
Corless, R. M., "Variations on a theme of Newton;' tions with one-sided derivatives;' Amer. Math.
Math. Mag. 71 (1998), 34-41. Monthly 93 (1986),471-475.
Flett, T. M., "A mean value theorem," Math. Gaz. 42 Rosenholtz, "Mere is no differentiable metric for W."
(1958), 38-39. Amer. Math. Monthly 86 (1979), 585-586.
Hall, W. S. and Newell, M. L., 'The mean value theorem Rotando, L. M. and Korn, H., "The indeterminate form
for vector valued functions;' Math. Mag. 52 (1979), 0°," Math. Mag. 50 (1977),41-42.
157-158. Thurston, H. A., "On the definition of the tangent line;'
Hartig, Donald, "L'Hospital's rule via integration;' Amer. Math. Monthly 71 (1964), 1099-1103.
Amer. Math. Monthly 98 (1991). 156-157. Tong, J. and Braza, P. A., "A converse of the mean-value
Langlois, W. E. and Holder, L. 1., "The relation of theorem;' Amer. Math. Monthly 104 (1997). 939-942.
The Riemann and
Riemann-Stieltjes Integral
6.1 The Riemann integral
6.2 Properties of the Riemann Integral
6.3 Fundamental Theorem of Calculus
6.4 Improper Riemann Integrals
6.5 The Riemann-Stieltjes integral
6.6 Numerical Methods
6.7 Proof of Lebesgue's Theorem
When Newton and Leibniz developed the calculus, both considered integration as the
inverse operation of differentiation. For example, in the De analysi. t Newton proved
that the area under the curve y = a''(m/n # -1) is given by
t
an
m+n
by using his differential calculus to prove that if A(x) represents the area from 0 to x
then A'(x) = ax'"/". Even though Leibniz arrived at the concept of the integral by using
sums to compute the area, integration itself was always the inverse operation of differ-
entiation. Throughout the eighteenth century, the definite integral of a function f(x) on
[a, b], denoted fa f(x)dx, was defined as F(b) - F(a) where F was any function whose
derivative was f (x). This remained the definition of the definite integral until the 1820s.
The modem approach to integration is again due to Cauchy, who was the first
mathematician to construct a theory of integration based on approximating the area un-
der the curve. Euler had previously used sums of the form J'=, f(xk _ ,)(xk - x,-,) to
approximate the integral of a function f(x) in situations where the function F(x) could
not be computed. Cauchy, however, used limits of such sums to develop a theory of in-
tegration that was independent of the differential calculus. One of the difficulties with
1. "The Mathematical Works of Isaac Newton;' edited by D. T. Whiteside, Johnson Reprint Corporation. Ness
York, 1964.
208 Chapter 6 The Riemann and Riemann-Stieltjes Integral
Cauchy's definition of the integral was that it was very restrictive: only functions that
were continuous, or continuous except at a finite number of points, were proved to be
integrable. However, one of Cauchy's key achievements was that using his definition,
he was able to prove the fundamental theorem of calculus; specifically. if f is continu-
ous on [a, bi, then there exists a function F on [a, hl such that F'(x) = f(x) for all
x E [a, b].
The modern definition of integration was developed in 1853 by Georg Bernhard
Riemann (1826-1866). Riemann was led to the development of the integral by trying
to characterize which functions were integrable according to Cauchy's definition. In the
process, he modified Cauchy's definition and developed the theory of integration that
bears his name. One of his achievements was providing necessary and sufficient condi-
tions for a real-valued bounded function to be integrable. In Section 6.1 we develop the
theory of the Riemann integral using the approach of Jean Gaston Darboux (1842-1917).
In this section we also include the statement of Lebesgue's theorem, which provides
necessary and sufficient conditions that a bounded real-valued function defined on
a closed and bounded interval be Riemann integrable. A self-contained proof of
Lebesgue's theorem will be given in Section 6.7. The equivalence of the Riemann and
Darboux approach will be proved in Section 6.2.
In Section 6.5 we will consider the more general Riemann-Stieltjes integral that
will give meaning to the following types of integrals:
+
f1
f(x) dx', f J '(x) d[xor f(x) da(x),
J0 ,
11
where a is a monotone increasing function on [a, b]. These types of integrals were de-
veloped by Thomas-Jean Stieltjes (1856-1894) and arise in many applications in both
mathematics and physics. The theory itself involves only minor modifications in the
definition of the Riemann integral; the consequences, however, are far-reaching. The
Riemann-Stieltjes integral permits the expression of many seemingly diverse results as
a single formula.
6.1 The Riemann Integral

There are traditionally two approaches to the theory of the Riemann integral: the orig-
inal method of Riemann and the method introduced by Darboux in 1875 using lower
and upper sums. I have chosen the latter approach to define the Riemann integral be-
cause of its easy adaptability to the Riemann-Stieltjes integral. We will, however, con-
sider both methods and show that they are, in fact, equivalent.
Upper and Lower Sums

Let [a, b], a < b, be a given closed and bounded interval in R. By a partition
91 of [a, b] we mean a finite set of points 91 = {xo, x1.. .. , such that
a=xo<x,< <xb.
There is no requirement that the points x, be equally spaced. For each i = 1. 2.... , n.
set
Ax, = x, - x;_,,
which is equal to the length of the interval [x,x,].
Suppose f is a bounded real-valued function on [a, b]. Given a partition
9 = {xo, x...... of [a, b], for each i = 1.2.. ... n. let
in, = inf{f(t):x,_, <_ t <_ x,}.
M, = sup{f(t):x, _, 5 t s x,}.
Since f is bounded, by the least upper bound property, the quantities m, and M, exist in
R. If f is a continuous function on [a, b], then by Corollary 4.2.9. for each i there exist
points t s; E [x,_,.x,] such that M, = f(t,) and m, = f(sj.
The upper sum °U(9, f) for the partition 9 and function f is defined by
1110"D = Y M;ox,.
Similarly, the lower sum Y(9, f) is defined by

n
20"D = I,8=1MAX,
Since m, <_ M, for all i = 1, . . . , n, we always have
s
for any partition 91 of [a, b]. The upper sum for a nonnegative continuous function f is
illustrated in Figure 6.1. In this case °lt(2P, f) represents the circumscribed rectangular
approximation to the area under the graph off. Similarly, the lower sum represents the
inscribed rectangular approximation to the area under the graph off.
Figure 6.1 Upper Sum U(9. j)

210 Chapter6 The Rlemann and Riemann-Stleltjes Integral
Upper and Lower Integrals

If the function f satisfies m : f(t) <- M for all t e [a, b], then
m(b - a) c --5 M(b - a) (1)
for any partition 9 of [a, b]. To see that inequality (1) holds, let iI' = {x°, ... , be
any partition of [a, b]. Since M; <- M for all i = 1, ... , n,
61491, f) M4Axi `- M(X, - xi -1) = M(b - a).

Similarly, -W(9, f) z m(b - a). Thus the set {°G.(91, f) : 91 is a partition of [a, b]} is
bounded above and below, as is the set {2(9', f)}.
6.1.1 DEFINITION Let f be abounded real-valued function on the closed and bounded
interval [a, b]. The upper and lower Integrals off, denoted fQ f and fa f respectively,
are defined by
b
f = inf{°U,(9, f) : 91 is a partition of [a, b]),
rb
f = sup{.T(91, f) : 9' is a partition of [a, b]}.
Ja
Since the sets {° L (91j )j )} and {2(9,f)} are nonempty and bounded, the lower and
upper integrals of a bounded function f : [a, b] -+ R always exist. Our first goal is to
prove that f; f s f. f for any bounded real-valued function f on [a, b]. To this end, we
make the f!1owing definition.
6.1.2 DEFINITION A partition 9* of [a, b) is a refinement of 9 if 9' C 9*.

A refinement of a given partition 9 is obtained by adding additional points to 91. If
911 and 9'2 are two partitions of [a, b], then 9, U 9'2 is a refinement of both 91, and 912.
6.1.3 LEMMA If 91* is a refinement of 9, then

2(91,f) 2(91*,f) s w.(91*,f) < °u(91,f).
Proof. Suppose 9' _ {x°, x1, ... , xa} and 91* = 9 U {x*}, where x* * x; for any
j = 0, 1, ... , n. Then there exists an index k such that xk_ < x* < xk. Let
1
Mk = sup{f(t) : t E [Xk_ 1, x*]},

Mk' = sup{f(t) : t E [x*, Xk]}.
Since f(t) s Mk for all t e [xk_,, xk], we have that f(t) s Mk for all t E (xk_ 1, x*] and
also for all t E [x*, xk]. Thus both Mk and Mk are less than or equal to Mk. Now
k-1 n
1149'*,f) _ M;Ax, + MM(X* - Xk-1) + Mk(xk - x*) + MMAzr
1=1 l-k-I
But
Mk(X* - Xk-1) + Mi(xk - X*) MkAXA.
Therefore,
The proof for the lower sum is similar. If P* contains k more points than 91, we need
only repeat the above argument k times to obtain the result.
6.1.4 THEOREM Let f be a bounded real-valued function on [a, b]. Then
t
jaf :S: f
Proof. Given any two partitions 9, 9 of [a, b].
2(91,f) S 2(9 U 9.,f) S 1(9 U 9.,f) t
Thus 2(91,f) < °ll.(a f) for any partitions 9, 9.. Hence
Ji= up X(O,f) S Gll.(a,f)

for any partition a. Taking the infimum over 9, gives the result. Q
The Riemann Integral

If f : [a, b] -+ R is bounded, then the lower and upper integrals off on [a, b] always
exist and satisfy fo f < f° f. As we will shortly see, there is a large family of functions
for which equality holds: such functions are said to be integrable.
6.1.5 DEFINITION Let f be a bounded real-valued function on the closed and bounded
interval [a, b]. If
then f is said to be Riemann integrable or integrable on [a. b]. The common value is
denoted by
Jb
or Jf(x)dx,
b
a f a
and is called the Riemann Integral or integral off over [a, b]. We denote by 9t[a, b]
the set of Riemann integrable functions on [a, b]. In addition, if f E 9t[a, b], we define
Jbaf -Jabs
212 Chapter6 The Riemann and Riemann-Stieltjes Integral
If f : [a, b) -> R satisfies m s f(t) M for all t e [a, b], then by inequality (1),
m(b-a): Jbf Ihf <_M(b-a).

a a
If, in addition, f E 91[a, b], then
m(b-a)<- Jf
b sM(b-a).
a
In particular, if f(x) ? 0 for all x E [a, b], then fa f >_ 0. If f E 9t[a, b] is nonnega-
tive, then the quantity fa f represents the area of the region bounded above by the graph
y = f(x), below by the x-axis, and by the lines x = a and x = b.
6.1.6 EXAMPLES
(a) The following function, attributed to Dirichlet, is the canonical example of a func-
tion that is not Riemann integrable on a closed interval. Let f be defined by
11, x E Q,
f(x) = 0, x0Q.
Suppose a < b. If 9 = {xo, ... , x} is any partition of [a, b], then m; = 0 and
M; = 1 for all i = 1, ... , n. Thus
2(9,.f) = 0 and °1L(9,f) = (b - a)
for any partition 9 of [a, b]. Therefore,
fb
1a
f=0 and J f=(b-a),
a
and as a consequence, f is not Riernann integrable on [a, b].

(b) Let f : [0, 1 ] -> l be defined by
0:5 x < 1
In this example we prove that f is integrable on [0, 1] with f, f = 2. Let

9 = {x°, x,, . .. , x,} be a partition of [0, 1 ], and let k E {1, ... , n} be such that
xk_, < I - xk (See Figure 6.2.) If mi and M; denote the infimum and supremum off on
[xi _ 1, x;] respectively, then
- f0, i = 1,.. ., k, and M;

- f0, i = 1,.. k - I,
f
m' 1, i=k+I,.. ,n, 1, i=k,.. n.
I Xk xn=1
2
Figure 6.2
Thus
n
x(91, f) = Axi = (1 - xk), and
i-k+t
ft
a(L(P,f) = I Oxi = (I - xk-1)

i-k
Since 1 - xk s 2 < 1 - xk_ 1, 2(91,f) s 2 <- °lt(91, f) for all partitions 91 of [0. 1 ].
Thus fo f s 2 s fo f. Hence if f is integrable on [0, 11, then fo f = 2. We conclude by
proving that f is indeed integrable on [0, 1). Since I - xk _ , = (1 - xk) + (xk - xk _ , ),
we have
`'ll(`',f) = 2(9',f) + (Xk - xk-1)

Let e > 0 be arbitrary. If 9 is any partition of [0, 1) with i xi < e for all i, then
nu(9P,f) < 2(91,f) + e. Thus
I
f f - ifk49-1f)s°uP1,f)
r1
s 2011f) + e c sup2(`9-,f)+e51 f+e,
0
where the infimum and supremum are taken over all partitions a of [0, 1). Since E > 0
was arbitrary, we have j' If = fo f. Thereforef is integrable on [0, 11 with f, f = 2.
(c) We now provide another example to illustrate how tedious even a trivial integral can
be if one relies only on the definition of the integral. Luckily, the fundamental theorem
of calculus (Theorem 6.3.2) will allow us to avoid such tedious computations. Let
f(x) = x, x E [a, b]. For the purpose of illustration we take a > 0 (Figure 6.3). Inter-
preting the integral as the area under the curve, we intuitively see that
tbx = -a)(b+a)=(b2-a22).
This is obtained from the formula for the area of a parallelogram. Let 91
{x0, x,, ... , be any partition of [a, b]. Since f(x) = x is increasing on [a, b],
m;=f(x;_,)=x;_, and M;=f(x;)=x;.
Figure 6.3
Therefore,
4911f) x; _ , Ox; and 949, f) x; Ax;-
For each index i,
x;-1 < 2(xi_, +x) <_x;.
If we multiply this by Ox; = x; - x;_, and sum from i = 1 to n, we obtain
l (xi - _') °d(g1,f)
But
;_, 2
(x?-xi_i)= 2 (xn-4o)= 2 (b2-a2).
6.1 The Riemann Integral 21S
Finally, if we take the supremum and infimum over all partitions 9' of [a, b], we have
J6
xdxs2(b2-a2):5 f xdx.
a
This in itself does not prove that f(x) = x is integrable on [a, b] with
Jb
xdx = 2 (b2 - a2).
a
To prove that f is integrable on [a, b], we note that for any partition 9' of [a, b],
fb
rb
0c xdx - J x dx <_ alt(9i, f) - T(91,f) = I (xi - xi -1)Axi
a a i-1
< (1ma5 Axi) (b - a).

Let e > 0 be given. If we choose a partition 9' such that Axi <. a/(b - a + 1) for all
i, then fax dx - fax dx < e. Since this holds for every e > 0, we conclude that
fa x dx = fax dx. Thus f(x) = x is Riemann integrable on [a, b] with fax dx =
2 (b2 - a2).
(d) Consider the function f(x) = x2, x E [0, 1]. For n E Ni, let 9' be the partition
{0, ,'- 2, ... , 1}. Since f is increasing on [0, 1 ], its infimum and supremum on each
interval [Y,11 are attained at the left and right endpoint respectively, with
m, = (i - 1)2/n2 and Mi = i2/n2. Since Axi = 1/n for all i,
2(9,,,f)°n3[12+22+.+(n- 1)2], and
n
Using the identity 12 + 22 + + m2 = 6m(m + lX2m + 1) (Exercise 1, Section
1.3), we have
491.1f) = 6 (2 - n and 1UP1,,,f) = 6 ( 1 -1

n/
n)(2+1).
C1
Thus sup.T(9D,,, f) = 1 and inf °lL(91,,, f) = 3. Since the collection 191.: n e N} is

a subset of the set of all partitions of [0, 1 ],
`-f( ',f) = Jx2 dx,

1
= sup 201., f) :5 and
3 0
3 = inf °lt(9,,, f) f °lt( P, f) = J.x2dx.

0
Therefore f(x) = x2 is integrable on [0, 1 ] with fo x2 dx = .
i
216 Chapter6 The Rietnann and Riemann-Stieltles Integral
Riemann's Criterion For Integrability

The following theorem, the original version of which is attributed to Riemann, provides
necessary and sufficient conditions for the existence of the Riemann integral.
6.1.7 THEOREM Abounded real-valued function f is Riemann integrable on [a, b] if and

only if for every E > 0, there exists a partition 9 of [a. b) such that
RL(9',f) - E. (2)
Furthermore, if 9 is a partition of [a, b] for which inequality (2) holds, then the in-
equality also holds for all refinements of 9.
Proof. Suppose inequality (2) holds for a given e > 0. Then
0C I f- Jf_<oL(P.f)_(f)<E.
b
a a
Thug f is integrable on [a, b].

Conversely, suppose f is integrable on [a, b]. Let e > 0 be given. Then there exist
partitions 9, and 92 of [a, b] such that
jbf
°lL(92,f) - <2 and tbf - 491'-f) < 2
Let 9 = 91 U 92. Then
jh
0149V)
90149V) L(92,f) <
u
f + 2 < `-P(('1Pi,f) + E E.
Therefore, °U.(9, f) - %(9,f) < e, which proves inequality (2). If D. is any refinement
of 9, then by Lemma 6.1.3,
0 <_ qt(L,f) - 2(a,f) :9 6u(P,f) - e(91,f) < E.
Thus inequality (2) is also valid for any refinement 2 of 9. Q
Integrability of Continuous and Monotone Functions

As an application of the previous theorem, we prove that every continuous teal-valued
function and every monotone function on [a, b] is Riemann integrable on [a. b]. As we
will see, both of these results will also follow from Lebesgue's theorem (Theorem 6.1.13).
6.1.8 THEOREM Lei f be a real-valued function on [a, b].

(a) If f is continuous on [a, b), then f is Riemann integrable on [a, b].
(b) If f is monotone on [a, b], then f is Riemann integrable on [a, b].
Proof. (a) Let e > 0 be given. Choose nl > 0 such that (b - a)rl < E. Since f is
continuous on [a, b], by Theorem 4.3.4,f is uniformly continuous on [a, b]. Thus there
exists a S > 0 such that
[f(x) - f(t) I < 77 (3)
for all x, t E [a, b] with Ix - tl < S. Choose a partition 91 of [a, b] such that Ox; < S
for all i = 1, 2, ... , n. Then by inequality (3),
Mi - misrl
for all i = 1, 2, ... , n. Therefore,
61(P,f)-201,f)= (M,-mi),%xi:S71 Ax,=rl(b-a)<e.

Thus by Theorem 6.1.7,f is integrable on [a, b].
(b) Suppose f is monotone increasing on [a, b]. For n E h1, set h = (b - a)/n.
. , n, set xi = a + ih. Then 9' = { x o , x,, ... ,
Also f o r i = 0, 1 , .. is a partition of
[a, b] that satisfies Axi = h for all i = 1, . . . , n. Since f is monotone increasing on
[a, b], mi = f(xi_ i) and Mi = f(x,). Therefore,
NOV) - `ê(911f) = [f(x;) -f(xi-1)]Ax,
=h (f (xi) - f (xi- 01
(b - a)
n [f(b) - f(a)].
Given e > 0, choose n E N such that
(b - a)
n
[f(b) - f(a)] < E.
For this n and corresponding partition 9, OIL.(91, f) - 2(91j) < e. Thus f is integrable
on [a, b].
The Composition Theorem

We next prove that the composition iP -f, of a continuous function gyp, with a Riemann
integrable function f, is again Riemann integrable. As an application of Lebesgue's the-
orem we will present a much shorter proof of this result later in the section.
6.1.9 THEOREM Let f be a bounded Riemann integrable function on [a, b] with Range
f C [c, d]. If ip is continuous on [c, d], then tp -f is Riemann integrable on [a, b].
Proof. Since 9 is continuous on the closed and bounded interval [c, d], ip is bounded
and uniformly continuous on [c, d]. Let K = sup{IQp(t)I: t E [c, d]}, and let e > 0 be
given. Set e' = e/(b - a + 2K).
Since ip is uniformly continuous on [c, d]. there exists 8, 0 < S < e', such that
19(s) - (P(01 < e' (4)
for all s, t E [c, d] with Is - tI < S. Furthermore, since f E 9%[a, b], by Theorem
6.1.7 there exists a partition 9 ' = {xo, . . . . . x}} of [a, b] such that
%(9',f) - _y(9', f) < S2.
218 Chapter6 The Rlemann and Riemann-Stieltjes Integral
To complete the proof we will show that
°1491, (P of) - 201, (P of) C E. (5)

By Theorem 6.1.7 it then follows that (p of E Jt(a, b].
For each k = 1, 2, . . . , n, let ink and Mk denote the infimum and supremum off
on [xk_ 1, xk). Also, set
mA = inf{cp(f(t)) : t E [x,-,, xJ) and

M,*t = sup{p(f(t)):1 E [xk_I,xk]}.
We partiton the set 11, 2, ... , n} into disjoint sets A and B as follows:
A={k:Mk-mk<6} and B={k:Mk-mka6}.
We first note that I f(t) - f(s)l 5 Mk - ink for all s, t E [xk_,, xk]. Hence if k E A.
then by inequality (4),
IcP(f(t)) - V(f(s))j < e'

for all s, t e [xk_,, xk]. But
Mk - m4 = sup{cp(f(t)) -'p(f(s)): s, t E [Xk_1,xk]}.
lift*
Therefore, - mk s e' for all k E A. On the other hand, if k E B. then
Mk* -mk < 2K. Thus
11491, (P -f) - 2(9"P of) _ (Mk - mk )AX* + (Mk - mk)AX&
kEA tea
5 E' 7, Axk + 2K T, AXk
kEA kEB
5 e'(b - a) + 2K Y, Axk.
kEB
But for k E B, 6 s Mk - Mk. Therefore,
Oxk 5 7' (Mk - mk)I&Xk s s (043V)

S kEB
- 49V)) < S < e'.
kEB
Hence by the above,

'k(91, V o f) - £('P, qp o f) 5 e'(b - a) + 2KE' = e.
This establishes inequality (5), and thus q) of E cJt[a, b]. U
As a consequence of the previous theorem, if f is Riemann integrable on [a, b],
then so are the functions If I and f2 . For emphasis we state this as a corollary.
6.1.10 COROLLARY If f E 9t[a, b], then If ! and f 2 are Riemann integrable on [a, b].
A natural question to ask is whether the composition of two Riemann integrable
functions is Riemann integrable. In Example 6.1.14(b) we will show that the answer is
an emphatic no!
6.1 The Riemann integral 219
Lebesgue's Theorem
In Theorem 6.1.8(a) we proved that every continuous function on [a, b] is Riemann in-
tegrable on [a, b]. By Exercise 16, this is also true for every bounded function on [a, b]
that is continuous except at a finite number of points. On the other hand, as a conse-
quence of Theorem 6.1.8(b), every monotone function on [a, bJ is Riemann integrable.
Hence, for example, if 1
is an enumeration of the rational numbers in [0, 1 ] and
c > 0 are such that T, c, converges, then by Theorem 4.4.10,
00
f(x) = 71 C. 1(x -
n=1
is monotone increasing on [0, 1 ], and thus is Riemann integrable on [0, 11. By Theo-
rem 4.4.10, the function f is continuous at every irrational number and discontinuous at
every, rational number in [0, 1].
We now state the beautiful result of Lebesgue that provides necessary and suffi-
cient conditions that a bounded real-valued function on [a, b) be Riemann integrable.
To properly state Lebesgue's result we need to introduce the idea of a set of measure
zero. The concept of measure of a set will be treated in detail in Chapter 10. The basic
idea is that the measure of an interval is its length. This is then used to define what we
mean by measurable set and the measure of a measurable set. At this point we only need
to know what it means for a set to have measure zero.
6 1.11 DEFINITION A subset E of P has measure zero if given any e > 0, there exists a
finite or countable collection {1n)n of open intervals such that
ECUJ and e,
n
where l(1n) denotes the length of the interval In.
6.1.12 EXAMPLES
(a) Every finite set E has measure zero. Suppose E = {x1, . . . , xN} is a finite subset of
R. For each n = 1, 2, ... , N, as in Figure 6.4 (with N = 6), let
/ E
Nxn+2N, .
Then
N N
ECU / and > l(!n) = e.
n=1 n=1
Therefore, E has measure zero.
it 14 12 13 16 l5
Xt
(0) (0)
X4 X2 X3 X6 XS
Figure 6.4
220 Chapter 6 The Riemann and Riemann-Stleltjea Integral
(b) Every countable subset of R has measure zero. Suppose E = {x,}00 I is a countable
subset of R. Let f > 0 be given. For each n E N. let
E E l
Since xn E 1 for all n, E C U. 11,,. Thus since 1(I4) = E12n,

00
j l(In) = E 2n = E.
w=1 n=1
As an example, the set Q of rational numbers has measure zero.

(c) The Cantor set Pin [0, 1] has measure zero (Exercise 21).
We now state the following theorem of Henri Lebesgue, the proof of which will be
given in Section 6.7. This result appeared in 1902 and provides the most succinct form
of necessary and sufficient conditions for Riemann integrability.
6.1.13 THEOREM (Lebesgue) A bounded real-valued function f on [a. b] is Riemann inte-

grable if and only if the set of discontinuities off has measure zero.
Remark. If f is continuous on [a, b], then clearly f satisfies the hypothesis of Theo-
rem 6.1.13 and thus is Riemann integrable. If f is a bounded function that is continuous
except at a finite number of points, then by Example 6.1.12(a) the set of discontinuities
off has measure zero. Hence f E 9t[a, b]. If f is monotone on [a, b], then by Corollary
4.4.8, the set of discontinuities off is at most countable, and thus by Example 6.1.12(b).
has measure zero. Hence again f E 9t[a, b].
As an application of Lebesgue's theorem we give the following short proof of The-

orem 6.1.9.
Proof of Theorem 6.1.9 Using Lebesguevs Theorem. As in Theorem 6.1.9,
suppose f E 9t[a, b] with Range f C [c, d], and suppose rp : [c, d] -, R is continuous.
Let
E = {x E (a, b): f is not continuous at x} and
F = {x E [a, b): rp of is not continuous at x}.
By Theorem 4.2.4, F C E. Since f is Riemann integrable on (a, b], the set E has mea-
sure zero, and as a consequence so does the set F. Therefore. V of E 9t[a, b]. Q
6.1.14 EXAMPLES
(a) As in Example 4.2.2(g), let f be defined on [0, 1] by
11, x=0,
f(x) =
1n, if x = n in lowest terms, x * 0.
Since f is continuous except at the rational numbers, which have measure zero, f is Rie-
mann integrable on [0, 1 ]. Furthermore, since $(P, f) = 0 for all partitions 9 of [0, 1 ],
Jf(x)dx = 0.
(b) Let f be the Riemann integrable function on [0, 1 ] given in (a), and let g : [0, 1 ] - R
be defined by
0, x0,
g(x) __ 11, x E (0, 1].
Since g is continuous except at 0, g E %[0, 1 ]. But for x E [0, 1 ],
(g of)(x) = J 1, if x is rational,
0, if x is irrational.
By Example 6.1.6(a), g of 6E Jt[O, 1).
EXERCISES 6.1
1. Let f (x) = 1 - x2, x E [ -1, 2]. Find Sf(9, f) and qt(9J) for each of the following partitions of [ -1, 2].
a. 9 = {-1,0, 1,2)
b.9={-1,-12,12,1,2,2}
2. Show that each of the following functions is Riemann integrable on [0, 2], and use the definition to find f02 I
1, 0, sx < 1
*a f(x) = 2, 15 x 5 2 ' b f (x) 3, 25 x<2
-2, sx 52
3. Show that each of the following functions are Riemann integrable
2 on [a, b] and find f; f.
0, a s x < c,
*a. f(x) = c (c a constant) b. f(x) = x = c.
2,
1, c<xsb
4. Use one of the methods of Examples 6.1.6 to find fo f for each of the following monotone functions f.
*a. f(x) = [3x], where (x) is the greatest integer function b. f(x) = x[2x]
c. f(x) - 2x + I d. f(x) = 1 - x2
5. Prove that f. *f = 0 for any real-valued function f on [a. a].
6 . *If f , g E 3t(a, b) with f ( x ) 5 g ( x ) f o r all x E [a, b], p r o v e that f f s f g.
7. a. Suppose f is continuous on (a. b] with f(x) z 0 for all x E [a, b]. If f. f = 0. prove that f(x) = 0 for all
x r= [a. b].
b. Show by example that the conclusion may be false if f is not continuous.
222 Chapter 6 The Rlemann and Riemann-Stieltjes integral
& *Let f : [(0, 11 - R be defined by
f()= {0, xEa,

X. xEQ
Compute fo f and fo f. Is f integrable on [0, 11?
9. Suppose f is a nonnegative Riemann integrable function on [a, b] satisfying f(r) = 0 for all r E Q fl [a, b].
Prove that f; f = 0.
10. *Use the method of Example 6.1.6(c) to show that
xdx= 3(b3-a3)(0:a<b).
Ja
11. Suppose f is monotone increasing on [a, b]. For n E N, set It = (b - a)/n. L e t 9n = {xo, x,, ... , xn} where for
each k = 0,. . ., n,xk = a + kh.
a. Prove that
(b a)
0 s Wt(9n,f) - f [f(b) -f(a)]
b. Prove that f; f = lim bU(9n, f).

12. Use the previous exercise to evaluate the following integrals.
b
I
(I (I
*a. x dx b. J (3x - 2) dx c. J x3 dx d. x3 dx
J0 -2 0
1a
13. Let f be a bounded function on [a, b]. Suppose there exists a sequence {9n} of partitions of [a, b) such that
lim SC(9n, f) = lim QU (9n, f) = L (L E 18).
ntiPO n-wo
Prove that f is Riemann integrable on [a, b] with fa f = L.
14. a. If f e 9t[a, b], prove directly (without using Theorems 6.1.9 or 6.1.13) that If I E 9t[a, b].
b. If Ill E 9t[a, b], is f e 9t[a, b]?
15. *a. If f e 91[a, b], prove directly that f2 E 9t[a, b].
b. Give an example of a bounded function f on [a. b] for which f2 E 9t[a, b], but f e 9t[a, b].
16. *Suppose f is a bounded real-valued function on [a, b] that has only a finite number of discontinuities. Prove di-
rectly that f is Riemann integrable on [a, b].
17. a. If E has measure zero, prove that every subset of E has measure zero.
b. If E,, E2 have measure zero, prove that E, U E2 has measure zero.
ao
c. If each En, n = 1, 2, ... , has measure zero, prove that U E. has measure zero.
n=I
18. Use Theorem 6.1.13 to prove that if f, g r= 9t[a, b], then f + g E 9t[a, b].
19. Prove directly that the function f of Example 6.1.14(a) is Riemann integrable on (a, b).
20. Let f : [a, b] -+ R be continuous. Suppose that for every Riemann integrable function g : (a, b] -+ R the product
fg is Riemann integrable and f; fg = 0. Prove that f(x) = 0 for all x E [a, b].
21. Prove that the Cantor set P in [0. 1) has measure zero.
S.2 Properties of the Riemann Integral

In this section we derive some basic properties of the Riemann integral. As in the pre-
vious section, [a, b], a < b, will be a closed and bounded interval in R, and dtra, b;
denotes the set of Riemann integrable functions on [a, b].
6.2.1 THEOREM Let f, g E aAt[a, b]. Then

rh b
(b
(a) f + g E 1.[a, b] with
Ja
(f + g) = J f + f g, a
a
b
(b) cf E It [a, b] for all c E with jci = c j f, and

a a
(c) fg E Ft[a. b].

Proof. (a) The integrability of f + g is actually a consequence of Theorem 6.1.13.
However, in establishing the formula for the integral of (f + g), we will also obtain the
integrability of f + gas a consequence. Let 91 = { x 0 . . . . . be a partition of [a, b].
For each i = 1, . .. , n, let
M;(f) = sup{f(t):tE [x,_I,x,]},
M,(g) = sup{g(t) :I E [x,- i, x,]}.
Then f(t) + g(t) s M,(f) + M, (g) for all t E [x,_ x,], and thus
sup{f(t) + g(t) : t E [x1_ 1,x1]} M,(f) + M,(g).
Therefore, for all partitions 9' of [a, b],
U49,f + g) c 'U(QP,f) + g).
Let e > 0 be given. Since f, g E 9t[a, b], there exist partitions 91f and 91, of
[a, b] such that
b
°lt(9f,.f) < jf+

a
2
and Jg±.
a
Let 9 = al'fU 2P8. Since 91 is a refinement of both al'f and 9914,

b
au(D,f + g) < 1u(9'f,f) + att(9g, g) < j f+ f g + e.

b
a a
Therefore,
b b fb
(f+g) < f+ ag+e.
f
Since the above holds for all e > 0,
b
jb(f+ g) _- jaf+ f bg
224 Chapter6 The Riemann and Riemann-Stieltjes Integral
A similar argument proves that
jn(f + g) 2: J f + f g.
Thus the lower integral of (f + g) is equal to the upper integral of (f + g), and as a con-
sequence, (f + g) E 9t[a, b] with
fb(f+g)=
ff+ fhg.
(b) The proof of (b) is left as an exercise (Exercise 1).
(c) To prove (c), we first note that
fg = 4[(f+g)2- (f-g)2].
By part (a) the functions (f + g) and (f - g) are integrable on [a, b]. and by Corollary
6.1.10, (f + g)'- and (f - g)2 are integrable. Hence by parts (a) and (b),fg is integrable
on [a, b].
6.2.2 THEOREM 1f f E 9t(a, b), then I f I E 9t[a, b] with
f E 9t[a, b], by Corollary 6.1.10, I f (E 9t[a, b]. Let c = ±1 be such

that I 5"f I = c fa f . Since cf(x) I f(# for all x E [a, b], °11(9, cf) <_ °U(9, If I)
for any partition 9 of [a, b]. Therefore, since both cf and If I are integrable,
fa cf f.' I f I. Combining. the above we have
jhj<
ff1 =cff- li I
6.2.3 THEOREM Let f be a bounded real-valued function on [a, b], and suppose
a < c < b. Then f E 9t[a, b] if and only if f e 9t[a, c] and f E 9t[c, b]. if this is the
case, then
Jf=faf+ff (6)
Proof. Let c E R satisfy a < c < b. We first prove that if f is a bounded real-valued
function on [a, b], then
ff= faf+ ff (7)
Suppose 9, and 92 are partitions of [a, c] and [c, b], respectively. Then 9 = 91 U 92
6.2 Properties of the ltiemann Integral 225
is a partition of [a, b] with c E 9. Conversely, if 9 is any partition of 'a, b with

c E 91, then 99 = 91, U 912 where 91, and 912 are partitions of [a, c] and [c, b]. respec-
tively. For such a partition °.P,
°tt(P,f) _ °u(P1,f) + aLt(92,f) > Ji + Ji.

a a
If 9. is any partition of [a, b], then 91 = 9. U {c} is a refinement of 9. containing c.

Therefore,
c b
'149-1f) ' 014911f) ' J f

a
+ J
c
f
Takin$ the infimum over all partitions a of [a, b] gives '
f
f-jaf+If
To prove the reverse inequality, let e > 0 be given. Then there exist partitions
91, and 91, of [a, c] and [c, b], respectively such that
%(91,.f)< (f+2 and °lt(P2, f) <

fb
E
f + 2*
a J
. Let 91 = 9', U 9'2. Then
J f S °u(oy,f) = °u(9'1,f) + gt(9D2, f) < J`f + J f+

a a c
Since e > 0 was arbitrary,
t f:5 fa'+ Jf
c
which when combined with the previous inequality proves equation (7). A similar ar-
gument also proves
f f=a a
+ f
'ff f"c
. (8)
If f is integrable on [a, c] and [c, b], then by equations (7) and (8),
b
J af+jf=lfslf=Jf+Jf
Therefore, f E Jt[a, b] and identity (6) holds. Conversely, if f E 9t[a, b], then by
equations (7) and (8),
Jf+if=lf+If
226 Chapter 6 The Riemann and Riemann-Stleltjes Integral
Since the lower integral off is always less than or equal to the upper integral off, the
ff
above holds if and only if
f-ff
11 b
and
a u
Hence f E k[a, c] and f E dt [c, b]. Q
Riemann's Definition of the Integral

We close this section by comparing the approach of Darboux with the original method
of Riemann.
6.2.4 DEFINITION Let f be abounded real-valued function on [a. b] and let 9

{xo. xl,....x} be a partition of [a, b]. For each i = 1, 2, ... , n. choose t, E
[xi _ 1, x,]. The sum
f(p.f) = ,if(oaxi
I
is called a Riemann sum off cith respect to the partition 91 and the points {t,}.
Figure 6.5 A Riemann Sum f(91. f) of f
In the Riemann approach to integration, one defines the integral of a bounded real-
valued function f as the limit of the Riemann sums off. Since 9(91,f) depends not only
on the partition °,D = {x0, xi,... , but also on the points t; E [x;_ 1, xi], we first
need to clarify what we mean by the limit of the Riemann sums &'(9, f ). For a partition
_ {x0, x1.... , xn} of [a, b]. set
I1911 max{oxi: i = 1,2,...,n}.
The quantity Il9 11 is called the norm or the mesh of the partition 9.
6.25 DEFINITION Let f be a bounded real-valued function on [a, b]. Then

lim -Y(9, f) =
119)1-'o
if given c > 0, there exists a S > 0 such that
l f (oaxi - I < E (9)

i=n[
for all partitions 9 of [a, bl with 11911 < S, and all choices of ti E [x, _ I, xi].
6.2.6 THEOREM Lei f be a bounded real-valued function on [a, b]. If

lim SP(9, f) = I,
11911-'0
then f E 9t[a, b] and fa f = 1. Conversely, if f E 9t[a, b], then 111im0 &(91,f) exists and
el-
rb
lim x(9,.1) = J f.
11911-.0
a
Proof. Suppose , la tprt0 P(9, f) = I. Let e > 0 be given, and let S > 0 be such that
inequality ( 9 ) holds f o r all partitions 9 = {x0, xl, ... , xn} of [a, b) with 11211 < S,
and all ti E [xi_ 1, xi]. By the definition of Mi, for each i = 1, ... , n, there exists
i E [xi_ 1, xi] such that f M, - e. Thus
n
OU(911f) = M4xi
i=1
< f(OAXi + E
//
AX,
i=1 i.l
< 1+ E + e(b -a]= 1+ E[ 1 + b -a].
Similarly T(21,f) > I - E[ 1 + b - a]. Therefore,
°U.(9, f) - 2(91,f) < 2E[ 1 + b - a].
Thus as a consequence of Theorem 6.1.7, f E 9t[a, b) with fa f = I.
Conversely, suppose f r= Jt[a, b]. Let M > 0 be such that I f(x)I s M for all
x E [a, b]. Let e > 0 be given. Since f E 9t[a, b], by Theorem 6.1.7 there exists a
partition . of [a, b] such that
Jb
e <,f) U(,.f)< f f+E.

228 Chapter 6 The Rlemann and Riemann-Stieltjes Integral
Suppose % = {xa, . . . , x,,,}. Let S = e/NM, and let 9 = ( ) , 0 ,...,y . ) be any parti-
tion of [a, b] with 11 9 II < S. As in the definition of the integral, let
M;=sup{f(x):xE[x;_1,x;]}, i= I,. . .. N.
Consider any interval [yk_ 1, yk], k = 1, . .. , n. This interval may or may not contain
points x. E a. Since 9 . contains N + I points, there are at most N - I intervals
[y,-,, yk] which contain an x, E a, i # 0, N. Suppose, as in Figure 6.6, {.x t, ... , xj+m}
C [Yk- 1, Yk] [f xx * yk_ 1, set M k = sup{ f (x) : x E [yk_,, xj]}. Similarly, if xl+m * yk,
set Mk = sup{f(x) : x E [x, +m, y'k]}.
Yk - 1 Y4
X1 xj+1 x142 X1
Figure 6.6
Let tk E [yk_ 1, Yk] be arbitrary. Since 1(t) - f(s)I s 2M for all t, s E [yk_,, yk],
f(tk) 2M + M1 s, s = 1, . .. , m, and
f(tk) 2M + Mk, i = 1, 2.
Therefore,
m
f(tk)AYk =f(rk)(x, Yk-1) + Y f(rk)OXI+, + f(1k)O'k - x,+m)

$=1
< 2M.Yk + Mk(x, - Yk-1) + M M,+,'&x,+, + M4(Yk - X,-m)

s=1
< 2M6 + 01L(9k, f ),

where 2 - P , = { y k -, x,, - , x;+., yk} is a partition of [yk- 1, Yk]. If the interval [Yk-,. yk]
contains no x; E a, i * 0, N. we simply let 9k = {yk_ 1, yk}. Let 9' = U. Then 9'
is a partition of [a, b] that is also a refinement of 2. Since at most N - I intervals
[Yk-1, Yk] contain a point of 9, other than xo and xN,
n n
f(9j) _ k=1
f(tk)DYk < 2M(N - 1)8 +
k-1
6u(9k,f)
b
<2e+04(9',f)<2e+Ru(a,f)<3e+ f. f.
a
A similar argument proves that
`9'(P,f) > jfb - 3e.

a
herefore,
f(P,f) - JiI < 3e.

6.2 Properties of The Riemann Integral 229
Since this holds for any partition QP _ {yo, ... , y,,} of [a, b] with 11 1,1P11 < a and any
choice of tk e k = 1, . . . , n,
rb
6.2.7 EXAMPLE In this example we will use the method of Riemann sums to evaluate
f; x dx. Since f(x) = x is Riemann integrable on [a, b],
rb
x dx = lim 9(91, f).
Ja
Since the limit exists for any t, E [x1_,, x,], we can take t; = (x;_, + x;)/2. With this
choice of t,,
1, R T R 2 2
fyfy /!11
"(.7,f) 7,(x,_, + xj)(x,
2j_j
1 - x,_,) 2 (x, - x?2 (b - Q ),
1
i -11
EXERCISES 6.2
1. Prove Theorem 6.2.1(b).
2. a. Use the method of Riemann sums to evaluate f; x2 d,
b. Use the method of Riemann sums to evaluate fa x" dx, n E N, n >_ 3.
3. a. Let fbe a real,valued function on [a, b] such that f(x) = 0 for all x # c, , ... , cR. Prove that f E A[a, b]
with f; f = 0.
*b. Let f, g E.9t[a, b] be such that f(x) = g(x) for all but a finite number of points in ra, b]. Prove that
faf=fog.
c. Is the result of part (a) still true if f (x) = 0 for all but countably many points in [a, b]?
4. Let f E 9t[-a, a], a > 0. Prove each of the following.
a. If f is even (i.e., f (-x) = f(x) for all x (E [ -a, a]), then f f = 25f.
b. Iff is odd (i.e.,f(-x) _ -f(x) for all x E [-a, a]), then f f = 0.
5. *Let f be a bounded real-valued function on [a, b] such that f E k[c, b] for every c, a < c < b. Prove that
fE f; f= lira, f b f
6. Let f be continuo6s on [0, 1]. Prove that
lim f(jf(x)dx.
7. Use the previous exercise to evaluate each of the following limits. (You may use any applicable methods from cal-
culus to evaluate the definite integrals.)
R R R
1 k n
a. lim
R-"OnYk k' b. lim Y c. lim
i n2 k2 R-rook=, n2 + k2
8. AS in Example 4.4.11(a), let f be the monotone function on [0, 1) defined by
f(x) _ 2" 1(x -
x = n/(n + 1), n E N. Find fo f(x) dx. Leave your answer in the form of an infinite series.
9. Suppose f E 9t[a, b] and c E R. Define f, on [a - c. b - c] by f.(x) = f(x + c). Prove that
ffEdt[a-c,b-c]with
Jf(x)ds = I f(x) dx.
6
10. Let f. g, h be bounded real-valued functions on [a. b] satisfying f(x) s g(x) S h(x) for all x E [a. b]. If
f. h E Il[a. b] with f. f = fu h = 1, prove that g E 'lt[a. b] with ff g = 1.
6.
I Fundamental Theorem of Calculus
In this section we will prove two well-known and very important results. Collectively
they are commonly referred to as the fundamental theorem of calculus. To Newton and
Leibniz, integration was the inverse operation of differentiation. In Leibniz's notation,
this result would simply be expressed as f f'(x) dx = f (x), where the symbol f (denot-
ing sum) was Leibniz's notation for the integral. This notation is still used in most cal-
culus texts to denote the indefinite integral of a function. Since the modern definition of
the integral is based on either Riemann or Darboux sums, we now need to prove that in-
tegration is indeed the inverse operation of differentiation. Both versions of the funda-
mental theorem of calculus presented here are essentially due to Cauchy, who proved
the results for continuous functions.
As was illustrated in Examples 6.1.6 and 6.2.7, the computation of the integral of
a function, using either Darboux's or Riemann's definition, can be extremely tedious.
For nontrivial functions, these computations are in most instances impossible. The first
version of the fundamental theorem of calculus provides a major tool for the evaluation
of Riemann integrals. We begin with the following definition.
6.3.1 DEFINITION Let f be a real-valued function on an interval L A function F on I is

called an antiderivative off on I if F'(x) = f(x) for all x E 1.
Remark. An antiderivative, if it exists, is not unique. If F is an antiderivative off,

then so is F + C for any constant C. Conversely, if F and G are antiderivatives off, then
F'(x) - G'(x) = 0 for all x E [a, b]. Thus by Theorem 5.2.9, G(x) = F(x) + C for
some constant C.
6.3.2 THEOREM (Fundamental Theorem of Calculus) if f E R[a, b] and if F is an

antiderivative off on [a, b], then
J f(x) dx = F(b) - F(a).2
2. The quantity F(b) - F(u} is usually denoted by F(x)l;.

6.3 Fundamental Theorem of Calculus 231
Proof. Let 9 = (x0, x 1 . . . . . xa} be any partition of [a, b]. If F is an antiderivative

off, then by the mean value theorem, for each i = 1, ... , n, there exists t, E (x,_ x,)
such that
F(x,) - F(x;_,) = f(t,)Ox,.
Therefore,
a a
FIf(y)Ox, _ [F(xi) - F(x;_1)] = F(b) - F(a).
Since
a
i=1
we have
f f (x) A 5 F(b) - F(a) < J f (x) dx.

Ja a
Thus if f is Riemann integrable on [a, b],
Ibf(x)dx=F(b)-F(a). U
a
Remark. The above version of the fundamental theorem of calculus is consider-

ably stronger than the version given in most elementary calculus texts in that it does
not require continuity of the function f. We only need that f is integrable and that it
has an antiderivative on [a, b]. We will illustrate this in part (b) of the following
example.
6.3.3 EXAMPLES
(a) If f (x) = x", n E N, then F(x) _ + 1
X"11 is an antiderivative off. Thus for any
a,bER,
fb
x" dx = I (bm+1 - an+1).

a
n+l
(b) Consider the function F on (0, 11 defined by
xZ sin X, x # 0,
F(x) =
A straightforward computation gives
-cos lx + 2x sin 1x , x # 0 '

f(x) - F'(x) = I
0, x=0.
232 Chapter6 The Riemann and Rlemann-Stieltjes Integral
Then f is bounded on [0, 1] and continuous everywhere except at x = 0. Thus

f E 9t[0, 1 ], and by the previous theorem,
i
f=F(1)-F(0)=sin1.
J
(c) Let f be defined on [0, 21 by

f(x)_ 1, 0 5x< 1,
x- 1, 1 5x 52,
and for x E [0, 2] let F(x) be defined by
rr
F(x) = 'f(r) dt.
J0
If 0 s x S 1, then F(x) = fo 1 dt = x. On the other hand, if 1 < x s 2, then
F(x)= 1 1dt+ 1(t- 1)dr= 2x2-x+Z.

o i
Thus
x, 0sxs1,
F(x) 2x2-x+2, 1 <x<-2.
The graphs off and F are given in Figure 6.7. Even though f is not continuous at x = 1,
the function F is continuous everywhere on [0, 2]. This in fact is always the case, as
will be proved in Theorem 6.3.4.
The following version of the fundamental theorem of calculus, also attributed to

Cauchy, proves that for continuous functions, integration is the inverse operation of dif-
ferentiation.
6.3.4 THEOREM (Fundamental Theorem of Calculus) Let f E 9t[a, b]. Define F on

[a, b] by
F(x) = 'f(t) dt.

Ja
Then F is continuous on [a, b]. Furthermore, if f is continuous at a point c E [a, b:,

then F is differentiable at c and
F'(c) = f(c).
Proof. The proof that F is continuous is left as an exercise (Exercise 1). Suppose f is
continuous at c E [a, b). We will show that F' (c) = f (c). Let h > 0. Then by Theo-
rem 6.2.3,
c+h c c+h
F(c + h) - F(c) J f(t) dt - f(t) dt = { f(t) A
f o c
Therefore,
F(c + h) - F(c) - f(c) - C c+h f(t) f (C)

dt -
h hJ
- 1
h
f" [f(t) - f(c)) dt.
Let E > 0 be given. Since f is continuous at c, there exists a S > 0 such that
NO - f(c) I < E
f o r all t, I t - c I < S. Therefore, if 0 < h < S,
F(c + h) - F(c) - f(c) I
Jc+h
h
If(t) - f(c)I dt
c
h Jc +h
<h Edt=E.
c
Thus
F(c + h)' - F(c) =

m Ac);
hli-,o h
i.e., F+' (c) = f (c). Similarly, if f is continuous at c E (a, b], F'_ (c) = f (c), which
proves the result. Q
Remarks
(a) If f is continuous on [a, b], then an antiderivative of f always exists; namely, the
function F given by
F(x) = -f(t) dt, a 5 x :S b.
Since f is continuous, F'(x) = f(x) for all x E [a, b]. As a consequence, we obtain
the following more elementary version of Theorem 6.3.2 normally encountered in the
study of calculus: If f is continuous on [a, b] and G is any antiderivative of f, then
f f = G(b) - G(a).
(b) Integrability of a function f on [a, b] does not imply the existence of an antideriv-
ative off. For example, if f is monotone increasing on [a, b] and F(x) = f` f, then for
any c E (a, b), F' (c) = f (c+) and F_(c) = f(c-) (Exercise 14). Thus if f is not con-
tinuous at c, the derivative of F does not exist at c.
The Natural Logarithm Function

As our first application of the fundamental theorem of calculus, we use the result to de-
fine the natural logarithm function In x.
6.3.5 EXAMPLE For x > 0, let L(x) be defined by
L(z) = dt.
i
Since f (t) = 1/t is continuous on (0, oo), by Theorem 6.3.4, L(x) satisfies L'(x) = l/x
for all x > 0. Furthermore, since L'(x) > 0 for all x E (0, oo), L is strictly increasing
on (0, oo).
We now prove that the function L(x) satisfies the usual properties of a logarithm
function; namely,
(a) L(ab) = L(a) + L(b) for all a, b > 0,
(b) L(e) _ -L(b), b > 0, and
(c) L(b') = rL(b), b > 0. r E R.
To prove (a), consider the function L(ax), x > 0. By the chain rule (Theorem 5.1.6),
d L(ax) =
dX
.a=X=L'(x).
ax
Thus by Theorem 5.2.9, L(ax) = L(x) + C for some constant C. From the definition of
L we have L(l) = 0. Therefore,
L(a) = L(1) + C = C.
Hence L(ax) = L(a) + L(x) for all x > 0, which proves (a). The proof of (b) proceeds
analogously. It is worth noting that for the proof of (a) and (b) we only used the fact
that L'(x) = l/x and L(1) = 0.
To prove (c), if n E N, then by (a) L(b") = nL(b). Also by (b),
L(b-") = L\\b/"/ nL(b) -nL(b).

=
Therefore, L(b") = nL(b) for all n E Z. Consider L('6) where n E N. Since
nL(-6) = L(b), L(-6) _ L L(b). Therefore,
L(b') = rL(b)
for all r E 0. Since L is continuous, the above holds for all r E R.
Our final step will be to prove that L(e) = 1, where e is Euler's number of Exam-
ple 2.3.5. To accomplish this we use the definition to compute the derivative of L at 1.
Since L'(1) exists,
L(l + L(1)
1 =L'(1)= ]im
n--,ao
i
= 1mnL(l+nl
= lim L(t I + !T) = L(e).
n
The last equality follows by the continuity of L and the definition of e. Therefore,
L(e) = 1 and the function L(x) is the logarithm function to the base e. This function is
usually denoted by log, x or In x, and is called the natural logarithm function.
Consequences of the Fundamental Theorem of Calculus

We now prove several other consequences of Theorem 6.3.4. Our first result is the mean
value theorem for integrals.
6.3.6 THEOREM (Mean Value Theorem for Integrals) Let f be a continuous real-valued
function on [a, b]. Then there exists c e [a, b] such that
b
f =f(c)(b - a).
fa
Proof. Let F(x) = fa f. Since f is continuous on [a, b], F'(x) = f(x) for all x E [a, b].
Thus by the mean value theorem (Theorem 5.2.6), there exists c E [a, b] such that
Jbf = F(b) - F(a) = F'(c)(b - a) =f(c)(b - a). 0

a
An alternative proof of the above can also be based on the intermediate value the-
orem using the continuity of f. This alternative method will be used in the proof of the
analogous result for the Riemann-Stieltjes integral.
6.3.7 THEOREM (Integration by Parts Formula) Let f, g be differentiable functions on

[a, b] with f, g' E 9t[a, b]. Then
Jb
fg' = f(b)g(b) - f(a)g(a) - gf'.
'Proof. Since f, g are differentiable on [a, b], they are continuous and thus also inte-
grable on [a, b]. Therefore by Theorem 6.2.1(c), fg' and gf' are integrable on [a, b].
Since
(fgY = gf' + fg',
the function (fg)' E 9t[a, b]. By the fundamental theorem of calculus (Theorem 6.3.2),
b Vg), b b
f(b)g(b) - f(a)g(a) = J = J gf + J fg',

a a a
from which the result follows. 0
6.3.8 THEOREM (Change of Variable Theorem) Let tp be differentiable on [a, b] with

rp' E &[a, b]. If f is continuous on I = rp([a, b]), then
h v(b)
f((p(t))w'(t) de f(x) dx.
a e(a)
Proof. Since tP is continuous, I = rp([a, b]) is a closed and bounded interval. Also,
since f a rp is continuous and tp' E 9t[a. b], by Theorem 6.2.1(c), (f o W)tp' E 9t[a, b].
If I = r([a, b]) is a single point, then tp is constant on [a. b]. In this case rp'(t) = 0 for
all t and both integrals above are zero. Otherwise, for x E I define
F(x) = J f(s) ds.

(a)
Since f is continuous, F'(x) = f(x) for all x E 1. By the chain rule,
d F(1P(t)) = F'((p(t))1P'(t) =f(tp(t))w'(t)
for all t E [a, b]. Therefore by Theorem 6.3.2,
Jbf((p(t))tp'(t) dt = F((p(b)) F(,p(a)) = J b)f(s) ds.

a (a)
Remark Another version of the change of variable theorem is given in Exercise 10.
6.3.9 EXAMPLE To illustrate the change of variable theorem, consider fo t/(1 + t2) di.
If we let (,p(t) = 1 + t2 and f(x) = 1/x, then
Joe
dt = 2 102f(`p(t))(p'(t) dt
1 + t2
which, by Theorem 6.3.8

_Js
xdx=21n5.
2
EXERCISES 6.3
1. *Let f E 9t[a, b]. For x E [a. b], set F(x) = f' f. Prove that F is continuous on [a, b].
2. For x E [0, 1 ], find F(x) = fo f(t) dt for each of the following functions f defined on [0, 11. In each case verify
that F is continuous on [0, 11. and that F'(x) = f(x) at all points where f is continuous.
1, 0 s x <2
a. f(x)=xz-3x+5 *b. f(x) =
-2, 1 5 x <-1
c. f(x) = x - [3x] d. f(x) = x[3x]
3. Let f(t) be defined by

0st<1,
f(t) {b'i2, 15t<-2,
and let F(x) be defined by F(x) = fo f(t) dt, 0 s x 5 2.
a. Find F(x).
b. For what value of bin the definition off is F(x) differentiable for all x E [0, 2]?
4. Let f be a continuous real-valued function on [a, b] and define Hon [a, b] by H(x) = f; f. Find H'(x).
3. Find F'(x) where F is defined on [0, 1 ] as follows.
*b. F(x) = fcosf2dr

1
a. F(x) _ dt
0 1+t
t
e. F(x) _ 1 + t3 dt d. F(x) = JIQ) dr, where f is continuous on [0, 1 ].
0
6. *Let L(x) be defined as in Example 6.3.5. Prove that g1/x) _ -L(x).

7. Suppose f: [a, b] -+R is continuous and g, h : [c, d] -+ [a, b] are differentiable. For x E [c. d] define
(AX)
H(x) = 1 f(t) dt.
Mx)
Find H'(x).
& Let f : R -+ R be continuous. For a > 0 define g on R by
g(x) = J f (t) dt.

a a
Show that g is differentiable and find g'(x).

9. *Let f be a continuous real-valued function on [a, b], g E 31[a, b] with g(x) 2z 0 for all x E [a. b]. Prove that
there exists c E [a, b] such that
Jef(x)g(x)dx =-f(c) J°g(x)dx.

o a
10. Prove the following change of variables formula: Let rp be a real-valued differentiable function on [a, b] with
(x) # 0 fof all x E [a, b]. Let 0 be the inverse function of rp on I = tp([a. b]). If f : 1-+ R is continuous on 1, then
(o(b) f(t)4x'(t)dt.
. f(rp(x))dx = -
a )v(a)
238 Chapter6 The Riemann and RIemann-Stieltjes Integral
11. Evaluate each of the following integrals. Justify each step.
a I
2 In x
X
dx
+b (4
iI vx
dx
(I x
c. J x In x d x d. J t at + b dr. a, b> 0
0 0
I
1/X
f j4
e, 10 dx dx
1+V I x
12. Suppose f is continuous on [0, 1 ]. Prove that
lim 'f (x") dx = f (O).

J0
13. Suppose f : (a. b] -> R is continuous. Let M = max{jf(x)j: x E [a, b]}. Show that
/b
tim I If(x)r dx = M.
14. Let f be a monotone increasing function on [a, b] and let F(x) = fo f(t) dt. Prove that F' (c) = f(c+) and
F_(c) = f(c-) for every c E (a, b).
15. Use Theorem 4.4.10 and the previous exercise to construct a continuous, increasing function F on [0, 1) that is
differentiable at every irrational number in [0, 11, and not differentiable at any rational number in (0, 1 ].
16. Cauchy-Schwarc Inequality for Integrals: Let f, g E &(a, b]. Prove that
(jb
J,bf(x)g(x) dx I 2 5 (Lb g2(x) dl/
f 2(x) dx/
(Hint: For a, 0 E R consider fa (af - /3g)2.)

17. The Exponential Function: As in Example 6.3.5, let L : (0, oo) - R be defined by
L(x) f!df.
I t
a. Show that L is strictly increasing on (0, oo) with Range L = R.

b. Let E : R -i (0, oo) denote the inverse function of L. Use Theorem 5.2.14 to prove that E'(x) = E(x) for an
x E R. (The function E is called the natural exponential function, and is often denoted by exp(x).)
c. Prove that E(x + y) = E(x)E(y) for all x, y E R.
d. Prove that E(x) = e` for all x E R. where e is Euler's number, and e` is as defined in Miscellaneous Exercise
3 of Chapter 1.
d.4 Improper Riemann Integrals

In the definition of the Riemann integral of a real-valued function f. we required that f
be a bounded function defined on a closed and bounded interval [a, b. If these two hy-
potheses are not satisfied, then the preceding theory does not apply, and we have to
modify the definition. In this section we will briefly consider the changes required if the
function j is unbounded at some point of its domain, or if the interval on which f is de-
fined is itself unbounded.
Unbounded Functions on Finite Intervals

If f E It[a. b] with If (x)I s M for all x E [a. b], then by Theorem 6.2.3, f E A [c, b]
for every c E (a, b), and
f of
- hf I = IJ f I < Ja` U1 < M(c - a).
As a consequence, if f E lt[a, b], then f E k[c, b] for every c E (a, b) and
e n
limf f=Jf.
C a
a bounded real-valued function defined only on (a, b] with

f E R[c, b] for every c E (a, b). If we define f(a) = 0, then f is a bounded function
on [a. b) satisfying f E 1t[c, b] for every c E (a. b). By Exercise 5. Section 6.2.
f E %[a, b] with
lim
C a- J
o f= f.
f a
Furthermore, by Exercise 3 of Section 6.2, the answer is independent of how we define

f(a).
Using the above as motivation, we extend the definition of the integral to include
the case in which f is unbounded at an endpoint. This extension of the integral is also
credited to Cauchy.
6.4.1 DEFINITION Let f be a real-valued function on (a, b) such that f E dt[c, b] for even
c E (a, b). The improper Riemann integral off on (a, b], denoted f,' f, is defined to be
b b
limf f,
Ca'
Ja f
provided the limit exists. If the limit exists. then the improper integral is said to be con-
vergent. Otherwise, the improper integral is said to be divergent.
240 Chapter6 TheRlemann and Riemann-Stleltjes Integral
A similar definition can be given if f is defined on (a, b) and is unbounded at b. If

f is unbounded at a point p, a < p < b, we consider the improper integrals off on the
intervals [a, p) and (p, b]. If each of the improper integrals exists, then we define the
improper integral off on [a, b] to be the sum
i, b
af+Jf.
l
6.4.2 EXAMPLES
(a) Consider the function f(x) = 1/x on (0, 1 ]. This function is clearly unbounded at
0. Since f is continuous on (0, 1 ],f e Jt[c, 1 ] for every c E (0, 1). By Example 6.3.5.
1
dx=Inc.
X
To evaluate lim In c we consider In r", where 0 < r < I and n E N. Since In r" _
C- Oo
n In r and In r < 0,
lim In r" = limn In r = -oo.
Therefore, since r" _+0 as n -+ oo and In x is monotone increasing, lirm In c = -oo.
Thus the improper integral of 1/x on (0, 1) diverges.
(b) For our second example we consider the improper integral of L(x) = In x on (0, 1 ].
Since L(x) is continuous on (0, 1), L E 9t[c, 1 ] for every c E (0, 1). Consider ff' In x dx.
If we take g(x) = x, then by the integration by parts formula,
j In x dx = L(x)g'(x) dx
J
= L(l)g(l) - L(c)g(c) - J L'(x)g(x) dx
=-cInc-1+c.
By the substitution c = 1/t, t > 0, and l'Hospital's rule,
-1n t
lira c In c = lim
1-.0o t
= -lim 1 = 0.
t-00 t
Therefore,
lim I In x dx = -1.
c-+O'
C
Hence the improper integral of In x converges on (0, 1 ) with
I lnxdx = -1.
0
(c) Consider the function f defined on [ -1, 1 ] by

0, -1 sx<-0.
f(x)= 1,
x
0<xsl.
Here f is unbounded at 0. For this example, the improper integral off over [ -1, 1' fails
to exist because the improper integral off over (0, 1 ] does not exist.
(d) There are some significant differences between the Riemann integral and the im-
proper Riemann integral. For example, if f E 9t[a, bJ, then by Theorem 6.2.1.
f2 E 9t[a, b]. This however is false for the improper Riemann integral! For example.
if f (x) = 1/Vx, x E (0,1 ], then
Jldx
x ,l° z
Thus the improper integral off converges on (0, 1]. However, f2(x) = t/x, and the im-
proper integral of 1/x on (0, 1 ] diverges. Also, if f E 9t[a, b], then [f I E 9t[a, b]. This
.again is false for improper Riemann integrals. In Exercise 4, we provide an example of
a function f for which the improper integral off converges, but the improper integral of
L f I diverges.
Infinite intervals
We now turn our attention to functions defined on infinite intervals.
6A.3 DEFINITION Lei f be a real-valued function on [a, co) that is Riemann integrable on
[a, c) for every c > a. The Improper Rlemann integral off on [a. oo), denoted f, f.
is defined to be
j= lim
C-400
J f,
n
provided the limit exists. If the limit exists, then the improper integral is said to be con-
vergent. Otherwise, the improper integral is said to be divergent.
If f is a real-valued function defined on (-oo, bJ satisfying f E 9t[c, b] for every

c < b, then the improper integral off on (-oo, b] is defined as
f f bf,
provided the limit exists. If f is defined on (-oo, oo), then the improper integral off is
defined as
xf+I 30f,
for some fixed p E R, provided that the improper integrals off on (-oo, p] and [ p, oo)
are both convergent. For a function f defined on (-oo, oo), care must be exercised in
computing the improper integral off. It is incorrect to compute
For example, if f (x) = x, then
lim 1 x dx = c--tlim 1 (c2

C-.0o oo 2
- (-c)2) = 0.
-c
However, the improper integral off on (-oo, oo) is divergent since

JOO fc
f lim
c-.oo
1
x dx = lim - c2 = oo.
c-.oo 2
Remark. If f is nonnegative on (a, oo) with f E Jt[a, c] for every c > a, then f.f
is a monotone increasing function of c on [a, oo). Thus lim f. f exists either as a real
number or diverges to no. For this reason, if f (x) 0 for C
all x E [a, oc), we use the notation
rb
Jf(x)dx < oo or J f(x) dx = 00
to denote that the improper integral off on [a, oo) converges or diverges to no respec-
tively.
6.4.4 EXAMPLES
(a) Let f (x) = 1/x2, x E [ 1, oo). Since f is continuous on [ 1, co), f e Jt[ 1, c) for
everyc > 1. Therefore,
Ji=l c-+ao Ix
imf'Idx=liml-1+1)=1.
c-.oo\ C
Thus the improper integral converges to the value 1.

(b) In this example we consider the function f (x) = (sin x)/x, x E [v, oo). Sincef is
continuous, f E 9t[a, c] for every c > v. This function has the property that the im-
proper integral off on [a, o0) converges, but the improper integral of VI diverges. The
proof of the convergence of the improper integral off is found in Exercise 7. Here we
will show that
Isin xl
IfI = r =oo.
For n E N, consider
sin xI ((k "i° Isin xl

J dx = J dx.
km1 x
a x ka
Since the integrand is nonnegative,
(k +3/4h. Isin
(k+I)" {sin xf
>
x J(k+1/4hr x
On the interval [(k + 1)1r, (k + ;)1r], sin xl \/2. Also,

I
>_ hr,+4)JrJ.
forallxErlk+4(k
x (k t )ir / \\
(k+3/4)n Isin x / ll ll
x Ia 22(k+1)rr
k+ 1'
"+
sin xl 1,/2
f x
dx 4 k_ik+l
By Example 2.7.4, the series 7,k 1 k diverges. Therefore,
(°° !sin ("+I)vIsinxI

J dx = ".4
lim dx = oo.
A
X
]r
x
As the previous example shows, the convergence of the improper integral off does
imply the convergence of the improper integral of If 1. If f is a real-valued function
on [a, oo) such that f e ,[a, c] for every c > a and the improper integral of If I con-
verges on (a, oo), then f is said to be absolutely integrable on [a, oo). An analogous
definition can be given for unbounded functions on a finite interval. We leave it as an
exercise to prove that if f is absolutely integrable on [a, oo), then the improper integral
off also converges on [a, oo) (Exercise 5).
We conclude this section with the following useful comparison test for improper
'integrals.
244 Chapter 6 The Riemann and Rlemann-Stleltjes Integral
6.4.5 THEOREM (Comparison T e s t ) Let g : [a, oo) -+ R be a nonnegative f u n c t i o n satis-

fying g E 9t[a, c] f o r every c > a and f g(x)dx < oo. If f : [a. oo) --+R satisfies
(a) f E Jt[a, c] for every c > a, and
(b) I f (x) 5 g(x) for all x E [a, oo),
then the improper integral off on [a, oo) converges, and
(x) dx :5 J g(x)dx.
J
Proof. The proof is left to the exercises (Exercise 6).
N EXERCISES 6.4
1. For each of the following functions f defined on (0, 1), determine whether the improper integral off converges. If it
converges, find fo f.
Inxx
'a f(x) = Xp> 0 f(x) _ x c. f(x) =
*d. f(x) = x In x e. f(x) = (I + x) I f. f(x) = tan (Zx)

+ x)
2. For each of the following, determine whether the improper integral converges or diverges. If it converges, evaluate
the integral.
*a. fee*dx b. C. Jx_'dx. p > I

o i x
°O In.r °D dx °G dx
d. X ax *e. E x l n x P> 1
2 2
x(l n x)P"
f'xZ+,dx f IV
g. h. f(x2+ 1)pdx. P> I I. 1)dx
(x2+ 1)(x+
3. For each of the following, determine the values of p and q for which the improper integral converges.
*a. f I/2xpjlnxjpdx b. c. Jaxp[ln(l +x)]°dx

0 2 0
4. Petfbedefined on(0, 1]by
f(x) = ±(X2 sin X2)- 2x sin X2 - cos X2.
Show that the improper Riemann integral off converges on (0, I ], but that the improper integral of VI diverges on
(0,1 1.
5. *If f is absolutely integrable on [a. oo) and integrable on [a, c] for every c > a, prove that the improper Riemann
integral off on [a, oo) converges.
7. Let f(x) _ (cos x E [1r,oo).

+a. Show that the improper integral of If I converges on [1r, oo).
*b. Use integration by parts on [ir, c], c > ir, to show that
sin x
J dx exists.
x
8. Show that fox-' sin x dx converges for all p, 0 0,set
The function r is called the Gamma function.

a. Show that the improper integral converges for all x > 0.
b. Use integration by parts to show that r(x + I) = xr(x), x > 0.
c. Show that r(1) = I.
d. For n r= Rl, prove that r(n + 1) = W.
The Riemann-Stieltjes Integral

In this section we consider the Riemann-Stieltjes integral, which, as we will see, is an
extension of the Riemann integral. To motivate the Riemann-Stieltjes integral we con-
sider the following example from physics involving the moment of inertia.
6.5.1 EXAMPLE Consider n-masses, each of mass m;, i = 1, ... , n, located along the
x-axis at distances r, from the origin with 0 < rt < < r (Figure 6.8). The mo-
ment of inertia 1, about an axis through the origin at right angles to the system of
masses, is given by
1 = ; r?m;.
0 HIM
Figure 6.8
3. Since the results of this section are not specifically required in subsequent chapters, this topic can be omit-
ted on first reading of the text.
246 Chapters The Riemann and Riemann-Stieltjes Integral
On the other hand, if we have a wire of length I along the x-axis with one end at the ori-
gin, then the moment of inertia 1 is given by
! = jxp(x)dx.
where for each x E [0, 1). p(x) denotes the cross-sectional density at x.
Although these two problems are totally different, the first being discrete and the
second continuous, the Riemann-Stieltjes integral will allow us to express both of these
formulas as a single integral. In the definition of the Riemann integral, we used the
length Axi of the ith interval to define the upper and lower Riemann sums of a bounded
function f. The only difference between the Riemann and Riemann-Stieltjes integral is
that we replace Axi by
Aai = a(xi) - a(x; _ 1),
where a is a nondecreasing function on [a, b]. Taking a(x) = x will give the usual
Riemann integral. Although the modification in the definition is only minor, the
consequences are far-reaching. Not only will we obtain a more extensive theory of
integration, but also an integral that has broad applications in the mathematical
sciences.
Definition of the Riemann-Stieltjes Integral

Let a be a monotone increasing function on [a, b], and let f be a bounded real-valued
function on [a, b]. For each partition 9 = {xo, x1 , .. . , of [a, b], set
Aai = a(xi) - a(x;_ 1), i= .. , n.
Since a is monotone increasing, Aai ? 0 for all i. As in Section 6.1, let

m, = inf{ f(t) : t E [x,_ i, xi]},
M, = sup{f(t):t E [xi_1,xi]}.
As for the Riemann integral, the upper Riemann-Stieltjes sum off with respect to a
and the partition 91, denoted all.(9, f, a), is defined by
kt(9, f, a) _ M;Aa,.
r=
Similarly, the lower Riemann-Stleltjes sum off with respect to a and the partition 9,
denoted 2(9, f, a), is defined by
-Y(9, f, a) _ miAai.
i=1
Since m; s M, and Aai ? 0, we always have 2(91,f a) < all(9, f, a). Further-
more, if m s f(x) s M for all x E [a, b], then
m[a(b) - a(a)] <_ Z(9, f, a) 5 Rt(9, f, a) < M[a(b) - a(a)) (10)
for all partitions 9 of [a, b]. Let 96 be any partition of (a, b). Since Mi s M for all i and
Aai ? 0,
n
Mi4ai -.5 Mla1 = M Lai = M[a(b) - a(a)).

i=1 i=1
Thus °U,(9), f, a) <_ M[a(b) - a(a)]. The other inequality follows similarly. In the above
we have used the fact that
n
I Aai = (a(xl) - a(x0)) + (a(x2) - a(xl)) + . + (a(xn) - a(xn-1))

i-
= a(xn) - a(xo) = a(b) - a(a)-

In analogy with the Riemann integral, the upper and lower Riemann-Stleltjes in-
tegrals off with respect to a over [a, b], denoted fQ f da and f; f da respectively, are
defined by
b
f da = inf{ott(91, f, a) : 9 is a partition of [a, b]},
Jfda = sup{f(9, f, a) : 9' is a partition of [a, b]},

n
By inequality (10) the set {qL(91, f, a) : 9' is a partition of [a, b]} is bounded below, and
thus the upper integral off with respect to a exists as a real number. Similarly, the lower
sums are bounded above, and thus the supremum defining the lower integral is also
finite. As for the Riemann integral, our first step is to prove that the lower integral is less
than or equal to the upper integral.
6.5.2 THEOREM Let f be abounded real-valued function on [a, b], and a a monotone
increasing function on [a, b]. Then
Jfda.
e
Jabfda:5
Proof. As in the proof of Lemma 6.1.3, if 9'* is a refinement of the partition 9, then
X(91,f a) <_ X(91*,f, a) :5 gL(96*,f a) c ah(y,f, a).
Thus if 91, 2, are any two partitions of [a, b],
f, a) s 2(9 U a j a) :S %(91 U ll, f, a) < 61L(,, f, a).
Therefore T(91, f, a) <_ 6tL(9;, f, a) for any partitions 9, 9.. Hence
Jfda = sp96,f,a) a)
a
for any partition 9.. Taking the infimum over a gives the result. O
248 Chapters The Riemann and Riemann-Stieltjes Integral
6.5.3 DEFINITION Let f be a bounded real-valued function on [a, bl, and a a monotone
increasing function on [a, b). If
h h
fda = Jfda.
f
then f is said to be Riemann-Stieltjes integrable or integrable with respect to a on

[a, b). The common value is denoted by
h (h
fda or ! f (x) da (x),
a a
and is called the Riemann-Stieltjes integral off with respect to a.
As was indicated previously, the special case a(x) = x gives the usual Riemann in-
tegral on [a, b).
6.5.4 EXAMPLES
(a) Fix a < c 5 b. As in Definition 4.4.9, let 1,.(x) = I (x - c) be the unit jump func-
tion at c defined by
= f0, x<c,
1`(x) t 1, x?c.
We now prove the following: If f is a bounded real-valued function on [a, b] that is con-
tinuous at c, a < c < b, then f is integrable with respect to 1, and
h 'f(x)
1 f dl,. = dl (x - c) = f (c).
a Ja
For convenience, we set a(x) = 1,.(x), which is clearly monotone increasing on [a, b).
Let 91 = {xo, x1..... x,} be any partition of [a, b]. Since a < c b, there exists an
index k, 1 5 k 5 n, such that
Xk- I < C 5 xk.
Then
Da* = a(xk) - a(xk_1) = I - 0 = 1, and
Aa;=0, foralli#k.
Therefore,
olt(91, f, a) = Mk&ak = Mk = sup{f(t) : xk_, <_ t 5 xk}, and

Z(91, f, a) = mk&ak = Mk = inf{f(t) : xk_ 15 t 5 xk}.
Since f is continuous at c, given e > 0, there exists a S > 0 such that
f(c) - e < f(t) <f(c)+E
f o r all t E [a, b] with I t - cI < S. If 91 is any partition of (a, b] with x, - xx_, < S
for all j, then
AC) - E S Mk c Mk :5 AC) + E.
Therefore, f (c) - E S Y(91, f , a) : 5U(91, f, a) <_ f(c) + e. As a consequence.
f(c) - E b f dc, _ I b f da f (c) + E.

Jn o
Since e > 0 was arbitrary, the upper and lower integrals off are equal, and thus f is in-
tegrable with respect to a on [a, b] with
Jb
f da = f (c).
(b) The function

f(x) _ 1 x E
0, x6EQ,
is not integrable with respect to any nonconstant monotone increasing function a.
Suppose a is monotone increasing on [a, b], a < b, with a(a) * a(b). If 91 =
(xo, x,, ... , is any partition of [a, b], then m; = 0 and M; = 1 for all i = 1, ... , n.
Therefore, f, a) = 0 and
Rt(91,f, a) i a; = a(b) - a(a).
Thus f is not integrable with respect to a.
Remark. It should be noted that in Example 6.5.4(a) only left continuity off at c is
required (Exercise 2(a)).
The following theorem, which is the analogue of Theorem 6.1.7 for the Riemann
integral, provides necessary and sufficient conditions for the existence of the Riemann-
Stieltjes integral. The proof of the theorem follows the proof of Theorem 6.1.7 verba-
tim and thus is omitted.
6.5.5 THEOREM Let a 'be a monotone increasing function on [a, b]. A bounded real-
valuedfunction f is Riemann-Stieltjes integrable with respect to a on [a, b] if and only
if for every e > 0; there exists a partition 9 of [a, b] such that
a) - a) < E.
Furthermore, if 9 is a partition of [a, b] for which the above holds, then the inequality
also holds for all refinements of P.
We now use the previous theorem to prove the following analogue of Theorem
6.1.8. Except for some minor differences, the two proofs are very similar.
6.5.6 THEOREM Let f be a real-valued function on [a, b) and a a monotone increasing

function on [a, b).
(a) If f is continuous on [a, b), then f is integrable with respect to a on [a, b].
(b) 1f f is monotone on [a, b] and a is continuous on [a, b], then f is integrable
with respect to a on [a, b].
Proof. (a) The proof of (a) is identical to the proof of Theorem 6.1.8(a) except that
given e > 0, we choose 71 > 0 such that
[a(b) - a(a)),q < e.
The remainder of the proof now follows verbatim the proof of Theorem 6.1.8(a).
(b) For any positive integer n, choose a partition 91 = {x°, x,.... , of [a, b]
such that
Aa; = a(x;) - a(x;_,) = n [a(b) - a(a)].

Since a is continuous, such a choice is possible by the intermediate value theorem. As-
sumef is monotone increasing on [a, b]. Then M; = f(xi) and m; = f(x,_1). Therefore,
1!t(91,f a) - 2(91,f, a) _ i (Ax) - f(x+-,)]Dal

i-t
[a(b) - a(a)]
n
U(x;) - Ax,-0]
i=1
[a(b) - a(a))
n U(b) - f(a)).
Given e > 0, choose n E N such that

[a(b) - a(a)]
[f(b) - f(a)) < e.
n
For this n and corresponding partition 91, °ll(?, f, a) - f, a) < e, which proves
the result. Q
Remark. In part (b) above, the result may be false if a is not continuous. For exam-
ple, if a < c < b, then the monotone function 1, is not integrable with respect to 1, on
[a, b] (Exercise 2b). The article by Kenneth Ross, listed in the Supplemental Read-
ing section, suggests an alternative approach to the Riemann-Stieltjes integral that
eliminates this difficulty.
Properties of the Riemann-Stieltjes Integral

We next consider several basic properties of the Riemann-Stieltjes integral. For rota-
tional convenience, we make the following definition.
6.5.7 DEFINITION For a given monotone increasing function a on [a, bJ, 9R(a) denotes
the set of bounded real-valued functions f on [a. b] that are Riemann-Stieltjes inte-
grable with respect to a.
6.5.8 THEOREM
(a) If f, g E gi(a), then f + g and cf are in %(a) for every c E R and
Jb b b
(f + g)da = fda + f g da,

a 1a a
b b
cf da = c fda.
a I
(b) If f E 9t(a;), i =Jfd(ai
1, 2, then f E 9R(a1 + a2) and
+ a 2) = Jfdai + Jfda2.
(c) If f E °Jt(a) and a < c < b, then f is integrable with respect to a on [a, c]
and [c, b] with
b b
a fda= f.cf da+ Jcfda.

f
(d) If f, g E t(a) with f (x) <_ g(x) for all x E [a, b), then
rb rb
Ja f da Ja g da.
(e) If I f(x)I 5 M on [a, b] and f E 9R(a), then If I E 9t(a) and

b f
f da < blf I da M[a(b) - a(a)].
Ja a
Proof. We provide the proofs of (b) and (e). The proofs of (a), (c), and (d), along with
other properties of the Riemann-Stieltjes integral. are left to the exercises.
(b) Since f E %(a;), given e > 0, there exists a partition 9D,, i = 1, 2, such that
-U.(gli, f, ai) - X(`l...f a.) < Z

Let 9' = 91, U 912. Since 91 is a refinement of both 91, and 912, inequality (11) is still
valid for the partition 91. Thus since
A(a1 + a2); = A(a,)1 + A(a2)i

forall i = 1,...,n,
61L(@1 j, al + a2) - £(9, J, al + a2)

= °lL.(9,f a,) - `-e(91,f, a,) + °1L(J,f a2) - `-e(g,f a2)
E e
< 2 + 2 = 6.
Therefore by Theorem 6.5.5, f E 9t(al + a2). Furthermore, for any partition 91 of [a, b),
2(91j, a, + a2) _ ZP,f al) + £(9j a2)
t((b rb
Jfda,+Jfda2
a a
< °d(91,f, a,) + °tt(9a,f, a2) a, + a2).

Thus since f E 9t(al + a2),
b b b
f f d(a, + a2) =
I f da, + f f dal.
(e) Suppose f E 9t(a) and 91 _ {x°, x,. ... . xa} is a partition of [a, b]. For each
i= 1,. ..,n,let
M; = sup{f(t): t E [xi_,,xi]}, M* = sup{I f(t)I:t e [xi_,,xi]},
mi = inf{f(t): t e [xi_,,x.]}, m* = inf{I f(t)I : t E [xi_,,xi]}.
If t, x E [xi_,, xi], then
II f(t) - If(x)II 1f(t) -f(x)I <_Mi - mi.
Thus M*i - m * 5 Mi - mi, for all i = 1, . .. , n, and as a consequence,
°u(91, If 1 , a) - ,x(91, I f I, a) s tlL(91, f, a) - X(Ol, f, a).
Therefore by Theorem 6.5.5, If I E 9t(a). Choose c = t l such that c f f da -e 0. Then
JQfda =cJafda= J cfdas jbIfIdcr MJ da
= M [a(b) - a(a)].
As for the Riemann integral, we also have the following mean value theorem and
integration by parts formula for the Riemann-Stieltjes integral.
6.5.9 THEOREM (Mean Value Theorem) Let f be a continuous real-valued function

on [a, b], and a a monotone increasing function on [a, b]. Then there exists c E [a, b]
such that
f da = f(c) [a(b) - a(a)J-

J
Proof. Let m and M denote the minimum and maximum off on [a. b} respectively.
Then by Theorem 6.5.8(d),
b
m[a(b) - a(a)] s
f fda <_ M[a(b) - a(a)].
a
If a(b) - a(a) = 0, then any c E [a, b] will work. If a(b) - a(a) # 0, then by the in-
termediate value theorem there exists c E [a, b) such that
b
f (c) = a(b) -1 a(a) J. fda.
6.5.10 THEOREM (Integration by Parts Formula) Suppose a and Pare monotone increasing
functions on [a, b]. Then a E 91.(/3) if and only if fl E 9t(a). If this is the case,
J b a d$ = a(b)S(b) - a(a)9(a) - J b/3 da.

a a
Proof. By Exercise 9, for any partition 91 of [a, b],

1u(91, a, (3) = a(b)(3(b) - a(a)f(a) - 2(91, (3, a), and
`-x(91, a, 0) = a(b)j3(b) - a(a)f(a) 111(91, (3, a).
Therefore,
11[(91, a, /3) - 201, a, /3) = 1401, /3, a) - 49. is, a).
From this identity it immediately follows by Theorem 6.5.5 that a E 9t(t) if and only
if /3 E M(a). Furthermore, if /3 E 9t(a), then given e > 0, there exists a partition 9' of
[a. b] such that
W(91, /3, a) >

Jb!3da_e.
a
Hence,
Jb/3
Jba d p 1U,(9', a, $) < a(b)f3(b) - a(a)$(a) - da + E.

a a
Since the above holds for any E > 0,

b b
a d( < a(b)R(b) - a(a)$(a) - J (3da.

f
A similar argument using the lower sum proves the reverse inequality.
254 chapter 6 The Riemann and Riemann-Stieltjes Integral
We conclude this section with two results that represent the extremes encountered
in Riemann-Stieltjes integration. As in Example 6.5.4(a), let I(x - c) be the unit jump
function at c E R. Suppose {s"},N,- I is a finite subset of (a, b] and are nonneg-
ative real numbers. Define the monotone increasing function a on [a, b] by
N
a(X) _ I cn I (X - Sn).
n-1
fa
f da =
N 1bN
If f is continuous on [a, b], then by Example 6.5.4(a) and Theorem 6.5.8(b),
c"
1
f(x) dl (x - sn) = "'1
I Cnf (sn) (12)
n=1 a
Suppose {s"}: I is a countable subset of (a, b] and {cn}: , is a sequence of nonnegative

real numbers for which E cn converges. As in Theorem 4.4.10, define a on [a, b] by
x
CO) _ I, c" 1(x - sn). (13)
"=I
Since 0 : I (x - sn) <_ 1 for all n, the series in equation (13) converges for every
x E [a, b], and a is a monotone increasing function on [a, b]. For such a function awe
have the following theorem.
6.5.11 THEOREM Let f be a continuous real-valued function on [a, b], and let a be the
monotone function defined by equation (13). Then f E Jt(a) and
rb oo+
( f da = G c"f(s")
fa n=1
Proof. Since f is continuous on [a, b], f E lt(a) (Theorem 6.5.6(a)). Let e > 0 be
given. Choose a positive integer N such that
00
I C. < e.
n-N+1
Define II and 02 as follows:

N
/ Cn1(x - sn).
/31(x) Cn1(x - Sn), /332(x)
nil n=N+I
Then a = 6, + $2, and by identity (12),

{(b N
J f d$l = I Cnf (sn)
fa n=1
Let M = max{I f(x)I : x (=- [a, b]). Then by Theorem 6.5.8(b) and (e),
1 Ja f da -n7, cnf(sn) I = J .fdf321 c M [$2(b) - /32(a)]

W
:5 Mlc,,<Me,
n=x+i
At the other extreme, if the monotone function a is also differentiable, then we
have the following result.
6.5.12 THEOREM Suppose f E 9t[a, b], and a is a monotone increasing differentiable

function on [a, b] with a' E 9t[a, b]. Then f E &(a) and
J f da = J f(x)a'(x) dx.
a a
Proof. Since both f and a' are Riemann integrable on [a, b], by Theorem 6.2.1(c),
fa' E 9t[a, b]. Let e > 0 be given. Since a' E 01[a, b], by Theorem 6.1.7 there ex-
ists a partition 91 of (a, b) such that
°U.(9, a') - Y(), a') < e. (14)
Let °. 2. _ {xo, ... , xn} be any refinement of 91. As in Theorem 6.2.6, for each i =
1, ... , n, we can choose s, E [xi_,, x,] such that
° L(°9 f, a) < f(s,)Aa; + e. (15)
By the mean value theorem, for each i = 1, ... , n, there exists t; E [xi _,, x,] such that
Dai = a(x;) - a(x,_1) = a'(ti)Axi.
Therefore,
n a
I f(s,)Aar = Yf(sr)a'(y)Ax,. (16)

i=i i=1
Let M = sup{ f(x)I :x E [a, b]}, and for i = 1, ... , n, let in, and M; denote the in-
fimtun and supremum respectively of a' over the interval [x_ ,, x,). Then
la'(si)-a'(t;)l:!-- Mi - m,
f o r all i = 1, ... , n. Therefore,
f(si)a'(ti),xi - In As;)a'(sAXi`
i 1 i-l i- 1f(si)I1a'(1i) - a'(sMAxi
`M(Mi - mi)ixi
= M(%(a, a') - -W(9, a')) < Me.
The last inequality follows by inequality (14) since .. is a refinement of 91. Therefore,
n
f(s;)a'(ti)Ax; S
i-i i- f(si)a'(s;)Ax, + Me <_ °U.(9, fa') + Me.
256 Chapter 6 The Riemann and Rlemann-Stieltjes Integral
Thus by equations (15) and (16).

'lt.(a, f, a) < 0U.(a. fa') + (M + 1)e.
Since fa' E 9t[a, b], there exists a partition ). of [a, b]. which is a refinement of 91.
such that
b
G(L(2, fa') < J fa' + E.
G
Thus
b
rh
f da < fa' + (M + 2)e.
t
Since this holds for any E > 0, fQ f da <_ f' fa'. A similar argument using lower sums
proves the reverse inequality. Thus f E J1(a) and
h b
J fda = I fa'.
a 1a
We now give several examples to illustrate the previous two theorems.
6.5.13 EXAMPLES
(a) First, we illustrate the finite version of Theorem 6.5.11; namely, identity (12). Consider
j exd(x].
Forx E [0, 2],
[x]=1(x- 1)+1(x-2).
Therefore since ex is continuous, it is integrable with respect to [x], and by identity (12),
J exd[x] = + e2.
(b) To illustrate Theorem 6.5.12, if a(x) = x2 on [0, 1), then for any Riemann inte-
grable function f on [0, 1 ],
ff(x)dx2 = 2 J fix) x dx.

0 0
In particular, if f(x) = sin lrx, then
1 f sin Trx dr' = 2 f x sin irx dx,

0 0
which by the integration by parts formula (Theorem 6.3.7)
X cos Trx I + n J cos irx dx

0 0
I
_ -- ITcos V + sin Vx U
?r
(c) As another illustration of Theorem 6.5.12,
13[x] 2 J ;[x]e2x dx
0 0
= 21 e - dx + 21 2e2' dx = 2e6 - e° - e2.

1 2
The Riemann-Stieltjes integral has important applications in a variety of areas. in-

cluding physics and probability theory. It allows us to express diverse formulas as a sin-
gle expression. To illustrate this, we again consider the moment of inertia problem of
Example 6.5.1.
6.5.14 EXAMPLE In this example we will show that both formulas of Example 6.5.1 can be
expressed as
I
I= x dm(x),
0
where m(x) denotes the mass of the wire or system from 0 to x. Clearly the function
m(x) is nondecreasing, and thus since x2 is continuous, the above integral exists.
In the first case, since 0 < rt < r2 < < r,,, and the masses m; are located at
r, (see Figure 6.8), m(x) is given by
m(x) m; l (x - r;).
Thus by Theorem 6.5.11,

r
x2dm(x) _ r?m;.
Io
On the other hand, if m(x) is differentiable and m'(x) is Riemann integrable, then by
Theorem 6.5.12,
Jx2 dm(x) = Jx2m'(x) dx.

0
It only remains to be shown that m'(x) is the density. The density p(x) of a wire (mass
per unit length) is defined as the limit of the average density. In the interval [x, x + Ax,
the average density is
m(x + Ax) - m(x)
Ax
Thus if m(x) is differentiable, p(x) = m'(x).
Rlemann-Stieltjes Sums
We conclude this section with a few remarks concerning Riemann-Stieltjes sums. As in
Definition 6.2.4, let f be a bounded real-valued function on [a, b], a a monotone in-
creasing function on [a, b], and 9' _ {xo, ... , a partition of [a, b]. For each
i = 1, 2, . , n, choosey E [x,_,, x;]. Then
F(`.'P, f a) = Ef(t,)Aa,
is called a Riemann-Stieltjes sum off with respect to a and the partition 9.

A natural question is whether the analogue of Theorem 6.2.6 is valid for the
Riemann-Stieltjes integral. Unfortunately, only part of the Theorem holds. If lim
MPH-+0
W(9, f, a) = I, then f e 911(a) on [a, b] and f. f da = I (Exercise 11). The converse,
as the following example will illustrate, is false! However, if f and a satisfy either of the
hypotheses of Theorem 6.5.6, then
1b
lim 9
f da = u9n-+O f, a).
a
The result where f is continuous and a is monotone increasing is given in Exercise 10.
6.5.15 EXAMPLE Let f and a be defined on [0, 21 as follows:

0, 0S x11,
f(x) 1, 1<x 2,
_ 0, 0 5 x < 1,
a(x) 1 c x < 2.
11,
Since f is left continuous at 1, by Example 6.5.4(a), fo fda = f(1) = 0. Suppose
= {xo, x,, ... , is any partition of [0, 2] with 11Z 91. Let k be such that
xk_, < I < xk. Then Aak = 1 and Aa; = 0 for all i # k. If tk E (1,xk],then
f, a) = 1. On the other hand, if tk = 1, then f(°,P, f, a) = 0. As a consequence,
n lim 9'(91,f, a)
does not exist.

EXERCISES 6.5
1. *Evaluate f ' f (x) da(x) wheref is bounded on (-1. 11 and continuous at 0, and a is given by
-1, x<0,
a(x) = 0, x = 0,
1, x > 0.
2. a. In Example 6.5.4(a), prove that if the function f is left continuous at c, then f is integrable with respect to 1, on
[a, b].
b. Show that 1, is not integrable with respect to 1,
3. Let a be nondecreasing on (a, b). Suppose f is bounded on [a. b] and integrable with respect to a on [a. b]. For
X E [a, b] set F(x) = f. f da.
*a. Prove that IF(x) - F(y)l s Mla(x) - a(y)l for some positive constant M and all x. y E [a, b].
b. Prove that if a is continuous at x E [a, b), then F is also continuous at x0.
c. Prove Theorem 6.5.8(d).
5. Use the theorems from the text to compute each of the following integrals.
W/2
*a. x d(sin x) b. J 3[x] dx2 *c. J x2 d[x]

J(o n o
d. 1([x] + x) da(x)
3 where a(x) - x2 + e'
I
4 I
e. ex d[x] f. 1 (x - [x]) dx3 g. J sin ax d[4x]

1 I 0
6. Verify the integration by parts formula with parts (b) and (c) of the previous exercise.
7. Find fu fda, where f is continuous on [0, 1 ] and
CO)
Leave your answer in the form of an infinite series.

& Let a be as in Exercise 7. Evaluate each of the following integrals. Leave your answers in the form of an infinite
series.
*a. Ix da(x) b. I 'a(x) dx

Jo Jo
9. Suppose a and j6 are monotone increasing on [a, b], and 91 is a partition of [a, b]. Prove that
11L(91, a, P) = a(b)$(b) - a(a)3(a) - (3, a).
10. Prove that if f is a continuous real-valued function on [a, bland a is monotone increasing on [a, b], then
imoY(P, f, a) exists and ii it,o pp, f, a) = f; f da.
a
11. Let f be a bounded real-valued function on [a, h] and a a monotone increasing function on (a. b]. Prove the fol-
lowing: If Arn0Y(91. f, a) = 1, then f E 9t(a) and f ."f da = 1.
12. *a. Let a be a monotone increasing function on [a, b]. If f E R(a), prove that j2 E Jt(a).
b. If f, g E llt(a), prove that f g E Jt(a).
9t(a) on [a, b].
13. If f E Jt(a) on [a, b] with Range f C [c, d], and rp is continuous on [c, d], prove that V of E
14. Suppose f is a nonnegative continuous function on [a, b], and a is nondecreasing on [a, b]. Define the function R
on [a, b] by 6(x) = fa f da. If g is continuous on [a, b], prove that
h b
1 g df = J fg da.
4,
Numerical Methods4
In this section, we will take a brief look at some elementary numerical methods that are
useful in obtaining approximations to Riemann integrals. Even though the fundamental
theorem of calculus provides an easy method for evaluating definite integrals, it is use-
ful only if we can find an antiderivative of the function being integrated. To illustrate
this, in Example 6.3.9 we showed that
z
dr = 2 In 5.
f 1 + 1--
JO
This, however, is not particularly useful if we do not know the value of In 5. By Exam-
ple 6.3.5,
ln5 = dr.
J
t
To obtain an approximation to In 5, we can choose from several available methods for
obtaining numerical approximations to the definite integral.
Approximations Using Riemann Sums

Since upper and lower sums are in general difficult to evaluate, Darboux's definition
of the Riemann integral is not particularly useful in obtaining numerical approxima-
tions. The most elementary numerical method is the use of Riemann sums. One of the
first mathematicians to use numerical methods was Euler, who considered sums of the
form
I J(x
k-1
.. t)(Xk - Xk-1)
4. The topics of this section may be omitted on first reading.

as an approximation to the integral. This is nothing but the Riemann sum for the parti-
tion 91 = {x0, x,, . . . , of [a, b] with tk = xk_, for all k.
In using Riemann sums to approximate the integral of f, it is convenient to take
equally spaced partitions. Let n E N, and set h = (b - a)/n. Define
xo=a, x,=a+h, (17)
Thus if 9 = {x0, x,, ... , with xi as defined, we always have .x; = h for all
i = 1, , n, and
n
Y,f(ti)Axi = h Yf(t,),
i-t i_,
where for each i = 1, ... , n, t; E [xi_,,xi). If we take t; = xi_ 1 for all i, then we ob-
tain Euler's formula above. Similarly, we could take t; to be the right endpoint xi of the
interval [xi_ 1, xi]. Another choice of t; would be the midpoint; i.e., t; = (xi_, + x;)/2.
For monotone functions, it is intuitively obvious that the midpoint gives a better ap-
proximation than either the right or left endpoint to the integral of f over the interval
[xi_ 1, xi](see Figure 6.9). With xi as defined by equation (17) and t; = (xi _, + x,)/2,
the above formula becomes
M (f) is called the nth midpoint \àpproximation to the integral off over
[a, b].
Regardless of the method used, it is important to be able to estimate the error be-
tween the true value and the approximate value. If f E 9t[a, b], then for any partition
Figure 6.9 Midpoint Approximation

of [a, b]. we always have

b
f f(x) dx - Y(P''f) t(P'f) - 20'f) (19)
1
for any choice of tk E [xk _ 1. xk ]. Inequality (19) follows from the fact that both the Rie-
mann sum of f and the integral of f lie between the lower and upper sum of f. If f is
monotone increasing on [a. b], then by the proof of Theorem 6.1.8(b),
GU(91rt,f) - T(91,,,f) = h[f(b) -f(a)].

Thus by inequality (19),
b
f(x)dx - h7,f(t;)` :5 hIf(b) -f(a)I (20)

If i-
for any choice of t; E [xi _ 1, x;]. Although we proved inequality (20) for a monotone in-
creasing function, the inequality is also valid for a monotone decreasing function. Thus
for monotone functions, inequality (20) provides an estimate of the error between the
true value and the approximate value.
6.6.1 EXAMPLE The function f(x) = 1/x is decreasing on [ 1, 5]. Thus by inequality (20),
for n E Id and h = (b - a)/n,
11n5 1f(ti)I :Sn15i 5 1
for any choice of t; E [xi _ 1, x;]. Thus with n = 8, the error is less than 0.4;
n = 160 only guarantees an error of less than slo = 0.02. We would be required to take
very large values of n to be guaranteed a sufficiently small error.
With n=8,h=Z andxi=I+i,i=0,1,...,8,andt,=xi_j=(1+i)/2
(the left endpoint),
8
1
9(Ps'f) = 1 +i
=2+3+4+5+6+7
I
8+9
= 1.8290 (to four decimal places).
To four decimal places, In 5 = 1.6094. Thus the error is 0.2196-less than the pre-
dicted error of 0.4.
If we use the midpoint approximation (18) (again with n = 8), then t; = 1 + (i - i) i'
which upon simplification equals (3 + 2i)/4. Thus
1 $ 4
M8(f)=2j3+2i
=2\5+7+9+ 111+ 13+ 15+ 17+ 19!

Using the midpoint approximation, the error is less than 0.01. This is considerably bet-
ter than predicted by inequality (20). As we will shortly see, this improved accuracy is
not a coincidence.
Inequality (20) provides an error estimate for monotone functions. For arbitrary
real-valued functions, under the additional hypothesis that the derivative is bounded, we
have the following theorem.
6.6.2 THEOREM Lei f be a real-valued differentiable function on [a, b] for which f'(x) is
bounded on [a, b]. Then for any partition 01 of [a, b],
jf(x)dx - II91IIM(b - a),

a
where M = sup{If'(x)I : x E [a, b]}.

Proof. Suppose 01 = {xp, x,, . . . , xj is a partition of [a, b]. Then
qt(o,f) - = G (Mi - mi)Ozi.

i=t
Since f is continuous on [a, b], for each i there exist si, si E [xi_,, xi] such that
Mi - mi = f(si) - f(si)
By the mean value theorem, f(si) - f(s;) = f'(ti)(si - s:) for some ti between si and s;.
Therefore, since If'(y)I Isi - s'i) s M11 91 II,
n
NOV) - 491,f) S MII91II 1 xi = II01IIM(b - a).

i-,
The result follows by inequality (19).
If f satisfies the hypothesis of the theorem, h = (b - a)/n, n E N, and the x; are
defined by equation (17), then
1a f(x)dx - hY f(ti) M(b - a)h
for any choice of ti E [x,_ 1, xi]. A slight improvement of this inequality is given in the
exercises (Exercise 6). If denotes the error between the true value and the ap-
proximate value, that is,
En() = f (x) dx - f ),
Ja
then the above inequality can simply be written as IE"(f)I s Ch, where C is a fixed
constant depending only on f and the interval [a, b]. Since the term "h" occurs to the
first power, this method is commonly referred to as a first order method.
In Example 6.6.1 we saw that the midpoint rule provided much greater accuracy
than using the left endpoint. This is not a coincidence as the following theorem proves.
6.6.3 THEOREM Iff is twice differentiable on [a, b] and f"(x) is bounded on [a, b], then
f (X)
I Ja
- M"(f) I <- M( 24 a)
b2'
where M = sup{I f"(x)I: x e [a, b]}.
Remark Since the error between the true value and the approximate value involves
the term h2, the midpoint approximation is a second order method.
Proof. To prove the result, we first prove that for each i = I, ... , n,
xXf(x)dx - hf(a+( i - ')h)I 4h' (21)
For t E [0, 2], consider the function

.+1
gi(t) = j
c, r
f(x)dx - 2tf(c;),
where ci = (x, _ I + xi)/2. Then
f(x)dx - hf
(a +(i-
gi(2)= h)).
Since f is continuous, by Theorem 6.3.4
g;(t) = f(ci + t) + f(ci - t) - 2 f'(c), and
g1(t) = f'(ci + t) - f'(ci - t).
By the mean value theorem,
f'(c, + t) - f'(c; - t) = f"(C)2t
for some C E (ci - t, ci + t). Since If"(l;)1 M, we obtain
Ig;(t)I <- 2Mt
for all t E [0, J. Since g;(0) = 0, by the fundamental theorem of calculus,
W(t)I = [fg7(x)dx I S
jIg7(x)Idr2Mfxdr1.rMt2.
0 0 0
Also, since gi(0) = 0, again by the fundamental theorem of calculus,

nr nr j8;(t)I Mr2)
J
0
8'(t)drl j0
dt = 24hj
which proves inequality (21). Therefore,
f(x)dx-M"(t)I = I j f(x)dx-hf1a+li-2)h)
j8i(2)I
jt,
82 24Mh3n = M(b - a)
h
24 h
2.
The Trapezoidal Rule

We now consider another common second order approximation method known as the
trapezoidal rule. In using Riemann sums, regardless of the choice of the points t, we
used rectangles to approximate the integral of the function f. In this method we will re-
place rectangles by trapezoids. Let f be a Riemann integrable function on [a, b] and let
91 = {x0, xr, .. , x"} be a partition of [a, b]. As in Figure 6.10, for each interval
.
[xi_,,x;], the area of the trapezoid formed by the points

(x;, f(xi)), (xi, 0) is given by
2[J(xi-I) +f(xi)] x.
xi
Figure 6.10
Summing these up gives
7, [f(xi-i) +f(x) ]Axi

2;_,
266 Chapter 6 The Riemann and Rlemann-Stleltjes Integral
as an approximation to the integral off on [a. b]. If, as previously, we set h = (b - a)In
and x; = a + ih, i = 0, ... , n, then the above sum becomes
h" h -I
2 7, [f(a + (i - 1)h) + f(a + ih)J = 2 {f(a) + 2 f(a + A) + f(b),.
If f is Riemann integrable on [a, b] and n E N, the quantity TT(f) defined by

n-I
h
Tn(f) = 2 [f(a) + 2 E f(a + ih) + f(b) J, (22)
where h = (b - a)/n, is called the nth trapezoidal approximation to the integral off
on [a, b). If we set yi = f(a + ih), then T"(f) can be expressed as
h l
TV) =2 Yo+22yi+Y.J.
The following theorem, under suitable hypothesis on f, provides an estimate of the
error of the trapezoidal approximation.
6.6A THEOREM 1f f is twice differentiable on [a, bJ and f"(x) is bounded on [a, b], then
b
M(b - a) h2
l f(x) dx - TT(f)
12
where M = sup{jf"(x)j : x E [a, b]}.

The above error estimate can also be expressed as follows: If f satisfies the hy-
pothesis of the theorem, then for n E N,
rb
I
J AX)
dx - TT(f) I - M(b - a)3 I
12 n2 (23)
In this form it is possible to determine the value of n required to guarantee predeter-

mined accuracy.
Proof. The proof of this theorem is very similar to the proof of Theorem 6.6.3. As a
consequence, we leave most of the details to the exercises (Exercise 5). The first step is
to prove that for each i = 1,...,n,
I f(x)dx-2[f(xi_l)+f(x;)]1 < h3.

.,-1 12
To accomplish this, consider the function

{('x ,+r
` f(x)dx 2[f(x,-1) +f(xi-1 + t))
Jx;_i
fort E [0, h]. Then g;(0) = 0 and
g,(h) = jf(x)dx - 2 [f(x;-1) +f(x,)]

By computing g;(t) and g"(t) one obtains from the hypothesis on f that Ig"(t)I S 1W.
Since g,(0) = g;(0) = 0, applying the fundamental theorem of calculus twice we ob-
tain, as in Theorem 6.6.3, that Jg,(h)I s ,2Mh3. The remainder of the proof is identical
to Theorem 6.6.3.
Simpson's Rule
The trapezoidal approximation T (f) amounts to approximating the function f with a
piecewise linear function which passes through the points {(x;, f(x;))}"=0 (see
Figure 6.11). Our intuition should convince us that one way to obtain a better approxi-
mation to the integral off over [a, b] is to use smoother curves. This is exactly what is
done in Simpson's rule, which uses parabolas to approximate the integral. To use qua-
dratic approximations we will need to use three successive points of the partition of
[a, b].This is because three points are required to uniquely determine a parabola.
Figure 6.11 Trapezoidal Approximation
Prior to deriving Simpson's rule, we first establish the following formula: If

p(x) = Ax2 + Bx + C is the quadratic function passing through the points (0, yo),
(h, y, ), (2h, Y2), then
( 2k h
J p(x) dx = (24)
0
s [Yo + 4Yi + Y2]
One way to derive this formula would be to first determine the coefficients A, B, C so
that p(O) = yo, p(h) = yi, p(2h) = y2, and then integrate p(x). This, however, is unnec-
essary. By integrating first,
J up(x) dx = 3 A(2h)3 + 2 B(2h)2 + C(2h)

0
= [8Ah2 + 6Bh + 6C)

3
= 3
[p(O) + 4p(h) + p(2h)] = 3
[Yo + 4Y. + Y2].
Let f E 9t[a, b] and let n E N be even. Set h = (b - a)/n. On each of the inter-
vals [a + 2(i - 1)h, a + 2ih], i = 1, ... we approximate the integral off by the
integral of the quadratic function that agrees with fat the points
Yo = f(a + 2(i - 1)h), y' = f(a + (2i - 1)h), y2 = f (a + 2ih).
By identity (24) this gives
[f(a + 2(i -- 1)h) + 4f(a + (2i - l)h) + f(a + 2ih)]

3
as an approximation to the integral off over the interval [a + 2(i - 1)h, a + 2ih].
Summing these terms from i = 1 to i gives
3[f(a)+4f(a+h)+2f(a+2h)
+4f(a+3h)+ +4f(a+(n-1)h)+f(b)]
as an approximation to the integral off over (a, b]. The quantity S (f) is called the nth
Simpson approximation to the integral off over [a, b]. If we set y; = f(a + ih), then
is given by
SM(f)= 3[Yo+4Yi +2Y2+4Y3+.
(25)
The following theorem, again under suitable restrictions on f, provides an error es-
timate for Simpson's rule.
6.6.5 THEOREM If f is four times differentiable on [a, b] and f t4i(x) is bounded on [a, b],
.
then for n E N even,

e
f(x)dx - Sr(f)1 < M(b - a) h4,
a
ISO
where M = sup{jf(4)(x)j : x E [a, b]}.
Remarks. Since the error term involves h4, Simpson's rule is a fourth order method.
Also, if f(x) is a polynomial of degree less than or equal to 3, then f(4i(x) = 0 and thus
f f (x) dx = S&).
0
If f satisfies the hypothesis of the theorem and n E RI is even, then the above error es-
timate can be expressed as
rb M(b - a)5 I
1, f (x) dx - S"(f) I `- (26)
180 n4
Proof. The proof of the theorem proceeds in an analogous manner to the proofs of
the previous two results. We first prove that for i = 1.... , ;,
I tc, f(x) dx - 3 [ f(c. - h) + 4 f(c.) + f(c. + h)] - _hs, (27)

ti
where c; = a + (2i - 1)h. To accomplish this, define g, on [0. h] by
gr(t) dx- 3 [f(c+ - t) + 4f(cr) +f(cr + t)).
To prove inequality (27), we are required to show that Ig,(h)I <- Mh5/90. Upon com-
puting the successive derivatives of g, we obtain
gi(t)=3[f'(c;-t)-f'(c,+t)]+3[f(ci-t)+f(cr+t)]-4f(ci).
g7(t)=3[ f"(c,--f"(c;+t)]+3[ t)+f'(c;+t)],
and
gr' (t) = 3 [f"'(C, - t) - f"'(c; + r)].
By the mean value theorem, f"(c, - t) t) = for some in

(c; - t, c; + t). Therefore.
I8,""(t) l s Mt g.
3
As in the previous two theorems, since g;(0) = g;(0) = g"(0) = 0, upon three integra-
tion we have Ig;(h)l s Mh590. This proves inequality (27). Finally,
n/2 nr_
1a f(x)dx - S"(f)I = Igi(x)I lgr(x)l
Mh5 n M(b - a)
h
90(2) 180
.6.6.6 EXAMPLE In this example we will use the trapezoidal rule and Simpson's rule with
n = 8 to obtain approximations to In 5. For f(x) = x-', f"(x) = 2x'3, and f(4kx) _
24x-'. Therefore,
sup{If"(x)I: x E [1, 5]} = 2 and sup{Ift4I(x)I: x E [1, 5]) = 24.
By Theorem 6.6.4 with h = Z. the error E8 in the trapezoidal approximation satisfies
6
IEa(f)I 212(-21 Y = = 0.16666
On the other hand, the error Es in using Simpson's rule is guaranteed to satisfy
a
IEsl)1 0.03333
2180 2 30 =
which is considerably better.
With x, = 1 + ih and y; = f(x;), i = 0, 1, .. . , 8,
2 1 2 1 2 1 2 1
Yo=1,Yi=3, Y6=4,Y7=9,Ye5
Therefore,
Ts(f)=4 Yo+2ly,+Ys
i=1 I
4+3+1+5+3+7+2+9+5
Since In 5 = 1.6094 to four decimal places, the error is less than 0.02, well within the
tolerance predicted by the theory. With Simpson's rule,
1
Ss(f)= 6[rrYo+4Yi +2Yz+4Y3+2Ya+4Ys+2Ye+4Y7 +Ya]
=I[I+3+1+8+3+g+2+9+I
With Simpson's rule, the actual error is less than 0.0014.
We can also use the results of the theory to determine how large n must be chosen
to guarantee predetermined accuracy. For the trapezoidal rule, since M = 2 and
(b - a) = 4, by inequality (23)
3
IEn(f)It212 n2'
Thus to obtain accuracy to within 0.001, we need
3
n2 > 212 10;,
which is accomplished with n ? 104. On the other hand, using Simpson's rule,
24.45 1
1< n4.
180
Thus to obtain accuracy to within 0.001, we need n E Rl even with
24.4'
nS 103.
> 180
This will be satisfied with n ? 20.
EXERCISES 6.6
1. a. Use the midpoint rule, trapezoidal rule, and Simpson's rule to approximate In 2 = f, 2(1/x) dx with n = 4. For each
method, determine the estimated error and compare your answer to In 2 = 0.69315 (to five decimal places).
b. Repeat part (a) with n = 8.
c. For each of the three methods (midpoint rule, trapezoidal rule, Simpson's rule) determine how large a must be
chosen to assure accuracy in the approximation of In 2 to within 0.0001.
2. a. Using the fact that
' dt it
f01+t 4
use the midpoint rule, trapezoidal rule, and Simpson's rule to approximate ,r/4 with n = 4. For each method,
determine the estimated error.
b. Repeat part (a) with n = 8.
c. For each of the three methods determine how large n must be chosen to obtain an approximation of n accurate
to within 0.0001.
3. Use Simpson's rule to obtain approximations of each of the following integrals accurate to at least four decimal
places.
2 i
a. dx "b. 10 1 + 2 dx c. to sin(x2) dx
1a 1 + aP
4. Determine how large a must be chosen so that the trapezoidal approximation T. approximates 5 a-'2dx with an er-
ror less than 10-6.
5. Fill in the details of the proof of Theorem 6.6.4.
6. Let f be a differentiable function on [a, b] with f(x) bounded on [a, b]. Let n E N. Prove that
e M(b - a)
I J f(x) dx - h f(a + ih) I S 2 h,
where h = (b - a)/n and M = sup{j f'(x)h : x E [a, b)).

7. a. Show that Ta,(f) = T"(f )).
i
b. Show that S5,(f) _ JM"(f) + 3
8. Prove the following variation of Theorem 6.6.4: If f is twice differentiable on [a. b] and f" is continuous on [a. b].
then there exists a point c E [a, b] such that
b
(b - a)h2
f(x) dx = 12 f"(c).
272 Chapter 6 The Rlemann and Riemann-Stieltjes integral
9. a. Use the previous exercise to show that ,O,Z < TA - In 2 < 384
b. Using part (a) and Exercise 1(b) show that .6915 < In 2 < .6938.
Proof of Lebesgue's Theorem

In this final section we present a self-contained proof of Lebesgue's characterization of
the Riemann integrable functions on [a, b]. Recall that a subset E of R has measure
zero if given any e > 0, there exists a finite or countable collection {In} of open inter-
vals with E C U. I, and En l(ln) < oo, where I (In) denotes the length of the interval
I. We begin with several preparatory lemmas.
6.7.1 LEMMA A finite or countable union of sets of measure zero has measure zero.
Proof. We will prove the lemma for the case of a countable union of sets of measure
zero. The result for a finite union is an immediate consequence. Suppose {En}FEN is a
countable collection of sets that measure zero. Set E = U E and let e > 0 be given.
Since each set E. is a set of measure zero, for each n E N there exists a finite or count-
able collection {1n.k}k of open intervals such that E.C and kl(I,,.k) < e/2".
Since we can always take In, k to be the empty set, there is no loss of generality in as-
suming that the collection {I , k}k is countable. Then {In, k}n. k is again a countable col-
lection of open intervals with E C U,,, k In, k.
Since N X N is countable, there exists a one-to-one function f of N onto N X N.
For each m E N, set J. = Then {J,n}T=, is a countable collection of open
intervals with EC Un, J,n. Since f is one-to-one, for each N E N, the set
F N = f({ 1, ... , N)) is a finite subset of N X N. Hence there exist positive integers N,
and KI such that for all (n, k) E FN we have I S n < N, and 1 <_ k <_ K1. Hence
N
11 (J.) = I(In.k) c Y, I(In.k)

m=1 (n.k)EP,, Isn<N,
IsksK,
But
N,++ K, N, x N,
t nsN,
I (In, k) = G I I (In. k) c 7.1 I (In. k) < Y
n=1k=1 n=Ik=I n-12
n< e.
15ksK,
Thus l (J,n) < e. Therefore E has measure zero. 0

6.7.2 LEMMA Suppose f is a nonnegative Riemann integrable function on [a, b] with
f, ,"'f = 0. Then {x E [a, b] : f (x) > 0} has measure zero.
Proof. We first prove that for each c > 0, the set EE _ {x E [a, b]: f(x) ? c} has
measure zero. Let e > 0 be given. Since fo f = 0, there exists a partition 9 =
{x0, x1,.., x,,} of [a, b) with °lt(9, f) < ce, where
and Mi = sup{ f (x) :x E [xi _ 1, xi] }. Let I= {i : E,0 (xi _ 1, x, ] # ¢}. If i E 1, then
Mi at c. Hence
CE > V(P,f) 2t I MiAxi CY1AX,
idl tEl
Thus J:1Ei1([x,_,, xi]) _ 7,iEi Axi < e. Finally, since EC Uiel [xi_ 1, xi], it follows
that Ec has measure zero.
To conclude the proof we note that {x E [a. b] : f(x) > 0} = U' 1 E. where for
each n E (\J,
En={xe[a.b]:f(x)!}.
By the above, each E. has measure zero. The conclusion now follows by Lemma 6.7. 1.
J
6.7.3 DEFINITION If E is a subset of IL the characteristic function of E, denoted Xe, is
the function defined by
XE(x) _ 1, x E E,
0, x 44 E.
As in Section 6.2, if 9 = {xo, xl, ... , xn} is a partition of [a, b], the norm 119111 of the
partition 9 is defined by 119I1 = max Axi. If f is a bounded real-valued function on [a, b],
15i5n
the infimum and supremum off on [xi_ 1, xi] are denoted by mi and Mi respectively; i.e..
m; = inf{f(x) : x E [xi_ 1.xi]} and Mi = sup{f(x) : x E [xi_ . xi]}.

The lower function L f and the upper function Uf for f and the partition P are defined
respectively by
Lf(x) and
i
i=1
Uf (x) = Mi X(x, _,. x,)(x)-
The graphs of the lower function L f and upper function Uf off are depicted in Figure
6.12. Since mi s f(x) c Mi for all x E [xi_ 1, xi),
L1(x) s f(x) s Uf(x) for all x E [a, b).
Also, since L f and Uf are continuous except at a finite number of points, they are Rie-
mann integrable on [a, b] (Exercise 1) with
J Lf=T(9P,f) and
Ja Uf=a13,(2,f)
a
Figure 6.12 Graphs of the Lower and Upper Functions off
THEOREM 6.1.13 (Lebesgue) Abounded real-valued function f on [a, b] is Rie-

mann integrable if and only if the set of discontinuities off has measure zero.
Proof.
1b1
Assume first that f is Riemann integrable on [a, b]. Then for each n E N,
there exists a partition 91, of [a, b] with
0:S
a
1/n, such that
and 0 5 O) (91,,,f) - f- f<n (28)
Since 2(91.,f) t 2(91,f) and V(91, f) s 61L(91, f) for any refinement 91 of a,,, the
partitions 9P,,, n = 1, 2, .. . , can be chosen so that SPn+I is a refinement of 91..
For each n E N, let L,, and U respectively be the lower and upper functions off
corresponding to the partition P,,. Then L,(x) s f(x) <_ U,(x) for all x E [a, b) and
1a
bLn = 49., f) and
f bUn = (29)
a
Since 91n+ l is a refinement of p,,, the functions L. and U. satisfy La(x) _< L,,, i(x) and
Un+ 1(x) s U,(x) for all x E [a. b]. Define L and U on [a, b] by
L(x) = lim L,(x) and U(x) = lim U,(x).
Then
L. (x) 5 L(x) 5 f(x) 5 U(x) s U,(x)
for all x E [a, b). Hence

rbUC bUC
16Ln bL: Jbf JbUn
a
JbL<
a 1a a Ja a a
But by equations (28) and (29).

!, h h
lim J
lim
L = Jim
n-.x
f I f.
u n a
Hence L and U are Riemann integrable on [a, b] with

b
I L = JbU= fbf
Since U(x) - L(x) ? 0 for all x E [a, b], by Lemma 6.7.2, {x E [a, b] : U(x) # L(x)}
has measure zero. Furthermore, since each 91 has measure zero, by Lemma 6.7.1, the set
E = {x E [a, b] : U(x) # L(x)} U (U 9'n)
also has measure zero. We conclude by showing that f is continuous on [a, b]\E.
Fix x, E [a, b]\ E, and let e > 0 be given. Since L(x,,) = U(x,), there exists an in-
teger k E N such that Uk (x,) - Lk(x,) < e. Also, since xa E 9'k, the functions Uk and
Lk are constant in a neighborhood of x0. Hence there exists a S > 0 such that
0 < Uk (x,,) - Lk (x) = Lk (x) - Lk (xo) < E
for all x E [a, b] with Ix - S. Finally, since Lk(x) <- f(x) Uk(x) for all
X E [a, b),
- E < Lk (x) - Uk (xo) C f (x) - f (x,,) < Uk (x) - Lk (xo) < E
for all x with Ix - x0I < S. Therefore f is continuous at x,,.
Conversely, suppose f is continuous on [a, b]\E, where E is a set of measure zero
and a < b. Let M > 0 be such that I f(x)I M for all x E [a, b], and let e > 0 be
given. Since the set E has measure zero, there exists a finite or countable collection
of open intervals such that E C U I and E 1(1,,) < E/4M. Also, since f is continuous
on [a, b]\E, for each x E [a, b] \ E there exists an open interval J, such that
1f (z) - f (y) I < E/2(b - a) for all y, z e J, fl [a, b]. The collection U {J, :
x E [a, b]\ E} is an open cover of [a, b]. Thus by compactness, a finite number, say
{Ik}x=, and {JX}; , also cover [a, b]. Let
9' = {a = to, t,, .. . , tN = b}
be the partition of [a, b] determined by those endpoints of Ik, k = 1. ... , n, and
J,,, j = 1, ... , m, that are in [a, b]. For each j, 1 s j c N, the interval (t,_,, t,) is
contained in some Ik or J,,. Let J = {j: (t, _ 1, t,) C Ik for some k}. Also, for each
j E {1, ...,NJ, let m, and M; denote the infimum and supremum off on [t,_,. t;] re-
spectively.Then M, - m, <- 2M for all j E J, and M, - m, < E/2(b - a) for all j 44 J.
Thus
R(9P,f)-X(P,.f)= I(Mi-mi)Oti+ I(M!-mi)At,

1CJ
52M:01 + E
2(b - a)1EJ
c1t!
,EJ f
N
E
:5 2M7, 1 (1k)+2(b-a 4r;
E E
52M\4M)+2(b-a)(b-a)=E.
Hence by Theorem 6.1.7 the function f is Riemann integrable on [a, b]. 1]
W EXERCISES 6.7
1. Let f be a bounded real-valued function on [a, b] and 91 a partition of [a, b]. Prove that the lower function Lt and
upper function Uf off for the partition ' are Riemann integrable on [a, b] with
rb
jb Lf = Y(9,f) and = W495,f)
0 1a
2. Let P be the Cantor set in [0, 1 ]. Show that Xr E %[(), 1) and find f.' kAX)dx.
3. Let f E R[a, b], and for x E [a, b], set F(x) = fo f(t) dt. Prove that there exists a subset E of [a, b] of measure
zero such that F'(x) = f(x) for all x E [a, b] \ E.
4. *Suppose f E %[a, b) and g is a bounded real-valued function on [a. b). If {x E [a, b) : g(x) # f(x)} has mea-
sure zero, is g Riemann integrable on [a, b)?
NOTES
The fundamental theorem of calculus is without question the Lebesgue's theorem is one of the most beautiful
key theorem of calculus; it relates the Cauchy-Riemann the- results of analysis. It provides very concise necessary
ory of integration to differentiation. For Newton, Leibniz, and sufficient conditions for Riemann integrability of a
and their successors, integration was the inverse operation of bounded real-valued function f. Although Riemann was
differentiation. When Cauchy, however, defined integration the first to provide such conditions, his result lacked the
independent of differentiation, the fundamental theorem of simplicity and elegance of Lebesgue's theorem. Unfortu-
calculus became a necessity. It was crucial in proving that nately for Riemann, the concept of measure had not yet
Cauchy's integral was the inverse of differentiation, thereby been developed when he stated and proved his result. In
providing both a convenient tool for the evaluation of defi- Chapter 10 we will develop the theory of measure and
nite integrals and proving that every continuous real-valued Lebesgue's extension of the Riemann integral.
function defined on an interval I has an antiderivative on /. The need for numerical methods for the evaluation of
Although we stated the fundamental theorem of calcu- definite integrals was recognized as early as the eighteenth
lus as two separate theorems, for continuous functions they century. Euler and Thomas Simpson (1710-1761), among
are the same. Specifically, if f is a continuous real-valued others, used numerical techniques to approximate the defi-
function on [a, b] and F is an antiderivative off, then for nite integral in problems where an antiderivative could not
any x E [a, b], be found. Even the error estimates developed in Section 6.6
date back to that era. With the availability of efficient cal-
F(x) = Fa) + 11f(t) dt. culators and high-speed computers, numerical methods
have increased in importance in the past few decades. This
Conversely, if f is continuous and F is defined as above. has led to the development of very sophisticated numerical
then F(x) = f(x). algorithms for the evaluation of definite integrals.
Although Newton and Leibniz are credited with differential calculus for finding tangents to a curve in
inventing the differential and integral calculus, many his 1670 treatise Lectiones geometriae. However, all these
mathematicians prior to their time knew formulas for early methods were developed using geometric arguments.
computing tangents and areas in particular instances. Newton and Leibniz developed the concepts, the notation.
Archimedes (around 200 a.c.), in his treatise Quadrature and the algorithms for making these computations for ar-
of the Parabola, used the method of exhaustion by in- bitrary functions. Most importantly, however, both men
scribed triangles to derive a formula for the area under a realized the inverse nature of the problems of tangents
parabolic segment. By the mid-1640s, Pierre de Fermat and areas. For these reasons they are credited with the de-
(1601-1665) had determined the formulas for the area velopment of the differential and integral calculus. Fur-
under any curve of the form y = x' (k * -1), and for ther information on the historical development of calculus
finding tangents to such curves. Isaac Barrow (1630- may be found in the text by Katz listed in the Bibliogra-
1677), a professor of gegmetry at Cambridge, developed phy and the article by Rosenthal listed at the end of the
an algebraic procedure that is virtually identical to the chapter.
A real-valued function f on [a, b] is a step function if there exist a finite number of disjoint intervals
(Ii}n- 1 with U Il = [a, b) such that f is constant on each of the intervals lt, j = 1, ... , n.
1. a. If f is a step function on [a, b), prove that f E 9t[a, b) with
b a
- f=7,
'-1
where ci is the value off on I;.

b , Iff E 9t[a, b] and e > 0 is given, prove that there exists a step function h on [a, b] such that
Ifaf - fahI< e.
c. If h is a step function on [a, b] and e > 0 is given, prove that there exists a continuous function g on [a, b]
such thatIfah - f°gI <e.
d. If f E 9t[a, b] and e > 0 is given, prove that there exists a continuous function g on [a, b] such that
I fa f - fa gl <e.
Let f be a real-valued function defined on (0, oo). The laplace transform off, denoted T(f ), is the function defined by
1(f)(s) =1 e "f(t) dt,whenever the improper Riemann integral exists.

0
2. Let f be defined on (0, oo). Prove that there exists a e R U {-oo, oo) such that f(f)(s) is defined for all
s e (a, oo), and that the integral defining Z(f)(s) diverges for all s E (-oo, a).
(Hint: First show that if `d;(f)(sa) exists for some sa, then T(f)(s) exists for all s > se.)
3. Suppose f E 9t[0, c] for every c > 0, and that there exists a positive constant C, and a E R. such that
I f(r) 1 s Ce- for all t ? 0. Prove that T(f)(s) exists for all s > a.
4. Compute the Laplace transform of each of the following functions. In each case, specify the interval on which
àf(f)(s) is defined
a. At) = 1
b. f (r)
c. f(t) = cos wt
d. f(t) = sin tot
e. f(t) = t", n E N (Use induction.)
I. f(t) = I (t - c), where J(t - c) is the unit jump function at t = c
g. f(r) = t°, a > - I (See Exercise 9, Section 6.4.)
5. Suppose f is differentiable on [0, oo) and a E R is such that `.f(f)(s) exists for all s > a. If lim e-"f(t) = 0 for all
s > a, prove that Y(f')(s) = s .(f)(s) - f(0).
® SUPPLEMENTAL READING
Bagby, R. J., "A convergence of limits," Math. Mag. 71 Jones, W. R. and Landau, M. D., "One-sided limits and
(1998),270-277. integrability," Math. Mag. 45 (1972), 19-21.
Bao-lin, Z., "A note on the mean-value theorem for inte- Klippert. J.. "On the right-hand derivative of a certain
grals." Amer. Math. Monthly 104 (1997), 561-562. integral function;' Amer. Math. Monthly 98 (1991).
Bartle, R. G., "Return to the Riemann integral;' Amer. 751-752.
Math. Monthly 103 (1996), 625-632. Kristensen, E., Poulsen, E. T. and Reich, E., "A charac-
Botsko, M. W., "An elementary proof that a bounded a.e. terization of Riemann integrabiliry"Amer. Math.
continuous function is Riemann integrable," Amer. Monthly 69 (1962), 498-505.
Math. Monthly 95 (1988), 249-252. Quadling, Douglas, "a in my eye," Math. Gaz. 77
Bullock. G. L., "A geometric interpretation of the Rie- (1993),314-321.
mann-Stieltjes integral:' Amer. Math. Monthly 95 Rickey, V. F. and Tuchinsky, P. M., "An application of
(1988), 448-455. geography to mathematics: History of the integral
Goel, S. K. and Rodriguez, D. M., "A note on evaluating of the secant," Math. Mag. 53 (1980), 162-166.
limits using Riemann sums:' Math. Mag. 60 Rosenthal, Arthur, "The history of calculus:' Amer.
(1987), 225-228. Math. Monthly 58 (1951), 75-86.
Grabinger, J. V., "Was Newton's calculus a dead end? Ross, K., "Another approach to Riemann-Stieltjes inte-
The Continental influence of Maclaurin's Treatise grals:' Amer. Math. Monthly 87 (1980).660-662.
of Fluxions," Amer. Math. Monthly 104 (1997), Stein, S. K., "The error of the trapezoidal method for a
393-410. concave curve:' Amer. Math. Monthly 83 (1976)
Hartman, F. and Sprows, D., "Oscillating sawtooth func- 643-645.
tions:' Math. Mag. 68 (1995), 211-213. Williams, K. S., "Note on 5 (sin xlx)dx," Math. Mag.
Jacobson, B., "On the mean-value theorem for inte- 44 (1971),9-11.
grals:' Amer. Math. Monthly 89 (1982), 300-301.
71 Series of Real Numbers
7.1 Convergence Tests
7.2 The Dirichlet Test
7.3 Absolute and Conditional Convergence
7.4 Square Summable Sequences
Although the study of series has a long history in mathematics,' the modem definition
of convergence dates back only to the beginning of the nineteenth century. In 1821,
Cauchy, in his text Cours d'Analyse, used his definition of limit to provide the first for-
mal definition of convergence of a series in terms of convergence of the sequence of
partial sums. The Cauchy criterion (Theorem 2.7.3) was the first significant result to
provide necessary and sufficient conditions for convergence of a series. Cauchy not
only stated and proved the result, he also applied his result to prove convergence and
divergence of given series. Many of the early convergence tests, such as the root and ra-
tio test, are due to him. Cauchy, with his formal development of series, placed the sub-
ject matter on a rigorous mathematical foundation.
In this chapter we will continue our study of series of real numbers. Our primary
emphasis in Section 7.1 will be on deriving several tests that are useful in determining
the convergence or divergence of a given series. In Section 7.3, we will study the con-
cepts of absolute convergence, conditional convergence, and rearrangements of series.
One of the key results of this section is that every rearrangement of an absolutely con-
vergent series not only converges, but converges to the same sum. As we will also see,
this fails dramatically if the series converges but fails to converge absolutely.
In Section 7.4, we give a brief introduction to the topic of square summable se-
quences. These play an important role in the study of Fourier series. One of the main
results of this section will be the Cauchy-Schwarz inequality for series. The section also
contains a brief introduction to normed linear spaces.
1. See the notes at the end of the chapter.
279
280 Chapter 7 Series of Real Numbers
Convergence Tests
In Section 2.7 we provided a very brief introduction to the subject of infinite series. In
the study of infinite series it is very useful to have tests available to help determine
whether a given series converges or diverges. For example, Corollary 2.7.5 is very use-
ful in determining divergence of a series. If the sequence {ak} does not converge to zero,
then the series 7- ak diverges. On the other hand, if lim ak = 0, then nothing can be as-
certained concerning convergence or divergence of the series I ak. In this section we
will state and prove several useful results that can be used to establish convergence or
divergence of a given series. Additional tests for convergence will be given in the exer-
cises and subsequent sections. With the exception of Theorem 7.1.1, all of our results
in this section will be stated for series of nonnegative terms.
As in Definition 2.7.1, given an infinite series Ja I ak of real numbers, I will
denote the associated sequence of partial sums defined by
n
sn _ Y, ak.
k=1
The series 7, ak converges if and only if the sequence of nth partial sums con-
verges in R. Furthermore, iffl-00
lim s = s, then s is called the sum of the series, and we
write
M
I ak = S.
k=1
If the sequence diverges, then the series I ak is said to diverge. Furthermore, if

lim s = oo (or -oo), then we write 7, ak = oo (-oo) to denote that the series di-
verges to 00 (or - oo).
If ak at 0 for all k, then by Theorem 2.7.6 the series 7, ak converges if and only if
lim s < oo. Thus for series of nonnegative terms we adopt the notation
cc
I ak < 00
k=1
to denote that the series converges.
Remarks
(a) Although we generally index a series by the positive integers N, it is sometimes
more convenient to start with k = 0 or with k = ka for some integer ko. In this case, the
resulting series are denoted as
I ak,
k=0
T, ak
k=k
Also, from the Cauchy criterion (Theorem 2.7.3) it is clear that I ak converges if
and only if 7, a k., ak converges for some, and hence every, k,, E N.
(b) Given any sequence {sn}.'= 1 of real numbers, we can always find a series I ak
whose nth partial sum is sn. If we set a1 = s1 and ak = sk - sk_ 1, k > 1, then
n
T. ak = Si + (SZ - S1) + . .
+ (Sn - S;_,) = Sn.
k=1
7.1.1 THEOREM If Yoko= , ak = a and 7.0ko- I bk = A, then
a
(a) Y, cak = ca, for any c E R, and
k-1
DO
(b) F; (ak+bk)=a+$3.
k-1
Proof. The proof of (a) is similar to (b) and thus is omitted. To prove (b), for each
n E N, let
n n
Sn = ak and to = bk.
k-1 k-1
Since the series converge to a and $3 respectively, tim s = a and lint t = $3. Therefore,
by Theorem 2.2.1, lim(sn + tn) = a + $3. But
n n
S. + to = ak + T, bk = 7, (ak + bk).
k=1 k-1 k-1
Therefore, sn + to is the nth partial sum of the - series Z(ak + bk). Since the sequence
{sn + t,,} converges to a + (3,
OD
I (ak+bk)=a+A
k-1
Comparison Test
One of the most important convergence tests is the comparison test. Though very ele-
mentary, it provides one of the most useful tools in determining convergence or diver-
gence of a series. It is useful both in applications and theory. Several of the proofs of
subsequent theorems rely on this test. In applications, by comparing the terms of a
given series with the terms of a series for which convergence or divergence is known,
we are then able to determine whether the given series converges or diverges.
7.1.2 THEOREM (Comparison Test) Suppose 7, ak and 7, bk are two given series of
nonnegative real numbers satisfying
0 < at Mbk
for some positive constant M and all integers k a kn, for some fixed k E N.
00
(a) If bk < oo, then Y, ak < oo
k=1 k-1
W x
(b)IfY, at = oo, then 7, bk = oo.
k=1 k-1
Proof. Suppose the terms {ak} and {bk} satisfy at 5 Mbk for all k > ko and some pos-
itive constant M. Then for n > m ? ko,
n n
05 1 akSM 7, bk.
k=m+I k=m+I
Suppose I bk converges. Then given e > 0, by the Cauchy criterion (2.7.3) there ex-
ists an integer n,, ? k such that
km+1 bk < M
for all n > m ? n,. Thus 0 <_ L+k=m+I at < e for all n > mat n0. Hence by the
Cauchy criterion E at converges. On the other hand, if I at diverges, then 1 bk must
also diverge.
As a corollary of the previous theorem we also have the following version of the
comparison test.
7.1.3 COROLLARY (Limit Comparison Test) Suppose E at and E bk are two given series
of positive real numbers.
(a) If liimW bn = L with 0 < L < oo, then 7. at converges if and only if E bk
converges.
(b) If limb = 0 and B bk converges, then I at converges.

-oo
Proof. The proof, the details of which are left to the exercises (Exercise 6), follows
immediately from the definition of the limit and the comparison test.
Remark. If Jim an/bn = 0 and I at converges, then nothing can be concluded about
the convergence of the series 7, bk. In 7.1.4(d) we provide an example of a divergent
series 7, bk and a convergent series E at for which lim 0. On the other hand,
n-+oo
in Exercise 23, given a convergent series E ak with ak > 0, you will be asked to con-
struct a convergent series I bk with bk > 0 such that lim a"Jb, = 0.
7.1.4 EXAMPLES
(a) As an application of the comparison test, consider the series
°O k
ksi 3
We will compare the given series with the convergent geometric series 7, Thus we
wish to show that there exists ko E Fk1 such that k/3k s 1/2k for all k a k0. By Theorem
2.2.6(d),
k
k k 3 = 0.
Thus by taking e = 1, there exists an integer ko such that k(3)k <_ 1 for all k ? k,. As a
consequence,
for all k ? k0.

3k s 2k
Since the geometric series E('-i)k converges (Example 2.7.2(a)), by the comparison test,
the given series also converges. A similar argument can be used to prove that I k"/ak
converges for any n E 7L and a E R with a > 1 (Exercise 2(1)). This is accomplished
by comparing the given series with 7, b-k, where 1 I (Exercise 11).
(b) As our second example, consider the series
°° +1
zk3+ 1
To use the comparison test, we first have to determine what series we want to compare
with. Since
k+ 1 _ 1 1 +k
and kltm.
1 k-1j,
2k +1 k 2+k 2+k 2
we will compare the given series with the series Y, 1/k. This series is known to diverge
(Example 2.7.4). If we take e = 4%52, then we can conclude that there exists ko E N
such that
1+k 11vF2
2+k:2- e
4
for all k at k0. Thus there exists a positive constant M and k, E N such that
k+1 I
2k3+1aMk
for all k ? ko. Since the series E 1/k diverges, by the comparison test the given series
also diverges.
(c) The divergence of the series
co
k+ 1
1 2k'+l
can also be obtained by the limit comparison test. Comparing the terms of the given se-
ries to the terms of the series Y, 1/k we have
?kT __ 1+k
=
lint
-"o 1
k
k-r+moo 2+k -2
Thus since the series 11/k diverges, by the limit comparison test the series
°O k+1
I 'V 2k' + 1
also diverges.
2_k
(d) Let ak = and bk = 1/k. Then 7, ak converges, 7, bk diverges, and
'- b,, aoo 2" = 0.
Thus if rim " = 0, convergence of the series 7, ak does not imply convergence of the
b
series bk.
Integral Test
Our second major convergence test is the integral test. Recall from Section 6.4, if f is a
real-valued function defined on [a, oo) with f E %[a, c] for every c > a, then the im-
proper Riemann integral off is defined by
dx = lint jf(x) dx,

u a
provided the limit exists. If f(x) at 0, we use the notation

00
f(x) dx < oo f f (x) dx = oo
a \ o 00
to denote that the improper integral off converges (diverges).
7.1.5 THEOREM (Integral Test) Let {ak} I be a decreasing sequence of nonnega-

tive real numbers, and let f be a nonnegative monotone decreasing function on( 1, oo)
satisfying f (k) = akfor all k E N. Then
F, ak < 00 if and only if Jf(x) dx < coo.

k-t
Proof. Since f is monotone on [ 1, oo), by Theorem 6.1.8 it is Riemann integrable on

[ 1, c] for everyc > 1. Let n E (\i, n ? 2, and consider the partition `2P = { 1, 2, .... n }
of [ 1 . n]. Since f is decreasing, f o r each k = 2, 3, ... , n (see Figure 7.1),
sup{f(t) : t E [k - 1, k]} = f(k - 1) = ak_ I, and
inf{f(t) t E [k - 1, k]} = f(k) = at.
Figure 7.1 Integral test
Therefore,
ak = X(9j) .f(x)dx C ak
k-2 I k-1
from which the result follows.
7.1.6 EXAMPLES
(a) As our first application of the integral test, we consider the p-series
When p = 1 this series is referred to as the harmonic series. If p 0, then {k-°} does
not converge to zero, and thus by Corollary 2.7.5, the series diverges. Suppose
p > 0, p 1. Let f(x) = xo, which is decreasing on [ 1, oo). Then
`
Jx-Pdr=---___[i
1
p-1 --Jc
1
1
Therefore,
rco 1 p > 1,
x-odx=lim x-°dx= p- 1
JI `~00
JI 00, p < 1.
Thus by the integral test, the series diverges for p 1. When
p = 1, then by Example 6.3.5,
f = Inc.
i
`Xdx
Since lim
X-W
In x lim+
1-.0
In t = oo (Example 6.4.2(a)), by the integral test the series
also diverges for p = I. Summarizing our results, we find
°O+ f converges, if p > 1,
G kr diverges, if p 1.
(b) As our second application of the integral test, we consider the series
I
00 1
k..,- kInk*
Let f(x) = (x In x)'', x E [2, oo). Since

(1+lnx)
f'(x) (x in x)2
f'(x) < 0 for all x > 1. Thus f is monotone decreasing on [2, oo). But
lim J c I dx = lim ln(ln c) - ln(ln 2) = oo.

C- 2 x In x
Thus by the integral test the given series diverges.

(c) As our final example, we consider the series
'° Ink
7 kp. pEIR.
We first consider the case p = 1. Since f(x) = (In x)/x is decreasing on [e, oo) and
Ìnx
lira - dx = lim0100
(in c)2 - 1 = oo,
C X
by the integral test the series I'k*-2 (In k)/k diverges.

Suppose now that p > 1. Write p = q + r, where q > 1 and r > 0. By I'Hospi-
tal's rule M On x)/x' = 0. Thus there exists k E N such that (In k)/k' < 1 for all
k ? ko. As a consequence,
Ink _ 1 Ink 1
kr kq k' kq
for all k ? k0. Since q > I the series I 1/kq converges. Hence by the comparison test,
the series I (In k)/kP also converges.
Finally, if p < 1, then again by the comparison test, the divergence of I 1/k' im-
plies that 7, (In k)/kP also diverges. This follows from the fact that (In k)/k" ? (In 2)/k"
for all k E N, k ? 2. Summarizing our results we find
ink f converges, if p > 1,
k__Z kP diverges, if p 5 1.
Ratio and Root Tests

We now consider the well-known ratio and root tests. Although useful in determining
the convergence or divergence of certain types of series, both of these tests are also very
important in the study of power series. The ratio test, attributed to Jean d'Alembert
(1717-1783), is particularly applicable to series involving factorials. The root test, at-
tributed to Cauchy, will be used in the next chapter to define the interval of convergence
of a power series. Because of the close similarity between these two tests we state them
together.
Prior to stating the ratio and root tests we recall the definition of the limit inferior
(lim) and limit superior ( lim) of a sequence of real numbers (Definition 2.5.1). If {sn}
is a sequence in R, then
lim s = sup inf{s : n ? k} = lim inf{s : n a k}, and

k6N k-+0o
lim s, kE
sup{s : n k} = slim sup(s : n !- k).
By Theorem 2.5.7, if E denotes the set of subsequential limits of {sn} in OI U {-oo, oo},
then
!Lm s = inf E and lim s,, = sup E.

n _CG n
In particular, if the sequence {sn} converges (or diverges to either - oo or oo), then
Urn S. = lira S. = lim S.
7.1.7 THEOREM (Ratio Test) Let I ak be a series of positive terms, and let
and r = lim
a"-+'.
R = lim ak'+
k-+oo ak k-+oo ak
00
(a) If R < 1, then I, ak < 00-
k-1
00
(b) If r > 1, then I ak = oo.

k-1
(c) If r 5 1 5 R, then the test is inconclusive.

7.1.8 THEOREM (Root Test) Let 7, ak be a series of nonnegative terms, and let
k
a = lim
k-0oo
ak
(a) If a < 1, then 7,

00 ak < oo.
k-I
00
(b) If a > 1, then ', ak = oo.

k=1
(c) If a = 1, then the test is inconclusive.
Before proving the theorems we give several examples that illustrate these two con-
vergence tests.
7.1.9 EXAMPLES
(a) Consider the p-series
°O 1
A kp, p E R.
k- I
limakk1=lim(1+k1 =
kroo a 1.
kyoo \\
Also, by Theorem 2.2.6,

r P
urn
oo k1 k-*w Ok) 1.
Therefore, for this series,

r=R=a=1
for any p E R; thus both tests are inconclusive. The series diverges, however, if p s 1,
and converges for p > 1.
(b) As our second example, we consider
pk
kl, 0<p<oo.
Here ak = pk/k!. Thus
p 4+I
+l
lim ak+1
k-+oo ak
= lim
k-boo
pk 1
(k + 1)! .
= p lim I
k-+= k + 1
= 0.
By the ratio test the series converges for all p, 0 < p < oo. As a consequence of Corol-
lary 2.7.5, we also obtain
k
limp =0
k-.oo k!
for all p E R. In this example, the presence of k! makes the root test difficult to use.
(c) Consider I a where
Zk, if n = 2k,
_
an l
3k. ifn=2k+1.
Here,
00
1+2+3+22+32+
-i
By computation,
2
a.+, (3) 1
a" 121, n=2k+l.

'1/3k
For the subsequence of {a,, with even n,
a2k+I lk
21
lim
k- oo a2k
= lim
k-,oo 3
= 0.
For the subsequence with odd n,
a2k+2 1 liM /3\k

lim 00.
k-+oo a2k+ t 2 k-G21
2 '
Thus {0, oo} is the set of subsequential limits of Consequently, r = 0 and
R = oo. Therefore, the ratio test is inconclusive. On the other hand,
1
n= 2k,
1 M )1/(2k+ t), n=2k+1.
By Theorem 2.2.6(b), urno (V)v(2k+ q = 1. Thus the sequence { } has two sub-
sequential limits; namely, 1/V and 1/V/3. Therefore, a = 1/N/2, and the series con-
verges by the root test.
Proof of Ratio Test. Suppose
lim
a= = R < 1.
n-.m a,,
Choose c such that R < c < 1. Then by Theorem 2.5.3(a), there exists a positive inte-
ger n, such that
an
±an1 s c for all n ? no.
In particular, ane, 1 s can . By induction on m, a,, ,,,, < C'"a,, Therefore, writing
n=no+m,mz0,
an <_ Mc" for all n at no,
where M = a,. /c"°. Thus since 0 < c < 1, by Example 2.7.2(a), the series E c" con-
verges. Therefore, E an also converges by the comparison test.
Suppose
lima"}=r> I.
nToo an
Again choose c so that r > c > 1. As above, there exists a positive integer n, such that
a,, z Mc" for some constant M and all n _> n,. But since c > 1,1 c" = oo, and thus by
the comparison test, B an = oo. Q
Proof of Root Test. Let
a = lim
n-oo
<a..
Suppose a < 1. Choose c so that a < c < 1. Again by Theorem 2.5.3, there exists a
positive integer n, such that
" an < c for all n a n,.
But then a,, s c" for all n ? n and E an < oo by the comparison test.
If a > 1, then > I for infinitely many n. Thus a" > 1 for infinitely many
n, and as a consequence, {a,} does not converge to zero. Hence by Corollary 2.7.5, the
series B a,, diverges. Q
Example 7.1.9(c) provides an example of a series for which the ratio test is incon-
clusive, but the root test worked. Thus it appears that the root test is a stronger test; a
fact which is confirmed by the following theorem.
7.1.10 THEOREM Let {an} be a sequence of positive numbers. Then
lim aa,
n-rm an
lim lim s lim a-`
n-+oo an
=oo
Remark As a consequence of the theorem, if a series E ak converges by virtue of the

ratio test, i.e., R < 1, then we also have a < 1. Similarly, if 7. ak diverges by virtue of
the ratio test, i.e., r > 1, then we also have a > 1. Thus if the ratio test proves con-
vergence or divergence of the series, so does the root test. The converse, however, is
false (as indicated above).
Proof. Let
F,ma-+i
R
n-+oo a"
If R = oo, then there is nothing to prove. Thus assume that R < oo, and let P > R be
arbitrary. Then there exists a positive integer n, such that
an
an
s0 for all n ? n,.
As in the proof of the ratio test, this gives an s MI" for all n ? n where M = a,.19"
Hence
"a"<_0u for all n=?n,.

Since lim = 1, we have
M-000
llm "a:s p.
n-,00
Since J3 > R was arbitrary, Tim s R, which proves the result. The inequality for
the limit inferior is proved similarly. Q
7.1.11 EXAMPLE If ak = k!, then lim(ak+I/ak) = oo. Thus as a consequence of the pre-
vious theorem,
lim 'N_! = oo. S

k-+oo
EXERCISES 7.1
1. If a and b are positive real numbers, prove that
W I
Al (ak + b)o
converges if p > I and diverges if p c 1.
2. Test each of the following series for convergence:
sA, 2
k' +I
k b.
M
kt,
_
k2+2k-1
01+
C. G2k
k2
00
k'e_k 3k
d. F, *e.
k1
k-1
(kl)2
*L
k+ 1 - f
k.
*9. (2k)!
J. (/ - 1)k k. 1. k_1 -j, a> 1, nEZ

k-1
*m.
k-1
sin
(1k)
p>0 p>0 0. ykln(l+k
k=1
3. For each of the following, determine all values of p E R for which the given geometric series converges, and find
the sum of the series.
*a 7, (sin 0 b. c. (l ± p) (p * 1)
3 k-11
4. Suppose ak a 0 for all k E N and I a,, < oo. For each of the following, either prove that the given series con-
verges, or provide an example for which the series diverges.
00 00 00
a,2, c.
4; 1 + ak k-1 k-I
M
*d. 7, kak e. V k ak f.
t-1 k-1
g. I a,,,, where (a j is a subsequence of

5. *Determine all values of p and q for which the following series converges:
I 1
k4(ln k)P'
(Hint: Consider the three cases q > 1, q = 1, q < 1.)
6. *Prove Corollary 7.1.3.
7. If I ak converges and B bk = oo, prove that E (ak + bk) = oo.
8. Suppose is a sequence in R with a > 0 for all n E N. For each k E 101 set
k
1
bk = - ± a,,.
k n=I
Prove that Jk , bk diverges.
9. Suppose that the series E ak converges and {n;} is a strictly increasing sequence of positive integers. Define the se-
quence {bk} as follows:
b,=a,+ +a ,
bk=a.,_,+,+...+a,,:
Prove that E bk converges and that $k , bk = k I ak. (This exercise proves that if the series E ak converges,
then any series obtained from Eat by inserting parentheses also converges to the same sum. The following exer-
cise shows that removing parentheses may lead to difficulties.)
10. Give an example of a series F. ak such that Ek , (a2k _, + a2k) converges, but E ak diverges.
11. Prove that
p(k)
k-1
converges for any polynomial p and a > 1.
12. *Suppose that the series E ak of positive real numbers converges by virtue of the root or ratio test. Show that the
series Yt , k"ak converges for all is E N.
13. *Show that the series
1 1
+23+32++...
1 I
converges, but that both the ratio and root tests are inconclusive.
14. Apply the root and ratio tests to the series I ak where
2k. when k is even,

at
2t+2' when k is odd.
15. Suppose ak 0 for all k E N. Prove that the series Ik , ak converges if and only if some subsequence {sj of the
sequence of partial sums converges.
16. *Cauchy Condensation Test: Suppose that a, z a2 z a3 a ? 0. Use the previous exercise to prove that
7, R , ak converges if and only if P 2kag converges.
17. Use the Cauchy condensation test to show that $4 , 1/n^ converges for all p > 1, and diverges for all p,
0<ps1.
18. Use the Cauchy condensation test to determine the convergence or divergence of each of the following series.
00+
00 1 O0 1
b.
*a. k_'
2 k Ink k_'2 k(In k)" (P > I) e' L k(ln kiln Ink)
19. *For k E N, let ck be defined by
Prove that {ck} is a monotone decreasing sequence of positive numbers that is bounded below. The limit c of the se-
quence is called Euler's constant. Show that c is approximately 0.577.
20. Raabe's Test: Let ak > 0 for all k r= N. Prove the following.
a. If ak+,/ak -- I - r/k for some r > 1 and all k a k k, E Id, then T. ak < oo. (Hint: Show that
(k - 1)ak - kak1 a (r - I)ak for all k ? k,.)
b. If ak+,/ak z I - 1/k for all k k k0 E N, then I ak = oo. (Hint: Show that {kak,1} is monotone increasing
fork z k,.)
21. *lf p, q > 0, show that the series
1)(p+2)... (p+k)
(p+
1)(q+2)...(q+k)
(q+
converges for q > p + I and diverges for q 5 p + 1.
22. For p > 0 consider the series

(2k
L I'+I2.4Ir+...+(1.2 4. 1) 1
+..
a. Show that the ratio test fails for this series.

b. Use Raabe's test to show that the series converges for p > 2, diverges for p < 2, and that the test is incon-
clusive when p = 2.
c. Prove that the series diverges for p = 2.
23. Let {an} be a sequence of positive real numbers.
*a. Suppose I ak converges. Construct a convergent series 71 bk with bk > 0 such that lim akbk = 0.
kroc
b. Suppose I ak diverges. Construct a divergent series E bk with bk > 0 such thatk m bb/ak = 0.
The Dirichlet Test

In this section we prove the Dirichlet convergence test, named after Peter Lejeune
Dirichlet (1805-1859), and apply it to both alternating series and trigonometric series.
The key to the Dirichlet test is the following summation by parts formula of Neils Abel
(1802-1829). This formula is the analogue for series of the integration by parts formula.
7.2.1 THEOREM (Abel Partial Summation Formula) Let {ak} and {bk} be sequences of
real numbers. Set
n
Ao=0 and A'= Iak
ifn?1.
k-I
Then if15p5q,
q 4-1
Y, akbk = E Akbk - bk+ 1) + Agbq - Ap-1b .
k=p k=p
Proof. Since ak = A k - Ak_ 1,

q q
I akbk = I (Ak - Ak-1) bk

k=p k=p
4 q-1
Akbk Akbk+1
k-p k=p-I
Akbk - bk+1) + Agbq - Ap-1b . O

k-p
As an application of the partial summation formula we prove the following theo-
rem of Dirichlet. Another application is given in Exercise 1.
7.2.2 THEOREM (Dirichlet Test) Suppose {ak} and {bk} are sequences of real numbers
satisfying the following:
(a) the partial sums A. _ Jk=, ak form a bounded sequence.
(b) b, a b2 a b3 a ? O, and
(c) k-occ
liimbk=0.
Then Jko' , akbk converges.
Proof. Since {A,} is a bounded sequence, we can choose M > 0 such that IA. 1 s M
for all n. Also, since b -+ 0, given e > 0, there exists a positive integer n, such that
b < e/2M for all n z n,. Thus, if n, s p s q, by the partial summation formula,
q- I
k.p akbkl =
L , Ak(bk - bk+1) + Aqbq - Ap_Ibp
k-p
1
qj IAkI(bk - bk+1) + I AqI bq + IAp-1lbp

k-p
Cq.P
-1
M7, (bk-bk+1)+bq+bp
2Mbp<E.
Hence by the Cauchy criterion (Theorem 2.7.3), Jk , akbk converges. 0
Alternating Series
Our first application of the Dirichlet test is to alternating series. An alternating series
is a series of the form Y, (-1)k bk or I(-1)k+ 1bk, with bk > 0 for all k
7.2.3 THEOREM (Alternating Series Test) If {bk} is a sequence of real numbers satis-
fying
(a) b1?b2z 0, and
(b) k bk = 0,
OD
then I (-1)k+lbkconverges.
k-1
Proof. Let ak = (-1)k+ 1. Then IA. I < 1 for all n, and the Dirichlet test applies. O
Remark The hypothesis (a) that {bk} is decreasing is required. If we only assume
that bk a 0 and Jim bk = 0, then the conclusion may be false (Exercise 2).
For an alternating series satisfying the hypothesis of Theorem 7.2.3, we can actu-
ally do better than just prove convergence. The following theorem provides an estimate
on the rate of convergence of the partial sums to the sum of the series.
296 Chapter? Series of Real Numbers
7.2.4 THEOREM Consider the series 7,(-1}k+Ibk, where the sequence {bk} satisfies the
hypothesis of Theorem 7.2.3. Let
00
Sn = (-1)k+Ibk and s = 7, (-1)k+1bk
k=I k=1
T h e n I s - s.1 < b.+ 1 for all n E N.
Proof. Consider the sequence {stn}. Since

2n
sz, = 7, (_ 1)k+Ibk = (b1 - b2) + ... + (b,,_ I -
k-1
and (bk_ 1 - bk) >_ 0 for all k, the sequence is monotone increasing. Similarly
{s2i+1} is monotone decreasing. Since {sq} converges to s, so do the subsequences
{s2,,} and {s2,+I}. Therefore, s2, s s <- s2,, +I for all n E N. As a consequence,
Is-SkI:5 ISk+I-SkI=bk+I for all kEN. E)
7.2.5 EXAMPLE Since the sequence fl/(2k - l)} decreases to zero, by Theorem 7.2.3, the
series
00 (-1)k+I
k- I 2k - l
converges. As we will see in Chapter 9 (Example 9.5.7), this series converges to 7r/4.
Thus by Theorem 7.2.4, if s is the nth partial sum of the series,
4 - s" 12n + 1
for all n E N. Although this can be used to obtain an approximation to ar, the conver-
gence is very slow. We would have to take n = 50 to be guaranteed accuracy to two
decimal places.
Trigonometric Series
Our next application of the Dirichlet test is to the convergence of trigonometric series.
These types of series will be studied in much greater detail in Chapter 9.
7.2.6 THEOREM (Trigonometric Series) Suppose {bk} is a sequence of real numbers sat-
isfying bI ? b2 ? ? 0 and kim bk = 0. Then
(a) Y,'I bk sin kt converges for all t E R, and

(b) Yk I bk cos ki converges for all t E P. except perhaps t = 2pir, p E Z.
Proof. To prove the result, we require the following two identities: For t * 2pir,
pE1,
cos Zt - cos(n + 2)t
71si n kt =
k=1 2 sin i tt
(1)
n
sin(n + !)t - sin i t
Ylcos kt =
k=,
2
2 sin gt
2 (2)
We will prove identity (1), leaving identity (2) for the exercises (Exercise 4). Set
A. = Ink= I sin kt. Using the trigonometric identity
sin A sin B = 2 (cos (A - B) - cos(A + B))
we obtain
(sm)An=
2 k=t
sin i t sin kt
2
=2kCosk2t-cosk+2 ItJ
=2fcos2t-cosn+21IJ.
Thus for t # 2pir, p E 7L,
Cos i t - cos(n + ;)t

An =
2sin 2t
Therefore,
lCos Its + jcos(n +

Z)t(
21 sin Zt{ sin 2tl '
which is finite provided t * 2pir, p e Z. Consequently by the Dirichlet test, the series
in (a) converges for all t # 2 pir, p E Z. If t = 2prr, p E Z. then sin kt = 0 for all k.
Thus the series in (a) converges for all t E R.
The proof of the convergence of the series in (b) is similar. However in this case, when
t = 2pir, cos kt = 1 for all k E N, and the given series may or may not converge. Q
7.2.7 EXAMPLE By Theorem 7.2.6, the series
Coskt
k-i k
converges for all t * 2pir, p E Z. When t = 2prr, p e Z, then cos kt = I for all k and
the series E k diverges. On the other hand, the series
00
1
- zcos ki
k-I k
converges for all t E R.
EXERCISES 7.2
1. *Abel's Test: Prove that if ak converges, and {bk} is monotone and bounded, then E akbk converges.
2. *Show by example that the hypothesis of Theorem 7.2.3 cannot be replaced by bk z 0 and ,, bk = 0.
3. If I ak converges, does E ak always converge?
4. *Prove that
sin(n + 2)t - sin 21

cos kt = , t * 2p7r, p G Z.
k:1 2 sin 2 t
S. Test each of the following series for convergence.
a, (-1)k+I' p>0
00 (-1)klnk (-Ir
kal k' b. k-2
7 k
*G
Mklnk
00 kk kk
sink
* d.
k-1 (k + 1) k
!. k-1D -1) k+1
(k + I)k+1
*f
k_Z In k
si kkt rER,>0 *h. k Cs kt IE R ,p > 0

6. Given that
w (-i)k+1 jr2
k.l k2 12'
determine how large n E N must be chosen so that I nrz/12 - s^ I < 10-', where s^ is the nth partial sum of the
series.
7. If p and q are positive real numbers, show that
00 (ln k)a
converges.
& *Suppose that E ak converges. Prove that
lim lkak=0.
0-00 n k.1
9. As in Exercise 19 of the previous section. let ck = I + ; + + - In k. Set
b = I - l2 + 1 - ... _ 1 = (-1)k+1
3 2n k=1 k
Show that lim b = In 2. (Hint: b, = c_ - c + In 2.)
7.3 Absolute and Conditional Convergence

In this section we introduce the concept of absolute convergence of a series. As we will
see in this and subsequent sections of the text, the notion of absolute convergence is
very important in the study of series. We begin with the definition of absolute and con-
ditional convergence.
7.3.1 DEFINITION A series ak of real numbers is said to be absolutely convergent (or

converges absolutely) if 1 ak I converges. The series is said to be conditionally con-
vergent if it is convergent but not absolutely convergent.
We illustrate these two definitions with the following examples.
7.3.2 EXAMPLES
(a) Since the sequence {l/k} decreases to zero, by Theorem 7.2.3 the series
(-1)k+ 1/k converges. However,
° (-I)k+I xI
II
k=1
k I =
k-1
Ik=oo.
Thus the series (-1)+'/k is conditionally convergent.

(b) Consider the series _ (-1)k+1/k2. By Theorem 7.2.3 the alternating series con-
verges. Furthermore, since 7, 1/k2 < oo, the series is absolutely convergent. U
Our first result for absolutely convergent series is as follows.
7.3.3 THEOREM If Z ak converges absolutely, then F_ ak converges and
akl
k=1 k-1 lakl.
Proof. Suppose Zak converges absolutely; i.e., `jak1 < oo. By the triangle in-
equality, for 1 < p s q,
7,ak) < y IakI

k-p k-p
Thus by the Cauchy criterion (Theorem 2.7.3) the series 2 , ak converges. Finally.
with p = I in the above,
kI ak l = 1im
l t a, I
:!s: l t lakl :9 kY" IakI. Q
Remark, To test a series I ak for absolute convergence we can apply any of the ap-
propriate convergence tests of Section 7.1 to the series 7, Iakl. There is, however, one
important fact which needs to be emphasized. If the series Iak I diverges by virtue of
the ratio or root test, i.e.,
r = lim Ian+ > 1 or a = lim

n-wo
» Ianl > 1,
w_-_ Ia.I
then not only does 71 1akl diverge, but I ak also diverges.
To see this, suppose a > 1. Then as in the proof of the root test, IakI > I for in-
finitely many k. Hence the sequence {ak} does not converge to zero, and thus by Corol-
lary 2.7.5, 1 ak diverges. Similarly, if r > 1 and if I < c < r, then as in the proof of
Theorem 7.1.7(b), there exists a positive integer n and constant M such that
Ianl y Mc"
for all n at no. Thus again, since c > 1, (an} does not converge to zero, and the series
71 ak diverges. We summarize this as follows.
7.3A THEOREM (Ratio and Root Test) Let I ak be a series of real numbers, and let
aslim .
Also, if ak # O for all k E N, let

a, Iak+, I
R= Tim
k-*oo ak
and r = lim
k+m ak
(a) If a 1 or r > 1, then the series 7, ak is divergent.
(c) If a = 1 or r 5 1 <_ R, then the test is inconclusive.
7.3.5 EXAMPLE To illustrate the previous theorem, we consider the series Z. -I P(k)o.
where p is a polynomial and b E R with I b I < 1. In this example, ak = p(k)b` and
thus
ak+, p(k + 1)
ak Ib1 I p(k) (
Since kim p(k + 1)/p(k) = 1 (verify this), Iak+,/a,I = IbI < 1. Thus by the ra-
,im
tio test the series is absolutely convergent.
Rearrangements of Series.
We next take up the topic of rearrangements of series. Loosely speaking, a series I a,.
is a "rearrangement" of the series 7- ak if all the terms in the original series 7, a4 ap-
pear exactly once in the series 7, ak, but not necessarily in the same order. For exam-
ple, the series
l 1z+
3-
+ 22+
4-
+
l2+
5 7-
k= k
The following provides a formal definition of this concept.
7.3.6 DEFINITION A series 7, ak is a rearrangement of the series ak if there exists a

one-to-one function j from N onto N such that ak = ate? for all k E N.
A natural question to ask is: "If the series 7, ak converges and I ak is a re-
arrangement of I ak, does the series ak' converge? If so, does it converge to the same
sum?" As we will see, the answer to both of these questions is no!
7.3.7' EXAMPLE Consider the series

1
k-l
which converges, but not absolutely. Consider also the following series, which is a re-
arrangement of the above:
+3-I+5+I-4+I+11-6 +... (3)

2 7 9
s= k
k=1
As in the proof of Theorem 7.2.4, s < for all n E N. In particular,
s<s3 =1-23
+
6'
Let s,, denote the nth partial sum of the series in (3). Then
1-3+4k-1 1 _ 1)
3'3- R 71
k -I 4k 2k)
Since
I + 1 1 - 8k-3
4k - 3 4k - 1 2k 2k(4k - 1)(4k - 3)'
we have
1 I _ 1 1
0<4k-3+4k-1
for some constant M. Thus the sequence {sin} is strictly increasing, and by the com-
parison test converges. Let s' = lim sin. Since
n-,00
A. +I -'S3n+ 4n + I and
1 1
S3n+2=S3,,+4n+ 1 +4n+3'
the sequences {sin+I} and {s',+2} also converge to s'. Therefore, lim sn = s'. Thus the
series (3) also converges. However, since
6=Si<Sg<Sy<
s' = lim s,, > 6. Thus the series (3) does not converge to the same sum as the original
n-i00
series. This, as we will see, is because the given series does not converge absolutely.
7.3.8 THEOREM If the series E ak converges absolutely, then every rearrangement of

7. ak converges to the same sum.
Proof. Let 7, ak be a rearrangement of I ak. Since lakl < oo, given e > 0, there
exists a positive integer N such that
lakl <E (4)

k-n
for all m z n a N. Suppose ak = atk), where j is a one-to-one function of N onto N.

Choose an integer p (p ? N) such that
{1, 2, . .. , N} C {j(1),j(2), ... ,j(p)}.
Such a p exists since the function j is one-to-one and onto N. Let
a n
Sn ak and s;, _ ak.
k-I k-I
Ifnap,
n rn
Sn - sn = ak - L./ a1(kN
k-! k-I
By the choice of p, the numbers a,, ... , aN appear in both sums and consequently will
cancel. Thus the only terms remaining will have index k or j(k) greater than or equal to
N. Thus, by inequality (4),
Is.-S;,I 52E
for all n a p. Therefore, lim s;, = lim sn, and thus the rearrangement converges to the
n-o0 n- 00
same sum as the original series. p
For conditionally convergent series we have the following very interesting result of
Riemann.
7.3.9 THEOREM Let ak be a conditionally convergent series of real numbers. Suppose

a E R. Then there exists a rearrangement I ak of 7, ak which converges to a.
Before proving the theorem, we illustrate the method of proof with the alternating
series
00
1
k-I
This series converges, but fails to converge absolutely. The positive terms of this series
are Pk = 1/(2k - 1), k E N; and the absolute values of the negative terms are
Qk = 1/2k, k E N. Since 7, 1 = oo, we also have 7, Pk = 7, Qk = oo. Suppose for
purposes of illustration a = 1.5. Our first step is to add "just enough" positive terms to
exceed a. More precisely, we let m, be the smallest integer such that P, + + P,,, > a.
For a = 1.5. m1 = 3; i.e., 1 + 3 + 3 > 1.5, whereas 1 + 3 < 1.5. Our next step is to
go in the other direction; namely, we let n 1 be the smallest integer such that
PI+...+pm,-Q,-. . .-Qn <a.
Again, for a = 1.5, n1 = 1, and 1 + 3 + s - ? < 1.5. We are able to do this since the
series 7, Pk and I Qk both diverge to oo. We continue this process inductively. Choos-
ing the smallest integers mk and nk at each stage of the construction will be the key in
proving that the resulting series converges to a.
Proof. Without loss of generality, we assume that ak * 0 for all k. Let pk and qk be
defined by
Pk = 2 (IakI + ak) and qk = 2 (IakI - ak).
Then pk - qk = ak and Pk + qk = I ak I I. Furthermore, if ak > 0, then qk = 0 and

Pk = ak ; if ak < 0, then Pk = 0 and qk = IakI.
We first prove that the series I Pk and 7, qk both diverge. Since
00 00
I(Pk+qk)= k-I
k-I
I IakI,
they cannot both converge. Also, since
n n n
I ak = 7,
k-I
Pk - 7, qk
k=I k-I
the convergence of one implies the convergence of the other. Thus they both must di-
verge.
Let P1, P;, P3,. .. denote the positive terms of 7, ak in the original order, and let
Q1, Q2, Q3, ... be the absolute values of the negative terms, also in the original order.
The series Pk and 7, Qk differ from I pk and I qk only by zero terms, and thus are
also divergent.
We will inductively construct sequences {mk} and {n*} such that the series
p1 + ... + Pm, - QI - . . . - Qn, + P.'+1
+ + Pm, - Qn..1 - - Q,, + . . . (S)
has the desired property. Clearly series (5) is a rearrangement of the original series.
Let m1 be the smallest integer such that
XI=PI+. ..+P,n'>a.
Such an m1 exists since 7, Pk = oo. Similarly, let n1 be the smallest integer such that
Y1
=X1-Q1-. . .-Qn, <a.
Suppose {m1, ... , mk} and {n1, . . . , nk} have been chosen. Let mk+ 1 and nk+ I be the
smallest integers greater than mk and nk respectively, such that
Xk+I = Yk + Pm,+l + . .. + Pm,., > a, and
Yk+I -Xk+1 -Qn,+I -- . - Qn,., <a.
Such integers always exist due to the divergence of the series I Pk and I Qk. Since
mk+1 was chosen to be the smallest integer for which the above holds,
Xk+I - Pm, 5a.
Therefore,
0<Xk+I -a-15 P.,.,
Similarly,
0<a-Yk+1:Qn,.,-
Since 7, an converges, we havek-.oc
lim Pk = lim Qk = 0. Therefore,
k-*oo
lim Xk = k-.oo
kroo
lim Yk = a.
Let S. be the nth partial sum of the series (5). If the last term of Sn is a Pn, then there
exists a k such that
Yk<Sn:!:-: Xi+I
If the last term of Sn is -Qn, then there exists a k such that
Yk+1 E S. < Xk+1.
In either case, we obtain 1 m Sn = a, which proves the result.
Remark. By a variation of the above proof one can show that if the series 7. ak is
conditionally convergent, then given a, 6 with
-oo<a 5 0<_00,
there exists a rearrangement 7, ak of 7, ak such that
lim S = a and lim S. = /3,

Atia
n1 m
where S. is the nth partial sum of E ak (Exercise 13).
EXERCISES 7.3
1. Prove the following:
a. If lim k°ak = A for some p > I, then F, ak converges absolutely.
b. If iimn k ak = A * 0, then Y, ak diverges.
k
c. If kim k ak = 0, then the test is inbonclusive.
2. 'Suppose I ak < oo and Mk < oo. Prove that the series V, ak bk converges absolutely.
3. a. Prove that if I ak converges and I bk converges absolutely, then 7, ak bk converges.
b. Show by example that the conclusion may be false if one of the two series does not converge absolutely.
4. Suppose that the sequence is monotone decreasing with lim b = 0. If {a,} is a sequence in R satisfying
I b- for all n E N, prove that ak converges absolutely.
5. a. If 7, ak converges absolutely, prove that ekak converges for every choice of ek E { -1. I }.
b. If Y_ ek ak converges for every choice of ek E { -1. 1 }, prove that 7, ak converges absolutely.
6. Test each of the following series for absolute and conditional convergence.
(_I)k+l
= 71
k-1 Nrk
b. I (-1)k
G N/k In(In k) sG
k-I kv
= (-I)k+Ikk
d.k-a-I(-I)kln(lnk)
In k
'e.
k
{-I)k
z k (In
k)P
P>0 k-1 (k + 1)
9. k-1
1 (k +
(-1)k+lkk
h.
00
,(-1+1 sin
() I I.
= (-1)k
,2kt+(_l)k
1)1` k-1 k
7. Test the series 1;k 1 pk/k°, p E R, for absolute and conditional convergence.
8. 'Show that the series I + - 3 + a + s - 1 + diverges.
9. Determine whether the series I - z - i + + s a i +s+s o ill + converges or diverges.
10. 'Prove that the series Ik 1 (sin k)/k is conditionally convergent.
11. Prove that if 7, ak is conditionally convergent, then there exists a rearrangement 7, ak of ak which diverges to c.
12. If ak a 0 for all k E N, and B ak = co, prove that 7, ak = oo for any rearrangement ak of I ak.
13. Suppose the series I ak is conditionally convergent. Given a,)9 with -oo s a <- rA s oo, show that there exists
14. Suppose every rearrangement of the series

ft
a rearrangement I ak of ak such that lim S. = a and lim S. where S. is the nth partial sum of 7, ak.
M-00
ak converges. Prove that ak converges absolutely.
7.41 Square Summable Sequences2

In this section we introduce the set l2 of square summable sequences of real numbers
and derive several useful inequalities. This set occurs naturally in the study of Fourier
series.
7.4.1 DEFINITION A sequence {ak}k 1 of real numbers is said to be in 12, or to be square

summable, if
0C
Iak<oo.
k-1
For {ak} E I2 set
II {ak} II 2 = Xak.
The set 12 is called the space of square summable sequences, and the quantity II{ak}II2
is called the norm of the sequence {ak}.
Remark. Since a sequence {ak} in R is by definition a function a from N into R with

ak = a(k), it is sometimes convenient to think of 12 as the set of all functions a: N - R
for which
II a Ill =
F 1
I a(k)I2 < C.O.
7.4.2 EXAMPLES
(a) For the sequence { l/k}x 1,
kk2.
H{'}1122
Since this is ap-series with p = 2. the series I 1/k2 converges and thus { 1/k} E 12. On
the other hand, since
l %Ilk II2 k-11k - '

we have { 1 / Vk} 4E l2.
2. The topic of square summable sequences, although important in the study of Fourier series, is not specif-
ically required in Chapter 9 and thus can be omitted on first reading. The concept of a nonmed linear space
occurs on several occasions in the discussion of subsequent topics in the text. At that point, the reader can
study this topic more carefully.
(b) For fixed q, 0 < q < oc, consider the sequence { I /k v }. Then
{ klq 22 k'
By Example 7.1.9(a) this series converges for all q > ; and diverges for all q s ;-Thus
{ 1/k°} E 12 if and only if q > ;
Cauchy-Schwarz Inequality
Our main goal in this section is to prove the Cauchy-Schwarz inequality for sequences
in l2. First, however, we prove the finite version of this inequality.
7.4.3 THEOREM (Cauchy-Schwarz Inequality) If n E NI, and at,.,., an and b1..... bn

are real numbers, then
n+ n
Iakbkl `- , bk
k=1 Fk= k=1
Proof. Let A E R and consider
0 ` 7,(Iakl - AIbkl)2 = 1 ak - 2A
n
` Iakbkl + A2 , bk.
n
k=1 k=1 k-1 k.1

The above can be written as
0 <_ A - 2AC + A2B,
where A = 7,k=, ak, C = 1k-1 Iakbkl. and B = 2;k=1 bk. If B = 0, then bk = 0 for
all k = 1, ... , n, and the Cauchy-Schwarz inequality certainly holds. If B 0, we
take k = C/B which gives
B2
OVA -
or C2 <_ AB; that is,
k lak 5
()(n)
Taking the square root of both sides gives the desired result. 0
As a consequence of the previous result we have the following corollary.
7A.4 COROLLARY (Cauchy-Schwarz Inequality) If {ak}, {bk} E 12, then 7,i 1 akbk is
absolutely convergent and
lakbkl s II{ak}112II{bk}II2.
k-1
Proof. For each n E N, by the previous theorem
Iakbkl ak bk II{ak}II2II{bk}II2-
k-1 k-l k-I
Letting n --> oo gives the desired result.
Minkowski's Inequality
Our next result shows that the norm 11 112 satisfies the "triangle inequality" on 12.
7A.5 THEOREM (Minkowski's Inequality) If {ak} and {bk} are in 12. then {ak + bk} E 12
and
II{ak + bk}II2 c II{ak}1I2 + II{bk}II2-
Proof. By hypothesis, each of the series 7, a2 and I bi' converge. Also, by Corol-
lary 7.4.4, the series E ak bk converges absolutely. Since
(ak + bk)2 = ak + 2akbk + bk <_ ak + 2lakbkl + Mk,
we have
00 -0
II{ak + bk}IIi = 2 (ak + bk)2 5 II{ak}IIz + 2 7, Iakbkl + II{bk}IIi
k-1 k=1
which by the Cauchy-Schwarz inequality,

II{ak}IIi + 211{ak}II2II{bk}II2 + II{bk}II2
S = (11{ak}II2 + 110J12)2-
Taking the square root of both sides gives the desired result.
Although not specifically stated, Minkowski's inequality is also true for finite
sums. In particular, if n E N and a,, . .. , a,,,'bi, . , ..
are real numbers, then
j(ak+bk)2S.Jj a,2k + bk. (6)

k-1 I k=1
In the following theorem we summarize some of the properties of the norm

112 on 12. These are very similar to the properties of the absolute value function on R.
II
As for the absolute value, the inequality

11{a4 + bk}112 < I1{ak}II2 + II{bk}II2
is also referred to as the triangle inequality for 12.
7A.6 THEOREM The norm 11 16 on 122 satisfies the following properties:

(a) ll{ak}112 ? 0/oral! {ak} E 12.
(b) II{ak}1I2 = 0 if and only if ak = O for all k E N.
(c) If {ak} E 12 and c E R, then {cak) E 12 and II{cak}II2 = IcI II{ak}II2.
(d) If {ak}, {bk} E l2, then {ak + bk} E 12 and II{ak + bk}IL II{ak)II2 + II{bk}II2.
Proof. The results (a) and (b) are obvious from the definition, and (d) is just a restatement
of Minkowski's inequality. The verification of (c) is left as an exercise (Exercise 5). 0
Euclidean Space U8"

The finite versions of both the Cauchy-Schwarz and Minkowski inequalities can be
used to provide some useful results about n-dimensional real euclidean space R". For
n E N, R" consists of all n-tuples (a,, a2, ... , aR) of real numbers. When n = 1, R' is
just the real line R. For notational convenience, we denote the above n-tuple by a; i.e..
a = (a,, a2, ... , a").
The point a is also referred to as the vector a with components a,, k = 1, 2, ... , n.
Two points a and b of R" are equal if and only if ak = bk for all k = 1, 2, ... , n. If
a = (a,, ... , a"), b = (b,,. . . , b") E R^ and c E R, we define a + b and ca by
a+b=(at+b,,...,a,+b"),
ca = (ca,, . . . , ca").
For two vectors a and b in IR", the Inner product of a and b, denoted (a. b). is de-
fined by
(a, b) akbk.
kai
Also, for a E l8" we define the norm or length of the vector a by
118112= \/(a, a)= ak.
As a consequence of the Cauchy-Schwarz inequality,

I(a, b)I s IIaI6IIbII2
for all a, b E W. The eucildean distance d2(a, b) between a and b is defined as
d2(a, b) = IIa - N12 = V 7, (ak - bk)2.

n
k=,
When n = 1, d2(a, b) simply becomes Ia - bI . By Minkowski's inequality (6) we have

d2(a, b) s d2(a, c) + d2(c, b)
for all a b. c E R".
Nonmed Linear Spaces

The space 12 as well as Q8" are both examples of vector spaces over It For complete-
ness we include the definition of a vector space.
7.4.7 DEFINITION A set X with two operations " + ", vector addition, and " ", scalar
multiplication, satisfying
x+ y E X for all x, y E X, and
c'xEX forallxEX,cER
is a vector space over R if the following are satisfied:
(a) x + y = y + x. (commutative law)
(b) x + (y + z) = (x + y) + z. (associative law)
(c) There is a unique element in X called the zero element, denoted 0, such that
x+0=x forallxEX.
(d) For each x E X, there exists a unique element -x E X such that
x+(-x)=0.
(e)
(f) +a'y.
(g) (a+b)'x=a'x+b'x.
(h) I'x=x.
It is clear that R" with + and ' defined by
a+b=(a,+b,,. . .,an+bn),
c' a = (cab. . .,Can)
is a vector space over R. Similarly, if addition and scalar multiplication of sequences
{ak}, {bk} in 12 are defined by
{ak} + {bk} = {ak + bk},
c {ak) = {cak}, c E R,
then it is easily shown that 12 is a vector space over R. The fact that {ak + bk} and {cak}
are in 12 is part of the statement of Theorems 7.4.5 and 7.4.6. The zero element 0 in l2
is the sequence {ak} where ak = 0 for all k E N.
In addition to being vector spaces, the spaces R" and 12 are also examples of
normed linear spaces. The concepts of a "normed linear space" and a "norm" are de-
fined as follows:
7.4.8 DEFINITION Let X be a vector space of R. A function !III : X -+ R satisfying

(a) llxll ? Ofor all x G X,
(b) IIxIl = 0 if and only if x = 0,
(c) llcxll = lcl llxll for all c E R, x E X, and
(d) llx+yll!5 llxll+IIYIIforallx,yEX
is called a norm on X. The pair (X, 11 11) is called a normed linear space.
Inequality (d) is called the triangle inequality for the norm 11 11. By Theorem 7.4.6,
is a nonmed linear space. Additional examples of normed linear spaces will
(12, II Il2)
occur elsewhere in the text and the exercises.
If (X, 11 11) is a normed linear space, the distance d(x. y) between two points
x, y E X is defined as d(x, y) = II x - y II. From the definition of the norm II 1. it im-
mediately follows that
(a) d(x, y) ? 0 for all x, y E X.
(b) d(x,y) = 0 if and only if x = y,
(c) d(x, y) = d(y, x), and
(d) d(x, y) s d(x, z) + d(z. y) for all x, y, z e X.
Thus d is a metric on the vector space X. As in Definition 2.1.6, fore > 0. p E X.
the e-neighborhood Nf(p) is defined as
For example, in 1I2. with norm II II2 an e-neighborhood of a = (a,, a2) is the set of
points x = (x,, x,) satisfying (x, - a,)2 + (x, - a2)2 < e-, that is. an open disc with
center a and radius a (See Figure 7.2). In 1183, an e-neighborhood of a is simply an open
ball with center a and radius e.
X2
a,
a,
Figure 7.2 An e-Neighborhood of a = (a,, a2) E R2
Since the notions of "convergence of a sequence" (2.1.7) and "limit point of a set"
(2.4.5) are defined in terms of e-neighborhoods, both of these concepts have analogous
definitions in a normed linear space (X, 11 11). Thus for example, a sequence in X
converges to x E X, if given e > 0, there exists n,, E IO' such that IIx - xfl < e for
all n E N. n ? no. If this is the case, we say that x,, converges in the norm II II to x.
This type of convergence, referred to as norm convergence, will be encountered again
later in the text.
As a general rule, many of the results involving sequences of real numbers are still
valid in the setting of a normed linear space. This is especially true for those theorems
whose proofs relied only on properties of the absolute value. Great care, however, must
be exercised with theorems that rely on the supremum property of R. For example. the
Bolzano-Weierstrass theorem fails for (12, 11 112) (Exercise 7). On the other hand, the
Bolzano-Weierstrass theorem still holds in (R". Ii IIz) (Exercise 6).
6 EXERCISES 7.4
1. Determine which of the following sequences are in 12.
1 ( ] " c I In k
d.
J !,in kj"
sa. b.
Ink k=z t Vk Ink I k- 2 Vk k-z k k-1
2. Determine all values of p E R such that the given sequence is in 12.
{k"Ik d. Vk-(Ink)P k.:

3. If {a,.} E 12, prove that tli m ak = 0.
4. *If {ak} E 12, prove that , at/k converges absolutely.

5. If {ak} E 12 and c E R. prove that {cak} E 12 and II{cak}IIz = Iclll{ak}IIz
6. a. Suppose that (%)-A., is a sequence in R", where f o r each k E N, pk = ( p , , . . . . . pk.,,) Prove that the se-
quence { pk} converges top = (p,, ... , p") in the norm 11112 if and only if m pk..; = p, for all i = 1..... n.
k-x
b. Prove that every bounded sequence in R" has a convergent subsequence.
7. For each n E N. let e" be the sequence in 12 defined by
e"(k)= (0, kn,
1. k=n.
Show that II e" - emit = V if n # m. (Remark: The sequence {e" } is a bounded sequence in 12 with no conver-
gent subsequence. Thus the Bolzano-Weierstrass theorem (2.4.11) fails in 12.)
8. Show that equality holds in the Cauchy-Schwarz inequality if and only if for all k E N. h, = cak for some c E R.
9. For a, b E R", let (a. b) denote the inner product of a and b. Prove each of the following.
a. (a,a)?0
b. (a, a) = 0 if and only if a = (0.. ... 0)
c. (a, b + c) _ (a, b) + (a, c)
d. (a, b) _ (b, a)
e. (a. b) = 12 (IIAV2
2 + Vblli - IIa - 15111).
10. *For a, b E R", use the law of cosines to prove that (a. b) = Ilal2llbf 2cos 0, where B is the angle between the
vectors a and b.
11. Let (X, II V) be a normed linear space. Prove that I lIxII - Ilyll l 5 Ox - yll for all x. y E X.
12. Let 1' denote the set of all sequences {ak} satisfying H{ak}ll, _ Ik , lakl < oo.
a. Prove that (1', II II,) is a normed linear space.
b. Prove that 1' C 12.
13. Determine all values of p E R for which each of the sequences in Exercise 2 is in 1'.
14. Let X be a nonempty set, and let S(X) denote the set of all bounded real-valued functions on X. For f E 31(X), set
Ilf L = sup{I f(x) I : x r= X}. Prove that (11(X), 11 IL) is a nonmed linear space.
Notes 313
NOTES
The geometric series is perhaps one of the most important gence. An excellent illustration of these eighteenth-century
series in analysis. It forms the basis for the proof of the techniques is Euler's derivation of the sum of the series
ratio and root test and thus also for the study of conver-
gence of power series. In the seventeenth and eighteenth 1I I
+3-+..
I
centuries, the convergence of many series was established

by comparison with the geometric series. The geometric The convergence of this series had been established ear-
series dates back to Euclid in the third century B.C. The lier by Johann Bernoulli. Using results from the theory of
formula for a finite sum of a geometric progression ap- equations and applying them to the power series expan-
peared in Euclid's Elements, and Archimedes. in his treatise sion of sin x, Euler was able to show that the sum of the
Quadrature of the Parabola, indirectly used the geomet- given series was 7r2/6.'
ric series to find the area under a parabolic arc,3 Even As power series were used more frequently to approx-
though Greek mathematicians knew how to sum a finite imate mathematical quantities, the emphasis turned to de-
geometric progression, they used reducto ad absurdum riving precise error estimates in approximating a function
arguments to avoid dealing with infinite quantities. by a finite sum of the terms of its power series. In 1768.
Infinite series first appeared in the middle ages. In the d'Alembert made a careful examination of the binomial
fourteenth century, Nicole Oresme (1323?-1382) of Italy series (I + r)", p E Q, discovered earlier by Newton.6
provided a geometric proof to the effect that the infinite D'Alembert computed the value of n for which the absolute
series value of the ratio of successive terms is less than 1, thereby
ensuring that the successive terms decrease. More impor-
1 2 n
2+22+
tantly, d'Alembert computed the bounds on the error made
2"
in approximating (I + r)" by a finite number of terms of
this series. His argument relied on a term-by-term compari-
had sum equal to 2, a result now quite easy to prove using
son with a geometric series, similar to the proof used in es-
power series. He was also the first to prove that the har- tablishing the ratio test. D'Alembert's result can easily be
monic series diverged, whereas the Italian mathematician converted into a proof of the convergence of the series.
Pietro Mengoli (1624-1686) was the first to show that the Computations analogous to those of d'Alembert
sum of the alternating harmonic series I - i + ; - o were also undertaken by Lagrange in his book, Theorie
+ is In 2. The results of Oresme and Mengoli, as des Foncrions Analvtiques, published in 1797. The re-
well as others of this era, were stated verbally, the infinite mainder term for Taylor's formula (to be discussed in the
sum notation did not appear until the seventeenth century. next chapter) first appeared in this text. Cauchy was un-
With the development of the calculus in the mid- doubtedly influenced by the works of d'Alembert and
seventeenth century, the emphasis shifted to the. study of Lagrange in his definition of convergence of a series.
power series expansions of functions and computations in- However, unlike the results of d'Alembert, which applied
volving power series. The expansion of functions in power only to the binomial series, and those of Lagrange for
series was, for Newton and his successors, an indispensable Taylor series, Cauchy's definition of convergence was en-
tool. In his treatise, Of the Method of Fluxions and Infinite tirely general: applicable to any series of real numbers.
series,s Newton includes a discussion of infinite series Using his definition of convergence. Cauchy proved
techniques for the solution of both algebraic and differen- that the statement now known as the Cauchy criterion was
tial equations. Both the geometric series and the binomial necessary for the convergence of a series of real numbers.
series were employed in many of his computations. This was also known earlier to Bolzano. However, without
During the seventeenth and eighteenth centuries, the completeness property of the real number system, nei-
mathematicians operated with power series in the same way ther Bolzano nor Cauchy was able to prove that the Cauchy
as with polynomials; often ignoring questions of conver- criterion was also sufficient for convergence of the series.
3. See p. 105 of the text by Katz.

4. The Mathematical Works of Isaac Newton, edited by D. T. Whiteside, Johnson Reprint Corporation. New York. 1964.
5. For details, see the article by J. Grabiner.
6. "REf exions sur les suites et sur les ravines imaginaires;' Opuscules Methematiques, vol. 5 (1768). pp. 171-215.
1. Given two series Ek oak and Boko.o bk, set c, = I , _ o ak b _ k, n = 0, 1.2.. . ..The series !.O c is called the
Cauchy product of E ak and I bk.
a. If I ak andd I bk converge absolutely, prove that E c converges absolutely and that
-oc"-\koak)(kobk).
b. Let ak = bk = (- I)k/, k = 0, 1, 2. ... Prove that the Cauchy product of F. ak and B bk diverges.
c. Prove that the result of part (a) is still true if only one of the two series converge absolutely: the other series
must still converge.
2. Let X be a nonempty set. If f is a real-valued function on X, define
II f 11, = sup S Y, I jr (x) I : F is a finite subset of X} .

.eF
In the above, the supremum is taken over all finite subsets F of X. Denote by l'(X) the set of all real-valued func-
tions f on X for which II f 11, < 00.
a. Suppose X is infinite. If jr E l'(X), prove that {x E X:f(x) # 0} is at most countable.
b. If {x E X:f(x) # 0} _ (x,: n E A}, where A C N, prove that 11f11, _ Z.EA I f(x.)I.
c. Prove that (1'(X), II II,)is a normed linear space.
3. For f E °t[a, b], set
(f(x))2dx]1/2
IIf112= I I
a. Prove that for f, g E R[a, b], f. I jr (x)g(x) I dx s 11 f 11211 g I1r
b. Forf, g E JL[a, b], prove that 11f + g 112 s 11f112 + 118112
c. Prove that 11 16 defines a norm on %[a, b], the space of continuous real-valued functions on [a, b).
d. is11 I6 anorm on Jt(a,b]?
4. As in the previous exercise, let b] denote the vector space of continuous real-valued functions on (a, b). For
f E 19[a, b], set IIf1I. = max{ If(x) I : x r= [a, b]}.
a. Prove that 11II. is a norm on %(a, b].
b. Prove that a sequence {f.) in 4[a, b] converges to f e 14[a, b] if and only if given e > 0. there exists
n. E N such that I f (x) - f (x) I < e for all x E [a, b] and all n E N, n a n,.
DEFINITION
(a) Let (X,11 11) be a normed linear space, and let E C X. A function f : E -+ R is continuous at p E E If given
e > 0, there exists a S > 0 such that If(x) - f(p) I < e for all x E E with IIx - pIi < S.
(b) Let X be a vector space over It A function f : X -+ R is linear if f(ar + by) = af(x) + bf(y) for all x, y E X
and a, b E R.
5. Let (X, II II)be a normed linear space and let f : X - R be a linear function. Prove that the following are equivalent.
a. f is continuous at some p E X.
b. f is continuous at 0.
c. There exists a positive constant M such that If(x)1 <- MIIxiifor all x E X.
d. There exists a positive constant M such that I f(x) - f(y)I <_ Mix - ylifor all x, y E X.
e. f is uniformly continuous on X.
X
6. For fixed b E 12, define 1' : 12 -+ R by r(a) = (a, b) = I a(k)b(k). Prove that r is a continuous linear function on l
k=1
7. As in Exercise 3, let b] denote the vector space of continuous real-valued functions on [a. b] with norm l N2.
For fixed g E 9t[a, b], prove that r defined by r(f) = fa f(x)g(x) dx is a continuous linear function on `g[a, b].
Behforooz, G. H.. "Thinning out the harmonic series;' Jungck, G., "An alternative to the integral test;' Math.
Math. Mag. 68 (1995), 289-293. Mag. 56 (1983), 232-235.
Boas, R. P., "Estimating remainders:' Math. Mag. 51 Katz, V. J "Ideas of Calculus in Islam and India," Math.
(1978), 83-89. Mag. 68 (1995),163-174.
Boas, R. R, "Partial sums of infinite series, and how they =I
grow," Amer. Math. Monthly 84 (1977), 237-258.
Kortam, R. A., "Simple proofs for 7 2 and
k_I k 6
Cowen, C. C., Davidson. K. R., and Kaufman, R. P., 00 x2
"Rearranging the alternating harmonic series," sin x =x l - a f ; 'Marh. Mag. 69 (1996),
k.1 kir'
Amer. Math. Monthly 87 (1980), 817-819. 122-125.
Efthimiou, C. J., "Finding exact values for infinite se-
ries:' Math. Mag. 72 (1999), 45-51. Maher, P., "Jensen's inequality by differentiation:' Math.
Goar, M., "Olivier and Abel on series convergence: An Gaz. 73 (1989). 139-140.
episode from early 19th century analysis:' Math. Tolsted, E., "An elementary derivation of the Cauchy,
Mag. 72 (1999), 347-355. Holder, and Minkowski inequalities from Young's
Grabinger, Judith V., The Origins of Cauchy's Rigorous inequality," Math. Mag. 37 (1964), 2-12.
Calculus, The MIT Press, Cambridge, Massachu- Young, R. M., "Euler's constant:' Math. Gazette 75
setts, 1981. (1991), 187-190.
pl Sequences and Series
J of Functions
8.1 Pointwise Convergence and Interchange of Limits
8.2 Uniform Convergence
8.3 Uniform Convergence and Continuity
8.4 Uniform Convergence and Integration
8.5 Uniform Convergence and Differentiation
8.6 The Weierstrass Approximation Theorem
8.7 Power Series Expansions
8.8 The Gamma Function
In this chapter we will study convergence properties of sequences and series of real-
valued functions defined on a set E. In most instances E will be a subset of R. Since we
are dealing with sequences and series of functions, there naturally arise questions in-
volving preservation of continuity, differentiability, and integrability. Specifically. is the
limit function of a convergent sequence of continuous, differentiable, or integrable func-
tions again continuous, differentiable, or integrable? We will discuss these questions in
detail in Section 8.1 and show by examples that the answer to all of these questions is
generally no. Convergence by itself is not sufficient for preservation of either continu-
ity, differentiability, or integrability. Additional hypotheses will be required.
In the 1850s Weierstrass made a careful distinction between convergence of a se-
quence or series of numbers and that of a sequence or series of functions. His concept
of uniform convergence is the additional hypothesis required for the preservation of
continuity and integrability. It was also Weierstrass who constructed a continuous but
nowhere differentiable function, and who proved that every continuous real-valued
function on a closed and bounded interval can be uniformly approximated by a poly-
nomial. As prominently as Cauchy is associated with the study of sequences and series
of numbers, Weierstrass is likewise associated with the study of sequences and series of
functions. For his many contributions to the subject area, Weierstrass is often referred
to as the father of modern analysis.
The study of sequences and series of functions has its origins in the study of power
series representation of functions. The power series of In (I + x) was known to Nicolaus
Mercator (1620-1687) by 1668, and the power series for many of the transcendental
317
318 Chapter 8 Sequences and Series of Functions
functions such as Arctan x, Arcsin x, among others, were discovered around 1670 by
James Gregory (1625-1683). All of these series were obtained without any reference to
calculus. Newton's first discoveries, dating back to the early months of 1665. resulted
from his ability to express functions in terms of power series. His treatise on calculus,
published posthumously in 1737, was appropriately entitled A treatise of the method of
fiuxions and infinite series. Among his many accomplishments. Newton derived the
power series expansion of (1 + x)" using algebraic techniques. This series and the
geometric series were crucial in many of his computations. Newton also displayed the
power of his calculus by deriving the power series expansion of ln(I + x) using term-
by-term integration of the expansion of 1/(1 + x). The mathematicians Colin Maclau-
rin (1698-1746) and Brooks Taylor (1685-1731) are prominent for being the first math-
ematicians to use the new calculus in determining the coefficients in the power series
expansion of a function.
8.1' Pointwise Convergence and Interchange of Limits

In this section we consider a number of questions involving sequences and series of func-
tions and interchange of limits. Some of these questions were incorrectly believed to be
true by many mathematicians prior to the nineteenth century. Even Cauchy in his text
Cours d'Anahse "proved" a theorem to the effect that the limit of a convergent sequence
of continuous functions was again continuous. As we will shortly see, this result is false!
To begin our study of sequences and series of functions we first define what we
mean by pointwise convergence of a sequence of functions. If E is a nonempty set and
if for each n E N, f is a real-valued function on E, then we say that { is a sequence
of real-valued functions on E. For each x E E, such a sequence gives rise to the se-
quence of real numbers, which may or may not converge. If the sequence
{ converges for all x E E, then the sequence is said to converge pointwise on
E, and by the uniqueness of the limit,
f(X) = lim L(x)
defines a function f from E into R. We summarize this in the following definition.
8.1.1 DEFINITION Let { f } , be a sequence of real-valued functions defined on a set E.

The sequence if.) converges pointwise on E if { converges for every x E E.
If this is the case, then f defined by
f(x) = I i m x E E.
defines a function on F. The fiaiction f is called the limit of the sequence If,).
In terms of e and n, the sequence {f,) converges pointwise (of if for each x E E,
given e > 0, there exists a positive integer n,, = e) such that
If,,(x) - f(x)I < e
8.1 Pointwise Convergence and Interchange of Limits 319
for all n ? na. The expression no = na(x, e) indicates that the positive integer n, may
depend both on e and x E E.
If as above, {fa}°a° is a sequence of real-valued functions on a nonempty set E,
1
then with the sequence {fa} we can associate the sequence {S0} of nth partial sums,
where for each n E N, S. is the real-valued function on E defined by
Sa(x) = (A + ... + fa)(x) = I fk(x)

k=1
The sequence is called a series of functions on E denoted by I* j fk or simply

E fk. The series 1 fk converges pointwise on E if for each x E E the sequence {Sa(x)}
of partial sums converges; that is, the series Z. I fk(x) converges for each x E E. If the
sequence {Sa} converges pointwise to the function S on E, then S is called the sum of
the series E fk and we write S = E. , fk, or if we wish to emphasize the variable x,
00
S(x) = 7, A(x), x E E.
k=1
Suppose fa : [a, b] -+ R for all n e Ni, and f(x) each x E [a, b].
Among the questions we want to consider are the following:
(a) if each fa is continuous at p E [a, b], is the function f continuous at p? Recall

that the function f is continuous at p if and only if
limf(t) = f(p).
Since f (x) = lim fa(x), what we are really asking is does
lim{ lim ff(t) = lim lim fa(t)

r-4p M-600(1-+p
(b) If for each n E N the function Ja is differentiable at p E [a, b], is f differen-

Sable at p? If so, does
f(p) = lim f,(p)?
n
(c) If for each n E N the function fa is Riemann integrable on [a, b], is f Riemann
integrable? If so, does
rb jb
f= a--.f f.?
a a
We now provide a number of examples to show that the answer to all of these ques-
tions is generally no.
8.12 EXAMPLES
(a) Let E = [0, 1 ], and for each x E E, n E N. let f"(x) = x". Clearly each f is con-
tinuous on E. The graphs of f, f,-, f4, f6, and f,i, are given in Figure 8.1. Since fn(1) = I
for all n, Jim 1. If 0 :5 x < 1, then by Theorem 2.2.6(e), lim fn(x) = 0. There-
n-r x
fore.
u m f,(x) = f(x) = S 0, 0 ` i < 1,
The function f, however, is not continuous on [0, 1 ].
Figure 8.1 Graphs of f,, f2,f4,ff,f10 a
(b) Consider the sequence {fk}j 0 defined by

,
f(x)=(1+ xE68.
For each k = 0, 1 , 2, .. . , fk is continuous on R. Consider the series 0 fi which for

each x E R is given by
x l k
fk(x)= x (1
L=O k=0 + x2
8.1 Pointwise Convergence and interchange of limits 321
We now show that this series converges for all x E I and also find its sum f. If x = 0.
then fk(O) = 0 for all k, and thus f(0) = 0. If x * 0, then 1/(1 + x2) < 1, and hence
by Example 2.7.2(a),
X2
71
(- = 1 + xz = f(x).
k=O
Therefore,
0, x = 0,
f(x) 1 +xz, x#0,
which again, is not continuous on R.
(c) Let {xk} be an enumeration of the rational numbers in [0, 1]. For each n E N,
define fn as follows:
J 0, if x = xk, 1 <_ k 5 n,
fn(x) = 1, otherwise.
Since each fn is continuous except at xi, ... , xn, fn is Riemann integrable on [0, 1 ]
with fn fn(x) dx = 1. On the other hand,
if x is rational,
lim f(x) = s10,1, if x is irrational,
which by Example 6.1.6(a) is not Riemann integrable on (0, 11.
(d) For x E [0, 1], n E N, let fn(x) = nx (I - x2)n. Since each fn is continuous, fn is
Riemann integrable on [0, 1 J. If 0 < x < 1, then 0 < 1 - x'' < 1, and thus by
Theorem 2.2.6,
lim nx(1 - X'-)n = 0.
nyoc
Finally, since 0, we have

f(x) = 1 m L(x) = 0 for all x E [0, 1 J.
Thus f is also Riemann integrable on (0, 1 ]. On the other hand,
1 n
dx = n x (1 - xZ)r dx =
10 0
1
2n+1
Therefore.
lim
n-ioc
f' frt(x) dx = 2 # 0 = I f (x) dx.
0 J11
(e) As our final example, consider fn(x) = (sin nx)/n, x E R. Since I sin nx -5 1 for all
x E=- R and n E N,
f (x) = lim
n-+oo
fn(x) = 0 for all x E R.
322 Chapcer8 Sequences and Series of Functions
Therefore, f'(x) = 0 for all x. On the other hand,

f;'(x) = cos nx.
In particular f;(0) = 1 so that lim f',(0) = 1 # f'(0). This example shows that in
general
lyltto fn(X)) # llm f n(x)

dx
Additional examples are given in the exercises.
EXERCISES 8.1
1. Find the pointwise limits of each of the following sequences of functions on the given set.
sin nx
*a. 11 _, }, xE(0,oo) b.
11+nx
. X E f 0. oo)
*c. {(cos x)Z"}, x E R d. {nxe-"}, x E R

2. For n E N, define f, : N -+ R by f, (m) = n/(m + n). Prove that
HIM lim f,(m)) # liml!m,fn(m)).

3. Consider the sequence {f.) with it 2, defined on [0, 1 ] by
nZx, 05x5 1,n

f(x) 2n-n2x, 1I
n
<xs 2n
0, ?<x<_1.
it
a. Sketch the graph of f, for it = 2, 3, and 4.
b. Prove that lim f,(x) = 0 for each x E [0, 1 ].
c. Show that fo fn(x) dx = I for all n = 2, 3.... .
4. Let g (x) = e-n`/n, x E 10, oo), it E N. Find lim
n-+oo
g,(x) and lim
ny0o
g;,(x).
5. Let f,(x) _ (x/n)e- `1n, X E [0.00).
a. Show that tim fn(x) = 0 for all x E [0, oo).
b. Given e > 0, does there exist an integer it, E N such that If,(x) I < e for all x E (0, oo) and all it z it,,.
(Hint: Determine the maximum of fn on [0, oo).)
c. Answer the same question as in pan (b) for x E [0, a], a > 0.
6. *If a,, m ? 0, it, m E N, prove that
cc cc - cc
71 71 an.,, = L an.,,,
n=Im=1 n=1
with the convention that if one of the sums is finite, so is the other, and equality holds. Conversely, if one is infi-
nite, so is the other.
8.2 Uniform Convergence

All of the examples in the previous section show that pointwise convergence by itself
is not sufficient to allow the interchange of limit operations; additional hypotheses are
required. It was Weierstrass who, in the 1850s, realized what additional assumptions
were needed to ensure that the limit function of a convergent sequence of continuous
functions was again continuous.
Recall from Definition 8.1.1, a sequence { f,} of real-valued functions defined on a
set E converges pointwise to a function f on E if for each x E E, given c- > 0, there ex-
ists a positive integer n = n,,(x, e) such that
Nx) - L WI < E
for all n ? n0. The key here is that the choice of the integer n,, may depend not only on e.
but also on x E E. If this dependence on x can be removed, then we have the following.
8.2.1 DEFINITION A sequence of real-valued functions {f} defined on a set E converges

uniformly to f on E. if for every e > 0, there exists a positive integer n such that
If,(x) - f(x)1 < E

for all x E E and all n ? n,,. Similarly, a series 7-4'- i fr of real-valued functions con-
verges uniformly on a set E if and only if the sequence of partial sums converges
uniformly on E.
The inequality in the definition can also be expressed as
f(x)- E <f(x)
for all x E E and n ? n,,. If E is a subset of R, then the geometric interpretation of the
above inequality is that for n ? n,,, the graph of y = f"(x) lies between the graphs of
y = f(x) - e and y = f(x) + E. This is illustrated in Figure 8.2 with E = [a, b].
8.2.2 EXAMPLES
(a) For x E [0, 1 ], n E t\i, let f"(x) = x". By Example 8.1.2(a), the sequence {f.1 con-
verges pointwise to the function
0sx< I,
f(x) _ {°, x= 1.
We now show that the convergence is not uniform. If the convergence were uniform,
then given E > 0, there would exist a positive integer n, such that I "(j) - f(x) 1 < e
for all n >_ n" and all x E [0, 1 ]. In particular,
x<e for all x E [0, 1).

324 Chapters Sequences and Series of Functions
a b
Figure 8.2
This, however, is a contradiction if e < 1. Even though the convergence is not uniform
on (0, 1 ], the sequence does converge uniformly to 0 on [0, a] for every a, 0 < a < 1.
This follows immediately from the fact that for x E [0, a], LAW I = I x' I -S a".
(b) Consider the series
00
k71
{ixe-k'' - (k - 0:9 x < 00.
Since the series is a telescoping series, the nth partial sum S"(x) is given by
S"(x) = nxe"";.
It is easily shown (Exercise Id, Section 8.1) that
S(x) = lim
M- 00
S"(x) = 0 for all x E (0, oo).
We now show that the convergence is not uniform. Suppose that the sequence {S"} con-
verges uniformly to 0 on [0, oo). Then if we take e = 1, there exists a positive integer
n0 such that
IS.(x) - S(x) l = S"(x) = nxe -"'_ < I
for all n a n, and x E [0, oo). However, for each n E N, by the first derivative test S.
has a maximum at x, = with
M. = max S"(x)
F2e_
This term, however, is greater than 1 for n =' 6. Thus the convergence is not uniform.
The graphs of S4, Ss, and S16 are given in Figure 8.3. Since the maximum of each S.
moves along the x-axis as n -* oo, such functions are often referred to as "sliding-
hump" functions. For this example, M. -+ oo as n -+ oo.
Figure 8.3 Graphs of S Se, S16
The Cauchy Criterion

Our first criterion for uniform convergence is the Cauchy criterion. The statement of
this result is very similar to the definition of Cauchy sequence.
8.2.3 THEOREM (Cauchy Criterion) A sequence of real-valued functions defined on

a set E converges uniformly on E if and only if for every e > 0, there exists an integer
no E NI such that
Ifn(x) - fm(x)I < E (1)
for all x E E and all n, m ? n0.

Proof. If (f.} converges uniformly to f on E, then the proof that inequality (1) holds
is similar to the proof that every convergent sequence is Cauchy. Conversely, suppose
that the sequence if.) satisfies inequality (1). Then for each x E E, the sequence { f (x)}
is a Cauchy sequence in R, and hence converges (Theorem 2.6.4). Therefore,
f(x) = lim .fi(x)
exists for every x E E.
We now show that the sequence {f,,} converges uniformly to f on E. Let e > 0 be
given. By hypothesis, there exists n, E NI such that inequality (1) holds for all x E E
and all n, m a no. Fix an m >_ n0. Then
f
LAX) - fm(x)I = ha Wx) - fm(x)I E
for all x E E. Since the above holds for all m ? n the sequence {f.} converges uni-
formly to f on E. Q
The analogous result for series is as follows.
326 Chap{er 8 Sequences and Series of Functions
8.2A COROLLARY The series lx I ft of real-valued functions on E converges uniformly

on E if and only if given e > 0. there exists a positive integer n,,, such that
kn
171-fx) I <E
nsl
for all x E E and all integers m > n ? no.

Proof. The proof of the corollary follows by applying the previous theorem to the
partial sums S"(x) of the series fk(x). Q
8.2.5 THEOREM Suppose the sequence (f.} of real-valued functions on the set E converges
pointwise to f on E. For each n E N, set
p
M. sup
I fn(x) - ! (x) I I.
Then I f.1 converges uniformly to f on E if and only if lim M. = 0.

n-oo
8.2.6 EXAMPLES
(a) To illustrate the previous theorem we consider the sequence
S"(x) = nxe-"XI, n = 1, 2, .. .
of Example 8.2.2(b). For this sequence, lim S"(x) = 0 for all x, 0 s x < oo. However,
n
M. = sup S"(x)
xE[0,oo)
which diverges to oo. Thus the convergence is not uniform on (0, oo). However, the se-
quence {S"} does converge uniformly to the zero function on [a, oo) for every a > 0
(Exercise 6).
(b) Consider the sequence { f} of Example 8.2-2(a) given by f"(x) = x", x E [0,1 ].
This sequence converges pointwise to the function f (x) = 0, 0 < x < 1, and f (l) = 1.
Since
Jx", 0x :S x < i,
I.f"(x) -f(x)I = l0, x= = 1,
we have
Mn = SU Ifnx) -f(x)I = 1.
:E[0, Ij
Thus since {M"} does not converge to zero, the sequence { f"} does not converge uni-
formly to f on [0, 1 ]. On the other hand, if 0 < a < I is fixed, then
Mn I fn(x)I = a".
xE[ Dal
Since lim an = 0, by Theorem 8.2.5 the sequence { fn} converges to zero uniformly on
n-400
[0, a] for every fixed a, 0 < a < 1.
The Welerstrass M-Test

The following theorem of Weierstrass provides a very useful test for uniform conver-
gence of a series of functions.
8.2.7 THEOREM (Weierstrass M-Test) Suppose { fk} is a sequence of rral-valued func-

tions defined on a set E, and {Mk} is a sequence of real numbers satisfying
k(x) L Mk for all x E E and k E F.

00 00
If Y, Mk < oo, then 7, k(x) converges uniformly and absolutely on E.
k-1 k-1
Pro of. Let S. (x) = Zk.I fk(x). Then for n > m,
n n n
ISn(x) - sm(x)l = Ik fk(x) IJi(X) < ± Mk

m+l k=m+1 k-m+I
Uniform convergence now follows by the Cauchy criterion. That 7, I fk(x) I also con-
verges is clear.
8.2.8 EXAMPLES
.(a) If Zak converges absolutely, then since Iak cos kxl s IakI for all x E R, by the
Weierstrass M-test, the series I ak cos kx converges uniformly on R. Similarly for the
series I ak sin kx. In particular, the series
00
cos kx sin kx
7
1
p>
k-1 kr k=1 k°
converge uniformly on R.
(b) Consider the series 7,I (x12)k. This is a geometric series that converges for all
x E R satisfying I x I< 2. If 0< a < 2 and I x s a, then
(2)k.
\2)k I
Since a/2 < 1 the series 7, (a/2)k converges. Thus by the Weierstrass M-test, the se-
ries I', (x/2)k converges uniformly on [-a, a] for any a, 0 < a < 2. The series,
however, does not converge uniformly on (-2, 2) (Exercise 11).
Although the Weierstrass M-test automatically implies absolute convergence, the
following example shows that uniform convergence as a general rule does not imply ab-
solute convergence.
8.2.9 EXAMPLE Consider the series

00
T, (-1)k+1 0 - x c 1.
k-1 k
For each k E N, set ak(x) = xk/k. For x E [0, 11, we have
a,(x) z a2(x) at ... ? 0 and li-m ak(x) = 0.
Thus by Theorem 7.2.3, the series I (-I)k "ak(x) converges for all x E [0, 1].
Let
S(x) _ (-1Y+'ak(x).
k=1
If S (x) is the nth partial sum of the series, then by Theorem 7.2.4
IS(x) - 1(x) s n + 1 for all x E (0. 1 ].
Thus converges uniformly to S on [0, 1 ]. However, the given series does not con-
verge absolutely when x = 1.
The converse is also false; absolute convergence need not imply uniform conver-
gence! As an example, consider the series Y,1 x2(1 + x2)-k of Example 8.1.2(b).
Since all the terms are nonnegative, the series converges absolutely to
1 0, x = 0,
f (x) = l + x2, x # 0,
on R. However, as a consequence of Corollary 8.3.2 of the next section, since f is not
continuous at 0, the convergence cannot be uniform on any interval containing 0.
The series I (- 1)t+ txk/k, x E [0, 1]. also provides an example of a series that
converges uniformly on [0, 1 ] but for which the Weierstrass M-Test fails.
EXERCISES 8.2
2. a. If (f.) and converge uniformly on a set E, prove that {f. + converges uniformly on E.
+b. If (f.1 and converge uniformly on a set E, and there exist constants M and N such that I f.(x)I <- M and
1g.(x)l s N for all n E N and all x r= E, prove that { converges uniformly on E.
c. Find examples of sequences f f.) and that converge uniformly on a set E, but for which { f does not
converge uniformly on E.
3. Show that if (f.1 converges uniformly on (a, b) and (f. (a)} and { ff (b)} converge, then (f.) converges uniformly
on (a, b].
4. Let f .(x) = nx(1 - x2r, 0 5 x 5 1. Show that does not converge uniformly to 0 on (0, 1
5. Let 0 s x 5 I.
x"'
a. Show that (f.} converges uniformly to 0 on [0, a] for any a, 0 < a < 1.
b. Does I f.1 converge uniformly on [0, fl?
6. Show that the sequence {nxe converges uniformly to 0 on [a. no) for every a > 0.
7. For each n E N. set
x
x + - sin nx, x r= R.
n
Show that the sequence { converges uniformly to f(x) = x for all x E (-a, a], a > 0. Does (f.1 converge uni-
formly to f on R?
8. Show that each of the following series converge uniformly on the indicated interval.
Ca. ; k-
00
k-1
,
+ xk-1
0 5 x < oc b, i e-k`xt, 0 5 x< oo
00 00 (-l)k*1
"c. Y, 1 5 x < on d. , 0 5 x< on
k-1 k-, k+x
9. Test each.of the following series for uniform convergence on the indicated interval.
sin 2kx 00 xk
'a.
00
x E G8 b F, x5I
(2k + 1 ?/2 ' k(In k)2 '
*c' k
1
(kx+2(k+1)x+2
1
05x51
00 (-1)k+lx2k+l
(xi<_1 e.
00
//\
Xsinikol, p> 1.JxIs2
d. k=O 2k+1
10. Show that each of the following series converge uniformly on (a, oo) for any a > 0, but do not converge uni-
formly on (0, oo).
M I I
'a. 2 I + k2x
00
b. I ii-+;
11. Show that the series E, , (x/2)k does not converge uniformly on (-2, 2).
12. If Jk o ak converges absolutely, prove that I-k o akxk converges uniformly on [ -1. I ].
13. If I .oak converges, prove that Iku akxk converges uniformly on [0, 1].
14. Let {ck) be a sequence of real numbers satisfying 7, Ickl < on, and let {xk} be a countable subset of (a, b]. Prove
that the' series lk , ck !(x - xk) converges uniformly on [a, b]. Here ! is the unit jump function defined in 4.4.9.
15. Dirichlet'17est for Uniform Convergence: Suppose {)k) and {gk} are sequences of functions on a set E satisfying
(a) the partial sums Ik., gk(x) are uniformly bounded on E; i.e., there exists M > 0 such that
<Mforalln E NlandxEE,
(b) k(x) >- fk I(x) > 0 for all k E hi and x E E, and
(c) jim fk(x) = 0 uniformly on E.
Prove that I fk(x)gk(x) converges uniformly on E.
16. Prove that
. sin kx . t os kz (p > 0)
k=1 kP k-1
converge uniformly on any closed interval that does not contain an integer multiple of 21r.
17. Define a sequence of functions If.) on [0, I ] by
if ii < x `
f(x)= 2" 2"
0, elsewhere.
Prove that 1 1 fn(x) converges uniformly on [0, 1 ], but that the Weierstrass M-test fails.
18. 'Let F0 be a bounded Riemann integrable function on [0. I ]. For n E 101, define F .(x) on [0, 1 ] by
F (x) = fo F.- 1(t) dt. Prove that 7,; o Fk(x) converges uniformly on (0. 1 ].
8.31 Uniform Convergence and Continuity

In this section we will prove that the limit of a uniformly convergent sequence of con-
tinuous functions is again continuous. Prior to proving this result, we first prove a
stronger result that will have additional applications later.
8.3.1 THEOREM Suppose is a sequence of real-valued functions that converges uni-

formly to a function f on a subset E of R. Let p be a limit point of E, and suppose that
for each n E N,
lim fn(x) = A,,.
Then the sequence {An} converges and

lim f(x) = lim A,,.
x-.p n-+co
Remark, The last statement can be rewritten as/
lim I lim
X_p \n-iW /I = lim I lim
n-.00 \X- p
It should be noted that p is not required to be a point of E; only a limit point of E.

Proof. Let e > 0 be given. Since the sequence I f.1 converges uniformly to f on E.
there exists a positive integer n such that
Ifn(x) - fm(x)I < E (2)
for all n, m _a n and all x E E. Since inequality (2) holds for all x E E. letting x -+p
gives
IAn - Amt s e for all n, m ? no.
Thus {An} is a Cauchy sequence in R, which as a consequence of Theorem 2.6.4 con-

verges. Let A= nti00
limA,
It remains to be shown that lim f(x) = A. Again, let e > 0 be given. First, by the
uniform convergence of the sequence { fn(x)} and the convergence of the sequence
there exists a positive integer m such that
1f(X) - fm(x)I < 3
for all x E E, and that
IA - Aml < 3.
Since lim fm(x) = Am, there exists a S > 0 such that
lfm(x) - Anl < 3 for all x E E, 0 <lx - pl < 8.
By the triangle inequality.

If(x) - At < If(x) -fm(x)I + Ifm(x) - Aml + IAm - Al
< e + I fm(x) - A, I
Thus if x E E with 0 < Ix - pi < 8,

If(x) - Al < E;
i.e., lim f (x) = A. []
8.3.2 COROLLARY Let E be a subset of R.

(a) If { is a sequence of continuous real-valued functions on E. and if {f.)
converges uniformly to f on E, then f is continuous on E.
(b) If { fn} is a sequence of continuous real-valued functions on E, and if I ,f
converges uniformly on E, then
S(x) _ 71 fn(x)
M=1
is continuous on E.
Proof. (a) If p E E is an isolated point, then f is automatically continuous at p. If
p E E is a limit point of E, then since f is continuous for each n E N,
Jim MX) = MP).
Thus by the previous theorem,
m f(x) = urn fn(P) = f(p).
s n-.oo
Therefore, f is continuous at p.
For the proof of (b), let
SS(x) = fk(X)
krl
Then for each n E N, S. is continuous on E. Since converges uniformly to S on E,

by part (a) S is also continuous on E.
8.3.3 EXAMPLE The sequence {xn},.,, x E [0, 1 ], of Example 8.1.2(a) does not converge
uniformly on [0, 1 ] since the limit function
o x < 1,
}.(x) =
J0,
1, x= 1,
is not continuous on [0, 11. Likewise, the series
00
( 1 __ 0, x = 0.
k ,1 + x2 Jk 1 + x2, X # 0,
of Example 8.1.2(b) cannot converge uniformly on any interval containing 0, since the
sum of the series is not continuous at 0. 0
Dini's Theorem'
A natural question to ask is whether the converse of Corollary 8.3.2 is true. Namely, if
f and f. are continuous for all it and f -+f pointwise, is the convergence necessarily
uniform? The following example shows that this need not be the case. However, in The-
orem 8.3.5 we will prove that with the additional assumption that the sequence { f (x)}
is monotone for all x, the convergence is indeed uniform.
8.3.4 EXAMPLE As in Example 8.2.2(b), for each n e N, let

S (x) = nxe-""s, x E [0. 1 ].
Then S is continuous on (0, 1 ] for each n, and lim SS(x) = S(x) = 0, which is also
"-OOD
continuous. However, since
max
Osssl S (x) = n ,
by Theorem 8,2.5 the convergence cannot be uniform.
8.3.5. THEOREM (Dins) Suppose K is a compact subset of R and (f.} is a sequence of con-
tinuous real-valued functions on K satisfying the following:
(a) {f,,) converges pointwise on K to a continuous function f, and
(b) f (x) ? f,,. 1(x) for all x E K and n E N.
Then converges uniformly to f on K.
1. This topic is not required in subsequent sections and thus can be omitted on first reading.
Proof. For each n E N let g = f - f. Then g is continuous on K, g (x)

gn+i(x) ? 0 for all x E K and all n E N, and
lim g (x) = 0 for each x E K.
Let e > 0 be given. For each n E Nl, let
K p be a limit point of K,,. By Theorem 2.4.7 there ex-

ists a sequence {xk} in K,, that converges to p. Since g is continuous and g (xk) >- a for
allkEN,
gn(P) = t m ga(xk) y E.
.Therefore, p E K,,. Thus K,, is closed, and as a consequence of Theorem 3.2.5, also
compact. Furthermore, since g (x) >_ g..#) for all x E K,
K,,. i C K,, foralln.
Finally, since g,,(x) -+ 0 for each x E K,
00
n-i
However, by Theorem 3.2.7 this can only be the case if K,,, = 4 for some n E N. Thus
for all n ? no,
for allxEK.
Therefore the sequence {g,} converges uniformly to 0 on K. Q
8.3.6 EXAMPLE We now provide an example to show that compactness is required. For
each n E N, set
I
nx+ V
0<x<l.
Then { f,(x)} monotonically decreases to f (x) = 0 foreach x E (0, 1). However, since
linof,,(x) = 1, by Theorem 8.3.1 the convergence cannot be uniform.
The Space `C[q b12

We conclude this section with a brief discussion of the space `t;[a, b) of all continuous
real-valued functions on the closed and bounded interval [a, b]. If f, g E 1.9[a, b] and
c E R, then by Theorem 4.2.3 the functions f + g and cf are also continuous on [a, b].
Thus c4[a, b] is a vector, space over R where for the zero element we take the constant
function 0; that is, the function given by f(x) = 0 for all x E [a, b]. We define a norm
on %[a, b] as follows.
2. This topic can be omitted on first reading. The concept of a complete normed linear space (Definition
8.3.10) is only required in Section 10.8.
8.3.7 DEFINITION For each f E'g[a, b], set

IIf(I. = max{l f(.r)I : x E [a, b]}.
The quantity II f M is called the uniform norm off on [a, b].
That II 11. is indeed a norm on b] is left to the exercises (Exercise 12). We now
take a closer look at uniform convergence of a sequence of continuous real-valued func-
tions and introduce the concept of convergence in norm. We first note that a sequence
of continuous real-valued functions on [a, b] is nothing but a sequence in the set
%[a, b]. Suppose that the sequence { in `P[a, b] converges uniformly to f on (a. b].
Then by Corollary 8.3.2, the function f E'F[a, b]. By the definition of uniform con-
vergence, given c > 0, there exists a positive integer n,, such that
I fn(x) - Ax) I < E
for all x E [a, b] and all n ? n,,. This is illustrated in Figure 8.2. Since f - f is con-
tinuous, if n z no, by Corollary 4.2.9
IIf,, - fllu = max{ I f0(x) - f(x)l : x E [a, b]} < e.
Therefore, 11f. - f Il u < e for all n ? no.
Conversely, suppose f, f E 'P[a, b] satisfy the following: For each e > 0 there ex-
ists a positive integer n,, such that N f - f j. < e for all n ? n0. But then If(x) - f (x)I
< E for all x E [a, b] and all n ? n0; i.e., { f on [a, b]. This
proves the following theorem.
8.3.8 THEOREM A sequence { f E %(a, b) if and only

if given e > 0, there exists na E N such that IIf - f,11. < e for all n ? no.
Using the above as motivation, we define convergence in a normed linear space as
follows.
8.3.9 DEFINITION Let (X, 11 11) be a normed linear space. A sequence in X converges
in norm if there exists x E X such that for every e > 0, there exists a positive integer
ne such that IIx - x.11 < e for all n ? no. If this is the case, we say that converges
in norm to x, and denote this by x x as n -+ oa.
i
From the definition it is clear that a sequence in X converges in norm to x E X
if and only if lim 11x - x11 = 0. Also, as in the proof of Theorem 2.1.10, if con-
verges in norm "&n its limit is unique. Using the norm it is also possible to define what
we mean by a Cauchy sequence in a normed linear space.
8.3.10 DEFINITION
(a) A sequence {x} in a normed linear space (X, ll h) is a Cauchy sequence if for
every e > 0, there exists a positive integer n, such that
Ilxn - xmll < E
for all integers n, m ? n,,.
(b) A normed linear space (X,11 II) is complete if every Cauchy sequence in X
converges in norm to an element of X.
As for sequences of real numbers, every sequence {x,,} in X that converges in norm
to x E X is a Cauchy sequence. In Theorem 2.6.4 we proved that the normed linear space
(R,I I) is complete. The following theorem proves that (%[a, b], II ll.) is also complete.
8.3.11 THEOREM The normed linear space (`P[a, b], II II.) is complete.
Proof. Let {fn} be a Cauchy sequence in %[a, b]; i.e., given e > 0. there exists a
positive integer no such that Ilfn - fmll. < E for all n,m >- no. But then
Ifn(x) - fm(x)I Il fn - fmllu < E
for all x E [a, b] and all n, m a n,. Thus by Theorem 8.2.3 and Corollary 8.3.2, the se-
quence { fn} converges uniformly to a continuous function f on [a, b]. Finally. since the
convergence is uniform, given e > 0, there exists an integer no such that
I fn(x) - f(x)l < E for all x E [a, b] and n >- n,.

As a consequence, we have 1Ifn - f 1j. < E for all n a no. Therefore, the sequence { fn}
converges to f in the norm 11 L. Q
Contraction Mappings
In Exercise 13 of Section 4.3 we defined the notion of a contractive function on a sub-
set E of R. We now extend this to normed linear spaces.
8.3.12 DEFINITION Let (X, N II) be a nonmed linear space. A mapping (function) T : X -+X
is called a contraction mapping (function) if there exists a constant c, 0 < c < 1.
such that
IIT(x) - T(y)iI < cllx - yll

for all x, y E X.
Clearly every contraction mapping on X is continuous, in fact uniformly continu-

ous on X. As in Exercise 13 of Section 4.3, we now prove that if T is a contraction map-
ping on a complete normed linear space (X, II 1). then T has a unique fixed point in X.
8.3.13 THEOREM Let (X, II II) be a complete normed linear space and let T : X -+ X be a
contraction mapping. Then there exists a unique point x E X such that T(x) = x.
Proof. Suppose T : X - X satisfies 11T (x) - T(y)p <- cll x - yq for fixed c,
0 < c < 1. and all x, y e X. We now define a sequence {xn} in X as follows: Let x, E X
be arbitrary. For n E N set xn = T(xn_,). That is, x, = T(x,), x2 = T(xi), etc. Since
IIxn+1 - xnll = lIT(xn) - T(xn-1)II cllxn - xn-111,
the sequence {xn} is a contractive sequence in X (see Definition 2.6.6). An argument
similar to the one used for a contractive sequence in Section 2.6 shows that
C.
Ilxn+m - xnll C 1 C Ilx, - xoll
for all n, m E N. The details are left as an exercise (Exercise 10). Since c" -* 0. the se-
quence {x"} is a Cauchy sequence in X. By completeness, the sequence {x"} converges
(in the norm) to some x E X. But by continuity of the mapping T,
x = lim x"+t = lim T(x") = T(x);
i.e., T(x) = x. Suppose y E X also satisfies T(v) = y. But then

Ily - Al = IIT(y) - T(x)ll cliv - xll.
Since 0 < c < 1, the above can be true if and only if Ilv - xii = 0; that is. v = x.
Thus x is unique.
EXERCISES 8.3
1. *Show that the series I'U x (1 - x)0 cannot converge uniformly for 0 s x 5 1.
2. For n E N. let f .(x) = x"/(I + x"), x (5 [0, 1 ]. Prove that the sequence f f.) does not converge uniformly on
[0, 1].
3. Give an example of a sequence f f.) of functions on [0, 1]. such that each f" is not continuous at any point of
[0, 11, but for which the sequence {f.) converges uniformly to a continuous function f on [0. 11.
4. *Suppose that f is uniformly continuous on R. For each n E N. set f"(x) = f(x + Prove that the sequence ff.)
converges uniformly to f on R.
S. Let (f.1 be a sequence of continuous real-valued functions that converges uniformly to a function f on a set
E C R. Prove that liim f"(x") = f(x) for every sequence {x"} C E such that x" -*x E E.
6. *Let E C R and let D be a dense subset of E. If (f.} is a sequence of continuous real-valued functions on E. and
if if.) converges uniformly on D. prove that (f.) converges uniformly on E. (Recall that D is dense in E if every
point of E is either a point of D or a limit point of D.)
7. Find a sequence {f"} in IC[0, I ] with II f.11. = I such that no subsequence of (f,) converges (in norm) in %[0. C.
8. Suppose {f.1 is a sequence of continuous functions on [a, h] that converges uniformly on [a, b]. For each
x E [a, b], set g(x) = sup{f" (x)}.
a. Prove that g is continuous on [a, b].
b. Show by example that the conclusion may be false if the sequence {f"} converges only pointwise on [a. b].
9. For each n E N and x E It, set f,(x) = (1 + )". Use Dini's theorem to prove that the sequence {f"} converges
uniformly to e' on [a, b] for any fixed a, b E R.
10. Let (X, II II) be a normed linear space and let T : X -* X be a contraction mapping with constant c, 0 < c < 1. If
{x,} is the sequence in X as defined in the proof of Theorem 8.3.13, prove that
11x... - x,, lI S (c"/(1 - c)) II x, - x.11 for all n, m E N.
11. Define T : [0, 1 ] -a %[0, 1 ] by (Tto)(x) = fo cp(t) dt, 0 S x S 1, v E %[0, 1 ], and set T2 = ToT.
a. Prove that I (T2V)(x)I s z X211011..
b. Show that T2 is a contraction mapping on `P[0, I ] and thus has a fixed point in `P[0. 1 ].
c. Prove that T has a fixed point in `4[0, 1 ].
12. Prove that b], f Q is a normed linear space.
13. Prove that (12, Ii 02) is a complete normed linear space.
8.4 Uniform Convergence and Integration 337
-8.4 Uniform Convergence and Integration

In this section we will prove that uniform convergence of a sequence If.) of Riemann
integrable functions is again sufficient for the limit function f to be Riemann integrable,
and for convergence of the definite integrals of fn to the definite integral off The anal-
ogous result for the Riemann-Stieltjes integral is left to the exercises (Exercise 2).
8.4.1 THEOREM Suppose fn E 9t[a, b] for all n E N. and suppose that the sequence {fn}
converges uniformly to f on [a, b]. Then f E gt[a, b] and
b Jb
f(x) dx = li m
f
a
n-M
a
fn(x)dx.
Proof. For each n E lkU, set

en =sEnlax,[!x)
a. bj V.
-f(x)I.
lim e. = 0. Also, for all

Since fn --).f uniformly on [a, b]. by Theorem 8.2.5, n-x
X E [Q, b],
fn(x) - En 5 f(x) t fn(x) + En.
Hence
b
rb
(fn-En) :
fb
f ff: a
Vn+En) (3)
Therefore,
0-
fa
b
f- rb
Ja
f<2En[b-a].
Since en -> 0, f E &[a, b]. Also by inequality (3),
rh rh
f(x) dx - Ja fn(x)dxI G En[b - Q],
Ja
and thus
b b
hm
fa
fn(x) dx = j f(x) dx.
a
8.4.2 COROLLARY /f fk E k[a, b] for all k E N, and if

00
f(x) fk(x), x E [a, b],

k-l
where the series converges uniformly on [a, b], then f E Jt[a, b] and
b m b
Ja
f(x) dx = 7, f A (x) dx.
k.1 a
Proof. Apply the previous theorem to S"(x) = 1,"t. I fk(x), which by Theorem 6.2.1
is integrable for each n E N. O
Although uniform convergence is sufficient for the conclusion of Theorem 8.4.1, it
is not necessary. For example, if f"(x) = x", x E [0, 1 ], then {f"} converges pointwise,
but not uniformly, to the function
J0, 0 : 5x<1 ,
f(x) = 1, x = 1.
The function f E 9t[0, 1 ] and
°n+ 1
=0= f f(x) dx.
In Section 10.6, using results from the Lebesgue theory of integration, we will be able
to prove a stronger convergence result that does not require uniform convergence of the
sequence If.}. However, it does require that the limit function f is Riemann integrable.
For completeness we include a statement of that result at this point.
8.4.3 THEOREM (Bounded Convergence Theorem) Suppose f and fn, n E N, are Rie-
mann integrable functions on (a, b] with lim f"(x) = f(x) for all x E [a, b]. Suppose
n-.oo
also that there exists a positive constant M such that [fn(x) I < M for all x E [a, b] and
allnE N. Then
lim f"(x) dx = f(x) dx.

n-iao ab Jab
It is easily checked that the sequence If.} on (0, 1], where for each n E N
fn(x) = x", satisfies the hypothesis of the prwtvious theorem. Also, since the limit func-
tion f is continuous except at x = 1, f E J.[0, 1].
EXERCISES 8.4
1. *If I I ak l < oo, prove that
f (akx*)dx=
0 7 I , k+1
ak
2. Let a be a monotone increasing function on [a, b]. Suppose f, E A(a) on [a, b) for all n E R;. and suppose that
the sequence converges uniformly to f on :a. b,. Then f E .R(o) and
(h (
f f da = lim f. f da.
3. For each n E N. let nx/(l + nx), x E [0, I ] . Show that the sequence { f } converges pointwise. but
not uniformly, to an integrable function f on i 0, 1 ]. and that lim f,' f;,(x) dx = fol f(.r) dr.
4. *If f is Riemann integrable on [0, 11. use the bounded convergence theorem to prove that Iim f r f(x) dr = 0.
5. Let (f,} be a sequence in :7i[a, b] that converges uniformly to f E .R[a, b'. For n E ) set F,(x) = f' f , and let
F(x) = f f, r E [a, b]. Prove that converges uniformly to F on [a, bl.
6. Suppose f : [0, I ] R is continuous. Prove that Iim f f(r") dr = f (0).
7. *Let {r"} be an enumeration of the rational numbers in [0, 11, and let f : [0, 11 -- R be defined by
Y I
f(x) = I Y2I (x - ri).

where / is the unit jump function of Definition 4.4.9. Prove that f E Jt [0. I
8. Define g on R by g(x) = x - [x]. where [x] denotes the greatest integer function. Prove that the function
K(nr)
f(x)
is Riemann integrable on [0, 1 ] and find f f(x) dx. This function was given by Riemann as an example of a
function that is not integrable according to Cauchy's definition.)
9. *Let g E R[a, b] and let {f} be a sequence of Riemann integrable functions on [a, b] that converges uniformly
to f on [n, b]. Prove that Iim f ,".ff, g = f b fg.
10. *Let g be a nonnegative real-valued function on ;0. oc) for which fo g(x) dr is finite. Suppose if,) is a se-
quence of real-valued functions on (0, oo) satisfying f;, E dt[0, c] for every c > 0 and ! f (x) <_ g(x) for all
x E [0. oo) and n E N. If the sequence (f.1 converges uniformly to f on [0. c] for every c > 0, prove that
liym fo .f (x) dx = fo .f(.r) Jr.
11. For 1fII = f,°1f(x)Idx.
a. Prove that (%[a, b], II J,) is a normed linear space.
*b. Show by example that the normed linear space (E[a. b], III ) is not complete.
i Uniform Convergence and Differentiation

In this section we consider the question of interchange of limits and differentiation.
Example 8.1.2(e) shows that even if the sequence (f J converges uniformly to f, this
is not sufficient for convergence of the sequence ff.'} of derivatives. Example 8.5.3
will further demonstrate dramatically the failure of the interchange of limits and dif-
ferentiation. There we will give an example of a series, each of whose terms has
derivatives of all orders, that converges uniformly to a continuous function f. but
for which f fails to exist at every point of R. Clearly, uniform convergence of the
sequence (f.1 is not sufficient. What is required is uniform convergence of the se-
quence If.}.
8.5.1 THEOREM Suppose is a sequence of differentiable functions on [a, b]. If
(a) {f,,) converges uniformly on [a, b], and

(b) { f (x,)} converges for some x E [a, b],
then If.} converges uniformly to a function f on [a, b], with
f'(x) = Jim f;,(x).

Remarks
(a) Convergence of { f (x,)} at some x, E [a, b] is required. For example, if we let
f (x) + n, then g"(x) = f"(x), but need not converge for any x E [a. b].
In Exercise I you will be asked to show that uniform convergence of (f.'} is also re-
quired; pointwise convergence is not sufficient.
(b) If, in addition to the hypotheses, we assume that f" is continuous on [a, b], then a
much shorter and easier proof can be provided using the fundamental theorem of cal-
culus. Since f" is continuous, by Theorem 6.3.2
f(x) = fn(xo) + Jf.(t) dt
I.X
for all x e [a, b]. The result can now be proved using Corollary 8.3.2 and Theorem
8.4.1. The details are left to the exercises (Exercise 2).
Proof. Let e > 0 be given. Since { f (x,)} converges and If.') converges uniformly,
time exists no E RI such that
fn(x,,) - fm(xo) I < 2 for all n, m z n (4)
and
a
I f"(t) - f",(t) I < 2(b a) for all t r= [a, b] and all n, m ? n,. (5)
Apply the mean value theorem to the functions f - fm with n, m ? n, fixed. Then for
x, y E [a, b], there exists t between x and y such that
I(ff(X) - f,,(X)) - (My) - fm&))I = I [f"(t) - f;'(t)](x - A.

Thus by inequality (5),
I(f (X) - fm(X)) - (AY) - fm&))I 2(b a)I X - yl < 2 (6)
Take y = x, in inequality (6). Then by inequalities (4) and (6). for all x E [a, b] and
n, ma n
If,(x) - fm(X) I I MW - fm(x)) - (ff(x0) - fm(X0))I + I f.(xo) - fm(X0) I
<2E +2E =E.

Hence by Theorem 8.2.3, the sequence (f,) converges uniformly on [a. b). Let
f(x) = .limff(x)
It remains to be shown that f is differentiable, and that
f'(x) = hm f;,(x)
for all x E [a, b]. Fix p E [a, b], and fort # p, t E [a, b], define
8n(t) = fn(t) - fe(P)' A() = f(t) - f(P)

t-p t- p
Then g(t) each t e [a, b], t * p, and for each n
1 m g.(t) = fe(P)
Let E = [a, b] \ {p}. Take y = p in inequality (6). Then for all t E E.
a
gm(t) I <_ 2(b for all n, m z no.
a)
Therefore, converges uniformly to g on E. Hence by Theorem 8.3.1,
f'(P) _r-P
lim g(t) =f.'(p) O
8.5.2 EXAMPLE To illustrate the previous theorem, consider the series
k-i .
Since 12-k sin kxl <_ 2-'for all x E R, by the Weierstrass M-test this series converges
uniformly to a function S on R. For n E N, let
sin kx
S, (x) =k,`
7 2
Then
S.'(X) k cos kx
k=1 2k
Since E k2 -k converges, by the Weierstrass M-test the sequence {S,,} converges uni-
formly on R. Thus by Theorem 8.5.1,
k cos kx
S'(x) = lim S; ,(x) _ 2k
A Continuous Nowhere Differentiable Function

We conclude this section with the following example of Weierstrass of a continuous
function that is nowhere differentiable. When this example first appeared in 1874, it as-
tounded the mathematical community.
342 Chapter8 Sequences and Series of Functions
8.5.3 EXAMPLE Consider the function f on R defined by

00 Cos akirx
f(x) = 2k
, (7)
k-0
where a is an odd positive integer satisfying

a>31r+2.
Since
cos ak9rx 1
2k 2k,
the series (7) converges uniformly on R, and hence f is continuous. The graphs of the
partial sums (with a = 13) Si(x) = cos(wx), S2(x) = Si(x) + 1 cos(13ax), and S3(x) _
S2(x) + 2, cos(1321rx) are illustrated in Figure 8.4.
2 2
S2
1
r
-1
Figure 8.4
We prove that f is nowhere differentiable by showing that for each x E R, there ex-
ists a sequence h -* 0 such that
f(x + hn) - f(x)
= 00.
1im I
11i00 hn
For n E N, set
Cos ak?rx
A-0 2'
cos aklrx
RR(x) = 0
k=n 2k
Then for h > 0,

f(x+h)-f(x)-SR(x+h)-SR(x)+RR(x+h)-RR(x)
It It It
Our first step will be to estimate

S"(x + h) - S"(x)
h
from above. By the mean value theorem,
Cos ak?r(X + h) - cos ak7rx k

- = -a it sln[a 1r(x + Q,
h
for some , 0 < < h. Since
y 15 aklr,
I -ak it sm[aklr(x + b)]
we obtain
S"(x + h) - S"(x) cos akvr(x + h) - cos akirx

It
I`2kI h
(8)
"-' (a)* __ n[l - (2)"]

:5 IT
k-o 2 1- 2
27r a
<a-2(2T*
We now proceed to obtain a lower estimate on the term involving R". To do so we
write
a"x = k" + B",
where k" is an integer, and -2 5 B" < . Set
h" -
1-B"
am
Since -'-z s B" < '-z, we have 2 >- 1 - B" > '-i. Therefore,
2 a" : < 2a". (9)

h"
Fork? n,
akir(X + h") = a-"a"ir(x + h")
ak_"ir(a"x
= + ak "Tr(k" + 1).
Since ak-" is odd and k" is an integer,
cos[akir(x + h")] = cos[ak-"ir(k" + 1)]
Also, akirx = ak-"a"Trx = ak-"ir(k,, + B"). Using the trigonometric identity
cos(A + B) = cos A cos B - sin A sin B,

and the fact that sin(k"ak-" ir) = 0, we have

cos aklrx = cos(a "k"1r) cos(ak "S"1r)
= (-I )k- cos(a-"S"1r).
Therefore,
cos a kir(x + h") - Cos akirx (- 1)k- cos(ak -"S"1r)

+ cos(a k- "5.1r)).
As a consequence,
1 R"(x + h") - R"(x) I - I °° I cos aklr(x + h - cos a`1rx

h,, k-1 2k h"
. (- I)t(. [ I + cos ak - "S"1r, I

k=" h" 2k
1 °° I + Cos ak "S"1r 1 1 - Cos 8"1r

2k
=h" k-n h" 2"
Since -Z s S" < Z, cos 8"1r 2t 0. Therefore, by inequality (9) and the above,
R"(x+h,,)-R"(x) " 1 3(21". (10)
Using the reverse triangle inequality, I a + b I ? I a I - I b I, we have

f (x + h") - f (x) I R& + h") - R"(x) (- I S"& x + h") - S"(x)
h" h" E h"
which by inequalities (8) and (10),
>2 a - 21r a
3 l2 a-2 2
2
(
\2J"[3
Since i > 1, we obtain
f(x + h") - f(x)
h f - 0o as n -+ oo,
n
provided a is an odd positive integer satisfying

a 21r 2
>0;
3
i.e., a > 31r + 2. Since 1r < 3.15, we need a z 13. N

Remark. The above proof is based on the proof of a more general result given in the
text by E. Hewitt and K. Stromberg. There it is proved (Theorem 17.7) that
f( x) = ' )°C cos akrrx

k=0
bk
has the desired property if a is an odd positive integer, and b is any real number with
b > I satisfying
6> 1+3ir.
The above function was carefully examined by G. H. Hardy [Trans. Amer. Math. Soc.,
17, 301-325 (1916)] who proved that the above f has the stated properties provided
1 :<.b s a.
These are by no means the only examples of such functions. A slightly easier con-
struction of a continuous function that is.nowhere differentiable is given in Exercise 7.
EXERCISES 8.5
1. For n E N, set f(x) = [0, 1 that the sequence converges on [0,
that f on that converge to
a on [a, b] for [a, b] n
converges on and for the of
that f on and that f'(x) = li m for all
xE
o a of real numbers satisfying I kjak I < oo. Show that the series Yk o akxk converges uni-
formly to a function f on jx] 5 1 and that f'(x) = joko-, kakxk- ' for all x, ] x 1 s 1.
4. *Let If.) be a sequence of differentiable real-valued functions on (a, b) that converges pointwise to a function f on
(a, b). Suppose the sequence {f;,} converges uniformly on every compact subset of (a, b). Prove that f is differen-
tiable on (a, b) and that f'(x) = lim f;,(x) for all x E (a, b).
S. State and prove an analogue of Theorem 8.5.1 for a series of functions 7, fk(x).
6. Show that each of the following series converge on the indicated interval and that the derivative of the sum can be
obtained by term-by-term differentiation of the series.
--1i, x E (O' OD)

b. a x E (0, oo)
M 0o xk
C. 2xk, ]x]<l d. k, x E (-oo, oo)
k=0
7. This exercise provides another construction of a continuous function f on R which is nowhere differentiable. Set
g(x) = x {, = 15 x is 1, and extend g to R to be periodic of period 2 by setting g(x + 2) = g(x). Define f on
R by
f(x) _ k=o(31
4
g(4kx).
a. Prove that f is continuous on R.

b. Fix xo E R and m E N. Set 8," Z4-'", where the sign is chosen so that no integer lies between 4'"x" and
4'"(xe + 8,"). Show that g(4"(x" + 8,.)) - 0 for all n > m.
c. Show that
s) - f (X.) > 2 (3"' + 1).
f (x° + m
Since 8," -+ 0 as m - oo, it now follows that f'(.t") does not exist.
8.6 The Weierstrass Approximation Theorem

In this section we will prove the following well-known theorem of Weierstrass.
8.6.1 THEOREM (Weierstrass) If f is a continuous real-valued function on [a, b], then

given e > 0, there exists a polynominal P such that
11(x) - P(x) I < E
for all x E [a, b].
An equivalent version, and what we will actually prove, is the following:
1f f is a continuous real-valued function on [a, b], then there exists a sequence {P"}
of polynomialf such that
f (x) = lim P"(x) uniformly on [a, b].
n- oc
Before we prove Theorem 8.6.1, we state and prove a more fundamental result that
will also have applications later. Prior to doing so, we need the following definitions.
8.6.2 DEFINITION A real-valued function f on R is periodic with period p if

f (x + p) = f (x) for all x r= R.
The canonical examples of periodic functions are the functions sin x and cos x, both
of which are periodic of period 2ir. The graph of a periodic function of period p is il-
lustrated in Figure 8.5. The graphs of a periodic function of period p on any two suc-
cessive intervals of length p are identical. It is clear that if f is periodic of period p, then
f(x+kp) =f(x) forallkEZ.
Another useful property of periodic functions follows.
8.6.3 THEOREM 1f f is periodic of period p and Riemann integrable on [0, p), then f is Rie-
mann integrable on [a, a + p] for every a E R. and
+P
f° f(x) dx = J r f (x) dx.
c J0
Proof. Exercise 2.
Figure 8.5 Graph of a Periodic Function of Period p
Approximate Identities
8.6A DEFINITION A sequence {Qn} of nonnegative Riemann iruegrable functions on
[-a, a] satisfying
(a) a QQ(t) dt = 1, and

J
(b) lim
n-roo
J dt = 0 for every S > 0,
{dsiro
is called an approximate Identity on [ -a, a].
An approximate identity {Qn} is sometimes also referred to as a Dirac sequence.
Remark. In (b), by the integral over the set {8 s I t 1} we mean the integral over the
two intervals [-a,-8] and [S, a].
As we will see in Theorem 8.6.5, and again in Chapter 9, approximate identities

play a very important role in analysis. An elementary example of such a sequence
Qnn=°° 1 is as follows:
n 1 1
n
Sxs n'
2'
QQ (t) =
0,
n
It is easily shown that the sequence {Q,,} is an approximate identity on [ -1, 1 ] (Exer-
cise 3). Other examples will be encountered in the proof of Theorem 8.6.1 and in the
exercises, and still others when we study Fourier series.
As a general rule, the Q. are usually taken to be even functions: i.e.,
QQ(-x) = Qn(x). T h e f a c t that the integrals over the set {t : f t I S} become small as
348 Chapter 8 - Sequences and Series of Functions
n --), oo seems to suggest that in some sense the functions themselves become small as
n becomes large. On the other hand, since the integrals over [-a, a] are always 1, by
property (b) of Definition 8.6.4
(8
lim J dt = 1
a
for every 8 > 0. This seems to indicate that the functions are concentrated near 0 and
must become very large near 0 (see Exercise 6). The graphs of the first few functions
Q1, Q2, and Q3 of a typical approximate identity are given in Figure 8.6.
-t -0.5 0 0.5
Figure 8.6 Graphs of Q Q2, Q3
8.6.5 THEOREM- Let bean approximate identity on [-I, I], and let f be abounded
real-valued periodic function on R of period 2 with f E 9t[ -1, 1 ]. For n e N, x E R,
define
S (x) = J l f (x + dt. (11)
1f f is continuous at x E R, then
lima Sn(x) = f(x).
Furthermore, if f is continuous on [ -1, 1 ], then

lim S (x) = f (x) uniformly on R.
Proof. We first note that since f is periodic and integrable on [-1. 1). f is integrable
on every finite subinterval of R. Thus the integral in equation (11) is defined for all
x E R. Also, since f is bounded, there exists a constant M > 0 such that L f (x)I s M
for all x E R.
Suppose first that f is continuous at x E R. By property (a) of Definition 8.6.4.
f(x) = Jf(x)Q0(t) dt.
Therefore,
IS.(x) -f(x)I = I f (f(x + t) -f(x)]Qn(t)dt

'
I f(x + t) - f(x)I Q,(t)dt. (12)

f
Let e > 0 be given. Since f is continuous at x, there exists a 8 > 0 such that
lf(X+t)-f(X)I < 2
for all t, I t I < S. Therefore,
Lf (x + r) - f (x) I Q. (t) dt < t a Q,,(t) dt

J a
2
5 J Q0(t) dt = 2. (13)
2
On the other hand,
J{851:I} If(x + t) - f(x)I Q,(t)dt 2M J

{85I'I}
Q.(t)tit.
Since {Qn) is an approximate identity , by property (b) there exists no E N such that
Qn(t) dt <
J{8s111} 4M
for all n a no. Thus if n a no,
f(f(x + t) - f(x)I Qn(t) dt < 2.

{8slr1}
Therefore, by equation (12) and the above,
IS. (X) -f(x)I s Ja If(x + t) -f(x)IQ0(t)dt
+ If(X + t) - f(X)I Qn(t) dt

{asIr1}
E E
<2+2=E
for all n z n,. Thus lim S0(x) = f(x).
n-Ow
Suppose f is continuous on [ -1. 1]. Since f is periodic, this implies that

f (-1) = f (I). By Theorem 4.3.4, f is uniformly continuous on [ -1, 11, and hence by
periodicity, also on R (Exercise 1). Thus given e > 0, there exists a 8 > 0 such that
I f(x + t) - f(x)I < 2
for all x E R and all t, I t l < 6. As a consequence, inequality (13) holds for all x E R.
Therefore, as above, there exists n E N such that
I f(x) I < E
for all x E R and all n ? no. This proves that the sequence converges uniformly to
f on R. Q
Proof of the Weierstrass Approximation Theorem. We now use Theorem
8.6.5 to prove the Weierstrass approximation theorem (Theorem 8.6.1). Suppose f is a
continuous real-valued function on [a, b]. By making a change of variable, i.e.,
g(x) = f((b - a)x + a), x E [0, 1 ],
we can assume that f is continuous on [0, 1 ]. Also, if we let
g(x)=f(x)-f(0)-x[f(l)-f(0)], xE[0,11,
then g(0) = g(l) = 0 and g(x) - f(x) is a polynomial. If we can approximate g by a
polynomial Q and set
P(x) = Q(x) + [f(x) - 8(x)]
then P is also a polynomial with I f(x) - P(x) I = Ig(x) - Q(x)I. Therefore, without
loss of generality, we can assume that f is defined on [0,1 ] satisfying
f(0) = f(1) = 0.
Extend f to [ -1, 1 ] by defining f (x) = 0 for all x E [ -1, 0). Then f is continuous
on [ -1, 1]. Finally, we extend f to all of R by defining
f (x) = f (x - 2k), k E Z,
where k E Z is chosen so that x - 2k E (-1, 1 ] (see Figure 8.7). Then f is continuous
and periodic of period 2 on all of R, and thus satisfies the hypothesis of Theorem 8.6.5.
f
-2 -1 o 2
Figure 8.7
Our next step is to find an approximate identity {Q"} on [ -1, 1 ] such that the cor-
responding functions S. of Theorem 8.6.5 defined by equation (11) are polynomials. To
accomplish this we let
Q,(t) = C"(1 - t2)",
where c > 0 is chosen such that
Q"(t) tit = 1.
12,
Thus the sequence {Q"} satisfies property (a) of Definition 8.6.4. To show that it also
satisfies (b) we need an estimate on the magnitude of c". Since
I I
1=c"J(l -t2rdt=2c"J(l -t2rdt
i 0
?2c"J (1-t2)"dt
o
2c,, (1-nt2)dt=2c" (
1
- 1
o 37 /
4c"
3\fn
we obtain
In the above we have used the inequality (1 - t2)" z 1 - nt2 valid for all t E [0, 1]
(Example 1.3.3(b)). Finally, for any 5, 0 < S < 1,
Q"(t)=c"(1-t2rsV(1-52r for all t,SsIt 1<-1.

Thus since 0 < (1 - 82) < 1, by Theorem 2.2.6(d), limo"(t) = 0 uniformly in
S <_ 1t1 <- 1. Therefore,
lim Q"(t) tit = 0.

"-'OD J{8 5 1 fl)
For x E [0, 1 ], set
P"(x) = j 1 f (x + t) Q"(t) tit.

This is the function S"(x) of Theorem 8.6.5, restricted to x E [0, 1]. Let x E [0, 1
Since f(t) = 0 for t E [ -1.0] U [ 1, 2],
f(x
Jt f(x + dr + t)QQ(t) dt,
c
which by the change of variables s = t + x gives
P"(x) = f 'f(s)Q. (s - x) ds.
Therefore, P"(x), for x E [0, 1 ], is a polynomial of degree less than or equal to 2n. As
a consequence of Theorem 8.6.5,
lim P"(x) = f(x)
n-M
uniformly on [0, 1).
thereby proving the result. O
Remark. The above proof of the Weierstrass approximation theorem is a variation of

a proof found in the text by Walter Rudin listed in the Bibliography.
EXERCISES 8.6
1. If f : lR - R is periodic of period 2 and continuous on [-1, 11, prove that f is uniformly continuous on R.
3. Forn E N, define Q. on [ -1, 1 ] as follows:
n -ln cx ln,
2
Q"(x) _
0, 1 <1xI sl.
Show that {Q"} is an approximate identity on [ -1, 1 ].
4. FornEN,setQ"(x)=c"(1 - IxI)",x E[-1,1].
a. Determine c" > 0 so that f', Q"(r) dr = 1.
b. Prove that with the above choice of c", the sequence {Q"} is an approximate identity on [ -1. 1 ].
c. Sketch the graph of Q"(x) for n = 2, 4, 8.
S. Forn E N, set Q"(x) = c,, x l e -', x E [ -1, 1 ].
a. Determine c" > 0 so that f', Q"(r) dr = 1.
b. Prove that with the above choice of c", the sequence {Q"} is an approximate identity on [ 1, 1 ].
6. *If {Q"} is an approximate identity on [ -1, 1 ], prove that Fim- x E [ -8. S)) = oo for every S > 0.
7. Let f be a continuous real-valued function on [0, 1). Prove that given e > 0, there exists a polynomial P with ratio-
nal coefficients such that 1f(x) - P(x) I < e for all x E [0, 1 ].
8.7 Power Series Expansions 353
8. Suppose f is a continuous real-valued function on [0, 11 satisfying
f f(x)x" dx = 0 for all n = 0. 1. 2....

0
Prove that f(x) = 0 for all x E [0, 1 ]. (Hint: First show that fo f(x)P(x) dx = 0 for every polynomial P. then use
the Weierstrass theorem to show that fo f2(x) dx = 0.)
8J Power Series Expansions

In this section we turn our attention to the study of power series and the representation
of functions by means of power series. Because of their special nature, power series
possess certain properties that are not valid for series of functions. We begin with the
following definition.
8.7.1 DEFINITION Let {ak}k_o be a sequence of real numbers, and let c E R. A series of
the form
00
I ak(X - c)k = a0 + a1(x - c) + a,(x - c)2 + aj(x - c)? +

k=0
is called a power series in (x - c). When c = 0, the series is called a power series in
x. The numbers at are called the coefficients of the power series.
Even though the study of representation of functions by means of power series dates
back to the mid-seventeenth century, the rigorous study of convergence is much more re-
cent. Certainly Newton and his successors were concerned with questions involving the
convergence of a power series to its defining function. It was Cauchy, however, who,
with his formal development of series, brought mathematical rigor to the subject. As an
application of his root and ratio test, Cauchy was among the first to use these tests to de-
termine the interval of convergence of a power series. This is accomplished as follows:
Consider a power series 7, ak(x - c)k. Applying the root test to this series gives
km Iakl Ix - cl A = Ix - cla.
m
k
where a = lim
k-.oo
I ak [ . Thus by Theorem 7.3.4, the series converges absolutely if
alx - cl < 1, and divergesifalx - cl > 1.Ifa=0,then alx-cl < I for all
x E R. If 0 < a < oo, then
aix - cl < I if and only if Ix - cl <
8.7.2 DEFINITION Given a power series 7- ak(x - c)k, the radius of convergence R is
defined by
R = litn.
1
If lim ° Ia oo we take R = 0, and if lim 'Ml q, = 0 we set R = oo.

354 chapter8 Sequences and Series of Functions
When R = 0, the power series E ak(x - c)k converges only for x = c. On the
other hand, if R = oo, then the power series converges for all x E R.
Remark If ak # 0 for all k and lim I ak+, I / I ak I exists, then by Theorem 7.1.10 the
radius of convergence of E ak(x - c)k is also given by
I lak+1I
= lim Iaj
R k-,oo
This formulation is particularly useful if the coefficients involve factorials.
8.7.3 THEOREM Given a power series 5:k'. O ak(x - c)k with radius of convergence R,
0 < R 5 oo, then the series
(a) converges absolutely for all x with Ix - cI < R, and
(b) diverges for all x with I x - c I > R.
(c) Furthermore, if 0 < p < R, then the series converges uniformly for all x with
Ix - ci :S P.
Proof. Statements (a) and (b) were proved in the discussion preceding the statement
of the theorem. Suppose 0 < p < R. Choose /3 such that p < 6 < R. Since
k
lim
k-roo = RI < I,
k { ak I < for all k at no.
Hence for k a n, and I x- c i 5 p,
lak(x - c)kj Iakjpk < (Q/k.
But (p//3) < I and thus 7,(p/j3)k < oo. Therefore, by the Weierstrass M-test, the se-
ries converges uniformly on Ix - c) s p. 0
The previous theorem provides no suggestion as to what happens when }x - c l = R.
As the following examples (with c = 0) illustrate, the series may either converge or di-
verge when I x I = R.
8.7.4 EXAMPLES
(a) The series
00
xk
71
k-1
has radius of convergence R = 1. This series diverges at both x = I and -1.

(b) The series J , xk/k also has radius of convergence R = 1. In this case, when
x = 1 the series diverges; whereas when x = -1, the series is an alternating series
which converges by Theorem 7.2.3.
(c) Consider the series 7,k'-, xk/k2. Again the radius of convergence is R = 1. In this
example the series converges at both x = I and - 1.
(d) Consider the series
I + 2x + 32x2 + 23x3 + 34x4 a,,xk

k=0
where
__ 3k, if k is even,
ak
2' if k is odd.
Hence Tim V1 akl = 3, and therefore, R = 3. The series diverges at both x = 3 and
X = -3.
(e) Finally, consider the series 7, k!xk. Here ak = k!, and
a541
lim- = kim(k + 1) = oo.
k--.oo ak
Thus by Theorem 7.1.10, k /l ak - oo and R = 0. Therefore, the power series con-

verges only for x = 0.
Abel's Theorem
Suppose we are given a power series Y, ak(x - c)k with radius of convergence R > 0.
By setting
AX) = Y,00 ak(x - c)k, (14)

k-0
we obtain a function that is defined for all x, f x - cI < R. Functions that are defined
in terms of a power series (as in series (14)) are usually referred to as ral.aoalytic
functions. Fix p with 0 < p < R. Since the series converges uniformly to f on
Ix - cI p, by Corollary 8.3.2, f is continuous on Ix - cI <- p. Since this holds for
all p < R. the function f is continuous on Ix - c I < R. If the series (14) also con-
verges at an endpoint, say at x = c + R, then f is continuous not only in (c - R, c + R)
but also at x = c + R. This follows from the theorem of Abel, which is shown next. For
convenience, we take c = 0 and R = 1.
8.7.5 THEOREM (Abel's Theorem) Suppose f(x) = Y-k o akxk has radius of conver-
gence R = 1, and that 7_' o ak converges. Then
00
f(x) = Y, ak.
k=0
Proof. Set s_I = 0, and f o r n = 0, 1, 2, ... let s = 7,1.0 ak. Then by the partial
summation formula (7.2.1),
akXk =
Sk(Xk - xk+I) + Sin
k-0 k=0 /
n-I
(1 - X) I Skxk +
k-0
Since the sequence {sn} converges, if we let n -a oo, then for all x. Ix I < 1,
x+
f (X) _ (1 - X) L, SkXk
k-0
Lets = lim s,,, and let e > 0 be given. Choose n,, E N such that I s - sn I < 1 e for
all n Z no. Since
00
(1 -x)7, xk= 1, IXI <
A-0
we have for all x, 0 < x < 1,
If(x) - SI = 0 - X)kI(Sk - S)xkl s (1 - X)k ISk - SIX'
(I-X)Sk-SI
k=0
+E(1-x),xk
2 k=n,+I
s(1-x)M+2
where M = Yk,olsk - sI. If we now choose 8 > 0 such that I - S < x < I im-
plies that (1 - x) M < e, then if(x) - s l < e for all x, 1 - 8 < x < I. Thus
lim f (x) = s. Q i
8.7.6 EXAMPLE To illustrate Abel's theorem, consider the series Icko-o (-1)ktk. This series
has radius of convergence R = 1. Furthermore, the series converges to f (t) = 1/(1 + t)
for all t, I t I < 1. Since the convergence is uniform on I t I s I x I where I x I < 1. by
Corollary 8.4.2,
dt °°
ln(1 + x) =
+
7(-1)J
= k.0 tk d
0 0
00
(-1)kxk+I = (-1)t+l xk
0
k.ok + 1 k.I k
for all x, I x I < 1. The series

(-1I xk
k-1 k
has radius of convergence R = 1, and also converges when x = 1. Thus by Abel's the-
orem,
001
n2=
k
1-2+3-4
1
Differentiation of Power Series

Suppose the power series Yw 0 ak(x - c)k has radius of convergence R > 0. If we dif-
ferentiate the series term-by-term we obtain the new power series,
0 00
2 kak(x - c)k-1 = G (k + 1)ak+1(x - c)k. (a5)
k-l k-0
The obvious question is what is the radius of convergence of the differentiated series
(15)? Furthermore, if f is defined by f(x) = I' oak(x - c)k, lx - cI < R, does the
series (15) converge to f'(x)? The answers to both of these questions are provided by
the following theorem.
8.7.7 THEOREM Suppose 7,- o ak(x - c)k has radius of convergence R > 0, and
Iak(x-c)k, Ix-cl <R.

k-0
Then
00
(a) 2 kak(x - c)k-1 has radius of convergence R, and
k.1
ao
(b) f'(x) = 7, kak(x - C)k-'for all x, Ix - cl < R.
k-l
Proof For convenience we take c = 0. Consider the differentiated series 7, kalxk -1.
By Theorem 2.2.6, lim \ = 1, and for x * 0,
lxI _
k
k
I k-1 = W. . By Exercise 10 of Section 2.5, for x * 0,
fi-m k IakIkIXIk-1
k
=lxliklakl.
Therefore, if R = oo, the differentiated series (15) converges for all x, and if 0 < R < oo,
the differentiated series converges for all x, IxI < R, and diverges for all x, Ixl > R.
Thus the radius of convergence of Joko- 1 kakxk-1 is also R.
Furthermore, for any p, 0 < p < R, by Theorem 8.7.3 the series 7, kakxk-' con-
verges uniformly for all x, I x 1 s p. Thus by Theorem 8.5.1, the series (15), obtained
by term-by-term differentiation, converges to f'(x), i.e.,
OQ
f'(x) = Y, kakxk-' for all x, I xI < R.

k=1
8.7.8 COROLLARY Suppose Ioko- o ak(x - c)k has radius of convergence R > 0, and
x
f (x) = Y, ak(x - c)k, I x - c I < R.
k=o
Then f has derivatives of all orders in Ix - c I < R, and for each n E IN,
x
ft")(x) _ ',k(k - 1) (k - n + 1)ak(x - c)k-". (16)
k-n
In particular,
ft")(c) = n! a,,. (17)
Proof. The result is obtained by successively applying the previous theorem to

f, f', f", etc. Equation (17) follows by setting x = c in equation (16).
8.7.9 DEFINITION A real-valued function f defined on an open interval I is said to be in-

finitely differentiable on I if f(x) exists on I for all n E N. The set of infinitely dif-
ferentiable functions on an open interval I is denoted by C'(1).
As a consequence of Corollary 8.7.8, if 7, ak(x - c)k has radius of convergence

R > 0 and if f is defined by f(x) = 7-'k-, ak(x - c)k for Ix - cI < R, then the func-
tion f is infinitely differentiable on (c - R, c + R) and its nth derivative is given by
equation (16). We illustrate this with the following example.
8.7.10 EXAMPLE For Ixi < 1,

00
xk.
1-x
Thus by the previous corollary,
00 co
1
(1 - x)2 = kYa
kz k-1 = 1( k + l)x k ,
00 00
2
(1 - x)3
= k k( k - 1)x k-2 = k ',( k + 2)( k + 1)xk ,
-2 -0
and for arbitrary n E N,
(n-1)!-
(1 -xY' J(k+n-1 )...(k+ 1)xk.
Uniqueness Theorem for Power Series

The following uniqueness result for power series is another consequence of Corol-
lary 8.7.8.
8.7.11 COROLLARY Suppose ak(x - c)k and 7, bk(x - c)k are two power series which
converge for all x, Ix - c < R, for some R > 0. Then
00 00
G ak(x - c)k = T, bk(x - c)k, 'x - ci < R,

k-0 k-0
i f and only i f ak = bk f o r all k = 0, 1, 2, ... .
Proof. Clearly, if ak = bk for all k, then the two power series are equal and converge
to the same function. Conversely, set
00
c)k
AX) = T,00 ak(x - and g(x) = Y, bk(x - c
k=0 k-0
If f (x) = g(x) for all x, x - c I < R, then r)(x) = gt") (x) for all n = 0, 1, 2.. . .
and all x, I x - c I < R. In particular, f")(c) = gO)(c) for all n = 0. 1, 2, .... Thus by
equation (17), a" = b" for all n.
Representation of a Function by a Power Series

Up to this point we have shown that if a function f is defined by a power series, that is
00
f(x) _ Y, ak(x - c)k, Ix - c) < R,
k-0
with radius of convergence R > 0, then by Corollary 8.7.8, f is infinitely differentiable

on (c - R, c + R) and the coefficients ak are given by ak = f clo(t)/k!. We now consider
the converse question. Given an infinitely differentiable function on an open interval I
and c E 1, canf be expressed as a power series in a neighborhood of the point c? Specif-
ically, does there exist an e > 0 such that
00
f (x) = G; ak(x - c)k

k-0
f o r all x, I x - c I < e, with ak = f ()(c)/k! f o r all k = 0, 1, 2, ... ? The following ex-
ample from Cauchy shows that this is not always possible.
8.7.12 EXAMPLE Let f be defined on R by
0, x=0.
Since X
m
e" = liim a-`' = 0, f is continuous at 0. For x * 0,
x3
When x = 0, we have
f'(0) = lim
h-*0
f (h)
-h A o) = lim
h-0
e
h
= lim
' '
t.
e'
= 0.
The last step follows from 1'Hospital's rule. Thus,
.f'(x) _ X, e '/'', x 0 0,
0, x = 0.
By induction, it follows as above, that for each n E N,

P(s)e_'fix,
f(n)(x) = 1 x # 0,
( 0, x0,
where P is a polynomial of degree 3n. The details are left to the exercises (Exercise 15).
Thus the function f is infinitely differentiable on R. If there exists R > 0 such that
f(x) = J o akxk for all x, IxI < R, then ak = 0 for all k. As a consequence, f cannot
be presented by a power series that converges to f in a neighborhood of 0.
Taylor Polynomials and Taylor Series

We now consider the problem of representing a function f in terms of a power series in
greater detail. Newton derived the power series expansion of many of the elementary
functions by algebraic techniques or term-by-term integration. For example, the series
expansion of 1/(1 + x) can easily be obtained by long division, which upon term-by-
term integration gives the power series expansion of In (1 + x). Maclaurin and Taylor
were among the first mathematicians to use Newton's calculus in determining the coef-
ficients in the power series expansion of a function. Both realized that if a function f (x)
had a power series expansion ak(x - c)k, then the coefficients ak had to be given by
f(c)/k!.
8.7.13 DEFINITION Let f be a real-valued function defined on an open interval 1, and let
c E I and n E N. Suppose f W(x) exists for all x E I. The polynomial
(k) (c)
r.( C)(x) _
kmo
f kf (x - c)k
is called the Taylor polynomial of order n off at the point c. If f is infinitely differen-
tiable on 1, the series
= fk)(c)
(x - c)k
k.0 k!
is called the Taylor series off at c.

For the special case c = 0, the Taylor series of a function f is often referred to as
the Maclaurin series. The first three Taylor polynomials, To, T1, T2, are given specifi-
cally by
T0(f, c)(x) = f(c),
T1(, c)(x) = f(c) + f'(c)(x - c),
f 2r) (x - c)2.
T2(f, c)(x) = f(c) + f'(c)(x - c) +
The Taylor polynomial T,(f, c) is the linear approximation to f at c; that is, the equa-
tion of the straight line passing through (c, f(c)) with slope f'(c).
In general, the Taylor polynomial T. of f is a polynomial of degree less than or
equal to n that satisfies
T (k)(f c)(c) = f tk)(c),
for all k = 0, 1, ... , n. Since ft">(c) might possibly be zero, T. (as the next example
shows) could very well be a polynomial of degree strictly less than n.
8.7.14 EXAMPLES In the following examples we compute the Taylor series of several func-
tions. At this stage nothing is implied about the convergence of the series to the function.
(a) Let f (x) = sin x and take c = Z. Then
f(2) = sin 2 = 1,
f'(2)=cos2=0,
f.("(2) = -sin 2 = -1,
J 3)(2) = -Cos Z = 0.
Thus
2
T3(f2)(x)=1-2t x-2/,
which is a polynomial of degree 2. In general, if n is odd, 0, and if n = 2k is
even, P 20(f) = (-1)k. Therefore, if n is even,
T ( f , 2)(x) = T. 1(J,
n12 (-I r
2)(x) = 0 (2k) \x - -2\zk
The Taylor expansion of f(x) = sin x about c = 2 is given by
00 (-1)k 2k
ir
()! IX
2)
(b) For the function f(x) = e-"X', by Example 8.7.12
0)(x) = 0 for all n E N.
Thus the Taylor series of f at c = 0 converges for all x E l ; namely, to the zero func-
tion. It, however, does not converge to f.
(c) In many instances, the Taylor expansion of a given function can be computed from
a known series. As an example, we find the Taylor series expansion of f(x) = 1/x about
c = 2. This could be done by computing the derivatives off and evaluating them at
c = 2. However, it would still remain to be shown that the given series converges to
f (x). An easier method is as follows: We first write
1 1 _l 1
x 2-(2-x) (2 2
For lwl < 1,

rwk.
°O
1-w k_o
Setting w = (2 - x)/2, we have

1 (2 - xk = 0O (-1)k
71 2k+1 (x - 2)k
X 2k O 2k k.O
for all x, I x - 21 < 2. By uniqueness, the given series must be the Taylor series of
f(x) = 1/x. In this instance, the power series also converges to the function f(x) for all
x satisfying Ix - 21 < 2. U
11
Remainder Estimates
To investigate when the Taylor series of a function f converges to f(x), we consider
R (x) = c)(x) = f (x) - c)(x). (18)
The function R. is called the remainder or error function between jr and c).
Clearly,
f (x) = Jim TT(f, c)(x) if and only if lim c)(x) = 0.
n
Since the Taylor polynomial T. is the nth partial sum of the Taylor series of f, the
Taylor series converges to f at a point x if and only if nliini c)(x) = 0. To emphasize
this fact, we state it as a theorem.
8.7.1 5 THEOREM Suppose f is an infinitely differentiable real-valued function on the open

interval I and c E I. Then for x E 1,
AX) = 00
fck)(c) (x - c) , k
k-O k.E
if and only if limn c)(x) = 0.

The formula
f(x) =f(c) +f'(c)(x - c) +f2(x - c)2 +

+
f n() (x - c)" + R"(f, x)(x)
is known as Taylor's formula with remainder. We now proceed to derive several for-
mulas for the remainder term R. These can be used to show convergence of T. to f.
Lagrange Form of the Remainder

Our first result, attributed to Joseph Lagrange (1736-1813), is called the Lagrange
form of the remainder. This result, sometimes also referred to as Taylor's theorem,
was previously proved for the special case n = 2 in Lemma 5.4.3.
8.7.16 THEOREM Suppose f is a real-valued function on an open interval 1. c E I and

n E N. If f (n+ ')(t) exists for every t E 1, then for any x E I, there exists a C between x
and c such that
Rn(x) = Rn(.f, c)(x) _ + 1) (x - C)n+l (19)
Remark. Continuity of f("+) is not required.

Proof. Fix x E 1, and let M be defined by
f(X) = T"(f, c)(x) + M(X - c)n+
To prove the result, we need to show that (n + 1)! M = f("+ ')(C) for some be-
tween x and c. To accomplish this, set
g(t) = f(t) - TTJ, c)(t) - M(t - On +,

= R"(t) - M(t - c)n+1
First, since T. is a polynomial of degree less than or equal to n,
g(n+Ikt) = f(n+t)(t) - (n + 1)! M.
Also, since Tne)(f, c)(c) = f (t)(c), k = 0, 1, ... , n,

g(c) = g'(c) = ... = g(")(c) = 0.
For convenience, let's assume x > c. By the choice of M, g(x) = 0. By the mean
value theorem applied to g on the interval [c, x], there exists x,, c < x, < x, such that
0 = g(x) - g(c) = g'(xi)(x - c).
Thus g'(x,) = 0. Since g'(c) = 0, by the mean value theorem applied to g' on the in-
terval [c, x, ], g"(x2) = 0 for some x2, c < x2 < x,. Continuing in this manner, we ob-
tain a point x" satisfying c < xn < x, such that g(")(xn) = 0. Applying the mean value
theorem once more to the function g(n) on the interval [c, xn], we obtain the existence of
a t E (c, xn) such that
0 = 8(n)(Xn) - g(n)(c) = Jn+ l)( )(X - C).
Thus 0; i.e., f(n+1)(C) - (n + 1)! M = 0, for some C between x and c. 0
In Example 8.7.20 we will give several examples to show how the remainder esti-
mates may be used to prove convergence of the Taylor series to its defining function. In
the following example we show how the previous theorem may be used to derive sim-
ple estimates and inequalities.
8.7.17 EXAMPLES
(a) In this example we use Theorem 8.7.16 with n = 2 to approximate f (x) = 1 + x,
x > -1. With c = 0 we find that
f(0) = 11 f'(0) = , f"(0) = -4 .
2
Therefore, T2(f, 0)(x) = 1 + 2 x - e x, and thus
1
1 +x = 1+2x-8x2+ RZV, 0x
1
) ( ).
By formula (19),
R2(f, 0)(x) = f (3)
L) X3
= 16
I (1 + )-SR x3
3! -
for some C between 0 and x. If x > 0, then > 0, and thus (1 + C)"s2 < 1. There-
fore, we have
V1 -+x - T2(f, 0) (x) I < x3

16
for any x > 0. If we let x = 0.4, then TZ(f, 0)(.4) = 1.18, and by the above,
I 1.4 - 1.181 < 0.004, so that two-decimal-place accuracy is assured. In fact, to five
decimal places 1.4 = 1.18322.
(b) The error estimates can also be used to derive inequalities. As in the previous
example,
1
-1/1+1= 1 + 2 x - 8x2 + R2(f, 0) (x).
For x > 0 we have 0 < R2(f, 0)(x) < 16x3. Thus
I+2x-8x2< VT +x<1+2x-gx2+ I x3
16
for. all x > 0.

Integral Form of the Remainder

Another formula for R.V, c) is given by the following integral form of the remainder.
This, however, does require the additional hypothesis that the (n + l )st derivative is
Riemann integrable.
8.7.18 THEOREM Suppose f is a real-valued function on an open interval 1, c E I and

n I-, N. If f ("+'kt) exists for every t E 1 and is Riemann integrable on every closed and
bounded subinterval of I, then
R"(x) = R"U c)(x) = n! { `f("+°(r)(x - t)"dt, x E 1. (20)

C
Proof. The result is proved by induction on n. Suppose n = 1. Then

R1(x) = f(x) - f(c) - f'(c)(x - c),
which by the fundamental theorem of calculus
= l " f'(t) dt - f'(c) dt = f J

[f'(t) - f '(c)) dt.
From the integration by parts formula (Theorem 6.3.7) with

u(t) = f'(t) - f'(c), v'(t) = 1,
u'(t) = f"(t), v(t) _ (t - jr),
we obtain
j CX
[f'(t) - f'(c)) dt = [f'(t) - f'(c))(r - x) I; - f"(t)(t - x) dt
JC C
Is (x - t) f"(i) dt.
To complete the proof, we assume that the result holds for n = k, and prove that
this implies the result for n = k + 1. Thus assume Rk(x) is given by equation (20). Then
Rk+ I(x) = f(x) - Tk+ IV, c)(x)

p(k+ (c)
f(x) - Tk(f c)(x) - (k + 1)! (x - c)k+i
f(k+I)(
=Rk(x)- (k+ 1)i) (x - cr+i
kt js (X - t)kf (k+')(t) dt - k' f (k+')(c) J I (x - t)k dt
1 k (k+I) (k+')
366 Chapter a Sequences and Series of Functions
As for the case n = 1, we again use integration by parts with
u(t) = f (k+'1(t) - f(k+'1(c) and v(t) = (x - 1)k-I,

k +
which upon simplification, gives
Rk+i(x) = 1 J s (x - t)k''f(k+'>(t)dt.
(k + 1)! J.
Cauchy's Form for the Remainder

Under the additional assumption of continuity of f we obtain Cauchy's form for
the remainder as follows.
8.7.19 COROLLARY Let f be a real-valued function on an open interval 1, c E 1 and n E N.

If f("+1) is continuous on I, then for each x E 1, there exists a C between c and x such
that
Rn(x) = R,(f, c)(x) =

P +14) (X - cxx - i)". (21)
n!
Proof Since f("+')(t)(x - t)" is continuous on the interval from c to x, by the mean
value theorem for integrals (Theorem 6.3.6), there exists a C between c and x such that
f fn+I)(:)(x - t/ dt = (x - c) f("+'kC)(x
l - C)".
C
The result now follows by equation (20).

We now compute the Taylor series for several elementary functions, and use the
previous formulas for the remainder to show that the series converges to the function.
8.7.20 EXAMPLES
(a) As our first example, we prove the binomial theorem (Theorem 2.2.5). For n E N
let f(x) = (1 + x)", x E R. Since f is a polynomial of degree n, if k > n then
f (k)(x) = 0 for all x E R. Therefore, by Theorem 8.7.16,
f(x) _
f (k)(0) = n!/(n - k)! f o r k = 0, 1, ... , n. Therefore,

n!
k=a k!(n - k)!

The series expansion of (1 + x)° for a E R with a < 0 is given in Theorem 8.8.4,
whereas the expansion for a > 0 is given in Exercise 7 of the next section. For ratio-
nal numbers a, the expansion of (1 + x)° was known to Newton as early as 1664.
(b) Let f (x) = sin x with c = 0. Then
f ("i(x) ( 1)k cos x, n = 2k + 1,

= (-1)ksinx, n = 2k.
Thus f(")(0) = 0 for all even n E N, and f ii(0) = (-1)'`, whenever n = 2k + 1, k =
0, 1, 2, .... Therefore, the Taylor series off at c = 0 is given by
00 (-1)k x2k+i
k=O (2k + 1)!
To show convergence of the series to sin x we consider the remainder term Rn(x).
By Theorem 8.7.16, for each x E R there exists a C such that
f(n+ 1 (SS)
n+l
R"{x) = (n + 1)! x
Since Ii (" ")(x) 1 -5 1 for all x, we have

Ixln+1
Rn(x),
(n + 1)!
By Theorem 2.2.6(f), lim jxI"+'/(n + 1)! = 0 for any x E R. As a consequence.
lim Rn(x) = 0 for all xnE+1R, and thus
00 (-+)
sinx = 7 x +', x E R.
k_o (2k + 1)`
The sine function, as well as the cosine function, can be defined strictly in terms of
power series. For further details, see Miscellaneous Exercise 3.
(c) As our third example, we derive the Taylor series for f (x) = ln(1 + x), where, as
in Example 6.3.5,
lnx=fix Idr, x>0,
denotes the natural logarithm function on (0, oo). Then f(0) = In (1) = 0, and by the
fundamental theorem of calculus, f '(x) = 1/(1 + x). Thus for n = 1, 2, ... ,
f("i(x) _ (-1)n+i (n - 1)!
0+x)"
In particular, f W(0) _ (-1)"+ 1(n - 1)!, and the Taylor series off at 0 becomes
. (-1)n+I
7'
n., n
x
Although we have already proved that this series converges to In (1 + x) for all
x, -1 < x 1 (Example 8.7.6), we will prove this again to illustrate the use of the
remainder formulas.
Suppose first that 0 < x <- 1. By Theorem 8.7.16,
(_ 1)"+zx"+l
R (x) = R If 0)(x1 =
(n + 1)(l + C)"''
for some C, 0 < < x. In this case, (1 + C) > 1, and thus
IR"(x) I n+ l x
"+l<
n+ 1
for all x, 0 < x s 1. Therefore,
1im R"(x) = 0 for all x e [0, 1 ].
n-Oc
We next consider the more difficult case -1 < x < 0. By the Cauchy form of the
remainder, if - I < x < 0, there exists ', x s s 0, such that
R"(x) = ft"+l'(C)(x l)"+2 (Ix+

)rx
n! - C)"x = (- n+l'
Therefore,
IR"(x)I
(II+I0 [(l + )]"'
Consider the function ap(t) defined on [x, 0] by
x
(P(t) = 1 + t
Then cp'(t) = (I + x)/(1 + t)2, which is positive on [x, 0] provided x > -1. There-
fore,
(P(t) q,(0) = -x = IxI
for all t, x s t <_ 0. Thus
I+I0lxln.
IR"(x)I `- (1
Since IxI < 1, lim R"(x) = 0. Therefore, the Taylor series converges to In (1 + x) for
all x,-1 <x -_ 1;i.e.,
00 (-I)n+l
ln(l+x)=
n=l n
x", -l<xsl.
The Taylor expansion of In(l + x) was first obtained in 1668 by Nicolaus Mercator.
Newton shortly afterward obtained the same expansion by term-by-term integration of
the series expansion of 1/(1 + x) (see Example 8.7.6).
(d) As our final example, we consider the natural exponential function E(x) = exp x
that is defined as the inverse function of the natural logarithm function L(x) = In x. The
domain of L is (0, oo) with range (-oo, oo). Since L'(x) = 1/x is strictly positive
on (0, oo), L is a strictly increasing function on (0, oo). The inverse function E(x) is de-
fined by
y = E(x) if and only if x = In y.
By the inverse function theorem (Theorem 5.2.14), E is differentiable on R with
E'(x) = E'(L(y)) = L'(y) = y = E(x).
Thus E'(x) = E(x) for all x E R. Since In ab = In a + In b, it immediately follows

that
E(x + y) = E(x)E(y) and E(-x) =

E(x)
Also, E(O) = 1, and by Example 6.3.5, E(1) = e where e is Euler's number defined in
Example 2.3.5. Therefore, E(n) = e" for every integer n. If r = m/n, m, n E 1 with
n * 0, then E(nr) = E(m) = e. But E(nr) _ (E(r)r. Therefore, if r E C4,
E(r) = e'.
For arbitrary x E R we define e" by e` = E(x). This definition of e' is consistent with
the definition given in Miscellaneous Exercise 3 of Chapter 1 (See Exercise 16).
Since E(")(x) = E(x) for all x E R, E(")(0) = E(0) = 1. Thus the Taylor series ex-
pansion of E(x) about c = 0 is given by
00
'X1
k-0 k
It is left as an exercise (Exercise 11) to show that this series converges to e' for all
x E R.
There is a more subtle question involving power series representation of functions
that we have not touched upon. The question concerns the following: How is the radius
of convergence R of the Taylor series off related to the function f? The full answer to
this question requires a knowledge of complex analysis and thus is beyond the scope of
this text. However, we will illustrate the question and provide a hint of the answer with
the following examples. If f(x) = 1/(1 + x), then the Taylor expansion off about
x = 0 is given by
1. (- I )kxk,
k=0
which has radius of convergence R = 1. This is expected since the given function f is
not defined at x = -1, and thus the series could not have radius of convergence R > 1.
If it did, this would imply that f would then have a finite limit at x = -1.
On the other hand, the function g(x) = 1 /(1 + x2) is infinitely differentiable on all
of R. However, the Taylor series expansion of g about c = 0 is given by
00
,(-I)kX2k,
k=0
which again has only radius of convergence R = 1. The reason for this is that even
though g is well-behaved on R, if we extend g to the complex plane C by
1
g(z) = 1 + z2
then g is not defined when z2 = -1; i.e., z = ±i, where i is the complex number that
satisfies i2 = -1.
EXERCISES 8.7
1. Find the radius of convergence of each of the following power series.
00 3k
a. D
Xk *b. x+12*
k-1 k k=0 4
00
1 - 1 xk
*d.
1 71 k
1 when k is even, 00 / kt
C. I,, akxk where ak - 2" . f. j k1 2)
k'0 when k is odd k1
2k +2
2. For each of the following, determine all values of x for which the given series converges.
l+x k
a.
k-0 X
k, (X * 0) tb G
k- 1 2(1
3kxk
x)"k
(X # 1) C. I 1-x
(x #1)
3. Using the power series expansion of 1/(1 - x) and its derivatives, And
00 00
+a. T', kxk, IXI < 1 b. I k2xk, IXI < t c. nkk

k-1 k_1 k-1
4. a. Use Theorem 8.7.16 to show that 1i - (I + 1 x - 9 x2) I < 81x3 for all x < 0.
b. Use the above inequality to approximate 'Mi, and provide an estimate of the error.
5. Determine how large n must be chosen so that I sin x - T (sin, 0)(x)l < .001 for all x, IXI s 1.
6. Use the Taylor series and remainder estimate of Example 8.7.20(c) to compute In 1.2 accurate to four decimal
places.
7. Suppose f(x) = Ik o ak(x - c)t has radius of convergence R > 0. For Ix - cl < R, set F(x) = f' f(r) dr.
Prove that
00 a,
k+1(x-c)k+i, Ix - cl < R.
F(x) = k
8. *a. Use the previous exercise and the fact that
Arctan x =
1 + x2
to obtain the Taylor series expansion of Arctan x about c - 0.
b Use pact (a) to obtain a series expansion for ir.
*c. How large must n be chosen so that the nth partial sum of the series in part (b) provides an approximation of
it correct to four decimal places?
9. Use Exercise 8 and Abel's theorem to prove that
(-1 a
k=o2k+1 4'
10. Find constants ao, a,, a2, a3, a4 such that
x4+3x2-2x+5=a4(x- 1)4+a3(x- 1)3+a2(x- 1)2+a,(x- 1)+ao.
11. Prove that the Taylor series of e` (with c = 0) converges toe` for all x E R.
12. Using any applicable method, find the Taylor series of each of the following functions at the indicated point, and
specify the interval on which the series converges to the function.
a. &) = cos x, c = 0 *b. f (x) = In x, c = 1
± \),
c. f(x) = In1 l x c=0 *d. f(x) = (I - x)-". c = 0
x
*e. f(x) = Aresinx, c = 0 L f(x) = 1 x2, c=0
g. f(x) = V X-, c = 1 h. f(x) = Arctanx, c = 1

13. Suppose f (x) _= Ik o akxk, I xl < R, where R > 0. Prove the following.
a. f (x) is even if and only if ak = 0 for all odd k.
b. M) is odd if and only if ak = 0 for all even k.
14. Suppose f(x) = 7.i oakxk, I x I < R1, and g(x) _ Zm o bk xk, I x I < R2. Prove that
f(x)g(x) _ Jk ockxk, lxi < min{R,, R2}, where ck = Yi-oaibk-i

15. Let f: R -> R be defined by f(x) = e "" for x * 0, and f(0) = 0. Prove that for each n E N,
P()e-V' x * 0,
Mx) 0, x = 0,
where P is a polynomial of degree 3n.
16. Suppose b > 1. For x E R define b(x) = E(x In b), where E is the natural exponential function.
a. Prove that b(r) _- b' for all r E Q.
b. For x E R, prove that b(x) = sup(b' : r E 0, r < x}.
8.81 The Gamma Function

We close this chapter with a brief discussion of the Beta and Gamma functions, both of
which are attributed to Euler. The Gamma function is closely related to factorials, and
arises in many areas of mathematics. The origin, history, and development of the Gamma
function are described very nicely in the article by Philip Davis listed in the supple-
mental reading. Our primary application of the Gamma function will be in the Taylor
expansion of (1 - x)-, where a > 0 is arbitrary.
8.8.1 DEFINITION For 0 < x < oo, the Gamma function r(x) is defined by
]F(x) = Jte1d:. (22)
When 0 < x < 1, the integral in equation (22) is an improper integral not only at
oo, but also at 0. The convergence of the improper integral defining f(x), x > 0, was
given as an exercise (Exercise 9) in Section 6.4. The graph of r(x) for 0 < x < 5 is
given in Figure 8.8. The following properties of the Gamma function show that it is
closely related to factorials.
6 I 2 3
Figure 8.8 Graph of f(x), 0 < x s 5
8.8.2 THEOREM
(a) For each x, 0 < x < oo, r(x + 1) = x r(x).
(b) For n E ICU, r(n + I) = n!.
Proof. Let 0 < c < R < oo. We apply integration by parts to
1 R t'e-' dt.
With u = tx and v' = e-',

R +xJR
txe-'dt= -rè-' tx -Ie-'di
R
Rx
e
Since im; cxe-` = 0 and m R*e-R = 0, taking the appropriate limits in the above
R
yields
r(x + 1) = J txe-'dt = x 0 1x-le-`dt = xI'(x).

0 0
This proves (a). For the proof of (b) we first note that
I'(1) = e-' dt = 1.
J000
Thus by induction, r(n + 1) = W.
8.8.3 EXAMPLE Since the value of I'(!) occurs frequently, we now show that I'('-2) = .
By definition,
r J = Joy t- 'ne-' dt.

2
With the substitution t = s2.
To complete the result, we need to evaluate the so-called probability integral f0 e-" ds.
This can be accomplished by the following trick using the change of variables theorem
from multivariable calculus. Consider the double integral
e-xl-''dxdy.
J= Joo f
0 0
By changing to polar coordinates

x = r cos 9, y = r sin 9,
with 0< r < oo, 0 E (0, z ),
WRe `rdrd9
J= 100
0 10
a (O0
= ?J e_,: rdr4. IT
0
On the other hand,

0o (oo >o
e-x:
2
J= J J e-:= a-)" dx dy = J dx
0 0 0
Therefore,
f0c
e-'2 dx = 2
0
from which the result follows.
The Binomial Series

As an application of the Gamma function, we will derive the power series expansion of
f(x) = (1 - x)-°, where a > 0 is real. The coefficients of this expansion are ex-
pressed very nicely in terms of the Gamma function. By Example 8.7.10, for n E N,
00
(1-x)-n=
(n - 1)! k.0 k!
which in terms of the Gamma function, gives
1 _ 1 r(k+n) k
x
(1 - xr 1'(n) k=o k!
We will now prove that this formula is still valid for all a E R with a > 0.
8.8.4 THEOREM (Binomial Series) For a > 0,

1 _ 1
oc r(n + a)xn, IxI
< 1.
x)° n!
Proof. We first show that the radius of convergence of the series is R = 1. Set
an = 17(n + a)/n!. Then
an+,r(n+1+a) n!
a (n + 1)! T(n + a)*
But by Theorem 8.8.2, I'(n + 1 + a) = (n + a)I'(n + a). Therefore,
a^+, =limn + a = 1.
lim
n-"o a -oo n + 1
and as a consequence of Theorem 7.1.10, we have R = 1.
To show that the series actually converges to (1 - x)-a, we set

1 00 r(n + a)
fa(xW = IxI < I.
r(a) n! x",
Since a power series can be differentiated term-by-term,
fv(x) = r(a)
1
I- nr(n + a) n-1
n!
x X.
Multiplying by (I - x) gives
' nr(n + a)
(1 - x)x rt-'
1
(l - x)fa(x) = r( a) Y' n!
_ 1 a r(n + a) n_i -° nr(n + a)

r(a) I (n - 1)! x n1 x
n-I
- 1 `F(n + I + a) nF(n + a) 1
L X.
F(a) o n! n!
But r(n + I + a) - nr(n + a) = ar(n + a). Therefore,

(1 - x)fa(x) = afa(x)
As a consequence,
d [(1 - x)afa(x)] = -a(1 - X)a-,fa(x) + (1 - x)af,(x)

_ -a(1 - 40-If (X) + all - x)°-'fa(x) = 0.
Therefore, (1 - x)a fa(x) is equal to a constant for all x, I x I < 1. But f.(0) = 1. Thus
(1 - x)afa(x) = 1; that is,
fa(x) = (1 - x)-a,
which proves the result. 0
The Beta Function
There are a number of important integrals that can be expressed in terms of the Gamma
function. Some of these, which can be obtained by a change of variables, are given in
the exercises. There is one integral, however, that is very important and thus we state it
as a theorem. Since the proof is nontrivial and would take us too far astray, we state the
result without proof. For a proof of the theorem, see Theorem 8.20 in the text by Rudin.
8.8.5 THEOREM For x > 0, y > 0,
r(x)F(y)
J' t'`-'(i - ty-' dt =
o F(x + y).
The function
r(x)r(y)
B(x' A T(x + y)' x, y > 0'
is called the Beta function.
EXERCISES 8.8
1. 'a. Compute r(z), r('2222).
b. Prove that for n E N,
CIn+2)= (2n)! n .
2. By making a change of variable, prove that
r(x)= f0(In- I dt, 0<x<oo.

3. Evaluate each of the following definite integrals.
*a I edt
0 t
b. f
dt, n E N.
0(
4. By making the change of variable t = sin 2 u in Theorem 8.8.5, prove that
(Ar 1 r(n)r(m)
Jo (sin u)2"`'(cos u)z `' du = n, m > 0.
2 r(n + m)'
5. Evaluate each of the following integrals.
a f+.R
J0
(sin x)2" dx, nEN b. f
0
n!1
(sin x)'' *' dx, n E N
6. Use the binomial series and term-by-term integration to find the power series expansion of
Arcsinx = f (1 - t') -'/2 dt.

0
7. Let a > 0. Set (g) = 1 and fork E N, set
(a a(a-1)(a-2) (a-k+l)
\k/ k!
Note, if m e N,
i
k k! (m - k)!'
k 5 m, and (k) = 0 fork > m.
a. Prove that the series I:-k-0(k) x' converges uniformly and absolutely for x E H, 11.
b. Prove that xt = (1 + x)°, x E [-1, 1 ].
NOTES
Without question the most important concept of this chap- entiable. Although this construction is much easier, the
ter is that of uniform convergence of a sequence or series partial sums of the series defining the function f are them-
of functions. It is the additional hypothesis required in selves not differentiable everywhere. Thus it is not so sur-
proving that the limit function of a sequence of continu- prising that f itself is not differentiable anywhere on It
ous or integrable functions is again continuous or inte- The proof of the Weierstrass approximation theorem
grable. As was shown by numerous examples, pointwise presented in the text is only one of the many proofs avail-
convergence is not sufficient. For differentiation, uniform able. A constructive proof by S. N. Bernstein using the so-
convergence of {f"} is not sufficient; uniform conver- called Bernstein polynomials can be found on page 107 of
gence of the sequence of derivatives (f' ) is also required. the text by Natanson listed in the Bibliography. The proof
The example of Weierstrass (Example 8.5.3) is inter- in the text, using approximate identities, was chosen be-
esting for several reasons. First, it provides an example of cause the technique involved is very important in analysis
a continuous function which is nowhere differentiable on and will be encountered later in the text. In Theorem
R. Furthermore, it provides an example of a sequence of 9.4.5 we will prove a variation of the Weierstrass ap-
infinitely differentiable functions that converges uniformly proximation theorem. At that point we will show that
on R, but for which the limit function is nowhere differ- every continuous real-valued function on [-vr, w] with
entiable. Exercise 7 of Section 8.5 provides another con- f(-ir) = f(r) can be uniformly approximated to within
struction of a continuous function f that is nowhere differ- a given e > 0 by a finite sum of a trigonometric series.
1. Using Miscellaneous Exercise I of Chapter 6 and the Weierstrass approximation theorem, prove the following: If
f E 9t.[a, b] and e > 0 is given, then there exists a polynomial P such that
tk PI < e.
£1_
2. DefinefonRby
{cex(i_:-). Ixi < I.

f(x) =
0, IxI 5 1.
where exp(x) = ex, and c > 0 is chosen so that f f(x) dx = 1. For A > 0, set fx(x) = i f(Ax).
a. Prove that fa E C(R) for all A > 0.
b. Prove that fa(x) = 0 for all x E R, I x I z A, and that f ,, fx (x) dx = 1.
c. Prove that for every 8 > 0, flasl1l} fa(r) dr = 0.
3. In this exercise we show how the trigonometric functions may be defined by means of power series. Define the
functions S and C on R by
_W (-1)k x2k,1 C _ (-1)k u

S(x)
k=o (2k + I)i (x) k-o (2k)! x
a. Show that the power series defining S and C converge for all x E R.
b. Show that S'(x) = C(x) and C'(x) = -S(x), x E R.
c. Show that S"(x) = -S(x) and C"(x) = -C(x).
d. Show that if f : R --)- R satisfies f "(x) = -f(x) with f(O) = 0, f'(0) = 1, then f (x) = S(x) for all x E R.
e. If f: R - R satisfies f"(x) = 1(x), prove that there exist constants c,, c2, such that f(x) = c,S(x) + c,C(x).
L Show that (S(x))2 + (C(x))2 = 1. (Hint: Consider the function f(x) _ (S(x))2 + (C(x))2.)
g. Show that C(x + y) = C(x) C(y) - S(x) S(y) and S(x + y) = S(x) C(y) + C(x) S(y) for all x, y E R.
Andrushkiw, J. W., "A note on multiple series of positive Mathf, P., "Approximation of Holder continuous func-
terms," Amer. Math. Monthly 68 (1961), 253-258. tions by Bernstein polynomials;' Amer. Math.
Billingsley, P., "Van der Waerden's continuous nowhere Monthly 106 (1999), 568-725.
differentiable function," Amer. Math. Monthly 89 Miller, K. S., "Derivatives of non-integer order," Math.
(1982), 691. Mag. 68 (1995), 183-192.
Blank, A. A., "A simple example of a Weierstrass func- Minassian, D. P. and Gaisser, J. W.. A simple Weier-
tion;' Amer. Math. Monthly 73 (1966), 515-519. strass function," Amer. Math. Monthly 91 (1984).
Boas, Jr., R. P., "Partial sums of infinite series and how 254-256.
they grow," Amer. Math. Monthly 84 (1977), Patin, J. M., "A very short proof of Stirling's formula,"
237-258. Amer. Math. Monthly % (1989),41-42.
Boas, Jr., R. P. and Pollard, H., "Continuous analogues Roy, Ranjan, "The discovery of the series formula for rr
of series;' Amer. Mark Monthly 80 (1973),18-25. by Leibniz, Gregory and Nilakantha," Math. Mag.
Cunningham, Jr., F., "faking limits under the integral 63 (1990),291-306.
sign;' Math. Mag. 40 (1967), 179-186. Sagan, H., "An elementary proof that Schoenberg's
Davis, P. J., "Leonhard Euler's integral: A historical pro- space filling curve is nowhere differentiable" Math.
file of the Gamma function;' Amer. Math. Monthly Mag. 65 (1992), 125-128.
66 (1959), 849-869. Schenkman, Eugene, "The Weierstrass approximation
French, A. P., " he integral definition of the logarithm theorem;' Amer. Math. Monthly 79 (1972),
and the logarithmic series;' Amer. Math. Monthly 65-66.
85 (1978), 580-582. Weinstock, Robert, "Elementary evaluations of
Kestleman, H., "Riemann integration of limit functions;' Ja a dr. Jo eos x2 dx, and Jo sin x2 dx; 'Amer.
Amer. Math. Monthly 77 (1970),182-187. Math. Monthly 97 (1990), 39-42.
Lewin, J. W., "Some applications of the bounded conver-
gence theorem for an introductory course in analy-
sis," Amer. Math. Monthly 94 (1987), 988-993.
Orthogonal Functions
y J and Fourier Series
9.1 Orthogonal Functions
9.2 Completeness and Parseval's Equality
9.3 Trigonometric and Fourier Series
9.4 Convergence in the Mean of Fourier Series
9.5 Pointwise Convergence of Fourier Series
In this chapter we consider the problem of expressing a real-valued periodic function of

period 2a in terms of a trigonometric series
00
2 ao + Y, (a cos nx + b sin nx),
n-l
where the an and bn are real numbers. As we will see, such series afford much greater
generality in the type of functions that can be represented as opposed to Taylor series.
The study of trigonometric series has its origins in the monumental work of Joseph
Fourier (1768-1830) on heat conduction in solids. His 1807 presentation to the French
Academy introduced a whole new subject area in mathematics while at the same time
providing very useful techniques for solving physical problems.
Fourier's work is the source of all modern methods in mathematical physics involv-
ing boundary value problems and has been a source of new ideas in mathematical analy-
sis for the past two centuries. To see how greatly mathematics has been influenced by the
studies of Fourier, one only needs to look at the two-volume work Trigonometric Series
by A. Zygmund (Cambridge University Press, 1968). In addition to trigonometric series.
Fourier's original method of separation of variables leads naturally to the study of or-
thogonal functions and the representation of functions in terms of a series of orthogonal
functions. All of these have many applications in mathematical physics and engineering.
Fourier initially claimed and tried to show, with no success, that the Fourier series
expansion of a function actually represented the function. Although his claim is false.
379
380 Chapter 9 Orthogonal Functions and Fourier Series
in view of the eighteenth-century concept of a function this was not an unrealistic ex-
pectation. Fourier's claim had an immediate impact on nineteenth-century mathemat-
ics. It caused mathematicians to reconsider the definition of "function." The question of
what type of function has a Fourier series expansion also led Riemann to the develop-
ment of the theory of the integral and the notion of an integrable function. The first sub-
stantial progress on the convergence of a Fourier series to its defining function is due to
Dirichlet in 1829. Instead of trying to prove, like Fourier, that the Fourier series always
converges to its defining function, Dirichlet considered the more restrictive problem of
finding sufficient conditions on the function f for which the Fourier series converges
pointwise to the function.
In the first section, we provide a brief introduction to the theory of orthogonal func-
tions and to the concept of approximation in the mean. In Section 9.2 we also introduce
the notion of a complete sequence of orthogonal functions and show that this is equiv-
alent to convergence in the mean of the sequence of partial sums of the Fourier series
to its defining function. The proof of the completeness of the trigonometric system
{ 1, sin nx, cos nx}' , will be presented in Section 9.4. In this section we also prove
Fejdr's theorem on the uniform approximation of a continuous function by the nth par-
tial sum of a trigonometric series. In the final section, we present Dirichlet's contribu-
tions to the pointwise convergence problem.
9.1 Orthogonal Functions

In this section we provide a brief introduction to orthogonal functions and the question
of representing a function by means of a series of orthogonal functions. Although these
topics have their origins in the study of partial differential equations and boundary value
problems,' they are closely related to concepts normally encountered in the study of
vector spaces.
If X is a vector space over R (see Definition 7.4.7), a function (,) : X X X -+ R is
an Inner product on X if
(a) (x, x) ? 0 for all x E X,
(b) (x, x) = 0 if and only if x = 0,
(c) (x, y) (y, x) for all x, y E X, and
(d) (ax + by, z) = a(x, z) + b(y, z) for all x, y, z E X and a, b E R.
In R", the usual inner product is given by
(a, b) = I, aab1
i=1
f o r a = (a,, ... , a") and b = (b,, ... , b") in R". If () is an inner product on X, then
two nonzero vectors x, y E X are orthogonal if (x, y) = 0. The term "orthogonal" is syn-
1. For a detailed treatment of this subject see the texts by Berg and McGregor or Weinberger listed in the
Bibliography.
onymous with "perpendicular;" and comes from geometric considerations in R". Two
nonzero vectors a and b in IB" are orthogonal if and only if they are mutually perpendic-
ular; that is, the angle 9 between the two vectors a and b is z or 90° (see Exercise 10, Sec-
tion 7.4).
In the study of analysis we typically encounter vector spaces whose elements are
functions. For example, in previous sections we have shown that the space 12 of square
summable sequences and the space 11[a, b] of continuous real-valued functions on
[a, b] are vector spaces over R. With the usual rules of addition and scalar multiplica-
tion, 9t[a, b], the set of Riemann integrable functions on (a, b), is also a vector space
over R. If for f, g E 9t[a, b] we define
U, g) = f f(x)g(x)
b dx,
I
then it is easily shown that (,) satisfies (a), (c), and (d) of the definition of an inner prod-
uct. It does not, however, satisfy (b). If a < b and ci, . . . , c" are a finite number of
points in [a, b], then the function
0, x # c,,
f(x) = 11 , x=ci,
is in 9t[a, b] satisfying (f, f) = 0, but f is not the zero function. Thus technically () is
not an inner product on 9t[a, b]; a minor difficulty which can easily be overcome by
defining two Riemann integrable functions f and g to be equal if f(x) = g(x) for all
x E [a, b] except on a set of measure zero. This will be explored in greater detail in
Chapter 10. Alternatively, if we restrict ourselves to the subset b] of 9t[a, b], then
(f, g) as defined above is an inner product on b] (Exercise 11).
Orthogonal Functions
We now define orthogonality with respect to the above inner product on 9t[a, b].
9.1.1 DEFINITION A finite or countable collection of Riemann integrable functions {¢n}

on [a, b] satisfying fa 40.' * 0 is orthogonal on [a, b] if
\Wn+ m) = &(X)4m(X) dx = 0 for all n * m.

1a
9.1.2 EXAMPLES
(a) For our first. example, we consider the two, functions ¢(x) = I and 4i(x) = x,
x E [ -1, 1]. Since
f f O(x)*(x) dx = f t X dx = 0,
the functions 46 and are orthogonal on the interval [-1, 11.

(b) In this example we show that the sequence of functions (sin nx}. , is orthogonal
on [-ir, ar]. By the trigonometric identity
sin A sin B = 2[cos(A - B) - cos(A + B)],
for n # m,
Jsin nx sin mx dx = J [cos(n - m)x - cos(n + m)x] dx

2 u
_ i sm(n - m)x sin(n + m)x _

0.
2 (n - m) (n+m)
For future reference, when n = m,
sine nx dx = 2 (I - cos 2nx) dx

-a
sin nXlI'
X- = Jr.
-p
2\
(c) As our final example, we consider the collection 11, sin ¶, cos T )"., on the in-
terval [ -L, L] where L > 0. As in (b), if n * m, then
1L
sinnlrx
L sin m7rx
L dx = 0.
L
Thus the collection (sin ¶ ) is orthogonal on [ -L, U. Also, by the trigonometric iden-
tities
cos A cos B = I [cos(A - B) + cos(A + B)]
sin A cos B = 2[sin(A - B) + sin(A + B)],
we have for n # m,
(L narx mirx (L nirx m7rx dx
cos L dx = 1 cos L
f
1 L cos
L
L sin
L
= 0.
Thus the functions in the collection {cos z} are all orthogonal on [-L, L] as are the
functions sin T and cos L" for all n, m E N with n * m. For m = n,
f sin nUx cos

L dx=
2n sin' "' ILL = 0.
This last identity shows that the functions sin Of and cos "LF are also orthogonal on
[-L. L] for all n E N. Finally, since
Jsindxr fcos
for all n E N, the constant function 1 is orthogonal to sin "lr` and cos " for all n E N.
In this example we also have
L
2 nWX L 2 ri?lX
dx = Co. .L dx = L.
f Lsin
L J
If in Example 9.1.2(b) we define ¢"(x) _ j sin nx, then the sequence {4"(x)}R ,
satisfies
{0, when n # in.
4)"(x)¢m(x) dx =
fA 1, when n = m.
Such a sequence of orthogonal functions is given a special name.
9.1.3 DEFINITION A finite or countable collection of Riemann integrable functions {di"}

is orthonormal on [a, b] if
Jb
10, when n * m,
d."(x)O,"(x) dx =
a
1; when n = m.
Given a collection {¢"} of orthogonal functions on (a, b], we can always construct
a family {fir"} of orthonormal functions on [a, b] by setting
where c" is defined by

b
c. = J ¢^(x) dx.
a
Approximation in the Mean

Let {O"} be a finite or countable family of orthogonal functions defined on an interval
[a, b]. F o r each N E N a n d c, ... , cN E R, consider the Nth partial sum
N
SN(x) _ c4,,(x) (1)
A natural question is, given a real-valued function f on [a, b], how must the coefficients
c" be chosen so that SN gives the best approximation to f on [a, b]? In the Weierstrass
approximation theorem we have already encountered one form of approximation; namely,

uniform approximation or approximation in the uniform norm. However, for the study
of orthogonal functions there is another type of norm approximation that turns out to be
more useful.
If X is a vector space over R with inner product (,), then there is a natural norm on
X associated with this inner product. If for x E X we define
IIxII= ,
then II II is a norm on X as defined in Definition 7.4.8. The details that II II is a norm is
left to the exercises (Exercise 12). The crucial step in proving the triangle inequality for
II II is the following version of the Cauchy-Schwarz inequality: For all x, y E X.
I(x,Y)I -s 11x11 11Y11.
The proof of this inequality follows verbatim the proof of Theorem 7.4.3. For the vec-
tor space R[a. b] with inner product (f, g) = fa f(x)g(x) dx, the norm of a function f,
denoted I I f 112, is given by
111112 = If
f E &[a, b], the problem to consider is, how must the constants c be chosen
in order to minimize the quantity
Ilf- SNII2 = f [f(x) - SN(x)]2dx?

a
This type of norm approximation is referred to as approximation in the mean or least

squares approximation. The following theorem specifies the choice of (ca) so that SN
provides the best approximation to f in the mean.
9.1A THEOREM Let f E gt[a, b] and let be a finite or countable collection of or-
thogonal functions on [a, b]. For N E N, let SN be defined by equation (1). Then the
quantity
b
[f(x) - SN(x)]2 dx
Ja
is minimal if and only if
fa f(x)-OR(x) dx
cn n = I , 2, ... N. (2)
fa 02(x) dx
Furthermore, for this choice of c,,,

N
rb
[.f(x) - SN(x)]2 dx =
jf2(x) dx - c2 I
b
.0,22(x) dx. (3)

eel a
Prior to proving the result, we give the following alternative statement of the pre-
vious theorem.
9.1.5 COROLLARY Let f E Ii(a, b] and let SN(x) = I:- I caaya(x) where the ca are de-
fined by equation (2). If TN(x) = 7.N_ 1 an E R, then
b
[f (x) - SN(X)]2 dx b [f(x) - TN(x)]2 dx,

f J
f o r any choice o f aa, n = 1, 2, ... , N.

Proof of Theorem 9.1.4. For fixed N E N,
b
0 5 J [f(x) - S,,(x)]2 dx
a
b f 2(x) dx - 2 J bf (x)SN (x) dx + J SN(x) dx. (4)

J a
By linearity of the integral (Theorem 6.2.1),

1bf N
(x)SN(x) dx cn bf (x)4in(X) dx.
a n°I Ja
Also,
hS2N(x)dx = bSN(x\ 'C4Y'n(x) f dx = c,J hSN(x)4n(x) dx.

J J / nI a
But
b N
SN(X)wn(X) = I ck I b 0k(X)0.(x) dx,

Ja a
which by orthogonality,
b
Cn 0(x) dX.
I
Therefore,
b
N= c .
jS(x)dx
N
n=1
2 +(
l Wn(X) dx.
JJJ
Upon substituting into equation (4) we obtain
0 G i [f(x) - SN(X)]2 dx
b
N C.
bf2(X) dX - 2 Cn bf(X)fn(X) dx + j- 6 'On(x) dx.
Ja nil 1a nil 1a
which upon completing the square
N
f 2(X) dx + b n(x) Ic b 2J L./ f bo-12
6 2
Ia -I aJ L
f. b n=1 fa n
The coefficients cn occur only in the middle term. Since this term is nonnegative, the
right side is a minimum if and only if
cn =
a n
With this choice of cn, we also obtain formula (3) upon substitution. U
9.1.6 EXAMPLE As was previously shown, the functions 01(x) = 1 and ¢2(x) = x are
orthogonal on [ -1, 11. Let f (x) = x3 + 1. Then
c f',f(x)01(x)
= l J (x3 + 1) dx = 1.
' f!14i(x)dx 2 I
and
f_',f(x)d2(x)dx (' 3
C2 = =- I3
(x,
+ x)dx = .
0z(x) 2 -, 5
Therefore, S2(x) = I + sx is the best approximation in the mean to f(x) = 1 + x3 on

[-I, 1]. The graphs off and S2 are given in Figure 9.1.
-1 -0.5 0 0.5 1
Figure 9.1 Graphs off and $2

9.1.7 DEFINITION Let I be a sequence of orthogonal functions on [a, b] and let

f E %[a, b]. For each n E 1, the number
f,,f(x)4n(x) dx
(5)
fa 4.'(x) dx
is called the Fourier coefficient off with respect to the system The series
I
00
=1
is called the Fourier series off. This is denoted by
00
AX) -- I cn r (x) (6)

n-1
Remark. The notation "--" in formula (6) means only that the coefficients in the
series are given by formula (5). Nothing is implied about convergence of the series'.
9.1.8 EXAMPLE In Example 9.1.2(b) it was shown that the sequence of functions
{sin nx},1 is orthogonal on [ -Tr, ir]. Since
n = 1, 2, .. ,
if f E Jt[ -ar, Tr], the Fourier coefficients c, n = 1. 2_ ... , off with respect to the or-
thogonal system {sin nx} are given by
c = ir- f (x) sin nx dx,
and the Fourier series off becomes
f (X) c sin nx.

=I
As indicated above, nothing is implied about convergence. Even
converge, it need not converge to the function f. Since the terms of the series are odd
functions of x, the series, if it converges, defines an odd function on [ -ir, vr]. Thus un-
less f itself is odd, the series could not converge to f. For example, if f(x) = 1, then
_ a
-if sinnxdx= - mIr cosnx
- =0.
f
1r 1
n
In this case, the series converges for all x. but clearly not to f (x) = 1.
Bessel's Inequality
For each N E N, let SN(x) denote the Nth partial sum of the Fourier series off, i.e..
N
SN(x) = 71 CIAn(x),
n=1
where the cn are the Fourier coefficients off with respect to the sequence of or-
thogonal functions on [a, b]. Then by identity (3) of Theorem 9.1.4.
b
f
N r0.
f2(x)dx- cnJ
Therefore,
fb
cn 0.2(x)dx < bf 2(X)dx.
J
a a
Since this holds for every N E J. by letting N -+ oo we obtain the following inequal-
ity.
9.1.9 THEOREM (Bessel's Inequality) If f E 9t[a, b] and {cn}R , are the Fourier coeffi-
cients off with respect to the sequence of orthogonal functions {tbn}". ,, then
ao ,!, rb
cn fa V'n(x) dx :5 f 2(x) dx.
n=I J
a
In Example 9.1.8 with f(x) = 1, f % f 2(x) dx = 2rr, and cn = 0 for all

n = 1, 2, .... Thus it is clear that equality need not hold in Bessel's inequality. How-
ever, there is one consequence of Theorem 9.1.9 that will prove useful later.
9.1.10 COROLLARY Suppose is a sequence of orthogonal functions on [a, b]. If

f E 9t[a, b], then
n-oo
lim f f (x)fn(x) dx = 0.
a
Ia (6.2(x) dx
Proof. Since f E 9t[a, b]. fa f2(x) dx is finite. Thus by Bessel's inequality, the series
c. f° d, converges. As a consequence,
Jim b
cn 46.(x) dx = 0,
n-+oo
a
and thus,
fboa f(x)On(x) dx = 0.
(x) -n faV
b 2(
n(x) dx
a
EXERCISES 9.1
1. *Let f(x) = sin rrx, ¢,(x) = 1, and 42(x) = x. Find c, and c2 so that S2(x) = c,4,(x) + c2O2(x) gives the best ap-
proximation in the mean to f on [ -1, 1 ].
2. a. Show that the polynomials Po(x) = 1, PI(x) = x. and P2(x) = 12x2 - 1 are orthogonal on [ -1, l ].
b. Let
(0, -1 <_ x < 0,
AX) 1, 0<-xs 1.
Find the constants co, c,, c2, such that S2(x) = cnPo(x) + c^(x) + c2P2(x) gives the best approximation in the
mean to f on [-1, 1].
3. +a. Let 00(x) = 1, ¢t,(x) = x - a,, 42(x) = x2 - a2x - a3. Determine the constants a,, a2, and a3, so that
{40, m,, 42} are orthogonal on [0, I ].
b. Find the polynomial of degree less than or equal to 2 that best approximates f(x) = sin rrx in the mean on
[0,1].
4. Let {4nr.- t be a sequence of orthogonal functions on [a, b]. For f, g E J1[a, b] with f - Man4n and g Ybn4n.
show that for a, jS E R,
(af + Pg) - 7, (aan + Pb.) ..

n" 1
S. a. Show that the sequence {sin is orthogonal on [0, err].
b. For f E 4t[0, in, show that the Fourier series off with respect to the sequence (sin nx) is given by
m
b. sin nx where b" = ? f (x) sin nx dx.
n=1 Ir o
'c. Find the Fourier series of f(x) = x on [0, ir] in terms of the orthogonal sequence {sin nx}.
6. a. Show that the sequence {1, cos nx}' , is orthogonal on [0, ir].
b. For f e %[0, +r], show that the Fourier series off with respect to the sequence {l, cos ax} is given by
I
ao + !
°O+
, an cos n x where
2 '
ao = - f (x)dx and a" f (x) cos nx d x , n = 1, 2, ..
2 nom, 7r Q 7r
o
'c. Find the Fourier series of f(x) - x on [0, zr] in terms of the orthogonal sequence 11. cos nx}.
7. If f E Jt[0, in], prove that lim
n-"D
fo f (x) sin nx dx = lim fo f (x) cos nx dx = 0.
n-+00
7p
8. Let {4n} be a sequence of orthogonal functions on [a, b]. If the series I an4n(x) converges uniformly to a func-
n"t
tion f(x) on [a, b], prove that for each n e N, a" is the Fourier coefficient of f.
9. Let {an} be a sequence in (0, 1) satisfying 0 < an+, < an < I for all n E N. Define 4" on [0, 1 ] by
0. 05x <an+i,
an+I x an+, an),
4,(x) =
-2(x - an)
, 1(an + an)
an - an+,
0, an<x:5 1.
Show that {4n} is orthogonal in 9t[0, l] and compute II4nII2 for each n E N.
.
10. Let Po(x) = 1, and for n E N let
P"(x) x2)", x E [-1, 1].

2"n! de (1 -
The polynomials P. are called the Legendre polynomials on [ -1, 1 ].
a. Find P1, P2, and P3.

b. Show that the sequence {Pn}moo is orthogonal on [-1, 1 J. (Hint: Use repeated integration by parts.)
11. For f.g E T[a, h] define (f, g) = fa f(x)g(x) dx. Prove that () is an inner product on b].
12. Let X be a vector space over l with inner product (,). Prove each of the following.
*a. I (x. y) l s 11 x I I 11 Y 11 for all x, y E X.
b. 11 11 is a norm on X.
13. Cauchy-Schwarz Inequality: Use the previous exercise to prove that if f, g E 3t[a. b], then
J hf (x) g(x) dx 12 < (t"12 g2(x) dx).

(x) d`) ( jah
14. Let X be a vector space over IIi with inner product Q. If {y,, ... , yn} are nonzero orthogonal vectors in X and
x E X. prove that the quantity II x - (c,y, + + cyn) II is a minimum if and only if
_ (x. Y;)
`' IIY;Ii-
for all i = 1.... , n. (Hint: Imitate the proof of Theorem 9.1.4.)
9.2 Completeness and Parseval's Equality

In this section we look for necessary and sufficient conditions on the sequence {0n} of
orthogonal functions on [a, b] for which equality holds in Bessel's inequality. To ac-
complish this it will be useful to introduce the notion of convergence in the mean.
9.2.1 DEFINITION A sequence of Riemann integrable function on [a, b] converges

in the mean to f E 9i[a, b] if
h
[f(x) -
lim
f
If we consider cJ&.[a, b] as a normed linear space with norm
f rh
111112 = f(x)dx
LJ
then convergence in the mean is nothing but convergence in norm as defined in Defi-
nition 8.3.9. Thus a sequence in 91t[a, b] converges to f E &[a, b] in the mean if
and only if lim 11f - f 112 = 0. Convergence in the mean is sometimes also referred to
as mean-square convergence.
It is natural to ask how convergence in the mean is related to pointwise or uniform
convergence. Our first theorem proves that uniform convergence implies convergence
in the mean. As should be expected, pointwise convergence is not sufficient (Exer-
cise 2). In the other direction, we will show in Example 9.2.3 that convergence in
the mean does not imply pointwise convergence, and thus certainly not uniform con-
vergence. There we construct a sequence (f.1 of Riemann integrable functions on
[0, 1 ] such that II f 11 2 -+ 0, but for which { f (x)} fails to converge for any
xE[0,1).
9.2.2 THEOREM I f f , f , , , n = 1, 2, . .. , are Riemann integrable on [a, b], and (f.) con-
verges uniformly to f on [a, b], then converges in the mean to f on [a, b].
Woof. Since the proof of this result is similar to the proof of Theorem 8.4.1, we leave
it as an exercise (Exercise 1).
9.2.3 EXAMPLE In this example we construct a sequence I f.1 on [0, 1] that converges to
zero in the mean, but for which { f (x)} does not converge for any x E [0, 1 ]. This se-
quence is constructed as follows: For each n E N, write n = 2k + j where k =
0, 1, 2, ... , and 0 s j < 2k For example, 1 = 21+0,2 = 2' +0,3 = 2' + 1, etc.
Define ff on [0, 1 ] by
1,
jksxsj+ k
1
2 2
0, otherwise.
of
The first four functions ff, f2, f3, and f4 are given as follows: fi(x) = 1, and
f 1, 0:5 z:S 2,
0<x<2,
fi(x)- {0,I, isx<
.fa(x) _ 1, 0 :5 :5 41
0, 1<xsl.
For each n e K f. E 3R[0, 1 ] with
J fn(X) dx = 1 dr = 2k
0 1
J/2'
Thus lim Jt; f n(x) dx = 0. On the other hand, if x E [0, 1 ], then the sequence { f (x)}
contains
n +00an infinite number of 0's and l's, and thus does not converge.
In the following theorem we prove that convergence in the mean of the partial sums
of the Fourier series is equivalent to equality in Bessel's inequality.
9.2.4 THEOREM Let {Wn}: I be a sequence of orthogonal functions on [a, b',. Then the
following are equivalent:
(a) For every f E 9t[a, b],
jb
- SN(X)]2 dx = 0,
limo
where SN is the Nth partial sum of the Fourier series off.

(b) For every f E 9t[a, b],
x b b
cn 1 4n(x) dx = J f2(x) dx. (Parseval's Equality)

n-I u a
where the cn are the Fourier coefficients off.

N
Proof. Suppose SN(x) _ cnd n(x) is the Nth partial sum of the Fourier series off.
n-1
Then by Theorem 9.1.4,
b b b
[f (x) - SN(x) ]2 dx = J f2(x) dx - Cn J 4' (x) dx.

Ja a n-1 a
From this it follows immediately that {SN} converges in the mean to f if and only if Par-
seval's equality holds. 0
9.2.5 DEFINITION A sequence {¢n}O I of orthogonal functions on [a, b] is said to be

complete if for every f E 9t[a, b],
ao
I Cn J 0n(x) dx = J f2(x) dx.

n-I u a
As a consequence of the previous theorem, the orthogonal sequence {0n} is com-

plete on [a, b] if and only if for every f E 9t(a, b], the sequence {SN} of the partial
sums of the Fourier series off converges in the mean to f. We now prove some addi-
tional consequences of completeness of an orthogonal sequence.
9.2.6 THEOREM If the sequence {¢n}f j of orthogonal functions on (a, b] is complete, and
if f is a continuous real-valued function on [a, b] satisfying
b
J f(x)4n(x) dx = 0 for all n = 1, 2, .. ,
a
then f (x) = O for all x e [a, b].

Proof. The hypothesis implies that the Fourier coefficients c off are zero for all
n E N. Thus by Parseval's equality,
fb
f2(x) dx = 0.
Ja
Since f2 is continuous and nonnegative, by Exercise 7, Section 6.1, this holds if and
only if f 2(x) = 0 for all x E [a, b]. Thus f (x) = 0 for all x E [a, b]. E)
There is a converse to Theorem 9.2.6. Since the proof of the converse requires a
knowledge of the Lebesgue integral, we only state the result. A sketch of the proof is
provided in the miscellaneous exercises (Exercise 3) of Chapter 10.
9.2.7 THEOREM If {4n}' , is a sequence of orthogonal functions on [a, b] for which the
only real-valued continuous function f on [a, b] satisfying
b
J f (x)4n(x) dx = 0 for all n = 1, 2, .
a
is the zero function, then the system is complete.

For the orthogonal system {sin nx}M-, on [ -vr, vr] and f (x) = 1, we have c = 0
for all n = 1, 2, .... Thus as a consequence of Theorem 9.2.6, the orthogonal system
{sin nx} is not complete on [ - vr, vr]. However, as we will see in Exercise 3 of Section
9.4, this system will be complete on [0, vr].
Another consequence of completeness is the following: Suppose {46,} is complete
on [a, b] and f, g are continuous real-valued functions on [a, b] satisfying
Jbf(x)On(x) dx = Jg(x)4a(x) dx
a a
for all n = 1, 2, ... , then f(x) = g(x) for all x E [a, b]. The above assumption sim-
ply means that f and g have the same Fourier coefficients. To prove the result, apply
Theorem 9.2.6 to h(x) = f(x) - g(x).
9.2.8 THEOREM Suppose is complete on [a, b], f, g E °k[a, b] with
f (x) 7, C.-On(x) and g(x) Y, b,,4 (x)

n-t n-I
Then
b oo b
f(x)g(x) dx = 7, cnbn J 46.1(x) dx.

f n-I a
Proof. Exercise 5. 0
El EXERCISES 9.2
2. For n E f0i, define the function f on [0, I ] by
f() =
v 0<x< !, n
0, elsewhere.
Show that converges to 0 pointwise but not in the mean.

nnx mrr '°
3. Consider the orthogonal system
I
1, cos
L
, sin
L _, on [ -L, L].
a. Show that if f E 9t[ -L, L], then the Fourier series off with respect to the above orthogonal system is given by
f(x)
I
2 ao + a cosnrx
L+b nirx
a,, = J f(x)cos nLx dx, n = 0. 1, 2, .... and b _ J f(x)sin' dx, n = 1.2.... .

L L L
b. Show that Bessel's inequality becomes

L
1
a-', + (a.' + b;) s 1 f2(x) dx.
L
2
4. *a. Assuming that the orthogonal system {sin nx} is complete on [0, in, show that Parseval's inequality becomes
Jnf(x)sinnxdx.
b2 _ 1 f 'f 2(x) dx, where b =
,,= 0
*b. Use Parseval's equality and the indicated function to find the sum of the given series.
I
(1) 7, (2k 1)2'
f(x) = I (ii) kI ji, f(x) = x
S. *Prove Theorem 9.2.8.

6. *Show by example that continuity off is required in Theorem 9.2.6.
9.3 1 Trigonometric and Fourier Series

In Section 9.1 we introduced Fourier series with respect to any system of
orthogonal functions on [a, b]. In this section we will emphasize the trigonometric
system
1, cos
nlrx,
L
--
sin
nTT x
L _,'
which by Example 9.1.2(c) is orthogonal on [-L, L]. For convenience we will take
L = Tr.
Any series of the form
2Ao + I (A cos nx + B. sin nx),

n=1
where the A. and B,, are real numbers, is called a trigonometric series. For example,
the series
00 sin nx
..2 In n
and I
MMI
cos nx
are both examples of trigonometric series. Since the coefficients
and
are nonnegative and decrease to zero, by Theorem 7.2.6 the first series converges for all
x E R, whereas the second converges for all x E R except x = 2pir, p E Z.
Fourier Series
For the orthogonal system {1, cos nx, sin nx} si on [-1r, 7r] we have
ff 4 n
sin 2 nx dx = cost nxdx = ir, and 12 dx = 21r.

f-n
I
-v 1 tr
Thus by Definition 9.1.7, the Fourier coefficients of a function f E 9t[ -a, ir] with re-
spect to the orthogonal system are defined as follows.
9.3.1 DEFINITION Let f E R[-7r, 1r]. The Fourier coefficients off with respect to the
orthogonal system (1, cos nx, sin nx} are defined by
ao = Jf(x)
-ir
dr,
1 "
an ,f (x) cos nx dx, n = 1, 2- ,
1 "
b.
b f (x) sin nx dx, n = 1, 2, ... .
J n
Also, the Fourier series off is given by
1 °O
AX) - 2ao + (a cos nx + b sin nx).
3% Chapter 9 Orthogonal Functions and Fourier Series
Remark. For the constant function 0o = 1, since f T, 020 = 21r, the term ao should be
defined as za f', f (x) dx, according to Definition 9.1.7. However, for notational conve-
nience it is easier to define ao as J°a f(x) dx and to include the constant 1 in the defi-
nition of the Fourier series.
For the orthogonal system ( 1, cos nx, sin nx), by Exercise 3(b) of the previous sec-
tion, Bessel's inequality 9.1.9 becomes
a
I ao +
00
(a + b2) <-L - -a
V(x))2 dx (Bessel's Inequality)
n-1
Thus for f E &[a, b] the sequences {a"} and {b,,} of Fourier coefficients off are square
summable sequences. In Theorem 10.8.7 we will prove that if {an} and {b"} are square
summable sequences, then there exists a Lebesgue integrable function f such that {a"}
and {b"} are the Fourier coefficients off.
Remark on Notation. In subsequent sections we will primarily be interested in real-

valued functions defined on all of N that are periodic of period 27r. As a consequence,
rather than defining our functions on [ -vr, vr], in the examples and exercises, we only
define the function f on [-vr, vr) with the convention that f(vr) = f(-r). This allows
us to extend f to all of R as a 2vr periodic function according to the following definition.
9.3.2 DEFINITION For a real-valued function f defined on( - ir, vr), the periodic exten-
sion (of period 21r) off to R is obtained by defining f (x) = f(x - 2kir). where k E 1
is such that x - 2kvr E [-vr, vr).
9.3.3 EXAMPLES
(a) Let f (x) tf0,

1,
-trsx<0,
0 <_x<V.
Then
ao=-faf(x)dx=1aldx= 1,
11
a 0
and for n = 1, 2, .. ,
fir
a" _
1
I
0
cos nx dx = n1
sin nx
Iff
0
= 0,
I
-I
IT
b" = 1yr sin nx dx =

nvr
cos nx -[1
n7r
- cosnvr] = n[1 - (-1)"].
I 0
0
In the above we have used the fact that cos nvr = (-1)". Thus the Fourier series off is
given by
00
1
f(x) ^- + na[1 - (-1)"]sinnx = 2 + yr T 2k + 1 sin(2k - I)x.
2 n -l k=0
If SN(x) denotes the Nth partial sum of the Fourier series, then S1 and S., are given by
2
S1(x) = + 2 sin x and SAX) = 2 + 2 [sin x + 3 sin 3x
7r ir
The graphs of f, S1. S3, S5, and S15 are given in Figure 9.2.
S1 S3 S5 S15
Figure 9.2 Graphs off, S1. S. S5, S15
(b) Let f (x) = x. Before we compute the Fourier coefficients, we will make several
observations that simplify this task. Recall that a function g(x) is even on [-a, a] if
g(-x) = g(x) for all x, and g(x) is odd if g(-x) = -g(x) for all x. By Exercise 4 of
Section 6.2, if g(x) is even on (-a, a], then
fa
g(x)dx = 21 ag(x)dx,
a 0
whereas if g(x) is odd,
a g(x) dx = 0.
! -a
The functions sin nx are all odd, whereas cos nx are even for all n. Therefore, since
f (x) = x is odd, x cos nx is odd and x sin nx is even. Thus. a" = 0 for all n = 0,
1,2,...,and
2
f
b= x sin nx dx,
a
which by an integration by parts

2
_ I n cosnxlr+ -IAcosnxdx
n
L a 0
=-ncosnar=n(-1)"+i
Therefore,
x, (-1)n+1
x -- 2 sin nx.
n=1 It
Riemann-Lebesgue Lemma
There is one additional result from the general theory that will be needed later. For the
orthogonal system { 1, cos nx, sin nx}, Corollary 9.1.10 is as follows.
93.4 THEOREM /f f E then
lim
n- 00
jf(x) cos nxdx = n-x
lim Jf(x) sin nxdx = 0.
Theorem 9.3.4 is commonly referred to as the Riemann-Lebesgue lemma. The

following example shows that integrability of the function f is required.
9.3.5 EXAMPLE Let f be defined on [-1T, Ir) as follows:

0, -irSx50,
f(x) 1, 0 < x < vr.
Ix
Then
(° A sin nx na sin x
f (x) sin nxdr = dx =
o o x
A
Hence,
lim J-"" f(x) sin nxdx = n-row

n-.00
lim f n" sinxdx = fO0sinx
x
dx 2
0 0
Is a Trigonometric Series a Fourier Series?

Since every Fourier series is a trigonometric series, an obvious question to ask is
whether every trigonometric series is a Fourier series. More specifically, given a
trigonometric series
00
2Ao + 7, (An cos nx + B. sin nx),
n-1
2. The value of a/2 for the improper integral is most easily obtained by contour integration and the theory
of residues of complex analysis. A real variables approach that computes the value of this integral is given in
the article by K. S. Williams listed in the supplemental readings.
with and converging to zero, does there exist a function f on [ -nr, a; such
that the coefficients A,, and B,, are given by Definition 9.3.1? As we will see, the answer
is no! First, however, in the positive direction, we prove the following.
9.3.6 THEOREM If the trigonometric series ;Ao + 7-(A cos nx + B. sin nx) converges
uniformly on [ -ir, a], then it is the Fourier series of a continuous real-valued function
on [ -?r, Tr].
Proof. For n E Nl, let
S (x) = 1 Ao + (Ak cos kx + Bk sin kx).
2 k-l
Since the series converges uniformly on [ -ir,1r], and S. is continuous for each n,
f (X) = urn S (x)
is a continuous function on [-ir, vr]. Form E N, consider
Jf(x) cos mxdx = L1( n-+w

lim S (x)) cos mxdx = lim A cos mxdx.
J
Since for each m, the sequence {S (x) cos mx} converges uniformly to f(x) cos mx on
[-ir, 1r], the above interchange of limits and integration is valid by Theorem 8.4.1. If
n > m, then
fSn(x)cosmx = 2Ao J cos mxdx + Ak r cos kx cos mxdx

k-1 i,
n a
+ Bk sin kx cos mxdx.
k-1
Thus by orthogonality,
S (x) cos mxdx = AmIr.
Letting n -> oo gives

91
A. = r
f (X) cos mx dx.
The analogous formula also holds for Bm, and thus the given series is the Fourier series
of f.
Remarks
(a) If 2 I Ak I and E I Bk I both converge, then by the Weierstrass M-test, the series
00
2 Ao + kY, (AA cos kx + Bk sin kx)
400 Chapter9 Orthogonal Functions and Fourier Series
converges uniformly on R, and thus is the Fourier series of a continuous function on

[ -ir, ir]. Convergence of the series I IAk I and 7, I Bk I is not necessary, however, for
uniform convergence of the trigonometric series. For example, the trigonometric series
1 sinn Innxn
°`
converges uniformly on 11 (Exercise 12), yet

x 1
= 00.
=2n In n
(b) In 1903, Lebesgue proved the following stronger version of Theorem 9.3.6: If
AX) = 2Ao + l ° t(Ak cos kx + Bk sin kx) for all x E (-1r, a), and if f is continuous
(in fact, measurable), then Ak and Bk are the Fourier coefficients of the function f.3 (See
also Miscellaneous Exercise 1.)
We now turn to the negative results. Consider the series
0 sin nx
a2 Inn (n
which by Theorem 7.2.6, converges for all x E R. However, there is no Riemann inte-
grable function f on [ -7r, or] such that
Inn =
b = If * f(x) sin nxdx.
n
If such a function f exists, then by Bessel's inequality we would have

00
n=2
hi
°,
=2 (Inn)'
' n
f2(x)dx.
jr, so is f2, and thus the integral is finite.

But since f is Riemann integrable on [-jr],
On the other hand,
°C
00,
n-2 (ln n)2
which gives a contradiction.
The above argument only shows that the series (7) is not obtained by means of
Definition 9.3.1 from a Riemann integrable function. There remains, however, the ques-
tion of whether this is still the case if we extend our definition to allow the class of
Lebesgue integrable functions to be introduced in Chapter 10. As we will see in Section
10.8, the answer to this is still no!
3. "Sur les series trigonometric:' Annales Scienriftques de I'Ecole Normale Supfrieure. (3) 20 (1903).
453-485.
Fourier Sine and Cosine Series

We close this section with a brief discussion of Fourier sine and cosine series. As we
have seen in Exercise 5 of Section 9.1, the sequence {sin nx}" I is orthogonal on [0, Tr].
Also, by Exercise 6 of Section 9.1, the same is true of the sequence {1, cos nx} -1. For
f E 9t[0, in, the Fourier series with respect to each of these two orthogonal systems
are called the Fourier sine and cosine series off respectively. Since
A
IT
sin2 nx dx = I cost nx dx =
0 0
the formulas of Definition 9.1.7 give the following.
9.3.7 DEFINITION For f E 9t[0, a], the Fourier sine series off is given by
00
f(x) 7, b sin nx,
n-I
where
rA
bn = ir J f (x) sin nxdx, n = 1, 2, ..
0
are the Fourier sine coefficients of f Similarly, the Fourier cosine series of
f E 9t[0, 7r] is given by
1
AX) - a0 + 00 a, cos nx,
n=
where
A
2 2
110 = f(x) dx and a=a f(x) cos nxdx, n = 1, 2, ..
0 0
are the Fourier cosine coefficients off.

There is a simple connection between Fourier series and the Fourier sine and co-
sine series. As in Example 9.3.3(b), we first note that if f is an even function on
[-ar, vr], then by Exercise 1,
00
1
AX) a0 + I a cos nx, where a _ jf(x)cosnxdx.
2 n-l 0
Similarly, if f is an odd function on [ -Tr, ir;, then
00 A
2
f (x) -r b,, sin nx, where b = f (x) sin nxdx.
n=1 77 I
0
Thus the coefficients and depend only on the values of the function f on
[0, ar]. Conversely, given a function f on [0, ir), the following definition extends f both
as an even and an odd function to the interval (-i, ir).
9.3.8 DEFINITION Let f be a real-valued function defined on [0, ir). The even extension of
f to (-it, it), denoted f is the function defined by
f(-x). -a < x < 0,
fe(x) =
f (x), 0 s x < V.
Similarly, the odd extension off to (-a, a), denoted f0, is the function defined by
1-f(-X). -T < x < 0,
f,(x) = 0, x = 0,
P X), 0 < x < V.
It is easily seen that the functions f,(x) and fo(x) are even and odd respectively on
(-1r, it), and agree with the given function f on (0, ar). For the function f given by the
graph in Figure 9.3, the graph of the even and odd extensions off are given in Figures
9.4(a) and (b) respectively. From the above discussion it is easily seen that if
f E 9t[0, ir], then the Fourier sine series off is equal to the Fourier series of f and the
Fourier cosine series off is equal to the Fourier series of f, (Exercise 2).
Figure 9.3
A -A
Figure 9.4
(a) Even Extension off (b) Odd Extension off
Remark, In the definition of the even and odd extension we only assumed that f was
defined on [0, a), and then defined f, and f, on (-a, Ir). If
Air-) = xlim_ f(x)
exists, then for the even extension we define f,(- v) = f(zr-). Thus f, is now defined
on [-ir, ir) and hence can be extended to all of TlR as a 2v-periodic function. For the
odd extension f we set f0(-w) = f(ir-), thereby defining f on [-1r, a).
EXERCISES 9.3
1. Let f E 9t(-a, ar). Prove the following.
'a. If f is even on [ -a, in, then the Fourier series off is given by
fv f (x) cos nxdx, it = 0, 1, 2, .
Ax) ^- 1 as + Y , a,, cos nx, where a
2 n-I o
b. If f is odd on [ -7, a], then the Fourier series off is given by

OD
f (x) ^- 7, by sin nx, where b _ ? f (x) sin nxdx, n = 1, 2, ..

N-1
2. Let f E 9t[0,ir]. Prove the following.
a. The Fourier sine series off is equal to the Fourier series of the odd extension f, off.
b. The Fourier cosine series off is equal to the Fourier series of the even extension f, off.
3. Find the Fourier series of each of the following functions f on [ -a, 1r):
-ar s x < 0, 0, -IT X :S 0,
a f(x)-{ 1,
1, 0sx<ir b f(x)-{x, 0<x<ar
'c. f(x) = lx( d. f(x) = x2
e. f(x) = I + x
0, -rr <_ x < -2,
1, 0 x< 2,
0, 2 x< in
4. On the interval [-2A, 2Tr) sketch the graph of the 2w-periodic extension of each of the functions in Exercise 3.
S. Find the Fourier sine and cosine series on [0, ar) of each of the following functions.
0 is
'a. f(x) = I b. f(x) = x c. f(x) = x2 'd. f(x) _ {0' x <a
2
f(x)-{-1, 0sx<2, f. f(x) = e-'

1, 25 x<ir
6. Let h be defined on [0, in) as follows:
cx, OsxsT2-
h(X) = 0 -4 2<x<a,
where c > 0 is a constant.
a. Sketch the graph of h on [0, or).

b. Sketch the graph of the even extension h, on (- 7r, 9r).
c. Find the Fourier series of h,
d. Sketch the graph of the odd extension h on (- IT. 7r).
e. Find the Fourier series of h,,.
7. a. Find the Fourier series of f (x) = sin x on [ - a, a].
b. Find the Fourier cosine series of f(x) = sin x on [0, zr).
8. *Suppose f, f are continuous on [-ir, Ir]. and f" E 91[ -7r, or]. Also, suppose that f(-v) = f(ir) and
f '(-vr) = f'(ar). Prove that the Fourier series off converges uniformly to f on [ -or, zr]. (Hint: Apply integration
by parts to show that Jail and Ib,l are less than C kfor some positive constant C.)
9. Show that the trigonometric series
sinnx
n=i n
converges for all x E IR but is not the Fourier series of a Riemann integrable function on [ -ir, zr].
10. If f is absolutely integrable on [ -it, zr] (see Section 6.4), prove that
n
n
lint J
n
f(x) cos nxdx = lim
J
f(x) sin nxdx = 0.
11. Using the previous exercise, show that each of the following hold.
a. lim J In x sin nxdx = nix

n-,x I
lim In x cos nxdx = 0
n 1o
b. lim J n xa sin nxdx = lim x° cos nxdx = 0 for all a > -1

n-.oc 10
0
12. Suppose an ? 0 for all n E N, and {nan} is monotone decreasing with lim nan = 0. Prove that J. ,an sin nx
converges uniformly. (Hint: Write I an sin nx as I (nan) (" n') and use the fact that the partial sums of ";"'
are uniformly bounded. See also Exercise 15 of Section 8.2.)
9.4 Convergence in the Mean of Fourier Series

Let f be a real-valued function defined on [ -1r, ir), and extend f to all of R to be peri-
odic of period 21T. Throughout this section we will assume that f e 9t[-v, ir]. Let
AX) ^ -ao + (an cos nx + bn sin nx)

2 n=1
be the Fourier Series off, and for each n E N, let

n
S,(x) = ao + (ak cos kx + b* sin kx)

2
be the nth partial sum of the Fourier series.

Our goal in this section will be to prove that if f E Jt[-a, irJ, then the sequence
converges in the mean to the function f on [ - ir, or]. By Theorem 9.2.4, this is
equivalent to completeness of the trigonometric system (1, cos nx, sin nx}. To investi-
gate mean-square convergence, and also pointwise convergence in the next section, it
will be useful to obtain an integral expression for S..
The Dirichlet Kernel

9.4.1. THEOREM Let f E 9t[-Tr, or]. Then for each n E N and x E R,
Sn(x) - V JA f(t)D,.(x - t) dt.

A
where D. is the Dirichlet kernel, given by
sin (n + j)t
2sin Z
t#2pir,pEZ,
D (t) = 1 + 7, Cos kt =
2 k-t
n+2, 1
t=2pir.
Proof. By the definition of the Fourier coefficients ak and bk (Definition 9.3.1),
S (x) = 1ao + 7, (ak cos kx + bk sin kx)

2 k-t
f 2 f (r) dt + ' (-'-j"f(r) cos kt dt) cos kx

',
+ k7, (1 f(t) sin kt dt sin kx
.1 ir J
It
= jA At) 12 +
'r j
1 1 "
f (t) - + cos k(x - t) J dt.
A
2 k-1
In the last step we have used the trigonometric identity

cos k(x - t) = cos kx cos kt + sin kx sin ki.
Set D (s) = z + eke I cos ks. Then by the above,
J t) dt.
A
To conclude the proof, it remains to derive the formula for If s = 2pir, p E Z.

then
Y'I=n+1.
2 ka, 2
By identity (2) of Theorem 7.2.6, for s # 2pa,
1 + sin(n + j)s - sin Z _ sin(n + Z)s

2 2 sin 2 2 sin 2
which establishes the result. L3
If the sequence I were an approximate identity, convergence results would

follow easily from Theorem 8.6.5. Unfortunately, the functions D are neither nonneg-
ative nor do they satisfy (b) of Definition 8.6.4. They do, however, still satisfy
I JDn(t) dt = 1,
A
(8)
a fact which will prove useful later. To see that equation (8) holds, it suffices to inte-
grate the sum defining term by term. The only nonzero term will be the integral
involving Z, and thus
Ltt=Lt= .
The graphs of D,, D3, and D5 are illustrated in Figure 9.5.
4T
Figure 9.5 Graphs of D,, D3, DS

The Fejer Kernel

To prove mean-square convergence, it turns out to be more useful to first consider the
arithmetic means of the partial sums S. For each n E N. set
o"(x) = n+1
By Exercise 14 of Section 2.2, if limS"(x) = S(x), then we also have
lim o"(x) = S(x).
However, it is possible that the sequence {S"(x)} diverges for a particularx, whereas the
sequence {o"(x)} may converge.
9.4.2 LEMMA For n E N, let S"(x) = iao + Y,k.1(ak cos kx + bk sin kx). Then
(a) o"(x) = Zao + J71 ( 1 - n + 1) (aj cos jx + bj sin jx)
and
A
(b) [1x) - S"(x)]2 dx < [1(x) - on(x)]2 dx.

J if j
Proof. The proof of (a) is left as an exercise (Exercise 1). Since o" is itself the par-
tial sum of a trigonometric series, the result (b) follows by Corollary 9.1.5.
Our next step is to express o"(x) as an integral analogous to that of S"(x) in the pre-
vious theorem.
9A.3 THEOREM Let f E Jt[-ir, zr]. Then (for each n E N and x E R,
o"(x) _ ! J f(t)F"(X - t) dt,
where F. is the Fejer kernel, given by
1 sin(n + I )L' 2 2p7r,
F"(t) = I r pk(t) -
n+lk,.o
2(n + 1)[
n+1
sin 2
t#
t - 2pzr.
2 '
The graphs of F1, F3, and F5 are illustrated in Figure 9.6.

3
Figure 9.6 Graphs of F,. F3. Fs
Proof. By Theorem 9.4.1,
fA s) ds,
where D. is the Dirichlet kernel. Therefore,
f(s)Fn(x - s) ds,
where
1 7, Dk(t).
n+ 1 k=0
If t = 2pir, p E Z, then Dk(2pvr) = k + , and thus
F 2or) =
"(P n+I2) =(n+1)
1 A0( k +
t
2
If t * 2p7r, p E Z, then
IT 1
k+ t
FF(t)
= 2(n + 1) sin 71 sin 2
sin sin + k + 2)t.

2(n + 1) sine 2' k=0 2
By the identity sin A sin B = Z(cos(B - A) - cos(B + A)],
r r
sin -sin k +
2
i2t
t=-
1
2k-0
(cos kt - cos(k + 1)t)
k-0
= (1 - cos(n + 1)t) = sin2(n + 1) 2

2
Therefore, for t * 2pir,
I sin(n + 1)_ 2
2(n + 1)L sin 2

O
We now prove the following properties of the Fejer kernel.
9.4.4 THEOREM
(a) F. is periodic of period 21r with
F (t) i? 0 for all t.
(c) IV
dt = 1.
1
(d) For 0 < 8 < ir, 1im 0 uniformly for all t, 8 s 1 tt c ir.
Proof. (a) Clearly Also, since
sin{ (n 1) (t + 2zr)) = sin( (n + 1) 2) cos(n + 1)ir = (- l)"'sin(n + 1) 2,

2
and sin 12(t + 21r) = -sin 2, substituting into the formula for F. gives
F(t). Therefore, F. is periodic of period 21r. The proof of (b) is obvious.
(c) Since f DD(t) dt = 1, we have
A
n
if F"(t) dt = Dk(t) dt = 1.
tr
n + 1 ka0 ;
To prove (d), we first note that sine 2 --- sine I for all t, 0 < 8 Itt ir. Also,
since Isin(n + 1);I s 1 for all t,
F (t) < 1 2a for all t, 8 :s Itt

2(n + 1) sin 1
Therefore, for any S, 0 < 8 < -tr, 1im F (t) = 0 uniformly on 6 s 1 tl <_ ir
410 Chapter 9 Orthogonal Functions and Fouler Series
As a consequence of (b), (c), and (d), the sequence FF(t)},°_, is an approximate

identity on [- or, or]. If f is a real-valued function on R that is periodic of period
2or, then
on(X) _ - I t)dt,
J n
which by the change of variable s = t - x,
=I w-zf(s + x)FF(s) ds.

I
Since the function s -+ f(s + x)F (s) is periodic of period 21r, by Theorem 8.6.3
on(x) f (s + x)FF(s) ds.

I
n
Thus if f is continuous on [-or, a] with f(-1r) = f(or), by Theorem 8.6.5,
limo,(x) = f(x)
uniformly on [ -7r, 1T). We summarize this in the following theorem of L. Fejer

(1880-1959).
9.4.5 THEOREM (Fej4r) If f is a continuous real-valued function on [-or, or] with

f (-or) = f (7r), then
f (x)
uniformly on [ -ot, or].
9A.6 COROLLARY If f is a continuous real-valued function on[-1r, or] with f(-v) = f(or),
then
n
lim [f(x) - SS(x)]Z dx = 0.
_n
Proof. The proof of the Corollary is an immediate consequence of Lemma 9.4.2(b).
Remark. There is a similarity between Theorem 9.4.5 and the Weierstrass approxi-
mation theorem. An alternative way of expressing. Theorem 9.4.5 is as follows: Let! be
continuous on [ -vr, vr] with f(-v) = f(ir). Given e > 0, there exists a trigonomet-
ric polynomial
n
T (x) _ Ao + (Ak cos kx + Bk sin kr)
2 k-I
such that
I1(x) - E
for all x E vr, vr]. The function is called a trigonometric polynomial since in
complex form it can be rewritten as
cke'kx,
T.(x) = L;
k--
where by De Moivre's formula e' = cos kx + i sin kx.
Convergence in the Mean

Corollary 9.4.6 only proves convergence in the mean for the case where f is a continuous
real-valued function on [ -vr, n]. We ate now ready to prove the main result of this section.
9A.7 THEOREM /f f E 9t[-vr, vr], then
Jim f [f(x) - S,(x)]2 dx = 0,
where S. is the nth partial sum of the Fourier series off.

For the proof of the theorem we require the following lemma.
9A.8 LEMMA, Let f E 3t[a, b] with I f(x)I s M for all x E [a, b). Then given e > 0,
there exists a continuous function g on [a, b] with g(a) = g(b) and I g(x)I <- M for all
x E [a, b], such that
b
f
a
If(x) g(x)I dx < e.
Before proving the lemma, we will first use the result to prove Theorem 9.4.7, then
consider some consequences of this. theorem.
Proof of Theorem 9.4.7. Suppose I f(x)I <_ M with M > 0, and let e > 0 be
given. By the lemma, there exists a continuous function g on [-n. vr] with
g(-vr) = g(vr) and Ig(x)I 5 M for all x E [-vr, vr] such that
A
f If(x) - g(x)I dx < 80 (9)

A
Let S (g)(x) be the function defined by
S0(gxx) = g(t)D0(x - t) dt,

aJ
where D. is the Dirichlet kernel. Then since g is continuous, by Corollary 9.4.6 the se-
quence {S (g)} converges in the mean to g on [ - ir, it ]. Thus there exists n,, E N such that
A [g(x) - S0(g)(x)]2 dx < 4

J
for all n no. Consider f ,[f(x) - S (g)(x)]2 dx. Since

f(x) - S0(gxx)l c I1(x) g(x)l + 1g(X) - S.(g)(x)l,
we have
If(x) - S0(g)(x)12 -_ 2[jf(x) - g(X)12 + lg(x) -
But by inequality (9),
11(x)-g(x)I2dx52MIf(x)-g(x)Idx<4
Therefore, for all n ? n,,,
[f(x) - Sn(g)(x)]2 dx < 2 + 2 A [g(x) - S0(g)(x)]2 dx < E.

1
However, for each n E N, S (g) is the nth partial sum of a trigonometric series. Thus if
S is the nth partial sum of the Fourier series off, by Corollary 9.1.5
[f(x) - S0(x)]2 dx s 1 A [.f(x) - S0(gxx)}2 dx.
Thus, given E > 0, there exists an integer n0, such that
[f(x) - E
for all n > n,; i.e., {S.) converges in the mean to f on [ -rr,1r). 0
As a consequence of Theorem 9.2.4 and Theorem 9.4.7 we have the following
corollary.
9.4.9 COROLLARY (Parseval's Equality) 1f f E %(-ir, ir], then
2 ao +
00
(a,2, + b.) fr f 2(X) dx.
I, I A
9.4.10 EXAMPLE By Example 9.3.3(b), the Fourier series of f(x) = x is given by

- 2(-1)".i
X_ sin nx.
n
Here an = 0 for all n, and b = 2(-1 r "/n. Thus by Parseval's equality,
00
4 2
_n X2dx=-a2,
n2 3
1
which gives
= 1r2
., _n2 6
0
Proof of Lemma 9.4.8. Let f E Jt[a, b] with if (x)I s M for all x E [a, b]. Since
f,(x) = f(x) + M is nonnegative, if we can prove the result for the function f,, the re-
sult also follows for the function f. Thus we can assume that f satisfies 0 <- f(x) s M
for all xE[a,b].
Let e > 0 be given. Since f is Riemann integrable on [a, b], there exists a partition
_ {xo, x,, ... , of [a, b] such that
0s jf(x)
b dx - f(9,f) < 2'
where
m;Ox;,
and m; = inf { f(t) : t E [x;_,, x;]}, i = 1 , 2, ... , n. For each i = 1, 2... , n, define
the functions h; on [a, b] by
mi, x..1 <_ x < X"
hi(x) =
0, elsewhere,
and let h(x) = "_, h,{x) (see Figure 9.7). The function h is called a step function
on [a, b]. Since his continuous except at a finite number of points, h is Riemann integ-
rable and
rb
h(x) dx n m;Ox; = Y(J', f).
J
Therefore,
0s I[f(x)-h(x)]dx<2.
b
a
m3 +
M5 t
M4 -4-
M2 t
mi t
mb + 0
I I I I i I I
xp=a X1 X2 X3 X4. X5 x6=b
Figure 9.7 Graph of the Step Function h
Also, since m; = inf{f(t) : t E [x;_ 1, x;]}, 0:5 h(x) s f(x) for all x E [a, b]. By tatt-
ing slightly shorter intervals and connecting the endpoints with straight line segments.
we leave it as an exercise (Exercise 9) to show that there exists a continuous function g
on [a, b] with g(a) = g(b) = 0, 0 s g(x) h(x), such that
Jb
[h(x) - g(x)]dx < 2.

a
Then 0 is g(x) s f(x), and

b b b
0 J [f(x) - g(x)] dx = J [f(x) - h(x)] dx + J [h(x) - g(x)] dx

a a a
<ZE+je=E. O
EXERCISES 9.4
1. Prove Lemma 9.4.2(a).
2. *Using the Fourier series of f(x) = x2 and Parseval's equality, find 1/n4.
3. *Prove that the orthogonal systems {sin nx} -I and {1, cos nx}; , are both complete on [0, a].
4. Show that the set {sin sin x, sin i, ...} is complete on [0, ir].
5. Let f E 9t[ -ar, ir], and let S. denote the nth partial sum of the Fourier series of f.
a. Using the Cauchy-Schwarz inequality for integrals (Exercise 13, Section 9.1), prove that
JIR
(Sa(x) - f(x))dxl 29rt Jp IS.(x) - f(x)I3dx
b. Prove that
A (!I
lim S.(x)dr = f(x)dx.

J
6. *Use the previous exercise to prove the following: Suppose f E 1t[ -7r, a] with
AX) - ao + (a cos nx + b sin nx).
For x E [-7r, ar], set F(x) = f af (t) dt - 1 Then F is continuous on [ v. ar] with F(-7r) = F(n). and
for all x E [-ir, ir],
F(x) = a,ir + (-fl sin nx - b" (cos nx - cos nIr)).

2 n n
where the convergence is uniform on [-a, Tr].

7. By integrating the Fourier series of f (x) = x (Example 9.3.3(b)), find the Fourier series of g(x) = x2.
8. Suppose f is a real-valued periodic function of period 2a on R with f E 1t[ -n, 1r], and .r E R is such
that f(x,-) and both exist. Prove that limo Z(f(xo-) + f(xo+)].
9. Complete the proof of Lemma 9.4.8; namely, given the step function h on (a. b] and e > 0. prove that there ex-
ists a continuous function g on [a, b] with 0 < g(x) <- h(x), g(a) = g(b) = 0, and fa [h(x) - g(x)] dx < e.
10. "Let f E &(a, b]. Given e > 0. prove that there exists a polynomial p(x) such that fa If(x) - p(x) I dx < e.
9. I Pointwise Convergence of Fourier Series

We now turn our attention to the question of pointwise convergence of a Fourier series
to its defining function. It is known that even if the function is continuous, this is not
always possible; additional hypothesis on the function f will be required. As was indi-
cated in the introduction, Dirichlet was the first to find sufficient conditions on the func-
tion f so that the Fourier series off converges to f. For the proof of his result we need
several preliminary lemmas.
9.5.1 LEMMA Let f be a periodic real-valued function on P of period 2Tr with

f E R[ -Tr, 7r]. Fix x E R. Then for any real number A,
v[f(x + t) f(x - t)
A=2J 2
- 1
d4
0
where D. is the Dirichlet kernel and S. is the nth partial sum of the Fourier series off.
Proof. By Theorem 9.4.1,
S. (x) J_/(tfl(X - t) dt,

which by the change of variables = x - t.

(-A +A
ds = 1 - f (x - s)D.(s) ds.
x+A A
Since both f and D are periodic of period 2a, by Theorem 8.6.3

I x4A 1 A
J - f (x - s)D.(s) ds = J f (x - s)D.(s) ds.

7r V
[ A A
Therefore,
S (x) _ f (x - s)D(s) ds
J 0
In the first integral set u = -s. Then since D (u),
1
Jo f(x - ds = -- J of(x + u)DD(u) du
A A
fn
= 1IT f(x + s)D.(s) ds.
0
Therefore,
S,(x) _ i1 A [f(x + s) + f(x - s))D.(s) ds.

0
Finally, since fo ds.= 1,
Z (1f(x+s) f(x- s)
Jo 2
which is the desired identity. U
9.5.2 LEMMA Ifg E 0Jt[0, ir), then
lim f ,g(t) sin(n + f t dt = 0.

2 0
Proof. Extend g to [-or, or) by defining g(x) = 0 for all x, -ors x < 0. Since
sin(n + J)t = cos i sin nt + sin i cos nt, we have
JA g(t) sinl n + 2) t dt = J Ag(t) sinl n + t dt

o
2)
rA A
=J gi(t) sin nt dt + g2(t) cos ,it dt,

A J -A
where g,(t) = g(t) cos 1 and g2(t) = g(t) sin 2. Since g is Riemann integrable on [-ar, ir;,
so are both g, and $2. Thus by Theorem 9.3.4,
A
(
lim 1 g,(i) sin nt dt = n-
lim g2(t) cos nt dt = 0,
nom= 00
A -A
from which the result follows. U
Dirichlet's Theorem
Before we state and prove Dirichlet's theorem, we briefly review some notation introduced
in Chapter 4. For a real-valued function f defined in a neighborhood of a given point x the
right and left limits off at x0, denoted f (x,+) and f (x,-) respectively, are defined by
f(x,+) = lim f(x) and f(x,-) = lim_ f(x),
provided, of course, that the limits exist.
9.5.3 THEOREM (Dirichiet) Let f be a real-valued periodic function on R of period 2ir

with f E Jt[-ir,1r]. Suppose x, E R is such that
(a) f (x,+) and f (x,-) both exist, and
(b) there exists a constant M and a S > 0 such that
If(x,+ t) - f(x,+)1 <- Mt and l f(x, - t) - f(x,-)I s Mt (10)
for all t,0<t<_a,then

limS.(xo) = 2[f(xo+) +f(xn-)]
Remark. If f is continuous at x then 2[f(x,+) + f(x,-)J = f(x,), and thus

1imoS (x,) = f (x,),
provided, of course, that the inequalities (10) hold.

Proof. Set A = 2[f(x,+) + f(x,-)]. By Lemma 9.5.1,
2IV[f(x°+t) 2
f(x°-t)
!I A[( (xo + t) -f(xo+))

0
lsin(n + ;)t
+ (f(x. - t) - f(xo-))1 2 sin i
dt
(A A
_ J g,(t) sin n + )tdt + I g2(t) sin (n + 2)tdt.

to
0 2
where
gi(t)
=f(x + t) -f(xo+)
and
=f(x,, - t) -f(x.-)
2sin1 g2(t)
2sin f
To prove the result, it suffices to show that the functions g, and g2 are Riemann inte-
grable on [0, Tr]. If this is the case, then by Lemma 9.5.2.
lim1gi(t)sinln+2ftdt=0, i = 1,2.
U
\
Therefore,
A = 2[f(xo+) +f(xo-)]-
To finish the proof we still have to show that g, E R[0, ar], i = 1, 2. We will prove
the result for the function g,, the proof for g2 being similar. Set h(t) = f(x, + t) -
f (x,+ ). Since f E Jt[ -vr, 1r] and is periodic of period 2ir, h is Riemann integrable on
[0, a]. Also, (sin Z)-' is continuous on (0, v] and thus Riemann integrable on [c. 1r] for
any c, 0 < c < ir. Therefore,
gilt) = 2 h(t)
sin
is Riemann integrable on [c, 17] for any c, 0 < c < a. Let S > 0 be as in hypothesis
(b). Then for 0 < t < S, using inequality (10),
If(; + t) - f(x,+)I
I g i (t)I = 12 stn 21
t
If(xo + t) -f(x,+)I
Itl Si SMlsinll.
Since
lirax-
X-,0' sin x
there exists a constant C such that
tt2
Z sC
sin 2t
for all t, 0 < t s ir. Therefore, g, is bounded on (0, S), and hence also on (0, w]. Thus
since g, is Riemann integrable on [c, ir] for any c, 0 < c < r, by Exercise 5 of Sec-
tion 6.2, g, is Riemann integrable on (0, w].
Piecewise Continuous Functions

Dirichlet's theorem required that f(x,-) and f(x,+) both exist at x0, and that f satisfy
the inequalities (10) at the point x0. The existence of both f(x,-) and f(x,+) means
that either f is continuous at x or in the terminology of Section 4.4, has a removable
or jump discontinuity at x0. Functions that have only simple discontinuities, and at most
a finite number on an interval [a, b], are said to be piecewise continuous on [a, b . We
make this precise with the following definition.
9.5A DEFINITION A real-valued function f is piecewise continuous on [a, b] if there

exist finitely many points a = x < x, < - < x = b such that
(a) f is continuous on (xi_,, x;) f o r all i = 1, 2, . .. , n, and
(b) for each i = 0, 1, 2, ... , n, f(x;+) and f(x;-) exist as finite limits.
For i = 0 and n, we of course only require the existence of the right and left limit
respectively. Also, we do not require that f be defined at xa, x1.... , The graph of a
typical piecewise continuous function is illustrated in Figure. 9.8.
xf b
Figure 9.8
In addition to piecewise continuous functions, it will also be useful to consider

functions for which the derivative is piecewise continuous. If f is piecewise continuous
on [a, b], then the derivative f' is piecewise continuous on [a. b), if there exist a finite
number of points a = xa < x, < . < x - b such that
(a) f(x) exists and is continuous on each interval (xi _ 1, xi), i = 1, 2,. .. , n, and
(b) for each i = 0, 1, 2, ... , n, the quantities f'(x;+) and f'(x;-) both exist as
finite limits. Again, for the endpoints a and b we obviously only require the
existence of f (a+) and f'(b-).
9.5.5 EXAMPLES
(a) Consider the function
x, 0<x<1,
f(x)- xZ+2, I <x<2.
Since f is continuous on (0, 1) and (1, 2) and f(0+), f(1-), f(1 +), and f(2-) all ex-
ist, f is piecewise continuous on [0, 2]. Also, since
f,(x)-{1, 0<x<1,
2x, 1 <x<2,
and f'(0+), f'(I-),f'(1 + ), and f'(2 -)all exist, f' is also piecewise continuous on [0, 2].
(b) As our second example, consider the function
0, x=0,
f(x) xI sin 1, O < x <- 1.
X
The function f is both continuous and differentiable on [0, 1 ] with
0, x=0,
'x
f 2x sin 1
x
cos 1,
x
0 < x < 1.
However, since f'(0+) does not exist, f' is not piecewise continuous on [0, 1].
Returning to Dirichlet's theorem, suppose f is such that both f'(x,-) and f'(x,+)
exist at the point x0. If f' is piecewise continuous on [ -Tr, in, then there exists x, > x
such that f and f' are continuous on (x x,). If f is not continuous at x redefine f at x,
as f (x,+ ). Then f (redefined if necessary) is continuous on [x, x,), and thus by Theo-
rem 5.2.11,
Ax. + Ax" +)
li o t) I
As a consequence, there exists a constant M and a S > 0 such that
If(xo + t) - f(x,+)I <- Mr
for all t, 0 < t < S. Similarly, the existence of f'(x,-) implies the existence of a con-
stant M and a S > 0 such that
If(x - t) -f(x,,-)I <_ Mr
for all t, 0 < t < S. Thus f satisfies the hypothesis of Dirichiet's theorem. On combin-
ing the above discussion with Theorem 9.5.3 we obtain the following corollary.
9.5.6 COROLLARY Suppose f is a real-valued periodic function on R of period 2a. If f

and f' are piecewise continuous on [ -Tr, Tr], then
xS.,(x) = 2[f(x+) +f(x-)]

for all x E R.
9.5.7 EXAMPLES
(a) Consider the function f defined on [-ir, ar) by
-ITsx<- 2'
-2. . 2,
2 <x<ar.
Extend f to all of U8 as a periodic function of period 2a. The function f is then continu-
ous except at x = 2 it + krr, k e Z. At each discontinuity x,,,
I[f(x"+) + f(xo-)] = 2
The graph of 2[f(x+) + f(x-)] on the interval [-27r, 21r] is given in Figure 9.9.
-O O-3 0 O-
-2A -3° -IT -° ° A 3° 2v

2 '- 2 2
Figure 9.9 Graph of 1[f(x+) + f(x-)]
To discuss convergence of the Fourier series off we first note that f is piecewise
continuous on [-1r, ar] and differentiable on (-1r, -2), (--, 1), and (Z, zr) with
f'(x) = 0 in the respective intervals. Thus f' is also piecewise continuous on [ -ir, ir].
As a consequence of Corollary 9.5.6, the Fourier series of f converges to
2[f(x+) +f(x-)] for all x E 08.
The Fourier series off is obtained as follows:
1
a2
ao=- J 3dx=3,
°/2
(2
6
a° 3cosnxdx=sin
n 2,n=1,2, ..
n/2
and since f is an even function, b = 0 for all n E Ni. Therefore,

3 6 nit
f (x)
2
+
Gnir- sin 2 cos nx.
n_l
If n is even, sin!' = 0. If n is odd, i.e., n = 2k + 1,
sin(2k + 1) = sin(ka + 2) sin = (-1)&.

Thus 2 2
I)k
AX) I cos(2k + 1) x.
+6 ka 2k
2
When x = 0, the series converges to f(O) = 3. As a consequence
3
3 6 `
°° (-1)t
0 2k+l'
which upon simplification gives

(-1)k - rr
00 2k + 1
k-o 4*
(b) Dirichlet's theorem can also be applied to the Fourier sine and cosine series of a
real-valued function f defined on [0,1r). As an example, consider the cosine series of
f(x) = x on (0, 7r). By Exercise 5(b) of Section 9.3, the cosine series of f(x) = x is
given by
x^ 2 4 00
I
k-I (2k 1
2 cos(2k - I)x.
1)'
Since the even extension f, off to [-a, Tr] is given by f,(x) x 1, the cosine series
of x is the Fourier series of I x I on [ -ir, 7r]. Since both f, and f; are piecewise con-
tinuous on [-ir, ir], the Fourier series converges to the 21r-periodic extension of I x
for all x E R. This function is illustrated in Figure 9.10.
-2n _p IT 2n
By Dirichlet's theorem, the Fourier series converges to I x I for all x E [-ir, rr].
Taking x = 0 gives
ir 4 00 1
0
2 k-l ( 2k- 1)2'
)
00
1 _ 7r -
.
.
or I (2k - 1)2 8
Remarks. Although the inequalities (10) of Theorem 9.5.3 are sufficient, they are by
no means necessary. There are variations of Dirichlet's theorem that provide sufficient
conditions on f to guarantee convergence of the series to 1 [ f (xo+) + f (xo-) ]. For ex-
ample, the inequalities of Theorem 9.5.3 can be replaced by the following:
If(xo + t) -f(xo+)1 < Mt° and If(xo - t) -.f(xo-)I c Mt° (11)

for all t, 0 < t < 8, and some a, 0 < a 5 1.1f f satisfies the above at x0, then the con-
clusion of Dirichlet's theorem is still valid. An even more general condition comes from
Ulisse Dini (1845-1918). who proved that if f E k[ -zr, zr] satisfies
o8lf(xo + t) - f(x"+)I
dt < oo and
J t
ff&If(x"-t)-.f(x.-)Idt<oo
to t
for some S > 0, then the Fourier series off converges to [f(x"+) + at x".
2
Both of the above hold if f satisfies inequalities (11) at x0.
Differentiation of Fourier Series

Our final result of this section involves the derivative of a Fourier series. Consider the
function f (x) = x. Since both f and f are continuous on [ - ir,1r], the Fourier series of
f converges to f for all x E (-,r, ir). Therefore, by Example 9.3.3(b),
2(-1)n''
x= sin nx, x E (-ii', Tr).
n=1 n
The differentiated series

00
2(-1)n+tcosnx
n=1
does not converge since its nth term fails to approach zero as n -> oo. The 21r-periodic
extension off has discontinuities at x = ±(2k - 1) it, k E Ni. It turns out that continu-
ity of the periodic function is important for differentiability of the Fourier series. Suffi-
cient conditions are given by the following theorem.
9.5.8 THEOREM Let f be a continuous function on [-ir, zr] with f(-ir) = f(v), and let
f be piecewise continuous on [-ir,,r]. If
Ax) = 1 ao + ',00(an cos nx + bn sin nx), x E [ -ir, it ],

2 n=t
is the Fourier series off then at each x E (-1r, ar) where f"(x) exists,
00
f'(x) = (-nan sin nx + nbn cos nx). (12)
Remark. At a point x where f"(x) does not exist but f"(x-) and f" (x+) both exist,
the series (12).converges to 2[f'(x-) + f'(x+)].
Proof. Suppose x E (-vr, ar) is such that f"(x) exists. Since fis continuous at x. by
Dirichlet's theorem
00
1
f' (x) = as + I (a" cos nx + P. sin nx).
2
Since f is continuous with f(-7T) = f(ir), and f E Jt.[-ar, zr],
as = f'(t)dt = 2 [f(l) - f(-a)] = 0.

2 J
Also, by integration by parts,
1
a" = - f'(t) cos nt dt
2ar
= 2[ f(ar) cos nar - f(-v) cos(-nir)] + tar 1 A f(t) sin nt dt
= nb".
Similarly, /3" = -na,,, which proves the result. O
9.5.9 EXAMPLE Consider the function f(x) = x2. Since f is even on (-ar, ar),
00
i
f (x) = 2 ao + a" cos nx.
Also, since f satisfies the hypothesis of Theorem 9.5.8,

00
f'(x) = Zr = 7, (-na" sin nx).
n.1
On the other hand, by Example 9.3.3(b),

4(-1)"+'
2x = "_I
7,
00
n
sin nx.
Therefore, a" = 4(-1)"/n2 for n = 1, 2, 3, . ... To find ao we use the definition. This
gives ao = 2ar2/3, and thus
.2f 2)"
x2 = + 4 001 ( cos nx, x E [-ar, 71.
3 "_1 n
Remarks. In closing this section, it should be mentioned that there exist continuous
functions f for which the Fourier series of f fails to converge at a given point. This was
first shown by P. du Bois Reymond. Other examples were subsequently constructed by
L. Fej6r and Lebesgue. The example of Fej6r can be found on p. 416 of the text by
E. C. Titchmarsh. Given any countable set E in (-7r, IT), it is possible to construct a

continuous function f on [-vr, 7r] such that the Fourier series off diverges on E and
converges on (-1r, 1r) \ E.4 In fact, it is possible to construct a continuous function on
[-7r, 7r] whose Fourier series diverges on an uncountable subset of (-a, rr).5 The ex-
istence of functions having such pathological behavior is due to the fact that for the
Dirichlet kernel D0,
(lr
lim I I Dn(1)I Lit = oc.
n- 00
0
The details are left to the miscellaneous exercises.

For all of the above examples, the Fourier series off converges to f except on a set
of measure zero. This is the case for the Fourier series of every Riemann integrable func-
tion f on [-ir, a]. In 1926, Kolmogorov showed that there exist Lebesgue integrable
functions whose Fourier series diverge everywhere.6 The biggest question about the con-
vergence of Fourier series was asked by Lusin: If a function f satisfies the hypothesis that
f2 is Lebesgue integrable on [-ir, ir], does the Fourier series off converge to f, except
perhaps on a set of measure zero? This problem remained unanswered for 50 years. The
first proof that this was indeed the case was provided by L. Carleson in 1966.'
EXERCISES 9.5
1. For each of the functions of Exercise 3, Section 9.3, sketch the graph of the function on the interval [ - 21r. 2v to
which the Fourier series converges.
2. ;Use the Fourier series of f(x) = x2 to find
CO (-1)M+1
Y,
n-t
n22 and I00
n=1
1
n2.
3. Using the Fourier cosine series of f(x) = sin x on [0, ar) (Exercise 7, Section 9.3), find each of the following
sums:
;a 1
sb (-1)"
4rt2 I Ls+t 4n2 - 1
4. a. Find the Fourier series of f(x) = (jr - IxI)2 on [-1r,1r].

b. On the interval [,-2r, 21r] sketch the graph of the function to which the series in part (a) converges.
c. Use the results of part (a) to find the sums of the following series:
00 CO
G 12 , a
4. See Chapter VIII of the text by A. Zygmund.

5. Ibid.
6. "Une s6rie de Fourier-Lebesgue divergente partout;' Compte Rendus, 183 (1926), 1327-1328.
7. "On convergence and growth of partial sums of Fourier Series" Acta Math., 116 (1966) 135-157.
S. a. Show that for -it < x < it,
e' =
eII
2
e-'[, - 2(cos;
(-1 nx - n sin nx)
'
2a nz+
b. Using part (a) find
W 1-11.-I W 1
G z and z
6. *a. Find the Fourier cosine series of e' on [0, ar).

*b. On the interval [-21r, 21r], sketch the graph of the function to which the series in part (a) converges.
7. Let f be a continuous function on [-ar, it] with f(-ir) = f(ir). If, in addition, f' is piecewise continuous on
[ -zr, a), prove that the Fourier series off converges uniformly to f on [ -tr, it].
& Suppose f E 9R[-ir,1r], and x, E (-it, ar) is such that f(x,-) and f(x,+) exist, and that
l f(x, f(xo+) . l f(xo - t)t f(x.- )
+ t)t
both exist as finite limits. Prove that f satisfies the hypothesis of Dirichlet's theorem (9.5.3).
9. Suppose f E 9t[-zr, ir] and x, E (-a,.it) is such that f(x,, ) and f(x,+) both exist If f satisfies the inequalities
(11) at x, for some a, 0 < a { 1, prove that the. Fourier series off at x, converges to x[f(x,+) + f(x,-)].
NOTES
There is no doubt about the significance of Fourier's con- The quest for an understanding of what types of functions
tributions to the areas of mathematical physics and ap- possessed Fourier series is partly responsible for the devel-
plied mathematics; one only needs to consult a text on opment of Lebesgue's theory of integration. The need for a
partial differential equations. The methods he developed more extensive theory of integration was illustrated by
relating to the theory of heat conduction are applicable to Lebesgue in 1903 in the paper Sur les series trigonometric.°
a large class of physical phenomena, including acoustics, In it he constructed an example of a function that is not Rie-
elasticity, optics, and the theory of electrical networks, mann integrable but that is representable everywhere by
among others. Fourier's work, however, is even more sig- its Fourier series. Such a function is f(x) _ -1a12 sin j1,
nificant in that it inaugurated a new area of mathematics. whose Fourier series is I. ,(cos kx)/k. This series con-
The study of Fourier series led to the development of verges everywere to the function f on [ - 7r, 1r), but since f is
the fundamental concepts and methods of what is now unbounded, it is not Riemann integrable on [ -ar, zr]. It is,
called real analysis. The study of the concept of a function however, integrable in the sense of Lebesgue. The article by
by Dirichlet and others was directly linked to their inter- Alan Gluchoff provides an excellent exposition on the influ-
est in Fourier series. The study of Fourier series by ence of trigonometric series to the theories of integration of
Riemann led to his development of the Riemann integral. Cauchy, Riemann, and Lebesgue.
He was concerned with the question of finding sufficient There are several important topics that we have not
conditions for the existence of the integrals that gave the touched upon in this chapter. One of these is the Gibbs
Fourier coefficients of a function f, that is, phenomenon, named after Josiah Gibbs (1839-1903).
To explain this phenomenon we consider the Fourier se-
a.= - f(x) cos nx dx and ries of f(x) = 0, x E [ -a, 0), and f(x) = 1, x E [0, or)
-II of Example 9.3.3 By Dirichlet's theorem we have
II
b= 1
f(x) sin nx dx. 8. Annales Scient s de PEcole Normale Supdrieure, (3) 20
-II (1903), 453-485.
Votes 427
0, -ar<x<0. of (0,7r] given by v , = j k. j = 0. 1, 2, ... , k. and

ti _;(Vi-,+vy),
2+-1( I 1) sin(2k-1)x= 2 , x0,
2i it l 3w
1, 0 < x < 7r. it sin + 3 sin
2k 2k
A careful examination of the graphs (see Figure 9.2) of (2k - 1)a I k
1
the partial sums S3, Ss, and S,s shows that near 0 each of
-I sin = R(r,) ;.
+ 2k 2k r
the functions S,, i = 3, 5, 15, has an absolute maximum at
a point t;, where the t; get closer to 0, but S,{t;) is bounded Therefore.
away from 1. n
We now consider the behavior of the partial sums S. (X
stz xx dr.
more closely. For n - (2k - 1), k E hl, 0
To approximate the integral we use the Taylor series ex-

S7k_,(x) = 2 + 2I sin x + sin 3x pansion of sin x. This gives
+
3
+ (2k 1) sin(2k - 1)x
x
sin x x2 x4 x6 xt 13t+st7t+9!-..
for all x E R. Therefore,

and
1 (" sin x
dt= Ir-3,31+S. St
1 Tr3
Sy,_,(x) = 2[cosx + cos3x +
ir
+ cos(2k - 1)x].
0 x
If we multiply SSt_,(x) by sin x and use the identity n-7 ir9
-7.T++9 91-
sinxcosjx = 2 [sin(j + I)x - sin(j - I)x], ,R2 ,T4 ,F6
=1 3.3!+5.5! 7.7!
we obtain
sin x Syk_,(x) = 1 sin kx.
- 1 - .54831 + .16235
From this it now follows that S2k_ t(x) has relative maxima
- .02725 + .00291
and minima at the points
.59 (to two decimal places).
2kx = ± 9r, ± 21r, .. , ± 2(k - 1)1r.
The notation'*-" denotes approximately equal to. There-
These points are equally spaced in (-,it, a). Consider the fore, 1.09. Even though ultimo S. (x) = 1
points xk = ar/(2k), at which each S3k_, has an (absolute) for all x E (0, Ir), if we approach zero along the points xt,
maximum with then S7k_, (xk) overshoots the value I by approximately
.09; i.e., kim LSy,_,(xk) - f(xt)l - 0.09. If the Fourier
series converges uniformly, this cannot happen.
Su_,(xk) = 2+ I sin 2k+ sin 2k
The above behavior, known as the Gibbs phenome-
3 non, is due to the fact that f has a jump discontinuity at
I (2k - 1)7r x0 = 0, and is typical of the behavior of the Fourier series
+ + Isin
2k of a piecewise continuous function at a jump discontinu-
ity. Furthermore, if f and f are piecewise continuous on
To find kiimaoS2k_,(xk), we write the above sum as a [ -a, 1r], then the amount of overshoot at a discontinuity
Riemann sum of the function g(x) = (sin x)/x. This func- x0, due to the Gibbs phenomenon, is approximately equal
tion is Riemann integrable on [0, a]. With the partition ' to 0.09[f(x0+) - The article by Shelupsky in
the supplemental readings provides a very nice discussion for all x E rr, rr]. where S. is the nth partial sum of the
of this phenomenon. series.
Another important question involves the uniqueness In 1870 Eduard Heine proved that if the sequence (S.}
of the representation of a function by a trigonometric se- converges uniformly to 0 on [ -ir, rr] then AA =Bk = 0 for
ries. Specifically, if all k. This is Theorem 9.3.6. The general uniqueness prob-
lem was solved by Cantor in the early 1870's. Furthermore,
1(x) _ 7, (Ak cos kx + Bk sin kx) he also proved uniqueness if equation (13) holds for all but
k-0 a finite number of x in [-rr, rr]. This then led Cantor to
(Ck cos kx + Dk sin kr) consider the uniqueness problem for infinite subsets of
k-0 [ - rr, rr]; specifically, if E C [ - ir, rr] is infinite and
for all x E [ -a, 'r], must Ak = Ck and Bk = Dk for all equation (13) holds for all x E [ -rr, a) \ E, must
Ak = Bk = 0 for all k? Since point set theory was undevel-
k = 0, 1, 2, . .? Alternatively, if
oped at this time, Cantor devoted much of his time and ef-
t
- Ao + T, (Ak cos kx + Bk sin kx) = 0 (13) fort to studying point subsets of R. For a thorough discus-
2 k-l sion of the uniqueness problem, refer to the article by
for all x E [ -rr, rr], must Ak = Bk = 0 for all k = 0, Marshall Ash listed in the supplemental readings, or to the
I.2, ... ? By equation (13) we mean that N-.OU 0 text by A. Zygmund listed in the Bibliography.
1. Suppose f is continuous on [ -rr, rr) and f(x) = i A0 + 7,k i (At cos kr + Bk sin kr). Let S. be the nth partial sum
of the series. If there exists a positive constant M such that IS.(x)I < M for alI x E [ -rr. rr] and n e N. prove
that Ak and Bk are the Fourier coefficients off.
2. Prove that lim j ID.(t)I dt = oo, where D, is the Dirichlet kernel. (See Example 6.4.4(b).)
3. Let f(x) = -In12 sin J. Show that the Fourier series of f is given by 1(cos kx)lk (Note: Since f is unbounded
at x = 0, the integrals defining ak and bk have to be interpreted as improper integrals.)
Ash, Marshall J., "Uniqueness of representation by Lanczos, Cornelius, Discourse on Fourier Series. Hafner
trigonometric series:' Amer. Math. Monthly % Pubi. Co., New York, 1966.
(1989),873-885. Mulcahy, C., "Plotting and scheming with wavelets;'
Askey, R. and Haimo. D. T., "Similarities between Math. Mag. 69 (1996), 323-343.
Fourier series and power series,' Amer. Math. Shelupsky, D., "Derivation of the Gibbs phenomenon;'
Monthly 103 (19%), 297-304. Amer. Math. Monthly 87 (1980). 210-212.
Gluchoff, Alan D., "Trigonometric series and theories of Shepp, L. A. and Kruskal, J. B., "Computerized tomog-
integration;' Math. Mag. 67 (1994). 3-20. raphy: The new medical X-ray technology;' Amer
Gonzalez-Velasco. Enrique A., "Connections in mathe- Math. Monthly 85 (1978), 420-439.
matical analysis: The case of Fourier series;' Amer. Simon, Barry, "Uniform convergence of Fourier series;'
Math. Monthly 99 (1992), 427-441. Amer. Math. Monthly 67 (1969), 55-56.
Halmos, R. "Fourier series;' Amer. Math. Monthly 85 Williams, K. S., "Note on f o (sin x/x) dx," Math. Mag.
(1978), 33-34. 44 (1971),9-11.
1 O Lebesgue Measure
and Integration
10.1 Introduction to Measure
10.2 Measure of Open Sets; Compact Sets
10.3 Inner and Outer Measure; Measurable Sets
10.4 Properties of Measurable Sets
10.5 Measurable Functions
10.6 The Lebesgue Integral of a Bounded Function
10.7 The General Lebesgue Integral
10.8 Square Integrable Functions
The concept of measure plays a very important role in the theory of real analysis. On
the real line, the idea of measure generalizes the length of an interval, in the plane, the
area of a rectangle, and so forth. It allows us to talk about the measure of a set in the
same way that we talk about the length of an interval. The development of the Riemann
integral of a bounded function on a closed and bounded interval [a, b depended very
much on the fact that we partitioned [a, b] into intervals. The notion of measure and
measurable set will play a prominent role in the development of the Lebesgue integral
in that we will partition [a, b] not into intervals, but instead into pairwise disjoint mea-
surable sets.
The theory of measure is credited to Henri Lebesgue (1875-1941) who, in his
famous 1902 thesis, defined measure of subsets of the line and the plane, as well as the
Lebesgue integral of a nonnegative function. Like Riemann, Lebesgue was led to the
development of his theory of integration while searching for sufficient conditions on
a function f for which the integrals defining the Fourier coefficients off exist. In this
chapter we will develop the theory of Lebesgue measure of subsets of I following
the original approach of Lebesgue using inner and outer measure. Although this ap-
proach is somewhat more tedious than the modern approach developed by Constantin
Carathdodory (1873-1950), it has the advantage of being more intuitive and easier to
visualize.
In the first section we will illustrate the need for the concept of measure of a set
and measurable function by considering an alternative approach to integration devel-
oped by Lebesgue in 1928. Although this ultimately will not be how we define the
429
430 Chapter 10 Lebesgue Measure and Integration
Lebesgue integral, the approach is instructive in emphasizing the concepts required for
the development of the Lebesgue theory of integration. In Section 10.2 we use the fact
that every open subset of R can be expressed as a finite or countable union of disjoint
open intervals to define the measure of open sets, and then of compact sets. These are
then used to define the inner and outer measure of subsets of R. A bounded subset of R
is then said to be measurable if these two quantities are the same.
In Section 10.6 we develop the theory of the Lebesgue integral of a bounded real-
valued function using upper and lower sums as in the development of the Riemann
integral. However, rather than using point partitions of the interval, we will use mea-
surable partitions. The key result of the section is that a bounded real-valued function
on [a, b] is Lebesgue integrable if and only if it is measurable. As we will see, the class
of Lebesgue integrable functions contains the class of Riemann integrable functions,
and for Riemann integrable functions, the two integrals coincide. One of the advantages
of the Lebesgue theory of integration involves the interchange of limits of sequences of
functions and integration. We will prove several very important and useful convergence
theorems, including the well-known bounded convergence theorem and Lebesgue's
dominated convergence theorem.
10.1 Introduction to Measure

In Definition 6.1.1 1 we defined what it means for a subset of lR to have measure zero.
In this chapter we will consider the concept of measure of a set in greater detail. When
introducing a new concept it is very natural to ask both "why," and "does it lead to
something useful?" Both of these questions were answered by Lebesgue in 1903 when
he exhibited a trigonometric series that converged everywhere to a nonnegative func-
tion f that was not Riemann integrable. ' The function f, however, is integrable accord-
ing to Lebesgue's definition and the trigonometric series is the Fourier series off.
In this section we will illustrate why it is necessary to consider the concept of mea-
sure of a set by considering an alternative approach to integration. As we will see later
in the chapter, this approach leads to greater generality in the types of functions that can
be integrated. In addition, Lebesgue's theory of integration allows us to prove inter-
change of limit and integration theorems without requiring uniform convergence of the
sequence of functions.
Let f be a bounded function on [a, b]. In developing the theory of the Riemann inte-
gral we partitioned the interval [a, b] and defined the upper and lower sums off corre-
sponding to the partition. Lebesgue's approach to integration is to partition the range of
the function, rather than the domain. For purposes of illustration, suppose f is nonnegative
with Range f C (0, /3). Let n E N, and partition [0, /3) into n disjoint subintervals
1U-1)nn j = 1, 2, , n.
See Figure 10.1 with n = 8.
1. Annales Scientifiques de 1'i`cole Normale Superieure. (3) 20 (1903). 453-485.

10.1 Introduction to Measure 431
Y
4
1$
1 ----- ----------------- -----

S: j I
{
11
6
{
1 1 I
{
x
a x1 ;2 -3 -'14 '1 b
Figure 10.1
For each j = 1, 2, ... , n, we let
Ej _ {x e [a,b]:(j-1) n <_ f(x) < ja}

n
In Figure 10.1,
E4= {xE [a, b] : 3 8 s f(x) < 4 8 } = [a, x1) U (xz, x3) U (xs, xs]
For a set such as E4, a finite union of disjoint intervals, it is reasonable to define the
measure of E,, denoted m(E,), as the sum of the length of the intervals, i.e.,
m(E4) = (x1 - a) + (x3 - x2) + (x5 - X,).
Assuming that we can do this for each of the sets Ep a lower approximation to the area
under the graph off is given by
f,
and an upper approximation is given by
i- j8m(Ei)
Taking limits as n -), oo, assuming that they exist and are equal, would in fact, provide
another approach to integration. That this -indeed is the case for a large class of func-
tions, including the Riemann integrable functions, will be proved in Section 10.6.
For nice functions, namely those for which the sets E, are finite unions of intervals,
the above is perfectly reasonable. However, suppose our function f is defined on [0, 1: by
f(x) = J0, x E a fl [0, 1],
x, elsewhere.
As above, for each j = 1, 2, . .. , n, let
E= {xE[O.l]:-1f(x)<i-}
n n
Thus E, = Q fl [0, 1 ] U {x irrational : 0 < x < ,-,}, and for j ? 2.
I
E, = x irrational : < x < i-
n n
Here the sets Ej are no longer unions of intervals, so what is meant by the measure of
the set is by no means obvious.
Our goal in the next two sections is to define a function A on a family .44 of subsets
of I8, called the measurable sets. The collection M will contain all the intervals and A
will satisfy the following:
(a) for an interval J, A(J) = length of J,

(b) A(E + x) = A(E) for all E E At and x E I8, where E + x = {y + x:y E E}.
and
(c) A(U Ek) _ A(Ek) for any finite or countable family {Ek} of pairwise disjoint
sets in .At.
10.2 Measure of Open Sets; Compact Sets

We begin our discussion of measure theory by first defining the measure of open and
compact subsets of R.
10.2.1 DEFINITION If J is an interval, we define the measure of J, denoted m(J), to be the

length of J.
Thus if J is (a, b), (a, b], [a. b), or [a, b], a, b E I8, then
m(J)=b-a.
If J is R, (a, oo), [a, oo), (-oo, b) or (-oo, b], we set m(J) = oo. In dealing with the
symbols oo and -oo, it is customary to adopt the following conventions:
(a) If x is real, then x + oo = oo, x - oo = -oo.
(b) If x > 0 then x oo = oo, x (-oo) -oo.
(c) If x < 0 then x - oo = - oo, x (- oo) = oo.
(d) Also, oo + co = oo, -oo -oo = =oo, oo (too) = ±oo,
-00*(±oo)=zoo.
The symbols oo - oo and -oo + oo are undefined, but we shall adopt the arbitrary
convention that 0 oo = 0.
Measure of Open Sets

Our first step will be to extend the set function m to the open subsets of R. Since this
extension relies on Theorem 3.1.13, we restate that result at this point.
THEOREM 3.1.13 If U is an open subset of R, then there exists a finite or count-

able collection {I } of pairwise disjoint open intervals such that
U=U In. R
Recall, the family {1"} is pairwise disjoint if and only if I" f l 1,n = 40 whenever n * m.
10.2.2 DEFINITION If U is an open subset of R with U = Unln, where {1"} is a finite or

countable collection of pairwise disjoint open intervals, we define the measure of U,
denoted m(U), by
m(U) m(I,).
Remarks
(a) For the empty set 0, we set m(¢) = 0.
(b) The sum defining m(U) may be either finite or infinite. If any of the intervals are of
infinite length, then m(U) = no. On the other hand, if
U=UI,,,
where the 1n are pairwise disjoint bounded open intervals, we may still have
00
m(U) = I m(1") = 00,
n-1
due to the divergence of the series to no. Since m(I,) ? 0 for all n, the sequence of par-
tial sums is monotone increasing and thus will either converge to a real number or di-
verge to no.
10.2.3 EXAMPLES
(a) Foreach n= 1,2,. ..,set
In=(n - 2",n+Zn
Then 11=(1 -2,1+1),12=(2-1,2+1),etc.

2 4 4 Since n+2-"< (n+1)-2''"+
for all n e N, the collection {I"}" 1 is pairwise disjoint. Let U = U' 0- 11n. Then
1 °D 1 1
m(U) = G m(I4) _ 2-2 = 2n
2.
n-1 n 11 n-0 12
The set U is an example of an unbounded open set with finite measure.

(b) Let J n = (n, n n = 1, 2, ... , and let V = U' I J,. Then
00 '° 1
m(V) m(J,) _ I - = oo.
n- n-In
(c) As in Section 3.3, let U denote the complement of the Cantor set in [0, 1]. Since
U is the union of the open intervals that have been removed, by Property 4 of the Can-
tor set,
m(U)= 1.
We now state and prove several results concerning the measure of open sets.
10.2.4 THEOREM If U and V are open subsets of U8 with U C V. then

m(U) 5 m(V).
Proof. The statement of the theorem appears to be so obvious that no proof seems to
be required. However, it is important to keep in mind how the measure of an open set
is defined. Suppose
U=UI, n
and V=UJm,
where {In}n and {Jm}m are finite or countable collections of pairwise disjoint open in-
tervals. Since U C V, each interval 4 C J. for some m. For each m, let
Since the collection {Jm}m is pairwise disjoint, so is the collection {Nm}m, and
U=U1.=UnU I,,.
Therefore,
m(U) _mnEN,
. m(I.).
But by Exercise 1,
I m(I4) c m(Jm),
FE N.
from which the result follows. Cl
Remark. As a consequence of the previous theorem, if U is an open subset of

(a, b), a, b E R, then
m(U)sb-a.
Thus every bounded open set has finite measure.
10.2.5 THEOREM If U is an open subset of R, then

m(U) kim m(Uk),
00
where for each k E 1N, U. = U fl (-k, k).

Proof. For each k, Uk is open, with
Uk C Uk+1 C U
for all k E N. By Theorem 10.2.4, the sequence {m(Uk)} is monotone increasing with
m(Uk) : m(U) for all k. Therefore,
rlim m(Uk) 5 m(U). (1)
oo
If U is bounded, then there exists ko E N such that
Ufl(-k,k)=U
for all k a ko. Hence m(Uk) = m(U) for all k k and thus equality holds in equa-
tion (1).
Suppose that U is an unbounded open subset of R with
U=UIi,n
where {In} is a finite or countable collection of pairwise disjoint open intervals. If

m(I,) = oo for some n, then m(U) = oo, and for that n, either In = R or In is an inter-
val of the form
(-oo, an) or (an, 00)
for some a E R. Suppose 1n = (an, oo). Choose k, E N such that k, a lad. Then for
all k at k0,
In fl (-k, k) = (an, k),
and thus
oo = 1im m(1, (1(-k, k)) s Item m(Uk) s m(U).
Therefore equality holds in equation (1). The other two cases follow similarly.
Suppose m(In) < oo for all n. Since U is unbounded, the collection {In} must be
infinite. If the collection were finite, then since each interval has finite length, each in-
terval is bounded, and as a consequence U must also be bounded. Let a E R with
a < m(U). Since
0
G m(I4) = m(U) > a,
n-1
there exists a positive integer N such that

N
> m(I4) > a.

M-1
Let V = UN 11,,. Then V is a bounded open set, and thus by the above.
m(V) =slim m(V fl (-k, k)).
Since m(V) > a, there exists k E N such that

m(V n(- k, k)) > a for all k a k0.
But v fl (-k, k) C Uk for all k E N. Hence by Theorem 10.2.4,
tn(V fl (-k, k)) s m(Uk)
and as a consequence
m(Uk) > a for all k ? ko.
If m(U) = oo, then since a < m(U) was arbitrary, we have m(Uk) -+ oo as k -> no.
If m(U.) < no, then given e > 0, take a = m(U) - e. By the above, there exists
k E N such that
m(U) - e < m(Uk) s m(U) for all k an k,,.
Therefore, lim m(Uk) = m(U).
Remark. In proving results about open sets, the previous theorem allows us to first
prove the result for the case where U is a bounded open set, and then to use the limit
process to extend the result to the unbounded case.
Our next goal is to prove the following.
10.2.6 THEOREM If { Un}n is a finite or countable collection of open subsets of R, then
tn(U U.) t n
m(U.).
For the proof of the theorem we require the following lemma.
10.2.7 LEMMA If {In}R i is a finite collection of bounded open intervals, then

N N
m( U In I M(In)-
n=1 n=1
It should be noted that the collection {In} is not assumed to be pairwise disjoint.
The lemma is most easily proved by resorting to the theory of the Riemann integral.
10.2.8 DEFINITION If E is a subset of fib, the characteristic function of E, denoted XE, is

the function defined by
1, x E E,
XE(x) = 0, x (4 E.
Suppose 1 is a bounded open interval. Choose a, b E R such that I C [a, b]. Since
Xf is continuous on [a, b] except at the two endpoints of 1, x, E k[a, b] with
Jb XI(x) dx = m(1).
It should be clear that this is independent of the interval [a, b] containing 1. If U is an

open subset of [a, b] with
m
U= UJk,
k=1
where the {Jk} are pairwise disjoint open intervals, then

m
Xu(x) = I Xj,(x)
k-I
and thus
n n
m(U) =7,, m(Jk) _I: Xj,(x) dx = J Xu(x) dx. (2)

k=I k=1 Ja a
P r o o f of lemma 10.2.7. Let {la} 1 be a finite collection of bounded open inter-

vals and let U = U =11,,. Choose a, b E I8 such that U C [a, b]. Then Xu is continu-
ous except at a finite number of points, and
N
Xu(x) t 7, X4(x).
n=1
Therefore by identity (2) and the above discussion,

e
m(U) = Jb Xu(x) dx <_

Ja
Xl.(x) dx =
Jb Xr,(x) _ , m(In) O
a n=I n=I a I
Proof of Theorem 10.2.6. Since the result for a finite collection follows obviously
from that of a countable collection, we suppose {Un}'° 1 is a countable collection of
open sets and
00
U= UIU,,.
Since U is open, U = U. J,n where {Jm}m is a finite or countable collection of pairwise

disjoint open intervals. Also, for each n,
U. = U In, k.
where for each n, {I,, k}k is a finite or countable collection of pairwise disjoint open in-
tervals.
We assume first that U is bounded with m(U) < oo. Let e > 0 be given. If the col-
lection {Jm}m is infinite, then since F'm m(Jm) < oo, there exists a positive integer N
such that
00
I m(J,) < e.
m-N+1
Thus
N
m(U) < Y, m(J,) + E.
M-1
If the collection {Jm}m is finite, the previous step is not necessary.

Consider the collection {Jm}m. 1. For each rn = 1, 2, ... , N. let K. be an open in-
terval such that
K. C Jm and m(Jm) <

N
Then
N N
M(U) < Y, m(J,,) + e < I m(K,) + 2e. (3)
M=1 m-I
Let A = U v=1 Km. Since each k. is closed and bounded, so is the set A. Thus A is com-
pact, and since
AC U=nU1(
the collection {I,,, k} is an open cover of A. Hence by compactness there exists a finite
number k,,, i = 1, ... , J. j = 1, . . . , mi, such that
ACUIn,.k
i.!
Since the intervals {Km} are pairwise disjoint
Y, m(K.) = m(U
m=I
Km) mU
aJ
m =1
which by Lemma 10.2.7

! m;
MV., k,) = 17, m(1 k, )

ij i-I !-I
.1 o0
`- m(U,,) s
!=1 n=I
Combining this with inequality (3) gives
m(U) < 2e.

n-I
Since e > 0 was arbitrary, the result follows for the case where U is bounded.
If U is unbounded, then for each k E N,
00 x
m(Ufl(-k,k))<_y m(u,, n(-k. k))52 m(U,).
n=1 n=1
The result now follows by Theorem 10.2.5. Q
10.2.9 THEOREM If U and V are open subsets of R, then

m(U) + m(V) = m(u U V) + m(u n v).
Proof. (a) If both U and V are unions of a finite number of bounded open intervals.
then so are u n V and U U V. Thus the functions
XU, Xv, Xuuv, Xunv
are all Riemann integrable on some interval [a, b] with
Xu(X) + Xv(x) = Xuuv(x) + Xunv(x)
for all x E [a, b]. Therefore by identity (2),
m(U) + m(V) = m(u U V) + m(u n v).
(b) For the general case, suppose
00
U= U
n-I
In and V = U Jn,
n=1
where the collections {In} and {J.} consist of pairwise disjoint open intervals respec-
tively. If one of m(U) or m(V) is oo, then by Theorem 10.2.4, m(U U V) = oo and the
conclusion holds. Thus we can assume that both m(U) and m(V) are finite.
Let e > 0 be given. Choose N E N such that
00 00
Y, m(I,) < e and m(JJ) < C.
n=N+I n-N+1
Let U * and U ** be defined by

N 00
U n- I In, U## In.

n N U+ I
Also let V* and V** be defined analogously. Then

m(U) = m(U*) + m(U**) and m(V) = m(V*) + m(V**).
Since m(U **) < e and m(V **) < e,
m(U) + m(V) < m(U*) + m(V*) + 2e.
Since the sets U* and V* are finite unions of open intervals, by part (a)
m(U*) + m(V*) = m(U* U V*) + m(U* n v*),
440 Chapter to Lebesgue Measure and Integration
which by Theorem 10.2.4

sm(UUV)+m(UnV).
The last inequality follows since both U * U V * and u * n V * are subsets of
U U V and u n V respectively. Hence
m(U)+m(V) <m(UUV)+m(UnV)+2e.
Since e > 0 was arbitrary, we have
m(U)+mm <_m(UUV)+m(UnV).
We now proceed to prove the reverse inequality. We first note that
U U V = (U * U V *) U (U ** U V **),
m(UUV)sm(U*UV*)+2e.
Also by the distributive law,
un v= (u- u u**)n(v* u v**)
= (U*nV*)U(U**nV*)U(U*nV**)U(U**nV**)
C(U*nV*)UU**UV**.
Therefore,
m(U n v) < m(U* n v*) + 2e.
Combining the above gives
m(UUV)+m(unv) < m(u* u v*) + m(u* n v*) + 4e,
which since U * and V * are finite unions of intervals
= m(U*) + m(V *) + 4e
s m(U) + m(V) + 4e.
Again since e > 0 was arbitrary, this proves the reverse inequality.
Measure of Compact Sets

We now define the measure of a compact subset of R. If K is a compact subset of R and
U is any bounded open set containing K, then
U=KU(U\K).
Using the fact that U.\ K is also open and bounded, and thus has finite measure, we de-
fine the measure of K as follows.
10.2.10 DEFINITION Let K be a compact subset of R. The measure of K, denoted m(K), is

defined to be
m(K) = m(U) - m(U \ K),
where U is any bounded open subset of R containing K.
10.2 Measure of Open Sets: Compact Sets 441
We first show that the definition of m(K) is independent of the choice of U.
10.2.11 THEOREM If K is compact, then m(K) is well defined.

Proof. Suppose U and V are any two bounded open sets containing K. Then by The-
orem 10.2.9,
m(U) + m(V\ K) = m(U U (V \ K)) + m(U n (V\ K))
=m(UUV)+m((UnV)\K).
In the above we have used the fact that
UU(V\K)= UUV and Un(V\K)=(UnV)\K.
Similarly,
m(U \ K) + m(v) = m(U U v) + m((U n V) \ K).
Therefore,
m(U) + m(V \ K) = m(V) + m(U \ K).
Since all the terms are finite,
m(U) - m(U \ K) = m(V) - m(V \ K).
Thus the definition of m(K) is independent of the choice of U. i.e., m(K) is well defined.
O
10.2.12 EXAMPLES
(a) In our first example we show that for a closed and bounded interval 1,a, b),
a, b E R, Definition 10.2.10 is consistent with Definition 10.2.1; namely,
m([a, b]) = b - a.
LetU=(a-E,b+ e),e>0.Then
U\[a,b] = (a - e,a)U(b,b + E).
Therefore,
m([a, bj) = m(U) - m(U \[a, b])
=(b-a)+2e-2e=b-a.
(b) If K = {xi, ... , with x; E R, then m(K) = 0. Choose 8 > 0 such that the in-
tervals
1j =(x,-S,xj +8), j=1,2,.. n,
are pairwise disjoint, and let U = U'.. 1;. Then
U \ K = 0(x,-6,x;)U(xx,x;+S).
Thus m(U) = m(U \ K); i.e., m(K) = 0. U

10.2.13 THEOREM
(a) If K is compact and U is open with K C U, then m(K) c m(U).

(b) If K, and K2 are compact with K, C K,, then m(K,) s m(K2).
Proof. The proof of (a) is an immediate consequence of the definition. For the proof
of (b), if U is a bounded open set containing K2, then since U \ K2 C U \ K,,
m(K,) = m(U) - m(U\K,) <- m(U) - m(U\K2) = m(K2). O
If I is an open interval and a, b E R, then i fl [a, b] is an interval and thus

my fl [a, b]) is defined with
m(I fl [a, b]) = m(1 fl (a, b)).
We extend this to open subsets of R as follows.
10.2.14 DEFINITION If U is an open subset of R with U = U,, 1,,, where is a finite

or countable collection of pairwise disjoint open intervals, and a, b E R, we define
m(U fl [a, b]) m(1 fl [a, b)).
Since my. fl [a, b)) = m(1 fl (a, b)) for all n, we have
m(U fl [a, b]) = m(U fl (a, b)).
Remark. What the above definition really defines is the measure of relatively open
subsets of [a, b) (see Definition 3.1.14). By Theorem 3.1.16, a subset G of [a, b] is
open in [a, b] if and only if there exists an open subset U of R such that G = U n [a, b].
Since the set U is not unique, we leave it as an exercise (Exercise 4) to show that if U.
V are open subsets of ltd with
u fl [a, b] = V fl [a, b],
then m(U fl [a, b]) = m(V fl [a, b]).
10.2.15 THEOREM If U is an open subset of lib and a, b e I8, then

m(U fl [a, b]) + m(U` fl [a, b]) = b - a.
Recall from Chapter 1 that If = R \ U = (x E H : x 44 U). If U is open, then U`

is closed and thus U` fl [a, b) is a compact subset of [a, b].
Proof. Suppose V D [a, b] is open. Let K = U` fl [a, b]. Then since V D K.
m(K) = m(V) - m(V \ K).
But
V \ K = V fl (U` fl (a, b])" = (V fl U) U (V fl [a, b]`) D V fl U.
Therefore, m(V fl U) <- m(V \ K). Since U fl [a, b) C U) v,
m(U fl [a, b]) + m(K) < m(U fl v) + m(V) - m(V \ K) s m(V).
Given e > 0, take V = (a - e, b + e). Then

m(Un[a,b])+m(U`n[a,b]):r= b-a+2e.
Since e > 0 is arbitrary, this proves
m(Un[a,b])+m(U`n[a,b])<--b-a.
To prove the reverse inequality,
let 1, = [a + e, b - e], where 0 < e <
2(b - a). Then
m(U n [a, b])+ m(u° n [a, b]) >- m(U n (a, b)) + m(U` n 4).
Since (a, b) is an open set containing U` n /E,
m(u` n 4) = b - a - m((a, b) \ (u° n f))

But
m((a, b) \ (U` n 4)) = m(((a, b) n U) U ((a, b) n /;))
= m(((a, b) Cl U) U (a, a - e) + (b - e, b)),

<- m(U Cl (a, b)) +
Therefore,
m(Un[a,b])+m(U`n[a,b])?b-a-2e.
Since e > 0 was arbitrary, the reverse inequality follows. Q
EXERCISES 10.2
1. If is a finite or countable collection of disjoint open intervals with U 1 C (a, b), prove that
m(4) m((a, b)).
2. *If U * io is an open subset of R, prove that m(U) > 0.

3. 'Let P denote the Cantor set in [0, 1 ]. Prove that m(P) = 0.
4. Suppose U, V are open subsets of R. a, b E R with U n [a, b] = V fl [a, b]. Prove that
m(U n [a, b]) = m(V n [a, b]).
5. If A. B are subsets of R. prove that
XAnB(x) = XA(x)XB(x),
XAUB(x) = XA(x) + XB(x) - XAnB(x),
XA'(x) = 1 - XA(x)
6. *If K, and Kz are disjoint compact subsets of R, prove that m(K, U K2) = m(K,) + m(K2).
444 Chapter io Lebesgue Measure and Integration
10.3 Inner and Outer Measure; Measurable Sets

Our goal in this section is to define a function A on a large family l1 of subsets of R.
called the measurable sets, which agrees with the function m on the open and compact
subsets of R. We begin with the definition of inner and outer measure of a set.
10.3.1 DEFINITION Let E be a subset of H. The Lebesgue outer measure of E, denoted

A*(E), is defined by
A*(E) = inf{m(U): U is open with E C U}.
The Lebesgue inner measure of E, denoted A*(E), is defined by
A*(E) = sup{m(K) : K is compact with K C E}.
10.3.2 THEOREM
(a) For any subset E of R, 0 < A*(E) t A*(E).
(b) If EI and E2 are subsets of a8 with El C E,, then
A*(EI) A*(E,) and A*(EI) < A*(E,).
Proof. (a) If K is compact and U is open with K C E C U, then

0 5 m(K) m(U).
If we fix K, then m(K) m(U) for all open sets U containing E. Taking the infimum
over all such U gives
0 tn(K) A*(E).
Taking the supremum over all compact subsets K of E proves (a). The proof of (b) is
similar and is left as an exercise (Exercise 7). 0
10.3.3 EXAMPLES
(a) If E is any countable subset of R. then
A*(E) = A*(E) = 0.
Suppose E = {xn}" I. Let E > 0 be arbitrary. For each n, let
e
In = Xn - 2n, Xn + 2n
and set U = U' I In. Then U is open with E C U. By Theorem 10.2.6,

z °°
E
m(U) m(In) F- T = 2e.
n=1 n=1
Therefore, A*(E) < 2e. Since e > 0 was arbitrary, A*(E) = 0. As a consequence, we
also have A*(E) = 0.
(b) if I is any bounded interval, then

A*(1) = A*(/) = m(1).
Suppose I = (a, b) with a, b E R. Since I itself is open,
A*(I) <_ m(1) = b - a.
On the other hand, if 0 < E < (b - a), then [a + $, b - fl ] is a compact subset of !.
and as a consequence
b -a - e=ml [ a+2,b-2Jj_A*(1).
Therefore,
b-a-ESA*(1)SA*(1)Sb-a.
Since e > 0 was arbitrary, equality holds. A similar argument proves that if 1 is any
closed and bounded interval, then
A*(I) = A*(!) = m(I).
As a consequence of Theorem 10.3.2(b), the result holds for any bounded interval 1.
(c) For any open set U,
A*(U) = A*(U) = m(U).
By definition, A*(U) = m(U). But m(U) _ where {I,} is a pairwise disjoint
collection of open intervals with U = Suppose a E R satisfies a < A*(U).
Since m(U) > a, there exist a finite number of intervals 11, ... ,1N such that
$N t m(1) > a. For each j, choose a closed and bounded interval Ji C I, such that
j tm(J,) > a. Let K = U t t J. Then K is a compact subset of U and thus
A*(U) > m(K). Finally, since the intervals {J% I are pairwise disjoint, by Exercise 6
of Section 10.2,
m(K) _ m(JJ) > a.

J=1
Therefore A*(U) > a. If A*(U) = oo, then by the above A*(U) > a for every a E R;
that is, A*(U) = oo. On the other hand, if A*(U) is finite, take a = A*(U) - E, where
e > 0 is arbitrary. But then A*(U) ? A* (U) > A*(U) - e for every e > 0. From this
it now follows that A*(U) = A*(U) = m(U).
Measurable Sets
In both of the previous examples, the inner and outer measures of the sets are equal. As
we shall see, all subsets of R built out of open sets or closed sets by countable unions,
intersections, and complementation will have this property. This includes most sets en-
countered in practice. In fact, the explicit construction of a set whose inner and outer
measures are different requires use of an axiom from set theory, the Axiom of Choice,
446 Chapter 1 0 Lebesgue Measure and Integration
which we have not discussed. The construction of such a set is outlined in the miscel-
laneous exercises.
10.3A DEFINITION
(a) A bounded subset E of ff is said to be LAbesgue measurable or measurable if
A*(E) = A*(E).
If this is the case, then the measure of E, denoted A(E), is defined as
A(E) = A*(E) = k*(E).
(b) An unbounded set E is measurable if E fl [a, b] is measurable for every
closed and bounded interval [a, b]. If this is the case, we define
A(E) = klint A(E fl [-k, kJ).
Remarks
(a) If E is unbounded and E fl l is measurable for every closed and bounded interval
1, then by Theorem 10.3.2 the sequence {A(E rl [-k, k])} t is nondecreasing, and as
a consequence A(E) = lim A(E fl [ -k, k]) exists.
(b) There is no discrepancy between the two parts of the definition. We will shortly
prove in Theorem 10.4.1 that if E is a bounded measurable set, then E fl 1 is measur-
able for every interval 1. Conversely, if E is a bounded set for which
A*(E fl [a, b]) = A*(E fl (a, b])

for all a, b E R, then by choosing a and b sufficiently large such that E C [a. b], we
have A*(E) = A*(E). The two separate definitions are required due to the existence of
unbounded nonmeasurable sets E for which
A* (E) = A*(E) = oo.
An example of such a set will be given in Exercise 5 of Section 10.4.
10.3.5 THEOREM Every set E of outer measure zero is measurable with A(E) = 0.
Proof. Suppose E C R with A*(E) = 0. Then for any closed and bounded interval 1.
A*(E fl I) A*(E) = 0.
Thus A* (En I) = A*(E fl I) = 0, and hence E fl I is measurable for every closed and
bounded interval I. Since A(E fl [ -k, k]) = 0 for every k E N, A(E) = 0. 0
As a consequence of the previous theorem and Example 10.3.3(a), every countable
set E is measurable with A(E) = 0. In particular, Q is measurable with A(Q) = 0. An-
other consequence of Theorem 10.3.5 is that every subset of a set of measure zero is
measurable.
10.3.6 THEOREM Every interval I is measurable with A(1) = m(l).

Proof. By Example 10.3.3(b), if 1 is a bounded interval, A*(l) = A*(l) = m(l). Thus
I is measurable with A(l) = m(1). On the other hand, if 1 is unbounded, then I fl [a, b]
is a bounded interval for every a, b E R, and thus measurable. In this case.
A(/) A(1 fl [-k, k]) = iim m(1 fl [-k, k]) = oo.
klym
10.3.7 THEOREM For any a, b E R and E C R,

A*(E fl [a, b]) + A* (E` fl [a, b]) = b - a.
Proof. Let U.be any open subset of R with E fl [a, b] C U. Then U` fl [a, b] is
compact with U` fl (a, b] C E` fl [a, b]. Therefore,
m(U) + A*(E` fl [a, b]) ? m(U fl [a, b]) + m(U` fl [a, b]) = b - a.
The last equality follows by Theorem 10.2.15. Taking the infimum over all open sets U
containing E` fl [a, b] gives
A*(E fl [a, b)) + A*(E` fl [a, b]) ? b - a.
To prove the reverse inequality, let K be a compact subset of E` fl [a, b]. Then K`
is open with K` fl [a, b] D E fl [a, b]. Therefore,
A*(E fl [a, b]) + m(K fl [a, b]) <_ m(Kc Cl [a, b]) + m(K fl [a, b])
= b - a.
The last equality again follows by Theorem 10.2.15 with U = K. Thus taking the
supremum over all compact subsets K of E` Cl [a, b] gives
A*(E fl [a, b]) + A*(E` fl [a, b]) s b - a,
which combined with the above, proves the result.
10.3.8 LEMMA For any subset E of R,

A*(E) = kiimn A*(E fl [-k, k]).
Proof. Since the proof is similar to that of Theorem 10.2.5 we leave it as an exercise
(Exercise 8).
10.3.9 THEOREM Suppose E,, E2 are subsets of R. Then

(a) A*(E1 U E2) + A*(E, Cl E2) <_ A*(E1) + A*(E2), and
(b) A*(E1 U E2) + A*(E, fl E2) ? A*(E1) + A*(E2).
Proof. (a) If A*(E;) = oo for some i, i = 1, 2, then inequality (a) certainly holds.
Thus suppose A*(E;) < oo for i = 1, 2. Let e > 0 be given. By the definition of outer
measure, for each i, we can choose an open set U. containing E, such that
m(U;) < A*(E;) + 2

Therefore,
e + A*(E,) + A*(E,) > m(U,) + m(U,),
=m(U,UU2)+m(U,nU2)
? A*(E, U E2) + A*(E, n E2).
The last inequality follows from the definition of outer measure. Since e > 0 was ar-
bitrary, inequality (a) follows.
(b) Let a, b E Il8 be arbitrary. By (a) applied to [a, b] n E;, we have
A*([a, b] n E,) + A*([a, b] n El)
? A*([a, b] n (E, U El)) + A*([a, b] n (E, n E;))
= A*([a, b] n (E, n E2)`) + A*([a, b] n (E, U E,)`).
But by Theorem 10.3.7, for any E C R,
A*([a, b] n E`) = (b - a) - A*(E n [a, b)).
Therefore,
A*(E, n [a, b]) + A*(E2 n [a, b])
<_ A*([a, b] n (E, n E2)) + A*([a, b] n (E, U E,)).
For each k E N. set 4 = [ -k, k]. By the above,
A*(E, n Ik) + A* (E2 n Ik) 5 A*((E, n E,) n 1k)
+A*((E, U E2) n 1k) s A*(E, n E2) + A*(E, U E2).
The result now follows by Lemma 10.3.8.
EXERCISES 10.3
1. a. If E C R. a, b E IR, prove that A*(E n (a. b)) = A*(E n [a, b]) and A*(E n (a, b)) = A*(E n [a. b]).
*b. If E C R, prove that A*(E + x) = A*(E) and A*(E + x) = A*(E) for every x E R, where
E+x=(a+ x: aEE).
2. Prove that every subset of a set of measure zero is measurable.
3. *Let E, C R with A*(E,) = 0. If E2 is a measurable subset of R, prove that E, n E. and E, U E2 are measurable.
4. Let E _ [0, 1 ] \ Q. Prove that E is measurable and A(E) = 1.
5. Let P denote the Cantor set in [0, 11. Prove that A*(P) = 0.
6. *If E C R, prove that there exists a sequence of open sets with E C U. for all n E N such that
-A*(E) = A*( n U,).
7. Prove Theorem 10.3.2(b).
& *Prove Lemma 10.3.8.
9. a. Prove that every open set U is measurable with A(U) = m(U).
b. Prove that every compact set K is measurable with A(K) = m(K).
c. Prove that every closed set is measurable.
10.4 Properties of Measurable Sets

In this section we will study some of the basic properties of measurable sets. Our first re-
sult proves that the union and intersection of two measurable sets are again measurable.
10.4.1 THEOREM If E, and E, are measurable subsets of R, then

E,flE, and E,UE,
are measurable with
A(E,) + A(E,) = A(E, U E,) + A(E, fl E,).
Proof. (a) We consider first the case where both E, and E, are bounded measurable
sets, in which case, E, fl E, and E, U E, are also bounded. Since
A(E) = A*(EE) = A*(E) i = 1, 2,
by Theorem 10.3.9,
A(E,) + A(E,) A*(E, U E,) + A*(E, fl E,)
A*(E, U E2) + A*(E, fl E,) : A(E,) + A(E2).
Therefore,
A*(E, U E,) + A*(E, fl E,) = A*(E, U E,) + A*(E, fl E,),
A*(E, U E,) - A*(E, U E2) = A*(E, fl E,) - A*(E, fl E,).
But for any bounded set E, A*(E) - A*(E) ? 0. Thus equality can hold in the above if
and only if both sides are zero; namely,
A*(E, U E,) = A*(E, U E,) and A*(E, fl E,) = A*(E, fl E,).
Therefore, E, fl E, and E, U E, are measurable, with
A(E,) + A(E,) = A(E, U E,) + A(E, fl E,).
(b) Suppose one or both of the measurable sets E, and E, are unbounded. Let
1= [a, b] with a. b E R. If both E, and E, are unbounded, then E, n/ and E, fl 1 are
measurable by definition. If one of the two sets, say E,, is bounded, then by part (a),
E, fl l is measurable. Thus in both cases, E, n l and E, fl 1 are bounded measurable
sets. But then
(E, f1 E,) fl I and (E, U E,) f11
are measurable for every closed and bounded interval I with
A(E, fl I) + A(E, fl 1) = A((E, fl E2) fl I) + A((E, U E2) fl 1). (4)
Since E, U E2 is unbounded, E, U E, is measurable by definition. Also, if E, fl E. is
unbounded, then it is measurable by definition. On the other hand, if E, fl E, is
bounded, choose I such that E, fl E2 C 1. In this case,
(E, fl E2)f11=E,flE2.
Therefore, El n E2 is measurable. Finally, to prove that

A(E1) + A(E,) = A(E1 U E2) + A(E, n E2),
take Ik = [ - k, k] in equation (4) and let k -+ oo. Q
Remark. As a consequence of the previous theorem, if E is a bounded measurable

subset of R, then £111 is measurable for every bounded interval 1. Also, if El, ... , E.
are measurable sets, then by induction
l Ek and U Ek
are measurable.
10.4.2 THEOREM A subset E of R is measurable if and only if

A*(E n [a, b]) + A*(E`' n [a, b]) < b - a
for every a, b E R.
Proof. Let E C R. Interchanging the roles of E and E` in Theorem 10.3.7 gives
A*(E n [a, bJ) + A*(E` n (a, b]) = b - a
for any a, b E R. If E is measurable, then E n [a, b] is measurable for every a, b E R.
and thus
A*(E n [a, b]) = A*(E n [a, b]).
Thus if E is measurable, A*(E n [a, b]) + A*(E` n [a, b)) = b - a.
Conversely, suppose E satisfies A*(E n [a, b]) + A*(E` n [a, b]) b - a for
every a, b E R. Since we always have
A*(E n [a, b]) + A*(E` n [a, b]) = b - a,
we obtain
A*(E n [a, b)) - A*(E n [a, b]) 5 0.
Since A*(E n [a, b]) c A*(E n [a, b]), the above can hold if and only if
A*(E Cl [a, b]) = A*(E n [a, b]).
Thus E n [a, b] is measurable for every a, b E R. Hence E is measurable. Q
10.4.3 COROLLARY A set E is measurable if and only if E` is measurable.
Proof. This is an immediate consequence of the fact that E satisfies Theorem 10.4.2
if and only if E' satisfies Theorem 10.4.2. Q
Our next goal is to show that the union and intersection of a countable collection
of measurable sets is again measurable. For the proof of this result we require the fol-
lowing theorem.
10.4A THEOREM
(a) If is a sequence of subsets of 18, then
A*(U En) s n=l

; A*(En)
(b) If {En}O,O- , is a sequence of pairwise disjoint subsets of R, then
k( U'
n=1
E- ?I
R-I
Proof. (a) If :, A*(En) = oo, then the conclusion in (a) is certainly true. Thus we
assume that Y,, A*(EE) < coo. Let e > 0 be given. For each n E N. there exists an
open set U. with En C Un such that
m(UU) < A*(EE) + 2n
Let U = U00 , Un. Then U is an open subset of R with E = U" , E. C U. Thus
A*(E) c m(U Un),
m(U,) < (A*(En) + 2n)

nil n=1
00
_ A*(E.) + e.
M=I
Since this holds for all e > 0, the result follows.

(b) Suppose the sets En, n = 1 , 2, ... , are pairwise disjoint. Since E, fl E2 = 46,
by Theorem 10.3.9(b),
A*(E,) + A*(E2) S A*(E, U E2).
By induction,
kF, A*(EE) 5 A*(U

k-I
E) <_ A"(r1 Ek t.
Since the above holds for all n E N, letting n --* cc gives the /desired result. O
10A.5 THEOREM Let {En}., be a sequence of measurable sets. Then
00 00
U En and nE
are measurable with
(a) A(U En) 5 n-1 A(E )

452 Chapter lo Lebesgue Measure and Integration
I f in addition, the s e t s ER, n = 1 , 2, . .. , are pairwise disjoint, then
(b) A(nUx En) = ,r A(En)

R=l
Proof. Let E = Un°_ I E. Without loss of generality we can assume that E (and
hence each of the sets ER) is bounded. If not, we consider
En[a,b] = RU (Enn(a,b])
for a, b E R.
Set A, = El, and for each n E N, n ? 2, set
An = ER \ (U
k=I
Ek) = (n EE)c
k-1
Since finite unions, intersections, and complements of measurable sets are again mea-
surable, An is measurable for each n E N. Furthermore, the sets An, n E N, are pair-
wise disjoint with
«;
U An = E.
Thus by Theorem 10.4.4(a) and (b),

a 00
T, A(An) g A*(E) . A*(E) < 7, A(An)
n=1 n=I
Therefore, A*(E) = A*(E), and since E is bounded, E is measurable. Furthermore, by

Theorem 10.4.4,
a
A(E) 5 7, A(E,,),
n
with equality if the sets En, n = 1, 2. .. , are pairwise disjoint.

By Corollary 10.4.3, E'0 is measurable for each n. Thus by the above, U',-, E' . is
measurable. Since
M
fEn=
n=1
U00
n-1
the intersection is also measurable. Q
10.4.6 THEOREM
(a) If {En}' 1 is a sequence of measurable sets with E, C E2 C , then
AI U En J = lim A(E0).
N=1 noc
(b) If {E.)' I is a sequence of measurable sets with E, E, D and

A(EI) < oo, then
/
ill (l En = lim A(E,).
n-+x
n I
Proof. (a) Let E = U w I En. By the previous theorem, E is measurable. If A(Ef) = 00

for some k, then A(E,,) = oo for all n z k and A(E) = oo, thus proving the result. Hence
we assume that A(E,,) < oo for all n.
Set Eo = ¢, and for n E N, let A. = E.\En Then each A. is measurable, the
collection {An} is pairwise disjoint, and
00
UAn=E.
n-1
Thus by Theorem 10.4.5(b),

N
A(E) = A(An) Nun T, A(En\ En_ I)
n=1 n-1
But E. = (En \ En _ I) U E. _ 1. Since the sets E. \ En _ 1 and En _ , are disjoint.

A(En) = A(EE\ En_1) + A(E._I).
Therefore,
N N
I A(En\En-1) _ 2 [A(En) - A(En-1)] = A(EN) - A(Eo) = A(EN),

n-1 n-1
from which the result now follows.

(b) Let E= no,*- I En. Again by the previous theorem. E is measurable. Since
M
EI = E U U (En \ En+ I),
n-I
which is a union of pairwise disjoint measurable sets, by Theorem 10.4.5(b),
A(E,) = A(E) +
n-I
N
= A(E) + lim T, [A(E.) - A(En+1)]
n=1
= A(E) + A(EI) - slim A(EN+I)
Since A(EI) < oo, A(E) = N A(EN+ 1).
The Sigma-Algebra of Measurable Sets

Let .At denote the collection of measurable subsets of R. By Theorem 10.3.6, every in-
terval I is in .AA, with A(I) = m(I). Since every open set U can be expressed as a finite or
countable union of pairwise disjoint open intervals, by Theorem 10.4.5 every open set
U is measurable with
A(U) = m(U).
Since the complement of every measurable set is measurable, At also contains all the
closed subsets of R. In particular, every compact set K is measurable with A(K) = m(K),
where m(K) is as defined in Definition 10.2.10. These by no means exhaust the mea-
surable sets. Any set obtained from a countable union or intersection of open sets or
closed sets, or of sets obtained in this manner, is again measurable.
The collection .,il. is very large. To illustrate just how large, we use the fact that the
Cantor set P has measure zero. Thus any subset of the Cantor set has outer measure
zero, and as a consequence of Theorem 10.3.5, is measurable. By Property 6 of the Can-
tor set (Section 3.3), P has the same cardinality (Definition 1.7.1) as the set of all
sequences of 0's and l's. By Miscellaneous Exercise 7 of Chapter 1, the set of all se-
quences of 0's and l's is equivalent to [0, 1 ], and this set has the same cardinality as all
of R. Thus the set of all subsets of P is equivalent (not equal) to the set of all subsets of
R. As a consequence, in the terminology of equivalence of sets, .At has the same cardi-
nality as the set of all subsets of P. However, nonmeasurable subsets of P do exist. The
construction of such a set will be outlined in the miscellaneous exercises.
We conclude this section by summarizing some of the properties of M.
10.4.7 THEOREM
(a) IfEEA,then EÈ.M..
(b) 0, H E A.
( e ) I f E E .M., n = 1 , 2, . , t h en
en
00 00
UEnEAM, and
M-1 n-1
(d) Every interval I E .At with A(I) = m(I).

(e) If E E At, then E + x E .At for all x E P with
A(E + x) = A(E).
Proof. The result (a) is Corollary 10.4.3, whereas (b) follows from Theorem 10.3.6
and Corollary 10.4.3. The statement (c) is Theorem 10.4.5, and (d) is Theorem 10.3.6.
The proof of (e) follows from Exercise lb of the previous section.
Any collection sd of subsets of a set X satisfying (a), (b), and (c) of the previous
theorem is called a sigma-algebra of sets. The o denotes that the collec-
tion s4 is closed under countable unions.
Remark An alternative approach to the theory of measure is credited to Constantin

Carathdodory. In this approach a subset E of P is said to be measurable if
A*(E fl T) + A*(E` fl T) = A*(T) (5)

for every subset T of R. Since T = (E fl T) U (E` fl T ). by Theorem 10.3.9 one

always has A*(T) s A*(E fl T) + A*(E, fl T ). Thus E satisfies equation (5) if and
only if
A*(E fl T) + A*(E, fl T) 55: A*(T).
The advantage to this approach is that it does not require the concept of inner measure.
and it includes both unbounded and bounded sets simultaneously.
If a subset E of Il8 satisfies equation (5), taking T = [a, b], a, b E R, gives
A*(E fl [a, b]) + A*(E` fl [a, b]) = A*([a, b]) = b - a.
Thus E satisfies Theorem 10.4.2 and hence is measurable. In Exercise 6. the reader will
be asked to prove that if E is measurable as defined in the text, then E satisfies equation
(5) for every subset T of Ft
D EXERCISES 10.4
1. Find a sequence of measurable sets with E, D E, D such that
OAI fl E I + lim
n-'
2. *If E is a measurable subset of I8, prove that given e > 0, there exists an open set U D E and a closed set F C E
such that A(U \ E) < e and A(E \ F) < e.
3. Let E be a bounded subset of R. Prove that E is measurable if and only if given e > 0. there exists an open set U
and a compact set K with K C E C U such that A(U \ K) < e.
4. *If E E, are measurable subsets of [0. I], and if A(E1) = 1. prove that A(E, n E,) = A(E2).
5. Suppose E, is a nonmeasurable subset of [0. I ]. Set E = E, U (l, oo). Prove that E is nonmeasurable but that
A*(E) = A*(E) = oo.
6. *Prove that a subset E of R is measurable if and only if A*(E fl T) + A*(E` fl T) = A*(T) for every subset T of R.
10.5 Measurable Functions

In our discussion of Lebesgue's approach to integration, we defined the sets
E) ={xE [a.b]:(j-1)2n -_ f(x)
where f : [a, b] -+ (0, 9) is a bounded real-valued function. As we saw in Sec-
tion 10.1, in order to define the integral off by partitioning the range, it is neces-
sary that the sets E;, j = 1, ... , n, be measurable. Thus we make the following defi-
nition.
456 Chapter 1 o Lebesgue Measure and Integration
10.5.1 DEFINITION Let f be a real-valued function defined on (a, b]. The function f is said
to be measurable if for every s E 18, the set
{xE[a,b]:f(x)>s}
is measurable. More generally, if E is a measurable subset of I8, a function f : E - R
is measurable if
{xEE:f(x)>s}
is measurable for every s e R.
Since f -'((s, oo)) = {x : f(x) > s}, f is measurable if and only if f -'((s, oo)) is a
measurable set for every s E R. We illustrate the idea of a measurable function with the
following examples.
10.5.2 EXAMPLES
(a) Let A be a measurable subset of I8 and let XA denote the characteristic function of
A. Then
R, s < 0,
{X: XA(x) > s} = A, 0:S s<1,
0, s ? 1.
Since each of the sets ¢, A, and l are measurable. XA is a measurable function on R.
(b) Let f : [0, 1) -+ I8 be defined by
f(x) = J0, x e Q n [0, ii,

x, x E [0, 1 ] \ Q.
Then
1
[0, 1], ifs < 0,
{x E [0, 1] : f(x) > s} = Q ` fl (s, l), if 0 <_ s < 1,
¢, if s?1.
Again, since each of the sets is a measurable subset of f8, f is measurable.
Properties of Measurable Functions

We now consider some properties of measurable functions. Our first result provides
several equivalent conditions for measurability.
10.5.3 THEOREM Let f be a real-valued function defined on a measurable set E. Then f is

measurable if and only if any of the following hold:
(a) {x : f (x) > s} is measurable for every s E R.
(b) {x : f(x) ? s} is measurable for every s E R.
(c) {x : f (x) < s} is measurable for every s E R.

(d) {x : f (x) < s} is measurable for every s E R.
Proof. The set of (d) is the complement of the set in (a). Thus by Corollary 10.4.3.
one is measurable if and only if the other is. Similarly for the sets of (b) and (c). Thus
it suffices to prove that (a) is equivalent to (b).
Suppose (a) holds. For each n E N, let
En = {x : f (x) > s - ,,}.
By (a), E. is measurable for all n E N. But
{x : AX) a s} = 1 ; Ea,
which is measurable by Theorem 10.4.5. Conversely, since

00
{x:f(x) > s} =R=1U{x:f(x)?s+n},
I
if (b) holds, then by Theorem 10.4.5, (a) also holds. 0
10.5A THEOREM Suppose f, g are measurable real-valued functions defined on a measur-

able set E. Then
(a) f + c and cf are measurable for every c E R,
(b) f + g is measurable,
(e) fg is measurable, and
(d) 11g is measurable provided g(x) # O for all x E E.
Proof. The proof of (a) is straightforward and is omitted.
(b) Lets E R. Then f(x) + g(x) > s if and only if f(x) > s - g(x). If x E E is
such that f(x) > s - g(x), then there exists r E 0 such that
f(x) > r > s - g(x).
Let {r,,},°,°__1 be an enumeration of Q. Then
{x : f(x) + g(x) > s} = U ({x : f(x) > ra} fl {x: ra > s - g(x)}).
Since f and g are measurable functions,

{x : f(x) > ra} and {x : ra > s - g(x)}
are measurable sets for every n E Ni. Thus their intersection and the resulting union is
also measurable. Therefore, f + g is measurable.
To prove (c) we first show that f2 is measurable. If s < 0, then
{x E E : f 2(x) > s} = E,
which is measurable. Assume s ? 0. Then
{x : f 2(x) > s} = {x : f (x) > \6} U {x : f (x) < -N/S-1-
458 Chapter 1o Lebesgue Measure and Integration
But each of these two sets are measurable. Thus their union is measurable. Since
f8 = 4[(f + 8)Z - (f - 8)2]1

the function fg is measurable.
The proof of (d) is left as an exercise (Exercise 5).
10.5.5 THEOREM Every continuous real-valued function on [a, b] is measurable.

Proof. Exercise 7.
A Property Holding Almost Everywhere

A very important concept in the study of measure theory involves the idea of a property
being true for all x except for a set of measure zero. This idea was previously encoun-
tered in the statement and proof of Lebesgue's theorem in Chapter 6; namely, a bounded
real-valued function if on [a, b] is Riemann integrable if and only if {x : f is not contin-
uous at x) has measure zero. An equivalent formulation is that f is continuous except on
a set of measure zero. In this section we will encounter several other properties that are
assumed to hold except on sets of measure zero.
10.5.6 DEFINITION A property P is said to hold almost everywhere (abbreviated a.e.) if the
set of points where P does not hold has measure zero, i.e.,
A({x : P does not hold}) = 0.
Remark The assertion that a set is of measure zero includes the assertion that it
is measurable. This, however, is not necessary. If instead we only require that
A*({x : P does not hold}) = 0, then by Theorem 10.3.5 the set {x: P does not hold) is,
in fact, measurable.
We will illustrate the concept of a property holding almost everywhere by means
of the following examples.
10.5.7 EXAMPLES
(a) Suppose f and g are real-valued functions defined on [a, b]. The functions f and g
are said to be equal almost everywhere, denoted f = g a.e., if
{x E [a, b] :f (x) g(x))
has measure zero. For example, if g(x) = I for all x E [0, 11 and
AX) = 11, x E [0, 1] \ 0,

l0, x E [0, ii fl a,
then {x E [0, 1] : f (x) # g(x)) = 0, 1] (l a, which has measure zero. Therefore, f = g a.e.
(b) In Theorem 10.5.4 we proved that if g is a real-valued measurable function on
[a, b] with g(x) # 0 for all x E [a, b], then 11g is also measurable on [a, b]. Suppose
we replace the hypothesis g(x) # 0 for all x E [a, b] with g # 0 a.e.; that is, the set
E={xE[a,b]:g(x)=0}
has measure zero. If we now define f by

g(x), x E [a, b] \ E,
f(x) 1, xEE,
then f(x) # 0 for all x E [a, b] and f(x) = g(x) except for x E E, which has measure
zero. Thus f = g a.e. on [a, b]. As a consequence of our next theorem, the function f
will also be measurable on [a, b].
(c) A real-valued function f on [a, b] is continuous almost everywhere if
{x E [a, b] : f is not continuous at x}
has measure zero. As in Example 6.1.14, consider the function f on [0,1 ] defined by
1(x) =
1 m.
if x = - in lowest terms, x # 0.
n n
As was shown in Example 4.2.2(g), the function f is continuous at every irrational num-
ber in [0, 1], and discontinuous at every rational number in [0, 1 ]. Therefore,
A({x E [0, 1 ] : f is not continuous at x}) = Ap fl [0, 11) = 0.
Thus f is continuous a.e. on [0, 1].
(d) Let f and f, n = 1, 2, ... , be real-valued functions defined on [a, b]. The se-
quence {f.) is said to converge almost everywhere to f, denoted f --+f a.e., if
{x E [a, b] : {f. (x)} does not converge to f (x)}
has measure zero. To illustrate this, consider the sequence (f,,} defined in Example
8.1.2(c) as follows: Let {xk} be an enumeration of 0 n [0, 1 ]. For each n E N, define
f, on [0, 11 by
AX) __ 0, if x = xk, 1 <_ k:5 n,

1, otherwise.
Then
0. x is rational,
hm f(x)
1, if x is irrational.
Thus {x E [0, 1 ] : { f (x)} does not converge to I) = an to, 1 ], which has measure
zero. Hence f -a 1 a.e. on [0, 1).
One of the key results needed in the sequel is the following.
10.5.8 THEOREM Suppose f and g are real-valued functions defined on a measurable set A.
!f f is measurable and g = f a.e., then g is measurable on A.
Pr oof. Let E = {x E A:g(x) # f(x)}. Then A(E) = 0. Let B = A\ E. Since E is
measurable, so is B. Also on B, g(x) = f(x). Fix s E R. Then
{xEA:g(x)>s}={xEB:g(x)>s}UIx EE:g(x)>s}
= {x E B:f(x) > s} U {x E E: g(x) > s}.
Since E, = {x E E: g(x) > s} is a subset of E and A (E) = 0, the set E, is measurable.

Also, since f is measurable, {x E B : f(x) > s} is measurable. Therefore, {x: g(x) > s},
and thus g, is measurable.
10.5.9 THEOREM Let f f.}:: , be a sequence of real-valued measurable functions defined on

a measurable set A such that (f (x)}' , is bounded for every x E A. Let
rp (x) = sup{ f" (x) : n E N) and Vi(x) = inf { fn (x) : n E NJ.
Then 9 and ifr are measurable on A.
Proof. The result follows by Theorem 10.4.5, and the fact that for every s E R.
a
{x:(p(x) > s) = U00 {x: f"(x) > s} and (x: 4,(x) < s} = n-i
U {x: f"(x) < s}.
n-i
10.5.10 COROLLARY Let be a sequence of real-valued measurable functions defined

on a measurable set A, and let f be a real-valued function on A. /f f" -+f a.e. on A, then
f is measurable on A.
Proof. Let E = {x: { f"(x)} does not converge tof(x)}. By hypothesis A(E) = 0. Set
{f"ax),
xEA\E,
gn(x) =
x E E.
Then g" = fn a.e., and thus is measurable. Also, lim gn(x) = g(x) exists for all x E A.
But
g(x) = lim g" (x) = lim gn (x) = inf sup{ fk(x) : k at n}.
n-oo n-"o n
By the previous theorem, each of the functions
F.(x) = sup{fk(x):k at n} and g(x) = inf{F,(x):n E N}
are measurable on A. Finally, since f = g a.e., f itself is measurable.
Suppose { fn} is a sequence of measurable functions on [a, b] such that fn --).f a.e.
Then by definition there exists a subset E of [a, b] such that A([a, b]\E) = 0, and
slim fn (x) = f (x) for all x E E. Exercise 15 provides a significant strengthening of
this result. There you will be asked to prove that given e > 0, there exists a measurable
set E C [a, b], such that A([a, b] \ E) < e, and {fn} converges uniformly to f on E. This
result is known as Egorov's theorem.
EXERCISES 10.5
1. *Let f be defined on [0, 1) by
0, x=0,
f(x)= x,
:.<'Prove
directly that f is measurable.

2. Let f be a real-valued function on a measurable set A with finite range, i.e., Range f = {ai, ... , aq}. a1 E R.
Prove that f is measurable if and only if f -I (a;) is measurable for all j = 1. ... , n.
3. Let A be a measurable subset of R, and let f : A - R be measurable. Prove that for each n E N, the function j
defined by
f (x) = ff(x), if If(x)I s n.

l n, if )f(x)I > n,
is measurable on A.
4. If f is measurable on [a. b], prove that f + c and cf are measurable on [a, b] for every c E R.
5, *If g is measurable on [a, b] with g(x) * 0 for all x E [a, b], prove that 1/g is measurable on [a, b].
6. Let f be a real-valued function on [a. b]. Prove that f is measurable if and only if f `(U) is a measurable subset
of [a, b] for every open subset U of R.
7. Prove that every continuous real-valued function f on [a, b] is measurable.
8, Let A be a measurable subset of R. If g :A -+ R is measurable and f : R -* R is continuous, prove that f o g is
measurable.
9. If f is monotone on [a, b] and g: R - [a, b] is measurable, prove that f e g is measurable.
10. Let E be a measurable subset of R, and let f be a measurable function on E. Define the functions f ' and f on E
as follows: f'(x) = max{f(x), 01, f -(x) = max{-f(x), 01.
*a. Prove directly that f' and f- are nonnegative measurable functions on E with f(x) = f' (x) - f- (x).
b. Prove that If(x)I = f'(x) + f-(x) and that If I is measurable.
*c. If f is a real-valued function on [a, b] such that I f I is measurable on (a, b), is f measurable on [a. b]?
d. If f(x) = + cos x, x E (0, 21r), find f' (x) and f-(x).
11. Let A be a measurable subset of R and let {f.) be a sequence of measurable functions on A. Let
E = {x E A : { f (x)}, .. i converges}. Prove that E is measurable.
12. Let f be a bounded measurable function on [0, 1] and let (f.1 be defined on [0. I ] by f .(x) = x"f(x). Prove that
each j is measurable and that (f.} converges to 0 almost everywhere.
13. Let {xk} be an enumeration of the rational numbers in [0, I J. For each n E N let
f(x)= I1, if x=xk,l5k<-n,

0, otherwise.
Show that f is measurable for each n E N. and that (f.1 converges to 0 almost everywhere on [0, I ].
14. *If f is differentiable on [a, b], prove that f' is measurable on [a, b].
15. Egorov's Theorem: Let {f} be a sequence of measurable functions on [a. b] such that f. --+f a.e. on [a. b].
Given e > 0. prove that there exists a measurable subset E of [a, h] such that A([a. b] \E) < e and {f} con-
verges uniformly to f on E.
16. Construct a sequence (J,} of measurable functions on [0. 1 ] such that { f (x)} converges for each x E [0, 1 j but
that (f.) does not converge uniformly on any measurable set E C [0. I ] with A([0, I ] \E) = 0.
10.61 The Lebesgue Integral of a Bounded Function

There are many different approaches to the development of the Lebesgue integral. One
is the method outlined in the first section of this chapter that is further pursued in Ex-
ercise 1. The drawback to this approach is that it a priori assumes that f is a measurable
function. The approach that we will follow is patterned on the Darboux approach to the
Riemann integral.
10.6.1 DEFINITION Let E be a measurable subset of R. A measurable partition of E is a

finite collection P = {E1, . .. , of pairwise disjoint measurable subsets of E such that
IDEA.=E.
k=1
Suppose 91 = {x,,, x1, ... , xn} is a point partition of [a, b] as considered in Chap-
ter 6. Set E1 = [x,,, x,], and f o r k = 2, .. . , n, set Ek = (xk_ 1, xk]. Then the collec-
tion 91 = {E,, . . . , En} is a measurable partition of [a, b]. A measurable partition of
[a, b], however, need not consist of intervals. For example, if E, = [a, b] n0 and
E2 = [a, b] \E1, then {E1, E,} is a measurable partition of [a, b].
As for the Riemann integral, if f is a bounded real-valued function on [a, b], and
01 = {E1, . .. , En} is a measurable partition of [a, b], we define the lower and upper
Lebesgue sums off with respect to 91, denoted L(9), f) and ql.L(91, f) respectively, by
n n
YLO' f) = 7, mkA(Ek) and 7, Mk A(EI),

k-1 k=I
where
mk = inf{ f (x) : x E Ek} and Mk = sup { f (x) : x E Ek}.
Clearly, 510, f) : OUL(O, f) for every measurable partition P of [a, b]. As for the
Riemann integral, we could now define the upper and lower Lebesgue integrals off, and
then define a function to be Lebesgue integrable if and only if these two quantities are
equal. The following theorem shows, however, that this is unnecessary.
10.6.2 THEOREM Let f be a bounded real-valued function on [a, b]. Then

sUUp Y.(2, f) = inf °11L(9., f ),
where the inftmum and supremum are taken over all measurable partitions 0 and of
[a, b], if and only if f is measurable on [a, b].
Remark. Although the previous theorem is stated for a closed interval [a, b], the re-
sult is also true for f defined on any measurable subset A of l with A (A) < cc.
As a consequence of the previous theorem, which we will shortly prove, we make
the following definition.
10.6.3 DEFINITION If f is a bounded real-valued measurable function on [a, b], the Le-
besgue Integral off over [a, b], denoted f(a bj f dA (or f. f dA), is defined by
t b)
f dA = sup `.EL(9, f),
where the supreemum is taken over all measurable partitions of [a, b]. If A is a measur-
able subset of [a, b], the Lebesgue integral off over A, denoted fA f dA, is defined by
f dA = fXA dA,
A J (a.bj
Remarks
(a) In defining fA f dA, it was implicitly assumed that f is defined on all of [a, b]. If f
is only defined on the measurable set A, A C [a, b], then fXA can still be defined on all
of [a, b] in the obvious manner; namely,
{f(x)'
(fxA)(x) _ x A.
Alternatively, if f is a bounded measurable function defined on a measurable subset A

of [a, b), we could define
jfdA = sup ZL (a, f ),

A
where the supremum is taken over all measurable partitions a of A. However, if

. _ {E,, ... , E} is any measurable partition of A, then
9' = {E,,... ,E,,, [a, b] 1A}
is a measurable partition of [a, b] with fL(Lf) = 2L(9,fXA)
Conversely, if 9 = {El, . . . , Ea} is any measurable partition of [a, b], then
9. = {E, fl A, ... , E. ('I A} is a measurable partition of A for which 21, (Q,f) _
2L(91,fXA). Thus the two definitions for fA f dA give the same value.
(b) To distinguish between the Lebesgue and Riemann integral of a bounded real-
valued function f on (a, b), the Riemann integral off, if it exists, will be denoted by
dx.
f bf(x)
a
464 Chapter to Lebesgue Measure and Integration
If in the Lebesgue integral we wish to emphasize the variable x, we will write

f.''d f(x) dA(x) to denote the Lebesgue integral off. The two different notations should
cause no confusion. In fact, in Corollary 10.6.8 we will prove that every Riemann inte-
grable function on [a, b] is also Lebesgue integrable, and the two integrals are equal.
Prior to proving Theorem 10.6.2 we first need the analogue of Theorem 6.1.4.
10.6.4 DEFINITION Let E be a measurable set and let l? be a measurable partition of E. A

measurable partition . of E is a refinement of 9 if every set in a is a subset of some
set in 91.
A useful fact about refinements is the following: If 9 = {E1,..., and 9 _
{A,, ... , Am} are measurable partitions of E, then the collection
{E ; fl
is a measurable partition of E that is a refinement of both 31 and 9.
10.6.5 LEMMA If 91,.q are measurable partitions of [a, b] such that 2 is a refinement of
91, then
-YL(a,f) c qLL(a.f)
As a consequence,
scup `.TL(91,f)
Proof. The proof of the lemma is almost verbatim the proof of Lemma 6.1.3 and The-
orem 6.1.4, and thus is left to the exercises (Exercise 2).
10.6.6 EXAMPLE In this example, we calculate the Lebesgue integral of what is commonly
called a simple function on [a, b]. A simple function on [a, b] is a measurable real-
valued function on [a, b] that assumes only a finite number of values.
Suppose s is a simple function on [a, b] with Range s = {a,, .. . , where
a, * of whenever i 0 j. For each i = I, . . . , n, set
E; = {x E [a, b] : s(x) = a;} = s-'({ail).
Since s is measurable, each E; is a measurable set, and
n++
S(x) _ !r a;XE,(x) (6)
Furthermore, since ai * a i if i * j, the sets E i = 1, ... , n, are pairwise disjoint

with U'. I E; = [a, b]. Equation (6) is called the canonical representation of s. If all
the sets E; are intervals, then s is a step function on [a, b].
We will now show that every simple function s is Lebesgue integrable on [a, b] and
compute the Lebesgue integral of s.
10.6.7 LEMMA If s is a simple function on [a, b] with

n
s = 7,, arXE,
where {E;}"= I are pairwise disjoint measurable subsets of [a, b] with U"= I E, = [a, b],
then s is Lebesgue integrable on [a. b] with
s dA aA(E,).
J(u, hl j=1
Proof. To show that s is Lebesgue integrable on [a, b], we will prove that
sup s) = inf 61L.L(9., s),
a .4
where the supremum and infmum are taken over all measurable partitions 9. of [a, b].
Since 9' = {E1..... ER} is a measurable partition of [a, b] and s(x) = a; for all
x E E.
-WL(9', s) = eILL(91, s) _ aiA(E;).
But then
n n
I ai A(Ej) !-__ sup $L(9., s) < inf altl(9 s) :5:- 1 ai A(Ei).
-1 a a ;-1
Therefore, s is Lebesgue integrable on [a, b] with f[, bis dA = I°=1 a;A(E;). 0
Remark Suppose f is a bounded real-valued measurable function on [a, b]. If
91 = {E ,, ... , En} is a measurable partition of [a, b], set
(x) = E mkXE,(x) and Ij!(x) = I MkXE,(x), (7)

k=1 k=1
where Mk = inf {f(x) : x E Ek} and Mk = sup{ f (x) : x E Ek}. Then tD and alt are simple
functions on [a, b] with O(x) f(x) s *(x) for all x E [a, b]. Furthermore, by Lemma
10.6.7
apdA = mkA(Ea) = 20',.f), and

J[a. bJ k-1
J[a. b] OdA = Y,
k-1
MkA(EE) = °LL(9,.f)-
Thus if f is a bounded real-valued measurable function on [a. b],
fdA = sup ,p dA,

a, b; f u. hJ
where the supremum is taken over all simple functions ap on [a, b] satisfying V(x) f(x)
for all x E [a, b].
Proof of Theorem 10.6.2. Suppose f is a measurable function on [a. b) with

m f (x) < M for all x E [a, b]. Let (3 = M - m, and for n E N, partition (m, M)
into n subintervals of l e n g t h 13/n. F o r each k = 1, ... , n, set
Ek= {xE [a,b]:m+(k-1 n -5f(X)

Then 91. = {E1, ... , E,,} is a measurable partition of [a, b). Also, if mk = inf{f(x) :
x E Ek} and Mk = sup{ f (x) : x E Ek}, then
mk?m+(k-1)Pn and Mksm+kR.

n
Therefore,
0 s innf °ltL(°.P, f) - suup Et(L,f) s °ut(9n,f) - XL(On,f)
L (Mk - mk)A(Ek)
k-1
n \\
[(m A(Ek)
= - J A(Ek) = (b - a).
n
Since n E N is arbitrary, by letting n -+ oo we obtain

suup L('Q, f) = inf VL(9,f) (8)
Conversely, suppose equation (8) holds. By taking a common refinement if neces-

sary, for each n E N, there exists a measurable partition !Pn of [a. b] such that
°1fL(9n,f) < innfG(LL(°,f) + , and
-TL(O'n,f) 5P
Since equality holds in equation (8),
'(LL(9'n,f) - 20n,f) < n

For the partition p,,, let cpn and On be simple functions on [a, b] as defined by equation
(7), satisfying (p,,(x) s f(x) s 11,,(x) for all x 6 [a, b], and
(p,, dA = 21fl'n, f ), 'Yn dA = °ltt{OImf)

J(a. b] 1(a. b)
Define (p and 11 on [a, b] by v(x) = sup,, (p,,(x) and 11(x) = infra By Theorem
10.5.9, the functions (p and 11 are measurable functions on [a, b], with
(p(x) s AX) < (G(x)
for all x E [a, b].
To complete the proof we will show that V = 41 a.e. on [a, b]. Then as a conse-
quence of Theorem 10.5.8, the function f will be measurable on [a, b]. Let
E _ {x E [a, b] : (p(x) < i/i(x)},
and for each k E Nl, let
Ek = Ix E [a, b] : V(x) < +l'(x) - k

Then E = Uk , Ek. If x E Ek, then v,,(x) < for all n E lJ For n, k E N. let
k
An. k = 1 x : 4p,(x) < 1P .(X) - k
If x E A , k, then 4#.(x) - `1. Consider the simple function

S.(x) = (4r,(x) Vn(X))XA..,(x).
Suppose the measurable partition 91 is given by {B,, .... BN}. Then

N
s (x) _ 71 (M, - mi)Xe nA,..,(x),

i-I
where Mi and mi denote the supremum and infimum off, respectively, over Bi. The collec-
tion
9 = {Bi flA,,,k}" 1 U {[a, b] \A,,,k}

is a measurable partition of [a, b]. If m* = E Bi fl A,, k}, then mj > 1/k
for all j = 1, ... , N. Also, since s (x) = 0 for all x E [a, b) \ A., k,
A(A. k)
L(9-, S.) _ m j A(B, n A,,, k) > 1 A(B f l A,,. k) =
i=1 k i., k
On the other hand,

N
(Mi - mi)A(Bi n A,,. k)

i=I
N+
t G(M1 - mi)A(Bi) _ ,(91.,f) <
XT, n
1=I
Combining the above two inequalities gives A(A,,. k) < k/n for all k, n E N. Since
Ek C A,,, k for all n; for each k,
A(Ek) < n
for all n E N. Therefore, A(Ek) = 0. Finally, since

m
A(E) <_ 1 A(Ek),
k-1
we have A(E) = 0. Thus V = 4 a.e. on [a. b), which proves the result.
Comparison with the Riemann Integral

The definition of the Lebesgue integral is very similar to that of the Riemann integral,
except that in the Lebesgue theory we use measurable partitions rather than point par-
titions. If 91 = {xo, xt, . . . , is a partition of [a, b], then
9'* _ {[x0, x1 ]} U {(xk_ 1, Xk]}R= 2
is a measurable partition of [a, b]. Furthermore, if f is a bounded real-valued function
on [a, b), then
2(9.f) < 2L(?*,f) and qt(QP.f) > qtL(?*.f)
Therefore, the lower Riemann integral off satisfies
h
i f = sup{J'(P, f) : 91 is a partition of (a, b]}

s sup{.TL(a, f) : I), is a measurable partition of [a, b]}.
Similarly, for the upper Riemann integral off we have
J f ?.inf {°ltt(a, f) : 1 is a measurable partition of [a, b)}.
If f is Riemann integrable on [a, b], then the upper and lower Riemann integrals off are
equal, and thus
J46f (x) dx 1:-: sup `.eL(a, f) ml 04L(2, f) s la bf (x) dx,
where the supremum and infimum are taken over all measurable partitions a of [a, b].
As a consequence of Theorem 10.6.2, this proves the following result.
10.6.8 COROLLARY 1f f is Riemann integrable on [a, b], then f is Lebesgue integrable on

[a, b], and
h
(
fdA =
+
f Ax) dx.
u
The converse, however, is false! This is illustrated by the following example.
10.6.9 EXAMPLE Let E = [0, I ] \ 0, and set

when x is irrational,
f(x) = XE(X) = {0, when x is rational.
By Example 6.1.6(a) the function f is not Riemann integrable. On the other hand, since
f is a simple function, f is Lebesgue integrable, and by Lemma 10.6.7,
f dA = A(E) = 1.
Jlo. l i
Properties of the Lebesgue Integral for Bounded Functions

The following theorem summarizes some basic properties of the Lebesgue integral for
bounded functions.
10.6.10 THEOREM Suppose f, g are bounded real-valued measurable functions on [a. b). Then
(a) for all a,,8 E R,
(af+ 6g)dA=a f fdA+0 f gdA.

a b] i b] ' a b]
(b) If A,. A2 are disjoint measurable subsets of [a, b], then
fdA= JfdA+1 fdA.

JA,UA2 A, A,
(c) If f a g a.e. on [a, b], then fdA ? J gdA.

I(a. b] [a, b]
(d) Iff=ga.e.on[a,b],then J fdA= gdA.

(a. b] f(a, b]
(e) fdA < If 1A.

Jla, b] J(a, b]
Proof. Since the proof of (a) is similar to the proof of the corresponding result for the
Riemann integral we leave it as an exercise (Exercise 4). For the proof of (b), by defi-
nition
f fdA = f f%A,UA, A.
A,UA2 ia.b]
Since A, fl A2 = fXA,UA, = fXA, + fXA2, and the result now follows by (a).
(c) Consider the function h(x) = f(x) - g(x). By hypothesis h ? 0 a.e. on [a, b i.
Let
E, = {x:h(x) >_ 0} and E2 = [a, b]\ E,.
Consider the measurable partition 9' = {Et, E2} of [a, b]. Then
h A ? XL(91, f) = m,A(E,) + m2A(E,).

a. b]
Since h(x) z 0 for all x E E,, m, = inf{h(x):x E E,} > 0. On the other hand, since
h ? 0 a.e., A(E2) = 0. Therefore, fa h dA ? 0. The result now follows by (a).
The result (d) is an immediate consequence of (c), and (e) is left for the exercises.
The measurability of f I follows from Exercise 8 or 10 of the previous section. Q
Bounded Convergence Theorem

One of the main advantages of the Lebesgue theory of integration involves the inter-
change of limits. If if.} is a sequence of Riemann integrable functions on [a, b] such
that fn(x) converges to a function f(x) for all x E [a, b], then there is no guarantee that
f is Riemann integrable on [a, b]. An example of such a sequence was given in Exam-
ple 8.1.2(c). For the Lebesgue integral, however, we have the following very useful
result.
10.6.11 THEOREM (Bounded Convergence Theorem) Suppose is a sequence of real-

valued measurable functions on (a. b) for which there exists a positive constant M such
that I fa(x)I M for all n E N. and all x E [a, b]. If
liim fn(x) = f (x) a.e. on [a, b],
then f is Lebesgue integrable on [a, b] and
1 f dA = lim fn dA.
n-4ooJ
[a. bj la. bj
Remark. Although we state and prove the bounded convergence theorem for a closed
and bounded interval [a, b], the conclusion is still valid if the sequence {fn) is defined
on a bounded measurable set A. The necessary modifications to the proof are left to the
exercises.
Proof. Since fn --> f a.e., f is measurable by Corollary 10.5.10, and thus Lebesgue
integrable. Let
E = {x e (a, b): fn(x) does not converge to f(x)).
Define the functions g and g,,, n E N, on [a, b] as follows:
x E (a, b) \ E, g(x) _ f(x), x E [a, b] \ E.
gn(x) = {f(x), and
0, xEE, 0, xEE.
Since A(E) = 0, gn = fn a.e. and g = f a.e. Therefore,
r6 b b b
"dA = fndA and Ja gdA= I fdA.
Ja 1a
Furthermore, gn(x) -+g(x) for all x E [a, b]. Let e > 0 be given. Form E N. set
E. _ {x E [a, b]:Ig(x) - gn(x)I < e forall n >_ m}.
Then E, C EZ C with U° , E. = [a, b]. Therefore,

a
fly Em = 45.
Here Em = [a, b] \ Em. Thus by Theorem 10.4.6, lim A(E,) = 0. Choose m E N such
that A(Em) < e. Then I g(x) - e for all7n*w m and all x E Em. Therefore,
f6fdA- jbfdAl = I bgdA

IQ - f bgndAl -'5 1 Ig gn)dA
fa. bJ
= jt - g d + t1& - gdA
< e A(E,) + 2M A(Ec) < [b - a + 2M].
Since e > 0 was arbitrary, we have lire fla b]
f dA = fla b, f dA. Q
Combining the bounded convergence theorem with Corollary 10.6.8, we obtain the
bounded convergence theorem for Riemann integrable functions previously stated in
Chapter 8. The theorem does require the additional hypothesis that the limit function f
is Riemann integrable.
THEOREM 8.4.3 Let f and f,,, n E N, be Riemann integrable functions on [a, b]

with lim f(x) all x E [a, b]. Suppose there exists a positive constant M
such that [ M for all x E [a, b] and all n E N. Then
fl-
lim
I
a a
f (x) dx.
10.6.12 EXAMPLES
(a) In the first example, we show that the conclusion of the bounded convergence the-
orem is false if the sequence f f,,) is not bounded; that is, there does not exist a finite
constant M such that I f (x)l <_ M for all n E N, and all x E [a, b]. For each n E N,
define f on [0, 11 by
f n, 0<x51n
(x) =
Mx) Sl
0, otherwise.
Then { is a sequence of measurable functions on [0, 1] that is not bounded, but

that satisfies
1im f (x) = f (x) = 0 for all x E [0, 1 ].
However, f, f. dA = nA((0, ]) = 1. Thus
lim f dA = 1 vk 0= fdA.
n-fM J [0.11 110,1:
(b) As our second example, we consider the sequence { of Example 9.2.3. For each
n E N, write n = 2k +j. where k = 0, 1, 2, .. . , and 0:!:-: _ j < 2k. Define f on [0, 1 ] by
j j+1
2k
`- x `- 2k
.f "(x)
0, otherwise.
The first few of these are as follows: f, = %,o_ :, f,- _ X;0.1/: .f3 - X.1/2. l: f4
X[o,1/41 .. For each n E N, f. E Jt[0, 1 ] with
1 1
dx = 2k
f
J0
Thus lim fo f (x) dx = 0. On the other hand, if x E [0, 1], then the sequence { f (x)}
contains an infinite number of 0's and 1's, and thus does not converge.
EXERCISES 10.6
1. *Let f be a bounded real-valued measurable function on [a, b) with m < f(x) < M for all x E [a, b]. Set
S=M-m.FornENandj=1....,n,letE1={xE[a,b]:m+(j-1)9,<_f(x)<m+jj}.The
Lebesgue sums for f are defined by mJ , (m + (j -1))A(E,). Prove that
lint fdA.
2. Prove Lemma 10.6.5.

3. Let f be a bounded measurable function on ;a, b]. If A is a measurable subset of (a, b) with k(A) = 0. prove that

b. Prove Theorem 10.6.10(e).
S. *Let f be a nonnegative bounded measurable function on [a, b]. If E, F are measurable subsets of (a, b) with
E C F. prove that fE f dA s f,,f dA.
6. Let f be a bounded measurable function on [a, b]. For each c > 0, prove that
A({x E [a, b]: I f(x)I > c}) s 1 1 fl dA.

Lbl
7. *Let f be a nonnegative bounded measurable function on [a, b] satisfying fta bl fdA = 0. Use the previous exer-
cise to prove that f = 0 a.e. on [a, b].
8. Fundamental Theorem of Calculus for the Lebesgue Integral: if f is differentiable on ia. bj and f' is bounded
on [a, h], then f' is Lebesgue integrable, and f,,.,, f' dA = f(b) - f (a).
9. Let.f be a bounded measurable function on [0. 11. Show that x"f(x) is measurable on [0. 11 for n = I. 2.... .
and find lint
f be a bounded measurable function on [a, b] such that f [ < I a.e. on [a, b]. Prove that lim f f " dA = 0.
11. Let be a sequence of nonnegative measurable functions on [a. h' satisfying /;,(.r) - M for all x E :a. b' and
n E N. If f f,) converges to f a.e. on [a. b], prove that
lint f
f e f- A = f fe-f dA.
" b:
12. *If f is a bounded real-valued measurable function on [a. bl, prove that there exists a sequence {s"} of simple
functions on [a, b] such that lint s"(x) = f(r) uniformly on [a, b].
13. Modify the proof of the bounded convergence theorem when the interval [a. b] is replaced by a bounded measur-
able set A.
14. Use Egorov's theorem (Exercise 15. Section 10.51 to provide an alternative proof of the bounded convergence the-
orem.
15. *If p is a simple function on (a. h] and e > 0 is given, prove that there is a step function b on :a. h_ such that
c(x) = h(x) except on a set of measure less than e.
16. Let S"(x) = A0 + Yi-i (Ax cos kx + BA sin kx). If , S"(x) I s M for all .r E [ - ar. a] and n E N. and
f(x) = lint S"(x) exists a.e. on [-a, a], prove that f is measurable and that the A4 and B5 are the Fourier coeffi-
cients off.
10.71 The General Lebesgue Integral

in this section we extend the definition of the Lebesgue integral to include both the
cases where the function f is unbounded, and where the domain of integration is un-
bounded. We will then prove the well-known results of Fatou and Lebesgue on the
interchange of limits and integration. We first consider the extension of the Lebesgue
integral to nonnegative measurable functions.
The Lebesgue Integral of a Nonnegative Measurable Function

Suppose first that A is a bounded measurable subset of R. and that f is a nonnegative
measurable function defined on A. For each n E N, consider the function f" defined on
A by
jf(t:), iff(x) n ,
f(x) n} =
I. n. if f(x) > n.
Then (fn) is a sequence of nonnegative bounded measurable functions defined on A.

with l fn(x) = f (x) for all x E A. Furthermore, if m > n, then
MX) 5 fm(x) 5 f(x)
for all x E A, and thus the sequence
JAfn
dAK
is monotone increasing, and converges either to a real number or diverges to oo. This
leads us to the following definition.
10.7.1 DEFINITION
(a) Let f be a nonnegative measurable function defined on a bounded measurable
subset A of R. The Lebesgue integral off over A, denoted fA f dt, is defused by
A A A
(b) If A is an unbounded measurable subset of R and f is a nonnegative

measurable function on A, the Lebesgue Integral off over A, denoted fA f dA,
is defined by
Jfr,J
A
-+oc
Afl[-n, n]
fdA.
In part (b) of the definition, the sequence
f dAl }nEN
fArj -n, n] J
is also monotone increasing, and thus converges either to a nonnegative real number or
diverges to oo. In the definition we do not exclude the possible value of oo for the inte-
gral off. If the integral is finite, however, we make the following definition.
10.7.2 DEFINITION A nonnegative measurable function f defined on a measurable subset

A of R is said to be (Lebesgue) Integrable on A if fA f dA < oo.
Remark If A is either a finite or infinite interval with endpoints a, b E R U (-00, oo),

then the integral of a nonnegative measurable function f on A is also denoted by f; f A.
10.7.3 EXAMPLES
(a) For our first example, we consider the function f(x) = I// defined on (0,1).
Then for each n E IN,
n, 0<x< 1
n-
fn(x) = min{f(x), n} =
1
Therefore,
f fndA= 1
0 0
ndx+
J
1 dx= l
n n
1
n
As a consequence,
I f dA = lim 1 f dA = liml 2 - 1) = 2.
The answer in this example corresponds to the improper Riemann integral of the func
tion f. This will always be the case for nonnegative functions for which the improper
Riemann integral exists (Exercise 18).
(b) Let g(x) = 1/x, 0 < x K 1. For the function g,
n, 0<x< 1,n
min{g(x), n} = 1
x'
-<x<
n
1
1.
Therefore,
1 I/n t 1
J min{g, n} dA = I n dx + J - dx = 1 + Inn.
0 0 1/fix
Thus
Jgdlt=lim(1
+Op
+ In n) = oo.
Since the Lebesgue integral of g is infinite, g is not integrable on (0, 11.

(c) As our final example, consider f (x) = 1.2 defined on A = (1, oc). In this example,
for n='2,Afl[-n,n] =(1,n),and
jfdA = jx2dx = I - 1.n

Therefore,
I` fdA=lim1-n=1.
I
Thus f is integrable on (1, oo).

The following theorem summarizes some of the basic properties of the Lebesgue
integral of nonnegative measurable functions. Integrability of the functions f and g is
not required.
10.7.4 THEOREM Let f, g be nonnegative measurable functions defined on a measurable

set A. Then
(a) j(f+g)dA= jfdA+ jgdA and

tcfdcJA1A
for all c > 0.
(b) If A,, A2 are disjoint measurable rsubsets of A, then
JA, UA.
fdA =
1A,
fdl + j fdA.
A,
(c) If f s g a.e. on A, then jfdA I g dA, with equality if f = g a.e. on A.

A A
Proof. We will indicate the method of proof by proving part of (a). The remaining
proofs are left to the exercises (Exercise 2). Suppose first that the set A is bounded. Let
h = f + g. Since
min{ f (x) + g(x), n} 5 min{ f (x), n) + min{g(x), n} <_ min{ f (z) + g(x), 2n},
we have h < f + g <_ h2R for all n E N. As a consequence,
!g,, dA _ Ih2.dA.
fAh- dAs
A A A
Suppose f, g are integrable on A. Then
mHf,, dA + fgndA=
A
) 1m
A
jf dA + Aliimm f g dA = JfdA+ JIdA.
A A A A
Therefore, since
f
lim h dA = lim J h2i dA = j(f + g) dA,
n-.Wf
A A A
the result follows. If one or both of the sequences { fA fn dA}, { f, gn dA} diverges to x,
then so does their sum. In this case, we obtain fA(f + g)dA = oo. If A is unbounded.
then by the above for each n E N,
j (f + g)dA = J
An;-n.._
fdA +
JAnl-n.r,
gdA,
Ant-n.n)
and the result again follows by letting n -+ co. Q

As a consequence of the previous theorem, if f and g are nonnegative integrable
functions on the measurable set A, then so is f + g and cf for every c > 0. Further-
more, if f <_ g a.e. and g is integrable, then so is f.
Fatou's Lemma
Our first major convergence theorem for integrals of nonnegative measurable functions
is the following result of Fatou.
10.7.5 THEOREM (Fatou's Lemma) If {fn} is a sequence of nonnegative measurable

functions on a measurable set A, and slim fn (x) = f(x) a.e. on A, then
1 f dA <_ lim JfndA.
Proof. Suppose first that the set A is bounded. For each k E N, let
hn(x) = min{ fn(x), k} and h(x) = min{ f (x), k}.
Then for each k E N, the sequence {hn} converges a.e. to It on A. Since I hn(x) 1 s k for
all x E A, by the bounded convergence theorem,
J min{ f, k) dA = lim f min{ fn, k} dA s lim Ji

A A °'A
Since the above holds for each k e N,
f f dA = lim J min{ f, k} dA s lim J f, dA.

A
k-ooo A noc A
If A is unbounded, then by the above for each k e N.
f dA s lim fn dA c lim fn dA.

fAn(-k. kJ W__-00 f W__00 f
A
Letting k -+oo will give the desired result. Q

10.7.6 EXAMPLE In this example we show that equality need not hold in Fatou's lemma.
Consider the sequence ( fn} on [0, 1 ] of Example 10.6.12(a). For each n E N, fn(x) = n
if 0 < x s 1, and fn(x) = 0 elsewhere. This sequence satisfies
0= f (lim
nix fn) dA < I =n-.oo
lim J f dA.
f 0, I j 0, l1
Remark. Fatou's lemma is often used to prove that the limit function f of a conver-
gent sequence of nonnegative Lebesgue integrable functions is Lebesgue integrable.
For if lim fA fn dA < oo and if fn -+f a.e. on A with fn ? 0 a.e. for all n, then by Fa-
tou's lemma, fA f dA < oo. Thus f is integrable on A.
The General Lebesgue Integral

We now turn our attention to the case where f is an arbitrary real-valued measurable
function defined on a measurable subset A of R. As in Exercise 10 of Section 10.5, we
define the functions f' and f " on A as follows:
f'(x) = max{f(x),0},f-(x) = max{ f(x),0}.
If f (x) > 0, then f +(x) = f (x) and f -(x) = 0. On the other hand, if f (x) < 0, then
f''(x) = 0 and f- (x) = -f(x). If f is measurable on A. then f+ and f- are nonnegative
measurable functions on A with
f(x) =f'(x) -f-(x) and lf(x)I =f`(x) + f (x)
for all x E A.
the integral off over A by
Our natural inclination is to definejf+dA
jfdA = - jrdA.
A A A
The only problem with this definition is that it is possible that fAf' dA = fA f- dA = co,
giving the undefined oo - oo in the above. However, if we assume that both f' and f-
are integrable on A, then the above definition makes sense. Furthermore, if f is measur-
able, and f' and f- are both integrable on A, then I f j is also integrable on A. Con-
versely, if f is measurable and f I is integrable on A, then since f' s If I and f - s if 1,
by Theorem 10.7.4(c) both f' and f - are integrable on A. Therefore, we make the fol-
lowing definition.
10.7.7 DEFINITION Let f be a measurable real-valued function defined on a measurable

subset A of R. The function f is said to be (Lebesgue) integrable on A if If I is integrable
on A. The set of Lebesgue integrable functions on A is denoted by 2(A). For f E 2(A),
the Lebesgue integral off on A is defined by
f f dA = f f+ dA - jf- dA.
A A A
Remark. The set Z(A) of Lebesgue integrable functions on A is often also denoted
by -T'(A).
The definition of the general Lebesgue integral is consistent with our definition of
the Lebesgue integral of a bounded function on [a, b'. If f is a bounded real-valued
measurable function on [a, b] then so are the functions f' and f-. Let E, =
{x: f(x) a 0} and E2 = {x: f(x) < 0}. Then E, and E2 are disjoint measurable subsets
of [a, b] with E, U E2 = [a, b]. Furthermore, AE, = f" and fXE, = -f -. By Theo-
rem 10.6.10(b)
f
jfdA = j fdA + jfdA
fb f bb b (b
fXE, dA + dA = j f + dA - 1 f - dA.
a a a a
Remark For a nonnegative measurable function, the Lebesgue integral and the improper
Riemann integral are the same, provided of course that the latter exists (Exercise 18). This,
however, is false for functions that are not nonnegative. For example, consider the function
f(x) = (sinx)/x, x E [7r,oo), of Example 6.4.4(b). By Exercise 7 of Section 6.4, the im-
proper integral off exists on [-rr, oo). However, as was shown in Example 6.4.4,
j
tn
sinxl
IfIdA= lintj "I x dx=00.
it
Thus f is not Lebesgue integrable on [ir,oo). Another such example for a finite interval
is given in Exercise 23. The crucial fact to remember is that a measurable function f is
Lebesgue integrable on a measurable set A if and only if IA If I dA < oo.
Our first result is the following extension of Theorem 10.7.4 to the class of inte-
grable functions.
10.7.8 THEOREM Suppose f and g are Lebesgue integrable functions on the measurable
set A. Then
(a) f + g a nd cf, c E R, are integrable on A with
j(f +g )dA= jf dA f g dA and f cfdA=cjfdA.

A A + A A A
(b) If f 5 g a.e. on A, then
jfdA 5 j g dA, with equality if! = g a.e.

A A
480 Chapter i0 Lebesgue Measure and Integration
(c) If A, and A, are disjoint measurable subsets of A, then
= J fdA+ 1 fdA.
JAU! A, A;
Proof. (a) The proof that cf is integrable and that fA cf dA = cfA fdA follows im-
mediately from the definition. Before proving the result about the sum, we first note that
the definition of the integral off on A is independent of the decomposition f = f - - f -.
Suppose that f = f, - f2 where f, and f, are nonnegative integrable functions on A.
Then
f'+f =f-+ f,,

and thus by Theorem 10.7.4,
jf+dA + jfdA = JA1A+ J f,dA.
Since all the integrals are finite,
J fdA = jfdA_ f f-dA = ffidA_ f f2 A.

A A A A A
If f and g are integrable on A, then by definition so are the functions f' + g` and
f + g-. Since f + g = (f+ + g+) - (f- + g-), by the above
j(f+ g)dA f (f++g+)dA- j(f- g)dA,

A = A A
= ji dA - ji- dA + jg dA - f g' dA
A A A A
= f fdA + jgdA.
A A
(b) If f s g a.e., then g f ?0 a.e. Therefore by part (a),
O-fJ(g)dA= jgdA - ffdA.

A A A
from which the result follows. The proof of (c) also follows from (a) and the fact that
since A, n At = b, fXA,UA, -- f XA, +.fXA;
Lebesgue's Dominated Convergence Theorem

Our second major convergence result is the following theorem of Lebesgue.
10.7.9 THEOREM (Lebesgue's Dominated Convergence Theorem) Let {fn} be a sequence

of measurable functions defined on a measurable set A such that Jim fn (x) = f(x)
exists a.e. on A. Suppose there exists a nonnegative integrable fiinction g on A such that
If. (x) I g(x) a.e. on A. Then f is integrable on A and
jfdA I'mJAfndA.
=
Proof. Since g is integrable on A, and I fn 15 g a.e. on A, by Theorem 10.7.4 each fn
is also integrable on A. Also, by Corollary 10.5.10, the function f is measurable on A.
Furthermore, by Fatou's lemma,
AA
jIfId!jIfnIdA s
Thus f is also integrable on A.
jgdA < oo.
A
.By redefining all the f,,, n E N. on a set of measure zero if necessary, we can with-
out loss of generality assume that If. (x) I <_ g(x) for all x E A. Consider the sequence
(g + fn}nEN of nonnegative measurable functions on A. By Fatou's lemma,
J (g + /) dA j(g+fn)dA
A = I A
= f gdA + lim
nix A
f fn A.
A
Therefore,
f f dA <_ lim f
n~x A
A.
A
S imilarly, by applying Fatou's lemma to the sequence {g - which is again a se-

quence of nonnegative functions on A,
f (g - f) dA s lim j (g - fn) dA = j g dA + Jiim j (-fn)dA.

A A A A
But lim fA ( fn)dA = -lim

n-oc
fA f A. Therefore.
n-*=
lim f fn dA.
f f dA > n-.x
A A
Combining the two inequalities gives the desired result. O

Remark. The hypothesis that there exists an integrable function g satisfying I fn I !f- g
a.e. is required in the proof in order to subtract f g dA in the above inequalities. This is
not possible if f gdA = oo. As the following example shows, if such a function g does
not exist, then the conclusion may be false.
10.7.10 EXAMPLE As in Example 10.6.12(a) consider the sequence { fn} on [0, 1] defined by
In, 0 <x< 1
fn(x) = n'
10, elsewhere.
For the sequence { fn} we have lim fn (x) = 0 for all x E [0,1 ] but f o If,, dA = 1 for
all n. We now show that any measurable function g satisfying g(i) ? fn (x) for all
X F: [0,I] and n E N satisfies f o 11gdA = oo. Since g(x) _' fn(x) f o r all x E (0,I]
we have g(x) ? n for all x E (n:11 ,1-, ]. Since the collection {(kk + 11, k]}k=t of intervals is
pairwise disjoint, by Theorem 10.7.4,
n
J gdA gdAa ,kA\I k+1 1 k1J/

[0 11 k=1 ft I k+I) 1-1 k=1
Since the series 7 k+lt diverges, we have f(o 1)gdA = 00.
EXERCISES 10.7
1. Let f be a nonnegative measurable function on a measurable set A. If f4 f dA = 0. prove that f = 0 a.e. on A.
2. *a. Prove Theorem 10.7.4(b).
3. Let A be a measurable subset of R.
a. If f is integrable on A and g is bounded and measurable on A, prove thatfg is integrable on A.
b. If f and g are integrable on A, is the function fg integrable on A?
4. *Let f,(x) = x-°, x E (0, 1). Prove that f, is integrable on (0, 1) for all p, 0 < p < 1, and that
1
(Q,,
J(-' 1)
1 --p'
1 p.
5. Define f on [ 1, oo) by
f(x) _ { V, if x E [n, n + l ln'"), n = 1, 2, .. ,
0, otherwise.
Show that f E X([ 1, oo)) but f2 `.P([ 1, oo)).
6. Let P denote the Cantor set of Section 3.3. Define f on [0, I ] as follows: f(x) = 0 for every x E P. and f(x) = k
for each x in the open interval of length 1/3k on [0, 1] 1 P. Prove that f is integrable on [0, 1 ] and that fo fdA = 3.
7. *Let f be a nonnegative integrable function on [a, b]. For each n E N, let E. = {x: n s f(x) < it + 1 }. Prove that
7, , nA(E,) < oo.
8. Evaluate( each of the following limits:
a, lim J (I - dx b. lim
n-.oo
(1 - x/n)nex"2 dx
1
(0. 1) L. W
9. Let f be an integrable function on (a, b]. Given e > 0, prove that there exists a bounded measurable function g
on [a, b] such that fla bl If - g I dA < e.
10. *Suppose f is integrable on a measurable set A with A (A) = oo. Given e > 0, prove that there exists a bounded
measurable set E C A such that LIE I f I dA < E.
11. Show by example that Fatou's lemma is false if the functions f,,, n E N, are not nonnegative.
12. Show by example that the bounded convergence theorem is false for a measurable set A with A (A) = oo.
13. *Let f be a Lebesgue integrable function on a measurable set A. Prove that given e > 0, there exists a 8 > 0 such
that fE If I dA < e for all measurable subsets E of A with A (E) < S.
14. Let f be amr integrable function on (a, b), where -oo <_ a < b s oo. Define F on (a, b) by
F(x) = 4 z f dA, x E (a, b).

J,
a. Prove that F is continuous on (a, b).
b. If f is continuous at x, E (a, b), prove that F is differentiable at x, with F'(x,) = f(x,).
15. If f E `.e([0, I ]), show that for every t E R the function x - sin (rf(x)) E `.e([0, 1]), and that
g(r) = flo it sin(tf(x))dA(x) is a differentiable function oft E R. Find g'(t).
16. If f G X(R), prove that lim JR f(x) sin nx dA = 0. (Hint: First prove the result for step functions.)
17. If f e T(R), prove that ,lim JR If(x + h) - f(x) I dA(x) = 0.
18. Let f be a nonnegative measurable function on (a, b] satisfying f E %[c, b] for every c, a < c < b. Prove that
fa fdA = Clinn+f" f(x) dx.
19. *Monotone Convergence Theorem: Let I f.1 be a monotone increasing sequence of nonnegative measurable
functions on a measurable set A. Prove that fA (lim dA = qim jA f,, dA.
20. Show that the monotone convergence theorem is false for a monotone decreasing sequence of measurable func-
tions.
21. *a. Let f be a nonnegative measurable function on a measurable set A, and let {A,} be a sequence of pairwise dis-
joint measurable subsets of A with U. A. = A. Prove that
= 7, jfdA.
00
n"I A,
b, Prove that the conclusion of part (a) is still valid for arbitrary f e T(A).
22. Let { f,} be a sequence of measurable functions on (a, b) satisfying I f,(x)I s g(x) a.e., where g is integrable on
[a, b]. If liimf,(x) = f(x) exist a.e. on [a, b], and h is any bounded measurable function on [a, b], prove that
(b
fh dA = lim f f,h dA.
1
111
a
23. *As in Exercise 4, Section 6.4, let f be defined on (0, 1) by
f(x) =±(xZ sin z) I
Prove that f is not L.ebesgue integrable on (0, 1).

24. Let f be a nonnegative measurable function on [a, b]. For each t 0, let
mf(t) = A({x E [a, b] : f(x) > t}).
a. Prove that m,(t) is monotone decreasing on (0, oo).
b. Prove that fa fdA = fo m1(t) dt.
484 Chapter 1 o Lebesgue Measure and Integration
10.8 Square Integrable Functions

In analogy with the space 12 of square summable sequences, we define the space 222 of
square integrable functions as follows.
10.8.1 DEFINITION Let A be a measurable subset of R. We denote by T2(A) the set of all
measurable functions f on A for which I f I2 is integrable on A. For f E f2(A), set
IIfII2=\JaIfI2dA/
The quantity IIf II2 is called the 2-norm or norm off. Clearly. IIf II2 >_ 0, and from
the definition it follows that if f E _T22(A) and c E 18, then cf E Y2(A) with II cfII2 =
IcIIIf1I2. We will shortly prove that if f, g E `.P2(A), then f + g E Y2(A) with
Ill + 8 112 < IVII2 + 119 112- Thus .T2(A) is a vector space over R.
If f E 22(A) satisfies IIf II2 = 0, then by Exercise 1, Section 10.7, 1 f 122 = 0 a.e., and
thus f = 0 a.e. on A. This does not mean that f (x) = 0 for all x E A; only that f = 0
except on a set of measure zero. Thus II 112 satisfies all the properties of a norm except
for IIf II2 = 0.if and only if f = 0. To get around this difficulty we will consider any two
functions f and g in X2 (A) for which f = g a.e. as representing the same function. For-
mally, we define two measurable functions f and g on A to be equivalent if f = g a.e. In
this way it is possible to define 22(A) as the set of equivalence classes of square inte-
grable functions on A. Rather than proceeding in this formal fashion, we will take the
customary approach of simply saying that two functions in f22 are equal if and only if
they are equal almost everywhere. With this definition Y2(A) is a normed linear space.
10.8.2 EXAMPLES
(a) For our first example, let f (x) = 1/Vc, x E (0, 1). By Exercise 4 of the previous
section, f is integrable on (0, 1) with 1(0 1)f dA = 2. Since f 2(x) = 1/x, by Example
10.7.3(b), f2 is not integrable on (0, 1) and thus f `12((0, 1)). On the other hand, if
g(x) = x- 'I, then g2(x) = x-21-', which by Exercise 4 of the previous section is inte-
grable. Thus g E `12((0, 1)) with
IIBII2 J x 2/'dx=3.
0
(b) Consider the function f (x) = 1/x for x E [ 1, oo). For any n E N, n ? 2,
J IfI2dA = J x2dx = rI
L i
- 1J.
n
Thus
IIfII2 = I hfI2dA = nix

lim J°IfI2dA = I.
I
Therefore, f E 12([ 1, oo)) with 11f112 = 1. It is easily shown that f E1([ 1, x)).
Cauchy-Schwarz Inequality
Our first result will be the analogue of the Cauchy-Schwarz inequality for 12. The fol-
lowing inequality is sometimes also referred to as Holder's inequality.
10.8.3 THEOREM (Cauchy-Sehwarz Inequality) Let A be a measurable subset of R. If

f, g E 5f2(A), then fg is integrable on A with
j
A
1181 dA s 111 1I211g112
Proof. By Theorem 10.5.4, the productfg is measurable, and for any x E A, we have
I f(x)g(x) 1:S
1 (I f(x) 12 + Ig(x)12).
2
Since by hypothesis f, g E `.E2(A), the function I fgI is integrable on A, and thusfg is in-
tegrable on A. As in the proof of Theorem 7.4.3, for y E R,
0s f (IfI -yIBI)2dA=IIf
A
j lfgI dA.
A
(9)
If II9II2 = 0, then by Exercise 1 of Section 10.7, g = 0 a.e. on A. As a consequence,

fAl fgl dA = 0, and the conclusion holds. If I19112 * 0, set y = (fAlfgldA)/Ilgll2. With
y as defined, equation (9) becomes
(fAIfgI dA)2
0 111112 -
IIg112
and thus (fA I fg I dA)2 11f 112118112, which proves the result. Q
Our next result is Minkowski's inequality for the space X22. Since the proof of this
is identical to the proof of Theorem 7.4.5, we leave the details to the exercises.
10.8.4 THEOREM (Minkowski's Inequality) Let A be a measurable subset of R. If

f, g E 2(A), then f+ g E `.C2(A) with
111+g112s111112+118112
The Normed Linear Space ,22([a, b])

If A is a measurable subset of R, the norm II Ikon 2(A) satisfies the following prop-
erties:
(a) Of 112 ? 0 for any f E _T2(A).
(b) 111112 = 0 if and only if f = 0 a.e. on A.
(c) IIcf 112 = I c 1111 112 for all f E W2(A) and c E R.
(d) IV + 9 112 111112 + IIgh12 for all f, g E e2(A).
Properties (a) and (c) follow from the definition, and (d) is Minkowski's inequality.
With the convention that two functions f and g are equal if and only if f = g a.e.. T2(A)
is a normed linear space.
By Definition 8.3.9, a sequence {/n} in ce2 converges in norm to a function f in `.EZ
if and only if Jim Of - fn112 = 0. Example 10.6.12(b) shows that convergence in Y2 does
not imply pointwise convergence of the sequence. As in Chapter 9, convergence in S2
is usually called convergence in the mean. We now prove that T22([a. b]) is a complete
normed linear space.
10.8.5 THEOREM The normed linear space (.T2([a, b]), 11 A2) is complete.
Remark. Although we state and prove the result for a closed and bounded interval.
the same method of proof will work for T2(A) where A is any measurable subset of R.
Before proving the theorem, we first state and prove the following lemma.
10.8.6 LEMMA Let A be a measurable subset of R. Suppose {f.1 is a monotone increasing

sequence of nonnegative measurable functions on A satisfying
for all n E IN,

A
and for some finite constant C. Then f (x) = lim is finite a.e. on A.
Proof. Since { f (x)} is monotone increasing for each x E A, the sequence either con-
verges to a real number or diverges to oo. Let f(x) = liymeOfn(x). and let
E_ {xEA:f(x) = oo}.
We will prove that A(E) = 0. For each k E N, let
EE= {xEA:f(x)>k}.
Then Ek D Ek+, for all k E tN with lkE Ek = E. For fixed k E IN, set
An, k= {x E A: fn(x) > k}, n E N.
Then A..k C An+I.k With U. EN A..k = Ek. Thus by Theorem 10.4.6(a),
A(Ek) = lim A(An.k)

But
A(A..k)= j 1dA5 1 j .fndA IjfndA!Ck

A.., k A.., kA
Therefore, A(Ek) s C/k for all k E N. Since A(E)) < oo, by Theorem 10.4.6(b),
A(E)= limA(EE)=0.
Therefore, f is finite a.e. on A. 0
Proof of Theorem 10.8.5. Let {fn} be a Cauchy sequence in 22. Then given
e > 0, there exists no E Ni such that Ilfn - fn II2 < e for all n, m ? no. For each k E NI,
let nk be the smallest integer such that Ilfn - fn 112 < 1/2k for all m, n ? nk. Then
n1cn2<...
CnkG...,and
IIJn,., ll2 <
2k.
For each k E N, set

gk = Ifn,l + Ifn_ -fn,I + .. ' + IA., -Jn.I.
By Minkowski's inequality,
gk dA = 118k112 llfn,ll2 + Ilfn

[a.b] \ j-1
(Ilf.,112 + 2k)2 (Ilfn,ll2 + 1)2.
Thus the sequence {gk} satisfies the hypothesis of Lemma 10.8.6. Therefore, urn k is
finite a.e. on [a, b]. But then lim gk is also finite a.e. on [a, b]. As a consequence, the
k-+oo
series
00
If, (x)I + Y, If.,.,(x) -fn,(x)I

j-1
converges a.e. on [a, b], and therefore so does the series
00
fn,(x) + Y,(fn,.,(x) -fn,(x))

j=1
But the kth partial sum of this series is Therefore, the sequence {fn,}kEh con-
verges a.e. on [a, b]. Let E denote the set of x E [a, b] for which this sequence con-
verges. Then A([a, b) \ E) = 0. Define
f(),x=- lim fn,(x)' x E E,
J 0, otherwise.
Then {f,} converges to f ae. on. [a, b].
It remains to be shown that f E 22, and that { fn} converges to f in 2 2. Since
IfnIf ..I
+
k
j=1
p
Ifnf., -fn,l = 8k
by Fatou's lemma
IfI2dA5 {im gkdA<oo.

'a.b] k-ico a b]
Thus f r= 22. Finally, since

f(x) - f., (x) = (f0,.,(x) - f., (x)) a.e.,
by Fatou's lemma again,

2
iab]
(f n'.' - f., 11 < \2
(a, b]
Therefore, II f - fn,112 < 1/2k for all k E N. Thus the subsequence { f,,,}kEN converges to
f in the norm of _y2. Finally, by the triangle inequality,
Of -f,112 < Of -fn,ll2 + Ilfn, -fnll2 < 2k + IIfn, -fnll2-
From this it now follows that the original sequence { fn} also converges to f in the norm
of 22.
Fourier Series
We close this chapter by making a few observations about Fourier series and the space
22([-ir, ir]). As in Definition 9.3.1, if f is Lebesgue integrable on [-a, ir]. the
Fourier coefficients of f with respect to the orthogonal system {1, cos nx, sin nx}.,
are given by
A
an = I f(x) cos nx dx, n = 0, 1, 2,. . ,

J
bn = 1 f(x) sin nx dx, n = 1, 2,
Since f is Lebesgue integrable, the functions f(x) sin nx and f(x) cos nx are measurable
on [-zr, -rr], and in absolute value less than or equal to I f(x)I. Thus the functions
f(x) cos nx and f(x) sin nx are all integrable on [ -ir, a].
The same method used in proving Bessel's inequality for Riemann integrable func-
tions proves the following (Exercise 8): If f E :E2([-ar, ir]) and {ak} and {bk} are the
Fourier coefficients off, then
00 i 7r
Za0+kY, (ak+bk)s f2dA. (Bessel's Inequality)

I
J A
Thus the sequences {ak}k o and {bk}' , are square summable. We now use complete-
ness of the space 22([-ir, it)) to prove the converse.
10.8.7 THEOREM If {ak}k=o and {bk}k , are any sequences of real numbers satisfying
1 00
2 ap + (ak + bk) < oo,
k=I
then there exists f E `.E2([-tr, rr]) whose Fourier coefficients are precisely {ak} and
{bk}.
Proof. For each n E t\!, set

n
S,(x) = 1 ao + I (ak cos kx + bk sin kx).

2 k-1
Since each S. is continuous, Sn is square integrable on [ -ir, a]. If m < n, then

A
11Sn - S'11'2 = J Ik=m+
7, I (ak cos kx + bk sin kr)dx,
2
n
which by orthogonality
= jr j, (a + b2E).
Since the series converges, the sequence {Sn] is a Cauchy sequence in `.C2([-Tr, a i).
Thus by Theorem 10.8.5, there exists a function f E X2([-zr, ar]) such that Sn con-
verges to f in 22; i.e., lim 11f - S,,92 = 0. If n > m, then
A A
Sn(x) cos mx dx = -cram and S. (x) sin mx dx = 1rbm.

A -A
Therefore,
Jf(x) COS mx dx - as 1 V (x) - S,(x)) cos mx dx.

1 A
A 1
--
which by the Cauchy-Schwarz inequality
11f- S.11211COS 'nX 112 = Ilf- S,112-
Since this holds for all n > m, letting n --> oo gives
1
am - jf(x)cosmxdr.
-A
A similar argument proves that the b,n are the sine coefficients off. O
Is Every Trigonometric Series a Fourier Series?

In Section 9.3 we showed that the series
sin nx
Inn
even though it converges for all x E R, is not the Fourier series of a Riemann integrable
function on [-a, ir]. Since 7, (in n ) -2 = oo, by Bessel's inequality neither is it the
Fourier series of a square integrable function. This, however, does not rule out the pos-
sibility that it is the Fourier series of a Lebesgue integrable function. The following
classical result is very useful in answering this question. Since the proof of the theorem
is beyond the level of this text, we state the result without proof.2
10.8.8 THEOREM
(a) If b > 0 for all n and 7,0,0-1 b,/n = oc, then

00
b, sin nx
is not the Fourier series of a Lebesgue integrable function.

(b) If {a,} is a sequence of nonnegative real numbers with lim a = 0. satisfying
a, c 1 (a, _, + a,, 1), then the series
00
7, a,cosnx
n-t
is the Fourier series of a nonnegative Lebesgue integrable function on [ -v, it 1.
Since the sequence { 1/(ln n)} satisfies hypothesis (a), the series
00 sin nx
n.2 Inn
is not the Fourier series of any integrable function on it, a]. However, it is interest-
ing to note that the sequence {1/(ln n)} also satisfies hypothesis (b) (Exercise 13), and
thus the series
cos nx
71 Inn
,-2
is the Fourier series of a nonnegative Lebesgue integrable function.
EXERCISES 10.8
1. *For x E (0, 1) let fo (x) = x'°, p > 0. Determine all values of p such that f, E Z2((0, I )).
2. For each n E N, let 1, = (n, n + =,). For a given sequence {c,} of real numbers, define f on 11. oo) by
f(x) = 7,W , c, Xtn (x). Show that f E 22([ 1. oo)) if and only if
00c2
a < oo.
n.l n
3. Find an example of a real-valued function f on (0, oo) such that f2 is integrable on (0, oo). but If V° is not in-
tegrable on (0, oo) for any p, 0 < p < oo, p * 2.
(Hint: Consider the function 8(x) l

x(1 + llnxl)2
2. A proof of the result can be found in Chapter V of the text by Zygmund.

Notes 491

S. Let A be a measurable subset of R with A(A) < oo. If f E f2(A), prove that f E Y(A) with
f If I dA < tlf 112 (A(A))1/2

A
6. Let { fa} be a sequence in 22([a. b]). Suppose {fa} converges in 22 to f E T2 and {fa} converges a.e. to some
measurable function g. Prove that f = g a.e. on [a, b].
7. 'Let If.) be a sequence in 22(A) that converges in cy2 to a function f r= If g E 222(A). prove that
lim
a-0D J
fag dA = jfdA.
A A
& If f E T2([-ir, ar]) and {ak} and {bk} are the Fourier coefficients of f, prove that
I
ao+ I(at+b})s-ar-° f2dA.
O° 1
2 k°1
9. Which of the following trigonometric series are Fourier series of an 22 function?

- cos nx sin nx C. cos nx
a. b
2 VnInn
10. Suppose E is a measurable subset of (-a, ir) with A(E) > 0. Prove that for each S > 0, there exist at most fi-
nitely many integers n such that sin nx z 8 for all x E E.
11. Let f r= 12([a, b]). Prove that given e > 0, there exists a continuous function g on [a, b] such that
11f - g 16 < e. (Hint First prove that there exists a simple function having the desired properties, and then use
Exercise 15, Section 10.6, and Lemma 9.4.8.)
12. For f E ?2([-w, ar]) let {ak} and {bk} be the Fourier coefficients off.
a. Prove Parseval's equality:
00
1
2ao+ (qk' +b,t)= f2dA.
A
b. If at = bk = 0 for all k, prove that f = 0 a.e.

13. *Show that the sequence (I /(In n)} satisfies the hypothesis of Theorem 10.8.8(b).
NOTES
Lebesgue's development of the theory of measure and in- convergence of the sequence (f.) is required. Otherwise,
tegration was one of the great mathematical achievements the limit function may not be Riemann integrable. On the
of the twentieth century. His proof that every bounded other hand, if { fa} is a sequence of measurable functions
measurable function is Lebesgue integrable was based on on [a, b], then its pointwise limit f is also measurable.
the new idea of partitioning the range of a function, rather Hence if f is also bounded, then f is integrable. The
than its domain. Lebesgue's theory of integration also bounded convergence theorem is notable for its simplicity
permitted him to provide necessary and sufficient condi- of hypotheses and proof. It only requires that (f.) be uni-
tions for Riemann integrability of a bounded function f. formly bounded and converge a.e. on [a, b]. This is suffi-
In addition to the fact that the Lebesgue integral en- cient to ensure that
larged the family of integrable functions, its power results
from the ease with which it handles the interchange of f, dA = (lim
limits and integration. For the Riemann integral, uniform n
Urn
-.m
a, b] o. bJ
.x f.) A.
With the additional hypothesis that the pointwise limit f is function is differentiable a.e. on [a, b).3 A slight general-
Riemann integrable, the bounded convergence theorem is ization of Theorem A then gives that
also applicable to a sequence if,) of Riemann integrable
functions.
F'dA = F(x) - F(a) = J fdA,
The bounded convergence theorem and the dominated
convergence theorem are the tools required to prove the
fundamental theorem of calculus for the Lebesgue integral. or that f.[ F' - f] dA = 0 for all x E [a. b]. As a conse-
quence of Miscellaneous Exercise 5 we have F' = f a.e.
THEOREM A 1ff is differentiable and f is bounded Newton and Leibniz realized the inverse relationship of
on [a, b], then f is Lebesgue integrable, and differentiation and integration. The above two versions of
fa f'dA = f(b) - f(a). the fundamental theorem of calculus provide a rigorous
formulation of this inverse relationship for a large class of
The proof of this result was requested in Exercise 8 of functions.
Section 10.6. It follows simply by applying the bounded The Lebesgue theory of integration also provides the
convergence theorem to the sequence defined by proper setting for the study of Fourier series. The bounded
g (x) = n[f(x + '-,) - f(x)]. Since f is differentiable, the convergence theorem was used by Lebesgue to prove uni-
sequence converges pointwise to f. Also, by the mean queness of a Fourier series. If
value theorem the sequence is uniformly bounded on
[a, b]. This then establishes the analogue of Theorem 6.3.2
for the Lebesgue integral. If instead of bounded, one as- f(x)=2Ao+ k-l (Ak cos kx + 8k sin kx).
sumes that f is Lebesgue integrable, then the result also
follows by Lebesgue's dominated convergence theorem.
for all x E [ -it. zr], then f, being the pointwise limit of a
The Riemann theory of integration allows us to
sequence of continuous functions, is automatically mea-
prove that if f E gt[a, b], and F is defined by F(x)
surable on [-rr, a]. If f is also bounded, then f is inte-
f.'f(t) dt, then F'(x) = f(x) at any x e [a, h] at which f
grable. If the sequence of partial sums of the trigono-
is continuous. Since f E 9l[a, b] if and only if f is con-
metric series is uniformly bounded, then by the bounded
tinuous a.e. on [a, b], we have that F'(x) = f(x) a.e. on
convergence theorem, the trigonometric series is the Fourier
[a, b]. Although not proved in the text, this result is still
series off. In his 1903 paper Sur les series trigonometric,'
valid for the Lebesgue integral.
Lebesgue showed that uniform boundedness of the partial
sums may be removed; that boundedness of the function f
THEOREM B Let f be Lebesgue integrable on (a, b),
itself was sufficient. This result was extended in 1912 by
and define F(x) = f, f dA. Then F'(x) = f(x) a. e. (a, b). de la VallEe-Poussins to the case where the function f is in-
By writing f = f * - f -, it suffices to assume that tegrable on (- rr, a ). See the article by Alan Gluchoff for
f 0. and thus F is monotone increasing on [a, b). It is an overview of how trigonometric series has influenced the
a fact independent of integration that every monotone various theories of integration.
1. Let A be a measurable subset of R. For f E 2(A) set lI f 11, = f,, I f I dA. Prove that (.T(A ), (I (I,) is a complete
normed linear space.
3. See page 208 of the text by Natanson.

4. Annales Scientifiques de 116cole Notmale Supdrieure. (3) 20 (1903).453-485.
5. "Sur l'unicitd du developpement trigonomEtrique;' Bull de t'Acad. Royale de Belgique (1912).702-718-.
see also Chapter 9 of the text by Zygmund.
2. Let A be a measurable subset of R, and let f, f, n = 1, 2, ... be measurable functions on A. The sequence If.) is
said to converge in measure to f if for every 6 > 0.
lim A({x E f(x)l 6}) = 0.
a-m
a. If is a sequence in :E(A), and converges in the norm of .Y(A) to f E `.E(A). prove that converges in
measure to f.
b. Find a sequence ff.) of measurable functions on a measurable set A that converges to a function fin measure.
but does not converge to f in norm.
c. If A(A) is finite and (f.} is a sequence of measurable functions that converges in measure to a measurable func-
tion f, prove that there exists a subsequence { fk} of {f} such that fk -+f a.e. on A.
d. Show that the sequence { f } of Example 10.6.12(b) converges to 0 in measure.
e. Find a sequence {f.) of measurable functions on a measurable set A such that f -a 0 everywhere on A but {
does not converge to 0 in measure.
3. Let be a sequence of orthogonal functions on [a, b] having the property that the zero function is the only con-
tinuous real-valued function f satisfying fa fkp, dA = 0 for all n E N. Prove that the system is complete.
(Hint: First use the hypothesis to prove that if f 2([a, b]) satisfies Ja' dA = 0 for all n E N. then f = 0
a.e. Next use completeness of `. b2 to prove that Parseval's equality holds for every f E `f ([a, b]).)
4. Construction of a Noumeasurable Set: For each x E [-12, Z], define the set K(x) by
K(x)={yE[-12,Z]:y-xE0}.
a. Prove that for any x, y E [ -s, ;], either K(x) n K(v) = 4 or K(.r) = KO,). (Note: K(x) = K(y) does not imply
that x = y; it only implies that x - y is rational.)
Consider the family & = {K(x):x E [ -12,121} of disjoint subsets of J. Choose one point from each
distinct set in this family and let A denote the set of points selected. The ability to choose such a point from
each of the disjoint sets requires an axiom from set theory known as the axiom of choice. Further information
about this very important axiom can be found in the text by Halmos.
Let rk, k = 0, 1, 2, ... be an enumeration of the rational numbers in [ -1. I ], with r1 = 0. For each
k=0,1,2..., setAkA+rk.
b. Show that the collection {Ak} is pairwise disjoint with [- i, ] CU u Ak C [- _, ;
c. Use the above to show that A*(A) = 0 and A*(A) > 0, thus proving that A is nonmeasurable.
S. Suppose f is Lebesgue integrable on [a, b]. If f. f dA = 0 for every x E [a, b], prove that f = 0 a.e.
Botts, Truman, "Probability theory and the Lebesgue in- Katznelson, Y. and Stromberg, K., "Everywhere differ-
tegral," Math. Mag. 42 (1969). 105-1 I1. entiable, nowhere monotone function;' Amer. Math.
Burkill, H., `"The periods of a periodic function;' Math Monthly 81(1974), 349-354.
Mag. 47 (1974),206-210. Kraft, R. L., "What's the difference between Cantor
Darst, R. B., "Some Cantor sets and Cantor functions;' sets?" Amer. Math. Monthly 101 (1994). 640-650.
Math. Mag. 45 (1972). 2-7. Ladder, G., "A simple introduction to integral equa-
Dressler, R. E. and Stromberg, K. R., '"The Tonelli inte- tions," Math. Mag. 69 (1996), 172-181.
gral," Amer. Math. Monthly 81 (1974), 67-68. Maligranda, L., "A simple proof of the Holder and
Gluchoff, Alan D.. 'Trigonometric series and theories of Minkowski inequality," Amer. Math. Monthly 102
integration;' Math. Mag. 67 (1994), 3-20. (1995), 256-259.
McShane, E. J., "A unified theory of integration;' Amer. Varberg, Dale E., "On absolutely continuous functions.
Math. Monthly 80 (1973), 349-359. Amer. Math. Monrhl' 72 (1965). 831-841.
Priestly, W. M., "Sets thick and thin;' Amer. Math. Wade. W. R., 'The bounded convergence theorem;'
Monthly 83 (1976), 648-650. Amer. Math. Monthly 81 (1974). 387-389.
Appendix:
Logic and Proofs
A.1 Propositions and Connectives
A.2 Rules of Inference
A.3 Mathematical Proofs
A.4 Use of Quantifiers
In searching for new mathematics, mathematicians use their imagination and ingenuity
to make a conjecture about new theories based on observations of particular cases or
phenomena. This approach is called inductive reasoning. Natural or social scientists
may test the conjecture by making further observations. If the results are incompatible
with the conjecture, then scientists must usually reject or modify the theory. Mathe-
maticians, on the other hand, attempt to change the conjecture into a theorem by prov-
ing (or disproving) that the conjecture follows logically from the accepted axioms, the
hypothesis, and proved theorems of other mathematicians. In constructing a proof
mathematicians are restricted to the use of deductive reasoning-the use of logic to
draw conclusions based on hypothesis and statements accepted as true.
In developing new mathematics, mathematicians have the advantage of not being
restricted to the study of observable phenomena or even axioms that have been accepted
as true. Euclid's fifth postulate (the parallel axiom) had been accepted as true for over
2200 years until the early nineteenth century when Janbs Bolyai (1802-1860), Carl
Friederich Gauss (1777-1855), and Nicolai Lobatchewsky (1793-1856) challenged
this axiom. The acceptance of alternatives to the parallel postulate led to the develop-
ment of new and interesting non-euclidean geometries, many of which are still very im-
portant today.
Deductive reasoning provides mathematics with an efficient means of both exposi-
tion and organization of its subject matter. In what follows we will attempt to explain
this formal, logical side of mathematics. Formal logic will not indicate the steps neces-
sary to prove a conjecture; it only provides a system of rules and procedures used to
495
496 Appendix Logic and Proofs
decide whether a given statement follows from some given set of statements. Intuition.
insight, and knowledge are still required to formulate the necessary steps in a proof.
A.1 1 Propositions and Connectives

The English language, as most languages, allows for a variety of sentences: interroga-
tory sentences (Will I pass this course?), exclamatory sentences (I got an A:). and sen-
tences that are either true or false. In the language of logic the latter are called proposi-
tions.
A.1.1 DEFINITION A proposition is a sentence or statement that is either true (T) or

false (F).
Some examples of propositions are
(a) "4 is an even number."
(b) "\ is rational.'
(c) "The author of this text was born in Frankfurt. Germany."
We certainly accept that (a) has the value T, whereas in the introduction to Chapter I it
is proved that (b) has the value F. For (c), however, even though you may not know
whether the statement is true or false, the statement does have either the value T or F.
The above are all examples of simple propositions; that is, propositions that con-
sist of a single statement. Simple propositions can be combined using the logical con-
nectives "and;' "or," and "not" to form compound propositions.
A.1.2 DEFINITION Given two propositions P and Q,

(I) The negation of P. denoted - P, is the proposition "not P." The proposition
P is true exactly when P is false.
(ii) The conjunction of P and Q, denoted P A Q, is the proposition "P and Q."
The proposition P A Q is true exactly when both P and Q are true.
(iii) The disjunction of P and Q, denoted P V Q, is the proposition "P or Q." The
proposition P V Q is true exactly when at least one of P or Q is true.
The truth values of the above connectives can be summarized very nicely by means
of the following truth table.
P Q -P PAQ PVQ
T T F T T
T F F F T
F T T F T
F F T F F
The connectives "and," "or;" and "not" are illustrated by the following examples.
Using (a), (b), and (c) above we have
-- (b) "V is not rational."
(a) A (b) "4 is an even number and V is rational,"
(a) V (c) "4 is an even number or the author of this text was born in Frankfurt, Ger-
many."
Since the statement "V is rational" is false, the proposition (a) A (b) is also false. On
the other hand, since (a) is true. (a) V (c) is true regardless of where the author was
born.
In logic the use of the word "or" is the inclusive or, meaning "one or the other, or
both" as opposed to the exclusive or, meaning "one or the other, but not both" For
propositions P and Q the exclusive or can be expressed in terms of A, V, and by
(P V Q) A -- (P A Q) (see Exercise 2). In addition to and, other common connectives
used in English are but, while, and although. Each of these has the same meaning as
and, and is represented symbolically by A. For example, if P is "It is cold" and Q is "It
is snowing;" then the statement "It is cold, but it is not snowing" would be expressed
symbolically by P A - Q.
In addition to the connectives A, V, and -, there are two other connectives that are
commonly used in logic; namely, the conditional and biconditional.
A.1.3 DEFINITION Given two propositions P and Q,

(i) The conditional sentence P => Q (read "P implies Q") is the proposition "If
P, then Q." The statement P Q is a true statement unless P is true and Q is
false, in which case it is a false statement.
(ii) The biconditional sentence P * Q is the proposition "P if and only if Q." The
sentence P q Q is true exactly when P and Q have the same truth values, oth-
erwise it is false.
The symbol " " is referred to as the implication or conditional symbol, whereas
t* is the biconditional symbol. In "P ' Q," the proposition P is the hypothesis or an.
tecedent and Q is the conclusion or consequent. The truth values for P ' Q and
P a Q are given in the following table.
P Q P=*Q PaQ
T T T T
T F F F
F T T F
F . F T T
In the truth table for P = Q, the only line where P Q is true and P is true is the
first line, where Q is also true. Thus the conditional statement P Q is often also ex-
pressed by saying that P is a sufficient condition for Q (if P is true then Q follows), or
that Q is a necessary condition for P (P cannot be true unless Q is true). That Q is a
necessary condition for P is sometimes also expressed by the phrase "P. only if Q"
Since the biconditional P,4* Q is true exactly when P and Q have the same truth val-
ues, this is often also verbally expressed by "P is necessary and sufficient for Q:'
To illustrate the truth values assigned in the conditional sentence, let us consider
the following example. Your professor agrees
"If you earn an A on the final, I will assign you an A for the course."
Here the antecedent P is "You earn an A on the final" and the consequent Q is "He as-
signs you an A for the course" The only case in which you have reason to be angry (the
sentence is false) is when P is true and Q is false. If both P and Q are false, you may
not be happy, but you have no cause to be angry with your professor. On the other hand,
if P is false and Q is true, you certainly will not be angry.
Closely related to the conditional sentence are the converse and the contrapositive
of P =*. Q.
A.l.4 DEFINITION For propositions P and Q, the converse of P . Q is Q =o. P, and the
contrapositive of P . Q is --Q' --P.
The truth values for each of these is given in the following table.
P Q P=#. Q Q .P --Q=:I, --P

T T T T T
T F F T F
F T T F T
F F T T T
A propositional formula is an expression involving finitely many logical connectives

(such as n, V, --, =, and e') and variables (such as P, Q, R, etc.). For example,
(P A (Q V R)) V -R
is a propositional formula; it becomes a proposition when the letters P, Q, and R repre-
sent propositions and thus have either the truth value T or F. Two propositional formu-
las are equivalent if and only if they have the same truth values for all assignments of
truth values to the simple propositions making up the propositional formulas. Simply
stated, two propositional formulas are equivalent if and only if they have the same truth
table. For example, the formulas P . Q and -P V Q are equivalent. This is verified by
the following truth table.
P Q P=Q -PvQ
T T T T
T F F F
F T T T
F F T T
Also, from the table for the contrapositive we see that -Q -P is equivalent to
P Q. To emphasize this we state it as a theorem.
A.1.5 THEOREM The proposition P Q is equivalent to its contrapositive -Q = -P.

Exercise I contains several pairs of equivalent propositional formulas. These are
important and should be memorized.
Some propositional formulas have the property that they are always true regardless
of the assignment of T or F to the simple propositions making up the formula. For ex-
ample, it is easily verified by a truth table that each of the propositional formulas
PV - P and (P A Q) r* P
have the value T for any assignment of T or F to P and Q. Such propositions are called
tautologies. A tautology is a propositional formula that is true for every assignment of
truth value to its components. A contradiction is the negation of a tautology. Thus
(P V -P) is a contradiction. By means of a truth table it is easily verified that
-- (P V -P) is equivalent to -P A P, which simply states that "P" and "not P" can-
not both be true simultaneously. This is sometimes referred to as the law of the excluded
middle. Exercise 3 contains several basic tautologies that will be very useful in our dis-
cussion on rules of inference.
From the definitions of equivalence of propositional formulas and the biconditional
it should be clear that two propositional formulas P and Q are equivalent if and only if
P q Q is a tautology. To emphasize its importance we state it as a theorem.
A.1.6 THEOREM Two propositional formulas P and Q are equivalent if and only if P q Q
is a tautology.
To illustrate this we consider the equivalent propositional formulas P Q and
-P V Q. The truth table for (P Q) a (-P V Q) is as follows:
P Q (P * Q) q (--P V Q)
T T T T T
T F F T F
F T T T T
F F T T T
Thus (P Q) q (-P V Q) is a tautology.
EXERCISES Al
,
1. For propositions P. Q, and R, verify by means of a truth table that each of the following pairs of propositional for-
mulas are equivalent.
a. ^(-P) and P b. P A Q and Q A P; P V Q and Q V P
c. PA (QVR)and(PA Q)V (PAR) d. P V (Q A R)and(P V Q) A (Q V R)
e. ^-(P A Q) and -P V --Q. f. ^-(P V Q) and ^-P A --Q
S. P4* Qand(P-r Q) A (Q=P)
Parts (c) and (d) above are referred to as the distributive laws for A and V, whereas parts (e)
and (f) are De Morgan's laws for A and V.
2. By means of a truth table show that (P V Q) A -- (P A Q) is true if either P or Q is true, and false otherwise.
3. Prove that each of the following is a tautology
a. P v -P (Excluded Middle)
b -(-P) q P (Double Negative)
c. P a [-P (R A --R)] (Contradiction)
d. (P A Q) - P (Conjunctive Simplification)
e. P (P V Q) (Disjunction)
f . [PA(P=* Q)] 'Q (ModusPonens)

g. [(P Q) A -- Q] ' -P (Modus Tollcns)
h. [(P= Q) A (Q =-,, R)] a (P R) (Transitivity)
4. For each of the following determine whether it is a tautology, a contradiction, or neither.
a.PA --P b.P''PA(PvQ)
c. (P A Q) A(-' PV - Q) d. -(PAQ)''(P '- Q)
e. IQ A (P=* Q)] 'P f. [P=* (Q A R)] '[(Q=R) V (R 'P))
g
S. In this exercise no knowledge about sequences is required. Let M, B, and C denote the following statements.
M: "The sequence is monotone."
B: "The sequence is bounded."
C: "The sequence converges:'
Express each of the following sentences symbolically, using the convention that "divergent' is the negation of
"convergent" and "unbounded" is the negation of "bounded" (In each of the statements the tern sequence refers to
a sequence of real numbers).
a. The sequence is monotone and bounded.
b. A convergent sequence is bounded.
c. A sequence converges, only if it is bounded.
d. In order that a sequence is bounded, it is necessary that it converges.
e. If the sequence diverges, then it is unbounded.
6. Provide an appropriate negation of each of the sentences in Exercise 5. Express your answer first in symbolic form
and then in English.
A.2J Rules of Inference

Given a propositional formula R, the method of truth tables provides a reliable means
to determine whether R is a tautology. This method can even be turned into a computer
program which would accept any formula R as an input and would determine whether
R is a tautology. However, checking formulas of even modest length, say twenty sym-
bols, turns out to consume inordinant amounts of time. What is worse, this situation
cannot be improved substantially by clever computer programming. It can be proven,
using the techniques of computational complexity-a branch of mathematical computer
science-that the problem of determining whether a propositional formula is a tautol-
ogy is'intractable in the sense that any program for its solution will place insurmount-
able demands on computational resources.
Where computational methods fail, mathematical ingenuity can still succeed. Con-
fronted with a propositional formula R it may be possible to offer a proof that it is a tau-
tology. Roughly speaking, a proof of R is a sequence of sentences or propositional for-
mulas, the last one being R, such that each of the sentences in the sequence is either an
axiom, a hypothesis, or a statement that follows from the previous sentences in the se-
quence by some principle of logical inference. To make this unambiguous. we have to

specify these principles of logical inference. A principle of logical inference may have
some premises (like the previous sentences referred to above) and a conclusion.
The Form of a Rule of Inference:

From P,,. . . , P, one can infer Q.
Symbolically this can be expressed as

P,
P1....,P. :. Q, or as
P.
Q,
Here the formulas P,, ... , P are the premises and Q is the conclusion. (The symbol
:. is used in mathematics to denote therefore.) We even allow the case when no
premises are present. In that case, Q can be regarded as an axiom. The most important
thing about a rule of inference is that it should be logically valid; that is,
P, A . . A P. = Q (or just Q if no premises are given)
should be a tautology. For example, to show that P, P Q :. Q is logically valid, it
suffices to show that (P A (P Q)) Q is a. tautology. This is easily verified by
means of a truth table as follows.
P Q P=* Q P A (P=Q)=:;- Q
T T T T T
T F F F T
F T T F T
F F T F T
Since it becomes impractical to always verify by means of a truth table whether

Pt A A P. Q is a tautology, we utilize tautologies to construct rules of inference.
We can take any tautology of the form above and convert it into a rile of inference. Of
course, we would gain nothing if we allowed ourselves to have a rule of inference for every
tautology. We still have to use truth tables to verify that each rule of inference comes from a
tautology. Fortunately, a handful of fairly simple tautologies is all that is needed. Each of the
following tautologies can be verified by the method of truth tables without too much effort
PV -P (Excluded Middle)
P PVQ (Disjunction)
[P A (P . Q)] Q (Modus Ponens)
[(P = Q) A --- Q) =! -- P (Modus Tollens)
P A Q . P (or Q) (Conjunctive Simplification)

[(P Q) A (Q R)} ' (P R) (Transitivity)
From these tautologies we obtain the rules of inference listed below. Some additional
useful and important tautologies are as follows.
--(--P) q P (Double Negative)
P V Q a Q V P, P n Q a Q n P (Commutative Laws)
Pv(QVR)a(PVQ)VR
(Associative Laws)
P A (Q A R) q (P A Q) A R
--(P n Q) Q
(De Morgan's Laws)
-(P v Q) q -P A Q
[P A (Q V R)] q [(P A Q) V (P A R)]
[Pv(QAR)]e*[(PVQ)A(PVR)] j
(Distributive Laws)
P=*Qa-Q=*--P (Contrapositive)
Pa[--P==O-(R A-R)] (Contradiction)
One of the most fundamental rules of inference, in fact usually taken as an axiom,
is the following.
Rule of the Excluded Middle:

One can infer P V P, for any statement P.
This rule simply states that the propositional formula P V --P can be inferred, for any
proposition P. For example, if an argument involves a real number x, one can assert at
any time that "Either x = 0 or x # 0."
Rule of Conjunction:
From P and Q, one can infer P A Q.
Rule of Disjunction:
From P, one can infer P v Q.
Modus Ponens Rule or Rule of Detachment:
From P and P Q, one can infer Q.
Modus Tollens Rule or Rule of Contrapositive Inference:
From -Q and P = Q, one can infer -P.
Rule of Conjunctive Simplification:

From P A Q, one can infer both P and Q.
Rule of Transitive Inference or Hypothetical Syllogism:
From P Q and Q . R, one can infer P . R.
The rule of conjunction, although not listed as a tautology, simply asserts that
P, Q :. P A Q is logically valid; that is, P A Q P A Q is a tautology. Likewise,
the rule of disjunction follows from the fact that P = P V Q is also a tautology. The
modus ponens rule is again a verbal statement of the modus ponens tautology
[P A (P ' Q)] Q. The implication P Q by itself, even if known to be true, infers
nothing about Q. The implication P Q is also true when both P and Q are false. How-
ever, if both P and P =*- Q are true, then Q must also be true.
The rule of conjunctive simplification is a restatement of the corresponding tautol-
ogy. However, it also follows as a consequence of the modus ponens rule. Since
(P A Q) P (or Q) is a tautology, by the rules of conjunction and modus ponens
(P A Q) A [(P A Q) P] P (or Q).
The modus tollens rule follows from the modus tollens tautology. It is also a simple
consequence of the contrapositive law and modus ponens. Since P Q is equivalent
to -Q -P, it is easily verified by a truth table that the formula -Q A (P Q) is
equivalent to -Q A (-Q -P). Thus by the modus ponens rule --P can be
inferred.
The above argument illustrates the use of the replacement rule. In the formula
-Q A (P = Q).
the expressions -Q and P =:;- Q are subformulas of the original formula. The new for-
mula -Q A (-Q =*- --P) was obtained from the original formula by replacing the sub-
formula P Q by its equivalent formula -Q = -P. The resulting formula is then
equivalent to the original. As a rule of inference, this is stated as follows.
Simple Replacement:
If P is a subformula of a formula P and P q Q', then from P one can infer
any formula Q that results from replacing an occurence of P in P with Q'.
There are several additional rules that are worthy of mention. The justification of
these rules is left to the exercises.
Disjunctive Syllogism:
From P V Q and --P, one can infer Q.
Rule of Inference by Cases:
From P Q and R Q. one can infer (P V R) Q.
Rules of Biconditional Inference:
From P Q and Q P, one can infer P q Q.
From P Q, one can infer P . Q and Q . P.
A more detailed discussion of the propositional proof system would have taken us
too far astray from our main goal; namely, to provide basic rules of inference with
which to construct valid arguments. A formal discussion of the propositional proof sys-
tem can be found in the text by Bums listed in the Supplemental Reading section. In
that text the author proves that every tautology, and only tautologies, can be derived
from the listed rules of inference.
T o p r o v e the validity of an inference PL, ... , P. .. Q. we simply have to verify
that using the rules of inference, we can infer the conclusion Q from the given premises
Pr, ... , P. Each statement in the proof should be either a premise or an axiom, or
should follow from previous statements by one of the accepted rules of inference. De-
vising a proof is a feat of mathematical ingenuity. However, once a purported proof is
in hand, it can be easily checked step by step for validity. We illustrate the use of the
rules of inference with the following examples.
All EXAMPLES
(a) As our first example we consider the following verbal argument.
If John is a Democrat, he associates with Democrats.
But John does not associate with Democrats.
Therefore, John is not a Democrat.
This argument can be written in symbolic form as
P=:0. Q
-P.
By the rule of contrapositive inference the argument is valid, whatever the truth or fal-
sity of the statements in it may be.
(b) As our second example we consider the following.
(a) P
(b) P = Q
(c) P=(QTR)
By the modus ponens rule, from (a) and (b) we can infer Q. Likewise from (a) and (c)
we can also infer Q R. But now from Q and Q R, we can infer R. Thus the infer-
ence is logically valid.
In symbolic form, the above proof can be written as follows:
1. P (premise)
2. P Q (premise)
3. P (Q R) (premise)
4. Q (modus ponens I & 2)
5. Q 'R (modus ponens l & 3)
.. R (modus ponens 4 & 5)
(c) For our final example we consider the following argument concerning Cauchy se-
quences. For the purposes of illustration we need to know nothing about sequences,
with the exception that they are usually denoted as and that divergence is the nega-
tion of convergence.
The sequence diverges.
A bounded sequence has a convergent subsequence.
Every Cauchy sequence that has a convergent subsequence converges.
Therefore, the sequence is not Cauchy.
In this particular example the conclusion happens to be true, but the argument is not
valid. To see this we will write the argument in symbolic form. Let P, Q, R, and S de-
note the following statements
P: "The sequence {p,} is Cauchy:'
Q: 'The sequence {p} converges:'
R: "The sequence is bounded:'
S: 'The sequence has a convergent subsequence."
In symbolic form the above argument is expressed as follows:
(a) - Q
(b) R=* S
(c) (P A S) Q
By (a) and (c) and the modus tollens rule we can infer -(P A S). But by De Morgan's
law, -(P A S) is equivalent to the statement -Pv - S. However, from -P V -- S we
can infer neither --P nor -S. If we knew that S (i.e., -(--5)) was true, then by the rule
of disjunctive syllogism we could infer -P. Unfortunately, from the given premises
nothing can be inferred about S. If S has the truth value F, then -P V -S is true re-
gardless of the truth value of P. This allows us to obtain an assignment of truth values
that make the premises true and the conclusion false. Thus the argument is not valid.
If the statement R had been included as a premise, then the resulting inference
would be valid. As is often the case in proofs written by the beginning student, state-
ment (b) of the proof is totally extraneous.
EXERCISES A.2
1. Construct a truth table for [(P' Q)A -.QJ -P to verify the validity of the argument in Example A.2.1(a).
2. a. Justify the rule of disjunctive syllogism.
b. Justify the rule of inference by cases.
3. Justify the rules of biconditional inference.
In Exercises 4-12 use the rules of inference or a truth table to test the validity of each of the following.
4. If L, is 11 to L2 and L2 is 11 to L3, then L, is 11 to L,.

L,istltoL2-
L2 is to L3.
L, is to L3-
5. If m is even, then 2 divides m.
If 2 divides m, then 4 divides m2.
If m is even and 4 divides m2, then m2 is even.
m2 is even.
6.
SQ
7. S
--P V -S
-.P
8. R
SaQ
(RVS)*Q
Q
9. P
PAR
RCS
(R A S)mo,Q
Q
10. P
R
PAS
(RAS) .Q
.. Q
11. P Q
R*S
P V R
QVS
12. The sequence diverges.

The sequence is bounded.
A bounded sequence has a convergent subsequence.
Every Cauchy sequence that has a convergent subsequence converges.
The sequence is not Cauchy.
A.3 Mathematical Proofs

In mathematics a proof is a logically valid deduction of a theorem from the premises
of the theorem, the axioms, or previously proved statements or theorems. The truth
of any statement in a proof should be traceable back to some initial set of axioms or
postulates that are assumed true. A proof should not be just a string of symbols.
Every step in a proof should express a complete sentence, including the justification
of the step.
In this section we look at several methods that are commonly used to prove a the-
orem. Most theorems in mathematics are stated in the form "If P, then Q"; that is.
P Q. Any theorem stated as a biconditional sentence "P, if and only if Q" is proved
by first proving P Q, and then Q ' P.
Direct Proof
The most straightforward type of proof of "P Q" is the direct proof; namely, we as-
sume the hypothesis P and use the axioms, computations, or other theorems and the
rules of logic to infer Q.
Direct Proof of P= Q
Proof. Assume P. Therefore Q. Thus P Q. t
We illustrate the method of direct proof and the use of the rules of inference in jus-
tifying the validity of an argument with the following examples.
A.3.1 EXAMPLES
(a) In Section 2.6 of the text the following theorem about Cauchy sequences is proved.
Theorem 2.6.4. Every Cauchy sequence of real numbers converges.
I In the text the symbol is used to mark the end of a proof. Some authors prefer to use QED. which is an
abbreviation of the Latin "quod eras demonstrandum." meaning "which was to be proved."
Let P and Q denote the following statements respectively.

P: 'The sequence is a Cauchy sequence of real numbers"
Q: 'The sequence converges."
The theorem to be proved is "If P, then Q"; that is, "If { is a Cauchy sequence of
real numbers, then the sequence converges:' Within Section 2.6 and in previous
sections the following related theorems are proved.
Thl "Every Cauchy sequence is bounded:'
Th2 "Every bounded sequence of real numbers has a convergent subsequence."
Th3 "If (p.} is a Cauchy sequence of real numbers that has a convergent
subsequence, then the sequence converges"
Let R and S denote the statements "The sequence is bounded" and "The sequence
has a convergent subsequence" respectively. Using P, Q, R, and 5, theorems Th l,
Th2, and Th3 can be written symbolically as
Thl P R,
Th2 RCS,
Th3 (PAS) Q.
From Th 1 and Th2, by the transitive rule we can infer P ' S. Thus from our assump-
tion P. by modus ponens we can infer S. Hence by the rule of conjunctive inference we
can infer P A S. But now by Th3 and the modus ponens rule we have
{(P A S) A [(P A S)*Q]}z*Q.
It is important to again emphasize that the fact that (P A S) Q is true does not allow
us to infer anything about Q. It is also required that P A S must be true.
In symbolic form, the above proof can be written as follows:
1. P (hypothesis)
2. PAR (Thl)
3. RCS (Th2)
4. P .S (transitive rule)
5. S (modus ponens)
6. PAS (conjunctive inference)
7. (PAS) = Q (Th3)
. Q (modus ponens)
The above provides a very,methodical argument illustrating the validity of the im-
plication P Q. For a short proof this works very nicely. However, there is not a sin-
gle proof in this text that is written in such detail using symbolic logic to proceed from
P to Q. Mathematical proofs should be written in complete sentences, including justifi-
cations. The truth of any statement in the proof must follow from the initial hypothesis,
the axioms, or previously proved theorems. A typical proof of the theorem, using the
facts introduced above, would proceed as follows. The comments in parenthesis are not
part of the proof; they are included as explanations of the statements.
Proof. Let be a Cauchy sequence of real numbers. (This asserts the truth of the hy-
pothesis P.) By Theorem 2.6.2 the sequence is bounded. (This asserts the implication
P R.) Thus by Corollary 2.4.1 the sequence 1p.1 has a convergent subsequence. (This is
the implication R'' S, which by the transitive rule gives P * S. In the next step we invoke
Theorem 2.6.3; namely, that P A S =:O- Q.) The result now follows by Theorem 2.6.3. Q
A better and more careful way, especially for the novice, to express the last sen-
tence would be as follows: "Since the Cauchy sequence has a convergent subse-
quence, by Theorem 2.6.3 the sequence converges."
(b) For our second example we prove the following statement.
"If m is an even integer, then m2 is divisible by 4:'
Let P be the statement "The integer in is an even integer," and Q the statement "The in-
teger m2 is divisible by 4." Thus we wish to prove that P Q.
Note: If m and n are integers, we say that m divides n, or n is divisible by in, if n = km
for some integer k.
Proof. (Assume P.) Suppose m is an even integer. Then m = 2k for some integer k.
(Here we use the definition of an even integer.) Then m2 = 4k2. Thus m2 is divisible by
4. (The conclusion Q.) U
Proof by Contraposition
Since the implication P Q is equivalent to its contrapositive -Q P, we can
prove the implication P Q by assuming --Q and deriving P. Such a proof is called
a contrapositive proof or proof by contraposition.
10,
Contrapositive Proof of P .Q
Proof. Assume -Q. Conclude - P by means of a direct proof. Thus

^-Q' --P. ThereforeP =:o-Q. Q
We illustrate the method of proof by contraposition with the following elementary

example.
A.3.2 EXAMPLE Let n be an integer. If n2 is even, then n is even.

Proof. Suppose n is not even Then n is odd. (Here, and below, we use the fact
that an integer is odd if and only if it can be expressed as 2k + I for some integer k.)
Thus n = 2k + 1 for some integer k. But then
n2=(2k+ 1)2=4k2+4k+ I =2(2k2+2k)+ 1.
(In the above we have used the rules concerning algebraic operations on the integers.
i.e., proved theorems.) Since m = 2k2 + 2k is an integer, we have
n2=2m+1
for some integer m. Thus n2 is odd (-P). U
Indirect Proof or Proof by Contradiction

Our next method of proof is the indirect proof or proof by contradiction. A proof by
contradiction makes use of the tautology
P .[--P=(RA-R)].
The statement RA - R is a contradiction. It is worthy of mention that R does not appear
anywhere on the left side of the tautology. Thus any proposition R that will do the job
will suffice. As opposed to our previous two methods of proof, proof by contradiction
can be applied to any statement "P." Direct proofs and contrapositive proofs only apply
to implications "P =;,- Q."
Proof of P by Contradiction
Proof. Assume -P. Therefore R. Therefore -R. Thus P is true.
A.3.3 EXAMPLE To illustrate the method of proof by contradiction we consider the fol-
lowing classical proof that is irrational. This proof is due to Pythagoras, and is the
first known proof using contradiction. The statement P to be proved is as follows:
P "V is irrational."
Proof. (We assume - P; that is, is not irrational, i.e., rational.) Suppose that
is rational. Then V2 = m/n, where m and n are integers, with m and n not both
even. (The sentence "m and n not both even" will be our statement R.) But then
m2 = 2n2.
Therefore m2 is even. Thus by Example A.3.2 the integer m is even. Since m is even, by
Example A.3.l(b) the integer m2 is divisible by 4. Since m2 = 2n2, 2n2 is divisible by 4.
Thus n2 is even, and by Example A.3.2 the integer n is also even. Thus m and n are both
even. (This is our statement -- R. The negation of "not both are even" is "both are
even.") This is a contradiction. Thus V is irrational. U
The method of proof by contradiction can also be used to prove the implication
P Q. Using the law of contradiction we have that
(P = Q) . [- (P =* Q) = (R A - R) ].
Thus to prove "P Q" by contradiction we must prove that -(P = Q) implies a con-
tradiction R A - R. Since P =:: Q a -P V Q, by De Morgan's law and double nega-
tion
-(P=* Q)q(PA-Q).
Thus in the proof of P Q by contradiction we must show that the assumption
PA ^ Q logically implies a contradiction RA R, for some appropriate statement R.
Proof of P= Q by Contradiction
Proof. Suppose PA Q. Therefore R. Therefore -R. Thus P Q
O
A.3.4 EXAMPLES
(a) To illustrate the method of proof by contradiction we will prove the following the-
orem about the integers.
Theorem. If a, b, c are integers satisfying a2 + b2 = c2, then a or b is an even
integer.
Let P denote the statement "a, b, c are integers satisfying a2 + b2 = c2" and Q denote
the statement "a or b is an even integer."
In the proof by contradiction we assume P A -- Q. Since a and b are assumed to be
integers, from the assumption PA -- Q we conclude that a and b are odd integers. Thus
we can write a = 2m + 1 and b = 2n + 1, where m and n are integers. But then
c2=a2+b2=(2m+ 1)2+(2n+ 1)2,
which upon simplification gives c2 = 4k + 2 for some integer k. Thus c2 is an even in-
teger. But then by Example A.3.2 the integer c itself is even. Hence c = 2p for some in-
teger p. This gives
c2=4p2=4k+2 or (p2-k)=1.
Since p and k are integers, p2 - k is an integer. (This is our statement R.) On the
other hand, 1 is not an integer. (This is our statement -R.) From our assumption
PA _ Q we derived a contradiction R A - R.
(b) For our second example we prove the theorem of Example A.3.2 by contradiction.
For an integer n, let P and Q denote "n2 is even" and "n is even" respectively.
To prove P Q we assume P and -Q; that is, n2 is even, and n is odd. Since n is
odd, n = 2k + I for some integer k. Therefore,
n2 = (2k + 1)2 = 2(2k2 + 2k) + 1.
Therefore n2 is odd (-P). This is a contradiction (PA P).
In the previous example we assumed PA - Q and showed that -Q leads to -P.

Thus P itself plays the role of R in the method of proof by contradiction. This is cer-
tainly permissible. However, if one assumes PA Q and derives -P without using P.
then the proof is in fact a proof by contraposition, rather than a proof by contradiction.
A WORD OF CAUTION! It has been my experience that students have a tendency to
overuse-and even misuse-proofs by contradiction. Quite often, in proving P Q,
the student will assume PA -Q and derive Q thereby obtaining the contradiction Q and
-Q. Inevitably, if the proof is correct, buried in the details is a direct proof of P Q.
Similarly, if the assumption PA -- Q leads to -P then the proof is in all likelihood a
proof by contraposition. This was the case in part (b) of the previous example.
Another problem with indirect proofs or proofs by contraposition is that they in-
volve the negation of statements. This is not always easy in analysis, especially if the
statements themselves are complicated and involve one or more quantifiers. (See Ex-
amples A.4.4 in the next section.) Before attempting an indirect proof or even a proof
by contraposition, the student is advised to first attempt to find a direct proof. A direct
proof often has the advantage of being more constructive. If the statement involves the
existence of a certain object, a direct proof may in fact provide a method for construct-
ing the given object.
Proof by Cases
The method of proof by cases is based on the rule of inference by cases. Thus to prove
that (P V Q) R it suffices to prove that P R and that Q - R. We illustrate this
with the following examples.
A.3.5 EXAMPLES
(a) For our first example we prove the following: "If n is a positive integer, then
n2 + n + l is odd:'
If n is a positive integer, then n can be either even or odd. Thus if P and Q repre-
sent the statements "n is an even positive integer" and "n is an odd positive integer" re-
spectively, then the statement we wish to prove is (P V Q) R, where R represents the
statement "n2 + n + 1 is odd:' By the rule of inference by cases it suffices to prove
P R and Q =* R. The details are left to the exercises (Exercise 6).
'(b) For our second illustration of the method of proof by cases we consider the fol-
lowing theorem.
Theorem. There are two irrational numbers a and b such that ab is rational.
Proof. First case: is rational. If this is the case we take a = b = V. Second
case: N/r2`r' is irrational. In this case we take a = \/'2-"2 and b = V2-. So
2(`/-2)'=
ab=(V "i)V2 _ 2=2,
which is rational. Therefore, there must be irrational numbers a and b so that ab is
rational.
In the proof, we let R denote the statement of the theorem and P the statement " V-2 % 2
is rational." In Case 1 we have P R, and in Case 2 we have -P = R. Thus by the rule
of inference by cases we have (P V -- P) R. But P V -- P is true regardless of P. Hence
R follows by the modus ponens rule. This example not only illustrates the method of proof
by cases, it also illustrates how P V -- P can always be asserted in a proof.
Counterexamples
Some conjectures in mathematics are simply statements that are either true or false. For
example, "V is irrational:' Other conjectures are general in that they assert something
about a whole class of objects. For example,
"Every Cauchy sequence of real numbers converges:'
If n is an even integer, then n2 is even."
The first makes an assertion about all Cauchy sequences, whereas the second makes an
assertion about all even integers. Both of these statements are true. If we are presented
with a statement about a class of objects, then such a statement is true if and only if it is
true for every object in the class. Thus to conclude that such a statement is false it suf-
fices to exhibit one object in the class for which the statement is not true. Such an object
is called a counterexample. To illustrate this we consider the following conjecture:
"If n is a positive integer, then n2 - n + 5 is prime."
With a little bit of thought, most students will immediately conclude that this conjec-
ture is false. If we check for n = 1, 2, 3, 4, then n2 - n + 5 becomes 5, 7, 11, and 17,
which are indeed all prime. However, when n = 5, n2 - n + 5 is equal to 25, which is
certainly not prime. Since we have exhibited an object (n = 5) for which the hypothe-
sis P is true but the conclusion Q is false, the implication P Q is false.
The previous example was very elementary, and to find a counterexample was not
very difficult. This, however, is not always the case. As an example, consider the fol-
lowing conjecture.
"Every continuous function on a closed and bounded interval is differentiable
except perhaps at a finite number of points:'
Most students who have completed a basic calculus sequence might be inclined to be-
lieve that this conjecture is true. Certainly, on the basis of most examples encountered
in calculus such a conjecture seems reasonable. In fact, many mathematicians through
the mid-nineteenth century accepted this, or a variation of it, as true. It was not until
1874, when Weierstrass constructed an example of a continuous function that was
nowhere differentiable (see Section 8.5), that the above conjecture was proved false.
Helpful Hints
The most common complaint heard from students when asked to prove a theorem is "I
don't know where to start" Unfortunately, there are no easy rules that can be used to
tell someone how to prove a theorem. Many of the problems and theorems in the exer-
cises follow from the definitions or from previous theorems. Some, however, require in-
sight and creativity, and for these the students must devise their own arguments. Some
helpful hints in constructing proofs follow.
(1) Make a list of all hypotheses and of what you want to prove. Do not ignore any
of the hypotheses. As a general rule they are all required. If you have not used all the
hypotheses then most likely your proof is incorrect.
(2) Refresh your memory with the pertinent definitions. If necessary, write them
out. This will help you to memorize them and also to understand them. It is very im-
portant that you know and understand all the definitions. They are the foundations upon
which the theory is built.
(3) Search for theorems that have similar hypotheses or similar conclusions.
Proved theorems are not just results; they are also tools that enable you to develop the
theory further. Suppose you are given P and R as hypotheses and are asked to prove Q.
If you have a theorem that P S and you can prove that (R A S) Q, then you have
your desired proof. An alternative approach is to work backward; that is. to find a the-
orem that has the same conclusion Q with given hypothesis S, and see if you can prove
S from the given hypotheses P and R.
(4) Learn the statements and the proofs of theorems. When I teach real analysis. I
always require students to memorize the statements of all the theorems and the proofs
of selected theorems. Contrary to student beliefs, this is not done to torture them. The
statements of the theorems are the tools used in proving other theorems; the proofs pro-
vide useful techniques that may be used elsewhere. They also provide a good model of
how to write a correct proof.
Some common errors committed by students include the following.
(1) Using theorems for which all the hypotheses are not satisfied.
(2) Making extra assumptions beyond those given in the statement of the problem
or theorem.
(3) When asked to prove something, for example all continuous functions f, the the-
orem is proved for a particular function such as f (x) = x-. Even though this is incorrect,
by attempting to prove the theorem for a special case, the student may in fact gain insight
on how to prove the result for the general case.
R EXERCISES A.3
1. As in Example A.3.l(a), write the proof of Example A.3.I(b) in symbolic form.
2. Same as the previous exercise for Example A.3.2.
3. Construct proofs of each of the following statements about the positive integers. Suppose k. m, and n are positive
integers.
a. If m and n are odd, then mn is odd. b. If m is odd, then m, is odd.
c. If m2 is odd, then m is odd. d. If k divides m and m divides n, then k divides n.
e. If n is odd, then n2 + 1 is even.
4. Consider the following statement. "If n is a positive integer, then > Prove the statement by pros iding
a. a direct proof, b. a proof by contradiction.
5. Prove that is irrational by contradiction.
6. Complete the details of the proof in Example A.3.5.
7. Prove that the equation a2 = 4b + 3 has no integer solutions.
S. Provide a counterexample to each of the following statements.
a. If n is a positive integer, then n2 - n + 41 is prime.
b. If n is a positive integer, then n! < 2'.
c. If n is a positive integer, then n' < (n + I)".
d. Every continuous function is differentiable.
_AAJ Use of Quantifiers2

We have already touched on the notion of quantifiers while discussing counterexamples
in the previous section. There we discussed the difference between sentences that are
simply statements that are either true or false, and sentences that make assertions about
a collection of objects. In this section we make the latter more precise.
Quantified Sentences
The sentence "x'- = 4" is not a proposition as it is neither true nor false. If we replace
x by specific values then the statement "x 2 = 4" becomes a proposition. For example,
the sentence is true for x = 2 and false for x = 3. Likewise, the sentence "x is a ratio-
nal number" is neither true nor false until x is replaced by a specific quantity. In the
sentences "x 2 = 4" and "x is a rational number," the 'x' is called a variable and the sen-
tences themselves are called formulas or open sentences in the variable x. Specifically,
a formula (in logic) is a statement containing one or more variables which becomes a
proposition when the variables are replaced by particular objects. W e will use the nota-
tion P(x) to denote that P is a formula in the variable x. Likewise, a formula in the vari-
ables x 1 ,. xk will be denoted by P(xt, .. . , xk). For example, "xi = x2 + x3' is a
formula in the variables xi, x2, x3.
Before the truth of a formula P(x) can be determined we must specify what objects
are available for discussion. This is called the universe U for P(x). For example. for the
formula x2 = 4 an appropriate choice for the universe U may be either the set of posi-
tive integers N, the set of all integers 71, or even the set of real numbers R. It is not
enough, however, to just specify the universe: For example, in the formula x - y = x.
a meaning must also be given to " " and If we are discussing 2 X 2 matrices.
then in addition to specifying the universe C' as the set of 2 X 2 matrices, we must also
2 Since this section requires some basic knowledge of the terminology of sets it is best postponed until Sec-
tion 1.1 has been read.
define matrix multiplication and equality of matrices. For logical considerations, we

must also require that the universe be nonempty.
For a formula P(x) with specified universe U, the truth set of P(x) is the set of all
x E U such that P(x) is a true proposition. In the notation of sets, the truth set is simply
{x E U: P(x)}. This is read as "the set of x in U such that P(x)." For example,
{x E N : x2 = 4} = {2}, whereas {x E TL : x2 = 4} = {-2, 2}.
Consider the two formulas P(x) and Q(x) given by "x2 = 4" and "(x + 1)2 =
x2 + 2x + 1" respectively. If we take as our universe the set of real numbers R, the
truth set for P(x) is nonempty. This is expressed by saying that there exists an x E 1R
such that P(x) is true. On the other hand, the truth set for Q(x) is all of R, and this is ex-
pressed by saying that Q(x) is true for all x E R. We make this precise with the fol-
lowing definition.
AA.1 DEFINITION Suppose P(x) is a formula in the variable x with universe U.

(1) The sentence (Vx) P(x) is read 'for all x, P(x)" and is true precisely when
{x E U : P(x)} = U. The symbol V is called the universal quantifier.
(ii) The sentence (3x) P(x) is read "there exists x such that P(x)" and is true
precisely when {x E U : P(x)} # 4. The symbol 3 is called the existential
quantifier.3
The expressions (Vx)P(x) and (3x)P(x) are called quantified sentences. The
phrase "for every" is synonymous with "for all" If we wish to emphasize the universe
U we write (Vx E U)P(x) and (3x E U)P(x). These are read as "for all x in U. P(x)-
and "there exists x in U, such that P(x)" respectively. For example, with the formulas
(x+ 1)2=x2+2x+landx2=4wehave
(`dxE R)[(x+ 1)2 = x2+ 2x+ I],
(3xCt8)(x2=4).
Most mathematicians avoid using V and 3 in publications. In fact, with the excep-
tion of this section, they are used nowhere else in the text. Expressing mathematical
statements in quantified form is first of all not always easy, and second makes for awk-
ward reading. Quantifiers, however, are crucial when it comes to negating complicated
statements.
AA.2 EXAMPLES
(a) As our first example we express the following theorem of the text as a quantified
sentence.
Theorem 2.6.4. Every Cauchy sequence of real numbers converges.
Clearly, the quantifier to be used is V. For our universe we take U to be the set of
all sequences of real numbers, and let C(x) and Q(x) denote the open sentences "x
is a Cauchy sequence" and "x converges" respectively. Consider first the sentence
3 The symbol 3! is often used to denote the existence of a unique x for which P(x) is true.
(b'x)[C(x) A Q(x)]. This sentence would be translated as follows: "For all sequences
x, x is a Cauchy sequence and x converges:' This sentence, however, is the same as
"every sequence is a convergent Cauchy sequence:' and this is clearly not the intent of
the original statement.
In other texts, Theorem 2.6.4 is sometimes expressed as "if is a Cauchy se-
quence of real numbers, then {an} converges" This statement can also be rewritten as
"For all sequences {an} of real numbers, if {an} is a Cauchy sequence, then {an} con-
verges:' This version, although somewhat awkward grammatically, is now easily writ-
ten in symbolic form as (Vx E U)[C(x) ' Q(x)l. In general, a sentence of the form
"All P(x) are Q(x)" is expressed in symbolic form as (Vx)[P(x) Q(x)].
(b) As another example consider the statement
"Some bounded sequences converge."
As in (a) let Q(x) be the sentence "x converges" and B(x) be the sentence "x is a
bounded sequence." Since "some" is taken to mean at least one, the proper quantifier is
3. However, should the statement be expressed symbolically as (3x E U)[B(x) Q(x)]
or as (3x E U)[B(x) A Q(x)]? The first would be interpreted as "There exists a se-
quence x, such that if x is bounded, then x converges." This clearly is not the intent of
the sentence. It does not ensure the existence of a bounded sequence that converges. The
second, (3x)[B(x) A Q(x)l, reads "There exists a sequence x, such that x is bounded
and x converges," and this is the correct interpretation. In general, the statement "Some
P(x) are Q(x)" is expressed symbolically as (3x)[P(x) A Q(x)].
(c) Most expressions in mathematics require the use of many quantifiers. To illustrate
this, consider the definition of convergence of a sequence of real numbers as given in
Section 2.1 of the text.
DefWtion 2.1.7. A sequence {pn} in Il8 is said to converge if there exists a point
p E R such that for every e > 0, there exists a positive integer n such that
Ipn - pI < efor all n?no.
This definition uses the quantifiers V and 3, not only once, but several times. We have
"3p E R," "3no E N," "(Ve)(e > 0)," and "(Vn)(n Z n.)." To write this statement in
symbolic form, we begin with (3p E R)[' ]. Consider the sentence "for every e > 0,
there exists ..:' This phrase, properly stated, should be expressed as "for all e E 13, if
e > 0, then ...,"which in symbolic form would be written as (Ve)[(e > 0) . (. )].
This leaves us with the final phrase "there exists a positive integer n, such that
Ipn - pI < e for all n n,,. This would be written as (3no E N) ("Ipn - p] < e for
all n ? no"). The statement "Ipn - pi < e for all n ? no," again if properly stated,
would read as "for all n E N, if n z n,,, then Ipn - pI < e," or in symbolic form,
(`tin E N)((n a I p, - p I < e). Combining all of the above finally gives "The se-
quence { pn} in R is said to converge" if
(3p e R)[(`de){(e > 0) ' [(3no E NI){(Vn E N)((n a no) Ipn - pI < e)}]}].
4 In mathematics the term "some" is taken to mean at least one. This differs from the colloquial interpreta-
tion where "some' is occasionally taken to mean two or more.
This can also be expressed as

(3p)(Ve)[e > 0 n = tp - pl < e)],
where the universe for each of the quantifiers is understood.
Negation of Quantified Sentences

The next thing we want to consider is how to negate a quantified statement. First, how-
ever, we need to define what it means for two quantified sentences to be equivalent. Recall
that two propositional formulas P and Q are equivalent if and only if P a Q is a tau-
tology. Suppose P(x) and Q(x) are two formulas with nonempty universe U. Then P(x)
and Q(x) are equivalent in U if and only if P(x) and Q(x) have the same truth value for
all x E U; that is, (Vx E U)(P(x) #' Q(x)). Two quantified sentences P(x) and Q(x) are
equivalent if and only if they are equivalent in every universe. For example. since
P Q is equivalent to P V Q, if P(x) and Q(x) are formulas or open sentences in x
with universe U, then
(Vx E U)[(P(x) Q(x)) a (-- P(x) V Q(x))).
Thus the quantified sentences (Vx)(P(x) Q(x)) and (Vx)(- P(x) V Q(x)) are equiv-
alent.
Consider now the quantified sentences (Vx)P(x) and (3x)P(x). The sentence
-- (Vx)P(x) is true in a given universe U if and only if (Vx)P(x) is false: that is, if and
only if the truth set {x E U: P(x)} is not equal to U. But this is true if and only if
{x E U : -- P(x)} is nonempty,5 i.e., (3x) P(x). Since this argument holds for any
universe U, the quantified statement (Vx)P(x) is equivalent to (3x) -- P(x). A similar
argument also proves that (3x)P(x) is equivalent to (Vx) - P(x). To emphasize these
we state them as a theorem.
A.4.3 THEOREM Suppose P(x) is a formula in the variable x. Then

(a) -(Vx)P(x) is equivalent to (3x) -P(x).
(b) -(3x)P(x) is equivalent to (Vx) - P(x).
In the following examples we find the negation of each of the quantified sentences of
Examples A.4.2.
A.4.4 EXAMPLES
(a) As in Example A.4.2(a) we consider the statement "Every Cauchy sequence con-
verges.' Using the same notation, in symbolic form this sentence was expressed as
(Vx)[C(x) .Q(x)]. The negation of the statement becomes (3x) - [C(x) Q(x)J.
Now ( C(x) Q(x)] is equivalent to (- C(x) V Q(x)), and by De Morgan's law the
latter is equivalent to C(x) A -- Q(x). Thus the negation of (Vx E U)[C(x) Q(x)] is
(3x E U)[C(x) A -- Q(x)]. This last statement would be read as "Mere exists a Cauchy
5 Here it is required that the universe U itself is not empty.

sequence in R that diverges." (Note: This .statement, however, is false in the real num-
ber system R, but true in the rational number system 0.)
(b) For our next example consider the negation of the statement "Some bounded se-
quences converge" In symbolic form this was expressed as (3x)[ B(x) A Q(x)] (see Ex-
ample A.4.2(b)). The negation of this statement becomes (Vx) -- [B(x) A Q(x)j. which
by De Morgan's law is equivalent to (Vx)[- B(x) V Q(x)]. But - B(x) V -- Q(x) is
equivalent to B(x) -- Q(x). Hence the negation of (3x)[B(x) A Q(x)] is equivalent to
(dx)[B(x) - Q(x)]. This last statement would read as "For all sequences of real
numbers, if is bounded, then diverges," or more simply as 'All bounded se-
quences in R diverge:'
(c) As our final example we undertake the negation of the definition of convergence
of a sequence of real numbers. In symbolic form, a sequence {pn} in R is said to con-
verge if
(3p (=- l)[(`de){(e > 0)' [(3n E N){(`dn E N)((n >_ Ip,, - pi <
We proceed to negate this sentence step by step. First, the negation of (3p E R)[ . }
becomes (Vp E 111) -- [ ]. Now inside the bracket [. ] we have (`de){P(e)w
Q(e, no, n)}, where P(e) denotes "e > 0" and Q(e, no, n) denotes "(3n0 E N)
{(V n E N)((n ? nn) a 1p. pi < e)}." But (Ve){P(e) Q(e, n,,, n)} is equiva-
lent to (3e){P(e) A -- Q(e, no, n)}. It is left as an exercise (Exercise 7) to show that the
negation of Q(e, no, n) becomes
(Vn,, E N){(3n E N)[(n ? n,) A (IPn - pi e)]}.
Combining all the above gives us the following. "A sequence in R is said to di-
verge" if
(Vp E R)(3e){(e > 0) A (Vn E N)[(3n E N)(n =' no) A (Ip - pi ? e)]},
or
(Vp E lR)(3e > 0)(Vn,, E N){(3n ? n,)IPp - pI ' e).

Translating everything into English gives À sequence in 18 is said to diverge if for
all p E R, there exists e > 0 such that for all n, E N, there exists n ? n,, such that
IPn - PI e.
0 EXERCISES A.4
1. Express each of the following sentences in symbolic form. Specify an appropriate universe for each.
a. All men are mortal.
b. Not all mortals are men.
C. Some isoceles triangles are equilateral triangles.
d. Some triangles are isoceles triangles and some are equilateral triangles.
e. Some triangles are isoceles and equilateral triangles.
f. Not all isoceles triangles are equilateral triangles.

g. Between any two distinct real numbers there is a rational number.
2. In the following, let the universe U be the set of all sequences of real numbers, and let M(x), B(x), and C(x) denote
the sentences "x is monotone, "x is bounded,' and "x converges" respectively. Express each of the following state-
ments in symbolic form using the quantifiers 3 and V.
a. All convergent sequences are bounded.
b. There exists an unbounded monotone sequence.
C. Some monotone sequences are unbounded.
d. Every bounded monotone sequence converges.
e. Not all bounded sequences converge.
L Every divergent sequence is unbounded.
3. Determine which of the following quantified sentences are equivalent.
a. (Vx)[P(x) A Q(x)] and [(Vx)P(x) A (Vx)Q(x)]
b. (Vx)[P(x) V Q(x)] and [(Vx)P(x) V (Vx)Q(x)]
c. (Vx)[P(x) Q(x)) and (Vx)P(x) (Vx)Q(x)
d. (Vx)(Vy)P(x. y) and (Vy)(Vx)P(x, y)
e. (3x)[(3Y)P(x, y)] and (3Y)[(3x)P(x, y)]
1. (3x)[(VY)P(x, y)] and (Vy)[(3x)P(x, y)]
4. Find the negation of each of the following quantified sentences.
a. (Vx)[P(x) V Q(x)]
b. (Vx)[P(x) V -- Q(x)]
e. (3x)[P(x)' . Q(x)]
d. (3x)[(P(x)* Q(x)) A (R(x)=* Q(x))]
e. (Vx)[{P(x) A (P(x) z* Q(x))} ' Q(x)]
f. (Vx)[(3Y)(P(x, y) ' Q(x, Y))]

5. Find the negation of each of the following quantified sentences.
a. (Vx E R)(2x > x)
b. (Vx E R)[(x > 0) .(2x > x))
C. (3x IE NX5x + 11 = 3x + 14)
d. (Vx E R)(3yE l8)(x + y = 0)
e. (Ve E R)[e > 0 ' (3nE I@!)(1, < e)]
6. Find the negation of each of the statements in Exercise 2. Express your answer first in symbolic form and then in
English.
7. Show that the negation of (3n,)((Vn)((n ? n,) p, - pI < e)) is (Vn,)(3n)((n 3 n,) A (Ip, - pI a e)].
8. In the following, f is a real-valued function defined on an open interval (a, b) and p is a point in (a, b). The defini-
tion of the limit of the function f at p is as follows: "The function f has a limit at p if there exists a number LE R
such that given any e > 0, there exists a 8 > 0 for which [f(x) - LI < e for all xE (a, b) with
0<]x-pI <S."
a. Express the definition of the limit of a function in symbolic form.
b. Find the negation of the definition of the limit of a function.
Burris, S. N., Logic for Mathematics and Computer Smith, D., Eggen, M., and St. Andre, R.. A Transition
Science, Prentice Hall, New Jersey, 1998. to Advanced Mathematics, Brooks/Cole. Pacific
Lemmon, E. J., Beginning Logic, Chapman & Grove, CA, 1997.
Hall/CRC, Boca Raton, Florida, 1998.
Bibliography
Berg, P. W. & McGregor, J. L., Elementary Partial Differential Equations, Holden-Day.

Oakland, CA, 1966.
Hewitt, E. & Stromberg, K., Real and Abstract Analysis, Springer-Verlag, New York, 1965.
Katz, Victor J., A History of Mathematics, 2d ed., Addison-Wesley, Reading, MA, 1998.
Natanson, I. P., Theory of Functions of a Real Variable, vol. 1, Frederick Ungar Publ. Co., New
York. 1964.
Rudin, W., Principles of Mathematical Analysis, McGraw-Hill, Inc., New York, 1976.
Titchmarch, E. C., The Theory of Functions, Oxford University Press, 1939.
Weinberger, H. F., Partial Differential Equations, John Wiley & Sons, New York. 1965.
Zygmund, A., Trigonometric Series, vol. I & 11, Cambridge University Press, London. 1968.
522
Hints and Solutions
to Selected Exercises
S CHAPTER 1
EXERCISES 1.1 page 5
2.(a)AfB=B,Af17L={-1,0,1,2,3,4,5},Bf1C={x:2:5x!53} (b)AXB={(x,y):-l<x<_5,
O!5 y<_3} 4. (a) Suppose xEAfl(Bf1C).ThenxEAandxEBflC.SincexEBf1C. xEBandxEC.
Thus x E A fl B and x E C. Therefore, x E (A f7 B) fl C. This proves An (an c) c (A fl B) fl C. The reverse
containment is proved similarly. 7. (a) Let x E A U (B fl C). Then x E A or x E s fl C. If x E A. then x E A U B
x E A U C, i.e., x E (A U B) fl (A U C). Thus A U (B n C C (A U B) fl (A U C). The reverse containment is

proved similarly. 8. If A = { 1, 2, 3} then .(A) = (d,. 11). {2}, (3), (1, 2), {1, 3), (2, 3), Al.

2. (a) No! The ordered pairs (0, 1) and (0, - I) are both elements of A. This, however, contradicts the definition of
function. 3. (a) f((1, 2, 3, 4)) = { 1. 3, 5, 71 and f '({1, 2, 3, 4}) = 1, 2). 4. (a) f(A) = (y : 0:5 Y S 9);
f-'(A) = {x : x' + I E A) = {x:-'5' s x s 1 } (c) f-'(x) = j x - 1, x r= R. 6. (a) Range f - R, f is one-
to-one, and x = f-'(y) = 3(y + 2). (c) Range f = {y : I s y < oo}, f is one-to-one, and x = (i - I)/y.
(g) Range f = l8, f is not one-to-one: If y, # y2. then (x, y,) # (x. y2) for any x, yet f(x, y,) = f(x, y2).
7. (a) Range f = {(x, y) E R X R : x2 + y2 = I } (b) f -'((1, 0)) = 0, f-'((0, -1)) = f 11. Assume f is not
one-to-one and show that this leads to a contradiction.
EXERCISES 13 page 19
1. (b) For n = 1, 1 = l2. Assume the result is true for n = k. Then for n = k + 1,
I +3+ . +(2k- 1)+(2(k+ 1) - 1)=k2+(2k+ 1)=(k+ 1)2. (f) When
n = 1, .r' - y2 = (x - y)(x + y) and equality holds. For n = k + 1, write
xk+2 xk-2 - Xyk-' + XY`-' - Yk+ = X(xk'' - k-')+ (x - y)yk+', and now apply the induction hypothesis.
2. (a) The result is true for n = 1. Assume that fork E N. 2k > k. Then by the induction hypothesis,
2k-' = 2k 2 > k 2 = k + k ? k + 1. (c) For n = 4,4! = 24 > 16 = 24. Thus the inequality is true when
n = 4. Assume that n! > 2" for some n z 4. Then (n + I)! = (n + I )n! > (n + 1)2" > 2 2" = 2"-'. Thus by the
modified principle of mathematical induction the inequality holds for all n E N. n ? 4. 4. For n E N let P(n) be
the statement f(n) = 3 2" + (-1)". Then P(n) is true for n = 1, 2. Fork ? 3, assume that P(j) is true for all
523
524 Hints and Solutions to Selected Exercises
j E N1, j < k. Use the fact that f(k) = 2f(k - 2) + f(k - 1) and the induction hypothesis to show that P(k) is true.
Thus by the second principle of mathematical induction, the result holds for all n E N. 5. (b) f(n) = n2
(d) f(n) = 0 if n is even, and f(n) _ (- 1)t"-'n/n! if n is odd. 7. For each n E N let S" = r + r2 + + r".
Then S. - rS" = r - r"+', from which the result follows. 8. Hint: Let A = '(a, + + a") and write a" xA
for some x a 0. Use the induction hypothesis to prove that (a, a" '') s x'1t"+')A.
Now use Bernoulli's inequality to prove that xj'j"'') s (n + x)/(n + 1). From this it now follows that
xii("+')A 5 n + x A =
n + 1
1
n + 1
(a,+.+a"+a"+,).
EXERCISES 1A page 26
4. Consider (a - b)2. S. (a) inf A = 0, sup A = I (e) inf C = -oo, sup C = 00 (e) inf E = 1, sup E = 3
(g) inf C = 0. sup C = no (i) inf I = - 2, sup ! = 2 14. (b) Since A and B are nonempty and bounded above,
a = supA and is = sup B both exist in R. Since a = supA we have a s a for all a E A. Similarly b :5 0 for all
b E B. Therefore, a + b 5 a + $ for all a E A, b E B. Thus a + /3 is an upper bound for A + B, and thus
y=sup(A+B)5a+J3.
To prove the reverse inequality, we first note that since y is an upper bound for A + B, a + b s y forall a E A.
b r= B. Let b E B be arbitrary, but fixed. Then a c y - b for all a E A. Thus y - b is an upper bound for A and
hence a <_ y - b. Since this holds for all b E B, we also have that b s y - a for all b E B. Thus S s a; i.e.,
a + / <- y. An alternative way to prove that y = a + /3 is by contradiction. Suppose y < a + P. Set
e = a + /3 - y. By assumption e > 0. Now consider a - Ie and 0 - ?e, and use Theorem 1.4.4to obtain a
contradiction. 15. (a) Let a = sup{f(x) : x E X}, )3 = sup{g(x) : x E X}. Since the range off and g am bounded,
a and P are finite with f(x) + g(x) a + 6 for every x E X. Therefore, a + 0 is an upper bound for
{f(x) + g(x) : x E X}. Thus sup{f(x) + g(x) : x E X} s a + /3. 16. (a) F(x) = 3x + 2. sup{F(x):x E [0, 1)} = 5
20. (a) F(x) = 3x + 2 and H(y) = 2y. Thus sup{H(y) : y E [0, 111 = 2 and inf{F(x) : x E [0, 1 ]} = 2.
EXERCISES 15 page 30
1. First prove that if p and q are positive integers, then there exists n E N such that np > q. If p > q then n = 1
works. If p s q, consider (q + I )p. 6. (a) Use the fact V2/2 is irrational. (b) Use Theorem 1.5.2 and (a).

1..0202020 2. (a) .0022 = + 51 + i + = 27 + 81 =
00 1
(d) .101010 _ +;+ +5 +,+ _i ar=13 1 1 ,] _- 3. .0101010

9

2. Let f : 101-s 0 be defined by f (n) = 2n - 1. 4. (a) g(x) = a + x(b - a) is a one-to-one mapping of (0, 1) quo
(a, b). 6. (a) Since A - X. there exists a one-to-one function It from A onto X. Similarly, there exists a one-to-one
function g from B onto Y. To prove the result, show that F : A X B --X x Y defined by F(a, b) = (h(a), g(b)) is
one-to-one and onto. 8. (a) U A" = R, f1 A. = {x : - I < x < 1) (c) U A. _ (-1, 2), fl A. _ [0, 1
18. (a) Consider the function on (0, 1) that for each n E N, n ? 2, maps to "=_ ,, and is the identity mapping
elsewhere. 19. For a polynomial p(x) = ax" + + a,x + ao, consider the height h of the polynomial defined by
h=n+Iaol+lat[+...+Ia"I
Prove that there are only a finite number of polynomials with integer coefficients of a given height It, and therefore
only a finite number of algebraic numbers arising from polynomials of a given height It.
22. If f is a function from A -+ '(A), show that f is not onto by considering the set {x E A : x 6E f(x)}.
23. For a, b E [0,1 ] with decimal expansion a = a, a2 and b = b, b2 , consider the function
f : [0 1) x [0, 1 ) - [0, 1) by f(a, b) = a, b, a2b2 ... .
CHAPTER 2
2. Ix I = Ix - y + y 1 5 Ix - y I + I y 1. Therefore, I x I- I y 15 I x - y 1. Interchanging x and y gives I y i- Ix 1 5
I y- x I= I x- y I Now use the definition of II x I- I y I I. 6. (a) -3:5,x:5 13/3 (c) -1 <x < 2 & (a) Let
a" = (3n + 5)/(2n + 7). Then I a. - 21 = 11/2(2n + 7) < 3/n. Given e > 0, choose n E N such that n, a 3/e.
Then for all n a n I a. - 2 I < e. Therefore, lim a" = 2(c) Set a" = (n2 + l )/2n2. Then I a. - III = < -.
Given e > 0, choose n, E N such that no a 1/2e. Then for all n a n", I a. - 121 < e.Thus lim a" = 2.
9. (a) If it is even, then n(1 + (-1)") = 2n. Thus the sequence is unbounded and diverges in R.
(c) When n is even, i.e., n = 2k, then sin i = sinker = 0. On the other hand, when it is odd, i.e., n = 2k + 1, then
sin If = sin(2k + 1) = (-1)k. Thus sin 9 assumes the values -1, 0, and 1 for infinitely many values of n. Therefore
the sequence diverges. 11. (a) Write b = 1 + a where a > 0. By Example 1.3.3(b), b" ? I + no. Now use the
previous problem. 12. First show that I a2. - a21 5 (I a" I + I a I) I a" - a I. Now use the fact that since {a"}
converges, there exists a positive constant M such that I a" 15 M for all n e N. 14. Consider a = 0 and a > 0
separately. For a > 0, V - V = (a" - a)/(V + Va-).
S. (b) If p > 1, let x" = -Vp - 1. Apply the inequality of Example 1.3.3(b) to (I + x")". 6. (a) Is (c) -1 (e) 2
7. (a) Converges to 1. (c) Since
1 n2 + 1 2 (1 + 1/n2)2
n2 \2n + 3 (2 + 3/n)2'
by use of the limit theorems the sequence converges to 4.
8. Use the fact that I cos x 1 5 1 for all x E R. 10. (a) Suppose lim a"+ ,/a" = L < 1. Choose e > 0 such that
L + e < 1. For this a there exists n. E 101 such that a"+ ,/a" < L + e for all n a n,. From this one obtains that for
i > no, 0 < a" <_ (L + M(L + e)", where M = a",(L + Since (L + e) < 1, by Theorem 2.2.6(e).
slim (L + e)" = 0. The result now follows by Theorem 2.2.4. 11. (a) With a" = n2a", 0 < a <
L = Iimoa". i/a" = a lim (I + n)2 = a. Thus since L = a < 1, the sequence converges. 12. Set
x" (a" - 1)/(a" + 1), and solve for a". 13. (b) Verify the result for it = 1. Assume the result holds for it = k. For
n = k + 1, (1 + a)(k+1) _ (1 + a)(1 + a)k, which by the induction hypothesis
=(1 +a) Li
+ k j- k
k k+1
k
).v.
a ±i a'+ 71
(- _ 1
I=a i=o i-t
Ck +
Using (a) show that the above is equal to I {a' 14. (a) By considering the sequence {ak - a} show first
that one can assume a = 0.
I
i j
2. Take I, _ [n oo). 3. (a) Set x" _ ( n2 + I)/n. Then x.2= 1 + , > 1 + " ; = x.'+,. Thus
x2. > x,2,+, > 1. From this it now follows that x" > x"+1 > 1 for all n E N. and that lim x" = 1. (c) Since
a > 1, a"+' = a a" > a". Thus {a"} is monotone increasing. Let or = sup{a" : n E N!}. If a < oo, then
a = lim a"+ = a lim a" = as > a. This, however, is a contradiction. Thus a" -4 oo. S. Use mathematical
induction to show that a" > 1 for all n (=- N. From the inequality 2ab s a2 + 62, a, b >_ 0, we have 2a" 5 a.2 + I or
a"+, = 2 - 1/a" s a". Therefore, {a"} is monotone decreasing. Finally, if a = Jim a", then
a = woo
lint a"+,="moo 1=2- 1.a
limit- a")
Therefore, a = 1. 7. (c) By induction, the sequence {a"} is monotone increasing and bounded above by 3. If
a = Jim a", then a2 = lim a"+, = lim(2a" + 3) = 2a + 3. Thus a is a solution of a2 - 2a - 3 = 0. The two solutions
are - I and 3. Since a must be positive, we have a = 3. (e) lim a" = 2. 9. (a) Use the inequality ab ! \ for all n. To show that {x"} is monotone decreasing, consider x",, - x" and simplify.
12. (a) e2 (e) e312 13. To show that {s"} is unbounded, show that s2. > I + "I. Hint: First show it for n = 1. 2. and
3; then use mathematical induction to prove the result for all n E N. 15. Hint: Fork = 2. t < k,k ! l) = k i - ;.
17. (b) n + n- (n - 1) (d) Hint: Write a = (1 + b) with b > 0. Now use the binomial theorem to
show that a"In° ? cn for some positive constant c and all n sufficiently large. 18. (c) The sequence is not monotone
increasing: If x" = n + (-1)'Vn-, then xu,, < x2,,. 23. This problem is somewhat tricky. It is not sufficient to just
choose a monotone increasing sequence in the set; one also has to guarantee that the sequence converges to the least
upper bound of the set. Let E be a nonempty subset of R that is bounded above. Let aqt denote the set of upper bounds of
E. Since E * 0, we can choose an element x, E E. Also, since E is bounded above, OIL * 46. Choose /3, E V. Let
a, = 12 (x, + 6, ), and consider the two intervals [x,, a, ] and (a,,19, ). Since x, E E, one or both of these intervals have
nonempty intersection with E. If (a,, /3,] fl E * 0. choose x2 E E such that a, < x2 s S,. In this case, set $2 = /3,. If
(a,, /3, ] fl E _ 0, choose x2 E E such that x, < x2 a,, and set /32 = a,. In this case, $2 E OU.. Proceeding
inductively, construct two monotone sequences {)"} and {/3"} such that (a) {x"} C E with x" 5 x"+, for all n, (b)
{/3"} C OIL with ,6. ? /3"+, for all n, and (c) 0 0" - x" <- 2-"+ 1(b, - x1). Assuming that every bounded monotone
sequence converges, let A = Jim /i". By (c) we also have /3 = lim x". It only remains to be shown that /3 = sup E.
24. Suppose A = {x" : n E N} is a countable subset of [0, 1 ]. To show that A C [0, l ] proceed as follows:
At least one of the three closed intervals [0, 13], [13.23). [j, 1 ] does not contain x,. Call it!,. Divide 1, into three closed
intervals of length 1/32. At least one of these, say 12, does not contain x2.

3. (a) {-1, 0. 1} (e) {1} 4. (a) e2. Set p" = (1 + i )3n. By Example 2.3.5, Jim p" = e. The answer now follows
since (I + 3 )6i = p.2. 7. (a) The set has only one limit point; namely { 11. Every point of the set is an isolated
point. 14. See the solution of Exercise 23 of Section 2.3.
1. (a) -oo, oo (c) -1, 1 (d) -3, j 3. 0, 1 6. (a) Let a = lim a" and y = I i(a" + b"). Since the sequences
are all bounded, a, y E R. Let e > 0 be given. Then by Theorems 2.5.3 and 2.5.4 there exists n° E N such that
a" > a - e/2 for all n ? n° and (a" + b") < y + e/2 for all n ? n°. Therefore. b" < y + e/2 - a" < y - a + e.
From this it now follows that limb" s y - a + e. Since e > 0 was arbitrary, we have limb" 5 y - a; i.e..
lim a" + Fim b" s lira (a" + Q. The other inequality is proved similarly. 8. 1. 1. Hint: Consider the subsequences
{s2.,} and {s2i,,+,}. 10. By Theorem 2.5.7 there exists a subsequence {a",} of {a"} such that a", --+a. Since {b"}
converges to b, a,,,b,,, -+ ab. Therefore, ab is a subsequential limit of Thus Tim z ab. The reverse
inequality follows similarly. The fact that b * 0 is crucial.
2. (a) The sequence {(n + 1)In} converges and thus is Cauchy. (d) Convergent sequence; thus Cauchy 4. (a) Show
that stn - s" Z. & For n 3, I a"+, - a" I < Ia, - a"_, I. Therefore, {a"} is contractive. If a = Iim a,,, then
;
0 < a < 1 and is a solution of a2 + 2a - 1 = 0. 11. (b) Since (a,,., - a") = (b - lxa" - a"_,), by induction
(a"_, - a") _ (b - 1)"-'(a2 - a,). Therefore,
n n-1
(ak+, - a*) _ (a2 - a,) Y, (b - 1)k = (a2 - a,) 2-
1(b-1)"
a.+l - al = b
.
k-S
Letting n - oo gives a = a, + i! b (a2 - a,).

1. Lets"= Sincek sk -1,k>2,forna2,s"{ 1 + Ik=2(k l - i)=2-2.
Therefore, {s"} is bounded above and hence converges by Theorem 2.7.6. 5. Use the inequality ab s 2 (a2 + b2),
a, b ? 0. 9. (b) Since the sequence (s"} of partial sums is monotone increasing, it suffices to show that some
subsequence is bounded. Consider the subsequence {s,,) where nk = 2k - I. and prove that s,,. < I + a + +
where a = 2-tP-1l. For example, when k = 2, nk = 3 and
s3=1+2P+
3P
=1+( +]1<1+P=1+2v -,-'
P
0 CHAPTER 3
2. By Corollary 2.4.8 a finite set has no limit points. Now apply Theorem 3.1.9. This can also be proved directly by
showing that the complement of a finite set is the finite union of open intervals. 6. (a) Let p E 0 = U,O0 be
arbitrary. Then p E 0, for some a E A. Since 0, is open, there exists e > 0 such that N,(p) C 0,: But then
N,(p) C 0; i.e., p is an interior point of 0. 8. (a) Let p E Int(E) be arbitrary. Since p is an interior point of E, there
exists an e > 0 such that N,(p) C E. To show that N,(p) C Int(E) it remains to be shown that every q E N,(p) is an
interior point of E. 9. (a) Since A U B is a subset of A U B (which is closed), by Theorem 3.1.11(c),
A U B C A U B. The reverse containment follows analogously. 11. (a) (a, b) 13. Use Exercise 6(b) of Section
1.5. 14. If p E D, set p, = p for all n E N. If p is a limit point of D, use Theorem 2.4.7. 17. Let {r,}'_, bean
enumeration of 0, and {t:,}- , an enumeration of the positive rational numbers. Take ' = {N,lr,) : j, n E N).
18. (a) Closed in X. (c) Neither open nor closed in X. 20. Suppose U C R is open, and suppose E C U is open in
U. Let p E E be arbitrary. Use the fact that E is open in U and that U is open to show that there exists an e > 0 such
that N,(p) C E. Thus p is an interior point of E. Since p E E was arbitrary, E is open in R. The converse is obvious.
23. (a) Let U = (0, 1) and V = (;, ;). 24. (a) Use the fact that A = A U A', and if U open satisfies u n A' # 0
then U fl A # 4,. 25. Let A be a nonempty subset of R. If A contains at least two points and is not an interval, then
there exist r, s E A with r < s and r E P with r < r < s. but r E A. The open sets U = (-no, r) and V = (r, oo) will
prove that A is not connected. Therefore, every connected set is an interval. Conversely, suppose A is an interval and A
is not connected. Then there exist disjoint open sets U and V with A fl U 0 ¢.. A fl V * 4,, and A C U U V. Suppose
a E A fl U and b E A fl V. By Theorem 3.1.13 applied to U and V, there exist disjoint open intervals f and J such
that a E I and b E J. Suppose a < b and J = (r, s). Show that t 14 U U V butt E A. This contradiction proves that
A is connected.

1. (b) Use the fact that 0 is a limit point of A. 4. (a) Let ° t = { U.J. be an open cover of A U B. Then 11 is also an
open cover of A and B respectively. Now use the compactness of A and B to obtain a finite subcover of A U B.
5. Since K is compact, by Theorem 3.2.9, K is closed and bounded. Let a = sup K. This exists since K is nonempty
and bounded. Now use the fact that K is closed to show that a E K.

4. If x E P with x = a, a2- , a, E (0, 2), set b, = )a,. Consider the function x --+.b,b,
CHAPTER 4
1. (a) If f(x) = 2x - 7, L = -3, then 1f(x) - L 1 = 21x - 21. Given e > 0 take S = e/2. Then for all x with
Ix - 21 < 8, 1 f (x) - L 1 = 2 ; x - 21 < 2S = e. (c) If f(x) = k/(1 + x), then
j(x) 21 21x+1111
< 2 lx- 1
for all x > 0. Hence given e > 0, choose S = min {2e, I ). With this choice of 8, x > 0 and thus
11(x) - I 0, x/I x 1 = 1, whereas for x < 0, x/I x 1
(c) The limit does not exist. Consider the sequence { I /nzr), which has limit 0 as n -+ no. 4. lim1 f (x) 3 5. (a) By
Figure 4.5, for 0 < t < -rr/2, sin t = length of PQ < length of arc PR = t. 7. (a) Use the inequality
IIf(x)I - ILII s If(x) - LI (see 2.1.4). (c) Use induction on n. S. (a) 0 (c) (e) J (g) 2. Note:
2( ). 9. Let L = lim f(x). By hypothesis L > 0. Take e = 1 L and use the definition of limit. 12. Since
x-r
g is bounded on E, there exists a positive constant M such that Ig(x) I s M for all x E E. Thus 1f(x) g(x) 15 MI f(x)'.
for all x E E. Now use the fact that lim f(x) = 0.
'v
16. Suppose lim f(x) = L. Let e > 0 be given. By definition there exists M > 0 such that I1(x) - L I < e for all
X-M
x E (a, oo) with x > M. Let S = 1/M. Then for all t E (0, 1/a) with t < S, 1 It E (a, oc) and I It > M. Therefore.
IS(r) - LI = If(',) - LI < e. The proof that limo g(r) = L implies iim f(x) = L is similar. 17. (a) 22 (c) 2
-M
(e) 2 (g) Limit does not exist. For all x > ;, cos ; > 1. Thus x cos x > 1 x and x cos ; - oo as x - oo.
1. (c) Since I - cos x = 2 sin2(x/2), for x # 0, g(x) sin2(x/2). Now use the fact that I sin t I s 11 1.
4.Ifp>0.
If(x) - f(P)I = If - VI = Ix - PI/(V + V P - ) < LIx - PI
Let e > 0 be given. Set S = m i n { p , V p e } . Then I x - p I < S implies that 11(x) - f(p)I < e. Therefore, f is
continuous at p. If p = 0, set S = e'-. Alternatively use Theorem 4.1.3 and Exercise 14 of Section 2.1.
6. (a) See Exercise 7a of Section 4.1. 7. Use Theorem 4.2.4, and the fact that x" is continuous on R for all
n E N. 9. (a) R \ (-2.0,2). (c) R.
12. (a) Use the fact that max {f(x), g(x)} =12(1(x) + g(x) + I f(x) - g(x) 1).
16. Consider g(x) = f(x) - f(x - 1), xE(O, I J. 18. Let p E E be a limit point of F. Use Theorem 2.4.7 and
continuity off to show that p E F. 20. (b) By induction, f(nx) = nf(x) for all n E h) and x E R. In particular,
f (n) = en where c = Al). Also, c = Al) = An -.) = nf(.1). Therefore, Al) = cln. Since f is continuous, letting
n -ioo gives f(0) = 0. From this it now follows that f(-x) = f(x) for all x E R. Thus f(n) = cn for all n E Z and
f(r) = cr for all r E 0 (write r = m/n, m, n E Z). Finally, by continuity Ax) = cx for all x E R. 22. Take
e = 1. Then for this choice of a there exists a 8 > 0 such that 11(x) - f(P)I s 1 for all x E Na(p) fl E. Show that
this implies that If(x)I c (If(P)I + 1) for all x E Na(p) () E. 25. Theorem 2.4.7 and Theorem 3.2.10 should
prove helpful. 29. By hypothesis, for each x E K there exists e, > 0 and M, > 0 such that I Ay) 1 5 M, for all
y E N1(x) fl K. The collection {N,,(x)},Ex is an open cover of K. Now use compactness of K to show that there exists
a positive constant M such that I f(y)I S M for all y E K.
2. (a) Suppose f(x) = x2 is uniformly continuous on [0, oo). Then with e = 1, there exists a S > 0 such that
J Ax) - f C Y ) I 1, then Iy" - x" I = I_11 < S for all n a n,,. But I f&.) -1(x")1 = 2+3, z 2 for all n. This is a contradiction!
3. (a) For all x, y E [0, oo),
If(x)-f(y)I = l+x l+y (I+x)(ly+Y)<Ix-y I.
Thus given e > 0, the choice 6 = e will work. (f) Set g(x) = sin x/x, x E (0. 1 ] and g(0) = 1. Then g is continuous
on [0, 1 ], and thus by Theorem 4.3.4 uniformly continuous on [0, 1 ]. From this it now follows that f is uniformly con
tinuous on (0, 1). 4. (a) Show that If(x) - f(y)I 5 1 x - yI for all x, y E [a, oo).
a3
5. (a) For x, y E [a, 00),

Imo- VY- I = Ix-yI/(V + VY- )52:rIx-yl.
7. (b) Suppose I f I and I g I are bounded by M, and M2, respectively. Then
I f(x)g(x) - f(y)g(y) I I AX) I I g(x) - g(y) I + I g(y) I I f (x) - f(y) I

5 MI I g(x) - g(y) I + M2I f(x) - f(y) I
Now use the uniform continuity off and g. 13. (a) Let x, E E be arbitrary. For it = 1 set x,,. = f(x.). Show that
the sequence {x"} is contractive.
2. (b) m. f(x) = ,lito. f(x) = 0 (f) lin f(x) = I. Hint: For <x '[ ] = n. 4. Use the fact that [.r; is
6
bounded near x,, = 2. 6. b = 14 7. For n E N set x = n and y = it + and show that I f (x.) - f(;)1 = sin 2.
12. If it E N, g(x) = x" is continuous and strictly increasing on (0, cc) with Range g = (0, oo). Therefore by
Theorem 4.4.12, its inverse function g-'(x) = x"" is also continuous on (0, oo). From this it now follows that f(x) is
continuous on (0, oo). 14. (a) Suppose first that U = (a, b) C 1. Then since f is strictly increasing and continuous
on 1. f((a, b)) = (f(a), f(b)), which is open. For an arbitrary open set U C 1, write U = where is a finite or
countable collection of open intervals, and use Theorems 1.7.14 and 3.1.6.
CHAPTER 5
EXERCISES S.1 page 174
1. (a) For f(x) = x3.
f(x + h) - f(x) - (x + h)3 - x' _
f '(X) = lim lim (3x2 + 3xh + h2) = 3x.
h ilm h =
(c) For h(x) = 1/x, x # 0,
h ,=
0
h '(.x) = .1im
.,
n-0 x(x + h) x'

(e)Forf(x)=x+1'x#-1'f(x)=ni0(x+1)(x+h+I) ),.
(x+l
2. If n E N, by the binomial theorem
2(n)
(x+h)"-x'= F,1n)hkx"-'l -I+h hr-'x"-4.
k-t k k 22
k
Dividing by h and taking the limit as h -- 0 proves the result for n E N. If n E 71 is negative, write x" = I /xf,
m E N. and use Theorem 5.1.5(c). 3. (a) Since cos x = sin(x + ), by the chain rule cos x = cos(x +
i (c) No. 7. (a) f'(x) exists for all x E 12\71.
-sin x. Alternatively, use the definition of the derivative. S. (a) Yes
Forx E (k, k + 1), k E Z,f(x) = x[x] = kr. Thus f'(x) = k for all x E (k, k+1), k E Z. (c) The function h is
differentiable at all x where sin x # 0. For x E (2kw, (2k + I ytr), k E 71, h'(x) = cos x. For x E ((2k - I )w, 2kwr),
k e Z, h'(x) -cos x. 9. (b) For x f 0, g'(x) = 2x sin - cos Since lynx cos does not exist. ii g'(x) does
0
not exist. 12. (a) f'(x) = 2/(2x + 1) (c) h'(x) = 3[L(x)]2/x 13. (b) f;(0) = hliw, ht''-'t sin ,which exists and
equals 0 if and only if (b - 1) > 0; i.e, b > I.

1. (a) Increasing on R. (c) Decreasing on (-oo, 0) and increasing on (0, oo), with an absolute minimum at x = 0.
4. (a) Show that the function f (x) = xv" - (x - 1)'r" is decreasing on the interval 1 <- x :5 alb. (b) Set
f(x) = ax - x°, x =' 0. Prove that f (x) ? f (l) to obtain x° s ax + (1 - a), x ? 0. Now take x = alb.
5. (a) If f"(c) > 0. then there exists a S > 0 such that f'(x)/(x - c) > 0 for all x. I x - c I < S. Therefore,
f (x) < 0 on (c - S, c) and f (x) > 0 on (c, c + S). Thus f has a local minimum at c. 7. Since P(2) = 0 we can
assume that P(x) = a(x - 2)2 + b(x - 2). Now use the fact that P must satisfy P(I) = I and P'(I) = 2 to determine a
and b. 12. (a) Let t, -c. Since f'(c) exists, lim (f(r") - f(c))/(t - c) = f'(c). Now apply the mean value theorem.
15. Since f+(a) = lira (f(x) - f(a))/(x - a;70, there exists a S > 0 such that (f(x) - f(a))/(x - a) > 0
for all x, a < x < a + S. 17. Hint: Consider f (x). 22. (a) For fixed a > 0 consider f(x) = L(ax), x E (0, x).
(c) By (a) and (b), L(b") = nL(b) for all n E Z and It E (0, oo). But then L(b) = L((b"")") = it L(V"). From this it
now follows that L(b') = rL(b) for all r E Q. Now use the continuity of L to prove that L(b") = xL(b) for all x E R.
where If = sup{b' : r E Q. r : x}. 23. (b) Since tan(Arctan x) = x. by Theorem 5.2.14 and the chain rule,
tan(Arctan x) = (sec(Arctan x))(' Arctan x) = I. The result now follows from the identity
sec2(Arctan x) = 1 + x2. To prove this, consider the right triangle with sides of length 1. 1 x (, 1 + x2 respectively.

f(x) -f(x,')/g(x) - g(xo)
2f(x)
.
g(x)
=
x - X. x - X.
Now use Theorem 4.1.6(c) and the definition of the derivative.
4. Use the fact that since ,ilitf(x) exists, f(x) is bounded on (a, a + 8) for some S > 0.
6. (a) lim
x3+2x-3 = lint 5x4+2 = 7 (e) By L'Hospital's rule. lim
Inx
- = =-
1
lim= -x = 0.
!- 2x3 - x2 - 1 =-+I 6x2 - 2x 4" +-'= x
(e) Make the substitution x = 1/t.

1. Let c, > 0 be arbitrary. By Newton's method c", (2c; + ar)/3c2, 2. (a) f(0) = I and f (1) _ -1. Therefore.
f has a zero on the interval (0, 1 ]. With c, = 0.5. c2 = 0.33333333, f(c2) = .037037037, c3 = 0.34722222,
f (c,3) = 0.000195587, c4 _ .34729635,f(c4) = 0.000000015.
CHAPTER 6
(a) f(x)
_ 1, 0Sx< 1, Let 91= xa
{ x j,
,, ... , x"} be any partition of [0, 2] and let k E {l.... , n} be such
2, 1 s x <- 2.
that x4 _, < I s x4. Then W(9D, f) = 4 - 3x4 and gt(9P, f) = 4 - 3x4_ ,. Thus 4k(9', f) - W(9', f) = 3(x4 - x4-,). By
Theorem 6.1.7 it now follows that f is Riemann integrable on (0, 2]. Also, since x,,- I < I S x4, 4914) 5 1 < 14911f)
for any partition P. Hence f f = 1. Alternatively, consider the partition 91 = {0, c, 1, 2} where 0 < c < I is
arbitrary. For this partition, 2(9j) = 1 and 411.(9', f) = 4 - 3c. The result now follows as above. 3. (a) If
91 = {xo, x,, .... x"} is a partition of [a, b], then inf{ f(t) : t E [x;_ x;]} = sup{f(t) : t E [x;_,, xj]} = c. Therefore,
20,J) = %4911f) = jt c(x; - x,-,) - c(b - a). 4. (a) Since f(x) _ [3x] is monotone increasing on (0, I ], f is
Riemann integrable on (0, 1]. For n 2:4 consider the partition P. _ {0, 1- t 1 }. For this partition
2(91, f) = I and 9t(9>", f) = I + I. From this it now follows that Jo [ 3x] dx = 1. 6. Let 91 = {xj,, x,, .... ..x.) be a
partition of [a, b). Since f(x) <- g(x) for all x E [a, b), sup{f(t) : t E [x,_,, x,]) :s sup{g(t) : t E [xj_, x,]) for all
i_= 1.... , n. As a consequence, alt(9P, f) s '9t(9', g) for all partitions 9 of (a, b). Ta ag the infimum over 9' gives
f f s " j g. The result now follows from the fact that f, g r= 9t[a, b]. 8. f f - 0, fo f = 3. 10. Since a z 0,
f is increasing on [a, b]. Thus if 91 = {xo, x,, ... , x"} is a partition of [a, b], and M, = x}. Therefore.
JI "
.T(9,e) = j zj-. £1 and `L(',f) _ x Axj.
ice!
Now show that x2_ i Axj :53(x3 - x3_,) s x,2Ax4. From this it will now follow that 2(9, f) s 3(b3 - a3) s
Since f is continuous, f E 9t(a, b) with f; f = 3(b3 - a3). 12. (a) With 91" =.-O.: k = 0, 1, ... , n} and f(x) = x,
VW)= " (k)l (=;k; k.
By Exercise la, Section 1.3, Zk_, k = 3n(n + 1). Therefore, fo x dx = him "V = 3. 15. (a) Since f is
bounded on [a, b], If(x)1 s M for all x E [a, b]. Let 9 - {xo, x...... x"} be any partition of [a, b). For each i let
M(f2) and m,(f2) denote the supremum and infimum of f'- respectively over [x,_,, x,] with an analogous definition for
M,(f) and m,(f). Let e > 0 be given. Then for each i there exists s t, E [x, _ x,] such that M,(f2) <f2(s,) + ?e and
m,(f2) >f2(t,) -/ `Z e. Therefore,
0:5 We) - m,V 2) <.f2(si) - f2(,,) + e 5 I As,) +f(ti) I I f(s,) - 1(ri) I + e

5 2M[M,(f) - m,(f)] + e
Since this holds for any e > 0, M,(f2) - m,(f2) <_ 2M[M,(f) - m,(f)]. Now use Theorem 6.1.7.
16. Assume that f is continuous on [a, b] except perhaps at a orb, or both. Since f is bounded, there exists M > 0
such that I f(x) I <_ M for all x E [a, b]. Let e > 0 be given. Choose y,, y2, a < y, < y2 < b so that y, - a < e/8M
and b - Y2 < s/8M. Then
[sup{f(t):t E [a, y,]} - inf{f(r):r E [a.y,]}](y, - a) 5 2M(y, - a) < 41 e.
Similarly for the interval [y2, b]. Since f is continuous on [y,, y2] there exists a partition 91 of : y y2] such that
ql(g, f) - $(9', f) < 2 E. Let 9* = 01 U (a, b). Then @1* is a partition of [a. b]. and by the above. 'tt(91*, f) -
2(91*. f) < E. Thus by Theorem 6.1.7, f E Jt[a, b]. Suppose f is continuous on [a, b] except at a finite number of
points c,, c2, ... , c with a s C1 < c2 < . . < c" 5 b. Apply the above to each of the intervals [a, c, ], [Cl- c2], .. .
[c", b] to obtain a partition of [a, b] for which Theorem 6.1.7 holds.

2. (a) Take t, = 3(x; + x,x,_, + x;_,). 3. (a) If c, * a, set co = a. Similarly, set c"., = b if c * b. First prove
that f E gt[c,_ c,], i = 1, 2, ... , n + I with f .,f = 0. Now use Theorem 6.2.3. (b) Set h = f - g.
5. Since f is bounded, I f(x)I s M for all x E [a, b] for some M > 0. Use equations (7), (8), and the fact that
f E 01 [c, b] for every c E (a, b), to prove that
(b (b ('c rc
O S J f- J f_ J f- f<- 2M(c - a).
J
Use this to show that f E i [a, bJ. 7. (a) lim k2 = x2 dx = -. 8. For each fixed c, 0 < c < 1, the
-n3 Ju 3
series defining f becomes a finite sum on [0, c]. Evaluate and use Exercise 5.

1. Sincef is bounded, I f(x)I <- M for all x E [a, b]. If x, y E [a, b] with x < y, then
IF(y)-F(x)I = JH
Thus F satisfies a Lipschitz condition on [a, b] and hence is uniformly continuous. 2. (b) F(x) = x for 0 5 x :s
F(x) = i - 2x for I < x 5 1. 2:
S. (b) F'(x) = cosx2 6. By the chain rule, -dO-.L(() = -1 = j[-L(x)]-Therefore, L(F) = -L(x) + C for some
constant C. Taking x = I shows that C = 0. 9. Let m and M denote the minimum and maximum off on [a. b],
respectively. Since g(x) z 0 for all x, mg(x) s f(x)g(x) 5 Mg(x). If fQ g > 0, then from the previous inequality one
obtains
Jb1g/Jbg
m SM.
a a
Nowapply the intermediate value theorem to f. If f, g = 0, use Theorem 6.2.2 to prove that fa fg = 0 and thus
the conclusion holds for any c E [a, b]. 11. (a) With rp(x) = In x. by Theorem 6.3.8 f '; dx = fo" 2 x dx = 2(In 2)2.
i
(b) Use integration by parts. (e) 2 In 2 - 1. Use Exercise 10 with (p (x) _ V. and f(t) = t/(1 + t).
S32 Hints and Solutions to Selected Exercises

1. (a) Since 0 < p < 1, jox-adz = oJim+ f'x-vdx = lim j[ 1 - c'-'] _ A Note: If p s 0, then x_a is ,
continuous on [0. 1], and if p z 1, then the improper integral diverges.
(d) Converges, with jox in x dx
2. (a) Converges, with fo e x dx = Jim fo e x dx = Jim 1 - e ` = 1. (e) f2 b dx = In (In c) - In (ln 2), which
diverges to coo as c -+ oo. (i) Use partial fractions to obtain
(
Jo (x2+1)(x+1)
x
+2 _1
c+11+2Arctanc-2.
Thus 1dx ar 3. (a) For p > -1, the improper integral converges for all q E R. and
Jo ( + l)(
xz+ ) 4 2.
for p < -1, the improper integral diverges for all q E R. When p = -1. the improper integral converges for all
q < -1, and diverges for all q ;2! -1. 5. Use the fact that 0 s I f(x) I - f(x) s 21 f(x) I
7 . (a) I;{ Idx 5 j;i dx = ; - Thus Jim J,,l fI < oo. (b) By integration by parts.
!; dx = -
s - j: r ZZ dx. By (a) and Exercise 5. Wn j*'X2 dx exists. Also, Win = 0.7herefore.
C-00
Jim f
C-W
dx exists. 9. (a) To show convergence of the improper integral, consider the integrals of t"' a-' over
the two intervals (0, 1 ) and [ 1, oo). If jr ? 1, then t'-'e-'is continuous on [0, 1). and thus the integral over [0, 1 ]
clearly exists. If 0 < x < 1, then since a-' s 1 fort E (0, 1 ],
Os
f t'-'e'dts I
C e
Thus 1 to + f'tx-'e-'dt < on. For tit 1, use l'Hospital's rule to show that there exists a t, ? 1 such that
tx'' 07 for all t 2t t,. Thus tx' a-' s e'"2 for all t Z t, and as a consequence, Pm. IR tx-'' tit < on. Thus the
improper integral defining r (x) converges for all x > 0.
1. 2f(0) 3. (a) See the hint to Exercise 1, Section 6.3. 5. (a) f - 1. Use Theorem 6.5.10. (C) 12 + 22 + 32.
For x r= [0, 3], [x] = 1(x - 1) + 1(x - 2) + I(x - 3), now use formula (12). & (a) jx , ii2. 12. (a) As in the
solution of Exercise 15a of Section 6.1, if 9 = {xo, x1, .... x"} is a partition of (a, b), Mi(f2) - m,(j2) s
2M[M,(f) - m,(f)] where M > 0 is such that I f(x)I s M for all x r= [a, b]. From this it now follows that
0 s ft (9,f2, a) - 2(9, f2, a) s 2M[olt(9, f, a) - T(9,f, a)].
Now apply Theorem 6.5.5.
.
2. (a) With n = 4, h = .25. Set x, _ .25i, y, = f(x,), i = 0, 1.2.3.4. Then yo = 1.00000, y, = 0.94118,
Y2 = 0.80000, y3 = 0.64000, y4 = 0.50000. Therefore,
T4(f) = Yo + 2Yi + 2Y2 + 2Y3 + Y4] = 0.782795.

25
S4(f) = 3
[Yo + 4y, + 2y2 + 4Y3 + Y4] = 0.785393.
By computation f"(t) - 2(3t2 - 1)/(1 + t)3. Using the first derivative test, f(t) has a local maximum of at t = 1.
Therefore, I f (t) I s((I for all t E [0, 1]. Thus by equation (23) with n = 4,
I4_T4(f)1s12\2)42=0.0026042.
Since 4 = 0.7853982 (to seven decimal places), I , - T4(f) I = 0.0026032. 3. (b) By computation, f (41(X)
3(4x2 - 1)(1 + X2)-712. By the first derivative test, the function f (") has a maximum on [0, 2] at x = v5/2. Thus
ft4>(x) I s 6/ (1 -+41Y < 1. If we choose n (even) so that fo f - S (f)j < 10-3, then we will be guaranteed
accuracy to four decimal places. By inequality (26) with M = 1, we need to choose n so that Ik'o -L < 10-3, or
n4 > 17,778. The value n = 12 will work. This value of n will guarantee that E12(f) < 0.0000086. Compare your
answer with the exact answer of V + 2ln(2 + V5).

4. No. Consider g = Xc on [0, 11 and let f be the zero function.
CHAPTER 7
2. (a) Diverges (c) Converges (e) Diverges (g) Converges by the ratio test (I) Converges (m) Converges for
p > 1; diverges for 0 < p s 1 3. (a) Converges to 1 /(1 - sin p) for all p E R for which sin p I < 1; that is, for
all p * (2k + 1)j, k E Z 4. (b) Since 7, ak converges, lim ak = 0. Thus there exists k, E N such that 0 s ak s 1
for all k -- k,. But then 0 5 ak s ak for all k a k and 7, ak converges by the comparison test. (d) Take ak = I/k2.
5. The series diverges for all q < 1, p E R, and converges for all q > 1, p E R. If q = 1, the series diverges for
p s 1 and converges for p > 1. 6. Suppose L, where 0 < L < oo. Take e = IL. For this e, there
exist k, E N such that #L 5 ak/bk s }L for all k z k,. The result now follows by the comparison test. 12. The proof
u++ses the fact that lim(LPY' Vk- = 1 for all n r= Z. 13. The given series is the sum of the two series
`k 1 Ijt and Mk t ?Jp, each of which converges. 16. let s, = a, + a2 + + a,,, and tk = a1 + 2a1 + +
2ka2-. By writing s, = ak + (a2 + a3) + (a4 - as + a6 + a7) + , show that if n < 22, then s s tk, and if is > 2k,
then s Z I'tk. From these two inequalities it now follows that F ak < oo if and only if I 2ka2, < oo.
1& (a) Diverges. If ak - 1/(k In k), then 2ka2. = 1/(k In 2). 19. Use Example 5.2.7 to show that ck - ck+ 1 a 0 for
all k. Thus {ck} is monotone decreasing. Use the definition of In k and the method of proof of the integral test to show
that ck =' 0 for all k. 21. Write ak+I/ak - I - xk/k where xk = (q - p)(k/(q + k + 1)). 22. (c) When p = 2,
k-fI.3...(2k- 1)12 Tk r =11`1-2,J
112
2.4... 2k ) zl-2k
( 1)a
Now use the fact that M9(1 + h)I"k = e. 23. (a) Set s, = L4., ak and let s = lim s,,. Consider the series E bk
where b, = (1< - W-771 ) and fork z 2, bk = (Vs_, - V).
1. If {b,} is monotone increasing to b. consider $00- , (b - bk)ak. 2. Take bk = 1/k fork odd, and bk = i/k2 for k
even. 4. If D. = E[SiO(k
1 coos kt, then
1 r \ 1
1 sin 2 I sin ?tJ.
2 2
5. (a) Converges (c) Converges (d) Diverges; kim , , = * 0 (f) Converges (h) Converges for all
t * 2nsr, n E Z. If t - 2nar, then the series converges for p > 1 and diverges for 0 < p s 1. & Use the partial
summation formula to prove that
n n-1 n
I kak = nA, - Y, Ak, where Ak = Y, al-

k-1 k-1 k-I
Now use Exercise 14 of Section 2.2.

2. Use the inequality lab 1 s 2 (a2 + b2), a, b E R. 6. (a) Converges conditionally (c) Converges absolutely for
p > I and conditionally for 0 1, and conditionally for 0 < p s I
S. Fast note that
n-I
S. - -
1 I
- I
3k + I + 3k + 2 3k + 3)'
Now show that Si,, -+oo as n --, oo. 10. By Theorem 7.2.6 the series converges. To show that S ' t = oo, show
that for any three consecutive integers, at least one satisfies I sin k I
1. (a) 11{l/(In k)}112 = Moko-2 1/(In k)2, which diverges (Exercise 5, Section 7.1) 2. (a) Ip I < 1 (c) p z 2
4. Since { 1/k} E 12, the result follows by the Cauchy-Schwarz inequality. 10. If we interpret the vectors a and b as
forming two sides of a triangle, with the third side given by b - a, then by the law of cosines, I(b - a(I2 = Ilbll2 +
11a112 - 211 a 112 11 b 112 cos 0. Now apply Exercise 9e.
CHAPTER 8
nx _ 0, x = 0, 0, x * k,r, k E Z,
1. (a) lim 1, x>0 (c) lira (cos x)2n __
"-.o, 1, x=kar,kEZ
3. (c)
rf n = j iM
n2x d x + f
2/
(2n - n2x) dx = i + = 1 S. (a) If x = 0, fn(0) = 0 for all n e N. If x > 0,
10 0 1/n
then 0 < fn(x) < x/n, f r o m which the result follows. (b) F o r each n E N& (x) has a maximum of e ' at x = n.
6. Use the fact that forN,MEN,
N(M M N M 00 00 00
"-I M-1
an. m
M-1 R-t
an. m
M-1 R=t
a, m ( 11-1
an. m
The above inequalities hold since a0, m > 0 for all n, m E N. Now first let M - oo, and then N -b oo, to obtain
an. m an. m
n=1 mil m=1 nil
The same argument also proves the reverse inequality.
2. (b) Suppose {f,} and {g,} converge uniformly to f and g respectively on E. Then I f"(x)gn(x) - f(x)g(x)1
I g"(x) I I f"(x) - f(x) I+ I f"(x) I I g" (x) - g(x) 1. By hypothesis, I S. (x) I s N for all x E E. n E N. Also, since
I f"(x) I s M for all /x E E, n E! /N, I f(x)1 fm M for all x E E. Therefore,
Ifn(x)8nx) - f(x)8(x)i N1 MX) - f(x)I + Ml 8.(x) - 8(x)I
Now use the definition of uniform convergence of { fn} and {g,} to show that given e > 0, there exists no E N such
that I fn (x) g" (x) - f(x)g (x) ( < e for all x E E and n a n0. 4. Find Mn = max{ fn (x) : x E [0, 111, and show that
M. -+ oo. S. (a) For x E [0, a], I f"(x)1 s a". If 0 < a < 1, then lim a" = 0. Thus given e > 0, there exists
n, E N so that d` < e for all n at n,; that is I f"(x) I < e for all x e[ -w
O, a], n z n,. Therefore, f f.) converges
uniformly to 0 on [0, a] whenever a < 1.
for all x E It Since .1 < co. the series 2

8. (a)
ii + X
2
+ converges uniformly by the Weierstrass
k' j?
M-test. (c) For x ,- 1, k2e-k: s k2(1/e)'t. Since 1/e < 1. the series 7, k2(1/e)k converges.
I
9. (a) 5 C R for all x E R. Since 7 < co, the given series converges
I
(2k +))3n 1 (2k 1)3/2
1 1
uniformly for all x E R by the Weierstrass M-test. (c) Hint: Let S,(x) = ku(kz + 2 - (k + 1)x + 2)
10. (a) For x ? a > 0. 1 + k2x ? ak2. Thus since 1/k2 < no. by the Weierstrass M-test, the given series
converges uniformly on (a, oo) for every a > 0. To show that it does not converge uniformly on (0. oo), consider
(S, - 5_t)(1/n2), where S. is the nth partial sum of the series. 1& Suppose I Fa(x) M for all x E [0, 1 J. Show
that I F,(x) I < M for all x E [0, 1], n E N. Now use the Weierstrass M-test.
n.

1. Show that kLx(1 - xyt = {?' Thus by Corollary 8.3.2, the convergence cannot be uniform on
[0, 1 ]. 4. Since f is uniformly continuous on R, given e > 0, there exists a S > O such that I f(x) - f(y) I < e for
all x, y E R, Ix - yI < S. Choose n, E N such that 1/n, < S. Then for all n ? n I f(x) - f,(x) I < e for all x E It
6. Let e > 0 be given. Since (f.1 converges uniformly on D. there exists n, E N such that I f,(x) - f(x) I < e for
all x E D. n, m > n,. Use continuity of the functions and the fact that D is dense in E to prove that
If(y)-fm(y)I Sefor all yCE E,n,in >n,.

1. By the Weierstrass M-test and the hypothesis on {ak}, the series akxk converges uniformly on [0, 1). Now apply
Corollary 8.4.2. 4. Since f E R[0, I ],f is bounded on (0, 1 ], i.e., I f (x) 15 M for all x e [0, 1 ]. Now apply the
bounded convergence theorem to g,(x) - x"f(x), which converges pointwise to g(x) = 0.0 5 x < 1, and f(1) when
x = 1. 7. For each k E N the function 2-kl(x - rk) is Riemann integrable on [0, 1] with fa 2-k 1(X - rk)dx =
2-k(1 - rk). By the Weierstrass M-test the series converges uniformly on [0, 1). Thus f E 31.[0, I) with fo f =
TOO 2-k(l - rk). 9. By Theorem 6.2.1, f g E 9t[a, b] for all n (=- N. Show that { f g} converges uniformly to fg on
[a, b] and apply Theorem 8.4.1. 10. Since If(x) 15 g(x) for all x E [0, oo). n E N, the same is true for I f(x) I. By
Exercise 5, Section 6.4, it now follows that the improper integrals of f n e N. and f converge. Since fog < on,
show that given e > 0, there exists c E R, c > 0, so that f,'g < Ie. Now show that
If f-Jf.1 5Jlf-fa[+2Js.
Use the uniform convergence of to f on [0, c] to finish the proof. 11. (b) To show that (19[a, b), n I6) is not
complete, it suffices to find a sequence If.) of continuous functions that converges in the norm II 1i to a Riemann
integrable function f that is not continuous.

2. By the fundamental theorem of calculus, f,(x) = f,(x,) + fx'f,(t) dt for all x r= [a, b]. If { f )converges
uniformly tog on [a, b], use Theorems 6.3.4 and 8.4.1 to prove that {f} converges uniformly to a function f on
[a, b) with f (x) = g(x) for all x E [a, b]. 4. Let x E [a, b] be arbitrary, and choose c, d such that
a < c < x < d < b. Now apply Theorem 8.5.1 to the sequence {f,} on [c, d], to obtain that f is differentiable at x
with f (x) = f;(x).
kx)_2
6. (a) Use the comparison test to show that the given series converges for all x > 0. Let S(x) _ 7,k , (1 + and
S (x) = Mk=, (1 + kx)-2. Then S.(x) = -228k= I k(1 + kx) -3. Use the Weierstrass M-test and the comparison test to
show that the sequences {S (x)} and IS' (x)) converge uniformly on [a. oo) for every a > 0. Thus by Theorem 8.5.1.
00
kx)_3
S'(x) = limS., (x) _ -2 ;k(1 +
k-I
for all x E (a, oo). Since this holds for every a > 0, the result holds for all x E (0, oo).
2. Let 91 _ {xo, x ... , be a partition of [a. a + p]. Set y1 = x, - a. Then 91* = {y y ... , is a partition
of [0, p]. If t E [x;_,, x;], then r = s + p for some s E [y; _,, y;]. Since f is periodic of period p,
f(t) = f(s + p) = f(s). Therefore, sup{f(t) : t E [x,x,]} = sup{f(s) : s E and as a consequence,
xlL(91, f) = q.L(9", f). From this it now follows that:
J.+P v
a
f=f'f
The proof for the lower integral is similar. Thus f E 9L[0,p] if and only if f E 9t[o,a + p].
4. (a) c = }(n + 1) 6. Set A,(8) = sup{Q (x) : x E [-S, S]). Then 0 < 8, < S2 implies A (S2).
Suppose tim A (S,) < oo for some 81 > 0. Then there exists a finite constant C. and n, E N such that A (S) s C for
all n ? n 0 < S 5 81. Use this fact to obtain a contradiction to the hypothesis that is an approximate identity.

1. (b) R = 2 (d) R = e 2. (b) By the root test the series converges absolutely for all x, - 2 < x < t and
1 1
diverges for all other x E R. 3 (a) x/(1 - x)2 S. (a) + x2 = 1 - (-x2) = (- l)kx2k, Ixl < 1. Use the
A-D
previous exercise and the fact that Arctan x = fo(l + t2)", dt to find the Taylor series expansion of Aretanx at c = 0.
(c) Use Theorem 7.2.4. 12. (b) By Example 8.7.20(c),
(- kk+I (x- l)k,
lnx=ln(l +(x - 1)) =
which converges for all x, 0 < x 5 2. (d) By computation, f O) (x) _ (k - })(t - x)-(k' 1R). Therefore, the
Taylor series expansion of (I - x) -'/' is given by
00 1.?2 . (k-2xk
2
1+ kl
k=I
For -1 < x s 0, use Theorem 8.7.16 to show that
(n + 1)!
Use convergence of the series
13 (n + )
IxIA.I. IxI < 1,
7 (n + l)!
to conclude that lim R0(x) = 0, - i < x 5 0. If 0 < x < 1, use Corollary 8.7.19 to show that
1.3...(n+ x fx-
n! (1-x)341-0)
for some {', 0 < C < x. Now use the method of Example 8.7.20(c) to show that lim R"(x) = 0 for all x, 0 < x < 1.
Thus the series converges to (I - x)-'R for all x, I x I < I. (e) Use the fact that '
Aresin x = 1 dt, Ix I < 1.

1 1 - t2

1. (a) r(2) = r(3 + 1) _ ]r(i) = }V;i 3. (a) foe-'r'ldt = r(2) = WW
ran i r(n + 21)x(21)
S. (a) (sinx)z'dx = 2
la r(n + 1)
CHAPTER 9
1. j' 0i = j- 1 = 2 and j 022 = j x2 = ;. Therefore, c, _ If 1, sin wx dx = 0 and c2 = 2 j' 1 x sin ax dx
Thus by Theorem 9.1.4. S2(x) _ ;x gives the best approximation in the mean to sinir x on [ -1, 1 ].
3. (a) a, 1, a2 = 1. a3 = -6 S. (c) b" = : fo x sin nx dx - - w cos mr = 2,(-1)"''. Therefore,
f(x) ^' 2 k)k+
sin kx.
k=1
6. (c) x - 2 - a E00 ( + l)2 cos(2k + 1)x 12. (a) As in the proof of Theorem 7.4.3, for A E R,
k-O
O :s II x - kyll2 = Kxil2 - 2A(x,y) + A=IIy112. If y * 0, take k = (x,y)11IYII2 to derive the inequality.

4. (a) For the orthogonal system {sin nx}.° , on [0, w], f sin2nx dx = 16. Therefore, b" = w f f(x) sin nx dx. Thus
Parseval's equality for the orthogonal system (sin nx) becomes
CC
M-1 fIr
0
2 2
(b) (1) g , (11)S. Use Parseval's equality and the fact that fg = i((f +'g)2 - f2 - g2] 6. Any function
6
that is identically zero except at a finite number of points will satisfy ja f dx = 0.

1. (a) If f is even on [ -zr,vr], then f (x) sin nx is odd and 1(x) cos nx is even. Thus b = O for all n = 1, 2... , and
;jof(x)cosnxdx,n=0,1,2....
2 (1 (-1)k)
3. (a) f(x) -- 1, k sin kx =
a k-0 2k + l
sin(2k + 1)x
IT
(c) I xI ^-
- ,rr F, (2k + 1)2
cos(2k + 1)x. (e) 1 + x -- I + 2 1 (- k sin kx
k=0
4 ao 00 1 k x
5. (a) 1,
1
a k<0 2k + 1
sin(2k + 1)x (d)
2 a k.o 2k + I
cos(2k + 1)x, -?
a"=o n LL
1 I (-1)" - cos I2 sin nx
6. (c) Since h, is even, the Fourier series of h, is the cosine series of h. Therefore.
(aR r*
a0 =
o x dx + I (ir - x) dx = itc, and
l
i
iran
(a/2 a
a" = x cos nx dx + -J (in - x)cos nx dx = Z l 2cos Z + (- I)"- ' - I
0 a/2 L
Thus h,(x) ^-
c-
4
c
a k-, k2
cos 2kx. & By integration by parts, bk = -
Since f" E fit[-ir, ir], it is bounded on [-ir, ir]; i.e.,I f" (x) I <- M. Therefore, I bk I S 2M/k2. Similarly for ak.
! Z1
irk
a
, f'(.x)sin kx dx.
Thus by the Weierstrass M-test, the Fourier series off converges uniformly on [-ir, ir).

a
2. 90 3. To show that {sin nx}.° , is complete on [0,ir] it suffices to show that Parseval's equality holds for every
f E 9t[0,ir]; i.e.,
0°
b,2, f2(x)dz.
^' 1 o
To accomplish this, let f^ denote the odd extension off to [ -ir,ir]. Since f" is odd, a^ = 0 for all n = 0, 1, 2. .
and b^ = A fo f(x)sin nxdx. Since the orthogonal system {I,cos nx,sin is complete on [-a, ir],
pp a a
b2, =7fI J fo(x)dx = 7r
? f2(x)dx.
0
^-1 a
6. Let S^(t) = lao + 1k-1(ak cos kt + bk sin kt). If x E [-ir, irthen
` f(t) dt - jx S^(t) dt s J If (1) - S^(t) I dt.

1
1 J s a a
k-i\1
Thus by Exercise 5a and Theorem 9.4.7, lim f'. S^(t) dt = f', f(t) dt, with the convergence being uniform on
[-ir,ir]. But
S^(t) dt = 2ao(x + ir) + ± I sin kx - kl (cos kx - cos kir) .
JJa k
from which the result follows. 10. Use Lemma 9.4.8 and the Weierstrass approximation theorem.

I) I)
3. (a) 1 (b) 1 -
((_1)ke"*
2. 1r 6. (a) + _ coskx (b) On [-ir,a),theseries
12 6 2 2 4 air IT a2 + k2
converges to el'i, and thus to the 2ir-periodic extension of e'l on alI of R.
CHAPTER 10
2. Since U is open and nonempty, there exists x E U and r > 0 such that (x - r, x + r) C U. Thus by Theorem
10.2.4, m(U) z m((x - r, x + r)) = 2r > 0. 3. First show that m(P^) = 2m(P^_1) for all n E N. From this it now
follows that m(P) s (3)' for all n E N.
6. First show that there exist disjoint bounded open sets U1, U2 with U, 1) K, and U2 J K2. Then
m(K,UK2)=m(U,U(.2)-m((U,UU2)\(K,UK2)).But(U,UU2)\(K,UK2)=(U,\K,)U(U2\K2).Now
use Theorem 10.2.9.

1. (b) First show that if U is any open set, then U + x is open and m(U + x) = m(U). Use this and the definition to
prove that A*(E + x) = A*(E). If K is compact and U is a bounded open set containing K. show that
(U + x) \ (K + x) = (U \ K) + x. Use this to show that m(K + x) = m(K) and A*(E + x) = A*(E). 3. Since
E, fl E2 C E, and A*(E,) = 0, A*(E, fl E2) = 0. Thus by Theorem 10.3.5, E, fl E2 is measurable. For E, U E2 apply
Theorem 10.3.9. 6. If A*(E) < oo, then for each k E N there exists an open set Uk with UA D E such that
m(Uk) < A*(E) + k. Now use the fact that E C fl U" C Uk for all k E N. S. Set Ek = E fl (-k, k), k E N. Then
{A*(Ek)} is monotone increasing with A*(Ek) <- A*(E) for all k E N. Let a = 51im A*(Ek). Suppose a < A*(E).
Choose jS E R such that a )3. Use this to show that there exists k G N such that A*(Et) > ,B for all k ? k, which is a contradiction-

2. If E is bounded, the result follows from the definition of A*(E) and A*(E), and Theorem 10.4.5(b) (for a finite
union). If E is unbounded, let E. = En(-n. n). Given e > 0, choose U. open such that E C U. and
A(U" \ E) < e/2". Let U = U U,;. Show that U \ E C U (U" \ E"). Now use Theorem 10.4.5 to show that
A(U \ E) < e. To obtain a closed set F C E satisfying A(E \ F) < e, apply the result for open sets to E.
4. First show that A(E, U E2) = 1; then use Theorem 10.4.1. 6. If E satisfies A*(E fl T) + A*(E` fl T) = A*(T) for
every T C R, then E satisfies Theorem 10.4.2 and thus is measurable. Conversely, suppose E is measurable and
T C R. If A*(T) - oo, the result is true. Assume A*(T) < oo. Let e > 0 be arbitrary. Then there exists an open set
U D T such that A(U) < A*(T) + e. Since E and U are measurable, E fl U and E` n U are disjoint measurable sets
with (E fl U) U (E` fl U) =U. Furthermore, E fl U D E fl T and E` fl U D E` fl T.Thus by Theorem 10.3.9,
A*(T) s A*(E fl T) + A*(E` fl T) S A(E fl U) + A(E` fl u) = A(U) < A*(T) + e.
Since the above holds for every e > 0, we have A*(E fl 7) + A*(E` fl T) = A*(T).

[0,1], if c < 0.
(0, 1 ], if OS c< 1,
1. {x : f(x) > c} a
(0, 1) U {1}, if l s c < 2.
(0, 1). if 2 s c.
5. If c > 0, then {x : > c} = {x : g(x) > 0} fl ix: g(x) < 1}. Since g is measurable, each of the sets {g(x) > 01
and {g(x) < <} is measurable. Thus their intersection is measurable. The case c < 0 is treated similarly. I0. (a) If
c z 0, then {x : f'(x) > c} _ {x : f(x) > c}, and if c < 0. then {x : f'(x) > c} = E. Since each of the sets
{f(x) > c} and E are measurable, f' is measurable. (c) Not in general. If E is a nonmeasurable set, consider the function
that is 1 on E and -1 on E`. 14. Since f is differentiable on [a, b], f'(x) - lim n(f(x + 1) - f(x) for all
x E [a, b]. For each n E N, g"(x) = n[f(x + ,'-,) - f(x)] is measurable (justify). Thus by Corollary 10.5.10. the function
f is measurable. 15. First show that given e, d > 0, there exists a measurable set E C [a. b] and n, E N such that
A([a, b] E) < e and I f"(x) - &) I < S for all x E E and n ? no. To accomplish this, for each k E N consider
Ak = {x : I f"(x) - f(x) I < S for all n z k}.
Now show that lim A(Ak) = 0. Here Ak = (a, b] \Ak. Complete the proof of Egorov's theorem as follows: By the
above, for each k E N, there exists a measurable set Ek and integer nk such that A(Ek) < e/2k and I f(x) - f,(x) I <
for all x E Ek and n z nk. The set E= fl Ek will have the desired properties.

1. For each n E N, let rp, = (m + (j -1)@ )XE. Then gyp, is a simple function on [a, b] with f. q,. dA = S,(f).
Furthermore, for each x E [a, b], 0 s f(x) - (p,(x) s $/n. Therefore, liimrp,(x) = f(x) for all x E [a, b]. Now apply
dAp=
the bounded convergence theorem. S. By Theorem 10.6.10(b), fF f fE f dA + fE. f dA >- fE f dA. 7. For
each n E N, let E. = {x : f(x) > 11. Then U E, = {x : f(x) > 0). Use the previous exercise to show A(E) = 0. Now
use Theorem 10.4.5. 12. The function rp, defined in the solution to Exercise 1 satisfies I f(x) - (p,(x) I < $/n for all
x E [a, b]. Thus .{9.1 converges uniformly to f on (a, b]. 15. Suppose first that rp = XA where A is a measurable
subset of [a, b]. By Exercise 2, Section 10.4, there exists an open set U ) A such that A(U \ A) < e/2. Use the set U
to show that there exists a finite number of disjoint closed intervals {J,}.N=, such that V = U J. C U and
A(U \ V) < e/2. Let h = -X ,.
j.. Then his a step function on [a, b] and {x : h(x) * rp(x)} C (U 1 V) U (U \ A). If
rp = JJ.,a;XAJ, where the Al are disjoint measurable subsets of [a, b), approximate each XA by a step function hi that
agrees with XA, except on a set of measure less than a/n.

2. (a) Assume first that A, and A2 are bounded measurable sets. For each n E N. set f, = min{f, n}. By Theorem
10.6.10(b),
( r
f,dA= J f,dA+ f,dA.
1A,UA, A, 1A,
Since each of the sequences { fA f,} ,, i = 1, 2 are monotone increasing, they converge either to a finite number, or
diverge to oo. In either case,
JAUA
= +4-+O
J
/
j of
dA = i(
A,
f,dA + J f,dA)
A,
JfdA
A.
j.fdA.
A
If either A, or A2 is unbounded, consider the integral off over (A, U A2) n[ - n, n], and use the above. 4. For
vv
f(x) = x -P, x E (0,1), f,(x) = min{f(x), n} = x'v n- o <_ x 0, lire f. f. dA = 1/(l -p).
7.05 nA(E,)5 JfdA..JfdAIbfdA<oo

UA=1E
4 o
10. Let A. = An[ -n, n). By definition, fA f dA = lim fA. f dA. Since f is integrable, given e > 0, there exists is E N
such that O :C- fA fdA - fA, fdA < e. By Theorem 10.7.4(b), fA f dA - fA f dA = fA%A. f dA. Thus E = A. is the
desired set. 13. Let f, = min{ I f I. n}. Then 0 < f, s f I. Since f is integrable, by Lebesgue's dominated
convergence theorem,
lim f, dA
1A
jIfIdA.
A
Therefore, given e > 0, there exists n E N such that fA(I f I - f,)dA < e/2. In particular, if E is any measurable
subset of A.
JIf I dA < jfdA+.

E
But if A(E) < oo, fEf dA : nA(E). Hence choose S > 0 so that nS < e/2.
19. Since {f"} is monotone increasing on A, f(x) = lim f"(x) exists, either as a finite number or as oc, for every
x E A. By Fatou's lemma, fA f dA s lim fA f dA. On the other hand, since f s f for all n E N.
lim fAf" dA t5 fA f dA. Combining the two inequalities proves the result.
21. (a) Use the monotone convergence theorem. 23. Hint: I f(x)I ? (I/x)Icot I/r=1 - 2x z x ' - 2x on each of
the intervals ((2n + 3)1r)-? s x <- ((2n - 3)rr)-'.
1. 0 < p < i 4. Since If + g 12 s 2(1 f 12 + I g I').f + g E 22(A). Assume II f + g 112 ;E 0. Then
Of +gIL= jii+gi2:5 jtf+llfi+jIf+IIi.

A A
which by the Cauchy-Schwarz inequality <- 11 f+ g Jz lI f lL + 11f+ g 16IIg II,. The result follows upon simplification.
7. Use the Cauchy-Schwarz inequality. 12. Let f(x) = I/ln x, x E (1, oo). Since f"(x) > 0 for all x E (1, ac),
f is convex on (1, oo) (see Miscellaneous Exercise 3, Chapter 5). Therefore f(lx + ,'-y) 5 1/(x) + ;-f(y) for all
x, y E (1, oo). Since n = 1(n - 1) + 11(n + I) the inequality follows.
Notation Index
CONNECTIVES AND QUANTIFIERS closed interval, 25

equivalent sets, 35
A conjunction, 496 indexed family of sets. 38
V disjunction, 496
sequence of sets. 38
negation, 496
conditional or implication, 497 U.
E.- UE.
=I union. 38
biconditional. 497 m
V for all, 516 (l e, flEe intersection. 38
3 there exists, 516
N.(P) s-neighborhood. 49
Int(E) interior of E, 94
SETS AND SET OPERATIONS set of limit points of E. 97
E'
N positive integers, 2 E closure of E. 97
N. [I,... ,n),35
Z integers, 2 FUNCTIONS
0 rational numbers, 3
f: A --i B function from A to B. 7
R real numbers, 3
n-dimensional euclidean space, 309 Domf domain of a function. 7
empty set, 2 Range f range of a function. 7
46
all subsets of A, 4 f (E) image of a set. 9
9(A).
AL sigma algebra of measurable sets, 453 f -'(H) inverse image of a set. 9
E element of, 2 Pg composition of functions, 13
0 not an element of, 2 f' inverse function, 12
AUB union. 3 ftg sum or difference of functions. 133
A f1 B intersection, 3 fg quotient of functions. 134
product of functions, 133
ACB subset, 3 fig
f*, f positive. negative parts of function. 461
AB proper subset, 3
lim f (x) limit of a function, 116
AFB relative complement, 3
complement, 3 li " f (x) limit at infinity. 126
A`
(a, b) ordered pair, 5 right, left limit at p. 148
AXB cartesian product, 5 lim f (x). lim f (X) limit superior, inferior. 162
inf A infimum of A. 22 (0 (f. p) oscillation off 162
sup A supremum of A, 22 f' derivative off 167
(a, b) open interval. 25 f+.fl right, left derivative. 168
(a, bl, [a, b) half open intervals, 25 ft^1 nth derivative. 168
544 Notation Index
B(.r. )-) Beta function. 376 Simpson's approximation. 268

Xe characteristic function. 273, 436 nt(J ) measure of an interval. 432
D. Dirichlet kernel. 405 ,n(U) measure of an open set, 433
E(r), e`. exp.t exponential function. 238 nt (K) measure of a compact set. 440
f(.r) Gamma function. 372 A,(E) inner measure of E. 444
[.r} greatest integer function. 151 outer measure of F. 444
1(x) unit jump function, 154 A(£) Lebesgue measure of E. 446
K. Fej6r kernel, 406 f, fdA. Lebesgue integral. 463
Y(f) Laplace transform. 277
!_(x), In x natural logarithm function. 234 SEQUENCES AND SERIES
c)(x). R.(-Y) remainder or error function. 362 sequence. 37
lim a limit of a sequence. 50
FUNCTION SPACES a convergence of I to a. 50
`G [a, h] continuous functions on [it. b], 333 lim lira it,, limit superior, inferior. 74
C`(!) infinitely differentiable functions
on !, 358 flat
c =.
infinite product. 90
di[a, b) Riemann integrable functions. 212 a, nth partial sum. 87
!R(a) Riemann-Stieltjes integrable tr,
functions. 251
I al series of real numbers, 87
12 square summable sequences, 306 1-1
[(at) 2 l2 norm, 306 sequence of functions. 318
norm. 310
euclidean norm, 309 series of functions, 319
inner product, 309
norm convergence. 334
II f II uniform norm. 334
Ii f I) 2' norm. 492 GREEK ALPHABET
..,f g. inner product in JtIa. bl, 381
alpha
E(A). Y'(A) Lebesgue integrable functions. 478 beta
12(A) square integrable functions. 484
chi
II f II2 f2 norm. 484 delta
epsilon
MEASURE AND INTEGRATION eta
91 partition of (a, 6), 208 gamma
IIphI norm of a partition, 227 iota
lower, upper sum. 209 kappa
f(p.f) Riemann sum, 226 lambda
f°ff,f lower, upper integral. 210 mu
no
fo1 f,f(x)dx Riemann integral, 211 omega
fa f(x)dv improper Riemann integral. 241 phi
'f(3,f,a), lower, upper Riemann-Stieltjes Pi
psi
3[(.9.f. a) sums. 246
sigma
f, "f da. f,, f da lower, upper. Riemann-Stieltjes tau
integral, 247 theta
f; f da Riemann-Stieltjes integral, 248 upsilon
midpoint approximation. 261 xi
T,(f) trapezoidal approximation. 266 zeta
Index
A Biconditional, 497 product, 314

Abel, Niels, 294 sentence, 497 sequence. 80.91. 334
Abel's partial summation formula. 294 Binary expansion, 31 Cauchy form of the remainder. 366
test, 298 Binomial theorem, 56. 366 Cauchy-Schwarz inequality. 112, 307.
theorem. 355 series, 374 384
Absolute convergence, of series. 299 Bisection, method of, 198 for integrals, 238,485
Absolute maximum (minimum), 176 Bolzano. Bernhard. 47.71. 105, 130 for series, 307
Absolute value, 45, 48 Bolzano-Weierstrass property, 104 Chain rule. 173
Absolutely integrable, 243 Bolzano-Weierstrass theorem: Change of variable theorem. 236. 237
Algebraic number, 43 for infinite sets, 71 Characteristic function. 273.436
Algebraic properties of Q, R, 20 for sequences, 71 Closed interval. 25
Alternating series test, 295 Borel, Emile, 105 in, 99
Almost everywhere. 458 Bound: set, 94
Antecedent, 497 lower, 21 Closure of a set. 97
Antiderivative, 230 upper, 21 Coefficients:
Approaches (tends to) infinity, 65, 190 Bounded above (below), 21 of Fourier series. 387.395.488
Approximate identity, 347 function, 124 of power series. 353
Approximation in the mean. 383 sequence, 51 Collection. 2
Approximation using Riemann sums, set, 21 Commutative. 20. 310.502
260 Bounded convergence theorem, 338,470 Compactness, preservation of, 136
Archimedes. 43, 277, 313 Compact set. 102, 111
Archimedean property. 28 C measure of. 440
Arithmetic-geometric mean Cantor, Georg. 1, 24, 35, 43, 90, Comparison test:
inequality, 20 99,428 for improper integrals. 244
Associative, 20,310, 502 Cantor ternary set, 108 for series. 282
Axiom of Choice, 493 function, 163 Complement of set. 3
Caratheodory, Constantin. 429, 454 Complete nonmed linear space. 334
B Cardinality. 35.44 Complete orthogonal sequence. 392
Barrow. Issac, 277 Cartesian product, 5 Completeness property of R, 24.81
Bernoulli. Jakob. 204 Cauchy, Augustin-Louis. 47, 88, 115. Complex number, 45
Bernoulli. Johann, 190, 204, 313 130, 166, 204, 207, 279, 313 absolute value of. 45
Bernoulli's inequality, 17 Cauchy condensation test, 293 conjugate of. 45
Bessel's inequality, 388, 396, 488 convergence criterion, 88. 325 imaginary part of. 45
Beta function, 376 mean value theorem. 181 real part of, 45
545
546 Index
Composition of functions, 13 0 Equivalent functions. 484

continuity of, 134 D'Alembert, Jean, 287, 313 propositions. 498
differentiability of. 173 Darboux, Jean Gaston. 208 sequences. 91
integrability of. 217 Decimal expansion, 41 sets, 35
Conclusion, 497 Decreasing function, 153 Error function, 362
Conditional, 497 sequence, 60 Error in numerical approximation. 263,
convergence, 299 Dedekind, Richard. 2. 24 282
sentence, 497 De Morgan's laws, 4, 39, 502 Euclidean distance, 49, 309
Conjunction. 496 Dense set. 29, 98 length (norm), 112, 309
Connected set. 99 Denumerable, set, 36 space, 111. 309
preservation of, 140, 143 Derivative, 167 Euler, Leonhard. 63. 130. 204
Consequent. 497 left, right, 168 Euler's constant, 293
Constant sequence, 50 second, 168 number(e). 63
Contained in, 3 Derivative of sums, products, and Even extension, 402
Continuous: quotients. 171 function, 229. 397
almost everywhere, 459 Differentiable function. 167 Existential quantifier, 516
at a point. 130, 314 Differentiation of Fourier series, 423 Exponential function, 186, 238, 369
inverse theorem, 158 of power series, 357
on a set, 130 Dini's theorem, 332 F
uniformly, 144 Dirac sequence. 347 Factorial, 18, 56
Continuous function, 130 Direct proof, 507 Family, 2
integrability of, 216, 250 Dirichlet, Peter Lejeune, 294. 380.415 Fatou's lemma, 477
Continuum hypothesis. 43 Dirichlet kernel , 405 FejEr , L.. 410, 424
Dirichlet test . 295 Fej r kernel . 407
Contractive function, 148, 335
for uniform convergence , 329 FejEr's, theorem . 410
sequence,84
Dirichlet's theorem , 417 Field, 20
Contradiction, 499, 502
Discontinuities: Fermat . Pierre de, 277
proof by , 15. 510.511
of first kind, 150 Finite expansion , 33
Contrapositive. 498, 502
jump, 150 set, 35
proof , 509
removable, 149 First derivative test , 183
Convergence:
of second kind, 150 First order method, 264
absolute, 299
simple, 150 Fixed point, 148
almost everywhere , 459
Disjoint sets. 3 Formula (in logic), 515
conditional . 299
Disjunction, 496 Fourier, Joseph. 379.426
of Fourier series, 415, 417, 420
Distance, 49, 111, 311 Fourier coefficients, 387, 395.488
of improper integrals, 239. 241 from a point to a set. 143 Fourier cosine series, 401
in the mean, 390,404, 411, 486 Distributive, 20 sine series, 401
in measure. 493 law(s), 4, 39, 502 Fourier series, 387, 395
in norm. 311. 334. 390 Divergence: differentiation of. 423
pointwise. 318 of improper integrals, 239, 241 integration of, 415
radius of. 353 of sequence, 50 pointwise convergence of. 415.417.
of sequence. 50 of series. 87, 280 420
of sequence of functions, 318 Diverges to infinity, 65. 280 Fourth order method. 268
of series. 87, 280 Domain, of function, 7 Function(s), 7
of series of functions. 319 Beta, 376
uniform, 323 E bounded,124
Converse, 498 e-neighborhood, 49, 311 Cantor ternary, 163
Convex functions, 205 Egorov's theorem, 462 characteristic. 273, 436
Cosine coefficient, 401 Element of a set. 2 composition of. 13
series. 401 Empty set, 2 continuous, 130. 314
Countable set, 36 Enumerable set, 36 contractive, 148. 335
Countability of Q, 41 Enumeration of a set, 37 convex, 205
Counterexample, 513 Equal almost everywhere, 458 decreasing, 153
Cover, open, 102 Equivalence class, 91 derivative of, 167
Index 547
difference of. 133 Q Integrable:

differentiable, 167 Gamma function, 245, 372 absolutely. 243
domain of, 7 Geometric series, 87 Lebesgue. 474. 478
even. 229. 397 Gibbs phenomenon, 426 Riemann, 211
exponential. 186. 238, 369 Graph, of function. 6 Riemann-Stieltjes, 249
Gamma. 245. 372 Greatest integer function, 151 Integral:
graph of. 6 Greatest lower bound, 22 improper, 239. 241
greatest integer, 151 property of R. 24 Lebesgue. 463, 474.478
identity, 9 Gregory, James, 318 lower (upper). 210. 247
increasing, 152 Riemann, 212
infinitely differentiable, 358 H Riemann-Stieltjes. 248
integrable. 211. 248. 474, 478 Half-closed (open) interval, 25 Integral form of the remainder. 365
inverse. 12 Harmonic series, 285 Integral test, for series, 284
jump of. 161 Heine. Eduard, 105, 428 Integration by parts. 235. 253
limit inferior of, 162 Heine-Borel-Bolzano-Weierstrass Interchange theorem:
limit of. 116 theorem, 106 relating to continuity. 331
limit superior of. 162 Heine-Bore) theorem, 105 relating to differentiation. 340
linear. 314 Hypothesis, 497 relating to integration, 337, 338.470.
Lipschitz. 145 477,481.483
measurable, 456 1 relating to limits. 330
monotone. 153 Identity function. 9 Interior. 94
natural logarithm, 234 Image (of a set), 9 point, 94
nondecreasing (increasing), 152 Improper integrals, 239, 241 of a set, 94
odd, 229, 397 convergent, 239, 241 Intermediate value theorem, 138
one-to-one, I I divergent. 239. 241 for derivatives, 184
onto, 7 Increasing function, 152 Intersection of sets. 3. 38
orthogonal, 381 sequence, 60 Interval(s), 25, 26
orthonormal. 383 Index set, 38 closed, 25
oscillation of, 162 Indexed family, 38 half-closed (open), 25
periodic, 148, 346 Indirect proof, 510 infinite, 25
periodic extension of, 396 Induction, mathematical, 15 measure of, 432
piecewise continuous, 419 Inequalities: nested. 61
polynomial, 124 arithmetic-geometric mean, 20 open, 25
product of. 133 Bernoulli, 17 Inverse function. 12
projection. 9 Bessel, 388, 396.488 continuity of, 158
quotient of, 134 Cauchy-Schwarz, 112,238. 307, 384, differentiability of. 185
range of. 7 485 Inverse function theorem. 185
rational. 135 Minkowski. 308, 485 Inverse image (of set), 9
real analytic. 355 triangle, 48, 111, 308, 310 Irrational number, 3
real-valued. 7 Infimum, 22 Isolated point. 69
sequence of, 318 property of R. 24
series of, 319 Infinite: J
simple. 464 expansion, 33 Jump discontinuity. 150
square integrable, 484 interval, 25
step. 277. 413 limit, 65, 190 L
strictly decreasing (increasing), product, 90 Lagrange, Joseph, 179, 203, 313, 363
153 series, 86 Lagrange. form of the remainder, 363
sum of. 133 set, 35 Laplace transform. 277
uniformly continuous, 144 Infinitely differentiable, 358 Largest element. 22
unit jump. 154 Inner measure, 444 Least square approximation, 384
value of, 7 Inner product, I11. 309, 380 Least upper bound. 22
Fundamental theorem of calculus, 230, Integers, 2 property of R. 24
232 Integrability theorems. 216. 220. 249. Lebesgue. Henri, 106, 219. 276. 426.
for Lebesgue integral. 473 250, 274, 462 429, 492
548 Index
Lebesgue integral, 463. 474, 478 inner, 444 Norm of partition. 227
of nonnegative function. 474 of interval. 432 Normed linear space. 310
properties of. 469.476, 479 of measurable set. 446 Nowhere differentiable function. 342
Lebesgue measure. 446 of open set. 433 Null sequence. 91
sums. 472 outer. 444 Number(s):
Lebesgue's dominated convergence zero. 219 algebraic, 43
theorem. 481 Measurable function, 456 complex, 45
Lebesguc's theorem, 220, 274 partition, 462 irrational. 3
Left continuous, 149 set. 446 natural. 2
derivative. 168 Mengoli. Pietro, 313 rational. 3
limit, 148 Mercator. Nicolaus. 317 real. 3
Legendre polynomials, 389 Mesh, of partition, 227
Leibniz, Gottfried. 115, 165. 207. 276 Metric. 111. 311 0
Leibniz's rule. 175 space, I I I Odd extension. 402
:Hospital, Marquis de. 190 Midpoint approximation, 261 function. 229. 397
Minimum: One-to-one function, II
L' l lospital's rule. 191, 194
absolute. 176 Onto function. 7
Limit:
Open:
of a function. 116 local, 176
cover. 102
inferior, 74, 162 Minkowski's inequality, 308, 485
in, 99
infinite. 65. 190 Modified principle of mathematical
interval, 25
at infinity. 126, 190 induction, 17
Open set. 94, 111
left (right). 148 Monotone convergence theorem, 483
measure of. 433
of a sequence, 50 Monotone function. 152
Order properties of S2, 21
of a sequence of functions, 318 integrability of, 216. 250
Ordered field, 21
subsequential, 67 Monotone sequence. 60
Ordered pair, 5
superior. 74. 162 convergence of. 61
equality of. 5
Limit comparison test, 282 Oresme. Nicole. 88, 313
Limit of sum, product. quotient: N Orthogonal. 380
of functions. 123 nth derivative, 168
functions. 381
of sequences, 53 nth partial sum, 87. 280, 319
Orthonormal functions. 383
Limit point, 69, 97 nth root, 29
Oscillation of a function. 162
Linear approximation. 361 nth term. 37, 50. 87 Outer measure. 444
Lipschitz condition. 145 Natural exponential function. 186, 238
function, 145 logarithm function, 234 P
Local maximum (minimum), 176 Natural numbers, 2 p-series. 285
Logarithm, natural, 234 Negation, 496 Pairwise disjoint, 98, 433
Logically valid, 501 of quantified sentence. 518 Parseval's equality. 392, 412. 491
Lower bound, 21 Neighborhood. 49. 311 Partial sum. 87, 280. 319
integral. 210, 247 Nested intervals property. 61 Partition. 208
sum, 209, 246,462 Newton, lssac, 115. 165. 197, 207, 276. measurable. 462
313. 318 norm (mesh) of. 227
M Newton's method. 197 refinement of, 210, 464
Maclaurin. Colin. 318 Nondecreasing (increasing): Periodic extension, 396
Maclaurin series, 361 function, 152 function, 148. 346
Mapping. See Function sequence.60 Piecewise continuous function, 419
Mathematical induction, 15, 17 Nonmeasurdble set. 493 Point:
Maximum: Nonnegative integers. 3 interior, 94
absolute, 176 Nonterminating expansion, 33 isolated. 69
local, 176 Norm. 310 limit, 69
Maximum element, 22 of integrable function. 492 Pointwise convergence. 318. 319
Mean-square convergence. 390 of square integrable function, 484 of Fourier series. 415. 417. 420
Mean value theorem, 179, 181 of square summable sequence, 306 Polynomial. 124
for integrals, 235.252 uniform. 334 degree of, 124
Measure: of vector, 112, 309 Positive integers, 2
of compact set, 440 Norm convergence. 311, 334. 390 rational numbers. 21
Index 549
real numbers. 20 Removable discontinuity, 149 nth term of. 37. 50

sequence, 92 Riemann, Georg Bernhard, 208, 276 nondecreasing (increasing). 60
Power series, 353 Riemann's criterion for integrability. 216 null. 91
differentiation of, 357 Riemann integral, 211 positive, 92
integration of, 371 properties of. 223 of sets, 38
radius of convergence of. 353 sum, 226 square summable, 306
uniform convergence of. 354 Riemann-Lebesgue lemma. 398 subsequence of. 67
uniqueness of. 359 Riemann-Stieltjes integral, 248 Sequence of functions. 318
Power set, 4 properties of, 250 limit of, 318
Preservation of compactness, 136 sum, 258 uniform convergence of. 323
of connectedness, 140, 143 Right continuous, 149 Sequential criterion for limits, 121
Principle of mathematical induction, 15 derivative, 168 Series. 86. 280
Projection function. 9 limit. 148 absolutely convergent. 299
Proof, 500 Rolle, Michel, 177 alternating. 295
by cases. 512 Rolle's theorem. 178 conditionally convergent. 299
by contradiction, 15, 510, 511 Root test, 288, 300 convergent. 87. 280
by contraposition, 509 Rules of inference: divergent. 87. 280
Proper subset, 3 biconditional, 504 Fourier. 387, 395
Proposition, 496 by cases, 504 of functions, 319
compound, 496 conjunction, 502 geometric. 87
negation of. 496 conjunctive simplification. 503 harmonic. 285
simple, 496 contrapositive inference. 502 Maclaurin. 361
Propositional formula, 498 disjunction, 502 nth partial sum. 87. 280
equivalent, 498 disjunctive syllogism, 504 nth term. 87
detachment, 502 power, 353
Q excluded middle, 502 rearrangements of. 301
Quadratic method, 202 hypothetical syllogism, 503 sum of. 87, 280
Quantified sentence, 516 modus ponens. 502 Taylor, 360
negation of, 518 modus tollens. 502 trigonometric, 296. 395
Quantifiers: simple replacement. 503 uniform convergence of, 323
existential. 516 transitive inference, 503 Set(s). 2
universal, 516 at most countable. 36
$ bounded, 21
R Secant line, 166 Cantor ternary. 108
Raabe's test, 293 Second derivative, 168 Cartesian product of, 5
Radius of convergence, 353 test, 188 closed. 94
Range, of function, 7 Second order method, 202. 264 closed in. 99
Raphson. Joseph. 197 Second principle of mathematical closure of, 97
Ratio test, 287, 300 induction, 17 compact. 102. 1 1 1
Rational number, 3 Sequence(s), 37.50 complement of. 3
function, 135 bounded, S I connected, 99
Real analytic function, 355 Cauchy, 80. 91, 334 contained in. 3
Real numbers, 3 constant. 50 countable. 35
order properties of, 21 contractive. 84 dense. 98
Real-valued function, 7 convergent. 50 denumerable. 36
Rearrangement of series, 301 decreasing. 60 disjoint. 3
Rearrangement theorem, 302 Dirac, 347 element of. 2
Recursive, 18 divergent. 50 empty. 2
Refinement, of partition, 210,464 divergent to infinity. 65 enumerable, 36
Relative complement, 3 equivalent. 91 enumeration of. 37
Relatively closed (open), 99 increasing. 60 equality of, 3
Remainder in Taylor's theorem, 362 limit inferior of, 74 equivalent, 35
Cauchy form, 366 limit of, 50 finite. 35
integral form, 365 limit superior of, 74 image of, 9
Lagrange form, 363 monotone, 60 indexed family of. 38
550 Index
infinite. 35 lower. 209. 246. 462 U

interior of. 94 partial, 87. 280.319 Uncountable set. 36
interior point of. 94 Riemann. 226 Uncountability of R. 41
intersection of. 3, 38 Riemann-Stieli jes. 258 Uniform convergence. 323
inverse image of, 9 of a series, 87, 280. 319 of power series. 354
isolated point of. 69 upper, 209. 246.462 of trigonometric series. 399.410
limit point of, 69. 97 Supremum. 22 Uniform continuity theorem. 146
measurable, 446 property of R. 24 Uniform norm. 334
nonmcasurable. 493 Uniformly bounded. 329
open. 94. 111 T continuous, 144
open in. 99 Tangent line. 166 differentiable. 188
pairwise disjoint, 98.433 Tautology. 499 Union of sets, 3.38
power. 4 Taylor. Brooks. 318 Uniqueness theorem for power series.
relative complement of. 3 Taylor polynomial. 360 359
sequence of, 38 series, 360
Unit jump function. 154
uncountable, 36 Taylor's theorem. 362
Universal quantilier. 516
union of. 3. 38 Terminating expansion. 33
Universe. 515
Sigma algebra. 454 Ternary expansion . 31
Upper bound. 21
Simple discontinuity. 150 function. 163
integral. 210. 247
Simple function , 464 Tests for convergence of series:
sum. 209. 246. 462
canonical representation of. 464 alternating series. 295
integral of, 465 Cauchy condensation. 293
Simpson, Thomas. 276 Cauchy criterion. 88. 325 V
Simpson's approximation, 268 comparison. 282 Value, of function. 7
rule, 267 Dirichlet, 295. 329 Vector(s). 112. 309
Sine coefficient. 401 integral. 284 norm (of), 112. 309
series. 401 limit comparison. 282 orthogonal, 380
Square integrable function. 484 Raabe, 293 Vector space. 310
summable sequence, 306 ratio, root, 287. 300
Square root. 13 of trigonometric series. 296
W
Step function . 277 . 413 Weierstrass M-test . 327
Weierstrass approximation theorem.
Stie)tjes . Thomas-Jean . 208 Tests for local maximum or minimum:
346
Strictly decreasing (increasing): first derivative, 183
M-test, 327
function. 153 second derivative , 188
Weierstrass , Karl . 71 . 93. 105 . 171 , 317 .
sequence. 60 Trapezodial approximation . 266
341 . 377
Subsequence. 67 rule, 265
Well-ordering principle . 15
Subsequential limit. 67 Triangle inequality. 48, 111. Mg, 310
Subset. 3 Trigonometric polynomial. 411
Sum: series. 296, 395 Z
Lebesgue. 472 Truth set. 516 Zero measure. 219

(MA4613&4623, MA6213&6223) Introduction To Real Analysis 2e PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(MA4613&4623, MA6213&6223) Introduction To Real Analysis 2e PDF

Uploaded by

Copyright:

Available Formats

Introduction to Real Analysis

(*, ) Manfred Stoll

1 L EI1 Pearson Education (fff ( J` I) X aa° 11 t T_ R Lh ' Wt * jf .

® iii J112 (CIP) as

q3W&*ffi4 CIPft*a*(2004) J059406'9,

2004*7)1 Z 1 NJO I11M*

A.IIji443, IIU k S MX, RX, dl It T7 R1JJ

i. J The Real Number System

1.3 Mathematical Induction 15

21 Sequences of Real Numbers

3 1 Structure of Point Sets 93

Q Limits and Continuity 115

The Riemann and 207

7J Series of Real Numbers

7.2 The Dirichlet Test 294

7.3 Absolute and Conditional Convergence 299

$ Sequences and Series of Functions

8.2 Uniform Convergence 323

9.3 Trigonometric and Fourier Series 394

Lebesgue Measure and Integration 429

APPENDIX: Logic and Proofs 495

Features New to the Second Edition

(1845-1917), both of whom published their results independently in 1872. Dedekind's

Sets and Operations on Sets

O = the set of rational numbers = { p/q : p, q E Z, q # 0}. and

AUB AnB BMA

1.1.1 THEOREM If A. B, and C are sets, then

Figure 1.2 The Cartesian Product A x B

b. Find each of the following: (A X B), (C x B), (A x B) n (C X B). (A n c) x B.

Figure 1.3 A Function as a Graph

Figure 1.4 A Function as a Mapping

(b) Let A and B be as in (a) and consider g defined by

(d) Let A be any nonempty set and let

In this example Dom p = A X B. Since p : (a, b) -* a, we denote this simply by

As was indicated in (c), if our function h is given by an equation such as

h(x) = x' + 2, Dom h = I8,

g(x)=x2+2, Domg={xEIB:-1 :5 x:5 2},

Image and Inverse Image

f(E) = {f(x) : x E E}.

f"'(y) _ {x E A : f(x) = y}.

h-'(E) ={xER:2x+3EE}= {x:_2 s x s - 1}

1.2.6 THEOREM Let f be a function from A to B. If B, and B2 are subsets of B, then

1.2.8 DEFINITION If f is a one-to-one function from A onto B. let

Figure 1.6 The Inverse Function

Figure 1.7 Composition of g with f

1.2.10 DEFINITION If f is a function from A to B and g is a function from B to C, then the

(g ofXx) = g(f(x)) = ( )'' = I + x,

(f o g)(x) =.f(g(x)) = 1 + x2, Dom f o g = R.

f-'(y)=y2- 1, Dom f'' = Range f = {y E R :Y z 0}.

(f f)(x) = f _' U W) = (f (x))2 - I = (\/j)2 - 1 = X.

e. Ax) _ + Dom f= {x E R : -1 5 x s 11 1. Ax) = sin x, Dom f = {x E R : 0!5 x :s r)

*g. f(x. y) = x. Dom f = R2

1.3 Mathematical Induction

1.3.1 THEOREM (Principle of Mathematical Induction) For each n E N, let P(n) be a

1.3.2 WELL-ORDERING PRINCIPLE Every nonempty subset of N has a smallest element.

which by the induction hypothesis

There is a second version of the principle of mathematical induction that is also

1.3.4 THEOREM (Second Principle of Mathematical Induction) For each n E N. let

f(3) = 0, f(4) = 5, f(5) = 0. f(6) =

From these values we conjecture that

n+l f( n+1)n-1 n+l IN

g' 1(2)+2(3)+ + n(n+1) n+1

* d f(1) = I f(2) = 0 and

q3W&ffi4 CIPfta*(2004) J059406'9,

20047)1 Z 1 NJO I11M