Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
IflTRODUCTIOII TO
MATtlFMATIC5
WITH
This page intentionally left blank
InTRODUCTlOn TO
M A T t l f MAT10
N E W JERSEY * L O N O O N

SINGAPORE * B E l J l N G
S H A N G H A I * HONG KONG
TAIPEI * C H E N N A I
Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA ofice: Suite 202, 1060 Main Street, River Edge, NJ 07661
UK ofice: 57 Shelton Street, Covent Garden, London WCZH 9HE
British Library CataloguinginPublication Data A catalogue record for this book is available from the British Library.
INTRODUCTION TO MATHEMATICS WITH MAPLE
Copyright 0 2004 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereoJ may not be reproduced in any form or by any means, electronic or mechanical, includingphotocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.
ISBN 98 12389318 ISBN 98 12560092 (pbk)
Printed by FuIsland Offset Printing (S) Pte Ltd, Singapore
To Narelle, Helen and Aia for their understanding and support while we were writing this book.
This page intentionally left blank
Preface
I attempted mathematics, and even went during the summer of 1828 with a private tutor to Barmouth, but I got on very slowly. The work was repugnant to me, chiefly from my not being able to see any meaning in the early steps in algebra. This impatience was very foolish, and in after years I have deeply regretted that I did not proceed far enough at least to understand something of the great leading principles of mathematics, for men thus endowed seem to have an extra sense. Charles Darwin, Autobiography (1876)
Charles Darwin wasn’t the last biologist to regret not knowing more about mathematics. Perhaps if he had started his university education at the beginning of the 21St century, instead of in 1828, and had been able to profit from using a computer algebra package, he would have found the material less forbidding. This book is about pure mathematics. Our aim is to equip the readers with understanding and sufficiently deep knowledge to enable them to use it in solving problems. We also hope that this book will help readers develop an appreciation of the intrinsic beauty of the subject! We have said that this book is about mathematics. However, we make extensive use of the computer algebra package Maple in our discussion. Many books teach pure mathematics without any reference to computers, whereas other books concentrate too heavily on computing, without explaining substantial mathematical theory. We aim for a better balance: we present material which requires deep thinking and understanding, but we also fully encourage our readers to use Maple to remove some of the laborious computations, and to experiment. To this end we include a large number of Maple examples.
vii
viii
Introduction to Mathematics with Maple
Most of the mathematical material in this book is explained in a fairly traditional manner (of course, apart from the use of Maple!). However, we depart from the traditional presentation of integral by presenting the KurzweilHenstock theory in Chapter 15.
Outline of the book
There are fifteen chapters. Each starts with a short abstract describing the content and aim of the chapter. In Chapter 1, “Introduction”, we explain the scope and guiding philosophy of the present book, and we make clear its logical structure and the role which Maple plays in the book. It is our aim to equip the readers with sufficiently deep knowledge of the material presented so they can use it in solving problems, and appreciate its inner beauty. In Chapter 2, “Sets”, we review set theoretic terminology and notation and provide the essential parts of set theory needed for use elsewhere in the book. The development here is not strictly axiomaticthat would require, by itself, a book nearly as large as this onebut gives only the most important parts of the theory. Later we discuss mathematical reasoning, and the importance of rigorous proofs in mathematics. In Chapter 3, “Functions”, we introduce relations, functions and various notations connected with functions, and study some basic concepts intimately related to functions. In Chapter 4, “Real Numbers)’, we introduce real numbers on an axiomatic basis, solve inequalities, introduce the absolute value and discuss the least upper bound axiom. In the concluding section we outline an alternative development of the real number system, starting from Peano’s axioms for natural numbers. In Chapter 5, “Mathematical Induction”, we study proof by induction and prove some important inequalities, particularly the arithmeticgeometric mean inequality. In order to employ induction for defining new objects we prove the socalled recursion theorems. Basic properties of powers with rational exponents are also established in this chapter. In Chapter 6, “Polynomials”, we introduce polynomials. Polynomial functions have always been important, if for nothing else than because, in the past, they were the only functions which could be readily evaluated. In this chapter we define polynomials as algebraic entities rather than func
Preface
ix
tions and establish the long division algorithm in an abstract setting. We also look briefly at zeros of polynomials and prove the Taylor Theorem for polynomials in a generality which cannot be obtained by using methods of calculus. In Chapter 7, “Complex Numbers”, we introduce complex numbers, that is, numbers of the form a bz where the number z satisfies z2 = 1. Mathematicians were led to complex numbers in their efforts of solving algebraic equations, that is, of the form a,xn unlxn’ a0 = 0, with the arc real numbers and n a positive integer (this problem is perhaps more widely known as finding the zeros of a polynomial). Our introduction follows the same idea although in a modern mathematical setting. Complex numbers now play important roles in physics, hydrodynamics, electromagnetic theory and electrical engineering, as well as pure mathematics. In Chapter 8, “Solving Equations”, we discuss the existence and uniqueness of solutions to various equations and show how to use Maple to find solutions. We deal mainly with polynomial equations in one unknown, but include some basic facts about systems of linear equations. In Chapter 9, “Sets Revisited”, we introduce the concept of equivalence for sets and study countable sets. We also briefly discuss the axiom of choice. In Chapter 10, “Limits of Sequences”, we introduce the idea of the limit of a sequence and prove basic theorems on limits. The concept of a limit is central to subsequent chapters of this book. The later sections are devoted to the general principle of convergence and more advanced concepts of limits superior and limits inferior of a sequence. In Chapter 11, “Series”, we introduce infinite series and prove some basic convergence theorems. We also introduce power seriesa very powerful tool in analysis. In Chapter 12, “Limits and Continuity of Functions”, we define limits of functions in terms of limits of sequences. With a function f continuous on an interval we associate the intuitive idea of the graph f being drawn without lifting the pencil from the drawing paper. The mathematical treatment of continuity starts with the definition of a function continuous at a point; this definition is given here in terms of a limit of a function at a point. We develop the theory of limits of functions, study continuous functions, and particularly functions continuous on closed bounded intervals. At the end of the chapter we touch upon the concept of limit superior and inferior of a function. In Chapter 13, “Derivatives”, we start with the informal description
+
+
+ +
Theorems.X Introduction to Mathematics with Maple of a derivative as a rate of change. we present the theory of integration introduced by the contemporary Czech mathematician J. We calculate derivatives of these functions and use these for establishing important properties of these functions. It does not have the same level of significance as a theorem. “Elementary Functions”. Note that these are set in slanting font. .1 (For illustrative purposes only!) This theorem is referred to only in the Preface. The Appendix contains some examples of Maple programs. and a general Index. outlined by a single line and with rounded corners. the book concludes with a list of References. we lay the proper foundations for the exponential and logarithmic functions. This concept is extremely important in science and applications.1 (What is mathematics?) There are almost as many definitions of what mathematics is as there are professional mathematicians living at the time. In this chapter we introduce derivatives as limits. Notes on notation Throughout the book there are a number of ways in which the reader’s attention is drawn to particular points. and for trigonometric functions and their inverses. ‘A lemma is sometimes known as an auxiliary theorem. as in Defintion 0. We also extend the Taylor Theorem from polynomials to power series and explore it for applications. as in Theorem 0. In Chapter 15. instead of the upright font used in the bulk of the book. Sometimes it is referred to as KurzweilHenstock theory. Definitions are set in the normal font. Our presentation generally follows Lee and Vfbornf (2000. and can safely be ignored when reading the rest of the book. and is usually proved separately to simplify the proof of the related theorem(s). lemmas’ and corollaries are placed inside rectangular boxes with double lines. Chapter 2). “Integrals”. Kurzweil. Finally. In Chapter 14. and are placed within rectangular boxes. an Index of Maple commands used in the book. establish their properties and use them in studying deeper properties of functions and their graphs.
e.’ instead of 0. flush with the right margin. like this. there are a number of Remarks scattered throughout the book. being an abbreviation for quod erut demonstrundum. translated from Latin. These relate to the immediately preceding text. we shall place the symbol 0 at the end of the proof.d.1 is referred to only in the Preface. There are also Examples of various kinds.. etc.e. definitions. as follows: Proof. to indicate the end of a proof. followed by the number of the theorem to which the corollary belongs then followed by the number of the individual corollary for that theorem. and are numbered consecutively (and separately) within each chapter. a special symbol is used to mark the end of a proof and the resumption of the main text.1) j u s t as helpful as the first corollary for the theorem! This is Corollaries are set in the same font as theorems and lemmas. means which was to be proved. used to illustrate a concept by providing a (usually simple) case which can show the main distinguishing points of a concept. as in Corollary 0. .d.” flush with the left margin.1 (Also for illustrative purposes only!) Since Theorem 0. Instead of saying that a proof is complete. which.1.2 (Second corollary for Theorem 0. Since the words of a proof are set in the same font as the rest of the book. This is not really a proof. or words to that effect. 0 Up to about the middle of the 20th century it was customary to use the letters ‘q. Most of the theorems. Corollaries are labeled by the number of the chapter. examples. and flush with the right margin. to help distinguish them.1. any of its corollaries can also be ignored when reading the rest of the book. To assist the reader. Remarks and Examples are set in sans serif font. Its main purpose is to illustrate the occurrence of a small hollow square. All proofs commence with the word “Proof. lemmas and corollaries in this book are provided with proofs.Preface x i The number before the decimal point in all of the above is the number of the chapter: numbers following the decimal point label the different theorems. Corollary 0. q.
but should attempt at least some. All such instances are clearly marked in the margin (see the outer margin of this page).xii Introduction to Mathematics with Maple Remark 0. Some of the exercises are challenging. We shall avoid the use of languages other than English in this book. As might be expected from the title. we must build on some previous knowledge of our readers. * * Exercises There are exercises to help readers to master the material presented.1 The first book published on calculus was Sir Isaac Newton's Philosophiae Naturalis Principia Mathernatica (Latin for The Mathematical Principles of Natural Philosophy). For instance. commonly referred to simply as Principia. We hope readers will attempt as many as possible. The idea here is to indicate readers can skip over these sections if they desire strict logical purity. rather than just reading. the beginning by scissors pointing into the book. and these are marked in the margin by the symbol 0. Exercises containing fairly important additional information.2 but then this example will be scissored. the end by scissors opening outwards. Starting with Chapter 3 we have tried to make sure that all proofs are in a strict logical order. On a few occasions we relax the logical requirements in order to illustrate some point or to help the reader place the material in a wider context. Scissors in the margin Obviously. We do not expect that readers will make an effort to solve all these challenging problems. it is obvious where the end of a Remark or an Example occurs. Chapter 1 and Chapter 2 summarise such prerequisites. Chapter 1 contains also a brief introduction to Maple. this was in Latin. are marked by 0. we might use trigonometric functions before they are properly introduced. and no special symbol is needed t o mark the return to the main text. Mathematics is learned by doing. 2Rather late in Chapter 14 . We recommend that these should be read even if no attempt is made to solve them. Since they use a different font. not included in the main body of text. and have additional spacing above and below.
We also thank those students we have taught over many years. As a small token of our love and appreciation we dedicate this book to them. . have helped us improve our teaching. who provided us with copies of Maple version 7 software. We thank the Mathematics Department at The University of Queensland for its support and resourcing. February 2004. Peter Adams K e n Smith Rudolf Vy’borny’ The University of Queensland. by their questions. Our greatest debt is to our wives. who.Preface xiii Acknowledgments The authors wish to gratefully acknowledge the support of Waterloo Maple Inc.
This page intentionally left blank .
. . . . . . . . . . . . . . 1. . . . . . . . . Algebra in Maple . . . . . .1 Basic mathematical operators . . . . . . . . . . . . .4 1. . . . . . . . . Examples of the use of Maple . . . . . .2 Useful inbuilt functions . . . . . . . . . . . . . . . . . . . . . . . Sets 33 xv . . . . . . . . . . . . . . . . . .2. . . . . . 1. . . . .7 Summary . . . 1. . . Introduction 1. . . .1 Help . . 1. .2.3 Worksheets in Maple . . . 1. . . . . . . . . . . . . . . . . . 1.5 1. . . .2 Special mathematical constants .3 Performing calculations .5 Stopping Maple . .4. 1. . . . . . . . . . . . . . . . . Introducing Maple . . . . . . . . . . . . .2. . . . . .3 1. . . . .5. . .3. . . . . . . . . . . . . . . . . . .5. . . . . . . . . . . . . vii 1 1 4 4 4 5 6 8 8 9 9 9 10 11 11 12 12 13 18 18 20 26 1. . . . . . . . 1. . . . Arithmetic in Maple .1 Assigning variables and giving names . . . . . . . . . . 1. . . . . . . . . . . . . . . . . 1. . . . . . . . . . Help and error messages with Maple . .4. . . . . . . . . .2. . . . . 1.2. . . . . . .2. . .4 Exact versus floating point numbers . . . . . .2 Starting Maple . . . . . . . . . . . . . . . . . .2. . .4. . . . . . . 1. . . . . . . . . . . . .3. . . . 1. . . . . . . .2 Ouraims . . .6 2 . . .Contents Preface 1. .1 What is Maple? . .4 Entering commands into Maple . . . . . . .2 Error messages . . . . . . . . . . . .6 Using previous results . . . . . . . . 1. . . .4.1 1. . . . . . . . 1.
75 3. . . . . . . . . . . . .6 4. . . .8 Fields . . . .. . . . . . . . . . . . . . . . . . . . . . . . . .6 3. Using Maple for solving inequalities . 77 3. . . . . . . . . . . . . . . . . . .1. .. . .5 Sets . . . . . . . . . . . . . . . . .2 4. . . . .3 2. . . 68 Functions in Maple . .2 2. . . . . . . . . . . . . . . . . . . . The least upper bound axiom .3. . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . .3:2 Defining functions in Maple . . .3. . . . . 129 135 .1. . . . . . . . . .2 Inductive reasoning . . . . . . . . . . . . . . . . . . . . . Real Numbers 4. . . . . . . .2 3. . . . . .1. . . . . . . . 89 Bijections .1 Library of functions . . . . . . . . . . . . . . . . . . . . . .3 63 3. . .3 Families of sets . . . .5 4. 91 Comments . . Comments and supplements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5. . . 63 Functions . . . Functions 3. . . Absolute value . . .5 Some common sets . . . . . . . . . . . . . . . . . Dedekind cuts . . . . . . . . . . 80 3. . . . . . . . . . . . . . . . . . . . . . Mathematical Induction 5. . .4 2. . . . . . . . . . . . . . .7 Relations . . . 2. . . Peano axioms. . . . . . . . . . . . . . . . . . . 2. . Propositions and their combinations . . . Indirect proof . . . 95 4 . . . . . . . . . . . . . . . . . .. . . . . Operation with real valued functions . . . . . . . . . .1 4. . . . . . . . Inductive sets . .3 Boolean functions . . . . . . . . . . . . .. . . .1 3. . . . . . . Supplement. . . . . . . . . . . . .1 Union. . . . . 97 97 100 105 108 112 114 120 121 129 5 . . 75 3. . . . . . . . . .4 Cartesian product of sets . . 81 Composition of functions . . 33 35 38 42 43 43 45 47 51 53 56 3. .4 Graphs of functions in Maple . .5 3. . . ... .. . . . . . . .xvi Introduction to Mathematics with Maple 2. . . . .1 Divisibility: An example of an axiomatic theory . . . . Order axioms . . . . . . . . Correct and incorrect reasoning . . 90 Inverse functions . . . . . . Aimhigh! . . . .3 4. . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . .1 5. . . . .1. 2.4 3. . . . .1. . . . .. . .2 Sets in Maple . . . . . . . . . . . . . . 2. . . . . . . . . . . . .4 4. .1 2. . . . . . . . . . . . . . . . .7 4. . . intersection and difference of sets . . . . . .
. . . .3 Maple and complex numbers . 7. . . 5. . . . . . . . . . 6. . . . . . . .6 The binomial theorem . . . . . . . . . . . . . . . . . . . .3 . . . . . . .1 Polynomial functions . . . . . . 5.1 8. . . . . . . . . . . . . . . . . . . Solving Equations General remarks . . . . . . . . . . . .6 Factorization . . . . . . . . . . . . Some important inequalities . . . . . . . . . . . .2 8.11 Roots and powers with rational exponents . . . . . . . . . . . . . . . . . Sets Revisited . . 6. Algebraic equations . . . . . . .5 The binomial equation . . . . .5 Inductive definitions . . . . . 5. . 8. . . . . . . . . . . . . . . . . . Polynomials 6. . . . . . . . .3. Comments . 8. . . . . . . .1 Field extensions . . . . .4 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . .2. Trigonometric form of a complex number . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . 5. .3. . . . . . . . . . . . . . . . . . . . . . . 6. . . .3 Long division algorithm . . . . . . . . . . . . . .2.8 5. . . . . . . . . .2 Square root of a complex number . 7. . . . . . . . . . . . . . . 7. . Maple commands solve and f solve . . . Complete induction . 7. .5 The Taylor polynomial . . . . . . . . . . . . . . 136 141 143 145 146 149 152 157 161 164 166 167 167 169 175 178 180 184 191 191 194 196 197 199 199 202 207 6 . . . . . . . . . . . . . . . . . . . . . .3 5. . . . . . 7. . .10 5. . . . . . . 8. . . . . . . . . . .4 Roots of polynomials . . . . . . . .2 Algebraic viewpoint . . . . . . 6. . .1 Sums in Maple . . . . 5. . . . . . . .4 Linear equations in several unknowns . . 7. . . . . . . . . . . .7 5. . . . . . . .1 Absolute value of a complex number . . . . . . . . . . . . . .9 5. . . . 6. . . . . . . . .2 Complex numbers . . . . . 225 . . . . Proof of the recursion theorem . . . . . . . . . . . . . . 228 231 9 . . . . . . . . 8. . . . .2 Products in Maple . Complex Numbers 7. 7.2.2. . . . . . . . . . . . . . . . . . . .2. . .1 Equations of higher orders and f solve .Contents xvii Notation for sums and products . .3. 209 207 213 . . . . . . . .4 Geometric representation of complex numbers. . . . . . . . .
.4 Continuity at a point . . . . . 13. . . . . . . . . . . . 12. .6 Comments and supplements . . . . Rearrangements . . Derivatives 313 319 322 329 332 338 353 357 13. . . . . . .7 10. . . . . . . . . . 11. . . . . . . . . . . . . . . .. . . . . . . . . . . 13. . . .. . . .7 11. . . .1 10. . .. .. . . . . . . .. . . . . . .. . .. . . . . Existence theorems . . .4 11. . . . . . . . . . . . . . . . .. . 357 362 . . .. .1 Limits .2 Basic theorems on derivatives . . . . .. .. .. . 12.. . . Limits and Continuity of Functions 12.5 10. . . . . . 12. . . . . . . .. . . . . . . 11.. . . . . . . ... Convergence tests . . .. . . . . . Basic theorems . . . . Power series .1 11.1 Limits of functions in Maple .. .1. . . . . . .. . 12. . . . . .4 Concluding comments . . . .6 10. . . . . .. .. . . .. . . . . 231 239 239 249 255 257 263 267 268 275 279 10. . Limits of sequences in Maple .. .. . . . . . . . . . 11. . . . .. . . . . 11. . .. . . . .. . . .. . . .. .8. . . . . .2 10. .. . . . . . .3 Multiplication of series. Limits of Sequences 10. . . . .. . . . . .8 . ...6 11. . . . .8.. . . . . . . . . . . . . .. . . . . . . . . .. .8... . . . . .. . . . . . . . .. . . .. . . . . .2 The Cauchy definition . . . . . . .. . . . . .. . . .. .5 11. . . . .. 11. .. . .. . . . Absolute and conditional convergence .. Muple and infinite series . . . . . . . Comments and supplements .. . .. . .. . .. . .. . . . .. . . . .. ... . . 12. . . . .. . .. . . . . . .. .. .2 Rearrangements revisited . . . .8 . . . . .. . . . . Infinite limits .. . ... . . . . .8. Subsequences .. . . . . . . .. . . . . . . . . . 279 285 . Definition of convergence . . .1 Introduction . . . . .. .. . .. .. . The concept of a limit Basic theorems . . . . .. . 12. . . . . . . . . . .. 289 290 295 297 300 303 303 306 307 309 313 .3 11. . . . . . . . . . .2 11. .xviii Introduction to Mathematics with Maple 9. .3 Infinite limits ..1 More convergence tests. .. .3 10. Series 11. .. . . . . . . . . .. . . . . .1 Equivalent sets . . . . . . . . . . . . . .. . . . . .. . 12.. . ... .4 10. . . . . . . .5 Continuity of functions on closed bounded intervals . . . Comments and supplements . . . Monotonic sequences . . . . . . .
. . . . . . . .l Muple Programming 450 455 457 458 469 477 481 488 492 498 501 501 501 Some Maple programs A. . . . .. . . . . .. . . . . . . . . . . . .12 Interchange of limit and integration . . . . . .. .. . . . . .1 Higher derivatives in Maple . . . . . . . . . . . . Appendix A A.2 The definition of the integral . . . . . . . . .. . . Integrals 15. . . . 13. Elementary Functions 14. ..4 Higher derivatives . . . . . . . . . . . . . .. .6 14. . . . . . . .Contents xix 13. . . . . . . . . . 13. . 425 430 Hyperbolic functions . . 15. . . . . . . . . . .. . . . . . . .10 The indefinite integral . . . . . . .8 Differentiation of power series . . 15. . . . . . . . . . . . . . . . .l... 14. . . .. . .4. . . . . . . . . . . .. . 407 The exponential function .. . . . 15.. . . . . . . . . . . . . . ... . . . . . . . . . 415 Trigonometric functions . . . . . . .. . . .. 15. .13 Comments and supplements . . . . . . . . 15. . . 15. . .11 Integrals over unbounded intervals . . . . . . .1 Intuitive description of the integral . .. . . 431 431 438 443 445 15. . . . . . . . . . . . . . . .2 Significance of the second derivative . . . . . . . . . . .5 Mean value theorems . . . . . . . . . . 15. . . . .. . . . . . . . . . . . .5 14. . . . . . . .. . . . . . . . .. . . . .4 BolzanoCauchy principle . . . . . . . . . 411 The general power . .. . .l Introduction . .. . . . . . . 13. . . . . .. .. . .2 14. . 15. 15. . . . .4 14. . . . . .3 14. . 408 The logarithm . . . . . . . . . . . . . . . .9 Remainder in the Taylor formula . . . . . . .. . . . .. . . . . . . . . .7 The fundamental theorem of calculus . . . . 13. . . . 13. . . . . . . .7 369 380 381 382 388 391 394 398 401 407 Introduction . . 418 Inverses to trigonometric functions. .. . .. . 15. . . . . . . . . . . .. . .5 Antiderivates and areas . . . . . . .. . . . . 15. . .8 Consequences of the fundamental theorem . . .1 Integration in Maple . . . . .4. . . . .. . . . . . . . . . . . . . . . .. 13.. . . . . . .3 Basic theorems . .. . . . . . . . . . . . . 15. . 13. . . . .9 Comments and supplements . . .. . . .. . . . . 15. . . .2. . . . .1 14. . . . . . . . 13. . . .6 Introduction to the fundamental theorem of calculus . . .6 The Bernoulli1’Hospital rule .3 Significance of the sign of derivative. . . .7 Taylor’s formula . . . . . . . . . .
. . . . . . . . .1. . . . . 504 Examples . .2 A. . . . .1. . . .2 The conditional statement . . . . 502 A. . . . . . . 506 511 513 519 References n this book Index of Maple commands used a Index . . . . . . . . .3 The while statement . .xx Introduction to Mathematics with Maple A . . . . . . . . . . . . . . . . . . .
We wish to learn how mathematics can be built. however. We place ourselves somewhere higher up in the lower half of this spectrum. We also wish to orientate the readers on the logical structure which forms the basis of this book. In modern mathematics we also use axioms but we have a different viewpoint.5. The axiomatic method is discussed later in section 2. the only means of ascertaining the validity of a statement is logic. Ideally. from the simple arithmetic operations a pupil learns in primary school to the sophisticated and difficult research which only a specialist can understand after years of long and hard postgraduate study. 1 . accepted their validity without questioning and developed all their mathematics therefrom. The ancient Greeks called these foundation stones axioms. In natural sciences the criterion of validity of a theory is experiment and practice. we try to make clear its logical structure and the role which Maple plays in the book. This can also be roughly described as where University mathematics starts. We wish to visit a small part of this cathedral of human ideas of quantities and space. However. teaching of mathematics would start with the axioms.Chapter 1 Introduction In this chapter we explain the scope and guiding philosophy of the present book. Mathematics is very different. 1 Our aims Mathematics can be compared to a cathedral. Mathematics is deductive. Experiment and practice are insufficient for establishing mathematical truth. Mathematics spans a very wide spectrum. this is hopelessly 1 . the chain of logical arguments cannot be extended indefinitely: inevitably there comes a point where we have to accept some basic propositions without proofs.
Ever since computers were invented they have been very powerful instruments for solving problems which required large scale numerical calculations. MATLAB.2x2 2x4. It can also give you an approximation of + + + 1. and the creators of Maple wanted to give it a name with a distinctive Canadian flavour. We shall aid our computations by a computer program called Maple. clearly its knowledge is useful. others include MACSYMA. Maple ‘knows’ more mathlThe notation of using scissors in the margin is explained in the preface. Some very basic facts about how to use Maple are explained in the rest of this Chapter. This book is not about Maple. however when we arrive at some point where Maple can be usefully employed we show the reader how to do it. will tell you that x6 1 divided by x2 + x 1 is x4 . The truth is that Maple was developed at the University of Waterloo in Canada. For instance. We hope that after readers finish reading this book it will be easy for them to adjust to another computer algebra system. Hence scissors in the margin1 appear only after Chapter 2. This program is supposed to be user friendly.1 with remainder 2.x 2 1 i2 2 by a polynomial of fourth degree as 1. So we patched it up. Chapter 2 contains all that the reader needs. MuMath. and Mathematica. These programs have various names. This poses a problem in that set theory should be established axiomatically and we are in no position to do so here. but not at a rigorous level. Maple is now widely used in engineering. We begin our serious work in Chapter 3. Maple. such a program can provide you with a partial quotient and remainder of division of two polynomials. set theoretical language and notation is not only convenient but desirable. We hope that at some later stage the reader will fill in this gap but we also give a warning not to do so now: axiomatic set theory is more difficult than anything we do here. in the the last decades there were programs invented that can manipulate a variety of symbolic expressions and operate on them. + . We believe that in our exposition. the name can be thought of as an acronym for MAthematics and PLEasure. Maple is one of the computer algebra systems.x3 x . if they need to. we shall call them computer algebra systems. or MAthematical Programming LanguagE. if properly asked. However. education and research. New concepts are introduced by rigorous definitions and theorems are proved.2 Introduction to Mathematics with Maple impractical at any level of instruction.
Hence readers should understand that scissors are applied automatically whenever Maple appears. computer algebra systems are only human inventions and can fail. It is therefore necessary to be careful and critical when looking at output from a computer algebra system. It can give us an answer in terms of more advanced mathematics than we know. this would be circuitous. Clearly. is clearly wrong. As an example we present a graph of the function f .1 Graph of x + sin 72 The graph. it helped solved research problems which were intractable by paper and pencil manipulation. we cannot use Maple when building our theory. but on another occasion it can be more insidious. as produced by Maple (see Figure 1. + ~ ~~ ~~ Fig. However. 1.1). Results are not necessarily correct simply . Here it is easy to spot the error. because of its irregularity. When we encounter this we shall show how to overcome it. Maple has proved itself as a very powerful tool. where f(x) = z sin7x.Introduction 3 ematics than we can ever hope of managing. It is like having a servant who is better educated than oneself. but this is not always an advantage.
1. 1 Introducing Maple What is Maple? Maple is a powerful mathematical computer program.2 1 . In particular. if you are connected to a Unix machine through . matrix operations.4 Introduction to Mathematics with Maple because they come from the computer. such as finding the derivative or integral of a function. It can do simple calculations. The use of computer in mathematics does not free anybody from the need of thinking. Maple is a mathematical computer program.2. 1. for your home PC you can obtain Maple from a scientific software supplier or from Waterloo Maple at http://www. It can also solve a variety of equations such as finding zeros of a polynomial or to solve linear as well as some nonlinear systems of equations. on a Unix or Linux machine you give the command maple after you login. Once the P C has has finished booting.2 2There is also the command maple. which will open Maple on a Unix machine without the graphical interface. Alternatively.html.maplesoft. designed to perform a wide variety of mathematical calculations and operations. However it can be useful. The differences between Maple use on a P C or on a Unix or Linux machine are minimal and we shall comment only on the essential ones. We do not recommend to use it because of the limited display.com/sales/student_iro. and double click the lefthand mouse button. At a large educational institution this will be almost automatic. you can begin your Maple session. To start Maple. for example. use the mouse to point at the Maple for Windows icon. designed to perform a variety of mathematical calculations and operations on symbolic as we11 as numeric entities. and even symbolic manipulations. be it their own PC or a large computer at an educational institution with a Unix operating system. Before you can start Maple you must be sure that Maple is installed on your machine. we presume that the readers know how to handle files and directories and do some simple editing of a file.2 Starting Maple We presume that the readers can handle basic tasks on a computer. graphing. 2 .
use the File menu. and so on). Usually. and where it has been saved (for example. To load an existing worksheet. That is. Edit. Again this is done in the standard way. If you use Linux then xmaple can be applied only if Xwindows is being used. and will not save a copy of the worksheet. when you use Maple. a window appears with a number of menus at the top (such as File. Near the top lefthand side of the screen is the Maple prompt. on a floppy disk. thus forming your worksheet. Save files from Maple in the usual way. If a worksheet has already been saved during a previous Maple session then you can load a copy of that worksheet into Maple.mws. Once the existing worksheet has been loaded. Maple opens an empty worksheet as soon as you run the program. you first need to know the name of the worksheet.2. or on the PC network). you can modify and execute commands.3 Worksheets in Maple An important concept in Maple is that of a worksheet. graphs and other pictures. Any commands you enter will be executed. and perform your choice of Maple operations. Sometimes you will not need to do this. double click the lefthand mouse button on the Maple for Windows icon. if you have not finished your work. the computer screen will display your commands (and comments). which is a > character. you may decide to save the worksheet to a floppy or hard disk on your computer. near the top of the screen. which is a I character. together with Maple output. note that the standard filename extension for a Maple worksheet file is . Most often in such a case the graphics are not available anyhow. On a Unix or Linux machine give the command xmaple. 1. However.Introduction 5 When Maple starts. you will be given the option of saving your worksheet to a file. you will be wanting to enter a new set of commands. When you have finished using Maple. By default. Of course. During a Maple session. rather than starting a new worksheet. To run Maple. Format. or if you wish to use the results at a later date. All of this material is collectively called a worksheet. you can reload an existing worksheet. and seIect the Open menu option. Next. when you have finished you may choose to save the modified a modem. followed by a flashing cursor.mw or . and modify it as required. . add comments.
If you end the command with a colon. then Maple performs the command.6 Introduction to Mathematics with Maple worksheet so you can use it again.) or a colon ( :). then Maple performs the command. ~ ~ Most Maple commands end with a semicolon or colon. but no result is displayed. Using arithmetic in Maple is discussed in detail in Section 1. it is waiting for you to type a command.2. but most of the time you will use a semicolon. Don’t forget! But if you do forget. . Nothing will happen. but you will also often need to enter your commands as text. Maple opens a new. For example. You will be asked if you wish to save your current worksheet before you exit Maple. and (almost always!) ending the command with a semicolon (. Whenever Maple displays the flashing I cursor. 1. empty worksheet by default. You can use menu items to get Maple to do the things you want. you will probably want to enter some new commands into your worksheet. and displays the result. until you enter a semicolon (on the next line). but the following examples should be fairly clear. This is done by typing your desired command. If you end the command with a semicolon. Press the Enter (or Return) key to execute the command. or reloading an existing one. you might wish to evaluate the sum 1+3 5. but you can load an existing worksheet using the File menu. type the semicolon or colon on the next line.4 Entering commands into Maple Whether you are creating a new worksheet. and press the enter key. The colon is useful for intermediate calculations. Sometimes you might forget the semicolon or colon.4. Your Maple session will appear as the following: + 1+3+5. A Maple worksheet is the name given to the text and graphics from a Maple session. followed by the enter key.
use the backspace key to correct the error.C . For example. You can interrupt execution (and return to the Maple prompt) by typing ControlC. and the result is displayed in the centre. You can also highlight something and remove it by using Cut from the Edit menu. Be particularly careful to use the correct brackets. 21° and 64 x 123. 210. + + > 1+3+5. sometim. and use the required punctuation. please note the following points: 0 0 0 0 0 0 Maple is case sensitive. Be careful to choose the correct case for your commands! Maple usually ignores spaces in your input commands. you press the c o n t r o l key. enter the following commands. If you make any mistakes while typing your command. (that is. (note the semicolon!). When using Maple commands. at the same time as pressing the C key). in the correct places.eswritten C t r l . 9 1024 7872 . Sometimes. provided you use a semicolon or colon at the end of each command. to evaluate 1 3 5. Be careful to spell the commands correctly. Maple displays your command near the lefthand side of the screen. Note that you can enter multiple commands on the same line. 64*123. a command you type may take a long time to execute.Introduction 7 9 ~~~ The command entered into Maple is 1+3+5. Most of the inbuilt Maple functions are in lowercase letters. Note the use of * for multiplication and for exponentiation. and the answer from Maple is 9.
The easiest is to select the Exit command in the File menu.5 Stopping Maple When you have finished. then 24 x 5 . select Exit from the File menu! I Using previous results Often. I 1 . 64 1. In older versions of Maple the double quote character 'I is used instead of %. 2 .8 Introduction to Mathematics with Maple If you forgot the semicolon and hit the enter key. two %% refers to the secondlast result.To find out which of the % or It to use on your machine type ?ditto at the Maple prompt. 6 To exit Maple. you can quit Maple in a number of ways. then 24 x 5 + 2 4 : 16 80 96 . you can complete the command on the next line by typing semicolon and the enter key there. you will want to use the results of immediately previous calculations in subsequent calculations. To evaluate 24.2. The percentage character % refers to the previous result. and %%% refers to the third last result.
3 . This can be done via the history function. 1 Help Maple contains a great deal of online help. You can use the ? command in a number of ways. Use the ? command for helpful information. ended with a semicolon or colon make sure the syntax. not just the last three. . Using help is discussed in detail in Section 1. but typing ?history will give you the required informat ion. you will need to use many previous results. This can be accessed via the help command. The following table summarises the more common uses (some of which will only become important after you have used Maple more extensively). 1.Introduction 9 Type %. Sometimes. Use ?name for help on the command name.2.3 Help and error messages with Maple 1 .3. You can use the help command to learn more about the history command. %% and %%% to use the results of the three most recent calculations.7 Summary In summary: double click on the Maple for Windows icon or give the command xmaple type your commands. spelling and punctuation are correct exit using Exit from the File menu 0 0 0 1. usually entered as the question mark ? character.
the expression which has been entered is not valid: we have missed out a number or a variable. These error messages are intended to be helpful: please read them. > 3*4+. . These commands all give more information about a specified function. usage. the Maple input cursor will be positioned in the command you entered. In each case. The following examples show some error messages from Maple. and think about what is being said. 1. numeric exception: division by zero. The results are surprising. A very convenient way of using help is to open the Help menu by clicking on Help located in the upper right corner of the screen.1 0 Introduction to Mathematics with Maple Command ~~ ? ?name ?index ?library ?datatypes ?expressions ?st at ement s Help Topic General help Help on the topic name. If you know exactly the name you are looking for choose Topic Search otherwise use Full Text Search from the Help menu. 3. which is (of course) impossible. at the place where Maple thinks the error occurred. we have tried to divide by 0. > 1/0. In the first case. or list all topics which begin with name An index of available help topics Standard library functions Basic data types Maple expressions Maple statements Other useful help information can be accessed via the commands info.’ unexpected. In the second case. Note that in this case. it usually displays an error message. Error. the message is reasonably simple to understand. related. Error. ’. 2 Error messages When Maple encounters an error. In the next two examples we failed to type * to indicate multiplication.
For example. unexpected number > 4(3+2). Error. 1. The following table summarises many of the common arithmetical operations.4.Introduction 11 > (3+2)4. and supports all of the usual mathematical operations. Maple follows the usual rules regarding order of calculations.4 Arithmetic in Maple 1. always try to enter your command precisely and carefully. in the second example Maple gave a wrong answer and did not issue an error message. Maple can be used for arithmetical calculations. Pay attention to the error messages! Maple tries to make them meaningful. A wrong input can return a wrong result without any warning.1 Basic mathematical operators As many of the previous examples have shown. Furthermore. However. Mat hematical Operation Mathematical example 1+3+5 953 2 x 5 Description Addition Subtraction Multiplication Division Exponentiation Factorial Square root 8/2 25 5! J25 usage 1+3+5 953 2*5 8/2 25 5! sqrt (25) Result 9 1 10 4 32 120 5 . 4 In the first instance Maple gave the correct error message. often with a pointer to the location of the problem. The moral of this is simple. multiplication is evaluated before addition. and so on.
For example. ensure that you have entered your expressions with appropriate precedence. you must be sure to: 0 0 0 0 0 include suitable brackets. Note the capital letter in P i . using a / character. 1. Two of the most important ones are shown in the following table. I Mathematical I name 7i I Description Area of the unit circle Exponential constant Maple usage Pi 1 Approximate 1 value 3.12 Introduction to Mathematics with Maple 1. sqrt (4)+5 is not the same as sqrt (4+5). whatever you want to take the square root of must be enclosed in round brackets. then the rest of the line is regarded as a comment by . use brackets. For example. For example.141592 2. you would enter (2*5+6) /(3*2). and special mathematical constants. 3x2 Always put the multiplication sign in! Always use brackets for function arguments! Be careful with precedence rules in calculations! Ensure that your brackets match up.4.3 Performing calculations When performing calculations. always use an asterisk character * for multiplication.7182818 1 ~~ e I I exp(1) I I Maple supports all of the common mathematical operations. 4*3+5 is different to 4* (3+5).2 Special mathematical constants Maple also knows a number of useful mathematical constants. To evaluate 2x&. and are correctly located! Consider the following Maple session. when using sqrt. enter divisions correctly. you must enter 2*x*sqrt (4*x). which can be used in your calculations. Note that if a # character appears anywhere in a line. match up your opening and closing brackets. To evaluate I6 . When in doubt.4.
rather than to 2. 4 Exact versus floating point numbers The Muple session given in the previous section shows that Maple evaluates 8 (2 * 6 6)/(3 * 2) to . The number of digits may be large but Muple will produce all the digits. > > > > # F o r g e t t i n g b r a c k e t s around f u n c t i o n arguments # g i v e s an e r r o r # Try t o f i n d t h e square r o o t of 4 s q r t 4. unexpected number. Error. and is ignored.66667. 3 > > # Evaluate (2*5+6) divided by (3*2) (2*5+6) / (3*2) . > > # The argument of t h e square r o o t f u n c t i o n # needs t o be placed i n parentheses # Evaluate t h e square r o o t of 4 > > sqrt(4). + . sqrt(4)+5. 4 . This is important.Introduction 13 Maple. 7 > > # Evaluate t h e square r o o t of (4+5) sqrt(4+5). You should use comments to make your calculations more readable. 1 . 2 > > # Evaluate ( t h e square r o o t of 4)+5. Muple 3 will (within the limitations of the current version of the software) give you the exact result rather than an approximation.
14 Introduction t o Mathematics with Maple > (9^9)^9. integers and fractions are exact numbers.1 are distinct entities.3 For sake of brevity we shall refer to numbers without decimal points as exact numbers. Any number which contains a decimal point is called in Maple a floating point number.and . It is important 1 to realize that for Maple. For example. 1966270504755529136180759085269121 162831034509442 1476692731541553\ 7966391196809 Note the backslash at the end of the line telling you that the number continues on the next line. . So are fi and T . Here is our first example with FP numbers. One way to do this is to type the numbers in your command with decimal points.0) .732050808 ~~ ~ 3Neither here nor anywhere else do we attempt t o explain the internal workings of Maple. 2. If only exact numbers enter a computation. . Maple will produce an exact result.~ + s q r (3. not because of 10 their appearance. but because of the way they are treated by Maple.1 into Maple you give tacitly some instructions which are absent 1 when entering .0) t .732050808 It is sufficient to write > l + s q r t (3. If you want an approximation you must ask for it. 10 > I. FP number for short. The name refers to the way in which these numbers are stored and handled by Maple. 2. If you enter .
489125293076057320 Use D i g i t s to custamise the number of significant figures given in the answer to a calculation. a similar thing can happen with any number of digits. ) .Introduction 15 However 1. This is not essential.0 +& produces a correct result but not the one you wanted. The next two examples show why exact arithmetic is preferable. Digits := 20 11.). 11. In the first example we set digits to three. You can control the precision of the decimal output via the D i g i t s command (note: capital D!). 10 > s q r t (132.7 fi .7 f i . s q r t ( 1 3 2 . which allows you to view and set the number of significant figures: > Digits. > Digits:=3: 1/(227*sqrt (2)7*sqrt (3)) . 1 22 .48912529 > Digits:=20.
99999993 The imperfection of the last two examples is not some fault in Maple's design.0)7*sqrt (3. 41.0)7*sqrt (3.'divisionby zero We did not ask for division by zero. If the above number appeared somewhere in a long calculation the process would stop. however. but the result using Maple's default precision of ten digits is correct only to seven digits. calculation with floating point numbers is not exact! > 27.92768468330373949420182744528363166911107461864767765 In the next example one would expect that after taking the root and exponentiation the original number would be returned: however. Error. We would encounter the same phenomena if we performed the calculations with paper and pencil. using the evalf 0 function.0)) .92768397 41. Increasing the number of digits helps.0)) . 1. Evalf stands .509803648 26. it is instead inherent in the nature of approximate calculations. This would not happen with exact arithmetic. within the precision of three decimals the denominator is approximately equal to zero and this causes Maple to issue an error message. you can still obtain a decimal point approximation. If all of the entries in a calculation are exact.0'(1/8). > Digits :4 0 :1/ (227*sqrt (2.16 Introduction to Mathematics with Maple > 1/(227*sqrt (2.
f r a c t i o n ) . you can use convert () to convert numbers or expressions to many different formats (such as binary. . 11. hexadecimal. > evalf ( s q r t (132) . The first argument to the evalf () function must be the calculation you want to be evaluated. > > > # Give P i t o 20 s i g n i f i c a n t d i g i t s . 11. Optionally.48912529 > > # Evaluate t h e square r o o t with 30 s i g n i f i c a n t d i g i t s .301 . > > > # Recall t h a t % g i v e s t h e r e s u l t of # t h e most r e c e n t l y e n t e r e d computation. 103993 33102 Use evalf () t o produce a floating point result. evalf ( s q r t (132) .4891252930760573197012229364 > sqrt(l32). Type ?convert.48912529 You can use the c o n v e r t 0 function to convert a FP number to an exact fraction. Use c o n v e r t 0 t o convert from one format t o another. evalf(%). and so on). 11. More generally. convert (evalf(Pi. a second argument tells Maple how many significant digits are required. for more information on this useful command.2O) . # and w r i t e t h e r e s u l t as an exact f r a c t i o n .Introduction 17 for EVALuate with Floating point arithmetic.
Maple will replace it with the expression or value which has been assigned to the variable. and then used in subsequent calculations. You can use any letters or combinations of letters (except for system constants such as P i ) as variables. you can find sums and products.1 Assigning variables and giving names The assignment operator := is used to assign values to variables. and Maple calculates an answer. expand and factorise expressions. and you can then manipulate algebraic expressions in many ways. a2. Maple has largely been used as a calculator: an expression is entered. For example. 25  27 4 > > # Use the values of a1 and a2 al*a2. This allows interim results to be calculated. a1 : = 2 5 a2 :=  27 4 > > # Look at the values of a1 and a2 al. 675 4 . However.18 Introduction to Mathematics with Maple 1.5 Algebra in Maple In the previous examples. and so on. Then whenever you use that variable. 1. ~~~ > > > # Assign the values 52 and ( 3 ' 3 ) / 4 # to the variables a1 and a2 a1 :=5^2. Maple is much more powerful than this. a2 :=3^3/4.5.
Introduction
19
Here are two more examples with assigning algebraic expressions rather than numbers.
>
u:=2*x+4;
u:=2~+4
3x+6
~
>
v:=x21;
v : = x 2 1
>
z:=xI;
z:=xl
x2  1
When a value is assigned to a variable, this value remains assigned until it is explicitly cleared. You must remember this: if you have assigned a value to a variable, each time you use that variable in a subsequent expression, the assigned value will be substituted into the expression. To clear the values of variables, you can use the restart command, which completely clears all variables (having the same result as if you exit Maple, then started it again). To clear the values of individual variables, you assign the variable its own name, with the righthand side enclosed in single quotes. It is a good habit to begin a second worksheet with the restart command, otherwise the variables from the previous worksheet would be still assigned.
~ ~ ~ ~ ~~ ~~~ ~~ ~
> >
#
Assign a value t o the variable a2 a2:=27/4;
20
Introduction to Mathematics with Maple
27 a2 := 
4
> > >
# Clear t h e variable a2, # f o r g e t t i n g its current value.
a2:=’a2’; a2 := a2
The assignment operator can be used to name almost any object which Maple recognizes. This can save a lot of typing by recalling the object by its assigned name either alone or as an argument of a command.
> p :=Pin3+5! ;
p := T 3
>
+ 120
eq: =xA2+5=p;
eq := x2
>
p01y:=xn5+x+1;
+ 5 = r3+ 120
poly := x5
>
+x + 1
 x2
f a c t o r (poly) ;
2.(
+ + 1)(2
+ 1)
Use := to assign values to variables. Use r e s t a r t to clear the values of all variables. Clear an individual variable by assigning the variable’s name to it, for example, x I : = ’ x l ’ ; . Use the assignment operator := to name mathematical objects.
1.5.2
Useful inbuilt functions
Maple provides you with a large number of very useful inbuilt functions. These can be used to manipulate expressions which you have defined. There is a command simplify. However, simplification can mean one thing on one occasion and something altogether different on another. Which is simpler,
Introduction
21
x4  1 or (x 1)(x  1)(x2 l)? Maple simplifies only when there cannot be any doubt: x is undoubtedly simpler than II: 0. Therefore the decision how to manipulate is often left with the user: Maple provides a range of options: expand, factor, normal, collect, sort and combine.
+
+
+
1.5.2.1 expand ( )
The expando function is used to distribute products over sums:
> >
Define A=x^21 and B=x^2+1 A:=x^21; B:=x^2+1;
#
A := x2  1
B := x2 + 1
> >
## Evaluate A*B
A*B;
( x2  1 )( x2
+ 1)
> >
#
Expand the previous result expand(%);
x4  1
> >
#
Expand (x21) (~^2+1)(~1)^2 expand((x"21)*(x'2+1)(xl~^2);
x4  2  x2 + 2 2
> # E x p a d (x+Y+z)2 (xyZ) 2 > expand((x+y+z)^2(xyz)^2);
4xy+4xz
> >
#
Expand (1+2x)^6 expand((l+2*x)^6) ;
22
Introduction to Mathematics with Maple
1.5.2.2 factor()
The factor0 function is used to factor a polynomial with integer, rational or algebraic number coefficients. factor0 is often used to reduce complicated expressions to a more manageable size.
> >
# Assign xn21 to
f, in other words name xn21 as f
f:=x^21;
f :z2l
> >
# Factorise f
factor(f);
(2l)(z+l)
> >
# Factorise xn4xn2+2x2
factor(xn4xn2+2*x2) ;
( z  1 ) (z3
+z2
+2)
1.5.2.3 simplify()
The simplify() function is used to write a complicated expression in a simpler form. Often Maple leaves the expression unchanged and leaves the decision how to proceed with the user.
> >
# Assign (sin x)^2 + (cos x12 to T T:=sin(x)^2+cos (x)^2;
T
> simplify(T) ;
:= sin( z ) 2
+ cos( x
) ~
Introduction
23
1
Note that the Maple command sin(x)2 is evaluated as (sin(x))^2 and not as sin( (x) 2). Here Maple follows the standard rules of precedence in evaluating functions before applying other mathematical operations. If you are in any doubt, insert additional parentheses to avoid possible errors.
> >
Assign (x+1)^3 to C C:=(x+1)^3;
#
> >
#
Assign (x1)^3 to Cl C1:=(~1)^3;
>
cCI;
> >
# Simplify the previous result
simplify(%) ;
6x2+2
1.5.2.4 normal() The normal (1 function is used to write expressions in normalised form; that is, a numerator over a common denominator:
221 x(21)
24
Introduction t o Mathematics with Maple
> # Normalise l/(xI) + x/(x1)^2 > normal(i/(xI) + x/(xI)^2) ;
221
1.5.2.5 collect ( )
The collect() function is used to collect coefficients of like powers of a specified variable; this variable must be indicated, otherwise Maple does not know what to collect. In the following example, the final expression has been collected into coefficients of the variable x:
> >
# Define f as (x+(xy)^2)^2
f :=(x+(xy)2)2;
f
> expand(f);
:= (x
+(x
 y)2)2
x2 + 2 x3  4 x2 y + 2 x y2
>
+ x4  4 x3 y + 6 x2y2  4 x y3 + y4
collect (%,XI ;
x4 + ( 2  4 y ) x3 + ( 4 y
+ 1 + 6 y2 ) x2 + ( 2 y2  4Y3 ) X
Y4
1.5.2.6 sort()
The sort (1 function is used to sort the terms of an expression into descending order of power of a specified variable.
> >
# Assign xA53x^3+x^7 to f
f:=x53*x3+xn7;
f :=x53x3+x7
> sort(f,x);
Introduction
25
x7+x53x3
> >
x^3y+x^4y^4+~^2+~^6y^2 to g g:=xn3*y+x*4*y^4+xn2+xn6*yA2;
# Assign
g := x3 y
+ x4 y4 + x2 + x6
y2
>
sort(g,y);
x4 y4
+ x6 y2 + x3 y + x2
1.5.2.7 combine()
The combine command reduces a given expression into a single term, if possible, or transforms it into a more compact form. Sometimes, similarly as with simplify, it leaves the original expression unchanged, leaving the user with the option to try another command. The following examples illustrate the use of combine.
>
combine(sqrt (12) *sqrt (3)) ;
6
>
combine(sqrt(3+sqrt(5))*sqrt(3~qrt(S)));
>
expand(%);
2
cos(2 x) The combine command reduces integer powers of sin and cos into linear expressions in terms of sin and cos of multiple arguments. The result is often more extensive than the original expression.
26
Introduction t o Mathematics with Maple
>
1 5  COS(32) 16 16 combine(cos(x)n2+sin(x)n3) ;
C O S ( ~ 2)
+
5 +8
COS(Z)
 cos(2 z)
1 2
1 1 3 +  sin(3 z) +  sin(z) 2 4 4
1.5.2.8 subs()
The subs command makes a substitution, it can be used to substitute a value or an expression. It does not simplify the result, it leaves it to the user to choose an appropriate command for further processing.
>
subs(x=l,sqrt(xn2+ll*x+24)) ;
J36
> simplify(%) ;
6
>
~~b~(~=t3,~^2+6*~+1);
(t  3)2 + 6 t  17
>
expand(%);
t2  8
combine(), expand(), factor(), simplify(), normal 0, collect (>, sort () and subs ()
are all useful for manipulating mat hematical expressions.
1.6
Examples of the use of Maple
Maple can simplify
Example 1.1 (simplify; rationalize; factor)
some quite complicated expressions
>
simplify(sqrt (2*sqrt (221)+3O)) ;
a+Im
Introduction
27
On some similar occasions Muple does not simplify, the reason being that there is another command more appropriate: for example, rationalize is one such command.
1
2 + f i
>
rationalize (%I; 2fi
~~
Another example of the same kind is:
>
simplify(2*a2+5*a*b+2*bn2);
2 a2
+ 5 a b + 2 b2
>
factor(%); (2a
+ b) (a + 2b)
Example 1.2 (solve) Maple can solve equations for you. The basic command is solve. We shall learn more about solving equations and the command solve in Chapter 8, but here is a simple example.
1 eq:=+x+l > solve(eq);
1 =1 21
La,1 + J z
We trust that the readers do not need Muple to solve such simple equations, the next example is more sophisticated. We use Maple to solve a Diophantine equation, this means that the solution must be in integers.
Example 1.3 (Diophantine equation)
equation is
An example of a Diophantine
1234567~  7654321~ = 1357924680.
The command for solving a Diophantine equation is isolve.
28
Introduction t o Mathematics with Maple
>
isolve(l234567*x765432l*y=l357924680);
( X = 4484627
+ 7654321Nl , y = 723149 + 1234567Nl }
The entry Nl indicates an arbitrary integer, there are infinitely many solutions. The amount of work with pencil and paper t o find this solution would be considerable. Maple’s answer is instantane~us.~ It is easy t o see that the Diophantine equation 3 x  2 y = nl where n is a given integer, has a solution (among infinitely many others) x = n, y = n. Maple’s answer is not very he1pfuI:
>
isolve (3*x2*y=n);
( x =  N l , n = 3 _ N l 2N2, y =  N 2 )
n = 3x
This means that if we choose x arbitrarily and y also then  2 y , which is correct but not helpful. What can we do with 12345673: 7654321y = n? As we might expect, blind use of isolve will
+
not help.
>
isolve (1234567*~765432l*y=n) ;

(x = Nl, y = N2, n = 1234567Nl
However we can solve the equation
7654321N2)
>
isolve (1234567*~765432l*v=l) ;
(U = 661375
+ 1234567Nl , u = 4100528 + 7654321Nl }
and then clearly
x
= un and y = u n
*Maple will even tell you how long it took t o solve an equation or to execute another command. See time in the Help file.
u2*v.xn2+x+l. + + + > rem(x^6+1 .x) . There are two nomials: we mentioned dividing x6 commands. it can be used with advantage.x). 5 > iquo(987654. 75973 .Introduction 29 We have seen that Maple can save a lot of tedious work. The commands are irem and iquo.1 The third argument in both commands is important as the next example shows.13) .13) . > rern(u^2+~^2.v) .~2*v. Note that three entries in brackets are required in both commands. > irem(987654. even in a situation when Maple is unable to solve the problem directly. However.x3 lx . The following Maple session shows how to use these commands. 5 v2 > rem(u^2+vA2. rem to give the remainder and quo to give the quotient. .xn2+x+1.u) . x4 . At the beginning of this chapter we said that Maple can divide poly1 by x2 x 1. 2 > quo(xn6+l.u2 5 4 Maple can also find the remainder and quotient for division of integers.
This helps with efficiency and speed of Maple. if we issue a command with(geometry1 Maple will produce a long list of various commands relating to geometry. and make these commands available. . 16. or the number of positive divisors using the command numtheoryCtau1 (note the use of [ and I). (1. [divisors] and then use the short form of the command thus > divisors (144) . To obtain the list of commands available in a package issue a command with(package) . 15 ~ Alternatively. divisors) . Using Help will tell us how to use this command to solve the problem of finding a circle which is tangent to three given circles. For instance. There are literally dozens of packages. In the dialog window which then appears fill in packages and click OK. 2. 24. If we know the name of the command and the package we can use the command on the command line. These commands reside in the main part of Maple called the kernel. 72. > numtheory[divisors] (144) . More specialized or not very frequently used commands are grouped in packages. 36. For instance. a problem formulated and solved by Apollonius from Perge in third century BC. In a similar fashion we can employ any command from the above list after the command with(geometry) has been issued. Another window opens and there you can click on index. the number theory package and the finance package. 4.package.30 Introduction to Mathematics with Maple So far all the commands we have used were available directly on the command line. 18. 8. In this section we shall use only two. 6. 3. The first command in the list is Apollonius. 144) > numtheory [tau] ( 144). 9. we can activate the command > with(numtheory. 12. 48. The list of packages can be obtained by first clicking on Help (located in the upper right corner of the screen) and choosing Topic Search from the menu which opens. we can find all positive divisors of a number using the number theory package numtheory and the command numtheory [divisors].
0. 6. He wishes to buy a house and stop payments a t 55. 165177. mortgage and annuity are the same.000.45190 Clearly. > w i t h ( f inance.06. 14385. 2. 4 (Mortgage) A 35 yearold man is confident t h a t he can save $12. In order to find out how much he can borrow he uses the command annuity from the finance package. as long as it is affordable. 48. it is far better to increase the cash payments rather than to increase the life of the mortgage. . 24. The current interest rate is 6%. 3. 8. A company advances the money for the mortgage and the consumer pays back installments. use the following commands. 36. 18. Example 1 . 16. The Help file contains details of how to adapt the annuity for a more realistic situation of monthly or fortnightly payments. 12.9738 This is a sufficient amount.20)=165000) . ~~ ~ ~~ ~ ~ > annuity(12000. However.20). paying the mortgage for ten more years is not an attractive idea.annuity) . The man decides to extend the life of the mortgage to see what happens. in an annuity.30). 144) Mathematics for finance is contained in the finance package. 72. 4. > solve(annuity(x.06. From a mathematical point of view.06. [annuity] > a~nuity(12000. . 9. in contrast. 137639.000 a year. To see by how much the installments must be increased for a mortgage of $165. when he plans to retire.0546 This amount seems insufficient for a house of the desired standard and quality.Introduction 31 (1.0. the individual makes a down payment and receives regular income from a finance institution.
039 sec]. Find the quotient and remainder for (1) 987654321.6. Solve the Diophantine equation ax by = c.6. X1O Exercise 1. X2O + 1. x9 + 2x + 7.6. 3100550. b.6.and 101ootrespectively.2 Exercise 1. 123456789.000 over a period of 40 years a t interest rate of 6.4 Find the quotient and the remainder for the polynomials (1) (2) + 52 + 7) x4 + 1. Determine how much time was needed.5 What are the monthly installments for a mortgage of $200.1 The command time(X) will tell you how much time (in seconds) was needed by Maple to execute X.32 Introduction to Mathematics with Maple Exercises @ Exercise 1. c to be 2100750.6. (2) 100000. [Hint: On the authors' machine using Maple 7 it was . + Exercise 1. Assign a.5%? . Exercise 1. 17.3 Use Maple to factorize xl' + x4 + 1.
if someone mentions the set of rational numbers greater than 1 he or she means the set of all such numbers. the M6. set of all persons who condemn this book. M3. for instance. But usually.. so we list a few sets and name them for ease of reference. University of Queensland. etc.Chapter 2 Sets In this chapter we review set theoretic terminology and notation. set of all persons who praise this book. set of all readers of this book. Table 2. At this stage the concept of a set is best illuminated by examples. in mathematics. It is customary to denote sets by capital letters like M . In these examples we have used the word ‘all’ to make it doubly clear that. 1 Sets The word set as used in mathematics means a collection of objects. the set of all rational numbers greater than 1. and this is 33 . Mz. the set of all positive even integers.1 Some examples of sets M I . the M7. 2. the M4. M I . below. the set of all buildings on the St Lucia campus of the the M5. every rational number greater than 1 belongs to M I . set of all even numbers between 1 and 5. Later we discuss mathematical reasoning.
This relation between A and B is denoted by A 2 B. whereas M I and M2 are infinite (that is. thus M7 = {4. whose elements . Thus M3 and M4 are finite. A set M is defined by specifying which elements lie in M .4}.1. This is a set which has no elements. and the other is a set with two elements. If an object x belongs to the set M . they are not finite). Also. It is also convenient to introduce the empty set. and we denote this by writing x E M . Thus in the above list it is obvious that Ad2 c M I . M7 = {2. .2.. } is irrelevant. . we say that x is an element of M or that x lies in M . Thus with a similar understanding we would say that M2 is the set of positive even integers. The symbol 8 is used to denote the empty set. and nothing else is allowed there.4.. Obviously A c A always. Thus the set A = (2. but in the world as it is this is not certain. At the first encounter the empty set may look a little strange but the introduction of the empty set has advantages.4. Two sets M and N are considered to be equal if they consist of the same elements. If all people were honest and rational it would be also true that M5 C M4 and M6 C M4. or they could be sets. x and y. .2} is the same set as Ad7 defined in Table 2.6. so that we could equally well write M7 = {4. {x.y}} has two elements..}. in the above list M4.2. If A c B and A is not equal to B we say that A is a proper subset of B. If all elements of a set A belong to a set B then we say that A is a part of B . . Furthermore. for example. any repetition of elements is ignored. . }. separated by commas. Ms and M 6 without worrying whether or not they lBut since you have read this footnote i t seems that M4 contains at least one element. It is worth pointing out that the elements in a set can be other sets. M5 and M6 need not contain any elements at all.4.2. for example.4}. one is the number 2. The essential part of the definition of M7 is that both 2 and 4 must lie inside {. We could also write 8 = {}. . whose elements are other sets. and we regard it as a finite set.34 Introduction to Mathematics with Maple automatically understood. M2 ={2.' The introduction of the empty set allows us to talk freely about M4.2. and enclose them in braces (curly brackets). This is denoted by A c B. A finite set has only a finite number of elements.8 . or that A is a subset of B. The relation A c B is often referred to as an inclusion. One way of describing a set is to list its elements. If z is not an element of M we write z @ M . These could be some mathematical entities. .. the order in which the elements are listed inside {.
Clearly M n N = N n M . We now introduce some further notation. MuN Fig. In particular cases later. intersection and diflerence of sets Given two sets M and N we can define a new set. x is a positive even integer}. for example.1 Union of two sets We define another new set by taking elements common to both M and N . consisting of all elements belonging to either M or N (or both). M2 = {x.l). Using the notation .Sets 35 contain an element. and clearly M U N =N UM. x has property P } . The set M U N is called the union of M and N (see Figure 2. where we prove that M = N we often do so by showing that M c N and N c M . denoted by M U N .2). 1 . We also agree that @ c M for any set M . If P is a certain property then the set of all elements having the property P is denoted by {x. 2 . x E M 2 . M7 = {x. 1 Union. 1 < x < 5). these form a set called the intersection of M and N which is denoted by M n N (see Figure 2. If M c N and N c M then clearly M and N consist of the same elements and hence M = N . 2.
unless M = N .2 Intersection of two sets Yet another set which can be formed from two sets is the difference. If M n N = 8 then we say that M and N are disjoint.3 Difference of two sets . If M and N are sets then the difference M \ N is. the set of elements lying in M but not in N . It is clear that M \ N # N \ M . x 6 N } . Fig. that is. M \ N = {x. x E M .36 Introduction t o Mathematics with Maple introduced above we have M U N = {x. in which case M \ N = 8. 2. This is illustrated in Figure 2. x E M or x E N } and M n N = {x. 2.3. Fig. by definition. x E M and x E N } .
The i=l number of sets so that u 00 notation is extended to refer to an infinite Ci is the set of those elements contained in at 00 i=l 1 least one of the sets C i C2?C. N .1. M } . together with 0 and 1. And if Dk is the set of integers less than or equal to I% then the intersection 0Dk consists of the set of negative k=l integers. . .1. For example. 2)}. It is easy to see that each of the new sets A. The set of subsets of A is called the power set of A .C . D may be denoted simply by A n B n C n D. i=l . [Hint: Use Help to find the command combinat [powerset J . . A\B = { x . for example. nAi. For example. 3. The corresponding ideas for 5 sets. xEAorxEB}.2 to prove ( A nB ) U ( A \ B ) = A . b. {I}. .B. Hence ( L U M ) U N = L U ( M U N ) and we write simply L U M U N .}. 3. and (7 Dk is the set of those elements k=l contained in all the sets Dk. M . u 7 Bi is the union of the seven sets B2. that is. if Ci is the set of positive proper fractions with denominator i then the union 00 u 00 Ci is the i=l set of rational numbers between 0 and 1. k= 1 Exercises Exercise 2. x E A. or any finite number. 2. The same convention is used for intersections.. AnB={x. . . 2 . M and N we can form two new sets A = ( L u M ) u N and B = L u ( M U N ) .Sets 37 Given three sets L . presents no new difficulties but some compact notation is needed. the power set of (1. .] .k = 1. 0. x 4 B}. {2}. . for B we form first the union M U N = Q and then use Q to form L U Q. 3. a. 2. . . 2 ) is {{}. 4 and in the same way. . (1.B consists of those elements belonging to at least one of the sets L .1.1 Use the definitions AUB={x. Use Maple to create the power set of (1. xEAandxEB}. @ Exercise 2. nDk 00 = (1. the intersection of four sets A. . Thus A1 n A2 n A3 nA4 is denoted by B1. . or 25. B7. For example.
2. 61.x} minus {I}).37. 1 .31.61. respectively. 43.29.47. 47. 8. { 19.b} union {x}). 37.37. B > := (23. 41. (23. 53. 37. 31. 37. 41. {31.41. 47. 59. 49. 2 Sets i n Maple Notation for sets in Maple is the same as in set theory. 29. . cannot be entered from the keyboard).67. 41. 29. 53. 49.53. Instead of the symbols U. 23.38 Introduction to Mathematics with Maple 2 . 25.71}. 67. Hence A > := (19. 55. 43) > A minus B. 71) > ({a. 71) A union B. 59. 47. 67. 55) B:={23. n. 25. intersect. 43. 25. 31. ({1.59. 67. 55) > B minus A . 29. 71) > A intersect B. 31. 61. 53. \ we write union.49. (19. 61. 43. 59.43. minus. The empty set is denoted by {I (the common mathematical symbol for the empty set.
x!. ({I} union (1.z] . set. The following are some examples of using sets. lists and # expression sequences.Z :=I. A := { a . consequently > ({I}intersect {Io}). it is a string of characters which can be entered into Maple.y . We know that Maple makes a distinction. z . 11 are all distinct. Sets.c}. > > > > # First we name a few sets. lists and expression sequences in Maple.2. A:={a. lists and expression sequences are all similar: the difference lies in the way Maple handles them. [2. Kai !Hy%.0. B:={john.b. An expression sequence is an ordered ntuple of expressions separated by commas. 0 2. lasts An expression is a basic entity in Maple. 1.0)). c} .Kathy}. For instance j o h .1. 2 and [I.1 Expression sequences. roughly speaking. b.b. a. An expression sequence enclosed in square brackets constitutes a list. 1 is an expression sequence. x+l/x. 2. You can combine two lists into a new list in which the first element of the second list follows the last element of the first list.3! . Order and repetition 1 . You can make an intersection of two sets but not of two lists.z . I.Sets 39 In ordinary life there is no distinction between 1 and 1. I. C := [x. 13 of expressions in lists or expression sequences do matter: [I.
# > > > > # the nops command counts the number of elements in a list or a set. 2 . john.x!.nops (C) . > > > # enclosing an expression sequence in appropriate symbols converts the sequence to a list or set.y.convert(C. Kathy} c := [x. 1. # # 4 5 . 2 .40 Introduction to Mathematics with Maple B := (6. a .list) . 21 Z := 1. # respectively. nops (B) .set). b > > > # the command convert changes a set into a list # or viceversa convert(A. but cannot be applied to an expression sequence.
64. 27. or some chosen ## elements from a list or a set.seq(nA2+n+7 .5) .4. Note that a set does not have an nth element. op(2.B[11 . AC21 . respectively. x!. [op(A).A) has the same effect as A [ 2 ] . and this need not agree # with what you might guess is the nth element. john. # 8. Kathy [a.3) . seq(i3 . # It can extract all elements. b.4.i=2. op(3. 125 .Z[31. 6 .john. . op(B). ## Some examples on the use of op follow... c. # However. # Maple will return a result.B)1. giving what it thinks # is the nth element. the op command is more powerful # and more flexible. Kathy] > > > The command seq can be used to generate # an expression sequence according to some formula.C). . b 6 1 # The command op(2.n=O.Sets 41 # the nth element can be extracted from a # list or an expression sequence by # attaching [ n ] to the list ## # or expression sequence.
Similarly. A family of sets can be indexed by any set. the set of all u 1EL elements belonging to at least one set Cl. If {Cl. A1 is the set of points on the straight line between the points (1.3. Use nops!] 2. . (1.5 Find the number of elements i n the power set of A. (13) Exercises 3. 19 > {seq(nn2+n+7.1. not necessarily by a subset of positive integers. [Hint: See Exercise 2. where A has 17 elements.1. Later we shall encounter situations where to each element l of a set L there is associated a set CZ.1.3 Families of sets Instead of saying that A is a set of sets we prefer to say that A is a system or a family of sets. that is.1. It is natural to denote this indexed family by {Cl. 13. Exercise 2. 0) and ( I . for instance.n=1. 1) and (0.l E L } is an indexed family then by Cl we understand the union of the family Cl. 30)). 1).4] into Combine the lists L = [l.3 a single list using the op command. Exercise 2.1.42 Introduction to Mathematics with Maple 7. 0). Exercise 2. In such a situation we say that the family of sets Cl is indexed by the set L.n=l.2. 0). If. the intersection A 1 is by definition the set of all elements n IEL belonging to every set Al. 1) in the plane and L is the set of real numbers between 0 and 1 then Al is the interior u 1EL of the square with vertices (0.4 Find the number of distinct elements in a list by first converting it to a set.2.l E L } .5] and M = [I.30)}intersect{seq(nn2+n+ll. . (1. 9.
x > 0 ) . . In the same way the idea of a Cartesian product of four or any finite number of sets should be clear. However this is not strictly correct since ( A x B ) x C is a set of pairs of things in which the first thing is a pair of elements and the second thing is a single element. For ease of reference we name here some subsets of reals as follows. Z= ( 0 . 3. at least on an intuitive basis. and for A x A x .) is systematically reviewed in Chapter 4. NO is the set of nonnegative integers. . q ) . 3. N is the set of natural numbers. d) are defined to be equal if and only if a = c and b = d. .1. 1. E Q= E. 2. 2.1. using terminology introduced so far.x A with n factors we write An. b ) . we have A x B = { ( a . (a.4 Cartesian product of sets If A and B are sets then the set of ordered pairs of elements. b) can be written (a. ( ( a . b).( a . It is natural to define A XB x C as the set of ordered triplets (a.2 the first from A and the second from B . . Z. One would perhaps expect that A x B x C = ( A x B ) x C = A x ( B x C). 2. Expressing this in symbols. For A x A we write A 2 . 1. c) with a E A . b) = (a. rationals.}. N o = ( 0 . b E B}. b. x E R. b). 3. Q is the set of rationals. . c) is its typical member. In this book we shall not make this fine distinction between ( A x B ) x C or A x ( B x C) and A x B x C. 2. for example. }. In fact this equation can be used t o define an ordered pair. P = {x. b) and (c. would read Q = { ( p . . Two ordered pairs ( a . The set of all real numbers will be denoted consistently by R. b E B and c E C.Sets 43 2. p {Q 2Using set notation.. c). the ordered pair (a.}. a E A . Z is the set of integers.3 P is the set of positive reals. familiar with real numbers. . 1. N = (1. b} }. q EN } . Warning: Realise that A x B generally is not equal to B x A . 2. whereas a typical member of A x B x C is a triplet ( a .. . is called the Cartesian product of A and B and denoted by A x B. 3. the definition of integers. q E N}. b. . p E Z. 3A more precise formulation of the definition of Q. The theory of real numbers (including.5 Some common sets We assume that the reader is. etc.
x < l } . (4) A U A = A . Prove the correct ones and give examples (strictly.x E 00 00 R. An.44 Introduction to Mathematics with Maple Exercises Exercise 2. B 1 = (2.) . For the sets listed in Table 2.1.9 Let An = {x.1 show that M I U M3nM4=O. N . Prove that. if L is an index set. x > O}. L = {x : x E Iw.1.1.8 In this exercise A. (3) ( A \ B ) n ( A \ C ) = A \ ( B u C ) . ( 5 ) ( A n C ) u B =( A u B ) n ( C u B ) . Exercise 2.1.12 Parts (2) and (3) of Exercise 2. ( 2 ) ( A\ B ) u ( A \ C ) = A \ ( B n C ) .10 u n u Bz.11 (1) ( A \ B ) n C = ( A n C ) \ ( B n C ) . Discover which of the following relations are correct. then (These relations are known as de Morgan’s Rules. Exercise 2. x a multiple of n}.11 can be generalized to families of sets. x E N. (3) A u ( B n C ) = ( A u B ) n ( A u C ) . B if AUB = AnB? Prove that @ Exercise 2. M2 Exercise 2. and X is any set.1. @ Exercise 2. Find Exercise 2. Al a family indexed by L .1. An. C are arbitrary sets. A n A = A .7 cN = Mi. P . (1) A n ( B u C ) = ( A n B ) u ( A n C ) . (2) ( A u B ) n C =( A n C ) u ( B u C ) .1.1. Q such that A 4 and neither P c Q nor Q c P. B .nB1.6 Give examples of sets M . n=2 n=2 EL EL What can you say about the sets A. counterexamples) to show that the others are false.
.’. giving that is. Sometimes a mathematical theorem is not expressed in the form ‘ I f ..b and overlooked the fact that c . Care is often needed. We appear to have proved that an arbitrary real number c must be zero. (C + b)(c . (2. Let b be an arbitrary real number.Sets 45 @ Exercise 2.. As this is absurd we must look for an explanation.b we have and therefore c = 0.13 L. 1EL 1EL 1EL Also show by an example that the inclusion cannot.2 Correct and incorrect reasoning The ability to reason correctly is essential in mathematics.1. The moral of the above example is that when applying a theorem one must make sure that the hypothesis is satisfied.. Consider the previous theorem again. The part of the sentence starting with ‘If’ is called the hypothesis (or assumption(s) or condition(s)) and the phrase starting with ‘then’ is called the conclusion (or assertion(s)).1) and hence. dividing by c . and involved an incorrect application of the following theorem.. but even with such a variation of form the hypothesis can be identified. We applied this for x = c . then . then . ’.1) to Equation (2. c = b. that is.b = 0. in general. 2.2). For example the theorem .b) = b(c  b). . It is of the form: ‘If . The error occurred when passing from Equation (2. and let c be the same number. Multiply this equation by c and subtract b2. be replaced by equality. . as the following fallacious argument shows. Prove that Let A1 and €31 be two families of sets indexed by a set U A E \ U B l c U(Al\Bl). If yx = xx and z # 0 then y = 2. .
and thus combine both theorems into one statement. Clearly S says something quite different from E. c satisfy a2 b2 = c2 then the triangle + is rightangled and c is its hypotenuse. which has a hypothesis and a conclusion. The next example is different. C: If the sides of a triangle a. can be rephrased to exhibit the hypothesis and the conclusion more clearly as T:If n is an integer then the number n(n + 1) is divisible by 2. The theorems P and C below are converses of each other. E: If an integer is divisible by six then it is an even number. In the above example it happens that both theorems are true. b. b. the converses of many important theorems are true. S: If an integer is even then it is divisible by six. For example. then one must prove it. In such cases instead of stating the two separate theorems we use the phrase ‘if and only if’. It is wrong to assume that the converse holds simply because the original theorem did.46 Introduction t o Mathematics with Maple T: The product of two consecutive integers is divisible by two. The converse is different in content from the original theorem. If one suspects that the converse is true and wishes to use it. b then a2 + b2 = c2. If these are interchanged a new theorem is obtained. P: If a rightangled triangle has hypotenuse c and other sides a . Summary: A piece of mathematical knowledge is usually stated in the form of a theorem. + The abbreviation ‘iff’ is sometimes used for ‘if and only if’. . moreover S is false while E is true. P and C together read PC: The triangle with sides a . which is called the conuerse of the original theorem. Actually. If the conclusion and the hypothesis are interchanged a new theorem is obtained. c is rightangled with hypotenuse c if and only if a2 b2 = c2.
then (1) The diagonals of a rhombus are perpendicular. 111.2 false. Give an example of a valid theorem whose converse is Exercise 2. Exercise 2.2. . But what are the ‘things’ in question? The first point is that they cannot be any phrase or sentence (in English) like (Hooray!’. and neither is B. ‘Go home!’. IV.3 Give an example of a true theorem with a true converse. The typical structure of a theorem has already been mentioned: ‘If something. If A and B are propositions. is not true.1 State the following theorems in the form ‘ I f .2. We shall call such sentences propositions. . A A A A is true. .2. and there are four possible cases in relation to the truth of both (see Table 2. Use the phrase ‘if and only if’ to combine both theorems into one statement.’ They must be capable of being true or false (but not both). Table 2. We now consider them more carefully. ‘The colour seven is tropical.2). but B is not. then each may be true or false.2 Truth and falsity of propositions I. 2. 11. but B is. then some other things’. (3) Every two positive integers have a greatest common divisor. (2) Every algebraic equation of degree one or higher has a solution.Sets 47 Exercises Exercise 2. is true.3 Propositions and their combinations What has been said in the last section is really only a vague introduction to certain ideas. (HOW are you?’. Consider the following example. (4) Grandfather’s knee pains whenever it rains. and so is B . is not true.
if D is the proposition the number six is even then 1 D is the number six is not even. For example. if John has all top grades he passes (so A + B ) .48 Introduction t o Mathematics with Maple A: John. The reader will remember from Section 2. B: John. For example. our interpretation of A + B means that for a false proposition A the implication A + B is true no matter what proposition B is. In our example. This is clear. The negation of a proposition C is a proposition which is true if C is false and false if C is true. Consequently. If A and B are propositions. achieved top grades. The last combination of two propositions A and B we consider is A or . the engineering student. Implication is an example of combining two propositions into a new one. we say that A implies B . if C means every blonde girl has a handsome suitor then 4 ’means it is not true that every blonde girl has a handsome suitor or. and case I1 does not occur. and we denote this by A + B. For example. so B + A is not true. has passed the year. The success of many proofs hinges upon a correct formulation of the proposition 7 C corresponding to a given C. By forming an implication A + B we combine two propositions A and B into a new one. the case I1 does not occur (neither does I). For instance. Such a relation between two propositions is called an implication. i f A then B. here A is the hypothesis and B is the conclusion. it is time to reflect and realize that in ordinary life many people do not understand implication the way mathematicians do.2 does not occur. or that B follows f r o m A . In this example case I1 cannot occur. since if A is a false proposition and B is any proposition. if A is 16 is even and B is 16 is positive then A and B means 16 is even and positive. the engineering student. and so is if 3 is less than 1 then not all numbers are equal. There are many ways of forming new propositions.2 that when A B then B need not imply A . The negation of C is usually denoted by not C or 1C. i f 3 is less than 1 then all numbers are equal is a true implication. This new proposition is true if case I1 in Table 2. Given two propositions A and B we can form a new proposition A and B. We shall agree that the proposition A and B is true if both propositions A and B are true. there is at least one blonde girl who either has n o suitor or her suitor is not handsome. We denote the proposition A and B by A A B. After having made precise the meaning of the implication A + B . We may express the implication A + B in other words by saying If A. in other words. but if he passes he need not have all top grades. then B.
2 does not occur. . y. t e a or coffee but not both’. We denote the proposition A or B by AVB.2 like this: For real numbers x.3) holds (that is. there is a snag in (2. z the implication (zy = z z and x # 0) J y = z (2. It is taken to mean that a t least one of the propositions A and B holds. Using the terminology from Section 2. We shall say that (2. some people may understand the proposition A (They would say ‘either A or B or both’ to cover the or B differentl~. for example. Clearly. Similarly. case I1 occurs.2 does not occur. the equivalence A H B is often proved by proving both A + B and 1 A + 1 B . for concise and convenient In this book we shall use the signs 3 or recordings of many theorems and mathematical statements. The relation A e B is called equivalence. y. but will not be used in this book.4).3) or (2.2 we say that A holds if and only if B’holds. For instance we can state the theorem from Section 2. the implication is true). In order to circumvent this difficulty we specify the precise meaning of (2. and A or B is true (by our convention). In mathematical logic this is known as the exclusive or. Unfortunately. case IV does not occur.4) (or a similar ‘implication’.) If A is three is greater than zero and B is three is equal to zero then A V B is three is greater than or equal to zero and is a true proposition. this is because it is not possible to decide whether or not z > 1 is true unless we know what x actually is.4). in this example.~ meaning of our ‘ A or B’. (2.Sets 49 B. Sometimes in statements like (2. . that is to say case IV in Table 2. in ordinary life. Instead of A or B we may sometimes say either A or B. Again.3) ) holds (is true) if it becomes a true implication after substitution of an arbitrary real number into (2.3) ). Strictly speaking x > 1 is not a proposition (in terms of our earlier definition). If A J B and B + A we say that A and B are equivalent and denote this by writing A H B.4) This certainly looks like a true implication.4) we shall substitute only natural *As in the question ‘Would you like tea or coffee?’ The implication here is ‘. .4) (or numbers z. z into (2. (2. Another example of this kind is x > 1J X > 0. The truth of the implication A + B is often proved by showing that 4 3 + A since both implications mean that case I1 of Table 2.
If we substitute either x = 1/2 or x = 1 into (2. (3) [ ( A B ) A ( B C ) ]+ ( A H C ) . and verify that the resulting implication is true in each case.4) then the hypothesis is false and the conclusion is true in the first case and false in the second.3 Prove that ( A @ B ) if and only if ( B H A). (4) B 3 A. but it will be clear from the context what substitutions are envisaged.6 Verify that the following implications are correct. (2) 1 A 3 4 3 .1 Let A be 2 is odd and B be 3 is odd.3. Exercise 2.) Exercise 2. (3) x = 0.4 Prove that for positive numbers a and b. (3) A + +. @ Exercise 2.3. y = z = 2. y = 2. (1) A + B.3.2 In implication (2. Exercise 2. however. * * * Exercise 2.3) in the text substitute (1) x = 1. stays always true. N are sets prove .2. ( 2 ) ’ ( A A B ) (1A) V (1B). ( u 5 b) (u2 5 b2) by showing ( u 5 b) + (u2 5 b2) and l ( u 5 b) 3 i ( u 2 5 b2). Now we are in a better position to appreciate the convention that false A + any B. * (1) A 3 (AVB).5 Explain why the word ‘implication’appears in quotation marks in one place on page 49.3.3. Decide whether or not the following implications hold.3. Exercises Exercise 2. (Inequalities are systematically treated in Chapter 4.7 If M . (2) x = 2. y = 3.3. z = 3. z = 9.50 Introduction t o Mathematics with Maple numbers or elements of a certain set. Exercise 2. Section 4. The implication.
8 If A and B are propositions state the negation of A V B and A + B.) Exercise 2. We briefly summarise the prerequisites for the proof. (1) There is an integer x such that for all integers y the equation xy = y holds. (2) M U N C N H M C N . This is known as indirect prooj In philosophical debate the method of indirect proof is known as reductio ad absurdum. 2. A defence lawyer can show the innocence of the accused by showing that the assumption that the accused committed the crime leads to the inescapable conclusion that the accused was at two distant places at the same timeobviously an untenable proposition.Sets 51 (1) ( M \ N ) U N = N H M c N . in words. (2) For all integers y there is an integer x such that the equation x y = y holds. This is used in proofs. (Example: the negation of A A B is 1 A V 1B. If A . The argument appears in one of Euclid’s books (written approximately 300 B.). since quite often the simplest way of proving that proposition A holds is to show that 1 A implies something manifestly wrong.2 is possible (since I1 is ruled out by the truth of A + B ) and consequently A is false.3. (3) For all integers y there is an integer x such that the equation x y = y2 holds. The next example is a classic example of an indirect proof and goes back to the ancient Greeks. the negations of these propositions. If someone says ‘If I were rich I would buy a new house’.4 Indirect proof + B and B is false then only case IV in Table 2. (3) ( M \ N ) U ( N \ M ) = 0 M = N . A theorem from elementary number theory says that every positive integer greater than or equal to 2 is divisible by some prime (possibly itself).9 Decide which of the following propositions are true and which are false. if 1 A implies something false then 1 A is false. * Exercise 2. there is a clear implication that the person is not rich.3. that is. A prime is an integer greater than or equal to 2 which has no divisors except 1 and itself. (4) There is an integer x such that for all integers y the equation xy = y2 holds. By the same reasoning.C. Also state. A is true.
3 Prove that for every prime p integer n such that either p = 6n . one of the pi. . the proposition that 1 is divisible by a prime (greater than 1) is obviously absurd. pa. For an indirect proof we assume the contrary.. In mathematics. However. Consequently 1 is divisible by pi since 1 = N ..p3 . 2 5 there exists a positive Q. the better’ for as soon as we reach real absurdity we have attained our goal (of course. that is.p3 . We have reached our goal: the theorem in italics has been proved. we often encounter situations where an indirect proof requires a number of fairly difficult steps and students are apt to start worrying in the middle of the proof because things look a little strange. however.* p n and both N and p l .b # 0 then a + b J Z is .4.52 Introduction to Mathematics with Maple Now we prove: There are infinitely m a n y primes. .4. If this happens they should realise that in an indirect proof ‘the stranger..4 irrational.4. Prove that if a E + 1. We have just given a simple example of a short indirect proof. Exercise 2.p2 p3 . p3.p n are divisible by pi (to make this rigorous we would also need to show that if a and b are divisible by p i . this is divisible by some prime. assuming that we didn’t make any errors). . .1 or p = 6n Exercise 2.p n 1.2 Prove that every prime other than 2 is odd. Now consider the integer formed by multiplying these together and adding 1: N = p l pa .b). then so is a . Exercise 2.p l .p2 . +  Exercises Exercise 2...b E Q.4. Then we can list all the primes as p l . ..1 Prove that each of the following numbers is not rational. p n .
If X is a set we may ask whether or not { X } E X . On the other hand it seems that the family F of all singletons has the property that { F } E F because { F } is a singleton.Sets 53 2. Consequently the primitive terms in the axiomatic system cannot be defined. then we simply agree to call the objects which satisfy the axioms sets. Any reasonable theory must be free of any paradox. named after its inventor. This seems innocent enough. if { Y }E Y then { Y } 4 Y by the defining property of the set Y . which is absurd. It can be objected that 4 4 4 . the socalled Russell’s Purudox. Unfortunately we have a paradox on our hands. Now we give an example showing the inadequacy of such a simple intuitive approach. If we set up an axiomatic system. it was only illustrated by examples. Indeed. On the other hand. This example is a slight modification of a famous paradox. In either case we have both { Y }E Y and { Y } Y .1.) Perhaps we may say dogmatically that constructions like the set of all singletons or the set of all sets are not legitimate in set theory. A set having only one element is called a singleton. Since set theory is most basic we have no choice but to build it axiomatically. Now define a set Y by Y = { { X } . we have { Y }E Y . The question of the truth of axioms in the modern interpretation of axiomatic theories simply does not arise.X } . Any mathematical discipline should be based on an already established branch of mathematics or founded axiomatically. say for set theory. Clearly. the British mathematician. Some may feel that axioms must be simple and selfevident. circular definition must be avoided. what is obvious and simple to a mathematician who has absorbed a great deal of knowledge accumulated over more than two millennia need not be obvious or simple to everybody. logician and philosopher Bertrand Russell. An axiomatically based discipline starts with a few basic propositions axiomswhich are not questioned for truth or validity.5 Comments and supplements The concept of a set was not introduced precisely in Section 2. Firstly. { X } f. However. How are we to get rid of the paradoxes of set theory? (There are others besides Russell’s. If z is an object then by {x} we denote the set consisting of one element x. if { Y } Y then again. and develops from these axioms by logical means. When using a property defining a set it is safe to consider only subsets of a set which we know (like the reals) or which can be constructed from known sets (like the Cartesian product R x R) . by the definition of Y . if X = 8 then { X } X because nothing is in X . Neither { Y } E Y nor { Y } 4 Y .
World Scientific Publishing. An interesting application of sets and axiomatic ideas was developed by the economist Kenneth Arrow. Lecture Thirty Five). is certainly not meaningless. Then these theorems are applicable to any objects which satisfy the axioms. A and 1 A . Communications in Pure and Applied Mathematics. because of its applicability to other sciences. in part for this work.5 To leave the primitive terms undefined has its advantage in that the undefined terms are capable of having several meanings. Wigner: The Unreasonable Effectiveness of Mathematics in the Natural Sciences. Imagine that we have proved some theorems on divisibility of integers as logical consequences of a few basic axioms. It may be a puzzle why mathematics is at all applicable but this is a rather philosophical question which we will leave aside. 1990. edited by Ronald E.54 Introduction t o Mathematics with Maple with such an interpretation axiomatic mathematics becomes a meaningless game. For example. The contributors to this book discuss aspects of Wigner’s paper. A fundamentally important requirement of any system of axioms is its consistency. vol.1 An example of an axiomatic theory which does not require any prerequisites is presented in Eves (1981. or the book Mathematics and Science. Most mathematicians now agree that in principle the only requirement a set of axioms must satisfy is consistency. how do the individual preferences of people for breakfast cereal or clothes styles relate to the cereals or clothes styles most in demand in society. This actually happens with objects called Gaussian integers. We show this in Subsection 2. The desire for an easy verification that a certain set of objects satisfies a list of axioms leads to the requirement of having as few axioms as possiblewithin reasonable limits: we don’t want to have so few axioms that the system is too weak to do any significant amount of mathematics. Mathematics. the system of axioms is called independent. pp. . XI11 (1960). Arrow became famous. An inconsistent system is obviously useless for any serious study. This is of considerable interest to economists and manufacturers. It was realised that his work had much wider implications than just economicsit applies 5The interested reader is referred to a famous paper by Eugene P. and also polynomials with real or rational coefficients. He was interested in the way the preferences of individual members in a society could be combined to produce an overall preference for that society. for example. Mickens. and received a Nobel Prize in Economics in 1972. 114. A system of axioms is consistent if it is impossible to prove from it two propositions contradicting one another. engineering and its prominence in the history of human thought.5. If none of the axioms (or parts of the axioms) is deducible from any of the others.
can be found in Kelly (1978). such efforts are doomed to failure. in the 20th century that consistency was proved by the American mathematician of Polish origin. without success. implying that the goal of finding a truly ‘democratic’ voting system is impossible. and then only for a section of Euclidean geometry called ‘elementary Euclidean geometry’. His work. The foundation of mathematics is not as rock solid as the layperson believes.Sets 55 to any area in which individual preferences need to be combined in some way. We may add that there is one somewhat controversial axiom in set 6Actually Godel proved more than this. Set theory ought to be developed from a system of axioms which prevent the occurrence of the paradoxes. Godel. inconsistent. The history of Euclidean geometry seemingly offers a way out of the problems besetting set theory. which is quite easy to follow. These propositions are called undecidable. Other people have tried. Tarski. proved that the consistency of the generally accepted axioms of set theory cannot be proved by mathematical means. In elections. It was. shortly after discoveries of other geometries. ‘The axioms are consistent’ is one such proposition. Some of these. Then we should aim at proving consistency for the axioms of set theory. voters (usually) have an order of preference for the candidates. Since their introduction noone seriously doubted the consistency of the axioms of Euclidean geometry. in fact. The critical study of the foundations of Euclidean geometry was undertaken towards the end of the 1gth century. by Euclid. Unfortunately. The geometry he created (or.A can be proved by mathematical means. The system of axioms set out by Arrow can be interpreted as a bare minimum statement for a voting system to be regarded as ‘democratic’. put on an axiomatic basis) is now called Euclidean geometry. The first scientific discipline which was axiomatized was geometry about 300 B. rather. An Austrian mathematician. . Euclid’s axioms were then augmented and Euclidean geometry perfected. however. A. then it is possible t o formulate a proposition A such that neither A nor . Arrow proved that the axioms he set out were.6 This creates a problem. It was found that Euclid used tacitly some propositions unwarranted by his axioms. He showed that if the axioms of set theory are consistent. only later. This was actually done in the first half of the 20th century. is set out in Arrow (1963). to find other axioms for a ‘democratic’ voting system. K. with proofs of inconsistency. Among the axioms of set theory is one which prohibits constructions like the one in Russell’s paradox. in that we do not have an entirely satisfactory philosophy for the foundation of mathematics.C.
7There are no problems about independence and consistency of the axiom of choice. The set of even integers as well as the set of numbers of the form 9m+15n with rn. The sum and product of even integers is even. (2) for every element T E R with T # 0. rather than which system of axioms should be used.1 Divisibility: A n example of an axiomatic theory This subsection requires more mathematical maturity than the rest of this chapter and uses some concepts explained only later in the book. (b) for every two elements x.n E Z form a ring.3 are satisfied. Godel showed that the axiom of choice is consistent with the other axioms of set theory.56 Introduction to Mathematics with Maple theory . some of which are contradicted by our i n t ~ i t i o n . to accept an unrestricted use of indirect proof. A set R is called a ring if. Not all mathematicians are prepared.the axiom of choice mentioned in Chapter 9. Mathematicians are more divided on which arguments are permissible and which are not. as they are for numbers of the form 9 m 15n. 2. and Cohen proved that it is independent of the other axioms. A ring R is called Euclidean if it has the following properties: + + (1) if x # 0 and y # 0 then xy # 0. Skipping this subsection will not affect understanding of the rest of the book. The interested reader is referred to Halmos (1974). y with y such that # 0 there are q and T in R and either T = 0 or N ( T )< N(y).5. however. It leads to rather surprising results. Here we leave axiomatics and set theory. but convenient. Set theoretical language plays only an auxiliary. there is a positive integer N ( T ) such that (a) N(xy) 2 N ( z ) if x # 0 and y # 0.3 are obviously satisfied for even integers. for every two elements x E R and y E R there is associated a sum x y E R and a product xy E R such that the axioms in Table 2. role in the mathematics considered here: much of the mathematics in this book was in existence long before set theory was born. for example. The axioms from Table 2. . When ~ the foundations of mathematics are studied there is also a need for an analysis of the logical means used in deriving conclusions.
Then la . If d is a divisor and D the greatest common divisor of two elements in iZ then D = m d for some integer m and consequently 1 0 12 Id[. Here i t is less obvious what q.* In a ringg x is divisible by y if there is q E R such that x = qy.5) should be.q2 I 5 1/2. If z = 0 or y = 0 then the greatest 8Another example of a Euclidean ring is the Gaussian integers: these are numbers of the form a bz with a. Let x = a zb. q and r can be found by the long division algorithm (see Section 6.Sets 57 Table 2. Denote by q1 and 42 the nearest integer to Q and p. "0 is divisible by any element of R and no element distinct from 0 is divisible by 0. In Z the greatest common divisor of two elements has the largest absolute value among all common divisors.We define N ( x ) = )xI2. If x is divisible by y we say that y is a divisor of 2. Set q = q1+ zq2 and T = x . Consequently have N ( T )= IrI2 = Iy(q1 . Polynomials with rational coefficients also form a Euclidean ring. The set of all greatest common divisors of a and b is denoted by gcd(a. y. r are the quotient and the remainder. respectively. N is the absolute value. It is common but perhaps a little confusing to write d = gcd(a. y is defined as a common divisor (that is x = m d and y = n d for some m.a ) y(q2 the Gaussian integers form a Euclidean ring. b) instead of d E gcd(a. For example. The integers form a Euclidean ring: q.that is N ( a + z b ) = a2 b2 and 5 lyI2(f f ) < IyI2 = N(y). 'Not necessarily a Euclidean ring.3).qy with T = T I + z T ~ . b). y = c zd. For N we take the degree of the polynomial. N ( r ) = Irl. + + + + + + + .l' The greatest common divisor d of two elements 2. D : x(y Z) = xy + + + XZ. T in (2. respectively. b E Z and z2 = 1.q1 I 5 1/2 and . b). by division of x by y.3 Ring axioms A1 A2 :X + Y = Y + X :x (y Z) = (X y) z A3 : There is an element 0 E R such that 0 x = x for all + + + + Mi : X Y = Y X M2 : x(?/z) = (XY)Z + x in R A4 : For every element x E R there exists an element (2) E R such that (x) x = 0. x/y = Q zp. n in the ring) which is divisible by any other common divisor of x. in Z the numbers 4 and 4 are the greatest common divisors of 8 and 12.
. P is a nonempty set of positive integers and has its smallest element N ( d ) (see Theorem 4. y in a Euclidean ring always have a greatest common divisor.58 Introduction to Mathematics with Maple common divisor of x. Consider the sets M = (mx + n y : m E R . We now show a constructive and effective way of finding the greatest common divisor. y) = y. We may assume for the rest of the proof that x # 0 and y # 0. 72. the last inequality is impossible since N ( d ) is the smallest element of P . Indeed. Consequently m n = 1 and m. n E R. The Theorem is obvious if x or y is zero. This can be done by the socalled Euclid’s algorithm. In Zthe divisors of 1 are 1 and 1. Consequently r = 0 and d is a divisor of x and y because both of these elements are in M . 1 Two elements x. and the corresponding natural numbers N(r1) > N(r2) > N ( r 3 ) > . . Theorem 2 . If . if d. r3 . If r1 # 0 then again 72 # 0 we continue the process and obtain elements r1. The process must end after at If r2 = 0 then it is easy to see that gcd(x. However.. . Otherwise two greatest common divisors differ by at most a factor which is a divisor of 1. The existence of a mathematical object is established without any laborious process of actually constructing the object.10). P = { N ( w ) : w E M . In the ring of polynomials with rational coefficients the divisors of 1 are all rational numbers distinct from 0. 0 The proof just given is a good example of an existence proof characteristic of modern mathematics. Hence d is the greatest common divisor. Proof. n E R } . and in particular d. Since R is Euclidean we have w = dq+r and either r = 0 or N ( r ) < N ( d ) . If y # 0 then we can find qo and r1 such that If r1 = 0 then gcd(x. w # o}. Let w E R be arbitrary. D are two greatest common divisors of x and y then d = m D and D = nd for some m.. y is 0. Every common divisor of z and y divides any element of M . n are divisors of 1. y) = T I .
Let us illustrate the Euclid algorithm by a simple example in Z. Maple can be employed to do the computations in the Euclid algorithm.%> .149017)by using MupZe.14. 119 = 14 x 8 gcd(133. below need not be retyped. The command for finding the remainder of division of integers z by y. . 133 = 119 X 1 I. Let us find gcd(215441.ll > irem(2I5441.Sets 59 most N(r1) steps when we obtain It can be shown by mathematical induction (see Chapter 5) that r n = gcd(z.y).66424) . 66424 > irem(149017. Start with x = 133 and y = 119. 16169 > > > # We can now use t h e same Maple command repeatedly # t o evaluate t h e successive remainders. 1748 "The command irem(%%. is irem(x.%> .149017).llg) = 7 + 7. as we know. # and use t h e s e as t h e d i v i s o r i n t h e next s t e p irem(%%. 3 ) . 14=7~2. you can use the menu of Edit to copy and paste.
22 1 Exercises Exercise 2.22 . z E R. Show that. The command is igcd if we are working in iZ and gcd if we are working in the ring of polynomials with rational coefficients.5.8x2 . One can do even better: Maple has an inbuilt facility for determining the greatest common divisor.3. Prove this and apply it to finding a greatest common divisor of the following polynomials: + (1) xlo0 ax2 bx 1.149017) .1 Define the greatest common divisor ofn elements.3. For example ~ ~~ ~~ ~ a > igcd(215441. n elements always have a greatest common divisor. in a Euclidean ring. ax2 bx 1.~24*~+4) .h E R with gcd(w.2 R is a ring. x2 . (2) 2 6 x5 3x4 . z ) .60 Introduction to Mathematics with Maple 437 > irem(%%. z ) = gcd(w hz. 437 > gcd(x25*~+6.2 2 .%) . + + + + + + + . Show that gcd(w.6x3 . h) = 1. Exercise 2. 0 Using Maple and the Euclid algorithm we found that 437 is the greatest common divisor of 215441 and 149017.5. w E R.
.x4 + x2  1. (3) 5xs . x5 . (2) 1234567.3 Use Maple to find the greatest common divisors oE (1) 108.5. 144.Sets 61 Exercise 2. 7654321.x3 + x2   1.2x7 x6 5 x 5 + + + 2x4 .
This page intentionally left blank .
The range of R is denoted by rgR and is the set of all second elements of 63 . 3). functions and various notations connected with functions. (2. Denoting by domR the domain of R we have domR = { a . if related means that the first person is the father of the second person then this relation consists of all pairs of the form (father. 2). and study some basic concepts intimately related to functions. (a. 1)). b) belongs to some set. For example. To say that two things are related is really the same as saying that the ordered pair ( a . (11.Chapter 3 Functions In this chapter we introduce relations. . b) has some property. This. We define the domain of the relation R to be the set of all first elements of pairs in R. This relation can be interpreted by saying that m and n are related (rncn) if n is the hour immediately following m on the face of a 12 hour clock. son). daughter) or (father. we also denote this by writing aRb. can be expressed by saying that the pair ( a . 3. If a set R is a relation and ( a . Another example of a relation is the set C = { (1. . b) E R then the elements a and b are related.1 Relations In mathematics it is customary to define new concepts by using set theory.1 (Relation) A relation is a set of ordered pairs. 12). b) E R } . . . We define: [ Definition 3. in turn. 1 This means that if A and B are sets then a relation is a part of the Cartesian product A x B. (12.
" f range R 0 I / domain R Fig. b) E R ) . so r g R = { b . . For the 'clock' relation C clearly d o m C = r g C = (1. as in relations R2 and R3.2 to 3.. The graph of a relation can cover a whole area. Note that the boundary is included in R2 but not included in R3.. (a. For the relation fatheroffspring the domain is the set of fathers. the range is the set of daughters and sons. The graph of R consists of all points P with first coordinate x and second coordinate y such that P z (2. 12). .1 Graph of a relation X The graphs of are shown in Figures 3. respectively. 3. y) E R..4. The domain and the range of a relation is schematically indicated in Figure 3. 2. We can often draw the graph of the relation R as in the diagram below.64 Introduction to Mathematics with Maple pairs in R.1.
3. 3. However. we wish to emphasise that our theory itself is independent of any geometry and can stand on its own feet without any support of graphical illustrations.Fzlnctions 65 n Fig.2 Graph of R 1 Fig. including basic analytical geometry. We shall use geometry freely to motivate or illuminate our theory. For this we presume that the reader is familiar with elementary geometry. it helps understanding. and often motivates a proof of a theorem or construction of a counterexample.3 Graph of R2 The graph of a relation or a geometric picture generally is a telling guide. Our ability to draw a graph is restricted by the unavoidable imperfection of the .
We can define a number of properties of relations. Here the imperfection of the drawing instruments makes it impossible to draw the graph of R4. or simply an equivalence.y ) . It is perhaps not clear from Figure 3.4 that the points lying on the parabola y = x2 do not belong to the graph of R3. Let R4 = ((2. x E Example 3. Let Mk = {(x.1 Z. A relation R is said to be transitive if x R y and y R a imply x R z . Either we will blacken the whole plane or the graph of R4 will be very incomplete. symmetric and transitive is called an equivalence relation.4 Graph of R 3 X drawing instruments. 0 0 0 A relation R is said to be reflexive if xRx for every x E dom R. x . A relation which is reflexive. y E Z. The next example is far more serious. A relation R is said to be symmetric if x R y implies y R x for all x E dom R and y E rg R. either x or y is rational}. 3.y divisible by k} for . y ) .66 Introduction to Mathematics with Maple Y I Fig. Obviously R 4 cannot be graphed.
7) (4. Similarly. and aRb if a and b differ by 6.1 Graph the following relations from R into R and decide which relations are (a) symmetric. Write this relation as a set of pairs.14) (3. Then = XMky. Examples of relations which are symmetric. If this sounds too abstract. . For congruence modulo 3 it is easy to see that K1 = K4 and K1 n K2 = 8. q ) and (r. say ( p . If R is an equivalence relation on A and a E A then the set K = { b . s) with q # 0. and Kb are either identical or disjoint. then the pairs (2.1. in other words elements in an equivalence class are considered equal. q).1.1. If two pairs of integers ( p . it is not.13. (b) reflexive. nor transitive. in arithmetic mod 3 we write 1 = 4 = 7 = . .1. . . how is this shown by the graph? Exercise 3. The ‘clock’ relation is not symmetric. we can emphasise this by writing K . (Prove this!) The equivalence Mk is often called congruence modulo k. Remark 3. nor reflexive. Sometimes all elements in an equivalence class are identified. aRb) is called an equivalence class and a is a representative of K . reflexive or transitive are given in Exercises 3. and describe the relation using the face of a clock. This is based on identification of elements in an equivalence class.2 Let dom R = { 1. this is the above equation of common fractions.21). s # 0 are defined to be equivalent if p s . and the y mod k (read x congruent to y modulo k) is used instead of notation 2 E k N. or (c) transitive. 12). Note that K depends on a.Functions 67 the relation Mk is reflexive.. symmetric and transitive. We are used from primary school to equations 2 4 6 like . .. 3. A common fraction p / q is just a pair of integers..= . instead of K .= .. If the relation is symmetric or reflexive.1 7 14 21 = = Exercises Exercise 3.qr = 0. 2.6. Two equivalence classes K .
called the domain of a function. (2) symmetric. clearly . 3 6 3 6 3. Show that this relation is an equivalence. the average daily temperature depends on the date. Define a relation S for subsets of R by S = { ( A .4 sets such that A = uA 1EL Let be a set and B 1 with 1 E L a family of disjoint B1.1.1.simply denotes the equivalence class h 2 4 2 4 containing the pair (a. However some doubts may arise as to what is meant by the word ‘rule’. for example. Exercise 3. but not symmetric or transitive.= . In one of the examples above the rule associates with each date a number equal to the average temperature on that date. Define a relation on A by declaring aRb if a and b lie in the same Bl. . (3) transitive.2 Functions The concept of a function is the mathematical abstraction of the notion of dependence in ordinary life. ( A \ B ) U ( B \A) is finite). This equivalence relation is a starting point for defining rational numbers in a terms of integers. the area of a square depends on the length of its side. A reasonably good description is to say that a function is a rule which associates with every element of a set.= .n)R(r. . B ) . but not reflexive or transitive. @ Exercise 3.68 Introduction to Mathematics with Maple Exercise 3.1. a uniquely determined element y called the value of the function.= . Prove that R is an equivalence. but not symmetric or reflexive. The price of a commodity depends on its quality. b).= . Using mathematical terminology we say. In this book we define functions using settheoretical language.5 @ Exercise 3.3 Define separate relations which are: (1) reflexive. Prove that S is an equivalence relation.6 Define a relation R with dom R = Zand rg R = (Z\{O}) as follows: (m. that the area of a square is a function of the length of its side. A c R. The fraction .s) if m s = nr. and the family of finite subsets of R is one equivalence class. . B c R.1.
(Read: ‘ x goes to y’ or ‘f sends x to y’. Not every relation is a function. . . . for which the prices are $20 and $10. we may denote the function { ( x . Since a function is a relation it is automatically clear what is d o m f . y) E f then y is denoted by f ( x ) (read f of 2 ) .2). Iff is a function and ( 2 . . The range of f is simply the set of all function values. . y) E F then y is the element associated with x by the rule for F . H . In the above example. x 2 ) .. The symbol f ( x ) is called the function value or the value of the function at x. then. $10)). . x E D } we shall often write + f:x++y. undesirable because it confuses function with function value. Functions are usually denoted by letters f. the function is simply {(good.’ It is possible to encounter in the literature expressions like ‘the function x2’. which is a set of pairs. y) E F and E F] +y =x for all x E d o m F and y.z ) [(z. the relation fatherRdaughter is not a function because there is (at least one) father who has two daughters. We shall try to avoid such abuse of terminology. . It is important to make a clear distinction between f . G. or F. h . while the output displayed when the program is run with a particular value of z as input is the function value. is a set of ordered pairs such that no distinct pairs have the same first element.Functions 69 Definition 3. respectively.) IThis can be illustrated by reference t o using a computer or programmable calculator t o evaluate some expression. x E R}.which is an element of rg f. if there are only two qualities of a merchandise. x / ( l n x 2 )). good and bad. . X E D . If F is a function and ( 2 . Definition 3. For example. The computer program is the function. are also used. rgf and the graph of f . Instead of writing f = { ( x . .2 ensures that every line parallel to the yaxis intersects the graph of f in at most one point.3). A function. $20).2 (Function) A function is a relation F such that (2. +. Sometimes letters of the Greek alphabet such as 4. (bad. There is a major difference between the program and the output of the program. This is a drastic and undesirable abbreviation for { ( x . and f ( x ) . The ‘clock’ relation C is a function because it does not contain two distinct pairs with the same first element. z E r g F . g. x E R} by fn. The relation R1 from the preceding section is not a function because the line x = 0 intersects the graph of R1 in two distinct points (see Figure 3. Suffices are employed too. .
4. We deal with graphs in Maple in Subsection 3. The notation f : A I+ B is used to indicate that domf = A and rgf c B. Not only does it carry out the required calculations. The table for the clock function C is given below. for example. For instance. This notation is convenient in theoretical discussions. involving a considerable amount of calculation. that is. but it has the disadvantage that it does not explicitly describe the values of f. but it also has excellent graphical capabilities. x w x / ( I + x). It is often convenient t o display the function pairs in a table. Two functions are equal if they are equal as sets. f : R 4 P means that f is defined and positive for all real x. For h it is R \ {1). To graph a relatively simple function (by hand) might be laborious.2 The function h : x H x/( 1 z) + x 3 2 2 3/2 3 4/3 4 1/2 1 1/4 1/3 0 1 1/2 2 2/3 J h(x) 3/2 0 The graph of a function usually conveys a good idea of what the function is like. h : x I+ x / ( 1 + x). Table 3. This is where Maple is of great help. In a situation like this it is understood that the domain of f is the largest subset of R for which f(x) makes sense.2.but otherwise f can be arbitrary. Figure 3. particularly if we wish to draw the graph of the function.70 Introduction t o Mathematics with Maple The notation f : x H f(x) is also used. Table 3.3. Equal functions have the same domain . Such an incomplete table for h : x I+ x / ( 1 + x) is given in Table 3. for instance. If there is no need to name a function we may even write x H y to describe a function.1 The clock function x 1 2 3 4 5 6 ’ 7 8 9 1 0 1 1 1 2 3 4 C(x) 2 5 6 7 8 9 10 11 12 1 If dom f is infinite (or simply very large) we may fill in only a few values for x.5 shows the graph of h obtained by Maple. if they consist of the same elements (ordered pairs).
Of course.5 Graph of l f x X and associate the same y with the same x. Using the arrow notation for functions we can write QI : x t+ x 2 . for example for h above . Other letters may be used freely. they then also have the same range. Note that a! is not the same as s : x I+ x 2 because s is tacitly defined on all of R (doms = R ) .Functions 71 1""""'"' 4 X / ~~ Fig. x E P. the letter z has been used for the first element of a pair belonging to a function. Example 3.2 Denote by a!(z)the area of a square whose side has length x. 3. So far.
s2).1 in Chapter 4) and when it is used for the zero function. There is an intimate connection between sets and characteristic functions: see Exercise 4. E D } . one should realize that 1 0 and the number 0 are different objects. x) : x E A } .5. It is customary to denote 1 0 by 0. h = {(w. however. x E A } U {(x. and f . If f is a function then we define f(A)and fl(B) by Obviously f(A) = (9. = {(s. By idA we denote the function {(x.l ( B ) is called the preimage of B. log) but not for the very often occurring function id. YEP. (2. y) E f and x E A } . It is called the identity and we denote it by id. clearly 1 0 : x H 0.7. x2 then Clearly rg f = f(dom f ) and dom f = fl(rg f). a!:yHy2. The function l p is sometimes called the Heaviside function in honour of the British mathematician and physicist Oliver Heaviside. cos.1 and the comments at the end of this chapter. like R in our case. s E Example 3. Clearly. The function l g is called the zero function.{I}}. If we do not specify otherwise we shall assume automatically that dom1A = R. In terms of ordered pairs 1~ = {(x. x @ A } . However.2. The context will (usually) make it clear when 0 is the number zero (see Axiom A3 in Table 4. The set f(A)is called the image of A under f . Some properties of the symbols f ( A ) and fl(B) are established in Exercise 3. Example 3. This notation is used by other authors too but it is by no means standard. a! V/(W + I)). If s : x . It is perhaps a little strange that there is a standard notation for many functions (like sin.2 we could equally write: h : t H t / ( t + 1). 0). In other words id : x H x or id(x) = x.v E R . and then it is understood that dom1A = X .3 The function x H x is perhaps the simplest function. 1).4 For a set A the characteristic function 1~of A is defined by l ~ ( x = ) 1 if x E A and l ~ ( x = ) 0 if x @ A. who used it when applying mathematics to solve problems in electrical circuit theory. in mathematics one often encounters a situation in which there is some general set X I fixed during a discourse.72 Introduction t o Mathematics with Maple and a from Example 3. l ~ ( z = ) 1 for every x E R. We can also write lw : x H 1.
particularly if either dom f or rg f is not a subset of R. However we may resort to it in order to conform with widely used notation. 1. functional. For example. if c E R. If g is a restriction of f to some set B .2. since there is little fear of confusion about the meaning. s and h : x It is a slight abuse of notation to denote by the same symbol a function and its restriction or a function and its extension. However for a given g and D there may be several extensions of g from domg to D x2. sl({l}) = 8. We may occasionally use the word ‘operator’ for a function whose domain is a set of functions. Given f and B . Other words are used with the same meaning as the word ‘function’. the function f l is ~ uniquely determined. Obviously. The terms operator.4}. f l ~ Exercises Exercise 3. s1({1}) = (1. For example. Describe the relation between the . There is some harmless ambiguity in this notation. there is often no such distinction as we emphasised made between a and s above.Functions 73 s({1.1p(x) are both extensions of a from P to R. Perhaps we should have written dom(flA) but on occasions like this we shall omit the parentheses. The function a from Example 3. The function s is the extension of cu to R. f : R + R and g : x H f(x . We may occasionally use the word ‘map’ or ‘mapping’ instead of the word ‘function’. If f : D H V and A c D then the function f l (read ~ f restricted to A ) is defined by is called the restriction of f to A. l}. transform are also used for some functions.2 is a restriction of s : x H x2 to P. always f l C ~ f and dornfl~ = A .c ) then T : f g is called the translation operator. Usually it is clear from the context what N is. Similarly the function f ++ f ( is ~ called the restriction operator. Some authors use the word ‘function’ only if the range is a set of numbers. g = ~ J B then . The symbol f ( N ) may denote the value of f at N E domf or it may denote the image of N under f for a set N . f is called the extension of g from B to D. For instance.1 Let f : R graphs off and g if + R. 2}) = {1. transformation.
A l . x Hl p . Let f be defined as follows: if x is irrational then Exercise 3.) (You should realise that f itself cannot be @ Exercise 3. Exercise 3. graphed . then f (F) = a. and graph 1 q the corresponding points.1~ < 0).with p E Z. f(Jz).2.f(l).2. if x = . A2.1 to graph x The function signum is defined by I+ 2x2 + @ Exercise 3. A1 for 1 E L be subsets of B 1 for 1 E L be subsets of Y . f(. Determine the values f ( l ) .l ( B l \ B2) = fl(B1) \ fl(B2). c B2 +fl(B1) c fl(B2).q E N and p .2 52 3. Graph the function sgn. give an example showing that equality (8) f . Let f : X I+ Y .5).2. (5) f ( A i ) \ f(A2) c f(A1 need not hold. A.74 Introduction to Mathematics with Maple Sketch the graphs for f = 1 p and f (x) = x2.5 X . give an example showing that equality need not hold. . (1) A1 (2) B1 B1. B2.2. Prove: c A2 =$ f(A1) c f(A2).3 sgn : z where A denotes {z. q relatively prime. + Use the results of Exercise 3. \ A2). and B. x E R.2.4 P f(x) = 0.
The following session illustrates use of the common logarithm. The reason is simply because mathematicians rarely use it.3 3. The arguments to the functions usually specify the point at which the function should be evaluated. give an example showing that A can be a proper subset of fl(f(A)). Some of the more common functions are shown in Table 3. Exercise 3 . and in these days of cheap pocket calculators for all sorts of specialised uses.Functions 75 (9) f(fl(B)) B. In( 100) In( 10) In( 100) In( lo) .3. The command ? l o g will tell you all you need about the common logarithm in Maple. which are passed to the function in parentheses. If you want to use a function you must know the name of the function and then you can ask Maple for help. 6 Let rg f = (0. * (11) A1 c A (12) B c y * f(A(A1) = f(A1).] 3. 2 . * (f(A)l(B)= A n fl(B). Show that there is a set A such that f(z)= ~ A ( z ) for every z E d o m f . The notation ldr unctions in Maple is similar to that used before for functions: these functions operate on one or more arguments. a negligible number of other people use logarithm tables. 11. f(fl(B)) = B B c f(X).1 Functions in Maple Library of functions Maple includes hundreds of mathemat.:al functions. [Hint: A = f1(1).3. and fairly cheap computers. You might wonder why the common logarithm (logarithm to base 10) is not in the above table. (10) fl(f(A)) 3 A.
but we can blame only ourselves for it because we did not take into account the fact that the first command simplify (%%%) produces a result. which now becomes the most recent result for the next command.3 Some Maple library functions Mathematical name Description Absolute value Square root Exponential Function Natural logarithm Trigonomet ric function Trigonometric function Trigonometric function Maximum of two or more numbers Minimum o f two or more numbers The greatest integer not exceeding x The smallest integer not preceding x Maple usage abs (x) 1 sqrt(x) exp (x) log(x) or h ( x ) sin(x) cos (XI tdx) max (x y z ) ) m i n h . The commands should have been .76 Introduction to Mathematics with Maple Table 3. namely 2.y z ) floor (x) ceil (x) > 2 ln(2) + ln(5) ~ ~~~~~~ There is an obvious error on the second last line.
and finally f ( 2 ) is evaluated. 5 26 In the next example.2 Defining functions in Maple In Maple you can define your own functions. Now the best way to recover is .3010299957 3.simplify(%%%) .simplify(%%%) . without having to retype the function value. The following example shows how to define a function f (x) = x2 1. f :=x+x2+1 > # Evaluate > f a t 2 and 5 f(2). This allows you to evaluate functions at various points. in the Maple definition of the function. The notation is similar to the one we employed for defining functions: for H. f(5). + > > # Define f (x) = x2 + 1 f:=x>x^2 + 1.Functions 77 > simplify(%%%) . type > as a hyphen followed by a greaterthan sign. and how to find the value of this function at x = 2 and at x = 5. .3. the variable a is assigned the value 3. then the function f (2)= ux2 is defined.
first define the function (say f ) . Look carefully at the following example: > > # Define f (x) = x+2 f:=x>x+2. 12 As shown above. and then specify the required point in parentheses (for example.  2x + I g := x + x2 . More generally. f :=x+x+2 > > # Define g(x) = x2 g:=x>xn22*x+1.2 x + 1 x22x+l . and Maple will evaluate the function with respect to those variables or expressions.78 Introduction to Mathematics with Maple a := 3 > f :=x>asx12. f (2)). when you want to evaluate a function at a particular point. you can place variables or expressions in the parentheses.
2 x + 1 ) 2 +2 In a similar way. f(z.y) >sqrt (x^2+y^2) .y) = x2 y2 i s defined by f := (x . y. > > # Define f (x. you can define functions of two or more variables. f := ( x. as shown below. For example: f(x) = x2 1 i s defined by f :=x>x^2+1. Use := and > to define functions.z) > (x2+y2+zn2)/ (x+y+z) .2 x + 1) ( x2 .z) = (x^2+~^2+~^2)/(X+y+Z) g: =(x.y) = sqrt(xA2+yn2) f := (x.y.y)>xA2+y^2.Functions 79 ( x + 2 ) (x2. Note the slightly different way of typing such functions. + + .y ) + sqrt( x 2 + Y2 ) 5 > > # Define g(x.
L := [7r.L) . . The command is select and we use the function big defined above. nobody can decide whether x is positive or not unless it is known what x actually represents.5.80 Introduction to Mathematics with Maple 3. 1 are used instead of false and true. however very often it is needed to select the elements by some property rather than by their position in the list.L) .1 how to select elements from a list by using the op command. is ( (sqrt (2) (3/2)>5/2). 1.sqrt(3). 0. Let us consider the list > < x) := x + is(2 < x 2 ) + is(2 L:=[Pi.1/31. We now define two functions which we shall use in selecting elements from a list. For instance > is (Pi<22/7) . .2.3. > big: =x>is (x>2) . > select (sqbig. [TI Similarly. > select (big. is (x>O) .5. a.sqbig:=x>is (xL2>2). we can create a list of elements with squares greater than 2. 1 1 3 from which we wish to create a list of elements greater than 2. big := x sqbig We have seen earlier in Subsection 2. P is a true propostion. namely true. indeed. or three elements.1. 2Sometimes 0.3 Boolean functions We call a function a Boolean function if its domain of definition is a set of propositions and its range has either two elements.l. false and fail2 We shall tacitly assume that if the value of a Boolean function at P is true then. Fail is used in Maple. true false FAIL The last answer is natural. on other occasions the word undecided is sometimes more appropriate. A Boolean function is realized in Maple by the command is. namely true and false.2. 2.
op(3. rat := x +is(x. The alternative for the ‘name’ in (3. If you do not like . like sin.] 3.3. [r. . or symbolically as sin(x) is. [Hint: Use the methods of the two previous Exercises. 7919. for example > a. It should not be confused with range of a function.5 in the list you can remove it manually. for removal of rationals from L.range.rational) . Exercises Exercise 3.L) . For instance.1) can be a name of a function. 1317237’ [Hint: is (x . [Hint: Make a list L of these integers. (3. > remove (rat .L).1) The name and range are always compulsory.1 Which of the following numbers are prime: 1979. define sq:=x>is(sqrt(x) . The form of the command for graphing a function is plot (name. This function value can be entered directly.] Exercise 3. we first define a function > rat:=x>is(x. We deal first with ‘name’= function value. as x2 .Functions 81 Other properties can also be used for selection or removal.51 Note that for Maple the floating point number .options) .3. a 1 Cop(1.L) .2 Find all complete squares between 100 and 200. There is .L)1.3z is. rational) and then use the command remove . below.3. The word ‘range’ is used here as in common English meaning the limits (for the first coordinate) within which the graph will be displayed.3.4 Graphs of functions in Maple Maple provides great power and flexibility in graphing various types of functions.prime)] Exercise 3.integer) and select (sq. The ‘name’ can be a general expression for a function value.3 Find all primes between 5000 and 5100.5 is not rational.
. .a2 Introduction t o Mathematics with Maple a great number of options.6 Graph of y = x2 . > > p1ot(xA23*x. 3.plot(sin(x) . We shall see later how to force Maple to produce the same scales. Fig. > > # First define the desired function f :=x>x+l/(x1) 2. which are all listed in the help file.and yaxes are different. It also saves typing if the function value is given by a long expression.x=Pi.1) is now in that the form x=a.b must not be used.6 and 3.x=l. but just the limits in the form a. The ones which are most likely to be needed by readers are explained as we go along.b.4) . The graphs of these two functions are shown in Figures 3. .7. The difference in ‘range’ in (3.Pi).32 The alternative for ‘name’=symbol of a function is perhaps more logically consistent.. Note that the scales on the x. We will illustrate this by an example with a function which we define our selves. . since we graph functions.
3/2. . > plot (f . > # The graph. In our example it is near the dip in the curve.. as must be expected with the infinite range of x.: Fig.Functions 83 0. looks a little rough. > > Restrict the range of x for the plot of the function plot(f. # The portion of the graph restricted to 1. is reasonably good but.1)2 Then p l o t the function. 3. The use of the infinite range is generally not recommended.5 5 x 5 4 is shown in Figure 3. .7 Graph of sinx f :=x+x+ 1 (x . however. shown in Figure 3.infinity) .infinity. It can. serve as a guide where to have a closer look at the graph.9. We now restrict the range of x to the interval where we expect the dip to occur.4).8.
restricted horizontal range + (x 1)2* ~ .5 2.5 3 3.9 Graph of x 1 . 3.8 Graph of x +A . 3.84 Introduction t o Mathematics with Maple infinity infiiity Fig.5 4 I Fig.unrestricted range (x 1 y  1 1.
which was shown in Section 1. Fig.6) will plot the function f for x between 0 and 6 and will not restrict the range of y. In order to obtain a good graph we can either limit the range of x or use the option numpoints to increase the number of points used to plot the graph. One possible way to help Maple to produce a good graph is to limit the range of the y values as well.5.10 Graph of x 1 +both ranges restricted (x 1 ) 2 ' *  The graph of S : x H x sin 7x.6). Maple will not be able to cope3 because f takes very large values near x = 1. and it 3Try it on your machine.Functions 85 If we specify the range of x as 1.5. .. Maple does just what you ask it to. 3. and must precede the range of values for y. > p l o t ( f . for instance between 0 and 6.1 (see page 3) was not satisfactory.0. As with other computer programs. The command p l o t (f .I. + . O . .. The result of restricting x to the range between 1 and 5 and y to the range between 0 and 6 is shown in Figure 3.10. Note that the range of values for x must also be given. By default Muple uses 50 points.
5) .11 Graph of x + sin 72: restricted range for x . . # Now p l o t t h e f u n c t i o n : # f i r s t with a r e s t r i c t e d range. respectively. But with functions with a large number of maxima and mimima within the range. 3. .5 . In most cases this is satisfactory.86 Introduction to Mathematics with Maple joins successive points by a straight line.20.12.plot(S. Fig. With 150 points the graph is fine. .numpoints=i50). The results of these two graphs are shown in Figures 3.11 and 3. more points are needed.20. > > > > # F i r s t d e f i n e t h e f u n c t i o n t o be graphed S :=x>x+sin(7*x) . > > > # and t h e second time over t h e o r i g i n a l range # but using more p o i n t s p l o t ( S .
x=4.4) . This sometimes results in an undesirable distortion as in the next graph of sinx. This is shown in Figure 3. c o s ( x ) . .Pi.14. _______~~ ~ It is possible to include multiple plots on the same set of axis by using set notation. so that all the function to be plotted are enclosed within braces { and 3. > plot (sin.Pi.13. . . The option scaling=CONSTRAINED ensures the same scale in both directions.O.12 Graph of x + sin 72: additional points included Maple adjusts the scales on each axis to fit the picture.. Try plot({sin(x) .Pi). 3. . The result of this is shown in Figure 3.scaling=CONSTRAINED).Functions 87 Fig.sin(2x)). > plot(sin.
7r] Fig.14 Graph of sinz using option scaling=CONSTRAINED . 3.13 Graph of sinz over the interval [0. 3.88 Introduction to Mathematics with Maple Fig.
It is easy to see that and we shall denote this double composition simply by f o g o h. + Composition of functions If f and g are two functions then the composition of f and g . see Exercise 3. 2 3 .1.Functions 89 Exercises Exercise 3. 3).2.4. x H cos(x/2). (2. What are domf o g and domg o f in (2)? . denoted by f o g (read f circle g) is defined by The domain of f o g is the set of all x E dom g for which g(x) E dom f .4.4 x H (x . 9 = ((1. see Exercise 3. 0 ) ) . (2) f = C(1.1. In particular dom f o g = dom g if rg g c dom f . 21. ( 0 . The same function can be expressed as a composition of two functions in many different ways.32.x H cos 22.4 Graph the following functions: x3 x . graph each group of three functions in one (1) x (2) x 3. g : x w x + l . H cos x. Exercises Exercise 3. H x3 . H 22.3.2.1 Find f o g and g o f i f (1) f : u H u 2 . (5) x (3) x (4) x H + Exercise 3. 4511.2)2. x H (x 2)2.3.4.5 picture: In each case.3x2 . Generally f o g # g o f .
1. A bijection F : X H Y is sometimes described as a onetoone correspondence between X and Y . fls is onetoone. 2 2 E dom f . An equivalent implication is f(x1) = f(x2) * x1 = x2. The function h from Example 3. x + 1. Y is called a bzjection of X onto Y if it is onetoone A function F : X and onto (or it is both injective and surjective). ifx2 then it follows easily by removing fractions that x1 = x2. Bijections A function f is said to be onetoone (or injective) if for every XI.6 The function h : x I+ x / ( 1 + x) is also onetoone. f :y g :x y2 . 1+x1 1+x2 If f(C) = D then we say that f maps C onto D . associate with each visitor the chair sat on. G : x H x . Let h : x H x2 + 22. Indeed. Indeed.90 Introduction to Mathematics with Maple Exercise 3.5 is soldout. or that f : C H D is surjective. then the function defined by the association between people and seats is surjective. This association defines a function which is onetoone: each chair has a t most one visitor sitting on it.5 onetoone o n a set S if In a theatre. A function f is said to be Example 3.2 A given function can generally be expressed as a composition of two functions in many different ways.6 is a bijection of X R\{1} ontoR\{l}. if y # 1 there is an x such that = yand x+l X for no x is = 1. For given X and Y there can be several bijections of x+1 .4. F : y~ y 2 + 4 y + 3 .5 is a bijection of the audience onto the set of chairs in the theatre.5 og =F o G.1. Show that h = f 3. Example 3. The function of Example 3. If the theatre in Example 3.
the identity function. 12) onto Exercises Exercise 3. In other words. The graph of f1 can be obtained in Maple by the command note the presence of [ and I in the command! In Figure 3.w = (u .Functions 91 X onto Y . x E R}. Every function. x E R}. 2. [Hint: 3.w) is the mirror image of (w. H1 : x x/2. Viewing the inverse function as a set of (reversed) pairs is useful in graphing. As a concrete example. u) across the line y = x. For instance. as a relation. x). The domain of f becomes the range of f1. = { (x. and similarly rg f = dom f1. If. for a function. 3.16 the graph of the inverse to x I+ &x3 x was obtained in this way. 42). Y E R l . the inverse relation f1 is a function then it is called the inverse function to f or simply the inverse of f .5.6 Inverse functions If S is a relation then the relation S1 = ((9.15. = {(y.1 + + x is onetoone. . (2. . the graph of S1 is simply the mirror image of the graph of S across this same line.w3) u . y) E S } is called the inverse relation to S or simply the inverse of S . The inverse function f1 exists if and only if that is. Since the point ( u .1 and the mapping from one number to the number diametrically opposite on the face of the clock are all bijections of (1. if H : x H 22 then H1 exists and H1 = ((2334. has an inverse: however this inverse relation need not be a function.1 Prove that f : x H A x 3 &(u3. y = x. These results are illustrated in Figure 3.w)A with A > 0.Y/2>. the clock function C shown in Table 3. This can also + . It follows from the definition of an inverse that f1 (f(x)) = x for every z E dom f and f (f1 (y) ) = y for every y E dom f1 = rg f . if f is onetoone. . itself. .
15 Graph of a onetoone function be written in the form f10 f = idA and f o f1 = idB where A = domf and B = r g f . If there is a unique solution to (3..2)has been solved.7 Let G : x ~ l / x x~:{x. so G1 : x H .. x = ... This is illustrated in Figure 3. Fig... .. write x instead of y and y instead of x. We have y = .17.. X Y X 1 1 1 However G and G1 ....O<x<l}. For finding the inverse of f we simply solve the equation with respect to x. Find G1. Since we know that it is irrelevant whether x or y or u is used in the notation f1 : y H fl(y) we can....... Example 3. after (3.2)then it is fl(y)... .92 Introduction to Mathematics with Maple Y .. 3...
Functions 93 . are two distinct functions: domG while = {x. The existence of f1 is guaranteed if f is onetoone.16 Inverse of x I+ &x3 + x.x > l}. domG1 = {x. . This should also be clear from the graphs of G and G1.0 < x < l}.TT 4t Fig. as can be readily checked. and is independent of our ability to find an explicit solution of the equation Y = f (4. It may happen that a function f has an inverse but solving the equation y = f(z)is difficult. 3.
The Muple command f solve can be used for numerical solution of Equation (3. The following Muple session illustrates this.3) and therefore for approximate evaluation of the inverse.3).1 that f : x H &x3 x is onetoone and has therefore an inverse.94 Introduction t o Mathematics with Maple Y 0 Fig.5.16.3) is not easy to solve4. 41n Chapter 8 we deal with solving equations generally and we also use Maple t o solve equations like (3. The equation Example 3. . Its graph was obtained in Figure 3. 3.8 + 1 3+x=y x 10 (3.17 Graph of a onetoone function and its inverse We have seen in Exercise 3.
Exercise 3.506628274 instead o f 6. x) > g(O>.594562117 2. 0 .1.g(l>.088730145 It is important to realize that g is only an approximation to f1. For instance.2 Exercise 3.9216989942 1. such a procedure was conceived by the great mathematician .Functions 95 g := y + fsolve(.6.l>.g(3). indeed.1 1 Let f : z H . x/(l + 1x1). . Exercises Exercise 3. g(f (sqrt (%Pi))) . Prove that f is onetoone and find 2x rg f and f1. gives 2.7 Comments In our approach we derived the concept of a function from that of a set.6. possible to choose the idea of a function as the most basic for all mathematics.x3 1 10 + x . It is. 3. This happens because the command f solve gives only the numerical solution.g(2).6. It may be argued that the concept of a function is more fundamental than that of a set and should be used as the most primitive notion.3 Use Maple to graph the inverse to f : x Find and graph the inverse o f f : x H H x5/(l + x4).y = 0.g(l.
We cannot a priori exclude the possibility that mathematics built on functions would then be distinct from the one based on sets. This does not seem to be the case: set theory and the rest of mathematics as we know it can be recaptured from von Neumann’s theory via characteristic functions. . and at our level of sophistication no mathematician would be inclined to adopt it. 1). The fact that functions can be used instead of sets for the foundation of mathematics is interesting but little known.96 Introduction t o Mathematics with Maple John von Neumann in the early 1920s. Naturally if we choose functions as the foundation stone their theory has to be established on an axiomatic basis. functions whose range is ( 0 . that is.
x # 0. A4 : For every element x E M 4 : For every element x E F . + Table 4. D : x(y z ) = xy xz. A set F together with two functions (x. solve inequalities. x = 0. 1 # 0. 4.1 Fields Real numbers satisfy three groups of axiomsfield axioms. such that lx = x for such that 0 x = z for all x in F all x in F .Chapter 4 Real Numbers In this chapter we introduce real numbers on an axiomatic basis.1 Field axioms ~~ A1 :X + Y = Y + X Mi :X ~ = Z J X A z : X+(Y+Z)=(Z+Y)+Z M2 : x(~z) = (XY)Z A 3 : There is an element 0 E F M3 : There is an element 1 E F . y.y) x y. order axioms and the least upper bound axiom. We discuss each group separately. there exists an elF there exists an element ement xl E F such that (x) E F such that (x) xlx = 1. introduce the absolute value and discuss the least upper bound axiom. (x. z in F . In the concluding section we outline an alternative development of the real number system.1 are satisfied for all x.y) xy from F x F into F is called a field if the axioms in Table 4. + + + + 97 .
(1)a = a.b and a bc are abbreviations for a ( 4 and )a (bc). In more usual language. We shall not dwell on such simple consequences of these axioms and hope that the 'usual rules' concerning addition and multiplication are known to the reader from elementary 1 arithmetic or algebra (for example. Setting x = 0 gives + + o=z+o =O+z  by axiom A l .b which means the same a as ab. y) I+ x y and (2. a2 . Hence every field is a ring. In cases where there is the possibility of ambiguity in this book. xy is the product of x and y. let z be such that x = x x for every x.0 = 0. which does not change x. y) I+ xy are called addition and multiplication. respectively.98 Introduction to Mathematics with Maple The functions (x. commonly called the zero element. and a ring in which axioms M3 and M4 are satisfied is a field. z A similar argument shows that there is exactly one element 1 and that x and x . axioms A2 and M2 are called the associative laws. Axiom D is called the distributive law. The fraction . In Chapter 2 we defined a ring as a set with addition and multiplication satisfying axioms A1A4.). Axioms A1 and M1 are called the commutative laws. for example.b2 = ( a b)(a . Because of axiom A2 we write a b c instead of either a ( b c) or ( a b) c. does not have an inverse under multiplication.l are uniquely determined by x. x + y is the sum of x and y. It follows easily that the element 0 from axiom A3 is uniquely determined. etc. Axioms A1A4 deal with addition. Ma and D. Indeed. similarly abc denotes either a(bc) or (ab)c. a . by axiom AS. A centred dot is sometimes used to indicate multiplication. while axioms M1M4 deal with multiplication. respectively. to produce the respective unit element.b). Many other conventions are used. MI. Derivation of some of these rules are to be found in the Exercises at the end of this section. a.or a / b is simply ab' and a b' denotes a (b'). including the rules for precedence of various operations. respectively. division by zero is forbidden. axioms A3 and M3 state the existence of a unit element under addition and multiplication. (a) = a . as in a . respectively. Note that the unit element under addition. (a') = a. parentheses + + + + + + + + + + + + . b We trust that the reader is familiar with these and similar conventions from elementary algebra. axioms A4 and M4 state the existence of an inwerse element under addition and multiplication.
uz = zu = z . not only theoretically but also in application.2 (1) a . z z = z . with z = 0.1.1. possibly. z denote elements of a fixed field F . Note that there are no other elements in this field. The addition and multiplication tables for this field. Let F = { z . u u = z . (4) (1)u = u. there exist fields with finitely many elements and these are important.1. T h e equation ax = b is uniquely solvable for x i f a # 0. a . These Exercises are stated as theorems: your task is to prove them.1. The structure of a field can be very different from that of the reals or rationals. more frequently used in the modern world than any other. + @ Exercise 4. (3) (al)' = a. It is very easy to check that F is a field. uu = u. There are many fields. Axiom M3 guarantees that any field must contain at least two elements. 0 and 1. O = O . Every time a computer carries out a parity check G F ( 2 )is invoked. . + + + + Table 4. u>.Real Numbers 99 will be inserted. z u = u z = u. where z # u and addition and multiplication satisfy z z = z . For example.a = O .3.2 Addition and multiplication in Add Mu1tiply + 0 1 0 0 1 1 1 0 0 1 It may be thought that any area of mathematics which contained 1 1 = 0 as a basic rule was of little use.are shown in the following table. u = 1. Exercises In Exercises 4. c . If we accept reals as known then we may exhibit the rationals also as a field. b. d . + @ Exercise 4.14. or a centred dot used. ( 5 ) U (b) = (a + + b). to clarify the precise meaning.1 T h e equation a x = b is uniquely solvable for x. (2) (a) = a . which is known as G F ( 2 ) . However this field is.
1. This means that the axioms A3. b E K the M5 : For every a . The existence of (a) and a . The sign of inequality points towards the smaller element. It is obvious that A5 and M5 are satisfied in any field. D and the following: A5 : For every a. equation a i.] + + + 4. solution.2 Order axioms A field F is ordered if the axioms in Table 4.100 Introduction to Mathematics with Maple (a)b = a(b) = (ab). @ Exercise 4. The inequalities a 5 b and b 5 c are shortened to a 5 b 5 c and a similar convention is used if either or both of the inequalities is strict. denote it by z and show that z b = b for every b E K .4 Let K be a set containing a t least two distinct elements. (3) ac = bc and c # 0 imply a = b. (2) a + c = b + c + a = b .3 (1) If ab = 0 then either a = 0 or b = 0. respectively. M4 can be replaced by A5 and M5. A4 and M3. If a < b we say that a is less (or smaller) than b. b. if b < a we may also write a > b and say that a is greater than b. @ Exercise 4. Some easy consequences of the axioms are stated in the next Theorem. A2. The inequality a < b is called strict inequality to distinguish it from the inequality a 5 b. and the definition of a field retains the same meaning. MI. and negative if a < 0. If 0 < a . .= b d bd ’ (6) (7) (8) * (11) al #o. c in F . (a)(b) = ab.x = b has a the equation ax = b has a solution. then a is said to be positive. Assume that addition and multiplication on K satisfy axioms Al.+ . The statement ‘a < b or a = b’ is recorded as a 5 b. Show similarly that the solution to ax = a with a # 0 is the multiplication unit. Prove that K is a field. a c ace =bd bd’ a c ad+bc (9) .3 are satisfied for all a. [Hint: First solve the equation a x = a. b E K .1. a # 0.l is obtained by solving the equation x a = 0 or ax = 1 respectively. M2.
(v) (0 < a < b.3 Order axioms 01 : There is a relation < on F such that exactly one of the following possibilities occurs: 0 2 0 3 0 4 a<b. (iii) Successive applications of 0 3 give + + + 0 4 that ac < bc and u+c<b+c and b+c<b+d.1) . (4. (ii) If c < 0 then by (i) 0 < c and it follows by by (i) bc < ac. A consequence of (vi) is that 1 > 0.c. (vi) a # ~ + . (iv) ( a < b. 0 < c < d ) + ac < bd.< . a 1 1 (viii) O < a < b + O < .a . c < d ) + a . d denote arbitrary elements in an ordered field F .Real Numbers 101 Table 4. : ( a < b) and (b < c ) 3 ( a < c ) .b < . b<a. b a + Since bc < ac means the same as ac > bc we have the following Memory Aid for (ii): Multiplication of an inequality by a negative number reverses the inequality. (iii) ( a < b. Theorem 4. b. a. (ii) ( a < b. 1 (vii) 0 < a + 0 < .. c < 0) +bc < ac.1 (i) a < b + . Proof.( a b) < b . a=b.d < b . : If 0 < c then a < b + ac < bc. + + For the remainder of this section. c.O < a ~ . b) gives (i) Using 0 3 with c = (a b = u . Each of the parts of the theorem is proved separately. c < d ) + a + c < b d. and x is an element of this field which is (usually) to be found.( a b) = a. : a < b + ( a c) < ( b c).
102 Introduction to Mathematics with Maple (iv) (v) (vi) (vii) (viii) Applying 0 2 to (4. x>4. We distinguish two cases: (a) x > 2.4) > (X . Consider. 1 1 1 Since a . for instance. Consequently 0 < .4~ = X(X . This inequality only makes sense if x # 4. Thus x satisfies (4.# 0 and by (vi) . .2) means finding all x satisfying (4. b by b c and c by b d) gives the result. Maple can be used with advantage to solve inequalities.3) We rewrite the inequality in an equivalent form (x . a" a" la 1 11 1 By (vii) we have ..2) Solving (4. Consequently a a2 la 1 1 0 = :0 < a: = .2) is satisfied if and only if x2 .(a) < (a) (a) = a2.4) . If a < 0 then 0 < a and 0 = 0 . 1 We wish t o solve x2 + 6 > 52.a < a . We deal with this in Section 4.a = a2. an appeal to 0 4 and then to 0 2 completes the proof.1) (with a replaced by a c.> 0. b ab ab a + + +  Theorem 4. X 24 > 1.82 + 1 6 . d < c by (i) and using (iii) with c replaced by d and d by c gives the result. (4. If 0 < a then 0 = 0 .3) > 0.4)2 = x2 .1 can advantageously be used to describe the set of x E F satisfying a given inequality. 42 > 16.4.= .2). We have successively ac < bc and bc < bd. Example 4 .> 0.2)(x .= 1it follows that .= .> 0. (4.and a b a b ab 1 1 . For x # 4 we have (x .2) if and only if x > 4.1 1 =a< b . (4.4)2 > 0 and it follows that (4.
(2) and ( 3 ) hold then this .) C 4 1 Y L 0 X Fig.1 A graph of x2 . Summarising: Inequality (4. ( 3 ) If x and y are in P then so is zy.3) is satisfied if and only if the point (x. multiplying by that Inequality (4. + If. either x or . in a field F .x is in P. that is. (for x = 2 the inequality is obviously false).Real Numbers 103 (b) x < 2.1. (2) If x and y are in P then so is x y. x # 0.4) is satisfied by x if and only if either x < 2 or x > 3. In case (a) we find by dividing by x .5 2 +6 The set P of positive elements in F has the following properties: (1) For every x E F. a set P is given such that (l). Inequality (4.5 x + 6 lies above the xaxis. and since this provides no further restriction all x < 2 satisfy (4.y) on the parabola y = x2 . 4.2. In case (b) we obtain similarly x < 3 .4). This result is obvious geometrically.4) is satisfied if and only if x 3 > 0 22 and it is not satisfied for 2<x<3. (See Figure 4.
2. m[ = {x : a 5 x}.' b . (A) (A) (B) (B) (C) (D) (E) (F) a [ = {x : x < a}. x denote elements of an ordered field. b. (3) ( a < b .a.2. We shall not use this.d < b . 00 or oo are used for convenience of notation and have no meaning by themselves. @ Exercise 4. 1 1 (4) O < a 5 b * O < . 00[ = {x : a < 2). c 2 0 ) * ac 5 bc. b) to denote an ordered pair of elements . ( 2 ) ( a < b. Exercises In these exercises a.< . This is emphasised by the use of m[ or 300 to indicate the end of the interval. 300. The length of [a. d . the interval (B) is open. b) = {x : a < x < b}. It is not difficult to guess what the halfopen (or halfclosed) interval [a. b] = {x : a 5 2 5 b}. b] or ] a . b[ is b . It should be emphasised that we have not introduced any new elements into F . c. ] a . b[ = {z : a < 2 < b}. are called intervals. [a.2 Solve the following inequalities: (1) 32 2 5 22 . It is also worth pointing out that other notations are used to indicate open intervals. The proof of this is left as an exercise. < < Exercise 4. ] a . a] = {x : x 5 a } .a' ( 5 ) 0 < a b += a2 b2.104 Introduction to Mathematics with Maple field becomes an ordered field by declaring a < b if b . c < d ) * a . + + .1 Prove the following: (1) ( a 5 b. The interval (A) is closed closed.5 5 42 3. c < 0 ) * ac 2 bc. In an ordered field the sets [a.c . 100. implying that there are no elements in the intervals actually equal to these symbols. One common notation is ( a .a E P. a and b are called the end points of the interval. b[ is. since we have used (a.
if a = b we take Max(a. (2) 22 J : X + l < (8) :+3 (9) x 2 x x+5 x+6' [Hint: x2 + + 1 > 0.3 Prove that 2ab 5 a2 a = b. I Theorem 4. a). 1 (3) x + 5 2.2 (i) (ii) (iii) (iv) (v) (vi) la1 2 0. a).1 + b2 with equality if and only if 4. [Hint: Consider z > 0 and z < 0 separately. Rewriting (iii) in the form la1 5 a we can combine (ii) and (iii) into . naturally.b)2 2 0.3 Absolute value By Max(a.1 (Absolute value) The absolute value of a is Max(a. + x + 1 = (x+ 1/2)2 + 3/4. la1 = Max(a. I * All assertions (i)(vi) are easy consequences of the definition.] (4) (z + l)(2x . b and.a1 = lal. Absolute value of a is denoted by lal. that is. a 5 lal. . b) we denote the larger of the two numbers a. a 2 0 la] = a.] @ Exercise 4.1) > 0. Definition 4. b) = a = b. a 5 O + la1 = a. a 5 la[. [Hint: ( a . (5) ( 5 ~2 ) ( 3 + ~ 1) 5 0.Real Numbers 105 + 6 5 x + 10 5 22 + 1.2.
+ b and (a + b) do not exceed la1 + lbl. both numbers a and hence la + b) I la1 + lbl. 1 + lbl .b gives that is. + b I la1 + lbl .a1 = la . Interchangllig a and b in this inequality gives lbl 1.1 (A) a 2 0.lbl) do not exceed la .bl + PI 7 (ii) Using (i) with a replaced by a . + b( = Max(a + b.106 Introduction to Mathematics with Maple (i) Combining (4. a This implies (a That is.1 5 J b . I4 I la .bl  (iii) The proof is simple when one distinguishes three cases: .5) with gives (la1 + PI) I a + b L1 . Both numbers la1 . (a + b ) ) 5 la1 + lbl . b 2 0.Ibll 5 .Ibl and (la1 . .bJ .bl. b < 0 . (B) a < 0 . consequently 1 1 4 .
Theorem 4. . a 2 0.4 inequality If x .6.5) we have It follows from Inequality (4.6) and (4.at.Real Numbers 107 (C) one number. a E [ . particularly in Chapter 10.9) with (4. E. E are elements of an ordered field then the .7) Proof.1 is satisfied if and only if a€ .a ) < E.6). and so I za 1 is the distance of z from a.6) < x < U S €. (4. is not positive.a and (x . We split the proof into two parts.a < €. a .8) gives € < x . In each of the cases (A)(C) it is easy to verify (iii) of Theorem 4.a1 < €.3 and we leave it to the reader as an exercise.9) Combining (4. b. say a.2) + .4 says that points x having distance from a less than e constitute the interval ] u . Let us assume Inequality (4. (See Figure 4.6) that € < Iz . 0 The next Theorem will be often used.UI <€ (4. and (x Both numbers x lz . I. we can say la1 is the distance of a from zero. . The Inequality (4. is not negative. 11. Theorem 4.7) states that xa<e. b 5 0. (4.7). which is equivalent to (4. By Inequality (4. and the other number.a ) are less than and consequently 0 Resorting to geometrical illustrations of elements of F on the number line.
4. (4) (5) 1x1 I I x .2 Intervals on the real line Exercises Exercise 4.2 which each of the inequalities holds): (1) Iz . Open( 6)). 4. RealRange( Open( 9). Another example is . unknown) . (3) 0 < 1 2 1 1 < 4. RealRange(Open( 3).51 < 2. Solve the following inequalities (describe the set for Exercise 4.108 Introduction to Mathematics with Maple Fig.4 Using Maple for solving inequalities Maple can save a lot of mechanical work when solving inequalities. 00[. (7) + + + + + + 4.a1 2 6 .SI < 2.2)/ for x = 0. 6[ and ] .3. 00) This answer means that the inequality is satisfied for all J: in the intervals ] .82 10. 3.1)(x . 32 1 < 1 x 1 < 32 1.5. 5. For instance ~ ~ ~ ~~ > solve ( (x+l) / (x+3)< ( x + 5 )/ ( x + 6 ) 1. 1. .9.1 Evaluate I(x .3. The form of the command is solve(inequality.21 1.3. (6) 1 x 1 1 2 . (2) I x1 1 2 < 2. The parameter unknown is an option which can be left out if only one variable occurs in the inequality.
We recommend that this option is always used when solving inequalities. RealRange( Open(4.0.(~)) .0) . but enter the numbers as floating point numbers. Most of our readers would find solving the next inequality difficult but Maple has no problems. > solve(abs(x~28*x+10.707106781 < x}.240370349)). this tells Maple to provide the solution as a set of inequalities.0. RealRange(Open(4 + h). The solution set consists again of two intervals. enclosing the variable in braces.707106781).7596296508 < x. {.7596296508). The sign ‘less or equal to’ is entered into Maple as <=. Open(4 + 2 h)). x < 3.Real Numbers 109 > solve(abs(2*x~216*x+21)~5). 5) (5. The order in which < and = appear is important! We do not expect a neat result and therefore enter the numbers as floating point numbers.5)~5. If we use the option unknown. Open(4 . Open(3.d)). ( x ) . In order to obtain a better idea of what these intervals are we repeat the question.4.6823278038 5 x} Solving inequalities involving polynomials of degree three might involve . Similarly the sign ‘greater than or equal to’ is entered as >=. {x < 7. RealRange(Open(4 .292893219)). > solve (abs (xn28*x+10. {.2 h).240370349. RealRange(Open( . > s o l v e (xA3+x>=i. Open(7.292893219}.
solve (b*x+3<2*b.3 below: the solution set is the set of x where the parabola lies above the line y = 1 and below the line y = 3x . {x}) . {x < 4 .6).6 . {signum(b) x < signum@)(3 b + 2 b) 1 From this it is easy to read the result: if b is positive the inequality is satisfied for x < (3+2b)/b.3*~6}.y=6. if b is negative it is satisfied for x > (3+2b)/b.2x .{XI) .6. > ineqs :=(1<xA22*x2. A system of inequalities like a < b < c must be entered in the unabbreviated form as a < b and b < c. It is.6 is satisfied if and only if 3 < x < 4. and neither can Maple. # It is advisable to enclose the unknown in braces. The next example shows that on occasions a small amount of additional work is needed. If there are several inequalities to solve they should be entered as a set. this topic is dealt with in Chapter 8.) In this particular case solving with paper and pencil is simpler. xA22*x2<3*x6}: > solve(ineqs. So far Maple has provided complete answers.. > plot({l. (For b = 0 it does not make sense to solve the inequality with respect to x.2x .~=6. however.6.. This denotes the set with one element.2 < 32 .110 Introduction t o Mathematics with Maple solving an algebraic equation of degree three. 3 < x } These results mean that the system 1 < x 2 . so the option ‘unknownJ is now important.2 < 3 2 .~~22*~2. the set of real x less than 1. surprising that Maple has difficulties with . The following example illustrates the result from Maple when we solve the system of inequalities 1 < x 2 . It is important to understand Maple’s answer like { x < 1). ~~~ ~ ~ A mathematician cannot solve every and all problems. > > > > # We a r e going to solve an inequality which involves a # parameter. The set { x < 1) should not be confused with { x : x < l}. This is illustrated in Figure 4. namely the inequality x < 1.
Later versions of Maple may be more powerful. for example. { x } ) .Real Numbers 111 Fig. > solve (xA2>a. . 3 2 . x2 .' In this particular instance this is caused by the presence of the parameter a. This is an incomprehensible result but if we help Maple by making an assumption on a we get the correct answer: lThis was true at the time of writing this book.3 Plot of 1.6 solving some very simple problems.22 .2. 4. x2 > a.
The latter should be used before the command to which it applies and then it remains valid for the rest of the worksheet. A set S inductive if (i) 1 E S . + Obviously F itself is inductive.2 Use Maple to solve Exercises 4. {Ix < Ja}. { x ) ) assuming a>O. from a strictly logical view natural numbers are defined + . Since 1 E N and 1 1 = 2 E N by (ii).2. However the command assuming is worth remembering. {Ix < G} The correct answer is that the inequality is always satisfied. We may describe this by saying that N is the smallest inductive set.3.{x}) assuming a<O.1 Exercise 4. Use Maple to find the unions and the intersections of 4. The intersection of all inductive sets in F is denoted by N. {x < &}.5 Inductive sets In this section F denotes an ordered field. (ii) x E S + (x cF is said to be + 1) E S. > solve(xn2>a. Exercises Exercise 4. and then again by (ii) 1 2 = 3 E N it is intuitively clear that N must contain all natural numbers.2 and 4. Maple knows also assume.6 1 Unfortunately the command assuming does not help in the next example. At this stage we recommend to our readers not to use Maple for solving inequalities which involve some parameter (like a in our case).2. {x < . However.4.112 Introduction to Mathematics with Maple > solve (xn2>a. the difference between assuming and assume is that the former applies only to the command which it follows.4.
We restate that N is the smallest inductive set as Theorem 4. then M = N. 1 E S .m 2 1for n E N}. that is. Since 1 < 1+ 1 we have 15 n + 1. We proved Theorems 4. by Theorem 4. M is inductive and Theorem 4. /I 1 E M and if Proof. It suffices to show that S = N. everything else is the work of man”. the natural numbers are fundamental.5 gives M = N. We define S = {m.7 I f n E N . mE N and m < n then n .5 (Principle of Mathematical Induction) a subset of N with the following properties (i) 1 E M . sincern E S i t follows that n1m 2 1. This means that n > 1a n d m < n1. Proof. As an example of its application we prove Theorem 4. 0 in turn. n 1 E M . n . n E N.5 then M = N.5 is almost a trivial consequence of the definition of N it can be useful. Obviously Theorem 4. + We are now in a position to start building the theory of natural numbers. means that 15 n for every n E N. The German mathematician L.m < n implies n . m E N. 1 5 x}. Firstly. Now let m E S and n E N.m 2 1.1 2 1).5 and 4. m 1 < n. there is no element of N between m and m + 1. This reflects two things. however we shall not pursue this.6 because they are very important for some fundamental results of the next section (especially Theorem 4. + + 1 5 n + 1. 0 This last inequality shows that m + 1 E S. For that we consider M = (1) U { n . and (ii) n E A 4 + (n I f M is + 1) E M for every n E M . N. Although Theorem 4.6 1 5 n for every n E I/ N. Kronecker once remarked “God made the natural numbers. We define M = {x : x E n E M then 1 5 n and 1 that is. It is easily verified that M is inductive. First we prove that 1 E S. We do this by showing that S is inductive. we cannot explain . In other words. but simply leave some indications in the exercises how this theory can be established in our framework.Real Numbers 113 as elements of N (in this book). This.10: The Well Ordering Principle).
bEN } . Secondly.5. Exercise 4.1 4. Peano and the development indicated above is presented in Landau (1951) and Youse (1972).5. Recall that we also set z = abl.5. the rationals and the real numbers themselves can be defined and their respective theories built on the foundation of the theory of natural numbers. Exercise 4. xEN}.5.] + m E N. namely the axiomatic system for the reals.8. However.6. after accepting natural numbers (for example. N o = N U (0) and P = {z. J: > 0 ) . [Hint: Use Theorem 4. on an axiomatic basis). Having defined N we can define Z and Q as Z=NU{O}U{z.] Exercise 4. the negative integers. Exercises Exercise 4.1 Prove: k E N and m E N imply k [Hint: Use mathematical induction. . we briefly outline the Peano approach in Section 4. The axiomatic theory of natural numbers was established by the Italian mathematician G . a E z.2 Prove: k E N and m E N imply k m E N.3 Prove: k E Z and m E Zimply k + m E Z and k m E Z.6 The least upper bound axiom We start with a definition. Q = {z. created for us by previous generations of mathematicians. [Hint: Use mathematical induction.4 Prove: n E N + n 5 n2. In this book we have chosen a different starting point.114 Introduction to Mathematics with Maple what they are.
the least upper bound.2 Let S = n2 5n 3 n c N } . Is there a smaller upper bound? It is easy to see that the answer is ‘yes’. Example 4. The number If we illustrate the elements of a set S as points on a vertical number line then we can imagine an upper bound K for S as a kind of obstruction which the elements of S cannot pass. Geometric intuition leads us to believe that K can be moved to its lowest position.2 (Bounded sets in ordered fields) A set S in ordered field F is said to be bounded above if there exists K E F such that 2 for all x E S. they stay below K . n2+n+I + = 3. S is said to be bounded if it is bounded above and bounded below. Taking into account that n 5 n2 we obtain s there an even smaller upper bound? This time The natural question now is: I the answer is ‘no’. set S is bounded if and only if there exists K1 such that for all x E S . n2+n+I’ n2+5n+3 51+5+3=9.Real Numbers 115 &finition 4. S is said to be bounded below if there exists k such that for all II: E S . in more usual words. The element K is called an upper bound for S and k is called a lower bound for S . For the ‘only if’ part choose K1 = Max( IK I Ikl). thus becoming the least upper bound. The ‘if’ part is obvious with K = K1 and k = K1. Clearly. n 2 + + M is bounded above since n2+5n+3 n2+n+I < Thus 9 is a n upper bound. because for n = 1 we have n2 + 5n 3 is the smallest or. .
with s1 < s2 then by (ii) there will be u E S such that s1 < u(< s2).116 Introduction to Mathematics with Maple Definition 4. while (ii) means that any number strictly less than s is not an upper bound.4 (Complete field) An ordered field is said to be plete if every nonempty set bounded above has a least upper bound. It can be proved that there is essentially only one complete Remark 4. Definition 4. For any x E R there always exists n E W such . because whatever we prove will be valid in all of them. because it was the largest element in S.5 (Real numbers as a complete ordered field) The real number system is a complete ordered field.3 (Least upper bound) An element s E F is said to be the least upper bound for S c F if (i) x 5 s for every x E S. Whenever a set M has a largest element then this element is the least upper bound. The symbol s u p s denotes the least upper bound (supremum) of a set S. x < 0) we obviously have s u p s = 0 and 0 4 S. Theorem 4. for a set S . say s1 and s2. Spivak (1967) contains a precise statement of this result as well as the proof on page 507. that is.2.5 as The Least Upper Bound Axiom Every nonempty set S is bounded above has a least upper bound. Definition 4. This last inequality contradicts our assumption that s1 is an upper bound for S. A set S can have at most one least upper bound. Property (i) states that s is an upper bound. However. cI R which The number 3 was the least upper bound for S in Example 4. As agreed previously the set of all reals is denoted by R. x E R. The phrase ‘least upper bound’ is sometimes abbreviated ‘lub’.8 that x < n. However the least upper bound of S need not belong to S: for the set S = {x. the least upper bound of S belongs to S then it is the largest element of S. if there were two. We reformulate Definition 4. we need not worry unduly now how many complete ordered fields there are. The Latin word supremum is used with the same meaning as least upper bound. s is the smallest possible upper bound.1 ordered field. (ii) if t < s then there exists a E S such that t < u(< s ) . Also if.
1. contradicting Theorem 4. The symbol inf S denotes the greatest lower bound (infimum) of a set S.10 (The Well Ordering Principle) empty set M c N has a smallest element. The Latin word infimum is used with the same meaning as greatest lower bound. Hence we would have i 5 n < k < i 1 and consequently Ic .6 (Greatest lower bound) An element m E F is said to be the greatest lower bound for S c F if (i) m 5 x for every x E S. 0 Proof.7. Theorem 4.2. + Theorem 4. that there is K E R such that n 5 K for every n E N. (ii) if m < t then there exists a E S such that m 5 a < t. An example of an ordered field which does not have the Archimedean property is given in Exercise 6.9 Every nonempty set S C R bounded below has a greatest lower bound. We now prove indirectly that Ic is the smallest element of M . Then N has a supremum s.8 can be rephrased by saying that N as a subset of R is not bounded. Such fields obviously cannot be complete. This is often referred to as the Archimedean property.1 < k It follows that s<k 1 E N. the proof is very similar to the above given proof of uniqueness of the least upper bound. + + . It is now easy to check that . 0 Proof. Definition 4. A set S c R can have at most one infimum. There are ordered fields which do not have the Archimedean property. It is not difficult to see that M is nonempty and bounded above. then there exists k E N such that s .n < 1. If it were not. Every non A 4 is bounded below and nonempty.Real Numbers 117 Assume. Define M = {x : x E S}. Choose t = s .and consequently s is not an upper bound for N. which is a contradict ion.s u p M is the greatest lower bound for S.10. 0 Theorem 4. contrary to what we want to prove. there would be n E M such that n < Ic. By the second property of the greatest lower bound there is a k E M such that i 5 k < i 1. Proof. therefore it has a greatest lower bound i. By the least upper bound axiom there exists s u p M . The phrase ‘greatest lower bound’ is sometimes abbreviated ‘glb’.
0 fi + 1x1 is also called the integer part of x. 151 [:I = 0. 0 Proof. .12 that a < r < b. It follows that a = .11 it is possible to prove that for every x E R there is a unique integer n such that n .12 is sometimes described by saying the set of rationals is dense. with n < f i . for if n 5 x < n 1 and 5 x < f i + 1. since we can take for k any of the infinite number of integers greater than 2 / ( b .118 Introduction to Mathematics with Maple Theorem 4. if x > 0 we take n = k . We obviously have = I. + Theorem 4.< .. I f a E R. etc. so we assume x 4 Z.n 5 x . This number is called the ceiling of x. denoted in Muple by ceil (x) . so we have a contradiction with Theorem 4. If x E Z there is nothing more to be proved. The mathematical symbol is 1x1. if x < 0 we take n = .n E N . Since N is not bounded above there is k E N such that k > * k Actually we have proved that there are infinitely m a n y rationals lying between any two reals. 101 = 0 .< .11 For every x E R there exists a unique n E Z such that n 5 x < n+ 1. The function x H 1x1 is called the greatest integer function. The number n (if it exists) is unique.2 Similarly as in Theorem 4.E Q and a < r < b. On the other hand.< k 2 k 2 a+b ba m+l 1m m+l a + b < b. = 1. we would have f i .n < 1. This number is called the floor of x. fi .< 2 2 k k k 2 k m We have that r = . Proof. Remark 4..k .1. 1.. Define m = ' k ] and we have .1 < x 5 n. For x E Z we have 1x1 = otherwise 1x1 1 = 1x1. it is denoted by f l o o r ( x ) in Maple. b E R and a < b there exists r E Q such /I The content of Theorem 4.7. because the points for Q are so densely packed. . .1.< . If we try to illustrate graphically the set Q on the number line we would not be able to distinguish between Q and the number line R itself. The set of natural numbers strictly greater than 1 x 1 has the smallest member k .a ) . 2 ba' 1 ba m a+b that is.and is denoted by 1x1.
3n2 . Use the same idea as in the proof of Theorem 4. and find sup S and inf S. + Exercise 4. . however. @ Exercise 4. (3) x (4) x ( x . (4) S = {x.5 nEN} n4 . Prove this.12 and apply the result of Exercise 2.6.7 Graph the following functions.6 Prove that if a < b there exists an irrational x such that a<x<b.1x1. Prove this.4. S1 # 8. preferably without the use of Maple. then sups1 5 sup S 2 and inf S 1 2 inf S 2 .Real Numbers 119 Exercises Exercise 4. 5’2 bounded.1 Exercise 4. Show that the set S = Exercise 4. x2 52 < 6.3 If S 1 and S2 are bounded then so is S 1U S 2 . for every x E X there exists a y E Y such that x 5 y then sup X 5 sup Y .lxJ)2. and also show that in the second case sup X could be greater than inf Y .6.6. If. 1x1 + 1 4 . [Hint. (2) X H X .1 Find sup S and inf S if n (2) s = 15) u {3)u10. Give an example of an S1 which is a proper subset of S 2 but for which sup S 1 = sup 5 ’ 2 .5 If X and Y are bounded nonempty subsets of R and for every x E X and every y E Y one has x 5 y then sup X 5 inf Y . (3) s = (5) u {3)u]4.6.6. Prove this.4 If S 1 c S 2 c R. 5[.6.4.5’ is bounded from above and from below.6. x E R. @ Exercise 4.2 n2 . (1) X H 1x1. @ Exercise 4.
6.< n 1. The definitions (4. It is customary to define the sum. not necessarily of R.10) f g :x f(x) .120 Introduction to Mathematics with Maple [Hint: For (1)(3) consider separately the intervals [n. For instance in 1 = n if n 5 . H ‘ 9 f(x) x E domf n domg n {x. LxJ + 1 .6.7 Operation with real valued functions In this section we assume that the ranges of all functions appearing are subsets of R. x E domf n domg. Even if you are using Maple part (6). If f < g and domf = domg then the graph of f lies entirely below that of g. g(x) # 0).x) is called the distance from x to the nearest integer.9 Graph he hump function h. during a discourse.] + I:] X + Exercise 4. 9(4’ Instead of ff we shall write f 2 . (4. Exercise 4. +1 1+ 1 x 1 + 111: 1 1 4. .we have look at the intervals separately. x E domf n domg. f(x)g(x). id2 : x H x2. n 1[ with n E N. x E d o m f n d o m g . Graph the following functions: (1) x I+ saw(x).g(x). difference. and denoted by saw.8 The function x H Min(x . product and quotient of functions as follows. We shall write f 5 g (or f < 9) to mean f ( x ) 5 g(x) (or f ( x ) < g(x)) for all x E domf n domg.10) make sense if the ranges of f and g are part of some field. f +g :x fg : x f :x f(x) + g(x). The other parts of the exercise should be treated similarly. If. h(x) = Iz with Maple and also without it. there is some underlying field. definitions (4. For example.LxJ. (2) x I+ saw(2x).10) are assumed to hold automatically.
. In other words the mapping x H x ' is onetoone. Dedekind cuts The axioms listed earlier in this Chapter.4 that the field axioms can be formulated differently. We did not employ it because it is long and laborious. Peano axioms. The set will be henceforth denoted N. diametrically opposite to ours: instead of considering the natural numbers as a subset of the reals it starts with the natural numbers and builds up to the reals. In words. 3 comes after 2. We have already seen in Exercise 4. etc.3 and the least Upper Bound axiom treated in Section 4. n 1 comes after n.1. Peano 2: For each x E N there exists one and only one element 2'.6 are not the only possible system of axioms for the reals. The element x' will be from now on called the successor of x: after each element of N there is one which immediately follows it.8 Supplement. Peano 4: If x ' = y' then x = y. Peano 3: . Different approaches to the real number system are possible. the order axioms listed in Table 4. This axiom describes in abstract the idea of counting: 2 comes after 1. x ' # 1 f o r every x E N. This means: there is at most one x for a given x'. There are five Peano's axioms. We briefly indicate this approach leaving many proofs aside.1.Real Numbers 121 Exercises 4. from a purely logical viewpoint. It starts with what are called the Peano axioms for the natural numbers. The one we shall consider in this section is. the axioms for a field given in Table 4. However the difference between the two sets of axioms there was minor. this approach may be preferable. the successor x' comes after + X. from a philosophical point of view. Peano 1: There is a nonempty set with a distinguished element 1. 1 is not the successor of any element in N. Perhaps.
We denote P pairs of elements of Z as p / q or . Q). y E N. We immediately recognize this axiom as Mathematical Induction: in our approach this is Theorem 4. noting in passing that it can be shown that the elements of Z thus defined satisfy the axioms for a ring (see Table 2.rather than ( p . it is symmetric. require proof. that is. and (ii) n E M + (n + 1) E M for every n E M . of course. in other words introducing 0 and the negative integers. Two such pairs $/i and p / q with i # 0 and q # 0 are declared equivalent if $9 = P i (4. For instance. Note that in our development Mathematical Induction was derived as a theorem.The definitions are . henceforth denoted x y such that + (i) x (ii) x + 1= x ‘ for every x E N. + y‘ = (x+ y)’ for every x. as well as the definitions of multiplication and order. which we omit but mention that axiom Peano 5 is essential for the proof. though what follows 9 here can be written in terms of ordered pairs if desired. then M = N. but pause at the next step. y there exists a uniquely determined natural number. As is customary we identify equivalent pairs: two equivalent pairs are simply equal.3). We skip over this. reflexive and transitive.1. while here it is one of the fundamental axioms. multiplication and order in N and establishing the ‘usual’ properties of natural numbers. This. the creation of Q. additionof x and y is defined as follows: To every pair of natural numbers x.5.11) This relation is indeed an equivalence relation according to Section 3. Then the theory moves to extend N to Z. With these five axioms at hand building up the theory of natural numbers proceeds by defining addition. The set of pairs p / q with p E Z and 0 # q E Z for which addition and multiplication is defined below is denoted by Q.122 Introduction to Mathematics with Maple Peano 5: If M is a subset of N with the following properties (i) 1 E M .
.Real Numbers 123 multiplication q s T =E. As a sample of such proofs let us prove axiom D.. There is an .pru +qsu qsu + pts = pr qsu qs pt + . representation of the pairs. the definitions are independent of the choice of p r 6 r p r fir and . It is obvious that addition and multiplication of elements in Q is commutative and there is no difficulty showing that all the other axioms from the Table 4.= .= q s q s q s 4s' To prove this. Firstly with q # 0 and s # 0 we have qs # 0. we first note + + It follows that From symmetry arguments we have also if rg = ?s. qu At various places in the above proofs use is made of the equivalence relation among pairs of rational numbers expressed in Equation (4.1 are also satisfied. qs These definition are meaningful.11). Secondly. This shows the correctness of the definition of addition. proceed in the same manner and the resulting field is called the quotient field of M. The construction of the field Q from the ring Z can be applied more generally: instead of starting with Z one can take a ring M. p ru + ts pru pts . The proof for multiplication is similar and left to the readers. If p / q = $14 then .
otherwise B nA # 0 and then BIB 4 AIA. of course. 2This is. *With F replaced by Q in 0 1 . A of Q is called a (Dedekind) cut if the following conditions are satisfied: C1 Both sets are nonempty and Q = A u A. We now prove the least upper bound theorem for cuts. Interpreting this may be easier if the reader makes a sketch. For cuts an order relation. C2 If a E A and ii E A then a < ii.4 We have now almost reached our goal: we have in Q an ordered field. q E Z and q # 0 is defined to be positive if p q E N.2 After the field axioms are shown to hold in Q derived in this fashion. AIA = BIB. If A = B then we have equality. In our development the existence of the least upper bound was guaranteed by an axiom: here it becomes a provable theorem. C3 If a E A then there exists a’ E A satisfying a < a‘. Such a cut is denoted by AIA. one and only one of the following is true:5 AIA 4 BIB. and A the upper section. 5AlA = BIB means naturally that A = B and A =B . and it remains to extend it to a complete ordered field. The set A is the called the lower section of the cut. which later will be renamed real numbers. Property C3 simply means that A does not have a largest element: A may or may not have a smallest element. the symbol I denoting the ‘cut’ in Q between A and A. BIB 4 AIA. it is time to introduce order in Q. A pair of subsets A . using the symbol 4. If B n A = 0 then A c B. namely ab = 0 must imply that either a or b is 0. This can be accomplished by what are known as Dedekind cuts.3. This order has the trichotomy property. If B n A # 0 then the first order relation holds. This is fairly intuitive: BIB is bigger than AIA if B reaches over A into A. is defined as follows: Order AIA 4 BIB if and only if B n /I # 0. 3See page 103. Axiom 01 in Table 4.a ) E P. The relation a is less than b is then defined by declaring a < b if and only if ( b .3: for two cuts A ) Aand BIB. satisfied in Z.124 Introduction to Mathematics with Maple additional requirement on M for this construction to work. This is done by defining the set P of positive element^:^ p / q with p . It is not difficult to prove that this relation satisfies the order axioms in Table 4.
we have X n S # 8 and this is impossible. Then. of course. It suffices to show that @ n S = 0. Step 2. We denote by 6 the cut 2 1 2with 2 = {z. Proof. is S. by the definition of order in G . Proof of (i). The cut 0 is the zero element for addition: for every cut AlA we have 'This. If s E S and S E S then s E X I for some X l l X l E 6 and 3 is in no X for X l X E G. but we omit the proof as we did with many other proofs in this section. . that is Q\ S . Step 6. then there exists a such that cut s ~ S (i) if X l X E G and X l X # SlS then X l X 4 SlS. that is X l X 4 MIA? for every X l X E G. Condition C2. By assumption S n T # 8 which means there is an X l X E 6 with X n # 0. By the definition of order in 6 this means a TIT 4 X ( X . z < 0) and 2 = Q \ 2. Consequently 5 E X 1 and therefore s < 3. Definition of SlS. Condition C1. 0 Addit ion Let Y = { a + b. Step 4. Assume contrary to what we want to prove that SlS 4 X l X . b E B } and define = Q\Y. If s E S then s E X2 for some X2lXz E G and there is some s" E X 2 c S with s < s". Let X l X E G and X l X # S ) S . Step 5. The complement of S . a E A . Proof of (ii).13 (The Least Upper Bound Theorem for cuts) I f 6 is a nonempty set of cuts and ~ ( is f such i that it exceeds every cut in G. must be proved. Step I . The set S consists of all rationals which belong to a lower section of a cut in 6.Real Numbers 125 Theorem 4. Step 3. We now define addition and multiplication for cuts. Then YIP is a cut' and we YIP Ef AIA + BIB. (ii) if TIT 4 SlS then there exists X l X E G such that TIT 4 XlX. If this were not true then n X O # 8 for some Xo(X0 E G but then MI@ 4 XolXo contradicting the assumption that X ( X 4 MI@ for every X ( X E 6. To make the proof easily readable we divide it into several easy steps. The next three steps prove that SlS is a cut.xE Q. since X C S. Condition C3.
12) (4. first for positive cuts and then generally. so it is a bijection. X l X . X ( X 4 YIY H x* < y*. it is possible to prove that the cuts form a complete ordered field.14) show that the mapping X l X H x* preserves addition. it can be shown that X l X + YIY I+ x* + y*. It is defined in stages.13) (4. Let X be the set consisting of negative rationals. and we define multiplication by Clearly the restriction to positive cuts was necessary.1). Then X l X is a cut7. if x 4 6 and 0 4 y. multiplication and order.0 4 (see Table 4. Multiplication: positive cuts Let AIA and BIB be positive cuts. b positive. Multiplication: general cuts If z. but unfortunately multiplication is not as easy. With these definitions. Define X = Q \ X .y are cuts then xy is defined by xy = 0 zy = ((x)y) xy = (z(y)) xy = (x)(y) i f x = O or y=O. This is not very difficult but could be tedious: it also involves defining the cut x (used in the definitions of general cuts) and the unit cut for multiplication.12) to (4. Two fields F and Fl are said to be 7Again. We say that a cut AJAis positive if 6 4 AIA. When is a cut a rational number? If X contains a smallest element x* then the mapping X l X I+ x* is clearly onetoone and onto Q. that is they satisfy Axiomss 0 2 . The definition of addition of cuts was simple enough. i.14) Relations (4. b E B and both numbers a.y*.126 Introduction to Mathematics with Maple AIA+0 = AIA. zero and all numbers of the form ab with a E A . (4. if x 4 6 and y 4 0. if 0 4 z and y 4 0. this needs to be proved We have already established Axiom 0 1 . which simply means that all elements of A are positive. otherwise the definition of X l X as a cut would not be meaningful.3) and all the field axioms (see Table 4.YIY H x* . Moreover.
neither is it possible to establish the consistency of the axioms of set theory. From a philosophical view this is the main difference between the two approaches. Our outline of an alternative route to reals has come to an end. Cuts and real numbers There remains the case where. as Godel showed. The bijection is called an isomorphism. so we now ‘have’ the reals. but not in substance. If Peano’s axioms are consistent then so are the axioms for reals. This is natural: in two isomorphic structures. They may be different in appearance. In accordance with this. Frequently in mathematics isomorphic structures are identified. Summary of Peano’s axioms It is not possible to establish the consistency of Peano’s axioms but. . However. We have constructed a complete. we say that those cuts for which the upper section contains a smallest element are the rationals.Real Numbers 127 isomorphic if there exists a bijection of F onto Fl which preserves addition and multiplication. the construction of the reals by cuts shows the relative consistency of the axioms for the reals. an irrational real number). From a mathematical point of view they are equally valid. for the cut X l X . the set X does not contain a smallest element. Two ordered fields are isomorphic if there is a bijection of one onto the other which preserves addition. We then use this cut to define a real number (in fact. ordered field which contains the rationals as an ordered subfield. multiplication and order. elements act the same way in both structures.
This page intentionally left blank .
The following example shows that we encounter inductive reasoning also in mathematics. an ornithologist watches birds of a certain species and then draws conclusions about the behaviour of all members of that species. 129 .1 Inductive reasoning The process of deriving general conclusions from particular facts is called induction.Chapter 5 Mat hernat ical Induction In this chapter we study proof by induction and prove some important inequalities.n is a multiple of 10. particularly the arithmeticgeometric mean inequality.n for the first few natural n n5n 1 2 0 30 240 1020 3120 7770 16800 3 4 5 6 7 It seems likely that for every n E N the number n5 . Basic properties of powers with rational exponents are also established in this chapter. General laws of motion were discovered from the motion of planets in the solar system.1 numbers: Let us consider the numbers n5 . 5. For example. It is often used in the natural sciences. Example 5. In order to employ induction for defining new objects we prove the so called recursion theorems.
54 1 The last inequality holds if and only if n < 625.1) we could spend all our life (or even all the life of the universe) testing successive particular cases for n and we would never discover that . ) . as we shall shortly see in Example 5. our conclusions are subject to further observations and experiments (devised to check the conclusions). that does not prove that it is true for all natural numbers. Example 5. It may not be true for n = 8.1. though you can easily check that it is.! j is positive.1) represents a statement valid for the first 625 natural numbers but not for all natural numbers. basing arguments on a finite number of examples is an uncertain procedure.1) holds for the natural number n if and only if 1 5 n 5 625. Indeed.1) * If we substitute for n quite a few of the early natural numbers we see t h a t the inequality holds.1) holds for all natural numbers n. Even if you have used a computer to check the first billion natural numbers.2 Consider the inequality 5n+$ 1 1 >1+5n 55 (5. but we made conclusions about the validity of the formula for infinitely many cases. and in the natural sciences.2. and there is yet another important difference. For natural numbers n the (rational) number 5 n . However it would be wrong to conclude that (5. In everyday life. If we considered the inequality instead of (5. and therefore Inequality (5.130 Introduction to Mathematics with Maple The reasoning in the above example does not give us the feeling of castiron certainty which mathematical arguments usually have.1 we observed a few particular cases. In Example 5.1+2 2 ( 5 : ) >n. and it can lead to serious mistakes. + . We summarise this as: Inequality (5.i ) and this inequality holds if and only if ( I t . In mathematics this additional check is missing.1) holds if and only if 5n+L 2> ( 5 n . Inequality (5.
1.3 clearly demonstrates the necessity of step 1.n is a multiple of 10. but example 5.3 Let us 'prove' by induction that for every n E N the number 2 n 1 is even.9 and 5.n is a multiple of 10 }. then it also holds for the next natural number. See also exercises 5. + + . We can now summarise: A proof by mathematical induction has two important parts: (I) A check that the proposition is valid for the natural number 1. particularly because it is usually the more difficult one. Something more than testing the first few (or the first few million) natural numbers is need to prove that n5. By expanding (n 1)5we have + + + (n+ 1)5 . the Principle of Mathematical Induction. Clearly 1 E M .2) is not true. The particular starting value in the first step need not always be 1 as the next example shows. then so is ( n 1)5. and so 5 n ( n 3 + 1) is indeed a multiple of 10. and 10(n3 + n2) is obviously a multiple of 10. Now we want to show that if n5 . The assumption made in the second part. It is not likely that somebody would make such a blunder as asserting that 2 n 1 is even for some particular n E N.n + 10(n3 + n 2 ) + 5 n ( n 3 + 1). + + + + + + + Of course.1. Let M = (n. namely that the statement holds for some natural number is often referred to as the induction hypothesis. The second step is often called inference from n to n 1. The key is Theorem 4. ( 2 ) A proof that if the proposition holds for any natural number.n is a multiple of 10 for every n E N. n5 . However. It is natural to concentrate on the second part. + Example 5. that is. We are assuming n5 . Assuming that 2n 1 is even. even if we consider a lot of particular cases.10. 2 n 1 is never even. 5. By the Principle of Mathematical Induction it will follow then that M = N.1.Mathematical Induction 131 there are natural numbers n for which (5.n is a multiple of 10.n is a multiple of 10. it suffices to show that 5 n ( n 3 + n) is a multiple of 10. neglect of the first part can also spell disaster.8.(n + 1) = n5 . We have seen in Example 5.(n 1).2 that the first part by itself is insufficient.n5 . we obtain then 2 ( n 1) 1 = ( 2 n 1 ) 2 is even because it is the sum of an even number 2 n 1 (by the induction hypothesis) and 2.5. n E N. One of the numbers n and n3 + 1 must be even.
and both x and y belong t o No. We now wish to emphasize the phrase any natural number in step 2. If. Clearly 1E S by the definition of S.3) has a solution X O . in turn. +1 The error was that the inference from n to n 1 was incorrect: if n E S then either n = n 1 or n = 1.4 We wish t o prove that the equation 2x+5y=n (5.n is divisible by 13 for every positive np n Exercise 5. if n = 1 there is no way of showing that n 1= 2 E S. + + + Exercises Exercise 5. Another amusing example with a similar twist is given in Exercise 5. Then n + 2 = (n+ 1)+ 1= n+ 1 . Now let n E S.3) is that any amount of whole dollars of a t least $4 can be paid with a mixture of $2 and $5 notes (or coins). Let S be the set containing 1and all natural numbers n such that n = n 1.2. Show that nI3 . then 2x0 = n 2 4. yo = 0.2 Prove that if p is a prime then for every positive integer n.1) is a solution of (5. on the other hand. n = n 1 . and consequently (n+ 1) E S. So.1 integer n.4) If yo 2 1 then it is easy t o check that + 3. that is.132 Introduction to Mathematics with Maple Example 5.1. Step 1: For n = 4 Equation (5. S = N. and so n must be even.3) has a solution x = 2.4) and both zo+ 3 and yo 1 are nonnegative integers. y = 0 .3 is divisible by p Prove the following formulae by mathematical induction: .4) and both 2x (20  belong t o No.1. yo in nonnegative integers for a particular value of n and consider the equation + 5y = n + 1. An everyday interpretation of (5. However.1.12. leading to xo . yo + 1 > 0 and (20 . Exercise 5. (5. n 2 4 such that x and y belong to No.and so all natural numbers are equal. by the Principle + + of Mathematical Induction. Step 2: Assume that (5.and this.5 We 'prove' by induction that all natural numbers are equal. yo . implies that n = n for every n E N. Consequently xo 2 2.3) has a solution for every n E N.1.2 2 0 . yo + 1) satisfy (5. Example 5.
1.Mathematical Induction 133 + + 32 + .1) (3) l 4+ z4 + 34 + .3+23. Prove that for every n E (15> N (l..). 4 (11) Suggest a generalisation of (9) and (lo). .4 Derive formulae (5)(8) in the previous Exercise from formulae (1) and (2).+ (272 .+ n4 = 30 n2(n+ 1 ) ~ ( 2 + n~ 2 n . Exercise 5. a2.n2). (8) 23 + + 63 + .( 2 n ..* . 2 + 2 ..+ n = (1) l 2 22 9 + + s2+ . + + + + + + + + Exercise 5.4+345+.] + + + + + + + + + @ Exercise 5.7 (1.+n(n+l)(n+2) = n(n + l ) ( n+ 2 ) ( n + 3 ) ..1)2 4(12 22 .+ 3 (7) l3 + 33 + s3+ ..1 . ( 9 ) I ...n(n 2k 1) .. (4) 1 5 + 2 5 + 3 5 + . + n(n + 1) = 3 (5) l2 32 9 12 n ( 2 n . by mathematical induction. : ) ] 2 .1).. and then prove the sug .5 If a l ..) Let k be a natural number... by using induction. .1 ) ( 2 n 1) (an .1. (2) i 3+ 23 + s3+ .+ 2 2 ” + ’ . ’ n(n + 1)(6n3 + 9n2 + n ..( 2 n ) 2 = l2 32 .(1 22”) = 1 2 2 ’ x3 * .4 + . an are real numbers then Prove this by using mathematical induction. 2 n ( n = + 7 (10) 1.6 Prove. + .(l.+ (2n)3 = 2 n 2 ( n + 112.1. .1)2= 7 3 1 ) ( 2 n 1) ...2. 7 7 gested formula.+ n2 = n(n + I)(%+ 1) ..+ + + ( 6 ) 22 + 42 + 62 + . [Hint: l2 2 ’ ..1 ) . + n3 = [ n ( n . ..q3= n 2 ( 2 n 2 .3 + 3 . Exercise 5. . n(n + l ) ( n+ 2 ) . where the right hand side contains all powers of x up to the highest shown.) > I .1. for x E R (1 2) (1 2 2 ) (1 2 4 ) . . .
Gk}. the induction hypothesis ensures that all these girls. . E Exercise 5. Combining this with the earlier result we see that all k 1 girls have blue eyes. . + Exercise 5.1. The ‘proof’proceeds by induction on the number of girls in the set. Prove this. x 2 r } . @ Exercise 5. Taking now the set (G2. 2 . Gs.1.13 Prove: if n E numbers q and r such that N o and rn E N then there exist unique n=mq+r. which completes the proof. which has only k elements and contains a girl with blue eyes. Prove this.1. Construct a formula valid for the first million natural numbers but not valid for all n E N. have blue eyes.m > 1 then n 5 mn. Consider now a set of k 1 blonde girls { G I . Now taking the set { G I .12 This puzzle is attributed to the Hungarian born American mathematician G.10 Use induction to decide for what values of n E N the inequality n 12 < n2 is correct. x Z.11 Ifm E N..8 The formula is valid for n = 1. 5 but not for n = 6. 4. has blue eyes. .1. and assume that at least one of them. M = {x. + + @ Exercise 5. @ Exercise 5. 3. Gk+1}. (2) k E M implies k 1 E M .9 Let IM c Z with the following properties: (1) M contains an integer r . . we deduce that all the girls in this set have blue eyes. . G2. The statement is obviously true if the set contains only one girl. .134 Introduction to Mathematics with Maple @ Exercise 5. Polya. Theorem: If a finite set of blonde girls contains one girl with blue eyes then all girls in the set have blue eyes. . + Then M contains all integers greater than or equal to r. Gk+l}. say GI. find the logical error on the above ‘proof’. Explain this paradox. . namely G2. Ga. . and using the induction hypothesis. . . including Gk+l.1. that is. . OLr<m.1.
If we try to prove the inequality 1 1 1 1. 2.1.4 + + 1 + . Clearly 1 E M . arc.4 1 * .2 21 . 1 . Let M = {n. 9}. 3. a + + n(n + 1) can easily be proved by induction... . .7) be true.4 + This proves that n 1 E M . a l ..7) holds for every n E N.5) by mathematical induction.13.2 Aim high! When proving theorems by mathematical induction.r E No..6) On the other hand. n E N.. with quotient q and remainder r . (This theorem expresses what you know from school as division of n by rn. and (5. . k . . such that E N then there exist numbers Ic E No withaiE(0.. then ( 5 . a + n(n + 1) < 1 1 (5.. ) @ Exercise 5. a2..3 + 31 . Let n E M .4 + * * + n ( n1 + 1) + (n + + 2) < 1. let (5. so by the principle of mathematical induction M = N and (5. 1 . i = O . we often encounter a seemingly paradoxical situation that a stronger theorem is easier to prove. .3 3. [Hint: use induction and Exercise 5.1.Mathematical Induction 135 and q E No. 5 ) is of no use for the proof that 1 +1. then 1 +12 1 23 +3 1 + .14 Prove that if n and ao.7) holds}.2 2. 1 1. the stronger statement that . (5.1 5.3 3. . 2 .. that is.2 1 + 21 + .
m < n we can first sum the numbers a l . and is sometimes referred to as a ‘dummy index’. The index i can be replaced by any other letter (or symbol) without altering the value of the sum or product. . and the numbers below and above the signs C and are called the limits of the sum or the product. . am and then the rest. The sum and product of the numbers ai are denoted respectively by The index i is called the summation or product index. i=l (5.136 Introduction to Mathematics with Maple 5. .3 Notation for sums and products Let n E N and a l . . . a2. . . a2. . so that we can write + n n n (5. so we have m C a i + i=l C i=m+l n n ai =C a i .9) from the left by setting i = m j and by changing the limits of summation accordingly. The upper limit is written simply n. The symbols C and also represent parentheses including any of the following terms which involve the summation or product index. but is to be interpreted as i = n.8) can also be rewritten in the form We obtain the right hand side of (5.8) The second sum on the left hand side of (5. Thus n n i=l n n t=l C a i = C a j = x a t .10) i= 1 I i=l . an be elements of a field. respectively. j=l If m E N.
Mathematical Induction 137 It follows from the distributive law that (5.16) . n. for j = 1. . 2. .11) If bl. . n and then sum these numbers and obtain n m (5. . . . and then sum these to obtain m m n (5. . . . . . m. 2. 2. . . .14) are the sum of the mn numbers aij it is obvious that m n n m (5. j = 1. For these numbers we can first form the sum . . .15) and either sum is often written (5. bn are elements of the same field then n n n (5. . ba. 2. . . i = 1.13) or we can first form the sum m j= 1 for i = 1.13) and (5. m.12) i= 1 i=l i=l Consider now the rnn numbers aij.14) Since both (5.
n i If we sum the columns in this expansion first and add these sums together we obtain n i n n Formula (5. j = l n aij (5.138 Introduction to Mathematics with Maple If m = n then (5. . for example.19) can be proved without explicitly writing out the terms as in (5.14) care should be taken to use distinctive letters for the summation index.16) can be further abbreviated to C i.18) by realising that each term in (5. For instance whereas The inner summation limit in a sum like (5.19) contains aij with i and j E N such that 1 5 j 5 i 5 n.14) can depend on the outer summation index. Denoting the summation indices by distinct symbols is also necessary when.17) When dealing with sums like (5. Thus we have. for example. two sums need to be multiplied.13) or (5.
we prove the next theorem.1 If n E N and a l . n. There are formulae for products similar to those for sums in (5.20) holds for some n E N. .9). (5. but. and should. .Mathematical Induction 139 as in Formulae (5. and leave a few similar proofs as exercises.20) is true with the equality sign. Since + + we have. . an are real numbers then (5. be proved by using mathematical induction.20) Proof. by Theorem 4. in fact. . 2. . n 1.12) are consequences of the field axioms. . and consider n 1 real numbers a k . .8). We do not wish to dwell on simple proofs such as these. a2. If n = 1 then (5. .11) and (5. Now assume (5.8). as an example of how these proofs could be arranged.2 and by the induction hypothesis. (5. k = 1. Theorem 5. .
140 Introduction to Mathematics with Maple (5.22) i=m+1 n n i=l n (5. i=l 0 (5. and to avoid continually having to consider these cases separately.12) and (5.27) and 0 IJ.21) i=l i=m+l n (5. . for instance.26) i=l i=l In order to preserve some formulae in limiting cases. . 2 . ai = c for i = 1.15).2 = 1. i=l i=l (5. . If. n then n e a i = e c = n c . it is convenient to introduce the void sum and product by defining Cai = 0 . .24) Care is needed in handling sums and products if the terms following the signs C or are independent of the summation or product index.28) i=l These formulae are purely conventional and are devoid of any deeper meaning. m n (5.25) Similarly n n (5. (5. respectively.23) i=l m n (5. .
Often when using the s u m 0 command.499205338 > > > > # Use a decimal p o i n t t o e v a l u a t e # the s u m of l / i from 1 t o 100. 4.5). .3.Mathematical Induction 141 5.1 Sums in Maple The s u m 0 function can be used to calculate sums.variable=m. the specified expression is summed over the given range for the specified variable. . .50). 55 > > # Evaluate t h e s u m of i n 2 + i from 1 to 5 sum(in2+i. 70 Recall that Maple gives exact solutions.100000). i = i . s u m 0 is being used on fractions. 13943237577224054960759 3099044504245996706400 > > > # Use e v a l f ( ) t o e v a l u a t e t h e same s u m # giving a decimal r e s u l t evalf ( s u m ( l / i .000 # g i v i n g a decimal r e s u l t s u m ( l .i=l.n) .. > # Evaluate t h e sum of i from 1 t o 10 > sum(i.i=l. you will need to use either evalf (1 to obtain a floating point result. i = l .lO). or a decimal point somewhere in the sum. In the following examples. > # Evaluate t h e sum of l / i from 1 t o 50 > s u m ( l / i . .50)). In the form s u m (expr . i = l . . . / i .
In the following examples.N).09014612 Rather than specifying numbers as the range for summation.i=l.142 Introduction to Mathematics with Maple 12.N)). > > # Evaluate the sum of i from I to N sum(i. you can also use variables.N 2 2 The above result is not in the form you will find quoted in textbooks or other reference works such as Gradshtein and Ryzhik (1996).27). so we try to simplify it. at least in cases where the second index is 1 less than the first index..i=l. the indefinite sum is calculated for i = 1 .(N 1 2 1 1 + q2. . For instance 0 0 . . N ) . # 1 1 1 3 N 3 + 5N 2 + s Maple honours our convention on void sums (5. > simplify(%) . . . 1N 2 + 2 1N 2 _ _ _ ~ ~ > > > Evaluate the sum of i2 from 1 to N # in simplified form simplify(sum(in2.
342847117 > > # Evaluate the product of (i+l)/(il) product((i+l)/(i1) ..i=I. > # Evaluate the product of l+l/i^2 from 1 to 10 > product (1+1/in2. > L:=[seq(ithprime(k) . 3.Mathematical Induction 143 Maple can sum elements of a list. The next few lines of Maple show how to evaluate a sum of numbers of the form l / p where p runs through the first 100 primes.. 1 N(N+l) 2 .lo) . 2. from 2 to 20 210 > > > > # # Evaluate the product of (i+l)/(il) for i ranging from 2 to N and reduce it # to what Maple thinks is the simplest form simplify(product((i+l)/(i1) .loo)] : > evalf(sum(l/L[i] . ..i=l..20) .2 Products in Maple The product0 function can be used t o calculate products. Its usage is very similar to that of the sum() function described above.3.N)).i=2.iOO)).106342121 5. > evalf(%). This enables evaluation of sums which would be difficult to form. 2200962205 658409472 Now turn this result into a decimal number.k=l.i=2.
3.1 Prove.2. 3 The required sum is now easily created > sum(product (a[chs [i] [ j ]1 . [I. 2.. Exercises @ Exercise 5. 4). 3 .3).144 Introduction to Mathematics with Maple Muple can produce sums and products which would not be so easy to produce otherwise.q E R n i=l n i= 1 (3) (q .2 notation.1 . . n. 3. Subscripts are created in Muple by brackets: ai is entered into Muple as a [ i l . by induction. chs := [[1.. i=l n 5 ai 5 bi for i = 1. . i=l .31. 3. 3 ) . . . 4 1 1 The second element in the third list [1. i=l n i=l Exercise 5. > chs :=combinat[choose] (4. 3. . The first step is to produce a list of all possible combinations of three elements from { 1. Prove (1) 0 5 ai (2) 0 < bi f o r i = 1. bi be real numbers for i = 1.3 Rewrite the formulae in Exercise 5. . for n E N.1. 2. 2. Assume we want to sum terms of the form aiajak where i.. n n .41.j=1. n implies n n ai 5 i=l i=l n bi .1) C qi = qn+l . 2. 2. Exercise 5.3.4]can be recalled by > chs[31 [ 2 ] .. j . We use the combinatorics package called combinat and the command is choose.41.3. .4) . .3. n implies n a i < r l [ b i . [ 2 . [I.3 using the Let ai. . k are all distinct and run from 1 to 4.i=l .
.Mathematical Induction n n 145 (3) k=l n 2 \ x a i ) (4) ‘ = i=l i. Derive the inequality alb2 albl I a2bl albl from which the lefthand inequality follows.25). The argument for the other inequality is similar.] 5. 2. y.5 Let a l l 132. Exercise 5.3. u are often used to denote sequences. A function whose domain is N is called a . .+an b. ..4 Prove that among n real numbers a l l a2. and bl.3.6 n n Show that [Hint: (1) Use (5. . @ Exercise 5. Then use induction to prove the lefthand inequality. a2. an there is a largest and a smallest. .3. be real numbers. . b. .19) and (5. b2.. The largest number is denoted by Max(a1. . .] + + 0 Exercise 5.3. j = l c n aiaj. .4 Sequences sequence. . . . . Let the minimum on the lefthand side be a l / b l . Letters a. an) and the smallest by Min(a1. bn be positive real numbers. The function value of a sequence u at n is called the nth t e r n of the sequence and is denoted by un. Displaying the first few terms often gives a good idea of the behaviour of the sequence although one must realize that the first few terms alone do not determine the sequence. .+b2+. a2. . an). . . However. z . Prove that Min (EL a2 b1’&”’’ ”> bn n ai+aa+*.+bn [Hint: Start with n = 2. (2) Simplify the double sum in (1) by using Exercise 5. . we may occasionally use the notation u(n) also. a. .1 (3).
. ( + 1) = f(2) 4. .1 For a given k define two different sequences which have the same first k terms. 5. we obtain + a2=a1+d=c+dl a3 = a2 a4 +d = c+ 2d. Example 5. in this case. Consider the following. for every n E N. Using the last relation successively for n = 1. we obtain Let Q = 99 = f(2)=f([g] f(3) = f +1) = f ( l ) + l = 2 ..4. = a3 + d = C + 3 d . . For example. Is it really certain that a.1 and we saw there that such a process needed to be made quite specific in mathematics.and define f(1) = 1. 7 f(4)=f([3 . the arithmetic sequence with first term c and difference d can be so defined. Setting n successively t o 1. f ( [ ( n+ l ) a ]+ 1) 100 f ( n )+ 1 .. There seems to be no difficulty in finding a. 2.146 Introduction to Mathematics with Maple Exercises Exercise 5. . 3. 2. by setting a1 = c and a. 3.5 Inductive definitions Sequences can often be defined inductively. .1 = 3 +1) = f ( 3 ) + 1 = 4 . d for n E N.6 . However this process of defining a sequence has some similarity with inductive reasoning discussed in Section 5. is defined for every n? The answer is ‘yes’. .+l = a.There seems t o be no difficulty in finding f ( n ) successively for every n E However this time there is no sequence f such that N. .
Then there exists a unique sequence f such that f(1) = a . Theorems concerned with sequences defined inductively (recursively) are usually proved by mathematical induction. f ( n ) = F(n) for every n E N and therefore f = F . if a E R.Mathematical Induction 147 and for every n E N. a E S and g : S 4S.9 at the end of this chapter. It is also possible to start with earlier or later terms than with n = 1. Then F(n 1) = g(F(n)) = g(f(n)) = f ( n 1). + + 1)aJ+ 1= 100. Theorem 5. an. n E N and F(n) = f ( n ) } . + . + + + We now use the Recursion Theorem to define the arithmetic sequence. For instance. and f(n + 1) = g(f(nNProof. Consequently A 4 = N. that n E M . for n = 100 we have L(n to f(lO0) = f(lO0) 1. Here we prove only the uniqueness and postpone the proof of existence to Section 5. leading Something more is needed to define a sequence than just checking that there is no difficulty in calculating the first few (or the first few billion) terms of the sequence. n E N o then we define an recursively as follows: a0 = 1. an+l = a . The solution is found in the Recursion Theorem. This means that n 1 E M .2 (Recursion Theorem) Let S be a set. Let F satisfy clearly 1 E M . Assume now Define M = {n. This is clearly impossible. Indeed. It is sufficient to take g : x H x d and f(1) = a1 = c.
Consequently an+' . 5. As a guide. b # 0 . such as the arrangements of branches on a tree. a > 1 am < an. n E No. . It occurs in a wide range of seemingly unrelated areas.1. Exercise 5. The statement is true for n = 0 and arbitrary m.am = a .5. b E R. this will also need to be justified.m E N o then anbn = (ab)".148 Introduction to Mathematics with Maple Theorem 5.4 can serve as a guide to more general recursive definitions.5.5. a E R and a # 0 . * This theorem is elementary and we leave the proof. .2 is not directly applicable to this definition. an . 2. anam = an+m. 5. If a similar recursive definition is needed relating more than two previous terms in the sequence. Exercises @ Exercise 5. (3) (ab)n = anbn.5.13. Proof. n E Z as Exercises 5. .4. the size of breeding populations of animals and in efficient procedures for sorting large amounts of data. 8.a n and by the induction hypothesis anam = an+m.5.5 and 5. By the recursive definition an+' = a . 0 5 a < b + an < b". am = a . and the required modification is stated in Exercise 5. 1. an+m = an+m+l. Assume (ii) and consider an+' am. 0 1. 13. n E N define a+ = (u')~. m E Z and n E Zprove: (I) an+m = anam. m < n.5. For . (am)n = amn. 21. A very general recursion theorem is proved later as Theorem 5.1 For a # 0 . generalised to the case where m E Z. here we prove item (ii) by induction.7. The sequence f is well defined and is called the Fibonacci sequence after the mathematician who first investigated it. 3. (2) (am)n= anm.3 (i) (ii) (iii) (iv) (v) If a E R. However Theorem 5.
3! = 1 2 3 = 6 . Z. n E Z. nE nE mE n N and 0 5 a < b imply an < b".5. b elements of a field.9 Prove this. @ Exercise 5. Prove this.5. Prove this. N and a < b imply a2n+1 < b2n+1. Prove the uniqueness o f f .7 this.5. xn+l + 5. defined for all n.5 @ Exercise 5.6 @ Exercise 5.5.5. @ Exercise 5.2! [Hint: Search for S.3 Prove that n (1) an .b) k=l an'bkl 2n k=O . Show that xn is not =3 1 x . The binomial coefficient ) (read ' n by k' or ' n choose k ' ) is defined by n! .] Let .10 2 1 .5. Prove E @ Exercise 5.5.5. Prove this.1 (Factorial and Binomial Coefficient) n factorial) is defined for n E NO by n! = nz. f ( 2 ) = b E S and f(n+2)= g(f(n)f .2 Define the geometric sequence (progression) inductively and prove the formula for the nth term and for the sum of n terms.5. l! = 1.4 Let S be a set and g : S x S + S . @ Exercise 5. n E Z. and explain why this does not contradict Theorem 5. ( n + 1) ) for n E N. Exercise 5.8 N and 0 5 a < b imply bn < an. n > m and a > 1 imply am < an.Mathematical Induction 149 Exercise 5. Then there exists exactlyone sequence f such that f ( 1 ) = a E S. x1 = Exercise 5. 2! = 2. m E Z. n > m and 0 < a < 1 imply am > an.6 The binomial theorem nt n! ( r e 2 finition 5.b " = ( a . n Consequently we have i=l O! = 1. for n E N and a . etc.
Assume (5.29) is called Pascal's Triangle.30) Remark 5. + + Proof. b are ele (a + b)" = 2 (L) k=O x a"'b'. The Pascal triangle shows that is an integer for n.150 Introduction to Mathematics with Maple It is a matter of simple calculation to check that successively as shown in the following table.a Y n summands for a in a ring and n E N. and the kth element in the nth row is (L I:) . In such a ring the product nu may not be defined for a in the ring and n E N .4 (Newton's Binomial Theorem) ments of a ring and n E N then If a . Usually multiplication by a natural number means repeated addition: by analogy we define nxa def = a+a+. For the proof of the Newton Theorem we also need the identity n x a m x a = (n m) x a. .. The last relation enables us to calculate 1 2 3 4 10 5 1 1 1 1 (5. The theorem is obviously true if n = 1.1 A ring need not contain all the natural numbers: for instance the ring of even numbers does not.30) and consider ( a + b)"+l. (5. 1 1 1 1 1 1 5 4 10 3 6 (L) + (Ic 1> (L 1:) = .. Theorem 5. The triangular table (5.29) Each number in this table is the sum of the two numbers which are situated to the left and the right in the row directly above it. k E N o with k 5 n.
1 cients: N Prove the following formulae for the binomial coeffi (1) k=O (&) = 2n1 where N = IT] n+l .) k=O 7 (5.31) and 0= g(l)k(.30) are 2n = 2 (.32) The binomial coefficient (3 is produced in Maple by binomial (n.6.). Exercises Exercise 5. k=O . k=O (5.k) .Mathematical Induction 151 (a + b)"" = (a + b)(a+ b)" Two easy consequences of (5.
7 Roots and powers with rational exponents A basic result of this section is the following: Theorem 5.] Exercise 5. (2) Expand ( 1 + ~ )and ~ " (l+x)"(l+x)" and compare the coefficient of x". k 5 n prove (1 + q < 1+ .6. n E positive number x such that xn = a. N then there exists exactly one . + h)" < 1+ Prove: ( a Exercise 5. a E R.5 [Hint:(l Prove: 0 k=O < h < 1. n. @ Exercise 5.4 Define the polynomial coefficients and prove the polynomial theorem r r for ai in some ring.6. n.6.] (1 + h)" < 1+ 2"h. n n2 n k k2 @ Exercise 5.6.6 + b)5 = a5 + b5 mod 5 5.k.6.31).32).3 For natural numbers k.] Exercise 5.152 Introduction to Mathematics with Maple [Hint: (1) Add (5.5 I f a > 0 .+ .2 For integers k .. n E N implies and use (5.31) and (5. 0 5 k 5 n prove n+l j=k [Hint: Use induction on m = n .
Since 0 < u < v implies un < vn there is at most one x satisfying zn = a.> 0.5 we have ~ ~ ( l + < h ) ~ xn 2nxnh. 1 (ii) We note that . . We define S = { u . a2ixy) (k) < .5.> 0.2 (nth root of a ) For n = 2 it is is called the nth root of a and is denoted by $. so that u $ S. since v 2 a implies wn 2 an > a (see Exercises 5.For a = 0 we set fi = 0. (ii) if a < xn there exists z E R such that 0 < z < x.5 5. For the existence proof we assume n > 1 since the case n = 1 is trivial. For the proof of Theorem 5. un < a}. 1. z < x with a < zn.7).5 we need the following Lemma. (A) a = 1 then x = 1. If a < xn then by (ii) of the same Lemma there exists x.5 and 5. X a 1 1 that yn < . We can now take z = . . We prove that xn = a by showing that each of xn < a and a < xn is impossible.1 there exists y E S . Lemma 5 . Proof. contradicting the definition of x as the supremum of S . y > x. Since (1+ h ) n < 1+2nh by Exercise 5. By the least upper bound axiom there exists x = s u p s . The set S is nonempty since 1 E S . and it is now sufficient to choose h such that 0 < h < 1 + and xn + 2"xnh 5 a. a Y We now take up the proof of Theorem 5. 1 Let a > 0. customary to write fi instead of fi. (i) We find h with 0 < h < 1such that ~ ~ ( l t <ha ) and ~ set y = x ( l + h).:  . We now distinguish three cases. Since x is the least upper bound of S there exists u E S. (i) if zn < a there exists y E R such that x < y.6.satisfies both of these By (i) there is y > . with z < u 5 2. If xn < a then by (i) of Lemma 5.Mathematical Induction 153 The number x from Theorem 5.5. This implies that a < zn < un.J: > 0. 2. y n < a. a < zn. u > 0. another contradiction. h = Min requirements.such X 1 0 Proof. (i.5. n E N. It is also bounded above by a . (B) a > 1.
1 ~ 1 . Every rational number r can be expressed in many different ways as P with p E Z and q E N. b > 0 and m E N then 6 $. 8 5. if a > 0. Hence for an odd n and a < 0 we have @ = . (iii) This follows from (ii) by setting a = 1. since it would follow that a = b in the first case and a > b in the second.33). = = 1 fis. b > 0.5) @a)n 0 a= Gvi.5 that = la1 (always). m. (i) is just a restatement of (5. since ( = ab we have (again by the uniqueness part of Theorem 5. (5.6 (i) (ii) (iii) = If a > 0 .fi. Consequently 0 need not be a (it is not if a < 0). We wish to define ar as i @ but before we 4 r =  do that we prove . Further. Indeed. (iv) @ = f i and fi > are both impossible.33) If n is odd then for every a E R there exists an x E R such that xn = a. Theorem 5. (iv) a < b =+ ~ < a.1 ~ 1 . however laI2 = a2 and it follows from the uniqueness part of Theorem 5. by its very definition. from which it follows that x = It is customary to denote this x also by fi.154 Introduction to Mathematics with Maple It is worth emphasizing that @ is. so that xn = a = . Proof. for a < 0 we have a = . positive or zero. (ii) ~ ffi i = by (i). Hence we must have * < ~.
m be integers. s E Q then (i) 1' = I. (ii) arbr = (ab)'. r E Q.8 If a . Proof.then q n fi= *.as ' (viii) (a < b. a ' > 0. s < T ) 3 as < a'. (5. (5. (iii> = aT br ' (iv) = b ' ' (v) aras = a r + s . For the rest of the proof we put r = . a ' = 1.5 we have x = y.= ar'.34) Let @ = x. b are positive. (ix) (1 < a.35) Remark 5. If . r E Q. .5. r > 0 ) 3 a ' < b'. @ = y.= .35) is the same as defined in Theorem 5. (vi) . as 1 (vii) ads = . q and n natural numbers and Theorem 5. O Definition 5. Then u p n = xQn and amq = ynQ. It follows that xnq = ynQ and by the uniqueness part of Theorem 5.r = E . a ' Proof. If r E Zthen ar defined by (5.3 (a' €or rational r ) we define ar = For a > 0.2 Section 5.Mathematical Induction 155 Let p . rn P with q E N and p E Z. s = Q rn E Z (it is a trivial exercise to show that r and s can be written with the same denominator). Q q EN fi. (i) is obvious.7 P m a > 0. (i)r 1.
6 (with a replaced by u p . r E Q and k E Z then then = urn. and therefore b ' (iv) This follows from (iii) by setting a = 1.3 .1 Prove that Theorem 5.156 Introduction t o Mathematics with Maple (ii) (by Theorem 5.) we have @ < @. b replaced by bP. Prove by induction: if a is positive.5.6 (i) ) = f/(ab)p (by Exercise 5.s > 0 and by using (viii) we obtain lr' < a'' = ar as the last equation following from (vi).6 remains valid for arbitrary real a .1) = (ab)'.7. (vii) Put r = 0 in (vi). (i)' (i)' = a'.7. etc. b provided n is odd and b # 0 in (ii) and (iii).3 (iv) ). (ix) We have r . q E N we have p E N. and by (iv) of Theorem 5. (vi) asars = ar by (v).2 @ Exercise 5. = ark. @ Exercise 5.7. Consequently a p < bP (by Theorem 5. ' 0 Exe rc ises @ Exercise 5. n E N o and r E Q Prove: i f a is positive. (iii) b ' = ( b i ) ' by (ii). (viii) Since r > 0.
Ll+(n+l)x. 3 6 ) .00. Exercise 5. a2.8 Let F : x +I x 2 for x < 1 and x H x for x 2 1. H v m + qr&. equality holds in ( 5 . If n = 1. . 5.36) holds for n and multiply it by 1 x 2 0. .7.8 Some important inequalities Mathematical induction is useful in proving general inequalities. By showing that the equation y = g ( x ) has a unique solution for y E [ f i 2 . W e have Hence 1 = 1. x 2 1. as the next three theorems show. Proof. If x E R. . 1[. ./m = (dn)' = = 1. Prove that F is onetoone.7 Let h : x onetoone and find h1.9 (Bernoulli's Inequality) then (1 x)" 2 1 n x . x 2 .36) is a trivial consequence of the Binomial 2 0.4 if x Inequality (5. an are nonnegative numbers then G= V a l a2 . + (1 x)"+' + 2 ( 1 + nz) (1 + 2 ) . domh =] . =1 + (n+ 1)z+ n x 2 . Assume (5.Mathematical Induction 157 @ Exercise 5. If a1 . / = fi = 1.7. Show that h is Exercise 5. Exercise 5.6 Let g : x H domg = [0.1].3 Theorem 5.5 d r m= . B u t we also have d r m = d m .7. . Theorem 5.36) Remark 5. Graph F . . Explain this paradox. an . nEN + + (5. and find rg F and F1. 1 prove the existence of 91 and find it.7.
Equality holds in (5. . .A is A. a n .40) i=l .a l ) ( a n + l A) >_ 0.A) or If this is combined with Inequality (5. . . n then (5. and this can be expanded to give A ( a l + a n + l . . Clearly. a1 an+l . A =  n+l is clear that a1 an+l.. an and an) A = . so we can arrange them in order of increasing size Let A be the arithmetic mean of these n + 1 numbers. . . . .A 2 0.158 Introduction to Mathematics with Maple is called the geometric mean of a l l u2.37) for n = 1.38) we get n1 (5. .37) i=l Proof. It + n1 i=l and hence From the obvious inequality a1 5 A 5 an+l it follows that ( A. .+ a 2 is called the arithmetic mean of a l . u3. . u2. The arithmetic mean of the n nonnegative numbers a2. . If Theorem 5. . .(a1 1 + +.. and applying the induction hypothesis to these numbers we have + + 1 i=l ai.10 (ArithmeticGeometric Mean Inequality) ui 2 0 for i = 1. an. 2. . rearranging the numbers doesn't affect the result.
A ) > 0. 2.43) Proof.44) .41).37) if and only if all ai are Proof. leading to strict inequality in (5. Assume now that the theorem is true for n E we have N. (2 k=l bi)1’2 = t n .42) This was established directly and easily in Exercise 4. The induction hypothesis (5. a = 6..37) becomes 2ab 5 a2 + b2. 0 For n = 2. We prove ‘the only’ part by showing that if not all ai are equal then nfl (5. ) 1 ’ 2 k=l k=l n = rn.38) gives (5. The ‘if’ part is obvious. (5. n then If ak and bk are in for (5.3 by considering the inequality ( a .41) a= 1 This is obvious if a1 = 0.2.43).Mathematical Induction 159 This is (5.39).b)2 2 0.Writing C akbk = sn. Multiplying this inequality with the positive number ai+l and taking into account Inequality (5. we now have a1 < A < an+l.1 Equality holds in (5. . If n = 1 equality holds in (5.10.37) with n replaced by n + 1.11 (Schwarz’s Inequality) k = 1. 0 Corollary 5. b= Jaz Inequality (5.al)(an+l . ( k a . Theorem 5. so let 0 < a1 < an+l With notation as before. . SO that ( A . .
Exercise 5.8. . substituting this in (5. .8.43) if and only if for a = 1. Ic E Show that S = N. N.5 For x 2 1 and n = r / s with r.1 There is equality in (5.8. = 1. . Exercise 5. xn be nonnegative. .s E N and r < s prove X that q G 5 1 + .47) 0 Corollary 5. (2) m E S implies Ic E S for all Ic < m. ar+l = 1. .. . n either bi = 0 or there exists c E R such that ai = cbi.10 on a1 = a2 = .10 to prove that among all rectangles of perimeter s the square has the largest area.7.4 Use Theorem 5. = n 1 + X .1 Let S c N with the followingproperties: (1) 2n E S whenever n E N. (5.8.1 to prove it for general n.45).160 Introduction to Mathematics with Maple implies we have.= a.3 n Let n X I ..8..8. @ Exercise 5. [Hint: Use Theorem 5. n n i= 1 i=l i=l i=l Prove these results. See Exercise 5. 0 Exercises Exercise 5.) Proof.2 Prove the arithmeticgeometric mean inequality for n = 2 m by induction on m. .11.22.8.1 . and then use Exercise 5. (ai = 0 is covered by the special case c = 0. . 2. Exercise 5.US . .
Proceed by induction. (c) ai = cbi for i = 1. By the well ordering . . that is. (b) t. Assume that S # N.7 Consequently equality holds throughout and Isn] = rntn.5 to obtain divide byI. bn+l = 0 . Distinguish three cases: (a) t . . Proof.1. N \ S # 8. n. The justification for such proofs is based on the following theorem. Theorem 5. Now use the induction hypothesis to complete the proof. = 0 .46 we must have Exercise 5. The ‘only if’ part is obvious if n = 1.Mathematical Induction 161 Exercise 5. If equality holds in (5.8. . bn+l # 0.12 Let S be a subset ofN with the followingproperties: (i) 1 E S .the foIIowing implication holds: ( k E S for all k < n) + n E S Then S = N. .8.6 r For 12: and n as in the previous exercise prove that 1 + [Hint: Use Exercise 5. [Hint: The ‘if’ part is easy.8.11. = 0 . 2. that the assertion is valid for all natural k less than n./? 2/72 +d p 5 1+ and then Prove Corollary 5. namely.9 Complete induction In some inductive proofs the assumption that the assertion is valid for n .45 and 5. A stronger induction hypothesis is employed.] 5.47) then from 5. (ii) for every n E N.1 is not strong enough to conclude that the assertion is valid for n.
8 Let x E R. . k < n and by the induction hypothesis k = 2t(2m .a1a2a3. . * .ala2ag.10) there is a smallest n E N \ S. By Theorem 5. say m. 10  a1 and proceed indefinitely. x > 0. This notation should not be confused with any product: it simply expresses the fact that for every n the number x satisfies n i=O n1 i=O If z > 0 and is expressed as a decimal fraction 10 102 uo.1). and S = N. 10 102 lon This defines a sequence an and we say that x is expressed as a decimal fraction a0. Complete induction can also be used for recursive definitions.12 is often useful as the next example shows. The next example shows this.12 S = N. Assume now that n is even and k E S for all k E N.. indeed every odd n E 5'. 1. k E N we therefore have k E S. and we denote this' by ao. which is a contradiction..l). so our initial assumption is wrong. There exists the largest integer not exceeding z. +an +5 X. Example 5. Example 5. and by (ii) we have m E S. k < n.162 Introduction to Mathematics with Maple principle (Theorem 4.1) with s E No and m E N}. For every n E N there exists s E N o and m E N such that Let S = (n. Then find the largest integer a1 such that ao+<x. n E N.. Since n is even we have n = 2 k . At the nth step find the largest integer an such that a 0 a1 + a2 + . 0 Theorem 5. n = 2'(2m . then * a1 a2 u u = a0 + ++* * an +lon 'In other words a0 = 1.1) and n E S. Clearly m # 1 since 1 E S by (i).7 n = Y(2m . n = 2t+1(2m . . k E N. We have 1 E S. For all k < rn.
..000.. and for b = 3 ternary fractions.1)>7 Proof. . If y < 0 we write x = y.. . f (21.2. find the decimal expansion of x in the form ao. a E S and snl H S for every n E N.Mathematical Induction 163 This. more precisely. If J: is a positive rational and of the form bl x=bo+ 10 b2 bn ++*+ 102 lon then the decimal fraction for x is bo . . is associated with the number 0. with y.ala2ag. .ala2a3. . For b = 2 we obtain what are called binary fractions. N o and. The above construction of decimal fractions or. together with the proof that ai E postponed to Exercises 5. Any other number b E N. . . in particular.3.2 and 5.b > 1 can be used instead of 10 in the above construction. Then there exists a unique function f : N +S such that gn : (ii) (9 f ( l >= a .. f (n>= g n ( f ( l > .9. Theorem 5.13 (Recursion Theorem) Let S be a set. and then associate the decimal fraction ao.9. The proof of existence is postponed to the last section of this chapter.. n > 1.f (n. 0 5 ai 5 9 is The decimal fraction 0. . of the sequence with nth term an can be made precise by using the following theorem. The proof of uniqueness is rather similar to the uniqueness proof of Theorem 5. bnOO0 .bl b2 . 0 The definitions of ai above can now be formally made: and .
E (0. [Hint: The proof is immediate from Exercise 5. otherwise it is divisible by a number smaller than 71..a1a2u3.4 Let x and y be positive with the same decimal fraction ao.is 0. We used this theorem as selfevident without proof in Chapter 2.13.. @ Exercise 5.. Prove that with our definition 0.a1a2a3. Prove it using complete induction! [Hint: If n is a prime there is nothing to prove. . . 3 Prove that the numbers s and rn from Example 5. 1.1 @ Exercise 5.3 A. is not a decimal fraction for any x E R. Prove that x = y.9. Prove that two distinct real numbers have distinct dec 5.7 are uniquely determined by n. respectively. Let a i .2 and 5..8 imal fractions. = C i=O n Let x > 0 and ao. @ Exercise 5. Also 1 show that the decimal fraction corresponding to .Prove that x = SUP(Y..9.2 Let a real number x > 0 be expressed as a decimal fraction ao.9.9.9. We need the following Lemma.9.6 Exercise 5.164 Introduction to Mathematics with Maple Exercises @ Exercise 5.3. We shall rectify this in Chapter 11.1 Every natural number is divisible by a prime.9. 1oi * @ Exercise 5..9. be its decimal fraction. 1) and a E (0. .9.. .10 Proof of the recursion theorem To complete our treatment of the Recursion Theorems 5. . y = An}. Exercise 5. Also prove that for binary and ternary fractions one has a.3333. 2}. . .7 Prove that if two positive real numbers x and y have distinct decimal fractions then they are distinct.999. Prove that 0 5 an 5 9 for every n E N.5 Our definition of decimal fractions is slightly different from the conventional one.13 we are going to prove the existence part of Theorem 5.3 Exercise 5.a1a2a3.
then for n 5 K .1 by the induction hypothesis and f d K ) = 9K(fK(1). Assume now f ~ .13.48) and (5. f K ( n . .52) and by (5. f1 : 1 H a.51) which means that fK(n) is uniquely determined for n = 1.50) and by (5. 2. fK(K.49).1)) (5. The first equation is easy. it exists. If f~ exists it must satisfy (5. Then we proceed by induction again. . In view of (5. . . .2. is uniquely determined. .1)) by (5.l ( l )= a. K ) H S such that fK(1) = a there exists a unique function f~ : (5. The function f l is uniquely determined. .48) * * * f K W = gn(fK(I). .2 For every K E (1. Assume f ~ . We use this piece of knowledge to define f ~ More . The proof is by induction. . .49). Then we have for n 5 K  1 and for n = K by (5.Mathematical Induction 165 Lemma 5.52). . First we prove that f ~ if .1exists and define f ~ ( nby ) (5. .is l unique.50). .49) Proof. n E N. f ~ ( 1= ) f ~ . We now take up the proof of Theorem 5. . fK(% for all n 5 K . K . It remains to show that f~ so defined satisfies (5.50) we have (5.52). precisely we define first fl(1) = a.
.. We define = f u 00 for every n E N. mathematical induction was one of the principal axioms. the recursion theorem or some modification of it is needed for the definition of addition and multiplication of natural numbers. Obviously f is a relation. The foundation for this was laid in the previous chapter in Theorems 4. 0 5. and it is sufficient to prove that it is a function. Consequently Proof.(s)= fs(s). which we sketched in Section 4.5 and 4. Q.166 Introduction to Mathematics with Maple f~ and emphasize that each fK is a set of K=l pairs (n. Similarly f(k) = f n ( k ) for all k 5 n. In Peano’s approach.11 Comments We devoted this chapter to mathematical induction and its applications and to recursive definitions. However the recursion theorem still must be proved and it plays an even more fundamental and central role than in our approach. . Then u = fp(s) and w = fq(s) for some natural numbers p . f K ( n ) ) . u ) E f and (s. k E N. by uniqueness. axioms of a complete ordered field. we have fp(s) = f. Since both the restriction of fp and the restriction of fq to (1. . Let (s. u = w. These theorems were established as consequences of the axioms for the reals. w) E f. that is. . . 2. that is. be equal to fs. In Peano’s axiomatic approach. s} must.10. Then f(1) = a since f(1) = fl(1).8.
6 . If A and B are two polynomials then the polynomials A B . 167 . 1 Polynomial functions If M is a ring and ao. a l . . This natural extension of the domain of definition is often understood without explicitly saying so. . A and AB are defined in the obvious way as + A + B : x H A(x)+ B ( x ) A : x A(x) AB : z H A(x)B(x) The coefficients of A+B are obvious. we also look briefly at zeros of polynomials and prove the Taylor Theorem for polynomials in a generality which cannot be obtained by using methods of calculus. . in the past. but the definition of A(x) makes sense for any x in a ring which contains M . establish the long division algorithm in an abstract setting. a polynomial with coefficients in M . The zero polynomial function is the zero function. if for nothing else than because. or rewrite Equation (6. In this chapter we define polynomials as algebraic entities rather than functions. an E M then a function of the form is called a polynomial. they were the only functions which could be readily evaluated. a2. one can add any number of zero coefficients. Obviously. they are the sums of the corresponding coefficients of A and B. The domain of definition of the polynomial is naturally M . .1) in ascending order of powers of z without changing the polynomial.Chapter 6 Polynomials Polynomial functions have always been important. or sometimes more explicitly.
The coefficients of AB are obtained by multiplying through. are not uniquely determined by the polynomial function. . 2 . In modern mathematics there is a need to regard polynomials as algebraic expressions. generally speaking.168 Introduction to Mathematics with Maple that is 10 : x H 0. + anibm There is a clear pattern to the formulae (6. Treating polynomials as algebraic expressions is needed because the coefficients of a polynomial function. all pl. the coefficients of . Then we can rewrite Equations (6. .2) as k pk = j=O akjbj (6. collecting terms with the same power of x and sorting them in descending (or ascending) powers of x. Similarly.A have opposite signs to the coefficients of A. . . . The precise meaning of this will be made clear in the next section. With these definitions of addition and multiplication polynomials themselves form a ring. In order to subsume them in a compact formula we set a k = 0 for k > n and bk = 0 for k > r n . For instance.3) for' k = 1 . the polynomial x I+ x2 x over GF(2) and the zero polynomial x I+ 0 + lFor Ic > n + rn.2). = 0. If and P then = AB.
An advantage of this notation is that it facilitates multiplication. and a different notation is used too.) are equal. replace X k X 1by Xk+' and collect terms with the same 'power' of X . Instead of (a. more precisely an unending polynomial over M .2. Later in this chapter we prove that if the ring M is a field which contains the field of rationals. it is a purely formal symbol indicating term number n 1 of the sequence.. by definition.or more explicitly Two unending polynomials (a. if and only if ak = bk for all k E No. Unending polynomials are also (more commonly) called formal power series.2. For instance + (1 2~ + + 3 x 2 + + (n + l)xn + . Addition of unending polynomials satisfies axioms .Polynomials 169 are equal as functions but they do not have the same coefficients. To indicate that we regard a sequence n I+ a. then indeed. equal polynomial functions have the same coefficients...2 Algebraic viewpoint A sequence of elements from M constitutes an unending polynomial. To multiply two unending polynomials.1 and 6. we multiply term by term.3).. The sum and product of two unending polynomials are defined as where pk are defined as in (6. 6.2 x + 3 x 2 + + (l)"(n + l>xn + ) = 1 + 2 x 2 + 3 x 4 + . For an unending polynomial (an) the elements ak are referred to as coefficients.) and (b. n E N o as an unending polynomial rather than as a mere sequence we write (an)." + Some examples on multiplication of unending polynomials are given in Exercises 6..) one writes One problem with this notation is that X " is not a power of an element from M .2. ) * (1 . The reason for this name becomes apparent later in this section.+ (n+ 1 * * * ) ~ ~ .
It follows that unending polynomials form a ring. It is easy to see that Axiom D from Table 4. In other words ( a n ) is a polynomial if there is a natural number N such that a k = 0 for k > N . . Polynomials are a special class of unending polynomials: therefore to show that they form a ring it suffices to show that the sum and product of two polynomials have only a finite number of coefficients distinct from zero. and each such term appears in the sum exactly once.. This is obvious for the sum.. Axiom M 1 is also obviously satisfied. Similarly to the notation (6. (a.1 is also satisfied.) with a k = 0 for k > n and ( b n ) with bk = 0 for k > m. The polynomials over a ring M themselves form a ring which is denoted by M [ X ] .7) and (6. The coefficient a0 is called the absolute term..8) consist of terms of the form a&c. This completes the proof that polynomials form a ring.B y = n. p and y a. since + + + k j=O + k j=O Now we prove that multiplication of unending polynomials is associative.9) briefly as A ( X ) . For the product of the polynomials.e nonnegative integers with a! . Using formula (6. If N > n + rn then (at least one of) aNi or bi is zero.170 Introduction to Mathematics with Maple A1 to A 4 from Table 4. The role of the zero element is played by the zero unending polynomial 0 OX .. the coefficient a N is said to be the leading coefficient. which we denote by 0 ..1. the general N term is i=O aNibi. Consequently the sums are equal and the multiplication is associative.OXn .. A polynomial is a formal power series which has only a finite number of coefficients distinct from zero.3) for multiplication of unending polynomials we have n ni Both sums in (6. where a .6) we write + + We shall denote the polynomial in (6.
0. 0.degB(X)) deg (A(X) .) + (b. . 1.) .) .11). The ring M need not have an element 1 (the unit element for multiplication) of Axiom M3 from Table 4. 0. . 0. ( a .10). More precisely.6.2. polynomials of degree zero behave exactly like elements of M . (6.46. .o. the zero polynomial which has all coefficients zero and which is denoted by 0 has no degree.. 0.). 0.0. . 0. (a. . In this spirit we say that the set of polynomials of degree zero together with the zero element is M . .. . the ring of even integers has no unit element. In the situation above. . If M does have a unit element then we can denote the polynomial (0. . . However.(bo. . .). Hence we have deg(A(X) . Exercise 6. b l . and if the elements of these structures behave. the coefficient of Xn+min this case must be different from zero. *.) = ( a . In abstract algebra and in this book we shall consider two isomorphic structures as identical. . 0. . If.. for example. (a. . The nonzero elements of M can be regarded as polynomials of degree zero. .. such as two fields or two rings. . as far as addition or multiplication is concerned.b . 0.). The case B ( X ) = 1 .. as far as addition and multiplication are concerned.12) For more details see Exercises 6.) = (a bo. we say that these structures are isomorphic. * If we have two algebraic structures. Indeed. .2. for a E M we have I ( a ) = ( a . It is easy to see that + + deg (A(X) + B(X)) 5 Max( degA(X). however. .Polynomials 171 If a.). . if M and M' are two rings (fields) we say that they are isomorphic if there is a bijection I : M I+ M' such that I(" b) = I ( a ) I ( b ) and I ( a b) = I ( a ) .. deg A(X) and deg B(X) are defined.I@). 0. + (6. a b l . M is a field then the equality holds in (6. .10) (6. The bijection I is called an isomorphism. 0. 0.B(X)) = deg A(X) + deg B(X).11): indeed. . . o. . . in symbols degA(X) = n.1.2. 0. . . 0. in the same way.o.11) provided deg (A(X) B(X)).B(X)) L deg A(X) + deg B(X).5 gives an example showing that strict inequality can occur in (6. # 0 and ak = 0 for k > n then the degree of A(X) is defined to be n.A(X) shows that strict inequality can occur in (6.) (b.) = (a + b. .
+ ao. The notation now becomes less formal and more meaningful. This notation is convenient because we have the following rule: 2This rule is false if F [ X ] is replaced by M [ X ] .13) If B ( X ) . The following rule looks obvious but still requires proof2. * .C ( X ) were not the zero polynomial then the leading coefficient on the right hand side of (6. . 1. 0.B ( X ) .. . (6. Cancellation rule: If F is a field and A ( X ) . 0.. Generally X” = (0. 0 . With the polynomial A ( X ) associate the polynomial function A defined on M A :x I+ anxn + anal xnl + . 1. Since F [ X ] is a ring we have A(X)(B(X) . * * (6. X is called the indeterminate and it behaves.13) 0 could not hold.13) would not be zero and Equation (6. ao.14). Each term akXk is a product of a polynomial ak of degree zero with the polynomial X k .). . .C ( X ) )= 0. . . with respect to addition and multiplication. and the original polynomial itself is a sum of polynomials a n X n .. 0 . A ( x ) is the value of this function at x and A ( X ) is the polynomial (6. a l X . C ( X ) E F [ X ] with A ( X ) not a zero polynomial and if A(X)B(X)=A(X)C(X) then B ( X ) = C ( X ) ..).9). . .14) Let us make the notation very clear: A is the function defined in (6. an1 xn1. Proof.. n zeros  . 0. like an element of M or an element of some ring containing M .172 Introduction to Mathematics with Maple by X and then X . Consequently B ( X ) = C ( X ) .. . X = X 2 is the polynomial (0.
.3 An easy way to multiply two polynomials A(x) and B(x) of degree m and n. + + + + 3The command convert is needed only for further computations: if you want only the result of multiplication you can omit it.1 and B(x) = xlo0 xg9 .ij = xyy.) (2) (1+2X+X2+0X3+~~ .4 The relation x = y mod m was introduced in Example 3. Show that these definitions are correct. It is however common to omit the tilde sign.. Show also that if M is a field then the polynomial B(X) exists if merely a0 # 0. respectively. Show also that this ring has a unit element satisfying axiom M3 from Table 4.m+n) . Proof. a n X n with a0 = 1 then there exists an unending polynomial B(X) such that A ( X ) . Prove this. For greater clarity we denoted the classes with a tilde sign. If A ( X ) = a0 a l X .Polynomials 173 Subst it ut ion rule: If A ( X ) and B ( X ) are two equal polynomials ouer M and z is an element of M or an element of a ring containing M then A(z) = B ( z ) . and denote the class by the number which it contains. if x = x' mod m and y = y' mod m then x y = x ' y' mod m.. Consequently the function values of A and B are evaluated in exactly the same 0 way and must be equal..2 If A 4 has a multiplicative unit element 1 then 1 O X . .1.2.xn . that is. + + + + + + @ Exercise 6.2. and each term of the form C k X k appears exactly once.1. +(l)"X"+.. Show that the equivalence classes modulo m with addition and multiplication j u s t defined form a ring. Exercises Exercise 6.1 * Multiply the following unending polynomials (1) (1 x . they have the same coefficients. + + + Exercise 6.x + x2 .1.....) * * * * * * * + + + + + + @ Exercise 6. is to use the Maple command3 taylor(A(x)B(x) ..+OX"+.+OX"+. ) (1 .2. with the corresponding relationship for the product.B ( X ) = 1.(lX+X2+. Denote the equivalence class containing an element x as 5.convert(%. Define Z ij = x y and 5 .2.polporn) : Veri& this for A(x) = x2 . A(X) and B ( X ) are equal polynomials which means that when both are written in order of ascending powers of X.is the multiplicative unit in the ring of unendingpolynomials: this unit is also denoted by 1.).(1)"Xn + . .
zy = 0 + y = 0. (b) if z. + @ Exercise 6. and (b) is the zero polynomial. B ( X ) )ef( k + I . We define addition and multiplication as follows: + + + + ( k . @ Exercise 6.2.2.aoXk . [Hint: The unit element of multiplication is (m.(1.2. Give your own example of M and polynomials of second and third degree for which the product (a) is of third degree.is isomorphic to R. Show that (1) a field is an integral domain. .22) holds.. Prove that the equivalence classes mod p with addition and multiplication defined in Exercise 6.X m ) .8 Let p be a prime. y E M then z # 0.X . (3) if M is an integral domain then so is M [XI.2.] 4Some authors do not include this axiom in the definition of an integral domain. B ( X ) ) are equal if and only if X z A ( X ) = X k B ( X ) as unending polynomials. A ( X ) .2.6 A ring M is called an integral domain if (a) it has a multiplicative unit element. A ( X ) ) = ( l . The set of elements of the form (1.4 as in axiom M3.174 Introduction t o Mathematics with Maple @ Exercise 6.5 If M is the ring of integers modulo 4 then the product of two polynomials 1 SX and SX of degree one is a polynomial of degree one. Exercise 6.15) can be introduced as pairs ( k . To find a multiplicative inverse use Exercise 6...B ( X ) ) .X . If the coefficients ai E R.7 Prove that an integral domain having only a finite number of elements is a field. Verify this.A ( X ) ) + ( I . A ( X ) ) where k E N and A ( X ) = ak ak+lX .A ( X ) ) with A ( X ) = a 0 . ( k .4 form a field.is an unending polynomial.9 Algebraic expressions of the form (6. + + Show that with these definitions of addition and multiplication the formal Lseries constitute a field.2. A ( X ) ).2.2. (2) if M is an integral domain then Equation (6. B ( X ) ) def ( k + I . we call such pairs formal Lseries. @ Exercise 6.. We say that two Lseries ( k . X z A ( X )+ X k 1 3 ( X ) ) .
10) and (6.17) If Q ( X ). If besides Q ( X ) and R ( X ) there are Q 1 ( X ) and R 1 ( X ) with the same properties then H ( X ) [ Q ( X) Q l ( X ) ]= R l ( X ) . + + 6.and n E N prove (1.R ( X ) are uniquely determined. Define (f g)(z) = f(z) g(z) and (f .12) .10 Let a be the first nonzero coefficient of an unending polynomial A ( X ) .A ( X ) ) to be positive if and only if a > 0.1 If P ( X ) .5. ~ Theorem 6. Show that the positive formal Lseries have the properties of positive elements mentioned on page 103 and conclude that formal Lseries constitute an ordered field.g)(z) = f(z). Show that if S contains at least two distinct elements then 5 is not an integral domain.. We have already used it in Section 1.. Proof.2 and readers have probably encountered it many times before.2.g(z).11 Let S be a nonempty set and 5 the set of functions f : S ++ R.3 Long division algorithm The next theorem describes socalled long division. The polynomials Q ( X ) .Polynomials 175 Exercise 6. that is A ( X ) = a x " am+1Xm+' .Q1 ( X )were not the zero polynomial then neither would R1 ( X )R ( X ) be and by (6. most likely only for F = R and without a proper proof. . [Hint: For A ( X ) = 1+ O X . First we prove uniqueness.16) and either deg R ( X ) < deg H ( X ) or R ( X ) is the zero polynomial.R ( X ) .] + + + @ Exercise 6. however. Show that with these definitions 5 becomes a ring which has a multiplicative unit.A ( X ) ) > n. +OX" .. Show that this field lacks the Archimedean property. In the field of formal Lseries (see the previous exercise) define (k.H ( X ) are polynomials over a field F and H ( X ) is not the zero polynomial then there exist polynomials Q ( X ) E F [ X ]and R ( X ) E F [ X ]such that P ( X >= H ( X ) Q ( X ) +R(X) (6.2. (6. .
' = P(X)/H(X). The commands are quo(P(X) . In the former case either by induction hypothesis or because degp(X) > degH(X) there exist polynomials G(X) and R(X) such that p(X) = H(X)G(X) R(X) with degR(X) < degH(X) or R(X) = 0. For instance . b E Q.X). If rn > n or P ( X ) = 0 then the theorem is true. as required. In the latter case R(X) = 0 and Q(X) = anhmlXnm. Maple works in the field which is implied by the coefficients. Let with pn # 0 and hm # 0.X) for the quotient and rem(P(X1 . In the remaining case we proceed by complete induction. The polynomial p(X) has been chosen so that degp(X) < degP(X) or possibly p(X) = 0. In the following example the field is formed by numbers of the form a b& with a.H(X) . in the smallest field which contains the coefficients of P(X) and H(X). It follows that P(X) = 0 H(X)[pnh. Let us now consider the polynomial p(X) = P(X) . Now we prove existence.X) for the remainder. P ( X ) and H(X) have integer coefficients then Q(X) and R(X) are found in Q[X]. However this argument must be enclosed in ' '. this means. either q or r.176 Introduction to Mathematics with Maple This contradicts (6. + 1 1 1 Jz+x 3 9 3 > rem(Xn2+sqrt(2)*X+sqrt (2) .3*X+l.pnhklXnmH(X). 3h + g 2 1 One can obtain both the remainder and the quotient by one command by adding an additional fourth argument. with Q(X) = 0 and R(X) = P(X). + + + Maple can find the polynomials Q(X) and R(X). The assertion is true if degP(X) = 0 with R(X) = 0 and Q(X) = poh. If.H(X) .lXnm G(X)] R(X). for example.17) and consequently Q(X) = Q1 (X) and then R(X) = R1 (X).
5 irem is then used to check the result.9l. 62’ + x8 + 6z7+ x6 + 6 x 5 + z4+ 6z3+ z2 + 6 z + 1 1 For polynomials of zero degree the Quo command can be used to find the multiplicative inverse of an element in the field of integers mod p . The next example shows how to find the multiplicative inverse of 91 mod 97. .x. there is no remainder.6*~1. where p is a prime. 5 0 f course.r. > Quo(l. but the coefficients can belong to the field of equivalence classes mod p .X) mod 97. > Quo(x~10. This is illustrated in the next example.’r’)mod 7.Polynomials 177 4 x+1 or The capitalized commands Quo and Rem work the same way.
16) is satisfied with two distinct polynomials Q ( X ).a ) Q ( X ) P(u).4 I f H ( X ) = X" h. 1 Exercises @ Exercise 6.1 If F is replaced by Z in Theorem 6.4 + + + Find the multiplicative inverse of 2 mod 101 using the Roots of polynomials If F is a field.Obviously.19) . Q l ( X ) = X and Q z ( X ) = 3X show that Equation (6.1 then polynomials Q ( X ) and R ( X ) need not exist. (See Exercise 6. Consider the following example: P ( X ) = X 2 + 1 and H ( X ) = 2 X .5 Quo command.2.ho and F is merely an integral domain then the assertion of Theorem 6. Theorem 6. in which case Equation (6.3.178 Introduction to Mathematics with Maple 16 > irem(l6*91.] @ Exercise 6. Prove it! Exercise 6. the required polynomials do not have integer coefficients.4. polynomials in Z[X] lie also in Q [ X ]and the theorem can be applied to Q [ X ] .1 remains valid.lX"' .) Considering P ( X ) = 2 X 2 .3.3 Let M be the ring of equivalence classes modulo 4. respectively. + (6. Prove all these assertions.97).3. A root of P ( X ) or of P is also called a zero of P ( X ) or of P . [Hint: Only Axiom M4 is not obvious. Prove that numbers of the form a and b E Q form a field. @ Exercise 6. P ( X ) E F [ X ]then a belonging to some field containing F is a root of the polynomial P ( X ) or of its associated function P if P ( a ) = 0.1 is often used for H ( X ) = X a. 6. However.2 +b d with a E Q Exercise 6.16) reads P ( X ) = ( X ..3.3.
the polynomial of the second degree X22X+1 has only one root. 0 If M is merely a ring then P ( X ) E M [ X ]of degree n can have more roots than n: one such example is given in Exercise 6.. The theorem is obviously correct for polynomials of degree zero.3 If F is a field and P ( X ) E F [ X ]is of the form P ( X )= p .1 roots and consequently P ( X ) has at most n roots. for example.4. By the induction hypothesis Q ( X ) has at most n .1.n. I f F has infinitely many elements and P and Q are two polynomial functions with P ( z )= Q(z) for everyz E F then they have the same coefficients. The proof cannot be given here but we shall make some remarks concerning this result about field extensions in Chapter 7. If it has a root a then by Equation (6.Q ( X ) must be the zero polynomial. 0 Exercises @ Exercise 6. A # S and assume that A has infinitely many elements. + p o and has n 1 distinct roots then p i = 0 for i = 0 .1. ~~ ~ Theorem 6. 1 .Polynomials 179 It follows that if a is a zero of P ( X ) then X . Show that the polynomial with B c A. Proof.a ) " Q ( X ) with Q ( a ) # 0 we say that a is a root of P ( X ) of multiplicity k . Let A c S .] ls\AX E g [ X ]has infinitely many roots. . It can be shown that for any polynomial in F [ X ]of degree at least one there exists a field F 3 F in which the polynomial has a root. Obviously a polynomial of degree n can have fewer than n roots.1 This exercise uses the notation of Exercise 6.a ) Q ( X ) . We proceed by induction.2.where Q ( X ) is of degree n .11. For k = 2 the root is called a double root.a divides P ( X ) . for k = 1 the root is simple.19) we have P ( X ) = ( X .4.2 has the following very important consequence. Theorem 6. If P ( X ) of degree n has no roots then there is nothing more to prove.2 If F is a field then the polynomial P ( X ) E F [ X ]of degree n has at most n distinct roots in any field which contains F . .. [Hint: X = 1~ . . . P ( X ) . If P ( X ) = ( X . From this we have Theorem 6. x n + . + Proof.
The whole computation can be arranged in the following array. qn2 can be found from the second equation. .. add up the elements in that column (by using equations (6. The coefficients of Q ( X ) = qnlXn' +..19) further.4.2 .21) aqn1 aq.20) in the form qn2 = pn1 aqnl. (6. etc..19) can be easily obtained by comparing the coefficients. It is important to write all coefficients. With this done we multiply each element of the third row by a .20) P i = aql Po = aqo + qo + P(a). place it in the next column in the row above. Pl Po (6.2 Show that if F is finite then there exists a polynomial which is not the zero polynomial and for which every element in F is a root. the quotient again by X ... One finds qn1 from the first equation. we write all coefficients of P ( X ) in the first row. Then we leave one row empty and copy p n in the first column of the third row.a and by continuing this process we form the following scheme.+qo and the value P ( a ) in Equation (6..5 The Taylor polynomial In this section we explore Equation (6. Firstly.) and successively fill in the whole array. 6. aq1 aqo We can divide Q ( X )by X . known as Horner's scheme.a .180 Introduction to Mathematics with Maple Exercise 6. + Pn Pn1 Pn2 . even the zero ones. and by continuing with this process we can find all qk and also P(a).
The right hand side of this equation is called the Taylor polynomial or the Taylor expansion of P ( X ) around the point X = a. by working backwards through the schemes (6.l = 92. which we denote by qk. To illustrate the procedure we expand X4 4 X 2 3 X 2 in powers of X .a.2 with the coefficients lying in the field of integers mod 5.23) we have g l ..23).Polynomials 181 (6.. .* .9n1. + + + .. can be obtained by continuation of the Horner scheme. n .22) and (6.n2 = .23) Finally.a. (6.o = 4n = P n .22) The coefficients of Q k ( x ) . we express P ( X ) as a sum of powers of X . In the scheme (6.1 = Qn.
a leads to We can take some inspiration from this formula and define a derivative of a polynomial (derivatives are treated from a calculus point of view in detail in Chapter 13) by The formula for q1 then becomes q1 = D P ( a ) . + The coefficients qk can be determined in other ways than by the scheme k (6. The following formulae are rather important. Collecting the absolute terms leads to the result qo = P ( u ) which we already know. Such a procedure might be advantageous if the degree of P(X) is very small. One can write Xk = ((X .2)4.4).2)3+ (X .23). collecting the coefficients of X .a ) a ) and use the binomial theorem (Theorem 5.2)2 + 3 ( X .2) + 3(X .182 Introduction to Mathematics with Maple 1 0 4 3 2 2 4 1 3 1 2 3 4 0 2 3 2 1 4 1 1 2 2 1 1 3 2 1 3 As a result we have x4+4x2+3x + 2 = (X . .
of course. If one has a computer and Maple available everything becomes easy.27) we write P l ( X ) = P I ( u ) ( D P l ( a ) ) ( X.X=a. Using induction it follows from (6. From these equations it is easy to find the coefficients q k .27) follows.25) and then substituting X = a we obtain D 2 P ( a )= 2 1.This can also be verified directly from (6.1 43.Polynomials 183 Equation (6. * (6.a ) + ( X . where we again denoted by W some polynomial. The command taylor(P(X) .a. the process is sometimes reversed: one finds D k P ( a )as q k by using the scheme (6. . happens in the most important cases when F = Q or R. however.c ) ~ .a)2W1(X. Defining D k P ( X )Ef D ( D k .23).29) The importance of this equation lies in the fact that it gives explicit formulae for the coefficients of powers of X .a ) .28) are distinct from zero6 only if F contains N. This.27) that D ( X c ) = ~ k ( X . where we abbreviated by W1 and W2 some polynomials whose form is unimportant for the proof.A ) + + ( X .4 2 .25). It follows that P l ( X ) P 2 ( X )= P1(a)P2(a) + [ P 2 ( 4 D P 1 ( 4+ P1(a)DP2(a)] ( X . To prove (6. P2(X)= P ' ( a ) + ( D P z ( a ) ) ( X.A ) + ( X .2 . D 3 P ( a )= 3 .n) 6For example n! = 0 mod p if n 2 p . In practice.' .l P ( X ) )and applying D successively to Equation (6.u ) ~ W ~ (X a).a ) . all coefficients of q k in (6.28) D"P(a) = n! qn. From this equation the formula (6.a ) 2 W ( x.26) follows easily from the definition of derivative. The Taylor formula then takes the form (6. Some caution is needed: generally speaking.
Exercise 6.184 Introduction to Mathematics with Maple produces expansion (6. We recapitulate some of them.polynom) .5.3 2b/3. It does that even if the coefficients of P ( X ) contain some parameters. (2) P ( X ) = x 5 + 2x4 + 3 x 2 + 2x + 1.or that P ( X ) is divisible by H ( X ) . For ~ 7With N ( r ) equal to the degree of the polynomial.23) or (6. Use Maple to find the Taylor expansion of X 2 at a = Decide in each cme which formula is more convenient to use. Exercise 6.5.1 of x3for a = 1. . If the polynomial D ( X ) is a divisor of both PI ( X ) and P 2 ( X ) it is called a common divisor: if every other common divisor divides D ( X ) then D ( X ) is called the greatest common divisor of P l ( X ) and P 2 ( X ) .2 polynomials (1) P ( X ) = Use the binomial theorem to find the Taylor expansion Find the Taylor expansion at a = 2 for the following x5. If not stated otherwise every polynomial is automatically assumed to belong to F [ X ] .30) The word factor is often used instead of divisor. Hence the complete command is then taylor(P(X) .X=a.6 Factorization In this section we shall assume that F always denotes a field.24) with the coefficients q k evaluated. The polynomial P ( X ) is a multiple of H ( X ) and H ( X ) is called a divisor of P ( X ) if there exists a polynomial Q ( X ) such that (6. If H ( X ) is a divisor of P ( X ) then we also say that H ( X ) divides P ( X ) . Exercises Exercise 6. In Section 2.29). as we mentioned earlier.n):convert(%.1 we established some results on divisibility in a Euclidean ring and applied it to polynomials with rational coefficients. (6. 6. Since F [ X ]is also a Euclidean7 ring these results remain valid.. to add the command convert. It is advisable.5.5.
Polynomials 185 any two polynomials P l ( X ) # 0 and P 2 ( X ) in F [ X ]the greatest common divisor always exists.but it is left as an exercise to prove indirectly that it is irreducible in Q [ X ] . and P ( X ) is divisible by some irreducible polynomial D ( X ) . A polynomial P ( X ) of degree n 2 1 is called reducible over (in) F if there exists a polynomial H ( X ) E F [ X ]of degree greater than zero but less than n such that Equation (6. It is important to realize that the concept of being reducible or irreducible depends not only on the polynomial P ( X ) but also on the field in question. + (a) If P ( X ) is irreducible in F [ X ]and 0 # c E F then c P ( X ) is irreducible. Multiplying by Q ( X ) gives + . X 2 .31) we have 1= D ( X ) A ( X ) H ( X ) B ( X ) for some A ( X ) and B ( X ) . It is the polynomial of the smallest degree which is of the form with H l ( X ) and H 2 ( X ) in F [ X ] .&)(X &) is reducible in R [ X ] .30) holds. For example. It follows from the definition that every polynomial of degree one is irreducible over F [XI.then D ( X ) is a divisor of either H ( X ) or Q ( X ) . Our aim is to show that any polynomial can be decomposed into a product of irreducible polynomials and that such a decomposition is in a certain sense unique. To this purpose we first list some properties of irreducible polynomials.30) implies that d e g H ( X ) is either n or 0. (c) If (6.The greatest common divisor is uniquely determined in the following sense: if D l ( X ) and D 2 ( X ) are two greatest common divisors of the same two polynomials then there exists an element in c E F such that D1 ( X ) = c D 2 ( X ) . A polynomial P ( X ) of degree n 2 1 is irreducible over F if Equation (6.2 = ( X .30) holds. (b) If H ( X ) is a divisor of an irreducible P ( X ) then either N ( X ) is of zero degree or H ( X ) = c P ( X ) for some nonzero element of F. If D ( X ) does not divide H ( X ) then the greatest common divisor of D ( X ) and H ( X ) is 1 and by (6.
. such that (6. . If Q l ( X ) is not of zero degree. irreducible over &"XI. by continuing with this process we must finally arrive at a polynomial of zero degree and we have P ( X ) = P1 ( X ) P 2 ( X ) P l ( X ) . . If P ( X )= Ql(X)Q2(X) .32) where all the polynomials Pi ( X ) are irreducible. 2.33) is another representation of P ( X ) with all Q i ( X )irreducible over F [ X ] then m = 1 and. Proof. divide one of the Q i ( X ) . . The theorem is true if P ( X ) is of degree one. Theorem 6. Let us now assume that the theorem is true for all polynomials of degree less than n and let P ( X ) be of degree n. (e) Every polynomial of degree at least 1 is divisible by some irreducible polynomial.4 For every polynomial P ( X ) E F [ X ] of degree at least one there exist polynomials P i ( X ) . If d e g P ( X ) > 0 then by (e) we have P ( X ) = P I ( X ) Q l ( X )with an irreducible P l ( X ) . because such polynomials are irreducible. (d) If P ( X ) = Q l ( X ) Q 2 ( X ) Q n ( X ) and P ( X ) is divisible by an irreducible F ( X ) then F (X ) is a divisor of one of the Qi ( X ). 2.186 Introduction to Mathematics with Maple The right hand side is divisible by D ( X ). The statement is true for polynomials of degree 1. we have Q l ( X ) = c l P l ( X ) with a nonzero c1 because Q l ( X ) is irreducible over F [ X ]and P ( X ) is of positive degree. 1. The proof is inductive. . i = 1. . Renumbering it Q l ( X ) (if required. .Qm(X) (6.*. there are nonzero ci E F such that P i ( X ) = c i Q i ( X ) for i = 1. An informal way of stating the theorem in an easy to remember form is: the decomposition of a polynomial into irreducible factors is unique. and so then is Q ( X ).32) holds. * * * (6. We proceed by complete induction. with appropriate renumbering (if necessary). I . Since P l ( X ) divides P ( X ) it must. by (d). 6 . . If the polynomial is not irreducible then it is a product of two polynomials of smaller degree and by the induction hypothesis one of these is divisible by an irreducible polynomial and consequently so is the original polynomial itself. This follows by an easy induction from (c). then again Q l ( X ) is divisible by some irreducible polynomial. Substituting this . and the others correspondingly).
. For option we discuss here three possibilities.Q m ( X ) * * Since the polynomials in this equation are of degree less than n the theorem can be applied: it follows that I . If o p t i o n is left out Muple tries to factorize the polynomial in the smallest field which contains the coefficients. consequently I = m and also P 2 ( X ) = c2Q2(X). If we require that the leading coefficients in all polynomials are 1and group together identical polynomials.Qm(X)) It follows that P 2 ( X ) * . The number or each number in the list or set should be a root of a polynomial with rational coefficients. even in (?[XI.1.. A shorter name can be given for P ( X ) using the assignment operator. (6.(PI (X))"' ( P .e ( X )= (ciQ2(X)) . Actual finding of the factors can be quite difficult.4 we established the existence and uniqueness of the decomposition of a polynomial into irreducible factors.(PS(X))"". 4 ( X ) = czQz(X). instead of fi as in (b) the option is RootOf ( z 2. Merely to decide whether or not a polynomial is irreducible can also be quite difficult.1 0 The theorem does not assert that the polynomials P i ( X ) are distinct.Polynomials 187 into Equation (6.2). the factorization takes the form P ( X ) = . The command to use is f a c t o r ( P ( X ) . a nonzero element of In Theorem 6. ( X ) ) n z * * . Similar to (b) but the numbers are given with RootOf.. . c.m. One types in the word real. . The polynomial is then factored in R [ X ] .1 = rn . . for example The factorization is carried out in the smallest field which contains the given number or all the numbers in the list or set.the natural numbers ni and F . are uniquely determined.. a. For example. A number or a list of numbers or a set of numbers is inserted for option. f i ( X ) = c I P I ( X ) Q ~ ( X . Muple can be of great help with these tasks. option) .34) The polynomials P i ( X ) . The following session illustrates the use of f a c t o r . respectively.32) we obtain P 1 ( X ) P 2 ( X ). Remark 6.. .
however. This means that the polynomial is irreducible over the rationals. p:=xn4+5*x3+5*x+1: then we try to factor it. Now we try to factor the polynomial in the smallest field which contains square roots of the first few primes.real).5)1. because the task is just too big. (x+ 5.9999999999) Two things should be noted. > factor(p. It is hard to guess. > > Then we try to factor again. the Maple command ithprime produces what it promises. # x4 + 5 x3 + 5 x + 1 Muple returned the polynomial to be factored. # 4 (2 x2 + 5 x + fih x + 2) (2 x2 + 5x  dii h x + 2) We were lucky and found a factorization. Finally.188 Introduction to Mathematics with Maple > > > > # to save typing we name the polynomial.1930799616) (x2. not advisable to make the list too long. It can take a long time or Muple can refuse t o do it. it is.3722813232 x + . factor(p1.. A natural question arises as to how many primes should we include in the list. First we produce a list of these roots. > L:=[seq(sqrt(ithprime(i)) yi=l.179201362) (x+ . Firstly.. the factorization uses floating point approximations and is not exact: it is not an exact factorization in .L). the ith prime. factor(p. we factor the polynomial over the reals.
Polynomials 189 the smallest possible subfield of reals.3x5 x3 2x2 .b)(2x2 + 5x . 51 4 4 4 JLZGZYZ.6.2x2 (3) X1O0 x2 + p x q. Secondly.1.  x + 2.f i h x + 2).x 3. Assume that the greatest common divisor of PI ( X ) and P2(X) is o f the form (6.6.a)(x . x3 .1 and degH2(X) 5 n . x2 + p x q.2 Find the greatest common divisor o f each of the following pairs of polynomials (1) x6 3x5 3x4 3x+4x2 62 4.1. the polynomial 2x2 + 52 + f i h x + 2 can be factorized by solving the quadratic equation and a complete exact factorization in the smallest possible subfield of reals obtained. The capitalized command F a c t o r works similarly: it factorizes a polynomial in the field of polynomials with coefficients integers modulo a prime. 5 4 1 4 fifi1 J 4 2 a 4 Denoting the first root as a and the second as b we then have p = (x .1 Let F be a field and Pl(X). Exercise 6. x4 (2) x6 . + + + + + + + + + + + + x3 . Prove that the polynomials H I (X) and H 2 ( X ) can be so chosen that deg H1 (X) <_ m .3x2 + 32 . > solve (2*xn2+5*x+1in(1/2) *3(1/2)*x+2) . P2(X) two polynomials in F I X ] o f degree n and m respectively. (x2 + 10) (x2 + 91) We are unable to prove it here but it is an interesting fact that the polynomial X4 1 is reducible in the field of integers mod p for every prime p and is irreducible over Q. > Factor(x^4+1) mod 101.31). . + Exercises Exercise 6.
6.190 Introduction to Mathematics with Maple in the field Q[x]and in the fields of equivalence classes mod 3 and mod 5.6.4 Show that the polynomial X4 1 is irreducible over Q and is reducible in the smallest field which contains + a.4q < 0.] Exercise 6. + + + + + + + + . Exercise 6. Exercise 6. Exercise 6.5 Find all integers a for which the polynomial x3 + ax + 1 is reducible in Q[x]. [Hint: fl must be the roots.6 Factorize X 8 X7 X 6 X 5 X4 X 3 X 2 X 1in Q [ X ]and over the field which contains the root of the polynomial X2+X+1.3 Prove that the polynomial x2 +px+ q is irreducible over the reals i f and only ifp 2 .6.6.
from rationals to reals. 191 . + + + 7.a0 = 0.1) has a solution.3. with a r c E R. hydrodynamics.2 = 0 is solvable? The obvious answer is yes: the reals. Complex numbers now play important roles in physics.1. This will be the field of complex numbers. from integers to rationals.1 Field extensions During the history of civilisation the concept of a number was unceasingly extended. Let us consider the following question. from positive numbers to negative numbers. electromagnetic theory. Is there a smaller field? The answer is again yes: there is a smallest field which contains rationals and namely the intersection of all fields which contain Q and the (real) number Can this field be constructed directly without using the existence of R? The answer is contained in Exercises 7. Our introduction follows the same idea although in a modern . a.xn anlxn' .mathematical setting. Let us consider the following problem: is it possible a. We now embark on an extension of reals to a field in which the equation J2+1=0 (7. electrical engineering as well as pure mathematics. etc. n E N.Chapter 7 Complex Numbers We introduce complex numbers.17.. Mathematicians were led to complex numbers in their efforts to solve socalled algebraic equations. that is numbers of the form a+ bz where the number z satisfies z2 = 1.. that is equations of the form u.1. Is it possible to extend the field of rationals to a larger field in which the equation E2 . The process of extension of a field can be most easily carried out in the case of a finite field.
192 Introduction to Mathematics with Maple to extend G F ( 2 ) in such a way that the equation' has a solution in the extended field. (See Multiply in Table 7. .) It is now easy to fill the rest of the addition table: remember 1 1 = 0 in G F ( 2 ) and this must also hold in the extended field.1 Addition and multiplication in G F ( 2 ) Add Multiply Denoting by k the solution of the equation we shall try to augment these tables by additional rows and columns in order to obtain addition and multiplication tables for the extended field. (See Table 7.2 Addition in G F ( 4 ) Add k l+k k l+k l+k k l+k 0 1 1 0 Filling the augmented multiplication table is a bit more involved and we use the Equation (7.2. addition and multiplication would be described by the augmented + + + 'It is easily checked the the equation is not solvable in G F ( 2 ) . let us denote it by 1 k and augment the table by another column and another row. The following Table 7.) If the extension existed. The entry in the second row and third column of the addition table cannot be equal to 0 or to 1 or to k .1 = k 1. So it is another 'new' element. we add a row and column headed by k . For instance. + + Table 7.3. (1 k ) k = k2 k = 1 = 1. k2 = k . Table 7.1 describes addition and multiplication in GF(2). Firstly.2) to do it.
b f i ) ( a 2 .1. k.2 We succeeded in our task of extending GF(2) in such a way that in the extended field the Equation (7. b E Q } c R.1 Prove: The set {a b f i . [Hint: ( a .1. 1. if you wish to emphasize that it is not the natural number 2). Q(a) + Prove that F a is a field. + Exercises @ Exercise 7.1.g.b) + a b f i ] Exercise 7. Then realize that 2 1 = 0 and 2 x 2 = 1 and the rest is easy. We now reverse the process.d) = (a (a.Complex Numbers 193 Table 7. For the four elements 0.c+ d). bc + ad).1. b) + @ Exercise 7.b2)l = 1 . + + k is this element itself. is easy and we leave it to the readers.] + + 2E.2. + 2bd. This field is often denoted by Q(&). which is called GF(4). a E Q. 1 k we define addition and multiplication by Tables 7.1) has a solution.1 is isomorphic to the field FJZfrom Exercise 7. This field contains Q and 4 and is the smallest field with this property.2 + (C.4 Write down the addition and multiplication tables for the field of three elements (denoted by G F ( 3 ) ) .1. the multi . 3 Let FG be the set of ordered pairs of rational numbers with the following definitions of addition and multiplication (a. The verification of the field axioms for the new field.3 Multiplication in GF(4) Multiply tables.2 and 7.3. the inverse for addition for the element 1 plicative inverse to k is 1 k . b)(c. [Hint: ( a + b J z ) ( a . @ Exercise 7. Denote it by 2 (or by 2.d ) = (ac + b.3 Show that the field from Exercise 7.1. [Hint: The third element must be equal to 1 1. together with the same addition and multiplication as in R constitutes a field.
4) + + + +a.194 Introduction to Mathematics with Maple Exercise 7. if P ( X ) is a polynomial with real coefficients and z a solution to the equation x 2 1 = 0 then all the numbers P(z) must be in C. namely Pi(%) P2(2) = (Pi P2)(2) and Pi(z)P2(2) =( P i P 2 ) ( 2 ) . Dividing P ( X ) by X 2 + 1 gives (7. Before we deal with following statement: M4 we prove the If P ( X ) E R[X] then P(z) = 0 if and only if X 2 (7.] 7. Before we construct it. A4. It follows that a = 0 and consequently P ( X ) is divisible by 1.1 need verification. we examine its structure.1.2 Complex numbers The field of complex numbers C is the smallest field which contain R and the root of the Equation (7.1). [Hint: Use the same method of extending the addition and multiplication tables which we used for construction of GF(4). b E R. If 2 is the zero polynomial then axiom A3 is satisfied with the zero element Z ( i ) . Let (7. Firstly. Now b = 0 otherwise z = a/b E R. Substituting4 X = z gives a bz = 0. In order to show that P(z) # 0 has a multiplicative inverse we note first that P ( X ) is not divisible by X 2 1. The inverse element to P(z) under addition is P(z). hence we have A4.1) is solvable. Verifying axiom M3 is also easy: if + then U ( i ) is the multiplicative unit. 4See Substitution rule on page 173 + + . Only axioms A3.3) hold. M 3 and M 4 from Table 4. the greatest common divisor of P ( X ) and X 2 1 is 1 and there exist x2+ + + + ~ 3Addition and multiplication of these elements is understood in the obvious way.1).3) + 1 divides P ( X ) .5 Extend GF(3) to a field of nine elements in which the Equation (7. with a. We now show that there are no other elements in C by showing that these numbers3 themselves form a field and so they form the smallest field which contains the reals and the solution to Equation (7. P ( X ) = Q ( X ) ( X 2 1) bX The ‘if’ part is obvious. Since X 2 1 is irreducible over R. assuming that C exists.
it is clear that with addition. no matter how the result appears.4) we have cx = P(z) = a bz. call its elements complex numbers and in the course of this section we shall prove that C is a field. in addition we replace the summands by equal5 elements the result will be only seemingly affected: by (7. We denote it by C. a crucial difference.H1 ( X ) such that P ( X ) H ( X ) ( X 2 + l ) H l ( X )= 1. Every element of C is of the form a b2 with a.(1) EE 0 mod ( X 2 l). however. we identify P ( X ) with Q ( X ) if P ( X ) = Q ( X ) mod ( X 2 1).1 that elements of an equivalence class are sometimes said to be equal.1) and as such we denote it by 2. Since we identified X 2 with 1 it follows that the indeterminate X itself is a solution to Equation (7. it will always consist of equal elements. a = %a. Motivated by this we introduce an equivalence relation into R [ X ]by saying that P ( X ) and Q ( X )are equivalent if their difference is divisible by X 2 + 1. b = %. by (7. Since R[X] is a ring. As we have just seen P(z) = Q(2) if and only if P ( X ) . multiplication and identification just explained the resulting algebraic structure is a ring.In the same vein we identify here X 2 with b 26 ' 1 since X 2 . by the equivalence for common a 2a fractions we identify . If. We denote that by P ( X ) = Q ( X ) mod ( X 2 + 1). We call a . + (7. and more generally. In symbols. Addition and multiplication in W[X] is consistent with equivalence mod ( X 2 1). It is worth noting that + + + + ~~~ 5That is. Similarly for multiplication.Q ( X ) is divisible by X 2 1. b the real part of Q! and the imaginary part of a . if P l ( X ) = & ( X ) mod ( X 2 1) and Q l ( X ) EE Q 2 ( X ) mod ( X 2 1)then + + + + We mentioned in Remark 3. For example. Polynomials are equal if and only if they have the same coefficients.5) Putting X = 2 shows that H(2) is the multiplicative inverse to P(2).Complex Numbers 195 polynomials H ( X ) . More precisely. Indeed. b E R. respectively. There is. equivalent mod ( X 2 + 1).with . . This relation is indeed an equivalence: it is obviously symmetric and reflexive. The proof that it is transitive is easy and left as an exercise.6). The similarity in addition and multiplication of numbers P(z) and polynomials with real coefficients leads naturally to the idea of using R [ X ]for the construction of the complex number field.
2.3 for the absolute value of real numbers. . p E C + It follows that if P ( X ) is a polynomial with real coefficients then P ( 6 ) = ( P a ) . am The first item is obvious. and we denote it by 6. It is easy to see that for a . If b # 0 then a is called an imaginary number6: if a = 0 then a is purely imaginary.bz is called the complex conjugate of a bz. Then we have 6This terminology comes from times when mathematicians regarded complex numbers as mysterious. Proof. To prove (iii) we first note that ap + tip is real and consequently it is equal to its real part. For (ii) we have laPl = d(ap)(aP)= = IalIPI.1 Absolute value of a complex number The number d m is called the absolute value of the complex number a = a bz. 7.l a 1 2 Sa and + l a 1=1 6 1 . It is always nonnegative and if b = 0 then a is real and the notation is consistent with that introduced in Section 4. This is very often used. We denote it by 101.196 Introduction to Mathematics with Maple the imaginary part of a complex number is a real number. These days an imaginary number is a concrete mathematical object. Obviously la1 2 %a. The number a .
3.1 Prove: If n E N and z E C then lznl = then this equation also holds for n E Z. Let us emphasize laI2 C is the set of numbers of the form a + bz.2 @ Exercise 7. By (i) we have ad = laI2 and if a tive inverse of a is . . 73:a = 0.a.2 to denote by fi the nonnegative number p such that p2 = a. + 2m. a  # 0 then a. ~ 1 ~ . If a < 0 then we define fi = Now we turn our attention to the case a = a zb with b # 0. C is a field. Now it is sufficient to take square roots.3 Prove (iv) of Theorem 7..1.1 For complex numbers 0 1 .. The multiplicalaI2 d consequently C is a field.3 we 0 leave it as an exercise. that is a number whose square is a.2. Exercises @ Exercise 7. . hence in our search for the square root of CY we can look for a p with a nonnegative real part. Let us recall that if a is real’ and nonnegative then we agreed in Definition 5. The following calculation yields the real part of p.2. prove: 7. 1 .2. If p2 = a then also (p)2 = a .Complex Numbers 197 In the last step we used (ii) and (i). a 2 .2.2 Square root of a complex number In this subsection we prove that every complex number a has a square root.If z # 0 @ Exercise 7. where a . [Hint: Mimic the proof of (ii) of Theorem 4. .= 1. Since the proof of (iv) is similar to the proof of (ii) of Theorem 4. b are real and z2 = 1.
Readers are warned that the rules which govern operations with square roots of positive reals as given in Theorem 5. ~ ~~ ~~~ ~~ ~~ ~ Remark 7 .6 are not valid for complex numbers.B 3 P = 3 P2 and then by using Equation (7.2) and fi: for a < 0 these are and For an imaginary number Q! = a zb with b # 0 there are also two square roots: one is given by Equation (7. w are complex numbers. lbIJ2 (7. although this symbol is commonly used.10) we have Consequently P=1 JZ Ja+JaZ+b2+2b J . 6 + z m z a .w .17) and the other is P.4 is false if z .1 J.2.4 and 7.198 Introduction t o Mathematics with Maple To determine the imaginary part of P we note first that 28. Some counterexamples are given in Exercises 7. [Hint: Choose z = w = 1.17) Let us summarize: Theorem 7.2.2 (Square root of a complex number) For a real number a # 0 there are two distinct square roots. For a > 0 these are (Definition 5. Exercises Give an example showing that the equation fifi = Exercise 7. 1 We deliberately avoided the use of the symbol f i for the square root of the complex number a .2.5.
returns 1 I and sqrt (2410*1) . g.2. We can also use Example A. gives 5 .1. The length of this segment is lal. 7.I. gives f i . Quite often Maple returns the question.4 Geometric representation of complex numbers. We can use a floating point approximation. . If the square root of a complex number is relatively simple expression then Maple provides the exact answer. The absolute value of a complex number z is entered into Maple as abs (z).2 we have . In this and the next subsection we rely on readers’ knowledge of trigonometric functions from their previous work. The plane in which the complex numbers are depicted is often referred to as the complex plane or the Gaussian plane or the Argand diagram. Or if we insist on obtaining an exact result we use the command evalc (sqrt (4+I) 1. a2). + + + + * 7. In such a case we have two options. The sum a+/3 with /3 = bl+zb2 is then represented by the diagonal of the parallelogram constructed with sides a and /3.= . With the notation as in Figure 7.To obtain Z enter conjugate (2).015329455 .0+1)yields 2.2.Cornpl ex Nurnbers 199 Exercise 7.17).3 Maple and complex numbers Maple uses the letter I for 2. sqrt (4. The arithmetic of complex number in Maple brings nothing new. The command evalc simplifies a complicated complex number to the standard form a + bz.2 in Appendix A which provides a Maple program which calculates the exact value of the square root of a complex number according to the formula (7. sqrt (4+I).5 Show that the equation 8. e. y as a segment with the initial point at the origin and end point ( a l .24809839341.is false for z = 1. For instance. conjugate ( (4+6*1) / (11)) . sqrt (2*I). see Figure 7. for instance abs (l+I) .2. Trigonometric f o m of a complex number A complex number a = a1 +za2 is represented in the plane with coordinate axis 2. For instance. A very useful representation of complex numbers is achieved with the help of trigonometric functions. They are rigorously defined and their properties established in Chapter 14. gives only d m . for instance 1 32 must be typed as 1+3*1. yields 1 51. Readers must not forget the operator * when entering a complex number.
18) if it is additionally required that T < 4 5 T and then it is called the argument of x. =~ z ~ ~ w ~ ( c o ssin4sin$ ~ c o s ~ + zcos+sin$ + zsin4cos$).1 Sum of two complex numbers z = IxI(cosq5 + zsinq5). The sum of the arguments can be used as the angle which . The absolute value of the product of two complex numbers is the product of absolute values.18) also uniquely determine 6.200 Introduction to Mathematics with Maple Y 0 X Fig. 7. The inequalities 0 5 4 < 27r together with Equation (7.If w = Iwl(cos$ zsin$) then we obtain for the product + zw = IzIIwI(cosq5 + zsin$)(cos$ + zsinq).18) For z # 0 the angle q5 is determined up to an integer multiple of 27r: it is uniquely determined by Equation (7. = lzllwl (cos(6 + $) + 2 sin($ + $)I. (7. In Muple we have argument ( z ) .
for n E N.7r. where for sake of simplicity we took Izl = I w I = 1. zn = 1x1” ( cos(n4) + z sin(n4)). . 7r]. For z = w we have and by simple induction. (7.19) This equation is called Moivre’s formula. if the sum lies in the interval ] .Complex Numbers 201 Y z I I I I I I I I I I I Izl sin+ I I I I I I I I I I I I 0 Fig. 8 0 n e has to be careful not t o say that that the argument of the product is the sum of the arguments of the factors: it is.8 This is illustrated in Figure 7.3. 7.2 Trigonometric form of a complex number is subtended by the representation of the product and the real axis. however.
In this subsection we assume a # 0.6 to express sin 54 in terms of sin 4 and cos 4.] 7. can also be used.3 Multiplication of complex numbers Exercises Use Equations (7. > solve (x13=1).20) with n E N is called the binomial equation.18).202 Introduction to Mathematics with Maple Y 0 . 7. Exercise 7. (7. say n = 3 and n = 10. since the case a = 0 is trivial.19) and the Binomial Theorem Exercise 7.2.7 Express sin4 4 i n terms of cos 2 4 and cos44. Let us try Maple for a = 1 and some n.2. .5 The binomial equation The equation of the form xn = a (7.2. The Maple command (combine( (sin(x))4) . 0 / I 1 X O I Fig. [Hint: Set z = cos 4 + z sin4 and then calculate ( z + 2)* in two different ways.
24) for some integer k . /W.i=i.x)J and employ seq(evalc(L[i]).I& 2 2 2 2 This is a nice result. but unfortunately for larger n Maple provide a complicated answer.lo). We use it for z = Izl(cos ++z sin 4) and write a = lal(cos a z sin a ) with 0 5 a < 27r. say.25) . L:=Csolve(x^(lO)l. .19). (7. The key is the Moivre formula (7.23) The last two equations in turn are satisfied if and only if 4 = (a n 1 + 2k7r). Fortunately there is a simple way of dealing with the binomial equation for any n./W /W.Complex Numbers 203 1 1 1 1 1. Obviously not all these values of 4 can be distinct: by the uniqueness of the trigonometric representation they will be if 05a + 2k7~< 2nn. m (7.J %1:=J C f i %2 := The result is complicated but can be simplified by using the command evalc. (7.21) (7.20) is satisfied if and only if / 4 4 dCT5 + 1 x 1= cos n$ = cos a sin n4 = sin a. One way to do it is to to put the solutions in a list first.22) (7. Equation (7.+I&.
There are n roots and in order to give a well defined meaning to the symbolg some choice must be made. Maple's choice might surprise you.20). For a = 1 they are denoted &k and by Moivre's formula &k = E ! . Now all the numbers (COS CY +n2k7r + sin "")> n are distinct and therefore they represent all solutions to Equation (7.2. Maple will made a choice for us! It finds the floating point approximation of the root for which the argument has the smallest absolute value and for which the imaginary part is nonnegative.26) Consequently k = 0. For instance. G G v v 'For n = 2 we write simply f i instead of G.(. One must be aware that if some choice is made for the symbol fi this choice might be inconsistent with previous definitions. 2.2. the answers 2 are (1) and . Maple simplifies what can be simplified but leaves 3 the choice of (1)t = to the readers.1) . We explained it because this notation is commonly used and because it helps to understand what Maple does with nth roots of a. Some other rules are left for Exercises 7. in particular with Definition 5. We shall call ~1 the nth primitive root of unity. This is undesirable and it is best to avoid the use of the symbol for a complex a as much as possible.2. The solutions to Equation (7. . It is possible to derive some rules for computation with fi.107.1. However. In the Gaussian plane the numbers &k are equidistantly placed on the unit circle: &k+l is obtained by rotating &k by an angle 27r/n.204 Introduction to Mathematics with Maple This leads to the condition (7. .One such obvious rule is ( fi)" =a for any complex a. 1. or root [31(8/27) . we agreed on Page 154 that would be 2/3 but if you ask Maple root [31(1) . The following Maple worksheet illustrates this. xk = . . n .20) are called nth roots of a and are denoted by @. . if we use the floating point approximations with the root command. In order to avoid confusion or even a mistake the following convention is made: whenever a choice f o r the symbol fi is made that choice must be kept consistently during a calculation or a particular investigation.12.
O) .2. provided ~1 denotes the nth primitive root of unity.11 Prove: If a . as shown by Maple.] Exercise 7. (3) a = 2 22 and n = 3.8 Show that Equations (7. . 2421 but > root [ 3 ] (2.2.2.0+2.] Exercise 7. [Hint: All the numbers in the set are solutions to Equation (7.20). [Hint: multiply Equation (7.23) by sina! and add.10 tion (7.3660254037 I This is an example where @ # a .23) are satisfied if and only if (7.22) and (7.000 because this root has the maximal argument and Maple chooses the approximation with the smallest argument. + Exercise 7. b are complex numbers.24)i holds.5000000001+ 3660254037 I The number 3660254037 is a floating point approximation to We 2 did not get the answer 1.Complex Numbers 205 The choice is up to us.20) for (1) a = 1 and n = 8. fi.20) and are distinct. > (1+1)3.366025404 + . 1. Exercises Exercise 7. > root [ 3 1 ( 1 .22) by COSQ and Equation (7. (2) a = 1 and n = 4.20) then Prove: If a # 0 and fi is a fixed solution to Equa is the set of all solutions to Equation (7. n E N and the choice of value for each of f i l and is fixed and if ~1 denotes the nth primitive root of unity then there exists a Ic E N such that fiG = ~ .9 Solve Equation (7.O*I).2.
2. n E N and the choice of value for each of G.] @ Exercise 7. Use the previous exercise.206 Introduction to Mathematics with Maple E.20).ab = 0.[Hint: Both numbers f i Gand m are solutions to the equation m.] %( . [Hint: x = f i G satisfies Equation (7. xn .12 Prove: If a . m. b are complex numbers. is fixed then it is possible to define && in such a way that f i G = Consider the case of a = b = 2.
Chapter 8 Solving Equations In this chapter we discuss existence and uniqueness of solutions to various equations and show how to use Maple t o find solutions.2) For greater clarity we printed the implication signs. starting with the equation itself and ending with an equation (or equations) of the form x = a.1).1) usually consists of a chain of implications.G. 8. 3 j x = 2' (8. It is a good habit always to check the solution by substituting the found value back into the original equation. For instance: 1 x+l 1 += 0 =2 x + 2 + x + 1 = 0.1). Solving an equation like (8.1) means finding all x in dom f which satisfy Equation (8.1 General remarks Given a function f . but implications are always understood automatically. This 'A more complicated equation of the form F ( z ) = G(z) can be reduced to the form given by taking f = F . solving the equation' f (4 =0 (8. 207 . Any x which satisfies Equation (8. In this chapter we assume that dom f is always part of R or C and it will be clear from the context which set is meant. We deal mainly with polynomial equations in one unknown and add only some basic facts about systems of linear equations.1) is called a solution of Equation (8. x+2 =2 22 + 3 = 0 .
1= J i T Z . x2 = 0. The conclusion is clear and easy: the original equation had no solution.2 2 + 1 = 5 . x = 2 or x = 2.22 + 1 = 1 .1. means the square root of a 31n this book unless otherwise stated the symbol nonnegative number y: if y contains an unknown or unknowns then the search is for quantities which make y nonnegative. (83) Substituting 0 in the original equation gives 1 = 1. (8. . in other words. If it does not then the chain of implication can lead to a wrong result.3): the second equation does not imply the first.22. a false proposition can lead t o anything. It is preferable to use equivalences rather than implications. it has a far more fundamental reason. it is assumed that a solution exists. Let us illustrate it by one more example.4) Exercises Exercise 8.1 = dG%. x = 0. Solving an equation starts with an assumption that the equation is satisfied for some x. x . If we do not use equivalences then the found value(s) need not be a solution and we must perform the check. It is not always possible to employ equivalences and this was the case in Equations (8.208 Introduction to Mathematics with Maple is not a mere verification of the calculations. x2 = 4.22.2 Consider the following example3 x . Back substitution shows that 2 is a solution and 2 is not. It is then possible to conclude that the value found in the last equation is indeed a solution. x2 . x2 . The verification is not necessary from the logical point of view but it is still useful for checking the calculations.1 Solve the following equations  x +(I) 21 X X 4x2 (2) 21 x += X t l 927 221. x+l 221' 2Remember.
.Solving Equations 209 Exercise 8. d x.1)2 = a . range). For example. unknowns). sometimes f solve gives a more useful result than solve.1 = 0 rather than solve it exactly.3).2 Equations (8. The structure for solve is solve (equation. 8.618033989 Both commands have some options and sometimes it is important to employ these options. . or if there are several equations and several unknowns solve(equations. 1.unknown).1.1 = &?%. unknown.6180339890. Although generally speaking it is better to obtain an exact solution. This is a natural result of computational inaccuracies.2 ( Solve the equation z . The following Maple session illustrates t h k 4 + 1. The structure for f solve is f solve (equation. Moreover. An exact solution can be converted to a floating point result by the command evalf.618033989.6180339887 > solve (xA2+xi=0). 1 1 1 +fil 1 fi 2 2 2 2 > evalf(%). for some equations this might take too long and for some equations it might even be impossible. There are two commands: solve provides an exact solution and fsolve gives a numerical solution. it is usually preferable to use f solve on an equation like z2 z . 4Note that the Maple floatign point results differ in low decimal places.2 d a . I [Hint: Mimic Maple commands solve and fsolve Maple can efficiently solve practically any equation. .
4).3) and (8. However. 51f there is one. x := 2 1=1 Maple has its own command eval for checking. Example 8. b are real numbers.x=2) . If there is only one unknown in the equation. Maple correctly did not return any result: there is no result t o be obtained. if there are several unknowns or equations they should be entered as sets. Eq := x . where a. or all complex solutions if the option complex is given. 2 This is very pleasing: Maple returned the correct answer and omitted the spurious 'solution' x=2. > x:=2.x) . Eq:=xl=sqrt(l2*~).b. The range can also be given by the word complex. > eval(eqn. for instance with { and 1 around them. If the option range is given as above. the equation has no solution. > eqn:=xi=sqrt (52*~) . The range could be given as a. or could be . As we know. .1 = JTZG > solve(Eq. this option can be omitted.eqn.1 = JTTEi > solve(eqn.i n f i n i t y . If the equation is a polynomial equation then f solve will find all real solutions. . There is additional useful and important information on solve and f solve in Help.1 > Now we use Maple for solving (8.x) . Maple will find a solution5 although there might be many solutions in that range.210 Introduction to Mathematics with Maple The options unknown or unknowns identify the unknown or unknowns for which the equation should be solved. i n f i n i t y . Maple can also check the solution. In this case. eqn := x . fsolve searches for a complex solution.
i2)=1O3OOO) .781236788 ~~ ~ An interest rate of 178% is clearly wrong. We shall look for an explanation shortly. 1. > fso~ve~10*(l+x)~(~2)l~103*x*(l+x~n(12)) .12).J: 10000 J:(l+J:)12 So the equation to be solved reads l O ( 1 +z)l2.6 > with(f inance) . We wish t o compare this offer with other investments and therefore want t o know the interest rate. .02432202661 Why was the first result wrong? The explanation lies in the formula for annuity. but first we try t o employ options with fsolve. 1 10000 . 1. .4.02432202661 ‘The finance package w a s introduced in Example 1.2 An agent offers an annuity of 12 years duration with yearly payments of $10 000 for $103 000. In order t o find it we use fsolve and annuity. hence we try .10 = 103x(1+ x ) 1 2 . O . We know that the rate of interest must be positive and less then loo%.x.Solving Equations 21 1 1=1 Example 8. > fsolve(annuity(1OOOO. > annuity(IOOOO.781236788. .x. This is a polynomial equation and f solve will provide all real solutions.
The annuity problem leads to a polynomial equation with three real solutions.158728473 .95284’78647 > X:=(%).3 This example was used by mathematicians to tease lay people.1 Grazing goat A goat is tied up a t the circumference of a circular meadow of radius one. > 1.22) + > > fso~ve(x*2*(cos (~))2+Pi/2~(1/2)*sin(2*x) = Pi/4.1 shows half of the circular yard. X := . provided we are certain in what range the desired solution lies.x. The length of the required rope is 2cosx and the area the goat can graze is 42 cos2 x n .22 . Example 8. The second attempt was successful. .sin(n .212 Introduction to Mathematics with Maple The discrepancy has disappeared. Using the notation from this figure we immediately recognize that the angle x = n/4 produces rope that is too long and that the angle x = n/3 produces rope which is too short. It is not difficult but it requires a numerical solution. .Pi/3) .9528478647 2*cos(X). 8.Pi/4. and this root was not the solution of the original problem. The first attempt picked up the negative root of the polynomial equation. with the goat tied a t the angle x. The moral of this story is that we should always use the options with fsolve. * 0 Fig. How long should the rope be which allows the goat to graze an area of half the circle? Figure 8.
The degree of P and the coefficients of P are also the degree of the equation and the coefficients of the equation. and of completeness of reals. pp.570796326 . is called algebraic. 1.) = 0 .5) where P is a polynomial. which is 7r/2.5) and is also called a root of Equation (8. * . Theorem 8.5). The name is slightly misleading: a better name would be “The Fundamental Theorem of algebra of complex numbers”. F.1 (The Fundamental Theorem of Algebra) An algebraic equation with complex coefficients and of degree at least one always has a solution within the field of complex numbers. > Pi4*X*(cos(X)) ^22*((Pi2*X)*(I/2)(I/2)*sin(Pi2*X)) 1. includes also the real and the rational coefficients. 8. A proof accessible with our level of knowledge can be found in Kurosh (1980. of course. > evalf ( P i / 2 ) . however all of them use the concept of continuity. There were many proofs given. but this name is still not perfect. (8. Gauss at the end of the eighteenth century who proved that algebraic equations always have solutions. 7This. respectively.570796327 This shows that we obtained the correct answer. We shall not prove the theorem here. From ancient times mathematicians tried to solve algebraic equations and it was only the German mathematician K. which will be considered in Chapter 12. These parts of mathematics are not considered to belong to algebra. we calculate the area which the goat cannot graze and compare it with the area of the halfcircle. 142151). since it requires too much preparation. We shall consider only equations with complex coefficient^.3 Algebraic equations An equation of the form P(.^ A zero of P is obviously a solution to (8.Solving Equations 213 The value for the length o f the rope looks about right: however t o be sure we found the right result.
37125*yn2 .4 > > First we define the polynomial f:=y>y^6+ 2*y^5 + 435*yn4 + 11025*~^3 .~ 2 then it is easy to see that W = a12 . However. q = 2122. provided each root is counted as many times as its multiplicity. ) ~ Rational roots It is clear (for instance from the Vieta formula for a n ) that if the equation P ( x ) = 0 has an integer root xi then zi divides a n .7) are called elementary symmetric polynomials. Multiplying through on the right hand side of Equation (8. We do not have much use for this otherwise important theorem. can by expressed as a polynomial in elementary symmetric polynomials. The next example shows how to use this for finding an integer root. that is a polynomial which is not changed by a permutation of indeterminates. say a . so we do not prove it. that is linear. if P ( X ) E C[X] is of degree at least two then it has a root. Indeed. Writing P ( z ) as a product of its irreducible. when P(x)= x 2 + p z p = (21 2 2 ) . we illustrate it by a simple example. 8Readers are probably familiar with these for n = 2.2122. + + q and . If W = ( 2 1 .6) and comparing coefficients leads to the socalled Vieta's formulae' The polynomials on the right hand sides of Equations (8. and is therefore divisible by X . Hence it is not irreducible. Example 8.253125*y + 759375. The socalled fundamental theorem on symmetric polynomials states that any symmetric polynomial.214 Introduction to Mathematics with Maple An important consequence of the Fundamental Theorem of Algebra is that the only irreducible polynomials in C[X] are linear. factors we have Not all xi need be distinct.a. however it follows that an algebraic equation of nth degree has n complex roots.
8) If p and q are relatively prime integers and p/q is a zero of Q then p divides an and q divides ao. Let ~ ( 2= ) aOzn + u1xnl + . 3125.9) . 1875. 75. 375. 6075.3 on page 80. 16875. 3375.151875. 51 Since there are very many very large entries in di t o do this work with pen and paper would be almost forbidding. 15. 675. > podi :=convert (divisors (759375). Therefore the rational zeros of Q can be found among fractions of the form divisor of an divisor of a0 (8. 9375.di) . This gives a long list di. Rational roots can be found similarly.op(podi)] . 625. 243. from which we select those z for which f(z)= 0. 81. 28125. divisors) . * * (8. For that we need the number theory package. which we put into a list we call podi (for positive divisors). 45. good := x + is(f(x) = 0) > select (good. podi := [l.253125 y + 759375 Next we use the Maple command divisors t o find the positive divisors of the absolute term. 135. Indeed.1215. > di:=[op(podi) . [3. 405. 2025.Solving Equations 215 f := y + y6 + 2 y5 + 435 y4 + 11025y3 . but with Maple it was easy. 50625.3. 225.37125 y2 . 10125. It follows similarly that q divides ao. we have Since p and q are relatively prime. 9. + an.list) . > with(numtheory. 84375. > good:=x>is(f (x)=O) . Maple provides the set of divisors. 7593751 The next step is to create a list of all divisors. 1125. 25.253125. p divides an. The procedure for selecting elements from a list according to a given condition was discussed in Subsection 3. We must not use letter D to denote the divisors. since it is protected. 30375. 27. 5. 5625.3. 125.
If there are no rational roots Maple returns an empty list.35.5 222 . multiplicity]. . + Example 8.1. We now compare roots with our previous results.5. but with Maple it is easier to implement.216 Introduction to Mathematics with Maple (See also Exercises 8. The next example shows the use of Maple for finding rational roots. We make a substitution x = y/uo and seek the integer roots of u.17x2 + Q:=x>6*xn317*xn2+22*x35.lQ(y).list) :div:=[op(L) . For justification see Exercise 8. consequently. 1 1 1 > roots(P(t1).3.35 > x :=y/6 :36*Q (XI . 1 1 . The method we employ is not recommended when using mere pen and paper. > Let us now find the rational roots of Q ( x ) = 6 x 3 . Maple has a readymade command for finding rational roots of polynomials. the rational root of Q is 7 / 3 . > P has a root 14.div) . Obviously.3 we have to check 12 possibilities to find that 3/2 and 1 / 2 are the rational roots.17y2 + 1329 .3.op(L)I : good:=t>is(P(t)=O) : select(good. “3.1260 > > > > ~:~~>~n317*~n2+132*~~1260~ with(numtheory. for a polynomial of higher degree with large coefficients the computer is needed.28. Q := II: + 6 x3 .and Maple answers with a list of pairs [root. It is roots. y3 .[5.) If Q(x) = 4x2 4z .divisors): L:=convert (divisors(l260) .3.17 x 2 + 22 x .
Solving Equations
217
Multiple roots It is important that solving of an algebraic equation can always be reduced to solving another algebraic equation with simple roots. Let k be a positive
integer and
with Q ( u ) # 0. Then by Equation (6.27)
D P ( x ) = k ( x  U)"'&(X)
+ ( X  u)'DQ(x) = ( X  u)"'F(x),
(8.11)
with F ( a ) = kQ(a) # 0. Consequently a is a root of D P ( x ) of multiplicity k  1. A root of P of multiplicity k is a root of multiplicity k  1 of the greatest common divisor of P and D P . Thus P ( x ) divided by the greatest common divisor will have only simple roots. > P:=X>X^~X^~+X^~X^~X^~X^~+X+~;
x7  x5 x6  x4  x3  x 2 + x + 1 To investigate the multiplicity of roots of P , we look at9
P
:= x
4
+
>
gcd(P(x) ,D(P) (x));
x3 x 2  x  1 This polynomial is of degree three, so P must have three multiple roots. To find whether or not some have multiplicity greater than two, we again look for the greatest common divisor of this polynomial and its derivative > gcd(%,3*~^2+2*~1) ;
x+l
1 is a triple root of P and it is easy to conclude that 1 is a double root.
+
We could also find by inspection that 1 was a root of P , and use the Taylor polynomial to determine its multiplicity.
>
taylor(P(x) ,x=l,8) :convert(%,polynom);
16 ( x  1)2 40 ( x  1)3 44 ( x  1)4 26 ( x  1)5 8 ( x  1)6 ( x  1)7
+
+
+
+
+
x5 3x4 1 is now an obvious root
>
gThe command D(P)
+
+ 4x3 + 4x2 + 3 2 + 1
taylor(%,x=l,6) :convert(%,polynom);
(XI
in Maple produces the derivative of P(x>.
218
Introduction t o Mathematics with Maple
2 (z 113  2 (z 114 (J: As before we obtained that 1 is a triple root.
+
+
+ +
115
Real roots
In this subsection we consider only polynomials with real coefficients. If a is an imaginary root of P then so is a. Indeed
since Ei = ai because the coefficients are real. This leads to
Theorem 8.2 If a is a root of multiplicity k of an algebraic equation with real coefficients then so is 6. The number of imaginary roots of an equation with real coefficients is even.
Proof. We have just seen that the theorem is true for k = 1. We proceed by complete induction. If P has an imaginary root a then P ( x ) = (x2 (a 6)x aii)Q(x),with Q having a as a root of multiplicity smaller than that of P. By the induction hypothesis the multiplicity of 6 for Q is the 0 same as that of a and the assertion follows.
+
+
Mathematicians throughout the ages found many theorems which help to determine the number of real roots of a given polynomial and isolate them in separate intervals. The best known are the theorems of BudanFourier, Sturm and the so called rule of Descartes. The interested reader is referred to Kurosh (1980). Maple has a tool for an application of the Sturm Theorem. We shall not discuss these matters here, but mention a Maple command realroot. It responds with a list of intervals each containing just one root, selects in some reasonable way the length of these intervals and sometimes returns an interval [a,a] which indicates that a is a root. Multiplicity of roots is not taken into account.
>
realroot (x^6+x^5xA2+1);
"1, 11, 127 1 1 1
Solving Equations
219
This means the polynomial has a root 1 and a real root in the interval [2,11. If there are no real roots the command returns an empty list.
> r e a l r o o t (xn4+I) ;
The concept of an algebraic solution
By an algebraic solution of an algebraic equation we understand formulae for the roots which use only the coefficients of the equation and only operations of addition, subtraction, multiplication, division and extraction of integer roots. This is a useful description but it is not precise enough. We shall say that we found an algebraic solution to the equation
with complex coefficientslO or that the equation is solvable in radicals if we define a finite chain of binomial equations
...
with the following properties:
(1) the numbers mi, i = 1,..,k are positive integers; (2) A1 is an element of a field F which is an extension of Q by the
coefficients of (AE); (3) Ai for i = 2, . . . , k is an element of the smallest field which contains F and the roots of previous binomial equations; (4) all the roots of Equation (AE) lie in the smallest field which contains F and roots of the binomial equations (A1)(Ak).
Example 8.6
Consider the equation (x3  a)2
+ 2 = 0. We see that the
"The concept of an algebraic solution to an equations with coefficients in an arbitrary field can be defined rather similarly but we shall not do it here.
220
Introduction to Mathematics with Maple
required chain of binomial equations is
It might be plausible to think that every algebraic equation is solvable in radicals: the equation was formed from the roots of the equation using addition, subtraction, multiplication, division and taking powers so it should be possible by using only addition, subtraction, multiplication, division and by extracting roots from complex numbers to untangle the equation and get back to the roots of the equation. Mathematicians for a long time held this false belief. It was the Italian mathematician Paulo Ruffini who lived in the second half of the eighteenth and first half of the nineteenth century who proved that there are equations of the fifth degree which cannot be solved algebraically. This was followed by important works of N.H. Abel and E. Galois. Today we have a theory, called the Galois theory (interested readers are referred for example to Hadlock (1978): however we add a warning that at this stage they might not be sufficiently equipped to handle it) by which it is possible to decide whether or not a given algebraic equation can be solved algebraically. The impossibility of solving an equation in radicals should not be confused with the existence of the solution. By the Fundamental Theorem of Algebra the solution always exists no matter whether or not a solution in radicals can be found. Solving an equation algebraically means finding the chain of Equations (A1)(Ak); that is, finding the roots of the equation by using very special equations, namely the binomial equations. From this point of view it would be surprising if every algebraic equation was solvable in radicals. Obviously it is easy to be wise after the event. Equations of degree up to four are always solvable in radicals, other equations might or might not be algebraically solvable. Maple is fantastic in this regard: if there is a solution in radicals, Maple will find it."
Quadratic equations
In this section we show that the 'usual' formula for the quadratic equation with which the readers are probably familiar from school is valid for a quadratic equation with complex coefficients. Rewriting the equation x 2
l1Within
+
reasonable limits of your computer capabilities.
Solving Equations
221
px + q = 0 in the form (x+ P )2
1 = D with D = p 2  4q we see immediately 2 4 that if D = 0 the equation has a double root 3 and if D > 0 then the 2 roots are
P 1 2 + 1? fand i a. 2 2 2
(8.12)
fiis understood in the sense of Definition 5.2, the square root of a positive real number. However, the Formulae (8.12) also hold if D is negative. The symbol then means Finally, if D is imaginary, we know that there are two complex numbers which squared equal to D. We denote one of them12 @ and the formulae (8.12) still hold. It is customary to choose fias p in Equation (7.17). For example, the roots of the equation x2 222 3 are z and 32, the roots of the equation x2  2x 1 2 2 are 2 z and 2. It is useful to know by heart the Formulae (8.12) for the solution but Maple provides the solution instantly, also for an equation of the form ax2 bz c = 0. However, for an equation with numerical coefficients the formulae for the solution should not be used but the command s o l v e applied directly to the given equation.
z m .
+ +
+
+
+ +
Cubic equations
Scipione del Ferro, who lived between 1496 and 1526, was the first to solve the general cubic equation. H. Cardano published formulae for the solution in his book Ars Mugnu in 1545 and the formulae for the solution are now often referred to as the Cardano formulae. Solving the cubic equation was undoubtedly a great intellectual achievement but the Cardano formulae themselves have little practical value. Often they give the solution in a rather complicated form and in case of an irreducible equation with real coefficients and three real roots the formulae involve cube roots of complex numbers and an effort to simplify them leads back to solving the equation. For this reason this case is called cusus irreducibilis (Latin for “the irreducible case”). Let us consider a cubic equation in more detail. An equation xn + a l X n  l
121t does not matter which.
+ + a,
* *
’
=0
(8.13)
222
Introduction to Mathematics with Maple
can be rewritten in the form
+ terms of degree less than n = 0. After the substitution y = x + a1/n, Equation (8.13) takes the form
(x In particular, for cubic equations, only the cubic x3 needs to be solved.
>
+):
n
+ px + q = 0
(8.14)
ceq:=xn3+p*x+q=0 ;
ceq := x3 + p z + q = o The Maple command solve can now provide the Cardano formulae, not in the original form, but Maple's form has an advantage. The Cardano formulae require the (possibly imaginary) cube root of 12p3 81q2; the correct choice of root has to be made. In the Maple formulae, this root can be chosen arbitrarily out of three possible values. However the same value must be kept throughout the formulae. The result is valid only for p # 0, but this is no obstacle since for p = 0 the equation is binomial.
+
>
Sol:=[solve(ceq,x)] ;
%1:= 108 q 12 J12p3 81 q2 For p = 3 and q = 4 the equation has an obvious root 1. But if we substitute p and q into the previous Maple result, we get:
>
+
+
p:=3;q:=4;Sol[l] ;
p := 3
q := 4
6 1 (432 + 12 d m ) ( ' I 3 ) 6 (432 + 12 di%%)(1/3)
Solving Equations
223
>
simplify(%) ;
(2 (2
+ &)(2/3)  1 + &)(1/3) + &)(2/3) +
(2
>
rationalize (%);
((2
+ &)(2/3) +
 1) (2
+ 6) + a)
>
simplify(%) ;
((2 &)(2/3)  1) (2 &)(2/3) (2 We still cannot obtain 1. Of course, our approach was clumsy. Applying solve directly to the equation with numeric coefficients immediately gives the correct answer.
>
solve (x^3+3*x4=0, x) ;
1 1 1 1 1,+Ifi,Ifi 2 2 2 2 If x1,x 2 , x 3 are roots of the equation then ( x 1  ~ 2 ) ~ (  ~ x3 1) 2 ( x 2  ~ 3 ) is called the discriminant of the equation. By the fundamental theorem on symmetric polynomials (see page 214) the discriminant can be expressed in terms of the coefficients. This is a laborious task and we let Maple do it for us. We use the command simplify in a new way, specifying eqs as a list of equations, which Muple should use in expressing SP in a new form, thus simplify (PS,eqs) ;
> >
~
eqns := [x [ll +x [2]+x [31=0,x [l] *x [2]+x [l] *X [3]+X [2]*X [3] =p,x[ll*x[2]*x[ 3]=qI;
eqns := [XI 2 2 x3 = 0 , xi x 2 xi ~3 > disc : = (x [l] x [2] ) ^2*(x [I] x
+ +
+
+ x3 == P,  2 1 w) 2* (x [21x [31)“2;
22 3 (2) 2 ~~ 3 ) ~
2 2 x 3 = Q]
disc := ( 2 1  ~ > simplify(disc,eqns) ;
2 (XI ) ~ ~
4p3  27q2
If p , Q are real then the equation has three real roots if and only if
the discriminant is positive, however the Cardano formulae then contains complex numbers (in a rather complicated way). But solve might also give the roots in a complex form. Consider the equation
224
Introduction to Mathematics with Maple
>
Eq:=x^33*~+1=0;
EQ := x3  3 z + 1 = 0
Looking at the graph or evaluating the discriminant 4p3  27q2 = 81 shows that this equation has three real roots. However
>
solve(Eq,x) ;
1 2 %1+ 2 4 I &i)(1/3) ’ (4 1 1 1 1  % 1 5 I & ( s %l4 (4 (4 4I fi)(1/3) 1 1 1 1  % 1  I %14 (4 4I4)(1/3) 2 (4
+
+
+
2
+ 4 I &)(1/3)
+
+ 4 I &)(ll3)1
2
One way to proceed is to use f solve. > fsolve(Eq,x) ;
 1.879385242
, .3472963553,1.532088886
If we want an exact solution we use the command evalc similarly as we did in Subsection 7.2.5 with the binomial equation. > L:=[solve(Eq)l :seq(simplify(evalc(L[i])) ,i=l. . 3 ) ;
2 2 2 2 2 2 cos( T ) , cos( n)  &sin( n),cos( T ) &isin( T ) 9 9 9 9 9 For an irreducible cubic equation it is almost invariably better to use f solve rather than solve, not only in the case of three real roots. On the other hand if the equation is reducible in the field implied by the coefficients, solve will find all roots in a neat form even if all roots are real.
+
>
solve (8*xn36*x+sqrt (2)) ;
4  l h + p , z l h  z & ,
1
1
1
1
1 ,Jz
The case of an irreducible cubic equation with three real roots is called cusus irreducibilis. The Cardano formulae (or solve) express the roots
Solving Equations
225
in a complicated form involving imaginary numbers and it can be shown that no amount of algebraic manipulation can simplify it to the standard form a bz. It is therefore very pleasing that with Muple you can obtain satisfactory result by using evalc together with solve. However Muple does not always do it ‘algebraically’: note that in the previous example trigonometric functions were used.
+
8 . 3 . 1
Equations of higher orders and fsolve
Although equations of fourth order can be solved algebraically and equations of order five or higher, generally speaking, cannot, we recommend, unless the equation has a strikingly simple form, to use f solve rather than solve. For comparison consider the next equation which can be solved algebraically (by solve).
>
solve (~5+x+i) ;
1 1 1
1
+I&,I&,%l2 2 2 2 6 3 (100 1 1 %l+ 12 (100
1
2
+
1 12 r n ) ( 1 / 3 )
+ 3’
2 
1
+
3 12 &%)(1/3)
(100
+ 12 4%)(1/3)
2 
3
1 %l+
12
(100
+
1 3 12 r n ) ( 1 / 3 )

(100
+ 12
3
rn)(1/3)
%1:= (100
>
12 &3)(1/3) f solve (x^5+x+i) ;
 ,7548776662
+
>
fsolve (xA5+x+i,x, complex) ;
. 7548776662,
.5000000000  .8660254038 I , .5000000000 .8660254038 I, .8774388331 .7448617666 I , .8774388331+ .7448617666 I
Although we recommend using f s o l v e for solving equations of degree higher than three, even with the aid of a computer, numerical calculations can have problems of their own. The next example illustrates this.
+
226
Introduction to Mathematics with Maple
>
Digits:=5;
Digits := 5
>
badpol:=expand( (x1.0001)n2*(x1.0002)n2*(xl.0003)n2) ;
>
badpol := x6  6.0012 x5 15.006 x4  20.012 x3 $15.012 x2  6.0060 x 1.0012 f solve (badpol) ;
+
+
l., l., l., l., l., 1.0012
The results says that the polynomial has a root of multiplicity 5 (which it does not) and the error for the last root is fairly large. This was caused by the number of digits used and rounding of all numbers to 5 digits. However similar phenomena can occur with any number of digits. It shows that the analysis of rounding errors can be important, but a lot of deep mathematics is needed for that. To see the reason why we obtained such an unsatisfactory result we graph the polynomial.
>
plot(badpol,x=0..2, scaling=CONSTRAINED);
iI
X
1. q relatively prime.OOO3)^2). Then Cseq(op(L) /K [il . + . > Digits:=30.00030000000000000000000000000 ~~ It should be realized that increasing the number of digits may not always be a viable option and.2 Test the numbers of the form (8.OOOl)~2*(xl. Digits := 30 > badpol:=expand((xl. could be meaningless.i=l .00020000000000000000000000000.Solving Equations 227 We see that the graph is almost indistinguishable from the x axis between 0.9) to find the rational roots of 2x6 x5 . > restart.3.72 6 and then compare your results with that obtained by the Maple command roots.00600058 x4 .00030000000000000000000000000.4 Use notation as i n the previous exercise with a = p / q and p .00600232043203860132 x + 1. a a rational root of S and S(x) = (x .6. [Hint: Compare coefficients!] 1 Exercise 8.001200580144019301320036 f solve (badpol) .6.l). Show that H has integer coefficients.1 Prove: A rational root of a polynomial with integer coefficients and leading coefficient 1 is an integer.3.3.nops(K) ) 1 is the list to be tested. + + > 1.4 and therefore to locate the zeros is difficult.a)H(x). Exercises ) Exercise 8.1 Exercise 8. [Hint: Make a list L of divisors of 2 and a list K of positive divisors of 6.20. when the numbers come from experimental data.6 and 1. 1.3 Let S be a polynomial with integer coefficients.0012 x5 15.00020000000000000000000000000. .9x4 . Prove that p .q and p q divide S( 1) and S (.00010000000000000000000000000.5x2 .3.0120034804320193 x2 . badpol := x6 .6x3 .] + + Exercise 8.1.1. By increasing the number of digits we obtain a perfect result.00010000000000000000000000000.012002320144 x3 15.OOO2)^2*(xl. 1. [Hint: q divides 1.
For instance.c ) .1 = 0. Consider the system (8. Exercise 8.15) (8.j ) 10/11 = 0 .4 n.b / ( l . + + Exercise 8.7 n%P.] + + 2x5 + x4 + + Exercise 8. So we The second case: The system has no solution.3.3. linalg and Linear Algebra. [Hint: q S ( 1 ) = ( q .16) Subtracting the first equation from the second gives ( c .228 Introduction to Mathematics with Maple respectively. It is easy to check that the found values are indeed a solution. So we have The first case: The system has a unique solution. 8.j ) + 1. y is uniquely determined as y = b / ( c . If c = 1 and b have # 0 then the second equations contradicts the first.y is a solution.!?. Increase the number of Digits to 30 and repeat.6 . Linear equations in several unknowns Linear equations in several unknowns are important because many problems in Science and engineering lead ultimately to solving large systems of linear equations.3.5 Use the previous exercise t o find the rational roots of S ( x ) = 6 x 5 l l x 4 .p ) H ( l ) . And if c = 1 and b = 0 then the second equation is superfluous: we can choose y arbitrarily and x = .l ) y = b and if c # 1 .(x. So we have .1 ) and x is then uniquely determined from the first equation x = . equations with over 40000 unknowns were encountered.6 2x3 + 2x2 + 1. This section is not meant even as a modest introduction to linear algebra or to these packages. The branch of mathematics which deals with linear equations is linear algebra and Maple has two useful packages.x3 5 x . Similarly for p + q. We start with a simple example which exhibits the behavior of linear systems. It is a simple guide on how to solve small systems with exact coefficients. in shipbuilding. Find the multiple roots of P ( x ) = x g 2x7 Use f s o l v e on the equations ( x .
{x.x+b*y+z=l}.w}) . b b i Y = . The first one would require a lot of work with pen and paper. caution is needed. > > restart.38 x = 1 21’ 21 + + + > so1ve(eqs. 7’ 7 Returning to our introductory example > eql :={x+y=O . 3 2 + 4 y + 5 2 6~ + V 2~ = 3 . 6 ~ y 2 2 3~ 4 v 5 w = .y. 2 6 ~~3 ~ + 4 ~ + 5 ~ + 6 w v= + 2. These three cases are typical for linear systems no matter how large the number of unknowns.{x. equations := {x + y + z = 1.v.x = 1 c’ l+c + Maple returned the answer only for the ‘general’ case.z.y}).y. 8 V = .10 z = .{x. eql : = { x + y = O . Linear sytems have another feature: the solution always lies in the same field as the coefficients.6 w = . This shows that even in simple examples when symbolic solution is obtained. We now look at some examples.Solving Equations 229 The third case: The system has infinitely m a n y solutions and some of the unknowns can be chosen arbitrarily. but using Maple the only work needed is to type the equations correctly. 4 ~ 5 y + 6 2 u 2 v 3 w = 4. equations:={x+y+z=l. eqs := 2 ~ + 3 ~ + 4 ~ + 5= ~1+ . x + b y + z = 1 .6 ) {X + + + + + + + + { + + + + + + + . 5 x 6 y z 2~ 3 v 4 w = 5 . x + y + a z = l } > solve(equations.x+y+a*z=l.32 u = 21’ 21 . .z}).u. 25 y = . x + c y = b } > solve(eqi.x+c*y=b} .
1)z = 0. However. 4 . Finally. if a = b = 1 any two unknowns can be chosen. If b = 1 and a # OI3 then z = 0.UI = 0. x z . + + + 2. (a . subtracting the first equation from the second and from the third leads to ( b. z = 0. In this example z cannot be chosen but z or y can. say x . x = y. 1 Use Maple to solve the systems (1) x . in this example.w = o . it did not give any hint as to which values of a. b are exceptional. y .z + w = 0 . 2 0.u + w = o .l ) y = 0. This illustrates two facts: Maple did not provide a complete solutions for all possible values of a and b and in the third case some unknowns can be chosen but the system might determine which unknowns cannot be chosen. y and the third one is then uniquely determined. (2) 2 .z = 0 . x . z = 1x.u+w=o. y . y = 0) Maple again provided the solution in the general case.u + 2) = 0.u = 0. = + + + + + + 13The case a = 1 and b # 0 is similar. ax by cz = d . .y. x . and y can be chosen arbitrarily.z + w = o . Unfortunately.y + v . a2x b2y c2z = d2. y u . (3) x y z = 1.2) = 0. y .230 Introduction to Mathematics with Maple (a: = 1 . Exercises Exercise 8 .
Indeed. Let f : n H 2n for n > 0. generalised to infinite sets. we then write A B. If the function f is a bijection of A onto B then it has an inverse and f1 is a bijection of B onto A .1 Equivalent sets Two finite sets A and B have the same number of elements if there exists a bijection of A onto B. so A B + B A . The required bijection is 2 We observe that the relation A B is reflexive.1 (a) The set  N and the set of even positive integers are equivalent. . and f : n H 2n 1 for n 5 0. Finally. Hence it is a bijection of Zonto N.. We can count the visitors in a soldout theatre by counting the seats. . . m[. We also discuss briefly the axiom of choice. 9. n H 2n is a bijection of N onto (2.1. which provides a bijection of A onto itself. X 1   231 . 6. Example 9. Definition 9. Reflexivity follows from the use of the function idA. I[10. Then f is obviously onto and it is easy t o check that it is onetoone. (b) Z N.1 A set A is said to be equivalent to a set B if there exists a bijection of A onto B. I+  + (c) 10. symmetric and transitive. The concept of equivalence for sets is an extension of the concept of two sets having the same number of elements.}. 4. i f f and g are bijections  .Chapter 9 Sets Revisited In this chapter we introduce the concept of equivalence for sets and study countable sets. and thus it is an equivalence relation.
. 3. . Warning: Some authors use the words ‘enumerable’ and ‘countable’ with slightly different meanings. 1.1). . if there exists a bijection of N onto A.1 The range of a sequence is countable. A set is called countable if it is finite or enumerable.10). Theorem 9.1 (b) we have N N N N 0. In view of the symmetry property we may put A B into words as ‘the sets A and B are equivalent’.. and if . which means that this sequence is a bijection of N onto Z. The theorem is obvious if r g u is finite. For example. n k have been defined let n k + 1 be the smallest integer n such that U n $ { un. unk }. so ( A B. n l . . . . this bijection is called an enumeration of A .. 2.13) and the well ordering principle (Theorem 4..that is.232 Introduction to Mathematics with Maple of B onto C and of A onto B . We list the members and cross out all repetitions when ‘moving along the sequence (9. . 722. 3. Note that ‘countable’ in mathematics differs from the meaning this word may have in other contexts.1. We are then left with either a finite sequence or we obtain an enumeration of rgu. 1. . 0 It is worth noting that in the course of this proof we have tacitly used the recursion theorem (Theorem 5. Also the word ‘denumerable’ is sometimes used instead of ‘countable’. . The set of even positive integers and Z are both enumerable by Example 9.2) from left to right’. Displaying the first few terms of a sequence often gives a good insight into the behaviour of the sequence. .B C) + A C. un. A set A is said to be enumerable if it is equivalent to N. The formal proof is as follows. Proof. then f 0 g is a bijection of A onto C. Let the sequence be n H un.1) Etom this it is clear that each n E Zappears in a uniquely determined place in (9. respectively. (9. Define n 1 = 1. Then the sequence k t+ U n k is an enumeration of rgu. for the sequence from Example 9. 2.
and hence B is countable by Theorem 9. 0 Proof. . = U N for n > N . In this case define a. . Then n I+ a. Theorem 9. a2.5 ~ N x N is countable. Now consider the identity function on B . If B = 0 or B = A there is nothing to prove. Proof.3. Clearly h : A I+ B is onto.2 sequence. and let h be the extension of this to A such that h ( ~= ) 6 for x E A \ B. 0 Theorem 9. The function f o u is a sequence and its range is B . If A is enumerable then A is the range of its enumeration. is a sequence and A is its range. This is rather intuitive: a subset of a countable set is countable. Hence B is countable by Theorem 9.4 If B c A and A is countable then so is B.Sets Revisited 233 Theorem 9. Otherwise choose some element b E B. If A is enumerable let u be the enumeration of A. . 0 Theorem 9.3 countable. I f A is countable and f : A H B is onto. A nonempty countable set is the range of some Proof.1. U N } for some N . idB. . then B is If A is finite then so is B. We can arrange the elements of N x N into an infinite table. . If A is finite then A = { a l .
. mE z}. Consequently u1 is an enumeration of N x N. k) H 2n1(2k1). By Theorem 9. However. Without loss of generality we can assume that each Ak is nonempty. Proof.3.6 u is onetoone and onto N. we give another proof. Since m ++ .7 The set Q is countable.5. Then n I+ ak(m)maps N onto u 00 Ak. 0 Theorem 9. the set Ak is countable by Theorem 9.9. Since Q = u Ak.1 this set is countable. For a reader who feels uncomfortable with this proof. Let u : (n. Theorem 9. Proof.2 each Ak is the range of some sequence.w : n H (m. Proof. 0 k=l Theorem 9. say ak: m + + a k ( m ) .maps k 00 m z 0 Ak. 1[ is not countable. because we have not explicitly defined the enumeration.8 10. we k=l see that Q is countable by Theorem 9.7 and Exercise 5. k ) . By Example 5. a formal proof based on the above ideas can also be given. Then u 00 Ak k=l A simple memory aid for Theorem 9. Let Ak be countable for every k E N.234 Introduction to Mathematics with Maple By following the arrows we can arrange the elements of N x N into a sequence and this sequence enumerates N x N.6 is: a countable union of countable sets is countable.6 is countable. and by Theorem 9. Let w be an enumeration of N x N. onto For k E N define Ak = (f.
6).8. as expounded in Example 5.. . . . (here n is an upper suffix. a2 = 0. . are marked. for example. I[ is countable.. .. a% # bi.4 a~ 4a 4~4 gt . ai2 .~ .1. 10. . = 1 if a: = 5 . not the index of some power). . then 10.ai1 . With each a. a4 = ~ .a1#2a3a4a5. .+ lom '  lo B 5 Bn+ 9 51 0 1" < Bn+. . . 1[is a range of some sequence a : n H a. a ~ a ~ & a . 2 2 2 2 2 . The proof uses decimal fractions for real numbers. a . # b2. 5 B . Proof. we associate its decimal fraction 0.. a: # b3. +&}. and so the decimal fraction . For m > n we b7n . Exercise 9. We now define a sequence b.aat .. . This process is illustrated in the following diagram where a+ # b l . = 5 if a: # 5 and b.#1a2a3a4a5.. showing the reason for the adjective "diagonal". . Assume 10. 1 1 1 1 1 a1 = 0.10 b2 b3 ++lo2 lo3 + .~. and is often used on other occasions (see. . . a3 = O .a: a ~.Sets Revisited 235 The method of proof used here is known as Cantor's diagonal process (or Cantor's diagonalization process). . . . B bl . . by b.1 From It follows from these inequalities that 10" the definition of B it is obvious that B. a 4 . Let B have = SUP { Bn. .aya. .a:. a. .
0 The phrase “continuing with this process indefinitely” in the above ‘proof’ looks suspect. is an enumeration of B. or. The decimal fraction for B differs in at least one place from the decimal fraction for any a. and choose a1 E A . a3.8. Also the set of irrationals. Arguments such as those used in the above ‘proof’ had been freely used in mathematics until it was realised that their acceptance hinges on the validity of the following statementwhich has come to be known as the axiom of choice. for every n E N. . . that the natural numbers are equivalent to any of the smallest infinite sets. . I[. a2. Theorem 9. It is also easy to see that 0 < B < 1.236 Introduction to Mathematics with Maple associated with B is 0.8 is that R is not countable. contradicting the assumption that 0 the range of a was 30. and since the theorem we are trying to prove refers to “every infinite set” there seems little hope of devising such functions. l[. . b[ onto 10. b[ is not countable either because b : x I+ is a bijection of ba ] a .} c A and n H a.. . Consequently B # a. The set A \ { a l } is nonempty (it is even infinite) and we can choose a2 E A \ { a l } . I/ Theorem 9. in other words.2 or gn as in Theorem 5.13 at our disposal.. [First attempt] Let A be infinite.blb2b3. The next theorem describes enumerable sets as the smallest infinite sets. An obvious consequence of Theorem 9. l Proof. Axiom of Choice If S is a family of nonempty sets then there exists a function G such that G ( S )E S for every S E S. Continuing with this process indefinitely we select Then B = { a l . R \ Q is not countable: if it were then R would be countable as the union of R \ Q and Q. Any xa interval ] a .8 shows that there is at least one infinite set which is ‘larger’ than the set of natural numbers. for the definition of decimal fraction see Example 5.. Can the proof be made impeccable by replacing the offending phrase by a reference to one of the recursion theorems? The trouble is that we have no function g as in Theorem 5. that is.9 Every infinite set contains an enumerable part.
each of the sets S in the family may also be infinite. Proof. However. we can provide an impeccable proof.9: this time. the difficulty with its acceptance in mathematics lies in its great generality: it asserts the existence of G. S and g in Theorem 5. furthermore. To see this we define f . We now return to the proof of Theorem 9. From this point of view the axiom of choice is very plausible.1 below. Define a bijection for each of the pairs o f sets given . for n E N. using the axiom of choice.Sets Revisited 237 The great German mathematician David Hilbert compared the family S and the function G to a group of societies each selecting its president. The sequences n H An and n an are uniquely determined by the recursion theorem. We now define and inductively.1. for any infinite family of sets S.13 as follows: Now a. and. From the axiom of choice there exists a function G which associates with every nonempty subset B c A an element G ( B )E B . becomes the first element in the couple f ( n )and n enumeration of a part of A . Exercises an is an 0 3 Exercise 9.
then so is u IEL 0 Exercise 9. A S . are countable. 0 Exercise 9. without using the axiom of choice.81. (2) {x: x 2 + p x + q = o . .1.1. 13. (2) (4) (6) (8) [o.1. then the set A1 x A2 x A3 x .9 Prove. 1 2 1 2 3 1 2 3 4 1 2 3 4 contains all rationals between 0 and 1. Prove this slight generalisation o f Theorem 9. 1 3 . I[ such that their decimal fractions do not contain the digit 7.4 I f n E N and A l .8 0 Exercise 9.238 Introduction to Mathematics with Maple (1) (3) N. 30.No (0. q E Q } . A. 1) as their ranges.1. Q x Q. Use this sequence t o define a bijection Q onto N.4 R. [d.8 to prove the following theorem.6 Use the diagonalization process from Theorem 9. N.x A. (1) { a + b&. Then S is not countable. [o.2 Exercise 9. (5) (7) zx z. [Hint: Most of the work has already been done in the proof of Theorem 9. Z.3 Prove that ifA is not countable and B is countable then Show that the following sets are countable: A \ B is not countable. 1 nEN } . x irrational. . Show that S is not countable. 0 < x < 1) has an enumerable part.1.7 0 Exercise 9.1. N. a E Q.5. I f L is countable and for every 1 E L the set A1 is A 1 . 2’5’3’ 3’ 3’ 4’4’ 4’4’5’5’5’5’. is also countable. b E Q}. @ Exercise 9.1. p E Q . 0 Exercise 9. [a. ] The range of the sequence 1. . Z x Z. [3. Let S be the set o f real numbers in 10. Prove this.1.. 0 Exercise 9. A2.5 countable. . . that the set B = {x. el.71. bl. Let S be the set o f all sequences having (0. I[.
3) (10. for (10.1)' (10.1) (10.1 indicate the number of throws. (10. The columns headed n in Table 10.4) it is 0. 10. The concept of a limit is central to subsequent chapters of this book.Chapter 10 Limits of Sequences In the early sections of this chapter we introduce the idea of the limit of a sequence and prove basic theorems on limits.2) (10.4) we see that as n becomes larger the terms of each sequence approach a certain number: for Sequences (10. In Sequences ( l O . (10.3) and (10.1 The concept of a limit If we observe the behaviour of several sequences.3) it is 1. The following table is a record of an experiment: casting of two dice and noting as a success whenever the sum is 7. l ) . and the other columns gives the values of s / n where s is the number of 239 . for example (10.4) there is a clear pattern in this approach. The later sections are devoted to the general principle of convergence and more advanced concepts of limits superior and limits inferior of a sequence. In the next example the terms of the sequence approach a certain number quite irregularly.2) and (10.2).
Table 10.166553834 . .166553001 . In Sequences (10.16666.166552003 There is no discernible pattern this time but (assuming the dice are unbiased) the terms of the sequence of values of s / n approach the number .166552167 .166552169 .166553169 .166551836 . . If we required higher accuracy we would have to keep casting the dice longer but the same phenomenon would occur.1665525 . This leads us to: Tentative temporary definition.166552336 .16655167 . all subsequent terms of the sequence are approximately equal t o . for any prescribed accuracy.240 Introduction to Mathematics with Maple successes in n trials. What is characteristic is that after one million terms.166553 . within three decimal places.166553334 .166553335 . after casting the dice sufficiently often the terms of the sequence will be approximately equal t o 1/6 within the required accuracy. In the example of casting two dice for accuracy to three decimal places the number N was one million. The sequence n I+ zn has a limit 1 if.166552669 .166551337 .166552668 .166553668 .166552502 . there exists N such that for n > N .166552334 .166553002 .1666.166552835 ..166553168 .1) and (10.166553502 1000016 1000017 1000018 1000019 1000020 1000021 1000022 1000023 1000024 1000025 1000026 1000027 1000028 1000029 1000030 .166552833 .166553001 .166552835 .166551503 .2) the limit is clearly zero and for accuracy of k decimal places the number N would be . the terms x .16655117 .166553501 . = 1/6.1 Results from throwing two dice n n 1000000 1000001 1000002 1000003 1000004 1000005 1000006 1000007 1000008 1000009 1000010 1000011 1000012 1000013 1000014 1000015 ..166552003 . are approximately equal t o 1.166552667 .
the tentative definition requires existence of some number N (with the required property) and soon as some N is found.1 The sequence n H xn = c has a limit c.3) the limit is again zero but it is not so obvious  what N should be.1Ok. The arrow in the notation xn + l indicating that I is the limit of the sequence x must not be confused with the arrow indicating the functional dependence as in n H x n .‘ ) n 2n will be approximately equal to zero to k decimal places if n > 2. A sequence which has a limit is called convergent.1 is often referred to as the EN definition of the limit.1 (Limit of a sequence) A sequence n H xn E CC is said t o have a limit I if for every positive E there exists a real number N such that whenever n > N . whereas xn + l states that I is the limit of the sequence n H 2.( . . and say that xn is approximately equal to I within E if / \ Definition 10. and 0 = lxn . Perhaps this N is unnecessarily large but that is of no importance. The phrases “the sequence converges to 1” or the “sequence approaches I” are used with the same meaning as “the sequence has a limit 1 ” . It is simpler to use a positive number.. If the sequence n H xn has a limit I we write n+m lim = I or xn + I.Limits of Sequences 241 2 lok. Since Ixn cc( = 0 for every n we can choose any number for N . Definition 10. Hence n H I defines a sequence with every term equal to I . the job is done. for instance N = 1. Since that 1 ’ + (1)”+’ 5 2n 1 < 1 . Example 10. It is undesirable and often would be inconvenient to use a decimal fraction to measure accuracy.I1 whenever n <& >N.10kwill do for N . say E . for n > 3 it is clear I. For Sequence (10. A sequence which is not convergent is called divergent. Hence 2.
If n > N then 1 1 1 0 < < . It is. Choose E = $1L . (B) If n H xn has a limit l then there are many N such that Ixn . however.< E . (C) The meaning of Definition 10.1 is not changed if. for example. ~~ 11 'I Assume.Generally. that the sequence > 0.L I < E whenever n > N1 and Ixn . Then by the definition of the limit there exist numbers N1. Let N = Max(N1. In particular 00 does not denote a number. Obviously this might be difficult or not feasible. If we can guess what the limit should be then we can use the definition to prove that our guess was correct. It provides no assistance with finding the value of 1. n00 It simply denotes the limit of the sequence x and does not introduce the symbol 00 in itself. possible to prove powerful theorems . If n > N then x has two distinct limits l . n+l N 1200 n 1 1 .2 1 The sequence n + 0. n+l is not bounded from above by Theorem H . or show that our guess was incorrect. N2 such that lxn . N + 13 and 2N are also suitable: any number greater than N is suitable. t h a t is + 1 I' Theorem 10. (D) The symbol lim xn is to be understood as an indivisible quantity.lI < E whenever n > N2. then. L . contrary to what we want to prove. 0 We now make some important comments concerning the definition of a limit. The set N n+l 4. As soon as we have a suitable N . instead of permitting N to be any real number we restrict it to N E N.1 Every sequence has a t most one limit. (Definition 10.N2).8 and consequently there exists N E N such that I/E < N . with IL . the larger the value of N . This proves that lim . (E) The definition makes it possible to check whether or not a number 1 is the limit of a specified sequence.lI Proof.Z I . This is manifestly wrong and we reached a contradiction.lI < E for n > N . N + 1. Let E > 0. the smaller the value of E.242 Introduction t o Mathematics with Maple Example 10.1 ( A ) The number N from the definition depends on E : this is sometimes emphasised by writing N ( E ) .0.IS convergent t o zero.
n+m k+m mmo Sometimes lim xn is abbreviated to limx. 1 E [ (see Figure 10.E .1 Limit of a real sequence For a complex sequence x .Limits of Sequences 243 concerning limits from the definition. see Figure 10.11 < E is equivalent by Theorem 4. (F) Consider now a real sequence x . where the terms of the sequence starting with 2 5 lie within the circle.4 to all terms of the sequence with an index greater than N lie in the interval 3 1 . the inequality lxn .1). a sequence whose terms are real numbers Xn. in the complex plane. Thus lim Xn = lim xk = lim Xm.11 < E means that the distance between xn and I is less than E . 10. Since the inequality Ixn . confusion can arise. a sequence xn E C for all n. that is. xn lies in the circle centered at I and of radius E . If several letters are involved in the formula for xn. It is not clear what is the mean ing of lim ( n n+m lim i). Some such theorems are given in the next section.. 1 E [ .E . and then to use these theorems to find limits. + + Fig. We shall not do n+m that in this book. It goes without saying that the letter n in the symbol lim Xn = I n+m can be replaced by any other letter without altering the meaning. Geometrically this means that.2. If the sequence has a limit 1 the points xn in the complex plane cluster near 1 in the circle centered at 1 and of radius E . ( 5 + i) + + n+2 =1 It is left as an exercise to show that n 2 (J> The notation xn ' + 1 has the same disadvantage " as abbreviatinc u . This means that outside this interval can lie only terms of the sequence with n 5 N . that is. If the terms xn are graphed as points on the number line then they cluster near 1 in the interval 11 . which are finitely many.
(10. In particular.6) shows that the sequence b is obtained from a by omitting the first term and renumbering.244 Introduction to Mathematics with Maple Complex plane 0 Fig. . a4. we have that Xn + 0 if and only if Ixnl + 0. a3.I1 + 0. . as for example in l / n + 0.5) and bl = a2. However. bn = an+l. Let a be a sequence of complex numbers and members of both sequences Listing a few (10.. for I = 0. we shall use this notation for its convenience. 10. . (K) It is easy to see by comparing the definitions that xn +I if and only if Ixn . if no confusion can arise.2 Limit of a complex sequence n+co lim xn to lim xn.
0 for n The sequence b is obtained from a by omitting the first term. Since bn = an+l. Theorem 10. by using Theorem 10. yn + 1 and there is an N1 such that Yn E R for all n E N.I1 < E for n > N. Since lirn = 0. For instance. < l + & . n+m an exists if and only Since the wording of the theorem includes 'if and only if' the proof comes in two parts.3) is convergent t o 1. Consider that .5) = 0 although the function a is not defined for n = 5 . Because of this we extend convergence. it follows that Jbn . where A is a finite set and we also call such functions sequences.Limits of Sequences 245 Theorem 10.2 repeatedly. if a n = l / ( n .1. 11. If lim an = 1 then for every E > 0 there exists N such that Ian .11 < E for n > N 1 + 1. that is. E R and X n + I .2 With the notation above lim n+m if lirn bn exists and then both limits are equal. n+m nmo Example 10.2.8) < x.3 n 1200 The Sequence (10.3 Assume that 5.11 < E Proof. for every E > 0 there exists N such that 0 < n > N . divergence and limits to functions defined not on all of N but only on N \ A . that omission or change of finite number of terms affects neither the existence nor the value of the limit.I1 < E n+m for n > N1. If lim bn = 1 then for every E > 0 there exists Nl such thatlb. . and N 3 such that (10.5 ) we would say that lim l/(n. n 4. If xn L zn L Yn for n (10.3 we used our knowledge of one sequence to establish the limit of another sequence. Consequently n 1 1 < E for n+l + for n >N.1 InP1 'I= n+m by Example 10. 1 1 lim .7) > N1 then Zn is convergent to 1. > N. This means that Ian . The same idea is contained in the next theorem. For every E > 0 there exist 1& N 2 Proof. It is clear. I. In Example 10.
and as xn and gn approach I . so does zn  Example 10. 0 We shall call Theorem 10. nk 1 1 n The Squeeze Principle dealt with real sequences. N3).ZI < E. It has.8) and (10.5 that for every k E Since lim .= 0. (10. Consequently and this implies IZn .3 the Squeeze Principle.9) hold. Let N = Max(N1. We have and Now we set Application of the Squeeze Principle completes the proof. .4 n To prove it we use the Squeeze principle but estimate dm k=l 1 first. If n > N then all Inequalities (10. an important consequence for complex sequences. Nz.noo lim . Example 10.7). however.246 Introduction to Mathematics with Maple for n > N2 and for n > N3. This is quite intuitive: zn is squeezed between xn and Yn.= 0 it follows from the Squeeze Principle nmo N.
Assume (10.11 < Yn for n E N and lim Yn = O then lirn Zn = 1. Yn = %Zn.pl < e/2 and gn . Hence ~ l A formal proof is similar t o the one given in Example 10. Since (10.we can.11 < lim IZn .Limits of Sequences 247 Corollary 10.Iz2k .10) if and only if nco lim lim xn yn =p = 4.1 1 = . The proof of Equation (10. choose yk = .4 Assume that z is a complex sequence and 1 a complex number with Xn = %Zn. respectively. I. . We prove it. Since 0 5 IZn .2.l .3.10). z 2 k = for k E N. p = % I and q = 9 1.11 we have 2‘ 2 k ) and then I z ~. 4’2’. n+w n4m n+m Proof.1 that Z k + 1. (10. 11. in applying the corollary.3.1 again and we have Equation (10.11).13) and the right hand side of (10. Then n+m lim zn =I (10.’ 2k 1 3 3 7 4 1 5 5 $7 16’p .13) approaches zero by (K) we can apply Corollary 10.1 1 5 yk for every k E Obviously y k + 0 and it follows from Corollary 10.4)< e/2 for n > N1 and n > N 2 . Theorem 10. ’ k N.3. N 2 such that lxn .12) n+co Proof.(since by Exercise 5.1 = 1 terms of the sequence are 1 Ic I5.11) (10.12) is similar.1 Let z be a complex sequence and y a real sequence.6 Let z 2 k .1.11 = 0 and by (K) lim n+m Yn zn the Squeeze Principle implies that = 1.1 1= 1 1 2 .’ k It is not difficult to see that zk + 1. Since l z 2 k . For e > 0 there are numbers N 1 .The first few k 2’2. 0 Example 10. If IZn .
that ( .21 . 1 . For instance the sequence n I+ (l)nis divergent.1. Exercises Exercise 10. .ZI < 1 for all n > N . 3 . n2 2 (3) lim = 0.248 Introduction to Mathematics with Maple This proves (10. Find the limit. has no limit.1. Consequently 2= I(ly .ZI < 2...21 < E for n > N.1 C Prove that (1) lim . Then for E = 1 there exists a number N such that l(l)n. fi nm n 3 +n + [Hint: For (1) mimic the proof of Example 10. [Hint: Simplify Ian . Choose a natural number k with k > N.= O for every c E C. This contradiction completes the proof that the sequence has no limit. Assume. For E > 0 find N such that lan .10).4 Show that the sequence with terms 1 . 1 . for an indirect proof.2.] Let ~ ~ = 1 and ~2~ = 4n+1 2n . Write down the first five terms of the sequence.z + z . Find N1 such that lan .ZI < 1. 0 Not every sequence has a limit. 2 . nE N } is not bounded for ( 2 ) .(1)"+'1 5 I(ly . 3 .3 Let a1 = 100 and an++'= (a1. Exercise 10.2 4n+ 1 Write down 2n 1 the first eight members of the sequence a.n)an for n E N.Then I(l)'"  ZI < 1 and I(l>"+'.1.. For (3) show that < . n+oo n 1 ( 2 ) lim n+W = O.] Exercise 10.1. + Exercise 10.and n apply the Squeeze principle. 2 . 3 .l ) n + 1. Use that the set n2+2 2 n31n { fi. 2 .ZI + I(l)"+' .21 < l o 3 for n > N l .
I [ < E .R. for n > N and some I E . Is the converse true? Let a E (C have the property that for every E > 0 the is true.ZI 5 E.I1 < 1. for .6 @ Exercise 10. [Hint: For an indirect proof @ Exercise 10. Prove that a = 0.5 unchanged if Prove that the meaning of Definition 10. . The function f is said to be bounded if it is bounded ~ ~~ This means that a complex sequence x is bounded if and only if there exists a number K such that Theorem 10. (2) the inequality l x .2 Basic theorems Definition 10.1. . It follows that = 1 there exists N such that Ixn . For > N .] Prove: If 1 E R and for every k E N there exists N such that LIOkxnJ= l l O k l ] for n > N then x . bounded below.1.1. (3) for some positive EO the requirement that E is positive is replaced by the requirement that 0 < E 5 E O .Limits of Sequences 249 @ Exercise 10.8 Let x be a sequence with the property that for every E > 0 there exists a number N such that Ix. This shows the importance of the order in which E and N appear in Definition 10.] 10.ZI < E is replaced by l x . [Hint: Use the previous exercise. bounded above.7 inequality la1 < E choose 2~ = la1 . . 0 Exercise 10. Prove that xn = I for n > N .5 Every convergent sequence is bounded. E n Proof.1 remains (1) the inequality n > N is replaced by n 2 N .1. + 1.2 bounded above on a s e t S if the set f ( s ) is bounded. respectively.1. Let xn + 1.
By taking c = 1 we obtain xn + I and then it follows from (ii)that xn .14) and S2 is bounded because it is finite. n 5 N } . = 1 it follows from (iv) t h a t l / y n +l/m.I [ + 0 and since follows from Corollary 10. (ii) For 7 > 0 there exists Nl and N2 such that IIXnI . Remark 10.+ IYn .1 that IZnl + 111.17) that E This proves (ii). Then for n > N we have from (10. It follows similarly from (iii) and (iv) t h a t y i k + mIC.250 Introduction to Mathematics with Maple for n > N . for n Max(N1.+ Yn xn 1 . (i) by (K) of Section 10. rgx = S1 U S2 is bounded.and for this particular r] find iV1 and 2 N2 as above.m. Consequently. By taking Y n = xn it follows from (iii) t h a t x: + Z 2 and by simple induction that xk + Z k for k E N. For x.1111 5 IXn ZI it and for n > N1 and n > N2.3. Consequently.provided m # 0. respectively. 0 . The set S 1 is bounded by (10. m An important special case of (iii) is Y n = c for every n E N. 0 (iv) if m # 0 then .1 Proof. choose 7 = .(1 + m)l I Ixn .1 I X n . N2) l(x. >N = + Y n ) .17) Now take > 0. Let S1 = { x .Y n + 1 . + CZ. This yields cx. n > N } and S2 = { x .m l < 27 & (10.
It follows that 1 2 . If n > N = Max( Nl . In particular Yn # 0. N2) then In view of Remark 10.16) we finish with (10.6.17). 1 1 (iv) It suffices to prove that .16) holds for n > N . M (independent of n) we shall regard the proof We now return to the rest of the proof of Theorem 10.+ .16) hold for n > N1 and n > N2. we have obviously that and it follows that For E > 0 we can find N1 and N2 such that (10.2 We now pause and reflect on the proof completed thus far. Let us Yn 1 choose E = . respectively. 2 Consequently for n > N . Starting with (10. that is there is a K such that Ixnl 5 K for all n E N. In a situation like this when we succeed in showing that there is N such that for al! n > N and some fixed of Xn + L as completed.15) and (10.then there exists N such that (10. This inequality is almost the one required by the definition of lim (2. Now. The fact that we used 7 instead of E is irrelevant.15) and (10. Only a simple adjustment was then needed to remove the factor 2.5. it was an arbitrary positive number. (iii) Since xn + I the sequence x is bounded by Theorem 10.2 this proves (iii).because as soon as this is esYn tablished (iv) follows by applying (iii) to Y n replaced by m . Proof.Limits of Sequences 251 Remark 10. except nmo + + that we have there 27 instead of 7. Yn) = I m.
7 I + .2n < .* YiTCFL .1 . then we estimate 1 Now < 1.18) above yields lim noo  1 . 1+n Using the Squeeze principle and Equation (10.2n First we realize that 1.2 l3 4 + . 1 m 0 Theorem 10..4 + .3.1. .8 We wish t o find lim nmc 1 . .n + 1 lim noo 2$ + $ limnjOO (2  + $) 2' Example 10.n n+l' (10. Now we have By Remark 10..2 I3 .+ Yn 1 . .252 Introduction to Mathematics with Maple for n >N.2 +3 4 + . Example 10.2 + 3 .2n ViTZi ~~ ~ = 1. Yn E R and 1 <m Xn < Yn.1 and by Corollary 10. for all n >N..2n $7XG .6 is important in itself for the further development of the theory but it is also a powerful tool for calculating limits.7 If X n I and Yn then there exists N such that = + +m with X n . Theorem 10.1 and (K) of Section 10.4 ? = lim 7200 n2+4n7 7 1imn+m(l+?$)  1  2n2 ..18) n lim = lim nmn+l noo 1 . = n.
we even have yn < xn for all n E N. The next two examples illustrate the role of strict and nonstrict inequalities in Theorems 10.1 for n > N . If ym + m > 0 then there exists N such that Y n > 0 Proof. Corollary 10. Let E. Y n E Iw and if xn IYn Proof. respectively.Limits of Sequences 253 Proof.contrary to our assumption.7. However it is not true t h a t xn 5 Y n for sufficiently large n. = m1 .1 and X n 5 m then 1 5 m.7 and 10.9 If X n = 0 and Y n = . + 1. It suffices to take x .7 is: Inequality between limits implies a similar inequality ultimately for the terms of the sequences. N2) we obtain 0 A memory aid for Theorem 10. .7. Example 10. For n from these inequalities > N = Max(N1. 0 Taking Y n =m yields If X n +1 Corollary 10.8.8 for n E N then 1 yn + m with Xn.7. = 1 = 0 for every n in Theorem 10. If m < 1 then by Theorem 10. then there exists iV1 and N2 such that 2 and l+m m&=< Y n < m+E. If xn I m. 2 for n > Nl and n > N2. 0 Theorem 10. It suffices to take Y n = m for every n in Theorem 10.8. 1 < m then there exists N such that Proof.l / n then 1 5 m since 1 = m = 0.7 (with the roles of x n and Y n interchanged) Y n < xn for some n E N. If xn + 0 X n Corollary 10.7.2 < m for n > N .
loon3 17' _ (3n 2)2 .3 Prove that [Hint: Use the inequalities LkxJ 5 kx.2.1 Find the following limits ( 2 ) lim 1000n3. ] yn + Exercise 10.] < [kxj and the Squeeze @ Exercise 10. B > 0 and q E N prove k=O . k x .5 Give examples of divergent sequences x and y such that (i) x y is convergent. + Exercise 10. Prove this.1 principle.4 If X n + Xn+1 I # 0 then + 1. Y n 4 Q. and (ii) x y is convergent.254 Introduction to Mathematics with Maple Example 10. (ii) Xn+l +c#1.2.10 If xn = . [Hint: Consider LncJ/n and LncJZ1/ ( n f i ) .2.2. Exercises Exercise 10. Also give X n examples of convergent sequences n I+ xn such that (1) X n Xn+1 diverges.2n 1 n* 1 10n 5. < yn for all n E N.2 C.l / n and Y n = l / n then xn However the inequality I < m is false since I = m = 0.(n I ) ~ (3) lim 7200 (3n 2)3 (n 1)3 1+2+3++n (4) lim 2 noo n+2 + ( + + + + + + '> @ Exercise 10.2.2.6 For A > 0. X n Exercise 10. Prove: For every c E Iw there are sequences xn + c with xn E Q.
d z ) .1 or the identity Max(a. @ Exercise 10.10 Prove: If x.n=inf inity). This can be done . Simply.3 Limits of sequences in Maple Maple can find limits of sequences quickly and efficiently even if the limit is hard to find otherwise.. [Hint: use either Definition 10.7 show that Using the previous exercise and the Squeeze principle 5.2. + m then Max(z. Use the previous exercise to prove that x : Find the following limits @ Exercise 10. instead of writing lim f(n) we issue n+w the command limit (f (n) .2. y .] Exercise 10.Limits of Sequences 255 @ Exercise 10.2. m) and Min(z. we will show how to use Maple to solve them in the next Section. b) = i(a+b+labl). m).2.. + 'I if 1 and r E Q. (a) lim n+w f7 8n21 ' [Hint: If you are finding some exercises difficult. ) + Max(l. yn) + Min(1.] 10.9 (1) n+m lim (diFTi . y are real sequences with zn + 1 and y . On most occasions this would be sufficient but the command is incomplete: we have to tell Maple that n (or m or whatever notation we use) is a positive integer.8 x. + 1 > 0 =+ 6+ 4 q 4 for q E N.
. We illustrate both possibilities.3). 1  > 3 limit (sqrt(nn2+2*n+3)sqrt (nn2+ni). For calculating just one limit it is more convenient to use assuming.m=infinity) assuming m::posint. 0 > limit ( (k2)* (k+surd( lk^3.‘I2 and lim ~ noo 3n2 7100 ( d   d m ) .k=inf inity) assuming k::posint.256 Introduction to Mathematics with Maple either with assume or assuming.9 part (ix).n=infinity). We wish to calculate lim (n 4. l. > assume(n: :posint) .l This result does not look right.3) ) ..  1 3 Neglecting the assuming part of the command can lead to incorrect results.n=infinity).n=infinity) . Now we calculate the limit from Exercise 10. > limit ( ( (n+2)2)/ (3*n2). For instance lim sin(n.2. ~ ~~ > limit(m+surd(lmn3. The command to use for odd roots is surd. In the next example we need a cubic root. We have to accept that we ask a wrong question and got a wrong answer.n) = 0 but n+oo ~ ~ ~~ ~~~ > limit (sin(n*Pi) . since using fractional exponents can bring in complex roots which are undesirable.  1 2 n := n It is a good habit to clear the assigned variable if it is not needed any more.
Sequences are functions defined on N.1 Use Maple to evaluate the limits in Exercise 10. that is. + + ncc 10.9 Every bounded monotonic sequence is convergent. A function which is either increasing or decreasing is called monotonic.3 applies to sequences as well.3.3 tion f is said to be increasing on a set S if for every 51. = and yn = q n 3 + 6 n 2 f 8 n 2n 1 q n 3 n2 . Monotonic sequences play an important role in the theory of limits. < Theorem 10. decreasing or strictly decreasing sequence. Find lim MaX(X.3 Let x.19) by 21 < 5 2 =+ f (21) < f ( 2 2 ) we obtain the definition of a strictly increasing f (on S ) . oo[. the characteristic function of the interval 10. x2 E S 21 5 x 2 =+ f ( X 1 ) I f(x2). the function x H x3 is strictly increasing and l p .2. Consequently Definition 10.+1 for every n E N. It should be clear what is meant by a strictly increasing.4 Monotonic sequences We shall assume in this section that all functions (and sequences) are real valued Definition 10.2 What happens if you use Maple to solve Exercise 10.9.. (10. oo[.3.9. Clearly a sequence x is increasing if and only if x.3. .3. By replacing the implication (10. [Hint: Maple cannot solve this problem directly.19) f is said to be increasing if it is increasing on dom f .] 3n Exercise 10.Limits of Sequences 257 Exercises Exercise 10. and by reversing the inequality f ( X I ) f (x2) or f (x1) < f(x2) we obtain the definition < For example the function x H x2 is strictly increasing on the interval 10. x. @ Exercise 10. yn).2. is increasing but not strictly increasing.
Obviously xn+1 = I b l X n . Example 10. hence the sequence n I+ X n is decreasing.1 0.E < X N 5 1.4 For an increasing sequence n H xn with limit 1 we shall write xn 7 1 .E < 1 there exists. and similarly if xn decreases with limit 1 we shall write xn 1 1. If it were then lcnl 1 1 and the same argument as in Example 10. The case a = 1 is obvious.1 it follows that bn + 0.9 and 10. Since xn 2 X N for n > N we have for n >N. which is clearly impossible. a strictly decreasing sequence is decreasing. By (K) of Section 10.10 sound theoretical their application often enables us to find limits. We define xn = lbl".12 We show that for a > 0. Since 1 . Consequently 1 = 0 and lbln .11 would show that cn + 0.5 If IcI > 1 then n H cn cannot be bounded. Although Theorems 10. Denote by 1 the supremum on the right hand side of (10. . it is sufficient to prove Theorem 10. Let E > 0.11 We show for b E CC that bn + 0 if lbl < 1. is decreasing and bounded below then 0 Remark 10. by definition of the least upper bound.3 Remark 10.10 it is convergent and If n H xn is increasing and bounded above then Proof. Remark 10.258 Introduction t o Mathematics with Maple Since a strictly increasing sequence is increasing. Passing t o the limit in the previous equation gives 1 = IbJl. Example 10. say 1.20). it is also bounded below by zero so it has a limit. If n H x . and x is increasing if x is decreasing. a number N such that 1 .
.. Example 10.33) and by what we have fi fi fi. Since ( l)n+l = ( 1)" we obtain by passage to the limit that I = I. We first prove that the sequence is bounded.+ . By the binomial theorem (1 + . This conclusion is false since the sequence in question has no limit..(lG) (1(1 < 1+ 1+ ..22) We next prove that the sequence is increasing.Limits of Sequences 259 Consider first a > 1. and therefore IcI 5 1. that is I = 0.).10) (10. This means that the sequence n + cn is bounded so clearly c = 1. Hence we have IcI 5 15 c 5 If 0 < a < 1 then 1/u = b > 1 .1 we have 1 .. By Equation (5. proving the b v The use of the existence assertion from Theorem 10.23) .)n = 2 (.9 or Theorem 10. consequently 1 5 cn 5 a.12 is essential.. +  2! n! 1 1 <1+1+~+ 22 +"'+2"1 < 3 (10. = just proved result. (l. By Remark 10.) n! > 2 we have n! > 2n1 i)+ $ A) :) +. In the arithmeticgeometric mean inequality (Theorem 5.13 The sequence n H (1 + i)n is convergent and its limit is denoted by e.+ For n and consequently 1 2! 1 3! * * 1 +n! (10.10 in Examples 10. Consider the sequence n H (l)n and denote its limit I . Hence it is convergent. fi j 1 .)$ k=Q = l+nL n + L(12! +L(l.+ 1. This number plays an important part in mathematical analysis.11 and 10. Then 1 5 c 5 and from below by 1.21) ' (1 +. J 1 1 < 1 + 1+ + . The sequence n + is decreasing and bounded 1 c.
yn + . say. ) .+ ..8.22) it is also bounded. Thus it has a limit. n!n+l I= n!n 1 1 1 1 .. Secondly. n E N.9) we have r1 21 r(r . + 1+ + + . Consequently e = E and hence =e. 2! 3! n! The sequence Yn (1  A) (1  .1) n for r...8 again we have e 1 n+m 1 1 + ). This example enables us to calculate e with high accuracy. Hence since r2 c n 1 <3 by (10. Firstly we have Yn < e. Firstly.. Y n E. we have Ym = yn 1 1 ++. let us now estimate 1 1 1 + 1+ + .14 Let yn = 1 is obviously increasing and by (10. Using Bernoulli’s inequality (Theorem 5. We are going to show that E = e.260 Introduction to Mathematics with Maple n+l n+l (10. Passing to the limit in the last inequality and using Theorem 10.. (1  +) from below. on using Theorem 10. that e 5 E.21). 1 < r 5 n.2! 3! 2 E. we have from (10. for m > n..22).24) Example 10.++ (n + I)! (n + 2)! ~ 1 m! I Yn + 7 .
namely that e cannot be a root a of an algebraic equation with rational coefficients. m E N and then + 1+ 1 + . With the commands Digits :=2OO and exp(i) you obtain e accurately for 200 decimal digits.Limits of Sequences 261 1 This shows that Y n .25) that 0 < e . n!n 10!10 1 The number l / ( n ! n )decreases rapidly. By multiplying the relevant equation by a sufficiently large integer it will be then a root of an equation with integer coefficients. Trivially every rational number a / b is algebraic as a root of the equation bx = a.7182818..2 The proof is out of our reach at this stage and the interested reader can find the proof in Spivak (1967.25) 3! n. 1 Prove that the sequence nw 1 n+l 1 + + . [Hint: Show that the sequence is increasing and bounded above by 1. This fact is expressed by saying that e is not algebraic or that e is transcendental.< 4 .yn 5 l / ( n ! n ) . We now prove that e is irrational. (10.is an upper bound for ym with m n!n 1 therefore e 5 yn . With the aid of computers e has been evaluated to billions of decimal places. If e were rational then e = m / n for some n.. . All concrete irrational numbers encountered so far in this book were algebraic and it is not easy to produce a transcendental number. It follows from Inequalities (10. The French mathematician C.. 3 1 1 culation will show that e 2.1 2A number is called algebraic if it is a root of an algebraic equation with rational coefficients.+ 7 + L. 4 . Hermite proved more. Chapter 20). Exercises Exercise 1 0 . for instance .+ 2approximates e with an error less than 2! l O! 108 hence The cal The middle term in this inequality is an integer and this is a contradiction..+ n+2  1 2n has a limit. Combining these results yields + > n and + n!n e m + 1 1+1+ 2! 1 +3! + <e n!  1 5 1+1+  1 2! 1 1 + +.
)n]2.] Exercise 10. (2) ( l + a ) 2 n = [(1+.= 0 for any c E C.5 Find the following limits . + . [Hint: Prove Exercise 10. [Hint: (1) (1+ .).262 Introduction to Mathematics with Maple Cn Exercise 10. n" L e.3 (1) nco lim (1 (2) + :In+'. n Use the Squeeze principle and Example 10.4 Find the following limits 5" j n 5n+1' 2n 3" (3) lim 1200 3n 5n' (2) lim n+m + + + [Hint: Divide by the dominating term.4.2 n! Icln/(n!) 1 0 b y an argument similar to that of Example 10.)n+5 = (1+ . then (1 + $) L $.)n (1+ .4.1 Exercise 10. Find the following limits $ir (1 + a)'".)"(1 (5) Show that (1 + a> . (3) (1(4) (1  $) n (l+) 1 n1 = (1 n .4. ) 5 .12.11 and then use (K) of Section 10.1 ncc Prove that lim .4.1.
:1 . X n + l = by the arithmeticthat lim x . for large n.4.8. 16. 4 . 16. . is obviously unbounded. (10. 1. . = ( n l)lal/n and consequently the sequence n ++ xn is decreasing. Then xn+l/x.27) (10. 1 .29) although this sequence. too.26) and (10.I 10. 2 ..27) increase without limit. It is natural to say that the terms of sequences (10.5 In this section we shall consider only real sequences.: nd.29) None of these sequences has a limit but (10. Consider the first few terms of the following sequences.26) (10. [Hint: use a similar procedure to that used in the previous exercise. We would not say that about (10.. . if x .9 Let p E N. . 2 . . .1. .6 Let a > 0. 1 . . . . Infinite limits +L then L = 1L. Prove @ Exercise 10. 2.11. . Show that the sequence is decreasing after the second term. . > 0 for n E N and %+' + 1 < 1 then X n x. [Hint: Use the method of Example 10.28) 3 5 7 n ~ b ..+1 n+m geometric mean inequality.2. . 4. then pass to the limit in the defining equation.] + @ Exercise 10.1.28) or (10. .26) and (10. + 0. [Hint: x.4. TL H a.4.7 Prove that nan + 0 for la1 < 1. . = nlaln. .Limits of Sequences 263 (3) lim q 2 n + 5 n + 13.4. exists and equals &.: 1 . 2. n+m n+m (4) Iim qi7F [Hint: For (2)(4) use the Squeeze principle. : 1. 2 2 2 nHc.4.1. [Hint: x.+1 < 2. (10.27) behave quite differently compared with (10. Prove that nPun + 0 for la1 < 1. .8 Exercise 10.29). 36.] Prove that if x .3.] @ Exercise 10.] ( 2 & 2 X n +: n ) . 51 = 1. Let x ..
28) is bounded and therefore cannot have an infinite limit.Xn > N .30) for all n . There is some similarity between the definitions of finite and infinite limits.5 (Negative infinite limits) A sequence x is said to have the infinite Eimit 00 (or just 00) if for every k there exists N such that Xn \ <k 00 ncc (10. particularly if one wants to emphasize that the limit is not infinite.31) for all n > N .26) has the infinite limit 00 since an = 2n1 2 n. Perhaps we should add that in mathematics infinite numbers are indeed introduced and used advantageously. We know that it has no finite limit either.29) is unbounded but has no limit either because all odd terms are eaual 1. Some theorems concerning finite limits can be extended to infinite x and the symbol n+oo 00 . but very little. finite or infinite. Sequence (10.264 Introduction to Mathematics with Maple Gefinition 10. The statement lim x n = 00 or 00 refers to the behaviour of the sequence itself has no meaning of its own. The limit in the sense of Definition 10. If x has the infinite limit we write lim xn = 00 or nmo + 00. Sequence (10. Sequence (10. / Definition 10. For a given K it is sufficient to set N = K . If x has the infinite limit Or X n + 03. if anything could be gained by introducing them here and now. we write lim x n = 00 Instead of saying that a sequence has limit 00 (or 00) we may occasionally say that the sequence diverges to 00 (or 00). We emphasize again that we have not introduced any infinite numbers. for we have bn 2 n / 2 and it is sufficient t o choose N = 2K.27) also has limit 00.4 (Positive infinite limits) A sequence x is s a i d 2 have the infinite limit +00 (or just 00) if for every K there exists N such that xn >K 00 (10.1 is sometimes called finite. Sequence (10. Obviously lim (n) = 00 whereas the sequence n H (1)"n has no n+m limit.
There exists M such that lynl < M for all n E N. Since n + xn is increasing it follows that xn > K for n > N .10. which proves that xn += 00. Also a sequence can have at most one limit. It suffices to consider an increasing sequence n I+ xn. un = n.33) (10. The next theorem is the analog of the Squeeze principle.12 need not hold.lYnl > K for n > N . It follows from the assumption that x n > K for n > N . wn = (2 (l)n)n. = 2n.13 infinite). 0 f Theorem 10. 0 + + + If n + Yn is not bounded (for instance if Y n + 00) then the conclusion of Theorem 10. Theorem 10.11 If xn 2 Y n for n E N and Y n then xn 4OC) (yn + 00) also. finite or infinite. Let 2. while + Xn Xn Xn xn + w. The proof for xn + 00 is similar. 0 . I Theorem 10.272. Since xn + 00 for an arbitrary K there exists N such that 2.35) + 00. + zn + + Yn + a. += 00 (xn + 00) Proof.32) (10. Y n = a . changing finitely many terms of the sequence affects neither the existence nor the value of the limit. Z n = 371. > K A 4 for n > N. If it is bounded then it has a finite limit by Theorem 10.34) (10. There exists N such that Y n > K for n > N. finite or infinite.12 Yn += 00 (00) If xn += 00 (00) and Y n is bounded. Then xn += 00 and all the other sequences diverge to 00.Limits of Sequences 265 limits but not all can and care is needed. then xn + Proof. (10. The proof for xn OO is similar. If it is not bounded then for every K there exists N E N such that X N > K . For instance. Let K be given. Consequently X n Yn > K M . Every monotonic sequence has a limit (finite or Proof. U n += +00 has no limit.
5.7 (1) lim nmc Prove: if zn + 00 or X n + 00 Prove: if Ixnl +.3 and + + 00. (5) it is additionally required that for some fixed number a (for example. Xn 00. Exercise 10. then .5. > 1 then an + Find the following limits 7 n2 + 3n .+ 0. + 00. If Give examples o f x . ( 2 ) for a given a E R it is true that xnyn + a. @ Exercise 10. and yn such that xn + 00.266 Introduction to Mathematics with Maple Exercises @ Exercise 10.5.5.5.5. finite or infinite. Xn .4 remains (1) the inequality n > N is replaced by n 2 N .2 (2) lim n00 n 2 1 O n .8 Exercise 10.5. a=O)K>a.6 Exercise 10.9 > 0 and f i + 1 > 1 then xn > 0 and xn+l I > 1 then xn + + 00.2 If xn + 00 and Y n + c and c > 0 then X n Y n c < 0 then X n Y n + 00. Yn +0 (1) X n Y n 00. Some of the following exercises are stated as theorems. @ Exercise 10. ( 3 ) the sequence n + xnyn has no limit. (2) it is additionally required that N E N i (3) it is additionally required that K E N.4 @ Exercise 10.5.5. The task is to prove them.1 unchanged if Prove that the meaning of Definition 10. Prove: if a 1 00 then 1xn1 + 00.1' + (3) for Q 2 O evaluate If xn If xn lim n" 1200 Exercise 10.5 @ Exercise 10. (4) the inequality X n > K is replaced by xn 2 K .10000 n+Jn 1 .
then the sequence k H X n .Limits of Sequences 267 10. is called a subsequence of the sequence n H xn.14 If a complex sequence x has a limit 1 then any subsequence k I+ xnk also has the limit 1.14 can often be used conveniently for proving that a given sequence is not convergent by finding two subsequences with distinct limits. For instance are terms of a subsequence of (10.l ) nhas two subsequences k 1 (1)2k + 2 and k H 1 (1)2k+1+ 0 and is therefore divergent.36). For E: l 0 Proof.Z I < E. This leads to I Theorem 10. 1xn . Theorem 10. For instance the sequence n H 1 ( .36). k > N then nk > 0 there exists N such that 2 k > N and therefore Ixn. One can prove similarly that if a sequence diverges to 00 (or to 00) the so does every subsequence. = 1. + + + Example 10.6 Subsequences If k nk I+ n k is a strictly increasing sequence of natural numbers. . If This proves that lim nca Xn.11 < E for n > N . that is E N and n k < n k + l .15 Since for n 23 . If we list the terms of the sequence (10. It is intuitive that if xn approaches a number then so does xnk.36) we obtain a record of the terms of a subsequence by omitting some members in (10.
l). 0 Exercise 10.6. Exercise 10.2 0 Exercise 10.] Existence theorems Monotonic sequences3 have the nice property that they always have limits. Let the limit be a. noo [Hint: Use Example 10.1.2 1since fi 2 1.12 we have ' a 6 = ( 'm)+ 2 a2 and also = fifi +a.39) 3Monotonic sequences are automatically assumed to be real. if the sequence is bounded above.38) an L xn I Pn (10. Moreover a . there is a decreasing sequence (10.15 to show that (1) an +O for 0 < a < 1.1)".2. [Hint: For K E N n + eK < n for large n. The subsequence k I+ has by Theorem 10.6. Exercises @ Exercise 10.268 Introduction to Mathematics with Maple the sequence is decreasing after the third term. It is also obviously bounded below.1 Use the method of Example 10. Consequently a = 1.7 Find lim n+oo (fi . With a sequence n I+ xn E R bounded below we can associate an increasing sequence n I+ a n as follows: Similarly. so we have a = a2.3 we have (1 + 10.6. It follows then K < n ( fi . hence convergent. By Remark 10.1) = 00.7 and Example 10. .] a> Prove that lim n (fi .14 the same limit.15 and the Squeeze principle. (2) VZ +1 for o < b.
We define: lim nmc an as lim inf xn and lim nmo noo Pn as lim sup xn. = 1. = noo if Xn E Iw and n+cc the sequence is not bounded 00. n+oo = liminf xn = 1. p = 1. hence limsupx. liminfz. x3n1 = 0. Similarly if xn liminf noo 2 . Clearly lim inf zn 5 lim SUP xn.2 =1 . We say limsupz.. 23n = 1 for n E N. noo These are called respectively the limit inferior and limit superior of the sequence n H 2.2 4121 1 displays z n . limsupz. pn = I . noo = = 1. .16 have that Let x3. Example 10. It is left as an exercise t o show that limsupx. Example 10. I 1 then a n = I . We for every n and consequently a n+oo = 1. a n and pn for the first few n.18 Let xn = (l)n+l + noo noo . = nmo 00 if lim xn = We employ similar conventions for the limit inferior.17 noo If Zn T 1 then clearly a n = Xn.Limits of Sequences 269 for every n E N. pn = xn and again limsupx. above. = 1 noo liminf xn = 1 and Example 10. The following Table 10. noo 00 n+oo We say that limsupx. = 1.
n+m 7100 A sequence is said to be a Cuuchy sequence or simply Cuuchy if for every E > 0 there exists N such that lzn x. Proof.270 Introduction to Mathematics with Maple Theorem 10. xn noo I. It is fairly .40) for all n > N and m > N . Consequently This proves that 11. In other words. 0 Remark 10. The limit of a sequence can be described informally as a number to which members of the sequence are ultimately approximately equal within a prescribed accuracy. If lim = 1 then for every E: > 0 there exists N such that for n > N . the Theorem remains valid if the word bounded is omitted.15 can be extended t o any real sequence. Again we separate the proof into two parts.6 Theorem 10. liminf x n = 00 if and only if lim x n = 00. A Cauchy sequence has the property that ultimately the members of the sequence are approximately equal to each other within any prescribed accuracy.15 and only if A real bounded sequence n lim sup xn = lim inf xn n00 nm I+ xn has a limit if and then lim sup x n = lim inf Xn = lim 1200 Xn n+co noo .1 <E (10. If limsup xn n+ca = liminf zn = 1 then lim noo noo xn = 1 by the Squeeze Principle since an 5 xn 5 pn. For instance.
and is proved below.zmI 5 I x .rOO = 1. . & & Now we have l x .+ . Let n H x. The converse is also true.11 <2 & for k > N. Since xn is Cauchy there exists N1 such that for n > N1 and m > N1.11 < .. 2 2 for n > N and m yielding ~~ ~~ > N. X..16 (BolzanoCauchy) A sequence with complex terms is convergent if and only if it is Cauchy. Theorem 10. every E there exists N such that k 1 . k 1. l Proof.Limits of Sequences 271 obvious that a convergent sequence is Cauchy.i ) / 2 . In (10. . There exists a natural N such that for n and m greater than N . Assume for an indirect proof that = lim inf x.  Set E = ( d .= E . + I then for . if x. Since the finite set of those xn with n 5 N is also bounded. We obtain + . This allows us to make use of Theorem 10. x n roo < lim sup x. be a Cauchy sequence. We first assume that xn are real..15. Indeed. This means that all terms of the sequence with TI > N are bounded by the (fixed) number Ixml 1. say Ixnl < K for n 5 N .41) we keep m fixed and run n through k . We need to prove the if part.11 + lxm . we have + for all n and hence the sequence is bounded. Choose a particular m rn = N 1) then + >n (for instance for n > N . .
272 Introduction to Mathematics with Maple By the definition of limit superior and by Corollary 10.~m+2). Firstly we have 1 + 1 2 = .19 Let x1 = 1. Since d E = (1 i)/2 it follows + This contradiction completes the proof in case of a real sequence. By Equation (10.E is a lower bound for xm for m > N 1 and therefore <inf{~rn. The BolzanoCauchy Theorem is often referred to as the general convergence principle or the Cauchy convergence principle. It is now clear by induction that for m 2 n 2 the number zm lies between xn+l and xn. If we knew 2 that the sequence were convergent with limit 1 we could find 1 easily. . = 3 x n . xn+2 = Xn + Xn+1 ~ n + 2 +1 TXn+1 = Xn+l+ Xn 1 2 = *. Indeed Example 10.q )= (l)n. 1.~rn+l. 22 = 2. Clearly Consequently. 1 = .42) Obviously zn+2 lies between xn+1 and 2.E 5 xm.1 on preservation of inequalities it follows i. the sequences n + un and n 4 vn are Cauchy and by what we have already proved for real sequences they converge.8. For a Cauchy sequence with complex terms xn we define U n = !RXn and 21. 2n 2n (10.We prove the convergence by showing that 5 3'  (x2 . 1 . Taking the limit as m +00 gives i E 5 I.x m l IlXn+1 znI = 3. and ~ n + 3 remains there by lying between ~ n + 2and xn+1. By Theorem 10. = 22 1 + x1 2 5 =2' and therefore I 5 2 z is Cauchy.4 0 it follows that the sequence x converges.42) we have + lxn ..
The sequence n + an is increasing and bounded from above by bl and therefore has a limit.9 The conclusion of the Theorem can be restated as 00 n=l n [an. bn] for n E N then Iy . B] contains infinitely many terms of the sequence. If also y E [an. Conthis interval by [a2. bn] = { c } Proof.an + 0. it follow^ that c 5 bn for all natural n.Limits of Sequences 273 for m > n. b2] and choose n tinuing with this process we obtain a sequence of nested intervals [ a k . say c and then an 5 c for all natural n. Let A 2 xn ~. We have now proved that an 5 c 5 bn for all n E N. Every bounded se We assume first that the sequence z is real. One of the intervals L [F. bn] be a sequence of closed bounded intervals such that bn .an + 0 and [an+l.CI 5 bn . bn] for all E N. Proof. Set a1 and = A . Theorem 10. Since bn+l 5 bn. For every E > 0 there exists N such that 2N+1 2n+1 + 0. we denote 2 >n 1 such that xn2 E [a2. Theorem 10. since bn = an (bn . that is y = c. bz]. < B for all n E N.18 (BolzanoWeierstrass) quence contains a convergent subsequence. Remark 10. It says that shrinking closed intervals have a point in common. bn+l] C [an. 0 + The proof of the next theorem contains a typical application of the Nested Intervals Lemma. The sequence n + bn is decreasing and has the same limit. Sending n + 00 gives Iy . Consequently for n > N and m > N we have < E since The next theorem is very plausible and is often used.an for all natural n.c I 5 0 .17 (Nested Intervals Lemma) Let [an.a n ) and bn . Then there exists a unique c such that c E [an. bl = B and n1 = 1. bn]. b k ] .
] + pn. U n k + &. + (bk .44) Give an example such that lim inf Y n n+oo < lim sup xn. 5 bk 5 c Xn. By the Nested Intervals Lemma there exists c E [ak. This completes the proof for real x. . By Theorem 10. Similarly.x n . has a convergent subsequence vnkj + ij.] a. Y n + l .+/~mxm11 5 7n182n to prove that it is Cauchy. n H Y n with (10. x3. Similarly for liminf. Use induction to prove that 2 3 5 xn 5 2 2 . Thenuse IznXn+lI+".2 Prove that for real sequences n X n 5 Yn for all n E N H xn.x n I 5 glxn . = yn/2.274 Introduction to Mathematics with Maple such that bk .1 Let x1 = 1. Then show that Ixn+1 . .8 Although we did not mention it explicitly we used the Recursion Theorem 5.a k ) . For the example . For a bounded complex x with un = 9222.2 in defining the sequence of intervals [an.4 it follows that 3 0 Remark 10. xn+l = 1 x i / 2 .43) (10. Otherwise set pn = SUP set Y n = 1 {yn. Exercises Exercise 10. Deduce lXn+1xn) 5 7n181n.a k ) 5 ak 5 xn. . = n+m 00 there is nothing to prove. the bounded sequence k H V n . bk] for all k E N and then C (bk . + Exercise 10.7. By what we have already proved there is a convergent subsequence. for n 2 3. n+m [Hint: If limsupy.7.a k = ( B .A)/2"' and X n k E [ a k . b k ] .} and use pn 5 (I)~ xn . The limit I must satisfy I = 1+Z2/2.i I . bn]. Prove that n H xn is Cauchy and zn 3 1 [Hint: Calculate 2 2 . and V n = 3 Z n we have that the sequence n H un is also bounded. The Squeeze principle implies lim kmo = c.
[.45) n+oo Give examples showing that all inequalities may be strict.ml 2 a.7.x n . For an example set Xn = Yn = (.I)nI Q) Exercise 10. Ic 2 n} sup { y k .00[= 0.8 Comments and supplements A number X is said to be an accumulation point of a sequence if there exists a subsequence converging to A. @ Exercise 10. For n > N obtain I .ml 2 4 10.5 Let x1 = 1. Prove State and prove a similar theorem for limit inferior. [Hint: 1 5 xn 5 2.7. 00 1 n=l n Similarly. (10.4 liminf nw xn n+oo Prove that for bounded sequences Yn + lim inf lim inf nca lim inf Xn n00 + lim sup n+oo n00 Yn lim sup ( x n nmo + Yn) 5 lim sup xn + lim sup Yn.7. Ic 2 n } 5 sup { x k .E Yn 5 xn Yn 5 1 E Yn then pass to the limit + + + + superior .1 + @ Exercise 10. [Hint: There is a positive number a and a subsequence n I+ xnj such that lxnj . Set y j = x n j . n=l Why this does not contradict Theorem 10.7 Prove: If n I+ xn is a bounded sequence and m is not the limit of this sequence then there exists a number 1 # r n and a subsequence converging to 1.7. k 2 n } then pass + + to the limit as n + 00. [Hint: For lim sup obtain sup { x k y k . IXn+l xnl = f i ( d ~f i ) . [Hint: Let xn + 1.n E N are nested but there is no n 00 [n. Prove that the sequence x is Cauchy and find its limit. = 2.6 cE Intervals 0.' I x n .Limits of Sequences 275 @ Exercise 10.Xn11 I Ixn .oo[are also nested but 13 . n]O.17.] @ Exercise 10. intervals [n. .3 that Let n H xn be convergent and n I+ Yn bounded. If there is a subsequence with an .l I / f i ) x n + 1 f i = X n J G = ' * .7. by the BolzanoWeierstrass Theorem there is a subsequence y j k + 1 and 11 . 52 = 2 and xn+1 = / .
8 clearly xjn lim SUP x n .1 > 0 for all n E N then lim inf nco % 5 lim inf xn noo %+I 65 lim sup 6 5 lim sup .5 an. It follows from Theorem 10. Xn n+oo nmo n+oo Xn 00. . we may write 0 = liminf nco Xn+1 zn+l< liminf 6 = 1 = limsup 6 < limsup = oo.37).< ~ j . The introduction of infinite limits superior and inferior enables simple formulation of some theorems.39) we have a n k 5 x n k 5 o n k . For every n E an  N there is a j n such that .46) We now show that lim sup x n is the largest accumulation point by defining nmo a subsequence converging to it. 4 P ~ ~ ~ ioc. we say also that 00 is an accumulation point of the sequence. (10. not necessarily bounded.4 A sequence can have many accumulation points.15.38) and(10. a sequence has a limit if and only i f its limit inferior and limit superior are equal. If a sequence. n400 1200 (10. using the notation (10. has only one accumulation point then it has a limit. for instance 00. b lor y . for instance. with this convention.276 Introduction to Mathematics with Maple infinite limit.4 asks for the proof of the following theorem: If 27. 1200 7200 (10.1= 1 and ~ 2 = . n then. Moreover.8 that for any accumulation point X lim inf xn 5 X 5 lim sup xn. n00 Similarly the limit inferior is the smallest accumulation point.47) Xn In inequalities like these we interpret the symbol oo as larger than any real number and oo as smaller than any real number. the sequence with the following terms 1. for instance ~ 2 n . n 1 By the definition of limit superior and Theorem 10. If. Exercise 10. For a bounded sequence it is clear that the set of all accumulation point is also bounded. By the BolzanoWeierstrass Theorem every bounded sequence has an accumulation point and if it has only one then it is convergent by Theorem 10.8. 5’3’ 3’ 4’4’4’5’ 5’p ” $7 1 1 2 1 2 3 1 2 3 4 has any number between 0 and 1 as an accumulation point.
. . 1/4. 5/3.4/3. .8. Consider separately Y = 0. [Hint: xnyn 5 YnPn.8. 1/3.1 0 Exercise 10.47).7.3 and 10.15.] + c:’” + (x: + 0Exercise 10. * * Xn n 5 lim sup xn.3/4. ] Prove: If yn noo 0 Exercise 10. 4/3.Limits of Sequences 277 Exercises @ Exercise 10. .7. let P + 00 using Exercises 10.5/3. .2/3. 1/2.11/4. . . 2/3.4 as well as Theorem 10.4 Prove Inequalities (10.1/4. noo [Hint:(l/(n p ) ) xk I ( l / ( n p ) ) xk pPn). 11/4.1/3.1 Define a sequence for which all real numbers are accumulation points. State a similar theorem for liminf. [Hint: 1/2..2 +Y 2 0 and n n+oo + xn is bounded then lim SUP Ynxn 5 Y lim SUP xn.8.3 n+co n+co Prove: 21 lim inf x n 5 lim inf + + + ~2 Xn n 5 limsup n+oo 21 + + + ~2 .8. Then let n + m.
This page intentionally left blank .
an is the nth term of (11. (11.1).1) i=l To indicate that we consider n I+ sn c rather than n H an we write ai. On the other hand. if ai = . with sn as in (11.2). For an arbitrary sequence n H an E C we can form another sequence by successive additions as follows: sn = a1 + a2 + .1 Definition of convergence Study of the behaviour of the terms of a sequence when they are successively added leads to infinite series. Sn is the nth partial sum of (11. We also introduce power seriesa very powerful tool in analysis.2) a “series” or an “infinite series”.2) is just an abbreviation for the sequence n Sn. 11.2) are..2). It is usually clear from the context what the partial sums for a series 1 like (11.2) The symbol (11.for some positive integer k ka 279 .Chapter 11 Series In this chapter we introduce infinite series and prove some basic convergence theorems. We shall call (11.+ an = z * n a i (11.
i=l (11.1).. or variants of these. For instance. but the notation has persisted ever since.. but the limit of the sequence of partial sums. Instead lWith sn given by (11. ‘the series has the sum S’ and ‘the series converges to S’. the same meaning as1 ncc lim sn = S.1 (Convergence) convergent to a sum S if n The series (11. +k k2 k” 1 1 1 However the symbol write a ka is ambiguous. of course..i . the geometric series. by Theorem 10. so we may speak of ‘the sum’ rather than ‘a sum’.3) The notation 00 Cai=S i=l (11. + + + + De finition 11.2) to be convergent to a sum S Obviously. for example we can write aj or arc instead of ai.3) has.1 a series can have at most one sum.+ .. The phrases ‘the series is convergent to the sum S’. The notation and terminology for infinite series comes from the times when the theory was rather vague.2) can be replaced by another letter without altering the meaning of the symbol.280 Introduction to Mathematics with Maple and every natural number i then sn = + .qn’. 21t comes from times when the sum of a series was not well defined but regarded vaguely. On occasions like this we would k2 rather than k . are used interchangeably. The usage of the term sum2 is a bit unfortunate because the sum of a series is not the usual sum.4) is used as an abbreviation for series (11. . The letter i in (11. Equation (11. that is the sequence n I+ sn where S n = 1 q q2 .2) is said to 2 n+ca lim Cai= S. is denoted by qa’. intuitively and imprecisely as a sum of infinitely many terms of the series.
1+ . For example.7). a3 (11. namely denoting the series and its sum by the same symbol. namely has two different meaningsit n H c n denotes the sequence of partial sums ak. Similarlyas with sequencesthe change of finitely many terms of a series does not affect its convergence or divergence.6) have the advantage of being rather indicative of the terms of the series: for instance one gets a better impression of the series (11. which.2) or (11.6) The notations of (11.Series 281 of (11. but changes the sum in the obvious way. the series .. .. = S. many definitions for sequences carry over to series.5) and (11.. diverges to 00 (or 00) respectively. When using (11. divergent t o 0 0 (or 00) according to whether the sequence of its partial sums diverges. a series is called divergent. (11..+ 1 +1 + . Perhaps we should add that most authors n n+ca ak and also lim k=l use the symbol c ca k=l ak instead of our symbol c Uk.1.6) one has to realize that the same symbol.5). In particular.+ 1+ .5) or a1 + a2 + a3 + . of course. Since a series is nothing but the sequence of its partial sums.4) the following suggestive notation is often used: a1 + a2 + + . has k=l the same disadvantage as (11.7) by writing 1 +1+ 3 1.+ 32 22 33 1 35 32 34 42 rather than (11.
and then or better still (11. . 1 1. .3 Let us consider the series . .1 It is an immediate consequence of Definition 11.' nws for q # 1 and lim sn = r 1 i since n+oo 00 lim qn = 0 for lql < 1 by Example 10.=1 ( l . Consequently C qi' i= 1 & Example 11...13.+1 + 1 . 1 1. for N E N does.2 The geometric series Cq2l converges t o = 1 1 if lql < 1.3 1 + 31 + + .3 + 31 .1)(2n We need some useful expression for the partial sum Sn = .5 + s e e + + I) ' By using the identity 1 (2kl)(2k$l) we obtain s.8) 00 By analogy with the finite sum we shall often write i=N+1 00 ui instead of i=l Example 11.282 Introduction to Mathematics with Maple converges if and only if. 3! 4! Example 11.7 1 (2n . sn = 1 + q + .1and the result of Example 10.14 that 1 1+1+ 2! 1 1 +++**. .1 +2 3 3 5 1 2n1 1 1 )=2n+1 2 1 2(2n+1)' .= e. . . = Indeed. + q n ..5 5.
however this representation is no longer unique.5 The series 1 1 + + +. with aN < 9.8 that IR.. and ai E {0. ' . if a0 . a1 + . . where G N = aN 1. For z > 0 it is customary. if z > 0 and z = a0 .9) Consequently. .. 10N 3 1 + k=N+1 10k Similar comments apply with appropriate changes made to binary and ternary fractions and obvious modifications should be made for negative reals.a1u2a3. + an < z < a0 + a1 + . for 4 .s. This is obvious because + aN + 10N c =alv. . 9 } for i E N...9) holds. We have and sn + 00.1. Jz& is divergent t o 00. alaza3. Every real z can then be represented by a decimal fraction.ala2a3..9). . Let ao. the decimal fraction for z if (11. a1a2a3. where a0 E No.1)(2k k=l c Example 11.4 z m + 1) 1  2 > 0. then also z = a0 . a. in contrast t o what we have done in Example 5.2. .IS 1.GNOOO. . .. For example. be the decimal fraction for z E We know from Example 5.. that is (11. then z is the sum of the series (11..+ an + 1 +10 10n 10 10n 10n It follows from the Squeeze Principle and Example 10.13 that n i=O ai 5 = z.. ..Series 283 Since sn 4 4 we have 1 (2k . is the decimal fraction for z. ula2as.aN999.33333. Example 11.8 to call a0 . .. instance the decimal expansion for ..
1 + . inserting parentheses into a convergent series is permissible.' (5) 1000 1001 1003 1 1 +++.284 Introduction to Mathematics with Maple We shall see in the next section that sums of series share many properties.1. (4) 12.5 1 2 ++ 4.1) + .1...2 Prove that if an = zn . . hence the series converges to zero. On the other hand. All the terms of this series are zero.3 2... .. To see this. .4 3'4. 4 4 . and find the sums of the convergent series.3 3. because this amounts to a selection of a subsequence of partial sums...10 1 1 1 ++.4 1 1 1 * (3) +++. n+cc @ Exercise 11. For example. with ordinary sums (that is sums over a finite number of terms). s2k = 0 .1. which is clearly divergent. with partial sums s1 = 1. Exercises. but not all... (1) (2) 1 * 1'2 2. By removing parentheses we obtain the series 1. s3 = 1.1)+ (1.. . and we know that every subsequence converges to the same limit as the original sequence. consider the series (I . it is not permissible to remove (infinitely many) parentheses in a series.. n=O 00 . s2k+l = 1.xn+1 and lim xn = I then the series C an is convergent and C an = XQ .. Exercise 11. s2 = 0 . 7 7...3. 1 .1 +1 . ++ (6) (7) x +C ( + ' $G  ~) with x > 0.1 Decide which of the following series are convergent and which are divergent.
4 00 If bk.11.7 Prove that (1) Partial sums of a convergent series form a bounded sequence.3 The series 1 ..on 2kk 2k' 2 k 2kk k=l k=l c three decimal places if k=N+1 0 Exercise 11.(1.1) . 5. since k+l 2 k+l O0 k + l is approximately equal to < .1. .(1. By inserting parentheses differently we obtain the series (1.1.1) . < b. k+l Exercise 11.6. Cak and converge.x)2 use Exercises 10. for some m E I V then k=l ak bk.1..1. especially Theorem 10.5 Find the approximate sum of the series .4. * converges to I. Explain this seeming paradox. and show that if in addition that 00 00 a.3. ( This exercise can be done easily with the help of Theorem 13.7.25 of Chapter 13.1. ( 2 ) If C U k converges then for every positive E there exists an M E N 11.1) + (1 .  which converges to zero.to 2% three decimal places.2 Basic theorems The following theorem is an immediate consequence of Definition 11. [Hint: i=n (1.6 and Example 10.Series 285 @ Exercise 11.8). @ Exercise 11. [Hint: use the previous exercise and (11.)] Prove that c &< c(i + 2 00 1 1 @ Exercise 11.the sum .1)+ C bk < k=l * .6 l)xi = for 1 x 1 < 1. and ak 5 bk for k E N 00 then k=l ak 5 k=l Prove this result.1 and theorems on limits of sequences.
.. ntm nmo 00 Sn = 0 Example 11. then there exists a K such that K 22 1 1x1  Then for k3K we have . a k = S .. . Let S n = C i = l a k and ~ ~ .6 1x Consider the series * + x3 + x4 + x6 . Since lim 7200 lim s n .+ 2 4 .l ) = 0. For 1 x 1 < 1 both of the series 1 + x3 + x6 + . k=l if J: # 0. ..x7 + . then 00 C ( a k + b k ) con k= 1 k=1 k=l Remark 11. . Indeed. and c E C.27 + .286 Introduction to Mathematics with Maple I( Theorem 11. = 1+ 2 3 + x 6 + . and 00 C bk 00 converge..1 I/ l (i) If C a k converges.7 The series x ( k ~converges ) ~ if and only if x = 0.x7 + x9 + X 1 O + 212 .1 = S we have lim an = lim (sn .2 If x a k converges then lim k+oo ak = 0..1 Using (i) with c = 1 and then applying (ii) gives Example 11. converge and by Remark 11. 1+x3' Theorem 11.1we have 1 + 2 3 + 2 4 + x 6 .. and x + x4 .s n . . and k=l k=l (ii) If both series C a k and verges. n+oo Proof. . then so does C c a k .27. .1 123 X * ..
by Theorems 10.l 5 ~ 2 it n follows from (11..2 is false.5. Proof.+2" 3" 1 .8 The series converges if a > 1.1 . diverges. Let 0°1 k=l = k . therefore I ( k ~ ) 2 ~ l1. Consequently. 1 kmo A Theorem 11. The assumption that the terms are nonnegative automatically implies that they are real. . + 7 ) 3" I 1 + 2" S2n. 1 which is called the harmonic series. lim kmo # 0. The sequence of partial sums is increasing because the terms are nonnegative. 4" 1 + L + + + + 2" (271)" Sn (24" 1 1 1 (1+ ( 2 n .2. Consequently.from Example 11.10) Since ~ 2 n . We will prove this indirectly. and the series must diverge by Theorem 11. For example.9 The series cE. by assuming the contrary.Series 287 ( k z [2 1.3.10) that the sequence of partial sums is bounded and the series converges by Theorem 11. 0 Example 11..3 A series with nonnegative terms converges if and only if the sequence of partial sums is bounded. 3" 5" 1 1 < 1 + 2" +.9 and 10.2 series The converse of Theorem 11. Hence it converges if and only if it is bounded..I + . Example 11.5 diverges although lim 1 A . the .+ . a E Q.= 0. for a > 1 we have 2"1 5 2"1  1' ( 11. Remark 11.+ . We have 1 1 =I+++.1)" 2" 2" 1 1 1 1 ++ ..
1 . Then.288 Introduction to Mathematics with Maple 1 S.. 1. 1 we have nQ 1 > n 1 and Exercises Exercise 11.+. + + .1 . < divergence follows from the comparison test. . The series Example 11. _ _ _ _ _ _ ~ /I Proof..+ 1 ..+1 1+ 1 S = l ++.1 Use the comparison test to prove the convergence or divergence of the following series..9.4 (Comparison test) If 0 5 a k 5 the convergence of C bk implies the convergence of ~ bk for k E N..) = T + s.+ 1 + . = 1 + 1+ 1 + . On the other hand. then its partial sums are bounded. let k=l 1+ 1 + . ..11) Theorem 11. S = 27'.. 1 s=1 +  2 3 Consequently. + . If C bk converges. which in turn implies that C a k converges.. hence this series 2k1 = T . The partial sums of 0 0 4 are bounded above by S. 0 Remark 11. 2k ..=I+1 ++ . . and by Theorem 11. This implies that the partial sums of C a k are also bounded.1. For a = 1 For a 1 this was established in Example 11..10 72" diverges for a 5 1.3 divergence of ak implies the divergence The conclusion of Theorem 11.then Cak.2.1 3 5 2 3 2 1 also converges.+ ? 1 ( l +1+ 1 + .4 can be rephrased as the of bk. a E Q.5 3 5 3 which is a contradiction. 3 5 2 4 6 2 3 1 1 1 1 < 1+ + . = 2T (11...
however. but C b.n=I. .3 If a. Generally speaking. ak 4 0 and the series diverges.3 but would require far more work.20). Use the comparison test. ./2 < a. it is difficult to find a sum of an infinite series exactly and on many occasions only a numerical approximation of the sum can be found. might diverge and C a. > 0. .3 5 =L > 0 b. < 31LIbn/2.2.i=l. ( 3 ) the partial sums are bounded. infinity) . implies that of C a.IO> . > 0 for n E N and lim n00 prove that the series a k and C bk either both converge or both diverge.. .] 11. b.lO). ( 2 ) the partial sums are bounded and the series diverges. What conclusion can be drawn if L = O? [Hint: There exists N E N such that ILlb. .2 Give examples of series C U k such that (1) ak + 0 and the series diverges. 1 32659200 Maple can find sums of infinite series for which we have not yet developed enough theory.i=i . (n+ 10) 1 can be found by the same method as in Example 11. .Series 289 @ Exercise 11. For instance > sum( I/n^2. Maple and infinite series For a few series it is easy to find the sum.2. Finding the sum of an infinite series in Maple is very similar to finding a sum of finitely many terms.infinity) . For example > sum(l/product((n+i). For instance the sum of c 1 00 (n + l)(n + 2 ) . . 0@ Exercise 11. If L = 0 then convergence of C b. converge. > 715357 23363003664000 sum(i/product ((n+i) .n=l. Maple can find sums of many series almost instantly.n=i.
[Hint: Digits=l5. Exercise 11.279585302 Exercises 00 Exercise 11. .4 Absolute and conditional convergence Theorem 10.7r2 1 6 In other situations we can even obtain a result hard to understand.n=l.3.evalf(sum(l/nA3.2 Find O01 i=l to fifteen decimal places. the socalled BolzanoCauchy convergence principle.290 Introduction to Mathematics with Maple . For example Mup2e returned the result in terms of the hypergeometric function but we need not to go into the study of this function: we can obtain a numerical approximation easily with the evalf command.16.3. leads to .] 11.infinity)).1 Find the sum of the series i=l from the first principles and by Maple. 1.
Condition (11.Series 291 Theorem 11.7 part (ii).1. then for every positive number E there exists an M such that for n > M l This result was proved directly from the definition of convergence in Exercise 11. the sequence of partial sums converges if and only if (11.1 I f a k converges. Hence.12) Ik=n+l I for every natural number p and n > N . there exists an N such that (11. I/ Corollary 11.5 The series a k converges i f and only if for every positive E . Proof. Remark 11. Here we provide another proof.5 is also referred to as the BolzanoCauchy Theorem. By taking limits as p + 00 we obtain Example 11.5.4 Theorem 11. Clearly W e prove again t h e divergence of the harmonic series 1 k=n+l 1 .12) is equivalent to the partial sums forming a Cauchy sequence. p E N.n > M .11 1 C k . 0 Proof. There exists an M such that for n E N.12) is satisfied. or as the general principle of convergence.
5 Example 11. s ~ a2. Consequently.+1 + 0.+2) 2 s2.+l] for every n E N..and s2.+ 1. 11. Moreover. a E Q. . Using Remark 11.1. = The bounded closed intervals [ s ~ ~ . The 'only if' part is an immediate consequence of Theorem 1 1 .  2 0 then the A series x(l)k'ak with series. Let I.~ > ~ .1 for every n E N.. The sequence n I+ s 2 . (a2. Since an + 0 for every positive E .6 (Leibniz test) If a 1 2 a 2 2 a3 2 series C ( ..+1 . the BolzanoCauchy condition is not satisfied for p = n and the series must diverge.. By Theorem 6. Indeed we have a 2 n + l . the error made .7. Uk 2 0 for every k E N is called an alternating Proof. ~ 2 ~ 4 = 1 s2.SZ. the series 1. . a2.+1 L 0 and san+l = ~ 2 ~ 1 a2n mn+l 5 s2.5 with a = 1 yields . and has the same sign.1 is decreasing. is increasing and the sequence n I+ s2..292 Introduction to Mathematics with Maple Consequently. sa.c 1 2 ~ + 2 2 0 and consequently s2. E = $ and Theorem 11. 1 I+ 2" 3" 4" converges for a > 0.a2. Thus + + + + are ~ nested + I ] and s2. I f rn > 2N 1 then + The sum of an alternating series with decreasing terms is trapped between ~ 2 .+1 . 3 . does not exceed in absolute in replacing S by a partial sum. = CL=l(l)klak. U 2 n t l L S 2 n . Similarly. s .l u k converges if and only i f a k + 0.3 there exists an S E [Q.. Remark 11. that is S ..s value the first omitted term.+2 = ~ 2 . a2.12 Since n" 1 0 for a > 0. there exists an N such that 0 5 a 2 ~ + 1< E .
by the BolzanoCauchy Theorem. Proof. for every positive E there exists an N such that k=n+l for n E N.2 (Absolute and Conditional Convergence) If the series lakl converges.Z . both of the series a: and x u .1. 0 _< a.6 The alternative proof of Theorem 11. l Proof. Then 0 5 a: 5 lakl.p E N and n > N ..Series 293 or. and then the convergence of C l a k l implies the convergence of C a: and C a.. _< l a k l . then so does C ak . since ak = a. by Remark 11. A convergent series which does not converge absolutely is said to be conditionally convergent.7 I f C lak I converges. By the BolzanoCauchy Theorem. 0 D 11. This in turn implies convergence of C ak. = 2 and n = 100 1 c 100 n=l (l)n+ln 2  < " n=l c (l)n+l 1 5 n2 c 100 n=l (ly+l 1 n2 1 +1012 1 Theorem 11. C C . 0 For series with real terms there is another proof of Theorem 11.a. = ak if a k 5 0. Remark 11. are divergent. with a. then C ak is said to be absolutely convergent. Since Ik=n+l the series I k=n+l ak converges.7 shows t h a t for a conditionally convergent series ak with real terms. Let a+  Iakl +ak 2 IC ak  = laklak 2 so that a: = ak if ak 2 0 and a.7 which employs an idea often used on other occasions.
+ 1 1 1.13.. is only conditionally convergent for 0 < a 5 Ica Absolutely convergent series have some properties in common with finite sums which are not shared by conditionally convergent series. are terms of a bounded sequence and ukbk converges absolutely. Prove that if CUEand chi are convergent then [Hint: lakbk) 5 1/2(u. .4.1 +1 + .13 The series (1)LtJ .3 does too. verges absolutely. by Example 11. 1 1 . prove that If converges absolutely prove that C bk Ic con Exercise 11. Exercise 11.)] + Prove that the series 1 6 1.converges by Example 11.1100 4 + .1 ++ 2 2 4 3 8 4 1 6 5 32 .1 Test for the absolute or the conditional convergence of the following series 1 2 1600 2100 2 3 4 5 ++. The series ka 1 (l)k+l converges ~ by > 0 and the series (l)k+l 1diverges for a 5 1.4.* (2) 2 a . k2 1 Example 11.1 3&1 4&1 5&1 600 .3.4..5 C ukbk converges absolutely.4 @ Exercise 11.. One such property is considered in the next section.that Ic2 is the series converges absolutely because .4.4. Exercises. 1.2 I f u.. .8.+ .14 Example 11.294 Introduction to Mathematics with Maple 1 Example 11. Consequently 1. . k+lak @ Exercise 11. b.12 for a Let Q! E Q.  @ Exercise 11.
. Given E > 0. . we have k=l N k=l K are distinct integers all greater than N .. 4 . n m }2 (1. The rearranged series contains all the terms of the original series and no other terms. then C IS . .. why not? Exercise 1 1 . and explain why this example does not n 1 1 n 2 ? ) = contradict the Leibniz test. then the series Can. . The formal proof is as follows: We write E k . . [Hint: k=2 f i . and the sum of all other terms in which they differ would be small by the BolzanoCauchy Theorem. .5 Rearrangements If k nk is a bijection of N onto itself. ak = S N 00 00 1 and c a k = S .l k = 2 A. . If we go far enough. Now there exists an mo such that (721.l diverges and a k +0.is called a C a k . The idea of the proof is simple. . . . the series whose terms are given by 1 a2k1 = I7Gi1 IW+l Prove this. u l . partial sums of both series will have a lot of terms in common.N } for all m 2 m o m and then (121. namely the first N . but contains them in different order. Proof. . . . Writing om = k=l .8 Every rearrangement of an absolutely convergent series converges absolutely and has the same sum.Series 295 has unbounded partial sums and therefore diverges. . 6 1  The series 1 l 5 . I+ rearrangement of Theorem 11. *1 q3+1 1 a2k = that is.. choose N so that lakl < ..UK} for some K 2 0 where ui an. . .1.1 E!+1 1 1 +. x( 11. N ..E ..++. nm} = (1.SNI k=l k=N+1 2 < + E . Does this contradict the Leibnitz test? If not.
. We have Consequently the partial sums of (11.+ 1+ . We now show that its rear. that is o m +S as m + 00. Exercises..13) diverges. 12 4 3 6 8 5 10 1 2 7 (11.15 The series x (. k O0 (++I k=l 1.296 Introduction to Mathematics with Maple so and hence for all m 2 mo. .1 + 11.14) 11 +11 .+1. . Show that the series .14.1 + 1+ 3 2 5 7 4 9 11 6 1 3 and (11.1 Let S = 1 .15) .1 + 1.5.fi ample 11.13) are not bounded and the series must diverge. 0 converges conditionally by Ex Example 11.+ 1 + 11 1 . .angement (11. 1 . @ Exercise 11...1 ) " 1 .
If we know that a series converges we can than perform operations which will help in evaluation of the sum. Theorem 11. is absolutely convergent.17) 1 for n >_ N then the series diverges.16) or (11. then the series a . In Section 11.6 Convergence tests Since the sum of a series cannot often be found it is important to have criteria of convergence which will allow us to establish at least that the sum exists. if however. (11.8 we consider several finer tests of convergence. [Hint: For the series in (11.17) is satisfied if lim n+m/ % an 1 = Z < 1 or if . For the series in (11.Series 297 converge and have sums and respectively. s= c (k=l k=l O 0 1 . In this section we prove two simple theorems which are based on comparison of a given series with the geometric series. ] 11.1 2k1 2k ) ’ 1 +4k1 1 4k3 1 4k2 a) ’ and from the first of these 1 2 k=l Now add the last two series.16) for n 2 N . Condition (11. Do not forget to prove the convergence of all the series involved in whatever procedure you use.9 (Ratio or d’Alembert’s Test) N and a q such that %+I 1+4<1 If there exists an (11.14) show that the series S can be written in either of the forms 2s is.15) proceed in a similar fashion.
9.8 and 11. z.298 Introduction to Mathematics with Maple =I > 1. Perhaps it is worth mentioning that x 1 = 1Theorem 11.9.1 (d’Alembert’s Limit Test) then If C an converges absolutely if 1 < 1and diverges if I > 1.9 is still applicable.16) By an easy induction. a. an+l This shows that no conclusion can be drawn regarding the convergence or divergence of the series C a n from .16 For the series C n x n we have By Corollary 11. + 0.17) holds. then 0 < l a ~ 5 I laN+ll 5 l a ~ + a l and it is not true that a. We may assume that N E N and obtain successively from (11. whereas Corollary 11.9.4the comparison test. l a ~ + k I 5 ( a ~ l q ~ We . now have that the series C l a ~ + k l converges by Theorem 11. so here we prove the theorem. Hence C a k converges absolutely. The series diverges by Theorem 11.9. For 1 x 1 = 1the series obviously diverges. 1 nm lim = 1 regardless of a.10). respectively.1 this series converges for 1 x 1 < 1 and diverges for 1 x 1 > 1. 0 Example 11. Proof.1 is not. If (11. for 1 with cy E Q converges for a > 1 and Example 11.17 The series diverges for a 5 1(see Examples 11.2. This corollary follows in a straightforward way from Theorem 11. Hence we have Corollary 11.
1 If limn.18) for n 2 N . ?k 1 Inequalities (10.9.10 to assert convergence.then C a n converges absolutely. if m 2 1 for n 2 N .9 is not applicable.2. Since 1 i/an is 25 for n odd and 35 for n even. Moreover.47) actually show that whenever Corollary 11. We now give the proof of Theorem 11.10. = I . then (11.1.19) holds. C an Remark 11.Series 299 1 l 1 Theorem 11.3).. Since q < 1. Again we may assume that N E N. Proof. If (11.17 illustrates this again.1.10. Theorem 11. in Corollary 11.m converges absolutely for I < 1 and diverges for I > 1.19) Can diverges. we can apply Theorem 11. as an+l/an > 1 for an even n and an+l/an < 1 for an odd n. then Ian1 2 1for n 2 N . Similar comments apply to Corollary 2. the lim can be replaced by limsup for the convergence and by liminf for divergence test.10 (Root Test or Cauchy’s Test) an N such that If there exists (11. If (11. On the other hand. 0 The series Example 11.3 2 32 22 33 is obviously convergent as a sum of two geometric series.18 1 1 1 1 1 1++++++.the series k=N lakl converges by the comparison test an (Theorem 11.10. so is Corollary 11.7 As with d’Alembert’s limit test. and consequently + 0 is false and the series diverges by Theorem 11.1 is applicable. . then the series Corollary 11. The series from Example 11.9.18) holds then Ian1 5 qn 00 for n 2 N . no conclusion can be drawn regarding convergence or divergence if I = 1 .
after polynomials. The next theorem says.1 Test the following series for convergence: 11. However this goes far beyond the framework of this book. that the domain of convergence of any power series is always a disc in the complex plane. A power series represents a function (11. Series of the form C Cnxn are called power series.300 Introduction t o Mathematics with Maple Exercises Exercise 11. the simplest functions of a complex variable. .6. Power series are. roughly speaking. For the power series from Example 11.16.20) n=O The domain of this function is the set of all x for which the series convergesthe socalled domain of convergence. In complex analysis many theorems on polynomials are extended to power series.7 Power series We encountered the series C n x n in Example 11.16 the domain of convergence was the unit disc in the complex plane.
If 1 x 1 > p the series cannot converge because the sequence n I+ Cnxn is unbounded. by the comparison test. this is case (i). the series C(l)n+lconverges conditionally for both x = 1 n X2n c. In either case it is possible to choose T E S with 1 x 1 < r and we select such r. Let p = s u p s . It is customary to say that p = 0 in case (i) and that p = 00 in case (iii) . there exists a positive p such that the series converges absolutely for 1 x 1 < p and diverges for 1 x 1 > p. i xn converges absolutely for all complex x with and x = 1 and the series converges conditionally for x = 1 and diverges for x = 1. there exists K such that IcnxnI 5 K for every n E N ) . if S is bounded.2 for all x # 0. the series 1x1 = p = 1. \ Definition 11. The series C n x n diverges for all complex x with 1 x 1=p = 1.11is called the radius of convergence of the series C c n x n .8 power series for Nothing can be asserted concerning the behaviour of the 1 x 1 = p. 0 Remark 11.Series 301 Theorem 11. Let S = {x. Hence we established convergence in cases (ii) and (iii).11 (Circle of convergence) C anxn there are only three possibilities: (i) (ii) (iii) For a power series it converges only for x = 0. The convergence or divergence of a given power series for complex x with 1x1 = p is a rather delicate matter which we shall not attempt t o solve. Proof. it converges absolutely for all x E @. Then we obtain The series C cnxn now converges absolutely. .3 (Radius of Convergence) The number p defined in Theorem 11. Obviously 0 E S and if there is no other element in S then the series diverges by Theorem 11. Let now x be arbitrary if S is unbounded and 1 x 1 < p if S is bounded.
a~ R. (3) p = 00 if a = 0. For instance. or there is a positive p such that (11. Exercises.21) via the substitution y = xXO.7.11 says that the series (11. (Show that this is also true for lim Icn/cn+l I = 0 or o a ) . the analogue of Theorem 11. and investigate the convergence for x = f p .21) are also called power series. prove that the radius of nco convergence o f (11. (2) p = O if a! = 00. .302 Introduction to Mathematics with Maple The series of the form (11.21) converges for 1x .7. The number p is then the radius of convergence of (11.21) converges for x = xo only. Prove that the radius of con vergence p is given by: (1) p = 1 a i f O < a . All statements for series (1) lead immediately to similar statements for series (11.3 = a. @ Exercise 11.7. Give an example o f a power series with posnco itive (and finite) radius of convergence for which lim ( C n / C n + l ( does not n+m exist .xol > p. Let limsup n+co 0@ Exercise 11. or for all x. Exercise 11.2 I f lim J C n / C n + 1 J exists.zol < p and diverges for I z .1 Find the radius of convergence p of the following power series.20) is this limit.21). 20 is the centre of this power series.
5 If there exists K E R and m E N such that lcnl 5 Kn" for all n E N. J. Cauchy and then forgotten until it was rediscovered by another French mathematician.Series 303 This theorem was originally proved by A. .. These tests are based on comparison with another series known to be convergent.8 Comments and supplements To find the sum of an infinite series in a simple closed form is often difficult or impossible. are rather crude.22) 1 +1 1 + . The theorem became rather important afterwards.7. =log2 12 3 4 (11. We mention here a few more tests. prove that the radius of convergence of (11. Hadamard. 1 + 1 ++ 22 32 and that = lr2 6 (1 1. . It can be shown that 1.7. @ Exercise 11. 11. The slower the convergence of this latter series.1 and 11.2. the finer the test derived. .22) and (11. In Sections 11. (Perhaps we should mention that so far we have not even defined lr or the function x I+ logx. The tests we have so far.23) However. for the first part use the the comparison test. [Hint: The radius of convergence of the last series is 1 by Exercise 11. this cannot be done by merely finding the limit of the partial sums. For series with nonnegative and decreasing terms the following theorem is rather strong but has a simple elementary proof. and many other tests for convergence exist..8. namely the root and ratio tests. L.4 and Using the previous result show that the series kCkXk have the same radius of convergence.) We have already pointed out that in cases where it is difficult or impossible to find the sum of a series it is important to establish that the series converges.] 11.7.23).20) is a t least 1. CkXk @ Exercise 11.1 M o r e convergence tests. more theory is needed before we can attempt to establish (11.2 we found sums of only a few series. Prove also that the radius of convergence of the series C nmxn is 1.
Assume C ai converges and put C ai = S.304 Ti Introduction t o Mathematics with Maple 11 Theorem 11. This latter series converges if and only if 2l" < 1. I. Example 11.8. Cai are bounded 00 and this series con 11. We obviously have i=O Consequently the partial sums of verges.12 (Cauchy's condensation test) If a1 a3 2 . Then i=l 2 a1 + f k+l i=O C 2aa2i i=2 22 C 2aa2i. that is.10). 22a2i are bounded and this series con Consequently the partial sums of verges. Proof. 11. then the series C ai converges if and only i f does.9 and 11... if and only if a! > 1. Assume 2 a2 2 Pa2i 2ia2i converges. Applying Cauchy's condensation test yields the result that C n" converges if and only if 2n(1")does.19 We consider again En" for a E Q (Examples 11.2 0 . .
1.1 If an > 0 and ~ C ai converges.26) by the Comparison Test that the series C a k + l also converges.25) implies that nun 5 (n l)an+l and consequently ( N l ) a ~ + 5 l nun for n 2 N . We know from Example 11. and Hence the sequence n H nun is strictly decreasing after its Nth term and it is also bounded below.24) that nun 2 (r + n)an+l > ( n + l ) a n + l . which by the result of Exercise 11. Proof. then the series exists an N 1 E N such that Cui is convergent.24) r > 1 and N E N such that n ( 3 . It follows from (11.25) 1 for all n > N1 then Cui is divergent. If.13 (Raabe’s test) Let an > 0.13. 0 + + Corollary 11.1 > 0 it follows from (11.1 ) %+I >_r for all n > N . Turning our attention to divergence.Series 305 Theorem 11. however. By comparing x a k with a multiple of the harmonic series we see that the former series diverges. we see that Inequality (11. Consequently it converges. If this limit is less than 1 then this series diverges. The . It follows that there is a c such that 0 < c < nun for all n.8 that the series C l / n 2 converges. If there exists an (11.2 implies the convergence of the series Since T . there n(&l)<l (11.
There are many more general tests. + an. then it follows only that ai is not absolutely convergent.8. namely the Leibniz test. of course.. and we refer the interested reader to Knopp (1956) for Abel's. but indicate the main idea of the proof instead. There is a famous theorem due to the German mathematician B.ai E R and the series Cui converges conditionally.306 Introduction to Mathematics with Maple ratio test is not conclusive for this series.2 Rearrangements revisited. We had only one test for nonabsolute (conditional) convergence.14 (Riemann's gence) If A.. a.. Thus + + + an. Riemann which states that a conditionally convergent series with >g real terms can be rearranged to have any sum. hence we have a. + anz + . be used for testing absolute convergence of a series. For instance converges but n(la.. Theorem 11. and by continuing with this process of jumping up and down over A we obtain the required rearrangement..+ll . .. We add up the positive terms of the series until we first exceed A. + an./a. anz . such that Czl ant = A.. .GI) 5 1.1) = . . On the other hand.13can. > A > an1 + anz + .+ll... Theorem 11. > A... . + a.. This can be done because the series of positive terms diverges to 03. We also had examples showing that a rearrangement of a conditionally convergent series may diverge or have a different sum. There Xi We shall not prove this theorem.1) 5 1.l/la. Dirichlet's and Dedekind's tests. We note that this test requires the terms to alternate in sign and to decrease in absolute value. We know from Section 11. if however n(la. n(&+n[(+)2l] = 2 t 1> 2 n and Raabe's test yields convergence. .5 that absolutely convergent series can be rearranged and the sum is not altered. + .GI/(. + u n s . then there exists a rearrangement a. Then we add positive terms to exceed A again. c(l)"+'/& c + 11. Then we add successively negative terms until the sum is smaller than A for the first time.
.20 If we multiply the series 1 x x2 with itself and arrange the resulting series by increasing powers of x we obtain + + +. On the other hand we have the following corollary to the Riemann Theorem. (rn. One of the series C an. This is clear from the following: If an.9 The content of this theorem is easy to remember: one multiplies the series C a n and Cb.and d k = anbm then the series d k converges absolutely and (2%) n=l ( m=l F b m ) = F k =dl k  (11. C bn must be only conditionally convergent. = 1+2x+3x2+= . bn E R.. we have the following theorem. If we apply the same rearrangement to C(an zbn) then the rearranged series cannot have S as its sum. More precisely. by multiplying each term of the first series by every term of the second series and then arranges the series with terms anbm arbitrarily.n) H k is a bijection from N x N onto N. C an converges absolutely and C bn converges conditionally then the series C( an zbn) converges conditionally but by Theorem 11.27) Remark 11. (1 +x+x2 + )(l+x+x2 + .) l + x + x + x 2 +x2 +x2 + .15 (Multiplication of series) If ai and bi converge absolutely. Infinite series which converge absolutely can be multiplied together.8 the real part of the rearranged series will have the same sum as the real part of the original series.Series 307 For series with complex terms the Riemann Theorem is not valid. + 11.8. bn E R and the series C ( a n zbn) converges conditionally to S then there is a rearrangement of the series which either has a sum distinct from S or diverges. + Corollary 11.1 (Rearrangement of series with complex terms) I f an. + ~~ ~~~ Proof. Theorem 11. Example 11. hence by Riemann’s Theorem can be rearranged to a different sum.3 Multiplication of series.14.
x)2' It is often convenient to arrange the product series in the following form 00 ck = albl k=l + (azh a1b2) 4...15.20 the Cauchy product of C xk' with itself was C k x k .1)"+1 C cn where n + l J a d m2 n+ca Jn with This series diverges.Consequently 1+2x+3x2+= ~ 1 (1 .l . After these remarks and examples we turn to the proof of Theorem 11.albj is sometimes called the Cauchy product of the series C a n and C bm.21 itself is The Cauchy product of the series = (. This example shows that the Cauchy product of two conditionally convergent series may diverge.1.aab2 4. Consider a sum n=l ldll 00 00 n=l + Id21 + + * * * where each dk above is of the form anbm. Proof.(Qbi 4..308 Introduction to Mathematics with Maple for 1 x 1 < 1. The so called Merten's Theorem asserts that the Cauchy product of an absolutely convergent series with a conditionally convergent series is convergent and has the obvious sum. In Example 11. indeed3 consequently lcnl 1 2 for 0 5 k < n .aib3) 4 * . therefore lim cn does not exist and the series C c n must diverge. Let A = C Ian] and B = C Ibnl. + + + Example 11. * The series C cj with cj = ajbl ajlb2 . and 2 1. Then clearly 3By 2ab 5 a2 + b2 . For the proof see Stromberg (1981) or Knopp (1956). Let N be the maximum of these n and M be the maximum of such m.
3. 11. The first who realized that the harmonic series diverges was most likely Nicole Oresme (13231382).11. Archimedes considered only (simple) series with positive terms but for these his definition was equivalent t o Definition 11. The areas of rectangles resting one on another 1 1 2).. An example of an infinite product is in Exercise 11. . His definition anticipated the modern one given by Cauchy one and a half millennia later. Consequently 2’ 4 1 + 2 3 + 4.( 2. . (4’3) and leaning on the yaxis with top right hand corners at4 (1. It was already Archimedes who realized that the sum of a series must be defined and not be left as (intuitively) obvious. Johann Bernoulli (16671748) is often credited with the divergence of the harmonic series.5. Consider Letting n + 00 completes the proof of Equation (11. 1 1 are 1. .8.1. ++ 4The first rectangle is shaded in Figure 11.1 in two different ways.. Some other examples of successive operations on the terms of a sequence and the taking of limits are in Exercises 11.8.1 . He calculated the area T of the infinite tower in Figure 11. Successive multiplication leads to infinite products.l).29) 2 4 8 1 6 by the following clever trick.27). To find the sum of C d k it is sufficient to find a sum of a particular rearrangement..4 Concluding comments In conclusion we add a few historical comments. .9. his proof is close to ours given in Example 11.8. 0 We were led to infinite series by successive addition of terms of a sequence..4 and 11.8. He employed the inequality we used in Example 11. Oresme found the sum of (1 1.Series 309 The partial sums of C ldkl are bounded and hence C d k converges absolutely.
2”+’] are n2n. 2 4 8 1 6 Consequently  . = 2 . 11. 1 2 3 +++ 4 + 2 4 8 1 6 .. hence the area of T is also 1 2 3 4 T=++++.29) is easy. 5The second step is shaded differently in Figure 11.1 . The history of infinite series is interesting.1 Oresme’s tower On the other hand the areas of the ‘ ~ t e p sabove ’ ~ the intervals [2”. it suffices to put IC = 1/2 in Example 11.20 (and multiply by 1/2).310 Introduction to Mathematics with Maple Fig.. This example illustrates two points: the intellectual power of the mathematicians of the distant past and that the wisdom accumulated in mathematical theories makes problems easier to solve.. For us to find the sum of the series in (11. unfortunately we have to leave it now.
l ) ( a. Give 2+ a recursive definition of an. (a .) Exercise 11.2).2 1 @ Exercise 11. k E a(a .Series 31 1 Exercises. . . Prove that lim xn exists and find it.k k! + 1) N.1 For which values of a is the series convergent? The symbol for a E R. Exercise 11.4 Let a1 = 2. Prove that lim an exists and find it.8. ncc .5 Let 2 Give a recursive definition of xn.. Using Raabe’s test prove the convergence of ( 2 n .1) . Exercise 11.8.8.3 Prove that nmo i=l @ Exercise 11.. (The n+m sequence n H an is an example of a continued fraction.8. .8. Continued fractions play an important part in number theory and numerical analysis. a2 = 2 1 +2’ a3 = 2 1 + 7.
This page intentionally left blank .
If f has a limit 1 at 2 from the right.Chapter 12 Limits and Continuity of Functions Limits of functions are defined in terms of limits of sequences. The limit of f at 2 from the left is denoted by limf(x). > 5 it follows that f(xn) + 1. In this chapter we shall develop the theory of limits of functions. is said to have a limit 1 at 2 from the right if for every sequence n xn for which 2.x) (Figure 12. and particularly functions continuous on closed bounded intervals.1). + 2 and x.1 The symbols x J. Xla: ( Definition 12. we write limf(x) = 1. 12.1 A function f . with domf c R. study continuous functions.2 If the condition xn > 2 is replaced by xn < 2 one obtains the definition of the limit of f a t 2 from the left. 2 and x 313 2 can be read as “x decreases . At the end of the chapter we touch upon the concept of limit superior and inferior of a function. Formally we define: Looking at the graph of f : x (l X efinition 12. this definition is given here in terms of a limit of a function at a point. With a function f continuous on an interval we associate the intuitive idea of the graph f being drawn without lifting the pencil from the drawing paper. Mathematical treatment of continuity starts with the definition of a function continuous at a point.1 Limits w . xt2 J Remark 12. it is natural to 1x1 say that the function value approaches 1 as x approaches 0 from the right.
x ) = 1 because (1 x t o 1x1 lxnl xn . It should be remembered that Definition 12.2) ignores the value of f a t 2 . . that is. according to Definition 12. 1.5  Fig. and since lim(1 .5 0.1 (Definition 12.3 It is clear from Definition 12.xn)  If Xn + X 0 and xn > 0 then (1 Xn lxn I .5 1 0.xn) = + 1. lim (I XlO X 1 x 1 Xn x) = 1. . the limit from the beginning of this section. respectively. Remark 12.1 We now prove. Similarly.2) t h a t two functions f and g which differ a t most for x 5 2 ( x 2 5) have the same limit from the right (from the left) a t 2 or they have no limits a t all.1 that f has a limit from the right a t 0 even though f was not defined a t 0. 12.1.314 Introduction to Mathematics with Maple to k” and “ x increases t o 2” .1 (or 12.1 + 1 if xn 4 0.2 Remark 12.xn) = xn 1 for xn <0 It was natural to say in Example 12. And it still would be natural to say so if f was defined a t 0 no matter what f(0) was.1 Graph of f(z)= (1 1x1 X x) Example 12. (1 .
1 A function f has a t most one limit from the right at 2 (at most one limit from the left at 2).then the values of sequence x2k1 = l / k and 221. = 2/(2k saw(l/sn) alternate between 0 and 1/2.2 The function f : II: H saw(l/z) has no limit from the right a t 0. Consequently I = L (by Theorem 10. + + x:To X Fig. If I and L are limits of f at 2 from the right. l/n] of the xaxis.) A partial graph of f is sketched in Figure 12.1 and 12. It consists of triangles of height 1/2 whose lateral sides are slightly curved' and base is formed by the interval [l/(n l). then by Definition 12. 0 The great advantage of Definitions 12.Limits and Continuity of Functions 315 Example 12. (The saw function is defined in Exercise 4. for k E N.1 f(zn)+ I and also f(zn) + L.2 is that they make it possible to use the already established theory of limits of sequences (see Chapter 10) for limits of functions. 'They are congruent to pieces of the graph of z I+ 1/x .1).6.1 and we shall exploit it further soon. Proof. This was clearly exploited in the proof of Theorem 12.2 Graph of f(z) = saw(l/z) Theorem 12. we state two more definitions first. However. if we choose to take as points of the I). Obviously the sequence n H f(zn) does not converge and the limit limsaw(l/z) does not exist.2.8 in Chapter 4. zn + 2 and zn > 2 . Clearly. 12.
1.4 then A function f has at most one limit at each o f 2 .5. We first write x + d z= xx2 . with x.2 +oo and 00.= 0. 1 lim . Example 12. x+x xlx xtx zt+Oo lim f(x) or lim f(x) can be replaced by any other letter without chang2400 ing the meaning of that symbol.5. We also may make the distinction verbally by saying t h a t the limit is in the real or the complex domain.3 A function f is said to have a limit I at 2 if for every sequence n H x.5 We find lim (X X+00 + d x 2 x). The meaning is changed according to whether x.3 is replaced by the symbol +oo or ca (and the inequality x. If Xn + a then G + +m and Theorem 10. x+m The letter x in any one of the symbols lim. EC we shall explicitly say so. # 2 and x. Example 12. + li. The proof is simple: if xn + 2 then f(x. If nothing is said to the contrary we shall assume that x. # 2 omitted) then one obtains the definition for the limit off at +oo or m respectively.3 x2 lim x2 = 4.IXldrn'* last expression is equal to d 1 for x < 0.4 We left the wording of Definition 12. l+Jm + oo then . it follows that The limit of f at 2 is denoted by lim.316 Introduction to Mathematics with Maple Definition 12. Example 12.x . The limit of f at +m or oo is denoted by xli~OOf(x) or lim f(x) respectively. limf(x). If domf c R and the symbol 2 in D e f i n i t i .) = xi converges to 4. For example lim t2 = lim h2 = 4.x2 + x This cable. Again the proof is simple: if xn ++oo x++m Xn 1 x + 0 by Exercise 10. limf(x). t2 h2 The next theorem can be proved in similar way to Theorem 12. 25 Remark 12.f(x). E R and if z .3 intentionally a little ambiguous.4 12. E R or C. Theorem 12.f(x). 6efinition 12.& r T . Now if x.6 concerning the limit of a sum is not appliX .
x + l 22 . As a sample we prove (iii). Let x.6 for the limit of a product of sequences we have f(xn)g(zn) + Z k which in turn implies by Definition 12.f (z). then f(xa) + 1 and g(xn) +k by Definition 12.4 x2 Theorem 12.5 An important case of assertion (iii) is g(z) = k for all [f(x)] = x. xx lim Min(f(x).4 that x+x limA f(x)g(x) = Zk. # 2 . k ) . k). Example 12. 22 X+2 22 Remark 12.7 In the previous example we used an analogue of Remark 12.4) = 0. All assertions (i)(vi) follow directly from Definition 12. Proof.3 directly since lim (x2 .3: if two functions differ a t most when x = 2 and one has a limit 1 a t 2 then so .lim.Limits and Continuity of Fzlnctions 317 1/x.6 We wish to find lim x+2 x2 3x x24 + + 2 .3. 22 (iii) lim f(x)g(x) = Zk.3 and corresponding theorems for limits of sequences. g(x)) = Min(l.3 remain valid if (a) 2 is replaced everywhere either by +oo or by m. For k = 1 we have limA x+x . Then lim kf(x) = k lim f(x). 22 (iv) if k (v) if Z 52 f(x) = 1 # O then limAg(x) k' > 0. r E Q then x+x lim [f(z)]' = Z'.3 If limA f(z) = 1 and lim. Since x+2 for x # 2 we have lim x+2 x2+3xt2 x+l = lim 22 1 x+2x2 4' 1 Remark 12.g(x) = k then X+X x+x (i) 52 lim lf(4 = 1% (ii) lim(f(z) f g(x)) = 1 f k .6 (b) x All assertions in Theorem 12. 22 (vi) limA Max(f(x). + 0 and 1 l+d + 1 Consequently lim (x X*oo 2' + d z )= Theorem 12. 112. + 2 is replaced everywhere by either x I 2 or by x T 2. 0 Remark 12. Using Theorem 10. g(x)) = Max(Z. + 2 and z . We cannot use (iv) of x2+3x+2 .
1) <I x . In order t o apply the theorem we note that. # 2 then there exists N such that 0 < I x .8. + 2 and x . 2 1 < 6 for n > N .6 with x replaced by 3x2 and n = 10.4 shows that the limit is 3/10. More generally. + . Remark 12. and limit is replaced by limit from the right (limit from the left) .8 Theorem 12. If x.2 1 < A then lim f(x) = 1 implies that limA g(x) = 1.7 y Y W 1 22 . This is clear from the definition of a limit.?I < S and lim g(z) = lim h ( z ) = 1 then f has a limit at 22 2 and lim f(z)= 1. h are automatically assumed to be real valued.3. x+x 22 The Squeeze Principle (see Theorem 10. by Exercises 5.4 If there exists S > 0 such that g(z> L f(x) 5 for O (12. Theorem 12. the functions g.5 1 < S is replaced by 5 < x < x S (x .) f f(%) L h(xn) 242 > N and it follows from the Squeeze Principle for sequences that 1.318 Introduction to Mathematics with Maple does the other.5 and 5. We wish t o calculate lim x+o 0 Example 12.b < x < 2).10' Now simple application of Theorem 12. if there exists A > 0 such that f(x) = g(x) for 0 < 1x . The proof follows the same pattern as the proof of Theorem 12. f.8. XKt x+x Since the formulation of the theorem involves inequalities. That in turn implies that lim f (x) = 1. Consequently g(xn) 5 for n f (x.3) is also valid for limits of functions.4 remains valid if (a) the inequality 0 < lx . ConsequentIy 3 10y/ 1qiT32l < 3 22 . Proof.
Then there exists n such that x. > limit(2*x+floor (x) . twosided so to speak.x=3. limf(l/y) = 1.5 z++m We have l i m f ( l / y ) = 1.5. Assume that lim f(x) = I and let yn > 0 and yn z++m + 0. I.right) . direction can be left. The following examples illustrate some possibilities.1.) + 1.5 and consequently f(l/y. 2 > limit(x/abs(x) . However f(l/y.Limits and Continuity of Fzlnctions 319 (b) one assumes the existence of K (k)such that inequality (12.1 Limits of functions in Maple Using Maple for evaluating limits is easy. right or complex. relevantpoint can be infinity or infinity as well as some numerical value. Assume limf(l/y) = 2 and let x n Y10 4 + I . then direction should be omitted. If the limit is in the real domain in the sense of Definition 12. +oo. The form of the command is limit (fvalue. fvalue stands for the value of the function for which the limit is sought.x=relevantpoint .x=O.left).direction) .) 11.5 and therefore f(l/y. For n > N set yn = l/xn. > 0 for n > N . There is a simple relation between the limit at +oo and the limit from the right at 0. Theorem 12.3. 0 12.) = f(xn) and hence f(xn) + 2 which proves z lim 4 t m f(s) = 1. + Y10 +oo by Exercise 10.5. Then l/y.1) is valid for x > K (x < k) and limits are replaced everywhere by limits at +oo (4. that is. 1 . YlO lim f(x) = 1 if and only if Proof. Then yn + 0 again by Exercise 10.
Exercises Exercise 12. (2) lim x2 x24 ’ 2x3 .x=inf i n i t y ) .3 x . it follows from the Squeeze Principle that.1 for x > 1. lim X+OO floor(x) 0.320 Introduction to Mathematics with Maple 1 > l i m i t ( x + s q r t (xn23*x) .x=inf i n i t y ) .2x2 i x . (3) lim x+l x 3 . 3 2 Sometimes Maple needs a little help > l i m i t ( l / f loor (x) .1. 1 Now there is an obvious question: is this limit exactly O? Since 1 o<<  1 1x1 . indeed. x24 ’ 2x2 . (1) lim z1 Find the following limits by paper and pencil method 2x2 . the limit is exactly 0.32 1 .X Q .32 1 .x .3 ’ (4) lim x1 X2& + + &1’ .1.1 and by Maple.
7 (4) lirn x Iq(x).Limits and Continuity of Functions 321 (5) x+1 lim (x + l)d2 x 221 7 . X++oO ª9) lim + + (x + 1)5 x +2 2 7 2)5 + + . h+O lirn f(2 + h ) = I. + ( x + 1000)5  7 x++oO Exercise 12. . Is this statement true if limits are replaced everywhere only if limA by limits from the right? From the left? [Hint: No.8.2 20 Use the Squeeze principle to prove: (1) lim Ll/xjx = 1.5.] task is to prove 3 Exercise 12.3 Y+Y If a # 0 and ij = a2 + b then x+x lim f ( a z + b) = 1 if and f ( y ) = 1. + IzllxJ X (3) lim x+o lVE21.1. State similar .1.5 @ Exercise 12.1 3 Exercise 12.1. 240 lim x+o Ye21. 7 (6) lim (7) JiT71. but remains true if a > 0.4 If a > 0 then lim f ( a x X++m What happens if a < O? 2++oO + b) = 1 @ y++m lim f ( y ) = 1. (8) lim 20 ymqi=%* 22 7 x5 10020 (10) lim x (&?Ti . y’m x+x lim f(z)= I @ theorems for the limit from the right and for the limit from the left.x). . (2) x+m x2 1 lim = 1.1. for (4) use I z l ~ ( z ) 5 I The following exercises are stated as theoremsthe them. Izl. x+o [Hint: For (3) use Exercise 5. 9 Exercise 12.1.6 lim f(x) = I @ lirn f(y) = 1.
1 E R and E . In other words. Theorem 12. Sometimes it is a little inconvenient to prove that f(z) + l for every sequence n I+ xn with x . Assume f has limit I but that it is not true that for every E > 0 there exists 6 such that the implication (12. which is similar to the EN definition of a limit of a sequence.3 is that the theory of limits of sequences carries over easily to limits of functions. @ Exercise 12. and consequently .2 1 < 6 * If(. xc 12.11 < E for n > N . S are positive numbers then we call the set an epsilondelta box centred at ( 2 .f(2))lies in the €4 box centred on ( 2 . This inconvenience is often overcome by using Theorem 12.3).6 below.8 If lim f(x) = 0 and there are positive 6 and M such x+c that 1g(x)l 5 M for O < I x .3). Let z n + 2 and x n # k . I. l ) (or briefly.lI < E .cI < 6 then lim f(x)g(x) = 0. l ) (see Figure 12.2) Proof. and in particular 6 = 1/n for n E N is not suitable.l f E ) .322 Introduction t o Mathematics with Maple @ Exercise 12.2 The Cauchy definition The great advantage of Definition 12. xx 11. This proves f(xn) +1 and hence limA f(x) = 1. (12. The function f has a limit 1 at 2 if and only if for Theorem 12.1. Let E > 0 and assume there exists 6 > 0 such that the implication (12. The graph of f enters the ES box at the left vertical side and leaves it at the right vertical sight of the box. then there exists N such that 0 <n 1 . +2 and xn # 2. an E6 box: see Figure 12. for every n E N there exists a number Z n If(?) . This €4 box is the interior of a rectangle with vertices at (2 f 6.2) holds. l ) with the possible exception of ( 2 .6 below states that f has a limit 1 at 2 if for every E > 0 there exists 6 > 0 such that the graph of f near the point ( 2 . This happens for every positive & no matter how small E is.1. This means there exists EO > 0 for which there is no suitable 6. 2 1 < S.2) holds.) .6 every E > 0 there exists 6 > 0 such that 0 < 132 . If 2.7 limf(x) = lim f ( 2 zlx t++m 1 + $.
2 1 < 1/n but I f ( x n ) . Hence I is not the limit of f at 2 .6 with the ES box makes sense only if domf c R and I E R. . respectively.6 Remark 12.10 Theorem 12.11 The modification of Theorem 12. + Remark 12.2 1 < S is replaced by the inequality 2 < x < 2 S and 2 .6 remains valid for the limit from the right and the limit from the left if the inequality 0 < Ix . However the theorem itself is valid in the complex domain as well.1.2) holds. a contradiction.9 The interpretation of Theorem 12.I1 2 EO. Remark 12.3.3 Interpretation of Theorem 12.2 but the sequence n H f(zn) does not converge to I . and so there must exist a 6 for every E such that the implication (12. 0 Fig. Now we have a sequence n H xn such that xn # ? and xn + ? by the Squeeze principle.6 < x < 2.6 for limits a t +oo or oo reads: f has a limit I a t +oo or oo if and only if for every E > 0 there exists K or Ic such that respectively. More precisely by Corollary 10.Limits and Continuity of Functions 323 such that 0 < lxn . 12.
Consequently convenient3form: (x+l. The proof that limf(x) = I is similar. and consequently limf(z) = 1. . Often it is also called the ~ . in conjunction with Remark 12. More precisely they say: ‘f has a limit 1 by definition if for every positive E there exists a positive S such that implication (12.6. 11.324 Introduction t o Mathematics with Maple Example 12.2) holds. Theorem 12. makes the following theorem fairly obvious.10.5) but the situation is simple enough for some elementary algebra. I( Theorem 12.I = x3 1 and we rewrite this expression in a more 3(x+ 1)2+3(x+1). (12.8 As an illustration of Theorem 12. 2 2 Let I be an open interval and f : I t+ R has a limit at 2 if and only if lim f(z)= lim f(z)and then lim f(z) = lim f(z) = XcTX XlX 1 xt? xi? Proof. If lim f(x) = I then for every E > 0 there exists S > 0 such that 5 2 that is. + + + + + + + + + + + x+1 + + Some authors use the implication (12.) .s1 < x < 2 * If(.3 11 5 7s < &. Definition 12.3 is named after the German mathematician Heine. I. (12.3) and 2 <z <2 + 62 * If(.1)and we have 1 < 6 +1.11 < E . If limf(z) = limf(x) = 1 then for every E > 0 there exists 61 > 0 and Xt? X@ XlX xtx 62 > 0 such that i .6 we prove t h a t lim z3= 1. In this case f(z) .’ Such a definition is often referred to as Cauchy’s definition of the limit.1)3 1 = (x+ 1)3 1 z 3 1 15 I z 11(1z 1 1 2 312 1 1 3). For E > 0 we can choose S = Min(~/7.7 lim f(z).2) as the definition of a limit of f at 2.6 definition of the limit.) .4) 3This really amounts to the use of the Taylor polynomial (see Section 6. if Iz 1 1 < 1 then 1 z 3 1 1 5 712 1 1 . Therefore.11 < &.
if lim.&). Sz).6 we prove: Theorem 12. then if 0 < I x2 1 < A then also 0 0 < Iz .g(x) = k > 1 then that f(x) < k when 0 < 1 25 there exists A > 0 such that g(x) > I when 0 < IJ: .L i m i t s and C o n t i n u i t y of Functions 325 respectively.) by (12.2 1 < 62.4). f (4 < g ( 4 when 0 < Iz . I+k 2 (12.& < g(x).6) Let A = Min(61. 0 Corollary 12. . and then by (12. We choose kZ 0 L .1 If lim f(z) = I < k then there exists A > 0 such x+x x2 1 < A. Also.g(x) = k and Z 22 Let r g f < k then limA f(x) c R. There exist 61 > 0 and 62 > 0 such that O<~x~~<sl~f(x)<z+&=2 l+k (12. If 22 there exists A > 0 such that cR and rgg = 1.61 < x < 2 or 2 < x < 2 62.2 1 < A.8 lim.2 1 < 61 and follows. Particularly important cases of this corollary are I = 0 or k = 0.  ZI < E 0 As another application of Theorem 12.7.8.6) < Iz . if the limit of f is positive at 2 then f is positive near 2 too (but not necessarily at 2 itself). &= Proof.3) or (12.5) and (12. The proof is analogous to the proof of Theorem 10.2 1 < A. Then we have: if 0 < I x2 1 < S then either 2 . and in either case we have + If(. Let S = Min(&.5) and 0 <I x2 1 < 6 2 * = k .
The proof that lim f(z) = sup {f(x).10 (Limits of monotonic functions) I f f :]a. Hence (12. Bounded monotonic sequences always have limits.8.then I 5 k. Let E > 0. Theorem 12. We shall assume that f is increasing: if it is not.b [ } xtb is very similar. Proofs of these are left for the exercises.9 If lim f(x) = I and limg(z) = k andifthere exists X+2 x+x a positive u such that f (4 I 9(4 for O < Iz .8 and 12. Analogues of Theorems 12. at +oo and at 00. Assume contrary to what we want to prove that I > k.8 (with the roles of f and g interchanged) f(z)> g(z) for 0 < Iz .2 1 < Min(a. (12. Proof. A).2 1 < u.2 1 < A for some A > 0.z € ] a .326 Introduction to Mathematics with Maple Theorem 12.9 as well as Corollary 12.7) Proof.7) does not hold if 0 0 < Iz .b[ such that . from the left. and this is a clear contradiction.b [ } .x € ] a . A similar theorem is valid for functions.1 are valid for limits from the right.b[~+R is monotonic and bounded then exist. The Cauchy definition is a convenient tool for the proof. the theorem for an increasing function can be applied to f. We prove that the first limit is equal to I = inf {f(x). By the definition of the greatest lower bound there exists 2 € ] a . Then by Theorem 12.
for all 11: with a < J: < a + S.11 The function f has a limit I at x + + O O 00.) X+oo lim f(z)= I if and only . For example limsgn(z) = 1 and limsgn(z) = 1. Conse+00 or 00: Theorem 12.cI/lxcl 2111: .cl/c2 for 1 x 1 > Ic/al.6 prove that + 3) = 2%+ 3.12 if for every E For the limit a t 00 we have > 0 there exists k such that x <k + If(.2. for (3) use the inequality 1 1 / 1 1 :. x4c [Hint: For (2) use the same trick as in Example 12.6 for limits at A monotonic function.(fI Remark 12. (2) ai_m. ).] < < Exercise 12. need not have a limit.2. if and only if for every E > 0 there exists K such that x > K + 1.2 Using Theorem 12. Exercises Exercise 12.11 < E .I1 < E. We now set K = 1/S and x = l / y and see that the above implication is 0 equivalent to 11: > K + . 1 I f(z)I f(z)< I + E. lim f(x) = 1.7. by monotonicity of f.1 (1) h ( 2 s 52 Using Theorem 12.Limits and Continuity of Functions 327 For g = a + 6 we now have.l/cl 1 x .5 lim X++OO f(11:) = 1 if and only if limf(l/y) = Y10 This latter equation holds if and only if for every E > 0 there exists 6 > 0 such that 0<y < 6 + Jf(l/y) . although having limits from right and left. (3) lim 1/x = 1/c provided c # 0. 0 xto 210 quently there is no limit at 0. Proof. lf(4. By Theorem 12.11 < E .x2 = u2. There is also an analogue to Theorem 12.6 prove that .4 < E .
2.S.] The following exercises are stated as theorems. State and prove similar theorems for the limits from the right.7 K such that If lim f(z)= I .8 If x E Q let x = p / q with p E Z and q E N and p and q relatively prime (the only positive integer dividing both is 1).2. (see Exercise 3. @ Exercise 12. @ Exercise 12.1 and limlzJ = n ( 5 ) lim(1zJ x+7 + LxJ) = 1.9 remains valid if x + k is replaced by x 2 1 < o by 2 .7. @ Exercise 12. then there State and prove an analogous theorem for the limit from the left.3 If limf(x) Xla: exists S > 0 such that = 1.5 If lim f(x) = I then there exists 6 bounded on ( 2 .1 ztn xtn for n E Z.3 for the definition of sgnz) z t o 210 (2) limsgn z = 1.2. limg(z) = k and I 212 < k . + @ Exercise 12. (4) limlzJ = n .2. Define .2 6). from the left and for limits a t +w or GO.328 Introduction to Mathematics with Maple (1) limsgnz = 1.2.4 exists K such that If lim f(z) = 1.2. (3) limlxj = n xln xln and 1imLzJ = n .o < x < 2 or x 2 or by x 2 and the inequalities 0 < I by 2 < x < 2 CT respectively. z++m lim g(x) = k and if there exists X++W then 1 5 k. + x+x @ Exercise 12.6 Theorem 12. for n E Z. [Hint: For ( 5 ) you can use the results of (3) and (4) together with Theorem 12. lim g(x) = k and 1 < k then there 2++00 z++m x >K + f(z)< g(x). 0 Exercise 12. > 0 such that f is State and prove an analogous theorem for the limit at w. but a direct proof is also easy.2. The task is to prove them.
while Limits from Definitions 12. 12.3. Iff has a limit +00 from the right at 2 . particularly if one wants to emphasize that the limit in question is not an infinite limit. at 2 . Generally speaking.= 00. 200 x+o lim f(x) = 00.= +oo. 210 x xto x or infinite. the 1etter.Limits and Continuity of Functions 329 f a s follows: f(x) = I/q if x E Q and f(x) = O if x lim f(x) = 0 for 2 E Q. 5+OO lim f(x) = 00. x+x # Q. limit at 2. A similar notation is used for the X+oo . The following definition covers the possibilities of a function having +oo or 00 as a limit. lim.Z denoting a real number is replaced by the symbol +00 or oo one obtains the definition of an infinite limit. x+o 2 2 1 1 lim . no .9 Clearly lirn x = +00. limit from the left at 2. at $00 or at 00.3.infinite limit 00.does not exist.= +00 but lim . /Definition 12.3 Infinite limits Functions.3 are x+x discussed in Exercises 12.3.5 If in the definitions of a limit from the right at x.3. For instance. we write limf(x) = +00. 1 1 lim. lim f(x) = +00 512 xT2 245 x + + O O 1 or lim f(x) = +00. Some extensions of Theorem 12.7.4.5 and 12. limf(x) = +00. Prove that 12.112. For the function f shown lirn f(x) = +00.4 are sometimes called finite limits. but not all can and care is needed. if limA f(z) = +00 then lim af(z) = +oo if a X+? > 0 and lim. finite 20 x Many theorems concerning finite limits can be extended to infinite limits. like sequences. respectively. limit at +00 or limit at 00. Example 12.a f ( x ) = 22 x+x 00 if a < 0 and ob viously lim*uf(z) = 0 if a = 0. may have infinite limits. lim f(x) = +00. x++00 lim x = X+oo 00. 2+m lim x2 = +00. Typical behaviour of functions which have infinite limits at 2 is sketched in Figure 12. from the left at 2 .
Then limA f(x) = +m if and only 22 if for any K there exists 6 > 0 such that 0 < 2 1 < 6 + f(x) > K. By (12.6 4 . Take any K and assume that there exists 6 > 0 such that (12. For questions of this kind see Exercises 12.3.8) Proof.7. .330 Introduction to Mathematics with Maple .4 Behaviour of functions with infinite limits conclusion can be made concerning limA(f(x) g(z)) if lim f(z)= +oo 22 22 + 22 and limA g ( x ) = m.8 . 2 1 < l / n but f(xn) 5 KO.4.f(x) = +oo but that for some KOthere does not exist 6 > 0 22 such that (12. + ? but .2  2 4 6 8 1 0 1 2 Fig. # 2 and x. This proves that lim f(z)= +oo. Assume lim.8). f(xn) > K for n > N . 12.6 and 12. Then for every n E N there exists a number x. (12. + 2. 22 11. An analogue of the Cauchy definition for infinite limits is the following: Theorem 12.8) holds. Let x .12 Let rg f c R. I. Then x.8) holds. # 2 and x. such that 0 < l x . then there exists N such that 0 < Iz.3.3.2 1 < 6 for n > N . 12.
7’ x3 1 (8) lim x+m 2 2 + 52 .J) = +GO. + Remark 12. which is a x+x contradict ion.14.1 (1) lim Find the following limits.5 2 .a  1 (2) lim (3) (4) lim (5) X T d w 2 . The Exercise 12.f(x) = +oo is false. from the left. and limits a t +oo and Example 12.12 that if and limAf(z)= +oo then limg(x) = +oo. x 1 ~ x2 2’ 1 .13 Theorem 12.6 < x < 5 and by 2 < x < 5 S respectively. It follows from the J definition of the floor function that x 2 1x1 and therefore 22 . result then follows from Example 12.7 ’ ZTO 2 n ’ 1 + + + (9) .n E N . 0 Remark 12.1. limits from the right.Limits and Continuity of Functions 331 f(xn) + +oo is false.12 remains valid for the limit from the right 1 < S is replaced by and for the limit from the left if the inequality 0 < Ix . Exercises 2 x.2’ 2x+1 st1 23 + 1’ lim xn. Consequently lim. (6) lim  lim .9 and Remark 12. This remains valid for x+x x+x 00. 210 x n ’ x3 1 (7) lim x++oO x2 52 .3.10 We have lim (22 x++oO 1. X+oO n E N.14 f 5g It is an immediate consequence of Theorem 12.
5 6 such that 2 < x < 2 + g(x) 2 m then limj(x)g(x) = +oo. Also prove this for lim and lim. [Hint: 2f(x) = IF(x)I + F ( z ) + g(x) = xx Deduce that lim(f(x) Prove: If lim f(z) = +oo and there are positive m and 6 XTX XT2 F ( 4 . Deduce that lim(f(z) g(x)) = +oo if lim f(z)= + XTX xtz + ztx +oo and limg(z) = I or +oo.6.332 Introduction to Mathematics with Maple 1 @ Exercise 12.312. 2 ) . [Hint: Use Exercise 10. it can equal anything. or xtx 1 or need not exist at all.8 State analogues of Exercises 12.6 < x < 2 then lim= +m. What XTa: XTj.1 @ Exercise 12.3.3.1 @ Exercise 12.6 < x and limg(x) = +oo then limf (4 = 0. 00.3. xf2 @ Exercise 12. limits from the right. limg(z) = oo and + g = F.4 f Let F be an arbitrary function with dom F = (00. Find functions f and g such that limf(z) = +oo.5.] fo' <2 for @ Exercise 12.3. limg(x) = 0 and f g = F.. can you say about limf(z)g(z)? [Hint: f(z)= IF(4I + 1 9 ( 4 = F(4 X t2 2 .2).7 Prove: I f f is bounded on some interval 2 . XTX >0 f (4 @ Exercise 12.6 Let F be an arbitrary function with dom F = (Loo. Formally. + g(z)) can be any number. 12. +oo. we make the following definition: .3.3. xt2 XTX @ Exercise 12.5.2 xtx xlx Prove: lim = 0 5 2 f (2) * x+x lim lf(z)I = +oo.3.3 Prove: If lim f (x) = +oo and g is bounded below on xtx some interval ( 2 .7 for limits. and limits at +oo and GO.3.4 Continuity at a point A function f is continuous at a point 2 if it is possible to find an approximate value of f(5) from an approximate value of 2.f(4. and if moreover f(x) xtx XTX 9 ( 4 g(x) 2 . 2 ) then lim(f (x) g(x)) = +a.3.X The limit need not exist but if it does. Find f and g such that lim f(x) = +oo.
Example 12.412. We do not do this in this book and.4. even in situations where infinite values are allowed. More generally. If f is discontinuous at 2 and limA f(z)exists. For example the floor function is discontinuous from the left at any integer and the function ceil is discontinuous from the right at any n E Z.4.6.8 the function x H x3 is continuous a t 3. xx . the function is considered to be discontinuous where lim f(x) = f(2)= f o o . xln Usually. for instance polynomials.Limits and Continuity of Functions 333 Definition 12. However. the greatest integer function is continuous from the right a t every point and the ceil function is continuous from the left a t every point. we say discontinuous instead of not continuous. The proof of this is beyond our reach at this stage. In mathematics it is sometimes convenient to consider functions which have values 00 or 00. every polynomial is continuous a t every 2 .6 the limit is understood as a finite limit. This is because the definition of f can be amended by declaring f(2) = lim.11 By Example 12. the discontinuity at 2 is 22 called removable. in particular. continuous from the right at 2 or continuous from the left at 2 if (respectively. The exercises contain examples of functions continuous at one point only. today some functions discontinuous at many or even all points of the domain of definition play an important part in mathematics and in applications. However. Some examples of functions with discontinuities are discussed in Exercises 12.f(x) and thus making the function continuous at 2. 22 In the past. mathematicians believed that for a function to be continuous everywhere was normal and to be discontinuous was an exception. in Definition 12. To a certain extent this is somewhat tenable even today in the sense that most functions encountered in applications are continuous apart from a few points.6 (Continuity) A function f is said to be continuous at 2. Obviously there are many functions continuous everywhere. the floor function is obviously continuous from the right a t n E N since liml(z)J = n = LnJ and similarly the function ceil is continuous from the left a t n E N. It is an interesting fact that there is no function continuous at all rational points and discontinuous at all irrational points. functions discontinuous everywhere and functions continuous a t all irrational points and discontinuous at all rational points.
334
Introduction to Mathematics with Maple
The changed function is, strictly speaking, a new function but it is customary to keep denoting it as f and identify it with the old f . Understanding of this point sometimes helps in dealing with some computer system algebra. For instance in Maple substituting x = 2 is, quite naturally, not allowed in h(x) = (x  2 ) 2 / ( x 2  5x 6) but the Maple solution of h(x) = 0 is x = 2. Maple automatically removes the singularity at x = 2 by setting h(2) = lim h(x) and solves the equation after that. x+2 Almost all statements about limits of functions translate into statements about continuity. Here we illustrate this by several corollaries to theorems or definitions which we need for future reference. One corollary to Theorem 12.7 is:
+
Corollary 12.7.1 (to Theorem 12.7) A function f is continuous at 2 i f and only i f it is continuous both from the left and from the right at 2.
The Cauchy definition of a limit translates into the Cauchy definition of continuity as follows.
Corollary 12.6.1 (to Theorem 12.6) f is continuous at 2 if and only if for every E > 0 there exists S > 0 such that
I x 2 1 < S 3 If(x)  f(?)I
< E.
(12.9)
It is a detail but an important one: the inequality 0 < I x2 1 from Theorem 12.6 is now omitted. The Corollary remains valid if the word continuity is replaced by continuity from the left or continuity from the right and the inequality I x 2 1 < 6 by 2  6 < x 5 2 or by 2 5 x 5 2 6, respectively.
+
f is continuous or continuous from the right or continuous from the left at ? i f and only i f for every xn + 2 or xn + 2, xn 2 2 or 2 , + 2, 2 , I ?, respectively, it is true that f(2n) + f(2).
Corollary 12.2.1 (to Definitions 12.1 and 12.2)
The only if part is not entirely obvious since now 2, = 2 for some n is possible. However the above Corollary can be proved using the Cauchy definition of continuity and the method of part I1 of the proof of Theorem 12.6. The next Corollary which is a counterpart of Theorem 12.3 is very useful for an instant recognition of whether or not a given function
Limits and Continuity of Functions
335
is continuous at some point. This helps in graphing functions in Maple. To obtain a correct graph of a function with a discontinuity use the option discont=true. The following Maple session illustrates this.
> >
plot(signum(x1) ,x=1. .3); plot(signum(x1) ,x=1. .3,discont=true);
I
Fig. 12.5 Muple plotting discontinuities
Corollary 12.3.1 (to Theorem 12.3) Iff and g are continuous (continuous from the right, continuous from the left) at 5 then so are the functions
336
Introduction to Mathematics with Maple
Corollary 12.8.2 (to Theorem 12.8) Let rg f c R and rg g c R. Iff and g are continuous at 3 and f(2) < g(3) then there exists a positive A such that
f (4 < d4
when Iz  3 1 < A. Remark 12.15 Important consequences of this Corollary are the cases g are constant. For instance, if f(2) < A 4 or g(2) > 0 then f(x) < M or g(z) > 0, respectively, for I z 3 1 < A. Similar statements hold for continuity from the right and from the left, for instance if f is continuous 4 from the right a t 3 and f(2)< M then there is a A > 0 such that f(x) < A for 2 5 x < ? + A .
when f or
A very important tool for establishing continuity is the next theorem on the continuity of composite functions. It says that a composition of two continuous functions is continuous. More precisely, we have
If g is continuous at 2 and f is continuous at Theorem 12.13 ij = g(2) then f o g is continuous at 3.
Proof. Let x, + 3. By Corollary 12.2.1 and continuity of g at 3 we have that y , = g(x,) + g(3), by the same Corollary and continuity of f at ij we also have that y , + f ( i j ) . In other words if 3, + 3 then 0 f o g(x,) + f o g(3). This, by Corollary 12.2.1, proves the theorem.
Example 12.12
A polynomial P is continuous everywhere and the square root function is continuous a t every ij > 0. Hence the function x :H is continuous a t every 3 for which P ( 2 ) > 0.
,/m
Theorem 12.13 is important and has a simple wording and a simple proof. In contrast t o theorems about continuity considered so far it is NOT valid if continuity is replaced by continuity from the right or left throughout: see Exercise 12.4.7. Also, in contrast t o previously stated theorems and corollaries it does not translate directly to limits: see Exercise 12.4.8.
Remark 12.16
Limits and Continuity of Functions
337
Exercises
@ Exercise 12.4.1
I t follows from Theorem 12.3 that a polynomial is continuous at every point y E R. Give a direct E  &proof. [Hint: P ( x ) P ( y ) = q(x,y)(x  y) with q(z,y) bounded for 1 x  yI < 1, say by K . Then, given E > 0, choose S = e/K.] Why are the functions f1, f2 and f3 with f i ( x ) = x3/1x1, f2(x) = x3/1x1 and f i ( 0 ) = 1, f3(x) = ~1x1 all distinct? Show that if removable singularities are cleared o f f then all three become equal. [Hint: All differ at 0 and two functions are equal if and only if they have the same domain of definition and have the same values for the same x. All three are equal for x # 0 and f3 is continuous at 0 . 1
@ Exercise 12.4.2
Exercise 12.4.3 Decide which of the followingfunctions have removable singularities at 0.
Prove the statements in the next three Exercises.
@ Exercise 12.4.4
@ Exercise 12.4.5
The function x H xlq(x) is continuous only at 0.
The function 1~ is discontinuous everywhere. (It is called the Dirichlet 's function.)
Exercise 12.4.6 Let p , q denote relatively prime positive integers, with p < q. For x = p / q let f(x) = l/q, f(0) = f(1) = 1 and f(x) = 0 for all other x E [0, I]. This function is continuous at all irrational points and discontinuous at all rational points of [0,1]. [Hint: If 2 is rational then by Exercise 4.6.6 there is a sequence n H xn # Q with xn 4 2 and f ( x n ) = 0. On the other hand, if 2 is irrational, for any E > 0 there are only finitely many rational points p / q with l / q 2 E . Consequently, there is a S > 0 such that the set { 2 ;I x2 1 < S} does not contain any of these points. For I x2 1 < 6 either f(x) = 0 or 0 < f(x) = f ( p / q ) < l/q < E.]
338
Introduction to Mathematics with Maple
@ Exercise 12.4.7
Both of the ffoorfunction and g : x H  x 2 are continuous from the right at 0 and a t g(0) = 0 but f o g is not continuous from the right at 0. Show this. Give an example of F, G continuous from the left a t G(l) and 1, respectively, such that F o G is not continuous from the left at 1. [Hint: f o g is 1 on [0,1[ and f o g(0) = 0. F ( x ) = 1x1,G ( x ) = 2  x.] The direct translation of Theorem 12.13 into limits is Exercise 12.4.8 false. Show that if lim g(x) = 6 and lim f (y) = I then lim f o g(z) need
1 ’ 2
0@
Y +Y
2 ’ 2
not be 1. [Hint: f (x)= [xl 1x1,g(x) = 0, lim f o g ( x ) = 0, lim f (y) = 1 . 1
x+2
Y+O
12.5
Continuity
Definition 12.7 A function interval [a,b] if it is continuous at every point c such that a < c < b, continuous
This definition can be rephrased by saying that f is continuous on [a,b] if it is continuous from the left at every point c E ] U , b] and continuous from the right at every point c E [a,b[. When drawing a graph of a function f continuous on [a,b] with f ( a ) < 0 and f ( b ) > 0 one has to cross the xaxis or lift the pencil from the paper. This leads to the following theorem.
Theorem 12.14 I f f : [a,b] R is continuous with f ( a ) < 0 and f ( b ) > 0 then there exists a point c € ] a ,b[ such that f (c) = 0.

Proof.
and
We denote [a,b] by [ao,bo] and consider the intervals [a,
[ a+b T , b]. I f f
(T)
$1
=
0 the proof ends, otherwise one of the
intervals will have the property that f is negative at the left end point and positive at the right end point. We denote this interval [ a l ,bl] and continue the process. Either we find a point c with f(c) = 0 in a finite number of steps, or we have constructed a sequence of intervals [a,, b,] ba such that [a,, bn] 3 [a,+l, &+I], and b,  a, = + . 0. By the Nested 2, Intervals Lemma (Theorem 10.17) there exists a number c such that a, 1 c
Limits and Continuity of Functions
339
and
bn
1c. By construction
and
f (bn) > 0.
(12.11)
By taking limits as n 4 00 in (12.11) and (12.10) we have f(c) 5 0 and 0 f(c) 2 0,and consequently f(c) = 0.
Despite its plausibility, Theorem 12.14 does require a proof. We have t o realize that intuition and the concept of continuity do not fully agree; for example there are continuous functions which behave so badly they cannot be graphed. Moreover, Theorem 12.14 has to be proved as a matter of principle: we are obliged t o prove it from the definition of continuity (and its established consequences) and not rely on drawings or our feelings.
Remark 12.17
The idea of the proof can be used to find the point c approximately. If we want to find c with an error not exceeding E , we find n so large that bn  an < E and then the approximate value ? i = (an bn)/2 differs from c by less than E . This can be easily programmed on a computer, and the method has the advantage that we have good control over the error. See Example A.3 in the Appendix.
+
Example 12.13 For f : x H x5 x 1 we have f(1) = 1 and f(0) = 1. Since f is continuous by Example 12.8, the equation x5 x 1 has a solution in the interval [1,0]. Since f is obviously strictly increasing this is
+ +
+ +
the only solution.
Example 12.14
Every cubic equation
f(x) = x3 + ax2 + bx + c = o
with a,b,c E R has a t least one real solution. Since
2+00
lim ( I + x a +>+x 
2 C)=l, 3
there exists k
< 0 such that
for
x 5 k , and consequently
f ( k ) < k3 1 < 0.
2
340
Introduction to Mathematics with Maple
Similarly there exists K
> 0 such that
f ( K ) > 5K3 > 0.
1
By Theorem 12.14 there exists c E [k, K ] such that f(c) = 0.
The following generalization of Theorem 12.14 is often used.
Theorem 12.15 (The Intermediate Value Theorem) Iff : [a,b]I+ R is continuous and a lies between f(a) and f ( b ) then there exists a point c E [a,b] such that f ( c ) = a.
Proof. If f ( a ) = f ( b ) then a = f ( a ) and the theorem is obvious. If f ( a ) < f ( b ) we define F : x H f(x)  a. Clearly F is continuous, F ( a ) < 0 and F(b) > 0. Therefore by Theorem 12.14 there exists a number c such that F(c) = 0; that is, f(c)  a = 0. If f ( a ) > f ( b ) one can apply the 0 already established result on f.
Example 12.15
The equation
has a solution in [1,2]. Indeed f(1) < 0 and f(2) value theorem applies with a = 0.
> 0 and the Intermediate
The intermediate value theorem plays an important role in our next theorem on continuity of the inverse function. We have:
Theorem 12.16 interval I then
Iff is strictly increasing and continuous on an
(i) the inverse f1 is strictly increasing on f(1); (ii) f(1) is an interval; (iii) f1 is continuous on f(1).
Remark 12.18
The theorem remains valid if the phrase strictly increasing
is replaced everywhere by strictly decreasing.
Y1 = f(fl(Y1))
2 y2 = f ( f  l ( Y 2 ) ) ,
which is a contradiction. This proves (i).
Limits and Continuity of Functions
34 1
For every n E N let bn = b if b, the righthand endpoint of I , belongs to I . Otherwise let bn t b if I is bounded from above and bn + 00 if I is not bounded above. Define a, similarly at the left of I . Then by the intermediate value theorem each [f(an, f ( b n ) ] is a part of f ( I ) and moreover
is an interval.
/I
0
V
I I
I I
I
I I
I
I I I
I
I
I
U = X  &
X
Fig. 12.6 Continuity of the inverse from the left. To prove (iii) we denote by ij a point of f ( I ) which is not a lefthand end point of f ( I ) and ij = f(2), that is 2 = fl(ij) . For a positive E let 6 = f(2  e )  f(2). (See Figure 12.6, which is provided for illustration, not as a proof.) For @  6 < y 5 ij we have, by the monotonicity of f1, that
The continuity from the right can be proved similarly.
0
342
Introduction to Mathematics with Maple
By Exercise 5.5.6 the function x I+ x n ,for an odd n E N, is strictly increasing on R. Obviously it is continuous everywhere. Hence the inverse function x H x; is continuous everywhere where it is defined, and that is on R. An alternative notation commonly used is x i = f i . In Maple it is necessary to use surd(x) , but the other two symbols can have different meanings in Maple.
Example 12.16
The theorem asserts that f ( I ) is an interval and if I is bounded and closed then so is f ( 1 ) . However, if I is bounded but not closed, the set f ( I ) need not be bounded. The function x H 1/x2, x ~ ] 0 , 1has ] as the inverse y H l/a, y E [l, m[. The original function had bounded domain but the domain of the inverse was unbounded. Exercise 12.5.5 shows the importance of the assumption that the domain o f f was an interval. The next few theorems extend local properties of functions to global properties. Typical of these theorems is Theorem 12.18, proof of which establishes boundedness of the function, a global property, from continuity of the function at every point, a local property. An important tool for the proofs is the Cousin Lemma below. A finite sequence of points t k , k = 0 , 1 , . . . ,n forms a division T of an interval [a,b] if
u = to
The intervals
[ t k , t1,+1]
< ti < t 2 <  . . < t ,
= b.
(12.14)
X I , in every subinterval
are called subintervals of the division. If there is an [ t k  l , t ~ ,for ] k = 1,. . . ,n then we call
a tagged division of [a,b]. The points XI, are called tags, and X I , tags the interval [tkl ,t k ] . We shall denote the tagged division 12.15 with dividing points T = { t o , t l , . . . , t n } and tags X = { X l , X 2,..., X,} by T X . A function is called a gauge if it is positive everywhere where it is defined. If 6 is a gauge on [a,b] then the tagged division T X is called &fine if
X I ,  XI,) < tlc1 < t k < Xk
+6(Xk)
for all
k = 1 , 2 , . . . ,n. (12.16)
In proofs, a gauge is often used to control the size of subintervals of a tagged division. Of course, any other letter can be used in place of 6 but use of this letter is traditional and a gauge usually enters a proof from some related E6 definition. We are now ready for4
4Lemma means an auxiliary theorem. History of mathematics knows a number of examples where a lemma becomes so important that it gains a status of a Theorem.
Limits and Continuity of Functions
343
Theorem 12.17 (The Cousin Lemma) For every gauge 6 there is a 6fine tagged division of a closed bounded interval [a,b].
Proof. We proceed indirectly. If there is no 6fine tagged division of [a,b] then there is no 6fine tagged division of one of the intervals [a,( a b)/2], [(a b)/2, b]. If it were then merging these two tagged divisions together would define a 6fine tagged division of [a, b]. Denote this interval [al,bl] and continue the process indefinitely. We obtain a sequence of nested closed bounded intervals [a,,b,] and none has a &fine tagged division. Since b,  a, = (b  a)/2, + 0 there is a point c E [a,, b,] for every N. For E = S(c) > 0 there is rn E N such that b,  a, < S(c). Consequently
+
+
c  S(C) < a,
< b, < c + 6 ( ~ ) .
This means that a , 5 c 5 b, is a &fine tagged division of [a,, b,], 0 contradicting the definition of [a,, b,] . Cousin’s Lemma looks obvious, but see Exercise 12.5.6. The first application of Cousin’s Lemma is:
Theorem 12.18 (Continuous functions are bounded) I f f is continuous on a closed and bounded interval [a,b] then the function f is bounded on [a,b].
is continuous on [a,b]. By Remark 12.15 there is a positive 6 such that lf(x)I < If(X)l 1 for I x  XI < 6. Let T X be a &fine tagged division 12.15 and A 4 = 1 Max(f(X1), f(X2), . . . f(X,)). If x E [a,b] then for some natural k the point x lies in the subinterval [ t k  l , t k ] and consequently I x  X k I < 6. It follows that
Proof. By Corollary 12.3.1 the function
If1
+
+
We emphasized that [a,b]was bounded and closed. The importance of this assumption is illuminated in Exercise 12.5.7. The next theorem says that a continuous f attains its maximum and minimum on [a,b].
This is so here.
For every natural n there is a point x.b]. enables to find this extremum.~ / 2 f(Xk) . such that f(xn) < m l/n. b] and I the piecewise linear function which agrees with f at all the points xk (see Figure 12. If x E [xzl.20 I f f : [a. which is E . + c as k + 00. Let be a 6fine tagged division of [a. I f ( . . + 51t is a consequence of the definition that a piecewise linear function 2 is continuous on [a. .7). a maximum value. ) . for all x E [a.b].b ] ) . Consequently I f ( .) + m. d in this interval such that f(c) I f(x) L f ( 4 for all x in [a. and by Definition of the greatest lower bound f(x) 2 m = f(c).1 we obtain f(xnk) + f(c) and by Theorem 12.19 (Weierstrass on extreme values) I f f is continuous on [a.x E [a.b] if there is a division D such that 1 is linear on all intervals [xkl. . By Corollary 12. xk] 5 for k = 1 .b] then there exist c. )..xk] then f(x) lies in the interval J k = ] f ( X k ). The next theorem shows that continuous functions can be well approximated by some simple functions.b] then for every positive E there exists a piecewise linear function I such that. Proof.344 Introduction to Mathematics with Maple Theorem 12.(fI . We shall call a function Z piecewise linear on [a. Hence f(c) = m. x.b].f(X)l By continuity. . However we shall see in the next chapter that the mere knowledge of the existence of.2..b] we find S = S ( X ) such that < ~ / for 2 I x XI < S ( X ) . . We denote by m = inf {f(x). By the BolzanoWeierstrass Theorem there is a convergent subsequence. 0 + This Weierstrass Theorem is theoretical in its character. b] H R is continuous on a bounded and closed interval [a. ~ / 2 and [ so does Z ( x ) because it lies between f(xk) and f(xk+l). 2 .l(x)I < E. ) Proof. Theorem 12. say xn. say.Z(x)I is 0 less then the length of J k .4 that f(x. The proof for d is similar or one can apply the already established part of the theorem to f. for every X E [a.
Fig. If f is continuous at a point y E [a.8 Piecewise linear approximation The theorem has a practical application in Maple. 12.Limits and Continuity of Functions 345 Y If I (xk1) i I I I f (xk+l) I I I I 0 xk1 xk xk+l 2 Fig.b] then for every E > 0 there exists . It was made with Maple: for making such graphs see Exercise 12. It has a large number of subintervals.8. A graph of a linear approximation to a function is depicted in Figure 12.5. 12.7 A detail of a piecewise linear approximation.11. so large that the eye sees a continuous curve rather than a polygonal line. When making a plot of a continuous function. the picture on the screen is a graph of a linear approximation to the function.
XI By continuity of f at X for every q there exists &(X)such that It . Then < S1 and I w .Iyl. We have found something important: 6 is the same for all y.19 on E .17 holds for all x . but does not depend on x (or y). In particular E & S1 = 2 d W &qF+ = 60 Iyl Now if y is restricted to the interval [0.17.17).8 We shall say that a function f is a set S if for every positive E there exists a positive S such 12. Proof.f ( X ) l < q/2 for (lu . This is easy to prove.) + If(u) .17 is 60 = .f(w)I < q.1] then 61 L E / ( 2 d X ) = 6 and so this 6 can be used in implication (12. As a consequence. Let 62 be chosen so that the above implication holds for 61 and q replaced by 6 2 and &/2.21 (Uniform continuity) A function continuous on a closed bounded interval K is uniformly continuous on K .XI < 6. . Theorem 12. with S = 6 the implication 12.346 Introduction to Mathematics with Maple a S > 0 such that This S depends generally speaking on y (and also E). d w The implication 12.17 holds for all y in [0. respectively. For instance for the function f : x H x2 a suitable 6 satisfying the implication 12.17 does not determine 6 uniquely: any positive number smaller than 60 can be used as a S in 12. y E S. 1].X I < 6.8 may (and usually does) depend Remark 12. T h e S in Definition 12. By Cousin’s Lemma there is a tagged division “Not merely for one particular y as before.6 12. If(t).
t k + l ] and then Conceptually and computationally.19) (12. t k ] for some k and w E [ t k .n) 2 If IuwI of 6 < 6 and u.18) The main idea of the proof consists in showing that if f can be approximated by a polynomial on some interval that it can also be approximated by a polynomial on a slightly larger interval. . We denote I = b . polynomials are the simplest functions.) .21) l .a and d = c . belongs to [ t k .  W I < ?5 (12. k = 1. says that continuous functions are close to polynomials. First we find a polynomial p which is between 0 and 1/2 on [a.f(4I < 5 < E & If u. d] and between 1/2 and 1 on [d. say u. very important from a theoretical point of view but also with some practical applications.. b].k O L U ( x ) < l for c . Define 1 6 = Min(&(Xk).22 (Weierstrass theorem on polynomial approximation) I f f : [a.Limits and Continuity of Functions 347 T X which is &fine. w do not belong to the same subinterval of T then the smaller one. w belongs to the same subinterval of T then by definition If(. The next theorem. For the proof we also need a polynomial approximation U of a function which is 1 on a part of the interval and 0 on another part of [a. b].l .~ < U ( x ) < l for a L x < c .1 I f a < c .. We state this as a lemma.20) (12.k < x < c 0 5 U ( z ) < v for c 5 x 5 b Proof. .Ic a polynomial U such that < c 5 b then for every q~ > 0 there exists (12. b].> for all x E [a. Theorem 12. b] H R is continuous and E > 0 then there exists a polynomial P such that If(. Lemma 12.k/2.
[2p(z)ln L 1 . from Equation (5.19). For a given E > 0 let S. 5 b such that there for a 5 J: 5 t.20) and (12. 0 We now proceed with the proof proper of the Weierstrass theorem. with the property that Proof. Obviously 0 5 U ( x )5 1 (12.to] and S. (12.348 Introduction to Mathematics with Maple This is easy p(x) = 1 1 5 + z ( x . gives U(J:) for a 5 x 1 . Clearly a < s 5 b. Consequently f can be approximated by a constant f ( a ) on [a. By continuity of f at a there exists t o such that for a 5 x 5 to. On the other hand we have for c 5 x Ib Since both 2p(ck) < 1 and 2p(c) > 1 it follows that both [2p(ck)ln and [ 2 p ( ~ ) ] + ~ 0.k.[P(41n)2n with some n E N which we choose suitably later.[2P(C .23) c .21) now follow from Equations (12. Then we define U ( 4 = (1.22) for a I x I b. By continuity of . Let s = sup S.k)ln (12. We can therefore find n E N such that 1 + = 0 Equations (12.24). (12. where n is replaced by 2n. # 8..d).22).23) and (12.36). Employing the Bernoulli inequality. respectively. be the set of all t exists a polynomial P.
19) and (12. Now we define and show that it satisfies (12.25) for a 5 z 5 c.28) we have 1 n [c. contrary to the definition of s.26) for s .s 6 would hold on [a.27) that on the interval [a.18) on [a. & Finally.18) on [c. First we have It follows from (12.PE(x)(.S] On the interval [ s . s] = [a.b]. b] and the proof is complete. (12.18) holds on [a. s . s a].Limits and Continuity of Functions 349 f at s there is S > 0 such that (12.26) and (12.(fI .S.This means there is a polynomial P.U ( X ) ) < E . This proves that s = b.27) Mq< 2E 3 (12.6 < c 5 s and s E SE. c] clearly ).a 5z + 5 c} and M so large that for all x E [a. satisfying Equation (12.S 5 x 5 s 6 and x 5 b.28) This is possible since m < E .P(.21). Let m = Max{lf(z) . 0 + + . b]. using (12. We apply the lemma for c . otherwise the inequality (12. b ] .k = s .)I L E U ( Z ) + 3 (1. Hence Equation (12.6 to find the function U with 0 < 7 < 1so small that (12. By the definition of the least upper bound there is a c with s .
30) Consequently. y is ) obtained by the long division algorithm from . Every polynomial is Lipschitz on any bounded interval.b] if there exists a constant L such that. f(u)).350 Introduction to Mathematics with Maple A function f is said to be Lipschitz on the interval [a. 12.b].29: the graph of f lies within the angle formed by lines of slopes L and L with vertex at (u. The geometrical meaning is illustrated in Figure 12. for y 2 a.9 Geometrical illustration of a Lipschitz function For instance.12 to show that this function is not Lipschitz on [O. It is also important to realize that this happens for every u E [a. this function is Lipschitz with L = l/@. if f(z) = .b]. Y L 0 X Fig. We leave it as Exercise 12. A Lipschitz function is continuous but the property of being Lipschitz is stronger: Lipschitz functions behave even better than continuous functions.5. 1 3 . Indeed The expression for q ( ~ . for all x . y E [u./Z then (12.
P(4I < Using Inequality (12. For given positive E we find.Limits and Continuity of finctions 351 the coefficients of P and from x . y by addition and multiplication. y E K . @ Exercise 12. Prove it. I].15. & Exercises Exercise 12.>.f(x2 t(z4 .17 is satisfied for S = 3L 0 and all x . Proof. Let P ( x ) = xn u1zn' . @ Exercise 12.3 + + + @ Exercise 12. Fkom this inequality it is clear that implication 12.5. I f f is a continuous bijection of [a. and Theorem 12.16 can be replaced by a more general assumption that f is onetoone. If the coefficients of P are real and an < 0 then P has a positive root.1 + + . and then As an application of the Weierstrass Theorem we now give a second proof of the theorem on uniform continuity.4 Prove the following.5..x z ) ) .5.d] then f is either strictly increasing or strictly decreasing on [a.b] onto [c.. so if x and y are restricted to a bounded interval q is bounded.32) we have 5. say by L. If q < 2 2 with f(q) < f(x2) and x3 < x4 with f(x3) > f ( ~ 4 ) use the auxiliary function F : t H f(x1 t(x3 . by the Weierstrass approximation theorem a polynomial P such that for all x E K If(.b]. Therefore the assumption that f is strictly increasing in Theorem 12. [Hint: Use an indirect proof.1 Show that the equation x17 + 1 + 2 4 + saw(x) = 200 199 has a solution.2 Prove that a polynomial with real coefficients and of odd degree has a real root. t E [0.an.5.21)) .
We shall encounter some in the next chapter. X := [0. .Pi/&2 * Pi/3.5. the piecewise linear function which takes the value yk at xk is called a h e a r spline. If it were then fi 5 L x for 0 5 x 5 1. 1 The function x H 112.5. [Hint: There is a closed bounded interval I such that f(x) < f ( a ) outside I .6 @ Exercise 12.1 + @ @ Exercise 12. t indicates the variable.14 Show that x H 1/x and x H x2 are not uniformly respectively.5. and the option l i n e a r is needed.a] such that IP(x)I > IP(0)I for x # I. .xn] and Y = [yl. a contradiction.2* Pi].8 @ Exercise 12.] @ Exercise 12.11 Given two lists X = [x~. Here. t . Reconcile this with Theorem 12. Prove that f attains its maximum on R.1] is not bounded. [Hint: Indirect proof.13 Show directly that x H 21x1 is uniformly continuous on R with 6 from implication 12. [Hint: id. Give an example of a function continuous on all of R which does not have a maximum or minimum value. Produce a graph of sin together with a linear spline approximation for Pi.yn] with xk strictly increasing. Exercise 12. [Hint: Since I P I + 00 for x + zkco there is a finite closed interval I = [a. [Hint:fl(y) = y for y < 0 and fl(y) = y 1 for y 2 0. The id function is continuous on the closed interval R and is not bounded.18. Let f(x) = x for x < 0 and f(x) = x .5.Y.10 x400 Suppose that f(a) > 0 for some a E R and lim f(x) = 0. y2.5.co[. @ Exercise 12.5.1].5. . . l i n e a r ) .. [Hint: The proofs continuous on the intervals]O.5. For d ( x ) = x there is no dfine 1 .. The maximum on I is the maximum on R.x2. .5.] @ Exercise 12.. The assumption that dom f is an interval in Theorem 12.5 Prove the following.1 for x 2 1 then the inverse f1 is discontinuous at 0. It can be made by the Maple command s p l i n e (X.] Exercise 12.7 Prove the following.9 If P is a polynomial show that the function I P I attains its minimum on R.l[ and [O. [Hint: If there was then X I 0 < d(X1) = XI L x1 .] tagged division of [0..17 equal to ~ / 2 . The minimum on I is the minimum on R.x E]O.16 is innocuous but essential.] lim f(x) = 1 ’ 0 0 @ Exercise 12.352 Introduction to Mathematics with Maple @ @ Exercise 12. 3 * Pi/2.12 Show that the function x H fi is not Lipschitz on the interval [0. as there are other splines of higher degree.5.
6 Comments and supplements Similarly to sequences. Consequently it has a limit from the right at 0 and this limit is. XTb lirn inf f (x) . In symbols.] E = 1 let x = 6. xla To illustrate these limits consider the function 1 f(z) : z H 2(sgn(z) + ) saw(l/z). We state the definition for the limit superior from the left and leave it to our readers to formulate by analogy the remaining five.Limits and Continuity of Functions 353 are indirect. y = 6/2 and x = 1/6. b [ ) : lirn sup f(z). y = x + 6/2. zTb 610 The notation for the other five limits is (with c ~ ] a . Assume that f : ] u . There are six of these ‘limits’: from the left. 12. the limit superior from the left of f at b. x+c liminf f (x). there are lim sup and lim inf for functions. 2 . from the right and twosided. x+c lim inf f (x) . b [ R~is bounded. since f is bounded. The function M is decreasing and bounded. lirn sup f (z) = lim M (6). by definition. xla lim sup f (z). For respectively.
S < xn for n > N . the BolzanoCauchy theorem applies to complex valued functions and also for limits in the complex domain. xcto 1 lim inf f (z) = .>.4 < 5 + 5’ whenever b. Proof. Theorem 12.S < x < b. then for every positive E there is a positive 6 such that.S < u < b and b6 < v < b. 210 Similarly as with sequences.34). If f has a limit from the left.S < u < b and b . liminf f(z) xtc liminf f(x) = Min x+c Limit superior and limit inferior make sense only for real valued functions. say I .23 (BolzanoCauchy) The function f has a limit from the left at b if and only if for every E > 0 there is a positive 6 such that whenever b .f(4I L If(4 . In contrast. There is a N such that b . E E . the limit exists if and only if the limit superior and inferior are equal and that is so also when the limit is taken from the right or left.354 Introduction to Mathematics with Maple It plausible that 3 limsup f(x) = XlO 2’ lim sup f (x)= 0.f(xm)l < E . This. let xn 3 b and xn < b.XTO 2’ liminf f(x) = 0. together with Inequality (12. f(z).ZI + If(4 . Taking x to be u and also v in the above inequality we have If(. for all x with b . implies that If(xn) . It is not difficult to prove that limsupf(x) = Max x+c . Turning to the second part of the proof.S < v < b.
) f(c+)[. Every 0 subsequence has the same limit. I] but is discontinuous at 0. For the limit in the complex domain the last inequalities are then required for complex u. . We have proved in Theorem 12. c < d. or 0 < Ic . w.Limits and Continuity of Functions 355 for n. The function f equal to saw(l/z) for z # 0 and f ( 0 ) = 0 has the intermediate value property on every subinterval of [1. So r is a bijection of the set of discontinuities onto a subset of rationals and hence the assertion. The inequalities b6 < u < b and b6 < w < b must be replaced by a < u < a+6 and a < w < a 6. The BolzanoCauchy theorem holds also for the limit from the right. respectively. There is a more dramatic example in Boas (1972. namely. say to 1. a monotonic function can be discontinuous only at countably many points. w > K imply If(u) . If c.@] if it attains every value between f(a) and f(P) somewhere in the interval. then f(c+) 5 f(d). hence f(yn) + 1. or the Darboux property. + The fact that monotonic functions have limits from the left and from the right has an interesting consequence. if. The function f is said to have the intermediate value property. This mapping is one to one: for distinct values of c.) . For an increasing and bounded function f we denote.wl < 6. .15 that a continuous function has the intermediate value property. Now we can associate with every point c of discontinuity a rational point r(c) ~ ] f ( c . It remains to be shown that if yn + b and yn < b then also f(y. The Darboux property was in the past confused with continuity but it does not characterize continuity.d are points of discontinuity of f then the open intervals ] f ( c . respectively. Indeed. f ( c + ) [ and ]f(d). If f is discontinuous at c then f(c) < f(c+). f(d+)[ are disjoint. By what we have already proved the sequence with terms also converges.f(w)I < E . By the BolzanoCauchy theorem for sequences 10.20 Similarly as with sequences. for instance. Its limit is I since for the subsequence f ( z n ) + 1. because they lie in distinct intervals. for brevity. on the interval [a.16. Remark 12. this implies that the sequence n I+ f ( z n ) is convergent. the BolzanoCauchy theorem reads: a finite limit of f a t 00 exists if and only if for every positive E there exists K such that u.uI < 6 and 0 < Ic . For limit of f a t 00. the corresponding values of r(c) are also distinct. by f(c) and f(c+) the limits from the left and from the right.) + 1. the BolzanoCauchy theorem makes it possible t o prove that a function has a limit without the actual knowledge of what the value of this limit is.m > N . limit and limit in the complex domain.
He was prevented from publishing some important results which were discovered much later by other mathematicians. Bolzano established the greatest lower bound theorem and what we called the BolzanoCauchy theorem for sequences. Secondly.1 intermediate value theorem. following custom rather than historical accuracy.356 Introduction to Mathematics with Maple p. . Exercises 0@ This exercise outlines the idea of Bolzano’sproof of the Exercise 12. Denote S the set of all x E [a.6. 71) of a function which has the intermediate value property on every subinterval of [0. Let f : [a.b] H R with f(a) < 0 and f ( b ) > 0. f(<) cannot be positive. in the course of his proof.b [ .1 1 . but did not contain any reference to intuition. Bolzano’s paper is remarkable. Bolzano was not only a mathematician but a moral philosopher and was politically persecuted. geometry or the concept of motion. f([) cannot be negative.b] for which f(x) > 0. which used only the definition and logical means. t = inf S exists and is in ]a. Show that 1. in particular by Weierstrass. By that he meant a proof. scientifically. 3. This was done by Bolzano in 1817 in a paper whose main purpose was to prove the intermediate value property for continuous functions. Mathematicians of the 17th and 18th centuries had an idea of continuity but were unable to formulate it precisely. The Bolzano definition was what we called the Cauchy definition. Firstly he set out his goal very clearly: he wanted to prove the theorem. 2.1] but is discontinuous at every point of [O. as he said.
In this chapter we introduce derivatives as limits.f ( t )and the average velocity will be + + f(t + h ) . Let the distance of this body from a fixed point be a known function of time.Chapter 13 Derivatives Derivative can be described informally as a rate of change and as such it is extremely important in Science and applications. The secant line joining the points (z. f(z h ) ) has the slope + + As h + 0 the points (z. Consider another example. t h] the body travels the distance f ( t h ) . f(z+ h ) )coalesce and the secant becomes a tangent. say f . During the time interval [t. f ( z h ) ) moves towards + + + 357 . This corresponds to moving the ruler aligned along the secant line carefully so that the point (z h. this time from geometry. see Figure 13. We also extend the Taylor Theorem from polynomials to power series and explore it for applications (within mathematics). establish their properties and use them in studying deeper properties of functions and their graphs. f(z)) and (z h.1 Introduction Let us think of a body moving along a straight line.1.f(t) h The velocity shown on the speedometer of a car or a plane is the instantaneous velocity and for our moving body this is the limit of the above expression as h + 0. 13. f(z)) and (z h.
Also Dx = (2)' = 1. . Indeed.1) (is called the derivative of f at x. Example 13. f(x)) until it reaches its limit position. and then the secant becomes a tangent. Using the binomial theorem it is easy to show that Dxn = nxnV1for n E N. if the tangent forms an angle $ with the xaxis then f'(x) = tan$. It is denoted by f'(x) or Df(x). (c)' = 0.358 Introduction to Mathematics with Maple (x. 13.1 The number (13. For the quadratic function we have Ox2= h+O lim(2x+ h) = 22.1 The derivative of a constant function is 0 at any point. Geometrically.1 Secant becoming a tangent Y (Definition 13. 0 x+h Fig.
Example 13. Indeed I+ 1 x 1 has a t zero derivative from the limlhl 0 = lim(1) = 1. Remark 13. If f ( x ) is abbreviated as y or u then f ’ ( x ) is denoted by y’ or u’. and fL(x). ) is sometimes calculated according to Equation (13. The limit in Equation (13. We shall comment on this in the last section of this chapter.1) can be understood as a limit in the complex domain or in the real domain. These two concepts are distinct! Usually it is clear from the context which derivative is meant. In the previous example. This is not true: Bolzano was aware of this and later Weierstrass published an example of a function continuous everywhere and differentiable nowhere.Derivatives 359 As h + 0 the second term on the right hand side tends to 0. make sense only as derivatives in the real domain. hTo h hTO This example shows that a continuous function need not have a derivative. In this book the word derivative means . respectively. For x # 0 we have Remark 13. Both onesided derivatives. and the derivative is then called an infinite derivative.f (4 xa The phrases ‘f has a derivative a t x’ and ‘f is differentiable a t x’ are used interchangeably.2 The limit in Equation (13. and from the left 1. If the limit in Equation (13. The derivative O f ( . The derivative from the right is denoted by f:(x).1 The notation f’(x) is more common.1 becomes definition of the derivative of f from the left or right.1) can be infinite. Geometric intuition might suggest that a function continuous on some interval would have derivative a t many points of the interval.2 The function f : x right 1.1) as D f ( a ) = lim x+a f (4. the derivative from the left and the derivative from the right. denotes the derivative from the left. all the formulae are valid for the derivative in the complex domain and then of course also for the derivative in the real domain. however.1) is replaced by the limit from the left or right then Definition 13. and Maple also uses D ( f ) ( x )rather than D f ( x ) .
Each notation has dx ’ its advantages and disadvantages. f :=x+x2+1 22. For instance. > restart. The symbol y‘ hides the x. it is clear that t should be regarded as a constant and from the notation d ( x 2 txc> dx x as a variable. we shall consider infinite derivatives only for functions for which the domain of definition is a part of R. that is.360 Introduction to Mathematics with Maple a finite derivative and if we allow the derivative to become infinite. We now consider some differentiation with Maple. Example 13.x2 tx =2x+t. so it cannot be used for denoting. for example. Moreover. say II: = 1.f ( x ) or by dx dx unambiguous.3 Let f ( x ) = &. Although dY does not hide x . f’(1). + d(x2 t x ) (x = lim h+O dx + + h)2+ t ( x + h) h . Then We introduce yet another notation for the derivative: we denote f ’ ( x ) by d d f ( x ) . but this latest notation has the advantage of indicating with respect to which variable the derivative is taken. The original notation is shorter and on most occasions .22 1 X 1 . g ( x ) = V G . dY Naturally if f ( x ) is denoted by y then f ’ ( x ) = y’ = . it is dx also unsuitable for denoting the derivative at a particular point. we shall explicitly say so.
x) .(1 x + x2) (1 2 2) (1 . (in simpl/abs) abs is not differentiable at 0 It is true that x H 1x1 is not differentiable at 0 nevertheless g is differentiable at zero. Error. However Maple can find the derivative as a limit 0 The main advantage of using Maple for differentiation lies in the fact that Muple easily calculates derivatives which otherwise would be laborious to do. the .1 derivative I/(Z./Z).x x2)2 + + Exercises Exercise 13.Derivatives 36 1 The command diff can also be used to calculate the derivative of a function with values f(x). 122 1x+x2 > . 1 22. Here is an example which would not be too difficult without Maple. Maple is unfortunately unable to calculate Dg(0).1. > d i f f ( ( l+xxA2)/ ( 1x+xn2) . Prove that the function x w & has. 2 1 2 2 (1 . but using Maple still has an advantage. for all x > 0.x x2)2  + + simplify(%).22 > g: =x>x*abs (x>. as we shall see later.
Remark 13.2 If a function has a finite derivative at a point then the function is continuous at this point.) i g’W7 (13.2 Basic theorems on derivatives If f is differentiable at z then the function h H D(h) = f ( x + h) .3) yields f(z so f is continuous at x.5) . 13. Equation (13. This theorem as well the next two are valid in the complex From this theorem it follows immediately: Theorem 13. If so then D(0) = f’(x).3 domain.1) and the result from Example 13.3) has a removable discontinuity at 0 and by defining D(0) = f’(z)becomes continuous.] Exercise 13.362 Introduction to Mathematics with Maple Exercise 13.3 If f’(x) and g’(z) exist and c is a constant then [cf(z)l’ = C f ’ b ) . + h ) . [Hint: Use equation (13.p z + l for z + o and n f N. which is theoretical in character but needed at many practical occasions.f ( 4 h (13. As h + 0.2 1 Show that DXn = n. namely that Dxn = nxnl.&). This leads to the next theorem. (13.4) [ f ( 4 + g(z)l’ = f’(.3 Use Maple to find D(1+ &)/(l .1.1.f(z)+ D(0)O = 0 / Theorem 13. I/ 0 Proof.1 The function f is differentiable at z if and only if there exists a function D continuous at zero such that Equation (13.3) holds.1. Theorem 13.
1 Prove by induction: @ Exercise 13. (u v)‘ = u ’ + v’. + Proof. since Dl(0) = f’(x) and D2(0) = g’(x).1.5).Derivatives 363 These formulae are more easily remembered as (cu)’ = CU‘.2. (uv)’ = u’v + UV’.1 f(z h) = f(z) Dl(h)h. Does this contradict Theorem 13. 1 for x = 0.1 do agree.2 Show that f g and f g might have derivatives even though neither f nor g has. Exercises @ Exercise 13.2 and if not why not. 0 Remark 13. By Theorem 13. > + + = (f(. Let us now turn to Equation (13. + + + + (13.2. (Hint: f(z) = 1x1. It is important to note that [ .4) and then by Equation (13.6. g ( x h) = g(x) D2(h)h.2.) follows from Equation (13. The Formulae (13.6).f ( x ) ] ’ = f’(3.25) and the definition of a polynomial according to Definition 13.f ( .4) and (13.5) are immediate consequences of theorems on limits and Definition 13.4 This Theorem together with the formula Dxn = nxnl makes i t clear that the definition of the derivative of a polynomial as given in Equation (6.g(x) =1. > g ( . .)az(h) + 9(4Dl(h) + w w 2 ( h > h > h (13.9) Consequently f(z h)g(z h) .1 + @ Exercise 13.8) (13.10) Dividing by h and then passing to the limit as h + 0 leads to 13.3 The function x sgn(z) has an infinite derivative +oo at 0 but is discontinuous at 0.
13) is easily du dx remembered as fractions which cancel. Example 13. We now extend the formula for negative integers. Example 13. Then by Equation (13.22.4 (The chain rule) If g is differentiable at x and f differentiable at g(z) then f o g is differentiable at x and (f (9(4) >‘ = f ’ b(4 >9’ (4 ( 13.1 we established the formula (xn)’ = nxnl for a positive integer n. We wish to calculate D ( l x2)20. .are indivisible but still Equation (13.4. Before we prove the theorem we illustrate it by two examples.11) The equation is often rewritten in the following way with y = f(z) and u = g(x> (13. dx dx d Y = 2 0 ~ =~ 4oz(i ~ 2 x2)19. Far more convenient is t o use Theorem 13.5 Set m = n then m is positive and Clearly this is valid for x # 0.13) dy du The symbols . We can expand the expression by the binomial theorem and then take the derivative. set u = 1 x 2 and y = f(u) = u2’. ~ + In Example 13.13).364 Introduction to Mathematics with Maple The chain rule The next theorem concerning the derivative of a composite function is one of the most important rules for differentiation.and .4 + + du _ . Theorem 13.
1 (13. By Theorem 13. This happens when the function is multiply composed.1. Proof.14) (13. Firstly. by Theorem 13. Example 13.6 to find +0 0 Sometimes the chain rule needs t o be applied t o the inner function.g(x) leads to + (13. passing to the limit in Equation (13.Derivatives 365 We now take up the proof of Theorem 13.1 that (&)I = 1/(2&).16) as h gives Equation 13. Setting u = g(x) and k = g(x h) .17) . Let us recall from Exercise 13.4 D&TF = and then 1 21/iT72x If f(u) = 1/u then the chain rule leads to Applying this and Equation (13.3 to the product 1 Theorem 13. Dl(0) = f'(u) and D2(0) = g'(x). We wish (x+ d m ) 5 .11.5 I f f and g have derivatives a t x and g'(z) # 0 then (13.6) from Theorem 13.4.15) with D1 and D 2 continuous at 0.16) Since k + 0 as h + 0.
3.a with the zaxis.f(x)) forms an acute angle a with the xaxis.2.4 Use theorems of this section to calculate the following derivatives and then check your results by Maple.2 provides a good motivation. It follows that 4 .366 Introduction to Mathematics with Maple This equation can be rewritten as ’ u’v (t> = . x) = (y.4 remains valid if the derivative is replaced by righthand (or lefthand) derivative. provided g is strictly increasing (decreasing).2. 1.7 Derivative of the inverse function Now we study the derivative of the inverse function. Figure 13. Give an example of this.5 Theorem 13.uv’ 212  Exercises Exercise 13. The tangent to the graph of f at (x.1 Prove that (f(aa: b))’ = af’(ax + + b). fl(y)) subtends an angle a with the yaxis and the angle 7r/2 . ( (22 + 3x+7 )I.4 is not valid if the derivative is replaced by righthand (or lefthand) derivative. @ Exercise 13.4. [Hint: Use @ Exercise 13. We consider a function f strictly monotonic on some interval I.2. Hence the tangent to the graph of f1 at (f(x). [Hint: g(x) = 2. x + 1)2 ’ @ Exercise 13.6 Theorem 13. f(x) = 1 x 1 at x = 0.1 Prove that Theorem 13.2. (=)I.
19) . (13.Derivatives 367 Fig.18) 1) The above formula is easily remembered as d Y1 dx dx d Y with the understanding that y denotes f(x).2 Derivative of the inverse We state this as %( 1 Theorem 13. 13.6 (Derivative of the inverse function) fXY)=f ‘b) * I f f is strictly monotonic and f has a derivative f’(x) # 0 a t an interior point x of the interval I then the inverse function has the derivative fLl(y) at the point y = f(z)and 1 (13.
20) for k ( h ) leads to Taking limit as h + 0 proves Equation (13. Define k ( h ) = fl(y continuity of the inverse.1 with V continuous at 0.18). Y +h =f ( x (13.21) (13.25) extends it t o n of the form l/rn with m E N. k is continuous a t 0 with k ( 0 ) = 0. The inverse’ fl(y) = yl/m has the derivative 1 1 dyh yn fl_. Equation (13.20) (13.22) gives (13. We now use the chain rule t o extend it for any rational m = p / q . p E Z and q E N.f ( x ) .24) Using Equation (13.22) h = f(z k ( h ) ). + Example 13. By the Theorem on the Proof.368 Introduction to Mathematics with Maple h) . By Theorem 13.25) We know t h a t the formula (xn)’ = nxnl holds for n E Z. + +W)).fl(y). o o [ .sign for a strictly decreasing f . (13. .7 Let f ( x ) = xm with m E N. for m odd on all of R and for m even on [ O . (x”)‘ = (x”p)’ ( ‘For even m we tacitly understand that f is restricted to [0. Denote y = f(x) then x = fl(y) and we obtain successively: + fl(Y + h) = x + k ( h ) .m[. Then f is increasing.(y) = dy mxml m 11 . Substituting for h from Equation (13. 0 A more detailed and careful examination of the proof will show that if f’(x) = 0 then f i l ( y ) = f o o with the sign valid for a strictly increasing f and .
We shall prove both theorems in this regard. is strictly decreasing in I .3 Significance of the sign of derivative.3.Derivatives 369 Exercises @ Exercise 13.b]. exists at c = f ( a ) and Prove it.9 Prove: I f f is strictly increasing on I and f'(c) = +oo for an interior point c E I then for d = f(c) the derivative fLl(d) = 0 13.1 @ Exercise 13.f is increasing in I . If a function has a positive derivative then the tangent forms an acute angle with the zaxis and it is plausible that then the function increases. Show that Theorem 13. (a) Z'(x) < 0 for all 2. It is easy to explicitly define a linear function Z with the following properties. then (f1). see Figure 13. Proof.6 can be applied to F and prove (f1): + (4= Pd. locally at a point or in an interval.5 everywhere. ' is replaced by f : (or fi) This theorem remains valid if f Remark 13. Then (i) (ii) (iii) (iv) (v) f'(z)2 O in I" +. is strictly increasing in I . It is this more general version which we prove.2. (c1. Theorem 13. So assume that fi(z)2 0 in I' and that there are points z1 < 2 2 with f(z1) > f(z2). f is strictly increasing and f $ ( a ) # 0.8 The validity of Theorem 13.6 can be extended to endpoints of I = [a.f f'(z) 5 0 in I" + f f'(z) = O in I" + f f'(z) > 0 in I" + f f'(z) < 0 in I" +.7 (On the sign of the derivative) Assume that a function f is continuous on the interval I and denote by I" the set of interior points of I . We prove first (i) indirectly. is decreasing in I .2. is constant in I . [Hint: Define F ( z ) = f(z) for z 2 a and F ( z ) = f ( a ) f:(a)(z a ) for x < a. . 14 for instance. We start with the case of the derivative having a constant sign on an interval.
. Since f(x2) Z(x2) < 0 by continuity of f 1 there is a positive S such that f(x) Z(x) < 0 for x2 . To prove (ii) it is sufficient to apply (i) to the function f.Z(zn) > 0 leads to f(<) 2 I(<). hence f+(x) = 0 for a < x < b. Then there are two points a . hence f(<) = Z(S) and e < < <+ Dividing by x  < and letting x + < gives This contradicts (i). Obviously f is constant on [a. By (i) f is increasing. Since f(z) . f(x) .Z(x) < 0 for x > <.6 < x 5 2 2 .370 Introduction t o Mathematics with Maple I I I I I I I I I I Fig. Using (b) and continuity of f . Consequently < 22. Passing to the limit as n + 00 in the inequality f(xn) . b with a < b in I such that f(a) = f(b).Z(x) > 0. for every n E N.there is xn with l/n > X n 2 5 and xn E S . After (i) and (ii) are established.3 A function which does not increase from 21 to x2 Let S = {x. The proof of (iv) is again indirect. 13. By the definition of the greatest lower bound. so assume it is not strictly increasing.Z it follows similarly that > z1.b]. which is a contradiction. XI 5 x 5 22) and = sup {S}.by continuity f(<)5 l(<). (iii) becomes obvious.
2). It is still important but at most occasion the intervals where f decreases or increases can be found by plotting the graph of f by Maple.O. . > plot(p.3). > p:=~>18*~^5125*~^3+385*~+1. 13.4 Graph of a quintic To get a better idea we plot p on [0. p := x + 18x5  125x3 + 3 8 5 x + 1 W e look at the graph of p > plot(p.8 W e wish to find the intervals where the following polynomial is monotonic. 21. part(v) follows from (iv) applied to f.. Fig. . Example 13.3. in Figure 13.Derivatives 371 Finally. 0 Before the age of computers this theorem w a s an indispensable tool for plotting functions.5.
4 0. 1.7 we calculate the points where the derivative is zero.5 Graph of a quintic.2 0. with restricted domain and range The exact location of points where f changes from increasing t o decreasing is still not clear. 1. 6. 6 Obviously p increases for x < then decreases till and so on.354006401. 13.527525232. 1. In order to apply apply Theorem 13. this is often so in examples like this.6 0. Between consecutive points the derivative does not change its sign and the function is therefore monotonic. .dz. 1.354006401. On reflection it was more convenient t o use f s o l v e rather then s o l v e .527525232 Since we were dealing with a polynomial.8 1 1 . 9 9 > fsolve(D(p) (x)=O).8 2 Fig. . > solve(D(p) (x)=O) . f s o l v e found all solutions t o .372 350 Introduction to Mathematics with Maple 300 250 200 150 100 50 0 0. 2 1. 6 1. 5 1 3 1 1 a. .4 1 .
9 Fig.1] and the dimensions of the largest rectangle are d and w .6. We needed the zero of A but the zero in itself was of secondary importance. If c < 0 then A’(x) < 0 on [0. We seek the maximum value of A(x) = (dx)(w1+x) on the interval [0.1.1].1 c. Similarly. Example 13.22 is positive for x < c = (d . If 0 < c < 1then the largest value of A is attained a t c and the dimensions of the largest rectangle are d . f > := x + 100 1 401 x3 . In more general situations.Derivatives 373 p(x) = 0 automatically.c and w . Find intervals of monotonicity of the following function.6 A rectangular plate The derivative A’(x) = d . the crucial importance was the sign of the derivative. f :=~>100/(401*~^3899*~^2+600*~) .w 1.w 1)/2 and negative for c > x.899 x2 + 600 x D(f)(x). .x)] . additional care is needed to find all solutions.1798 x 600 (401 x3 .1]. if c > 1 then A has the largest value for x = 1.899 x 2 600 x ) ~ + + > L:=[solve(D(f) (x)=O.10 > > > restart. + + + Example 13. 13. see Figure 13. The problem consists of finding a rectangle of largest area cut from a metal plate with the lower left hand corner removed. the function is strictly decreasing and A attains its maximum value a t the left endpoint of [0. 100 1203 x2 .
w[. .374 Introduction to Mathematics with Maple L := [899 1203  1  1203 m' 1 +&ail 1203 1203 899 .a[. . Hence f is strictly decreasing on these intervals.2). On the remaining interval f strictly increases.5 2 1  Fig.00.O[ and 30.w.5 1 X 1.7 Graph of f(x) = 100/(401x3 .8 9 9 ~ ' 6002) + . 2 local maximum Y 1 0. For large x the derivative is obviously also negative. Near 0 the derivative is negative. by continuity it is negative on ] .y=2. Our result is confirmed by the graph > plot(f(x). 13. w[.2.b] and [b. We are left with the intervals ] . hence f strictly decreases on [b.x=1.7 cannot be applied t o any interval containing zero. [a. . hence Theorem 13.9916379636 b= 899 + L 1203 1203 d m f is discontinuous a t 0. O[ and ]O. a ] .5029588776.
f ( 4 I f(c)* (13.9 I f f is continuous a t c and f ' W >0 on some interval ] a .29) is reversed then f has a local minimum and if the reversed inequality is strict then f has a strict local minimum. We say that f has a local maximum at a / 3 .x) has a local minimum at 0. an interior point of I . . a It is important to be aware that the converse of Theorem 13.c[ and (13.29) If this inequality is strict then the maximum is also strict. Since the derivative exists f'(c) = f i ( c ) = fL(c) and consequently f'(c) = 0. if there exists a positive 6 such that. b [ then f has a strict local maximum at c. Theorem 13. since f(x)f(c) 5 0 to the right of c. Then. Similarly f:(c) 2 0. The function x I+ x3 has derivative 0 at x = 0 but is strictly increasing on R and therefore cannot have a local extremum at 0. + 1) Theorem 13.30) and (13.30) f'(4 < 0 (13. the function x I+ x2(x .Derivatives 375 We observe from the graph in Figure 13.8 is false. More precisely finition 13.31) then f has a strict local minimum a t c. for 0 < I x c I < S. If inequality (13. it follows that f i ( c ) 5 0. The next Theorem guarantees the existence of a local extremum. If reversed inequalities hold in (13.1) has a strict local maximum at 0. For definiteness let us assume f has a local maximum.2 A function f : I H R is said to have a 1 0 2 maximum at c.8 I f f has a local extremum a t c and f'(c) exists then 1 Proof.7 that at x = &/3 the function is larger than at nearby points although obviously f is larger still for z positive and close to 0. The common name for a local maximum or a local minimum is a local extremum. The function x H (x 1x1)(2 .31) on some interval ] c .
30). The proof for the local minimum is similar. The reason why Maple uses this alternative notation need not concern us.1) +22+2 (1 Ix . > L :=solve (g .x2)l0x2O.1) denotes sgn(x . Example 13.x) .11 Let f(x) = (1 . Since and f ' changes sign a t both x = 1 / d and x = l / f i the function has strict local maxima a t these points. the function f increases to the left of c. The next example shows how to use Maple in such a situation. x . f := x + 15 1 +x2+2x1 + Ix .376 Introduction to Mathematics with Maple This Theorem is easily remembered as: If f ' changes sign at c then it has a local extremum at c. Hence it is smaller to the left of c than it is at c. or we can apply the already proved 0 part of the theorem to f. Example 13. By Theorem 13. Similarly. 1. f decreases to the right of c so it is smaller there than at c. f . defined > restart. However we note that f is not differentiable a t x = 1.2]. x . Many practical problems in applications require finding the largest or the smallest value of a given function on a given interval.7 and by inequality (13.1 1 I 1 g := 15 abs(1. Proof. Obviously.l . > f:=x>15/(l+abs(xl))+xn2+2*~1. O . the largest or the smallest value is either a local extremum or the value of the function at one of the ends of the interval. . on the interval [0. It is clear without any differentiation that f has strict local minima a t .12 We wish t o find the largest and smallest value of the function below. However there are other extrema.11)2 + abs(1.1).
1. is(13/2=f (0)). 13 29 K l := (17.Derivatives 377 For better understanding we convert these numbers t o decimal fractions. 13/2 is attained a t 0.11084811. using the Muple command map to map the set { K } onto the set {Kl}. 2) In order to apply max and min we convert K 1 t o an expression sequence. We had better check it: > is(i7=f (1)). K := 0. 1. > Kl:=map(f. the points where the derivative does not exist (there is just one) and zeros of the derivative which lie in [0.Ll[l]. 13 29 K2 := 17. > K:=O.I.2]. Ll := 1. 2 ' 2 17 17 is the value of f a t x = 1.674571001 We evaluate the function f a t these points.583911177 Now we put together the endpoints. 1. 14.674571001.11084811.{K}).2. 2. true true . 14. > L1 :=evalf (1) . 2.
. for instance t h a t solve or fsolve did not find all solutions to O f ( . > plot(f ..... Here the easiest check is to plot f.. it is always advisable to check the results. .... Moreover..6 0. We get a more definite answer if the derivative is positive.....32) for 0 < h < 6. 16 14 12 10 . If the inequalities (13.... it is easy to overlook some subtle feature of Maple.. A function f is said to be increasing at a point c if there is a S > 0 such that f ( c .8 1 1....32) for f are reversed the function is said to be decreasing at c...6 1.2 1.8 2 Fig....2).....4 0... However. and have found the maximum and minimum values..... when using a computer algebra system.o. 0 0......8 Graph of the function f If the derivative is 0 at some point then the function might or might not have a local extremum at that point... A small typing error can cause an enormous error in the final result. ) = 0.....h) < f(c) and f ( c + h) > f ( c ) ..2 0. (13..4 1..378 Introduction to Mathematics with Maple We succeeded.. 13.
[Hint: Indirect proof. Define an auxiliary function F : z H Lx . Theorem 13. 1 v m . The proof is entirely similar if Df(c) < 0.10 F is increasing at x1 and decreasing at 2 2 . On the interval [XI.11 I f f ‘ exists on an open interval I then it has the intermediate value property. By Theorem 13. 221 the function F attains its maximum. For f(z)= 1 x 1 we have f i ( 1 ) = 1.15 we have D ( h ) > 0 for Ihl < 6.3 + .(1) = 1 but fi(2)is never zero on ]0.32 hold. The derivative of a function has the intermediate property even if it is not continuous. If Df(c) > 0 then by Remark 12. < 0 then f is increasing a t with D(0) = Df(c) and D continuous at 0. cm[. assume that f(d) < f(c) and find a local maximum between c and d. 1 .3.1[. Ezercises Exercise 13.3. respectively. By Theorem 13.1 Find the local maxima and minima of [Hint: The answer is 1.] The previous exercise is particulaxly useful if I is either infinite or open.Derivatives 379 ~ Theorem 13.10 If Df(c) > 0 or Df(c) c or decreasing at c. f. Find the minimum value of 2 1/x2 on 10. This Theorem is not valid for onesided derivatives. Consequently c # 2 1 and c # 2 2 and the maximum at c is local. Proof. Consequently Equations 13.O.f(2).3.2 Prove that if a continuous f has only one local minimum a t c on an interval I and no local maximum then f (c) is the smallest value o f f on I . @ Exercise 13. say at c.8 we have F’(c) = 0 and 0 consequently f’(c) = L. @ Exercise 13. For sake of definiteness let f’(s1) < L < f‘(22) with 2 1 < 2 2 .
b[ If further. b > a. then it has at least one local maximum at some point c € ] a . .34) The notation with D2 = D D or generally D" is also used.3.6 A truck travels between Sydney and Brisbane. at the uniform speed of x km per hour and does not exceed the speed limit of lOOkm per hour.3. that the cost of the crew is c dollars per hour and the cost of diesel per litre is $0.7 to prove: I f f ' is constant on some open interval I then f is linear on I . Prove it.3. Show also that the assumption that f continuous is essential.3. If this function has a derivative at x then it is called the second derivative of f at x and is denoted by f"(x) or D2f(x). For what c is the limit speed of 100km/hour the most economical? + Exercise 13.7 Among all circular cones of surface area S find the one which has the largest volume. a distance of 700km. Assuming that the consumption of diesel for 1OOkm is given by a bx2 of litres.3. Exercise 13. f has no local minimum inside ] a . the third derivative.4 Prove: If a continuous f has local minima at a and b. @ Exercise 13.8 Find the circular cone of largest volume inscribed in a sphere of radius R. denoted by f "'.74. The second derivative can again have a derivative.4 Higher derivatives If a function f has a derivative at every point of an interval I then the function x H f'(z) is denoted naturally as f '.3. The nth derivative f(") is defined inductively (13.5 I f f is increasing on I with rgg c I then g and f have the same extrema (local or otherwise). og Exercise 13. For instance (x7)" = 7 (x6)' = 42x5 = D2x7. 13.9 Use Theorem 13.380 Introduction to Mathematics with Maple @ Exercise 13.b[ then f attains its largest value at c. @ Exercise 13.
f :=x+x3 > > D(f)(x). Similarly do)= w. > > f:=x>x^3. and is similar to the proof of the binomial theorem 5. + 1)(1+ x ) ~ .n + 1).4. . . x .x) . Df : = x + 3 x 2 > D(Df)(2).4 and is left to the readers as an exercise. = m(m .12 .(m .39) In this formula u(O). . 3 x2 Df :=unapply(%.36) (13.12 > diff ( x 6 3 .the zero derivative of u.Derivataves 38 1 or if f(x) = (1 + x)" with m E Q then (13.1 ) .39) is by induction.1) . This convention is not only used here but generally throughout calculus.n + (13. The proof of Equation (13. + x)" = m(m . 13.1)(1+ 2 y 2 P ( 1 x)" = m(m .1 Higher derivatives in Maple restart.( m." .38) For x = 0 we obtain f'"'(0) The Leibniz formula for the nth derivative of the product reads (13. is to be understood as u itself.35) (13.37) D(1+ x ) " = m(1 D2(1 + x)ml. x ) .
9 and 13. f is said to be concave or strictly concave on I if f is convex or strictly convex. this is easily %/loo!.%).40) is strict the function is strictly convex.ll):diff(x"(lOO). If f"(c) > 0 then f has a strict local minimum at c. 1 13. if f"(c) < O then f has a strict local maximum at C.10 to the second derivative gives Theorem 13. . If the inequality in Equation (13.100):diff(I/(l+x). The terms concave up (upwards) and concave down (downwards) are used by some authors instead of convex and concave. for m = 1 this number is equal to confirmed: > loo!.382 Introduction to Mathematics with Maple If we have to differentiate many times it is convenient to produce a sequence of x's first. This theorem deals with the second derivative at a point. > dnu:=seq(x. Equivalent forms of inequality (13.4. We are now going to study the behaviour of a function if the second derivative is of a definite sign on some interval.12 Assume f'(c) = 0. respectively.40) for 21 < x2 < 5 3 in I .dnu).k=l.d~:subs(x=O. . 9332621544394415268169923885626670049071596826438162146859\ 2963895217599993229915608941463976156518286253697920827223\ 758251185210916864000000000000000000000000 By Equation (13.38).k=l. 5653408585997652480000z 8 ' > dnum:=seq(x. A function f is said to be convex on an interval I if (13.2 Significance of the second derivative Applying Theorems 13.40) are .
45) We denote . I t is strictly convex if and only if f ’ is strictly increasing on I .43) Taking the limit again in inequality (13. The next theorem extends this monotonicity to the derivative. subject only to the condition 2 1 < 3 3 .42). leads to (13. Since 2 1 and x 3 were arbitrary. this time as 22 + 2 3 .43) and (13. Let f be convex.13 A function differentiable on an open interval I is convex if and only if f ’ is increasing on I . we have proved the monotonicity of f If f is not convex then there are three points 21 < 2 2 < 2 3 such that inequality (13. Proof.1 A function differentiable on an open interval I is concave if and only i f f ‘ is decreasing on I . f ( x 3 ) ) to ( 2 2 . I t is strictly concave if and only i f f ’ is strictly decreasing on I .Derivatives 383 The geometrical meaning of Equation (13.42) gives (13. Sending 2 2 + z1 in inequality (13. f ( 2 1 ) ) to ( 2 3 . Theorem 13. f(zz)). f ( ~ 2 ) ) .44) leads to f’(x1) _< f ’ ( 2 3 ) . f(x1)) with ( 2 2 . In other words ’.40) is: The point ( 2 2 .44) Combining Inequalities (13. f ( 2 3 ) ) is smaller than the slope of the line joining ( 2 3 . f ( x 2 ) ) lies below the straight line joining ( 2 1 .40) is false. (13. Corollary 13. This means that the difference quotients are monotonically increasing. whereas Equation (13.42) indicates that the slope of the line joining (XI .13.
2 O in I” + f in in in in (ii) (iii) (iv) (v) f”(x) 5 O f”(x) = O f”(x) > O f”(x) < O I” =+ f I” + f I” =+ f I” 3 f Example 13. together with Theorem 13.119 x)(’/~) 2 (123x2 . leads to the next Theorem.119 x ) ( 2 / 3 ) + 238 123 x’ 119 f2 := 9 (41 x3 .46) would be zero at 21.x3 [. is concave in I . is strictly convex in I . At these points F’(x) = 0 and consequently at these points f ‘ would be equal to Q. x2. If f is strictly convex then f’ is still increasing. f ‘ would not be strictly increasing.119 x)(5/3) (41 x3 . hence not strictly convex. f := x + (41 x3 . is strictly concave in I . and then f would be linear on this subinterval by Exercise 13. which is easier to apply. If f is strictly increasing but not strictly convex then F defined in Equation (13. is convex in I . if it were not strictly increasing it would be constant on some subinterval of I . This proves f ‘ is not increasing. is linear on I .45) we obtain F(x1)= 0 < F(x2). So.7.7. by Theorem 13.119)’ 82 x 9 (41 x3 . there is a point J12 between x1 and 2 2 such that f’(J12) > Q. Reasoning similarly leads to J23 between 22 and x3 for which f’(C23) < Q.9. f :=x>(4l*x^3119*x)(l/3). 2 3 and would have local extrema in 3x1.13 > > restart. 22 [ and 1x2.3. 0 This Theorem. consequently. ~~ Theorem 13.14 Then (i) f”(x) Assume that f” exists on an open interval I .119 x)(5/3) + .384 Introduction to Mathematics with Maple and chose an auxiliary function From inequality (13.
41 dam] 41 > a:=L[3] .X 2. When plotting f the command surd must be used rather than a fractional exponent because cube roots of negative numbers are involved. L := [O. An immediate consequence of this definition is Theorem 13. The graph is shown in Figure 13. 1 a := .&35 41 1 b : = dB% 41 f is convex on 10. for some positive 6.6 < x < c and also for c Consequently F is strictly increasing and < x < c+6 . u [ and concave on the two remaining intervals.b:=L[21 . Iff has an inflection at c and f”(c) exists then An inflection point is of significance for the graph of f since the graph of f attaches itself tightly to the tangent at the inflection point: the graph crosses the tangent from one side to the other. A function f is said to have an inflection point at c if f‘ has a strict local extremum at c. the function has a positive derivative for c . If. To find where the derivative is positive and where it is negative we find the zeros of the denominator. F ( z ) > 0 for x > c. . so the sign is determined by the denominator.00. b[ and 3 . f ‘ has a strict local maximum at c then.x)l. > > restart. plot(surd(41*xn3119*~.3) . We denote 1 1 &mi.48) F ( z ) < 0 = F ( c ) for x < c.9. 2). (13.15 f”(c) = 0.Derivatives 385 The numerator is positive. > ~:=[so~ve(4i*x~3i~~*x=O. for instance.47) (13.
47) means that the graph of f lies first below the tangent and then by inequality (13. 13.386 Introduction to Mathematics with Maple Fig. if f ' has a strict . Similarly. cs C C + 6 Fig.48) above the tangent. 13. concavity f t.10 An inflection point a t x = c Inequality (13.9 Convexity.
7. 0 Example 13. + + + @ Exercise 13.3)6 . 2. b] if and only iff. If f’’(c) = 0 and f”‘(c) # 0 then f has an inffec Proof.4.lox3 6 x .16. on an interval I . g(z) = (x .12. 3. h(x) = x3/(x2 1). Theorem 13.4. + Exercises Exercise 13. then the graph crosses from above to below the tangent. If f”(c) = 0 then f might have an inflection point at c but need not.4 Prove: A differentiable function has an inflection between two consecutive local extrema.6 Prove: f i s convex on [a. I .Derivatives 387 local minimum. Why? [Hint: Consider f ’ ( x ) for x > 0 and for x < 0. or fL . The next Theorem guarantees the existence of an inflection point. @ Exercise 13. b] and is increasing.4.4. 2ax3 + bx2 has an @ Exercise 13.14 The function f : x H 72 x3 has the second derivative equal to zero at x = 0. the third derivative f”’(0) = 6. but not strictly convex.4. 3 is ~zero ~ at x = 0 but there is no inflection. H Exercise 13.2 Prove: I f f is convex.1 @ Exercise 13.3 Find a and b such that f : x inflection point at x = 1.1 Find intervals where the given functions are convex and where they are concave. This is an immediate consequence of the definition of an inflection point and Theorem 13. exists on [a. f (x) = 3x5 . Consequently f has an inflection point at IL: = 0 by Theorem 13. however it is not true that f”(0) = 0.32 4. (x4)” = 4 . then there is a subinterval of I on which f is linear.16 tion point at c.4.5 The function f : x H ZIZI has an inflection point a t x = 0.
12 exhibits the same phenomenon for a parametrically given curve {[f(t).49) I I I a 01 c Fig.11 The mean value theorem b We state this as 1 Theorem 13. that is (13. f ( c ) )on the graph off where the tangent is parallel to the line joining the endpoints.b] I+ R is continuous on [a. Figure 13.b] and differentiable for a < IC < b then there exists a point c E [a.17 (The Lagrange mean value theorem) I f f : [a.g ( t ) ] .11suggests that there is a point (c.49) holds.5 Mean value theorems Figure 13.a 5 t 5 b } .388 Introduction to Mathematics with Maple 13.b] such that Equation (13. 13. We state this as .
The term mean value theorems is not really appropriate. for instance. for constant f the number c can be any number in [a.52) The advantage in using the notation with 0 is that for negative h.50) that2 g(b) . The importance of the mean value theorem comes from the fact that it allows conclusions to be made about the function from knowledge of the derivative. equations like (13. If g’(z) # 0 for a < x < b then it follows from (13. b] H R and g : [a. b] and differentiable for a < t < b then there exists a point c with a < c < b such that The Lagrange mean value theorem is a special case of the Cauchy mean value theorem for g ( s ) = x. The dependence of 0 on f .g(a) # 0 and Equation (13. x and h in the Lagrange mean value theorem is studied in Exercises 13. One writes x and x+ h instead of a and b.49). the emphasis should lie on the increment of the function or functions.1 and 13.18 (The Cauchy mean value theorem) I f f : [a. It suffices to prove the Cauchy theorem. not on the mean value c.50) takes a more convenient form (13.50) and (13.51) Formulae (13. b] H R are continuous on [a. Then c becomes x O h with 0 < 0 < 1. even if nothing more is known about 0 than 0 < O < 1.1[.5.5. The numbers c or 0 are not uniquely determined. . For the proof of the Cauchy theorem we denote 2By choosing f(s)= s.2.51) are often written in a different but equivalent form. but in Englishspeaking mathematics the terminology is firmly established.Derivatives 389 Theorem 13.51) then becomes + (13. b] and 0 any number in ]0. Equation (13. (13.52) hold without any need to change the inequality for 0.
Show that 0 + 1/2 as h + 0. b].3 The Lagrange mean value theorem becomes false i f f fails to be differentiable at one point.22 the function F attains its extrema in [a. (f(z).390 Introduction to Mathematics with Maple 0 Fig.2 Find 0 from the Lagrange mean value theorem if f ( t ) = t 3 . g ( b ) ) .l . Proof. By Theorem 13. [a.5.8 the derivative F’(c) = 0 and this is equivalent to Equation (13.ll. @ Exercise 13. F is zero at the ends of the interval [a. 0@ Exercise 13. Since F ( a ) = F(b) = 0 either the maximum or the minimum is attained at an interior point c.1 Show: If f ( t ) = t2 then 0 from the Lagrange mean value theorem is 1/2.g(a)).12 The Cauchy mean value theorem Readers familiar with analytic geometry will recognize that IF(x)l is twice the area of a triangle with vertices ( f ( a ) . By Weierstrass’ theorem 12. 0 Exercises @ Exercise 13. independent of x. 13. g(x)).50). ( f ( b ) . b]. b] = [1.5.5. [Hint: f(x) = 1x1.
Remark 13. In the above we have two variants. The following theorem describes the general situation. + 13. For complex valued functions the mean value theorems are false. Consider the following counterexample to the Lagrange theorem: f (x)= x3 .z2) and [a. However they were discovered by J.5 Prove: Between two real zeros of a differentiable function there is a zero of the derivative.6 The Bernoulli1’Hospital rule Theorems for calculating limits by means of derivatives are known as 1’Hospital rules.6 If f ( a ) = g(a) = 0 then h provided f’(a). g’(a) exist and g’(a) # 0. However the limit can be calculated as f(x) = (13. who promptly published them under his own name. In the second the limit is equal t o the limit of the ratio of the derivatives a t that point.5. b] = [0. point t o recognise that these two procedures are distinct. In the first the required limit is equal to the ratio of the derivatives a t the point in question. The combination of these two methods is important for extending the rule t o the case where the first few derivatives of f and g are zero a t the point. If we try to apply this result to f i .1 and g(x) = for a = 1 we encounter a difficulty: g’(1) does not exist.5.x 2(x3 . Bernoulli who communicated them to Marquis I’Hospital.1]. We first make a remark indicating that the rule is plausible. It is a rather subtle. but significant. Their importance for practical evaluation of limits is diminished when limits can be found easily by Maple.4 @ Exercise 13. we need the rule in the next section.54) with the result that x+1 lim fi1 ~ = lim  3 215 v m = 0.Derivatives 391 @ Exercise 13. . However.
9(4 (13. . Consequently lim x+a fo = 1.respectively.56) g(a) = g’(a) = . Using the Cauchy mean value theorem gives for some O with 0 < 0 < 1. we now have (i). f ( a ) = f’(a) = * * * = f(. Since lim x+a f (4= lim f” (4 g(x) 5a g”(z) (13.)(a) = 0. gn+l(a) ’ Proof.60) ) by the induction hypothesis.+l (4 respectively. 3f(0) and g(O) are understood as f and 9. + 0 then4 Oh. We proceed by induction now.55) (13. f(”) and g(n) are continuous at a.= g(.392 Introduction to Mathematics with Maple Theorem 13. a an interior point of I and f : I R and g : I H R.)(a) = 0.19 Let I be an interval. 40depends on n but this dependence has no influence on the proof.)/g’(a+Oh.) + 1.57) I exists and is equal to 1 or to f .. and EITHER (i) lim xa f(n+l) (x) g(n+l)(2) =I OR (ii) both f then (a)and gn+l ( a ) # 0 exist xa f (4 lirn 9(x) (13..59) by assumption and by what we have already established. We commence the proof of (i) for3 n = 0. Let h. If n is a nonnegative integer. f’(a+Oh. (13. + 0 and by assumption (i) and the definition of a limit.
20 I f f : I H R is continuous a t an interior point of intervd I and lim f’(x) = I then f is differentiable at a and f’(a)= 1. Theorem 13. The next Theorem has a very similar proof. The derivative can exist at a even if lim f’(z) does not. For sake of completeness we state the next theorem in which the symbol Lim stands consistently throughout the theorem for any of lim xa or lim or lim or lim or lim . we have by (i) that Equation (13. + 0 and the right hand side tends (by assumption) to 1.21 f ‘(4 If Lirn = I then g‘(4 f (41 Lim 9(4 if EITHER Lim f(z) = Lim g ( x ) = 0 OR Lim 1g(z)1= 00 ~~ This theorem remains true if I denotes 00 or 00. Consequently. It says.Derivatives 393 For (ii). By the Lagrange mean value theorem with 0 < 0 < 1.60) is true and by Equa0 tion (13. If h n + 0 then Oh. that the derivative cannot have “jump discontinuities”. .53) applied to f n and g n in place of f and g we obtain (ii). For instance. x+a Proof. 2 # x4a 0 and f ( 0 ) = 0 then f has a discontinuous The Bernoulli1’Hospital rule can also be applied in situation when g ( x ) + 00. if f(z) = x2saw(1/z) for derivative at 0. roughly speaking. the left hand side has also limit I and 0 this means that the derivative exists and is 1 . xTa xJa x+m x+m Theorem 13.
Exercise 13. Prove: If g is monotonic.4 The Bernoulli1’Hospital rule can be extended to onesided derivatives. the theorem does not really give any means of finding this approximation effectively and constructively.7 Taylor’s formula Often it is convenient and sometimes even necessary to approximate a given function by another function.1 in Exercise 12.1.6. but need not do so for a more complicated function.1.5. We have already encountered polynomial approximation in Weierstrass’ Theorem 12.2 x1 Reckless application of the BernoulliI’Hospital rule can lead t o wrong results. .] 13.394 Introduction to Mathematics with Maple Exercises Exercise 13.1)) = 3/2. they have derivatives of all orders and their function values can be computed by multiplication and addition only.1)) = lim(2x/9z2) = 2/9 but clearly lim ((z2 x+l + 2)/(3z3 .22 which deals with a uniform approximation of a continuous function on an interval. For example. which have many simple good properties. Use the methods of this section to evaluate the limits @ Exercise 13. Even computers can handle polynomials better than most other functions: Maple finds all zeros of a polynomial by f solve.6. generally speaking. However. 0@ 0@ Exercise 13. For instance. [Hint: xla For the proof and discussion of theorems of this type see Vyborny and Nester (1989). limf(z) = limg(z) = 0 and xla xla xla limf:(z)/g$(z) xla = l or limfL(x)/gl(x) = l then limf(x)/g(x) = 1. It is easy to work with polynomials. It is constructive and effective. lim ((x2 2)/(3x3 . We encountered Taylor polynomial before in Section 6.20 can be extended t o onesided limits and onesided derivatives . If f is a poly‘However.6. the approximation can be good everywhere.6.3 Show that Theorem 13. which is simpler or has better properties. and given some luck. but approximates the given function only5near a given point. Our study in this section will deal with a different approximation. x+1 + Where is the mistake.
k) :convert (% .Derivatives 395 nomial of degree n then is a polynomial equal to f for all x. In this section we shall be satisfied with the next Theorem 13. t G t 8 . can always be formed. end do: A few approximations. > > > k:=l. in Figure 13.x=a. would be a good approximation of f and then the estimate of would determine how good the approximation is. We illustrate the nature of approximation by plotting some graphs in Maple.x=O. The following lines of Maple code produce the first eight approximations.62) 6Explained earlier for polynomials but applicable generally.polynom) .k). for the function defined below. together with f. It is natural to expect that T. Maple uses the symbol h ( z ) = 0(4(z)) and g(z) = O ( + ( z ) ) to indicate that (13. Let as recall that the command6 taylor (f (x) . > > restart.14 the third and eighth approximations are graphed. while k<9 do t l Ik:=convert(taylor(F(x).22. . which says that Rn is of an order of magnitude smaller than (z .polynom) k:=k+l. In Figure 13.a). regardless of whether or not f is a polynomial.. produces the Taylor polynomial for f of degree k and it can be then used for plotting. are depicted in the next two Figures. If f is n times differentiable at a then T.13 the second and third approximations are plotted. In the last chapter of this book we shall derive powerful but reasonably simple estimates of &+I. F:=x>l/sqrt(l+x).
396 Introduction to Mathematics with Maple X Fig. 13.14 Taylor approximations .13 Taylor approximations X Fig. 13.
= f'"l'(a) = 0. .8. .T n ( x ) = Rn+l(z)= o((z Proof. f(z). f(z)= f(a) + ( 9 + a ( x ) ) (z .+~(u =O ) ~ ' g ( a =O > By Theorem 13. We are now ready for the main theorem of this section. .q.1. I t is a maximum if f n ( a ) < 0 and a minimum if f n ( a ) < 0.1 there is a positive 6 such that f n ( a ) a ( z )> 0 for Iz .63) D~R.22 I f f has the nth order derivative at a then . + . If f ' ( a ) = f " ( a ) = . These symbols do not reveal the point at which ( 4 ( z ) ) the limit in Equation (13. then f has a local extremum a t a.a)n).21 for k = 1 . but they do not imply that o(z) = 0. n .23 Let f : [a.which means there is a local 7The relevant point is 0. . . Theorem 13.u ) ~Clearly .. .22 assuming f n ( a ) > 0.Derivatives 397 or that g(z) is bounded.62) is taken or the interval on which 0 is bounded. We use it for a proof of a theorem on local extrema.b] I+ Iw and n be even. . Dng(a) = n! This theorem is called the Taylor formula with Peano remainder. f '"'(4 #O. with a ( z ) +0. w + + Theorem 13. (13. By Corollary 12. Denote g(z) = (z . This must be understood from the context. Proof.n .a1 < 6 and then f(z)< f ( a ) . but unfortunately the behavior of these symbols might look bizarre to the uninitiated: for instance the relations7 o(z) O(z) = O ( z )or o(z) o(z) = o(z) are correct. These symbols are convenient and widely used. We apply Theorem 13.. for k = 1 ..
24: I f n is even then there is no inflection at a . ) = c1+ 2 ~ 2 (U ~) + + + 3 ~ 3 ( 1 1: * * TZC"(Z .65) This is plausible but it does require a proof. ) = . Theorem 13. . l / d W .8 Differentiation of power series If a function is given by a power series then it is easy to find the derivative. . 13. If f " ( .8 81t follows Vyborny (1987). 2.398 Introduction to Mathematics with Maple minimum at a. 2 ( 1  4 3 . xm(l . (13.1 @ Exercise 13. we skip it. The case f " ( a ) < 0 can be handled similarly.3 for each of the following functions: I. Find the local maxima.b]H R and nbe odd. minima and inflection points Exercise 13.23: I f n is odd then there is no local extremum at a.x ) ' with rn E N and I E N. Exercises @ Exercise 13. then f has an inflection at a. or we can 0 apply the already proved part to f.7.7. with differentiation like that for polynomials: f ' ( . . = j("f'(a>= 0.2 Prove the following addition to Theorem 13. 3. Prove the following addition to Theorem 13. The next Theorem has very similar proof.a)"' + * * .7.24 Let f : [a. f(")(a) # 0.
We need Lemma 13.64) have a positive radius of convergence R (possibly 00). For a given x with 1 x 1 < R choose H such that 1 x 1 Multiplying inequality (13..67) follows from inequality (13.1 .Derivatives 399 Theorem 13. Then the power series d(x) = c1+ 2c2(z .25. h E C and < lhl I H (13. I H [2c.67) Proof..66) ( n ~ " . Without loss of generality we take a = 0.xnll +H < R. this in turn implies that the radius of convergence for d is R.' lI 1 H [2(1x1 + H)" + IxI"]. ~ gives lnc.67) by I c .1 0 Let n denote a positive integer.a ) + 3c3(x  + . (13.66) with h = H .4. 0 Proof.+ ncn(x .66) we use the binomial Theorem 5. (1x1+ H)" 1 + cnlxl"]. Since Inequality (13.< R f'(x) = d(x).25 Let the power series (13. We can now prove Theorem 13. For the proof of inequality (13. For x E C.a)"' + . By the comparison theorem the power series for d converges. It follows from inequal . has the same radius of convergence R and for all complex x with .
Here is an example in a positive direction.69). Substituting x = a into Equation (13.1. being again a power series.70) has the name of Maclaurin series. This leads to n! This series is called the Taylor series of f at a.65) gives c1 = f’(a). A convergent power series has derivatives of all orders. In Chapter 15 we shall prove some general theorems guaranteeing the validity of Equation (13. It is important to realize that we obtained these formulae under the assumption that f was representable by a power series. 0 An important consequence of our theorem is that the derivative. it might happen that a function has derivatives of all orders but is not equal to its Taylor series. Such a function is defined in Example 14. Theorem 13. Example 13.15 The formula is valid for 1x1 < 1. as a power series. that is (13.and by differentiating ntimes leads to Cn = f‘n’(a). consequently f’(z) = d(z). is differentiable. and so on. However. is again differentiable.25 gives . differentiating and substituting again gives c2 = f ” ( a ) / 2 . The special case a = 0. Then the second derivative.400 Introduction to Mathematics with Maple ity (13.66) that The sum on the right hand side is a welldefined finite number and as h + 0 the right hand side tends to 0. This assumption must be justified for concrete functions.
(13.71) It is convenient t o extend the definition of the binomial coefficient to any real n and positive integer k by (13.1 times (1)"'(m (1 x)m + .m + 2 ) ~ ~ .Derivatives 401 and by differentiating m .73) k=O This equation is called the binomial expansion: it resembles Formula (5.9 Comments and supplements The concepts of righthand and lefthand derivative can be further generalized. The difference is that now the formula is correct only for 1 x 1 < 1and the expansion is infinite.~ + ' . in a similar way as onesided limits were to limits superior and inferior. The following four numbers .1)  e . in the following form (1 x )  + = c 00 (13.23 These definitions allow us to rewrite the Maclaurin series for ( l + ~ ) m ~ E. (IC .l)! = C k=ml 00 IC(IC.72) for instance 3/  1. N.30) in the Newton's binomial theorem for a = 1 and b = x. 13.
for a continuous f . The first one is the upper righthand Dini derivate. 13. continuous on some interval.20. There is a theorem for Dini derivates similar to Theorem 13. It can be shown that. So is the set where Df (c) < D+f (c). is at most countable. with similar terminology for the other two.15 All Dini derivates are distinct at c The behaviour of a function with all Dini derivates distinct at c is indicated in Figure 13. D+ Y f / D I I I I 0 C X Fig.9.402 Introduction to Mathematics with Maple are called the Dini derivates. the third is the lower lefthand Dini derivate. A consequence of this is: the set where both the righthand derivative and the lefthand derivative exist and are not equal is countable. the set of points c such that D+f ( c ) < Df ( c ) is at most countable. If f and one of its Dini derivates is continuous on an open interval I then all Dini derivates are equal to f' on I . There is another theorem in similar vein: A set of points of strict local maxima (or minima) of a function. Bolzano constructed a function continuous everywhere and differentiable .
v are consecutive points of D. It follows that f. The functions f. v are consecutive dividing points of D.1]. If f . each piecewise linear and continuous on [0.u > 0 then the intervals [u. with 1 = v . f4 are shown in Figure 13. u + :l] [u ~ 3 lu .74).16 The first approximation to the Bolzano function Now we sketch his construction. 11 1 2 + + [u + 51. The Bolzano’s function f is the limit of . 1 u + 11 7 8 [u +g 7 l .+l(v) = fn(v). = fA = on the interiors of the first and third interval of (13. 13. has been defined and u. very quickly become difficult to graph. f3.74) become subintervals of Dn+l. on the interiors of second and fourth interval of (13.1]. Consequently fn+l is continuous on [0. 13. Let f1 : x H x.1] as follows: D1 consists of [0. if u.Derivatives 403 nowhere.74). First we define a sequence of divisions of [0. We proceed by defining a sequence of functions f . X Fig. then let f n + l ( U ) = f n ( 4 and fA+l fA+l 5 5 f. The graphs of f2.17. v] (13.
1]. accessible to our readers. that is for z E [0. 13. We consider now a very simple example from McCarthy (1953).2 1 0. a relatively simple example due to Van der Waerden.17 The first three approximations to Bolzano's function fn.8 0 . 6 Fig. A famous example is by Weierstrassg. The function is given by gThe definition of the Weierstrass function is also in McCarthy (1953) . For details and proofs we refer to Viborni (2001).404 Introduction to Mathematics with Maple 1. is in Spivak (1967).
A finite derivative cannot % Exercises 0 Exercise 13. 9 . therefore it is possible to choose h N positive or negative in order that x and x h N are in the interval on which g N is linear.g i ( x > = o for i > N since 22ih~ is an integerlo Using this we have The righthand side goes to infinity as N exist.Derivatives 405 where The proof that F is continuous is easy and left for Exercise 1 3 .] 0 Exercise 13. 1 .9. Then we have + (i) g i ( x 1 (ii) g N ( x k h N ) . Then use continuity of 1 C 2n 1 k gn(x).2 Bolzano function. [Hint: Choose k such that C < &/3. twice as long as I h N l .g N ( z ) = .9. Use Maple to plot the third approximation to the 'OThe function saw has period 1 and hence saw(u + m) = saw(u) for m E N. .' 4' (iii) + h N ) . + 00. Let N be a positive integer and h N = &2222N.1 O 0 1 gn(z) i=k+l 2n Prove that F is continuous. The function g N is h e a r N on intervals of length 22 / 2 .
This page intentionally left blank .
However. Theorem 13. 14.1 Introduction We have already mentioned that some theorems from the previous chapter are valid for differentiation in the complex domain.7 makes no sense in the complex domain since the concept of an increasing function applies only to functions which have real values. 13. lz . If z1 and z2 are in S so is 2.4. 0 An easy consequence of this Theorem is: f : C H C is a polynomial of degree at most n if and only if f(n+l)(z) = 0 for z E C. By (iii) of Theorem 13. Let F : t I+ f(Z).7 both F1 and F2 are constant in [0.hence f(z1) = f(z2) and f is constant in S. Then F’(t) = f’(Z)(z1 . Consequently F i ( t ) = Fi(t) = 0. Theorem 14.7 can be extended as follows.2.1] let 2 = tzl (1 . 407 + . this is so for Theorem 13.Chapter 14 Elementary Functions In this chapter we lay the proper foundations for the exponential and logarithmic functions.1].1 and 13. In contrast. We calculate derivatives of these functions and use this for establishing important properties of these functions. I f f’(z) = 0 for d l z E S then f is constant in S .1 For a E C and R > 0 denote by S the disc { z .5 and for the chain rule. for trigonometric functions and their inverses. For t E [0.a 1 < R } or the whole complex plane C. Theorem 13. Proof. part (iii) of Theorem 13.t)z2.3.2 2 ) = 0. for basic rules of differentiation in Theorems 13. F l ( t ) = %ZF(t) and f 2 ( t ) = Q F ( t ) . Specifically.
+ n .= E(+ + E has the defining property of exp.408 Introduction t o Mathematics with Maple 14. The graph of exp is shown in Figure 14. This series converges by the ratio test.2 The exponential function A natural question to ask is: Is there a function which is not changed by differentiation? An obvious and uninteresting answer is.n ! + ..1) in C then there is a constant c such that E(x) = cexp(x). Theorem 14.1) + Y) = exp(z) exP(Y) (14.9. Corollary 11.. the zero function. for all complex x. If exp existed then all derivatives at 0 would be exp(0) = 1 and the Maclaurin expansion of exp would be x2 x3 1 + x + + + * .2) (iii) exp(x) # 0 for all x E C..2 The equation E’(x)= E(x) has a solution exp : C H C satisfying the condition exp(0) = 1. .25 x x2 xnl E’(x) = 1 + 2 2! + 3 . (iv) The restriction of exp to R is strictly increasing and for a nonnegative integer n If E is any solution to Equation (14. Hence a better question is: Does a function E exist such that E’ = E and E’(a) # 0 for some a? The answer is contained in the next theorem.3! . yes. 2! 3! Xn +n! + a * . . Obviously E(0) = 0 and by Theorem 13.1 Proof. so we call it exp and we have a very . say E. (ii) For z . y E C exp(z (14.1. . it therefore defines a function. The function exp has the following properties (i) exp(1) = e.
Hence we have that exp(x+u) exp(x) = exp(u).+ . that is. Since exp(x)’ = exp(z) > 0 the function exp is increasing.. Since x and u are arbitrary complex numbers we can set z = z u and x = y. By (ii) we have exp(z)exp(x) = exp(0) = 1.14. exp(0) = 1 by (ii) and by the intermediate value theorem exp(x) > 0 for all x E R.1 The exponential function important expansion x2 x3 Xn exp(x)= l + x + . To prove (ii) consider H : x H exp(x u)exp(x)..2) follows. consequently H(x) = H ( 0 ) = exp(u). hence (iii). by Examples 10.+ .Elementary finctions 409 1  0 Fig.+ . + + + .+ ..* . Then u = z x and Equation (14. 2! 3! n! (14.13 and 10. 14.4) (i) is obvious from the definition of e. We have that H’(x) = 0.
It follows that f ( k ) ( 0 ) =0 for all k E N as asserted.4) with x replaced by 1/x2. 20 . 0 Example 14.+and therefore (n + l)! >(n l)! .1) in C then [E(z)/exp(z)]' = 0. the derivatives = 0 for all k E N.410 Introduction to Mathematics with Maple For x > 0 we have xn+l exp(x)>l+x+. it is zero for all x and hence it is n o t equal t o f(x) for any x # 0. Equation (14.6) then lim fck)(x)= 0.3) could also been proved using Theorem 13. we have Consequently the limit in Equation (14.6) that f'(0) = 0.1 f(')(O) For f(x) = exp(l/x2) with f(0) = 0. This example is interesting because the Maclaurin series of f converges for all x.6) Since. by Equation (14. It follows from this equation with k = 1 and from Equation (14.p+l + The second limit relation in (14. consequently this function is constant and then E(x) = E(0)exp(x).5) It will be convenient t o consider 1 (14..19.3) follows from the first since 1 If E is a solution to Equation (14. Simple induction shows that this is also true for any derivative f(")(z). Firstly (14.6) is zero. where P is a polynomial. For x # 0 the derivative f'(x) = 2exp(l/x2)/x3 is of the form P(l/x)exp(1/x2).By Equation (14.
Then use it repeatedly to show exp(r) = er for r E Q.3 The logarithm We have seen that the restriction of x H exp(x) to R is strictly increasing on R.l)/x show that the discontinuity at x = 0 is removable and find the Maclaurin expansion of 9. This is justified by the fact that for r E Q.6 Find the following limits 1 (1) lim exp( ). The inverse function to it exists and is called the natural logarithm or simply the logarithm and is denoted by In. The graph is shown in Figure 14. We summarise the properties of In in the next theorem.2 Prove: exp(x)exp(2) = 1. Graph this function. Proof of this is left to Exercise 14.2.1 @ Exercise 14.2.2.] 14. er = exp(r). [Hint: Consider three possibilities for u.2. Prove: exp(nx) = (exp(x))". points of inflection. [Hint: Use the expansion for exp(y) with y = z2. Exercises @ Exercise 14.2.2. Exercise 14.] Find local extrema. @ Exercise 14.5 For g ( x ) = (exp(z) . (2) lim "10 1 exp( l/x) ' x 1020exp(x) . Exercise 14.2.4 convexity and concavity for f(x) = exp(x2).2.Elementary f i n c t i o n s 411 It is customary to write e" instead of exp(s) for any x E C. intervals of Exercise 14. xyo x exp(l/x) .102 200' + + + + (4) 2+00 lim exp(ax) for a E R. (3) lim 2+m 2 3 2 2 . .3 For f(z)= x1"exp(x2) find f'104'(0) and f(1003)(0).2.
In x < 0 for 0 < z < 1. 00[ ontoR. Y Proof.412 Introduction to Mathematics with Maple Fig. (iii) In 1 = O and In z (iv) lim l n z X+OO > 0 for x > 1. In e = 1.8) In . . 14.= l n z . (v) For every x > 0 and y > 0 ln(zy) = In z X + In y. By the inverse function theorem rgln = domexp = R. (i) We already know that In is strictly increasing.3 The logarithm has the following properties (i) it is defined on 10. .is strictly increasing and maps 30.lny. 20 lim l n z = 00. = 00. (14. oo[.7) (14.2 The graph of the logarithm Theorem 14.
For instance. exp(1) = e hence lne = 1. the inequalities for In x follow from the fact that In is increasing... 1 1 In (f(x)) = .In (1 .Elementary Functions 413 (ii) By Theorem 13. (iv) follows from (i).. and since L(0) = In 1 = 0 we have that lnx = L(x) for 1x1 < 1. 0 The second formula follows by writing x/y instead of x.6 on the derivative of the inverse function . Using the fact that in an + + .. We now show t h a t the expansion holds also for x = 1. Example 14. .+ .10) has.In (1 2 2 + x2) f’(4 . Then Inxy = In (exp(u) exp(w)) = In (exp(u + w)) = u + w = lnx + In y.f(x) and after simplification X X (122) (1+x2)1 f’(x) =  _. I 2x (1 x2)di=? + Example 14.. (iii) exp(0) = 1 hence In 1 = 0.x2) .. In other words the right hand side of Equation (14.3 (Maclaurin series for the logarithm) derivative of ln(1 + x) to obtain the expansion We use the The function x 2 x3 L(x) = x . for 1 x 1 < 1. (v) Set x = exp(u) and y = exp(v).2 (Logarithmic differentiation) Sometimes it is easier to evaluate Dln(f(x)) = f’(x)/f(x) rather than Df(x) and then find f’(x). the same derivative as ln(1 x).10) is the Maclaurin expansion of ln(1 x) for 1 x 1 < 1. 2 3 (14.
The series (14.1 Find the Maclaurin expansion of ln( 1 x 1 x2 = (1 .x3)/(1 . 1 + + + + x2). A To achieve accuracy of five decimal places we need Irk1 5 0. converges very slowly. A better convergent series is obtained by replacing x with x in Equation (14. a forbidding task.3.3 Show that In is convex everywhere where it is defined. we have (for 5 > 0) (14. we have (14. ) I m. .12) This interesting series. Find the following limits without using Maple.4 . . (14.5 rough estimate suggests k = 4 or 5 and a calculator or Maple would confirm t h a t k = 5 suffices. The remainder after the kth nonzero term r k can be estimated as follows: Irk1 32k1 32k+l 5 2 2 k + 1 ( l + $ + .414 Introduction to Mathematics with Maple alternating series with descending terms the remainder does not exceed the first omitted term.3.13) To calculate In2 we would set x = 1/3.11) Using continuity of In and sending x + 1 first and then n + 00.3.10) and subtracting the resulting series.13) was used in the past for calculating logarithmic tables and although there is no need to have logarithmic tables now the series is still useful.2 Exercise 14. To obtain In2 with accuracy of five decimal places we would have to add 100000 terms. Exercises Exercise 14. [Hint: @ Exercise 14. . called the Leibniz series.
Elementary finctions 415 (4) lim z+e . .1 14. For a = e we have et =exp(tIne) = exp(t). The symbol ar represents the exponential with base a and exponent r. A simple calculation shows that f'(x) = 0 and therefore f(z)= f(1) for x > 0.14) leads to lnar = ln(exp(r1na)) = r l n a .4 The general power If a > 0 and r E Q then ar = exp(r1na). and Ea :xI+ ax for a > 0. as an example we prove part (v) of Theorem 5.2.1.x . There are two functions naturally associated with Equation (14.14) to define the meaning of ar for a > 0 and any r E R.14) To prove this we define f(x) = xr exp(r In x) for J: > 0.e lnx .8 becomes valid for r E R and s E R. We now use Equation (14. The proofs are easy. that is aras = exp(rIna)exp(slna) = exp ( ( r + s) In a ) = ar+'. r E R.14). (et)' = exp(rln(et)) = exp(rt) = ert. x E R. Theorem 5. For the In function the Equation (14. p : x + + x r for x > 0. (14. Equation (14.14) follows by setting x = a and using Exercise 14. With this definition.8.
~ . Using the chain rule A consequence of this formula is that Equation (13. (14.1) . + z)"M(x) = M ( 0 ) = 1..15) and the definition of binomial coefficients in (13. ( Consequently M satisfies (1 ((1 2)"M(z))' Hence (1 + + z ) M ' ( z )= rnM(z).n + 1)(1+ x ) ~ .1.16) and we want t o show that M ( x ) = (1 x ) ~ The .4 (Maclaurin expansion of (1 x ) ~ )Using Equation (14. '> + )I. and finally .416 Introduction to Mathematics with Maple We are familiar with p for a rational r and have already mentioned that the usual properties expressed in Theorem 5.38)' namely Dn(I + x)" = m(m . Using this we have = (1 + q m .:( =m J .15) now holds for any real m..72) we are led t o the following Maclaurin expansion of (1 + + x)m (14.(m . series on the righthand x 1 < 1 and diverges for 1 x 1 > 1 by the ratio test. Example 14.l [mlM(z) + (1+ x ) M ' ( x ) ] = 0. n E N and x > .8 extend to real r. Differentiating M gives + The coefficient of x k equals rn [(".16) converges for 1 Hence we assume 1x1 < 1 for the rest of this example. side of (14. We prove only the 'new' formula ( a ' ) ' = rz'l.
14. Employing the chain rule again (ax)'= [exp(zIn a)]' = exp(z In a ) In a = ax In a The rest follows from (i) routinely. 0 The function inverse to E. It exists for a > 0.3. The graphs of E. for a = 5/9 and a = 9/5 are shown in Figure 14.Elementary Functions 417 We now turn our attention to the function Ea.3 The general exponential function ~ ~~ ~~~~~~~~ Theorem 1 4 . I 0 Fig. Proof. 4 The function x I+ ax has the following properties (i) (ax)'= ax In a (ii) It is strictly increasing for a > 1 and strictly decreasing for a < 1. For a = 10 it is called the common . (iii) it is convex. a # 1. is called logarithm with base a and denoted by log..
log. x > 0 prove log. To obtain a formula for log. see Exercises 14. A . (14. b # 1.= Inx 1. Compare with + Find the limit of x" as x + 0 and as x 00.1 Exercise 10. @ Exercise 14.19) X X' = xx((Inx+ 1).3 and 14. . Readers are most likely familiar with this function whose defining property is x = 10'ogx.5 Find Dxfi. We have InX = x l n x .4.5 W e employ logarithmic differentiation t o find the derivative of X : x H xx.5 14. are similar to and proved by the properties of In. x .418 Introduction to Mathematics with Maple logarithm and denoted by log. xy = log.4. indeed. Exercise 14.4 If x. a # 1. : t . prove + log.4. a log.4. x. Y. x X # 1. Example 14.4. r E R.z =  @ Exercise 14.3 If a > 0. y.17) (14.4. b > 0. x' = r log. we solve the equation y = ax with respect to x. Clearly lny = x l n a and consequently log. In t In a The properties of the function log. a are positive.4. it is periodic. Trigonometric functions There was a good reason for restricting exp to R when dealing with logarithm.13.4. Exercise 14. + Exercises @ Exercise 14.2 Find the limit of (1 + x ) ' / ~for x + 0.= log. X' . log. Y. Y log.18) (14. The function exp is not injective on C .
It follows from Equation (14.2] with cost0 = 0. Therefore I exp(it)12 = I exp(it)(lexp(it)( = I exp(0)l = 1. (v) For t E R the function t H exp(zt) maps I R onto the unit circle in the complex plane. t o is uniquely determined and it is the smallest positive zero of of the function + ‘Readers are most likely used to a different definition. There exists a to E [0. (ii) For t = 2 the series (14. For 0 < t < 2 the derivative (cos t)’ = . (ii) There exist a smallest positive number.4) that exp(it) and exp(it) are complex conjugates. Therefore. which shall be called T.21) (sin t)‘ = cost (cost)’ = . In the rest of this section t denotes a real number.20) (14. hence cost strictly decreases. (14. + + It is easy to see that if f is periodic then f(z mp) = f(z)for m E Z.22) (14.20) alternates and the terms decrease.2 16/24 < 0.Elementary Fbnctions 419 function f is said to be periodic with period p # 0 if (i) z E dom f + z p E dom f . Consequently cos 2 < 1 . (iii) z H exp(z) is periodic with period 2x2.sint.23) Proof.5 erties The exponential function has the following prop (i) I exp(it)I = 1 for t E R. such that exp(z~/2) = 2. . later in this section we reconcile our definition with the ‘school’ definition. by the intermediate value theorem. The periodicity of exp is a part of the next theorem.sin t < 0. (iv) exp(2mz) = 1 if and only if z is an integer. It follows from the expansion of exp that (14. However. + Theorem 14. (ii) f ( z p ) = f(x) for every x E dom f. Before we start proving the theorem we definelcost = %exp(it) and sin t = 53 exp(it).
n/2) > 0. Since cos is continuous and decreasing on [0. If exp(2tnz) = 1 then exp(tnz) = exp(tnz) and since these are complex conjugates it follows that Sexp(tnz) = sin(tn) = 0. b 2 0. + ~~ 2cos(t . n/2] and by (i) we have sinn/2 = 1.420 Introduction to Mathematics with Maple cos.18). for2 t > 1/2 also sintn = cos(t .2.2) that exp(nz) = zz = 1 and exp(2nz) = (l)(I) = 1. n/2] there exists t o with cos t o = a. (v) Let u = a zb. By Equation (14. since exp is increasing on R. (iii) now follows from Equation (14.n/2)= exp((t .4 and in particular from Equation (7. Since y = LyJ t with 0 5 t < 1. Since it follows that x = 0.n/2) + zsin(t . (v2) Let a < 0. Now we have filled the gap from that subsection by proving the existence and uniqueness of argument (2). + + 0 Given a2 + b2 = 1 we proved. So we have exp(zn/2) = 2.n/2)2) = zcost ~~ + sint. the existence of $ such that cos$ = a sin$ = b n<$<n Part (iv) asserted the uniqueness of $. For a complex number z # 0 we have z = IzI(cos 4 zsin$). For an integer n we have exp(2nnz) = exp(4nm/2) = 24n = 1. b 2 0. Then b = sinto > 0 and a zb = exp(zt0) = u. . By (vl) there exists tl with def + Consequently u = exp(z(t1 n/2)). The only possibility is t = 0. In 4 we recognize the argument of z from Section 7. it suffices to show that exp(2tnz) = 1 implies t = 0. in part (v). + + (vl) Consider first the case a 2 0.22). (v3) If b < 0 then by (vl) or (v2) there exists t 2 with u = exp(zt2) and then u = exp(z(t2 n)). It follows by successively using Equation (14.2). IuI = 1. the function sin increases on [O. y = S z . However sintn > 0 or 0 < t 5 1/2. (iv) Let x = %z. We denote it n/2.
We know that eit and eit are complex conjugates. The latter is far more advantageous and we leave it as an exercise to compare both solutions.24) This is an interesting relation which ties together four of the most important numbers of mathematics. We have also proved sin2 t + cos2t = 1. either Theorem 13.5. so it follows that eat + ezt cost = eat .eat 2 ' (14.rra  1. (14.4.sin z sin y + z(sin z cos y + cos z sin y) 2 + sin z)(cos y 2 (14. The graph of sin and cos (with other trigonometric functions) are shown in Figure 14.26) sint = 22 These are known as the Euler formulae. Example 14.27) The formulae for sin(z y) and cos(z comparing real and imaginary parts in cos(z + + y) with z. z and 1. We have established that sin and cos are periodic with period 27r.Elementary Functions 421 We have proved that e2.6 by Limit of (sin(2z3 + z 5 ) / z ( l . 7r. (14. e. otherwise an insidiously wrong result is obtained.19 or power series can be used.25) (14. 31t is important to enclose the whole denominator in (). y E IW are obtained by + y) + zsin(z + y) = exp(z(z + 9)) = exp(zz) exp(zy) = (cos z + sin Y) = cos z cos y .1. .28) For some other formulae for trigonometric functions see Exercise 14.cosz) can be easily found 4 Without Muple. All properties of trigonometric functions follow easily from our definitions or from what we have already proved.
lim tant = 00 t+.422 Introduction to Mathematics with Maple 3 = tanx sin x 0 K cotx / 27r = cosz Fig. 14. The graph is shown in Figure 14.4 Trigonometric functions The function tan can now be defined as sin x tanx = cos x ’ and all its properties developed using sin and cos.lr/2 and tiIr/z lim tant = 00 .4.
5 with Figure 14.O) in the anticlockwise direction we mark a point B on the circle with the length of the circular arc between A and B being a!.tan(a: .6. t ] into n . The school definition In the x. where exp(it) and sint and cost are depicted according to our definitions.5 The school definition Comparing Figure 14. sin a Fig. see Figure 14. We divide the interval [ O . We call this the school definition.y plane we construct the unit circle (see Figure 14.4.n/2) the graph of cot is the reflection of the graph of tan on the yaxis and a shift by n/2 to the right. 14. Since cot x = . it is sufficient to show that t is the length of the circular arc between 1 and exp(tz). Then we define cos a! to be the xcoordinate of B and sin a! to be the ycoordinate of B.Elementary finctions 423 Further 1 (tanz)' = 1 tan2x = cos2 x * + We define cotx = and have (cot x)' = (I def cos x sin x + cot2 x) = I/ sin2x. we see that in order to show that both definitions agree. Moving from A = (1.5).
n. .424 Introduction to Mathematics with Maple Fig. The length of the polygonal line inscribed in the circle between 1 and exp(it) is (k . . .6 Definition of sin and cos. the absolute value on the right hand side of the last equation can be omitted.54) that lim x+o =t x sin xt and consequently In = 2nsin (k) + t as n + oo. It is an easy consequence of the definition of sin or of Equation (13. 14.exp = 2 lsin Since 0 < t/2n < 7r/2 for large n . 2 . As n increases without limit the length of the polygonal line In tends to the length of the circular arc between 1 and exp(tz). k = 1 .1)tz k=l Since = lexp (E &)  z l Iexp (2) (g) I (&)1 . . parts by dividing points kt/n.
for JiT2 1 x 1 < 1. Its inverse is denoted by arcsin. Graphs of sin and arcsin are shown in Figure 14. xy x+y cos sinx .1 Prove the following formulae for x . (14. cos 2t = cos2 t . We now prove 1 Darcsinx = . .7. y E R: sin 2t = 2 sin t cos t . .sin y = 2sin 2 ' 2 x+y 2y c o s x . Prove it! 14.3 The function f : t H t2 sin( l/t) for t # 0 with f(0) = 0 has a derivative for all real t.5.c o s y = 2sinsin. The function sin is strictly increasing and continuous on [7r/2. The restriction4 of it to this interval is also continuous and strictly increasing and maps [ 7r/2. When dealing with the inverse it is important to realize that it is an inverse to the restriction of sin to [n/2.1]./2].1 dx dsiny COSY' d Y 41t is a common practice to denote this restriction also by sin.29) By the theorem on the derivative of the inverse we have 1 d arcsin x .sin2 t . 7r/2].1. /dn /F.5.Elementary Functions 425 Exercises @ Exercise 14.7r/2] onto [ .5. We have sin(arcsinx) = x for every x but arcsin(sinx) = x only for x E [7r/2.7~/2]. 2 2 Exercise 14.6 Inverses to trigonometric functions.cos t 1 cost @ Exercise 14.2 Find lim t4 cos 5t . The derivative is discontinuous a t 0.
continuous. 1 dc?. continuous.7r/2.7r/2[and continuous. Its inverse is called arccos.~/2. For the derivative we have D arccos z = .8. The function tan is strictly increasing on ] . The graphs of arccos and cos are shown in Figure 14.29). The graphs of tan . hence cosy = & d m .7r/2[. and maps Iw onto ] .29) is proved. The function cos restricted to [O. T ] . It is strictly increasing. hence cosy = d m and (14.n] is strictly decreasing. The inverse to tan restricted to this interval is called arctan.1] onto [0.Since 7r/2 < y < 7r/2 it follows that cosy > 0.7 sinz and arcsinz Now siny = z. and maps [1. 14. which is strictly decreasing.426 Introduction to Mathematics with Maple lr  2 y = arcsinz 7T  2 0 1 $ 1 Fig. The proof is similar to establishing (14.
x6 + .10. .n[is strictly decreasing.31) x3 x5 x+ + .cos2y X d tan y dY = 1 1 1 tan2 y 1 22' + + (14. Darctanx = 1 .8 cosz and arccosz and arctan are shown in Figure 14. To obtain the Maclaurin expansion of arctanx is easy. (14.7r[. .30) The function cot restricted to ]O. continuous. which is strictly decreasing. Its inverse is called arccot.9. As we have done before we use the expansion for the derivative. The graphs are shown in Figure 14. and maps R onto ]0.x2 The power series + x4 . For the derivative we have the formula d arctan x _. x7 3 5 7 (14.32) . 14.1 .Elementary finctions 1 I 1 0 1 1 Fig.
428 Introduction to Mathematics with Maple 7r  2 Fig... .6 contains a guideline for efficient evaluation of 7r. obtaining + + 7r 1 ..9 tan x and arctan x has.. 14.6. 4 3 5 7 (14.1 1.34) This series converges very slowly. (14. Exercise 14. the same derivative as arctan and both function are equal at x = 0.33) 3 5 7 By the same method we used to prove the Leibniz series for log2 we can prove that this equation also holds for x = 1. =1+ +. for 1x1 < 1. . x3 x5 x7 arctanx = IC . so they must be equal for 1x1 < 1.. . ..
14. for x for < 0.6.4 Prove that: Darccot x = l/(l Find f' for + x2).6. for x > 0.Elementary Functions 429 . y=cotx 0 Fig. .3 Exercise 14.2 arcsinx IT + arccosx = for 2 1 7 r arctanx + arctan() = x 2 1 7r arctanx + arctan() = 2 X 1 5 x 5 1. Prove the following identities  @ Exercise 14.1 Use Maple to graph arcsin(sin(x)) on [37r. @ Exercise 14.6. 3 ~ 1 . arctanx = arcsin X ~ diT2 x E R.10 cot x and arccot x Exercises Exercise 14.6.
2 Find D ( cosh x sinh2 x + In (tanh z)) . cosh2x Maple knows hyperbolic functions. . (tanhz)' = .5 Prove lim x+o . and the series for the arctan would converge quickly. 2 sinh(x) cosh(x) Exercises Exercise 14.7.7r/4 = 4 arctan 1/5 .6. To evaluate 7 ~ / 4to ten decimal places it suffices to take seven terms in the series for x = 1/5 and three terms for x = 1/239. (It can be expected that . 14. From a purely mathematical point of view they offer nothing new. + > expand(sinh(2*x)) . although in some formulae the sign and . Exercise 14. 1 (cosh 2) = sinh 2.7r/4.O = 4 a .430 Introduction to Mathematics with Maple @ Exercise 14.7.e" 2 ' def ex e" tanh = ex ex' cosh2x sinh2 x = C O S sinhx = def cosh x = def ex + e" 2 ' + + cosh2x . Prove this and and confirm it numerically with Maple.sinh2x = 1.O and tan . Set a = arctan 1/5 and .are opposite to what they are in corresponding formulae for trigonometric functions. interestingly they display striking similarities with trigonometric functions. We have + (sinhx)' = coshx.arctan 1/239. For instance ~ ~ X .1 Show that sinh is strictly increasing on R and find an explicit formula for its inverse.6.O would be small.) Show that tan @ = 1/239. x arctan x 0@ Exercise 14.1. ex .7 Hyperbolic functions The following functions are often used in applications.6 By the previous Exercise 4 x arctan 1/5 E 4/5 M n/4.
Initially Riemann’s theory was a great success. E [ t k . Henstock is a contemporary British mathematician who made significant contributions. Sometimes it is referred t o as KurzweilHenstockl theory. L. Later it was found deficient in many aspects.1 Intuitive description of the Integral We considered tagged division of intervals in Section 12. 2We recommend this book for further reading on the theory of the integral.l . unfortunate as it may be. Kurzweil. Cauchy. b] a = to < tl < t. Our presentation follows. Lee and Vfbornq (2000.15). particularly in Inequalities (12. However. The symbol T X will denote the tagged division with dividing points T and tags X . in parts. Chapter 2). 431 . t k ] . =b and X the set of tags X I . it is this theory which is usually presented to first year classes and dealt with in books on calculus.Chapter 15 Integrals In this chapter we shall present the theory of integration introduced by the contemporary Czech mathematician J. His definition of an integral was similar but more innovative than the one developed previously by A.2 15. The first modern integration theory was developed by the German mathematician B. Riemann made a decisive steps forward by considering the totality of functions which can be integrated and building a systematic theory.5 of Chapter 12. Throughout this chapter T will be a set of dividing points of an interval [a. Riemann.
a 5 z 5 b. If no confusion can arise we shall abbreviate R ( f .ti1) This sum is called a Riemann sum. . as a tagged division becomes finer and finer the sums R approach the area of the set {(z. for a nonnegative f . We shall only do this if the objects left out in the abbreviated notation (like T X in R ( f ) ) are fixed during the discussion. Our intuition leads us to believe that. Fig.432 Introduction to Mathematics with Maple i Definition 15. 0 5 y 5 f(z)}.T X ) = 1f ( X i ) ( t i i=l n i .y).b]H C and a tagged division T X we associate the sum R(f. 15.1 for a nonnegative function f .1 (Riemann Sum) With a function f : [u. The sum R is simply the area of the rectangles which intersect the graph of f . or more explicitly the Riemann sum corresponding to the function f and the tagged division T X .1 A Riemann sum The geometric meaning of R is indicated by Figure 15. T X ) to R ( f ) or R ( T X ) or even just R.
a}. where f : x H x2 t.” + + Example 15. in concrete situations when the function f is given by a formula. For any tagged division T X . 0 5 y 5 l} despite the fact that points (x. it is fairly clear what the area of the interior of a simple polygon should be. If we try to graph the set S we would blacken the whole square Q = {(x. namely b one writes la f(x) d x instead of f .” sa + is the variable of integration and t is a constant. namely 1.T X ). Hence f(x)dx = f ( u )d u = Jab f ( a )d a . 0 5 y 5 1. The notation f is unambiguous and in theoretical discussions preferable. A tentative attempt at definition of the integral We shall say that the function f is “integrable” if the sums R ( f . the concept of an area of an arbitrary set in R2 has to be defined and the theory of area of sets in R2 (or R”) should be developedwe cannot merely rely on our intuition.Integrals 433 A word of caution is needed here. Also.T X ) approach a limit as T becomes finer and finer can be studied independently of their geometric meaning.” s.” s.” s. Ja (x2 t ) d x denotes f .y) with irrational y do not lie inside S. We shall see later that this notation facilitates the use of an b important theorem.” s. b s.I becomes arbitrarily small when T X is sufficiently fine. for instance Jab x3 d x. for example. for instance f(x) = x3.1 Let f be a constant function f(x) = c on [u. . one may conjecture that the area of S is the same as the area of Q. y E The set S is fairly complicated and some readers may have no idea as to what the area of S should or could be. However. The question of how and when the sums R ( f .” s.~) 0. 0 5 II: 5 1.b].” s. It can be replaced by other letters without altering the meaning of the symbol. in (x2 t ) d x. T X ) approach a number I as T X becomes finer and finer. Such a conjecture would be completely false. the symbol d x indicates that x s. However. Just reflect on the following question: what is the area of the set S = {(II:.another notation is more common. 5 x 5 1. Firstly. One can also say that R ( f . I is called the integral of f from u to b and is denoted by f . The reasons for using the notation f(x) d x are historical and of convenience. The letter x in the symbol f(x)d x is called the variable of integration.y). Since S is so dense in Q. That is the goal of this chapter.
regardless of the division T and b b f = c(b .a). .x=0. right sum and middlesum for numerical evaluation of Riemann sums.. rightend and in the middle of the subintervals. dz=ba.2) . They are leftbox. Maple has several commands for Riemann sums.a ) .x=O. > with(student1: leftbox(x^3.so clearly haveJaldz=Ja b b s .rightbox(x^3.434 Introduction to Mathematics with Maple n1 n1 i=O i=O All the sums R ( f . If the optional number of intervals is not given Maple subdivides into four subintervals. s . All six commands require the package student to be loaded. T X ) are equal t o c(b . The basic interval must be given and. For c = 1 we the tags X .a ) or c d x = c(b . optionally. 15. rightbox and middlebox for plotting the function and the rectangles representing the Riemann sums. > X X Fig. the number of subintervals as well. right and middle indicates that the tags are at the leftend.2). There are corresponding commands leftsum.The left.2 Leftbox and a rightbox for a Riemann sum .
.100) ) . .x=O . 6 4 2 0.6 18 2 O' 02 04 06 08 1 X 12 14 18 IS 2 Fig.2). > rniddlebox(x^3. Increasing the number of intervals from four to ten improves the approximation.4 0. .2 0.6 0.4 1. Let us try a division with 100 subintervals.999800000 In the next example we look at general tagged division with arbitrary tags.980000000 The correct value is 4. 15.3 Middleboxes for Riemann sums The command middlesum evaluates the Riemann sum. > middlesum(x^3.2.Integrals 435 For a monotonic function the middlesum lies between the leftsum and the rightsum and gives a better approximation. .2. 3.x=O.x=O.8 1 X 12 1.2. 3. > evalf (middlesum(x3.10) .10).x=O. .middlebox(x'3. > evalf(%>.
a2) . . I].2 If f(x) = x on [a.t?l) (tf . . We estimate R ( f . The Riemann sum R ( f .a2). We define X = Max((ti . for a specially chosen XI say X = Y with it is easy to evaluate R ( f T . We now show that if T is sufficiently fine then R(f. TX) approaches i ( b 2 .ti 2 1 2 + t. Y ) = (b2 . Let f(x) = 1 [+ 1x2] . then clearly 1% . . .u 2 ) . and [a.ti& i = 1 . R(fT .(yZ .t i .&)(ti . T X ) = C:==. n ) . Y )= f = = 2 i=l .(b2 .tn2 2 2 + t.1 ) . T Y ).b] then the evaluation of R ( f .3 = [xdx =Z 1( b 2 .u2)/2. However.TX)for any Xis arbitrarily close to (b2 . We shall try to obtain more insight into the approach of the Riemann sum to a limit by considering one more example. 2 .tnl) .436 Introduction to Mathematics with Maple Example 15. Y ) for any T . T X ) depends on whether or not 0 is a tag. We . T X ) is not obvious and R depends on the tagged division T X .u2)/2.Xi1 5 X and If X is small then R ( f TX) . Hence [f Example 15. Since f(x) 1 is 1except for x = 0.R ( f . . geometric intuition tells us that Jl f = 2. As X becomes smaller the tagged division becomes finer and R ( f .  tf + * * * .b] = [1. is close to R ( f T . 1 c (t?.
3 X i#O for only the interval [ t k .101 < E holds. t~. = XI. is]tagged by zero.1] by (15.a2)1 < E .1) S(x) = 2 S(0) = 2 E for x#0 If the tagged division T X is Sfine then the length of the interval which is tagged 2 Equation (15.4.1 With notation as in Example 15. Use the middlesum with 500 subintervals. Define a gauge 6 on [1. O X ) . Exercises @ Exercise 15.2 Use Maple’s boxes for sin x on the interval [0. a gauge S such that for every Sfine tagged division T X the inequality I%L(f.1.] . T X ) .3 Given E > 0.+^ = 0. If one looks a t Figure 15. 101.$(b2 . for the function f : x 1x1 .1x1 and the interval [0. find.2 1 < E. & Exercise 15. (iii) X I . zero tags two adjacent intervals. (ii) XI. How can we ensure by one simple condition that I R ( f . When the function t o be integrated is unbounded the requirement that some intervals in the tagged division are substantially smaller than others becomes essential.Integrals 437 distinguish three cases: (i) X I . [Hint: For an integer n let S(n)= e/22. = 0 no subinterval is tagged by zero. This is the key for making our tentative definition of the integral precise.1.l . A tagged division is by zero is less than ~ / and sufficiently fine if it is 6fine.1) is satisfied. Remark 15. 27r].2 if X < show ba that I R ( f . We shall see this later in Example 15.1. 0 Exercise 15.1 such a requirement is quite natural. # 0.1 In the above example it was convenient to have some intervals smaller than others. T X ). i # k.
Assume that for every positive E there exists a gauge 61 such that I n ( f . I Remark 15.3) . (See also Exercise 15.438 Introduction to Mathematics with Maple 15. with the function 6 defined in that example.2 it was good enough to interpret the phrase T X is suficiently fine to mean that the length X of the largest subinterval [ti+l. The difficulty lies in making precise the phrase suficiently fine. In Example 15. T X ). T X ).3 it 2(b . ~ /Definition 15.1). We shall abbreviate the saying that T X is &fine to T X << 6. sufficiently fine meant &fine with a suitable gauge 6. or just integrable on [a. T X ) . b of course. namely I = Ja f = s .1 There is at most one number I satisfying the condition from Definition 15.I . So we can say that in all our examples.1. ti] was smaller than & / ( b. We shall.I I < & whenever T X << 61.JI <& (15.a ) .2. In Example 15. Proof.2 (Integral) The number I is said to be the integral of f from a to b if for every positive E there exits a gauge 6 such that \ I n ( f J X )I I < E whenever T X << 6.a) was convenient to require that some of the subintervals of a tagged division were substantially smaller than others and this was ensured by requiring that the tagged division was &fine.: f (4d xf is said to be Riemann integrable. This can be rephrased as: the tagged division & is 6fine with 6 = . and also that there exists a gauge 62 such that (15.I ( < E for all sufficiently fine tagged divisions T X .b].2 function then If the gauge in the above definition is a constant positive Theorem 15.2 The definition of the integral When attempting to define the limit of Riemann sums it is easy to make precise that part of the informal statement concerning the smallness of R . If a number I satisfying this condition exists then f is said to be Kurzweil integrable.2) l R ( f . We can say: The number I is the integral of f if for every positive E we have l R ( f . keep the notation introduced earlier.
Define S as follows S(x) = 1 E if x is irrational. T ~ .4 Let f ( x )= 0 for x E [0.a2). (15.4) by Example 15.a ) as defined in Example 15. Then By letting E + 0 we obtain 1 1 . and t h a t Ja 1~ = 0.3 that [ cdx = c(b .3 for (15.5) by Example 15. Let 6 ( x ) = Min(61(2).4) (15.2) and (15. . for k E otherwise. Note that IQ differs from a function identically equal to 0 only a t rational points. Example 15.5) (15. .Integrals 439 whenever T X << 62.We show that f Jt f = = 0. . for 6(fk) = 2k+2 i = 1 . Clearly we have that S < 2~ some positive integer N . Let T ~ .6).1] and positive E choose S(l/k) = 2k2k2~ if x = l / k . x # l / k . S(x) = for (15.1.2 formally it is sufficient to choose S(x) = 1 for & (15. 2 . that is I = J . .. For every N.62(2)) and T X both Equations (15. T z .JI 5 0.a ) .2 and 15. Let E be a positive number.1]..3) hold and consequently S.1. 0 It follows from considerations in Examples 15. b]. . . be an enumeration of all rational numbers in [a.and S(x) = 1 where S is the sum of finitely many numbers which have the form f ( X i ) ( t i &I).2. k E N and f ( l / k ) is integrable on [0. each is zero or is less than k22S(l/k) 5 2"'~ and there are a t most two terms for the same k. b Example 15.5 We prove that 1~ is integrable. 15.6) [ x d x = (b2 To satisfy Definition 15. 1 . and S 2(b . Let TX << 6 then E k2 for k N. Consequently S <E and So1 f = 0.
440 Introduction to Mathematics with Maple If T X << S we have 0 5 R(TX) = c n lQ(Xi)(ti .. Note that this choice of S implies t l . (2 + .+ ’> 2N .A].7) i=l The right hand side of this equation is a sum of finitely many numbers which have the form l~(Xi)(ti . Similarly to the previous example we have that S <E Hence R(TX)5 E . This together with Inequality (15. each is zero or smaller than 2 ~ / 2 ~ for + ~ some k . There are a t most two terms for the same k . nobody w a s able to explicitely define one. This happens because authors of those textbooks use the Riemann definition of integral. . s. However.t o < 2~x1 < XI). so there is no chance of encountering one in applications. Define 6(z) = E X for x # 0 and 4> E > 0 and be a Sfine partition of [O.til).6 (An integral of an unbounded function) to prove t h a t We wish (15. A great advantage of the Kurzweil integral is that practically every bounded function is integrable. (15.for 1 S(0) = E ~ Let .ti1).8) where f(z) = .for some positive integer N. fi z # 0 and f(0) = 0.7) proves t h a t exists and equals 0.” IQ The function from this example is often quoted in textbooks as an example of a nonintegrable function. Example 15.Since X1 = 0 (if X1 > 0 then X1 5 tl = and 31t can be shown with the axiom of choice that there is a bounded function which is not Kurzweil integrable.
E).9) For an estimate from below note that for i 2 2.f OS Y 0. 15.Integrals 441 0. we have (15. (15.V i=2 G ) = 2JiE(&i .4 A Riemann sum for a &fine tagged division for Xi # 0.10) . Hence n i=2 n 2 2drEX(&.l 02 0 02 04 06 08 1 Fig.
9) and (15. For a better plot we used f / l O instead of f.4.3 If a < c < b and 6 is a gauge on [a.7 Let f be defined by f(x) = . Maple divided the interval into 43 subintervals and calculated the Riemann sum R = 0. b] then there exists a 6fine tagged division having c as one of its dividing points. say a = t j for some j.for x # 0 and f(0) = 0.8) follows from Inequalities (15. On the other hand Xi 5 Xj 5 a for i 5 j and therefore 1 and since t l < X1 + S(X1) < 2a/4 = I > K . 2a This contradicts Equation (15. a / 4 ) for x E [0. 22 We show that f is not integrable on [0.32.1788903398 Since Remark 15. 1 Example 15.190776588 and 0. Choose a such that 0 < a < 1.1] containing a as one of the dividing points.442 Introduction t o Mathematics with Maple E is arbitrary.11) and proves that f is not integrable. The graph is shown in Figure 15. and l/a > 2K. Then TX is also 61 fine and therefore Equation (15.c] and [c. Then there exists a number K and a gauge 81 such t h a t whenever TX << 61. for an indirect proof.1]. Equation (15. Let TX << 6 be a tagged division of [0. and define S(x) = Min(61(x). b] and then merge them together. By contrast middlesum and rightsum produced with 43 subintervals 0. Assume.198134. To see this it is sufficient t o find 6fine tagged divisions of [a.10).0001 and 6(x) = 0. We programmed Maple t o construct a &fine partition and the corresponding Riemann sum with 6(0) = 0.11) holds.1]. . that it is.
Maple displays the expression to be integrated with an integral sign.x=O.x=0. then Maple does so. and does no calculations. 8 > int ( (x3)*exp(xn2).1 Integration in Maple Maple evaluates integrals efficiently and with ease.5163007634 .e4 Sometimes Maple cannot calculate the integral exactly. That is. In the next Maple session we calculate JdzT . > int(3*xn2.2. . I 2 3x2dx. so Maple does no calculations.2) . We can use the evalf 0 function to obtain a numerical answer. the integral is rewritten.Integrals 443 15. Use the evalf 0 function to give a numerical answer.I). If the integral can be calculated exactly. The int 0 function can be used to calculate integrals. Consider following example. If Maple cannot calculate the integral exactly. then rather than giving an answer.2). This integral cannot be evaluated int(i/(x+exp(x)) . . The range of integration needs to be specified and then the integral is calculated. . We wish to evaluate 3 2 exactly. 2 sin2x d x and x3ex2dx . > evalf(%). > [A d x .. the integral must be calculated numerically. Instead.. x=1 .
2.2.cosyil 5 Ixi .2.>I # c and @ Exercise 15.for 22 J: # 0. [Hint: This is obvious if f(c) = 0. Prove that 9: x H f ( A x ) is integrable on [ a / A .7.1 Exercise 15. and a 5 c 5 b.2 Prove that the characteristic function of an interval [a.2.2.b] is integrable and its integral from a to b is p .6 integral. x # p. State a similar result for A < 0 .1 Prove that g : x H f ( x + A ) is integrable on [aA. S(a) = S(p) = e/4.b/A] if and only if f is integrable on [a.1 Let A > 0.b]. 1 1 .3 @ Exercise 15.a.1 Let f ( x ) = 0 for all x E [a. bA] if and only i f f is integrable on [a. [Hint: Define S(x) = Min(1x .PI) for x # a .4 f = 0.sin a.p] C [a. Il n ~ ( C O S X cosyi)(ti ~ .5 l cos z dx = sin b . [Hint: Do not work hard. use Example 15. Show that f is not integrable on [1.Ix . and f(0) = 0.b] except possibly for x = c.2.444 Introduction to Mathematics with Maple Exercises Exercise 15. Prove that f is integrable and l @ Exercise 15. Since I cosxi . otherwise take S ( x ) = 1 for x S(C) = A.b]. [Hint: Use S for f to define S for 9.] 1 Let f ( x ) = .ti1) . [Hint: Use S for F to define 6 for 9. Prove the following formula using the definition of @ Exercise 15.] 2(lf (.a [ .yil by the .
b] and S”.2 If f is integrable on [a. There exists a whenever TX << S.f cS” = a a f. and (15. Theorem 15. Let c # 0.b] then so is f +g (15. It follows that Remark 15. (15.3 and Iff and g are integrable on [a.13) Proof. # 0 then f is also integrable and Theorem 15.12) Proof. that For every positive E there exist positive functions 61 and 62 such (15.15) . positive function S such that E > 0.b]and c f R.Integrals 445 15.4 If cf is integrable and c Equation (15.14) for T X << 61. then cf is integrable on [a.3 Basic theorems The next few theorems are similar to theorems on limits of sequences or functions. If c = 0 the theorem is obvious.12) holds.
Let n H sn be an enumeration of S.f ) . Define C S(sn) = 2”+2(1 f ( S J + 1)’ and S(12.ti1). + .ti1) the sum of terms for which Xi 4 S.TX) = C ’ f ( X Z ) ( t i . (15.14) and (15.5. In a Riemann sum denote by f ( X i ) ( t i&I) the sum of terms for which X i E S . Obviously we can assume that S is enumerable.4 and15.ti1) = 0 and therefore R(f. TX).” The proof combines ideas from Examples 15. T X ) = R( we have + Remark 15. Since g = f (g . I t is an easy exercise to extend Formula (15.4 If S is countable and f(x) = 0 for x E [a. this theorem is an immediate consequence 0 of Theorems 15. (otherwise we enlarge it by joining it with an arbitrary enumerable part of [a. s.b]\ S then f is integrable and f = 0.15) hold and since R(f g .clearly C ” f ( X i ) ( t i .446 Introduction to Mathematics with Maple for T X << 62. Then g is integrable and Proof.)= 1 for x 4 S.b ] ) .b] and let g be a function which differs from f only at points of a countable set S C [a. If TX << S then (si can tag two adjacent intervals) This proves f is integrable and J : f = 0. 0 Theorem 15.5 of n functions.3 and 15.5 Let f be an integrable function on [a.4.b].13) to a sum Theorem 15. TX) + R ( g . If T X << 6 then Inequalities f . Define 6(x) = Min(&(x). Proof. and C ” f ( X i ) ( t i. 62(2)).
Theorem 15. For example.b] and f 5 g then [f Proof.Integmls 447 Remark 15. s . Let 6(x) = Min(&(x).1 Iff and g are integrable and f (x)5 g(x) for all x E [a. . it follows from (15. Since R ( f . and [g&<R(g. f 1 z d x= 1x1 l2 1 f(x)dx. it also allows us to existence nor the value of assign a meaning to the integral of a function which is not defined for all x in the basic interval.6 I f f and g are integrable on [a. “change in the definition of the function” defines a new function. If T X is 6fine then (15. This is used very often. 0 Corollary 15. I[ 9 .17 hold simultaneously. T X ) .62(x)). where f(x) = dx = 1. J_: i .16) and (15.17) whenever T X is &fine.b] except on a countable set then *Strictly speaking. Therefore. For every positive E there exist gauges 61 and 62 such that whenever T X is &fine.16) and (15. T X ) 5 R ( g .6.for x # 0 and 1x1 X f (0) = 0 (or something else).6 The last theorem can also be expressed as follows: A change in the definition of a function4 f a t countably many points affects neither the b f.TX)< (15. but such points are usually ignored.17) that Letting E + 0 completes the proof.
An example of such a function is given later in Example 15. then for a &fine tagged division of [a.6. a clear contradiction.x for x < b.b].b] there exists an integer j such that X j = c. IM(ba).CI for x # c and a < c < b.2 I f f is integrable and f 5 M on [a. then for any &fine Ptagged division of [a.t o < S(X1) 5 X1 . Similarly. . if d(x) < b .a for x > a then the first tag X 1 of any &fine tagged division must be a.b]for some M E R. if S(x) 5 I x . assuming that a = t o < X1 I t l . is equal to b.3 then I f both the functions f and If I are integrable Proof. Corollary 15. f . we have XI . We shall refer t o choosing S in such a way that c becomes a tag of any &fine tagged division as anchoring the tagged division on c. m(bfor an integrable f satiseing f 4 I J.a .a I tl . then [f Similarly.17 Remark 15.448 Introduction to Mathematics with Maple Corollary 15. Indeed.b] the last tag X . Further.7 If 6(x) I x . b 2 m on [a.6. the function If I need not be. We have If1 I f I If1 and therefore as required 0 Warning: Even if f is integrable.
b] and Proof. Clearly. 6(c) = Min(X(c)J(c)). b] then (15. Theorem 15. ~and ] TX is a Jfine tagged division of [c. (x .18) and <Let & 2' (15.b] then f is integrable over [a. such that every 6fine tagged division T X is anchored on Sby that we mean that all points of S become tags.Integrals 449 It is now easy to see t h a t if S is a finite set in [a.c] and [c. For every positive E there exist positive functions 4 and such that if T X is a &fine tagged division of [ a .x)).c)).19) 6 = Min(&(z). .b] then there exists a positive function 6 . for c < z 5 b. 6 = Min(z(z).7 If f is integrable over [a. for a 5 z < c.(c .
A function f is called a step function on [a.b]. This was possible because we anchored the tagged division on c. .1 integrable on [l. 1.7 and integrability of constant functions.4.2 Iff is integrable on [al. b].< i?n = b } such that f is constant on ] t k .3. If f is integrable on [a. ~and ] another for [c. Every step function is a linear combination of characteristic functions of intervals and onepoint sets. The next theorem is a direct consequence of Theorems 15.an].n . 31. This trick is very useful and we will use it without further explanation in future. 15.8 Every step function is integrable.2.a3].b] if there exists a division T E { a = to < tl < . Theorem 15.3. t k + l [ for k = 0. .18) and (15.. Prove this and use it to show that f (z) = LzJ is integrable over any interval [a. .1 Show that the functions 1x1 and g(z) = (z . one for the interval [ a . . an] then it is integrable on [ a l .4 BolzanoCauchy principle We encountered the BolzanoCauchy principle for limits of sequences and functions. For the integral it reads . A simple proof uses the BolzanoCauchy principle which we consider in the next section. .7 we split the Riemann sum into two. 15. Exercises Exercise 15. .3.450 Introduction to Mathematics with Maple and Hence by (15. .b]. ~ 1 are )~ Exercise 15. [anl.1. 15. 15.19) In the proof of Theorem 15. b] then it is integrable on any subinterval of [a.[az.a2].
p] C [a.Integrals 45 1 Theorem 15. . Proof. T n X n ) is Cauchy. b] and [a. otherwise we just replace d.9 A function f : [a.and T X << 6 ~ Then . b] then (i) T X and SY are anchored on c. Let 6.. so it has a limit.6n(z)).Z V X N ) .16 and we therefore omit it.+l(z) by Min(dl(z). .R ( f T . say I .R(f. A proof for a = a and p < b is similar and the general case follows by a combination of these special cases. PI. 2 &+I. The proof of the necessity of the condition is very similar to the proof of Theorem 10. For n > N we have IW. For each n let US choose TnXn << 6. SY>I< E . Proof. . Theorem 15.10 I f f is integrable on [a. T X ) .20) This implies that the sequence with terms equal to R ( f . We can assume that 6 . Letting n + 00 gives For a given positive E let N > . b] H C is integrable if and only if for every positive E there exists a gauge 6 such that for any two tagged divisions T X << 6 and SY << 6 IWf. & 2 With the BolzanoCauchy principle it is easy to prove integrability on subintervals. For every positive E there is a gauge 6 with the following property: if T X and SY are 6fine tagged divisions of [a. b] then f is integrable on [a. nXJ < N 1 (15. be the gauge associated with E = 1/n by the condition of the theorem. . We prove the theorem for a < a = c and P = b.
R ( S Y ) (< E Let (15. Theorem 15.For both tagged divisions. b].23) Proof. (15.. 0 The BolzanoCauchy principle is the necessary and sufficient condition for integrability. b] except possibly a countable set and LbHl then f is integrable on [a. R ( T X ) << 6 and R ( S Y )<< 6.. b]. b].R ( V 2 )= R ( T X ) . which we may denote R ( T X ) . R(TX)+ R(UW) is a Riemann sum for a tagged division of [a.452 Introduction to Mathematics with Maple (ii) I R ( T X ) . . .21) a = t o 5 X1 I tl 5 .R ( S Y ) It follows from Inequality (15.R(VZ)I < &. Theorem 15. Clearly + R(UW). . Similarly R(=) R ( V 2 )= R ( S Y ) .. Since change of a function on a countable set does not affect either the existence or the value of the integral. I X n 5 tn = b to I X1 5 tl 5 .11 below gives only a sufficient condition but it is easier to apply.21) and by the definition of 6 that IR(UW). tm and let UW and V Z be 6fine tagged divisions of [c.11 If for every positive E there exist integrable functions h. we can and shall assume . b h l E . H such that for all x E [a. I X m 5 tm = Denote by T X the tagged division C I Xm+l 5 .
1 5 t < t k . let f be increasing. R ( h . H . For E gauge 6 such that all inequalities > 0 there exists a R(H. bl. Let T be a division of [a.SY)<L H+E.O X ) . D X ) > L hE.12 [a. O X ) < S.” H J. O X ) .1 ) for t k . R(H. The last theorem is often applied to step functions h. Then .22) holds everywhere on [a. R ( h .R ( f S . b] into n equal parts with dividing points t k . R(f.R ( f ( S Y )I R ( H . S Y ) > L hE. The Riemann sums for f are smaller than Riemann sums for H and they are bigger than Riemann sums for h.” h +  2e < 3E. hold whenever D X << 6 and SY << 6. H + b b b b E. Without loss of generality. I f f is monotonic on [a.R ( h .Integrals 453 that Inequality (15. Y ) > 3e.O X ) . b] then it is integrable on Proof.S Y ) 5 Similarly J. Theorem 15. Hence R(f. b]. Define H ( t ) = f ( t k ) and h(t) = f ( t k .
13 (by Example 15.b[. defined by f(z)= 1/x2 for x ~ ] 0 . 1 and ] f(0) = 0 is decreasing on 10.1.8 then it is integrable because f has finite one sided limits a t the end points and therefore differs a t a t most two points from its monotonic extension to [a.n .The function f . b] such that f restricted to ]ti.1.b] if there exists a division T of [a. If.9 A function f is called piecewise monotonic on [a. Remark 15.11. I] but it is not integrable on [0. . Integrability of continuous functions is proved by the same method as Theorem 15. f is monotonic and bounded on ] a . however.a ) ( f ( b ).22) and (15.7). .b] then it is integrable on 1 > 0 let 6 be a gauge with the property that For a 6fine tagged division TX let Both Inequalities (15.13 Proof.23) are satisfied and f is integrable by Theorem 15.b]. If f is piecewise monotonic on [a. 1 Theorem 15. for n > ( b .23) is satisfied. 0 .b] and bounded then it is integrable.12.f ( u ) ) / & Inequality (15.454 Introduction to Mathematics with Maple Clearly. For E Iff is continuous on [a. b].&+I[ is monotonic for i = 0. 0 It is important in the above theorem that f is monotonic on a finite and closed interval [a. .. Remark 15.
oo["=]O. The function F is said to be an antiderivative o f f on I if F is continuous on I and F'(x) = f ( x ) for every point x E I".4. OO[.LxJ is integrable on @ Exercise 15. I]" =]O. 15. b]. To prove this consider H = F . for example [O. Show that the function x H x . 0 5 y 5 f(x)}. Give an example of a nonintegrable f which fails to be continuous only at one point. We denote by I" the set of all interior points of an interval I . If F and G are antiderivatives of f on I then there is a constant c such that F ( x ) = G(x) + c. [0. Let us denote by A(w) the area of For h positive. I[. b] \S is integrable on [a.A(w) is the area of the set v h.Integrals 455 Exercises Exercise 15. Therefore if 0 < h < S then .~) 21 .G. b].4. Let us now consider a function f which is continuous and nonnegative on [a.1 any interval [a. then H' = 0 on I" and consequently H is constant on I" and therefore on I . Prove that a function bounded and continuous on [a.b] and let F be an antiderivative of f on [a. Clearly + + ((2.2 Let S be a countable set. The word "primitive" is used interchangeably for the word antiderivative.5 x 5 Since f is continuous at v.x1 < 6.b]. A(w h) .5 Antiderivates and areas In this section we shall assume that we know intuitively what the area of a planar set is. for every E > 0 there exists S > 0 such that for I v .
A: = c(w . By the previous discussion the areas A1 and A2 of the sets { (2. Example 15. and to every interval [u.u) if f(x) = c for u 5 x 5 w. . y). then A:(f) For any reasonable theory of area of planar sets. = F(w) . 0 5 y 5 x} and { (z. x2 5 y 5 x}. .F ( a ) .25). 0 5 x 5 1 . Consequently A(w) = F(w) c. We have discovered a way to compute the area of the set (15. In particular. similarly. O I y 5 x2} are respectively 5x5 Al(v) =  W2 2’ = 57 1 w3 A~(w =)Obviously the area of S is 3’ 1 A2(1) = 3’ 1 6 A more detailed analysis of our discussion would show that we have actually proved the following theorem.F ( u ) .F ( a ) . one can show that A/_(w) = f ( w ) for a < w 5 b.456 Introduction to Mathematics with Maple This proves A/.”(f) = A r ( f ) for all a 5 u 5 w I w 5 b. A: + A. the area A: (f)of the set {(x.y). A:(f) = F(b) . A : 5 A:(g) if f(x) 5 g ( x ) for u I x 5 w.y).and which has an antiderivative F .25) by (15. 0 1. By setting x = a it follows that c = F(a) and hence A(x) = F ( x ) . 0 L x 5 1. for x = b we have + A(b) = F ( b ) .F ( a ) . 0 5 y 5 f(x)} must satisfy the requirements (i)(iv).(w) = f(w) for a 5 w < b. y).b] a number AE(f) in such a way that (i) (ii) (iii) (iv) A: 2 0. the difference of the values of an antiderivative to f at b and a. u 5 x 5 v. Theorem 15. w] c [a.8 Consider the area of the set s= {(x. It follows that A’(w) = f(w) and A is an antiderivative of f .14 If it is at all possible to assign to every function f which is continuous and nonnegative on [u.b]. In particular.
our intuition led us to believe that this area is All this indicates that s. a 5 x 5 b.27) 5Provisional until readers become familiar with measure theory . 15. The difference F(b).26) where F is an antiderivative o f f .26) is valid for every function f possessing an antiderivative F is called the Fundamental Theorem of Calculus. for nE N. area and volume is measure theory.Integrals 457 Our theorem asserts that AE(f) is uniquely determined and AE(f) = F ( v )F ( u ) . We shall not attempt to expound it and refer our readers to Lee and Vfbornf (2000)."f. On the other hand.y). or by [ F ( x ) ] Hence (15. We shall prove it in the next section. Example 15. Example 15. In other words: No matter how the area of a planar set is defined.9 By the NewtonLeibniz formula 16 xn dx = (bn+l 1 n+l . This result can also be established by using Maple.1.F ( a ) is often denoted by F(x)l: or FI:.1 we found that was always the difference between the values of the antiderivative F of f at b and a. Theorem 15. where F is any primitive of f . In Theorem 15.6 Introduction to the fundamental theorem of calculus In Examples 15.10 By the Fundamental Theorem of Calculus we have (15. as long as requirements (i)(iv) are satisfied we always have AE(f)= F ( v ) . (15.14 we found that the area of the set {(x.1 and 15. 0 5 y 5 f ( x ) } was the same difference.14 is a kind of justification5 for evaluating areas by antider ivat ives. This is by no means so by chance.26) is often referred to as the NewtonLeibniz formula. The theory which deals with the concepts of length.26) can be rewritten in the form laf = F(x)l:.an+l).F ( u ) . The theorem which states that Formula (15. b :.1 and in Exercise 15. Equation (15."f s.
ti1). However this need not hold for i = 1 or i = n.F ( a ) .15 (The Fundamental Theorem of Calculus.ti1) + f(Xn)(b.7 that the function f. Then f is (Kurzweil) integrable and s.F(ti1) is very close to f ( X i ) ( t i .458 Introduction to Mathematics with Maple if either a < b < 0 or 0 < a < b. We write R ( f . For that we employ the relation F’(Xi) = f(Xi). ( X i can be either a or b and F need not have a derivative at a or b ) . X ) is close to F(b) .F ( a ) + c i=2 [F(ti). We are going to do this by showing that F ( t i ) .F ( a ) = c n i=l ( F ( t i ). that is. For any tagged division T we have F(b) . .tn1) F(b) .D f = F(b) . During the proof we shall consider only tagged divisions anchored at a and b.F ( a ) = F(t1) .F ( a ) .1 ) . 15.a ) and + n1 c i=2 n1 f ( X i ) ( t i. We can always ensure this by demanding that the tagged division is 6fine for some gauge. We know from Example 15. (15. Therefore we split the sum (15. (15.F(t21)).) Let F be an a antiderivative of f on [a.T X ) into three summands and estimate the differences between them separately.b].q n .F(ti1)I + F(b) .7 The fundamental theorem of calculus Theorem 15.29) and R ( f . we shall have X1 = a and X n = b.29) We wish to show that R ( f T . T X )= f(Xl)(tl .28) We commence with a plan of the proof. defined by f(x) = 1/x2 for x # 0 and f(0) = 0 is not integrable on any interval containing zero.
33) I w < X + q(X).Integrals 459 We denote A5 = C ( F ( t 2 ).ti1)) r i=2 l * and b . tla (15.a ) = 0 and limf(b)(b . Since F’(x) = f(x) for x E ( a .F(t)I < E .b) we can find a positive q = q(x) such that and whenever X .q(X) < u 5 X ties (15.32) Since limf(a)(t .t ) = 0 there exists a positive ttb w such that whenever a 5 t < a + w .F ( t 2 1 ) f ( X i ) ( t i . Combining Inequali .y < t 5 b implies that IF@) .34) and (15. and whenever b .w < t 5 b.
. w = ti.40) Combining (15.a < E. It can be shown that (15.a) 0 Remark 15. (15.n .36).a ) .31) with t = tl and we have ( F ( t 1 ).30) with (15. at b).w) 6(x) = Min ( ~ ( x (x ) .F ( a ) . (15.39) and (15. .( F ( b ).36) ( F ( b ).40) gives IR(f.7 that (x.” f = F(b) .6.f(XZ)(tZ . Similarly (15.F(ti1) .b[ except possibly a finite set then f is integrable and s.38).460 Introduction to Mathematics with Maple Now define 6 by &(a)= 6(b) = Min(y. (Recall from Remark 15.1. Since tl .2)).10 If F is continuous on [ q b ] and F’(x) = f(x) for all x E ] a . (See Exercise 15.28) holds if there is a countable set S such that F’(x) = f(x) for every x E [a. This follows immediately by dividing [a.) .a < 6(X1) = &(a)I y we can use (15.37) < & ( X I )5 w we can use (15.7.F ( u ) (= A1 < E .b] into a finite number of intervals.37). we have A5 < ~ ( ba).t i 4 ) l I &(ti . for x E ( a .F(a))l < E ( 4 which completes the proof since E was arbitrary. . ( b .b) For the rest of the proof.b]. where the Fundamental Theorem is applicable.ti1) Summing these inequalities for i = 2 .x) appear in the definition of 6 to anchor X1 at a and X.b]S and F is continuous on [a. and u = ti1 and we obtain JF(t2) . +b . (15. (15. . (15.T X ) .35) with X = X i . let T X by &fine.F(tnl)l = A2 Since tl .7) with t = tl and we have Similarly We can use (15.a ) and ( b . .
Table 15. However if you want to have a primitive. The symbol f(x) d x denotes a primitive of f but not necessarily the same function at each occurrence. Gradshtein and Ryzhik (1996)). for instance.7 easily. Maple will find the integral instantly. These are easily verified 1 GTi Example 15. Sometimes the primitive can be found by using Table 15. way that the primitive to id is x H x2 or a function differing from it by a constant function.7 summarises such formulae. for example. It is denoted by However. there is no point in finding a primitive with Maple for evaluating an integral by the Fundamental Theorem. but this requires some caution.11 it does not the primitive cannot be expressed by a simple formula anyhow. Here are some examples: s are partly new.oo. 2 (X At many occasions the primitive can be easily obtained by reading a formula for differentiation backwards. . Then we do not have to write C in the equations. A primibiwe to a given function is not uniquely determined. Of course.XI. so easily that it is not worth opening the computer. Maple will find the primitive with the command int (f (XI . Maple will find it for you on most occasions. emphasizing in this J J interval ] . For instance. J x d x = x2/2 + C. Now.6 Maple also omits the constant C as we do. which we established earlier by differentiation. two primitives to the same function differ at most by a constant. we have both J f(x) d x. We prefer to regard equations involving primitives as equivalences: two functions are equivalent if they differ by a constant. however. The formulae for by differentiation.Integrals 46 1 Direct integration For evaluation of an integral of f we need a primitive of f .O[)and (X (x+ 1)2 + 1)dx = 2 ' X2 + 1)dx = + X. Extensive tables of primitives have been compiled (see. Some authors write.
00[ for . a E R . x # 0 for n < O X E P .462 Introduction to Mathematics with Maple Table 15. x unrestricted for n 2 0. O[ or x €10.00[ [exdx = ex / a x d x = ax no restriction acP. . k E Z x dx = arcsinx dx = arctanx €1 . k E Z I I J sinxdx =  cos x dx = sin x dx = tanx cos2 x + s ' sin2 x 1 dx = cot x any interval not containing kr. I[ or 31.00.a#1 cosx no restriction no restriction any interval not containing (2k l ) r / 2 . n # 1. comments / x n d x = n+l xn+l n E Z.1 A short table of primitives Range of validity.1.1[ no restriction no restriction for 1 + sign. a # 1 either x E] 00.sign ] .
d x cos x sin 2x 2 sin x cos x 2 cos x 2 sin x 1 1 = In I cosxl . before Theorem 15. F ( A X + B)A = AX + B ) . If for instance f(x) lf(x)I = f(x) and by the chain rule 1 [h(f(x))]’ = (f’(x)) = < 0 then f (4 f ‘(4 f (4 Example 15.sin x + + J X 2 2 . but it is convenient to consider it here. by the chain rule (iF(A2 Example 15.13 sin2xdx = 1 + + B))‘ = .tan xcot cos2 x sin2 x cos2 x x. J1czos2x x dx= 1 4 sin2x.2 ) 2 1 = . If F is the primitive of f on an interval [a.Integrals 463 Example 15.7 Indeed.4x +5 dx=L/ 2x4 dx 2 x2445+5 d x + 7 / 1 (x .ln(x2 .17 .14 cosx / L d x = / sin2x cos2 x d x = J e d x + J . b] then x I+ F(Ax B ) is A the primitive of x H f(Ax + B) on the interval with endpoints Aa + B and Ab + B. 2 2 J tanxdx= J d x = lnlcosxl. 2 + + + We now illustrate the use of the F’undamental Theorem by several examples 7This is a simple special case of change of variables.2).12 sin2 x + cos2x dx dx J sin2x dx = J dx= Jz J + z z . There is another simple formula very useful in direct integration. namely valid on any interval on which f(x) # 0.42 5) 7 arctan(x . .In I sinxl.
for the validity of the theorem it is essential that F is continuous on the closed interval [a.b[. Example 15.464 Introduction to Mathematics with Maple Example 15. By the Fundamental Theorem of Calculus F’ is integrable on [0. & For E + 0 the right hand side diverges to infinity . and fi = 0 otherwise. IF’[ 2 F’I = IF(bi)F(ai)l = ai a i 1 +: 2 + 1 1 2 2 . IS the function F defined by F ( x ) = n=O n t l ‘ a primitive of f by the . If the integral l i ( l / x 2 ) d x existed then (with 0 < E < 1) Jd’$dx 2 l1 fdx= 1 . Then l l ]l/Jm.a contradiction.7. For an indirect proof assume it is and define fi to be IF’( on 1/&[ (we denote this interval ]ai. Jd’ 1 Ibi 1 bi bi fi = fi = a.b]. However IF’I is not integrable. Note that fi is not differentiable a t zero. 2 2 + 1 For every n E N and every x E [0.17 Let F be defined by F ( x ) = x2cos(n/x2) for x # 0 and F ( 0 ) = 0.1.18 Let f ( x ) = n=O CnXn 00 with radius of convergence r . Example 15. 00 00 we have a Example 15.15 We have Just apply the NewtonLeibniz formula.16 Using the Fundamental Theorem gives an instant solution to Example 15. In this example we can see how convenient it was t o assume in the theorem the existence of the derivative only in the open interval ] a . However.1] and F’ = 1.1] we have n i=l Consequent Iy Since the right hand side of this inequality diverges as n + contradiction. Then CnXn+l .bi[).
.x 2 we obtain Example 15. For instance > int(x^lo*exp(x^2) . Jdx et2 dt plays an important role in probability and 1 . @(x) = The function areas of applied mathematics such as heat conduction. However. Equation (15. e x p ( .41) gives a meaningful expression for <p and a method for calculating it. a t least for small x.2).x=O. It cannot be expressed in terms of more familiar functions' but other integrals or quantities of interest can be expressed in terms of erf. Theorem 13. . Extensive tables for the function @ have been produced. exponential. It is interesting to see how Maple handles this integral.Integrals 465 theorem on differentiation of power series (that is.19 for every x > 0. The answer.fierf(x) 2 If you look in Help for the meaning of erf(x) Muple will tell you that erf(x) = $J. Using the previous example with f(x) = e . the use of erf makes sense: erf is a higher transcendental function which is encountered often. a t the first glance. looks silly: it says that the integral in question equals to itself. Hence by the Fundamental Theorem of Calculus 00 provided that r < a < /3 < r . ~ 16201 16 e(4) 945 +fierf(2) 64 Like rational.25). logarithmic or trigonometric functions.t 2 ) d t .
18 to f(x) =  In( 1 . tn However. if you need a numerical value of the last expression you can obtain it with the evalf command. and then we have We cannot set t = 1 directly in this equation because the radius of convergence of the power series is 1and Examplel5.F ( 0 ) = " 1 n=l .18 is not applicable with [a. then F will be an antiderivative of f on [0.1].3 that . n 2 To prove that F is continuous from the left a t 1we write tn1 n=l 00 tn .< . we can find N such that n2 n=N+l .1] and by applying the Fundamental Theorem of Calculus we obtain  l1 dz = F(1).ln(1 . lim tt1 tn 1 C= 0.and show that F is continuous a t 1 from  n=O n2 the left.x) 2 .converges.20 Noting from Example 14. n=l N n 2 .1 Since " 1 n=l 00 . if we set F ( t ) = . p] = [0. Example 15.Then n2 1 & 4' With N now fixed.x) X n=l we can apply Example 15.466 Introduction t o Mathematics with Maple Of course.
For Abel's theorem we refer to Lee and V$born$ (2000.Integrals 467 Consequently there exists a 6 such that for 1. The Maple solution is . 1 6 An elementary proof of the equation O0 .n=l. The continuity of F from the right follows immediately from the Abel theorem.can be found in Matsuoka n2 1 7T2 1 6 (1961).7r2 '.7r2 This answer only looks different from ours: the sum of the series is 7r2/6.= . .2). Theorem A. .6 < t 5 1we have Combining these results we have which completes the proof.as can be confirmed by Muple. This theorem asserts that a power series with radius of convergence r is continuous from the left a t r if the power series converges for x = r . 1 6 > sum(l/nn2. i n f i n i t y ) .5..
S and let q ( x ) = 1 otherwise.468 Introduction to Mathematics with Maple Exercises Exercise 15. J m d x . 1. f 2 JE1 2 ..] J’. J Cot2x d x .35) for x E [a. 3. [Hint: J’.fLyT Exercise 15.lne. Let n I+ xn be an enumeration of S . J t a n 2 x d x .] Prove that if @ Exercise 15. 2x+3 6.3 Show that for any x > 0 Exercise 15. + bx3)4d x . 7.7.5 x+2 dx. Find guess the primitive. J ( a Find 2.b] then [ f ( x ) d x = F(b) . Also give a simpler proof of Example 15. J x2 + 72 .20. and Exercise 15. x J d9 +d42  22’ dx d x 2 4x + + 29’ Let f(z) = 2xsin & .1 1. [Hint.4 Show that if r > 0 and the series on the right hand side converges absolutely.7. [Hint: It is not difficult to f(0) = 0.1].7.7. 4. Define 4 . Jaxb2x d x .b] .1 @ Exercise 15.nxln cos & for z # 0. J dx .2 f ( x )dx and f ( x )dx.sinx’ dx x2 16 5.zn1 < 6(xn) we have IF(x) F(zn)l < ~ / and 2 If(xn)l ~ < ~ / and 2 let ~ 6(z) = 1 for x S.5 Using the Fundamental Theorem of Calculus prove that f ( x ) = 1 / x is not integrable on [0.F ( a ) .7.S with countable S. and (ii) F is continuous on [a. [Hint: Use the same method as in Example 15. + 11.7.b] .7. define 6(xn) > 0 to be such that for I X . Find q ( x ) as in (15.6 (i) F’(x) = f(x) for all x E [a.
~ . Equation (15. (15.43) as the following (zX ) ~ C O S Z .7. However F(z)g(z) . 91f one is prepared to use Exercise 15.G ( z ) f (z) = 4zl is not integrable.42) follows.b]. Equation (15. For F and G chosen as above.8 Consequences of the fundamental theorem The rule for differentiation of a product and the chain rule lead to theorems on integration by parts and substitution.b] except on a finite setg then . G = and F ( 0 ) = G(0) = 0 neither Fg nor Gf exists because then both would exist by Equation (15. 0 Provided that one of the integrals in Equation (15.42) holds but Formula (15.43) does not make sense.42) Proof. F' = f and G' = g on [a.Integrals 469 6(x) = Min(q(x). ] 15.11 We illustrate the use of integration by parts in three cases.F(a)G(a) [ fG.43) below exists we obtain useful formula [ Example 15. ( i ) integrating a product in which one factor has a simple derivative and the other is not too difficult t o integrate. b] and by the Fundamental Theorem. We have (FG)' = Fg+Gf except on a finite set. (15.16 (Integration by parts) If F and G are continuous on [a.6 this set can be countable.44) Ja Ja The assumptions of the theorem alone do not guarantee the existence of either integral in Equation (15. . Moreover this example also shows that the product of an integrable function f and a continuous function G need not be integrable.F(u)G(u).43) and so would their difference. the function F G is continuous on [a. $ so Remark 15.h(x)) and proceed as in the proof of the Fundamental Theorem of Calculus. For F ( z ) = ~ ~ s i n z .43) The above formula is often rewritten in the easily remembered form (15. Theorem 15.21 Fg = F(b)G(b).~for z # 0 example shows.
.( a 1)2  + As a teaching tool. 1 2 .1 lx"lnxdx= a + l l n t . Besides the integral and the limits of integration. > student[intparts] (Int(x^(alpha)*ln(x) . v = Inz and have ta+l ta+l ta+l . a E R. (iii) by integrating by parts we obtain an equation from which the integral can be found (sometimes to obtain such an equation one must integrate by parts twice)..44).22 We illustrate (i) with the integral J : za l n z d z .c U + l lnt .l s d x = . Example 15. We deal with the integral from Example 15.44) we set u' = xa. Example 15.23 An example illustrating (ii) is arctan z d z = . Q! # 1. in other words the function v in Formula (15.470 Introduction to Mathematics with Maple (ii) by inventing a factor which is easy to integrate (this is actually a subcase of (i)). t ) . Maple has a command intparts in the student package.24 to illustrate (iii) n 1 . If we want just the integral Maple would compute it instantly with the i n t o command.x=l.22 for a = 1 consequently J = . Maple also requires the factor to be differentiated.1n2. > simplify(%) .In2 t. In applying Equation (15.Example 15. l n ( x ) ) . The command intparts is useful in seeing how integration by parts works or as a practical introduction to integration by parts but a better insight into the method is obtained by working with pen and paper.
v = sin'' x and have z cos2x d x SI. Similarly 1 = 1. l7r 2 n . It follows from these formulae that lim f i S n = n+m m.1S2n2 =2n 2n 2n2 22 since SO= 7r/2.sink' x cos xl. = .25 (Wallis' formula) By integration by parts we diminish k in the integral SI.Sin2n+lz 5 sin2nz. Application of the chain rule leads to . established by the British mathematician Wallis.3 . Such formulae are called reduction formulae.1) 1 TI2 This formula leads to an interesting limit for 7r.. In Equation (15.44) we set u' = sin x.45) and (15.1 2n .l2 + (k .. and by the Squeeze principle 2n 2 Since + 1 we also have 2n+1 + Equations (15. and consequently This implies S2n+2 5 1 and it follows that 5 S Z n + l 5 SZn.7r/2] we have 0 5 sinz since S sin2n+2x < .46) are the Wallis formulae. = Jo5 sin'xdx.Integrals 471 Example 15. For an even k = 2 n E N we have S2n = 2 n . For z E [0.
b[.48) < a we define [f = /af b (15. B). integration by parts (15. On the other hand the formula lf must be modified to in order to be valid for any pair a . B] with F’(z) = f(x) for x E ( A . Remark 15.472 Introduction to Mathematics with Maple Theorem 15. Therefore we extend the definition of the integral first. B].) If (i) the function 4 : [a. continuous on [A.17 (Change of variables. this is not guaranteed by the assumption of the theorem. b] H [A. if b < a. For instance. b]. = f(4(t))@(t for ) every t E ( a .49) Most of what we learned about the integral extends easily to the case of b < a.3 If a = b we define LUf= o and if b (15. then ~~ The theorem is often referred to as integration by substitution. b) and Proof. (iv) there is a function F . Noting that [F(4(t))]’ applying the Fundamental Theorem to the integral on the righthand side .12 The theorem as stated suggests that 4 ( a ) < 4 ( b ) .43) is valid without any change. However. 6efinition 15. (iii) 4 has a derivative on ] a . (ii) 4 is continuous on [a. b.
when making the substitution x = 4(t) one has to substitute not only 4(t) for x but also dx = +'(t)dt.47) are equal. Example 15.47) are zero.47) is F($(b))F(~(u) and ) both sides of Equation (15.= #'(t). Very important for the use of the theorem is to change not only the variables but also the limits of integration. often an expert creates that factor by some manipulation of the integrand and achieves considerable simplification. In either case.To apply the theorem we set x = $(t)= r sint with t E [7r/2. and we have + l' d n d x represents the area of a l ' d m d x = r2 cos2t d t 7rr2 .47) from dt right to left an experienced mathematician recognizes the factor #(t)dt for substitution as dx. If we wish to change variables x = 4(t)in the integral f we need to find a and b such that +(u) = a and 4(p) = b. This equation is easy to remember: it is obtained from dx . 0 : 1 Equation (15.47) can be used in two different ways: to evaluate the integral on the left hand side by finding the value of the integral on the right hand side.26 The integral quarter circle { (x. a = $ .47) we have If $(t)= $(b) then both sides of Equation (15. This is illustrated in Example 15. 7r/2] and dx = r cos t d t. This is easy if 4 has an inverse. since then. If 4(a) < +(b) then by the Fundamental Theorem the lefthand side of (15. When using Equation (15. The strength of this theorem lies in the fact that no assumption is made concerning the monotonicity of 4 or existence of 41.27.~ ( aand ) b = 4(p).Integrals 473 of (15. If 4 ( a ) > 4(b) then the last equation holding by the Fundamental Theorem. y). x2 y2 5 r 2 } . or by going in the opposite direction.as if by multiplication by dt. When x = 0 or x = r then t = 0 or t = n/2.
x=O. .t>.2&. Part of the skill of using change of variables is anticipation of the form of the new integral and it is our belief that such an anticipation is acquired by working examples on change of variables with pen and paper. So we give two more examples working without Muple.. (Int(f(z).b). Maple has a command changevar for changing variables.474 Introduction to Mathematics with Maple We have recaptured the geometric significance of r . 20 h Note that $ is not monotonic on [2.3]. tdt + L ~t " z l 6 =z 1J 2 5 d x = &jZO 25 = 5 . We give two examples of changing variables in > > student [changevar] (x+sqrt (1+xn2)=t. We originally defined 7r/2 as the smallest positive root of the equation cosz = 0. When making the substitution t2 16 = $(t)= x we recognize in t d t half of d x and we have In order to calculate s_. lfi+2f dt > > student [changevar] (x=sin(t> .47) from left to right.t). Int(l/sqrt(l+xn2). but it is clear that the integrand is 1 ..sint and so the integral can be easily evaluated. x = a. new variable) . for the second time. The structure of the command is as follows: changevar(equati0n defining the substitution.1) . It resides in the student package. Example 15. sin(t) + 1 cos(t) d t Maple will not easily simplify the integrand in the last example.x=0.27 we use Equa@Ti6 tion (15. Int(sqrt ((ix>/(i+x>> .2). .
We shall prove that later in . We must look for an explanation. d x = dt and have dx dt 1+8cos2t =1 dx 1+8cos2x' Consequently ?r dx l+8cos2x dx '1+8cos2x First we show thatlo bm/2 lim Indeed 1 0 dx 1+8cos2x =1 tan b dx 1+8cos2z' 4 Now we evaluate (with b dx 5~ / 2 b. "If f f is integrable on [a. violating assumption (ii) in Theorem 15. so the theorem cannot be applied. b] then lim tTb Section15.10 Jdt Jdb f = f . The function tan is discontinuous a t n/2. l+8cos2x < 7r/2) 1 dx tan b arctan b (3) tan . The correct evaluation is as follows: Jd ?r dx l+8cos2x =I =L12 dx ?r dx 1 + 8 c 0 s ~ x +1 ~+/8~ cos2x' For the second integral we apply the substitution x = 7r .47.t.28 integral I = ?r Using the substitution y = t a n x and d y = in the cos2 x dx dx leads to 1 8cos2x + The result is obviously wrong: an integral of a continuous positive function must be positive.Integrals 475 Example 15.
x=O. Although mechanical use of change of variables could easily lead to a wrong result. Exercises Use integration by parts to evaluate the following inExercise 15. "Iff is continuous on [a.47) holds in the following sense: If one of the integrals appearing in (15..21. Theorem 15.x=2. Theorem 2.476 Introduction to Mathematics with Maple The limit of the right hand side for b t 7r/2 is 7r/6. 1 4 int (x/sqrt(xA2+l6) .r) assuming r > O . . > int (sqrt(r2x2). > restart. However there is an additional assumption that 4 is monotonic. Maple was sophisticated enough to avoid any mistake.8.. r27r > 5 .1 tegrals. Finally we have the result dx 1+8cos2x =2= 7 r 7 r 6 3' It is interesting to compare our work in the last three examples with the ease with which Maple evaluates these integrals. 7r 1 3 The last example on Maple use is particularly pleasing.47) is valid.17 and if q5 is strictly monotonic then Formula (15.8).18 (Monotonic change of variables) If we awume (i)(iii) of Theorem 15.11 The next theorem does not have this weakness.1.b] then (iv) holds because of Corollary 15. with and without Maple.7.Pi). Assumption (iv) in Theorem 15. On some occasions we might like to change the variables because we do not know F but then it might be difficult to justify the use of (15.17 is a little inconvenient. . For the proof we refer to Lee and Vfbornf (2000.2 6 > int(l/(l+8*(cos(x>)'2) .47).x=O.3).47) exists then so does the other and (15.
+ sin x d x.8.a)" n ( a ) T + %+l.9 Remainder in the Taylor formula The Peano form of the remainder in the Taylor formula says that the remainder is smaller than the last term by an order of magnitude. . provided 5 < a < x < 5. f'. Theorem 15.50) (15. Sometimes a more precise estimate is needed..19 (Integral remainder) If f . (4) J 0 rX tsintcostdt. .. b and f ( n + l ) exists on I except possibly a finite subset of I . then the function t H f("+')(t)(bt)"/n! is integrable on I and lI) f ( b ) = f(a) where + f'(a)(b . . It is provided by the next theorem. (5) J.Integrals 477 (1) (2) { J X ! X ! z ex dx. (15. Change the variables in the following integrals + Exercise 15.2 (1) SU(l + 1 x y o 0 dx. %d t . f(") are continuous on an interval I with endpoints a. . t3tdt.a ) + . t = 1 x. 15.+ f ( b . t = cos x assuming n/2 < x < n/2.51) .
12 For n = 1 we have Proof.t)("+')/(n l)!we have (the existence of the first integral is guaranteed by the induction hypothesis) + This means Rn+l = f ("+l)(U) ( b .6 then the excepRemark 15.a)"+' (n I)! + + Rn+2. By induction.478 Introduction to Mathematics with Maple If one is prepared to use Exercise 15. For n = 0 the result is just the Fundamental Theorem. by Equation (15. from a strictly logical view this step is not needed. Equation (15.50) is often given a different form These ways of writing the formula have the advantage that they do not suggest that h is positive or that x > a.13 tional set in the above theorem can be countable. a Turning to the induction hypothesis.t ) f " ( t ) d t . .44) with u = t .b and w ' = f " we have J" (ff ( s ) d s ) a a = J"(b .7. Employing integration by parts (15.44) with u = f(n+l) and v(t) = ( b . 12The case n = 1 is considered as motivation for the rest of the proof.
29 > taylor(tan(x) . u := .x=0. XI > t3:=diff (tan(x) .0099500000 > M:=2+8*a^2+6*an4. This can be .000792079 Finally .x. In concrete application a precise estimate of K can be important. b] then (15. for some constant K . Using the Maple command taylor for obtaining the Taylor approximation gives Example 15.001 or 1000. > a:=0.3). x + 0(x3> This tells us that ltanx . t3 := 2 (I + tan(z>2>2 4 tan(^)^ (1 + tan(x>2> > + simplify(%) . M := 2.Integrals 479 A particularly useful estimate of Rn+l follows from from the remainder formula (15. If Ifn+'(x)l 5 M for x E [a. We use Inequality (15.01n2)/2. 2 + 8 tan(^)^ + 6 tan(^)^ We need some preliminary estimate of tan.51) if f n + l is bounded on the interval [a.52) We wish to approximate tan J: just by x. b].x.52).x).5 KJ2J3.01/l(0.
b].01.8504949 ~~ for 1 x 1 5 0. that f ( n + l ) is bounded. Then .53) For n = 0 this formula reduces to the Lagrange mean value theorem and it is called the Lagrange remainder. In the next example we wish to approximate exp(tanx)) by a polynomial of fifth degree. > taylor(exp(tan(x)) 13. Such The error is smaller than 1012M/(6!) < .t6).b]. > 185. Let rn.00055 12 e*0099 > M: =evalf (%).026 high accuracy is rarely needed but it was easy to achieve with MapZe.x3 + g x4 + x5 + O(x6) 2 2 120 t6 := 272 tan(x) (1 + tan(^)^)^ etan(z)+ 136 (1 + tan(^)^)^ etan(z) + 416 tan(^)^ (1 + tan(x)2)2etan(z)+ 1168tan(z)2(1 + tan(^)^)^ etan(z) + 480 tan(x) (1+ tan(^)^)^ etan(%) + 40 (1+ tan(^)^)^ etan(z) + 32 tan(^)^ (1 + tan(x)2)etan(z)+ 496 tan(^)^ (1+ tan(x)2)2etan(z) + 720 tan(^)^ (1 + tan(^)^)^ etan(’)+ 260 tan(x)2 (1+ tan(^)^)^ etan(z) + 30 tan(x) (1 + tan(^)^)^ etan(z) + (1 + tan(x)2)6etan(z) subs(tan(x)=a.b[ such that If fn+’(x) exists for all (15. 1 1 3 37 +x2 + .x . b] then there exists a c € ] a . Ad be the greatest lower bound or the least upper bound of f ( n + l ) on [a.480 Introduction to Mathematics with Maple for 1x1 5 .14 (The Lagrange remainder) x E [a.6). M := 186.01. This form is easy to remember: the remainder looks like the next term in the expansion but the derivative is evaluated a t a shifted point c.  Remark 15. In proving it we assume. The Lagrange remainder is valid if f ( n + l ) exists on [a. for sake of simplicity.x=0.
103.011.2 Give an example showing that the Lagrange remainder is not valid if f ( n + l ) fails to exist at one point of ]a. a 5 c with the property that < d 5 b and 6 a gauge (15. 0.rr/6.. or simply the indefinite integral. b[.9. If T X is a tagged division then by T X I : we denote the tagged division Lemma 15. + =1x [o.1 given intervals: Find the accuracy of the following formulae on the (I) sinx 1 Mx (2) 1 xl+ x3 (3) cosx = 1 [.f s.54) is called the indefinite integral of f . [Hint: Go back to the mean value theorem.10 The indefinite integral The function F(x) = Lx.] 15. @ Exercise 15.” (15.rr/61. Whatever we say or prove about F applies with little or no change to F ( z ) = f .9.1 Iff is integrable on [a.Integrats 48 1 By the intermediate value property of the derivative there exists a c with Exercises Exercise 15. We shall study its properties in this section.1031. [. 61.55) .
Continuity form the left at a point which is not the left endpoint of [a. d ] . c] and [d. We shall not make a distinction between a gauge and its restriction. Using (15. .57) leads to Sending q + 0 gives Equation (15.l~ The sum R(f 7 sy> + Wf.c) = 0 and there exists w > 0 such that f ( c ) ( t. I Proof.b]. For E > 0 find a gauge S such that (15.U V ) is a Riemann sum for a 6fine tagged division of f on [a. For r) > 0 let 61 5 6 and 62 5 6 be gauges such that tagged divisions SY << 61 and U V << 62 of [a. should be omitted.56). iP3 0x1: << 6 (15. 0 Theorem 15. a 5 c < b.c ) < the lemma can be ~ / for 2 c < t < c = w. b]. b] is proved similarly.58) Obviously lim f(c)(t . D X If. respectively.482 Introduction to Mathematics with Maple then.56) Proof.Lc Let c < t < Min(w. t.55) and (15. + R(f. We prove continuity of F from the right at c. b] then its indefinite integral is continuous on [a.20 I f f is integrable on [a. 141f a = c or d = b terms involving R ( S Y )or R ( U V ) .G(c)) then 13We should have said: if 0x1: is fine with the restriction of 6 t o [c. re~pective1y. b].
Generally speaking.21 I f f is integrable on [a. By continuity off from the left at c for E > 0 there exists q > 0 such that . If f = IQ.1x1 then forO<x< 1 forl<xL2 Hence FL (1) = 1. (iii) F’(c) = f(c) i f f is continuous at c.21. The next question is: if F is differentiable. G’+(b)= f ( b ) . a < c < b. For instance. Proof. then F(x) = F’(x) = 0 # l ~ ( x = ) 1 if x is rational. b] then it has a primitive. Corollary 15. a < c < b. is differentiability.59) < x < b and GL(a)= f ( a ) . b] then (i) FL(c) = f(c) i f f is continuous from the left at c. the indefinite integral need not be differentiable on all of ] a . However the next theorem shows that F is differentiable at points of continuity of f and then F’(x) = f(x). (ii) F$(c) = f(c) i f f is continuous from the right at c. after continuity. Theorem 15. F$ (1) = 0 and F is not differentiable at 1.Integrals 483 Consequently IF@) . is F’(x) = f(x)? The answer is again no. We prove only (i) since (ii) is entirely similar and (iii) follows from (i) and (ii). a 5 c < b. if f : x H x .b[.1 I f f is continuous on [a. more precisely there exists a function G such that G’(4 = f ( 4 for a (15.F(4 The next natural question “0 ask.
E ) ( C . Jl Theorem 15.). for an integrable f .) with c. Similarly Remark 15.1 In t d t = .1 . c] for every c with a li$Lc < c < b and (15.484 Introduction to Mathematics with Maple It follows by integrating this inequality that (f( c ) . that 1imSbf = F(b) .22 that s. If we knew that the limit of the left hand side is In t d t then we have evaluated the integral. we can use it profitably for evaluation of integrals. (15. Take a strictly decreasing sequence {c. b] for [f exists and equals A b (15. cn11 then Proof.F ( z ) 5 (f(c) + . b] and Ja f = A .60) then f is integrable on [a. For every positive E there exists S.15 If f is integrable on [a.61) f exists and equals A b ja f then f is integrable on [a. The right hand side has limit equal to 1 as z + 0. We established in Example 15. 4 a and co = b.62) . The next theorem gives an affirmative answer. Theorem 15. F F(c) . such that if T X ( n )is a &fine tagged division of [ C n .z In z + 2.z) 0 l f If we knew without assuming the integrability of f that this equation holds. b] and = A.22 (Hake's theorem) every c with a < c < b and lim cla Iff is integrable on [c.F ( a ) = tla t .)(c and FL(c) = f(c). We can and shall assume without loss of generality that f(a)= 0.20 says.
66) tl < t o and consequently by (15. (15. By Inequality (15.~ 2 Let now T X be a Sfine partition of [a. (15.67) Relations (15. .63) whenever a < c < r . cn11.Integrals 485 According to the hypothesis there exists r such that (15. for all n E N.)). .63) + & ( X I )= a + &(a) < r Since it suffices to show that (15. Let N be the first integer for which c ~ + 1 5 tl.)(t. l 5 S(Z) I Sn(Z) for Z E [cn.) 5 Min(S. b]. Now define a gauge 6 which has.).( b . In addition we also require that d(a) = r . 1 S(b) = .66) 1 ) . the following properties. The Riemann sum R ( T X ) starts with the term f (X.64) and (15.68) 15The first inequality implies that S(c.t l ) .a. &+I (c.(c.65) imply that X 1 = a and hence f ( X I ) = 0.
f is integrable on [c. .C N the term c k tags the subinterval in which it lies. . . . . For k = 0.N } each [til. . for every c.65)l6 ~ ] c j cj . . It remains to prove the claim.+ l ] for some j 5 k. a < c < b. 1. g integrable on [a. Then by (15. Applying Lemma 15. b]. 0 Theorem 15.69).62. . For k E { 0 .22 yields the next theorem Theorem 15. . The possibility X i < ck can be ruled out similarly.68). The claim has been proved. Then ck belongs to If c k < X i then X i (15. . b] except a finite set then f is integrable on [a. 1 .n}. . b] and there are functions G. . l6If x z = Cj+l .3.C N .486 Introduction to Mathematics with Maple Claim: The Sfine partition T X anchors on C O .64) or by contradicting Inequality (15. . ti] for some i E {2.23 (Integrability test) H.1 with the roles of a and p played by tl and C N we have Thus we have Equation (15. c 1 . Accepting the claim for a moment we have N n=l and consequently The sum on the right hand side is smaller then ~ / by 2 Inequalities 15. b] and such that g(4 I f ( 4 L G(4 for all z E [a. .
The limits b b liml g zla and l xi la ml G exist by Theorem 15. a" < d.30 is called elliptic. Consequently and liiL b f exists by the BolzanoCauchy principle.Integrals 487 Proof. for every positive E there is a d such that whenever a < a'.20. It is defined for p > 0 and q > 0 by the .311028777 In the intermediate result Maple employed the function Beta. We show that integrable on [0. Maple can evaluate the integral l / d G to 1. Factorizing. It cannot be easily evaluated by the Fundamental Theorem.z)(l x)(1 (1 x)(1+ x2) 2 1 for 0 5 x 5 1 we have + + + z2). By the BolzanoCauchy convergence principle for limits of functions.23 can be applied with g ( x ) = 0 and G ( x ) = show integrability.1]. as the primitive is not an elementary function. 1 / d m is Since 05 1 diT? < 1 diTi and Theorem 15. The integral 0 Example 15.x4 = (1 . also called Euler's integral of the first kind. 1 .
2 Use the previous exercise to show the integrability of l/JZE and nonintegrabilty of 1/ sinx on [0. respectively. many other integrals occurring in applications can be expressed in terms of the Beta function. For instance 00 appear in physics and probability.10. Except for some special values of p and q the Beta function cannot be expressed in terms of elementary functions of p and q.23. 13.x)q'd x. We remedy that now.71) whenever T X is a &fine tagged division of a bounded interval [A.q) = 1 1 x*'( 1 . However. Exercises @ Exercise 15. B] 3 I n [K. [Hint: 40 Use Theorem 15.1] if and only if a < 1.1 Prove: I f f is integrable on [c. 1 3 for 0 < c < 1 and limxaf(x) = L # 0 then f is integrable on [0.4 A number I is the integral o f f over I if for there are a gauge S and a positive number K such that IWTX) 4< E (15. K ] .1 Exercise 15. B] with I 3 [A. Integral of f over I is denoted by . We denote by I any of the following intervals positive E 15.11 Integrals over unbounded intervals In applications one encounters integrals over unbounded intervals.488 Introduction to Mathematics with Maple equation B(p. So far we have not defined integrals over unbounded intervals. 15.10. as in the above example.
or by similar symbols where f is replaced by f (2) f(u) d u .72) 15< x i ti1 x i x i S(Xi) & . Let T X be a 6fine tagged division of [l.3(1+ E ) <1+ 3+2& 3' . We choose 6 ( z j= and K > 3 / ~ . There is at most one number I satisfying the requirements of the definition: in other words. Motivated by the Fundamental Theorem and by the fact that s.~ we shall approximate (ti Xf 1 We wish to show that I C .1 and is omitted. Example 15. > z' = I C . b] then this definition has the same meaning as Definition 15. The proof is rather similar to the proof of Theorem 15.tn] K . If I is one of the intervals in (15.3' tn 1 & (15.~ ~ =Z1..2. the number I is welldefined.Integrals 489 If I = [a.  = 1 ." ti1) by . The geometric meaning behind the definition above is that for a nonnegative function f on an unbounded interval I the area under the graph of f and above the xaxis is approximated by Riemann sums for &fine tagged divisions of sufficiently large bounded intervals.70) then is denoted by d IC or by respectively.31 m/3(1 with + E) d  dx t.> 1. etc. ti1 ti 1 1 FirstI y 1L secondIy $ (A ) .
73) J ! Remark 15. b+oo 1 > = 1. 1 I 1% $1 (& ). The integral f and for s”.24 (Hake) if f is KHintegrable on b+oo lim [a.20 and 15. f =J ! o o f + if it exists but the limit on the right hand side might exist and the integral might not. (15.72) gives A combined version of Theorems 15.22 for an infinite interval reads Theorem 15. b . By the Fundamental Theorem boo Jim [ + d z = lim (1 X . Using these estimates on the Riemann sum yields Combining this inequality with (15. for instance. .490 Introduction to Mathematics with Maple This leads to L t. If. 003 if and only [f exists and then b+oo lim J.32 We give a one line solution to Example 15. .31.of =L m f . f(z)= sinz or z then the limit is zero but the integral does not exist.   5 (ti_l 1 ) .16 Similar theorems hold for m f . Example 15..
It is defined its r(s)= for s lm exp(x)zsl d 2.33 by parts gives W e prove integrability of (sinx)/x on [ l . for a >K and b >a By the BolzanoCauchy Theorem the limit lim b+m Theorem 15. g) H B(p.Int egmls 491 Example 15. g such that dx) I I f (4I G(4 for all x E [a.74) > 0. For an application of this integrability test see Exercise 15. Example 15. a < c < 00. 001. Integrating Consequently.75) .a). For a positive integer s = n it becomes (15. 1 1 to conclude integrability of exp(x2) on ] .25 (Integrability test) I< for every c.00. also called Euler’s integral of the second kind. An analogous test holds for c + 00 if f is integrable on [c. (15. f is integrable on [a.00[ ] . 00[. Since exp(x2) 5 1/x2 for x # 0 the integrability test and can be applied with g(z) = 0 and G(x) = 1/x2 on both intervals [l. One of the most important higher transcendental functions is the function. c] and there are integrable functions G.10.00[.00[ then f is integrable on [a.00. The proof is very similar to the proof of the integrability test in Section 15.1.34 In this chapter we encountered the special functions erf and the Euler integral ( p .11. Theorem 15.23. b] for every c < b.
11. o o [ I ~ R be integrable on [a. Show that if a! > 1 then not exist.2 f exists and i f a! 5 1 then the integral does Use Maple to evaluate s". c] for every 2+m lim x"f(x) = L # 0.76) is often needed.. Since the sum of a series is nothing but the limit of partial sums.76) translates into . where we evaluate Jom 2 &csc( 2 1 9 7r) sin(.2066317466 Exercises @ Exercise 15.1 c > a and Let f : [ a . . interchange of limit and integration. 15. Equation (15.rr) qE) 5 10 3 r(5) > evalf(%).12 Interchange of limit and integration In pure and applied mathematics.11.492 Introduction to Mathematics with Maple Many integrals can be expressed in terms of the Gamma function as in our cos 22 T d x next example. according to the equation (15.. Exercise 15.exp(x2)dx.
6 for all n > N . hence lim nmo 1 1 fn = 1. Under uniform convergence Equation (15. Proof. r n = l/n. 0 This theorem has simple wording and a very easy proof. The theorem is not applicable because the convergence is not uniform.5 uniformly convergent on I to f if r n + 0.$[(z) = 0 for every z E [0.26 [a.76) is valid with [a. Equation (15. we have If the functions f n are integrable on an interval Theorem 15.Integrals 493 It would be wrong to assume that Equation (15.% Let fn fn = 0. b] = [0. For instance. If f n = n l ~ o .1] and consequently Jo' . b] and the convergence fn + f is uniform on [a. the sequence n + (sinnz)/n is uniformly convergent to the zero function on R.11. By Theorem 15. If j n ( x ) = xn then. More precisely. It follows from the definition of uniform convergence that for every E > 0 there exists a number N such that for all x E [a. On the other hand lim nllo. b] then f is integrable and Equation (15.76) holds. since.1]..76) is valid. the condition of uniform convergence is sometimes too restrictive. The integrability of f follows from these inequalities and Theorem 15. and f be functions defined on an interval I . l / then ~[ Jof n n+m 1 = l'for every n E N. The theorem is no longer valid .76) always holds. in this case. It is a useful theorem. however. b] and all n > N . Define The sequence of functions n H f n is said to be Definition 15. as is easily checked by direct evaluation.
nevertheless.2. Theorem 15. The following inequality is obvious First we find n such that r. Proof. important. are functions continuous at c € ] a . For instance. if f ./n then fn converge uniformly to zero but The concept of uniform convergence is. b[ and converging uniformly on [a. as the next theorem shows. . 0 We now return to the interchange of limit and integration. b] to f then f is continuous at c. such that < &/3and then with this n fixed we find S > 0 The continuity of f at c follows. b] is replaced by an unbounded interval.27 If f . = ln.494 Introduction to Mathematics with Maple if the interval [a.
Then (15.35 Denote by iI( the length of the interval I . Example 15.2).78) provided the series converges.5. Theorem 3. . Then 1 exists. Here we illustrate the theorem by two examples. I f it is finite then f is integrable and if it is infinite then f is not integrable. For the proof we refer to Lee and V9born9 (2000.6.78) Example 15.36 (The Laplace integral) Integral exp(x2)dx is called the Laplace integral. Example 6. however they require knowledge which is not yet available to us.EZ1Ik. (iii) f n ( x ) L f n + l ( x ) for n E N and all 2 E I .Integrals 495 Theorem 15. then l ~converge . increasingly t o 1~ and Taking limits as n + 00 on both sides of this equation and applying the Monotone Convergence Theorem proves Equation (15. (ii) nmo lim f n ( x ) = f(x) for every x E I . Let where the intervals are pairwise disjoint. Now we show how this can be proved. or f n ( x ) 2 f n + l ( s ) for n E N and all x E I .28 (Monotone Convergence Theorem) that Assume (i) the functions f n are integrable on an interval I for every n E N. For a short proof see Lee and Vfbornf (2000.17 From the power series expansion of 17There are more elegant proofs. Let An = U. Maple already told us that it equals to &/2.9).
(15.81) Altogether. "nzn =exp(x 2 ).79) We evaluate the integral I n on the right hand side by changing variables x = f i c o s t . d x = .25 where we also proved t h a t &Sn + Hence m. Using the arithmeticgeometric mean inequality. . is increasing for every x.24) that the sequence. it can be proved similarly in Inequalities (10. with terms fn(2) (1 .23) and (10. (15. (15.81) Another important theorem on interchange of limit and integration is the next theorem.496 Introduction to Mathematics with Maple In( 1 x) it follows that In( 1 ax) =a x + 20 lim + and consequently lim n+oo ( 1.80) and (15. It follows from the Monotone Convergence Theorem that exp(z2)dx= nmo lim lfi(l:)n.s i n t d t and have We evaluated the integral S n in Example 15. we have by (15.x2/n)n for 0 5 x 5 & and fn(x) = 0 otherwise.79).
Integrals
497
Theorem 15.29 (Dominated Convergence Theorem) that
Assume
(i) the functions f n are integrable on an interval I for every n E N, (ii) n4co lim f n ( x ) = f (x)f o r every x E I , (iii) there exists integrable functions g and G such that
g(x) I fn(x) IG(x)
for all x E I and all n E N.
Then f is integrable on I and
1 I
Again, we only illustrate this theorem by an example.
Example 15.37 The r function was defined in Equation (15.74). We wish to prove the formula
r(t)= lim
noo
t(t
+ I )  (t + n ) '
a
n! nt
for
t > 0. It is not difficult t o show that
ln
Since
(1  X ) n u t  ' d u
= nt
1
1
(1  y)nytldy.
The first formula can be proved by induction, the second by substitution, with u = n y and d u = n d y. Hence we want to show that
we seek an integrable nonnegative function G such t h a t 0
5 f n 5 G, where
498
Introduction to Mathematics with Maple
The inequality ln(1  x / n ) 5  x / n follows from the power series expansion of ln(1 x ) and it yields (1  x / n ) n 5 e” for 0 5 x 5 n. W e can take G ( x ) = e”xtl, this function is integrable on [0, m[ because it is continuous and
+
(15.83)
The first inequality guarantees integrability of Ic > t 1 on [I,m[.
+
G on [0,1], the second with
15.13
Comments and supplements
The basic ideas for integration can be found in antiquity, in the exhaustion method and in the work of Archimedes. There is a long list of mathematicians contributing to the subject before Newton and Leibniz, including Fermat, Cavalieri, Kepler; see Eves (1981). Surprisingly, the Fundamental Theorem of Calculus was discovered by Barrow, a predecessor of Newton at Cambridge. Newton and Leibniz developed systematic methods for calculating derivatives and integrals and open ways to apply calculus to problems in other sciences. The logical foundations of Calculus had some important critics. The mathematicians of the seventeenth and eighteenth century were aware of the shortcomings of the logical foundations but had different priorities. D’Alembert even said: “Let’s move ahead, the confidence will arrive later.” The first definition of integral which can stand up to modern scrutiny was Riemann’s. Shortly before him, Cauchy proved that the rightsums (and the leftsums) converge to a limit when the maximal length of the subintervals tended to zero. Riemann not only allowed the tag to float anywhere in the subinterval but made a decisive step forward by considering the set of (Riemann) integrable functions, namely those functions for which the Riemann sums had a limit as the maximal length of subintervals tended to zero. Riemann’s definition was originally a success, but as time went by, it was realized that the definition was far from ideal. One of the deficiencies of Riemann theory is that none of the monotone convergence theorem, the dominated convergence theorem or the Hake theorems are valid for the Riemann integral. Lebesgue set himself a goal to create an integration theory such that:
(i) To every function bounded on the interval [a, b] there is a number
Integrals
499
I associated with it such that the following axioms are satisfied. If I = s,” f = s,” f(z)d z then
(ii)
(iii) for all a , b, c
[f +
(iv) If f 2 0 and a
JC
b
f+
la f
= 0;
< b then
(v) For all h, a , b
(vi)
l1
b
l d x = 1;
(vii) If f n ( x ) 5 fn+l and f n ( z )+ f ( x ) for every x in [a, b] then
. . . J , J f.
fn
b
=
In Lebesgue’s requirement (vii) we recognize the monotone convergence theorem. Lebesgue created the integration satisfying his axioms. His theory had an immediate and beneficial impact on mathematical analysis and is still the integration theory predominantly used by professional mathematicians. Lebesgue, however, did not define the integral for every bounded function. Instead, he created a class of summable, or Lebesgue integrable functions. These include all Riemann integrable functions and also many unbounded functions. However, in Lebesgue theory, Hake’s theorems are still not valid and neither is the Fundamental Theorem, at least not in such a simple and powerful form as in Theorem 15.15. Lots of effort was devoted to making Lebesgue theory accessible to nonspecialist physicists and engineers. Some of the users of integration can and do master Lebesgue theory. Experience shows, however, that Lebesgue theory can be too subtle and
500
Introduction to Mathematics with Maple
too difficult for general consumption. In contrast, the Kurzweil integration theory is relatively simple and has all the advantages of Lebesgue theory. This is the reason why we made perhaps an unusual step in adopting it for an elementary exposition. Kurzweil’s definition is equivalent to two definitions proposed during the time after Lebesgue and before Kurzweil: one by Perron and another by Denjoy. Theories based on these definitions are even more demanding that the Lebesgue theory. In recent years, several publications presented Kurzweil theory in detail: we recommend Lee and Vfbornf (2000) and Bartle (2001), where further references can be found.
Appendix A
Maple Programming
A.l
Some Maple programs
Introduction
A. 1.1
It is not our aim to expound Maple programming. However, some examples can be well illuminated by exposing the entire structure of the computation by a Maple program. So we give here the mere rudiments of Maple programming, just what we need in these examples. One occasion when readers might find it profitable to write their own program is when they need to carry out the same operation1 many times. A Maple procedure starts with its name, followed by the assignment operator :=, the word proc and parameters to be operated on, enclosed in parenthesis. Hence it looks like
name := proc(parameters)
Wide choice is allowed for the parameters. They can be integers, real or complex numbers, polynomials or even sets. You must end the procedure either with end proc : or end proc ;.2 The semicolon is better: Maple comes back and prints the procedure or gives an error message. The program itself consists of commands operating on the parameters, each individual command ending with a ; character or a : character. The symmetric difference of two sets is the set which contain all elements which lie in one of the sets but not in both. Our first procedure computes symmetric difference of two sets.
>
s d i f f :=proc(A,B)
'Not available in Maple itself. 21n older versions of Maple, use just end;
501
502
Introduction to Mathematics with Maple
> >
(A union B) minus (A intersect B); end proc;
sdiff := proc(A, B ) (A union B ) minus (A intersect B ) end proc
Let us test our program
>
A:={1,2,3);
B:={2,3,4);
A
:= (1, 2, 3)
B := (2, 3, 4)
>
sdiff (A,B) ;
A.1.2
The conditional statement
When programming, it is important that the computation be able to branch according to some condition which can only be tested midway during the computation itself. This is achieved in Maple by the statement if. The most basic form of this statement has the structure
if condition(s)
then cornrnand(s) else commands end if;
It is important that you end each if statement with end if;.3 The use of if is best explained by our next example in which the procedure picks up from two sets the one which has the smaller number of elements. In the example we use the command nops(A) which counts the number of elements of A.
> > > >
smaller :=proc (A, B) if nops(A)<nops(B) else B; end if; end proc; then A;
31n older versions of Maple you have to use f i instead of end i f .
Maple Programming
503
smaller := proc(A, B) if nops(A)
< nops(B) then A else B end if end proc
Testing the procedure, we have
> smaller(A,B) ;
Note that our program picked up B as the smaller set; this is as expected. If the number of elements of A is not smaller than the number of elements of B then the else alternative is activated. Let us say that we overlooked this and that we want both sets printed when the number of elements in the sets is the same. The next procedure does that, and provides an example of nested if statements. Note the indentation of the nested statements, which improves readability. If you need to use nested if statements more often, then you might like to learn the elif statement from Help.
>
> > > > > > >
Smaller:=proc (A ,B) if nops(A)<nops(B) then A; else if nops(A)=nops(B) then {A,B}; else B; end if end if; end proc;
Smaller := proc(A, B )
if nops(A) < nops(B) then A elseif nops(A) = nops(B) then { B , A} else B end if end if end proc
Testing again gives
504
Introduction to Mathematics with Maple
>
Smaller ( A , B) ;
A.1.3
The while Statement
The while statement allows one or more commands to be executed multiple times (as a loop). Commands are executed while a stated condition is satisfied, with execution terminating when the condition fails. The while statement has the structure
while c o n d i t i o n do command(s)
end do;
Do not forget to end the statement with end do; !4 The following procedure halves a number a until the quotient becomes smaller than In
the procedure we assign q to be a , in order to prevent this assignment from taking effect outside the procedure, we use the command local 9;. Actually, Maple automatically makes any variable introduced in a procedure, local to that procedure.
> >
>
pulka:=proc(a) local q; q:=a; while 10A5*q>i do
> >
>
q:=q/2;
end do; q;
>
> end proc: Let us test our program for a = 3. >
pulka(3); 3 524288
401der versions of Maple use od instead of end do.
Maple Programming
505
In the old days when computations were limited to pen and paper the following formula was used for efficiently computing the square root of a.
Some algebraic manipulation leads to the estimate
/Z within k decimal places we let the computation Hence in order to obtain , run while 2(xn  x,+1) 2
> >
rt2:=proc(a,k::nonnegint)
x:=a:
oldx:=x+lO*(k) : x: =O. 5* (x+a/x)
> while 2*(oldxx)>(1/2) *lO*(k) do >
>
oldx: =x: end do: end:
>
Warning, 'x' is implicitly declared local Warning, 'oldx' is implicitly declared local
We test our procedure and compare it with the square root procedure provided by Maple: > rt2(2,9) ;sqrt (2.0); rt2(7225,8) ; > sqrt (7225.0) ;
1.414213563 1.414213562
85.00000001
85.00000000
The discrepancies at the last decimal place are not caused by some imperfection of our program but by Maple using only ten significant digits
506
Introduction to Mathematics with Maple
(as a default) for the computation. Increasing the number of significant digits to, say twelve, will remove the blemish. Of course, we wrote our program as an illustration on the use of while and not for calculating the square root.
A.2 Examples
In our first example we automate the calculation of the greatest common divisor of two positive5 integers which we did on Page 59 by repeated use of the command irem.
Example A. 1 (Greatest common divisor, the Euclid algorithm) The idea is to run irem by using while as long as the result is not zero.
> > > > > > > >
> >
mygcd:=proc(m::posint,n::posint) local a,b,r; a :=m; b := n; while b > 0 do r := irem(a,b); a := b; b := r; end do; a; end;
mygcd := proc(m::posint, n::posint)
locala, b, r; a := m ; b := n ; while0 < bdor := irem(a, b) ; a := b ; b := rend do; a end proc
Let us test our procedure and compare the results with Maple igcd
> mygcd(987654321123456789,123456789987654321); > igcd(987654321123456789,123456789987654321) ;
1222222221 1222222221
5The case when one of the numbers is zero need no computing.
m.2 can be modified that it covers any complex number a+zb. t: a:=Re(A). for x <0 for x 2 0 (A. .EnvsignumO:=l. This is done by replacing b/lbl by S(b) where S is defined as follows S ( x )= {. > t :=sqrt (an2+bn2) . > sqrt( (t+a)/2)+I*signum(b)*sqrt( > end proc: (ta)/2) Here are a few numerical examples > compsqrt (1) . Example A. .2) The easy verification is left to readers. > compsqrt (1) . We can now write a procedure for computing the exact value of the square root of a complex number. we call it compsqrt > > compsqrt :=proc ( A ) local a.Maple Programming 507 The following example provides a program for calculating the exact value of the square root of a complex number. The function S differs from the Maple function signum only at zero. However the command hvsignumO= a. 21 Now we save the procedure compsqrt in a file with the extension . Note the quotation marks around the file name. We call the file Ourprocs . not only the case b # 0.m. b:=Im(A). redefines signum at zero to be a. b. I > compsqrt (34*1).2 The equation from Theorem 7. > > .
expression. expressions or programs can be saved in a file. expression.mII . In each case. The next Example illustrates the practical use of the idea in the proof of Theorem 12. In a future session you can recall the file and the function by the command read. .3 (Bisection method for finding roots) We halve the interval ( A . The structure of the command is save function. 6Function. . B ) to locate a zero of a given f within ten significant digits. "myfi1e. The following Maple worksheet illustrates this. and then the procedure can be used.z 1 Ifi 2 Other Maple objects can be saved the same way we saved compsqrt. There is a similarity to calling a package via the command with(package) .6 > restart.mn. .15.508 Introduction to Mathematics with Maple > save compsqrt. First we call the file with the read command then we use the procedure for finding the exact value of the square root > > read "Ourprocs.m" Note the comma after the words function and expression. 1f i . > compsqrt (1) .. "Ourprocs. . one has to recall the file first. . Example A. Several functions.
end proc. evalf(l/2 * a 1/2 * b) end if end proc + + + + + . b := b else b := 1/2 * a 1/2 * b . # (a+b) /2.B) > local a. ifevalf(f(A)) < 0 theng := f e l s e g := fend if.10000000000 * a) do if evalf ( g ( 1/2 * a 1/2 * b)) < 0 t h e n a := 1/2 * a 1/2 * b . end if. while 1 < evalf (10000000000* b . b:=B. # +(1/2)*b. else b:=(l/2)*a+(1/2)*b. b. B ) locala. a := a end if end do. g. end do. end if. a:=a. A.Maple Programming 509 > bisect:=proc(f . ifevalf(g(B)) < 0 thenError('f(A) * f ( B ) must be negative') else a :=A. while evalf ( 10nlO* (ba)) > l do if O>evalf(g(a/2+b/2)) then a:=(1/2)*a b:=b. evalf ( (a+b) /2) .g.A. #it is easier to read it in decimal form. > > > > > > > > > > > > > > > > > > > > > > > if O>evalf ( f( A ) ) then g:=f else g:=f end if. bisect := proc(f. 1/2 * a 1/2 * b. # #now we make sure that f takes values #of opposite signs at the ends of the interval # if evalf ( g ( B ) ) < O then Error ( ' f (A) *f (B) must be negative' ) else a:=A.b. b:= B .
Other methods might become unstable if more than one zero of f is present in [A.895494267 68719476736 save bisect. > 41462209955 T . > f:=x>(1/2)*xsin(x).510 Introduction to Mathematics with Maple An advantage of this method is that it always finds a zero.532088886 274877906944 ' The use of bisect is not restricted to polynomials.mtt .7] and [2. 1.sin(x) 1 2 bisect (f . f :=x+x33x+1 The polynomial has zeros in [2. Let us see what happens for [3.B].879385242 137438953472 421 137386305 1.Pi/2.13 and [l. 21. [0. We now illustrate the use of the procedure on several examples. 1. 258300740771. f := x > +.5]. > f:=x>x^33*x+l. 11.x . "Bisect.Pi) . .
An everywhere continuous nowhere differentiable function. (2000). (1963). Chelsea. K. P. Spivak. xE1 511 . E. (1956). R. K. P. Amer. Mir. (1978). Foundations of analysis. 68. Matsuoka. Rhode Island: American Mathematical Society. The Mathematical Association of America. Foundations of analysis. J. R. New York. (1978). An elementary proof of the formula l/k2 = 7r2/6. Field theory and its classical problems. Gradshtein. (1967). Y. & Vfbornf. M. 485487. Eves. Cambridge. Lee. The Mathematical Association of America) McCarthy. Arrow impossibility theorems. Math. Boas. The integral: An easy approach after Kurxweil and Henstock. P. Knopp. (1996). R. C. Math. Infinite sequences and series. G. Amsterdam: W. (1980). Amer. (1951). Social choice and individual values. Y. I. Monthly. & Ryzhik. J. California: Wadsworth. New York: Academic Press. UK: Cambridge University Press. Benjamin. A primer of real functions. R. Naive set theory.. Kelly. A. Hadlock. (1981). 60. Landau. Table of integrals. Monthly. (1972). (1974). Great moments in mathematics (after 1650). H. A modern theory of integration. Kurosh. (1961). Halmos. Providence. (2001). R.References Arrow. Stromberg.A. Mathematical Association of America. I. Springer. 709. Higher algebra. R. New Haven: Yale University Press. (Reprinted in Selected papers on calculus. K. Boston: Academic Press.. Inc. (1953). Belmont. Mathematical Association of America. series and products (computer file). New York: Dover. Bartle. Calculus. (1981).
Bolzano’s anniversary. Monthly. (1987). Differentiation of power series. Amer. R. 44. V$born$. (1989). R. R. L’HGpital rule. (1972). Math. . B. Youse. 116121. 28. 369370. R. The Australian Mathematical Society Gazette. Introduction to real analysis. (2001). 177183. Elemente der Mathematik. & Nester.. 94. Boston: Allyn and Bacon.512 References Vyborny. a counterexample. Vyborny.
Maple commands. 26 37 combine conjugate 25. such as collect. such as Exit. but are provided here to assist in running a Muple session.l: List of Maple commands used in this book Command abs (x) annuity Apollonius Description The absolute value of the number x Calculates quantities relating to annuities: requires the finance package to be loaded Calculates and graphs the eight circles which touch three specified circles: requires the geometry package to be loaded Make an assumption about a variable: holds for the remainder of the worksheet (or until restart) Make an assumption about a variable: holds only for the command it follows Evaluates the binomial coefficient The ceiling of x (the smallest integer not less than 2) Carry out a change of variables in an integral Collect similar terms together A collection of commands for solving problems in combinatorial theory: loaded by using the command with(combinat) Combine expressions into a single expression The complex conjugate B of a complex number z Continued on Pages 75 31 30 assume 112 assuming binomial ceil (x) changevar collect combinat 111 151 75. Table A. Any which are set in normal type.Index of Maple commands used in this book The following table provides a brief description of the Maple commands used in this book. 26 199 next page 513 . strictly. Most of the commands in the following table are set in the type used for Maple commands throughout this book. and some of the pages on which they are used. are not. 118 474 24. together with the pages on which they are described.
141 8 12 + 75 21.118 94. 187 189 finance 31 f l o o r (x) f solve geometry 75. 173. 203 16. 17.. The exponential function ex Expand a n expression into individual terms Factorise a polynomial The same as f a c t o r except that calculations are carried out in a field modulo a prime p A collection of commands for financial calculations: loaded by using the command with (f inance ) The floor of z (the largest integer not greater than x) Find numerical solution(s) of an equation Pages 17. 26 22. 215 465 199.7182818284. 25. 17. 2 10 30 A collection of commands for solving problems in geometry: loaded by using the command with (geometry) Continued on next page . 16 30. 26. 209. 184 75 36 1 359 15. .514 Index of Maple commands List of Maple commands (continued) Command convert cos (x) dif f D(f ( a > Digits divisors e r f (x) evalc evalf Exit exp(1) exp (XI expand factor Fact o r Description Convert from one form to another The trigonometric function cos x Differentiate a function Evaluates the derivative of the function f at x=a Set the number of digits used in calculations Find all the divisors of an integer: requires the numtheory package to be loaded The error function: Changes a complex valued expression into the form a bz Evaluate an expression in numerical terms Exit from Maple session The base of natural logarithms: e = 2.
26 30 OP Pi plot powerset 41. 176 next page . Plot a function Finds the sets comprising the power set of a given set: requires the combinat package to be loaded The product of a number of terms Quotient when one polynomial is divided by another Continued on Pages int intersect intparts iquo irem is isolve ithprime I n (x> log (XI max(x. y.z> minus nops normal numthe ory 443 38 470 29 29 80 27. s) m i n h . 28 188 75 75 75 377 75 38 40 23. .1415926535 . 81 12 81 37 product QUO 143 29. .z) map(f.Index of Maple commands 515 List of Maple commands (continued) Command Description Integrate a function Intersection of two sets Carry out an integration by parts The quotient when one integer is divided by another The remainder when one integer is divided by another Test the value of a Boolean function Find integer solutions of equations Finds the ith prime number The logarithmic function In x The logarithmic function logx (the same as In x) The maximum of two or more real numbers Map elements of the set S into another set by the function f The minimum of two or more real numbers Difference of two sets Counts the number of items in a list or set Collects a sum of fractions over a common denominator A collection of commands for solving problems in number theory: loaded by using the command with(numtheory) Select one or more items from a list or set The value of 7r: 3.y.
142. 26 141. 176 177 81 19 216 80 41 22. 13. 75. 143 75 108. 209 24.26. or a Taylor series Pages 177 27 29. 183 Quo rationalize rem R e m remove restart roots select seq simplify s i n (x) solve sort spline sqrt(x) subs SUm surd t d x ) tau taylor Continued on next page . 289 256 75 30 173. 12. into another expression The sum of a series Finds the real value of an odd root of a real number The trigonometric function tan z The number of factors of an integer Used to find the product of two polynomials. 26 352 11.516 Index of Maple commands List of Maple commands (continued) Command Description The same as quo except that calculations are carried out in a field modulo a prime p Rationalise a result with a complicated denominator Remainder when one polynomial is divided by another The same as rem except that calculations are carried out in a field modulo a prime p Remove one or more items from a set or a list Reset the Maple environment to its starting setting with no variables defined Finds rational roots of a polynomial Selects some elements from a list Used to generate an expression sequence Simplify a complicated expression The trigonometric function sin x Find a solution of an equation Sort a sequence of terms Approximate a function by straight lines or curves The square root function Substitutes a value. 143. or an expresseion.
6 .Index of Maple commands 517 List of Maple commands (continued) Command union with Description Union of two sets The command with(newpackage1 loads the additional Maple commands available in the package newpackage Using worksheets in Maple Pages 38 30 worksheets 5.
This page intentionally left blank .
115 bounded. 8 . 221 Cartesian product. 11 /. 115 bounded below. 13 %. Georg diagonal process. 9 #. 11 Arrow. 392 Bernoulli. 7 Barrow. 236 axioms. 306. 115 Cantor.1. 100 Peano. 451 bound lower. 235 Cardano H. 427 Archimedes. 455 a r c c o t . 76 algebra. 98 0. 6 :=. 116 of choice. 498 arcsin. 1 8 antiderivative. 99. 54 assigning variables. 11 Abel. Kenneth. 11 *. 467 abs. 390. see sets. 270.Index !. . 304 .. 220. 18 ?.xi 11 +. 115 bounded above. 273 function. 97 GF(2). 53 GF(4). Cartesian product case sensitivity. 11 8 1. 199 arithmetical operators. 425 arctan. 11 . 192 order. 121 ring. 309 inequality. J . 157 rule. 498 Bernoulli. 402 BolzanoCauchy principle. 18 axiom least upper bound. 309 condensation test. 114. 271. 390 binomial coefficient.. 400 Bolzano. 193 519 backspace. 72 'I. 55 field. 309. 426 Argand. 7 Cauchy. 57. 115 upper. Euclid. 1.
16 exact numbers. 13 floor. 359 from the right. 500 derivative definition. 51. 171 floating point number. 332 from the left. 196 imaginary part. 488 formulae. 498 ceil. 308 Cavalieri. 153 degree of a polynomial. 196 complex conjugate. 343 D’Alembert. 488. 333 of the inverse. 297 Darboux property. 400 Cohen. 324 mean value theorem.. 358 limit.A. 196 geometric representation. 430 cot. 171 Denjoy. 476 characteristic function. 124 definition Cauchy. 401 derivates. 13. 498 limit test. 14 floating point numbers. 298 test. 197 trigonometric form.. 427 Cousin lemma. 76 change of variables. 340 on intervals. Augustus. 199 imaginary. 293 nth root of a. 388 product. 169 formula . 359 Descartes R. 272 definition. 358 from the left. 184 domain. 44 decimal fraction. 199 constants in Maple. 72 coefficient binomial. 63 e. 421 Euler ’s integral. 493 cos. 306 divisor. 218 Dini. 221 field isomorphic. 76 formal power series. 313 limit from the right. 39 Fermat. 13 complex number absolute value. 56 comments. 195 square root. 498 Ferro del S. 324 derivative. 401 derivate. 472. 306 Dedekind cut. 76 cosh. 401 Dirichlet. 346 convergence uniform.520 Introduction to Mathematics with Maple convergence principle. 196 real part. 12 Euclid. 316 limit from the left. Paul. 313 definitions Absolute and Conditional Convergence. 195 purely imaginary. 76 expression sequence. 14 exp. 355 de Morgan. 492 evalf 0 . 124. 162 Dedekind. 333 from the right. 338 uniform. 12 continuity at a point. 55 Euler.
499 Leibniz. 237 Horner. 75 limit at infinity. 7 interval. 4 8 0 mean value theorem. 419 step. C. 504 log. 180 if statement. 498 formula. Oliver. 4 8 0 remainder. 218 fraction decimal. 498. 499 integration. 385 integral. 91 onetoone.Index 521 reduction. 4 8 9 integral domain. 481 Laplace.. Leopold. 131 inequality arithmeticgeometric mean. 55. 171 Kepler. 104 end points. 39 In. 380 series. 90 decreasing. 498 axioms. 484. 157 inference. 431 Hermite. Kurt. 174 integration by parts. 75 Godel. 104 isomorphic. 492 function. 476 interrupting commands. 490 Hake ’ s theorem. 472 definition. 69 bijection. 468 by substitution. 56 Galois. 7 6 Map 1 e . 484. 104 closed. 9 Henstock. 495 integral. 414 test. 261 Hilbert. 480. 162 function. 438 Lagrange. 450 surjective. 4 9 8 Kronecker. 438 over unbounded intervals. 470 formulae Euler. 3 4 6 Functions in MupZe. 199 Gauss K. 113 Kurzweil integrable. 72 help. 7 6 local. 257 injective. 158 Bernoulli. 213 greatest lower bound. F. 90 inverse. 471 Wallis. 90 periodic.. Henri. 117 Hake. 502 increasing at a point. 438. R. 353 list. 90 uniformly continuous. 171 isomorphism. 387. 316 limit inferior. 220 Gamma. 131 infimum. 499 Heaviside. 378 induction hypothesis. 104 open. 421 Fourier J. 490. 292 Library of functions. David. 495 Lebesgue. 353 limit superior. 117 inflection. 492 Gauss plane. 90 onto. 472. 257 increasing.
7 Polya. 150 Leibniz formula. 171 of polynomials. 23 simplify(). 22 int 0 . 170 unending. 43 Raabe test. 169 polynomials multiplication. 167. 66 transitive. 501 products in Maple. 397 Peano. Bertrand. 63 reducible. 12 R . 477 axiomatic approach. 143 Q . 24 Maclaurin. 375 Moivre. 43 Newton. 301 range. 43 quitting Map'le. 257 . 76 minimum local.522 Introduction to Mathematics with Maple commands. 43 Peano. 15 expando. 220 Russell. the set of positive real numbers. 166 axioms. 17 collect0. 180 semicolons. 170 root multiple. 498 binomial theorem. the set of real numbers. 66 symmetric. 6 sequence. the 43 set of natural numbers. 53 scheme Horner. 63 inverse. 170 previous results. G. 173 ring of. 185 relation. 309 I F ' . the set of nonnegative integers. the set of rational numbers. 127 remainder.0. 157 min. 432 Riemann. 360 Digitso. 500 pi. 443 normal 0 . 457 nops. 214 Ruffini Paulo. 375 mean arithmetic. 217 primitive of unity. 304 radius of convergence. 21 factor0. 76 maximum local. 400 max. 145 decreasing. 22 sort(). 158 geometric. 170. 443 Using convert0 . No. 134 polynomial. 513 integration. 8 N. 306 ring isomorphic. 438 sum. 66 Riemann integrable. 204 rational. 91 reflexive.24 diffo. Bernard. 114 Perron. 201 multiple commands. 8 proc. Giuseppe. 121. 502 Oresme.
76 starting Maple. 124 monotone convergence. 340 Weierstrass. 171 . 148 increasing. 181 Taylor theorem. 181 Taylor polynomial. 273 dominated convergence. 299 theorem binomial. 74 signum.Index 523 Fibonacci. 495 square root. 53 sinh. 232 difference. 502 step function. 257 series formal power. 344. 471 What is Maple?. 497 greatest lower bound. 36 enumerable. 477 test D’Alembert. 150 BolzanoCauchy. 342. 437 tan. 4 statement if. 141 tagged division. 117 least upper bound. 347 Vieta. 304 root. 271. 470 formula. 297 D’Alembert limit. 450 stopping Maple. 292 Raabe. 232 sets Cartesian product. the set of integers. 43 zero polynomial. 430 spaces. 76 singleton. 214 von Neumann. 354 BolzanoWeierstrass. 55 Taylor. 431 &fine. 298 Leibniz. 35 sgn. 4 Z. John. 477 The Intermediate Value Theorem. Alfred. 218 sums in Maple. 7 sqrt. 76 Tarski. 180 Taylor expansion. 291. 43 countable. 35 union. 198 Taylor. 169 set countable. 74 sin. 400 polynomial. 8 Stusm. 96 Wallis. 232 intersection. 11.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.