(Andrei Bourchtein, Ludmila Bourchtein) CounterExa PDF

Mathematics
TEXTBOOKS in MATHEMATICS TEXTBOOKS in MATHEMATICS
CounterExamples
CounterExamples
This book provides a one-semester undergraduate introduction to
counterexamples in calculus and analysis. It helps engineering, natu-
ral sciences, and mathematics students tackle commonly made er- From Elementary Calculus to
roneous conjectures. The book encourages students to think criti-
cally and analytically, and helps to reveal common errors in many the Beginnings of Analysis
examples.
In this book, the authors present an overview of important concepts
and results in calculus and real analysis by considering false state-
ments, which may appear to be true at first glance. The book cov-
ers topics concerning the functions of real variables, starting with
elementary properties, moving to limits and continuity, and then to
differentiation and integration. The first part of the book describes
single-variable functions, while the second part covers the functions
of two variables.
The many examples presented throughout the book typically start
at a very basic level and become more complex during the develop-
ment of exposition. At the end of each chapter, supplementary exer-
cises of different levels of complexity are provided, the most difficult
of them with a hint to the solution.
This book is intended for students who are interested in develop-
ing a deeper understanding of the topics of calculus. The gathered
Bourchtein
counterexamples may also be used by calculus instructors in their Bourchtein
classes.
Andrei Bourchtein
K23601
Ludmila Bourchtein
K23601_cover.indd 1 8/1/14 9:05 AM

CounterExamples
From Elementary Calculus to
the Beginnings of Analysis
K23601_FM.indd 1 8/4/14 12:49 PM

TEXTBOOKS in MATHEMATICS
Series Editors: Al Boggess and Ken Rosen
PUBLISHED TITLES
RISK ANALYSIS IN ENGINEERING AND ECONOMICS, SECOND EDITION
Bilal M. Ayyub
COUNTEREXAMPLES: FROM ELEMENTARY CALCULUS TO THE BEGINNINGS OF ANALYSIS
Andrei Bourchtein and Ludmila Bourchtein
INTRODUCTION TO THE CALCULUS OF VARIATIONS AND CONTROL WITH MODERN APPLICATIONS
John T. Burns
MIMETIC DISCRETIZATION METHODS
Jose E. Castillo
AN INTRODUCTION TO PARTIAL DIFFERENTIAL EQUATIONS WITH MATLAB®, SECOND EDITION
Mathew Coleman
RISK MANAGEMENT AND SIMULATION
Aparna Gupta
ABSTRACT ALGEBRA: AN INQUIRY-BASED APPROACH
Jonathan K. Hodge, Steven Schlicker, and Ted Sundstrom
QUADRACTIC IRRATIONALS: AN INTRODUCTION TO CLASSICAL NUMBER THEORY
Franz Holter-Koch
GROUP INVERSES OF M-MATRICES AND THEIR APPLICATIONS
Stephen J. Kirkland
AN INTRODUCTION TO NUMBER THEORY WITH CRYPTOGRAPHY
James Kraft and Larry Washington
REAL ANALYSIS AND FOUNDATIONS, THIRD EDITION
Steven G. Krantz
ELEMENTS OF ADVANCED MATHEMATICS, THIRD EDITION
Steven G. Krantz
APPLYING ANALYTICS: A PRACTICAL APPROACH
Evan S. Levine
ADVANCED LINEAR ALGEBRA
Nicholas Loehr
PUBLISHED TITLES CONTINUED
DIFFERENTIAL EQUATIONS WITH MATLAB®: EXPLORATION, APPLICATIONS, AND THEORY

Mark A. McKibben and Micah D. Webster
APPLICATIONS OF COMBINATORIAL MATRIX THEORY TO LAPLACIAN MATRICES OF GRAPHS
Jason J. Molitierno
ABSTRACT ALGEBRA: AN INTERACTIVE APPROACH
William Paulsen
ADVANCED CALCULUS: THEORY AND PRACTICE
John Srdjan Petrovic
COMPUTATIONS OF IMPROPER REIMANN INTEGRALS
Ioannis Roussos
K23601_FM.indd 3 8/4/14 12:49 PM

This page intentionally left blank
TEXTBOOKS in MATHEMATICS
CounterExamples
From Elementary Calculus to
the Beginnings of Analysis
Andrei Bourchtein
Ludmila Bourchtein
Pelotas State University, Brazil
K23601_FM.indd 5 8/4/14 12:49 PM

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2015 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works

Version Date: 20140804
International Standard Book Number-13: 978-1-4822-4668-1 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a photo-
copy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
To Haim and Maria with love and gratitude;
to Maxim with love and pride
Contents
Preface xiii
List of Figures xv
Introduction xix
0.1 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
0.1.1 On the structure of this book . . . . . . . . . . . . . . xix
0.1.2 On mathematical language and notation . . . . . . . . xxi
0.2 Background (elements of theory) . . . . . . . . . . . . . . . . xxii
0.2.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxii
0.2.2 Functions . . . . . . . . . . . . . . . . . . . . . . . . . xxvii
I Functions of one real variable 1

1 Elementary properties of functions 3
1.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Function definition . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 Periodicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Even/odd functions . . . . . . . . . . . . . . . . . . . . . . . 19
1.6 Monotonicity . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7 Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2 Limits 35

2.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3 Elementary properties (arithmetic and comparative) . . . . . 44
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3 Continuity 53

3.2 Local properties . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.3 Global properties: general results . . . . . . . . . . . . . . . . 62
ix
x Contents
3.4 Global properties: the famous theorems . . . . . . . . . . . . 69

3.4.1 Mapping sets . . . . . . . . . . . . . . . . . . . . . . . 69
3.4.2 Weierstrass theorems . . . . . . . . . . . . . . . . . . . 72
3.4.3 Intermediate Value theorem . . . . . . . . . . . . . . . 74
3.5 Uniform continuity . . . . . . . . . . . . . . . . . . . . . . . 80
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4 Differentiation 89

4.2 Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Local properties . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.4 Global properties . . . . . . . . . . . . . . . . . . . . . . . . 106
4.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.5.1 Tangent line . . . . . . . . . . . . . . . . . . . . . . . 112
4.5.2 Monotonicity and local extrema . . . . . . . . . . . . . 113
4.5.3 Convexity and inflection . . . . . . . . . . . . . . . . . 118
4.5.4 Asymptotes . . . . . . . . . . . . . . . . . . . . . . . . 123
4.5.5 L’Hospital’s rule . . . . . . . . . . . . . . . . . . . . . 126
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5 Integrals 133

5.2 Indefinite integral . . . . . . . . . . . . . . . . . . . . . . . . 140
5.3 Definite (Riemann) integral . . . . . . . . . . . . . . . . . . . 144
5.4 Improper integrals . . . . . . . . . . . . . . . . . . . . . . . . 152
5.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6 Sequences and series 175

6.2 Numerical sequences . . . . . . . . . . . . . . . . . . . . . . . 180
6.3 Numerical series: convergence and elementary properties . . 185
6.4 Numerical series: convergence tests . . . . . . . . . . . . . . . 189
6.5 Power series . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
II Functions of two real variables 205

7 Limits and continuity 207

7.2 One-dimensional links . . . . . . . . . . . . . . . . . . . . . 212
7.2.1 Concepts and local properties . . . . . . . . . . . . . . 213
Contents xi
7.2.2 Global properties . . . . . . . . . . . . . . . . . . . . . 217

7.3 Multidimensional essentials . . . . . . . . . . . . . . . . . . . 223
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
8 Differentiability 237
8.1 Elements of Theory . . . . . . . . . . . . . . . . . . . . . . . 237

8.2 One-dimensional links . . . . . . . . . . . . . . . . . . . . . . 242
8.2.1 Concepts and local properties . . . . . . . . . . . . . . 242
8.2.2 Global properties and applications . . . . . . . . . . . 247
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
9 Integrability 287

9.2 One-dimensional links . . . . . . . . . . . . . . . . . . . . . . 293
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
Bibliography 323
Symbol Description 325
Index 327
Preface
In this manuscript we present counterexamples to different false state-

ments, which frequently arise in the calculus and fundamentals of real anal-
ysis, and which may appear to be true at first glance. A counterexample is
understood here in a broad sense as any example that is counter to some
statement. The topics covered concern functions of real variables. The first
part (chapters 1-6) is related to single-variable functions, starting with ele-
mentary properties of functions (partially studied even in college), passing
through limits and continuity to differentiation and integration, and ending
with numerical sequences and series. The second part (chapters 7-9) deals
with function of two variables, involving limits and continuity, differentiation
and integration.
One of the goals of this book is to provide an outlook of important concepts
and theorems in calculus and analysis by using counterexamples. We restricted
our exposition to the main definitions and theorems of calculus in order to
explore different versions (wrong and correct) of the fundamental concepts and
to see what happens a few steps outside of the traditional formulations. Hence,
many interesting (but more specific and applied) problems not related directly
to the basic notions and results are left out of the scope of this manuscript.
The selection and exposition of the material are directed, in the first place,
to those calculus students who are interested in a deeper understanding and
broader knowledge of the topics of calculus. We think the presented material
may also be used by instructors that wish to go through the examples (or their
variations) in class or assign them as homework or extra-curricular projects.
In order to make the majority of the examples and solutions accessible to
calculus students, we tried to avoid more advanced notions and results of real
analysis. At the same time, we believe that some examples, despite having
a simple formulation, are not so simple to solve, and can be of interest to
students of real analysis as well.
It is assumed that a reader has knowledge of college algebra. For reading
of the first two chapters, a familiarity with the university calculus courses is
not required. The chapters 3-6 are accessible for students who have learned
basic concepts of the first two semesters of calculus, and the chapters of the
second part require some knowledge of functions of several variables. The
logical sequence of the material just follows the chapter sequence, that is, for
xiii
xiv Preface
reading of a specific chapter one does not need any knowledge of the next
chapter subjects.
This book is not appropriate as the main textbook for a course, but rather,
it can serve as a supplement that can help students to master important
concepts and theorems. So we think the best way to use this book is to read
its parts while taking a respective calculus course: the chapters 2-4 correspond
roughly to the topics of the first semester calculus, chapters 5-6 - to the second
semester, and chapters 7-9 - to the third semester. On the other hand, the
students already familiarized with the subjects of calculus sequence can find
here deeper interpretation of the results and finer relation between concepts
than in standard presentations. Also, more experienced students will better
understand provided examples and ideas behind their construction.
To facilitate the reading of the main text (containing counterexamples)
and make the text self-contained, and also to fix terminology, notation and
concepts, we gather the relevant definitions and results in the introductory sec-
tions of each chapter. Additionally some basic notions of the theory of sets and
functions used everywhere in the book are presented in the Introduction chap-
ter. For many examples, we make explicit references to the concepts/theorems
to which they are related.
A representative (but by no means exhaustive) bibliography can be found
at the end of the book, including both collections of problems and textbooks
in calculus/analysis. On the one hand, these references are the sources of dif-
ferent examples collected here, although it was out of our scope to trace all
the original sources. On the other hand, they may be used for finding further
information (examples and theory) on various topics. Some of these references
are classic collections of the problems, such as that by Demidovich [3] and by
Gelbaum and Olmstead [5]. Our preparation of this text was inspired in the
first place by the latter book. We tried to extend its approach to a broader
and less specialized range of topics in calculus and analysis by providing an
alternative set of counterexamples, which require less training in mathemat-
ics and are accessible not only for mathematics majors, but also for natural
science and engineering students. We hope that both mathematics and non-
mathematics students will find our book as stimulating, challenging and useful
as we found the above two books during our student days and as we continue
to find it until now.
List of Figures
1.2.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.4 Example 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.5 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2.6 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.2 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.4.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.5.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.6.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.6.4 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.6.5 Example 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.7.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.7.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.7.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.2.2 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.4 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.1 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3.2 Example 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3.3 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.1 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2.2 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.3.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.4 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
xv
xvi List of Figures
3.3.5 Example 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.4.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.3 Example 9, Remark 1 . . . . . . . . . . . . . . . . . . . . . . 73
3.4.4 Example 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.4.5 Example 14, Remark 1 . . . . . . . . . . . . . . . . . . . . . 77
3.4.6 Example 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.4.7 Example 16 . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.5.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.5.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.5.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.5.4 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.2.1 Example 3, graph of the function . . . . . . . . . . . . . . . 95

4.2.2 Example 3, graph of the derivative . . . . . . . . . . . . . . 96
4.2.3 Example 4, Remark 2 and Example 5, the function f (x) . . 97
4.2.4 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.3.3 Example 6 and Example 7 . . . . . . . . . . . . . . . . . . . 104
4.4.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.4.2 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.4.3 Example 9, graph of the first function . . . . . . . . . . . . 111
4.5.1 Example 2, graph of the second function . . . . . . . . . . . 113
4.5.4 Example 10, graph of the function . . . . . . . . . . . . . . 116
4.5.8 Example 14, graph of the second derivative . . . . . . . . . 120
4.5.9 Example 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.5.11 Example 16, graphs of the first and second derivatives . . . 123
4.5.12 Example 17 . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.5.13 Example 18, the first counterexample . . . . . . . . . . . . . 125
4.5.14 Example 18, the second counterexample . . . . . . . . . . . 125
5.2.1 Example 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.3.1 Example 9, Remark 1 . . . . . . . . . . . . . . . . . . . . . . 149
5.4.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.4.2 Example 1, Remark 1 . . . . . . . . . . . . . . . . . . . . . . 154
5.4.3 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.4.4 Example 2, Remark 1 . . . . . . . . . . . . . . . . . . . . . . 156
5.4.5 Example 2, Remark 2 and Example 3 . . . . . . . . . . . . . 157
List of Figures xvii

5.5.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.5.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.5.3 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.5.4 Example 3, Remark 2 . . . . . . . . . . . . . . . . . . . . . . 166
5.5.5 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.5.6 Example 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

6.3.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
6.4.1 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
6.4.2 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
6.4.3 Example 9, the first counterexample . . . . . . . . . . . . . 193
6.4.4 Example 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.4.5 Example 12, Remark 1 . . . . . . . . . . . . . . . . . . . . . 197
6.4.6 Example 12, Remark 2 . . . . . . . . . . . . . . . . . . . . . 197
6.4.7 Example 13 . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
6.5.1 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6.5.2 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
7.2.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

7.2.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
7.2.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
7.2.4 Example 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
7.2.5 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
7.2.6 Example 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
7.2.7 Example 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.3.2 Example 1, the function on some paths through the origin . 224
7.3.4 Example 2, the function on some paths through the origin . 225
7.3.6 Example 5, the function on some paths approaching infinity 228
7.3.7 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
7.3.8 Example 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
7.3.9 Example 12 . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

8.2.2 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
8.2.3 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
8.2.4 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
8.2.5 Example 10, graph of the second function . . . . . . . . . . 249
8.2.6 Example 11, graph of the second function . . . . . . . . . . 249
8.3.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
xviii List of Figures
8.3.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

8.3.4 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
8.3.6 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
8.3.7 Example 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
8.3.8 Example 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
8.3.9 Example 17 . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
8.3.10 Example 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
8.3.12 Example 23 . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
8.3.13 Example 24 . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
9.2.1 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296

9.2.2 Example 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
9.2.3 Example 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
9.3.1 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
9.3.2 Example 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
9.3.3 Example 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
9.3.4 Example 6, division of the domain S . . . . . . . . . . . . . 308
9.3.5 Example 6, graph of the function f . . . . . . . . . . . . . . 309
9.3.6 Example 11, right rectangle . . . . . . . . . . . . . . . . . . 315
9.3.7 Example 11, left rectangle . . . . . . . . . . . . . . . . . . . 315
Introduction
0.1 Comments
0.1.1 On the structure of this book
This book consists of the Introductory chapter and nine chapters of coun-
terexamples divided into two parts. In the first part (chapters 1-6), the sub-
jects of single-variable functions, corresponding approximately to the first two
terms of the standard university calculus sequence, are considered:
1) in chapter 1 - elementary properties of functions, such as elements of defi-
nition, periodicity, even/odd functions, monotonicity and extrema;
2) in chapter 2 - limits, with analysis of limit definition and properties;
3) in chapter 3 - continuity, with an emphasis on global properties;
4) in chapter 4 - differentiability, considering elementary properties, main
theorems and applications;
5) in chapter 5 - different kinds of integrals (indefinite, definite and improper)
with applications;
6) in chapter 6 - infinite sequences and series, including both basic and finer
convergence/divergence tests.
The second part (chapters 7-9) contains the topics of the functions of two
variables, which are quite representative for the most topics of the multi-
variable functions. All the chapters of the second part include a few examples
similar to those presented in the first part in order to show how the ideas ap-
plied in one-dimensional case can be generalized/extended to many variables.
However, the main part of the chapters 7-9 is devoted to examples that high-
light a specificity of concepts and results for multi-variable functions. The
subjects of the second part correspond to the third term of the university
calculus courses:
xix
xx Introduction
1) in chapter 7 - limits and continuity, with details on relation between general,

partial and iterated limits;
2) in chapter 8 - differentiation, including partial derivatives, differentiability,
directional derivative and their applications;
3) in chapter 9 - double, iterated and line integrals with applications.
Each chapter is divided in sections corresponding to the (conditional) divi-
sion of the material in main topics. Examples in each section usually start at
a very basic level and attain certain level of complexity during the developing
of exposition. At the end of each chapter supplementary exercises of different
levels of complexity are provided, the most difficult of them with a hint to the
solution.
The logical sequence of the material just follows the chapter sequence,
which means that the reading of a specific chapter does not require any knowl-
edge of the next chapter subjects. At the same time, different concepts, results
and examples considered and analyzed in earlier chapters/sections are used in
the subsequent sections.
Since the use of counterexamples in this book is aimed to analyze impor-
tant concepts and theorems in calculus and analysis, the selection of the ma-
terial is restricted by those examples that have direct connection with main
concepts. Therefore, more specific and applied problems are left out of the
scope of this manuscript. The interested reader can find these kinds of prob-
lems in different collections of problems on real analysis. Some well-known
books of this type, containing problems of different levels of difficulty, are
indicated in the bibliography: [2], [3], [7], [8], [9], [10].
The following structure is chosen for the presentation of each example:

1) False statement.
2) Solution (counterexample(s)). It provides a complete analytic solution to
the posed problem.
3) Remark(s) (optional). It provides additional explanations, extensions of the
counterexamples, comparisons with other similar situations, and also it gives
links to a general theory and makes comparisons with the correct statements.
4) Figure(s) (optional, for some examples involving more geometric and com-
plex constructions). It provides geometric representation of the analytic solu-
tions and also it serves to develop a geometric intuition.
This form of presentation offers first a prompt for the reader to find out
what is wrong with the proposed statement, substantiating a guess with an-
alytic proof and looking for the corresponding correct results. If the reader
wants to receive hints to the solution or corresponding results of a theory,
he/she may proceed directly to items 2)-4) of each example.
In the choice of figures for illustration of the analytic solutions, we don’t
try to provide as many pictures as possible or to keep an equal distribution
of pictures among chapters, sections, subsections, etc. Instead we follow the
principle of choosing the figures that can be most helpful for understanding
the presented counterexamples (at least from our point of view). In the first
Introduction xxi
place, it is related to the nature of each counterexample (if it is more analytic

or geometric) and to the complexity and singularity of the functions employed
in solution.
All the false statements in this text are consistently placed within quota-
tion marks.
0.1.2 On mathematical language and notation
The language in this manuscript, just as in the majority of mathematical

books, is a mixture of a natural language with mathematical terminology and
symbolism used traditionally for a concise expression of numerous concepts,
results and logical reasoning. Depending on the topic, the use of mathemati-
cal symbols and terms can be more or less concentrated. In order to facilitate
access to these symbols/terms inside the text, the list of notations and sub-
ject index are provided. Nevertheless, we find it useful to give below a few
preliminary explanations of some common symbols used most frequently:
1) The symbol of equality = has multiple (albeit close) meanings: it can mean
equality between two objects (numbers, sets, functions, etc.), for example,
2 · 9 = 3 · 6; it can be used to define a new object by means of other known
ones, for example, a3 = a · a · a; or it can be used for notation of an object,
for instance, R = (the set of all real positive numbers)
2) The symbol → (or ⇒ ) also has multiple meanings: it can mean a logical
implication, as in “the number a is a multiple of 4” → “the number a is a
multiple of 2”; or it can mean a tendency of a variable quantity, for example,
x → a means that the variable x approaches the value a; it is also used to
denote a transformation, for example, f : X → Y indicates that a function f
maps a set X into Y .
3) The symbol {. . .} means collection of elements that form a set, for example,
N = {1, 2, 3, . . .} is the set of natural numbers, and Z = {0, ±1, ±2, . . .} is the
set of integers.
4) The symbol ∈ (∈ / ) denotes that an element belongs (does not belong) to a
set, for example, −1 ∈ Z (−1 ∈ / N). Similarly, the symbol ⊂ (̸⊂ ) denotes that
one set is (is not) subset of another, for example, Z ⊂ R (Z ̸⊂ N).
5) The symbol ∀ means, depending on context, “any”, “every”, “each”, “for
any”, “for all”, for example, ∀x ∈ Z can mean every element (or for all ele-
ments) of the set Z, that is, every (or for all) integer(s).
6) The symbol ∃ (̸ ∃) means existence (non-existence), for example, ∃x ∈ N
such that x is even or ̸ ∃x ∈ N such that x is negative.
Throughout this work we try to follow the standard terminology and no-
tation used in calculus/analysis books. To avoid any misunderstanding and
ambiguity, the main concepts and results, along with the corresponding ter-
minology and notation, are set out in the supplementary parts of the text
containing background material.
xxii Introduction
0.2 Background (elements of theory)
The background material is split in several parts around the manuscript.

The first part, the general background section presented below, contains some
basic concepts and results from the set theory and function theory common
for real-valued functions of one and several real variables and used in different
sections of the manuscript. Other parts appear at the beginning of each chap-
ter of counterexamples and bring the material relevant for the specific topics
treated there. All the concepts and results presented can be found in almost
every calculus and analysis book. In particular, one can consult the classical
books on calculus and analysis indicated in the bibliography list, which treat
the subjects with different levels of strictness, abstraction and generalization:
[1], [4], [6], [11], [12], [13], [14].
0.2.1 Sets
General sets
Set. We consider a set to be a primary non-definable concept, which can

be described as a collection of objects (elements) gathered by using some
criterion. To distinguish between sets and their elements when they appear
together we will use upper case for the former and lower case for the latter. If
element a is in (belongs to) A we write a ∈ A; otherwise we write a ∈ / A.
The set which contains no element is called the empty set (usual notation
∅).
Subset. If every element of a set A is an element of a set B, we say that
A is a subset of B (or A is contained in B) and write A ⊂ B.
If A ⊂ B and B ⊂ A, then the two sets coincide and we write A = B.
Otherwise A ̸= B.
Operations on sets.
The union of the sets A and B (denoted A ∪ B) is defined to be the set C
such that c ∈ C if c ∈ A or c ∈ B.
The intersection of the sets A and B (denoted A ∩ B) is defined to be the
set C such that c ∈ C if c ∈ A and c ∈ B.
The difference of the sets A and B (denoted A\B) is the set of all elements
a ∈ A such that a ∈ / B.
Introduction xxiii
The complement of B ⊂ A with respect to A is the set of all elements

a ∈ A such that a ∈
/ B.
The Cartesian product A × B is the set of all the ordered pairs (a, b) where
a ∈ A and b ∈ B.
Basic properties : A∪B = B∪A, A∩B = B∩A, (A ∪ B)∪C = A∪(B ∪ C),
(A ∩ B) ∩ C = A ∩ (B ∩ C), A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C), A ∩ (B ∪ C) =
(A ∩ B) ∪ (A ∩ C), A ∪ ∅ = A, A ∩ ∅ = ∅.
Numerical sets
Natural numbers. The set N of the natural numbers is the set of all
numbers used for counting: N = {1, 2, 3, ...} .
Integers. The set Z of the integers is the set including all the natural
numbers, zero and the negatives of the natural numbers: Z = {0, ±1, ±2, . . .}.
Rational numbers. The set Q of the rational numbers is composed of all
the fractions pq such that p ∈ Z, q ∈ N.
Decimal fractions. A decimal fraction is the number represented in the
form (±) a0 , a1 a2 a3 . . ., where a0 ∈ N ∪ {0} and the digits ai ∈ N ∪ {0} , 0 ≤
ai ≤ 9, i = 1, 2, 3, . . . . This is the decimal representation of every point on a
coordinate line.
A decimal fraction is called finite if ai = 0 for all i ≥ m, where m is some
natural number. A decimal fraction is called periodic if ak+i = ak+i+p for
some k ∈ N ∪ {0}, p ∈ N, and all i ∈ N, that is, there exists a finite string of
digits (of length p), that is repeated infinitely many times starting from the
(k + 1)-th digit. Otherwise, a decimal fraction is called aperiodic.
Remark 1. A finite fraction can be considered as a special case of a periodic
fraction.
Remark 2. The set of the rational numbers can be equivalently defined as
the set of all decimal periodic fractions.
Irrational numbers. The set I of the irrational numbers is composed of
all decimal aperiodic fractions.
Real numbers. The set R of the real numbers is the set of all decimal
fractions.
Remark. Although I is not a common notation for the set of irrational
numbers, we will use this symbol for brevity of notation (another, more com-
mon option is R\Q).
Basic properties: N ⊂ Z ⊂ Q ⊂ R, I ⊂ R, Q ∪ I = R, Q ∩ I = ∅.
Remark 1. Real numbers are also called real points (or simply points) due
to existence of a one-to-one correspondence between the set R and the set of
all points on a coordinate line called the real axis (see more about one-to-one
correspondence in section 1.2 )
Remark 2. Frequently it is convenient to consider the extended real axis,
including positive (+∞) and negative (−∞) infinities. In this case all real
xxiv Introduction
(geometric) points are called the finite points, and infinities are called the
infinite points.
Remark 3. The Cartesian product R × R = R2 is the set of all the ordered
pairs of real numbers. Due to existence of a one-to-one correspondence between
the set R2 and the set of all points on a plane, with the above pairs being the
Cartesian coordinates of the points, these elements of R2 are also called the
points.
Sets in Rn
The space Rn . The n-dimensional space Rn can be considered as the

set of all the ordered n-tuples (x1 , . . . , xn ), where xi ∈ R, i = 1, . . . , n. It
can be shown that there exists a one-to-one correspondence between Rn and
the set of all the points x with the Cartesian coordinates (x1 , . . . , xn ) in the
“geometric” n-dimensional space. In the case n = 2 and n = 3 it means one-to-
one correspondence between all the ordered pairs and all the points in a plane,
and between all the ordered triples and all the points in (three-dimensional)
space, respectively. For this reason, n-tuples can be conveniently thought of
as the points and the i-th element in n-tuple is called the i-th coordinate of
the point x = (x1 , . . . , xn ).
The space Rn can be considered as the Cartesian product of n one-
dimensional real spaces, R × R × . . . × R = Rn that explains the used notation.
In the case n = 1, the used notation of the space is simply R.

In the case n = 1, n = 2 and n = 3, for the sake of simplicity, the notations
of the coordinates can be changed to (x) (or simply x), (x, y) and (x, y, z),
respectively.
Distance between two points. The distance between two points

√∑ (x1 , . . . , xn ) and y = (y1 , . . . , yn ) is defined by the formula d (x, y) =
x =
n 2
i=1 (xi − yi ) . Another usual notation is d (x, y) = |x − y|. In the case
n = 1, n = 2 and n = 3, it corresponds to the well-known Euclidean distance
between two points on a line d (x, y) = |x − y|, where √the vertical lines are the
2 2
symbols of the absolute value, on a plane d (x, y) = (x1 − y1 ) + (x2 − y2 ) ,
√
2 2 2
and in a space d (x, y) = (x1 − y1 ) + (x2 − y2 ) + (x3 − y3 ) .
The following inequality holds:
√
|xi − yi | ≤ d (x, y) ≤ n max |xi − yi | ,
1≤i≤n
which means, in particular, that the distance between x and y is small if, and
only if, all the coordinates get close.
Sphere. The set of all points equidistant from a given point c is called a
Introduction xxv
sphere , that is, for a given centerpoint c and radius r > 0, the sphere Sc,r is
defined by the analytical relation Sc,r = { x ∈ Rn : d (x, c) = r}.
Open ball. The set of all points located inside the sphere Sc,r is called an
open ball . The usual notation is Bc,r and the analytical description is Bc,r =
{ x ∈ Rn : d (x, c) < r} . In the case n = 1, n = 2 and n = 3, the open ball
2 2
is the open interval (c − r, c + r), the open disk (x1 − c1 ) + (x2 − c2 ) < r2 ,
2
and the open ball (in the common geometrical sense of the word) (x1 − c1 ) +
2 2
(x2 − c2 ) + (x3 − c3 ) < r , respectively.
2
Closed ball. The set of all points located both inside and on the sphere
SC,r is called a closed ball . The notation and analytical description is B̄c,r =
{x ∈ Rn : d (x, c) ≤ r} .
Neighborhood. The open ball Bc,δ , δ > 0 is called a δ-neighborhood
of the point c. Frequently the term is shortened to neighborhood of c. The
number δ is called the radius of the neighborhood.
Deleted neighborhood. The δ-neighborhood with the deleted central
point, that is, the set Bc,δ \ {c}, δ > 0 is called the deleted neighborhood of c.
Limit point. A point c is a limit point of the set S if in any deleted

neighborhood of c there exists at least one point of S. (Of course it implies
that in any neighborhood of c there are infinitely many points of S.) Limit
points are also often referred to as accumulation points.
Isolated point. If c ∈ S and c is not a limit point of S, then c is called
an isolated point of S.
Interior point. A point c is an interior point of S if there is a neighbor-
hood of c contained in S.
Exterior point. A point c is an exterior point of S if there is a neighbor-
hood of c which does not contain any point of S. (Equivalently, an exterior
point of S is an interior point of the complement of S in Rn .)
Boundary point. A point c is a boundary point of S if any neighborhood
of c contains both points of S and out of S. (In other words, c is a boundary
point of S if it is neither an interior nor an exterior point of S.)
Open set. A set S is open if every point of S is its interior point.

Closed set. A set S is closed if every limit point of S is a point of S.
Bounded set. A set S is bounded if there exists a ball that contains S.
Otherwise, a set is unbounded .
Compact set. A set S is called compact if it is closed and bounded.
Closure of set. If L is the set of all limits points of a given set S, then
S̄ = S ∪ L is called the closure of S.
Dense set. A set S ⊂ D is dense in D if S̄ = D.
Separated sets. Two sets A and B are separated if Ā ∩ B = ∅ and
A ∩ B̄ = ∅ .
Disconnected/connected set. A set S is disconnected if it can be repre-
sented as S = A ∪ B, where A and B are nonempty separated sets. Otherwise
a set is called connected .
xxvi Introduction
Curves. The set of points x = x (t), t ∈ [a, b] is called a curve with

endpoints x (a) and x (b). A curve is continuous if x (t) is continuous on [a, b].
A continuous curve is simple if it does not intersect itself anywhere between
its endpoints. A continuous curve is closed if its endpoints coincide.
Linearly connected set. A set S is linearly connected if any two of its
points can be joined by a (continuous) curve which lies in S. If S is linearly
connected then it is connected. For open sets the converse is also true.
Simply connected set. A connected set S is simply connected if every
simple closed curve in S encloses only points that are in S.
General Remark. Any characterization of a set (such as connectedness,

closeness, etc.) is related to the space where this set is considered. For example,
an open disk is an open set in R2 , while it is neither open nor closed set in
R3 .
Special sets in R
Interval. The interval is a set of points x in R, which satisfy one of

the following inequality: a < x < b (open interval ), or a ≤ x ≤ b (closed
interval ), or a ≤ x < b, or a < x ≤ b (the last two intervals are called half-
open or half-closed ). The common notations are (a, b), [a, b], [a, b) and (a, b],
respectively.
Remark. In the case of a strong inequality, the point a can be −∞ and
the point b can be +∞. In the former case, the formal inequality −∞ < x < b
(or −∞ < x ≤ b ) just means x < b (or x ≤ b ), and similar interpretation is
used for the latter case.
Neighborhood. The interval (a − δ, a + δ), δ > 0 is called a δ-
neighborhood of the point a. Frequently the term is shortened to neighborhood
of a. The number δ is called the radius of the neighborhood.
One-sided neighborhood. The right-hand (left-hand) neighborhood of
a is the interval (a, a + δ) ((a − δ, a)).
Deleted neighborhood. The δ-neighborhood with the deleted central
point, that is, the set (a − δ, a) ∪ (a, a + δ), δ > 0 is called the deleted neigh-
borhood of a .
Theorem. Any open set is the union of open intervals.
Bounded set. A set S is bounded above (below) if there exists a real
number M (m) such that a ≤ M (a ≥ m) for all a ∈ S. The number M (m)
is called an upper (lower) bound of S. A set is bounded if it is bounded above
and below. Otherwise, a set is unbounded .
Supremum/infimum. If a set S is bounded above, then there exists
the least upper bound of S, which is called supremum and denoted sup S.
Similarly, if a set S is bounded below, then there exists the greatest lower
bound of S, which is called infimum and denoted inf S.
Bounded interval. An open interval (a, b) is bounded when a and b are
Introduction xxvii
finite points, that is a and b are some points on the real axis. If a or b are
infinite points, the interval is unbounded . A closed interval [a, b] is supposed
to be always bounded.
Remark. Sometimes a bounded interval is called a finite interval and un-
bounded - infinite. Of course, the number of the points in any non-singular
(a < b) interval is always infinite. Usually, the specific meaning of the term
“finite/infinite” as applied to intervals is clear in the used context.
Theorem. Any connected set in R is an interval.
0.2.2 Functions
Concepts
Function, domain, image, pre-image. A correspondence f between

two sets X and Y such that with each element x ∈ X is associated exactly one
element y ∈ Y is called a function from X to Y and is denoted by y = f (x) :
X → Y , or in short notation f (x) or even f . The set X is called the domain
of f (we also say f is defined on X ) and the set Y is called the codomain.
The part of Y that contains all values f (x) for x ∈ X is called the image (or
range) of f and the usual notation is f (X). Respectively, the image of any
set S ⊂ X is denoted by f (S).
The set of all x ∈ X such that f (x) ∈ P , P ⊂ Y is called the inverse
image (or pre-image) of P and the usual notation is f −1 (P ).
Remark 1. Here we consider only such functions that X ⊂ Rn and Y ⊂ R.
Remark 2. Since a general codomain is of a little interest (it can always
be chosen to be R), when it is possible we will consider that in the definition
of a function y = f (x) : X → Y the set Y is the image f (X), that is, the
smallest codomain.
Graph of a function. If f is defined on X, then the graph of f is the

set of the points (x, f (x)) ∈ Rn+1 such that x ∈ X.
Level surface. The surface f (x) = k, where the constant k ∈ f (X), is
the level surface.
Mapping onto. If a function f (x) defined on X has image Y , that is

f (X) = Y , then f (x) maps X onto Y . Another common name is surjective
mapping (or surjective function).
Mapping one-to-one. If for any pair x1 , x2 ∈ X, x1 ̸= x2 it follows
that f (x1 ) ̸= f (x2 ), then f (x) is called a one-to-one mapping on X. Another
common name is injective mapping (or injective function).
One-to-one correspondence. If f (x) is a one-to-one mapping of X
xxviii Introduction
onto Y , then f (x) is said to be a one-to-one correspondence between X and

Y . Another common name is bijective mapping (or bijective function).
Remark. Frequently the term one-to-one function is used for one-to-one
correspondence.
Composition of functions. Let f (x) maps X into Y , and g (y) maps

Y in to Z. Then the composition of f and g is the function h with domain
X and codomain Z defined by the formula h (x) = g (f (x)), ∀x ∈ X. The
standard notation is h (x) = g (f (x)) or h (x) = g ◦ f (x).
Inverse function. Let f (x) be a one-to-one function with domain X
and image Y . In this case, it is possible to define the inverse function f −1 (y),
with domain Y and image X, that assigns to each y ∈ Y the only element
x ∈ X such that f (x) = y.
Restriction of function. If f (x) is defined on X and S ⊂ X, the

restriction of f (x) to S is the function g (x) defined on S such that g (x) =
f (x), ∀x ∈ S.
Extension of function. If f (x) is defined on X and S ⊃ X, the extension
of f (x) onto S is the function g (x) defined on S such that g (x) = f (x),
∀x ∈ X.
Equivalent sets. Two sets are called equivalent if there exists a one-to-
one correspondence between them.
Finite/infinite sets. A set is called finite if it has a finite number of
elements. Otherwise a set is infinite. The empty set is considered to be finite.
Countable/uncountable set. A set is countable if it is equivalent to N.
A set is uncountable if it is neither finite nor countable.
Elementary properties
Bounded function. A function f (x) is bounded above (below) on a set

S if there exists a real number M (m) such that f (x) ≤ M (f (x) ≥ m) for
all x ∈ S. A function is bounded if it is bounded above and below. Otherwise,
a function is unbounded.
Global maximum. A point x0 is a global maximum (strict global max-

imum) point of f (x) on a set S if for any x ∈ S, x ̸= x0 it follows that
f (x) ≤ f (x0 ) (f (x) < f (x0 )).
Global minimum. A point x0 is a global minimum (strict global min-
f (x) ≥ f (x0 ) (f (x) > f (x0 )).
Global extremum. A point x0 is a global extremum (strict global ex-
tremum) if it is global maximum or global minimum (strict global maximum
or strict global minimum).
Local maximum. A point x0 is a local maximum (strict local maximum)
Introduction xxix
point of f (x) if there exists a neighborhood of x0 such that for any x ̸= x0 ,

which belongs to this neighborhood and to the function domain, it follows
that f (x) ≤ f (x0 ) (f (x) < f (x0 )).
Local minimum. A point x0 is a local minimum (strict local minimum)
that f (x) ≥ f (x0 ) (f (x) > f (x0 )).
Local extremum. A point x0 is a local extremum (strict local extremum)
if it is local maximum or local minimum (strict local maximum or strict local
minimum).
Remark 1. There is an ambiguity in the terminology of extremum points,
because the extremum point x0 , the extremum value f (x0 ) and the extremum
pair (x0 , f (x0 )) all of them can be called the extremum point. However, usu-
ally it does not generate a misunderstanding regarding the use of this term
within a specific context.
Remark 2. Sometimes a global extremum (minimum/maximum) is also
called an absolute extremum (minimum/maximum), and a local extremum
(minimum/maximum) - a relative extremum (minimum/maximum).
Part I
Functions of one real

variable
1
Chapter 1
Elementary properties of functions
1.1 Elements of theory
Concepts
Function, domain, image, pre-image. A correspondence f between

two sets X and Y such that with each element x ∈ X is associated exactly one
element y ∈ Y is called a function from X to Y and is denoted by y = f (x) :
X → Y , or in short notation f (x) or even f . The set X is called the domain
of f (we also say f is defined on X) and the set Y is called the codomain.
The part of Y that contains all values f (x) for x ∈ X is called the image (or
range) of f and the usual notation is f (X). Respectively, the image of any
set S ⊂ X is denoted by f (S).
The set of all x ∈ X such that f (x) ∈ P , P ⊂ Y is called the inverse
image (or pre-image) of P and the usual notation is f −1 (P ).
Remark. In chapters 1-6 we consider only such functions that X ⊂ R and
Y ⊂ R.
Remark. Since a general codomain is of a little interest (it can always be
chosen to be R), when it is possible we will consider that in the definition of a
function y = f (x) : X → Y the set Y is the image f (X), that is, the smallest
codomain.
Graph of a function. If f is defined on X, then the graph of f is the

set of all the ordered pairs (x, f (x)) ∈ R2 such that x ∈ X.
Mapping onto. If a function f (x) defined on X has image Y , that is
f (X) = Y , then f (x) maps X onto Y . Another common name is surjective
mapping (or surjective function).
Mapping one-to-one. If for any pair x1 , x2 ∈ X, x1 ̸= x2 it follows
that f (x1 ) ̸= f (x2 ), then f (x) is called a one-to-one mapping on X. Another
common name is injective mapping (or injective function).
One-to-one correspondence. If f (x) is a one-to-one mapping of X
3
4 Counterexamples: From Calculus to the Beginnings of Analysis
onto Y , then f (x) is said to be a one-to-one correspondence between X and

Y . Another common name is bijective mapping (or bijective function).
Remark. Frequently the term one-to-one function is used for one-to-one
correspondence.
Composition of functions. Let f (x) maps X into Y , and g (y) maps Y

into Z. Then the composition of f and g is the function h with domain X and
codomain Z defined by the formula h (x) = g (f (x)), ∀x ∈ X. The standard
notation is h (x) = g (f (x)) or h (x) = g ◦ f (x).
Inverse function. Let f (x) be a one-to-one function with domain X
and image Y . In this case, it is possible to define the inverse function f −1 (y),
with domain Y and image X, that assigns to each y ∈ Y the only element
x ∈ X such that f (x) = y.
Equivalent sets. Two sets are called equivalent if there exists a one-to-
one correspondence between them.
Finite/infinite sets. A set is called finite if it has a finite number of
elements. Otherwise a set is infinite. The empty set is considered to be finite.
Countable/uncountable set. A set is countable if it is equivalent to N.
A set is uncountable if it is neither finite nor countable .
Bounded function. A function f (x) is bounded above (below) on a set

S if there exists a real number M (m) such that f (x) ≤ M (f (x) ≥ m) for
all x ∈ S. A function is bounded if it is bounded above and below. Otherwise,
a function is unbounded.
Even function. A function f (x) defined on X is called even if for any
x ∈ X it holds that f (−x) = f (x).
Odd function. A function f (x) defined on X is called odd if for any
x ∈ X it holds that f (−x) = −f (x).
Periodic function. A function f (x) defined on X is called periodic with
period T ̸= 0 if for any x ∈ X it holds that f (x + T ) = f (x). The smallest
positive number T (if it exists) for which this property holds is called the
fundamental period.
The primarily (geometric) properties of even functions (follow directly from
the definition):
1) the domain of an even function is symmetric with respect to the origin;
2) the graph of an even function is symmetric with respect to the x-axis.
The primarily properties of odd functions (follow directly from the defini-
tion):
1) the domain of an odd function is symmetric with respect to the origin;
2) the graph of an odd function is symmetric with respect to the origin.
Elementary properties of functions 5
The primarily properties of periodic functions (follow directly from the

definition):
1) the domain of a periodic function is not bounded either at the left or at
the right;
2) the graph of a periodic function can be obtained by an infinite many trans-
lations of its part defined on an interval of length T .
Increasing function. A function f (x) is called increasing (strictly in-

creasing) on a set S if for any x1 , x2 ∈ S, x1 < x2 it follows that f (x1 ) ≤ f (x2 )
(f (x1 ) < f (x2 ) ).
Decreasing function. A function f (x) is called decreasing (strictly de-
creasing) on a set S if for any x1 , x2 ∈ S, x1 < x2 it follows that f (x1 ) ≥ f (x2 )
(f (x1 ) > f (x2 ) ).
Monotone function. A function f (x) is called monotone (strictly mono-
tone) on a set S if it is increasing or decreasing (strictly increasing or strictly
decreasing) on this set.
Global maximum. A point x0 is a global maximum (strict global max-
f (x) ≤ f (x0 ) (f (x) < f (x0 )).
Global minimum. A point x0 is a global minimum (strict global min-
f (x) ≥ f (x0 ) (f (x) > f (x0 )).
Global extremum. A point x0 is a global extremum (strict global ex-
tremum) if it is global maximum or global minimum (strict global maximum
or strict global minimum).
Local maximum. A point x0 is a local maximum (strict local maximum)
that f (x) ≤ f (x0 ) (f (x) < f (x0 )).
Local minimum. A point x0 is a local minimum (strict local minimum)
that f (x) ≥ f (x0 ) (f (x) > f (x0 )).
Local extremum. A point x0 is a local extremum (strict local extremum)
if it is local maximum or local minimum (strict local maximum or strict local
minimum).
Remark. There is an ambiguity in the terminology of extremum points,
because the extremum point x0 , the extremum value f (x0 ) and the extremum
pair (x0 , f (x0 )) all of them can be called the extremum point. However, usu-
ally it does not generate a misunderstanding regarding the use of this term
within a specific context.
1.2 Function definition
Example 1. “If A and B are two subsets of the domain X of f (x), then
f (A ∩ B) = f (A) ∩ f (B).”
Solution.
Consider the function f (x) = sin x and choose A = [0, 2π] e B = [π, 3π].
Then f (A ∩ B) = f ([π, 2π]) = [−1, 0], while f (A)∩f (B) = [−1, 1]∩[−1, 1] =
[−1, 1].
Remark 1. The following relation is true: f (A ∩ B) ⊂ f (A) ∩ f (B).
Remark 2. For the union the statement is true, that is, f (A ∪ B) =
f (A) ∪ f (B).
FIGURE 1.2.1: Example 1
Example 2. “If A is a subset of the domain X of f (x), then f (X\A) =

f (X) \f (A).”
Solution.
Consider the function f (x) = x2 and choose A = [0, +∞). Then f (R\A) =
f ((−∞, 0)) = (0, +∞), while f (R) \f (A) = [0, +∞) \ [0, +∞) = ∅.
Remark 1. The following relation is true: f (X\A) ⊃ f (X) \f (A).
Remark 2. The statement is true for one-to-one functions.
Remark to Examples 1 and 2. The first two examples of this section show
that some properties of domains cannot be extended to their images.
Example 3. “If f (x) maps X onto X, then f (x) is one-to-one on X.”

Solution.
Consider the function f (x) = cos 2πx and choose X = [−1, 1]. Then
f (X) = [−1, 1] = X, but f (x) is not one-to-one (it gets the same values
in the points x and −x for ∀x ∈ [0, 1], since f (x) is even).
Example 4. “If f (x) maps X onto X and is one-to-one, then f (x) = x

or f (x) = −x.”
Solution.
The function f (x) = x3 transforms X = [−1, 1] onto [−1, 1] and is one-to-
one correspondence.
Example 5. “If f (x) maps X onto Y ⊂ X, Y ̸= X, then there is no

inverse function to f (x) on Y .”
Solution. [ ] [ ]
The function f (x) = x3 transforms X = − 12 , 12 onto Y = − 18 , 18 ⊂ X,
√
but the inverse function f −1 (y) = 3 y is well defined on Y .
Remark. This example shows that the existence of an inverse does not
related to the “sizes” of the domain and image.
Example 6. “For any non-constant function defined on R there exists an

interval where the function admits the inverse.”
Solution. {
1, x ∈ Q
Dirichlet’s function D (x) = is defined on R, but it has
0, x ∈ I
repeated values in any interval.
Remark. It is seen that the existence of an inverse is not guaranteed even
on a small (but still representative) part of domain.
Example 7. “If f (x) is injective on A and B, then f (x) is injective on

A ∪ B.”
Solution.
Let f (x) = x on A = [0, 1] and f (x) = x − 2 on B = [2, 3]. Then f (x)
is injective separately on A and B, but it is not injective on A ∪ B, because
f (A) = f (B).
Another kind of counterexample is f (x) = x − [x] on A = [0, 1) and
B = [1, 2). In this case A and B are not separated and A ∪ B = [0, 2) is
connected set. However, the result is the same: f (A) = f (B).
One more example is f (x) = x2 on A = [−1, 0] and B = [0, 1]. In this case
neither the union A ∪ B nor the function f (x) have “jumps”: more strictly,
the set A ∪ B is connected and f (x) is continuous (continuous functions will
be considered in chapter 3). However, again f (A) = f (B), which means that
f (x) is not injective on A ∪ B.
Remark 1. The following strengthened version of the statement is still
false: “if f (x) is injective on a neighborhood of each point of its domain
X, then f (x) is injective on X”. The corresponding counterexample can be
f (x) = |x| or f (x) = x2 considered on X = R\ {0}.
Remark 2. The statement in Remark 1 (but not the original one) becomes
true if f (x) is continuous on a connected domain X (continuous functions will
be considered in chapter 3).
Remark 3. Similar statement for the intersection of sets is true: if f (x) is
injective on A and B (and A ∩ B ̸= ∅), then f (x) is injective on A ∩ B.
Example 8. “If f (x) is surjective on a pair A → C and on B → D, then

f (x) is surjective on the pair A ∩ B → C ∩ D.”
Solution.
Let f (x) = |x|: A = [0, 1] → C = [0, 1] and B = [−1, 0] → D = [0, 1].
Evidently, the conditions of the statement hold, but not the conclusion: f (x) =
|x|: A ∩ B = {0} → C ∩ D = [0, 1] is not surjective.
Remark. Similar statement for the union of sets is true: if f (x) is surjective
on a pair A → C and on B → D, then f (x) is surjective on the pair A ∪ B →
C ∪ D.
Remark to Examples 7 and 8. These two examples show that the injective
property does not extend on the union of domains (although it does extend
on the intersection) and the surjective property does not extend on the inter-
section of domains (while it does extend to the union).
Example 9. “If f (x) : X → Y is surjective and g (y) : Y → Z is injective,

then g (f (x)) : X → Z is surjective.”
Solution.
Let f (x) = |x|: X = [−1, 1] → Y = [0, 1], and g (y) = y : Y = [0, 1] → Z =
[−1, 1]. Evidently, the conditions of the statement hold, but the composition
g (f (x)) = |x| : X = [−1, 1] → Z = [−1, 1] is not surjective, because there is
no element in X = [−1, 1] that corresponds to any negative number in Z.
Remark 1. The same counterexample can be used for the false statement:
“if f (x) : X → Y is surjective and g (y) : Y → Z is injective, then g (f (x)) :
X → Z is injective”.
Remark 2. The change of the order of the functions does not change
the situation, more precisely, the following statement is also false: “if f (x) :
X → Y is injective and g (y) : Y → Z is surjective, then g (f (x)) : X → Z
is injective or surjective”. Consider the injective function f (x) = x : X =
[−1, 1] → Y = [−2, 2], and the surjective g (y) = y 2 : Y = [−2, 2] → Z =
[0, 4]. Their composition g (f (x)) = x2 : X = [−1, 1] → Z = [0, 4] is not
injective neither surjective, because g (f ([−1, 0])) = g (f ([0, 1])) = [0, 1].
Remark 3. If both f (x) and g(x) are injective then their composition is
also injective; and if f (x) and g(x) are surjective then their composition is
also surjective.
Example 10. “If g (f (x)) : X → Z is bijective composition of f (x) :

X → Y and g (y) : Y → Z, then both f (x) and g (y) are bijective.”
Solution.
Let f (x) = x: X = {0} → Y = R, and g (y) = 0 : Y = R → Z =
{0}. With this choice both functions are not bijective: f (x) is not surjective
(although it is injective) and g (y) is not injective (although it is surjective).
However, their composition is a bijective function: g (f (x)) = 0 : X = {0} →
Z = {0}. [ ]
Another simple counterexample is non-surjective f (x) = x: X = 0, π2 →
Y = [0, π], followed by non-injective g (y) = sin y : Y = [0, π] → Z = [0, 1].
However,
[ π] their composition is a bijective function: g (f (x)) = sin x : X =
0, 2 → Z = [0, 1].
Remark. If the composition g (f (x)) is bijective, it implies that f (x) is
injective and g (y) is surjective.
1.3 Boundedness
Example 1. “The sum of two unbounded functions defined on the same

domain is also an unbounded function on this domain.”
Solution.
Let f (x) = x and g (x) = −x. Both functions defined and unbounded on
R, but h (x) = f (x) + g (x) = 0 is a bounded function on R.
Remark 1. If one is interested in a compact domain (such as a closed
interval
{ [a, b]),( then a) similar example with the functions f (x) = −g (x) =
tan x , x ∈ − π2 , π2 [ ]
defined on − π2 , π2 works.
0, x = −2, 2
π π
Remark 2. The property of boundedness is maintained by the sum, differ-

ence and product of two bounded functions (at least for the functions defined
on the same domain). It is not so for unbounded functions.
Example 2. “A bounded function always attains its global extremum.”

Solution.
The function f (x) = x considered on (0, 1) is bounded (for example, with
lower bound 0 and upper bound 1), but it does not achieve either its global
minimum or global maximum. Indeed, since f (x) assumes positive values,

its global maximum, if it existed, would be positive. Since all the values of
f (x) are less than 1, any number in [1, +∞) cannot be its maximum. On
the other hand, for any x0 ∈ (0, 1) the value of the function f (x1 ) = x1 at
the point x1 = x02+1 ∈ (0, 1) is larger than f (x0 ) = x0 , implying that any
x0 ∈ (0, 1) cannot be a maximum point. Since we exhausted all options for
a global maximum, it remains to conclude that f (x) = x does not attain
its maximum on (0, 1). A similar reasoning leads to the conclusion about the
absence of a global minimum.
Remark 1. Analogous result is true for a compact domain. For instance,
in the case of the closed
{ interval [0, 1], a slight modification of the above coun-
x , x ∈ (0, 1)
terexample f (x) = 1 shows a bounded function without
2 , x = 0, x = 1
global extrema.
Remark 2. In both examples (for open and closed intervals) the chosen
bounded functions have no local extremum also.
Example 3. “If a function is bounded in some neighborhood of each point

of (a, b), then it is bounded on (a, b).”
Solution. ( ) ( )
Consider f (x) = tan x on − π2 , π2 . For each point x0 ∈ − π2 , π2 there
is a neighborhood where f (x) is bounded: for example, in the neighborhood
(x1 , x2 ), where x1 = x0 −π/2
2 , x2 = x0 +π/2
2 all the values of f ((x) lie inside
)
the interval (f (x1 ) , f (x2 )), due to strict increasing of tan x on( − π2 , )π2 , i.e.
f (x) is bounded on (x1 , x2 ). However, tan x is unbounded on − π2 , π2 .
Remark 1. If an open interval (a, b) is substituted by closed [a, b], then
the statement will be true.
Remark 2. The boundedness is a “global” property, that is, the property
defined on a given set, unlike a “local” property defined in a neighborhood of
each point. For “local” properties their extension to the entire set is frequently
defined as a fulfillment of “local” property at each point of a set. It is not so
for boundedness: local boundedness does not imply global boundedness.
Example 4. “If a function is defined on R , then it is locally bounded at

least in one point.”
Solution.
Let us suppose below that any rational number is represented in the form
x= m m
n , where m is an integer, n is a natural number, and the fraction n is
in lowest terms (the last means that m and n have no common factor). In
this case the numbers
{ m and n are uniquely determined, and the following
n , x ∈ Q, x = mn
function f (x) = is well-defined. It can be shown that
0 , x ∈ I and x = 0
this function is unbounded in any neighborhood of an arbitrary real point
x0 . In effect, if f (x) was bounded in some neighborhood of x0 , then for all
rational points x = m n in this neighborhood the denominators n would be
bounded, and hence the numerators m would be bounded too. But this would
permit only finitely many rational numbers in such neighborhood. It leads
to a contradiction with the known property that in any neighborhood of an
arbitrary real point there exist infinitely many rational numbers.
1.4 Periodicity
Example 1. “If a non-constant function f (x) is periodic, then it has a

fundamental period.”
Solution. {
1, x ∈ Q
Consider Dirichlet’s function D (x) = . Any rational number
0, x ∈ I
is a period of this function (due to the properties of the real numbers, the sum
of two rational numbers is again rational, and the sum of a rational number
with an irrational one is always irrational).
Remark 1. The statement will turn true if additionally f (x) is continuous
in at least one point (but proof is not simple and it involves the concept of
continuity, which will be considered later).
Remark 2. D (x) is an example of a function for which any rational number

is a period and any irrational number is not (the latter is because D (0) = 1 ̸=
0 = D (x) for any irrational number x). Contrary to this situation, there is no
function for which any irrational number is a period and any rational number
is not (supposing, per absurd, √ that there exists
√ such a function, we pick up
two irrational periods T1 = 2 and T2 = 1 − 2 and immediately arrive to
contradiction, because their sum T1 + T2 = 1 should be a period of the same
function, but it is a rational number).
Example 2. “If both functions f (x) and g (x) have a fundamental period
T, then f (x) + g (x) has the same fundamental period.”
Solution.
If f (x) = sin x and g (x) = 1 − sin x (both functions have the fundamental
period T = 2π), then h (x) = f (x) + g (x) = 1 and it does not have a
fundamental period (although T = 2π is one of the periods of h (x)).
Remark 1. Another simple example when a fundamental period of the
sum exists, but is smaller than T : f (x) = sin x and g (x) = sin 2x − sin x have
the fundamental period T = 2π, but h (x) = f (x) + g (x) = sin 2x has the
fundamental period π.
Remark 2. Similar examples can be constructed for other arithmetic oper-
ations. For example, the false statement for the product has the form: “if both
functions f (x) and g (x) have a fundamental period T , then f (x)·g (x) has the
same fundamental period”. The corresponding example can be f (x) = sin x,
g (x) = 2 cos x (both functions have the fundamental period T = 2π), and
h (x) = f (x) · g (x) = sin 2x has the fundamental period π.
Remark 3. The correct statement asserts that if f (x) and g (x) have a
fundamental period T , then f (x) + g (x) has the same period T . It just may
be a non-fundamental period for the sum.
Example 3. “If both functions f (x) and g (x) are periodic, then f (x) +
g (x) is also periodic.”
Solution.
Consider f (x) = sin x with the fundamental period Tf = 2π and g (x) =
sin πx with the fundamental period Tg = 2. Let us show that h (x) = f (x) +
g (x) = sin x + sin πx has no period. In fact, if there existed a period T of the
function h (x), then the following would hold for an arbitrary x:
h (x + T ) = sin (x + T ) + sin π (x + T ) = sin x + sin πx = h (x) ,
or
sin (x + T ) − sin x = − sin π (x + T ) + sin πx,
or ( ) ( )
T T T T
sin cos x + = − sin π cos π x + .
2 2 2 2
Now, we specify the two values of x in the last equation. First, let us set
x1 = π2 − T2 . Then the last equation simplifies to sin π T2 cos π π2 = 0. Since the

second multiplier is not zero (cos x = 0 only at the points xk = π2 +kπ, k ∈ Z),
we have sin π T2 = 0, which gives the solutions Tn = 2n, n ∈ Z\ {0} (sin x = 0
only at the points xn = nπ, n ∈ Z, and zero is eliminated by the definition of
a period).
On the other hand, we can set x2 = 12 − T2 . Then the same equation
simplifies to sin T2 cos 12 = 0, imposing the condition sin T2 = 0, that is, Tm =
2mπ, m ∈ Z\ {0}. Since all the numbers Tn = 2n are the integers and Tm =
2mπ are irrational, there is no common number between two sets of solutions,
i.e., it does not exist a non-zero number that satisfies the considered equation
for a period T both for x1 and x2 . It means that there is no period for the
function h (x).
Remark 1. Another interesting example can be provided with the functions
f (x) = cos x and g (x) = x − [x] (the last has the fundamental period Tg = 1).
In this case, again supposing, per absurd, that h (x) = f (x) + g (x) = cos x +
x − [x] has a period T , we obtain the relevant equation of periodicity in the
form
cos (x + T ) + x + T − [x + T ] = cos x + x − [x]
or ( )
T T
−2 sin sin x + = −T + [x + T ] − [x] .
2 2
[ ] [ ]
Setting first x1 = − T2 , we obtain −T + T2 − − T2 = 0, which implies that T
is an integer, because if it is not so then we readily

[ ] arrive
[ ] to a contradiction
[ ] by
rewriting the last equation in the form, −T + T2 + T2 +1 = 0, or T2 = T −1 2 .
On the other hand, by setting x2 = 0, we obtain −2 sin2 T2 = −T +[T ]. Since T
should be an integer, the right-hand side is zero and, consequently, sin T2 = 0,
which is not possible for integers. Therefore, the function h (x) does not possess
a period.
Remark 2. For the statement to be true, it is necessary to add the condition
T
that the periods of f (x) and g (x) are commensurate (it means that Tfg ∈ Q).
Remark 3. The sum of two periodic functions with non-commensurate
periods still can be a periodic function, but such examples are more complex.
Remark 4. For continuous functions, the condition that periods of two
functions are commensurate is necessary and sufficient for the sum to be a
periodic function also.
Remark 5. Similar examples can be constructed for other arithmetic oper-
ations. For instance, for the false statement: “if both functions f (x) and g (x)
are periodic, then f (x) · g (x) is also periodic”, the following counterexample
works:f (x) = sin x, g (x) = sin 2π x.
Example 4. “A nonconstant function cannot possess both rational and

irrational periods.”
Solution. √
Let us construct a function that has two periods T1 = 1 and T2 = 2. First,
{ √ }
we define the following set of real numbers S = x = p + 2q, p, q ∈ Q .
Then we define the characteristic function of S, that is,
{
1, x ∈ S
fS (x) = .
0, x ∈/S
It is easy
( to ) that T1 and T2 are periods of fS (x): if x ∈ S then √
√ see (x + 1) ∈ S
and x + 2 ∈ S, and therefore fS (x) = fS (x + 1) = fS (x + 2) = 1; if
( √ )
x∈ / S and x + 2 ∈
/ S then (x + 1) ∈ / S, and therefore fS (x) = fS (x + 1) =
√
fS (x + 2) = 0.
Remark 1. This statement becomes true if the condition of the continuity
of function is added.
Remark 2. It is worth to notice that in the provided
√ counterexample any
rational number and also any number of the form 2q, q ∈ Q are periods of
the presented function.
Example 5. “If the sum of two functions is periodic, then each of the
functions is also periodic.”
Solution.
The functions f (x) = x and g (x) = − [x] are not periodic, but their sum
f (x) + g (x) = x − [x] is periodic with the fundamental period 1.
Remark. Similar examples can be constructed ( for other
) arithmetic op-
erations. For example, the functions f (x) = x2 + cos x cos x and g (x) =
x2 + cos x are not periodic, but their ratio fg(x)
(x)
= cos x is periodic with the
fundamental period 2π (note that x + cos x ̸= 0 for any x ∈ R, so the ratio
2
is defined on R).
Example 6. “The composition of two non-periodic functions is again a

non-periodic function.”
Solution.
√
The functions f (x) = x2 : R → [0, +∞) and g (y) = cos y : [0, +∞) →
[−1, 1] are not periodic: the former is strictly increasing on the infinite interval
[0, +∞), so it cannot repeat its values there, and the latter has the domain
[0, +∞) bounded at the left, which cannot happen with a periodic function.
Nevertheless, the composition g (f (x)) = cos |x| = cos x : R → [−1, 1] is a
periodic function.
Remark 1. Of course, a composition can be non-periodic also. For instance,
just use another order for composition √ of2 the above functions to obtain the
non-periodic function f (g (x)) = (cos x) : [0, +∞) → [0, { 1].
−1, x < 0
Remark 2. Another interesting example is f (x) = and
1, x ≥ 0
g (y) = |y|. In this case, both compositions give the constant function:
g (f (x)) = f (g (x)) = 1.
Remark 3. For composition of a non-periodic and periodic functions the
result depends on the order of composition. For instance, if f (x) is periodic
and g (y) is not, then the function g (f (x)) is always periodic having the same
period as that of f (x) (of course, we consider here only the cases when the
composition is defined), but the function f (g (x)) can be periodic or not,
depending on a specific choice of √ functions. If one chooses f (x) = sin x and
√
g (y) = y, then f (g (x)) = sin x is a non-periodic function, but for f (x) =
sin x and g (y) = y the composition f (g (x)) = sin x is periodic.
1.5 Even/odd functions
Example 1. “The sum of an even function and an odd function is odd.”

Solution.
Consider f (x) = x {(odd function) and g (x) = |x| (even function). Then
0, x < 0
h (x) = f (x) + g (x) = is neither even nor odd.
2x, x ≥ 0
Remark. A similar simple example can be given for the difference. Actually
the sum of an even function and an odd function is always neither even nor
odd, unless one of these functions is zero (zero is the only function which is
both even and odd). The product or ratio of an even function and an odd
function is always an odd function (when it is defined).
Example 2. “If the sum of two functions is even, then each of these two
functions is also even.”
Solution.
Let f (x) = x2 + x and g (x) = x2 − x. Both functions are neither even nor
odd, but h (x) = f (x) + g (x) = 2x2 is an even function.
Remark 1. Similar simple examples can be given for the difference, product
and ratio. For example, the product of the functions f (x) = x − 1 and g (x) =
x + 1, which are neither even nor odd, gives the even function h (x) = f (x) ·
g (x) = x2 − 1.
Remark 2. The converse is correct: the sum, difference, product and ratio
of two even functions is again an even function (when it is defined).
Example 3. “If the sum of two functions is odd, then each of these two
functions is also odd.”
Solution. { {
x, x < 1 x, x < 1
Let f (x) = and g (x) = . Both functions are
−x, x ≥ 1 3x, x ≥ 1
neither even nor odd, but h (x) = f (x) + g (x) = 2x is an odd function.
Remark 1. A similar simple example can be given for the difference.
Remark 2. The converse is correct: the sum and difference of two odd
functions is again an odd function (when it is defined).
Remark 3. For the product and ratio, if each of two functions is odd then
the result (when defined) is an even function. Therefore, it is hard even to
suppose that the product (or the ratio) of two odd functions will give another
odd function.
Example 4. “If the product of two functions is odd, then one of these
functions is even and another is odd.”
Solution. { {
x, x < 1 1, x < 1
Let f (x) = and g (x) = . Both functions are
−x, x ≥ 1 −1, x ≥ 1
neither even nor odd, but h (x) = f (x) · g (x) = x is an odd function.
Remark 1. A similar simple example can be given for the ratio.
Remark 2. The converse is correct: the product and ratio of an even
function and an odd function is an odd function (when it is defined).
Example 5. “If the absolute value of a function is even, then the function
itself is also even.”
Solution.
If f (x) = x (odd function), then |f (x)| = |x| is even function.
Remark
{ 1. Of course, the original function can be neither even nor odd:
x, x < 1
f (x) = gives the same even function |f (x)| = |x|.
−x, x ≥ 1
Remark 2. The same counterexamples can be used to the following false
statement: “if the square of a function is even, then the function itself is also
even”.
Remark 3. The converse is true: if a function is even, then its absolute
value and its square are also even functions.
1.6 Monotonicity
Example 1. “If both functions f (x) and g (x) are increasing (decreasing)
on the same domain, then f (x) − g (x) is also increasing (decreasing) on this
domain.”
Solution.
Let f (x) = x and g (x) = x2 with both functions defined and strictly
increasing on [0, +∞). However, h (x) = f (x)−g (x) = x−x2 is not monotone
on [0, +∞), as it is seen from the following evaluation:
( )
h (x2 ) − h (x1 ) = x2 − x1 − x22 − x21 = (x2 − x1 ) (1 − x2 − x1 ) ,
and notice that for x1 < x2 the first parenthesis is positive, but the second
changes its sign on [0, +∞) (for 0 ≤ x1 < x2 ≤ 12 the second expression is
positive, but for 12 ≤ x1 < x2 it is negative).
For the functions defined on R, one can use the counterexample with
f (x) = x and g (x) = x3 . Evidently, both functions are strictly increasing
on R. However, h (x) = f (x) − g (x) is not monotone on R, as it is seen from
the following evaluation:
( ) ( )
h (x2 ) − h (x1 ) = x2 − x1 − x32 − x31 = (x2 − x1 ) 1 − x22 − x2 x1 − x21 ,
and notice that for x1 < x2 the first parenthesis is positive, but the second
changes its sign on R (for − 21 < x1 < x2 < 12 the second expression is positive,
but for 1 < x1 < x2 it is negative).
Remark. A similar statement for the sum of two increasing (decreasing)
functions is true.
Example 2. “If both functions f (x) and g (x) are increasing (decreasing)
on the same domain, then f (x) · g (x) is also increasing (decreasing) on this
domain.”
Solution.
Let f (x) = x1 and g (x) = x1 be considered on (−∞, 0). Evidently, both
function are strictly decreasing on (−∞, 0). However, h (x) = f (x)·g (x) = x12
is strictly increasing on (−∞, 0).
Another pair of functions can produce non-monotone product. For in-
stance, f (x) = x and g (x) = x are defined and strictly increasing on R.
However, h (x) = f (x) · g (x) = x2 is not monotone function on R.
Remark. In the particular case when f (x) = g (x) the above statement
is transformed as follows: “if a function f (x) is increasing (decreasing) on its
domain, then f 2 (x) is also increasing (decreasing) on this domain”. The same
function f (x) = x disproves the last statement.
Example 3. “If both functions f (x) and g (x) are monotone on the same
domain, then f (x) + g (x) is also monotone on this domain.”
Solution.
Let f (x) = x + sin x and g (x) = −x (both defined on R). Evidently, g (x)
is strictly decreasing on R, and f (x) is strictly increasing on R, because for
an arbitrary pair x1 , x2 ∈ R, x1 < x2 , we have
f (x2 ) − f (x1 ) = x2 − x1 + sin x2 − sin x1

x2 − x1 x2 + x1 x2 − x1
= x2 − x1 + 2 sin cos > x2 − x1 − 2 =0
2 2 2
(here, we used two elementary properties of trigonometric functions: sin α < α,
∀α > 0 and |cos β| ≤ 1, ∀β). At the same time, the function h (x) = f (x) +
g (x) = sin x is not monotone on R (and on any interval of the length greater
than π).
Example 4. “If f (x) is not monotone and g (x) is monotone on the same
domain, then f (x) + g (x) is monotone on this domain.”
Solution.
Let f (x) = |x| and g (x) = x2 be considered on R. Evidently, both functions
satisfy
{ x the above conditions, but the resulting function h (x) = f (x) + g (x) =
−2 , x < 0
is not monotone on R.
2 , x≥0
3x
Remark. The statement with the opposite conclusion is also false. The
corresponding counterexample can be constructed with the functions f (x) =
|x| and g (x) = 2x defined on R. Although f (x) is not monotone { and g (x) is
x, x < 0
monotone on R, the resulting function h (x) = f (x) + g (x) =
3x , x ≥ 0
is monotone on R.
Example 5. “If both functions f (x) and g (x) are non-monotone on the
same domain, then f (x) + g (x) is also non-monotone on this domain.”
Solution. {
3x , x < 0
Let f (x) = |2x| and g (x) = x − |2x| = be defined on R.
−x , x ≥ 0
Evidently, both function are non-monotone on R, but the resulting function

h (x) = f (x) + g (x) = x is strictly increasing on R.
Remark to Examples 3-5. All the statements presented above for the sum
can also be used for the difference, product and ratio of two functions. For
instance, the false statement for the product, similar to Example 5, can be
formulated as follows: “if both functions f (x) and g (x) are non-monotone
on the same domain, then f (x) · g (x) is also non-monotone on this domain”.
The counterexample
{ 1 can be constructed using the functions f (x) = x2 and
g (x) = x , x ̸= 0 , both defined and non-monotone on R. Their product
0, x = 0
h (x) = f (x) · g (x) = x is strictly increasing on R.
Example 6. “If f (x) is monotone on R and f (x) ̸= 0 , ∀x ∈ R, then f (x) 1
is monotone on R.”
Solution. {
x − 1, x ≤ 0
Let f (x) = . Evidently, this is a strictly increasing func-
x + 1, x > 0
{ 1
tion on R. However, the function g (x) = f (x) =
1 x−1 , x ≤ 0 is not mono-
1
x+1 , x > 0
tone on R. Indeed, this function is strictly decreasing on separate intervals
(−∞, 0] and (0, +∞), since for any x1 < x2 ≤ 0 it follows that
1 1 x1 − x2
g (x2 ) − g (x1 ) = − = <0
x2 − 1 x1 − 1 (x2 − 1) (x1 − 1)
(the numerator is negative and the denominator is positive) and for any x2 >
x1 > 0 one obtains
1 1 x1 − x2
g (x2 ) − g (x1 ) = − = <0
x2 + 1 x1 + 1 (x2 + 1) (x1 + 1)
(again the numerator is negative and the denominator is positive). However,
g (x) is not decreasing on R: if one chooses x1 = −1 and x2 = 1, then g (x2 ) −
1
g (x1 ) = 1+1 − −1−1
1
= 1 > 0.
Example 7. “If a function f (x) is not monotone on (a, b), then its square
also is not monotone on (a, b).”
Solution.
{ Let f (x) be defined on [0, 2] in the following way: f (x) =
−x , x ∈ [0, 1]
. Evidently, f (x) is not monotone on [0, 2]. However,
x , x ∈ (1, 2]
f 2 (x) = x2 is strictly increasing on [0, 2] .
Example 8. “The composition of two non-monotone functions is again a

non-monotone function.”
Solution. { 1
The functions f (x) = x , x ̸= 0 , g (y) ≡ f (y) are not monotone on R,
0, x = 0
but their composition g (f (x)) = f (g (x)) = x is a strictly increasing function
on R.
A similar
 counterexample can be given for a finite interval: the functions
 −1 − x, x ∈ (−1, 0)
f (x) = 0, x = 0 , g (y) ≡ f (y) are not monotone on (−1, 1),

1 − x, x ∈ (0, 1)
but their composition g (f (x)) = f (g (x)) = x is a strictly increasing function
on (−1, 1).
Remark 1. Of course, a composition can give a non-monotone function
also. For instance, if f (x) = x2 , g (y) ≡ f (y), then g (f (x)) = f (g (x)) = x4 .
All the functions are non-monotone on R.
Remark 2. The following strengthened version is also false: “The compo-
sition of two functions, each of which is non-monotone anywhere on its do-
main, is again a non-monotone function.” In this formulation “non-monotone
anywhere” means that a function is not monotone in any neighborhood of
an
{ arbitrary point of the domain. The corresponding example is f (x) =
x, x ∈ Q
, g (y) ≡ f (y) (considered on any infinite or finite interval
−x, x ∈ I
symmetric with respect to the origin) and g (f (x)) = f (g (x)) = x.
Remark 3. A composition of monotone functions is always a monotone (we

consider only the cases when the composition is defined). If both functions are
strictly monotone, then its composition also is strictly monotone.
Example 9. “If a function f (x) has an inverse on [a, b], then f (x) is
monotone on [a, b]. ”
Solution.
{ Let f (x) be defined on [0, 2] in the following way: f (x) =
x , x ∈ [0, 1]
. It is easy to see that f (x) is one-to-one correspon-
1 − x , x ∈ (1, 2]
dence between{[0, 2] and [−1, 1] with the inverse function defined as follows:
y , y ∈ [0, 1]
x = f −1 (y) = . On the other hand, f (x) is not mono-
1 − y , y ∈ [−1, 0)
tone on [0, 2].
Remark 1. The same function can be used if the closed interval is sub-
stituted by open. In the last case, if one wishes to obtain an open inter-
val
{ as the image of f (x), then the function should be modified to f (x) =
x , x ∈ (0, 1)
.
1 − x , x ∈ [1, 2) {
x, x ∈ Q
Remark 2. A slightly more complex function f (x) = on
1 − x, x ∈ I
[0, 1] gives the counterexample both for the above statement and the following
strengthened statement (also false): “if f (x) has an inverse on [a, b], then there
exists a subinterval of [a, b] on which f (x) is monotone” (see explanations in
Example 10).
Remark 3. The converse is true if the monotonicity is strict.
Example 10. “If f (x) is a one-to-one correspondence between intervals

[a, b] and [c, d], then this function is monotone at least in some subintervals of
[a, b].”
Solution. {
x, x ∈ Q
Let f (x) be defined on [0, 1] in the following way: f (x) = .
1 − x, x ∈ I
It is easy to see that f (x) is a one-to-one correspondence between [0, 1]
−1
and
{ [0, 1] with the inverse function defined as follows: x = f (y) =
y, y ∈ Q
. At the same time, in any interval (c, d) ⊂ [0, 1] there are both
1−y, y ∈ I
rational and irrational points. If we choose two rational points x1 , x2 ∈ (c, d),
x1 < x2 then f (x1 ) < f (x2 ). On the other hand, for two irrational points
x3 , x4 ∈ (c, d), x3 < x4 it follows f (x3 ) > f (x4 ). Of course, the same con-
siderations are true for closed or half-open intervals. Therefore, f (x) is non-
monotone in any subinterval in [0, 1].
Remark 1. This is a finer version of the previous Example 9.

Remark 2. The same function can be used if the closed intervals are
substituted by open.
Remark 3. Due to the theorem about an inverse of a strictly monotone
function, it may appear that condition of monotonicity is somehow attached
to a possibility to find an inverse, but actually there is no requirement on
monotonicity for the existence of an inverse function.
1.7 Extrema
Example 1. “A function cannot have infinitely many strict local extrema.”

Solution.
The function f (x) = cos x has infinitely many strict local maxima xk =
2kπ, ∀k ∈ Z and minima xn = π + 2nπ, ∀n ∈ Z. These strict local extrema
are also non-strict global extrema on R.
A more interesting situation occurs when a bounded interval is considered.
In this case the counterexample can be provided with the function f (x) =
cos x1 considered on (0, 1): its infinite set of strict local maxima consists of
the points xk = 2kπ1
, ∀k ∈ N and strict local minima are xn = π+2nπ1
, ∀n ∈
N ∪ {0}. Again these strict local extrema are also non-strict global extrema
on (0, 1).
Remark. Although a local extremum is referred to a property of a function
on a small interval, one may think that a sufficient finite number of small inter-
vals always covers the given domain (for example, interval) and, consequently,
a function can possess only a finite number of local extrema. However, a given
interval (domain) can be partitioned in infinitely many small subintervals each
of which contains a strict local extremum of a function.
Example 2. “If a function has an infinite set of strict local maxima, then
there is the largest function value among these maximum function values.”
Solution.
For an infinite interval, we can consider f (x) = x cos x on (0, +∞). All
the points xk = 2kπ, ∀k ∈ N are strict local maxima and the function values
at these points f (xk ) = xk = 2kπ increase infinitely as k approaches infinity,
so there is no largest number among these values. In the same way, there is
no smallest number among the minimum values f (xn ) = −xn = −π − 2nπ at
the minimum points xn = π + 2nπ, ∀n ∈ N ∪ {0}.
For a finite interval, we can choose f (x) = x1 cos x1 defined on (0, 1). Its
1
strict local maxima are the points xk = 2kπ , ∀k ∈ N where the function takes
the values f (xk ) = xk = 2kπ, k ∈ N, and there is no largest value among
1
these values. In the same way, there is no smallest value among the minimum
values f (xn ) = − x1n = −π − 2nπ that the function takes at the minimum
1
points xn = π+2nπ , ∀n ∈ N ∪ {0}.
Example 3. “If a function f (x) is non-constant on any interval of its

domain and has a global maximum at x0 , then there exists a neighborhood of
x0 such that f (x) < f (x0 ) for every x ̸= x0 in this neighborhood.”
Solution. {
1, x ∈ Q
Consider Dirichlet’s function D (x) = . Evidently, each ra-
0, x ∈ I
tional point is a (non-strict) global maximum of D (x) and each irrational
point is a (non-strict) global minimum. However, there is no neighborhood
of a rational point x0 such that D (x) < D (x0 ) for every x ̸= x0 of this
neighborhood (in any neighborhood there exist rational points x̃ such that
D (x̃) = D (x0 ) = 1).
1
Example 4. “If f (x) has a minimum at x0 , then f (x) has a maximum at
x0 .”
Solution.
First, for local extrema, we can consider the function f (x) = |x|, that has
1 1
a local (and global) minimum at x0 = 0, but the function f (x) = |x| has no
maximum at 0 (it does not defined at 0).
For global extrema, the function f (x) = x2 − 1 has a global minimum at
1 1
x0 = 0, but the function f (x) = x2 −1 has a local, but not global, maximum
at x0 = 0.
Example 5. “If f (x) and g (x) have a maximum at x0 , then f (x) · g (x)
also has a maximum at x0 .”
Solution.
The functions f (x) = g (x) = − |x| have a local (and global) maximum at
x0 = 0, but f (x) · g (x) = x2 has a local (and global) minimum at this point.
{ Another interesting counterexample is f (x) = − |x| and g (x) =
x + 1, x ≤ 0
. Both functions have a local (and global) maximum at
−x , x > 0 { 2
x + x, x ≤ 0
x0 = 0, but their product h (x) = f (x) · g (x) = is strictly
x2 , x > 0
[ ]
increasing at x0 = 0, since x2 + x is strictly increasing on − 12 , 0 , and x2 is
strictly increasing on [0, +∞).
Remark. The statement is true for the sum of two functions.
Example 6. “If a function has more than one local minima on an interval,
then it is not invertible on this interval.”
Solution.
The function f (x) = x − 2 [x] has local minimum at each integer point. In
fact, on an interval (n − 1, n + 1), for any fixed n ∈ Z, the function is defined
{
x − 2 (n − 1) , x ∈ (n − 1, n)
as follows: f (x) = , and f (n) = −n is the
x − 2n , x ∈ [n, n + 1)
minimum value on this interval. Therefore, f (x) has infinitely many local
minima on R. However, f (x) is invertible with the inverse given by formula
f −1 (y) = y + 2 [y]: on each interval [n, n + 1) the function is defined by the
formula y = f (x) = x − 2n, which is simply invertible x = f −1 (y) = y + 2n,
and the image of each interval [n, n + 1) is [−n, −n + 1), which is different
for different values of n; therefore gluing the inverses for each n we obtain the
general inverse f −1 (y) = y + 2 [y].
Remark. Of course, the same is true for local maxima.
Exercises
1. Give one more counterexample to the statement in Example 1, section

1.2.
2. Show that f (x) = 2x2 − 1 is a counterexample to Example 3 in section
1.2 (choose X properly).
3. Give one more counterexample to Example 5, section 1.2 using the
function 2x (choose X properly).
4. Give one more counterexample to the statement in Remark 2 to Example
9, section 1.2.
5. Analyze if the following statement is true: “if g (f (x)) : X → Z is
injective composition of f (x) : X → Y and g (y) : Y → Z, then both
f (x) and g (y) are injective”. If not, provide a counterexample. What about
surjective functions?
6. Find a counterexample to the following statement: “if f (x) is bounded
1
and non-zero on (a, b), then f (x) is bounded on the same interval”.
7. It is simple to show that if f (x) is bounded on any open interval con-
tained in [a, b], then it is bounded on [a, b]. Show that a similar statement: “if
f (x) is bounded on any closed interval contained in (a, b), then it is bounded
on (a, b)” is false by providing a counterexample.
8. Show that the following statement “a non-constant function cannot have
infinitely many irrational periods” is false by providing a counterexample.
Consider a finer version of the above statement: “there is no function which has
no rational period, but has infinitely many positive irrational periods smaller
then any given positive number” and construct corresponding counterexample.
How is this situation related to Remark 2 in Example 1.4.1? (Hint: the first
counterexample is simple - just choose a basic trigonometric function, like
sin x; but the second is a bit{harder - try
√ a version of Dirichlet’s function with
1 , x = 2 · q, ∀q ∈ Q
a scaled variable - D (x) = .)
0 , otherwise
9. Construct a counterexample to the statement similar to Example 2,
section 1.4: “if both functions (defined on the same domain) have the same
fundamental period, then their ratio has the same fundamental period”.
10. Construct a counterexample to the following statement: “if both func-
tions (defined on the same domain) have the same fundamental period, then
their sum has a fundamental period.”
11. Prove that the counterexample in Remark 5 to Example 3, section 1.4
works.
12. Give a counterexample to the statement in Example 5, section 1.4,
where the sum is substituted by the product.
13. Give a counterexample to statement in Example 2, section 1.5, where
the sum is substituted by the ratio.
14. Give a non-trivial example that shows that the situation described
in Remark 3 to Example 3 in section 1.5 is feasible, albeit is rare. That is,
construct a counterexample to the following false statement: “if the product
of two odd functions is odd then both functions are zero” (Hint: Since the
product of two odd functions is always an even function and the condition
says that the result is an odd function, it follows that the product is zero.
Now, to decompose zero into two non-zero odd factors, divide all the domain,
say R, into two parts symmetric with respect to zero. Set the first function to
be zero and the second to be any non-zero odd function, say x, on the first
part, say (−1, 1), and vice-versa on the remaining part R\(−1, 1). )
15. Verify if the following statement is true: “if the composite function
g(f (x)) is odd then at least one of the functions is odd”. If not, provide a
counterexample. What about even functions?
16. Show that the difference of two functions monotone on the same domain
can be both a monotone and non-monotone function. Formulate these results
in the form of counterexamples.
17. Provide a counterexample to the statement in Example 4, section 1.6,
with the product instead of sum of functions.
18. Provide a counterexample to the statement in Example 5, section 1.6,
with the ratio instead of sum of functions.
19. Verify if Dirichlet’s function provides a counterexample for Example 1
in section 1.7.
20. Consider a refined version of the statement in Example 4, section 1.7:
1
“if f (x) has a minimum at x0 and f (x) is defined on the same domain as f (x),
1
then f (x) has a maximum at x0 .” Is this statement still false? If so, provide a
counterexample.
21. Consider the following statement: “if f (x) has a maximum at x0 and
g (x) has a minimum at x0 , then f (x) /g (x) has a maximum at x0 .” Analyze
if the statement is true. If not, provide a counterexample.
22. Consider the following stronger version of the statement in Example
6, section 1.7: “if a function has more than one local minima and maxima on
an interval, then it is not invertible on this interval”. Verify if the statement
is true and construct a counterexample if it is not.
23. Analyze the following statement: “if f (x) and g(x) attain their global
maximum values A and B on domain X, then h(x) = f (x) + g(x) has the
global maximum value A + B on X”. If it is false, provide a counterexample.
Chapter 2
Limits
Concepts
Limit (general limit). Let f (x) be defined on X and a be a limit

point of X. We say that the limit of f (x), as x approaches a, exists and
equals A if for every ε > 0 there exists δ > 0 such that for all x ∈ X such
that 0 < |x − a| < δ it follows that |f (x) − A| < ε. The usual notations are
lim f (x) = A and f (x) → A.
x→a x→a
Remark. In calculus, a non-essential simplification that f (x) is defined in
some deleted neighborhood of a is frequently used.
Partial limit. Let f (x) be defined on X and a be a limit point of X. The

partial limit of f (x) at a is the limit of f (x), as x approaches a, on a subset
of X. Among partial limits, one of the most important for the functions of
one variable are the one-sided limits.
One-sided limits. Let f (x) be defined on X and a be a limit point of
X. We say that the right-hand (left-hand) limit of f (x), as x approaches a,
exists and equals A if for every ε > 0 there exists δ > 0 such that for all x ∈ X
such that 0 < x − a < δ (−δ < x − a < 0) it follows that |f (x) − A| < ε. The
standard notations are lim f (x) = A ( lim f (x) = A ) and f (x) → A
x→a+ x→a− x→a+
(f (x) → A).
x→a−
Infinite limit. Let f (x) be defined on X and a be a limit point of X.
We say that the limit of f (x), as x approaches a, is +∞ (−∞), if for every
E > 0 there exists δ > 0 such that for all x ∈ X such that 0 < |x − a| < δ it
follows that f (x) > E (f (x) < −E). The usual notations are lim f (x) = +∞
x→a
( lim f (x) = −∞) and f (x) → +∞ (f (x) → −∞).
x→a x→a x→a
Sometimes it is also considered a “general” infinite limit in the following
sense:
35
The limit of f (x), as x approaches a, is ∞, if for every E > 0 there exists

δ > 0 such that for all x ∈ X such that 0 < |x − a| < δ it follows that
|f (x)| > E. The usual notations are lim f (x) = ∞ and f (x) → ∞.
x→a x→a
If any of the above three infinite limits is admitted at the same time the
notations used are lim f (x) = (±) ∞.
x→a
Limit at infinity. Let f (x) be defined on X, which is unbounded above
(below). We say that the limit of f (x), as x approaches +∞ (−∞), exists and
equals A , if for every ε > 0 there exists D > 0 such that for all x ∈ X such
that x > D (x < −D) it follows that |f (x) − A| < ε. The usual notations are
lim f (x) = A ( lim f (x) = A) and f (x) → A (f (x) → A).
x→+∞ x→−∞ x→+∞ x→−∞
Unicity. If the limit exists, then it is unique.
Remark 1. In the properties below it is supposed that f (x) and g (x) are
defined on the same domain.
Remark 2. The point a can be finite or infinite.
Comparative properties
1) If lim f (x) = A, lim g (x) = B and f (x) ≤ g (x) for all x ∈ X in a deleted
x→a x→a
neighborhood of a, then A ≤ B.
2) If lim f (x) = A, lim g (x) = B and A < B, then f (x) < g (x) for all
x→a x→a
x ∈ X in a deleted neighborhood of a.
Arithmetic (algebraic) properties

If lim f (x) = A , lim g (x) = B, then
x→a x→a
1) lim (f (x) + g (x)) = A + B
x→a
2) lim (f (x) − g (x)) = A − B
x→a
3) lim (f (x) · g (x)) = A · B
x→a
f (x)
4) lim = A
(under the additional condition B ̸= 0)
x→a g(x) B
If lim f (x) = A , then
x→a
5) lim |f (x)| = |A|
x→a
α
6) lim (f (x)) = Aα (under the condition α ∈ N ; or under the condition
x→a
α ∈ Z and A ̸= 0; or under the condition A > 0)
7) lim αf (x) = αA (under the assumption α > 0)
x→a
The squeeze theorem. If lim f (x) = lim g (x) = A, and the inequality
x→a x→a
Limits 37
f (x) ≤ h (x) ≤ g (x) holds for all x ∈ X in a deleted neighborhood of a, then

lim h (x) = A.
x→a
2.2 Concepts
Example 1. “If f (x) is defined on R, then there is at least one point

where lim f (x) exists.”
x→a
Solution. {
1, x ∈ Q
Consider Dirichlet’s function D (x) = . It is easy to show that
0, x ∈ I
D (x) has no limit at any real point. In fact, in any deleted neighborhood of
an arbitrary point a there exist both rational and irrational points. Therefore,
the number 1 cannot be the limit of D (x) at a, because for ε0 = 1 and
any small δ there exist points x (irrational ones) such that 0 < |x − a| < δ,
but |f (x) − 1| = |0 − 1| ≥ ε0 . Similarly, the number 0 cannot be the limit,
because in any δ-neighborhood of a there exist points x (rational ones) such
that 0 < |x − a| < δ, but |f (x) − 0| = |1 − 0| ≥ ε0 . Finally, any real number
A ̸= 0 cannot be the limit of D (x), because for ε0 = |A| in any δ-neighborhood
of a there exist points x (irrational ones) such that 0 < |x − a| < δ, but
|f (x) − A| = |0 − A| ≥ ε0 . Hence, the definition of limit is not satisfied at
arbitrary point a, whatever the value of A we try.
Example 2. “If ̸ ∃ lim f (x) then ̸ ∃ lim |f (x)|.”

x→a x→a
Solution. {
1, x ∈ Q
Consider a slight modification of Dirichlet’s function D̃ (x) = .
−1 , x ∈ I

Evidently, D̃ (x) has no limit at any point just like D (x). However, D̃ (x) = 1
and this function has the limit 1 at any point.
Remark 1. This is the false converse to the right statement: if ∃ lim f (x)
x→a
then ∃ lim |f (x)|.
x→a
Remark 2. The situation with the square is quite similar. The statement:
“if ∃ lim f 2 (x), then ∃ lim f (x)” is the false converse to the right statement:
x→a x→a
if ∃ lim f (x) then ∃ lim f 2 (x). The corresponding counterexample can be
x→a x→a
provided by the same function D̃ (x).
Example 3. “If a sequence yn converges to A and a function f (x) is such

that f (n) = yn , ∀n ∈ N, then lim f (x) = A.”
x→+∞
Solution.
Consider the function f (x) = sin πx, which in the points xn = n, ∀n ∈ N

is equal to 0. The corresponding sequence yn = f (xn ) = sin πn = 0,
∀n ∈ N converges to 0, but the function has no limit as x approaches in-
finity (f (x) has partial limits that take any value in [−1, 1], for example,
lim f (xn ) = lim sin πn = 0 for xn = n, ∀n ∈ N and lim f (xk ) =
xn →+∞ ( )
n→+∞ xk →+∞
lim sin π 2k + 12 = 1 for xk = 2k + 12 , ∀k ∈ N).
k→+∞
Remark. The converse is true: if lim f (x) = A and the sequence is
x→+∞
defined by the relation yn = f (n) , ∀n ∈ N, then lim yn = A.
n→+∞
Example 4. “If there exists a sequence xn such that lim f (xn ) = A,

xn →a
then lim f (x) = A.”
x→a
Solution. {
1, x ∈ Q
Let us consider Dirichlet’s function D (x) = and a = 0.
0, x ∈ I
Using the rational points in the form xn = n1 , n ∈ N we form the sequence
such that xn → 0 when n → +∞. Calculating the corresponding limit we
have lim D (xn ) = lim 1 = 1. However, a general limit does not exist (see
xn →0 xn →0
Example 1 for details).
Another interesting function is f (x) = sin x1 defined on R\ {0}. First, using
1
the points xk = kπ , k ∈ N we have the sequence such that xk → 0 when k →
+∞. For this sequence we get lim f (xk ) = lim sin (kπ) = lim 0 = 0.
xk →0 k→+∞ k→+∞
1
Another sequence can be chosen as follows: xn = π/2+2nπ , n ∈ N with the
same property xn → 0 when n → +∞. For this sequence lim f (xn ) =
( ) xn →0
lim sin π2 + 2nπ = lim 1 = 1. Since two partial limits are different, a
n→+∞ n→+∞
Limits 39
general limit does not exist. Actually, it can be easily shown that the set of
all the partial limits of this function covers the entire interval [−1, 1].
Remark 1. Of course, under the above conditions, if lim f (x) exists then
x→a
it is equal to A.
Remark 2. This is a weakened version of the Cauchy-Heine criterion:
lim f (x) = A if, and only if, lim f (xn ) = A for all sequences xn such that
x→a xn →a
xn → a, xn ̸= a.
n→+∞
Remark 3. Evidently, the following more general statement is also false:
“if a finite or an infinite number of partial limits exist and coincide, then a
general limit also exists”. Both counterexamples provided above work in this
case too.
Remark 4. The existence and equality of partial limits can guarantee the
existence of a general limit only in the case when the subsets of real numbers
used in these limits cover all the points of the function domain in a deleted
neighborhood of the limit point a. For example, if f (x) is defined in some
deleted neighborhood of a, then this is the case of two one-sided limits. Also
this is the case of two partial limits based first on all rational numbers and
second on all irrational numbers in a deleted neighborhood of a.
Remark 5. In practice, when an evaluation of a general limit is difficult, the
comparison of partial limits can be a useful way to get an idea about a general
limit and, in the case of inexistence of some partial limits or inequality between
two partial limits, it leads to the immediate conclusion about inexistence of a
general limit.
Example 5. “If a function has both one-sided limits, then it has a general
limit too.”
Solution. 
 −1 , x < 0
Let f (x) = sgn x = 0 , x = 0 . Then both one-sided limits exist:

1, x > 0
lim sgn x = lim (−1) = −1 and lim sgn x = lim 1 = 1, but since they
x→0− x→0− x→0+ x→0+
are different, the general lim f (x) does not exist.
x→0
Remark. The well-known in Calculus result states that a general limit
exists if, and only if, both one-sided limits exist and coincide (this statement
is true both for finite and infinite limits). Notice, however, that this result is
true in the context of the Calculus simplified supposition that the function is
defined in a deleted neighborhood of the limit point a. In general, this result
is not valid as is shown in Example 7 (although it is still true that when both
one-sided limits exist and coincide then a general limit exists also and has the
same value.)
Example 6. “If the one-sided limits at point a coincide, then there exists
a finite limit at a.”
Solution.
The function f (x) = x12 has both one-sided limits that are equal:
lim x12 = lim x12 = +∞. The general limit also exists, but it is infinite:
x→0− x→0+
lim 12 = +∞.
x→0 x
Example 7. “If f (x) is not defined in the left/right-hand neighborhood

of a, then the limit of f (x) at a does not exist.”
Solution.
Limits 41
√
If f (x) = x and a = 0 then the function is not defined in any left-hand
neighborhood of a. However,
√ applying the
√ general definition of the limit one
can easily obtain that lim x = lim x = 0. Indeed, for ∀ε > 0 choosing
x→0 x→0+
δ = ε2 we obtain for all x such that 0 < x < δ (that is, according to the
general definition, for all x of the domain of f (x)√ that √are located in the
deleted δ-neighborhood of zero) that |f (x) − 0| = x < δ = ε. Therefore,
the definition is satisfied.
Remark. In the study of limits in Calculus, the simplified assumption
that a function is defined in a deleted neighborhood of the limit point a is
frequently used. It leads to the well-known Calculus result about equivalence
between existence of a general limit and existence and equality of two one-
sided limits. This example shows that it is not so for a general definition of
the limit.
Remark to Examples 8-10. To focus on some important elements in the

definition of the limit, in the three following Examples we will use the Calculus
assumption that a function is defined in a deleted neighborhood of the limit
point a.
Example 8. “The definition of the limit can be reformulated as follows: the

limit of f (x) as x approaches a is A if for every ε > 0 there is a corresponding
δ > 0 such that if |f (x) − A| < ε then |x − a| < δ .”
Solution.
At first glance, the above “definition” conveys an impression that it has
at least some relation to the correct definition, because the inequalities of the
correct definition, meaning proximity among function values and among argu-
ment values, are involved here. Nevertheless, this statement has nothing to do
with the correct definition. In fact, if f (x) is defined on a bounded interval,
say (a, b), then the above statement does not imply any restriction on f (x),
because by choosing δ = b − a the required inequality |x − a| < δ is satisfied
for an arbitrary x ∈ (a, b) and for an arbitrary ε > 0. On the other hand, if
f (x) is defined on an unbounded domain, say on R, then the{ above “defini-
ex , x ∈ Q
tion” also fails. In fact, let us consider the function f (x) =
−ex , x ∈ I
on R. Let us choose any A > 0 and consider ε-neighborhood of A: |y − A| < ε,
ε ≤ A2 . The corresponding δ-neighborhood of the point a = ln A is defined as
|x − a| < δ with δ = max {ln (A + ε) − a, a − ln (A − ε) } = a − ln (A − ε).
It can be checked now that for all the values of f (x) that satisfied the
condition |f (x) − A| < ε it follows that |x − a| < δ (more precisely, x ∈
(ln (A − ε) , ln (A + ε) ) ⊂ (a − δ, a + δ) ). However, f (x) has no limit at a,
because for any x ∈ I (in particular for x ∈ I in the δ-neighborhood of a = ln A)
one obtains |f (x) − A| = |−ex − A| > A.
Remark. For an unbounded domain, say R, the above “definition” can also
be too restrictive. For example, if f (x) = sin x, A = 0 and ε = 3, then the
inequality |x − a| < δ holds for an arbitrary x (and a) whatever δ is. It means

that the implication of the “definition” is not true even for some continuous
functions. (We are acting here in advance, because the concept of continuity
will be considered in the next chapter.)
Example 9. “The definition of the limit can be reformulated as follows: the

limit of f (x) as x approaches a is A if for every δ > 0 there is a corresponding
ε > 0 such that if |f (x) − A| < ε then |x − a| < δ .”
Solution.
Actually this condition means that there exists a partial limit A of f (x)
as x approaches a (that is, there is a sequence of points xk such that
lim f (xk ) = A). However, it is not sufficient to guarantee the existence
xk →a { 2
x , x∈Q
of the general limit. For example, if f (x) = , a = 0 and A = 0,
1, x ∈ I
2
then by choosing 2ε = δ for any δ < 1 one guarantees that for all f (x) such

that |f (x)| = x < ε we obtain |x| < δ for the corresponding values of
x. However, the general limit does not exist, because there are two different
partial limits: lim f (x) = 0 and lim f (x) = 1.
x→0, x∈Q x→0, x∈I
Remark. If we assume that the inverse function exists, then the above
statement is the definition of the limit (equal to a) of the inverse function
x = f −1 (y) when the argument y approaches A.

Limits 43
Example 10. “The definition of the limit can be reformulated as follows:

the limit of f (x) as x approaches a is A if for every δ > 0 there is a corre-
sponding ε > 0 such that if |x − a| < δ then |f (x) − A| < ε.”
Solution.
The above “definition” seems to be rather natural and tempting, because
it apparently corresponds to a vague idea about limits: when x approaches a
the function values f (x) approach A. However, this similarity is only appear-
ing, because the inversion of the dependence between δ and ε (in comparison
with the correct definition) leads to possibility to attribute to ε arbitrary large
values, that destroys the requirement that f (x) should approach A. For ex-
ample, Dirichlet’s function D (x), which actually does not have a limit at any
point (see Example 1), satisfies the above “definition” with an arbitrary a and
A = 0 if one chooses ε = 2 for every δ > 0.
Remark. The condition of the statement implies boundedness of f (x) in
any deleted neighborhood of the point a.
Example 11. “If for a specific function f (x) and point a the dependence
of δ from ε in the limit definition cannot be expressed in the form δ = cε ,
where c is a positive constant, then f (x) does not have a limit at a.”
Solution. √
For f (x) = 3 x considered in a neighborhood
√ of the point a = 0, if δ = ε3
then the limit definition is satisfied: lim 3 x = 0. However, there is no constant
x→0
√ that δ = cε. Indeed, if 0 < |x| < δ = cε, then we can obtain
c√> 0 such
| 3 x| < 3 cε, but for any fixed c the expression in the right-hand side is
greater than ε, if ε √is sufficiently small. Therefore, we will not obtain the
required evaluation | 3 x| < ε if δ = cε.
Remark. Of course, similar statements with requirement of any other
specific law δ (ε) are also false.
Example 12. “If for a specific function f (x) and point a there are two
different ways to determine δ (ε) in the limit definition, then f (x) does not
have a limit at a.”
Solution.
Actually, if the limit definition is satisfied, then always there are infinitely
many ways to determine δ (ε). For example, if f (x) = x and a = 0 then an
evident choice to satisfy the definition of lim x = 0 is δ = ε. But it means
x→0
that for any{constant
} 0 < c < {1 the}law δ = cε is also suitable, as well
as δ = min ε, ε2 , or δ = min ε, ε3 , or more exotic δ = ln (1 + ε) and
δ = 1 − e−ε , or many others.
2.3 Elementary properties (arithmetic and comparative)
Example 1. “If ̸ ∃ lim f (x) and ̸ ∃ lim g (x) then ̸ ∃ lim (f (x) + g (x)).”
x→a x→a x→a
Solution.
Let f (x) = −g (x) = D (x). Then for an arbitrary a the limit of f (x)
and g (x) does not exist (see Example 1, section 2.2 for details). However,
h (x) = f (x) + g (x) = 0 has zero limit at any point a.
Remark 1. Similar simple examples can be given for other arithmetic
operations. For example, for the product we can use{ the following modification
1, x ∈ Q
of Dirichlet’s function: f (x) = −g (x) = D̃ (x) = , with h (x) =
−1 , x ∈ I
f (x) · g (x) = −1.
Remark 2. This is a false “negative” version of the sum rule for limits.
Remark 3. This “negative” formulation is equivalent to the direct con-
verse to the sum rule, that is the following statement is equally false: “if
∃ lim (f (x) + g (x)) then ∃ lim f (x) and ∃ lim g (x)”.
x→a x→a x→a
Example 2. “If both limits lim f (x) and lim g (x) exist and the former
x→a x→a
equals zero, then lim fg(x)
(x)
= 0.”
x→a
Solution.
Let f (x) = g (x) = x and a = 0. Then lim f (x) = lim g (x) = 0, but
x→0 x→0
f (x) x x
lim = lim = lim 1 = 1. (The function x is defined in a deleted neigh-
x→0 g(x) x→0 x x→0
borhood of zero, which is sufficient to consider the concept of the limit, and
in any such neighborhood xx =1, which justify the equality lim xx = lim 1).
x→0 x→0
Remark 1. This is a false extension of the ratio rule for limits.
Remark 2. It is important (and interesting) to note that if lim f (x) = 0
x→a
and lim g (x) = 0 then the result for lim fg(x)(x)
is unpredictable in a general
x→a x→a
case, and it can lead to a bunch of different conclusions depending on a specific
choice of functions f (x) and g (x). This is why this general situation is called
an indeterminate form and is denoted by symbol 00 . To specify different sit-
uations that can occur under the general assumption that lim f (x) = 0 and
x→a
lim g (x) = 0, let us briefly consider the following examples with the limit
x→a
point a = 0:
1) if f (x) = x2 and g (x) = x then lim f (x) = lim g (x) = 0 and lim fg(x)
(x)
=
x→0 x→0 x→0
lim x = 0.
x→0
2) if f (x) = cx (c = const) and g (x) = x then lim f (x) = lim g (x) = 0 and
x→0 x→0
f (x)
lim = lim c = c.
x→0 g(x) x→0
Limits 45
f (x)
3) if f (x) = x and g (x) = x3 then lim f (x) = lim g (x) = 0 and lim =
x→0 x→0 x→0 g(x)
lim 12 = +∞.
x→0 x
f (x)
4) if f (x) = −x and g (x) = x3 then lim f (x) = lim g (x) = 0 and lim =
x→0 x→0 x→0 g(x)
lim −12 = −∞.
x→0 x {
x, x ∈ Q
5) if f (x) = and g (x) = x then lim f (x) = lim g (x) = 0 and
0, x ∈ I x→0 x→0
f (x) f (x)
g(x) = D (x) in R\ {0}; hence lim g(x) does not exist.
x→0
Therefore, depending on a specific choice of functions f (x) and g (x), an
indeterminate form 00 can result in an arbitrary constant (including zero),
infinity or even nonexistence of the limit. This makes false any statement
about a specific property of lim fg(x)
(x)
- existence/non-existence, assuming a
x→a
specific value or be infinite - and the above specifications of f (x) and g(x) are
counterexamples to such statements.
Remark 3. The indeterminate form 00 is one of seven indeterminate forms,
∞
others being denoted by 0·∞, ∞ , ∞−∞, 1∞ , ∞0 , 00 . It is easy to specify what
each symbol means just comparing with the description for 00 . For example, an
indeterminate form 0·∞ means that lim f (x) = 0 and lim g (x) = (±) ∞ and
x→a x→a
we are considering the limit lim f (x)· g (x). Of course, similar statements and
x→a
corresponding examples can be constructed for each of indeterminate forms.
Remark 4. All the above considerations are true for both finite and infinite
limit points a.
Remark 5. The condition of non-zero limit of denominator is the important
condition for the validity of the ratio rule. At the same time, a mere fact that
the denominator has zero limit does not mean that the limit of the ratio does
not exist. This last observation is of extreme importance for one of the central
concepts in Calculus and applications - derivative, which is dealing with the
indeterminate form 00 .
Example 3. “If lim f (x) = A, lim g (x) = B and A ≤ B, then there is a

x→a x→a
deleted neighborhood of a where f (x) ≤ g (x).”
Solution.
Let f (x) = x, g (x) = 2x and a = 0. Then lim f (x) = 0 ≤ lim g (x) = 0,
x→0 x→0
but in any left-hand neighborhood of a = 0 it holds f (x) > g (x).
Remark 1. This statement is false for both finite and infinite limit points
a.
Remark 2. This is a wrong version of the correct property: if lim f (x) = A,
x→a
lim g (x) = B and A < B, then there is a deleted neighborhood of a where
x→a
f (x) < g (x).
Example 4. “If lim f (x) = A, lim g (x) = B and there is a deleted

x→a x→a
neighborhood of a where f (x) < g (x) then A < B.”
Solution.
Let f (x) = −x2 , g (x) = x2 and a = 0. Then in any deleted neighborhood
of zero f (x) < g (x) and both limits exist, but lim f (x) = 0 = lim g (x).
x→0 x→0
Remark 1. This statement is false for both finite and infinite limit points a.
For example, for a = +∞ this statement assumes the form: “if lim f (x) =
x→+∞
A, lim g (x) = B and f (x) < g (x) for all x > 0 then A < B”. The
x→+∞
corresponding counterexample is: f (x) = x1 , g (x) = 2
x with f (x) < g (x) for
all x > 0 and lim f (x) = lim g (x) = 0.
x→+∞ x→+∞
Remark 2. This is a wrong version of the correct property: if lim f (x) = A,
x→a
lim g (x) = B and there is a deleted neighborhood of a where f (x) ≤ g (x)
x→a
then A ≤ B.
Remark to Examples 3 and 4. The two last Examples show that slight mod-
ifications of the signs in formulations of the comparative properties (changing
strong inequality to non-strong and vice-versa) lead to wrong results.
Example 5. “If lim f (x) = A and lim g (x) = (±) ∞, then lim f (x) ·
x→a x→a x→a
g (x) = (±) ∞.”
Limits 47
Solution.
1
Let f (x) = x2 , g (x) = x2 and a = 0. Then lim f (x) = 0, lim g (x) = +∞,
x→0 x→0
but lim f (x) · g (x) = 1.
x→0
a.
Remark 2. Of course the simplest example is with f (x) ≡ 0, but it can
give a wrong impression that f (x) ≡ 0 is the only function when the limit of
the product exists.
Remark 3. The statement will be true if to add that A ̸= 0.
Remark 4. For the sum the corresponding statement is true: if lim f (x) =
x→a
A and lim g (x) = (±) ∞, then lim (f (x) + g (x)) = (±) ∞. The same is for
x→a x→a
the difference.
Example 6. “If lim f (x) = A and lim g (x) does not exist, then lim f (x)·
x→a x→a x→a
g (x) does not exist.”
Solution.
Let f (x) = x, g (x) = D (x) and a = 0. Then lim f (x) = 0 and lim g (x)
x→0 x→0
does
{ not exist (see Example 1, section 2.2), but h (x) = f (x) · g (x) =
x, x ∈ Q
and lim h (x) = 0 (since lim h (x) = lim x = 0 and
0, x ∈ I x→0 x→0,x∈Q x→0,x∈Q
lim h (x) = lim 0 = 0 and Q ∪ I = R).
x→0,x∈I x→0,x∈I
a.
Remark 2. Of course the simplest example is with f (x) ≡ 0, but it can
give a wrong impression that f (x) ≡ 0 is the only case when the limit of the
product exists.
Remark 3. The statement will be true if to add that A ̸= 0.
Remark 4. The existence of finite limits of both functions f (x) and g(x)
guarantees the existence of the limit of their product (and also sum and differ-
ence). On the other hand, the existence of one of the limits and non-existence
of another one assures the non-existence of the limit of the sum and difference,
but not of the product.
Remark 5. The corresponding statement for the ratio is also false.
Example 7. “If lim f (x) = A and lim g (x) = B , then lim g (f (x)) =
x→a x→A x→a
B.”
Solution.
If f (x) = 0, ∀x ∈ R and a = 0 then lim f (x) = 0 = A. Further, if
{ x→0
0, x ̸= 0
g (x) = then lim g (x) = 0 = B. However, g (f (x)) = 1, ∀x ∈ R
1, x = 0 x→0
and hence lim g (f (x)) = 1 ̸= B.
x→0
Remark 1. The statement will be true if the following condition is added:

f (x) ̸= A when x ̸= a.
Remark 2. In chapter 3 we will consider continuous functions and their
properties, but acting in advance we can recall the following Composite
function theorem: if lim f (x) = b, g (x) is continuous at b and the com-
x→a
posite function g (f (x)) is defined in a deleted neighborhood of a, then
lim g (f (x)) = g (b). It is quite tempting to relax a bit the condition of con-
x→a
tinuity in this theorem by using instead the condition of the existence of a
finite limit. But this Example shows that the result of this simple substitution
is disastrous.
Example 8. “If a function f (x) is non-negative and unbounded in a

neighborhood of a, then f (x) does not vanish in a deleted neighborhood of
a.”
Solution. { 1
sin 1 , x ̸= 0
Consider f (x) = x x . This non-negative function is un-
0, x = 0
1
bounded in a neighborhood of a = 0, since for xk = π/2+2kπ , k ∈ N we have
(π ) (π ) ( )
lim f (xk ) = lim 2 + 2kπ sin 2 + 2kπ = lim π2 + 2kπ = +∞.
xk →0 k→+∞ k→+∞
On the other hand, for xn = nπ 1
, n ∈ N we have xn → 0 and f (xn ) =
n→∞
nπ sin nπ = 0 for any n ∈ N.
Remark 1. Of course the point{zero has nothing special for this example.
1 1
x−a sin x−a , x ̸= a
A similar counterexample f (x) = can be given for
0, x = a
any a. Moreover, a can be infinite. In the last case, f (x) = x2 (1 − cos x) is a
suitable function for the counterexample.
Remark 2. If f (x) is non-constant on any interval, non-negative and
bounded on R, the statement
{ still remains false. The counterexample with
1 + sin x1 , x ̸= 0
the function f (x) = and a = 0 works in this case.
0, x = 0
Example 9. “If f (x) is not bounded in any neighborhood of a point a,
then at least one of the limits lim |f (x)| or lim |f (x)| is infinite.”
x→a− x→a+
Solution.
Let us consider the function f (x) = x1 cos x1 defined in R\ {0}. Since
for xk = 2kπ 1
, k ∈ Z\ {0} we get lim f (xk ) = lim 2kπ cos (2kπ) =
xk →0 k→±∞
lim 2kπ = ±∞, that is, the function is unbounded in any neighborhood
k→±∞
of zero. At the same time, the general limit does( not exist,
) (because )for
1
xn = π/2+nπ , n ∈ Z we obtain lim f (xn ) = lim π2 + nπ cos π2 + nπ =
xn →0 n→±∞
lim 0 = 0.
n→±∞
Remark. Under the conditions of the statement one can only guarantee
that there exists a partial infinite limit at a.
Limits 49

Remark to Examples 8 and 9. The last two Examples illustrate an “un-

expected” behavior of unbounded functions. It happens that unboundedness
and vanishing are quite combined properties even in any small interval, and
that unboundedness in any small neighborhood of a chosen point does not
imply a general tendency to infinity at that point.
Limits 51
Exercises
1. Show that cos x1 considered at the limit point a = 0 is another coun-

1
terexample to the statement in Example 4 in section 2.2. What about the e x
at a = 0?
2. Provide a counterexample to the statement: “if lim f (x) = ∞ and
x→a
lim g (x) exists then lim (f (x) + g (x)) = ∞.”
x→a x→a
3. Provide a counterexample to the statement: “if f (x) is bounded in a
neighborhood of a and lim f (x) = A, then left-hand limit also exists”.
x→a+
4. Show that the following “definition” of lim f (x) = A is not correct:
x→+∞
“for every D > 0 there exists a corresponding ε > 0 such that x > D implies
that |f (x) − A| < ε ”.
5. Show that the following “definition” of lim f (x) = +∞ is not correct:
x→a
“for every δ > 0 there exists a corresponding E > 0 such that 0 < |x − a| < δ
implies that f (x) > E ”.
6. Verify if the following “definition” of the limit is correct: “for every ε > 0
there exists δ such that for all x ∈ X such that 0 < |x − a| < δ it follows that
|f (x) − A| < ε”. If not, provide a counterexample.
7. Suppose f (x) is defined on X, a is a limit point of X, and k ∈ N is
a fixed number. Show that the following definition of the limit is false: “if
∀ε ≥ 10−k there exists δ(ε) > 0 such that ∀x ∈ X, 0 < |x − a| < δ it
follows that |f (x) − A| < ε, then lim f (x) = A”. (Hint: use the function
x→a
f (x) = 10−(k+1) [x], X = R, a = 1, A = 10−(k+1) .)
8. Give a counterexample for the statement in Example 1, section 2.3 with
the difference and ratio of functions instead of the sum.
9. Show that 0 · ∞ is the indeterminate form, that is, under the condition
lim f (x) = 0 and lim g (x) = ∞ any statement about existence/non-existence
x→a x→a
or (in the case of the existence) about a specific value (0, non-zero constant,
infinity) of the lim f (x) · g (x) is not valid. (Hint: follow the construction of
x→a
counterexamples in Remark 2 to Example 2, section 2.3)
Do the same for the indeterminate form 00 .
10. Give a counterexample to Example 3, section 2.3 such that f (x) >
g(x), ∀x ∈ R\{a}.
11. Construct a counterexample to the following statement “if sin f (x) has
a limit at a, then f (x) also has a limit at this point”. (Notice that the converse
is true.)
12. Provide the counterexample to the statement: “if lim f (x) = A,
x→a
lim g (x) = B and A2 < B 2 , then there is a deleted neighborhood of a where
x→a
f (x) < g (x)”.
13. Suppose that f (x) and g(x) are defined on X, a is a limit point of X,
and lim f (x) does not exist (either finite or infinite) while lim g (x) = ∞.
x→a x→a
Show that under the above conditions the following statements are false:
1) “f (x) + g(x) has no limit at a”
2) “f (x) + g(x) has the infinite limit at a”
3) “f (x) · g(x) has no limit at a”
4) “f (x) · g(x) has the infinite limit at a”
14. Show that the functions f (x) = x2 cos πx and g(y) = sgn2 y with the
limit point a = 0 provide another counterexample to the statement in Example
7, section 2.3.
15. Use the function in Remark 2 to Example 8 in section 2.3 to disprove
the statement in that Remark.
16. Give a counterexample to the following statement: “if f (x) is defined
on (a, +∞), bounded on any finite interval (a, b), and lim f (x)x = A, then
x→+∞
lim (f (x + 1) − f (x)) = A”. (The converse is true: if f (x) is defined on
x→+∞
(a, +∞), bounded on any finite interval (a, b), and lim (f (x+1)−f (x)) = A,
x→+∞
then lim f (x) = A.)
x→+∞ x
17. Provide a counterexample to the following false modification of the
squeeze theorem: “if lim f (x) = A ≤ B = lim g (x), and the inequality
x→a x→a
f (x) ≤ h (x) ≤ g (x) holds for all x ∈ X in a deleted neighborhood of a, then
A ≤ lim h (x) ≤ B”.
x→a
Chapter 3
Continuity
General Remark. Due to the close connection between the concepts of limit
and continuity, the majority of the examples of chapter 2 can be reformulated
for continuous functions. A few of them are described in this chapter to ex-
emplify the modifications required. Notice that all these properties are of a
local character, and they appear in section 3.2 together with some additional
examples of local properties. The greatest attention in this chapter is paid
to global properties of functions related to continuity, because these issues
were not discussed so far and they are important both in mathematics and
applications.
Concepts
Continuity at a point. A function f (x) defined on X is continuous at

a point a ∈ X if for every ε > 0 there exists δ > 0 such that whenever x ∈ X
and |x − a| < δ it follows that |f (x) − f (a)| < ε.
Remark. In calculus, a non-essential simplification that f (x) is defined in
some neighborhood of a is frequently used. In this case the continuity can be
rewritten in the form lim f (x) = f (a) or f (x) → f (a).
x→a x→a
Actually the original definition of continuity and the definition by limit
differ only in one “pathological” situation of an isolated point a of X, because
the limit definition requires a to be a limit point of X, while the general
definition of continuity does not contain such a requirement. It means that
at any isolated point a of X, a function f (x) is continuous according to the
general definition, but not continuous according to the definition by limit.
Since the behavior of a function at an isolated point is hardly of any interest,
53
we will usually consider the definition of continuity by limit as a complete

definition.
One-sided continuity. A function f (x) defined on X is right-hand (left-

hand) continuous at a point a ∈ X if for every ε > 0 there exists δ > 0 such
that whenever x ∈ X and 0 ≤ x − a < δ (−δ < x − a ≤ 0) it follows that
|f (x) − f (a)| < ε.
Continuity on a set. A function f (x) defined on X is continuous on a

set S ⊂ X, if f (x) is continuous at every point of S.
Remark. It is worth to notice that in the case when endpoints belong
to the interval, for example a closed interval [a, b], continuity on [a, b] means
continuity at every point c ∈ (a, b) plus right-hand continuity at a and left-
hand continuity at b.
Discontinuity point. Let a be a limit point of the domain of f (x). If

f (x) is not continuous at a, then a is a point of discontinuity of f (x) (or
equivalently, f (x) has a discontinuity at a).
Remark. Sometimes it is required that a should be a point of the domain
of f (x). We will not impose this restriction.
Classification of discontinuities.
1) Removable discontinuity. f (x) has a removable discontinuity at a, if the
finite limit of f (x), as x approaches a, exists, but the function is not defined at
a or has the value different from the value of the limit. It is called removable,
because the proper redefinition of the original function just at the point a
removes this discontinuity.
2) Discontinuity of the first kind (or jump discontinuity). f (x) has a jump
discontinuity at a, if both one-sided limits of f (x), as x approaches a, exist
and finite, but they have different values (that is, a general limit does not
exist).
3) Discontinuity of the second kind (or essential discontinuity). f (x) has an
essential discontinuity at a, if at least one of the one-sided limits does not
exist or is infinite.
Remark. There are other variants (more and less detailed) of the classifi-
cation of discontinuities.
Local properties
Remark. The properties of functions that characterize the function behav-

ior in an arbitrary small neighborhood of a point are called the local properties
as opposed to the global properties related to the function behavior on a chosen
fixed set (in particular, on the entire domain).
Continuity 55
1) If f (x) and g (x) are continuous at a, and f (x) ≤ g (x) for all x ∈ X in a
deleted neighborhood of a, then f (a) ≤ g (a).
2) If f (x) and g (x) are continuous at a, and f (a) < g (a), then f (x) < g (x)
for all x ∈ X in a neighborhood of a.

If f (x) and g (x) are continuous at a, then
1) f (x) + g (x) is continuous at a
2) f (x) − g (x) is continuous at a
3) f (x) · g (x) is continuous at a
4) fg(x)
(x)
is continuous at a (under the additional condition g (a) ̸= 0)
If f (x) is continuous at a, and α ∈ R, then
5) |f (x)| is continuous at a
α
6) (f (x)) is continuous at a (under the condition α ∈ N; or under the
condition f (a) > 0)
7) αf (x) is continuous at a (under the assumption α > 0)
Composite function theorem. If lim f (x) = b, g (x) is continuous at

x→a
b and the composite function g (f (x)) is defined in a deleted neighborhood of
a, then lim g (f (x)) = g (b).
x→a
Corollary. If f (x) is continuous at a, g (x) is continuous at f (a), and the
composite function g (f (x)) is defined in a neighborhood of a, then g (f (x))
is continuous at a.
Global properties
Remark. The properties of functions that characterize the function behav-

ior on a chosen set (in particular, on the entire domain) are called the global
properties as opposed to the local properties that describe the behavior in a
small neighborhood of a point.
Open Set Characterization. f (x) is continuous on a domain X if, and

only if, the inverse image f −1 (S) of any open set S ⊂ f (X) is open.
Closed Set Characterization. f (x) is continuous on a domain X if,
and only if, the inverse image f −1 (S) of any closed set S ⊂ f (X) is closed.
The Compact Set Theorem. If f (x) is continuous on a compact
set S, then its image f (S) is also a compact set. In particular, if f (x) is
continuous on [a, b], then its image is [fmin , fmax ], where fmin = min f (x),
x∈S
fmax = max f (x).
x∈S
The First Weierstrass Theorem. If f (x) is continuous on a compact

set, then f (x) is bounded on this set. In particular, if f (x) is continuous on

[a, b], then it is bounded on this interval.
The Second Weierstrass Theorem. If f (x) is continuous on a compact
set, then f (x) attains its global maximum and minimum values on this set. In
particular, if f (x) is continuous on [a, b], then it attains its global maximum
and minimum values on this interval.
Remark. Evidently, the Second Weierstrass Theorem includes the result
of the First Theorem. However, by historical reasons, these two theorems are
frequently considered in the given above sequence.
Intermediate Value property. A function f (x) satisfies the interme-

diate value property on [a, b], if for any pair x1 , x2 ∈ [a, b], x1 < x2 , the
image of [x1 , x2 ] under f (x) contains [f (x1 ) , f (x2 )] if f (x1 ) ≤ f (x2 ) or
[f (x2 ) , f (x1 )] if f (x1 ) ≥ f (x2 ) (that is, if f (x) takes on two values some-
where on [a, b], it also takes on every value in between).
The Intermediate Value Theorem. If f (x) is continuous on a con-
nected set S, then f (x) satisfies the intermediate value property on S. In
particular, if f (x) is continuous on a connected set S, then its image f (S) is
also a connected set.
Uniform continuity
Uniform Continuity. f (x) is uniformly continuous on a set S if for every

ε > 0 there exists δ > 0 such that whenever x1 , x2 ∈ S and |x1 − x2 | < δ it
follows that |f (x1 ) − f (x2 )| < ε.
Nonuniform Continuity. f (x) fails to be uniformly continuous on a

set S if there exists ε0 > 0 such that for any δ > 0 can be found two points
xδ and x̃δ in S such that |xδ − x̃δ | < δ, but |f (xδ ) − f (x̃δ )| ≥ ε0 .
Criterion for Nonuniform Continuity. f (x) fails to be uniformly
continuous on a set S if there exists ε0 > 0 and two sequences xn and x̃n in
S such that |xn − x̃n | → 0, but |f (xn ) − f (x̃n )| ≥ ε0 .
n→∞
The Cantor Theorem. If f (x) is continuous on a compact set S, then

f (x) is uniformly continuous on S.
Continuity 57
3.2 Local properties
Example 1. “If a function has a limit at a, then it is continuous at a.”

Solution.
Let f (x) = sinx x , a = 0. In this case, lim f (x) = 1 , but f (x) is not defined
x→0
at a = 0, so f (x) is not continuous at a = 0. This kind of discontinuity is
called removable, because the redefinition of { the original function just at a
x , x ̸= 0 is a continuous
sin x
single point can remove discontinuity: f˜ (x) =
1,x=0
function at a = 0 (and at any other real point).
Remark 1. Another example of the same removable discontinuity is when
f (x) is defined at a, but lim f (x) ̸= f (a), for instance (in line with the
{ x→a
x , x ̸= 0 , a = 0 has a removable discontinuity
sin x
above example), fˆ (x) =
0,x=0
at a = 0.
Remark 2. Besides a removable discontinuity presented in the above coun-
terexamples, there are two more kinds of discontinuities (actually there are
variations of the classification of discontinuities). Let us briefly list these sit-
uations below:
1)f (x) = sgn x, a = 0. In this case, lim f (x) does not exist, and consequently
x→0
f (x) is not continuous at a = 0, but both one-sided limits exist: lim f (x) =
x→0−
−1 and lim f (x) = 1. This is an example of a discontinuity of the first kind
x→0+
(or jump discontinuity).
2) f (x) = x12 , a = 0. In this case, lim f (x) = +∞ and this is an example of
x→0
a discontinuity of the second kind (or essential discontinuity). The same kind
of discontinuity occurs when{ the limit does not exist (either finite or infinite),
sin x1 , x ̸= 0
as for the function f (x) = at the point a = 0.
0,x=0
Example 2. “If a function is continuous at a, then it has a finite limit at
a.”
Solution. { √
x, x ≥ 0
For f (x) = the point a = −1 is isolated (that is, there is
0 , x = −1
a deleted neighborhood of a = −1 where there is no point of the domain of
f (x)). For this reason, a = −1 is not a limit point for the domain of f (x),
and hence the limit of f (x) does not exist at a = −1. On the other hand,
f (x) is continuous at a = −1 by definition.
Remark to Examples 1 and 2. While the fact that the existence of a

finite limit does not imply a continuity at the same point is pointed out in all
Calculus books (see Example 1), the converse statement presented in Example
2 is generally accepted as correct, although there are some nuances regarding
this.
Example 3. “Whatever function f (x) is considered, if only an informa-

tion on its domain is available, it is not possible to make any conclusion about
continuity of this function.”
Solution.
By definition, any function with the domain Z is continuous at every point
of its domain (every point of Z is isolated and the definition of the continuity
is satisfied).
Remark. Of course, any function with the domain Z has no limit due to
the absence of limit points in the domain.
Remark to Examples 2 and 3. The situation with an isolated point of the

domain of f (x) is a kind of pathological situation with no importance for
understanding of the continuity, and apparently with no interest in applica-
tions. For this reason, in calculus this situation is not even considered and
it usually assumed that a function f (x) is continuous at a if , and only if,
lim f (x) = f (a). So usually we will avoid this hardly useful specific situation
x→a
of an isolated point and will focus on the cases when a is a limit point of the
domain of a function f (x).
Example 4. “If f 2 (x) is continuous at a point a, then f (x) is also

continuous at this point.”
Solution.
We can use the same example as for the limits (Example 2, section 2.2):
Continuity 59
{
1, x ∈ Q
f (x) = D̃ (x) = is discontinuous (essential discontinuity) at
−1 , x ∈ I
every real point, while f 2 (x) = 1 is continuous on R.
Remark 1. Other simple examples with different kinds of discontinuity
can be provided:
{
1 , x ̸= 0
1) f (x) = has a removable discontinuity at a = 0, but
−1 , x = 0
f 2 (x) = 1 {is continuous on R.
1, x < 0
2) f (x) = has a jump discontinuity at a = 0, but f 2 (x) = 1
−1 , x ≥ 0
is continuous on R.
Remark 2. This is the wrong converse to the correct statement: if f (x) is
continuous at a point a, then f 2 (x) is also continuous at this point.
Remark 3. The three last counterexamples can also be used to disprove a
similar false statement about the absolute value: “if |f (x)| is continuous at a
point a, then f (x) is also continuous at this point”. The converse is correct:
if f (x) is continuous at a point a, then |f (x)| is also continuous at this point.
Example 5. “If f (x) is continuous, positive and unbounded in any right-

1
hand side neighborhood of a, then f (x) is bounded in some right-hand side
neighborhood of a.”
Solution.
The function f (x) = x1 sin x1 + 2x is positive, continuous on (0, 1) (check
it by applying the arithmetic and composition rules) and unbounded in any
right-hand side neighborhood of a = 0, because for xn = (1+2n)π 2
, n ∈ N we
have ( )
(1 + 2n) π 4
lim f (xn ) = lim + = +∞.
xn →0+ n→+∞ 2 (1 + 2n) π
1 x
At the same time, g (x) = f (x) = |sin 1/x|+2x2 is also unbounded, because for
1
xk = kπ , k ∈ N we obtain
xk kπ
lim g (xk ) = lim = lim = +∞.
xk →0+ k→+∞ 2x2k k→+∞ 2
Remark. A similar example can be given for an infinite point a. The false
statement: “if f (x) is continuous, positive and unbounded on (0, +∞), then
1
f (x) is bounded on (0, +∞)”, can be disproved by using counterexample with
function f (x) = x sin x + x + x1 .
Example 6. “If f (x) is continuous at 0 and f (0) = 0, then the inequality

|f (x)| ≤ |x| holds at least in some neighborhood of 0.”
Solution. √ √ √
If f (x) = 3 x then f (x) is continuous at 0 (since lim 3 x = 0 = 3 0),
√ x→0
however 3 x > x for any x ∈ (0, 1).
Remark 1. This Example shows that the continuity does not prescribe
any rate of approximation of a function to its value at the continuity point.
The only thing that is guaranteed is the existence of approximation (but not
its form).
Remark 2. The converse statement is true.
Example 7. “If f (x) satisfies lim [f (a + h) − f (a − h)] = 0, then f (x)

h→0
is continuous at a.”
Solution.
If f (x) = x12 and a = 0, then the condition of the statement is satisfied,
because x12 is an even function. However, f (x) is not continuous at zero (it is
not even defined there and besides lim x12 = +∞).
x→0
Remark 1. The continuity condition is satisfied if a function approaches
its value f (a) whatever the form of tendency of x to a is chosen. Therefore,
any specification of the way that x approaches a can destroy the concept of
continuity. The above Example shows how it happens when the pairs of points
approaching a are symmetric with respect to a.
Remark 2. The converse is true.
Example 8. “If f (x) and g (x) are discontinuous at a point a, then

f (x) + g (x) is also discontinuous at this point.”
Solution.
Continuity 61
We can use the same example as for the limits (Example 1, section 2.3):
f (x) = −g (x) = D (x). Since Dirichlet’s function does not have a limit at
any point, it is discontinuous at any point. However, h (x) = f (x) + g (x) = 0
has zero limit at any point a.
Remark 1. This is a false “negative” version of the sum rule for continuous
functions.
Remark 2. Since there are different kinds of discontinuity, it is worth
noting that a counterexample can be constructed with discontinuity of any
kind: { sin x {
x , x ̸= 0 and g (x) =
0 , x ̸= 0
1) if f (x) = , both with a re-
0, x = 0 1, x = 0
movable discontinuity at a = 0, then the function h (x) = f (x) + g (x) =
{
x , x ̸= 0
sin x
is continuous at the origin.
1, x = 0
{ 2 {
x + 1, x < 0 x, x < 0
2) if f (x) = and g (x) = , both with a
x, x ≥ 0 x2 + 1 , x ≥ 0
jump discontinuity at a = 0, then the function h (x) = f (x)+g (x) = x2 +x+1
is continuous{at zero. {
x2 , x ̸= 0 and g (x) = x − x12 , x ̸= 0
1
3) if f (x) = , both having the
0, x = 0 0, x = 0
infinite limits at a = 0, that is, an essential discontinuity at a = 0, then the
function h (x) = f (x) + g (x) = x is continuous at zero.
Remark 3. If f (x) and g (x) have different kinds of discontinuities at a
point a then the statement of this Example is true.
Remark 4. Similar simple examples can be given for other arithmetic op-
erations. For instance, for the product we can use the same modified Dirichlet
function as in Example 2, {section 2.2 of the limits:
1, x ∈ Q
f (x) = −g (x) = D̃ (x) = , with h (x) = f (x) · g (x) = −1.
−1 , x ∈ I
Example 9. “If f (x) is continuous at a point a and g (x) is discontinuous
at a, then f (x) · g (x) is discontinuous at this point.”
Solution.
We can use an example similar to that applied in limits
{ (Example 6, section
0, x ∈ Q
2.3): f (x) = x is continuous and g (x) = 1 − D (x) = is discon-
1, x ∈ I
tinuous
{ at a = 0 (an essential discontinuity). However, h (x) = f (x) · g (x) =
0, x ∈ Q
is continuous at a = 0, because lim h (x) = 0 = h (0).
x, x ∈ I x→0
Remark 1. Other simple examples can be provided with other types of
discontinuity: {
sin x, x ̸= 0
1) f (x) = x is continuous and g (x) = has a removable dis-
1, x = 0
continuity at a = 0, but f (x) · g (x) = x sin x is continuous at a = 0.
2) f (x) = x is continuous and g (x) = sgn x has a jump discontinuity at a = 0,
but f (x) · g (x) = |x| is continuous at a = 0.
{
2 1/x, x ̸= 0
3) f (x) = x is continuous and g (x) = has an essential dis-
1, x = 0
continuity at a = 0, but f (x) · g (x) = x is continuous at a = 0.
Remark 2. The statement will be true if to add that f (a) ̸= 0.
Remark 3. The corresponding statement for the ratio is also false, but the
statement for the sum and difference is true.
Example 10. “If f (x) is defined in a neighborhood of a point a, is in-

creasing on the left and decreasing on the right of a, then a is a local maximum
point of f (x).”
Solution. { 1
The function f (x) = x2 , x ̸= 0 satisfies the conditions of the state-
0, x = 0
ment in a neighborhood of a = 0, but f (0) = 0 < x12 = f (x) for ∀x ̸= 0.
Remark. This statement will be true if the condition of continuity at the
point a would be added.
3.3 Global properties: general results
Example 1. “If f 2 (x) is continuous on [a, b], then f (x) is also continuous
on this interval.”
Solution. {
1, x ∈ Q
The modified Dirichlet function D̃ (x) = is not continuous
−1 , x ∈ I
at any point (it does not even have a limit - see Example 2, section 2.2), but
D̃2 (x) = 1 is continuous on an arbitrary interval [a, b] .
Remark 1. The same counterexample works also for (a, b).
Remark 2. The same counterexample works also for following false state-
ment about the absolute value of a function: “if |f (x)| is continuous on [a, b]
(or (a, b)) then f (x) is also continuous on this interval”.
Example 2. “If f (x) is continuous on [a, b] and is not anywhere constant,

then it can not take its global maximum and minimum values infinitely many
times. ”
Solution. {
−x + x cos x1 , x ̸= 0
Let us consider f (x) = on [0, 1]. Due to arith-
0, x = 0
metic and composition rules this function is continuous on (0, 1]. It is also
continuous at 0, because lim f (x) = 0 = f (0) (the result for the limit follows
x→0
from the following evaluation |f (x)| = |−x| · 1 − cos 1 ≤ 2 |x| → 0). There-
x x→0
fore, f (x) is continuous on [0, 1]. Also there is no interval inside [0, 1] where
Continuity 63
f (x) is constant. However, this( function takes ) its global maximum infinitely
many times. Indeed, f (x) = x −1 + cos x1 ≤ 0 for ∀x ∈ [0, 1] and f (0) = 0,
so 0 is the global maximum value of f (x). It is easy to see that the function
takes this value in all points xn = 2nπ 1
, n ∈ N. Since all these points lie in
[0, 1], f (x) takes its global maximum infinitely many times.
Remark 1. A similar example can be constructed for a global mini-
mum or for both minimum and maximum. In the former case one can just
use the function { symmetric to the shown above with respect to the x -axis:
x − x cos x1 , x ̸= 0
g (x) = −f (x) = on [0, 1]. In the latter case, the func-
 0, x = 0
 −x + x cos πx , x ∈ (0, 1]
tion h (x) = −x + (2 − x) cos 2−x π
, x ∈ (1, 2) considered on [0, 2] shows

0 , x = 0 ; −2 , x = 2
that continuous anywhere non-constant function can take both the global
maximum and minimum values infinitely many times.
Remark 2. Since the behavior of the continuous on an interval function
is supposed to be sufficiently “smooth”, it may appear that it cannot have
too many oscillations within a finite interval in order to attain infinitely many
extreme values. However, it is possible.
Example 3. “If a function is defined on R, it cannot be continuous at

only one point.”
Solution: {
x, x ∈ Q
The function f (x) = is discontinuous at any point, except
−x , x ∈ I
for 0. In fact, in any neighborhood of every point x0 there are rational and
irrational points. Hence, we can consider two partial limits at x0 :
lim f (x) = lim x = x0

x→x0 , x∈Q x→x0 , x∈Q
and
lim f (x) = lim (−x) = −x0 .
x→x0 , x∈I x→x0 , x∈I
Since these partial limits are different for any x0 ̸= 0, it follows that the limit
of f (x) does not exist at any x0 ̸= 0, and therefore, f (x) is not continuous
at any x0 ̸= 0. At the point x0 = 0 we get lim f (x) = 0 = f (0), so f (x) is
x→0
continuous at x0 = 0.
Remark. Of course, the statement is also false for an arbitrary domain.
For example, if the domain is the interval (−a, a), then one can choose the
same function in the counterexample.
Example 4. “A function cannot have infinitely many continuity points

and infinitely many discontinuity points at the same time.”
Solution.
A trivial counterexample can be constructed by just gluing a continuous
Continuity 65
and discontinuous
{ functions defined on different sets. For instance, the func-
1 , x ∈ [−1, 0]
tion f (x) = defined on [−1, 1] is continuous at every
D (x) , x ∈ [0, 1]
point in [−1, 0) and discontinuous at every point in [0, 1].
Much more interesting counterexample arises if we interpret the inexact
term “at the same time” in a more strict sense: on any interval in the function
domain. In this case, one can use the Riemann function defined as follows.
Suppose that any rational number is represented in the form x = m n , where
m is integer, n is natural, and the fraction m n is in lowest terms (the last
means that m and n have no common factor). In this case the numbers m
and
 1 n are uniquelymdetermined, and the Riemann function f (x) = R (x) =
 n , x ∈ Q, x = n
0, x ∈ I is well-defined on R. It can be shown that this function

1, x = 0
is continuous at every irrational point and discontinuous at every rational
point. Indeed, in any neighborhood of a rational point x1 ∈ Q there exist
irrational ones. Therefore, the partial limit lim f (x) = 0 is different
x→x1 , x∈I
from f (x1 ) which is equal n1 if x1 = m
n or 1 if x1 = 0. It means that f (x) is
discontinuous at every rational point in R. Now, let us consider an arbitrary
irrational point x2 ∈ I. First, approaching x2 by irrational points we readily
obtain lim f (x) = 0 = f (x2 ). If we choose any set of rational points
x→x2 , x∈I
x = m n approaching x2 , then we can note that the denominator n should
approach infinity, because if it would not be so, that is, if the denominator
n would be bounded, say n ≤ n0 , then it will be only finitely many numbers
of the form x = m n , n ≤ n0 in a neighborhood of the point x2 . However, it
will contradict the well-known fact that in any neighborhood of an arbitrary
real point there are infinitely many rational points. But if the condition that
the rational points x = mn approach x2 implies that n → +∞, then it follows
that lim f (x) = lim n1 = 0 = f (x2 ). Since the two considered partial
x→x2 , x∈Q n→+∞
limits involve all the real points (Q∪I = R), we can conclude that lim f (x) =
x→x2
0 = f (x2 ), i.e., f (x) is continuous at every irrational point in R.
Remark. It can be shown that there does not exist a function continuous
at every rational point and discontinuous at every irrational point.
Remark to Examples 3 and 4. The number of points where a defined on

R function can be continuous and discontinuous is largely variable. Examples
3 and 4 illustrate two possible distributions of continuity and discontinuity
points, the first being simple while the second being quite non-trivial.
Example 5. “If f (x) is continuous on R and lim f (n) = A, then

n→+∞
lim f (x) = A.”
x→+∞
Solution.
The function f (x) = sin πx is continuous on R and lim f (n) =

n→+∞
lim sin πn = 0, but this is only one of the partial limits when x approaches
n→+∞ ( ) ( )
+∞. Another partial limit lim f 2n + 12 = lim sin 2πn + π2 = 1 gives
n→+∞ n→+∞
a different result, so lim f (x) does not exist.
x→+∞
Remark 1. The converse is true: if lim f (x) = A and f (x) is defined
x→+∞
on N, then lim f (n) = A.
n→+∞
Remark 2. This Example is a strengthened form of Example 3 in section
2.2: it shows that even continuity on R is not sufficient to make a conclusion
about the behavior at infinity by knowing only a specific partial limit at
infinity.
Example 6. “If f (x) is continuous on an interval A and on an interval

B, then it is also continuous on A ∪ B.”
Solution. {
x , x ∈ [0, 1]
The function f (x) = is continuous on A = [0, 1] and
x + 1 , x ∈ (1, 2)
on B = (1, 2), but it is not continuous on A ∪ B = [0, 1] ∪ (1, 2) = [0, 2),
because f (x) has a jump discontinuity at the point 1.
Remark. The statement will be true if A ∩ B ̸= ∅.
Continuity 67
Example 7. “If f (x) satisfies the condition lim [f (x + h) − f (x − h)] =

h→0
0, ∀x ∈ R, then f (x) is continuous on R.”
Solution. {
1, x ∈ Z
If f (x) = then the statement property is satisfied: for
0, x ∈/Z
∀x ∈ / Z there is a neighborhood where all the points are not integers, so
lim [f (x + h) − f (x − h)] = 0; for ∀x ∈ Z in a deleted neighborhood of ra-
h→0
dius 1 all the points are not integers, so again lim [f (x + h) − f (x − h)] = 0.
h→0
However, for ∀x ∈ Z it follows that lim f (x + h) = 0 ̸= 1 = f (x), that is,
h→0
f (x) is not continuous at any integer.
Remark 2. This Example is a replica of Example 7 in the previous section
reformulated for an entire domain of a function. A simple even function does
not provide a counterexample in this case, but the symmetry in the choice of
points in the suggested “definition” still lead to violation of the continuity.
1
Example 8. “If function f (x) is continuous on an interval, then f (x) is
also continuous on this interval.”
Solution.
The function f (x) = sin x is continuous on R, but f (x)
1
= sin1 x is discon-
tinuous at all points xk = kπ, k ∈ Z.
Remark. The statement will be true with the additional condition f (x) ̸=
0 for all points of the interval.
Example 9. “If f (x) is continuous and bounded on an interval and

f (x) ̸= 0 for every x in this interval, then f (x)
1
is also continuous and bounded
on this interval.”
Solution.
Consider the continuous function f (x) = x which is bounded and different

1
from zero on (0, 1). Evidently, f (x) = x1 is unbounded on (0, 1).
1
Remark. Under the statement conditions, the function f (x) maintains
continuity on the chosen interval.
Example 10. “If f (x) is continuous and unbounded on (0, +∞), and
f (x) > 0 , ∀x ∈ (0, +∞), then lim f (x)
1
= 0.”
x→+∞
Solution.
Consider f (x) = x sin x+x+1 on (0, +∞). Evidently, f (x) is continuous on
(0, +∞) due to the arithmetic properties of continuous functions and f (x) =
x (sin x + 1) + 1 > 0 for ∀x ∈ (0, +∞). Also, choosing xn = nπ, n ∈ N one
obtains the following partial limit
lim f (xn ) = lim (nπ (sin nπ + 1) + 1) = +∞,

xn →+∞ n→+∞
implying that f (x) is unbounded on (0, +∞). On the other hand, for the same
sequence xn = nπ we obtain
1 1
lim = lim = 0,
xn →+∞ f (xn ) n→+∞ nπ (sin nπ + 1) + 1
Continuity 69
while for xk = − π2 + 2kπ, k ∈ N the partial limit is different:
1 1
lim = lim
xk →+∞ f (xk ) k→+∞ (2kπ − π/2) (sin (2kπ − π/2) + 1) + 1
1
= lim = 1.
k→+∞ 0 + 1
1
It means that the limit lim does not exist.
x→+∞ f (x)
Remark 1. We can also consider a modified statement: “if f (x) is contin-
uous and unbounded on (0, +∞), and f (x) > 0 , ∀x ∈ (0, +∞), then f (x)1
is
bounded on (0, +∞)”. A similar counterexample is provided by the function
f (x) = x sin x + x + x1 on (0, +∞).
Remark 2. Similar statements for bounded intervals are also false.
3.4 Global properties: the famous theorems
3.4.1 Mapping sets
General Remarks. In this subsection the question of what properties of

the sets are preserved under a continuous mapping is considered. One knows
that a continuous function preserves openness and closeness of any set in its
image. Naturally the question arises if analogous property is true for sets
contained in its domain. The answer to this question is given in the first two
Examples. On the other hand, it is well-known that a continuous function
preserves compactness of any set from its domain, but it does not do so for
sets from its image, as shown in Example 3. Further, boundedness is one of
the components of compactness, but this property alone is not preserved by
a continuous function as shown in Example 4.
Example 1. “If f (x) is continuous on (a, b), then its image is also an
open interval.”
Solution.
Evidently, the function f (x) = x2 is continuous on (−1, 1), but its image
f ((−1, 1)) is the interval [0, 1), which is not open.
Remark 1. This false statement can be reformulated in a more general
form (which is also false): “if f (x) is continuous on an open set S then its
image is also an open set”.
Remark 2. This is a wrong reformulation of the following result (frequently
called the Open Set Characterization of continuous functions): f (x) is contin-

uous on a domain X if, and only if, the inverse image f −1 (S) of any open set
S contained in the image f (X) is open. (Recall that for a function f (x) with
domain X and image Y = f (X) the inverse image f −1 (S) of a set S ⊂ Y is
the set of all points x ∈ X such that f (x) ∈ S).
Example 2. “If f (x) is continuous on a closed set [a, +∞), then its image
is also a closed set.”
Solution.
1
The function f (x) = 1+x 2 is continuous on [0, +∞) (just apply the arith-
metic properties), but its image f ([0, +∞)) is the interval (0, 1], which is not
closed (to determine the image notice that f (0) = 1, f (x) is strictly decreas-
ing and positive on [0, +∞) and lim f (x) = 0).
x→+∞
Remark 1. This false statement can be reformulated in a more general
form (which is also false): “if f (x) is continuous on a closed set S then its
image is also a closed set”.
Remark 2. This is a wrong reformulation of another characterization of
continuous functions: f (x) is continuous on a domain X if, and only if, the
inverse image f −1 (S) of any closed set S contained in the image f (X) is
closed.
Continuity 71
Example 3. “If f (x) is continuous on S and its image f (S) is a compact

set, then S is also a compact set.”
Solution.
The function f (x) = sin x is continuous on S = (0, 2π) and the image
f (S) = [−1, 1] is the compact, but the original set S is not a compact set.
Example 4. “If f (x) is continuous on a bounded set, then its image is

also a bounded set.”
Solution. ( ) (( ))
The function f (x) = tan x is continuous on 0, π2 , but its image f 0, π2
is the interval (0, +∞), which is not bounded.
Remark to Examples 1-4. These statements can be viewed as wrong mod-

ifications of the correct statement on the preservation of compactness under a
continuous mapping: if f (x) is continuous on a compact set, then its image is
also a compact set. (Recall that in R a compact set is a bounded and closed
set, for example [a, b], where a and b are finite points.) This important global
property sometimes is called the Compact Set theorem.
3.4.2 Weierstrass theorems
General Remarks. Examples of this subsection analyze different elements

in the formulation of the Weierstrass theorems: the first three Examples deal
with the first theorem, and the remaining three examples consider the second
theorem. Examples 5, 6 and 8 show that the kind of a set considered in both
theorems is important for validity of the results: if only one of the features
of a compact set - closeness or boundedness - is violated, the results of the
theorems will not be true. Examples 7 and 10 investigate attempts to formulate
the converses to the Weierstrass theorems. Finally, Example 9 illustrates what
happens if the continuity condition is violated just in one point.
Example 5. “If f (x) is continuous on (a, b), then it is bounded on this

interval.”
Solution. ( )
The function f (x) = tan x is continuous on − π2 , π2 , but it is not bounded
there ( lim tan x = +∞).
x→π/2−
Remark 1. This function is also counterexample for the following state-
ment: “if f (x) is continuous on (a, b) then it attains its global minimum and
maximum on this interval”.
Remark 2. Of course, the interval (a, b) can be substituted by [a, b) or
(a, b] with the [same)false statement. For instance, for [a, b) the same function
considered on 0, π2 works.
Example 6. “If f (x) is continuous on a closed set, then it is bounded on

this set.”
Solution.
The function f (x) = x is continuous on the closed set [0, +∞), but it is
not bounded there.
Remark to Examples 5 and 6. These statements are wrong modifications

of the correct result representing an important global property of continuous
functions: if f (x) is continuous on [a, b], then it is bounded on this interval.
The same result holds for any compact set in R. This result is frequently called
the first Weierstrass theorem.
Example 7. “If f (x) is bounded on [a, b], then it is continuous on this

interval.”
Solution.
Dirichlet’s function is bounded on any interval [a, b], but it does not even
have a limit at any point (see Example 1, section 2.2 for details).
Remark 1. The same example works also for (a, b).
Remark 2. Another simple example with a less “pathological” function
Continuity 73
is f (x) = sgn x considered on such [a, b], that 0 ∈ (a, b) (it has a jump
discontinuity at the point zero).
Remark 3. The converse is true and represents the first Weierstrass theo-
rem.
Example 8. “If f (x) is continuous and bounded on (a, b), then it attains
its global minimum and maximum on this interval.”
Solution.
The function f (x) = x is continuous and bounded on (−1, 1), but it does
not attain either minimum or maximum values.
Remark. For an infinite interval, for example (−∞, +∞), the function
f (x) = arctan x provides a counterexample.
FIGURE 3.4.3: Example 9, Remark 1
Example 9. “If f (x) is defined on [a, b] and continuous on [a, b] except

for one point c ∈ [a, b], then it attains its global extrema on [a, b].”
Solution. {
|x| , x ∈ [−1, 0) ∪ (0, 1]
The function f (x) = is defined on [−1, 1] and
1, x = 0
continuous on [−1, 0) and (0, 1], but not at 0 (it has a removable discontinuity
at 0: lim f (x) = 0, but f (0) = 1). At the same time this function does not
x→0
attain its minimum, because lim f (x) = 0, but all the function values are
x→0
positive.
Remark 1. Another counterexample when function, which satisfies the
above conditions, does not attain
 either global minimum or maximum is pro-
 x + 1, x ∈ [−1, 0)
vided by the function f (x) = x − 1, x ∈ (0, 1] on [−1, 1].

0, x = 0
Remark 2. Of course, the discontinuity point c can be one of the endpoints.
For instance, the function of the above counterexample considered on [0, 1],
has a discontinuity at the left-hand endpoint.
Remark to Examples 8 and 9. These statements are wrong modifications

of the following important global property of continuous functions: if f (x) is
continuous on [a, b] then it attains global maximum and minimum values on
this interval. The same result holds for any compact set in R. This property
is frequently called the second Weierstrass theorem.
Example 10. “If f (x) attains its global minimum and maximum on [a, b],
then it is continuous on this interval.”
Solution.
The counterexamples proposed in Example 7 (including Remarks 1 and 2)
work here.
Remark. The converse statement is true and represents the second Weier-
strass theorem.
3.4.3 Intermediate Value theorem
General Remarks. The examples in this subsection aim to analyze different

attempts to relax, generalize and convert the conditions of the Intermediate
Value theorem. Example 11 stresses the importance of the continuity over all
the considered set. A failure to be continuous at only one point invalidates the
theorem result. Example 12 shows the importance of connectedness of the set:
the theorem is not true even on the union of two closed bounded intervals.
The next Example examines if the known corollary to the Intermediate Value
theorem remains true when the condition of continuity is omitted. Finally,
Examples 14-16 analyze attempts to construct the converse statement to the
theorem. It happens, that simple converse is very far to be true (Example
14), and even together with the additional conditions of extreme properties
the converse fails to assure a continuity (Remark 2 to Example 14). Other
versions of supposed converses fail to guarantee continuity even in a single
point (Examples 15 and 16).
Example 11. “If f (x) is defined on [a, b] and continuous on (a, b), then for
any constant C between f (a) and f (b) there is c ∈ (a, b) such that f (c) = C.”
Continuity 75
Solution.
The function f (x) = sgn x considered on [0, 1] is continuous on (0, 1), and
consequently satisfies the conditions. However, for C = 1/2, 0 = f (0) < C <
f (1) = 1, there is no point in (0, 1) (neither in [0, 1]) such that f (c) = C.
Remark. Of course, continuity on (a, b) can be substituted by continuity
on [a, b) or (a, b] with the same effect. For instance, for (a, b] = (0, 1] the above
function provides a counterexample.
Example 12. “If f (x) is continuous on [a, b] ∪ [α, β], where b < α, then
for any constant C between f (a) and f (β) there is a point c ∈ [a, b] ∪ [α, β]
such that f (c) = C.”
Solution.
Let us consider f (x) = sgn x on the set S ≡ [−2, −1] ∪ [1, 2]. Evidently,
this function is continuous on S, but for C = 0, −1 = f (−2) < C < f (2) = 1,
there is no point in S such that f (c) = C.
Remark 1. The values f (a) and f (β) can be substituted for any other
pair with the same effect.
Remark 2. The closed intervals can be substituted for any other kind
of intervals (and more generally by non-connected set). For instance, for the
set (−2, 1) \ {0} = (−2, 0) ∪ (0, 1) the same functions and C = 0 provide a
counterexample.

Remark to Examples 11 and 12. These statements are wrong modifica-

tions of the following important global property called the Intermediate Value
theorem: if f (x) is continuous on [a, b], then f (x) satisfies the intermediate
value property on [a, b] (that is, if f (x) takes on two values somewhere on
[a, b], it also takes on every value in between). In more general form this the-
orem states: if f (x) is continuous on a connected set, then its image is also
a connected set (recall that the connected sets on R are intervals, that is the
result holds for [a, b], [a, b), (a, b] and (a, b), where a and b can be finite or
infinite).
Example 13. “Between two local minima there is a local maximum.”

Solution.
The function f (x) = sec2 x has infinitely many local minima on R and no
local maximum.
Remark 1. Another simple counterexample with a bounded function is
f (x) = x − [x] on R.
Remark 2. The statement is true for a continuous function.
Example 14. “If f (x) satisfies the intermediate value property on [a, b]
(that is, if f (x) takes on two values somewhere on [a, b], it also takes on every
value in between), then f (x) is continuous on [a, b].”
Solution. {
sin x1 , x ̸= 0
The function f (x) = is continuous at any point different
0,x=0
from 0 (as a composition of continuous functions), but it is not continuous
at 0, because the limit as x approaches 0 does not exist (see Example 4,
section 2.2 for details). So, this function is not continuous on [0, 1], but it
does satisfy the intermediate value property on [0, 1]. The latter is evident
for (0, 1], because f (x) is continuous there, and therefore the intermediate
value property is satisfied. For any pair 0 = x1 < x2 ≤ 1 the property is also
satisfied, because on any interval (0, x2 ) there are points x̄k = kπ 1
, k ∈ N such
that f (x̄k ) = f (0) = 0 and x̄k ∈ (0, x2 ) for sufficiently large k. Again, on the
interval [x̄k , x2 ] the function f (x) is continuous and consequently it assumes
all the values between f (x̄k ) and f (x2 ), or equivalently, between f (0) and
f (x2 ).
Remark 1. It is possible to construct a function with the intermediate value
property, which has infinitely many points of discontinuity on a finite interval.
The above counterexample can be { used as a base for construction of such a
π
sin x−a , x ̸= a
function. First notice that fa (x) = keeps the properties of
0,x=a
the above function f (x) on [a, a + δ] for ∀δ > 0. Then construct a collection of
{ π
sin x−1/(n+1) , x ̸= n+1
1
such functions fn (x) = 1 each of which is defined on
0 , x = n+1
[ ]
its own domain Xn = n+1 1
, n1 , ∀n ∈ N. Finally notice that ∪ Xn = (0, 1]
n∈N
Continuity 77
and define the resulting function on (0, 1] as follows: f˜ (x) = fn (x), x ∈ Xn ,

∀n ∈ N. Using the reasoning for the function f (x) it is straightforward to
show that f˜ (x) satisfies the intermediate value property on (0, 1]. Indeed, for
an arbitrary pair of points 0 < x̃1 < x̃2 ≤ 1 there are two possibilities: one
of the points xn = n+1 1
, n ∈ N lies in [x̃1 , x̃2 ) or none of these points lies in
this interval. In the former case, the image of f˜ (x) is [−1, 1] already on the
interval (xn , x̃2 ); and in the latter case f˜ (x) is continuous on [x̃1 , x̃2 ]. So in
both cases f˜ (x) satisfies the intermediate value property. At the same time,
f˜ (x) is not continuous at the points xn = n+1 1
, ∀n ∈ N.
Remark 2. The same counterexample works even for the strengthened
false statement: “if f (x) satisfies the intermediate value property on [a, b] and
attains its global minimum and maximum on [a, b], then f (x) is continuous
on this interval”. This is actually a strengthened version for this Example and
for Example 10 as well, with the conditions including the conclusions of the
Weierstrass and Intermediate Value theorems. It shows that these theorems
are not convertible: even when a function satisfies the resulting properties of
both theorems, it is not sufficient to guarantee the function continuity. It is
necessary to add some additional condition to arrive to correct “converse”
result. One of the versions of such “converse” theorems is: if f (x) satisfies the
intermediate value property on [a, b] and if additionally f (x) takes on each
value only finitely many times, then f (x) is continuous on [a, b]. Another
version is as follows: if f (x) is monotone on an interval I and the image of I
is also an interval, then f (x) is continuous on I.

Example 15. “If f (x) attains its global minimum and maximum on
[a, b] and the image of this interval is [f (a) , f (b)], then f (x) is continuous on
[a, b].”
Solution {
x, x ∈ [0, 1] ∪ [2, 3]
The function f (x) = attains the minimum (f (0) =
3 − x , x ∈ (1, 2)
0) and maximum (f (3) = 3) at the endpoints of the domain [0, 3]. It also
assumes any value between 0 and 3, because the image of [0, 1], defined by
the function f (x) = x, is [0, 1], the image of (1, 2), defined by the function
f (x) = 3−x, is (1, 2), and the image of [2, 3], defined by the function f (x) = x,
is [2, 3]. So the total image is the interval [0, 3] = [f (0) , f (3)]. However, f (x)
has a jump discontinuity at the points 1 and 2: lim f (x) = 1, lim f (x) = 2,
x→1− x→1+
lim f (x) = 1, lim f (x) = 2.
x→2− x→2+
Remark. Another example of a function, which satisfies
{ the above con-
sin πx , x ̸= 0
ditions and has an essential discontinuity, is f (x) = on the
0,x=0
interval [−2, 2].
Example 16. “If f (x) attains its global minimum A and maximum B
on [a, b] and its image is [A, B], then f (x) is continuous at least at one point
on [a, b]. ”
Continuity 79
Solution 
 x, x ∈ Q, x ̸= 0, x ̸= 1
Let us consider f (x) = −x, x ∈ I on [−1, 1]. Evidently,

1, x = 0 ; 0, x = 1
all function values are located between −1 and 1, this function achieves its
global minimum (f (−1) = −1) and maximum (f (0) = 1) and it assumes
every value in [−1, 1], that is its image is the interval [−1, 1]. Therefore, the
conditions of the statement are satisfied. However, there is no point in [−1, 1]
where function is continuous. In fact, in any neighborhood of a point x0 there
are rational points different from 0 and 1 (denote them Q1 ≡ Q\ {0, 1}) and ir-
rational points. Let us consider the two partial limits at x0 : lim f (x) =
x→x0 , x∈Q1
lim x = x0 and lim f (x) = lim (−x) = −x0 . Since these par-
x→x0 , x∈Q1 x→x0 , x∈I x→x0 , x∈I
tial limits are different for any x0 ̸= 0, it follows that the limit of f (x) does
not exist at any x0 ̸= 0, and therefore, f (x) is not continuous at any x0 ̸= 0.
Finally, at the point x0 = 0 the limit exists lim f (x) = 0, but it is different
x→0
from the function value f (0) = 1. Therefore,f (x) is not continuous at x0 = 0
also. Hence, we have exhausted all possibilities and there is no point in [−1, 1]
where f (x) is continuous.
Remark 1 to Examples 15 and 16. The last two Examples are strengthened
versions of Example 10, with the conditions including the conclusions of the
Weierstrass and Compact Set theorems. It shows that these theorems are
not convertible: even if a function satisfies the resulting properties of both
theorems, this is not sufficient to guarantee the function continuity, even at a
single point.
Remark 2 to Examples 15 and 16. Note that the conditions on images
in these Examples do not mean that a function satisfies the intermediate
value property, because these conditions require only that the function should
assume all the values between the minimum and maximum on the entire in-
terval, but there is no requirement for its subintervals, which is important in
the intermediate value property.
3.5 Uniform continuity
Example 1. “If f (x) is continuous and bounded on (a, b), then f (x) is
uniformly continuous on (a, b).”
Solution.
The function f (x) = sin πx is bounded and continuous on (0, 1) (as a
composition of continuous functions), but it does not possess uniform con-
tinuity on (0, 1). Indeed, let us consider two sequences of points in (0, 1):
xn = n1 , ∀n ∈ N, n > 1 and x̃n = 1+4n 2
, ∀n ∈ N. For sufficiently large n
the distance between the corresponding
points of the two sets gets as close as
1 2
we wish: |xn − x̃n | = n − 1+4n → 0, but |f (xn ) − f (x̃n )| = |0 − 1| = 1,
n→+∞
∀n ∈ N, n > 1. It means that for ε < 1 whatever δ > 0 we choose, there exist
pairs of points xn , x̃n ∈ (0, 1) with the distance between them less than δ,
such that the difference between the corresponding function values is greater
than ε. Therefore, we have a contradiction to the definition of uniform conti-
nuity.
Remark 1. Another interesting example for an infinite interval is the func-
tion f (x) = sin x2 continuous but not uniformly continuous on R.
Remark 2. If f (x) is continuous on a compact set, in particular, on [a, b],
then f (x) is uniformly continuous on [a, b] according to the Cantor theorem.
Example 2. “If f (x) is unbounded on R, then f (x) is not uniformly

continuous on R.”
Solution.
Consider the function f (x) = x, which is unbounded on R. According to
the definition of uniform continuity, for ∀ε > 0 we can choose δ = ε such
that for any pair of points x1 , x2 such that |x1 − x2 | < δ it follows that
|f (x1 ) − f (x2 )| = |x1 − x2 | < ε . Hence, the definition is satisfied.
Remark 1. The conditions of the Cantor theorem guarantee, in particular,
Continuity 81
that the function is bounded on the considered compact set. However, it is

shown above that the boundedness of a function is not a necessary condition
for its uniform continuity.
Remark 2. For a finite interval (a, b) this statement is true.
Example 3. “If f (x) is uniformly continuous on (a, c) and (c, b), then
f (x) is uniformly continuous on (a, c) ∪ (c, b).”
Solution. {
x − 1, x ∈ (−1, 0)
Let f (x) = . This function is uniformly continuous
x + 1, x ∈ (0, 1)
on (−1, 0) and (0, 1), but it does not keep uniform continuity on (−1, 0)∪(0, 1).
In fact, for (0, 1) we get: for an arbitrary pair x1 , x2 such that |x1 − x2 | < δ it
follows that |f (x1 ) − f (x2 )| = |x1 − x2 | < δ = ε, that is the definition holds.
The same is true for (−1, 0). However, for (−1, 0) ∪ (0, 1), if x1 ∈ (0, 1) and
x2 ∈ (−1, 0), then we obtain |f (x1 ) − f (x2 )| = |2 + x1 − x2 | > 2 whatever
close points x1 , x2 are chosen.
Remark 1. A similar statement for continuous functions is also false (see
Example 6 in section 3.3).
Remark 2. The statement is also false for a pair of intervals (a, c] and
(c, b), and for a pair of intervals (a, c) and [c, b). For instance, to construct a
counterexample for the first pair it is sufficient to define the above function
additionally at the point 0 as follows: f (0) = −1. For these cases, a similar
statement for continuous functions is also false.
Remark 3. For a pair of intervals (a, c] and [c, b), and also for closed
intervals [a, c] and [c, b], a similar statement is correct both for uniformly
continuous and continuous functions.
Example 4. “If f (x) is uniformly continuous on R, then f 2 (x) is also

uniformly continuous on R.”
Solution:
Evidently, the function f (x) = x is uniformly continuous on R (see Ex-
ample 2 for details), but f 2 (x) = x2 is not, since for the two sequences
xn = n + n1 , ∀n ∈ N and x̃n = n , ∀n ∈ N the distance between the cor-
responding points of the two sets
gets as close aswe wish for sufficiently large
n: |xn − x̃n | = n1 → 0, but f 2 (xn ) − f 2 (x̃n ) = 2 + n12 > 2, ∀n ∈ N.
n→+∞
Example 5. “If f (x) and g (x) are uniformly continuous on a set S, then
f (x) · g (x) is also uniformly continuous on S.”
Solution.
Evidently, the function f (x) = x is uniformly continuous on R (see Ex-
ample 2 for details). The uniform continuity of g (x) = sin x on R is also of an
Continuity 83
elementary proof:

x1 − x2 x1 + x2
|sin x1 − sin x2 | = 2 sin cos ≤ |x1 − x2 | < δ = ε,
2 2
for ∀x1 , x2 ∈ R such that |x1 − x2 | < δ (here, we used two elementary prop-
erties of the trigonometric functions: |sin α| ≤ |α|, ∀α and |cos β| ≤ 1, ∀β).
However, for the function h (x) = f (x) · g (x) = x sin x we can choose xn =
2nπ + n1 , ∀n ∈ N and x̃n = 2nπ , ∀n ∈ N such that |xn − x̃n | = n1 → 0,
n→+∞
but
( ) ( )
1 1 1 21 2

|h (xn )−h (x̃n )| = 2nπ+ sin > 2nπ+ = 4+ 2 > 4, ∀n ∈ N
n n n πn πn
( )
(here the trigonometric inequality sin x > π2 x for x ∈ 0, π2 was applied). It
means that h (x) is not uniformly continuous on R.
Remark 1. If S is a bounded domain, then the statement is true.
Remark 2. The sum of uniformly continuous functions on a set S is uni-
formly continuous function on S.
Example 6. “If f (x) is uniformly continuous and f (x) ̸= 0 on a set S,

1
then f (x) is also uniformly continuous on S.”
Solution.
The function f (x) = x satisfies the conditions of the statement on (0, 1):
f (x) ̸= 0 for x ∈ (0, 1) and it is uniformly continuous on R (see the proof
1
in Example 2). However, the function g (x) = f (x) = x1 is not uniformly
continuous on (0, 1). In fact, choosing the points in (0, 1) in the form xn =
1 1
n , x̃n = n+1 , ∀n ∈ N\ {1}, we obtain: |xn − x̃n | = n − n+1 → 0, but
1 1
n→∞
|g (xn ) − g (x̃n )| = |n − (n + 1)| = 1.
Remark to Examples 4-6. Notice that analogous properties for continuous

functions are true.
Example 7. “If f (x) and g (x) are not uniformly continuous on a set S,
then f (x) · g (x) also is not uniformly continuous on S.”
Solution.
{ Let us consider a slight modification to the function of { Example 3: f (x) =
x − 1, x ∈ (−1, 0) −1, x ∈ (−1, 0)
, and also the function g (x) = .
x + 1, x ∈ [0, 1) 1, x ∈ [0, 1)
These functions are uniformly continuous on (−1, 0) and [0, 1), but they are
not uniformly continuous on (−1, 0) ∪ [0, 1) = (−1, 1), because both have a
jump discontinuity at the point 0. At the same time,
{
−x + 1, x ∈ (−1, 0)
h (x) = f (x) · g (x) = = 1 + |x|
x + 1, x ∈ [0, 1)
is a uniformly continuous function on (−1, 1). In fact, for any pair x1 , x2 ∈

(−1, 1) such that |x1 − x2 | < δ it follows that
|h (x1 ) − h (x2 )| = |1 + |x1 | − 1 − |x2 || ≤ |x1 − x2 | < δ = ε.
Example 8. “If f (x) is continuous on X, and g(x) is uniformly continuous

on f (X), then the composite function g(f (x)) is uniformly continuous on X.”
Solution.
Consider the functions f (x) = x1 defined on X = (0, +∞) and g(x) = sin x
defined on f (X) = (0, +∞). The first function is not uniformly continuous
on X, since for xn = n1 and x̃n = n+1 1
, ∀n ∈ N one has |xn − x̃n | = | n1 −
n+1 | → 0, while |f (xn ) − f (x̃n )| = |n − (n + 1)| = 1 > 0. The second
1
n→∞
function is uniformly continuous on R (see Example 5 in this section) and,
consequently, it is uniformly continuous on (0, +∞). At the same time, the
composite function h(x) = g(f (x)) = sin x1 is not uniformly continuous on
(0, +∞), because for xk = 2kπ 1 2
and x̃k = (4k+1)π , one gets |xk − x̃k | =
| 2kπ − (4k+1)π | → 0, but |h(xk ) − h(x̃k )| = |1 − 0| > 0, ∀k ∈ N.
1 2
k→∞
Remark 1. The statement with the opposite conclusion is also false. Indeed,
the function f (x) = x1 defined on X = (0, 1) is non-uniformly continuous (just
choose the same sequences of points xn = n1 and x̃n = n+11
as above), and the
Continuity 85
function g(x) = x1 defined on X̃ = (1, +∞) is uniformly continuous, since for

∀ε > 0 and any pair of points x1 , x2 ∈ (1, +∞) such that |x1 − x2 | < δ = ε it
follows that

1 1 |x1 − x2 |

|g (x1 ) − g (x2 )| = − = < ε.
x1 x2 |x1 | · |x2 |
At the same time, the composite function h(x) = g(f (x)) = x is uniformly
continuous on R and, therefore, on (0, 1).
Remark 2. The change of order of a non-uniform continuous and uniformly
continuous functions in the composite function leads to the same results.
Remark 3. The well-known result on uniform continuity of composite func-
tion states that if f (x) is uniformly continuous on X, and g(x) is uniformly
continuous on f (X), then the composite function g(f (x)) is uniformly contin-
uous on X. If one tries to weaken the conditions on one of the functions, then
the result may be wrong.
Remark 4. Of course, the composite function is continuous on X as com-
position of continuous functions.
Exercises
1. Show that the following statement is false: “if f (x) is continuous at 0

and f (0) = 0, then the inequality |f (x)| ≤ |x|α holds for some α > 0 at least
in some neighborhood of 0”.
2. Provide a counterexample to the statement “if f (x) + g(x) is continu-
ous at a, then both f (x) and g(x) are continuous at a”. What about other
arithmetic operations?
3. Give a counterexample to the statement: “if cos f (x) is continuous at a,
then f (x) is also continuous”. What about cosh f (x)?
4. Verify if the following statement is true: “if g(f (x)) is continuous at a,
then f (x) is also continuous”. What if the condition of continuity of g(y) at
f (a) is added? (Hint: in the last case try f (x) = sin x1 {and g(y) = 0 at a = 0;
0 , y ∈ [−1, 1]
to not involve constant functions one can use g(y) = ;
y 2 − 1 , otherwise
still another counterexample is f (x) = D̃(x) and g(y) = y 2 at an arbitrary
point a.)
5. Verify if the following statement is true: “if g(f (x)) and f (x) are con-
tinuous at a, then g(y) is continuous at f (a)”.
6. Suppose f (x) and g(x) are such that both f (g(x)) and g(f (x)) are
defined in a neighborhood of a and f (g(x)) is continuous at a. Is g(f (x)) also
continuous?
7. Give a counterexample to the statement that “a continuous on [a, b]
function can not possess infinitely many local extrema on this interval”. (This
is a weakened version of the statement in Example 2, section 3.3.)
8. Demonstrate that the counterexamples suggested in Remark to Example
2, section 3.3, work.
9. Provide a counterexample to the following statement: “if f (x) is in-
creasing and continuous on the interval [c, a] and on the interval (a, b], then
it is increasing on [c, b]”.
10. Give a counterexample to the statement: “a function defined on R
cannot be continuous at each point x ∈ Z and discontinuous at all other
points”.
11. It is known that the convexity condition on an open interval (a, b) -
f (αx1 + (1 − α)x2 ) ≤ αf (x1 ) + (1 − α)f (x2 ), ∀x1 , x2 ∈ (a, b) and ∀α ∈ (0, 1)
- implies continuity of f (x) on (a, b). Show that{this implication fails in the
0 , x ∈ (−1, 1)
case of a closed interval. (Hint: consider f (x) = )
1 , x = 0, x = 1
12. It is known that any function f (x) that maps continuously a closed
interval on itself has a fixed point, that is, such point c in the closed interval
that f (c) = c. Show that this is not true for an open interval. (Hint: consider
f (x) = x2 on (0, 1).)
Continuity 87
13. Provide a counterexample to the following statement: “if f (x) is con-

tinuous and invertible on domain X, then the inverse{ function f −1 (y) is con-
x, x ≤ 0
tinuous on Y = f (X)”. (Hint: consider f (x) = )
x − 1, x > 1
14. Provide a counterexample to the following statement: “if f (x) is
continuous on X ⊂ R, X ̸= R, then there is a continuous extension of
f (x) from X onto R, that is, such function g(x) continuous on R, that
g(x) = f (x), ∀x ∈ X”. (Hint: use f (x) = sin x1 , X = (0, +∞); notice that the
statement is true for a closed set X.)
15. Demonstrate that the function f (x) = x sin x + x + x1 proposed in
Remark 1 to Example 10, section 3.3 is a counterexample to the corresponding
statement.
16. Provide counterexamples for Remark 2 to Example 10 in section 3.3.
17. Give a counterexample to the statement in Example 1 in section 3.4
such that an open interval is transformed onto the closed interval under con-
tinuous mapping.
18. Show that the statement “if f (x) is continuous and has a compact
image, then its domain is a bounded set” is false. (This is a strengthened
version of the statement in Example 3, section 3.4.)
19. Show that the following statement is false: “if f (x) is continuous and
bounded on a closed set, then it attains its global minimum and maximum on
this set”.
20. Provide a counterexample to the following statement: “if f (x) is con-
tinuous on [a, b] except for only one point, then f (x) is bounded on [a, b]”.
21. Give a counterexample to the statement: “if f (x) is defined, but not
continuous on [a, b], it cannot attain its global extrema on this interval”.
22. Give a counterexample to the statement in Example 11, section 3.4
with interval (a, b) substituted by [a, b).
23. According to the Intermediate Value theorem, any connected set has
a connected image under continuous mapping. Verify if the same is true in
the “inverse” direction: “if f (x) is continuous on R, then any connected set in
f (X) has a connected inverse image”. If not, construct a counterexample.
24. Verify if the following “definition” of the uniform continuity is correct:
“for every ε > 0 and every δ > 0 whenever x1 , x2 ∈ S and |x1 − x2 | < δ it
follows that |f (x1 ) − f (x2 )| < ε”. If not, provide a counterexample.
25. Demonstrate that the function sin x2 in Remark 1 to Example 1 of
section 3.4 is a counterexample to the statement in that Example.
26. Show that the statement “if f (x) is uniformly continuous and g(x)
is not uniformly continuous on X, then h(x) = f (x) · g(x) is not uniformly
continuous on X” is false by providing a counterexample. Even a strengthened
version with non-continuous function g(x) is false. (Hint: try f (x) = |x| and
g(x) = sgn x on X = R.)
27. Provide counterexamples to the statements mentioned in Remark 2 to
Example 8 in section 3.4, that is, to the following statements: “if f (x) is uni-
formly continuous on X, and g(x) is continuous on f (X), then the composite
function g(f (x)) is uniformly continuous on X” and “if f (x) is uniformly con-
tinuous on X, and g(x) is continuous on f (X), then the composite function
g(f (x)) is not uniformly continuous on X”.
28. Show that the following statement is false: “if f (x) is uniformly con-
tinuous on any interval [a, b], then it is uniformly continuous on R”.
29. Since the violation of uniform continuity for one of the functions in
the composition can lead to loss of uniform continuity by the composite func-
tion, it may be tempting to state that “if f (x) is non-uniformly continuous on
X, and g(x) is non-uniformly continuous on f (X), then the composite func-
tion g(f (x)) is non-uniformly continuous on X”. Show that it is not true by
providing a counterexample.
30. A function f (x) is called Lipschitz-continuous on an interval I if there
exists a constant C ≥ 0 such that |f (x) − f (y)| ≤ C|x − y| for ∀x, y ∈ I.
It is straightforward to see that the Lipschitz-continuity implies the uniform
continuity. What about the converse?
Chapter 4
Differentiation
Concepts
Differentiability at a point. A function f (x) defined on X is called

differentiable at a point a ∈ X, if for any point x in a neighborhood of a the
following representation holds: f (x) = f (a) + A · (x − a) + α · (x − a), where
A is a constant and α is a function of x such that α → 0 .
x→a
Remark 1. The same definition is frequently written in the terms of the
increments of the argument ∆x = x − a and the function ∆f = f (x) − f (a)
as follows: ∆f = A · ∆x + α · ∆x, where A is a constant and α is a function
of ∆x such that α → 0 .
∆x→0
Differential. The expression A · ∆x ≡ df is called the differential of f (x)
at a point a, and the difference γ = ∆f − df is called the remainder. Note
that the term df is the linear (with respect to ∆x) part of ∆f , and this is
also the main part of ∆f in a small neighborhood of a if A ̸= 0. Thus, the
differential represents a linear approximation to the function increment in a
neighborhood of a.
Derivative at a point. Given a function f (x) defined on a neighborhood
of a point a, the derivative of f (x) at a (denoted f ′ (a) or dx
df
(a) ) is defined
by f ′ (a) = lim f (x)−f
x−a
(a)
provided this limit exists and is finite.
x→a
Theorem. A function f (x) is differentiable at a point a if, and only if,
there exists f ′ (a). Moreover, in the definition of differentiability A = f ′ (a).
Remark 1. Sometimes the equality A = f ′ (a) is considered to be a part
of the definition of differentiability.
Remark 2. Due to the last Theorem, a function possessing a derivative at
a is also called differentiable at a.
89
One-sided derivatives. Given a function f (x) defined at a point a

and on a right-hand (left-hand) neighborhood of a, the right-hand (left-hand)
′ ′
derivative of f (x) at a is defined by f+ (a) = lim f (x)−f
x−a
(a)
(f− (a) =
x→a+
f (x)−f (a)
lim x−a ) provided this limit exists and is finite.
x→a−
Differentiation on a set. If f ′ (x) is defined at every point of a set S,

then f (x) is differentiable on S.
Remark . It is worth to notice that in the case when endpoints belong to
the interval, for example a closed interval [a, b], the differentiability on [a, b]
means the differentiability at every point c ∈ (a, b) plus the existence of the
right-hand derivative at a and the left-hand derivative at b.
Continuous differentiability. If the derivative f ′ (x) is continuous on

a set S, then f (x) is continuously differentiable on S.
Higher-order derivatives. If the derivative g (x) = f ′ (x) is defined

on a neighborhood of a point a, and the derivative of g (x) exists at a, then
it is called the second derivative of f (x) at a and the standard notation is
2
f ′′ (a) = g ′ (a) ( or ddxf2 (a) ). In a similar way, consecutively the n-th derivative
of f (x) at a is defined as the derivative of the (n − 1)-th derivative of f (x)
n
at a, and the standard notation is f (n) (a) or ddxnf (a).
Tangent line. Let f (x) be continuous in a neighborhood of a point a,

and b be a point in such a neighborhood. The secant line through the points
Pa = (a, f (a)) and Pb = (b, f (b)) is defined by the formula: y − f (a) =
f (b)−f (a)
b−a (x − a). The tangent line is the limit position of the secant lines (if it
exists) when Pb tends to Pa (in other words, when b approaches a). Since each
secant line is characterized but its slope f (b)−f
b−a
(a)
, to find the limit position
of secants is equivalent to find the limit of their slopes. If this limit exists
and is finite, then lim f (b)−f
b−a
(a)
= f ′ (a) and the equation of the tangent is
b→a
y − f (a) = f ′ (a) (x − a). If lim f (b)−f (a)
b−a = (±) ∞, then the equation of
b→a
f (b)−f (a)
the tangent is x = a. Sometimes the case when lim b−a = +∞ and
b→a−
f (b)−f (a)
lim b−a = −∞ (or vice-versa) is considered to be a different situation
b→a+
called cusp or corner.
Differentiation 91
Basic properties
Theorem. If f (x) is differentiable on a set, then it is continuous on this

set.

If f (x) and g (x) are differentiable at a point x, then
′
1) f (x) + g (x) is differentiable at x and (f (x) + g (x)) = f ′ (x) + g ′ (x);
′
2) f (x) − g (x) is differentiable at x and (f (x) − g (x)) = f ′ (x) − g ′ (x);
′
3) f (x) · g (x) is differentiable at x and (f (x) · g (x)) = f ′ (x) · g (x) + f (x) ·
g ′ (x);
( )′ ′
(x)·g ′ (x)
4) fg(x)
(x)
is differentiable at x and fg(x)
(x)
= f (x)·g(x)−f
g 2 (x) (under the
additional condition g (x) ̸= 0).
Chain rule. If f (x) is differentiable at a point x, and g (y) is differentiable

at the point y = f (x), then the composite function g (f (x)) is differentiable
′
at the point x and (g (f (x))) = g ′ (y) · f ′ (x).
Rolle’s Theorem. If f (x) is continuous on [a, b] and differentiable on

(a, b) with f (a) = f (b), then there exists a point c ∈ (a, b) where f ′ (c) = 0.
The Mean Value Theorem (Lagrange’s Theorem). If f (x) is con-
tinuous on [a, b] and differentiable on (a, b), then there exists a point c ∈ (a, b)
where f ′ (c) = f (b)−f
b−a
(a)
.
Darboux’s Theorem. If f (x) is differentiable on a connected set S, then

f ′ (x) satisfies the intermediate value property on S. (In particular, f ′ (S) is
also a connected set.)
Theorem. If f (x) is differentiable on a connected set S, then f ′ (x)
cannot have removable or jump discontinuities on S. (However, f ′ (x) can
have discontinuities of the second kind.)
Inverse Function Theorem.

The first formulation. If y = f (x) is strictly monotone and continuous
in a neighborhood of x0 , and there exists f ′ (x0 ) ̸= 0, then in a neighbor-
hood of the point y0 = f (x0 ) there exists the inverse function x = f −1 (y)
−1
( )−1
differentiable at y0 and such that dfdy (y0 ) = dxdf
(x0 ) .
The second formulation. If y = f (x) is continuously differentiable in
a neighborhood of x0 and f ′ (x0 ) ̸= 0, then in a neighborhood of the point
y0 = f (x0 ) there exists the inverse function x = f −1 (y) differentiable at y0
−1
( )−1
and such that dfdy (y0 ) = dx df
(x0 ) .
Applications
Criterion of a constant function. A differentiable on (a, b) function

f (x) is constant on (a, b) if, and only if, f ′ (x) = 0, ∀x ∈ (a, b).
Criterion of monotonicity. A differentiable on (a, b) function f (x)
is increasing (decreasing) on (a, b) if, and only if, f ′ (x) ≥ 0 (f ′ (x) ≤ 0),
∀x ∈ (a, b).
Sufficient condition of strict monotonicity. A differentiable on (a, b)
function f (x) is strictly increasing (decreasing) on (a, b) if f ′ (x) > 0 (f ′ (x) <
0), ∀x ∈ (a, b).
Critical (stationary) point. If f ′ (c) = 0, then c is called a critical

(stationary) point.
Fermat’s Theorem. If f (x) is differentiable at a local extremum point
c, then f ′ (c) = 0 .
The first derivative test for local extremum. Let f (x) be differen-
tiable in a neighborhood of a point c. Then
1) if f ′ (x) > 0 in a left-hand neighborhood of c and f ′ (x) < 0 in a right-hand
neighborhood of c, then c is a strict local maximum;
2) if f ′ (x) < 0 in a left-hand neighborhood of c and f ′ (x) > 0 in a right-hand
neighborhood of c, then c is a strict local minimum;
3) if f ′ (x) < 0 (f ′ (x) > 0) in both one-sided neighborhoods of c, then c is
not a local extremum.
The second derivative test for local extremum. Let f (x) be twice
differentiable at a point c and f ′ (c) = 0. Then
1) if f ′′ (c) < 0, then c is a strict local maximum;
2) if f ′′ (c) > 0, then c is a strict local minimum.
Upward concavity. A function f (x) defined on (a, b) is upward con-

cave (strictly upward concave) on (a, b) if the inequality f (α1 x1 + α2 x2 ) ≤
α1 f (x1 ) + α2 f (x2 ) (f (α1 x1 + α2 x2 ) < α1 f (x1 ) + α2 f (x2 ) ) holds for any
points x1 , x2 ∈ (a, b) and arbitrary parameters α1 , α2 ≥ 0, α1 + α2 = 1.
Another common term is convex (strictly convex) function.
Geometrically it means that the line segment joining (x1 , f (x1 )) and
(x2 , f (x2 )) for ∀x1 , x2 ∈ (a, b) lies above the graph of f (x).
Downward concavity. A function f (x) defined on (a, b) is down-
ward concave (strictly downward concave) on (a, b) if the inequality
f (α1 x1 + α2 x2 ) ≥ α1 f (x1 ) + α2 f (x2 ) (f (α1 x1 + α2 x2 ) > α1 f (x1 ) +
α2 f (x2 )) holds for any points x1 , x2 ∈ (a, b) and arbitrary parameters
α1 , α2 ≥ 0, α1 + α2 = 1.
Geometrically it means that the line segment joining (x1 , f (x1 )) and
(x2 , f (x2 )) for ∀x1 , x2 ∈ (a, b) lies below the graph of f (x).
Inflection point. Let f (x) be continuous in a neighborhood of a point
Differentiation 93
a. If f (x) is concave upward (downward) in a left-hand neighborhood of a

and concave downward (upward) in a right-hand neighborhood of a, then the
point a is called the inflection point.
Theorem. If f (x) is concave on (a, b), then it is continuous on (a, b).
Theorem. Let f (x) be differentiable on (a, b). f (x) is strictly upward

(downward) concave on (a, b) if, and only if, the graph of f (x) lies above
(below) each tangent line on (a, b) except at the point of contact.
Theorem (criterion of concavity). Let f (x) be twice differentiable on
(a, b). f (x) is upward (downward) concave on (a, b) if, and only if, f ′′ (x) ≥ 0
(f ′′ (x) ≤ 0) on (a, b).
Theorem (sufficient condition of concavity). Let f (x) be twice
differentiable on (a, b). If f ′′ (x) > 0 (f ′′ (x) < 0) on (a, b), thenf (x) is strictly
upward (downward) concave on (a, b).
Theorem. If f (x) is twice differentiable at an inflection point c, then

f ′′ (c) = 0.
Theorem. Let f (x) be twice differentiable in a neighborhood of a point
c. Then
1) if f ′′ (x) > 0 in a left-hand neighborhood of c and f ′′ (x) < 0 in a right-
hand neighborhood of c (or f ′′ (x) < 0 in a left-hand neighborhood of c and
f ′′ (x) > 0 in a right-hand neighborhood of c), then c is an inflection point;
2) if f ′′ (x) keeps the same sign in both one-sided neighborhoods of c, then c
is not an inflection point.
Asymptotes
The line y = A is a horizontal asymptote of f (x) if either lim f (x) = A
x→−∞
or lim f (x) = A, or both.
x→+∞
The line y = Ax + B is an oblique (slant) asymptote of f (x) if either
lim (f (x) − Ax − B) = 0 or lim (f (x) − Ax − B) = 0, or both.
x→−∞ x→+∞
The line x = a is a vertical asymptote if lim f (x) = (±) ∞ or
x→a−
lim f (x) = (±) ∞, or both.
x→a+
4.2 Concepts
Example 1. “If f (x) is not differentiable at a point a, then it is not

Solution.
The function f (x) = |x| is continuous at a = 0, but it is not differentiable

there. In fact, lim f (x) = lim x = 0 and lim f (x) = lim (−x) = 0, so
x→0+ x→0+ x→0− x→0−
lim f (x) = 0 = f (0). At the same time, lim f (x)−f
x−0
(0)
= lim x
= 1, while
x→0 x→0 + x→0+ x
f (x)−f (0) −x
lim x−0 = lim = −1. Since the one-sided derivatives are different,
x→0− x→0− x
the derivative at a = 0 does not exist.
Remark 1. This is the false converse to the theorem stating that if f (x)
is differentiable at a then it is continuous at this point.
Remark 2. There is the famous Weierstrass function continuous on R and
nowhere differentiable, as well as other examples of this kind. However, the
study of such functions requires the knowledge of the uniform convergence of
series of functions, which is out of the scope of this book.
Example 2. “If f (x) is differentiable in both right-hand and left-hand

neighborhoods of a point a and lim f ′ (x) = lim f ′ (x), then f (x) is differ-
x→a+ x→a−
entiable at a.”
Solution.
For any point a ̸= 0 the function f (x) = sgn x has zero derivative: for
every x < 0, f (x) = −1 and so f ′ (x) = 0; for every x > 0, f (x) = 1 and
then f ′ (x) = 0. Therefore, lim f ′ (x) = lim f ′ (x) = 0. However, f (x) is
x→0+ x→0−
not differentiable at a = 0 because it is not even continuous at this point.
Remark 1. The same counterexample works for a strengthened (but still
false) version of this statement: “if f (x) is differentiable at every x ̸= a and
lim f ′ (x) = A, then f (x) is differentiable at a”.
x→a
Remark 2. The following counterexample shows√another kind of non-
differentiability for the original statement: f (x) = 3 x is differentiable at
every point a ̸= 0, and lim f ′ (x) = lim √ 1
3 2 = +∞. Moreover, the function
x→0 x→0 3 x
is continuous at a = 0, and nevertheless the derivative does not exist at a = 0.
Notice that this counterexample does not work for the statement in Remark
1, where the finite limit is required.
Remark 3. If we put aside the case of infinite limits in the original state-
ment, then both statements (original and in Remark 1) will be true if we add
the condition of continuity of f (x) at a.
Remark 4. A “negative” version is true and applicable in practice: if f (x)
is differentiable in both right-hand and left-hand neighborhoods of a point a
and lim f ′ (x) ̸= lim f ′ (x), then f (x) is not differentiable at a.
x→a+ x→a−
Example 3. “If f (x) is differentiable on a neighborhood of a point a, then

its derivative is continuous at this point.”
Solution. { 2
x sin x1 , x ̸= 0
The function f (x) = is differentiable at any x and its
0, x = 0
Differentiation 95
{
2x sin x1 − cos x1 , x ̸= 0
derivative is f ′ (x) = . Indeed, for any x ̸= 0 the
0, x = 0
derivative is readily found by applying the arithmetic and chain rules, while
for x = 0 the derivative can be calculated
by appealing to the definition where
f (x)−f (0) 1

we first evaluate the expression x−0 = x sin x ≤ |x| for x ̸= 0 and then
letting x → 0 we get f ′ (0) = 0. However, the function f ′ (x) is discontinuous
at zero: lim 2x sin x1 = 0, but lim cos x1 does not exist (it can be shown in the
x→0 x→0
same way as for sin x1 in Example 4, section 2.2).
Remark . Darboux’s theorem specifies what types of discontinuities are
admissible by a derivative function. It states that if f (x) is differentiable on
(a, b), then f ′ (x) cannot have a removable or jump discontinuity on (a, b).
(But f ′ (x) still may have discontinuities of the second kind like in the shown
counterexample.)
FIGURE 4.2.1: Example 3, graph of the function
Example 4. “If a function f (x) is not differentiable at a point a, then

f 2 (x) also is not differentiable at a.”
Solution.
The function f (x) = |x| is not differentiable at a = 0 (see Example 1 in
this section), but f 2 (x) = x2 is differentiable on R.
Remark 2. This Example can be easily extended to a set of points in
the following form: “if a function f (x) is not differentiable at any point in
FIGURE 4.2.2: Example 3, graph of the derivative
(a, b), then f 2 (x) is not differentiable at least at one point in (a, b)”. This
statement can be disproved
{ by the counterexample with the modified Dirichlet
1, x ∈ Q
function D̃ (x) = , which has no limit (and therefore is not
−1 , x ∈ I
differentiable) at any point (see Example 2, section 2.2), but its square is the
constant function D̃2 (x) = 1 differentiable on R. {
x, x ∈ Q
Another simple counterexample is the function f (x) = that
−x , x ∈ I
is not differentiable at any x ∈ R. In fact, it was shown in Example 3, section
3.3, that f (x) is discontinuous at any x ̸= 0, so it is not differentiable at any
x ̸= 0, and for x = 0 two partial limits in the definition of derivative give
different results:
f (x) − f (0) x
lim = lim =1
x→0,x∈Q x−0 x→0,x∈Q x
and
f (x) − f (0) −x
lim = lim = −1,
x→0,x∈I x−0 x→0,x∈I x
so the derivative does not exist at x = 0. However, f 2 (x) = x2 is differentiable

on R.
Example 5. “If a function f (x) is not differentiable at a point a, then

|f (x)| also is not differentiable at this point.”
Differentiation 97
FIGURE 4.2.3: Example 4, Remark 2 and Example 5, the function f (x)
Solution.
The same
{ functions in Remark 2 to the
{previous Example 4 work here: both
1, x ∈ Q x, x ∈ Q
D̃ (x) = and f (x) = are not differentiable
−1 , x ∈I −x , x ∈ I

at any x ∈ R, but D̃ (x) = 1 is differentiable on R and |f (x)| = |x| is
differentiable on R\ {0}.
Remark 1. The following extended statement is also false: “if a function
f (x) is not differentiable at any point in (a, b), then |f (x)| is not differen-
tiable at least at one point in (a, b)”. The same functions can be used for the
counterexamples: the function D̃ (x) on R, and the function f (x) on (0, +∞).
Remark 2. The converse is also false. The corresponding example is f (x) =
x, which is differentiable on R, but |f (x)| = |x| is not differentiable at a = 0 .
Remark 3. It is worth to note that relation between f (x) and |f (x)| in
terms of differentiability is different from that for continuous functions. For
continuous functions the continuity of f (x) implies the continuity of |f (x)|,
while the converse implication does not hold. For differentiability both impli-
cations are not true.
Example 6. “If a function f (x) is infinitely differentiable on R and f (0) =

f ′ (0) = f ′′ (0) = . . . = 0, that is, f (n) (0) = 0 , ∀n ∈ N ∪ {0}, then f (x) ≡ 0
on R.”
Solution.
{ −1
e x2 , x ̸= 0
Let us consider the function f (x) = . This function is
0, x = 0
infinitely differentiable at every point x ̸= 0 according to the arithmetic and
chain rules. At the point x = 0 we use the definition:
e− x 2
1
t 1
f ′ (0) = lim = lim t2 = lim 2 = 0
x→0 x t→∞ e t→∞ 2tet
(applying the substitution x = 1t and L’Hospital’s rule). Therefore, we

{
2 − x12
′ x3 e , x ̸= 0
have f (x) = . Similarly, for the second derivative we get
0, x = 0
{ ( )
− x64 e− x2 + x23 e− x2 , x ̸= 0
1 2 1
′′
f (x) = . Generalizing, it can be shown
0, x = 0
{ ( )
P3n x1 e− x2 , x ̸= 0
1
by mathematical induction that f (n) (x) = , ∀n ∈ N,

0, x = 0
where P3n (t) is a polynomial of degree 3n. We have already deduced this
formula for n = 1 and n = 2. Let us suppose now that the formula is true for
n = k and prove that it holds for n = k + 1. At every point x ̸= 0 we have
( ( ) )′
1
e− x 2
1
f (k+1) (x) = P3k
x
( ) ( ) ( )
1 − x12 2 1 − x12 1
e− x 2 .
1
= P3k e · 3 + P3k+1 e = P3(k+1)
x x x x
For the point x = 0, applying the substitution x = 1t and L’Hospital’s rule,
we obtain:
( )
P3k x1 e− x2
1
(k+1) f (k) (x)−f (k) (0) P3k+1 (t)

f (0) = lim = lim = lim = 0.
x→0 x−0 x→0 x t→±∞ e t2
Therefore, the formula for the derivatives is proved. Hence, f (x) satisfied the
conditions of the statement, but f (x) is equal to zero only at the origin.
Example 7. “If lim f (x+h)−f (x−h)

2h = A at a point x, then f ′ (x) = A.”
h→0
Solution.
The function f (x) = |x| satisfies the above condition at x = 0, but it does
not differentiable at this point.
Remark 1. Actually for x = 0 any even function non-differentiable at
x = 0 provides a counterexample to this statement. For instance, one can
choose even Dirichlet’s function, which does not have a limit at any point.
Remark 2. If the differentiability of f (x) at x is added to the condition,
then the statement becomes true.
Differentiation 99
Example 8. “If lim f (x+h)−2f (x)+f (x−h)

h2 = A at a point x, then f ′′ (x) =
h→0
A.”
Solution.
The function f (x) = sgn x satisfies the above condition at x = 0, but it
does not even have a limit at this point.
Remark 1. Actually for x = 0 any odd function discontinuous at the origin
provides a counterexample to this statement.
Remark 2. If the second derivative f ′′ (x) exists, then the statement be-
comes true.
Example 9. “If f (x) is continuous in a neighborhood of a point a and is

differentiable at a, then there is a neighborhood of a where f (x) is differen-
tiable.”
Solution. { 2
x sin πx , x ̸= 0
The function f (x) = is continuous at x ̸= 0 accord-
0, x = 0
ing to the composition and arithmetic properties of continuous functions, and
at x = 0 one gets lim f (x) = 0 = f (0). Therefore, f (x) is continuous on R.
x→0
The definition of this function can be rewritten as follows:
 2
 x sin x , 2n+1 < x < 2n , n ∈ Z
π 1 1
f (x) = −x2 sin x , 2n < x < 2n−1

π 1 1
, n∈Z .

0 , x = 0, x = n , n ∈ Z\ {0}
1
Then for x ̸= 0 and x ̸= n1 the derivative can be found by applying the

arithmetic and chain rules:
{
′ 2x sin πx − π cos πx , 2n+1
1 1
< x < 2n , n∈Z
f (x) = .
−2x sin x + π cos x , 2n < x < 2n−1
π π 1 1
, n∈Z
Let us show that the derivative does not exist at any point xn = 1
2n ,n ∈
Z\ {0}. In fact,
( π π)
lim f ′ (x) = lim 2x sin − π cos = −π,
x→xn− x→xn− x x
while ( π π)
lim f ′ (x) = lim −2x sin + π cos = π.
x→xn+ x→xn+ x x
Note that xn = 1
→ 0,
2n n→±∞ but, at the same time the function is differentiable
at x = 0:
f (x) − f (0) π
f ′ (0) = lim = lim x sin = 0.
x→0 x−0 x→0 x
4.3 Local properties
Example 1. “If h (x) = f (x) + g (x) is differentiable at a point a, then

both f (x) and g (x) are differentiable at a.”
Solution.
The functions f (x) = |x| and g (x) = − |x| are not differentiable at the
point a = 0 (see Example 1, section 4.2), but their sum h (x) = f (x)+g (x) = 0
is differentiable at 0.
Remark 1. Similar statements and counterexamples can be given for other
arithmetic operations. For example, for the ratio the corresponding counterex-
ample is f (x) = (|x| + 1) cos x and g (x) = |x| + 1. At the point a = 0 the
one-sided derivatives of f (x) are different:
′ (x + 1) cos x − 1 x cos x − 2 sin2 x/2

f+ (0) = lim = lim =1
x→0+ x x→0+ x
and
′ (−x + 1) cos x − 1 −x cos x − 2 sin2 x/2

f− (0) = lim = lim = −1.
x→0− x x→0− x
Differentiation 101
The situation is similar for the one-sided derivatives of g (x):
′ (x + 1) − 1 ′ (−x + 1) − 1
g+ (0) = lim = 1 and g− (0) = lim = −1.
x→0+ x x→0− x
Hence, the functions f (x) and g (x) are non-differentiable at a = 0. At the

same time, the ratio h (x) = fg(x)
(x)
= cos x is differentiable on R.
Remark 2. The reverse is true (it is one of the arithmetic rules of differ-
entiation).
Example 2. “If h (x) = f (x) · g (x) is differentiable at a point a and f (x)

is differentiable at the same point, then g (x) is also differentiable at a.”
Solution. √
The function f (x) = x is differentiable at a = 0 and g (x) =√ 3 x is
3
not differentiable at a = 0 (the corresponding limit is infinite lim x−0 x =
1
√3
x→0
lim √ 4
3 2 = +∞), but their product h (x) = f (x)·g (x) = x is differentiable
x→0 x
√
3 4 √
at 0, because lim xx −0 = lim 3 x = 0.
x→0 x→0
Remark 1. The above counterexample shows also that the following state-
ment is not correct: “if f (x) is differentiable at a point a, and g (x) is non-
differentiable at a, then f (x) · g (x) is non-differentiable at a”. On the other
hand, the statement: “if f (x) is differentiable at a point a, and g (x) is non-
differentiable at a, then f (x) · g (x) is differentiable at a” is not true either.
The counterexample is f (x) = 1 and g (x) = |x|, a = 0. It means that the
situation under the conditions of the statements is indeterminate.
Remark 2. Similar statement can be given for the ratio and the corre-
sponding counterexample can be f (x) = (|x| + 1) sin x and g (x) = |x| + 1 at
a = 0.
Remark 3. For the sum or difference the corresponding statement is true:
if h (x) = f (x) ± g (x) is differentiable at a point a and f (x) is differentiable
at the same point, then g (x) is also differentiable at a. Or, equivalently, if
f (x) is differentiable at a point a, and g (x) is non-differentiable at a, then
f (x) ± g (x) is non-differentiable at a.
Example 3. “If f (x) is not differentiable at a point a, and g (x) is differ-

entiable at the point f (a), then g (f (x)) is not differentiable at a.”
Solution.
The function f (x) = |x| is not differentiable at a = 0 and g (x) = x2 is
differentiable at f (0) = 0. However, g (f (x)) = x2 is differentiable at a = 0.
Remark 1. If we interchange the definitions of the functions - setting
f (x) = x2 and g (x) = |x| - and keep the same a = 0, then we obtain the
counterexample to a similar statement: “if f (x) is differentiable at a point a,
and g (x) is not differentiable at the point f (a), then g (f (x)) is not differen-
tiable at a”.
Remark 2. Of course, the opposite conclusion is also not correct: if f (x) is
not differentiable at a point a, and g (x) is differentiable at the point f (a), it

may happen that g (f (x)) is not differentiable at a. For example, f (x) = |x|
and g (x) = x, a = 0. It means that the situation under the conditions of the
statement is indeterminate.
Remark 3. This is a false “negative” version of the chain rule.
Example 4. “If f (x) is not differentiable at a point a, and g (x) is not

differentiable at the point f (a), then g (f (x)) is not differentiable at a.”
Solution. {
1/x, x ̸= 0
The function f (x) = g (x) = is not differentiable at a = 0
0, x = 0
(it has a discontinuity of the second kind at 0), but the function g (f (x)) = x
is differentiable at 0.
Remark 1. Another interesting example {involves two functions non-
x, x ∈ Q
differentiable at any point: f (x) = −g (x) = . Their compo-
−x , x ∈ I
sition g (f (x)) = −x is differentiable on R.
Remark 2. Of course, it may happen that under the above conditions the
function g (f (x)) will not be differentiable at a, for example, f (x) = |x| and
g (x) = |x|, a = 0. It means that the situation under the conditions of the
statement is indeterminate.
Example 5. “If f (x) is differentiable on R, then f ′ (x) is bounded in a

neighborhood of each point.”
Solution. { 2 /
x sin 1 x2 , x ̸= 0
The function f (x) = is differentiable on R. Indeed,
0, x = 0
for any x ̸= 0, applying the arithmetic and chain rules, we obtain f ′ (x) =
2x sin x12 − x2 cos x12 , and at x = 0, appealing to the definition, we get
/
′ x2 sin 1 x2 − 0 1
f (0) = lim = lim x sin 2 = 0
x→0 x−0 x→0 x

(the last {result follows from the evaluation x sin x12 ≤ |x|, x ̸= 0). Hence,
2x sin x12 − x2 cos x12 , x ̸= 0
f ′ (x) = . Now, choosing xk = √2kπ 1
, k ∈N
0, x = 0 ( √ )
we obtain lim f ′ (xk ) = lim −2 2kπ cos 2kπ = −∞, that is, f ′ (x) is
xk →0 k→+∞
unbounded in any neighborhood of 0.
Example 6. “There is no function differentiable only at one point of its

domain.”
Solution. { 2
x ,x∈Q
The function f (x) = is not differentiable at any point a ̸= 0.
0, x ∈ I
In fact, this function is not even continuous at any a ̸= 0: lim f (x) =
x→a, x∈Q

lim x2 = a2 ̸= 0 and lim f (x) = lim 0 = 0, that is, two partial

x→a, x∈Q x→a, x∈I x→a,x∈I
limits are different. However, f (x) is differentiable at the point a = 0, because
lim f (x)−f
x−0
(0)
= lim x = 0 and lim f (x)−f x−0
(0)
= lim 0 = 0,
x→0, x∈Q x→0, x∈Q x→0, x∈I x→0,x∈I
that is, two partial limits coincide. Since these partial limits involve all real
numbers in a neighborhood of 0, it follows that ∃f ′ (0) = lim f (x)−f
x−0
(0)
= 0.
x→0
FIGURE 4.3.3: Example 6 and Example 7
Example 7. “If f (x) is differentiable at a point a, then it is continuous

at least in some neighborhood of a.”
Solution.
The function of the previous Example 6 works here.
Example 8. “If there is a tangent line to the graph of f (x) at a point c,

then f (x) is differentiable at c.”
Solution. √
The function f (x) = 3 x has a vertical tangent line at the point c = 0,
but it is not differentiable
√
at this point. In fact, the derivative is infinite:
f (x)−f (0) 3 x
lim x−0 = lim x = +∞, so f (x) is not differentiable at c = 0, but
x→0 x→0
x = 0 is a vertical tangent line to the graph of f (x).
Example 9. “If f (x) is not differentiable at x0 , then either the inverse

function does not exist in a neighborhood of the point y0 = f (x0 ) or (if it

exists) it is not differentiable at y0 .”
Solution. √
The function y = f (x) = 3 x is strictly monotone and continuous in
a neighborhood of x0 = 0, but its derivative does not exists at the origin.
However, the inverse function x = f −1 (y) = y 3 is defined and differentiable
on R, including at y0 = 0.
Remark . This example shows that the conditions of the Inverse Function
Theorem are just sufficient.
Example 10. “If f (x) is differentiable on R, f ′ (x) is bounded in a neigh-

borhood of a point x0 and f ′ (x0 ) ̸= 0, then in a neighborhood of the point
y0 = f (x0 ) there exists the inverse function x = f −1 (y).”
Solution. {
x + 2x2 sin x1 , x ̸= 0
The function f (x) = is differentiable at every
0, x=0
real point. In fact, for any x ̸= 0 the derivative can be calculated by using
the arithmetic and chain rules: f ′ (x) = 1 + 4x sin x1 − 2 cos x1 . For x = 0, we
should appeal to definition:
( )
x + 2x2 sin 1/x 1
f ′ (0) = lim = lim 1 + 2x sin = 1.
x→0 x x→0 x
Hence,
{
′ 1 + 4x sin x1 − 2 cos x1 , x ̸= 0
f (x) = and f ′ (0) ̸= 0.
1, x=0
Besides, f ′ (x) is bounded on [−1, 1]:

1 + 4x sin 1 − 2 cos 1 ≤ 1 + 4 |x| + 2 ≤ 7.
x x
Thus, all the conditions of the statement are satisfied. Nevertheless, one can
show that there is no neighborhood of x0 = 0 on which f (x) is a one-to-one
correspondence. Let us consider, for simplicity, a right-hand neighborhood of
0. On such a neighborhood sin x1 takes all the values in [−1, 1], and therefore,
f (x) oscillates between the parabolas x − 2x2 and x + 2x2 . The apex of the
parabola x−2x2 is at the point x = 41 . Consider ∀δ > 0, δ < 14 and δ1 = πδ < δ.
Let us find a point x1 such that f (x1 ) = x1 − 2x21 and x1 < δ1 . From the
first condition, sin x11 = −1, that gives xk = 2kπ−π/2 1
, k ∈ N, and the second
condition imposes restriction on k: 2k − 2 > δ , that is, k > 41 + 2δ
1 1 1
. Let us
choose some fixed k1 , which satisfies the last condition, and the corresponding
1
x1 = 2k1 π−π/2 ∈ (0, δ1 ) ⊂ (0, δ). Note that the function value at the auxiliary
1
point x2 = 2k1 π+π/2 is larger than at x1 :
( π) ( π)
f (x2 ) − f (x1 ) = x2 + 2x22 sin 2k1 π + − x1 − 2x21 sin 2k1 π −
2 2
1 2 1 2
= + − +
2k1 π + π/2 (2k1 π + π/2)2 2k1 π − π/2 (2k1 π − π/2)2
/
(4 − π) · 4k1 π 2 + (4 + π) · π 2 4
= 2 2 > 0.
(2k1 π + π/2) (2k1 π − π/2)
Consider one more point x3 = 2(k1 +1)π−π/21
< x1 < 41 . Since 2x − x2 is
( 1)
increasing on 0, 4 , it follows that f (x3 ) < f (x1 ). Hence, we have the three
points x3 < x2 < x1 < δ such that f (x3 ) < f (x1 ) < f (x2 ) . Applying the
Intermediate Value Theorem to the continuous function f (x) on the interval
[x3 , x2 ], we conclude that for ∀A such that f (x3 ) < A < f (x2 ), in particular,
for A = f (x1 ), there exists a point xA ∈ [x3 , x2 ] at which f (xA ) = A = f (x1 ).
Thus, f (x) takes the same value f (xA ) = f (x1 ) at two different points of the
interval (0, δ), that is, f (x) is not injective on (0, δ). Since it holds for ∀δ > 0,
the function f (x) is not one-to-one in any neighborhood of x0 = 0 and an
inverse does not exist in a neighborhood of the origin.
Remark. This false statement is a weakened version of the Inverse Function
Theorem where the condition of the continuity of the derivative is replaced
by the weaker condition of its boundedness. In fact, the derivative of f (x) is
discontinuous at 0: for xn = 2nπ+π 1
→ 0 (n ∈ N) one has f ′ (xn ) = 1 + 2 =
n→∞
3 ̸= 1 = f ′ (0).
4.4 Global properties
Example 1. “If both functions f (x) and g (x) are differentiable on R and
f (x) > g (x) on R, then there exists at least one point in R where f ′ (x) >
g ′ (x).”
Solution.
Evidently, the functions f (x) = e−x and g (x) = −e−x are differentiable
on R and f (x) > g (x) on R, but f ′ (x) = −e−x < e−x = g ′ (x) on R.
Example 2. “If a function is differentiable and monotone on R, then its

derivative is also monotone on R.”
Solution.
The function f (x) = 2x + sin x is differentiable and strictly increasing on
R (f ′ (x) = 2 + cos x > 0, ∀x ∈ R), but its derivative is not monotone on R.
Remark . The following variation of the statement is also false: “if a
function is differentiable and non-monotone on R then its derivative is also
non-monotone on R”. It can be disproved by the counterexample f (x) = x2 .
Example 3. “If f (x) is differentiable on [a, b] then f ′ (x) is bounded on

[a, b].”
Solution.
This is a “global” version of the statement in Example 5, sec-
tion 4.3, and we can {use the same function for a counterexample.
x2 sin x12 , x ̸= 0
The function f (x) = has the derivative f ′ (x) =
0, x = 0
{
2x sin x12 − x2 cos x12 , x ̸= 0
(see Example 5, section 4.3 for details). How-
0, x = 0
ever, f ′ (x) is unbounded on [−1, 1], because lim 2x sin x12 = 0, and for
x→0
xk = √1
2kπ
, k ∈ N it follows
2 1 √
lim cos 2 = lim 2 2kπ cos 2kπ = +∞,
xk →0 xk xk k→+∞
so lim f ′ (xk ) = −∞.

xk →0
{
x4/3 cos x1 , x ̸= 0
Another similar counterexample is f (x) = consid-
0, x = 0
ered on [−1, 1]. This function is differentiable on [−1, 1]:
{ 4 1/3
′ 3x cos x1 + x−2/3 sin x1 , x ̸= 0
f (x) = .
0, x = 0
2
However, the derivative is not bounded, since for xk = (1+4k)π , k ∈ N one
gets
( )2/3
−2/3 1 (1 + 4k) π
lim x sin = lim = +∞,
xk →0 k xk k→+∞ 2
so lim f ′ (xk ) = +∞.
xk →0
Example 4. “If f (x) is continuous on [a, b], has a tangent line at each
point of its graph on (a, b) and f (a) = f (b), then at least one of these tangent
lines is horizontal.”
Solution. √
The function f (x) = x2 is continuous on [−a, a], ∀a > 0, satisfies the
3
condition f (−a) = f (a), and has the derivative f ′ (x) = 3 √ 3 x ̸= 0 for every
2
x ̸= 0. Therefore, a tangent line at any point (x, f (x)), x ∈ [−a, 0) ∪ (0, a],
exists, but it is √
not horizontal. At the point x = 0 the derivative does not
3 2
exist, since lim xx = ∞, but the tangent line at this point exists, although
x→0
it is vertical and not horizontal.
Remark . This statement is a wrong weakened geometric version of Rolle’s
theorem: if f (x) is continuous on [a, b], differentiable on (a, b) and f (a) = f (b)
then there is a point c ∈ (a, b) such that f ′ (c) = 0. (It implies that the tangent
line at the point (c, f (c)) is horizontal.)

Example 5. “If f (x) is bounded on [a, b], differentiable on (a, b) and

f (a) = f (b) = 0, then its derivative is zero at some point in (a, b). ”
Solution.
The function f (x) = x − [x] is bounded on [0, 1], differentiable on (0, 1)
(f (x) = x on (0, 1) ) and f (0) = f (1) = 0, but f ′ (x) = 1 for every x ∈ (0, 1).
Remark . This is a wrong version of Rolle’s theorem with dropped condi-
tion of the continuity of f (x) on [a, b].
Example 6. “If f (x) is continuous on [a, b] and differentiable on (a, b),

then for any c ∈ (a, b) there are two points x1 , x2 ∈ (a, b) such that f ′ (c) =
f (x2 )−f (x1 )
x2 −x1 .”
Solution.
The function f (x) = x3 is continuous on [−1, 1] and differentiable on
(−1, 1). However, for c = 0 there is no pair x1 , x2 ∈ (−1, 1) such that f ′ (0) =
0 = f (xx22)−f
−x1
(x1 )
, because f (x) = x3 is strictly increasing on [−1, 1] .
Remark 1. This is a wrong reformulation of the Medium Value theorem:
if f (x) is continuous on [a, b] and differentiable on (a, b), then for any x1 , x2 ∈
[a, b] there is a point c ∈ (a, b) such that f ′ (c) = f (xx22)−f
−x1
(x1 )
. (It implies that
the tangent line at the point (c, f (c)) is parallel to the line passing through
the points (x1 , f (x1 )) and (x2 , f (x2 )).)
Remark 2. The above statement can be also considered as a wrong ex-
tension of the following theorem: if f (x) is convex on [a, b] and differentiable
on (a, b), then for any c ∈ (a, b) there are two points x1 , x2 ∈ (a, b) such that
f ′ (c) = f (xx22)−f
−x1
(x1 )
.
Example 7. “If a function f (x) is differentiable on (a, b) and f ′ (x) is

unbounded on (a, b), then f (x) is also unbounded on (a, b).”
Solution. √
The function f (x) = 3 x is differentiable on (0, 1) and its derivative
f ′ (x) = √ 1
3 is unbounded on (0, 1), but the function values are in (0, 1).
3 x2
Remark. This is the false converse to the following correct statement:
if a function f (x) is differentiable on a bounded interval (a, b) and f (x) is
unbounded on (a, b), then f ′ (x) is also unbounded on (a, b). If (a, b) is un-
bounded, then the last statement is false. The corresponding counterexample
is: f (x) = x on R.
Example 8. “If a function f (x) is differentiable on a finite interval (a, b)

and lim f (x) = ∞, then lim f ′ (x) = ∞.”
x→a+ x→a+
Solution.
1
The function f (x) = x + cos x1 is differentiable on (0, 1) and lim f (x) =
x→0+
+∞. However, the derivative f ′ (x) = − x12 + x12 sin x1 does not have a limit
2
(finite or infinite) when x approaches 0. Indeed, for xn = (4n+1)π , n ∈ N we
have lim f ′ (xn ) = lim 0 = 0, but for xk = 2

(4k−1)π , k ∈ N it follows that
xn →0+
(n→∞ )
′
lim f (xk ) = lim − x22 = −∞.
xk →0+ k→∞ k
Example 9. “If f (x) is differentiable on (a, +∞) and both the function
and its derivative are bounded on (a, +∞), then the existence of lim f (x)
x→+∞
implies the existence of lim f ′ (x), and vice-versa.”
x→+∞
Solution.
sin x2 2
First, consider f (x) = x on (1, +∞). Since sinxx ≤ 1
x for x > 1, it
2
′ sin x2
follows that lim sin x = 0. However, f (x) = 2 cos x − 2
x2 has no limit at
x→+∞ x
infinity. It is sufficient√to compare two partial limits: the first corresponding
to the sequence xn = 2nπ, n ∈ N gives
(√ )
lim f ′ (xn ) = lim f ′ 2nπ = lim 2 = 2,
xn →+∞ n→+∞ n→+∞
√
and the second related to xk = π + 2kπ, k ∈ N results in
(√ )
lim f ′ (xk ) = lim f ′ π + 2kπ = lim (−2) = −2.
xk →+∞ k→+∞ k→+∞
Since two partial limits are different, the limit lim f ′ (x) does not exist.
x→+∞
Second, consider f (x) = cos (ln x) on (1, +∞). The derivative
f ′ (x)
=
sin(ln x) ′ sin(ln x)
− x has zero limit at infinity lim f (x) = 0, because − x ≤ x1
x→+∞
for x > 1. However, the function has no limit at infinity, since for the partial
limit with the points xn = e2nπ , n ∈ N we have
( )
lim f (xn ) = lim f e2nπ = lim 1 = 1,
xn →+∞ n→+∞ n→+∞
while for xk = eπ+2kπ , k ∈ N the result is different

( )
lim f (xk ) = lim f eπ+2kπ = lim (−1) = −1.
xk →+∞ k→+∞ k→+∞
Remark 1. Other interesting counterexamples are f (x) = e−x cos ex √ for

the case when lim f (x) exists, but lim f ′ (x) does not, and f (x) = sin 3 x
x→+∞ x→+∞
for the opposite case.
Remark 2. It is simple to modify the above examples in such a way that the
function or its derivative will be unbounded. For example, the function f (x) =
e−x cos e2x is bounded on (0, +∞) and lim f (x) = 0, but its derivative
x→+∞
f ′ (x) = −e−x cos e2x − 2ex sin e2x has no limit at infinity and is unbounded:
for xn = 12 ln (2nπ) , n ∈ N one gets
( ) 1
lim f ′ (xn ) = lim −e−xn cos e2xn = − lim √ = 0,
xn →+∞ n→+∞ n→+∞ 2nπ
( )
while for xk = 12 ln π2 + 2kπ , k ∈ N one obtains
√
′
( ) π
lim f (xk ) = lim −2e sin e xk 2xk
= − lim 2 + 2kπ = −∞.
xk →+∞ k→+∞ k→+∞ 2
√ √
In the opposite direction, the function f (x) = 3 x sin 3 x has no limit at
( )3
infinity and is unbounded: choosing xn = π2 + 2nπ , n ∈ N one gets
(π )
lim f (xn ) = lim + 2nπ = +∞.
xn →+∞ n→+∞ 2
√ √
However, its derivative f ′ (x) = √ 3
1 1
sin 3 x + 3 √
3 x cos
3
x is bounded and
3 x2
has limit at infinity equal to zero.
FIGURE 4.4.3: Example 9, graph of the first function
Example 10. “If f ′ (x) = 0 on a set S then f (x) = const on this set.”
Solution. {
1, x ∈ (0, 1)
Let f (x) = with the set S = (0, 1) ∪ (1, 2). Evidently,
2, x ∈ (1, 2)
′
f (x) = 0 at every point x ∈ S, but f (x) is not a constant function.
Remark . The statement will be correct if the set S is connected (that is,
S is an interval).
4.5 Applications
4.5.1 Tangent line
Example 1. “A line is tangent to a curve at a point if it touches, but does

not cross the curve at that point.”
Solution.
The tangent to the graph of the function f (x) = x3 at the origin is the
x-axis, which crosses the graph at the same point.
Example 2. “A line is tangent to a curve if it touches or intersects the

curve at exactly one point.”
Solution.
The tangent to the graph of the function f (x) = cos x at the point x = 0
is y = 1 and it touches the graph of cos x at all points xn = 2nπ, ∀n ∈ Z.
The tangent to the graph of the function f (x) = x sin x at the origin is
y = 0 and it crosses the graph of x sin x at all points xn = nπ, ∀n ∈ Z. At
the same time, at the point x = π2 this function has the tangent line y = x,
which touches the graph of f (x) at all points xk = π2 + 2kπ, ∀k ∈ Z and
also crosses the graph at the origin. The tangent line y = −x at the point
x = − π2 has similar properties: it touches the graph of f (x) at all points
xm = − π2 + 2mπ, ∀m ∈ Z and crosses the graph at the origin.
Example 3. “If a function f (x) is continuous on (a, b) and at some point

c ∈ (a, b) the tangent line to the graph of the function is vertical, then f (x)
does not have a local extremum at c.”
Solution. √
The function f (x) = x2 is continuous on R and
3
f (x) − f (0) 1
lim = lim √ = +∞,
x→0+ x−0 x→0+ 3
x
f (x) − f (0) 1
lim = lim √ = −∞,
x→0− x−0 x→0− 3
x
which means that x = 0 is the vertical tangent to the graph of f (x). At the
same time c = 0 is a strict local (and global) minimum of f (x).
Remark. Let us consider the statement with the opposite conclusion: “if
f (x) is continuous on (a, b) and at some point c ∈ (a, b) the tangent line to
the graph of the function is vertical, then f (x) has a local extremum √at c”.
This statement is also false and a simple counterexample is f (x) = 3 x at
c = 0.
FIGURE 4.5.1: Example 2, graph of the second function
Example 4. “If f (x) is continuous on (a, b) and at some point c ∈ (a, b)

a tangent line does not exist, then f (x) does not have a local extremum at
c.”
Solution.
The function f (x) = |x| does not have a tangent at c = 0, but this point
is a strict local (and global) minimum of this function.
Remark . Let us consider the statement with the opposite conclusion: “if
f (x) is continuous on (a, b) and at some point c ∈ (a, b) a tangent line does
not exist, then f (x) has a local extremum
{ at c”. This statement is also false
x, x < 0
and a simple counterexample is f (x) = at c = 0.
3x, x ≥ 0
4.5.2 Monotonicity and local extrema
Example 5. “If f (x) is strictly increasing and differentiable on (a, b), then
f ′ (x) > 0 on (a, b).”
Solution.
The function f (x) = x3 is strictly increasing and differentiable on (−1, 1),
but f ′ (0) = 0.
Remark 1. This is the wrong converse to the following statement: if

f ′ (x) > 0 on (a, b), then f (x) is strictly increasing on (a, b). This is also
the wrong extension of the statement: if f (x) is increasing and differentiable
on (a, b), then f ′ (x) ≥ 0 on (a, b).
Remark 2. Of course, the same is true for decreasing functions.
Example 6. “If f ′ (c) > 0, then f (x) is increasing in a neighborhood of

the point c.”
Solution. {
x + 2x2 sin x1 , x ̸= 0
0, x=0
{
1 + 4x sin x1 − 2 cos x1 , x ̸= 0
real point, and its derivative is f ′ (x) = and
1, x=0
′
f (0) > 0 (see Example 10, section 4.3 for details). However, in every neigh-
borhood of zero f ′ (x) has both positive and negative values: for the points
2
xk = π(4k+1) , k ∈ Z, we obtain f ′ (xk ) = 1 + π(4k+1)
8
> 0; while for the points
′
xn = 2nπ , n ∈ Z\ {0}, we obtain f (xn ) = 1 − 2 < 0. It means that f (x)
1
cannot be increasing in any neighborhood of the point c = 0.

Remark 1. Of course, the same is true for the relation between the condi-
tion f ′ (c) < 0 and decreasing of the function.
Remark 2. If one adds the condition of continuity of f ′ (x) at the point c,
then the statement becomes true.
Example 7. “If f ′ (c) = 0, then the point c is a local extremum of the

function f (x).”
Solution.
The function f (x) = x3 is differentiable at any real point and f ′ (0) = 0,
but c = 0 is not a local extremum, because f (x) = x3 is a strictly increasing
function on R.
Example 8. “If c is a local extremum of f (x), then f ′ (c) = 0 .”

Solution.
The function f (x) = |x| is not differentiable at c = 0, but this point is a
local (and global) minimum of this function. √
Remark . Another good example is the function f (x) = |x| at c = 0.
Remark to Examples 7 and 8. Each of these Examples is a wrong modifi-

cation of the necessary condition for a local extremum: if f (x) is differentiable
at a point c and c is a local extremum, then f ′ (c) = 0.
Example 9. “If c is a local extremum of f (x) and f (x) is twice differen-

tiable at c, then f ′′ (c) ̸= 0 .”
Solution.
The function f (x) = x4 is twice differentiable at c = 0 and the origin is a
local (and global) minimum of this function, but f ′′ (0) = 0.
Remark . This is the false converse to the Second Derivative test: if f (x)
is twice differentiable at a stationary point c and f ′′ (c) ̸= 0, then c is a local

extremum of f (x).
Example 10. “If the derivative f ′ (x) is defined in a neighborhood of a

point c, and it does not change the sign passing through this point, then c is
not a local extremum.”
Solution. {
2x2 + x2 cos x1 , x ̸= 0
0, x=0
real point. In fact, for any x ̸= 0 the derivative can be calculated by using the
arithmetic and chain rules: f ′ (x) = 4x + 2x cos x1 + sin x1 . For x = 0 applying
the definition we obtain
( )
2x2 + x2 cos 1/x 1
f ′ (0) = lim = lim 2x + x cos = 0.
x→0 x x→0 x
{
′ 4x + 2x cos x1 + sin x1 , x ̸= 0
Hence, f (x) = . Now, using the points xk =
0, x=0
′
π(4k+1) , k ∈ Z, we obtain f (xk ) = π(4k+1) + 1 > 0, k ∈ Z. At the same time,
2 8
for the points xn = π(4n−1) , n ∈ Z, we get f ′ (xn ) = π(4n−1)

2 8
− 1 < 0, n ∈ Z.
Note that in any neighborhood of the point c = 0 there exist both points xk
and xn (for k and n sufficiently large), that is, in any neighborhood of zero
there are both positive and negative values of f ′ (x). Therefore, the derivative
does not preserve the sign in one-sided neighborhoods of zero, so f ′ (x) does
not change its sign passing through the point ( c = 0. On
) the other hand, the
following evaluation is satisfied: f (x) = x2 2 + cos x1 > 0 = f (0), ∀x ̸= 0.
Therefore, by the definition, c = 0 is a strict local (and global) minimum of
f (x).
Remark 1. This is a warning to use with caution the following popular
“rule”: “if f ′ (x) exists in a neighborhood of a point c, but it does not change
sign passing through this point, then c is not a local extremum”. This “rule”
is a non precise shortened version of a part of the so-called First Derivative
test: if f ′ (x) exists in a neighborhood of a point c, and f ′ (x) keeps the same
sign on both sides of c, then c is not a local extremum.
Remark 2. An analogous false “rule”, also disproved by this counterex-
ample, is the following: “if f (x) is continuous in a neighborhood of a local
minimum point c, then f (x) is decreasing in a left-hand neighborhood of c and
increasing in a right-hand neighborhood of c” (and a similar “rule” for a local
maximum). This “rule” is a non precise shortened version of another part of
the First Derivative test: if f ′ (x) exists in a neighborhood of a point c, and
f ′ (x) changes sign passing through the point c, then c is a local extremum.
4.5.3 Convexity and inflection
Example 11. “If f (x) is strictly concave upward and twice differentiable
on (a, b), then f ′′ (x) > 0 on (a, b).”
Solution.
The function f (x) = x4 is strictly concave upward and twice differentiable
on (−1, 1), but f ′′ (0) = 0.
Remark 1. This is the wrong converse to the following statement: if
f ′′ (x) > 0 on (a, b), then f (x) is strictly concave upward on (a, b). This
is also the wrong extension of the statement: if f (x) is concave upward and
twice differentiable on (a, b), then f ′′ (x) ≥ 0 on (a, b).
Remark 2. Of course, the same is true for strictly concave downward
functions.
Example 12. “If f (x) is twice differentiable in a neighborhood of a point

c and f ′′ (c) = 0, then the point c is an inflection point.”
Solution.
The function f (x) = x4 is twice differentiable on R and f ′′ (0) = 0, but
c = 0 is not an inflection point, since f (x) is concave upward on R (f ′′ (x) =
12x2 > 0 for ∀x ̸= 0).
Example 13. “If c is an inflection point of f (x), then f ′′ (c) = 0 .”

Solution. {
−x2 , x < 0
The function f (x) = is not twice differentiable at c = 0
x4 , x ≥ 0
(f ′′ (x) = −2 for ∀x < 0, f ′′ (x) = 12x2 for ∀x > 0, so lim f ′′ (x) = −2 ̸=
x→0−
0 = lim f ′′ (x)), but this point is an inflection point for f (x), because the
x→0+
function has a downward concavity at negative points (f ′′ (x) = −2 < 0 for
∀x < 0) and an upward concavity at positive points (f ′′ (x) = 12x2 > 0 for
∀x > 0).
Remark to Examples 12 and 13. Each of these Examples is the wrong

converse to the necessary condition of an inflection point: if f (x) is twice
differentiable at a point c and the point c is an inflection point, then f ′′ (c) = 0.
Example 14. “If a function f (x) is twice differentiable in a deleted neigh-

borhood of a point c, and a tangent line at the point c lies above the graph
of f (x) in a left-hand neighborhood of c and below the graph of f (x) in a
right-hand neighborhood of c, then c is an inflection point.”
Solution. {
2x3 + x3 sin x1 , x ̸= 0
The function f (x) = is differentiable on R and
0, x=0
twice differentiable on R\ {0}. In fact, for any x ̸= 0 the first and second

FIGURE 4.5.8: Example 14, graph of the second derivative
derivatives can be calculated by using the arithmetic and chain rules:

1 1
f ′ (x) = 6x2 + 3x2 sin − x cos ,
x x
1 1 1 1
f ′′ (x) = 12x + 6x sin − 4 cos − sin .
x x x x
For x = 0, applying the definition, we obtain
( )
′ 2x3 + x3 sin 1/x 2 2 1
f (0) = lim = lim 2x + x sin = 0,
x→0 x x→0 x
2 2
but the second derivative does not exist, because lim 6x +3x sin 1/x−x
x
cos 1/x
( ) x→0
does not exist ( lim 6x + 3x sin x1 = 0 and lim cos x1 does not exist). Since
x→0 x→0
f ′ (0) = 0, (the tangent
) line at c = 0 exists and has (the form )y = 0. Besides,
f (x) = x3 2 + sin x1 < 0, ∀x < 0 and f (x) = x3 2 + sin x1 > 0, ∀x > 0,
which means that the tangent line at c = 0 (the x-axis) lies above the graph of
f (x) for x < 0 and below the graph for x > 0. At the same time, the second
derivative does not keep the sign in any one-sided neighborhood of c = 0.
Indeed, for xk = 2kπ 1
, k ∈ Z\ {0} we obtain f ′′ (xk ) = 2kπ
12
− 4 < 0, while
for xn = 2nπ+π , n ∈ Z it follows f ′′ (xn ) = 2nπ+π + 4 > 0. Therefore, f (x)
1 12
does not have any kind of concavity in any one-sided neighborhood of c = 0.

Hence, the point c = 0 is not an inflection point.
Remark . This is a warning to use with caution the following too simplified
“rule”: “if the tangent line at the point c lies above (below) the graph of f (x) in
a left-hand neighborhood of c and below (above) the graph of f (x) in a right-
hand neighborhood of c, then c is an inflection point”. Or even more simplified
“definition”: “c is called an inflection point of f (x) if the tangent line to the
graph of f (x) at (c, f (c)) crosses the graph”. These “rules” are non precise
shortened versions of the following theorem: if f (x) is twice differentiable in a
deleted neighborhood of a point c, and f ′′ (x) changes its sign passing through
the point c, then c is an inflection point.
Example 15. “If f (x) is continuous on (a, b) and is not differentiable at

a point c ∈ (a, b), then c is not an inflection point of the graph of f (x).”
Solution. { 2
x + x, x < 0
The function f (x) = is continuous on R and does not
2x − x2 , x ≥ 0
have a derivative at c = 0, since the one-sided derivatives are different:
′ f (x) − f (0)
f− (0) = lim = lim (x + 1) = 1,
x→0− x−0 x→0−
′ f (x) − f (0)
f+ (0) = lim = lim (2 − x) = 2.
x→0+ x−0 x→0+
At the same time, f ′′ (x) = 2 for x < 0, and f ′′ (x) = −2 for x > 0, that
is, the function is concave upward on (−∞, 0) and downward on (0, +∞).
Therefore, c = 0 is an inflection point.
Remark . Let us consider the statement with the opposite conclusion: “if
f (x) is continuous on (a, b) and is not differentiable at a point c ∈ (a, b), then
c is an inflection point of the graph {of f (x)”. This statement is also false and
x2 + x, x < 0
a simple counterexample is f (x) = at c = 0.
x2 − x, x ≥ 0
Example 16. “If f (x) is continuous in a neighborhood of a point c and

f ′ (c) = f ′′ (c) = 0, then the point c is either a local extremum or an inflection
point of f (x).”
Solution. { 4
x sin x1 , x ̸= 0
Let f (x) = . Notice that in any deleted neighbor-
0, x=0
hood of x = 0 there exist points xn = (4n+1)π 2
, ∀n ∈ Z where f (xn ) =
4
xn > 0 = f (0), and also there exist points xk = (4k−1)π 2
, ∀k ∈ Z such
that f (xk ) = −xk < 0 = f (0), that is, x = 0 is not a local extremum
4
of f (x). On the other hand, the derivative can be easily found using the
arithmetic and chain rules for x ̸= 0 and the definition of the derivative for
x = 0 (see Example 5, section { 4.3 and Example 14 in this section for analo-
′ 4x3 sin x1 − x2 cos x1 , x ̸= 0
gous calculations): f (x) = . Furthermore, the
0, x=0
FIGURE 4.5.11: Example 16, graphs of the first and second derivatives
second derivative { also 2exists and can be found by applying similar calcula-
12x sin 1
− 6x cos x1 − sin x1 , x ̸= 0
tions: f ′′ (x) = x . At the points xn =
0, x=0
′′
(4n+1)π , ∀n ∈ Z\ {0} the second derivative is negative f (xn ) = 12xn −1 < 0,
2 2
and at the points xk = (4k−1)π 2

, ∀k ∈ Z\ {0} the second derivative is positive
′′
f (xk ) = −12xk + 1 > 0. So the function has no concavity in any one-sided
2
neighborhood of x = 0. Therefore, x = 0 is not an inflection point of f (x).
4.5.4 Asymptotes
Example 17. “If y = c is a tangent line to the graph of f (x), then y = c

cannot be an asymptote to the same graph.”
Solution. {
sin πx, x < 1
The function f (x) = has the tangent line y = 1 at the
x+1 , x ≥ 1
x−1
( )
point x = 12 , since f ′ 12 = π cos π2 = 0, and the same line is the horizontal
asymptote, because lim f (x) = lim x−1 x+1 = 1.
x→+∞ x→+∞
Remark 1. The statement is also false
{ 1 for vertical tangents and asymptotes.
A simple counterexample is f (x) = x, x < 0
√ at the point x = 0.
3
x, x ≥ 0
Remark 2. The statement is true for vertical tangents and asymptotes of

continuous functions.
Example 18. “The graph of a function cannot cross or touch its asymp-
tote.”
Solution.
The function f (x) = sinx x has the horizontal asymptote y = 0 (since
lim f (x) = lim sinx x = 0) and the graph of f (x) crosses this asymptote
x→+∞ x→+∞
infinitely many times at the points xn = nπ , ∀n ∈ N.
Another function f (x) = e−x (1 − cos x) has the horizontal asymptote
y = 0 (since lim f (x) = lim e−x (1 − cos x) = 0) and the graph of f (x)
x→+∞ x→+∞
touches this asymptote infinitely many times at the points xn = 2nπ , ∀n ∈ N
(at all these points, y = 0 is the tangent line to the graph of f (x)).
Example 19. “If f (x) and g (x) are continuous in a neighborhood of a

point a, and g (a) = 0, then the function fg(x)
(x)
has a vertical asymptote at the
point a.”
Solution.
Although the functions f (x) = sin x and g (x) = x are continuous on R,
and g (x) = x equals 0 at the point x = 0, the ratio h (x) = fg(x)(x)
= sinx x has a
finite limit at 0: lim sinx x = 1. Hence, h (x) has no asymptote at x = 0.
x→0
FIGURE 4.5.13: Example 18, the first counterexample
FIGURE 4.5.14: Example 18, the second counterexample

Remark. The statement becomes true under the additional condition that
f (a) ̸= 0. Notice, however, that without the continuity of f (x) and g (x),
the last condition does not guarantee the existence{of a vertical asymptote. A
x + 1 , x ̸= 0
simple counterexample is: f (x) = x + 1, g (x) = , which at
0, x = 0
the point x = 0 gives lim fg(x)
(x)
= lim x+1
x+1 = 1.
x→0 x→0
Example 20. “If y = Ax + B is a slant asymptote to f (x), then the slope

of the graph of f (x) approaches A, as x approaches infinity.”
Solution.
3
The
function
f (x) = sinxx has the horizontal asymptote y = 0, since
3 3
0 ≤ sinxx ≤ |x|
1
→ 0, and consequently, lim sinxx = 0. However, its
x→+∞ x→+∞
3
derivative f ′ (x) = 3x cos x3 − sinx2x does not approach 0 at infinity, be-
cause the second term tends to 0 (just like the function
√ f (x)), but the first
term is unbounded at infinity:√ if one choose x n = 3
2nπ, ∀n ∈ N, then
lim 3xn cos x3n = lim 3 3 2nπ = +∞.
xn →+∞ n→+∞
Remark 1. To obtain a slant asymptote with non-zero slope, just use a
3
similar function f (x) = x + sinxx .
Remark 2. This example is related to Example 9 in section 4.4, and it
shows that a tempting try to apply derivative inside the limit is not correct:
if lim (f (x) − Ax − B) = 0 it does not mean that lim (f ′ (x) − A) = 0.
x→+∞ x→+∞
4.5.5 L’Hospital’s rule
Example 21. “If f (x) and g (x) are differentiable in a deleted neighbor-
′
hood of a point c, g ′ (x) ̸= 0 in this neighborhood and lim fg′ (x)
(x)
exists, then
x→c
′
f (x)
lim = lim fg′ (x)
(x)
.”
x→c g(x) x→c
Solution.
The functions f (x) = x + 2 and g (x) = x + 1 are differentiable on R, and
′
g ′ (x) = 1 ̸= 0 on R. However, lim fg(x)
(x)
= 2, while lim fg′ (x)
(x)
= 1.
x→0 x→0
Remark 1. It happens because the important condition lim f (x) =
x→c
lim g (x) = 0 in L’Hospital’s rule is missing here.
x→c
Remark 2. For an indeterminate form ∞ ∞ a similar false modification of
L’Hospital’s rule can be constructed and the same functions f (x) and g (x)
can be used for a counterexample.
Remark 3. For both indeterminate forms 00 and ∞ ∞ the point c can also
be infinite. For instance, for the latter form the following counterexample can
be given: if f (x) = 3 − x12 and g (x) = 1 − x1 on (0, +∞), then f ′ (x) = x23 ,
′
g ′ (x) = 1
x2 ̸= 0 and lim f (x) = 3 ̸= 0 = lim f ′ (x) . It happens because
x→+∞ g(x) x→+∞ g (x)
the important condition lim f (x) = lim g (x) = ∞ in L’Hospital’s rule

x→+∞ x→+∞
is missing here.
Example 22. “If f (x) and g (x) are differentiable in a deleted neighbor-
hood of a point c, g ′ (x) ̸= 0 in this neighborhood and lim f (x) = lim g (x) =
x→c x→c
′
0, then lim fg(x)
(x)
= lim fg′ (x)
(x)
.”
x→c x→c
Solution.
The functions f (x) = x2 cos x1 and g (x) = sinx are differentiable in R\ {0},
′
g (x) = cos x ̸= 0 in (−1, 1) and lim x2 cos x1 = lim sin x = 0 (for the first

x→0
x→0
limit we can apply the evaluation x2 cos x1 ≤ x2 ). Nevertheless, lim fg(x)
(x)
̸=
x→0
f ′ (x)
lim ′ In fact, since lim sinx x = 1 and lim x cos x1 = 0 (applying a similar
.
x→0 g (x) x→0 x→0
evaluation x cos x1 ≤ |x|), we can calculate the original limit just applying
the arithmetic rules:
f (x) x2 cos x1 x 1
lim = lim = lim · lim x cos = 1 · 0 = 0.
x→0 g (x) x→0 sinx x→0 sin x x→0 x
At the same time,
f ′ (x) 2x cos x1 + sin x1

lim = lim
x→0 g ′ (x) x→0 cos x
does not exist, because lim 2x cos x1 = 0 and lim cos x = 1, but the limit
x→0 x→0
lim sin x1 does not exist.
x→0
Remark 1. This statement is a wrongly “weakened” version of L’Hospital’s
rule for an indeterminate form 00 where the condition of the existence of
′
lim f ′ (x)
(finite or infinite) is omitted.
x→c g (x)
∞
Remark 2. For an indeterminate form ∞ a similar false modification of
L’Hospital’s rule can be constructed: “if f (x) and g (x) are differentiable in a
deleted neighborhood of a point c, g ′ (x) ̸= 0 in some neighborhood of c, and
′
lim f (x) = lim g (x) = ∞, then lim fg(x)(x)
= lim fg′ (x)
(x)
”. The corresponding
x→c x→c x→c x→c
counterexample can be given in the following form: for f (x) = 2x − sin x and
g (x) = 2x+sin x, the limits at infinity are infinite lim f (x) = lim g (x) =
x→+∞ x→+∞
′ f (x)
+∞, both functions are differentiable on R (with g (x) ̸= 0) and lim =
x→+∞ g(x)
′
1, but lim fg′ (x)
(x)
= lim 2−cos x does not exist, since partial limits are
x→+∞ x→+∞ 2+cos x
different
2 − cos x 1 1 2 − cos x
lim = lim = ̸= 3 = lim 3 = lim .
x→+∞,x=2nπ 2 + cos x n→+∞ 3 3 k→+∞ x→+∞,x=π+2kπ 2 + cos x
Example 23. “If f (x) and g (x) are differentiable on R, lim f (x) =
x→+∞
′ ′
lim g (x) = +∞ and lim f ′ (x) exists, then lim f (x) = lim f ′ (x) .”
x→+∞ x→+∞ g (x) x→+∞ g(x) x→+∞ g (x)
Solution.
The functions f (x) = 2x + sin 2x and g (x) = (2x + sin2x) esin x are differ-
entiable on R and approaches infinity, because |sin 2x| ≤ 1 and esin x ≥ e−1 on
R, while 2x approaches infinity. The limit of the ratio of the derivatives exists
and can be calculated as follows:
f ′ (x) 2 + 2 cos 2x
lim = lim
x→+∞ g ′ (x) x→+∞ (2 + 2 cos 2x) esin x + (2x + sin 2x) cos xesin x
4 cos x
= lim =0
x→+∞ (4 cos x + 2x + sin 2x) esin x
(again 4 cos x and sin 2x are bounded on R, esin x ≥ e−1 on R, and 2x

approaches infinity). Nevertheless, the limit of the ratio of the functions
− sin x
lim f (x) = lim esin 1
x does not exist, since partial limits of e , as
x→+∞ g(x) x→+∞ [ −1 ]
x approaches infinity, assume all the values in e , e .
Remark 1. This statement is a wrongly “weakened” version of L’Hospital’s
rule for an indeterminate form ∞ ′
∞ where the condition that g (x) ̸= 0 in some
neighborhood of infinity is omitted.
Remark 2. A similar example can be provided for an indeterminate form
∞
∞ at a finite point c.
Exercises
1. Show that the following “definition” of differentiability at a point is

not correct: “a function f (x) defined on X is called differentiable at a point
a ∈ X, if for any point x in a neighborhood of a the following representation
holds: f (x) = f (a) + A · (x − a) + α · (x − a), where A is a constant and
|α| ≤ C|x| for some C > 0”. (Hint: this “definition” is too restrictive, try the
differentiable function f (x) = x4/3 at a = 0.)
2. Provide a counterexample to the statement “if f (x) is continuous at a,
{ of the one-sided derivatives at a”.
then there exists at least one
x − 1, x < 0
3. Show that f (x) = is another counterexample to the
x, x ≥ 0
statement in Example 2, section 4.2.
4. Show that f (x) = sin x is a counterexample to the statement “if f (x)
is differentiable then |f (x)| is also differentiable” where the differentiability
of |f (x)| does not hold at infinitely many points (compare with Remark 2 to
Example 5, section 4.2).
5. Analyze if the following statement is true: “if f (x) is infinitely differen-
tiable on R and f (n) (x) = 0 for x ∈ (−∞, 0), then f (x) ≡ 0 on R.” (Hint:
compare this statement with that of Example 6, section 4.2 and modify the
function of that Example.)
6. Verify if the following statement is true: if at some x fixed f ′ (x) =
A, then lim f (x+2h)+f4h(x)−2f (x−h)
= A. What about the converse statement?
h→0
Provide a counterexample to the false statement.
Formulate similar statements for lim 4f (x+h)−3f2h
(x)−f (x+2h)
= A in the
h→0
form of counterexample?
7. Verify if the following statement is true: if at some x fixed f ′′ (x) =
A, then lim f (x+2h)−2fh2(x+h)+f (x) = A. What about the converse statement?
h→0
Provide a counterexample to the false statement.
Formulate similar statements for lim f (x+2h)−f (x+h)−f
2h2
(x)+f (x−h)
= A in
h→0
the form of counterexample?
8. Give a counterexample to the statement in Example 1, section 4.3, with
the sum substituted by the product of two functions.
9. Provide a counterexample to the following statement: “if f ′ (x) ex-
ists and is bounded on (a, b), then f ′ (x) cannot have infinitely many
points
 of discontinuity”.
( (Hint:
] consider on (−1, 1) the function f (x) =

 x 2
g (−x) , x ∈ − 1
, − 1
 n n n+1 ( )
x−1/(n+1)
0, x = 0 [ ) , where gn (x) = g , g(x) =


n(n+1)
 x2 gn (x) , x ∈ 1 , 1
n+1 n
x(x − 1)2 sin (x−1)
π
2 .)
10. It is well-known that the derivative of a differentiable periodic func-

tion is also periodic with the same period. Show that the converse is false.
Formulate the result as a counterexample.
11. Provide a counterexample to the statement “if f (x) is differentiable
and periodic with a fundamental period T , then f ′ (x) is also periodic with
the fundamental period T ”.
12. Give a counterexample to the statement: “if f (x) is differentiable on
an interval, then it is uniformly continuous on the same interval”. What about
the converse? Consider separately open and closed intervals.
13. It is well-known that a function f (x) that has a bounded derivative on a
(finite or infinite) interval is uniformly continuous on that interval. Show that
the uniform continuity and differentiability on √ an interval do not guarantee
boundedness of derivative. (Hint: try f (x) = x on (0, 1).)
14. Recall that a function f (x) is Lipschitz-continuous on an interval I if
there exists a constant C ≥ 0 such that |f (x) − f (y)| ≤ C|x − y| for ∀x, y ∈ I.
Show that Lipschitz-continuity does not imply differentibility. What about the
converse? What if f (x) is infinitely differentiable on I?
15. Any critical point of f (x) is also a critical point of f 2 (x). Is the converse
correct?
16. There are two apparently natural and similar definitions of a increasing
at a point a function. The first says that f (x) is increasing at a point a if there
exists a neighborhood of a where f (x) is increasing; and the second determines
that f (x) is increasing at a point a if there exists a neighborhood of a such that
f (x) ≤ f (a) for ∀x < a and f (x) ≥ f (a) for ∀x > a. (Analogous definitions
can be specified for a strictly increasing function.) Show that these definitions
are not equivalent. (Hint: use the function { of Example 6, section 4.5.)
x4 sin2 x1 , x ̸= 0
17. Verify that the function f (x) = provides a coun-
0, x=0
terexample to the following intuitive “rule”: “if f (x) is a smooth function,
say twice differentiable, in a neighborhood of a minimum point a, then f (x)
should be decreasing in some interval to the left of a and increasing in some
interval to the right”. Compare the result with Example 10 in section 4.5.
18. Explain why the following “definition” of upward concavity is wrong:
“a function f (x) is upward concave on (a, b) if the inequality f (α1 a + α2 b) ≤
α1 f (a) + α2 f (b) holds for arbitrary parameters α1 , α2 ≥ 0, α1 + α2 = 1”.
Give an example of function satisfying this property, but not upward concave.
19. Provide a counterexample to the statement: “if f (x) and g(x) are down-
ward concave on an interval, then their product is also a downward concave
function on this interval”. Show that the conclusion that the product is an
upward concave function is also wrong. What if f (x) and g(x) are upward
concave?
20. If f (x) is strictly upward concave and g(x) is strictly downward concave
on R, then both f (x) + g(x) and f (x) · g(x) may maintain the same type of
concavity over R. Show this by example and formulate this statement in the
form suitable for counterexamples.
21. It is well-known that the convexity on an interval (a, b) implies the

continuity on the same interval. Does it imply differentiability on (a, b) ? Give
examples for both strict and non-strict cases.
22. A function f (x) is continuous on [a, b] and strictly convex on [a, c] and
[c, b], for some c ∈ (a, b). Does it convex on [a, b]?
23. The theorem states that if lim f ′ (x) = A and lim (f (x) − xf ′ (x)) =
x→+∞ x→+∞
B, then y = Ax + B is a slant asymptote of f (x). Show that the converse is
not true. (Hint: compare with Example 20 in section 4.5.)
24. Explain why a direct application of L’Hospital’s rule does not work
in the case f (x) = (2x + 1)2/3 , g(x) = (x + 2)2/3 , x → +∞. Show that
(2x+1)2/3
lim 2/3 = 22/3 and formulate this as a counterexample.
x→+∞ (x+2)
25. Show that f (x) = x2 sin x1 and g(x) = x provide another counterexam-
ple to the statement in Example 22, section 4.5. ( )
26. Show that f (x) = x2 − 14 sin 2x and g (x) = x2 − 41 sin 2x (cos x + 2)2
represent another counterexample to the statement in Example 23, section
4.5.
Chapter 5
Integrals
Indefinite integral
Antiderivative and Indefinite Integral. A function f (x) is integrable

on a set S if there exists a differentiable on S function F (x) such that F ′ (x) =
f (x) for ∀x ∈ S. The function F (x) is called an antiderivative (or a primitive)
of the function f (x) on S. If F (x) is an antiderivative of f (x) on S, then the
set of all antiderivatives is called an indefinite integral of f (x) on S and it
has the form F (x) + C, ∫where C is an arbitrary real constant. The standard
notation is F (x) + C = f (x) dx.
Basic properties: ∫
1) if f (x) is differentiable on S, then f ′ (x) dx = f (x) + C on S.
(∫ )′
2) if f (x) is integrable on S, then f (x) dx = f (x) on S.
3) if f (x) and g (x) are integrable on S, then f (x) + g (x) is also integrable
on S and ∫ ∫ ∫
f (x) + g (x) dx = f (x) dx + g (x) dx.
4) if f (x) is integrable on S and c is non-zero constant, then cf (x) is also

integrable on S and ∫ ∫
cf (x) dx = c f (x) dx.
Change of variable.
If φ (t) is differentiable on T and f (x) is integrable on S = φ (T ), then the
133
function f (φ (t)) φ′ (t) is integrable on T and

∫ ∫
f (x) dx = f (φ (t)) φ′ (t) dt.
Integration by parts.
If f (x) and g (x) are differentiable on S, and f (x) g ′ (x) is integrable on
S, then f ′ (x) g (x) is also integrable on S and
∫ ∫
f (x) g (x) dx = f (x) g (x) − f (x) g ′ (x) dx.
′
Definite (Riemann) integral
Riemann Integral. Let f (x) be defined on [a, b]. A partition P is a

finite ordered set of points x0 , x1 , . . . , xn−1 , xn in [a, b] such that
P = {xi : a = x0 < x1 < . . . < xn−1 < xn = b} .
We denote the length of a subinterval [xi−1 , xi ] by ∆xi = xi − xi−1 and the

partition diameter by ∆ = max ∆xi . The Riemann sum of f (x) correspond-
1≤i≤n
ing to the specific partition P and the specific choice of points ci ∈ [xi−1 , xi ]
is
∑
n
S (f ; P, c) = f (ci ) · ∆xi ,
i=1
in the abbreviated form S (f ). If there exists a finite limit of Riemann sums

of f (x), as ∆ approaches 0, which does not depend on a choice of partitions
and points ci , then the function f (x) is Riemann integrable on [a, b] and the
value of the limit is called the Riemann (definite) integral of f (x) on [a, b]:
∫ b
f (x) dx = lim S (f ) .
a ∆→0
∫b ∫b
Additionally, if a = b, then a f (x) dx = 0, and if a > b, then a f (x) dx =
∫a
− b f (x) dx .
Remark 1. If there is no confusion with an indefinite integral, the Riemann
integrable function can be also called an integrable function and the Riemann
integral just an integral. We will use these abbreviations in this section.
Remark 2. All results below are formulated for a < b.
Theorem. If f (x) is integrable on [a, b], then it is bounded there.

Integrals 135
Theorem. If f (x) is continuous on [a, b], then it is integrable there.

Theorem. If f (x) is bounded on [a, b] and is continuous on [a, b] except
at a finite number of points, then f (x) is integrable on [a, b].
Theorem. If f (x) is bounded on [a, b] and is continuous on [a, b] except
at a countable set of points, then f (x) is integrable on [a, b].
Lebesgue’s Criterion. f (x) is integrable on [a, b] if, and only if, f (x)
is bounded on [a, b] and the set of its discontinuity points has (Lebesgue)
measure zero.
Remark. Any finite or countable set has Lebesgue measure zero.
Comparative Properties
1) If f (x) and g (x) are integrable on [a, b] and f (x) ≤ g (x) on [a, b] , then
∫ b ∫ b
f (x) dx ≤ g (x) dx.
a a
2) If f (x) is integrable on [a, b], finf = inf f (x), fsup = sup f (x), then
x∈[a,b] x∈[a,b]
∫ b
finf · (b − a) ≤ f (x) dx ≤ fsup · (b − a).
a
In particular, if f (x) is continuous on [a, b], fmin = min f (x), fmax =

x∈[a,b]
max f (x), then
x∈[a,b]
∫ b
fmin · (b − a) ≤ f (x) dx ≤ fmax · (b − a) .
a

1) If f (x) and g (x) are integrable on [a, b], then f (x) + g (x) is also integrable
on [a, b] and
∫ b ∫ b ∫ b
f (x) + g (x) dx = f (x) dx + g (x) dx.
a a a
2) If f (x) is integrable on [a, b] and c is an arbitrary constant, then cf (x) is

also integrable on [a, b] and
∫ b ∫ b
cf (x) dx = c f (x) dx.
a a
3) If f (x) is integrable on [a, b], then |f (x)| is also integrable on [a, b] and
∫ ∫
b b

f (x) dx ≤ |f (x)| dx.
a a
4) If f (x) and g (x) are integrable on [a, b], then f (x) g (x) is also integrable
on [a, b].
5) f (x) is integrable on [a, b], if and only if, f (x) is integrable on [a, c] and
[c, b], where c is some point in [a, b]. Furthermore,
∫ b ∫ c ∫ b
f (x) dx = f (x) dx + f (x) dx.
a a c
The Mean Value Theorem for integral. If f (x) is continuous on

∫b
[a, b], there exists a point c ∈ [a, b] such that a f (x) dx = f (c) · (b − a).
The Fundamental Theorem of Calculus.

The first formulation. If f (x) is Riemann integrable on [a, b] and has
an antiderivative on [a, b], that is there exists F (x) such that F ′ (x) = f (x)
∫b
on [a, b], then a f (x) dx = F (b) − F (a).
The second formulation (a weaker version). If f (x) is continuous
∫b
on [a, b], then a f (x) dx = F (b) − F (a), where F (x) is one of antiderivatives
of f (x) on [a, b].
Remark. If f (x) is continuous on [a, b], then it has an antiderivative F (x)
on this interval.
Change of variable.
The first formulation. If f (x) is integrable on [a, b], and φ (t) is contin-
uously differentiable and strictly increasing (decreasing) on [α, β] with image
[a, b], then
∫ b ∫ β ∫ a ∫ β
f (x) dx = f (φ (t)) φ′ (t) dt ( f (x) dx = f (φ (t)) φ′ (t) dt ).
a α b α
The second formulation. If f (x) is continuous on [a, b], and φ (t) is

continuously differentiable on [α, β] with image [a, b] and φ (α) = a, φ (β) = b
(φ (α) = b, φ (β) = a), then
∫ b ∫ β ∫ a ∫ β
f (x) dx = f (φ (t)) φ′ (t) dt ( f (x) dx = f (φ (t)) φ′ (t) dt ).
a α b α
The first formulation. If f (x) and g (x) are differentiable on [a, b] and
f (x) g ′ (x) is integrable on [a, b], then f ′ (x) g (x) is also integrable on [a, b]
and
∫ b ∫ b
f ′ (x) g (x) dx = f (b) g (b) − f (a) g (a) − f (x) g ′ (x) dx.
a a
The second formulation (a weaker version). If f (x) and g (x) are

Integrals 137
continuously differentiable on [a, b], then

∫ b ∫ b
f ′ (x) g (x) dx = f (b) g (b) − f (a) g (a) − f (x) g ′ (x) dx.
a a
Improper integrals
Improper integral of the first kind. Let f (x) be integrable on every

interval [a, b], where a is fixed and b is arbitrary, b > a. The following limit is
called an improper integral of the first kind on [a, +∞):
∫ +∞ ∫ b
f (x) dx = lim f (x) dx.
a b→+∞ a
In the same way one can define an improper integral of the first kind on
(−∞, b]:
∫ b ∫ b
−∞ a→−∞ a
If the limit exists and is finite, then the improper integral is called convergent.
Otherwise it is called divergent.
Additionally, an improper integral of the first kind on (−∞, +∞) is defined
as follows: ∫ +∞ ∫ a ∫ +∞
f (x) dx = f (x) dx + f (x) dx
−∞ −∞ a
(here a is an arbitrary real number). If both integrals in the right-hand side

are convergent, then the integral on (−∞, +∞) is called convergent.
Remark 1. We do not consider here the principal value of the last integral.
∫ +∞Remark 2. We present below some results for the improper integral
a
f (x) dx. Similar results are true for other improper integrals of the first
kind.
∫ +∞
Cauchy criterion. An improper integral a f (x) dx is convergent if,
∫ if, for ∀ε > 0 there is B > a such that for any pair b2 > b1 > B it
and only
b
holds b12 f (x) dx < ε.
General Comparison theorem. Let 0 ≤ f (x) ≤ g (x) on [a, +∞).
∫ +∞ ∫ +∞
If a g (x) dx converges, then a f (x) dx also converges (equivalently, if
∫ +∞ ∫ +∞
a
f (x) dx diverges, then a g (x) dx also diverges).
Specific Comparison theorem. If 0 ≤ f (x) ≤ xcp on [a, +∞), where
∫ +∞
c > 0 is a constant and p > 1, then a f (x) dx converges. If f (x) ≥ xcp on
∫ +∞
[a, +∞), where c > 0 is a constant and p ≤ 1, then a f (x) dx diverges.
Change of variable.
Let f (x) be continuous on [a, +∞), and φ (t) be continuously differentiable
and strictly increasing on [α, +∞] with image [a, +∞). Then the improper
∫ +∞ ∫ +∞
integrals a f (x) dx and α f (φ (t)) φ′ (t) dt are both convergent or both
divergent and ∫ ∫+∞ +∞
f (x) dx = f (φ (t)) φ′ (t) dt.
a α
Let f (x) and g (x) be continuously differentiable on [a, +∞), and
∫ +∞
lim f (x) g (x) = B. Then the improper integrals a f (x) g ′ (x) dx and
x→+∞
∫ +∞ ′
a
f (x) g (x) dx are both convergent or both divergent and
∫ +∞ ∫ +∞
′
f (x) g (x) dx = B − f (a) g (a) − f (x) g ′ (x) dx.
a a
Improper integral of the second kind. Let f (x) be defined on [a, b),
where a and b are finite points, integrable on every interval [a, c], where c
is arbitrary between a and b, a < c < b, and unbounded in any left-hand
neighborhood of b (that is b is a singular point of f (x)). The following limit
is called an improper integral of the second kind at b:
∫ b ∫ c
a c→b− a
In the same way one can define another improper integral of the second
kind at a singular point a:
∫ b ∫ b
a c→a+ c
If the limit exists and is finite, then the improper integral is called convergent.
Otherwise it is called divergent.
Additionally, if d ∈ (a, b) is an internal singular point, then an improper
integral of the second kind is defined as follows:
∫ b ∫ d ∫ b
f (x) dx = f (x) dx + f (x) dx, a < d < b,
a a d
where the integrals in the right-hand side are the improper integrals at d. If
both integrals in the right-hand side are convergent, then the integral in the
left-hand side is called convergent.
Remark 1. We do not consider here the principal value of the last integral.
Integrals 139
Remark 2. If f (x) is integrable on [a, b], then an improper integral of the

second kind coincides with the Riemann integral.
Remark 3. Improper integrals of the second kind can be transformed to
improper integrals of the first kind by a change of variable.
Applications
Rectifiable curve. Let f (x) be continuous on [a, b]. For an arbitrary

partition
P = {xi : a = x0 < x1 < . . . < xn−1 < xn = b}
of [a, b] one can define the length of a polygonal curve inscribed in the graph
of f (x):
n √
∑ 2 2
L (f, P ) = (xi − xi−1 ) + (f (xi ) − f (xi−1 )) .
i=1
If a finite limit of these lengths, as the partition diameter ∆ approaches 0,
exists and does not depend on a choice of partitions, then it is called the length
of f (x) on [a, b]: L (f ) = lim L (f, P ). In this case the curve of f (x) on [a, b]
∆→0
is called rectifiable.
Theorem. If f (x) is continuously differentiable on [a, b], then the curve
of f (x) on √[a, b] is rectifiable and its length can be calculated by the formula
∫b 2
L (f ) = a 1 + (f ′ (x)) dx.
Area of a region. Quadrable figure. Let f (x) and g (x) be defined on

[a, b]. Consider the plane figure that lies between two curves f (x) and g (x)
and between two vertical lines x = a and x = b. For an arbitrary partition
P = {xi : a = x0 < x1 < . . . < xn−1 < xn = b}
of [a, b] and an arbitrary choice of points ci ∈ [xi−1 , xi ] one can define the sum
of the areas of approximating rectangles as follows:
∑
n
A (f, g, P ) = |f (ci ) − g (ci )| · (xi − xi−1 ) .
i=1
If there exists a finite limit of these sums as the partition diameter ∆ ap-
proaches 0, and this limit does not depend on a choice of partitions and
points ci , then it is called the area of the considered plane figure: A (f, g) =
lim A (f, g, P ). In this case the plane figure is called quadrable or squarable.
∆→0
Remark . Actually, this is a simplified definition of a quadrable plane
figure. The complete definition is much more complex, involving sets of in-
scribed/circumscribed polygons and the concepts of infimum/supremum, and
it is out of scope of this text.
Theorem. If f (x) and g (x) are continuous on [a, b], then the correspond-
ing plane figure is quadrable and its area can be calculated by the formula
∫b
A (f, g) = a |f (x) − g (x)| dx.
Solid of revolution. A solid of revolution is a space figure obtained by

rotating a plane region around a straight line (the axis of rotation), which lies
on the same plane and does not contain interior points of the region.
Theorem. If f (x) is continuous on [a, b], the plane region lies between
f (x) and the x-axis and between vertical sides x = a and x = b, and the
x-axis is the axis of rotation, then the surface area and the volume of the solid
of revolution are well-defined and can be calculated by the formulas
∫ b √ ∫ b
2
S (f ) = 2π |f (x)| 1 + (f ′ (x)) dx and V (f ) = π f 2 (x) dx,
a a
respectively.
Remark . Variants of these formulas in the cases when a plane figure is
defined by two curvilinear boundaries or when the axis of rotation is differ-
ent of the x-axis, etc., can be easily deduced from these basic formulas (or
consulted in textbooks).
5.2 Indefinite integral
Remark . In this section we will say that a function f (x) is integrable

when f (x) has an antiderivative, i.e. when its indefinite integral exists.
Example 1. “If f (x) is integrable on an interval, then it is bounded

there.”
Solution.
For an open interval (a, b) (a counterexample
) is very simple. The function
f (x) = cos12 x is integrable on − π2 , π2 , with one of the antiderivatives being
F (x) = tan x, but f (x) is not bounded on this interval.
For a{closed interval [a, b] the situation is not so simple, but the function
2x sin x12 − x2 cos x12 , x ̸= 0
f (x) = considered in Example 5, section 4.3
0, x = 0
can provide a counterexample. For convenience we reproduce here similar con-
siderations of that Example using { the integral terminology. One of antideriva-
x2 sin x12 , x ̸= 0
tives of f (x) on R is F (x) = . This is easy to check by
0, x = 0
applying the arithmetic and chain rules of differentiation for any x ̸= 0 and
appealing to the definition for x = 0: lim F (x)−F x−0
(0)
= lim x sin x12 = 0, since
x→0 x→0
Integrals 141

x sin 12 ≤ |x|. However, f (x) is not bounded in any neighborhood of the ori-
x ( √ )
gin since for xk = √2kπ1
, ∀k ∈ N one gets lim f (xk ) = lim −2 2kπ =
xk →0 k→+∞
−∞. Therefore, the function f (x) is integrable, but unbounded on [−1, 1] (or
on any closed interval containing 0).
Example 2. “If f (x) is bounded on an interval, then it is integrable

there.”
Solution.
The function f (x) = sgn x is bounded on R, but it is not integrable on any
interval containing the origin, because the derivative cannot contain a jump
discontinuity (according to Darboux’s theorem).
{ Another (more radical) example is Dirichlet’s function D (x) =
1, x ∈ Q
, which is bounded on an arbitrary interval, but it is not in-
0, x ∈ I
tegrable on any interval.
Example 3. “If f (x) and g (x) are not integrable, then f (x) + g (x) is
not integrable too.”
Solution.
The functions f (x) = x − sgn x and g (x) = sgn x are not integrable on
any interval containing the origin, because they have jump discontinuities at
zero, but their sum f (x) + g (x) = x is integrable on R.
Remark . Similar false statements can be formulated for other arithmetic
operations. For instance, the statement for the product: “if f (x) and g (x)
are not integrable, then f (x) · g (x) is not integrable too”.
{ It can be disproved
x − 1, x < 0
by the counterexample with the functions f (x) = and
{ x + 1, x ≥ 0
−1 , x < 0
g (x) = . They have no antiderivative on (−1, 1), but their
1, x ≥ 0
product f (x)·g (x) = 1+|x| is integrable on R, with one of the antiderivatives
being x + 21 x |x|.
Example 4. “If f (x) is integrable and g (x) is not integrable on an interval,

then f (x) · g (x) is not integrable on this interval.”
Solution.
The function f (x) = x is integrable on R, and g (x) = sgn x is not inte-
grable on any interval∫ containing the origin, but their product f (x)·g (x) = |x|
is integrable on R: |x| dx = 12 x |x| + C , C = const.
Remark . The statement would be true if instead of the product one
considers the sum or difference of an integrable and non-integrable functions.
Example 5. “If f (x) is not integrable on an interval, then |f (x)| is also

not integrable on this interval.”
Solution.
{
−1, x < 0
The function f (x) = is not integrable on any interval con-
1, x ≥ 0
taining the origin, but the function |f (x)| = 1 is integrable on R.
{ Another (more radical) example is the modified Dirichlet function
D̃ (x) =
1, x ∈ Q
, which is not integrable on any interval, but D̃ (x) = 1 is
−1 , x ∈ I
integrable on R.
Remark 2. Similarly, the following statement on the square is false: “if
f (x) is not integrable on an interval, then f 2 (x) is also not integrable on this
interval”. The same functions can be used for counterexample.
Example 6. “If f (x) satisfies the intermediate value property on an in-

terval, then it is integrable there.”
Solution. {
2x sin x1 − cos x1 , x ̸= 0
The function f (x) = is integrable on [−1, 1]
0, x = 0
{ 2
x sin x1 , x ̸= 0
with an antiderivative F (x) = (see Example 3, section
0, x = 0
{
2x sin x1 , x ̸= 0
4.2 for details), and g (x) = is integrable on [−1, 1]
0, x = 0
since
{ it is continuous. Therefore, their difference h(x) = g(x) − f (x) =
cos x1 , x ̸= 0
is also integrable on [−1, 1]. Using the same reasoning as
0, x = 0
in Example
{ 14,1 section 3.4, it can be shown that the more general function
cos x , x ̸= 0
h̃(x) = has the intermediate value property on [−1, 1] for
α, x = 0
any α ∈ [−1, 1] (for α = 0, that is, for h (x) it also follows from Darboux’s
theorem). However, h̃ (x) { is not integrable for any α ̸= 0, since otherwise the
0 , x ̸= 0
function h̃(x) − h (x) = would be also integrable on [−1, 1],
α, x = 0
which is not possible because a derivative cannot contain a removable discon-
tinuity.
Remark 1. In the above counterexample the interval [−1, 1] can be sub-
stituted by any interval containing 0.
Remark 2. { Another similar example can be provided with the func-
2
x cos x
1
2 , x ̸= 0
tion h̃(x) = considered on any interval containing 0.
α, x = 0
This function has the intermediate value property for an arbitrary α, but
{ is2 not 1integrable for α ̸= 0, because the corresponding function h(x) =
it
x cos x2 , x ̸= 0 is integrable as the difference between the two integrable
0, x = 0
Integrals 143
{
2x sin x12 − x2 cos x12 , x ̸= 0
functions f (x) = (see Example 1 in this sec-
0, x = 0
{
2x sin x12 , x ̸= 0
tion) and g (x) = (the latter is continuous).
0, x = 0
Remark 3. This is a wrong converse to Darboux’s theorem: if f (x) is inte-
grable on a connected set S, then it satisfies the intermediate value property
on S.
Example 7. “The product of two integrable functions is also integrable.”

Solution.
Let us consider
{ the three integrable functions of { the previous Example
2x sin x1 − cos x1 , x ̸= 0 2x sin x1 , x ̸= 0
6: f (x) = , g (x) = and
0, x = 0 0, x = 0
{
cos x1 , x ̸= 0
h(x) = g(x) − f (x) = . It can be shown, that the func-
0 , )x = 0
{ 1(
2 1 + cos x , x ̸= 0
2
tion h2 (x) = is not integrable on any interval
0, x = 0
containing 0. In fact, let us suppose, per absurd, that h2 (x) is integrable
and denote { its primitive by H̃(x). If H(x) is a primitive of h(x), then
( ( x ))′
2 cos x , x ̸= 0
1 2
H 2 = (applying the chain rule for x ̸= 0 and the
0, x = 0
( { 1
( x ) )′
definition for x = 0). Then H̃ (x) − H 2 ≡ ĥ (x) = 2 , x ̸= 0 , but
0, x = 0
this is not possible since (according to Darboux’s theorem) a derivative cannot
have a removable discontinuity.
Remark. A similar statement for Riemann integrable functions is true.
Example 8. “If f (x) is even and integrable, then its antiderivative is

odd.”
Solution.
The even function f (x) = cos x has the antiderivative F (x) = sin x + 1,
which is not odd (nor even).
Remark 1. The correct statement is that among all the antiderivatives of
an even function there exists one antiderivative that is odd.
Remark 2. For an odd function the statement is true: any antiderivative
of an odd function is even.
Example 9. “If f (x) is periodic and integrable, then there is a periodic

antiderivative of f (x).”
Solution.
For the periodic function f (x) = cos x+1, the set of all the antiderivatives
(indefinite integral) has the form F (x) + C = sin x + x + C, C = const, where
no one of the functions is periodic.
Remark. The correct statement asserts that any antiderivative of a peri-
odic function is the sum of a periodic function and a linear function.
5.3 Definite (Riemann) integral
Remark. In this section we will say that a function f (x) is integrable

when there exists an indefinite integral of f (x), and say that f (x) is Riemann
integrable when there exists a definite (Riemann) integral of f (x).
Example 1. “If f (x) is bounded on [a, b], then f (x) is Riemann integrable
on [a, b].”
Solution. {
1, x ∈ Q
Dirichlet’s function D (x) = is bounded but not Riemann
0, x ∈ I
integrable on any interval [a, b]. Indeed, for an arbitrary partition
P = {xi : a = x0 < x1 < . . . < xn−1 < xn = b; ∆xi = xi − xi−1 , i = 1, . . . , n}
of [a, b] in each of subintervals [xi−1 , xi ] there are both rational and irrational
Integrals 145
points. Choosing rational points ci ∈ [xi−1 , xi ] in the definition of Riemann

sums, we obtain
∑
n ∑
n
Srat (f ) = f (ci ) · ∆xi = 1 · ∆xi = b − a,
i=1 i=1
while for irrational points ci , we have

∑
n ∑
n
Sirr (f ) = f (ci ) · ∆xi = 0 · ∆xi = 0.
i=1 i=1
Therefore, the limits for these sums, as the partition diameter ∆ = max ∆xi
1≤i≤n
approaches 0, are different
lim Srat (f ) = b − a ̸= 0 = lim Sirr (f ) ,
∆→0 ∆→0
which means that the Riemann integral does not exist.

Remark 1. For the reader accustomed to the upper and lower Riemann in-
tegrals, the proof can be made in the following way. For an arbitrary partition
P of [a, b], the upper Riemann sum is
∑
n ∑
n
S̄ (f ) = sup f (x) · ∆xi = 1 · ∆xi = b − a,
i=1 x∈[xi−1 ,xi ] i=1
while the lower sum is

∑
n ∑
n
S (f ) = inf f (x) · ∆xi = 0 · ∆xi = 0.
x∈[xi−1 ,xi ]
i=1 i=1
Therefore, the upper and lower Riemann integrals are different

lim S̄ (f ) = b − a ̸= 0 = lim S (f ) ,
∆→0 ∆→0
which means that the Riemann integral does not exist.

Example 2. “If f (x) is Riemann integrable on [a, b], then f (x) is contin-
uous on [a, b].”
Solution.
The function f (x) = sgn x is Riemann integrable but non-continuous on
[0, 1]. In fact, f (x) has a removable discontinuity at 0 (as the function is
considered on [0, 1]). Let us show that f (x) is Riemann integrable by the
definition. For an arbitrary partition of [0, 1] and an arbitrary choice of the
points ci ∈ [xi−1 , xi ], the following evaluation of the Riemann sums S (f )
holds:
∑
n ∑
n ∑
n
1 − ∆x1 = 0 · ∆x1 + 1 · ∆xi ≤ S (f ) = f (ci ) · ∆xi ≤ 1 · ∆xi = 1.
i=2 i=1 i=1
Since ∆x1 → 0 when ∆ → 0, it follows that the limit of the Riemann sums,
as ∆ approaches 0, exists and does ∫ 1 not depend on the choice of partition and
points ci : lim S (f ) = 1, that is 0 sgn xdx = 1.
∆→0
Remark 1. For those who use to apply the upper and lower Riemann
integrals, the proof can be done as follows. For an arbitrary partition of [0, 1]
the upper Riemann sum is
∑
n ∑
n
S̄ (f ) = sup f (x) · ∆xi = 1 · ∆xi = 1,
i=1 x∈[xi−1 ,xi ] i=1
while the lower sum is

∑
n ∑
n
S (f ) = inf f (x) · ∆xi = 0 · ∆x1 + 1 · ∆xi = 1 − ∆x1 .
x∈[xi−1 ,xi ]
i=1 i=2
∫1
Therefore, lim S̄ (f ) = 1 = lim S (f ), that is 0
sgn xdx = 1.
∆→0 ∆→0
Remark 2. The same considerations
{ can be applied to show that the
0, x < 0
Heaviside function H (x) = is Riemann integrable on [−1, 0] and
1, x ≥ 0
∫0
−1
H (x) dx = 0. It is just a bit more difficult to show that f (x) = sgn x and
H (x) are Riemann integrable but discontinuous on any interval containing 0.
Remark
 3. Another interesting counterexample is the Riemann function
 1/n , x ∈ Q, x = m/n
R (x) = 0, x ∈ I (recall that m ∈ Z, n ∈ N and the ratio m
 n
1, x = 0
is in lowest terms), defined in Example 4, section 3.3 where it was shown that
R (x) is discontinuous at all rational points (and continuous at all irrational
points). It happens that this function is also Riemann integrable. This can
be shown by applying the definition, but the proof is somewhat involved. To
avoid such complications, the reader familiarized with the Lebesgue measure
can appeal to the Lebesgue criterion that asserts that a bounded function is
Riemann integrable if the set of the points of its discontinuity is of Lebesgue
measure zero, which is the case of the Riemann function that has countably
many discontinuities on any interval [a, b]. Further, we can find the value of
the integral by noting that all the lower Riemann sums are equal to 0:
∑
n ∑
n
S (f ) = inf f (x) · ∆xi = 0 · ∆xi = 0
x∈[xi−1 ,xi ]
i=1 i=1
(since any subinterval of an arbitrary partition contains irrational points).

Since the Riemann integral exists, its value is equal to the lower Riemann
∫b
integral, hence a R (x) dx = 0.
Example 3. “If a function f (x) is integrable on [a, b], then it is Riemann

integrable on this interval.”
Integrals 147
Solution. {
2x sin x12 − x2 cos x12 , x ̸= 0
The function f (x) = is integrable and
0, x = 0
unbounded on [−1, 1] (see Example 1, section 5.2 for details). So it is not
Riemann integrable on this interval.
Example 4. “If a function f (x) is Riemann integrable on [a, b], then it is

integrable on [a, b].”
Solution.
The function f (x) = sgn x is Riemann integrable on [0, 1] (see Example 2
in this section), but it has no antiderivative on [0, 1], because the derivative
of any function cannot have removable or jump discontinuities.
Example 5. “If |f (x)| is Riemann integrable on [a, b], then f (x) is also
Riemann integrable on [a, b].”
Solution. {
1, x ∈ Q
The modified Dirichlet function D̃ (x) = is not Riemann
−1 , x ∈ I

integrable on any interval (it can be shown like in Example 1), but D̃ (x) = 1
is Riemann integrable on any interval.
Remark 2. The corresponding statement for the square of a function is
also false: “if f 2 (x) is Riemann integrable on [a, b], then f (x) is also Rie-
mann integrable on [a, b]”. The same modified Dirichlet function D̃ (x) gives
a counterexample. Again the converse is true.
Example 6. “If f (x) · g (x) is Riemann integrable on [a, b], then f (x) and
g (x) are Riemann integrable on [a, b].”
Solution.
If f (x) = g (x), then the problem is reduced to Remark 2 of the previous
Example 5.
{ Of course, the functions may be { different. For example, if f (x) =
tan x , x ∈ (0, π/2) cot x , x ∈ (0, π/2)
and g (x) = , both functions
1 , x = 0,[ x =] π/2 1 , x = 0, x = π/2 [ ]
defined on 0, π2 , then f (x) and g (x) are unbounded on 0, π2 , and conse-
quently, are not Riemann [ integrable.
] However, h (x) = f (x) · g (x) = 1 is
Riemann integrable on 0, π2 .
Remark. The converse is true.
∫b
Example 7. “If f (x) is Riemann integrable on [a, b] and a f (x) dx = 0,
then f (x) ≡ 0 on [a, b].”
Solution.
Consider non-zero function f (x) = x on [−1, 1]. Evidently f (x) is Rie-
∫1
2 1
mann integrable and −1 xdx = x2 = 0.
−1

Remark 2. The following generalized version of the statement is also
∫b
false: “if f (x) and g (x) are Riemann integrable on [a, b] and a f (x) dx =
∫b
a
g (x) dx, then f (x) = g (x) on [a, b]”. The corresponding counterexample
can be constructed with the functions f (x) = x and g (x) = sin x on [−a, a],
∀a > 0. Again, the converse is true.
∫b
Example 8. “If f (x) ≥ 0 on [a, b] and a f (x) dx = 0, then f (x) ≡ 0 on
[a, b].”
Solution. ∫1
The Riemann function R (x) is non-negative on [0, 1] and 0 R (x) dx = 0
(see details in Remark 3 to Example 2), but it is positive at every rational
point.
Remark. The result is correct for continuous on [a, b] functions.
∫b
Example 9. “If f (x) is Riemann integrable on [a, b] and a f 2 (x) dx = 0,
then f (x) ≡ 0 on [a, b].”
Solution.
For the non-zero integrable function f (x) = 1 − sgn x considered on [0, 1]
∫1
the integral of f 2 (x) = (1 − sgn x)2 = 1 − sgn x equals 0: 0 1 − sgn xdx =
∫1
0
0dx = 0. Here we have applied the property that the value of the Riemann
integral does not depend on the definition of a function at a finite number of
points.
Remark 1. To construct a counterexample with a function assuming pos-
itive and negative values at a countable set of points we can use the modified
Riemann function:
 √
 1/ √ n , x ∈ Q, x = m/n , n = 2k
f (x) = R̃ (x) = −1/ n , x ∈ Q, x = m/n , n = 2k + 1

0 , x ∈ I ; 1, x = 0
(here m ∈ Z, n ∈ N, k ∈ N and the ratio m n is in lowest terms). Since

2
∫b
R̃ (x) = R (x) and a R (x) dx = 0 on any interval [a, b] (see Remark 3 to
Example 2), the conditions of the statement are satisfied.
Remark 2. Under the conditions of the statement, it is true that f (x) = 0
at all points of [a, b] where f (x) is continuous.
∫b
Example 10. “If a f (x) dx ≥ 0, then f (x) ≥ 0 on [a, b].”
Solution.
The function f (x) = x2 − 1 is negative on [0, 1) and positive on (1, 2], and
∫2( 2 )
0
x − 1 dx = 23 > 0.
Remark 1. The converse is true if f (x) is Riemann integrable.
Remark 2. The statement can be generalized to the following form (also
Integrals 149
∫b ∫b
false): “if a f (x) dx ≥ a g (x) dx, then f (x) ≥ g (x) on [a, b] ”. The corre-
sponding counterexample can be provided with the functions f (x) = x2 and
g (x) = 1 on [0, 2]. Again, the converse is true.
Example 11. “If f (x) + g (x) is Riemann integrable on [a, b], then both
f (x) and g (x) are Riemann integrable on [a, b].”
Solution.
f (x) = D (x) and g (x) = 1 − D (x) is a simple counterexample for this
case.
Remark. Similar statements for other arithmetic operations are also false.
For example, for the ratio the corresponding counterexample can be provided
using the modified Dirichlet function: f (x) = g (x) = D̃ (x).
Example 12. “The composition of two Riemann integrable functions is

also Riemann integrable.”
Solution.
Let f (x) be the Riemann function f (x) = R (x) on [0, 1] and g (x) = sgn x
on R. Both functions are Riemann integrable on any interval (see Remark 3 to
Example
{ 2). }However, their composition is the Dirichlet function g (f (x)) =
1, x ∈ Q
= D (x) on [0, 1], that is not Riemann integrable on [0, 1].
0, x ∈ I
Example 13. “If F (x) is an antiderivative of f (x) on [a, b], then

∫b
a
f (x) dx = F (b) − F (a).”
Solution. {
2x sin x12 − x2 cos x12 , x ̸= 0
The function f (x) = in Example 1, sec-
0, x = 0
tion 5.2 has an antiderivative on [−1, 1], but f (x) is not Riemann integrable
on [−1, 1], since it is not bounded on this interval.
∫b
Example 14. “If f (x) is Riemann integrable on [a, b], then a f (x) dx =
F (b) − F (a), where F (x) is an antiderivative of f (x) on [a, b].”
Solution. ∫1
The function f (x) = sgn x is Riemann integrable on [0, 1] and 0 sgn xdx =
1, but there is no antiderivative of f (x) on [0, 1] (see Example 2 in section 5.2
and Example 4 in this section for details).
Remark to Examples 13 and 14. The statements in Examples 13 and 14 are

weakened (and false) versions of one of the formulations of the Fundamental
Theorem of Calculus: if f (x) is Riemann integrable on [a, b] and F (x) is an
∫b
antiderivative of f (x) on [a, b], then a f (x) dx = F (b) − F (a).
Example 15. “If f (x) is Riemann integrable on [a, b], then there exists
∫b
c ∈ [a, b] such that a f (x) dx = f (c) · (b − a).”
Solution. {
0, x < 0
The Heaviside function H (x) = is Riemann integrable (see
1, x ≥ 0
∫1
Remark 2 to Example 2) on [−1, 1] and −1 H (x) dx = 1, but there is no point
where H (x) is equal to 12 .
Remark. The additional condition of the continuity of f (x) on [a, b] will
turn this false statement in one of the versions of the Mean Value Theorem
for integrals.
Example 16. “If f (x) is continuous on [a, b], and φ (t) is continuously
∫β ∫b
differentiable on [α, β], then α f (φ (t)) φ′ (t) dt = a f (x) dx, where a =
φ (α) and b = φ (β).”
Solution.
Let us consider the following calculations
2 applying this “rule”. First, cal-
∫2 2 ∫2 2 t3
culate −1 t dt directly: −1 t dt = 3 = 3. Next, consider the change of
−1
the variable using the continuously differentiable function x = φ (t) = t2 with
φ′ (t) = 2t. The lower limit α = −1 corresponds to a = φ (−1) = 1 and the
upper β = 2 to b = φ (2) = 4. Now expressing all elements in the terms of
∫4 √
variable x we get the right-hand side integral in the form 1 12 xdx with the
√
integrand f (x) = 12 x continuous on [1, 4]. Calculating the last integral we
Integrals 151
∫4 √ 4
obtain: 1 12 xdx = 31 x3/2 1 = 73 . Notice, that the conditions of the statement
hold, but two results are different.
Remark. The correct formulation of the change of variable rule is: if f (x) is
continuous on [a, b], and φ (t) is continuously differentiable on [α, β] with image
∫β ∫b
[a, b] and φ (α) = a, φ (β) = b, then α f (φ (t)) φ′ (t) dt = a f (x) dx. In the
false statement of this Example, the condition of the connection between the
domain of f (x) and the image of φ (t) was omitted, and it gives a wrong result
illustrated by the presented counterexample, where the image of φ (t) = t2
when t ∈ [−1, 2] is [0, 4], while the interval [φ (−1) , φ (2)] = [1, 4] is only a
part of the image. Of course, the correct rule provides sufficient conditions,
and in some cases when these conditions are not satisfied the formula may
still be true, but it may also happen that the formula will not work, as in the
counterexample above.
Example 17. “If f (x) and g (x) are continuous on [a, b] and con-
∫b
tinuously differentiable on (a, b), then a f (x) g ′ (x) dx = f (x) g (x)|a −
b
∫b ′
a
f (x) g (x) dx.”
Solution. √
The functions f (x) = x and g (x) = − cos x are continuous on [0, 1]
and continuously differentiable on (0, 1) (with the derivatives f ′ (x) = 2√1
x
,
′
g (x) = sin x), but the formula
∫ 1 ∫
√ √ 1 1 1 cos x

x sin xdx = − x cos x 0 + √ dx
0 2 0 x
is not valid, because the integral in the right-hand side does not exist (the
function cos√ x is not bounded on [0, 1], and so it is not Riemann integrable).
x
Remark 1. One of the correct statements of the integration by parts rule
is the following: if f (x) and g (x) are continuously differentiable on [a, b], then
∫ b ∫ b
′
f ′ (x) g (x) dx.
b
f (x) g (x) dx = f (x) g (x)|a −
a a
While these are the sufficient conditions, we can see that a weak violation
of them at one of the endpoints can lead to a wrong result. Of course, since
the conditions of the rule are sufficient, it may happen that they are violated,
√
but the rule is still working. For instance, using the same f (x) = x and
changing the second function to g (x) = sin x, we keep the same properties of
two functions on [0, 1], but the formula of the integration by parts holds in
this case: ∫ 1 ∫
√ √ 1 1 1 sin x
x cos xdx = x sin x0 − √ dx.
0 2 0 x
sin
√x
Although x
is not defined at 0, but it is Riemann integrable, because this
sin
√x
function is continuous on (0, 1] and lim x
= 0, which indicates just a re-
x→0
movable discontinuity at 0.
∫1
Remark 2. The integral 0 cos √ x dx still exists if it is considered as an
x
improper integral. In this generalized context of the integrability, the formula
∫ 1 ∫ 1
√ √ 1 cos x
x sin xdx = − x cos x0 + √ dx
0 0 2 x
is correct. (Improper integrals will be considered in section 5.4.)
Example 18. “If f (x) is Riemann integrable on [a, b], then the value of
the Riemann integral does not change if one alters the values of f (x) at a
countable number of points.”
Solution. {
1, x ∈ Q
The functions f (x) ≡ 0 and D (x) = differ only at rational
0, x ∈ I
points, that is on a countable set, however f (x) is Riemann integrable on any
[a, b] and D (x) is not.
Remark. The statement becomes true if both functions are Riemann in-
tegrable.
Example 19. “If f (x) is continuous on [a, b] except at only one point,
then f (x) is Riemann integrable on [a, b].”
Solution.
The function f (x) = tan x has the only discontinuity point x = π2 on [0, π],
but f (x) is not Riemann integrable because it is not bounded on [0, π].
Remark. The statement becomes true if f (x) is additionally bounded on
[a, b].
5.4 Improper integrals

∫ +∞
Example 1. “If a
f (x) dx converges, then lim f (x) = 0.”
x→+∞
Solution.
Consider
√ f (x) = cos x2 on [1,√+∞). By choosing two sequences of points
- xk = 2kπ, k ∈ N and xn = 2nπ + π, n ∈ N - we obtain two different
partial limits at infinity:
lim f (xk ) = lim cos 2kπ = 1

xk →+∞ k→+∞
and
lim f (xn ) = lim cos (2nπ + π) = −1.
xn →+∞ n→+∞
Integrals 153
∫ +∞
Therefore, lim f (x) does not exist. However, 1
f (x) dx converges. In-
x→+∞
deed, representing
( )′
2 sin x2 sin x2
f (x) = cos x = + ,
2x 2x2
we obtain for any b > 1:

∫ b ∫ b ( )′ ∫
sin x2
sin x2 sin b2 sin 1 1 b sin x2
cos x2 dx = dx = + − + dx.
1 1 2x2
2x 2b 2 2 1 x2
∫ +∞ 1
2
The last integral converges, because sinx2x ≤ x12 and the integral 1 x2 dx

2 2
converges. Also lim sin2bb = 0, because sin2bb ≤ 2b 1
→ 0 . Therefore, the
b→+∞ b→+∞
∫ +∞
integral 1 cos x2 dx converges.
∫ +∞
Remark 1. The statement: “if a f (x) dx converges, then f (x) is
bounded at infinity (that is f (x) is bounded on [c, +∞) for some c > a) ”, is
also false. Evidently, if lim f (x) = 0 then f (x) is bounded at infinity, and
x→+∞
the converse is not true. So this is a strengthened (with the same conditions,
but a weakened conclusion) version of the original statement. The function
f (x) = x cos x4 on [1, +∞) provides a counterexample to the last statement.
Indeed, by changing the variable t = x2 the improper
∫ +∞ integral of x∫cos x4 can be
+∞
reduced to the improper integral of cos x : 1 x cos x4 dx = 21 1 cos t2 dt.
2
√
Hence, the integral converges, √ but for the sequence xk = 2kπ, ∀k ∈ N it
4
follows lim f (xk ) = lim 4 2kπ cos 2kπ = +∞, that is, f (x) = x cos x4
xk →+∞ k→+∞
is unbounded at infinity.
Remark 2. The readers familiar with infinite series can note the contrast
between this statement and the known fact that a general term of a convergent
infinite series must tend to zero. (Infinite series will be considered in section
6.3.)
∫ +∞
Remark 3. The converse statement: “if lim f (x) = 0, then a f (x) dx
x→+∞
converges”, is also false. This is a more elementary result and
∫ +∞a 1simple
counterexample is f (x) = x1 on [1, +∞): lim x1 = 0, but 1 x dx =
x→+∞
lim ln b = +∞ diverges.
b→+∞
Remark 4. The related correct result states that if lim f (x) = A ̸= 0,
x→+∞
∫ +∞
then a f (x) dx diverges.
Example 2. “If f (x) is non-negative, continuous and unbounded at infin-

∫ +∞
ity (that is f (x) is unbounded on [c, +∞) for any c > a), then a f (x) dx
diverges.”
Solution.
First let us construct an auxiliary function with a small non-zero support.

Integrals 155
Consider an arbitrary n ∈ N and choose the interval An = [n − εn , n + εn ]

centered at the point n, where εn > 0 is the radius of An . The auxiliary
function with the support An is defined as

 (x − n + εn ) · εn , x ∈ [n − εn , n]
n
gn = − (x − n − εn ) · εnn , x ∈ [n, n + εn ] ,

0, x ∈
/ An
that is gn has the form of a tooth of a saw, with the support An and height n,
located above the x-axis. It is easy to evaluate the area of this tooth (triangle):
∫ ∫ n+εn
gn (x) dx = gn (x) dx = εn n.
An n−εn
1
In particular, if we choose εn = n·10 n , that is, the support radius fast decrease
∫ n+εn
along the x-axis, then n−εn gn (x) dx = 101n . Now we construct the function
{
gn , x ∈ An , ∀n ∈ N ∞
defined on [0, +∞): f (x) = , where A = ∪ An . Evi-
0, x ∈/A n=1
dently, f (x) is non-negative, continuous (as gluing together the line segments)
and unbounded at infinity (since f (n) = n). However, the improper integral
∫ +∞
0
f (x) dx converges. In fact, due to the evaluation of the tooth areas we
have: ∫ +∞ ∑∞
1 1
f (x) dx = n
=
0 n=1
10 9
(the sum of the infinite geometric progression), that is, the improper integral
converges.
Remark 1. If the conditions of the statement are weakened, then of course
the statement will remain false and the constructed counterexample will work,
but at the same time a simpler counterexample can arise. For instance, for
the false statement: “if f (x) is a positive and unbounded at infinity, then
∫ +∞
a
f (x) dx diverges” (the condition
{ of continuity is dropped here), there is
e−x , x ∈/N
a simpler counterexample: f (x) = .
n, x ∈ N
Remark 2. It is not difficult to construct a counterexample to a strength-
ened (but still false) version of the above statement: “if f (x) is positive,
∫ +∞
continuous and unbounded at infinity, then a f (x) dx diverges”. For ex-
ample, taking the advantage of the{constructed counterexample, we can sum
gn , x ∈ An , ∀n ∈ N
the constructed function f (x) = with any positive
0, x ∈/ A = ∪n∈N An
function, say g (x) = x21+1 , continuous on [0, +∞) and whose improper inte-
gral
∫ +∞converges. According to the arithmetic properties of improper integrals,
0
h (x) dx, h (x) = f (x) + g (x), is also convergent and h (x) is a positive,
continuous and unbounded at infinity function.

Integrals 157
FIGURE 5.4.5: Example 2, Remark 2 and Example 3
Example 3. “If f (x) is positive and continuous on [a, +∞), and

∫ +∞
a
f (x) dx converges, then f (x) is bounded at infinity.”
Solution.
This statement is closely related to that of Example 2, and the function
{
gn , x ∈ An , ∀n ∈ N 1
h (x) = f (x) + g (x) , f (x) = , g (x) = 2
0, x ∈/ A = ∪n∈N An x +1
on [0, +∞) constructed in Remark 2 of Example 2 provides a counterexample
to the current statement also (the reasoning is the same).
Remark. Of course, any weakened version of this statement is also false.
For instance, for the statement: “if f (x) is positive and continuous on [a, +∞),
∫ +∞
and a f (x) dx converges, then lim f (x) = 0”, the same counterexample
x→+∞
can be applied (evidently, if lim f (x) = 0 then f (x) is bounded at infinity,
x→+∞
so the last statement is weaker than the original).
∫ +∞ ∫ +∞
Example 4. “If a f (x) dx converges, then a |f (x)| dx also con-
verges.”
Solution. [ )
Let us consider f (x) = sinx x on π4 , +∞ . Integrating by parts we obtain
∫ +∞ +∞ ∫ +∞
sin x 1 cos x
dx = − cos x − dx.
π/4 x x π/4 π/4 x2
x ∫
Since the last integral converges ( cos
x2
≤ 12 and +∞ 12 dx converges) and
x π/4 x
∫ +∞
lim cosx x = 0 ( cosx x ≤ x1 → 0), it follows that π/4 sinx x dx converges.
x→+∞ x→+∞
Now let us evaluate the improper integral for |f (x)| = sinx x :
∫ ∫ ∫
+∞ sin x 3π/4
1/2 7π/4
1/2
dx ≥ dx + dx + . . .
x x x
π/4 π/4 5π/4
∫ 3π/4 ∫ 5π/4 ∫ 7π/4 ∫∫ +∞9π/4

1/4 1/4 1/41/4 1/4
≥ dx+ dx+ dx+
dx+. . . = dx.
π/4 x 3π/4 5π/4x 7π/4 x x π/4 x
∫ +∞
Since the last improper integral diverges, it follows that π/4 sinx x dx is di-
vergent.
Remark. The converse is correct.
∫ +∞
Example 5. “If a f (x) dx converges and g (x) is bounded on [a, +∞),
∫ +∞
then a f (x) g (x) dx also converges.”
Solution.
We can reduce this case to that of Example 4 by choosing g (x) in such
a way that it will mimic the action of the absolute value. Let f (x) = sinx x
[ ) ∫ +∞
on π4 , +∞ . It was already shown in Example 4 that π/4 sinx x dx converges.
Integrals 159
Choose now
{
1, x ∈ [2kπ, π + 2kπ) , k ∈ N
g (x) = .
−1, x ∈ [π + 2kπ, 2π + 2kπ) , k ∈ N
∫ +∞
Then f (x) g (x) = sinx x and it was shown above that π/4 sinx x dx diverges.
∫ +∞
Remark. If additionally a |f (x)| dx converges, then the statement will
be true.
∫ +∞ ∫ +∞
Example 6. “If both a f (x) dx and a g (x) dx converge, then
∫ +∞
a
f (x) g (x) dx also converges.”
Solution.
The counterexample is: f (x) = cos √ x and g (x) = cos√
3 x
x
on [1, +∞). Fol-
x
∫ +∞ cos x
lowing the reasoning of Example 4, it can be shown that 1 √ dx and
∫ +∞ cos x ∫ +∞ cos x cos x x
1
√
3 x dx are convergent. At the same time, 1
√
x
· √3 x dx is divergent
2
due to the following considerations: cos x
= 2x15/6 + cos 2x
, and the integral
∫ +∞ 1 ∫ +∞ cos 2xx5/6 2x5/6
dx diverges, while 1 dx converges.
1 2x5/6 2x5/6 ∫ +∞
Remark. The following statement is also false: “if a f (x) dx converges,
∫ +∞ 2
then a f (x) dx also converges”. The corresponding counterexample can
be given with the function f (x) = cos
√ x considered on [1, +∞).
x
∫ +∞
Example 7. “If a f (x) dx diverges and g (x) is unbounded on [a, +∞),
∫ +∞
then a f (x) g (x) dx also diverges.”
Solution. √ ∫ +∞ 1
Let us consider f √(x) = x1 and g (x) = x sin x on [1, +∞). 1 x dx
diverges and g (x) = x sin x is unbounded on [1, +∞), since for xn = π2 +
√π
2nπ, n ∈ N it follows that lim g (xn ) = lim 2 + 2nπ = +∞. So the
xn →+∞ n→+∞
conditions of the statement are satisfied. Now, let us evaluate the improper
integral of the product:
∫ +∞ ∫ +∞ +∞ ∫
sin x cos x 1 +∞ cos x
f (x) g (x) dx = √ dx = − √ − dx.
1 1 x x 1 2 1 x3/2
∫ +∞ cos x cos x
√ x √1 ≤ 3/2
Since lim cos √ x = 0 ( cos
x x ≤ x ) and 1 x 3/2 dx converges ( x3/2 x
1
x→+∞
∫ +∞ 1 ∫ +∞ 1 √
and 1 x3/2
dx converges), we can conclude that 1 x x sin xdx converges.
∫ +∞ ∫ +∞
Example 8. “If a f (x) + g (x) dx converges, then a f (x) dx and
∫ +∞
a
g (x) dx also converge.”
Solution.
∫ +∞ 1 ∫ +∞ −1 ∫ +∞ 1 −1 ∫ +∞
Both 1 x dx and 1 x dx diverge, while 1 x + x dx = 1 0dx =
0 converges.
Remark. Similar examples can be given for other arithmetic operations.
∫ +∞ 1 −1
For instance, for the product the same functions can be used: 1
∫ +∞ 1 x · x dx =
− 1 x2 dx is a convergent integral.
Remark. We have focused here on improper integrals of the first kind, that
is the integrals considered on infinite intervals on the x-axis (the integrand do-
main is unbounded). Another kind (the second) of improper integrals is related
to infinite intervals on the y-axis (the integrand image is unbounded), while the
domain is bounded. Since two kinds of improper integrals are intimately con-
nected (there is a simple change of variable that transforms the second kind of
improper integrals to the first kind and vice-versa), we do not repeat here sim-
ilar counterexamples, which can be constructed for improper integrals of the
second kind as well. To take the advantage of the counterexamples constructed
for integrals of the first kind, just recall the formulas that connect two kinds of
integrals: if f (x) is continuous on [a, +∞) (without a loss of generality we can
assume that a > 0), then by changing the variable t = x1 , an improper integral
∫ +∞
of the first kind a f (x) dx can be reduced to the improper integral of the
∫ b (1) 1
second kind 0 f t t2 dt , b = a1 , where the lower limit is a singular point
(that is the integrand is unbounded in any right-hand neighborhood of 0, but it
is bounded on any interval [c, b], 0 < c < b). Moreover, both integrals converge
∫ +∞ ∫b ( )
(or diverge) simultaneously and are equal: a f (x) dx = 0 f 1t t12 dt . Us-
ing these formulas, some of the presented examples can be easily reformulated
for improper integrals of the second kind. For instance, Example 1 can be re-
∫b
formulated as follows: “if 0 g (t) dt converges, then lim g (t)·t2 = 0”, with the
t→0+
corresponding counterexample g (t) = t12 cos t12 defined on (0, 1]. At the same
time, some examples assume simpler form when solved directly for improper
integrals of the second kind. For instance, Example 6 can be reformulated in
∫b ∫b ∫b
the form: “if both 0 f (t) dt and 0 g (t) dt converge, then 0 f (t) g (t) dt also
1
converges”. However, the counterexample with the functions f (t) = t3/2 cos 1t
1
and g (t) = t5/3 cos 1t , obtained by transformation of the functions in Example
6, is somewhat involved. A much simpler counterexample for this statement
can be provided with the functions f (t) = √1t and g (t) = √ 1
3 2 on (0, 1].
t
5.5 Applications
Example 1. “If a function is continuous on [a, b] and continuously differ-

entiable on (a, b), then its graph is a rectifiable curve.”
Solution. {
x sin x1 , x ∈ (0, 1]
The function f (x) = is differentiable on (0, 1)
0, x = 0
Integrals 161
according to the arithmetic and chain rules, and the derivative f ′ (x) =
sin x1 − x1 cos x1 is continuous on (0, 1), according to the arithmetic and com-
position rules. The function is also continuous at the endpoints of [0, 1] :
at 1 by the arithmetic and composition rules and at 0 by the definition -
lim x sin x1 = 0 = f (0). Hence, f (x) satisfies the conditions of the statement
x→0+
on [0, 1]. However, the graph of f (x) is not a rectifiable curve. Indeed, the
graph represents infinitely many oscillations with decreasing heights toward
to 0, accumulating near the y-axis. The locations of the crests of these os-
cillations are given by the formula xn = π+4nπ2
, n ∈ N and their heights are
hn = xn , n ∈ N. It is obvious that the length of one oscillation Ln (the length
of the part of the curve between two crests) is greater than the corresponding
wave height hn . So the sum of the lengths of the first N oscillations can be
evaluated as follows:
∑
N ∑
N ∑
N
2 2 ∑1
N
Ln > hn = > .
n=1 n=1 n=1
π (1 + 4n) 5π n=1 n
Choosing N = 2k , we obtain the following evaluation for the last sum:
∑N
1 1 1
SN = = 1 + + ... + k
n=1
n 2 2
( ) ( )
1 1 1 1 1 1 1
=1+ + + + + + + + ...
2 3 4 5 6 7 8
( )
1 1 1
+ + + . . . +
2k−1 + 1 2k−1 + 2 2k
1 1 1 1 1
>1+ + 2 · + 4 · + . . . + 2k−1 · k = 1 + k → +∞.
2 4 8 2 2 k→∞
Therefore, the chosen subsequence of the sums SN∑ , N = 2k , is divergent, that
N 2
implies the divergence of the sequence SN . Since n=1 Ln > 5π SN , it follows
∑N
that n=1 Ln → +∞, that is the length of the graph is infinite and the
N →∞
curve is not rectifiable. (The sums SN are the partial sums of the harmonic
series; see more about numerical series in section 6.3).
{ 2Remark 1. Another interesting counterexample is the function g (x) =
x sin x12 , x ∈ (0, 1]
, which is even smoother than f (x) above, because it
0, x = 0
is differentiable on [0, 1] { and continuously differentiable on (0, 1). However
′ 2x sin x12 − x2 cos x12 , x ∈ (0, 1]
the derivative g (x) = is unbounded near
0, x = 0
0, that opens the possibility for the curve to be non-rectifiable, and really
the curve has an infinite length, that can be shown by applying the same
arguments as for f (x).
Remark 2. This statement is a wrongly “weakened” version of one of the
theorems on rectifiable curves: if f (x) is continuously differentiable on [a, b],

then its graph is a rectifiable curve. Of course, this is a sufficient condition, but
we can see that in some cases the violation of the differentiability condition
only at one of the endpoints can result in a non-rectifiable curve.
Example 2. “If the graph of a continuous on [a, b] function has a finite

length, then this function is continuously differentiable on [a, b].”
Solution.
First of all, a very simple counterexample is f (x) = |x| on [−1, 1], with
the function non-differentiable
√ at 0, but the length of the graph is well-defined
and equal to 2 2.
A much more interesting example occurs if we require that a function
be non-differentiable at infinitely many points in [a, b]. To construct such
example, consider a “saw” function located in the interval [0, 1] with “teeth”
of variable height and width. More precisely, the n-th “tooth” is defined by
the formula:
{ ( )
xn−1/2 −xn (x − xn ) , x ∈ xn , (xn−1/2
hn
fn (x) = ) ,
xn−1/2 −xn−1 (x − xn−1 ) , x ∈ xn−1/2 , xn−1
hn
where hn is the height, [xn , xn−1 ] is the base (xn−1 − xn being the width),
and xn−1/2 is the centerpoint (note that xn−1 > xn ). The entire func-
tion is defined as follows: consider an infinite partition of the interval [0, 1]
Integrals 163
in the form 1 = x0 > x1 > . . . > xn > xn+1 > . . . > 0, with the
only limit point 0 ( lim xn = 0); define the saw function in the form
{ n→∞
fn , x ∈ [xn , xn−1 ] , n ∈ N
f (x) = . By construction, f (x) is continuously
0, x = 0 ( ) ( )
differentiable on every interval xn , xn−1/2 and xn−1/2 , xn−1 , ∀n ∈ N, but
it is not differentiable at the points xn−1/2 and xn , ∀n ∈ N. Moreover, if
a sequence of the heights hn is chosen in such a way that hn → 0, then
n→∞
f (x) is continuous on [0, 1]. To specify the saw function we choose xn−1 = n1 ,
xn−1/2 = xn−12+xn , and hn = 2(n+1) 1
2 , ∀n ∈ N. Let us show that the graph
of f (x) is a rectifiable curve. Noting that the half-width of each tooth is

∆n = xn−1/2 − xn = xn−1 − xn−1/2 = 2n(n+1) 1
, we obtain for the length of
each tooth:
√( )2 √( )2
Ln = xn−1 − xn−1/2 + hn + 2 xn−1/2 − xn + h2n
√ √
√ 1 1 2
= 2 ∆2n + h2n =
2 2 + 4 ≤ n (n + 1) .
n (n + 1) (n + 1)
The sum of the lengths of the first N teeth is readily evaluated in the form:
√ ( )
∑
N ∑
N
2 √ 1 √
Ln ≤ = 2 1− ≤ 2.
n=1 n=1
n (n + 1) N +1
Since the inequality

√ holds for ∀N ∈ N, it implies that the length of the graph
is limited by 2, and we conclude that the curve is rectifiable.
Remark. We see that f (x) is not even piecewise smooth function (that is
a function which is continuously differentiable on all the closed subintervals,
which form a finite partition of [a, b]).
Example 3. “If a plane figure is quadrable (that is the area of a plane

figure exists and is finite), then its boundary is a rectifiable curve.”
Solution.
A simple example is related to improper integrals, involving a figure un-
bounded in the horizontal (improper integral of the first kind) or vertical
(improper integral of the second kind) direction. Let us consider f (x) = e−x
on [0, +∞). The graph of f (x) is an unbounded curve, which has an infinite
length:
∫ +∞ √ ∫ √2
−2x
t2
L= 1+e dx = dt
0 1 t2 − 1
( )√2 √
1 t − 1 √ 1 2−1 1 t − 1
= t + ln = 2 − 1 + ln √ − lim ln = +∞
2 t + 1 1 2 2 + 1 2 t→1+ t + 1
(to calculate this integral we have used the change of the variable: t =
√ ( )
1 + e−2x , x = − 12 ln t2 − 1 , dx = − t2tdt
−1 ). However, the area of the fig-
ure below the graph of f (x) (above the x-axis and at the left of the y-axis) is
finite: ∫ +∞
( )+∞
A= e−x dx = −e−x 0
= 1.
0
More interesting, albeit a bit more involved example, is related to a
bounded figure, defined by a bounded function on a finite interval [a, b]. Let
us consider the saw function of the previous Example 2 with the following
specification of the heights and widths of teeth: xn−1 = n1 , xn−1/2 = xn−12+xn ,
and hn = 2n1
, ∀n ∈ N. Like in Example 2, f ((x) is continuous
) (on [0, 1] and )is
continuously differentiable on every interval xn , xn−1/2 and xn−1/2 , xn−1 ,
∀n ∈ N, but it is not differentiable at the points xn−1/2 and xn , ∀n ∈ N. Let us
show that the graph of f (x) is a non-rectifiable curve. Indeed, for the length
of each tooth we have: Ln ≥ 2hn = n1 , and consequently for the sum of the
∑N ∑N
lengths of the first N teeth we obtain: n=1 Ln ≥ n=1 n1 . Using the eval-
∑N
uation made in Example 1 for the sequence SN = n=1 n1 we conclude that
this sequence diverges lim SN = +∞. Hence, the constructed saw curve is
N →∞
non-rectifiable. At the same time, continuity and non-negativity of the func-
tion f (x) guarantee that the area between the graph of f (x) and the x-axis
(with the lateral sides x = 0 and x = 1) exists and can be expressed in the
∫1
form of the Riemann integral A = 0 f (x) dx.
Integrals 165
Remark 1. For those familiar with infinite series or willing to consult the
required results in the theory of series, the last integral can be ( easily calcu-
)
∫1 ∑∞
lated: A = 0 f (x) dx = n=1 An , where An = 21 2∆n · hn = 14 n12 − n(n+1) 1
∑∞ 2 ∑∞
is the area of the n-th tooth. Since n=1 n12 = π6 and n=1 n(n+1) 1
= 1, we
∑∞ ( 2
)
get A = n=1 An = 4 6 − 1 . (Infinite series will be considered in section
1 π
6.3.)
Remark 2. Another interesting
{ example is a slight modification of the func-
x sin2 x1 , x ∈ (0, 1]
tion in Example 1: f (x) = . Just like in Example 1, it
0, x = 0
can be readily shown that this function is continuous on [0, 1] and differentiable
on (0, 1]. Also the graph of f (x) has the crests located at xn = π+2nπ 2
,n∈N
and heights hn = xn , n ∈ N. So the same evaluation for the sum of the lengths
of the first N oscillations can be used:
∑
N ∑
N ∑
N
2 2 ∑1
N
Ln > hn = > .
n=1 n=1 n=1
π (1 + 2n) 3π n=1 n
∑N
Recalling that SN = n=1 n1 → +∞, we conclude that the graph of this
n→∞
function is a non-rectifiable curve. On the other hand, the area bounded by
the graph of f (x) and the x-axis (with the lateral sides x = 0 and x = 1)
∫1
exists and is equal to A = 0 f (x) dx, because the function f (x) is continuous
and non-negative on [0, 1]. It is easy to see that this area is less then 1/2, since
the figure is contained within the triangle with the vertices (0, 0), (1, 0) and
(1, 1).
Remark 3. According to the topics of this section we confine ourselves
to the cases when a plane figure is defined by a real function of one variable
f (x). However, there is a variety of other examples related to the definition of
plane curves in another form (parametric, geometric, etc.). For example, the
Koch snowflake, T-Square, Dragon curve and many other fractals have these
properties, some of them (for example, the Sierpinski gasket) even have a zero
area along with infinite boundary length.
Example 4. “If f (x) is continuous on [a, b], then the area between the
graph of f (x) and the x-axis (and the lateral walls x = a and x = b) can be
∫b
found by the formula A = a f (x) dx.”
Solution.
A simple counterexample is f (x) = x on [−1, 1]. In this case the area of
the figure is the sum of the areas of two equal triangles: A = 2 · 21 = 1, but
∫1
the Riemann formula gives −1 xdx = 0.
Remark. The condition of non-negativity of f (x) is missing in the
statement conditions, or without non-negativity the correct formula is A =
∫b
a
|f (x)| dx.

Integrals 167
Example 5. “If a plane figure has a finite area, then the corresponding
solid of revolution has a finite surface area.”
Solution.
Let us consider the plane figure between the graph of the function f (x) =
− ln x considered on (0, 1], the segment [0, 1] of the x-axis and the y-axis. This
unbounded figure has a finite area:
∫ 1 ∫ 1
1
A= − ln xdx = −x ln x|0 + dx
0 0
ln x 1/x
= 1 + lim x ln x = 1 + lim = 1 + lim / =1
x→0+ x→0+ 1/x x→0+ −1 x2
(the integration by parts was applied to the first integral and L’Hospital’s rule
was applied to calculate an indeterminate form). At the same time, the solid
of revolution, obtained by rotating this plane figure about the x-axis, has an
infinite surface area:
∫ 1 √ ∫ 1
1 ln x
S = 2π (− ln x) 1 + 2 dx > −2π dx
0 x 0 x
1
= −π ln2 x0 = π lim ln2 x = +∞.
x→0+
Remark 1. It is simple to show that the volume of this solid of revolution

is finite:
∫ 1 ∫ 1 ∫ 1

2 1

2 1 1
V = π ln xdx = π x ln x 0 −2π ln xdx = π x ln x 0 −2π x ln x|0 +2π dx
2
0 0 0
ln2 x − 2 ln x 2 ln x · 1/x − 2 · 1/x

= 2π − π lim = 2π − π lim /
x→0+ 1/x x→0+ −1 x2
ln x − 1 1/x
= 2π − 2π lim = 2π − 2π lim / 2 = 2π − 2π lim x = 2π
x→0+ −1/x x→0+ 1 x x→0+
(the integration by parts was applied twice to find the integral, and then
L’Hospital’s rule was used twice to calculate an indeterminate form).
Remark 2. It is easy to see that the length of the corresponding curve is
infinite:
∫ 1√ ∫ 1 ∫ 1 ∫ 1
1 x2 + 1 x dx
L= 1 + 2 dx = √ dx = √ dx + √
0 x 2
0 x x +1 0
2
x +1 0 x x +1
2
√ 1 ∫ +∞ √ ( √ )+∞
dt
= x + 1 +
2 √ = 2 − 1 + ln t + t2 + 1
0 2
t +1 1
1
√ ( √ ) ( √ )
= 2 − 1 − ln 1 + 2 + lim ln t + t2 + 1 = +∞
t→+∞
(the change of variable x = 1/t was applied to one of the integrals).
Example 6. “If a solid of revolution has a finite volume, then the plane
figure, which gave rise to this solid, is quadrable.”
Solution.
Let us consider f (x) = x1 on [1, +∞). For any b > 1, the curve on the
∫b√
subinterval [1, b] has the length L (b) = 1 1 + x14 dx, and the area of the
∫b
region between the graph and the x-axis is A (b) = 1 x1 dx = ln b. For the same
subinterval [1, b], the surface area and the volume of the solid of revolution
(obtained by rotating this region around the x-axis) are, respectively:
∫ b √
1 1
S (b) = 2π 1 + 4 dx > 2π ln b
1 x x
and ∫ ( )
b
1 1
V (b) = π dx = π 1 − .
1 x2 b
Letting b → +∞, we obtain the infinite plane region (with an infinite length of
two boundaries) with an infinite area A = lim A (b) = +∞, which generates
b→+∞
the solid of revolution with an infinite surface area S = lim S (b) = +∞,
b→+∞
Integrals 169
but with the finite volume V = lim V (b) = π. (This solid of revolution is
b→+∞
called Gabriel’s horn or Torricelli’s trumpet).
Remark 1. The integral of the surface area can be calculated directly:
∫ +∞ √ ∫
1 1 π 1 1√
S = 2π 1 + 4 dx = 1 + tdt
1 x x 2 0 t
∫ √ ( )√2
2
u2 1 u − 1
=π du = π u + ln
1 u2 − 1 2 u + 1 1
√
(√ ) π 2 − 1 π u − 1
= +∞.
=π 2 − 1 + ln √ − lim ln
2 2 + 1 2 u→1+ u + 1
Here, in the first integral we have applied the change of the variable t = x14 ,
dx = − 14 t−5/4 dt and in the second integral one more change of the variable
√
u = 1 + t, t = u2 − 1, dt = 2udu.
Remark 2. Similar
( ] results can be obtained for the family of functions
f (x) = x1α , α ∈ 12 , 1 on [1, +∞).

Exercises
1. Suppose that the Riemann integral of f (x) on [a, b] is defined as follows:

∫b
a
f (x) dx = lim S (f ), where S(f ) is the Riemann sum corresponding
n→+∞
to a partition with n subintervals. Is this definition correct? If not, give a
counterexample. For what kind of partitions are the tendencies n → +∞ and
∆ → 0 equivalent?
2. Suppose that the condition of independence of the limit of the Riemann
sums from the choice of points ci ∈ [xi−1 , xi ] is not required. Show by example
that this will violate the definition of the Riemann integral.
3. Give a counterexample to the statement: “if f (x) is Riemann integrable
1
and non-zero on [a, b], then f (x) is also integrable on [a, b]”. What condition
should be added to make this statement correct? Generalize the statement,
counterexample and additional correctness condition on the case of the ratio
of two Riemann integrable functions.
4. The known theorem states that a continuous on [a, b] function f (x) is
∫b
identically zero on [a, b] if, and only if, a f (x)g(x)dx = 0 for any continuous on
[a, b] function g(x). Will this result remain true if the choice of g(x) is restricted
to the polynomials of degree less or equal n? If not, provide a counterexample.
What if g(x) are chosen among the polynomials of all degrees?
5. It is well-known that for a periodic with a period T and Riemann inte-
∫ c+T ∫T
grable on any [a, b] function f (x) it holds c f (x) dx = 0 f (x) dx for any
c. Disprove the converse statement using a counterexample. What if f (x) is
additionally continuous on R?
6. Provide a counterexample to the statement: ∫x “if f (x) is a periodic and
continuous on R function, then F (x) = 0 f (t) dt is also periodic”. What
minimum condition can be added to make this statement correct?
7. It is well-known
∫c that for an odd, continuous on [−c, c], c > 0 function
f (x) it holds −c f (x) dx = 0. Disprove the converse statement using a coun-
terexample. What if this property holds for any c > 0?
8. Construct a counterexample to the following formulation of the change
of variable: “if f (x) is continuous on [a, b], and φ (t) is differentiable and
∫b ∫β
increasing on [α, β] with image [a, b] , then a f (x) dx = α f (φ (t)) φ′ (t) dt.
(Hint: consider f (x) ≡ 1 on [0, 1] and φ(t) defined on [0, 1] as follows

 0, t = 0( ]
3n , t ∈ 2n , 2n−1n − n3nn−1 ,n∈N
1 1 1 1
φ(t) = .
 1 2 πn3 (t−1/2 ) ( 1 ]
3n + 2
3n cos 2 , t ∈ 2n−1 − 1
,
n3n 2n−1
1
, n ∈ N
It can be shown that φ(t) is continuous, differentiable and increasing on [0, 1],
however f (φ(t))φ′ (t) = φ′ (t) is not integrable since it is not bounded on [0, 1].)
9. It is well-known that for a continuous on [a, b] function f (x) the integral
Integrals 171
∫x
with the variable upper limit F (x) = a f (t)dt is a differentiable function on
[a, b] and F ′ (x) = f (x). Show that if the condition of continuity is relaxed to
integrability on [a, b], then the statement is not true anymore. On the other
hand, if f (x) is not continuous at c ∈ [a, b], it still may happen that F (x) is
differentiable at c. Formulate both results as counterexamples. (Hint: use the
Heaviside function
{ on [−1, 1] and show that F (x) is not differentiable at 0;
0, x = 0
use f (x) = on [−1, 1] and show that F (x) is differentiable at
cos x1 , x ̸= 0
0 .)
10. Use sgn x on a corresponding interval as another counterexample to
the statement in Example 15, section 5.3.
11. The generalized Mean Value theorem states that if f (x) is continuous
and g(x) is integrable and sign-preserving on [a, b], then there exists c ∈ [a, b]
∫b ∫b
such that a f (x)g(x)dx = f (c) a g(x)dx. Show that the condition of conti-
nuity of f (x) cannot be relaxed and the condition of sign-preserving cannot
be omitted even for continuous g(x). Formulate these results in the form of
counterexamples. (Hint: for continuous but not sign-preserving function use
g(x) = sin x, f (x) = x on [−π, π].)
12. It is well-known that for a Riemann integrable on [a, b] function f (x)
∫c ∫b
there is a point c ∈ [a, b] such that a f (x)dx = c f (x)dx. Use an example
to show that it is not always possible to choose c ∈ (a, b). (Hint: consider
f (x) = sin x on [0, 2π].)
( ) ∫b
13. If f (x) is a convex function on [a, b] then f a+b
2 (b − a) ≤ a f (x)dx ≤
f (a)+f (b)
(b − a). Show that the converse is not true.
2 ∫ +∞
14. Show that the statement “if a f (x) dx diverges, then lim f (x) =
x→+∞
A ̸= 0 ” is not correct.
15. Suppose f (x) is continuous and monotone on [a, +∞). Show that the
∫ +∞
condition lim xf (x) = 0 does not imply the convergence of a f (x) dx.
x→+∞
(Notice that the converse is true.)
16. Suppose f (x) is continuous and monotone on (0, b] and 0 is a singular
point of f (x). Show that the condition lim xf (x) = 0 does not imply the
x→0+
∫b
convergence of 0 f (x) dx. (Notice that the converse is true.) Compare with
the previous exercise.
17. Provide a counterexample to the following statement: “if f (x) ≤ g(x)
∫ +∞ ∫ +∞
on [a, +∞) and a g(x)dx converges, then a f (x)dx also converges”.
What if the condition f (x) ≤ g(x) is substituted by |f (x)| ≤ |g(x)| on
[a, +∞)? Compare with the General Comparison theorem. (Hint: for the first
case choose f (x) = −x and g(x) = x12 on [1, +∞); and for the second try
f (x) = | sin
2x
x|
and g(x) = sinx x on [1, +∞).)
18. Provide a counterexample to the following statement: “if f (x) and
∫b
g(x) are integrable on any [c, b] ⊂ (a, b] , f (x) ≤ g(x) on (a, b] and a g(x)dx
∫b
converges, then a f (x)dx also converges”. What if the condition f (x) ≤ g(x)
is substituted by |f (x)| ≤ |g(x)| on (a, b]? Give a correct formulation of the

General Comparison theorem for the improper integrals of the second type.
(Hint: for simplicity suppose that f (x) and g(x) are continuous on (a, b] and
follow the suggestions in the previous exercise.)
19. Show that the following statement is false: “let f (x) be integrable on
∫b
any [c, b], c ∈ (0, b) and 0 be a singular point of f (x); if 0 f (x)dx converges,
then |f (x)| < x1 ” (Hint: try f (x) = sin(1/x)x3/2
on (0, 1].)
20. Show that the following statement is false: “let f (x) be integrable on
∫b
any [c, b], c ∈ (0, b) and 0 be a singular point of f (x); if 0 f (x)dx diverges,
then |f (x)| > x1 ” (Hint: consider f (x) = x ln 1 1
x on (0, e .) ∫b
21. Suppose f (x) is continuous on (a, b], the improper integral a f (x) dx
converges and g(x) is continuous on [a, b]. Is it sufficient to guarantee the
∫b
convergence of a f (x)g(x)dx ? (Hint: consider f (x) = √1x3 sin x1 and g(x) =
{
0,
√ x = 01 on [0, 1].)
x sin x , x ̸= 0
∫b
22. Does the convergence of the improper integral a f (x)dx imply the
∫b
convergence of a |f (x)|dx? If not, formulate in the form suitable for a coun-
terexample and provide one.
23. Show by example that the convergence of the improper integrals
∫b ∫b ∫b
a
f (x)dx and a g(x)dx does not guarantee the convergence of a f (x)g(x)dx.
24. Show by example that the divergence of the improper integrals
∫b ∫b ∫b
a
f (x)dx and a g(x)dx does not imply the divergence of a f (x)g(x)dx.
(Hint: choose x ∈ (0, 1] and consider the functions
{
fn (x), x ∈ [xn − δn , xn + δn ]
f (x) =
0, otherwise
and {
gn (x), x ∈ [xn+1/2 − δn , xn+1/2 + δn ]
g(x) = ,
0, otherwise
1 1 xn +xn+1
where xn = n, δn = 8n2 , xn+1/2 = 2 ,
π
fn (x) = 2nπ cos( (x − xn )) · H(x − xn + δn )H(xn + δn − x),
2δn
π
gn (x) = 2nπ cos( (x − xn+1/2 )) · H(x − xn+1/2 + δn )H(xn+1/2 + δn − x),
2δn
∀n ∈ N, and H(x) is the Heaviside function.)
∫ +∞
25. By definition, the principal value of the improper integral −∞ f (x)dx
∫a ∫ +∞
is lim −a f (x)dx . It is evident, that the convergence of −∞ f (x)dx imply
a→+∞
the convergence of its principal value. Show by example that the converse is
not true. Construct a similar definition for the principal value of the improper
integral of the second type and give a corresponding counterexample.
Integrals 173
26. Elaborate technical details of the proof for the counterexample in Ex-
ample 1, section 5.5. {
x cos x1 , x ∈ (0, 1]
27. Show that the graph of the function f (x) = on
0, x = 0
[0, 1] has an infinite length, but the area of the corresponding plane figure
between the graph and x-axis is finite. Thus, this is one more counterexample
to the statement in Example 3, section 5.5.
28. Show that the plane region between the graph of f (x) = √1x and x-axis
on the interval (0, 1] has the finite area, but the corresponding solid formed
by revolving this region around x-axis has an infinite volume. Formulate this
result as a counterexample. Show that this solid of revolution has an infinite
surface area. Therefore, it also provides a counterexample to the statement in
Example 5, section 5.5 (although in Example 5 the volume is finite).
29. Show that the graph of f (x) = x2x+1 considered on [0, +∞) generates
the plane figure (between the graph and x-axis) with infinite area and corre-
sponding solid of revolution (rotating around x-axis) with infinite surface area,
but finite volume. Thus, this is one more counterexample to the statement in
Example 6, section 5.5.
Chapter 6
Sequences and series
Numerical sequences
Remark. We will consider here only infinite real sequences.
Sequence definition. A sequence is a function whose domain is N. The

∞
standard notation is: a1 , a2 , . . . , an , . . . or {an }n=1 , or an , n ∈ N or simply an
when this notation is clear within a specific context.
Remark. The definition is usually extended to domains which include
additional integers or exclude some naturals. The point in these extensions is
to maintain the domain with the properties of ordering of N: all the elements
of a sequence should be indexed (with an integer index), the first element
with an initial index must exist, and each following element has the index
equal to the index of the preceding element plus one. (Evidently the number
of non-natural indices is finite or empty).
Convergent sequence. A sequence an is convergent if there exists a

finite limit lim an = A. Otherwise a sequence is divergent.
n→∞
Remark 1. A standard concept of the limit at infinity, slightly adopted
to sequences, is used here: for every ε > 0 there exists a natural number
(index) N such that for all natural numbers (indices) n > N it follows that
|an − A| < ε.
Remark 2. For the sequences the only limit point in domain is positive
infinity, so the notation n → +∞ is frequently abbreviated to n → ∞.
175
Subsequences. The subsequence of a sequence an is a new sequence

composed of some elements of the original sequence, keeping their order.
Monotonicity and boundedness. In the same way as for general func-

tions, the properties of monotonicity (increasing/decreasing) and bounded-
ness/unboundedness are defined for sequences.
The usual properties (arithmetic and comparison) of the limits

are valid for the convergent sequences (see the list of properties in subsection
2.1.2).
Additional (specific) properties of convergent sequences:

1) If lim f (x) = A and an = f (n), then lim an = A.
x→+∞ n→∞
2) If a sequence an is increasing (decreasing) and bounded above (below), then
it is convergent.
3) A convergent sequence an is bounded.
4) Any bounded sequence has a convergent subsequence.
Cauchy’s criterion. A sequence an converges if, and only if, for any ε > 0
there exists N such that for all n > N and all p ∈ N it holds |an+p − an | < ε.
Numerical series: convergence and elementary properties
Remark. We will consider here only infinite real series.
Series definition. The sum ∑ of all the elements

∑+∞ of a sequence is called
+∞
series. The standard
∑ notation is: n=1 a n or n=i an , where i is the initial
index, or simply an .
Partial sums. The sum∑ of the first k terms of a series is called the k-th
∑k+i−1
k
partial sum. Notation: Sk = n=1 an or Sk = n=i an .
Remark. Even when a series
∑k starts from the index i, the k -th partial
sum can be defined as Sk = n=i an . The choice of one of two options for
nominating the partial sums has no influence on the main properties of series,
including convergence/divergence.
Convergence. A series is convergent if the sequence of its partial sums
is convergent, that is lim Sk = S, where S is called the sum of the series.
k→∞ ∑
The used notation for a convergent series: an = S. Otherwise, the series is
divergent.
∑
Divergence test. If an is convergent, then lim an = 0.
n→∞
∑ ∑
Arithmetic
∑ properties. If an∑and bn are convergent, then so are
the series can (c is a constant) and (an + bn ) and, furthermore,
Sequences and series 177
∑ ∑
1) ∑ can = c an∑ ; ∑
2) (an + bn ) = an + bn .
Theorem. Removal or addition of a finite number of terms does not affect

the convergence or divergence of a series.
∑
Cauchy criterion. A series an is convergent if, and only if, for ∀ε > 0
there is a natural
number N such that for ∀n > N and ∀p ∈ N it follows that
∑n+p
k=n+1 ak < ε.
Numerical series: convergence tests
∑
Convergence of geometric series. A geometric series q n (q = const)
is convergent if |q| < 1 and divergent if |q| ≥
∑1.1
Convergence of p-series. A p-series np (p = const) is convergent if
p > 1 and divergent if p ≤ 1.
The Integral test. Suppose f (x) is positive and decreasing function on

∫ +∞
[1, +∞) and an = f (n) , n ∈ N. Then ∑ the improper integral 1 f (x) dx is
convergent if, and only if, the series an is convergent.
∑ ∑
The Comparison test. Suppose an and bn are positive series (that
is, series
∑ with positive terms). Then ∑
1) if ∑ bn is convergent and an ≤ bn , n ∈ N, then∑ an is also convergent;
2) if bn is divergent and an ≥ bn , n ∈ N, then∑ an is also
∑ divergent.
The Limit Comparison test. Suppose an and bn are positive
series. If lim abnn = c, where c is a positive constant, then either both series
n→∞
converge or both diverge.
The Leibniz (alternating series) test. If an alternating series

∑ n−1
(−1) bn , bn > 0, n ∈ N satisfies the two conditions
1) bn+1 ≤ bn , n ∈ N
2) lim bn = 0,
n→∞
then the series is convergent.
∑
Absolute convergence. A ∑ series an is called absolutely convergent if
the series of the absolute values |an | is convergent.
Theorem. If a series is absolutely convergent, then it is convergent.
Conditional convergence. A series is conditionally convergent if it is

convergent, but not absolutely convergent.
∑ 1
The Cauchy (root) test. Let an be a positive series and lim ann =
n→∞
C. Then: ∑
1) if C < 1, then ∑ an converges;
2) if C > 1, then an diverges;
3) if C = 1 or the limit does not exist, then the test is inconclusive.
∑
The D’Alembert (ratio) test. Let an be a positive series and
lim aan+1
n
= D. Then:
n→∞ ∑
1) if D > 1, then ∑ an is convergent;
2) if D < 1, then an is divergent;
3) if D = 1 or the limit does not exist, then the test is inconclusive.
∑
The( Raabe )test. Consider a series an , an > 0 and suppose that
lim n · an+1 − 1 = R. Then:
an
n→∞ ∑
1) if R > 1, then ∑ an is convergent;
2) if R < 1, then an is divergent;
3) if R = 1 or the limit does not exist, then the test is inconclusive.
∑
The Bertrand
( ( test.
) Consider
) a positive series an and assume that
lim ln n · n · aan+1
n
− 1 − 1 = B. Then:
n→∞ ∑
1) if B > 1, then ∑ an converges;
2) if B < 1, then an diverges;
3) if B = 1 or the limit does not exist, then the test is inconclusive.
Remark 1. All the tests for positive series can also be used for checking
the absolute convergence of arbitrary series.
Remark 2. In all convergence results, the properties required to be satisfied
for all indices, can be actually satisfied for all indices starting from some index
N.
Power series
∑∞ n
Power series. The series n=0 cn (x − a) , where cn , n ∈ N ∪ {0} and
a are real numbers, is called a power series centered at a (or simply a power
series). The numbers cn are called the coefficients and a the centerpoint of the
series.
∑∞ n
Convergence theorem. For a given power series n=0 cn (x − a) there
are only three options:
1) the series converges only when x = a,
2) the series converges for all x,

3) there is a positive number R such that the series converges for |x − a| < R
and diverges for |x − a| > R.
The number R is called the radius of convergence. By convention, the
radius of convergence is R = 0 in the first case and R = ∞ in the second case.
∑∞ n
Differentiation and integration. If the power series n=0 cn (x − a)
has
∑∞radius of convergence R > 0, then the function f (x) defined by f (x) =
n
n=0 cn (x − a) is differentiable on (a − R, a + R) and
∞
∑
f ′ (x) =
n−1
ncn (x − a) ,
n=1
∫ ∞
∑ n+1
(x − a)
f (x) dx = cn + C.
n=0
n+1
The last two formulas show that the power series can be differentiated
and integrated term by term within the interval of convergence. The radius
of convergence of the last two series is R. The immediate consequence of
this result is that f (x) represented by a convergent power series is infinitely
differentiable inside the interval of convergence.
∑+∞ ∑+∞
n n
Theorem. If two power series n=0 cn (x − a) and n=0 bn (x − a)
are convergent on the same interval and the set of the points of this interval,
on which they take the same values, has a finite limit point contained within
the interval of convergence, then an = bn , ∀n.
Theorem. A function can have only one expansion in a power series
centered at the chosen point. ∑+∞ ∑+∞
Theorem. If the radii of convergence of n=0 an xn and n=0 bn xn are
∑+∞
Ra and Rb , respectively, then the radius of convergence of n=0 (an + bn ) xn
is
1) Rc = min {Ra , Rb } if Ra ̸= Rb ;
2) Rc ≥ min {Ra , Rb } if Ra = Rb .
∑∞ n
Taylor coefficients. If f (x) = n=0 cn (x − a) is convergent for
(n)
f (a)
|x − a| < R, R > 0, then cn = n! , ∀n ∈ N ∪ {0}.
∑∞ f (n)(a) n
Taylor series. The power series representation f (x) = n=0 n ! (x−a)
is called the Taylor series of f (x) (centered at a).
6.2 Numerical sequences
Remark. A numerical sequence is a function with a special domain. There-

fore, many properties of sequences related to monotonicity, boundedness and
limits (considering limits at infinity) are just repetitions of the corresponding
properties of general functions. Hence, many counterexamples presented for
functions in sections 1.3, 1.6, 2.2 and 2.3 can be easily adopted for the case
of sequences. For this reason, we will not present in this section all such re-
formulations, but just give a few of them and also provide some additional
examples specific for the case of sequences.
Example 1. “If a sequence is bounded, then it is convergent.”

Solution.
n
Consider the sequence an = (−1) , n ∈ N. This is a bounded sequence
n
(|(−1) | ≤ 1, ∀n ∈ N), but it has no limit, because all even elements of the
sequence are equal to 1 and odd elements to −1. So whatever number a would
be chosen as a candidate for the limit value, there are elements of the sequence
with arbitrary large indices that have distance to a greater than ε = 1/2, that
is the definition of the limit is not satisfied.
Another proof of the non-existence of limit can be made by using the
relation between a general limit of a sequence and limits of its subsequences
(partial limits): since the limits of two subsequences are different lim a2n =
n→∞
lim 1 = 1 and lim a2n+1 = lim (−1) = −1 (the first containing the even
n→∞ n→∞ n→∞
terms and the second - the odd terms), the limit of the original sequence does
not exist.
Remark. The converse statement is true: if a sequence has a finite limit,
then it is bounded.
Example 2. “If a sequence an is unbounded, then lim an = (±) ∞.”

n→∞
(see Example 9, section 2.3 for comparison)
Solution.
n
The sequence an = (1 + (−1) ) n is unbounded, because lim a2n =
n→∞
lim 4n = +∞, but it has no limit (either finite or infinite), because another
n→∞
partial limit is zero: lim a2n+1 = lim 0 = 0.
n→∞ n→∞
Example 3. “If both sequences an and bn are divergent, then the sequence
an + bn is also divergent.”
(see Example 1, section 2.3 for comparison)
Solution.
n n+1
Let us choose an = (−1) and bn = (−1) . Then both sequences diverge
(see Example 1 for details), but the sequence cn = an + bn = 0 is convergent
to zero.
Remark 1. Analogous constructions can be made for other arithmetic
operations. In particular, the product of the same divergent sequences is the
n n+1
convergent sequence dn = (−1) · (−1) = −1.
Remark 2. Under the conditions of the statement, the opposite conclusion
n
is also false. For instance, if an = bn = (−1) , then the sequence cn = an +bn =
n
2 · (−1) is also divergent.
Example 4. “If a sequence an diverges and bn converges, then the sequence

an · bn is divergent.”
(see Examples 5 and 6, section 2.3 for comparison)
Solution.
n
For the divergent an = (−1) and convergent bn = n1 sequences, their
n
product cn = (−1) n is a convergent sequence, since the even and odd partial
limits are equal:
1 −1
lim c2n = lim = 0 = lim = lim c2n+1 .
n→∞ n→∞ 2n n→∞ 2n + 1 n→∞
Remark 1. The simplest example is when bn = 0.

Remark 2. The statement will be true if to add that lim bn ̸= 0.
n→∞
Remark 3. Actually, it is easy to prove that for any choice of convergent
and divergent sequences their sum is always divergent. The same is true for
the difference, but not for the product and ratio as it is shown in the provided
counterexample.
Example 5. “If the ranges of two convergent sequences an and bn coincide,

then lim an = lim bn .”
n→∞ n→∞
Solution. { {
0, n = 1 1, n = 1
The sequences an = and bn = are convergent
1, n > 1 0, n > 1
and have the same ranges, but lim an = 1 ̸= 0 = lim bn .
n→∞ n→∞
Example 6. “If a sequence an is bounded, positive and (an+1 − an ) → 0,

n→∞
then an is convergent.”
Solution. √
The sequence an = 2 + sin n π2 is bounded (located between 1 and 3) and
positive, and
√ π √ π
an+1 − an = sin n + 1 − sin n
2 2
π (√ √ ) π (√ √ )
= 2 sin n + 1 − n cos n+1+ n
4 4
π 1 π (√ √ )
= 2 sin √ √ cos n+1+ n → 0
4 n+1+ n 4 n→∞
(the first term sin π4 √n+1+

1 √
approaches 0, when n approaches infinity, and
π
(√ n
√ )
the second term cos 4 n + 1 + n is bounded). At the same time, this
2
sequence is divergent, because for n = (2k) , k ∈ N one has an = 2+sin 2k π2 =
2
2, while for n = (4m + 1) , m ∈ N one gets an = 2 + sin (4m + 1) π2 = 3.
Remark. The converse is true: if an is convergent, then an is bounded and
(an+1 − an ) → 0.
n→∞
Example 7. “If a sequence an is bounded, positive and aan+1

n
→ 1, then
n→∞
an is convergent.”
Solution. √
The sequence an = 2 + sin n π2 of the previous Example 6 is bounded and
positive, and satisfies the property
√ √
an+1 sin n + 1 π2 − sin n π2
√
an − 1 = 2 + sin n π2

(√ √ )
2 sin π √ 1 √ cos π n + 1 + n
π 1
√ → 0.
4 n+1+ n 4
= √ π ≤ 2 sin √
2 + sin n 2 4 n + 1 + n n→∞
However, as it was shown in Example 6, this sequence is divergent.
Remark. The converse is also false. Indeed, the sequence an =
{
1/n , n = 2k
, k ∈ N is convergent to zero and positive. However,
2/n, n = 2k + 1
the limit of an does not exist, because lim aa2k+1
an+1
2k
2·2k
= lim 2k+1 = 2 and
k→∞ k→∞
a2k 2k−1
lim = lim = 12 .
k→∞ a2k−1 k→∞ 2k·2
Example 8. “If lim an bn = 0, then lim an = 0 or lim bn = 0 (or

n→∞ n→∞ n→∞
both).”
Solution.
n n
The sequences an = 1 + (−1) and bn = 1 − (−1) have no limit (see
a similar Example( 1 for details),
) but the sequence an bn approaches zero:
2n
lim an bn = lim 1 − (−1) = 0.
n→∞ n→∞
Example 9. “If a sequence an has the property that lim (an+p − an ) =

n→∞
0 , ∀p ∈ N (p is fixed), then an converges.”
Solution.
Consider an = 1 + 21 + . . . + n1 . Let us evaluate the following subsequence
1 1
a2k = 1 +
+ ... + k
2 2
( ) ( )
1 1 1 1 1 1 1
=1+ + + + + + + + ...
2 3 4 5 6 7 8
( )
1 1 1
+ + + ... + k
2k−1 + 1 2k−1 + 2 2
1 1 1 1 1
>1+ + 2 · + 4 · + . . . + 2k−1 · k = 1 + k → +∞.
2 4 8 2 2 k→∞
Therefore, the chosen subsequence is divergent, that implies the divergence
of the original sequence. At the same time, the condition of the statement is
satisfied:
( )
1 1 1
lim (an+p − an ) = lim + + ... + =0
n→∞ n→∞ n + 1 n+2 n+p
for any fixed natural number p.

Remark
√ 1. Another interesting example, involving a bounded sequence, is
an = cos π n. For any fixed p ∈ N one has
( √ √ √ √ )
n+p− n n+p+ n
lim (an+p − an ) = lim −2 sin π sin π = 0,
n→∞ n→∞ 2 2
since the last multiplier is bounded and

√ √ √ √ √ √
n+p− n ( n + p − n) ( n + p + n)
lim sin π = lim sin π √ √
n→∞ 2 n→∞ 2 ( n + p + n)
pπ
= lim sin √ √ = 0.
n→∞ 2 ( n + p + n)
However, the sequence is divergent because for n = 4m2 , m ∈ N one has
2
lim a4m2 = lim cos 2mπ = 1, while for n = (2m + 1) , m ∈ N it follows
m→∞ m→∞
that lim a(2m+1)2 = lim cos (2m + 1) π = −1.
m→∞ m→∞
Remark 2. This is a wrong reformulation of Cauchy’s criterion: a sequence
an converges if, and only if, for any ε > 0 there exists N such that for all n > N
and all p ∈ N it holds |an+p − an | < ε.
Example 10. “If lim an = 0 and lim bn = +∞, then lim an bn = 0.”
n→∞ n→∞ n→∞
(see Examples 2 and 5, section 2.3 for comparison)
Solution.
Let an = n1 and bn = n. Then lim an = 0 and lim bn = +∞, but
n→∞ n→∞
lim an bn = lim n1 n = 1.
n→∞ n→∞
Remark 1. This is a false extension of the product rule for limits.
Remark 2. This Example is similar to Example 2, section 2.3, and an
indeterminate form 0 · ∞ treated here is one of seven indeterminate forms
mentioned in Remark 3 to that Example. Let us see that the situation is really
undefined, that is the limit of the product strongly depends on a specification
of the sequences an and bn :
1) if an = nc (c = const) and bn = n, then lim an = 0, lim bn = +∞ and
n→∞ n→∞
lim an bn = lim c = c.
n→∞ n→∞
2) if an = n1 and bn = n2 , then lim an = 0, lim bn = +∞ and lim an bn =
n→∞ n→∞ n→∞
lim n = +∞.
n→∞
3) if an = − n1 and bn = n2 , then lim an = 0, lim bn = +∞ and lim an bn =
n→∞ n→∞ n→∞
lim (−n) = −∞.
n→∞
(−1)n
4) if an = n and bn = n, then lim an = 0, lim bn = +∞ and
n→∞ n→∞
n
lim an bn = lim (−1) does not exist.
n→∞ n→∞
So, just like in Example 2, section 2.3, an indeterminate form 0 · ∞ can
result in an arbitrary constant (including zero), infinity or even non-existence
of the limit.
6.3 Numerical series: convergence and elementary prop-

erties
∑
Example 1. “If lim an = 0, then an converges.”
n→∞
Solution. ∑1
The harmonic series lim n1 = 0. The divergence can
n diverges, but n→∞
∑N
be shown by evaluating the partial sums HN = n=1 an with the indices
N = 2k :
1 1
H2k = 1 + + . . . + k
2 2
( ) ( )
1 1 1 1 1 1 1
=1+ + + + + + + + ...
2 3 4 5 6 7 8
( )
1 1 1
+ + + ... + k
2k−1 + 1 2k−1 + 2 2
1 1 1 1 1
>1+ + 2 · + 4 · + . . . + 2k−1 · k = 1 + k → +∞.
2 4 8 2 2 k→∞
Remark 1. The converse is true and represents the necessary condition
of convergence for∑numerical series. Formulated in an equivalent form: if
lim an ̸= 0, then an diverges, the converse is called the Divergence test.
n→∞
Remark 2. The following strengthened version is also false: “if lim an =
∑ n→∞
0 and the sequence an is decreasing, then an is convergent”. The same
harmonic series gives a counterexample.
Example 2. “If lim an = 0 and the sequence of the partial sums is

∑ n→∞
bounded, then an is convergent.”
Solution.
Let us consider the following sequence:
1 1
a1 = 1 , a2 = a3 = − , a4 = a5 = a6 = ,
2 3
1 n−1
. . . , a n(n−1) +1 = . . . = a n(n+1) = (−1)
,... .
2 2 n
We have lim an = 0 and the sequence of the partial sums Sk is bounded
n→∞ ∑
by 0 and 1, but the series an diverges, because the partial sums have
no limit: for the partial sums with indices k = 1, 6, . . . , (2n−1)2n
2 , . . . , ∀n ∈
N one gets Sk = 1 → 1, while for the partial sums with indices k =
k→∞
2n(2n+1)
3, 10, . . . , 2 , . . . , ∀n ∈ N one obtains Sk = 0 → 0.
k→∞
Remark 2. For positive series, the condition of boundedness of partial

sums is necessary and sufficient for series convergence.
∑
Example 3. “If lim nan = 0, then an converges.”
n→∞
Solution.
1
Consider an = n ln lim nan = lim ln1n = 0, but
n , n > 2. Evidently, n→∞
∑ 1 n→∞
1
the series n ln n is divergent according to the integral test: f (x) = x ln x is
positive and decreasing on [3, +∞) and the improper integral
∫ +∞ ∫ +∞
1 +∞
f (x) dx = dx = ln (ln x)|3 = +∞
3 3 x ln x
is divergent.
∑
Example 4. “If an converges , then lim nan = 0.”
n→∞
Solution. ∑ n 1
The alternating series (−1) √n is convergent (both conditions of the
n√
Leibniz test are satisfied), but lim nan = lim (−1) n = ∞ (the limit does
n→∞ n→∞
not exist if one considers the specific infinity - positive or negative).
Remark 1. The statement is still false if the condition of the positivity
of an is added. In this case the counterexample is provided by the sequence
{
/ 2 n = k , k ∈ N . Indeed, lim nan does not exist, since for n =
2
1/n,
an =
1 n , otherwise n→∞
k , k ∈ N, we have lim nan = lim k k2 = 1, while for n = k 2 + 1, k ∈ N,
2 2 1
n→∞( )k→∞1
we get lim nan = lim k 2 + 1 (k2 +1) 2 = 0. At the same time, the following
n→∞ k→∞
evaluation for partial sums holds:
1 1 1 1 1 1
Sk 2 = 1 + + 2 + + ... + ( )2 + . . . + 2 +
22 3 4 2
(k − 1) + 1 (k 2 − 1) k2
∑k 2
∑
k ∑∞
1 1 1 π2
< + < 2 = 2 .
n=1
n2 n=1 n2 n=1
n2 6
2
Since the series is positive, it follows that Sk < π3 , ∀k ∈ N, that implies
convergence of the series.
Remark 2. The statement becomes true if an is a non-negative decreasing
sequence.
∑ ∑ ∑
Example 5. “If an and bn are convergent, then an bn is also con-
vergent.”
Solution.
n n 1
If an = (−1) √1n and bn = (−1) √ 3 n , then the alternating series
∑ n 1 ∑ n 1
(−1) √n and (−1) √ 3 n are convergent since both conditions of the
Leibniz test are satisfied for these two series:

1 1 1
lim |an | = lim √ = 0, |an+1 | = √ < √ = |an | , ∀n ∈ N,
n→∞ n→∞ n n+1 n
and
1 1 1
lim |bn | = lim √ = 0, |bn+1 | = √3
< √ = |bn | , ∀n ∈ N.
n→∞ n→∞ 3 n n+1 3
n
∑ ∑ 1
However the series an bn = n5/6
is divergent (it is a p-series with p = 56 ).
Remark 1. As a particular case of this ∑ example, one can also ∑ formu-
late the following false statement: “if a series an converges then a2n also
converges”. The corresponding counterexample ∑ is inn line with the above: if
n
an = (−1) √1n , then the alternating series (−1) √1n is convergent, but
∑1
the series of the squares n is harmonic, that is divergent.
Remark 2. For positive series this statement is true.
∑ ∑ ∑
Example 6. “If an and bn are divergent, then an bn is also diver-
gent.”
Solution. ∑ ∑
If an = n1 and bn = √1n , then an and bn are divergent (p-series with
∑ 1
p ≤ 1), but n3/2
is convergent (a p-series with p > 1).
Remark. As a particular case of this

∑ example, one can ∑also formulate the
following false statement: “if a series an diverges then a2n also diverges”.
The
∑ 1 corresponding counterexample is similar to the ∑ above: the harmonic series
1
n is divergent , but the series of the squares n 2 is convergent (this is a
p-series with p > 1).
∑ ∑ ∑
Example 7. “If an and bn are divergent, then (an + bn ) is also
divergent.”
Solution. ∑ ∑ ∑ ∑
If an = −bn = n1 , then an and bn diverge, but (an + bn ) = 0
converges.
Remark
∑ 1. This ∑is a wrong “negative” version
∑ of the following arithmetic
rule:∑if an and ∑bn are ∑ convergent, then (an + bn ) is also convergent
and (an + bn ) = an + bn .
Remark 2. If additionally both series are positive, then the statement is
true.
∑ ∑ ∑
Example 8. “If an converges and bn diverges, then an bn di-
verges.”
Solution. ∑ 1
If an = n12 and bn = √1n , then n2 converges (a p-series with p > 1), and
∑ 1 ∑ 1
√ diverges (a p-series with p ≤ 1), but converges (a p-series with
n n5/2
p > 1).
Remark 1. The statement with the ∑ opposite conclusion is also false: if
n n n
an =∑ (−1) n1 and bn = (−1) , then (−1) n1 converges (by Leibniz test),
∑1
n
and (−1) diverges (the general term does not approach zero), but n
diverges (the harmonic series).
Remark
∑ 2. The corresponding
∑ statement for∑the sum and difference is
true: if an converges and bn diverges, then (an ± bn ) diverges.
∑ ∑
Example 9. “If an diverges and bn is unbounded, then an bn di-
verges.”
Solution. ∑
n√
Let an = n1 and bn = (−1) n. The harmonic series √ n1 is divergent and
the sequence bn is unbounded, because lim |bn | = lim n = +∞. However
∑ ∑ (−1)n n→∞ n→∞
the series an bn = √ is a convergent alternating series (according to
n
the Leibniz test).
∑∞ ∑∞
Example 10. “Let us consider
∑∞ two series n=1 an and ∑k=1 bk , where
∞
bk = a2k−1 + a2k , ∀k ∈ N. If k=1 bk converges, then so does n=1 an .”
Solution. ∑∞ ∑∞ n+1
Consider the divergent series ∑∞an = ∑
n=1 n=1 (−1) . The elements
∞
bk = 1 − 1 = 0, ∀k ∈ N. So the series k=1 bk = k=1 0 = 0 converges.
Remark 1. This counterexample shows that the associative property does
not
∑∞ work for series in the same way as for finite sums. However, if series
n=1 an is positive, then the statement is true.
∑ ∑
∑ Example 11. “If an and bn diverge, and an ≤ cn ≤ bn , ∀n ∈ N, then
cn also diverges.”
Solution. ∑ ∑
If an = − n1 and bn = 1, then the series an and bn diverge. Choosing
cn = n12 the last
∑ 1condition of the statement is also satisfied: − n1 ≤ n12 ≤ 1,
but the series n2 is convergent.
Remark 1. Under the given conditions, the statement with the opposite
∑ an = − n and bn = 1,
1
conclusion is also false. For instant, keeping the same
√1
but choosing cn = n , all the conditions holds, but cn diverges.
Remark 2. This statement ∑ is a wrong
∑ “negative” reformulation of the
following correct
∑ property: if a n and b n converge, and an ≤ cn ≤ bn ,
∀n ∈ N, then cn also converges.
6.4 Numerical series: convergence tests

∑ ∑
Example 1. “If an converges and bn diverges, then an < bn , ∀n.”
Solution. ∑ 1
If∑an = n12 and bn = − n1 , then n2 is a convergent p-series (with p > 1)
−1
n is the divergent negative harmonic series, but n2 > − n , ∀n. ∑
1 1
and
Remark. The statement is true for positive series (that is, series an
with positive terms an > 0, ∀n), and in this case it represents a part of the
Comparison test.
∑ ∑
Example 2. “If an converges and bn diverges, then |an | < |bn |, ∀n.”
Solution. ∑
n n
If an = (−1) √1n and bn = n1 , then the series (−1) √1n converges
(both conditions of the Leibniz test are satisfied: lim |an | = lim √1n = 0
∑ n→∞ n→∞
and |an+1 | = √n+1
1
< √1n = |an |, ∀n ∈ N) and 1
n diverges (the harmonic
series), but √1n ≥ n1 , ∀n.
Remark. The statement is true for positive series, and in this case it
represents a part of the Comparison test.
∑ n+1
Example 3. “Let (−1) an , an > 0 be an alternating series. If
lim an = 0, then the series converges.”
n→∞
Solution.
{
2/k, n = 2k − 1
Consider the alternating series with an = , k ∈ N.
1/k, n = 2k
Since lim a2k−1 = lim k2 = 0 and lim a2k = lim k1 = 0, the condition
k→∞ k→∞ k→∞ k→∞
lim an = 0 is satisfied. However, the series is divergent, because the even
n→∞
partial sums are equal to those of the harmonic series:
∑2k ( ) ( ) ( )
n+1 2 1 2 1 2 1
S2k = (−1) an = − + − + ... + −
n=1
1 1 2 2 k k
1 1 ∑
k
1
=1+ + ... + = = Hk → +∞.
2 k n=1 n k→∞
Remark. This is a weakened false version of the convergence test for al-
ternating series (the Leibniz test) where the condition an+1 < an , ∀n ∈ N is
missing.
∑ n+1
Example 4. “Let (−1) an , an > 0 be an alternating series. If
an+1 < an , ∀n ∈ N, then the series converges.”
Solution.
Consider the alternating series with an = 1+ n1 , n ∈ N. Although 1+ n+1
1
<
1 + n , ∀n ∈ N, but the series is divergent, because
1
( the
) necessary condition of
convergence is not satisfied: lim an = lim 1 + n1 = 1 ̸= 0.
n→∞ n→∞
Remark. This is a weakened false version of the convergence test for al-
ternating series (the Leibniz test) where the condition lim an = 0 is missing.
n→∞
∑ n
Example
∑ 5. “If (−1) an is convergent and 0 ≤ bn ≤ an for all n ∈ N,
n
then (−1) bn is also convergent.”
Solution. ∑
∞ (−1)n
The series √ converges according to the Leibniz test: an =
n=1 n
√1
→
n n→∞
0 and a n+1 = √ 1
n+1
≤ an = √1n for each n ∈ N. Also, 0 ≤ bn =
∑∞ (−1)n
√ 1
2 n+(−1)n
≤ √1n = an , for all n ∈ N. Nevertheless, the series n=1 2√n+(−1) n
diverges due to the following arguments: first rewrite this series in the form
∞
∑ n ∑∞ ( n )
(−1) 1 (−1) 1
√ n = √ −√ √ n ,
n=1
2 n + (−1) n=1
2 n n (2 n + (−1))
∑∞ (−1)n ∑∞ 1 √
and next note that √ converges, while
n=1 n n=1 2n+(−1)n n diverges
according to the comparison test for positive series, 1 √
because 2n+(−1) n
n
≥ 3n
1
for all n ∈ N.
Remark. This is a wrong extension of the comparison test (valid for posi-
tive series) onto the case of alternating series.

Example 6. “If lim (an+1 + an+2 + . . . + an+p ) = 0 for p = 1, 2, 3, . . .,

∑ n→∞
then the series an converges.”
Solution. ∑1 ( )
1 1 1
The harmonic series n is divergent, but n→∞
lim n+1 + n+2 + . . . + n+p =0
for p = 1, 2, 3, . . .
Remark. This is a wrong reformulation of Cauchy’s criterion for conver-
gence of series.
∑ ∑
Example 7. “If an converges and lim abnn = 1 , then bn also con-
n→∞
verges.”
Solution. ∑ ∑ (−1)n ∑
Let us consider the two alternating series an = √ and bn =
∑ ( (−1)n ) n
√ 1
n
+ n . The former is convergent according to the Leibniz test, and
the latter is divergent, since its general term is the sum of the terms
( of thencon-
)
vergent and divergent series. At the same time, lim an = lim 1 + (−1)
bn √
n
=
n→∞ n→∞
1.
Remark. The statement is true for positive series, and in this case it
represents a particular form of the Limit Comparison test.
∑
Example 8. “If an is a positive series (that is an > 0, ∀n) and lim aan+1
n
n→∞
does not exist, then the series is divergent.”
Solution. ∑ ∑ 2+(−1)n
The series an = n2 is positive and convergent (an ≤ n32
∑ 3 a2n+1 (2n)2
n2 is a convergent p-series). However a2n = 3(2n+1)2 → 3 , while
1
and
n→∞
2
a2n
a2n−1 =
3(2n−1)
→ 3, that is a general limit lim aan+1
(2n)2 n→∞ n
does not exist.
n→∞
Remark 1. The conclusion about convergence is also false. A similar coun-
∑ ∑ 2+(−1)n
terexample for this case is an = n which is a positive divergent
series, whose limit lim aan+1
n
does not exist.
n→∞
Remark 2. The correct statement is a part of the D’Alembert (ratio) test
asserting that under the given conditions no conclusion can be drawn with
regard to convergence/divergence of series.
Remark 3. The situation is the same for the Cauchy
∑ (root)
∑ test when
√ 1
lim n an does not exist. For example, the series an = n n is
(3+(−1) )
n→∞ ∑ 1
positive and convergent (an ≤ 21n and is a convergent geometric series),
2n √
√ √ 1
but a general limit lim n an does not exist, because 2n a2n = 2n 42n → 1,
n→∞√ n→∞ 4
√ ∑
1
while 2n+1 a2n+1 = 2n+1 22n+1 → 12 . On the other hand, the series an =
∑ n n
n→∞
n n
(3 + (−1) ) is positive and divergent (the general term an = (3 + (−1) )
√
does not approach zero), but again a general limit lim n an does not exist,
√ √ √ √
n→∞
42n → 4, while 2n+1 a2n+1 = 22n+1 → 2.
2n 2n+1
since 2n a2n =
n→∞ n→∞
∑
Example 9. “Let an be a positive series. If the D’Alembert (ratio) test
is not applicable, then the Cauchy (root) test does not work as well.”
Solution. ∑+∞
Let us consider the series n=1 2(−1) −n . The ratio test does not work,
n
an+1
because the limit in the test lim an does not exist: two partial se-
n→∞
a2n 21−2n a2n+1
quences provide different results - a2n−1 = 2−1−2n+1 = 2 > 1 and a2n =
2−1−2n−1 √
21−2n = 18 < 1. At the same time, the root test is applicable: lim n an =
n→∞
lim 2((−1) −n)/n = 12
n
< 1, that is the series converges.
n→∞
∑+∞ n−(−1)n
Another example can be given for the divergent series n=1 2 .
Again the D’Alembert test is not applicable, because lim aan+1
n
does not exist:
n→∞
a2n 22n−1 1 a2n+1 22n+1+1
a = 22n−1+1 = 2 < 1 and a2n = 22n−1 = 8 > 1. However, the Cauchy
2n−1
√ n
test shows divergence of the series: lim n an = lim 2(n−(−1) )/n = 2 > 1.
n→∞ n→∞
Remark. The converse is true: if the Cauchy test is not conclusive, then
neither is the D’Alembert test. This is because the existence of lim aan+1 =a
√ n→∞ n
implies that lim n an = a, but the converse does not hold.

n→∞
FIGURE 6.4.3: Example 9, the first counterexample
∑
Example 10. “Let an be a positive series. If the ratio and root tests
are not conclusive, then neither is any other test.”
Solution. ∑+∞
Let us consider the convergent p-series (p = 2) n=1 n12 . The limits of
n2
the ratio and root tests are equal to 1: lim aan+1 = lim (n+1)2 = 1 and
√ n→∞ n n→∞
√
lim n an = lim n n12 = 1. So both tests are inconclusive. However, con-
n→∞ n→∞
vergence can be established by applying the integral test with the positive
decreasing function f (x) = x12 , x ∈ [1, +∞), whose improper integral is con-
vergent: +∞
∫ +∞ ∫ +∞
1 1
f (x) dx = dx = − = 1.
1 1 x2 x 1
Or one can use the Raabe test, which also reveals convergence:
( ) ( ) ( )
an (n + 1)2 2 1
lim n − 1 = lim n − 1 = lim n + = 2 > 1.
n→∞ an+1 n→∞ n2 n→∞ n n2
Remark. Similar example
∑+∞can be provided for divergent series: divergence
of the p-series (p = 1/2) n=1 √1n can be shown by applying the integral test
with f (x) = √1x , x ∈ [1, +∞) or the Raabe test
( ) (√ ) √
an n+1 n 1
lim n − 1 = lim n √ − 1 = lim √ √ = < 1,
n→∞ an+1 n→∞ n n→∞ n+1+ n 2
an+1 √
but both the ratio and root tests do not work: lim an = lim n an = 1.
n→∞ n→∞
∑
Example 11. “Let an be a positive series. If the Cauchy, D’Alembert,
Raabe and Integral tests are not conclusive, then neither is any other test.”
Solution.
Let us consider an = e−(1+ 2 +...+ n−1 ) and the corresponding series
1 1
∑+∞ −(1+ 1 +...+ 1 )

n=2 e
2 n−1 . The limit of the D’Alembert test gives
e−(1+ 2 +...+ n )
1 1
an+1
= lim e− n = 1,
1
lim = lim
n→∞ an n→∞ −(1+ 12 +...+ n−1
1
) n→∞
e
that is the D’Alembert test is inconclusive. Since this limit exists, so does
√
the limit of the Cauchy test and the former and latter coincide lim n an =
n→∞
lim an+1 = 1, which means that the Cauchy test is also inconclusive. More-
n→∞ an
over, the Raabe test also fails:
( ) ( 1 )
an ex − 1
lim n · − 1 = lim n e n − 1 = lim+ = lim+ ex = 1.
n→+∞ an+1 n→+∞ x→0 x x→0
Finally, the Integral test is not applicable for this series (one cannot generate
an integrable function f (x) such that f (n) = an ). Nevertheless, the Bertrand
test shows that the series diverges:
( ( ) ) ( 1 )
an
lim ln n · n · − 1 − 1 = lim ln n ne n − (n + 1)
n→+∞ an+1 n→+∞
( )
1 x 1 ex − 1 − x
= lim+ (− ln x) e − − 1 = − lim+
x→0 x x x→0 x(ln x)−1
ex − 1 1 ex − 1
= − lim+ −1
· lim+ −1
= − lim+
x→0 (ln x) x→0 1 − (ln x) x→0 (ln x)−1
(ln x)2 2(ln x) 2.x

= lim ex · lim = lim −1
= lim =0<1
x→0+ x→0+ 1/x x→0 + −x x→0 + 1
(here we change the discrete index n by the continuous variable x in order to
apply L’Hospital’s rule).
Remark. There is the well-known result in the theory of series stating that
there is no definitive test for verification of convergence or divergence of all
series.
Example 12. “Let f (x) be continuous and non-negative on [1, +∞). If

∫ +∞ ∑∞
the improper integral 1 f (x) dx converges, then the series n=1 f (n) also
converges.”
Solution {
fn (x) , x ∈ Xn , n ∈ N
Consider the continuous function f (x) = with
[ ] 0, otherwise
fn (x) defined on Xn = n − 2n1 2 , n + 2n1 2 in the form
{ ( ) [ ]
2n2 x + 1(− 2n3 ,) x ∈ n[ − 2n1 2 , n ]
fn (x) = .
−2n2 x + 1 + 2n3 , x ∈ n, n + 2n1 2
The improper integral converges:

∫
1 ∑
+∞ +∞
1 1 1
f (x) dx = · 1 · + 2· ·1· 2
1 2 2 n=2 2 2n
) (
1 ∑ 1
+∞
1 1 π2 π2 1
= + 2
= + −1 = − .
4 n=2 2n 4 2 6 12 4
∑∞ ∑∞
However, the corresponding series is divergent: n=1 f (n) = n=1 1.
Remark 1. The function in the statement can even be made positive: for
the positive continuous function g (x) = f (x) + x12 the improper integral
∫ +∞ ∫ +∞ ∫ +∞
1
g (x) dx = f (x) dx + dx
1 1 1 x2
converges
∑∞ as the
∑∞sum ( of 1two
) convergent integrals, but the series diverges
n=1 g (n) = n=1 1 + n , since the general term does not approach 0.
2
Remark 2. Analogous statement about divergence: “if f (x) is continu-

∫ +∞
ous and non-negative on∑[1, +∞), and the improper integral 1 f (x) dx
∞
diverges, then the series n=1 f (n) also diverges”, is also false. The follow-
ing counterexample can be considered. Let f (x) = cos (π + 2πx) + 1. Then
∑∞ ∑∞
the series is convergent n=1 f (n) = n=1 0 = 0, but the improper integral
diverges: ∫ ∫
+∞ +∞
f (x) dx = 1 + cos (π + 2πx) dx
1 1
( )
1
= lim x+ sin (π + 2πx) − 1 = +∞.
x→+∞ 2π
Again the positivity of f (x) does not save the situation - just consider the
same modification as in Remark 1: g (x) = f (x) + x12 = cos (π + 2πx) + 1 + x12 .
Remark 3. It is possible to construct counterexamples to the main state-
ment and to the statement in Remark 2 with functions unbounded on [1, +∞).
Remark 4. The main statement and the statement in Remark 2 are weak-
ened false versions of the Integral test missing the condition of monotonicity
of the function f (x) on [1, +∞).
Example 13. “Let f (x) be continuous and non-negative on [1, +∞). If

∫ +∞ ∑∞
the improper integral 1 f (x) dx converges and the series n=1 f (n) also
converges, then f (x) → 0.”
x→+∞
Solution.
Consider the function constructed in Example 12 and move it the distance
of a half of unit to the right along the x-axis. The resulting function is deter-

{
˜ fñ (x) , x ∈ X̃n , n ∈ N
mined analytically as follows: f (x) = with fñ (x)
0, otherwise
[ ]
defined on X̃n = n − 2n1 2 + 12 , n + 2n1 2 + 12 in the form
{ ( ) ( ) [ ]
˜ 2n2 x(− 12 + ) 1(− 2n3 ,) x ∈ n[ − 2n1 2 + 12 , n + 12 ]
fn (x) = .
−2n2 x − 12 + 1 + 2n3 , x ∈ n + 12 , n + 2n1 2 + 12
The improper integral converges
∫
1 ∑ 1 1∑ 1
+∞ +∞ +∞
π2
f˜ (x) dx = + 2
= 2
=
1 2 n=2 2n 2 n=1 n 12
∑∞ ∑∞ ( )
as well as the series n=1 f˜ (n) = ˜ 1
n=1 0 = 0. However, f n + 2 =
1 ̸→ 0.
n→+∞
Remark 1. The function in the statement can even be made positive: for
the positive continuous function g̃ (x) = f˜ (x) + x12 the improper integral
∫ +∞ ∫ +∞ ∫ +∞
˜ 1
g̃ (x) dx = f (x) dx + 2
dx
1 1 1 x
converges
∑∞ as the
∑∞sum1 of two convergent integrals, and the p-series (p = 2)
n=1 g̃ (n) = n=1 n2 is also convergent.
Remark 2. It is possible to construct a counterexample to this statement
with a function unbounded on infinity.

6.5 Power series

∑+∞
Example 1. “A power series n=0 cn xn always has an interval of conver-
gence.”
Solution. ∑+∞
The power series n=1 nn xn converges only at the origin.
∑+∞
Example 2. “If series n=0 an xn converges at x = c, then it also con-
verges at x = −c.”
Solution. ∑
+∞ n
The series n=0 xn converges at x = −1, but diverges at x = 1.
Example 3. “If a function f (x) is infinitely differentiable on R, then the

∑+∞ (n)
power series n=0 f n !(0) xn converges on R.”
Solution.
2 is infinitely differentiable on R, but its Taylor
1
The function f (x) = 1+x
∑+∞ n 2n
series n=0 (−1) x converges only on (−1, 1).
Example 4. “If a function f (x) is infinitely differentiable on R, and the

∑+∞ (n) ∑+∞ (n)
series n=0 f n !(0) xn converges on R, then n=0 f n !(0) xn = f (x) , ∀x ∈ R.”
Solution. {
e−1/x , x ̸= 0 is infinitely differentiable on R
2
The function f (x) =

0, x = 0
and f (n) (0) = 0, ∀n ∈ N (see Example 6, section 4.2 for details). The Taylor
∑+∞ (n) ∑+∞
series n=0 f n !(0) xn = n=0 0 · xn = 0 converges on R, but it coincides with
f (x) only at the origin.
Remark. The same counterexample works also for the following false state-
ment: “if a function f (x) is infinitely differentiable on R, then there exists at
∑+∞ (n)
least one interval where n=0 f n !(0) xn = f (x) ”.
∑+∞ n
∑+∞ n
Example 5. “If two power series n=0 an x and n=0 bn x take the
same values at infinitely many points, then an = bn , ∀n.”
Solution. ∑+∞ ∑+∞
n x2n+1 n (2x)2n+1
The series sin x = n=0 (−1) (2n+1) ! and sin 2x = n=0 (−1) (2n+1) !
converge on R and take the same zero values at all points xk = kπ, ∀k ∈ Z.
However, their coefficients are different and the series do not coincide.
∑+∞Remark. n
The
∑+∞correct statement is the following: if two power series
n
n=0 a n x and n=0 bn x are convergent on the same interval (−R, R), and
the set of the points of this interval, on which they take the same values, has
a limit point in (−R, R), then an = bn , ∀n.

Example 6. “If two power series, convergent on the same interval, have
different coefficients, then they assume different values at least at one point.”
Solution. ∑+∞ xn ∑+∞ (x−1)n
Two power series n=0 n ! and n=0 e n ! converge on R and have
different coefficients, but they take the same values at every x ∈ R (they are
the Taylor series expansions for ex centered at 0 and 1, respectively).
Remark. The correct statement asserts that a function can have only one
representation as a power series centered at the same point.
∑+∞ ∑+∞
Example 7. “If two power series n=0 an xn and n=0 bn xn have the
∑+∞
same radius of convergence, then the series n=0 (an + bn ) xn also has this
radius of convergence.”
Solution. ∑ ∑+∞
+∞ n n+1 n
The series n=0 (−1) xn and n=0 (−1) x have the radius
of conver-
an bn
gence equal to 1: Ra = lim an+1 = 1 and Rb = lim bn+1 = 1. However,
∑+∞ n→∞
∑+∞ n→∞
the series n=0 (an + bn ) xn = n=0 0·xn is convergent on R, that is Rc = ∞.
Remark 1. Of course, the resulting series can ∑+∞ be different (from zero.
) nFor
n+1
example, choosing the second series in the form n=0 (−1) 1 − 3
1
n x we
bn 1−1/3n
keep the same radius of convergence Rb = lim bn+1 = lim 1−1 3n+1 = 1,
n→∞ n→∞ /
∑+∞ ∑+∞ (−1)n n
n + bn ) x = n=0 3n ·x has a different radius
n
but the third series n=0 (a
cn
of convergence Rc = lim cn+1 = 3.
n→∞
Remark
∑+∞ 2. The corresponding
∑+∞ correct result is as follows: if power se-
n n
ries a x and b x have the radii of convergence Ra and Rb ,
n=0 n n=0 n ∑+∞
respectively, then the radius of convergence of the series n=0 (an + bn ) xn
satisfies the inequality Rc ≥ min {Ra , Rb }. If additionally Ra ̸= Rb , then
Rc = min {Ra , Rb }.
Exercises
1. Show that the following condition: “there exists ε > 0 such that ∀n ∈ N
it holds |an − A| < ε” does not define A as a limit of an . What property of
an is defined by this condition?
2. Provide a counterexample to the statement: “any subsequence of an
unbounded sequence is also unbounded”. (Recall that any subsequence of a
bounded sequence is bounded.)
3. Give an example of a divergent sequence an such that |an | and a2n
converges. Formulate as a counterexample.
4. Show that lim an = +∞ does not imply lim (an+1 −an ) = +∞ neiher
n→∞ n→∞
lim an+1 = +∞. State in the form of counterexamples.
n→∞ an )
5. Give a counterexample to the statement “if both sequences an and bn
are divergent, then the sequence abnn is also divergent”. Compare with Example
3 in section 6.2.
6. Give a counterexample to the statement “if sequences an is convergent
and bn is divergent, then the sequence abnn is divergent”.
7. Illustrate the indeterminate form ∞ − ∞ with different examples of
sequences, that is, show that when lim an = +∞ and lim bn = +∞, then
n→∞ n→∞
any statement about existence/non-existence or (in the case of the existence)
about a specific value of lim (an − bn ) is not true.
n→∞
8. Consider the sequence of arithmetic means yn = x1 +...+x n
n
of the first
n terms of a sequence xn . Show that the statement “if yn converges then xn
converges also” is false. (Notice, that the converse is true: if xn converges then
yn converges to the same limit.)
√
9. Consider the sequence of geometric means zn = n x1 · . . . · xn of the first
n terms of a positive sequence xn . Provide a counterexample to the statement
“if zn converges then xn converges also”. (Notice, that the converse is true: if
xn converges then zn converges to the same limit.)
10. Let xn be a positive sequence. What can be said about the relation
between
{ convergence of arithmetic and geometric means? (Hint: consider xn =
(n+1)2 , n − impar .)
1
n2 , n − par
What about simultaneous convergence of arithmetic and geometric means?
Does it guarantee the convergence of the original sequence? (Hint: consider
xn = 2 + (−1)n .) ∑
11. Show that the∑ following statement is false: “if an converges and
bn ≤ an , ∀n, then bn also converges”. Does the statement become true
if the condition bn ≤ an is substituted by |bn | ≤ |an |, ∀n? Compare with
Examples 1 and 2 in section 6.4.
∑ ∑√
12. Give an example of a divergent series an such that an an+1 is
convergent. ∑ ∑
∑√ 13. Give an example of divergent positive series an and bn such that
a{n bn is convergent. Compare{with the previous exercise. (Hint: consider
n , n − impar and b = n , n − par
1 1
an = .)
n3 , n − par n3 , n − impar
1 n 1
(
∑ 1 )
(−1)n
14. Show that the series √ − is another counterexample to
n n
15. Provide∑ a counterexample
∑ to the following statement: ∑ “if both posi-
tive series an and bn are divergent, then the series min(an , bn ) also
diverges”.
∑+∞ (2n)!!
16. Show that the divergence of the series n=1 (2n+1)!! cannot be estab-
lished by the root, ratio and Integral tests, but it can be proved by the Raabe
test. Compare this result with Examples 10 and∑ 11 in section 6.4.
+∞ 1
17. Show that the convergence of the series n=2 (n−√n) log2 n
cannot be
established by the root, ratio and Raabe tests, but it can be proved by the
Integral or Bertrand tests. Compare this result with Examples 10 and 11 in
section 6.4.
∑+∞ 18. Show that the root, ratio, Raabe and Bertrand tests fail for the series
(−1)n −n
n=1 2 . However, the simple application of the Comparison test gives
the result. The last three exercises illustrates the conclusion in Remark to
Example 11,∑section 6.4.
19. Let an be an arbitrary series with partial sums Sn . Consider the
sequence of arithmetic means of the∑partial sums Cn = S1 +...+S n
n
. If the last
sequence converges then the series an is called convergent by Cesàro. It is
well-known that the convergence of the original series implies the convergence
by Cesàro. Show that the converse is not true. ∑
20. Give a counterexample to the statement “if an is positive and
the
{ 1 sequence na n is unbounded, then the series diverges”. (Hint: try an =
√ , n = k4
n , k ∈ N ) (Notice, that the following statement is true: if
2 , n ̸= k
1 4
∑ n
an is positive and lim nan = A , 0 < A ≤ +∞, then the series diverges.)
n→∞ ∑ √
21. Show that the statement “if an is positive and lim nan = 0 , then
n→∞
the series is convergent” is false. ∑
22. Show that the statement “if an is positive and lim n log nan = 0 ,
n→∞
1
then the series is convergent” is false. (Hint: consider an = n log n log log n .)
23. Provide counterexamples mentioned in Remark 3 to Example 12, sec-
tion 6.4.
24. Provide a counterexample mentioned in Remark 2 to Example 13,
section 6.4. ∑+∞
25. Suppose f (x) can be developed in the series n=0 an xn on (−1, 1).
Show by example that the following statement is false: “if lim f (x) = A, then
x→1−
∑+∞ 1
the numerical series n=0 an converges to A”. (Hint: consider f (x) = 1+x .)
Part II
Functions of two real

variables
205
Chapter 7
Limits and continuity
General Remark. Part II (chapters 7-9) contains the topics of the func-
tions of two variables, which are quite representative for the most topics of
the multivariable functions. The counterexamples presented in this part of
the work are separated into two groups. The first one includes those examples
that have intimate connections with counterexamples of one-variable functions
considered in Part I. They show how the ideas applied in the one-dimensional
case can be generalized/extended to many variables. Accordingly, the exam-
ples of this group are placed in the sections titled “one-dimensional links” in
each of chapters 7-9. The examples of the second group have a weak or no
connection with one-dimensional case, highlighting a specificity of concepts
and results for multivariable functions. Some of them illustrate the situations
that are feasible for two-variable functions but cannot happen in the case of
one-variable functions. In each chapter, all the examples of the second group
are collected in sections titled “multidimensional essentials”.
Limits. Concepts
Limit (general limit). Let f (x, y) be defined on X and (a, b) be a limit

point of X. We say that the limit of f (x, y), as (x, y) approaches (a, b), exists
and equals A if√for every ε > 0 there exists δ > 0 such that for all (x, y) ∈ X
2 2
such that 0 < (x − a) + (y − b) < δ it follows that |f (x, y) − A| < ε. The
usual notations are lim f (x, y) = A and f (x, y) → A.
(x,y)→(a,b) (x,y)→(a,b)
Remark. In calculus, a non-essential simplification that f (x, y) is defined
in some deleted neighborhood of (a, b) is frequently used.
207
Partial limit. Let f (x, y) be defined on X, S be a subset of X and (a, b)

be a limit point of S. The partial limit of f (x, y) on X at (a, b) is the limit of
f (x, y), as (x, y) approaches (a, b), on a subset S.
Infinite limit. Let f (x, y) be defined on X and (a, b) be a limit point of

X. We say that the limit of f (x, y), as (x, y) approaches (a, b), is +∞ (−∞)
every E > 0 there exists δ > 0 such that for all (x, y) ∈ X such that
if for√
2 2
0 < (x − a) + (y − b) < δ it follows that f (x, y) > E (f (x, y) < −E).
The usual notations are lim f (x, y) = +∞ ( lim f (x, y) = −∞)
(x,y)→(a,b) (x,y)→(a,b)
and f (x, y) → +∞ (f (x, y) → −∞).
(x,y)→(a,b) (x,y)→(a,b)
Sometimes it is also considered a “general” infinite limit in the following
sense:
The limit of f (x, y), as (x, y) approaches (a, b), is ∞ if for every
E
√ > 0 there exists δ > 0 such that for all (x, y) ∈ X such that 0 <
2 2
(x − a) + (y − b) < δ it follows that |f (x, y)| > E. The standard no-
tations are lim f (x, y) = ∞ and f (x, y) → ∞.
(x,y)→(a,b) (x,y)→(a,b)
If any of the above three infinite limits is admitted at the same time the
notation used is lim f (x, y) = (±) ∞.
(x,y)→(a,b)
Limit at infinity. Let f (x, y) be defined on an unbounded set X. We

say that the limit off (x, y), as (x, y) approaches ∞, exists and equals A , if
√ every ε > 0 there exists D > 0 such that for all (x, y) ∈ X such that
for
x2 + y 2 > D it follows that |f (x) − A| < ε. The standard notations are
lim f (x, y) = A and f (x, y) → A.
(x,y)→∞ (x,y)→∞
Iterated limits. Let f (x, y) be defined on X, and (a, b) be a limit point

of X. Let f (x, y) possesses the limits lim f (x, y) = g (y) for any fixed y close
x→a
to b, y ̸= b, and such that (x, y) ∈ X. In this way, a new function g (y) is
determined. If the following limit exists lim g (y) = A , then it is called the
y→b
iterated limit of the original function f (x, y) and denoted lim lim f (x, y) = A.
y→b x→a
Similarly can be defined the second iterated limit lim lim f (x, y) = B.
x→a y→b
Path limits. The important case of partial limits for the functions of
two variables is the path limits, when the partial limit is considered along a
chosen curve, which allows a reduction of the analysis of multivariate limits
to the investigation of one-dimensional limits. Let (a, b) be a limit point of
the domain X of the function f (x, y), and (x, y) = (φ (t) , ψ (t)), t ∈ [α, β]
be a parametric definition of the curve that belongs to X and passes through
(a, b), that is, (φ (t0 ) , ψ (t0 )) = (a, b) for some t0 in [α, β]. Then the limit
lim g (t) ≡ lim f (φ (t) , ψ (t)) is called the path limit as (x, y) approaches
t→t0 t→t0
Limits and continuity 209
(a, b) along the chosen curve. Usually, it is sufficient to consider continuous

one-to-one curves passing through (a, b), and we will consider path limits
under this restriction.
Limits. Elementary properties
Unicity. If the limit exists, then it is unique.

Remark 1. In the properties below it is supposed that f (x, y) and g (x, y)
are defined on the same domain.
Remark 2. The point (a, b) can be finite or infinite.
If lim f (x, y) = A, lim g (x, y) = B then
(x,y)→(a,b) (x,y)→(a,b)
1) if f (x, y) ≤ g (x, y) for all (x, y) ∈ X in a deleted neighborhood of (a, b),
then A ≤ B.
2) if A < B, then f (x, y) < g (x, y) for all (x, y) ∈ X in a deleted neighborhood
of (a, b).

If lim f (x, y) = A, lim g (x, y) = B, then
(x,y)→(a,b) (x,y)→(a,b)
1) lim (f (x, y) + g (x, y)) = A + B
(x,y)→(a,b)
2) lim (f (x, y) − g (x, y)) = A − B
(x,y)→(a,b)
3) lim (f (x, y) · g (x, y)) = A · B
(x,y)→(a,b)
f (x,y)
4) lim = A
(under the additional condition B ̸= 0)
(x,y)→(a,b) g(x,y) B
If lim f (x, y) = A, then

(x,y)→(a,b)
5) lim |f (x, y)| = |A|
(x,y)→(a,b)
α
6) lim (f (x, y)) = Aα (under the condition α ∈ N; or under the con-
(x,y)→(a,b)
dition α ∈ Z and A ̸= 0; or under the condition A > 0)
7) lim αf (x,y) = αA (under the assumption α > 0)
(x,y)→(a,b)
The squeeze theorem. If lim f (x, y) = lim g (x, y) = A

(x,y)→(a,b) (x,y)→(a,b)
and the inequality f (x, y) ≤ h (x, y) ≤ g (x, y) holds for all x ∈ X in a deleted
neighborhood of (a, b), then lim h (x, y) = A.
(x,y)→(a,b)
Continuity. Concepts
Continuity at a point. A function f (x, y) defined on X is continuous

∈ X if for every ε > 0 there exists δ > 0 such that whenever
at a point (a, b)√
2 2
(x, y) ∈ X and (x − a) + (y − b) < δ it follows that |f (x, y) − f (a, b)| <
ε.
Remark 1. Frequently it is helpful to consider the definition of
the continuity written in the limit form lim f (x, y) = f (a, b) or
(x,y)→(a,b)
f (x, y) → f (a, b). Actually the original definition of continuity and
(x,y)→(a,b)
the definition by limit differ only in one “pathological” situation of an iso-
lated point (a, b) of X, because the limit definition requires (a, b) to be a limit
point of X, while the general definition of continuity does not contain such a
requirement. It means that at any isolated point (a, b) of X, a function f (x, y)
is continuous according to the general definition, but not continuous according
to the definition by limit. Since the behavior of a function at an isolated point
is hardly of any interest, we will usually consider the definition of continuity
by limit as a complete definition.
Remark 2. In calculus, a simplification that f (x, y) is defined in some
neighborhood of (a, b) is frequently used. Evidently, in this case the general
and limit definitions of continuity coincide.
Continuity with respect to a separate variable. A function f (x, y)

defined on X is continuous in x at a point (a, b) ∈ X if the function of one
variable f (x, b) is continuous at a. Continuity in y is defined similarly.
Continuity on a set. A function f (x, y) defined on X is continuous on

a set S ⊂ X, if f (x, y) is continuous at every point of S.
Remark. It is worth to notice that in the case when a boundary point (a, b)
belongs to set S, continuity on S implies continuity at (a, b) for the original
function considered only on S.
Discontinuity point. Let (a, b) be a limit point of the domain of f (x, y).
If f (x, y) is not continuous at (a, b), then (a, b) is a point of discontinuity of
f (x, y) (or equivalently, f (x, y) has a discontinuity at (a, b)).
Remark 1. Sometimes it is required that (a, b) should be a point of the
domain of f (x, y). We will not impose this restriction.
Remark 2. Classification of discontinuities is usually not considered for
functions of several variables.
Continuity. Local properties
If f (x, y) and g (x, y) are continuous at (a, b) then
1) if f (x) ≤ g (x) for all (x, y) ∈ X in a deleted neighborhood of (a, b), then
f (a, b) ≤ g (a, b).
2) if f (a, b) < g (a, b), then f (x, y) < g (x, y) for all (x, y) ∈ X in a neighbor-
hood of (a, b).

If f (x, y) and g (x, y) are continuous at (a, b), then
1) f (x, y) + g (x, y) is continuous at (a, b)
2) f (x, y) − g (x, y) is continuous at (a, b)
3) f (x, y) · g (x, y) is continuous at (a, b)
4) fg(x,y)
(x,y)
is continuous at (a, b) (under the additional condition g (a, b) ̸= 0)
If f (x, y) is continuous at (a, b), and α ∈ R, then
5) |f (x, y)| is continuous at (a, b)
α
6) (f (x, y)) is continuous at (a, b) (under the condition α ∈ N; or under the
condition f (a, b) > 0)
7) αf (x,y) is continuous at (a, b) (under the assumption α > 0)
Composite function theorem. If f (x, y) and g (x, y) are continuous

at (a, b), h (p, q) is continuous at (c, d) = (f (a, b) , g (a, b)), and the compos-
ite function h (f (x, y) , g (x, y)) is defined in a neighborhood of (a, b), then
h (f (x, y) , g (x, y)) is continuous at (a, b).
Continuity. Global properties
Open Set Characterization. f (x, y) is continuous on a domain X if,

and only if, the inverse image f −1 (S) of any open set S ⊂ f (X) is open.
Closed Set Characterization. f (x, y) is continuous on a domain X if,
and only if, the inverse image f −1 (S) of any closed set S ⊂ f (X) is closed.
The Compact Set Theorem. If f (x, y) is continuous on a compact set
S, then its image f (S) is also a compact set.
The First Weierstrass Theorem. If f (x, y) is continuous on a compact

set, then f (x, y) is bounded on this set.
The Second Weierstrass Theorem. If f (x, y) is continuous on a com-
pact set, then f (x, y) attains its global maximum and minimum values on
this set.
Intermediate Value property. A function f (x, y) satisfies the inter-

mediate value property on a set S ⊂ X, if f (x, y) takes on every value in
between f (x1 , y1 ) and f (x2 , y2 ) for any pair (x1 , y1 ) , (x2 , y2 ) ∈ S (that is,
if f (x, y) takes on two values somewhere in S, it also takes on every value in
between).
The Intermediate Value Theorem. If f (x, y) is continuous on a con-
nected set S, then f (x, y) satisfies the intermediate value property on S. In
particular, if f (x, y) is continuous on a connected set S, then its image f (S)
is also a connected set.
Uniform Continuity. f (x, y) is uniformly continuous on a set S if for

every
√ ε > 0 there exists δ > 0 such that whenever (x1 , y1 ) , (x2 , y2 ) ∈ S and
2 2
(x1 − x2 ) + (y1 − y2 ) < δ it follows that |f (x1 , y1 ) − f (x2 , y2 )| < ε.
Nonuniform Continuity. f (x, y) fails to be uniformly continuous on
a set S if there exists ε0 > 0 such that for
√ any δ > 0 one can found two
2 2
points (xδ , yδ ) and (x̃δ , ỹδ ) in S such that (xδ − x̃δ ) + (yδ − ỹδ ) < δ, but
|f (xδ , yδ ) − f (x̃δ , ỹδ )| ≥ ε0 .
Criterion for Nonuniform Continuity. f (x, y) fails to be uni-
formly continuous on a set S if there √ exists ε0 > 0 and two sequences
2 2
(xn , yn ) and (x̃n , ỹn ) in S such that (xn − x̃n ) + (yn − ỹn ) → 0, but
n→∞
|f (xn , yn ) − f (x̃n , ỹn )| ≥ ε0 .
The Cantor Theorem. If f (x, y) is continuous on a compact set S,
then f (x, y) is uniformly continuous on S.
7.2 One-dimensional links
General Remark. In this section, we gather the examples that have di-
rect connections with those considered in chapters 2 and 3 for functions of
one variable. It means that both the formulation of examples and their solu-
tions can be found as a straightforward generalization of the corresponding
one-dimensional cases. The aim is to show interconnection between the func-
tions of one and two variables, and illustrate different ways of how the one-
dimensional constructions and results can be extended to multidimensional
case. Each example in this section has the references to the corresponding
one-dimensional counterpart.
7.2.1 Concepts and local properties
Example 1. “If lim f (x, y) does not exist, then lim |f (x, y)|
(x,y)→(a,b) (x,y)→(a,b)
does not exist as well.”
(compare with Example 2, section 2.2)
Solution.
Consider the √ two-dimensional modified Dirichlet’s function f (x, y) =
{
1, √ x2 + y 2 ∈ Q
D̃ (x, y) = . The restriction of this function to x-axis
−1 , x2 + y 2 ∈ I
gives the one-dimensional modified Dirichlet’s function, which does not have a
limit at any point in R (see Example 2 in section 2.2). By the same arguments
as in one-dimensional case, the function
D̃
(x, y) does not have a limit at any

point in R . On the other hand, D̃ (x, y) = 1 and this function has the limit
2
1 at any point in R2 .
Remark 1. The Remarks to Example 2 in section 2.2 adjusted to the case
of R2 are applicable here also.
Remark 2. There are many other two-dimensional modifications of Dirich-
let’s function
{ suitable for this counterexample.
{ For instance, the functions
1, x ∈ Q 1 , |x| + |y| ∈ Q
D̄ (x, y) = and D̂ (x, y) = can be used.
−1 , x ∈ I −1 , |x| + |y| ∈ I

Example 2. “If there exists a sequence (xn , yn ) such that

lim f (xn , yn ) = A, then lim f (x, y) = A.”
(xn ,yn )→(a,b) (x,y)→(a,b)
(compare with the second counterexample in Example 4, section 2.2)
Solution.
1
Let us consider the function f (x, y) = sin |x|+|y| defined on R2 \ {(0, 0)}.
1
On the one hand, using the points with the coordinates xk = yk = 2kπ , k∈N
one has the sequence (xk , yk ) → (0, 0) such that lim f (xk , yk ) =
k→+∞ (xk ,yk )→(0,0)
lim sin (kπ) = 0. On the other hand, the points with the coordinates xn =
k→+∞
1
yn = π+4nπ , n ∈ N form the sequence approaching the origin such that
( )
lim f (xn , yn ) = lim sin π2 + 2nπ = 1. Therefore, a general limit
(xn ,yn )→(0,0) n→+∞
does not exist.
Remark. All Remarks to Example 4 of section 2.2 are applicable here with
the corresponding adjustments to the case of R2 .
Example 3. “If both limits lim f (x, y) and lim g (x, y) exist
(x,y)→(a,b) (x,y)→(a,b)
f (x,y)
and the former equals zero, then lim = 0.”
(x,y)→(a,b) g(x,y)
Solution.
√
Let f (x) = g (x) = x2 + y 2 and (a, b) = (0, 0). Then lim f (x, y) =
(x,y)→(0,0)
f (x,y)
lim g (x, y) = 0, but lim = 1.
(x,y)→(0,0) (x,y)→(0,0) g(x,y)
corresponding adjustments to the case of R2 . In particular, for an indetermi-
nate form 00 , one can construct specific functions f (x, y) and g (x, y) such that
the limit of their ratio will be an arbitrary chosen a priory constant (including
zero) or infinity, or this limit will not exist.
Example 4. “If lim f (x, y) = A, lim g (x, y) = B and there

(x,y)→(a,b) (x,y)→(a,b)
is a deleted neighborhood of (a, b) where f (x, y) < g (x, y), then A < B.”
Solution.
Let f (x, y) = −x2 − y 2 , g (x, y) = x2 + y 2 and (a, b) = (0, 0). Then both
limits exist and in any deleted neighborhood of the origin f (x, y) < g (x, y),
but lim f (x, y) = lim g (x, y) = 0.
(x,y)→(0,0) (x,y)→(0,0)
corresponding adjustments to the case of R2 .
Example 5. “If f (x, y) is unbounded in any neighborhood of a point

(a, b), then at least one of the path limits of |f (x, y)|, as (x, y) approaches
(a, b), is infinite.”
Solution.
Let us consider the function f (x) = √ 21 2 cos √ 21 2 defined in
x +y x +y
R2 \ {(0, 0)}. First, notice that the restriction of this function to the x-axis
coincides (except for the absolute value) with the function in Example 9, sec-
tion 2.3. Therefore, choosing for this restriction the same points as in Example
9, one can show that( the function
) is unbounded in any neighborhood of the
1
origin: for (xk , 0) = 2kπ , 0 , k ∈ Z\ {0} one gets lim |f (xk , yk )| =
(xk ,yk )→(0,0)
lim |2kπ| cos |2kπ| = +∞. Next, let us show that no path limit of |f (x, y)|, as
k→∞
(x, y) approaches the origin, is infinite. For an arbitrary (continuous, one-to-
one) curve (x, y) = (φ (t) , ψ (t)), t ∈ [α, β] such that (φ (t0 ) , ψ (t0 )) = (0, 0),
t0 ∈ (α, β),
√ it is possible to find a set of points (xn , yn ) whose distance to the
1
origin is x2n + yn2 = π/2+nπ , n ∈ N, n > N1 . Indeed, the distance function
√
2 2
d (t) = φ (t) + ψ (t) is continuous on [α, β], that implies that d (t) assumes
all the values between d (α) > 0 and d (t0 ) = 0. In particular, choosing N1
1
large enough for the inequality π/2+N 1π
< d (α) to be satisfied, one ensures
that for any n ∈ N, n > N1 there are points (x√ n , yn ) = (φ (tn ) , ψ (tn )) on the
1
chosen curve (φ (t) , ψ (t)) such that d (tn ) = x2n + yn2 = π/2+nπ < d (α).
The use of these points along the chosen path, leads to the following result:
(π ) (π )

lim |f (xn , yn )| = lim + nπ cos + nπ = lim 0 = 0,
(xn ,yn )→(0,0) n→+∞ 2 2 n→+∞
which contradicts to the statement.

Remark. This is a refined version of Example 9 in section 2.3, because
under the same conditions we have shown here that none of the path limits
can be infinite (while in Example 9 it was shown only for the two types of
path limits allowable on R - one-sided limits along x-axis).
Example 6. “If f (x, y)√is continuous at the origin and f (0, 0) = 0, then
the inequality |f (x, y)| ≤ x2 + y 2 holds at least in some neighborhood of
the origin.”
Solution. √
If f (x, y) = 3 x2 + y 2 then √ f (x, y) is continuous at the origin
√ (since
lim f (x, y) = lim 3
x + y = 0 = f (0, 0)), however x + y 2 >
2 2 3 2
√
(x,y)→(0,0) (x,y)→(0,0)
x2 + y 2 for any (x, y) in any neighborhood of the origin of the radius less
than 1.
Example 7. “If f (x, y) is defined in a neighborhood of a point (a, b) and is

decreasing with respect to the distance to (a, b), then (a, b) is a local maximum
point of f (x, y).”
Solution. {
x2 +y 2 , x + y ̸= 0
1 2 2
The function f (x, y) = 2 2 satisfies the conditions of
0, x + y = 0
1
the statement in any neighborhood of the origin, but f (0, 0) = 0 < x2 +y 2 =
f (x, y) for ∀ (x, y) ̸= (0, 0).

Remark. This statement will be true if the condition of continuity at the
point (a, b) would be added.
7.2.2 Global properties
Example 8. “If f (x, y) is continuous on a connected bounded set S and is

not anywhere constant, then it can not take its global maximum and minimum
values infinitely many times.”
Solution. { ( )
x cos x+y1
− 1 − y , (x, y) ̸= (0, 0)
Let us consider f (x, y) = on the
0 , (x, y) = (0, 0)
unit square S = [0, 1] × [0, 1]. Due to arithmetic and composition rules this
function is continuous on S\ {(0, 0)} . It is also continuous at (0, 0), be-
cause lim f (x, y) = 0 = f (0, 0) according to the squeeze theorem:
(x,y)→(0,0)
|f (x, y)| ≤ 2 |x| + |y| → 0. Therefore, f (x, y) is continuous on S.
(x,y)→(0,0)
Also there is no connected open subset of S where f (x, y) is constant. How-
ever, this function
( takes )its global maximum infinitely many times. Indeed,
f (x, y) = x cos x+y1
− 1 − y ≤ 0 for ∀ (x, y) ∈ S and f (0, 0) = 0, so zero
is the global maximum value of f (x, y). Furthermore, the function takes this
value at all the points in S with xn = 2nπ1
and yn = 0, n ∈ N.
Remark. A similar example can be constructed for a global minimum or
for both minimum and maximum.
Example 9. “If a function is defined on R2 , it cannot be continuous at

only one point.”
Solution. {
|x| + |y| , |x| + |y| ∈ Q
The function f (x, y) = is discontinuous at
− |x| − |y| , |x| + |y| ∈ I
any point, except for the origin. In fact, in any neighborhood of every point
(x0 , y0 ) there are points (x, y) satisfying both the condition |x| + |y| ∈ Q and
|x| + |y| ∈ I. Hence, using the former points, one gets the partial limit
lim f (x, y) = lim |x| + |y| = |x0 | + |y0 | ,
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
while using the latter one arrives to

lim f (x, y) = lim − |x| − |y| = − |x0 | − |y0 | .
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
The two results are different for any (x0 , y0 ) ̸= (0, 0), that means that the
general limit does not exist (and the function is not continuous) at any point
different from the origin. If (x0 , y0 ) = (0, 0), then any partial limit gives the
same value zero, which coincide with f (0, 0). This means that the function is
continuous at the origin.
Remark 1. The function d˜(x, y) = |x| + |y| represents another kind of a
distance from the origin (norm) in R2 . For this distance, the points equidistant
from the origin are located on the sides of a square centered at the origin and
parallel to the bisectrices of the coordinate quadrants. In these terms, the
function f (x, y) equals to the distance from the origin, when this distance is
a rational number, and to the minus distance, otherwise.
Remark 2. The statement is also false for an arbitrary domain.
Example 10. “If f (x, y) is continuous on R2 and lim f (n, n) = A,

n→+∞
then lim f (x, y) = A.”
(x,y)→∞

Solution.
We can use almost the same function as in Example 5, section 3.3.
The function f (x, y) = sin πx is continuous on R2 and lim f (n, n) =
n→+∞
lim sin πn = 0, but this is only one of the partial limits when x approaches
n→+∞ ( ) ( )
+∞. Another partial limit lim f 2n + 12 , n = lim sin 2πn + π2 = 1
n→+∞ n→+∞
gives a different result, so lim f (x, y) does not exist.
(x,y)→∞
Remark. The converse is true: if lim f (x, y) = A and f (x, y) is defined
(x,y)→∞
on N × N, then lim f (n, n) = A.
n→+∞
General Remark to the next examples on global properties of continuous

functions. Each characterization of a set (such as connectedness, closeness,
etc.) is related to the space where this set is considered, that is, any charac-
terization of a domain (or its part) of a function is considered in R2 , while a
property of its image is considered in R.
Example 11. “If f (x, y) is continuous on connected sets S and P , then

it is also continuous on S ∪ P .”
Solution.
We can use almost the same { function as in Example 6, section 3.3. Let us
x , (x, y) ∈ B̄1 (0, 0)
consider the function f (x, y) = , where B̄1 (0, 0)
x + 1 , (x, y) ∈ B1 (2, 0)
and B1 (2, 0) are the closed and open unit balls, centered at (0, 0) and (2, 0),
respectively. This function is continuous on S = B̄1 (0, 0) and P = B1 (2, 0),
but it is not continuous on S ∪ P , since at the point (1, 0) there are two partial
limits, approaching (1, 0) along x-axis at the left and at the right, which give
two different values 1 and 2, respectively.
Remark. The statement will be true if S ∩ P ̸= ∅.
Example 12. “If f (x) is continuous on an open set S, then its image is
also an open set.”
Solution.
The application of the counterexample in Example 1, Section 3.4 is
straightforward: the function f (x, y) = x2 + y 2 is continuous on (−1, 1) ×
(−1, 1), but its image is the half-open interval [0, 2).
Another simple counterexample is the function f (x, y) = sin (x + y): it is
continuous on the open set (0, 2π) × (0, 2π), but its image on this set is the
closed interval [−1, 1].
Remark. The Remarks to Example 1 of section 3.4 are applicable here
with corresponding adjustments to the case of R2 .
Example 13. “If f (x, y) attains its global minimum and maximum on a
compact S, then it is continuous on S.”

Solution.
The extension of one-dimensional counterexamples is straightforward.
For instance,
√ the two-dimensional version of Dirichlet’s function D (x, y) =
{
1 , √x2 + y 2 ∈ Q
attains its global minimum 0 and maximum 1 on any
0 , x2 + y 2 ∈ I
compact set, containing interior points, but it does not have a limit at any
point.
Another simple counterexample is f (x, y) = sgn (x + y) considered on any
compact set containing the origin as the interior point (for instance, S =
[−1, 1] × [−1, 1] ).
Remark. The converse statement is true and represents the second Weier-
strass theorem.
Example 14. “If f (x, y) is continuous on S, then f (x, y) assumes any

value between f (x1 , y1 ) and f (x2 , y2 ), (x1 , y1 ), (x2 , y2 ) ∈ S on the points of
S.”
Solution.
The function from Example 12, Section 3.4 can be easily adapted to
the two-dimensional case. Let us consider f (x, y) = sgn x on the set S =
B1 (−1, 0) ∪ B1 (1, 0), where B1 (−1, 0) and B1 (1, 0) are the open unit balls,
centered at (−1, 0) and (1, 0), respectively. Evidently, this function is contin-
uous on S, but it does not assume any value between f (−1, 0) = −1 and
f (1, 0) = 1.
Remark. The intermediate value property is satisfied for the functions
continuous on a connected sets.
Example 15. “Between two local minima there is a local maximum.”

Solution.
The function in Example 13, section 3.4 is easily adaptable to this case:
f (x, y) = sec2 x has infinitely many local minima on R2 and no local maxi-
mum.
Remark. This statement is true for continuous functions with domain
in R. For multivariate functions the statement is false even when f (x, y) is
continuous (see Example 11 in section 7.3).
Example 16. “If f (x, y) is continuous and bounded on a bounded con-

nected set S, then f (x, y) is uniformly continuous on this set.”
Solution.
The function f (x, y) = sin πx (adapted from Example 1, section 3.5) is
bounded and continuous on S = (0, 1) × (−1, 1), but it does( not) possess
uniform continuity on S. In fact, for the sequences (xn , 0) = n1 , 0 , ∀n ∈
( )
N, n > 1 and (x̃n , 0) = 1+4n2
, 0 , ∀n ∈ N we have the property that the
distance between the corresponding points of the two sets tends to zero
√
1 2
2 2
(xn − x̃n ) + (0 − 0) = |xn − x̃n | = − → 0,
n 1 + 4n n→+∞
but
|f (xn , 0) − f (x̃n , 0)| = |0 − 1| = 1, ∀n ∈ N, n > 1.
Therefore, f (x, y) is not uniformly continuous on S.
Remark. The statement will be correct if S would be compact (according
to the Cantor theorem).
Example 17. “If f (x, y) is uniformly continuous on R2 , then f 2 (x, y) is

also uniformly continuous on R2 .”
Solution.
The function f (x, y) = x (adapted from Example 4, section 3.5) is uni-
) R , but f (x, y) = x is not, since for the two sequences
2 2 2
formly continuous
( on
(xn , 0) = n + n , 0 , ∀n ∈ N and (x̃n , 0) = (n, 0) , ∀n ∈ N the distance be-
1
tween the corresponding points of the two sets gets as close as we wish for
sufficiently large n: |xn − x̃n | = n1 → 0, but
n→+∞
2
f (xn ) − f 2 (x̃n ) = 2 + 1 > 2, ∀n ∈ N.
n2
7.3 Multidimensional essentials
Example 1. “If a function approaches zero, as a point approaches the

origin along any of the coordinate axis, then its limit is zero at the origin.”
Solution. { xy
x2 +y 2 , x + y ̸= 0 has both coordinate lim-
2 2
The function f (x, y) =
0, x2 + y 2 = 0
its equal to zero at the origin: lim f (x, 0) = lim x02 = 0 and lim f (0, y) =
x→0 x→0 y→0
0
lim 2 = 0. However, it does not possess the general limit, since the par-
y→0 y
tial limits along straight lines y = kx, k ∈ R, depend on the chosen line:
kx2 k
lim f (x, kx) = lim x2 +k 2 x2 = 1+k 2 .
x→0 x→0
Remark. For the continuity, this false statement can be reformulated as
follows: “if a function is continuous with respect to each of variables at some
point, then it is continuous at this point”. The same function provides coun-
terexample.
Example 2. “If a function approaches zero, as a point approaches the

FIGURE 7.3.2: Example 1, the function on some paths through the origin
origin along any straight line passing through the origin, then its limit is zero
at the origin.”
Solution. {
x2 y
x4 +y 2 , x + y ̸= 0
2 2
The function f (x, y) = 2 2
satisfies the conditions
0, x + y = 0
of the statement. Indeed, for any straight line y = kx, k ∈ R, we have
kx3 kx
lim f (x, kx) = lim x4 +k 2 x2 = lim x2 +k 2 = 0, and for the straight line x = 0
x→0 x→0 x→0
the limit is also zero: lim f (0, y) = lim y02 = 0. However, along the parabola
y→0 y→0
( ) 4
y = x2 the result is different: lim f x, x2 = lim x4x+x4 = 12 .
x→0 x→0
Remark. Analogous false statement for the continuity is: “if a function is
continuous along any straight line passing through a chosen point, then it is
continuous at this point”. The same function provides counterexample.
Example 3. “If f (x, y) approaches zero, as a point approaches the origin

along any algebraic curve y = cxm/n , where c ∈ R and m, n ∈ N (x ≥ 0 in
case the exponential fraction is in lowest terms and n is even), then the limit
of f (x, y) is zero at the origin.”
Solution. { 2
e−1/x y
−2/x 2 , x ̸= 0
Let us show that the function f (x, y) = e +y 2 satisfies
0, x = 0
FIGURE 7.3.4: Example 2, the function on some paths through the origin
the conditions of the statement. For y = 0 (c = 0) this is evident,

( and)for
an arbitrary algebraic curve y = cxm/n , c ̸= 0, we have lim f x, cxm/n =
x→0
−1/x 2 m/n
lim e 2 cx . By substitution x = 1
the last limit can be calculated as
x→0 e−2/x +c2 x2m/n t
follows:
( ) ce−t (1/t)
2 m/n
ctm/n /et
2
lim f x, cxm/n = lim 2m/n

= lim 2 = 0,
x→0 t→∞ e−2t2 + c2 (1/t) t→∞ t2m/n /e2t + c2
since
tm/n m tm/n−1 m tm/n−2
lim 2 = lim 2 = lim 2 = ... = 0
t→∞ et t→∞ n 2tet t→∞ n 2et
by applying L’Hospital’s rule repeatedly until the power of t in the numerator

becomes negative or zero. Nevertheless, the limit of f (x, y) at the origin does
not exist, because along the curve y = e−1/x the partial limit is different:
2
( ) e−1/x · e−1/x
2 2
1
lim f x, e−1/x = lim −2/x2
2
−2/x 2 = .
x→0 x→0 e +e 2
Remark 1. Of course, the family of paths(curves) that assures the same
value for the partial limits can be even broader(wider) than the set of algebraic
curves, but still the general limit may not exist.
Remark 2. For the continuity, a similar false statement is: “if a function is
continuous along any algebraic curve passing through a chosen point, then it
is continuous at this point”. The same function provides the counterexample.
Example 4. “If f (x, y) = QPm n (x,y)

(x,y) , where Pn (x, y) and Qm (x, y) are
polynomials of degree n and m, respectively, and the latter polynomial is zero
at some point, then f (x, y) does not have the general limit at this point.”
Solution. 2
−y 2
The function f (x, y) = xx+y has the required form and it also has the
x2 −y 2
limit at the origin: lim = lim (x − y) = 0.
(x,y)→(0,0) x+y (x,y)→(0,0)
Remark 1. The statement is also false if the numerator or denominator
sin(x2 +y 2 )
functions are not polynomials. For example, f (x, y) = x2 +y2 has the limit
2 2
x +y
equal to 1 at the origin, as well as f (x, y) = arctan(x 2 +y 2 ) .
Remark 2. The statement will be true if Pn (x, y) is not equal to zero at

the limit point and only the finite limits are considered.
Example 5. “If a function approaches zero along any ray x = t cos α,

y = t sin α, 0 ≤ t < +∞, as t approaches infinity, then it approaches zero at
infinity.”
Solution.
The function f (x, y) = x2 e−(x −y) satisfies the conditions of the state-
2
ment. Indeed, if cos α = 0 (α = ± π2 ), then f (t cos α, t sin α) = 0. For any

other α we have
t2 cos2 α
lim f (t cos α, t sin α) = lim 2 2
t→+∞ t→+∞ et cos α−t sin α
and applying L’Hospital’s rule twice we obtain:

2t cos2 α
lim f (t cos α, t sin α) = lim t2 cos2 α−t sin α · (2t cos2 α − sin α)
t→+∞ t→+∞ e
2 cos2 α
= lim ( ) = 0.
2
t→+∞
et2 cos2 α−t sin α · (2t cos2 α − sin α) + 2 cos2 α
However, approaching infinity along the parabola y = x2 , we obtain
( )
lim f x, x2 = lim x2 e−(x −x ) = lim x2 = +∞, which means that a
2 2
x→+∞ x→+∞ x→+∞

general limit at infinity does not exist.
Example 6. “If iterated limits of f (x, y) exist, then the general (double)
limit also exists.”
Solution.
For the function f (x, y) = x−y
x+y both iterated limits exist at the point
(0, 0):
x−y x x−y −y
lim lim = lim = lim 1 = 1, lim lim = lim = lim (−1) = −1.
x→0 y→0 x + y x→0 x x→0 y→0 x→0 x + y x→0 y x→0
FIGURE 7.3.6: Example 5, the function on some paths approaching infinity
x−y
However, the general limit lim does not exist, because two partial
(x,y)→(0,0) x+y
x−x
limits are different: the partial limit along the straight line y = x is lim =
x→0 x+x
0, while the partial limit along the straight line y = 2x is lim x+2x = − 3 .
x−2x 1
{x→0
x2 −y 2
, x2 +y 2 ̸= 0
Another counterexample of the same kind is f (x, y) = x2 +y22 2 .
0, x +y = 0
At the point (0, 0) both iterated limits exist:
x2 − y 2 x2 x2 − y 2 −y 2
lim lim 2 2
= lim 2 = 1, lim lim 2 2
= lim 2 = −1,
x→0 y→0 x + y x→0 x y→0 x→0 x + y x→0 y
but the general limit does not exist since the partial limit along the straight
line y = x, lim xx2 −x
2 2
x→0 +x2 = 0, is different from the partial limit along the straight
x2 −4x2
line y = 2x, lim 2 2 = − 35 .
x→0 x +4x
Remark. Both functions are also counterexamples to the following false
statement: “if iterated limits exist, then they are equal”.
Example 7. “If both iterated limits exist and are equal, then the general
(double) limit also exists.”
Solution. { xy
x2 +y 2 , x + y ̸= 0
2 2
The function of the Example 1, f (x, y) = 2 2 has
0, x + y = 0
both iterated limits at the point (0, 0) equal to 0:

xy 0 xy 0
lim lim = lim 2 = 0, lim lim 2 = lim 2 = 0.
x→0 y→0 x2 +y 2 x→0 x y→0 x→0 x +y 2 x→0 y
xy
However, the general limit lim 2 2 does not exist since the partial
(x,y)→(0,0) x +y
limits along the straight lines y = kx are different (see Example 1): for ex-
2
ample, along the straight line y = x one has lim x2x+x2 = 12 , while along the
x→0
2x2
straight line y = 2x one gets lim x2 +4x2 = 25 .
x→0
Another counterexample of the same kind is the function
{
x2 y 2 2
2 y 2 +(x−y)2 , x + y
2
̸= 0
f (x, y) = x
2 2
.
0, x + y = 0
In fact, both iterated limits are equal to 0:

x2 y 2 0 x2 y 2 0
lim lim 2 = lim = 0, lim lim = lim 2 = 0,
x→0 y→0 x2 y 2 +(x−y) x→0 x 2 y→0 x→0 x2 y 2 +(x−y)2 x→0 y
x4
but the general limit does not exist, since for y = x one obtains lim 4 = 1,
x→0 x
while for y = 0 the result is different lim 02 = 0.
x→0 x

Example 8. “If the general limit of f (x, y) exists, then iterated limits
also exist.”
Solution. { 1
(x + y) sin xy , xy ̸= 0
The function f (x, y) = has the general limit
0, xy =
0
1
at the origin. In fact, the inequality (x + y) sin xy ≤ |x| + |y| leads to the
evaluation |f (x, y)| ≤ |x| + |y| for an arbitrary point (x, y). Therefore, due
to the squeeze theorem, the fact that lim (|x| + |y|) = 0 implies that
(x,y)→(0,0)
1
lim f (x, y) = 0. At the same time, lim (x + y) sin xy does not exist
(x,y)→(0,0) y→0
for any x ̸= 0, because for any given x0 ̸= 0 there is a sequence of the points
yn = 2nπ1 x0 → 0 such that lim (x0 + yn ) sin x01yn = 0 and another sequence
n→∞ yn →0
yk = 1
→ 0 such that lim (x0 + yk ) sin x01yk = x0 . It means, that
(2kπ+ 2 )x0 k→∞
π
yk →0
the iterated limit lim lim f (x, y) does not exist. Due to the same reasons, the
x→0 y→0
second iterated limit lim lim f (x, y) does not exist either.
y→0 x→0
{ Similar counterexample is provided with the function f (x, y) =

x sin y1 + y sin x1 , xy ̸= 0
. The general limit is equal to 0 due to inequality
0, xy = 0

x sin y1 + y sin x1 ≤ |x| + |y| and application of the squeeze theorem. Nev-
ertheless, the iterated limit lim lim f (x, y) does not exist, because the in-
x→0 y→0
terior(internal) limit lim f (x, y) is not defined for any fixed x ̸= 0: for the
y→0
( )
sequence yn = nπ → 0 one gets lim x sin y1n + yn sin x1 = 0, while for
1
n→∞
( yn →0 )
yk = (4k+1 )π → 0 one has lim x sin y1k + yk sin x1 = x. Consideration of
2
k→∞ yk →0
another iterated limit is analogous.
Example 9. “If the general limit of f (x, y) exists and one of the iterated
limits also exists, then the second iterated limit exists as well.”
Solution. {
y + x sin y1 , y ̸= 0
The function f (x, y) = has the zero limit at the ori-
0, y = 0
gin, because |f (x, y)| ≤ |x| + |y| → 0. Besides, for any fixed y ̸= 0 we
( ) (x,y)→(0,0)
obtain lim y + x sin y1 = y, and consequently lim lim f (x, y) = lim y = 0.

x→0 y→0 x→0 y→0
Hence, the general limit and one of the iterated limits exist and are equal. Nev-
ertheless, the second iterated limit does not exist, since the limit lim f (x, y)
y→0
is not defined for any fixed x ̸= 0: for the sequence yn = 1
→
nπ n→∞ 0 one gets
( )
lim yn + x sin y1n = 0, while for yk = 2
→ 0
(4k+1 )π k→∞ the result is different
yn →0
( )
lim yk + x sin y1k = x.
yk →0
{
x + y sin x1 , x ̸= 0
Remark. The function f˜ (x, y) = provides the coun-
0, x = 0
terexample with the interchanged variables.
Example 10. “If the general limit and one of the iterated limits of f (x, y)
do not exist, then the second iterated limit also does not exist.”
Solution. { xy
x2 +y 2 + x sin y , y ̸= 0 does not have a general
1
0, y = 0
limit, because two
( 2partial limits
) give different results: along the line y = x

one obtains lim 2x2 + x sin x = 2 (since x sin x1 ≤ |x| → 0), while along
x 1 1
x→0 ( ) x→0
2x2 1 2
the line y = 2x one has lim x2 +4x2 + x sin 2x = 5 . Moreover, the limit
x→0
lim lim f (x, y) also does not exist, because lim f (x, y) is not defined for
x→0 y→0
(
y→0
)
any fixed x ̸= 0: if yn = nπ → 0 then lim x2xy
1 n
+y 2 + x sin 1
yn = 0, but
yn →0
n→∞
( ) n
)π → 0 then lim
2 xyk
if yk = (4k+1 x2 +y 2
+ x sin y1k = x. Nevertheless, the
k→∞ yk →0 k
second iterated limit exists:

( )
xy 1
lim lim f (x, y) = lim lim + x sin = lim 0 = 0.
y→0 x→0 y→0 x→0 x2 + y 2 y y→0
{ xy
x2 +y 2 + y sin x , x ̸= 0 provides the
1
Remark. The function f˜ (x, y) =
0, x = 0
counterexample with the interchanged variables.
Example 11. “If a continuous on a connected set S function f (x, y)

possesses two local maxima in S, then there exists a local minimum of f (x, y)
in S.”
Solution.
The function f (x, y) = sin x−y 2 considered on S( = (−2π, ) 2π)×(−1,(1) has
)
two local (and global) maxima at the points P1 = − 3π 2 , 0 and P 2 = π
2 0 .
,
Indeed, the function sin x assumes its maximum values at the points − 3π 2 and
π
2 in the interval (−2π, 2π) and the function −y 2
has the maximum value at
the point 0 of the interval (−1, 1). However, there is no minimum point of
f (x, y), since along the line x = x0 the original function is reduced to one-
variable function f (x0 , y) = sin x0 − y 2 , which does not have a minimum on
all its domain y ∈ R.
Remark. The same statement is true for one-variable functions.

Example 12. “If a function f (x, y) is continuous in a neighborhood of

the point (x0 , y0 ), and both f (x, y0 ) and f (x0 , y) have local maxima at this
point, then f (x, y) possesses a local maximum at (x0 , y0 ).”
Solution.
The function f (x, y) = −x2 + 4xy − y 2 is continuous on R2 , in particular,
in any neighborhood of (0, 0). The functions f (x, 0) = −x2 and f (0, y) = −y 2
have local (and global) maxima at the origin. However, the function f (x, y)
does not have a local maximum at (0, 0), because along the line y = x we get
f (x, x) = 2x2 and for this function the origin is the local minimum.
Exercises
1. Show that if a function approaches zero, as a point approaches the

origin along any straight line and any parabola, it does not mean that this
function has{ a limit at the origin. Formulate this result as a counterexample.
2x3 y
x6 +y 2 , x + y ̸= 0 )
2 2
(f (x, y) = 2 2
0, x + y = 0
2. Show that the function f (x, y) = 3xx4 −5y
4 2
+y 2 has the iterated limits (albeit

different), but does not have the general limit at the origin. Notice that this
is another counterexample to Example { 6, section 7.3.
x3 y
x4 +y 4 , x + y ̸= 0 (see Example
2 2
3. Show that the function f (x, y) =
0, x2 + y 2 = 0
1 section 7.3) has the equal iterated limits, but does not have the general limit.
Notice that this is another counterexample to Example 7, section 7.3.
4. Prove that
x2 + y 2 x2 + y 2
lim lim = lim lim ,
x→∞ y→∞ 1 + (x − y)4 y→∞ x→∞ 1 + (x − y)4
x2 +y 2
but the general limit lim 1+(x−y) 4 does not exist. Formulate as a coun-
(x,y)→∞
terexample. Compare with Example{ 7, section 7.3.
1
(2x2 + y 2 ) cos xy , xy ̸= 0
5. Prove that f (x, y) = is another coun-
0, xy = 0
terexample to Example 8, section { 7.3.
x2 y 2
4 +y 4 + x cos y 2 , y ̸= 0
2 1
6. Prove that f (x, y) = x is another coun-
0, y = 0
terexample to Example 10, section 7.3.
7. Give an example of a function f (x, y) discontinuous at every point of the
square [0, 1] × [0, 1], but continuous with respect to x for any fixed y ∈ [0, 1].
Formulate as a counterexample. (Hint: use f (x, y) = D(y).)
8. Show that the continuity of a function does not imply that
small increments of function values produce small increments of inde-
pendent
{ variables. Formulate as a counterexample. (Hint: try f (x, y) =
0, x2 + y 2 ≤ 1
)
x2 + y 2 − 1, x2 + y 2 > 1
9. Show that the condition that small increments of function val-
ues lead to small increments of independent variables does not guarantee
the continuity. Formulate as a counterexample. (Hint: consider f (x, y) =
{
x2 + y 2 , x ≥ 0
and notice that both small and large increments the
x2 + y 2 − 1, x < 0
independent variables correspond to small increments of the function values.)
10. Differently from the theorem on composition of two continuous func-

tions, the composition of two discontinuous functions can result in a continu-
ous one. Formulate as a counterexample. (Hint: use u = D(x−y), v = D(x+y)
and z = D(u + v), where D(t) is Dirichlet’s function.)
11. “If u = f (x, y) and v = g(x, y) are discontinuous functions and z =
h(u, v) is a continuous function, then z = h(f (x, y), g(x, y)) is a discontinuous
function.” Prove by example that this is a false statement. (Hint: use u =
D(x − y), v = D(x + y) and z = (u − 21 )2 + (v − 12 )2 , where D(t) is Dirichlet’s
function.)
12. Provide a counterexample to the following false statement: “if f (x, y) is
continuous on a compact set S, except for a one point, then f (x, y) is bounded
on S”. What if additionally the function is defined on S?
13. Provide a counterexample to the following false statement: “if f (x, y)
is bounded on a compact set S and is continuous on S, except for a one point,
then f (x, y) attains its global extrema on S”.
14. Prove by example that the following statement is false: “if f (x, y) is
continuous on a closed set S, then its image is also a closed set”.
15. Prove by example that the following statement is false: “if f (x, y) is
continuous on a bounded set S and f (S) is a compact, then S is also a compact
set”.
16. Show by example that the following statement is false: “if f (x, y) sat-
isfies the intermediate value property on a connected set S, then f (x, y) is
continuous on S”.
is uniformly continuous on S and Q, then f (x, y) is uniformly continuous on
S ∪ Q”.
and g(x, y) are uniformly continuous on S, then f (x, y) · g(x, y) is uniformly
continuous on S”. What about two non-uniformly continuous function? Should
their product be non-uniformly continuous also?
is uniformly continuous on any compact set then it is uniformly continuous
on R2 ”.
20. Show by example that the following statement is false “if f (x, y) is uni-
1
formly continuous and positive on S, then f (x,y) is also uniformly continuous
on S”.
Chapter 8
Differentiability
8.1 Elements of Theory
Concepts
Partial derivatives. Suppose function f (x, y) is defined on X ⊂ R2 .

Let us keep one of the independent coordinates, let say y, fixed and consider
the function of one variable g (x) = f (x, y0 ). If this function is differentiable
at x0 , then the derivative g ′ (x0 ) is called the partial derivative of f (x, y)
with respect to x at the point (x0 , y0 ). The following notations are common:
fx (x0 , y0 ) ≡ ∂x f (x0 , y0 ) ≡ ∂1 f (x0 , y0 ) ≡ ∂f
∂x (x0 , y0 ). The specification of the
point is frequently omitted if it is clear from the context. In the same way can
be defined the partial derivative with respect to y.
Geometrically the partial derivative fx (x0 , y0 ) represents the slope of the
tangent line at (x0 , y0 ) to the trace of the function graph on the plane x = x0
in R3 .
Partial derivatives on a set. A function f (x, y) defined on X has
partial derivative on a set S ⊂ X, if fx exists at every point of S.
Higher (order) partial derivatives. If there exists a partial derivative

of the function h (x, y) = fx (x, y) with respect to x at the point (x0 , y0 ),
then this partial derivative is called the second order (or simply the second)
partial derivative of the original function f (x, y) with respect to x, x at the
point (x0 , y0 ). The usual notations are (omitting the point indication): fxx ≡
2
∂xx f ≡ ∂11 f ≡ ∂∂xf2 . In the same way can be defined other three partial
derivative of the second order: fxy , fyx and fyy .
Similarly, if there exists the partial derivative with respect to x of the func-
tion h (x, y) = fxm (x, y) representing the m-th partial derivative of f (x, y),
237
then this partial derivative is called the (m + 1)-th partial derivative of the
original function f (x) with respect to x, ..., x. The standard notations are:
m+1
fxm+1 ≡ fx...x ≡ ∂1...1 f ≡ ∂∂xm+1f . In the same way can be defined all others
2m+1 − 1 partial derivative of the (m + 1)-th order.
Clairaut’s Theorem. If the mixed derivatives fxy and fyx are continuous
at (x0 , y0 ), then they are equal at (x0 , y0 ).
Remark. A similar result holds for the mixed m-th partial derivatives:
if the two mixed corresponding m-th partial derivatives are continuous at a
point (x0 , y0 ), then they are equal at (x0 , y0 ).
Differentiation at a point. A function f (x, y) defined on X is called

differentiable at a point (x0 , y0 ) ∈ X, if for any point (x, y) in a neighbor-
hood of (x0 , y0 ) the following representation holds: f (x, y) = f (x0 , y0 ) + a ·
(x − x0 ) + b · (y − y0 ) + α · (x − x0 ) + β · (y − y0 ), where a, b are constants, and
α, β are the functions of x, y such that α, β → 0.
(x,y)→(x0 ,y0 )
Remark 1. The same definition is frequently written in the terms of the
increments of the arguments ∆x = x−x0 , ∆y = y −y0 and the function ∆f =
f (x, y)−f (x0 , y0 ) as follows: ∆f = a·∆x+b·∆y+α·∆x+β·∆y, where a, b are
constants and α, β are the functions of ∆x, ∆y such that α, β → 0.
(∆x,∆y)→(0,0)
Remark 2. The same definition can be formulated in the following equiv-
alent form: a function f (x, y) is called differentiable at a point (x0 , y0 ), if for
any point in a neighborhood of (x0 , y0 ) the following representation holds:
f (x, y) = f (x0 , y0 ) + a · (x − x0 ) + b · (y − y0 ) + γ (x, y), where a, b are
constants, √
and γ is the function of x, y that approaches zero faster than
2 2 γ(x,y)
ρ (x, y) = (x − x0 ) + (y − y0 ) , that is → 0
ρ(x,y) ρ→0 .
Differential and remainder. The expression a·∆x+b·∆y ≡ df is called

the differential of f (x, y) at a point (x0 , y0 ) and the difference γ = ∆f − df is
called the remainder. Note that the term df is the linear (with respect to ∆x,
∆y) part of ∆f , and this is also the main part of ∆f in a small neighborhood of
(x0 , y0 ) if a2 + b2 ̸= 0. Thus, the differential represents a linear approximation
to the function increment in a neighborhood of (x0 , y0 ).
Theorem. If f (x, y) is differentiable at (x0 , y0 ), then there exist both
first partial derivatives fx , fy at (x0 , y0 ) and a = fx , b = fy .
Gradient. The vector of the first partial derivatives (fx , fy ) is called the
gradient. The common notations are ∇f and gradf .
Remark. According to the last theorem and the gradient definition, the
definition of the differentiation can be rewritten using the dot product in the
form ∆f = ∇f · ∆r + (α, β) · ∆r, where ∆r = (∆x, ∆y) and α, β satisfy the
required for differentiation properties. Sometimes the last formula is used as a
Differentiability 239
definition of the differentiation. Accordingly, the definition of the differential

can be rewritten as df = ∇f · ∆r.
Differentiation on a set. If f (x, y) is differentiable at every point of a

set S, then f (x, y) is differentiable on S.
Continuous differentiability. If the differential g (x, y) ≡ df =
fx (x, y) ∆x + fy (x, y) ∆y is a continuous function at (x0 , y0 ) (keeping the
increments ∆x, ∆y fixed), then the original function f (x, y) is continuously
differentiable at (x0 , y0 ). If the differential df is continuous on a set S, then
f (x, y) is continuously differentiable on S.
Higher (order) differentiation at a point. Let a function f (x, y)

be differentiable in a neighborhood of a point (x0 , y0 ). If the differential
g (x, y) ≡ df = fx (x, y) ∆x+fy (x, y) ∆y is a differentiable function at (x0 , y0 )
(keeping the increments ∆x, ∆y fixed), then the original function f (x, y) is
twice differentiable and its second (order) differential is (by definition) equal
to the differential of g (x, y):
d2 f ≡ dg = gx ∆x + gy ∆y = fxx ∆x∆x + fxy ∆x∆y + fyx ∆x∆y + fyy ∆y∆y.
In the case when the mixed partial derivatives are equal, the second differential
can be simplified to the form
d2 f = fxx ∆x∆x + 2fxy ∆x∆y + fyy ∆y∆y.

2
The following symbolic form is helpful: d2 f = (∆x∂x + ∆y∂y ) f .
Similarly, if a function f (x, y) is m times differentiable in a neighborhood
m
of a point (x0 , y0 ) and its m-th differential h (x) ≡ dm f = (∆x∂x + ∆y∂y ) f
is a differentiable function at (x0 , y0 ) (keeping the increments ∆x, ∆y fixed),
then the original function f (x, y) is m + 1 times differentiable at (x0 , y0 )
and its (m + 1)-th differential can be found by the formula: dm+1 f =
m+1
(∆x∂x + ∆y∂y ) f.
Directional derivative. Let us consider a half-straight line defined by

the initial point (x0 , y0 ) and the direction (non-zero) vector v = (a, b): x =
x0 +at, y = y0 +bt, t ≥ 0 (t is the parameter). Along this ray a function f (x, y)
is the function of the one variable t: h (t) ≡ f (x0 + at, y0 + bt). If the function
h (t) is differentiable at t0 = 0, then its derivative is the directional derivative
Dv f (x0 , y0 ) of the original function f (x, y), that is, Dv f (x0 , y0 ) = h′ (0).
Tangent plane. If f (x, y) is differentiable at a point (x0 , y0 ), then the

following tangent plane to the graph of f (x) at (x0 , y0 ) can be defined: z −
f (x0 , y0 ) = fx (x0 , y0 ) · (x − x0 ) + fy (x0 , y0 ) · (y − y0 ).
It follows immediately from the definition that the tangent plane contains
the tangent lines at (x0 , y0 ) to the two traces of the function graph - on the
plane x = x0 and on y = y0 .
Notice, that the linear approximation of a function through its tangent

plane corresponds to the linear approximation of its increment through differ-
ential: z − f (x0 , y0 ) = df (x0 , y0 ).
Hessian. The Hessian

( matrix)(Hessian) is the matrix of the second partial
fxx fxy
derivatives H(f ) = .
fxy fyy
Basic properties
Theorem. If f (x, y) is differentiable at (x0 , y0 ), then it is continuous at

(x0 , y0 ).
Similarly, if f (x, y) is differentiable on a set, then it is continuous on this
set.
Theorem. f (x, y) is continuously differentiable at (x0 , y0 ) if, and only

if, the first partial derivatives fx , fy are continuous at (x0 , y0 ).
Theorem. f (x, y) is continuously differentiable on S if, and only if, the
partial derivatives fx , fy are continuous on S.

If f (x, y) and g (x, y) are differentiable at a point (x, y), then
1) f (x, y) + g (x, y) is differentiable at (x, y) and d (f + g) = df + dg
2) f (x, y) − g (x, y) is differentiable at (x, y) and d (f − g) = df − dg
3) f (x, y) · g (x, y) is differentiable at (x,
( y)) and d (f · g) = df · g + f · dg
4) g(x,y) is differentiable at (x, y) and d fg = df ·g−f
f (x,y)
g2
·dg
(under the additional
condition g (x, y) ̸= 0)
Chain rule. If p = f (x, y) and q = g (x, y) are differentiable func-

tions at a point (x0 , y0 ), and h (p, q) is a differentiable function at the
point (p0 , q0 ) = (f (x0 , y0 ) , g (x0 , y0 )), then the composite function φ (x, y) ≡
h (f (x, y) , g (x, y)) is differentiable at the point (x0 , y0 ) and its partial deriva-
tives can be calculated as follows: φx = hp px + hq qx , φy = hp py + hq qy .
Theorem. If f (x, y) is differentiable at a point (x0 , y0 ), then f (x, y) has

a directional derivative at this point in the direction of any vector v = (a, b)
f a+f b
and the following formula holds Dv f = ∇f · |v| v
= √xa2 +by2 .
Extremum properties of the gradient:
1) If f (x, y) is differentiable and ∇f ̸= 0, then the maximum and minimum
values of Dv f are |∇f | and − |∇f |, that occur in the directions ∇f and −∇f ,
respectively.
2) If f (x, y) is differentiable and ∇f ̸= 0, then ∇f is the normal vector to
level curves of f (x, y), that is, ∇f is the normal vector to the tangent lines
of the level curves.
Theorem. If f (x, y) is differentiable at a point (x0 , y0 ), then the tangent

plane at (x0 , y0 ) contains all the tangent lines passing through (x0 , y0 ).
Mean Value Theorem. If f (x, y) is continuously differentiable in a

neighborhood of a point (x0 , y0 ), then for any point (x, y) in this neighborhood
there exists a point (x̃, ỹ) = (x0 + θ (x − x0 ) , y0 + θ (y − y0 )), θ ∈ (0, 1) in the
same neighborhood such that ∆f (x0 ) = df (x̃).
Remark. The condition of continuous differentiability can be substituted
by the condition of continuity of the first order partial derivatives.
Taylor’s Theorem. If f (x, y) is n + 1 continuously differentiable in a

neighborhood of a point (x0 , y0 ), then for any point (x, y) in this neighborhood
the following formula holds:
1 1
f (x, y) = f (x0 , y0 ) + df (x0 , y0 ) + d2 f (x0 , y0 ) + . . .
1! 2!
1 n 1
+ d f (x0 , y0 ) + dn+1 f (x̃, ỹ) ,
n! (n + 1) !
where (x̃, ỹ) is the point located on a straight-line segment between
(x0 , y0 ) and (x, y), that is, there exists θ ∈ (0, 1) such that (x̃, ỹ) =
(x0 + θ (x − x0 ) , y0 + θ (y − y0 )).
Remark 1. The continuous differentiability can be substituted by the ex-
istence of continuous partial derivatives of n + 1-th order.
Remark 2. Taylor’s theorem is a generalization of the Mean Value Theo-
rem.
Applications
Criterion of a constant function. A differentiable on a connected set

S function f (x, y) is constant on S if, and only if, df (x, y) = 0, ∀ (x, y) ∈ S.
Critical (stationary) point. If df (x0 , y0 ) = 0, then (x0 , y0 ) is called a

critical (stationary) point.
Theorem (necessary condition of a local extremum). If f (x, y) is

differentiable at a local extremum point (x0 , y0 ), then df (x0 , y0 ) = 0 .
Second differential test (sufficient condition of a local ex-

tremum). Let f (x, y) be twice differentiable at a critical point (x0 , y0 ). Then
1) if d2 f (x0 , y0 ) maintains the same sign for small increments ∆x, ∆y, then
(x0 , y0 ) is a strict local extremum; more specifically, if d2 f (x0 , y0 ) > 0, then
(x0 , y0 ) is a strict local minimum, and if d2 f (x0 , y0 ) < 0, then (x0 , y0 ) is a
strict local maximum;
2) if d2 f (x0 , y0 ) changes its sign for small increments ∆x, ∆y, then (x0 , y0 ) is
not a local extremum.
Second derivatives test (sufficient condition of a local ex-

tremum). Let f (x, y) be twice continuously differentiable at a critical point
(x0 , y0 ). Then ( )
fxx fxy
1) if the Hessian matrix H (f ) = is definite at (x0 , y0 ), that is,
fxy fyy
if the determinant of the Hessian matrix is positive at (x0 , y0 ), then (x0 , y0 )
is a strict local extremum; more specifically, if H (f ) is positive definite, that
is, if fxx > 0 and fxx fyy − fxy 2
> 0, then (x0 , y0 ) is a strict local minimum;
if H (f ) is negative definite, that is, if fxx < 0 and fxx fyy − fxy 2
> 0, then
(x0 , y0 ) is a strict local maximum;
2) if the Hessian matrix is indefinite at (x0 , y0 ), that is, if fxx fyy − fxy 2
< 0 at
(x0 , y0 ), then (x0 , y0 ) is not a local extremum.
General Remark. Like in section 7.2, the examples in this section are mod-
ified versions of some examples for one-variable functions presented in Part I
and, correspondingly, the majority of the provided below counterexamples use
the functions of the chapter 4 adapted to the case of R2 . Each example in this
section has the reference to the corresponding one-dimensional counterpart.
8.2.1 Concepts and local properties
Example 1. “If f (x, y) is not differentiable at a point (a, b), then it is not
Solution. √
The function f (x, y) = x2 + y 2 is continuous at the origin, since
lim f (x, y) = 0 = f (0, 0). However, the restriction of f (x, y) to x-
(x,y)→(0,0)
axis is the absolute value function f (x, 0) = |x|, which is not differentiable at
0. Therefore, f (x, y) does not possess the partial derivative in x at the origin,
that implies that f (x, y) is not differentiable at the origin.
Remark. The Remarks to Example 1 in Section 4.2 are applicable here
Example 2. “If a function f (x, y) is not differentiable at a point (a, b),

then f 2 (x, y) also is not differentiable at (a, b).”
Solution. √
The function f (x, y) = x2 + y 2 of Example 1 is not differentiable at the
origin, but its square g (x, y) = f 2 (x, y) = x2 + y 2 is differentiable function
on R2 , since the partial derivatives gx (x, y) = 2x and gy (x, y) = 2y are the
continuous function on R2 .
Remark. The Remarks to Example 4 in section 4.2 are applicable here
Example 3. “If h (x, y) = f (x, y) + g (x, y) is differentiable at a point

(a, b), then both f (x, y) and g (x, y) are differentiable at (a, b).”
Solution. √ √
The functions f (x, y) = x2 + y 2 and g (x, y) = − x2 + y 2 are not
differentiable at (0, 0) (see Example 1), but their sum h (x, y) = f (x, y) +
g (x, y) ≡ 0 is differentiable on R2 .
Example 4. “If h (x, y) = f (x, y) · g (x, y) is differentiable at a point (a, b)

and f (x, y) is differentiable at the same point, then g (x, y) is also differen-
tiable at (a, b).”
Solution.
√ The function f (x, y) = x2 + y 2 is differentiable at (0, 0) and√g (x, y) =
3
x2 + y 2 is not differentiable at (0, 0) (for the same reason as x2 + y 2 ).
( )4/3
However, their product h (x, y) = f (x, y) · g (x, y) = x2 + y 2 is differen-
tiable at the origin, because both partial derivatives exist at the origin
f (x, 0) − f (0, 0) x8/3
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) y 8/3

fy (0, 0) = lim = lim =0
y→0 y−0 y→0 y
and the remainder in the definition of differentiability

( )4/3
α (x, y) = f (x, y) − (f (0, 0) + fx (0, 0) · x + fy (0, 0) · y) = x2 + y 2
satisfies the required evaluation

α (x, y) ( 2 )5/6
lim = lim x + y2 = 0.
(x,y)→(0,0) ρ (x, y) (x,y)→(0,0)

Example 5. “If f (x, y) is not differentiable at a point (a, b) and g (t) is

not differentiable at the point f (a, b), then g (f (x, y)) is not differentiable at
(a, b).”
Solution. {
x2 +y 2 , (x, y) ̸= (0, 0) is not differentiable at
1
0 , (x, y) = (0, 0)
{ 1
(0, 0) and the function g (t) = t , t ̸= 0 is not differentiable at 0, but the
0, t = 0
function g (f (x, y)) = x2 + y 2 is differentiable at (0, 0).
Example 6. “If f (x, y) is differentiable on R2 , then at least one of the

partial derivatives is bounded in a neighborhood of each point.”
Solution.
{ ( 2 )
2 , (x, y) ̸= (0, 0)
1
x + y 2 sin x2 +y
The function f (x, y) = is differen-
0 , (x, y) = (0, 0)
tiable on R2 . Indeed, for any (x, y) ̸= (0, 0), both partial derivatives
1 2x 1
fx (x, y) = 2x sin − 2 cos 2 ,
x2 + y 2 x + y2 x + y2
and
1 2y 1
fy (x, y) = 2y sin − 2 cos 2
x2 + y 2 x + y2 x + y2
exist and are continuous according to the arithmetic and chain rules, that
implies continuous differentiation on R2 \ {(0, 0)}. For the origin, first we es-
tablish the existence of the partial derivatives:
/
f (x, 0) − f (0, 0) x2 sin 1 x2
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
/
f (0, y) − f (0, 0) y 2 sin 1 y 2
fy (0, 0) = lim = lim = 0,
y→0 y−0 y→0 y
and then evaluate the remainder in the definition of differentiability:
( 2 ) 1
α (x, y) ∆f − df x + y 2 sin x2 +y 2
lim = lim = lim √
(x,y)→(0,0) ρ (x, y) (x,y)→(0,0) ρ (x, y) (x,y)→(0,0) x2 + y 2
√ 1
= lim x2 + y 2 sin = 0.
(x,y)→(0,0) x2 + y2
Therefore, f (x, y) is differentiable at the origin as well, and the statement
condition
( )is satisfied. However, by choosing the sequence of the points (xk , 0) =
√ 1
, 0 , k ∈ N which approach the origin, we obtain lim fx (xk , 0) =
2kπ
( √ ) xk →0
lim −2 2kπ cos 2kπ = −∞, that is, fx (x, y) is unbounded in any neigh-
k→+∞
borhood of( the origin.
) In a similar way, approaching the origin by the points
(0, yn ) = 0, 2nπ , n ∈ N, it can be shown that fy (x, y) is also unbounded
√ 1
in any neighborhood of the origin.
Example 7. “There is no function differentiable only at one point of its

domain.”
Solution.
Using the same reasoning {as in Example √ 9, section 7.2, one can prove
x2 +√ y 2
, x2 + y 2 ∈ Q
that the function f (x, y) = is not continuous
0 , x + y2 ∈ I
2
(and consequently not differentiable) at any point (x, y) ̸= (0, 0). However,
at the origin, both partial derivatives exist fx (0, 0) = 0 and fy (0, 0) = 0,
that means √ that the remainder assumes the form α (x, y) = ∆f − df =

{ 2
x + √ y 2
, x2 + y 2 ∈ Q
and for all points in a neighborhood of the ori-
0 , x2 + y 2 ∈ I
gin it can be evaluated as follows: |α (x, y)| ≤ x2 + y 2 . Therefore, |α(x,y)|
ρ(x,y) ≤
√ α(x,y)
2 2
x + y and according to the squeese theorem lim
(x,y)→(0,0)ρ(x,y) = 0, that is,
f (x, y) is differentiable at the origin.
8.2.2 Global properties and applications
Example 8. “If both functions f (x, y) and g (x, y) are differentiable on

R2 and f (x, y) > g (x, y), ∀(x, y) ∈ R2 , then there exists at least one point in
R2 where fx (x, y) > gx (x, y) or fy (x, y) > gy (x, y).”
Solution.
Both functions f (x, y) = e−x + e−y and g(x, y) = −e−x − e−y are differen-
tiable on R2 and f (x, y) > g(x, y), ∀(x, y) ∈ R2 . However, fx (x, y) = −e−x <
e−x = gx (x, y) and fy (x, y) = −e−y < e−y = gy (x, y), ∀(x, y) ∈ R2 .
Example 9. “If df (x, y) = 0 on a set S then f (x, y) = const on this set.”

Solution. {
1 , (x, y) ∈ B1 (0, 0)
Consider f (x, y) = with the set S = B1 (0, 0) ∪
2 , (x, y) ∈ B1 (2, 0)
B1 (2, 0), where B1 (0, 0) and B1 (2, 0) are the open unit balls, centered at
(0, 0) and (2, 0), respectively. This function is differentiable at each point of S
and df (x, y) = 0 on S, but the function is not a constant.
Remark 1. The statement will be correct if the set S is connected.
Remark 2. A generalization of this Example can be found in Example 17,
section 8.3.
Example 10. “A tangent plane touches or intersects the surface at exactly

one point.”
Solution.
The tangent plane to the graph of the function f (x, y) = cos x at the
origin is the plane z = 1, which touches the graph of f (x, y) at all the points
(xn , y) = (2nπ, y) , ∀n ∈ Z, ∀y.
The tangent plane to the graph of the function f (x, y) = x sin x at the
origin is the plane z = 0, which crosses the graph (at all) points (xn , y) =
(nπ, y) , ∀n ∈ Z , ∀y. At the same time, at the point π2 , 0 this function has
the tangent( plane z =) x, which touches the graph of f (x, y) at all the points
(xk , y) = π2 + 2kπ, y , ∀k ∈ Z, ∀y and also ( crosses
) the graph at the origin.
The tangent plane z = −x at the point −( π2 , 0 has similar ) properties: it
touches the graph at all the points (xm , y) = − π2 + 2mπ, y , ∀m ∈ Z, ∀y and
crosses the graph at the origin.
Example 11. “If df (x0 , y0 ) = 0, then the point (x0 , y0 ) is a local ex-
tremum of the function f (x, y).”
Solution.
The function f (x, y) = x3 is differentiable on R2 and df (0, 0) = 0, but
(0, 0) is not a local extremum, since in any neighborhood of the origin f (x, y)
assumes both negative (for x < 0) and positive (for x > 0) values.
Another counterexample, classical for two-variable functions, is the exam-
ple of the hyperbolic paraboloid f (x, y) = x2 − y 2 that has a saddle point at
the origin. Again, f (x, y) is differentiable on R2 and df (0, 0) = 0, but (0, 0)
is not a local extremum, since in any neighborhood of the origin f (x, y) as-
sumes both negative (on y-axis) and positive (on x-axis) values. Furthermore,
f (x, 0) = x2 is upward parabola with the minimum point at (0, 0), while
f (0, y) = −y 2 is downward parabola with the maximum point at (0, 0). Such
points are called saddle points.
Example 12. “If f (x, y) has a strict local extremum at (x0 , y0 ), then
fx (x0 , y0 ) = fy (x0 , y0 ) = 0.”

Solution.
The function f (x, y) = |x| + |y| has a strict local (and global) minimum
at the origin, since for any (x, y) ̸= (0, 0) it holds f (x, y) = |x| + |y| >
0 = f (0, 0). But the partial derivatives does not exist at the origin, because
lim f (x,0)−f
x−0
(0,0)
= lim xx = 1 while lim f (x,0)−f
x−0
(0,0)
= lim −xx = −1;
x→0+ x→0+ x→0− x→0−
and the same considerations are true for the derivative in y. √
Another similar counterexample is the function f (x, y) = x2 + y 2 ,
which has a strict local (and global)√ minimum at the origin, since for any
(x, y) ̸= (0, 0) it holds f (x, y) = x2 + y 2 > 0 = f (0, 0). However, √the par-
tial derivatives does not exist at the origin: lim f (x,0)−f (0,0) 2
x−0 = lim xx = 1,
x→0+ x→0+
√
f (x,0)−f (0,0)
while lim x−0 = lim x2
x = −1; and the same considerations are
x→0− x→0−
true for the derivative in y.
Remark. This is a wrong modification of the necessary condition for a
local extremum: if f (x, y) is differentiable at a point (x0 , y0 ) and this point is
a local extremum, then fx (x0 , y0 ) = fy (x0 , y0 ) = 0.
Example 13. “If (x0 , y0 ) is a local extremum of f (x, y) and f (x, y) is

twice differentiable at (x0 , y0 ), then d2 f (x0 , y0 ) ̸= 0 .”
Solution.
The function f (x, y) = x4 + y 4 is twice differentiable at the origin, which

is its local (and global) minimum. The second differential has the form
d2 f = fxx ∆x2 + 2fxy ∆x∆y + fyy ∆y 2 = 12x2 ∆x2 + 12y 2 ∆y 2
and it vanishes at the origin, that contradicts the above statement.

Remark. This is the false converse to the Second Differential test: if f (x, y)
is twice differentiable at a critical point (x0 , y0 ) and d2 f (x0 , y0 ) ̸= 0, then
(x0 , y0 ) is a local extremum of f (x, y).
Example 1. “If a function has partial derivatives at some point, then it

is continuous at this point.”
Solution. { xy
x2 +y 2 , x + y ̸= 0
2 2
The function of Example 1, section 7.3, f (x, y) = 2 2
0, x + y = 0
has partial derivatives at the origin:
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim =0
x→0 x−0 x→0 x
and
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Moreover, both partial derivatives exist at each point in R2 and for (x, y) ̸=
y (y 2 −x2 )
(0, 0) can be calculated by applying the arithmetic rules: fx (x, y) = (x2 +y2 )2
x(x2 −y 2 )
and fy (x, y) = (x2 +y2 )2 . Nevertheless, f (x, y) is not continuous at the origin
(see proof in Example 1, section 7.3).
A similar example
{ can be provided with the function of Example 2, section
x2 y
x4 +y 2 , x + y ̸= 0 . Again, the partial derivatives at the
2 2
7.3, f (x, y) = 2 2
0, x + y = 0
origin should be evaluated by definition:
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0,
y→0 y−0 y→0 y
2xy (y 2 −x4 )
and at any other point one can use the arithmetic rules: fx (x, y) = (x4 +y 2 )2
,
x2 (x4 −y 2 )
fy (x, y) = (x4 +y2 )2 . However, it was shown in Example 2, section 7.3 that
f (x, y) is not continuous at the origin.
Remark 1. In both examples, although the partial derivatives exist on
R2 , it can be shown that they are discontinuous at the origin. For example,
the partial limit of fx (x, y) of the first function along the path y = 2x is
2
lim fx (x, 2x) = lim 2x·3x (5x2 )2
= ∞.
x→0 x→0
Remark 2. For functions of one variable, the existence of derivative implies
the continuity (at a point or on a set). The situation is different for functions
of several variables.
Example 2. “If all the directional derivatives and partial derivatives of

f (x, y) exist and are equal at some point, then f (x, y) is continuous at the
same point.”
Solution. {
x6
(y−x2 )2 +x6
, x2 + y 2 ̸= 0
The function f (x, y) = 2 2
has zero partial
0, x + y = 0
derivatives at the origin:
f (x, 0) − f (0, 0) x6 x
fx (0, 0) = lim = lim 4 = lim =0
x→0 x−0 x→0 (x + x6 ) x x→0 1 + x2
and
f (0, y) − f (0, 0) 0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
The directional derivatives are also zero:

f (αt, βt) − f (0, 0) α 6 t6 1
Du f (0, 0) = lim = lim
t→0+ t−0 t→0+ t2 (β − α2 t)2 + α6 t6 t
α 6 t3
= lim 2 =0
t→0+ (β − α2 t) + α6 t4
in any direction u = (α, β). Nevertheless, f (x, y) is not continuous at the
origin, because along the path y = x2 one gets
( ) x6
f x, x2 = = 1 ̸→ 0 = f (0, 0) .
0 + x6 x→0
Remark 1. Consequently, this function is not differentiable at the origin.

Remark 2. This is a strengthened version of Example 1.
Example 3. “If a function is continuous at a point (x0 , y0 ) and has partial

derivatives at this point, then it is differentiable at (x0 , y0 ).”
Solution.
√
The function f (x, y) = |xy| satisfies the conditions
√ of the statement at
the origin. In fact, lim f (x, y) = lim |xy| = 0 and
(x,y)→(0,0) (x,y)→(0,0)
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
However, f (x, y) is not differentiable at the origin, because the remainder in
the definition of differentiability
√
α (x, y) = f (x, y) − (f (0, 0) + fx (0, 0) · x + fy (0, 0) · y) = |xy|
√
does not approach zero faster than ρ = x2 + y 2 : along the path y = x one
gets
α |x| 1
lim = lim √ = √ ̸= 0.
ρ→0 ρ x→0 2 |x| 2
√ Another counterexample can be given with the function f (x, y) =

3
x3 + y 3 considered in a neighborhood of the origin. This function is contin-
uous and has partial derivatives at the origin, since
√3
lim f (x, y) = lim x3 + y 3 = 0
(x,y)→(0,0) (x,y)→(0,0)
and √
f (x, 0) − f (0, 0)
3
x3
fx (0, 0) = lim = lim = 1,
x→0 x−0 x→0 x
√
f (0, y) − f (0, 0) 3
y3
fy (0, 0) = lim = lim = 1.
y→0 y−0 x→0 x
To check differentiability at the origin, let us evaluate the difference between

f (x, y) and its linear approximation near the origin:
√
α (x, y) = f (x, y) − (f (0, 0) + fx (0, 0) · x + fy (0, 0) · y) = 3 x3 + y 3 − x − y.
Along the line y = x we have

√ √
2x3 − 2x 2−2
3 3
α (x, x)
lim = lim √ = lim √ sgn x ̸= 0
x→0 ρ (x, x) x→0 2 |x| x→0 2
α(ρ)
(the last limit does not exist), therefore lim ρ ̸= 0 as it should be for a
ρ→0
differentiable function. Hence, f (x, y) is not differentiable at the origin.
Example 4. “If f (x, y) is continuous and has bounded partial derivatives

on R2 , then f (x, y) is differentiable on R2 .”
Solution. {
√ xy , x2 + y 2 ̸= 0
The function f (x, y) = x +y 2
2
is continuous at any
0, x + y 2 = 0
2
point (x, y) ̸= (0, 0) due to arithmetic rules, and at the origin we should
appeal to the definition:
|xy| 2 |xy| 1 √ 2 2 1√ 2
|f (x, y)−f (0, 0)| = √ = 2 2· x +y ≤ x + y2 → 0.
2
x +y 2 x +y 2 2 (x,y)→(0,0)
In the same way, the partial derivatives can be calculated at any point (x, y) ̸=
(0, 0) according to the arithmetic rules:
y3 x3
fx (x, y) = 3/2
, fy (x, y) = 3/2
(x2 + y2 ) (x2 + y 2 )
and at the origin by using the definition:
f (x, 0) − f (0, 0) 0−0
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Additionally, the partial derivatives are bounded on R2 :
3
y y2 |y|
|fx (x, y)| = 3/2
= 2 + y2
· 1/2
≤ 1,
2
(x + y )2 x (x + y 2 )
2
and similarly |fy (x, y)| ≤ 1. Nevertheless, the function is not differentiable at
the origin. In fact, the difference between the function and its linear approxi-
mation is
xy
α (x, y) = f (x, y) − (f (0, 0) + fx (0, 0) · x + fy (0, 0) · y) = √ .
x2 + y 2
α(x,x) x2 1
Therefore, along the path y = x one has lim = lim 2 = 2, which
x→0 ρ(x,x) x→0 2x
contradicts to the definition of differentiability.
Remark. The partial derivatives of the function are continuous at each
point different from the origin, which implies that the function is differentiable
at each point different from the origin.
Example 5. “If f (x, y) is continuous and possesses partial derivatives and

all the directional derivatives at some point, then f (x, y) is differentiable at
this point.”
Solution. √
It was shown in Example 3 that the function f (x, y) = |xy| is continuous
and possesses partial derivatives at the origin, but it is not differentiable at
the origin. It remains to prove that f (x, y) has all the directional derivative.
Indeed, choosing an arbitrary unitary direction vector u = (α, β), it follows
by definition that
√
f (αt, βt) − f (0, 0) |αβ| t2 − 0 √
Du f (0, 0) = lim = lim = |αβ|.
t→0+ t−0 t→0+ t
{ One can show that a similar example is provided by the function f (x, y) =
√ xy , x2 + y 2 ̸= 0
x2 +y 2 from Example 4.
0, x2 + y 2 = 0
Example 6. “If f (x, y) is continuous and has bounded partial deriva-
tives on R2 , and for any differentiable curve x (t) , y (t) passing through
the origin and such that (x′ (0)) + (y ′ (0)) > 0 the composite function
2 2
g (t) = f (x (t) , y (t)) is differentiable for ∀t ∈ R, then f (x, y) is differentiable

at the origin.”
Solution. {
x3
x2 +y 2 , x + y ̸= 0 satisfies
2 2
Let us show that the function f (x, y) = 2 2
0, x + y = 0
the conditions of the statement. First, this function is continuous at any point
(x, y) ̸= (0, 0) due to arithmetic rules. At the origin we should appeal to the
definition:
x2
|f (x, y) − f (0, 0)| = |x| ≤ |x| → 0,
x2 + y2 (x,y)→(0,0)
so lim f (x, y) = 0 = f (0, 0). Then the partial derivatives can be calcu-
(x,y)→(0,0)
lated at any point but the origin by applying the arithmetic rules:
( )
x2 x2 + 3y 2 −2x3 y
fx (x, y) = 2 , fy (x, y) = 2.
(x2 + y 2 ) (x2 + y 2 )
At the origin we apply the definition:
f (x, 0) − f (0, 0) x
fx (0, 0) = lim = lim = 1,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0

fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Moreover, the partial derivatives are bounded on R2 :
( ) ( )
x2 x2 + 3y 2 x2 x2 + y 2 2x2 y 2
|fx (x, y)| = 2 = 2 + 2 ≤1+1=2
2
(x + y )2 2 2
(x + y ) (x + y 2 )
2
and
2 x3 y x2 2 |xy|
|fy (x, y)| = 2 = · ≤ 1 · 1 = 1.
(x2 + y2 ) (x2 + y 2 ) (x2 + y 2 )
Let us prove the additional condition of this example. Since the given
curve x (t) , y (t) is differentiable and the partial derivatives of f (x, y) exist,
the composite function g (t) = f (x (t) , y (t)) can be differentiated by applying
the chain rule: for any t ̸= 0 we obtain
x4 + 3x2 y 2 2x3 y
g ′ (t) = fx (x, y)·x′ (t)+fy (x, y)·y ′ (t) = 2 ·x′ (t)− 2 ·y
′
(t) ,
(x2 + y2 ) (x2 + y 2 )
and for t = 0 we have
g ′ (0) = fx (0, 0) · x′ (0) + fy (0, 0) · y ′ (0) = 1 · x′ (0) + 0 · y ′ (0) = x′ (0) .
Hence, all the conditions of the statement hold. Nevertheless, f (x, y) is not
differentiable at the origin. Indeed, the difference between the function and
its linear approximation in a neighborhood of the origin is as follows:
x3 xy 2
α (x, y) = f (x, y)−(f (0, 0)+fx (0, 0)·x+fy (0, 0)·y) = − x=− .
x2 +y 2 x2 +y 2
Then, along the path y = x the ratio √α(x,y)

2 2
does not tend to zero:
x +y
α (x, x) −x3 −1
lim = lim √ = lim √ sgn x ̸= 0.
x→0 ρ (x, x) x→0 2 2 |x| x2 x→0 2 2
Therefore, f (x, y) is not differentiable at the origin.
Remark to Examples 3-6. These examples represent a sequence of refined,

but still insufficient, conditions for differentiation.
Remark to Examples 1-6. These examples show that when the partial
derivatives are discontinuous at some point, then the function may be non-
differentiable at this point. Continuity of the partial derivatives guarantees
the differentiability, but this is a sufficient condition as it is shown in the next
Example 7.
Example 7. “If f (x, y) is differentiable at a point (x0 , y0 ), then its partial

derivatives are continuous at this point.”
Solution. { ( 2 ) 1
x + y 2 sin x2 +y 2
2 , x + y
2
̸= 0
The function f (x, y) = 2 2 satisfies the
0, x + y = 0
conditions of the statement at the origin. First, let us find its partial deriva-
tives. For any point (x, y) ̸= (0, 0), applying the arithmetic rules we obtain:
1 2x 1
fx (x, y) = 2x sin − 2 cos 2 ,
x2 + y 2 x + y2 x + y2
1 2y 1
fy (x, y) = 2y sin − 2 cos 2 .
x2 + y 2 x + y2 x + y2
At the origin we should appeal to the definition:
f (x, 0) − f (0, 0) x2 sin x12

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) y 2 sin y12

fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Second, the remainder in the definition of differentiability at the origin takes
the form
( ) 1
α (x, y) = f (x, y) − (f (0, 0)+fx (0, 0) · x+fy (0, 0) · y) = x2 +y 2 sin ,
x2 +y 2
and consequently,
α (x, y) √ 1
lim = lim x2 + y 2 sin 2 =0
(x,y)→(0,0) ρ (x, y) (x,y)→(0,0) x + y2
due to the following evaluation

√ 1 √ 2
x2 + y 2 sin ≤ x + y2 → 0.
x2 + y 2 (x,y)→(0,0)
Therefore, f (x, y) is differentiable at the origin. (Notice, that f (x, y) is also

differentiable at each point different from the origin, because its partial deriva-
tives are continuous there.) Now let us show that the partial derivatives are dis-
continuous at the origin. Indeed, choosing the sequence of the points (xn , yn ),
xn = √2nπ1
→ 0, yn = 0 → 0, we obtain
n→∞ n→∞
2 2
fx (xn , yn ) = 2xn sin 2nπ − cos 2nπ = − → −∞,
xn xn n→∞
which means that fx (x, y) is discontinuous (and even unbounded) at the
origin. Similarly for (xk , yk ), xk = 0 → 0, yk = √2kπ
1
→ 0, one gets
k→∞ k→∞
fy (xk , yk ) → −∞, that is, fy (x, y) is also discontinuous and unbounded
k→∞
at the origin.
Another counterexample is provided by the function


 x2 sin x1 + y 2 sin y1 , xy ̸= 0

x2 sin x1 , x ̸= 0 , y = 0
f (x, y) =

 y 2 sin y1 , x = 0 , y ̸= 0

0, x = y = 0
Remark. The considered counterexample shows that the above statement

can be reformulated in the following (still false) form: “if f (x, y) is differen-
tiable on R2 , then its partial derivatives are continuous at the origin”.
Example 8. “If f (x, y) is differentiable at a point (x0 , y0 ) and one of its

partial derivatives is continuous at this point, then another partial derivative
is also continuous at (x0 , y0 ).”
Solution. {
4x6 y 2
(x4 +y 2 )2
, x2 + y 2 ̸= 0
Let us consider the function f (x, y) = in a
0, x + y 2 = 0
2
neighborhood of the origin. The partial derivatives can be found by applying

the arithmetic rules for any point different from the origin:
24x5 y 4 − 8x9 y 2 8x10 y − 8x6 y 3

fx (x, y) = 3 , fy (x, y) = 3 ;
(x4 + y2 ) (x4 + y 2 )
and by using the definition at the origin:
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Since both partial derivatives are zero at the origin, the remainder in the
6 2
definition of differentiability takes the form α (x, y) = f (x, y) = (x4x y
4 +y 2 )2 , and
consequently,
( )2
α (x, y) 4x6 y 2 2x2 y |x|
= √ = ·√ · |x| ≤ |x| → 0.
ρ (x, y) (x4 +y 2 )2 x2 +y 2 x4 +y 2 x2 +y 2 (x,y)→(0,0)
Therefore, f (x, y) is differentiable at the origin. Notice that due to the arith-
metic rules of continuity both partial derivatives are continuous at each point
different from the origin. It guarantees differentiability of f (x, y) on R2 . Fi-
nally, the continuity of fx (x, y) at the origin follows from the evaluation:

24x5 y 4 − 8x9 y 2
|fx (x, y) − fx (0, 0)| = 3
(x4 + y 2 )
( )2 ( )2
2x2 y y2 2x2 y x4
≤6 · 4 · |x| + 2 · · |x|
x4 + y 2 x + y2 x4 + y 2 x4 + y 2
≤ 6 |x| + 2 |x| = 8 |x| → 0.

(x,y)→(0,0)
Hence, all the conditions of the statement hold. However, fy (x, y) is not con-
tinuous at the origin, because along the path y = 2x2 we obtain
( ) 16x12 − 64x12 48
lim fy x, 2x2 = lim 3 =− ̸= 0 = fy (0, 0) .
x→0 x→0 4
(5x ) 125
{ Another counterexample can be given with the function f (x, y) =

x2 sin x1 + y 2 , x ̸= 0
. The partial derivatives are readily calculated by
y2 , x = 0
applying the arithmetic rules at each point out of the y-axis:
1 1
fx (x, y) = 2x sin − cos , fy (x, y) = 2y.
x x
For the points of the y-axis we appeal to the definition: at the origin we have
f (x, 0) − f (0, 0) x2 sin 1/x − 0 1

fx (0, 0) = lim = lim = lim x sin = 0,
x→0 x−0 x→0 x x→0 x
f (0, y) − f (0, 0) y2 − 0
fy (0, 0) = lim = lim = 0,
y→0 y−0 y→0 y
and for any other point (0, y0 ) , y0 ̸= 0 we obtain
f (x, y0 )−f (0, y0 ) x2 sin x1 +y02 −y02 1

fx (0, y0 ) = lim = lim = lim x sin = 0,
x→0 x−0 x→0 x x→0 x
f (0, y) − f (0, y0 ) y 2 − y02
fy (0, y0 ) = lim = lim = 2y0 .
y→y0 y − y0 y→y0 y − y0
Since both partial derivatives are continuous on R2 \ (0, y), f (x, y) is dif-
ferentiable on R2 \ (0, y). Let us verify differentiability of f (x, y) on the y-
axis. At the{origin the remainder of the linear approximation is α (x, y) =
x2 sin x1 + y 2 , x ̸= 0
f (x, y) = . For the points of the y-axis we have
y2 , x = 0
2
α(0,y)
ρ(0,y) =
√y 2 = |y| → 0 and for the points outside of the y-axis we
y (x,y)→(0,0)
obtain

α (x, y) x2 sin 1/x + y 2 x2 + y 2 √

ρ (x, y) = √x2 + y 2 ≤ √x2 + y 2 = x + y (x,y)→(0,0)
→
2 2 0.
Therefore, f (x, y) is differentiable at the origin. For any other point

(0, y0 ) , y0 ̸= 0 the remainder is
α (x, y) = f (x, y) − (f (0, y0 ) + fx (0, y0 ) · x + fy (0, y0 ) · (y − y0 ))

{ 2
x sin x1 + y 2 , x ̸= 0 ( )
= 2 − y02 + 2y0 (y − y0 ) .
y , x=0
Therefore, for the points of the y-axis we have
2
α (0, y) y 2 + y 2 − 2y0 y (y − y0 )
= √ 0 =√ → 0,
ρ (0, y) 2 2 (x,y)→(0,0)
(y − y0 ) (y − y0 )
and for the points outside of the y-axis we obtain

α (x, y) x2 sin 1/x + (y − y0 )2
= √
ρ (x, y)
x + (y − y0 )
2 2
2 √
x2 + (y − y0 ) 2
≤√ = x2 + (y − y0 ) → 0.
2 (x,y)→(0,y0 )
x2 + (y − y0 )
It implies that f (x, y) is differentiable at any point (0, y0 ) , y0 ̸= 0. Summa-
rizing, f (x, y) is differentiable on R2 . Notice that the formulas found for the
partial derivative in y can be rewritten in the form fy (x, y) = 2y on R2 , so
this derivative is continuous on R2 . Hence, all the conditions of the statement
hold. However, fx (x, y) is discontinuous at each point of the y-axis, because
1
approaching a point (0, y) by the sequence of the points (xn , y), xn = 2nπ ,
∀n ∈ Z\ {0}, we obtain
( )
1
lim fx (xn , y) = lim sin 2nπ − cos 2nπ = −1 ̸= 0 = fx (0, y) .
xn →0 n→∞ nπ
Remark. Notice that both functions are also counterexamples to the fol-
lowing version of the false statement: “if f (x, y) is differentiable on R2 and one
of its partial derivatives is continuous on R2 , then another partial derivative
is also continuous on R2 ”.
Example 9. “If f (x, y) has partial derivatives at some point, then f (x, y)
has all the directional derivatives at the same point.”
Solution. { xy
2 , x +y ̸= 0
2 2
2
It was shown in Example 1 that the function f (x, y) = x +y 2 2
0, x + y = 0
has zero partial derivatives at the origin (notice that its partial derivatives
exist at every point in R2 ). However, f (x, y) has no directional derivative at
the origin, except for the directions along the coordinate axes. In fact, let
u = (α, β) be an arbitrary unitary direction vector not parallel to coordinate
axes, that is, αβ ̸= 0. Then evaluating the directional derivative by definition
one obtains
αβt2
f (αt, βt) − f (0, 0) (α2 +β 2 )t2 −0 αβ
Du f (0, 0) = lim = lim = lim = ∞,
t→0+ t−0 t→0+ t t→0+ t
that is, these derivatives do not exist.
Example 10. “If f (x, y) is continuous and possesses all the directional
derivatives at some point, then there exist a partial derivative of f (x, y) at
this point.”
Solution.

√
The graph of the function f (x, y) = x2 + y 2 is an infinite cone, which
visually seems to be continuous at the origin and has all the directional deriva-
tives at the origin because the restriction of f (x, y) to any ray starting from
the origin is just a half-line. The analytical proof is also simple. First, f (x, y)
is continuous because it is a irrational function defined on R2 . Second, in the
direction of an arbitrary unitary vector u = (α, β), the calculation of the
directional derivatives gives:
√
f (αt, βt) − f (0, 0) α2 t2 + β 2 t2 − 0
Du f (0, 0) = lim = lim = 1.
t→0+ t−0 t→0 + t
However, both partial derivatives do not exist at the origin, since on the
coordinates axis x and y the function assumes the form f (x, 0) = |x| and
f (0, y) = |y|, respectively.
Remark. Consequently, this function is not differentiable at the origin.
Example 11. “If f (x, y) possesses second-order partial derivatives, then

the mixed derivatives are equal.”
Solution. {
xy xx2 −y
2 2
+y 2 , x + y ̸= 0 . Apply-
2 2
Let us consider the function f (x, y) =
0, x2 + y 2 = 0
ing arithmetic rules for partial derivative calculation and simple algebra for
simplification we can obtain the following expressions for partial derivatives

at any point different from the origin:
x4 y + 4x2 y 3 − y 5 −xy 4 − 4x3 y 2 + x5

fx (x, y) = 2 , fy (x, y) = 2 .
(x2 + y2 ) (x2 + y 2 )
At the origin both partial derivatives can be calculated by the definition:
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Therefore, fx (x, y) and fy (x, y) are defined on R2 . To calculate the mixed
second-order derivative at the origin we should again apply the definition:
fx (0, y) − fx (0, 0) −y − 0
fxy (0, 0) = lim = lim = −1,
y→0 y−0 y→0 y
fy (x, 0) − fy (0, 0) x−0

fyx (0, 0) = lim = lim = 1.
x→0 x−0 x→0 x
Hence, both mixed derivatives exist, but they are different. It is simple to
show that the second-order mixed derivatives exist on R2 and coincide in all
{ also exist on R .
2
points but the origin. Other two second order derivatives
xy , |y| ≤ |x|
Another counterexample is the function f (x, y) = . For
−xy, |y| > |x|
|y| < |x| we readily get fx (x, y) = y, fy (x, y) = x, and for |y| > |x|: fx (x, y) =
−y, fy (x, y) = −x. For the points with |y| = |x| ̸= 0 the first-order derivatives
do not exist, because approaching any fixed point (x0 , y0 ) |y0 | = |x0 | ̸= 0 from
the side where |y| < |x| we obtain
f (x, y0 ) − f (x0 , y0 ) xy0 − x0 y0

lim = lim = y0 ,
x→x0 x − x0 x→x0 x − x0
while approaching the same point from the side with |y| > |x| we get
f (x, y0 ) − f (x0 , y0 ) −xy0 − x0 y0

lim = lim = ∞,
x→x0 x − x0 x→x0 x − x0
and similarly for derivative with respect to y. Finally, at the origin both deriva-
tives can be obtained by the definition:
f (x, 0) − f (0, 0) 0−0

fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
Hence, the first-order derivatives are defined on R2 , including the origin, but
excluding the two lines y = x and y = −x (without the origin). Since both
partial derivatives are defined at the origin, and the last is a limit point for
their domains, we can consider the second-order partial derivatives at the
origin. In particular, for the mixed derivatives we obtain:
fx (0, y) − fx (0, 0) −y − 0
fxy (0, 0) = lim = lim = −1,
y→0 y−0 y→0 y
fy (x, 0) − fy (0, 0) x−0
fyx (0, 0) = lim = lim = 1.
x→0 x−0 x→0 x
Therefore, the mixed derivatives are different at the origin.
Remark 1. The condition that guarantees the equality of the mixed deriva-
tives is their continuity. So in this Example the mixed derivatives are discon-
tinuous at the origin. Let us show this explicitly for the first function. Applying
arithmetic rules for derivatives and simple algebra we can obtain the following
results for any (x, y) ̸= (0, 0):
x6 + 9x4 y 2 − 9x2 y 4 − y 6 x6 + 9x4 y 2 − 9x2 y 4 − y 6

fxy (x, y) = 3 , fyx (x, y) = 3 ,
(x2 + y2 ) (x2 + y 2 )
that is, the mixed derivatives are continuous and coincide on R2 \ (0, 0). How-
ever, at the origin these derivatives are discontinuous. Indeed, approaching
the origin along the line y = x we obtain:
lim fxy (x, x) = 0 ̸= −1 = fxy (0, 0) and lim fyx (x, x) = 0 ̸= 1 = fyx (0, 0) .
x→0 x→0
Remark 2. The continuity of the mixed derivatives is a sufficient, but not

necessary, condition for their equality, as it is shown in the next Examples 12,
13 and 14.
Example 12. “If f (x, y) possesses first- and second-order partial deriva-
tives at some point (or on R2 ) and the mixed derivatives coincide, then f (x, y)
is continuous at this point (or on R2 ).”
Solution. { 2 2
x4 +y 4 , x + y ̸= 0 . Its first-
x y 2 2
Let us consider the function f (x, y) = 2 2
0, x + y = 0
order derivatives exist on R and can be found by using arithmetic rules at
2
any point different from the origin:
2xy 6 − 2x5 y 2 2x6 y − 2x2 y 5

fx (x, y) = 2 , fy (x, y) = 2 ,
(x4 + y4 ) (x4 + y 4 )
and by applying definition at the origin:
f (x, 0) − f (0, 0) 0−0
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
In the same mode we can calculate the second-order derivatives on R2 . In
particular, for the mixed derivatives we obtain
24x5 y 5 − 4x9 y − 4xy 9
fxy (x, y) = fyx (x, y) = 3
(x4 + y 4 )
for (x, y) ̸= (0, 0), and fxy (0, 0) = fyx (0, 0) = 0. Hence, the conditions of the
statement are satisfied. However, the function is not continuous at the origin,
since along the line y = x we get
x4 1
lim f (x, x) = lim = ̸= 0 = f (0, 0) .
x→0 x→0 2x4 2
Actually, the function even does not have a limit at the origin, since along
4x4 4
another line y = 2x we obtain different result: lim f (x, 2x) = lim 17x 4 = 17 .
x→0 x→0
Remark. Evidently, at least one of the first-order derivatives is discon-
tinuous at the origin, because otherwise the function would be differentiable
(and therefore, continuous) at the origin. It is simple to show explicitly: along
7
−2x7
y = x we get lim fx (x, x) = lim 2x 4x 8 = 0, while along y = 2x we obtain:
x→0 x→0
128x7 −8x7
lim fx (x, 2x) = lim 289x 8 = ∞.
x→0 x→0
Example 13. “If f (x, y) possesses partial derivatives of all orders at some
point (or on R2 ) and the mixed derivatives do not depend on the order of
differentiation, then f (x, y) is continuous at this point (or on R2 ).”
Solution. { ( 2 2
)
exp − xy2 − xy 2 , xy ̸= 0
Let us consider the function f (x, y) = and
0, xy = 0
show that it is discontinuous at the origin. Indeed, along the line x = y one
gets ( 2 )
x x2
lim f (x, x) = lim exp − 2 − 2 = e−2 ̸= 0 = f (0, 0) .
x→0 x→0 x x
Even being discontinuous this function possesses all partial derivatives. We
start with an arbitrary point out of the coordinate axes. In this case, applying
the arithmetic and chain rules we readily obtain
( 2 ) ( )
x y2 2x 2y 2
fx (x, y) = exp − 2 − 2 · − 2 + 3 ,
y x y x
( 2 ) ( )
x y2 2y 2x2
fy (x, y) = exp − 2 − 2 · − 2 + 3 .
y x x y
Apart from the form of these derivatives, we can see that any derivative has
the form ( 2 )
x y2
fxm yn (x, y) = exp − 2 − 2 · P (x, y) ,
y x
where m + n = k is the order of the derivative, and P (x, y) is a polynomial

( x 2y , i,2j) ∈ Z. Indeed, the partial derivatives of each term
i j
with respect to
h (x, y) = exp − y2 − xy 2 · xi y j are
x
( 2 ) ( )
x y2 2x i j 2y 2 i j
hx (x, y) = exp − 2 − 2 · − 2 x y + 3 x y + ix y
i−1 j
y x y x
and
( 2 ) ( )
x y2 2y i j 2x2 i j
hy (x, y) = exp − 2 − 2 · − 2 x y + 3 x y + jx y
i j−1
.
y x x y
So their sum will give again the conjectured form. Notice also that all these
derivatives are continuous when xy ̸= 0 and it implies that the corresponding
mixed derivatives are equal.
Now let us consider the remaining points for which xy = 0. Using the
method of mathematical induction we will prove that at each of these point
the partial derivative of any order is zero. First, we check this for the first-order
derivatives. For fx we obtain the following results at the origin:
f (x, 0) − f (0, 0) 0−0
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
at the points (x0 , 0), x0 ̸= 0 :
f (x, 0) − f (x0 , 0) 0−0
fx (x0 , 0) = lim = lim = 0,
x→x0 x − x0 x→x 0 x − x0
and at the points (0, y0 ), y0 ̸= 0 :

( 2 )
f (x, y0 ) − f (0, y0 ) 1 x y2
fx (0, y0 ) = lim = lim exp − 2 − 02
x→0 x−0 x→0 x y0 x
( 2) ( 2)
x 1 y
= lim exp − 2 · exp − 02
x→0 y0 x x
( 2)
x ( )
= lim exp − 2 · lim t exp −y02 t2 = 1 · 0 = 0
x→0 y0 t→∞
(the substitution t = x1 has been used on the penultimate step). Summa-

rizing, fx (x, y) = 0 at each point such that xy = 0. Absolutely the same
reasoning shows that fy (x, y) = 0 at the same points. Therefore, the base of
the induction is proved.
Let us suppose now that at each point of the coordinate axes fxm yn (x, y) =
0 for all the derivatives of the order m+n = k and show that fxm yn x (x, y) = 0
and fxm yn y (x, y) = 0 at these points. Again, it is sufficient to analyze only
the former derivative, because all considerations for the latter are the same.
At the origin we have
fxm yn (x, 0) − fxm yn (0, 0) 0−0
fxm yn x (0, 0) = lim = lim = 0.
x→0 x−0 x→0 x
At the points (x0 , 0), x0 ̸= 0 the situation is also simple:

fxm yn (x, 0) − fxm yn (x0 , 0) 0−0
fxm yn x (x0 , 0) = lim = lim = 0.
x→x0 x − x0 x→x0 x − x0
For the points (0, y0 ), y0 ̸= 0 analysis is more involved. First, we transform

the definition to the following form:
fxm yn (x, y0 ) − fxm yn (0, y0 )
fxm yn x (0, y0 ) = lim
x→0 x−0
( 2) ( 2)
x 1 y
= lim exp − 2 · P (x, y0 ) exp − 02 ,
x→0 y0 x x
where P (x, y) is the polynomial from the definition of( fxm y)n (x, y), xy ̸= 0.
2
Since the limit of the first multiplier is simple, lim exp − xy2 = 1, we should
x→0 0
analyze the remaining terms. Notice that just like P (x, y0 ), the function
i j
x P (x, y0 ) is also a polynomial with the terms x y0 , i, j ∈ Z. If i ∈ N, then
1
such term tends to zero ( as 2x) → 0. If i = 0, then such term is constant, but
y
it is multiplied by exp − x02 inside the limit, and the last function tends to
zero as x → 0. Finally, for i < 0 we can apply the following calculations:
( 2)
( 2) exp
y
− x02
y0 t−i
lim x exp − 2 = lim
i
= lim
x→0 x x→0 x−i t→∞ exp (y 2 t2 )
0
−i · t−i−1 −i · t−i−2
= lim = lim = . . . = 0.
t→∞ 2ty 2 exp (y 2 t2 ) t→∞ 2y 2 exp (y 2 t2 )
0 0 0 0
Here, we used the substitution x = 1t and then applied L’Hospital’s rule

repeatedly until the power of t in the numerator becomes negative ( 2or) zero.
y
Based on this analysis, we can conclude that lim x1 P (x, y0 ) exp − x02 = 0,
x→0
and consequently, fxm yn x (0, y0 ) = 0 for each point (0, y0 ), y0 ̸= 0. Concluding,
we have shown that at each point of the coordinate axes the partial derivatives
of any order are zero: fxm yn (x, y) = 0. In particular it means that the mixed
derivatives do not depend on the order of differentiation.
Hence, we have proved, that partial derivatives of an arbitrary order can
be expressed in the form:
{ ( 2 2
)
exp − xy2 − xy 2 · P (x, y) , xy ̸= 0
fxm yn (x, y) = ,
0, xy = 0
with the polynomial P (x, y) specified above, and the corresponding mixed
derivatives are equal.
Example 14. “If f (x, y) is continuous and possesses first- and second-
order partial derivatives on R2 , and the mixed derivatives coincide on R2 ,
then f (x, y) is differentiable on R2 .”
Solution. {
x3 y 2
, x2 + y 2 ̸= 0
x4 +y 4
Let us consider the function f (x, y) = . First, the
0, x + y 2 = 0
2
continuity of f (x, y) at any (x, y) ̸= (0, 0) follows directly from the arithmetic
rules for continuous functions and at the origin it can be shown by the following
evaluation:
3 2
x y 2 2
|f (x, y) − f (0, 0)| = 4 = |x| x y ≤ 1 |x| → 0.
x + y4 x4 + y 4 2 (x,y)→(0,0)
Second, the first-order derivatives can be found by using the arithmetic rules
at any point (x, y) ̸= (0, 0):
3x2 y 6 − x6 y 2 2x7 y − 2x3 y 5

fx (x, y) = 2 , fy (x, y) = 2 ,
(x4 + y4 ) (x4 + y 4 )
and by applying the definition at the origin:
f (x, 0) − f (0, 0) 0−0
fx (0, 0) = lim = lim = 0,
x→0 x−0 x→0 x
f (0, y) − f (0, 0) 0−0
fy (0, 0) = lim = lim = 0.
y→0 y−0 y→0 y
In the same way one can calculate the second-order derivatives: at any (x, y) ̸=
(0, 0) one obtains
−24x5 y 6 + 2x9 y 2 + 6xy 10 2x11 − 24x7 y 4 + 6x3 y 8

fxx (x, y) = 3 , fyy (x, y) = 3 ,
(x4 + y 4 ) (x4 + y 4 )
24x6 y 5 − 6x2 y 9 − 2x10 y

fxy (x, y) = fyx (x, y) = 3
(x4 + y 4 )
and at the origin
fxx (0, 0) = 0, fyy (0, 0) = 0, fxy (0, 0) = fyx (0, 0) = 0.
Hence, the conditions of the statement hold. Nevertheless, the function is not
differentiable at the origin. In fact, the difference between the function and
its linear approximation at the origin has the form:
x3 y 2
α (x, y) = f (x, y) − (f (0, 0) + fx (0, 0) · x + fy (0, 0) · y) = .
x4 + y4
α(x,y) x3√ y2
Along the path y = x, x > 0 the ratio ρ(x,y) = does not
(x4 +y 4 ) x2 +y 2
approach zero:
α (x, x) x5 1
lim = lim √ = √ ̸= 0.
x→0 ρ (x, x) x→0 2 2x5 2 2
Therefore, f (x, y) is not differentiable at the origin.

Remark 1. For any (x, y) ̸= (0, 0) the first-order derivatives are continuous,
which ensures differentiability of the function.
Remark 2. Since f (x, y) is not differentiable at the origin, at least one of
the first-order derivatives is not continuous there, and consequently at least
one of the mixed derivatives is also discontinuous at the origin. It can be shown
explicitly that both fx (x, y) and fy (x, y) are discontinuous at the origin, and
so are fxy (x, y) and fyx (x, y).
Example 15. “If f (x, y) is differentiable at some point, then its partial
derivatives exist in a neighborhood of this point.”
Solution. { 2
x + y 2 , x, y ∈ Q
Let us consider the function f (x, y) = . First, we
0, otherwise
show that f (x, y) is continuous at the origin. Approaching the origin by the
points with rational coordinates, we obtain
( 2 )
lim f (x, y) = lim x + y 2 = 0 = f (0, 0) .
(x,y)→(0,0) (x,y)→(0,0)
x,y∈Q
Approaching the origin by other points we have
lim f (x, y) = lim 0 = 0 = f (0, 0) .

(x,y)→(0,0) (x,y)→(0,0)
x∈Q
/ or y ∈Q
/
Since the points used in these two partial limits cover all the points in a neigh-
borhood of the origin, it follows that lim f (x, y) = f (0, 0), i.e., f (x, y)
(x,y)→(0,0)
is continuous at the origin. Second, let us consider the partial derivatives at
the origin. If x ∈ Q, then
f (x, 0) − f (0, 0) x2 − 0
lim = lim = 0,
x→0, x∈Q x−0 x→0 x
and for x ∈ I we obtain the same result:
f (x, 0) − f (0, 0) 0−0
lim = lim = 0.
x→0, x∈I x−0 x→0 x
Since the union of the rational and irrational points is equal to all the real
points, it follows that fx (0, 0) = lim f (x,0)−f
x−0
(0,0)
= 0. For any point (x0 , y0 ) ̸=
x→0
(0, 0) with y0 ∈ I, fx also exists:
f (x, y0 ) − f (x0 , y0 ) 0−0

fx (x0 , y0 ) = lim = lim = 0.
x→x0 , y0 ∈I x − x0 x→x 0 x − x0
On the other hand, for (x0 , y0 ) ̸= (0, 0) with y0 ∈ Q, the derivative fx does
not exist:
1) if x0 ∈ Q, then for x ∈ Q one has
f (x, y0 ) − f (x0 , y0 ) x2 + y02 − x20 − y02

lim = lim = 2x0 ,
x→x0 , x∈Q x − x0 x→x0 x − x0
while for x ∈ I one obtains
f (x, y0 ) − f (x0 , y0 ) 0 − x20 − y02
lim = lim = ∞,
x→x0 , x∈I x − x0 x→x0 x − x0
f (x,y0 )−f (x0 ,y0 )
and consequently, lim x−x0 does not exist;
x→x0
2) if x0 ∈ I, then again two approaches give different results -
f (x, y0 ) − f (x0 , y0 ) x2 + y02 − 0

lim = lim =∞
x→x0 , x∈Q x − x0 x→x0 x − x0
and
f (x, y0 ) − f (x0 , y0 ) 0−0
lim = lim =0
x→x0 , x∈I x − x0 x→x0 x − x0
f (x,y0 )−f (x0 ,y0 )

showing non-existence of lim x−x0 .
x→x0
Summarizing, fx (x0 , y0 ) = 0 at (x0 , y0 ) = (0, 0) and at (x0 , y0 ) with y0 ∈ I.
For all other points fx does not exist. Due to exactly the same reasoning,
fy (x0 , y0 ) = 0 if (x0 , y0 ) = (0, 0) or (x0 , y0 ) has x0 ∈ I, and fy does not exist
at all other points. Consequently, at each point (x, y) ̸= (0, 0) with rational
coordinates both partial derivatives do not exist, and at each point with ir-
rational coordinates both partial derivatives exist. Since any neighborhood of
the origin contains both the points with rational and irrational coordinates, it
follows that in any neighborhood of the origin there points where both partial
derivatives do not exist.
Finally, let us prove that f (x, y) is differentiable at the origin. Since
fx (0, 0) = fy (0, 0) = 0, the difference between the function and its linear
approximation is α (x, y) = f (x, y). Using the points with rational coordi-
nates we obtain
α (x, y) x2 + y 2
lim = lim √ = 0,
(x,y)→(0,0) ρ (x, y) (x,y)→(0,0) x2 + y 2
and for the rest of points the limit is the same:
α (x, y) 0
lim = lim √ = 0.
(x,y)→(0,0) ρ (x, y) (x,y)→(0,0) x + y2
2
α(x,y)
Therefore, the general limit exists and is equal to zero: lim = 0,
(x,y)→(0,0) ρ(x,y)
which means that the function is differentiable at the origin.
Example 16. “If f (x, y) is differentiable at some point, then it is contin-

uous in a neighborhood of this point.”
Solution. { 2
x + y 2 , x, y ∈ Q
We can use the function f (x, y) = of the previous
0, otherwise
Example. It was already shown above that this function is differentiable (and
therefore, continuous) at the origin. Let us check if f (x, y) is continuous at
(x, y) ̸= (0, 0). First, we consider the points of the y-axis, that is, (0, y0 ),
y0 ̸= 0. If y0 ∈ Q, then approaching the chosen point by the points (x, y) with
x ∈ I, we obtain
lim f (x, y) = lim 0 = 0 ̸= y02 = f (0, y0 ) ,

(x,y)→(0,y0 ) (x,y)→(0,y0 )
x∈I
that is, f (x, y) is not continuous at such points. If y0 ∈ I, then we choose for
approximation the points (x, y) with x, y ∈ Q, and obtain
( 2 )
lim f (x, y) = lim x + y 2 = y02 ̸= 0 = f (0, y0 ) ,
(x,y)→(0,y0 ) (x,y)→(0,y0 )
x,y∈Q
that is, f (x, y) is not continuous at these points. Therefore, f (x, y) is discon-
tinuous at all points (0, y0 ), y0 ̸= 0.
Using the same reasoning we can show that f (x, y) is also discontinuous
at all points (x0 , 0), x0 ̸= 0. In a similar way we can show that f (x, y) is not
continuous at an arbitrary point (x0 , y0 ), x0 ̸= 0, y0 ̸= 0 either. In fact, if
x0 , y0 ∈ Q, then approaching the chosen point by the points (x, y) with x ∈ I,
we obtain
lim f (x, y) = lim 0 = 0 ̸= x20 + y02 = f (x0 , y0 ) ,
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
x∈I
that is, f (x, y) is not continuous at such points. If x0 ∈ I or y0 ∈ I (or both),

then we choose for approximation the points (x, y) with x, y ∈ Q, and obtain
( 2 )
lim f (x, y) = lim x + y 2 = x20 + y02 ̸= 0 = f (x0 , y0 ) ,
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
x,y∈Q
that is, f (x, y) is also discontinuous. Summarizing, we can conclude that

f (x, y) is discontinuous at each point of R2 except at the origin.
Remark. Actually, this counterexample refute the strengthened statement:
“if f (x, y) is differentiable at a point P0 = (x0 , y0 ), then there is at least one
point different from P0 where f (x, y) is continuous”.
Example 17. “If f (x, y) is continuously differentiable in a domain D ⊂ R2

and fy (x, y) = 0, ∀ (x, y) ∈ D, then f (x, y) does not depend on y in D.”
Solution.
Let us consider the first quadrant D1 = {(x, y) : x > 0, y > 0} and the
domain D = R2 \L, where L = {(x,{y) : x ≥ 0, y = 0} is the ray, and let us
x3 , (x, y) ∈ D1
define on D the function f (x, y) = . For ∀ (x, y) ∈
0, (x, y) ∈ D2 = D\D1
D1 we have the continuous partial derivatives: fx (x, y) = 3x2 , fy (x, y) = 0.
For ∀ (x, y) ∈ D2 \L1 , where L1 = {(x, y) : x = 0, y > 0}, the derivatives are
also continuous: fx (x, y) = 0, fy (x, y) = 0. Finally, in the points of L1 we
should use the definitions both for existence and continuity of the partial
derivatives. For fx (0, y0 ) there are two different ways of approaching a point
(0, y0 ) , y0 > 0: using the points in D1 , we obtain
f (x, y0 ) − f (0, y0 ) x3 − 0
fx (0, y0 ) = lim = lim = 0,
x→0 x−0 x→0 x − 0
(x,y)∈D1
and for the points in D2 we have

f (x, y0 ) − f (0, y0 ) 0−0
fx (0, y0 ) = lim = lim = 0.
x→0 x−0 x→0 x − 0
(x,y)∈D2
Since both limits are zero, we conclude that fx (0, y0 ) = 0.

For fy (0, y0 ) all the points in the definition lie on the positive part of the
y-axis, and the calculation is straightforward:
f (0, y) − f (0, y0 ) 0−0
fy (0, y0 ) = lim = lim = 0.
y→y0 y − y0 y→y0 y − y0
Next, we verify the continuity of the partial derivatives on L1 . Again, there

are two different ways of approaching a point (0, y0 ) , y0 > 0: for the paths in
D1 , we obtain
lim fx (x, y) = lim 3x2 = 0,
(x,y)→(0,y0 ) x→0
(x,y)∈D1
and for the paths in D2 we get
lim fx (x, y) = lim 0 = 0.

(x,y)→(0,y0 ) x→0
(x,y)∈D2
Therefore, lim fx (x, y) = 0 = fx (0, y0 ), that is fx (0, y0 ) is continuous

(x,y)→(0,y0 )
on L1 . Finally, since fy (x, y) = 0 at each point of D (including L1 ), it fol-
lows that this derivative is continuous on D. Hence, all the conditions of the
statement hold. Nevertheless, f (1, 1) = 1, while f (1, −1) = 0, that is, f (x, y)
does depend on y.
Remark 1. If the intersection of a region D with any line parallel to the
y-axis is an interval, then the statement will be true.
Remark 2. Recall that for a function g (x) of a single variable there is the
following result: if g ′ (x) = 0 on a connected set (i.e. interval), then g (x) =
const on this set, that is, g (x) does not actually depend on x. We see that
a direct generalization of this result for functions of several variables is not
true.

Example 18. “If ∇f exists and does not vanish at (x0 , y0 ), then the
maximum rate of change of f (x, y) occurs in the direction of ∇f .”
Solution. {
x2 y
x4 +y 2 + y , x + y ̸= 0
2 2
Let us consider the function f (x, y) = 2 2
. In
0, x + y = 0
Example 2, section 7.3, it was shown that a similar function is not continuous
at the origin. Using the same arguments, we can arrive to the same conclusion
regarding the above function: the path limits along y = 0 and y = x2 give
different results -
( ) x4 1
lim f (x, 0) = lim 0 = 0, while lim f x, x2 = lim + x2 = .
x→0 x→0 x→0 x→0 x4 + x4 2
At the same time, both partial derivatives exist at the origin:
f (x, 0) − f (0, 0)
fx (0, 0) = lim = 0,
x→0 x−0
f (0, y) − f (0, 0) y
fy (0, 0) = lim = lim = 1,
y→0 y−0 y→0 y
and ∇f (0, 0) = (0, 1) ̸= (0, 0). Since the function is not differentiable at the
origin, we appeal to the definition of the directional derivative: choosing an
arbitrary unitary direction vector u = (α, β), we get
f (αt, βt) − f (0, 0)

Du f (0, 0) = lim
t→0+ t−0
( ) ( )
1 α2 βt3 α2 β
= lim + βt = lim +β .
t→0+ t α 4 t4 + β 2 t2 t→0+ α 4 t2 + β 2
Therefore:
1) if α = 0, then Du f (0, 0) = lim β = ±1 ;
t→0+
2) if β = 0, then Du f (0, 0) = lim 0 = 0 ; and
t→0+
3) if α ̸= 0, β ̸= 0, then
( ( ) ) ( )
1 − β2 β 1 − β2 β 1
Du f (0, 0) = lim 2 +β = +β = .
t→0+ (1 − β 2 ) t2 + β 2 β 2 β
Thus, the directional derivative exists in any direction. However, for any
u = (α, β), α ̸= 0, β ̸= 0, the following inequality holds: |Du f (0, 0)| = |β|
1
>
|∇f (0, 0)|, that is, almost in any direction the rate of change is greater than
1
in the direction of the gradient. Moreover, since lim |β| = +∞ there is no
β→0
direction of the maximum rate of change.
Remark. This result is a consequence of discontinuity (and therefore non-

differentiability) of f (x, y) at the origin. So the theorem on the extremum
properties of the gradient is not applicable here.
Example 19. “If ∇f (x0 , y0 ) = 0 , then all the directional derivatives of

f (x, y) vanishes at (x0 , y0 ).”
Solution. √
It was shown in Examples 3 and 5 that the function f (x, y) = |xy|
possesses zero partial derivatives and all √ the directional derivatives at the
origin, which have the form Du f (0, 0) = |αβ| in the direction of an arbitrary
unitary vector u = (α, β). Therefore, these derivatives are not zero in any
direction different from those of the coordinate axes.
One can { prove that another kind of the counterexample is the function
x
xy 2
2 +y 2 , x + y
2
̸= 0
f (x, y) = 2 2 . In Examples 1 and 9 it was shown that
0, x + y = 0
∇f (0, 0) = 0 and, at the same time, there is no directional derivative, except
in the direction of the coordinate axes.
Remark. The cause of the statement incorrectness is non-differentiability
of f (x, y) at the origin.
Example 20. “If f (x, y) is differentiable on R2 and has two or more local
maxima, then it has also a local minimum.”
Solution.
Let us show that the infinitely differentiable on R2 function f (x, y) =
(1 + ey ) cos x−yey has infinitely many local maxima and none local minimum.
First, solving the system for critical points
{
fx (x, y) = − sin x (1 + ey ) = 0
,
fy (x, y) = ey (cos x − 1 − y) = 0
we obtain sin x = 0, i.e., xn = nπ , ∀n ∈ Z from the first equation, and,

substituting these values of x in the second equation, we get yn = cos nπ − 1,
i.e., y2n = 0 and y2n+1 = −2. Hence, there are the following critical points:
P2n = (2nπ, 0) and P2n+1 = ((2n + 1) π, −2), ∀n ∈ Z. To check the sufficient
conditions of local extrema at the found critical points we need the second-
order derivatives:
fxx (x, y) = − (1 + ey ) cos x, fxy (x, y) = −ey sin x,
fyy (x, y) = ey (cos x − 2 − y) .

( )
fxx fxy
If the Hessian matrix is positive/negative definite at a crit-
fxy fyy
ical point then that point is a local minimum/maximum; if the determi-
nant of the matrix is negative at a critical point then that point is not a
local extremum. For the points P2n = (2nπ, 0) the Hessian matrix assumes
( )
−2 0
the form , that is the matrix is negative definite, and conse-
0 −1
quently, all the points P2n = (2nπ, 0) are local maxima. On the other hand,
for
( the −2 ) = ((2n + 1) π, −2) the corresponding Hessian matrix
points P2n+1
1+e 0
is not definite (its determinant is negative) and there-
0 −e−2
fore, all the points P2n+1 = ((2n + 1) π, −2) are not local extrema.
( )2 ( )2
The function f (x, y) = x2 − 1 + x2 y − x − 1 gives another coun-
terexample.
2
The function f (x, y) = x4 + y 4 − (x + y) gives one more counterexample.
Remark 1. A similar example for continuous functions is given in Example
11, section 7.3.
Remark 2. Similar statement for one-variable(single-variable) functions is
true.
Example 21. “If f (x, y) is differentiable on R2 and its restriction to an

arbitrary straight line passing through the origin has a strict local minimum
at the origin, then f (x, y) has a local minimum at the origin.”
Solution.
) the infinitely differentiable on R function f (x, y) =
2
( Let2 )us
( consider
y−x y − 3x . Its restriction to the x-axis is the function f (x, 0) = 3x4 ,
2
which has the strict local minimum at the origin. The restriction to the y-axis,
f (0, y) = y 2 , also possesses the strict local minimum at the origin. Finally, the
restriction to an arbitrary line y = kx, k = const ̸= 0, results in the function
g (x) = f (x, kx) = k 2 x2 − 4kx3 + 3x4 . The critical points of the last function,
obtained as solutions √
of the equation g ′ (x) = 2k 2 x − 12kx2 + 12x3 = 0, are
3± 3
x1 = 0, x2,3 = 6 k. We are interested in only the first point, because our
analysis is for the origin. Since the second derivative g ′′ (x) = 2k 2 −24kx+36x2
is positive at x1 = 0, we conclude that x1 = 0 is the local minimum of f (x, kx).
Hence, f (x, y) satisfies the statement conditions. However, it does not have
local extrema at the origin, since in any neighborhood of the origin f (x, y)
assumes both positive and negative values:
( )
f (x, 0) = 3x4 > 0 = f (0, 0) , f x, 2x2 = −x4 < 0 = f (0, 0) .
) is the infinitely differentiable on R func-

2
Another similar
( counterexample
)(
tion f (x, y) = x − y 2
2x − y . Its restriction to the x-axis, f (x, 0) = 2x2 ,
2
has the strict local minimum at the origin. The restriction to the y-axis,
f (0, y) = y 4 , also possesses the strict local minimum at the origin. Finally, the
restriction to an arbitrary line y = kx, k = const ̸= 0, gives g (x) = f (x, kx) =
k 4 x4 − 3k 2 x3 + 2x2 . Solving the equation g ′ (x) √= 4k 4 x3 − 9k 2 x2 + 4x = 0,
we find the critical points: x1 = 0, x2,3 = 9±8k217 . The only point of our
concern is x1 = 0, which is the strict local minimum of f (x, kx), because
g ′′ (x) = 12k 4 x2 − 18k 2 x + 4 is positive there. Hence, f (x, y) satisfies the
statement conditions. However, it does not have local extrema at the origin,
since in any neighborhood of the origin f (x, y) assumes both positive and
negative values:
( )
2 2 1
f (x, 0) = 2x2 > 0 = f (0, 0) , f y , y = − y 4 < 0 = f (0, 0) .
3 9
Remark. This example is a strengthened form of the statement in Example

12, section 7.3 for continuous functions.
Example 22. “If f (x, y) is differentiable on R2 and its restriction to an

arbitrary algebraic curve y = cxm/n , where c ∈ R and m, n ∈ N (x ≥ 0 in case
the exponential fraction is in lowest terms and n is even), has a strict local
minimum at the origin, then f (x, y) has a local minimum at the origin.”
Solution.
Let us consider the infinitely differentiable on R2 function
{ ( )( )
y − e−1/x y − 3e−1/x , x ̸= 0
2 2
f (x, y) = .
y2 , x = 0
For the restriction to the x-axis we have f (x, 0) = 3e−2/x > 0 = f (0, 0),
2
that is, x = 0 is the strict local minimum. For the restriction to the y-axis,
f (0, y) = y 2 , y = 0 is the strict local minimum. Finally, the restriction to
y = cxm/n , c ̸= 0, leads to the function

( ) ( )( )
g (x) = f x, cxm/n = cxm/n − e−1/x cxm/n − 3e−1/x = x2m/n · h (x) ,
2 2
where
h (x) = c2 − 4ce−1/x · x−m/n + 3e−2/x · x−2m/n .
2 2
To evaluate the behavior of g (x), we need to calculate the auxiliary limit:
tm/n m tm/n−1 m tm/n−2

lim e−1/x · x−m/n = lim
2
2 = lim 2 = lim 2 = . . . = 0.
x→0 t→∞ et t→∞ n 2tet t→∞ n 2et
Here, we used the substitution x = 1t and then applied L’Hospital’s rule

sucessively until the power of t in the numerator becomes negative or zero.
Based on this result, we readily conclude that
( ) c2
lim h (x) = lim c2 − 4ce−1/x · x−m/n + 3e−2/x · x−2m/n = c2 > .
2 2
x→0 x→0 2
It means, that there exists a neighborhood (−δ, δ) of x = 0 such that for
2
∀x ∈ (−δ, δ) we have h (x) > c2 . Notice that x2m/n > 0 for ∀x ̸= 0. Therefore,
∪
g (x) = x2m/n · h (x) > 0 for ∀x ∈ (−δ, 0) (0, δ). Since g (0) = 0, this means
that x = 0 is the strict local minimum of g (x). Hence, f (x, y) satisfies the
statement conditions. At the same time, in any neighborhood of the origin,
f (x, y) assumes
( both
) positive and negative values: f (0, y) = y 2 > 0 = f (0, 0),
−1/x2 −2/x2
while f x, 2e = −e < 0 = f (0, 0). Therefore, it does not have
local extrema at the origin.
Remark. Example 12 in section 7.3 and the last two Examples 21 and 22
represent a chain of refined, but still insufficient, conditions for a local maxi-
mum/minimum in terms of extremum properties of the function restrictions.
Example 23. “If f (x, y) is differentiable on R2 and has only one critical
point c, which is a local maximum, then this point is a global maximum of
f (x, y).”
Solution.
Let us consider the infinitely differentiable on R2 function f (x, y) = 5xey −
x − e5y . The
5
{ only critical point is found by solving the system of necessary
fx = 5ey − 5x4 = 0
conditions: , which has the only solution x = 1, y =
fy = 5xey − 5e5y = 0
0. Therefore, the only critical point is c = (1, 0). The second-order derivatives
fxx = −20x3 , fxy = 5ey , fyy = 5xey − 25e5y give the positive determinant of
the Hessian matrix at the point c = (1, 0):
( ) ( )
fxx fxy −20 5
det (c) = det = 375 > 0,
fxy fyy 5 −20
which means that the critical point is a local extremum. Since fxx (1, 0) =
−20 < 0, one concludes that c = (1, 0) is a local maximum of f (x, y). However,
the point c = (1, 0) is not a global maximum, because f (1, 0) = 3, while
f (−2, 0) = 21.
The function f (x, y) = 3xey − x3 − e3y provides a similar counterexample.
Remark. Similar statement for one-variable functions is true.
Example 24. “If f (x, y) is differentiable on R2 , has only one local mini-
mum and is unbounded below, then f (x, y) achieves a local maximum some-
where.”
Solution.
Let us consider the infinitely differentiable on R2 function f (x, y) =
3
x2 + y 2 (1 + x) . The system of necessary conditions for critical points
{ 2
fx = 2x + 3y 2 (1 + x) = 0
3 has the only solution x = 0, y = 0. The
fy = 2y (1 + x) = 0
second-order derivatives
2 3
fxx = 2 + 6y 2 (1 + x) , fxy = 6y (1 + x) , fyy = 2 (1 + x)
give the positive determinant of the Hessian matrix at the critical point c =
(0, 0): ( ) ( )
fxx fxy 2 0
det (c) = det = 4 > 0,
fxy fyy 0 2
which means that the critical point is a local extremum. Since fxx (0, 0) = 2 >
0, it follows that c = (0, 0) is a local minimum. Besides, f (x, y) is unbounded
below since for x = −2 one has f (−2, y) = 4 − y 2 → −∞. (It is also un-
y→∞
bounded above, since for x = 0 one has f (0, y) = y 2 → +∞. ) However,
y→∞
f (x, y) has no local maximum, because c = (0, 0) is the only critical point.
The function f (x, y) = −3xey +x3 +e3y provides another counterexample.
Remark. Similar statement for one-variable functions is true.
Remark to Examples 20-24. The same statements are false for any type
of a local extremum, either minimum or maximum.
Exercises
1. Let f (x, y) be differentiable at (a, b). Show by example that the assertion
that “in this case the remainder satisfies the inequality |γ| = |∆f − df | ≤ Cρ2 ,
where C is a non-negative constant and ρ2 = ∆x2 + ∆y 2 ” is not true.
2. If a function f (x, y) has partial derivatives in a neighborhood of (a, b)
it does not guarantee that f (x, y) is differentiable or (at least) continuous at
(a, b). Provide an example to this statement. Formulate this result in the form
of a counterexample. Compare with Examples 1 and 2 in section 8.3.
3. If a function f (x, y) has bounded partial derivatives in a neighborhood
of (a, b) it does guarantee the continuity of f (x, y) in the same neighborhood,
but it is not yet sufficient for differentiability even at (a, b). Provide an example
to the last part of this statement. Formulate in the form of a counterexample.
Compare with Examples 3 and 4 in section 8.3.
4. Give an example of a function f (x, y) that has continuous mixed deriva-
tive fyx , but does not have the first derivative fx . Formulate as a counterex-
ample.
5. Give an example of a function f (x, y) that is not differentiable at a point
(a, b), but |f (x, y)| is differentiable. What about a function differentiable (a, b)
with the absolute value |f (x, y)| not differentiable? Formulate these results in
the form of counterexamples.
6. Show that the differentiability of h(x, y) = f (x, y) · g(x, y) at (a, b) does
not imply the differentiability of f (x, y) and g(x, y). Compare with Examples
3 and 4 in section 8.2.
7. Show that each of the functions f (x, y) = D(x − y), g(x, y) = D(x + y)
and h(u, v) = D(u + v), where D(t) is Dirichlet’s function, is not differentiable
anywhere, but the composed function h(f (x, y), g(x, y)) is differentiable ev-
erywhere.
8. Construct a counterexample to the following statement: “if there exists
a tangent plane to the graph of f (x, y) at (a, b), then f (x, y) is differentiable
at (a, b)”.
9. Construct a counterexample to the following statement: “if f (x, y) is
differentiable and unbounded on a set S, then at least one of the partial
derivatives fx (x, y) or fy (x, y) is unbounded on S”.
10. Construct a counterexample to the following statement: “if
f (x, y) is differentiable and both partial derivatives are unbounded on a
bounded set S, then f (x, y) is unbounded on S”. (Hint: use f (x, y) =
{ 1
(x2 + y 2 ) cos x2 +y 2
2, x + y
2
̸= 0
2 2 on S = x2 + y 2 ≤ 1 .)
0, x + y = 0
11. Give a counterexample to the statement: “if f (x, y) is continuous in a
neighborhood of (a, b) and a tangent plane does not exist at (a, b), then f (x, y)
does not have a local extremum at (a, b)”.
12. Show that the function f (x, y) = |xy| is another counterexample to

Example 15 in section 8.3.
13. Construct a differentiable at (a, b) function that does not possess partial
derivatives in any other point. Formulate as a counterexample. (Hint: f (x, y) =
(x2 + y 2 ) · D(x + y).)
14. Give an example of a function not differentiable at (a, b), but possessing
a tangent plane at (a, b).
15. Construct a function that is not differentiable
√ anywhere, √ but has a
tangent plane at some points. (Hint: f (x, y) = x2 · D(x2 + y 2 ) + x2 · (1 −
3 5
D(x2 + y 2 )).)
16. Show that f (x, y) = sin x + y 2 is a counterexample to the following
statement similar to that in Example 20, section 8.3: “if f (x, y) is differentiable
on R2 and has two or more local minima, then it has also a local maxima”.
Use the functions in Example 20 to obtain other counterexamples to the last
statement.
17. Show that f (x, y) = (x3 − y)(y − 3x3 ) is a counterexample to the
following statement similar to that in Example 21, section 8.3: “if f (x, y) is
differentiable on R2 and its restriction to an arbitrary straight line passing
through the origin has a strict local maximum at the origin, then f (x, y) has
a local maximum at the origin”. Use the functions in Example 21 to obtain
other counterexamples to the last statement.
18. Provide a counterexample to the following statement: “if d2 f (a, b) ≥ 0
at a critical point (a, b), then f (x, y) has a local minimum at (a, b)”. (Hint:
try f (x, y) = x2 + y 3 at the {
origin.)
xy xx4−2y
4 4
+y 4 , x + y ̸= 0 is another counterex-
2 2
19. Show that f (x, y) =
0, x2 + y 2 = 0
ample to Example 11, section 8.3.
20. Give a counterexample to the following statement: “if f (x, y) is con-
tinuously differentiable on R2 and fx (x, y) → 0, fy (x, y) → 0, then
(x,y)→∞
√ (x,y)→∞
f (x, y) → const ”. (Hint: consider f (x, y) = cos x + y .)
3 2 2
(x,y)→∞
21. Give a counterexample to the following statement: “if f (x, y) is contin-
uously differentiable on R2 and f (x, y) → 0, then fx (x, y) → 0 and
{
(x,y)→∞ (x,y)→∞
2 3
1
sin (x +y ) , x2 +y 2 ̸= 0
2
fy (x, y) → 0 ”. (Hint: try f (x, y) = x2 +y22 .)
(x,y)→∞ 0, x + y 2 = 0
{ 3
|x |, y ≥ 0
22. Show that f (x, y) = considered on D = {y ∈ [0, 1], x ∈
x3 , y < 0
[−y, 1]} ∪ {y ∈ [−1, 0], x ∈ [y, 1]} is another counterexample to Example 17 in
section 8.3.
23. Is the continuity of fxx and fyy is sufficient for the twice differentia-
bility of f (x, y)? If not, provide a counterexample. (Hint: consider f (x, y) =
xy ln(− ln(x2 + y 2 )) in a neighborhood of the origin.)
24. The implicit function theorem (one of its versions) states that if F (x, y)
is continuously differentiable in a neighborhood of a point (a, b) at which

F (a, b) = 0 and Fy (a, b) ̸= 0, then near a the equation F (x, y) = 0 determines
the only function y = f (x) such that b = f (a), and f (x) is continuously dif-
ferentiable with the derivative defined by the formula f ′ (x) = − F x (x,y)
Fy (x,y) . Show
that the continuity of F (x, y) in a neighborhood of (a, b) is not a necessary
condition, but its violation may invalidate the results of the above theorem.
(Hint: the equation F (x, y) = 0 considered { in a neighborhood of the origin
x + y, x ̸= y
defines the only function if F (x, y) = , but it defines two
{ −x, x = y
x + y, x ̸= y
functions if F (x, y) = .)
0, x = y
25. Show that the condition F (a, b) = 0 cannot be dropped in the formu-
lation of the implicit function theorem.
26. Show that the condition Fy (a, b) ̸= 0 in the implicit function theorem
is not a necessary condition, but its violation may invalidate the results of the
theorem. (Hint: the equation F (x, y) = 0 considered in a neighborhood of the
origin defines the only function if F (x, y) = x3 − y 3 , but it does not define
any function if F (x, y) = x − y 2 . )
Chapter 9
Integrability
Double (Riemann) Integral
1. Intergrability over a rectangle. Let f (x, y) be defined on the

rectangle R = [a, b] × [c, d] ⊂ R2 . A partition P of R into a set of elementary
rectangles
Rij = [xi−1 , xi ] × [yj−1 , yj ] , i = 1, . . . , n, j = 1, . . . , m
is made by using the partitions
Px = {xi : a = x0 < x1 < . . . < xn−1 < xn = b}
and
Py = {yj : c = y0 < y1 < . . . < ym−1 < ym = d}
of the intervals [a, b] and [c, d], respectively. The corresponding increments are
denoted ∆xi = xi −x √ i−1 and ∆yj = yj −yj−1 , the length of a diagonal of Rij is
denoted by ∆ij = ∆x2i + ∆yj2 and the partition diameter by ∆ = max ∆ij .
i,j
The (double) Riemann sum of f (x, y) corresponding to the ∑nspecific
∑m partition P
and the specific choice of points (ξi , ηj ) ∈ Rij is S (f ) = i=1 j=1 f (ξi , ηj ) ·
∆xi ∆yj . If there exists a finite limit of the Riemann sums of f (x, y), as ∆
approaches 0, which does not depend on a choice of partitions and points
(ξi , ηj ), then this limit is called the double
∫∫ (Riemann) integral∫∫ of f (x, y) over
R, is denoted as follows: lim S (f ) = R f dA or lim S (f ) = R f dxdy, and
∆→0 ∆→0
the function f (x, y) is called (Riemann) integrable over R.
2. Integrability over a general set. Let f (x, y) be defined on a bounded
287
closed set D ⊂ R2 whose boundary ∂D has zero area. (The last means that
for any ε > 0 the boundary ∂D can be covered by a set of rectangles with the
total area less than ε. Those familiar with the measure theory will immediately
recognize the definition of the plane Lebesgue measure zero. In particular,
any rectifiable curve has zero area.) Consider an arbitrary rectangle R =
[a, b] × [c, d]
{ containing the set D and define the following function on R:
f (x, y) , (x, y) ∈ D
F (x, y) = . If F (x, y) is integrable over R, then f (x, y)
0, otherwise ∫∫ ∫∫
is integrable over D and D f dA = R F dA.
Remark. Hereinafter in this chapter we will consider integrability over sets
D that represent the closures of bounded connected open sets with zero area
boundary.
Theorem. If f (x, y) is integrable over D, then it is bounded on D.

Theorem. If f (x, y) is continuous on D, then it is integrable over D.
Theorem. If f (x, y) is bounded on D and is continuous on D except at
a finite number of smooth curves of zero area, then f (x, y) is integrable over
D.
Theorem. If f (x, y) is bounded on D and is continuous on D except at
a countable set of curves of zero area, then f (x, y) is integrable over D.
Lebesgue’s Criterion. f (x, y) is integrable over D if, and only if, f (x, y)
is bounded on D and the set of its discontinuity points has plane (Lebesgue)
measure zero.
1) If f (x, y) and g (x, y) are integrable over D and f (x, y) ≤ g (x, y) on D,
then ∫∫ ∫∫
f dA ≤ gdA.
D D
2) If f (x, y) is integrable on D, then

∫∫
finf · A (D) ≤ f dA ≤ fsup · A (D) ,
D
where finf = inf f (x, y), fsup = sup f (x, y) and A (D) is the area of
(x,y)∈D (x,y)∈D
D. In particular, if f (x, y) is continuous on D, then
∫∫
fmin · A (D) ≤ f dA ≤ fmax · A (D) ,
D
where fmin = min f (x, y), fmax = max f (x, y).

(x,y)∈D (x,y)∈D

1) If f (x, y) and g (x, y) are integrable over D, then f (x, y) + g (x, y) is also
Integrability 289
integrable over D and

∫∫ ∫∫ ∫∫
(f + g) dA = f dA + gdA.
D D D
2) If f (x, y) is integrable over D and c is an arbitrary constant, then cf (x, y)

is also integrable over D and
∫∫ ∫∫
cf dA = c f dA.
D D
3) If f (x, y) is integrable over D, then |f (x, y)| is also integrable over D and
∫ ∫ ∫∫

f dA ≤ |f | dA.

D D
4) If f (x, y) and g (x, y) are integrable over D, then f (x, y) g (x, y) is also
integrable over D.
5) f (x, y) is integrable over D, if and only if, f (x, y) is integrable over D1
and D2 , where D1 ∪ D2 = D and D1 ∩ D2 is a zero area curve. Furthermore,
∫∫ ∫∫ ∫∫
f dA + f dA = f dA.
D1 D2 D
The Mean Value Theorem for double integral. If f (x, y) is in-

tegrable
∫∫ on D, then there exists a constant γ, finf ≤ γ ≤ fsup , such that
D
f dA = γ · A (D). In particular, if f (x, y) is continuous
∫∫ on a connected set
D, then there exists a point (a, b) ∈ D such that D f dA = f (a, b) · A (D).
Iterated Integrals
1. Rectangles. Let f (x, y) be defined on the rectangle R = [a, b]×[c, d] ⊂

∫d
R2 and for each x ∈ [a, b] there exists the Riemann integral c f (x, y) dy =
∫b
g (x). If the Riemann integral a g (x) dx also exists, it is called the iterated
∫b ∫b ∫d
integral of f (x, y) and denoted as follows: a g (x) dx ≡ a dx c f (x, y) dy. In
∫d ∫b
the same way can be defined the second iterated integral c dy a f (x, y) dx.
2. General domains. Let f (x, y) be defined on a bounded closed set
D ⊂ R2 whose boundary ∂D has zero area. Additionally, suppose that each
line parallel to y-axis crosses ∂D at two points at most, let say y1 (x) and
y2 (x), y1 (x) ≤ y2 (x), or along the entire segment [y1 (x), y2 (x)] ⊂ ∂D. In this
case the set D can be represented as D = [a, b] × [y1 (x) , y2 (x)] and it is fre-
quently called the type I region. If for each x ∈ [a, b] there exists the Riemann
∫ y (x) ∫b
integral y12(x) f (x, y) dy = g (x) and the Riemann integral a g (x) dx also ex-
ists, the latter is called the iterated integral of f (x, y) and denoted as follows
∫ b ∫ y2 (x)
a
dx y1 (x) f (x, y) dy. In a similar way can be defined the second iterated in-
∫ d ∫ x (y)
tegral c dy x12(y) f (x, y) dx on the type II region D = [x1 (y) , x2 (y)] × [c, d].
Fubini’s Theorem. Let f (x, y) be defined on a type I region D. If f (x, y)

is Riemann integrable over D and for each x ∈ [a, b] there exists the Riemann
∫ y (x) ∫ b ∫ y (x)
integral y12(x) f (x, y) dy, then the iterated integral a dx y12(x) f (x, y) dy ex-
ists and is equal to the double integral:
∫∫ ∫ b ∫ y2 (x)
f (x, y) dA = dx f (x, y) dy.
D a y1 (x)
A similar result is true for type II regions.

In the special case when f (x, y) is continuous on D, the existence of double
and both iterated integrals is guaranteed and the three integrals give the same
value.
Change of variables. Let f (x, y) be defined on a bounded closed

connected set D. Consider two functions x = φ (u, v), y = ψ (u, v) de-
fined on (u, v) ∈ D̃ and such that this transformation (the change of vari-
ables) is one-to-one correspondence between D̃ and D. If both functions φ
and ψ have continuous partial derivatives on D̃ and the Jacobian ∂(φ,ψ)∂(u,v) ≡
( )
φu φv
det ̸= 0 on D̃, then for f (x, y) integrable over D the following
ψu ψv
formula holds:
∫∫ ∫∫
∂ (φ, ψ)
f (x, y) dxdy =
f (u, v) dudv.
D D̃ ∂ (u, v)
Line Integrals
Line integral. Let C be a plane rectifiable simple curve given in the

parametric form x = φ (t), y = ψ (t) , t ∈ [a, b], where functions φ(t), ψ(t)
are continuous on [a, b], and let f (x, y) be defined on C. (Recall that a curve
is called simple if it does not have self-intersections, except probably for the
initial and final points.) Consider an arbitrary partition
P = {ti : a = t0 < t1 < . . . < tn−1 < tn = b}
of [a, b] with the diameter ∆ = max ∆ti , ∆ti = ti − ti−1 . A partition P

i
corresponds to a certain partition of C by the points (xi , yi ) = (φ (ti ) , ψ (ti )),
i = 1, . . . , n, and an increment ∆ti is associated with the following increments
Integrability 291
of coordinates and length of the line segment joining two successive points
(xi−1 , yi−1 ) and (xi , yi ):
√
2 2
∆xi = xi − xi−1 , ∆yi = yi − yi−1 , ∆li = (xi − xi−1 ) + (yi − yi−1 ) .
Choose an arbitrary point τi ∈ [ti−1 , ti ] in each partition interval, which cor-

responds to the point (x (τi ) , y (τi )) on the arc of C between (xi−1 , yi−1 ) and
(xi , yi ), and form the three integral sums:
∑
n
Sl (f ) = f (x (τi ) , y (τi )) · ∆li ,
i=1
∑
n ∑
n
Sx (f ) = f (x (τi ) , y (τi )) · ∆xi , Sy (f ) = f (x (τi ) , y (τi )) · ∆yi .
i=1 i=1
If finite limits of these sums, as the partition diameter ∆ approaches 0, exist

and do not depend on a choice of partitions and points τi , then they are called
the line integral of f (x, y) along C and are denoted as follows:
∫
lim Sl (f ) = f (x, y) dl,
∆→0 C
∫ ∫
lim Sx (f ) = f (x, y) dx, lim Sy (f ) = f (x, y) dy.
∆→0 C ∆→0 C
In order to distinguish one line integral from another, the first one is called
the line integral with respect to arc length, the second - with respect to x, and
the third - with respect to y. Frequently, the first integral is called the line
integral of the first kind, and two others - the line integrals of the second kind.
The sum of two second kind integrals
∫ ∫ ∫
f (x, y) dx + g (x, y) dy ≡ f (x, y) dx + g (x, y) dy
C C C
is called a general line integral of the second kind.

Remark 1. The value of the line integral of the first kind does not depend
on the parametrization of C.
Remark 2. The absolute value of the line integrals of the second kind does
not depend on the parametrization of C.
Remark 3. If C is a simple closed curve representing the boundary of a
bounded set D, then the positive orientation of C is such that D is always on
the left under the traversal of C.
Theorem. If C is a smooth curve without singular points (that is, in the

curve definition x = φ (t), y = ψ (t) are continuously differentiable on [a, b],
and (φ′ (t)) + (ψ ′ (t)) ̸= 0, ∀t ∈ [a, b]), and if f (x, y) is continuous on C,
2 2
then the line integrals of f (x, y) along C exist and can be calculated by the
following formulas:
∫ ∫ b √
2 2
f (x, y) dl = f (φ (t) , ψ (t)) (φ′ (t)) + (ψ ′ (t)) dt,
C a
∫ ∫ b
f (x, y) dx = f (φ (t) , ψ (t)) φ′ (t) dt,
C a
∫ ∫ b
f (x, y) dy = f (φ (t) , ψ (t)) ψ ′ (t) dt,
C a
where the integrals in the right-hand sides are the Riemann integrals.
Remark. The theorem is true also for a piecewise smooth curve, that is,
for a continuous curve represented as a union of a finite number of smooth
curves.
The line integrals have the same comparative and arithmetic properties as
the double (and “ordinary”) Riemann integral.
∫
Path Independence.
∫ The
∫ line integral C f dx + gdy is path independent
in a set D if C1 f dx + gdy = C2 f dx + gdy for any two smooth curves (paths)
in D that have the same initial and final points.
Theorem. If f (x, y) and g (x, y) are continuous and have continuous
partial derivatives fy and gx on an open simply-connected set D, then the
following four conditions
∫ are equivalent:
1) the line integral ∫ C f dx + gdy is path independent in a set D
2) the line integral C f dx + gdy = 0 for every closed path in D (that is, for
every path whose initial and final points coincide)
3) the expression f dx + gdy represents the differential of a function η (x, y) on
D
4) fy = gx on D.
Remark 1. The conditions 1) and 2) are equivalent for any line integrable
(for example, continuous) on D functions f and g.
Remark 2. The conditions 1) and 3) are equivalent for any continuous
functions f (x, y) and g (x, y) on an open connected set D.
Remark 3. The condition 4) follows from 1) (or 2), or 3)) for any open
connected set D.
Remark 4. The condition 3) can be rewritten in the following form: there
exists a function η (x, y) such that ∇η = (f, g) in D. In this case the vector
field F = (f, g) is called conservative (or potential) on D and the function η
is called a potential function of F .
Green’s formulas. If f (x, y) and g (x, y) have continuous partial deriva-

tives on a closed bounded set D with a smooth boundary ∂D, then the fol-
Integrability 293
lowing formulas hold:

∫∫ ∫ ∫∫ ∫
fy dA = − f dx, gx dA = gdy,
D ∂D D ∂D
∫∫ ∫ ∫∫ ∫
(gx − fy ) dA = f dx + gdy, (fx + gy ) dA = −gdx + f dy,
D ∂D D ∂D
where all the line integrals have the positive orientation, that is, a single
positive traversal of ∂D.
Integration by parts. If f (x) and g (x) have continuous partial deriva-

tives on a closed bounded set D with a smooth boundary ∂D, then the fol-
lowing formulas hold:
∫∫ ∫∫ ∫
fy gdA = − f gy dA − f gdx,
D D ∂D
∫∫ ∫∫ ∫
fx gdA = − f gx dA + f gdy,
D D ∂D
where all the line integrals have the positive orientation.
General Remark. Like in sections 7.2 and 8.2, the examples in this section
have direct connections with some examples for one-variable functions given in
Part I and, correspondingly, the majority of the provided below counterexam-
ples are adapted to R2 versions of the functions introduced in chapter 5. Each
example in this section has the reference to the corresponding one-dimensional
counterpart.
Example 1. “If f (x, y) is bounded on D, then there exists the double

integral of f (x, y) over D.”
Solution.
{ √
The two-dimensional version of Dirichlet’s function
1 , √x2 + y 2 ∈ Q
D (x, y) = is bounded but not integrable over any
0 , x2 + y 2 ∈ I
bounded closed set D. Let us consider for simplicity a rectangle R =
[a, b] × [c, d]. For an arbitrary partition of R, each elementary rectangle
Rij = [xi−1 , xi ] × [yj−1 , yj ] contains the points with rational and irrational
distance from the origin. Choosing rational distance points in the definition
of Riemann sums, one obtains Srat (f ) = (b − a) (d − c), while for irrational

distance points one gets Sirr (f ) = 0. Therefore, the limits for these sums are
different:
lim Srat (f ) = (b − a) (d − c) ̸= 0 = lim Sirr (f ) ,
∆→0 ∆→0
which means that the double integral does not exist.

Remark. The graph of this function is similar to that of the two-
dimensional modified Dirichlet’s function shown in Example 1, section 7.2.
Example 2. “If f (x, y) is integrable over D, then f (x, y) is continuous

on D.”
Solution.
The function f (x, y) = sgn x is discontinuous on D = [0, 1] × [0, 1], since
lim f (x, y) = lim sgn x = 1 ̸= 0 = f (0, y0 ) .

(x,y0 )→(0,y0 ) x→0
On the other hand, for an arbitrary choice of partition and points (ξi , ηj ) ∈
Rij , the following evaluation of the Riemann sums for f (x, y) is true:
∑
n ∑
m
1 − ∆x1 = 1 · ∆xi ∆yj ≤ S (f )
i=2 j=1
∑
n ∑
m ∑
n ∑
m
= f (ξi , ηj ) · ∆xi ∆yj ≤ 1 · ∆xi ∆yj = 1.
i=1 j=1 i=1 j=1
Since ∆x1 → 0 when ∆ → 0, it follows that the limit of the Riemann sums,
as ∆ approaches 0, exists and ∫∫ does not depend on the choice of partition and
points (ξi , ηj ): lim S (f ) = D sgn xdA = 1, that is, f (x, y) is integrable over
∆→0
D.
Example 3. “If |f (x, y)| is integrable over D, then f (x, y) is also inte-
grable over D.”
Solution.
{ √
The two-dimensional modified Dirichlet’s function
1, √ x2 + y 2 ∈ Q
D̃ (x, y) = is not integrable over any set D, in partic-
−1 , x2 + y 2 ∈ I
ular, it is not integrable over a rectangle
R = [a, b] × [c, d] (it can be shown

like in Example 1), but D̃ (x, y) = 1 is integrable over any set D.
Remark 1. The Remarks to Example 5 of section 5.3 are applicable here
Remark 2. The graph of this function is shown in Example 1, section 7.2.
Integrability 295
∫∫
Example 4. “If f (x, y) is integrable over D and D f 2 dA = 0, then
f (x, y) = 0 on D.”
Solution.
According to the arithmetic rules for integrals, the function f (x, y) =
1 − sgn x is integrable on D = [0, 1] × [0, 1] as a difference of two integrable
functions (see Example 2 for integrability of sgn x) and
∫∫ ∫∫ ∫∫ ∫∫
f 2 dA = 1 − sgn xdA = 1dA − sgn x dA = 1 − 1 = 0.
D D D D
However, f (x, y) ̸= 0 on D since f (0, y) = 1.

Remark. The Remarks to Example 9 of section 5.3 are applicable here
Example 5. “If f (x, y) + g (x, y) is integrable over D, then both f (x, y)

and g (x, y) are integrable over D.”
Solution.
f (x, y) = D (x, y) and g (x, y) = 1 − D (x, y) is a simple counterexample
for an arbitrary set D.
Remark. Similar statements for other arithmetic operations are also false.
Example 6. “If double integral and both iterated integrals of f (x, y) exist
and are equal, then these integrals can be calculated using antiderivtives of
f (x, y) with respect to x (when y is fixed) or with respect to y (when x is
fixed).”
Solution.
The function f (x, y) = sgn(xy) is integrable on [0, 1]×[0, 1] and its iterative
integrals also exist (to show this one can use the arguments similar to those
∫1 ∫1
applied in Example 2). However, the first iterated integral 0 dx 0 sgn(xy)dy
∫1
requires first the evaluation of 0 sgn(x0 y)dy for every fixed x0 ∈ [0, 1], that is
∫1
possible to make obtaining the result 0 sgn(x0 y)dy = sgn x0 , but not through
antiderivatives because there is no antiderivative for sgn(x0 y) for any x0 ̸= 0.
∫1
Additionally, for the same reason, the subsequent calculation of 0 sgn xdx is
also not possible through antiderivative. Due to the function symmetry, the
same considerations are true for the second iterative integral.
Example 7. “If ∫∫
f (x, y) is integrable over D, then there exists a point
(a, b) ∈ D such that D f dA = f (a, b) · A (D).”
Solution. {
0, x < 0
The Heaviside function H (x, y) = is integrable over D =
1, x ≥ 0
[−1, 1] × [−1, 1] , since it is a bounded function on D and it is continuous on

D with the only exception being the simple line x = 0, y ∈ [−1, 1]. Due to the
arithmetic rules for double integrals
∫∫ ∫∫ ∫∫
HdA = 0dA + 1dA = 0 + 2 = 2,
D D1 D2
where D1 = [−1, 0] × [−1, 1] and D2 = [0, 1]∫∫

× [−1, 1]. At the same time , there
HdA
is no point (a, b) ∈ D such that H (a, b) = A(D)
D
= 12 .
Remark. The additional condition of the continuity of f (x, y) on D will
turn this false statement in the Mean Value Theorem for double integrals.
Example 8. “If f (x, y) is integrable (or even continuous) over D, then

the value of the integral does not change if one alters the values of f (x, y) at
a countable number of points.”
Solution.
The functions f (x, y) ≡ 0 and
{
1 , x ∈ Q, y ∈ Q
D̂ (x, y) = D(x) · D(y) =
0 , otherwise
(here D(x) and D(y) are one-dimensional Dirichlet’s functions) differ only at
Integrability 297
points with rational coordinates, that is on a countable set, however f (x, y)

is Riemann integrable and D̂ (x, y) is not integrable over any set D (it can be
shown in the same way as in Example 1 of this section).
Remark. The statement becomes true if both functions are integrable.
Example 9. “If f (x, y) is continuous on D except at only one point, then

f (x, y) is integrable over D.”
Solution. ( )
The function f (x, y) = cot x2 + y 2 has the only discontinuity point (0, 0)
on D = [0, 1]×[0, 1], but f (x, y) is not integrable over D because it approaches
infinity as (x, y) approaches (0, 0).
Remark. The statement becomes true if f (x, y) is additionally bounded
on D.
Example 10. “If f (x, y) is continuous on D, then the volume of the

figure between the graph of f (x, y) and the base D in xy-plane ∫∫ (and the
corresponding lateral walls) can be found by the formula V = D f dA.”
Solution.
A simple counterexample is f (x, y) = x + y on [−1, 1] × [−1, 1]. In this case
the volume of the figure is the sum of the volumes of two equal tetrahedrons:
4 4
V = V1 + V2 = 3 + 3 = 83 , but the integral formula gives
∫∫ ∫ ∫ ∫ ( )1 ∫ 1
1 1 1
y 2
f dA = dx (x + y) dy = xy + dx = 2xdx = 0.
D −1 −1 −1 2 −1 −1
Remark. The condition of the non-negativity of f (x, y) is missing in the

statement
∫∫ conditions, or without the non-negativity the correct formula is
V = D |f | dA.
Example 1. “Denote rij = ∆xi · ∆yj the area of an elementary rectangle

Rij in partition of a rectangle R and r = max rij . Then the definition of
i,j ∫∫
the double integral over a rectangle R can be given as follows: R f dA =
lim S (f ), where the limit does not depend on a choice of partitions and points
r→0
(ξi , ηj ).”
Solution.
Integrability 299
Let us consider the function f (x, y) = x on the rectangle R = [0, 1] × [0, 1]

and choose the partition into a set of elementary rectangles
R1j = [0, 1] × [yj−1 , yj ] , j = 1, . . . , m,
that is, keep the entire interval in x-axis. Then lim S (f ) depends on the choice
r→0
of points (ξi , ηj ): if one use the points (0, ηj ), then f (0, ηj ) = 0 and
∑
1 ∑
m
S (f ) = f (0, ηj ) · ∆xi ∆yj = 0,
i=1 j=1
while for (1, ηj ) one has f (1, ηj ) = 1 and
∑
1 ∑
m
S (f ) = f (1, ηj ) · ∆xi ∆yj = 1.
i=1 j=1
Therefore, the double integral of f (x, y) = x does not exist according to

the proposed “definition” with the parameter r. However, the existence of the
double integral can be immediately confirmed using the properties of integrals
(f (x, y) = x is continuous on R = [0, 1] × [0, 1]) or directly by the definition.
In the last case, one can employ the following evaluations:
∑
n ∑
m ∑
n ∑
m
S (f ) = f (ξi , ηj ) · ∆xi ∆yj ≤ xi · ∆xi ∆yj
i=1 j=1 i=1 j=1
∑
m ∑
n ∑
n
= ∆yj · xi · ∆xi = xi · (xi − xi−1 )
j=1 i=1 i=1
1∑ 1∑
n n
= (xi + xi−1 )(xi − xi−1 ) + (xi − xi−1 )(xi − xi−1 )
2 i=1 2 i=1
1∑ 2 1∑ ∑
n n n
1 1
= (xi − x2i−1 ) + ∆x2i ≤ (x2n − x20 ) + max ∆xi · ∆xi
2 i=1 2 i=1 2 2 i i=1
1 2 1 1 1
≤ (x − x20 ) + ∆ = + ∆
2 n 2 2 2
and
∑
n ∑
m ∑
n ∑
m
S (f ) = f (ξi , ηj ) · ∆xi ∆yj ≥ xi−1 · ∆xi ∆yj
i=1 j=1 i=1 j=1
∑
m ∑
n ∑
n
= ∆yj · xi−1 · ∆xi = xi−1 · (xi − xi−1 )
j=1 i=1 i=1
1∑ 1∑
n n
= (xi + xi−1 )(xi − xi−1 ) − (xi − xi−1 )(xi − xi−1 )
2 i=1 2 i=1
1 2 1 ∑n
1 1
≥ (xn − x20 ) − max ∆xi · ∆xi ≥ − ∆.
2 2 i i=1
2 2
Thus, for every partition and an arbitrary choice of the points (ξi , ηj ) it holds
1 1 1 1
− ∆ ≤ S(f ) ≤ + ∆
2 2 2 2
and, consequently, ∫∫
1
lim S (f ) = f dA =
.
∆→0 R 2
Remark. The problem with the use of the maximum area r of elementary
rectangles as a parameter for the limit of the Riemann sums is that approach-
ing r to 0 does not necessarily require a refinement of partitions in both
directions. For instance, in the above counterexample, when r approaches 0
for the chosen set R1j = [0, 1] × [yj−1 , yj ], j = 1, . . . , m, the refinement oc-
curs only in y-direction, while in x-direction always is used the entire interval
[0, 1]. It should be noted that the same situation can not occur with the par-
tition diameter ∆. Indeed, for the partitions provided in the counterexample
the diameter cannot approach 0, because ∆ > 1 for each of the above par-
titions (∆ approaches 1 when r approaches √ 0). In a general case, according
to its definition ∆ = max ∆ij , ∆ij = ∆x2i + ∆yj2 , the condition that ∆
i,j
approaches 0 means that the maximum diagonal of the elementary rectangles
should approach 0, which in its turn implies that the maximum length of each
side of elementary rectangles should approach 0, that is, the refinement is
compulsorily in both directions.
Example 2. “If f (x, y) is discontinuous on a countable set of curves in a

set D, then f (x, y) is not integrable on D.”
Solution.
Let D = R = [0, 1] × [0, 1] and
1
E = ∪k=1
+∞
Ek , Ek = {(x, y) : x = , y ∈ [0, 1]}, ∀k ∈ N.
2k
{
1, (x, y) ∈ E
Consider the following function on D: f (x, y) = that is
0, (x, y) ∈ D\E
discontinuous at every point of E. To show that f (x, y) is integrable over D
let us estimate the integral Riemann sums. Since f (x, y) ≥ 0 on D, an integral
sum is non-negative for an arbitrary partition P of R into a set of elementary
rectangles
Rij = [xi−1 , xi ] × [yj−1 , yj ] , i = 1, . . . , n, j = 1, . . . , m
and for an arbitrary choice of the points (ξi , ηj ) ∈ Rij :
∑
n ∑
m
S (f ) = f (ξi , ηj ) · ∆xi ∆yj ≥ 0.
i=1 j=1
Integrability 301
On the other hand, choosing the points (ξi , ηj ) to be in E each time when
Rij contains at least one point of E, we get upper estimation for any integral
sum:
∑
n ∑m
S (f ) = f (ξi , ηj ) · ∆xi ∆yj
i=1 j=1
∑ ∑
≤ S1 + S2 = 0 · ∆xi ∆yj + 1 · ∆xi ∆yj ,
i,j i,j
where the first sum corresponds to all Rij that have no common point with E
and the second - to the rectangles containing points of E. Notice that S1 = 0
and each term in S2 equals to the area of the corresponding rectangle Rij .
Therefore, we should now evaluate the area of all the rectangles containing
the points of E.
Choose an arbitrary ϵ > 0 and find a corresponding K such that 21K <
4 ≤ 2K−1 (such K exists, because 2k → 0). Then 2k < 4 for ∀k ≥ K, which
ϵ 1 1 1 ϵ
k→∞
means that all segments Ek , k ≥ K lie in the strip 0 < x < 4ϵ , 0 ≤ y ≤ 1 and
only K −1 segments Ek , k = 1, . . . , K −1 lie out of this strip. Consider now an
ϵ
arbitrary partition P with the diameter ∆ < 4K and choose the smallest group
of the leftmost rectangles Rij , i = 1, . . . , I, which contain all the segments
Ek , k ≥ K. Evidently, all these rectangles, except probably for those in the
rightmost column, belong to the the strip [0, 4ϵ ] × [0, 1]. However, the area
of the last column is easily estimated in terms of ϵ: 1 · ∆xI < 1 · ∆ < 4K ϵ
.
Therefore, the overall area of these rectangles can be evaluated as follows:
ϵ ϵ ϵ
A1 < + ≤ .
4K 4 2
It remains to estimate the area of those Rij , which contain all the points of
Ek , k = 1, . . . , K−1. Notice that each of these Ek lie inside a strip [xik −1 , xik ]×
[0, 1] of the width ∆xik = xik − xik −1 or belongs to a border x = xik , y ∈ [0, 1]
between two subsequent strips. Therefore, the overall area of these rectangles
satisfies the following evaluation:
∑
K−1
ϵ
A2 ≤ (∆xik + ∆xik +1 ) < 2∆ · (K − 1) < .
2
k=1
ϵ
Summarizing, for an arbitrary partition with the diameter ∆ < 4K , the
overall area of the elementary rectangles Rij , which contain all the segments
Ek , ∀k ∈ N, has the following evaluation:
ϵ ϵ
A = A1 + A2 < + = ϵ.
2 2
Since S2 ≤ A < ϵ, it follows from the above estimates of the integral sums
that 0 ≤ S(f ) ≤ S2 < ϵ. The last inequality implies that lim S(f ) = 0, that
∫∫ ∆→0
is, f (x, y) is integrable over R and R f dA = 0.
Remark 1. Since the constructed set of the curves

1
Ek = {(x, y) : x = , y ∈ [0, 1]}, ∀k ∈ N
2k
has the plane measure zero and the function f (x, y) is bounded on D, the
existence of the double integral follows from Lebesgue’s Criterion of integra-
bility. However, we think that an explicit evaluation of the Riemann sums is
instructive to prove integrability in this case.
Remark 2. The above function f (x, y) differs from 0 on a countable set
of segments, which is a larger set then a countable set of points, like that
used in Example 8, section 9.2. Nevertheless, the function in Example 8 is
not integrable while the above function gives the double integral equal to
0. Therefore, one can conclude that not only the “quantity” of discontinuity
points is important for integrability, but also their distribution. In Example 8
the discontinuities were introduced in such a way that every point in D became
a discontinuity point, while in the last Example the set of discontinuity points
is not dense in D.
Example 3. “If f (x, y) is discontinuous at each point of a bounded set

S, then both iterated integrals over S do not exist.”
Solution.
Consider the function f (x, y) defined on S = [0, 1] × [0, 1] as follows:
Integrability 303

 0, x ∈ Q
f (x, y) = y − 21 , x ∈ I, y ̸= 12 . This function is discontinuous at every

−1 , x ∈ I, y = 21
point of S: at every point (x0 , y0 ) , x0 ∈ Q, y0 ̸= 12 the function has two
different partial limits
1
lim f (x, y) = 0, lim f (x, y) = y0 − ̸= 0;
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 ) 2
x=x0 x∈I, y=y0
at the points (x0 , y0 ) , x0 ∈ Q, y0 = 1

2 two different partial limits are
lim f (x, y) = 0, lim f (x, y) = −1 ̸= 0;

(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
x=x0 x∈I, y=y0
for each (x0 , y0 ) , x0 ∈ I, y0 ̸= 1

2 two different partial limits are
1
lim f (x, y) = 0, lim f (x, y) = y0 − ̸= 0;
(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 ) 2
x∈Q, y=y0 x=x0
finally, for (x0 , y0 ) , x0 ∈ I, y0 = 1

2 the different partial limits are
lim f (x, y) = 0, lim f (x, y) = −1 ̸= 0;

(x,y)→(x0 ,y0 ) (x,y)→(x0 ,y0 )
x∈Q, y=1/2 x∈I, y=1/2
∫1 ∫1
Nevertheless, the iterated integral 0 dx 0 f (x, y) dy exists. In fact, for
∫1 ∫1
any fixed xrac ∈ Q one has 0 f (xrac , y) dy = 0 0dy = 0. Further, for
any fixed xirr ∈ I the function f (xirr , y) is continuous with respect to y
on [0, 1] except at the only point y = 21 where f (xirr , y) has a remov-
able discontinuity, that is, f (xirr , y) is Riemann integrable on [0, 1] and
∫1 ∫1 ∫1
0
f (x, y) dy = 0 y − 21 dy = 0. Therefore, g (x) = 0 f (x, y) dy = 0 for
every fixed x ∈ [0, 1]. Consequently, the iterated integral is defined and
∫1 ∫1 ∫1
dx 0 f (x, y) dy = 0 g (x) dx = 0.
0 ∫1 ∫1
Remark 1. Notice that the second iterated integral 0 dy 0 f (x, y) dx
∫1
does not exist, since 0 f (x, y) dx does not exist: for any fixed y, the function
f (x, y) represents a Dirichlet’s like function, which is not Riemann integrable.
Remark 2. Evidently, f (x, y) is not integrable in the sense of double
integral, because the number of its discontinuities is not of the measure 0.
Example 4. “If a region D is divided into regions D1 and D2 by a curve

of an infinite length, then the arithmetic property 5) in section 9.1 does not
hold.”
Solution.
Consider the rectangle D = [−1, 1] × [−2, 2] divided into two regions by
{
g( x), x ∈ [0, 1],
the saw-like curve defined as follows: h(x) = is the
−g(−x), x ∈ [−1, 0]
function composed of the left “negative” saw { (the second sentence) and the
gn , x ∈ [ 21n , 2n−1
1
], ∀n ∈ N
right “positive” saw (the first sentence) g(x) = ,
0, otherwise
containing an infinite number of teeth in the form
{ n+1
2 x − 2, x ∈ [ 21n , 2n+1
3
]
gn (x) = , ∀n ∈ N.
−2n+1 x + 4, x ∈ [ 2n+1 , 2n−1
3 1
]
Notice, that each point of the segment [−1, 1] on y-axis is the limit point of
the curve h(x).
Consider first the “positive” saw g(x) located in the rectangle [0, 1]×[−2, 2].
Since each tooth gn (x) has the length greater than 2 and the number of these
teeth is infinite, the curve g(x) has an infinite length. Since the “negative”
saw −g(−x) is symmetric to g(x) with respect to the origin, its length is also
infinite. Nevertheless, the plane (Lebesgue) measure of the curve h(x) is zero.
In fact, one can show that this curve can be covered by a finite set of rectangles
whose overall area is less than a given positive number. To this end, consider
an arbitrary ϵ > 0 and choose the number N such that 21N < 24 ϵ
≤ 2N1−1
(such number always exists, because 2N → 0). Construct now the rectangle
1
N →∞
Rϵ = [− 24 ϵ ϵ
, 24 ] × [− 32 , 23 ] of the area Aϵ = 24
2ϵ
· 3 = 4ϵ , which contains all
the teeth of the saw h(x) with n ≥ N (both of the left negative and right
positive parts) and also all the points of the segment [−1, 1] on y-axis. The
remaining N − 1 teeth of the “positive” saw g(x) represent a rectifiable curve
and, consequently, have the plain measure zero, that is, this part of saw can
be covered by a finite set of rectangles with the overall area less than 4ϵ . The
same is true for the remaining N − 1 teeth of the “negative” saw −g(−x).
Therefore, all the saw h(x) can be covered by a finite number of rectangles
with the overall area A < 4ϵ + 4ϵ + 4ϵ < ϵ, which means that the curve h(x)
has zero area. Since the applied splitting of D is such that D1 ∪ D2 = D
and D1 ∩ D2 is the zero area curve h(x), all the conditions of the arithmetic
property 5) in section 9.1 are satisfied and for any integrable on D function
f (x, y) it holds ∫∫ ∫∫ ∫∫
f dA + f dA = f dA.
D1 D2 D
For instance, if f (x, y) ≡ 1, then one obtains the area of D at the right-hand
side of the last formula
∫∫ ∫ 1 ∫ 2
f dA = dx dy = 8
D −1 −2
and the areas of D1 and D2 at the left-hand side

∫∫ ∫∫
f dA + f dA = 4 + 4 = 8
D1 D2
Integrability 305
(due to the symmetry of D1 and D2 and the symmetry of f (x, y) each area is
just a half of the area of D).
Example 5. “If both iterated integrals exist, then they are equal.”
Solution. {
x−y
(x+y)3
, (x, y) ̸= (0, 0)
Let us consider the function f (x, y) = defined
0, (x, y) = (0, 0)
on the unit square S = [0, 1] × [0, 1]. For any x ̸= 0 we have
∫ ∫ ( )1
1
x−y 1
−x − y
2x 1 x
I1 (x) = 3 dy = + 3 dy = −
0 (x + y) 0 (x + y)
3
(x + y) x + y (x + y)2
0
( )
1 1 1 1 1
= − −x 2 − = 2.
x+1 x (x + 1) x2 (x + 1)
Since the last function is continuous (and consequently integrable) on [0, 1]
and the value of the definite integral does
∫ 1not ∫depend on the function definition
1
in a single point, the iterated integral 0 dx 0 f (x, y) dy exists and is equal
∫ ∫ ∫ 1
1 1 1
1 1 1
dx f (x, y) dy = 2 dx = − x + 1 0
= .
2
0 0 0 (x + 1)
Due to the antisymmetry of the function with respect to the ∫interchange of

1 x−y
the variables, for the second iterated integral we obtain I2 (y) = 0 (x+y)3 dx =
∫ 1 ∫ 1
− (y+1)2 , ∀y ̸= 0 and 0 dy 0 f (x, y) dx = − 2 . Therefore, both iterated in-
1 1
tegrals exist, but they are different. Note that this result is caused by un-
boundedness (and therefore discontinuity) of the function at the origin. This
violates the conditions of Fubini’s theorem and may lead to inequality between
the iterated integrals as this counterexample shows.
 Remark 1. Another interesting counterexample is the function f (x, y) =
 y −2 , 0 < x < y < 1
−x−2 , 0 < y < x < 1 considered on the unit square S = [0, 1] × [0, 1].

0, otherwise
Remark 2. For the above functions, the double integral does not exist.
∫b ∫d ∫d ∫b
Example 6. “If a dx c f (x, y) dy = c dy a f (x, y) dx, then f (x, y) is
continuous on [a, b] × [c, d].”
Solution.
The function
 −1
 y , (x, y) ∈ D1 : 0 < x < y < 1
f (x, y) = x−1 , (x, y) ∈ D2 : 0 < y < x < 1

0, otherwise
defined on the unit square S = [0, 1] × [0, 1] has discontinuities on the square
Integrability 307
boundary and also on the line y = x. In fact, on the boundary points (x0 , y0 )
we have the following cases:
1) if y0 = 0, 0 < x0 ≤ 1, then (x, y) ∈ D2 and
1 1
lim f (x, y) = lim = ̸= 0 = f (x0 , 0) ;
(x,y)→(x0 ,0) x→x0 x x0
2) if x0 = 0, 0 < y0 ≤ 1, then (x, y) ∈ D1 and

1 1
lim f (x, y) = lim = ̸= 0 = f (0, y0 ) ;
(x,y)→(0,y0 ) y→y0 y y0
3) if y0 = 1, 0 < x0 < 1, then (x, y) ∈ D1 and

1
lim f (x, y) = lim = 1 ̸= 0 = f (x0 , 1) ;
(x,y)→(x0 ,1) y→1 y
4) if x0 = 1, 0 < y0 < 1, then (x, y) ∈ D2 and

1
lim f (x, y) = lim = 1 ̸= 0 = f (1, y0 ) ;
(x,y)→(1,y0 ) x→1 x
5) if (x0 , y0 ) = (0, 0), then the partial limit for (x, y) ∈ D1 is

1
lim f (x, y) = lim = ∞;
(x,y)→(0,0) y→0 y
6) if (x0 , y0 ) = (1, 1), then the partial limit for (x, y) ∈ D1 is

1
lim f (x, y) = lim = 1 ̸= 0 = f (1, 1) .
(x,y)→(1,1) y→1 y
Hence, all the boundary points are the discontinuity points. Besides, for the
points (x0 , x0 ), 0 < x0 < 1, the partial limit when (x, y) ∈ D1 is
1 1
lim f (x, y) = lim = ̸= 0 = f (x0 , x0 ) .
(x,y)→(x0 ,x0 ) y→x0 y x0
Therefore, all the points on the line y = x are also discontinuity ones.
Nevertheless, the iterated integrals exist and are equal:
∫ 1 ∫ 1 ∫ 1 (∫ x ∫ 1 )
1 1
dx f (x, y) dy = dy + dy dx
0 0 0 0 x x y
∫ 1 ( y x ) ∫ 1
1 1
= + ln y|x dx = (1 − ln x) dx = (2x − x ln x)|0 = 2;
0 x 0 0
∫ 1 ∫ 1 ∫ 1 (∫ y ∫ 1 )
1 1
dy f (x, y) dx = dx + dx dy
0 0 0 0 y y x
∫ 1 ( y ) ∫ 1
x 1 1
= + ln x| dy = (1 − ln y) dy = (2y − y ln y)|0 = 2.
0 y 0 y
0
Remark 1. This shows that the conditions in Fubini’s theorem are sufficient
and even if they are not satisfied, the iterated integrals may exist and be equal.
Remark 2. The double integral in the above counterexample does not
exist, because the function is unbounded on S.
Remark 3. A similar counterexample can be given with the function
 −1
 y , 0<x<y<1
g (x, y) = −x−1 , 0 < y < x < 1

0, otherwise
defined on the unit square S = [0, 1] × [0, 1]. The principal difference between
f (x, y) and g (x, y) is that along the line y = x the former has removable
discontinuity (just redefine the function as f (x, x) = x1 for 0 < x < 1), while
the latter has a “wilder” discontinuity of the jump type.
FIGURE 9.3.4: Example 6, division of the domain S
Example 7. “If both iterated integrals exist and have the same value,
then the double integral also exists.”
Solution.
Let us consider the function f (x,{ y) defined on the unit square S =
1 , (x, y) ∈ T
[0, 1] × [0, 1] as follows: f (x, y) = , where T consists of
0, (x, y) ∈/T
Integrability 309
FIGURE 9.3.5: Example 6, graph of the function f
( )
the points of S such that (x, y) = m p , p , m, n, p ∈ N, where both fractions
n
are in lowest terms (note that the denominator is the same). First, let us show
∫1 ∫1
that the iterated integral 0 dy 0 f (x, y) dx exists. In fact, if we choose any
∫1
irrational y0 , then f (x, y0 ) = 0 and 0 f (x, y0 ) dx = 0. On the other hand,
for a rational y0 = np (with the fraction in lowest terms), f (x, y0 ) is different
from zero only in the finite number of the points, because for a given natural
p there are at most p numbers of the form x = m p , m ∈ N, in [0, 1]. Therefore,
∫1
again 0 f (x, y0 ) dx = 0. Since the inner integral is zero for any y0 , the iter-
∫1 ∫1
ated integral exists and 0 dy 0 f (x, y) dx = 0. Due to the symmetry of the
function and domain with respect to interchange
∫ 1 ∫ 1 of variables, the same result
is true for the second iterated integral: 0 dx∫∫0 f (x, y) dy = 0.
Now let us prove that the double integral S f (x, y) dA does not exist. To
this end, let us consider the special uniform partitions obtained by division
of each side of S in p equal parts, where p is a prime number. Note that
the diameter of such partitions approaches 0 as p approaches infinity. Each
elementary square
Rij = [xi−1 , xi ] × [yj−1 , yj ] , i, j = 1, . . . , p
contains the points (ci , dj ) ∈

/ T and also the points (ci , dj ) ∈ T . The possibility
of the former choice is evident, since in any interval [xi−1 , xi ] (or [yj−1 , yj ] )
there are irrational points. The latter choice can be made by using the points
of the upper right corner in each Rij :

( )
i j
(ci , dj ) = (xi , yj ) = , , i, j = 1, . . . , p − 1,
p p
except for the elementary squares of the upper row (j = p) and the right
column (i = p), for which we choose
( )
i p−1
(ci , dp ) = , , i = 1, . . . , p − 1,
p p
( )
p−1 j
(cp , dj ) = , , j = 1, . . . , p − 1,
p p
and ( )
p−1 p−1
(cp , dp ) = , .
p p
The corresponding double Riemann sums
p ∑
∑ p
S (f ; P ) = f (ci , dj ) · ∆xi ∆yj
j=1 i=1
are equal 0 in the case (ci , dj ) ∈

/ T and their partial limit is 0, but these sums
are equal 1 if (ci , dj ) ∈ T and the corresponding partial limit is 1. Since two
partial limits of the Riemann sums are different, there is no limit of the double
Riemann sums as the partition diameter approaches 0, which means that the
double integral does not exist.
Remark. The statement of this example is a strengthened (but still false)
version of the statement in Example 6.
Example 8. “If f (x, y) is bounded and integrable on a bounded set S,

then both iterated integrals over S exist and are equal to the double integral.”
Solution.
{ 1Let us define f m (x, y) on the square S = [0, 1] × [0, 1] as follows: f (x, y) =
n , y ∈ Q, x = n ∈ Q , where m, n ∈ N and the fraction m is in lowest
0 , otherwise n
terms (i.e., m and n are relatively prime integers). It can be shown that this
function is discontinuous at every point in (0, 1] × [0, 1] with both rational co-
ordinates and continuous at all remaining points. Indeed, in any neighborhood
of a rational point (x0 , y0 ), x0 , y0 ∈ Q, x0 ̸= 0, there are points with irrational
coordinates, and the partial limit along these points lim f (x, y) = 0
(x,y)→(x0 ,y0 )
(m ) x∈I or y∈I
is different from f (x0 , y0 ) = f n , y0 = n1 . It means that f (x, y) is dis-
continuous at every point with rational coordinates in (0, 1] × [0, 1]. For an
arbitrary point (x0 , y0 ) such that x0 ∈ I ∪ {0} or y0 ∈ I (or both), we split all
the paths of approaching (x0 , y0 ) into two groups. First, if a path consists of
Integrability 311
points (x, y) such that x ∈ I ∪ {0} or y ∈ I (or both), then we readily obtain
lim f (x, y) = 0 = f (x0 , y0 ). In the second group we place all paths
(x,y)→(x0 ,y0 )
x∈I or y∈I
containing only rational points with x ̸= 0. For any path of the second group,
since rational abscissas x = mn approach x0 , it implies that the denominator n
should approach infinity, because if it would not be so, that is, if the denomina-
tor n would be bounded, say n ≤ n0 , then it will be only finitely many numbers
of the form x = m n , n ≤ n0 in a neighborhood of the point x0 . However it will
contradict to the well-known fact that in any neighborhood of an arbitrary
real point there are infinitely many rational points. Since n → ∞ when x = m n
approach x0 , it follows that lim f (x, y) = lim n1 = 0 = f (x0 , y0 ).
(x,y)→(x0 ,y0 ) n→∞
x∈Q and y∈Q
The two considered groups of paths involve all the points in S, and therefore
lim f (x, y) = 0 = f (x0 , y0 ), i.e., f (x, y) is continuous at every point
(x,y)→(x0 ,y0 )
(x0 , y0 ), x0 ∈ I ∪ {0} or y0 ∈ I (or both).
Now let us show that f (x, y) is integrable on S (the boundedness of the
function is evident). Recall that a countable union of countable sets is again
a countable set. For any fixed rational y the number of points with rational
abscissas is countable. Since the number of rational ordinates y is countable
too, the set of points with both rational coordinates is countable (it is true both
for S and R2 ). Therefore it has the two-dimensional Lebesgue measure zero
and, according to the Lebesgue criterion, f (x, y) is integrable on S. Further,
we can find the value of the integral by noting that all the lower Riemann
sums are equal to 0:
∑
n ∑
m ∑
n
S (f ) = inf f (x, y) · ∆xi ∆yj = 0 · ∆xi ∆yj = 0
Rij
j=1 i=1 i=1
(since any elementary rectangle of an arbitrary partition contains the point

with irrational coordinates). Since the Riemann
∫∫ integral exists, its value is
equal to the lower Riemann integral, hence S f (x, y) dA = 0.
∫1 ∫1
For any fixed y ∈ I we have 0 f (x, y) dx = 0 0dx = 0, and for any fixed
y ∈ Q the given function is the one-dimensional Riemann function, which
is known to be integrable (see Remark 3 to Example 2 in section 5.3) and
∫1
0
f (x, y) dx = 0. So the first iterated integral exists and equals zero:
∫ 1 ∫ 1 ∫ 1
dy f (x, y) dx = 0dy = 0.
0 0 0
However, the second iterated

∫1 integral does not exist, because for any fixed
x ∈ Q\ {0} the integral 0 f (x, y) dy does not exist. In fact, for x = m
n (with
m, n being fixed natural numbers) we have
(m ) {
1/n , y ∈ Q, 1
φ (y) = f ,y = = D (y) ,
n 0 , otherwise n
where D (y) is Dirichlet’s function, which is known to be non-integrable in the

sense of the Riemann integral (see Example 1 in section 5.3). Therefore, φ (y)
is also non-integrable and the second iterated integral is not defined.
Remark. A straightforward generalization of this counterexample gives
the function whose{ 1double integral exist, but both iterated integrals does
n + 1q , y = pq ∈ Q, x = m n ∈ Q , defined on the square
not: g (x, y) =
0 , otherwise
S = [0, 1] × [0, 1], m, n, p, q ∈ N and the fractions m p
n and q are in lowest
terms. As above, it can be shown that this function is discontinuous at ev-
ery point in (0, 1] × (0, 1] with both rational coordinates and continuous at
all remaining points. Again, just like for f (x, y), the number of discontinu-
ity points
∫∫ is countable and consequently the double integral of g (x, y) exists
and S g (x, y) dA = 0. The difference occurs when we study the( first) it-
erated integral: for any fixed y = pq ∈ Q the function ψ (x) = g x, pq =
{ 1
n + q, x = n ∈ Q
1 m
is discontinuous at each point of the interval [0, 1],
0 , otherwise
∫1
because 1q is a fixed positive number. Therefore, the integral 0 g (x, y) dx
does not exist and the first iterated
∫ 1 integral is not defined. Similarly, for any
fixed x = m ∈ Q, the integral g (x, y) dy does not exists, since the func-
n { 1 0
(m ) n + 1
q , y = pq ∈ Q
tion φ (y) = g n , y = is discontinuous at each point
0 , otherwise
of the interval [0, 1]. Consequently, the second iterated integral is also not
defined.
Example 9. “If f (x, y) is continuous on R2 , then the double integral can

be calculated exactly using any of the iterated integrals.”
Solution. ( 2 )
The function f (x, y) = y ex y + e−x y is continuous on R2 according to
2
the arithmetic and composition rules. Since the conditions of Fubini’s theorem
are satisfied for any bounded set S in R2 , the double integral and both iterated
integrals exist and are equal. However, if we choose the square S = [−a, a] ×
[−b, b], then it is very simple to calculate the double integral using one of the
iterated integrals
∫∫ ∫ a ∫ b ( ) ∫ a
x2 y −x2 y
f (x, y) dA = dx y e +e dy = 0dx = 0
S −a −b −a
(since f (x, y) is odd with respect to y), but )it cannot be done by using another
∫b ∫ a ( x2 y
+ e−x y dx, because the original function
2
iterated integral −b dy −a y e
does not have any helpful symmetry with respect to x and does not have an
antiderivative in the class of elementary functions. The last integral still can
be calculated by expending the function in power series, but it will result
(after integration) in a numerical series whose sum usually cannot be found
exactly.
Integrability 313
Example 10. “If a transformation x = φ (u, v), y = ψ (u, v) between sets

D̃ and D is not a one-to-one mapping and/or its Jacobian vanishes at some
points (u, v) ∈ D̃, then the corresponding change of variables in the double
integral is not applicable.”
Solution.
For simplicity
∫∫ we assume that the function f (x, y) ≡ 1, that is, the double
integral D f dxdy in the Cartesian coordinates (x, y) represents the area of
the figure in D. Let us consider the transformation { of the rectangle
} D̃ =
[0, 1] × [0, 2π] onto the disk D = B̄1 (0, 0) = (x, y) : x2 + y 2 ≤ 1 defined
by the formulas x = u cos v, y = u sin v, representing the relation between
the (Cartesian ) and polar coordinates. The Jacobian of this transformation
xu xv
det = u vanishes at the points where u = 0. It can also be noticed
yu yv
that the entire line segment u = 0 , v ∈ [0, 2π] in D̃ corresponds to the only
point (x, y) = (0, 0) in D, and that two different segments u ∈ [0, 1] , v = 0
and u ∈ [0, 1] , v = 2π in D̃ are transformed onto the same segment x ∈
[0, 1] , y = 0. Therefore, this correspondence is not one-to-one. However, the
result on the change of variables is still valid and can be justified by the
following reasoning. Consider a slight smaller rectangle D̃δ = [δu , 1] × [δv , 2π],
δu , δv > 0, which is transformed under the same { mapping to the disk } D
with two peaces removed { - small disk Bδ (0, 0) = (x, y) : x2 + y 2 }< δu2 and
small sector Cδ (0, 0) = (x, y) : x2 + y 2 ≤ 1 , 0 < y < tan δv · x , that is
Dδ = D\ (Bδ (0, 0) ∪ Cδ (0, 0)). Under this transformation, both problems of
vanishing of the Jacobian and non-uniqueness correspondence between the
points are eliminated, and all the conditions of the theorem on change of
variables are satisfied, that makes possible the use of the formula:
∫∫ ∫∫ ∫∫
∂ (φ, ψ)
1dxdy = dudv = |u| dudv

Dδ D̃δ ∂ (u, v) D̃δ
∫ ∫
2π 1
1 − δu2
= dv udu = (2π − δv ) .
δv δu 2
Since this result is true for any small δu , δv > 0, applying the limit, as δu , δv
approaches 0, to the both sides of the formula, one obtains:
∫∫ ∫∫ ∫ 2π ∫ 1
∂ (φ, ψ)
1dxdy = dudv = dv udu = π.

D D̃ ∂ (u, v) 0 0
Remark 1. This counterexample shows that the conditions of the theorem

on the change of variables are sufficient, but not necessary. In practice, we
frequently need to apply the theorem on change of variables in the cases
when the conditions of non-zero Jacobian and one-to-one correspondence are
violated at some points. The use of the popular polar coordinates (like in the
counterexample above) is just one of typical examples of such situation. In
general, if the two required conditions are not satisfied at a finite number of
points or on a finite number of simple rectifiable curves, then the theorem
on the change of variables is still applicable, with justification similar to the
presented above.
Remark 2. The next example shows that when one of the conditions
is violated on a large part of a set, then the change of variables may be
unjustifiable.
Example 11. “If f (x, y) ≡ 1 and a transformation x = φ (u, v), y =

ψ (u, v) between sets D̃ and D has continuous partial derivatives on D̃ and
the Jacobian is positive on D̃, except possibly at the only point, then the
result of the theorem on the change of variables is valid.”
Solution.
Let us consider the transformation x = u2 − v 2 , y = 2uv defined and con-
tinuously differentiable on D̃ = [−1, 1] × [−1, 1] with the image D being the
y2 y2
set enclosed by the( parabolas ) x = 1 − 4 and x = 4 − 1. The correspond-
xu xv ( )
ing Jacobian det = 4 u2 + v 2 is positive at every point, but
yu yv
at the origin. The last fact alone does not impede the use of the change of
variables (see justification in the previous Example 10), and applying formally
the change formula we obtain
∫∫ ∫∫ ∫ 1 ∫ 1
∂ (φ, ψ) ( ) 32
1dxdy = dudv = dv 4 u2 + v 2 du = .
∂ (u, v) 3
D D̃ −1 −1
However, the alternative evaluation of the area of figure D through the Rie-
mann integral gives a different result:
∫ 2( ) ( 2 )
y2 y 16
1− − − 1 dy = .
−2 4 4 3
The problem with the change of variables in this case is that the used trans-
formation is not one-to-one mapping between D̃ and D. It happens that a
right “half” of D̃, D̃r = [0, 1] × [−1, 1] already covers all the image D, and
the correspondence between D̃r and D is one-to-one with the only excep-
tion on the boundary points u = 0 , v ∈ [−1, 1], which cover twice the
segment y = 0 , x ∈ [−1, 0] in D. The same is true for the left “half”
D̃l = [−1, 0] × [−1, 1]. In this way, the transformation from D̃ to D cov-
ers the image twice, violating the condition of one-to-one correspondence on
the entire domain. This is the cause of the wrong use of the change of vari-
ables above. The correct mode is to select one of the “halves” of D̃, say D̃r ,
and then to apply (justifiably) the change of variables, obtaining the correct
result:
∫∫ ∫∫ ∫ 1 ∫ 1
∂ (φ, ψ) ( ) 16
1dxdy = 4 u2 + v 2 du =
∂ (u, v) dudv = dv
3
D D̃r −1 0
Integrability 315
(the use of D̃l gives the same result and this is why the wrong result obtained
for D̃ is twice as much as the correct one).
FIGURE 9.3.6: Example 11, right rectangle
FIGURE 9.3.7: Example 11, left rectangle

Example 12. “If P (x, y) and Q (x, y) are ∫continuously differentiable, and
Py = Qx in a domain D, then the line integral C P dx+Qdy along any smooth
closed curve C lying in D is equal zero.”
Remark. In all the line integrals along a closed curve, we use the convention
that a curve is traversed once in the positive (counterclockwise) direction.
Solution.
Consider the functions P (x, y) = − x2 +y y x
2 and Q (x, y) = x2 +y 2 on do-
main D = R \ (0, 0). According to the arithmetic rules, both functions are
2
continuous in D, and their partial derivatives
2xy y 2 − x2 y 2 − x2 2xy
Px = 2 , Py = 2 , Q x = 2 , Qy = − 2
(x2 + y 2 ) 2 2
(x + y ) 2
(x + y )2 (x + y 2 )
2
are also continuous functions in D. Therefore, all the statement conditions

hold. Since Py = Qx , in a simply connected part of D it is easy to find the
function F (x, y) = − arctan xy , y ̸= 0, whose exact differential is P dx + Qdy.
Notice, however, that P ,Q, Py and Qx are not continuous at the origin:
( ) ( )
1 1
lim P (0, y) = lim − = ∞, lim Q (x, 0) = lim = ∞,
y→0 y→0 y x→0 x→0 x
1
lim Py (0, y) = lim Qx (0, y) = lim = ∞,
y→0 y→0 y→0 y2
and this causes violation of the integral formula in the statement for any
curve
{ 2 C2 going} around the origin. For instance, if C is the unit circumference
x + y = 1 , then using the polar coordinates r, θ and θ as the integration
parameter, we obtain:
∫ ∫ 2π
P dx + Qdy = P (θ) xθ dθ + Q (θ) yθ dθ
C 0
∫ 2π ∫ 2π
= (− sin θ) (− sin θ) dθ + cos θ cos θdθ = dθ = 2π ̸= 0.
0 0
Remark 1. The integral does not vanish because one of the conditions of
the corresponding theorem on the line integral is not satisfied: the domain
D should be simply-connected in order to guarantee that the line integral
vanishes. If we add the origin to the domain D, then the new domain will be
simply connected, but the functions P , Q, Py and Qx will not be continuous
in the new domain, because they are not continuous at the origin.
Remark 2. If we choose a simply connected domain D, which does not
contain the origin, then the statement will be true.
Remark 3. Notice, that the condition of continuity of P , Q, Py and Qx
is only sufficient in the referred theorem, but not necessary as shown in the
next example.
Integrability 317
Example 13. “If at least one of the functions P, Q, Py = Qx has a discon-

tinuity at a point (x0 , y0 ) of a simply connected domain D, then there exists a
smooth closed
∫ curve C lying in D and not passing through the point (x0 , y0 )
such that C P dx + Qdy ̸= 0.”
Solution.
Set D = R2 and consider the functions P (x, y) = (x22xy +y 2 )2
and Q(x, y) =
y 2 −x2
(x2 +y 2 )2
, which are continuous in R2 \ (0, 0) together with their partial deriva-
−6xy 3 2
tives Py = Qx = 2x(x2 +y 2 )3
(according to the arithmetic rules). All these func-
tions have a discontinuity at the origin:
1 −1
lim P (x, x) = lim = ∞, lim Q (x, 0) = lim 2 = ∞,
x→0 y→0 2x2 x→0 x→0 x
2
= ∞.
lim Py (x, 0) = lim Qx (x, 0) = lim
x→0 x→0 x3 x→0
Now let us choose a smooth closed curve C that does not pass through the
origin. If the point (0, 0) lies outside the domain enclosed by C, it is always
possible to construct an auxiliary simply connected domain, which contains
C and does not contain (0, 0).∫ Therefore, for such domain the conditions of
the theorem are satisfied and C P dx + Qdy = 0. If the point (0, 0) lies inside
the domain D̃ enclosed by C, then the { 2 situation is
} more complex. Let us
2 2
introduce
∫ a small circumference
∫ C a = x + y = a inside D̃ and note that
C
P dx + Qdy = Ca
P dx + Qdy, since P dx + Qdy is an exact differential in
R2 \ (0, 0) (i.e. P dx + Qdy = df ). Calculating the last integral in the polar
coordinates r, θ, with θ being the integration parameter, we obtain:
∫ ∫ 2π
P dx + Qdy = P (θ) xθ dθ + Q (θ) yθ dθ
Ca 0
∫ ( )
2π
2a2 cos θ sin θ a2 sin2 θ − cos2 θ
= a · (− sin θ) + a · cos θdθ
0 a4 a4
∫
1 2π
=− cos θdθ = 0.
a 0
Therefore, the line integral along any smooth closed curve C lying in D and
not passing through the origin is zero.
Example 14. “If f (x, y) and g(x,

∫∫ y) have the sets∫of discontinuity points
dense in D, then Green’s formula D (gx − fy ) dA = ∂D f dx + gdy does not
hold.”
Solution.
Consider f (x, y) = 2xy + R(x) and g(x, y) = x2 − y 2 + R(y) on D = [0, 1] ×
[0, 1] (here R(x) and R(y) are the one-dimensional Riemann functions). In
Example 4, section 3.3 it was shown that R(x) is continuous at every irrational
point and discontinuous at every rational point. Therefore, the function f (x, y)
is discontinuous at each point of the segments y ∈ [0, 1], ∀x ∈ Q∩[0, 1] and the
function g(x, y) is discontinuous at each point of the segments x ∈ [0, 1], ∀y ∈
Q∩[0, 1]. Let us denote the sets of discontinuity points of the functions f (x, y)
and g(x, y) by Ef and Eg , respectively. Evidently, both sets are dense in D,
that is, Ēf = D and Ēg = D. Nevertheless, Green’s formula
∫∫ ∫
(gx − fy ) dA = f dx + gdy
D ∂D
is satisfied. To show this, let us evaluate the left- and right-hand sides
of the formula. First, gx = 2x = fy for ∀(x, y) ∈ D, and consequently
∫∫
D
(gx − fy ) dA = 0. The line integral along ∂D can be divided in the four
line integrals along the sides of the square D:
∫ ∫ ∫ ∫ ∫
f dx + gdy = f dx + gdy + f dx + gdy + f dx + gdy + f dx + gdy,
∂D AB BC CK KA
where
AB = {(x, y) : y = 0, 0 ≤ x ≤ 1}, BC = {(x, y) : x = 1, 0 ≤ y ≤ 1},
KC = {(x, y) : y = 1, 0 ≤ x ≤ 1}, AK = {(x, y) : x = 0, 0 ≤ y ≤ 1}.

Calculating each line integral separately, we obtain:
∫ ∫ 1 ∫ 1
f dx + gdy = f (x, 0)dx = 2x · 0 + R(x)dx = 0,
AB 0 0
∫ ∫ ∫
1 1 1
f dx + gdy = − f (x, 1)dx = − 2x · 1 + R(x)dx = − x2 0 = −1,
CK 0 0
∫ ∫ ∫ )1 (
1 1
y 3 2
f dx + gdy = g(1, y)dy = 1 − y 2 + R(y)dy =
= , y−
BC 0 0 3 0 3
∫ ∫ 1 ∫ 1
y 3
1
1
f dx + gdy = − g(0, y)dy = − −y 2 + R(y)dy = = 3.
KA 0 0 3 0
(Integrability of the Riemann function and the fact that its definite integral
is zero was shown in Remark 3 to Example 2, section 5.3.) Joining all the
calculations, we get
∫
2 1
f dx + gdy = 0 − 1 + + = 0,
∂D 3 3
that is, Green’s formula holds in this case.

Integrability 319
Exercises
1. Show by example that the following statement is false: “if f∫∫ (x, y) is
bounded on a rectangle R = [a, b] × [c, d], then the double integral R f dA
can be defined by partitioning R in elementary rectangles Rij of the same size
and choosing the points (ξi , ηj ) in the Riemann sum to be the centerpoints of
Rij ”. What if f (x, y) is integrable on R?
2. Give an example that shows that the condition of independence of the
limit on the choice of points (ξi , ηj ) is important in the definition of the dou-
ble integral.{ (Hint: use the two-dimensional version of Dirichlet’s function
1, x, y ∈ Q
D̂(x, y) = and apply two different rules for the choice of
0, otherwise
the points (ξi , ηj ): choose the points with rational coordinates and then with
irrational coordinates in each Rij . )
3. Show that the partition diameter ∆ in the double integral definition
cannot be substituted by any combination of the numbers of intervals n and
m in x- and y-direction. In particular, construct a counterexample to the
following false definition of the double integral of f (x, y) over a rectangle R:
“if there exists a finite limit of the Riemann sums, as n and m approach ∞,
which does not depend on a choice of points (ξi , ηj ), then this limit is called
the double integral of f (x, y) over R”. (Hint: divide each side [a, b] and [c, d]
of the rectangle R = [a, b] × [c, d] in two equal parts and then apply refinement
only two one of the obtained halves of each interval.)
Show that the limits lim S (f ) and lim S (f ) are equal if one uses the
∆→0 m,n→∞
uniform partitions in x- and y-directions.
4. Show by example that the converse to the comparative property 1) in
section 9.1 is not true, that is, the∫∫
following statement
∫∫ is false: “if f (x, y) and
g (x, y) are integrable over D and D f dA ≤ D gdA, then f (x, y) ≤ g (x, y)
on D”. {
1, x, y ∈ Q
5. Show that D̂(x, y) = is another counterexample to
0, otherwise
the statement in Example 1,{section 9.2.
1, x, y ∈ Q
6. Show that D̃(x, y) = is another counterexample to
−1, otherwise
7. Provide a counterexample to the following statement: “if f (x, y) is
discontinuous at a countable set of points in a set D, then it is not inte-
grable over D”. Compare this result with Example 2, section 9.3. (Hint: con-
sider D = [0, 1] × [0, 1], E = ∪+∞
{ n=1 en , en = (1/n, 1/n), ∀n ∈ N, f (x, y) =
1, (x, y) ∈ E
)
0, (x, y) ∈ D\E
8. Show that another counterexample to the statement in Example 4, sec-
tion 9.3. can be composed{as follows: the region D = [−1, 1] × [−2, 2] and the
sin πx , x ∈ [−1, 0) ∪ (0, 1]
dividing curve is h(x) = . Use the reasoning of
0, x = 0
Example 4 to show that h(x)  is a zero area curve.
1
 y2 , 0 < x < y < 1
9. Show that f (x, y) = − 1 , 0 < y < x < 1 considered on [0, 1] × [0, 1]
 x2
0, otherwise
is another counterexample for Example 5 in section 9.3.
10. Show that f (x, y) = D(x) · R(y), where D(x) and R(y) are one-
dimensional Dirichlet’s and Riemann’s functions, is another counterexample
for Example 8 in section 9.3.
2
−v 2
11. Verify that the transformation x = (uu2 +v 2 )2 , y = − (u2 +v 2 )2 has contin-
2uv
uous partial derivatives and positive Jacobian on the set D̃ located between
the circles u2 + v 2 = 12 and u2 + v 2 = 1 and bounded by the segments
2 ≤ u ≤ 1, v = 0 and −1 ≤ v ≤ − 2 , u = 0 (this is three-quarters of the ring
1 1
without the part located in u > 0, v < 0 ), but the theorem on the change of
variables does not hold even for f (x, y) ≡ 1. Find out what is the problem
with this transformation and formulate the result as a counterexample.
12. Provide a counterexample to the statement: “if a transformation
x = φ (u, v), y = ψ (u, v) between sets D̃ and D has zero Jacobian on a
subset D̃1 ⊂ D̃ of non-zero area, then the corresponding change of variables
in the double integral is not applicable”. Compare this result with Exam-
ple
{ 10. 2(Hint: consider f (x, y) ≡ 1 and the transformation x = φ (u, v) =
0, u + v 2 < 1
, y = ψ (u, v) = u · v on the set D̃ enclosed
(u2 + v 2 − 1)2 , u2 + v 2 ≥ 1
by the rays u = v, u = −v for u > 0 and by the arc of the circumference
u2 + v 2 = 2 .)
13. Show that P (x, y) = − (5x+2y)2+(3x+2y)
y x
2 and Q(x, y) = (5x+2y)2+(3x+2y)2
considered on D = R give another counterexample for Example 12, section

2
9.3.
14. Prove that Green’s formula does not hold for the functions of Exam-
ple 12, section 9.3, P (x, y) = − x2 +y y x
2 and Q(x, y) = x2 +y 2 considered on
the disk D = x2 + y 2 ≤ 1 or any domain containing the origin. Prove also

that Green’s formula is satisfied for the functions of Example 13, section 9.3,
2
−x2
P (x, y) = (x22xy
+y 2 )2
and Q(x, y) = (xy2 +y 2 )2 on the same unit disk or any do-
main containing the origin. Notice that in both cases the conditions of Green’s
formula are satisfied at all points except only for the origin. Compare results
and formulate them as counterexamples.
2 and Q(x, y) = x2 +y 2 considered on D = R
x y 2
15. Show that P (x, y) = x2 +y
give another counterexample for Example 13, section 9.3.
16. Provide a counterexample to the following statement: “if∫∫ f (x, y) has
the∫ set of discontinuity points dense in D, then Green’s formula D fy dA =
− ∂D f dx does not hold”. Compare this result with Example 14 in section
9.3. (Hint: consider f (x, y) = x2 + y 2 + R(x), where R(x) is the Riemann
Integrability 321
function, on the set D bounded by the curves y = x2 + 1, y = −x2 , x = 0 and

x = 1.)
17. Provide a counterexample to the following statement: “if∫∫g(x, y) has
the
∫ set of discontinuity points dense in D, then Green’s formula D gx dA =
∂D
gdy does not hold”. Compare this result with Example 14 in section 9.3
and with the previous exercise. (Hint: consider g(x, y) = 3xy 2 + R(y), where
R(y) is the Riemann function, on the set D bounded by the curves x = y 2 ,
x = y 2 + 2, y = 0 and y = 1.)
Bibliography
[1] S. Abbott. Understanding Analysis. Springer, New York, NY, 2002.
[2] P. Biler and A. Witkowski. Problems in Mathematical Analysis. Marcel

Dekker, New York, NY, 1990.
[3] B. Demidovich. Problems in Mathematical Analysis. Mir Pub., Moscow,

1989.
[4] G.M. Fichtengolz. Differential- und Integralrechnung, Vol.1-3. V.E.B.

Deutscher Verlag Wiss., Berlin, 1968.
[5] B.R. Gelbaum and J.M.H. Olmsted. Counterexamples in Analysis. Dover

Pub., Mineola, NY, 2003.
[6] V.A. Ilyin and E.G. Poznyak. Fundamentals of Mathematical Analysis,

Vol.1,2. Mir Pub., Moscow, 1982.
[7] W.J. Kaczor and M.T. Nowak. Problems in Mathematical Analysis,

Vol.1-3. AMS, Providence, RI, 2001.
[8] B.M. Makarov, M.G. Goluzina, A.A. Lodkin, and A.N. Podkorytov. Se-
lected Problems in Real Analysis. AMS, Providence, RI, 1992.
[9] G. Polya and G. Szego. Problems and Theorems in Analysis, Vol.1-2.

Springer, Berlin, 1998.
[10] T.L.T. Radulesku, V.D. Radulesku, and T. Andreescu. Problems in Real

Analysis: Advanced Calculus on the Real Axis. Springer, New York, NY,
2009.
[11] W. Rudin. Principles of Mathematical Analysis. McGraw-Hill, New York,

NY, 1976.
[12] M. Spivak. Calculus. Publish or Perish, Houston, TX, 2008.

[13] J. Stewart. Calculus: Early Transcendentals. Brooks/Cole, Belmont, CA,
2012.
[14] V.A. Zorich. Mathematical Analysis I, II. Springer, Berlin, 2004.
323
Symbol Description
= equality, definition, notation sup supremum, least upper

→,⇒ logical implication, ten- bound
dency, correspondence inf infimum, greatest lower
{. . .} collection of elements bound
∈ (∈ / ) belongs (does not belong) to max maximum
⊂ (̸⊂ ) is contained (is not con- min minimum
tained) in f (x) function
∀ any, every, each, for all f: X → Y function, correspondence,
∃ (̸ ∃) existence (non-existence) mapping from X to Y
N set of natural numbers f −1 (x) inverse function
Z set of integers D(x) Dirichlet’s function
R(x) Riemann function
Q set of rational numbers
lim f limit of f (x) as x approaches
R set of real numbers, one- x→a
a
dimensional real space
f → limit of f as x approaches a
Rn n-dimensional real space x→a
∅ empty set lim f right-hand limit of f as x
x→a+
D̄ closure of set D approaches a
∂D boundary of set D f → right-hand limit of f as x
x→a+
∪ union of sets approaches a
∩ intersection of sets lim f left-hand limit of f as x
\ difference of sets x→a−
approaches a
× Cartesian product of sets
f → left-hand limit of f as x
d(x, y), |x − y| distance between two x→a−
points approaches a
Sc,r sphere with radius r and ∆x increment of variable x
centerpoint c ∆f increment of function f (x)
Bc,r open ball with radius r and f ′ , dx
df
derivative of f
′
centerpoint c f+ right-hand derivative of f
′
B̄c,r closed ball with radius r and f− left-hand derivative of f
2
centerpoint c f ′′ , ddxf2 second derivative of f
n
(a, b) open interval f n , ddxnf n-th derivative of f
[a, b] closed interval df differential of f
(a, b], [a, b) half-open (half-closed) in- d2 f second order differential of f
tervals dm f m-th order differential of f
325
326 Symbol Description
∫
f dx integral, indefinite integral lim lim f iterated limit of function of
∫b x→a y→b
a
f dx definite (Riemann) integral, two variables
improper integral of the sec- fx ,∂x f partial derivative in x
ond kind ∂f
∂x partial derivative in x
S(f ; P, c),S(f ) Riemann sum fxx ,∂xx f second-order partial deriva-
∆ partition diameter tive in x, x
∫ +∞ ∫b ∂2f
f dx, −∞f dx improper integral of ∂x2 second-order partial deriva-
a
the first kind tive in x, x
∫ +∞ fxm ,fx...x m-th order partial deriva-
−∞
f dx improper integral of the first
kind tive in x, . . . , x
∂mf
∞
{an }n=1 ,an numerical sequence ∂xm m-th order partial derivative
in x, . . . , x
lim an limit of sequence an
∑
n→∞ ∇f ,gradf gradient
an sum of sequence an , numer- Dv f directional derivative
ical series H(f ) Hessian matrix
∑+∞ ∫∫
∑∞ n=1 an numerical series ∫∫D f dA double integral over D
n f dxdy double integral over D
n=0 cn (x−a) power series
∫ bD ∫ d
lim f limit of function of two a
dx c f dy iterated integral
(x,y)→(a,b) ∂(φ,ψ)
variables Jacobian
(
∂(u,v) )
f → limit of function of two φu φv
(x,y)→(a,b) det Jacobian
∫ ψu ψv
variables
f dl line integral in arc length
lim lim f iterated limit of function of ∫C f dx line integral in x
y→b x→a ∫C
two variables C
f dy line integral in y
Mathematics
TEXTBOOKS in MATHEMATICS TEXTBOOKS in MATHEMATICS
CounterExamples
CounterExamples
This book provides a one-semester undergraduate introduction to
counterexamples in calculus and analysis. It helps engineering, natu-
ral sciences, and mathematics students tackle commonly made er- From Elementary Calculus to
roneous conjectures. The book encourages students to think criti-
cally and analytically, and helps to reveal common errors in many the Beginnings of Analysis
examples.
In this book, the authors present an overview of important concepts
and results in calculus and real analysis by considering false state-
ments, which may appear to be true at first glance. The book cov-
ers topics concerning the functions of real variables, starting with
elementary properties, moving to limits and continuity, and then to
differentiation and integration. The first part of the book describes
single-variable functions, while the second part covers the functions
of two variables.
The many examples presented throughout the book typically start
at a very basic level and become more complex during the develop-
ment of exposition. At the end of each chapter, supplementary exer-
cises of different levels of complexity are provided, the most difficult
of them with a hint to the solution.
This book is intended for students who are interested in develop-
ing a deeper understanding of the topics of calculus. The gathered
Bourchtein
Bourchtein
counterexamples may also be used by calculus instructors in their
classes.
Andrei Bourchtein
K23601
Ludmila Bourchtein
K23601_cover.indd 1 8/1/14 9:05 AM

(Andrei Bourchtein, Ludmila Bourchtein) CounterExa PDF

Uploaded by

Copyright:

Available Formats

(Andrei Bourchtein, Ludmila Bourchtein) CounterExa PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Andrei Bourchtein, Ludmila Bourchtein) CounterExa PDF

Uploaded by

Copyright:

Available Formats

What are some of the main concepts discussed in the text?

What are some of the main concepts discussed in the text?

What are some examples of different types of sets mentioned?

What are some examples of different types of sets mentioned?

Mathematics

TEXTBOOKS in MATHEMATICS TEXTBOOKS in MATHEMATICS

K23601_cover.indd 1 8/1/14 9:05 AM

K23601_FM.indd 1 8/4/14 12:49 PM

DIFFERENTIAL EQUATIONS WITH MATLAB®: EXPLORATION, APPLICATIONS, AND THEORY

K23601_FM.indd 3 8/4/14 12:49 PM

K23601_FM.indd 5 8/4/14 12:49 PM

No claim to original U.S. Government works

International Standard Book Number-13: 978-1-4822-4668-1 (eBook - PDF)

I Functions of one real variable 1

1.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 35

3.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 53

3.4 Global properties: the famous theorems . . . . . . . . . . . . 69

4.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 89

5.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 133

6 Sequences and series 175

6.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 175

II Functions of two real variables 205

7.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 207

7.2.2 Global properties . . . . . . . . . . . . . . . . . . . . . 217

8.1 Elements of Theory . . . . . . . . . . . . . . . . . . . . . . . 237

9.1 Elements of theory . . . . . . . . . . . . . . . . . . . . . . . . 287

Symbol Description 325

In this manuscript we present counterexamples to diﬀerent false state-

4.2.1 Example 3, graph of the function . . . . . . . . . . . . . . . 95

5.2.1 Example 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.4.6 Example 4 and Example 5 . . . . . . . . . . . . . . . . . . . 158

6.2.1 Example 6 and Example 7 . . . . . . . . . . . . . . . . . . . 182

7.2.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

8.2.1 Example 1 and Example 2 . . . . . . . . . . . . . . . . . . . 243

8.3.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

9.2.1 Example 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296

0.1.1 On the structure of this book

1) in chapter 7 - limits and continuity, with details on relation between general,

The following structure is chosen for the presentation of each example:

place, it is related to the nature of each counterexample (if it is more analytic

0.1.2 On mathematical language and notation

The language in this manuscript, just as in the majority of mathematical

0.2 Background (elements of theory)

The background material is split in several parts around the manuscript.

Set. We consider a set to be a primary non-deﬁnable concept, which can

The complement of B ⊂ A with respect to A is the set of all elements

The space Rn . The n-dimensional space Rn can be considered as the

In the case n = 1, the used notation of the space is simply R.

Distance between two points. The distance between two points

Limit point. A point c is a limit point of the set S if in any deleted

Open set. A set S is open if every point of S is its interior point.

Curves. The set of points x = x (t), t ∈ [a, b] is called a curve with

General Remark. Any characterization of a set (such as connectedness,

Interval. The interval is a set of points x in R, which satisfy one of

Function, domain, image, pre-image. A correspondence f between

Graph of a function. If f is deﬁned on X, then the graph of f is the

Mapping onto. If a function f (x) deﬁned on X has image Y , that is

onto Y , then f (x) is said to be a one-to-one correspondence between X and