Ca

Ca lcu
ag
ai l on a
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on ce a gain Ca lcu
Ca lcu us,
l o ce ga l c l us
Ca lcu u s, nc ag in C a ul ,
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
,o c g l
n e a ain Ca lcu us,
l on
on ce a gain Ca lcu us,
ce ga l c lu on ce
C s
ag in a ul , o ce
n a
Ca lcu us,
ag ain l on ce a gain
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on ce a gain Ca lcu
Ca lcu us,
l on ce a gain l
Ca lcu us,
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce g l
Ca lcu us,
nc
on e a ain
ag ain l on
Ca lcu us,
ce g l on ce
Ca lcu us,
ag ain l on ce a
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on e a ain c g
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on ce a gain Ca lcu
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce ga l
n a in Ca lcu us,
l on
on ce a gain Ca lcu us,
ce g l on ce
ag ain Ca lcu us,
l on e c
Ca lcu us,
ag ain l on ce a
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on e a ain c g
Ca lcu
Ca lcu us,
l on ce a gain l
Ca lcu us,
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce g l
Ca lcu us,
nc
on e a ain
ag ain l o n
Ca lcu us,
ce g l on ce
Ca lcu us,
ag ain l on ce a
Ca lcu us,
ag ain l on ce a gain
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l o ce ga
Ca lcu us, n a in
l on ce a gain Ca
Ca lcu us,
l on ce a gain Ca lcu
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce ga l c lu
n a in Ca ul s ,o
on ce a gain Ca lcu us, nc
on ce a gain l on e
Ca lcu us,
ce g l on ce
Ca lcu us,
ag ain l on ce a
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on e a ain c g
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on ce a gain Ca lcu
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce ga l
n a in Ca lcu us,
l
l u on
on ce a gain
ce ga
Ca cu s, c
ag i n C a
l c l us onc e
ul ,o
Ca lcu us, n ea
ag ain l on ce a gain
Ca lcu us,
ag ain
ai l on ce a gain
n Ca lcu us,
l on ce a gain
Ca lcu us,
l on ce a gain Ca
Ca lcu us,
l on ce a ain g
Ca lcu
Ca lcu us,
l on ce a gain l
Ca lcu us,
Ca lcu us, c g l
Ca lcu us,
lc lus onc e ag ain
ul l
us , on e a ain Ca lcu us,
, o ce g l
Ca lcu us,
nc
on e a ain
ag ain l on
Ca lcu us,
ce g l on ce
Ca lcu us,
ag ain
ai l on ce a
n Ca lcu us,
l on ce a gain
Ca lcu us, c g
lc lus onc e ag ain
ul , e a
us on i
, o ce aga n C

November 6, 2007 Version

nc
e
ag in a
Ca lcu
ag ain
ai l
n Ca lcu us,
l
Ca lcu us,
l c l us
ul
us , on
, o ce
nc
e
[email protected]
David A. SANTOS
ii

Copyright © 2007 David Anthony SANTOS. Permission is granted to copy, distribute and/or modify this docu-
ment under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by
the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A
copy of the license is included in the section entitled “GNU Free Documentation License”.
GNU Free Documentation License

Version 1.2, November 2002

Copyright © 2000,2001,2002 Free Software Foundation, Inc.

51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA

Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.

Preamble
The purpose of this License is to make a manual, textbook, or other functional and useful document “free” in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either
commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others.
This License is a kind of “copyleft”, which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software.
We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License
is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.

1. APPLICABILITY AND DEFINITIONS

This License applies to any manual or other work, in any medium, that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. Such a notice grants a world-wide, royalty-free license,
unlimited in duration, to use that work under the conditions stated herein. The “Document”, below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as “you”. You accept the license if you copy, modify
or distribute the work in a way requiring permission under copyright law.
A “Modified Version” of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language.
A “Secondary Section” is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document’s overall subject (or to related matters) and
contains nothing that could fall directly within that overall subject. (Thus, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection
with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them.
The “Invariant Sections” are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License. If a section does not fit the above definition
of Secondary then it is not allowed to be designated as Invariant. The Document may contain zero Invariant Sections. If the Document does not identify any Invariant Sections then there are none.
The “Cover Texts” are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License. A Front-Cover Text may be at most 5 words, and a
Back-Cover Text may be at most 25 words.
A “Transparent” copy of the Document means a machine-readable copy, represented in a format whose specification is available to the general public, that is suitable for revising the document straightforwardly with generic text editors or
(for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters.
A copy made in an otherwise Transparent file format whose markup, or absence of markup, has been arranged to thwart or discourage subsequent modification by readers is not Transparent. An image format is not Transparent if used for any
substantial amount of text. A copy that is not “Transparent” is called “Opaque”.
Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML, PostScript or PDF
designed for human modification. Examples of transparent image formats include PNG, XCF and JPG. Opaque formats include proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the
DTD and/or processing tools are not generally available, and the machine-generated HTML, PostScript or PDF produced by some word processors for output purposes only.
The “Title Page” means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as
such, “Title Page” means the text near the most prominent appearance of the work’s title, preceding the beginning of the body of the text.
A section “Entitled XYZ” means a named subunit of the Document whose title either is precisely XYZ or contains XYZ in parentheses following text that translates XYZ in another language. (Here XYZ stands for a specific section name
mentioned below, such as “Acknowledgements”, “Dedications”, “Endorsements”, or “History”.) To “Preserve the Title” of such a section when you modify the Document means that it remains a section “Entitled XYZ” according to this definition.
The Document may include Warranty Disclaimers next to the notice which states that this License applies to the Document. These Warranty Disclaimers are considered to be included by reference in this License, but only as regards
disclaiming warranties: any other implication that these Warranty Disclaimers may have is void and has no effect on the meaning of this License.

2. VERBATIM COPYING
You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced
in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept
compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3.
You may also lend copies, under the same conditions stated above, and you may publicly display copies.

3. COPYING IN QUANTITY
If you publish printed copies (or copies in media that commonly have printed covers) of the Document, numbering more than 100, and the Document’s license notice requires Cover Texts, you must enclose the copies in covers that carry,
clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full
title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be
treated as verbatim copying in other respects.
If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages.
If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a computer-network
location from which the general network-using public has access to download using public-standard network protocols a complete Transparent copy of the Document, free of added material. If you use the latter option, you must take reasonably
prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or
through your agents or retailers) of that edition to the public.
It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.

4. MODIFICATIONS
You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the
Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version:

A. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same
title as a previous version if the original publisher of that version gives permission.

B. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors,
if it has fewer than five), unless they release you from this requirement.

C. State on the Title page the name of the publisher of the Modified Version, as the publisher.

iii
iv

D. Preserve all the copyright notices of the Document.

E. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices.

F. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below.

G. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document’s license notice.

H. Include an unaltered copy of this License.

I. Preserve the section Entitled “History”, Preserve its Title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section Entitled “History” in the
Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence.

J. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be
placed in the “History” section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission.

K. For any section Entitled “Acknowledgements” or “Dedications”, Preserve the Title of the section, and preserve in the section all the substance and tone of each of the contributor acknowledgements and/or dedications given therein.

L. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles.

M. Delete any section Entitled “Endorsements”. Such a section may not be included in the Modified Version.

N. Do not retitle any existing section to be Entitled “Endorsements” or to conflict in title with any Invariant Section.

O. Preserve any Warranty Disclaimers.

If the Modified Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant.
To do this, add their titles to the list of Invariant Sections in the Modified Version’s license notice. These titles must be distinct from any other section titles.
You may add a section Entitled “Endorsements”, provided it contains nothing but endorsements of your Modified Version by various parties–for example, statements of peer review or that the text has been approved by an organization as
the authoritative definition of a standard.
You may add a passage of up to five words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover
Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may
not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one.
The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modified Version.

5. COMBINING DOCUMENTS
You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the
original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice, and that you preserve all their Warranty Disclaimers.
The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title
of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in
the license notice of the combined work.
In the combination, you must combine any sections Entitled “History” in the various original documents, forming one section Entitled “History”; likewise combine any sections Entitled “Acknowledgements”, and any sections Entitled
“Dedications”. You must delete all sections Entitled “Endorsements”.

6. COLLECTIONS OF DOCUMENTS
You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection,
provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects.
You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding
verbatim copying of that document.

7. AGGREGATION WITH INDEPENDENT WORKS

A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, is called an “aggregate” if the copyright resulting from the compilation
is not used to limit the legal rights of the compilation’s users beyond what the individual works permit. When the Document is included in an aggregate, this License does not apply to the other works in the aggregate which are not themselves
derivative works of the Document.
If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one half of the entire aggregate, the Document’s Cover Texts may be placed on covers that bracket the Document
within the aggregate, or the electronic equivalent of covers if the Document is in electronic form. Otherwise they must appear on printed covers that bracket the whole aggregate.

8. TRANSLATION
Translation is considered a kind of modification, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but
you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License, and all the license notices in the Document, and any Warranty Disclaimers,
provided that you also include the original English version of this License and the original versions of those notices and disclaimers. In case of a disagreement between the translation and the original version of this License or a notice or disclaimer,
the original version will prevail.
If a section in the Document is Entitled “Acknowledgements”, “Dedications”, or “History”, the requirement (section 4) to Preserve its Title (section 1) will typically require changing the actual title.

9. TERMINATION
You may not copy, modify, sublicense, or distribute the Document except as expressly provided for under this License. Any other attempt to copy, modify, sublicense or distribute the Document is void, and will automatically terminate your
rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance.

10. FUTURE REVISIONS OF THIS LICENSE

The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems
or concerns. See http://www.gnu.org/copyleft/.

Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License “or any later version” applies to it, you have the option of following the terms and conditions

either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as

a draft) by the Free Software Foundation.

Que a quien robe este libro, o lo tome prestado y no lo devuelva, se le convierta en una serpiente en las
manos y lo venza. Que sea golpeado por la parálisis y todos sus miembros arruinados. Que languidezca
de dolor gritando por piedad, y que no haya coto a su agonía hasta la última disolución. Que las polillas
roan sus entrañas y, cuando llegue al final de su castigo, que arda en las llamas del Infierno para siempre.
-Maldición anónima contra los ladrones de libros en el monasterio de San Pedro, Barcelona.
Contents

GNU Free Documentation License iii

1. APPLICABILITY AND DEFINITIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
2. VERBATIM COPYING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
3. COPYING IN QUANTITY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
4. MODIFICATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
5. COMBINING DOCUMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
6. COLLECTIONS OF DOCUMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
7. AGGREGATION WITH INDEPENDENT WORKS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
8. TRANSLATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
9. TERMINATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
10. FUTURE REVISIONS OF THIS LICENSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

Preface viii

1 Preliminaries 1
1.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Numerical Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Injective and Surjective Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Algebra of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Inverse Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.4 Inverse Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Groups and Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Addition and Multiplication in R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Order Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6.1 Absolute Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7 Classical Inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7.1 Triangle Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7.2 Bernoulli’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.7.3 Rearrangement Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.7.4 Arithmetic Mean-Geometric Mean Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.7.5 Cauchy-Bunyakovsky-Schwarz Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.7.6 Minkowski’s Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.8 Completeness Axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.8.1 Greatest Integer Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

v
vi CONTENTS

2 Topology of R 34
2.1 Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.2 Dense Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3 Open and Closed Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Interior, Boundary, and Closure of a Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.5 Connected Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.6 Compact Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7 R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.8 Lebesgue Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.9 The Cantor Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3 Sequences 47
3.1 Limit of a Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Convergence of Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Classical Limits of Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4 Averages of Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5 Orders of Infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.6 Cauchy Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.7 Topology of sequences. Limit Superior and Limit Inferior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4 Series 70
4.1 Convergence and Divergence of Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2 Convergence and Divergence of Series of Positive Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3 Summation by Parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4 Alternating Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.5 Absolute Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

5 Real Functions of One Real Variable 81

5.1 Limits of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.3 Algebraic Operations with Continuous Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.4 Monotonicity and Inverse Image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.5 Convex Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.5.1 Graphs of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.6 Classical Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.1 Affine Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.2 Quadratic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.3 Polynomial Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.4 Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.5 Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6.6 Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.6.7 Inverse Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.7 Continuity of Some Standard Functions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.7.1 Continuity Polynomial Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
CONTENTS vii

5.7.2 Continuity of the Exponential and Logarithmic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.7.3 Continuity of the Power Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.8 Inequalities Obtained by Continuity Arguments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.9 Intermediate Value Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.10 Variation of a Function and Uniform Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.11 Classical Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

6 Differentiable Functions 109

6.1 Derivative at a Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2 Differentiation Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.3 Rolle’s Theorem and the Mean Value Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.4 Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.5 Convex Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.6 Inequalities Obtained Through Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.7 Asymptotic Preponderance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.8 Asymptotic Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.9 Asymptotic Expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

7 Integrable Functions 139

7.1 The Area Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
7.2 Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.3 Riemann-Stieltjes Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.4 Euler’s Summation Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

8 Sequences and Series of Functions 156

8.1 Pointwise Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8.2 Uniform Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
8.3 Integrals and Derivatives of Sequences of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.4 Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
8.5 Maclaurin Expansions to know by inspection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.6 Comparison Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.7 Taylor Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
8.8 Abel’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
Homework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

A Answers and Hints 159

Answers and Hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Preface

For many years I have been lucky enough to have students ask for more: more challenging problems, more illuminating
proofs to different theorems, a deeper look at various topics, etc. To those students I normally recommend the books in
the bibliography. Some of the same students have complained of not finding the books or wanting to buy them, but being
impecunious, not being able to afford to buy them. Hence I have decided to make this compilation.

Here we take a semi-rigorous tour through Calculus. We don’t construct the real numbers, but we examine closer the
real number axioms and some of the basic theorems of Calculus. We also consider some Olympiad-level problems whose
solution can be obtained through Calculus.

The reader is assumed to be familiar with proofs using mathematical induction, proofs by contradiction, and the me-
chanics of differentiation and integration.
David A. SANTOS
[email protected]

viii
Chapter 1

Preliminaries

Why bother? We will use the language of set theory throughout these notes. There are various elementary
results that pop up in later proofs, among them, the De Morgan Laws and the Monotonicity Reversing of Com-
plementation Rule.
The concept of a function lies at the core of mathematics. We will give a brief overview here of some basic
properties of functions.

1.1 Sets
This section contains some of the set notation to be used throughout these notes. The one-directional arrow =⇒ reads
“implies” and the two-directional arrow ⇐⇒ reads “if and only if.”

1 Definition We will accept the notion of set as a primitive notion, that is, a notion that cannot be defined in terms of more
elementary notions. By a set we will understand a well-defined collection of objects, which we will call the elements of the
set. If the element x belongs to the set S we will write x ∈ S, and in the contrary case we will write x 6∈ S.1 The cardinality of
a set is the number of elements the set has. It can either be finite or infinite. We will denote the cardinality of the set S by
card(S).

☞ Some sets are used so often that merit special notation. We will denote by

N = {0,1,2,3,...}

the set of natural numbers, by

Z = {... ,−3,−2,−1,0,1, 2, 3, . ..}2
3
by Q the set of rational numbers , by R the real numbers, and by C the set of complex numbers. We will occasionally also use
αZ = {... ,−3α,−2α,−α,0, α, 2α, 3α,. ..}, etc.

We will also denote the empty set, that is, the set having no elements by ∅.

2 Definition The union of two sets A and B is the set

A ∪ B = {x : (x ∈ A) or (x ∈ B )}.

This is read “A union B .” See figure 1.1. The intersection of two sets A and B is

A ∩ B = {x : (x ∈ A) and (x ∈ B )}.
1
Georg Cantor(1845-1918), the creator of set theory, said “A set is any collection into a whole of definite, distinguishable objects, called elements, of our
intuition or thought.”
2
Z for the German word Zählen meaning “integer.”
3
Q for “quotients.”

1
 Sets

A B A B A B

Figure 1.1: A ∪ B Figure 1.2: A ∩ B Figure 1.3: A \ B

This is read “A intersection B .” See figure 1.2. The set difference of two sets A and B is

A \ B = {x : (x ∈ A) and(x 6∈ B )}.

This is read “A set minus B .” See figure 1.3.

3 Definition Two sets A and B are disjoint if A ∩ B = ∅.

4 Example Write A ∪ B as the disjoint union of three sets.

Solution: Observe that

A ∪ B = (A \ B ) ∪ (A ∩ B ) ∪ (B \ A),

and that the sets on the dextral side are disjoint.

5 Definition A subset B of a set A is a subcollection of A, and we denote this by B j A. 4 This means that x ∈ B =⇒ x ∈ A.

☞∅ and A are always subsets of any set A.

Observe that
A = B ⇐⇒ (A ⊆ B ) and (B ⊆ A).

We use this observation on the next theorem.

6 T HEOREM (De Morgan Laws) Let A,B ,C be sets. Then

A \ (B ∩ C ) = (A \ B ) ∪ (A \ C ), A \ (B ∪ C ) = (A \ B ) ∩ (A \ C ).

Proof: We have

x ∈ A \ (B ∪ C ) ⇐⇒ x ∈ A and x 6∈ (B or C )
⇐⇒ (x ∈ A) and ((x 6∈ B ) and (x 6∈ C ))
⇐⇒ (x ∈ A and x 6∈ B ) and (x ∈ A and x 6∈ C )
⇐⇒ (x ∈ A \ B ) and (x ∈ A \ C )
⇐⇒ x ∈ (A \ B ) ∩ (A \ C ).

Also,
x ∈ A \ (B ∩ C ) ⇐⇒ x ∈ A and x 6∈ (B and C )
⇐⇒ (x ∈ A) and ((x 6∈ B ) or (x 6∈ C ))
⇐⇒ (x ∈ A and x 6∈ B ) or (x ∈ A and x 6∈ C )
⇐⇒ (x ∈ A \ B ) or (x ∈ A \ C )
⇐⇒ x ∈ (A \ B ) ∪ (A \ C )

❑
4
There seems not to be an agreement here by authors. Some use the notation ⊂ or ⊆ instead of j. Some see in the notation ⊂ the exclusion of equality.
In these notes, we will always use the notation j, and if we wished to exclude equality we will write á.

2
 Chapter 1

7 T HEOREM (Monotonicity Reversing of Complementation) Let A,B , X be sets. Then

A j B ⇐⇒ X \ B j X \ A.

Proof: We have
AjB ⇐⇒ (x ∈ A) =⇒ (x ∈ B )
⇐⇒ (x 6∈ B ) =⇒ (x 6∈ A)
⇐⇒ (x ∈ X and x 6∈ B ) =⇒ (x ∈ X and x 6∈ A)
⇐⇒ X \ B j X \ A.
❑

8 Definition Let A 1 , A 2 ,... , A n , be sets. The Cartesian Product of these n sets is defined and denoted by

A 1 × A 2 × · · · × A n = {(a 1 , a 2 ,... , a n ) : a k ∈ A k },

that is, the set of all ordered n-tuples whose elements belong to the given sets.

☞ In the particular case when all the A k are equal to a set A, we write

A1 × A2 × · · · × An = An .

If a ∈ A and b ∈ A we write (a,b) ∈ A 2 .

p p
9 Example The Cartesian product is not necessarily commutative. For example, ( 2,1) ∈ R × Z but ( 2,1) 6∈ Z × R. Since
R × Z has an element that Z × R does not, R × Z 6= Z × R.

10 Example Prove that if X × X = Y × Y then X = Y .

Solution: Let x ∈ X . Then (x, x) ∈ X × X , which gives (x, x) ∈ Y × Y , so y ∈ Y . Hence X ⊆ Y .

Similarly, if y ∈ Y then (y, y) ∈ Y × Y , which gives (y, y) ∈ X × X , so y ∈ X . Hence Y ⊆ X .

Thus X ⊆ Y and Y ⊆ X gives X = Y .

Homework

Problem 1.1.1 For a fixed n ∈ N put A n = {nk : k ∈ N}. Problem 1.1.5 Prove the following associative laws:
1. Find A 2 ∩ A 3 .
A ∩ (B ∩C ) = (A ∩ B ) ∩C , A ∪ (B ∪C ) = (A ∪ B ) ∪C .
\∞
2. Find An .
n=1
Problem 1.1.6 Prove that
∞
[
3. Find An .
n=1 A ∩ B = A ⇐⇒ A ⊆ B .

Problem 1.1.2 Prove the following properties of the empty set: Problem 1.1.7 Prove that

A ∩ ∅ = ∅, A ∪ ∅ = A. A ∪ B = A ⇐⇒ B ⊆ A.

Problem 1.1.3 Prove the following commutative laws: Problem 1.1.8 Prove that
A ∩ B = B ∩ A, A ∪ B = B ∪ A. A ⊆ B =⇒ A ∩C ⊆ B ∩C .

Problem 1.1.4 Prove by means of set inclusion the following dis- Problem 1.1.9 Prove that
tributive law:
(A ∪ B ) ∩C = (A ∩C ) ∪ (B ∩C ). A ⊆B and C ⊆ B =⇒ A ∪C ⊆ B .

3
 Numerical Functions

Problem 1.1.10 Prove the following distributive laws: Problem 1.1.13 Let A,B , and C be sets. Shew that
A ∩ (B ∪C ) = (A ∩ B ) ∪ (A ∩C ), A ∪ (B ∩C ) = (A ∪ B ) ∩ (A ∪C ).

A × (B \ C ) = (A × B ) \ (A ×C ).
Problem 1.1.11 Is there any difference between the sets ∅, {∅} and
{{∅}}? Explain.
Problem 1.1.14 Prove that a set with N ∈ N elements has exactly 2N
Problem 1.1.12 Is the Cartesian product associative? Explain. subsets.

1.2 Numerical Functions

¡ ¢ ¡ ¢
11 Definition By a (numerical) function f : Dom f → Target f we mean the collection of the following ingredients:

➊ a name for the function. Usually we use the letter f .

¡ ¢
➋ a set of real number inputs called the domain of the function. The domain of f is denoted by Dom f j R.

➌ an input parameter , also called independent variable or dummy variable. We usually denote a typical input by the
letter x.

➍ a set of possible
¡ ¢ real number outputs of the function, called the target set of the function. The target set of f is denoted
by Target f j R.

➎ an assignment rule or formula, assigning to every input a unique output. This assignment rule for f is usually de-
noted by x 7→ f (x). The output of x under f is also referred to as the image of x under f , and is denoted by f (x).

The notation5 ¡ ¢ ¡ ¢
Dom f
→ Target f
f :
7 → x f (x)
¡ ¢ ¡ ¢
read “the function f , with domain Dom f , target set Target f , and assignment rule f mapping x to f (x)” conveys all
the above ingredients.

☞ Oftentimes we will only need to mention the assignment rule of a function, without mentioning its domain or target set.
In such instances we will sloppily say “the function f ” or more commonly, “the function x 7→ f (x)”, e.g., the square function
x 7→ x 2 .6

¡ ¢
12 Definition The image Im f of a function f is its set of actual outputs. In other words,
¡ ¢ ¡ ¢
Im f = { f (a) : a ∈ Dom f }.
¡ ¢ ¡ ¢
Observe that we always have Im f ⊆ Target f . For a set A, we also define

f (A) = { f (a) : a ∈ A}.

13 T HEOREM Let f : X → Y be a function and let A j X , A 0 j X . Then

1. A j A 0 =⇒ f (A) j f (A 0 )

2. f (A ∪ A 0 ) = f (A) ∪ f (A 0 )

3. f (A ∩ A 0 ) j f (A) ∩ f (A 0 )

4. f (A) \ f (A 0 ) j f (A \ A 0 )

Proof:
5
Notice the difference in the arrows. The straight arrow −→ is used to mean that a certain set is associated with another set, whereas the arrow 7→ (read
“maps to”) is used to denote that an input becomes a certain output.
6
This corresponds to the even sloppier American usage “the function f (x) = x 2 .”

4
 Chapter 1

1. x ∈ A =⇒ x ∈ A 0 and hence f (x) ∈ f (A) =⇒ f (x) ∈ f (A 0 ) =⇒ f (A) j f (A 0 )

2. Since A j A ∪ A 0 and A 0 j A ∪ A 0 , we have f (A) j f (A ∪ A 0 ) and f (A 0 ) j f (A ∪ A 0 ), by part (1) and thus
f (A) j f (A 0 ) j f (A ∪ A 0 ). Moreover, if y ∈ f (A ∪ A 0 ), then ∃x ∈ A ∪ A 0 such that y = f (x). Then either x ∈ A
and so f (x) ∈ f (A) or x ∈ A 0 and so f x ∈ f (A 0 ). Either way, f (x) ∈ f (A) ∪ f (A 0 ) and

y ∈ f (A ∪ A 0 ) =⇒ y ∈ f (A) ∪ f (A 0 ) =⇒ f (A ∪ A 0 ) j f (A) ∪ f (A 0 ).

Hence
f (A ∪ A 0 ) = f (A) ∪ f (A 0 ).
3. Let y ∈ f (A ∩ A 0 ). Then ∃x ∈ A ∩ A 0 such that f (x) = y. Thus we have both x ∈ A =⇒ f (x) ∈ f (A) and
x ∈ A 0 =⇒ f (x) ∈ f (A 0 ). Therefore f (x) ∈ f (A) ∩ f (A 0 ) and we conclude that f (A ∩ A 0 ) j f (A) ∩ f (A 0 ).
4. Let y ∈ f (A) \ f (A 0 ). Then y ∈ f (A) and y ∉ f (A 0 ). Thus ∃x ∈ A such that f (x) = y. Since y ∉ f (A 0 ), then
x ∉ A 0 . Therefore x ∈ A \ A 0 and finally, y ∈ f (A \ A 0 ). This means that f (A) \ f (A 0 ) j f (A \ A 0 ) as claimed.

1.2.1 Injective and Surjective Functions

14 Definition A function is injective or one-to-one whenever two different values of its domain generate two different values
in its image. A function is surjective or onto if every element of its target set is hit, that is, the target set is the same as the
image of the function. A function is bijective if it is both injective and surjective.

15 Example The function

R → R
a:
x 7→ x2
is neither injective nor surjective.

The function h h
R → 0 ;+∞
b:
x 7→ x2
is surjective but not injective.

The function h h
0 ;+∞ → R
c:
x 7→ x2
is injective but not surjective.

The function h h h h
0 ;+∞ → 0 ;+∞
d:
x 7→ x2
is a bijection.

A bijection between two sets essentially tells us that the two sets have the same size. We will make this statement more
precise now for finite sets.

16 T HEOREM Let f : A → B be a function, and let A and B be finite. If f is injective, then card(A) ≤ card(B ). If f is surjective
then card(B ) ≤ card(A). If f is bijective, then card(A) = card(B ).

Proof: Put n = card(A), A = {x1 , x2 ,... , xn } and m = card(B ), B = {y 1 , y 2 ,... , y m }.

If f were injective then f (x1 ), f (x2 ),... , f (xn ) are all distinct, and among the y k . Hence n ≤ m.

If f were surjective then each y k is hit, and for each, there is an xi with f (xi ) = y k . Thus there are at least m
different images, and so n ≥ m. ❑

5
 Numerical Functions

1.2.2 Algebra of Functions

¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢
17 Definition Let f : Dom f → Target f and g : Dom g → Target g . Then Dom f ± g = Dom f ∩ Dom g and the
sum (respectively, difference) function f + g (respectively, f − g ) is given by
¡ ¢ ¡ ¢ ¡ ¢
Dom f ∩ Dom g → Target f ± g
f ±g : .
x 7→ f (x) ± g (x)

In other words, if x belongs both to the domain of f and g , then

( f ± g )(x) = f (x) ± g (x).

¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢
18 Definition Let f : Dom f → Target f and g : Dom g → Target g . Then Dom f g = Dom f ∩ Dom g and the
product function f g is given by ¡ ¢ ¡ ¢ ¡ ¢
Dom f ∩ Dom g → Target f g
fg: .
x 7 → f (x) · g (x)
In other words, if x belongs both to the domain of f and g , then

( f g )(x) = f (x) · g (x).

¡ ¢ ¡ ¢ ¡ ¢
19 Definition
¡ ¢ Let g : Dom g → Target g be a function. The support of g , denoted by supp g is the set of elements in
Dom g where g does not vanish, that is
¡ ¢ ¡ ¢
supp g = {x ∈ Dom g : g (x) 6= 0}.

µ ¶
¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ f ¡ ¢ ¡ ¢
20 Definition Let f : Dom f → Target f and g : Dom g → Target f . Then Dom = Dom f ∩ supp g and the
g
f
quotient function is given by
g ¡ ¢ ¡ ¢ ¡ ¢
Dom f ∩ supp g → Target f /g
f
: f (x) .
g x 7→
g (x)
f f (x)
In other words, if x belongs both to the domain of f and g and g (x) 6= 0, then (x) = .
g g (x)

¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢
21 Definition Let f : Dom f → Target f , g : Dom g → Target g and let U = {x ∈ Dom g : g (x) ∈ Dom f }. We define
the composition function of f and g as
¡ ¢
U → Target f ◦ g
f ◦g : . (1.1)
x 7 → f (g (x))
We read f ◦ g as “ f composed with g .”

1.2.3 Inverse Image

22 Definition Let X and Y be subsets of R and let f : X → Y be a function. Let B j Y . The inverse image of B by f is the set

f −1 (B ) = {x ∈ X : f (x) ∈ B }.

If B = {b} consists of only one element, we write, abusing notation, f −1 ({b}) = f −1 (b). It is clear that f −1 (Y ) = X and
f −1 (∅) = ∅.

23 Example Let
{−2,−1,0,1,3} → {0,1,4,5,9}
f : .
x 7 → x2

Then f −1 ({0,1}) = {0,−1,1}, f −1 (1) = {−1,1}, f −1 (5) = ∅, f −1 (4) = 2, f −1 (0) = 0, etc. Notice that we have abused notation
in all but the first example.

6
 Chapter 1

24 T HEOREM Let f : X → Y be a function and let B j Y , B 0 j Y . Then

1. B j B 0 =⇒ f −1 (B ) j f −1 (B 0 )

2. f −1 (B ∪ B 0 ) = f −1 (B ) ∪ f −1 (B 0 )

3. f −1 (B ∩ B 0 ) = f −1 (B ) ∩ f −1 (B 0 )

4. f −1 (B ) \ f (B 0 ) = f −1 (B \ B 0 )

Proof:

1. Assume x ∈ f −1 (B ). Then there is y ∈ B j B 0 such that f (x) = y. But y is also in B 0 so x ∈ f −1 (B 0 ). Thus

f −1 (B ) j f −1 (B 0 ).
2. Since B j B ∪ B 0 and B 0 j B ∪ B 0 , we have f −1 (B ) j f −1 (B ∪ B 0 ) and f −1 (B 0 ) j f −1 (B ∪ B 0 ), by part (1).
Thus f −1 (B )∪ f −1 (B 0 ) j f −1 (B ∪B 0 ). Now, let x ∈ f −1 (B ∪B 0 ). There is y ∈ B ∪B 0 such that f (x) = y. Either
y ∈ B and so y ∈ B =⇒ x ∈ f −1 (B ) or y ∈ B 0 and so y ∈ B =⇒ x ∈ f −1 (B 0 ). Either way, x ∈ f −1 (B )∪ f −1 (B 0 ).
Thus f −1 (B ∪ B 0 ) j f −1 (B ) ∪ f −1 (B 0 ). We conclude that f −1 (B ∪ B 0 ) = f −1 (B ) ∪ f −1 (B 0 ).
3. Let x ∈ f −1 (B ∩ B 0 ). Then ∃y ∈ B ∩ B 0 such that f (x) = y. Thus we have both y ∈ B =⇒ x ∈ f −1 (B ) and y ∈
B 0 =⇒ x ∈ f −1 (B 0 ). Therefore x ∈ f −1 (B ) ∩ f −1 (B 0 ) and we conclude that f −1 (B ∩ B 0 ) j f −1 (B ) ∩ f −1 (B 0 ).
Now, let x ∈ f −1 (B ) ∩ f −1 (B 0 ). Then x ∈ f −1 (B ) and x ∈ f −1 (B 0 ). Then f (x) ∈ B and f (x) ∈ B 0 . Thus
f (x) ∈ B ∩ B 0 and so x ∈ f −1 (B ∩ B 0 ). Hence f −1 (B ) ∩ f −1 (B 0 ) j f −1 (B ∩ B 0 ) also, and we conclude that
f −1 (B ) ∩ f −1 (B 0 ) = f −1 (B ∩ B 0 ).
4. Let x ∈ f −1 (B ) \ f −1 (B 0 ). Then x ∈ f −1 (B ) and x ∉ f −1 (B 0 ). Thus f (x) ∈ B and f (x) ∉ B 0 . Thus f (x) ∈
B \ B 0 and therefore x ∈ f −1 (B \ B 0 ), giving f −1 (B ) \ f −1 (B 0 ) j f −1 (B \ B 0 ). Now, let x ∈ f −1 (B \ B 0 ). Then
f (x) ∈ B \ B 0 , which means that f (x) ∈ B but f (x) ∉ B 0 . Thus x ∈ f −1 (B ) but x ∉ f −1 (B 0 ), which gives
x ∈ f −1 (B ) \ f −1 (B 0 ) and so f −1 (B \ B 0 ) j f −1 (B ) \ f −1 (B 0 ). This establishes the desired equality.

25 T HEOREM Let f : X → Y be a function. Let A × B j X × Y . Then

1. A j ( f −1 ◦ f )(A)

2. ( f ◦ f −1 )(B ) j B

Proof: We have

1. Let x ∈ A. Then ∃y ∈ Y such that y = f (x). Thus y ∈ f (A). Therefore x ∈ f −1 ( f (A)).

2. y ∈ ( f ◦ f −1 )(B ). Then ∃x ∈ f −1 (B ) such that f (x) = y. Thus x ∈ f −1 (y). Hence f (x) ∈ B . Therefore y ∈ B .

1.2.4 Inverse Function

26 Definition Let A × B j R2 . A function F : A → B is said to be invertible if there exists a function F −1 (called the inverse of
F ) such that F ◦ F −1 = Id B and F −1 ◦ F = Id A . Here Id S is the identity on the set S function with rule Id S (x) = x.

The central question is now: given a function F : A → B , when is F −1 : B → A a function? The answer is given in the next
theorem.

27 T HEOREM Let A×B j R2 . A function f : A → B is invertible if and only if it is a bijection. That is, f −1 : B → A is a function
if and only if f is bijective.

7
 Numerical Functions

Proof: Assume first that f is invertible. Then there is a function f −1 : B → A such that

f ◦ f −1 = Id B and f −1 ◦ f = Id A . (1.2)

Let us prove that f is injective and surjective. Let s, t be in the domain of f and such that f (s) = f (t). Applying
f −1 to both sides of this equality we get ( f −1 ◦ f )(s) = ( f −1 ◦ f )(t). By the definition of inverse function, ( f −1 ◦
f )(s) = s and ( f −1 ◦ f )(t) = t. Thus s = t . Hence f (s) = f (t) =⇒ s = t implying that f is injective. To prove that
f is surjective we must shew that for every b ∈ f (A) ∃a ∈ A such that f (a) = b. We take a = f −1 (b) (observe that
f −1 (b) ∈ A). Then f (a) = f ( f −1 (b)) = ( f ◦ f −1 )(b) = b by definition of inverse function. This shews that f is
surjective. We conclude that if f is invertible then it is also a bijection.

Assume now that f is a bijection. For every b ∈ B there exists a unique a such that f (a) = b. This makes the rule
g : B → A given by g (b) = a a function. It is clear that g ◦ f = Id A and f ◦ g = Id B . We may thus take f −1 = g .
This concludes the proof. ❑

Homework

Problem 1.2.1 Find all functions with domain {a,b} and target set Problem 1.2.7 Let f satisfy f (n + 1) = (−1)n+1 n − 2f (n),n ≥ 1 If
{c,d }. f (1) = f (1001) find

f (1) + f (2) + f (3) + · · · + f (1000).

Problem 1.2.2 Let A, B be finite sets with card (A) = n and
card (B ) = m. Prove that
Problem 1.2.8 If f (a) f (b) = f (a + b) ∀ a,b ∈ R and f (x) > 0 ∀ x ∈
• The number of functions from A to B is m n .
R, find f (0). Also, find f (−a) and f (2a) in terms of f (a).
• If n ≤ m, the number of injective functions from A to B is
m(m −1)(m −2) · · · (m −n +1). If n > m there are no injective
functions from A to B . R \ {−1} → R \ {1}
Problem 1.2.9 Prove that f : x − 1 is a bijection
x 7→
x +1
Problem 1.2.3 Let A and B be two finite sets with card (A) = n and and find f . −1
card (B ) = m. If n < m prove that there are no surjections from A to
B . If n ≥ m prove that the number of surjective functions from A to
Problem 1.2.10 Let f [1] (x) = f (x) = x + 1, f [n+1] = f ◦ f [n] ,n ≥ 1.
B is [n]
à ! à ! à ! à ! Find a closed formula for f
m m m m
mn − (m−1)n + (m−2)n − (m−3)n +· · ·+(−1)m−1 (1)n . h i
1 2 3 m −1
Problem 1.2.11 Let f , g : 0 ;1 → R be functions. Demonstrate
h i2 1 ¯ ¯
Problem 1.2.4 Let h : R → R be given by h(1 − x) = 2x. Find h(3x). that there exist (a,b) ∈ 0 ;1 such that ≤ ¯ f (a) + g (b) − ab ¯ .
4

Problem 1.2.5 Consider the polynomial Problem 1.2.12 Demonstrate that there is no function f : R \
2 4 2003 2 8012 {1/2} → R such that
(1 − x + x ) = a0 + a1 x + a2 x + · · · + a8012 x .
µ µ ¶¶
x −1
Find x ∈ R \ {1/2} =⇒ f (x) f = x2 + x + 1
2x − 1
➊ a0
➋ a0 + a1 + a2 + · · · + a8012 Problem 1.2.13 Find all functions f : R \ {−1,0} → R such that
➌ a0 − a1 + a2 − a3 + · · · − a8011 + a8012 µ ¶
−1
➍ a0 + a2 + a4 + · · · + a8010 + a8012 x ∈ R \ {−1,0} =⇒ f (x) + f = 3x + 2.
x +1
➎ a1 + a3 + · · · + a8009 + a8011
Problem 1.2.14 Let f [1] (x) = f (x) = 2x, f [n+1] = f ◦ f [n] ,n ≥ 1.
Problem 1.2.6 Let f : R → R, be a function such that ∀x ∈]0;+∞[, Find a closed formula for f [n]
p
[ f (x 3 + 1)] x = 5,
Problem 1.2.15 Find all functions g : R → R that satisfy g (x + y ) +
find the value of q g (x − y ) = 2x 2 + 2y 2 .
" Ã !# 27
27 + y 3 y
f
y3 Problem 1.2.16 Find all the functions f : R → R that satisfy
for y ∈]0;+∞[. f (x y ) = y f (x).

8
 Chapter 1

1
Problem 1.2.17 Find all functions f : R \ {0} → R for which Problem 1.2.19 Let f [1] = f be given by f (x) = . Find
1−x
µ ¶
1 (i) f [2] (x) = ( f ◦ f )(x),
f (x) + 2f = x.
x (ii) f [3] (x) = ( f ◦ f ◦ f )(x), and
(iii) f [69] = (f ◦ f ◦··· f ◦ f ) (x).
| {z }
Problem 1.2.18 Find all functions f : R \ {−1} → R such that 69 compositions with itself
µ ¶
1−x
( f (x))2 · f = 64x. Problem 1.2.20 Let f : A → B and g : B → C be functions. Shew
1+x
that (i) if g ◦ f is injective, then f is injective. (ii) if g ◦ f is surjective,
then g is surjective.

1.3 Countability
28 Definition A set X is countable if either it is finite or if there is a bijection f : X → N, that is, the set X has as many
elements as N.

Any countable set can be thus enumerated a sequence

x1 , x2 , x3 ,... .

Thus the strictly positive integers can be enumerated as customarily:

1,2,3,... .

Another possible enumeration7 is the following

3,5,7,9,... , ,2 · 3,2 · 5,2 · 7,2 · 9,... ,22 · 3,22 · 5,22 · 7,22 · 9,... ,... 24 ,23 ,22 ,2,1,

that is, we start with the odd integers in increasing order, then 2 times the odd integers, 22 times the odd integers, etc., and
at the end we put the powers of 2 in decreasing order.

29 L EMMA Any subset X j N is countable.

Proof: If X is finite, then there is nothing to prove. If X is infinite, we can arrange the elements of X increasing
order, say,
x1 < x2 < x3 < · · · .

We then map the smallest element x1 ∈ S to 1, the next smallest x2 to 2, etc. ❑

☞ Hence, even though 2N á N, the sets 2N and N have the same number of elements. This can also be seen by noticing that
f : N → 2N given by xn = 2n is a bijection.

30 L EMMA A set X is countable if and only if there is an injection f : X → N.

Proof: The assertion is evident if X is finite. Hence assume X is infinite. If f : X → N is an injection then f (X ) is
an infinite subset of N. Hence there is a bijection g : f (X ) → N by virtue of Lemma 29. Thus (g ◦ f ) : X → N is a
bijection. ❑

☞ An obvious consequence of the above lemma is that if X 0 is countable and there is an injection f : X → X 0 then X is
countable.

31 T HEOREM Z is countable.
7
Which is relevant in chaos theory, for Sarkovkii’s Theorem.

9
 Groups and Fields

Proof: One can take, as a bijection between the two sets, for example, f : Z → N,
(
2x + 1 if x ≥ 0
f (x) =
−2x if x < 0.

32 T HEOREM Q is countable.

Proof: Consider f : Q → N given

³a ´
f = 2|a| 3b 51+signum(a) ,
b
a
where is in least terms, and b > 0. By the uniqueness of the prime factorisation of an integer, f is an injection.
b
❑

☞ The above theorem means that there as many rational numbers as natural numbers. Thus the rationals can be enumer-
ated as
q1 , q2 , q3 ,... ,

33 T HEOREM (Cantor’s Diagonal Argument) R is uncountable.

Proof: Assume R were countable so that its complete set of elements may be enumerated, say, as in the list

r 1 = n 1 .d 11 d 12 d 13 ...
r 2 = n 2 .d 21 d 22 d 13 ...
r 3 = n 3 .d 31 d 32 d 33 ... ,

where we have used decimal notation. Define the new real r = 0.d 1 d 2 d 3 ... by d i = 0 if d i i = 6 0 and d i = 1 if
d i i = 0. This is real number (as it is a decimal), but it differs from r i in the i th decimal place. It follows that the
list is incomplete and the reals are uncountable. ❑
i h
34 T HEOREM The interval − 1 ;1 is uncountable.

i h πx
Proof: Observe that the map f : − 1 ;1 → R given by f (x) = tan is a bijection. ❑
2

Homework

i h
Problem 1.3.1 Prove that there as many numbers in [0;1] as in any Problem 1.3.2 Prove that there as many numbers in −∞ ;+∞ as
i h
interval [a;b] with a < b.
in 0 ;+∞ .

1.4 Groups and Fields

Here we observe the rules of the game for the operations of addition and multiplication in R.

35 Definition Let S,T be sets. A binary operation is a function

S ×S → T
⊗: .
(a,b) 7 → ⊗(a,b)

We usually use the “infix” notation a ⊗ b rather than the “prefix” notation ⊗(a,b). If S = T then we say that the binary
operation is internal or closed and if S 6= T then we say that it is external.

10
 Chapter 1

36 Example Ordinary addition is a closed binary operation on the sets N, Z, Q, R. Ordinary subtraction is a binary operation
on these sets. It is not closed on N, since for example 1 − 2 = −1 6∈ N, but it is closed in the remaining sets.

37 Example The operation ⊗ : R × R → R given by a ⊗ b = 1 + a · b, where · is the ordinary multiplication of real numbers is
commutative but not associative. To see commutativity we have

a ⊗ b = 1 + ab = 1 + ba = b ⊗ a.

Now,

1 ⊗ (1 ⊗ 2) = 1 ⊗ (1 + 1 · 2) = 1 ⊗ (3) = 1 + 1 · 3 = 4, but (1 ⊗ 1) ⊗ 2 = (1 + 1 · 1) ⊗ 2 = 2 ⊗ 2 = 1 + 2 · 2 = 5,

so the operation is not associative.

38 Definition Let G be a non-empty set and ⊗ be a binary operation on G × G. Then 〈G,⊗〉 is called a group if the following
axioms hold:

G1: ⊗ is closed, that is,

∀(a,b) ∈ G 2 , a ⊗ b ∈ G,

G2: ⊗ is associative, that is,

∀(a,b,c) ∈ G 3 , a ⊗ (b ⊗ c) = (a ⊗ b) ⊗ c,

G3: G has an identity element, that is

∃e ∈ G such that ∀a ∈ G, e ⊗ a = a ⊗ e = a,

G4: Every element of G is invertible, that is

∀a ∈ G, ∃a −1 ∈ G such that a ⊗ a −1 = a −1 ⊗ a = e.

☞ From now on, we drop the sign ⊗ and rather use juxtaposition for the underlying binary operation in a given group.
Thus we will say a “group G” rather than the more precise “a group 〈G,⊗〉.”

39 Definition A group G is abelian if its binary operation is commutative, that is, ∀(a,b) ∈ G 2 , a ⊗ b = b ⊗ a.

40 Example 〈Z,+〉, 〈Q,+〉, 〈R,+〉, 〈C,+〉 are all abelian groups under addition. The identity element is 0 and the inverse of
a is −a.

41 Example 〈Q \ {0},·〉, 〈R \ {0},·〉, 〈C \ {0},·〉 are all abelian groups under multiplication. The identity element is 1 and the
1
inverse of a is .
a

42 Example 〈Z \ {0},·〉 is not a group. For example the element 2 does not have a multiplicative inverse.

43 Example Let V4 = {e, a,b,c} and define ⊗ by the table below.

⊗ e a b c
e e a b c
a a e c b
b b c e a
c c b a e

It is an easy exercise to check that V4 is an abelian group, called the Klein Viergruppe.

44 T HEOREM Let G be a group. Then

11
 Groups and Fields

1. There is only one identity element, the identity element is unique.

2. The inverse of each element is unique.

3. ∀(a,b) ∈ G 2 we have
(ab)−1 = b −1 a −1 .

Proof:

1. Let e and e 0 be identity elements. Since e is an identity, e = ee 0 . Since e 0 is an identity, e 0 = ee 0 . This gives
e = ee 0 = e 0 .
2. Let b and b 0 be inverses of a. Then e = ab and b 0 a = e. This gives

b = eb = (b 0 a)b = b 0 (ab) = b 0 e = b 0 .

3. We have
(ab)(b −1 a −1 ) = a(bb −1 )a −1 = a(e)a −1 = aa −1 = e.
Thus b −1 a −1 works as a right inverse for ab. A similar calculation shews also that it works as a left inverse.
Since inverses are unique, we must have
(ab)−1 = b −1 a −1 .

This completes the proof. ❑

45 Definition Let n ∈ Z and let G be a group. If a ∈ G, we define

a 0 = e,

a |n| = |a · a{z· · · a},

|n| a 0 s

and
a −|n| = |a −1 · a −1 −1
{z · · · a } .
|n| a −10 s

☞ If (m,n) ∈ Z2 , then by associativity

(a n )(a m ) = (a m )(a n ) = a m+n .

46 Definition Let F be a set having at least two elements 0F and 1F (0F 6= 1F ) together with two binary operations · (field
multiplication) and + (field addition). A field 〈F,·,+〉 is a triplet such that 〈F,+〉 is an abelian group with identity 0F ,
〈F \ {0F },·〉 is an abelian group with identity 1F and the operations · and + satisfy

a · (b + c) = (a · b) + (a · c),

that is, field multiplication distributes over field addition.

☞ We will continue our practice of denoting multiplication by juxtaposition, hence the · sign will be dropped.

47 Example 〈Q,·,+〉, 〈R,·,+〉, and 〈C,·,+〉 are all fields. The multiplicative identity in each case is 1 and the additive identity
is 0.

Homework

12
 Chapter 1

Problem 1.4.1 Is the set of real irrational numbers closed under ad- where juxtaposition means ordinary multiplication and + is the or-
dition? Under multiplication? dinary addition of real numbers. Prove that 〈Q∩] − 1;1[,⊗〉 is an
abelian group by following these steps.
Problem 1.4.2 Let 1. Prove that ⊗ is a closed binary operation on Q∩] − 1;1[.

S = {x ∈ Z : ∃(a,b) ∈ Z2 , x = a 3 + b 3 + c 3 − 3abc}.
2. Prove that ⊗ is both commutative and associative.
Prove that S is closed under multiplication, that is, if x ∈ S and y ∈ S
then x y ∈ S. 3. Find an element e ∈ Q∩]−1;1[ such that (∀a ∈ Q∩]−1;1[) (e⊗
a = a).
Problem 1.4.3 (Putnam, 1971) Let S be a set and let ◦ be a binary
operation on S satisfying the two laws 4. Given e as above and an arbitrary element a ∈ Q∩] − 1;1[,
solve the equation a ⊗ b = e for b.
(∀x ∈ S)(x ◦ x = x),

and
(∀(x, y, z) ∈ S 3 )((x ◦ y ) ◦ z = (y ◦ z) ◦ x). Problem 1.4.6 Let G be a group satisfying (∀a ∈ G)
Shew that ◦ is commutative.
a 2 = e.

Problem 1.4.4 (Putnam, 1972) Let S be a set and let ∗ be a binary Prove that G is an abelian group.
operation of S satisfying the laws ∀(x, y ) ∈ S 2
Problem 1.4.7 Let G be a group where (∀(a,b) ∈ G 2 )
x ∗ (x ∗ y ) = y, (1.3)
((ab)3 = a 3 b 3 ) and ((ab)5 = a 5 b 5 ).

(y ∗ x) ∗ x = y. (1.4) Shew that G is abelian.

Shew that ∗ is commutative, but not necessarily associative.
Problem 1.4.8 Suppose that in a group G there exists a pair (a,b) ∈
Problem 1.4.5 On Q∩] − 1;1[ define the binary operation ⊗ by G 2 satisfying
(ab)k = a k b k
a +b
a ⊗b = , for three consecutive integers k = i ,i + 1,i + 2. Prove that ab = ba.
1 + ab

1.5 Addition and Multiplication in R

Since R is a field, it satisfies the following list of axioms, which we list for future reference.

48 Axiom (Arithmetical Axioms of R) 〈R,·,+〉—that is, the set of real numbers endowed with multiplication · and addition
+—is a field. This entails that + and · verify the following properties.
R1: + and · are closed binary operations, that is,

∀(a,b) ∈ R2 , a + b ∈ R, a · b ∈ R,

R2: + and · are associative binary operations, that is,

∀(a,b,c) ∈ R3 , a + (b + c) = (a + b) + c, a · (b · c) = (a · b) · c

R3: + and · are commutative binary operations, that is,

∀(a,b) ∈ R2 , a + b = b + a, a · b = b · a,

R4: R has an additive identity element 0, and a multiplicative identity element 1, with 0 6= 1, such that

∀a ∈ R, 0 + a = a + 0 = a, 1 · a = a · 1 = a,

R5: Every element of R has an additive inverse, and every element of R \ {0} has a multiplicative inverse, that is,

∀a ∈ R, ∃(−a) ∈ R such that a + (−a) = (−a) + a = 0,

∀b ∈ R \ {0}, ∃b −1 ∈ R \ {0} such that b · b −1 = b −1 · b = 1,

13
 Addition and Multiplication in R

R6: + and · satisfy the following distributive law:

∀(a,b,c,) ∈ R3 , a · (b + c) = a · b + a · c.

Since + and · are associative in R, we may write a sum a 1 + a 2 + · · · + a n or a product a 1 a 2 · · · a n of real numbers without
risking ambiguity. We often use the following shortcut notation.

49 Definition For real numbers a i we define

X
n Y
n
a1 + a2 + · · · + an = ak and a1 a2 · · · an = ak .
k=1 k=1

☞ By convention
X

k∈∅
ak = 0 and
Y

k∈∅
ak = 1 .

50 T HEOREM (Lagrange’s Identity) Let a k ,b k be real numbers. Then

à !2 à !à !
X
n X
n X
n X
ak bk = a k2 bk2 − (a k b j − a j bk )2 .
k=1 k=1 k=1 1≤k< j ≤n

Proof: For j = k, a k b j − a j bk = 0, so we may relax the inequality in the last sum. We have
X X
(a k b j − a j bk )2 = (a k2 b 2j − 2a k bk a j b j + a 2j bk2 )
1≤k< j ≤n 1≤k≤
Xj ≤n 2 2 X X
= ak b j − 2 ak bk a j b j + a 2j bk2
1≤k≤ j ≤n 1≤k≤ j ≤n 1≤k≤ j ≤n
à !2
X
n X
n X
n
= a k2 b 2j − ak bk ,
k=1 j =1 k=1

proving the theorem. ❑

Recall that the factorial symbol ! is defined by
0! = 1; k! = k(k − 1)! if k ≥ 1.

à ! à !
n n
51 Definition (Binomial Coefficients) Let n ∈ N We define =1= and for 1 ≤ k ≤ n,
0 n
à !
n n!
= .
k k!(n − k)!
à !
n
If k > n we take = 0.
k

52 L EMMA (Pascal’s Identity) For n ≥ 1 and 1 ≤ k ≤ n,

à ! à ! à !
n n −1 n −1
= + .
k k k −1

Proof: We have à ! à !
n −1 n −1 (n − 1)! (n − 1)!
+ = +
k k −1 k!(n − 1 − k)! (k − 1)!(n − k)!
µ ¶
(n − 1)! 1 1
= +
(k − 1)!(n − 1 − k) µ k n − k¶
(n − 1)! n
=
(k − 1)!(n − 1Ã−!k) k(n − k)
n! n
= = .
k!(n − k)! k

14
 Chapter 1

Using Pascal’s Identity we obtain Pascal’s Triangle.

à !
0
0
à ! à !
1 1
0 1
à ! à ! à !
2 2 2
0 1 2
à ! à ! à ! à !
3 3 3 3
0 1 2 3
à ! à ! à ! à ! à !
4 4 4 4 4
0 1 2 3 4
à ! à ! à ! à ! à ! à !
5 5 5 5 5 5
0 1 2 3 4 5
.. .. .. .. .. ..
. . . . . .

When the numerical values are substituted, the triangle then looks like this.

1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
.. .. .. .. .. ..
. . . . . .

à !WeÃsee from
! Pascal’s Triangle that binomial coefficients are symmetric. This symmetry is easily justified by the identity
n n
= . We also notice that the binomial coefficients tend to increase until they reach the middle, and that then they
k n −k
decrease symmetrically.

53 T HEOREM (Binomial Theorem) For n ∈ N,

à !
Xn n
n
(x + y) = x k y n−k .
k=0 k

Proof: The theorem is obvious for n = 0 (defining (x + y)0 = 1), n = 1 (as (x + y)1 = xà +!y), and n = 2 (as
Xn n
(x + y)2 = x 2 + 2x y + y 2 ). Assume n ≥ 3. The induction hypothesis is that (x + y)n = x k y n−k . Then we
k=0 k

15
 Addition and Multiplication in R

have
(x + y)n+1 = (x + y)(x + y)n
à à ! !
Xn n
k n−k
= (x + y) x y
k
à ! k=0 à !
Xn n
k+1 n−k
Xn n
= x y + x k y n−k+1
k k
k=0 Ã ! k=0 Ã !
n+1
X n k+1 n−k X
n−1 n n
= x + x y + x k y n−k+1 + y n+1
k=0 Ã k k
! k=1 Ã !
n+1
Xn n k n−k+1
Xn n
= x + x y + x k y n−k+1 + y n+1
k=1 ÃÃ k − 1 k
! Ã !! k=1
n+1
Xn n n k n−k+1
= x + + x y + y n+1
k=1 Ã k − 1 k
!
n+1
Xn n +1
= x + x k y n−k+1 + y n+1
k
à k=1 !
X n + 1 k n−k+1
n+1
= x y ,
k=0 k
proving the theorem. ❑

54 L EMMA If a ∈ R, a 6= 1 and n ∈ N \ {0}, then

1 − an
1 + a + a 2 + · · · a n−1 = .
1−a

Proof: For, put S = 1 + a + a 2 + · · · + a n−1 . Then aS = a + a 2 + · · · + a n−1 + a n . Thus

S − aS = (1 + a + a 2 + · · · + a n−1 ) − (a + a 2 + · · · + a n−1 + a n ) = 1 − a n ,

and from (1 − a)S = S − aS = 1 − a n we obtain the result. ❑

55 T HEOREM Let n be a strictly positive integer. Then

y n − x n = (y − x)(y n−1 + y n−2 x + · · · + y x n−2 + x n−1 ).

x
Proof: By making the substitution a = in Lemma 54 we see that
y
³ ´n
µ ¶2 µ ¶n−1 1 − x
x x x y
1+ + +··· + =
y y y 1 − xy

we obtain µ ¶µ µ ¶2 µ ¶n−1 ¶ µ ¶n
x x x x x
1− 1+ + +··· + = 1− ,
y y y y y
or equivalently,
µ ¶µ ¶
x x x2 x n−1 xn
1− 1 + + 2 + · · · + n−1 = 1 − n .
y y y y y
Multiplying by y n both sides,
µ ¶ µ ¶ µ ¶
x n−1 x x2 x n−1 xn
y 1− y 1 + + 2 + · · · + n−1 = y n 1 − n ,
y y y y y

which is
y n − x n = (y − x)(y n−1 + y n−2 x + · · · + y x n−2 + x n−1 ),
yielding the result. ❑

16
 Chapter 1

n(n + 1)
56 T HEOREM 1 + 2 + · · · + n = .
2

First Proof: Observe that

k 2 − (k − 1)2 = 2k − 1.

From this
12 − 02 = 2·1−1
2 2
2 −1 = 2·2−1
32 − 22 = 2·3−1
.. .. ..
. . .
n 2 − (n − 1)2 = 2·n −1

Adding both columns,

n 2 − 02 = 2(1 + 2 + 3 + · · · + n) − n.

Solving for the sum,

n(n + 1)
1 + 2 + 3 + · · · + n = n 2 /2 + n/2 = .
2

Second Proof: We may utilise Gauss’ trick: If

An = 1 + 2 + 3 + · · · + n

then
A n = n + (n − 1) + · · · + 1.

Adding these two quantities,

An = 1 + 2 + ··· + n
An = n + (n − 1) + ··· + 1
2A n = (n + 1) + (n + 1) + ··· + (n + 1)
= n(n + 1),

n(n + 1)
since there are n summands. This gives A n = , that is,
2

n(n + 1)
1 +2 +··· +n = .
2

Applying Gauss’s trick to the general arithmetic sum

(a) + (a + d ) + (a + 2d ) + · · · + (a + (n − 1)d )

we obtain
n(2a + (n − 1)d )
(a) + (a + d ) + (a + 2d ) + · · · + (a + (n − 1)d ) = (1.5)
2
❑

n(n + 1)(2n + 1)
57 T HEOREM 12 + 22 + 32 + · · · + n 2 = .
6

Proof: Observe that

k 3 − (k − 1)3 = 3k 2 − 3k + 1.

17
 Addition and Multiplication in R

Hence
13 − 03 = 3 · 12 − 3 · 1 + 1
23 − 13 = 3 · 22 − 3 · 2 + 1
33 − 23 = 3 · 32 − 3 · 3 + 1
.. .. ..
. . .
n 3 − (n − 1)3 = 3 · n2 − 3 · n + 1

Adding both columns,

n 3 − 03 = 3(12 + 22 + 32 + · · · + n 2 ) − 3(1 + 2 + 3 + · · · + n) + n.

n(n + 1)
From the preceding example 1 + 2 + 3 + · · · + n = ·n 2 /2 + n/2 = so
2

3
n 3 − 03 = 3(12 + 22 + 32 + · · · + n 2 ) − · n(n + 1) + n.
2
Solving for the sum,
n3 1 n
12 + 22 + 32 + · · · + n 2 = + · n(n + 1) − .
3 2 3
After simplifying we obtain
n(n + 1)(2n + 1)
12 + 22 + 32 + · · · + n 2 = .
6
❑

Homework

Problem 1.5.1 Prove that for n ≥ 1, Problem 1.5.7 Prove that

à ! à ! à ! à ! à !
n n n n
X n k
X X n X n X n
2n = ; 0= (−1)k , 2n−1 = = . k p (1 − p)n−k = np.
k
k=0 k k=0 k 0≤k≤n k 1≤k≤n k k=1
k even k odd

Problem 1.5.8 Prove that

Problem 1.5.2 Given that 1002004008016032 has a prime factor à !
Xn n k
p > 250000, find it. k(k − 1) p (1 − p)n−k = n(n − 1)p 2 .
k=2 k

Problem 1.5.3 Prove that (a + b + c)2 = a 2 + b 2 + c 2 + 2ab + 2bc +

2ca. Problem 1.5.9 Demonstrate that
à !
n
X n k
(k − np)2 p (1 − p)n−k = np(1 − p).
Problem 1.5.4 Let a,b,c be real numbers. Prove that k
k=0

a 3 + b 3 + c 3 − 3abc = (a + b + c)(a 2 + b 2 + c 2 − ab − bc − ca).

Problem 1.5.10 Let x ∈ R \ {1} and let n ∈ N \ {0}. Prove that

Problem 1.5.5 Prove that n

X 2k 1 2n+1
= − n+1 .
à ! à ! k=0 x
2k +1 x −1 x 2 +1
n n n −1
= .
k k k −1
Problem 1.5.11 Consider the n k k-tuples (a1 , a2 ,... , ak ) which
can be formed by taking ai ∈ {1,2,... ,n}, repetitions allowed.
Problem 1.5.6 Prove that
Demonstrate that
à ! à !
n n n −1 n −2 X
= · · . min(a1 , a2 ,... , ak ) = 1k + 2k + · · · + n k .
k k k −1 k −2 a i ∈{1,2,...,n}

18
 Chapter 1

1.6 Order Axioms

Vocabulary Alert! We will call a number x positive if x ≥ 0 and strictly positive if x > 0. Similarly, we will call
☞ a number y negative if y ≤ 0 and strictly negative if y < 0. This usage differs from most Anglo-American books,
who prefer such terms as non-negative and non-positive.

We assume R endowed with a relation > which satisfies the following axioms.

58 Axiom (Trichotomy Law) ∀(x, y) ∈ R2 exactly one of the following holds:

x > y, x = y, or y > x.

59 Axiom (Transitivity of Order) ∀(x, y, z) ∈ R3 ,

if x > y and y >z then x > z.

60 Axiom (Preservation of Inequalities by Addition) ∀(x, y, z) ∈ R3 ,

if x > y then x + z > y + z.

61 Axiom (Preservation of Inequalities by Positive Factors) ∀(x, y, z) ∈ R3 ,

if x > y and z > 0 then xz > y z.

☞ x < y means that y > x. x ≤ y means that either y > x or y = x, etc.

62 T HEOREM The square of any real number is positive, that is, ∀a ∈ R, a 2 ≥ 0. In fact, if a 6= 0 then a 2 > 0.

Proof: If a = 0, then 02 = 0 and there is nothing to prove. Assume now that a 6= 0. By trichotomy, either a > 0 or
a < 0. Assume first that a > 0. Applying Axiom 61 with x = z = a and y = 0 we have

aa > a0 =⇒ a 2 > 0,

so the theorem is proved if a > 0.

If a < 0 then −a > 0 and we apply the result just obtained:

−a > 0 =⇒ (−a)2 > 0 =⇒ 1 · a 2 > 0 =⇒ a 2 > 0,

so the result is true regardless the sign of a. ❑

Theorem 62 will prove to be extremely powerful and will be the basis for many of the classical inequalities that follow.

63 T HEOREM If (x, y) ∈ R2 ,
x > y ⇐⇒ x − y > 0.

Proof: This is a direct consequence of Axiom 60 upon taking z = −y. ❑

64 T HEOREM If (x, y, a,b) ∈ R4 ,

x>y and a ≥ b =⇒ x + a > y + b.

Proof: We have
x > y =⇒ x + a > y + a, y + a ≥ y + b,
by Axiom 60 and so by Axiom 59 x + a > y + b. ❑

19
 Order Axioms

65 T HEOREM If (x, y, a,b) ∈ R4 ,

x > y > 0 and a ≥ b > 0 =⇒ xa > yb.

Proof: Indeed
x > y =⇒ xa > y a, y a ≥ yb,
by Axiom 61 and so by Axiom 59 xa > yb. ❑

66 T HEOREM 1 > 0.

Proof: By definition of R being a field 0 6= 1. Assume that 1 < 0 then 12 > 0 by Theorem 62. But 12 = 1 and so
1 > 0, a contradiction to our original assumption. ❑

67 T HEOREM x > 0 =⇒ −x < 0 and x −1 > 0.

Proof: Indeed, −1 < 0 since −1 6= 0 and assuming −1 > 0 would give 0 = −1 + 1 > 1, which contradicts Theorem
66. Thus
−x = −1 · x < 0.
Similarly, assuming x −1 < 0 would give 1 = x −1 x < 0. ❑

68 T HEOREM x > 1 =⇒ x −1 < 1.

Proof: Since x −1 6= 1, assuming x −1 > 1 would give 1 = xx −1 > 1 · 1 = 1, a contradiction. ❑

1.6.1 Absolute Value



 −1 if x < 0,
69 Definition (The Signum (Sign) Function) Let x be a real number. We define signum(x) = 0 if x = 0,


+1 if x > 0.

¡ ¢ ¡ ¢
70 L EMMA The signum function is multiplicative, that is, if (x, y) ∈ R2 then signum x · y = signum(x) signum y .

Proof: Immediate from the definition of signum. ❑

71 Definition (Absolute Value) Let x ∈ R. The absolute value of x is defined and denoted by

|x| = signum (x) x.

72 T HEOREM Let x ∈ R. Then

(
−x if x < 0,
1. |x| =
x if x ≥ 0.

2. |x| ≥ 0,

3. |x| = max(x,−x),

4. |−x| = |x|,

5. − |x| ≤ x ≤ |x|.
p
6. x 2 = |x|

7. |x|2 = |x 2 | = x 2

8. x = signum (x) |x|

20
 Chapter 1

Proof: These are immediate from the definition of |x|. ❑

73 T HEOREM (∀(x, y) ∈ R2 ), ¯ ¯ ¯ ¯
¯ x y ¯ = |x| ¯ y ¯ .

Proof: We have ¯ ¯ ¡ ¢ ¡ ¢¡ ¡ ¢ ¢ ¯ ¯
¯x y ¯ = signum x y x y = signum(x) x signum y y = |x| ¯ y ¯ ,

where we have used Lemma 70. ❑

74 T HEOREM Let t ≥ 0. Then

|x| ≤ t ⇐⇒ −t ≤ x ≤ t .

Proof: Either |x| = x or |x| = −x. If |x| = x,

|x| ≤ t ⇐⇒ x ≤ t ⇐⇒ −t ≤ 0 ≤ x ≤ t .

If |x| = −x,
|x| ≤ t ⇐⇒ −x ≤ t ⇐⇒ −t ≤ x ≤ 0 ≤ t .
❑
¯ ¯ ¯ ¯
2 x + y + ¯x − y ¯ x + y − ¯x − y ¯
75 T HEOREM If (x, y) ∈ R , max(x, y) = and min(x, y) = .
2 2

Proof: Observe that max(x, y) + min(x, y) = x + y, since one of these quantities must be the maximum and the
other the minimum, or else, they are both equal.
¯ ¯ ¯ ¯
Now, either ¯ x − y ¯ = x − y, and so x ≥ y, meaning that max(x, y)−min(x, y) = x − y, or ¯x − y ¯ = −(x − y) =
¯ y − x,
¯
which means that y ≥ x and so max(x, y)−min(x, y) = y −x. In either case we get max(x, y)−min(x, y) = ¯ x − y ¯.
Solving now the system of equations

max(x, y) + min(x, y) = x+y

¯ ¯
max(x, y) − min(x, y) = ¯x − y ¯ ,

for max(x, y) and min(x, y) gives the result. ❑

Homework

Problem 1.6.1 Let x, y be real numbers. Then Problem 1.6.4 Let x, y, z be real numbers. Prove that
2 2
0 ≤ x < y ⇐⇒ x < y .
max(x, y, z) = x+y +z−min(x, y )−min(y, z)−min(z, x)+min(x, y, z).
Problem 1.6.2 Let t ≥ 0. Prove that
|x| ≥ t ⇐⇒ (x ≥ t ) or (x ≤ −t ). Problem 1.6.5 Let a < b. Demonstrate that

Problem 1.6.3 Let (x, y ) ∈ R2 . Prove that max(x, y ) = a +b

|x − a| < |x − b| ⇐⇒ x < .
−min(−x,−y ). 2

1.7 Classical Inequalities

1.7.1 Triangle Inequality
76 T HEOREM (Triangle Inequality) Let (a,b) ∈ R2 . Then

|a + b| ≤ |a| + |b|. (1.6)

21
 Classical Inequalities

Proof: From 5 in Theorem 72, by addition,

−|a| ≤ a ≤ |a|
to
−|b| ≤ b ≤ |b|
we obtain
−(|a| + |b|) ≤ a + b ≤ (|a| + |b|),
whence the theorem follows by applying Theorem 74. ❑
By induction, we obtain the following generalisation to n terms.

77 C OROLLARY Let x1 , x2 ,... , xn be real numbers. Then

|x1 + x2 + · · · + xn | ≤ |x1 | + |x2 | + · · · + |xn | .

Proof: We apply Theorem 76 n − 1 times

|x1 + x2 + · · · + xn | ≤ |x1 | + |x2 + · · · xn−1 + xn |

≤ |x1 | + |x2 | + |x3 + · · · xn−1 + xn |
..
.
≤ |x1 | + |x2 | + · · · + |xn−1 + xn |
≤ |x1 | + |x2 | + · · · + |xn−1 | + |xn | .

78 C OROLLARY Let (a,b) ∈ R2 . Then

||a| − |b|| ≤ |a − b| . (1.7)

Proof: We have
|a| = |a − b + b| ≤ |a − b| + |b|,
giving
|a| − |b| ≤ |a − b|.
Similarly,
|b| = |b − a + a| ≤ |b − a| + |a| = |a − b| + |a|,
gives
|b| − |a| ≤ |a − b| =⇒ − |a − b| ≤ |a| − |b| .
Thus
− |a − b| ≤ |a| − |b| ≤ |a − b| ,
and we now apply Theorem 74. ❑

79 T HEOREM Let b i > 0 for 1 ≤ i ≤ n. Then

µ ¶ µ ¶
a1 a2 an a1 + a2 + · · · + an a1 a2 an
min , ,... , ≤ ≤ max , ,... , .
b1 b2 bn b1 + b2 + · · · + bn b1 b2 bn

Proof: For every k, 1 ≤ k ≤ n,

µ ¶ µ ¶ µ ¶ µ ¶
a1 a2 an ak a1 a2 an a1 a2 an a1 a2 an
min , ,... , ≤ ≤ max , ,... , =⇒ bk min , ,... , ≤ a k ≤ bk max , ,... , .
b1 b2 bn bk b1 b2 bn b1 b2 bn b1 b2 bn
Adding all these inequalities for 1 ≤ k ≤ n,
µ ¶ µ ¶
a1 a2 an a1 a2 an
(b1 + b2 + · · · + bn ) min , ,... , ≤ a 1 + a 2 + · · · + a n ≤ (b1 + b2 + · · · + bn ) max , ,... , ,
b1 b2 bn b1 b2 bn
from where the result is obtained.❑

22
 Chapter 1

1.7.2 Bernoulli’s Inequality

80 T HEOREM If 0 ≤ a < b, n ≥ 1 ∈ N
bn − an
na n−1 < < nb n−1 .
b−a

Proof: By Theorem 55,

bn − an
= b n−1 + b n−2 a + b n−3 a 2 + · · · + b 2 a n−3 + ba n−2 + a n−1
b−a
< b n−1 + b n−1 + · · · + b n−1 + b n−1
= nb n−1 ,

from where the dextral inequality follows. The sinistral inequality can be established similarly. ❑

81 T HEOREM (Bernoulli’s Inequality) If x > −1, x 6= 0, and if n ∈ N \ {0} then

(1 + x)n > 1 + nx.

Proof: Set b = 1 + x, a = 1 in Theorem 80 and use the sinistral inequality. ❑

☞ If x > 0 then Bernoulli’s Inequality is an easy consequence of the Binomial Theorem, as

à ! à ! à !
n n n 2 n
(1 + x) = 1 + x+ x +··· > 1+ x = 1 + nx.
1 2 1

1.7.3 Rearrangement Inequality

82 Definition Given a set of real numbers {x1 , x2 ,... , xn } denote by

x̌1 ≥ x̌2 ≥ · · · ≥ x̌n

the decreasing rearrangement of the xi and denote by

x̂1 ≤ x̂2 ≤ · · · ≤ x̂n

the increasing rearrangement of the xi .

83 Definition Given two sequences of real numbers {x1 , x2 ,... , xn } and {y 1 , y 2 ,... , y n } of the same length n, we say that they
are similarly sorted if they are both increasing or both decreasing, and differently sorted if one is increasing and the other
decreasing..

1 1 1
84 Example The sequences 1 ≤ 2 ≤ · · · ≤ n and 12 ≤ 22 ≤ · · · ≤ n 2 are similarly sorted, and the sequences ≥ ≥ ··· ≥ 2
12 22 n
and 13 ≤ 23 ≤ · · · ≤ n 3 are differently sorted.

85 T HEOREM (Rearrangement Inequality) Given sets of real numbers {a 1 , a 2 ,... , a n } and {b 1 ,b 2 ,... ,b n } we have
X X X
ǎ k b̂k ≤ ak bk ≤ â k b̂k .
1≤k≤n 1≤k≤n 1≤k≤n
X
Thus the sum a k bk is minimised when the sequences are differently sorted, and maximised when the sequences are
1≤k≤n
similarly sorted.

23
 Classical Inequalities

☞ Observe that X X X X
ǎk b̂k = âk b̌k and âk b̂k = ǎk b̌k .
1≤k≤n 1≤k≤n 1≤k≤n 1≤k≤n

Proof: Let {σ(1),σ(2),... ,σ(n)} be a reordering of {1,2,... ,n}. If there are two sub-indices i , j , such that the
sequences pull in opposite directions, say, a i > a j and bσ(i ) < bσ( j ) , then consider the sums

S = a 1 bσ(1) + a 2 bσ(2) + · · · + a i bσ(i ) + · · · + a j bσ( j ) + · · · + a n bσ(n)

S0 = a 1 bσ(1) + a 2 bσ(2) + · · · + a i bσ( j ) + · · · + a j bσ(i ) + · · · + a n bσ(n)

Then
S 0 − S = (a i − a j )(bσ( j ) − bσ(i ) ) > 0.
This last inequality shews that the closer the a’s and the b’s are to pulling in the same direction the larger the sum
becomes. This proves the result. ❑

1.7.4 Arithmetic Mean-Geometric Mean Inequality

86 T HEOREM (Arithmetic Mean-Geometric Mean Inequality) Let a 1 ,... , a n be positive real numbers. Then their geometric
mean is at most their arithmetic mean, that is,
p
n
a1 + · · · + an
a1 · · · an ≤ ,
n
with equality if and only if a 1 = · · · = a n .

We will provide multiple proofs of this important inequality. Some other proofs will be found in latter chapters.

First Proof: Our first proof uses the Rearrangement Inequality (Theorem 85) in a rather clever way. We may
assume that the a k are strictly positive. Put
a1 a1 a2 a1 a2 · · · an
x1 = , x2 = , ... , xn = = 1,
(a 1 a 2 · · · a n )1/n (a 1 a 2 · · · a n )2/n (a 1 a 2 · · · a n )n/n

and
1 1 1
y1 = , y2 = , ... , yn = = 1.
x1 x2 xn
Observe that for 2 ≤ k ≤ n,

a1 a2 · · · ak (a 1 a 2 · · · a n )(k−1)/n ak
xk y k−1 = · = .
(a 1 a 2 · · · a n )k/n a 1 a 2 · · · a k−1 (a 1 a 2 · · · a n )1/n

The xk and y k are differently sorted, so by virtue of the Rearrangement Inequality we gather

1 +1 +··· +1 = x1 y 1 + x2 y 2 + · · · + xn y n
≤ x1 y n + x2 y 1 + · · · + xn y n−1
a1 a2 an
= + +··· + ,
(a 1 a 2 · · · a n )1/n (a 1 a 2 · · · a n )1/n (a 1 a 2 · · · a n )1/n
or
a1 + a2 + · · · + an
n≤ ,
(a 1 a 2 · · · a n )1/n
from where we obtain the result. ❑

Second Proof: This second proof is a clever induction argument due to Cauchy. It proves the inequality first for
powers of 2 and then interpolates for numbers between consecutive powers of 2.

Since the square of a real number is always positive, we have, for positive real numbers a,b

p p p a +b
( a − b)2 ≥ 0 =⇒ ab ≤ ,
2

24
 Chapter 1

proving the inequality for k = 2. Observe that equality happens if and only if a = b. Assume now that the in-
equality is valid for k = 2n−1 > 2. This means that for any positive real numbers x1 , x2 ,... , x2n−1 we have
¡ ¢1/2n−1 x1 + x2 + · · · + x2n−1
x1 x2 · · · x2n−1 ≤ . (1.8)
2n−1
Let us prove the inequality for 2k = 2n . Consider any any hpositive real numbers
i y 1 , y 2 ,... , y 2n . Notice that there
are 2n − 2n−1 = 2n−1 (2 − 1) = 2n−1 integers in the interval 2n−1 + 1 ;2n . We have
q
¡ ¢1/2n ¡ ¢1/2n−1 ¡ ¢1/2n−1
y 1 y 2 · · · y 2n = y 1 y 2 · · · y 2n−1 y 2n−1 +1 · · · y 2n
¡ ¢1/2n−1 ¡ ¢1/2n−1
y 1 y 2 · · · y 2n−1 + y 2n−1 +1 · · · y 2n
≤
2
y 1 + y 2 + · · · + y 2n−1 y 2n−1 +1 + · · · + y 2n
+
≤ 2n−1 2n−1
2
y 1 + · · · + y 2n
= ,
2n
where the first inequality follows by the Case n = 2 and the second by the induction hypothesis (1.8). The theorem
is thus proved for powers of 2.

Assume now that 2n−1 < k < 2n , and consider the k positive real numbers a 1 , a 2 ,... , a k . The trick is to pad this
collection of real numbers up to the next highest power of 2, the added real numbers being the average of the
existing ones. Hence consider the 2n real numbers

a 1 , a 2 ,... , a k , a k+1 ,... , a 2n

a1 + a2 + · · · + ak
with a k+1 = ... = a 2n = . Since we have already proved the theorem for 2n we have
k
³ a + a +··· + a ´
1 2 k
µ ³ a + a + · · · + a ´2n −k ¶1/2n a 1 + a 2 + · · · + a k + (2n − k)
a1 a2 · · · ak
1 2 k
≤ k ,
k 2n

whence
a1 + a2 + · · · + ak ³ a + a +··· + a ´
1 2 k
³ ´
n a 1 + a 2 + · · · + a k 1−k/2
n k + (2n − k)
k k
(a 1 a 2 · · · a k )1/2 ≤ ,
k 2n
which implies
n
³ a + a + · · · + a ´1−k/2n ³ a + a + · · · + a ´
1 2 k 1 2 k
(a 1 a 2 · · · a k )1/2 ≤ ,
k k
a1 + a2 + · · · + ak
Solving for gives the desired inequality. ❑
k
Third Proof: As in the second proof, the Case k = 2 is easily established. Put
a1 + a2 + · · · + ak
Ak = , G k = (a 1 a 2 · · · a k )1/k .
k
Observe that
a k+1 = (k + 1)A k+1 − k A k .
The inductive hypothesis is that A k ≥ G k and we must shew that A k+1 ≥ G k+1 . Put

a k+1 + (k − 1)A k+1 ³ ´1/k

A= , G = a k+1 A k−1
k+1 .
k
By the inductive hypothesis A ≥ G. Now,

(k + 1)A k+1 − k A k + (k − 1)A k+1

+ Ak
A + Ak k
= = A k+1 .
2 2

25
 Classical Inequalities

Hence
A + Ak
A k+1 =
2
≥ (A A k )1/2
≥ (GG k )1/2 .
³ ´1/2k
k+1 k−1
= G k+1 A k+1

We have established that ³ ´1/2k

k+1 k−1
A k+1 ≥ G k+1 A k+1 =⇒ A k+1 ≥ G k+1 ,

completing the induction. ❑

Fourth Proof: We will make a series of substitutions that preserve the sum

a1 + a2 + · · · + an

while strictly increasing the product

a1 a2 · · · an .
At the end, the a i will all be equal and the arithmetic mean A of the numbers will be equal to their geometric
a1 + a2 + · · · + an nA
mean G. If the a i where all > A then > = A, impossible. Similarly, the a i cannot be all
n n
< A. Hence there must exist two indices say i , j , such that a i < A < a j . Put a i0 = A, a 0j = a i + a j − A. Observe
that a i + a j = a i0 + a 0j , so replacing the original a’s with the primed a’s does not alter the arithmetic mean. On
the other hand, ¡ ¢ ¡ ¢
a i0 a 0j = A a i + a j − A = a i a j + a j − A (A − a i ) > a i a j
since a j − A > 0 and A − a i > 0.

This change has replaced one of the a’s by a quantity equal to the arithmetic mean, has not changed the arith-
metic mean, and made the geometric mean larger. Since there at most n a’s to be replaced, the procedure must
eventually terminate when all the a’s are equal (to their arithmetic mean). Strict inequality then holds when at
least two of the a’s are unequal. ❑

1.7.5 Cauchy-Bunyakovsky-Schwarz Inequality

87 T HEOREM (Cauchy-Bunyakovsky-Schwarz Inequality) Let xk , y k be real numbers, 1 ≤ k ≤ n. Then
¯ ¯ Ã !1/2 Ã !1/2
¯Xn ¯ X
n X
n
¯ ¯ 2 2
¯ x y ¯≤ xk yk ,
¯k=1 k k ¯ k=1 k=1

with equality if and only if

(a 1 , a 2 ,... , a n ) = t(b1 ,b2 ,... ,bn )
for some real constant t.

First Proof: The inequality follows at once from Lagrange’s Identity

à !2 à !à !
X
n X
n X
n X
xk y k = xk2 y k2 − (xk y j − x j y k )2
k=1 k=1 k=1 1≤k< j ≤n
X
(Theorem 50), since (xk y j − x j y k )2 ≥ 0. ❑
1≤k< j ≤n

X
n X
n X
n
Second Proof: Put a = xk2 , b = xk y k , and c = y k2 . Consider the quadratic polynomial
k=1 k=1 k=1

X
n X
n X
n X
n
at 2 + bt + c = t 2 xk2 − 2t xk y k + y k2 = (t xk − y k )2 ≥ 0,
k=1 k=1 k=1 k=1

26
 Chapter 1

where the inequality follows because a sum of squares of real numbers is being summed. Thus this quadratic
polynomial is positive for all real t, so it must have complex roots. Its discriminant b 2 − 4ac must be negative,
from where we gather
à !2 à !à !
X
n X
n
2
X
n
2
4 xk y k ≤ 4 xk yk ,
k=1 k=1 k=1
which gives the inequality ❑

For our third proof of the CBS Inequality we need the following lemma.

88 L EMMA For (a,b, x, y) ∈ R4 with x > 0 and y > 0 the following inequality holds:

a 2 b 2 (a + b)2
+ ≥ .
x y x+y
a b
Equality holds if and only if = .
x y

Proof: Since the square of a real number is always positive, we have

(a y − bx)2 ≥ 0 =⇒ a 2 y 2 − 2abx y + b 2 x 2 ≥ 0
=⇒ a 2 y(x + y) + b 2 x(x + y) ≥ (a + b)2 x y
a 2 b 2 (a + b)2
=⇒ + ≥ .
x y x+y
Equality holds if and only if the first inequality is 0.❑

☞ Iterating the result on Lemma 88,

a12 a2 a2 (a1 + a2 + · · · + an )2
+ 2 +··· + n ≥ ,
b1 b2 bn b1 + b2 + · · · + bn
a1 a2 an
with equality if and only if = = ··· = .
b1 b2 bn

Third Proof: By the preceding remark, we have

x12 y 12 x2 y2 x2 y2
x12 + x22 + · · · + xn2 = 2
+ 2 22 + · · · + n 2 n
y1 y2 yn
(x1 y 1 + x2 y 2 + · · · + xn y n )2
≥ ,
y 12 + y 22 + · · · + y n2

and upon rearranging, CBS is once again obtained.❑

1.7.6 Minkowski’s Inequality

89 T HEOREM (Minkowski’s Inequality) Let xk , y k be any real numbers. Then
à !1/2 à !1/2 à !1/2
X
n
2
X
n X
n
(xk + y k ) ≤ xk2 + y k2 .
k=1 k=1 k=1

Proof: We have
X
n X
n X
nXn
(xk + y k )2 = xk2 + 2
xk y k + y k2
k=1 k=1 k=1
à k=1
!1/2 Ã !1/2
Xn
2
X
n
2
Xn
2
X
n
≤ xk + 2 xk yk + y k2
k=1 k=1 k=1 k=1
ÃÃ !1/2 Ã !1/2 !2
X
n X
n
= xk2 + y k2 ,
k=1 k=1
where the inequality follows from the CBS Inequality.❑

27
 Classical Inequalities

Homework

Problem 1.7.1 Let (a,b,c,d ) ∈ R4 . Prove that Problem 1.7.11 Prove that ∀(a,b,c) ∈ R3 , with a ≥ 0, b ≥ 0, c ≥ 0,
the following inequalities hold:
||a − c| − |b − c|| ≤ |a − b| ≤ |a − c| + |b − c| .
a 3 + b 3 + c 3 ≥ max(a 2 b + b 2 c + c 2 a, a 2 c + b 2 a + c 2 b),

Problem 1.7.2 Let (x1 , x2 ,... , xn ) ∈ Rn be such that a 3 + b 3 + c 3 ≥ 3abc,

1³ 2 ´
a 3 + b3 + c 3 ≥ a (b + c) + b 2 (c + a) + c 2 (a + b) .
x12 + x22 + · · · + xn2 = x13 + x23 + · · · + xn3 = x14 + x24 + · · · + xn4 . 2

Prove that xk ∈ {0,1}. Problem 1.7.12 (Chebyshev’s Inequality) Given sets of real num-
bers {a1 , a2 ,... , an } and {b1 ,b2 ,... ,bn } prove that
à !à !
Problem 1.7.3 Let n ≥ 2 an integer. Let (x1 , x2 ,... , xn ) ∈ Rn be such 1 X 1 X 1 X 1 X
that ǎk b̂k ≤ ak bk ≤ â b̂ .
n 1≤k≤n n 1≤k≤n n 1≤k≤n n 1≤k≤n k k

x12 + x22 + · · · + xn2 = x1 x2 + x2 x3 + · · · + xn−1 xn + xn x1 .

Problem 1.7.13 If x > 0, from
Prove that x1 = x2 = · · · = xn . p p 1
x +1− x = p p ,
x +1+ x
Problem 1.7.4 If b > 0 and B > 0 prove that
prove that
1 p p 1
a A a a+A A p < x +1− x < p .
< =⇒ < < . 2 x +1 2 x
b B b b +B B
Use this to prove that if n > 1 is a positive integer, then
Further, if p and q are positive integers such that
p 1 1 1 p
2 n +1 −2 < 1+ p + p +··· + p < 2 n −1
7 p 11 2 3 n
< < ,
10 q 15
Problem 1.7.14 If 0 < a ≤ b, shew that
what is the least value of q?
1 (b − a)2 a + b p 1 (b − a)2
· ≤ − ab ≤ ·
8 b 2 8 a
Problem 1.7.5 Prove that if r ≥ s ≥ t then
Problem 1.7.15 Shew that
r 2 − s 2 + t 2 ≥ (r − s + t )2 . 1 3 5 9999 1
· · ··· < .
2 4 6 10000 100
Problem 1.7.6 Assume that ak ,bk ,ck ,k = 1,... ,n, are positive real
numbers. Shew that Problem 1.7.16 Prove that for all x > 0,

à !4 à !à !à !2 n
X 1 1 1
n
X n
X n
X n
X < − .
ak bk ck ≤ ak4 bk4 ck2 . k=1 (x + k)
2 x x +n
k=1 k=1 k=1 k=1
X¯ ¯ X
Problem 1.7.17 Let xi ∈ R such that ¯xi ¯ = 1 and xi = 0.
Problem 1.7.7 Prove that for integer n > 1, i =1 i =1
Prove that ¯ ¯
µ ¶ ¯X n x ¯ 1µ ¶
n +1 n ¯ i ¯ 1
n! < . ¯ ¯≤ 1− .
2 ¯i =1 i ¯ 2 n

Problem 1.7.8 Prove that for integer n > 2, Problem 1.7.18 Let n be a strictly positive integer. Let xi ≥ 0. Prove
that
Yn n
X
n n/2 < n!. (1 + xk ) ≥ 1 + xk .
k=1 k=1

Problem 1.7.9 Prove that for all integers n ≥ 0 the inequality n(n − When does equality hold?
1) < 2n+1 is verified.
Problem 1.7.19 (Nesbitt’s Inequality) Let a,b,c be strictly positive
3 real numbers. Then
Problem 1.7.10 Prove that ∀(a,b,c) ∈ R ,
a b c 3
+ + ≥ .
a 2 + b 2 + c 2 ≥ ab + bc + ca. b +c c +a a +b 2

28
 Chapter 1

Problem 1.7.20 Let a > 0. Use mathematical induction to prove Equality occurs if and only if ad = bc.
that s r q p
p h i
1 + 4a + 1
a + a + a +··· + a < , Problem 1.7.27 Let xk ∈ 0 ;1 for 1 ≤ k ≤ n. Demonstrate that
2
where the left member contains an arbitrary number of radicals. Ã !
n
Y n
Y 1
min xk , (1 − xk ) ≤ n .
k=1 k=1 2
Problem 1.7.21 Let a,b,c be positive real numbers. Prove that

(a + b)(b + c)(c + a) ≥ 8abc.

Problem 1.7.28 If n > 0 is an integer and if ak > 0, 1 ≤ k ≤ n are
real numbers, demonstrate that
Problem 1.7.22 (IMO, 1978) Let ak be a sequence of pairwise dis-
à !2
tinct positive integers. Prove that Xn a n X
X n a j ak
k
≤ .
n a
X Xn 1
k k=1 k j =1 k=1 j + k − 1
2
≥ .
k=1 k k=1 k
Problem 1.7.29 Let n be a strictly positive integer, let ak ≥ 0, 1 ≤
Problem 1.7.23 (Harmonic Mean-Geometric Mean Inequality) k ≤ n be real numbers such that a1 ≥ a2 ≥ · · · ≥ an , and let bk , 1 ≤
Let xi > 0 for 1 ≤ i ≤ n. Then k ≤ n be real numbers. Assume that for all indices k ∈ {1,2,... ,n},
n
≤ (x1 x2 · · · xn )1/n , k
X k
X
1 1 1
+ +··· + ai ≤ bi .
x1 x2 xn i =1 i =1
with equality iff x1 = x2 = · · · = xn .
Prove that
n
X n
X
Problem 1.7.24 (Arithmetic Mean-Quadratic Mean Inequality) ai2 ≤ bi2
i =1 i =1
Let xi ≥ 0 for 1 ≤ i ≤ n. Then
à 2 !1/2
x1 + x2 + · · · + xn x1 + x22 + · · · + xn2 Problem 1.7.30 Let n ≥ 2 an integer and let ak , 1 ≤ k ≤ n be real
≤ ,
n n numbers such that a1 ≤ a2 ≤ · · · ≤ an . Prove that there is an index
k ∈ {1,2,... ,n} such that
with equality iff x1 = x2 = · · · = xn .
12
(ak+1 − ak )2 ≤ (a12 + a22 + · · · an2 ).
Problem 1.7.25 Given a set of real numbers {a1 , a2 ,... , an } prove n(n 2 − 1)
that there is an index m ∈ {0,1,... ,n} such that
¯ ¯ Problem 1.7.31 (AIME 1991) Let P = {a1 , a2 ,... , an } be a collec-
¯ X X ¯ ¯ ¯
¯ ¯
¯ ak − ak ¯ ≤ max ¯ak ¯ . tion of points with
¯ ¯ 1≤k≤n
1≤k≤m m<k≤n

If m = 0 the first sum is to be taken as 0 and if m = n the second one 0 < a1 < a2 < · · · < an < 17.
will be taken as 0.
Consider
n q
X
Problem 1.7.26 Give a purely geometric proof of Minkowski’s In- S n = min (2k − 1)2 + ak2 ,
P k=1
equality for n = 2. That is, prove that if (a,b),(c,d ) ∈ R2 , then
q p p where the minimum runs over all such partitions P . Shew that ex-
(a + c)2 + (b + d )2 ≤ a 2 + b 2 + c 2 + d 2 . actly one of S 2 ,S 3 ,... ,S n ,... is an integer, and find which one it is.

1.8 Completeness Axiom

Why bother?We saw that both Q and R are fields, and hence they both satisfy the same arithmetical axioms.
Why the need then for R? In this section we will study a property of R that is not shared with Q, that of com-
pleteness. It essentially means that there are no ‘holes’ on the real line.

90 Definition A number u is an upper bound for a set of numbers A ⊆ R if for all a ∈ A we have a ≤ u. The smallest such
upper bound is called the supremum or least upper bound of the set A, and is denoted by sup A. If sup A ∈ A then we say
that A has a maximum and we denote it by max A(= sup A). Similarly, a number l is a lower bound for a set of numbers
B ⊆ R if for all b ∈ B we have l ≤ b. The largest such lower bound is called the infimum or greatest lower bound of the set B ,
and is denoted by inf B . If sup B ∈ B then we say that B has a minimum and we denote it by inf B (= inf B ).

29
 Completeness Axiom

☞ We define inf(R) = −∞, sup(R) = +∞, inf(∅) = +∞ and sup(∅) = −∞.

91 Definition A set of numbers A is said to be complete if every non-empty subset of A which is bounded above has a
supremum lying in A.

92 Axiom (Completeness of R) Any non-empty set of real numbers which is bounded above has a supremum. Any non-
empty set of real numbers which is bounded below has a infimum.

93 T HEOREM (Approximation Property of the Supremum and Infimum) Let A 6= ∅ be a set of real numbers possessing a supre-
mum sup A. Then
∀ε > 0 ∃a ∈ A such that sup A − ε ≤ a.

Let B 6= ∅ be a set of real numbers possessing an infimum inf B . Then

∀ε > 0 ∃b ∈ B such that inf A + ε ≥ b.

Proof: If ∀a ∈ A, sup A − ε > a then sup A − ε would be an upper bound smaller than the least upper bound,
a contradiction to the definition of sup A. Hence there must be a rogue a ∈ A such that sup A − ε ≤ a.

If ∀b ∈ A, inf B + ε < b then inf B + ε would be a lower bound greater than the greatest lower bound, a contra-
diction to the definition of inf B . Hence there must be a rogue b ∈ B such that inf B + ε ≥ b.
❑
☞ The above result should be intuitively clear. sup A sits on the fence, just to the right of A, so that going just a bit to the
left should put sup A − ε within A, etc.

94 T HEOREM (Monotonicity Property of the Supremum and Infimum) Let ∅ á A ⊆ B j R and suppose that both A and B
have a supremum and an infimum. Then sup A ≤ sup B and inf B ≤ inf A.

Proof: Assume B is bounded above with supremum sup B . Suppose x ∈ A. Then x ∈ B and so x ≤ sup B . Thus
sup B is an upper bound for the elements of A, and so A and so by definition, sup A ≤ sup B .

Assume B is bounded below with infimum inf B . Suppose x ∈ A. Then x ∈ B and so x ≥ inf B . Thus inf B is a
lower bound for the elements of A and so by definition, inf A ≥ inf B . ❑

95 L EMMA Let a,b be real numbers and assume that for all numbers ε > 0 the following inequality holds:

a < b + ε.

Then a ≤ b.

a −b
Proof: Assume contrariwise that a > b. Hence > 0. Since the inequality a < b + ε holds for every ε > 0 in
2
a −b
particular it holds for ε = . This implies that
2
a −b
a <b+ or a < b.
2
Thus starting with the assumption that a > b we reach the incompatible conclusion that a < b. The original
assumption must be wrong. We therefore conclude that a ≤ b. ❑

96 T HEOREM (Additive Property of the Supremum) Let ∅ á A ⊆ R, and B j R. Put

A + B = {x + y : (x, y) ∈ A × B }

and suppose that both A and B have a supremum. Then A + B has also a supremum and
sup(A + B ) = sup A + sup B .

30
 Chapter 1

Proof: If t ∈ A + B then t = x + y with (x, y) ∈ A × B . Then t = x + y ≤ sup A + sup B , and so sup A + sup B is an
upper bound for A + B . By the Completeness Axiom, A + B is bounded. Thus sup(A + B ) ≤ sup A + sup B .

We now prove that sup A + sup B ≤ sup(A + B ). By the approximation property, ∀ε > 0 ∃a ∈ A and b ∈ B such
ε ε
that sup A − < a and sup B − < b. Observe that a + b ∈ A + B and so a + b ≤ sup(A + B ). Then
2 2
sup A + sup B − ε < a + b ≤ sup(A + B ),

and by Lemma 95 we must have

sup A + sup B ≤ sup(A + B ).
This completes the proof.❑

97 T HEOREM (Archimedean Property of the Real Numbers) If (x, y) ∈ R2 with x > 0, then there exists a natural number n
such that nx > y.

Proof: Consider the set

A = {nx : n ∈ N}.
Since 1 · x ∈ A, A is non-empty. If ∀n ∈ N we had nx ≤ y, then A would be bounded above by y. By the Complete-
ness Axiom, A would have a supremum sup A. Thus ∀n ∈ N, nx ≤ sup A. Since (n + 1)x ∈ A, we would also
have
(n + 1)x ≤ sup A =⇒ nx ≤ sup A − x.
This means that sup A − x is an upper bound for A which is smaller than its supremum, a contradiction Thus
there must be an n for which nx > y. ❑

98 C OROLLARY N is unbounded above.

Proof: This follows by taking x = 1 in Theorem 97. ❑

The Completeness Axioms tells us, essentially, that there are no “holes” in the real numbers. We will see that this prop-
erty distinguishes the reals from the rational numbers.
p
99 L EMMA [Hipassos of Metapontum] 2 is irrational.

m
Proof: Assume there is s ∈ Q such that s 2 = 2. We can find integers m,n 6= 0 such that s = . The crucial part
n
of the argument is that we can choose m,n such that this fraction be in least terms, and hence, m,n must not be
both even. Now, m 2 s 2 = n 2 , that is 2m 2 = n 2 . This means that n 2 is even. But then n itself must be even, since
the product of two odd numbers is odd. Thus n = 2a for some non-zero integer a (since n 6= 0). This means that
2m 2 = (2a)2 = 4a 2 =⇒ m 2 = 2a 2 . This means once again that m is even. But then we have a contradiction,
since m and n were not both even. ❑

100 T HEOREM Q is not complete.

Proof: We must shew that there is a non-empty set of rational numbers which is bounded above but that does
not have a supremum in Q. Consider the set A = {r ∈ Q : r 2 ≤ 2} of rational numbers. This set is bounded above
by u = 2. For assume that there were a rogue element of A, say r 0 such that r 0 > 2. Then r 02 > 4 and so r 0 would
not belong to A, a contradiction. Thus r ≤ 2 for every r ∈ A and so A is bounded above. Suppose that A had a
supremum s, which must satisfy s ≤ 2. Now, by Lemma 99 we cannot have s 2 = 2 and thus s 2 < 2. By Theorem
1 1
97 there is an integer n such that 2 − s 2 > n . Put t = s + n−1 , a rational number and observe that since s ≤ 2
10 10
one has
2s 1 2s 1 5 1
t 2 = s 2 + n−1 + 2n−2 < s 2 + n−1 + n−1 ≤ s 2 + n−1 < s 2 + n < 2.
10 10 10 10 10 10
Thus t ∈ A and t > s, that is t is an element of A larger than its least upper bound, a contradiction. Hence A does
not have a least upper bound. ❑

31
 Completeness Axiom

1.8.1 Greatest Integer Function

101 T HEOREM Given y ∈ R there exists a unique integer n such that

n ≤ y < n + 1.

Proof: By Theorem 97, the set {n ∈ Z : n ≤ y} is non-empty and bounded above. We put

T y U = sup{n ∈ Z : n ≤ y}.

☞ ∀x ∈ R, Tx U ≤ x < Tx U + 1.

102 Definition The unique integer in Theorem 101 is called the floor of x and is denoted by Tx U.

The greatest integer function enjoys the following properties:

103 T HEOREM Let α,β ∈ R, a ∈ Z,n ∈ N. Then

1. Tα + a U = TαU + a

α TαU
2. T U=T U
n n
3. TαU + TβU ≤ Tα + βU ≤ TαU + TβU + 1

Proof:

1. Let m = Tα + a U. Then m ≤ α + a < m + 1. Hence m − a ≤ α < m − a + 1. This means that m − a = TαU,

which is what we wanted.
2. Write α/n as α/n = Tα/n U + θ,0 ≤ θ < 1. Since n Tα/n U is an integer, we deduce by (1) that

TαU = Tn Tα/n U + nθU = n Tα/n U + TnθU.

Now, 0 ≤ TnθU ≤ nθ < n, and so 0 ≤ TnθU/n < 1. If we let Θ = TnθU/n, we obtain

TαU α
=T U + Θ, 0 ≤ Θ < 1.
n n
This yields the required result.
3. From the inequalities α−1 < TαU ≤ α,β−1 < TβU ≤ β we get α+β−2 < TαU+TβU ≤ α+β. Since TαU+TβU
is an integer less than or equal to α+β, it must be less than or equal to the integral part of α+β, i.e. Tα+βU.
We obtain thus TαU + TβU ≤ Tα + βU. Also, α + β is less than the integer TαU + TβU + 2, so its integer part
Tα+βU must be less than TαU + TβU +2, but Tα+βU < TαU + TβU +2 yields Tα+βU ≤ TαU + TβU +1. This
proves the inequalities.

104 Definition The ceiling of a real number x is the unique integer Vx W satisfying the inequalities

Vx W − 1 < x ≤ Vx W.

105 Definition The fractional part of a real number x is defined and denoted by

{x} = x − Tx U.

Observe that 0 ≤ {x} < 1.

Homework

32
 Chapter 1

Problem 1.8.1 Let A and B be non-empty sets of real numbers. Put Problem 1.8.6 Let x ∈ R and let n be a strictly positive integer. Prove
that
−A = {−x : x ∈ A}, A − B = {a − b : (a,b) ∈ A × B }. n−1
X k
Tnx U = T x + U.
Prove that k=1 n
1. If A is bounded above, then −A is bounded below and sup A =
−inf(−A).
Problem 1.8.7 (Putnam 1948) If n is a positive integer, demon-
2. If A and B are bounded above then A ∪ B is also bounded strate that
above and sup(A ∪ B ) = max(sup A,sup B ). p p p
T n + n + 1U = T 4n + 2U.
3. If A is bounded above and B is bounded below, then A − B is
bounded above and sup(A − B ) = sup A − infB .
Problem 1.8.8 Find a formula for the n-th non-square.

Problem 1.8.2 Assume that A is a subset of the strictly positive real

1 Problem 1.8.9 Prove that if a,b are strictly positive integers then
numbers. Prove that if A is bounded above, then the set A −1 = { :
x
1
x ∈ A} is bounded below and that sup A = . a2 (a + 2b)2
inf A −1 < 2 =⇒ < 2.
b2 (a + b)2
Problem 1.8.3 Let n ≥ 2 be an integer. Prove that
Prove, moreover, that
à !
X n2
max (x j − xi ) = T U. (a + 2b)2 a2
0≤x 1 ≤x 2 ≤···≤x n ≤1 1≤i < j ≤n 4 −2 < 2− .
(a + b)2 b2

Problem 1.8.4 Find a non-zero polynomial P (x, y ) such that (a + 2b)2 a2

This means that is closer to 2 than is.
P (T2t U, T3t U) = 0 (a + b)2 b2

for all real t .

p 2x + 5
Problem 1.8.10 Shew that ∀x > 0, x is farther from 5 than
x +2
Problem 1.8.5 Prove that the integers is.
³ p ´n
T 1+ 2 U
Problem 1.8.11 (Existence of n-th Roots) Let a > 0 and let n ∈ R,
with n a positive integer, are alternately even or odd. n ≥ 2. Prove that there is a unique b ∈ R such that b n = a.

33
Chapter 2

Topology of R

2.1 Intervals
Why bother? In this section we give a more precise definition of what an interval is, and establish the interesting
property that between any two real numbers there is always a rational number.

106 Definition An interval I is a subset of the real numbers with the following property: if s ∈ I and t ∈ I , and if s < x < t,
then x ∈ I . In other words, intervals are those subsets of real numbers with the property that every number between two
elements is also contained in the set. Since there are infinitely many decimals between two different real numbers, intervals
with distinct endpoints contain infinitely many members.

☞ The empty set ∅ is trivially an interval.

We will now establish that there are nine types of intervals.

Interval Notation Set Notation Graphical Representation

[a;b] {x ∈ R : a ≤ x ≤ b}1
a b
]a;b[ {x ∈ R : a < x < b}
a b
[a;b[ {x ∈ R : a ≤ x < b}
a b
]a;b] {x ∈ R : a < x ≤ b}
a b
]a;+∞[ {x ∈ R : x > a}
a +∞
[a;+∞[ {x ∈ R : x ≥ a}
a +∞
] − ∞;b[ {x ∈ R : x < b}
−∞ b
] − ∞;b] {x ∈ R : x ≤ b}
−∞ b
] − ∞;+∞[ R
−∞ +∞
Table 2.1: Types of Intervals. Observe that we indicate that the endpoints are included by means of shad-
ing the dots at the endpoints and that the endpoints are excluded by not shading the dots at
the endpoints.

☞ h i
If x ∈ R, then {x} = x ; x .

34
 Chapter 2

107 T HEOREM The only kinds of intervals are those sets shewn in Table 2.1, and conversely, all sets shewn in this table are
intervals.
h i
Proof: The converse is easily established, so assume that I j R possesses the property that ∀(a,b) ∈ I 2 , a ;b j
i ∅ is an
I . Since i interval one may assume that h I 6= ∅h. Let a ∈ I be a fixed element of I and put M a = {x ∈ I : x ≤
a} = − ∞ ; a ∩ I and N a = {x ∈ I : x ≥ a} = a ;+∞ ∩ I .

h h
If N a is not bounded above, then ∀b ∈ a ;+∞ , ∃c ∈ N a such that b ≤ c. Since a ≤ b ≤ c, this entails that
h h
b ∈ N a . Thus N a = a ;+∞ .

h i h h
If N a is bounded above, then it has supremum s = sup(N a ) and N a j a ; s . By Theorem 93, ∀b ∈ a ; s , c∈
N a such that b ≤ c, and since a ≤ b ≤ c, this entails that b ∈ N a . Thus
h h h i
a ; s j Na j a ; s ,
h h h i
and so N a = a ; s or N a = a ; s .

h h h i h h
Thus N a is one among three possible forms: a ;+∞ , a ; s , or a ; s . Applying a similar reasoning, one obtains
i i i i h i
gathers that M a is of one of the forms − ∞ ; a , l ; a , or l ; a , where l = inf(M a ). Since I = M a ∪ N a , there are
3 choices for M a and 3 for N a , hence there are 3 · 3 = 9 choices for I . The result is established. ❑

\∞ h 1 1i
108 Example Determine 1 − k ;1 + .
k=1 2 k

Solution: Observe that the intervals are, in sequence,

h1 i h3 3i h7 4i
;2 ; ; ; ; ; ... .
2 4 2 8 3
\∞ h 1 1i
We claim that 1 − k ;1 + = 1. For we see that
k=1 2 k

1 1 1
∀k ≥ 1, ≤ 1 − k < 1 < 1 + ≤ 2,
2 2 k
1
so 1 is in every interval. Could this intersection contain a number smaller than 1? No, for if ≤ a < 1, then we
2
can take k large enough so that
1
a < 1− ,
2k
for example
1
a < 1− =⇒ k > − log2 (1 − a),
2k
so taking k ≥ T− log2 (1 − a)U + 1 will work. Could the intersection contain a number b larger than 1? No, for if
1 < b < 2, then we can take k large enough so that

1
1+ < b,
k
for example
1 1
1+ < b =⇒ k > ,
k b −1
1
so taking k ≥ T U + 1 will work. Hence the only number in the intersection is 1.
b −1

35
 Dense Sets

2.2 Dense Sets

109 Definition A set B j R is dense in A j R if ∀(a 1 , a 2 ) ∈ A 2 ,
a1 < a2 , ∃b ∈ B such that a 1 < b < a 2 , that is, between any
two different elements of A one can always find an element of B .

110 T HEOREM Q is dense in R.

Proof: Let x, y be real numbers with x < y. Since there are infinitely many positive integers, there must be a
1 m
positive integer n such that n > by the Archimedean Property of R. Consider the rational number r = ,
y −x n
where m is the least natural number with m > nx. This means that

m > nx ≥ m − 1.
m m
We claim that x < < y. The first inequality is clear, since by choice x < . For the second inequality observe
n n
that, again
1 m 1 1 m 1 1 m
nx ≥ m − 1 and y − x > =⇒ x > − and y > x + =⇒ y > − + = .
n n n n n n n n
m
Thus is a rational number between x and y. ❑
n

111 T HEOREM R \ Q is dense in R.

a b
Proof: Let a < b be two real numbers. By Theorem 110, there is a rational number r with p < r < p . But
p p 2 2
then a < 2r < b, and the number 2r is an irrational number.❑

112 T HEOREM (Dirichlet) For any real number θ and any integer Q ≥ 1, there exist integers a and q, 1 ≤ q ≤ Q, such that
¯ ¯
¯ ¯
¯θ − a ¯ ≤ 1 .
¯ q ¯ qQ

Proof: For 1 ≤ n ≤ Q, let · ·

n −1 n
In = ; .
Q Q
Thus these Q intervals partition the interval [0;1[. The Q + 1 numbers

{0θ},{1θ},{2θ},... ,{Qθ}

lie in [0;1[. Hence by the pigeonhole principle there is an n such that I n contains at least two of these numbers,
say
{q 1 θ} ∈ I n , {q 2 θ} ∈ I n , 0 ≤ q 1 < q 2 ≤ Q.
Put q = q 2 − q 1 , a = [q 2 θ] − [q 1 θ]. Since {q 1 θ} ∈ I n ,{q 2 θ} ∈ I n we must have
¯ ¯
¯{q 2 θ} − {q 1 θ}¯ < 1 .
Q
But
{q 2 θ} − {q 1 θ} = q 2 θ − [q 2 θ] − q 1 θ + [q 1 θ] = qθ − a,
whence the result. ❑

a
113 C OROLLARY If θ is irrational prove that there exist infinitely many rational numbers , gcd(a, q) = 1, such that θ lies
q
ia 1 a 1 h
in the open intervals − 2
; + 2 .
q q q q

36
 Chapter 2

¯ ¯
¯ a r ¯¯ 1 ar
¯
Proof: Suppose that ¯θ − < for 1 ≤ r ≤ R. Since the differences θ − are non-zero, we may choose Q so
q r ¯ q r2 q
¯ r ¯
¯ a¯ 1
large in Theorem 112 that none of these rational numbers is a solution of ¯¯θ − ¯¯ < . Since this latter inequal-
q qQ ¯ ¯
¯ a ¯¯ 1
¯
ity does have a solution, the R given rational approximations do not exhaust the set of solutions of ¯θ − ¯ < 2 .
q q
❑

Homework

\ h i
Problem 2.2.1 Determine k ;1001 − k . Problem 2.2.8 Let
1≤k≤500
p p
Q + 2Q = {a + 2b : (a,b) ∈ Q2 }
∞ h
[ 1i
Problem 2.2.2 Determine 1 ;1 + .
k=1 k and define addition on this set as

∞ h i p p p
[ (a + 2b) + (c + 2d ) = (a + c) + 2(b + d ),
Problem 2.2.3 Determine − k ;k .
k=1
and multiplication as
∞ h
\ 1i
Problem 2.2.4 Determine 1 ;1 + . p p p
k=1 k (a + 2b)(c + 2d ) = (ac + 2bd ) + 2(ad + bc).

p
∞ h
\ h Then 〈Q + 2Q,·,+〉 is a field.
Problem 2.2.5 Determine k ;+∞ .
k=1
Problem 2.2.9 Put D = {x : x = q 2 or x = −q 2 , q ∈ Q}. Prove
∞ i
\ 1i that D is dense in R.
Problem 2.2.6 Determine 1 ;1 + .
k=1 k

h i h i Problem 2.2.10 A dyadic rational is a rational number of the form

m
Problem 2.2.7 Let I = a ;b , and I 0 = a 0 ;b 0 be closed intervals , where m ∈ Z, n ∈ N. Prove that the set of dyadic rationals is
2n
in R. Prove that I j I 0 if and only if a 0 ≤ a and b ≤ b 0 . dense in R.

2.3 Open and Closed Sets

Why bother? Many of the properties that we will study in these notes generalise to sets other than R. To better
understand what is it from the features of R that is essential for a generalisation, the language of topology is
used.

114 Definition The open ball Bx0 (r ) centred at x = x0 and radius ε > 0 is the set
i h
Bx0 (ε) = x0 − ε ; x0 + ε .

115 Definition A set Nx0 j R is an open neighbourhood of a point x0 if ∃ε > 0 such that Bx0 (r ) j Nx0 , that is, there is a
sufficiently small open ball containing x0 completely contained in Nx0 .

116 Definition A set U j R is said to be open in R if ∀ x ∈ U there is an open neighbourhood Nx0 such that Nx0 j U . A set
F ⊆ R is said to be closed in R if its complement U = R \ F is open in R.

117 T HEOREM Every open ball is open.

Proof: Let Bx0 (r ) with r > 0 be an open ball and let x ∈ Bx0 (r ). We must shew that there is a sufficiently small
neighbourhood of x completely within Bx0 (r ) . That is, we search for ε > 0 such that y ∈ Bx (ε) =⇒ y ∈ Bx0 (r ).

37
 Open and Closed Sets

Now,
y ∈ Bx (ε) =⇒ y ∈ Bx0 (r ) ⇐⇒ |y − x| < ε =⇒ |y − x0 | < r.
By the Triangle Inequality
|y − x0 | ≤ |y − x| + |x − x0 | < ε + |x − x0 |.
So, as long as
ε + |x − x0 | < r,
we will be within Bx0 (r ). One can take
r − |x − x0 |
ε= .
2
❑
i h i h i h i h
118 Example The open intervals a ;b , a ;+∞ , − ∞ ;b , − ∞ ;+∞ , are open in R.
h i h h i i i h
The closed intervals {a}, a ;b , a ;+∞ , − ∞ ;b , − ∞ ;+∞ = R, are closed in R.

The sets ∅ and R are simultaneously open and closed in R.

i i h h
The intervals a ;b and a ;b are neither open nor closed in R.

119 T HEOREM The union of any (finite or infinite) number of open sets in R is open in R. The union of a finite number of
closed in R sets is closed in R.

The intersection of a finite number of open sets in R is open in R. The intersection of any (finite or infinite) number of
closed sets in R is closed in R.

[
∞
Proof: Let U1 ,U2 ,... , be a sequence of open sets in R (some may be empty) and consider x ∈ Un . There is an
i n=1 h
index N such that x ∈ U N . Since U N is open in R, there is an open neighbourhood of x x − ε ; x + ε j U N , for
ε > 0 small enough. But then
i h [
∞
x − ε ; x + ε j UN j Un ,
n=1
and so given an arbitrary point of the union, there is a small enough open neighbourhood enclosing the point
and within the union, meaning that the union is open.

\
∞
If F n is an arbitrary intersection of closed sets, then there are open sets Un = R \ F n . By the De Morgan Laws,
n=1

\
∞ \
∞ [
∞
Fn = (R \Un ) = R \ Un ,
n=1 n=1 n=1

[
∞ \
∞
and since Un is open by the above paragraph, F n is the complement of an open set, that is, it is closed.
n=1 n=1

\L
Let U1 ,U2 ,... ,UL be a sequence of open sets in R and consider x ∈ Un . Then x belongs to each of the Uk and
i h n=1
so there are εk > 0 such that x ∈ x − εk ; x + εk j Uk . Let ε = min εk be the smallest one of such. But then for
1≤k≤L
all k,
i h i h i h \L
x − ε ; x + ε j x − εk ; x + εk j Uk , =⇒ x − ε ; x + ε j Un ,
n=1
and so given an arbitrary point of the intersection, there is a small enough open neighbourhood enclosing the
point and within the intersection, meaning that the intersection is open.

Using the De Morgan Laws and the preceding paragraph, the remaining statement can be proved. ❑

38
 Chapter 2

120 Example The intersection of an infinite number of open sets may not be open. For example
∞ i
\ 1 1 h h h
1− ;2 − = 1 ;2 ,
k=1 n +1 n +1

which is neither open nor closed.

121 T HEOREM (Characterisation of the Open Sets of R) A set A j R is open if an only if it is the countable union of open
sets of R.

2.4 Interior, Boundary, and Closure of a Set

122 Definition Let A j R. The interior of A is defined and denoted by
[
Å = Ω,
ΩjA
Ω open

that is, the largest open set inside A. The points of Å are called the interior points of A.

123 Definition Let A j R. The closure of A is defined and denoted by

[
A= Ω,
ΩkA
Ω closed

that is, the smallest closed set containing A. The points of A are called the adherence points of A.

☞ One always has Å j A j A. A set U is open if and only if U = Ů . A set F is closed if and only if F = F .

124 Definition Let A j R. The boundary of A is defined and denoted by

Bdy (A) = A − Å.

The elements of Bdy (A) are called the boundary points of A.

125 Example We have

i˜˚ i i h i i h i ³i i´
1. 0 ;1 = 0 ;1 , 0 ;1 = 0 ;1 , Bdy 0 ;1 = {0,1}

— ˚
2. {0,1} = ∅, {0,1} = {0,1}, Bdy ({0,1}) = {0,1}

3. Q̊ = ∅, Q = R, Bdy (Q) = R

126 T HEOREM Let A j R. Then

— ˚
R \ Å = R \ A, R\ A = R \ A.

Proof: The theorem follows from the De Morgan Laws, as

[ \ \ \
R \ Å = R \ Ω= (R \ Ω) = (R \ Ω) = F = R \ A,
ΩjA ΩjA R\A jR\Ω R\A jF
Ω open Ω open Ω open F closed

and
\ [ [ \ ˚
R\ A = R\ F= (R \ F ) = (R \ F ) = —
Ω=R \ A.
F kA F kA R\A kR\F R\A kΩ
F closed F closed F closed Ω open

39
 Interior, Boundary, and Closure of a Set

127 T HEOREM x ∈ A ⇐⇒ ∀Nx , Nx ∩ A 6= ∅. That is, x is an adherent point if and only if every neighbourhood of x has a
nonempty intersection with A.

i h i h i h
Proof: Assume x ∈ A and let r > 0. If x −r ; x +r ∩ A = ∅, then x −r ; x +r j R \ A. Since x −r ; x +r is open,
i h
— ˚
we have—in particular— x − r ; x + r j R \ A = R \ A by Theorem 126. This means that x 6∈ A, a contradiction.

—˚
Conversely, assume that for all neighbourhoods Nx of x we have Nx ∩ A 6= ∅. If x 6∈ A then x ∈ R \ A = R \ A.
i h i h
˚ ˚
—
Since R \ A is open there is an r 0 > 0 such that x − r 0 ; x + r 0 j R
— \ A j R \ A. But then x − r 0 ; x + r 0 ∩ A = ∅, a
contradiction. ❑

128 T HEOREM Let ∅ & A j R be bounded above. Then sup A ∈ A. If, moreover, A is closed then sup(A) ∈ A.

¯ ¯
Proof: Let r >i 0. By Theorem 93, there
h exists a ∈ A such that sup(A) − r < a, which gives ¯sup(A) − a ¯ < r .
This shews that sup A − r ;sup A + r ∩ A 6= ∅ regardless of how small r > 0 might be and, hence, sup(A) ∈ A by
Theorem 127. If A is closed, then A = A. ❑

129 Definition Let A j R. A point x ∈ A is called an isolated point of A if there exists an r > 0 such that Bx (r ) ∩ A = {x}. The
set of isolated points of A is denoted by A ∗ .

A point y ∈ R is called an accumulation point of A in R if

∀Nx , (Nx \ {x}) ∩ A 6= ∅,

that is, if any neighbourhood of x meets A at a point different than x. The set of accumulation points of A is called the
derived set of A and is denoted by Acc (A).

130 Example We have

h i
1. 0 is an isolated point of the set A = {0} ∪ 1 ;2 .
½ ¾
1 1 1
2. Every point of the set A = 1, , ,... is isolated. This is because we may take r = n+2 in the definition of isolated
2 3 ½ ¾ 2
i1 1 1 1 h 1 1 1 1 1 1 1
point, and then − n+2 ; + n+2 ∩ A = . Observe that − = and − = and
n 2 n 2 n n n + 1 n(n + 1) n − 1 n n(n − 1)
that 2n+2 > max(n(n + 1),n(n − 1)).
½ ¾
1 1
3. 0 is an accumulation point of A = 1, , ,... .
2 3

131 T HEOREM x is an accumulation point of A if and only if every neighbourhood of x in R has an infinite number of points
of A.

Proof: Suppose x ∈ A 0 . Suppose

³i a neighbourhood
h ´ of x had only finitely many elements of A, say {y 1 , y 2 ,... , y n }.
¯ ¯
Take 2r = min ¯ y k − x ¯. Then x − r ; x + r \ {x} ∩ A = ∅ contradicting the fact that every neighbourhood of x
1≤k≤n
meets A at a point different from x.

Conversely if every neighbourhood of x in R has an infinite number of points of A, then a fortiori, any intersection
of such a neighbourhood with A will contain a point different from x, and so x ∈ Acc(A). ❑

132 T HEOREM A set is closed if and only if it contains all its accumulation points.

40
 Chapter 2

i h
Proof: If A is closed then R \ A is open. If c ∈ R \ A then there exists r > 0 such that c − r ;c + r j R \ A, a
neighbourhood that clearly does not contain any points of A, which means c 6∈ Acc (A).

Conversely, suppose a set Acc (A) j A. We will h a fortiori, x 6∈ Acc(A).

i prove thath R \ A is open. If ix ∈ R \ A, then
This means that there is an r > 0 such that x − r ; x + r ∩ A = ∅. Hence x − r ; x + r j R \ A, and so R \ A is
open. ❑

☞ One has
A ∗ j A, A − A j Acc (A) , A ∗ ∩ Acc (A) = ∅, A ∗ ∪ Acc (A) = A.

2.5 Connected Sets

133 Definition A set X j R is connected if, given open sets U ,V of R with U ∪ V = X , U ∩ V = ∅, either U = ∅ or V = ∅.

134 T HEOREM If X j R is connected, and if (a,c) ∈ X 2 , b ∈ R, are such that a < b < c then b ∈ X .

135 C OROLLARY The only connected sets of R are the intervals. In particular, R is connected.

2.6 Compact Sets

[
∞
136 Definition A sequence of open sets U1 ,U2 ,... is said to be an open cover for A j R if A j Un . U1 ,U2 ,... has a subcover
n=1
[
∞
Uk1 ,Uk2 ,... of A if A j Ukn .
n=1

137 Definition A set of real numbers is said to be compact in R if every open cover of the set has a finite subcover.2

[i h i h
138 Example Since R = n − 1 ;n + 1 , the sequence of intervals n − 1 ;n + 1 , n ∈ Z is a cover for R.
n∈Z

h i
139 T HEOREM Let a,b be real numbers with a ≤ b. The closed interval a ;b is compact in R.

h i h i h i
Proof: Let U1 ,U2 ,... be an open cover for a ;b . Let E be the collection of all x ∈ a ;b such that a ; x has a
finite subcover from the Ui . We will shew that b ∈ E .

[
∞ h i
Since a ∈ Ui , there exists Ur such that a ∈ Ur . Thus {a} = a ; a j Ur and so E 6= ∅. Clearly, b is an upper
i =1
bound for E . By the Completeness Axiom, sup E exists. We will shew that b = sup E .

h i [ ∞
By Theorem 128, sup E ∈ a ;b j Ui , hence there exists U s such that sup E ∈ U s . Since U s is open, there exists
i i =1h
ε > 0 such that sup E − ε ;sup E + ε j U s . By Theorem 93 there is x ∈ E such that sup E − ε < x ≤ sup E . Thus
h i [n
there is a finite subcover from the Ui , say, U p 1 U p 2 , . . . , U p n such that a ; x j Uki .
i =1

We thus have à !
h i h i[i h [
n
a ;sup E j a ; x sup E − ε ;sup E + ε j Uki ∪U s ,
i =1

2
This definition is appropriate for R but it is not valid in general. However, it very handy for one-variable calculus, hence we will retain it.

41
 Compact Sets

a finite subcover. This means that sup E ∈ E .

1
Suppose now that sup E < b, and consider y = sup E + min(b − sup E ,ε). Then
2
à !
h i h i h i [
n
sup E < y, a ; y = a ;sup E ∪ sup E ; y j Uki ∪U s ,
i =1

whence y ∈ E , contradicting the definition of sup E . This proves that sup E = b and finishes the proof of the
theorem. ❑

140 T HEOREM (Heine-Borel) A set A of R is closed and bounded if and only if it is compact.

2
Proof: Let A be i and bounded in R, and let U1 ,U2 ,... , be an open cover for A. There exist (a,b) ∈ R , a ≤ b,
h closed
such that A j a ;b . Since
h i [
∞
a ;b j (R \ A) ∪ Ui ,
i =1

by Theorem 139 there is a finite subcover of the Ui , say, Uki such that
h i [
∞
a ;b j (R \ A) ∪ Uki .
i =1

Therefore
h i h i [∞
A = A ∩ a ;b j a ;b j Uki ,
i =1

and so A admits an open subcover.

i h
Conversely, suppose that every open cover of A admits a finite subcover. The open cover − n ;n ,n ∈ R of A must
i h
admit a finite subcover by our assumption, hence there is N ∈ N such that A j − N ; N , meaning that A is
bounded. Let us shew now that R \ A is open.

Let x ∈ R \ A. We have
µ h ¶
[ 1 1i \h 1 1i
R\ x − ;x + = R\ x − ;x + = R \ {x} k A,
n≥1 n n n≥1 n n

since x 6∈ A. By hypothesis there is N ∈ N and n 1 ,n 2 ,... ,n N such that

µ h ¶
[
N 1 1 i h 1 1 i
Aj R\ x − ;x + jR\ x − ;x + ,
k=1 nk nk nm nm
h 1 1 i
where m = max(n 1 ,n 2 ,... ,n N ). This gives x − ;x + j R \ A, meaning that R \ A is open, whence A is
nm nm
closed.
❑

141 C OROLLARY (Cantor’s Intersection Theorem) Let

h i h i h i
a 1 ;b1 k a 2 ;b2 k a 3 ;b3 k ...

be a sequence of non-empty, bounded, nested closed intervals. Then

∞ h
\ i
a j ;b j =6 ∅.
j =1

42
 Chapter 2

h i \∞ h i
Proof: Assume that a 1 ;b1 ∩ a j ;b j = ∅. Then
j =2

h i ∞ h
\ i [∞ ³ h i´
a 1 ;b1 j R \ a j ;b j = R \ a j ;b j .
j =2 j =2
h i h i
The R \ a j ;b j for an open cover for a 1 ;b1 , which is closed and bounded. By Theorem 7 we have
h i h i h i h i
a j ; a j j a i ;bi =⇒ R \ a i ;bi j R \ a j ;b j .

By the Heine-Borel Theorem 140 there is a finite subcover, say

h i [N ³ h i´ h i
a 1 ;b1 j R \ a n j ;bn j j R \ a n N ;bn N .
j =1
h i h i h i h i
But then a n N ;bn N j R \ a1 ;b1 , which contradicts an N ;bn N j a1 ;b1 , and the proof is complete.❑

142 T HEOREM (Bolzano-Weierstrass) Ever bounded infinite set of R has at least one accumulation point.

Proof: Let A be a bounded set of R with Acc(A) = ∅. Then A ∗ = A = A. Notice that then every element of A is an
isolated point of A, and hence,
i h
∀a ∈ A, ∃r a > 0, such that a − r a ; a + r a ∩ A = {a}.

Observe that [i h
Aj a − ra ; a + ra ,
a∈A
i h
and so the a − r a ; a + r a form an open cover for A. Since A = A, A is closed. By the Heine-Borel Theorem 140 A
i h
has a finite subcover from among the a − r a ; a + r a and so there exists an integer N > 0 and a i such that

N i
[ h
Aj ai − r ai ; ai + r ai .
i =1

Since
N i
[ h [N
A= A∩ ai − r ai ; ai + r ai = {a i },
i =1 i =1

A has only N elements and thus it is finite. ❑

143 T HEOREM Let X j R. Then the following are equivalent.

1. X is compact.

2. X is closed and bounded.

3. every infinite set of X has an accumulation point.

4. every infinite sequence of X has a converging subsequence in X .

Homework

Problem 2.6.1 Give an example shewing that the union of an infi- Problem 2.6.2 Prove that a set A j R is dense if and only if A = R.
nite number of closed sets is not necessarily closed.

Problem 2.6.3 For any set A j R prove that Bdy (A) = Bdy(R \ A).

43
 R

Problem 2.6.4 Let A 6= ∅ be a subset of R. Assume that A is bounded zero number (which may be assumed positive); or (iii) or all reals.
above. Prove that sup(A) = sup(A).
Problem 2.6.7 Let A ∈ R. Prove the following
Problem 2.6.5 Demonstrate that the only subsets of R which are si-
multaneously open and closed in R are ∅ and R. One codifies this by 1. A = A 6. A ∩ B j A ∩ B
saying that R is connected. 2. Å˚ = Å
˜ ˚
3. A j B =⇒ A j B 7. Å ∪ B̊ j A ∪B
Problem 2.6.6 Prove that the closed additive subgroups of the real 4. A j B =⇒ Å j B̊
˜ ˚
numbers are (i) just zero; or (ii) all integral multiples of a fixed non- 5. A ∪ B = A ∪ B 8. A ∩ B = Å ∩ B̊

2.7 R
Why bother? The algebraic rules introduced here will simplify some computations and statements in subse-
quent chapters.

Geometrically, each real number can be viewed as a point on a straight line. We make the convention that we orient the real
line with 0 as the origin, the positive numbers increasing towards the right from 0 and the negative numbers decreasing
towards the left of 0, as in figure 2.1.

-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7

−∞ +∞

Figure 2.1: The Real Line.

We append the object +∞, which is larger than any real number, and the object −∞, which is smaller than any real
number. Letting x ∈ R, we make the following conventions.

(+∞) + (+∞) = +∞ (2.1)

(−∞) + (−∞) = −∞ (2.2)

x + (+∞) = +∞ (2.3)

x + (−∞) = −∞ (2.4)

x(+∞) = +∞ if x > 0 (2.5)

x(+∞) = −∞ if x < 0 (2.6)

x(−∞) = −∞ if x > 0 (2.7)

x(−∞) = +∞ if x < 0 (2.8)

x
=0 (2.9)
±∞
Observe that we leave the following undefined:
±∞
, (+∞) + (−∞), 0(±∞).
±∞

44
 Chapter 2

h i
144 Definition We denote by R = − ∞ ;+∞ the set of real numbers such with the two symbols −∞ and +∞ appended,
obeying the algebraic rules above. Observe that then every set in R has a supremum (it may as well be +∞ if the set is
unbounded by finite numbers) and an infimum (which may be −∞).

2.8 Lebesgue Measure

i h ³i h´
145 Definition Let (a,b) ∈ R2 . The measure of the open interval a ;b is b − a. We denote this by µ a ;b = b − a. If
∞ i
[ h X
∞
G= a k ;bk is a union of disjoint, bounded, open intervals, then µ (G) = (bk − a k ).
k=1 k=1

146 Definition Let E j R be a bounded set. The outer measure of E is defined and denoted by

µ (E ) = inf µ (O) .
E jO
O open

147 Definition A set E j R is said to be Lebesgue measurable if ∀ε > 0, ∃G k E open such that µ (G \ E ) < ε. In this case
µ (E ) = µ (E ).

2.9 The Cantor Set

148 Definition (The Cantor Set) The Cantor set C is the canonical example of an uncountable set of measure zero. We
construct C as follows. h i i1 2h
Begin with the unit interval C 0 = 0 ;1 , and remove the middle third open segment R1 := ; . Define C 1 as
3 3
h 1i[h2 i
C 1 := C 0 \ R1 = 0 ; ;1 (2.10)
3 3
Iterate this process on each remaining segment, removing the open set
i 1 2h[i 7 8 h
R2 := ; ; (2.11)
9 9 9 9
to form the four-interval set h 1i[h2 1i[h2 7i[h8 i
C 2 := C 1 \ R2 = 0 ; ; ; ;1 (2.12)
9 9 3 3 9 9
Continue the process, forming C 3 ,C 4 ,... Note that C k has 2k pieces.
At each step, the endpoints of each closed segment will remain in the set. See figure 2.2.
The Cantor set is defined as
\
∞ [∞
C := Ck = C0 \ Rn (2.13)
k=1 n=1

149 T HEOREM (Cardinality of the Cantor Set) The Cantor Set is uncountable.

Proof: Starting with the two pieces of C 1 , we mark the sinistral segment “0” and the dextral segment “1”. We then
continue to C 2 , and consider only the leftmost pair. Again, mark the segments “0” and “1”, and do the same for
the rightmost pair. Successively then, mark the 2k−1 leftmost segments of C k “0” and the 2k−1 rightmost segments
“1.” The elements of the Cantor Set are those with infinite binary expansions. Since there uncountable many such
expansions, the Cantor Set in uncountable.❑

150 T HEOREM (Measure of the Cantor Set) The Cantor Set has (Lebesgue) measure 0.

45
 The Cantor Set

C0
0 1
C1
0 1 2 1
3 3
C2
0 1 2 1 2 7 8 1
9 9 3 3 9 9
.. ..
. .

Figure 2.2: Construction of the Cantor Set.

Proof: Using the notation of Definition 148, observe that

2 1 1
µ(R1 ) = − = (2.14)
µ3 3 ¶ 3 µ ¶
2 1 8 7 2
µ(R2 ) = − + − = (2.15)
9 9 9 9 9
..
. (2.16)
X
k n−1
2
µ(Rk ) = n
(2.17)
n=1 3

Since the R’s are disjoint, the measure of their union is the sum of their measures. Taking the limit as k → ∞,
µ ¶ ∞ n−1
[
∞ X 2
µ Rn = n
= 1. (2.18)
n=1 n=1 3

Since clearly µ(C 0 ) = 1, we then have

µ ¶
[∞ X∞ 1
µ(C ) = µ C 0 \ Rn = µ(C 0 ) − n
= 1 − 1 = 0. (2.19)
n=1 n=1 2

151 T HEOREM The Cantor set is closed and its interior is empty.

Proof: Each of C 0 ,C 1 ,C 2 ,..., is closed, being the union of a finite number of closed intervals. Thus the Cantor Set
is closed, as it is the intersection of closed sets.

m
Now, let I be an open interval. Since the numbers of the form n , (m,n) ∈ Z are dense in the reals, there is exists
3
m
a rational number n ∈ I . Expressed in ternary, this rational number has a finite expansion. If this expansion
3
contains the digit “1”, then this number does not belong to Cantor Set, and we are done. If not, since I is open,
m 1
there must exist a number k > n such that n + k ∈ I . By construction, the last digit of the ternary expansion of
3 3
this number is also “1”, and hence this number does not belong to the Cantor Set either.❑

46
Chapter 3

Sequences

3.1 Limit of a Sequence

Why bother? The limit concept is at the centre of calculus. We deal with discrete quantities first, that is, with
limits of sequences.

152 Definition A (numerical) sequence is a function a : N → R. We usually denote a(n) by a n .1

☞ We will use the notation {an }ln=k to denote the sequence ak , ak+1 ... , al . For example

{an }10
n=0 = {a 0 , a 1 , a 2 ,... , a 10 },

{bn }6n=4 = {b4 ,b5 ,b6 },

½µ ¶ ¾
1 n +∞ 9 64
1+ = {2, , ,... ,},
n n=1 4 27
etc.

153 Example The Harmonic sequence is

1 1
1, , , ... ,
2 3
1
or a n = for n ≥ 1.
n

154 Definition A sequence {a n }+∞

n=1 is bounded if there exists a constant K > 0 such that ∀n,|a n | ≤ K . It is increasing if for
all n > 0, a n ≤ a n+1 and decreasing if for all n ≥ 0, a n ≥ a n+1 .

3.2 Convergence of Sequences

155 Definition A sequence {a n }+∞
n=1 is said to converge if

∃L ∈ R,∀ε > 0, ∃N > 0 such that ∀n ∈ N, n ≥ N =⇒ |a n − L| < ε.

In other words, eventually2 the differences

|a n − L| ,|a n+1 − L| ,|a n+2 − L| ,...

remain smaller that an arbitrarily prescribed small quantity. We denote the fact that the sequence {a n }+∞
n=1 converges to L
as n → +∞ by
lim a n = L, or by a n → L as n → +∞.
n→+∞
1
It is customary to start at n = 1 rather than n = 0. We won’t be too fuzzy about such complications, but we will be careful to write sense.
2
A good word to use in informal speech “eventually” will mean “for large enough values” or in the case at hand ∀n ≥ N for some strictly positive integer
N.

47
 Convergence of Sequences

A sequence that does not converge is said to diverge. Thus a sequence diverges if

∀L ∈ R,∃ε > 0,∀N ∈ N,∃n ∈ N such that n > N and |a n − L| ≥ ε.

☞ Given a sequence sequence {an }+∞

n=1 and L ∈ R,

an → L as n → +∞ ifandonlyif lim inf an = lim sup an = lim an = L.

156 Definition A sequence {b n }+∞

n=1 diverges to plus infinity if ∀M > 0, ∃N > 0 such that ∀n ≥ N , bn > M. A sequence
{cn }+∞
n=1 diverges to minus infinity if ∀M > 0, ∃N > 0 such that ∀n ≥ N , cn < −M. A sequence that diverges to plus or
minus infinity is said to properly diverge. Otherwise it is said to oscillate.

157 Definition Given a sequence {a n }+∞

n=1 , we say that lim a n exists it is either convergent or properly divergent.
n→+∞

158 Example The constant sequence

1,1,1,1,...
converges to 1. Similarly, if a sequence is eventually stationary, that is, constant, it converges to that constant.

159 Example Consider the sequence

1 1 1
1, , ,... , ,... ,
2 3 n
1 1
We claim that → 0 as n → +∞. Suppose we wanted terms that get closer to 0 by at least .00001 = . We only need
n 105
1 1
to look at the 100000-term of the sequence: = . Since the terms of the sequence get smaller and smaller, any
100000 105
term after this one will be within .00001 of 0. We had to wait a long time—till after the 100000-th term—but the sequence
eventually did get closer than .00001 to 0. The same argument works for any distance, no matter how small, so we can
eventually get arbitrarily close to 0. A rigorous proof is as follows. If ε > 0 is no matter how small, we need only to look at
1
the terms after N = T + 1U to see that, indeed, if n > N , then
ε
1 1 1
sn = < = 1 < ε.
n N T ε + 1U

Here we have used the inequality

t − 1 < Tt U ≤ t , ∀t ∈ R.

160 Example The sequence

0,1,4,9,16,... ,n 2 ,...

p sequence gets arbitrarily large. A rigorous proof is as follows. If M > 0 is no matter how large, then
diverges to +∞, as the
the terms after N = T M U + 1 satisfy (n > N )
p
tn = n 2 > N 2 = (T M U + 1)2 > M.

161 Example The sequence

1,−1,1,−1,1,−1,... ,(−1)n ,...
has no limit (diverges), as it bounces back and forth from −1 to +1 infinitely many times.

162 Example The sequence

0,−1,2,−3,4,−5,... ,(−1)n n,... ,
has no limit (diverges), as it is unbounded and alternates back and forth positive and negative values..

We will now see some properties of limits of sequences.

48
 Chapter 3

163 T HEOREM (Uniqueness of Limits) If a n → L and a n → L 0 as n → +∞ then L = L 0 .

|L − L 0 |
Proof: The statement only makes sense if both L and L 0 are finite, so assume so. If L 6= L 0 , take ε = > 0 in
2
the definition of convergence. Now

lim a n = L =⇒ ∃N 1 > 0, ∀n ≥ N 1 |a n − L| < ε,

n→+∞
¯ ¯
lim a n = L 0 =⇒ ∃N 2 > 0, ∀n ≥ N 2 ¯a n − L 0 ¯ < ε.
n→+∞

If n > max(N 1 , N 2 ) then by the Triangle Inequality (Theorem 76) then

¯ ¯ ¯ ¯ ¯ ¯
¯L − L 0 ¯ ≤ |L − a n | + ¯a n − L 0 ¯ < 2ε = ¯L − L 0 ¯ ,

a contradiction, so L = L 0 . ❑

164 T HEOREM Every convergent sequence is bounded.

Proof: Let {a n }+∞

n=1 converge to L. Using ε = 1 in the definition of convergence, ∃N > 0 such that

n ≥ N =⇒ |a n − L| < 1 =⇒ L − 1 < a n < L + 1,

hence, eventually, a n does not exceed L + 1. ❑

| | | | | | |
x0 x1 x2 . . xn . . s
. .

Figure 3.1: Theorem ??.

When is it guaranteed that a sequence of real numbers has a limit? We have the following result.

165 T HEOREM Every bounded increasing sequence {a n }+∞

n=0 of real numbers converges to its supremum. Similarly, every
bounded decreasing sequence of real numbers converges to its infimum.

Proof: The idea of the proof is sketched in figure 3.1. By virtue of Axiom 92, the sequence has a supremum s.
Every term of the sequence satisfies a n ≤ s. We claim that eventually all the terms of the sequence are closer to s
than a preassigned small distance ε > 0. Since s − ε is not an upper bound for the sequence, there must be a term
of the sequence, say a n0 with s − ε ≤ a n0 by virtue of the Approximation Property Theorem 93. Since the sequence
is increasing, we then have

s − ε ≤ a n0 ≤ a n0 +1 ≤ a n0 +2 ≤ a n0 +2 ≤ ... ≤ s,
which means that after the n 0 -th term, we get to within ε of s.

To obtain the second half of the theorem, we simply apply the first half to the sequence {−a n }+∞
n=0 . ❑

166 T HEOREM (Order Properties of Sequences) Let {a n }+∞

n=1 be a sequence of real numbers converging to the real number
L. Then

1. If a < L then eventually a < a n .

2. If L < b then eventually a n < b.

3. If a < L < b then eventually a < a n < b.

4. If eventually a n ≥ a then L ≥ a.

49
 Convergence of Sequences

5. If eventually a n ≤ b then L ≤ b.

6. If eventually a ≤ a n ≤ b then a ≤ L ≤ b.

Proof: We apply the definition of convergence repeatedly.

1. Taking ε = L − a in the definition of convergence, ∃N 1 > 0 such that

∀n ≥ N 1 , |a n − L| < L − a =⇒ ∀n ≥ N 1 , a − L < a n − L < L − a =⇒ ∀n ≥ N 1 , a < an ,

that is, eventually a < a n .

2. Taking ε = b − L in the definition of convergence, ∃N 2 > 0 such that

∀n ≥ N 2 , |a n − L| < b − L =⇒ ∀n ≥ N 2 , L − b < a n − L < b − L =⇒ ∀n ≥ N 2 , a n < b,

that is, eventually a n < b.

3. It suffices to take N = max(N 1 , N 2 ) above.
4. If, to the contrary, L > a, then by part (1) we will eventually have a n > a, a contradiction.
5. If, to the contrary, L < b, then by part (2) we will eventually have a n < b, a contradiction.
6. If either L < a or b < L we would obtain a contradiction to parts (4) or (5).

167 T HEOREM (Sandwich Theorem) Let {a n }+∞ +∞ +∞

n=1 , {un }n=1 , {v n }n=1 be sequences of real numbers such that eventually

un ≤ a n ≤ v n .

If for L ∈ R, un → L and v n → L then a n → L.

Proof: For all ε > 0 there are N 1 > 0, N 2 > 0 such that

∀n ≥ max(N 1 , N 2 ), |un − L| < ε, |v n − L| < ε =⇒ −ε < un − L < ε, −ε < v n − L < ε.

Thus for such n,

−ε < un − L ≤ a n − L ≤ v n − L < ε, =⇒ −ε < a n − L < ε =⇒ |a n − L| < ε,

from where {a n }+∞

n=1 converges to L. ❑

168 T HEOREM Let {a n }+∞

n=1 be a sequence of real numbers such that a n → L. Then |a n | → |L|.

Proof: From Corollary 77, we have the inequality ||a n | − |L|| ≤ |a n − L| from where the result follows. ❑

169 T HEOREM Let {a n }+∞ +∞

n=1 be a sequence of real numbers such that a n → 0, and let {b n }n=1 be a bounded sequence. Then
a n bn → 0.

Proof: Eventually |a n | < ε. Assume that eventually |bn | ≤ U . Then

|a n bn | ≤ U |a n | < U ε,

from where the result follows. ❑

1 1
170 T HEOREM If b n → l 6= 0 then b n is eventually different from 0 and → .
bn l

50
 Chapter 3

|l |
Proof: By Theorem 169, |bn | → |l |. Using ε = > 0 in the definition of convergence, we have that eventually
2
|l | |l | |l | |l |
||bn | − |l || < =⇒ |l | − < |bn | < |l | + =⇒ < |bn | ,
2 2 2 2
1 1 2
That is, eventually |bn | is strictly positive and so makes sense. Also, eventually, < Now, for sufficiently
bn |bn | |l |
large n, ¯ ¯ ¯ ¯
¯ 1 1 ¯¯ ¯¯ l − bn ¯¯ |bn − l | 2ε
¯ − = = < ,
¯b l ¯ ¯ |bn ||l | ¯ |bn | |l | |l | |l |
n

from where the result follows.❑

171 T HEOREM (Algebraic Properties of Sequences) Let k ∈ R. If {a n }+∞ +∞ 0

n=1 converges to L and {b n }n=1 converges to L then

lim (k a n + bn ) = kL + L 0 , lim (a n bn ) = LL 0 .
n→+∞ n→+∞

Moreover, if L 0 6= 0 then µ ¶
an L
lim = 0.
n→+∞ bn L

Proof: The trick in all these proofs is the following observation: If one multiplies a bounded quantity by an
arbitrarily small quantity, one gets an arbitrarily small quantity. Hence once considers the absolute value of the
desired terms of the sequence from the expected limit.

¯ ¯
Given ε > 0 there are N 1 > 0 and N 2 > 0 such that |a n − L| < ε and ¯bn − L 0 ¯ < ε. Then
¯ ¯ ¯ ¯ ¯ ¯
¯(k a n + bn ) − (kL + L 0 )¯ = ¯(k a n − kL) + (bn − L 0 )¯ ≤ |k| |a n − L| + ¯bn − L 0 ¯ < ε(k + 1),

and so the sinistral side is arbitrarily close to 0, establishing the first assertion.

For the product, observe that by Theorem 164 there exists a constant K > 0 such that |bn | < K . Hence
¯ ¯ ¯ ¯ ¯ ¯
¯a n bn − LL 0 ¯ = ¯(a n − L)bn + L(bn − L 0 )¯ ≤ |a n − L| |bn | + |L| ¯bn − L 0 ¯ < εK + |L| ε = ε(K + |L|),

and again, the sinistral side is made arbitrarily close to 0.

1 1
Finally, if L 0 6= 0 then by Theorem 170, bn is eventually 6= 0 and → 0 . We now simply apply the result we
bn L
obtained for products, giving µ ¶
1 L
an bn → L = 0.
L0 L

Homework

Problem 3.2.1 If ∀n > 0, an > 0 and {an }+∞

n=1 converges to L must Problem 3.2.4 Prove that if an → +∞ and if there exists K > 0 such
it be the case that L > 0? that eventually bn ≥ K , then an bn → +∞.

Problem 3.2.2 Prove that if an → +∞ and if {bn }+∞

n=1 is bounded, Problem 3.2.5 Prove that if an → +∞ and bn → +∞ is bounded,
then an + bn → +∞. then an bn → +∞.

Problem 3.2.3 Prove that if an → +∞ and bn → +∞ is bounded, Problem 3.2.6 Prove that if an → +∞ and if {bn }+∞
n=1 is bounded,
then an + bn → +∞. then an + bn → +∞.

51
 Convergence of Sequences

µ ¶
1 1 1 1
Problem 3.2.7 Prove that if an → +∞ then → 0. 2. lim + +··· + ,
an n→+∞ 1 · 2 2·3 n(n + 1)
µ ¶
1 1 1
3. lim + +··· + ,
Problem 3.2.8 Prove that if an → 0 and if eventually an > 0, then n→+∞ 1 · 2 · 3 2·3·4 n(n + 1)(n + 2)
1
→ +∞.
an Problem 3.2.22 What reasonable meaning can be given to
1
n
X n ?
Problem 3.2.9 Prove that → 1 as n → +∞. 1
2 1+
i =1 n + i 1
1+
1
1+
1 ..
Problem 3.2.10 Prove that → 0. .
(n!)1/n
Problem 3.2.23 Let K ∈ N \ {0}, and let a1 ,... , aK ,λ1 ,... ,λK be
2n strictly positive real numbers. Prove that
Problem 3.2.11 Prove that → 0.
n! à !1/n à !−1/n
K
X K
X
lim λk akn = max ak , lim λk ak−n = min ak .
Problem 3.2.12 Let x1 , x2 ,... be a bounded sequence of real num- n→+∞ 1≤k≤K n→+∞ 1≤k≤K
k=1 k=1
s 2
bers, and put s n = x1 + x2 + · · · + xn . Suppose that n2 → 0. Prove
n ½ ¾
sn an +∞
that → 0. Problem 3.2.24 Prove that if is a monotonic sequence,
n b
½ ¾+∞ n n=1
a1 + a2 + · · · + an
then the is also monotonic in the same
Problem 3.2.13 Prove rigorously that the sequence {sinn}+∞
n=0 is di-
b1 + b2 + · · · + bn n=1
sense.
vergent.

Problem 3.2.25 Let a,b,c be real numbers such that b 2 − 4ac < 0.
Problem 3.2.14 Prove that (n!)1/n → +∞ as n → +∞.
Let {X n }+∞ +∞
n=1 , {Yn }n=1 be sequences of real numbers such that

Problem 3.2.15 A sequence of real numbers a1 , a2 ,... satisfies, for a X n2 + bX n Yn + cYn2 → 0, as n → +∞.
all m,n, the inequality Prove that X n → 0 and Yn → 0 as n → +∞.
1
|am + an − am+n | ≤ .
m +n Problem 3.2.26 (Gram’s Product) Prove that
Prove that this sequence is an arithmetic progression. n k3 − 1
Y 2
lim 3
= .
n→+∞
k=2 k + 1 3
p p
Problem 3.2.16 Prove rigorously that n + 1 − n → 0 as n → +∞.
1
Problem 3.2.27 Prove that the sequence {xn }+∞
n=1 with xn = 1+ 2 +
1 1 1 2
Problem 3.2.17 Prove that the sequence Hn = 1 + + + · · · + 1 1 1
2 3 n +· · ·+ 2 satisfies xn ≤ 2− for n ≥ 1. Hence deduce that it con-
diverges to +∞. 32 n n
verges.

Problem 3.2.18 Find

Problem 3.2.28 Prove the convergence of the sequence xn =
K
p Xn 1
X (n − 1)!
lim p p p p . , n ≥ 1.
K →+∞ n=1 (1 + 1)(1 + 2)(1 + 3) · · · (1 + n) k=1 n + k

Problem 3.2.19 What reasonable meaning can be given to Problem 3.2.29 Prove the convergence of the sequence, x1 = a, x2 =
xn + xn−1
s b, xn+1 = , n ≥ 2 and (a,b) ∈ R2 , a 6= b. Also, find its
r q 2
p limit.
1+ 1+ 1+ ··· ?

Problem 3.2.30
µ Prove
¶ the convergence of the sequence, x1 = a,
Problem 3.2.20 Prove that 1 b
xn+1 = xn + , n ≥ 1 and (a,b) ∈ R2 , a > 0,b > 0. Also, find its
1+2+··· +n 1 2 xn
→ , as n → +∞. limit.
n2 2

Problem 3.2.21 Calculate the following limits: Problem 3.2.31

µ Prove
¶ the convergence of the sequence, x1 = a,
à ! 1 b
12 22 (n − 1)2 xn+1 = xn + , n ≥ 1 and (a,b) ∈ R2 , a < 0,b > 0. Also, find its
1. lim + 2 +··· + , 2 xn
n→+∞ n 2 n n2 limit.

52
 Chapter 3

a +b 1. {an }+∞
Problem 3.2.32 Let (a,b) ∈ R2 , a > b > 0. Set a1 = , b1 = n=1 is monotonically decreasing,
p 2
ab. If for n > 1, 2. {bn }+∞
n=1 is monotonically inreasing,
an + bn p
an+1 = , bn = an bn , 3. both sequences converge,
2
Prove that 4. their limits are equal.

3.3 Classical Limits of Sequences

Why bother? In this section we will obtain various classical limits. In particular, we define the constant e and
obtain a few interesting results about it.

172 T HEOREM Let r ∈ R be fixed. If |r | < 1 then r n → 0 as n → +∞. If |r | > 1 then r n → +∞ as n → +∞.

¯ ¯
¯1¯
Proof: Taking x = ¯¯ ¯¯ − 1 in Bernoulli’s Inequality (Theorem 81), we find
r
¯ ¯n µ¯ ¯ ¶ µ¯ ¯ ¶
¯1¯ ¯ ¯ ¯ ¯
¯ ¯ > 1+n ¯1¯−1 > n ¯1¯−1 .
¯r ¯ ¯r ¯ ¯r ¯

Therefore
|r |
|r |n < → 0,
n(1 − |r |)
1
as n → +∞, since → 0 as n → +∞.
n

If |r | > 1, again by Bernoulli’s Inequality

|r |n = (1 + |r | − 1)n > 1 + n(|r | − 1),

and the dextral side can be made arbitrarily large.❑

173 T HEOREM Let |r | < 1. Then

1
1 +r +r 2 +··· +r n → , as n → +∞.
1−r

Proof: If S n = 1 + r + r 2 + · · · + r n then r S n = r + r 2 + r 3 + + · · · + r n+1 and

1 − r n+1
S n − r S n = 1 − r n+1 =⇒ S n = .
1−r
Then apply Theorem 172. ❑

☞ An estimating trick that we will use often is the following. If 0 < r < 1 then the truncated sum is smaller than the infinite
sum and so, for all positive integers k:

1
1+r +r 2 +··· +r k < 1+r +r 2 +··· +··· = .
1−r

174 T HEOREM Let a ∈ R, a > 0, be fixed. Then a 1/n → 1 as n → +∞.

Proof: If a > 1 then a 1/n > 1 and by Bernoulli’s Inequality,

a −1
a = (1 + (a 1/n − 1))n > 1 + n(a 1/n − 1) =⇒ 0 ≤ a 1/n − 1 < ,
n

whence a 1/n − 1 → 0 as n → +∞.

53
 Classical Limits of Sequences

1
If 0 < a < 1 then b = > 1 and so by what we just proved,
a
1
b 1/n → 1 =⇒ → 1 =⇒ a 1/n → 1,
a 1/n
proving the theorem.❑

an
175 T HEOREM Let a ∈ R, a > 1, k ∈ N \ {0}, be fixed. Then → +∞ as n → +∞.
nk

Proof: Observe that a 1/k > 1. We have, using the Binomial Theorem,
à !
³ ´n ³ ´n X n n
1/k 1/k
a = 1 + (a − 1) = (a 1/k − 1)i .
i =0 i

Since each term of the above expansion is ≥ 0, we gather that

¡ 1/k ¢n ¡ 1/k ¢n
³ ´n n(n − 1) a (n − 1) 1/k a an
1/k 1/k 2 2
a ≥ (a − 1) =⇒ ≥ (a − 1) =⇒ → +∞ =⇒ k → +∞,
2 n 2 n n
as desired. ❑

☞ In particular
2n
n
→ +∞ as n → +∞.

an
176 T HEOREM Let a ∈ R, , be fixed. Then → 0 as n → +∞.
n!

Proof: Put N = T|a|U + 1 and let n ≥ N . Then

¯ n¯ µ ¶µ ¶ µ N ¶µ ¶
¯a ¯
¯ ¯ = |a| · |a| · · · |a| |a|
·
|a|
· · ·
|a|
≤
|a|
1 · 1 · · · 1 ·
|a|
→ 0,
¯ n! ¯ 1 2 N N +1 N +2 n N! n
as n → +∞. ❑

177 T HEOREM The sequence µ ¶

1 n
en = 1 + ,n = 1,2,...
n
µis a bounded
¶ increasing sequence, and hence it converges to a limit, which we call e. Also, for all strictly positive integers n,
1 n
1+ < e.
n

Proof: By Theorem 80
b n+1 − a n+1
≤ (n + 1)b n =⇒ b n [(n + 1)a − nb] < a n+1 .
b−a
1 1
Putting a = 1 + , b = 1 + we obtain
n +1 n
µ ¶ µ ¶
1 n 1 n+1
en = 1 + < 1+ = e n+1 ,
n n +1
1
whence the sequence e n ,n = 1,2,... increases. Again, by putting a = 1, b = 1 + we obtain
2n
µ ¶ µ ¶
1 n 1 2n
1+ < 2 =⇒ 1 + < 4 =⇒ e 2n < 4.
2n 2n
Since e n < e 2n < 4 for all n, the sequence is bounded above. In view of Theorem 165 the sequence converges to
a limit. Weµ call this
¶ limit e. It follows also from this proof and from Theorem 166 that for all strictly positive
1 n
integers n, 1 + < e. ❑
n

54
 Chapter 3

µ ¶ µ ¶
☞ Another way of obtaining 1 +
1 n
n
< 1+
1 n+1
n +1
is as follows. Using the AM-GM Inequality with x1 = 1, x2 = · · · =
1
xn+1 = 1 + we have
n
µ ¶
1
µ ¶n/(n+1) 1 + n 1 + µ ¶ µ ¶
1 n 1 n/(n+1) n +2 1
1+ < =⇒ 1 + < = 1+
n n +1 n n +1 n +1

from where the desired inequality is obtained.

½µ ¶n+1 ¾+∞ µ ¶
1 1 n+1
178 T HEOREM The sequence 1+ is strictly decreasing and 1 + → e. Also, for all strictly positive inte-
n n=1 n
µ ¶
1 n+1
gers n, 1 + > e.
n

Proof: By Theorem 80
b n+1 − a n+1
≥ (n + 1)a n .
b−a
1 1
Putting a = 1 + , b = 1 + we obtain
n +1 n
µ ¶ µ ¶ µ ¶
1 n+1 1 n+2 n 3 + 4n 2 + 4n + 1
1+ > 1+ .
n n +1 n(n + 2)2
µ ¶
n 3 + 4n 2 + 4n + 1
The result will follow as long as > 1. But
n(n + 2)2

n 3 + 4n 2 + 4n + 1
n(n + 2)2 = n(n 2 + 4n + 4) = n 3 + 4n 2 + 4n < n 3 + 4n 2 + 4n + 1 =⇒ > 1.
n(n + 2)2

1
Thus the sequence is a sequence of strictly decreasing sequence of real numbers. Putting a = 1, b = 1 + in
n
b n+1 − a n+1
≥ (n + 1)a n we get
b−a µ ¶
1 n+1
1+ > 1 + n(n + 1) > 2,
n
so the sequence is bounded below. In view of Theorem 165 the sequence converges to a limit L. To see that L = e
observe that µ ¶ µ ¶ µ ¶
1 n+1 1 n 1
1+ = 1+ 1+ → e · 1 = e.
n n n
µ ¶
1 n+1
It follows also from this proof and from Theorem 166 that for all strictly positive integers n, 1 + > e. ❑
n
µ ¶ µ ¶
☞ The inequality 1 +
1 n+2
n +1
< 1+
1 n+1
n
can be obtained by the Harmonic Mean-Geometric Mean Inequality by
1
putting x1 = 1, x2 = x2 = · · · = xn+2 = 1 +
n
µ ¶
n +2 1/(n+2) n +2 1 (n+1)/(n+2)
1 1 1
≤ (x x
1 2 · · · x n+2 ) =⇒ ³ n ´ < 1 + .
n
x 1 + x 2 + · · · + x n+2 1 + (n + 1)
n +1

179 T HEOREM 2 < e < 3.

Proof: By the Binomial Theorem à !

µ ¶
1 n X n n 1
1+ = · k.
n k=0 k n

55
 Classical Limits of Sequences

Now, for 2 ≤ k ≤ n,
à ! µ ¶µ ¶ µ ¶
n 1 1 n(n − 1)(n − 2) · · · (n − k + 1) 1 1 2 k −1 1 1
· k = · k
= · (1) · 1 − 1 − · 1 − ≤ ≤ k−1 .
k n k! n k! n n n 2 ·3···k 2

Thus
µ ¶n à !
1 Xn n 1 1 1 1 1 1 1
1+ = · k ≤ 1 + 1 + + + · · · + n−1 < 1 + 1 + + + · · · + n−1 + · · · < 1 + 2 = 3,
n k=0 k n 2 4 2 2 4 2

1
by Theorem 173 (with r = ), and so the dextral inequality is proved. The sinistral inequality follows from Theo-
2
rem 177. ❑

☞ e = 2.718281828459045235360287471352 . .. .
µ ¶
1 1 1 1
180 T HEOREM e = lim 1+ + + +··· + .
n→+∞ 1! 2! 3! n!

1 1 1 1 © ª+∞
Proof: Put y k = 1 + + + + · · · + . Clearly y k+1 > y k so that y k k=1 is an increasing sequence. We will
1! 2! 3! k!
prove that it is bounded above with supremum e. By the Binomial Theorem
µ ¶ Ã ! Ã ! Ã ! Ã ! Ã ! Ã !
1 n X n n 1 n 1 n 1 n 1 n 1 n 1
1+ = · j = 1+ +··· + k
+··· + n
≥ 1+ +··· + ,
n j =0 j n 1 n k n n n 1 n k nk

for 0 < k < n. Now let j be fixed, 0 < j < n. Taking limits as n → +∞,
à ! µ ¶µ ¶ µ ¶ à !
n 1 1 n(n − 1)(n − 2) · · · (n − k + 1) 1 1 2 j −1 n 1 1
· = · = · (1) · 1 − 1− · 1− =⇒ lim · j = .
j nj j! nj j! n n n n→+∞ j n j!

Hence, taking limits as n → +∞,

µ ¶ Ã ! Ã !
1 n n 1 n 1 1 1 1 1
1+ ≥ 1+ +··· + =⇒ e ≥ 1 + + + + · · · + = yk ,
n 1 n k nk 1! 2! 3! k!

or renaming,
1 1 1 1
e ≥ 1+
+ + +··· + = yn . (3.1)
1! 2! 3! n!
à ! µ ¶µ ¶ µ ¶
n 1 1 1 2 k −1 1 1
Moreover, since · k = · (1) · 1 − 1− · 1− ≤ ≤ , we have
k n k! n n n 2 · 3 · · · k k!

µ ¶ Ã ! Ã ! Ã !
1 n n 1 n 1 n 1
1+ = 1+ +··· + +··· +
n 1 n k n k n nn
1 1 1
≤ 1+ +··· + +··· +
1! k! n!
= yn . (3.2)

In conclusion, from 3.1 and 3.2 we get

µ ¶n
1
1+ ≤ y n ≤ e,
n
and by taking limits and using the Sandwich Theorem, we get that y n → e as n → +∞. ❑

1 1 1 1 1
181 L EMMA Let n,m be strictly positive integers and let 1 + + + +··· + = y n . Then y m+n − y n < .
1! 2! 3! n! n · n!

56
 Chapter 3

Proof: We have
1 1 1 1
y m+n − y n = + + +··· +
(n + 1)! (n + 2)! (n + 3)! (n + m)!
µ ¶
1 1 1 1
< 1+ + · · · +
(n + 1)! n + 2 (n + 2)2 (n + 2)m−1
µ ¶
1 1 1
< 1+ + + · · · + · · ·
(n + 1)!  n + 2  (n + 2)2

1  1 
=  
(n + 1)!  1 
1−
n +2
1 n +2
= · .
(n + 1)! n + 1
Here the second inequality follows by using the estimating trick deduced from Theorem 173. Observe that this
bound is independent of m. ❑

1 1 1 1 1
182 L EMMA Let 1 + + + +··· + = y n . Then 0 < e − y n < .
1! 2! 3! n! n!n

Proof: From Lemma 181,

1 n +2
0 < y m+n − y n < · .
(n + 1)! n + 1
Taking the limit as m → +∞ we deduce

1 n +2
0 < e − yn ≤ · .
(n + 1)! n + 1

(The first inequality is strict by Theorem 180.) We only need to shew that for integer n ≥ 1

1 n +2 1
· < .
(n + 1)! n + 1 n!n

But working backwards (which we are allowed to do, as all quantities are strictly positive),

1 n +2 1
· < ⇐ n!n(n + 2) < (n + 1)!(n + 1)
(n + 1)! n + 1 n!n
⇐ n(n + 2) < (n + 1)(n + 1)

⇐ n 2 + 2n < n 2 + 2n + 1

⇐ 0 < 1,

and the theorem is proved. ❑

183 T HEOREM e is irrational.

p
Proof: Assume e is rational, with e = , where p and q are positive integers and the fraction is in lowest terms.
q
Since qe = p, an integer, q!e must also be an integer. Also q!y q must be an integer, since
µ ¶
1 1 1 1
q!y q = q! 1 + + + + · · · + .
1! 2! 3! q!

But by Lemma 182,

1 1
0 < e − yq < =⇒ 0 < q!(e − y q ) < ≤ 1.
q!q q
That is, the integer q!(e − y q ) is strictly between 0 and 1, a contradiction. ❑

57
 Classical Limits of Sequences

n o+∞
184 T HEOREM The sequence n 1/n is decreasing for n ≥ 3. Also, n 1/n → 1 as n → +∞.
n=1

Proof: Consider the ratio

µ ¶
(n + 1)n 1 n 1 e
= 1+ · < .
n n+1 n n n

Thus for n ≥ 3,
(n + 1)n
< 1 =⇒ (n + 1)1/(n+1) < n 1/n .
n n+1
Hence we have
31/3 > 41/4 > 51/5 > · · · .

Clearly, if n > 1 then n 1/n > 11/n = 1. Also, by the Binomial Theorem, again, if n > 1,
à r !n à !Ãr !1 à ! Ãr !2 à !Ãr !2 µ ¶
2 n n 2 n 2 n 2 n(n − 1) 2
1+ =1 + + +··· > 1 + = 1+ = n.
n 1 n 2 n 2 n 2 n

We then conclude that

r
1/n 2
1<n < 1+ ,
n

and that n 1/n → 1 follows from the Sandwich Theorem. ❑

☞ 21/2 = 41/4 .

Homework

½ ¾+∞
Problem 3.3.1 What’s wrong with the following? Since the product Problem 3.3.4 Prove that the sequence 1 − 1 + 1 − 1 + · · · + 1 − 1
of the limits is the limit of the product, 2 3 4 2n − 1 2n n=1
converges to log 2.
µ ¶n µ ¶ µ ¶ µ ¶
1 1 1 1
e = lim 1 + = lim 1 + · lim 1 + · · · lim 1 + =1| · 1·
{z· · 1} = 1.
n→+∞ n n→+∞ n n→+∞ n n→+∞ n
| {z } Problem
n times 3.3.5 Let n be a strictly positive integer and let xn denote
n times the unique real solution of the equation x n +x +1. Prove that xn → 1
as n → +∞.
Problem 3.3.2 Demonstrate that for all strictly positive integers n:
s r Problem 3.3.6 Let
q
π 1 p v
cos n+1 = 2 + 2 + 2 + · · · + 2, u s r
2 2| u q
{z } t p
n radicands an = n + (n − 1) + (n − 2) + · · · + 2 + 1,
s r q p 1
π 1 p
sin n+1 = 2− 2+ 2+··· + 2. for n ≥ 1. Prove that an − n → .
2 2| 2
{z }
n radicands
Hence deduce Viète’s Formula for π: Problem 3.3.7 Prove that e is not a quadratic irrational.
s r q n µ ¶
p Y k
π = lim 2n 2− 2+ 2+··· + 2. Problem 3.3.8 Find lim 1+ .
n→+∞ n→+∞ n
| {z } k=1
n radicands

( )+∞ Problem 3.3.9 A quadratic integer is any number x that satisfies an

2n 1
X equation
Problem 3.3.3 Prove that the sequence converges to
x 2 + mx + n = 0, (m,n) ∈ Z2 .
k=n k n=1
log 2. Prove that the real quadratic integers are dense in the reals.

58
 Chapter 3

3.4 Averages of Sequences

Why bother? In this section we will examine some classical results that allow us to compute more complicated
limits. Had we the language of matrices, most results here could be deduced from a classical result of Toeplitz.
Since we don’t, we will develop ad hoc methods which are interesting by themselves.

We start with the following discrete analogues of L’Hôpital’s Rule.

© ª+∞
185 T HEOREM Let {xn }+∞
n=1 , y n n=1 , be two sequences of real numbers such that xn → 0, y n → 0. Suppose, moreover, that
the xn are eventually strictly decreasing. Then
xn − xn−1 xn
lim = lim ,
n→+∞ y n − y n−1 n→+∞ yn
provided the sinistral limit exists (be it finite or +∞).

xn−1 − xn xn − xn−1
Proof: Assume first that = → L, a finite real number. Then, given ε > 0 we can find N > 0
y n−1 − y n y n − y n−1
such that for n > N ,
xn−1 − xn
L −ε < < L + ε, y n < y n−1 .
y n−1 − y n
Thus (L − ε)(y n−1 − y n ) < xn−1 − xn < (L + ε)(y n−1 − y n ), and repeating this inequality for n + 1,n + 2,... ,n + m,

(L − ε)(y n − y n+1 ) < xn − xn+1 < (L + ε)(y n − y n+1 ),

(L − ε)(y n+1 − y n+2 ) < xn+1 − xn+2 < (L + ε)(y n+1 − y n+2 ),
..
.

(L − ε)(y m+n−1 − y m+n ) < xm+n−1 − xm+n < (L + ε)(y m+n−1 − y m+n ).

Adding columnwise,
(L − ε)(y n − y m+n ) < xn − xm+n < (L + ε)(y n − y m+n ).
Letting m → +∞, and since the y n are strictly positive,
xn xn
(L − ε)y n < xn < (L + ε)y n =⇒ L − ε < < L + ε =⇒ →L
yn yn
as n → +∞.

xn−1 − xn
If diverges to +∞ then for all M > 0 we can find N 0 > 0 such that for all n ≥ N 0 ,
y n−1 − y n
xn−1 − xn
> M =⇒ xn−1 − xn > M(y n−1 − y n ).
y n−1 − y n
Reasoning as above, for positive integers m ≥ 0,

xn − xm+n > M(y n − y m+n ).

Taking the limit as m → +∞,

xn xn
xn ≥ M y n =⇒ ≥ M =⇒ → +∞.
yn yn
❑

186 T HEOREM (Stolz’s Theorem) Let {a n }+∞ +∞ +∞

n=1 , {b n }n=1 , be two sequences of real numbers. Suppose that {b n }n=1 is strictly
increasing for sufficiently large n and that bn → +∞ as n → +∞. Then
a n − a n−1 an
lim = lim ,
n→+∞ bn − bn−1 n→+∞ bn
provided the sinistral side exists (be it finite or infinite).

59
 Averages of Sequences

a n − a n−1
Proof: Assume first that → L, finite. Then for every ε > 0 there is N > 0 such that (∀)n ≥ N ,
bn − bn−1
a n+1 − a n
L −ε < < L + ε, bn+1 > bn .
bn+1 − bn

This means that

(L − ε)(bn+1 − bn ) < a n+1 − a n < (L + ε)(bn+1 − bn )

By iterating the above relation for N + 1, N + 2,... ,m + N we obtain

(L − ε)(b N +1 − b N ) < a N +1 − a N < (L + ε)(b N +1 − b N ),

(L − ε)(b N +2 − b N +1 ) < a N +2 − a N +1 < (L + ε)(b N +2 − b N +1 ),

..
.

(L − ε)(bm+N − bm+N −1 ) < a m+N − a m+N −1 < (L + ε)(bm+N − bm+N −1 ).

Adding columnwise,
¯ ¯
¯ a m+N − a N ¯
(L − ε)(bm+N − b N ) < a m+N − a N < (L + ε)(bm+N − b N ) =⇒ ¯¯ − L ¯¯ < ε.
bm+N − b N

Now, µ ¶µ ¶
a m+N a N − Lb N bN a m+N − a N
−L = + 1− −L ,
bm+N bm+N bm+N bm+N − b N
so by the Triangle Inequality
¯ ¯ ¯ ¯ ¯ ¯¯ ¯
¯ a m+N ¯ ¯ a N − Lb N ¯ ¯ ¯¯ ¯
¯ − L ¯≤¯ ¯ + ¯1 − b N ¯ ¯ a m+N − a N − L ¯ .
¯b ¯ ¯ b ¯ ¯ ¯ ¯
bm+N bm+N − b N ¯
m+N m+N

a N − Lb N bN
Since N is fixed, → 0 and → 0 as m → +∞ Thus the dextral side is arbitrarily small, proving
bm+N bm+N
am
that → L as m → +∞.
bm

a n − a n−1
Assume now that → +∞. For sufficiently large n thus a n −a n−1 > bn −bn−1 . Thus the a n are eventually
bn − bn−1
bn
increasing and a n → +∞ as n → +∞. Applying the results above to the we obtain
an

bn bn − bn−1
lim = lim =0
n→+∞ a n n→+∞ a n − a n−1
an
and so lim = +∞ too. ❑
n→+∞ bn

187 T HEOREM (Cèsaro) If a sequence of real numbers converges to a number, then its sequence of arithmetic means con-
x1 + x2 + · · · + xn
verges to the same number, that is, if xn → a then → a.
n

First Proof: Let a n = x1 + x2 + · · · + xn and bn = n in Stolz’s Theorem. ❑

Second Proof: It is instructive to give an ad hoc proof of this result. Given ε > 0 there exists N > 0 such that if
n ≥ N then |xn − a|. Then
¯ x + x +··· + x ¯ ¯¯ (x − a) + (x − a) + · · · + (x − a) ¯¯ |(x − a)| + |(x − a)| + · · · + |(x − a)|
¯ 1 2 n ¯ 1 2 n 1 2 n
¯ − a ¯ = ¯¯ ¯≤
¯ .
n n n

60
 Chapter 3

Now we run into a slight problem. We can control the differences |xk − a| after a certain point,p but the early
differences need to be taken
p care of. To this end we consider the differences x k − a with k ≤ T n U or k > Tn U
where n is so large that T n U ≥ N . We have
¯ ¯
¯ ¯
|(x1 − a)| + |(x2 − a)| + · · · + |(xn − a)| |(x1 − a)| + |(x2 − a)| + · · · + ¯(xTpn U − a)¯
=
n ¯ ¯ n
¯ p ¯
¯(xT n U+1 − a)¯ + |(x2 − a)| + · · · + |(xn − a)|
+
p n p
T n U max1≤k≤Tpn U |xk − a| (n − T n U)ε
< + .
n n

These two last quantities tend to 0 as n → +∞, from where the result follows. ❑

1 + 21/2 + 31/3 + · · · + n 1/n

188 Example Since n 1/n → 1, → 1.
n

1 1 1
1 1 + + +··· +
189 Example Since → 0,
2 3 n → 0.
n n

µ ¶ µ ¶ µ ¶ µ ¶
1 1 1 2 1 3 1 n
µ ¶n 1+ + 1+ + 1+ +··· + 1 +
1 1 2 3 n
190 Example Since 1 + → e, → e.
n n

191 Example The converse of Cèsaro’s Theorem is false. For, the sequence a n = (−1)n oscillates and does not converge. But
1 − 1 + 1 − 1 + · · · + (−1)n
its sequence of averages is bn = → 0 as n → +∞ since the numerator is either 0 or −1.
n

192 T HEOREM If a sequence of positive real numbers converges to a number, then its sequence of geometric means con-
verges to the same number, that is, if ∀n > 0, xn ≥ 0 and xn → a then (x1 x2 · · · xn )1/n → a.

Proof: The proof mimics Cèsaro’s Theorem 187. Since xn → a, for all ε > 0 there is N > 0 such that for all n ≥ N ,

|xn − a| < ε =⇒ a − ε < xn < a + ε.

Then
à !Tpn U/n à !Tpn U/n
³ ´1/n ³ ´1/n
1/n
min
p xk xTpn U+1 · · · xn ≤ (x1 x2 · · · xTpn U xTpn U+1 · · · xn ) ≤ max
p xk xTpn U+1 · · · xn .
1≤k≤T n U 1≤k≤T n U

p
This gives, for T n U ≥ N ,
à !Tpn U/n à !Tpn U/n
p p
(n−T n U)/n
min
p xk (a − ε) ≤ (x1 x2 · · · xTpn U xTpn U+1 · · · xn )1/n ≤ max
p xk (a + ε)(n−T n U)/n
.
1≤k≤T n U 1≤k≤T n U

à !Tpn U/n à !Tpn U/n

Now, both minp xk and maxp xk converge to 1 as n → +∞ by virtue of Theorem 174,
1≤k≤T n U 1≤k≤T n U
and again by the same theorem,
p p p p
(a − ε)(n−T n U)/n
= (a − ε) (a − ε))T n U/n
→ a − ε, (a + ε)(n−T n U)/n
= (a + ε) (a + ε))T n U/n
→ a +ε

as n → +∞. This gives the result. ❑

61
 Orders of Infinity

µ ¶n
n +1
193 Example Since e n = → e, then by the Theorem 192
n
µµ ¶1 µ ¶2 µ ¶3 µ ¶ ¶1/n µ ¶
2 3 4 n +1 n (n + 1)n 1/n
(e 1 e 2 · · · e n )1/n = ··· = → e.
1 2 3 n n!

n n n +1
This gives = · → 1 · e = e.
(n!)1/n n + 1 (n!)1/n

Homework

ÃÃ !!1/n
Problem 3.4.1 If {an }+∞
n=1 is a sequence of strictly positive real num- 2n
an Problem 3.4.3 Prove that lim = 4.
bers such that → a > 0. Prove that n→+∞ n
an−1

an p 1
lim = lim n an . Problem 3.4.4 Prove that lim (n(n + 1) · · · (n + n))1/n = 4e.
n→+∞ a n−1 n→+∞ n→+∞ n

1 2
Problem 3.4.2 Let xn → a and y n → b. Prove that Problem 3.4.5 Prove that lim (1 · 3 · 5· · · (2n − 1))1/n = .
n→+∞ n e
x1 y n + x2 y n−1 + · · · + xn y 1
→ ab.
n ¶ µ
(3n)! 1/n 2
1
Problem 3.4.6 Prove that lim = .
n→+∞ n 2 n! e

3.5 Orders of Infinity

© 2 ª+∞
Why bother? It is clear that the sequences {n}+∞
n=1 and n n=1 both tend to +∞ as n → +∞. We would like now
to refine this statement and compare one with the other. In other words, we will examine their speed towards
+∞.

194 Definition We write a n = O (b n ) if ∃C > 0, ∃N > 0 such that ∀n ≥ N we have |a n¯ | ≤ C

¯ |bn |. We then say that a n is Big Oh
¯ an ¯
of bn , or that a n is of order at most bn as n → +∞. Observe that this means that ¯¯ ¯¯ is bounded for sufficiently large n.
bn
The notation a n << bn , due to Vinogradov, is often used as a synonym of a n = O (bn ).

☞ A sequence {an }+∞

n=1 is bounded if and only if a n << 1.

An easy criterion to identify Big Oh relations is the following.

an
195 T HEOREM If lim = c ∈ R, then a n << bn .
n→+∞ bn

½ ¾+∞
an
Proof: By Theorem 164, a convergent sequence is bounded, hence the sequence is bounded: so for
¯ ¯ bn n=+1
¯ an ¯
sufficiently large n, ¯¯ ¯¯ < C for some constant C > 0. This proves the theorem. ❑
bn

☞ ³ ´
The = in the relation an = O (bn ) is not a true equal sign. For example n 2 = O n 3 since lim
n2
n→+∞ n 3
= 0 and so n 2 << n 3
n3
by Theorem 195. On the other hand, lim = +∞ so that for sufficiently large n, and for all M > 0, n 3 > Mn 2 , meaning
³ ´ n→+∞ n 2
that n 3 6= O n 2 . Thus the Big Oh relation is not symmetric.3

3
One should more properly write an ∈ O (b n ), as O (b n ) is the set of sequences growing to infinity no faster than b n , but one keeps the = sign for
historical reasons.

62
 Chapter 3

© ª+∞ n o+∞
196 T HEOREM (Lexicographic Order of Powers) Let (α,β) ∈ R and consider the sequences n α n=1 and n β . Then
n=1
n α << n β ⇐⇒ α ≤ β.

nα
Proof: If α ≤ β then lim is either 1 (when α = β) or 0 (when α < β), hence n α << n β by Theorem 195.
n→+∞ nβ

If n α << n β then for sufficiently large n, n α ≤ C n β for some constant C > 0. If α > β then this would mean that
for all large n we would have n α−β ≤ C , which is absurd, since for a strictly positive exponent α − β, n α−β → +∞
as n → +∞.❑

197 Example As n → +∞,

n 1/10 << n 1/3 << n << n 10/9 << n 2 ,
for example.

198 T HEOREM If c ∈ R \ {0} then O (c a n ) = O (a n ), that is, the set of sequences of order at most c a n is the same set at those
of order at most a n .

Proof: We prove that bn = O (c a n ) ⇐⇒ bn = O (a n ). If bn = O (c a n ) the there are constants C > 0 and N > 0 such
that |bn | ≤ C |c a n | whenever n ≥ N . Therefore, |bn | ≤ C 0 |a n | whenever n ≥ N , where C 0 = C |c|, meaning that
bn = O (a n ). Similarly, if bn = O (a n ) the there are constants C 1 > 0 and N 1 > 0 such that |bn | ≤ C 1 |a n | whenever
C1 C1
n ≥ N 1 . Since c 6= 0 this is equivalent to |bn | ≤ (c |a n |) = C 00 (c |a n |) whenever n ≥ N 1 , where C 00 = , meaning
c c
that bn = O (c a n ). Therefore, O (a n ) = O (c a n ). ❑

199 Example As n → +∞,

µ 3¶
¡ ¢ n ¡ ¢
O n3 = O = O 4n 3 .
3

¡ ¢ ¯ ¯
200 T HEOREM (Sum Rule) Let a n = O (xn ) and b n = O y n . Then a n + b n = O(max(|xn | , ¯ y n ¯)).

Proof: There exist strictly positive constants C 1 , N 1 ,C 2 , N 2 such that

¯ ¯
n ≥ N 1 , =⇒ |a n | ≤ C 1 |xn | and n ≥ N 2 , =⇒ |bn | ≤ C 2 ¯ y n ¯ .

Let N 0 = max(N 1 , N 2 ). Then for n ≥ N , by the Triangle inequality

¯ ¯
|a n + bn | ≤ |a n | + |bn | ≤ C 1 |xn | + C 2 ¯ y n ¯ .

Let C 0 = max(C 1 ,C 2 ). Then ¯ ¯ ¯ ¯

|a n + bn | ≤ C 0 (|xn | + ¯ y n ¯) ≤ 2C 0 max(|xn | , ¯ y n ¯),
whence the theorem follows. ❑

m m−1
201 C OROLLARY Let a n =
¡ k 0¢n + k 1 n + k 2 n m−2 + · · · + k m−1 n + k n be a polynomial of degree m in n with real number
coefficients. The a n = O n m , that is, a n is of order at most its leading term.

Proof: By the Sum Rule Theorem 200 the leading term dominates.❑

202 T HEOREM (Transitivity Rule) If a n = O(b n ) and b n = O(cn ), then a n = O (cn ).

Proof: There are strictly positive constants C 1 ,C 2 , N 1 , N 2 such that

n ≥ N 1 , =⇒ |a n | ≤ C 1 |bn | and n ≥ N 2 , =⇒ |bn | ≤ C 2 |cn | .

If n ≥ max(N 1 , N 2 ), then |a n | ≤ C 1 |bn | ≤ C 1 C 2 |cn | = C |cn | , with C = C 1 C 2 . This gives a n = O (cn ). ❑

63
 Orders of Infinity

¡ ¢ ¡ ¢ ¡ ¢
203 Example By Corollary 201, 5n 4 − 2n 2 + 100n − 8 = O 5n 4 . By Theorem 198, O 5n 4 = O n 4 . Hence
¡ ¢
5n 4 − 2n 2 + 100n − 8 = O n 4 .

¡ ¢
204 T HEOREM (Multiplication Rule) If a n = O(xn ) and b n = O(y n ), then a n b n = O xn y n .

Proof: There are strictly positive constants C 1 ,C 2 , N 1 , N 2 such that

¯ ¯
n ≥ N 1 , =⇒ |a n | ≤ C 1 |xn | and n ≥ N 2 , =⇒ |bn | ≤ C 2 ¯ y n ¯ .
¯ ¯ ¯ ¯ ¡ ¢
If n ≥ max(N 1 , N 2 ), then |a n bn | ≤ C 1C 2 ¯ xn y n ¯ = C ¯xn y n ¯, with C = C 1 C 2 . This gives a n bn = O xn y n . ❑

© ª+∞
205 T HEOREM (Lexicographic Order of Exponentials) Let (a,b) ∈ R, a > 1, b > 1, and consider the sequences a n n=1 and
© n ª+∞
b n=1 . Then a n << b n ⇐⇒ a ≤ b.

a
Proof: Put r = , and use Theorems 172 and 195. ❑
b

1
206 Example << 1 << 2n << e n << 3n .
2n

207 L EMMA Let a ∈ R, a > 1, k ∈ N \ {0}. Then n k << a n .

nk
Proof: By Theorem 175, lim = 0. Now apply Theorem 195. ❑
n→+∞ a n

208 T HEOREM (“Exponentials are faster than powers”) Let a ∈ R, a > 1, α ∈ R. Then n α << a n .

Proof: Put k = max(1, TαU + 1). Then by Theorem 196, n α << n k . By Lemma 207, n k << a n , and by the
Transitivity of Big Oh (Theorem 202), n α << n k << a n . ❑

209 Example
n 100 << e n .

210 T HEOREM (“Logarithms are slower than powers”) Let (α,β) ∈ R2 , α > 0. Then (logn)β << n α .

Proof: If β ≤ 0, then (logn)β << 1 and the assertion is evident, so assume β > 0. For x > 0, then log x < x. Putting
x = n α/β , we get
βn α/β ββ n α
logn α/β < n α/β =⇒ logn < =⇒ (logn)β < ,
α αβ
whence (logn)β << n α . ❑

By the Multiplication Rule (Theorem 204) and Theorems 196, 208, 210, in order to compare two expressions of the type
a n n b (log)c and u n n v (log)w we simply look at the lexicographic order of the exponents, keeping in mind that logarithms
are slower than powers, which are slower than exponentials.

211 Example In increasing order of growth we have

1 1 1 1 q n
10
<< << = << 1 << (loglogn) << logn << << n << n logn << e n .
en 2n n 2 logn logn

212 Example Decide which one grows faster as n → +∞: n log n or (logn)n .

64
 Chapter 3

2
Solution: Since n log n = e (log n) and (logn)n = e n loglog n , and since (logn)2 << n loglogn, we conclude that
n logn << (logn)n .

We now define two more fairly common symbols in asymptotic analysis.

an
213 Definition We write a n = o (b n ) if → 0 as n → +∞, and say that a n is small oh of bn , or that a n grows slower than bn
bn
as n → +∞.

214 Definition A sequence {a n }+∞

n=1 is said to be infinitesimal if a n = o (1), that is, if a n → 0 as n → +∞.

☞ We know from above that for a > 1 lim

¡ γ¢
nα
n→+∞ a n
¡ ¢
= 0, and so n α = o a n . Also, for γ > 0, lim
n→+∞
(log n)β
nγ
= 0, and so
β
(log n) = o n .

an
215 Definition We write a n ∼ b n if → 1 as n → +∞, and say that a n is asymptotic to bn .
bn

Asymptotic sequences are thus those that grow at the same rate as the index increases.

¡ ¢ ¡ ¢
f =O g g =O f

¡ ¢ f ∼g ¡ ¢
f =o g g =o f

Figure 3.2: Diagram of O relations.

© ª+∞ © 2 ª+∞
216 Example The sequences n 2 − n sin n n=1 , n + n − 1 n=1 are asymptotic since
sin n
n 2 − n sin n 1−
= n → 1,
n2 + n − 1 1 1
1+ − 2
n n
as n → +∞.

217 T HEOREM Let {a n }+∞ +∞

n=1 and {b n }n=1 be two properly diverging sequences. Then a n ∼ b n ⇐⇒ a n = b n (1 + o (1)).

Proof: Since the limit is 1 > 0, either both diverge to +∞ or both to −∞. Assume the former, and so, eventually,
bn will be strictly positive. Now,

an an
lim =1 ⇐⇒ ∀ε > 0,∃N > 0,1 − ε < < 1+ε
n→+∞ bn bn

⇐⇒ bn − bn ε < an < bn + bn ε

⇐⇒ |a n − bn | < bn ε

⇐⇒ a n − bn = o (bn ) .

65
 Cauchy Sequences

The relationship between the three symbols is displayed in figure 3.2.

Homework

³ ´ ³ ´
Problem 3.5.1 Prove that e n << n!. that an = O n 2 and an = o n 2 . Which of the two statements con-
veys more information?
Problem 3.5.2 Prove that O (O (an )) = O (an ).
Problem 3.5.6 True or false: an = O (n) =⇒ an = o (n).
Problem 3.5.3 Let k ∈ R be a constant. Prove that k + O (an ) =
O (k + an ) = O (an ). Problem 3.5.7 True or false: an = o (n) =⇒ an = O (n).

³ ´
Problem 3.5.4 Let k ∈ R, k > 0, be a constant. Prove that (an +
Problem 3.5.8 True or false: an = o n 2 =⇒ an = O (n).
bk )k << ank + bn
k
.

³ ´
Problem 3.5.5 For a sequence of real numbers {an }+∞
n=1 it is known Problem 3.5.9 True or false: an = o (n) =⇒ an = O n 2 .

3.6 Cauchy Sequences

218 Definition A sequence of real numbers {a n }+∞
n=1 is called a Cauchy Sequence if

∀ε > 0, ∃N > 0, such that ∀n,m ≥ N |a n − a m | < ε.

219 T HEOREM Cauchy sequences are bounded.

Proof: Let {a n }+∞

n=1 be Cauchy. Take N > 0 such that for all n ≥ N , |a n − a N | < 1 . Then a n is bounded by

max (|a 1 |,|a 2 | ,... ,|a N |) + 1.

220 L EMMA If a Cauchy sequence of real numbers has a convergent subsequence, then the parent sequence converges, and
it does so to the same limit as the subsequence.

© ª+∞
Proof: Let {a n }+∞
n=1 be a Cauchy sequence of real numbers, and suppose that its subsequence a n k k=1 converges
to the real number a. Given ε > 0, take N > 0 sufficiently large such that
¯ ¯
∀m,n,n k ≥ N , |a n − a m | < ε, and ¯ a n − a ¯ < ε.
k

By the Triangle Inequality,

¯ ¯ ¯ ¯
|a n − a| ≤ ¯ a n − a nk ¯ + ¯ a nk − a ¯ < ε + ε = 2ε,

whence a n → a.❑

221 T HEOREM (General Principle of Convergence) A sequence of real numbers converges if and only if it is Cauchy.

Proof:

(⇒) If a n → a, given ε > 0, choose N > 0 such that |a n − a| < ε for all n ≥ N .
Then if m,n ≥ N ,
|a n − a m | ≤ |a n − a| + |a m − a| ≤ ε + ε = 2ε.

Since 2ε > 0 can be made arbitrarily small, a n is Cauchy.

66
 Chapter 3

h i
(⇐) Suppose a n is Cauchy. By virtue of Theorem 219 it is bounded, say that for all n > 0, a n ∈ α ;β . Put

S = {s : an ≥ s for infinitely many n}.

As α ∈ S , S 6= ∅. S is bounded above by β. By the Completeness Axiom, S has a supremum, a = sup S .

Given ε > 0, a − ε < a and so there is s ∈ S such that a − ε < s. By definition of S , there are infinitely many
n with a n ≥ s > a −ε. a +ε > a, so that a +ε ∉ S and so there are only finitely many n for which a n ≥ a +ε.
Thus there are infinitely many n with a n ∈ (a − ε, a + ε).
Choose N > 0 such that |a n − a m | < ε for all m,n ≥ N . We can find m ≥ N with a m ∈ (a − ε, a + ε) ie
|a m − a| < ε. Then if n ≥ N ,
|a n − a| ≤ |a n − a m | + |a m − a| < ε + ε = 2ε

As 2ε can be made arbitrarily small this shews a n → a.

Homework

3.7 Topology of sequences. Limit Superior and Limit Inferior

222 T HEOREM A set X j R is dense in R is and only if for every x ∈ R there is a sequence {xn }+∞
n=1 of elements of X \ {x} that
converges to x.

Proof:
1
=⇒ For each positive integer n, since X is dense in R, there exists xn ∈ X \ {x} such that |xn − x| < n . But then
2
xn → x as n → +∞.
⇐ Let x ∈ R and let {xn }+∞
n=1 of elements of X \ {x} that converges to x. Then ∀ε > 0, ∃N ∈ N such that ∀n ≥ N ,
|xn − x| < ε. But then we have found elements of X \ {x} which are arbitrarily close to x, meaning that X is
dense in R.

223 T HEOREM Let X j R. A point x ∈ R is an accumulation point of X if and only if there exists a sequence of elements of
X \ {x} converging to x.

Proof:

=⇒ If x is an accumulation point of X , every closed interval I n := [x −1/n; x +1/n], n ∈ N, satisfies I n ∩(X \{x}) 6=
1
∅, thus ∀n ∈ N, ∃xn ∈ I n ∩ (X \ {x}). Since |xn − x| < , we conclude that lim xn = x.
n
⇐ Suppose now that {xn }+∞ 1 is an infinite sequence of points of X \ {x} converging to x. If x ∉ Acc (X ), then
x ∉ Acc(x1 , x2 ,...). Thus there is a neighbourhood of x, Nx such that Nx ∩ {x1 , x2 ,...}. Thus there is a ε > 0
such that ]x − ε; x + ε[j Nx . For this ε and for none of the xn it is true then that |xn − x| < ε, contradicting
the fact that lim xn = x.
n→+∞

224 Definition Given a sequence {a n }+∞

1 , the new sequence

bk = inf a n = inf{a k , a k+1 , a k+2, }, k ≥ 1,

n≥k

satisfies bk ≤ bk+1 , that is, it is increasing, and hence it converges to its supremum. We then put

lim inf = sup inf a k .

n→+∞ n≥1 k≥n

67
 Topology of sequences. Limit Superior and Limit Inferior

Similarly, the new sequence

ck = sup a n = sup{a k , a k+1 , a k+2, }, k ≥ 1,
n≥k
satisfies ck ≥ ck+1 , that is, it is decreasing, and hence it converges to its infimum. We then put

lim sup = inf sup a k .

n→+∞ n≥1 k≥n

We now prove the following theorem for future reference.

225 T HEOREM For any sequence {a n }+∞

n=0 of strictly positive real numbers
a n+1 p p a n+1
lim inf ≤ lim inf n a n ≤ lim sup n a n ≤ lim sup .
n→+∞ an n→+∞ n→+∞ n→+∞ an

Proof: We will prove the last inequality. The first is quite similar, and the two middle ones are obvious.
a n+1
Put r = lim sup . If r = +∞ then there is nothing to prove. For r < +∞ choose r 0 > r . There is N ∈ N such
n→+∞ an
that
a n+1
∀n ≥ N , ≤ r 0.
an
Hence,
a N +1 ≤ r 0 a N , a N +2 ≤ r 0 a N +1 , a N +3 ≤ r 0 a N +2 ,... a N +t ≤ r 0 a N +t −1 ,
and so, upon multiplication and cancelling,

a N +t ≤ a N (r 0 )t ,

and putting n = N + t,
q
p n p
a n ≤ a N (r 0 )−N (r 0 )n =⇒ n
an ≤ r 0 a N (r 0 )−N =⇒ lim sup n
an ≤ r 0,
n→+∞
q
n
since a N (r 0 )−N is a fixed real number (does not depend on n), and so, a N (r 0 )−N → 1 by Theorem 174.
❑
The following theorem is an easy exercise left to the reader.

226 T HEOREM Let {a n }+∞

1 be a sequence of real numbers. Then
1. if limsup a n = +∞, then {a n }+∞
1 has a subsequence converging to +∞.
n→+∞

2. if limsup a n = −∞, then lim a n = −∞.

n→+∞ n→+∞

3. if limsup a n = a ∈ R, then
n→+∞
∀² > 0, ∃n 0 such that a n < a + ² whenever n ≥ n 0
and also, there are infinitely many a n such that a − ² < a n .

4. if liminf a n = −∞, then {a n }+∞

1 has a subsequence converging to −∞.
n→+∞

5. if liminf a n = +∞, then lim a n = +∞.

n→+∞ n→+∞

6. if liminf a n = a ∈ R, then
n→+∞
∀² > 0, ∃n 0 such that a − ² < a n whenever n ≥ n 0
and there are infinitely many a n such that a n < a + ².

7. liminf a n ≤ limsup a n is always verified, and furthermore, liminf a n = limsup a n if and only if lim a n exists, in
n→+∞ n→+∞ n→+∞ n→+∞ n→+∞
which case liminf a n = lim a n = limsup a n .
n→+∞ n→+∞ n→+∞

Homework

68
 Chapter 3

a
Problem
p 3.7.1 Identify the set of accumulation points of the set Clearly, the fraction in this enumeration occupies the a +
p b
{ a − b : (a,b) ∈ N2 }. b(b + 1)
-th place. For each integer k ≥ 1, cover the k-th fraction
2
a a
Problem 3.7.2 Consider the following enumeration of the proper by an interval of length 2−k centred at . Shew that the point
b
p b
fractions 2
0 1 0 1 2 0 1 2 3 does not belong to any interval in the cover.
, , , , , , , , ... . 2
1 1 2 2 2 3 3 3 3

69
Chapter 4

Series

4.1 Convergence and Divergence of Series

227 Definition Let {a n }+∞
n=1 be a sequence of real numbers. A series is the sum of a sequence. We write
X
n
sn = a1 + a2 + · · · + an = ak .
k=1

Here sn is the n-th partial sum. Observe in particular that

a n = sn − sn−1 .

228 Definition If the sequence {sn }+∞

n=1 has a finite limit S, we say that the series converges to S and write
X
+∞
a k = lim sn = S.
n→+∞
k=1

Otherwise we say that the series diverges.

X
+∞
Observe that a n converges to S if ∀ε > 0, ∃N such that ∀n ≥ N ,
n=1
¯Ã ! ¯
¯ X ¯
¯ ¯
¯ a k − S ¯ = |sn − S| < ε.
¯ k≤n ¯

Now, since à ! à ! à !
X X X X
ak − S = ak − ak = ak ,
k≤n k≤n k≥1 k>n
we see that a series converges if and only if its “tail” can be made arbitrarily small. Hence, the reader should notice that
adding or deleting a finite amount of terms to a series does not affect its convergence or divergence. Furthermore, since
the sequence of partial sums of a convergent series must be a Cauchy sequence we deduce that a series is convergent if and
only if ∀ε > 0, ∃N > 0 such that ∀m ≥ N ,n ≥ N , m ≤ n,
¯ ¯
¯Xn ¯
¯ ¯
|sn − sm | = ¯ a k ¯ < ε. (4.1)
¯k=m ¯

X
∞
229 T HEOREM (n -th Term Test for Divergence) If a n converges, then a n → 0 as n → +∞.
n=1

X
n
Proof: Put sn = a k . Then
k=1

lim sn = S =⇒ a n = sn − sn−1 → S − S = 0.
n→+∞

70
 Chapter 4

In general, the problem of determining whether a series converges or diverges requires some work and it will be dealt
with in the subsequent sections. We continue here with some other examples.

µ ¶n µ ¶
X
+∞ 2 2 n
230 Example The series 1+ diverges, since its n-th term 1 + → e2.
n=1 n n

1 X
+∞ 1
231 Example We will prove that the harmonic series diverges, even though → 0 as n → +∞. Thus the condition in
n=1 n n
Theorem 229 though necessary for convergence is not sufficient. The divergence of the harmonic series was first demon-
strated by Nicole d’Oresme (ca. 1323-1382), but his proof was mislaid for several centuries. The result was proved again
by Pietro Mengoli in 1647, by Johann Bernoulli in 1687, and by Jakob Bernoulli shortly thereafter. Write the partial sums in
dyadic blocks,
2M 1
X X
M 2m
X 1
= .
n=1 n m=1 n=2m−1 +1 n

As 1/n ≥ 1/N when n ≤ N , we deduce that

2m
X 1 2m
X 1
≥ 2−m = (2m − 2m−1 )2−m =
n=2m−1 +1
n n=2m−1 +1
2

Hence,
2M 1
X M
≥
n=1 n 2
so the series diverges in the limit M → +∞.

The following theorem says that linear combinations of convergent series converge.

X
+∞ X
+∞ X
+∞
232 T HEOREM Let a n = A and bn = B be convergent series and let γ ∈ R be a real number. Then the series (a n + γbn )
n=1 n=1 n=1
converges to A + γB .

Proof: For all ε > 0 there exist N , N 0 such that for all n ≥ max(N , N 0 ),
¯ ¯ ¯ ¯
¯X ¯ ε ¯X ¯ ε
¯ ¯ ¯ ¯
¯ ak − A¯ < , ¯ bk − B ¯ < .
¯k≤n ¯ 2 ¯k≤n ¯ 2(|γ| + 1)

|γ|
Hence, by the triangle inequality and by the obvious inequality ≤ 1, we have
|γ| + 1
¯Ã ! ¯ ¯ ¯ ¯ ¯
¯ X ¯ ¯X ¯ ¯X ¯ ε ε ε ε
¯ ¯ ¯ ¯ ¯ ¯
¯ a k + γbk − (A + γB )¯ ≤ ¯ a k − A ¯ + |γ| ¯ bk − B ¯ ≤ + |γ| < + = ε.
¯ k≤n ¯ ¯k≤n ¯ ¯k≤n ¯ 2 2(|γ| + 1) 2 2

233 Definition A geometric series with common ratio r and first term a is one of the form

X
+∞
a + ar + ar 2 + ar 3 + · · · = ar n .
n=0

By Theorem 173, if |r | < 1 then the series converges and we have

X
+∞ a
a + ar + ar 2 + ar 3 + · · · = ar n = .
n=0 1−r

71
 Convergence and Divergence of Series

234 Example A fly starts at the origin and goes 1 unit up, 1/2 unit right, 1/4 unit down, 1/8 unit left, 1/16 unit up, etc., ad
infinitum. In what coordinates does it end up?

Solution: Its x coordinate is

1
1 1 1 2
− + − · · · = 2 −1 = .
2 8 32 1− 4 5

Its y coordinate is
1 1 1 4
1− + −··· = −1
= .
4 16 1− 4 5

Therefore, the fly ends up in µ ¶

2 4
, .
5 5
Here we have used the fact the sum of an infinite geometric progression with common ratio r , with |r | < 1 and
first term a is
a
a + ar + ar 2 + ar 3 + · · · = .
1−r

X
N
235 Definition A telescoping sum is a sum where adjacent terms cancel out. That is, a n is a telescoping sum if we can
n=0
write a n = bn+1 − bn and then

X
N
a n = a 0 + a 1 + · · · + a N = (b1 − b0 ) + (b2 − b1 ) + · · · + (b N +1 − b N ) = b N +1 − b0 .
n=0

236 Example We have

X
N 1 XN µ1 1
¶ µ
1 1
¶ µ
1 1
¶ µ
1 1
¶
1
= − = − + − +··· + − = 1− .
n=1 n(n + 1) n=1 n n + 1 1 2 2 3 N N + 1 N +1

Thus µ ¶
X
+∞ 1 XN 1 1
= lim = lim 1 − = 1.
n=1 n(n + 1) N →+∞ n=1 n(n + 1) N →+∞ N +1

237 Example We have

X
N 1 1 XN µ 1 1
¶
1
µµ
1 1
¶ µ
1 1
¶ µ
1 1
¶¶
= − = − + − +··· + −
n=1 n(n + 1)(n + 2) 2 n=1 n(n + 1) (n + 1)(n + 2) 2 1·2 2·3 2·3 3·4 N (N + 1) (N + 1)(N + 2)
µ ¶
1 1 1
= − .
2 2 (N + 1)(N + 2)

Thus µ ¶
X
+∞ 1 XN 1 1 1 1 1
= lim = lim − = .
n=1 n(n + 1)(n + 2) N →+∞ n=1 n(n + 1)(n + 2) N →+∞ 2 2 (N + 1)(N + 2) 4

Homework

∞
X 2n +∞
X 1
Problem 4.1.1 Find the sum of n+1
. arctan 2 .
n=3 e n=0 n + n +1

+∞
X Problem 4.1.4 Find the exact numerical value of the infinite sum
1
Problem 4.1.2 Find the sum of the series . +∞
p
2 X (n − 1)!
n=2 4n − 1 p p .
n=1 (1 + 1) · · · (1 + n)

Problem 4.1.3 Find the exact numerical value of the sum

72
 Chapter 4

Problem 4.1.5 Shew that Problem 4.1.8 The Fibonacci Numbers f n are defined recursively
as follows:
n
X k 1 n2 + n
4 2
= · ,
k=1 k + k + 1 2 n2 + n + 1
f 0 = 1, f 1 = 1, f n+2 = f n + f n+1 , n ≥ 0.
n
X k
and thus prove that 4 2
converges. +∞
X fn 3
k=1 k + k + 1 Prove that = .
3n 5
n=1

Problem 4.1.6 Let b(n) denote the number of ones in the binary
X X
expansion of the positive integer n, for example b(3) = b(112 ) = 2. Problem 4.1.9 Let an be a convergent series and let bn be a
X b(n) n≥0 n≥0
Prove that = log 4. X
n=1 n(n + 1) divergent series. Prove that (an + bn ) diverges.
n≥0

Problem 4.1.7 Find X

1 1 1 1 1 1 1 1 Problem 4.1.10 Prove that if an is a series of positive terms and
1+ + + + + + + + +··· , n≥1X
2 3 6 8 9 12 16 18 that its partial sums are bounded, then an converges. Shew that
which is the sum of the reciprocals of all positive integers of the form X n≥1
this is not necessarily true if an is not a series of positive terms.
2n 3m for integers n ≥ 0,m ≥ 0. n≥1

4.2 Convergence and Divergence of Series of Positive Terms

We have several tools to establish convergence and divergence of series of positive terms. We will start with some simple
comparison tests.

238 T HEOREM (Direct Comparison Test) Let {a n }+∞ +∞ +∞

n=0 , {b n }n=0 , {c n }n=0 , be sequences
X of positive real numbers.
X Suppose that
eventually a n ≤ bn , that is, that ∃N ≥ 0 such that ∀n ≥ N there holds a n ≤ bn . If bn converges, then a n converges.
X X n≥0 n≥0
If eventually a n ≥ cn , and cn diverges, then a n also must diverge.
n≥0 n≥0

Proof: The theorem is clear from the inequalities

X X X X
an ≤ bn , an ≥ cn .
n≥N n≥N n≥N n≥N
X X
If bn converges, then its tail can be made as small as we please, and so the tail of a n can be made as small
n≥0 X n≥0
as we please. Similarly if cn diverges, because it is a series of positive terms, its tail grows without bound and
X n≥0
so the tail of a n grows without bound. ❑
n≥0

☞ Call a divergent series of positive terms a “giant” and a converging series of positive terms a “midget.” The comparison
tests say that if a series is bigger than a giant it must be a giant, and if a series is smaller than a midget, it must be a midget.

X 1
239 Example From example 236, converges. Since for n ≥ 1,
n≥1 n(n + 1)

1 1
n(n + 1) < (n + 1)2 =⇒ < ,
(n + 1)2 n(n + 1)
we deduce that the series
X 1 X 1
2
= 2
n≥1 (n + 1) n≥2 n
X 1 X 1
converges. Since adding a finite amount of terms to a series does not affect convergence, we deduce that 1+ 2
= 2
n≥2 n n≥1 n
converges.

X
+∞ 1 1 1 X 1
+∞
240 Example n
converges. For n ≥ 2 we have n
≤ 2
and the series converges by direct comparison with 2
.
n=1 n n n n=1 n

73
 Convergence and Divergence of Series of Positive Terms

X 1 X 1
241 Example From example 230, diverges. Since for n ≥ 1, logn < n, we deduce that diverges. Notice that
n≥1 n n≥2 logn
here we start the sum at n = 2 since the logarithm vanishes at n = 1.

242 Example Prove that

X 1
p p
p prime

diverges.

Solution: We will prove this by contradiction. Let p 1 = 2, p 2 = 3, p 3 = 5, . . . be the sequence of primes in

ascending order and assume that the series converges. Then there exists an integer K such that
X 1 1
< .
m≥K +1 p m 2

Let P = p 1 p 2 · · · p K and consider the numbers 1+nP for n = 1,2,3,.... None of these numbers has a prime divisor
in the set {p 1 , p 2 ,... , p K } and hence all the prime divisors of the 1 + nP belong to the set {p K +1 , p K +2 ,...}. This
means that for each t ≥ 1, Ã !s
X
t 1 X X 1 X 1
≤ ≤ s
= 1,
n=1 1 + nP s≥1 m≥K +1 p m s≥1 2

X
t 1
that is, , a series of positive terms, has bounded partial sums and so it converges. But since 1+nP ∼ nP
n=1 1 + nP
1 X 1
as n → +∞ and diverges, we obtain a contradiction.
P n≥1 n

Since the convergent behaviour of a series depends of its tail, the following asymptotic comparison tests should be clear,
and its proof follows the same line of reasoning as Theorem 238.

243 T HEOREM (Asymptotic Comparison Test) Let {a n }+∞ +∞

}+∞
n=0 , {b n }n=0 , {c nXn=0 , be sequences
X of real numbers which are X
even-
tually positive. Suppose that a n << bn , and that cn << a n . Then both a n and bn converge together, and both an
X n≥0 n≥0 X n≥0
and cn diverge together. Moreover, if {bn }+∞
n=0 is eventually a strictly positive sequence and a n ∼ b n , then a n and
X n≥0 n≥0
bn converge or diverge together.
n≥0

In order to effectively use the comparison tests we must have a ready catalogue of series whose convergence or diver-
gence we know. In the subsequent lines we will develop such a catalogue. We start with the following consequence of the
comparison tests.

244 T HEOREM (Cauchy Condensation Test) Let {a n }+∞

n=0 be a sequence of positive real numbers which is monotonically de-
X
∞ X
∞
creasing. Then a n converges if and only if the sum 2n a 2n converges.
n=0 n=0

Proof: Since the sequence {a n }+∞

n=0 is monotonically decreasing and positive,

2n+1
X−1 2n+1
X−1 2n+1
X−1 2n+1
X−1
a 2n+1 −1 ≤ ak ≤ a 2n =⇒ 2n a 2n+1 −1 ≤ a k ≤ 2n a 2 n .
k=2n k=2n k=2n k=2n

The second inequality yields

2NX
+1
−1 X
N 2n+1
X−1 X
N 2NX
+1
−1 X
N
an = ak ≤ 2n a 2n =⇒ lim a n ≤ lim 2n a 2 n .
n=0 n=0 k=2n n=0 N →+∞ n=0 N →+∞ n=0

X
+∞ X
+∞
Thus if 2n a 2n converges so does an .
n=0 n=0

74
 Chapter 4

The first inequality yields

2n+1
X−1 2n+1
X−1
2n a 2n+1 −1 ≤ ak =⇒ (2n+1 − 1)a 2n+1 −1 ≤ 2 a k − a 2n+1 −1
k=2n k=2n
X
N N 2n+1
X X−1 X
N 2NX
+1
−1 X
N
=⇒ (2n+1 − 1)a 2n+1 −1 ≤ 2 ak − a 2n+1 −1 = 2 an − a 2n+1 −1 .
n=0 n=0 k=2n n=0 n=0 n=0

As an application of Cauchy’s Test, we obtain the following important result.

X
+∞ 1
245 T HEOREM (p -series Test) If p > 1 then ζ(p) = p
converges, but diverges when p ≤ 1.
n=1 n

Proof: If p ≤ 0, divergence follows from Theorem 229. If p > 0, then using the fact that x 7→ x p is monotonically
increasing, we may use Theorem 244. Since

X 2k X³ ´k
= 2(1−p)
k≥0 2pk k≥0

is a geometric series with ratio 21−p , it converges by Theorem 173 when

21−p < 1 =⇒ (1 − p) log2 2 < log2 1 =⇒ 1 − p < 0 =⇒ p > 1,

and diverges for p > 1. The case p = 1 has been shewn to diverge in example 230. ❑

p X
+∞ 1
246 Example Since 2 > 1, the series p converges.
n=1 n 2

247 Example Since p p

n 2 + (loglogn)2007 n 2 1
∼ 3 = p
n 3 + n(logn)5 + 1 n n 3− 2
p
p X n 2 + (loglogn)2007
and 3 − 2 > 1, the series 3 5
converges.
n≥1 n + n(logn) + 1

248 C OROLLARY (De Morgan’s Logarithmic Scale) If p > 1 then all of

X
+∞ 1 X
+∞ 1 X
+∞ 1 X
+∞ 1
p
; p
; ; ; ...
n=1 n n≥e n(logn) n≥e e n(logn)(loglogn)p n≥e e
e n(logn)(loglogn)(logloglogn)p

converge, but diverge when p ≤ 1.

Proof: The theorem is proved inductively by successive applications of Cauchy’s Condensation Test. We will
X
+∞ 1 X 1
+∞
prove how the case for p
follows from the case p
and leave the rest to the reader. We see that
n≥e n(logn) n=1 n

X 2k 1 X 1
= ,
k≥1 2k (log2k )p (log2) k≥1 k p
p

and so this case follows from Theorem 245. ❑

X
+∞ (logn)100
249 Example Determine whether converges.
n=4 n 3/2 loglogn

75
 Convergence and Divergence of Series of Positive Terms

(logn)100 (logn)100 (logn)100 1

Solution: Since (logn)100 << n 1/4 , eventually << 1. We have << ·
n 1/4 n 3/2 loglogn n 1/4 n 5/4 loglogn
X
+∞ 1 X
+∞ (logn)100
and since < +∞, we have < +∞, that is, the series converges.
n=4 n 5/4 loglogn n=4 n 3/2 loglogn

The reader should be aware that the value of the exponent in Theorems 245 and 248 is fixed. The following examples should
dissuade him that “having an exponent higher than 1” implies convergence.

X
∞ 1
250 Example Test 1+1/n
for convergence by comparing it to a suitable p-series. Use the direct comparison test.
n=1 n

1 1
Solution: By induction n < 2n =⇒ n 1/n < 2 and so n 1+1/n < 2n =⇒ < . So the series diverges by
2n n 1+/1n
X∞ 1
direct comparison to .
n=1 2n

X
∞ 1
251 Example Test for convergence by comparing it to a suitable p-series. Use the direct comparison test
n=2 n 1+1/ logn

1
Solution: We have n = e log n =⇒ n logn = e and so n 1+1/ logn = en, n > 1.. So the series diverges by direct
X∞ 1
comparison to .
n=2 en

X
∞ 1
252 Example Test for convergence by comparing it to a suitable p-series. Use the direct comparison test.
n=2 n 1+1/ loglog n

x2
Solution: By considering the monotonicity of f (x) = e x − (see Theorem 385) or otherwise, we can prove that
2
x2
ex > for x > 0. Now,
2
1/ loglogn logn (logn)2
n 1/ log log n = e log n = e loglogn > .
2(loglogn)2
This gives
2(loglogn)2 1
2
> 1
.
n(logn) 1+
n log logn
Now,
X
+∞ 2(loglogn)2
2
n=2 n(logn)
can be shewn to converge by comparing to a series in the De Morgan logarithmic scale.

X
+∞ 1
253 Example Prove that the series diverges.
n=1 n 1.8+sinn

h p
4 5 i π 3
Solution: For k ∈ Z, the interval I k = (2k + )π ;(2k + )π has length > 1 and x ∈ I k =⇒ sin x ≤ − .
3 3 3 2
5π
The gap between I k and I k+1 is < < 6. Hence, among any seven consecutive integers, at least one must fall
3 p
3
into I k and for this value of n we must have 1.8 + sin n < 1 − < 1. This means that
2
X
+∞ 1 X
+∞ X
n=7m+7 1 X
+∞ 1
1.8+sin n
= 1.8+sinn
≥ ,
n=1 n m=0 n=7m+1 n m=0 7m +7

and since the rightmost series diverges, the original series diverges by the direct comparison test.

76
 Chapter 4

The following result puts the harmonic series at the “frontier” of convergence and divergence for series with monotonically
decreasing positive terms.
X
254 T HEOREM (Pringsheim’s Theorem) Let a n be a converging series of positive terms of monotonically decreasing
µ ¶ n≥1
1
terms. Then a n = o .
n

Problem 4.2.1 Since the series converges, its sequence of partial sums is a Cauchy sequence and by 4.1, given ε > 0, ∃m > 0,
such that ∀n ≥ m,
X
n
a k < ε.
k=m+1
Because the series decreases monotonically, each of a m+1 , a m+2 ,... , a n is at least a n and thus
X
n
(n − m)a n ≤ a k < ε.
k=m+1

ε
Again, since the series converges, a n → 0 as n → +∞ we may choose n large enough so that a n < . In this case
m
2ε
(n − m)a n < ε =⇒ na n < ε + ma n < 2ε =⇒ a n < ,
n
which proves the theorem.

The disadvantage of the comparison tests is that in order test for convergence, we must appeal to the behaviour of an
auxiliary series. The next few tests provide a way of testing the series against its own terms.

X
+∞
255 T HEOREM (Root Test) Let a n be a series of positive terms. Put r = limsup(a n )1/n . Then the series converges if r < 1
n=1
and diverges if r > 1. The test is inconclusive if r = 1.

Proof: If r < 1 choose r 0 with r < r 0 < 1. Then there exists N ∈ N such that
p
∀n ≥ N , n a n ≤ r 0 =⇒ a n ≤ (r 0 )n .
X
+∞ X
+∞
But then a n converges by direct comparison to the converging geometric series (r 0 )n .
n=0 n=0
If r > 1 then there is a sequence {n k }+∞
k=1 of positive integers such that
p
nk a
n k → r.

This means that a n will be > 1 for infinitely many values of n, and so, the condition a n → 0 necessary for conver-
gence, does not hold.
X 1
+∞ X 1
+∞
By considering , which diverges, and 2
, which converges, one sees that r = 1 may appear in series of
n=1 n n=1 n
different conditions. ❑

X
+∞ a n+1
256 T HEOREM (Ratio Test) Let a n be a series of strictly positive terms. Put r = limsup . Then the series converges
n=1 an
if r < 1 and diverges if r > 1. The test is inconclusive if r = 1.

Proof: If r < 1, then there exists N ∈ N such that

a N +1 ≤ r a N , a N +2 ≤ r a N +1 , a N +3 ≤ r a N +2 ,... a N +t ≤ r a N +t −1 .

Multiplying all these inequalities together,

a N +t ≤ a N r t .

77
 Convergence and Divergence of Series of Positive Terms

Putting N + t = n we deduce that

a n ≤ a N r −N r n .
X
+∞
Since a N r −N is a constant, we may use direct comparison between a n and the converging geometric series
n=1
X
+∞ X
+∞
a N r −N r n , concluding that a n converges.
n=1 n=1
If r > 1 then a n+1 ≥ a n ≥ a N for all n ≥ N . This means that the condition a n → 0, necessary for convergence, does
not hold.
X 1
+∞ X 1
+∞
By considering , which diverges, and 2
, which converges, one sees that r = 1 may appear in series of
n=1 n n=1 n
different conditions. ❑

☞ The root test is more general than the ratio test, as can be seen from Theorem 225.

257 Example Since

(n+1)!
(n+1)n+1 1 1
=µ ¶ → <1
n! 1 n e
nn 1+
n
X
+∞ n!
the series n
converges.
n=1 n

258 Example Since

µ ¶1/n
(n!)n n!
= →0
nn
2
nn
X
+∞ (n!)n
the series 2
converges.
n=1 nn

Homework

+∞
X
Problem 4.2.2 True or False: If the infinite series an of strictly Problem 4.2.6 Use the comparison tests to shew that if an > 0 and
X∞ X∞ a
n=1 n
+∞ an converges, then converges.
X n
positive terms, converges, then an2 must necessarily converge. n=1 n=1
n=1

Problem 4.2.7 Give an example of a series converging to 1 with n-

+∞
X 1
Problem 4.2.3 True or False: If the infinite series an of strictly th term an > 0 satisfying an << 2 . (That is, the n-th term goes to
n=1 n
+∞
X zero faster than the reciprocal of a square.)
positive terms converges, then sin(an ) must necessarily con-
n=1
verge. Problem 4.2.8 Give an example of a converging series of strictly
+∞
X +∞
X
positive terms an such that (an )1/n also converges.
+∞
X n=1 n=1
Problem 4.2.4 True or False: If the infinite series an of strictly
n=1
+∞
X Problem 4.2.9 Give an example of a converging series of strictly
positive terms converges, then tan(an ) must necessarily con- +∞ +∞
X X
verge.
n=1 positive terms an such that (an )1/n diverges.
n=1 n=1

+∞
X
Problem 4.2.5 True or False: If the infinite series an converges, Problem 4.2.10 Give an example of a converging series of strictly
n=1 positive terms an such that lim (an )1/n does not exist.
+∞
X n→+∞
then cos(an ) must necessarily converge.
n=1 ∞ 3n
X
Problem 4.2.11 Test 2n
using both direct comparison and
n=1 n

78
 Chapter 4

the root test. Problem 4.2.14 Let p n be the n-th prime. Thus p 1 = 2, p 2 = 3,
p 3 = 5, etc. Put a1 = p 1 and an+1 = p 1 p 2 · · · p n + 1 for n ≥ 1. Find

+∞
X 1
Problem 4.2.12 Let S be the set of positive integers none of whose .
n=1 a n
digits in its decimal representation is a 0:
X
S = {1,2,3,4,5,6, 7, 8, 9,11, 12,13, 14,15, 16,17, 18,19, 21, · · · }. Problem 4.2.15 Determine whether an converges, when an is
n≥2
given as below.
X 1
Prove that the series converges. µ ¶
n∈S
n 1 n 2 · 4 · 6· · · (2n)
1. 1+ − e. 7. .
n nn

Problem 4.2.13 Let d (n) be the number of strictly positive divisors 2. coshα n − sinhα n. 1! + 2! + · · · + n!
p 8. .
of the integer n. Prove that d (n) ≤ 2 n. Use this to prove that (n 3 + 1)2 (n + 2)!
3. log 2 .
(n + 1)3 1! − 2! + · · · ± n!
X d (n) p
n p 9. .
4. n + 1 − n n. (n + 1)!
2 Ã !
n≥1 n n3 + 1 (log n)n
5. arccos 3 . 10. .
converges. n +2 n log n
an 1
6. . 11. .
1 + a 2n (log n)log n

4.3 Summation by Parts

In this section we consider series whose terms have arbitrary signs. We first need the following result.
X
259 T HEOREM (Summation by Parts) Let A n = a k , A −1 = 0. Then for p ≤ q,
0≤k≤n
X X
ak bk = A k (bk − bk+1 ) + A q b q − A p−1 b p .
p≤k≤q p≤k≤q −1

Proof: Changing a subindex,

X X
ak bk = (A k − A k−1 )bk
p≤k≤q p≤k≤q
X X
= Ak bk − A k−1 bk
p≤k≤q p≤k≤q
X X
= Ak bk − A k bk+1 + A q b q − A p−1 b p .
p≤k≤q −1 p≤k≤q −1

giving the result. ❑

☞ An alternative and more symmetric formulation will be given once we introduce Riemann-Stieltjes integration.

We now obtain a convergence test.

X
260 T HEOREM (Dirichlet’s Test) The series a k bk converges if
k≥0

X
n
1. the partial sums A n = a k are bounded
k=0

2. bn ≥ bn+1

3. bn → 0 as n → +∞

Proof:
❑

79
 Alternating Series

4.4 Alternating Series

X
+∞
A series of the form (−1)n a n where the all the a n ≥ 0 is called an alternating series.
n=1

X
+∞
261 T HEOREM (Leibniz’s Alternating Series Test) The alternating series (−1)n a n converges if all the following condi-
n=1
tions are met

• the a n eventually decrease, that is, a n+1 ≤ a n for all n ≥ N .

• an → 0

X
+∞ 1
262 Example The series (−1)n+1 converges by Leibniz’s Test. In fact, one can prove that it equals log2.
n=1 n

4.5 Absolute Convergence

X
+∞ X
+∞
If |a n | converges then a n converges. The converse is not true.
n=1 n=1

¯ sin n ¯ 1 +∞ X sin n ¯¯ X sin n

+∞
¯ ¯
263 Example Since ¯ 2
¯ ≤ 2
, | 2 ¯ converges by the comparison test. Thus 2
converges absolutely and so
n n n=1 n n=1 n
it converges.

X∞ (−1)n sin(3n) X∞ (−1)n n

264 Example Determine whether the following two infinite series converge: (I) , (II) .
n=2 n2 2
n=1 n + 2

Solution: We have ¯
¯ sin 3n ¯¯ 1
¯(−1)n ¯≤ 2,
n2 n
x n
so (I) converges absolutely. As for number (II), f (x) = 2 is decreasing (take the first derivative) 2 → 0,
x +2 n +2
so it converges by Leibniz’s Test.

80
Chapter 5

Real Functions of One Real Variable

5.1 Limits of Functions

i h i h
265 D EFINITION -P ROPOSITION (Cauchy-Heine, Sinistral Limit) Let f : a ;b → R and let x0 ∈ a ;b . The following are equiv-
alent.

1. ∀ε > 0, ∃δ > 0 such that ¯ ¯

x0 − δ < x < x0 =⇒ ¯ f (x) − L ¯ < ε.
i h
2. For each sequence {xn }+∞
n=1 of points in the interval a ;b with xn < x0 , xn → x0 =⇒ f (xn ) → L.

If either condition is fulfilled we say that f has a sinistral limit f (x0 −) as x increases towards x0 and we write

f (x0 −) = lim f (x) = lim f (x).

x→x 0 − x%x 0

Proof:

1 =⇒ 2 Suppose that ∀ε > 0, ∃δ > 0 such that

¯ ¯
x0 − δ < x < x0 =⇒ ¯ f (x) − L ¯ < ε.

Let xn < x0 , xn → x0 . Then

|xn − x0 | < δ =⇒ x0 − δ < xn < x0 + δ

¯ n. But¯ we are assuming that xn < x0 , so in fact we have x0 − δ < xn < x0 . By our
for sufficiently large
assumption then ¯ f (xn ) − L ¯ < ε, and so 1 =⇒ 2.
i h
2 =⇒ 1 Suppose that for each sequence {xn }+∞ n=1 of points in the interval a ;b with xn < x0 , xn → x0 =⇒ f (xn ) →
L. If it were not true that f (x) → L as x → x0 , then there exists some ε0 > 0 such that for all δ > 0 we can
find x such that ¯ ¯
0 < |x − x0 | < δ =⇒ ¯ f (x) − L ¯ ≥ ε0 .
In particular, for each strictly positive integer n we can find xn satisfying

1 ¯ ¯
0 < |xn − x0 | < =⇒ ¯ f (xn ) − L ¯ ≥ ε0 ,
n
a contradiction to the fact that f (xn ) → L.

In an analogous manner, we have the following.

i h i h
266 D EFINITION -P ROPOSITION (Cauchy-Heine, Dextral Limit) Let f : a ;b → R and let x0 ∈ a ;b . The following are equiv-
alent.

81
 Limits of Functions

i h
1. For each sequence {xn }+∞
n=1 of points in the interval a ;b with xn > x0 , xn → x0 =⇒ f (xn ) → L.

2. ∀ε > 0, ∃δ > 0 such that ¯ ¯

x0 < x < x0 + δ =⇒ ¯ f (x) − L ¯ < ε.

If either condition is fulfilled we say that f has a dextral limit f (x0 +) as x decreases towards x0 and we write

f (x0 +) = lim f (x) = lim f (x).

x→x 0 + x&x 0

Upon combining Propositions 265 and 266 we obtain the following.

i h i h
267 D EFINITION -P ROPOSITION (Cauchy-Heine) Let f : a ;b → R and let x0 ∈ a ;b . The following are equivalent.

1. f (x0 −) = f (x0 +)
i h
2. For each sequence {xn }+∞
n=1 of points in the interval a ;b different from x0 , xn → x0 =⇒ f (xn ) → L.

3. ∀ε > 0, ∃δ > 0 such that ¯ ¯

0 < |x − x0 | < δ =⇒ ¯ f (x) − L ¯ < ε.

If either condition is fulfilled we say that f has a (two-sided) limit L as x decreases towards x0 and we write

L = lim f (x).
x→x 0

We now prove analogues of the theorems that the proved for limits of sequences.

268 T HEOREM (Uniqueness of Limits) Let X j R, a ∈ R, and f : X → R. If lim f (x) = L and lim f (x) = L 0 then L = L 0 .
x→a x→a

¯ ¯
Proof: If L 6= L 0 then take 2ε = ¯L − L 0 ¯ in the definition of limit. There is δ > 0 such that
¯ ¯ ¯ ¯
¯ ¯ ¯L − L 0 ¯ ¯ ¯ ¯L − L 0 ¯
¯
0 < |x − a| < δ =⇒ f (x) − L < ¯ , ¯ 0¯
f (x) − L < .
2 2
By the Triangle Inequality
¯ ¯ ¯ ¯
¯ ¯ ¯ ¯ ¯ ¯ ¯L − L 0 ¯ ¯L − L 0 ¯ ¯ ¯
¯L − L 0 ¯ ≤ ¯L − f (x)¯ + ¯ f (x) − L 0 ¯ < + = ¯L − L 0 ¯ ,
2 2
¯ ¯ ¯ ¯
but ¯L − L 0 ¯ < ¯L − L 0 ¯ is a contradiction. ❑

269 T HEOREM (Local Boundedness) Let X j R, a ∈ R, and f : X → R. If lim f (x) = L exists and is finite, then f is bounded
x→a
in a neighbourhood of a.

Proof: Take ε = 1 in the definition of limit. Then there is a δ > 0 such that
¯ ¯ ¯ ¯
0 < |x − a| < δ =⇒ ¯ f (x) − L ¯ < 1 =⇒ ¯ f (x)¯ < 1 + |L| ,

and so f is bounded on this neighbourhood. ❑

270 T HEOREM (Order Properties of Limits) Let X j R, a ∈ R, and f : X → R. Let lim f (x) = L exist and be finite. Then
x→a

1. If s < L then there exists a neighbourhood Na of a contained in X such that ∀x ∈ Na , s < f (x).

2. If L < t then there exists a neighbourhood Na of a contained in X such that ∀x ∈ Na , f (x) < t.

3. If s < L < t then there exists a neighbourhood Na of a contained in X such that ∀x ∈ Na , s < f (x) < t.

4. If there exists a neighbourhood Na j X such that ∀x ∈ Na , s ≤ f (x), then s ≤ L.

82
 Chapter 5

5. If there exists a neighbourhood Na j X such that ∀x ∈ Na , f (x) ≤ t, then L ≤ t.

6. If there exists a neighbourhood Na j X such that ∀x ∈ Na , s ≤ f (x) ≤ t, then s ≤ L ≤ t.

Proof: We have

1. Take ε = L − s > 0 in the definition of limit. There is δ > 0 such that

¯ ¯
0 < |x − a| < δ =⇒ ¯ f (x) − L ¯ < L − s =⇒ s − L + L < f (x) < 2L − s =⇒ s < f (x),

as claimed.
2. Take ε = t − L > 0 in the definition of limit. There is δ > 0 such that
¯ ¯
0 < |x − a| < δ =⇒ ¯ f (x) − L ¯ < t − L =⇒ L − t + L < f (x) < t − L + L =⇒ f (x) < t ,

as claimed.
3. This follows by (1) and (2).
4. If on the said neighbourhood Na we had, on the contrary, L > s then (1) asserts that there is a neighbourhood
of Na0 j Na such that f (x) > s, a contradiction to the assumption that ∀x ∈ Na , s ≤ f (x).
5. If on the said neighbourhood Na we had, on the contrary, L < t then (2) asserts that there is a neighbourhood
of Na0 j Na such that f (x) < t, a contradiction to the assumption that ∀x ∈ Na , t ≥ f (x).
6. This follows by (4) and (5).

Analogous to the Sandwich Theorem for sequences we have

271 T HEOREM (Sandwich Theorem) Assume that a,b,c are functions defined on a neighbourhood Nx0 of a point x0 except
possibly at x0 itself. Assume moreover that in Nx0 they satisfy the inequalities a(x) ≤ b(x) ≤ c(x). Then

lim a(x) = L = lim c(x) =⇒ lim b(x) = L.

x→x 0 x→x 0 x→x 0

Proof: For all ε > 0 there is δ > 0 such that

0 < |x − x0 | < δ =⇒ |a(x) − L| < ε and |c(x) − L| < ε =⇒ L − ε < a(x) < L + ε and L − ε < c(x) < L + ε.

If we now consider x ∈ Nx0 ∩ {x : 0 < |x − x0 | < δ} then

L − ε < a(x) ≤ b(x) ≤ c(x) < L + ε =⇒ L − ε < b(x) < L + ε =⇒ |b(x) − L| < ε,

whence lim b(x) = L. ❑

x→x 0

272 T HEOREM Let X j R, a ∈ R, and f , g : X → R. Let (L,L 0 ,λ) ∈ R3 . Then

¯ ¯
1. lim f (x) = L =⇒ lim ¯ f (x)¯ = |L|.
x→a x→a
¯ ¯
2. lim f (x) = 0 ⇐⇒ lim ¯ f (x)¯ = 0.
x→a x→a

3. lim f (x) = L, lim g (x) = L 0 =⇒ lim ( f (x) + λg (x)) = L + λL 0 .

x→a x→a x→a

4. lim f (x) = L, lim g (x) = L =⇒ lim ( f (x)g (x)) = LL 0 .

0
x→a x→a x→a

5. If lim f (x) = 0 and if g is bounded on a neighbourhood Na of a, then lim f (x)g (x) = 0.

x→a x→a
µ ¶
f (x) L
6. lim f (x) = L, lim g (x) = L 0 6= 0 =⇒ lim = 0.
x→a x→a x→a g (x) L

83
 Limits of Functions

Proof:
¯¯ ¯ ¯ ¯ ¯
1. This follows from the inequality ¯¯ f (x)¯ − |L|¯ ≤ ¯ f (x) − L ¯.
¯ ¯ ¯ ¯ ¯ ¯
2. This follows from the inequalities − ¯ f (x)¯ ≤ f (x) ≤ ¯ f (x)¯ and min(− f (x), f (x)) ≤ ¯ f (x)¯ ≤ max(− f (x), f (x))
and the Sandwich Theorem.
3. For all ε > 0 there are δ1 > 0 and δ2 > 0 such that
¯ ¯ ¯ ¯
0 < |x − a| < δ1 =⇒ ¯ f (x) − L ¯ < ε, and 0 < |x − a| < δ2 =⇒ ¯ g (x) − L 0 ¯ < ε.

Take δ = min(δ1 ,δ2 ). Then

¯ ¯ ¯ ¯ ¯ ¯
0 < |x − a| < δ =⇒ ¯ f (x) + λg (x) − (L + λL 0 )¯ ≤ ¯ f (x) − L ¯ + |λ| ¯ g (x) − L 0 ¯ < (1 + |λ|)ε.

Since the dextral side can be made arbitrarily small, the assertion follows.
4. For all ε > 0 there are δ1 > 0 and δ2 > 0 such that
¯ ¯ ¯ ¯
0 < |x − a| < δ1 =⇒ ¯ f (x) − L ¯ < ε, and 0 < |x − a| < δ2 =⇒ ¯ g (x) − L 0 ¯ < ε.

Also, by Theorem 269, g is locally bounded and so there exists B > 0, and δ3 > 0 such that
¯ ¯
0 < |x − a| < δ3 =⇒ ¯ g (x)¯ < B .

Take δ = min(δ1 ,δ2 ,δ3 ). Then

¯ ¯ ¯ ¯ ¯ ¯¯ ¯ ¯ ¯ ¯ ¯
¯ f (x)g (x) − LL 0 ¯ = ¯( f (x) − L)g (x) + L(g (x) − L 0 )¯ ≤ ¯ f (x) − L ¯ ¯g (x)¯ + |L| ¯g (x) − L 0 ¯ < (B + ¯L 0 ¯)ε.

As the dextral side can be made arbitrarily small, the result follows.
5. For all ε > 0 there are δ1 > 0, B > 0, and δ2 > 0 such that
¯ ¯ ¯ ¯
0 < |x − a| < δ1 =⇒ ¯ f (x)¯ < ε, and 0 < |x − a| < δ2 =⇒ ¯ g (x)¯ < B .

Take δ = min(δ1 ,δ2 ). Then ¯ ¯ ¯ ¯

¯ f (x)g (x)¯ ≤ |B | ¯ f (x)¯ < B ε.

As the dextral side can be made arbitrarily small, the result follows.
¯ 0¯
¯ ¯ ¯ 0¯ ¯L ¯
¯ ¯ ¯ ¯
6. First g (x) → L as x → a by part (1). Hence, for ε = ¯¯ ¯¯ > 0 there is a sufficiently small δ0 > 0 such that
2
¯ ¯ ¯ ¯ ¯ ¯ ¯ 0¯ ¯ ¯
¯¯ ¯ ¯ ¯¯ ¯L 0 ¯ ¯ 0 ¯ ¯L 0 ¯ ¯ ¯ ¯ 0 ¯ ¯L 0 ¯ ¯L ¯ ¯ ¯ 3 ¯L 0 ¯
¯¯g (x)¯ − ¯L 0 ¯¯ < ¯
=⇒ L − ¯ ¯ ¯ ¯
< g (x) < L +¯ =⇒ ¯ ¯
< g (x) < ,
2 2 2 2 2
¯ ¯
that is, ¯g (x)¯ is bounded away from 0 x sufficiently close to a. Now, for all ε > 0 there are δ1 > 0 and δ2 > 0
such that ¯ ¯ ¯ ¯
0 < |x − a| < δ1 =⇒ ¯ f (x) − L ¯ < ε, and 0 < |x − a| < δ2 =⇒ ¯ g (x) − L 0 ¯ < ε.
For δ = min(δ1 ,δ2 ,δ0 ),
¯ 0¯ ¯ ¯
¯L ¯ ¯ ¯ 3 ¯L 0 ¯
0 < |x − a| < δ =⇒ L − ε < f (x) < L + ε, ¯ ¯
< g (x) < , and L 0 − ε < g (x) < L 0 + ε.
2 2
Hence
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯¯ ¯ ¯ ¯ ¯ ¯
¯ f (x) L ¯ ¯ L 0 f (x) − Lg (x) ¯ ¯ L 0 ( f (x) − L) − L(g (x) − L 0 ) ¯ ¯L 0 ¯ ¯ f (x) − L ¯ + |L| ¯ g (x) − L 0 ¯ 2(¯L 0 ¯ + |L|)ε
¯ ¯ ¯ ¯=¯ ¯≤ ¯ ¯ ,
¯ g (x) − L 0 ¯ = ¯ g (x)L 0 ¯ ¯ g (x)L 0 ¯ ¯g (x)¯ |L 0 |
<
|L 0 | |L 0 |

which gives the result.

In the manner of proof of Proposition 265, we may prove the following two propositions.

84
 Chapter 5

i h
273 D EFINITION -P ROPOSITION (Cauchy-Heine, Limit at +∞) Let f : a ;+∞ → R The following are equivalent.
i h
1. For each sequence {xn }+∞
n=1 of points in the interval a ;+∞ ,

xn → +∞ =⇒ f (xn ) → L.

2. ∀ε > 0, ∃M, M > max(0, a), such that ¯ ¯

x ≥ M =⇒ ¯ f (x) − L ¯ < ε.

If either condition is fulfilled we say that f has a limit L as x tends towards +∞ and we write
L = lim f (x).
x→+∞

i h
274 D EFINITION -P ROPOSITION (Cauchy-Heine, Limit at −∞) Let f : − ∞ ; a → R The following are equivalent.
i h
1. For each sequence {xn }+∞
n=1 of points in the interval − ∞ ; a ,

xn → −∞ =⇒ f (xn ) → L.

2. ∀ε > 0, ∃M, M < min(0, a), such that ¯ ¯

x ≤ M =⇒ ¯ f (x) − L ¯ < ε.

If either condition is fulfilled we say that f has a limit L as x tends towards −∞ and we write
L = lim f (x).
x→−∞

275 Definition We write lim f (x) = +∞ or lim f (x) = +∞ if ∀M > 0, ∃δ > 0 such that
x→a+ x&a
i h
x ∈ a ; a + δ =⇒ f (x) > M.

Similarly, we write lim f (x) = +∞ or lim f (x) = +∞ if ∀M > 0, ∃δ > 0 such that
x→a− x%a
i h
x ∈ a − δ ; a =⇒ f (x) > M.

Finally, we write lim f (x) = +∞ if ∀M > 0, ∃δ > 0 such that

x→a
i h
x ∈ a − δ ; a + δ =⇒ f (x) > M.

276 Definition We write lim f (x) = −∞ or lim f (x) = −∞ if ∀M < 0, ∃δ > 0 such that
x→a+ x&a
i h
x ∈ a ; a + δ =⇒ f (x) < M.

Similarly, we write lim f (x) = −∞ or lim f (x) = −∞ if ∀M < 0, ∃δ > 0 such that
x→a− x%a
i h
x ∈ a − δ ; a =⇒ f (x) < M.

Finally, we write lim f (x) = −∞ if ∀M < 0, ∃δ > 0 such that

x→a
i h
x ∈ a − δ ; a + δ =⇒ f (x) < M.

277 T HEOREM Let X ,Y be subsets of R, a ∈ X and b ∈ Y , f : X → R, g : Y → R such that f (X ) j Y , and let L ∈ R. Then

lim f (x) = a and lim g (x) = L =⇒ lim (g ◦ f )(x) = L.

x→a x→b x→a

Proof:
❑

Homework

85
 Continuity

1
Problem 5.1.1 Prove that lim sin does not exist. lim f (x) = +∞. Demonstrate that
x→0 x x→a

1. If lim g (x) = +∞, then lim ( f (x) + g (x)) = +∞.

x→a x→a
Problem 5.1.2 Let m,n be strictly positive integers. Prove that 2. If lim g (x) = L ∈ R, then lim ( f (x) + g (x)) = +∞.
xn − 1 n x→a x→a
lim = . 3. If lim g (x) = +∞, then lim ( f (x)g (x)) = +∞.
x→1 x m − 1 m x→a x→a
4. If lim g (x) = L > 0, then lim ( f (x)g (x)) = +∞.
x→a x→a
Problem 5.1.3 Let X j R, a ∈ R, and f , g : X → R. If f (x) → +∞
and there exists a neighbourhood Na j X of a where f (x) ≤ g (x),
prove that g (x) → +∞. Problem 5.1.5 (Cauchy Criterion for Functional Limits) Let X j
R, a ∈ X , and f : X → R. Prove that f has a limit at a (finite ¯ or ¯infi-
nite) if and
¯ only if for all
¯ ε > 0 there is a δ > 0 such that ¯x 0 − x 00 ¯ < δ
Problem 5.1.4 Let X j R, a ∈ R, and f , g : X → R. Suppose that implies ¯ f (x 0 ) − f (x 00 )¯ < ε.

5.2 Continuity
i h i h
278 Definition A function f : a ;b → R is said to be continuous at the point x0 ∈ a ;b , if we can exchange limiting
operations, as in µ ¶
lim f (x) = f lim x (= f (x0 )).
x→x 0 x→x 0

In other words, a function is continuous at the point x0 if

¯ ¯
∀ε > 0,∃δ > 0, such that |x − x0 | < δ =⇒ ¯ f (x) − f (x0 )¯ < ε.

h i
279 Definition A function f : a ;b → R is said to be right continuous at a, if

f (a) = f (a+).

It is said to be left continuous at b, if

f (b) = f (b−).

In view of the above definitions and Proposition 267, we have the following

280 T HEOREM The following are equivalent.

i h i h
1. The function f : a ;b → R is continuous at the point x0 ∈ a ;b .

2. f (x0 −) = f (x0 ) = f (x0 +).

i h
3. If {xn }+∞
n=1 , and for all n, xn ∈ a ;b , then xn → x0 =⇒ f (xn ) → f (x0 ).

281 Example What are the points of discontinuity of the function

h h
0 ;+∞ → R

 1 h h p
f : 
 ?
 if x ∈ Q ∩ 0 ;+∞ , x = , in lowest terms
x 7→ p +q q

 h h

 0 if x ∈ 0 ;+∞ \ Q

h h h h
Solution: Let a ∈ Q. Since 0 ;+∞ \ Q is dense in 0 ;+∞ , there exists a sequence {a n }+∞ n=1 of points in
h h
0 ;+∞ \ Q such that a n → a as n → +∞. Observe that f (a n ) = 0 but f (a) 6= 0. Hence a n → a does not imply
h h
f (a n ) → f (a) and f is not continuous at a. On the other hand, let n ∈ 0 ;+∞ \ Q. Then f (b) = 0. Let {bn }+∞
n=1 be
h h pn
a sequence in 0 ;+∞ ∩ Q converging to b, bn = in lowest terms. By Dirichlet’s Approximation Theorem we
qn

86
 Chapter 5

1
must have p n → +∞ and q n → +∞. Hence → 0. So f is continuous at b. In conclusion, f is continuous
h h p n + qn h h
at every irrational in 0 ;+∞ and discontinuous at every rational in 0 ;+∞ .

¡ ¢
282 D EFINITION -P ROPOSITION (Oscillation of a function at a point) Let f be bounded. The function ω : Dom f → [0;+∞[,
called the oscillation of f at x and given by
¯ ¯
ω( f , x) = lim sup{¯ f (a) − f (b)¯ : |a − x| < δ,|b − x| < δ}
δ→0+

is well-defined. Moreover, f is continuous at x if and only if ω( f , x) = 0.

Proof: Observe that in fact

¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
ω( f , x) = lim sup{¯ f (a) − f (b)¯ : |a − x| < δ,|b − x| < δ} = inf sup{¯ f (a) − f (b)¯ : |a − x| < δ,|b − x| < δ} ≤ ¯ f (a) − f (b)¯ ≤ 2 ¯ f ¯ < +∞
δ→0+ δ>0

This says that ω( f , x) is well-defined.

❑
h i i h
283 Definition We say that a function f is continuous on the closed interval a ;b if it is continuous everywhere on a ;b ,
continuous on the right at a and continuous on the left at b. If X j R, then f : X → R is said to be continuous on X (or
continuous) if it is continuous at every element of X .

284 T HEOREM Let X j R. A function f : X → R is continuous if and only if the the inverse image of an open set is open in X .

Proof:

=⇒ Let A j R be an open set. We must shew that

i f −1 (A) is openhin X . Let a ∈ f −1 (A). Since f (a) ∈ A and A is
open in R, there exists an r > 0 such that f (a) − r ; f (a) + r j A. Since f is continuous at a, there exists a
δ > 0 such that
¯ ¯ i h i h
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < r, that is, x ∈ a + δ ; a − δ =⇒ f (x) ∈ f (a) − r ; f (a) + r ,
i h ³i h´
that is, x ∈ a + δ ; a − δ =⇒ x ∈ f −1 f (a) − r ; f (a) + r ,
i h ³i h´
that is, a + δ ; a − δ j f −1 f (a) − r ; f (a) + r
³i h´ i h
Since f −1 f (a) − r ; f (a) + r j f −1 (A), we have shewn that a + δ ; a − δ j f −1 (A), which means that
for any a, a neighbourhood of a lies entirely in f −1 (A), that is, f −1 (A) is open.
⇐ Given ε > 0, we must find a δ > 0 such that for all a ∈ X ,
¯ ¯
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < ε.

Now ¯ ¯ i h ³i h´
¯ f (x) − f (a)¯ < ε =⇒ f (x) ∈ f (a) − ε ; f (a) + ε =⇒ x ∈ f −1 f (a) − ε ; f (a) + ε .
i h ³i h´
Now, f (a) − ε ; f (a) + ε j R is open in R, and so, by assumption, so is f −1 f (a) − ε ; f (a) + ε . This
³i h´
means that if t ∈ f −1 f (a) − ε ; f (a) + ε then there is a r > 0 such that
i h ³i h´
t − r ; t + r j f −1 f (a) − ε ; f (a) + ε .
³i h´
But clearly a ∈ f −1 f (a) − ε ; f (a) + ε , and hence there is a δ > 0 such that
i h ³i h´
a − δ ; a + δ j f −1 f (a) − ε ; f (a) + ε .

87
 Continuity

Thus i h ³i h´
x ∈ a − δ ; a + δ =⇒ x ∈ f −1 f (a) − ε ; f (a) + ε ,

or equivalently,
i h
|x − a| < δ =⇒ f (x) ∈ f (a) − ε ; f (a) + ε ,

that is,
¯ ¯
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < ε,

as we needed to shew.

285 T HEOREM Let X j R. A function f : X → R is continuous if and only if the the inverse image of a closed set is closed in
X.

Proof: Let F j R be a closed set. Then R \F is open. By Theorem 284 f −1 (R \F ) is open in X , and so X \ f −1 (R \F )
is closed in X . But X \ f −1 (R \ F ) = f −1 (F ), proving the theorem. ❑

286 T HEOREM If two continuous functions agree on a dense set of the reals, then they are identical. That is, if X j R is dense
in R and if f : R → R and g : R → R satisfy f (x) = g (x) for all x ∈ X , then f (x) = g (x) for all x ∈ R.

Proof: Let a ∈ R \ X . Since X is dense in R, there is a sequence {xn }+∞

n=1 j X such that xn → a as n → +∞. Notice
that since xn ∈ X , we have f (xn ) = g (xn ). By continuity
³ ´ ³ ´
f (a) = f lim xn = lim f (xn ) = lim g (xn ) = g lim xn = g (a),
n→+∞ n→+∞ n→+∞ n→+∞

proving the theorem. ❑

287 T HEOREM (Cauchy’s Functional Equation) Let f be a continuous function defined over the real numbers that satisfies
the Cauchy functional equation:
∀(x, y) ∈ R2 , f (x + y) = f (x) + f (y).

Then f is linear, that is, there is a constant c such that f (x) = c x.

Proof: Our method of proof is as follows. We first prove the assertion for positive integers n using induction. We
then extend our result to negative integers. Thence we extend the result to reciprocals of integers and after that to
rational numbers. Finally we extend the result to all real numbers by means of Theorem 286.

We prove by induction that for integer n ≥ 0, f (nx) = n f (x). Using the functional equation,

f (0 · x) = f (0 · x + 0 · x) = f (0 · x) + f (0 · x) =⇒ f (0 · x) = 0 f (x),

and the assertion follows for n = 0. Assume n ≥ 1 is an integer and that f ((n − 1)x) = (n − 1) f (x). Then

f (nx) = f ((n − 1)x + x) = f ((n − 1)x) + f (x) = (n − 1) f (x) + f (x) = n f (x),

proving the assertion for all strictly positive integers.

Let m < 0 be an integer. Then −m > 0 is a strictly positive integer, for which the result proved in the above
paragraph holds, and thus and by the above paragraph, f (−mx) = −m f (x). Now,

0 = f (0) =⇒ 0 = f (mx + (−mx)) = f (mx) + f (−mx) =⇒ f (mx) = − f (−mx) = −(−m f (x)) = m f (x),

and the assertion follows for negative integers. We have thus proved the theorem for all integers.

88
 Chapter 5

a ³ a´ ³a´
Assume now that x = , with a ∈ Z and b ∈ Z \ {0}. Then f (a) = f (a · 1) = a f (1) and f (a) = f b =bf by
b b b
the result we proved for integers and hence
³a´ ³a ´ ³a ´
a f (1) = b f =⇒ = f = f (1) .
b b b

We have established that for all rational numbers x ∈ Q, f (x) = x f (1).

We have not used the fact that the function is continuous so far. Since the rationals are dense in the reals the
extension of the result now follows from Theorem 286.❑

Homework

Problem 5.2.1 Find all functions f : R → R, continuous at x = 0 Problem 5.2.6 What are the points of discontinuity of the function
such that ∀x ∈ R, f (x) = f (3x).
R → R

f : 
 ?
Problem 5.2.2 Find all functions  0 if x ∈ Q
µ ¶ f : R → R, continuous at x = 0
x x 7→
such that ∀x ∈ R, f (x) = f . 

1 + x2  1 if x ∈ R \ Q

Problem 5.2.3 Determine the setqof points of discontinuity of the Problem 5.2.7 What are the points of discontinuity of the function
function f : R → R, f : x 7→ Tx U + x − Tx U.
R → R

f : 
 ?
Problem 5.2.4 What are the points of discontinuity of the function  cos x if x ∈ Q
x 7→


R → R  sin x if x ∈ R \ Q

f : 
 ?
 x if x ∈ Q
x 7→ Problem 5.2.8 Find all functions f : R → R, continuous at x = 1


 0 if x ∈ R \ Q such that ∀x ∈ R, f (x) = −f (x 2 ).

Problem 5.2.9 Let a ∈ R be fixed. Find all functions f : R → R, con-

Problem 5.2.5 What are the points of discontinuity of the function
tinuous everywhere such that ∀(x, y ) ∈ R2 , f (x − y ) = f (x) − f (y ) +
R → R ax y .

f : 
 ? h h h h
 0 if x ∈ Q
x 7→ Problem 5.2.10 Let f : 0 ;+∞ → 0 ;+∞ , x 7→

 r q
 x if x ∈ R \ Q p
x + x + x + · · ·. Is f right-continuous at 0?

5.3 Algebraic Operations with Continuous Functions

i h i h
288 T HEOREM (Algebra of Continuous Functions) Let f , g : a ;b → R be continuous a the point x0 ∈ a ;b . Then

1. f + g is continuous at x0 .

2. f g is continuous at x0 .

f
3. if g (x0 ) 6= 0, is continuous at x0 .
g

Proof: This follows directly from Theorem 272. ❑

289 T HEOREM Let X ,Y be subsets of R, a ∈ X and b ∈ Y , f : X → R, g : Y → R such that f (X ) j Y . If f is continuous at a

and g is continuous at f (a), then g ◦ f is continuous at a.

89
 Monotonicity and Inverse Image

Proof: This follows at once from Theorem 277. ❑

290 T HEOREM Let f : I → R be a monotone function, where I j R is a non-empty interval. Then the set of points of discon-
tinuity of f is either finite or countable.

With Theorems 288 and 289 we can now demonstrate the

5.4 Monotonicity and Inverse Image

291 Definition Let X and Y be subsets of R. Let f : X → Y , and assume that X has at least two elements. Then f is said to
be
f (b) − f (a)
• increasing if ∀(a,b) ∈ X 2 , a < b =⇒ f (a) ≤ f (b). Equivalently, if the ratio ≥ 0.
b−a
f (b) − f (a)
• strictly increasing if ∀(a,b) ∈ X 2 , a < b =⇒ f (a) < f (b). Equivalently, if the ratio > 0.
b−a
f (b) − f (a)
• decreasing if ∀(a,b) ∈ X 2 , a < b =⇒ f (a) ≥ f (b). Equivalently, if the ratio ≤ 0.
b−a
f (b) − f (a)
• strictly decreasing if ∀(a,b) ∈ X 2 , a < b =⇒ f (a) > f (b). Equivalently, if the ratio < 0.
b−a
f is said to be monotonic if it is either increasing or decreasing, and strictly monotonic if it is either strictly increasing or
strictly decreasing.

☞ Observe that if f is increasing, then −f is decreasing, and conversely. Similarly for strictly monotonic functions.

292 T HEOREM Let X j R and let f : X → R be strictly monotone. Then f is injective.

Proof: Recall that f is injective if x 6= y =⇒ f (x) 6= f (y). If f is strictly increasing then x < y =⇒ f (x) < f (y)
and if f is strictly decreasing then x < y =⇒ f (x) > f (y). In either case, the condition for injectivity is fulfilled.
❑

293 T HEOREM Let I j R be an interval and let f : I → f (I ) be strictly monotone. Then f −1 is strictly monotone in the same
sense as f .

Proof: Assume first that f is strictly increasing and put x = f −1 (a), y = f −1 (b) and that a < b. If x ≥ y, then,
since f is strictly increasing, f (x) ≥ f (y). But then, f ( f −1 (a)) ≥ f ( f −1 (b)) =⇒ a ≥ b, a contradiction.

A similar argument finishes the theorem for f strictly decreasing.

❑

The following theorem is remarkable, since it does not allude to any possible continuity of the function in question.

294 T HEOREM Let I j R be an interval and let f : I → f (I ) be strictly monotone. Then f −1 is continuous.

Proof: Let b ∈ f (I ), b = f (a), and ε > 0. We must shew that there is δ > 0 such that
¯ ¯ ¯ ¯
¯ y − b ¯ < δ =⇒ ¯ f −1 (y) − b ¯ < ε.
i h
If a is not an endpoint of I , there is an α > 0 such that a − α ; a + α j I . Put ε0 = min(ε,α). Since both f and
f −1 are both strictly monotone
¯ −1 ¯
¯ f (y) − b ¯ < ε0 =⇒ b −ε0 < f −1 (y) < b +ε0 =⇒ f (b −ε0 ) < f ( f −1 (y)) < f (b +ε0 ) =⇒ f (b −ε0 ) < y < f (b +ε0 ).

90
 Chapter 5

Since f is strictly increasing and a −ε0 < a, f (a −ε0 ) < f (a) = b. Thus there must be an η > 0 such that f (a −ε0 ) =
b − η < b. Similarly, there is an η0 such that b < b + η0 = f (a + ε0 ). Putting η00 = min(η,η0 ), we have that for all
y ∈ f (I ),
¯ ¯
¯ y − b ¯ < η00 =⇒ b − η00 < y < b + η00

=⇒ b − η < y < b + η0

=⇒ a − ε0 < f −1 (y) < a + ε0

¯ −1 ¯
=⇒ ¯ f (y) − f −1 (b)¯ < ε0 ,

finishing the proof for when a is not an endpoint. If a were an endpoint, the above proof carries by suppressing
one of η or η0 . ❑
h i ³h i´
295 T HEOREM A continuous function f : a ;b → f a ;b is invertible if and only if it is strictly monotone.

Proof:

=⇒ Assume f is continuous and invertible. Since f is injective, f (a) 6= f (b). Assume that f (a) < f (b), if
0 0 0 0
f (a) > f (b) the argument
h is isimilar. We would like to shew that if a < b =⇒ f (a ) < f (b ). Consider the
continuous function g : 0 ;1 → R,

g (t) = f ((1 − t)a + t a 0 ) − f ((1 − t)b + tb 0 ).

We have
g (0) = f (a) − f (b) < 0 and g (1) = f (a 0 ) − f (b 0 ).
0 0 0 0
If g (1) = 0, then we must have
i a =h b , contradicting a < b . If g (1) > 0, then by the Intermediate Value
Theorem there must be an s ∈ 0 ;1 such that g (s) = 0. This entails

(1 − s)a + sa 0 = (1 − s)b + sb 0 =⇒ 0 > (1 − s)(a − b) = s(b 0 − a 0 ) > 0,

absurd. This entails that g (1) < 0 =⇒ f (a 0 ) < f (b 0 ), as wanted.

⇐ Trivially, f is surjective. If f is strictly monotone, then f is injective by Theorem 292, and thus f is invertible,
by Theorem 27.

5.5 Convex Functions

296 Definition Let A × B j R2 . A function f : A → B is convex in A if ∀(a,b,λ) ∈ A 2 × [0;1],

f (λa + (1 − λ)b) ≤ f (a)λ + (1 − λ) f (b).

It is strictly convex if the inequality above is strict. Similarly, a function g : A → B is concave in A if ∀(a,b,λ) ∈ A 2 × [0;1],

g (λa + (1 − λ)b) ≥ g (a)λ + (1 − λ)g (b).

It is strictly concave if the inequality above is strict.

5.5.1 Graphs of Functions

297 Definition Given a function f , its graph is the set on the plane

Γ f = {(x, y) ∈ R2 : y = f (x)}.

91
 Classical Functions

298 Example Figures ?? through ?? shew the graphs of a few standard functions, with which we presume the reader to be
familiar.

5.6 Classical Functions

5.6.1 Affine Functions
299 Definition An affine function is one with assignment rule of the form x 7→ ax + b, where a,b are real constants.

300 T HEOREM The graph of an affine function is a line on the plane. Conversely, any non-vertical straight line on the plane
is the graph on an affine function.

5.6.2 Quadratic Functions

5.6.3 Polynomial Functions
5.6.4 Exponential Functions
n³ x ´n o+∞
301 D EFINITION -P ROPOSITION Let x ∈ R be fixed. The sequence 1+ is bounded and strictly increasing. Thus it
n n>−x
converges and we define the natural exponential function by
³ x ´n
exp : R → R, exp(x) := lim 1+ .
n→+∞ n

x x
Proof: Observe that 1 + > 0 for n > −x. Using the AM-GM Inequality with x1 = 1, x2 = · · · = xn+1 = 1 +
n n
³ x´
³ 1+n 1+
x ´n/(n+1) ³ ´ ³
n = 1 + x =⇒ 1 + x n < 1 + x n+1 ,
´
1+ <
n n +1 n +1 n n +1
whence the sequence is increasing.

³ µ ¶
x ´n 1 n
For 0 < x ≤ 1 then 1 + ≤ 1+ < e, by Theorem 177.
n n

If x > 1 then by the already proved monotonicity,

³ µ ¶ µ ¶
x ´n Tx U + 1 n Tx U + 1 n(Tx U+1)
1+ ≤ 1+ < 1+ < e Tx U+1 .
n n n(Tx U + 1)

x ³ x ´n
If x ≤ 0 then 1 + ≤ 1 and so 1 + ≤ 1. ❑
n n

☞ By Theorem 177, exp(1) = e. We will later prove, in ????, that for all x ∈ R, exp(x) = e x .

5.6.5 Logarithmic Functions

5.6.6 Trigonometric Functions
i πh
302 T HEOREM Let x ∈ 0 ; . Then sin x < x < tan x.
2

Proof:
❑

92
 Chapter 5

Homework

Problem 5.6.1 How many solutions does the equation have?

x
sin x =
100 Problem 5.6.4 How many solutions does the equation
have?
x
sin(sin(sin(sin(sin(x))))) =
3
Problem 5.6.2 Prove that
2 h πi have?
x ≤ sin(x) ≤ x,∀ x ∈ 0 ; .
π 2
Problem 5.6.5 (Chebyshev Polynomials)
Problem 5.6.3 How many solutions does the equation

sin x = log x Problem 5.6.6 (Cardano’s Formula)

5.6.7 Inverse Trigonometric Functions

5.7 Continuity of Some Standard Functions.

5.7.1 Continuity Polynomial Functions
303 L EMMA Let K ∈ R be a constant. The constant function f : R → R, f (x) = K is everywhere continuous.

Proof: Given a ∈ R and ε > 0, take δ = ε. Then clearly

¯ ¯
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < ε,

since f (x) = f (a) = K and the quantity after the implication is 0 < ε and we obtain a tautology. ❑

304 L EMMA The identity function f : R → R, f (x) = x is everywhere continuous.

Proof: Given a ∈ R and ε > 0, take δ = ε. Then clearly

¯ ¯
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < ε,

since the quantity after the implication is |x − a| < δ and we obtain a tautology. ❑

305 L EMMA Given a strictly positive integer n, the power function f : R → R, f (x) = x n is everywhere continuous.

Proof: By Lemma 304, the function x 7→ x is continuous. Applying this Lemma and the product rule from
Theorem 288 n times, we obtain the result. ❑

306 T HEOREM (Continuity of Polynomial Functions) Let n be a fixed positive integer. Let a k ∈ R, 0 ≤ k ≤ n be constants.
Then the polynomial function f : R → R, f (x) = a 0 + a 1 x + a 2 x 2 + · · · + a n x n is everywhere continuous.

Proof: This follows from Lemma 305 and the sum rule from Theorem 288 applied n + 1 times. ❑

5.7.2 Continuity of the Exponential and Logarithmic Functions

307 L EMMA Let a > 1. The exponential function R → R, x 7→ a x is continuous at x = 0.

Proof: For integral n > 0 we know that lim a 1/n = 1 by virtue of Theorem 174. We wish to shew that a x → 1 as
n→+∞
1
x → 0. Observe first that lim a −1/n = lim = 1 also. Thus given ε > 0, and since a > 1, there is N > 0 such
n→+∞ n→+∞ a 1/n
that
1 − ε < a −1/N < a 1/N < 1 + ε.

93
 Continuity of Some Standard Functions.

i 1 1h
If x ∈ − ; then,
N N
a −1/N < a x < a 1/N .

By the above, this implies that

¯ ¯ ¯ ¯
1 − ε < a x < 1 + ε =⇒ ¯a x − 1¯ < ε =⇒ ¯ a x − a 0 ¯ < ε,

finishing the proof. ❑

i h
308 T HEOREM (Continuity of the Exponential Function) Let a > 0, a 6= 1. The exponential function f : R → 0 ;+∞ , x 7→ a x
is everywhere continuous.

Proof: Assume first that a > 1. Let us shew that it is continuous at an arbitrary u ∈ R. If x → u then x − u → 0.
Thus
lim a x = a u lim a x−u = a u lim a x−u = a u lim a t = a u · 1 = a u ,
x→u x→u x−u→0 t →0

by Lemma 307, and so the continuity is established for a > 1.

1 1
If 0 < a < 1 then > 1 and by what we have proved, x 7→ x is continuous. Then
a a

1 1
lim a x = lim = = au ,
x→u x→u 1 1
ax au

proving continuity in the case 0 < a < 1.❑

i h
309 L EMMA Let a > 0, a 6= 1. Then 0 ;+∞ → R, x 7→ loga x is everywhere continuous.

i h
Proof: Its inverse function R → 0 ;+∞ , x 7→ a x , is everywhere continuous and strictly monotone. The result
then follows from Theorem 294. ❑

5.7.3 Continuity of the Power Functions

i h i h
310 T HEOREM Let p ∈ R. Then 0 ;+∞ → 0 ;+∞ , x 7→ x p is everywhere continuous.

Proof: This follows by the continuity of compositions: x p = e p log x . ❑

Homework

h i h i
Problem 5.7.1 Prove the continuity of the function R → − 1 ;1 , Problem 5.7.4 Prove the continuity of the function − 1 ;1 →
h i
x 7→ sin x. 0 ;π , x 7→ arccos x.
h i h
Problem 5.7.2 Prove the continuity of the function − 1 ;1 → − π
π πi Problem 5.7.5 Prove the continuity of the function R \ (2Z +1)
2
→
; , x 7→ arcsin x.
2 2 R, x 7→ tan x.

h i i π πh
Problem 5.7.3 Prove the continuity of the function R → − 1 ;1 , Problem 5.7.6 Prove the continuity of the function R → − ; ,
2 2
x 7→ cos x. x 7→ arctan x.

94
 Chapter 5

5.8 Inequalities Obtained by Continuity Arguments

The technique used Theorem 287, of proving results in a dense set of the real numbers and extending the result by continuity
can be exploited in a variety of situations. We now use it to give a generalisation of Bernoulli’s Inequality.

311 T HEOREM (Generalisation of Bernoulli’s Inequality) Let (α, x) ∈ R2 with x ≥ −1. If 0 < α < 1 then

(1 + x)α ≤ 1 + αx.
i h i h
If α ∈ − ∞ ;0 ∪ 1 ;+∞ then
(1 + x)α ≥ 1 + αx.

Equality holds in either case if and only if x = 0.

m
Proof: Let α ∈ Q, 0 < α < 1. Then α = for integers m,n with 1 ≤ m < n. Since x + 1 ≥ 0, we may use the
n
AM-GM Inequality to obtain

(1 + x)α = (1 + x)m/n
¡ ¢1/n
= (1 + x)m · 1n−m
m(1 + x) + (n − m) · 1
≤
n
n + mx
=
n
m
= 1+ x
n
= 1 + αx.

Equality holds when are the factors are the same, that is, when 1 + x = 1 =⇒ x = 0.

Assume now that α ∈ R \ Q with 0 < α < 1. We can find a sequence of rational numbers {a n }+∞
n=1 j Q such that
a n → α as n → +∞. Then
(1 + x)a n ≤ 1 + a n x,

whence by the continuity of the power functions (Theorem 310),

(1 + x)α = lim (1 + x)a n ≤ lim (1 + a n x) = 1 + αx,

n→+∞ n→+∞

giving the result for all real numbers α with 0 < α < 1, except that we need to prove that equality holds only for
x = 0. Take a rational number r with 0 < α < r < 1, and recall that we are assuming that α is irrational. Then
³ α ´r
(1 + x)α = (1 + x)α/r )r ≤ 1 + x .
r
³ α ´r
Since the exponent on the right is rational, by what we have proved above 1 + x ≤ 1 + x with equality if and
r
only if x = 0. Hence the full result has been proved for the case α ∈ R with 0 < α < 1.

Let α > 1. If 1 + αx < 0, then obviously (1 + x)α > 0 > 1 + αx, and there is nothing to prove. Hence we will assume
1
that αx ≥ −1. By the first part of the theorem, since 0 < < 1,
α
1
(1 + αx)1/α ≤ 1 + · αx = 1 + x =⇒ 1 + αx ≤ (1 + x)α ,
α

with equality only if x = 0. The theorem has been proved for α > 1.

95
 Inequalities Obtained by Continuity Arguments

Finally, let α < 0. Again, if 1+αx < 0, then obviously (1+ x)α > 0 > 1+αx, and there is nothing to prove. Assume
thus αx ≥ −1. Choose a strictly positive integer n satisfying 0 < −α < n. Now,

α2 2 ³ α ´³ α ´ 1 α
1≥ 1− 2
x = 1 − x 1 + x =⇒ α ≥ 1 + x,
n n n 1− x n
n
and so by the first pat of the theorem

α 1
(1 + x)−α/n ≤ 1 − x =⇒ (1 + x)α/n ≥ α
n 1− x
n
α/n α
=⇒ (1 + x) ≥ 1+ x
n
³ α ´n
α
=⇒ (1 + x) ≥ 1 + x ,
n
³ α ´n α
and since n is a positive integer, 1 + x ≥ 1 + n · x = 1 + αx and so (1 + x)α ≥ 1 + αx also when α < 0. This
n n
finishes the proof of the theorem. ❑

312 T HEOREM (Monotonicity of Power Means) Let a 1 , a 2 ,... , a n be strictly positive real numbers and let (α,β) ∈ R2 be such
that α · β 6= 0 and α < β. Then
µ ¶1/α Ã β β β !1/β
a 1α + a 2α + · · · + a nα a1 + a2 + · · · + an
≤ ,
n n
with equality if and only is a 1 = a 2 = · · · = a n .

µ ¶1/α µ ¶α
a 1α + a 2α + · · · + a nα ak
Proof: Assume first that 0 < α < β. Put cα = and d k = . Observe that
n cα
µ ¶β µ ¶β µ ¶β 1/β
a1 a2 an Ã
+ +··· + β/α β/α β/α !1/β
cβ 
 cα cα cα

 d1 + d2 + · · · + dn
=  = ,
cα  n  n

and that
µ ¶1/α µ ¶1/α
d1 + d2 + · · · + dn 1 a 1α + a 2α + · · · + a nα
= = 1 =⇒ d 1 + d 2 + · · · + d n = n.
n cα n
Put d k = 1 + xk . Then x1 + x2 + · · · + xn = 0. By Theorem 311,

β/α β
dk = (1 + xk )β/α ≥ 1 + xk . (5.1)
α
Letting k run from 1 through n and adding,

β/α β/α β/α β

d1 + d2 + · · · + dn ≥n+ (x1 + x2 + · · · + xn ) = n.
α
Hence
β/α β/α β/α
d1 + d2 + · · · + dn cβ
≥ 1 =⇒ ≥ 1,
n cα
proving the theorem when 0 < α < β.

β/α β/α β/α

β d + d2 + · · · + dn
If α < β < 0, then 0 < < 1. The inequality in (5.1) is reversed, giving 1 ≤ 1, and since
α n
β < 0,
à β/α β/α β/α !1/β
cβ d1 + d2 + · · · + dn
= ≥ 11/β = 1,
cα n

96
 Chapter 5

proving the theorem when α < β < 0.

Finally, we tackle the case α < 0 < β. By the AM-GM Inequality, putting G = (a 1 a 2 · · · a n )1/n
a 1α + a 2α + · · · + a nα
G α = (a 1α a 2α · · · a nα )1/n ≤ .
n
Raising the quantities at the extreme of the inequalities to the power −1/α and remembering that −1/α > 0, we
gather that
µ α ¶
a 1 + a 2α + · · · + a nα 1/α
≤ G.
n
In a similar manner,
β β β
β β β a + a2 + · · · + an
G β = (a 1 a 2 · · · a n )1/n ≤ 1 ,
n
and
à β β β !1/β
a1 + a2 + · · · + an
G≤ ,
n
since β > 0. This finishes the proof. ❑

313 L EMMA Let α, a, x be real numbers with α > 1, a > 0, and x ≥ 0. Then
³ a ´α/(α−1)
x α − ax ≥ (1 − α) .
α

Proof: By Theorem 311, since α > 1,

(1 + z)α ≥ 1 + αz, z ≥ −1,
with equality only is z = 0. Putting z = 1 + y,

y α ≥ 1 + α(y − 1) =⇒ y α − αy ≥ 1 − α, y ≥ 0,

with equality only if y = 1. Let c > 0 be a constant. Multiplying the above inequality by c α we obtain

(c y)α − αc α−1 (c y) ≥ (1 − α)c α , for y ≥ 0.

Putting x = c y and a = αc α−1 , we get

³ a ´α/(α−1)
x α − ax ≥ (1 − α) ,
α
³ a ´α/(α−1)
with equality if and only if x = c = .
α
❑

1 1
314 T HEOREM (Young’s Inequality) Let p > 1 and put + = 1. Then for (x, y) ∈ ([0;+∞[)2 we have
p q
xp yq
xy ≤ + .
p q

Proof: Put α = p, a = p y in Lemma 313, obtaining

µ ¶
p y p/(p−1)
x p − (p y)x ≥ (1 − p) = (1 − p)y p/(p−1) .
p
1 p −1 p p
Now, = =⇒ q = and p − 1 = . Hence
q p p −1 q
p q
x p − (p y)x ≥ (1 − p)y p/(p−1) =⇒ (1 − p)y p/(p−1) ≥ − y ,
q
and rearranging gives the result sought. ❑

97
 Inequalities Obtained by Continuity Arguments

We now derive a generalisation of the Cauchy-Bunyakovsky-Schwarz Inequality.

1 1
315 T HEOREM (Hölder Inequality) Let x j , y k , 1 ≤ j ,k ≤ n, be real numbers. Let p > 1 and put + = 1. Then
p q
¯ ¯ à !1/p à !1/q
¯Xn ¯ Xn Xn ¯ ¯q
¯ ¯ p ¯ yk ¯
¯ x y ¯≤ |xk | .
¯k=1 k k ¯ k=1 k=1

X
n X
n ¯ ¯
Proof: If either |xk |p = 0 or ¯ y k ¯q = 0 there is nothing to prove, so assume otherwise. From Young’s
k=1 k=1
Inequality we have

|xk | |y k | |xk |p |y k |q
¡Pn ¢ ¡
p 1/p Pn
¯ ¯q ¢1/q ≤ ¡Pn p
¢ + ¡Pn ¯ ¯q ¢ .
¯y ¯ q
|xk | ¯y ¯ |x | p
k=1 k k=1 k
k=1 k=1 k

Adding, we deduce

X
n |xk | |y k | 1 X
n 1 Xn
¡Pn ¢ ¯ ¯q ¢1/q ≤ ¡Pn ¢ |xk |p + ¡Pn
¯ ¯q ¢ |y k |q
|xk | p 1/p ¡Pn ¯y ¯ |x k |p
p ¯ y ¯ q
k=1 k=1 k=1 k k=1
k=1 k=1 k Pn ¡Pn ¯ ¯q k=1¢
|x k | p ¯ yk ¯ q
= ¡Pnk=1 p ¢ + ¡Pk=1 ¯ ¯q ¢
|x | p n ¯y ¯ q
k=1 k k=1 k
1 1
= +
p q

= 1.

This gives
à !1/p à !1/q
X
n X
n X
n ¯ ¯
|xk y k | ≤ |xk | p ¯ y k ¯q .
k=1 k=1 k=1

The result follows by observing that

¯ ¯ à !1/p à !1/q
¯Xn ¯ Xn Xn X
n ¯ ¯q
¯ ¯ p ¯ yk ¯
¯ x y ¯≤ |x y | ≤ |xk | .
¯k=1 k k ¯ k=1 k k k=1 k=1

Finally, we derive a generalisation of Minkowski’s Inequality.

316 T HEOREM (Generalised Minkowski Inequality) Let p ∈]1;+∞[. Let x j , y k , 1 ≤ j ,k ≤ n, be real numbers. Then the fol-
lowing inequality holds
à !1/p à !1/p à !1/p
X
n ¯ ¯p X
n X
n ¯ ¯p
¯xk + y k ¯ ≤ |xk | p
+ ¯ yk ¯ .
k=1 k=1 k=1

Proof: From the triangle inequality for real numbers

¡ ¢
|xk + y k |p = |xk + y k ||xk + y k |p−1 ≤ |xk | + |y k | |xk + y k |p−1 .

Adding
X
n X
n X
n
|xk + y k |p ≤ |xk ||xk + y k |p−1 + |y k ||xk + y k |p−1 . (5.2)
k=1 k=1 k=1

98
 Chapter 5

By the Hölder Inequality

à !1/p à !1/q
X
n
p−1
X
n
p
X
n
(p−1)q
|xk ||xk + y k | ≤ |xk | |xk + y k |
k=1
Ãk=1 !1/p Ãk=1 !1/q (5.3)
Xn
p
Xn
p
= |xk | |xk + y k |
k=1 k=1

In the same manner we deduce

à !1/p à !1/q
X
n X
n ¯ ¯ X
n ¯ ¯
|y k ||xk + y k | p−1
≤ ¯ y k ¯p ¯ x k + y k ¯p . (5.4)
k=1 k=1 k=1

Hence (5.2) gives

à !1/p à !1/q!1/q Ãà !1/q
X
n X
n X
n ¯ ¯ X
n ¯ ¯q Xn ¯ ¯p
|xk + y k | p
≤ |xk | p ¯ xk + y k ¯p+ ¯ yk ¯ ¯ xk + y k ¯
k=1
ÃÃk=1 !1/p k=1
à !1/p ! à k=1 !1/qk=1
Xn X
n ¯ ¯ X
n ¯ ¯
= |xk |p + ¯ y k ¯p ¯ x k + y k ¯p
k=1 k=1 k=1

from where we deduce the result. ❑

Homework

Problem 5.8.1 Prove that if α > 0 and n > 0 an integer then Deduce that
1α + 2α + · · · + n α 1
n 1+α − (n − 1)1+α (n + 1)1+α − n 1+α lim = .
< nα < . n→+∞ n 1+α 1+α
1+α 1+α

5.9 Intermediate Value Property

317 T HEOREM (Intermediate Value Theorem) Let I j R and let (a,b) ∈ I 2 . Let f : I → R be a continuous function such that
f (a) ≤ f (b). Then f attains every intermediate value between f (a) and f (b), that is,
h i
∀t ∈ f (a) ; f (b) ,∃c ∈ I , such that f (c) = t .

h i
Proof: Suppose on the contrary that there is a t ∈ f (a) ; f (b) such that for all c ∈ I , f (c) 6= t. Hence f (a) < t <
f (b). Assume, without loss of generality, that a < b. Consider the sets
i h n h i o i h ³i h i h´
U = − ∞ ; a ∪ x ∈ a ;b : f (x) < t = − ∞ ; a ∪ f −1 − ∞ ; t ∩ a ;b ,

and i h n h i o i h ³i h i h´
V = b ;+∞ ∪ x ∈ a ;b : f (x) > t = b ;+∞ ∪ f −1 t ;+∞ ∩ a ;b .

Then U ,V are open sets of R by virtue of Theorem 284. But then R = U ∪ V and U ∩ V = ∅, U 6= ∅, V 6= ∅,
contradicting the fact that R is connected. Thus there must exist a c such that f (c) = t. ❑

318 C OROLLARY A continuous function defined on an interval maps that interval into an interval.

Proof: This follows at once from the Intermediate Value Theorem and the definition of an interval. ❑
h i i h
319 T HEOREM (Bolzano’s Theorem) If f : u ; v → R is continuous and f (u) f (v ) < 0, then there is a w ∈ u ; v such that
f (w ) = 0.

99
 Intermediate Value Property

Proof: This follows at once from the Intermediate Value Theorem by putting a = min( f (u), f (v )) < 0 and
b = max( f (u), f (v )) > 0 . ❑

320 C OROLLARY Every polynomial p(x) ∈ R[x] with real coefficients and odd degree has at least one real root.

Proof: Let p(x) = a 0 + a 1 x + a 2 x 2 + · · · + a n x n , with a n 6= 0 and n odd. Since p has odd degree, lim p(x) =
x→−∞
(−∞) signum(a n ) and lim p(x) = (+∞) signum(a n ), which are of opposite sign. The polynomial must then
x→+∞
attain positive and negative values and between values of opposite sign, it will have a real root. ❑

321 C OROLLARY If f is continuous at the point a and f (a) 6= 0, then there is a neighbourhood of a where f (x) has the same
sign as f (a).

¯ ¯
¯ f (a)¯
Proof: Take ε = > 0 in the definition of continuity. There is a δ > 0 such that
2
¯ ¯ ¯ ¯ ¯ ¯
¯ ¯ ¯ f (a)¯ ¯ f (a)¯ ¯ f (a)¯
|x − a| < δ =⇒ ¯ f (x) − f (a)¯ < =⇒ f (a) − < f (x) < f (a) + ,
2 2 2
from where the result follows. ❑

322 T HEOREM A continuous function defined on a compact set maps that compact set into a compact set.

© ª+∞
Proof: Let f : X → R be continuous and X j R compact. Let y n n=1 j f (X ) be an infinite sequence of f (X ).
There are xn ∈ X such that xn = f (y n ). Since {xn }+∞
n=1 j X is an infinite sequence of X and X is compact, it has a
© ª+∞
convergent subsequence in X , say, xnk k=1 with xnk → x ∈ X , by virtue of Theorem 143. Since f is continuous

xnk → x =⇒ f (xnk ) → f (x).

© ª+∞ © ª+∞
Clearly f (x) ∈ f (X ). Thus the arbitrary sequence y n n=1 j f (X ) has the convergent subsequence y nk k=1 in
f (X ), and one more appeal to Theorem 143 proves compactness. ❑
h i h i
323 T HEOREM (Weierstrass Theorem) A continuous function f : a ;b → R attains a maximum and a minimum on a ;b .

h i
Proof: By Theorem 322, f ( a ;b ) is compact, and so, by the Heine-Borel Theorem, it is closed and bounded.
Thus there exists (m, M) ∈ R2 such that m = hinf i f (x) and M = sup
h i f (x). We must prove that these are
x∈ a ;b x∈ a ;b
h i h i h i
attained in a ;b , i.e., that there exist µ ∈ a ;b and µ0 ∈ a ;b such that f (µ) = m and f (µ0 ) = M. By
h i
the Approximation Property of the Infimum and the Supremum, we may find sequences {mn }+∞n=1 j a ;b , and
h i
{Mn }+∞
n=1 j a ;b such that m ≤ m n and m n → m, and also, Mn ≤ M, and Mn → M as n → +∞. By the In-
h i h i
termediate Value Theorem, there exist µn ∈ a ;b and µ0n ∈ a ;b such that mn = f (µn ) and Mn = f (µ0n ). By
h i © ª+∞ h i © ª+∞ h i
the compactness of a ;b the sequences µn n=1 j a ;b and µ0n n=1 j a ;b have convergent subsequences
© ª+∞ h i n o+∞ h i h i h i
µnk k=1 j a ;b and µ0nk j a ;b such that µnk → µ ∈ a ;b and µ0nk → µ0 ∈ a ;b . By continuity and
k=1
the uniqueness of limits,

µnk → µ =⇒ mnk = f (µnk ) → m = f (µ), and µ0nk → µ0 =⇒ Mnk = f (µ0nk ) → M = f (µ0 ),

h i
and so f attains both extrema in a ;b .❑

h i h i
324 T HEOREM (Fixed Point Theorem) Let f : a ;b → a ;b be continuous. Then f has a fixed point, that is, there is
h i
c ∈ a ;b such that f (c) = c.

100
 Chapter 5

Proof: If either f (a) = a or f (b) = b we are done. Assume then that f (a) > a and f (b)i< b. hPut g (x) = f (x) − x.
Then g is continuous, g (a) > 0 and g (b) < 0. By Bolzano’s Theorem, there must be a c ∈ a ;b such that g (c) = 0,
that is, f (c) − c = 0, finishing the proof. ❑

Homework

Problem 5.9.1 Let p(x), q(x) be polynomials with real coefficients Problem 5.9.7 Let I be a closed and bounded interval on the line
such that and let f be continuous on I . Suppose that for each x ∈ I , there ex-
p(x 2 + x + 1) = p(x)q(x). ists a y ∈ I such that
1
Prove that p must have even degree. | f (y )| ≤ | f (x)|.
2
Prove the existence of a t ∈ I such that f (t ) = 0.
Problem 5.9.2 A function f defined over all real numbers is contin-
uous and for all real x satisfies
¡ ¢ ¡ ¢ Problem 5.9.8 Find all continuous functions that satisfy the func-
f (x) · ( f ◦ f )(x) = 1. tional equation µ ¶
x+y
Given that f (1000) = 999, find f (500). f (x) + f (y ) = f ,
1−xy
for all −1 < x, y < 1.
Problem 5.9.3 Let f : R → R be a continuous function such that
lim f (x) = 0 = lim f (x). If f is strictly negative somewhere on
x→−∞ x→+∞ Problem 5.9.9 (Putnam 1947) A real valued continuous function
R then f attains a finite absolute minimum on R. If f is strictly pos-
satisfies for all real x, y the functional equation
itive somewhere on R then f attains a finite absolute maximum on
R. q
f ( x 2 + y 2 ) = f (x) f (y ).
h i h i
Problem 5.9.4 Let f : 0 ;1 → 0 ;1 be continuous. Prove that 2
h i Prove that f (x) = ( f (x))x .
there is no c ∈ 0 ;1 such that f −1 ({c}) has exactly two elements.
h i h i
Problem 5.9.10 Suppose that f : 0 ;1 → 0 ;1 is continuous.
h i h i
Problem 5.9.5 Let f , g be continuous functions from 0 ;1 to Prove that there is a number c in 0 ;1 such that f (c) = 1 − c.
h i
0 ;1 such that
h i Problem 5.9.11 (Universal
h Chord
i Theorem) Suppose that f is a
∀x ∈ 0 ;1 f (g (x)) = g ( f (x)). continuous function of 0 ;1 and that f (0) = f (1). Let n be a
h i
h i
strictly positive integer. Prove that there is some number x ∈ 0 ;1
Prove that f and g have a common fixed point in 0 ;1 .
such that f (x) = f (x + 1/n).

Problem 5.9.6 A continuous function f : R → R satisfies

∀x ∈ R f (x + f (x)) = f (x). Problem 5.9.12 Under the same conditions of problems 5.9.11
prove that there are no universal chords of length a,0 < a < 1, a 6=
Prove that f is constant. 1/n.

5.10 Variation of a Function and Uniform Continuity

h i
325 Definition A partition P of the interval a ;b is any finite set of points x0 , x1 ,... , xn such that

a = x0 < x1 < · · · < xn = b.

h i
A partition P 0 of a ;b is said to be finer than the partition P if P j P 0 .

326 Definition The mesh or norm of P is ¯¯ ¯¯

¯¯ ¯¯
¯¯P ¯¯ = max |xk − xk−1 | .
1≤k≤n

☞ ¯¯ ¯¯ ¯¯ ¯¯
¯¯ ¯¯ ¯¯ ¯¯
If P j P 0 then clearly ¯¯P 0 ¯¯ ≤ ¯¯P ¯¯, since the finer partition has probably more points which will make the corre-
sponding subintervals smaller.

101
 Variation of a Function and Uniform Continuity

h i h i
327 Definition Let f be a bounded function on an interval a ;b and let I j a ;b be a subinterval. The oscillation of f
on I is defined and denoted by
ω( f , I ) = sup f (x) − inf f (x).
x∈I x∈I

h i h i
328 T HEOREM Let f : a ;b → R be a continuous function. Given ε > 0 there exists a partition of a ;b into a finite number
of subintervals of equal length such that the oscillation of f on each of these subintervals is at most ε.

h i
Proof: Let P ε mean the following: there is an ε > 0 such that for all partitions of a ;b into a finite number
h i
of intervals of equal length, the oscillation of f is ≥ ε. By bisecting a ;b , at least one of the halves must have
h i h i h i
property P ε , say a 1 ;b1 . If a ;b we to have property P ε , then by bisecting a 1 ;b1 , at least one of the halves
h i
must have property P ε , say a 2 ;b2 . Continuing in this way we have constructed a sequence of imbricated
intervals h i h i h i h i
a ;b k a 1 ;b1 k a 2 ;b2 k · · · k a n ;bn k · · ·
h i b−a
where the length of a n ;bn is bn − a n = n → 0 as n → +∞. By the Cantor Intersection Theorem, there is a
2 h
∞ h
\ i i h i
point c ∈ a n ;bn . Moreover, we have ω( f , a n ;bn ) ≥ ε. Since c ∈ a ;b , f is continuous at c. Hence there
n=1
is a δ > 0 such that i h ¯ ¯ ε
x ∈ c − δ ;c + δ =⇒ ¯ f (x) − f (c)¯ <
2
i h2
. Taking (x 0 , x 00 ) ∈ c − δ ;c + δ we have
¯ ¯ ¯ ¯ ¯ ¯
¯ f (x 0 ) − f (x 00 )¯ ≤ ¯ f (x 0 ) − f (c)¯ + ¯ f (c) − f (x 00 )¯ < ε,

whence h i i h
ω( f , a ;b ∩ c − δ ;c + δ ) < ε.
h i
Now, if there was an ε > 0 such that for all partitions of a ;b into a finite number of intervals of equal length,
h i
the oscillation of f is ≥ ε, then by taking n large enough above we could find one of the a n ;bn completely
inside one of the subintervals of the partition. By the above, the oscillation there would be < ε, a contradiction. ❑
h i
329 T HEOREM Let f : a ;b → R be a continuous function. Given ε > 0 there exists a δ > 0 such that on any subinterval
h i
I j a ;b having length < δ the oscillation of f on I is < ε.

b−a
Proof: Let δ = . By Theorem 328 we may choose n so large that the oscillation of f on each of
n
h i h i h i
a ;a +δ , a + δ ; a + 2δ , ... , a + (n − 1)δ ;b , (5.5)

ε h i
is < . Let I j a ;b be any subinterval of length < δ and let x 0 ∈ I be the point where f achieves its largest value
2
and x 00 ∈ I be the point where f achieves its smallest value. Then x 0 and x 00 either belong to the same interval in
¯ ¯ ε
5.5—in which case ¯ f (x 0 ) − f (x 00 )¯ < —or since I has length smaller than δ, to two consecutive subintervals
2
h i h i
a + ( j − 1)δ ; a + j δ , a + j δ ; a + ( j + 1)δ .

In this case
ε ε
f (x 0 ) − f (x 00 ) = ( f (x 0 ) − f (a + j δ)) + ( f (a + j δ) − f (x 00 )) < + = ε.
2 2
The theorem now follows.❑

102
 Chapter 5

h i
330 Definition A function f is said to be uniformly continuous on a ;b if ∀ε > 0 there exists δ > 0 depending only on ε
h i2
such that for any (u, v ) ∈ a ;b ,
¯ ¯
|u − v | < δ =⇒ ¯ f (u) − f (v )¯ < ε.

h i
331 T HEOREM If f : a ;b → R is continuous, then f is uniformly continuous.

Proof: This follows from Theorem 329. ❑

332 T HEOREM (Heine’s Theorem) If f : X → R is continuous and X is compact, then f is uniformly continuous.

Proof: This follows from Theorem 331. ❑

i h
333 T HEOREM Let f be an increasing function on an open interval a ;b . Then, for any x satisfying a < x < b,

i h f (t) = f (x−) ≤ f (x) ≤

sup iinf h f (t) = f (x+).
t ∈ a ;x t ∈ x ;b

Moreover, if a < x < y < b, then f (x+) ≤ f (y−).

Proof: The set { f (t) : a < u < x} is bounded above by f (x) and hence it has a supremum sup
i h f (t) = A and
t ∈ a ;x

clearly A ≤ f (x) as f is increasing. Let us shew that A = f (x−). By the Approximation Property of the Supremum,
there is δ > 0 such that a < x − δ < x and A − ε < f (x − δ) ≤ A. But as f is increasing,
¯ ¯
x − δ < t < x =⇒ f (x − δ) ≤ f (t) < A =⇒ ¯ f (x) − A ¯ ,

whence f (x−) = A.

A similar reasoning gives iinf h f (t) = f (x+).

t ∈ x ;b

Now, if a < x < y < b, then by what has already been proved we obtain

f (x+) = inf f (t) = inf f (t),

x<t <b x<t <y

again, remembering that f is increasing. Similarly,

f (y−) = sup f (t) = inf f (t),

a<t <y x<t <y

from where f (x+) ≤ f (y−). ❑

h i
334 T HEOREM Let f be an increasing function defined on the interval a ;b and let

a = x0 < x1 < x2 < · · · < xn = b

be n + 1 points partitioning the interval. Then

X¡
n−1 ¢
f (xk +) − f (xk −) ≤ f (b) − f (a).
k=1

103
 Variation of a Function and Uniform Continuity

i h
Proof: Let y k ∈ xk ; xk+1 . For 1 ≤ k ≤ n − 1, by Theorem ??,

f (xk +) ≤ f (y k ) and f (y k−1 ) ≤ f (xk −) =⇒ f (xk +) − f (xk −) ≤ f (y k ) − f (y k−1 ).

Adding,
X¡
n−1 X¡
¢ n−1 ¢
f (xk +) − f (xk −) ≤ f (y k ) − f (y k−1 ) = f (y n−1 ) − f (y 0 ).
k=1 k=1
The proof is completed upon noticing that f (y n−1 ) − f (y 0 ) ≤ f (b) − f (a). ❑
h i
335 T HEOREM Let f : a ;b → R be a monotone function, Then the set of points of discontinuity of f is either finite or
countable.

Proof: Assume f is increasing, for if f were decreasing, we may apply the same argument to − f . Let m > 0 be
an integer, and let ½ ¾
i h 1
Sm = x ∈ a ;b : f (x+) − f (x−) ≥ .
m
If x1 < x2 < · · · < xn are in Sm then by Theorem 334,
n
≤ f (b) − f (a),
m
h i [
∞
which implies that Sm is a finite set. The set of discontinuities of f in a ;b is Sm , the countable union of
m=1
finite sets, and hence it is countable. ❑
h i
336 Definition Let f be a function defined on the interval a ;b and let

a = x0 < x1 < x2 < · · · < xn = b

be n + 1 points partitioning the interval. If there exists V > 0 such that
X
n ¯ ¯
¯ f (xk ) − f (xk−1 )¯ ≤ V
k=1
h i h i
for all partitions of a ;b , the we say that f is of bounded variation on a ;b .

h i h i
337 T HEOREM If f is monotonic on a ;b , then f is bounded variation on a ;b .

Proof: Let
a = x0 < x1 < x2 < · · · < xn = b
h i
be any partition of a ;b . Then

X
n ¯ ¯
¯ f (xk ) − f (xk−1 )¯ = max( f (b) − f (a), f (a) − f (b)),
n=1
¯ ¯
the first choice occurring when f is increasing and the second when f is decreasing. Then V = ¯ f (b) − f (a)¯
satisfies the definition of bounded variation for an arbitrary partition. ❑
h i h i
338 T HEOREM If f is of bounded variation on a ;b then f is bounded on a ;b .

i h h i
Proof: Let x ∈ a ;b and consider the partition a < x < b of a ;b . Since f is of bounded variation there is a
V > 0 such that ¯ ¯ ¯ ¯
¯ f (a) − f (x)¯ + ¯ f (x) − f (b)¯ ≤ V .
But then ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
¯ f (x)¯ ≤ ¯ f (x) − f (a)¯ + ¯ f (a)¯ ≤ V + ¯ f (a)¯ .
¯ ¯
and so f is bounded by the constant quantity V + ¯ f (a)¯. ❑

104
 Chapter 5

Homework
i h i h
Problem 5.10.1 Shew that 0 ;+∞ → 0 ;+∞ , x 7→ x 2 , is not uni- formly continuous.

5.11 Classical Limits

339 L EMMA If 0 < x ≤ 1 then
exp(x) − 1
1≤ ≤ 1 + x(e − 2).
x
1
If − ≤ x < 0 then
2
exp(x) − 1 x
1+x ≤ ≤ 1+ .
x 4

Proof:
³ x ´n
Since 1 + ≤ exp(x) for n > −x by Proposition 301, we have 1 + x ≤ exp(x) for all x > −1. Now, for n ≥ 2 and
n
0 < x ≤ 1,
à ! à ! à ! à !
³ x ´n n x n x2 n x3 n xn
1+ = 1+ + 2
+ 3
+··· +
n 1 n 2 n 3 n n nn
µ µ ¶µ ¶ µ ¶µ ¶µ ¶ µ ¶µ ¶µ ¶ µ ¶ ¶
1 1 1 1 1 1 2 1 1 1 2 n − 1 n−2
= 1 + x + x2 1− + 1− 1− x +··· + 1− 1− ··· 1 − x
2! n n 3! n n n n! n n n n
µ ¶
1 1 1
≤ 1 + x + x2 + +··· +
2! 3! n!

< 1 + x + x 2 (e − 2),

upon using Theorem 180. This proves the first set of inequalities.

x2 ³ x ´2 1
For x > −2, 1 + x + = 1+ ≤ exp(x) by Proposition 301. Now we assume that − ≤ x ≤ 0. As before,
4 2 2

³ µ µ ¶µ ¶ µ ¶µ ¶µ ¶ µ ¶µ ¶µ ¶ µ ¶ ¶
x ´n 1 1 1 1 1 1 2 1 1 1 2 n − 1 n−2
1+ = 1 + x + x2 1− + 1− 1− x +··· + 1− 1− ··· 1 − x .
n 2! n n 3! n n n n! n n n n

1
Since x k ≤ 0 for odd k and x k ≤ for even k we may delete the odd terms from the dextral side and so
2k
³ µ
µ ¶µ ¶ µ ¶µ ¶µ ¶ µ ¶ ¶
x ´n 1 1 1 1 1 1 2 2k − 1 2k
1+ ≤ 1 + x + x2 1− +0 +··· + 1− 1− ··· 1 − x +···
n 2! n n n! n n n n
µ ¶
2 1 1
≤ 1+x +x + +···
2 22
≤ 1 + x + x2.

1
On taking limits exp(x) ≤ 1 + x + x 2 for − ≤ x ≤ 0. Thus we have
2

1 x2 exp(x) − 1 x
− ≤ x < 0 =⇒ 1 + x + ≤ exp(x) ≤ 1 + x + x 2 =⇒ 1 + x ≤ ≤ 1+ ,
2 4 x 4
since division by negative x reverses the sense of the inequalities. ❑

exp(x) − 1
340 T HEOREM lim = 1.
x→0 x

105
 Classical Limits

exp(x) − 1 exp(x) − 1
Proof: We prove that lim = 1 and that lim = 1. Let us start with the first assertion. For
x→0+ x x→0− x
0 < x ≤ 1 we have, by the Sandwich Theorem, and Lemma 339,

exp(x) − 1 exp(x) − 1
1≤ ≤ 1 + x(e − 2) =⇒ lim = 1,
x x→0+ x

proving the first assertion.

1
For − ≤ x ≤ 0 we have, by the Sandwich Theorem and Lemma 339,
2

x2 x exp(x) − 1 exp(x) − 1
1+x + ≤ exp(x) ≤ 1 + x + x 2 =⇒ 1 + ≤ ≤ 1 + x =⇒ lim = 1,
4 4 x x→0− x

proving the second assertion. ❑

341 L EMMA For 0 < x ≤ 1,

x(e − 2) log(1 + x)
1− ≤ ≤1
1+x x
1
and for − ≤ x ≤ 0,
2
log(1 + x) x
1≤ ≤ 1− .
x 1+x

Proof: Since x 7→ log(1 + x) is strictly increasing, we have by Lemma 339 for 0 < x ≤ 1,

1 + x ≤ exp(x) ≤ 1 + x + x 2 (e − 2) =⇒ log(1 + x) ≤ x ≤ log(1 + x + x 2 (e − 2)).

Notice that we have established that log(1 + x) ≤ x for 0 < x ≤ 1. Now

µ ¶ µ ¶
2 x 2 (e − 2) x 2 (e − 2)
log(1 + x + x (e − 2)) = log(1 + x) 1 + = log(1 + x) + 1 + .
1+x 1+x

x2 x 2 (e − 2) e − 2
Since for x > 0, x 7→ is strictly increasing, < < 1 for 0 < x < 1. Thus we may use log(1+y) ≤ y,
1+x 1+x 2
2
x (e − 2)
0 ≤ y ≤ 1 with y = obtaining
1+x
µ ¶
x 2 (e − 2) x 2 (e − 2)
log 1 + ≤ .
1+x 1+x

Hence
x 2 (e − 2)
x ≤ log(1 + x + x 2 (e − 2)) ≤ log(1 + x) + .
1+x
In conclusion,

x 2 (e − 2) x(e − 2) log(1 + x)
0 < x ≤ 1 =⇒ log(1 + x) ≤ x ≤ log(1 + x) + =⇒ 1 − ≤ ≤ 1.
1+x 1+x x
1
Similarly, for − ≤ x < 0, by Lemma 339,
2
µ ¶
x2 2 x2
1+x + ≤ exp(x) ≤ 1 + x + x =⇒ log 1 + x + ≤ x ≤ log(1 + x + x 2 ).
4 4

Since x 7→ log(1 + x) is increasing, plainly

µ ¶
x2
log(1 + x) ≤ log 1 + x + ≤ x.
4

106
 Chapter 5

1 x2 1
Now observe that − ≤ x < 0 =⇒ 0 < ≤ < 1 and so
2 1+x 2
µ ¶
x2 x2 x2
log(1 + x + x 2 ) = log(1 + x) + log 1 + ≤ log(1 + x) + =⇒ x ≤ log(1 + x) + .
1+x 1+x 1+x

In conclusion,

1 x2 log(1 + x) x
− ≤ x < 0 =⇒ log(1 + x) ≤ x ≤ log(1 + x) + =⇒ 1 ≤ ≤ 1− ,
2 1+x x 1+x

since division by negative x reverses the sense of the inequalities. ❑

log(1 + x) − x
342 T HEOREM lim = 0.
x→0 x

Proof: By Lemma 341, for 0 < x ≤ 1,

x(e − 2) log(1 + x) log(1 + x)

1− ≤ ≤ 1 =⇒ lim = 1,
1+x x x→0+ x

by the Sandwich Theorem. Again, by Lemma 341 and the Sandwich Theorem,

1 log(1 + x) x log(1 + x)
− ≤ x ≤ 0 =⇒ 1 ≤ ≤ 1− =⇒ lim = 1.
2 x 1+x x→0− x

Combining both results, the theorem follows.❑

B
b

C
b

b θ b
A
O

Figure 5.1: Theorem 344.

(1 + x)a − 1
343 T HEOREM If a ∈ R, then lim = a.
x→0 x

Proof: This is evident for a = 0. Assume now a 6= 0. Since x 7→ exp(x) is continuous and since a log(1 + x) → 0 as
x → 0, by Theorems 340 and 342,

(1 + x)a − 1 exp(a log(1 + x)) − 1 log(1 + x)

lim = a lim · lim = a · 1 · 1 = a.
x→0 x x→0 a log(1 + x) x→0 x

sin θ
344 T HEOREM lim = 1.
θ→0 θ

107
 Classical Limits

sin θ sin θ sin θ

Proof: We first prove that lim = 1. Since θ 7→ is an even function it will also follow that lim =
θ→0+ θ θ θ→0− θ
1.

θ
Assume 0 < θ < and consider 4O AB right-angled at A, with O A = 1 and ∠BO A = θ. C is the point where line
2
OB meets the unit circle with centre at O and D is its perpendicular projection. The area of 4O AC is smaller
than the area of the circular sector O AC , which is smaller than the area of 4O AB . Hence

1 θ 1 1 sin θ sin θ
sin θ < < tanθ =⇒ < < 1 =⇒ lim =1
2 2 2 cos θ θ θ→0+ θ

by the Sandwich Theorem, proving the theorem.❑

108
Chapter 6

Differentiable Functions

6.1 Derivative at a Point

345 Definition Let I be an interval, a ∈ I˚, and f : I → R. We say that f is differentiable at a if the limit
f (x) − f (a) f (a + h) − f (a)
lim = lim
x→a x −a h→0 h
df
exists and is finite. In such a case we denote this limit by f 0 (a), D f (a), or (a) and we call this quantity the derivative of
dx
f at a.

346 Definition Let I be an interval, a ∈ I˚, and f : I → R. If

f (x) − f (a) f (a + h) − f (a)
lim = lim
x→a+ x −a h→0+ h
exists and is finite we say that f is differentiable at a on the right and write f +0 (a) for this limit. If
f (x) − f (a) f (a + h) − f (a)
lim = lim
x→a− x −a h→0− h
exists and is finite we say that f is differentiable at a on the left and write f −0 (a) for this limit.

347 T HEOREM Let I be an interval, a ∈ I˚, and f : I → R. Then f is differentiable at a if and only if both f + (a) and f − (a) exist
and are equal. In this case f + (a) = f 0 (a) = f − (a).

Proof: Obvious. ❑

348 T HEOREM Let I be an interval, a ∈ I˚, and f : I → R. If f is differentiable at a then it is continuous at a.

Proof: We have
µ ¶ µ ¶µ ¶
f (a + h) − f (a) f (a + h) − f (a)
lim f (a + h) − f (a) = lim h = lim lim h = f 0 (a) · 0 = 0.
h→0 h→0 h h→0 h h→0

Thus lim f (a + h) − f (a) = 0 =⇒ lim f (a + h) = f (a) and so f is continuous. ❑

h→0 h→0

349 T HEOREM Let I j R be an interval. If f : I → R is identically constant, then f 0 (I ) = 0.

f (x) − f (c) K −K
Proof: Assume that f (I ) = K , a constant. Let c ∈ I˚. Then f 0 (c) = lim = lim = 0. If c is an
x −c
x→c x→c x −c
endpoint of I , then the argument is modified to be either the left or right derivative. ❑

Homework

109
 Differentiation Rules

Problem 6.1.1 Let f : R → R, Problem 6.1.2 Let f : R → R, x 7→ |x|. Prove that f is not differen-
 tiable at x = 0 and that for x 6= 0, f 0 (x) = signum(x).


 x +1 if x ∈ Q
f (x)


 2−x if x ∈ R \ Q
Problem 6.1.3 Let f : R → R, x 7→ x |x|. Determine whether f 0 (0)
Prove that f is nowhere differentiable. exists.

6.2 Differentiation Rules

350 T HEOREM Let I be an interval, a ∈ I˚, λ ∈ R a constant, and f , then, g : I → R. If f and g are differentiable at a then

1. (Linearity Rule) f + λg is differentiable at a and ( f + λg )0 (a) = f 0 (a) + λg 0 (a)

2. (Product Rule) f g is differentiable at a and ( f g )0 (a) = f 0 (a)g (a) + f (a)g 0 (a)

µ ¶0
1 1 g 0 (a)
3. if g (a) 6= 0, is differentiable at a and (a) = −
g g (g (a))2
µ ¶0
f f f 0 (a)g (a) − f (a)g 0 (a)
4. (Quotient Rule) if g (a) 6= 0, is differentiable at a and (a) =
g g (g (a))2

Proof:

1. This follows by the linearity of limits.

2. We have
( f g )(a + h) − ( f g )(a)
( f g )0 (a) = lim
h→0 h
g (a + h)( f (a + h) − f (a)) + f (a)(g (a + h) − g (a))
= lim
h→0 h
( f (a + h) − f (a)) (g (a + h) − g (a))
= lim g (a + h) lim + lim f (a) lim
h→0 h→0 h h→0 h→0 h
= g (a) f 0 (a) + f (a)g 0 (a),

as desired.
3. We have
1 1
µ ¶0 −
1 g (a + h) g (a)
(a) = lim
g h→0 h
g (a) − g (a + h)
g (a + h)g (a)
= lim
h→0 h
g (a) − g (a + h) 1
= lim lim
h→0 h h→0 g (a + h)g (a)
µ ¶
¡ 0 ¢ 1
= −g (a)
g (a)g (a)
g 0 (a)
= − ,
g (a)2
as desired.
4. Using (2) and (3),
µ ¶0 µ ¶ µ ¶0
f 1 1
(a) = f 0 (a) (a) + f (a) (a)
g g g
f 0 (a) f (a)g 0 (a)
= −
g (a) g (a)2
f (a)g (a) − f (a)g 0 (a)
0
= ,
(g (a))2

110
 Chapter 6

as desired.

351 T HEOREM (Chain Rule) Let I , J be intervals of R, with a ∈ I . Let f : I → R and g : J → R be such that f (I ) j J . If f is
differentiable at a and g is differentiable at f (a), then g ◦ f is differentiable at a and (g ◦ f )0 = g 0 ( f (a)) f 0 (a).

Proof: Put b = f (a), and 


 g (y) − g (b)

 if y 6= b
ϕ(y) = y −b

 0

 g (b) if y = b

Since g is differentiable at b, ϕ is continuous at y = b. Now, for x 6= a,

g ( f (x)) − g ( f (a)) f (x) − f (a)

= ϕ( f (x)) .
x −a x −a
(If f (x) 6= f (a) this follows directly from the definition of ϕ. If f (x) = f (a), both sides of the equality are 0.)
By the continuity of f at a and of ϕ at b,

lim ϕ( f (x)) = ϕ( f (a)) = g 0 ( f (a)),

x→a

whence
g ( f (x)) − g ( f (a))
(g ◦ f )0 (a) = lim
x→a x −a
f (x) − f (a)
= lim ϕ( f (x))
x→a x−a
= g 0 ( f (a)) f 0 (a),

as desired.
❑

352 T HEOREM (Inverse Function Rule) Let I be an interval of R, with a ∈ I . Let f : I → R be strictly monotonic and contin-
uous over I . If f is differentiable at a and f 0 (a) 6= 0, then the inverse f −1 : f (I ) → R is differentiable at f (a) and

1
( f −1 )0 ( f (a)) = .
f 0 (a)

Proof: Put b = f (a). Observe that lim f −1 (y) = a, and by the composition rule for limits,
y→b

f −1 (y) − f 1 (b) f −1 (y) − a 1

lim = lim −1
= 0 ,
y→b y −a y→b f ( f (y)) − a f (a)

proving the theorem. ❑

☞ Once it is known that ( f −1 )0 exists, we may proceed as follows. Since f −1 ( f (x)) = x, differentiating on both sides, using
the Chain Rule on the sinistral side,
( f −1 )0 ( f (x)) f 0 (x) = 1,
from where the result follows.

353 Definition Let I be an interval of R. Let f : I → R be differentiable at every point of I . The function f 0 : I → R, x 7→ f 0 (x)
is called the derivative function or derivative of the function f .

354 T HEOREM Let n ≥ 0 be an integer. Let f : R → R, x 7→ x n . Then f is everywhere differentiable and f 0 : R → R is given by
x 7→ nx n−1 .

111
 Differentiation Rules

Proof: Assume first n is strictly positive. By Theorem 55,

xn − an (x − a)(x n−1 + ax n−2 + a 2 x n−3 + · · · + a n−2 x + a n−1 )

lim = lim
x→a x −a x→a x −a
= lim (x n−1 + ax n−2 + a 2 x n−3 + · · · + a n−2 x + a n−1 )
x→a

= na n−1 .

Observe that this is true for all a ∈ R.

If n = 0 then f is constant, say f (x) = K for all x and so

f (x) − f (a) K −K
lim = lim = 0.
x→a x−a x→a x − a

❑
i h i h 1 i h
355 T HEOREM Let n > 0 be an integer and f : 0 ;+∞ → 0 ;+∞ , x 7→ n
. Then f 0 exists everywhere in 0 ;+∞ and
i h i h x
n
f 0 : 0 ;+∞ → 0 ;+∞ is given by f 0 (x) = − n+1 .
x

Proof: We use the result above, part (3) of Theorem 350, and the Chain Rule, to get

d 1 nx n−1 n
n
= − n 2 = − n+1 ,
dx x (x ) x

and the theorem follows.❑

i h i h i h
356 L EMMA Let q ∈ Z, q > 0 be an integer, and f : 0 ;+∞ → 0 ;+∞ , x 7→ x 1/q . Then f 0 exists everywhere in 0 ;+∞ and
i h i h x 1/q −1
f 0 : 0 ;+∞ → 0 ;+∞ is given by f 0 (x) = .
q

Proof: We have ( f (x))q = x. Using the Chain Rule q f 0 (x)( f (x))q −1 = 1. Since f (x) 6= 0,

1 1 1
f 0 (x) = = = x 1/q −1 .
q( f (x))q −1 q(x 1/q )q −1 q

❑
i h i h i h i h
357 T HEOREM Let r ∈ Q and let f : 0 ;+∞ → 0 ;+∞ , x 7→ x r . Then f 0 exists everywhere in 0 ;+∞ and f 0 : 0 ;+∞ →
i h
0 ;+∞ is given by f 0 (x) = r x r −1 .

a
Proof: Let r = , where a,b are integers, with b > 0. We use the Chain Rule, Lemma 356, and Theorem 355.
b
Then
d a/b d 1/b a 1 a
x = (x ) = a(x 1/b )a−1 · x 1/b−1 = x a/b−1 = r x r −1 ,
dx dx b b
proving the theorem.
❑

358 T HEOREM (Derivative of the Exponential Function) Let exp : R → R, x 7→ e x . Then exp is everywhere differentiable and
exp0 : R → R is given by x 7→ e x .

112
 Chapter 6

Proof: Using Theorem 340, we have, with h = x − a,

ex − ea e x−a − 1
lim = e a lim
x→a x − a x→a x − a
eh − 1
= e a lim
h→0 h
= ea · 1

= ea.

i h i h
359 T HEOREM (Derivative of the Logarithmic Function) Let f : 0 ;+∞ → −∞ ;+∞ , x 7→ log x. Then f 0 exists everywhere
i h i h 1
in 0 ;+∞ and f 0 : 0 ;+∞ → R \ {0} is given by f 0 (x) = .
x

x
Proof: Let a > 0. Then, with h = − 1, and using Theorem 342,
a

x
log x − log a log
lim = lim a
x→a x−a x→a x − a ³ ´
x
1 log 1 + − 1
a
= · lim x
a x→a −1
a
1 log(1 + h)
= · lim
a h→0 h
1
= ·1
a
1
= .
a

i h i h i h
360 T HEOREM (Power Rule) Let t ∈ R and let f : 0 ;+∞ → 0 ;+∞ , x 7→ x t . Then f 0 exists everywhere in 0 ;+∞ and
i h i h
f 0 : 0 ;+∞ → 0 ;+∞ is given by f 0 (x) = t x t −1 .

Proof: Using the Chain Rule,

d t d ¡ ¢ t ¡ ¢ t
x = exp(t log x) = · exp(t log x) = · x t = t x t −1 .
dx dx x x

361 T HEOREM (Derivative of sin) . Let sin : R → R, x 7→ sin x. Then sin is everywhere differentiable and sin0 : R → R is given
by x 7→ cos x.

113
 Differentiation Rules

Proof: We make a change of variables, and use Theorem 344,

sin x − sin a sin(x − a + a) − sin a

lim = lim
x→a x −a x→a x −a
sin(x − a) cos a + cos(x − a) sin a − sin a
= lim
x→a x −a
sin(x − a) cos(x − a) − 1
= (cos a) lim + (sin a) lim
x→a x −a x→a x −a
sin h cos h − 1
= (cos a) lim + (sin a) lim
h→0 h h→0 h
cos2 h − 1
= (cos a) · 1 + (sin a) lim
h→0 h(cos h + 1)
− sin2 h
= (cos a) · 1 + (sin a) lim
h→0 h(cos h + 1)
sin h − sin h
= (cos a) + (sin a) lim · lim
h→0 h h→0 cosh + 1

= cos a,

and the theorem follows. ❑

362 T HEOREM (Derivatives of the Goniometric Functions)

d
1. sin x = cos x x ∈R
dx
d
2. cos x = sin x x ∈R
dx
d π
3. tan x = sec2 x x ∈ R \ (2Z + 1)
dx 2
d π
4. sec x = sec x tan x x ∈ R \ (2Z + 1)
dx 2
d
5. csc x = − csc x cot x x ∈ R \ Zπ
dx
d
6. cot x = − csc2 x x ∈ R \ Zπ
dx

Proof: (1) is Theorem 361. To prove (2), observe that

d d ³π ´ ³π ´
cos x = sin − x = − cos − x = − sin x.
dx dx 2 2
To prove (3), we use the Quotient Rule,
d d sin x (cos x)(cos x) − (− sin x)(sin x) 1
tan x = = = = sec2 x.
dx dx cos x cos2 x cos2 x
To prove (4), we use once again the Quotient Rule,

d d 1 (0)(cos x) − (− sin x)(1) sin x

sec x = = = = sec x tan x.
dx dx cos x cos2 x cos2 x
To prove (5), observe that
d d ³π ´ ³π ´ ³π ´
csc x = sec − x = − sec − x tan − x = − csc x cot x.
dx dx 2 2 2
To prove (6), observe that
d d ³π ´ ³π ´
cot x = tan − x = − sec2 − x = − csc2 x.
dx dx 2 2
❑

114
 Chapter 6

363 Definition (Higher Order Derivatives) Let I be an interval of R and let f : I → R. For a ∈ I we define the successive
derivatives of f at a, inductively. Put f (a) = f (0) (a). If n ≥ 1,

f (n) (a) = f 0 ( f (n−1) (a)),

provided f is differentiable at f (n−1) (a).

☞ We usually write f 00 instead of f (2) .

364 T HEOREM (Leibniz’s Rule) Let n be a positive integer.

à !
(n)
Xn n
(f g ) = f (k) g (n−k)
k=0 k

Proof: This is a generalisation of the Productà Rule.

! The proof is by induction on n. For n = 0 and n = 1 the
Xn n
assertion is obvious. Assume that ( f g )(n) = f (k) g (n−k) . Observe that
k=0 k

( f g )(n+1) = (( f g )(n) )0
à à ! !0
Xn n
(k) (n−k)
= f g
k
k=0Ã !
X
n n
= ( f (k+1) g (n−k) + f (k) g (n−k+1) )
k=0 Ãk ! Ã !
X
n n
(k+1) (n−k)
Xn n
= f g + f (k) g (n−k+1)
k=0 k ÃÃ ! Ã k=0 k
!!
(0) (n+1)
Xn n n
= f g + + f (k) g (n+1−k) + f (n+1) g (0)
k k + 1
à ! k=0
X n + 1 (k) (n+1−k)
n+1
= f g ,
k=0 k

proving the statement.

❑

Homework

Problem 6.2.1 Prove that Problem 6.2.4 Demonstrate that if for all x ∈ R there holds the
identity
2 1 1 Xn n
X
= −
x2 − 1 x −1 x +1 ak (x − a)k = bk (x − b)k ,
k=0 k=0
2 Ã !
n
X n
and use this result to find the 100th derivative of f (x) = .
x2 − 1 then ak = b j (a − b) j −k .
j =k j

Problem 6.2.2 Find the 100-th derivative of x 7→ x 2 sin x.

Problem 6.2.5 Let p be a polynomial of degree r and consider the
polynomial F with
Problem 6.2.3 Demonstrate that the polynomial p(x) ∈ R[x] has a
zero at x = a of multiplicity k if and only if F (x) = p(x) + p 0 (x) + p 00 (x) + · · · + p (r ) (x).

p(a) = p 0 (a) = · · · = p (k−1) (a) = 0. Prove that ¡ ¢

d F (x) exp(−x)
= −exp(−x)p(x).
dx

115
 Rolle’s Theorem and the Mean Value Theorem

6.3 Rolle’s Theorem and the Mean Value Theorem

h i h i
365 T HEOREM (Rolle’s Theorem) Let (a,b) ∈ R2 such that a < b, f : a ;b → R be such that f is continuous on a ;b and
i h i h
differentiable in a ;b , and f (a) = f (b). Then there exists c ∈ a ;b such that f 0 (c) = 0.

h i
Proof: Since f is continuous on a ;b , by Weierstrass’ Theorem 323,

m= hinf i f (x), M= sup

h i f (x),
x∈ a ;b x∈ a ;b

exist. If m = M, then f is constant and so by Theorem 349, f 0 is identically 0 and there is nothing to prove. Assume
that m < M. Since f (a) = f (b), one may not simultaneously
i h have M = f (a) and m = f (a). Assume thus without
loss of generality that M 6= f (a). Then there exists c ∈ a ;b such that f (c) = M. Now

f (x) − f (c) f (x) − f (c)

lim ≥ 0, lim ≤ 0,
x→c− x −c x→c+ x −c

whence it follows that f 0 (c) = 0, proving the theorem. ❑

h i h i
366 T HEOREM (Mean Value Theorem) Let (a,b) ∈ R2 such that a < b, f : a ;b → R be such that f is continuous on a ;b
i h i h f (b) − f (a)
and differentiable on a ;b . Then there exists c ∈ a ;b such that f 0 (c) = .
b−a

Proof: Put h i f (b) − f (a)

g : a ;b → R, g (x) = f (x) − x.
b−a
h i i h
Then g is continuous on a ;b and differentiable on a ;b , and g (a) = g (b). Since g satisfies the hypotheses of
i h
Rolle’s Theorem, there is c ∈ a ;b such that

f (b) − f (a) f (b) − f (a)

g 0 (c) = 0 =⇒ f 0 (c) − = 0 =⇒ f 0 (c) = ,
b−a b−a
proving the theorem.❑

367 T HEOREM If f : I → R is continuous on the interval I , differentiable on I˚, and if ∀x ∈ I˚, f 0 (x) = 0 then f is constant on
I.
i h
Proof: Let (a,b) ∈ I 2 , a < b. By the Mean Value Theorem, there is c ∈ a ;b such that

f (b) − f (a) = f 0 (c)(b − a) = 0 · (b − a) =⇒ f (b) = f (a),

thus any two outputs have exactly the same value and f is constant. ❑

368 T HEOREM If f : I → R is continuous on the interval I , and differentiable on I˚. Then f is increasing on I if and only if
∀x ∈ I˚, f 0 (x) ≥ 0 and f is decreasing on I if and only if ∀x ∈ I˚, f 0 (x) ≤ 0.

Proof:

=⇒ Suppose f is increasing. Let x0 ∈ I˚. If h 6= 0 is so small that x0 + h ∈ I˚, then

f (x0 + h) − f (x0 ) f (x0 + h) − f (x0 )

≥ 0 =⇒ lim ≥ 0 =⇒ f 0 (x0 ) ≥ 0.
h h→0 h

If f is decreasing we apply what has just been proved to − f .

116
 Chapter 6

i h
⇐ Suppose that for all x ∈ I˚, f 0 (x) ≥ 0. Let (a,b) ∈ I 2 , a < b. By the Mean Value Theorem, there is c ∈ a ;b
such that
f (b) − f (a) = (b − a) f 0 (c) ≥ 0,
and so f is increasing. If for all x ∈ I˚, f 0 (x) ≤ 0 we apply what we just proved to − f .

369 T HEOREM If f : I → R is continuous on the interval I , and differentiable on I˚. Then f is strictly increasing on I if and
only if ∀x ∈ I , f 0 (x) ≥ 0 and the set {xß̊
∈ I ◦ : f 0 (x) = 0} = ∅. Also, f is strictly decreasing on I if and only if ∀x ∈ I , f 0 (x) ≤ 0
and {xß̊
∈ I ◦ : f 0 (x) = 0} = ∅.

Proof:

=⇒ Suppose f is strictly increasing. From Theorem 368 we know that ∀x ∈ I˚, f 0 (x) ≥ 0. Assume that {xß̊∈ I ◦ : f 0 (x) = 0} 6=
i h i h
∅. Then there is c ∈ {xß̊
∈ I ◦ : f 0 (x) = 0} and ε > 0 such that c −ε ;c +ε j I and ∀x ∈ c −ε ;c +ε , f 0 (x) = 0.
i h
By Theorem 367, f must be constant on c − ε ;c + ε and so it is not strictly increasing, a contradiction. If
f is strictly decreasing, we apply what has been proved to − f .

⇐ Conversely, suppose that ∀x ∈ I , f 0 (x) ≥ 0. and the set {xß̊ ∈ I ◦ : f 0 (x) = 0} = ∅. From Theorem 368, f is
2
increasing hon I . iSuppose that there exist i(a,b)h∈ I , a < b such that f (a) = f (b). Since f is increasing, we
have ∀x ∈ a ;b , f (x) = f (a). But then a ;b j {x ∈ I ◦ : f 0 (x) = 0}, a contradiction, since this last set was
assumed empty. If f 0 (x) ≤ 0 we apply what has been proved to − f .

Homework

Problem 6.3.1 Shew, by means of Rolle’s Theorem, that 5x 4 − 4x + Problem 6.3.6 Let n ≥ 1 be an integer and let f : [0;1] → R be differ-
1 = 0 has a solution in [0;1]. entiable and such that f (0) = 0 and f (1) = 1. Prove that there exist
distinct points 0 < a0 < a2 < · · · < an−1 < 1 such that
Problem 6.3.2 Let a0 , a1 ... , an be real numbers satisfying n−1
X 1
a1 a2 an 0 (a )
= n.
a0 + + +··· + = 0. k=0 f k
2 3 n +1
Shew that the polynomial
Problem 6.3.7 (Putnam 1946) Let p(x) is¯a quadratic
¯ polynomial
a0 + a1 x + · · · + an x n with real coefficients satisfying max
h i ¯ f (x)¯ ≤ 1. Prove that
i h x∈ −1 ;1
has a root in 0 ;1 . ¯ 0 ¯
max
h i ¯ f (x)¯ ≤ 4.
x∈ −1 ;1
Problem 6.3.3 Let a,b,c be three functions such that a 0 = b, b 0 = c,
and c 0 = a. Prove that the function a 3 + b 3 + c 3 − 3abc is constant.
Problem 6.3.8 h(Generalised
i Mean Value Theorem)
i h Let f ,g be
h i continuous of a ;b and differentiable on a ;b . Then there is
Problem 6.3.4 Suppose that f : 0 ;1 → R is differentiable, f (0) = i h
i h i h c ∈ a ;b such that
0 and f (x) > 0 for x ∈ 0 ;1 . Is there a number d ∈ 0 ;1 such that

2f 0 (c) f 0 (1 − c) ( f (b) − f (a))g 0 (c) = (g (b) − g (a)) f 0 (c).

= ?
f (c) f (1 − c)
Problem 6.3.9 (First L’Hôpital Rule) Let I be an open interval (fi-
Problem 6.3.5 Let n ≥ 1 be an integer and let f : [0;1] → R be differ- nite or infinite) having c has an endpoint (which may be finite or
entiable and such that f (0) = 0 and f (1) = 1. Prove that there exist infinite). Assume f , g are differentiable on I , g and g 0 never vanish
distinct points 0 < a0 < a2 < · · · < an−1 < 1 such that f 0 (x)
on I and that lim f (x) = 0 = lim g (x). Prove that if lim 0 =L
x→c x→c x→c g (x)
n−1
X
f 0 (ak ) = n. f (x)
(where L is finite or infinite), then lim =L
k=0 x→c g (x)

117
 Extrema

Problem 6.3.10 (Second L’Hôpital Rule) Let I be an open interval Problem 6.3.11 If f 0 exists on an interval containing c, then
(finite or infinite) having c has an endpoint (which may be finite
0 f (c + h) − f (c − h)
or infinite). Assume f , g ¯are differentiable
¯ ¯ on
¯ I , g and g never f 0 (c) = lim .
¯ ¯ ¯ ¯
vanish on I and that lim f (x) = lim g (x) = +∞. Prove that if h→0 2h
x→c x→c
f 0 (x) f (x)
lim 0 = L (where L is finite or infinite), then lim =L Problem 6.3.12 If f 00 exists on an interval containing c, then
x→c g (x) x→c g (x)

f (c + h) + f (c − h) − 2c
f 00 (c) = lim .
h→0 h2

6.4 Extrema
370 Definition Let X j R, f : X → R.

1. We say that f has a local maximum at a if there exists a neighbourhood of a, Na such that ∀x ∈ Na , f (x) ≤ f (a).

2. We say that f has a local minimum at a if there exists a neighbourhood of a, Na such that ∀x ∈ Na , f (x) ≥ f (a).

3. We say that f has a strict local maximum at a if there exists a neighbourhood of a, Na such that ∀x ∈ Na , f (x) < f (a).

4. We say that f has a strict local minimum at a if there exists a neighbourhood of a, Na such that ∀x ∈ Na , f (x) > f (a).

5. We say that f has a local extremum at a if f has either a local maximum or a local minimum at a.

6. We say that f has a strict local extremum at a if f has either a strict local maximum or a strict local minimum at a.
The plural of extremum is extrema.

371 T HEOREM If f : I → R is continuous on the interval I , differentiable on I˚, and if f has a local extremum at a ∈ I˚, then
f 0 (a) = 0.

Proof: Suppose f admits a local maximum at a. Let h 6= 0 be so small that a + h ∈ I . Now

f (a + h) − f (a) f (a + h) − f (a)
h > 0 =⇒ ≤ 0, h < 0 =⇒ ≥ 0.
h h

Upon taking limits as h → 0, f 0 (a) ≤ 0 and f 0 (a) ≥ 0, whence f 0 (a) = 0. ❑

372 Definition Let f : I → R. The points x ∈ I where f 0 (x) = 0 are called critical points or stationary points of f .

h i i h
373 T HEOREM Let f : a ;b → R be a twice differentiable function having a critical point at c ∈ a ;b . If f 00 (c) < 0 then f
has a relative maximum at x = c, and if f 00 (c) > 0 then f has a relative minimum at x = c.

Proof: Assume that f 0 (c) = 0 and f 00 (c) < 0. Since

f 0 (x) f 0 (x) − f 0 (c)

lim = lim = f 00 (c) < 0,
x→c x − c x→c x −c
0 0
there exists δ > 0 such
i that f h(x) > 0 when c − δ < x < c iand f (x)
h > 0 when c < x < x + δ. Consequently, f is
strictly increasing on c − δ ;c and strictly decreasing on c ;c + δ . Hence

|x − c| < δ =⇒ f (x) ≤ f (c),

and so x = c is a local maximum. If f 00 > 0 then we apply what has been proved to − f . ❑
h i
374 T HEOREM (Darboux’s Theorem) Let f be differentiable on a ;b and suppose that f 0 (a) < C < f 0 (b). Then there exists
i h
c ∈ a ;b such that f 0 (c) = C .

118
 Chapter 6

h i
Proof: Put g (x) = f (x) − C x. Then g is differentiable on a ;b . Now g 0 (a) = f 0 (a) − C < 0 so g is strictly
increasing at x = a. Similarly, g 0 (b) = f 0 (b) − Ci < 0 hso g is strictly decreasing at x = b. Since g is continuous, g
must have a local maximum at some point c ∈ a ;b , where g 0 (c) = f 0 (c) − C = 0, proving the theorem. ❑

Homework

Problem 6.4.1 Let f be a polynomial with real coefficients of degree Problem 6.4.2 Put f (0) = 1, f (x) = x x for x > 0. Find the mini-
n such that ∀x ∈ R f (x) ≥ 0. Prove that mum value of f .
∀x ∈ R f (x) + f 0 (x) + f 00 (x) + · · · + f (n) (x) ≥ 0.

6.5 Convex Functions

375 Definition Let I j R be an interval. A function f : I → R is said to be convex if
h i
∀(a,b) ∈ I 2 ,∀λ ∈ 0 ;1 , f (λa + (1 − λ)b) ≤ λ f (a) + (1 − λ) f (b).

We say that f is concave if − f is convex.

☞ f is convex if given any two points on its graph, the straight line joining these two points lies above the graph of f . See
figure 6.1.

b
b

b
b

b
b
b
b

Figure 6.1: A convex curve Figure 6.2: A concave curve.

h i X
n
376 Definition Let (x1 , x2 ,... , xn ) ∈ Rn and let λk ∈ 0 ;1 be such that λk = 1. The sum
k=1

X
n
λk xk
k=1

is called a convex combination of the xk .

h in h i
377 T HEOREM If (x1 , x2 ,... , xn ) ∈ a ;b , then any convex combination of the xk also belongs to a ;b .

h i X
n
Proof: Assume λk ∈ 0 ;1 be such that λk = 1. Since the λk ≥ 0 we have
k=1

a ≤ xk ≤ b =⇒ λk a ≤ λk xk ≤ λk b.
X
n
Adding, and bearing in mind that λk = 1,
k=1
à ! à !
X
n X
n X
n X
n
λk a ≤ λk xk ≤ λk b =⇒ a ≤ λk xk ≤ b,
k=1 k=1 k=1 k=1

proving the theorem. ❑

119
 Convex Functions

378 T HEOREM (Jensen’s Inequality) Let I j R be an interval and let f : I → R be a convex function. Let n ≥ 1 be an integer,
h i X
n
xk ∈ I , and λk ∈ 0 ;1 be such that λk = 1. Then
k=1
à !
X
n X
n
f λk xk ≤ λk f (xk ).
k=1 k=1

h i2
Proof: The proof is by induction on n. For n = 2 we must shew that given (x1 , x2 ) ∈ a ;b ,

f (λ1 x1 + λ2 x2 ) ≤ λ1 f (x1 ) + λ2 f (x2 ).

As λ1 + λ2 = 1, we may put λ = λ2 = 1 − λ1 and so the above inequality becomes

f (λx1 + (1 − λ) x2 ) ≤ λ f (x1 ) + (1 − λ) f (x2 ),

retrieving the definition of convexity.

à ! à !
X
n−1 X
n−1 X
n−1 i h X
n
Assume now that f µk x k ≤ µk f (xk ), when µk = 1, µk ∈ 0 ;1 . We must prove that f λk xk ≤
k=1 k=1 k=1 k=1
X
n X
n i h
λk f (xk ), when λk = 1, λk ∈ 0 ;1 .
k=1 k=1

X λk
n−1
If λn = 1 the assertion is trivial, since then λ1 = · · · = λn−1 = 0. So assume that λn 6= 1. Observe that =
¡Pn ¢ k=1 1 − λn
λ − λn
k=1 k 1 − λn X λk
n−1
= = 1 so that xk is a convex combination of the xk and hence also belongs to
1 − λn 1 − λn k=1 1 − λn
h i
a ;b , by Theorem 377. Since f is convex,

à ! à !
X
n X
n−1
f λk xk = f λk xk + λn xn
k=1 Ãk=1 !
X
n−1
λk
= f (1 − λn ) xk + λn xn
1 − λn
Ãk=1 !
X λk
n−1
≤ (1 − λn ) f xk + λn f (xn )
k=1 1 − λn

λk
By the inductive hypothesis, with µk = = 1,
1 − λn
à !
X λk
n−1 X
n−1 λk
f xk ≤ f (xk ) .
k=1 1 − λn k=1 1 − λn

Finally, we gather, Ã ! Ã !
X
n X
n−1
λk
f λk xk ≤ (1 − λn ) f xk + λn f (xn )
k=1 k=1 1 − λn
X λk
n−1
≤ (1 − λn ) f (xk ) + λn f (xn )
k=1 1 − λn
X
n−1
= λk f (xk ) + λn f (xn )
k=1
Xn
= λk f (xk ) ,
k=1

proving the theorem. ❑

120
 Chapter 6

379 T HEOREM Let I j R be an interval and let f : I → R. For a ∈ I we put

I \ {a} → R
Ta : .
f (x) − f (a)
x 7→
x −a
Then f is convex if and only if ∀a ∈ I , T a is increasing over I \ {a}.

Proof: Let a < b < c as in figure 6.3. Consider the points A(a, f (a)), B (b, f (b)), and C (c, f (c)). The slopes

f (b) − f (a) f (c) − f (b) f (c) − f (a)

m AB = , mBC = , mC A = ,
b−a c −b c −a
satisfy
m AB ≤ m AC , m AC ≤ mBC , m AB ≤ mBC ,
and the theorem follows.
h iAn analytic
h proof
i may be obtained byhobserving
i that from Theorem 377, any λa + (1 − λ)c
lies in the interval a ;c for λ ∈ 0 ;1 . Conversely, given b ∈ a ;c , we may solve for λ the equation

c −b h i
b = λa + (1 − λ)c =⇒ λ = ∈ 0 ;1 .
c −a
Hence
c −b b−a f (b) − f (a) f (c) − f (b)
f (λa + (1 − λ)c) ≤ λ f (a) + (1 − λ) f (c) ⇐⇒ f (b) ≤ f (a) + f (c) ⇐⇒ ≤ .
c −a c −a b−a c −b
(6.1)
This gives
f (b) − f (a) f (c) − f (a) f (c) − f (b)
≤ ≤ (6.2)
b−a c −a c −b
from where the theorem follows.
❑ C

A
B

a b c

Figure 6.3: Theorem 379.

380 T HEOREM Let I j R be an interval and let f : I → R be a convex function. Then f is left and right differentiable on every
point of I˚ and for (a,b,c) ∈ I 3 with a < b < c,
f (b) − f (a) f (c) − f (b)
≤ f − (b) ≤ f + (b) ≤ .
b−a c −b

I \ {b} → R
Proof: Since f is convex, ∀b ∈ I˚, Tb : is increasing, by virtue of Theorem 379. Thus
f (x) − f (b)
x 7→
h h h h x −b
∀u ∈ a ;b , ∀v ∈ b ;c
Tb (a) ≤ Tb (u) ≤ Tb (v ) ≤ Tb (c).

121
 Convex Functions

h h
This means that Tb is increasing on b ;c and bounded below by Tb (u). It follows by Theorem 333 that Tb (b+)
exists, and so f is right-differentiable at b. Moreover,

Tb (a) ≤ Tb (u) ≤ f +0 (b) ≤ Tb (c).

Similarly, Tb is increasing and bounded above by f +0 (b). Appealing again to Theorem 333, f is left-differentiable
at b and
Tb (a) ≤ f −0 (b) ≤ f +0 (b) ≤ Tb (c).
❑

381 C OROLLARY If f is convex on an interval I , then f is continuous on I˚.

Proof: Given b ∈ I˚, we know that f is both left and right differentiable at b (though we may have f −0 (b) < f +0 (b)).
Regardless, this makes f left and right continuous at b: hence both f (b−) = f (b) and f (b+) = f (b). But then
f (b−) = f (b+) and so f is continuous at b. ❑

382 T HEOREM Let I j R be an interval and let f : I → R be differentiable on I . Then f is convex if and only if f 0 is increasing
on I .

Proof:

=⇒ Assume f is convex. Let a < x < c. By (6.2),

f (x) − f (a) f (c) − f (a) f (c) − f (x)

≤ ≤ .
x −a c −a c −x
Taking limits as x → a+,
f (c) − f (a)
f +0 (a) ≤ .
c −a
Taking limits as x → c−,
f (c) − f (a)
≤ f −0 (c).
c −a
Thus f +0 (a) ≤ f −0 (c). Since f is differentiable, f +0 (a) = f 0 (a) and f −0 (c) = f 0 (c), and so f 0 (a) ≤ f 0 (c) proving
that f 0 is increasing.
i h
⇐ Assume f 0 is increasing and that a < x < b. By the Mean Value Theorem, there exists α ∈ a ; x and
i h
α0 ∈ x ;b such that
f (x) − f (a) f (b) − f (x)
= f (α), = f (α0 ).
x −a b−x
Since f 0 (α) ≤ f (α0 ) we must have
f (x) − f (a) f (b) − f (x)
≤ ,
x−a b−x
and so f is convex in view of (6.1).

383 C OROLLARY Let I j R be an interval and let f : I → R be twice differentiable on I . Then f is convex if and only if f 00 ≥ 0.

Proof: This follows from Theorems 368 and 382. ❑

384 Definition An inflexion point is a point on the graph of a function where the graph changes from convex to concave or
viceversa.

Homework

122
 Chapter 6

Problem 6.5.1 (Putnam 1991) A there any polynomials p(x) with

[0;+∞[ → R
real coefficients of degree n ≥ 2 all whose n roots are distinct real Problem 6.5.3 By considering f :
numbers and all whose n − 1 zeroes of p 0 (x) are the midpoints be-
x 7→ x k − k(x − 1)
tween consecutive roots of p(x)?
for 0 < k < 1 and using first and second derivative arguments, ob-
x tain a new proof of Young’s Inequality 314.
Problem 6.5.2 Prove that the inflexion points of x 7→ are
tan x
aligned.

6.6 Inequalities Obtained Through Differentiation

x2
385 T HEOREM Let x > 0. Then < exp(x).
2

x2
Proof: Let f (x) = exp(x) − . Then f 0 (x) = exp(x) − x and f 00 (x) = exp(x) − 1. Since x > 0, f 00 (x) > 0 and so f 0
2
is strictly increasing. Thus f 0 (x) > f 0 (0) = 1 > 0 and so f is increasing. Thus

x2
f (x) > f (0) =⇒ exp(x) − > 0,
2
proving the theorem.❑

x
386 T HEOREM lim = 0.
x→+∞ exp(x)

Proof: From Theorem 385, for x > 0,

x 2 x 2
0< < =⇒ 0 ≤ lim ≤ lim = 0,
exp(x) x x→+∞ exp(x) x→+∞ x

and the theorem follows from the Sandwich Theorem. ❑

xα
387 T HEOREM Let α ∈ R. Then lim = 0.
x→+∞ exp(x)

Proof: If α < 1 then

xα x
= · x α−1 → 0 · 0,
exp(x) exp(x)
by Lemma 386. If α ≥ 1 then
µ ¶α
xα αx
= α−α → α−α · 0α = 0,
exp(x) exp(αx)
by continuity and by Lemma 386. ❑

388 T HEOREM Let x > 0. Then log x < x.

1
Proof: Put f (x) = x − log x. Then f 0 (x) = 1 − . For x < 1, f 0 (x) < 0, for x = 1, f 0 (x) = 0, and for x > 1, f 0 (x) > 0,
x
which means that f has a minimum at x = 1. Thus

f (x) > f (1) =⇒ x − log x > 1.

Since x − log x > 1 then a fortiori we must have x − log x > 0 and the theorem follows.❑

log x
389 L EMMA lim = 0.
x→+∞ x

123
 Inequalities Obtained Through Differentiation

Proof: From Theorem 388, log x 2 < x 2 . For x > 1, log x > 0 and hence,

log x 1
x > 1 =⇒ 0 < < ,
x 2x

log x
whence lim = 0 by the Sandwich Theorem. ❑
x→+∞ x

i h log x
390 T HEOREM Let α ∈ 0 ;+∞ . Then lim = 0.
x→+∞ xα

Proof: If α > 1 then

log x log x 1−α
= ·x → 0 · 0,
xα x
by Lemma 389. If 0 < α ≤ 1 then
log x log x α 1
= → · 0 = 0,
xα αx α α
by continuity and by Lemma 389. ❑

i πh
391 T HEOREM For x ∈ 0 ; , sin x < x < tan x.
2

Proof: Observe that we gave a geometrical iargument for this inequality in Theorem 344. First, let f(x)= sin x −
0 πh
x. Then f (x) = cos x − 1 < 0, since for x ∈ 0 ; , the cosine is strictly positive. This means that f is strictly
i πh 2
decreasing. Thus for all x ∈ 0 ; ,
2

f (0) > f (x) =⇒ 0 > sin x − x =⇒ sin x < x,

giving the first half of the inequality.

i πh
For the second half, put g (x) = tan x − x. Then g 0 (x) = sec2 x − 1. Now, since |cos x| < 1 for x ∈ 0 ; , sec2 x > 1.
2
Hence g 0 (x) > 0, and so g is strictly increasing. This gives

g (0) < g (x) =⇒ 0 < tan x − x =⇒ x < tan x,

obtaining the second inequality.❑

π/2

Figure 6.4: Jordan’s Inequality

i πh 2
392 T HEOREM (Jordan’s Inequality) For x ∈ 0 ; , x < sin x < x.
2 π

124
 Chapter 6

³π ´
Proof: This inequality says that the straight line joining (0,0) to ,1 lies below the curve y = sin x for x ∈
i πh 2 µ ¶
sin x x − tan x
0; . See figure 6.4. Put f (x) = for x 6= 0 and f (0) = 1. Then f 0 (x) = (cos x) < 0 since
2 i πxh i πh x2
cos x > 0 and x − tan x < 0 for x ∈ 0 ; . Thus f is strictly decreasing for x ∈ 0 ; and so
2 2
³π´ sin x 2
f (x) > f =⇒ > ,
2 x π
proving the theorem.❑

393 Definition If w 1 , w 2 ,... , w n are positive real numbers such that w 1 +w 2 +· · ·+w n = 1, we define the r -th weighted power
mean of the xi as:

r
¡ ¢1/r
Mw (x1 , x2 ,... , xn ) = w 1 x1r + w 2 x2r + · · · + w n xnr .
1
When all the w i = we get the standard power mean. The weighted power mean is a continuous function of r , and
n
taking limit when r → 0 gives us
0 w w w
Mw = x1 1 x2 2 · · · w n n .

394 T HEOREM (Generalisation of the AM-GM Inequality) If r < s then

r s
Mw (x1 , x2 ,... , xn ) ≤ M w (x1 , x2 ,... , xn ).

Proof: Suppose first that 0 < r < s are real numbers, and let w 1 , w 2 ,... , w n be positive real numbers such that
w 1 + w 2 + · · · + w n = 1.

s
Put t = > 1 and y i = xir for 1 ≤ i ≤ n. This implies that y it = xis . The function f :]0;+∞[→]0;+∞[, f (x) = x t is
r
1
strictly convex, since its second derivative is f 00 (x) = x t −2 > 0 for all x ∈]0;+∞[. By Jensen’s inequality,
t(t − 1)

(w 1 y 1 + w 2 y 2 + · · · + w n y n )t = f (w 1 y 1 + w 2 y 2 + · · · + w n y n )
≤ w 1 f (y 1 ) + w 2 f (y 2 ) + · · · + w n f (y n )
= w 1 y 1t + w 2 y 2t + · · · + w n y nt .

s
with equality if and only if y 1 = y 2 = · · · = y n . By substituting t = and y i = xir back into this inequality, we get
r

(w 1 x1r + w 2 x2r + · · · + w n xnr )s/r ≤ w 1 x1s + w 2 x2s + · · · + w n xns

with equality if and only if x1 = x2 = · · · = xn . Since s is positive, the function x 7→ x 1/s is strictly increasing, so
raising both sides to the power 1/s preserves the inequality:

(w 1 x1r + w 2 x2r + · · · + w n xnr )1/r ≤ (w 1 x1s + w 2 x2s + · · · + w n xns )1/s ,

which is the inequality we had to prove. Equality holds if and only if all the xi are equal.

The cases r < 0 < s and r < s < 0 can be reduced to the case 0 < r < s.❑

Homework

Problem 6.6.1 Complete the following steps (due to George Pólya) 2. Put
in order to prove the AM-GM Inequality (Theorem 86).

nak
1. Prove that ∀x ∈ R, x ≤ e x−1 . Ak = ,
a1 + a2 + · · · + an

125
 Asymptotic Preponderance

and G n = a1 a2 · · · an . Prove that 3. Deduce that

³ a + a + · · · + a ´n
nnGn 1 2 n
A1 A2 · · · An = , Gn ≤ .
(a1 + a2 + · · · + an )n n
and that
A 1 + A 2 + · · · + A n = n. 4. Prove the AM-GM inequality by assembling the results above.

6.7 Asymptotic Preponderance

395 Definition Let I j R be an interval, and let a ∈ I . A function α : I → R is said to be infinitesimal as x → a if lim α(x) = 0.
x→a
We say that α is negligible in relation to β as x → a or that β is preponderant in relation to α as x → a, if ∀ε > 0,∃δ > 0 such
that i h ¯ ¯
x ∈ a − δ ; a + δ =⇒ |α(x)| ≤ ε ¯β(x)¯ .
¡ ¢
We express the condition above with the notation α(x) = ox →a β(x) (read “α of x is small oh of β of x as x tends to a”).
¡ ¢ ¡ ¢
i α is Big Ohhof β around¯ x = ¯a—written α(x) = Ox→a β(x) , or α(x) <<x→a β(x) —if ∃C > 0 and
Finally, we say that
∃δ > 0 such that ∀x ∈ a − δ ; a + δ , |α(x)| ≤ C ¯β(x)¯.

☞ ¡ ¢ ¡ ¢
Notice that a above may be finite or ±∞. If a is understood, we prefer to write α(x) = o β(x) rather α(x) = ox→a β(x) .
Also
¡ ¢ α(x)
α = ox→a β ⇐⇒ lim = 0 and β(a) = 0 =⇒ α(a) = 0.
x→a β(x)

396 Example sin : R → [−1;1] is infinitesimal as x → 0, since lim sin x = 0.

x→0

1 1
397 Example f : R \ {0} → R, x 7→ is infinitesimal as x → +∞, since lim = 0.
x x→+∞ x

398 Example We have x 2 = o (x) as x → 0 since

x2
lim = lim x = 0.
x→0 x x→0

¡ ¢
399 Example We have x = o x 2 as x → +∞ since

x 1
lim 2
= lim = 0.
x→+∞ x x→+∞ x
¡ ¢ ¡ ¢ ¡ ¢
400 Definition We write α(x) = γ(x) +
¡ o β(x)
¢ as x → a if α(x) − γ(x) = o β(x) as x → a. Similarly, α(x) = γ(x) + O β(x)
as x → a means that α(x) − γ(x) = O β(x) as x → a.

401 Example We have sin x = x + o (x) as x → 0 since

sin x − x sin x
lim = lim − lim 1 = 1 − 1 = 0.
x→0 x x→0 x x→0

402 T HEOREM Let f , g ,α,β,u, v be real-valued functions defined on an interval containing a ∈ R. Let λ ∈ R be a constant.
Let h be a real valued function defined on an interval containing b ∈ R. Then
¡ ¢ ¡ ¢
1. f = o g =⇒ f = O g .

2. f = o (α) =⇒ λ f = o (α).

3. f = o (α) , g = o (α) =⇒ f + g = o (α).

¡ ¢ ¡ ¢
4. f = o (α) , g = o β =⇒ f g = o αβ .

126
 Chapter 6

5. f = O (α) =⇒ λ f = O (α).

6. f = O (α) , g = O (α) =⇒ f + g = O (α).

¡ ¢ ¡ ¢
7. f = O (α) , g = O β =⇒ f g = O αβ .
¡ ¢ ¡ ¢
8. f = O (α) , g = o β =⇒ f g = o αβ .
¡ ¢ ¡ ¢
9. f = O (α) ,α = O β =⇒ f = O β .
¡ ¢ ¡ ¢
10. f = o (α) ,α = O β =⇒ f = o β .
¡ ¢ ¡ ¢
11. f = O (α) ,α = o β =⇒ f = o β .

12. f = o (α) , lim h(x) = a =⇒ f ◦ h = ox→b (α ◦ h).

x→b

13. f = O (α) , lim h(x) = a =⇒ f ◦ h = Ox→b (α ◦ h).

x→b

Proof: These statements follow directly from the definitions.

¡ ¢
1. If f = o g then ∀ε > 0 there exists δ > 0 such that
i h ¯ ¯
¯ f (x) ¯ ¯ ¯ ¯ ¯ ¡ ¢
¯
x ∈ a − δ ; a + δ =⇒ ¯ − 0¯¯ < ε =⇒ ¯ f (x)¯ < ε ¯ g (x)¯ =⇒ f = O g ,
g (x)

using C = ε in the definition of Big Oh.

2. This follows by Theorem 272.
3. This follows by Theorem 272.
f (x) g (x) f (x)g (x) f (x) g (x) ¡ ¢
4. Both lim = 0 and lim = 0. Hence lim = lim · lim = 0 =⇒ f g = o αβ .
x→a α(x) x→a β(x) x→a α(x)β(x) x→a α(x) x→a β(x)
5. If f = O (α) then there is δ > 0 and C > 0 such that
i h ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
x ∈ a − δ ; a + δ =⇒ ¯ f (x)¯ ≤ C ¯ g (x)¯ =⇒ ¯λ f (x)¯ ≤ C |λ| · ¯ g (x)¯ =⇒ λ f = O (α)

6. There exists δ1 > 0,δ2 > 0 and C 1 > 0, C 2 > 0 such that
i h ¯ ¯ i h ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ C 1 |α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ ¯ g (x)¯ ≤ C 2 |α(x)| .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯ ¯ ¯
x ∈ a −δ ; a +δ =⇒ ¯ f (x) + g (x)¯ ≤ ¯ f (x)¯ + ¯ g (x)¯ ≤ C 1 α(x)+C 2 α(x) = (C 1 +C 2 )α(x) =⇒ f + g = O (α) .

7. There exists δ1 > 0,δ2 > 0 and C 1 > 0, C 2 > 0 such that
i h ¯ ¯ i h ¯ ¯ ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ C 1 |α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ ¯ g (x)¯ ≤ C 2 ¯β(x)¯ .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯¯ ¯ ¯ ¯ ¯ ¯ ¡ ¢
x ∈ a−δ ; a+δ =⇒ ¯ f (x)g (x)¯ = ¯ f (x)¯ ¯ g (x)¯ ≤ C 1 |α(x)|·C 2 ¯β(x)¯ = (C 1C 2 ) ¯α(x)β(x)¯ =⇒ f g = O αβ .

8. There exists δ1 > 0,δ2 > 0 and C 1 > 0, such that ∀ε > 0
i h ¯ ¯ i h ¯ ¯ ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ C 1 |α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ ¯g (x)¯ ≤ ε ¯β(x)¯ .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯¯ ¯ ¯ ¯ ¯ ¯ ¡ ¢
x ∈ a − δ ; a + δ =⇒ ¯ f (x)g (x)¯ = ¯ f (x)¯ ¯g (x)¯ ≤ C 1 |α(x)| · ε ¯β(x)¯ = ε(C 1 ) ¯α(x)β(x)¯ =⇒ f g = o αβ .

127
 Asymptotic Preponderance

9. There exists δ1 > 0,δ2 > 0 and C 1 > 0, C 2 > 0 such that
i h ¯ ¯ i h ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ C 1 |α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ |α(x)| ≤ C 2 ¯β(x)¯ .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯ ¡ ¢
x ∈ a − δ ; a + δ =⇒ ¯ f (x)¯ ≤ C 1 |α(x)| ≤ C 1 C 2 ¯β(x)¯ =⇒ f = O β .

10. There exists δ1 > 0,δ2 > 0 and C > 0, such that ∀ε > 0
i h ¯ ¯ i h ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ ε|α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ |α(x)| ≤ C ¯β(x)¯ .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯ ¡ ¢
x ∈ a − δ ; a + δ =⇒ ¯ f (x)¯ ≤ ε|α(x)| ≤ C ε ¯β(x)¯ =⇒ f = o β .

11. There exists δ1 > 0,δ2 > 0 and C > 0, such that ∀ε > 0
i h ¯ ¯ i h ¯ ¯
x ∈ a − δ1 ; a + δ1 =⇒ ¯ f (x)¯ ≤ C |α(x)| and x ∈ a − δ2 ; a + δ2 =⇒ |α(x)| ≤ ε ¯β(x)¯ .

Thus if δ = min(δ1 ,δ2 ),

i h ¯ ¯ ¯ ¯ ¡ ¢
x ∈ a − δ ; a + δ =⇒ ¯ f (x)¯ ≤ C |α(x)| ≤ C ε ¯β(x)¯ =⇒ f = o β .

12. There exists δ1 > 0,δ2 > 0 such that ∀ε > 0

i h ¯ ¯ i h i h
x ∈ a−δ1 ; a+δ1 =⇒ ¯ f (x)¯ ≤ ε|α(x)| and x ∈ b−δ2 ;b+δ2 =⇒ |h(x) − a| ≤ ε =⇒ h(x) ∈ a−ε ; a+ε .

Thus if δ = min(δ1 ,δ2 ,ε),

i h ¯ ¯
x ∈ b − δ ;b + δ =⇒ ¯( f ◦ h)(x)¯ ≤ ε|(α ◦ h)(x)| =⇒ f ◦ h = ox→b (α ◦ h) .

13. There exists δ1 > 0,δ2 > 0,C > 0 such that ∀ε > 0
i h ¯ ¯ i h i h
x ∈ a−δ1 ; a+δ1 =⇒ ¯ f (x)¯ ≤ C |α(x)| and x ∈ b−δ2 ;b+δ2 =⇒ |h(x) − a| ≤ ε =⇒ h(x) ∈ a−ε ; a+ε .

Thus if δ = min(δ1 ,δ2 ,ε),

i h ¯ ¯
x ∈ b − δ ;b + δ =⇒ ¯( f ◦ h)(x)¯ ≤ C |(α ◦ h)(x)| =⇒ f ◦ h = Ox→b (α ◦ h) .

☞ In the above theorem, (8), (10), and (11) essentially say that O (o) = o (O) = o (o) = o and (9) says that O (O) = O.

The following corollary is immediate.

403 C OROLLARY Let α and β be infinitesimal functions as x → a. Then the following hold.

1. The sum of two infinitesimals is an infinitesimal:

¡ ¢ ¡ ¢ ¡ ¢
o β(x) + o β(x) = o β(x) .

2. The difference of two infinitesimals is an infinitesimal:

¡ ¢ ¡ ¢ ¡ ¢
o β(x) − o β(x) = o β(x) .
¡ ¢ ¡ ¢
3. ∀c ∈ R \ {0},o cβ(x) = o β(x) .
¡ ¢ ³ ´
4. ∀n ∈ N,n ≥ 2,1 ≤ k ≤ n − 1, o (β(x))n = o (β(x))k .

128
 Chapter 6

¡ ¡ ¢¢ ¡ ¢
5. o o β(x) = o β(x) .
¡ ¢ ¡ ¢
6. ∀n ∈ N,n ≥ 1, (β(x))n o β(x) = o (β(x))n+1 .
¡ ¢
o (β(x))n ¡ ¢
7. ∀n ∈ N,n ≥ 2, = o (β(x))n−1 .
β(x)
¡ ¢
o β(x)
8. = o (1).
β(x)
à !
X
n
k
¡ ¢
9. If ck are real numbers, then o ck (β(x)) = o β(x) .
k=1
¡ ¢
10. (αβ)(x) = o (α(x)) and (αβ)(x) = o β(x) .
¡ ¢
11. If α ∼ β, then (α − β)(x) = o (α(x)) and (α − β)(x) = o β(x) .

404 T HEOREM (Canonical small oh Relations) The following relationships hold

³ ´
1. ∀(α,β) ∈ R2 , x α = ox→+∞ x β ⇐⇒ α < β.
³ ´
2. ∀(α,β) ∈ R2 , x α = ox→0+ x β ⇐⇒ α > β.

3. log x = ox→+∞ (x).

³ ´
4. ∀(α,β) ∈ R2 ,β > 0, (log x)α = ox →+∞ x β .
¯ ¯α ³ ´
5. ∀(α,β) ∈ R2 ,β < 0, ¯log x ¯ = ox →0+ x β .
¡ ¢
6. ∀(α, a) ∈ R2 , a > 1, x α = ox→+∞ a x
¡ ¢
7. ∀(α, a) ∈ R2 , a > 1, a x = ox→−∞ |x|α

Proof:

1. Immediate.
2. Immediate.
3. This follows from Lemma 389.
4. If α = 0 then eventually (log x)α = 1 and so the assertion is immediate. If α < 0 the assertion is also imme-
diate, since then (log x)α → 0 as x → +∞. If α > 0, by Theorem 390,
log x
→ 0,
x β/α
whence µ ¶α
(log x)α log x
= → 0α = 0.
xβ x β/α
1
5. If x → 0+ then → +∞. Hence by the preceding part and by continuity, as x → 0+ and for γ > 0,
x
µ¯ ¯¶α
¯ ¯
¯log 1 ¯
¯ x¯
µ ¶γ → 0.
1
x
But µ¯ ¯¶α
¯ ¯
¯log 1 ¯ ¡¯ ¯¢
¯ x¯ ¯− log x ¯ α ¯ ¯α
µ ¶γ = µ ¶γ = x γ ¯log x ¯ ,
1 1
x x
¯ ¯α ¡ −γ ¢ ¯ ¯α ³ ´
¯ ¯
and so log x = ox→0+ x , and so putting β = −γ < 0 we have ¯log x ¯ = ox →0+ x β .

129
 Asymptotic Equivalence

6. For α < 1 we have

xα x log a x α−1
= ¡ ¢ · → 0 · 0,
a x exp x log a log a
x log a x α−1
since ¡ ¢ → 0 by continuity and Theorem 387, and → 0 since α − 1 < 0. If α > 1 then
exp x log a log a
 α
à !α log a
α
α α  x  α
x
= ¡
x
¢x = ·  µ α ¶  → α · 0α = 0,
a x
a 1/α (log a)α  log a  (log a)α
exp x
α

by continuity and Theorem 387.

7. If α > 0, a > 1 then |x|α → +∞ but a x → 0 as x → −∞, hence there is nothing to prove. If α = 0, again the
result is obvious. Assume α < 0. If x → −∞ then −x → +∞ and so by the preceding part

|x|−α
→0
a −x
since the above result is valid regardless of the sign of α. Now

ax |x|−α
α = −x ,
|x| a
proving the result.

405 Example In view of Corollary 403 and Theorem 404, we have

¡ ¢
o −2x 3 + 8x 2 = o (x) ,

as x → 0.

406 Example In view of Corollary 403 and Theorem 404, we have

¡ ¢ ¡ ¢
o −2x 3 + 8x 2 = o x 4 ,

as x → +∞.

Homework

Problem 6.7.1 Which one is faster as x → +∞, (log log x)log x or (log x)log log x ?

6.8 Asymptotic Equivalence

407 Definition Let I j R be an interval,
¡ ¢ and let a ∈ I . We say that α is asymptotic to a function β : I → R as x → a, and we
write α ∼ β, if α ∼ β ⇐⇒ α − β = oa β .

☞ If in a neighbourhood Na of a β 6= 0 then


 α

 ∼1
α ∼ β ⇐⇒ β



 β(a) = 0 =⇒ α(a) = 0

sin x
408 Example We have sin x ∼ x as x → 0, since lim = 1.
x→0 x

130
 Chapter 6

x2 + x
409 Example We have x 2 + x ∼ x as x → 0, since lim = 1.
x→0 x

x2 + x
410 Example We have x 2 + x ∼ x 2 as x → +∞, since lim = 1.
x→+∞ x 2

411 T HEOREM 

 ¡ ¢
 α=O β
α ∼ β =⇒


 β = O (α)

¡ ¢
Proof: If α − β = o β there is a neighbourhood Na of a such that
¯ ¯ ¯ ¯
∀ε > 0, x ∈ Na =⇒ ¯α(x) − β(x)¯ ≤ ε ¯β(x)¯ .
1
In particular, for ε = , we have
2
¯ ¯ 1¯ ¯
x ∈ Na =⇒ ¯α(x) − β(x)¯ ≤ ¯β(x)¯ .
2
Hence
¯ ¯ ¯ ¯ ¯ ¯ 3¯ ¯ ¡ ¢
x ∈ Na =⇒ |α(x)| = ¯α(x) − β(x) + β(x)¯ ≤ ¯α(x) − β(x)¯ + ¯β(x)¯ ≤ ¯β(x)¯ =⇒ α = O β ,
2
and
¯ ¯ ¯ ¯ ¯ ¯ 1¯ ¯ ¯ ¯
x ∈ Na =⇒ ¯β(x)¯ = ¯β(x) − α(x) + α(x)¯ ≤ ¯β(x) − α(x)¯+|α(x)| ≤ ¯β(x)¯+|α(x)| =⇒ ¯β(x)¯ ≤ 2 |α(x)| =⇒ β = O (α) .
2

412 T HEOREM The relation of asymptotic equivalence ∼ is an equivalence relation on the set of functions defined on a
neighbourhood of a.

Proof: We have

Reflexivity α − α = 0 = o (α).
¡ ¢
Symmetry α − β = o β =⇒ β = O (α) by Theorem 411. Now by (10) of Theorem 402,
¡ ¢
α − β = o β andβ = O (α) =⇒ α − β = o (α) =⇒ β − α = o (α) ,

whence β ∼ α.
¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢
Transitivity Assume α − β = o β and β − γ = o γ . Then ¡ ¢ we also have β
¡ ¢ by Theorem 411 ¡ =¢ O γ . Hence α − β = o γ
by (10) of Theorem 402. Finally α − β = o γ and β − γ = o γ give α − γ = o γ by (3) of Theorem 402.

❑
The relationship between o,O, and ∼ is displayed in figure 6.5.

413 T HEOREM The relation of asymptotic equivalence ∼ possesses the following properties.



 α∼β
1. =⇒ αγ ∼ βδ.


 γ∼δ

131
 ¡ ¢ ¡ ¢ Asymptotic Equivalence
f =O g g =O f

¡ ¢ f ∼g ¡ ¢
f =o g g =o f

Figure 6.5: Diagram of Big Oh relations.




 α∼β
2. =⇒ αn ∼ βn


 n ∈ N \ {0}

1 1
3. if α ∼ β and if there is a neighbourhood Na of a where ∀x ∈ Na \ {a},β(a) 6= 0, then and are defined on Na \ {a}
α β
1 1
and ∼a .
α β


 ¡ ¢
 α=o β ¡ ¢
4. =⇒ α = o γ .


 β∼γ




 α∼β ¡ ¢
5. =⇒ α = o γ .

 ¡ ¢
 β=o γ

6. if α ∼ β and if there is a neighbourhood Na of a where ∀x ∈ Na \ {a},β(a) > 0, and if r ∈ R then αr ∼a βr .

7. (Dextral Composition) If α ∼a β and if lim γ(x) = a, then α ◦ γ ∼a β ◦ γ.

x→b

Proof: We prove the assertions in the given order.

¡ ¢ ¡ ¢
1. Since α − β = o β and γ − δ = o (δ) then α = O β , and so
¡ ¢ ¡ ¢ ¡ ¢
αγ − βδ = α(γ − δ) − δ(β − α) = O β o (δ) − δo β = o βδ .

2. This follows upon applying the preceding product rule n − 1 times, using γ = α and δ = β.
3. Observe that µ ¶
1 1 β − α o (α) 1
− = = =o ,
α β αβ αβ β
upon using β − α = o (α) and (8) of Corollary 403.
¡ ¢ ¡ ¢ ¡ ¢
4. We have α = o β and β − γ = o γ . This last implies that β = O γ by Theorem 411. Hence
¡ ¢ ¡ ¡ ¢¢ ¡ ¢
α=o β =o O γ =o γ .
¡ ¢ ¡ ¢ ¡ ¢
5. We have α − β = o β and β = o γ . This last implies that α = O β by Theorem 411. Hence
¡ ¢ ¡ ¡ ¢¢ ¡ ¢
α=O β =O o γ =o γ .

α
6. Since β is eventually strictly positive, so is α. Hence α ∼ β ⇐⇒ (x) → 1 as x → a. Since the function
i h β
x 7→ x r is continuous in 0 ;+∞ ,

α αr
(x) → 1 =⇒ r (x) → 1 =⇒ αr ∼ βr .
β β

132
 Chapter 6

α(x) − β(x)
7. We have → 0 as x → a. Now if γ(x) → a as x → b then as x → b,
β(x)

α(γ(x)) − β(γ(x))
→ 0.
β(γ(x))

414 T HEOREM (Exponential Composition) exp(α) ∼a exp(β) ⇐⇒ α − β ∼a 0.

Proof: We have

¡ ¢
exp(α) ∼a exp(β) ⇐⇒ exp(α) − exp(β) = o exp(β)
¡ ¢¡ ¢ ¡ ¢ ¡ ¢
⇐⇒ exp(−β) exp(α) − exp(β) = exp(−β) o exp(β)

⇐⇒ exp(α − β) − 1 = o (1)

⇐⇒ α − β = o (0) .

☞ The above theorem does not say that α ∼ β =⇒ exp(α) ∼ exp(β). That this last assertion is false can be seen from the
following counterexample: x + 1 ∼ x as x → 0, but exp(x + 1) = e exp(x) is not asymptotic to exp(x).

415 T HEOREM (Logarithmic Composition) Suppose there is a neighbourhood of a Na such that

h i
∀x ∈ Na \ {a},β(x) > 0. Suppose, moreover, that α ∼a β and that lim β(x) = l with l ∈ 0 ;+∞ \ {1}. Then log◦α ∼a
x→a
log◦β.

i h
Proof: Either l ∈ 0 ;+∞ \ {1} or l = +∞ or l = 0.

In the first case, logα(x) → logl and logβ(x) → logl as x → a hence

logα ∼ logl ∼ logβ, as x → a.

In the second case β(x) > 1 eventually, and thus logβ(x) 6= 0. Hence

α(x)
log
logα(x) logα(x) − log β(x) β(x) log1
−1 = = → = 0,
logβ(x) logβ(x) log β(x) +∞

α(x)
since → 1 and logβ(x) → +∞ as x → +∞.
β(x)

1 1
The third case becomes the second case upon considering and . ❑
α β

416 T HEOREM (Addition of Positive Terms) If α ∼ β and γ ∼ δ and there exists a neighbourhood of a Na such that ∀x ∈
Na \ {a},β(x) > 0,δ(x) > 0 then
α + γ ∼ β + δ.

133
 Asymptotic Equivalence

¡ ¢
Proof: We have α − β = o β and γ − δ = o (δ). Hence

(α + γ) − (β + δ) = (α − β) + (γ − δ)
¡ ¢
= o β + o (δ)
¡ ¢
= o β+δ ,

which means α + γ ∼ β + δ. ❑

417 T HEOREM The following asymptotic expansions hold as x → 0 :

1. exp(x) − 1 ∼ x and thus exp(x) = 1 + x + o (x)

2. log(1 + x) ∼ x and thus log(1 + x) = x + o (x)

3. sin x ∼ x and thus sin(x) = x + o (x)

4. tan x ∼ x and thus tan(x) = x + o (x)

5. arcsin x ∼ x and thus arcsin(x) = x + o (x)

6. arctanx ∼ x and thus tan(x) = x + o (x)

7. for α ∈ R constant, (1 + x)α − 1 ∼ αx and thus (1 + x)α = 1 + αx + o (x)

x2 x2 ¡ ¢
8. 1 − cos x ∼ and thus cos(x) = 1 − + o x2
2 2

Proof: Results 1—7 follow from the fact that

f (x) − f (a)
f 0 (a) 6= 0, → f 0 (a) =⇒ f (x) − f (a) ∼ f 0 (a)(x − a).
x −a
x
Property 8 follows from the identity 1 − cos x = 2 sin2 .❑
2

418 Example Since tan x = x + o (x), we have

µ 2¶
x2 x2 x x2 ¡ ¢
tan = +o = + o x2 ,
2 2 2 2

as x → 0. Also, ¡ ¢ ¡ ¢
(tan x)3 = (x + o (x))3 = x 3 + 3x 2 o (x) + 3xo x 2 + (o (x))3 = x 3 + o x 3 .

x2 ¡ ¢
419 Example Since cos x = 1 − + o x 2 , we have
2

9x 4 ¡ ¢
cos 3x 2 = 1 − + o x4 .
2

¡ ¢
420 Example Find an asymptotic expansion of cot2 x of type o x −2 as x → 0.

Solution: Since tan x ∼ x we have

1
cot2 x ∼ .
x2
µ ¶
1 1
We can write this as cot2 x = 2
+ o .
x x2

134
 Chapter 6

421 Example Calculate

x2
sin sin tan
lim 2 .
x→0 logcos3x

Solution: We use theorems 417 and 403.

µ 2 ¶
x2 x ¡ ¢
sin sin tan = sin sin + o x2
2 2
µ 2 µ 2 ¶¶
x ¡ ¢ x ¡ ¢
= sin + o x2 + o + o x2
2 2
µ 2 ¶
x ¡ 2¢
= sin +o x
2
2 ¡ ¢
x
= + o x2 ,
2
and µ ¶
9x 2 ¡ 2¢
log cos3x = log 1 − +o x
2
2 µ ¶
9x ¡ ¢ 9x 2 ¡ ¢
= − + o x2 + o − + o x2
2 2!
9x 2 ¡ 2¢
= − +o x
2
The limit is thus equal to
x2 ¡ ¢ 1
+ o x2 + o (1) 1
lim 2 2 = lim 2 =− .
x→0 9x ¡ ¢ x→0 9 9
− + o x2 − + o (1)
2 2

2
422 Example Find lim (cos x)(cot x)
.
x→0

µ ¶
1 1
Solution: By example 420, we have cot2 x = 2
+ o . Also,
x x2
µ ¶
x2 ¡ 2¢ x2 ¡ ¢
logcos x = log 1 − +o x =− + o x2 .
2 2

Hence
2x ¡ ¢
(cos x)cot = exp (cot2 x) logcos x
µµ µ ¶¶ µ 2 ¶¶
1 1 x ¡ 2¢
= exp +o 2 − +o x
x2 x 2
1
= exp(− + o (1))
2
→ e −1/2 ,

as x → 0.

Homework
µ ¶
log(1 + 2tan x) 1 x
Problem 6.8.1 Prove that → 2 as x → 0. Problem 6.8.2 Prove that 1 + → e as x → +∞.
sin x x

π
Problem 6.8.3 Prove that (tan x)cot 4x → e 1/2 as x → .
4

135
 Asymptotic Expansions

6.9 Asymptotic Expansions

423 Definition Let n ∈ N and let f : N0 → R where N0 is a neighbourhood of 0. We say that f admits an asymptotic expan-
sion of order n about x = 0 if there exists a polynomial p of degree n such that
¡ ¢
∀x ∈ N0 , f (x) = p(x) + o0 x n .

The polynomial p is called the regular part of the asymptotic expansion about x = 0 of f .

424 T HEOREM If f admits an asymptotic expansion about 0, its regular part is unique.

¡ ¢ ¡ ¢
Proof: Assume f (x) = p(x) + o0 x n and f (x) = q(x) + o0 x n , where p(x) = p n x n + · · · + p 1 x + p 0 and q(x) =
q n x n +· · ·+q 1 x+q 0 are polynomials of degree n. If p 6= q let k be the largest k for which p k 6= q k . Then subtracting
both equivalencies, as x → 0,
¡ ¢ ¡ ¢ ¡ ¢
p(x)−q(x) = o x n =⇒ (p n −q n )x n +(p n−1 −q n−1 )x n−1 +· · ·+(p 1 −q 1 )x = o x n =⇒ (p k −q k )x k +· · ·+· · · = o x n .
³ ´
But (p k − q k )x k + · · · + · · · O x k as x → 0, a contradiction, since k ≤ n. ❑

425 Definition Let n ∈ N, a ∈ R, and let f : Na → R where Na is a neighbourhood of a. We say that f admits an asymptotic
expansion of order n about x = a if there exists a polynomial p of degree n such that
¡ ¢
∀x ∈ Na , f (x) = p(x − a) + oa (x − a)n .

The polynomial p is called the regular part of the asymptotic expansion about x = a of f .

426 Definition Let n ∈ N, and let f : N+∞ → R where N+∞ is a neighbourhood of +∞. We say that f admits an asymptotic
expansion of order n about +∞ if there exists a polynomial p of degree n such that
i h µ ¶ µ ¶
1 1
∀x ∈ Na ∩ 0 ;+∞ , f (x) = p + o+∞ n .
x x

The polynomial p is called the regular part of the asymptotic expansion about +∞ of f .

¡ ¢
427 T HEOREM Let f : N0 → R be a function with an asymptotic expansion f (x) = p(x) + o0 x n , where p is a polynomial.
Then, if f is even, then p is even and if f is odd, then p is odd.

¡ ¢ ¡ ¢
Proof: Let f (x) = p(x) + o x n as x → 0, where p is a polynomial of degree n. Then f (−x) = p(−x) + o x n . If f
is even then
¡ ¢ ¡ ¢
p(x) + o x n = f (x) = f (−x) = p(−x) + o x n ,

and so by uniqueness of the regular part of an asymptotic expansion we must have p(x) = p(−x), so p is even.
Similarly if f is odd then
¡ ¢ ¡ ¢
−p(x) + o x n = − f (x) = f (−x) = p(−x) + o x n ,

and so by uniqueness of the regular part of an asymptotic expansion we must have −p(x) = p(−x), so p is odd.
❑

We want to expand the function f in powers of x − a:

f (x) = a 0 + a 1 (x − a) + a 2 (x − a)2 + · · · + a n (x − a)n + · · · ,

and that we will truncate at the n-th term, obtaining thereby a polynomial of degree n in powers of x−a. We must determine
what the coefficients a k are, and what the remainder

f (x) − a 0 − a 1 (x − a) − a 2 (x − a)2 − · · · − a n (x − a)n = R(x)

136
 Chapter 6

¡ ¢
is. We hope that this remainder is oa (x − a)n . The coefficients a k are easily found. For 0 ≤ k ≤ n since f is n + 1 times
differentiable, differentiating k times,

f (k) (a)
f (k) (x) = k!a k + ((k + 1)(k) · · · 2)a k+1 (x − a) + ((k + 2)(k + 1) · · · 3)a k+2 (x − a)2 + · · · + R (k) (x), =⇒ = ak ,
k!

as long as R(a) = R 0 (a) = R 00 (a) = · · · = R (n) (a) = 0. We write our ideas formally in the following theorems.

428 T HEOREM (Taylor-Lagrange Theorem) Let I j R, I 6= ∅ be an interval of R and let f : I → R be n + 1 times differentiable
in I . Then if (x, a) ∈ I 2 ,, there exist c with inf(x, a) < c < sup(x, a) such that

f 00 (a) f (3) (a) f (n) (a) f (n+1) (c)

f (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + (x − a)3 + · · · + (x − a)n + (x − a)n+1 .
2! 3! n! (n + 1)!

Proof: If x = a then there is nothing to prove. If x < a then replace x 7→ f (x) with x 7→ f (−x), which then verifies
thehsameihypotheses given in the theorem. Thus it remains to prove the theorem for x > a. Consider the function
φ : a ; x → R with
X
n (x − t)k (x − t)n+1
φ(t) = f (x) − f (k) (t) −R ,
k=0 k! (n + 1)!
where R is a constant. Observe that φ(x) = 0. We now choose the constant
h R iso that φ(a) = 0. Observe that
i φ ish
differentiable and that it satisfies the hypotheses of Rolle’s Theorem on a ; x . Therefore, there exists c ∈ a ; x
such that φ0 (c) = 0. Now
à !
0
X
n
(k+1) (x − t)k (k) (x − t)k−1 (x − t)n (x − t)n (n+1) (x − t)n
φ (t) = − f (t) − f (t) +R =− f (t) + R ,
k=1 k! (k − 1)! n! n! n!

from where we gather, that R = f (n+1) (c) and the theorem follows. ❑

429 C OROLLARY (Taylor-Young Theorem) Let f : Na → R be n +1 times differentiable in Na . Then f admits the asymptotic
expansion of order n about a:

f 00 (a) f (3) (a) f (n) (a) ¡ ¢

f (x) = f (a) + f 0 (a)(x − a) + (x − a)2 + (x − a)3 + · · · + (x − a)n + oa (x − a)n .
2! 3! n!

Proof: Follows at once from Theorem 428. ❑

The following theorem follows at once from Corollary 429.

430 T HEOREM Let x → 0. Then

x3 x5 x 2n+1 ¡ ¢
1. sin x = x − + − · · · + (−1)n + o x 2n+2 .
3! 5! (2n + 1)!

x2 x4 x 2n ¡ ¢
2. cos x = 1 − + − · · · + (−1)n + o x 2n+1 .
2! 4! (2n)!

x 3 2x 5 ¡ ¢
3. tan x = x + + + o x5 .
3 15
x2 x3 xn ¡ ¢
4. e x = 1 + x + + +··· + + o xn
2! 3! n!
x2 x3 xn ¡ ¢
5. log(1 + x) = x − + − · · · + (−1)n+1 + o xn .
2 3 n

137
 Asymptotic Expansions

τ(τ − 1) 2 τ(τ − 1)(τ − 2)(τ − 3) · · · (τ − n + 1) n ¡ ¢

6. (1 + x)τ = 1 + τx + x +··· + x + o xn .
2 n!

¡ ¢
431 Example Find an asymptotic development of log(2 cos x + sin x) around x = 0 of order o x 4 .

Solution: By theorem 430,

µ ¶ µ ¶
x2 x4 ¡ 5¢ x3 ¡ 4¢
2 cos x + sin x = 2 1− + +o x + x− +o x
2 24 6
3 4 ¡ ¢
x x
= 2 + x − x2 − + + o x4
µ 6 12 ¶
x x2 x3 x4 ¡ ¢
= 2 1+ − − + + o x4 ,
2 2 12 24

and so,
µ ¶
x x2 x3 x4 ¡ ¢
log(2 cos x + sin x) = log2 1 + − − + + o x4
2 2 12 24
µ ¶
x x2 x3 x4 ¡ ¢
= log2 + log 1 + − − + + o x4
2 2 12 24
µ ¶
x x2 x3 x4 ¡ 4¢
= log2 + − − + +o x
2 2 12 24
µ ¶
1 x x 2
x 3
x 4 ¡ ¢ 2
− − − + + o x4
2 2 2 12 24
µ ¶
1 x x2 x3 x4 ¡ ¢ 3
+ − − + + o x4
3 2 2 12 24
µ ¶
1 x x2 x3 x4 ¡ 4¢ 4 ¡ ¢
− − − + +o x + o x4
4 2 2 12 24
µ ¶ µ ¶
x x2 x3 x4 1 x2 x3 x4
= log2 + − − + − − +
2 2 12 24 2 4 2 6
µ ¶
1 x 3 3x 4 1 x4 ¡ 4¢
+ − − · +o x
3 8 8 4 16
x 5x 2 5x 3 35x 4 ¡ ¢
= log2 + − + − + o x4
2 8 24 192
as x → 0.

Homework

Problem 6.9.1 Prove that the limit exists. The constant

µ ¶
1 1 1
γ = lim 1+ + +··· + − log n
n→+∞ 2 3 n
µ ¶ is called the Euler-Mascheroni constant. It is not known whether γ
1 1 1
lim 1+ + +··· + − log n, is irrational.
n→+∞ 2 3 n

138
Chapter 7

Integrable Functions

7.1 The Area Problem

h i h i
432 Definition Let f : a ;b → R be bounded, say with m ≤ f (x) ≤ M for all x ∈ a ;b . Corresponding to each partition
h i
P = {x0 , x1 ,... , xn } of a ;b , we define the upper Darboux sum

X
n
U (f ,P) = ( sup f (x))(xk − xk−1 ),
k=1 x k−1 ≤x≤x k

and the lower Darboux sum

X
n
L( f , P ) = ( inf f (x))(xk − xk−1 ).
x k−1 ≤x≤x k
k=1

Clearly
L( f , P ) ≤ U ( f , P ).

Finally, we put
Zb
f (x)dx = inf h i U ( f , P ),
a
P is a partition of a ;b

which we call the upper Riemann integral of f and

Zb
f (x)dx = sup h i L( f , P ).
a
P is a partition of a ;b

which we call the lower Riemann integral of f .

h i Zb Zb
433 Definition Let f : a ;b → R be bounded. We say that f is Riemann integrable if f (x)dx = f (x)dx. In this case,
a a
Zb h i
we denote their common value by f (x)dx and call it the Riemann integral of f over a ;b .
a

h i h i
434 T HEOREM Let f be a bounded function on a ;b and let P j P 0 be two partitions of a ;b . Then

L( f , P ) ≤ L( f , P 0 ) ≤ U ( f , P 0 ) ≤ U ( f , P ).

Proof: Clearly is enough to prove this when P 0 has exactly one more point than P . Let

P = {x0 , x1 ,... , xn }

139
 The Area Problem

with a = x0 < x1 < · · · < xn−1 < xn = b. Let P 0 have the extra point x∗ with xi < x∗ < xi +1 . Observe that we
have both inf f (x) ≤ inf f (x) and inf f (x) ≤ inf f (x) since the larger interval may contain
x i ≤x≤x i +1 x i ≤x≤x ∗ x i ≤x≤x i +1 x ∗ ≤x≤x i +1
smaller values of f . Then

inf f (x)(xi +1 − xi ) = inf f (x)(xi +1 − x∗ + x∗ − xi )

x i ≤x≤x i +1 x i ≤x≤x i +1

= inf f (x)(x∗ − xi ) + inf f (x)(xi +1 − x∗ )

x i ≤x≤x i +1 x i ≤x≤x i +1

≤ inf f (x)(x∗ − xi ) + inf f (x)(xi +1 − x∗ ).

x i ≤x≤x ∗ x ∗ ≤x≤x i +1

Thus

L( f , P ) = ( inf f (x))(x1 − x0 ) + · · · + ( inf f (x))(xi +1 − xi ) + · · · + ( inf f (x))(xn − xn−1 )

x 0 ≤x≤x 1 x i ≤x≤x i +1 x n−1 ≤x≤x n

≤ ( inf f (x))(x1 − x0 ) + · · · + ( inf f (x))(x∗ − xi ) + ( inf f (x))(xi +1 − x∗ ) + · · · + ( inf f (x))(xn − xn−1 )

x 0 ≤x≤x 1 x i ≤x≤x ∗ x ∗ ≤x≤x i +1 x n−1 ≤x≤x n

= L( f , P 0 ).

A similar argument shews that U ( f , P 0 ) ≤ U ( f , P ). The we have

L( f , P ) ≤ L( f , P 0 ) ≤ U ( f , P 0 ) ≤ U ( f , P )

proving te theorem.❑
h i h i
435 T HEOREM Let f be a bounded function on a ;b and let P1 and P2 be any two partitions of a ;b . Then

L( f , P1 ) ≤ U ( f , P2 )

Proof: Let P 0 = P1 ∪ P2 be a common refinement for P1 and P2 . By Theorem 434,

L( f , P1 ) ≤ L( f , P1 ∪ P2 ) ≤ U ( f , P1 ∪ P2 ) ≤ U ( f , P1 ),

and
L( f , P2 ) ≤ L( f , P1 ∪ P2 ) ≤ U ( f , P1 ∪ P2 ) ≤ U ( f , P2 ),
whence the theorem follows.❑

h i Zb Zb
436 T HEOREM Let f be a bounded function on a ;b . Then f (x)dx ≤ f (x)dx.
a a

Proof: By Theorem 435,

Zb
L( f , P1 ) ≤ U ( f , P2 ) =⇒ f (x)dx = sup h i L( f , P1 ) ≤ U ( f , P2 ),
a
P1 is a partition of a ;b

and so Zb
f (x)dx ≤ U ( f , P2 .
a
Taking now the infimum,
Zb Zb
f (x)dx ≤ inf h i U ( f , P2 ) = f (x)dx,
a a
P2 is a partition of a ;b

and the result is established.❑

140
 Chapter 7

h i
437 T HEOREM Let f be a bounded function on a ;b . Then f is Riemann integrable if and only if ∀ε > 0, ∃P a partition
h i
of a ;b such that
U ( f , P ) − L( f , P ) < ε.

Proof:

⇐ If for all ε > 0, U ( f , P ) − L( f , P ) < ε then by Theorem 436,

Zb Zb Zb Zb
L( f , P ) ≤ f (x)dx ≤ f (x)dx ≤ U ( f , P ) =⇒ 0 ≤ f (x)dx − f (x)dx < ε,
a a a a

Zb Zb
and so f (x)dx = f (x)dx, which means that f is Riemann-integrable.
a a
=⇒ Suppose f is Riemann integrable. By the Approximation property of the supremum and infimum, for all
ε > 0 there exist partitions P1 and P2 such that
Zb Zb
ε ε
U ( f , P2 ) − f (x)dx < , f (x)dx − L( f , P1 ) < .
a 2 a 2

Hence by taking P = P1 ∪ P2 then

Zb
ε
U ( f , P ) ≤ U ( f , P2 ) < f (x)dx + < L( f , P1 ) + ε < L( f , P ) + ε,
a 2

from where U ( f , P ) − L( f , P ) < ε.

❑
(
0 x irrational,
438 Example • f (x) = x ∈ [0;1]
1 x rational.
Then U ( f , P ) = 1,L( f , P ) = 0, for any partition P , and so f is not Riemann integrable.

0 x irrational,
• f (x) = 1 p x ∈ [0;1]
 x rational = in lowest terms.
q q
is Riemann integrable with
Z1
f (x) dx = 0
0

h i h i
439 Definition Let f be a bounded function on a ;b and let P = {x0 , x1 ,... , xn } be a partition of a ;b . If tk are selected
so that xk−1 ≤ tk ≤ xk , put
X
n
S( f , P ) = f (tk )(xk − xk−1 ),
k=1

is the Riemann sum of f associated with P .

h i h i h i
440 T HEOREM Let f 1 , f 2 ,... , f m be Riemann integrable over a ;b , and let f : a ;b → R. If for any subinterval I j a ;b
there exists strictly positive numbers a 1 , a 2 ,... , a m such that

ω( f , I ) ≤ a 1 ω( f 1 , I ) + a 2 ω( f 2 , I ) + · · · + a m ω( f m , I ),

then f is also Riemann integrable.

141
 The Area Problem

h i
Proof: Let P = {a = x0 < x1 < · · · < xn = b} be a partition of a ;b selected so that for all j ,

ε
U ( f j , P ) − L( f j , P ) < .
a1 + a2 + · · · + am

Using the notation of the preceding theorem,

U ( f , P ) − L( f , P ) = Z (f ,P)
X
n h i
= ω( f , xk−1 ; xk )(xk − xk−1 )
k=1
Xn Xm h i
≤ a j ω( f j , xk−1 ; xk )(xk − xk−1 )
k=1 j =1
Xm Xn h i
= aj ω( f j , xk−1 ; xk )(xk − xk−1 )
j =1 k=1
Xm ¡ ¢
= a j U ( f j , P ) − L( f j , P )
j =1

< ε,

and the theorem follows from Theorem 437. ❑

h i
441 T HEOREM (Algebra of Riemann Integrable Functions) Let f and g be Riemann integrable functions on a ;b and let
λ ∈ R be a constant. Then the following are also Riemann integrable

1. f + λg
¯ ¯
2. ¯ f ¯

3. f g
¯ ¯ 1
4. provided inf ¯g (x)¯ > 0, also
x∈[a;b] g
¯ ¯ f
5. provided inf ¯g (x)¯ > 0, also
x∈[a;b] g

Proof: Since
¯ ¯ ¯ ¯ ¯ ¯ ¯¯ ¯ ¯¯¯ ¯ ¯
¯ f (x) + λg (x) − f (t) − λg (t)¯ ≤ ¯ f (x) − f (t)¯ + |λ| ¯ g (x) − g (t)¯ , and ¯¯ f (x) − ¯ f (t)¯¯¯ ≤ ¯ f (x) − f (t)¯ ,

we have ¯ ¯
ω( f + λg , I ) ≤ ω( f , I ) + |λ| ω(g , I ) and ω(¯ f ¯ , I ) ≤ ω( f , I ),
from where the first two assertions follow, upon appealing to Theorem 440.

¯ ¯ ¯ ¯
To prove the third assertion, put a 1 = sup ¯ f (u)¯ and a 2 = sup ¯g (u)¯
u∈[a;b] u∈[a;b]

¯ ¯ ¯ ¯
¯ f (x)g (x) − f (t)g (t)¯ = ¯ f (x)(g (x) − g (t)) + g (t)( f (x) − f (t))¯

¯ ¯¯ ¯ ¯ ¯¯ ¯
≤ ¯ f (x)¯ ¯ g (x) − g (t)¯ + ¯ g (t)¯ ¯ f (x) − f (t)¯
à ! à !
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
≤ sup f (u) g (x) − g (t) + sup g (u) ¯ f (x) − f (t)¯
¯ ¯ ¯ ¯ ¯ ¯
u∈[a;b] u∈[a;b]
¯ ¯ ¯ ¯
= a 1 ¯g (x) − g (t)¯ + a 2 ¯ f (x) − f (t)¯ ,

142
 Chapter 7

which gives
ω( f g , I ) ≤ a 1 ω( f , I ) + a 2 ω(g , I ),
and so the third assertion follows from Theorem 440.
¯ ¯
To prove the fourth assertion, with a = inf ¯g (x)¯ > 0, observe that we have
x∈[a;b]
¯ ¯
¯ 1 1 ¯¯ 1 ¯ ¯
¯ − = ¯ ¯ ¯g (x) − g (t)¯
¯ g (x) g (t) ¯ ¯ g (x)g (t)¯
1 ¯¯ ¯
≤ 2
g (x) − g (t)¯ ,
a
1 1
and this gives ω( , I ) ≤ 2 ω(g , I ). The fourth assertion now follows by again appealing to Theorem 440.
g a

The fifth assertion follows from the third and the fourth.❑
h i
442 T HEOREM Let f and g be Riemann integrable functions on a ;b and let λ ∈ R be a constant. Then
Zb Zb Zb
( f (x) + λg (x))dx = f (x)dx + λ g (x)dx.
a a a

h i h i
Proof: Let P = {a = x0 < x1 < · · · < xn = b} be a partition of a ;b and choose tk such that tk ∈ xk−1 ; xk .
Then for any ε > 0 there exist δ > 0 and δ0 > 0 such that
¯ Zb ¯
¯Xn ¯ ε ¯¯ ¯¯
¯ ¯ ¯¯ ¯¯
¯ f (tk )(xk − xk−1 ) − f (x)dx ¯ < if ¯¯P ¯¯ < δ,
¯k=1 a ¯ 2
¯ Zb ¯
¯ X n ¯ ε ¯¯ ¯¯
¯ ¯ ¯¯ ¯¯
¯λ g (tk )(xk − xk−1 ) − λ g (x)dx ¯ < if ¯¯P ¯¯ < δ0 .
¯ k=1 a ¯ 2
¯¯ ¯¯
¯¯ ¯¯
Hence, if ¯¯P ¯¯ < min(δ,δ0 ),
¯ Zb Zb ¯
¯Xn ¡ ¢ ¯
¯ ¯
¯ f (tk ) + λg (tk ) (xk − xk−1 ) − f (x)dx − λ g (x)dx ¯
¯k=1 a a ¯
¯ Zb ¯ ¯ Zb ¯
¯X n ¯ ¯ X n ¯
¯ ¯ ¯ ¯
≤¯ f (tk )(xk − xk−1 ) − f (x)dx ¯ + ¯λ g (tk )(xk − xk−1 ) − λ g (x)dx ¯
¯k=1 a ¯ ¯ k=1 a ¯

<ε

proving the theorem.❑

h i h i
443 T HEOREM Let f and g be Riemann integrable functions on a ;b with f (x) ≤ g (x) for all x ∈ a ;b . Then
Zb Zb
f (x)dx ≤ g (x)dx.
a a

Proof: The function h = g − f is positive for all x ∈ [a;b] and hence L(h, P ) ≥ 0 for all partitions P . It is also
integrable by Theorem 442. Thus
Zb Zb
h(x)dx = h(x)dx ≥ 0.
a a
But Zb Zb Zb Zb
h(x)dx ≥ 0 =⇒ 0 ≤ (g (x) − f (x))dx = g (x)dx − f (x)dx,
a a a a
Zb Zb
and so f (x)dx ≤ g (x)dx, as claimed. ❑
a a

143
 The Area Problem

h i
444 T HEOREM (Triangle Inequality for Integrals) Let f be a Riemann integrable function on a ;b . Then

¯Zb ¯ Zb
¯ ¯ ¯ ¯
¯ f (x)dx ¯≤ ¯ f (x)¯ dx.
¯ ¯
a a

¯ ¯ ¯ ¯ ¯ ¯
Proof: By Theorem 441, ¯ f ¯ is integrable. Now, since − ¯ f ¯ ≤ f ≤ ¯ f ¯ we just need to apply Theorem 443 twice. ❑

h i i h
445 T HEOREM (Chasles’ Rule) Let f be a Riemann integrable function on a ;b and let c ∈ a ;b . Then f is Riemann
h i h i
integrable function on a ;c and c ;b . Moreover,

Zb Zc Zb
f (x)dx = f (x)dx + f (x)dx.
a a c
i h h i h i h i
Conversely, if c ∈ a ;b and f is Riemann integrable on a ;c and c ;b then f is Riemann integrable on a ;b and

Zb Zc Zb
f (x)dx = f (x)dx + f (x)dx.
a a c

Proof: Consider the partitions

P = {a = x0 < x1 < · · · < xm = c < xm+1 < · · · < xn = b}, P 0 = {a = x0 < x1 < · · · < xm = c}, P 00 = {c = xm < xm+1 < · · · < xn = b}.

where by virtue of Theorem 437, given ε > 0, we choose P so that

U ( f , P ) − L( f , P ) < ε.

It follows that
¡ ¢ ¡ ¢
U ( f , P 0 ) − L( f , P 0 ) + U ( f , P 00 ) − L( f , P 00 ) = U ( f , P ) − L( f , P ) < ε.

Hence f is Riemann-integrable over both [a;c] and [c;b]. Observe that

Zc Zb
0 ≤ U ( f , P 0) − f (x)dx < ε, 0 ≤ U ( f , P 00 ) − f (x)dx < ε,
a c

Zc Zb
0≤ f (x)dx − L( f , P 0 ) < ε, 0≤ f (x)dx − L( f , P 00 ) < ε,
a c

and upon addition,

µZc Zb ¶
0 ≤ U (f ,P) − f (x)dx + f (x)dx < 2ε,
a c
µZc Zb ¶
0≤ f (x)dx + f (x)dx − L( f , P ) < 2ε,
a c

whence
Zb Zc Zb
f (x)dx = f (x)dx + f (x)dx,
a a c

as required. ❑

446 T HEOREM (Converse of Chasles’ Rule) Let f be a function defined on the interval [a;b] and let c ∈]a;b[. If f is Riemann-
integrable on [a;c] and [c;b] then it is also Riemann integrable in [a;b] and
Zb Zc Zb
f (x)dx = f (x)dx + f (x)dx.
a a c

144
 Chapter 7

Proof: Since f is Riemann-integrable on both subintervals, it is bounded there, and so it is bounded on the
larger subinterval. By Theorem 437, given ε > 0 there exist partitions P 0 and P 00 such that

U[a;c] ( f , P 0 ) − L [a;c] ( f , P 0 ) < ε, U[c;b] ( f , P 00 ) − L [c;b] ( f , P 00 ) < ε.

The above inequalities also hold in the refinement P = P 0 ∪ P 00 , and

U ( f , P ) = U[a;c] ( f , P ) +U[c;b] ( f , P ), L( f , P ) = L [a;c] ( f , P ) + L [c;b] ( f , P ).

We then deduce that

¡ ¢ ¡ ¢
U ( f , P ) − L( f , P ) = U[a;c] ( f , P ) +U[c;b] ( f , P ) − L [a;c] ( f , P ) + L [c;b] ( f , P )
¡ ¢ ¡ ¢
= U[a;c] ( f , P ) − L [a;c] ( f , P ) − U[a;c] ( f , P ) − L [c;b] ( f , P )

< 2ε,

and so f is Riemann integrable in [a;b] by virtue of Theorem 437. Now

Zb
f (x)dx ≤ U (f ,P)
a

< L( f , P ) + ε

= L [a;c] ( f , P ) + L [c;b] ( f , P ) + ε
Zc Zb
≤ f (x)dx + f (x)dx + ε,
a c

and similarly
Zb
f (x)dx ≥ L( f , P )
a

> U (f ,P) − ε

= U[a;c] ( f , P ) +U[c;b] ( f , P ) − ε
Zc Zb
≥ f (x)dx + f (x)dx − ε,
a c

hence Zc Zb Zb Zc Zb
f (x)dx + f (x)dx − ε ≤ f (x)dx ≤ f (x)dx + f (x)dx + ε
a c a a c

giving the desired equality between integrals.❑

h i
447 T HEOREM Let f be Riemann integrable over [a;b] and let g : inf f (u) ; sup f (u) → R be continuous. Then g ◦ f
u∈[a;b] u∈[a;b]
is Riemann integrable on [a;b].

h i
Proof: Since g is uniformly continuous on the compact interval inf f (u) ; sup f (u) , for given ε > 0 we
u∈[a;b] u∈[a;b]
may find δ0 such that
h i2 ¯ ¯
(s, t) ∈ inf f (t) ; sup f (u) ; |s − t| < δ0 =⇒ ¯ f (s) − f (t)¯ < ε.
t ∈[a;b] u∈[a;b]

Let δ = min(δ0 ,ε). Since f is Riemann-integrable, we may choose a partition P = {a = x0 < x1 < · · · < xn = b}
such that
U ( f , P ) − L( f , P ) < δ2 , (7.1)

145
 The Area Problem

by virtue of Theorem 437. Let

mk = inf f (x); Mk = sup f (x);
x k−1 ≤x≤x k x k−1 ≤x≤x k

mk∗ = inf (g ◦ f )(x); Mk∗ = sup (g ◦ f )(x).

x k−1 ≤x≤x k x k−1 ≤x≤x k

We split the set of indices {1,2,... ,n} into two classes:

A = {k : 1 ≤ k ≤ n, Mk − mk < δ}; B = {k : 1 ≤ k ≤ n, Mk − mk ≥ δ}.

If k ∈ A and xk−1 ≤ x ≤ y ≤ xk , then

¯ ¯ ¯ ¯
¯ f (x) − f (y)¯ ≤ Mk − mk < δ ≤ δ0 =⇒ ¯(g ◦ f )(x) − ( f ◦ g )(y)¯ < ε,

whence Mk∗ − mk∗ ≤ ε. Therefore

X¡ ¢ X
n
Mk∗ − mk∗ (xk − xk−1 ) ≤ ε (xk − xk−1 ) = ε(b − a).
k∈A k=1

If k ∈ B then Mk − mk ≥ δ and by virtue of (7.1),

X X X
δ (xk − xk−1 ) ≤ (Mk − mk )(xk − xk−1 ) ≤ (Mk − mk )(xk − xk−1 ) = U ( f , P ) − L( f , P ) < δ2 ,
k∈B k∈B 1≤k≤n

whence X
(xk − xk−1 ) < δ ≤ ε.
k∈B
¯ ¯
sup ¯ ¯
Upon assembling all these inequalities, and letting M = h i g (t) , we obtain
t ∈ infu∈[a;b] f (u) ;supu∈[a;b] f (u)

X¡ ¢ X¡ ∗ ¢
U (g ◦ f , P ) − L(g ◦ f , P ) = Mk∗ − mk∗ (xk − xk−1 ) + Mk − mk∗ (xk − xk−1 )
k∈A k∈B
X
≤ ε(b − a) + 2M (xk − xk−1 )
k∈B

≤ ε(b − a) + 2Mε

= ε(b − a + 2M),

whence the result follows from Theorem 437. ❑

Zb Za Za
448 Definition If b < a we define f (x)dx = − f (x)dx. Also, f (x)dx = 0.
a b a

h i h i
449 T HEOREM A function f on a ;b is Riemann integrable on a ;b if and only if its set of discontinuities forms a set of
Lebesgue measure 0.

Proof:

=⇒ Given γ > 0 and δ > 0, put ε = γδ. Let f be Riemann integrable. There is a partition P = {a = x0 < x1 <
· · · < xn = b} such that
U ( f , P ) − L( f , P ) < ε.
i h
Let x ∈ xi ; xi +1 be such that ω( f , x) ≥ γ. Then

i sup h f (x) − i inf h f (x) ≥ γ.

x i ;x i +1 x i ;x i +1

146
 Chapter 7

Now observe that

à !
[ i h
{x ∈ [a;b] : ω( f , x) ≥ δ} = xi ; xi +1 ∪ {x0 , x1 ,... , xn }.
sup f −inf f ≥γ

Hence
¡ ¢ X
µ {x ∈ [a;b] : ω( f , x) ≥ γ} ≤ |xi +1 − xi |
sup]x f (x)−inf]xi ;xi +1 [ f (x)≥γ
i ;xi +1 [  
1X  
≤ |xi +1 − xi |  
i sup h f (x) − i inf h f (x)
γ i
x i ;x i +1 x i ;x i +1

1¡ ¢
≤ U ( f , P ) − L( f , P )
γ
ε
<
γ

= δ.
¡ ¢ ¡ ¢
Letting δ → 0+ and γ → 0+ we get µ {x ∈ [a;b] : ω( f , x) ≥ 0} = 0, and in particular, µ {x ∈ [a;b] : ω( f , x) > 0} =
0 which means hat the set of discontinuities is a set of measure 0.
¡ ¢
⇐ Conversely, assume µ {x ∈ [a;b] : ω( f , x) > 0} = 0. We can write

[ 1
{x ∈ [a;b] : ω( f , x) > 0} = {x ∈ [a;b] : ω( f , x) > }.
K ≥1 K
µ ¶
1 1
Fix K large enough so that < ε. Since µ {x ∈ [a;b] : ω( f , x) ≥ } = 0, we can find open intervals I j (K )
K K
such that
1 [ X ¡ ¢
{x ∈ [a;b] : ω( f , x) ≥ } j I j (K ), µ I j (K ) < ε.
K j ≥1 j ≥1
½ ¾
1
It is easy to shew that x ∈ [a;b] : ω( f , x) > is closed and bounded and hence compact, so we may find
K
a finite subcover with
1
{x ∈ [a;b] : ω( f , x) > } j I j 1 ∪ I j 2 ∪ · · · ∪ I j N .
K
Now h i ¡ ¢
a ;b \ I j 1 ∪ I j 2 ∪ · · · ∪ I j N

1
is a finite disjoint union of closed intervals, say J 1 ∪ J 2 ∪ · · · ∪ J M . If x ∈ J i then ω( f , x) < . Thus on each of
K
1
the J i we may find so fine a partition that ω( f ,L) < for every interval such partition. All these partitions
K
and the endpoints of the I j k form a partition, say P . Write S = S 1 ∪ S 2 ∪ · · · ∪ S M for the intervals of the
1
partition P that are not the I j k . Observe that ω( f ,S k ) < . Then
K
à ! à !
X ¡ ¡ ¢¢ X ¡ ¢
U ( f , P ) − L( f , P ) = sup f − inf f µ I j k + sup f − inf f µ (S k )
I jk I jk I jk Sk Sk Sk

¯ ¯ X
N ¡ ¢ 1X
≤ 2 sup ¯ f ¯ µ I jk + µ (S k )
[a;b] k=1 K Sk
¯ ¯
≤ 2 sup ¯ f ¯ ε + (b − a)ε
à [a;b] !
¯ ¯
= ¯ ¯
2 sup f + (b − a) ε.
[a;b]

This proves the theorem.

147
 Integration

❑
h i
450 C OROLLARY Every continuous function f on [a;b] is Riemann integrable on a ;b .

Proof: This is immediate from Theorem 449. ❑

h i h i
451 C OROLLARY Every monotonic function f on a ;b is Riemann integrable on a ;b .

Proof: Since a countable set has measure 0, and since the set of discontinuities of a monotonic function is
countable (Theorem 335), the result is immediate.❑

2
p
f (x) = 2 · x · cos (log x) · sin x
1

0
1 2 3 4 5 6 7 8 9
−1

−2

Homework

h i h i
Problem 7.1.1 Let f be a bounded function on a ;b . Then f is Problem 7.1.3 Let f be bounded on a ;b . Then f is Riemann
h i
Riemann integrable
h ifi and only if ∀ε > 0, ∃δ > 0 such that for all integrable on a ;b if and only if for every ε > 0,ε0 > 0 there is a
partitions P of a ;b , h i
¯¯ ¯¯ partition P of a ;b such that
¯¯ ¯¯
¯¯P ¯¯ < δ =⇒ U ( f , P ) − L( f , P ) < ε.
n
X
h i (xk − xk−1 )χ{x∈[a;b]:ω( f ,[xk−1 ;xk ])≥ε0 } < ε.
Problem 7.1.2 Let f be a bounded function on a ;b . Then f is k=1
h i
Riemann integrable on a ;b if and only if

¯¯ S( f , P )
¯¯ lim Here χ(.) is the indicator function defined on a set E as
¯¯ ¯¯
¯¯P ¯¯→0

exists and is finite. ¯¯ S( f , P ) =
In this case we write ¯¯ lim 

¯¯ ¯¯  1 if x ∈ E
¯¯P ¯¯→0 χE (x) = .
Zb 

 0 if x 6∈ E
f (x)dx.
a

7.2 Integration
h i h i
452 T HEOREM (First Fundamental Theorem of Calculus) Let f : a ;b → R be Riemann integrable on a ;b . If there exists
h i
a differentiable function F : a ;b → R such that F 0 = f then

Zb
f (x)dx = F (b) − F (a).
a

148
 Chapter 7

Proof: Given ε > 0, in view of Theorem 437, there is a partition P = {a = x0 < x1 < · · · < xn = b} such that

U ( f , P ) − L( f , P ) < ε.

Since F is differentiable on [a;b], it is continuous on [a;b]. Applying the Mean Value Theorem to each partition
subinterval [xk−1 ; xk ], we obtain ck ∈]xk−1 ; xk [ such that

F (xk ) − F (xk−1 ) = f (ck )(xk − xk−1 ).

This gives
X X
F (b) − F (a) = (F (xk ) − F (xk−1 )) = f (ck ) (xk − xk−1 ) ,
1≤k≤n 1≤k≤n

and since inf f (u) ≤ f (ck ) ≤ sup f (u), we deduce that

u∈[x k−1 ;x k ] u∈[x k−1 ;x k ]

L( f , P ) ≤ F (b) − F (a) ≤ U ( f , P ).
Zb
Furthermore, we know that L( f , P ) ≤ f (x)dx ≤ U ( f , P ). Hence, combining these two last inequalities,
a
¯ Zb ¯
¯ ¯
¯F (b) − F (a) − f (x)dx ¯ < ε,
¯ ¯
a

and the theorem follows.❑

h i h i
453 T HEOREM (Second Fundamental Theorem of Calculus) Let f : a ;b → R be Riemann integrable on a ;b , and let

Zx h i
F (x) = f (t)dt , x ∈ a ;b .
a
h i i h
Then F is continuous on a ;b . Moreover, if f is continuous at c ∈ a ;b , then F is differentiable at c and F 0 (c) = f (c).

¯ ¯ ¯ ¯
Proof: There is M > 0 such that ∀x ∈ [a;b], ¯ f (x)¯ ≤ M. Now, if a ≤ x < y ≤ b with ¯ x − y ¯ < ε , then
M
¯Z y ¯ Zy Zy
¯ ¯ ¯ ¯ ¯ ¯
¯F (y) − F (x)¯ = ¯ f (t)dt ¯¯ ≤ ¯ f (t)¯ dt ≤ Mdt = M(y − x) < ε
¯
x x x

Thus F is continuous on [a;b] and by Heine’s Theorem, uniformly continuous on [a;b]. Now, take u ∈]a;b[, and
observe that Zx
F (x) − F (u) 1
x 6= u =⇒ = f (t)dt .
x −u x −u u
Moreover, Zx
1
f (u) = f (u)dt ,
x −u u

and therefore,
Zx
F (x) − F (u) 1 ¡ ¢
− f (u) = f (t) − f (u) dt .
x −u x −u u

Since f is continuous at u, there is δ > 0 such that

¯ ¯
t ∈ [a;b],|t − u| < δ =⇒ ¯ f (t) − f (u)¯ < ε.

This gives ¯ ¯
¯ F (x) − F (u) ¯
¯ − f (u) ¯<ε
¯ x −u ¯

for x ∈]a;b[ with |x − u| < δ. From this it follows that F 0 (u) = f (u).❑

149
 Integration

h h
454 T HEOREM (Young’s Inequality for Integrals) Let f be a strictly increasing continuous function on 0 ;+∞ and let f (0) =
0. If A > 0 and B > 0 then ZA ZB
AB ≤ f (x)dx + f −1 (x)dx.
0 0

Proof: The inequality is evident from Figure 7.1. The rectangle of area AB fits nicely in the areas under the
curves y = f (x), x ∈ [0; A] and x = f −1 (y), y ∈ [0;B ]. ❑

1 1 h i
455 T HEOREM (Hölder’s Inequality for Integrals) Let p > 1 and put = 1 − . If f and g are Riemann integrable on a ;b
q p
then ¯Zb ¯ µZb ¶1/p µZb ¶1/q
¯ ¯ ¯ ¯ ¯ ¯
¯ f (x)g (x)dx ¯≤ ¯ f (x)¯p dx ¯g (x)¯q dx .
¯ ¯
a a a

¯ ¯ ¯ ¯p ¯ ¯q
Proof: First observe that all of ¯ f g ¯ , ¯ f ¯ and ¯g ¯ are Riemann-integrable, in view of Theorem 441. Now, with
f (x) = x p−1 in Young’s Inequality (Theorem 454), we obtain,

Ap B 1/(p−1)+1 Ap B q
AB ≤ + = + . (7.2)
p 1/(p − 1) + 1 p q
Zb ¯ ¯
If any of the integrals in the statement of the theorem is zero, the result is obvious. Otherwise put A p = ¯ f (x)¯p dx,
Zb a
¯ ¯
Bq = ¯ g (x)¯p dx. Then by (7.2),
a
¯ ¯ ¯ ¯p ¯ ¯q
¯ f (x)g (x)¯ A −p ¯ f (x)¯ B −q ¯ g (x)¯
≤ + .
AB p q
Integrating throughout the above inequality,
Zb ¯ ¯ Zb ¯ ¯ Zb ¯ ¯
1 ¯ f (x)g (x)¯ dx ≤ 1 ¯ f (x)¯p dx + 1 ¯g (x)¯q dx = 1 + 1 = 1,
AB a p
pA a q
qB a p q
whence the theorem follows. ❑

Figure 7.1: Young’s Inequality (Theorem 454).

h i
456 T HEOREM Let f : a ;b → R. Then
h i h i h i Zb
1. If f is continuous on a ;b , ∀x ∈ a ;b , f (x) ≥ 0, ∃c ∈ a ;b with f (c) > 0 then f (x)dx > 0.
a
h i h i h i Zb
2. If f , g are continuous on a ;b , ∀x ∈ a ;b , f (x) ≤ g (x), and ∃c ∈ a ;b with f (c) < g (c) then f (x)dx <
Zb a

g (x)dx.
a

150
 Chapter 7

Proof: The second part follows from the first by considering f − g . Let us prove the first part.

Assume first that c ∈]a;b[. Then there is a neighourhood ]c − δ;c + δ[j]a;b[ of c, with δ > 0, such that ∀x ∈
f (c)
]c − δ;c + δ[, f (x) ≥ . Therefore
2
Zb Zc+δ Zc+δ
f (c)
f (x)dx ≥ f (x)dx > dx = δ f (c) > 0.
a c−δ c−δ 2

If c = a then we consider a neighbourhood of the form ]a; a + δ[, and similarly if c = b, we consider a neighbour-
hood of the form ]b − δ;b[ ❑
h i h i
457 T HEOREM (First Mean Value Theorem for Integrals) Let f , g be continuous on a ;b , with g of constant sign on a ;b .
i h
Then there exists c ∈ a ;b such that
Zb Zb
f (x)g (x)dx = f (c) g (x)dx.
a a

Proof: If g is identically 0, there is nothing to prove. Similarly, if f is constant in [a;b] there is nothing to prove.
Otherwise, g is always strictly positive or strictly negative in the interval [a;b]. Let

m = inf f (x); M = sup f (x).

x∈[a;b] x∈[a;b]

Then Rb
a f (x)g (x)dx
m< Rb < M.
a g (x)dx
By the Intermediate Value Theorem, there is c ∈]a;b[ such that
Zb
f (x)g (x)dx
a
f (c) = Zb ,
g (x)dx
a

proving the theorem.❑

458 T HEOREM (Integration by Parts) Let f , g be differentiable functions on [a;b] with f 0 and g 0 integrable on [a;b]. Then
Zb Zb ¯b
¯
f 0 (x)g (x)dx + f (x)g 0 (x)dx = f (x)g (x)¯ = f (b)g (b) − f (a)g (a).
a a a

Proof: This follows at once from the Product Rule for Derivatives and the Second Fundamental Theorem of
Calculus, since
Zb Zb Zb
¡ ¢0
( f g )0 = f 0 g + f g 0 =⇒ f (b)g (b) − f (a)g (a) = f (x)g (x) dx f 0 (x)g (x)dx + f (x)g 0 (x)dx.
a a a

459 C OROLLARY (Repeated Integration by Parts) Let n ∈ N. If the n + 1-th derivatives f (n+1) and g (n+1) are continuous on
[a;b] then
Zb ³ ´ ¯b Zb
(n+1) (n) (n−1) (n−1) (n) ¯
f (x)g (x)dx = f (x)g 0
(x) − f (x)g (x) + f (x)g00 n
(x) − · · · + (−1) f (x)g (x) ¯ +(−1) n+1
f (n+1) (x)g (x)dx.
a a a

Proof: Follows by inducting on n and applying Theorem 458. ❑

151
 Integration

460 T HEOREM (Integration by Substitution) Let g be a differentiable function on an open interval I such that g 0 is continu-
ous on I . If f is continuous on g (I ) then f ◦ g is continuous on I and ∀(a,b) ∈ I 2 ,
Zb Zg (b)
( f ◦ g )(x)g 0 (x)dx = f (u)du.
a g (a)

Zx
Proof: Fix c ∈ I and put F (x) = f (u)du. By The Second Fundamental Theorem of Calculus, ∀x ∈ I ,F 0 (x) =
c
f (x). Furthermore, let t(x) = F (g (x)). By The Chain Rule, t 0 = (F 0 ◦ g )g 0 = ( f ◦ g )g 0 . Therefore
Zb Zb
( f ◦ g )(x)g 0 (x)dx = t 0 (x)dx
a a

= t(b) − t(a)

= F (g (b)) − F (g (a))
Zg (b) Z g (a)
= f (u)du − f (u)du
Zc g (b) c

= f (u)du,
g (a)

as was to be shewn. ❑
h i h i
461 T HEOREM (Second Mean Value Theorem for Integrals) Let f , g be continuous on a ;b , with g monotonic on a ;b .
i h
Then there exists c ∈ a ;b such that
Zb Zc Zb
f (x)g (x)dx = g (a) f (x)dx + g (b) f (x)dx.
a a c

Zx
Proof: Put F (x) = f (t)dt. Then F 0 (x) = f (x). Hence
a
Zb Zb ¯b Zb
¯
f (x)g (x)dx = F 0 (x)g (x)dx = F (x)g (x)¯ − F (x)g 0 (x)dx
a a a a

and therefore
Zb Zb
f (x)g (x)dx = F (b)g (b) − F (a)g (a) − F (x)g 0 (x)dx.
a a
By the First Mean Value Theorem for Integrals and by the First Fundamental Theorem of Calculus, there is a
c ∈]a;b[ such that
Zb Zb
F (x)g 0 (x)dx = F (c) g 0 (x)dx = F (c)(g (b) − g (a)).
a a
Assembling all the above,
Zb
f (x)g (x)dx = F (b)g (b) − F (a)g (a) − F (c)(g (b) − g (a))
a

= g (b)(F (b) − F (c)) + g (a)(F (c) − F (a))

Zb Zc
= g (b) f (x)dx + g (a) f (x)dx,
c a

as desired.❑

462 T HEOREM (Generalisation of the AM-GM Inequality) Let a i ≥ 0, p i ≥ 0 with p 1 + p 2 + · · · + p n = 1. Then

p p p
G = a 1 1 a 2 2 · · · a n n ≤ p 1 a 1 + p 2 a 2 + · · · + p n a n = A.

(Here we interpret 00 = 1.)

152
 Chapter 7

Proof: There is a subindex k such that a k ≤ G ≤ a k+1 . Hence

X
k ZG µ ¶
X
n Za i µ ¶
1 1 1 1
pi − dx + pi − dx ≥ 0,
i =1 ai x G i =k+1 G G x

as all the integrands are ≥ 0. Upon rearranging

X
n ZG X n Za i X n X n
1 1 ai − G A
pi dx ≤ pi dx =⇒ p i (log a i − logG) ≤ pi · =⇒ 0 ≤ − 1,
i =1 ai x i =1 G G i =1 i =1 G G

obtaining the inequality ❑

Homework

Z1
Problem 7.2.1 Let p be a polynomial of degree at most 4 such that Problem 7.2.10 Find the value of max(x 2 ,1 − x)dx.
p(−1) = p(1) = 0 and p(0) = 1. If p(x) ≤ 1 for x ∈ [−1;1], find the 0
Z1
largest value of p(x)dx. h i
−1
Problem 7.2.11 Let a > 0. Let f be a continuous function on 0 ; a
h i
Z3 such that f (x) + f (a − x) does not vanish on 0 ; a . Evaluate
Za
Problem 7.2.2 Compute x Tx Udx. f (x)dx
0 .
0 f (x) + f (a − x)

Problem 7.2.3 Let f be a differentiable function such that

Problem 7.2.12 Let a > 0. Let F be a differentiable
Za function such
h i
f (x + h) − f (x) = e x+h − h − e x 0 0
that ∀x ∈ 0 ; a F (a − x) = F (x). Evaluate F (x)dx.
0
and f (0) = 3. Find f (x).
Problem 7.2.13 Let n ≥ 0 be an integer. Let a be the unique differ-
Problem 7.2.4 Let f be aZ continuous function such that f (x) f (a − entiable function such that ∀x ∈ R
a 1
x) = 1 and let a > 0. Find dx. (a(x))2n+1 + a(x) = x.
0 f (x) + 1
Zx
Evaluate a(t )dt .
Problem 7.2.5 Let f be a Riemann integrable function over ev- 0
ery bounded interval and such that f (a + b) = f (a) + f (b) for all
(a,b) ∈ R2 . Demonstrate that f (x) = x f (1). Zπ/2
sin xdx
Problem 7.2.14 Find .
Z2 0 sin x + cos x
Problem 7.2.6 Compute x Tx 2 Udx.
0 Zπ/2
1dx
Problem 7.2.15 Find p .
Z2 0 1 + (tan x) 2
Problem 7.2.7 Find |x 2 − 1| dx.
−1 Z
1
Problem 7.2.16 Find p dx.
Problem 7.2.8 Let n be a fixed integer. Let f : R → R be given by x x2 − 1

 Z

 1
 x if x ≤ 0 Problem 7.2.17 Find p dx.
f (x) = 1+ x +1


 2n if 2n − 2n−2 < x ≤ 2n+1 − 2n−1
Z
x 1/2
Z2n Z2n Problem 7.2.18 Find dx.
n x 1/2 − x 1/3
Prove that f (x)dx = xdx = 22 −1 .
0 0
Z
a 2x
Problem 7.2.9 (Putnam 1938) Evaluate the limit Problem 7.2.19 Find p dx, a > 0.
ax + 1
Zt
(1 + sin 2x)1/x dx Z
1
lim 0 . Problem 7.2.20 Find dx.
t →0 t (e x − e −x )2

153
 Integration

Z5 Z
bxc
Problem 7.2.21 Prove that dx = 4log(5) − 3log(2) − log(3). Problem 7.2.39 Find sin(log x)dx.
1 x

Z
x Z
Problem 7.2.22 Find e e +x dx. log log x
Problem 7.2.40 Find dx.
x
Z
Problem 7.2.23 Find tan x log(cos x) dx. Z
Problem 7.2.41 ( sec xdx in three ways) A traditional indefinite

Z integral is Z
log log x
Problem 7.2.24 Find dx. sec xdx = log(tan x + sec x) +C .
x log x
Justify this formula.
Z 18
x −1
Problem 7.2.25 Find dx. 1 cos x cos x
x3 − 1 Now, prove that = + . Use this to
cos xZ 2(1 + sin x) 2(1 − sin x)
Z find a second formula for sec xdx.
1
Problem 7.2.26 Find dx.
x8 + x
Z A third way is as follows. Using sin 2θ = 2sin θ cosθ shew that
x π
Z csc xdx = log | tan | + C . Now use csc( + x) = sec x to find yet
4x 2
Z 2
Problem 7.2.27 Find dx.
2x + 1 another formula for sec xdx.

Z Z
x2
Problem 7.2.28 Find dx. Problem 7.2.42 Find (arcsin x)2 dx.
(x + 1)10

Z
1 Z
Problem 7.2.29 Find dx. dx
1 + ex Problem 7.2.43 Find p p .
x +1+ x −1
Z
1
Problem 7.2.30 Find dx. Z
1 − sin x Problem 7.2.44 x arctan xdx.

Zp
Problem 7.2.31 Find 1 + sin 2x dx. Z
p
Problem 7.2.45 Find tan xdx.
Z
x
Problem 7.2.32 Find p dx. Z
1 − x4 2x + 1
Problem 7.2.46 Find dx.
x 2 (x − 1)
Z
Problem 7.2.33 Find sec4 xdx.
Z
p
Problem 7.2.47 Find log(x + x)dx.
Z
Problem 7.2.34 Find sec5 xdx.
Z
1
Problem 7.2.48 Find dx.
Z x4 + 1
1/3
Problem 7.2.35 Find ex dx.
Z
1
Problem 7.2.49 Find dx.
Z x3 + 1
Problem 7.2.36 Find log(x 2 + 1)dx.

Problem 7.2.50 Demonstrate that for all strictly positive integers n,

Z
µ ¶ µ ¶ µ ¶ µ ¶
Problem 7.2.37 Find xe x cos xdx. 1 n 1 1 n 1
1+ 1+ < e < 1+ 1+ ,
n 4n n 2n
Z that is, e is contained in the second quarter of the interval
Problem 7.2.38 Find x 2/3 log xdx. hµ ¶ µ
1 n
¶
1 n+1 i
1+ ; 1+ .
n n

154
 Chapter 7

7.3 Riemann-Stieltjes Integration

7.4 Euler’s Summation Formula

155
Chapter 8

Sequences and Series of Functions

8.1 Pointwise Convergence

© ª+∞
463 Definition We say that a sequence of functions f n n=1 f n : I → R defined on an interval I j R converges pointwise to
the function f if ∀x ∈ I , ∀ε > 0 ∃N > 0 (depending on ε and on x) such that
¯ ¯
n ≥ N =⇒ ¯ f n (x) − f (x)¯ < ε.

h i
464 Example The sequence of functions x 7→ x n ,n = 1,2,... converges pointwise on the interval 0 ;1 to the function
h i
f : 0 ;1 → {0,1} with
 h h


 0 if x ∈ 0 ;1
f (x) =


 1 if x = 1

8.2 Uniform Convergence

© ª+∞
465 Definition We say that a sequence of functions f n n=1 f n : I → R defined on an interval I j R converges uniformly to
the function f if ∀x ∈ I , ∀ε > 0 ∃N > 0 (depending only on ε) such that
¯ ¯
n ≥ N =⇒ ¯ f n (x) − f (x)¯ < ε.
unif
In this case we write f n −→ f .

© ª+∞
466 T HEOREM Let f n n=1 be a sequence of functions defined over a common domain I . If there exists a numerical se-
quence {a n }+∞
n=1 with a n → 0 as n → +∞, and a function f defined over I such that eventually
¯ ¯
¯ f n (x) − f (x)¯ ≤ a n ,

unif
then f n −→ f .

© ª+∞
467 T HEOREM If the sequence of continuous functions f n n=1 f n : I → R defined on an open interval I j R converges
uniformly to f on I , then f is continuous on I . Moreover, if x0 ∈ I then we may exchange the limits, as in
µ ¶ ³ ´
lim lim f n (x) = lim lim f n (x) = lim f (x).
n→+∞ x→x 0 x→x 0 n→+∞ x→x 0

© ª+∞
468 T HEOREM If the sequence of integrable functions f n n=1 f n : I → R defined on an open interval I j R converges
uniformly to f on I , then f is integrable on I . Moreover, if (a,b) ∈ I 2 then we may exchange the limit with the integral, as
in µZb ¶ Zb ³ Zb
´
lim f n (x)dx = lim f n (x) dx = f (x)dx.
n→+∞ a a n→+∞ a

156
 Chapter 8

8.3 Integrals and Derivatives of Sequences of Functions

8.4 Power Series

A power series about x = a is a series of the form
X
+∞
f (x) = a n (x − a)n .
n=0

This is a function of x, and truncating it gives polynomial approximations to f . The goal is to approximate “decent” func-
tions about a given point x = a.

These expansions don’t necessarily make sense for all x. The region where the power series converges is called the
interval of convergence.

X∞ 2n (x − 3)n
469 Example Find the interval of convergence of the series p .
n=1 n

Solution: By the ratio test, the series will converge if

¯ 2n+1 (x − 3)n+1 p ¯ r
¯ n ¯ n
¯ p · n ¯ = 2 |x − 3| → r < 1,
n +1 2 (x − 3)n n +1
that is when
5 7
2|x − 3| < 1 =⇒ <x< .
2 2
5 7 5 X∞ (−1)n
The series converges absolutely when < x < . We must also test the endpoints. At x = the series is p ,
2 2 2 n=1 n
5 X∞ 1
which converges conditionally by Leibniz’s Test. At x = the series is p , which diverges.
2 n=1 n

8.5 Maclaurin Expansions to know by inspection

•
x2 x3
ex = 1 + x + + +···
2! 3!
• The sine is an odd function:
x3 x5 x7
sin x = x − + − +···
3! 5! 7!
• The cosine is an even function:
x2 x4 x6
cos x = 1 − + − +···
2! 4! 6!
• If a is a real constant,
a(a − 1) 2 a(a − 1)(a − 2) 3 a(a − 1)(a − 2)(a − 3) 4
(1 + x)a = 1 + ax + x + x + x +···
2! 3! 4!

470 Example Expand f (x) = cos x around x = 1.

Solution: We have

cos x = cos(x − 1 + 1)

= cos(x − 1) cos 1 − sin(x − 1) sin 1

µ ¶ µ ¶
(x − 1)2 (x − 1)4 (x − 1)3 (x − 1)5
= (cos 1) 1 − + − · · · − (sin 1) (x − 1) − + −···
2! 4! 3! 5!

157
 Comparison Tests

Homework

Problem 8.5.1 Given a finite collection of closed squares of total Problem 8.5.2 Given a finite collection of closed squares of total
area 3, prove that they can be arranged to cover the unit square. 1
area , prove that they can be arranged to cover the unit square,
2
with no overlaps

8.6 Comparison Tests

Homework

h i ∞
X
Problem 8.6.1 Let {an }∞ n=1 be a sequence of real numbers satisfying exists and is bounded on 0 ;1 . Prove that if f (an ) converges,
∞
X X∞
n=1
0 < an < 1 for all n. Assume that an diverges but an2 con- ∞ ¯
X ¯
n=1
h i n=1 so does ¯ f (an )¯.
verges. Let f be a function defined on 0 ;1 whose second derivative n=1

8.7 Taylor Polynomials

Homework

Z1
Problem 8.7.1 Evaluate (log x)(log(1 − x))dx. Problem 8.7.3 Find the sum of the infinite series
0

∞
X 2 1 1 1 1 1
Problem 8.7.2 Evaluate the infinite series arctan 2 . 1− + − + − +··· .
n=1 n 4 6 9 11 14

8.8 Abel’s Theorem

Homework

Problem 8.8.1 Put Problem 8.8.3 Evaluate the sum

1 1 1 (−1)n+1
an = 1 − + − + · · · + − log 2. ∞ µ ¶
2 3 4 n X 1 1 1
∞
+ − .
X n=0 4n + 1 4n + 3 2n + 2
Prove that an converges and find its sum.
n=1

Problem 8.8.4 Evaluate the limit

Problem 8.8.2 Evaluate the sum
1 1 (1 Zπ
X∞ 1+ + +··· + 1
2 3 n lim
· tan (αsin x) dx.
. α→0 α 0
n=1 n(n + 1)

158
Appendix A

Answers and Hints

1.1.1 Observe that A n = {0, n, 2n, 3n, . . .}.

1. A6 .

2. N.

3. {0}.

1.1.4 We have,

x ∈ (A ∪ B ) ∩ C ⇐⇒ x ∈ (A ∪ B ) and x ∈ C

⇐⇒ (x ∈ A or x ∈ B ) and x ∈ C

⇐⇒ (x ∈ A and x ∈ C ) or (x ∈ B and x ∈ C )

⇐⇒ (x ∈ A ∩ C ) or (x ∈ B ∩ C )

⇐⇒ x ∈ (A ∩ C ) ∪ (B ∩ C ),

which establishes the equality.

1.1.13 We check the two statements

x ∈ A × (B \ C ) ⇐⇒ x ∈ (A × B ) \ (A × C ).

Let us prove first =⇒. By definition of ×, x = (a, b), where a ∈ A, b ∈ B , b ∉ C . Thus x ∈ A × B but x ∉ A × C . By definition of \ we are done. Now we prove the assertion ⇐=. By definition of × and \, x = (a, b) where a ∈ A, b ∈ B . Since
x ∉ A × C , we observe that b ∉ C . Thus a ∈ A, b ∈ B \ C , and we gather that x ∈ A × (B \ C ).

1.1.14 Attach a binary code to each element of the subset, 1 if the element is in the subset and 0 if the element is not in the subset. The total number of subsets is the total number of such binary codes, and there are 2
N in number.

1.2.1 There are 2

2 = 4 such functions, namely:

¡ ¢
➊ f 1 given by f 1 (a) = f 1 (b) = c. Observe that Im f 1 = {c}.
¡ ¢
➋ f 2 given by f 2 (a) = f 2 (b) = d . Observe that Im f 2 = {d }.
¡ ¢
➌ f 3 given by f 3 (a) = c, f 3 (b) = d . Observe that Im f 3 = {c, d }.
¡ ¢
➍ f 4 given by f 4 (a) = d , f 4 (b) = c. Observe that Im f 4 = {c, d }.

1.2.2 Each of the n elements ofA must be assigned an element of B , and hence there are m · m · · · m = m n possibilities, and thus m n functions.If a function from A to B is injective then we must have n ≤ m in view of Theorem 16. If
| {z }
n factors
to different inputs we must assign different outputs then to the first element of A we may assign any of the m elements of B , to the second any of the m − 1 remaining ones, to the third any of the m − 2 remaining ones, etc., and so we have
m(m − 1) · · ·(m − n + 1) injective functions.

1.2.4 Rename the independent variable, say h(1 − s) = 2s. Now, if 1 − s = 3x then s = 1 − 3x. Hence

h(3x) = h(1 − s) = 2s = 2(1 − 3x) = 2 − 6x.

1.2.5 Put
p(x) = (1 − x 2 + x 4 )2003 = a0 + a1 x + a2 x 2 + · · · + a8012 x 8012 .

Then

➊ a0 = p(0) = (1 − 02 + 04 )2003 = 1.

➋ a0 + a1 + a2 + · · · + a8012 = p(1) = (1 − 12 + 14 )2003 = 1.

➌
a0 − a1 + a2 − a3 + · · · − a8011 + a8012 = p(−1)

= (1 − (−1)2 + (−1)4 )2003

= 1.

p(1) + p(−1)
➍ The required sum is = 1.
2

p(1) − p(−1)
➎ The required sum is = 0.
2

159
 Answers and Hints

1.2.6 We have r r
" à !# 27 " õ ¶ !#3 3
27 + y 3 y 3 3 y
f = f +1
y3 y
 r 3
" õ ¶ !# 3

 3 3 y 

=  f +1 
y

= 53

= 125.

1.2.7 We have

f (2) = (−1)2 1 − 2 f (1) = 1 − 2 f (1)

f (3) = (−1)3 2 − 2 f (2) = −2 − 2 f (2)

f (4) = (−1)4 3 − 2 f (3) = 3 − 2 f (3)

f (5) = (−1)5 4 − 2 f (4) = −4 − 2 f (4)

. . . . .
. . . . .
. . . . .

f (999) = (−1)999 998 − 2 f (998) = −998 − 2 f (998)

f (1000) = (−1)1000 999 − 2 f (999) = 999 − 2 f (999)

f (1001) = (−1)1001 1000 − 2 f (1000) = −1000 − 2 f (1000)

Adding columnwise,
f (2) + f (3) + · · ·+ f (1001) = 1 − 2 + 3 − · · ·+ 999 − 1000 − 2( f (1) + f (2) + · + f (1000)).

This gives
2 f (1) + 3( f (2) + f (3) + · · ·+ f (1000)) + f (1001) = −500.

Since f (1) = f (1001) we have 2 f (1) + f (1001) = 3 f (1). Therefore

500
f (1) + f (2) + · · · + f (1000) = − .
3

1.2.8 Set a = b = 0. Then ( f (0))2 = f (0) f (0) = f (0 + 0) = f (0). This gives f (0)2 = f (0). Since f (0) > 0 we can divide both sides of this equality to get f (0) = 1.

1
Further, set b = −a. Then f (a) f (−a) = f (a − a) = f (0) = 1. Since f (a) 6= 0, may divide by f (a) to obtain f (−a) = .
f (a)
Finally taking a = b we obtain ( f (a))2 = f (a) f (a) = f (a + a) = f (2a). Hence f (2a) = ( f (a))2

1.2.9 To prove that f is injective, we prove that f (a) = f (b) =⇒ a = b . We have

a −1 b −1
f (a) = f (b) =⇒ =
a +1 b +1
=⇒ (a − 1)(b + 1) = (a + 1)(b − 1)

=⇒ ab + a − b − 1 = ab − a + b − 1

=⇒ 2a = 2b

=⇒ a = b,

whence f is injective. To prove that f is surjective we must prove that any y ∈ R \ {1} has a pre-image a ∈ R \ {−1} such that f (a) = y . That is,

a −1 1+y
= y =⇒ a − 1 = y a + y =⇒ a − y a = 1 + y =⇒ a(1 − y) = 1 + y =⇒ a = .
a +1 1−y

µ ¶
1+y 1+x
Thus f = y , and f is surjective. This also serves to prove that f −1 (x) = .
1−y 1−x

1.2.10 We have f [2] (x) = f (x + 1) = (x + 1) + 1 = x + 2, f [3] (x) = f (x + 2) = (x + 2) + 1 = x + 3 and so, recursively, f [n] (x) = x + n.

1.2.14 We have f [2] (x) = f (2x) = 22 x, f [3] (x) = f (22 x) = 23 x and so, recursively, f [n] (x) = 2n x.

1.2.15 Let y = 0. Then 2g (x) = 2x 2 , that is, g (x) = x 2 . Let us check that g (x) = x 2 works. We have

g (x + y) + g (x − y) = (x + y)2 + (x − y)2 = x 2 + 2x y + y 2 + x 2 − 2x y + y 2 = 2x 2 + 2y 2 ,

which is the functional equation given. Our choice of g works.

1.2.16 Let x = 1. Then f (y) = y f (1). Since f (1) is a constant, we may let k = f (1). So all the functions satisfying the above equation satisfy f (y) = k y.

1 1 x 1
1.2.17 From f (x) + 2 f ( ) = x we obtain f ( ) = − f (x). Also, substituting 1/x for x on the original equation we get
x x 2 2

f (1/x) + 2 f (x) = 1/x.

Hence µ ¶
1 1 1 1 x 1
f (x) = − f (1/x) = − − f (x) ,
2x 2 2x 2 2 2

2 x
which yields f (x) = − .
3x 3

160
 Appendix A

1.2.18 We have µ ¶
1−x
( f (x))2 · f = 64x,
1+x
whence
µ µ ¶ ¶2
1−x
( f (x))4 · f ) = 642 x 2 (I )
1+x
1−x
Substitute x by . Then
1+x
µ ¶ µ ¶
1−x 2 1−x
f f (x) = 64 . (I I )
1+x 1+x
Divide (I) by (II),
µ ¶
1+x
f (x)3 = 64x 2 ,
1−x
from where the result follows.

1 x −1
1.2.19 We have (i) f [2] (x) = ( f ◦ f )(x) = f ( f (x)) = = .
1 − 1−x1 x
µ ¶
[3] [2] x −1 1
(ii) f (x) = ( f ◦ f ◦ f )(x) = f ( f (x))) = f = = x.
x 1 − x−1
x
(iii) Notice that f
[4] (x) = ( f ◦ f [3] )(x) = f ( f [3] (x)) = f (x) = f [1] (x). We see that f is cyclic of period 3, that is, f [1] = f [14] = f [7] = . . ., f [2] = f [5] = f [8] = . . . , f [3] = f [6] = f [9] = . . .. Hence f [69] (x) = f [3] (x) = x.

1.2.20 To see (i) observe that

f (a) = f (b) =⇒ g ( f (a)) = g ( f (b)) =⇒ a = b,
whence f is injective. (The first implication is clear, the second implication follows because g ◦ f is injective.)

To see (ii), given y ∈ C , ∃x ∈ A such that g ( f (x)) = y , since g ◦ f is surjective. But then, letting a = f (x) ∈ B we have g (a) = y and g is surjective.

x −a
1.3.1 The map f : [0; 1] → [a; b] f (x) = is a bijection.
b−a
i h i h
1.3.2 The map f : − ∞ ; +∞ → 0 ; +∞ f (x) = e x is a bijection.

p
1.4.1 Both answers are “no.” If a = −b = 2, which we will prove later on to be irrational, we have a + b = 0, rational, and ab = −2, also rational.
p
1 3 2 3
1.4.2 Let ω = − +i . Then ω + ω + 1 = 0 and ω = 1. Then
2 2
x = a 3 + b 3 + c 3 − 3abc = (a + b + c)(a + ωb + ω2 c)(a + ω2 b + cω),

y = u 3 + v 3 + w 3 − 3uv w = (u + v + w)(u + ωv + ω2 w)(u + ω2 v + ωw).

Then
(a + b + c)(u + v + w) = au + av + aw + bu + bv + bw + cu + c v + c w,

(a + ωb + ω2 c)(u + ωv + ω2 w) = au + bw + c v

+ω(av + bu + c w)

+ω2 (aw + bv + cu),

and

(a + ω2 b + ωc)(u + ω2 v + ωw) = au + bw + c v

+ω(aw + bv + cu)

+ω2 (av + bu + c w).

This proves that

xy = (au + bw + c v)3 + (aw + bv + cu)3 + (av + bu + c w)3

−3(au + bw + c v)(aw + bv + cu)(av + bu + c w),

which proves that S is closed under multiplication.

1.4.3 We have

x◦y = (x ◦ y) ◦ (x ◦ y)

= [y ◦ (x ◦ y)] ◦ x

= [(x ◦ y) ◦ x] ◦ y

= [(y ◦ x) ◦ x] ◦ y

= [(x ◦ x) ◦ y] ◦ y

= (y ◦ y) ◦ (x ◦ x)

= y ◦ x,

proving commutativity.

1.4.4 By (1.4)
x ∗ y = ((x ∗ y) ∗ x) ∗ x.
By (1.4) again
((x ∗ y) ∗ x) ∗ x = ((x ∗ y) ∗ ((x ∗ y) ∗ y)) ∗ x.
By (1.3)
((x ∗ y) ∗ ((x ∗ y) ∗ y)) ∗ x = (y) ∗ x = y ∗ x,
which is what we wanted to prove.

To shew that the operation is not necessarily associative, specialise S = Z and x ∗ y = −x − y (the opposite of x minus y ). Then clearly in this case ∗ is commutative, and satisfies (1.3) and (1.4) but

0 ∗ (0 ∗ 1) = 0 ∗ (−0 − 1) = 0 ∗ (−1) = −0 − (−1) = 1,

and
(0 ∗ 0) ∗ 1 = (−0 − 0) ∗ 1 = (0) ∗ 1 = −0 − 1 = −1,
evincing that the operation is not associative.

161
 Answers and Hints

1.4.5 1. Clearly, if a, b are rational numbers,

|a| < 1, |b| < 1 =⇒ |ab| < 1 =⇒ −1 < ab < 1 =⇒ 1 + ab > 0,

a +b a +b
whence the denominator never vanishes and since sums, multiplications and divisions of rational numbers are rational, is also rational. We must prove now that −1 < < 1 for (a, b) ∈] − 1; 1[2 . We have
1 + ab 1 + ab

a +b
−1 < <1 ⇔ −1 − ab < a + b < 1 + ab
1 + ab
⇔ −1 − ab − a − b < 0 < 1 + ab − a − b

⇔ −(a + 1)(b + 1) < 0 < (a − 1)(b − 1).

2
Since (a, b) ∈] − 1; 1[ , (a + 1)(b + 1) > 0 and so −(a + 1)(b + 1) < 0 giving the sinistral inequality. Similarly a − 1 < 0 and b − 1 < 0 give (a − 1)(b − 1) > 0, the dextral inequality. Since the steps are reversible, we have
a +b
established that indeed −1 < < 1.
1 + ab

a +b b+a
2. Since a ⊗ b = = = b ⊗ a , commutativity follows trivially. Now
1 + ab 1 + ba

µ ¶
b +c
a ⊗ (b ⊗ c) = a
µ1 + bc ¶
b +c
a+
1 + bc
= µ ¶
b +c
1+a
1 + bc
a(1 + bc) + b + c a + b + c + abc
= = .
1 + bc + a(b + c) 1 + ab + bc + c a

One the other hand,

µ ¶
a +b
(a ⊗ b) ⊗ c = c
µ 1 + ab¶
a +b
+c
1 + ab
= µ ¶
a +b
1+ c
1 + ab
(a + b) + c(1 + ab)
=
1 + ab + (a + b)c
a + b + c + abc
= ,
1 + ab + bc + c a

whence ⊗ is associative.

a +e
3. If a ⊗ e = a then = a , which gives a + e = a + ea 2 or e(a 2 − 1) = 0. Since a 6= ±1, we must have e = 0.
1 + ae

a +b
4. If a ⊗ b = 0, then = 0, which means that b = −a , that is, a −1 = −a .
1 + ab

1.4.6 We must shew that ∀(a, b) ∈ G

2 we have ab = ba. But

ab = e(ab)e

= (b 2 )(ab)(a 2 )

= b((ba)(ba))a

= b(ba)2 a

= b(e)a

= ba,

whence the result follows.

1.4.7 We have

(ab)3 = a 3 b 3 =⇒ ab(ab)ab = a(a 2 b 2 )b

=⇒ baba = a 2 b 2

=⇒ (ba)2 = a 2 b 2 .

Similarly

(ab)5 = a 5 b 5 =⇒ (ba)4 = a 4 b 4 .

But we also have

(ba)4 = ((ba)2 )2 = (a 2 b 2 )2 = a 2 (b 2 a 2 )b 2 ,

and so
a 2 (b 2 a 2 )b 2 = (ba)4 = a 4 b 4 =⇒ b 2 a 2 = a 2 b 2 .

We have shewn that ∀(a, b) ∈ G

2

((ba)2 = a 2 b 2 ) and (b 2 a 2 = a 2 b 2 ).

Hence

(ba)2 = a 2 b 2 = b 2 a 2 =⇒ baba = b 2 a 2

=⇒ ab = ba,

proving that the group is abelian.

162
 Appendix A

1.4.8 Since
(ab)i +2 = (ab)(ab) · · · (ab) = a(ba)i +1 b,
| {z }
i +2 times

multiplying by a
−1 on the left and by b −1 on the right the equality

(ab)i +2 = a i +2 b i +2 (A.1)

we obtain
(ba)i +1 = (a)i +1 (b)i +1 . (A.2)

By hypothesis
(ab)i +1 = (a)i +1 (b)i +1 . (A.3)

Hence (A.2) and (A.3) yield

(ab)i +1 = (ba)i +1 . (A.4)

Similarly, from (A.3) we obtain

(ab)i = (ba)i , (A.5)

from which
(ab)−i = (ba)−i . (A.6)

Multiplying (A.4) and (A.6) together, we deduce

ab = ba,

which is what we wanted to shew.

1.5.1 The first two follow immediately from the Binomial Theorem, the first by putting x = y = 1 and then x = −y = 1. The third follows by adding the first two and dividing by 2. The fourth follows by subtracting the second from the first and
then dividing by 2.

1.5.2 If a = 103 , b = 2 then

a6 − b6
1002004008016032= a 5 + a 4 b + a 3 b 2 + a 2 b 3 + ab 4 + b 5 = .
a −b
This last expression factorises as
a6 − b6
= (a + b)(a 2 + ab + b 2 )(a 2 − ab + b 2 )
a −b
= 1002 · 1002004 · 998004

= 4 · 4 · 1002 · 250501 · k,

where k < 250000. Therefore p = 250501.

1.5.4 From the Binomial Theorem,

(A + B )3 = A 3 + 3A 2 B + 3AB 2 + B 3 =⇒ A 3 + B 3 = (A + B )3 − 3AB (A + B ).

Then

a 3 + b 3 + c 3 − 3abc = (a + b)3 + c 3 − 3ab(a + b) − 3abc

= (a + b + c)3 − 3(a + b)c(a + b + c) − 3ab(a + b + c)

= (a + b + c)((a + b + c)2 − 3ac − 3bc − 3ab)

= (a + b + c)(a 2 + b 2 + c 2 − ab − bc − c a).

1.5.5 Ã ! Ã !
n n! n (n − 1)! n n −1
= = · = .
k k!(n − k)! k (k − 1)!(n − k)! k k −1

1.5.6 Ã ! Ã !
n n! n(n − 1) (n − 2)! n n −1 n −2
= = · = · · .
k k!(n − k)! k(k − 1) (k − 2)!(n − k)! k k −1 k −2

à ! à !
n n −1
1.5.7 We use the identity k =n . Then
k k −1
à ! à !
n
X n k n
X n −1 k
k p (1 − p)n−k = n p (1 − p)n−k
k k −1
k=1 k=1 Ã !
n−1
X n − 1 k+1
= n p (1 − p)n−1−k
k
k=0 Ã !
n−1
X n −1 k
= np p (1 − p)n−1−k
k
k=0
= np(p + 1 − p)n−1

= np.

1.5.8 We use the identity

à ! à !
n n −2
k(k − 1) = n(n − 1) .
k k −2

Then à ! à !
n
X n k n
X n −2 k
k(k − 1) p (1 − p)n−k = n(n − 1) p (1 − p)n−k
k k −2
k=2 k=2 Ã !
n−2
X n − 2 k+2
= n(n − 1) p (1 − p)n−1−k
k
k=0 Ã !
n−2
X n −1 k
= n(n − 1)p 2 p (1 − p)n−2−k
k
k=0
= n(n − 1)p 2 (p + 1 − p)n−2

= n(n − 1)p 2 .

163
 Answers and Hints

1.5.9 We use the identity

(k − np)2 = k 2 − 2knp + n 2 p 2 = k(k − 1) + k(1 − 2np) + n 2 p 2 .

Then
à !
n
X n k n
X
(k − np)2 p (1 − p)n−k = (k(k − 1) + k(1 − 2np)
k
k=0 k=0 Ã !
n k
+n 2 p 2 ) p (1 − p)n−k
k
à !
n
X n k
= k(k − 1) p (1 − p)n−k
k
k=0 Ã !
Xn n k
+(1 − 2np) k p (1 − p)n−k
k
k=0
à !
n n
X
+n 2 p 2 p k (1 − p)n−k
k
k=0
= n(n − 1)p 2 + np(1 − 2np) + n 2 p 2

= np(1 − p).

1.5.11 Observe that the number of k -tuples with min((a1 , a2 , . . ., ak )) = t is (n − t + 1)k − (n − t)k .

1.7.2 The given equalities entail t

n
X
(x 2 − xk )2 = 0.
k
k=1

A sum of squares is 0 if and only if every term is 0. This gives the result.

1.7.3 The given equality entails that

1³ ´
(x1 − x2 )2 + (x2 − x3 )2 + · · · + (xn−1 − xn )2 + (xn − x1 )2 = 0.
2

A sum of squares is 0 if and only if every term is 0. This gives the result.

a a+A a+A A
1.7.4 Since aB < Ab one has a(b + B ) = ab + aB < ab + Ab = (a + A)b so < . Similarly B (a + A) = aB + AB < Ab + AB = A(b + B ) and so < .
b b +B b +B B

We have
7 11 7 18 11 7 25 18 11
< =⇒ < < =⇒ < < < .
10 15 10 25 15 10 35 25 15

25 5 5 11 4 7
Since = , we have q ≤ 7. Could it be smaller? Observe that > and that < . Thus by considering the cases with denominators q = 1, 2, 3, 4, 5, 6, we see that no such fraction lies in the desired interval. The smallest
35 7 6 15 6 10
denominator is thus 7.

1.7.5 We have
(r − s + t)2 − t 2 = (r − s + t − t)(r − s + t + t) = (r − s)(r − s + 2t).

Since t − s ≤ 0, r − s + 2t = r + s + 2(t − s) ≤ r + s and so

(r − s + t)2 − t 2 ≤ (r − s)(r + s) = r 2 − s 2

which gives

(r − s + t)2 ≤ r 2 − s 2 + t 2 .

n
X
1.7.6 Using the CBS Inequality (Theorem 87) on (ak bk )ck once we obtain
k=1
à !1/2 à !1/2
n
X n
X n
X
ak b k c k ≤ a2 b2 c2 .
k k k
k=1 k=1 k=1
à !1/2
n
X
Using CBS again on a2 b2 we obtain
k k
k=1
à !1/2 à !1/2
n
X n
X n
X
ak b k c k ≤ a2 b2 c2
k k k
k=1 Ãk=1 !1/4 Ã k=1 !1/4 Ã !1/2
n
X n
X Xn
≤ a4 b4 c2 ,
k k k
k=1 k=1 k=1

which gives the required inequality.

1.7.7 This follows directly from the AM-GM Inequality applied to 1, 2, . . ., n :

1+2+···+n n +1
n!1/n (1 · 2 · · ·n)1/n < = ,
n 2

where strict inequality follows since the factors are unequal for n > 1.

1.7.8 First observe that for integer k , 1 < k < n , k(n − k + 1) = k(n − k) + k > 1(n − k) + k = n . Thus

n!2 = (1 · n)(2 · (n − 1))(3 · (n − 2)) · · ·((n − 1) · 2)(n · 1) > n · n · n · · · n = n n .

1.7.9 From the Binomial Theorem, for n ≥ 2,

à ! à ! à ! à ! à !
n n n n n n(n − 1)
2n = (1 + 1)n = + + +···+ > = =⇒ 2n+1 > n(n − 1).
0 1 2 n 2 2

This establishes the inequality for n ≥ 2. For n = 0, 0 = 0(0 − 1) < 2

0+1 and for n = 1, 0 = 1(1 − 1) < 21+1 , so the inequality is true for all natural numbers.

164
 Appendix A

1.7.10 Assume without loss of generality that a ≥ b ≥ c . Then a ≥ b ≥ c is similarly sorted as itself, so by the Rearrangement Inequality

a 2 + b 2 + c 2 = aa + bb + cc ≥ ab + bc + c a.

This also follows directly from the identity

µ ¶
b +c 2 3
a 2 + b 2 + c 2 − ab − bc − c a = a − + (b − c)2 .
2 4

One can also use the AM-GM Inequality thrice:

a 2 + b 2 ≥ 2ab; b 2 + c 2 ≥ 2bc; c 2 + a 2 ≥ 2c a,

and add.

1.7.11 Assume without loss of generality that a ≥ b ≥ c . Then a ≥ b ≥ c is similarly sorted as a 2 ≥ b 2 ≥ c 2 , so by the Rearrangement Inequality

a 3 + b 3 + c 3 = aa 2 + bb 2 + cc 2 ≥ a 2 b + b 2 c + c 2 a,

and
a 3 + b 3 + c 3 = aa 2 + bb 2 + cc 2 ≥ a 2 c + b 2 a + c 2 b.

Upon adding
1³ 2 ´
a 3 + b 3 + c 3 = aa 2 + bb 2 + cc 2 ≥ a (b + c) + b 2 (c + a) + c 2 (a + b) .
2

Again, if a ≥ b ≥ c then
ab ≥ ac ≥ bc,

thus
a 3 + b 3 + c 3 =≥ a 2 b + b 2 c + c 2 a = (ab)a + (bc)b + (ac)c ≥ (ab)c + (bc)a + (ac)b = 3abc.

This last inequality also follows directly from the AM-GM Inequality, as
a3 + b3 + c 3
(a 3 b 3 c 3 )1/3 ≤ ,
3

or from the identity

a 3 + b 3 + c 3 − 3abc = (a + b + c)(a 2 + b 2 + c 2 − ab − bc − c a),

and the inequality of problem 1.7.10.

1.7.12 We apply n times the Rearrangement Inequality

ǎ1 b̂1 + ǎ2 b̂2 + · · · + ǎn b̂n ≤ a1 b 1 + a2 b 2 + · · · + an b n ≤ â1 b̂1 + â2 b̂2 + · · · + ân b̂n

ǎ1 b̂1 + ǎ2 b̂2 + · · · + ǎn b̂n ≤ a1 b 2 + a2 b 3 + · · · + an b 1 ≤ â1 b̂1 + â2 b̂2 + · · · + ân b̂n

ǎ1 b̂1 + ǎ2 b̂2 + · · · + ǎn b̂n ≤ a1 b 3 + a2 b 4 + · · · + an b 2 ≤ â1 b̂1 + â2 b̂2 + · · · + ân b̂n
.
.
.

ǎ1 b̂1 + ǎ2 b̂2 + · · · + ǎn b̂n ≤ a1 bn + a2 b1 + · · · + an bn−1 ≤ â1 b̂1 + â2 b̂2 + · · · + ân b̂n

Adding we obtain the desired inequalities.

1.7.14 Use the fact that (b − a)

2 = (pb − pa)2 (pb + pa)2 .

1.7.15 Let
1 3 5 9999
A= · · ···
2 4 6 10000

and
2 4 6 10000
B= · · ··· .
3 5 7 10001

Clearly, x
2 − 1 < x 2 for all real numbers x . This implies that
x −1 x
<
x x +1

whenever these four quantities are positive. Hence

1/2 < 2/3

3/4 < 4/5

5/6 < 6/7

. . .
. . .
. . .

9999/10000 < 10000/10001

As all the numbers involved are positive, we multiply both columns to obtain
1 3 5 9999 2 4 6 10000
· · ··· < · · ··· ,
2 4 6 10000 3 5 7 10001

or A < B . This yields A 2 = A · A < A · B . Now

1 2 3 4 5 6 7 9999 10000 1
A·B = · · · · · · ··· · = ,
2 3 4 5 6 7 8 10000 10001 10001

and consequently, A
2 < A · B = 1/10001. We deduce that A < 1/p10001 < 1/100.

165
 Answers and Hints

1.7.16 Observe that for k ≥ 1, (x + k)2 > (x + k)(x + k − 1) and so

1 1 1 1
< = − .
(x + k)2 (x + k)(x + k − 1) x +k −1 x +k

Hence
1 1 1 1 1 1 1 1 1 1
+ + +···+ + < + + +···+ +
(x + 1)2 (x + 2)2 (x + 3)2 (x + n − 1)2 (x + n)2 x(x + 1) (x + 1)(x + 2) (x + 2)((x + 3)) (x + n − 2)(x + n − 1) (x + n − 1)(x + n)
1 1 1 1 1 1 1 1 1 1
= − + − + − +···+ − + −
x x +1 x +1 x +2 x +2 x +3 x +n −2 x +n −1 x +n −1 x +n
1 1
= − .
x x +n

1.7.17 For 1 ≤ i ≤ n , we have

¯ ¯ µ µ ¶¶2 µ ¶ µ ¶
¯2 ¯
¯ − 1 − 1 ¯ ≤ 1 − 1 ⇐⇒ 2 − 1 + 1 1 2 4 4 1 4 (i − n)(i − 1)
¯i ≤ 1− ⇐⇒ − 1+ + ≤ 0 ⇐⇒ ≤ 0.
n¯ n i n n i2 i n n i 2n
Thus ¯ ¯ ¯
¯X n x ¯ 1¯X n µ2 µ ¶¶ ¯¯
¯ i ¯ ¯ 1 ¯
¯ ¯= ¯ − 1+ xi ¯ ,
¯ ¯ ¯
2 i =1 i ¯
i =1 i n
X
as xi = 0. Now
i =1 ¯
¯X n µ2 µ ¶¶ ¯¯ n ¯ ¯ µ ¶ n µ ¶
¯ 1 ¯ X ¯¯ 2 1 ¯¯ ¯¯ ¯¯ 1 X ¯ ¯
¯x ¯ = 1 − 1 .
¯ − 1+ xi ¯ ≤ ¯ − 1 − n ¯ xi ≤ 1 − n i
¯ ¯
i =1 i n i =1 i i =1 n

1.7.18 Expanding the product

n
Y n
X n
X n
X
(1 + xk ) = 1 + xk + xi x j + · · · ≥ 1 + xk ,
k=1 k=1 1≤i < j ≤n k=1

n
X
since the x k ≥ 0. When n = 1 equality is obvious. When n > 1 equality is achieved when x i x j = 0.
1≤i < j ≤n

1.7.19 Assume a ≥ b ≥ c . Put s = a + b + c . Then

1 1 1
−a ≤ −b ≤ −c =⇒ s − a ≤ s − b ≤ s − c =⇒ ≥ ≥
s−a s −b s −c
1 1 1
and so the sequences a, b, c and , , are similarly sorted. Using the Rearrangement Inequality twice:
s −a s −b s −c

a b c a b c a b c a b c
+ + ≥ + + ; + + ≥ + + .
s−a s −b s −c s −c s−a s −b s−a s −b s −c s −b s −c s−a

Adding these two inequalities

µ ¶
a b c b +c c +a c +a
2 + + ≥ + + ,
s−a s −b s −c s−a s −b s −c
whence µ ¶
a b c
2 + + ≥ 3,
b +c c +a a +b
from where the result follows.

1.7.20 Let s r q p
p 1 + 4a + 1
P (n) : a+ a+ a +···+ a < .
2
| {z }
n radicands
Let us prove P (1), that is
p
p 1 + 4a + 1
∀a > 0, a< .
2
1
To get this one, let’s work backwards. If a >
4 p
p 1 + 4a + 1 p p
a< ⇐⇒ 2 a < 1 + 4a + 1
2
p p
⇐⇒ 2 a − 1 < 4a + 1
p p
⇐⇒ (2 a − 1)2 < ( 4a + 1)2
p
⇐⇒ 4a − 4 a + 1 < 4a + 1
p
⇐⇒ −2 a < 0.

1 p p
all the steps are reversible and the last inequality is always true. If a ≤ then trivially 2 a − 1 < 4a + 1. Thus P (1) is true. Assume now that P (n) is true and let’s derive P (n + 1). From
4
s r s r s
q p q p
p 1 + 4a + 1 p 1 + 4a + 1
a+ a+ a +···+ a < =⇒ a + a + a + · · · + a < a + .
2 2
| {z } | {z }
n radicands n+1 radicands

we see that it is enough to shew that

s p p
1 + 4a + 1 1 + 4a + 1
a+ = .
2 2
But observe that s
p p
p p 1 + 4a + 1 1 + 4a + 1
( 4a + 1 + 1)2 = 4a + 2 4a + 1 + 2 =⇒ = a+ ,
2 2
proving the claim.

1.7.21 From the AM-GM Inequality,

p p p
a + b ≥ 2 ab; b + c ≥ 2 bc ; c + a ≥ 2 c a,

and the desired inequality follows upon multiplication of these three inequalities.

166
 Appendix A

1.7.22 By the Rearrangement inequality

n a
X n ǎ
X n 1
X
k k
≥ ≥ ,
2 2 k
k=1 k k=1 k k=1

as ǎk ≥ k , the a ’s being pairwise distinct positive integers.

1.7.23 By the AM-GM Inequality,

1 1 1
µ ¶ + +···+
1 1 1 1/n x x2 xn
··· ≤ 1 ,
x1 x2 xn n

whence the inequality.

1.7.24 By the CBS Inequality, ³ ´³ ´

¡ ¢
1 · x 1 + 1 · x 2 + · · · + 1 · x n 2 ≤ 12 + 12 + · · · + 12 x 1
2 + x2 + · · · + x2 ,
2 n

which gives the desired inequality.

1.7.25 Put X X
Tm = ak − ak .
1≤k≤m m<k≤n
Clearly T0 = −Tn . Since the sequence T0 , T1 , . . ., Tn changes signs, choose an index p such that T p−1 and T p have different signs. Thus either T p−1 − T p = 2|a p | or T p − T p−1 = 2|a p |. We claim that

³¯ ¯ ¯ ¯´ ¯ ¯
¯ ¯
min ¯T p−1 ¯ , ¯T p ¯ =≤ max ¯ak ¯ .
1≤k≤n

For ¯ ¯
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
For, if contrariwise both ¯T p−1 ¯ > max ¯ak ¯ and ¯T p ¯ > max ¯ak ¯, then 2|a p | = |T p−1 − T p | > 2 max ¯ak ¯, a contradiction.
1≤k≤n 1≤k≤n 1≤k≤n

1.7.26 It is enough to prove this in the case when a, b, c, d are all positive. To this end, put O = (0, 0), L = (a, b) and M = (a + c, b + d ). By the triangle inequality OM ≤ OL + LM , where equality occurs if and only if the points are
collinear. But then q q q
(a + c)2 + (b + d )2 = OM ≤ OL + LM = a 2 + b 2 + c 2 + d 2 ,

a c
and equality occurs if and only if the points are collinear, that is = .
b d

1.7.31 Use Minkowski’s Inequality and the fact that 17

2 + 1442 = 1452 . The desired value is S
12 .

1.8.3 We have
X X X
(x j − xi ) = xj − xi
1≤i < j ≤n 1≤i < j ≤n 1≤i < j ≤n
Xn n−1
X
= ( j − 1)x j − (n − 1)xi
j =2 i =1
n−1
X
= −(n − 1)x1 + ((k − 1) − (n − k))xk + (n − 1)xn
k=2
= −(n − 1)x1 − (n − 3)x2 − · · · + (n − 3)xn−1 + (n − 1)xn .

This sum is maximal when the negative coefficients of the x i are 0 and the positive coefficients of the x i are equal to 1. If n is even the maximum is

1 + 3 + · · · + (n − 1).

If n is odd, the maximum coefficient is

2 + 4 + · · · + (n − 1).

The result follows thus.

1.8.4 We claim that 3T2t U − 2T3t U = 0, ±1 or −2. We can then take

P (x, y) = (3x − 2y)(3x − 2y − 1)(3x − 2y + 1)(3x − 2y + 2).

In order to prove the claim, we observe that Tx U has unit period, so it is enough to prove the claim for t ∈ [0, 1). We divide [0; 1[ as

[0, 1[= [0; 1/3[∪[1/3; 1/2[∪[1/2; 2/3[∪[2/3; 1[.

If t ∈ [0, 1/3[, then both T2t U and T3t U are = 0, and so 3T2t U −2T3t U = 0. If t ∈ [1/3; 1/2[ then T3t U = 1 and T2t U = 0, and so 3T2t U −2T3t U = −2. If t ∈ [1/2; 2/3[, then T2t U = 1, T3t U = 1, and so 3T2t U −2T3t U =
1. If t ∈ [2/3; 1[, then T2t U = 1, T3t U = 2, and 3T2t U − 2T3t U = −1.

1.8.5 By the Binomial Theorem

à !
p p X n
(1 + 2)n + (1 − 2)n = 2 (2)k := 2N ,
2k
0≤k≤n/2
p p p p p p p
2 < 0, it must be the case that (1− 2)n is the fractional part of (1+ 2)n or (1+ 2)n +1 depending on whether n is odd or even, respectively. Thus for odd n , (1+ 2)n −1 < (1+ 2)n +(1− 2)n <
an even integer. Since −1 < 1−
p p p p p p p p
(1 + 2)n , whence (1 + 2)n + (1 − 2)n = T(1 + 2)n U, always even, and for n even 2N := (1 + 2)n + (1 − 2)n = T(1 + 2)n U + 1, and so T(1 + 2)n U = 2N − 1, always odd for even n.

471 Example Prove that the first thousand digits after the decimal point in
p
(6 + 35)1980

are all 9’s.

Solution: Reasoning as in the preceding problem,

p p
(6 + 35)1980 + (6 − 35)1980 = 2k,
p 1 p p
an even integer. But 0 < 6 − 35 < 1/10, (for if < 6 − 35, upon squaring 3500 < 3481, which is clearly nonsense), and hence 0 < (6 − 35)1980 < 10−1980 which yields
10

1 p
2k − 1 + 0.9 . . .9 = 2k − < (6 + 35)1980 < 2k,
| {z } 101980
1979 nines

This proves the assertion of the problem.

167
 Answers and Hints

1.8.7 By squaring, it is easy to see that

p p p p
4n + 1 < n + n + 1 < 4n + 3.

Neither 4n + 2 nor 4n + 3 are squares since squares are either congruent to 0 or 1 mod 4, so
p p
T 4n + 2U = T 4n + 3U,

and the result follows.

1.8.8
2 2 2 2
Let Tn be the n -th non-square. There is a natural number m such that m < Tn < (m + 1) . As there are m squares less than Tn and n non-squares up to Tn , we see that Tn = n + m. We have then m < n + m < (m + 1) or
1 1 p 1
m 2 −m < n < m 2 +m +1. Since n, m 2 −m, m 2 +m +1 are all integers, these inequalities imply m 2 −m + < n < m 2 +m + , that is to say, (m −1/2)2 < n < (m +1/2)2 . But then m = T n + U. Thus the n -th non-square
p 4 4 2
is Tn = n + T n + 1/2U.

1.8.9 Assume on the contrary that

(a + 2b)2 a2
≥ 2 =⇒ a 2 + 4ab + 4b 2 ≥ 2(a 2 + 2ab + b 2 ) =⇒ 2b 2 ≥ a 2 =⇒ ≥ 2,
(a + b)2 b2

a contradiction. By adding,
a2 (a + 2b)2 a2 (a + 2b)2 (a + 2b)2 a2
< 2, < 2 =⇒ + < 4 =⇒ −2 < 2− .
b2 (a + b)2 b2 (a + b)2 (a + b)2 b2

1.8.10 It needs to be proved that ¯ ¯

¯ 2x + 5 p ¯ ¯¯ p ¯¯
¯ ¯
¯ x + 2 − 5¯ < ¯x − 5¯ .

1.8.11 = {x : x > 0, x n < a}. Shew that E is bounded above with supremum b = sup E . Then shew that b n = a by arguing by contradiction first against b n < a and then against b n > a . In the first case it may be
Consider the set E
µ ¶ µ ¶
a − bn n b n −n
advantageous to prove b + < a for N large enough and use the Binomial Theorem to establish the inequality. In the second case consider b n 1 + > a , for integral M sufficiently large, again using the Binomial Theorem
N Ma
to establish the inequality.

h i
2.2.1 500 ; 501 .

h i
2.2.2 1 ;2 .

2.2.3 R.

2.2.4 {1}.

2.2.5 ∅.

2.2.6 ∅.

p
2.2.8 Closure is immediate. Most of the other axioms are inherited from the larger set R. Observe 0F = 0, 1F = 1 and the multiplicative inverse of a + 2b, (a, b) 6= (0, 0) is

p p
p 1 a − 2b a 2b
(a + 2b)−1 = p = = − .
a + 2b a 2 − 2b 2 a 2 − 2b 2 a 2 − 2b 2

Here a
2 − 2b 2 6= 0 since p2 is irrational.

2 p p p p 2
Assume (a, b) ∈ R with a < b . If ab < 0, then 0 ∈ D is between a and b . If 0 < a < b then a < b , and since Q is dense in R, there is a rational number r such that a < r < b =⇒ a < r < b . If a < b < 0, then
2.2.9
p p p p
−b < −a , and since Q is dense in R, there is a rational number s such that −b < s < −a =⇒ −b < s 2 < −a =⇒ a < −s 2 < b .

2 with a < b . There is a strictly positive integer n such that n > 1

2.2.10 Assume (a, b) ∈ R . Thus
b−a

1 1
0 < n < < b − a.
2 n

Put m = T2
n a U + 1, and so by definition m − 1 ≤ 2n x < m . Hence
m 1 1
a < n ≤ a + n < a + < a + b − a = a.
2 2 n

2.6.6 For the proof of this let G be such a set (so that x + y is in G if x, y are, and G is closed), and suppose that we are not in cases (i) or (ii). Then it is enough to show that G contains arbitrarily small positive numbers, for then multiples of
these will be dense in R , but G being closed forces G = R. To achieve this let I = inf{x : x ∈ G, x > 0}. If I = 0 we are done; but if I > 0 there cannot be numbers x ∈ G arbitrarily close to and greater than I , for then x − I would run
through small positive members of G , in particular smaller than I , contradicting its definition. This means that I belongs itself to G , and from there it is easy to see that we are in case (ii) contrary to the assumption. Hence indeed I = 0,
G = R.

1
3.2.1 No. Take an = . Then an > 0 always, but L = 0.
n

3.2.9 We have for n > 1,

n2 n n n
X n n n n2
= +···+ < < +···+ = ,
n2 + n n2 + n n2 + n 2
i =1 n + i n2 + 1 n2 + 1 n2 + 1
| {z } | {z }
n times n times

and the result follows by the Sandwich Theorem since each of the sequences on the extremes converges to 1.

3.2.10 Evidently n! ≤ n
n . By problem 1.7.8, if n > 2 then n n/2 ≤ n!. Thus
1 1 1
≤ ≤
n (n!)1/n n 1/2

and the result follows by the Sandwich Theorem.

3.2.11 For n ≥ 2 we have

2n 2 2 2 2 2 4
= · · · · · ≤ 2 · 1 · 1 · · ·1 · = → 0.
n! 1 2 3 n n n

168
 Appendix A

3.2.12 There is a positive integer m with m

2 ≤ n < (m + 1)2 . Consider
¯s ¯
¯ m2 sn ¯
¯ ¯.
¯ 2 − n ¯
m

3.2.13 Since −1 ≤ sin n ≤ 1, any possible limit must be finite. By way of contradiction assume that sin n → a as n → +∞. Then

lim sin n = a =⇒ lim sin(n + 2) = a,

n→+∞ n→+∞

whence
lim (sin(n + 2) − sin n) = a − a = 0.
n→+∞
Now,
sin(n + 2) − sin n = 2(sin 1) cos(n + 1) =⇒ cos(n + 1) → 0, as n → +∞.

From
cos(n + 1) = cosn cos1 − sin n sin 1

we obtain
1 1
sin n = (cosn cos 1 − cos(n + 1)) → (0 · cos1 − 0) = 0,
sin 1 sin 1
and so a = 0. But then
1 = sin2 n + cos2 n → 02 + 02 = 0,

a contradiction.

3.2.14 By problem 1.7.8, (n!)

1/n > pn for n ≥ 3. Hence, for all M > 0, as long as n > M 2 we will have

p
(n!)1/n > n > M,

giving the result.

3.2.16 We have
p p n +1−n 1 1
n +1− n = p p = p p < p .
n +1+ n n +1+ n 2 n

1 1
Hence, as long as p < ε that is, as long as n > we will have
2 n 4ε2
¯p p ¯¯ 1
¯
¯ n + 1 − n ¯ < p < ε.
2 n

3.2.17 Write
M
2X M m
2X
1 X 1
= .
n=1 n m=1 n=2m−1 +1 n

Since 1/n ≥ 1/N when n ≤ N , we gather that

m
2X m
2X
1 1
≥ 2−m = (2m − 2m−1 )2−m = .
m−1 n m−1 2
n=2 +1 n=2 +1
Thus
M
2X 1 M
≥
n=1 n 2

and the sequence can be made arbitrarily large.

3.2.18 Observe that for n ≥ 2, p p

(n − 1)! (n)!
p p p p − p p p p
(1 + 1)(1 + 2)(1p+ 3) · · ·(1 + n − 1) µ(1 + 1)(1
p +¶ 2)(1 + 3) · · ·(1 + n)
(n − 1)! n
= p p p p 1− p
(1 + 1)(1 + p2)(1 + 3) · · ·(1 + n − 1) 1+ n
(n − 1)!
= p p p p .
(1 + 1)(1 + 2)(1 + 3) · · ·(1 + n)

Therefore
p p
K
X (n − 1)! K!
p p p p =1− p p p .
n=1 (1 + 1)(1 + 2)(1 + 3) · · ·(1 + n) (1 + 1)(1 + 2) · · ·(1 + K )
p
K!
Now prove that uK = p p p decreases to 0.
(1 + 1)(1 + 2) · · ·(1 + K )

p
3.2.19 Put x 1 = 1, xn+1 = 1 + xn , n ≥ 0. We claim that the sequence {xn }+∞
n=1 is increasing and bounded above. By Theorem 165 the sequence must have a limit L . To prove that the sequence is increasing consider xn+1 − xn (fill
in this gap). To prove that the sequence is bounded, we claim that for all n ≥ 1, x n
< 4. For this is clearly true for n = 1. So assume that xn < 4. Then

p p p
xn+1 = 1 + xn < 1 + 4 = 5 < 4,

and so the assertion follows by induction.

Since we have shewn that L exists we now may compute

p
p p p 1+ 5
L= lim x = lim 1 + xn = 1 + L =⇒ L = 1 + L =⇒ L 2 − L − 1 = 0 =⇒ L = ,
n→+∞ n+1 n→+∞ 2

where we have chosen the positive root as the sequence is clearly strictly positive.

n2 + n
3.2.20 By Theorem 56, 1 + 2 + · · · + n = , and the desired result follows.
2

1 1
3.2.21 ; 1; .
3 4

169
 Answers and Hints

1
3.2.22 Put x 1 = 1, xn+1 = , n ≥ 0. We claim that the sequence {xn }+∞
n=1 is increasing and bounded above. By Theorem 165 the sequence must have a limit L . To prove that the sequence is increasing consider xn+1 − xn (fill in
1 + xn
this gap). To prove that the sequence is bounded, we claim that for all n ≥ 1, prove by induction that x n < 4 (fill in this gap).

Since we have shewn that L exists we now may compute

p
1 1 1 5−1
L= lim x = lim = =⇒ L = =⇒ L 2 + L − 1 = 0 =⇒ L = ,
n→+∞ n+1 n→+∞ 1 + xn 1+L 1+L 2

where we have chosen the positive root as the sequence is clearly strictly positive.

½ ¾+∞
an
3.2.24 Assume that is increasing. Then
bn n=1
a1 a an a
≤ 2 ≤ ··· ≤ ≤ n+1 .
b1 b2 bn bn+1

Using Theorem 79,

a1 + a2 + · · · + an an a a + a2 + · · · + an a + a2 + · · · + an+1 a
≤ ≤ n+1 =⇒ 1 ≤ 1 ≤ n+1 ,
b1 + b2 + · · · + bn bn bn+1 b1 + b2 + · · · + bn b1 + b2 + · · · + bn+1 bn+1
½ ¾ ½ ¾+∞ ½ ¾
a1 + a2 + · · · + an +∞ an an +∞
proving that is also increasing. If were decreasing, − is increasing and we apply what we just have proved.
b1 + b2 + · · · + bn n=1 bn n=1 bn n=1

3.2.26 We have
n k3 − 1
Y n k −1 Y
Y n k2 + k + 1
= .
3 k +1 2
k=2 k + 1 k=2 k=2 k − k + 1

Now
n k −1
Y (n − 1)! 2
= = .
k +1 (n+1)! n(n + 1)
k=2 2

By observing that (k + 1)
2 − (k + 1) + 1 = k 2 + k + 1, we gather that

n k2 − k + 1
Y 32 + 3 + 1 42 + 4 + 1 52 + 5 + 1 n2 + n + 1 n2 + n + 1
= · · ··· = .
k 2 +k +1 2 2 − 2 + 1 32 + 3 + 1 42 + 4 + 1 (n − 1) 2 + (n − 1) + 1 3
k=2

Thus
n k3 − 1
Y 2 n2 + n + 1 2
= · → ,
3 3 n(n + 1) 3
k=2 k + 1

as n → +∞.

1 1 1
3.2.27 Clearly x n < xn + = xn+1 , and so the sequence is strictly increasing. By shewing that xn < 2 − < 2 we will be shewing that it is bounded above, and hence convergent by Theorem 165. For n = 1, x1 = 1 = 2 − and
(n + 1)2 n 1
1
so the assertion is true. Assume that x n <2− . Then
n

1 1 1 n − (n + 1)2 n2 + n + 1 n2 + n 1
xn+1 = xn + <2− + = 2+ =2− <2− = 2− ,
(n + 1)2 n (n + 1)2 n(n + 1)2 n(n + 1)2 n(n + 1)2 n +1

and the claimed inequality follows by induction. We will prove later on a result of Euler:
1 1 1 π2
1+ + +···+ +··· = .
22 32 n2 6

3.3.1 The product rule for limits only applies to a finite number of factors. Here the number of factors grows with n .

3.3.3 From Theorem 177, and since x 7→ log x is increasing,

µ ¶ µ ¶ µ ¶ µ ¶
1 k+1 1 k+1 1 1
1+ < e < 1+ =⇒ (k + 1) log 1 + < 1 < (k + 1) log 1 + .
k +1 k k +1 k

Rearranging,
k +2 1 k +1
log < < log .
k +1 k +1 k

Summing from k = n − 1 to k = 2n − 1,
2n−1
X k +2 2n−1
X 1 2n−1
X k +1 2n + 1 1 1 1 2n
log < < log =⇒ log < + +···+ < log
k +1 k +1 k n n n +1 2n n −1
k=n−1 k=n−1 k=n−1 µ ¶ µ ¶
1 1 1 1 2
=⇒ log 2 + < + +···+ < log 2 +
n n n +1 2n n −1

and the result follows from the Sandwich Theorem.

3.3.4 Observe that µ ¶

1 1 1 1 1
1+ + + +···+ +
2µ 3 4 2n − 1 ¶ 2n
1 1 1 1
−2 + + +···+
2 4 6 2n µ ¶
1 1 1 1 1
= 1+ + + +···+ +
2 3µ 4 2n − 1 2n¶
1 1 1 1 1
−2 · 1+ + + +···+
µ 2 2 3 4 n ¶
1 1 1 1 1
= 1+ + + +···+ +
2µ 3 4 2n − 1 ¶ 2n
1 1 1 1
− 1+ + + +···+
2 3 4 n
1 1 1
= + +···+ ,
n +1 n +2 2n

and use the result of problem 3.3.3.

170
 Appendix A

3.3.7 We begin by looking at the Taylor series for e :

x
∞ xk
X
ex = .
k!
k=0
∞ 1
X ∞
This converges for every x ∈ R, so e = and e
−1 = X (−1)k 1 . Arguing by contradiction, assume ae 2 + be + c = 0 for integers a , b and c . That is the same as ae + b + ce −1 = 0.
k! k!
k=0 k=0
Fix n > |a| + |c|, then a, c | n! and ∀k ≤ n , k! | n! . Consider

∞ 1
X ∞
X 1
0 = n!(ae + b + ce −1 ) = an! + b + cn! (−1)k
k! k!
k=0 k=0
n
X n! ∞
X n!
=b+ (a + c(−1)k ) + (a + c(−1)k )
k! k!
k=0 k=n+1

Since k! | n! for k ≤ n , the first two terms are integers. So the third term should be an integer. However,
¯ ¯
¯ X∞ n! ¯¯ ∞ n!
X
¯
¯ (a + c(−1)k ) ¯ ≤ (|a| + |c|)
¯ k! ¯ k!
k=n+1 k=n+1
∞
X 1
= (|a| + |c|)
(n + 1)(n + 2) · · ·k
k=n+1
∞
X
≤ (|a| + |c|) (n + 1)n−k
k=n+1
∞
X
= (|a| + |c|) (n + 1)−t
t=1
1
= (|a| + |c|)
n

∞
X 1
is less than 1 by our assumption that n > |a| + |c|. Since there is only one integer which is less than 1 in absolute value, this means that (a + c(−1)k ) = 0 for every sufficiently large n which is not the case because
k!
k=n+1

∞
X 1 ∞
X 1 1
(a + c(−1)k ) − (a + c(−1)k ) = (a + c(−1)n+1 )
k! k! (n + 1)!
k=n+1 k=n+2
is not identically zero. The contradiction completes the proof.

p
3.3.9 Apply Problem 2.6.6 We can apply this to the stated problem by observing that for a fixed d , a positive integer without square factors, the numbers a + b d are quadratic integers if a, b are rational integers, and that the set of such
numbers is an additive group of reals. Clearly the closure of this group (it, together with its set of limit points) is a group too, for if x n → x and y n → y then x n + y n → x + y . The new group is not of form (i) or (ii), hence must be all reals, and
the proof (of a slightly stronger theorem) is complete.

³ ´ an ³ ´ an
3.5.5 an = o n 2 does, since this says that lim = 0, whereas an = O n 2 says that is bounded by some positive constant.
n→+∞ n 2 n2

an an
3.5.6 False. Take an = 2n , for example. Then an << n , = 2, and so 9 0.
n n

an
3.5.7 True. → 0 and so by Theorem 195, an << n .
n

an
3.5.8 False. Take an = n 3/2 . Then → 0 but an 6= O (n).
n2

an
3.5.9 True. → 0 and so by Theorem 195, an << n . Since n << n 2 , the assertion follows by transitivity.
n

2
4.1.1 This is a geometric series with common ratio |r | = < 1, so it converges. We have
e

23
∞
X 2n 23 24 e4 8
= + +··· = = .
n=3 e
n+1 e4 e5 1− 2e e 4 − 2e 3

4.1.2 Observe that

1 1 1
= − .
4n 2 − 1 2(2n − 1) 2(2n + 1)
Hence
+∞ µ ¶ µ ¶ µ ¶
X 1 1 1 1 1 1 1 1 1
= − + − + − +··· = = .
2 2(1) 2(3) 2(3) 2(5) 2(5) 2(7) 2(1) 2
n=2 4n − 1

tan x − tan y 1 π
4.1.3 Since tan(x − y) = , observe that arctan = arctan(n + 1) − arctan n . Hence the series telescopes to lim arctan(n + 1) − arctan 1 = .
1 + tan x tan y n2 + n + 1 n→+∞ 4

4.1.7 By unique factorisation of the integers, the desired sum is µ ¶µ ¶

1 1 1 1 1 1 1 1
1+ + + +··· 1+ + + +··· = · = 3.
2 22 23 3 32 3 3 1 1
1− 1−
2 3
X X
4.1.9 Since the sum of two convergent series is convergent by Theorem 232, if (an + bn ) then from the identity bn = (an + bn ) − an we would deduce that bn converges, a contradiction.
n≥0 n≥0
X
4.1.10 Put s N = an . There is a positive constant M such that ∀N > 0, s N ≤ M . Observe that because the terms are positive
1≤n≤N

s N +1 = s N + a N +1 ≥ s N ,

©
ª+∞
and so the sequence s N
N =1 is a monotonically increasing bounded above sequence and so it converges X
by Theorem 165.

This is not necessarily true if the series does not have positive terms. For example, the series (−1)n+1 has bounded partial sums, in fact they are either 1 or 0. But the sequence of partial sums then is
n≥1

1, 0, 1, 0, 1, 0, . . .

which does not converge.

171
 Answers and Hints

4.2.2 True. For, we must have an

2 ≤ a and the series of squares converges by direct comparison to the original series.
→ 0 and so eventually 0 < an ≤ 1. This means that eventually an n

sin x
4.2.3 True. Since an → 0, we must have sin an → an and so the series converges by asymptotic comparison to the original series. (Recall that lim = 1.)
x→0 x

tan x
4.2.4 True. Since an → 0, we must have tan an → an and so the series converges by asymptotic comparison to the original series. (Recall that lim = 1.)
x→0 x

4.2.5 False. Since an → 0, we must have cos an → 1 and so the series diverges by the n -th Term Test.

an
4.2.6 Only the fact that ≤ an is needed here.
n

1 1 +∞
X 1
4.2.7 Take an = n . Then an <<
2 n2
and
n = 1.
n=1 2

1 1
4.2.8 Take an = or an = .
2n
2 n 2n

1 1
4.2.9 Take an = n or an = n .
2 n

1 1
4.2.10 For even n ≥ 0 take an = n and for odd n ≥ 1 take an = n . Then
2 3
+∞
X +∞
X 1 +∞
X 1
an = + ,
2n 2n−1
n=0 n=0 2 n=1 3

and both series on the right are geometric convergent series. However if n is even, (an )
1/n = 1 and if n is odd (a )1/n = 1 meaning that lim (a )1/n does not exist.
n n
2 3 n→+∞

4.2.11 By the root test

µ ¶1/n
1/n = 3n 3
an = → 0 < 1,
n 2n n

and the series converges. By direct comparison, for n ≥ 3 we have

3n 3n 1 1 1
= · ≤1 n ≤ ,
n 2n nn nn n n3
∞
X 1
and the series converges by direct comparison to .
3
n=1 n

4.2.12 We divide the sum into decimal blocks. There are 9

k k -digit integers in the interval [10k ; 10k+1 [ that do not have a 0 in their decimal representation. Thus

µ ¶
X1 +∞ X X 1 +∞ X k 1
= ≤ 9 = 10.
n n 10 k
n∈S k=0 n∈[10k ;10k+1 [∩S k=0

4.2.15

e
1. an ∼ − =⇒ diverges. 6. converges iff |a| 6= 1.
2n

α 7. Converges.
2. an ∼ e n(α−2) =⇒ converges iff α < 2.
2α−1
(n − 1)(n − 1)! + n! 2
3 8. an ≤ ≤ =⇒ converges.
3. an ∼ − =⇒ converges. (n + 2)! (n + 1)(n + 2)
n2
9. Converges.
1
4. an ∼ =⇒ converges.
n2 10. an 6−→ 0 =⇒ diverges.
s
2 1
5. an ∼ =⇒ converges. 11. an = =⇒ converges.
n3 n log logn

1 1 1 1
5.1.1 Put an = µ ¶ , bn = µ ¶ for integer n ≥ 1. Then an → 0 and bn → 0, but sin → −1 and sin → +1, so the limit does not exist in view of Proposition 267.
1 1 an bn
2n − π 2n + π
2 2

p
1 + 1 + 4x
5.2.10 f (0) = 0, but for x > 0, f (x) = , so f is not right-continuous at x = 0.
2

5.6.2 Consider a unit circle and take any point P on the circumference of the circle.
Drop the perpendicular from P to the horizontal line, M being the foot of the perpendicular and Q the reflection of P at M . (refer to figure)
Let x = ∠ POM.
π
For x to be in [0, ], the point P lies in the first quadrant, as shown.
2
The length of line segment P M is sin(x). Construct a circle of radius MP , with M as the center.
Length of line segment PQ is 2 sin(x).
Length of arc P AQ is 2x .
Length of arc P BQ is πsin(x).
Since PQ ≤ length of arc P AQ (equality holds when x = 0) we have 2 sin(x) ≤ 2x . This implies
sin(x) ≤ x
π
Since length of arc P AQ is ≤ length of arc P BQ (equality holds true when x = 0 or x = ), we have 2x ≤ πsin(x). This implies
2

2
x ≤ sin(x)
π

Thus we have
2 π
x ≤ sin(x) ≤ x, ∀ x ∈ [0, ]
π 2

172
 Appendix A

5.9.1 If p had odd degree, then, by the Intermediate Value Theorem it would have a real root. Let α be its largest real root. Then

0 = p(α)q(α) = p(α2 + α + 1)

meaning that α
2 + α + 1 > α is a real root larger than the supposedly largest real root α, a contradiction.

1 1
5.9.2 Observe that f (1000) f ( f (1000)) = 1 =⇒ f (999) = . So the range of f include all numbers from to 999. By the intermediate value theorem, there is a real number a such that f (a) = 500. Thus
999 999

1
f (a) f ( f (a)) = 1 =⇒ f (500) = .
500

5.9.5

i h
5.9.10 If either f (0) = 1 or f (1) = 0, we are done. So assume that 0 ≥ f (0) < 1 and 0 < f (1) ≤ 1. Put g (x) = f (x)+ x −1. Then g (0) = f (0)−1 < 0 and g (1) = f (1) > 0. By Bolzano’s Theorem there is a c ∈ 0 ; 1 such that g (c) = 0,
that is, f (c) + c − 1 = 0, as required.

h i
5.9.11 Consider g (x) = f (x) − f (x + 1/n), which is clearly continuous. If g is never 0 in 0 ; 1 then by Corollary 321 g must be either strictly positive or strictly negative. But then

µ µ ¶¶ µ µ ¶ µ ¶¶ µ µ ¶ µ ¶¶ µ µ ¶ ³ n ´¶
1 1 2 2 3 n −1
0 = f (0) − f (1) = f (0) − f + f −f + f −f +···+ f −f .
n n n n n n n

The sum of each parenthesis on the right is strictly positive or strictly negative and hence never 0, a contradiction.

h i h i sin 2πx
a
5.9.12 Consider the function f : 0 ; 1 → 0 ; 1 , x 7→ − x.
sin 2π
a

6.2.1 Observe that that

1 1 (x + 1) − (x − 1) 2
− = = .
x −1 x +1 (x − 1)(x + 1) x2 − 1

If f (x) = (x − 1)−1 then

f 0 (x) = −1(x − 1)−2 ; f 00 (x) = (−1)(−2)(x − 1)−3 ; (−1)(−2)(−3)(x − 1)−4 ; . . .; f (100) (x) = 100!(x − 1)−101 .

Similarly, if g (x) = (x + 1)
−1 then

g 0 (x) = −1(x + 1)−2 ; g 00 (x) = (−1)(−2)(x + 1)−3 ; (−1)(−2)(−3)(x + 1)−4 ; . . .; g (100) (x) = 100!(x + 1)−101 .

Hence
d100 2
= f (100) (x) − g (100) (x) = 100!(x − 1)−101 − 100!(x + 1)−101 .
dx 100 x 2 − 1

6.2.2 We use Leibniz’s Rule and the observation that the third derivative of x 7→ x 2 is 0. Also (sin x)(4n) = sin x , (sin x)(4n+2) = − sin x , (sin x)(4n+1) = cosx , and (sin x)(4n+3) = − cos x , Then

! Ã Ã ! Ã !
d100 2 100 2 100 2 0 100 2 00
x sin x = x (sin x)(100) + (x ) (sin x)(99) + (x ) (sin x)(98) = x 2 sin x − 200x cosx − 9900 sin x.
dx 100 0 1 2

6.3.1 Put f (x) = x 5 − 2x 2 + x . Then f (0) = f (1) = 0 and by Rolle’s Theorem there is c ∈]0; 1[ such that f 0 (c) = 5c 4 − 4c + 1 = 0.

6.3.2 Set
a x2 a x3 an x n+1
f (x) = a0 x + 1 + 2 +···+ ,
2 3 n +1
and use Rolle’s Theorem.

i h
6.3.4 Set g (x) = f (x)2 f (1 − x). Since g (0) = g (1) = 0, g satisfies the hypotheses of Rolle’s Theorem. There is a c ∈ 0 ; 1 such that

g 0 (c) = 0 =⇒ 2 f 0 (c) f (c) f (1 − c) − f (c)2 f 0 (1 − c) = 0.

Since by assumption f (c) f (1 − c) 6= 0 we must have, upon dividing by every term by f (c)2 f (1 − c), the assertion.

hk k +1 i i k k +1 h
6.3.5 For 0 ≤ k ≤ n − 1, consider the interval ; . By the Mean Theorem, there are ak ∈ ; such that
n n n n
µ ¶ µ ¶
k +1 k
f −f µ µ ¶ µ ¶¶
n n k +1 k
f 0 (ak ) = =n f −f .
1 n n
n

Summing from k = 0 to k = n − 1 and noting that the dextral side telescopes,

n−1 n−1 µ µ ¶ µ ¶¶
X 0 X k +1 k
f (ak ) = n f −f = n( f (1) − f (0)) = n.
n n
k=0 k=0

h i i
6.3.6 Let ki ∈ 0 ; 1 be the smallest number such that f (ki ) = , 1 ≤ i ≤ n − 1. Put k0 = 0, kn = 1. The existence of the ki is guaranteed by the Intermediate Value Theorem. Moreover, since the ki are chosen to be the first time f is
n
i
, once again, by the Intermediate Value Theorem we must have
n
0 < k1 < k2 < · · · < kn−1 < 1.
i h
Hence, by the Mean Value Theorem, there exists ai ∈ ki ; ki +1 , 0 ≤ i ≤ n − 1, such that

f (ki +1 ) − f (ki ) 1 1
f 0 (ai ) = = =⇒ = n(ki +1 − ki ).
ki +1 − ki n(ki +1 − ki ) f 0 (ai )

Summing,
n−1
X 1 n−1
X
=n (ki +1 − ki ) = n(kn − k0 ) = n.
0
k=0 f (ak ) k=0

173
 Answers and Hints

µ ¶1/e
1
6.4.2 We have f 0 (x) = x x (log x + 1) whence f 0 (x) = 0 =⇒ x = e −1 . Since f 0 (x) < 0 for 0 < x < e −1 and f 0 (x) > 0 for x > e −1 , x = e −1 is a local (relative) minimum. Thus f (x) ≥ f (e −1 ) = .
e

6.5.3 Let 0 < k < 1, and consider the function

[0; +∞[ → R
f : .
x 7→ x k − k(x − 1)

1 ap
Then 0 = f 0 (x) = k x k−1 − k ⇔ x = 1. Since f 00 (x) = k(k − 1)x k−2 < 0 for 0 < k < 1, x ≥ 0, x = 1 is a maximum point. Hence f (x) ≤ f (1) for x ≥ 0, that is x k ≤ 1 + k(x − 1). Letting k = and x = we deduce
p bq

µ p ¶
a 1 a
≤ 1+ −1 .
b q/p p bq

Rearranging gives
a p b 1+p/q−p b 1+p/q
ab ≤ b 1+p/q + −
p p

from where we obtain the inequality.

6.6.1 We have:

1. Put f : R → R, f (x) = e x−1 − x. Clearly f (1) = e 0 − 1 = 0. Now,

f 0 (x) = e x−1 − 1,

f 00 (x) = e x−1 .

If f 0 (x) = 0 then e x−1 = 1 implying that x = 1. Thus f has a single minimum point at x = 1. Thus for all real numbers x

0 = f (1) ≤ f (x) = e x−1 − x,

which gives the desired result.

2. Easy Algebra!

3. Easy Algebra!

4. By the preceding results, we have

A 1 ≤ exp(A 1 − 1),

A 2 ≤ exp(A 2 − 1),

.
.
.

A n ≤ exp(A n − 1).

Since all the quantities involved are positive, we may multiply all these inequalities together, to obtain,

A 1 A 2 · · · A n ≤ exp(A 1 + A 2 + · · · + A n − n).

In view of the observations above, the preceding inequality is equivalent to

nn Gn
≤ exp(n − n) = e 0 = 1.
(a1 + a2 + · · · + an )n

We deduce that µ ¶
a1 + a2 + · · · + an n
Gn ≤ ,
n
which is equivalent to
a1 + a2 + · · · + an
(a1 a2 · · · an )1/n ≤ .
n
Now, for equality to occur, we need each of the inequalities A k ≤ exp(A k − 1) to hold. This occurs, in view of the preceding lemma, if and only if A k = 1, ∀k , which translates into a1 = a2 = . . . = an . This completes the proof.

6.7.1 (loglog x)log x = exp((log x)(logloglog x)) and (log x)log log x = exp((loglog x)2 ). Now, lexicographically,

(loglog x)2 << (log x)(logloglog x) =⇒ exp((loglog x)2 ) << exp((log x)(logloglog x))

and thus (loglog x)

log x is faster.

7.1.1 ⇐ This follows directly from Theorem 437.

=⇒ If f is Riemann integrable, let ε > 0 and let P 0 = {a = y 0 < y 1 < . . . < y m = b} be a partition with m + 1 points such that

ε
U ( f , P 0 ) − L( f , P 0 ) < .
2
h i ¯ ¯ ε ¯¯ ¯¯
¯¯ ¯¯ 00 0
As f is bounded, there is M > 0 such that ∀x ∈ a ; b , ¯ f (x)¯ ≤ M . Take δ = and consider now an arbitrary partition P = {a = x 0 < x 1 < . . . < x n = b} with norm ¯¯P ¯¯ < δ. Put P = P ∪ P . Arguing
8mM
as in Theorem 434, we obtain ¯¯ ¯¯
¯¯ ¯¯ ε
L( f , P 00 ) − L( f , P ) < 2mM ¯¯P ¯¯ < 2mMδ = .
4

Since by Theorem 435 L( f , P

0 ) ≤ L( f , P 00 ) we gather
ε
L( f , P 0 ) − L( f , P ) < .
4
In a similar fashion we establish that
ε
U ( f , P ) −U ( f , P 0 ) < ,
4
and upon assembling the inequalities,
ε
U ( f , P ) − L( f , P ) < U ( f , P 0 ) − L( f , P 0 ) + < ε,
2

since we had assumed that U ( f , P

0 ) − L( f , P 0 ) < ε .
2

174
 Appendix A

7.1.2 =⇒ Assume f is Riemann-integrable. For ε > 0 let δ > 0 be chosen so that the conditions of Theorem ?? be fulfilled. By definition of a Riemann sum,

L( f , P ) ≤ S( f , P ) ≤ U ( f , P ),

and therefore
Zb Zb
U ( f , P ) < L( f , P ) + ε ≤ f (x)dx + ε = f (x)dx + ε
a a

and
Zb Zb
L( f , P ) > U ( f , P ) − ε ≥ f (x)dx − ε = f (x)dx − ε.
a a
These inequalities give ¯ ¯
¯ Zb ¯
¯ ¯
¯S( f , P ) − f (x)dx ¯ < ε,
¯ a ¯

Zb
whence ¯¯ lim
¯¯ S( f , P ) = f (x)dx.
¯¯ ¯¯ a
¯¯P ¯¯→0

¯¯ ¯¯
¯¯ ¯¯
⇐ Suppose that ¯¯ lim
¯¯ S( f , P ) = L , existing and finite. Given ε > 0 there is δ > 0 such that ¯¯P ¯¯ < δ implies
¯¯ ¯¯
¯¯P ¯¯→0

ε ε
L− < S( f , P ) < A + . (A.7)
3 3
h i
Now, choose P = {a = x 0 < x1 < · · · < xn = b}. By letting tk range over xk−1 ; xk we gather, from (A.7)

ε ε
L− ≤ L( f , P ) ≤ U ( f , P ) ≤ L + ,
3 3

whence
2
U ( f , P ) − L( f , P ) ≤ ε < ε,
3
h i
meaning that f is Riemann-integrable over a ; b by Theorem 437. Thus
Zb
L( f , P ) ≤ f (x)dx ≤ U ( f , P ),
a
Zb
and so ¯¯ lim
¯¯ S( f , P ) = f (x)dx .
¯¯ ¯¯ a
¯¯P ¯¯→0

h i
7.1.3 =⇒ Let P = {a = x0 < x1 < · · · < xn = b} be a partition of a ; b . Set

n
X h i
Z ( f , P) = ω( f , xk−1 ; xk )(xk − xk−1 ) = U ( f , P ) − L( f , P ), Ω= sup f (x) − inf f (x).
k=1 x∈[a;b] x∈[a;b]

Let
n
X
δ= (xk − xk−1 )χ{x∈[a;b]:ω( f ,[x 0 .
k=1 k−1 ;xk ])≥ε }

0
Then Z ( f , P ) ≥ δε . Since we are assuming that f is Riemann-integrable, there exists a partition P (by Theorem 437) such that

Z ( f , P ) ≤ ε0 ε.

Thus we have δε
0 < εε0 from where δ < ε.

⇐ Assume there is a partition P for which δ < ε. In the intervals I = [xk−1 ; xk ] where ω( f , I ) ≥ ε0 the oscillation of f is at most Ω, and in the remaining intervals (the sum of which is b − a − δ, the oscillation is less than ε0 . Hence

Z ( f , P ) ≤ δΩ + (b − a − δ)ε0 .

Choose now
ε00 ε00
ε0 = , δ= .
2(b − a) 2Ω

Since b − a − δ ≤ b − a ,
ε00 ε00
Z ( f , P ) ≤ δΩ + (b − a − δ)ε0 ≤ + = ε00 ,
2 2
whence f is Riemann-integrable by Theorem 437.

8
7.2.1
5

7.2.2
Z3 Z1 Z2 Z3
x Tx U dx = x Tx U dx + x Tx U dx + x Tx U dx
0 0Z Z 1 Z 2
1 2 3
= 0 x dx + 1 x dx + 2 x dx
0 1 2
x 2 ¯¯2 ¯3
¯
= ¯ + x2 ¯
2 1 2
1
= (2 − ) + (9 − 4)
2
13
= .
2

7.2.3 We have
f (x + h) − f (x) e x+h − h − e x e x+h − e x h
f 0 (x) = lim = lim = lim − lim = e x − 1,
h→0 h h→0 h h→0 h h→0 h

whence f (x) = e x − x + C . Since 3 = f (0) = e 0 − 0 + C =⇒ C = 2, we deduce that f (x) = e x − x + 2.

175
 Answers and Hints

Za
1
7.2.4 Put I = dx . We have
0 f (x) + 1
Za Za Za Z0 Za
1 f (u) f (a − u) f (a − u) f (v) f (u)
I= du = du = du = − dv = du,
0 f (u) + 1 0 f (u) + f (u) f (a − u) 0 1 + f (a − u) a 1 + f (v) 0 1 + f (u)
whence Za Za Za
f (u) f (a − u) 2 + f (u) + f (a − u)
2I = du + du = du = a,
0 1 + f (u) 0 1 + f (a − u) 0 2 + f (u) + f (a − u)
a
and so I = .
2

7.2.5 Observe first that f (0 + 0) = f (0) + f (0) and so f (0) = 0. Integrate f (u + y) = f (u) + f (y) for u ∈ [0; x]keeping y constant, getting

Zx Zx Zx Zx
f (u + y)du = f (u)du + f (y)du = f (u)du + x f (y).
0 0 0 0

Also, by substitution,
Zx Z y+x Z y+x Zy
f (u + y)du = f (u)du = f (u)du − f (u)du.
0 y 0 0

Hence Z y+x Zy Zx
x f (y) = f (u)du − f (u)du − f (u)du. (A.8)
0 0 0
Exchanging x and y :
Z y+x Zx Zy
y f (x) = f (u)du − f (u)du − f (u)du. (A.9)
0 0 0
f (x) f (y) f (x)
From (A.8) and (A.9) we gather that x f (y) = x f (y). If x y 6= 0 then = . This means that for is constant, and so for x 6= 0, f (x) = c x for some constant c . Since f (0) = 0, f (x) = c x for all x . Taking x = 1, f (1) = c .
x y x

7.2.7 We have
Z2 Z1 Z2
|x 2 − 1| dx = (1 − x 2 ) dx +(x 2 − 1) dx
−1 −1 1
3
x ¯1¯ x 3 ¯2
¯
= (x − )¯ + ( − x)¯
3 −1 3 1
1 8 1
= 2(1 − ) + ( − 2) − ( − 1)
3 3 3
4 2 2
= + +
3 3 3
8
=
3
q
dx xdx udu
7.2.16 Put u = x 2 − 1; u 2 = x 2 − 1 so that 2udu = 2xdx and = = . Thus
x x2 u2 + 1

Z Z Z q
1 u 1
p dx = du = du = arctan u + C = arctan x2 − 1 + C .
x x2 − 1 (u 2 + 1)u u2 + 1

p
7.2.17 Put u = x + 1; u 2 = x + 1; from where dx = 2udu . Whence

Z Z Zµ ¶
p p
1 2u 2
p dx = du = 2− du = 2u − 2 log |1 + u| + C = 2 1 + x − 2 log|1 + 1 + x| + C .
1+ x +1 1+u 1+u

7.2.18 Put x = u 6 ; dx = 6u 5 du , giving

Z Z
x 1/2 (u 3 )(6u 5 )
dx = du
x 1/2 − x 1/3 3 2
Z u 6− u
6u
= du
Zuµ− 1 ¶
1
= 6 u5 + u4 + u3 + u2 + u + 1 + du
à u −1 !
u6 u5 u4 u3 u2
= 6 + + + + + u + log |u − 1| + C
6 5 4 3 2
6x 5/6 3x 2/3
= x+ + + 2x 1/2 + 3x 1/3 + 6x 1/6 + 6 log|x 1/6 − 1| + C .
5 2

7.2.19 Put u
2 = a x + 1; 2udu = (log a)a x dx and so

Z Z Z
a 2x 2u(u 2 − 1) 2u 2 − 2 2u 3 2u 2(a x + 1)3/2 2(a x + 1)1/2
p dx = du = du = − +C = − +C.
ax + 1 u log a log a 3 log a log a 3 loga log a

7.2.20 Observe that (e

x − e −x )2 = (e −x (e 2x − 1))2 = e −2x (e 2x − 1)2 , and so

Z Z Z
1 e 2x 1 1 1
dx = dx = du = − +C = − +C,
(e x − e −x )2 (e 2x − 1)2 2u 2 2u 2(e 2x − 1)

on putting u = e
2x − 1.

7.2.21 We have
Z5 Z2 Z3 Z4 Z5
bxc bxc bxc bxc bxc
dx = dx + dx + dx + dx
1 x 1
Z2 x Z3 2 x Z4 3 x Z5 4 x
1 2 3 4
= dx + dx + dx + dx
1 x 2 x 3 x 4 x
= (log2 − log1) + 2(log3 − log2) + 3(log4 − log3) + 4(log5 − log4)

= 4 log(5) − 3 log(2) − log(3).

7.2.22 Put u = e x , etc. Z Z Z

x x x x
e e +x dx = e x e e dx = e e de x = e e + C

176
 Appendix A

7.2.23 Put u = log(cosx), etc.

Z Z
(log(cos x))2
tan x log(cos x)dx = (log(cosx))d(− log(cos(x))) = − +C
2

7.2.24 Put u = log log x , etc.

Z Z
loglog x ¡ ¢ log log x
dx = loglog xd loglog x = +C
x log x 2

7.2.25 Carry out the long division.

Z 18 Z
x −1 x 16 x 13 x 10 x7 x4
dx = (x 15 + x 12 + x 9 + x 6 + x 3 + 1)dx = + + + + + x +C
x3 − 1 16 13 10 7 4

7.2.26 After an algebraic trick, put u = 1 + x −7 , etc.

Z Z Z
1 x −8 1 d(1 + x −7 ) 1
dx = dx = − = − log|1 + x −7 | + C
x8 + x 1 + x −7 7 1 + x −7 7

7.2.27 Put u = 2x + 1 Z Z Z
2x 2x 1 2x 1 u −1 1 ¡ ¢ 1 ¡ x ¢
dx = d(2x + 1) = du = u − log |u| + C = 2 + 1 − log|2x + 1| + C
2x + 1 log2 2x + 1 log 2 u log2 log2

7.2.28 Put u = x + 1. Then x 2 = (u − 1)2 = u 2 − 2u + 1, and hence

Z Z 2
x2 u − 2u + 1
dx = du
(x + 1)10 Z u 10
= u −8 − 2u −9 + u −10 du
u −7 u −8 u −9
= − + − +C
7 4 9
(x + 1) −7 (x + 1)−8 (x + 1)−9
= − + − +C
7 4 9

7.2.29 Algebraic trick, and then u = e

−x + 1, etc.
Z Z Z
1 e −x 1
dx = dx = − d(e −x + 1) = − log|e −x + 1| + C
1 + ex e −x + 1 e −x + 1

7.2.30 Z Z Z Z
1 1 + sin x 1 + sin x
dx = dx = dx = sec2 x + sec x tan xdx = tan x + sec x + C
1 − sin x 1 − sin2 x cos2 x

7.2.31 Zp Zq
1 + sin 2xdx = sin2 x + 2 sin x cos x + cos2 xdx
Zq
= (sin x + cosx)2 dx
Z
= | sin x + cosx|dx

= ∓ cos x ± sin x + C

7.2.32 Put u = x 2 , etc. Z Z

x 1 1 1 1
q dx = p du = arcsin u + C = arcsin x 2 + C
2 1 − u2 2 2
1 − (x 2 )2

7.2.33 We have Z Z
sec4 xdx = sec2 x(tan2 x + 1)dx
Z Z
= sec2 x tan2 xdx + sec2 xdx
Z Z .
= (tan x)2 d(tan x) + sec2 xdx
tan3 x
= + tan x + C .
3

7.2.34 We have Z Z
sec5 xdx = sec3 x sec2 xdx
Z
= sec3 xd(tan x)
Z
= sec3 x tan x − tan xd(sec3 x)
Z
= sec3 x tan x − 3 tan2 x sec2 x sec xdx
Z
= sec3 x tan x − 3 (sec2 x − 1) sec3 xdx
Z Z
= sec3 x tan x − 3 sec5 xdx + 3 sec3 xdx

The above implies that

Z Z
tan x sec3 x 3
sec5 xdx = + sec3 xdx
4 4
3
tan x sec x 3 tan x sec x 3
= + + log| sec x + tan x| + C ,
4 8 8
upon recalling from class that Z
tan x sec x 1
sec3 xdx = + log| sec x + tan x| + C
2 2

7.2.35 First put t = x 1/3 , then t 3 = x =⇒ 3t 2 dt = dx . Thus

Z Z
1/3
ex dx = 3t 2 e t dt

= 3t 2 e t − 6te t − 6e t + C
1/3 1/3 1/3
= 3x 2/3 e x − 6x 1/3 e x − 6e x +C,

where the penultimate step results from tabular integration by parts.

177
 Answers and Hints

7.2.36 We have Z Z
log(x 2 + 1)dx = x log(x 2 + 1) − xd(log(x 2 + 1))
Z
x2
= x log(x 2 + 1) − 2 dx
2
Z x2 + 1
x + 1 − 1
= x log(x 2 + 1) − 2 dx
Zµ x 2 + 1 ¶
2 1
= x log(x + 1) − 2 1− dx
x2 + 1
= x log(x 2 + 1) − 2(x − arctan x) + C

7.2.37 Put Z
I= xe x cos x := (Ax + B )e x cos x + (C x + D)e x sin x + K .

Differentiating both sides,

xe x cos x = Ae x cosx + (Ax + B )e x cosx − (Ax + B )e x sin x + C e x sin x + (C x + D)e x sin x + (C x + D)e x cos x.

Equating coefficients,

xe x cosx : 1 = A +C

xe x sin x : 0 = −A + C

e x cosx : 0 = A +B +D

e x sin x : 0 = −B + C + D

1 1 1 1 1
From the first two equations C = , A = . Then the third and fourth equations become − = B + D; − = −B + D , whence D = − , and B = 0. We conclude that
2 2 2 2 2

Z µ ¶
x x x −1 x
xe x cosx = e cos x + e sin x + K .
2 2

7.2.38 We will do this one two ways: first, by making the substitution
t = log x =⇒ e t = x =⇒ e t dt = dx.

Observe also that x

2/3 = e 2t/3 . Then
Z Z
x 2/3 log xdx = te 2t/3 e t dt
3t 5t/3 9 5t/3
= e − e +C
5 25
3(logx) 5/3 9 5/3
= x − x +C.
5 25
Aliter: By directly integrating by parts,
Z Z Ã !
3x 5/3
x 2/3 log xdx = log xd
5
Z
3x 5/3 3
= log x − x 5/3 d(log x)
5 5 Z
3(log x) 5/3 3
= x − x 2/3 dx
5 5
3(log x) 5/3 9 5/3
= x − x +C,
5 25
as before.

7.2.39 This integral can be done multiple ways. For example, you may integrate by parts directly and then “solve” for the integral. Another way is the following. Start by putting

t = log x =⇒ e t = x =⇒ e t dt = dx.

Then Z Z
sin(log x)dx = e t sin tdt,

an integral that we found in class. We will find it again, using a method similar of problem 7.2.37. Put

Z
I= e t costdt := Ae t cos t + B e t sin t + K .

Differentiating both sides

e t cos t = Ae t cost − Ae t sin t + B e t sin t + B e t cos t.

Equating coefficients,

e t cos t : 1 = A +B

e t sin t : 0 = −A + B

1
and so A =B = . We have thus
2 Z Z
sin(log x)dx = e t sin tdt
1 t 1
= e cos t + e t sin t + K
2 2
1 1
= x cos log x + x sin log x + K .
2 2

t t
7.2.40 Put t = loglog x =⇒ e e = x =⇒ e t e e dt = dx . Hence
Z Z t
loglog x te t e e
dx = dt
x t
ee
= te t − e t + C

= (log x)(loglog x) − (log x) + C ,

where the penultimate equality follows from a tabular integration by parts.

178
 Appendix A

7.2.41 Observe that

Z Z Z
sec x tan x + sec2 x ¡ ¢
sec xdx = dx = d log(tan x + sec x) = log(tan x + sec x) + C ,
tan x + sec x

For the second way, simple algebra will yield the identity. We have
Z Z Z
cos x cos x
sec xdx = dx + dx
2(1 + sin x) 2(1 − sin x)
1 1
= log|1 + sin x| − log|1 − sin x| + C .
2 ¯ 1 + sin x ¯ 2
1 ¯ ¯
= log ¯ ¯+C
2 1 − sin x

For the third way, we have

Z Z
1
csc xdx = dx
Z sin x
1
=
2 sin x x dx
2 cos 2
Z cos x
= 2 dx
2 sin x 2 x
2 cos 2
Z sec2 x
= 2 dx
2 tan x2
u=tan x Z
2 du
=
u
x
= log | tan | + C .
2
Thus Z Z Z ¯
π π π ¯ π x ¯¯
sec xdx = csc( + x)dx = csc( + x)d( + x) = log ¯ tan( + )¯ + C .
2 2 2 4 2

7.2.42 Putting t = arcsin x we have

sin t = x =⇒ costdt = dx,

whence Z Z
(arcsin x)2 dx = t 2 costdt

= t 2 sin t + 2t cos t − 2 sin t + C

= (arcsin x)2 x + 2(arcsin x) cos(arcsin x) − 2x + C

q
= (arcsin x)2 x + 2(arcsin x) 1 − x 2 − 2x + C

7.2.43 We have
Z Z p p
dx ( x + 1 − x − 1)dx
p p =
x +1+ x −1 2 .
1 1
= (x + 1)3/2 − (x − 1)3/2 + C
3 3

7.2.44 We have à !
Z Z
x2
x arctan xdx = arctan xd
2
Z
x2 x2
= arctan x − d(arctan x)
2 2
Z
x2 1 x2 .
= arctan x − dx
2 2 1 + x2
Z
x2 1 x2 + 1 − 1
= arctan x − dx
2 2 1 + x2
x2 x 1
= arctan x − + arctan x + C
2 2 2

p
7.2.45 Put u = tan x and so u 2 = tan x , 2udu = sec2 xdx = (tan2 x + 1)dx = (u 4 + 1)dx . Hence the integral becomes

Z Z
p u2
tan xdx = 2 du.
u4 + 1

To decompose the above fraction into partial fractions observe (Sophie Germain’s trick) that u
4 + 1 = u 4 + 2u 2 + 1 − 2u 2 = (u 2 + u p2 + 1)(u 2 − u p2 + 1) and hence

Z Z
p u2
tan xdx = 2 du
p u 4Z+ 1 p Z
2 u 2 u
= − p du + p du
2 u2 − u 2 + 1 p
p2 u +u 2+1 p2 p
2 p 2 p 2 p 2 p
= − log(u 2 + u 2 + 1) + log(u 2 − u 2 + 1) + arctan( 2u + 1) − arctan(− 2u + 1) + C
p4 4 p 2 2
2 p 2 p
= − log(tan x + 2 tan x + 1) + log(tan x − 2 tan x + 1)
4 p p4
2 p 2 p
+ arctan( 2 tan x + 1) − arctan(− 2 tan x + 1) + C
2 2

7.2.46 Put
2x + 1 A B C
= + + =⇒ 2x + 1 = Ax(x − 1) + B (x − 1) + C x 2 .
x 2 (x − 1) x x2 x −1

Letting x = 1 we get 3 = C . Letting x = 0 we get 1 = −B =⇒ B = −1. To get A observe that equating the coefficients of x 2 on both sides we get 0 = A + C , whence A = −3. Thus

Z Z Z Z
2x + 1 1 1 1
dx = −3 dx − dx + 3 dx
x 2 (x − 1) x x2 x −1
1
= −3 log|x| + + 3 log|x − 1| + C
¯ ¯x
¯ x −1 ¯ 1
= 3 log ¯ ¯+ +C.
x x

179
 Answers and Hints

7.2.47 Integrating by parts,

Z Z
p p p
log(x + x)dx = x log(x + x) − xd log(x + x)
1
Z x(1 + p )
p 2 x
= x log(x + x) − p dx
Zµ x + x p ¶
p 1 x
= x log(x + x) − 1− · p dx
Z2 p x+ x
p 1 x
= x log(x + x) − x + p dx
2 x+ x
p Z
u= x p u2
= x log(x + x) − x + du
2
p Z u +u
u= x p 1
= x log(x + x) − x + 1 − du
u +1
p
= x log(x + x) − x + u − log(u + 1) + C
p p p
= x log(x + x) − x + x − log( x + 1) + C

7.2.48 We use Sophie Germain’s trick to factor

p p
x 4 + 1 = x 4 + 2x 2 + 1 − 2x 2 = (x 2 + 1)2 − 2x 2 = (x 2 − 2x + 1)(x 2 + 2x + 1),

and seek the partial fraction decomposition

1 Ax + B Cx +D p p
= p + p =⇒ 1 = (Ax + B )(x 2 + 2x + 1) + (C x + D)(x 2 − 2x + 1).
x4 + 1 x 2 − 2x + 1 x 2 + 2x + 1
Equating coefficients

x3 : 0 = A +C
p
x2 : 0 = B + D + 2(A − C )
p
x : 0 = A + C + 2(B − D)

x0 : 1 = B +D

1 1
From the first and third equation it follows that A = −C and that B = D . From the fourth equation B = D = and from the second equation A = − p = −C . Hence we must integrate
2 2 2

Z Z p Z p
1 2x + 2 2x − 2
dx = p dx − p dx
x4 + 1 p 4(x
Z
2 + 2x + 1)
p 4(x
Z
2 − 2x + 1)
p Z p Z
2 2x + 2 1 1 2 2x + 2 1 1
= p dx + p dx − p dx + p dx
2 2 2 x 2 − 2x + 1
p8 x + 2x + 1 p4 x + 2x + 1 8
Z x − 2x + 1 Z
4
2 p 2 p 1 dx 1 dx
= log(x 2 + x 2 + 1) − log(x 2 − x 2 + 1) + p + p
p8 p8 2
p (x 2 + 1)2 + 1 2p (−x 2 + 1)2 + 1
2 p 2 p 2 p 2 p
= log(x 2 + x 2 + 1) − log(x 2 − x 2 + 1) + arctan(x 2 + 1) − arctan(−x 2 + 1) + C
8 8 4 4

7.2.49 We begin by observing that

1 A Bx +C
= + =⇒ 1 = A(x 2 − x + 1) + (B x + C )(x + 1).
x3 + 1 x + 1 x2 − x + 1

1 2 2 1
Letting x = −1 we obtain 1 = 3A =⇒ A = . Letting x = 0 we obtain 1 = A + C =⇒ C = 1 − A = . Finally, we must have A + B = 0, since the coefficient of x must be zero. thus B = − . We must then integrate
3 3 3

Z Z Z x− 1 Z
dx x −2 1 2 1 1
− dx = log |x + 1| − +
3(x + 1) 3(x 2 − x + 1) 3 1 )2 + 3
3(x − 2 2 (x − 1 )2 + 3
4 Z 2 4
1 1 1 3 2 1
= log |x + 1| − log|(x − )2 + | +
3 6 2 4 3 4 (x − 1 )2 + 1
p3 2
1 1 1 3 2 3 1
= log |x + 1| − log|(x − )2 + | + · arctan(x − )
3 6 2 4p 3 2 2
1 1 3 2 1
= log |x + 1| − log|x 2 − x + 1| + arctan p (x − )
3 6 3 3 2

8.8.1

8.8.2

8.8.3

8.8.4

180
Bibliography

[Apo] Apostol, T. M., Calculus, Vol 1 & 2, 2nd ed., Waltham: Xerox, 1967.

[Har] Hardy, G. H., Pure Mathematics, 10th ed., New York: Cambridge University Press, 1952.

[Kla] Klambauer, Gabriel, Aspects of Calculus, New York: Springer-Verlag, 1986.

[Lan] Landau, E., Differential and Integral Calculus, New York, Chelsea Publishing Company, 1950.

[Olm] Olmstead, J. M. H., Calculus with Analytic Geometry, Vol 1 & 2, New York: Appleton-Century-Crofts, 1966.

[Spi] Spivak, Michael Calculus, 3rd ed., Houston, Texas: Publish or Perish, Inc., 1994.

181