This action might not be possible to undo. Are you sure you want to continue?
an introduction to mathematical thought process
daniel solow
east' westf'rn rf'St'nf' university
john wUey &: sons
new york chichester brisbane toronto singapore
it is hoped that we,., of this book
will kindly remove (lnd return the questionnaire
ill('lude(1 as its last page. to tile address that
('all bf> left visible after folding and stapling
('opyri(da' © 19B2 by johll IfJUt>y & .0 .... , illC'.
rf>protlu("ioll or '", .... Indoll 01 tilly pan 01 'his lfJork
bt'yolltl ,ht" pf>rmit't'tl by .t'('rio .... 107 or lOB 01 1M
1976 Ullitf>tl .'a't'. ('opyri(da' acl lfJithoUl ,he permit.oll
01 'M ('opyrilJla' olfJlU'r is un/alfJjiJ  rt'qut'.,. lor
pt'rmit.ioll or fur,hf>r illlo""",ioll .hould bt' atltirt'''t'tl 10
,/w pt'rmit.iOll. tlt'pantlle .. " johll UJUt>y & .0'1$, illC'.
isb .. 0471 S6645 S
10 9
to my late father • • •
ANATOLE A. SOWW
and to my mother • • •
RUTH SOLOW
foreword
In a related article, "Teaching Mathematics with
Proof Techniques," the author has \vritten, "'The
inability to communicate proofs in an understandable
manner has plagued students and teachers in all
branches of mathem8tics." All of those who have had
the exrerience of teaching mathematics and most of
those who had the eXferience of trying to learn
it must surely agree that acquiring an understanding
of what constitutes a sound mathematical is a
major stumbling block for the stunent. stunents
attempt to circumvent the obstacle by avoiding
ittrusting to the indulgence of the examiner not to
inclune any proofs on the test. 'This collusion
between stunent ano teacher mey avoid some of the
unpleasant consequencesfor both student and
teacherof the student's lack of mastery, but it
does not aJter the fact that a key element in
methematics, arguably its most characteristic
feature, has not entered the rerertoire.
rr. Solow believes that it is possible to teach
the student to understann the nature of a proof by
systematizing it. He argues his case cogently with a
wealth of netail and in this pamphlet, and I
do not doubt that his ideas deserve attention,
discussion, and, above all, Gne of
his principal aims is to teach the student how to
the proofs offered in textbooks. These proofs
are, to be sure, not fresented in a systematic form.
Thus, much attention is paidparticularly in the two
append icesto showing the reader how to recogni ze
the standard ingredients of a mathematical argument
in an informal rresentation of a froof.
There is a valid analogy here with the role of
the traditional algorithms in elementary arithmetic.
v
vi
FCREWORD
It is important to acquire familiarity with them and
to understand why they work and to what problems they
could, in principle, be applied. Eut once all of
this has been learned, one would not slavishly
execute the algorithms in reallife situations (even
in the absence of a calculator!). So, the author
contends, it is with proofs. Understand and analyze
their structureand then you will be able to rean
and understand the more informal versions you will
find in textbooksand finally you will be able to
create your own proofs. Dr. Solow is not claiming
that mathematicians creAte their proofs by
consciously and deliberately applying the
·forwardbackward methoo·; he is suggesting that we
have a far better chance of teaching an appreciation
of proofs by systematizing them than by our present,
rather haphazard, procedure baseo on the hOFe that
the stucents can learn this difficult art by osmosis.
One must agree with Dr. Solow that, in this
country, students begin to grapple with the ideas of
mathematical proof far too late in their student
careersthe appropriate stage to be into
these ideas is, in the judgement of many, no later
than eighth grade. However, it would be wrong for
university and college teachers merely to excuse
their own failures by a comforting reference to
defects in the student's precollege education.
Today, mathematics is genera] ly recognized as a
subject of fundamental importance because of its
ubiquitous role in contemporary life. To be userl
effectively, its methods must be understood
properlyotherwise we cast ourselves in the roles of
(inefficient) robots when we try to use mathematics,
and we place undue strain on our naturally imferfect
memor ies. Dr. Solow has given much thought to the
question of how an understanding of a mathematical
proof can be acquired. students tocay do not
acquire that understanding, and Cr. Solow'S to
remeoy this very unsatisfactory situation deserves a
fair trial.
Louis D. Eeaumont University Professor
Case Western Reserve
Clevelanc'l, Ohio
rETFR flILTON
to the student
finishing my undergraduate degree, I began
to wonder why learning theoretical mathematics had
been so difficult. As I progressed through my
graduate work, I realized that mathematics possessed
many of the aspects of a game: a game in which the
rules had been partially concealed. Imagine trying
to play chess before you know how all of the pieces
move! It is no that so many students have had
trouble with abstract mathematics.
This pamphlet describes some of the rules by
which the game of theoretical mathematics is played.
It has been my experience that virtually anyone who
is motivated and who has a of high school
can learn these rules. roing so will
greatly the time (and frustration) involveC' in
learning abstract I hope this pamphlet
serves that purrose for you.
To play chess, you must first learn how the
individual rieces move. Only after these rules have
entered your subconscious can your mind turn its full
attention to the more creative issues of strategy,
tactics, and the like. So it appears to be with
mathematics. Hard work is required in the beginning
to learn the fundamental rules presented in this
pamphlet. In fact, your goal should be to absorb
this material so that it becomes second nature to
you. Then you will fine that your mind can focus on
the creative aspects of mathematics. These rules are
no substitute for creativity, and this pamphlet is
not meant to teach creativity. I do believe
that it can rrovide you with the tools needed to
express your creativity. Equally important is the
fact that these tools will enable you to understand
and to appreciate the creativity of others.
vii
vIII
'10 THE STUt'ENT
You are about to learn a key part of the
mathematical thought process. 'As you study the
material and solve problems, be conscious of your
thought process. Ask questions and seek answers.
Remember, the only unintelligent question is the one
that goes unasked.
Cleveland, Ohio
June 1981
SOLOW
to the instructor
The inability to communicGlte proofs in an
understandable manner has plagued students and
teachers in all branches of mathematics. result
has been frustrated frustrated teachers,
and, oftentimes, a "watered down" course to enable
the students to follow at least some of the material,
or a test that protects students [rom the
consequences of this deficiency in their mathematical
understand ing.
One might conjecture that most students simply
cannot understand abstract mathematics, but my
experience indicates otherwise. seems to have
been lacking is a prorer method for exrlaining
theoretical mathematics. In this pamphlet I have
developed a method for communicating proofs: a
common language can be taught by professors and
understood by students. In essence, this pamphlet
categorizes, icentifies, and explains (Glt the
student's level) the various techniques that are used
repeatedly in virtually all rroofs.
Cnce the students understand the techniques, it
is then possible to explain any proof as a sequence
of applications of these techniques. In fact, it is
advisable to do so because the process reinforces
what the students have in the pamphlet.
Explaining a proof in terms of its component
techniques is not difficult, as is illustrated in the
examples of this pamrhlet. Pefore each "condensed"
proof is an outline of the proof explaining the
methodology, thought process, and techniques that are
being used. Teachin9 proofs in this manner requires
nothing more than Freceding each step of the proof
ix
x
TO THr
with an indication of which technique is about to be
used, and why.
When discussing a Froof in class, I actively
involve the students by soliciting their heIr in
choosing the techniques and in <'lesigning the proof.
I have been pleasantly surprised by the quality of
their comments and It has been my
exrerience that once the students become comfortable
with the proof techniques, their minds tend to
address the more imrortant issues of
such as why a Froof is done in a particular way and
why the riece of mathematics is important in the
first place. This pamrhJet is not meant to teach
creativity, but r do believe that it does describe
many of the necessary underl ying ski lIs whose
aCQuisition will free the stucent's mind to focus on
the creative aspects. I have also found thtlt, by
using this approach, it is possible to teach
subsequent matheMatical at a much more
sorhisticaten level without lOSing the students.
In any event, the message is clear. I am
suggesting that there ere many benefits to be gained
by teaching mathematical thought process in addition
to mathematical material. This pamFhlet is designed
to be a major step in the right direction by making
abstract mathematics and enjoyable to
the students and by providing you with a Method for
communicating with them.
Cleveland, Chio
ISel
C'.lIt!IEL SOLOW
acknowledgments
For helping to get this work known in the
mathematics community, my deepest gratitude goes to
Peter Bilton, an outstanding mathematician and
educator. r also thank Paul .. a1mos, whose timely
recognition and support greatly facilitated the
dissemination of the knowledge of the existence of
this pamphlet and teaching method. I am also
grateful for discussions with Gail Young and Ceorge
Fol ya.
Regarding the preparation of the manuscript, no
single person had more constructive comments than 10m
Butts. He not only contributed to the mathematical
content but also corrected many of the grammatical
and stylistic mistakes in a version. I
suppose that I should thank his mother for being an
Fnglish teacher. I also Charles Wells
for reading commenting on the first handwritten
draft and for encouraging me to pursue the project
further. other people substantive
suggestions, inclue ing Jl.lan SChoenfelr., Samuel
Goldberg, and Fllpn Stenson.
Of all the people involved in this project, none
deserve more credit than my students. It is because
of their voluntary efforts that this parnrhlet has
been prepared in such a short time. 1hanks
especially to John Cemocko [or acting in the capacity
of senior editor while concurrently trying to
complete his Ph.!:'. program. Also, I appreciate the
help that r received from Creiling and Robert
Wenig in data basing the text ann preparing the
exercises. Michael has worked on this project almost
as long as r have. "spec ial word of thanks CJoes to
Greg for coordinating the secone rewriting of
xi
xi i
.a C II'N 01'11 LEI:GNI ENTS
the document, for adding useful comments, and
general, for keeping me very organized.
responsibilities have subsequently been assumed
Robin Symes.
in
Vis
by
In addition, I am grateful to F:avi f(umar for the
long hours he spent on the computer preparing the
final version of the manuscrirt, to Fetty Tracy and
Bognar for their professional anc flawless
typing assitance, and to Virginia Penade for her
technical editing. I am also indebted to my class of
for preparing the soJutions to the
r acknowlefoge Cary Ostedt, the mathematics
editor of John t·riley and for his help in
speeding the publication rrocess. I thank the
following professors for refereeing the manuscript
anc for recommending its publication: Tucker,
I:avic Singer, Poward lnton, anc Niven.
Last but not least, r am grateful to my wife,
.audrey, for her heIr in proof reading and for her
patience yet another of My projects.
['. s.
contents
1. The Truth of It
2. The ForwardBackward 8
3. On Cefinitions and Terminology 23
Cuantifiers Part I: The Construction 34
5. Cuantifiers rart II: The Choose 40
6. Cuantifiers Part III: Induction 50
7. Cuantifiers Part IV: Specialization SS
The Contradiction
9. The ContraFositive 72
1(,. t1ots or Nots Lead to Knots 78
II. Ppecial Proof Techn iques
12. 8umrnary
Appendix Putting It J\11 Together:
Append ix El: rutt ing It HI Together:
Solutions to rxercises
Glossary of Mathemetica1 Symbols
Index
Part I
Fart II
83
101
1C'g
ll7
11'7
IG9
xIIi
xlv
CONTENTS
TABLES
Table 1. The Truth
of" "
Impl ies E" 5
Table 2. Proof of Example 1 15
Table 3. The Truth Table for
"NOT B Implies NOT A" 31
Table 4. Summary of Proof Techniques 96
DEFINITIONS
Definitions 1  10 24
Definition 11 24
Defini tion 12 35
Definition 13 ~ 5
Definition 14 t.2
Definition 15 42
Definition 16
6('
Definition 17
1 ('II
Definition H ~ 109
how to read and do proofs
an introduction to mathematical thought process
1 the truth of it aU
The objective of mathematicians is to discover
ane to communicate certain truths. is
the language of mathematicians, and a proof is a
method of communicating a mathematical truth to
another Ferson who also "speaks" the language. A
remarkable property of the language of mathematics is
its precision. Properly presented, a proof will
contain no ambiguity: there will be no doubt as to
its correctness. Unfortunately, many proofs that
appear in textbooks and journal articles are not
presented properly; more appropriately stated, the
rroofs are rresented properly for someone who already
knows the language of mathematics. Thus, to
understand and/or present a proof, you must learn a
new language, a new methoo of thought. 'This pamphlet
explains much of the basic "gramMar" you will need,
but as in learning any new language, a lot of
practice on your part will be needed to become
fl uent.
The approach of this pamphlet is to categorize
and to explain the techniques that used
in proofs. One objective is to teach you how to read
and how to understand a written proof by ieentifying
the techniques that have been used. Learning to do
so will enable you to study almost any mathematical
subject on your own, a desirable goal in itself.
second objective of this is to teach
you to develop and to communicate your own proofs of
known mathematical truths. Coing so requires that
you use a certain amount of creativity, intuition,
and experience. Just as there are many ways to
express the same idea in any language, so are there
different proofs for the same mathematical fact. 'The
1
2
THE TRUTH OF IT ALL
proof techniques presented here are designed to get
you started and to guide you through a proof.
Consequently, this pamphlet describes not only how
the proof techniques work, but also, when each
technique is likely to be used, and why. It is often
the case that a correct technique can be chosen based
on the very form of the problem under consideration.
Therefore, when attempting to create your own proof,
learn to select a proof technique consciously before
wasting hours trying to figure out what to 00. 'the
more aware you are of your thought process, the
bet te r i tis.
The ultimate objective, however, is to use your
newly acquired skills and language to discover anf to
communicate previously unknown mathematical truths.
While the goal is an admirable one, it is extremely
difficult to attain. The first step in this
direction is to reach the level of being able to read
proofs and to develop your own rroofs of already
known facts. This alone will give you a much deeper
and richer understanding of the matheMatical universe
around you.
The basic material on proof techniques is
presented in the next eleven chapters. The twelfth
chapter is a complete summary and it is followed by
two appendices that illustrate the various techniques
with several examples.
The pamphlet is designed to be read by. anyone
with a good knowledge of high school mathematics.
Advanced students who have seen proofs before can
read the first two chapters, skip to the summary
chapter, and subsequently read the two appendices to
see how all of the techniques fit together. The
remainder of this chapter explains the types of
relationships to which proofs can be applied.
Given two statements A a n ~ E, each of which may
be either true or false, a fundamental problem of
interest in Mathematics is to show that if ~ is true
then E is true. ,. proof is a formal method for
accomplishing this task. As you will soon discover,
the particular form of ,. and E can often inficate a
way to proceed. Some exaJllples of statements are:
THE TRUTH OF IT ALL
1. Two different lines in a plane are
either parallel or else they intersect
in exactly one point.
'J. 1 = o.
'2. '2x = 5 and y :II 1.
x is not> O.
5. There is an angle t such that
cos(t) = t.
3
Observe that statement (1) is always true, (2) is
always false, and statements (3) and (4) can be
either true or false depending on the value of a
variable.
It is perhaps not as obvious that statement (5)
is always true. It therefore becomes necessary to
have some meth00 for "proving" that such statements
are true. In other words, a mathematical proof is a
convincing argument that is expressed in the language
of mathematics. Thus, a proof should contain enough
mathematical detail so as to be convincing to the
person(s) to whom the proof is addressed. For
instance, a proof of statement (5) aimed at
convincing a mathematics professor might consist of
nothing more than Figure 1. Cn the other hand, a
proof directed toward a high school student would
probably require greater detail, perhaps even the
nefinition of cosine. It is the lack of sufficient
detail that can often make a proof difficult to read
and to understand. cne objective of this pamFhlet is
to teach you to decipher these "condensed" rroofs
that are likely to appear in textbooks and other
mathematical literature.
In order to do a proof, you must know exactly
what it means to show that "if is true then E is
true." The statement A is often called the
and B the For brevity, the statement "if
A is true then P is true" is shortened to "if then
8" or simply"' implies E." are often
very lazy when it comes to writing. such, they
have developed a symbolic "shorthand." For instance,
a mathematician would often write => B" instead or
"A implies E." For the most part, textbooks do not
use the symbolic notation, but teachers often do, and
eventually you may find it useful too. Therefore
4
THF TRUTH OF IT ALL
this document will include the aFproFriate symbols
but will not use them in proofs. complete list of
the symbols can be found in the glossary at the end
of this pamphlet.
Fig. 1. proof that there is an angle t such that
cos(t) ::: t.
It seemS reasonable that the conditions under
wh ich ",.. impl ies E" are true wi 11 depend on whether
and I? themselves are true. Consequently, there are
four possible cases to consider:
1. A is true and E is true.
2. A is true and B is false.
A is false and B is true.
4. is false and E is false.
Suppose, for example, that your friend made the
statement: "rf it rains then brings her
umbrella." Here, the statement is "it rains" and E
is "Mary brings her umbrella." To determine When the
statement "1 implies B" is false, ask yourself in
which of the four cases would you be willing to call
your friend a liar. In the first caSe (i.e., when it
does rain and does bring her your
fr iend has told the truth. In the second case, it
has rained, and yet did not bring her umbrella,
THE TRUTH OF IT ALL 5
as your friend said she would. Pere your friend has
not told the truth. Finall y, in cases (3) an(1 (/I),
it does not rain. You would not really want to call
your friend a liar in the case of no rain because,
your friend only said that something happen if
it did rain. Thus, the statement imrlies E" is
true in each of the four cases except the secon(1 one,
as summarized in Table l.
Tablo I is an example of a table. truth
table is a method for determining when a comrlex
statement (in this case, "l. implies P") is true by
examining all rossible truth values of the in(1ividual
statements (in this case, an(1 E). ether examples
of truth tables will appear in Chapter 3.
According to Table I, when trying to show that
"A impl ies E" is true, you can assume that the
statement to the 1 eft 0 f the word "imrl ies· (namel y
11) is true. Your goal is to concl ude that the
statement to the right (namely E) is true. Note that
a proof of the statement "A imrlies E" is not an
attempt to verify whether A anc F themselves are
true, but rather, to show that B is a logical result
of hav ing assumed that is true.
In general, your ability to show that n is true
will oepend very on the fact that you have
assumed to be true, ultimately, you will have
to discover the linking relationship between ann B.
I:'oing so will require a certain amount of creativity
on your part. The rroof techniques rresented here
Table 1. The Truth of ",.. Illlrlies P"
True
True
False
false
B
True
False
True
False
II IJI'lFl ies E
True
False
'!rue
True
*************************************
6
THE TRUTH OF IT ALL
are designed to get you started and guide you along
the path.
Hereafter, and E will be statements which are
either true or false. The problem of interest will
be that of showing implies E."
Exercises
1.1. of the following ere statements? (Recall
that a must be either true or false.)
(a) ax
2
+ bx + c = C
Cb) (b + " b
2
 /2a
(c) Triangle YYZ is simil?r to PST.
«(I) 3 + n + n
2
(e) sin(Jt/2) < sin(Jt/Ll)
(f) For every angle t, sin2(t) + cos
2
(t) = 1.
1.2. For each of the following rrob!ems, irentHy the
hypothesis the conclusion.
(a) If the right triangle XYZ with siees of
lengths x and y, and hypotenuse of length z,
has an area of z2/., then the triangle XY:
i s i so sc e 1 e s •
(b) n is an even integer => n
2
is an even
integer.
(c) If a, b, c, c, e, and f are real numbers
the prorerty that ad  be 0, then tho
two linear equations ax + by = e and
cx + = f Cen he for x end y.
(<") The sum of the first n positive integers is
n(n + 1)/2.
(e) r is real and satisfies r2 = 2 imrlies r is
irrational.
(f) If P Q are positive real numbers
ypq F (r + q)/? then F q.
rq) x is A real number, the value
of x(x  1) is at least 1/
4
•
1.3. If you are trying to frove thC'lt "I il'lrl jes P" is
true ant! you know thClt r is fa] se, do you want
to show that l is true or false? Fxr1ain.
THE TRUTH OF IT ALL
7
1.4. Us ing Table 1, c'letermine the cone i tions. under
which the following statements are true or
fal se , and give your reason.
(a) If 2 > 7 then 1 > ::.
( b) If '2 < 7 then
]
< 3.
( c) If x :3 then
]
<
2.
(d) If x
=
3 then 1 > 2.
1.5. Prepare a truth t a b l ~ for each of the following
statements.
(a) A implies (B implies C).
(b) (,P. impJ ies P) imrl ies C.
2 the forwardbltckward method
The purpose of this chapter is to describe one
of the fundamental rroof techn j ques: the 6oltwa'td
backwaltd method. fpecial emrhasis is 0iven to
material of this chaFter because all of the other
proof techn iquE's wi 1 J USE> the forwar(lbackwarc
methof.
The first steF in proof requires recognizing
the statements /'l anI" I? In general, everything after
the word "if" and before the "then" cOMprises
statement I, whilE' everything after the word "then"
constitutes statement r. everything
that you are assuming to be true (Le., the
hypothesis) is "i everything you are trying to
Frove (i.e., the conclusion) is r. Consider the
following excmr1e.
Example 1. If the right trianqle YYZ with sides of
lengths x an,' y, imd hypotenuse of length z, has .::In
area of then the triangle YYZ is isosceles (see
Figure ?).
x
x Y
Fig. 2. The ri':lht triangle XYZ.
8
9
Outline of proof.
statements:
In this example one has the
A: right triangle XYZ with sides of
lengths x and y, and hypotenuse of length z
has an area of z2/ 4•
8: The triangle XYZ is isosceles.
Recall that, when proving "A implies e," you can
assume that A is true and you must somehow use this
information to reach the conclusion that B is true.
In attempting to figure out just how to reach the
conclusion that B is true, you will be going through
a Cn the other hand, when you make
specific use of the information contained in A, you
will be going through a Eoth of
these processes will be described in detail.
In the hackward process you begin hy asking "How
or when can r conclude that the statement P is true?"
The very manner in which you phrase this question is
critical since you must eventually be able to answer
it. The question should be posed in an abstract way.
For Example 1, the correct abstract question is "How
can I show that a triangle is isosceles?" it is
true that you want to show that the particular
triangle XYZ is isosceles, by asking the abstract
question, you calIon your general knowledge of
triangles, clearing away irrelevant (such as
the fact that the triangle is called XYZ instead of
thus allowing you to focus on those aspects of
the problem that really seem to matter. The question
obtained from statement B in such problems will be
called the A properly posed
abstraction question should contain no symbols or
other notation from the specific problem under
consideration. The key to many proofs is formulating
a correct abstraction question.
In any event, once
have posed the
abstraction question, the next steF in the hackward
process is to answer it. Returning to the example,
how can
show that a triangle is isosceles?
one way is to show that two of its sides
10
FCRWARrBACRWARD
have equal length. Referring to Figure 2, you should
show that x  y. Cbserve that answering the
abstraction question is a phase process. First
you give an abstract answer: to show that d triangle
is isosceles, show that two of sides have equal
length. Next, you apply this answer to the specific
situation: in this case, to show that two of its
sides have equal length means to show that x = y, not
that x = z or y = z. The process of asking the
abstraction question, answering it abstractly, and
then applying that answer to the specific situation
will be referred to as the
The abstraction process has given you a new
statement, El, with the property that if you could
show that Bl is true then B would be true. For the
example above, the new statement is:
81: x = y
If you can show that x = y, then the triangle XYZ is
isosceles. Once you have the statement El, all of
your efforts must now be directed toward reaching the
conclusion that El is true, for then it will follow
that F is true. How can you show that Bl is true?
Eventually you will have to make use of the
assumption that A is true, and when solving this
problem, you would most likely do so now, but for the
moment, let us continue working backwarc by repeating
the abstraction process on the new statement Pl.
This will illustrate some of the difficulties that
can arise in the backward process. Can you pose the
new abstraction question?
Since x and yare the lengths of two sides of a
triangle, a reasonable abstraction question would
appear to be "How can I show that the lengths of two
sides of a triangle are equal?" second perfectly
reasonable abstraction cuestion would be "How can I
show that two real numbers are equal?" all, x
and yare also real numbers. Cne of the difficulties
that can arise in the abstraction process is the
possibility of more than one abstraction question.
Choosing the correct one is more of an art than a
science. In fortunate circumstances, there will be
11
only one obvious abstraction question. In other
cases, you may have to proceed by trial and error.
This is where your intuition, insight, creativity,
experience, diagrams, and graphs can play an
important role. One general guideline is to let the
information in A (which you are assuming to be true)
help you to choose the question, as will be done in
this case.
Regardless of which question you finally settle
on, the next step will be to answer it, first in the
abstract and then in the specific situation. Can you
do this for the two abstraction questions above? ror
the first one, you might show that two sides of a
triangle have equal length by showing that the angles
opposite them are equal. For the triangle XYZ of
Figure 2, this would mean that you have to show that
angle X equals angle Y. A cursory examination of the
contents of statement A does not seem to provide much
information concerning the angles of triangle XYZ.
For this reason, the other abstraction question will
be chosen.
Now one is faced with the question "How can I
show that two real numbers (namely x and y) are
equal?" One answer to this question would be to show
that the difference of the two numbers is c.
this answer to the specific statement Bl
means you would have to show that (x  y) = O.
Unfortunate1y, there is another perfectly acceptable
answer: show that the first number is less than or
equal to the second number and also that the second
number is less than or equal to the first number.
Applying this answer to the specific statement Fl,
you would have to show that x s y and y s x. Thus, a
second difficulty can arise in the backward process.
Fven if you choose the correct abstraction question,
there may be more than one answer to it.
you might choose an answer that will not permit you
to complete the proof. For instance, associated with
the abstraction question "How can I show that a
triangle is isosceJes?" is the answer "Show that the
triangle is equilateral." Of course it will be
to show that triangle XYZ of Example 1 is
equilateral, since one of its angles is 90 degrees.
12
Returning to the abstraction question "How can I
show that two real numbers (namely x and y) are
equal?" suppose, for the sake of argument, that you
choose the answer of showing that their difference is
O. Cnce again, the abstraction process has given you
a new statement, 82, with the property that if you
could show that 82 is true, then in fact PI would be
true, and hence so Specifically, the new
statement is:
82: x  y = 0
Now all of your efforts must be directed toward
reaching the conclusion that 82 is true. You must
ultimately make use of the information in P, but for
the moment, let us continue once more with the
abstraction process applied to the new statement P2.
One abstraction question is "How can r show that
the difference of two real numbers is At this
point it may seem that there is no reasonable answer
to this question. Yet another problem can arise in
the abstraction process. The abstraction question
might have no apparent answer! Do not despair: all
is not lost. Remember that when proving "A implies
B" you are allowed to assume that A is true. Nowhere
have you made use of this fact. It is time to do so
through the forward process.
The forward process involves starting with the
statement which you assume to be true and deriving
from it some other statement AI, which you know to be
true as a result of A being true. It should be
emphasized that the statements derived from A are not
haphazard. Rather, they are directed toward linking
up with the last statement derived in the backward
process. This last statement should act as the
guiding light in the forward process. Let us return
to Fxample l, keeping in mind the fact that the last
statement obtained in the backward process was
"x  y = 0."
For the example above, the statement is "The
right triangle XYZ with sides of length x and y, and
hypotenuse of length z, has an area of z2/ 4." Cne
fact that you know (or should know) as a result of A
FCRWJ.lRI:'EJlCKWARD
13
being true is that xy/2 = z2/ 4, because the area of a
right triangle is onehalf the base times the height,
in this C('lse xy/2. So you have obtained the neW
statement:
Another statement follows froM ].I by the
Pythagorean theorem, so you also have:
A2: (x
2
+ y2) = z2
The forward can also combine and use the new
statements to produce more true statements. For
instance, it is possibJe to combine and ].12 by
replacing z2 in '1 with (x
2
+ y2) from A2 obtaining
the statement:
A3: xy/2
One of the problems with the forward process is
that it is also possible to generate some useless
statements, for instance, "angle X is less than 90
degrees." there are no specific guidelines for
producing new statements, keep in mind the fact that
the forward process is directed toward obtaining the
statement P2: x  y = 0, which was the last one
derived in the backward process. It is for this
reason that z2 was eliminated from ].I] anc A7.
Continuing with the forward process, you should
attempt to rewrite so as to make it look more like
E2. For instance, you can multiply both of 1:
by t and subtract ?xy from both siees to obtain:
A4: (x
2
_ 2xy + y2) ('
By factoring you can obtain:
AS: (x  y)2 "" 0
Cne of the most common steps of the forward process
is to rewrite statements in different forms, as was
done in obtaining,.t (1n(1 '5. For Fxam[le], the
final step in the forward (and in the entire
14
FeRlA1 R[)
proof) is to take the square root of both sides of
the equality in A5, thus obtaining precisely the
statement E2: x  y = O. The proof is now complete
since you started with the assumption that A is true
and used it to derive the conclusion that E2, and
hence S, is true. 'Ihe steps and reasons are
summClrized in Table 2.
rt is interesting to note that the forward
process ultimately produced the elusive answer to the
abstraction question associated with P:?: "Pow can r
show that the difference of two real numbers is
r?"which was to show that the square of the
is 0 (see in Table 2).
Finally, you shoulr.'1 realize that, in general, it
will not be practical to write clown the entire
thought process that goes into a proof, for this
woulr'l requi re fClr too much time, effort, and space.
Father, a highly version is usually
presented and often makes little or no reference to
the backware process. ror the rroblem above it might
go something like this.
Proof of Example 1. FrOM the hypothesis the
formula for the area of a right triangle, the area of
XYZ = xy/2 = Py the Pythagorean theorem,
(x
2
+ y2) = Z2, anCl on substituting (x
2
+ y2) for z2
and performing some algebraic manipulations one
obtains (x  y)2 = r. Pence x = y and the triangle
XYZ is isosceles.11 (The "II" or some equivalent
symbol is usually used to indicate the enCl of the
proof. Sometimes the letters C.E.C. are used as they
stand (or the Latin words quod
meaning was to be demonstrated.")
Sometimes the shortened
backward anCl partly jorward.
proof will be partly
For example:
Proof of Example 1. The statement will be proved by
establishing that x = y, which in turn is done by
showing that (x  y)2 = (x
2
 + y2) = O. But the
area of the triangle is (l/:?)xy = (1/4)z2, so that
2xy = z2. p.y the Pythagorean theorem, z2 = (x
2
+ y2)
and hence (x
2
+ y2) = 2xy, or (x
2
 :?xy + y2) = C, as
required.11
FORWARDBACKWARD
15
Table 2. Proof of Example 1
* * * * * * * * * * * * * * * * ~ * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Statement
A : Area of XYZ is z2/4
~ 1 : xy/2 == z2/ 4
A2: x
2
+ y2 = z2
A3: xy/2 = (x
2
+ y2)/4
A4: x
2
 2xy + y2 = 0
AS: (x  y)2 == 0
B2: (x  y) = 0
B1: x == y
B : XYZ is isosceles
Reason
Given
Area == Cbase) (height)/2
Pythagorean theorem
Substitute A2 into Al
Algebra
Factoring M
Take square root in AS
Add Y to both sides of e2
Since 81 is true
*****************************************************
The proof can also be written entirely from the
backward process. Although this version is slightly
unnatural, nonetheless, it is worth seeing.
Proof of Example 1. To reach the conclusion, it will
be
2
shown that x = y by verifying t ~ a t (x _ y)2 =
(x  2xy + y2) = 0, or equivalently, that (x
2
+ y2)
= 2xy. This can be established by showing that
2xy = z2, for the Pythagorean theorem states 2that
(x
2
+ y2) = z2. Tn order to see that 2xy = z , or
equivalently, that (1/2)xy = C1/4)z2, note that
(1/2)xy is the area of the triangle and it is equal
to (l/4)z2 by hypothesis, thus completing the
proof.//
Proofs found in research articles are often very
condensed, giving little more than a hint of how to
do the proof. For example:
18
Proof of Example 1. The hypothesis tOjether with the
Pythagorean theorem yield (x
2
+ y ) • 2xy, hence
(x  y)2 • O. Thus the triangle is isosceles as
required.11
Note that the word whence
w
effectively conceals
the reason that (x  y) • o. Was it algebraic
manipulation (as we know it was) or something else?
Unfortunately, these shortened versions are typically
given in mathematics books, and it is this fact that
makes proofs so hard to read. You should strive
toward the ability to read and to dissect a condensed
proof. To do so, you will have to figure out which
proof technique is being used (since the
forwardbackward method is not the only one
available). Then, from what is written, you will
have to discover the thought process that went into
the proof and, finally, be able to verify all of the
steps involved. The more condensed the proof, the
harder this process will be. Some examples of how to
read condensed proofs appear in the two appendices.
In this pamphlet, your life will be made
substantially easier because an outline describing
the proof technique, methodology, and reasoning that
was involved will precede each condensed proof. Out
of necessity, these outlines will be more succinct
than the one given in Example 1.
A summary of the forwardbackward method for
proving implies B
W
is in order. Begin with the
statement e, which you are trying to conclude is
true. Through the abstraction process of asking and
answering the abstraction question, derive a new
statement, Bl, with the property that if Bl is true,
then so is B. efforts are now directed toward
establishing that Bl is true. To that end, apply the
abstraction process to Bl obtaining a new statement,
B2, with the property that if B2 is true, then so is
Bl (and hence B). Remember that the abstraction
process is motivated by the fact that A is assumed to
be true. Continue in this manner until either you
obtain the statement (in which case the proof is
finished) or until you can no longer pose andlor
answer the abstraction question fruitfully. In the
latter case, it is time to start the forward process,
in which you derive a sequence of statements from A
17
that are necessarily true as a result of A being
assumed true. Remember that the goal of the forward
process is to obtain precisely the last statement you
had in the backward process, at which time you will
have successfully completed the proof.
The forward and backward processes can easily be
remembered by thinking of the statement B as a needle
in a haystack. When you work forward from the
assumption that A is true, you start somewhere on the
outside of the haystack and try to find the needle.
In the backward process, you start at the needle and
try to work your way out of the haystack toward the
statement A (see Figure 3) •
...
,.. .. ' ,
A "
/0=
I Forward • B ,\
( \
Fig. 3. Finding a needle in a haystack.
Another way of remembering the forwardbackward
method is to think of a in which is the
starting point and B is the desired ending point (see
Figure 4). It might be necessary to alternate
several times between the forward and backward
processes before you succeed, for there are likely to
be several false starts and blind alleys.
18 FCRWARCBACKWARD
As a general rule, the forwardbackward method
is probably the first technique to tryon a problem
unless you have reason to use a different approach
based on the form of B, as will be described shortly.
In any case, you will gain much insight into the
relationship between ~ and 8.
A Forward
•
I
I I
I ~ I I


~

U
B Backward
•

~ 
~



F19 4. The maze.
FORWARDBACKWARD 19
Exercises
Note: All proofs should contain an outline of proof
as well as a condensed version.
2.1. Explain the difference between the forward and
backward processes. Cescribe how each one works
and what can go wrong. How are the two
processes related to each other?
2.2. Consider the problem of proving that If x is a
real number, then the maximum value of
x
2
+ 2x + I is ~ 2." Which of the following
abstraction questions is incorrect, and why?
(a) How can I show that the maximum value of a
parabola is ~ to a number?
(b) How can I show that a number is s to the
maximum value of a polynomial?
(c) How can I show that the maximum value of the
function _x
2
+ 2x + I is ~ to a number?
(d) How can I show that a number is s to the
maximum of a quadratic function?
2.3. Consider the problem of showing that "If
R = {real numbers x: x
2
 x s e},
S .. {real numbers x: fxl) (x3) ~ OJ, and
T = {real numbers x: x ~ I},
then R intersect S
the following is
question, and why?
the other choices.
is a subset of T.  ~ ' h i c h of
the most correct abstraction
Explain what is wrong with
(a) How can I show that a set is a subset of
another set?
(b) How can I show that the set R intersect S is
a subset of T?
(c) How can I show that every point in R
intersect Sis ~ to l?
(d) How can I show that the intersection of two
sets has a point in common with another set?
2.4. For each of the following problems, list as many
abstraction questions as you can (at least two).
20
Be sure that your questions contain no symbols
or notation from the specific problem.
(a) If 1 and 1 are the tangent lines to a
circle C at \he two endpoints e, and e.
2
of a
diameter d, respectively, then 1, and 12 are
parallel.
(b) If f and 9 are continuous functions then the
function f + g is continuous.
(Note: Continuity is a property of a
function.)
(c) If n is an even integer then n
2
is an even
integer.
(d) If n is a given integer that satisfies
3n
2
+ 2n + 8 = 0, then 2n2  3n = 2.
2.5. For each of the following abstraction questions,
list as many answers as you can (at least
three).
(a) How can r show that two real numbers are
equa l?
(b) How can I show that two triangles are
cong ruent?
(c) How can I show that two lines are parallel?
(d) How can r show that a quadrilateral is a
rectangle?
2.6. For each of the following problems, (1) pose an
abstraction question, (2) answer it abstractly,
and (3) apply your answer to the specific
problem.
(a) If a, b, and c are real numbers for which
a > 0, b < P, and b
2
 4ac = C, then the
solution to the equation ax
2
+ bx + c = 0 is
positive.
(b) In the following c:'Iiagram, if SU is a
perpendicular bisector of RT, and RS = 2RU,
then triangle RST is equilateral.
21
2.7. For each of the following hypotheses, list as
many statements as you can (at least three) that
are a result of applying the forward process for
2.8.
precisely one step. 2
(a) The real number x satisfies x  3x + 2 < O.
(b) The sine of X in triangle XYZ of
Figure 2 is 1/,2.
(c) The circle C consists of ail values 2 for x
and y that satisfy (x  3) + (y  2) = 25.
(d) The triangle UVW is equilateral.
Consider the problem of proving
are real numbers such that x
2
+
y2 + X = 3, then Iyl = 2." Tn
from the hypothesis, which of
not val id, and why?
(a) y2 = 3  x
(b) y2 == 25/6  (XLi6)2
(c) (3  y2)2 +  25 == 0
((I) ex + 5) == 6y /(x  5)
that "If x and y
6y2  25 == 0 and
working forward
the following is
2.9. Consider the problem of proving that "Tf x and y
are nonnegative real numbers that satisfy
x + y == 0, then x = 0 and y = 0."
(a) For the following condensed proof, write an
outline of the proof indicating the forward
and backward steps, and the abstraction
questions and answers.
Proof. First it will be shown that x s 0,
for then, since x 0 by the hypothesis, it
must be that x == O. To see that x S 0, by
the hypothesis, x + y = 0, so x == yo Also,
since y (\, it follows that y s 0 and
hence x == y s ('\. rinally, to see that
y == 0, since x = 0 and x + y = 0, it must be
that 0 + y = y = 0.//
(b) Rewrite the condensed proof of part (a)
entirely from the backward process.
2.10. Consider an alphabet consisting of the two
letters .6 and t, together wi th the following
rules for creating new words from old ones. The
rules.can be applied in any order.
22
1. Double the current word (e.g.,
could become ) •
2. Erase tt from the current word (e.g.,
could become
Replace in the current word by t
(e.g., could become
Add the letter t at the ri9ht end of
the current word if its last letter is
• (e.g., could become
(a) Use the forward process to derive all of
the possible words that can be obtained in
three steps by repeatedly appl ying the
above rules to the initial word
(b) Apply the backward process one step to the
word Specifically, list all of the
words for which an application of one of
the above rules would result in
(c) Prove that "If then
(d) Prove that "If then
2.11. Prove that if the right triangle XYZ of
Figure 2 is 2isosceles, then the area of the
triangle is z /4.
2.12. Prove that the statement in 2.6(b) is true.
3 on definitions and mathematical terminology
In the previous chapter you learned the
forwardbackward method and saw the importance of
formulating and answering the abstraction question.
One of the simplest yet most effective ways of
answering an abstraction question is through the use
of a definition, as will be explained in this
chapter. In addition, you will learn some of the
"vocabulary" of the language of mathematics.
A de6inition is nothing more than a statement
that is agreed on by all parties concerned. You have
already come across a definition in Chapter 1. There
we defined what it means for the statement "A implies
8" to be true. Specifically, we agreed that it is
true in all cases except when A is true and e is
false. Nothing says that you must accept this
definition as being correct. If you choose not to,
then we will be unable to communicate regarding this
particular idea.
refinitions are not made randomly. Usually they
are motivated by a mathematical concept that occurs
repeatedly. In fact, a definition can be viewed as
an abbreviation that is agreed on for a particular
concept. 'rake, for example, the notion of "a
positive integer greater than one that is not
divisible by any positive integer other than one and
itself ," that is abbreviated (or defined) as a
"prime." Surely it is easier to say "prime" than "a
positive integer greater than one ,"
especially if the concept comes up frequently.
Several other examples of definitions would be:
23
24 DEFINITIONS ANO TERMINOLOGY
Definition 1. An integer n divides an integer m
(written nlm) if m· kn for some
integer k.
Definition 2. A positive integer p > 1 is prime if
the only positive integers that divide
pare 1 and p.
Definition 3. A triangle is isosceles if two of its
sides have equal length.
Definition 4. Two pairs of real numbers
(x
2
'Ya_) are equal if
~ • Y2·
and
and
Definition 5. An integer n is even if and only if its
remainder on division by 2 is c.
Definition 6. An integer n is odd if and only if
n • 2k + 1 for some integer k.
Definition 7. A real number r is a rational number if
and only if r can be expressed as the
ratio of two integers p and q in which
the denominator q is not o.
Definition 8. Two statements A and e are equivalent
if and only if A implies e and B
implies A.
Definition 9. The statement A AND B (written A 1\ B)
is true if and only if ~ is true and B
is true.
Definition 10. The statement A OR B (written A \I B)
is true in all cases except when A is
false and B is false.
Observe that the words if and only if have
been used in some of the definitions, but, in
general, if tends to be used instead of if and
only if. Some terms, such as set and point, are
left undefined. One could possibly try to define a
set as a collection of objects, but to do so is
impractical because the concept of an object is too
DEFINITIONS AND TERMINOLOGY
25
vague. One would then be led to ask for the
definition of an "object," and so on, and so on.
Such philosophical issues are beyond the scope of
this document.
In the proof of Example 1, a definition was
already used to answer an abstraction question.
Recall the very first one, which was "How can I show
that a triangle is isosceles?" Using Definition 3, in
order to show that a triangle is isosceles, one shows
that two of its sides have equal length. Definitions
are equally useful in the forward process. For
instance, if you know that an integer n is odd, then
by Definition 6 you would know that n = 2k + 1 for
some integer k. Using definitions to work forward
and backward is a common occurrence in proofs.
It is often the case that there are two possible
definitions for the same concept. Take, for example,
the notion of an even integer that was introduced in
Definition 5. A second plausible definition for an
even integer is "an integer that can be expressed as
two times some integer." Cf course there can be only
one definition for a particular concept, so when more
possibilities exist, how do you select the definition
and what happens to the other alternatives? Since a
definition is simply something agreed on, anyone of
the alternatives can be agreed on as the definition.
Once the definition has been chosen, it would be
advisable to establish the "equivalence" of the
definition and the alternatives.
For the case of an even integer, this would be
accomplished by using Cefinition 5 to create the
statement A: "n is an integer whose remainder on
division by 2 is 0." Using the alternative concept,
one then creates the statement B: "n is an integer
that can be expressed as two times some integer." To
establish the fact that the definition is equivalent
to the alternative, you must show that "A implies S"
and "P implies A" (see refinition 8). Then you would
know that if ~ is true (i.e., n is an integer whose
remainder on division by 2 is 0), then P is true
(i.e., n is an integer that can be expressed as two
times some integer). ~ o r e o v e r , if S is true then A
is true too.
26
DEFINITIONS AND TERMINOLOCY
The statement that ~ is equivalent to e is often
written "A is true if and only if B is true" or, more
simply, "1 if and only if B." In mathematical
notation one would write "A iff S" or "" <=> B."
Whenever you are asked to show that "A if and only if
e," you must show that "A implies B" and "B implies
A. "
It is quite important to be able to establish
that R definition is equivalent to an alternative.
Suppose, for example, that, in some proof, you derive
the abstraction question "How can I show that an
integer is even?" As a result of having obtained the
equivalence of the two concepts, you now have two
possible answers at your ,fingertips. One is obtained
directly from the definition, so one way to show that
an integer is even is to show that its remainder on
division by 2 is 0. The second answer comes from the
alternative: you can show that the integer can be
expressed as two times some integer. Similarly, in
the forward process, if you know that n is an even
integer, then you would have two possible statements
that are true as ~ result of this: the original
defini tion and the al ternRtive. While the abil i ty to
answer an abstraction question (or to go forward) in
more than one way can be a hindrance, as was the case
in Example 1, it can also be advantageous, as shown
in the next example.
Example 2. If n is an even integer then n
2
is an even
integer.
OUtline of proof. Proceeding by the forwardbackward
method, you are led immediately to the abstraction
question "How can r show that an integer (namely n
2
)
is even?" By choosing the alternative over the
definition, you can answer this question by showing
that n
2
can b ~ expressed as two times some integer,
the only question being which integer. The answer
comes from the forward process.
Since n is an even integer, using the
alternative, n can be expressed as two times some
integer, say k (i.e., n = 2k). So
n
2
= (n) (n) = (2k) (2k) = 1k2 = 2(2k
2
).
CEFINITIONS AND TERMINOLOCY 27
Thus, it has been shown that n
2
can be written as two
times some integer, that integer being 2k2, and this
completes the proof. Of course this problem could
also have been solved by using Definition 5, but it
is much harder that way.
Proof of Example 2. Since n is an even integer, there
is an integer k for which n = 2k. Consequently
n
2
= (2k)2 = 2 (2k
2
), and so n
2
is an even integer./ /
A definition is one common method for working
forward and for answering certain abstraction
questions. The more statements that you can show are
equivalent to the definition, the more ammunition you
will have available for the forward and backward
processes; however, a large number of equivalent
statements can also make it to know exactly
which one to use.
In dealing with proofs, there are four terms
that you will often come across in mathematics:
proposi tion, theorem, lemma, and corollary. A
is a true statement of interest that you
are trying to prove. All of the examples that have
been presented here are propositions. Some
propositions are (subjectively) considered to be
extremely important and these are referred to as
The proof 0 f a theorem can be very long,
and it is often easier to communicate the proof in
pieces. For example, in proving the statement A
implies B, it may first be necessary to show that A
implies C," then that C implies C, and finally that
D implies B." Each of these supporting propositions
might be presented separately and would be referred
to as a lemma. In other words, a lemma 1s a
preliminary proposition that is to be used in the
proof of a theorem. Once a theorem has been
established, it is often the case that certain
propositions follow almost immediately as a of
knowing that the theorem is true. These are called
In summary, a proFosition is a true
statement that you are trying to prove, and a theorem
is an important proposition. Finally, a lemma is a
preliminary proposition that is to be used in the
proof of a theorem, and a corollary is a proposit1on
that follows from a theorem.
28 tEFINITIONS AND TERMINOLOGY
Just as there are certain mathematical concepts
that are accepted without a formal definition, so are
there certain propositions that are accepted without
a formal proof. These unproved propositions are
called a x i o m ~ . One example of an axiom is the
statement: the shortest distance between two points
is a straight line. ~ further discussion of axioms
is beyond the scope of this document.
Just as a definition can be used in the forward
and backward processes, so can a (previously proven)
proposition, as will be shown in the next example.
Example 3. If the right triangle RST with sides of
lengths rand s, and hypotenuse of length t satisfies
t = (2rs, then the triangle RST is isosceles (see
Figure 5).
R
Fig. s. The right triangle RST.
Outline of proof. The forwardbackward method gives
rise to the abstraction auestion How can I show that
a triangle (namely RST) is isosceles? One answer is
to use Definition 3, but a second answer is also
provided by the conclusion of Fxamp1e 1, which states
that the triangle XYZ is isosceles. Perhaps the
current triangle RST is also isosceles for the same
reason as triangle YYZ. In order to find out, it is
necessary to see if RST also satisfies the hypothesis
of Example I, as did triangle XYZ, for then RST will
also satisfy the conclusion, and hence be isosceles.
DEFINITIONS AND TERMINOLOGY 29
In verifying the hypothesis of Example 1 for the
triangle RST, it is first necessary to "match up" the
current notation with that of Example 1. To be
specific, the corresponding lengths are x • r, y • s,
and z = t. Thus, to check the hypothesis of
Example 1 for the current problem, you must see if
the area of triangle RST equals 1/4(t
2
), or
equivalently, since the area of triangle RST is
1/2'rs), you must see if l/2(rs) = 1/4(t
2
).
The fact that 1/2(rs) = 1/4 (t
2
) will be
established by working forward from the current
hypothesis that t = (1FS. To be specific, on
squaring both 2sides and dividing by 4, one obtains
1/2(rs) = 114ft), as desired. Do not forget to
observe that the hypothesis of Example 1 also
requires that the triangle RST be a right triangle,
which of course it is, as stated in the current
hypothesis.
Notice how much more difficult it would have
been to match up the notation if the current triangle
had been labeled WXY with sides of length wand x,
and hypotenuse of length y. Lnfortunately this
"overlapping" notation can (and will) arise, and,
when it does, it is particularly important to keep
the symbols straight.
In the condensed proof that follows, note the
complete lack of reference to the matching of
notation.
Proof of Example 3. By the hypothesis, t = {!FS, so
t
2
= 2rs, or equivalently, 1/4(t
2
) = 1/2(rs)2 Thus,
the area of the right triangle RST = l/41t ). As
such, the hypothesis, and hence the conclusion, of
Example 1 is true. Consequently, the triangle RST is
isosceles.11
Using the conclusion of a previous proposition
to answer an abstraction question is quite common.
Do not forget that you must match up the current
notation to that of the previous proposition so that
the hypothesis of the previous proposition can be
verified.
30
DEFINITIONS ANC TERMINOLOGY
Associated with a statement A is the statement
NOT A (sometimes written ~ ~ ) . The statement NOT A
is true when A is false, and vice versa. More will
be said about the NOT of a statement in Chapter 10.
Given two statements A and B, you have already
learned the meaning of the statement A implies B.
There are many other ways of saying the statement A
implies B, for example:
1. Whenever A is true, B must also be true.
2. B follows from A.
:. B is a necessary consequence of A.
4. A is sufficient for B.
5. A only if B.
Three other statements related to A implies B are:
]. B impl ies A (called the c.onvwe)
2. NOTA implies NOTB (called the tnvwe)
3. NOT B implies NOT A (called the c.ontJr..a.po.6Uive)
Table 1 can be used to determine when each of
these three statements is true. For instance, the
contrapositive statement, NOT B implies NOT A, is
true in all cases except when the statement to the
left of the word implies (namely NOT B) is true and
the statement to the right of the word implies
(namely NOT A) is false. In other words, the
contrapositive statement is true in all cases except
when B is false and A is true, as shown in Table 3.
Note, from Table 3, that the statement NOT B
implies NOT A is true under the same conditions as
A implies B,· that is, in all cases except when A is
true and B is false. This observation gives rise to
a new proof technique known as the contrapositive
method that will be described in Chapter 9. Truth
tables similar to Table 3 can be derived for the
converse and inverse statements and are left as
exercises.
This chapter has explained the meaning of many
of the terms used in the language of mathematics.
More importantly, it showed how definitions and how
DEFINITIONS AND TERMINOLOGY
31
Table 3. The Truth Table for wNOT 8 Implies NOT AW
*****************************************************
A 8 NOT B NOT A A
=> B NOT B
=>
NOT A

True True False False True True
True False True False False False
False True False True True True
False False True True True True
*****************************************************
previous propositions can often be used in the
forwardbackward method. Now it is time to learn
more proof techniques.
Exercises
Note: All proofs should contain an outline of proof
as well as a condensed version.
3.1. For each of the following conclusions, pose an
abstraction question. Then use a definition to
(1) answer the question abstractly and (2) apply
the answer to the specific problem.
(a) If n is an odd integer, then n
2
is an odd
integer.
(b) If sand t are rational numbers with t ~ 0,
then sit is rational.
(c) Suppose that a, b, c, d, e, and f are real
numbers. If (x
1
'Y1) and (x
2
'Y2) are real
numbers satisfying:
32
(d)
(e)
DEFINITIONS AND TERMINOLOGY
aX
1
+ bY1 • e, cx
1
+ dY
1
• f,
aX
I
+ bY
I
• e, ex I + dy
z
• f,
then (X
1
'Y'l) equals (xI 'Ya ).
If n is a PQsitive integer greater than 1
for which 2"  I is prime, then n is prime.
If (nl), n, and (n+l) are three consecutive
integers, then 9 divides the sum of their
cubes.
3.2. For each of the following hypotheses, use a
definition to work forward one step.
(a) If n is an odd integer then n
2
is an odd
integer.
(b) If sand t are rational numbers with t ~ 0,
then sit is rational.
(c) If triangle RST is equilateral then the area
of the triangle is (3/4 times the square of
the leng th of a side.
(d) If the right triangle YYZ of Figure 2
satisfies sin(X) • cos(X), then triangle XYZ
is isosceles.
(e) If a, b, and c are integers for which alb
and blc, then ale.
3.3. Write truth tables for the following statements.
(a) The converse of A implies B.
(b) The inverse of A implies B.
How are (a) and (b) related?
(c) A CR B
(d) A AND B
(e) A AND NOT B
(f) (NOT A) OR 8
How is (f) reI ated to  A implies 8?
3.4. For each of the following propositions, write
down the converse, inverse, and contrapositive
statements. 2
(a) If n is an integer for which n is even then
n is even. 2
(b) If r is a real number such that r ~ 2, then
r is not rational.
(c) If the quadrilateral ABCt is a parallelogram
with one right angle, then the quadrilateral
ABCt is a rectangle.
DEFINITIONS AND TERMINOLOCY
33
fd) If t is an angle for which sin(t) = cos(t)
and 0 < t < _, then t = _/4.
l.S. Prove that if n is an odd integer, then n
2
is an
odd integer.
3.6. Prove that if n is an odd integer and m is an
odd integer, then mn 1s an odd integer.
3.7. Prove that if "A implies F" and"e implies C,"
then "A implies Co"
3.8. Prove that if "A implies e," "8 implies C," and
·C implies A,· then A is equivalent to B and A
is equivalent to C.
3.9. Suppose that you have a definition in the form
of a statement A together with three possible
alternative definitions, say S, r, and r.
(a) How many proofs would it require to show
that is equivalent to each of the three
al ternatives?
(b) How many proofs it require to show
that implies 8,· ·e implies C,· ·C
implies D," and "0 implies
(c) Explain why the approach in part (b) is
sufficient to establish that each of the
alternatives is equivalent to the original
definition (and to each other) •
3.10. Prove that if the right triangle UVW with sides
of lengths u and v, and hypotenuse of length w
satisfies sin(U) = then the triangle UVW
is isosceles, by:
(a) Using the definition of an isosceles
tr iangle.
(b) Verifying the hypotheSis of Example 1.
(c) Verifying the hypothesis of Fxample 3.
4 quantifiersl: the construction method
In the previous chapter you saw that a
definition could successfully be used to answer an
abstraction question. The next four chapters provide
you with several other techniques for formulating and
answering an abstraction question that arises when e
has a special form.
Two particular forms of E appear repeatedly
throughout all branches of mathematics. They can
always be identified by certain key words that appear
in the statement. The first one has the words "there
is" ("there are," "there exists"), whereas the second
one has "for all" ("for each," "for every," "for
any"). These two groups of words are referred to as
and each one will give rise to its own
proof technique. The remainder of this chapter deals
with the "there is" and with
the corresponding proof technique called the
method. The "for
all" and its associated proof technique is discussed
in the next chapter.
The quantifier "there is" arises quite naturally
in many mathematical statements. Recall Cefinition 7
for a rational number as being a real number that can
be expressed as the ratio of two integers in which
the denominator is not zero. This definition could
just as well have been written using the quantifier
"there are."
Definition 11. real number r is rational if and
only if there are integers p and q
with q F 0 such that r = p/q.
34
CONSTRUCTION
35
Another such example arises from the alternative
definition of an even integer, that being an integer
that can be expressed as the product of two times
some integer. Using a quantifier to express this
statement, one obtains:
Definition 12. An integer n is even if
there is an integer
n = 2k.
and only if
k such that
It is imFortant to observe that the quantifier "there
is" allows for the possibility of more than one such
object, as is shown in the next definition.
Definition 13. integer n is a square if 2there is
an integer k such that n = k •
Note that if an integer n (SClY, for example, n = 9)
is square, then 2there are usually two values of k
that satisfy n = k (in this case, k = 3 or 3).
will be said in Chapter 11 about the issue of
uniqueness (i.e., the existence of only one such
object) •
There are many other instances where an
existential quantifier can and will be usen, but from
the examples above, you can see that such statements
always have the same basic structure. Fach time the
quantifier "there is," "there are," or "there exists"
appears, the statement will have the following bClsic
form:
There is an "object" with a "certain property"
such that "something happens."
The words in quotation marks depend on the particular
statement under consideration, and you must learn to
read, to identify, and to write each of the three
components. Consider these examples.
1. is an integer x > 2 such that
(x  5x + 5) = O.
Cbject: integer x
Certain property: x >22
Something happens: (x  5x + 6) 0
38 CONSTRUCTION METHOD
2. There are real numbers x and y both > 0 such that
(2x + 3y)  8 and (5x  y) • 3.
Object: real numbers x and y
Certain property: x > 0, y > 0
Something happens: (2x + 3y)  8 and (5x  y)  3
Mathematicians often use the symbol "3" to
abbreviate the words "there is" ("there are," etc.)
and the symbol " ~ " for the words "such that" ("for
which," etc.). The use of the symbols is illustrated
in the next example.
~ . 3 an angle t ~ cos(t) = t.
Cbject: angle t
Certain property: none
Something happens: cos(t) = t
Observe that the words "such that" (or equivalent
words like "for which") always precede the something
that happens. Practice is needed to become fluent at
reading and writing these statements.
During the backward process, if you ever come
across a statement having the quantifier "there is,"
then one way in which you can proceed to show that
the statement is true is through the construction
method. The ioea is to construct (guess, produce,
devise an algorithm to produce, etc.) the desired
object. Cf course you must show that the object has
the certain property and that the something happens.
Bow you actually construct the desired object is not
at all clear. Sometimes it will be by trial and
error; sometimes an algorithm can be designed to
produce the desired object. It all depends on the
particular problem, nonethless, the information in
statement A will surely be used to help accomplish
the task. Indeed, the appearance of the quantifier
"there is" strongly suggests turning to the forward
process to produce the desired object. The
construction method was subtly used in Example ~ , but
another example will serve to clarify the process.
Example 4. If a, b, c, d, e, and f are real numbers
with the property that (ad  bc) , 0, then the two
linear equations (ax + by) = e and (cx + dy) = f can
be solved for x and y.
CONSTRUCTION METHOD 37
OUtline of proof. On starting the backward process,
you should recognize that the statement B has the
form discussed above, even though the quantifier
there are
w
does not appear explicitly. Observe that
the statement B can be rewritten to contain the
quantifier explicitly, for example, wThere are real
numbers x and y such that (ax + by) • e and
(ex + dy) = f. Statements containing Whidden
quantifiers occur frequently in problems and you
should watch for them.
Proceeding with the construction method, the
issue is how to construct real numbers x and y such
that (ax + by) = e and (cx + dy) = f. If you are
clever enough to Wguess
W
that x • (de  bf)/(ad  be)
and y = (af  ce)/(ad  bc), then you are very
fortunate, but you must still show that the something
happens, in this case, that (ax + by) • e and
(ex + dy) • f. This, of course, is not hard to do.
Also observe that, by guessing these values for x and
y, you have used the information in A since the
denominators are not O.
While this Wguess and check
w
approach is
perfectly acceptable for producing the desired x and
y, it is not very informative as to how these
particular values were produced. A more instructive
proof would be desirable. For example, to obtain the
values for x and y, you could start with the two
equations (ax + by) • e and (ex + dy) = f. On
multiplying the first equation by d and the second
one by b and then subtracting the second one from the
first one, you obtain (ad  bc) x • (de  bf). If you
then use the information in A, it is possible to
divide this last equation by (ad  bc) since, by
hypothesis, this number is not '0, thus obtaining
x • tde  bf)/(ad  bc). A similar process can be
used to obtain y. taf  ce)/(ad  be). Recall here
that a proof is a convincing argument. As such, the
statement Wa similar process can be used to obtain
y • taf  ce)/(ad  bc)W might not be very convincing
to you, in which case a proof directed at you would
have to contain the details of how y is obtained. In
any event, you must still show that, for these values
of x and y, (ax + by) = e and (ex + dy) • f, and that
will complete the proof.
38 CONSTRUCTION METHOD
Proof of Example 4. On multiplying the equation
Cax + by) • e by d, and the equation Ccx + dy) • f by
b, and then subtracting the two equations one obtains
(ad  bc)x = (de  bf). By using the hypothesis,
Cad  bc) , 0, and so dividing by (ad  bc) yields
x  (de  bf)/Cad  bc). A similar argument shows
that y • (af  ce)/(ad  bc), and it is not hard to
check that, for these particular values of x and y,
(ax + by) = e and Ccx + dy) • f.//
The construction method is not the only
technique available for dealing with statements
having the quantifier "there is," but it often works
and should be considered seriously. To be successful
with the construction method, you must become a
"builder," and use your creative ability to construct
the desired object having the certain property.
Also, you must not forget to show that the something
happens. Your "building supplies" consist of the
information contained in A.
Exercises
Note: ~ l l proofs should contain an outline of proof
as well as a condensed version.
4.1. For each of the following statements, identify
the objects, the certain property, and the
something that happens.
Ca) In the Himalayas, there is a mountain over
20,000 feet high that is taller than every
other mountain in the world.
Cb) There exists an integer x that satisfies
x
2
 5x/2 + 3/2 = o.
(c) Through a point P not on a line i, there is
a line ~ . through P parallel to i.
(d) There exists an angle t between 0 and ~ / 2
such that sin(t) • cos(t).
(e) Between the two real numbers x and y, there
are distinct rational numbers rand s for
which Ir  sl < 0.001.
CONSTRUCTION METHOD 39
4.2. Reword the following statements using the
symbols "3" and " ~ . "
(a) A triangle XYZ is isosceles if two of its
sides have equal length.
(b) Given an angle t, one can find an angle t'
whose tangent is larger than that of t.
(c) At a party of n people, at least two of the
people have the same number of friends.
(d) A polynomial of degree n, say p(x), has
4.3. (a)
(b)
exactly n complex roots, say r
1
' ,r
n
,
for which p(r
1
) = ••• = p(r
n
) = o.
Prove that there is an integer x such that
x
2
 5x/2 + 312 = O. Is the integer unique?
Prove that there is a real number x such
that x
2
 5x/2 + 312 = O. Is the real
number unique?
4.4. Prove that if a, b, and c are integers for which
alb and blc, then ale.
4.5. Prove that if sand t are rational numbers and
t ~ 0, then sIt is a rational number.
5 quantifiersll: the choo.e method
This chapter develops the c . h o o ~ e method, a proof
technique for dealing with statements containing the
quantifier "for all." Such statements arise quite
naturally in many mathematical areas, one of which is
set theory. Some time will be devoted to this topic
right here and now because it will demonstrate the
use of the quantifier "for all."
A ~ e t is nothing more than a collection of
items. For example, the numbers 1, 4, and 7 can be
thought of as a collection of items and hence they
form a set. Each of the individual items is called a
membe4 or element of the set and each member of the
set is said to be tn or belong to the set. 'I'he set
is usually denoted by enclosing the list of its
members (separate" by commas) in braces. Thus, the
set consisting of the numbers 1, 4, and 7 would be
written {1,4,7}. To indicate that the number 4
belongs to this set, mathematicians would write
"4 E {1,4,7}," where the symbol "E" stands for the
words "is a member of." Similarly, to indicate that 2
is not a member of {1,4,7}, one would write
"2 ~ {l,4,7}."
While it is certainly desirable to make a list
of all of the elements in a set, sometimes it is
impractical to do so because the list is simply too
long. For example, imagine having to write down
every integer between 1 and 100,000. When a set has
an infinite number of elements (such as the set of
real numbers that are greater than or equal to 0) it
will actually be impossible to make a complete list,
even if you wanted to. Fortunately there is a way to
40
CHOOSE ft1ETHOD 41
describe such "large" sets through the use of what is
known as 6et It involves using a
verbal and mathematical description for the members
of the set. Consider the example of the set of all
real numbers that are greater than or equal to O.
One would write S := {real numbers x: x:!: OJ, where
the ":" stands for the words "such that." Everything
following the ":" is referred to as the
of the set. The question that one always
has to be able to answer is "How do I know if a
particular item belongs to the set or not?" To answer
such a question, you need only check if the item
satisfies the defining property. If so, then it is
an element of the set, otherwise it is not. For the
example above, to see if the real number 1 belongs to
S, you simply replace x by 2 and see if
the defining property is true. In this case 3 does
belong to S because 3 is;;: c.
Sometimes part of the defining property appears
to the left of the ":" as well as to the right, and,
when trying to determine if a particular item belongs
to the set, you must be sure to verify this portion
of the defining property too. For example, if
T = {real numbers x :!: 0: (x
2
 x  2) :!: OJ, then 1
does not belong to T even though it satisfies the
defining property to the right of the ":". The
reason is that it does not satisfy the defining
property to the left of the":" since 1 is not:!: C.
From a proof theory point of view, the defining
property plays exactly the same role as a definition
did: it is used to answer the abstraction question
"How can I show that an item belongs to a particular
set?" One answer is to check that the item satisfies
the defining property.
While discussing sets, observe that it can
happen that no item satisfies the defining property.
Consider, for example,
{real numbers x :!: 0: (i + 3x + 2) = OJ.
The only real numbers for which (x
2
+ 3x + 2) = 0 are
x • 1 and x = 2. Neither of these satisfies the
defining property to the left of the ":". Euch a set
42
CHOOSE METHOD
is said to be empty, meaning that it has no members.
The special symbol " ~ " is used to denote the empty
set.
To motivate the use of the quantifier "for all,"
observe that it is usually possible to write a set in
more than one way, for example, the sets
S = {real numbers x: (x
2
 ~ x + 2) s C} and
T z {real numbers x: ) s x s 2},
where 1 s x ~ 2 means that 1 s x and x s 2.
Surely for two sets Sand T to be the same, each
element of S should appear in T and vice versa.
Using the quantifier "for all," a definition can now
be made.
Definition 14. 'A set S is said' to be a subset of a
set T (written S ~ T) if and only if
for each element x in S, x is in 7.
Definition 15. Two sets Sand T are said to be equal
(written S = T) if and only if S is a
subset of T and T is a subset of S.
Like any definition, these can be used to answer
an abstraction question. The first one answers the
question "How can I show that a set (namely S) is a
subset of another set (namely T)?" by requiring you
to show that for each element x in S, x is also in T.
'As you will see shortly, the choose method will
enable you to "show that for each element x in E, x
is also in T." The second definition answers the
abstraction question "How can J show that the two
sets Sand T are equal?" by requiring you to show
that S is a subset of T and T is a subset of S.
In addition to set theory, there are many other
instances where the quantifier "for all" can and will
be used, but, from the above example, you can see
that all such statements appear to have the same
consistent structure. When the quantifiers "for
all," "for each," "for every," or "for any" appear,
the statement will have the following basic form
(which is similar to the one you saw in the previous
chapter) :
CHOOSE METHOD
For every "object" with a "certain Froperty,"
"something happens."
43
The words in quotation marks depend on the particular
statement under consideration, and you must learn to
read, to write, and to identify the three components.
Consider these examples.
1. For every angle t, sin
2
(t) + COS
2
(t) = 1.
Object: angle t
Certain property: none
Something happens: sin
2
(t) + COS
2
(t) = 1
Mathematicians often use the symbol "V" to abbreviate
the words "for all" ("for each," etc.). The use of
symbols is illustrated in the next example.
2. V real numbers y > a, 3 a real number x ~ , y.
Object: real numbers y
Certain property: y > 0
Something happens: 3 a real number x ~ 2 = Y
Observe that a comma always precedes the something
that happens. Sometimes the quantifier is "hidden,"
for example, the statement "the cosine of any angle
strictly between 0 and ~ / 4 is larger than the sine of
the angle" could be phrased equally well as "for
every angle t with r < t < ~ / 4 , cos(t) > sin(t)."
Practice is needed to become fluent at reading and
writing these statements.
Curing the backward process, if you ever come
acroSS a statement having the quantifier "for all" in
the form discussed above, then one way in which you
might be able to show that the statement is true is
to make a list of all of the objects having the
certain property. Then, for each one, you could try
to show that the something happens. When the list is
finite, this might be a reasonable way to proceed.
However, more often than not, it will not be
practicable because the list is too long, or even
infinite. You have already dealt with this type of
obstacle in set theory where the problem was overcome
by using the defining property to describe the set.
Here, the choose method will allow you to circumvent
the difficulty.
44
CHOOSE METHOD
The choose method can be thought of as a proof
machine that, rather than actually checking that the
something happens for each and every object having
the certain property, has the capability of doing so.
If you had such a machine, then there would be no
need to check the whole (possibly infinite) list
because you would know that the machine could always
do so. The choose method shows you how to design the
inner workings of the proof machine.
In
Proof Cut
machine ~           ~
Fig. 6. The proof machine for the choose method.
To understand the mechanics of the choose
method, put yourself in the role of the proof machine
and keep in mind that you need to have the capability
of taking any object with the certain property and
concluding that the something happens (see Figure 6).
As such, pretend that someone gave you one of these
objects, but remember, you do not know precisely
which one. ~ l l you do know is that the particular
object does have the certain property, and you must
somehow be able to use the property to reach the
conclusion that the something happens. This is most
easily accomplished by working forward from the
certain property and backward from the something that
happens. In other words, with the choose method, you
choose one object that has the certain property.
Then, by using the forwardbackward method, you must
conclude that, for the chosen object, the something
happens. Then your proof machine will have the
capability of repeating the proof for any of the
objects having the certain property.
CHOOSE METHOD 45
Suppose, for example, that, in some proof, you
n:eded to show that, for all real numbers x with
x  3x + 2 !!S 0, ] 1: X 1: 2. With the choose method
you would choose one of these real numbers, say x',
that
2
does have the certain property (in this case,
(x')  3x' + 2 !!S 0). Then, by working forward from
the fact that (X,)2  3x' + 2 ~ 0, you must reach the
conclusion that, for x', the something happens, that
is, 1 :s x' 1: 2.
Here, the symbol x' has been used to
distinguish the chosen object from the general
object, x. This distinction is often ignored (i.e.,
the same symbol is used for both the general object
and the chosen one), and you must be careful to
interpret the symbol correctly. Consider the
following example.
Exaaple 5. If
S .. {real numbers x: (x
2
 3x + 2) 1: O}, and
T .. {real numbers x: 1 1: X 1: 2},
then S .. T.
OUtline of proof. The forwardbackward method gives
rise to the abstraction question How can I show that
two sets (namely Sand T) are equal? Definition 15
can provide the answer that you must show that S is a
subset of T and T is a subset of S, so first try to
establish that S is a subset of T, and afterward,
that T is a subset of S.
To show that S is a subset of T, you obtain the
abstraction question How can I show that a set
(namely S) is a subset of another set (namely T)?
Again, using the definition leads to the answer that
you must show that, for all x in S, x is in T. This
new statement Bl clearly has the form described
above, thus indicating that you should proceed by the
choose method. To do so, you must choose an object
having the certain property and show that the
something happens. In this case that means you
should choose an element, say x, in S, and, using the
fact that x is in S (i.e., that it satisfies the
46
CHOOSE METHOD
defining property of S) together with the information
in A, you must show that x is in T. Note that you do
not want to pick one specific element in S, say 3/2.
Also, note the double use of the symbol x for both
the general and the chosen object.
The statement x is in T now becomes B2 and you
should apply the abstraction process to B2, obtaining
the abstraction question How can I show that an item
(namely x) belongs to a set (namely T)? together
with the answer that you must show that x satisfies
the defining property of T (i.e., that 1 s x s 2), to
which your efforts must be directed.
Turning now to the forward process, you can make
use of the information in A to show that 1 s x s 2,
because you have assumed that A is true. However,
there is additional information available to you.
Recall that, during the backward process, you made
use of the choose method, at which time you chose x
to be an element in S.Now is the time to use the fact
that x is in Sspecifically, since x is in S, from
the defining property of the set S, one has
(x
2
 3x + 2) ~ O. Then, by factoring, one obtains
ex  2) ex  1) s O. The only way that the product of
the two numbers (x  2) and (x  1) can be ~ 0 is for
one of them to be ~ 0 and the other ~ O. In other
words, either (x  2) ~ 0 and ex  1) ~ 0, or else
(x  2) ~ 0 and (x  1) ~ O. The first situation can
never happen because if it did, x would be ~ 2 and x
would be ~ 1, which is impossible. Thus the second
condition must happen (i.e., x ~ 2 and x ~ 1), but
this is precisely the last statement obtained in the
backward process, and hence it has successfully been
shown that S is a subset of T. Do not forget that
you still have to show that T is a subset of S in
order to complete the proof that S = T. This part
will be left as an exercise.
Observe that, when you use the choose method,
you obtain additional information that is added to
the assumption that A is true. Invariably, in the
forward process, you will use the extra information.
Proof of Example 5. To show that S = T it will be
shown that Sis a subset of T and Tis a subset of S.
CHOOSE MFTHOD
47
To see that S is a subset of T, let x be in S (the
use of the word "let" frequently indicates that the
choose method has been invoked). Consequently,
(x
2
 3x + 2) s 0, and so one has (x  2) (x  1) s O.
This means that either (x  2) ~ 0 and (x  1) s 0,
or else, (x  2) s 0 and (x  1) ~ O. The former
canno t happen because if it did, x ~ 2 and x s 1.
Hence it must be that x s 2 and x ~ 1, which means
that x is in T. The proof that T is a subset of S is
omitted.11
The choose method is a viable approach for
dealing with a statement that contains the quantifier
"for all." Proceed by choosing an object having the
certain property. Add this information to that in A
and attempt to show that the something happens by
using the forwardbackward method.
Exercises
Note: All proofs should contain an outline of proof
as well as a condensed version.
5.1. For each of the following definitions, identify
the objects, the certain property, and the
something that happens.
(a) The real number x· is a maximum of the
function f if, for every real number x,
f (x) s f(x·).
(b) Suppose that f and g are functions of one
variable. Then g is ~ f on the set S of
real numbers if, for every element x in S,
g(x) ~ f(x).
(c) The real number u is an upper bound for a
set S of real numbers if, for all x in S,
x s u.
(d) The real number u is a least upper bound for
a set S of real numbers if u is an upper
bound for S and V real numbers t > 0,
3 xES ~ x > u  t.
48
Ie)
(f)
(g)
(h)
CHOOSE METHOD
The set C of real numbers is convex if, for
every element x and y in C and for every
real number t between 0 and 1, tx + (1  t)y
is an element of C.
The function f of one variable is a convex
function if, for all real numbers x and y
and for all real numbers 0 s t s ], it
follows that
f(tx + n  t)y) s tf(x) + (1  t)f(y).
The function f of one variable is continuous
at the point x if, for every real number
& > 0, there is a real number 8 > 0 such
that, for all real numbers y with
I x  y I < 8, If (x)  f (y) I < &.
Suppose that x, x1, x2,... are real
numbers. The sequence x', x
2
,... converges
to x if, V real numbers & > 0, 3 an integer
k' ~ V integers k > k', Ixk  xl < &.
5.2. For each of the parts in Exercise 5.], describe
how the choose method would be aFplied. Use a
different symbol to distinguish the chosen
object from the general object. For instance,
for Exercise 5.l(a), to show that x· is the
maximum of the fUnction f, one would choose a
real number, say x', for which it must then be
shown that f(x') s f(x·). Thus, to apply the
choose method in Exercise 5.]la), one would say
"Let x I be a real number. It wi 11 be shown that
f (x ') s f (x·) ."
5.3. Consider the problem of showing that "For every
object x with a certain property, something
happens." Discuss why the approach of the choose
method is the same as that of using the
forwardbackward method to show that "If x is an
object with the certain property, then the
something happens." How are the t ~ ~ statements
in quotation marks related?
5.4. Reword the following statements using the
appropriate symbols V, 3, ~ wherever necessary.
(a) Some mountain is taller than every other
mountain.
(b) If t is an angle then it follows that
sin(2t) = 2sin(t)cos(t). (Hint: Make use
CHOOSE METHOD 49
of Fxercise 5. 'Y.)
(c) The square root of the product of any two
nonne9ative real numbers p and q is always
their sum divided by 2.
(d) If x and yare real numbers such that x < y,
then there is a number r such that
x < r < y.
5.5. For each of the followin9 statements, indicate
which proof techniques you woulc use (choose
and/or construction) .3nd in which order. Also,
explain how the technique would be applied to
the particular problem, that is, what would you
construct, what would you choose, etc.
(a) There is a real number > 0 such that, for
all elements x in the set S of real numbers,
Ixl
(b) For all real numbers M > 0, there is an
element x in the set S of real numbers such
that Ixl > M.
(c) V real numbers & > 0, 3 a real number
8 > 0 V real numbers x anCl y wi th
Ix  yl < 8, If(x)  fey) I < & (where f is a
function of one variable) •
5.6. For the sets S anCl T of Fxamrle 5, prove that
T S; s.
5.7. Prove that for every real number x > 2, is
a real number y < r. such that x = 2y/(1 + y).
5.8. Prove that if m and b are real numbers, and f is
a function defined by f(x) = mx + b
,
then f is
convex. (Hint: Use the definition in Exercise
5.1(f).)
6 quantifierslll: induction
In the previous chapter you learned how to use
the choose method when the quantifier "for all"
appears in the statement P. There is one very
spec ial form of B for wh ich a separate proof
techn i que known as mathe.mati.c.af. btduc.ti.o tt has been
developed. Induction should seriously be considered
(even before the choose method) when E has the form:
For every integer n 2: 1, "something happens"
where the something that happens
that depends on the integer n.
the statement:
is some statement
.n.n example would be
n
For all integers n 2: J, L k = n{n+l)/2,
k=l
n
where L k = 1+ ••• +n.
k=l
When considering induction, the key words to look for
are "integer" and "2:1."
One way to attempt proving such statements would
be to make an infinite list of problems, one for each
of the integers starting from n = 1, and then prove
each statement separately. While the first few
prohlems on the list are usually easy to verify, the
issue is how to check the nth one and beyond. For
the example above, the list would be:
P (1)
50
1
Lk=1(l+1)/2
k=l
or 1 = 1
2
P(2)
Lk
k=l
3
P(3)
Lk
k=l
pen)
n+1
P(n+1): Lk
k=l
=
INDUCTION
51
2(2+1)/2 or 1+2
=
3
3 (3+1)/2 or 1+2+3
= 6
n(n+l)/2
(n+lH(n+l) + 11/2 (n+1) (n+2)/2
Induction is a clever method for proving that
each of these statements in the infinite list is
true. As wi th the choose method, induction can be
thought of as an automatic problemsolving machine
that starts with P(l) and works its way progressively
down the list proving each statement as it proceeds.
Here is how it works. You start the machine by
verifying that P(l) is true, as can easily be done
for the example above. '!"hen you feed P(l) into the
machine. It uses the fact that P(l) is true and
automatically proves that P(2) is true. You then
take P(2) and put it into the machine. Cnce again,
it uses the fact that P(2) is true to reach the
conclusion that P(3) is true, and so on (see Figure
7). Observe that, by the time the machine is going
to prove that P(n+l) is true, it will already have
shown that Pen) is true (from the {::revious step).
Thus, in designing the machine, you can assume that
PCn) is true, and your job is to make sure that
P(n+l) will also be true. Co not forget that, in
order to start the whole rrocess, you must also
verify that pel) is true.
52 INDUCTION
Verify that PO) is true
I
I
Tn
Proof machine
Out
Fig. 7. The proof machine for induction.
INDUCTION
53
To repeat, a proof by induction consists of two
steps. The first step is to verify that the
statement pel) is true. To do so, you simply replace
n everywhere by 1. Usually, to verify that the
resulting statement is true, you will only have to do
some minor rewriting.
The second step is much more challenging. It
requires that you reach the conclusion that P(n+l) is
true by using the assumption that Pen) is true.
There is a very standard way of doing this. Begin by
writing down the statement P(n+l). Since you are
allowed to assume that Pen) is true and you want to
conclude that Pfn+l) is true, you should somehow try
to rewrite the statement P(n+l) in terms of P(n)as
will be illustrated in a momentfor then you will be
able to make use of the assumption that Pen) is true.
On establishing that rCn+l) is true, the proof will
be com pI ete.
n
Example 6. For every integer n l!: 1, E k = n(n+l)/2
k=l
Outline of proof. When you are using the method of
induction, it is helpful to write down the statement
PCn), in this case:
n
PIn): E k
k=l
n(n+l)/2
The first step in a proof by induction is to verify
PCl). n everywhere by in P(n), you
obta in
1
P (1 ): E k = 1 (1 + 1) /2
k=l
a small amount of rewriting, it is easy to
verify this statement since
1
E k 1 = 1 f1 + 1) /2
k=l
54 INDUCTION
This step is often so easy that it is virtually
omitted in the condensed proof by simply saying "The
statement is clearly true for n = 1."
The second step is more invoJved. You must use
the assumption that r (n) is true to reach the
conclusion that P(n+l) is true. The best way to
proceed is to write down the statement P(n+l) by
carefully replacing n with (n+l) everywhere in P(n),
and rewriting a bit, if necessary. In this case
n+l
P (n+l): E k = (n+l) [(n+1)+l]/2 == (n+l) (n+2)/2
k=l
To reach the conclusion that P(n+l) is true, begin
with the left side of the equality in P(n+l) and try
to make it look like the right side. In so doing,
you should use the information in Pen) by relating
the left side of the equality in P(n+l) to the left
side of the equality in P(n), for then you will be
able to use the right side of the equality in P(n).
In this example,
n+l n
E k ==
k=l
(Ek) + (n+l)
k=l
Now you can use the assumption that pen) is true by
replacing
with n(n+l)/2, obtaining
n+l
E k = rn(n+l)/2] + (n+1)
k=l
All that remains is
[n(n+l)/2 + (n+l)]
the right side of
algebraic steps are:
a bit of algebra to rewrite
as [(n+l) (n+2)/2], thus obtaining
the equality in P(n+l). The
INDUCTION
2
n(n+1)/2 + (n+1) :: (n + n)/2 + 2(n+1)/2
In summary,
:: (n
2
+ 3n + 2)/2
:: (n+1) (n+2)/2
1+ ••• +(n+1) :: (1+ ••• +n) + (n+l)
:: n(n+l)/2 + (n+l)
= (n
2
+ 3n + 2)/2
:: (n+1) (n+2)/2
55
Your ability to relate P(n+l) to pen) so as to use
the induction hypothesis that pen) is true will
determine the success of the proof by induction. If
you are unable to relate P(n+l) to PIn), then you
might wish to consider a different proof technique.
Proof of Example 6. The statement is clearly true for
n :: 1. Assume that it is true for n (i.e., that
l+ ••• +n :: n(n+l)/2). Then
1+ ••• +(n+l):: 11+ ••• +n) + (n+l)
which is P(n+l).//
:: n(n+l)/2 + 2(n+l)/2
:: (n
2
+ 3n + 2)/2
= (n+l) (n+2)/2
When using the method of induction, it is not
necessary that the first value for n has to be 1.
For instance, the statement "for all integers n ~ 5,
2
n
> n
2
" can also be proved by induction. The only
modification is that, in order to start the proof
machine, you must verify PIn) for the first possible
value of n. In this case, that first value would be
n = 5, so you will have to check that 2
5
> 52 (which,
58
of course, it is
second step 0 f
same. You would
true (i.e., 2
n
>
2
n
+
1
> (n+l)2).
fact that n 5,
INDUCTION
since 2
5
= 32 while 52 = 25). The
the induction proof would remain the
still have to show that if P (n) is
n
2
), then P (n+l) is al so true (I.e.,
In so doing, you can also use the
if necessary.
Another modification to the basic induction
method arises when you are having difficulty relating
P(n+l) to P(n). Suppose, however, that you can
relate P(n+l) to p.(j), where j < n. In this case,
you would like to use the fact that Plj) is true, but
can you assume that prj) is, in fact, true? The
answer is yes! To see why, recall the anology of the
proof machine, and observe that, by the time the
machine has to show that P (n+l) is true, it wi 11
already have established that all of the statements
P(1), ••• ,Pfj), ••• ,Pln) are true (look again at Figure
7). Thus, when trying to show that PCn+1) is true,
you can assume that PCn) and all of the preceding
statements are true. Such a proof is referred to as
lnductlon.
Induction is a very powerful technique when
applicable; however, it is important to realize that
induction does not help you to discover the correct
form of the statement P Cn). Induction only ver ifies
that a given statement P(n) is true for all integers
n some initial one. The key to its success rests
in your ability to relate PCn+l) to PCn) or to some
previous statement, but do not forget to verify that
the statement is also true for the very first
possible value of n.
Exercises
Note: Proofs in this chapter need not contain an
outline of proof.
INDUCTION
57
6.1. For which of the following statements woulC'
induction be directly applicable? When it is
not applicable, explain why.
(a) For every positive integer n, P civides
5
n
+ 2·3
n

1
+ 1.
(b) There is an integer n ~ 0 such that 2n > n
2
•
(c) For every integer n ~ 1, it follows that
1(11)+ ••• +n(n!) = (n+l)!  1. (Recall that
n! = n (nl ) ••• 1. )
(d) For every integer n ~ 4, n! > n
2
•
(e) For every real number n ~ 1, n
2
~ n.
6.2. (a) Why and when would you want to use induction
instead of the choose method?
(b) Wny is it not possible to use induction on
statements of the form: For every "object"
with a "certain property," "something
happens?"
6.3. Prove, by induction, that, for every integer
n ~ 1, l(1!)+ ••• +n(nl) = (n+l)!  1.
6.4. Prove, by induction, that, for every integer
n ~ 5, 2 n > n
2
•
6.5. Prove, by induction, that a set of n ~ 1
elements has 7
n
subsets (including the empty
set) •
6.6. Prove, without using induction, that, for any
integer n ~ 1, l+ ••• +n = n(n+l)/2.
6.7. Prove that for every integer n ~ 1, E civides
n
3
 n, by showing that (1) the statement is
true for n = 1, and (2) if the statement is true
for (nl), then it is also true for n.
6.8.
t'escribe a "mod i fied" induction procedure that
could be used to prove statements of the form:
(a)
For every integer s some initial one,
something happens.
( b) For every integer, something happens.
( c) For every posi tive odd integer, something
happens.
58
JNCUCTION
6.9. What is wrong with the following proof that all
horses have the same color?
Proof. Let n be the number of horses. When
n = 1, the statement is clearly true, that is,
one horse has the same color, whatever color it
is. Assume that any group of n horses has the
same color. consider a group of (n+l)
horses. Taking any n of them, the induction
hypothesis states that they all have the same
color, say brown. The only issue is the color
of the remaining "uncolored" horse. Consider,
therefore, any other group of n of the (n+l)
horses that contains the uncolored horse.
by the induction hypothesis, all of the
horses in the new group must have the same
color. Then, since all of the colored horses in
this group are brown, the uncolored horse must
also be brown.11
7 quantifiersIV: specialization
In the previous three chapters you discovered
how to proceed when a quantifier appeared in the
statement B. This chapter will develop a method for
exploiting quantifiers that appear in the statement
A. When the statement A contains the quantifier
"there is" in the standard form:
There is an "object" with a "certain property"
such that "something happens"
you can use this information in a straightforward
way. When showing that "A implies P" by the
forwardbackward method, you are assuming A is true,
and in this case that means you can assume that
indeed there is an object with the certain property
such that the something happens. In doing the proof,
you would say: "Let x be an object with the certain
property and for which the something happens •••• "
The existence of this object will somehow be used in
the forward process to obtain the conclusion that B
is true.
The more interesting situation occurs when the
statement ~ contains the quantifier "for all" in the
standard form:
For all "objects" with a "certain property,"
"something happens."
To use this information, one typical method emerges
and it is referred to as 6pecialization As a
result of assuming A is true, you know that, for all
objects with the certain property, something happens.
If, at some point in the backward process, you were
to come across one of these objects that ~ o e s have
59
60
SPECIALIZATION
the certain property, then you can use the
information in A by being able to conclude that, for
this particular object, the something does indeed
happen, and that should help you to conclude that B
is true. In other words, you will have specialized
the statement A to one particular object having the
certain property. For instance, if you know that,
for every angle t, sin
2
(t) + cos
2
(t) = 1, then, in
particular, for one angle, say t you can
conclude that sin
2
+ = 1. example
demonstrates the proper use of specialization.
Definition 16. A real number u is an upper bound for
a set of real numbers! if for all
elements t in T, t s u.
Example 7. If R is a subset of a set S of real
numbers and u is an upper bound for S, then u is an
upper bound for R.
Outline of proof. The forwardbackward method gives
rise to the abstraction question "How can I show that
a real number (namely u) is an upper bound for a set
of real numbers (namely R)?" Definition 16 is used to
answer the question. Thus, i. t must be shown that,
for all elements r in R, r s u. The appearance of
the quantifier "for all" in the backward process
suggests proceening with the choose method, whereby
one chooses an element, say r, in R for which it must
be shown that r s u.
Turning now to the forward process, you will see
how specialization is used to obtain the desired
conclusion that r s u. From the hypothesis that R is
a subset of S, and by Definition ]4, you know that
each element in R is also in S. In the backward
process you came across theparticular element r in
R, and hence you can use specialization to conclude
that r is in S.
from the hypothesis, you know that u is an
upper bound for S. By Definition IE this means that,
for every element s in S, s s u. the
appearance of the quantifier "for every" in the
forward process suggests using specialization. In
SPECIALIZATION 81
particular, r is an element of S, as was shown in the
previous paragraph. So, by specialization, you can
conclude that r s u. Since the statement "r s u" was
the last one obtained in the backward process, the
proof is now complete.
Note that, when using special ization, you must
be very careful to keep your notation and symbols in
order. Also, be sure that the particular object to
which you are specializing does satisfy the certain
property, for only then c ~ n you conclude that the
something happens.
In the condensed proof that follows, note the
lack of reference to the forwardbackward, choose,
and specialization methods.
Proof of Example 7. To show that u is an upper bound
for P, let r be an element of R (the word "let"
indicates t h ~ t the choose method has been used). 8y
hypothesis, R is a subset of S and so r is also an
element of S (here is where specialization has been
used). Furthermore, by hypothesis, u is an upper
bound for S, thus, every element in S is s u. In
particular, r is an element of S, so r s u (again
specialization has been used) .11
This and the previous three chapters have
provided various techniques for dealing with
quantifiers that can appear in either A or 8. As
always, let the form of the statement guide you.
When 8 contains the quantifier "there is," the
construction method can be used to produce the
desired object. The choose method is associated with
the quantifier "for all," except when the statement e
is supposed to be true for every integer starting
from some initial one. In the latter case, induction
is 1 ikely to be successful, provided that you can
relate the statement for (n+l) to the one for n.
Finally, if the quantifier "for all" appears in the
statement A, specialization can often be exploited.
When using specialization, be sure that the
particular object under consideration does satisfy
the certain property, for only then will the
something happen.
82 SPECIALIZATION
All of the material thus far has been organized
around the forwardbackward method. Now it is time
to see some other techniques for showing that "A
implies e."
Exercises
Note: All proofs should contain an outline of proof
as well as a condensed version.
7.1. Explain the difference between the choose and
the specialization methods.
7.2. For each of the following statements and given
objects, what properties must the object satisfy
in order to be able to apply specialization,
and, given that it does satisfy the properties,
what can you conclude about the object? 2
(a) Statement: Y integers n ~ 5, 2
n
> n •
Given object: m
(b) Statement: For every element x in the set S
with Ixl < 5, x is in the set T.
Given object: y
(c) Statement: Y e > 0, :I 8 > 0 ~ Y Y with
I x  y I < 8, I f (x)  f (y) I < e
(where 8, e, x, and yare real
numbers, and f is a function of
one variable).
Given object: e'
(d) Statement: Any rectangle whose area is
onehalf the square of the length
of a diagonal is a square.
Given object: the quadrilateral eRST
(e) Statement: For any angle t with 0 < t < x/4,
cos(t) > sin(t).
Given object: Angle S of the triangle RST
7.3. Prove that if R is a subset of Sand S is a
subset of T, then R is a subset of T.
SPECTALTZATI0N
63
7.4. Prove that if Sand T
Exercise 5.] (e», then S
set.
are convex sets (see
intersect T is a convex
7.5. Prove that if f is a convex function of one
variable (see Exercise S.l(f», then for all
real numbers s ~ fI, the function sf is convex
(where the value of the fUnction sf at any point
x is sf(x».
7.6. Prove that if f is a convex function of one
variable (see Exercise 5.1(f» and y is a real
number, then the set C = {real numbers x:
f (x) s y} is convex.
7.7. Prove that 1 is a least upper bound of the set
S = {1  1/2, 1  1/3, 1  1/4, ••• } (see
Exercise S.l(d». (Hint: The set S can be
written as {real numbers x: there is an integer
n ~ 2 such that x = I  l/n}.)
8 the contradiction method
As powerful as the forwardbackward method is,
you may well find yourself unable to complete a proof
for one reason or another, as is shown in the next
example.
Example 8. If n is an integer and n
2
is even, then n
is even.
Outline of proof. The forwardbackward method gives
rise to the abstraction question "How can I show that
an integer (namely n) is even?" One answer is to show
that there is an integer k sllch that n :0: 2k. The
appearance of the quantifier "there is" suggests
proceening with the construction method, and so the
forward process will be used in an attempt to produce
the desired integer k.
Working forward from the hypothesis that 2n2 is
even, there is an integer, say M, such that n 2m.
Since the objective is to produce an integer k for
which n = 2k, it is natural to take the sauare root
of both sides of the equation n
2
= 2m to obtain
n :0: f2m, but how can you rewrite f2m to look like 2k?
It would seem that the forwarabackwarry method has
fa il ed!
Proof of Example 8. The technique that you are about
to learn will lead to a simple proof of this problem,
and it is left as an exercise.11
Fortunately, there are several other techniques
that you might want to try before you give up. In
this chapter, the c.ontlla..dic.tioYl method is described
64
CONTRAtICTION METHOD
65
together with an indication of how and when it should
be used.
With the contradiction method, you begin by
assuming that A is true, just as you would in the
forwardbackward method. However, to reach the
desired conclusion that B is true, you proceed by
asking yourself a simple question: "Why can't B be
false?" After all, if B is supposed to be true, then
there must be some reason why B cannot be false. The
objective of the contradiction method is to discover
that reason. In other words, the idea of a proof by
contradiction is to assume that A is true and B is
false, and see why this cannot happen. So what does
it mean to "see why this cannot happen?" Suppose, for
example, that as a result of assuming that A is true
and e is false (hereafter written as NOT B), you were
somehow able to reach the conclusion that 0 = I!?!
Would that not convince you that it is impossible for
A to be true and B to be false simultaneously? Thus,
in a proof by contradiction, you assume that A is
true and that NOT B is true, and, somehow, you must
use this information to reach a contradiction to
something that you absolutely know to be true.
Another way of viewing the contradiction method
is to recall that the statement "A implies B" is true
in all cases except when A is true and B is false.
In a proof by contradiction, you rule out this one
unfavorable case by actually assuming that it does
happen, and then reaching a contradiction.
At this point, several very natural questions
arise:
1. What contradiction should you be looking for?
2. Exactly how do you use the assumption that A is
true and B is false to reach the contradiction?
3. Why and when should you use this approach instead
of the forwardbackward method?
The first question is, by far, the hardest to
answer because there are no specific guidelines.
Each problem gives rise to its own contradiction, and
it usually takes creativity, insight, persistence,
and luck to produce a contradiction.
66
Method
Forward
Backward
Contra
diction
CONTRADICTION METHOD
Assume Conclude
forward backward
A}>
<B
At forward
BJ> ••• * (contradiction)
NOT
Fig. 8. ForwardBackward method versus
contradiction method.
As to the second question, one common approach
to finding a contradiction is to work forward from
the assumption that A and NOT B are true, as will be
illustrated in a moment.
The discussion above also indicates why you
might wish to use contradiction instead of the
forwardbackward method. With the forwardbackward
method you only assume that A is true, while in the
contradiction method, you can assume that both A and
NOT B are true. Thus, you get two statements from
which to reason forward instead of just one (see
Figure 8). On the other hand, you have no definite
knowledge of where the contradiction will arise.
As a general rule, use contradiction when the
statement NOT B gives you some useful information.
There are at least two recognizable instances when
the contradiction method should be considered.
Recall the statement B associated with Example 8: "n
is an even integer." Obviously an integer can only be
odd or even. When you assume that B is not true
(i.e., that n is not an even integer), then it must
be the case that n is an odd integer. Here, the
statement NOT B has given you some useful
information. In general, when the statement e is one
of two possible alternatives (as in Example C), the
contradiction method is likely to be effective
because, by assuming NOT S, you will know that the
CONTRACICTION METHOD
67
other case must happen, and that should help you to
reach a contradiction.
A second instance when the contradiction method
is likely to be successful is when the statement B
contains the word "not," as is shown in the next
example.
Example 9. If r is a real number such that r2 = 2,
then r is irrational.
Outline of proof. It is important to note that the
conclusion of Example 9 can be rewritten so as to
read "r is not rational," and as such, the appearance
of the word "not" now suggests using the
contradiction method, whereby you can assume that A
ancl NOT B are both true. In this case, that means
you can assume that r2 = 2 and that r is a rational
number. Using this information, a contradiction must
now be reached.
Working forward and using Definition 7 for a
rational number, there are integers p and q with
q # 0 such that r = p/q. There is still the
unanswered question of where the contradiction
arises, and this takes a lot of creativity. A
crucial observation here will really help. It is
possible to assume that p and q have no common
divisor (i.e., no integer that divides both p and q),
for if they did, you could divide this integer out of
both the numerator p and the denominator q. Now a
contradiction can be reached by showing that 2 is a
common divisor of p and q! This will be done by
showing that p and q are even, and hence 2 divides
them both.
Working forward, since r = p/q, it follows that
r2 = p2/q2. But, from the hypothesis, you also know
that r2 = 2, so 2 = p2/q2. The rest of the forward
process is mostly rewriting 2 = p2/q2 via algebraic
manipulations to reach the desired conclusion that
both p and q are even integers.
multiplying both sides of the last equation by q you
obtain 2q2 = p2. From 2q2 = p2, you can certainly
say that 2q2 is even no matter what kind of integer q
is, and since p2 = 2q2, p2 must also be even.
88
CONTRADICTION METHOD
Continuing with the forward process, what useful
information can be derived from the fact that p2 is
even? Well, the only way for the integer p times
itself to be even is for p to be even, and this was
the desired conclusion. Recall, once again, that a
proof is supposed to be a convincing argument, and
this last sentence might not convince you that p is
even. In this case it would be necessary to provide
more details. ".s always, a proof should be written
with the audience in mind. If you need further
convincing that p is even, look at Example 8.
Meanwhile, it is still necessary to show that q
is even too. Continuing with the forward process,
since p is even, by the alternative definition of an
even integer, there is an integer k such that p = 2k.
Now, recall ing 2 that 2q2 = p2, it follows that
2q2 = C2kt = 4k, which says that q2 = 2k
2
• Cnce
again, 2k is even no matter what type of integer k
is, and since ql = 2k2, q2 is even. Finally, note
that the only way for an integer q times itself to be
even is for q to be even (see Example 8). Thus, both
p and q are even and the desired contradiction has
been reached.
Proof of Example 9. Assume, to the contrary, that r
is a rational number of the form p/q (where p and q
are integers with q F 0) and that r2 = 2.
Furthermore, it can be assumed that p and q have no
common divisor for, if they did, this number could be
canceled from both the numerator p and the
denominator
2
q.2 Since r2 = 2 and r = p/q, it follows
that 7 = P Iq, or equivalently, 2q2 = p2. Noting
that 2q2 is even, p2, and hence p, must be even.
Consequently, there is an integer k such that p = 2k.
On substituting this value for p, one obtains
2q2 = p2 = f2k)2 = 4k2, or equivalently, q2 = 2k
2
•
From this it then follows that q2, and hence q, must
be even. Thus it has been shown that both p and q
are even and have the common divisor 2. This
contradiction establishes the claim.11
This proof was discovered in ancient times by a
follower of Pythagoras, and it epitomizes the use of
contradiction. Try to prove the statement by some
other method!
CONTRAnrCTION ~ E T H O D
69
There are several other valuable uses for the
contradiction method. Recall that, when the
statement B contains the quantifier "there is," the
construction method is recommended in spite of the
difficulty of actually producing the desired object.
The contradiction method opens up a whole new
approach. Instead of showing that there is an object
with the certain property such that the something
happens, why not proceed from the assumption that
there is no such object? Now your job is to use this
information to reach some kind of contradiction. H o ~
and where the contradiction arises is not at all
clear, but it may be a lot easier than producing or
constructing the object. Consider the following
example.
Imagine that you wish to show that there are at
least two people in the world who have exactly the
same number of hairs on their heads. If the
construction method is used, then you would actually
have to go out and find two such people. To save you
the time and trouble, contradiction can be used. In
so doing, you can assume that no two people have the
same number of hairs on their heans, or equivalently,
that everyone has a different number of hairs on
their heads. Now, assign numbers to the people in
such a way that the person with the fewest hairs
receives number 1, the person with the next fewest
hairs receives number 2, and so on. Recall that each
person is assumed to have a different number of
hairs. Thus, the person whose number is 2 must have
at least one more hair than the person whose number
is ], and so on. Consequently, the person whose
number is one billion must have at least one billion
more hairs than the person whose number is ]!
Clearly no person can have a billion more hairs than
someone else, and so a contradiction has been
established.
This example illustrates a subtle but very
significant difference between a proof using the
construction method and one that uses contradiction.
If the construction method is successful, then you
will have produced the desired object, or at least
indicated how it might be produced, perhaps with the
aid of a computer. Cn the other hand, if you
70
CONTRADICTION METHOD
establish the same result by contradiction, then you
will know that the object exists but will have no way
of physically constructing it. For this reason, it
is often the case that proofs done by contradiction
are quite a bit shorter and easier than those done by
construction because you do not have to create the
desired object. You only have to show that its
nonexistence is impossible! This difference has led
to some great philosophical debates in mathematics.
~ o r e o v e r , an active area of current research consists
of finding constructive proofs where previously only
proofs by contradiction were known.
As you have seen, the contradiction method can
be a very useful technique when the statement B
contains the word "not" in it. You work forward from
the assumption that A and NOT E are true to reach a
contradiction. One of the disadvantages of the
method is that you do not know exactly what the
contradiction is going to be. The next chapter
describes another proof technique in which you
attempt to reach a very specific contradiction. As
such, you will have a "guiding light" since you will
know what contradiction you are looking for.
Exercises
Note: All rroofs should contain an outline of proof
as well as a condensed version.
8.1. When applying the contradiction method to the
following propositions, what should you assume?
(a) If R" m, and n are 3 consecutive integers,
then 24 does not nivide R,2 + m
2
+ n
2
+ 1.
(b) If the matrix ~ is not singular then the
rows of r are not linearly deFendent.
(c) If f and g are two functions such that (1)
g !:!: f and (2) f is unbounded above, then g
is unbounden above.
CCNTRAVICTION METHOD
71
8.2. Reword each of the following statements so that
the word "not" appears explicitly.
(a) There are an infinite number of primes.
(b) The set S of real numbers is unbounded.
(c) The only positive integers that divide the
positive integer pare J and p.
(d) The I ines ~ and t' are parallel.
(e) The real number x is < 5.
8.3. Prove, by contradiction, that if n is an integer
and n
2
is even, then n is even.
8.4. Prove, by contradiction, that no chord of a
circle is longer than a rliameter.
8.5. Prove, by contradiction, that if ~ 1 and t2 are
two lines in a plane that are both perpendicular
to a third 1 ine ~ in the plane, then t1 and t2
are parallel.
8.6. Prove, by contradiction, that, at a party of
n ( ::!: 2) people, there are at least two people who
have the same number of frienrls at the party.
8.7. Prove, by contradiction, that there do not exist
three consecutive positive integers such that
the cube of the largest is equal to the sum of
the cubes of the remaining two.
8.8. Prove, by contradiction, that there are an
infinite number of primes. (Hint: Assume that
n is the largest prime. T h ~ n consider any prime
number p that divides n! + 1. Pow is p related
to n?)
8.9. For each of the following statements, indicate
which proof techniques you would use, and in
which order. Specifically, state what you would
assume and what you would try to conclude.
(Note: Throughout, Sand T are sets of real
numbers, and all of the variables refer to real
numbers. )
(a) 3 s e s ~ seT.
(b) 'f/ s in S, ~ t in T such that s > t.
( c) j M > 0 s uc h t hat, 'f/ x inS, I x I < f.l.
9 the contrapositive method
The previous chapter described the contradiction
method in which you work forward from the two
statements and NCT B to reach some kind of
contradiction. In general, the difficulty with this
method is that YOli do not know what the contrad iction
is going to be. As will be seen in this chapter, the
method has the of directing
you toward one specific type of contradiction.
The contrapositive method is similar to
contrad iction, in that you beg in by assuming that ".
and NOT B are true. Unlike contradiction however,
you do not work forward from both A and NOT B.
Instead, you work forward only froM t'OT P. Your
objective is to reach the contradiction that A is
false (hereafter written NOT A). ran you ask for a
better contradiction than that? How can be true
and false at the same time?
To repeat, in the contrapositive method, you
assume that and NOT B are true and you work forward
from the statement NOT B to reach the contradiction
that is false. such, the contrapositive method
can be thought of as a more "passive" form of
contradiction in the sense that the assumFtion that
is true passively provides the contradiction. Tn the
contradiction method however, the that A
is true is actively used to reach a contradiction
(see Figure 9).
From Figure 9 you can also see the advantages
and disadvantages of the contrapositive method over
the contradiction method. The of the
72
CONTRAPOSJTJVE METHOD
73
contrapositive method is that you work forward from
only one statement (namely NOT B) instead of two. On
the other hand, the advantage is that you know
precisely what you are looking for (namely NOT A).
Because of this, you can often apply the abstraction
process to the statement NOT A in an attempt to work
backward. The option of working backward is not
available in the contradiction method because you do
not know what contradiction you are looking for.
Method
Con tra
positive
Assume
forward
NOT B}>
Conclude
backward
<NOT A
Contra
diction ••• * (contradiction)
NOT
Fig. 9. Contrapositive method versus
contradiction method.
The next example demonstrates the contrapositive
method.
Example 10. If P and q are positive real numbers such
that ypq is not equal to (p+q)/2, then p is not equal
to q.
Outline of proof. The appearance of the word "not" in
the conclusion should suggest usin9 either the
contrad iction or contraposi tive method. Here, the
contrapositive method will be whereby you will
work forward from the statement NOT B and backward
from the statement NOT A. In this case, NOT B is the
statement "p = q" while NOT A is "{Pq = (p+q)/7." The
objective is to work forward from the fact that p • q
to reach the desired conclusion that ypq = (p+q)/2.
Try to do this for yourself.
74
CONTRAPOSITIVE METHOV
Since you want to conclude that ypq:: (p+q)/2
and you are assuming that p :: q, why not replace q
everywhere by p? In other words, since p:: q,
ypq :: rn = p, and also (p+q) 12 = ,p+p) 12 = p.
Thus, working forward from the assumption that p = q,
it is easy to See that ypq = (p+q) 12 :: p. Note that
the assumption that p > 0 was used in claiMing that
y"pP":: p.
In the following condensed version of the proof,
note that nowhere in the proof does it formally say
that the statement NOT A has been reached.
Proof of Example 10. Assume, to the contrary, that
p :: q. Since p is positive, it then follows that
(p'q = fPP = p = (p+p) 12 = fp+q) 12
and the proof is complete.11
It is also quite interesting to compare the
contrapositive and forwardbackward methods. In the
forwardbackward method you work forward from A and
backward from B, while in the contrapositive method,
you work forward from NOT B and backward from NOT A
(see Figure 10).
If you understand Figuce 1(', then it is not hard
to See why the contrapositive method might be better
than the forwardbackward method. Perhaps you can
obtain more useful information by working NOT B
forward rather than A. Also, it might be easier to
perform the abstraction process on the statement NOT
A rather than on B, as would be done in the
forwardbackward method.
The forwardbackward method arose from
conSidering what happens to the truth of "A implies
8" when A is true and when A is false (recall
Table 1). The contrapositive method arises from
similar considerations regarding B. Specifically, if
B is true, then according to Table 1, the statement
"A implies E" is true. Hence, there is no need to
consider the case when B is true. So suppose B is
false. In order to ensure that "A implies B" is true,
according to Table 1, you would have to show that A
Method
Forward
Backward
Contra
positive
CONTRAPOSITIVE METHOD
Assume
forward
A}>
A
forward
NOT B}>
75
Conclude
backward
(B
backward
(NOT A
Fig. 10. ForwardBackward method versus
contrapositive method.
is false. the contrapositive method has you
aSSume that B is false and try to conclude that A is
false.
Indeed, the statement "A implies B" is logically
equivalent to "NOT B implies NOT A" (see Table 3).
As such, the contrapositive method can be thought of
as the forwardbackward method applied to the
statement "NOT B implies NOT A." condensed
proofs that use the contrapositive method make little
or no to the contradiction method.
In general, it is difficult to know whether the
forwardbackward, contradiction, or contrapositive
method will be more effective for a given problem
without trying each one. However, there is one
instance that often indicates that the contradiction
or contrapositive method should be chosen, or at
least considered seriously. This occurs when the
statement B contains the word "not" in it, for then
you will usually find that the statement NOT B has
some useful information.
In the contradiction method, you work forward
from the two statements and NOT B to obtain a
contradiction. In the contrapositive method you also
reach a contradiction, but you do so by working
forward from NOT B to reach the conclusion NOT A. You
should also, of course, work backward from A. A
76
Method
Forward
Backward
Contra
diction
Contra
positive
CONTRAPOSITIVE METHOD
Assume
forward
A}>
Concl ude
backward
(B
• • • * (contrad iction)
A
forward
NOT B}>
backward
(NOT A
Fig. 11. ForwardBackward method versus
contradiction method versus
contrapositive method.
comparison of the forwardbackward, contradiction,
and contrapositive methods is given in Figure 11.
Both the contrapositive and contradiction
methods require that you be able to write down the
NOT of a statement. The next chapter shows you how
to do so when the statements contain quantifiers.
Exercises
Note: ~ l l proofs should contain an outline of proof
as well as a condensed version.
9.1. If the contrapositive method of proof is to be
used on the following propositions, then what
statement(s) will you work forward from and what
statement should you concl ude?
(a) If n is an integer for which n
2
is even,
then n is even.
CONTRAPOSTTTVE METHOD
77
(b) Suppose that S is a subset of the set T of
real numbers. If S is not bounded then T is
not bounded.
( c) If x and yare real numbers with x y, then
f(x) f (y) (where f is a function of one
variable) •
(d) If the matr ix ft1 is not singular, then the
rows of ft1 are not linearly dependent.
9.2. In a proof by the contrapositive method of the
proposition "Tf r is a real number with r > 1,
then there does not exist a real number t
between 0 and x/4 such that sin(t) = r cos(t),"
which of the following is a result of the
forward process?
(a) r  1 :s; 0
fb) sin
2
(t) = r2(1  sin
2
(t»
(c) 1  r < 0
f d ) ta n ( t) = 1/ r
9.3. If the contrapositive method is used to prove
the proposition "Tf the of the
function f at the point x is not equal to 0,
then x is not a local minimum of f," then which
of the following is the correct abstraction
question? is wrong with the other choices?
(a) How can I show that the point x is a local
minimum of the function f?
(b) How can r show that the derivative of the
function f at the point x is O?
(c) How can I show that a point is a local
minimum of a function?
(d) How can I show that the derivative of a
fUnction at a point is O?
9.4. Prove that if c is an odd integer then the
equation n
2
+ n  c = 0 has no odd integer
solution for n.
9.5. Suppose that m and b are real numbers with m 0
and let f be the function that is defined by
f (x) = mx + b. Prove that, for all x f: y,
f(x) f(y).
9.6. Prove, by the contrapositive method, that if no
angle of a quadrilateral PSTU is obtuse, then
the quadrilateral RSTU is a rectangle.
10 nots of nots lead to knots
As you saw in the prev ious chapter, the
contrapositive method is a valuable proof technique.
To use it, however, you must be able to wrice down
the statement NeT B so that you can work it forward.
Similarly, you have to know exactly what the
statement NOT 'Po is so that you can apply the
abstraction process to it. Tn some instances, the
NOT of a statement is to find. For example, if
A is the statement "the real number x is > n," then
the NOT of is "it is not the case that the real
number x is > 0," or equivalently, "the real number x
is not > 0." Tn fact, the word "not" can be
el iminated al together by incorporating it into the
statement to obtain "the real number x is C."
A more challenging situation arises
statement contains qUAntifiers. For
suppose that the statement B contains the
"for all" in the form:
when the
instance,
quantifier
For all "objects" with "certain Froperty,"
"something haFpens."
Then NOT B is the statement:
It is not the case that, for all "objects" with
the "certain property," "something happens"
which really means that there is an object
certain property for which the
harpen. Eimilarly, if the statement E
contain the quantifier "there is" in the
form:
78
wi th the
does not
were to
standard
NOTS OF NOTS 79
There is an ·object· with the ·certain property·
such that "something happens,·
then NOT 8 is the statement:
It is not the case that there is an "object" with
the "certain property· such that "something
happens,"
or, in other words, for all objects with the certain
property, the something does not happen.
In general, there are three easy steps to
finding the NOT of a statement containing one or more
quantifiers:
Step 1. Put the word NOT in front of the entire
statement.
Step 2. If the word NOT appears to the left of
a quantifier, then move it to the right
of the quantifier and place it just
before the something that happens. As
you do so, you mag ically change the
quantifier to its opposite, so that
"for all· becomes ·there is· and "there
is" becomes "for all."
Step 3. When all of the quantifiers appear to
the left of the NOT, eliminate the NOT
by incorporating it into the statement
that appears immediately to its right.
These steps will be demonstrated with the following
examples.
1. For every real number x ~ 2, (x
2
+ x  6) ~ o.
Step 1. NOT for every real
(x
2
+ x  6) ~ o.
number x ~ 2,
Step 2. There is a real number x ~ 2 such that
NOT (x
2
+ x  6) ~ o.
step 3. There is a real number x
ex2 + x  6) < o.
~ 2 such that
80 NOTS OF NOTS
Note in Step '2 that, when the NOT is passed from
left to right, the quantifier changes but the certain
property (namely x ~ 2) does not! Also, since the
quantifier "for every" is changed to "there exists,"
it becomes necessary to replace the "," by the words
"such that." In a completely analogous manner, if the
quantifier "there exists" is changed to "for all,"
then the words "such that" are removed and a "," is
inserted, as is illustrated in the next example.
'2. There is a real number x ~ 2 such that
(x
2
+ x  6) ~ O.
Step 1. NOT there is a real number x ~ 2 such
that (x
2
+ x  6)
~
o.
Step '2. For all real numbers x ~ 2, NOT
(x
2
+ x  6) ~ O.
Step ? For all real numbers x ~ 2,
(x 2 +
X  6) < o.
Finally, if the original statement contains more
than one quantifier, then Step 2 will have to be
repeated until all of the quantifiers appear to the
left of the NOT, as is demonstrated in the next two
examples.
? For every real number x between 1 and 1, there
is a real number y between 1 and 1 such that
(x
2
+ y2) S 1.
Step 1. NOT for every real number x between 1
and 1, there is a real number y between
1 and 1 such that (x
2
+ y2) :s 1.
Step 2. There is a real number x between 1 and
1 such that, NOT there is a real number
y between 1 and 1 such that
(x
2
+ y2) :s 1.
Step 2. There is a real number x between 1 and
1 such that, for all real numbers y
between 1 and 1, NOT (x
2
+ y2) :s 1.
Step ? There is a real number x between 1 and
1 such that, for all 2real numbers y
between 1 and 1, (x
2
+ Y ) > 1.
NOTS OF NOTS 81
4. There is a real number x between 1 and 1 such
that, for all real numbers y between 1 and 1,
(x
2
+ y2) :;; 1.
Step 1. NOT there is a real number x between 1
and 1 such that, for all real numbers y
between 1 and 1, (x
2
+ y2) :;; 1.
Step 2. For all real numbers x between 1 and 1,
NOT for all real numbers y between 1
and 1, (x
2
+ y2) :s 1.
Step 2. For all real numbers x between 1 and 1,
there is a real number y between 1 and
1 such that NOT (x
2
+ y2) :s 1.
Step Z. For all real numbers x between 1 and 1,
there is a real number y between 1 and
1 such that (x
2
+ y2) > 1.
Another situation where you must be careful is
in taking the NOT of a statement containing the words
AND or CR. Just as the quantifiers are interchanged
when taking the NOT of the statement, so the words
AND and OR interchange. Specifically, NOT fA AND 8]
becomes [NOT A] OR [NOT 8]. Similarly, NOT [A OR 8]
becomes [NOT A] AND fNOT 8). For example:
5. NOT f x 2: 3 AND Y < 2] is [x < 3] OR [y 2: 2].
6. NOT [x 2: 3 OR Y < 2] is [x < 3] AND fy 2: 2].
Remember that, when you use the contrapositive method
of proof, the first thing to do is write down the
statements NOT 8 and NOT A.
Exercises
Note: All proofs should contain an outline of proof
as well as a condensed version.
10.1. Write the NOT for each of the definitions in
82 NeTS OF NOTS
Exercise 5.1. For instance, 5.1 fa) would read
"the real number x· is not a maximum of the
function f if there is a real number x such
that f(x) > f(x·)."
10.2. Reword the following statements so that the
word "not" appears explicitly. For example,
the statement "x > 0" can be reworded to read
"x is not sO."
(a) For each element x in the set S, x is in T.
(b) There is an angle t between 0 and X/2 such
that sin(t) = cos(t).
(c) For every "object" wi th a "certain
property," "something happens."
(d) There is an "object" with a "certain
property" such that "something happens."
10.3. If the contradiction method is used to prove
the following statements, then what should you
assume?
(a) For every integer n ~ 4, n! > n
2
•
(b) A impl ies (8 OR C) •
(c) A implies (e A ~ D C).
(d) If f is a convex function of one variable,
x· is a real number, and there Is a real
number 8 > 0 such that, for all real
numbers x that satisfy the property that
'x  x·, < 8, f(x) ~ f(x·), then for all
real numbers y, fey) ~ f(x·).
10.4. If the contrapositive method is used to prove
each of the following propositions, then what
statement(s) will you work forward from and
what statement(s) will you work backward from?
(a) 'A impl ies (e OR C).
(b) A impl ies (8 AND C) •
(c) If n is an even integer and m is an odd
integer, then either mn is divisible by 4
or n is not divisible by ~ .
10.5. Prove, by contradiction, that if x
real numbers such that x ~ 0,
x + y = 0, then x = 0 and y = o.
and yare
y ~ 0, and
11 special proof techniques
You now have three major proof techniques to
help you in proving that "A impl ies S": the
forwardbackwa rd, the contrapos 1 tive, and the
contradiction methods. In addition, when B has
quantifiers, you have the choose and construction
methods. There are several other special forms of E
that have wellestablished and usually successful
proof techniques associated with them. Three of
these will be developed in this charter.
The first is referred to as the
method and it is associated with a statement B that
not only wants you to show that there is an object
with a property such that something happens,
but also that the object is unique (i.e., it is the
only such object). You will know to use the
uniqueness method when the statement P contains the
word "unique" as well as the quantifier "there is."
In such a case, your first job is to show that
the desired object does exist. This can be done by
either the construction or the contradiction method.
The next step will be to ShO\Ol un iqueness in one of
two standard ways. 'The fi rst approach has you aSSume
that there are two objects having the certain
property and for which the something happens. If
there really is only one such object, then, using the
certain rroperty, the something that and
perhaps the information in you must conclude that
the two objects are one and the same (i.e., that they
are really equal). 'The forwardbackward method is
usually the best way to prove that they are equal.
The process is illustrated in the next examrle.
83
84 SPECIAL
Example 11. If a, b, c, d, e, and f are real numbers
such that (ad  be) 0, then there are unique real
numbers x and y such that (ax + by) = e and
(cx + dy) = f.
Outline of proof. The existence of the real numbers x
and y was established in example 4 viD the
construction method. Here, the unlqueness will be
establishec." by the method described above. Thus, you
assume that (xl 'Yl) and (x
2
'Y2) are two objects wi th
the certain property and for which the something
harpens. Hence one obtains a(x
1
) + b(Yl) = e and
c(x
l
) + d(Yl) = f, and also that a(x
2
) + b(Y2) = e
and c(x
2
) + a(Y2) = f. Using these four equations
and the assumption that A is true, it will be shown,
by the forwardbackward method, that the two objects
(xl 'Yl) and (x
2
'Y2) are equal. Specifically, the
abstraction question is "Bow ci'ln I show that two
pairs of real numbers (namely (xl 'Yl) ant' (x
2
'Y2»
are equal?" Using the definition of equality of
orc."ered pairs (see refinition one answer is to
show that both xl = x
2
and y = y , or equivalently,
that (x,  x
2
) = 0 anc'l (Yl 1. Y2)2= (1. Eoth of these
statements are obtained from the forward process by
applying Some algebraic manipulations to the four
equations and by using the fact that (ad  bc) (1.
Proof of Example 11. The existence of the real
numbers x and y was established in Example 4 via the
construction method. Hence, only the issue of
uniqueness will be addressed. To that end, assume
that (x ,y ) C'lnd (x ,y ) are real numbers satisfying
1 1 2 2
(1) a(x
l
) + b(Yl) e,
( 2) c(x )
1
+ d(Yl)
f,
(3 ) a(x )
2
+ b(Y2)
e,
(4 ) c (X
2
)
+ d(Y2) =
f.
fubtracting
en
from
(1) anC'
)
from ( 2) yielc'ls
X 2) + b (y 1  Y 2 ) ) (), c nc
X
2
) + d(Yl  Y2») = o.
SPECIAL TECHNIQUES 85
On multiplying the first of these equations by d and
the second one by b, and then subtracting the second
equation from the first one, it follows that
[(ad  bc) (x,  x
2
)] = O. From the hypothesis that
(ad  bc) , 0, one has (x,  x
2
) = 0, and hence
x, = x2' A similar sequence of algebraic
manipulations will establish that y = y, and thus
the uniqueness is proved.// '2
The second method for showing uniqueness has you
assume that there are two different objects having
the certain property and for which the something
happens. Now supposedly this cannot happen, so, by
using the certain property, the something that
happens, the information in and especially the
fact that the objects are different, you must then
reach a contradiction. This process is demonstrated
in the next example.
Example 12. If r is a positive real number then there
is a unique real number x such that x
3
= r.
OUtline of The appearance of the quantifier
"there is" In the conclusion suggests using the
construction method to produce a real number x such
that x
3
= r. This part of the proof will be omitted
so that the issue of uniqueness can be addressed. To
that end, suppose that x and yare two different real
numbers such that x
3
= rand y3 = r. Using this
information together with the hypothesis that r is
positive, and especially the fact that x, y, a
contradiction will be reached by showing that r = 0,
thus contradicting the hypothesis that r is positive.
To show that r = 0, one can work forward. In
particular, since x
3
= rand y3 = r, it follows that
x
3
= y3. Thus (x
3
 y3) = 0, and on one
obtains [(x  y) (x
2
+ xy + y2)] = O. Here is where
you can use the fact that x y to divide by (x  y) ,
obtaining (x
2
+ xy + y2) = O. Thinking of this as a
quadratic equation of the form (ax
2
+ bx + c) = 0 in
which a = 1, b = y, and c = y2, the quadratic formula
states that
x = =
2 2
88
TECHNIQUES
Since x is real and the above formula for x requires
taking the square root of _3y2, it must be that
y • 0, and if y. 0, then r z y3 = 0, and the
contradiction has been reached.
Proof of Example 12. Only the issue of uniqueness
will be addressed. To that end, assume that x and y
are two different real numbers for which x
3
= rand
y3 '" r. Hence, it follows that 0 II: (x
3
_ y3) ..
rex  y) (x
2
+ xy + y2)]. Since x y, it must be
that (x
2
+ xy + y2) '" O. By the quadratic formula,
it follows that
y+ VCy24y2)
x = =
2 2
Since x is real, it must be that y = C. But then
r = y3 = 0, thus contracicting the hypothesis that r
is positive./ /
special proof technique, called the
method, arises when B is of the form
"either r is true, or else D is true" (where C and C
are statements). In other \'lords, the either/or
method is to be used when trying to show that the
statement "A implies r on t" is true. Applying the
forwardbackward method, you would begin by assuming
A is true and you would like to conclude that either
C is true or else 0 is true. Suppose that you were
to make the additional assumption that C is not true.
Clearly it had better turn out that, in this case, 0
is true. Thus, with the either/or method, you can
assume that A is true and C is false and you must
then conclude that C is true, as is illustrated in
the next example.
Example 13. If x
2
 5x + 6 :!: 0, then x :s 2 or x :!: 3.
Outline of proof. From the above discussion, you can
assume that (x
2
 5x + 6) :!: 0 and x > 2. It is your
job to conclude that x :!: 3. forward from the
assumption that (x
2
 5x + 6) 0, it fo11o,""s that
[(x  2)(x  3») :!: O. Since x > 2, (x  2) > (I, and
hence (x  3) :!: (1 Ci.e., x a 3, as desired).
SPECIAL TECHNIQUES
Proof of Example 13. Assume that (x 2  5x
and x > 2. It follows that (ex  2)(x
and, since (x  2) > 0, it must be that
desired.//
87
+ 6) ~ 0
 3)] ~ 0,
x :::: 3 as
It is worth noting that the either/or method
could have been done equally well by assuming that A
is true and D is false, and then concluding that C is
true. Try this approach on Example 13.
The final proof technique to be developed in
this chapter is the ma.'x./m.i.n method that arises in
problems dealing with maxima and minima. Suppose
that S is a nonempty set of real numbers having both
a largest and a smallest member. For a given real
number x, you might be interested in the position of
the set S relative to the number x. For instance,
you might want to prove one of the following
statements:
1. All of S is to the right of x (see Figure 12(a».
2. Some of S is to the left of x (see Figure 12(b».
3. All of S is to the left of x (see Figure 12 fc) ) •
4. Some of S is to the right of x (see Figure 12fd».
In mathematical problems these four statements
are likely to appear, respectively,
(a) min{s: s is in S} :::: x
(b) min{s: s is in S} :s x
(c) max { s: sis inS} :s x
(d) max { s: sis inS} :::: x
as:
The proof technique associated with the first two are
discussed here and the remaining two are left as
exercises. Nonetheless, the idea behind the max/min
technique is to convert the given problem into an
equivalent problem containing a quantifier. Then,
the appropriate choose or construction method can be
used.
Consider, therefore, the problem of determining
if the smallest member of S is :::: x. An equivalent
problem containing a quantifier can be obtained by
considering the corresponding statement (1) above.
Since "all" of S should be to the right of x, you
88 SPECIAL TECHNICUES
need to show that, for all elements s in S, x s, as
is illustrated in the next example.
S
Real line
o x
Pig. 12(a). All of S to the right of x.
S
[
3
Real line
o x
Pig. 12(b). of S to the left of x.
S
+1+[+3+ neal line
o x
Pig. 12(c). of S to the left of x.
s
neal line
o x
rig. 12(<.'1). Eome of S to the right of x.
SPECIAL TECHNICUES
89
Example 14. If R is the set of all real numbers, then
min{x(x  1): x is in Il} ;:: 1/4.
Outline of proof. From the form of B, the maximin
method will be used. ~ s per the discussion above,
the statement P can be converted to "for all real
numbers x, x(x  1) ~ 1/4." Once in this form, it
becomes clear that the choose method should be used
to choose a real number x for which it must be shown
that x(x  1) :! 1/4 or, that (x
2
 x + I/Ll) ;:: O.
But, since (x 2  X + 1/4) = ex  1/2)2 , this number
is always;:: r, and the proof is complete.
Proof of Example 14. In order to conclude that
min{x(x  1): x is in Il} ;:: 1/4, let x be any real
number. Then it follows that x(x  1) is ~ 1/4
because (x
2
 x + lit') = ex  1/2)2, which is ;:: C.II
Turning now to the problem of showing that the
smallest member of ~ is s x, the approach is slightly
different. 70 proceed, consider the corresponding
statement (2) above. Since "some" of S should be to
the left of x, an equivalent problem is to show that
there is an element s in S such that s s x. !hen,
the construction or contradiction method could be
used.
This chapter has described three srecial proof
techniques that are appropriate when E has the
corresponding special form. The final chapter
provides a comrlete summary.
Exercises
Note: All proofs shoulo contain an outline of proof
as well as a concensed version.
11.1. Prove that if x is a real number that is > 2,
then there is a unique real number y < 0 such
that x = 2y/(1 + y) •
90 SFECIAL TECHNICUE5
11.2. Prove, by using the second uniqueness method,
that if m and b are real nUMbers with m ; 0,
then there is a unique number x such that
mx + b = O.
11.3. Frove that if a and b are real numbers, at
least one of which is not r, and i = y:I, then
there is a unique complex number, say c + di,
such that (a + bi) (c + di) = 1.
11.4. What would be the advantages and disadvantages
of using the contradiction instead of
the either/or method to prove that "A implies
(E OR C)?"
11.5. Prove that if n is an even integer and m is an
odd integer, then either 4 divides mn or 4
not divide n.
11.6. Consider the proposition "If x is a real number
that satisfies x
3
+ 3x
2
  27 0, then
Ixl 3."
(a) Reword the rroposition so that it is of the
form "A implies OR C."
(b) Prove the proposition by assuming that A
and NOT B are true.
(c) Prove the proposition by assuming that
and C are true.
11.7. Convert the following max/min problems into the
appropr iate Quanti fier statement. (Note: S is
a set of real numbers and x is a given real
number. )
(a) maxIs: s is in S} !!i x
(b) maxIs: s is in 5} x
In the remainder of this problem, a, b, c, and
u are given real numbers, and x is a variable.
(c) min{cx: aX!!i b and x OJ:s u
(d) max{cx: ax b and x OJ u
(e) min{ax: b:S x :s c} u
(f). max{bx: a!!i x :s cJ :s u
11.8. Prove that if S is a nonemFty subset of a set T
of real numbers and t· is a real number that
satisfies the Froperty that for each element t
in T, t t*, then minIs: s is in S} t"'.
SPECIAL TECHNICUES 91
11.9. Suppose that a, b, and c are given real numbers
and that x and u are variables. Prove that
min{cx: ax ::: b, x ::: OJ is at least as large as
max {ub: ua '5 c, u ::: C J •
12 summary
The list of proof techniques is now complete.
The techniques presented here are by no means the
only ones, but they do constitute the basic set.
Undoubtedly you will come across others as you are
exposed to more mathematics. Perhaps you will
develop some of your own. In any event, there are
many fine points and tricks that you will pick up
with experience. A final summary of how and when to
use each of the various techniques for proving the
proposition " ~ implies B" is in order.
With the forwardbackward method, you can assume
that A is true and your job is to prove that B is
true. Through the forward process, you will derive
from A a sequence of statements AI, A2, • • • which
are necessarily true as a result of A being assumed
true. This sequence is not random. It is guided by
the backward process whereby, through asking and
answering the abstraction question, you derive from B
a statement, El, with the property that if it is
true, then so is B. This abstraction process can
then be applied to Bl, obtaining a new statement, B2,
and so on. The objective is to I ink the forward
sequence to the backward sequence by generating a
statement in the forward sequence that is precisely
the same as the last statement obtained in the
backward sequence. Then, like a column of dominoes,
you can do the proof by going forward along the
sequence from A all the way to B. When obtaining the
sequence of statements, watch for quantifiers to
appear, for then the construction, choose, induction,
and/or specialization methods may be useful in doing
the proof.
92
SUMMARY
93
For instance, when the quantifier "there is"
arises in the backward process in the standard form:
There is an "object" with a "certain property"
such that "something happens"
you should consider using the construction method to
actually produce the desired object. With the
construction method, you work forward from the
assumption that A is true to construct (produce, or
devise an algorithm to produce, etc.) the object. Be
sure to check that it satisfies the certain property
and also that the something happens.
On the other hand, when the quantifier "for all"
arises in the backward process in the standard form:
For all "objects" with a "certain property,"
"something happens"
you should consider using the choose method. Here,
your objective is to design a proof machine that is
capable of taking any object with the certain
property and proving that the something happens. 70
do so, you select (or choose) an object that does
have the certain You must conclude that,
for that object, the something happens. Once you
have chosen the object, it is best to proceed by
working forward from the fact that the chosen object
does have the certain property (together with the
information in A, if necessary) and backward from the
something that happens.
The induction method should be considered (even
before the choose method) when the statement 8 has
the form:
For every integer n greater than or equal to
some initial one, a statement P(n) is true.
The first step of the induction method is to verify
that the statement is true for the first possible
value of n. The second step requires you to show
that if Prn) is true, then P(n+l) is true. Remember
that the success of a proof by induction rests on
your ability to relate the statement P(n+l) to P(n)
94 SUMMARY
so that you can make use of the that Pen)
is true. In other words, to perform the second step
of the induction proof, you should write down the
statement pen), replace n everywhere by (n+l) to
obtain Pfn+l), and then see if you can express pen+l)
in terms of pen). Only then will you be able to use
the assumption that pen) is true to reach the desired
conclusion that Pfn+l) is also true.
Finally, when the quantifier "for all" arises in
the forward process in the standard form:
For all "objects" with a "certain property,"
"something happens"
you will probably want to use the specialization
method. To do so, watch for one of these objects to
arise in the backward process, for then, by using
specialization, you can conclude that the something
does happen for the particular object that is under
consideration. That fact should then be helpful in
reaching the conclusion that B is true. When using
specialization, be sure to verify that the particular
object does satisfy the certain property, for only
then will the something happen.
When the original statement B contains the word
"not," or when the forwardbackward method fails, you
should consider the contrapositive or contradiction
method, with the former being chosen first. To use
the contrapositive approach, you must immediately
write down the statements NOT B and NOT A, using the
techniques of Chapter 10 if necessary. Then, by
beginning with the assumption that NOT B is true,
your job is to conclude that NOT A is true. This is
best accomplished by applying the forwardbackward
method, working forward from the statement NOT Band
backward from the statement NOT A. Once again,
remember to watch for quantifiers to appear in the
forward or backward process, for if they do, then the
corresponding construction, choose, induction, and/or
specialization methods may be useful.
In the event that the contrapositive method
fails, there is still hope with the contradiction
method. With this approach, you are allowed to
SUMMARY 95
assume not only that A is true but also that B is
false. This gives you two facts from which you must
derive a contradiction to something that you know to
be true. Where the contradiction arises is not
always obvious, but it will be obtained by working
the statements A and NOT B forward.
In trying to prove that "A implies B," let the
form of e guide you as much as possible.
Specifically, you should scan the statement e for
certain key words, as they will often indicate how to
proceed. For example, if you come across the
quantifier "there is," then consider the construction
method, whereas the quantifier "for all" suggests
using the choose or induction method. When the
statement B has the word "not" in it, you will
probably want to use either the contrapositive or
contradiction method. Other key words to look for
are "uniqueness," "either ••• or ••• ," and
"maximum" and "minimum," for then you would use the
corresponding uniqueness, either/or, and max/min
methods. If you are unable to choose an approach
based on the form of B, then you should proceed with
the forwardbackward method. 1able 4 provides a
complete summary.
You are now ready to "speak" mathematics. Your
new "vocabulary" and "grammar" is complete. You have
learned the three major proof techniques for proving
propositions, theorems, lemmas, and corollaries.
They are the forwardbackward, contrapositive, and
contradiction methods. You have come to know the
quantifiers and the corresponding construction,
choose, induction, and specialization methods. For
special situations, your bag of proof techniques
includes the uniqueness, either/or, and max/min
methods. If all of these proof techniques fail, you
may wish to stick to Greekafter all, it's all Greek
to me.
96
SUMMARY
Table 4. Summary of Proof Techniques.
*****************************************************
Proof
Techn ique
Forward
Backward
(page 8)
Contra
posi tive
(page 72)
Contra
diction
(page 64)
Construc
tion
(page 34)
Choose
(page 40)
Induction
(page SO)
When to Use It
As a first attempt,
or when B does not
have a recognizable
form
When B has the
word "not" in it
When B has the
word "not" in it,
or when the first
two methods fail
When E has the term
"there is," "there
exists," etc.
When B has the term
"for all," "for
each," etc.
When B is true
for each integer
beginning with an
initial one, say no
What to Assume
NOT B
A and
NOT B
11, and choose
an object with
the certain
property
The statement
is true for n
*****************************************************
SUMMARY 97
*****************************************************
What to Conclude
B
NOT 'A
Some contra
diction
There is the
desired object
That the some
thing happens
The sta temen t
is true for
n + 1. 'Also show
it true for no'
HoW to Do It
Work forward from ~ and apply
the abstraction process to B
Work forward from NOT Band
backward from NOT 'A
Work f o r w ~ r d from ~ and NOT B
to reach a contradiction
Guess, construct, etc., the
object having the certain
property and show that tre
something happens
Work forward from A and the
fact that the object has the
certain property. 'Also work
backward from the something
that happens.
First substitute no_ everywhere
and show it true. Then invoke
the induction hypothesis for
n to prove it true for n + 1.
*****************************************************
98
SUMfillARY
Table 4. Summary of Proof Techniques (continued).
*****************************************************
Proof
Technique
Speclal
1 zation
(page 5 ~ )
Unique
ness 1
(page 83)
Unique
ness 2
(page 85)
Either/
or 1
(page 86)
Either/
or 2
(page 87)
JYlax/min 1
(page R7)
fIIIax/min 2
(page 89)
When to Use It
When }. has the
term "for all,"
"for each," etc.
When B has the
word "unique"
in it
When E has the
word "unique"
in it
When B has the
form "C OR 0"
When E has the
form "C OR 0"
When B has the
form "max S ~ x"
or "min S 2: x"
When E has the
form "max S 2: x"
or "min S ~
x"
What to Assume
There are two
such obj ec ts,
and A
There are two
different
objects, and A
A and NOT C
A and NOT D
Choose an
s in S,
and A
,..
*****************************************************
SUMMARY
99
*****************************************************
What to Conclude
B
'J'he two objects
are equal
Some contra
diction
D
C
s :s x
or
s ?: x
Construct s in
S so that
s ::: x
or
s :s x
How to Do It
Work forward by specializing
A to one particular object,
the one obtained in the
backward process
Work forward using A and the
properties of the objects.
Also work backward to show
the objects are equal.
Work forward from A using the
properties of the two objects
and the fact that they are
different
Work forward from A and
NOT C, and backward from r
Work forward from A and
NOT C, and backward from C
Work forward from A and the
fact that s is in S. ~ l s o
work backward.
Use ~ and the construction
method to produce the
des ired sin S
*****************************************************
100 SUMMARY
Exercises
12.1. For each of the following statements, indicate
which proof technique you would use to begin
the proof and explain why.
(a) If p and ~ are odd integers then the
equation x + 2px + 2q = 0 has no rational
solut ion for x.
(b) For every integer n ~ 4, nl > n
2
•
ec) If f and 9 are convex functions then f + g
is a convex function.
(d) If a, b, and c are real numbers, then the
maximum value of ab + bc + ca subject to
the condition that a
2
+ b
2
+ c
2
= I is s 1.
(e) In a plane, there is one and only one line
perpendicular to a given line t through a
point P on the line.
(f) If f and 9 are two functions such that (1)
for all real numbers x, f(x) 5 g(x) and (2)
there is no real number ~ such that, for
all x, f(x) s M, then there is no real
number M > 0 such that, V x, g(x) 5 M.
(g) If f and 9 are continuous functions at the
point x, then so 1s the function f + g.
(h) If f and 9 are continuous functions at the
point x, then for every real number £ > 0
there is a real number 8 > 0 such that, for
all real numbers y with Ix  yl < 8,
If(x) + g(x)  (f(y) + g(y»1 < £.
(i) If f is a function of one variable defined
by f(xl = 2· + x
2
/2, then there is a real
number x· between 0 and I such that, for
all y, f(x·) S fey).
12.2. Describe how you would use each of the
following proof techniques to prove that "for
every integer n ~ d, nl > n
2
" State what you
would assume and what you would conclude.
(al Induction method
(b) Choose method
(cl ForwardBackward method (Hint: Convert the
problem to an equivalent one of the form
"if ••• then. .")
(d) Contradiction method
A putting it aU togetherl
Through the use of a specific example, this
appendix will show you how to read and how to
understand a written proof as it might appear in a
textbook. Subsequently, some general problemsolving
strategies will be ~ i s c u s s e d and demonstrated on a
problem.
Due to the way in which most proofs are
currently written, it is necessary for you to figure
out which proof techniques are being used and why.
Often, what makes a written proof so difficult to
understand is the fact that the author has omitted
part of the thought process, thus forcing you to
reconstruct it. Doing so requires that you identify
which proof technique is being used and how it
applies to the particular problem. The next example
demonstrates how to read a proof. Unlike previous
examples in this pamphlet, the condensed version of
the proof will precede the outline, which then
explains how to read the condensed proof.
The example deals with the concept of an
"unbounded" set of real numbers, meaning that there
are elements in the set that are "arbitrarily" far
away from O. A formal definition of a bounded set
will be given first.
Definition 17. A set S of real numbers is bounded if
there is a real number M > 0 such
that, for all elements x in S,
I x I < M.
Example 15. Jf S is a subset of a set T of real
numbers and S is not bounded, then T is not bounded.
Proof of Example 15. (For reference purposes, each
sentence of the proof will be written on a separate
line.)
81. Suppose that S is a subset of T and assume to the
contrary that T 1s bounded.
101
102 APPENDIX A
82. Hence there is a real number
MI
> 0
such that,
for a11 x in T,
I xl <
MI.
83. It will be shown that S is bounded.
84. To that end, let
Xl
be an element of s.
85. 9ince 9 is a subset of T, it follows that
Xl
i,s
in T.
86. But then I Xl I
< MI
and so 9 is bounded, which is
a contradiction.11
OUtline of proof. An interpretation of each of the
statements 91 through 96 will be given.
Interpretation of 81. The author is indicating that
the proof is going to be done by contradiction and
appropriately is assuming that (part of) A is true,
that is, that S is a subset of T, and also that NOT B
is true, that is, that T is bounded. Note that the
author has not explicitly stated that 9 is not
bounded, although it is being assumed true.
The use of the contradiction method should not
come as a surprise since the conclusion of the
example contains the word "not." Sometimes it is
beneficial to scan through the proof to find the
contradiction. In 96 it says that S is bounded, and
this contradicts the hypothesis that 9 is not
bounded. Hopefully, statements 92 through S6 will
indicate how the author has worked forward from A and
NOT B to reach the contradiction.
Interpretation of 82. The author is working
from the statement NOT B (i.e., that T is a
set) via Definition 17 to claim that there
real number MI > 0 such that, for all
I x I < ,.,1.
forward
bounded
is the
x in T,
Interpretation of 83. It is here that the author
states the desired objective of showing that S is
bounded. Observe that there is no indication that
the reason for wanting to show that 9 is bounded is
to reach a contradiction.
APPENDIX A
103
Interpretation of S4. This statement might seem
cryptic at first, and the reason is that the author
has omitted several steps of the thought process.
What has happened is that the author has implicitly
posed the abstraction question "How can I show that a
set (namely S) is bounded?" and has then used
Definition 17 to answer it, whereby it must be shown
that there is a real number > 0 such that, for all
x in S, Ixl < M.
Recognizing the quantifier "there is," the
construction method should be used to produce the
desired value What the author has failed to
tell you is that the value for f.4 (see S2).
Given that this is the case, it must be shown that
has the desired properties, i.e., that > 0 'which
it is) and that, for all x in S, Ixl < M'.
Recognizing the appearance of the quantifier
"for all" in the backward process, one should then
proceed by the choose method. This is precisely what
the author has done in S4, where it states:
" ••• let x' be an element of S." Once the element
x' in S has been chosen, it must then be shown that
Ix'i < M', which the author has not explicitly said
he would do. Observe that the word "let" in S4
indicates that the choose method has been used.
What makes 84 so difficult to understand is the
lack of sufficient detail, particularly the omission
of the abstraction question and answer, and several
other steps. Such details are often left for you to
figure out for yourself.
Interpretation of SSe In S5, the author has suddenly
switched to the forward process and is working
forward from].. (1.e., from the fact that 8 1S a
subset of T) by using the definition of a subset to
conclude that, for all elements x in S, x is in T
(see Definition 14). Then, without stating so, the
author has specialized this statement to the
particular value of x' that is in S, thus concluding
that x, is in T. Note once again that part of the
author'S thought process has been omitted and left
for you to figure out. At this po int, it is not
clear why it is necessary to show that Xl is in T.
104
APPENDIX A
The reason should be related to
Ix' I < .....
showing that
Interpretation of S6. It is here that the author
finally concludes that Ix' I < M', thus showing that S
is bounded and reaching a contradiction to the
hypothesis that S is not bounded. The only question
is how did the author reach the conclusion that
Ix'i < "". Once again, the specific proof technique
has not been mentioned. Actually, the author has
used specialization. Specifically, the statement S2
has been specialized to the particular value of x,.
However, recall that, when using specialization, it
is necessary to verify that the particular object
under consideration (x' in this case) has the certain
property (that x' is in T). The fact that x' is in T
was established in ~ 5 , but observe that, at the time,
the author did not tell you why it was being shown
that x' is in T.
Read the condensed proof again. Observe that
what makes it difficult to understand is the lack of
reference to the specific techniques that were used,
especially the abstraction process and the
specialization method. In addition, the author has
omitted part of the thought process. To understand
the proof you have to fill in the missing steps.
In summary, when reading proofs, expect to do
some work of your own. Learn to identify the various
techniques that are being used. Begin by trying to
determine if the forwardbackward or contradiction
method is the primary technique. Then try to follow
the methodology associated with that technique. Be
watchful for quantifiers to appear, for then the
corresponding choose, induction, construction, and/or
specialization methods are likely to be used. Be
particularly careful of notational complications that
can arise with the choose and specialization methods.
For instance, in the condensed proof of Example 15,
the author could have used the symbols "x" and " ~ "
instead of "x'" and "M'." If this had been done, then
you would have had to distinguish between such
statements as "for all x with ~ certain property,
something happens," in which "x" refers to a general
object, and a statement of the form "let x have a
APPENDIX A 105
certain in which refers to a specific
object. The double use of a symbol is confusing but
common in proofs, as is illustrated in the example in
Appendix S.
In any event, when you are unable to follow a
particular step of a written proof, it is most likely
due to the lack of suffic ient detail. 70 fill in the
gaps, learn to ask yourself how you would proceed to
do the proof. Then try to see if the written proof
matches your thought process.
Now tha t yo u know how to read a proof, it is
time to see how to do a proof on your own. In the
example that follows, pay particular attention to how
the form of the statement under consideration
dictates the technique to be used.
Example 16. If T is a bounded subset of real numbers,
then any subset of T is bounded.
Outline of proof. Looking at the statement S, you
should recognize the quantifier any" and thus
you should begin with the choose method.
you would write: "Let S be a subset of
T. It will be shown that S is bounded." Then the new
statement to be proved is:
B1: S is bounded.
Not recognizing any special form to Bl, it is
probably best to proceed with the forwardbackward
method. The idea is to work backward from Bl until
it is no longer fruitful to so. Then, the forward
process will be applied to the hypothesis that T is a
bounded set of real numbers.
Working backward from Bl, can you pose an
abstraction question? One such question is "How can
I show that a set (namely S) is bounded?" 1\
reasonable way to answer an abstraction question is
by a definition. r'oing so in this case means you
must show that:
B2: There is a real number > 0 such that, for all
elements x in S, Ixl <
108 APPENDIX A
The appearance of the quantifier "there is"
suggests using the construction method and also
indicates that you should turn to the forward process
to produce the desired value Working forward
from the hypothesis that T is bounded, is there
anything that you can say is true? At least one
possibility is to use the definition, whereby you can
conclude that:
AI: There is a real number > 0 such that, for all
elements x in T, Ixl <
Observe that the symbol has been used instead of
because was already used in the backward
process.
Perhaps is the desired value When
using the construction method, 1 t Is generally
advisable to try to construct the desired object in
the most "obvious" way. In this case, that means
setting M = If this "guess" is correct, then you
must still show that the certain property holds
(i.e., M > 0) and also that the something happens
(I.e., that for all x in S, Ixl < It is not hard
to see that > 0 because > 0 and = M'. Thus it
remains only to show that:
83: For all elements x in S, I xl <
The appearance of the quantifier "for all"
suggests proceeding with the choose method, whereby
you would choose an element, say x', in S for which
you must show that:
84: Ix'i or, since M = M', Ix'i <
Observe that the symbol "x'" has been used for a
particular object to avoid confusion with the symbol
"x." At this point, it might not be clear how to
reach the conclusion that Ix'i < M', but look
carefully at the statement AI. Observe that you
could conceivably use specialization to reach the
desired conclusion that Ix'i < M', provided that you
can specialize the statement to x = x'. In order to
be able to do so, you must make sure that the
particular object under consideration (namely x')
APPENDIX A
107
does satisfy the certain property (namely that x' is
in T). Your objective, therefore, is to show that x'
is in T.
Recall that x' was chosen to be an element of S
and that S was chosen to be a subset of T. How can
you use this information to show that x' is in T?
Working forward from the definition of a subset, you
know that, for a11 x in S, x is in T. Thus, you can
specialize this statement to x = x', which you know
is in S, and reach the desired conclusion that x' is
in T, completing the proof.
Proof of Example 16. Let S be a subset of T. To show
that S is bounded, a real number ~ > 0 will be
produced with the property that, for all x in S,
I x I < ~ .
By hypothesis, l' is bounded, and so there is a
real number ~ , > 0 such that, for all x in T,
Ixl < ~ ' . Set ~ = M'. To see that ~ has the desired
properties, let x' be an element of S. Since £ is a
subset of 1', x' is in T. i?ut then it fo11ows that
Ix'i < ~ = M', as desired.11
Problem solving is not a precise science. Some
general suggestions will be given here, but a more
detailed discussion can be found in George Polya's
book How to Solve It (Princeton University Press,
Princeton, NJ, 1915) orin Wayne Wickelgren's book
How to Solve Problems (W. H. Freeman, San Francisco,
CA, 1 9 7 ~
\\'hen trying to prove that "'II implies B,"
consciously choose a technique based on the form of
B. If no form is apparent, then it is probably best
to proceed with the forwardbackward method. If
unsuccessful, there are several avenues to pursue
before giving up. You might try asking yourself why
B cannot be false, thus leading you into the
contradiction (or contrapositive) method.
The point is that, when you are unable to
complete a proof for one reason or another, it is
advisable to seek another proof technique
consciously. Sometimes the form of the statement B
108
APPENDIX A
can be manipulated so as to induce a different
technique. For instance, suppose that the statement
B contains the quantifier "for all" in the standard
form:
For all "objects" with a "certain property,"
"something happens."
If the choose method fails, then you might try
rewriting B so that it reads:
There is no "object" with the "certain property"
such that the "something" does not happen.
In this form, the contradiction method suggests
itself because of the appearance of the word "no,"
whereby you would assume that there is an object with
the certain property such that the something does not
happen. Your objective would be to reach a
contradiction.
When the statement B does have a recognizable
form, it is usually best to try the corresponding
proof technique first. Remember that, as you proceed
through a proof, different techniques will be needed
as the form of the statement under consideration
changes.
If you are really stuck, it can sometimes be
advantageous to leave the problem for a while, for
when you return to it, you might see a new approach.
Undoubtedly you will learn many tricks of your own as
you solve more and more problems.
B putting it aU togetherll
This appendix will give you more practice in
reading and doing proofs. The examples presented
here are more difficult than those in Appendix A.
The additional difficulty is due, in part, to the
fact that the statements under consideration contain
three quantifiers as opposed to two.
The first example is designed to teach you how
to read a condensed proof as it might appear in a
textbook. As such, the condensed version of the
proof will precede the detailed outline which then
explains how to read the condensed proof. The
example deals with the concept of a "continuous"
function of one variable, meaning that if the value
of the variable is changed "slightly" then the value
of the function does not change "radically." Another
way of saying that a function is continuous is that
its graph can be drawn without lifting the pencil.
To proceed formally, a definition of continuity is
given.
Definition 18. A function f of one variable is
continuous at the point x if for every
real number & > 0, there is a real
number 8 > 0 such that, for all real
numbers y with the property that
Ix  yl < 8, it then follows that
If(x)  fey) I < &.
Example 17. If f and 9 are two functions that are
continuous at x, then the function f + 9 is also
continuous at x. (Note: The function f + 9 is the
function whose value at any point y is fCy) + g(y».
Proof of Example 17. (For reference purposes, each
sentence of the proof will be written on a separate
line. )
S1. To see that the function f + 9 is continuous at
x, let & > o.
109
110
APPENDIX B
S2. It will be shown that there is a 8 > 0 such that,
for all y with Ix  yl < 8, If(x) + g(x)
(f(y) + g(y»1 < &.
83. Since f is continuous at x,
such that, for all y
If(x)  f(y) I < &/2.
there is a 8
,
> 0
wi th I x  y I < 8
,
'
S4. Similarly, since 9 is continuous at x, there is a
8
2
> 0 such that, for all y with Ix  yl < 8
2
,
Ig(x)  g(y) I < & /2.
SSe Let 8 ,. min{ 8
"
8
2
} (which is > 0) and let y have
the property that I x  y I < 8 •
S6.Then Ixyl<8
,
and Ixyl<8
2
, and so
If(x) + g(x) ff(y) + g(y» I :s If(x)  f(y) I +
Ig(x)  g(y) I < &/2 + &/2 = &, and the proof is
complete.//
OUtline of proof. An interpretation of each of the
statements 81 through 86 will be given.
Interpretation of Sl. This statement indicates that
the forwardbackward method is going to be used
because the statement "to see that f + 9 is
continuous at x" means that the author is about to
show that the statement e is true. The question is
why did the a uthor say" ••• let & > O?" Hopefully
it is related to showing that f + 9 is continuous at
x. What has happened is that the author has
implicitly posed the abstraction question "How can I
show that a function (namely f + 9) is continuous at
a point (namely x)?" and has used Definition 12 to
answer it, whereby it must be shown that, for every
& > 0, "something happens."
Recognizing the appearance of the quantifier
"for every" in the backward process, one should then
proceed with the choose method, whereby a specific
value of & > 0 would be chosen. The author has done
this ~ n e n he says" ••• let & > 0." Observe that
the symbol "&" appearing in the definition of
continuity refers to a general value for & while the
symbol "&" appearing in 51 refers to a specific value
reSUlting from the choose method. The double use of
APPENDIX B 111
a symbol can be confusing but is common in written
proofs.
What makes 51 so difficult to understand is that
the author has omitted part of the thought process,
namely, the abstraction question and answer. 5uch
details are often left for you to figure out for
yourself •
Interpretation of 82. The author is
what he is going to do, but why
show that there is a 8 > 0 such
happens?" Recall that, in 51, the
used to select an £ > 0 and so, it
that, for that &, the something
something is "there is a 8 > 0
82 (see Definition Ie).
stating precisely
should he want to
that "something
choose method was
must be shown
happens. Tha t
" as stated in
Interpretation of 83. The author has turned to the
forward process (having recognized from 82 that the
quantifier "there is" suggests using the construction
method to produce the desired 8 > 0). The author is
working forward from the hypothesis that f is
continuous at x by using Definition Ill. What the
author has failed to tell you is that, in the
definition, the quantifier "for all" arises, and the
statement has been specialized to the value of &/2,
where & is the one that was chosen in Sl. Note,
again, that the symbol "&" in Definition le refers to
a general value for & while the symbol "&" in 51
refers to a specific value for &. /it this moment, it
is not clear why the author decided to specialize to
the value of &/2 as opposed to &. The reason will
become evident in 56.
Interpretation of 84. The author is working forward
from the fact that 9 is continuous at x in precisely
the same way that was done in 53. Keep in mind that,
from 52, the author is trying to construct a value of
8 > 0 for which something happens.
Interpretation of 85. It is here that the desired
value of 8 > 0 is produced. The values of 8
1
and 8
2
,
from 83 and S ~ , respectively, are combined to give
the desired 8, namely, 8 = mint 8 , 8
2
}. It is not yet
clear why 8 has been constructed
1
in this way. In any
112
APPENDIX B
event, it is the author's responsibility to show that
8 > 0 and also that 8 satisfies the desired property
that, for all y with Ix  yl <8, If(x) + g(x) 
If ey) + 9 ey) ) I < I: (see S2).
In S5 the author merely remarks that 8 > 0
(which is true because both 8
,
and 8
2
are > 0).
Also, the reason that the author has said • ••• let
y have the property that Ix  yl < 8" is because the
choose method is being used to show that, for all y
with Ix  yl < 8, If(x) + g(x)  (f(y) + gey» I < 1:.
Note, once again, the double use of the symbol "y."
Nonetheless, to complete the proof, it must be shown
that If(x) + g(x)  (f(y) + g(y» I < 1:.
Interpretation of S6. It is here that the author
concludes that If(x) + g(x) (f(y) + g(y» I < 1:.
The only question is how. It is not hard to see that
If(x) + g(x) (f(y) + g(y» I :s If(x)  fey) I +
Ig(x)  g(y) I because the absolute value of the sum
of two numbers, namely f(x)  fey) and g(x)  g(y),
is always :s. the sum of the absolute values of the two
numbers, but why is If(x)  fey) I + Ig(x)  g(y) I <
1:/2 + 1:/2? Once again the author has omitted part of
the thought process.
What has happened is that specialization has
been used to claim that If(x)  fey) I < 1:/2 and
Igex)  g(y) I < 1:/2. Specifically, statements S3 and
S4 have both been specialized to the particular value
of y that was chosen in SS. Recall that, when using
specialization, one must be sure that the particular
object under consideration (namely y) satisfies the
certain property(in this particular case, Ix  yl < 8
,
and I x  y I < 8
2
). Note that, in S6, the author does
indeed claim tnat Ix  yl < 8
,
and Ix  yl < 8
2
, but
can you see why this is true? Look again at 55.
Since y. was chosen so that I x  yl < 8 and 8 =
mint 8
1
, 8
2
}, it must be that both 8:s 8
,
and 8:s 8
2
,
so indeed I x  y I < 8 and I x  y I < 8
2
, Pence, tHe
author is justified in
'
specia1izing S2 and S4 to this
particular value of y and hence can conclude that
If(x)  fey) I < 1:/2 and Ig(x)  g(y) I < 1:/2. In
other words, the author is justified in claiming that
Ifex) + g(x) (f(y) + g(y»1 :s If(x)  fey)1 +
Ig(x)  g(y) I < 1:/2 + 1:/2 = 1:. Perhaps now it is
APPENDIX B 113
also clear why the author specialized (in 53 and S ~ )
to t/2 instead of t.
Again, note that what makes 5f so hard to
understand is the omission of part of the thought
process. You must learn to fill in the missing links
by determining which proof technique is being used
and consequently what has to be done.
The next example illustrates how to go about
doing your own proof. Pay particular attention to
how the form of the statement under consideration
leads to the correct technique.
Example 18. If f is a function of one variable that,
at the point x, satisfies the property that there are
real numbers c > 0 and 8' > 0 such that, for all y
with Ix  yl < 8', If(x)  fey) I :s clx  yl, then f
is continuous at x.
OUtline of proof. Begin by looking at the statement B
and trying to select a proof technique. Since B does
not have a particular form, the forwardbackward
method should be used. Thus, you are led to the
abstraction question "How can I show that a function
(namely f) is continuous at a point (namely x)?" As
usual, a definition is available to provide an
answer, that being to show that:
81: For all t > 0, there is a 8 > 0 such
all y with Ix  yl < 8, If(x)  fey) I
that,
< t.
for
Looking at Bl, you should recognize the
quantifier "for all" and hence use the choose method
to select a particular value of t > o. It is
advisable to use a symbol other than "t" for the
specific choice so as to avoid confusion. In the
condensed proof you would write "Let t' > 0 •••• "
Once you have selected t' > 0, the choose method
requires you to show that the something happens, in
this cllse:
821 There is a 8 > 0 such that, for all y with
Ix  yl < 8, If(x)  fey) I < t'.
Loo king at B 2 , yo u should recogni ze the
114 APPENDIX B
quantifier "there is" and hence use the construction
method to produce the desired value of 8. When the
construction method is used, it is advisable to turn
to the forward process to produce the desired object.
Working forward from the hypothesis that there
are real numbers c > 0 and 8' > 0 for which something
happens, you should attempt to construct 8. Perhaps
8 = 8' tor perhaps 8 = c). Why not "guess" that
8 = 8' and see if it is correct? If not, then maybe
you will discover what the proper value for 5 should
be. To check if the current guess of 8 = 8' is
correct, it is necessary to see if the certain
property in 82 holds for 8' and also if the something
happens. From B2, the certain property is that
8 > 0, but since 8 = 8' and 8' > 0, it must be that
8 > O. Thus, it remains only to verify that, for
8 = 8':
83: For all y with Ix  yl < 8, If(x)  f(y)1 < c'.
Looking at B ~ , you should recognize the
quantifier "for all" in the backward process and you
should use the choose method to select a particular
value of y, say y', with Ix  y' I < 8. The choose
method then requires you to show that the something
happens, in this CBse:
84: If(x)  f(y')1 < c'.
To see if B ~ is true, the hypothesis still has
some unused information. Specifically, the number
c > 0 has not been used, nor has the statement:
Al: For all y with Ix  yl < 8', If(x)  fey) I s
clxyl.
Upon recognlzlng the quantifier "for all" in the
forward process, perhaps you should consider using
specialization; the only question is which value of
y to specialize to. Why not try the value of y' that
was obtained in the backward process? Fortunately,
y' was chosen to have the certain property in Al
(i.e., Ix  y'l < 8 and 8 = 8', so Ix  y'l < 8'),
and therefore, you can specialize Al to y',
obta in ing :
APPENJ)IX B
115
A2: If(x)  fey') 1$ clx  y' I.
Recall that the last statement obtained in the
backward process was Bb, so the idea is to see if A2
can be made to look more like B ~ . For instance,
since y' was chosen so that Ix  y' I < 8, A2 can be
rewritten as:
A3: If(x)  f(y')1 s c8.
B4 would be obta ined if c 8 were known to be
< e'. Unfortunately, 8 = 8', but you do not know
that c8' < e'. This means that the original guess of
S = 8' was incorrect. Perhaps 8 should have been
chosen> 0 so that c8 < e', or equivalently, since
c > 0, why not choose C < 8 < e' /c?
In order to see if the new "guess" of 0 < 8 <
e'/c is correct, it is necessary to check if, for
this value of 8, the certain property in B2 holds and
also whether the something happens. Clearly 8 has
been chosen to be > C and thus has the certain
property. It remains only to verify that the
something happens, in this case, that E2 is true.
As before, one would proceed with the choose
method to select y' wi th I x  y' I < 8, and again one
is led to showing that E.e1 is true. The previous
approach was to specialize III to y'. This time,
however, there is a problem because y' is not known
to satisfy the certain property in AI, i.e., that
I x  y' I < 8'. Unfortunately, all you c10 know is
that Ix  y'l < 8 and 0 < 8 < e'/c. If only you
could apply specialization to AI, then you would
obtain precisely'll:, and finally, since c8 < e', you
could reach the desired conclusion that E4 is true.
Can you figure out how to choose 8 so that both
c8< e' and so that Al can be specialized to y'? The
answer is to choose 0 < 8 < min{8', e'/c}, for then,
when y' is chosen wi th I x  y' I < 8, it wi) 1 follow
that Ix  y'l < 8' because 8 < 8'. Thus, it will be
possible to specialize Al to y'. Also, from '112 and
the fact that 8 < e'/c, it will be possible to
conclude that If(x)  fey') I s clx  y'l < c8 s
c(e'/c) = e', thus completing the proof.
116
APPENDIX B
Proof of Example 18. To show that f is continuous at
x, let e' > O. By the hypothesis, it is possible to
construct 0 < 8 < min{8', el/cl. Then, for yl with
Ix  yl I < 8, it follows that Ix  y'l < 8
1
and so,
by the hypo th e sis, I f ( x )  f ( Y I ) I sci x  y I I < c 8.
Furthermore, since 8 < el/c, one has If(x)  f(y') I <
c8s c(el/c) = e', and the proof is complete.//
As with any language,
speaking come only with
techniques are designed to get
right direction.
reading, writing, and
practice. The proof
you started in the
solution.s to exercises
Chapter 1
1.1. (a), (c), (e), and (f) are statements.
1.2. (a) Hypothesis: The right triangle XYZ with
sides of lengths x and y, and hypotenuse of
length z, has an area of z2/4.
Conclusion: The triangle XYZ is isosceles.
(b) Hypothesis: n is an even integer.
Conclusion: n
2
is an even integer.
(c) Hypothesis: a, b, c, d, e, and f are real
numbers that satisfy the property that
ad  be ; o.
Conclusion: The two linear equations ax + by
= e and ex + dy = f can be solved for x and
y.
(d) Hypothesis: n is a positive integer.
Conclusion: The sum of the first n integers
is n(n + 1)/2.
(e) Hypothesis: r is a real number and satisfies
r2 = 2.
Conclusion: r is an irrational number.
(f) Hypothesis: p and q are positive real
numbers with fPij ; (p + q)/2.
Conclusion: p ; q.
'g) Hypothesis: x is a real number.
Conclusion: The minimum value of x(x  1) is
at least 1/4.
1.2. If we want to prove that ftA => S" is true and we
know that S is false, then A should also be
false. The reason is that if.,. is false, then
it does not matter whether B is true or false,
and Table 1 ensures that ".,. => S" is true. On
the other hand, if A is true and S is false,
then "A => B" w o u 1 ~ be false.
117
118 SOLUTIONS
1.4. (a) True, because 'A is false.
1. 5.
(b) True, because A and B are both true.
(c) True, because B is true. The truth of ~
does not matter.
(d) True if x ~ 3 because then A is f.alse.
False if x • 3 because then A is true.
(a) (T = true, F = fal se)
A B C (B=>C) A=>(B=>C)

T T T T T
T T F F F
T F T T T
T F F T T
F T T T T
F T F F T
F F T T T
F F F T T

fb) (T
= true, F
=
fal se)
A B C (A=>B) (A=>B)=>C

T T T T T
T T F T F
T F T F T
T F F F T
F T T T T
F T F T F
F F T T T
F F F T F

Chapter 2
2.1. The forward process is a process that makes
specific use of the information contained in
statement 'A, the hypothesis. The backward
process is a process that tries to find a chain
SOLUTIONS
119
of statements leading
statement 8, the
Incorporated in the
abstraction process
generating this chain
time.
that directly to the fact
conclusion, is
backward process
that is directed
true.
is the
toward
at a of statements, one
Specifically, the backward process works by
beginning with the statement B that we are
trying to conclude is true. Then we use the
abstraction process to ask and answer the
abstraction questions, deriving a sequence of
new statements with the property that if the
sequence of new statements is true, then B is
true. The abstraction process will continue in
this manner until either we obtain the statement
A, or until we can no longer ask and/or answer
the abstraction question. There can be
difficulties in the backward process, such as
the possibility of more than one abstraction
question. If this occurs, then we must use the
information in statement A to help us choose the
appropriate one. ~ second difficulty is that
there may be more than one answer to the
abstraction question and some of these answers
may not lead to the completion of the proof.
The forward process begins with the
statement A, which is assumed to be true, and
derives from it a sequence of new statements
that are true as a result of statement A being
true. Fvery new statement der ived from A should
be directed toward linking up with the last
statement derived in the backward process. The
last statement of the backward process acts as
the guiding light in the forward process, just
as the last statement in the forward process
helps us to choose the right abstraction
question and the right abstraction answer.
2.2. (c) is incorrect because it uses the specific
notation given in the problem.
2.3. (a) is correct because it asks, abstractly, how
the statement E can be proven true. (b) and (c)
are not valid because they use the specific
120 SOLUTIONS
notation in the problem. Statement (d) is an
incorrect question for this problem.
2.4. (a) How can I show that two lines are parallel?
How can I show that two lines do not
intersect? How can I show that two lines
tangent to a circle are parallel? How can I
show that two tangent lines passing through
the endpoints of the diameter of a circle
are parallel?
(b) How can I show that a function is
continuous? How can I show that the sum of
two functions is continuous? How can J show
that the sum of two continuous functions is
continuous?
(c) How can I show that an integer is even? How
can I show that a number is an even integer?
How can I show that an integer is not odd?
How can I show that an integer squared is
even?
(d) 1I0w can
de9ree
How can
can I
share a
I show that the solution to a second
polynomial is a specific integer?
I solve a quadratic equation? How
show that two quadratic equations
common root?
2.5. (a) Show that their difference equals zero.
Show that one is s the other and vice
versa. Show that their ratio is one. Show
that they are both equal to a third number.
(b) Show that correspondin9 sideangIesides are
equal. Show that their corresponding
an9lesidean9Ies are equal. Show that
corresponding sidesidesides are equal.
Show that they are both congruent to a third
trian9le.
Cc) Show that they lie in the same plane and do
not intersect. Show that they are both
perpendicular to a third line. Show that
they have equal slopes. Show that their
corresponding equations are identical or
have no common solution. Show that they are
each parallel to a third line.
Cd) Show that it has three 90 degree angles.
Show that it is a square. Show that it is a
parallelogram with one right angle.
SOLUTIONS
121
2.6. (a) (1) How can I show that the solution to a
quadratic equation is positive?
(2) Show that the quadratic formula gives a
positive solution.
(3) Show that the solution b/a is positive.
(b) (1) How can I show that a triangle is
equilateral?
(2) Show that three sides have equal length
or, show that three angles are equal.
(3) Show that R'f = ~ = ~ , or show that
angle R = angle S = angle T.
2.7. (a) (x  2)(x  1) < o.
x(x  3) <  2.
x
2
+ 3x  2 > O.
(b) x/z = 1/Y2.
( c)
Angle X has 45 degrees.
cos (X) = 1/{2.
The circle has its center at (3,
The circle has a radius of 5.
The circle crosses the y axis at
(0, 2).
x
2
 6x
+ ~ + y2 _ 4y + A
= 25.
:» •
(0, 6) and
(d) OV = :ow = OW.
Angle U = angle W = angle V = 60 cegrees.
The triangle is isosceles.
The height of the triangle is Y3i2 times the
length of a side.
2.8. (d) is not val id because ·x " 5" is not stated
in the hypothesis, and so, "if x = 5, then it
will not be possible to divide by x  5.
:>.9. Ca) Outline of proof. An abstraction question
associated with the conclusion is "How can I
show that a real number (namely x) is O?" To
show that x = 0, it will be established that
both x :!Ii 0 and x ~ O. Working forward from
the hypothesis immediately establishes that
x ~ O. To see that x :!Ii 0, it will be shown
that x = y and y:!li C. Both of these
statements follow by working forward from
the hypothesis that x + y = 0 (so x = y)
122
SOLUTIONS
and y l!: 0 (so y s 0) • It remains only to
show that y = 0, which follows by working
forward from the fact that x = 0 and the
hypothesis that x + y = o. Specifically,
o = x + y = 0 + y = y.
(b) Proof. To see that both x = 0 and y = 0, it
will first be shown that x l!: 0 ,which is
given in the hypothesis) ana x s o. The
latter is accomplished by showing that
x = y anc that y s r. To see that x = y,
observe that the hypothesis states that
x + y = O. Similarly, y s r., because the
hypothesis states that y ~ r. Thus, x = o.
Finally, to see that y = 0, one can use
the fact that x = 0 and the hypothesis that
x + y = r to reach the desired conc1usion.11
2.10. (a) The number to the left of each 1 ine in the
diagram below indicates which rule is being
used.
J / ~
J ; S ~ s t ~
ssss sst stst
1//31 ~ 1\ 1\
ssssssss ts st sssst sstsst stststst
(b) The number to the left of each line in the
diagram below indicates which rule is being
used.
2.11.
SOLUTIONS
123
(c) A : s given
AI: ss rule 1
A 2: ssss rule 1
B 1: sssst rule 4
B tst rule 3
(d) A s given
AI: tst frOM part c
A,}: tsttst rul e 1
A:I·
tsst rule
'}
114: tssttsst rule 1
B1: tsssst rule 2
B : ttst rule 3
Outline of proof. In this problem one has:
A: The right triangle XYZ is isosceles.
B: The area of triangle XYZ is z2/ 4•
An abstraction question for E is
show that the area of a triangle
particuler value?" Cne answer is
formula for computing the area of
show that
El: z2/ 4 = xy/2
"How can I
is equal to a
to use the
a triang Ie to
Working from the hypothesis that
triangle XYZ is isosceles, one has:
A] : x = y
12: x  y 0
Since XYZ is a right triangle, from the
Pythagorean theorem, we obtain
z2 = x
2
+ y2
From the statement A2, if we square both sides
of the equation and perform some algebraic
manipulations, we obtain
A4: (x  y)2 = 0
A5: x
2
 2xy + y2 : 0
124 SOLUTIONS
A6: X
2
+ Y 2 = 2xy
Substituting A: into A6 yields
",7: z2 = 2xy
both sides by 4 finally yields the
desired result:
A8: z2 = xy/2
Proof. From the hypothesis, one knows that
x = y, or equivalently, that x  y = o.
Performing some algebraic manipulations, one
obtains x
2
+ y2 = 2xy. By the Pythagorean
theorem, z2 = x
2
+ y2 and on substituting z2
for x
2
+ y2, one obtains z2 = 2xy, or
equivalently, z2/
A
= y.y/2. From the formula
for the area of a right triangle, the area of
XYZ = xy/2. Hence z2/ 4 is the area of the
triangle.//
2.12. Outline of proof. abstraction question
associated with the conclusion is "How can I
show that a triangle is equilateral?" One
answer is to show that all three sides have
equal leng_th, specifically, = = IT''f. To
see that = one can work forward from the
hypothesis to establish that triangle RSU is
congruent to triangle SUT. Specifically, by
the hypothesis, 1m = IJ'f and, in addition,
angle RUS = angle SUT = 90 degrees. Of course
= so the sideangIeside theorem states
that the two triangles are congruent. Finally,
to see that work forward from the
to obtain the conclusion that
RS = 21m = 1m + O'T = rrf.
Proof. To see that triangle RST is
equilateral, it will actually be shown that
= To that end, the hypothesiS that
SU is a perpendicular bisector of RT ensures
(by the sideangIeside theorem) that triangle
RSU is congruent to triar:!9le fliT.
RS ... ST. To see that R'S = JTIf, by the
!:!'ypothesis,. one can easily conclude that
= 21m = in} + tJ'f' =
Chapter 3
3" 1. ( a) Ab s " en :
Abs. Ans:
Spec. Ans:
(b) ·Abs. Cn:
Abs. Ans:
Spec. Ans:
Ic) Abs. en:
Abs. Ans:
Spec. Ans:
(d) Abs. en:
.abs. Ans:
Spec • .ans:
(e) Abs. On:
Abs. Ans:
Spec. Ans:
SOLUTIONS
125
How can I show that an integer is
odd?
Show that the integer equals two
times some integer plus one.
Show that n
2
• 2k + 1.
How can I show that a real number
is a rational number?
Show that the real number is
equal to the ratio of two
integers, where the integer in
the denominator is not equal to
zero.
Show that sit • plq, where p and
q are integers and q ~ o.
How can I show that two pairs of
real numbers are equal?
Show that the first and second
elements of one pair of real
numbers are equal to the first
and second elements of the other
pair, respectively.
Show that Xj = x
2
and Y1 = Y2"
How can I Show tnat an integer is
pr ime?
Show that the
positive, greater
only be divided
one.
integer is
than I, and can
by itself and
Show that n > 1 and, if p is an
integer that divides n, then
p :: 1 or p = n.
How can I show that an integer
divides another integer?
Show that the second integer
equals the product of the first
integer with another integer, and
that the first integer is not
equal to zero.
Show that the following
expression holds true for some
integer
3
k: 3
(n  1) + n
3
+ (n + 1) • 9k.
128 SOLUTIONS
3.2. (A is the hypothesis and Al is the result of
working forward one step).
(a) A : n is an odd integer.
AI: n '" 2k + 1, where k is an integer.
(b) A : sand t are rational numbers with t ~ o.
AI: s '" p/q, where p and q are integers with
q ~ O. Also, t • alb, where a and bare
integers with b ~ o.
(c) A : Triangle RST is equilateral.
(d)
( e)
AI: 1m = S"f = ~ , and
A :
AI:
A :
AI:
angle(R) • angle(S) = angle(T).
,6.
sin(X) '" cos(X).
x/z = y/z (or, x = y).
a, b, c are integers for
divides band b divides c.
b = pa and c '" qb, where p and
integers and, a ~ 0, b ~ o.
which a
q are two
3.3. (T '" true, F = false)
(a) Truth table for converse of " ~ implies 8."
A B
T T
T F
F T
F F
"A Impl ies e"
T
F
T
T
"B Impl ies A"
T
T
F
T
(b) Truth table for inverse of "A implies B."
A B
T T
T F
F T
F F
NOT A
F
F
T
T
NOT B "NOT A Implies NOT 8"
F
T
F
T
T
T
F
T
Note: Converse and inverse of "A implies B" are
SOLUTIONS 127
equivalent. Both are false if and only if
A is false and B is true.
(e) Truth table for "A OR B.
"
A B "'It OR B"

T T T
T F T
F T T
F F F
'd) Truth table for "A AND B."
T
T
F
F
B
T
F
T
F
T
F
F
F
(e) Truth table for "A AND NOT E"
B
T T
T F
F T
F F
NOT B
F
T
F
T
"A AND NOT B"
F
T
F
F
(f) Truth table for "(NOT A) OR B."
NOT A B (NOT A) OF B "A Implies B"
T
T
F
F
F
F
T
T
T
F
T
F
T
F
T
T
T
F
T
T
Note:"" implies 8" and "NOT A OR B" are
equivalent. Both are false when 'It is true
and B is false.
128
? 4. (a)
( b)
(c)
Converse
Inverse
Contra
posi tive
Converse
Inverse
Contra
posi tive
Inverse
Contra
posi tive
(d) Conver se
Inverse
Contra
posi tive
SOLUTIONS
if n is an even integer, then
n
2
is even.
if n is an integer for which
n
2
is not even, then n is odd.
if n is an odd integer, then n
2
is odd.
if r is not rational, then r is
a real number such that r2 = 2.
if r is a real number such
that r2 2, then r is
rational.
if r is rational, then r is a
real number such that r2 f 2.
if the quadrilateral is a
rectangle, then the quadrilateral
ABCI:' is a parallelogram with
one right angle.
if the quadrilateral ABCI:' is
not a parallelogram with one
right angle, then the
quadrilateral ABCD is not a
rectangle.
if the quadrilateral is
not a rectangle, then ABCD is
not a Farallelogram with one
right angle.
if t = x/4, then t is an angle
for which sin(t) = cos(t) and
o < t < x.
if t is an angle for which
sin(t) f cos(t) and
o < t < x, then t f x/4.
if t f x/4, then t is an
angle for which 0 < t < x, and
sin ( t) f co s ( t) •
3.5. Outline of proof. The forwardbackward method
gives rise to the abstraction question "How can
I show an integer (namely n
2
) is odd?" The
definition for an odd integer is used to answer
the abstraction question which means that we
have to show that n
2
= 2k + 1 for some
k. We now turn toward the hypothesis and work
SOLUTIONS
128
forward to reach the desired conclusion.
Since n is an odd integer,
n = 2m + 1 for some integer m.
by definition,
'Iherefore,
n
2
= (2m + 1)2
= 4m
2
+ 4m + 1
= 2(2m
2
+ 2m) + 1
Thus the desired value for k is (2m
2
+ 2m), and
since m is an integer, (2m
2
+ 2m) is an integer.
Hence it has been shown that n
2
can be expressed
as 2 times some integer plus 1, which completes
the proof.
Proof. Since n is an odd integer, there is an
integer m for which n = 2m + 1, and therefore,
n
2
=
(2m +
1)2
=
4m
2
+ 4m + 1
=
2(2m
2
+ 2m) + 1,
and hence n
2
is an odd integer./ /
3.6. Outline of proof. Proceeding by the
forwardbackward method, one is led to the
abstraction question "Pow can I show an integer
(namely mn) is odd?" By using the definition of
an odd integer, this question can be answered by
showing that mn can be expressed as 2 times some
integer plus 1. We now turn to the forward
process to find out which integer.
Since m is an odd integer and n is an odd'
integer, one has m = 2b + 1 for some integer b,
and n = 2c + 1 for integer c. 'Iherefore,
mn = (2b + 1) (2c + 1)
= 4bc + 2b + 2c + 1
= 2(b + c + 2bc) + 1.
130 SOLUTIONS
Thus it has been shown that mn can be written as
2 times another integer plus 1, that other
integer being (b + c + 2bc), and this completes
the proof.
Proof. Since
m .. 2b + 1 and
integers. Thus
m and n are odd
n .. 2c + 1, where
mn .. (2b + 1) (2c + 1)
= 2(b + c + 2bc) + 1.
Hence mn is an odd integer.11
integers,
band care
3.7. Outline of proof. Using the forwardbackward
method, one is led to the abstraction question
"How can I show that one statement (namely A)
implies another statement (namely C)?" According
to Table I, the answer is to assume that the
statement to the left of the word "implies" is
true, and then reach the conclusion that the
statement to the right of the word "implies" is
true. In this specific case, we assume A to be
true and we try to prove that C is also true.
Working forward from the information given
in the hypothesis, we find that since " ~ implies
S" is true and A is true, B must be true. Since
S is true, and "B implies C" is true, C must
also be true. Hence the proof is complete.
Proof. To conclude that "A implies C" is true,
assume that A is true. Sy the hypothesis, "A
impl ies B" is true, so B must be true.
Finally, since "S implies C" is true, it must
be that C is true, thus completing the proof.11
3.8. Outline of proof. The forwardbackward method
gives rise to the abstraction question "How can
I show that two statements (namely A and B) are
equivalent?" Definition 8 indicates that one
must show that "A implies B" is true (which is
given in the hypothesis), and "B implies A" is
true. The hypothesis that "B implies C" and "C
implies A," together with Exercise 3.7, ensures
SOLUTIONS 131
that "B implies A" is true. The proof that A is
equivalent to C is similar.
Proof. In order to show that A is equivalent to
B, one need only show that "B implies A," since
the hypothesis states that "A implies B." The
fact that "B impl ies f." follows from the
hypothesis that "8 implies C" and "C implies A,"
together with Exercise 3.7. The proof that A is
equivalent to C is omitted.11
3.9. (a) We will require six proofs to show that A
is equivalent to each of the three
alternatives. The six proofs are: A => B,
P, => A, " => C, C => A, A => D, and D => A.
(b) We would require only four proofs, namely,
A => S, B => C, C => D, and D => A.
(c) If the four statements in part (b) above
are true, then you can show that A is
equivalent to any of the alternatives by
using Exercise 3.7. For instance, to show
that A is equivalent to D, one already
knows that "D impl ies A." Also, by
Exercise 3.7, since "A implies B," "B
implies C," and"C implies D," it follows
that "A implies D."
3.10. (a) Outline of proof. The forwardbackward
method gives rise to the abstraction
question "How can I show that a triangle is
isosceles?" Using the definition of an
isosceles triangle, we have to show that
two of its sides are equal, which in this
specific case means that we have to show
that u = v. Working forward fromtb..e
hypothesis, we know that sin(U) = Yu/2v.
By the definition of sine, sin(U) = ulw, so
ru.72V = ulw, and by applying algebraic
manipulations, one obtains that w
2
= 2uv.
Furthermore, by the Pythagorean theorem,
u
2
+ v
2
= w
2
• On substituting for w
2
, one
has u 2 + v
2
= 2uv, and consequen tl y,
u
2
 2uv + v
2
= o. On factoring, and then
taking the square root of both sides of the
equation, it follows that u  v = 0,
or, u = v, completing the proof.
132
SOLUTIONS
Proof. Since sin{U)" ~ and also
sin(U) • u/w, Yii12V .. u/w, or w
2
= 2uv.
Now, by Pythagorean theorem, w'2 = u
2
+ v
2
,
and on substituting 2uv for w
2
, and then
performing some algebraic manipulations,
one obtains u .. v.//
(b) Outline of proof. In order to verify the
hypothesis of Example 1 for the current
triangle UVW, it is necessary to match up
the notation. Specifically, x = u, y .. v,
and z .. w. Then it must be shown that
uv/2 = w
2
/4. Working forward from the
current hypothesis that sin(U) • yu/2v, and
since sin(U). u/w, one has ~ = u/w,
or, u/2v • u
2
/w2, or, w
2
• 2uv, and finally
that uv/2" w
2
/4. Observe also that
triangle UVW is a right triangle.
Proof. By the hypothesis, sin(U) = fU72V,
and from the definition of sine,
sin (U) .. u/w, thus YIi12V. u/w. By
applying algebraic manipulations, one
obtains uv/2 • w
2
/4. Hence, the hypothesis
of Example 1 holds for the current right
triangle UVW, and consequently the triangle
is isosceles.//
(c) Outline of proof. In order to verify the
hypothesis of Example 3 for the current
triangle UVW, it is necessary to match up
the notation. Specifically, r • u, s = v,
and t • w. Thus it must be shown that
w .. Y2uv but, as in the proof of part (b),
Yu/2v = u/J!..Lso u/2v = u
2
/w2, or, w
2
• 2uv,
or, w· r2uv. Observe also that triangle
UVW is a right triangle.
Proof. By the hypothesis, sin(U) = YU72V,
and from the definition of sine,
sin(U) • u/w, thus one has yu/2v = u/w. By
algebraic manipulations, one obtains
w = Y2uv. Hence the hypothesis of Example
3 holds for the current right triangle UVW,
and consequently the triangle must be
isosceles./ /
SOLUTIONS
133
Chapter 4
4.1.
(a)
( b)
Obj ect
mountain in
Himalyas
integer x
Ic) line l'
(d)
(e)
angle t
rational
numbers r,s
Certain
property
Something happens
over 20,000 taller than every
feet other mountain in
the world
none x
2
 5x/2 + 3/2
= 0
through P
not on 1
l' parallel to R.
r < t < X/2 sin(t) = cos(t)
between x Ir  sl < 0.(101
and y
4.2. (a) A triangle )(YZ is isosceles if ::I two sides !)
their lengths are equal.
4.3.
(b) Given an angle t, ::I an angle t' ~ tangent of
t' is greater than tangent of t.
(c) At a party of n people, ::I at least two
people ~ they have the same number of
fr iends.
(d) For a polynomial of degree n, say P(x), ::I
exactly n complex roots, r
1
' • , rn ~
p(r
1
) == ••• = p(r
n
) = O.
Outline of proof. Using the
method, we will find values of
x
2
5x/2 + 3/2 = O. From the
formula, the values of x are:
construction
x such that
quad ratic
(5/2 +"'25/4  12/2)/2 = (5/2 ~ 1/2)/2,
or, x equals I or 3/2. Now if we substitute
these values of x into x
2
 5x/2 + 3/2 = O,we
see that the quadratic equation is satisfied.
(a) Proof. Using the quadratic formula for
x
2
 5x/2 + 3/2 = 0, we find that x = 1 or
134
( b)
SOLUTIONS
x = 3/2. Thus, there exists an
namely x = 1, such that x
2
 5x/2 +
The integer is unique.11
integer,
3/2 = o.
Proo f • Us i ng the quad ra t ic
x
2
 5x/2 + 312 = 0, we find
x = 3/2. Thus there exists a
for x such that x
2
 Sx/2 +
real number x is not unique.11
formula on
that x = 1 or
real number
3/2 :: O. The
4.4. Outline of proof. The forwardbackward method
9ives rise to the abstraction question "How can
I show that an integer (namely a) divides
another integer (namely c)?" By Definition 1,
one answer is to show that there is an integer k
such that e = ak. The appearance of the
quantifier "there is" suggests turning to the
forward process to construct the ~ e s i r e d k.
From the hypothesis that alb and blc, and
by Definition 1, there are integers p and q such
that b = ap and c = bq. Therefore, it follows
that c = bq = Cap)q = a(pq), and the desired
integer k is k = pq.
Proof. Since alb and blc, by c'lefinition, there
are integers p and q for which b = ap and
c = bq. Put then c = bq = lap)q = a(pq). Thus,
alc.11
4.5. Outline of proof. The forwardbackward method
gives rise to the abstraction question "How can
r show that a real number (namely sit) is
rational?" By Definition 11, one answer is to
show that there are integers p and g with q ~ 0
such that sit = p/q. The appearance of the
quantifier "there are" suggests turning to the
forward process to construct the ~ e s i r e d p and
q.
From the hypothesis that s and t are
rational numbers, and by Definition 11, there
are integers a, b , c, and d with
b ~
0 and
d ~ 0 such that s = alb
and t c/d.
Furthermore, since t ~ 0, c ~ 0, and thus
be ~ o. Hence sit
(alb) I (c/d)
::
ad/bc. So
the ~ e s i r e d integers p and q are p = ad and
SOLUTIONS
135
g = bc. Observe that since b ~ 0 and c ~ 0,
g ~ 0, and furthermore, sIt ~ pIg.
Proof. Since sand t are rational, by
definition, there are integers a, b, c, and d
with b ~ 0 and d ~ 0 such that s = alb and
t = c/d. Furthermore, since t ~ 0, c ~ o.
Constructing p = ad and q = bc, and noting that
g ~ 0, one has sIt = ta/b)/(c/d) ad/bc = pIg,
and hence sIt is rational.11
Chapter 5
5.1. (a) Object: real number x
Certain property: none
Something happens: f(x) s f(x·)
(b) Object: element x
Certain property: x in S
Something happens: g(x) ! f(x)
(c) Object: element x
Certain property: x in S
Something happens: x s u
(d) Object: real number &
Certain property: & > 0
Something happens: there is x in S such that
x > u  &
te) Object: elements x and y, and real
Certain property: x and y in
o s t s 1
Something happens: tx +
(l

t) y
element of C
(f) Object: real numbers x, y, and t
Certain property: Cst s 1
Something happens:
numbers t
C, and
is an
f(tx + n  t)y) s tf(x) + (1  t)f(y)
136
(g)
(h)
SOLUTIONS
Object: real number E
Certain property: E > 0
Something happens: there is a real
8 > 0 such that, for all real numbers
I x  yl < 8, I f(x)  fey) I < E
number
y with
Object: real number E
Certa in prorerty: E > 0
Something happens: there
such that for all
Ixkxl < E
is an integer k'
integers k > k',
5.2. (a) Let x' be a real number. It will be shown
that f(x') ~ f(x·).
(b) Let x' be an element in S. It will be shown
that g(x') i:!: f(x').
(c) Let x' be an element in S. It will be shown
that x' ~ u.
(d) Let E' be a real number such that E' > O.
It will be shown that there is an x in S
such that x > u  E'.
(e) Let x', y' be elements in C, and let t' be
a real number between 0 and 1. It will be
shown that text + ,I  t')y' is an element
of C.
(f) Let x', y', t' be real numbers such that
C ~ t' ~ 1. It will be shown that
f(t'x' + (lt')y') ~ t'f(x') + (lt')f(y').
(g) Let E' be a real number such that E' > O.
It will be shown that there is a real number
8 > 0 such that, for all real numbers y with
Ix  yl < 8, If(x)  f(y)1 < E'.
(h) Let E' be a real number such that E' > O.
It will be shown that there is an integer k'
such that, for all integers k > k',
I xk  x I < E'.
5 . ~ . When using the choose method to show that "for
all objects with a certain property, something
happens," you would choose one particular object
that does have the certain prorerty. You would
then work forward from the certain property to
reach the conclusion that the something happens.
This is precisely the same as using the
forwardbackward method to show that "if x is an
object with certain property, then the something
SOLUTIONS
137
happens,w whereby you would work forward from
the fact that x is an object with the certain
property, and backward from the something that
happens. In other words, a statement containing
the quantifier wfor all
w
can be converted into
an equivalent statement having the form
"if • then •••• w
5.4. (a) 3 a mountain ~ Yother mountains, this one
is taller than the others.
(b) Y angle t, sin(2t) = 2sin(t)cos(t).
(c) Y two nonnegative real numbers p and
q, ffi s (p + q)/2.
(d) Y two real numbers x and y with x < y, 3 a
rational number r ~ x < r < y.
5.5. (a) First you would use the construction method
to construct H > 0, and then use the choose
method to choose x' in S for which it must
be shown that I x' I s fII.
(b) First you would use the choose method and
choose H' > 0, and then use the construction
method to construct an x in S for which
I x I > M'.
(c) First you would use the choose method to
choose t' > 0, then use the construction
method to construct 0 > 0, and then again
use the choose method to choose x' and y'
with lx'  y' I < 0, for which it must be
shown that If(x')  f(y')1 < t'.
5.6. Outline of proof. The forwardbackward method
gives rise to the abstraction question WHow can
I show that a set (namely T) is a subset of
another set (namely S)?W The definition provides
the answer that one must show that
81: for all t in T, t is in S.
The appearance of the quantifier wfor all
w
in
the backward process suggests using the choose
method to choose an element t' in T for which it
must be shown that
82: t' is in S.
138
SOLUTIONS
This, in turn, is done by showing that t'
satisfies the defining property of S, i.e., that
B ~ : (t,)2  3t' + 2 s O.
Working forward from the fact that t' is in
T, i.e., that it satisfies the defining property
of T, one knows that
AI: 1 s tIs 2.
Consequently, (t'  1) ~ 0 and (t'  2) s 0, so
(t,)2  3t' + 2 .. Ct'  1) (t
l
 2) s 0, thus
establishing B ~ and completing the proof.
Proof. To show that T ~ S, let t
l
be an element
of T. It will be shown that t' is in S. Since
tl is in T, 1 s t' s 2, so (t'  1) ?! 0 and
( t I  2) sO. Th us ,
(t,)2  3t' + 2 .. Ct'  1) (t'  2) s 0
and so t' is in S.II
5.7. Outline of proof. The key word "for every" in
the conclusion suggests uSing the choose method.
To that end, let Xl > 2 be a real number. To
construct the desired y < 0, we want
x' = 2yl (1 + y)
or
x'
+ x'y .. 2y
or
x'
"" 2y 
x'y = y(2
 Xl)
or
y = x'/(2  x').
Hence the desired y is x'/(2  x'), and it is
easily shown that, for this value of y,
x'  2y/(1 + y). However, it must also be shown
that y < 0, which it is, since x' > 2.
Proof. Let x' > 2, and
construct y = x'/(2  x').
and it is also easy
x' = 2y/(1 + y) .11
therefore we can
Since x' > 2, y < 0,
to verify that
SOLUTIONS
139
~ . P . Outline of proof. The forwardbackward method
gives rise to the abstraction question "How can
I show a function is convex?" Using the
definition in Exercise 5.1 (f), one must show
that
B1: For all real numbers x and y, and for all t
wi th 0 ~ t ~ 1,
f(tx + (1  t)y) ~ tf(x) + (1  t)f(y).
The appearance of the quantifier "for all,"
suggests using the choose method. Therefore, we
choose real numbers x' and y', and a real number
t' that satisfies 0 ~ t' ~ 1. We must show that
f(t'x' +
n
_ t')y')
~ t'f(x') +
n  t')f(y').
But by the hypothesis,
f(t'x' +
(1 _ t')y')
=
m(t'x' + ( 1
_ t')y')
+ b
=
mt'x' + my'  mt'y'+ b
=
mt'x' + bt' + my'  mt'y' + b  bt'
= t' (mx' + b) + n  t') ( my' + b)
= t'f(x') + (1  t')f(y').
Thus the desired inequality holds.
Proof. To show that f is convex, we must show
that for all real numbers x and y, and for all t
satisfying C ~ t ~ 1,
f(tx + (1  t)y) ~ tf(x) + (1  t)f(y).
Let x' and y' be real numbers, and let t'
satisfy C ~ t' ~ 1, then
f(t'x' + (1  t')y') = m(t'x' + (1  t')y') + b
= mt'x' + my'  mt'y'+ b
t'(mx' + b) + (1  t') (my' + b)
140
SOLUTIONS
= t'f(x') + (1  t')f(y').
Thus the inequality holds.11
Chapter 6
6.1. (a) Applicable.
(b) Not applicable because the statement
contains the quantifier "there is" instead
of "for all."
(c) Applicable.
(d) Applicable.
(e) Not applicable because in this statement, n
is a real number, and induction is
applicable only to integers.
6.2. (a) The choose method is used when the following
form appears in statement B: "for every
object with a certain property, something
happens." Induction is used whenever the
object is an integer and the certain
property is that of being greater than some
initial integer. Induction is used in such
cases because it is often easier to show
that the something happens for n + 1 given
that it happens for n, rather than to show
that it happens for n, given the certain
property, as would be done with the choose
method.
(b) It is not possible to use induction when the
object is not an integer because showing
that P(n) implies pen + 1) may "skip over"
many values of the object. } ~ a result, the
statement will not have been proved for such
values.
6.3. Proof. First we show that P Cn) is true for
n = 1. Replacing n by 1, we must show that
lfll) = (1 + 1)1  1. But this is clear since
1 11 I) = 1 = (1 + 1) 1  1.
6.4.
SOLUTIONS 141
Now we assume that P In) is true and use
that fact to show that Pin + 1) is true. So
PIn) IO!) + • + n(n!)
::
In + I)!  1
PIn + 1 ) In!) + • + (n + 1) (n + I)! =
(n+2)!1
Starting with the hft side of Pin + 1), and
then by using PIn), one has
In!) + • . . + n(n!) + (n + 1) (n + 1) !
= [lO!) + + n(n!)] + (n + l)(n + 1)1
= [In+ I)! 1) + In+ 1)(n+ I)!
= (n+ 1)1[1 + (n+ 1)]1
:: (n + I)! (n + 2)  1
(n + 2)!  1.
Thus PIn + 1) is true, compJ eting the proof.1 I
Proof. First we show that Pen) is true for
n = 5. But :2
5
= 32 ano
52
= 25, so 2
5
> 52.
Hence we see that it is true for n = 5.
Assuming that PIn) is true, we then have to
prove that Pen + 1) is true. So
pen) 2 n > n
2
PIn + 1)
:2
n
+
1
> In + 1)2
Starting with the left side of PIn + 1), and
using the fact that Pin) is true, one has:
2
n
+
1
= 212n) > 21n2)
To obtain PIn + 1), it must be shown that for
n > 5, :2n
2
> (n + 1)2 = n
2
+ 2n + 1, or, by
subtracting n from both sides, that n 2 > 2n + 1.
This last statement is true for n = 5 (but not
for n = 1 or 2), and it can be shown to be true
for n > 5 by induction.11
142 SOLUTIONS
6.5. Proof. The statement is true for a set
consisting of one element, say x, because its
subsets are {x} and ~ , i.e., there are 21 = 2
subsets. Assume that for a set with n elements,
the number of subsets is 2". We will show that
for a set with (n + 1) elements, the number of
subsets is 2"+1. For a set S with (n + 1)
elements we can construct the subsets by listing
all of the subsets that are possible using the
first n elements, and then, to each such subset,
we can add the last element of S. By the
induction hypothesis, there are 2" subsets using
n elements, and an additional 2" subsets are
created by a d ~ i n g the last element of S to each
of the subsets of n elements. Thus, the total
number of subsets of S is 2 "+ 2" = 2"+1, and
the statement has been established for
(n + 1).//
6.6. Proof. Let
S
=
1
+
2
+ + n
Then
S
=
n + (n  1 )
+ + 1
and on adding the two equa tions one obtains
2S = n(n + 1)
i • e • , S
=
n(n + 1)/2.//
6.7. Proof. Checking for n = 1, we find that
n
3
 n 1  1 = 0, and six divides 0 since
o =(6) ee). Assuming that the statement is true
for (n  1), we know that 6 divides
(n1)3  'n1),or, (n1)3 (n1) =6k
for some integer k. We have to show that
n
3
 n = om for some integer m. To relate pen)
to Pen + 1) we have
(n  1)3  (n  1) = n
3
 3n
2
+ 3n  1  n + 1
n 3 _ 3n 2 + 2n
= n 3  n  (3n
2
 3n)
SOLUTIONS
143
So n
3
 n = rCn  1)3  In  1)] + 3n
2
 3n. By
the induction hypothesis, [en  1)3  (n  I)J
can be divided by 6, so we have to show that
3n
2
 3n can also be divided by 6, or
equivalently, that n
2
 n can be divided by 2.
However, since n
2
 n = n(n  1) = the product
of two consecutive integers, either n or (n  1)
must be even, and so indeed, n(n  1) can be
divided by 2. Consequently, it follows that
?n
2
 3n • 6p for some integer p. Thus, the
desired value for the integer m is k + p, for
then
n
3
 n = rCn  1)3  (n  1)] + 3(n
2
 n)
= 6k + 6p
= 6Ck + p) .11
6.8. (a) Verify that the statement is true for the
initial value. Then, assuming that the
statement is true for n, prove that it is
true for n  1.
(b) Verify that the statement is true for some
integer. Assuming that the statement is
true for n, prove that it is true for n + 1,
and that it is also true for n  1.
(c) Verify that the statement is true for n = 1.
Assuming that the statement is true for
2n + 1, prove that it is also true for
2n + 3.
6.9. The mistake occurs in the very last sentence
where it states that "Then, since all of the
colored horses in this (second) group are brown,
the uncolored horse must also be brown." How do
you know that there is a colored horse in the
second group? In fact, when the original group
of (n + 1) horses consists of exactly 2 horses,
the second group of n horses will not contain a
colored horse! The entire difficulty is caused
by the fact that the statement should have been
verified for the initial integer n = 2, not
n = I! This, of course, you will be unable to
do.11
1« SOLUTIONS
Chapter 7
7.1. Use the choose method when the quantifier wfor
all
w
appears in the backward process. Use
specialization when the quantifier wfor all
appears in the forward process. In other words,
use the choose method when you want to show that
wfor all objects with a certain property,
something happens.
w
Use specialization when you
know that wfor all objects with a certain
property, something happens.
w
7.2. (a)
(b)
(c)
(d)
(e)
m must be an integer ~ 5, and if it is, then
2
m
> mi.
y must be an element of S with Iyl < 5, and
if it is, then y is an element of the set T.
e' must be > 0, and if it is, then 3 8 > 0
:l Y Y wi th I x  y I < 8, If (x)  f (y) I < &'.
The quadrilateral ORST must have its area
equal to the square of the length of a
diagonal, and if it does, then QRST is a
square.
Angle S of triangle RST must be strictly
between 0 and ~ / 4 , and if it is, then
cos(S) > sineS).
7.? Outline of proof. The forwardbackward method
gives rise to the abstraction question How can
I show that a set (namely R) is a subset of
another set (namely T)1
W
One answer is by the
definition, so we show that
~ l : for all r in R, r is in T
The appearance of the quantifier wfor all in
the backward process indicates that the choose
method should be us en • So choose an r' in R
and, using that information and the hypothesis,
it must be shown that r' is in T.
Turning to the forward process, the
hypothesis says that R is a subset of Sand S is
a subset of T which, by definition, means,
respectively, that
SCLUTIONS 145
Al: for all r in R, r is in S, and
A2: for all s in ~ , s is in T.
Specializing Al to r' 'which is in R), one has
that r' is in S. Specializing A'2 to r' 'which
is in S), one has that r' is in T, which is Bl,
thus completing the proof.
Proof. To show that R is a subset of T, we must
show that for all r in R, r is in T. Let r' be
in R. By hypothesis, R is a subset of S, so r'
is in S. Also, by hypothesis, S is a subset of
T, so r' is in T.II
7 . ~ . Outline of proof. The forwardbackward method
gives rise to the abstraction question wHow can
J show that a set (namely S intersect T) is
convex?W One answer is by the definition,
whereby it must be shown that
Bl: for all x and y in S intersect T, and for
all 0 ~ t ~ 1, tx + t1  t) y is in S
intersect T.
The appearance of the quantifier for all in
the backward process suggests using the choose
method to choose x' and y' in S intersect T, and
a t' with 0 ~ t' ~ 1, for which it must be shown
that
P'2: t'x' + t1  t')y' is in S intersect T.
By working forward from the hypothesis and the
above information, !2 will be established by
showing that t'x' + '1  t')y' is in both Sand
T. Specifically, from the hypothesis that S is
convex, and by the definition, it follows that
Al: for all x and y in S, and for all 0 s t ~ 1,
tx + '1  t)y is in S.
Specializing this statement to x', y', and t'
yields that t'x' + t1  t')y' is in S. A
similar argument shows that t'x' + fl  t')y' is
also in T, thus completing the proof.
146
SOLUTIONS
Proof. To see that S intersect T is convex, let
x' and y' be in S intersect T, and let
0:5 t' :5 1. It will be established that
t'x' + (1  t')y' is in S intersect T. From the
hypothesis that S is convex, it follows that
t'x' + (1  t')y' is in S. In a similar manner,
t'x' + (1  t')y' is in T. Thus, one has that
t'x' + f1  t')y' is in S intersect T, and so S
intersect T is convex.//
7.S. Outline of proof. The appearance of the
quantifier "for all" in the conclusion indicates
that the choose method should be used to choose
a real number s' ?!: 0 for which it must be shown
that the function s'f is convex. An associated
abstraction question is "How can I show that a
function (namely s' f) is convex?" Us ing the
definition in Exercise 5.1(f), one answer is to
show that
Bl: for all real numbers x and y, and for all
0:5t:51,
s'f(tx + (1  t)Y):5 ts'f(x) + tl  t)s'f(y).
The appearance of the quantifier "for all"
in the backward process suggests using the
choose method to choose real numbers Xl and yl,
and (l :5 t I :5 1 for which it must be shown that
82: s'f(t'x' + (1  t')y') :5
t's'f(x') + (1  t')s'f(y').
The desired resul t is obtained by working
forwarCl from the hypothesis that f is a convex
function. By the definition in Exercise 5.1 If),
one knows that
~ 1 : for all real numbers x and y, and for all
0:5t:51,
f(tx + 11  t)y) :5 tf(x) + 11  t)f(y).
Specializing this statement to x', y', and t'
(noting that C :5 t' :5 1) yielc"s
A'2: f(t'x' + 11  t')y') :5
t'f(x') + (1  t')f(y').
SOLUTIONS 147
The desired statement 82 can be obtained by
multiplying both sides of the inequality in A2
by the nonnegative number s', thus completing
the proof.
Proof. To show that s'f is convex, let x' and
y' be real numbers, and let 0 s t' s 1. It will
be shown that s'f(t'x' + (1  t')y') s
t's'f(x') + (1  t')s'f(y·).
Since f is a convex function by hypothesis,
it then follows from the definition that
f(t'x' + (l  t')y') s t'f(x') + (1  t')f(y').
The desired result is obtained by multiplying
both sides of this inequality by the nonnegative
number s'./ /
7.6. Outline of proof. The forwardbackward method
gives rise to the abstraction question How can
I show that a set (namely C) is convex? Using
the definition in Exercise 5.1fe), one answer is
to show th,at
81: for all x and z in the set C, and for all
o s t s 1, tx + (1  t)z is in C.
The appearance of the quantifier for all
in the backward process suggests using the
choose method to choose x' and z' in C, and a
o s t' s 1. Using this informat ion and the
hypothesis, it must be shown that
82: t'x' + n  t')z' is in C.
This, in turn, is done by showing that the point
t'x' + (1  t')z' satisfies the defining
property of C, i.e.,
B3: f(t'x' + II  t')z') s y
Turning to the forward process, by
hypothesis, f is a convex function. So by the
definition in Exercise 5.1(f), it must be that
AI: for all x and z, and for all O!5: t S 1,
f (tx + (1  t) z) s tf ( x) + (l  t) f ( z) •
148
SOLUTIONS
Recognizing the quantifier for all
forward process, specialization will
Specifically, Al can be specialized to
and t' (noting that () s t' :! 1), so
A2: f(t'x' + (1  t')z') s
t'f(x') + (1  t')f(z')
in the
be used.
x', z',
Finally, to obtain B2, we will use the fact that
x' and z' are in the set C, I.e., they satisfy
the defining property of C, thus, f(x') s y and
f ( z') s y. So, from A 2 ,
f(t'x' + n  t')z') s t'f(x') + (1  t')f(z')
s t'y + fl  t')y
= y,
and hence B2 is true, completing the proof.
Proof: To show that C is convex, let x' and z'
be elements of C, and let t' satisfy 0 S t' S 1.
Hence f(x') s y and f(z') s y. By the
hypothesis, f is a convex function and so
f(t'x'+ (1  t')z') s t'f(x') + (1  t')f(z')
S tty + (1  t')y
= y
and consequently, t'x' + (1  t')z' is in C.II
7.7. Outline of proof. The forwardbackward method
gives rise to the abstraction question How can
I show that a number (namely 1) is a least upper
bound for a set (namely S)? Using the
definition in Exercise 5.1 fe), one answer is to
show that
Bl: is an upper bound for
t > 0, there is an x
x > 1  t.
S, and, for
in S such
all
that
To show that the first part of Bl is true, one
SOLUTIONS
149
is led to the abstraction question How can I
show that a number (namely 1) is an upper bound
for a set (namely 8)? Again, the definition in
Exercise S.l(c) can be used to provide the
answer that one must show that
B2: for all x in 8, x s 1.
Recognizing the quantifier for all in the
backward process, the choose method is used to
choose an element x in 8 for which it must be
shown that
B ~ : x S 1.
To establish B3, we now make use of the fact
that x is in S. To do so, it is important to
observe that the set 8 can be written as
S • {real numbers x: there is an integer n ~ 2
with x • 1  lin}.
Since x is in S, there is an integer n ~ 2 for
which x • 1  lin, and since n::: 2,
x • 1  lin s l, thus B3 is true.
Returning to Bl, one must still show that
B4: for all e > 0, there is an x in S such that
x > 1  c.
Again, the choose method is used to select an
e > 0 for which it must be shown that
B5: there is an x in S such that x > 1  e.
Turning to the forward process, the desired x in
S will be constructed by finding an integer
n ::: 2 for which 1  lin > 1  e, for then one
can construct x = 1  lin. The desired n is any
integer > lie, for then 1  lin > 1  e, thus
completing the proof.
Proof. To see that lis an upper bound for S,
let x be in S. Hence, there is an integer n ::: 2
such that x = 1  lin. So x s 1, and therefore 1
150 SOLUTIONS
is an upper bound for S. To complete the proof,
let e > O. An element x in S will be produced
for which x > 1  e. Specifically, let n > lie.
Then x • 1  lin satisfies x > 1  e.11
Chapter 8
8.1. (a) Assume: t, m, and n are three consecutive
integers, and that 24 divides
R. 2 + m 2 + n
2
+ 1.
(b) Assume: Matrix ~ is not singular, and the
rows of. ~ are linearly dependent.
(c) Assume f and 9 are two functions such that
9 ! f, f is u n b o u n ~ e d above, and 9 is
bo und ed above.
8.2. (a) The number of primes is not finite.
(b) The set of real numbers is not bounded.
(c) The positive integer p cannot be divided by
any positive integer other than 1 and p.
(d) The lines i and it do not intersect.
(e) The real number x is not ! 5.
8.3. Outline of proof. To use the contradiction
method, we start with the assumptions
}\: n is an integer for which n
2
is even.
NOT B: n is not even, i.e., n is odd.
We work forward from these assumptions using the
definition of an odd integer to reach the
contradiction that n
2
is odd. Thus, there
exists an integer k such that n = 2k + 1. Then
n
2
• (2k + 1) 2
4k2 + 4k + 1
= 2(2k
2
+ 2k) + 1
SOLUTIONS 151
2
which 2is of the form n = 2p + 1, where
p • 2k + 2k. '!'hus n
2
is odd, and this
contradiction establishes the result.
Proof. Assume, to the contrary, that n is odd
and n
2
is even. Hence, there is an integer k
such that, n = 2k + 1. Consequently,
n
2
= f2k + l) 2
= ~ k 2 + 4k + 1
= 2 f2k2 + 2k) + l.
and hence n
2
is odd, contradicting the initial
assumpt ion./ /
8.4. Outline of Proof. We proceed by assuming that
there is a chord of a circle that is longer than
its diameter. Starting with this assumption,
and using the properties of a circle, we should
arrive at a contradiction in order to provide a
valid proof.
Let AC be the chord of the circle that is
longer than the diameter of the circle. Let AB
be a diameter of the circle. '!'his construction
is valid since, by definition, a diameter is a
line passing through the center terminating at
the perimeter of the circle.
Now it follows that angle ACB has 90 degrees
(since it is an angle in a semicircle). Hence
the triangle ABC is a right triangle in which AB
is the hypotenuse. '!'hen the desired
contradiction is that the hypotenuse of a right
triangle is shorter than one side of the
triangle, which is impossible.
152
SOLUTIONS
Proof. Assume that there does exist a chord,
say AC, of a circle that is longer than a
diameter. We construct a diameter that has one
of its ends coinciding with one end of the
chord. If we join the other ends, then we
obtain a right triangle in which the diameter AB
is the hypotenuse. Hence ~ > ~ , which
contradicts the initial assumption.//
8.5. Outline of proof. To use the contradiction
method, we assume that the two lines ~ and t
2
,
which are both perpendicular to a third line t
in the plane, are not parallel. Working
forward, we can conclude that they do intersect
at some point in the plane. Hence the three
lines t" t2 and t wi 11 form a tr tangle.
The sum of the three angles of a triangle is 180
degrees. Since two of the angles are 90
degrees, it must be that the third angle of the
triangle has zero degrees, an obvious
contrad ic tion.
Proof. Assume that t, and t2 are not parallel,
and that they are both perpendicular to the same
line t. Hence t, and ta. must intersect, and
form a triangle in which two of the angles are
90 degrees. Since the sum of the three angles
of a triangle is 180 degrees, we are forced to
the conclusion that the third angle has zero
degrees, a contradiction.//
8.6. OUtline of proof. To use contradiction, we
proceed by assuming that no two people have the
same number of friends, i.e., everybody has a
different number of friends. Since there are n
SOLUTIONS
153
people, each of whom has a different number of
friends, we can list the number of friends each
person has according to an increasing sequence.
In other words,
person number 1 has
person number 2 hes
person number 3 has
no friends
1 friend
2 fr iends
person number n has n  1 friends.
By doing so, we arrive at the contradiction that
the last person will be friends with all the
other (n  1) people, including the first one,
who has no friend!
Proof. Assume, to the contrary, that no two
people have the same number of friends. The
people at the party can be numbered In such a
way that
person 1 has 0
person 2 has 1
person n has n  1
friends
friend
fr iends.
It then follows that the person with (n  1)
friends is a friend of the person who has no
friends, a contradiction.//
8.7. Outline of proof. By the contradiction method,
it can be assumed that there do exist three
consecutive numbers (n  1), n, and (n + 1) such
that the cube of the largest is equal to the sum
of the cubes of the remaining two. Now, we can
arrive at a contradiction by showing that no
integer satisfies this equation. Specifically,
we have that
(n  1)3 + n
3
= tn + 1)3
or
3 2 3 3
3n
2
n 3n + 3n  1 + n
= n
+ + 3n + 1
or
n
3
_
6n
2
 2 •
0
or
2
n (n 
6) ..
2
154
SOLUTIONS
2
No w for n ~ 6, n (n  6) ~ C, wh 11 e fo r n > 6,
n
2
(n  6) > 2. Hence there is no integer n for
wh i c h n
2
(n  6) = 2.
Proof. Assume that (n  1), n, and (n + 1) are
three consecutive integers such that
(n  1)3 + n
3
= (n + 1) 3 •
'!ben,
n
3
 3n
2
+ 3n  1 + n
3
= n
3
+ 3n
2
+ 3n + 1
or
n
3
_ 6n
2
= 2
or
n
2
(n  6) = 2.
Fo r n ~ 6 , n
2
(n  6) ~ 0, wh 11 e for n > 6,
n
2
(n  6) > 2. Hence, there is no integer n for
which n
2
(n  6) = 2.//
B.P. Outline of proof. By the contradiction methoc,
it can be assumed that the number of primes is
finite. If so, t ~ e n there will be a prime
number that is larger than all the other prime
numbers. Let n be such a number. Consider the
number n! + 1, and let p be any prime number
that divides n! + 1. ~ contradiction will be
reached by showing that p ~ nand p > n.
Since n is the largest prime, p ~ n. We
use the fact that p divides n! + 1 to show that
p ~ 1, P ~ 2, ••• , p ~ n, and hence it must
be that p > n. When nl + 1 is divided by 2,
there is a remainder of 1 since,
nl + 1 = rn(n1) . . .
(2)(1)] + 1
Similarly, when n! + 1 is divided by r, where
1 < r ~ n, there is a remainder of 1. Indeed,
one has that
(n! + l)/r = n!/r + l/r
Hence n! + 1 has no prime factor between and
n. But by assumption, it has a pr ime factor.
SOLUTIONS
155
Hence the prime factor is 9reater than n, which
contradicts the assumption that n is the largest
prime number.
Proof. Assume, to the contrary, that there are
a finite number of primes. Let n be the largest
prime. Then n! + I is not a prime since,
n! + I > n. Therefore, n! + I has a prime
factor p less than or equal to n. But nl + 1
cannot be divided by any number between land n.
Hence p > n and we therefore arrive at a
contradiction.11
8.9. (a) Construction method is to be used.
Construct an element s in S and show that s
is in T.
(b) Choose and contradiction methods, in that
order.
Choose Sl in S and conclude that there is no
t in T such that s I > t. To reach the
conclusion, use the contradiction method to
assume that there is a t in T such that
Sl > t. Now reach a contradiction.
(c) Contradiction method is to be used.
Assume that there exists an M > 0 such that,
for all x in S, Ixl < M, and try to reach a
contrad iction.
Chapter 9
9.1. (a) Work fo rwa rd from: n is an odd integer.
Try to conclude: n
2
is an odd integer.
(b) Work forward from: S is a subset of T, and
T is bounded.
Try to conclude: Sis bounded.
(c) Work forward from: f (x) = f (y) •
Try to conclude: x = y.
(d) Work fo rwa rd from: the rows of matrix M are
linearly dependent.
Try to conclude: M is singular.
158
9.2.
SOLUTIONS
Statement (b) is a
because yo u can
number t between
sin(t) = r cos(t).
the answer in part
result of the forward process
assume that there is a real
o and "/4 such that
Upon squaring both sides,
(b) resul ts.
s. 'J. To use the contraposi tive method, we need to
work forward from NOT B and backward from NOT A.
(a) Incorrect, the abstraction process should
not be applied to NOT B. This abstraction
question also contains specific notation
from the problem.
(b) Incorrect, because the abstraction question
contains specific notation from the problem.
(c) Incorrect, because the abstraction process
should not be applied to NOT B.
(d) Correct.
9 . ~ . Outline of proof. With the contrapositive
method, one can assume that there is an odd
integer solution, say m, to the equation
n
2
+ n  c = o. It must be shown that c is not
odd, i.e., that c is even. But from the
equation, one can write that c = m + m
2
• To see
that c is even, observe that m is odd, and so
m
2
is also odd (see Exercise 3.5). Hence
c = m + m
2
= odd + odd = even, and the proof is
compl ete.
Proof. Assume that there is an odd integer
solution, say m, to the equation n
2
+ n  c = O.
It wi 11 be shown that c is not odd. But
c = m + m
2
, and since m is odd, m + m
2
is even,
and so c is even.11
9.5. Outline of proof. The appearance
quantifier "for all" in the conclusion
using the choose method, whereby one
real numbers x and y with x F y, for
must be shown that
el: f(x) F fey).
of the
suggests
chooses
which it
The word "not" in e 1 now suggests proceed ing
with the contradiction or contrapositive method.
Here, the contrapositive method will be used,
SOLUTIONS
157
and accordingly, it can be assumed that
f(x) • fey). Now it must be shown that x· y.
To reach the desired conclusion, work forward
from the fact that f(x) = f(y). Specifically,
mx + b • my + b, and subtracting the right side
from both sides of the equation, one obtains
m(x  y) = o. Using the hypothesis that m ~ 0,
it follows that x  y = 0, and consequently
x = y.
Proof. ~ s s u m e that x and yare real numbers for
which f(x) • f(y). It will be shown that x = y.
But since f(x)  f(y), it then follows that
mx + b = my + b, or, m(x  y) • O. By the
hypothesis that m ~ 0, one reaches the desired
conclusion that x • y.11
9.6. Outline of proof. By the contrapositive method,
one can assume that the quadrilateral RSTU is
not a rectangle. It must be shown that there is
an obtuse angle. The appearance of the
quantifier ·there is· suggests turning to the
forward process to produce the desired obtuse
angle.
Working forward, one can conclude that at
least one angle of the quadriatera1 is not 90
degrees, say it is angle R. If angle R has more
than 90 degrees, then it is the desired angle.
Otherwise, angle R has less than 90 degrees.
This, in turn, means that the remaining angles
of the quadrilateral must add up to more than
270 degrees (because the sum of all the angles
in RSTU is 360 degrees). Among these three
angles that add up to more than 270 degrees, one
of them must be greater than 90 degrees, and
that is the desired obtuse angle.
Proof. ~ s s u m e that the quadrilateral RSTU is
not a rectangle, and hence, one of its angles,
say R, is not 90 degrees. An obtuse angle will
be found. If angle R has more than 90 degrees,
then it is the desired angle. Otherwise, the
remaining three angles add up to more than 270
degrees. Therefore, one of the remaining three
angles is obtuse.11
158
SOLUTIONS
Chapter 10
10.1. (a)
(b)
(c)
(d)
(e)
( f)
(g)
(h)
The real number x· 1s not a maximum of the
function f if there is a real number. x such
that f(x) > f(x·).
Suppose that f and 9 are functions of one
variable. ~ e n g is not ~ f on the set S
of real numbers if there exists an element
x in S such that g (x) < f(x).
The real number u is not an upper bound for
a set S of real numbers if there exists an
x in S such that x > u.
The real number u is not a least upper
bound for a set S of real numbers if u is
not an upper bound for S, or if there
exists a real number e > 0 such that for
all x in S, x s u  e.
The set C of real numbers is not convex if
there exist elements x and y in C and there
exists a real number t between zero and one
such that tx + (1  t)y is not an element
of C.
The function f of one variable is not
convex if there exist real numbers x and y
and t, where 0 s t s 1, such that
f(tx + n  t)y) > tf(x) + n  t)f(y).
The function f of one variable is not
continuous at the point x if there exists
real number £ > 0 such that for all real
numbers 8 > 0, there is a y with
Ix  yl < 8 such that If(x)  fey) I ~ e.
Suppose that x, x
1
, x
2
, ••• are real
numbers. The sequence x 1, x2, ••• does
not converge to x if 3 a real number e > 0
3 Y integers k', 3 an integer k > k' such
that I xk  x I ~ L
10.2. (a) There does not exist an element x in the
set S such that x is not in T.
(b) It is not true that for every angle t
between 0 and x/2, sin(t) ~ cos(t) •
(c) There does not exist an ·object· with the
·certain property· such that the
·something· does not happen.
10.3.
1 0.4.
SOLUTIONS
159
(d) It is not true that for every ·object· with
the ·certain property,· the ·somethin9·
does not happen.
(a)
(b)
( c)
(d)
(a)
( b)
(c)
Assume that there is an integer n ~ 4 such
that nJ :s n
2
•
Assume A, NOT B AND NOT C.
Assume A, NOT B OR NOT C.
Assume that f is a convex function of one
variable, x· is a real number, and there is
a real number 8 > 0 such that for all real
numbers x with the property that
Ix  x·1 < 8, f(x) ~ f(x·). Finally,
assume that there is a real number y such
that fey) < f(x·).
Forward from: (NOT B) AND (NOT C)
Backward from: NOT A
Forward from: (NOT B) OR (NOT C)
Backward from: NOT A
Forward from: mn is not divisible by 4
and n is divisible by 4.
Backward from: n is an odd integer or m
is an even integer.
10.5. Outline of proof. When usin9 the contradiction
method, it can be assumed that x ~ 0, y ~ 0,
x + y = 0, and that either x # 0 or y # o. So
suppose first that x # 0 and hence, since
x ~ 0, it must be that x > O. A contradiction
to the fact that y ~ 0 will be reached by
showin9 that y < o. Specifically, since
x + y = 0, it follows that y • x but x < 0
and so the contradiction has been reached. A
similar ar9ument can be used for the case where
y # o.
Proof. Assume that x ~ 0, y ~ 0, x + y = 0,
and that either x # 0 or y # O. If x # 0, then
x > 0 and y = x < 0, but this contradicts the
fact that y ~ o. Similarly, if y # 0, then
y > 0, and x • y < 0, but this contradicts the
fact that x ~ 0.//
180
SOLUTIONS
Chapter 11
11.1. Outline of proof. The first uniqueness method
will be used, wherein one must first construct
the real number y. This, however, was already
done in Exercise 5.7. Thus, it remains only to
show the uniqueness by assuming that y and z
are both real numbers with y < 0, z < 0,
x • 2y/(1 + y), and x = 2z/(1 + z). By working
forward via algebraic manipulations, it will be
shown that y • z. Specifically,
x • 2y/(1 + y) • 2z/(1 + z)
or equivalently,
y + yz = z + yz
and, on subtracting yz from both sides, one
obtains the desired conclusion that y = z.//
Proof. The existence of the
established in Exercise 5.7.
unique, suppose that y and
z < 0, x = 2y/(1 + y), and x
then it follows that 2y/(1
and so y + yz = z + yz,
desired.//
real number y was
To show that y is
z satisfy y < 0,
= 2z/(1 + z). But
+ y) = 2z/(1 + z),
or, y = z, as
11.2. Outline of proof. According to the second
uniqueness method, we must first construct a
real number x for which mx + b = O. But since
the hypothesis states the m ; 0, the desired x
is b/m, because then it follows that
mx + b • m(b/m) + b = b + b = o.
To establish the uniqueness, suppose that
x and y both satisfy mx + b • 0 and my + b • 0,
and also that x; y. A contradiction is
reached by showing that m = o. Specifically,
since mx + b = 0 = my + b, it follows that
m(x  y) = o. On dividing by the nonzero
number x  y, it follows that m • o.
SOLUTIONS 181
Proof. To construct the number x for which
mx + b = 0, let x • b/m (noting that m ~ 0).
Then mx + b .. m(b/m) + b = O.
Now suppose that y ~ x and also satisfies
my + b = o. Then mx + b .. my + b, and so
m(x  y) = O. But since x  y ~ 0, it must be
that m .. 0, and this contradicts the hypothesis
that m f" 0.11
11.2. Outline of proof. The issue of existence must
be addressed first. To construct the desired
complex number (c + di) that satisfies
(a + bi) (c + di) = 1, let c = a/(a
2
+ b
2
) and
d .. b/(a
2
+ b
2
) (noting that the denominator
is not 0 since, by the hypothesis, at least one
of a or b is not 0). But then,
(a + bi) (c + d i) = ac  bd + tbc + acl) i
= (a 2 + b 2) I (a 2 + b 2) + 0 i
1.
To see the uniqueness, suppose that e + fi
is another complex number that also satisfies
(a + bi) (e + fi) .. 1. It will be shown that
c + di equals e + fi. Working forward and
multiplying both sides of the equation above by
e + fi, one obtains:
or,
(e + fi)[ Ca + bi)(c + di)] = fe + fi)(l),
[Ce + fi)(a + bi)](c + di) .. e + fi,
and since (e + fi) (a + bi) .. 1, it follows that
c + di = e + fi.
Proof. Since either a f" 0 or b f" 0, it follows
that a 2 + b
2
f" 0, and so it is possible to
construct the complex number 2c + di in which
c = a/(a
2
+ b
2
) and d = b/(a + b
2
), for then
(a + bi) (c + d i) = ac  be + tbc + ad) i
= 1.
182 SOLUTIONS
To see the uniqueness, suppose that e + fi
also satisfies (a + bi) (e + fi) '" 1. By the
rules of complex mul tiplication, it follows
that
(e + fi)[(a + bi)(c + di)] '" fe + f1)(l)
so c + di = e + fi and the
established.//
un iqueness is
11.4. Advantages: You have three statements to work
forward from, namely, NOT B, and NOT C.
With the either/or method, you would only have
two statements to work forward from.
Disadvantages: You cannot work backward because
you do not know what the contradiction will be.
In the either/or method you can work backward
from the statement (B OR C) •
11.5. Outline of proof. By the either/or method, we
can assume that n is an even integer and m is
and odd integer, and also that 4 divides n. We
must conclude that 4 divides mn.
Working backward, one is led to the
abstract,ion question "How can I show one
integer (namely 4) another (namely
mn)?" Using the definition leads to the, answer
that one must show that there is an integer k
such that mn '" 4k.
Turning to the forward process, we must
construct the value for k. Since m is odd,
there is an integer j such that m = 2j + 1.
Since 4 divides n, there is and integer p such
that n = 4p. Therefore
mn = f2j + 1)(4p) = 4'2jp + p)
and the desired k is k 2jp + p.
Proof. that n is even, m is odd, and
that 4 divides n. It will be shown that 4
divides mn, or equivalently, that there is an
integer k such that mn = 4k. But since m is
SOLUTIONS
163
odd, there is an integer j such that
m = 2j + 1. Since 4 divides n, there is and
integer p such that n == 4p. Thus,
mn = (2j + 1) (4p) ::: 4(2jp + p).
Hence, the desired integer is k = 2jp + p, thus
completing the proof.//
Note: This proof could also be done by assuming
n is an even integer and m is an odd
integer, and that 4 does not divide mn,
and then concluding that does not
divide n.
11.6. (a) If x is a real number that
x
3
+ 3x
2
 9x  27 0, then
satisfies
x :s 3 or
x ?: 3.
(b) Outline of proof.
either/or method,
x
3
+ 3x
2
 9x  27
must show that x ?: 3.
to the
we can that
o and that x > 3. We
Now one has
X 3 + 3 x
2
 9 x  27 = (x  3) (x + 3) 2 ?: O.
Since x > 3, (x + 3)2 is positive, and so
we must have x  3 ?: 0, or, x ?: 3.
Proof. Assume
and that x > 3.
that x 3 + 3x 2  9x 27 0
Then it follows that
(x  3) (x + 3) 2 ?: O.
Since x > 3, (x + 3)2 is positive, so
x  3 ?: 0, or equivalently, x ?: 3.//
(c) The proof is similar to part (b) exceFt you
assume that x < 3.
11. 7. ( a ) Fo r a 11 sin S , s :s x.
(b) There is an s in S such that s ?: x.
(c) There is an x with ax :s b and x?: 0 such
that cx :s u.
184 SOLUTIONS
(d) There is an x with ax b and x 0 such
that cx u.
(e) For all x with b x c, ax :!: u.
(f) Far all x with a x bx u.
11.8. of proof. The max/min method can be
used to convert the conclusion into the
equivalent quantifier statement:
B: for all s in S, s :!: t*.
The appearance of the quantifier "for all" in
the backward process suggests using the choose
method to choose an Sl in S, for which it must
be shown that
Ell: Sl :!: t*.
The desired conclusion
working forward and
Specifically, since S is
follows that
can be obtained by
using specialization.
a sub se t 0 f T, it
AI: for all s in S, s is in T.
Specializing Al to s = s' (which is in S), it
follows that Sl is in T. Also, the hypothesis
states that
A2: for all t in T, t :!: t*.
Again, applying specialization to 1'.2 with
t = Sl (which is in T), it follows that
Sl:!: t*, and thus the proof is complete.
Proof. To reach the conclusion, let Sl be in
S. It will be shown that Sl :!: t*. Ey the
hypothesis that S is a subset of T, it follow
that Sl is in T. But then, the hypothesis
ensures that Sl :!: t*.//
11.9. Outline of proof. The max/min method can be
used to convert the conclusion into the
equivalent quantifier statement:
SOLUTIONS 185
Bl: for all x O!: 0 with ax ~ b, and for all
u .a: 0 wi th ua:S c, cx ~ ub.
The appearance of the quantifier Rfor all
R
in
the backward process suggests choosing an
x' ~ 0 with ax' ~ b, and also choosing u' ~ 0
with u'a:os: c, for which it must be shown that
cx' ~ u' b. To accomplish the goal, it wi 11 be
shown that CX'I:!: u'ax' and that u'ax' l!: u'b.
Specifically, upon multiplying both sides of
the inequality uta ~ c by the nonnegative
number x', it follows that c x ' ~ u'ax'.
Similarly, upon mul tipl ying the inequality
aX'1:!: b on both sides by the nonnegative number
u', it follows that u'ax' ~ u'b, and the prQof
is complete.
Proof. To reach the desired conclusion,
X'I:!: 0 with ax' ~ b, and let u ' ~ 0
uta s c. It then follows that u'ax' l!: u'b
cx' ~ u'ax', and therefore one has
cx' ~ u'ax' l!: u'b, and consequently the
is complete.11
let
with
and
that
proof
Chapter 12
12.1.
(a) Contraposi tive or contrad iction method,
since the word
Rno
R
appears in the
conclusion.
(b) Induction method, since statement B is true
for every integer n l!: 4.
( c) ForwardBackward method, since there is no
apparent form to B.
(d) Maximin method, since B has the word
maximum
R
in it.
(e) Uniqueness method, since there is supposed
to be one and only one line.
( f) Contrad iction or contraposi tive method,
since the word
Rno
R
appears in the
conclusion.
188 SOLUTIONS
(g) ForwardBackward method, since there is no
apparent form to B.
(h) Choose method, since the first quantifier
in B is "for every."
(i) Construction method, since the first
quantifier in B is "there is."
12.2. (a) Using the induction method, we would assume
nJ > n 2 and n 2: 4, and try to show that
(n + 1)1> (n + 1)2. We would, of course,
also have to show that 4! > 4
2
•
(b) Using the choose method, we would choose an
integer n I for wh ich n' ?: 4. We would try
to show that (n')1 > In,)2.
(c) Converting the statement to the form "if
••• then " one obtains "if n is an
integer 2: 4 then nf > n
2
". We would
therefore assume that n is an integer 2: 4
and try to show that nl > n
2
•
(d) Using the contradiction method, we would
assume that there is an integer n ~ 4 such
that nl ~ n
2
, and try to reach a
contrad iction.
Glossary of Mathematical Symbols
Symbol Meaning page
>
implies 3
<=> (iff) if and only if ,6
E is an element of 40
subset 42
S
empty set 42
not 30
V for all (for each, etc.) 43
3
there is (there are, etc.) 3fi
such that 36
A
and 24
V
or 24
Q.E.D.
quod
(which was to be demonstrated) 14
167
index
·A implies B," 3, 30
abstraction process, 10
abstraction question, 9
alternative definition, 25
AND, 24, 81
assumption,S, 51, 53, ~ ~ , 72
axiom, 28
backward process, 9
certain property, 35, 43
choose method, 40
conclusion, 3
condensed proof, 15, 109
construction method, 34
contradiction method, ~ 4
contrapositive method, 72
contrapositive statement, 30
converse statement, 30
corollary, 27
definition, 23
defining property (of a set), 41
doing proofs, 3
either/or method, ~ ~
empty set 42
equivalence, 2r;
equivalent definitions, 25
equivalent statements, 2 ~
existence, 35
existential quantifier, 3d
169
170
INDEX
forwardbackward method, 8, 16
forward process, 9, 12
"for all," for any," "for each, "for every, 34, 40
for which, 36
generalized induction method, ~ 6
hypothesis, ?
if and only if, 24
implication, ?
implies, ?
induction method, 50
infinite list, 40
infinite set, 40
inverse statement, ~ o
lemma, 27
matching up of notation, 29
mathematical induction, 50
mathematical terminology, 23
mathematics as a language, 1
max/min method, 'e7
membership in a set, ~ o
negation, 78
negation in quantifiers, 7e
"no, not in conclusion, 67
NOT A, :0, 7f'
NOT B, 7P
NOT B implies NOT A," 30
"NOT A implies NOT B," 30
NOTS of quantifiers, 7P
only if, ?O
OR, el
proof, 1
proof by contradiction, 64
proof machine, 44, 52
proof techniques
INDEX
choose, 40
construction, :4
contradiction, 64
contrapositive, 72
either/or, e6
forwardbackward, e
induction, 50
max/min, f!7
specialization, S9
uniqueness, P3
proposi tion, 27
Q.E.D., 1.IJ
quantifiers
choose method (for all), ~ o
construction method (there exists), :4
existential, :4
"hidden," 37, i13
induction method, 50
specialization method, ~ 9
universal, ?i1
reading proofs, 16, 101, 109
set, .lJO
empty, ~ 2
set builder notation, ~ l
setdefining property, ~ l
set equality, 42, 45
something happens, :5, 43
specialization method, 59
special proof techniques, 83
statement, 2, f!
changing form of, IP7
subset, i12
symbolic notation, :
theorem, n
"there are," "there exists," "there is," 34
truth table, 5
"A implies F," 5
"NOT B implies NOT A," 31
171
172
INDEX
uniqueness, 83
uniqueness method, 83
universal quantifier, ~ 4
working backward, g
working forward, 9, 12
how to read and do proofs
an introduction to mathematical thought process
daniel solow
east' westf'rn rf'St'nf' university
john wUey
&:
sons
new york chichester brisbane toronto singapore
it is hoped that we,., of this book will kindly remove (lnd return the questionnaire ill('lude(1 as its last page. to tile address that ('all bf> left visible after folding and stapling
('opyri(da' © 19B2 by johll IfJUt>y & .0.... , illC'.
rf>protlu("ioll or '", ....Indoll 01 tilly pan 01 'his lfJork bt'yolltl pf>rmit't'tl by .t'('rio.... 107 or lOB 01 1M 1976 Ullitf>tl .'a't'. ('opyri(da' acl lfJithoUl ,he permit.oll 01 'M ('opyrilJla' olfJlU'r is un/alfJjiJ  rt'qut'.,. lor pt'rmit.ioll or fur,hf>r illlo""",ioll .hould bt' atltirt'''t'tl 10 ,/w pt'rmit.iOll. tlt'pantlle .." johll UJUt>y & .0'1$, illC'.
,ht"
isb.. 0471 S6645 S
10 9
SOWW and to my mother • • • RUTH SOLOW .to my late father • • • ANATOLE A.
.
ex~erimentation. above all. v . ~any stunents attempt to circumvent the obstacle by avoiding ittrusting to the indulgence of the examiner not to inclune any proofs on the test." All of those who have had the exrerience of teaching mathematics and most of those who ~ave had the eXferience of trying to learn it must surely agree that acquiring an understanding of what constitutes a sound mathematical ~roof is a major stumbling block for the stunent. discussion. arguably its most characteristic feature." the author has \vritten. and. "Teaching Mathematics with Proof Techniques. to be sure. rr. 'This collusion between stunent ano teacher mey avoid some of the unpleasant consequencesfor both student and teacherof the student's lack of mastery. There is a valid analogy here with the role of the traditional algorithms in elementary arithmetic. These proofs are. much attention is paidparticularly in the two append icesto showing the reader how to recogni ze the standard ingredients of a mathematical argument in an informal rresentation of a froof. and I do not doubt that his ideas deserve attention. Solow believes that it is possible to teach the student to understann the nature of a proof by systematizing it. not fresented in a systematic form. but it does not aJter the fact that a key element in methematics. "'The inability to communicate proofs in an understandable manner has plagued students and teachers in all branches of mathem8tics. He argues his case cogently with a wealth of netail and ex~mfle in this pamphlet. has not entered the stu~ent's rerertoire. Gne of his principal aims is to teach the student how to rea~ the proofs offered in textbooks. Thus.foreword In a related article.
Eut once all of this has been learned. Dr. Today. and Cr. However.vi FCREWORD It is important to acquire familiarity with them and to understand why they work and to what problems they could. Solow'S rl~n to remeoy this very unsatisfactory situation deserves a fair trial. procedure baseo on the hOFe that the stucents can learn this difficult art by osmosis. Solow has given much thought to the question of how an understanding of a mathematical proof can be acquired. students begin to grapple with the ideas of mathematical proof far too late in their student careersthe appropriate stage to be initi~ted into these ideas is. To be userl effectively. rather haphazard. Solow is not claiming that mathematicians creAte their proofs by consciously and deliberately applying the ·forwardbackward methoo·. in principle. he is suggesting that we have a far better chance of teaching an appreciation of proofs by systematizing them than by our present. Solow that. So. One must agree with Dr. one would not slavishly execute the algorithms in reallife situations (even in the absence of a calculator!). Louis D. be applied. it would be wrong for university and college teachers merely to excuse their own failures by a comforting reference to defects in the student's precollege education. in the judgement of many. no later than eighth grade. mathematics is genera] ly recognized as a subject of fundamental importance because of its ubiquitous role in contemporary life. Understand and analyze their structureand then you will be able to rean and understand the more informal versions you will find in textbooksand finally you will be able to create your own proofs. ~ost students tocay do not acquire that understanding. it is with proofs. Dr. Ohio rETFR flILTON . in this country. the author contends. and we place undue strain on our naturally imferfect memor ies. Eeaumont University Professor Case Western Reserve l~iversity Clevelanc'l. its methods must be understood properlyotherwise we cast ourselves in the roles of (inefficient) robots when we try to use mathematics.
In fact. tactics. I hope this pamphlet serves that purrose for you. Equally important is the fact that these tools will enable you to understand and to appreciate the creativity of others. Imagine trying to play chess before you know how all of the pieces move! It is no ~onder that so many students have had trouble with abstract mathematics. As I progressed through my graduate work. I do believe that it can rrovide you with the tools needed to express your creativity. These rules are no substitute for creativity. and the like. To play chess. Then you will fine that your mind can focus on the creative aspects of mathematics. It has been my experience that virtually anyone who is motivated and who has a knowle~ge of high school mathem~tics can learn these rules. you must first learn how the individual rieces move. I began to wonder why learning theoretical mathematics had been so difficult. Only after these rules have entered your subconscious can your mind turn its full attention to the more creative issues of strategy. and this pamphlet is not meant to teach creativity. your goal should be to absorb this material so that it becomes second nature to you. roing so will greatly re~uce the time (and frustration) involveC' in learning abstract mathem~tics. vii . This pamphlet describes some of the rules by which the game of theoretical mathematics is played. Hard work is required in the beginning to learn the fundamental rules presented in this pamphlet.to the student ~fter finishing my undergraduate degree. So it appears to be with mathematics. ~owever. I realized that mathematics possessed many of the aspects of a game: a game in which the rules had been partially concealed.
Remember. Cleveland. Ask questions and seek answers.vIII '10 THE STUt'ENT You are about to learn a key part of the mathematical thought process. the only unintelligent question is the one that goes unasked. be conscious of your o~n thought process. Ohio June 1981 r::'A~IEL SOLOW . 'As you study the material and solve problems.
and. Teachin9 proofs in this manner requires nothing more than Freceding each step of the proof ix . In essence. icentifies. as is illustrated in the examples of this pamrhlet.to the instructor The inability to communicGlte proofs in an understandable manner has plagued students and teachers in all branches of mathematics. it is then possible to explain any proof as a sequence of applications of these techniques. Cnce the students understand the techniques. ~~at seems to have been lacking is a prorer method for exrlaining theoretical mathematics. and techniques that are being used. this pamphlet categorizes. One might conjecture that most students simply cannot understand abstract mathematics. In this pamphlet I have developed a method for communicating proofs: a common language t~at can be taught by professors and understood by students. frustrated teachers. Pefore each "condensed" proof is an outline of the proof explaining the methodology. but my experience indicates otherwise. oftentimes. In fact. ~he result has been frustrated stu~ents. and explains (Glt the student's level) the various techniques that are used repeatedly in virtually all rroofs. it is advisable to do so because the process reinforces what the students have learne~ in the pamphlet. thought process. a "watered down" course to enable the students to follow at least some of the material. Explaining a proof in terms of its component techniques is not difficult. or a test that protects students [rom the consequences of this deficiency in their mathematical understand ing.
such as why a Froof is done in a particular way and why the riece of mathematics is important in the first place. but r do believe that it does describe many of the necessary underl ying ski lIs whose aCQuisition will free the stucent's mind to focus on the creative aspects. Cleveland. I have also found thtlt.lIt!IEL SOLOW ~'une . it is possible to teach subsequent matheMatical m~terial at a much more sorhisticaten level without lOSing the students. When discussing a Froof in class. This pamrhJet is not meant to teach creativity. the message is clear. It has been my exrerience that once the students become comfortable with the proof techniques. and why. I actively involve the students by soliciting their heIr in choosing the techniques and in <'lesigning the proof. This pamFhlet is designed to be a major step in the right direction by making abstract mathematics un~erstancable and enjoyable to the students and by providing you with a Method for communicating with them. In any event. by using this approach.x TO THr I~STRUCTCR with an indication of which technique is about to be used. Chio ISel C'. their minds tend to address the more imrortant issues of m~them~tics. I am suggesting that there ere many benefits to be gained by teaching mathematical thought process in addition to mathematical material. I have been pleasantly surprised by the quality of their comments and ~uestions.
1hanks especially to John Cemocko [or acting in the capacity of senior editor while concurrently trying to complete his Ph. Of all the people involved in this project.lan SChoenfelr.!:'. none deserve more credit than my students. r also thank Paul . I suppose that I should thank his mother for being an Fnglish teacher. a1mos. ~any other people ma~e substantive suggestions.. whose timely recognition and support greatly facilitated the dissemination of the knowledge of the existence of this pamphlet and teaching method. I also acknowle~ge Charles Wells for reading an~ commenting on the first handwritten draft and for encouraging me to pursue the project further. "spec ial word of thanks CJoes to Greg ~adey for coordinating the secone rewriting of xi . Samuel Goldberg.acknowledgments For helping to get this work known in the mathematics community. Michael has worked on this project almost as long as r have. It is because of their voluntary efforts that this parnrhlet has been prepared in such a short time. no single person had more constructive comments than 10m Butts. I am also grateful for discussions with Gail Young and Ceorge Fol ya. I appreciate the help that r received from ~ichael Creiling and Robert Wenig in data basing the text ann preparing the exercises. inclue ing Jl. Regarding the preparation of the manuscript. and Fllpn Stenson. He not only contributed to the mathematical content but also corrected many of the grammatical and stylistic mistakes in a prelimin~ry version. an outstanding mathematician and educator. program.. Also. my deepest gratitude goes to Peter Bilton.
r acknowlefoge ~r. I am grateful to F:avi f(umar for the long hours he spent on the computer preparing the final version of the manuscrirt. Cary Ostedt.a CII'N 01'11 LEI:GNI ENTS the document. Vis responsibilities have subsequently been assumed by Robin Symes. . the mathematics editor of John t·riley and ~ons. In addition. r am grateful to my wife. s. Poward lnton. Last but not least. I am also indebted to my class of l~el for preparing the soJutions to the home~orks. for his help in speeding the publication rrocess.xi i . for her heIr in proof reading and for her patience ~uring yet another of My projects. ['. for adding useful comments. anc rv~n Niven. I:avic Singer. . to Fetty Tracy and ~artha Bognar for their professional anc flawless typing assitance. ~lso.audrey. and in general. for keeping me very organized. I thank the following professors for refereeing the manuscript anc for recommending its publication: ~lan Tucker. and to Virginia Penade for her technical editing.
8 Terminology ~ethod ~athematical 23 34 40 50 SS 6~ Cuantifiers Part I: The Construction 5. Ppecial Proof Techn iques 12.. t1ots or Nots Lead to Knots II. Cuantifiers Part IV: Specialization ~. 8umrnary Appendix ~: Putting It J\11 Together: Part I 101 1C'g Append ix El: rutt ing It HI Together: Fart II Solutions to rxercises Glossary of Mathemetica1 Symbols Index ll7 11'7 IG9 xIIi . Cuantifiers Part III: Induction ~ethod 7. The ContraFositive ~ethod 72 78 83 ~2 1(. The ForwardBackward 3. On Cefinitions and ~. The Truth of It ~11 ~ethod 2.contents 1. The Contradiction ~ethod 9. Cuantifiers rart II: The Choose 6.
xlv CONTENTS TABLES Table 1. The Truth Table for "NOT B Implies NOT A" Table 4.2 42 6(' 1 ('II 109 . Summary of Proof Techniques 5 15 31 96 DEFINITIONS Definitions Definition Defini tion Definition Definition Definition Definition Definition Definition 1 . The Truth of" " Impl ies E" Table 2.10 11 24 24 35 ~5 12 13 14 15 16 17 H~ t. Proof of Example 1 Table 3.
how to read and do proofs an introduction to mathematical thought process .
.
Coing so requires that you use a certain amount of creativity. many proofs that appear in textbooks and journal articles are not presented properly. a new methoo of thought. ~athematics is the language of mathematicians. you must learn a new language. but as in learning any new language. Thus. and a proof is a method of communicating a mathematical truth to another Ferson who also "speaks" the language. ~ second objective of this pa~phlet is to teach you to develop and to communicate your own proofs of known mathematical truths. The approach of this pamphlet is to categorize and to explain the v~rious techniques that a~e used in proofs. Learning to do so will enable you to study almost any mathematical subject on your own. Just as there are many ways to express the same idea in any language. 'This pamphlet explains much of the basic "gramMar" you will need.1 the truth of it aU The objective of mathematicians is to discover ane to communicate certain truths. to understand and/or present a proof. a lot of practice on your part will be needed to become fl uent. intuition. Properly presented. A remarkable property of the language of mathematics is its precision. 'The 1 . and experience. a proof will contain no ambiguity: there will be no doubt as to its correctness. a desirable goal in itself. the rroofs are rresented properly for someone who already knows the language of mathematics. more appropriately stated. One objective is to teach you how to read and how to understand a written proof by ieentifying the techniques that have been used. so are there different proofs for the same mathematical fact. Unfortunately.
proof is a formal method for accomplishing this task. The twelfth chapter is a complete summary and it is followed by two appendices that illustrate the various techniques with several examples. and why. when attempting to create your own proof. As you will soon discover. Advanced students who have seen proofs before can read the first two chapters. 'the more aware you are of your thought process. this pamphlet describes not only how the proof techniques work. The ultimate objective..2 THE TRUTH OF IT ALL proof techniques presented here are designed to get you started and to guide you through a proof. but also.. . The basic material on proof techniques is presented in the next eleven chapters. the particular form of . however. Consequently. anyone with a good knowledge of high school mathematics. a fundamental problem of interest in Mathematics is to show that if ~ is true then E is true. Given two statements A an~ E. and subsequently read the two appendices to see how all of the techniques fit together. and E can often inficate a way to proceed. Some exaJllples of statements are: . It is often the case that a correct technique can be chosen based on the very form of the problem under consideration. The remainder of this chapter explains the types of relationships to which proofs can be applied. The first step in this direction is to reach the level of being able to read proofs and to develop your own rroofs of already known facts. the bet te r i tis. While the goal is an admirable one. learn to select a proof technique consciously before wasting hours trying to figure out what to 00. The pamphlet is designed to be read by. is to use your newly acquired skills and language to discover anf to communicate previously unknown mathematical truths. skip to the summary chapter. each of which may be either true or false. This alone will give you a much deeper and richer understanding of the matheMatical universe around you. when each technique is likely to be used. it is extremely difficult to attain. Therefore.
Two different lines in a plane are either parallel or else they intersect in exactly one point. In other words. '2. 'J.THE TRUTH OF IT ALL 1. :II 3 Observe that statement (1) is always true." For the most part. ~s such." ~athematicians are often very lazy when it comes to writing. a proof directed toward a high school student would probably require greater detail. the statement "if A is true then P is true" is shortened to "if ~ then 8" or simply"' implies E. textbooks do not use the symbolic notation." For instance. a proof of statement (5) aimed at convincing a mathematics professor might consist of nothing more than Figure 1. but teachers often do. a mathematician would often write "~ => B" instead or "A implies E. It therefore becomes necessary to have some meth00 for "proving" that such statements are true. and statements (3) and (4) can be either true or false depending on the value of a variable. For brevity. It is the lack of sufficient detail that can often make a proof difficult to read and to understand. In order to do a proof. Therefore . a proof should contain enough mathematical detail so as to be convincing to the person(s) to whom the proof is addressed. '2x = 5 and y 1. (2) is always false. they have developed a symbolic "shorthand. x is not> O. ~. Thus." The statement A is often called the hypothe4~4 and B the conctu4~on. There is an angle t such that cos(t) = t. and eventually you may find it useful too. a mathematical proof is a convincing argument that is expressed in the language of mathematics. you must know exactly what it means to show that "if ~ is true then E is true. It is perhaps not as obvious that statement (5) is always true. cne objective of this pamFhlet is to teach you to decipher these "condensed" rroofs that are likely to appear in textbooks and other mathematical literature. perhaps even the nefinition of cosine. 5. 1 = o. For instance. Cn the other hand.
~ complete list of the symbols can be found in the glossary at the end of this pamphlet. and I? themselves are true. ask yourself in which of the four cases would you be willing to call your friend a liar. . A is ~. Consequently. impl ies E" are true wi 11 depend on whether ~. Fig. Suppose. for example. it has rained. that your friend made the statement: "rf it rains then ~ary brings her umbrella." To determine When the statement "1 implies B" is false..e. A is 2. ~ is true true false false and and and and E B B E is is is is true. false. and yet ~ary did not bring her umbrella. In the second case. A is 4.. the statement ~ is "it rains" and E is "Mary brings her umbrella. false. t such that It seemS reasonable that the conditions under wh ich "." Here. 1. true.4 THF TRUTH OF IT ALL this document will include the aFproFriate symbols but will not use them in proofs.. In the first caSe (i. ~ proof that there is an angle cos(t) ::: t. when it does rain and ~ary does bring her umbrell~) your fr iend has told the truth. there are four possible cases to consider: 1.
an~ ultimately. as summarized in Table l. it does not rain. Thus.. ~ truth table is a method for determining when a comrlex statement (in this case. implies P") is true by examining all rossible truth values of the in(1ividual statements (in this case. Pere your friend has not told the truth. "A In general. but rather. the statement "~ imrlies E" is true in each of the four cases except the secon(1 one. According to Table I. The Truth of ". Illlrlies P" ***~*~*******~*********************~* B II IJI'lFl ies E True True False false True False True False True False '!rue True ************************************* . "l. I:'oing so will require a certain amount of creativity on your part. Note that a proof of the statement "A imrlies E" is not an attempt to verify whether A anc F themselves are true. ether examples of truth tables will appear in Chapter 3. your ability to show that n is true will oepend very he~vily on the fact that you have assumed ~ to be true. You would not really want to call your friend a liar in the case of no rain because. ~ an(1 E). to show that B is a logical result of hav ing assumed that ~ is true. The rroof techniques rresented here Table 1. you will have to discover the linking relationship between ~ ann B.. Your goal is to concl ude that the statement to the right (namely E) is true.THE TRUTH OF IT ALL 5 as your friend said she would. when trying to show that impl ies E" is true. Tablo I is an example of a t~uth table. in cases (3) an(1 (/I). you can assume that the statement to the 1 eft 0 f the word "imrl ies· (namel y 11) is true. your friend only said that something woul~ happen if it did rain. Finall y.
be ~ 0.) (a) ax 2 + bx + c = C 2 Cb) (b + " b .2. you along Hereafter. (<") The sum of the first n positive integers is n(n + 1)/2.~ac) /2a (c) Triangle YYZ is simil?r to tri~ng)e PST. then tho two linear equations ax + by = e and cx + ~y = f Cen he solve~ for x end y. the minimu~ value of x(x . «(I) 3 + n + n 2 (e) sin(Jt/2) < sin(Jt/Ll) (f) For every angle t.1) is at least 1/ 4 • 1. ~hich of the following ere statements? (Recall that a state~~nt must be either true or false.6 THE TRUTH OF IT ALL are designed to get you started and guide the path. ~ and E will be statements which are either true or false." Exercises 1. e. do you want to show that l is true or false? Fxr1ain. rq) ~ben x is A real number. . c. has an area of z2/.3. sin2(t) + cos 2 (t) = 1. 1. and f are real numbers ~ith the prorerty that ad . irentHy the hypothesis an~ the conclusion. For each of the following rrob!ems. (f) If P an~ Q are positive real numbers ~ith ypq F (r + q)/? then F ~ q. b. If you are trying to frove thC'lt "I il'lrl jes P" is true ant! you know thClt r is fa] se. (a) If the right triangle XYZ with siees of lengths x and y. and hypotenuse of length z.1. The problem of interest will be that of showing "~ implies E. then the triangle XY: i s i so sc e 1 e s • (b) n is an even integer => n 2 is an even integer.. c. (c) If a. (e) r is real and satisfies r2 = 2 imrlies r is irrational.
(b) (. impJ ies P) imrl ies C.THE TRUTH OF IT ALL 7 1. following . and give your reason.P. (a) A implies (B implies C). Us ing Table 1.5. c'letermine the cone i tions. under which the following statements are true or fal se .4. (a) If 2 > 7 then 1 > ::. (d) If x = 3 then 1 > 2. Prepare a truth tabl~ for each of the statements. ( c) If x :3 then ] < 2. ( b) If '2 < 7 then ] < 3. 1.
2. fpecial emrhasis is 0iven to th~ material of this chaFter because all of the other proof techn iquE's wi 1 J USE> the forwar(lbackwarc methof. of The first steF in ~ny proof requires recognizing the statements /'l anI" I? In general. imd hypotenuse of length z.::In area of z2/~. everything after the word "if" and before the ~ord "then" cOMprises statement I. x x Y Fig.' y. If the right trianqle YYZ with sides of lengths x an. the hypothesis) is everything t~1at you are trying to Frove (i. "i Example 1. has .2 the forwardbltckward method The purpose of this chapter is to describe one the fundamental rroof techn j ques: the 6oltwa'tdbackwaltd method. 8 . Consider the following excmr1e..e. ~lternatively. whilE' everything after the word "then" constitutes statement r.. the conclusion) is r. The ri':lht triangle XYZ. everything that you are assuming to be true (Le. then the triangle YYZ is isosceles (see Figure ?).
the next steF in the hackward process is to answer it. how can ~u show that a triangle is isosceles? of its sides Certain~y one way is to show that two . For Example 1. clearing away irrelevant ~etails (such as the fact that the triangle is called XYZ instead of ~eC). you will be going through a 6o~wa~d p~oce~~. and hypotenuse of length z has an area of z2/ 4• 8: The triangle XYZ is isosceles. Eoth of these processes will be described in detail. Cn the other hand.FCRWARVB~CKWAPV 9 Outline of proof. A properly posed abstraction question should contain no symbols or other notation from the specific problem under consideration. In attempting to figure out just how to reach the conclusion that B is true. when you make specific use of the information contained in A. The question obtained from statement B in such problems will be called the ab~t~act~on que~t~on. posed the In any event. In the hackward process you begin hy asking "How or when can r conclude that the statement P is true?" The very manner in which you phrase this question is critical since you must eventually be able to answer it. when proving "A implies e. thus allowing you to focus on those aspects of the problem that really seem to matter. you will be going through a backwa~d p~oce~~. once have ~u abstraction question. by asking the abstract question. statements: A: In this example one has the T~e right triangle XYZ with sides of lengths x and y. you calIon your general knowledge of triangles. the correct abstract question is "How can I show that a triangle is isosceles?" ~bile it is true that you want to show that the particular triangle XYZ is isosceles. The key to many proofs is formulating a correct abstraction question." you can assume that A is true and you must somehow use this information to reach the conclusion that B is true. Returning to the example. The question should be posed in an abstract way. Recall that.
with the property that if you could show that Bl is true then B would be true. El. x and yare also real numbers. Choosing the correct one is more of an art than a science. there will be . show that two of ~t~ sides have equal length. and then applying that answer to the specific situation will be referred to as the db~t~act~on p~oce~~. How can you show that Bl is true? Eventually you will have to make use of the assumption that A is true. you should show that x . This will illustrate some of the difficulties that can arise in the backward process. In fortunate circumstances. Can you pose the new abstraction question? Since x and yare the lengths of two sides of a triangle. for then it will follow that F is true. answering it abstractly. The abstraction process has given you a new statement.y.10 FCRWARrBACRWARD have equal length. the new statement is: 81: x =y If you can show that x = y. not that x = z or y = z. you would most likely do so now. Cbserve that answering the abstraction question is a tw~ phase process. and when solving this problem. but for the moment. a reasonable abstraction question would appear to be "How can I show that the lengths of two sides of a triangle are equal?" ~ second perfectly reasonable abstraction cuestion would be "How can I show that two real numbers are equal?" ~fter all. Referring to Figure 2. Once you have the statement El. Next. then the triangle XYZ is isosceles. First you give an abstract answer: to show that d triangle is isosceles. you apply this answer to the specific situation: in this case. let us continue working backwarc by repeating the abstraction process on the new statement Pl. For the example above. The process of asking the abstraction question. to show that two of its sides have equal length means to show that x = y. Cne of the difficulties that can arise in the abstraction process is the possibility of more than one abstraction question. all of your efforts must now be directed toward reaching the conclusion that El is true.
A cursory examination of the contents of statement A does not seem to provide much information concerning the angles of triangle XYZ. you may have to proceed by trial and error.y) = O. Thus.FORWARrBACKW~RD 11 only one obvious abstraction question. For instance. ~pplying this answer to the specific statement Bl means you would have to show that (x . Now one is faced with the question "How can I show that two real numbers (namely x and y) are equal?" One answer to this question would be to show that the difference of the two numbers is c. there is another perfectly acceptable answer: show that the first number is less than or equal to the second number and also that the second number is less than or equal to the first number. Applying this answer to the specific statement Fl. this would mean that you have to show that angle X equals angle Y. ~oreover. insight. creativity. diagrams. Can you do this for the two abstraction questions above? ror the first one. associated with the abstraction question "How can I show that a triangle is isosceJes?" is the answer "Show that the triangle is equilateral. experience. as will be done in this case. since one of its angles is 90 degrees. . Regardless of which question you finally settle on. Unfortunate1y. you might choose an answer that will not permit you to complete the proof. the next step will be to answer it. For the triangle XYZ of Figure 2. One general guideline is to let the information in A (which you are assuming to be true) help you to choose the question. For this reason. first in the abstract and then in the specific situation. the other abstraction question will be chosen. In other cases. a second difficulty can arise in the backward process. there may be more than one answer to it. you would have to show that x s y and y s x. This is where your intuition." Of course it will be im~ossible to show that triangle XYZ of Example 1 is equilateral. you might show that two sides of a triangle have equal length by showing that the angles opposite them are equal. Fven if you choose the correct abstraction question. and graphs can play an important role.
The forward process involves starting with the statement ~. and hypotenuse of length z. that you choose the answer of showing that their difference is O. the statement ~ is "The right triangle XYZ with sides of length x and y." Cne fact that you know (or should know) as a result of A . and hence so would~.12 FCRWAR~8ACKWARr Returning to the abstraction question "How can I show that two real numbers (namely x and y) are equal?" suppose. which you know to be true as a result of A being true. Yet another problem can arise in the abstraction process. You must ultimately make use of the information in P. One abstraction question is "How can r show that the difference of two real numbers is ~?" At this point it may seem that there is no reasonable answer to this question. 82. for the sake of argument. then in fact PI would be true. which you assume to be true and deriving from it some other statement AI. It should be emphasized that the statements derived from A are not haphazard. keeping in mind the fact that the last statement obtained in the backward process was "x . but for the moment. Rather. has an area of z2/ 4. let us continue once more with the abstraction process applied to the new statement P2. This last statement should act as the guiding light in the forward process. with the property that if you could show that 82 is true. the abstraction process has given you a new statement.y = 0. Specifically. The abstraction question might have no apparent answer! Do not despair: all is not lost. It is time to do so through the forward process.y = 0 Now all of your efforts must be directed toward reaching the conclusion that 82 is true." For the example above. Let us return to Fxample l. Cnce again. the new statement is: 82: x . Nowhere have you made use of this fact. they are directed toward linking up with the last statement derived in the backward process. Remember that when proving "A implies B" you are allowed to assume that A is true.
because the area of a right triangle is onehalf the base times the height. Continuing with the forward process. it is possibJe to combine ~l and ].lRI:'EJlCKWARD 13 being true is that xy/2 = z2/ 4 ." t~'hiJe there are no specific guidelines for producing new statements. For instance. "angle X is less than 90 degrees.y = 0. in this C('lse xy/2.I by the = z2 The forward ~rocess can also combine and use the new statements to produce more true statements. you can multiply both si~es of 1: by t and subtract ?xy from both siees to obtain: A4: (x 2 _ 2xy + y2) (' By factoring you can obtain: AS: (x  y)2 "" 0 Cne of the most common steps of the forward process is to rewrite statements in different forms. So you have obtained the neW statement: Another u~eful statement follows froM Pythagorean theorem. as was done in obtaining. for instance.. For instance. It is for this reason that z2 was eliminated from ].I] anc A7. so you also have: A2: (x 2 + y2) ]. which was the last one derived in the backward process.t (1n(1 '5. you should attempt to rewrite ]. For Fxam[le].FCRWJ.12 by replacing z2 in '1 with (x 2 + y2) from A2 obtaining the statement: A3: xy/2 One of the problems with the forward process is that it is also possible to generate some useless statements.I~ so as to make it look more like E2. the final step in the forward ~rocess (and in the entire . keep in mind the fact that the forward process is directed toward obtaining the statement P2: x .
Py the Pythagorean theorem. is true. and hence S.CJ<W''' R[) proof) is to take the square root of both sides of the equality in A5. (x 2 + y2) = Z2.y)2 = (x 2 . The statement will be proved by establishing that x = y. and space. Pence x = y and the triangle XYZ is isosceles. as required. Finally. meaning "w~ich was to be demonstrated. thus obtaining precisely the statement E2: x . a highly condense~ version is usually presented and often makes little or no reference to the backware process.:?xy + y2) = C. Proof of Example 1. are used as they stand (or the Latin words quod e~at demon6t~andum. proof will be For example: partly Proof of Example 1.C. FrOM the hypothesis an~ the formula for the area of a right triangle.y the Pythagorean theorem. z2 = (x 2 + y2) and hence (x 2 + y2) = 2xy.11 .'1 realize that. anCl on substituting (x 2 + y2) for z2 and performing some algebraic manipulations one obtains (x .E. rt is interesting to note that the forward process ultimately produced the elusive answer to the abstraction question associated with P:?: "Pow can r show that the difference of two real numbers is r?"which was to show that the square of the ~ifference is 0 (see l~ in Table 2). the area of XYZ = xy/2 = z2/~. 'Ihe steps and reasons are summClrized in Table 2. Father.~xy + y2) = O. which in turn is done by showing that (x . ror the rroblem above it might go something like this. for this woulr'l requi re fClr too much time. The proof is now complete since you started with the assumption that A is true and used it to derive the conclusion that E2. But the area of the triangle is (l/:?)xy = (1/4)z2. p. in general. or (x 2 . it will not be practical to write clown the entire thought process that goes into a proof. so that 2xy = z2.y = O.14 FeRlA1 AR[)S~.11 (The "II" or some equivalent symbol is usually used to indicate the enCl of the proof.y)2 = r. you shoulr.") Sometimes the shortened backward anCl partly jorward. effort. Sometimes the letters C.
2xy + y2) = 0. Proof of Example 1 15 ****************~************************************ Statement A : Area of XYZ is z2/4 ~1: xy/2 == z2/ 4 Reason Given Area == Cbase) (height)/2 A2: x 2 + y2 = z2 A3: xy/2 = (x 2 + y2)/4 A4: x 2 . Although this version is slightly unnatural. Tn order to see that 2xy = z . thus completing the proof. or equivalently. giving little more than a hint of how to do the proof.2xy + y2 = 0 AS: (x  Pythagorean theorem Substitute A2 into Al Algebra Factoring M Take square root in AS Add Y to both sides of e2 Since 81 is true y)2 == 0 B2: (x B1: x == y y) =0 B : XYZ is isosceles ***************************************************** The proof can also be written entirely from the backward process. it will be shown that x = y by verifying t~at (x _ y)2 = (x 2 . To reach the conclusion. that (x 2 + y2) = 2xy.// Proofs found in research articles are often very condensed. note that (1/2)xy is the area of the triangle and it is equal to (l/4)z2 by hypothesis. for the Pythagorean theorem states 2that (x 2 + y2) = z2. or equivalently. Proof of Example 1. it is worth seeing. This can be established by showing that 2xy = z2. For example: . that (1/2)xy = C1/4)z2. nonetheless.FORWARDBACKWARD Table 2.
you will have to figure out which proof technique is being used (since the forwardbackward method is not the only one available). the harder this process will be. apply the abstraction process to Bl obtaining a new statement.y) • o. You should strive toward the ability to read and to dissect a condensed proof. Was it algebraic manipulation (as we know it was) or something else? Unfortunately. with the property that if Bl is true. To do so. Some examples of how to read condensed proofs appear in the two appendices. Thus the triangle is isosceles as required. which you are trying to conclude is true. Bl. these shortened versions are typically given in mathematics books. and reasoning that was involved will precede each condensed proof. then so is Bl (and hence B). be able to verify all of the steps involved. your life will be made substantially easier because an outline describing the proof technique. and it is this fact that makes proofs so hard to read. Through the abstraction process of asking and answering the abstraction question. A summary of the forwardbackward method for proving w~ implies BW is in order. finally.11 Note that the word whence w effectively conceals the reason that (x . from what is written. The more condensed the proof. Begin with the statement e. In the latter case. it is time to start the forward process. B2. Continue in this manner until either you obtain the statement ~ (in which case the proof is finished) or until you can no longer pose andlor answer the abstraction question fruitfully. hence (x . in which you derive a sequence of statements from A .18 FORWARDBAC~ARD Proof of Example 1. In this pamphlet. you will have to discover the thought process that went into the proof and.y)2 • O. derive a new statement. The hypothesis tOjether with the Pythagorean theorem yield (x 2 + y ) • 2xy. then so is B. with the property that if B2 is true. Remember that the abstraction process is motivated by the fact that A is assumed to be true. To that end. Out of necessity. methodology. ~ll efforts are now directed toward establishing that Bl is true. these outlines will be more succinct than the one given in Example 1. Then.
~~ /0= I ( A ~. for there are likely to be several false starts and blind alleys. . In the backward process. you start at the needle and try to work your way out of the haystack toward the statement A (see Figure 3) • .~ . BaCkwar~ )~ • B . Remember that the goal of the forward process is to obtain precisely the last statement you had in the backward process. . When you work forward from the assumption that A is true. Another way of remembering the forwardbackward method is to think of a m~ze in which ~ is the starting point and B is the desired ending point (see Figure 4)... you start somewhere on the outside of the haystack and try to find the needle. It might be necessary to alternate several times between the forward and backward processes before you succeed....FORWARDBAC~ARD 17 that are necessarily true as a result of A being assumed true. The forward and backward processes can easily be remembered by thinking of the statement B as a needle in a haystack. at which time you will have successfully completed the proof. Finding a needle in a haystack. 3. ' .\ " Forward \ Fig.
A Forward • I II   I~II B Backward ~ U ~ •    ~ F19. The maze. . the forwardbackward method is probably the first technique to tryon a problem unless you have reason to use a different approach based on the form of B.4. you will gain much insight into the relationship between ~ and 8.18 FCRWARCBACKWARD As a general rule. In any case. as will be described shortly.
and T = {real numbers x: x ~ I}." Which of the following abstraction questions is incorrect. and why? Explain what is wrong with the other choices.3. Consider the problem of showing that "If R = {real numbers x: x 2 . then R intersect S is a subset of T. S . of proof 2. For each of the following problems.1.x s e}. Explain the difference between the forward and backward processes.~'hich of the following is the most correct abstraction question.4. . {real numbers x: fxl) (x3) ~ OJ. list as many abstraction questions as you can (at least two). .2. and why? (a) How can I show that the maximum value of a parabola is ~ to a number? (b) How can I show that a number is s to the maximum value of a polynomial? (c) How can I show that the maximum value of the function _x 2 + 2x + I is ~ to a number? (d) How can I show that a number is s to the maximum of a quadratic function? 2.FORWARDBACKWARD Exercises 19 Note: All proofs should contain an outline as well as a condensed version. then the maximum value of x 2 + 2x + I is ~ 2. Cescribe how each one works and what can go wrong. (a) How can I show that a set is a subset of another set? (b) How can I show that the set R intersect S is a subset of T? (c) How can I show that every point in R intersect Sis ~ to l? (d) How can I show that the intersection of two sets has a point in common with another set? 2. Consider the problem of proving that If x is a real number.. How are the two processes related to each other? 2.
i f SU is a perpendicular bisector of RT. and c are real numbers for which a > 0. (a) If a. (a) How can r show that two real numbers are equa l? (b) How can I show that two triangles are cong ruent? (c) How can I show that two lines are parallel? (d) How can r show that a quadrilateral is a rectangle? 2. then the solution to the equation ax 2 + bx + c = 0 is positive. (1) pose an abstraction question.6. (2) answer it abstractly. then 1. respectively. and 12 are parallel. and b 2 . list as many answers as you can (at least three). (a) If 1 and 1 are the tangent lines to a circle C at \he two endpoints e. 2. and (3) apply your answer to the specific problem. and e. (Note: Continuity is a property of a function. b < P. (d) If n is a given integer that satisfies 3n 2 + 2n + 8 = 0.20 FORWARCB~CKW~RD Be sure that your questions contain no symbols or notation from the specific problem. b.) (c) If n is an even integer then n 2 is an even integer. (b) In the following c:'Iiagram. 2 of a diameter d. (b) If f and 9 are continuous functions then the function f + g is continuous. .4ac = C. For each of the following abstraction questions. For each of the following problems. then 2n2 . then triangle RST is equilateral.3n = 2.5. and RS = 2RU.
2 (a) The real number x satisfies x .FORW~RDBACKWARC 21 2. Consider the problem of proving are real numbers such that x 2 + y2 + X = 3. since x = 0 and x + y = 0.// (b) Rewrite the condensed proof of part entirely from the backward process." Tn from the hypothesis. and the abstraction questions and answers. (d) The triangle UVW is equilateral. it must be that x == O.3x + 2 < O. First it will be shown that x s 0.3) + (y . for then.x (b) y2 == 25/6 . and why? (a) y2 = 3 .8.6 and t.10.25 == 0 ((I) ex + 5) == 6y /(x . together wi th the following rules for creating new words from old ones.(XLi6)2 (c) (3 . since y ~ (\. since x ~ 0 by the hypothesis. Proof.2) = 25.y2)2 + 6~2 .can be applied in any order. which of not val id. list as many statements as you can (at least three) that are a result of applying the forward process for precisely one step.2. x + y = 0. so x == yo Also. (c) The circle C consists of ail values 2 for x and y that satisfy (x . then x = 0 and y = 0. to see that hence x == y s ('\. it follows that y s 0 and rinally. then Iyl = 2. 2.7. by the hypothesis. For each of the following hypotheses. write an outline of the proof indicating the forward and backward steps.5) that "If x and y 6y2 .9. (a) 2. y == 0. To see that x S 0.25 == 0 and working forward the following is 2. Consider an alphabet consisting of the two letters . The rules. ." (a) For the following condensed proof. Consider the problem of proving that "Tf x and y are nonnegative real numbers that satisfy x + y == 0. (b) The sine of an~le X in triangle XYZ of Figure 2 is 1/. it must be that 0 + y = y = 0.
~.U could become ~tut~ ) • 2. Specifically. ~. t~. then the area of the triangle is z /4.12.22 FORW~RDBACKW~RD 1. Erase tt from the current word (e. list all of the words for which an application of one of the above rules would result in t~t. ~tt~ could become ~~). could become t~.." (d) Prove that "If ~ then tt~t." 2.g. ~t~~~ could become ~tt). Prove that if the right triangle XYZ of Figure 2 is 2isosceles.6(b) is true.. Double the current word (e. (b) Apply the backward process one step to the word t~t. Replace u~ in the current word by t (e.. (c) Prove that "If ~ then t~t.g.g. ~. Add the letter t at the ri9ht end of the current word if its last letter is • (e. Prove that the statement in 2. (a) Use the forward process to derive all of the possible words that can be obtained in three steps by repeatedly appl ying the above rules to the initial word ~. 2.g.11. ..t).
In addition. If you choose not to. 'rake. A de6inition is nothing more than a statement that is agreed on by all parties concerned. for example. Nothing says that you must accept this definition as being correct.3 on definitions and mathematical terminology In the previous chapter you learned the forwardbackward method and saw the importance of formulating and answering the abstraction question. then we will be unable to communicate regarding this particular idea. the notion of "a positive integer greater than one that is not divisible by any positive integer other than one and itself . Usually they are motivated by a mathematical concept that occurs repeatedly. we agreed that it is true in all cases except when A is true and e is false." that is abbreviated (or defined) as a "prime. There we defined what it means for the statement "A implies 8" to be true. Several other examples of definitions would be: 23 . One of the simplest yet most effective ways of answering an abstraction question is through the use of a definition. as will be explained in this chapter. you will learn some of the "vocabulary" of the language of mathematics. In fact." especially if the concept comes up frequently. Specifically. refinitions are not made randomly." Surely it is easier to say "prime" than "a positive integer greater than one . a definition can be viewed as an abbreviation that is agreed on for a particular concept. You have already come across a definition in Chapter 1.
The statement A OR B (written A \I B) is true in all cases except when A is false and B is false.are left undefined.24 DEFINITIONS ANO TERMINOLOGY 1. A positive integer p > 1 is prime if the only positive integers that divide pare 1 and p. of its Definition Definition Definition Definition Definition Definition Definition 4. An integer n is even if and only if its remainder on division by 2 is c..tends to be used instead of if and only if. A triangle is isosceles if two sides have equal length. An integer n is odd if and only n • 2k + 1 for some integer k. 3. if 7. One could possibly try to define a set as a collection of objects. A real number r is a rational number if and only if r can be expressed as the ratio of two integers p and q in which the denominator q is not o. but.is too .and B implies A. The statement A AND B (written A 1\ B) is true if and only if ~ is true and B is true. Definition Definition Definition 10.and point. 8. in general. Two pairs of real numbers and and (x 2 'Ya_) are equal if ~ • Y2· 5. but to do so is impractical because the concept of an object.have been used in some of the definitions. 6. such as set.Some terms. 2.9. if.. Observe that the words if and only if. Two statements A and e are equivalent if and only if A implies e. An integer n divides an integer m (written nlm) if m· kn for some integer k.
how do you select the definition and what happens to the other alternatives? Since a definition is simply something agreed on. n is an integer whose remainder on division by 2 is 0). One would then be led to ask for the definition of an "object. Using definitions to work forward and backward is a common occurrence in proofs." Using the alternative concept. In the proof of Example 1. one then creates the statement B: "n is an integer that can be expressed as two times some integer. you must show that "A implies S" and "P implies A" (see refinition 8). A second plausible definition for an even integer is "an integer that can be expressed as two times some integer. the notion of an even integer that was introduced in Definition 5. Recall the very first one. this would be accomplished by using Cefinition 5 to create the statement A: "n is an integer whose remainder on division by 2 is 0. if you know that an integer n is odd. one shows that two of its sides have equal length. anyone of the alternatives can be agreed on as the definition. so when more possibilities exist. then by Definition 6 you would know that n = 2k + 1 for some integer k. which was "How can I show that a triangle is isosceles?" Using Definition 3. Take. then P is true (i." and so on. in order to show that a triangle is isosceles. Once the definition has been chosen." To establish the fact that the definition is equivalent to the alternative. a definition was already used to answer an abstraction question. for example.. For the case of an even integer. it would be advisable to establish the "equivalence" of the definition and the alternatives.DEFINITIONS AND TERMINOLOGY 25 vague.e. For instance. and so on. if S is true then A is true too. It is often the case that there are two possible definitions for the same concept. Such philosophical issues are beyond the scope of this document.e. . ~oreover. Then you would know that if ~ is true (i. n is an integer that can be expressed as two times some integer)." Cf course there can be only one definition for a particular concept. Definitions are equally useful in the forward process..
say k (i. One is obtained directly from the definition." Whenever you are asked to show that "A if and only if e. in the forward process. the only question being which integer. it can also be advantageous. Since n is an even integer. The second answer comes from the alternative: you can show that the integer can be expressed as two times some integer. in some proof. n = 2k).26 DEFINITIONS AND TERMINOLOCY The statement that ~ is equivalent to e is often written "A is true if and only if B is true" or. using the alternative. you are led immediately to the abstraction question "How can r show that an integer (namely n2 ) is even?" By choosing the alternative over the definition. so one way to show that an integer is even is to show that its remainder on division by 2 is 0. you can answer this question by showing that n2 can b~ expressed as two times some integer. for example. as was the case in Example 1. "1. you derive the abstraction question "How can I show that an integer is even?" As a result of having obtained the equivalence of the two concepts. Proceeding by the forwardbackward method. more simply. While the abil i ty to answer an abstraction question (or to go forward) in more than one way can be a hindrance. as shown in the next example. i f you know that n is an even integer.i f and only if B. that. Similarly. Suppose.. . then you would have two possible statements that are true as ~ result of this: the original defini tion and the al ternRtive. you now have two possible answers at your . OUtline of proof. So n2 = (n) (n) = (2k) (2k) = 1k2 = 2(2k 2 ). If n is an even integer then n 2 is an even integer. The answer comes from the forward process." In mathematical notation one would write "A iff S" or "" <=> B. n can be expressed as two times some integer. Example 2." you must show that "A implies B" and "B implies A.fingertips.e. " It is quite important to be able to establish that R definition is equivalent to an alternative.
however. a proFosition is a true statement that you are trying to prove. The more statements that you can show are equivalent to the definition. and so n 2 is an even integer./ / A definition is one common method for working forward and for answering certain abstraction questions. and it is often easier to communicate the proof in pieces.CEFINITIONS AND TERMINOLOCY 27 Thus. Finally. theorem. it has been shown that n 2 can be written as two times some integer. a lemma is a preliminary proposition that is to be used in the proof of a theorem. These are called c. there is an integer k for which n = 2k. it is often the case that certain propositions follow almost immediately as a ~esult of knowing that the theorem is true. and a theorem is an important proposition. in proving the statement A implies B. and a corollary is a proposit1on that follows from a theorem. but it is much harder that way. a large number of equivalent statements can also make it ~ifficult to know exactly which one to use.o~olla~~e~... a lemma 1s a preliminary proposition that is to be used in the proof of a theorem. there are four terms that you will often come across in mathematics: proposi tion." then that C implies C." Each of these supporting propositions might be presented separately and would be referred to as a lemma. Consequently n 2 = (2k)2 = 2 (2k 2 ). Some propositions are (subjectively) considered to be extremely important and these are referred to as theo~em~.it may first be necessary to show that A implies C. that integer being 2k2. . In summary. lemma. A p~opo~~t~on is a true statement of interest that you are trying to prove.For example. Proof of Example 2. All of the examples that have been presented here are propositions.. the more ammunition you will have available for the forward and backward processes. In other words. Since n is an even integer. and this completes the proof. Of course this problem could also have been solved by using Definition 5. and corollary. Once a theorem has been established. The proof 0 f a theorem can be very long.and finally that D implies B. In dealing with proofs.
which states that the triangle XYZ is isosceles. it is necessary to see if RST also satisfies the hypothesis of Example I. and hypotenuse of length t satisfies t = (2rs. but a second answer is also provided by the conclusion of Fxamp1e 1.. One example of an axiom is the statement: the shortest distance between two points is a straight line. s. In order to find out. and hence be isosceles.28 tEFINITIONS AND TERMINOLOGY Just as there are certain mathematical concepts that are accepted without a formal definition. as did triangle XYZ. Outline of proof. The right triangle RST.One answer is to use Definition 3. The forwardbackward method gives rise to the abstraction auestion How can I show that a triangle (namely RST) is isosceles?. . R Fig. for then RST will also satisfy the conclusion. Perhaps the current triangle RST is also isosceles for the same reason as triangle YYZ. as will be shown in the next example. If the right triangle RST with sides of lengths rand s.~ further discussion of axioms is beyond the scope of this document. so are there certain propositions that are accepted without a formal proof. These unproved propositions are called axiom~. Just as a definition can be used in the forward and backward processes. Example 3. then the triangle RST is isosceles (see Figure 5). so can a (previously proven) proposition.
In the condensed proof that follows. when it does. and hypotenuse of length y. By the hypothesis. the corresponding lengths are x • r. Do not forget that you must match up the current notation to that of the previous proposition so that the hypothesis of the previous proposition can be verified.DEFINITIONS AND TERMINOLOGY 29 In verifying the hypothesis of Example 1 for the triangle RST. the hypothesis. Consequently. Thus. note the complete lack of reference to the matching of notation. and hence the conclusion. or equivalently. 1/4(t 2 ) = 1/2(rs)2 Thus. As such. Do not forget to observe that the hypothesis of Example 1 also requires that the triangle RST be a right triangle. . To be specific. the triangle RST is isosceles. To be specific. on squaring both 2sides and dividing by 4. one obtains 1/2(rs) = 114ft). t = {!FS. you must see if l/2(rs) = 1/4(t 2 ). and. y • s. as stated in the current hypothesis. since the area of triangle RST is 1/2'rs). Lnfortunately this "overlapping" notation can (and will) arise. it is particularly important to keep the symbols straight. and z = t. you must see if the area of triangle RST equals 1/4(t 2 ). Proof of Example 3. the area of the right triangle RST = l/41t ). to check the hypothesis of Example 1 for the current problem. as desired. The fact that 1/2(rs) = 1/4 (t 2 ) will be established by working forward from the current hypothesis that t = (1FS. Notice how much more difficult it would have been to match up the notation if the current triangle had been labeled WXY with sides of length wand x. so t 2 = 2rs. it is first necessary to "match up" the current notation with that of Example 1. which of course it is. of Example 1 is true.11 Using the conclusion of a previous proposition to answer an abstraction question is quite common. or equivalently.
More will be said about the NOT of a statement in Chapter 10.6Uive) Table 1 can be used to determine when each of these three statements is true.· that is. and vice versa.(called the tnvwe) 3. Whenever A is true.are: ].30 DEFINITIONS ANC TERMINOLOGY Associated with a statement A is the statement NOT A (sometimes written ~~). B impl ies A(called the c. Given two statements A and B.is true in all cases except when the statement to the left of the word implies. For instance. Note.ontJr. A is sufficient for B. as shown in Table 3. B follows from A. it showed how definitions and how ..a. More importantly. in all cases except when A is true and B is false. This observation gives rise to a new proof technique known as the contrapositive method that will be described in Chapter 9.for example: 1.is true under the same conditions as A implies B.onvwe) 2. that the statement NOT B implies NOT A. NOT B implies NOT A. A only if B. 5. you have already learned the meaning of the statement A implies B. The statement NOT A is true when A is false. 4. NOTA implies NOTB..(namely NOT B) is true and the statement to the right of the word implies(namely NOT A) is false. Three other statements related to A implies B. from Table 3. NOT B implies NOT A. This chapter has explained the meaning of many of the terms used in the language of mathematics. :.(called the c.po. B must also be true. 2. B is a necessary consequence of A.There are many other ways of saying the statement A implies B. In other words. the contrapositive statement. Truth tables similar to Table 3 can be derived for the converse and inverse statements and are left as exercises.. the contrapositive statement is true in all cases except when B is false and A is true.
c. Then use a definition to (1) answer the question abstractly and (2) apply the answer to the specific problem. (b) If sand t are rational numbers with t ~ 0. Exercises Note: All proofs should contain an outline as well as a condensed version. of proof 3. For each of the following conclusions. If (x 1 'Y1) and (x 2 'Y2) are real numbers satisfying: . then sit is rational. pose an abstraction question. (c) Suppose that a. e.DEFINITIONS AND TERMINOLOGY Table 3. Now i t is time to learn more proof techniques. and f are real numbers. d. b. (a) If n is an odd integer.1. then n 2 is an odd integer. The Truth Table for wNOT 8 Implies NOT AW 31 ***************************************************** True True False False True False True False False True False True False False True True True False True True True False True True A 8 NOT B NOT A A => B NOT B => NOT A ***************************************************** previous propositions can often be used in the forwardbackward method.
2 (b) If r is a real number such that r ~ 2. then ale. (a) The converse of A implies B.2.3.How are (a) and (b) related? (c) A CR B (d) A AND B (e) A AND NOT B (f) (NOT A) OR 8 How is (f) reI ated to . write down the converse.32 DEFINITIONS AND TERMINOLOGY aX 1 + bY1 • e. (a) If n is an odd integer then n2 is an odd integer. 3.4. and (n+l) are three consecutive integers. and contrapositive statements. Write truth tables for the following statements. (c) If the quadrilateral ABCt is a parallelogram with one right angle. (e) If a. 2 (a) If n is an integer for which n is even then n is even. . then 9 divides the sum of their cubes.(b) The inverse of A implies B. (c) If triangle RST is equilateral then the area of the triangle is (3/4 times the square of the leng th of a side. For each of the following hypotheses. If (nl). If n is a PQsitive integer greater than 1 for which 2" . (b) If sand t are rational numbers with t ~ 0. use a definition to work forward one step. then (X 1 'Y'l) equals (xI 'Ya ). then triangle XYZ is isosceles. 3.I is prime. aX I + bYI • e. then r is not rational. (d) If the right triangle YYZ of Figure 2 satisfies sin(X) • cos(X). then n is prime. (d) (e) cx 1 + dY1 • f. then sit is rational. then the quadrilateral ABCt is a rectangle. ex I + dyz • f. n.A implies 8?3. For each of the following propositions. inverse. b. and c are integers for which alb and blc.
" and ·C implies A. . Prove that if n is an odd integer.· ·e implies C." 3. 3.· ·C implies D. (b) Verifying the hypotheSis of Example 1. by: (a) Using the definition of an isosceles tr iangle. Prove that if "A implies F" and"e then "A implies Co" implies an C.9. and r. Suppose that you have a definition in the form of a statement A together with three possible alternative definitions. then the triangle UVW is isosceles. and hypotenuse of length w satisfies sin(U) = ~. (c) Verifying the hypothesis of Fxample 3. Prove that if n is an odd integer and m is odd integer. (a) How many proofs would it require to show that ~ is equivalent to each of the three al ternatives? (b) How many proofs woul~ it require to show that . then n 2 is an odd integer.S.8.· then A is equivalent to B and A is equivalent to C. l. 3.10. then mn 1s an odd integer. Prove that if the right triangle UVW with sides of lengths u and v. say S. then t = _/4." and "0 implies ~?" (c) Explain why the approach in part (b) is sufficient to establish that each of the alternatives is equivalent to the original definition (and to each other) • 3. 3.6.~ implies 8.DEFINITIONS AND TERMINOLOCY 33 fd) If t is an angle for which sin(t) = cos(t) and 0 < t < _." "8 implies C.7. Prove that if "A implies e. r.
These two groups of words are referred to as qUdntl6~e~~.4 quantifiersl: the construction method In the previous chapter you saw that a definition could successfully be used to answer an abstraction question. . Recall Cefinition 7 for a rational number as being a real number that can be expressed as the ratio of two integers in which the denominator is not zero. whereas the second one has "for all" ("for each.t qUdnt~6le~ "for all" and its associated proof technique is discussed in the next chapter.t qUdntl6le~ "there is" and with the corresponding proof technique called the con~t~uctlon method. The unlve~~d. They can always be identified by certain key words that appear in the statement. Two particular forms of E appear repeatedly throughout all branches of mathematics. This definition could just as well have been written using the quantifier "there are. and each one will give rise to its own proof technique. The remainder of this chapter deals with the ex." "for any"). The first one has the words "there is" ("there are.l~tentld. The quantifier "there is" arises quite naturally in many mathematical statements." Definition 11." "there exists")." "for every. The next four chapters provide you with several other techniques for formulating and answering an abstraction question that arises when e has a special form. 34 ~ real number r is rational if and only if there are integers p and q with q F 0 such that r = p/q.
you can see that such statements always have the same basic structure. the statement will have the following bClsic form: There is an "object" with a "certain property" such that "something happens. Cbject: integer x Certain property: x >22 Something happens: (x .." "there are. Using a quantifier to express this statement. the existence of only one such object) • There are many other instances where an existential quantifier can and will be usen. as is shown in the next definition. then 2there are usually two values of k that satisfy n = k (in this case. ~n integer n is a square if 2there an integer k such that n = k • is Note that i f an integer n (SClY.e. for example. Consider these examples.CONSTRUCTION METHO~ 35 Another such example arises from the alternative definition of an even integer. Fach time the quantifier "there is. and k only i f such that It is imFortant to observe that the quantifier "there is" allows for the possibility of more than one such object. Definition 13." The words in quotation marks depend on the particular statement under consideration.5x + 6) 0 Th~re . k = 3 or 3). that being an integer that can be expressed as the product of two times some integer. to identify. 1. and to write each of the three components. An integer n is even if there is an integer n = 2k. one obtains: Definition 12. but from the examples above. ~ore will be said in Chapter 11 about the issue of uniqueness (i. is an integer x > 2 such that (x . and you must learn to read.5x + 5) = O." or "there exists" appears. n = 9) is square.
0. the appearance of the quantifier "there is" strongly suggests turning to the forward process to produce the desired object. Cbject: angle t Certain property: none Something happens: cos(t) = t Observe that the words "such that" (or equivalent words like "for which") always precede the something that happens. d. Indeed. Sometimes it will be by trial and error.bc) . and f are real numbers with the property that (ad .38 CONSTRUCTION METHOD There are real numbers x and (2x + 3y) . the information in statement A will surely be used to help accomplish the task. Practice is needed to become fluent at reading and writing these statements.) and the symbol "~" for the words "such that" ("for which. The construction method was subtly used in Example ~. y > Something happens: (2x + 3y) y both > 0 such that 3. nonethless.y) . . Cf course you must show that the object has the certain property and that the something happens. e. b.y) • Object: real numbers x and y Certain property: x > 0. Mathematicians often use the symbol "3" to abbreviate the words "there is" ("there are. 3 an angle t ~ cos(t) = t. sometimes an algorithm can be designed to produce the desired object.3 2. It all depends on the particular problem. ~. then the two linear equations (ax + by) = e and (cx + dy) = f can be solved for x and y. 0 . The use of the symbols is illustrated in the next example.8 and (5x .) the desired object. c. produce. etc.8 and (5x ." etc. If a. The ioea is to construct (guess." then one way in which you can proceed to show that the statement is true is through the construction method. During the backward process." etc. if you ever come across a statement having the quantifier "there is. Bow you actually construct the desired object is not at all clear. Example 4.). but another example will serve to clarify the process. devise an algorithm to produce.
it is not very informative as to how these particular values were produced.bc) x • (de . Recall here that a proof is a convincing argument. for example.ce)/(ad . .be) and y = (af . even though the quantifier there are w does not appear explicitly.bf)/(ad . it is possible to divide this last equation by (ad . (ex + dy) • f. to obtain the values for x and y.bc).ce)/(ad . If you are clever enough to Wguess W that x • (de . (ax + by) = e and (ex + dy) • f.CONSTRUCTION METHOD 37 OUtline of proof. A similar process can be used to obtain y. you could start with the two equations (ax + by) • e and (ex + dy) = f. Proceeding with the construction method. In any event. by guessing these values for x and y. and that will complete the proof.bf)/(ad . As such.bc) since.bf).ce)/(ad . the statement Wa similar process can be used to obtain y • taf . taf . by hypothesis. then you are very fortunate. the issue is how to construct real numbers x and y such that (ax + by) = e and (cx + dy) = f.be). you obtain (ad . that This. but you must still show that the something (ax + by) • e and happens. you should recognize that the statement B has the form discussed above.bc). in which case a proof directed at you would have to contain the details of how y is obtained. thus obtaining x • tde . If you then use the information in A. is not hard to do. A more instructive proof would be desirable. While this Wguess and check w approach is perfectly acceptable for producing the desired x and y. for these values of x and y. For example.quantifiers occur frequently in problems and you should watch for them.bc)W might not be very convincing to you. you must still show that. of course. Observe that the statement B can be rewritten to contain the quantifier explicitly. in this case. this number is not '0. On starting the backward process. Also observe that. you have used the information in A since the denominators are not O. On multiplying the first equation by d and the second one by b and then subtracting the second one from the first one. wThere are real numbers x and y such that (ax + by) • e and Statements containing Whidden(ex + dy) = f.
(de .// The construction method is not the only technique available for dealing with statements having the quantifier "there is.5x/2 + 3/2 = o.1.bf). you must not forget to show that the something happens. Cad . (e) Between the two real numbers x and y. 0.bc).38 CONSTRUCTION METHOD Proof of Example 4.bc). and the something that happens. and then subtracting the two equations one obtains (ad . the certain property.sl < 0. Your "building supplies" consist of the information contained in A.000 feet high that is taller than every other mountain in the world. you must become a "builder. (ax + by) = e and Ccx + dy) • f. On multiplying the equation Cax + by) • e by d. Also.001. (d) There exists an angle t between 0 and ~/2 such that sin(t) • cos(t).ce)/(ad . (c) Through a point P not on a line i." and use your creative ability to construct the desired object having the certain property.bc) . identify the objects. there is a mountain over 20.bf)/Cad . A similar argument shows that y • (af . through P parallel to i.bc)x = (de . Cb) There exists an integer x that satisfies x 2 . Ca) In the Himalayas. By using the hypothesis.bc) yields x . To be successful with the construction method. Exercises Note: ~ll proofs should contain an outline as well as a condensed version. for these particular values of x and y. and it is not hard to check that. of proof 4. there are distinct rational numbers rand s for which Ir . and the equation Ccx + dy) • f by b. and so dividing by (ad . . there is a line ~." but it often works and should be considered seriously. For each of the following statements.
rn .5x/2 + 312 = O. then ale. has exactly n complex roots. Prove that if sand t are rational numbers t ~ 0. b. Reword the following statements using the symbols "3" and "~. Prove that if a. then sIt is a rational number. (a) (b) Prove that there is an integer x such that x2 .5. Is the real number unique? 4. 1 4. (b) Given an angle t. and . one can find an angle t' whose tangent is larger than that of t.CONSTRUCTION METHOD 39 4. say p(x). Is the integer unique? Prove that there is a real number x such that x2 . (c) At a party of n people. and c are integers for which alb and blc. say r1 ' . (d) A polynomial of degree n. for which p(r ) = • • • = p(rn ) = o.3.2. 4." (a) A triangle XYZ is isosceles if two of its sides have equal length.4. at least two of the people have the same number of friends.5x/2 + 312 = O.
7}." While it is certainly desirable to make a list of all of the elements in a set. Fortunately there is a way to 40 .7}.4. Thus. For example." Such statements arise quite naturally in many mathematical areas. a proof technique for dealing with statements containing the quantifier "for all.e method This chapter develops the c. 4. Each of the individual items is called a membe4 or element of the set and each member of the set is said to be tn or belong to the set. imagine having to write down every integer between 1 and 100. one would write "2 ~ {l. the set consisting of the numbers 1. one of which is set theory." A ~et is nothing more than a collection of items.7}.hoo~e method.4. and 7 would be written {1.7}. 4. sometimes it is impractical to do so because the list is simply too long. mathematicians would write "4 E {1. For example.4. 'I'he set is usually denoted by enclosing the list of its members (separate" by commas) in braces. the numbers 1." where the symbol "E" stands for the words "is a member of." Similarly. Some time will be devoted to this topic right here and now because it will demonstrate the use of the quantifier "for all.5 quantifiersll: the choo. even if you wanted to.000. When a set has an infinite number of elements (such as the set of real numbers that are greater than or equal to 0) it will actually be impossible to make a complete list. and 7 can be thought of as a collection of items and hence they form a set. to indicate that 2 is not a member of {1. To indicate that the number 4 belongs to this set.4.
" Everything following the ":" is referred to as the de6~n~ng pMpe~ty of the set. otherwise it is not. Consider the example of the set of all real numbers that are greater than or equal to O. The question that one always has to be able to answer is "How do I know if a particular item belongs to the set or not?" To answer such a question.. For the example above. the defining property plays exactly the same role as a definition did: it is used to answer the abstraction question "How can I show that an item belongs to a particular set?" One answer is to check that the item satisfies the defining property. you simply replace x ~verywhere by 2 and see if the defining property is true. for example. One would write S := {real numbers x: x:!: OJ. While discussing sets.CHOOSE ft1ETHOD 41 describe such "large" sets through the use of what is known as 6et bu~tde~ notat~on.: c. when trying to determine if a particular item belongs to the set. If so.2) :!: OJ. you need only check if the item satisfies the defining property. Sometimes part of the defining property appears to the left of the ":" as well as to the right. The only real numbers for which (x 2 + 3x + 2) = 0 are x • 1 and x = 2. where the ":" stands for the words "such that. i f T = {real numbers x :!: 0: (x 2 . observe that it can happen that no item satisfies the defining property. For example. you must be sure to verify this portion of the defining property too. Consider. It involves using a verbal and mathematical description for the members of the set. Euch a set . then it is an element of the set. From a proof theory point of view. Neither of these satisfies the defining property to the left of the ":". The reason is that it does not satisfy the defining property to the left of the":" since 1 is not:!: C.x . then 1 does not belong to T even though it satisfies the defining property to the right of the ":". In this case 3 does belong to S because 3 is. {real numbers x :!: 0: (i + 3x + 2) = OJ. and. to see if the real number 1 belongs to S.
where 1 s x ~ 2 means that 1 s x and x s 2. the sets S = {real numbers x: (x 2 . there are many other instances where the quantifier "for all" can and will be used. from the above example." The second definition answers the abstraction question "How can J show that the two sets Sand T are equal?" by requiring you to show that S is a subset of T and T is a subset of S. To motivate the use of the quantifier "for all. the choose method will enable you to "show that for each element x in E. you can see that all such statements appear to have the same consistent structure. In addition to set theory. 'As you will see shortly. Definition 14. Like any definition. The first one answers the question "How can I show that a set (namely S) is a subset of another set (namely T)?" by requiring you to show that for each element x in S." "for each. each element of S should appear in T and vice versa. x is also in T. x is also in T. but. Using the quantifier "for all. meaning that it has no members. Two sets Sand T are said to be equal (written S = T) if and only if S is a subset of T and T is a subset of S. x is in 7. the statement will have the following basic form (which is similar to the one you saw in the previous chapter) : . 'A set S is said' to be a subset of a set T (written S ~ T) if and only if for each element x in S." a definition can now be made. The special symbol "~" is used to denote the empty set. When the quantifiers "for all. for example. Definition 15." "for every.42 CHOOSE METHOD is said to be empty." observe that it is usually possible to write a set in more than one way.~x + 2) s C} and T z {real numbers x: ) s x s 2}. these can be used to answer an abstraction question. Surely for two sets Sand T to be the same." or "for any" appear.
you could try to show that the something happens. However. it will not be practicable because the list is too long. 1. if you ever come acroSS a statement having the quantifier "for all" in the form discussed above. cos(t) > sin(t)." "something happens. for each one.CHOOSE METHOD 43 For every "object" with a "certain Froperty. and to identify the three components. You have already dealt with this type of obstacle in set theory where the problem was overcome by using the defining property to describe the set. Object: angle t Certain property: none 2 Something happens: sin (t) + COS 2 (t) = 1 Mathematicians often use the symbol "V" to abbreviate the words "for all" ("for each.). Here. the choose method will allow you to circumvent the difficulty. Object: real numbers y Certain property: y > 0 Something happens: 3 a real number x ~ 2. 3 a real number x ~ . to write. then one way in which you might be able to show that the statement is true is to make a list of all of the objects having the certain property." for example. V real numbers y > a. 2." The words in quotation marks depend on the particular statement under consideration. Curing the backward process." etc. When the list is finite. more often than not.= Y Observe that a comma always precedes the something that happens." Practice is needed to become fluent at reading and writing these statements. Then. The use of symbols is illustrated in the next example. . Sometimes the quantifier is "hidden. or even infinite. this might be a reasonable way to proceed. sin 2 (t) + COS 2 (t) = 1. For every angle t. the statement "the cosine of any angle strictly between 0 and ~/4 is larger than the sine of the angle" could be phrased equally well as "for every angle t with r < t < ~/4. Consider these examples.y. and you must learn to read.
In other words. The proof machine for the choose method. ~ll you do know is that the particular object does have the certain property. 6. has the capability of doing so. pretend that someone gave you one of these objects. Then. but remember. then there would be no need to check the whole (possibly infinite) list because you would know that the machine could always do so. The choose method shows you how to design the inner workings of the proof machine. put yourself in the role of the proof machine and keep in mind that you need to have the capability of taking any object with the certain property and concluding that the something happens (see Figure 6). for the chosen object. In Proof machine Cut ~~ Fig.44 CHOOSE METHOD The choose method can be thought of as a proof machine that. with the choose method. To understand the mechanics of the choose method. the something happens. . Then your proof machine will have the capability of repeating the proof for any of the objects having the certain property. As such. and you must somehow be able to use the property to reach the conclusion that the something happens. If you had such a machine. you do not know precisely which one. rather than actually checking that the something happens for each and every object having the certain property. by using the forwardbackward method. you must conclude that. This is most easily accomplished by working forward from the certain property and backward from the something that happens. you choose one object that has the certain property.
(x 2 1  3x + 2) 2}. using the definition leads to the answer that you must show that. for x'.3x' + 2 ~ 0.e. This new statement Bl clearly has the form described above. Consider the following example. ] 1: X 1: 2. that T is a subset of S.. using the fact that x is in S (i.Definition 15 can provide the answer that you must show that S is a subset of T and T is a subset of S. in some proof. in S. that.3x' + 2 !!S 0).This distinction is often ignored (i.. If S . thus indicating that you should proceed by the choose method. the something happens. by working forward from the fact that (X. the same symbol is used for both the general object and the chosen one).has been used to distinguish the chosen object from the general object. In this case that means you should choose an element. so first try to establish that S is a subset of T. The forwardbackward . that is. x.. Here. To show that S is a subset of T. and you must be careful to interpret the symbol correctly.. for all x in S.e. (x') . for all real numbers x with x . that 2 does have the certain property (in this case.. {real numbers x: T . and. say x'. {real numbers x: then S .3x + 2 !!S 0. 1 :s x' 1: 2. and afterward. x is in T. for example.CHOOSE METHOD 45 Suppose. and 1: X 1: method gives rise to the abstraction question How can I show that two sets (namely Sand T) are equal?. To do so. you obtain the abstraction question How can I show that a set (namely S) is a subset of another set (namely T)?Again. With the choose method you would choose one of these real numbers. you must choose an object having the certain property and show that the something happens. you must reach the conclusion that. you n:eded to show that. 1: O}.)2 . that it satisfies the OUtline of proof. the symbol x'. Exaaple 5.. say x. T. Then.
or else (x . The statement x is in T. either (x . during the backward process. but this is precisely the last statement obtained in the backward process. note the double use of the symbol x.1) ~ O. because you have assumed that A is true. Proof of Example 5. you can make use of the information in A to show that 1 s x s 2. ..Now is the time to use the fact that x is in Sspecifically. Turning now to the forward process. ex . in the forward process. you made use of the choose method.e. This part will be left as an exercise.1) ~ 0. you will use the extra information. which is impossible. that 1 s x s 2). when you use the choose method. x would be ~ 2 and x would be ~ 1. Do not forget that you still have to show that T is a subset of S in order to complete the proof that S = T.2) and (x .for both the general and the chosen object. Also.3x + 2) ~ O. Recall that.2) ex .together with the answer that you must show that x satisfies the defining property of T (i. you obtain additional information that is added to the assumption that A is true. one has Then. since x is in S. say 3/2.now becomes B2 and you should apply the abstraction process to B2. Thus the second condition must happen (i. one obtains (x 2 . from the defining property of the set S. The first situation can never happen because if it did. x ~ 2 and x ~ 1). by factoring. obtaining the abstraction question How can I show that an item (namely x) belongs to a set (namely T)?.46 CHOOSE METHOD defining property of S) together with the information in A. In other words. to which your efforts must be directed.. The only way that the product of the two numbers (x . there is additional information available to you. Observe that. However.1) s O. you must show that x is in T. and hence it has successfully been shown that S is a subset of T. To show that S = T it will be shown that Sis a subset of T and Tis a subset of S. Note that you do not want to pick one specific element in S. Invariably.2) ~ 0 and ex .e.1) can be ~ 0 is for one of them to be ~ 0 and the other ~ O. at which time you chose x to be an element in S.2) ~ 0 and (x .
2) s 0 and (x . the certain property. The proof that T is a subset of S is omitted.1) s 0. Then g is ~ f on the set S of real numbers if. (a) The real number x· is a maximum of the function f if. The former canno t happen because i f it did. for all x in S. let x be in S (the use of the word "let" frequently indicates that the choose method has been invoked).1) s O. Add this information to that in A and attempt to show that the something happens by using the forwardbackward method.1. f (x) s f(x·).2) ~ 0 and (x . for every real number x. Hence it must be that x s 2 and x ~ 1. (c) The real number u is an upper bound for a set S of real numbers if. (b) Suppose that f and g are functions of one variable. Consequently.1) ~ O. 5.11 The choose method is a viable approach for dealing with a statement that contains the quantifier "for all. or else. 3 xES ~ x > u  t. (x . This means that either (x .2) (x . which means that x is in T. x ~ 2 and x s 1.3x + 2) s 0. of proof identify the objects. For each of the following definitions. Exercises Note: All proofs should contain an outline as well as a condensed version. for every element x in S. and the something that happens. x s u." Proceed by choosing an object having the certain property. . (d) The real number u is a least upper bound for a set S of real numbers if u is an upper bound for S and V real numbers t > 0. and so one has (x .CHOOSE MFTHOD 47 To see that S is a subset of T. (x 2 . g(x) ~ f(x).
... there is a real number 8 > 0 such that.y I < 8. ~ wherever necessary.3. for all real numbers x and y and for all real numbers 0 s t s ].2.. Use a different symbol to distinguish the chosen object from the general object. to apply the choose method in Exercise 5. for which it must then be shown that f(x') s f(x·).48 CHOOSE METHOD for every element x and y in C and for every real number t between 0 and 1. (b) If t is an angle then it follows that sin(2t) = 2sin(t)cos(t).l(a).]la). say x'.t)y is an element of C. for all real numbers y with I x . one would say "Let x I be a real number.]. Thus.f (y) I < &. Ixk . 3.. It wi 11 be shown that f (x ') s f (x·) ..t)y) s tf(x) + (1 . For each of the parts in Exercise 5." Discuss why the approach of the choose method is the same as that of using the forwardbackward method to show that "If x is an object with the certain property. then the something happens. it follows that f(tx + n . Ie) The set C of real numbers is convex if. describe how the choose method would be aFplied.t)f(y). are real (h) Suppose numbers. something happens." 5. x 2 . If (x) . The sequence x'. (a) Some mountain is taller than every other mountain. x1. one would choose a real number.4. that x. (Hint: Make use . 3 an integer k' ~ V integers k > k'. for every real number & > 0. (f) The function f of one variable is a convex function if. tx + (1 . for Exercise 5. (g) The function f of one variable is continuous at the point x if.xl < &. Reword the following statements using the appropriate symbols V. V real numbers & > 0. x2. to show that x· is the maximum of the fUnction f." How are the t~~ statements in quotation marks related? 5. For instance. Consider the problem of showing that "For every object x with a certain property. 5. converges to x if.
such that x = 2y/(1 + y). prove that 5. For the sets S anCl T of Fxamrle T S. 3 a real number 8 > 0 ~ V real numbers x anCl y wi th Ix .8. Ixl ~ ~. and f is a function defined by f(x) = mx + b .fey) I < & (where f is a function of one variable) • 5. 5.5. explain how the technique would be applied to the particular problem. etc. 'Y. For each of the followin9 statements. th~re is a real number y < r. (c) V real numbers & > 0.1(f). indicate which proof techniques you woulc use (choose and/or construction) . then there is a r~tional number r such that x < r < y. that is. (b) For all real numbers M > 0. Prove that if m and b are real numbers. what would you choose. (Hint: Use the definition in Exercise 5.7. there is an element x in the set S of real numbers such that Ixl > M. 5.6. what would you construct. 5. then f is convex. Prove that for every real number x > 2. Also. (d) If x and yare real numbers such that x < y.yl < 8. for all elements x in the set S of real numbers.) (c) The square root of the product of any two nonne9ative real numbers p and q is always ~ their sum divided by 2. (a) There is a real number ~ > 0 such that.) . If(x) .3nd in which order. s.CHOOSE METHOD 49 of Fxercise 5.
6 quantifierslll: induction
In the previous chapter you learned how to use the choose method when the quantifier "for all" appears in the statement P. There is one very spec ial form of B for wh ich a separate proof techn i que known as mathe.mati.c.af. btduc.ti.o tt has been developed. Induction should seriously be considered (even before the choose method) when E has the form: For every integer n
2:
1, "something happens"
where the something that happens is some statement that depends on the integer n. .n.n example would be the statement: n For all integers n 2: J, k = n{n+l)/2, k=l
L
where
k=l When considering induction, the key words to look for are "integer" and "2:1." One way to attempt proving such statements would be to make an infinite list of problems, one for each of the integers starting from n = 1, and then prove each statement separately. While the first few prohlems on the list are usually easy to verify, the issue is how to check the nth one and beyond. For the example above, the list would be:
1
P (1)
L
n
k = 1+ ••• +n.
Lk=1(l+1)/2 k=l
or
1
=1
50
INDUCTION 2 Lk k=l
51
P(2)
2(2+1)/2
or
1+2
=3
P(3)
3 Lk = 3 (3+1)/2 k=l
or
1+2+3 = 6
pen) n+1 Lk k=l
n(n+l)/2
P(n+1):
(n+lH(n+l) + 11/2
(n+1) (n+2)/2
Induction is a clever method for proving that each of these statements in the infinite list is true. As wi th the choose method, induction can be thought of as an automatic problemsolving machine that starts with P(l) and works its way progressively down the list proving each statement as it proceeds. Here is how it works. You start the machine by verifying that P(l) is true, as can easily be done for the example above. '!"hen you feed P(l) into the machine. It uses the fact that P(l) is true and automatically proves that P(2) is true. You then take P(2) and put it into the machine. Cnce again, it uses the fact that P(2) is true to reach the conclusion that P(3) is true, and so on (see Figure 7). Observe that, by the time the machine is going to prove that P(n+l) is true, it will already have shown that Pen) is true (from the {::revious step). Thus, in designing the machine, you can assume that PCn) is true, and your job is to make sure that P(n+l) will also be true. Co not forget that, in order to start the whole rrocess, you must also verify that pel) is true.
52
INDUCTION
Verify that PO) is true
I
I
Tn
Proof machine Out
Fig. 7. The proof machine for induction.
INDUCTION
53
To repeat, a proof by induction consists of two steps. The first step is to verify that the statement pel) is true. To do so, you simply replace n everywhere by 1. Usually, to verify that the resulting statement is true, you will only have to do some minor rewriting. The second step is much more challenging. It requires that you reach the conclusion that P(n+l) is true by using the assumption that Pen) is true. There is a very standard way of doing this. Begin by writing down the statement P(n+l). Since you are allowed to assume that Pen) is true and you want to conclude that Pfn+l) is true, you should somehow try to rewrite the statement P(n+l) in terms of P(n)as will be illustrated in a momentfor then you will be able to make use of the assumption that Pen) is true. On establishing that rCn+l) is true, the proof will be com pI ete. Example 6. For every integer n
l!:
1,
k=l
E k = n(n+l)/2
n
Outline of proof. When you are using the method of induction, it is helpful to write down the statement PCn), in this case:
PIn):
k=l
Ek
n
n(n+l)/2
The first step in a proof by induction is to verify PCl). ~eplacing n everywhere by in P(n), you obta in
P (1 ):
~ith
k=l
E k = 1 (1 +1) /2
it is easy to
1
a small amount of rewriting, verify this statement since
k=l
Ek
1
1
= 1 f1 +1) /2
begin with the left side of the equality in P(n+l) and try to make it look like the right side." The second step is more invoJved. and rewriting a bit. obtaining n+l E k = rn(n+l)/2] + (n+1) k=l All that remains is a bit of algebra to rewrite [n(n+l)/2 + (n+l)] as [(n+l) (n+2)/2].54 INDUCTION This step is often so easy that it is virtually omitted in the condensed proof by simply saying "The statement is clearly true for n = 1. n+l Ek k=l n == (Ek) + (n+l) k=l true by Now you can use the assumption that pen) is replacing with n(n+l)/2. The algebraic steps are: . In so doing. In this case n+l P (n+l): E k k=l = (n+l) [(n+1)+l]/2 == (n+l) (n+2)/2 To reach the conclusion that P(n+l) is true. you should use the information in Pen) by relating the left side of the equality in P(n+l) to the left side of the equality in P(n). The best way to proceed is to write down the statement P(n+l) by carefully replacing n with (n+l) everywhere in P(n). thus obtaining the right side of the equality in P(n+l). In this example. i f necessary. You must use the assumption that r (n) is true to reach the conclusion that P(n+l) is true. for then you will be able to use the right side of the equality in P(n).
that first value would be n = 5. For instance. If you are unable to relate P(n+l) to PIn). n 2 > n 2" can also be proved by induction.// When using the method of induction. Proof of Example 6. The statement is clearly true for n :: 1. it is not necessary that the first value for n has to be 1. Assume that it is true for n (i.. you must verify PIn) for the first possible value of n. The only modification is that. .INDUCTION n(n+1)/2 + (n+1) :: (n + n)/2 + 2(n+1)/2 2 + 3n + 2)/2 :: (n :: (n+1) (n+2)/2 In summary.e. that l+ ••• +n :: n(n+l)/2). 1+ ••• +(n+1) 2 55 :: (1+ ••• +n) + (n+l) :: n(n+l)/2 + (n+l) = (n 2 + 3n + 2)/2 :: (n+1) (n+2)/2 Your ability to relate P(n+l) to pen) so as to use the induction hypothesis that pen) is true will determine the success of the proof by induction. so you will have to check that 2 5 > 52 (which. then you might wish to consider a different proof technique. Then 1+ ••• +(n+l):: 11+ ••• +n) + (n+l) :: n(n+l)/2 + 2(n+l)/2 :: (n 2 + 3n + 2)/2 = (n+l) (n+2)/2 which is P(n+l). in order to start the proof machine. the statement "for all integers n ~ 5. In this case.
5 of course. then P (n+l) is al so true (I. when trying to show that PCn+1) is true. Exercises Note: Proofs in this chapter outline of proof. you can also use the i f necessary. it is second step 0 f same. but do not forget to verify that the statement is also true for the very first possible value of n.. in fact.Pfj). but can you assume that prj) is. it is important to realize that induction does not help you to discover the correct form of the statement P Cn).e. ••• .e. where j < n. you can assume that PCn) and all of the preceding statements are true. Such a proof is referred to as gene~allzed lnductlon. 2 n > 2n+1 > (n+l)2). it wi 11 already have established that all of the statements P(1).58 INDUCTION since 2 = 32 while 52 = 25). that you can relate P(n+l) to p.Pln) are true (look again at Figure 7). need not contain an . recall the anology of the proof machine. true? The answer is yes! To see why. however. Induction only ver ifies that a given statement P(n) is true for all integers n ~ some initial one. In so doing.(j). The key to its success rests in your ability to relate PCn+l) to PCn) or to some previous statement. ••• . and observe that. Another modification to the basic induction method arises when you are having difficulty relating P(n+l) to P(n). Suppose. The the induction proof would remain the still have to show that i f P (n) is n 2). Thus. You would true (i. however. fact that n ~ 5.. Induction is a very powerful technique when applicable. you would like to use the fact that Plj) is true. by the time the machine has to show that P (n+l) is true. In this case.
( c) For every posi tive odd integer.1. For which of the following statements woulC' induction be directly applicable? When it is not applicable.8. . (b) There is an integer n ~ 0 such that 2n > n 2 • (c) For every integer n ~ 1. and (2) if the statement is true for (nl). for any 6. 2 n > n 2 • that. l(1!)+ ••• +n(nl) = (n+l)! . that a set of n ~ 1 n elements has 7 subsets (including the empty set) • 6. t'escribe a "mod i fied" induction procedure that could be used to prove statements of the form: (a) For every integer s some initial one. something happens." "something happens?" 6. that.4. 6. Prove.7. (a) For every positive integer n. by induction. for every integer integer 6. n 2 ~ n. Prove.2. Prove that for every integer n ~ 1. that. without using induction. explain why. then it is also true for n. ) (d) For every integer n ~ 4.5. 6. by showing that (1) the statement is true for n = 1. P civides 5n + 2·3 n . Prove. l+ ••• +n = n(n+l)/2. (Recall that n! = n (nl ) ••• 1. E civides n 3 .1. something happens. Prove. 6.n. integer n ~ 1.1 + 1.3. by induction. by induction.INDUCTION 57 6. n ~ 5. (a) Why and when would you want to use induction instead of the choose method? (b) Wny is it not possible to use induction on statements of the form: For every "object" with a "certain property. for every n ~ 1. n! > n 2 • (e) For every real number n ~ 1. ( b) For every integer.6. something happens.1. it follows that 1(11)+ ••• +n(n!) = (n+l)! .
one horse has the same color.11 . since all of the colored horses in this group are brown. the statement is clearly true. Let n be the number of horses. any other group of n of the (n+l) horses that contains the uncolored horse. ~gain.58 JNCUCTION all 6. whatever color it is. the induction hypothesis states that they all have the same color. Assume that any group of n horses has the same color. therefore. ~ow consider a group of (n+l) horses. Consider. the uncolored horse must also be brown. The only issue is the color of the remaining "uncolored" horse. When n = 1. Then. by the induction hypothesis.9. that is. all of the horses in the new group must have the same color. say brown. What is wrong with the following proof that horses have the same color? Proof. Taking any n of them.
one typical method emerges and it is referred to as 6pecialization As a result of assuming A is true.7 quantifiersIV: specialization In the previous three chapters you discovered how to proceed when a quantifier appeared in the statement B." "something happens. you are assuming A is true. for all objects with the certain property. In doing the proof. and in this case that means you can assume that indeed there is an object with the certain property such that the something happens. something happens. When the statement A contains the quantifier "there is" in the standard form: There is an "object" with a "certain property" such that "something happens" you can use this information in a straightforward way. When showing that "A implies P" by the forwardbackward method. you would say: "Let x be an object with the certain property and for which the something happens • • • • " The existence of this object will somehow be used in the forward process to obtain the conclusion that B is true. at some point in the backward process. The more interesting situation occurs when the statement ~ contains the quantifier "for all" in the standard form: For all "objects" with a "certain property. you were to come across one of these objects that ~oes have 59 . you know that. If. This chapter will develop a method for exploiting quantifiers that appear in the statement A." To use this information.
the appearance of the quantifier "for every" in the forward process suggests using specialization. If R is a subset of a set S of real numbers and u is an upper bound for S. and by Definition ]4. in particular. ~galn. for every element s in S. Definition 16. From the hypothesis that R is a subset of S. you will have specialized the statement A to one particular object having the certain property. In other words. and that should help you to conclude that B is true. In . if you know that. sin 2 (t) + cos 2 (t) = 1. and hence you can use specialization to conclude that r is in S. for this particular object. for every angle t. r s u. Example 7. for all elements r in R. you know that each element in R is also in S. in R for which it must be shown that r s u. say t ~ ~/4. you will see how specialization is used to obtain the desired conclusion that r s u. for one angle. then u is an upper bound for R. The appearance of the quantifier "for all" in the backward process suggests proceening with the choose method. A real number u is an upper bound for a set of real numbers! if for all elements t in T. Turning now to the forward process. s s u. By Definition IE this means that. say r. you can conclude that sin 2 (~/4) + coS2(~/~) = 1. ~lso. ~n example demonstrates the proper use of specialization. whereby one chooses an element. In the backward process you came across theparticular element r in R. t must be shown that. then you can use the information in A by being able to conclude that.60 SPECIALIZATION the certain property. t s u. you know that u is an upper bound for S. For instance. Thus. Outline of proof. The forwardbackward method gives rise to the abstraction question "How can I show that a real number (namely u) is an upper bound for a set of real numbers (namely R)?" Definition 16 is used to answer the question. the something does indeed happen. then. from the hypothesis. i.
Also. every element in S is s u. R is a subset of S and so r is also an element of S (here is where specialization has been used). When 8 contains the quantifier "there is. by hypothesis. In the latter case. let the form of the statement guide you. provided that you can relate the statement for (n+l) to the one for n. you must be very careful to keep your notation and symbols in order. r is an element of S. proof is now complete. 8y hypothesis. Furthermore. so r s u (again specialization has been used) . The choose method is associated with the quantifier "for all." except when the statement e is supposed to be true for every integer starting from some initial one. u is an upper bound for S. Since the statement "r s u" the last one obtained in the backward process. In the condensed proof that follows. you conclude that r s u. note the lack of reference to the forwardbackward. 81 the can was the Note that. induction is 1 ikely to be successful. Proof of Example 7. for only then will the something happen. choose. Finally.11 This and the previous three chapters have provided various techniques for dealing with quantifiers that can appear in either A or 8. thus. if the quantifier "for all" appears in the statement A. as was shown in previous paragraph. specialization can often be exploited. for only then c~n you conclude that the something happens. by specialization. . So. As always. let r be an element of R (the word "let" indicates th~t the choose method has been used). and specialization methods. be sure that the particular object under consideration does satisfy the certain property. when using special ization. To show that u is an upper bound for P. be sure that the particular object to which you are specializing does satisfy the certain property. r is an element of S. In particular. When using specialization.SPECIALIZATION particular." the construction method can be used to produce the desired object.
and. the of choose proof and 7.2. then R is a subset of T. what can you conclude about the object? 2 (a) Statement: Y integers n ~ 5. Explain the difference between the specialization methods. given that it does satisfy the properties.f (y) I < e (where 8. x is in the set T. and f is a function of one variable)." Exercises Note: All proofs should contain an outline as well as a condensed version. 7. and yare real numbers. For each of the following statements and given objects.1. Given object: y (c) Statement: Y e > 0. what properties must the object satisfy in order to be able to apply specialization. I f (x) . Prove that if R is a subset of Sand subset of T. e. 2n > n • Given object: m (b) Statement: For every element x in the set S with Ixl < 5. Given object: Angle S of the triangle RST 7. :I 8 > 0 ~ Y Y with I x . Given object: e' (d) Statement: Any rectangle whose area is onehalf the square of the length of a diagonal is a square.82 SPECIALIZATION All of the material thus far has been organized around the forwardbackward method.y I < 8. S is a . Now it is time to see some other techniques for showing that "A implies e. x. Given object: the quadrilateral eRST (e) Statement: For any angle t with 0 < t < x/4. cos(t) > sin(t).3.
Prove that 1 is a least upper bound of the set S = {1 . 1 .l(f».l/n}. the function sf is convex (where the value of the fUnction sf at any point x is sf(x».6.4.1/4. • • • } (see Exercise S. Prove that if f is a convex function of one variable (see Exercise 5. (Hint: The set S can be written as {real numbers x: there is an integer n ~ 2 such that x = I . Prove that if f is a convex function of one variable (see Exercise S.5. then S intersect T is a convex set.l(d». 1 . 7. then for all real numbers s ~ fI.1(f» and y is a real number. then the set C = {real numbers x: f (x) s y} is convex. 7.) .] (e».7. 7.1/3. Prove that if Sand 63 T are convex sets (see Exercise 5.SPECTALTZATI0N 7.1/2.
there is an integer. The forwardbackward method gives rise to the abstraction question "How can I show that an integer (namely n) is even?" One answer is to show that there is an integer k sllch that n 2k.ontlla. then is even.11 Fortunately. If n is an integer and n 2 is even. In this chapter. and it is left as an exercise. Example 8. The appearance of the quantifier "there is" suggests proceening with the construction method.tioYl method is described 64 . Since the objective is to produce an integer k for which n = 2k. the c. :0: Working forward from the hypothesis that 2n2 is even.8 the contradiction method As powerful as the forwardbackward method is. it is natural to take the sauare root of both sides of the equation n 2 = 2m to obtain n f2m. and so the forward process will be used in an attempt to produce the desired integer k. such that n 2m.. you may well find yourself unable to complete a proof for one reason or another. say M. n Outline of proof.dic. but how can you rewrite f2m to look like 2k? It would seem that the forwarabackwarry method has fa il ed! :0: Proof of Example 8. there are several other techniques that you might want to try before you give up. The technique that you are about to learn will lead to a simple proof of this problem. as is shown in the next example.
So what does it mean to "see why this cannot happen?" Suppose. then there must be some reason why B cannot be false. to reach the desired conclusion that B is true. you rule out this one unfavorable case by actually assuming that it does happen. the idea of a proof by contradiction is to assume that A is true and B is false. somehow. you proceed by asking yourself a simple question: "Why can't B be false?" After all. Exactly how do you use the assumption that A is true and B is false to reach the contradiction? 3. persistence. What contradiction should you be looking for? 2. With the contradiction method. for example. Another way of viewing the contradiction method is to recall that the statement "A implies B" is true in all cases except when A is true and B is false. you must use this information to reach a contradiction to something that you absolutely know to be true. just as you would in the forwardbackward method. by far. In other words. . The objective of the contradiction method is to discover that reason.CONTRAtICTION METHOD 65 together with an indication of how and when it should be used. you were somehow able to reach the conclusion that 0 = I!?! Would that not convince you that it is impossible for A to be true and B to be false simultaneously? Thus. insight. and. and it usually takes creativity. the hardest to answer because there are no specific guidelines. Why and when should you use this approach instead of the forwardbackward method? The first question is. Each problem gives rise to its own contradiction. that as a result of assuming that A is true and e is false (hereafter written as NOT B). In a proof by contradiction. and luck to produce a contradiction. and see why this cannot happen. At this point. several arise: very natural questions 1. you assume that A is true and that NOT B is true. if B is supposed to be true. in a proof by contradiction. you begin by assuming that A is true. However. and then reaching a contradiction.
Here. Recall the statement B associated with Example 8: "n is an even integer." Obviously an integer can only be odd or even. when the statement e is one of two possible alternatives (as in Example C). On the other hand. 8. There are at least two recognizable instances when the contradiction method should be considered. As to the second question. ForwardBackward method versus contradiction method. you can assume that both A and NOT B are true. The discussion above also indicates why you might wish to use contradiction instead of the forwardbackward method. As a general rule. With the forwardbackward method you only assume that A is true. one common approach to finding a contradiction is to work forward from the assumption that A and NOT B are true. that n is not an even integer). the contradiction method is likely to be effective because. then it must be the case that n is an odd integer. use contradiction when the statement NOT B gives you some useful information. you have no definite knowledge of where the contradiction will arise. When you assume that B is not true (i..e.66 CONTRADICTION METHOD Assume forward A}> Conclude backward Method ForwardBackward Contradiction NOT <B * (contradiction) At forward BJ> • • • Fig. as will be illustrated in a moment. the statement NOT B has given you some useful information. In general. you get two statements from which to reason forward instead of just one (see Figure 8). Thus. while in the contradiction method. by assuming NOT S. you will know that the .
. whereby you can assume that A ancl NOT B are both true. But. Working forward and using Definition 7 for a rational number. Working forward. there are integers p and q with q # 0 such that r = p/q. Outline of proof. so 2 = p2/q2. and since p2 = 2q2. It is important to note that the conclusion of Example 9 can be rewritten so as to read "r is not rational. you 67 to A second instance when the contradiction method is likely to be successful is when the statement B contains the word "not. it follows that r2 = p2/q2. from the hypothesis. From 2q2 = p2. such that r2 = 2. A crucial observation here will really help. Now a contradiction can be reached by showing that 2 is a common divisor of p and q! This will be done by showing that p and q are even. In this case. a contradiction must now be reached. you can certainly say that 2q2 is even no matter what kind of integer q is. Example 9. If r is a real number then r is irrational. since r = p/q. . It is possible to assume that p and q have no common divisor (i. p2 must also be even. The rest of the forward process is mostly rewriting 2 = p2/q2 via algebraic manipulations to reach the desired conclusion that both p and q are even integers.CONTRACICTION METHOD other case must happen. that means you can assume that r2 = 2 and that r is a rational number. and hence 2 divides them both. ~pecificallY'2upon multiplying both sides of the last equation by q you obtain 2q2 = p2. Using this information. and this takes a lot of creativity.e." as is shown in the next example. There is still the unanswered question of where the contradiction arises. for if they did. you also know that r2 = 2. no integer that divides both p and q). and that should help reach a contradiction. the appearance of the word "not" now suggests using the contradiction method. you could divide this integer out of both the numerator p and the denominator q." and as such.
q2 = 2k 2 • From this it then follows that q2. it follows that 7 = P2Iq. Noting that 2q2 is even. Try to prove the statement by some other method! . recall ing 2 that 2q2 = p2. This contradiction establishes the claim. In this case it would be necessary to provide more details. Thus. Meanwhile. look at Example 8.2 Since r2 = 2 and r = p/q. once again. Furthermore. and this last sentence might not convince you that p is even. that a proof is supposed to be a convincing argument. and since ql = 2k2. there is an integer k such that p = 2k. Consequently. must be even. Thus it has been shown that both p and q are even and have the common divisor 2. p2. or equivalently. it follows that 2q2 = C2kt = 4k. 2k is even no matter what type of integer k is. by the alternative definition of an even integer. it can be assumed that p and q have no common divisor for. a proof should be written with the audience in mind. must be even. ". and hence p. q2 is even. Continuing with the forward process. it is still necessary to show that q is even too. Assume. to the contrary. If you need further convincing that p is even. and it epitomizes the use of contradiction. since p is even. Now. both p and q are even and the desired contradiction has been reached.88 CONTRADICTION METHOD Continuing with the forward process. On substituting this value for p.s always. this number could be canceled from both the numerator p and the denominator q. if they did. what useful information can be derived from the fact that p2 is even? Well. that r is a rational number of the form p/q (where p and q are integers with q F 0) and that r2 = 2. the only way for the integer p times itself to be even is for p to be even. Proof of Example 9. or equivalently. note that the only way for an integer q times itself to be even is for q to be even (see Example 8). there is an integer k such that p = 2k. one obtains 2q2 = p2 = f2k)2 = 4k2. 2q2 = p2. which says that q2 = 2k 2 • Cnce again. and this was the desired conclusion. and hence q. Finally. Recall.11 This proof was discovered in ancient times by a follower of Pythagoras.
Imagine that you wish to show that there are at least two people in the world who have exactly the same number of hairs on their heads." the construction method is recommended in spite of the difficulty of actually producing the desired object. perhaps with the aid of a computer. and so on. Ho~ and where the contradiction arises is not at all clear. Consider the following example. when the statement B contains the quantifier "there is.CONTRAnrCTION ~ETHOD 69 There are several other valuable uses for the contradiction method. assign numbers to the people in such a way that the person with the fewest hairs receives number 1. Cn the other hand. contradiction can be used. or equivalently. that everyone has a different number of hairs on their heads. and so a contradiction has been established. If the construction method is used. Instead of showing that there is an object with the certain property such that the something happens. or at least indicated how it might be produced. Now. why not proceed from the assumption that there is no such object? Now your job is to use this information to reach some kind of contradiction. but it may be a lot easier than producing or constructing the object. and so on. the person whose number is 2 must have at least one more hair than the person whose number is ]. Consequently. the person with the next fewest hairs receives number 2. then you will have produced the desired object. the person whose number is one billion must have at least one billion more hairs than the person whose number is ]! Clearly no person can have a billion more hairs than someone else. This example illustrates a subtle but very significant difference between a proof using the construction method and one that uses contradiction. Recall that each person is assumed to have a different number of hairs. you can assume that no two people have the same number of hairs on their heans. Recall that. The contradiction method opens up a whole new approach. then you would actually have to go out and find two such people. In so doing. Thus. if you . To save you the time and trouble. If the construction method is successful.
As such. then 24 does not nivide R. Exercises Note: All rroofs should contain an outline as well as a condensed version. You work forward from the assumption that A and NOT E are true to reach a contradiction. (b) If the matrix ~ is not singular then the rows of r are not linearly deFendent. ~oreover. then you will know that the object exists but will have no way of physically constructing it.1. and n are 3 consecutive integers. One of the disadvantages of the method is that you do not know exactly what the contradiction is going to be. an active area of current research consists of finding constructive proofs where previously only proofs by contradiction were known. For this reason.70 CONTRADICTION METHOD establish the same result by contradiction. you will have a "guiding light" since you will know what contradiction you are looking for. (c) If f and g are two functions such that (1) g !:!: f and (2) f is unbounded above. . As you have seen. When applying the contradiction of proof method to the following propositions. the contradiction method can be a very useful technique when the statement B contains the word "not" in it. it is often the case that proofs done by contradiction are quite a bit shorter and easier than those done by construction because you do not have to create the desired object.2 + m2 + n2 + 1. 8. what should you assume? (a) If R" m. You only have to show that its nonexistence is impossible! This difference has led to some great philosophical debates in mathematics. then g is unbounden above. The next chapter describes another proof technique in which you attempt to reach a very specific contradiction.
state what you would assume and what you would try to conclude.8. 'f/ x inS. Prove. Prove. (b) 'f/ s in S. Prove. Th~n consider any prime number p that divides n! + 1. (e) The real number x is < 5. 8. by contradiction. 8.6. I x I < f.5. by contradiction.2. by contradiction. ~ t in T such that s > t. that if chord of a ~1 and t2 are two lines in a plane that are both perpendicular to a third 1 ine ~ in the plane. the 8. that. by contradiction. Pow is p related to n?) 8. Reword each of the following statements so 71 that the (a) (b) (c) word "not" appears explicitly.3. then n is even. by contradiction. Specifically.4. ( c) j M > 0 s uc h t hat. There are an infinite number of primes. The only positive integers that divide positive integer pare J and p.7. (Note: Throughout. and all of the variables refer to real numbers. Sand T are sets of real numbers.9. ) (a) 3 s e s ~ seT. that there do not exist 8. that no circle is longer than a rliameter. three consecutive positive integers such that the cube of the largest is equal to the sum of the cubes of the remaining two. indicate which proof techniques you would use. and in which order. that if n is an integer and n2 is even. Prove. by contradiction. then t1 and t2 are parallel. at a party of n ( ::!: 2) people. there are at least two people who have the same number of frienrls at the party. (Hint: Assume that n is the largest prime. 8. that there are an infinite number of primes. Prove. (d) The I ines ~ and t' are parallel.CCNTRAVICTION METHOD 8. . The set S of real numbers is unbounded. For each of the following statements. Prove. 8.l.
the contrapositive method can be thought of as a more "passive" form of contradiction in the sense that the assumFtion that ~ is true passively provides the contradiction. you assume that ~ and NOT B are true and you work forward from the statement NOT B to reach the contradiction that ~ is false. ran you ask for a better contradiction than that? How can ~ be true and false at the same time? To repeat. the ~ont~apo~itive method has the a~vantage of directing you toward one specific type of contradiction. and NOT B are true. Tn the contradiction method however. In general. From Figure 9 you can also see the advantages and disadvantages of the contrapositive method over the contradiction method. you work forward only froM t'OT P. The contrapositive method is similar to contrad iction. the difficulty with this method is that YOli do not know what the contrad iction is going to be. Instead. the ~ssumption that A is true is actively used to reach a contradiction (see Figure 9). in that you beg in by assuming that ". in the contrapositive method. Unlike contradiction however. you do not work forward from both A and NOT B. ~s such.9 the contrapositive method The previous chapter described the contradiction method in which you work forward from the two statements ~ and NCT B to reach some kind of contradiction. Your objective is to reach the contradiction that A is false (hereafter written NOT A). The ~isadvantage of the 72 . As will be seen in this chapter.
Here. . Outline of proof. The next example demonstrates the contrapositive method. Method Con trapositive Assume Conclude forward NOT B}> backward <NOT A Contradiction NOT AB}:~~~:~~> ••• * (contradiction) Fig. 9.CONTRAPOSJTJVE METHOD 73 contrapositive method is that you work forward from only one statement (namely NOT B) instead of two. NOT B is the statement "p = q" while NOT A is "{Pq = (p+q)/7. then p is not equal to q. In this case. Example 10. the advantage is that you know precisely what you are looking for (namely NOT A). On the other hand. The appearance of the word "not" in the conclusion should suggest usin9 either the contrad iction or contraposi tive method. whereby you will work forward from the statement NOT B and backward from the statement NOT A. The option of working backward is not available in the contradiction method because you do not know what contradiction you are looking for." The objective is to work forward from the fact that p • q to reach the desired conclusion that ypq = (p+q)/2. the contrapositive method will be use~. If P and q are positive real numbers such that ypq is not equal to (p+q)/2. Try to do this for yourself. you can often apply the abstraction process to the statement NOT A in an attempt to work backward. Contrapositive method versus contradiction method. Because of this.
working forward from the assumption that p = q. Specifically. The forwardbackward method arose from conSidering what happens to the truth of "A implies 8" when A is true and when A is false (recall Table 1). and also (p+q) 12 = . If you understand Figuce 1('. The contrapositive method arises from similar considerations regarding B. ypq :: = p. you would have to show that A . Since p is positive. In the forwardbackward method you work forward from A and backward from B. Assume. if B is true. why not replace q everywhere by p? In other words. as would be done in the forwardbackward method. according to Table 1. So suppose B is false. that p :: q. there is no need to consider the case when B is true. then it is not hard to See why the contrapositive method might be better than the forwardbackward method. the statement "A implies E" is true. Proof of Example 10. Hence. rn In the following condensed version of the proof. while in the contrapositive method.p+p) 12 = p. Perhaps you can obtain more useful information by working NOT B forward rather than A. it is easy to See that ypq = (p+q) 12 :: p. Note that the assumption that p > 0 was used in claiMing that y"pP":: p. to the contrary.74 CONTRAPOSITIVE METHOV Since you want to conclude that ypq:: (p+q)/2 and you are assuming that p :: q. it then follows that (p'q = fPP = p = (p+p) 12 = fp+q) 12 and the proof is complete. it might be easier to perform the abstraction process on the statement NOT A rather than on B. Thus. In order to ensure that "A implies B" is true. you work forward from NOT B and backward from NOT A (see Figure 10).11 It is also quite interesting to compare the contrapositive and forwardbackward methods. note that nowhere in the proof does it formally say that the statement NOT A has been reached. since p:: q. then according to Table 1. Also.
In the contradiction method. but you do so by working forward from NOT B to reach the conclusion NOT A. Indeed. You should also. the statement "A implies B" is logically equivalent to "NOT B implies NOT A" (see Table 3). A . In general.CONTRAPOSITIVE METHOD Method ForwardBackward Contrapositive Assume forward A}> 75 Conclude backward (B backward (NOT A A forward NOT B}> Fig. you work forward from the two statements ~ and NOT B to obtain a contradiction. of course. contradiction. In the contrapositive method you also reach a contradiction. the contrapositive method has you aSSume that B is false and try to conclude that A is false. for then you will usually find that the statement NOT B has some useful information." ~ost condensed proofs that use the contrapositive method make little or no r~ference to the contradiction method. This occurs when the statement B contains the word "not" in it. As such. However. it is difficult to know whether the forwardbackward. ~us. there is one instance that often indicates that the contradiction or contrapositive method should be chosen. is false. the contrapositive method can be thought of as the forwardbackward method applied to the statement "NOT B implies NOT A. ForwardBackward method versus contrapositive method. 10. work backward from ~OT A. or at least considered seriously. or contrapositive method will be more effective for a given problem without trying each one.
The next chapter shows you how to do so when the statements contain quantifiers. and contrapositive methods is given in Figure 11. 11. then n is even. then what statement(s) will you work forward from and what statement should you concl ude? (a) If n is an integer for which n 2 is even. ForwardBackward method versus contradiction method versus contrapositive method. contradiction.1. Both the contrapositive and contradiction methods require that you be able to write down the NOT of a statement. of proof 9.76 CONTRAPOSITIVE METHOD Assume forward A}> Concl ude backward Method ForwardBackward Contradiction (B • • •* (contrad iction) Contrapositive A forward NOT B}> backward (NOT A Fig. . Exercises Note: ~ll proofs should contain an outline as well as a condensed version. comparison of the forwardbackward. If the contrapositive method of proof is to be used on the following propositions.
Prove that.r < 0 ta n ( t) = 1/ r r  1 :s. 9. Prove that if c is an odd integer then the equation n 2 + n . 9. If the contrapositive method is used to prove the proposition "Tf the deriv~tive of the function f at the point x is not equal to 0. Suppose that m and b are real numbers with m ~ 0 and let f be the function that is defined by f (x) = mx + b." which of the following is a result of the forward process? (a) (c) fb) sin 2 (t) fd ) = r2(1 1 .6.sin 2 (t» 9.CONTRAPOSTTTVE METHOD 77 (b) Suppose that S is a subset of the set T of real numbers. then the quadrilateral RSTU is a rectangle. by the contrapositive method. that if no angle of a quadrilateral PSTU is obtuse.c = 0 has no odd integer solution for n.3.2. then x is not a local minimum of f. then there does not exist a real number t between 0 and x/4 such that sin(t) = r cos(t). If S is not bounded then T is not bounded." then which of the following is the correct abstraction question? ~~at is wrong with the other choices? (a) How can I show that the point x is a local minimum of the function f? (b) How can r show that the derivative of the function f at the point x is O? (c) How can I show that a point is a local minimum of a function? (d) How can I show that the derivative of a fUnction at a point is O? 9.5. 9. Prove. . 0 . In a proof by the contrapositive method of the proposition "Tf r is a real number with r > 1. then the rows of ft1 are not linearly dependent. ( c) If x and yare real numbers with x ~ y. f(x) ~ f(y). for all x f: y.4. then f(x) ~ f (y) (where f is a function of one variable) • (d) If the matr ix ft1 is not singular.
you must be able to wrice down the statement NeT B so that you can work it forward. if A is the statement "the real number x is > n. however." ~ "certain Froperty. the word "not" can be el iminated al together by incorporating it into the statement to obtain "the real number x is ~ C. For example. Tn some instances. For instance. Similarly." Tn fact. To use it." or equivalently. the contrapositive method is a valuable proof technique." A more challenging situation arises when the statement contains qUAntifiers. if the statement E were to contain the quantifier "there is" in the standard form: 78 . "the real number x is not > 0. suppose that the statement B contains the quantifier "for all" in the stan~ar~ form: For all "objects" with "something haFpens. you have to know exactly what the statement NOT 'Po is so that you can apply the abstraction process to it. Eimilarly." then the NOT of ~ is "it is not the case that the real number x is > 0." "something happens" with which really means that there is an object wi th the certain property for which the som~thing does not harpen. for all "objects" the "certain property. the NOT of a statement is e~sy to find.10 nots of nots lead to knots As you saw in the prev ious chapter." Then NOT B is the statement: It is not the case that.
If the word NOT appears to the left of a quantifier.6) ~ o. (x 2 + x . 2 2 such such that that ~ . then move it to the right of the quantifier and place it just before the something that happens." Step 3. eliminate the NOT by incorporating it into the statement that appears immediately to its right. NOT (x 2 + x . As you do so.NOTS OF NOTS 79 There is an ·object· with the ·certain property· such that "something happens. the something does not happen. There is a real number x NOT (x 2 + x . There is a real number x ex2 + x . the following For every real number x ~ 2.6) ~ o. 1. ~ for every o. real number ~ x 2. Step 2." or.6) < o.· then NOT 8 is the statement: It is not the case that there is an "object" with the "certain property· such that "something happens. These steps will be demonstrated with examples. When all of the quantifiers appear to the left of the NOT. Step 2. Put the word NOT in front of the entire statement. there are three easy steps to finding the NOT of a statement containing one or more quantifiers: Step 1. certain In general. step 3.6) ~ Step 1. so that "for all· becomes ·there is· and "there is" becomes "for all. you mag ically change the quantifier to its opposite. in other words. for all objects with the property.
NOT there is a real number between 1 and 1 such that y (x 2 + y2) :s 1. For all numbers real (x 2 + x . . for all 2real numbers y between 1 and 1. (x 2 + Y ) > 1. numbers 2." then the words "such that" are removed and a ". if the original statement contains more than one quantifier. Step 1. there is a real number y between 1 and 1 such that (x 2 + y2) :s 1." it becomes necessary to replace the ".6) ~ O. NOT for every real number x between 1 and 1. There is a real number x between 1 and 1 such that. Step ? For (x 2 + all X  2.6) ~ real O. as is illustrated in the next example. ? For every real number x between 1 and 1. when the NOT is passed from left to right. there is a real number y between 1 and 1 such that (x 2 + y2) S 1. Step ? There is a real number x between 1 and 1 such that. NOT x ~ real 6) < o. the quantifier changes but the certain property (namely x ~ 2) does not! Also. if the quantifier "there exists" is changed to "for all." In a completely analogous manner. Step 2. number x ~ 2 such x x ~ ~ that 2 such Step 1. NOT there is a real that (x 2 + x . for all real numbers y between 1 and 1." by the words "such that. There is a real number x between 1 and 1 such that.80 NOTS OF NOTS Note in Step '2 that." is inserted. '2. There is a (x 2 + x . since the quantifier "for every" is changed to "there exists.6) ~ o. Finally. NOT (x 2 + y2) :s 1. then Step 2 will have to be repeated until all of the quantifiers appear to the left of the NOT. Step 2. number Step '2. as is demonstrated in the next two examples.
NOT there is a real number x between 1 and 1 such that. 6. for all real numbers y between 1 and 1. 10. OR Y < 2] is [x < 3] AND fy Remember that. the first thing to do is write down the statements NOT 8 and NOT A. 2]. there is a real number y between 1 and 1 such that (x 2 + y2) > 1. (x 2 + y2) :. Another situation where you must be careful is in taking the NOT of a statement containing the words AND or CR. Step 1. 1.1. (x 2 + y2) :. for all real numbers y between 1 and 1. Similarly. For all real numbers x between 1 and 1.. Just as the quantifiers are interchanged when taking the NOT of the statement. Step Z. NOT fx NOT [x 2: 2: 3 AND Y < 2] is [x < 3] 3 OR [y 2: 2: 2]. 1. NOT for all real numbers y between 1 and 1.. NOT fA AND 8] becomes [NOT A] OR [NOT 8].NOTS OF NOTS 4. Step 2. Write the NOT for each of the of proof in definitions . so the words AND and OR interchange. 81 There is a real number x between 1 and 1 such that. For all real numbers x between 1 and 1. Specifically. there is a real number y between 1 and 1 such that NOT (x 2 + y2) :s 1. when you use the contrapositive method of proof. (x 2 + y2) :s 1. Step 2. NOT [A OR 8] becomes [NOT A] AND fNOT 8). Exercises Note: All proofs should contain an outline as well as a condensed version. For all real numbers x between 1 and 1. For example: 5.
(c) For every "object" wi th a "certain property.1 fa) would read number x· is not a maximum of the f if there is a real number x such > f(x·). then either mn is divisible by 4 or n is not divisible by ~. x + y = 0.4. that if x real numbers such that x ~ 0. by contradiction." (a) For each element x in the set S. n! > n 2 • (b) A impl ies (8 OR C) • (c) A implies (e A~D C). x is in T. For instance. and yare y ~ 0. and . For example." (d) There is an "object" with a "certain property" such that "something happens. (b) A impl ies (8 AND C) • (c) If n is an even integer and m is an odd integer. (d) If f is a convex function of one variable." "something happens.5." 10. 10. then what should you assume? (a) For every integer n ~ 4. 5. the statement "x > 0" can be reworded to read "x is not sO." 10. then x = 0 and y = o. f(x) ~ f(x·).2. Prove. x· is a real number.1.82 NeTS OF NOTS Exercise "the real function that f(x) 5. then what statement(s) will you work forward from and what statement(s) will you work backward from? (a) 'A impl ies (e OR C).3. then for all real numbers y. Reword the following statements so that the word "not" appears explicitly. fey) ~ f(x·). and there Is a real number 8 > 0 such that. (b) There is an angle t between 0 and X/2 such that sin(t) = cos(t).x·. If the contrapositive method is used to prove each of the following propositions. < 8. for all real numbers x that satisfy the property that 'x . 10. If the contradiction method is used to prove the following statements.
but also that the object is unique (i.e. you have the choose and construction methods. that they are really equal).. You will know to use the uniqueness method when the statement P contains the word "unique" as well as the quantifier "there is. the something that ha~pens. your first job is to show that the desired object does exist. The process is illustrated in the next examrle. This can be done by either the construction or the contradiction method. 'The fi rst approach has you aSSume that there are two objects having the certain property and for which the something happens. 'The forwardbackward method is usually the best way to prove that they are equal. Three of these will be developed in this charter.e. In addition. then. There are several other special forms of E that have wellestablished and usually successful proof techniques associated with them." method and In such a case. the contrapos 1tive. The next step will be to ShO\Ol un iqueness in one of two standard ways. using the certain rroperty. If there really is only one such object. and the contradiction methods. when B has quantifiers. 83 . The first is referred to as the un~quene~~ it is associated with a statement B that not only wants you to show that there is an object with a cert~in property such that something happens. you must conclude that the two objects are one and the same (i..11 special proof techniques You now have three major proof techniques to help you in proving that "A impl ies S": the forwardbackwa rd. it is the only such object). and perhaps the information in ~.
= f. The existence of the real numbers x and y was established in example 4 viD the construction method.Y2») = o.y ) are real numbers satisfying 1 1 2 2 (1) a(x l ) + b(Yl) ( 2) c(x ) + d(Yl) 1 a(x ) + b(Y2) 2 (4 ) c (X ) + d(Y2) 2 fubtracting from (3 ) e. Proof of Example 11. such that numbers x (cx + dy) = If a. and f are real numbers (ad . To that end. Hence one obtains a(x 1 ) + b(Yl) = e and c(x l ) + d(Yl) = f. c. Here. only the issue of uniqueness will be addressed. assume that (x ." by the method described above. Hence. d. and also that a(x 2) + b(Y2) = e and c(x 2) + a(Y2) = f. The existence of the real numbers x and y was established in Example 4 via the construction method. Outline of proof. Thus. or equivalently. it will be shown. by the forwardbackward method. that (x. ( 2) yielc'ls X 2) X + b (y1 c nc 2) + d(Yl . then there are unique real and y such that (ax + by) = e and f. e. e. (1)  en anC' Y2 ) ) (~ ) from (). b. f. the abstraction question is "Bow ci'ln I show that two pairs of real numbers (namely (xl 'Yl) ant' (x 2 'Y2» are equal?" Using the definition of equality of orc. Y2)2= (1.x 2) = 0 anc'l (Yl 1.84 SPECIAL TEC~NICUES Example 11.bc) ~ (1. the unlqueness will be establishec. . Eoth of these statements are obtained from the forward process by applying Some algebraic manipulations to the four equations and by using the fact that (ad ."ered pairs (see refinition ~). one answer is to show that both xl = x2 and y = y .be) ~ 0.y ) C'lnd (x . . Using these four equations and the assumption that A is true. that the two objects (xl 'Yl) and (x 2'Y2) are equal. Specifically. you assume that (xl 'Yl) and (x 2 'Y2) are two objects wi th the certain property and for which the something harpens.
suppose that x and yare two different real numbers such that x 3 = rand y3 = r. OUtline of pr~of. and c = y2. If r is a positive real number then there is a unique real number x such that x3 = r. you must then reach a contradiction. by using the certain property. To show that r = 0. From the hypothesis that (ad . and on f~ctoring. Thinking of this as a quadratic equation of the form (ax 2 + bx + c) = 0 in which a = 1.bc) .y) (x 2 + xy + y2)] = O. To that end. and thus the uniqueness is proved. the information in ~. The appearance of the quantifier "there is" In the conclusion suggests using the construction method to produce a real number x such that x 3 = r.// '2 The second method for showing uniqueness has you assume that there are two different objects having the certain property and for which the something happens. Using this information together with the hypothesis that r is positive. y. a contradiction will be reached by showing that r = 0. it follows that [(ad . and especially the fact that the objects are different. This part of the proof will be omitted so that the issue of uniqueness can be addressed.x 2)] = O. and then subtracting the second equation from the first one. the quadratic formula states that x = 2 = 2 . so. . Here is where you can use the fact that x ~ y to divide by (x . Now supposedly this cannot happen. 0. and especially the fact that x. In particular.SPECIAL TECHNIQUES On multiplying the first of these equations by d 85 and the second one by b. This process is demonstrated in the next example. since x 3 = rand y3 = r. thus contradicting the hypothesis that r is positive. b = y. one has (x. the something that happens. .x 2) = 0. one can work forward. Thus (x 3 . one obtains [(x .bc) (x. and hence x. obtaining (x 2 + xy + y2) = O. Example 12.y) .y3) = 0. = x2' A similar sequence of algebraic manipulations will establish that y = y. it follows that x3 = y3.
it fo11o. ~rorking forward from the assumption that (x 2 . Applying the forwardbackward method. in this case.5x + 6) :!: 0 and x > 2. the either/or method is to be used when trying to show that the statement "A implies r on t" is true.5x + 6 :!: 0.e. it must be that y • 0.3») :!: O. But then r = y3 = 0. and hence (x . 0. Outline of proof. and if y. By the quadratic formula. Only the issue of uniqueness will be addressed. you can assume that (x 2 .5x + 6) ~ 0.~ 2 Since x is real. To that end. Suppose that you were to make the additional assumption that C is not true. rex . Since x > 2. . it must be that y = C. assume that x and y are two different real numbers for which x 3 = rand y3 '" r.3) :!: (1 Ci.88 SPECI~L TECHNIQUES Since x is real and the above formula for x requires taking the square root of _3y2. you can assume that A is true and C is false and you must then conclude that C is true. From the above discussion. 0 is true. thus contracicting the hypothesis that r is positive. e~the~/o~ ~nother method. (x . It is your job to conclude that x :!: 3. Thus.. Hence. you would begin by assuming A is true and you would like to conclude that either C is true or else 0 is true. If x 2 . as desired). and the contradiction has been reached. In other \'lords. x a 3. or else D is true" (where C and C are statements)./ / special proof technique. it must be that (x 2 + xy + y2) '" O. called the arises when B is of the form "either r is true.2)(x . as is illustrated in the next example.2) > (I. with the either/or method. Clearly it had better turn out that. Example 13.""s that [(x . then x :s 2 or x :!: 3. Since x ~ y. it follows that 0 (x 3 _ y3) .. Proof of Example 12.y) (x 2 + xy + y2)]. it follows that II: x = y+ VCy24y2) 2 = y:!:. then r z y3 = 0.
as: (a) (b) (c) (d) min{s: min{s: max {s: max {s: s is s is sis sis in S} in S} inS} inS} :: :s :s :: x x x x statements The proof technique associated with the first two are discussed here and the remaining two are left as exercises.// It is worth noting that the either/or method could have been done equally well by assuming that A is true and D is false.'x.2)(x . 4. and then concluding that C is true. you might want to prove one of the following statements: 1. and. 3. 12(b». the problem of determining if the smallest member of S is : : x. In mathematical problems these four are likely to appear.2) > 0. For instance. the idea behind the max/min technique is to convert the given problem into an equivalent problem containing a quantifier.3)] ~ 0.n method that arises in problems dealing with maxima and minima. Consider. Assume that (x 2 . Nonetheless. Then. All Some All Some of of of of S is to S is to S is to S is to the the the the right left left right of of of of x x x x (see (see (see (see Figure Figure Figure Figure 12(a». 2. respectively. It follows that (ex .5x + 6) ~ 0 and x > 2. Try this approach on Example 13. Since "all" of S should be to the right of x. the appropriate choose or construction method can be used. Suppose that S is a nonempty set of real numbers having both a largest and a smallest member. 12 fc) ) • 12fd».SPECIAL TECHNIQUES 87 Proof of Example 13. you might be interested in the position of the set S relative to the number x./m. since (x . it must be that x :::: 3 as desired. An equivalent problem containing a quantifier can be obtained by considering the corresponding statement (1) above. For a given real number x. you . The final proof technique to be developed in this chapter is the ma.i. therefore.
12(<.88 SPECIAL TECHNICUES ~ need to show that. 12(c). x is illustrated in the next example. ~1~[~3+.Real line S o x Pig. ~ll of S to the left of x. of S to the left of x. All of S to the right of x.'1). s. 12(a). for all elements s in S. +1+[+3+o x S neal line Pig. as ++[13+. Eome of S to the right of x.neal line o x s rig. . S o [ x ~ome 3 Real line Pig. 12(b).
Prove that if x is a real number that is > 2.:: O.SPECIAL TECHNICUES 89 Example 14.1): x is in Il} . ~s per the discussion above. i t becomes clear that the choose method should be used to choose a real number x for which it must be shown that x(x . 70 proceed.1) :! 1/4 or. Proof of Example 14. let x be any real number. Exercises Note: All proofs shoulo contain an outline as well as a concensed version.x + I/Ll) . Then it follows that x(x . since (x 2 . the approach is slightly different.X + 1/4) = ex . and the proof is complete.:: 1/4. an equivalent problem is to show that there is an element s in S such that s s x. But. !hen. of proof 11.1/2)2 . consider the corresponding statement (2) above. then min{x(x . the statement P can be converted to "for all real numbers x.1) ~ 1/4. Outline of proof.:: r. that (x 2 . the construction or contradiction method could be used. The final chapter provides a comrlete summary.II Turning now to the problem of showing that the smallest member of ~ is s x. the maximin method will be used. In order to conclude that min{x(x .x + lit') = ex .1/2)2.:: 1/4.1) is ~ 1/4 because (x 2 . From the form of B. Since "some" of S should be to the left of x.1. x(x .:: C.1): x is in Il} . this number is always." Once in this form. then there is a unique real number y < 0 such that x = 2y/(1 + y) • . which is . This chapter has described three srecial proof techniques that are appropriate when E has the corresponding special form. If R is the set of all real numbers.
b. Prove." (b) Prove the proposition by assuming that A and NOT B are true. Frove that if a and b are real numbers.5." (a) Reword the rroposition so that it is of the form "A implies ~ OR C. What would be the advantages and disadvantages of using the contradiction m~thod instead of the either/or method to prove that "A implies (E OR C)?" 11.8.7. then minIs: s is in S} ~ t"'. t ~ t*. a.4. such that (a + bi) (c + di) = 1. (c) Prove the proposition by assuming that ~ and ~OT C are true. 0. then there is a unique number x such that mx + b = O. say c + di. 11. and i = y:I.90 SFECIAL TECHNICUE5 11. Prove that if S is a nonemFty subset of a set T of real numbers and t· is a real number that satisfies the Froperty that for each element t in T. (c) min{cx: aX!!i b and x ~ OJ:s u (d) max{cx: ax ~ b and x ~ OJ ~ u (e) min{ax: b:S x :s c} ~ u (f). Consider the proposition "If x is a real number that satisfies x3 + 3x 2 . and x is a variable. by using the second uniqueness method. (Note: S is a set of real numbers and x is a given real number. ) (a) maxIs: s is in S} !!i x (b) maxIs: s is in 5} ~ x In the remainder of this problem. c. then Ixl ~ 3. at least one of which is not r. Prove that if n is an even integer and m is an odd integer.6.2. and u are given real numbers. max{bx: a!!i x :s cJ :s u 11.27 ~ 0. . Convert the following max/min problems into the appropr iate Quanti fier statement.~x .3. 11. 11. that if m and b are real nUMbers with m . 11. then there is a unique complex number. then either 4 divides mn or 4 ~oes not divide n.
SPECIAL TECHNICUES 91 11. b.9. Prove that min{cx: ax ::: b. u ::: CJ • . Suppose that a. and c are given real numbers and that x and u are variables. x ::: OJ is at least as large as max {ub: ua '5 c.
When obtaining the sequence of statements. you can assume that A is true and your job is to prove that B is true. with the property that if it is true. watch for quantifiers to appear. through asking and answering the abstraction question. 92 . The objective is to I ink the forward sequence to the backward sequence by generating a statement in the forward sequence that is precisely the same as the last statement obtained in the backward sequence. you can do the proof by going forward along the sequence from A all the way to B. Through the forward process. for then the construction. A final summary of how and when to use each of the various techniques for proving the proposition "~ implies B" is in order. then so is B. Perhaps you will develop some of your own. and so on.12 summary The list of proof techniques is now complete. A2. obtaining a new statement. you derive from B a statement. Then. you will derive from A a sequence of statements AI. It is guided by the backward process whereby. like a column of dominoes. Undoubtedly you will come across others as you are exposed to more mathematics. B2. there are many fine points and tricks that you will pick up with experience. choose. With the forwardbackward method. The techniques presented here are by no means the only ones. but they do constitute the basic set. In any event. El. • • • which are necessarily true as a result of A being assumed true. and/or specialization methods may be useful in doing the proof. This sequence is not random. This abstraction process can then be applied to Bl. induction.
when the quantifier "for all" arises in the backward process in the standard form: For all "objects" with a "certain property. if necessary) and backward from the something that happens. Remember that the success of a proof by induction rests on your ability to relate the statement P(n+l) to P(n) . You must conclude that. Once you have chosen the object. Here. With the construction method. On the other hand. or devise an algorithm to produce. then P(n+l) is true. a statement P(n) is true. when the quantifier "there is" arises in the backward process in the standard form: There is an "object" with a "certain property" such that "something happens" you should consider using the construction method to actually produce the desired object." "something happens" you should consider using the choose method. it is best to proceed by working forward from the fact that the chosen object does have the certain property (together with the information in A. Be sure to check that it satisfies the certain property and also that the something happens. The second step requires you to show that if Prn) is true. etc. your objective is to design a proof machine that is capable of taking any object with the certain property and proving that the something happens. 70 do so.) the object.SUMMARY 93 For instance. for that object. you select (or choose) an object that does have the certain ~roperty. you work forward from the assumption that A is true to construct (produce. the something happens. The first step of the induction method is to verify that the statement is true for the first possible value of n. The induction method should be considered (even before the choose method) when the statement 8 has the form: For every integer n greater than or equal to some initial one.
That fact should then be helpful in reaching the conclusion that B is true. remember to watch for quantifiers to appear in the forward or backward process. for then. watch for one of these objects to arise in the backward process. Once again. induction. using the techniques of Chapter 10 if necessary. you must immediately write down the statements NOT B and NOT A. by using specialization. your job is to conclude that NOT A is true. to perform the second step of the induction proof. replace n everywhere by (n+l) to obtain Pfn+l). choose. When using specialization. To use the contrapositive approach. be sure to verify that the particular object does satisfy the certain property. Only then will you be able to use the assumption that pen) is true to reach the desired conclusion that Pfn+l) is also true. To do so." or when the forwardbackward method fails. with the former being chosen first.94 SUMMARY so that you can make use of the as~umption that Pen) is true. Then. With this approach. when the quantifier "for all" arises in the forward process in the standard form: For all "objects" with a "certain property. and then see if you can express pen+l) in terms of pen). you should write down the statement pen). In the event that the contrapositive method fails. working forward from the statement NOT Band backward from the statement NOT A. When the original statement B contains the word "not. you can conclude that the something does happen for the particular object that is under consideration. Finally. In other words. there is still hope with the contradiction method. you should consider the contrapositive or contradiction method. for if they do. for only then will the something happen. and/or specialization methods may be useful." "something happens" you will probably want to use the specialization method. then the corresponding construction. This is best accomplished by applying the forwardbackward method. you are allowed to . by beginning with the assumption that NOT B is true.
" "either • • • or • • • . either/or." let the form of e guide you as much as possible." and "maximum" and "minimum. choose." then consider the construction method. This gives you two facts from which you must derive a contradiction to something that you know to be true. In trying to prove that "A implies B. Where the contradiction arises is not always obvious. and corollaries. lemmas. either/or. as they will often indicate how to proceed. then you should proceed with the forwardbackward method. and max/min methods. and max/min methods." for then you would use the corresponding uniqueness. . induction. Your new "vocabulary" and "grammar" is complete. theorems. your bag of proof techniques includes the uniqueness. Other key words to look for are "uniqueness. you may wish to stick to Greekafter all. For special situations. whereas the quantifier "for all" suggests using the choose or induction method. 1able 4 provides a complete summary. and specialization methods. They are the forwardbackward. you should scan the statement e for certain key words. If all of these proof techniques fail. you will probably want to use either the contrapositive or contradiction method. contrapositive. For example. Specifically. You have learned the three major proof techniques for proving propositions. When the statement B has the word "not" in it. it's all Greek to me. If you are unable to choose an approach based on the form of B. You are now ready to "speak" mathematics. if you come across the quantifier "there is. You have come to know the quantifiers and the corresponding construction. and contradiction methods. but it will be obtained by working the statements A and NOT B forward.SUMMARY 95 assume not only that A is true but also that B is false.
" etc. or when B does not have a recognizable form When B has the word "not" in it NOT B What to Assume Contraposi tive (page 72) Contradiction (page 64) When B has the word "not" in it." "for each. Summary of Proof Techniques. and choose an object with the certain property The statement is true for n Induction (page SO) When B is true for each integer beginning with an initial one." etc. or when the first two methods fail When E has the term "there is. say no ***************************************************** .96 SUMMARY Table 4. 11. A and NOT B Construction (page 34) Choose (page 40) When B has the term "for all. ***************************************************** Proof Techn ique ForwardBackward (page 8) When to Use It As a first attempt." "there exists.
That the something happens The sta temen t is true for n + 1. construct. 'Also work backward from the something that happens. Then invoke the induction hypothesis for n to prove it true for n + 1. the object having the certain property and show that tre something happens Work forward from A and the fact that the object has the certain property. etc.. 'Also show it true for no' ***************************************************** . First substitute no_ everywhere and show it true.SUMMARY 97 ***************************************************** What to Conclude B HoW to Do It Work forward from ~ and apply the abstraction process to B NOT 'A Work forward from NOT Band backward from NOT 'A Some contradiction Work forw~rd from ~ and NOT B to reach a contradiction There is the desired object Guess.
. and A Uniqueness 2 (page 85) Either/ or 1 (page 86) Either/ or 2 (page 87) JYlax/min 1 (page R7) fIIIax/min 2 (page 89) When E has the word "unique" in it When B has the form "C OR 0" When E has the form "C OR 0" When B has the form "max S ~ x" or "min S 2: x" When E has the form "max S 2: x" or "min S ~ x" There are two different objects." etc.. and A . and A A and NOT C A and NOT D Choose an s in S. What to Assume Uniqueness 1 (page 83) When B has the word "unique" in it There are two such obj ec ts." "for each. ***************************************************** Proof Technique Speclal1 zation (page 5~) When to Use It When }. Summary of Proof Techniques (continued). has the term "for all.98 SUMfillARY Table 4. ***************************************************** .
Work forward from A using the properties of the two objects and the fact that they are different Work forward from A and NOT C.SUMMARY 99 ***************************************************** What to Conclude B How to Do It Work forward by specializing A to one particular object. and backward from C 'J'he two objects are equal Some contradiction D C s s :s x or ?: x Work forward from A and the fact that s is in S. the one obtained in the backward process Work forward using A and the properties of the objects. Also work backward to show the objects are equal. ~lso work backward. and backward from r Work forward from A and NOT C. Use ~ and the construction method to produce the des ired sin S Construct s in S so that s ::: x or s :s x ***************************************************** .
b.2. f(x·) S fey). Describe .1.(f(y) + g(y»1 < £. then the maximum value of ab + bc + ca subject to the condition that a 2 + b 2 + c 2 = I is s 1. g(x) 5 M. then there is a real number x· between 0 and I such that. f(x) s M. For each of the following statements. and c are real numbers. how you would use each of the following proof techniques to prove that "for every integer n ~ d. V x. (f) If f and 9 are two functions such that (1) for all real numbers x.yl < 8. (a) If p and ~ are odd integers then the equation x + 2px + 2q = 0 has no rational solut ion for x. (i) If f is a function of one variable defined by f(xl = 2· + x 2/2. nl > n 2 • ec) If f and 9 are convex functions then f + g is a convex function. (g) If f and 9 are continuous functions at the point x.") (d) Contradiction method 12.100 SUMMARY Exercises 12. (al Induction method (b) Choose method (cl ForwardBackward method (Hint: Convert the problem to an equivalent one of the form "if • • • then. indicate which proof technique you would use to begin the proof and explain why. then for every real number £ > 0 there is a real number 8 > 0 such that. (b) For every integer n ~ 4. nl > n2 " State what you would assume and what you would conclude. (h) If f and 9 are continuous functions at the point x. (e) In a plane. then there is no real number M > 0 such that. for all x. for all y. then so 1s the function f + g. . f(x) 5 g(x) and (2) there is no real number ~ such that. If(x) + g(x) . (d) If a. there is one and only one line perpendicular to a given line t through a point P on the line. for all real numbers y with Ix .
) 81. it is necessary for you to figure out which proof techniques are being used and why. which then explains how to read the condensed proof. then T is not bounded. The example deals with the concept of an "unbounded" set of real numbers. the condensed version of the proof will precede the outline. Proof of Example 15. each sentence of the proof will be written on a separate line. thus forcing you to reconstruct it. Suppose that S is a subset of T and assume to the contrary that T 1s bounded. this appendix will show you how to read and how to understand a written proof as it might appear in a textbook.A putting it aU togetherl Through the use of a specific example. Subsequently. Due to the way in which most proofs are currently written. Often. what makes a written proof so difficult to understand is the fact that the author has omitted part of the thought process. M. A set there that. Example 15. Jf S is a subset of a set T of real numbers and S is not bounded. The next example demonstrates how to read a proof. Definition 17. some general problemsolving strategies will be ~iscussed and demonstrated on a problem. Doing so requires that you identify which proof technique is being used and how it applies to the particular problem. meaning that there are elements in the set that are "arbitrarily" far away from O. A formal definition of a bounded set will be given first. Unlike previous examples in this pamphlet. (For reference purposes. I xI < S of real numbers is bounded if is a real number M > 0 such for all elements x in S. 101 .
The author is working forward from the statement NOT B (i. Hopefully.. statements 92 through S6 will indicate how the author has worked forward from A and NOT B to reach the contradiction.. of the Interpretation of 81. that is. The use of the contradiction method should not come as a surprise since the conclusion of the example contains the word "not. and this contradicts the hypothesis that 9 is not bounded. 84. i. Hence there is a real number for a11 x in T. Note that the author has not explicitly stated that 9 is not bounded. Xl 85. that S is a subset of T. it follows that in T." Sometimes it is beneficial to scan through the proof to find the contradiction. although it is being assumed true. Interpretation of 82.102 APPENDIX A MI 82. To that end. It is here that the author states the desired objective of showing that S is bounded. But then I Xl I < MI and so 9 is bounded. It will be shown that S is bounded. An interpretation of each statements 91 through 96 will be given. that T is a bounded set) via Definition 17 to claim that there is the real number MI > 0 such that. . and also that NOT B is true. for all x in T.1. that is. > 0 such that.e. 9ince 9 is a subset of T.s is 86. In 96 it says that S is bounded. The author is indicating that the proof is going to be done by contradiction and appropriately is assuming that (part of) A is true. which a contradiction. Observe that there is no indication that the reason for wanting to show that 9 is bounded is to reach a contradiction. that T is bounded.11 OUtline of proof.. I xI < . 83. let Xl be an element of s. Interpretation of 83. I xl < MI.
Ixl < M'.. one should then proceed by the choose method. for all elements x in S. has the desired properties. for all x in S. What the author has failed to tell you is that the value for f. which the author has not explicitly said he would do. Recognizing the appearance of the quantifier "for all" in the backward process.e. Ixl < M. Then. This statement might seem cryptic at first." Once the element x' in S has been chosen. it must be shown that ~. x is in T (see Definition 14). Such details are often left for you to figure out for yourself.. This is precisely what the author has done in S4. the author has suddenly switched to the forward process and is working forward from]. for all x in S. What makes 84 so difficult to understand is the lack of sufficient detail. is in T. What has happened is that the author has implicitly posed the abstraction question "How can I show that a set (namely S) is bounded?" and has then used Definition 17 to answer it.4 is~' (see S2). particularly the omission of the abstraction question and answer. it must then be shown that Ix'i < M'. At this po int. Interpretation of SSe In S5. Observe that the word "let" in S4 indicates that the choose method has been used.. whereby it must be shown that there is a real number ~ > 0 such that. > 0 'which it is) and that. Note once again that part of the author'S thought process has been omitted and left for you to figure out. i. and the reason is that the author has omitted several steps of the thought process.APPENDIX A 103 Interpretation of S4." the construction method should be used to produce the desired value for~. thus concluding that x. where it states: " • • • let x' be an element of S.e. (1. Given that this is the case. Recognizing the quantifier "there is. . that ~. and several other steps. from the fact that 8 1S a subset of T) by using the definition of a subset to conclude that. the author has specialized this statement to the particular value of x' that is in S. it is not clear why it is necessary to show that Xl is in T. without stating so.
Learn to identify the various techniques that are being used. Once again. Begin by trying to determine if the forwardbackward or contradiction method is the primary technique.. In summary. expect to do some work of your own. the author has omitted part of the thought process. when reading proofs.. Actually. Read the condensed proof again." If this had been done.. Observe that what makes it difficult to understand is the lack of reference to the specific techniques that were used.104 APPENDIX A should be related to showing that The reason Ix' I < . construction. The only question is how did the author reach the conclusion that Ix'i < "". when using specialization. the specific proof technique has not been mentioned. induction. The fact that x' is in T was established in ~5.. for then the corresponding choose. at the time. For instance. especially the abstraction process and the specialization method. recall that. but observe that. then you would have had to distinguish between such statements as "for all x with ~ certain property. Then try to follow the methodology associated with that technique. the author has used specialization.." in which "x" refers to a general object. and a statement of the form "let x have a . To understand the proof you have to fill in the missing steps. Specifically. the author could have used the symbols "x" and "~" instead of "x'" and "M'. in the condensed proof of Example 15. the author did not tell you why it was being shown that x' is in T. Be watchful for quantifiers to appear. something happens. thus showing that S is bounded and reaching a contradiction to the hypothesis that S is not bounded. It is here that the author finally concludes that Ix' I < M'. In addition. However. the statement S2 has been specialized to the particular value of x. it is necessary to verify that the particular object under consideration (x' in this case) has the certain property (that x' is in T). Interpretation of S6. Be particularly careful of notational complications that can arise with the choose and specialization methods. and/or specialization methods are likely to be used.
If T is a bounded subset of real numbers.APPENDIX A 105 certain property. r'oing so in this case means you must show that: B2: There is a real number elements x in S. it is time to see how to do a proof on your own. Now tha t yo u know how to read a proof. you would write: "Let S be a subset of T. the forward process will be applied to the hypothesis that T is a bounded set of real numbers. learn to ask yourself how you would proceed to do the proof. Not recognizing any special form to Bl. as is illustrated in the example in Appendix S. for all . when you are unable to follow a particular step of a written proof. it is probably best to proceed with the forwardbackward method. In any event. can you pose an abstraction question? One such question is "How can I show that a set (namely S) is bounded?" 1\ reasonable way to answer an abstraction question is by a definition. ~ccordingly. In the example that follows. Outline of proof. Looking at the statement S." Then the new statement to be proved is: B1: S is bounded. Ixl < ~ ~. The double use of a symbol is confusing but common in proofs. It will be shown that S is bounded. Then.~ in which ~x~ refers to a specific object. Example 16. > 0 such that. then any subset of T is bounded. 70 fill in the gaps. it is most likely due to the lack of suffic ient detail. pay particular attention to how the form of the statement under consideration dictates the technique to be used. you should recognize the quantifier ~for any" and thus you should begin with the choose method. The idea is to work backward from Bl until it is no longer fruitful to ~o so. Working backward from Bl. Then try to see if the written proof matches your thought process.
whereby you would choose an element. that means setting M = ~'. M > 0) and also that the something happens (I. that for all x in S. but look carefully at the statement AI. since M = M'. It is not hard to see that ~ > 0 because ~. Perhaps ~. When using the construction method. > 0 and ~ = M'. it might not be clear how to reach the conclusion that Ix'i < M'. in S for which you must show that: 84: Ix'i <~. 1 t Is generally advisable to try to construct the desired object in the most "obvious" way. I xl < ~. Observe that you could conceivably use specialization to reach the desired conclusion that Ix'i < M'. provided that you can specialize the statement to x = x'. Thus it remains only to show that: 83: For all elements x in S. > 0 such that.. whereby you can conclude that: AI: There is a real number elements x in T. The appearance of the quantifier "for all" suggests proceeding with the choose method. is there anything that you can say is true? At least one possibility is to use the definition. for all ~'. Ixl < ~. Ix'i < ~'. Observe that the symbol "x'" has been used for a particular object to avoid confusion with the symbol "x.e. If this "guess" is correct.108 APPENDIX A The appearance of the quantifier "there is" suggests using the construction method and also indicates that you should turn to the forward process to produce the desired value for~. In order to be able to do so. Working forward from the hypothesis that T is bounded. you must make sure that the particular object under consideration (namely x') . Ixl < ~).e. Observe that the symbol "~'" has been used instead of "~" because "~" was already used in the backward process. is the desired value of~.. or. then you must still show that the certain property holds (i." At this point. In this case. say x'.
Recall that x' was chosen to be an element of S and that S was chosen to be a subset of T. If no form is apparent. it is advisable to seek another proof technique consciously.11 Problem solving is not a precise science. Let S be a subset of T. CA. You might try asking yourself why B cannot be false. but a more detailed discussion can be found in George Polya's book How to Solve It (Princeton University Press. x' is in T. therefore. thus leading you into the contradiction (or contrapositive) method. If unsuccessful. then it is probably best to proceed with the forwardbackward method. Princeton. when you are unable to complete a proof for one reason or another. How can you use this information to show that x' is in T? Working forward from the definition of a subset. To see that ~ has the desired properties. 1915) orin Wayne Wickelgren's book How to Solve Problems (W. Freeman. a real number ~ > 0 will be produced with the property that. let x' be an element of S. The point is that. which you know is in S. for a11 x in S. there are several avenues to pursue before giving up. H. completing the proof. you can specialize this statement to x = x'. I x I < ~. Since £ is a subset of 1'. for all x in T. as desired. Set ~ = M'. NJ. l' is bounded. Ixl < ~'.APPENDIX A 107 does satisfy the certain property (namely that x' is in T). and so there is a real number ~. Sometimes the form of the statement B . To show that S is bounded. for all x in S. you know that. i?ut then it fo11ows that Ix'i < ~ = M'. San Francisco. x is in T. Your objective. Some general suggestions will be given here. Proof of Example 16. By hypothesis. 197~ \\'hen trying to prove that "'II implies B. Thus. and reach the desired conclusion that x' is in T. > 0 such that. is to show that x' is in T." consciously choose a technique based on the form of B.
For instance." If the choose method fails. . you might see a new approach. as you proceed through a proof. the contradiction method suggests itself because of the appearance of the word "no. different techniques will be needed as the form of the statement under consideration changes. rewriting B so that it reads: then you might try There is no "object" with the "certain property" such that the "something" does not happen. Your objective would be to reach a contradiction." "something happens. Remember that. When the statement B does have a recognizable form. suppose that the statement B contains the quantifier "for all" in the standard form: For all "objects" with a "certain property. it can sometimes be advantageous to leave the problem for a while.108 APPENDIX A can be manipulated so as to induce a different technique. for when you return to it. In this form. If you are really stuck." whereby you would assume that there is an object with the certain property such that the something does not happen. Undoubtedly you will learn many tricks of your own as you solve more and more problems. it is usually best to try the corresponding proof technique first.
there is a real number 8 > 0 such that. meaning that if the value of the variable is changed "slightly" then the value of the function does not change "radically. a definition of continuity is given. continuous at 109 .B putting it aU togetherll This appendix will give you more practice in reading and doing proofs. the condensed version of the proof will precede the detailed outline which then explains how to read the condensed proof. The additional difficulty is due. Definition 18. To proceed formally. in part.fey) I < &. to the fact that the statements under consideration contain three quantifiers as opposed to two. ) S1. The example deals with the concept of a "continuous" function of one variable. (For reference purposes. A function f of one variable is continuous at the point x if for every real number & > 0. it then follows that If(x) . The first example is designed to teach you how to read a condensed proof as it might appear in a textbook. (Note: The function f + 9 is the function whose value at any point y is fCy) + g(y». The examples presented here are more difficult than those in Appendix A.yl < 8. each sentence of the proof will be written on a separate line. Proof of Example 17. If f and 9 are two functions that are continuous at x. for all real numbers y with the property that Ix . As such. then the function f + 9 is also continuous at x. let & > o. Example 17. To see that the function f + 9 is x." Another way of saying that a function is continuous is that its graph can be drawn without lifting the pencil.
Since f is continuous at x. there is a 8 2 > 0 such that. The double use of . for all y with Ix ." Recognizing the appearance of the quantifier "for every" in the backward process. Similarly. The author has done this ~nen he says" • • • let & > 0. SSe Let 8 . such that. of the Interpretation of Sl. > 0 wi th I x ." Observe that the symbol "&" appearing in the definition of continuity refers to a general value for & while the symbol "&" appearing in 51 refers to a specific value reSUlting from the choose method.f(y) I + Ig(x) . since 9 is continuous at x. If(x) + g(x) (f(y) + g(y»1 < &.yl < 8.. and so If(x) + g(x) ff(y) + g(y» I :s If(x) . there is a 8. whereby a specific value of & > 0 would be chosen. whereby it must be shown that. and the proof is complete.y I < 8 • S6.110 APPENDIX B S2. 2 Ig(x) . An interpretation of each statements 81 through 86 will be given. ' S4. "something happens. What has happened is that the author has implicitly posed the abstraction question "How can I show that a function (namely f + 9) is continuous at a point (namely x)?" and has used Definition 12 to answer it.yl < 8 . for all y with Ix . It will be shown that there is a 8 > 0 such that.g(y) I < &/2 + &/2 = &.f(y) I < &/2.Then Ixyl<8 .// OUtline of proof. This statement indicates that the forwardbackward method is going to be used because the statement "to see that f + 9 is continuous at x" means that the author is about to show that the statement e is true.g(y) I < & /2.y I < 8. for all y If(x) . 83. The question is why did the a uthor say" • • • let & > O?" Hopefully it is related to showing that f + 9 is continuous at x. min{ 8 8 2 } (which is > 0) and let y have " the property that I x . one should then proceed with the choose method. and Ixyl<8 2 . for every & > 0.
where & is the one that was chosen in Sl. the something happens. What the author has failed to tell you is that. The values of 81 and 8 2 . Note. Keep in mind that. 8 }. The author has turned to the forward process (having recognized from 82 that the quantifier "there is" suggests using the construction method to produce the desired 8 > 0). namely. it must be shown that. 5uch details are often left for you to figure out for yourself • Interpretation of 82. It is not yet clear why 8 has been constructed 1 in 2 this way. /it this moment. The author is working forward from the fact that 9 is continuous at x in precisely the same way that was done in 53. The reason will become evident in 56. Tha t something is "there is a 8 > 0 " as stated in 82 (see Definition Ie). the quantifier "for all" arises. It is here that the desired value of 8 > 0 is produced. The author is working forward from the hypothesis that f is continuous at x by using Definition Ill. the abstraction question and answer. are combined to give the desired 8. but why should he want to show that there is a 8 > 0 such that "something happens?" Recall that. in 51. and the statement has been specialized to the value of &/2. in the definition. the author is trying to construct a value of 8 > 0 for which something happens. from 83 and S~. it is not clear why the author decided to specialize to the value of &/2 as opposed to &. Interpretation of 85. Interpretation of 84. 8 = mint 8 . namely. again. The author is stating precisely what he is going to do. In any . for that &. respectively. can 111 be confusing but is common in written What makes 51 so difficult to understand is that the author has omitted part of the thought process. that the symbol "&" in Definition le refers to a general value for & while the symbol "&" in 51 refers to a specific value for &. Interpretation of 83.APPENDIX B a symbol proofs. the choose method was used to select an £ > 0 and so. from 52.
If(x) + g(x) If ey) + 9 ey) ) I < I: (see S2).(f(y) + gey» I < 1:. in S6.yl < 8" is because the choose method is being used to show that.yl < 8 2 . it is the author's responsibility to show that property that.g(y) I because the absolute value of the sum of two numbers. Ix . it must be shown that If(x) + g(x) . 82 }. and I x . Pence. namely f(x) .g(y).y I < 8 and I x .y I < 8 2 . Specifically. was chosen so that I x . to complete the proof.g(y) I < 1:/2. statements S3 and S4 have both been specialized to the particular value of y that was chosen in SS. The only question is how. If(x) + g(x) .fey) I < 1:/2 and Igex) . it must be that both 8:s 8. Note.yl < 8 . Interpretation of S6. Recall that. Since y.yl < 8 and 8 = mint 81 .g(y) I < 1:/2 + 1:/2 = 1:.fey) I + Ig(x) . and 8 2 are > 0). and Ix .yl < 8. In S5 the author merely remarks that 8 > 0 (which is true because both 8 . the double use of the symbol "y." Nonetheless.y I < 82 ). the author is justified in claiming that Ifex) + g(x) (f(y) + g(y»1 :s If(x) . the author does indeed claim tnat Ix . one must be sure that the particular object under consideration (namely y) satisfies the certain property(in this particular case.g(y) I < 1:/2 + 1:/2? Once again the author has omitted part of the thought process. but why is If(x) . Also. tHe author is justified in specia1izing S2 and S4 to this particular value of ' y and hence can conclude that If(x) .g(y) I < 1:/2.fey) and g(x) . Perhaps now it is . but can you see why this is true? Look again at 55.yl < 8.fey)1 + Ig(x) . and 8:s 82 . is always :s. In other words.fey) I + Ig(x) .112 APPENDIX B 8 > 0 and also that 8 satisfies the desired event. the sum of the absolute values of the two numbers. It is not hard to see that If(x) + g(x) (f(y) + g(y» I :s If(x) . the reason that the author has said • • • • let y have the property that Ix . once again. What has happened is that specialization has been used to claim that If(x) .yl <8. Note that. It is here that the author concludes that If(x) + g(x) (f(y) + g(y» I < 1:. so indeed I x . when using specialization. for all y with Ix . for all y with Ix .(f(y) + g(y» I < 1:.fey) I < 1:/2 and Ig(x) .
Loo king at B2. in this cllse: 821 There is a 8 > 0 such that. all y with Ix .yl. OUtline of proof.APPENDIX B also clear why the author specialized (in 53 and to t/2 instead of t. yo u should all y with the recogni ze . for all y with Ix . If(x) . then f is continuous at x. If(x) . for Looking at Bl. satisfies the property that there are real numbers c > 0 and 8' > 0 such that. Thus. Begin by looking at the statement B and trying to select a proof technique.yl < 8.fey) I :s clx .yl < 8. If f is a function of one variable that. a definition is available to provide an answer.fey) I < t.fey) I < t'. at the point x. It is advisable to use a symbol other than "t" for the specific choice so as to avoid confusion. note that what makes 5f so hard to understand is the omission of part of the thought process. that being to show that: 81: For all t > 0. The next example illustrates how to go about doing your own proof. Since B does not have a particular form. there is a 8 > 0 such that. the forwardbackward method should be used. you should recognize the quantifier "for all" and hence use the choose method to select a particular value of t > o. 113 S~) Again. for Ix . You must learn to fill in the missing links by determining which proof technique is being used and consequently what has to be done. Example 18. In the condensed proof you would write "Let t' > 0 • • • • " Once you have selected t' > 0. If(x) . the choose method requires you to show that the something happens. Pay particular attention to how the form of the statement under consideration leads to the correct technique. you are led to the abstraction question "How can I show that a function (namely f) is continuous at a point (namely x)?" As usual.yl < 8'.
you should attempt to construct 8. When the construction method is used. with Ix . Why not try the value of y' that was obtained in the backward process? Fortunately. it must be that 8 > O.yl < 8'.y'l < 8 and 8 = 8'. To check i f the current guess of 8 = 8' is correct. Perhaps 8 = 8' tor perhaps 8 = c).y' I < 8. From B2. in this CBse: 84: If(x) . it is advisable to turn to the forward process to produce the desired object. you can specialize Al to y'. for 8 = 8': 83: For all y with Ix .y'l < 8'). the number c > 0 has not been used.f(y')1 < c'. nor has the statement: Al: For all y clxyl. and therefore.e. with Ix . the only question is which value of y to specialize to.114 APPENDIX B quantifier "there is" and hence use the construction method to produce the desired value of 8. it is necessary to see if the certain property in 82 holds for 8' and also if the something happens. obta in ing : . If(x) .. If(x) .fey) I s Upon recognlzlng the quantifier "for all" in the forward process. perhaps you should consider using specialization. Working forward from the hypothesis that there are real numbers c > 0 and 8' > 0 for which something happens. so Ix . Ix . Why not "guess" that 8 = 8' and see if it is correct? If not. the hypothesis still has some unused information. To see if B~ is true.f(y)1 < c'. y' was chosen to have the certain property in Al (i. The choose method then requires you to show that the something happens.yl < 8. say y'. Looking at B~. Thus. but since 8 = 8' and 8' > 0. the certain property is that 8 > 0. Specifically. you should recognize the quantifier "for all" in the backward process and you should use the choose method to select a particular value of y. it remains only to verify that. then maybe you will discover what the proper value for 5 should be.
so the idea is to see if A2 can be made to look more like B~.f(y')1 s c8. for this value of 8. it wi) 1 follow that Ix . If only you could apply specialization to AI. Unfortunately.y' I. the certain property in B2 holds and also whether the something happens. For instance. Clearly 8 has been chosen to be > C and thus has the certain property.y'l < 8' because 8 < 8'. you could reach the desired conclusion that E4 is true. e'/c}.y' I < 8'. in this case.e. A2 can be rewritten as: A3: If(x) .y' I < 8. . Thus.fey') 1$ clx . This means that the original guess of S 8' was incorrect. This time. since y' was chosen so that Ix . As before. B4 would be obta ined if c 8 were known to be Unfortunately. that E2 is true.y' I < 8. but you do not know that c8' < e'. when y' is chosen wi th I x . thus completing the proof. from '112 and the fact that 8 < e'/c.y' I < 8. The previous approach was to specialize III to y'. 115 Recall that the last statement obtained in the backward process was Bb.fey') I s clx . It remains only to verify that the something happens. one would proceed with the choose method to select y' wi th I x . 8 = 8'. since c > 0.y'l < c8 s c(e'/c) = e'. Also. and finally. all you c10 know is that Ix . there is a problem because y' is not known to satisfy the certain property in AI.. however. Can you figure out how to choose 8 so that both c8< e' and so that Al can be specialized to y'? The answer is to choose 0 < 8 < min{8'. or equivalently. since c8 < e'. Perhaps 8 should have been chosen> 0 so that c8 < e'. it will be possible to conclude that If(x) . for then. i.APPENJ)IX B A2: If(x) .y'l < 8 and 0 < 8 < e'/c. = In order to see if the new "guess" of 0 < 8 < e'/c is correct. it is necessary to check if. why not choose C < 8 < e' /c? < e'.e1 is true. it will be possible to specialize Al to y'. and again one is led to showing that E. then you would obtain precisely'll:. that I x .
I f ( x ) . and speaking come only with practice.116 APPENDIX B Proof of Example 18. one has If(x) . reading. it is possible to construct 0 < 8 < min{8'. Furthermore. To show that f is continuous at x. The proof techniques are designed to get you started in the right direction.yl I < 8. writing.y'l < 8 1 and so. let e' > O. for yl with Ix . el/cl. . since 8 < el/c. By the hypothesis. Then. and the proof is complete.f ( YI ) I s c i x .f(y') I < c8s c(el/c) = e'.y I I < c 8. it follows that Ix .// As with any language. by the hypo th e sis.
(c).. (a). Conclusion: p . c.2. Conclusion: The two linear equations ax + by = e and ex + dy = f can be solved for x and y.. d. and Table 1 ensures that ". On the other hand. then it does not matter whether B is true or false. 1.s to exercises Chapter 1 1.. Conclusion: The sum of the first n integers is n(n + 1)/2. e. (e). 1. Conclusion: The minimum value of x(x . 'g) Hypothesis: x is a real number. then "A => B" wou1~ be false. If we want to prove that ftA = 2. q. Conclusion: n 2 is an even integer. is false. (f) Hypothesis: p and q are positive real numbers with fPij .solution. then A should also be false. Hypothesis: n is a positive integer. The reason is that if. (p + q)/2. b. and hypotenuse of length z. and f are real numbers that satisfy the property that ad . => S" is true. Hypothesis: n is an even integer. Hypothesis: r is a real number and satisfies r2 Conclusion: r is an irrational number. Hypothesis: a.2. o. 117 . i f A is true and S is false.1.1) is at least 1/4.be . => S" is true and we know that S is false. Conclusion: The triangle XYZ is isosceles.. and (f) are statements. (a) Hypothesis: The (b) (c) (d) (e) right triangle XYZ with sides of lengths x and y. has an area of z2/4.
F = fal se) A T T T T F T T T F F B C (B=>C) T F T T F T T A=>(B=>C) T F T T F F T T F T F T F F F T T F F T T T T fb) (T = true. (a) (T = true. because 'A is false. because A and B are both true. False if x • 3 because then A is true.118 SOLUTIONS 1. the hypothesis. 1. (d) True if x ~ 3 because then A is f. F A T T T = fal se) (A=>B) T T T T F T F B C T F T F (A=>B)=>C T F T T F T F T F F F T F F  F F F T T T T T T F T F Chapter 2 2. The forward process is a process that makes specific use of the information contained in statement 'A. (c) True. because B is true.4. (b) True. The truth of ~ does not matter. (a) True. The backward process is a process that tries to find a chain . 5.1.alse.
backward process is the that is directed toward of statements. is true. (b) and (c) are not valid because they use the specific . abstractly. deriving a sequence of new statements with the property that if the sequence of new statements is true. If this occurs. such as the possibility of more than one abstraction question. (a) is correct because it asks. ~ second difficulty is that there may be more than one answer to the abstraction question and some of these answers may not lead to the completion of the proof. The abstraction process will continue in this manner until either we obtain the statement A. 119 directly to the fact that conclusion.SOLUTIONS of statements leading statement 8. then B is true. and derives from it a sequence of new statements that are true as a result of statement A being true. There can be difficulties in the backward process. Then we use the abstraction process to ask and answer the abstraction questions. The forward process begins with the statement A. or until we can no longer ask and/or answer the abstraction question. the specific 2. Fvery new statement der ived from A should be directed toward linking up with the last statement derived in the backward process. the Incorporated in the abstraction process generating this chain time.2. which is assumed to be true. then we must use the information in statement A to help us choose the appropriate one.3. one at a Specifically. the backward process works by beginning with the statement B that we are trying to conclude is true. how the statement E can be proven true. (c) is incorrect because it uses notation given in the problem. The last statement of the backward process acts as the guiding light in the forward process. just as the last statement in the forward process helps us to choose the right abstraction question and the right abstraction answer. 2.
Show that they are both congruent to a third trian9le. Show that their ratio is one. Statement (d) is an incorrect question for this problem.5.4. Show that they are both perpendicular to a third line. (a) How can I show that two lines are parallel? How can I show that two lines do not intersect? How can I show that two lines tangent to a circle are parallel? How can I show that two tangent lines passing through the endpoints of the diameter of a circle are parallel? (b) How can I show that a function is continuous? How can I show that the sum of two functions is continuous? How can J show that the sum of two continuous functions is continuous? (c) How can I show that an integer is even? How can I show that a number is an even integer? How can I show that an integer is not odd? How can I show that an integer squared is even? (d) 1I0w can I show that the solution to a second de9ree polynomial is a specific integer? How How can I solve a quadratic equation? can I show that two quadratic equations share a common root? 2. . 2. Show that corresponding sidesidesides are equal. Show that it is a square.120 SOLUTIONS notation in the problem. Show that they are each parallel to a third line. Show that it is a parallelogram with one right angle. Show that one is s the other and vice versa. (b) Show that correspondin9 sideangIesides are equal. Show that they are both equal to a third number. Show that their corresponding an9lesidean9Ies are equal. Show that their corresponding equations are identical or have no common solution. Cd) Show that it has three 90 degree angles. Show that they have equal slopes. Cc) Show that they lie in the same plane and do not intersect. (a) Show that their difference equals zero.
The triangle is isosceles. (b) (1) How can I show that a triangle is equilateral? (2) Show that three sides have equal length or. :» • The circle has a radius of 5. (3) Show that the solution b/a is positive. or show that angle R = angle S = angle T. (d) is not val id because ·x " 5" is not stated in the hypothesis. show that three angles are equal. The circle crosses the y axis at (0. Both of these statements follow by working forward from the hypothesis that x + y = 0 (so x = y) .3) < . (b) x/z = 1/Y2.2 > O. An abstraction question associated with the conclusion is "How can I show that a real number (namely x) is O?" To show that x = 0. Angle X has 45 degrees.SOLUTIONS 2. To see that x :!Ii 0. Working forward from the hypothesis immediately establishes that x ~ O. cos (X) = 1/{2.2)(x . (a) 121 (1) How can I show that the solution to a quadratic equation is positive? (2) Show that the quadratic formula gives a positive solution. (d) and OV Angle U = angle W = angle V = 60 cegrees. it will be established that both x :!Ii 0 and x ~ O. x2 .6x + ~ + y2 _ 4y + A = 25. x 2 + 3x . ( c) The circle has its center at (3. and so.9. (a) (x . Ca) Outline of proof.6. it will be shown that x = y and y:!li C. x(x .7. 2. 2. "if x = 5.2.8. (3) Show that R'f = ~ =~. = :ow = OW.5. The height of the triangle is Y3i2 times the length of a side. then it will not be possible to divide by x . 6) (0. :>. 2).1) < o.
122 SOLUTIONS and y l!: 0 (so y s 0) • It remains only to show that y = 0. it will first be shown that x l!: 0 . The latter is accomplished by showing that x = y anc that y s r. (a) The number to the left of each 1 ine in the diagram below indicates which rule is being used. Similarly. Specifically. which follows by working forward from the fact that x = 0 and the hypothesis that x + y = o.10.which is given in the hypothesis) ana x s o. To see that both x = 0 and y = 0. because the hypothesis states that y ~ r. Thus. To see that x = y. Finally.. . J. o = x + y = 0 + y = y. to see that y = 0.11 2.S~ ssssssss J/~ sst st~ stst stststst 1//31 ~ ts st ssss sssst sstsst 1\ 1\ (b) The number to the left of each line in the diagram below indicates which rule is being used. one can use the fact that x = 0 and the hypothesis that x + y = r to reach the desired conc1usion. y s r. observe that the hypothesis states that x + y = O. x = o. (b) Proof.
A: B: The right triangle XYZ is isosceles.SOLUTIONS (c) A : s AI: ss A 2: ssss B1: sssst B tst A s AI: tst A. Outline of proof.: from the z2 = x2 + y2 From the statement A2. if we square both sides of the equation and perform some algebraic manipulations.11. The area of triangle XYZ is z2/ 4• An abstraction question for E is "How can I show that the area of a triangle is equal to a particuler value?" Cne answer is to use the formula for computing the area of a triang Ie to show that El: z2/ 4 = xy/2 that Working forwar~ from the hypothesis triangle XYZ is isosceles.}: tsttst A:I· tsst 114: tssttsst B1: tsssst B : ttst given rule 1 rule 1 rule 4 rule 3 given frOM part c rul e 1 rule '} rule 1 rule 2 rule 3 In this problem one has: 123 (d) 2. one has: A] : x = y x . we obtain A4: (x y)2 =0 A5: x2  2xy + y2 : 0 . Pythagorean theorem.y 0 12: Since XYZ is a right triangle. we obtain ~.
by the hypothesis. Of course ~ = ~.. one obtains z2 = 2xy.12. z2 = x 2 + y2 and on substituting z2 for x 2 + y2. by the !:!'ypothesis.// Proof. To see that R'S = JTIf. the hypothesiS that SU is a perpendicular bisector of RT ensures (by the sideangIeside theorem) that triangle RSU is congruent to triar:!9le fliT. or equivalently. one can easily conclude that ~ = 21m = in} + tJ'f' = ~. specifically. work forward from the ~pothesis to obtain the conclusion that RS = 21m = 1m + O'T = rrf.124 SOLUTIONS A6: X 2 + Y2 = 2xy Substituting ". angle RUS = angle SUT = 90 degrees. ~ = ~ = IT''f. To equilateral.// 2. one can work forward from the hypothesis to establish that triangle RSU is congruent to triangle SUT..7: A: into A6 yields z2 = 2xy 4 finally yields the ~ividing both sides by desired result: A8: z2 /~ = xy/2 Proof. RS . or equivalently. Performing some algebraic manipulations. 1m = IJ'f and. ~:ence. Hence z2/4 is the area of the triangle.. to see that ~ =~. Finally. see that triangle RST is it will actually be shown that ~ = ~ =~. By the Pythagorean theorem. one knows that x = y.y/2.y = o. in addition. z2/ A = y. To see that ~ = ~. ST. . Specifically. that x . From the formula for the area of a right triangle. ~n abstraction question associated with the conclusion is "How can I show that a triangle is equilateral?" One answer is to show that all three sides have equal leng_th. so the sideangIeside theorem states that the two triangles are congruent. Outline of proof. one obtains x2 + y2 = 2xy. To that end. From the hypothesis. the area of XYZ = xy/2.
(b) Spec. ( a) Ab s " en : Ans: Abs. Ans: Ic) Abs. then p :: 1 or p = n. and can only be divided by itself and one. On: Ans: Spec. en: Ans: Abs.ans: (e) Abs. where the integer in the denominator is not equal to zero. Ans: Abs. Cn: Abs. (d) Spec.SOLUTIONS Chapter 3 125 3" 1. Show that sit • plq. Show that n 2 • 2k + 1. How can I show that an integer divides another integer? Show that the second integer equals the product of the first integer with another integer. Ans: Spec. Ans: Spec • . respectively.1) + n + (n + 1) • 9k.abs. . Ans: How can I show that an integer is odd? Show that the integer equals two times some integer plus one. en: . Abs. How can I show that two pairs of real numbers are equal? Show that the first and second elements of one pair of real numbers are equal to the first and second elements of the other pair. greater than I. Show that Xj = x 2 and Y1 = Y2" How can I Show tnat an integer is pr ime? integer is Show that the positive. where p and q are integers and q ~ o. Ans: ·Abs. Show that n > 1 and. Show that the following expression holds true for some integer3 k: 3 3 (n . if p is an integer that divides n. How can I show that a real number is a rational number? Show that the real number is equal to the ratio of two integers. and that the first integer is not equal to zero.
and angle(R) • angle(S) = angle(T).3. b ~ o. . (a) A : n is an odd integer.6. where p and q are integers with q ~ O.128 SOLUTIONS 3. "~ 3. where a and bare integers with b ~ o. F = false) (a) Truth table for converse of A B implies 8. c are integers for which a divides band b divides c. t • alb. AI: x/z = y/z (or. where p and q are two integers and. x = y). b. where k is an integer. (b) A : sand t are rational numbers with t ~ o. AI: n '" 2k + 1. AI: b = pa and c '" qb. (A is the hypothesis and Al is the result of working forward one step). (c) A : Triangle RST is equilateral. AI: s '" p/q. A : a. Also. (d) ( e) A : sin(X) '" cos(X). a ~ 0." A T T F F B T F T NOT A F F T T NOT B "NOT A Implies NOT 8" F T F T T T F T F Note: Converse and inverse of "A implies B" are . (T '" true." "A Impl ies T e" "B Impl ies A" T T F T T T F F T F T F F T T (b) Truth table for inverse of "A implies B. AI: 1m = S"f = ~.2.
(e) Truth table for "A OR B.SOLUTIONS 127 equivalent. " A T T T T F F F B "'It OR B" T T F T F 'd) Truth table for "A AND B." NOT A T T F F F F B T F T F (NOT A) OF B T F T T "A Implies B" T F T T T T Note:"" implies 8" and "NOT A OR B" are equivalent. Both are false when 'It is true and B is false. . Both are false if and only if A is false and B is true." B T T F F T F T F T F F F (e) Truth table for "A AND NOT E" B NOT B F T F T "A AND NOT B" F T F F T T F F T F T F (f) Truth table for "(NOT A) OR B.
then r is a real number such that r2 = 2.128 SOLUTIONS if n2 if n2 if is n is an even integer. if r is rational. then t is an angle for which sin(t) = cos(t) and o < t < x. n is an odd integer. then ABCD is not a Farallelogram with one right angle. then r is a real number such that r2 f 2. (a) Converse Inverse Contraposi tive ( b) Converse Inverse Contraposi tive (c) Conv~rse if r is not rational. if the quadrilateral ~BCD is a rectangle. Outline of proof. n is an integer for which is not even.5. then the quadrilateral ABCI:' is a parallelogram with one right angle. if the quadrilateral A~CI:' is not a rectangle. and sin ( t) f co s ( t) • Inverse Contraposi tive (d) Conver se Inverse Contraposi tive 3. if t is an angle for which sin(t) f cos(t) and o < t < x. then t is an angle for which 0 < t < x. if r is a real number such that r2 ~ 2. We now turn toward the hypothesis and work . then n 2 odd. if the quadrilateral ABCI:' is not a parallelogram with one right angle. then t f x/4. then n is odd. then is even. then r is rational. if t f x/4. ? 4. The forwardbackward method gives rise to the abstraction question "How can I show an integer (namely n 2 ) is odd?" The definition for an odd integer is used to answer the abstraction question which means that we have to show that n 2 = 2k + 1 for some inte~er k. if t = x/4. then the quadrilateral ABCD is not a rectangle.
Hence it has been shown that n 2 can be expressed as 2 times some integer plus 1. Proof./ / 3. + 1 for some integer m. there is an integer m for which n = 2m + 1.SOLUTIONS forward to reach the desired conclusion. one is led to the abstraction question "Pow can I show an integer (namely mn) is odd?" By using the definition of an odd integer. We now turn to the forward process to find out which integer. and therefore. one has m = 2b + 1 for some integer b. n2 = (2m + 1)2 = 4m 2 + 4m + 1 + 2m) + 1 = 2(2m 2 Thus the desired value for k is (2m 2 + 2m). . Since m is an odd integer and n is an odd' integer. n 128 = 2m Since n is an odd integer. mn = (2b + 1) (2c + 1) = 4bc = 2(b + 2b + 2c + 1 + c + 2bc) + 1. Since n is an odd integer. which completes the proof. and since m is an integer. Proceeding by the forwardbackward method. n2 = (2m = 4m 2 + 1)2 + 4m + 1 2 = 2(2m + 2m) + 1. by definition. 'Iherefore. Outline of proof. this question can be answered by showing that mn can be expressed as 2 times some integer plus 1. 'Iherefore.6. and n = 2c + 1 for integer c. and hence n 2 is an odd integer. (2m 2 + 2m) is an integer.
2b + 1 and n . one is led to the abstraction question "How can I show that one statement (namely A) implies another statement (namely C)?" According to Table I.. The hypothesis that "B implies C" and "C implies A." together with Exercise 3.7. we find that since "~ implies S" is true and A is true. Outline of proof. Outline of proof. "A impl ies B" is true. thus completing the proof. odd Proof. Finally. Since m and n are m . that other integer being (b + c + 2bc). To conclude that "A implies C" is true.11 3. 2c + 1. since "S implies C" is true. we assume A to be true and we try to prove that C is also true. and this completes the proof. ensures . Hence mn is an odd integer.8. and "B implies A" is true. and then reach the conclusion that the statement to the right of the word "implies" is true. integers.130 SOLUTIONS Thus it has been shown that mn can be written as 2 times another integer plus 1. the answer is to assume that the statement to the left of the word "implies" is true. (2b + = 1) (2c + 1) 2(b + c + 2bc) + 1. Sy the hypothesis. In this specific case.. B must be true.7. Working forward from the information given in the hypothesis. Thus mn . and "B implies C" is true. Using the forwardbackward method. assume that A is true. Proof. Since S is true. C must also be true. it must be that C is true.11 3. Hence the proof is complete. where band care integers.. The forwardbackward method gives rise to the abstraction question "How can I show that two statements (namely A and B) are equivalent?" Definition 8 indicates that one must show that "A implies B" is true (which is given in the hypothesis). so B must be true.
11 3.SOLUTIONS that "B implies A" is true. one obtains that w2 = 2uv. In order to show that A is equivalent to B. equivalent to C is similar. A => D. B => C. and by applying algebraic manipulations. one has u 2 + v 2 = 2uv. C => D.7. The proof that A is equivalent to C is omitted. we know that sin(U) = Yu/2v.e hypothesis. On factoring." follows from the hypothesis that "8 implies C" and "C implies A. Furthermore. and then taking the square root of both sides of the equation. completing the proof.7. by the Pythagorean theorem. For instance. Working forward fromtb." it follows that "A implies D. and D => A. it follows that u . (a) Outline .9." since the hypothesis states that "A implies B.10. The six proofs are: A => B. by Exercise 3. namely. (a) We will require six proofs to show that A is equivalent to each of the three alternatives. P. A => S." "B implies C." together with Exercise 3. By the definition of sine. u2 + v 2 = w2• On substituting for w2 . which in this specific case means that we have to show that u = v. and D => A. (c) If the four statements in part (b) above are true. we have to show that two of its sides are equal. one need only show that "B implies A.. 3.2uv + v 2 = o. sin(U) = ulw. so ru." of proof.72V = ulw. or." and"C implies D. and consequen tl y. " => C.v = 0. u = v. The forwardbackward method gives rise to the abstraction question "How can I show that a triangle is isosceles?" Using the definition of an isosceles triangle. 131 The proof that A is Proof. C => A. then you can show that A is equivalent to any of the alternatives by using Exercise 3. since "A implies B. (b) We would require only four proofs. one already knows that "D impl ies A." Also. to show that A is equivalent to D.7. => A. u 2 ." The fact that "B impl ies f.
By applying algebraic manipulations.. one obtains w = Y2uv. Hence the hypothesis of Example 3 holds for the current right triangle UVW. or. and consequently the triangle is isosceles. r • u. sin(U) = YU72V. and consequently the triangle must be isosceles.// .// (b) Outline of proof. Then it must be shown that uv/2 = w2/4. thus YIi12V. u/w.. u/2v • u 2 /w2... Observe also that triangle UVW is a right triangle. one obtains uv/2 • w2/4. and since sin(U). and from the definition of sine. x = u. Observe also that triangle UVW is a right triangle. Yu/2v = u/J!. Hence. s = v. and t • w. one obtains u . Since sin{U)" ~ and also sin(U) • u/w.. w· r2uv.// (c) Outline of proof. v. sin (U) . or. Y2uv but. and finally that uv/2" w2/4. y . it is necessary to match up the notation. Proof. by Pythagorean theorem. w2 • 2uv. u/w... v. w2 • 2uv. or.Lso u/2v = u 2/w2. Specifically. u/w. and z . sin(U) • u/w. By the hypothesis. By the hypothesis. and on substituting 2uv for w2. it is necessary to match up the notation. Now. u/w. one has ~ = u/w.132 SOLUTIONS Proof. By algebraic manipulations. or w2 = 2uv. In order to verify the hypothesis of Example 3 for the current triangle UVW. Specifically. or. Proof. and then performing some algebraic manipulations. Working forward from the current hypothesis that sin(U) • yu/2v. thus one has yu/2v = u/w. w'2 = u 2 + v 2. and from the definition of sine. In order to verify the hypothesis of Example 1 for the current triangle UVW. the hypothesis of Example 1 holds for the current right triangle UVW. sin(U) = fU72V. w. Thus it must be shown that w . as in the proof of part (b). Yii12V .
::I exactly n complex roots.5x/2 + 3/2 = O. From the formula. (c) At a party of n people. rn ~ p(r ) == 1 ••• = p(rn ) = O. we find that x = 1 or .(101 4. we will find values of x such that quad ratic x2 5x/2 + 3/2 = O. (a) A triangle )(YZ is isosceles if ::I two sides !) their lengths are equal.sl < 0. Using the quadratic formula for x 2 . Ic) (d) (e) l' parallel to sin(t) r < t < X/2 between x and y = cos(t) Ir . (d) For a polynomial of degree n. Now if we substitute these values of x into x 2 .SOLUTIONS Chapter 4 133 4. ::I an angle t' ~ tangent of t' is greater than tangent of t. Obj ect Certain property over 20.3.we see that the quadratic equation is satisfied. Outline construction of proof. 4.5x/2 + 3/2 = 0.000 feet none through P not on 1 Something happens (a) mountain in Himalyas integer x line l' angle t rational numbers r.2. say P(x).1.s taller than every other mountain in the world ( b) =0 x2  5x/2 + 3/2 R. x equals I or 3/2. (b) Given an angle t. Using the method. the values of x are: (5/2 +"'25/4 . ::I at least two people ~ they have the same number of fr iends. r 1 ' • . (a) Proof. or.12/2)/2 = (5/2 ~ 1/2)/2.
134
SOLUTIONS x = 3/2. Thus, there exists an integer, namely x = 1, such that x 2  5x/2 + 3/2 = o. The integer is unique.11 Proo f • Us i ng the quad ra t ic formula on x 2  5x/2 + 312 = 0, we find that x = 1 or x = 3/2. Thus there exists a real number for x such that x 2  Sx/2 + 3/2 :: O. The real number x is not unique.11
( b)
4.4. Outline of proof. The forwardbackward method 9ives rise to the abstraction question "How can I show that an integer (namely a) divides another integer (namely c)?" By Definition 1, one answer is to show that there is an integer k such that e = ak. The appearance of the quantifier "there is" suggests turning to the forward process to construct the ~esired k. From the hypothesis that alb and blc, and by Definition 1, there are integers p and q such that b = ap and c = bq. Therefore, it follows that c = bq = Cap)q = a(pq), and the desired integer k is k = pq. Proof. Since alb and blc, by c'lefinition, there are integers p and q for which b = ap and c = bq. Put then c = bq = lap)q = a(pq). Thus, alc.11 4.5. Outline of proof. The forwardbackward method gives rise to the abstraction question "How can r show that a real number (namely sit) is rational?" By Definition 11, one answer is to show that there are integers p and g with q ~ 0 such that sit = p/q. The appearance of the quantifier "there are" suggests turning to the forward process to construct the ~esired p and
q.
From the hypothesis that s and t are rational numbers, and by Definition 11, there are integers a, b , c, and d with b ~ 0 and and t such that c/d. s = alb d ~ 0 and thus Furthermore, since t ~ 0, c ~ 0, (alb) I (c/d) :: ad/bc. Hence sit So be ~ o. the ~esired integers p and q are p = ad and
SOLUTIONS g g
135
~
~
= bc.
Observe that since b ~ 0 and c 0, and furthermore, sIt ~ pIg.
0,
Proof. Since sand t are rational, by definition, there are integers a, b, c, and d with b ~ 0 and d ~ 0 such that s = alb and t = c/d. Furthermore, since t ~ 0, c ~ o. Constructing p = ad and q = bc, and noting that g ~ 0, one has sIt = ta/b)/(c/d) ad/bc = pIg, and hence sIt is rational.11
Chapter 5
5.1. (a) Object: real number x Certain property: none Something happens: f(x) s f(x·) (b) Object: element x Certain property: x in S Something happens: g(x) ! f(x) (c) Object: element x Certain property: x in S Something happens: x s u (d) Object: real number & Certain property: & > 0 Something happens: there is x in S such that x > u  &
te)
Object: elements x and y, and real numbers t Certain property: x and y in C, and os t s 1 Something happens: tx + (l t) y is an element of C

(f) Object: real numbers x, y, and t Certain property: Cst s 1 Something happens: f(tx + n  t)y) s tf(x) + (1  t)f(y)
136
SOLUTIONS
(g) Object: real number Certain property: E
8 >
(h)
Something happens: there is a real number 0 such that, for all real numbers y with I x  yl < 8, I f(x)  fey) I < E
>0
E
Object: real number E Certa in prorerty: E > 0 Something happens: there is an integer k' such that for all integers k > k', Ixkxl < E
It will be shown that f(x') ~ f(x·). Let x' be an element in S. It will be shown that g(x') i:!: f(x'). Let x' be an element in S. It will be shown that x' ~ u. Let E' be a real number such that E' > O. It will be shown that there is an x in S such that x > u  E'. Let x', y' be elements in C, and let t' be a real number between 0 and 1. It will be shown that text + ,I  t')y' is an element of C. Let x', y', t' be real numbers such that C ~ t' ~ 1. It will be shown that f(t'x' + (lt')y') ~ t'f(x') + (lt')f(y'). Let E' be a real number such that E' > O. It will be shown that there is a real number 8 > 0 such that, for all real numbers y with Ix  yl < 8, If(x)  f(y)1 < E'. Let E' be a real number such that E' > O. It will be shown that there is an integer k' such that, for all integers k > k', I xk  x I < E'.
5.2. (a) Let x' be a real number.
(b) (c) (d)
(e)
(f) (g)
(h)
5.~.
When using the choose method to show that "for all objects with a certain property, something happens," you would choose one particular object that does have the certain prorerty. You would then work forward from the certain property to reach the conclusion that the something happens. This is precisely the same as using the forwardbackward method to show that "if x is an object with certain property, then the something
Outline of proof. 5. for which it must be shown that If(x') . (a) First you would use the construction method to construct H > 0.f(y')1 < t'. The appearance of the quantifier wfor all w in the backward process suggests using the choose method to choose an element t' in T for which it must be shown that 82: t' is in S. then use the construction method to construct 0 > 0. 5. and backward from the something that happens. sin(2t) = 2sin(t)cos(t).5. . (c) Y two nonnegative real numbers p and q. (b) First you would use the choose method and choose H' > 0. and then use the choose method to choose x' in S for which it must be shown that I x' I s fII. In other words. a statement containing the quantifier wfor all w can be converted into an equivalent statement having the form "if • then • • • • w 5. and then use the construction method to construct an x in S for which I x I > M'.4.y' I < 0.6. 3 a rational number r ~ x < r < y.SOLUTIONS 137 happens. t is in S. (a) 3 a mountain ~ Yother mountains. this one is taller than the others. The forwardbackward method gives rise to the abstraction question WHow can I show that a set (namely T) is a subset of another set (namely S)?W The definition provides the answer that one must show that 81: for all t in T. (b) Y angle t. ffi s (p + q)/2.w whereby you would work forward from the fact that x is an object with the certain property. (d) Y two real numbers x and y with x < y. and then again use the choose method to choose x' and y' with lx' . (c) First you would use the choose method to choose t' > 0.
7.x'). in turn.e. one knows that AI: 1 s tIs 2..2) s 0 and so t' is in S. 2y or or x' "" 2y .x').2) s 0. The key word "for every" in the conclusion suggests uSing the choose method. Outline of proof. y < 0. i. so (t' .)2 . we want x' = 2yl (1 + y) or x' + x'y . Since tl is in T. let Xl > 2 be a real number.1) (t l . (t' . thus establishing B~ and completing the proof.2) sO.11 . and it is also easy to verify that x' = 2y/(1 + y) .II 5. that it satisfies the defining property of T.2) s 0..1) ?! 0 and ( t I . Proof. so (t. and it is easily shown that. is done by showing that t' satisfies the defining property of S. that B~: (t. Ct' . Th us . for this value of y.. Ct' . which it is. and therefore we can construct y = x'/(2 .Xl) = x'/(2 . Working forward from the fact that t' is in T. (t. Let x' > 2. Consequently..3t' + 2 s O.x'). To that end. It will be shown that t' is in S. To construct the desired y < 0.1) (t' .3t' + 2 .138 SOLUTIONS This.)2 . i.e. However. 1 s t' s 2. let t l be an element of T. Proof. Since x' > 2. To show that T ~ S.1) ~ 0 and (t' .x'y y = y(2 .2y/(1 + y). it must also be shown that y < 0.3t' + 2 . x' . Hence the desired y is x'/(2 . since x' > 2..)2 .
then numbers. The forwardbackward method gives rise to the abstraction question "How can I show a function is convex?" Using the definition in Exercise 5. To show that f is convex. Proof.mt'y'+ b (1  = mt'x' t'(mx' + b) + t') (my' + b) .1 (f). t The appearance of the quantifier "for all." suggests using the choose method. f(tx + (1 .mt'y' + b . one must show that B1: For all real numbers x and y. Therefore. 139 Outline of proof. and for all wi th 0 ~ t ~ 1.t')f(y'). we must show that for all real numbers x and y.SOLUTIONS ~. and let t' Let x' and y' satisfy C ~ t' f(t'x' + (1  be real 1. Thus the desired inequality holds. and a real number t' that satisfies 0 ~ t' ~ 1. we choose real numbers x' and y'.mt'y'+ b + bt' + my' . f(tx + (1  t)y) ~ ~ tf(x) + (1  t)f(y).bt' = t' (mx' + b) + n  t') ( my' + b) t'f(x') + (1  t')f(y'). We must show that f(t'x' + n (1 _ t')y') ~ t'f(x') + n .t)y) ~ tf(x) + (1 .t')y') + b + my' . f(t'x' + _ t')y') = m(t'x' + ( 1 _ t')y') + b = mt'x' = mt'x' = + my' .t)f(y). But by the hypothesis. and for all t satisfying C ~ t ~ 1.P. t')y') = m(t'x' + (1 .
and induction is applicable only to integers. (b) It is not possible to use induction when the object is not an integer because showing that P(n) implies pen + 1) may "skip over" many values of the object." Induction is used whenever the object is an integer and the certain property is that of being greater than some initial integer. 6. Induction is used in such cases because it is often easier to show that the something happens for n + 1 given that it happens for n. 6. (e) Not applicable because in this statement.1. rather than to show that it happens for n. something happens. 1 11 I) = 1 = (1 + 1) 1 that P Cn) is true for by 1. (a) Applicable." (c) Applicable.11 Chapter 6 6.3. n is a real number. }~ a result. (b) Not applicable because the statement contains the quantifier "there is" instead of "for all. Replacing n lfll) = (1 + 1)1 . .140 SOLUTIONS = t'f(x') + (1  t')f(y'). (d) Applicable. the statement will not have been proved for such values. First we show n = 1.1. as would be done with the choose method. (a) The choose method is used when the following form appears in statement B: "for every object with a certain property. Thus the inequality holds. Proof. given the certain property. we must show that But this is clear since 1.2.
so 2 5 > 52. it must be shown that for n > 5. by subtracting n from both sides. Thus PIn + 1) is true. Proof.4.. that n 2 > 2n + 1. n = 5. one has In!) + • .SOLUTIONS 141 Now we assume that P In) is true and use that fact to show that Pin + 1) is true. and it can be shown to be true for n > 5 by induction. :2n 2 > (n + 1)2 = n 2 + 2n + 1. This last statement is true for n = 5 (but not for n = 1 or 2). First we show that Pen) is true for But :2 5 = 32 ano 52 = 25. + n(n!) + (n + 1) (n + 1) ! = [lO!) + + n(n!)] + (n + l)(n + 1)1 = [In+ I)! 1) + In+ 1)(n+ I)! = :: (n+ 1)1[1 + (n+ 1)]1 (n (n + I)! (n + 2) . or. = and Starting with the hft side then by using PIn). So pen) PIn + 1) 2n > n2 :2 n +1 > In + 1)2 Starting with the left side of PIn + 1).1 + 2)! .1 In!) + • (n+2)!1 + (n + 1) (n + I)! of Pin + 1). one has: n 2 + 1 = 212n) and > 21n2) To obtain PIn + 1). we then have to prove that Pen + 1) is true.11 . compJ eting the proof.1 I 6.1. using the fact that Pin) is true. Assuming that PIn) is true. Hence we see that it is true for n = 5. So PIn) PIn + 1 ) IO!) + • + n(n!) :: In + I)! .
1 = 0. i.7.5.. (n1)3 (n1) =6k for some integer k. and then. Proof. Thus.1).1 .3n 2 + 3n . to each such subset.n = om for some integer m.// 6.// Checking for n = 1. For a set S with (n + 1) elements we can construct the subsets by listing all of the subsets that are possible using the first n elements.n + 1 n3 _ 3n 2 + 2n n  =n3  (3n 2  3n) . the number of subsets is 2"+1.1) = n 3 . By the induction hypothesis.'n1).n = n(n + 1)/2. Proof. Proof. We have to show that n 3 . we can add the last element of S.e. The statement is true for a set consisting of one element.(n .1 ) + and on adding the two equa tions one obtains 2S = n(n + 1) i •e • . Let S Then S =1 =n + 2 + + n + 1 + (n . we know that 6 divides (n1)3 . n3 .1)3 . there are 21 = 2 subsets.or. there are 2" subsets using n elements.142 SOLUTIONS 6. we find that 1 .6. Assuming that the statement is true for (n . because its subsets are {x} and ~. S 6. the total number of subsets of S is 2 "+ 2" = 2"+1. We will show that for a set with (n + 1) elements. To relate pen) to Pen + 1) we have (n . and an additional 2" subsets are created by ad~ing the last element of S to each of the subsets of n elements. and the statement has been established for (n + 1). say x. Assume that for a set with n elements. and six divides 0 since o =(6) ee). the number of subsets is 2".
you will be unable to do.I)J can be divided by 6. the desired value for the integer m is k + p. prove that it is true for n + 1. Thus. prove that it is true for n . However. of course.3n • 6p for some integer p.SOLUTIONS 143 So n3 .n = n(n .1) must be even. when the original group of (n + 1) horses consists of exactly 2 horses." How do you know that there is a colored horse in the second group? In fact. prove that it is also true for 2n + 3. [en . for then n3  n = rCn  1)3  (n . either n or (n . By the induction hypothesis.1)] + 3(n 2  n) = 6k + 6p = 6Ck + p) . the uncolored horse must also be brown. Consequently.8. 6.9.1)3 .n can be divided by 2. since n 2 .1)] + 3n 2 . Assuming that the statement is true for n.11 . so we have to show that 3n 2 .1) can be divided by 2. and so indeed. Assuming that the statement is true for 2n + 1.11 6. or equivalently.1.In .1) = the product of two consecutive integers.1. (b) Verify that the statement is true for some integer. i t follows that ?n 2 . assuming that the statement is true for n. Then.(n . the second group of n horses will not contain a colored horse! The entire difficulty is caused by the fact that the statement should have been verified for the initial integer n = 2. and that it is also true for n . since all of the colored horses in this (second) group are brown.n = rCn .1)3 .3n. (c) Verify that the statement is true for n = 1.3n can also be divided by 6. n(n . The mistake occurs in the very last sentence where it states that "Then. (a) Verify that the statement is true for the initial value. that n 2 . not n = I! This.
1« SOLUTIONS Chapter 7 7. (d) The quadrilateral ORST must have its area equal to the square of the length of a diagonal. w 7. The forwardbackward method gives rise to the abstraction question How can I show that a set (namely R) is a subset of another set (namely T)1 W One answer is by the definition. and if it is. ~ 5. means. something happens. w Use specialization when you know that wfor all objects with a certain property.y I < 8. use the choose method when you want to show that wfor all objects with a certain property. and if it is. respectively. using that information and the hypothesis. If (x) . and if it does. (a) m must be an integer m 2 y must be an element of S with Iyl < 5.2. then cos(S) > sineS). (b) > mi.? Outline of proof. In other words. the hypothesis says that R is a subset of Sand S is a subset of T which. so we show that ~l: for all r in R. Use specialization when the quantifier wfor allappears in the forward process.in the backward process indicates that the choose method should be us en • So choose an r' in R and. r is in T The appearance of the quantifier wfor all. Use the choose method when the quantifier wfor all w appears in the backward process. it must be shown that r' is in T. then y is an element of the set T.1. then 3 8 > 0 :l Y Y wi th I x . (e) Angle S of triangle RST must be strictly between 0 and ~/4. and if it is. (c) e' must be > 0. something happens. then QRST is a square. and if it is. that . then 7. by definition. Turning to the forward process.f (y) I < &'.
for which it must be shown that P'2: t'x' + t1 . we must show that for all r in R. By working forward from the hypothesis and the above information. by hypothesis. and by the definition. r is in T. and t' yields that t'x' + t1 . Also.t')y' is also in T. one has that r' is in S. y'. A2: for all s in ~. A similar argument shows that t'x' + fl . Outline of proof. 145 and s is in T.~. tx + t1 . and a t' with 0 ~ t' ~ 1. Let r' be in R.SCLUTIONS Al: for all r in R. S is a subset of T. ~ 1. By hypothesis. so r' is in T. !2 will be established by showing that t'x' + '1 . thus completing the proof.t')y' is in both Sand T. Proof. The forwardbackward method gives rise to the abstraction question wHow can J show that a set (namely S intersect T) is convex?W One answer is by the definition. and for all 0 ~ t ~ 1. The appearance of the quantifier for all. Specializing A'2 to r' 'which is in S). Specifically. which is Bl. so r' is in S. To show that R is a subset of T.t)y is in S.t')y' is in S. .t) y is in S intersect T. whereby it must be shown that Bl: for all x and y in S intersect T. and for all 0 s t tx + '1 .in the backward process suggests using the choose method to choose x' and y' in S intersect T. one has that r' is in T. from the hypothesis that S is convex. R is a subset of S.II 7. Specializing Al to r' 'which is in R). r is in S. it follows that Al: for all x and y in S. Specializing this statement to x'. thus completing the proof.t')y' is in S intersect T.
// 7. By the definition in Exercise 5.t)f(y).t)Y):5 ts'f(x) + tl .t)s'f(y). To see that S intersect T is convex.1 If). t'x' + (1 . Outline of proof. let x' and y' be in S intersect T. y'.t')y' is in S intersect T. and let 0:5 t' :5 1. (noting that C :5 t' :5 1) yielc"s A'2: f(t'x' + 11 .t')y' is in T.t')s'f(y'). t' . The appearance of the quantifier "for all" in the backward process suggests using the choose method to choose real numbers Xl and yl.t')y') :5 t'f(x') + (1 . The appearance of the quantifier "for all" in the conclusion indicates that the choose method should be used to choose a real number s' ?!: 0 for which it must be shown that the function s'f is convex. The desired resul t is obtained by working forwarCl from the hypothesis that f is a convex function. and (l :5 t I :5 1 for which it must be shown that 82: s'f(t'x' + (1 . and so S intersect T is convex. one answer is to show that Bl: for all real numbers x and y.t')f(y'). From the hypothesis that S is convex. and all Specializing this statement to x'.t')y' is in S intersect T.1(f).S.146 SOLUTIONS Proof.t')y' is in S. it follows that t'x' + (1 . s'f(tx + (1 . An associated abstraction question is "How can I show that a function (namely s' f) is convex?" Us ing the definition in Exercise 5. and for 0:5t:51. In a similar manner. f(tx + 11 . It will be established that t'x' + (1 . and for all 0:5t:51. one has that t'x' + f1 . one knows that ~1: for all real numbers x and y.t')y') :5 t's'f(x') + (1 . Thus.t)y) :5 tf(x) + 11 .
B3: f(t'x' + II .t')z') s y Turning to the forward process.t')z' is in C. The forwardbackward method gives rise to the abstraction question How can I show that a set (namely C) is convex?. tx + (1 . it must be that AI: for all x and z. f is a convex function.1fe).1(f).t) z) s tf ( x) + all (l  O!5: t S 1.. i. t) f ( z) • . So by the definition in Exercise 5.t)z is in C.t')y') s t's'f(x') + (1 . Since f is a convex function by hypothesis. is done by showing that the point t'x' + (1 . it then follows from the definition that f(t'x' + (l .t')y') s t'f(x') + (1 . one answer is to show th. To show that s'f is convex.Using the definition in Exercise 5. and a o s t' s 1. and let 0 s t' s 1. in turn.at 81: for all x and z in the set C. It will be shown that s'f(t'x' + (1 .6. Using this informat ion and the hypothesis. This.t')f(y'). it must be shown that 82: t'x' + n . by hypothesis. Proof. let x' and y' be real numbers. for all The appearance of the quantifier for allin the backward process suggests using the choose method to choose x' and z' in C. Outline of proof. thus completing the proof.e. and o s t s 1. and for f (tx + (1 .SOLUTIONS 147 The desired statement 82 can be obtained by multiplying both sides of the inequality in A2 by the nonnegative number s'./ / 7.t')s'f(y·).t')z' satisfies the defining property of C. The desired result is obtained by multiplying both sides of this inequality by the nonnegative number s'.
t')y = y. specialization will be used.t')z') s t'f(x') + (1  t')f(z') s t'y + fl . Specifically. and t' (noting that () s t' :! 1)..7. one . and hence B2 is true. I. for all there is an x in S such that t. Hence f(x') s y and f(z') s y. z'. and. t'x' + (1 .t')z') s t'f(x') + (1 . f is a convex function and so f(t'x'+ (1  t')z') s t'f(x') + S (1  t')f(z') tty + (1  t')y =y and consequently. f(t'x' + n .t')f(z') Finally. let x' and z' be elements of C. The forwardbackward method gives rise to the abstraction question How can I show that a number (namely 1) is a least upper for a set (namely S)?. to obtain B2. thus. By the hypothesis.1 fe). f(x') s y and f ( z') s y.Using the bound definition in Exercise 5. Outline of proof.in the forward process.II 7. x >1 t is an upper bound for S. To show that the first part of Bl is true. So. Proof: To show that C is convex. from A2 . so A2: f(t'x' + (1 .e. they satisfy the defining property of C. we will use the fact that x' and z' are in the set C. Al can be specialized to x'. and let t' satisfy 0 S t' S 1. one answer is to show that Bl: > 0. completing the proof.148 SOLUTIONS Recognizing the quantifier for all.t')z' is in C.
lin > 1 . Returning to Bl. for then 1 . we now make use of the fact that x is in S. x s 1. for then one can construct x = 1 . there is an integer ~ n ~ 2 Since x is in S.in the backward process.lin. Again. thus B3 is true. Hence. The desired n is any integer > lie. there is an integer n ::: 2 such that x = 1 .c. Turning to the forward process. To see that lis an upper bound for S. To establish B3. and therefore 1 that an .Again.lin. there is an x in S such x > 1 . 2 for n::: 2. let x be in S.lin}.lin. the definition in Exercise S. So x s 1. Recognizing the quantifier for all. the desired x in S will be constructed by finding an integer n ::: 2 for which 1 . the choose method is used to select e > 0 for which it must be shown that B5: there is an x in S such that x > 1 . and since x • 1 . To do so.lin > 1 . Proof. there is an integer n which x • 1 .e.SOLUTIONS 149 is led to the abstraction question How can I show that a number (namely 1) is an upper bound for a set (namely 8)?.e. thus completing the proof. the choose method is used to choose an element x in 8 for which it must be shown that B~: x S 1. it is important to observe that the set 8 can be written as S • {real numbers x: with x • 1 . one must still show that B4: for all e > 0.l(c) can be used to provide the answer that one must show that B2: for all x in 8.lin s l.e.
e. Then x • 1 . (a) Assume: t . we start with the assumptions }\: n is an integer for which n 2 is even. NOT B: n is not even. 8. i.1.e..lin satisfies x > 1 . An element x in S will be produced for which x > 1 . Specifically. We work forward from these assumptions using the definition of an odd integer to reach the contradiction that n 2 is odd. ~ are linearly dependent. Then n2 • (2k + 1) 2 4k2 + 4k + 1 = 2(2k 2 + 2k) + 1 .11 Chapter 8 8. and that 24 divides R.3. let n > lie. f is unboun~ed above. (d) The lines i and it do not intersect. and the rows of. and n are three consecutive integers. let e > O. by 8.150 SOLUTIONS is an upper bound for S. (c) Assume f and 9 are two functions such that 9 ! f. n is odd. (a) The number of primes is not finite. (b) Assume: Matrix ~ is not singular. To complete the proof. (b) The set of real numbers is not bounded. and 9 is bo und ed above.2. (c) The positive integer p cannot be divided any positive integer other than 1 and p.e. there exists an integer k such that n = 2k + 1. (e) The real number x is not ! 5. 2 + m2 + n2 + 1. To use the contradiction method. Outline of proof. Thus. m.
Now it follows that angle ACB has 90 degrees (since it is an angle in a semicircle). which is impossible. Starting with this assumption. '!'hen the desired contradiction is that the hypotenuse of a right triangle is shorter than one side of the triangle. n = 2k + 1. We proceed by assuming that there is a chord of a circle that is longer than its diameter. we should arrive at a contradiction in order to provide a valid proof. contradiction establishes the result. Hence./ / 8. where '!'hus n 2 is odd. contradicting assumpt ion. Assume. and using the properties of a circle. Proof. the initial and hence n 2 is odd. '!'his construction is valid since.SOLUTIONS 2 151 which 2is of the form n = 2p + 1. Let AB be a diameter of the circle. Outline of Proof. there is an integer k such that. to the contrary. n2 = f2k + = ~k2 + = 2 f2k2 l) 2 4k + 1 + 2k) + l. that n is odd and n 2 is even. Let AC be the chord of the circle that is longer than the diameter of the circle. a diameter is a line passing through the center terminating at the perimeter of the circle. by definition.4. . Consequently. Hence the triangle ABC is a right triangle in which AB is the hypotenuse. and this p • 2k + 2k.
Proof. i. it must be that the third angle of the triangle has zero degrees. We construct a diameter that has one of its ends coinciding with one end of the chord. Outline of proof. are not parallel.. and that they are both perpendicular to the same line t. we can conclude that they do intersect at some point in the plane.5. Since the sum of the three angles of a triangle is 180 degrees. we proceed by assuming that no two people have the same number of friends.152 SOLUTIONS Proof. of a circle that is longer than a diameter. and ta. must intersect. and t2 are not parallel. Assume that t. If we join the other ends. which contradicts the initial assumption.// 8. OUtline of proof. The sum of the three angles of a triangle is 180 degrees. Working forward. which are both perpendicular to a third line t in the plane.6. then we obtain a right triangle in which the diameter AB is the hypotenuse. say AC. Assume that there does exist a chord. Hence the three lines t" t2 and t wi 11 form a tr tangle. and form a triangle in which two of the angles are 90 degrees. To use the contradiction method. Since two of the angles are 90 degrees. everybody has a different number of friends. Since there are n . To use contradiction. Hence ~ >~. an obvious contrad ic tion. Hence t. we assume that the two lines ~ and t 2 .e.// 8. we are forced to the conclusion that the third angle has zero degrees. a contradiction.
By doing so..1 friends friend fr iends. including the first one. who has no friend! Proof. 2 . n.1) people.// 8. a contradiction.SOLUTIONS 153 people.2 n 2 (n . Assume. person number 1 has person number 2 hes person number 3 has no 1 2 friends friend fr iends person number n has n .1). we can arrive at a contradiction by showing that no integer satisfies this equation.1 + n = n3 0 + 3n 2 + 3n + 1 n3 _ 6n 2 . we have that (n 1)3 + n3 3 = • tn + 1)3 or or or n 3 3n 2 + 3n . The people at the party can be numbered In such a way that person person person 1 has 2 has n has 0 1 n . and (n + 1) such that the cube of the largest is equal to the sum of the cubes of the remaining two. By the contradiction method. each of whom has a different number of friends. that no two people have the same number of friends. In other words. Outline of proof. we arrive at the contradiction that the last person will be friends with all the other (n .6) . It then follows that the person with (n . Specifically. it can be assumed that there do exist three consecutive numbers (n .1) friends is a friend of the person who has no friends.7. we can list the number of friends each person has according to an increasing sequence. Now.1 friends. to the contrary.
there is a remainder of 1 since. n 2 (n . Hence there is no integer n for wh i c h n 2 (n . + l/r (2)(1)] + 1 Similarly.1). and hence it must be that p > n. and (n + 1) three consecutive integers such that (n . Indeed. Consider the number n! + 1. p ~ n. P ~ 2.. one has that (n! + l)/r = n!/r and Hence n! + 1 has no prime factor between But by assumption. ~ contradiction will be reached by showing that p ~ nand p > n. n. t~en there will be a prime number that is larger than all the other prime numbers.3n 2 + 3n . Let n be such a number. Since n is the largest prime. where 1 < r ~ n. Outline of proof.6) ~ C. there is a remainder of 1. wh 11 e for n > 6.6) = 2. • • • . n 3 .1 + n 3 or or n 3 _ 6n 2 n (n 2 are = n3 =2 = 2. p ~ n. there is no integer n for which n 2 (n . n.// B.154 2 SOLUTIONS No w for n ~ 6..6) > 2. when n! + 1 is divided by r. nl + 1 = rn(n1) . it has a pr ime factor. n 2 (n .1)3 + n3 = (n + 1) 3 • '!ben.6) ~ 0. By the contradiction methoc. wh 11 e fo r n > 6. it can be assumed that the number of primes is finite. Proof.P. Hence. n 2 (n . and let p be any prime number that divides n! + 1. + 3n 2 + 3n + 1 6) Fo r n ~ 6 . We use the fact that p divides n! + 1 to show that p ~ 1. Assume that (n .6) = 2.6) > 2. . When nl + 1 is divided by 2. n (n . If so.
for all x in S. To reach the conclusion. . Assume. (b) Choose and contradiction methods. and T is bounded. Choose Sl in S and conclude that there is no t in T such that s I > t. n! + I has a prime factor p less than or equal to n. Proof. (a) Work fo rwa rd from: Try to conclude: (b) Work forward from: Try (c) Work Try (d) Work Try to conclude: forward from: to conclude: fo rwa rd from: to conclude: n is an odd intege r. in that order. use the contradiction method to assume that there is a t in T such that Sl > t. that there are a finite number of primes. n 2 is an odd integer. But nl + 1 cannot be divided by any number between land n. Then n! + I is not a prime since. (c) Contradiction method is to be used. S is a subset of T. Ixl < M. M is singular.1. n! + I > n.11 8. (a) Construction method is to be used. to the contrary. f (x) = f (y) • x = y. Let n be the largest prime. the rows of matrix M are linearly dependent.9. and try to reach a contrad iction. which contradicts the assumption that n is the largest prime number. Construct an element s in S and show that s is in T.SOLUTIONS 155 Hence the prime factor is 9reater than n. Sis bounded. Hence p > n and we therefore arrive at a contradiction. Assume that there exists an M > 0 such that. Therefore. Chapter 9 9. Now reach a contradiction.
say m. (a) Incorrect. (d) Correct. and since m is odd. Statement (b) is a result of the forward process because yo u can assume that there is a real number t between o and "/4 such that Upon squaring both sides.158 SOLUTIONS 9. It must be shown that c is not odd.. (c) Incorrect. To use the contraposi tive method.~. Here. . observe that m is odd.5. the answer in part (b) resul ts. Outline of proof. and the proof is compl ete. of the 9.c = o. It wi 11 be shown that c is not odd. Hence m2 c = m + m2 = odd + odd = even. (b) Incorrect.2. But from the equation. With the contrapositive method.11 9. 'J. that c is even. This abstraction question also contains specific notation from the problem. because the abstraction process should not be applied to NOT B.e. sin(t) = r cos(t). Assume that there is an odd integer solution. one can write that c = m + m2 • To see that c is even. to the equation n 2 + n .5). and so is also odd (see Exercise 3. m + m2 is even. i. The word "not" in e 1 now suggests proceed ing with the contradiction or contrapositive method.c = O. say m. the abstraction process should not be applied to NOT B. for which it must be shown that el: f(x) F fey). Outline of proof. The appearance quantifier "for all" in the conclusion suggests using the choose method. the contrapositive method will be used. one can assume that there is an odd integer solution. s. But c = m + m2 . because the abstraction question contains specific notation from the problem. to the equation n2 + n . whereby one chooses real numbers x and y with x F y. we need to work forward from NOT B and backward from NOT A. and so c is even. Proof.
~ssume that x and yare real numbers for which f(x) • f(y). If angle R has more than 90 degrees. it then follows that mx + b = my + b.y) • O. it follows that x . and hence. Therefore. It will be shown that x = y. An obtuse angle will be found. say R.f(y). the remaining three angles add up to more than 270 degrees. it can be assumed that f(x) • fey). m(x . Specifically. The appearance of the quantifier ·there is· suggests turning to the forward process to produce the desired obtuse angle. This. mx + b • my + b. is not 90 degrees. means that the remaining angles of the quadrilateral must add up to more than 270 degrees (because the sum of all the angles in RSTU is 360 degrees). and that is the desired obtuse angle. By the contrapositive method. and consequently x = y. one of the remaining three angles is obtuse. Outline of proof. one of them must be greater than 90 degrees. and subtracting the right side from both sides of the equation. in turn. By the hypothesis that m ~ 0. one can conclude that at least one angle of the quadriatera1 is not 90 degrees. If angle R has more than 90 degrees. one reaches the desired conclusion that x • y. Among these three angles that add up to more than 270 degrees. It must be shown that there is an obtuse angle.y = 0. ~ssume that the quadrilateral RSTU is not a rectangle. Proof. one can assume that the quadrilateral RSTU is not a rectangle. Now it must be shown that x· y. or. Proof. work forward from the fact that f(x) = f(y). Otherwise. angle R has less than 90 degrees.6. then it is the desired angle. one obtains m(x . say it is angle R. Otherwise. Using the hypothesis that m ~ 0. But since f(x) .y) = o.SOLUTIONS 157 and accordingly.11 . then it is the desired angle. To reach the desired conclusion.11 9. one of its angles. Working forward.
yl < 8 such that If(x) . such that f(tx + n . The real number x· 1s not a maximum of the function f if there is a real number. (e) The set C of real numbers is not convex if there exist elements x and y in C and there exists a real number t between zero and one such that tx + (1 .x I ~ L (a) 10.t)y is not an element of C. The sequence x 1. or if there exists a real number e > 0 such that for all x in S. (a) There does not exist an element x in the set S such that x is not in T. x 1 .e. ~en g is not ~ f on the set S of real numbers if there exists an element x in S such that g (x) < f(x). (h) Suppose that x. (b) It is not true that for every angle t between 0 and x/2.2. (g) The function f of one variable is not continuous at the point x if there exists real number £ > 0 such that for all real numbers 8 > 0.t)y) > tf(x) + n . there is a y with Ix . ( f) The function f of one variable is not convex if there exist real numbers x and y and t. • • • does not converge to x if 3 a real number e > 0 3 Y integers k'. (d) The real number u is not a least upper bound for a set S of real numbers if u is not an upper bound for S. x such that f(x) > f(x·). • • • are real numbers. x2. sin(t) ~ cos(t) • (c) There does not exist an ·object· with the ·certain property· such that the ·something· does not happen. (b) Suppose that f and 9 are functions of one variable. 3 an integer k > k' such that I xk . (c) The real number u is not an upper bound for a set S of real numbers if there exists an x in S such that x > u.158 SOLUTIONS Chapter 10 10. . x s u .1.fey) I ~ e.t)f(y). x2 . where 0 s t s 1.
SOLUTIONS 159 (d) It is not true that for every ·object· with the ·certain property. Assume that f is a convex function of one variable. y ~ 0. and that either x # 0 or y # o.· the ·somethin9· does not happen. (a) Assume that there is an integer n (b) ( c) (d) ~ 4 such that nJ :s n 2 • Assume A. but this contradicts the fact that y ~ o. x· is a real number. assume that there is a real number y such that fey) < f(x·). Assume A. but this contradicts the fact that x ~ 0. since x + y = 0. Assume that x ~ 0. x + y = 0. f(x) ~ f(x·). and there is a real number 8 > 0 such that for all real numbers x with the property that Ix . If x # 0.4.5. NOT B OR NOT C. 10. and x • y < 0. n is an odd integer or m is an even integer. NOT B AND NOT C. ( b) Forward from: Backward from: (c) Forward from: Backward from: 10. it can be assumed that x ~ 0. A similar ar9ument can be used for the case where y # o. Specifically. Similarly. x + y = 0.x·1 < 8. Finally. So suppose first that x # 0 and hence. (a) Forward from: Backward from: (NOT B) AND (NOT C) NOT A (NOT B) OR (NOT C) NOT A mn is not divisible by 4 and n is divisible by 4. 1 0. A contradiction to the fact that y ~ 0 will be reached by showin9 that y < o. it follows that y • x but x < 0 and so the contradiction has been reached. since x ~ 0. and that either x # 0 or y # O.3. Outline of proof. it must be that x > O. then x > 0 and y = x < 0. When usin9 the contradiction method.// . y ~ 0. Proof. if y # 0. then y > 0.
it follows that m • o. To establish the uniqueness. however. on subtracting yz from both sides. Specifically. x = 2y/(1 + y). y + yz = z + yz and. it will be shown that y • z. But since the hypothesis states the m . we must first construct a real number x for which mx + b = O.y) = o. The first uniqueness method will be used. unique.2. and x then it follows that 2y/(1 and so y + yz = z + yz. since mx + b = 0 = my + b. 0. as 11. According to the second uniqueness method.y. This. was already done in Exercise 5. But + y) = 2z/(1 + z). because then it follows that mx + b • m(b/m) + b = b + b = o.// real number y was To show that y is z satisfy y < 0. or.180 SOLUTIONS Chapter 11 11. A contradiction is reached by showing that m = o. suppose that y and z < 0. Outline of proof. Thus. y = z. z < 0. Specifically. By working forward via algebraic manipulations. . wherein one must first construct the real number y.7. x • 2y/(1 + y). x • 2y/(1 + y) • 2z/(1 + z) or equivalently. it remains only to show the uniqueness by assuming that y and z are both real numbers with y < 0. and x = 2z/(1 + z).// Proof. and also that x. = 2z/(1 + z). On dividing by the nonzero number x . the desired x is b/m. one obtains the desired conclusion that y = z. y. Outline of proof.7. suppose that x and y both satisfy mx + b • 0 and my + b • 0. it follows that m(x . The existence of the established in Exercise 5. desired.1.
But then.11 11. Now suppose that y ~ x and also satisfies my + b = o. it follows that c + di = e + fi. 0. Since either a f" 0 or b f" 0. Then mx + b . for then (a + bi) (c + d i) = ac = 1. b/(a 2 + b 2) (noting that the denominator is not 0 since.be + tbc + ad) i . one obtains: (e + fi)[ Ca + bi)(c + di)] or. 1. Outline of proof. and this contradicts the hypothesis that m f" 0. by the hypothesis.y) = O. [Ce + fi)(a + bi)](c + di) . The issue of existence must be addressed first. It will be shown that c + di equals e + fi..SOLUTIONS 181 Proof.2. my + b. = fe + fi)(l). Working forward and multiplying both sides of the equation above by e + fi. 1. But since x . Then mx + b . suppose that e + fi is another complex number that also satisfies (a + bi) (e + fi) .. and so it is possible to construct the complex number 2c + di in which c = a/(a 2 + b 2) and d = b/(a + b 2). and so m(x . at least one of a or b is not 0). let c = a/(a 2 + b 2) and d . it follows that a 2 + b 2 f" 0. and since (e + fi) (a + bi) . To see the uniqueness.. Proof.bd + tbc + acl) i = (a 2 + b 2) I (a 2 + b 2) + 0 i 1..y ~ 0. it must be that m . let x • b/m (noting that m ~ 0). (a + bi) (c + d i) = ac .. To construct the desired complex number (c + di) that satisfies (a + bi) (c + di) = 1. e + fi. ... m(b/m) + b = O. To construct the number x for which mx + b = 0.
~. we construct the value for k. that there is an integer k such that mn = 4k.5.4. and NOT C. With the either/or method. and that 4 divides n. + 1. it follows that (e + fi)[(a + bi)(c + di)] '" fe + f1)(l) so c + di = e + fi established. and also that 4 divides n. Since m is there is an integer j such that m = 2j Since 4 divides n. Outline of proof. you would only have two statements to work forward from. But since m is . Disadvantages: You cannot work backward because you do not know what the contradiction will be.182 SOLUTIONS To see the uniqueness.ion question "How can I show one integer (namely 4) ~ivides another (namely mn)?" Using the definition leads to the. there is and integer p that n = 4p. one is led to the abstract. By the rules of complex mul tiplication. answer that one must show that there is an integer k such that mn '" 4k. namely. or equivalently. Proof. such = f2j + 1)(4p) = 4'2jp + p) and the desired k is k 2jp + p. In the either/or method you can work backward from the statement (B OR C) • 11. Therefore mn must odd. Working backward. we can assume that n is an even integer and m is and odd integer.// and the un iqueness is 11. suppose that e + fi also satisfies (a + bi) (e + fi) '" 1. It will be shown that 4 divides mn. NOT B. We must conclude that 4 divides mn. By the either/or method. Advantages: You have three statements to work forward from. ~ssume that n is even. m is odd. Turning to the forward process.
x ?: 3. Hence. x 3 + 3x 2 . (a) If x is a real number that satisfies x 3 + 3x 2 . or equivalently. thus Note: This proof could also be done by assuming n is an even integer and m is an odd integer. Then it follows that (x  3) (x + 3) 2 ?: O. (x + 3)2 is positive. the desired integer is k completing the proof. and that 4 does not divide mn.3 ?: 0.3 ?: 0. ( a ) Fo r a 11 sin S . We must show that x ?: 3.27 = (x . X 3 + 3x Since x > 3. (x + 3)2 is positive. x ?: 3. there is an integer j such that m = 2j + 1. so (b) Outline Proof. (c) There is an x with ax :s b that cx :s u.3) (x + 3) 2 ?: O. Now one has 2 . and then concluding that ~ does not divide n. x .9 x .SOLUTIONS 163 odd. Since 4 divides n.9x . Since x > 3. and we must have x .27 ~ 0. mn = (2j + 1) (4p) ::: 4(2jp + p). there is and integer p such that n == 4p. of proof.9x 27 ~ 0 and that x > 3. and x?: 0 such . 11. 7. 11. then x :s 3 or x ?: 3. Assume that x 3 + 3x 2 .9x .27 ~ o and that x > 3. (b) so exceFt you (b) There is an s in S such that s ?: x. s :s x.// = 2jp + p. Thus. ~ccording to the we can ~ssume that either/or method.// (c) The proof is similar to part assume that x < 3. or.6.
applying specialization to 1'. with that Again. let Sl be in S. of proof. Specializing Al to s = s' (which is in S).// 11. and thus the proof is complete. Ey the hypothesis that S is a subset of T. The max/min method can be used to convert the conclusion into the equivalent quantifier statement: . for which it must be shown that Ell: Sl :!: t*. s :!: Outlin~ t*. ~ 0 such x x ~ c. it follows that Sl is in T. it follows Sl:!: t*. The desired conclusion can be obtained by working forward and using specialization. It will be shown that Sl :!: t*. ~ 11. Proof. the hypothesis states that A2: for all t in T.8. But then. To reach the conclusion. ax bx ~c. The appearance of the quantifier "for all" in the backward process suggests using the choose method to choose an Sl in S. The max/min method can be used to convert the conclusion into the equivalent quantifier statement: B: for all s in S. the hypothesis ensures that Sl :!: t*.9. (e) For all x with b (f) Far all x with a ~ ~ ~ b and :!: x u.2 t = Sl (which is in T).184 SOLUTIONS (d) There is an x with ax that cx ~ u. Specifically. it follow that Sl is in T. u. it follows that AI: for all s in S. Also. Outline of proof. t :!: t*. s is in T. since S is a sub se t 0 f T.
since statement B is true for every integer n l!: 4.1. since there is supposed to be one and only one line. (d) Maximin method. Rno R appears in the since the word conclusion. it follows that u'ax' ~ u'b. Proof. ( f) Contrad iction or contraposi tive method. upon multiplying both sides of the inequality uta ~ c by the nonnegative number x'. and also choosing u' ~ 0 with u'a:os: c. it wi 11 be shown that CX'I:!: u'ax' and that u'ax' l!: u'b.SOLUTIONS Bl: for all x O!: 0 with ax u . and for all ub. it follows that cx'~ u'ax'. since there is no apparent form to B. since B has the word maximum R in it. (b) Induction method. and therefore one has that cx' ~ u'ax' l!: u'b. To reach the desired conclusion. (a) Contraposi tive or contrad iction Rno R appears in the since the word conclusion. Specifically. for which it must be shown that cx' ~ u' b. and consequently the proof is complete. It then follows that u'ax' l!: u'b and cx' ~ u'ax'. upon mul tipl ying the inequality aX'1:!: b on both sides by the nonnegative number u'. The appearance of the quantifier Rfor all R in the backward process suggests choosing an x' ~ 0 with ax' ~ b. ( c) ForwardBackward method. Similarly. and the prQof is complete.a: 0 wi th ua:S c. cx ~ ~ 185 b. and let u'~ 0 with uta s c. let X'I:!: 0 with ax' ~ b. 12.11 Chapter 12 method. To accomplish the goal. (e) Uniqueness method. .
since there is no apparent form to B. we would assume that there is an integer n ~ 4 such that nl ~ n 2 . we would choose an integer n I for wh ich n' ?: 4. (c) Converting the statement to the form "if • • • then " one obtains "if n is an integer 2: 4 then nf > n 2 ". also have to show that 4! > 4 2 • (b) Using the choose method. and try to show that (n + 1)1> (n + 1)2. and try to reach a contrad iction.)2. . We would. since the first quantifier in B is "for every. of course. We would therefore assume that n is an integer 2: 4 2 and try to show that nl > n • (d) Using the contradiction method. (a) Using the induction method." (i) Construction method. (h) Choose method. we would assume nJ > n 2 and n 2: 4.2. We would try to show that (n')1 > In." 12.188 SOLUTIONS (g) ForwardBackward method. since the first quantifier in B is "there is.
) there is (there are. etc.E.D.6 40 42 42 30 S ~ V 3 ~ 43 3fi 36 A V Q. 24 24 quod e~at demon6t~andum (which was to be demonstrated) 14 167 .Glossary of Mathematical Symbols Symbol Meaning implies if and only if is an element of subset empty set not for all (for each.) such that and or page 3 > <=> (iff) E ~ . etc.
index
·A implies B," 3, 30 abstraction process, 10 abstraction question, 9 alternative definition, 25
AND, 24, 81
assumption,S, 51, 53, axiom, 28 backward process, 9
~~,
72
certain property, 35, 43 choose method, 40 conclusion, 3 condensed proof, 15, 109 construction method, 34 contradiction method, ~4 contrapositive method, 72 contrapositive statement, 30 converse statement, 30 corollary, 27 definition, 23 defining property (of a set), 41 doing proofs, 3 either/or method, ~~ empty set 42 equivalence, 2r; equivalent definitions, 25 equivalent statements, 2~ existence, 35 existential quantifier, 3d
169
170
INDEX
forwardbackward method, 8, 16 forward process, 9, 12 "for all," for any," "for each, "for every, 34, 40 for which, 36 generalized induction method, hypothesis, ? if and only if, 24 implication, ? implies, ? induction method, 50 infinite list, 40 infinite set, 40 inverse statement, ~o lemma, 27 matching up of notation, 29 mathematical induction, 50 mathematical terminology, 23 mathematics as a language, 1 max/min method, 'e7 membership in a set, ~o negation, 78 negation in quantifiers, 7e "no, not in conclusion, 67 NOT A, :0, 7f' NOT B, 7P NOT B implies NOT A," 30 "NOT A implies NOT B," 30 NOTS of quantifiers, 7P only if, ?O OR, el proof, 1 proof by contradiction, 64
~6
?i1 reading proofs." "there exists. i13 induction method. S9 uniqueness. 1.IJ quantifiers choose method (for all). n "there are. 72 either/or. :4 "hidden." 5 "NOT B implies NOT A. . IP7 subset. f! changing form of. 43 specialization method. 44. :4 contradiction. e induction. ~9 universal. :4 existential. 101. f!7 specialization. 40 construction. 45 something happens. 83 statement.E. 27 Q." 34 truth table. 52 proof techniques choose. 2. 59 special proof techniques." "there is. : theorem. ~l setdefining property. e6 forwardbackward. 5 "A implies F. 50 max/min. 109 set. ~2 set builder notation.D. i12 symbolic notation. 42. :5. 16. ~o construction method (there exists).. 64 contrapositive." 31 . P3 proposi tion. 50 specialization method.INDEX 171 proof machine." 37. ~l set equality.lJO empty.
12 ~4 . working backward. 83 universal quantifier.172 INDEX uniqueness. 9. g working forward. 83 uniqueness method.