A Gentle Introduction To The Art of Mathematics

A Gentle Introduction to the Art of
Mathematics
Joseph Fields
Southern Connecticut State University

ii
Contents
1 Introduction and notation 1

1.1 Basic sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Definitions: Prime numbers . . . . . . . . . . . . . . . . . . . 11
1.3 More scary notation . . . . . . . . . . . . . . . . . . . . . . . 19
1.4 Definitions of elementary number theory . . . . . . . . . . . . 22
1.4.1 Even and odd . . . . . . . . . . . . . . . . . . . . . . . 22
1.4.2 Decimal and base-n notation . . . . . . . . . . . . . . . 23
1.4.3 Divisibility . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.4.4 Floor and ceiling . . . . . . . . . . . . . . . . . . . . . 26
1.4.5 Div and mod . . . . . . . . . . . . . . . . . . . . . . . 27
1.4.6 Binomial coefficients . . . . . . . . . . . . . . . . . . . 28
1.5 Some algorithms . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.6 Rational and irrational numbers . . . . . . . . . . . . . . . . . 45
1.7 Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2 Logic and quantifiers 57

2.1 Predicates and Logical Connectives . . . . . . . . . . . . . . . 57
2.2 Implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.3 Logical equivalences . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4 Two-column proofs . . . . . . . . . . . . . . . . . . . . . . . . 88
2.5 Quantified statements . . . . . . . . . . . . . . . . . . . . . . 92
iii
iv CONTENTS
2.6 Deductive reasoning and Argument forms . . . . . . . . . . . . 100

2.7 Validity of arguments and common errors . . . . . . . . . . . . 109
3 Proof techniques I 117

3.1 Direct proofs of universal statements . . . . . . . . . . . . . . 117
3.2 More direct proofs . . . . . . . . . . . . . . . . . . . . . . . . 129
3.3 Contradiction and contraposition . . . . . . . . . . . . . . . . 133
3.4 Disproofs of universal statements . . . . . . . . . . . . . . . . 139
3.5 By cases and By exhaustion . . . . . . . . . . . . . . . . . . . 144
3.6 Existential statements . . . . . . . . . . . . . . . . . . . . . . 151
4 Sets 159
4.1 Basic notions of set theory . . . . . . . . . . . . . . . . . . . . 159
4.2 Containment . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4.3 Set operations . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
4.4 Venn diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.5 Russell’s Paradox . . . . . . . . . . . . . . . . . . . . . . . . . 188
5 Proof techniques II — Induction 193

5.1 The principle of mathematical induction . . . . . . . . . . . . 193
5.2 Formulas for sums and products . . . . . . . . . . . . . . . . . 202
5.3 Other proofs using PMI . . . . . . . . . . . . . . . . . . . . . 214
5.4 The strong form of mathematical induction . . . . . . . . . . . 222
6 Relations and functions 225

6.1 Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
6.2 Properties of relations . . . . . . . . . . . . . . . . . . . . . . 235
6.3 Equivalence relations . . . . . . . . . . . . . . . . . . . . . . . 244
6.4 Ordering relations . . . . . . . . . . . . . . . . . . . . . . . . . 253
6.5 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
CONTENTS v
6.6 Special functions . . . . . . . . . . . . . . . . . . . . . . . . . 276
7 Proof techniques III — Combinatorics 287

7.1 Counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
7.2 Parity and Counting arguments . . . . . . . . . . . . . . . . . 304
7.3 The pigeonhole principle . . . . . . . . . . . . . . . . . . . . . 319
7.4 The algebra of combinations . . . . . . . . . . . . . . . . . . . 325
8 Cardinality 337
8.1 Equivalent sets . . . . . . . . . . . . . . . . . . . . . . . . . . 337
8.2 Examples of set equivalence . . . . . . . . . . . . . . . . . . . 343
8.3 Cantor’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . 355
8.4 Dominance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
8.5 CH and GCH . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
9 Proof techniques IV — Magic 379

9.1 Morley’s miracle . . . . . . . . . . . . . . . . . . . . . . . . . . 381
9.2 Five steps into the void . . . . . . . . . . . . . . . . . . . . . . 389
9.3 Monge’s circle theorem . . . . . . . . . . . . . . . . . . . . . . 398
9.4 Pythagorean triples . . . . . . . . . . . . . . . . . . . . . . . . 400
References 402
Index 404
vi CONTENTS
List of Figures
1.1 The sieve of Eratosthenes. . . . . . . . . . . . . . . . . . . . . 13

1.2 Pascal’s triangle. . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.3 A small example in pseudocode and as a flowchart . . . . . . . 38
1.4 The division algorithm in flowchart form. . . . . . . . . . . . . 40
1.5 The Euclidean algorithm in flowchart form. . . . . . . . . . . . 42
2.1 A schematic representation of a transistor. . . . . . . . . . . . 62

2.2 Series connections implement and. . . . . . . . . . . . . . . . . 63
2.3 Parallel connections implement or. . . . . . . . . . . . . . . . 63
2.4 Parenthesizations expressed as digital logic circuits. . . . . . . 65
2.5 Disjunctive normal form. . . . . . . . . . . . . . . . . . . . . . 67
3.1 A four-color map. . . . . . . . . . . . . . . . . . . . . . . . . . 145

3.2 A 3 × 3 determinant. . . . . . . . . . . . . . . . . . . . . . . . 149
3.3 A Z-module. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.1 An example of a relation. . . . . . . . . . . . . . . . . . . . . . 226

6.2 An example of the “divides” relation. . . . . . . . . . . . . . . 227
6.3 The graph of the “less than” relation. . . . . . . . . . . . . . . 231
6.4 The graph of the divisibility relation. . . . . . . . . . . . . . . 231
6.5 Some simple Hasse diagrams. . . . . . . . . . . . . . . . . . . 256
6.6 Hasse diagram for (P({1, 2, 3}), ⊆). . . . . . . . . . . . . . . . 257
vii
viii LIST OF FIGURES
6.7 Hasse diagram of divisors of 72. . . . . . . . . . . . . . . . . . 258

6.8 The sets related to an arbitrary function. . . . . . . . . . . . . 264
7.1 Full houses in Yahtzee. . . . . . . . . . . . . . . . . . . . . . . 290

7.2 Königsberg, Prussia. . . . . . . . . . . . . . . . . . . . . . . . 308
7.3 Königsberg, Prussia as a graph. . . . . . . . . . . . . . . . . . 309
7.4 A desk with pigeonholes. . . . . . . . . . . . . . . . . . . . . . 320
8.1 Cantor’s snake. . . . . . . . . . . . . . . . . . . . . . . . . . . 347

8.2 Equivalent intervals. . . . . . . . . . . . . . . . . . . . . . . . 351
8.3 An interval is equivalent to a semi-circle. . . . . . . . . . . . . 352
8.4 Binary representations in the unit interval. . . . . . . . . . . . 357
8.5 Setup for proving the C-B-S theorem. . . . . . . . . . . . . . . 368
8.6 An A-stopper in the proof of C-B-S. . . . . . . . . . . . . . . . 371
9.1 The setup for Morley’s Miracle. . . . . . . . . . . . . . . . . . 382

9.2 The first Morley triangle. . . . . . . . . . . . . . . . . . . . . . 383
9.3 Conway’s puzzle proof. . . . . . . . . . . . . . . . . . . . . . . 386
9.4 Scaling in Conway’s puzzle proof. . . . . . . . . . . . . . . . . 387
9.5 An infinite army in the lower half-plane. . . . . . . . . . . . . 390
9.6 Moving one step into the void is trivial. . . . . . . . . . . . . . 392
9.7 Moving two steps into the void is more difficult. . . . . . . . . 393
9.8 Moving three steps into the void takes 8 men. . . . . . . . . . 394
9.9 The taxicab distance to (0, 5). . . . . . . . . . . . . . . . . . . 396
List of Tables
2.1 Converse, inverse and contrapositive. . . . . . . . . . . . . . . 73

2.2 Basic logical equivalences. . . . . . . . . . . . . . . . . . . . . 85
2.3 The rules of inference. . . . . . . . . . . . . . . . . . . . . . . 105
3.1 The definitions of elementary number theory restated. . . . . . 122
4.1 Basic set theoretic equalities. . . . . . . . . . . . . . . . . . . . 174
6.1 Properties of relations. . . . . . . . . . . . . . . . . . . . . . . 236
ix
x LIST OF TABLES
To the student
You are at the right place in your mathematical career to be reading this
book if you liked Trigonometry and Calculus, were able to solve all the prob-
lems, but felt mildly annoyed with the text when it put in these verbose,
incomprehensible things called “proofs.” Those things probably bugged you
because a whole lot of verbiage (not to mention a sprinkling of epsilons and
deltas) was wasted on showing that a thing was true, which was obviously
true! Your physical intuition is sufficient to convince you that a statement
like the Intermediate Value Theorem just has to be true – how can a function
move from one value at a to a different value at b without passing through
all the values in between?
Mathematicians discovered something fundamental hundreds of years be-
fore other scientists – physical intuition is worthless in certain extreme sit-
uations. Probably you’ve heard of some of the odd behavior of particles
in Quantum Mechanics or General Relativity. Physicists have learned, the
hard way, not to trust their intuitions. At least, not until those intuitions
have been retrained to fit reality! Go back to your Calculus textbook and
look up the Intermediate Value Theorem. You’ll probably be surprised to
find that it doesn’t say anything about all functions, only those that are
continuous. So what, you say, aren’t most functions continuous? Actually,
the number of functions that aren’t continuous represents an infinity so huge
that it outweighs the infinity of the real numbers!
xi
xii TO THE STUDENT
The point of this book is to help you with the transition from doing math
at an elementary level (which is concerned mostly with solving problems)
to doing math at an advanced level (which is much more concerned with
axiomatic systems and proving statements within those systems). There are
several books that already propose to aid with this transition, all have various
strengths and weaknesses. In this text we hope to avoid some of those pitfalls
by keeping certain major themes in mind:
1. Mathematics is alive! Math is not just something to be studied from

ancient tomes. A mathematician must have a sense of playfulness.
One needs to “monkey around” with numbers and other mathematical
structures, make discoveries and conjectures and uncover truths.
2. Math is not scary! There is an incredibly terse and compact language

that is used in mathematics – on first sight it looks like hieroglyphics.
That language is actually easy to master, and once mastered, the power
that one gains by expressing ideas rigorously with those symbols is truly
astonishing.
3. Good proofs are everything! No matter how important a fact one dis-
covers, if others don’t become convinced of the truth of the statement
it does not become a part of the edifice of human knowledge. It’s been
said that a proof is simply an argument that convinces. In mathe-
matics, one “convinces” by using one of a handful of argument forms
and developing one’s argument in a clear, step-by-step fashion. Within
those constraints there is actually quite a lot of room for individual
style – there is no one right way to write a proof.
4. You have two cerebral hemispheres – use them both! In perhaps no

other field is the left/right-brain dichotomy more evident than in math.
Some believe that mathematical thought, deductive reasoning, is syn-
xiii
onymous with left-brain function. In truth, doing mathematics is often

a creative, organic, visual, right-brain sort of process – however, in
communicating one’s results one must find that linear, deductive, step-
by-step, left-brain argument. You must use your whole mind to master
advanced mathematics.
Also, there are amusing quotations at the start of every chapter.

xiv TO THE STUDENT
Preface: for Instructors
At many universities and colleges in the United States a course which pro-
vides a transition from lower-level mathematics courses to those in the major
has been adopted. Some may find it hard to believe that a course like Calcu-
lus II is considered “lower-level” so let’s drop the pejoratives and say what’s
really going on. Courses for Math majors, and especially those one takes
in the Junior and Senior years, focus on proofs — students are expected to
learn why a given statement is true, and be able to come up with their own
convincing arguments concerning such “why”s. Mathematics courses that
precede these typically focus on “how.” How does one find the minimum
value a continuous function takes on an interval? How does one determine
the arclength along some curve. Et cetera. The essential raison d’etre of this
text and others like it is to ease this transition from “how” courses to “why”
courses. In other words, our purpose is to help students develop a certain
facility with mathematical proof.
It should be noted that helping people to become good proof writers –
the primary focus of this text – is, very nearly, an impossible task. Indeed,
it can be argued that the best way to learn to write proofs is by writing a lot
of proofs. Devising many different proofs, and doing so in various settings,
definitely develops the facility we hope to engender in a so-called “transitions”
course. Perhaps the pedagogical pendulum will swing back to the previous
tradition of essentially throwing students to the wolves. That is, students
xv
xvi FOR INSTRUCTORS
might be expected to learn the art of proof writing while actually writing
proofs in courses like algebra and analysis1 . Judging from the feedback I
receive from students who have completed our transitions course at Southern
Connecticut State University, I think such a return to the methods of the
past is unlikely. The benefits of these transitions courses are enormous, and
even though the curriculum for undergraduate Mathematics majors is an
extremely full one, the place of a transition course is, I think, assured.
What precisely are the benefits of these transitions courses? One of my
pet theories is that the process one goes through in learning to write and
understand proofs represents a fundamental reorganization of the brain. The
only evidence for this stance, albeit rather indirect, are the almost universal
reports of “weird dreams” from students in these courses. Our minds evolved
in a setting where inductive reasoning is not only acceptable, but advisable in
coping with the world. Imagine some Cro Magnon child touching a burning
branch and being burned by it. S/He quite reasonably draws the conclusion
that s/he should not touch any burning branches, or indeed anything that is
on fire. A Mathematician has to train him or herself to think strictly by the
rules of deductive reasoning – the above experience would only provide the
lesson that at that particular instant of time, that particular burning branch
caused a sensation of pain. Ideally, no further conclusions would be drawn
– obviously this is an untenable method of reasoning for an animal driven
by the desire to survive to adulthood, but it is the only way to think in the
artificial world of Mathematics.
While a gentle introduction to the art of reading and writing proofs is the
primary focus of this text, there are other subsidiary goals for a transitions
course that we hope to address. Principle among these is the need for an
introduction to the “culture” of Mathematics. There is a shared mythos
1
At the University of Maryland, Baltimore County, where I did my undergraduate
work, these courses were actually known as the “proofs” courses.
xvii
and language common to all Mathematicians – although there are certainly

some distinct dialects! Another goal that is of extraordinary importance is
impressing the budding young Mathematics student with the importance of
play. My thesis adviser2 used to be famous for saying “Well, I don’t know!
Why don’t you monkey around with it a little . . . ” In the course of monkeying
around – doing small examples by hand, trying bigger examples with the
aid of a computer, changing some element of the problem to see how it
affected the answer, and various other activities that can best be described as
“play,” eventually patterns emerged, conjectures made themselves apparent,
and possible proof techniques suggested themselves. In this text there are
a great many open-ended problems, some with associated hints as to how
to proceed (which the wise student will avoid until hair-thinning becomes
evident), whose point is to introduce students to this process of mathematical
discovery.
To recap, the goals of this text are: an introduction to reading and writ-
ing mathematical proofs, an introduction to mathematical culture, and an
introduction to the process of discovery in Mathematics. Two pedagogical
principles have been of foremost importance in determining how this mate-
rial is organized and presented. One is the so-called “rule of three” which is
probably familiar to most educators. Propounded by (among others) Hughes,
Hallett, et al. in their reform Calculus it states that, when possible, infor-
mation should be delivered via three distinct mechanisms – symbolically,
graphically and numerically. The other is also a “rule of three” of sorts, it
is captured by the old speechwriter’s maxim – “Tell ’em what you’re gonna
tell ’em. Tell ’em. Then tell ’em what you told ’em.” Important and/or
difficult topics are revisited at least three times in this book. In marked
contrast to the norm in Mathematics, the first treatment of a topic is not
rigorous, precise definitions are often withheld. The intent is to provide a
2
Dr. Vera Pless, to whom I am indebted in more ways than I can express.
xviii FOR INSTRUCTORS
bit of intuition regarding the subject material. Another reason for providing
a crude introduction to a topic before giving rigorous detail revolves around
the way human memory works. Unlike computer memory, which (excluding
the effects of the occasional cosmic ray) is essentially perfect, animal memory
is usually imperfect and mechanisms have evolved to ensure that data that
are important to the individual are not lost. Repetition and rote learning
are often derided these days, but the importance of multiple exposures to a
concept in “anchoring” it in the mind should not be underestimated.
A theme that has recurred over and over in my own thinking about the
transitions course is that the “transition” is that from inductive to deductive
mental processes. Yet, often, we the instructors of these courses are ourselves
so thoroughly ingrained with the deductive approach that the mode of in-
struction presupposes the very transition we hope to facilitate! In this book I
have, to a certain extent, taken the approach of teaching deductive methods
using inductive ones. The first few times a concept is encountered should
only be viewed as providing evidence that lends credence to some mathe-
matical truth. Most concepts that are introduced in this intuitive fashion
are eventually exposited in a rigorous manner – there are exceptions though,
ideas whose scope is beyond that of the present work which are nonetheless
presented here with very little concern for precision. It should not be forgot-
ten that a good transition ought to blend seamlessly into whatever follows.
The courses that follow this material should be proof-intensive courses in ge-
ometry, analysis and algebra. The introduction of some material from these
courses without the usual rigor is intentional. Despite the intentionality, I
should note that this is, quite frankly, an experiment. Nevertheless, it is an
experiment that I am convinced will be beneficial.
Chapter 1
Introduction and notation
Wisdom is the quality that keeps you from getting into situations where you
need it. –Doug Larson
1.1 Basic sets

It has been said1 that “God invented the integers, all else is the work of
Man.” This is a mistranslation. The term “integers” should actually be
“whole numbers.” The concepts of zero and negative values seem to many
to be unnatural constructs. Indeed, otherwise intelligent people are still
known to rail against the concept of a negative quantity – “How can you
have negative three apples?” The concept of zero is also incredibly profound.
Probably most will agree that the natural numbers are a natural con-
struct. We will take as given that you know what the natural numbers are –
the numbers we use to count things. Traditionally, the natural numbers are
denoted N.
1
Usually attributed to Kronecker – “Die ganze Zahl shuf der liebe Gott, alles Übrige
ist Menschenwerk.”
1
2 CHAPTER 1. INTRODUCTION AND NOTATION
N = {1, 2, 3, 4, . . .}
Perhaps the best way of saying what a set is, is to do as we have above.
List all the elements. Of course, if a set has an infinite number of things in
it, this is a difficult task – so we satisfy ourselves by listing enough of the
elements that the pattern becomes clear.
Taking N for granted, what is meant by the “all else” that humankind
is responsible for? The basic sets of different types of “numbers” that every
mathematics student should know are: N, Z, Q, R and C. Respectively: the
naturals, the integers, the rationals, the reals, and the complex numbers.
The use of N, R and C is probably clear to an English speaker. The integers
are denoted with a Z because of the German word zahlen which means “to
count.” The rational numbers are probably denoted using Q, for “quotients.”
Etymology aside, is it possible for us to provide precise descriptions of the
remaining sets?
The integers (Z) are just the set of natural numbers together with the
negatives of naturals and zero. We can use a doubly infinite list to denote
this set.
Z = {. . . − 3, −2, −1, 0, 1, 2, 3, . . .}
To describe the rational numbers precisely we’ll have to wait until Sec-
tion 1.6. In the interim, we can use an intuitively appealing, but somewhat
imprecise definition for the set of rationals. A rational number is a fraction
built out of integers. This also provides us with a chance to give an example
of using the main other way of describing the contents of a set – so-called
set-builder notation.
a
Q={ a ∈ Z and b ∈ Z and b 6= 0}
b
1.1. BASIC SETS 3
This is a good time to start building a “glossary” – a translation lexicon

between the symbols of mathematics and plain language. In the line above we
are defining the set Q of rational numbers, so the first symbols that appear
are “Q =.” It is interesting to note that the equals sign has two subtly
different meanings: assignment and equality testing, in the mathematical
sentence above we are making an assignment – that is, we are declaring that
from now on the set Q will be the set defined on the remainder of the line.2
Let’s dissect the rest of that line now. There are only 4 characters whose
meaning may be in doubt, {, }, ∈ and . The curly braces (a.k.a. french
braces) are almost universally reserved to denote sets, anything appearing
between curly braces is meant to define a set. In translating from “math” to
English, replace the initial brace with the phrase “the set of all.” The next
arcane symbol to appear is the vertical bar. As we will see in Section 1.4.3
this symbol has (at least) two meanings – it will always be clear from context
which is meant. In the sentence we are analyzing, it stands for the words
“such that.” The last bit of arcana to be deciphered is the symbol ∈, it
stands for the English word “in” or, more formally, “is an element of.”
Let’s parse the entire mathematical sentence we’ve been discussing with
an English translation in parallel.
Q = {
The rational numbers are defined to be the set of all
a
b
fractions of the form a over b such that
2
Some Mathematicians contend that only the “equality test” meaning of the equals
sign is real, that by writing the mathematical sentence above we are asserting the truth
of the equality test. This may be technically correct but it isn’t how most people think of
things.
a∈Z and b∈Z
a is an element of the integers and b is an element of the integers
and b 6= 0 }
and b is nonzero. (the final curly brace is silent)
It is quite apparent that the mathematical notation represents a huge

improvement as regards brevity.
As mentioned previously, this definition is slightly flawed. We will have
to wait ’til later to get a truly precise definition of the rationals, but we invite
the reader to mull over what’s wrong with this one. Hint: think about the
issue of whether a fraction is in lowest terms.
Let’s proceed with our menagerie of sets of numbers. The next set we’ll
consider is R, the set of real numbers. To someone who has completed Cal-
culus, the reals are perhaps the most obvious and natural notion of what is
meant by “number.” It may be surprising to learn that the actual definition
of what is meant by a real number is extremely difficult. In fact, the first
reasonable formulation of a precise definition of the reals came around 1858,
more than 180 years after the development of the Calculus3 . A precise def-
inition for the set R of real numbers is beyond the scope of this book, for
the moment consider the following intuitive description. A real number is a
number that measures some physical quantity. For example, if a circle has di-
ameter 1 then its circumference is π, thus π is a real number. The points (0, 0)
p √
and (1, 1) in the Cartesian plane have distance (0 − 1)2 + (0 − 1)2 = 2,
√
thus 2 is a real number. Any rational number is clearly a real number –
slope is a physical quantity, and the line from (0, 0) to (b, a) has slope a/b. In
ancient Greece, Pythagoras – who has sometimes been described as the first
3
Although it was not published until 1736, Newton’s book (De Methodis Serierum et
Fluxionum) describing both differential and integral Calculus was written in 1671.
1.1. BASIC SETS 5
pure Mathematician, believed that every real quantity was in fact rational, a
√
belief that we now know to be false. The numbers π and 2 mentioned above
are not rational numbers. For the moment it is useful to recall a practical
method for distinguishing between rational numbers and real quantities that
are not rational – consider their decimal expansions. If the reader is unfamil-
iar with the result to which we are alluding, we urge you to experiment. Use
a calculator or (even better) a computer algebra package to find the decimal
√
expansions of various quantities. Try π, 2, 1/7, 2/5, 16/17, 1/2 and a few
other quantities of your own choice. Given that we have already said that
the first two of these are not rational, try to determine the pattern. What is
it about the decimal expansions that distinguishes rational quantities from
reals that aren’t rational?
Given that we can’t give a precise definition of a real number at this point
it is perhaps surprising that we can define the set C of complex numbers with
precision (modulo the fact that we define them in terms of R).
C = {a + bi a ∈ R and b ∈ R and i2 = −1}
Translating this bit of mathematics into English we get:
C = {
The complex numbers are defined to be the set of all
a + bi
expressions of the form a plus b times i such that
a∈R and b∈R
a is an element of the reals and b is an element of the reals

and i2 = −1 }
and i has the property that its square is negative one.
We sometimes denote a complex number using a single variable (by con-

vention, either late alphabet Roman letters or Greek letters. Suppose that
we’ve defined z = a + bi. The single letter z denotes the entire complex
number. We can extract the individual components of this complex number
by talking about the real and imaginary parts of z. Specifically, Re(z) = a
is called the real part of z, and Im(z) = b is called the imaginary part of z.
Complex numbers are added and multiplied as if they were binomials

(polynomials with just two terms) where i is treated as if it were the variable
– except that we use the algebraic property that i’s square is -1. For example,
to add the complex numbers 1 + 2i and 3 − 6i we just think of the binomials
1 + 2x and 3 − 6x. Of course we normally write a binomial with the term
involving the variable coming first, but this is just a convention. The sum of
those binomials would be 4−4x and so the sum of the given complex numbers
is 4 − 4i. This sort of operation is fairly typical and is called component-wise
addition. To multiply complex numbers we have to recall how it is that we
multiply binomials. This is the well-known FOIL rule (first, outer, inner,
last). For example the product of 3 − 2x and 4 + 3x is (3 · 4) + (3 · 3x) +
(−2x · 4) + (−2x · 3x) this expression simplifies to 12 + x − 6x2 . The analogous
calculation with complex numbers looks just the same, until we get to the
very last stage where, in simplifying, we use the fact that i2 = −1.
1.1. BASIC SETS 7
(3 − 2i) · (4 + 3i)
= (3 · 4) + (3 · 3i) + (−2i · 4) + (−2i · 3i)
= 12 + 9i − 8i − 6i2
= 12 + i + 6
= 18 + i
The real numbers have a natural ordering, and hence, so do the other
sets that are contained in R. The complex numbers can’t really be put into
a well-defined order — which should be bigger, 1 or i? But we do have a
way to, at least partially, accomplish this task. The modulus of a complex
number is a real number that gives the distance from the origin (0 + 0i) of
the complex plane, to a given complex number. We indicate the modulus
using absolute value bars, and you should note that if a complex number
happens to be purely real, the modulus and the usual notion of absolute
value coincide. If z = a + bi is a complex number, then its modulus, ka + bik,
√
is given by the formula a2 + b2 .
Several of the sets of numbers we’ve been discussing can be split up based
on the so-called trichotomy property: every real number is either positive,
negative or zero. In particular, Z, Q and R can have modifiers stuck on so
that we can discuss (for example) the negative real numbers, or the positive
rational numbers or the integers that aren’t negative. To do this, we put
superscripts on the set symbols, either a + or a − or the word “noneg.”
So
Z+ = {x ∈ Z x > 0}
and
Z− = {x ∈ Z x < 0}
and
Znoneg = {x ∈ Z x ≥ 0}.
Presumably, we could also use “nonpos” as a superscript to indicate non-

positive integers, but this never seems to come up in practice. Also, you
should note that Z+ is really the same thing as N, but that Znoneg is different
because it contains 0.
We would be remiss in closing this section without discussing the way the
sets of numbers we’ve discussed fit together. Simply put, each is contained
in the next. N is contained in Z, Z is contained in Q, Q is contained in R,
and R is contained in C. Geometrically the complex numbers are essentially
a two-dimensional plane. The real numbers sit inside this plane just as the
x-axis sits inside the usual Cartesian plane – in this context you may hear
people talk about “the real line within the complex plane.” It is probably
clear how N lies within Z, and every integer is certainly a real number. The
intermediate set Q (which contains the integers, and is contained by the reals)
has probably the most interesting relationship with the set that contains it.
Think of the real line as being solid, like a dark pencil stroke. The rationals
are like sand that has been sprinkled very evenly over that line. Every point
on the line has bits of sand nearby, but not (necessarily) on top of it.
1.1. BASIC SETS 9
Exercises — 1.1
1. Each of the quantities indexing the rows of the following table is in one
or more of the sets which index the columns. Place a check mark in a
table entry if the quantity is in the set.
N Z Q R C
17
π
22/7
−6
e0
1+i
√
3
i2
2. Write the set Z of integers using a singly infinite listing.
3. Give a description of the set of rational numbers whose decimal ex-

pansions terminate. (Alternatively, you may think of their decimal
expansions ending in an infinitely-long string of zeros.)
√ √
4. Find the first 20 decimal places of π, 3/7, 2, 2/5, 16/17, 3, 1/2 and
42/100. Classify each of these quantity’s decimal expansion as: termi-
nating, having a repeating pattern, or showing no discernible pattern.
5. Consider the process of long division. Does this algorithm give any in-
sight as to why rational numbers have terminating or repeating decimal
expansions? Explain.
6. Give an argument as to why the product of two rational numbers is

again a rational.
7. Perform the following computations with complex numbers
(a) (4 + 3i) − (3 + 2i)
(b) (1 + i) + (1 − i)
(c) (1 + i) · (1 − i)
(d) (2 − 3i) · (3 − 2i)
8. The conjugate of a complex number is denoted with a superscript star,

and is formed by negating the imaginary part. Thus if z = 3 + 4i then
the conjugate of z is z ∗ = 3 − 4i. Give an argument as to why the
product of a complex number and its conjugate is a real quantity. (I.e.
the imaginary part of z · z ∗ is necessarily 0, no matter what complex
number is used for z.)
1.2. DEFINITIONS: PRIME NUMBERS 11
1.2 Definitions: Prime numbers
You may have noticed that in Section 1.1 an awful lot of emphasis was placed
on whether we had good, precise definitions for things. Indeed, more than
once apologies were made for giving imprecise or intuitive definitions. This
is because, in Mathematics, definitions are our lifeblood. More than in any
other human endeavor, Mathematicians strive for precision. This precision
comes with a cost – Mathematics can deal with only the very simplest of
phenomena4 . To laypeople who think of math as being a horribly difficult
subject, that last sentence will certainly sound odd, but most professional
Mathematicians will be nodding their heads at this point. Hard questions
are more properly dealt with by Philosophers than by Mathematicians. Does
a cat have a soul? Impossible to say, because neither of the nouns in that
question can be defined with any precision. Is the squareroot of 2 a rational
number? Absolutely not! The reason for the certainty we feel in answering
this second question is that we know precisely what is meant by the phrases
“squareroot of 2” and “rational number.”
We often need to first approach a topic by thinking visually or intuitively,
but when it comes to proving our assertions, nothing beats the power of hav-
ing the “right” definitions around. It may be surprising to learn that the
“right” definition often evolves over the years. This happens for the simple
reason that some definitions lend themselves more easily to proving asser-
tions. In fact, it is often the case that definitions are inspired by attempts to
prove something that fail. In the midst of such a failure, it isn’t uncommon
for a Mathematician to bemoan “If only the definition of (fill in the blank)
were . . . ”, then to realize that it is possible to use that definition or a modi-
fication of it. But! When there are several definitions for the same idea they
4
For an intriguing discussion of this point, read Gian Carlo Rota’s book Indiscrete
Thoughts [14].
had better agree with one another!

Consider the definition of a prime number.
Definition. A prime number is a positive integer, greater than 1, whose only

factors are 1 and itself.
You probably first heard this definition in Middle School, if not earlier.
It is a perfectly valid definition of what it means for an integer to be prime.
In more advanced mathematics, it was found that it was necessary to define
a notion of primality for objects other than integers. It turns out that the
following statement is equivalent to the definition of “prime” we’ve just given
(when dealing with integers), but that it can be applied in more general
settings.
Definition. A prime is a quantity p such that whenever p is a factor of some

product ab, then either p is a factor of a or p is a factor of b.
Exercise. The number 1 is not considered to be a prime. Does 1 satisfy the

above definition?
Often, it is the case that we can formulate many equivalent definitions

for some concept. When this happens you may run across the abbreviation
TFAE, which stands for “The following are equivalent.” A TFAE proof
consists of showing that a host of different statements actually define the
same concept.
Since we have been discussing primes in this section (mainly as an ex-
ample of a concept with more than one equivalent definition), this seems
like a reasonable time to make some explorations relative to prime numbers.
We’ll begin in the third century B.C.. Eratosthenes of Cyrene was a Greek
Mathematician and Astronomer who is remembered to this day for his many
accomplishments. He was a librarian at the great library of Alexandria. He
made measurements of the Earth’s circumference and the distances of the
Sun and Moon that were remarkably accurate, but probably his most re-
membered achievement is the “sieve” method for finding primes. Indeed, the
sieve of Eratosthenes is still of importance in mathematical research. Basi-
cally, the sieve method consists of creating a very long list of natural numbers
and then crossing off all the numbers that aren’t primes (a positive integer
that isn’t 1, and isn’t a prime is called composite). This process is carried
out in stages. First we circle 2 and then cross off every number that has 2 as
a factor – thus we’ve identified 2 as the first prime number and eliminated
a whole bunch of numbers that aren’t prime. The first number that hasn’t
been eliminated at this stage is 3, we circle it (indicating that 3 is the sec-
ond prime number) and then cross off every number that has 3 as a factor.
Note that some numbers (for example, 6 and 12) will have been crossed off
more than once! In the third stage of the sieve process, we circle 5, which
is the smallest number that hasn’t yet been crossed off, and then cross off
all multiples of 5. The first three stages in the sieve method are shown in
Figure 1.1.
1 2 3 4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19 20
21 22 23 24 25 26 27 28 29 30
31 32 33 34 35 36 37 38 39 40
41 42 43 44 45 46 47 48 49 50
Figure 1.1: The first three stages in the sieve of Eratosthenes. What is the
smallest composite number that hasn’t been crossed off?
It is interesting to note that the sieve gives us a means of finding all the
primes up to p2 by using the primes up to (but not including) p. For example,
to find all the primes less than 132 = 169, we need only use 2, 3, 5, 7 and 11
in the sieve.
Despite the fact that one can find primes using this simple mechanical
method, the way that prime numbers are distributed amongst the integers
is very erratic. Nearly any statement that purports to show some regularity
in the distribution of the primes will turn out to be false. Here are two such
false conjectures regarding prime numbers.
Conjecture 1. Whenever p is a prime number, 2p − 1 is also a prime.
Conjecture 2. The polynomial x2 − 31x + 257 evaluates to a prime number

whenever x is a natural number.
In the exercises for this section, you will be asked to explore these state-
ments further.
Prime numbers act as multiplicative building blocks for the rest of the
integers. When we disassemble an integer into its building blocks we are
finding the prime factorization of that number. Prime factorizations are
unique. That is, a number is either prime or it has prime factors (possibly
raised to various powers) that are uniquely determined – except that they
may be re-ordered.
On the next page is a table that contains all the primes that are less than
5000. Study this table and discover the secret of its compactness!
T 0 1 2 3 4 5 6 7 8 9
H
0 2 3 5 7 1 3 7 9 3 9 1 7 1 3 7 3 9 1 7 1 3 9 3 9 7
1 1 3 7 9 3 7 1 7 9 9 1 7 3 7 3 9 1 1 3 7 9
2 1 3 7 9 3 9 1 1 7 3 9 1 7 1 3 3
3 7 1 3 7 1 7 7 9 3 9 7 3 9 3 9 7
4 1 9 9 1 1 3 9 3 9 7 1 3 7 9 7 1 9
5 3 9 1 3 1 7 7 3 9 1 7 7 3 9
6 1 7 3 7 9 1 1 3 7 3 9 1 3 7 3 1
7 1 9 9 7 3 9 3 1 7 1 9 3 7 7
8 9 1 1 3 7 9 9 3 7 9 3 7 1 3 7
9 7 1 9 9 7 1 7 3 7 1 7 3 1 7
10 9 3 9 1 1 3 9 9 1 1 3 9 7 1 3 7
11 3 9 7 3 9 1 3 3 1 1 7 3
12 1 3 7 3 9 1 7 9 9 7 9 3 9 1 7
13 1 3 7 9 1 7 1 7 3 1 9
14 9 3 7 9 3 9 7 1 3 9 1 1 3 7 9 3 9
15 1 3 1 3 9 3 9 7 1 9 3 7
16 1 7 9 3 9 1 7 7 7 3 7 9 3 7 9
17 9 1 3 3 1 7 3 9 7 3 7 9
18 1 1 3 1 7 1 7 1 3 7 9 9
19 1 7 3 1 3 9 1 3 9 7 3 7 9
20 3 1 7 7 9 9 3 3 9 1 3 7 9 9
21 1 3 9 1 7 1 3 3 1 9
22 3 7 3 1 7 9 3 1 7 9 3 1 7 3 7
23 9 1 3 9 1 7 1 7 1 7 1 3 9 3 9
24 1 7 3 7 1 7 9 7 3 7
25 3 1 1 9 3 9 1 7 9 1 3
26 9 7 1 3 7 7 9 3 1 7 3 7 9 3 9
27 7 1 3 9 9 1 1 9 3 7 7 9 1 7
28 1 3 9 3 7 3 1 7 1 9 7 7
29 3 9 7 7 9 3 7 3 9 1 9
30 1 1 9 3 7 1 9 1 7 9 3 9
31 9 9 1 7 3 7 9 1 7 1
32 3 9 7 1 9 1 3 7 9 1 9
33 1 7 3 9 3 9 1 3 7 9 1 1 3 9 1
34 7 3 3 9 7 1 3 7 9 1 9
35 1 7 7 9 3 9 1 7 7 9 1 1 3 3
36 7 3 7 3 1 7 3 9 1 3 7 1 7
37 1 9 9 7 3 9 1 7 9 9 3 7
38 3 1 3 3 7 1 3 3 7 1 9
39 7 1 7 9 3 9 1 3 7 7 9
40 1 3 7 3 9 1 7 9 1 7 3 9 1 3 9
41 1 7 9 3 9 3 7 9 7
42 1 1 7 9 9 1 1 3 3 9 1 1 3 3 9 7
43 7 7 9 9 7 3 3 1 7
44 9 1 3 1 7 1 7 3 1 3 3
45 7 3 7 9 3 7 9 1 7 3 1 7
46 3 1 7 9 3 9 1 7 3 3 9 1
47 3 1 3 9 3 1 9 3 7 9 3 9
48 1 3 7 1 1 1 7 9
49 3 9 9 1 3 7 3 1 7 7 9 3 7 3 9
Exercises — 1.2
1. Find the prime factorizations of the following integers.
(a) 105
(b) 414
(c) 168
(d) 1612
2. Use the sieve of Eratosthenes to find all prime numbers up to 100.
1 2 3 4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19 20
21 22 23 24 25 26 27 28 29 30
31 32 33 34 35 36 37 38 39 40
41 42 43 44 45 46 47 48 49 50
51 52 53 54 55 56 57 58 59 60
61 62 63 64 65 66 67 68 69 70
71 72 73 74 75 76 77 78 79 80
81 82 83 84 85 86 87 88 89 90
91 92 93 94 95 96 97 98 99 100
3. What would be the largest prime one would sieve with in order to find
all primes up to 400?
4. Complete the following table which is related to Conjecture 1.

p 2p − 1 prime? factors
2 3 yes 1 and 3
3 7 yes 1 and 7
5 31 yes
7 127
11
5. Characterize the prime factorizations of numbers that are perfect squares.
6. Find a counterexample for Conjecture 2.
7. Use the second definition of “prime” to see that 6 is not a prime.

In other words, find two numbers (the a and b that appear in the
definition) such that 6 is not a factor of either, but is a factor of their
product.
8. A famous conjecture that is thought to be true (but for which no proof

is known) is the Twin Prime conjecture. A pair of primes is said to be
twin if they differ by 2. For example, 11 and 13 are twin primes, as
are 431 and 433. The Twin Prime conjecture states that there are an
infinite number of such twins. Try to come up with an argument as to
why 3, 5 and 7 are the only prime triplets.
9. Another famous conjecture, also thought to be true – but as yet un-

proved, is Goldbach’s conjecture. Goldbach’s conjecture states that
every even number greater than 4 is the sum of two odd primes. There
is a function g(n), known as the Goldbach function, defined on the pos-
itive integers, that gives the number of different ways to write a given
number as the sum of two odd primes. For example g(10) = 2 since
10 = 5 + 5 = 7 + 3. Thus another version of Goldbach’s conjecture is
that g(n) is positive whenever n is an even number greater than 4.
Graph g(n) for 6 ≤ n ≤ 20.
1.3. MORE SCARY NOTATION 19
1.3 More scary notation
It is often the case that we want to prove statements that assert something
is true for every element of a set. For example, “Every number has an addi-
tive inverse.” You should note that the truth of that statement is relative,
it depends on what is meant by “number.” If we are talking about natural
numbers it is clearly false: 3’s additive inverse isn’t in the set under con-
sideration. If we are talking about integers or any of the other sets we’ve
considered, the statement is true. A statement that begins with the English
words “every” or “all” is called universally quantified. It is asserted that the
statement holds for everything within some universe. It is probably clear
that when we are making statements asserting that a thing has an additive
inverse, we are not discussing human beings or animals or articles of clothing
– we are talking about objects that it is reasonable to add together: numbers
of one sort or another. When being careful – and we should always strive to
be careful! – it is important to make explicit what universe (known as the
universe of discourse) the objects we are discussing come from. Furthermore,
we need to distinguish between statements that assert that everything in the
universe of discourse has some property, and statements that say something
about a few (or even just one) of the elements of our universe. Statements
of the latter sort are called existentially quantified.
Adding to the glossary or translation lexicon we started earlier, there

are symbols which describe both these types of quantification. The symbol
∀, an upside-down A, is used for universal quantification, and is usually
translated as “for all.” The symbol ∃, a backwards E, is used for existential
quantification, it’s translated as “there is” or “there exists.” Lets have a look
at a mathematically precise sentence that captures the meaning of the one
with which we started this section.
∀x ∈ Z, ∃y ∈ Z, x + y = 0.
Parsing this as we have done before with an English translation in parallel,

we get:
∀x ∈Z ∃y
For every number x in the set of integers there is a number y
∈Z x+y =0
in the integers having the property that their sum is 0.
Exercise. Which type of quantification do the following statements have?
1. Every dog has his day.
2. Some days it’s just not worth getting out of bed.
3. There’s a party in somebody’s dorm this Saturday.
4. There’s someone for everyone.
A couple of the examples in the exercise above actually have two quanti-
fiers in them. When there are two or more (different) quantifiers in a sentence
you have to be careful about keeping their order straight. The following two
sentences contain all the same elements except that the words that indicate
quantification have been switched. Do they have the same meaning?
For every student in James Woods High School, there is some

item of cafeteria food that they like to eat.
There is some item of cafeteria food that every student in James

Woods High School likes to eat.
1.3. MORE SCARY NOTATION 21
Exercises — 1.3
1. How many quantifiers (and what sorts) are in the following sentence?
“Everybody has some friend that thinks they know everything about
some sport.”
2. The sentence “Every metallic element is a solid at room temperature.”

is false. Why?
3. The sentence “For every pair of (distinct) real numbers there is another
real number between them.” is true. Why?
4. Write your own sentences containing four quantifiers. One sentence in

which the quantifiers appear (∀∃∀∃) and another in which they appear
(∃∀∃∀).
1.4 Definitions of elementary number theory
1.4.1 Even and odd
If you divide a number by 2 and it comes out even (i.e. with no remainder)
the number is said to be even. So the word even is related to division. It
turns out that the concept even is better understood through thinking about
multiplication.
Definition. An integer n is even exactly when there is an integer m such

that n = 2m.
You should note that there is a “two-way street” sort of quality to this
definition – indeed with most, if not all, definitions. If a number is even, then
we are guaranteed the existence of another integer half as big. On the other
hand, if we can show that another integer half as big exists, then we know
the original number is even. This two-wayness means that the definition is
what is known as a biconditional, a concept which we’ll revisit in Section 2.2.
A lot of people don’t believe that 0 should be counted as an even number.
Now that we are armed with a precise definition, we can answer this question
easily. Is there an integer x such that 0 = 2x ? Certainly! let x also be 0.
(Notice that in the definition, nothing was said about m and n being distinct
from one another.)
An integer is odd if it isn’t even. That is, amongst integers, there are only
two possibilities: even or odd. We can also define oddness without reference
to “even.”
Definition. An integer n is odd exactly when there is an integer m such that

n = 2m + 1.
1.4. DEFINITIONS OF ELEMENTARY NUMBER THEORY 23
1.4.2 Decimal and base-n notation

You can also identify even numbers by considering their decimal representa-
tion. Recall that each digit in the decimal representation of a number has
a value that depends on its position. For example, the number 3482 really
means 3 · 103 + 4 · 102 + 8 · 101 + 2 · 100 . This is also known as place notation.
The fact that we use the powers of 10 in our place notation is probably due to
the fact that most humans have 10 fingers. It is possible to use any number
in place of 10. In Computer Science there are 3 other bases in common use:
2, 8 and 16 – these are known (respectively) as binary, octal and hexadeci-
mal notation. When denoting a number using some base other than 10, it is
customary to append a subscript indicating the base. So, for example, 10112
is binary notation meaning 1 · 23 + 0 · 22 + 1 · 21 + 1 · 20 or 8 + 2 + 1 = 11. No
matter what base we are using, the rightmost digit of the number multiplies
the base raised to the 0-th power. Any number raised to the 0-th power is
1, and the rightmost digit is consequently known as the units digit. We are
now prepared to give some statements that are equivalent to our definition of
even. These statements truly don’t deserve the designation “theorem,” they
are immediate consequences of the definition.
Theorem 1.4.1. An integer is even if the units digit in its decimal repre-
sentation is one of 0, 2, 4, 6 or 8.
Theorem 1.4.2. An integer is even if the units digit in its binary represen-
tation is 0.
For certain problems it is natural to use some particular notational sys-

tem. For example, the last two “theora” would tend to indicate that binary
numbers are useful in problems dealing with even and odd. Given that
there are many different notations that are available to us, it is obviously
desireable to have means at our disposal for converting between them. It is

possible to develop general rules for converting a base-a number to a base-
b number (where a and b are arbitrary) but it is actually more convenient
to pick a “standard” base (and since we’re human we’ll use base-10) and
develop methods for converting between an arbitrary base and the “stan-
dard” one. Imagine that in the not-to-distant future we need to convert
some numbers from the base-7 system used by the Seven-lobed Amoebazoids
from Epsilon Eridani III to the base-12 scheme favored by the Dodecatons
of Alpha-Centauri IV. We will need a procedure for converting base-7 to
base-10 and another procedure for converting from base-10 to base-12. In
the School House Rock episode “Little Twelve Toes” they describe base-12
numeration in a way that is understandable for elementary school children –
the digits they use are {1, 2, 3, 4, 5, 6, 7, 8, 9, δ, }, the last two digits (which
are pronounced “dec” and “el”) are necessary since we need single symbols
for the things we ordinarily denote using 10 and 11.
Converting from some other base to decimal is easy. You just use the
definition of place notation. For example, to find what 4516637 represents in
decimal, just write
4·75 +5·74 +1·73 +6·72 +6·7+3 = 4·16807+5·2401+1·343+6·49+6·7+3 = 79915.
(Everything in the line above can be interpreted as a base-10 number,

and no subscripts are necessary for base-10.)
Converting from decimal to some other base is harder. There is an algo-
rithm called “repeated division” that we’ll explore a bit in the exercises for
this section. For the moment, just verify that 3δ2712 is also a representation
of the number more conventionally written as 79915.
1.4.3 Divisibility
The notion of being even has an obvious generalization. Suppose we asked

whether 3 divided evenly into a given number. Presumably we could make
a definition of what it meant to be threeven, but rather than doing so (or
engaging in any further punnery) we shall instead move to a general defini-
tion. We need a notation for the situation when one number divides evenly
into another. There are many ways to describe this situation in English, but
essentially just one in “math,” we use a vertical bar – not a fraction bar. In-
deed the difference between this vertical bar and the fraction symbol (|versus
/) needs to be strongly stressed. The vertical bar when placed between two
numbers is a symbol which asks the question “Does the first number divide
evenly (i.e. with no remainder) into the second?” On the other hand the
fraction bar asks you to actually carry out some division. The value of 2 | 5
is false, whereas the value of 2/5 is .4
As was the case in defining even, it turns out that it is best to think of
multiplication, not division, when making a formal definition of this concept.
Given any two integers n and d we define the symbol d | n by
Definition. d | n exactly when ∃k ∈ Z such that n = kd.
In spoken language the symbol d | n is translated in a variety of ways:
• d is a divisor of n.
• d divides n evenly.
• d is a factor of n.
• n is an integer multiple of d.
1.4.4 Floor and ceiling
Suppose there is an elevator with a capacity of 1300 pounds. A large group

of men who all weigh about 200 pounds want to ascend in it. How many
should ride at a time? This is just a division problem, 1300/200 gives 6.5 men
should ride together. Well, obviously putting half a person on an elevator is
a bad idea – should we just round-up and let 7 ride together? Not if the 1300
pound capacity rating doesn’t have a safety margin! This is an example of
the kind of problem in which the floor function is used. The floor function
takes a real number as input and returns the next lower integer.
Suppose after a party we have 43 unopened bottles of beer. We’d like
to store them in containers that hold 12 bottles each. How many containers
will we need? Again, this is simply a division problem – 43/12 = 3.58333.
So we need 3 boxes and another 7 twelfths of a box. Obviously we really
need 4 boxes – at least one will have some unused space in it. In this sort of
situation we’re dealing with the ceiling function. Given a real number, the
ceiling function rounds it up to the next integer.
Both of these functions are denoted using symbols that look very much
like absolute value bars. The difference lies in some small horizontal strokes.
If x is a real number, its floor is denoted bxc, and its ceiling is denoted
dxe. Here are the formal definitions:
Definition. y = bxc exactly when y ∈ Z and y ≤ x < y + 1.
Definition. y = dxe exactly when y ∈ Z and y − 1 < x ≤ y.
Basically, the definition of floor says that y is an integer that is less than
or equal to x, but y + 1 definitely exceeds x. The definition of ceiling can be
paraphrased similarly.
1.4.5 Div and mod

In the next section we’ll discuss the so-called division algorithm – this may
be over-kill since you certainly already know how to do division! Indeed, in
the U.S., long division is usually first studied in the latter half of elementary
school, and division problems that don’t involve a remainder may be found
as early as the first grade. Nevertheless, we’re going to discuss this process
in sordid detail because it gives us a good setting in which to prove relatively
easy statements. Suppose you are setting-up a long division problem in which
the number n is being divided by a divisor d.
q
d n
..
.
r
Recall that the answer consists of two parts, a quotient q, and a remainder
r. Of course, r may be zero, but also, the largest r can be is d − 1. The
assertion that this answer uniquely exists is known as the quotient-remainder
theorem:
Theorem 1.4.3. Given integers n and d, there are unique integers q and r
such that n = qd + r and 0 ≤ r < d.
The words “div” and “mod” that appear in the title of this subsection
provide mathematical shorthand for q and r. Namely, “n mod d” is a way of
expressing the remainder r, and “n div d” is a way of expressing the quotient
q.
If two numbers, m and n, leave the same remainder when you divide them
by d, we say that they are congruent modulo d. One could express this by
writing n mod d = m mod d, but usually we adopt a shorthand notation
n≡m (mod d).

If one is in a context in which it is completely clear what d is, it’s accept-

able to just write n ≡ m.
The “mod” operation is used quite a lot in mathematics. When we do
computations modulo some number d, (this is known as “modular arith-
metic” or, sometimes, “clock arithmetic”) some very nice properties of “mod”
come in handy:
x + y mod d = (x mod d + y mod d) mod d
and
x · y mod d = (x mod d · y mod d) mod d.
These rules mean that we can either do the operations first, then reduce
the answer mod d or we can do the reduction mod d first and then do the
operations (although we may have to do one more round of reduction mod
d).
For example, if we are working mod 10, and want to compute 87·96 mod
10, we can instead just compute 7 · 6 mod 10, which is 2.
1.4.6 Binomial coefficients
A “Binomial” is a polynomial with 2 terms, for example x + 1 or a + b. The

numbers that appear as the coefficients when one raises a binomial to some
power are – rather surprisingly – known as binomial coefficients.
Let’s have a look at the first several powers of a + b.
(a + b)0 = 1
(a + b)1 = a + b
(a + b)2 = a2 + 2ab + b2
To go much further than the second power requires a bit of work, but try
the following
Exercise. Multiply (a + b) and (a2 + 2ab + b2 ) in order to determine (a + b)3 .

If you feel up to it, multiply (a2 +2ab+b2 ) times itself in order to find (a+b)4 .
Since we’re interested in the coefficients of these polynomials, it’s impor-

tant to point out that if no coefficient appears in front of a term that means
the coefficient is 1.
These binomial coefficients can be placed in an arrangement known as
Pascal’s triangle 5 , which provides a convenient way to calculate small bino-
mial coefficients
1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
Figure 1.2: The first 5 rows of Pascal’s triangle (which are numbered 0
through 4 . . . ).
5
This triangle was actually known well before Blaise Pascal began to study it, but it
carries his name today.
Notice that in the triangle there is a border on both sides containing 1’s
and that the numbers on the inside of the triangle are the sum of the two
numbers above them. You can use these facts to extend the triangle.
Exercise. Add the next two rows to the Pascal triangle in Figure 1.2.
Binomial coefficients are denoted using a somewhat strange looking sym-

bol. The number
in the k-th position in row number n of the triangle is
n
denoted . This looks a little like a fraction, but the fraction bar is miss-
k
ing. Don’t put one in! It’s supposed to be missing. In spoken English you
n
say “n choose k” when you encounter the symbol .
k
There is a formula for the binomial coefficients – which is nice. Otherwise
we’d need to complete
a pretty huge Pascal triangle in order to compute
52
something like . The formula involves factorial notation. Just to be
5
sure we are all on the same page, we’ll define factorials before proceeding.
The symbol for factorials is an exclamation point following a number.
This is just a short-hand for expressing the product of all the numbers up
to a given one. For example 7! means 1 · 2 · 3 · 4 · 5 · 6 · 7. Of course, there’s
really no need to write the initial 1 — also, for some reason people usually
write the product in decreasing order (7! = 7 · 6 · 5 · 4 · 3 · 2).
The formula for a binomial coefficient is

n n!
= .
k k! · (n − k)!
For example

5 5! 1·2·3·4·5
= = = 10.
3 3! · (5 − 3)! (1 · 2 · 3) · (1 · 2)
A slightly more complicated example (and one that gamblers are fond of)
is

52 52! 1 · 2 · 3 · · · · 52
= =
5 5! · (52 − 5)! (1 · 2 · 3 · 4 · 5) · (1 · 2 · 3 · · · · 47)
48 · 49 · 50 · 51 · 52
= = 2598960.
1·2·3·4·5
The reason that a gambler might be interested in the number we just cal-
culated is that binomial coefficients do more than just give us the coefficients
in the expansion of a binomial. They also can be used to compute how many
ways one can choose a subset of a given size from a set. Thus 52

5
is the
number of ways that one can get a 5 card hand out of a deck of 52 cards.
Exercise. There are seven days in a week. In how many ways can one choose
a set of three days (per week)?
Exercises — 1.4
1. An integer n is doubly-even if it is even, and the integer m guaranteed

to exist because n is even is itself even. Is 0 doubly-even? What are
the first 3 positive, doubly-even integers?
2. Dividing an integer by two has an interesting interpretation when using

binary notation: simply shift the digits to the right. Thus, 22 = 101102
when divided by two gives 10112 which is 8 + 2 + 1 = 11. How can you
recognize a doubly-even integer from its binary representation?
3. The octal representation of an integer uses powers of 8 in place notation.

The digits of an octal number run from 0 to 7, one never sees 8’s or 9’s.
How would you represent 8 and 9 as octal numbers? What octal number
comes immediately after 7778 ? What (decimal) number is 7778 ?
4. One method of converting from decimal to some other base is called

repeated division. One divides the number by the base and records
the remainder – one then divides the quotient obtained by the base
and records the remainder. Continue dividing the successive quotients
by the base until the quotient is smaller than the base. Convert 3267
to base-7 using repeated division. Check your answer by using the
meaning of base-7 place notation. (For example 543217 means 5 · 74 +
4 · 73 + 3 · 72 + 2 · 71 + 1 · 70 .)
5. State a theorem about the octal representation of even numbers.
6. In hexadecimal (base-16) notation one needs 16 “digits,” the ordinary

digits are used for 0 through 9, and the letters A through F are used to
give single symbols for 10 through 15. The first 32 natural number in
hexadecimal are: 1,2,3,4,5,6,7,8,9,A,B,C,D,E,F,10,11,12,13,14,15,16,
17,18,19,1A, 1B,1C,1D,1E,1F,20.
Write the next 10 hexadecimal numbers after AB.
Write the next 10 hexadecimal numbers after F A.
7. For conversion between the three bases used most often in Computer
Science we can take binary as the “standard” base and convert using a
table look-up. Each octal digit will correspond to a binary triple, and
each hexadecimal digit will correspond to a 4-tuple of binary numbers.
Complete the following tables. (As a check, the 4-tuple next to A in
the table for hexadecimal should be 1010 – which is nice since A is
really 10 so if you read that as “ten-ten” it is a good aid to memory.)
hexadecimal binary
0 0000
1 0001
2 0010
octal binary 3
0 000 4
1 001 5
2 6
3 7
4 8
5 9
6 A
7 B
C
D
E
F
8. Use the tables above to make the following conversions.
(a) Convert 7578 to binary.

(b) Convert 10078 to hexadecimal.
(c) Convert 1001010101102 to octal.
(d) Convert 1111101000110101 to hexadecimal.
(e) Convert F EED16 to binary.
(f) Convert F F F F F F16 to octal.
9. It is a well known fact that if a number is divisible by 3, then 3 divides

the sum of the (decimal) digits of that number. Is this result true in
base 7? Do you think this result is true in any base?
10. Suppose that 340 pounds of sand must be placed into bags having a
50 pound capacity. Write an expression using either floor or ceiling
notation for the number of bags required.
11. Assuming the symbols n,d,q and r have meanings as in the quotient-
remainder theorem (Theorem 1.4.3). Write expressions for q and r, in
terms of n and d using floor and/or ceiling notation.
12. Calculate the following quantities:
(a) 3 mod 5
(b) 37 mod 7
(c) 1000001 mod 100000
(d) 6 div 6
(e) 7 div 6
(f) 1000001 div 2
1.5. SOME ALGORITHMS 35
13. Calculate the following binomial coefficients:
3

(a) 0
7

(b) 7
13

(c) 5
13

(d) 8
52

(e) 7
14. An ice cream shop sells the following flavors: chocolate, vanilla, straw-
berry, coffee, butter pecan, mint chocolate chip and raspberry. How
many different bowls of ice cream – with three scoops – can they make?
1.5 Some algorithms of elementary number

theory
An algorithm is simply a set of clear instructions for achieving some task.
The Persian mathematician and astronomer Al-Khwarizmi6 was a scholar at
the House of Wisdom in Baghdad who lived in the 8th and 9th centuries A.D.
He is remembered for his algebra treatise Hisab al-jabr w’al-muqabala from
which we derive the very word “algebra,” and a text on the Hindu-Arabic
numeration scheme.
Al-Khwarizmi also wrote a treatise on Hindu-Arabic numerals.

The Arabic text is lost but a Latin translation, Algoritmi de nu-
mero Indorum (in English Al-Khwarizmi on the Hindu Art of
Reckoning) gave rise to the word algorithm deriving from his
name in the title. [12]
6
Abu Ja’far Muhammad ibn Musa al-Khwarizmi
While the study of algorithms is more properly a subject within Computer

Science, a student of Mathematics can derive considerable benefit from it.
There is a big difference between an algorithm description intended for hu-
man consumption and one meant for a computer7 . The two favored human-
readable forms for describing algorithms are pseudocode and flowcharts. The
former is text-based and the latter is visual. There are many different mod-
ules from which one can build algorithmic structures: for-next loops, do-while
loops, if-then statements, goto statements, switch-case structures, etc. We’ll
use a minimal subset of the choices available.
• Assignment statements
• If-then control statements
• Goto statements
• Return
We take the view that an algorithm is something like a function, it takes

for its input a list of parameters that describe a particular case of some gen-
eral problem, and produces as its output a solution to that problem. (It
should be noted that there are other possibilities – some programs require
that the variable in which the output is to be placed be handed them as an
input parameter, others have no specific output, their purpose is achieved as
a side-effect.) The intermediary between input and output is the algorithm
instructions themselves and a set of so-called local variables which are used
7
At least as of this writing – the whole history of Computer Science could be described
as the slow advance whereby computers have become able to utilize more and more ab-
stracted descriptions of algorithms. Perhaps in the not-too-distant future machines will
be capable of understanding instruction sets that currently require human interpretors.
much the way scrap paper is used in a hand calculation – intermediate calcu-
lations are written on them, but they are tossed aside once the final answer
has been calculated.
Assignment statements allow us to do all kinds of arithmetic operations
(or rather to think of these types of operations as being atomic.) In actuality
even a simple procedure like adding two numbers requires an algorithm of
sorts, we’ll avoid such a fine level of detail. Assignments consist of evaluating
some (possibly quite complicated) formula in the inputs and local variables
and assigning that value to some local variable. The two uses of the phrase
“local variable” in the previous sentence do not need to be distinct, thus
x = x + 1 is a perfectly legal assignment.
If-then control statements are decision makers. They first calculate a
Boolean expression (this is just a fancy way of saying something that is either
true or false), and send program flow to different locations depending on
that result. A small example will serve as an illustration. Suppose that in
the body of an algorithm we wish to check if 2 variables, x and y are equal,
and if they are, increment x by 1. This is illustrated in Figure 1.3 both in
pseudocode and as a flowchart.
Notice the use of indentation in the pseudocode example to indicate the
statements that are executed if the Boolean expression is true. These ex-
amples also highlight the difference between the two senses that the word
“equals” (and the symbol =) has. In the Boolean expression the sense is
that of testing equality, in the assignment statements (as the name implies)
an assignment is being made. In many programming languages this dis-
tinction is made explicit, for instance in the C language equality testing is
done via the symbol “==” whereas assignment is done using a single equals
sign (=). In Mathematics the equals sign usually indicates equality testing,
when the assignment sense is desired the word “let” will generally precede
the equality.
No
Is x equal to y?
If x = y then
x=x+1
Yes
End If
.
.
.
Let x = x + 1.
Figure 1.3: A small example in pseudocode and as a flowchart
While this brief introduction to the means of notating algorithms is by no

means complete, it is hopefully sufficient for our purpose which is solely to
introduce two algorithms that are important in elementary number theory.
The division algorithm, as presented here, is simply an explicit version of the
process one follows to calculate a quotient and remainder using long division.
The procedure we give is unusually inefficient – with very little thought one
could devise an algorithm that would produce the desired answer using many
fewer operations – however the main point here is purely to show that divi-
sion can be accomplished by essentially mechanical means. The Euclidean
algorithm is far more interesting both from a theoretical and a practical per-
spective. The Euclidean algorithm computes the greatest common divisor
(gcd) of two integers. The gcd of of two numbers a and b is denoted gcd(a, b)
and is the largest integer that divides both a and b evenly.
A pseudocode outline of the division algorithm is as follows:

Algorithm: Division
Inputs: integers n and d.
Local variables: q and r.
Let q = 0.
Let r = n.
Label 1.
If r < d then
Return q and r.
End If
Let q = q + 1.
Let r = r − d.
Goto 1.
This same algorithm is given in flowchart form in Figure 1.4.

Note that in a flowchart the action of a “Goto” statement is clear because
an arrow points to the location where program flow is being redirected. In
pseudocode a “Label” statement is required which indicates a spot where
flow can be redirected via subsequent “Goto” statements. Because of the
potential for confusion in complicated algorithms that involve multitudes of
Goto statements and their corresponding Labels, this sort of redirection is
now deprecated in virtually all popular programming environments.
Before we move on to describe the Euclidean algorithm it might be useful
to describe more explicitly what exactly it’s for. Given a pair of integers, a
and b, there are two quantities that it is important to be able to compute,
the least common multiple or lcm, and the greatest common divisor or gcd.
The lcm also goes by the name lowest common denominator because it is the
smallest denominator that could be used as a common denominator in the
process of adding two fractions that had a and b in their denominators. The
Input: integers n & d

Local: integers q & r
Let q = 0 and r = n.
No
Is r > d?
Yes
Let r = r − d. Return:
Let q = q + 1. q&r
Goto
Figure 1.4: The division algorithm in flowchart form.
gcd and the lcm are related by the formula
ab
lcm(a, b) = ,
gcd(a, b)
so they are essentially equivalent as far as representing a computational chal-

lenge.
The Euclidean algorithm depends on a rather extraordinary property of

the gcd. Suppose that we are trying to compute gcd(a, b) and that a is the
larger of the two numbers. We first feed a and b into the division algorithm
to find q and r such that a = qb + r. It turns out that b and r have the same
gcd as did a and b. In other words, gcd(a, b) = gcd(b, r), furthermore these
numbers are smaller than the ones we started with! This is nice because
it means we’re now dealing with an easier version of the same problem. In
designing an algorithm it is important to formulate a clear ending criterion, a
condition that tells you you’re done. In the case of the Euclidean algorithm,
we know we’re done when the remainder r comes out 0.
So, here, without further ado is the Euclidean algorithm in pseudocode.
A flowchart version is given in Figure 1.5.
Algorithm: Euclidean
Inputs: integers a and b.
Local variables: q and r.
Label 1.
Let (q, r) = Division(a, b).
If r = 0 then
Return b.
End If
Let a = b.
Let b = r.
Goto 1.
Input: integers a & b

Local: integers q & r
Let (q, r) = Division(a, b).
Yes
Is r = 0?
No
Let a = b. Return:
Let b = r. b
Goto
Figure 1.5: The Euclidean algorithm in flowchart form.

It should be noted that for small numbers one can find the gcd and lcm
quite easily by considering their factorizations into primes. For the moment
consider numbers that factor into primes but not into prime powers (that
is, their factorizations don’t involve exponents). The gcd is the product of
the primes that are in common between these factorizations (if there are no
primes in common it is 1). The lcm is the product of all the distinct primes
that appear in the factorizations. As an example, consider 30 and 42. The
factorizations are 30 = 2 · 3 · 5 and 42 = 2 · 3 · 7. The primes that are
common to both factorizations are 2 and 3, thus gcd(30, 42) = 2 · 3 = 6.
The set of all the primes that appear in either factorization is {2, 3, 5, 7} so
lcm(30, 42) = 2 · 3 · 5 · 7 = 210.
The technique just described is of little value for numbers having more
than about 50 decimal digits because it rests a priori on the ability to find
the prime factorizations of the numbers involved. Factoring numbers is easy
enough if they’re reasonably small, especially if some of their prime factors
are small, but in general the problem is considered so difficult that many
cryptographic schemes are based on it.
Exercises — 1.5
1. Trace through the division algorithm with inputs n = 27 and d = 5,

each time an assignment statement is encountered write it out. How
many assignments are involved in this particular computation?
2. Find the gcd’s and lcm’s of the following pairs of numbers.

a b gcd(a, b) lcm(a, b)
110 273
105 42
168 189
3. Formulate a description of the gcd of two numbers in terms of their

prime factorizations in the general case (when the factorizations may
include powers of the primes involved).
4. Trace through the Euclidean algorithm with inputs a = 3731 and

b = 2730, each time the assignment statement that calls the division
algorithm is encountered write out the expression a = qb + r. (With
the actual values involved !)
1.6. RATIONAL AND IRRATIONAL NUMBERS 45
1.6 Rational and irrational numbers

When we first discussed the rational numbers in Section 1.1 we gave the
following definition, which (it was admitted) was slightly flawed.
a
Q={ a ∈ Z and b ∈ Z and b 6= 0}
b
We are now in a position to fix the problem.
So what was the problem after all? Essentially this: there are many ex-
pressions formed with one integer written above another (with an intervening
fraction bar) that represent the exact same rational number. For example
3 14
6
and 28
are distinct things that appear in the set defined above, but we
all know that they both represent the rational number 21 . To eliminate this
problem with our definition of the rationals we need to add an additional
condition that ensures that such duplicates don’t arise. It turns out that
what we want is for the numerators and denominators of our fractions to
have no factors in common. Another way to say this is that the a and b
from the definition above should be chosen so that gcd(a, b) = 1. A pair of
numbers whose gcd is 1 are called relatively prime.
We’re ready, at last, to give a good, precise definition of the set of rational
numbers. (Although it should be noted that we’re not quite done fiddling
around; an even better definition will be given in Section 6.3.)
a
Q={ a, b ∈ Z and b 6= 0 and gcd(a, b) = 1}.
b
As we have in the past, let’s parse this with an English translation in
parallel.
Q = {
The rational numbers are defined to be the set of all

a
a, b ∈ Z
b
fractions of the form a over b such that a and b are integers
and b 6= 0 and gcd(a, b) = 1 }
and b is non-zero and a and b are relatively prime.
Finally, we are ready to face a fundamental problem that was glossed-

over in Section 1.1. We defined two sets back then, Q and R, the hidden
assumption that one makes in asserting that there are two of something is
that the two things are distinct. Is this really the case? The reals have been
defined (unrigorously) as numbers that measure the magnitudes of physical
quantities, so another way to state the question is this: Are there physical
quantities (for example lengths) that are not rational numbers?
The answer is that yes there are numbers that measure lengths which
are not rational numbers. With our new and improved definition of what
is meant by a rational number we are ready to prove that there is at least
one length that can’t be expressed as a fraction. Using the Pythagorean
theorem it’s easy to see that the length of the diagonal of a unit square is
√ √
2. The proof that 2 is not rational is usually attributed to the followers
of Pythagoras (but probably not to Pythagoras himself). In any case it
is a result of great antiquity. The proof is of a type known as reductio ad
absurdum 8 . We show that a given assumption leads logically to an absurdity,
a statement that can’t be true, then we know that the original assumption
must itself be false. This method of proof is a bit slippery; one has to first
assume the exact opposite of what one hopes to prove and then argue (on
purpose) towards a ridiculous conclusion.
√
Theorem 1.6.1. The number 2 is not in the set Q of rational numbers.
8
Reduction to an absurdity – better known these days as proof by contradiction.
Before we can actually give the proof we should prove an intermediary

result – but we won’t, we’ll save this proof for the student to do later (heh,
heh, heh. . . ). These sorts of intermediate results, things that don’t deserve to
be called theorems themselves, but that aren’t entirely self-evident are known
as lemmas. It is often the case that in an attempt at proving a statement we
find ourselves in need of some small fact. Perhaps it even seems to be true
but it’s not clear. In such circumstances, good form dictates that we first
state and prove the lemma then proceed on to our theorem and its proof.
So, here, without its proof is the lemma we’ll need.
Lemma 1.6.2. If the square of an integer is even, then the original integer
is even.
Given that thoroughness demands that we fill in this gap by actually

proving the lemma at a later date, we can now proceed with the proof of our
theorem.
√
Proof: Suppose to the contrary that 2 is a rational number.
Then by the definition of the set of rational numbers, we know
that there are integers a and b having the following properties:
√ a
2 = and gcd(a, b) = 1.
b
√ a
Consider the expression 2 = . By squaring both sides of this
b
we obtain
a2
2= .
b2
This last expression can be rearranged to give
a2 = 2b2
An immediate consequence of this last equation is that a2 is an

even number. Using the lemma above we now know that a is an
even integer and hence that there is an integer m such that a =
2m. Substituting this last expression into the previous equation
gives
(2m)2 = 2b2 ,
thus,
4m2 = 2b2 ,
so
2m2 = b2 .
This tells us that b2 is even, and hence (by the lemma), b is even.
Finally, we have arrived at the desired absurdity because if a and
b are both even then gcd(a, b) ≥ 2, but, on the other hand, one
of our initial assumptions is that gcd(a, b) = 1.
Q.E.D.
Exercises — 1.6
1. Rational Approximation is a field of mathematics that has received

much study. The main idea is to find rational numbers that are very
good approximations to given irrationals. For example, 22/7 is a well-
known rational approximation to π. Find good rational approximations
√ √ √
to 2, 3, 5 and e.
2. The theory of base-n notation that we looked at in sub-section1.4.2

can be extended to deal with real and rational numbers by introducing
a decimal point (which should probably be re-named in accordance
with the base) and adding digits to the left of it. For instance 1.1011
is binary notation for 1 · 20 + 1 · 2−1 + 0 · 2−2 + 1 · 2−3 + 1 · 2−4 or
1 1 1 11
1+ + + =1 .
2 8 16 16
Consider the binary number .1010010001000010000010000001 . . ., is this
number rational or irrational? Why?
3. If a number x is even, it’s easy to show that its square x2 is even.

The lemma that went unproved in this section asks us to start with a
square (x2 ) that is even and deduce that the unsquared number (x) is
even. Perform some numerical experimentation to check whether this
assertion is reasonable. Can you give an argument that would prove
it?
√ √
4. The proof that 2 is irrational can be generalized to show that p is
irrational for every prime number p. What statement would be equiva-
lent to the lemma about the parity of x and x2 in such a generalization?
√
5. Write a proof that 3 is irrational.
1.7 Relations
One of the principle ways in which mathematical writing differs from ordinary
writing is in its incredible brevity. For instance, a Ph.D. thesis for someone
in the humanities would be very suspicious if its length were less than 300
pages, whereas it would be quite acceptable for a math doctoral student to
submit a thesis amounting to less than 100 pages. Indeed, the usual criteria
for a doctoral thesis (or indeed any scholarly work in mathematics) is that
it be “new, true and interesting.” If one can prove a truly interesting, novel
result in a single page – they’ll probably hand over the sheepskin.
How is this great brevity achieved? By inserting single symbols in place
of a whole paragraph’s worth of words! One class of symbols in particular has
immense power – so-called relational symbols. When you place a relational
symbol between two expressions, you create a sentence that says the relation
holds. The period at the end of the last sentence should probably be pro-
nounced! “The relation holds, period!” In other words when you write down
a mathematical sentence involving a relation, you are asserting the relation
is True (the capital T is intentional). This is why it’s okay to write “2 < 3”
but it’s not okay to write “3 < 2.” The symbol < is a relation symbol and
you are only supposed to put it between two things when they actually bear
this relation to one another.
The situation becomes slightly more complicated when we have variables
in relational expressions, but before we proceed to consider that complication
let’s make a list of the relations we’ve seen to date:
=, <, >, ≤, ≥, | , and ≡ (mod m).
Each of these, when placed between numbers, produces a statement that

is either true or false. Ordinarily we wouldn’t write down the false ones,
instead we should express that we know the relation doesn’t hold by negating
1.7. RELATIONS 51
the relation symbol (often by drawing a slash through it, but some of the
symbols above are negations of others).
So what about expressions involving variables and these relation symbols?
For example what does x < y really mean? Okay, I know that you know
what x < y means but, philosophically, a relation symbol involving variables
is doing something that you may have only been vaguely aware of in the past
– it is introducing a supposition. Watch out for relation symbols involving
variables! Whenever you encounter them it means the rules of the game are
being subtly altered – up until the point where you see x < y, x and y are
just two random numbers, but after that point we must suppose that x is
the smaller of the two.
The relations we’ve discussed so far are binary relations, that is, they go
in between two numbers. There are also higher order relations. For example,
a famous ternary relation (a relationship between three things) is the notion
of “betweenness.” If A, B and C are three points which all lie on a single
line, we write A ? B ? C if B falls somewhere on the line segment AC. So the
symbol A ? B ? C is shorthand for the sentence “Point B lies somewhere in
between points A and C on the line determined by them.”
There is a slightly silly tendency these days to define functions as being
a special class of relations. (This is slightly silly not because it’s wrong –
indeed, functions are a special type of relation – but because it’s the least
intuitive approach possible, and it is usually foisted-off on middle or high
school students.) When this approach is taken, we first define a relation to
be any set of ordered pairs and then state a restriction on the ordered pairs
that may be in a relation if it is to be a function. Clearly what these Algebra
textbook authors are talking about are binary relations, a ternary relation
would actually be a set of ordered triples, and higher order relations might
involve ordered 4-tuples or 5-tuples, etc. A couple of small examples should
help to clear up this connection between a relation symbol and some set of
tuples.
Consider the numbers from 1 to 5 and the less-than relation, <. As a set
of ordered pairs, this relation is the set
{(1, 2), (1, 3), (1, 4), (1, 5), (2, 3), (2, 4), (2, 5), (3, 4), (3, 5), (4, 5)}.
The pairs that are in the relation are those such that the first is smaller
than the second.
An example involving the ternary relation “betweenness” can be had from
the following diagram.
A
F B
C
E D
The betweenness relation on the points in this diagram consists of the

following triples.
{(A, B, C), (A, G, D), (A, F, E), (B, G, E), (C, B, A), (C, G, F ), (C, D, E),
(D, G, A), (E, D, C), (E, G, B), (E, F, A), (F, G, C)}.
Exercise. When thinking of a function as a special type of relation, the

pairs are of the form (x, f (x)). That is, they consist of an input and the
1.7. RELATIONS 53
corresponding output. What is the restriction that must be placed on the

pairs in a relation if it is to be a function? (Hint: think about the so-called
vertical line test.)
Exercises — 1.7
1. Consider the numbers from 1 to 10. Give the set of pairs of these
numbers that corresponds to the divisibility relation.
2. The domain of a function (or binary relation) is the set of numbers

appearing in the first coordinate. The range of a function (or binary
relation) is the set of numbers appearing in the second coordinate.
Consider the set {0, 1, 2, 3, 4, 5, 6} and the function f (x) = x2 (mod 7).
Express this function as a relation by explicitly writing out the set of
ordered pairs it contains. What is the range of this function?
3. What relation on the numbers from 1 to 10 does the following set of

ordered pairs represent?
{(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6), (1, 7), (1, 8), (1, 9), (1, 10),
(2, 2), (2, 3), (2, 4), (2, 5), (2, 6), (2, 7), (2, 8), (2, 9), (2, 10),
(3, 3), (3, 4), (3, 5), (3, 6), (3, 7), (3, 8), (3, 9), (3, 10),
(4, 4), (4, 5), (4, 6), (4, 7), (4, 8), (4, 9), (4, 10),
(5, 5), (5, 6), (5, 7), (5, 8), (5, 9), (5, 10),
(6, 6), (6, 7), (6, 8), (6, 9), (6, 10),
(7, 7), (7, 8), (7, 9), (7, 10),
(8, 8), (8, 9), (8, 10),
(9, 9), (9, 10),
(10, 10)}
4. Draw a five-pointed star, label all 10 points. There are 40 triples of

these labels that satisfy the betweenness relation. List them.
1.7. RELATIONS 55
5. A function f (x) is said to be invertible if there is another function g(x)

such that g(f (x)) = x for all values of x. (Usually, the inverse function,
g(x) would be denoted f −1 (x).) Suppose a function is presented to
you as a relation – that is, you are just given a set of pairs. How
can you distinguish whether the function represented by this list of
input/output pairs is invertible? How can you produce the inverse (as
a set of ordered pairs)?
Chapter 2
Logic and quantifiers
If at first you don’t succeed, try again. Then quit. There’s no use being a
damn fool about it. –W. C. Fields
2.1 Predicates and Logical Connectives

In every branch of Mathematics there are special, atomic, notions that defy
precise definition. In Geometry, for example, the atomic notions are points,
lines and their incidence. Euclid defines a point as “that which has no part” –
people can argue (and have argued) incessantly over what exactly is meant by
this. Is it essentially saying that anything without volume, area or length of
some sort is a point? In modern times it has been recognized that any formal
system of argumentation has to have such elemental, undefined, concepts –
and that Euclid’s apparent lapse in precision comes from an attempt to hide
this basic fact. The notion of “point” can’t really be defined. All we can do is
point (no joke intended) at a variety of points and hope that our audience will
absorb the same concept of point that we hold via the process of induction1 .
1
inference of a generalized conclusion from particular instances – compare DEDUC-
TION
57
58 CHAPTER 2. LOGIC AND QUANTIFIERS
The atomic concepts in Set Theory are “set”, “element” and “member-
ship”. The atomic concepts in Logic are “true”, “false”, “sentence” and
“statement”.
Regarding true and false, we hope there is no uncertainty as to their
meanings. Sentence also has a well-understood meaning that most will agree
on – a syntactically correct ordered collection of words such as “Johnny was
a football player.” or “Red is a color.” or “This is a sentence which does
not refer to itself.” A statement is a sentence which is either true or false.
In other words, a statement is a sentence whose truth value is definite, in
more other words, it is always possible to decide – one way or the other –
whether a statement is true or false. The first example of a sentence given
above (“Johnny was a football player”) is not a statement – the problem is
that it is ambiguous unless we know who Johnny is. If it had said “Johnny
Unitas was a football player.” then it would have been a statement. If it
had said “Johnny Appleseed was a football player.” it would also have been
a statement, just not a true one.
Ambiguity is only one reason that a sentence may not be a statement.
As we consider more complex sentences, it may be the case that the truth
value of a given sentence simply cannot be decided. One of the most cel-
ebrated mathematical results of the 20th century is Kurt Gödel’s “Incom-
pleteness Theorem.” An important aspect of this theory is the proof that in
any axiomatic system of mathematical thought there must be undecidable
sentences – statements which can neither be proved nor disproved from the
axioms2 . Simple sentences (e.g. those of the form subject-verb-object) have
little chance of being undecidable for this reason, so we will next look at ways
of building more complex sentences from simple components.
Let’s start with an example. Suppose I come up to you in some window-
2
There are trivial systems that are complete, but if a system is sufficiently complicated
that it contains “interesting” statements it can’t be complete.
2.1. PREDICATES AND LOGICAL CONNECTIVES 59
less room and make the statement: “The sun is shining but it’s raining!” You
decide to investigate my claim and determine its veracity. Upon reaching a
room that has a view of the exterior there are four possible combinations
of sunniness and/or precipitation that you may find. That is, the atomic
predicates “The sun is shining” and “It is raining” can each be true or false
independently of one another. In the following table we introduce a con-
vention used throughout the remainder of this book – that true is indicated
with a capital letter T and false is indicated with the Greek letter φ (which
is basically a Greek F, and is a lot harder to mistake for a T than an F is.)
The sun is shining It is raining

T T
T φ
φ T
φ φ
Each row of the above table represents a possible state of the outside
world. Suppose you observe the conditions given in the last row, namely
that it is neither sunny, nor is it raining – you would certainly conclude
that I am not to be trusted. I.e. my statement, the compounding of “The
sun is shining” and “It is raining” (with the word “but” in between as a
connector) is false. If you think about it a bit, you’ll agree that this so-called
compound sentence is true only in the case that both of its component pieces
are true. This underscores an amusing linguistic point: “but” and “and”
have exactly the same meaning! More precisely, they denote the same thing,
they have subtly different connotations however – “but” indicates that both
of the statements it connects are true and that the speaker is surprised by
this state of affairs.
In Mathematics we distinguish two main connectives for hooking-up sim-
ple sentences into compound ones. The conjunction of two sentences is the
compound sentence made by sticking the word “and” between them. The
disjunction of two sentences is formed by placing an “or” between them.
Conjunctions are true only when both components are true. Disjunctions
are false only when both components are false.
As usual, mathematicians have developed an incredibly terse, compact
notation for these ideas.3 First, we represent an entire sentence by a single
letter – traditionally, a capital letter. This is called a predicate variable. For
example, following the example above, we could denote the sentence “The
sun is shining” by the letter S. Similarly, we could make the assignment
R = “It is raining.” The conjunction and disjunction of these sentences can
then be represented using the symbols S ∧ R and S ∨ R, respectively. As a
mnemonic, note that the connective in S ∧ R looks very much like the capital
letter A (as in And).
To display, very succinctly, the effect of these two connectives we can use
so-called truth tables. In a truth table we list all possible truth values of
the predicate variables and then enumerate the truth values of some com-
pound sentence. For the conjunction and disjunction connectors we have
(respectively):
A B A∧B A B A∨B
T T T T T T
T φ φ and T φ T .
φ T φ φ T T
φ φ φ φ φ φ
In addition to these connectors we need a modifier (called negation) that

acts on individual sentences. The negation of a sentence A is denoted by ¬A,
and its truth value is exactly the opposite of A’s truth value. The negation
3
One begins to suspect that mathematicians form an unusually lazy sub-species of
humanity.
of a sentence is also known as the denial of a sentence. A truth table for the
negation operator is somewhat trivial but we include it here for completeness.
A ¬A
T φ
φ T
These three simple tools (and, or & not) are sufficient to create extraor-
dinarily complex sentences out of basic components. The way these pieces
interrelate is a bit reminiscent of algebra, in fact the study of these logical
operators (or any operators that act like them) is called Boolean Algebra4 .
There are distinct differences between Boolean and ordinary algebra however.
In regular algebra we have the binary connectors + (plus) and · (times), and
the unary negation operator −, these are certainly analogous to ∧, ∨ & ¬,
but there are certain consequences of the fact that multiplication is effec-
tively repeated addition that simply don’t hold for the Boolean operators.
For example, there is a well-defined precedence between · and +. In parsing
the expression 4 · 5 + 3 we all know that the multiplication is to be done first.
There is no such rule governing order of operations between ∧ and ∨, so an
expression like A ∧ B ∨ C is simply ambiguous – it must have parentheses
inserted in order to show the order, either (A∧B)∨C or A∧(B ∨C). Another
distinction between ordinary and Boolean algebra is exponentiation. If there
were exponents in Boolean algebra, we’d need two different kinds – one for
repeated conjunction and another for repeated disjunction.
Exercise. Why is it that there is no such thing as exponentiation in the

algebra of Logic?
While there are many differences between Boolean algebra and the usual,
garden-variety algebra, there are also many similarities. For instance, the
4
In honor of George Boole, whose 1854 book An investigation into the Laws of Thought
inaugurated the subject.
associative, commutative and distributive laws of Algebra all have versions

that work in the Boolean case.
A very handy way of visualizing Boolean expressions is given by digital
logic circuit diagrams. To discuss these diagrams we must make a brief
digression into Electronics. One of the most basic components inside an
electronic device is a transistor, this is a component that acts like a switch
for electricity, but the switch itself is controlled by electricity. In Figure 2.1
we see the usual schematic representation of a transistor. If voltage is applied
to the wire labeled z, the transistor becomes conductive, and current may
flow from x to y.
x y
Figure 2.1: A schematic representation of a transistor.
Suppose that two transistor are connected as in Figure 2.2 (this is called
a series connection). In order for current to flow from x to y we must have
voltage applied to both the wires labeled z and w. In other words, this circuit
effectively creates the and operation – assuming voltage is always applied to
x, if z and w are energized then the output at y will be energized.
When two transistors are connected in parallel (this is illustrated in Fig-
ure 2.3) current can flow from x to y when either (or both) of the wires at
z and w have voltage applied. This brings up a point which is confusing
for some: in common speech the use of the word “or” often has the sense
known as exclusive or (a.k.a. xor), when we say “X or Y” we mean “Either
z w
x y
Figure 2.2: The connection of two transistors in series provides an imple-

mentation of the and operator.
X or Y, but not both.” In Electronics and Mathematics, or always has the

non-exclusive (better known as inclusive) sense.
z w
x y
Figure 2.3: The connection of two transistors in parallel provides an imple-

mentation of the or operator.
As a sort of graphical shorthand, electronics engineers use the symbols

below to indicate and-gates, or-gates & not-gates (better known as negators).
And (∧) Or (∨)
Not (¬)
Using this graphical “language” one can make schematic representations

of logical expressions. Some find that tracing such diagrams makes under-
standing the structure of a Boolean expression easier. For example, in Fig-
ure 2.4 we illustrate 2 of the possible ways that the conjunction of four
predicate variables can be parenthesized. In fact, when a multitude of pred-
icates are joined by the same connective, the way in which the expression
is parenthesized is unimportant, thus one often sees a further shorthand —
gates with more than 2 inputs.
A common task for an electronics designer is to come up with a digi-
tal logic circuit having a prescribed input/output table. Note that an in-
put/output table for a logic circuit is entirely analogous with a truth table
for a compound sentence in Logic — except that we use 0’s and 1’s rather
than T’s and φ’s.
Suppose that we wanted to design a circuit that would have the following
input/output table.
B
((A ∧ B) ∧ (C ∧ D))
C
C
(((A ∧ B) ∧ C) ∧ D)
D
Figure 2.4: Two of the possible ways to parenthesize the conjunction of four
statement variables – expressed as digital logic circuits.
x y z out
0 0 0 0
0 0 1 0
0 1 0 0
0 1 1 1
1 0 0 0
1 0 1 0
1 1 0 1
1 1 1 1
A systematic method for accomplishing such a design task involves a
notion called disjunctive normal form. A Boolean expression is in disjunctive
normal form if it consists of the disjunction of one or more statements, each
of which consists entirely of conjunctions of predicate variables and/or their
negations. In other words, the or of a bunch of ands. In terms of digital logic
circuits, the ands we’re talking about are called recognizers. For example,
the following 3-input and gates recognize the input states in the 4th, 7th and
8th rows of the i/o table above. (These are the rows where the output is
supposed to be 1.)
x x
y y
z z
z
In Figure 2.5 we illustrate how to create a circuit whose i/o table is as

above using these recognizers.
out
Figure 2.5: A digital logic circuit built using disjunctive normal form. The
output of this circuit is (¬x ∧ y ∧ z) ∨ (x ∧ y ∧ ¬z) ∨ (x ∧ y ∧ z).
Exercises — 2.1
1. Design a digital logic circuit (using and, or & not gates) that imple-
ments an exclusive or.
2. Consider the sentence “This is a sentence which does not refer to itself.”
which was given in the beginning of this chapter as an example. Is this
sentence a statement? If so, what is its truth value?
3. Consider the sentence “This sentence is false.” Is this sentence a state-

ment?
4. Complete truth tables for each of the sentences (A∧B)∨C and A∧(B ∨
C). Does it seem that these sentences have the same logical content?
5. There are two other logical connectives that are used somewhat less
commonly than ∨ and ∧. These are the Scheffer stroke and the Peirce
arrow – written | and ↓, respectively — they are also known as NAND
and NOR.
The truth tables for these connectives are:
A B A|B A B A↓B
T T φ T T φ
T φ T and T φ φ
φ T T φ T φ
φ φ T φ φ T
Find an expression for (A ∧ ¬B) ∨ C using only these new connectives

(as well as negation and the variable symbols themselves).
2.2 Implication
Suppose a mother makes the following statement to her child: “If you finish
your peas, you’ll get dessert.”
This is a compound sentence made up of the two simpler sentences P =
“You finish your peas” and D = “You’ll get dessert.” It is an example of
a type of compound sentence called a conditional. Conditionals are if-then
type statements. In ordinary language the word “then” is often elided (as is
the case with our example above). Another way of phrasing the “If P then
D.” relationship is to use the word “implies” — although it would be a rather
uncommon mother who would say “Finishing your peas implies that you will
receive dessert.”
As was the case in the previous section, there are four possible situations
and we must consider each to decide the truth/falsity of this conditional
statement. The peas may or may not be finished, and independently, the
dessert may or may not be proffered.
Suppose the child finishes the peas and the mother comes across with the
dessert. Clearly, in this situation the mother’s statement was true. On the
other hand, if the child finishes the hated peas and yet does not receive a
treat, it is just as obvious that the mother has lied! What do we say about
the mother’s veracity in the case that the peas go unfinished? Here, Mom
gets a break. She can either hold firm and deliver no dessert, or she can
be a softy and give out unearned sweets – in either case, we can’t accuse
her of telling a falsehood. The statement she made had to do only with the
eventualities following total pea consumption, she said nothing about what
happens if the peas go uneaten.
A conditional statement’s components are called the antecedent (this is
the “if” part, as in “finish your peas”) and the consequent (this is the “then”
part, as in “get dessert”). The discussion in the last paragraph was intended
2.2. IMPLICATION 71
to make the point that when the antecedent is false, we should consider the
conditional to be true. Conditionals that are true because their antecedents
are false are said to be vacuously true. The conditional involving an an-
tecedent A and a consequent B is expressed symbolically using an arrow:
A =⇒ B. Here is a truth table for this connective.
A B A =⇒ B
T T T
T φ φ
φ T T
φ φ T
Exercise. Note that this truth table is similar to the truth table for A ∨ B
in that there is only a single row having a φ in the last column. For A ∨
B the φ occurs in the 4th row and for A =⇒ B it occurs in the 2nd
row. This suggests that by suitably modifying things (replacing A or B by
their negations) we could come up with an “or” statement that had the same
meaning as the conditional. Try it!
It is fairly common that conditionals are used to express threats, as in

the peas/dessert example. Another common way to express a threat is to
use a disjunction – “Finish your peas, or you won’t get dessert.” If you’ve
been paying attention (and did the last exercise), you will notice that this
is not the disjunction that should have the same meaning as the original
conditional. There is probably no mother on Earth who would say “Don’t
finish your peas, or you get dessert!” to her child (certainly not if she expects
to be understood). So what’s going on here?
The problem is that “Finish your peas, or you won’t get dessert.” has
the same logical content as “If you get dessert then you finished your peas.”
(Notice that the roles of the antecedent and consequent have been switched.)
And, while this last sentence sounds awkward, it is probably a more accurate
reflection of what the mother intended. The problem really is that people are
incredibly sloppy with their conditional statements! A lot of people secretly
want the 3rd row of the truth table for =⇒ to have a φ in it, and it
simply doesn’t! The operator that results if we do make this modification is
called the biconditional, and is expressed in English using the phrase “if and
only if” (which leads mathematicians to the abbreviation “iff” much to the
consternation of spell-checking programs everywhere). The biconditional is
denoted using an arrow that points both ways. Its truth table follows.
A B A ⇐⇒ B
T T T
T φ φ
φ T φ
φ φ T
Please note, that while we like to strive for precision, we do not necessarily
recommend the use of phrases such as “You will receive dessert if, and only
if, you finish your peas.” with young children.
Since conditional sentences are often confused with the sentence that
has the roles of antecedent and consequent reversed, this switched-around
sentence has been given a name: it is the converse of the original statement.
Another conditional that is distinct from (but related to) a given conditional
is its inverse. This sort of sentence probably had to be named because of a
very common misconception, many people think that the way to negate an
if-then proposition is to negate its parts. Algebraically, this looks reasonable
– sort of a distributive law for logical negation over implications – ¬(A =⇒
B) = ¬A =⇒ ¬B. Sadly, this reasonable looking assertion can’t possibly
be true; since implications have just one φ in a truth table, the negation of
an implication must have three – but the statement with the ¬’s on the parts
of the implication is going to only have a single φ in its truth table.
2.2. IMPLICATION 73
To recap, the converse of an implication has the pieces (antecedent and

consequent) switched about. The inverse of an implication has the pieces
negated. Neither of these is the same as the original implication. Oddly, this
is one of those times when two wrongs do make a right. If you start with
an implication, form its converse, then take the inverse of that, you get a
statement having exactly the same logical meaning as the original. This new
statement is called the contrapositive.
This information is displayed in Table 2.1
converses
A =⇒ B B =⇒ A
inverses
¬A =⇒ ¬B ¬B =⇒ ¬A
Table 2.1: The relationship between a conditional statement, its converse,

its inverse and its contrapositive.
One final piece of advice about conditionals: don’t confuse logical if-then
relationships with causality. Many of the if-then sentences we run into in
ordinary life describe cause and effect: “If you cut the green wire the bomb
will explode.” (Okay, that one is an example from the ordinary life of a
bomb squad technician, but . . . ) It is usually best to think of the if-then
relationships we find in Logic as divorced from the flow of time, the fact that
A =⇒ B is logically the same as ¬A ∨ B lends credence to this point of
view.
Exercises — 2.2
1. The transitive property of equality says that if a = b and b = c then

a = c. Does the implication arrow satisfy a transitive property? If so,
state it.
2. Complete a truth table for the compound sentences A =⇒ B and

¬A ∨ B.
3. Complete a truth table for the compound sentence A =⇒ (B =⇒ C)

and for the sentence (A =⇒ B) =⇒ C. What can you conclude
about conditionals and the associative property?
4. Determine a sentence using the and connector (∧) that gives the nega-
tion of A =⇒ B.
5. Rewrite the sentence “Fix the toilet or I won’t pay the rent!” as a
conditional.
6. Why is it that the sentence “If pigs can fly, I am the king of Mesopotamia.”
true?
7. Express the statement A =⇒ B using the Peirce arrow and/or the

Scheffer stroke. (See Exercise 5 in the previous section.)
8. Find the contrapositives of the following sentences.
(a) If you can’t do the time, don’t do the crime.
(b) If you do well in school, you’ll get a good job.
(c) If you wish others to treat you in a certain way, you must treat
others in that fashion.
(d) If it’s raining, there must be clouds.

2.2. IMPLICATION 75
P∞
(e) If an ≤ bn , for all n and n=0 bn is a convergent series, then
P∞
n=0 an is a convergent series.
9. What are the converse and inverse of “If you watch my back, I’ll watch
your back.”?
10. The integral test in Calculus is used to determine whether an infinite

series converges or diverges: Suppose that f (x) is a positive, decreasing,
real-valued function with limx−→∞ f (x) = 0, if the improper integral
R∞
f (x) has a finite value, then the infinite series ∞
P
0 n=1 f (n) converges.
The integral test should be envisioned by letting the series correspond

to a right-hand Riemann sum for the integral, since the function is
decreasing, a right-hand Riemann sum is an underestimate for the falue
of the integral, thus
∞
X Z ∞
f (n) < f (x).
n=1 0
Discuss the meanings of and (where possible) provide justifications for

the inverse, converse and contrapositive of the conditional statement
in the integral test.
2.3 Logical equivalences

Some logical statements are “the same.” For example, in the last section,
we discussed the fact that a conditional and its contrapositive have the same
logical content. Wouldn’t we be justified in writing something like the fol-
lowing?
A =⇒ B = ¬B =⇒ ¬A
Well, one pretty serious objection to doing that is that the equals sign
(=) has already got a job; it is used to indicate that two numerical quanti-
ties are the same. What we’re doing here is really sort of a different thing!
Nevertheless, there is a concept of “sameness” between certain compound
statements, and we need a symbolic way of expressing it. There are two no-
tations in common use. The notation that seems to be preferred by logicians
is the biconditional ( ⇐⇒ ). The notation we’ll use in the rest of this book
is an equals sign with a bit of extra decoration on it (∼
=).
Thus we can can either write
(A =⇒ B) ⇐⇒ (¬B =⇒ ¬A)
or
A =⇒ B ∼
= ¬B =⇒ ¬A.
I like the latter, but use whichever form you like – no one will have any
problem understanding either.
The formal definition of logical equivalence, which is what we’ve been
describing, is this: two compound sentences are logically equivalent if in a
truth table (that contains all possible combinations of the truth values of
the predicate variables in its rows) the truth values of the two sentences are
equal in every row.
2.3. LOGICAL EQUIVALENCES 77
Exercise. Consider the two compound sentences A ∨ B and A ∨ (¬A ∧ B).

There are a total of 2 predicate variables between them, so a truth table with 4
rows will suffice. Fill out the missing entries in the truth table and determine
whether the statements are equivalent.
A B A∨B A ∨ (¬A ∧ B)
T T
T φ
φ T
φ φ
One could, in principle, verify all logical equivalences by filling out truth
tables. Indeed, in the exercises for this section we will ask you to develop a
certain facility at this task. While this activity can be somewhat fun, and
many of my students want the filling-out of truth tables to be a significant
portion of their midterm exam, you will probably eventually come to find it
somewhat tedious. A slightly more mature approach to logical equivalences
is this: use a set of basic equivalences – which themselves may be verified via
truth tables – as the basic rules or laws of logical equivalence, and develop
a strategy for converting one sentence into another using these rules. This
process will feel very familiar, it is like “doing” algebra, but the rules one is
allowed to use are subtly different.
First we have the commutative laws, one each for conjunction and disjunc-
tion. It’s worth noting that there isn’t a commutative law for implication.
The commutative property of conjunction says that A ∧ B ∼ = B ∧ A. This
is quite an apparent statement from the perspective of linguistics. Surely it’s
the same thing to say “the weather is cold and snowy” as it is to say “the
weather is snowy and cold.” This commutative property is also clear from
the perspective of digital logic circuits.
A
A∧B
B
A
B∧A
B
The commutative property of disjunctions is equally transparent from the

perspective of a circuit diagram.
A
A∨B
B
A
B∨A
B
The associative laws also have something to do with what order oper-
ations are done. One could think of the difference in the following terms:
Commutative properties involve spatial or physical order and the associative
properties involve temporal order. The associative law of addition could be
used to say we’ll get the same result if we add 2 and 3 first, then add 4, or if
we add 2 to the sum of 3 and 4 (i.e. that (2+3)+4 is the same as 2+(3+4).)
Note that physically, the numbers are in the same order (2 then 3 then 4) in
both expressions but that the parentheses indicate a precedence in when the
plus signs are evaluated.
The associative law of conjunction states that A ∧ (B ∧ C) ∼

= (A ∧ B) ∧ C.
In visual terms, this means the following two circuit diagrams are equivalent.
B (A ∧ B) ∧ C
B A ∧ (B ∧ C)
The associative law of disjunction states that A ∨ (B ∨ C) ∼

= (A ∨ B) ∨ C.
Visually, this looks like:
B (A ∨ B) ∨ C
B A ∨ (B ∨ C)
C
Exercise. In a situation where both associativity and commutativity pertain

the symbols involved can appear in any order and with any reasonable paren-
thesization. In how many different ways can the sum 2 + 3 + 4 be expressed?
Only consider expression that are fully parenthesized.
The next type of basic logical equivalences we’ll consider are the so-called
distributive laws. Distributive laws involve the interaction of two operations,
when we distribute multiplication over a sum, we effectively replace one in-
stance of an operand and the associated operator, with two instances, as is
illustrated below.
2*(3+4)=(2*3)+(2*4)
The logical operators ∧ and ∨ each distribute over the other. Thus we
have the distributive law of conjunction over disjunction, which is expressed
in the equivalence A ∧ (B ∨ C) ∼ = (A ∧ B) ∨ (A ∧ C) and in the following
digital logic circuit diagram.
B A ∧ (B ∨ C)
B
(A ∧ B) ∨ (A ∧ C)
We also have the distributive law of disjunction over conjunction which is

given by the equivalence A ∨ (B ∧ C) ∼= (A ∨ B) ∧ (A ∨ C) and in the circuit
diagram:
B A ∨ (B ∧ C)
B
(A ∨ B) ∧ (A ∨ C)
C
Traditionally, the laws we’ve just stated would be called left-distributive

laws and we would also need to state that there are right-distributive laws
that apply. Since, in the current setting, we have already said that the
commutative law is valid, this isn’t really necessary.
Exercise. State the right-hand versions of the distributive laws.
The next set of laws we’ll consider come from trying to figure out what
the distribution of a minus sign over a sum (−(x + y) = −x + −y) should
correspond to in Boolean algebra. At first blush one might assume the analo-
gous thing in Boolean algebra would be something like ¬(A ∧ B) ∼= ¬A ∧ ¬B,
but we can easily dismiss this by looking at a truth table.
A B ¬(A ∧ B) ¬A ∧ ¬B
T T φ φ
T φ T φ
φ T T φ
φ φ T T
What actually works is a set of rules known as DeMorgan’s laws, which

basically say that you distribute the negative sign but you also must change
the operator. As logical equivalences, DeMorgan’s laws are
¬(A ∧ B) ∼
= ¬A ∨ ¬B
and
¬(A ∨ B) ∼
= ¬A ∧ ¬B.
In ordinary arithmetic there are two notions of “inverse.” The negative

of a number is known as its additive inverse and the reciprocal of a number
is its multiplicative inverse. These notions lead to a couple of equations,
x + −x = 0
and
1
x· = 1.
x
Boolean algebra only has one “inverse” concept, the denial of a predicate
(i.e. logical negation), but the equations above have analogues, as do the
symbols 0 and 1 that appear in them. First, consider the Boolean expression
A ∨ ¬A. This is the logical or of a statement and its exact opposite; when
one is true the other is false and vice versa. But, the disjunction A ∨ ¬A, is
always true! We use the symbol t (which stands for tautology) to represent a
compound sentence whose truth value is always true. A tautology (t) is to
Boolean algebra something like a zero (0) is to arithmetic. Similar thinking
about the Boolean expression A ∧ ¬A leads to the definition of the symbol c
(which stands for contradiction) to represent a sentence that is always false.
The rules we have been discussing are known as complementarity laws:
A ∨ ¬A ∼
= t and A ∧ ¬A ∼
= c
Now that we have the special logical sentences represented by t and c we

can present the so-called identity laws, A ∧ t ∼
= A and A ∨ c ∼ = A. If you
“and” a statement with something that is always true, this new compound
has the exact same truth values as the original. If you “or” a statement
with something that is always false, the new compound statement is also
unchanged from the original. Thus performing a conjunction with a tautology
has no effect – sort of like multiplying by 1. Performing a disjunction with a
contradiction also has no effect – this is somewhat akin to adding 0.
The number 0 has a special property: 0 · x = 0 is an equation that holds
no matter what x is. This is known as a domination property. Note that
there isn’t a dominance rule that involves 1. On the Boolean side, both the
symbols t and c have related domination rules.
A∨t∼
=t and A∧c∼
=c
In mathematics the word idempotent is used to describe situations where

a power of a thing may be equal to that thing. For example, because (−1)3 =
−1, we say that −1 is an idempotent. Both of the Boolean operations have
idempotence relations that just always work (regardless of the operand). In
ordinary algebra idempotents are very rare (0, 1 and −1 are the only ones
that come to mind), but in Boolean algebra every statement is equivalent to
its square – where the square of A can be interpreted either as A ∧ A or as
A ∨ A.
A∨A∼
=A and A∧A∼
=A
There are a couple of properties of the logical negation operator that

should be stated, though probably they seem self-evident. If you form the
denial of a denial, you come back to the same thing as the original; also the
symbols c and t are negations of one another.
¬(¬A) ∼
=A and ¬t ∼
=c
Finally, we should mention a really strange property, called absorption,

which states that the expressions A ∧ (A ∨ B) and A ∨ (A ∧ B) don’t actually
have anything to do with B at all! Both of the preceding statements are
equivalent to A.
A ∧ (A ∨ B) ∼
=A and A ∨ (A ∧ B) ∼
=A
In Table 2.2, we have collected all of these basic logical equivalences in

one place.
Conjunctive Disjunctive Algebraic

version version analog
Commutative
A∧B ∼
=B∧A A∨B ∼
=B∨A 2+3=3+2
laws
Associative A ∧ (B ∧ C) A ∨ (B ∨ C) 2 + (3 + 4)
∼
= (A ∧ B) ∧ C ∼
= (A ∨ B) ∨ C = (2 + 3) + 4
laws
Distributive A ∧ (B ∨ C) ∼
= A ∨ (B ∧ C) ∼
= 2 · (3 + 4)
laws (A ∧ B) ∨ (A ∧ C) (A ∨ B) ∧ (A ∨ C) = (2 · 3 + 2 · 4)
DeMorgan’s ¬(A ∧ B) ¬(A ∨ B)

none
laws ∼
= ¬A ∨ ¬B ∼
= ¬A ∧ ¬B
Complementarity A ∧ ¬A ∼
= c A ∨ ¬A ∼
= t 2 + (−2) = 0
Identity
A∧t∼
=A A∨c∼
=A 7+0=7
laws
Domination A∧c∼
=c A∨t∼
=t 7·0=0
Idempotence A∧A∼
=A A∨A∼
=A 1·1=1
Absorption A ∧ (A ∨ B) ∼
=A A ∨ (A ∧ B) ∼
=A none
Table 2.2: Basic logical equivalences.

Exercises — 2.3
1. There are 3 operations used in basic algebra (addition, multiplication

and exponentiation) and thus there are potentially 6 different distribu-
tive laws. State all 6 “laws” and determine which 2 are actually valid.
(As an example, the distributive law of addition over multiplication
would look like x + (y · z) = (x + y) · (x + z), this isn’t one of the true
ones.)
2. Use truth tables to verify or disprove the following logical equivalences.
(a) (A ∧ B) ∨ B ∼
= (A ∨ B) ∧ B
(b) A ∧ (B ∨ ¬A) ∼
= A∧B
(c) (A ∧ ¬B) ∨ (¬A ∧ ¬B) ∼
= ((A ∨ ¬B) ∧ (¬A ∨ ¬B)
(d) The absorption laws.
3. Draw pairs of related digital logic circuits that illustrate DeMorgan’s

laws.
4. Because a conditional sentence is equivalent to a certain disjunction,

and because DeMorgan’s law tells us that the negation of a disjunc-
tion is a conjunction, it follows that the negation of a conditional is a
conjunction. Find denials (the negation of a sentence is often called its
“denial”) for each of the following conditionals.
(a) “If you smoke, you’ll get lung cancer.”

(b) “If a substance glitters, it is not necessarily gold.”
(c) “If there is smoke, there must also be fire.”
(d) “If a number is squared, the result is positive.”
(e) “If a matrix is square, it is invertible.”
5. The so-called “ethic of reciprocity” is an idea that has come up in

many of the world’s religions and philosophies. Below are statements
of the ethic from several sources. Discuss their logical meanings and
determine which (if any) are logically equivalent.
(a) “One should not behave towards others in a way which is disagree-
able to oneself.” Mencius Vii.A.4 (Hinduism)
(b) “None of you [truly] believes until he wishes for his brother what
he wishes for himself.” Number 13 of Imam “Al-Nawawi’s Forty
Hadiths.” (Islam)
(c) “And as ye would that men should do to you, do ye also to them
likewise.” Luke 6:31, King James Version. (Christianity)
(d) “What is hateful to you, do not to your fellow man. This is the law:
all the rest is commentary.” Talmud, Shabbat 31a. (Judaism)
(e) “An it harm no one, do what thou wilt” (Wicca)
(f) “What you would avoid suffering yourself, seek not to impose on
others.” (the Greek philosopher Epictetus – first century A.D.)
(g) “Do not do unto others as you expect they should do unto you.
Their tastes may not be the same.” (the Irish playwright George
Bernard Shaw – 20th century A.D.)
2.4 Two-column proofs
If you’ve ever spent much time trying to check someone else’s work in solving
an algebraic expression, you’d probably agree that it would be a help to know
what they were trying to do in each step. Most people have this fairly vague
notion that they’re allowed to “do the same thing on both sides” and they’re
allowed to simplify the sides of the equation separately – but more often than
not, several different things get done on a given line, mistakes get made, and
it can be nearly impossible to figure out what went wrong and where.
Now, after all, the beauty of math is supposed to lie in its crystal clarity,
so this sort of situation is really unacceptable. It may be an impossible goal
to get “the average Joe” to perform algebraic manipulations with clarity,
but those of us who aspire to become mathematicians must certainly hold
ourselves to a higher standard. Two-column proofs are usually what is meant
by a “higher standard” when we are talking about relatively mechanical
manipulations – like doing algebra, or more to the point, proving logical
equivalences. Now don’t despair! You will not, in a mathematical career, be
expected to provide two-column proofs very often, in fact, in more advanced
work one tends to not give any sort of proof for a statement that lends itself
to a two-column approach. But, if you find yourself writing “As the reader
can easily verify, Equation 17 holds. . . ” in a paper, or making some similar
remark to your students, you are morally obligated to being able to produce
a two-column proof.
So what, exactly, is a two-column proof? In the left column you show
your work, being careful to go one step at a time. In the right column you
provide a justification for each step.
We’re going to go through a couple of examples of two-column proofs
in the context of proving logical equivalences. One thing to watch out for:
if you’re trying to prove a given equivalence, and the first thing you write
2.4. TWO-COLUMN PROOFS 89
down is that very equivalence, it’s wrong! This would constitute the logical
error known as “begging the question” also known as “circular reasoning.”
It’s clearly not okay to try to demonstrate some fact by first asserting the
very same fact. Nevertheless, there is (for some unknown reason) a powerful
temptation to do this very thing. To avoid making this error, we will not
put any equivalences on a single line. Instead we will start with one side or
the other of the statement to be proved, and modify it using known rules of
equivalence, until we arrive at the other side.
Without further ado, let’s provide a proof of the equivalence A ∧ (B ∨
¬A) ∼
= A ∧ B.5
A ∧ (B ∨ ¬A)
distributive law
∼
= (A ∧ B) ∨ (A ∧ ¬A)
complementarity
∼
= (A ∧ B) ∨ c
identity law
∼
= (A ∧ B)
We have assembled a nice, step-by-step sequence of equivalences – each

justified by a known law – that begins with the left-hand side of the statement
to be proved and ends with the right-hand side. That’s an irrefutable proof!
In the next example we’ll highlight a slightly sloppy habit of thought that
tends to be problematic. People usually (at first) associate a direction with
the basic logical equivalences, this is reasonable for several of them because
one side is markedly simpler than the other. For example, the domination
rule would normally be used to replace a part of a statement that looked
like “A ∧ c” with the simpler expression “c”. There is a certain amount of
5
This equivalence should have been verified using truth tables in the exercises from the
previous section.
strategization necessary in doing these proofs, and I usually advise people

to start with the more complicated side of the equivalence to be proved. It
just feels right to work in the direction of making things simpler, but there
are times when one has to take one step back before proceeding two steps
forward. . .
Let’s have a look at another equivalence: A∧(B∨C) ∼
= (A∧(B∨C))∨(A∧
C). There are many different ways in which valid steps can be concatenated
to convert one side of this equivalence into the other, so a subsidiary goal is
to find a proof that uses the least number of steps. Following my own advice,
I’ll start with the right-hand side of this one.
(A ∧ (B ∨ C)) ∨ (A ∧ C)
distributive law
∼
= ((A∧B)∨(A∧C))∨(A∧C)
associative law
∼
= (A∧B)∨((A∧C)∨(A∧C))
idempotence
∼
= (A ∧ B) ∨ (A ∧ C)
distributive law
∼
= A ∧ (B ∨ C)
Note that in the example we’ve just done, the two applications of the
distributive law go in opposite directions as far as their influence on the
complexity of the expressions are concerned.
2.4. TWO-COLUMN PROOFS 91
Exercises — 2.4
1. Write two-column proofs that verify each of the following logical equiv-
alences.
(a) A ∨ (A ∧ B) ∼
= A ∧ (A ∨ B)
(b) A ∨ B ∼
= A ∨ (¬A ∧ B)
(c) A ∼
= A ∧ ((A ∨ ¬B) ∨ (A ∨ B))
(d) A ∼
= A ∧ (A ∨ (A ∧ (B ∨ C)))
(e) ¬(A ∧ B) ∧ ¬(A ∧ C) ∼
= ¬A ∨ (¬B ∧ ¬C)
2.5 Quantified statements

All of the statements discussed in the previous sections were of the “com-
pletely unambiguous” sort; that is, they didn’t have any unknowns in them.
As a reader of this text, it’s a sure bet that you’ve mastered Algebra and are
firmly convinced of the utility of x and y. Admittedly, we’ve used variables
to refer to sentences (or sentence fragments) themselves, but we’ve said that
sentences that had variables in them were ambiguous and didn’t even deserve
to be called logical statements. The notion of quantification allows us to use
the power of variables within a sentence without introducing ambiguity.
Consider the sentence “There are exactly 7 odd primes less than 20.”
This sentence has some kind of ambiguity in it (because it doesn’t mention
the primes explicitly) and yet it certainly seems to have a definite truth
value! The reason its truth value is known (by the way, it is T) is that the
sentence is quantified. “X is an odd prime less than 20.” is an ambiguous
sentence, but “There are exactly 7 distinct X’s that are odd primes less than
20.” is not. This example represents a fairly unusual form of quantification.
Usually, we take away the ambiguity of a sentence having a variable in it by
asserting one of two levels of quantification: “this is true at least once” or
“this is always true”. We’ve actually seen the symbols (∃ and ∀) for these
concepts already (in Section 1.3).
An open sentence is one that has variables in it. We represent open
sentences using a sort of functional notation to show what variables are in
them.
Examples:
x
i) P (x) = “22 + 1 is a prime.”
ii) Q(x, y) = “x is prime or y is a divisor of x.”
iii) L(f, c, l) = “The function f has limit l at c, if and only if, for every
2.5. QUANTIFIED STATEMENTS 93
positive number , there is a positive number δ such that whenever

|x − c| < δ it follows that |f (x) − l| < .”
That last example certainly is a doozey! At first glance it would appear

to have more than three variables in it, and indeed it does! In order of
appearance, we have f , l, c, , δ and x – the last three variables that appear
(, δ and x) are said to be bound. A variable in an open sentence is bound if
it is in the scope of a quantifier. Bound variables don’t need to be mentioned
in the argument list of the sentence. Unfortunately, when sentences are given
in natural languages the quantification status of a variable may not be clear.
For example in the third sentence above, the variable δ is easily seen to be
in the scope of the quantifier ∃ because of the words “there is a positive
number” that precede it. Similarly, is universally quantified (∀) because
the phrase “for every positive number” appears before it. What is the status
of x? Is it really bound? The answers to such questions may not be clear at
first, but after some thought you should be able to decide that x is universally
quantified.
Exercise. What word in example iii) indicates that x is in the scope of a ∀

quantifier?
It is not uncommon, in advanced Mathematics, to encounter compound

sentences involving dozens of variables and 4 or 5 levels of quantification.
Such sentences seem hopelessly complicated at first sight – the key to under-
standing them is to determine each variable’s quantification status explicitly
and to break things down into simpler subparts.
For instance, in understanding example iii) above, it might be useful to
define some new open sentences:
D(x, c, δ) = “|x − c| < δ”
E(f, x, l, ) = “|f (x) − l| < ”
Furthermore, it’s often handy to replace an awkward phrase (such as “the

limit of f at c is l”) with symbols when possible.
Example iii) now looks like
lim f (x) = l ⇐⇒ ∀ > 0 ∃δ > 0 ∀x D(x, c, δ) =⇒ E(f, x, l, ).

x→c
The sentence D(x, c, δ) is usually interpreted as saying that “x is close to

c” (where δ tells you how close.) The sentence E(f, x, l, ) could be expressed
informally as “f (x) is close to l” (again, serves to make the word “close”
more exact).
It’s instructive to write this sentence one last time, completely in symbols
and without the abbreviations we created for saying that x is near c and f (x)
is near l:
lim f (x) = l ⇐⇒ ∀ > 0 ∃δ > 0 ∀x (|x − c| < δ) =⇒ (|f (x) − l| < ).
x→c
It would not be unfair to say that developing the facility to read, and
understand, this hieroglyph (and others like it) constitutes the first several
weeks of a course in Real Analysis.
Let us turn back to another of the examples (of an open sentence) from
x
the beginning of this section. P (x) = “22 + 1 is a prime.”
In the 17th century, Pierre de Fermat made the conjecture6 that ∀x ∈
N, P (x). No doubt, this seemed reasonable to Fermat because the numbers
given by this formula (they are called Fermat numbers in his honor) are
all primes – at first! Fermat numbers are conventionally denoted with a
n
subscripted letter F, Fn = 22 + 1, the first five Fermat numbers are prime.
0
F0 = 22 + 1 = 3
1
F1 = 22 + 1 = 5
6
Fermat’s more famous conjecture, that xn +y n = z n has no non-trivial integer solutions
if n is an integer with n > 2 was discovered after his death.
2
F2 = 22 + 1 = 17
3
F3 = 22 + 1 = 257
4
F4 = 22 + 1 = 65537
Fermat probably computed that F5 = 4294967297, and we can well imag-

ine that he checked that this number was not divisible by any small primes.
Of course, this was well before the development of effective computing ma-
chinery, so we shouldn’t blame Fermat for not noticing that 4294967297 =
641 · 6700417. This remarkable feat of factoring can be replicated in seconds
on a modern computer, however it was done first by Leonhard Euler in 1732!
There is quite a lot of literature concerning the primeness and/or compos-
iteness of Fermat numbers, so far, all the Fermat numbers between F5 and
F32 (inclusive) have been shown to be composite. One might be tempted to
conjecture that only the first five Fermat numbers are prime, however this
temptation should be resisted . . .
Let us set aside, for the moment, further questions about Fermat numbers.
Suppose we define the set U (for ‘Universe’) by U = {0, 1, 2, 3, 4}. Then the
assertion, “∀x ∈ U, P (x).” is certainly true. You should note that the only
variable in this sentence is x, and that the variable is bound – it is universally
quantified. Open sentences that have all variables bound are statements, it is
possible (in principle, and in finite universes, in practice) to check the truth
value of such sentences. Indeed, the sentence “∀x ∈ U, P (x)” has the same
logical content as “P (0) ∧ P (1) ∧ P (2) ∧ P (3) ∧ P (4)”. Both happen to be
true, but the real point here is to note that a universally quantified sentence
can be thought of instead as a conjunction.
Exercise. Define a new set U by U = {0, 1, 2, 3, 4, 5}. Write a sentence

using disjunctions that is equivalent to “∃x ∈ U, ¬P (x).”
Even when we are dealing with infinite universes, it is possible to think

of universally quantified sentences in terms of conjunctions, and existentially
quantified sentences in terms of disjunctions. For example, a quick look at

the graphs should be sufficient to convince you that “x > ln x” is a sentence
that is true for all x values in R+ . There is a notation, reminiscent of so-
called sigma notation for sums, that can be used to express this universally
quantified sentence as a conjunction.
∀x ∈ R+ , x > ln x ∼
^
= x > ln x
x∈R+
A similar notation exists for disjunctions. Purely as an example, consider

the following problem from recreational math: Find a four digit number that
is an integer multiple of its reversal. (By reversal, we mean the four digit
number with the digits in the opposite order – for example, the reversal of
1234 is 4321.) The sentence7 that states that this question has a solution is
∃abcd ∈ Z, ∃k ∈ Z, abcd = k · dcba
This could be expressed instead as the disjunction of 9000 statements, or

more compactly as
_
∃k ∈ Z, abcd = k · dcba.
1000≤abcd≤9999
Exercise. The existential statement above is true because 8712 = 4 · 2178.

There is one other solution – find it!
An important, or at least useful, talent for a Mathematics student to

develop is the ability to negate quantified sentences. The major reason for
this is the fact that the contrapositive of a conditional sentence is logically
7
This sentence uses what is commonly referred to as an “abuse of notation” in order
avoid an unnecessarily complex problem statement. One should not necessarily avoid such
abuses if one’s readers can be expected to easily understand what is meant, any more than
one should completely eschew the splitting of infinitives.
equivalent to it. This leads to a method known as “proof by contraposition”

which can be expressed succinctly by the advice:
“If you get stuck, try writing down the contrapositive.”
Since writing down the contrapositive will often involve finding the nega-
tion of a quantified sentence, let’s try a few.
Our universe of discourse8 will be P = {Manny, Moe, Jack}. Consider
the sentence “∀x ∈ P, x starts with M.” The equivalent sentence expressed
conjunctively is
(Manny starts with M)∧

(Moe starts with M)∧
(Jack starts with M).
The negation of this sentence (by DeMorgan’s law) is a disjunction:
(Manny doesn’t start with M)∨

(Moe doesn’t start with M)∨
(Jack doesn’t start with M)
Finally, this disjunction of three sentences can be converted into a single

sentence, existentially quantified over P :
“∃x ∈ P, ¬(x starts with M).”
The discussion in the previous paragraphs justifies some laws of Logic
which should be thought of as generalizations of DeMorgan’s laws:
¬(∀x ∈ U, P (x)) ∼
= ∃x ∈ U, ¬P (x)
8
The Pep Boys – Manny, Moe and Jack – are hopefully known to some readers as the
mascots of a chain of automotive supply stores.
and
¬(∃x ∈ U, P (x)) ∼
= ∀x ∈ U, ¬P (x).
It’s equally valid to think of these rules in a way that’s divorced from
DeMorgan’s laws. To show that a universal sentence is false, it suffices to
show that an existential sentence involving a negation of the original is true.9
9
To show that it is not the case that every Pep boy’s name starts with ‘M’, one only
needs to demonstrate that there is a Pep boy (Jack) whose name doesn’t start with ‘M’.
Exercises — 2.5
1. There is a common variant of the existential quantifier, ∃!, if you write

∃! x, P (x) you are asserting that there is a unique element in the uni-
verse that makes P (x) true. Determine how to negate the sentence
∃! x, P (x).
2. The order in which quantifiers appear is important. Let L(x, y) be

the open sentence “x is in love with y.” Discuss the meanings of the
following quantified statements and find their negations.
(a) ∀x ∃y L(x, y).

(b) ∃x ∀y L(x, y).
(c) ∀x ∀y L(x, y).
(d) ∃x ∃y L(x, y).
3. Determine a useful denial of:

∀ > 0 ∃δ > 0 ∀x (|x − c| < δ) =⇒ (|f (x) − l| < ).
The denial above gives a criterion for saying limx→c f (x) 6= l.
4. A Sophie Germain prime 10 is a prime number p such that the corre-

sponding odd number 2p+1 is also a prime. For example 11 is a Sophie
Germain prime since 23 = 2 · 11 + 1 is also prime. Almost all Sophie
Germain primes are congruent to 5 (mod 6), nevertheless, there are
exceptions – so the statement “There are Sophie Germain primes that
are not 5 mod 6.” is true. Verify this.
10
Marie-Sophie Germain (1776 - 1831) was a French mathematician who made major
contributions to number theory, acoustics and elasticity [12].
2.6 Deductive reasoning and Argument forms
Deduction is the process by which we determine new truths from old. It

is sometimes claimed that nothing truly new can come from deduction, the
truth of a statement that is arrived at by deductive processes was lying
(perhaps hidden somewhat) within the hypotheses. This claim is something
of a canard, as any Sherlock Holmes aficionado can tell you, the statements
that can sometimes be deduced from others can be remarkably surprising.
A better argument against deduction is that it is a relatively ineffective way
for most human beings to discover new truths – for that purpose inductive
processes are superior for the majority of us. Nevertheless, if a chain of
deductive reasoning leading from known hypotheses to a particular conclusion
can be exhibited, the truth of the conclusion is unassailable. For this reason,
mathematicians have latched on to deductive reasoning as the tool for, if not
discovering our theorems, communicating them to others.
The word “argument” has a negative connotation for many people be-
cause it seems to have to do with disagreement. Arguments within math-
ematics (as well as many other scholarly areas), while they may be impas-
sioned, should not involve discord. A mathematical argument is a sequence
of logically connected statements designed to produce agreement as to the
validity of a proposition. This “design” generally follows one of two possibil-
ities, inductive reasoning or deductive reasoning. In an inductive argument a
long list of premises is presented whose truths are considered to be apparent
to all, each of which provides evidence that the desired conclusion is true.
So an inductive argument represents a kind of statistical thing, you have all
these statements that are true each of which indicates that the conclusion is
most likely true. . . A strong inductive argument amounts to what attorneys
call a “preponderance of the evidence.” Occasionally a person who has been
convicted of a crime based on a preponderance of the evidence is later found
2.6. DEDUCTIVE REASONING AND ARGUMENT FORMS 101
to be innocent. This usually happens when new evidence is discovered that

incontrovertibly proves (i.e. shows through deductive means) that he or she
cannot be guilty. In a nutshell: inductive arguments can be wrong.
In contrast a deductive argument can only turn out to be wrong under
certain well-understood circumstances.
Like an inductive argument, a deductive argument is essentially just a
long sequence of statements; but there is some additional structure. The last
statement in the list is the conclusion – the statement to be proved – those
occurring before it are known as premises. Premises may be further subdi-
vided into (at least) five sorts: axioms, definitions, previously proved theo-
rems, hypotheses and deductions. Axioms and definitions are often glossed
over, indeed, they often go completely unmentioned (but rarely unused) in
a proof. In the interest of brevity this is quite appropriate, but conceptu-
ally, you should think of an argument as being based off of the axioms for
the particular area you are working in, and its standard definitions. A rote
knowledge of all the other theorems proved up to the one you are working
with would generally be considered excessive, but completely memorizing
the axioms and standard definitions of a field is essential. Hypotheses are a
funny class of premises, they are things which can be assumed true for the
sake of the current argument. For example, if the statement you are trying
to prove is a conditional, then the antecedent may be assumed true (if the
antecedent is false, then the conditional is automatically true!). You should
always be careful to list all hypotheses explicitly, and at the end of your
proof make sure that each and every hypothesis got used somewhere along
the way. If a hypothesis really isn’t necessary then you have proved a more
general statement (that’s a good thing).
Finally, deductions – I should note that the conclusion is also a deduction
– obey a very strict rule: every deduction follows from the premises that
have already been written down (this includes axioms and definitions that
probably won’t actually have been written, hypotheses and all the deductions
made up to this point) by one of the so-called rules of inference.
Each of the rules of inference actually amounts to a logical tautology that
has been re-expressed as a sort of re-writing rule. Each rule of inference will
be expressed in the sequel as a list of logical sentences that are assumed to
be among the premises of the argument, a horizontal bar, followed by the
symbol ∴ (which is usually voiced as the word “therefore”) and then a new
statement that can be placed among the deductions.
For example, one (very obvious) rule of inference is
A∧B
∴B
This rule is known as conjunctive simplification, and is equivalent to the

tautology (A ∧ B) =⇒ B.
The modus ponens rule11 is one of the most useful.
A
A =⇒ B
∴B
Modus ponens is related to the tautology (A ∧ (A =⇒ B)) =⇒ B.

Modus tollens is the rule of inference we get if we put modus ponens
through the “contrapositive” wringer.
¬B
A =⇒ B
∴ ¬A
11
Latin for “method of affirming”, the related modus tollens rule means “method of
denying.”
Modus tollens is related to the tautology (¬B ∧ (A =⇒ B)) =⇒ ¬A.

Modus ponens and modus tollens are also known as syllogisms. A syl-
logism is an argument form wherein a deduction follows from two premises.
There are two other common syllogisms, hypothetical syllogism and disjunc-
tive syllogism.
Hypothetical syllogism basically asserts a transitivity property for impli-
cations.
A =⇒ B
B =⇒ C
∴ A =⇒ C
Disjunctive syllogism can be thought of as a statement about alternatives,
but be careful to remember that in Logic, the disjunction always has the
inclusive sense.
A∨B
¬B
∴ A
Exercise. Convert the A ∨ B that appears in the premises of the disjunctive
syllogism rule into an equivalent conditional. How is the new argument form
related to modus ponens and/or modus tollens?
The word “dilemma” usually refers to a situation in which an individual

is faced with an impossible choice. A cute example known as the Crocodile’s
dilemma is as follows:
A crocodile captures a little boy who has strayed too near the
river. The child’s father appears and the crocodile tells him
“Don’t worry, I shall either release your son or I shall eat him.
If you can say, in advance, which I will do, then I shall release
him.” The father responds, “You will eat my son.” What should
the crocodile do?
In logical arguments the word dilemma is used in another sense having to

do with certain rules of inference. Constructive dilemma is a rule of inference
having to do with the conclusion that one of two possibilities must hold.
A =⇒ B
C =⇒ D
A∨C
∴ B∨D
Destructive dilemma is often not listed among the rules of inference be-
cause it can easily be obtained by using the constructive dilemma and re-
placing the implications with their contrapositives.
A =⇒ B
C =⇒ D
¬B ∨ ¬D
∴ ¬A ∨ ¬C
In Table 2.3, the ten most common rules of inference are listed. Note
that all of these are equivalent to tautologies that involve conditionals (as
opposed to biconditionals), every one of the basic logical equivalences that
we established in Section 2.3 is really a tautology involving a biconditional,
collectively these are known as the “rules of replacement.” In an argument,
any statement allows us to infer a logically equivalent statement. Or, put
differently, we could replace any premise with a different, but logically equiv-
alent, premise. You might enjoy trying to determine a minimal set of rules
of inference, that together with the rules of replacement would allow one to
form all of the same arguments as the ten rules in Table 2.3.
Name Form
A
Modus ponens A =⇒ B
∴B
¬B
Modus tollens A =⇒ B
∴ ¬A
A =⇒ B
Hypothetical syllogism B =⇒ C
∴ A =⇒ C
A∨B
Disjunctive syllogism ¬B
∴A
A =⇒ B
C =⇒ D
Constructive dilemma
A∨C
∴ B∨D
Table 2.3: The rules of inference.

Name Form
A =⇒ B
C =⇒ D
Destructive dilemma
¬B ∨ ¬D
∴ ¬A ∨ ¬C
A∧B
Conjunctive simplification
∴A
A
Conjunctive addition B
∴ A∧B
A
Disjunctive addition
∴ A∨B
A =⇒ B
Absorption
∴ A =⇒ (A ∧ B)
Table 2.3: The rules of inference. (continued)

Exercises — 2.6
1. In the movie “Monty Python and the Holy Grail” we encounter a me-
dieval villager who (with a bit of prompting) makes the following ar-
gument.
If she weighs the same as a duck, then she’s made of wood.

If she’s made of wood then she’s a witch.
Therefore, if she weighs the same as a duck, she’s a witch.
Which rule of inference is he using?
2. In constructive dilemma, the antecedants of the conditional sentences

are usually chosen to represent opposite alternatives. This allows us to
introduce their disjunction as a tautology. Consider the following proof
that there is never any reason to worry (found on the walls of an Irish
pub).
Either you are sick or you are well.

If you are well there’s nothing to worry about.
If you are sick there are just two possibilities:
Either you will get better or you will die.
If you are going to get better there’s nothing to worry about.
If you are going to die there are just two posibilities:
Either you will go to Heaven or to Hell.
If you go to Heaven there is nothing to worry about. If you go
to Hell, you’ll be so busy shaking hands with all your friends
there won’t be time to worry . . .
Identify the three tautologies that are introduced in this “proof.”

3. For each of the following arguments, write it in symbolic form and

determine which rules of inference are used.
(a) You are either with us, or you’re against us. And you don’t
appear to be with us. So, that means you’re against us!
(b) All those who had cars escaped the flooding. Sandra had a car –
therefore, Sandra escaped the flooding.
(c) When Johnny goes to the casino, he always gambles ’til he goes
broke. Today, Johnny has money, so Johnny hasn’t been to the
casino recently.
(d) (A non-constructive proof that there are √irrational numbers a

√ 2
and b such that √ab is rational.) Either 2 is rational or it is
√ 2 √
irrational. If 2 is rational, we let a √= b = 2. Otherwise, we
√ √2 √ √ √2 2
let a = 2 and b = 2. (Since 2 = 2, which is rational.)
It follows that in either case, there are irrational numbers a and b
such that ab is rational.
2.7. VALIDITY OF ARGUMENTS AND COMMON ERRORS 109
2.7 Validity of arguments and common errors

An argument is said to be valid or to have a valid form if each deduction
in it can be justified with one of the rules of inference listed in the previous
section. The form of an argument might be valid, but still the conclusion
may be false if some of the premises are false. So to show that an argument
is good we have to be able to do two things: show that the argument is valid
(i.e. that every step can be justified) and that the argument is sound which
means that all the premises are true. If you start off with a false premise,
you can prove anything!
Consider, for example the following “proof” that 2 = 1.
Suppose that a and b are two real numbers such that a = b.
by hypothesis, a and b are

equal, so
2
a = ab
subtracting b2 from both
sides
a2 − b2 = ab − b2
factoring both sides
(a + b)(a − b) = b(a − b)
canceling (a − b) from both
sides
a+b=b
Now let a and b both have a particular value, a = b = 1, and we

see that 1 + 1 = 1, i.e. 2 = 1.
This argument is not sound (thank goodness!) because one of the premises
– actually the bad premise appears as one of the justifications of a step – is
false. You can argue with perfect logic to achieve complete nonsense if you
include false premises.
Exercise. It is not true that you can always cancel the same thing from
both sides of an equation. Under what circumstances is such cancellation
disallowed?
So, how can you tell if an argument has a valid form? Use a truth table.
As an example, we’ll verify that the rule of inference known as “destructive
dilemma” is valid using a truth table. This argument form contains 4 pred-
icate variables so the truth table will have 16 rows. There is a column for
each of the variables, the premises of the argument and its conclusion.
A B C D A =⇒ B C =⇒ D ¬B ∨ ¬D ¬A ∨ ¬C
T T T T T T φ φ
T T T φ T φ T φ
T T φ T T T φ T
T T φ φ T T T T
T φ T T φ T T φ
T φ T φ φ φ T φ
T φ φ T φ T T T
T φ φ φ φ T T T
φ T T T T T φ T
φ T T φ T φ T T
φ T φ T T T φ T
φ T φ φ T T T T
φ φ T T T T T T
φ φ T φ T φ T T
φ φ φ T T T T T
φ φ φ φ T T T T
Now, mark the lines in which all of the premises of this argument form
are true. You should note that in every single situation in which all the
premises are true the conclusion is also true. That’s what makes “destructive
dilemma” – and all of its friends – a rule of inference. Whenever all the
premises are true so is the conclusion. You should also notice that there are
several rows in which the conclusion is true but some one of the premises isn’t.
That’s okay too, isn’t it reasonable that the conclusion of an argument can be
true, but at the same time the particulars of the argument are unconvincing?
As we’ve noted earlier, an argument by deductive reasoning can go wrong
in only certain well-understood ways. Basically, either the form of the ar-
gument is invalid, or at least one of the premises is false. Avoiding false
premises in your arguments can be trickier than it sounds – many state-
ments that sound appealing or intuitively clear are actually counter-factual.
The other side of the coin, being sure that the form of your argument is valid,
seems easy enough – just be sure to only use the rules of inference as found
in Table 2.3. Unfortunately most arguments that you either read or write
will be in prose, rather than appearing as a formal list of deductions. When
dealing with that setting – using natural rather than formalized language –
making errors in form is quite common.
Two invalid forms are usually singled out for criticism, the converse error
and the inverse error. In some sense these two apparently different ways
to screw up are really the same thing. Just as a conditional statement and
its contrapositive are known to be equivalent, so too are the other related
statements – the converse and the inverse – equivalent. The converse error
consists of mistaking the implication in a modus ponens form for its converse.
The converse error:
B
A =⇒ B
∴A
Consider, for a moment the following argument.
If a rhinoceros sees something on fire, it will stomp on it.

A rhinoceros stomped on my duck.
Therefore, the rhino must have thought that my duck was on fire.
It is true that rhinoceroses have an instinctive desire to extinguish fires.

Also, we can well imagine that if someone made this ridiculous argument
that their duck must actually have been crushed by a rhino. But, is the
conclusion that the duck was on fire justified? Not really, what the first
part of the argument asserts is that “(on fire) implies (rhino stomping)” but
couldn’t a rhino stomp on something for other reasons? Perhaps the rhino
was just ill-tempered. Perhaps the duck was just horrifically unlucky.
The closer the conditional is to being a biconditional, the more reason-
able sounding is an argument exhibiting the converse error. Indeed, if the
argument actually contains a biconditional, the “converse error” is not an
error at all.
The following is a perfectly valid argument, that (sadly) has a false
premise.
You will get an A in your Foundations class if and only if you

read Dr. Fields’ book.
You read Dr. Fields’ book.
Therefore, you will get an A in Foundations.
Suppose that we try changing the major premise of that last argument
to something more believable.
If you read Dr. Fields’ book, you will pass your Foundations class.
You did not read Dr. Fields’ book.
Therefore, you will not pass Foundations.
This last argument exhibits the so-called inverse error. It is by no means

meant as a guarantee, but nevertheless, it seems reasonable that if someone
reads this book they will pass a course on this material. The second premise
is also easy to envision as true, although the “you” that it refers to obviously
isn’t you, because you are reading this book! But even if we accept the
premises as true, the conclusion doesn’t follow. A person might have read
some other book that addressed the requisite material in an exemplary way.
Notice that the names for these two errors are derived from the change
that would have to be made to convert them to modus ponens. For example,
the inverse error is depicted formally by:
¬A
A =⇒ B
∴ ¬B
If we replaced the conditional in this argument form by its inverse (¬A =⇒

¬B) then the revised argument would be modus ponens. Similarly, if we re-
place the conditional in an argument that suffers from the converse error by
its converse, we’ll have modus ponens.
Exercises — 2.7
1. Determine the logical form of the following arguments. Use symbols to

express that form and determine whether the form is valid or invalid. If
the form is invalid, determine the type of error made. Comment on the
soundness of the argument as well, in particular, determine whether
any of the premises are questionable.
(a) All who are guilty are in prison.

George is not in prison.
Therefore, George is not guilty.
(b) If one eats oranges one will have high levels of vitamin C.
You do have high levels of vitamin C.
Therefore, you must eat oranges.
(c) All fish live in water.

The mackeral is a fish.
Therefore, the mackeral lives in water.
(d) If you’re lazy, don’t take math courses.

Everyone is lazy.
Therefore, no one should take math courses.
(e) All fish live in water.

The octopus lives in water.
Therefore, the octopus is a fish.
(f) If a person goes into politics, they are a scoundrel.

Harold has gone into politics.
Therefore, Harold is a scoundrel.
2. If we allow quantifiers and open sentences in an argument form we get

arguments that are termed “universal” and “particular.”
∀x, A(x) =⇒ B(x)

For example A(p) is the particular form of modus
∴ B(p)
ponens (here, p is not a variable – it stands for some particular element
∀x, A(x) =⇒ B(x)
of the universe of discourse) and ∀x, ¬B(x) is the universal
∴ ∀x, ¬A(x)
form of modus tollens.
Reexamine the arguments from problem (1), determine their forms (in-
cluding quantifiers) and whether they are universal or particular.
Chapter 3
Proof techniques I — Standard

methods
Love is a snowmobile racing across the tundra and then suddenly it flips over,
pinning you underneath. At night, the ice weasels come. –Matt Groening
3.1 Direct proofs of universal statements

If you form the product of 4 consecutive numbers, the result will be one less
than a perfect square. Try it!
1 · 2 · 3 · 4 = 24 = 52 − 1
2 · 3 · 4 · 5 = 120 = 112 − 1
3 · 4 · 5 · 6 = 360 = 192 − 1
It always works!
117
118 CHAPTER 3. PROOF TECHNIQUES I
The three calculations that we’ve carried out above constitute an induc-
tive argument in favor of the result. If you like we can try a bunch of further
examples,
13 · 14 · 15 · 16 = 43680 = 2092 − 1
14 · 15 · 16 · 17 = 571200 = 2392 − 1
but really, no matter how many examples we produce, we haven’t proved the
statement — we’ve just given evidence.
Generally, the first thing to do in proving a universal statement like this
is to rephrase it as a conditional. The resulting statement is a Universal
Conditional Statement or a UCS. The reason for taking this step is that the
hypotheses will then be clear – they form the antecedent of the UCS. So,
while you won’t have really made any progress in the proof by taking this
advice, you will at least know what tools you have at hand. Taking the
example we started with, and rephrasing it as a UCS we get
∀a, b, c, d ∈ Z, (a,b,c,d consecutive) =⇒ ∃k ∈ Z, a·b·c·d = k 2 − 1
The antecedent of the UCS is that a, b, c and d must be consecutive. By

concentrating our attention on what it means to be consecutive, we should
quickly realize that the original way we thought of the problem involved a
red herring. We don’t need to have variables for all four numbers; because
they are consecutive, a uniquely determines the other three. Finally we have
a version of the statement that we’d like to prove that should lend itself to
our proof efforts.
Theorem 3.1.1.
∀a ∈ Z, ∃k ∈ Z, a(a + 1)(a + 2)(a + 3) = k 2 − 1.

3.1. DIRECT PROOFS OF UNIVERSAL STATEMENTS 119
In this simplistic example, the only thing we need to do is come up with

a value for k given that we know what a is. In other words, a “proof” of this
statement involves doing some algebra.
Without further ado. . .
Proof: Suppose that a is a particular but arbitrarily chosen

integer. Consider the product of the 4 consecutive integers, a,
a + 1, a + 2 and a + 3. We would like to show that this product
is one less than the square of an integer k. Let k be a2 + 3a + 1.
First, note that
a(a + 1)(a + 2)(a + 3) = a4 + 6a3 + 11a2 + 6a.
Then, note that
k 2 − 1 = (a2 + 3a + 1)2 − 1
= (a4 + 6a3 + 11a2 + 6a + 1) − 1
= a4 + 6a3 + 11a2 + 6a.
Q.E.D.
Now, if you followed the algebra above, (none of which was particularly
difficult) the proof stands as a completely valid argument showing the truth of
our proposition, but this is very unsatisfying! All the real work was concealed
in one stark little sentence: “Let k be a2 + 3a + 1.” Where on Earth did
that particular value of k come from? The answer to that question should
hopefully convince you that there is a huge difference between devising a
proof and writing one. A good proof can sometimes be somewhat akin to a
good demonstration of magic, a magician doesn’t reveal the inner workings

of his trick, neither should a mathematician feel guilty about leaving out
some of the details behind the work! Heck, there are plenty of times when
you just have to guess at something, but if your guess works out, you can
write a perfectly correct proof.
In devising the proof above, we multiplied out the consecutive numbers
and then realized that we’d be done if we could find a polynomial in a
whose square was a4 + 6a3 + 11a2 + 6a + 1. Now, obviously, we’re going
to need a quadratic polynomial, and because the leading term is a4 and the
constant term is 1, it should be of the form a2 + ma + 1. Squaring this gives
a4 + 2ma3 + (m2 + 2)a2 + 2ma + 1 and comparing that result with what we
want, we pretty quickly realize that m had better be 3. So it wasn’t magic
after all!
This seems like a good time to make a comment on polynomial arith-
metic. Many people give up (or go searching for a computer algebra system)
when dealing with products of anything bigger than binomials. This is a
shame because there is an easy method using a table for performing such
multiplications. As an example, in devising the previous proof we needed to
form the product a(a + 1)(a + 2)(a + 3), now we can use the distributive law
or the infamous F.O.I.L rule to multiply pairs of these, but we still need to
multiply (a2 + a) with (a2 + 5a + 6). Create a table that has the terms of
these two polynomials as its row and column headings.
a2 5a 6
a2
a
Now, fill in the entries of the table by multiplying the corresponding row
and column headers.
a2 5a 6
2 4 3
a a 5a 6a2
a a3 5a2 6a
Finally add up all the entries of the table, combining any like terms.
You should note that the F.O.I.L rule is just a mnemonic for the case
when the table has 2 rows and 2 columns.
Okay, let’s get back to doing proofs. We are going to do a lot of proofs
involving the concepts of elementary number theory so, as a convenience,
all of the definitions that were made in Chapter 1 are gathered together in
Table 3.1.
Even
n is even ⇐⇒ ∃k ∈ Z, n = 2k
Odd
n is odd ⇐⇒ ∃k ∈ Z, n = 2k + 1
Divisibility
d|n ⇐⇒ ∃k ∈ Z, n = kd
Floor
y = bxc ⇐⇒ y ∈Z ∧ y ≤x<y+1
Ceiling
y = dxe ⇐⇒ y ∈Z ∧ y−1<x≤y
Quotient-remainder theorem, Div and Mod

∀n, d ∈ Z, ∃!q, r ∈ Z, n = qd + r ∧ 0 ≤ r < d
n div d = q
n mod d = r
Prime
∀ p > 1 ∈ Z+ , p is prime ⇐⇒
∀x, y ∈ Z+ , p = xy =⇒ x = 1 ∨ y = 1
Table 3.1: The definitions of elementary number theory restated.

In this section we are concerned with direct proofs of universal state-

ments. Such statements come in two flavors – those that appear to involve
conditionals, and those that don’t:
Every prime greater than two is odd.
versus
For all integers n, if n is a prime greater than two, then n is odd.
These two forms can readily be transformed one into the other, so we will
always concentrate on the later. A direct proof of a UCS always follows a
form known as “generalizing from the generic particular.” We are trying to
prove that ∀x ∈ U, P (x) =⇒ Q(x). The argument (in skeletal outline) will
look like:
Proof: Suppose that a is a particular but arbitrary ele-

ment of U such that P (a) holds.
..
.
Therefore Q(a) is true.

Thus we have shown that for all x in U , P (x) =⇒ Q(x).
Q.E.D.
Okay, so this outline is pretty crappy. It tells you how to start and end a
direct proof, but those obnoxious dot-dot-dots in the middle are where all the
real work has to go. If I could tell you (even in outline) how to fill in those
dots, that would mean mathematical proof isn’t really a very interesting ac-
tivity to engage in. Filling in those dots will sometimes (rarely) be obvious,
more often it will be extremely challenging; it will require great creativity,
loads of concentration, you’ll call on all your previous mathematical experi-

ences, and you will most likely experience a certain degree of anguish. Just
remember that your sense of accomplishment is proportional to the difficulty
of the puzzles you attempt. So let’s attempt another. . .
In Table 3.1 one of the very handy notions defined is that of the floor of
a real number.
y = bxc ⇐⇒ (y ∈ Z ∧ y ≤ x < y + 1).
There is a sad tendency for people to apply old rules in new situations
just because of a chance similarity in the notation. The brackets used in
notating the floor function look very similar to ordinary parentheses, so the
following “rule” is often proposed
bx + yc = bxc + byc
Exercise. Find a counterexample to the previous “rule.”
What is (perhaps) surprising is that if one of the numbers involved is an

integer then the “rule” really works.
Theorem 3.1.2.
∀x ∈ R, ∀n ∈ Z, bx + nc = bxc + bnc
Since the floor of an integer is that integer, we could restate this as

bx + nc = bxc + n.
Now, let’s try rephrasing this theorem as a UCS: If x is a real number
and n is an integer, then bx + nc = bxc + n. This is bad . . . it appears that
the only hypotheses that we can use involve what kinds of numbers x and
n are — our hypotheses aren’t particularly potent. The next most useful
ally in constructing proofs are the definitions of the concepts involved. The
quantity bxc appears in the theorem, lets make use of the definition:
a = bxc ⇐⇒ a ∈ Z ∧ a ≤ x < a + 1.
The only other floor function that appears in the statement of the theorem
(perhaps even more prominently) is bx + nc, here, the definition gives us
b = bx + nc ⇐⇒ b ∈ Z ∧ b ≤ x + n < b + 1.
These definitions are our only available tools so we’ll certainly have to
make use of them, and it’s important to notice that that is a good thing; the
definitions allow us to work with something well-understood (the inequalities
that appear within them) rather than with something new and relatively
suspicious (the floor notation). Putting the proof of this statement together
is an exercise in staring at the two definitions above and noting how one can
be converted into the other. It is also a testament to the power of naming
things.
Proof: Suppose that x is a particular but arbitrary real number

and that n is a particular but arbitrary integer. Let a = bxc.
By the definition of the floor function it follows that a is an
integer and a ≤ x < a + 1. By adding n to each of the parts
of this inequality we deduce a new (and equally valid) inequality,
a + n ≤ x + n < a + n + 1. Note that a + n is an integer and the
inequality above together with this fact constitute precisely the
definition of a + n = bx + nc. Finally, recalling that a = bxc (by
assumption), and rewriting, we obtain the desired result
bx + nc = bxc + n.
Q.E.D.
As we’ve seen in the examples presented in this section, coming up with

a proof can sometimes involve a bit of ingenuity. But, sometimes, there is
a “follow your nose” sort of approach that will allow you to devise a valid
argument without necessarily displaying any great leaps of genius! We close
this section with a few pieces of advice.
• Before anything else, determine precisely what hypotheses you can use.
• Jot down the definitions of anything in the statement of the theorem.
• There are 26 letters at your disposal (and even more if you know Greek)
(and you can always throw on subscripts!) don’t be stingy with letters.
The nastiest mistake you can make is to use the same variable for two
different things.
• Please write a rough draft first. Write two drafts! Even if you can write
beautiful, lucid prose on the first go around, it won’t fly when it comes
to organizing a proof.
• The statements in a proof are supposed to be logical statements. That

means they should be Boolean (statements that are either true or false).
An algebraic expression all by itself doesn’t count, an inequality or an
equality does.
• Don’t say “if” when you mean “since.” Really! If you start a proof
about rational numbers like so:
Proof: Suppose that x is a particular but arbitrary rational

number. If x is a rational number, it follows that . . .
people are going to look at you funny. What’s the point of supposing
that x is rational, then acting as if you’re in doubt of that fact by
writing “if”? You mean “since.”
• Mark off the beginning and the end of your proofs as a hint to your
readers. In this book we start off a proof by writing Proof: in italics
and we end every proof with the abbreviation Q.E.D.1
1
Quod erat demonstrandum or “(that) which was to be demonstrated.” some authors
prefer placing a small rectangle at the end of their proofs, but Q.E.D. seems more pompous.
Exercises — 3.1
1. Every prime number greater than 3 is of one of the two forms 6k + 1

or 6k + 5. What statement(s) could be used as hypotheses in proving
this theorem?
2. Prove that the sum of two rational numbers is a rational number.
3. Prove that if the sum of two integers is even, then so is their difference.
4. Prove that if x is an odd integer, then x2 is of the form 4k + 1.
5. Define the evenness of an integer n by:
evenness(n) = k ⇐⇒ 2k | n ∧ 2k+1 - n
State and prove a theorem concerning the evenness of products.
6. Suppose that a, b and c are integers such that a | b and b | c. Prove that
a | c.
7. Suppose that a, b, c and d are integers with a 6= c. Further, suppose

that x is a real number satisfying the equation
ax + b
= 1.
cx + d
Show that x is rational. Where is the hypothesis a 6= c used?
8. Show that if two positive integers a and b satisfy a | b and b | a then

they are equal.
3.2. MORE DIRECT PROOFS 129
3.2 More direct proofs

In creating a direct proof we need to look at our hypotheses, consider the
desired conclusion, and develop a strategy for transforming A into B. Quite
often you’ll find it easy to make several deductions from the hypotheses,
but none of them seems to be headed in the direction of the desired con-
clusion. The usual advice at this stage is “Try working backwards from the
conclusion.”
There is a lovely result known as the “arithmetic-geometric mean inequal-
ity” whose proof epitomizes this approach. Basically this inequality compares
two different ways of getting an “average” between two real numbers. The
arithmetic mean of two real numbers a and b is the one you’re probably used
to, (a + b)/2. Many people just call this the “mean” of a and b without using
the modifier “arithmetic” but as we’ll see, our notion of what intermediate
value to use in between two numbers is dependent on context. Consider the
following two sequences of numbers (both of which have a missing entry)
2 9 16 23 37 44
and
3 6 12 24 96 192.
How should we fill in the blanks?

The first sequence is an arithmetic sequence. Arithmetic sequences are
characterized by the property that the difference between successive terms
is a constant. The second sequence is a geometric sequence. Geometric se-
quences have the property that the ratio of successive terms is a constant.
The blank in the first sequence should be filled with the arithmetic mean
of the surrounding entries (23 + 37)/2 = 30. The blank in the second se-
quence should be filled using the geometric mean of its surrounding entries:
√
24 · 96 = 48.
Given that we accept the utility of having two inequivalent concepts of
mean that can be used in different contexts, it is interesting to see how
these two means compare to one another. The arithmetic-geometric mean
inequality states that the arithmetic mean is always bigger.
a+b √
∀a, b ∈ R, a, b ≥ 0 =⇒ ≥ ab
2
In proving this statement we have little choice but to work backwards
from the conclusion because the only hypothesis we have to work with is
that a and b are non-negative real numbers – which isn’t a particularly potent
tool. But what should we do? There isn’t a good response to that question,
we’ll just have to try a bunch of different things and hope that something
will work out. When we finally get around to writing up our proof though,
we’ll have to rearrange the statements in the opposite order from the way
they were discovered. This means that we would be ill-advised to make any
uni-directional inferences, we should strive to make biconditional connections
between our statements (or else try to intentionally make converse errors).
The first thing that appeals to your humble author is to eliminate both
the fractions and the radicals. . .
a+b √
≥ ab
2
√
⇐⇒ a + b ≥ 2 ab
⇐⇒ (a + b)2 ≥ 4ab
⇐⇒ a2 + 2ab + b2 ≥ 4ab
3.2. MORE DIRECT PROOFS 131
One of the steps above involves squaring both sides of an inequality. We

need to ask ourselves if this step is really reversible. In other words, is the
following conditional true?
√ √
∀x, y ∈ Rnoneg , x ≥ y =⇒ x≥ y
Exercise. Provide a justification for the previous implication.
What should we try next? There’s really no good justification for this
but experience working with quadratic polynomials either in equalities or
inequalities leads most people to try “moving everything to one side,” that
is, manipulating things so that one side of the equation or inequality is zero.
a2 + 2ab + b2 ≥ 4ab
⇐⇒ a2 − 2ab + b2 ≥ 0
Whoa! We’re done! Do you see why? If not, I’ll give you one hint: the
square of any real number is greater than or equal to zero.
Exercise. Re-assemble all of the steps taken in the previous few paragraphs
into a proof of the arithmetic-geometric mean inequality.
Exercises — 3.2
1. Suppose you have a savings account which bears interest compounded

monthly. The July statement shows a balance of $ 2104.87 and the
September statement shows a balance $ 2125.97. What would be the
balance on the (missing) August statement?
2. Recall that a quadratic equation ax2 + bx + c = 0 has two real solutions

if and only if the discriminant b2 − 4ac is positive. Prove that if a and
b have the same sign and c has a different sign (from a and b), then the
quadratic equation has (two) real solutions.
3. Prove that if x3 − x2 is negative then 3x + 4 < 7.
1
4. Show that if x is a positive real number, then x + x
≥ 2.
3.3. CONTRADICTION AND CONTRAPOSITION 133
3.3 Indirect proofs: contradiction and con-

traposition
Suppose we are trying to prove that all thrackles are polycyclic 2 . A direct
proof of this would involve looking up the definition of what it means to be
a thrackle, and of what it means to be polycyclic, and somehow discerning
a way to convert whatever thrackle’s logical equivalent is into the logical
equivalent of polycyclic. As happens fairly often, there may be no obvious
way to accomplish this task. Indirect proof takes a completely different
tack. Suppose you had a thrackle that wasn’t polycyclic, and furthermore,
show that this supposition leads to something truly impossible. Well, if it’s
impossible for a thrackle to not be polycyclic, then it must be the case that
all of them are. Such an argument is known as proof by contradiction.
Quite possibly the sweetest indirect proof known is Euclid’s proof that
there are an infinite number of primes.
Theorem 3.3.1. (Euclid) The set of all prime numbers is infinite.
Proof: Suppose on the contrary that there are only a finite

number of primes. This finite set of prime numbers could, in
principle, be listed in ascending order.
{p1 , p2 , p3 , . . . , pn }
Consider the number N formed by adding 1 to the product of all

of these primes.
n
Y
N =1+ pk
k=1
2
Both of these strange sounding words represent real mathematical concepts, however,
they don’t have anything to do with one another.
Clearly, N is much larger than the largest prime pn , so N cannot

be a prime number itself. Thus N must be a product of some
of the primes in the list. Suppose that pj is one of the primes
that divides N . Now notice that, by construction, N would leave
remainder 1 upon division by pj . This is a contradiction since we
cannot have both pj | N and pj - N .
Since the supposition that there are only finitely many primes
leads to a contradiction, there must indeed be an infinite number
of primes.
Q.E.D.
If you are working on proving a UCS and the direct approach seems to be
failing you may find that another indirect approach, proof by contraposition,
will do the trick. In one sense this proof technique isn’t really all that indirect;
what one does is determine the contrapositive of the original conditional and
then prove that directly. In another sense this method is indirect because
a proof by contraposition can usually be recast as a proof by contradiction
fairly easily.
The easiest proof I know of using the method of contraposition (and
possibly the nicest example of this technique) is the proof of the lemma we
√
stated in Section 1.6 in the course of proving that 2 wasn’t rational. In case
you’ve forgotten we needed the fact that whenever x2 is an even number, so
is x.
Let’s first phrase this as a UCS.
∀x ∈ Z, x2 even =⇒ x even
Perhaps you tried to prove this result earlier. If so you probably came
across the conceptual problem that all you have to work with is the evenness
of x2 which doesn’t give you much ammunition in trying to show that x is

even. The contrapositive of this statement is:
∀x ∈ Z, x not even =⇒ x2 not even
Now, since x and x2 are integers, there is only one alternative to being
even – so we can re-express the contrapositive as
∀x ∈ Z, x odd =⇒ x2 odd.
Without further ado, here is the proof:
Theorem 3.3.2.
∀x ∈ Z, x2 even =⇒ x even
Proof: This statement is logically equivalent to
∀x ∈ Z, x odd =⇒ x2 odd
so we prove that instead.

Suppose that x is a particular but arbitrarily chosen integer such
that x is odd. Since x is odd, there is an integer k such that
x = 2k + 1. It follows that x2 = (2k + 1)2 = 4k 2 + 4k + 1 =
2(2k 2 + 2k) + 1. Finally, we see that x2 must be odd because it
is of the form 2m + 1, where m = 2k 2 + 2k is clearly an integer.
Q.E.D.
Let’s have a look at a proof of the same statement done by contradiction.
Proof: We wish to show that
∀x ∈ Z, x2 even =⇒ x even.
Suppose to the contrary that there is an integer x such that x2

is even but x is odd.3 Since x is odd, there is an integer m such
that x = 2m + 1. Therefore, by simple arithmetic, we obtain
x2 = 4m2 + 4m + 1 which is clearly odd. This is a contradiction
because (by assumption) x2 is even.
Q.E.D.
The main problem in applying the method of proof by contradiction is

that it usually involves “cleverness.” You have to come up with some reason
why the presumption that the theorem is false leads to a contradiction –
and this may or may not be obvious. More than any other proof technique,
proof by contradiction demands that we use drafts and rewriting. After
monkeying around enough that we find a way to reach a contradiction, we
need to go back to the beginning of the proof and highlight the feature that
we will eventually contradict! After all, we want it to look like our proofs are
completely clear, concise and reasonable even if their formulation caused us
some sort of Gordian-level mental anguish.
We’ll end this section with an example from Geometry.
Theorem 3.3.3. Among all triangles inscribed in a fixed circle, the one with
maximum area is equilateral.
Proof: We’ll proceed by contradiction. Suppose to the contrary

that there is a triangle, 4ABC, inscribed in a circle having maxi-
mum area that is not equilateral. Since 4ABC is not equilateral,
there are two sides of it that are not equal. Without loss of gen-
erality, suppose that sides AB and BC have different lengths.
3
Recall that the negation of a UCS is an existentially quantified conjunction.
Consider the remaining side (AC) to be the base of this trian-

gle. We can construct another triangle 4AB 0 C, also inscribed
in our circle, and also having AC as its base, having a greater
altitude than 4ABC — since the area of a triangle is given by
the formula bh/2 (where b is the base, and h is the altitude), this
triangle’s area is evidently greater than that of 4ABC. This is a
contradiction since 4ABC was presumed to have maximal area.
We leave the actual construction 4AB 0 C to the following exer-
cise.
Q.E.D.
Exercise. Where should we place the point B 0 in order to create a triangle

4AB 0 C having greater area than any triangle such as 4ABC which is not
isosceles?
B
A C
Exercises — 3.3
1. Prove that if the cube of an integer is odd, then that integer is odd.
2. Prove that whenever a prime p does not divide the square of an integer,
it also doesn’t divide the original integer. (p - x2 =⇒ p - x)
3. Prove (by contradiction) that there is no largest integer.
4. Prove (by contradiction) that there is no smallest positive real number.
5. Prove (by contradiction) that the sum of a rational and an irrational

number is irrational.
6. A Pythagorean triple is a set of three natural numbers, a, b and c, such

that a2 + b2 = c2 . Prove that, in a Pythagorean triple, at least one
of a and b is even. Use either a proof by contradiction or a proof by
contraposition.
3.4. DISPROOFS OF UNIVERSAL STATEMENTS 139
3.4 Disproofs of universal statements

The idea of a “disproof” is really just semantics – in order to disprove a
statement we need to prove its negation.
So far we’ve been discussing proofs quite a bit, but have paid very little
attention to a really huge issue. If the statements we are attempting to
prove are false, no proof is ever going to be possible. Really, a prerequisite
to developing a facility with proofs is developing a good “lie detector.” We
need to be able to guess, or quickly ascertain, whether a statement is true or
false. If we are given a universally quantified statement the first thing to do
is try it out for some random elements of the universe we’re working in. If we
happen across a value that satisfies the statement’s hypotheses but doesn’t
satisfy the conclusion, we’ve found what is known as a counterexample.
Consider the following statement about integers and divisibility:
Conjecture 3.
∀a, b, c ∈ Z, a | bc =⇒ a | b ∨ a | c.
This is phrased as a UCS, so the hypothesis is clear, we’re looking for

three integers so that the first divides the product of the other two. In the
following table we have collected several values for a, b and c such that a | bc.
a b c a|b ∨ a|c ?
2 7 6 yes
2 4 5 yes
3 12 11 yes
3 5 15 yes
5 4 15 yes
5 10 3 yes
7 2 14 yes
Exercise. As noted in Section 1.2 the statement above is related to whether

or not a is prime. Note that in the table, only prime values of a appear. This
is a rather broad hint. Find a counterexample to Conjecture 3.
There can be times when the search for a counterexample starts to feel re-
ally futile. Would you think it likely that a statement about natural numbers
could be true for (more than) the first 50 numbers a yet still be false?
Conjecture 4.
∀n ∈ Z+ n2 − 79n + 1601 is prime.
Exercise. Find a counterexample to Conjecture 4
Hidden within Euclid’s proof of the infinitude of the primes is a sequence.

Recall that in the proof we deduced a contradiction by considering the num-
ber N defined by
n
Y
N =1+ pk .
k=1
Define a sequence by
n
Y
Nn = 1 + pk ,
k=1
where {p1 , p2 , . . . , pn } are the actual first n primes. The first several values
of this sequence are:
n Nn
1 1 + (2) = 3
2 1 + (2 · 3) = 7
3 1 + (2 · 3 · 5) = 31
4 1 + (2 · 3 · 5 · 7) = 211
5 1 + (2 · 3 · 5 · 7 · 11) = 2311
.. ..
. .
Now, in the proof, we deduced a contradiction by noting that Nn is much

larger than pn , so if pn is the largest prime it follows that Nn can’t be prime
– but what really appears to be the case (just look at that table!) is that Nn
actually is prime for all n.
Qn
Exercise. Find a counterexample to the conjecture that 1 + k=1 pk is itself
always a prime.
Exercises — 3.4
1. Find a polynomial that assumes only prime values for a reasonably

large range of inputs.
2. Find a counterexample to Conjecture 3 using only powers of 2.
3. The alternating sum of factorials provides an interesting example of a

sequence of integers.
1! = 1
2! − 1! = 1
3! − 2! + 1! = 5
4! − 3! + 2! − 1! = 19
et cetera
Are they all prime? (After the first two 1’s.)
4. It has been conjectured that whenever p is prime, 2p − 1 is also prime.

Find a minimal counterexample.
5. True or false: The sum of any two irrational numbers is irrational.

Prove your answer.
6. True of false: There are two irrational numbers whose sum is rational.
Prove your answer.
7. True or false: The product of any two irrational numbers is irrational.

Prove your answer.
8. True or false: There are two irrational numbers whose product is ra-
tional. Prove your answer.
9. True or false: Whenever an integer n is a divisor of the square of an

integer, m2 , it follows that n is a divisor of m as well. (In symbols,
∀n ∈ Z, ∀m ∈ Z, n | m2 =⇒ n | m.) Prove your answer.
3.5 Even more direct proofs: By cases and

By exhaustion
Proof by exhaustion is the least attractive proof method from an aesthetic
perspective. An exhaustive proof consists of literally (and exhaustively)
checking every element of the universe to see if the given statement is true
for it. Usually, of course, this is impossible because the universe of discourse
is infinite; but when the universe of discourse is finite, one certainly can’t
argue the validity of an exhaustive proof.
In the last few decades the introduction of powerful computational assis-
tance for mathematicians has lead to a funny situation. There is a growing
list of important results that have been “proved” by exhaustion using a com-
puter. Important examples of this phenomenon are the non-existence of a
projective plane of order 10[10] and the only known value of a Ramsey num-
ber for hypergraphs[13].
Proof by cases is subtly different from exhaustive proof – for one thing a
valid proof by cases can be used in an infinite universe. In a proof by cases
one has to divide the universe of discourse into a finite number of sets4 and
then provide a separate proof for each of the cases. A great many statements
about the integers can be proved using the division of integers into even and
odd. Another set of cases that is used frequently is the finite number of
possible remainders obtained when dividing by an integer d. (Note that even
and odd correspond to the remainders 0 and 1 obtained after division by 2.)
A very famous instance of proof by cases is the computer-assisted proof
of the four color theorem. The four color theorem is a result known to map
makers for quite some time that says that 4 colors are always sufficient to
color the nations on a map in such a way that countries sharing a boundary
4
It is necessary to provide an argument that this list of cases is complete! I.e. that
every element of the universe falls into one of the cases.
3.5. BY CASES AND BY EXHAUSTION 145
are always colored differently. Figure 3.1 shows one instance of an arrange-
ment of nations that requires at least four different colors, the theorem says
that four colors are always enough. It should be noted that real cartogra-
phers usually reserve a fifth color for oceans (and other water) and that it is
possible to conceive of a map requiring five colors if one allows the nations to
be non-contiguous. In 1977, Kenneth Appel and Wolfgang Haken proved the
four color theorem by reducing the infinitude of possibilities to 1,936 sepa-
rate cases and analyzing each of these with a computer. The inelegance of a
proof by cases is probably proportional to some power of the number of cases,
but in any case, this proof is generally considered somewhat inelegant. Ever
since the proof was announced there has been an ongoing effort to reduce the
number of cases (currently the record is 633 cases – still far too many to be
checked through without a computer) or to find a proof that does not rely
on cases. For a good introductory article on the four color theorem see[6].
Figure 3.1: The nations surrounding Luxembourg show that sometimes 4

colors are required in cartography.
Most exhaustive proofs of statements that aren’t trivial tend to either

be (literally) too exhausting or to seem rather contrived. One example of a
situation in which an exhaustive proof of some statement exists is when the
statement is thought to be universally true but no general proof is known –
yet the statement has been checked for a large number of cases. Goldbach’s
conjecture is one such statement. Christian Goldbach [4] was a mathemati-
cian born in Königsberg Prussia, who, curiously, did not make the conjecture5
which bears his name. In a letter to Leonard Euler, Goldbach conjectured
that every odd number greater than 5 could be expressed as the sum of three
primes (nowadays this is known as the weak Goldbach conjecture). Euler
apparently liked the problem and replied to Goldbach stating what is now
known as Goldbach’s conjecture: Every even number greater than 2 can be
expressed as the sum of two primes. This statement has been lying around
since 1742, and a great many of the world’s best mathematicians have made
their attempts at proving it – to no avail! (Well, actually a lot of progress
has been made but the result still hasn’t been proved.) It’s easy to verify
the Goldbach conjecture for relatively small even numbers, so what has been
done is/are proofs by exhaustion of Goldbach’s conjecture restricted to finite
universes. As of this writing, the conjecture has been verified to be true of
all even numbers less than 2 × 1017 .
Whenever an exhaustive proof, or a proof by cases exists for some state-
ment it is generally felt that a direct proof would be more esthetically pleas-
ing. If you are in a situation that doesn’t admit such a direct proof, you
should at least seek a proof by cases using the minimum possible number of
cases. For example, consider the following theorem and proof.
Theorem 3.5.1. ∀n ∈ Z n2 is of the form 4k or 4k + 1 for some k ∈ Z.
Proof: We will consider the four cases determined by the four

possible residues mod 4.
case i) If n ≡ 0 (mod 4) then there is an integer m such that n =

4m. It follows that n2 = (4m)2 = 16m2 is of the form 4k
where k is 4m2 .
5
This conjecture was discussed previously in the exercises of Section 1.2
case ii) If n ≡ 1 (mod 4) then there is an integer m such that n =

4m + 1. It follows that n2 = (4m + 1)2 = 16m2 + 8m + 1 is
of the form 4k + 1 where k is 4m2 + 2m.
case iii) If n ≡ 2 (mod 4) then there is an integer m such that n =

of the form 4k where k is 4m2 + 4m + 1.
case iv) If n ≡ 3 (mod 4) then there is an integer m such that n =

of the form 4k + 1 where k is 4m2 + 6m + 2.
Since these four cases exhaust the possibilities and since the de-
sired result holds in each case, our proof is complete.
Q.E.D.
While the proof just stated is certainly valid, the argument is inelegant
since a smaller number of cases would suffice.
Exercise. The previous theorem can be proved using just two cases. Do so.
We’ll close this section by asking you to determine an exhaustive proof

where the complexity of the argument is challenging but not too impossible.
Consider the game called “binary determinant tic-tac-toe”6 which is played
by two players who alternately fill in the entries of a 3 × 3 array. Player
One goes first, placing 1’s in the array and player Zero goes second, playing
0’s. Player One’s goal is that the final array have determinant 1, and player
Zero’s goal is that the determinant be 0. The determinant of a 3 × 3 array is
6
This question was problem A4 in the 63rd annual William Lowell Putnam Math-
ematics Competition (2002). There are three collections of questions and answers from
previous Putnam exams available from the MAA [1, 7, 9]
computed by summing a bunch of products. The products are of the entries

along diagonal swaths through the array (diagonals that run off the edge of
the array reappear on the other side) – with a sign convention: diagonals
that go down to the right are counted as positive and diagonals that go up
to the right are negative. This sounds a bit confusing, but a diagram and an
example should clear that up.
The first confusing issue is that the diagonals can “wrap around” the
array, this can be taken care of by replicating the first and second columns of
the array, to the right of the original array. Let’s follow an example. suppose
we wished to find the determinant of the following 3 × 3 array (note that this
array isn’t binary – so we can follow the computation more easily.)
 
1 2 3
 
 4 5 6 
 
7 8 9
In Figure 3.2 we see the following work being performed. First, the array
is expanded by copying the first two columns to its right. Next, the products
along the six diagonals are computed. Finally, the determinant is computed
by adding up these numbers, but with minus signs on the products from the
back diagonals.
The determinant of our example is 96 + 84 + 45 − 105 − 48 − 72 = 0.
In “binary determinant tic-tac-toe” all the computations are done mod 2, so
the answers are either 0 or 1.
Now that we’ve explained the setup, here’s the question: In ordinary tic-
tac-toe, there is a non-losing strategy for the player who goes first – is this
the case in this new game?
Exercise. Show that player Zero can always win a game of binary determi-
nant tic-tac-toe by the method of exhaustion.
1 2 3
4 5 6
7 8 9
1 2 3 1 2
4 5 6 4 5
7 8 9 7 8
1 2 3 1 2
4 5 6 4 5
7 8 9 7 8
105 96
48 84
72 45
Figure 3.2: Finding the determinant of a 3 × 3 array.

Exercises — 3.5
1. Prove that if n is an odd number then n4 (mod 16) = 1.
2. Prove that every prime number other than 2 and 3 has the form 6q + 1
or 6q + 5 for some integer q. (Hint: this problem involves thinking
about cases as well contrapositives.)
3. Show that the sum of any three consecutive integers is divisible by 3.
4. A vampire number is a 2n digit number v that factors as v = xy where

x and y are n digit numbers and the digits of v are the union of the
digits in x and y in some order. The numbers x and y are prosaically
known as the “fangs” of v. To eliminate trivial cases, pairs of trailing
zeros are disallowed.
Show that there are no 2-digit vampire numbers.
Show that there are seven 4-digit vampire numbers.
5. Lagrange’s theorem on representation of integers as sums of squares

says that every positive integer can be expressed as the sum of at most
4 squares. For example, 79 = 72 + 52 + 22 + 12 . Show (exhaustively)
that 15 can not be represented using fewer than 4 squares.
6. Show that there are exactly 17 numbers x in the range 1 ≤ x ≤ 100

that can’t be represented using fewer than 4 squares.
7. The trichotomy property of the real numbers simply states that every
real number is either positive or negative or zero. Trichotomy can be
used to prove many statements by looking at the three cases that it
guarantees. Develop a proof (by cases) that the square of any real
number is non-negative.
3.6. EXISTENTIAL STATEMENTS 151
3.6 Proofs and disproofs of existential state-

ments
From a certain point of view, there is no need for the current section. If
we are proving an existential statement we are disproving some universal
statement. (Which has already been discussed.) Similarly, if we are trying
to disprove an existential statement, then we are actually proving a related
universal statement. Nevertheless, sometimes the way a theorem is stated
emphasizes the existence question over the corresponding universal – and so
people talk about proving and disproving existential statements as a separate
issue from universal statements.
Proofs of existential questions come in two basic varieties: constructive

and non-constructive. Constructive proofs are conceptually the easier of the
two – you actually name an example that shows the existential question is
true. For example:
Theorem 3.6.1. There is an even prime.
Proof: The number 2 is both even and prime.

Q.E.D.
Exercise. The Fibonacci numbers are defined by the initial values F (0) = 1
and F (1) = 1 and the recursive formula F (n + 1) = F (n) + F (n − 1) (to get
the next number in the series you add the last and the penultimate).
n F (n)
0 1
1 1
2 2
3 3
4 5
5 8
.. ..
. .
Prove that there is a Fibonacci number that is a perfect square.
A nonconstructive existence proof is trickier. One approach is to argue

by contradiction – if the thing we’re seeking doesn’t exist that will lead to an
absurdity. Another approach is to outline a search algorithm for the desired
item and provide an argument as to why it cannot fail!
A particularly neat approach is to argue using dilemma. This is my
favorite non-constructive existential theorem/proof.
Theorem 3.6.2. There are irrational numbers α and β such that αβ is

rational.
√
√ 2 √
Proof: If 2is rational
√
then we are done. (Let α = β = 2.)
√ 2 √
Otherwise,
√
let α = 2 and β = 2. The result follows because
√ √2 2 √ (√2√2) √ 2
2 = 2 = 2 = 2, which is clearly rational.
Q.E.D.
Many existential proofs involve a property of the natural numbers known

as the well-ordering principle. The well-ordering principle is sometimes ab-
breviated WOP. If a set has WOP it doesn’t mean that the set is ordered
in a particularly good way, but rather that its subsets are like wells – the
kind one hoists water out of with a bucket on a rope. You needn’t be con-
cerned with WOP in general at this point, but notice that the subsets of the
natural numbers have a particularly nice property – any non-empty set of
natural numbers must have a least element (much like every water well has
a bottom).
Because the natural numbers have the well-ordering principle we can
prove that there is a least natural number with property X by simply finding
any natural number with property X – by doing that we’ve shown that the
set of natural numbers with property X is non-empty and that’s the only
hypothesis the WOP needs.
For example, in the exercises in Section 3.5 we introduced vampire num-
bers. A vampire number is a 2n digit number v that factors as v = xy where
x and y are n digit numbers and the digits of v are the union of the digits in
x and y in some order. The numbers x and y are known as the “fangs” of v.
To eliminate trivial cases, pairs of trailing zeros are disallowed.
Theorem 3.6.3. There is a smallest 6-digit vampire number.
Proof: The number 125460 is a vampire number (in fact this

is the smallest example of a vampire number with two sets of
fangs: 125460 = 204 · 615 = 246 · 510). Since the set of 6-digit
vampire numbers is non-empty, the well-ordering principle of the
natural numbers allows us to deduce that there is a smallest 6-
digit vampire number.
Q.E.D.
This is quite an interesting situation in that we know there is a smallest

6-digit vampire number without having any idea what it is!
Exercise. Show that 102510 is the smallest 6-digit vampire number.

There are quite a few occasions when we need to prove statements in-
volving the unique existence quantifier (∃!). In such instances we need to do
just a little bit more work. We need to show existence – either constructively
or non-constructively – and we also need to show uniqueness. To give an
example of a unique existence proof we’ll return to a concept first discussed
in Section 1.5 and finish-up some business that was glossed-over there.
Recall the Euclidean algorithm that was used to calculate the greatest
common divisor of two integers a and b (which we denote gcd(a, b)). There
is a rather important question concerning algorithms known as the “halting
problem.” Does the program eventually halt, or does it get stuck in an
infinite loop? We know that the Euclidean algorithm halts (and outputs the
correct result) because we know the following unique existence result.
∀a, b ∈ Z+ , ∃! d ∈ Z+ such that d = gcd(a, b)
Now, before we can prove this result, we’ll need a precise definition for
gcd(a, b). Firstly, a gcd must be a common divisor which means it needs to
divide both a and b. Secondly, among all the common divisors, it must be the
largest. This second point is usually addressed by requiring that every other
common divisor divides the gcd. Finally we should note that a gcd is always
positive, for whenever a number divides another number so does its negative,
and whichever of those two is positive will clearly be the greater! This allows
us to extend the definition of gcd to all integers, but things are conceptually
easier if we keep our attention restricted to the positive integers.
Definition. The greatest common divisor, or gcd, of two positive integers

a and b is a positive integer d such that d | a and d | b and if c is any other
positive integer such that c | a and c | b then c | d.
∀a, b, c, d ∈ Z+ d = gcd(a, b) ⇐⇒ d | a ∧ d | b ∧ (c | a ∧ c | b =⇒ c | d)
Armed with this definition, let’s return our attention to proving the
unique existence of the gcd. The uniqueness part is easier so we’ll do that
first. We argue by contradiction. Suppose that there were two different num-
bers d and d0 satisfying the definition of gcd(a, b). Put d0 in the place of c
in the definition to see that d0 | d. Similarly, we can deduce that d | d0 and
if two numbers each divide into the other, they must be equal. This is a
contradiction since we assumed d and d0 were different.
For the existence part we’ll need to define a set – known as the Z-module
generated by a and b – that consists of all numbers of the form xa + yb where
x and y range over the integers.
This set has a very nice geometric character that often doesn’t receive the
attention it deserves. Every element of a Z-module generated by two numbers
(15 and 21 in the example) ) corresponds to a point in the Euclidean plane.
As indicated in Figure 3.3 there is a dividing line between the positive and
negative elements in a Z-module. It is also easy to see that there are many
repetitions of the same value at different points in the plane.
Exercise. The value 0 clearly occurs in a Z-module when both x and y are
themselves zero. Find another pair of (x, y) values such that 21x + 15y is
zero. What is the slope of the line which separates the positive values from
the negative in our Z-module?
In thinking about this Z-module, and perusing Figure 3.3, you may have
noticed that the smallest positive number in the Z-module is 3. If you hadn’t
noticed that, look back and verify that fact now.
Exercise. How do we know that some smaller positive value (a 1 or a 2)

doesn’t occur somewhere in the Euclidean plane?
What we’ve just observed is a particular instance of a general result.
Theorem 3.6.4. The smallest positive number in the Z-module generated by

a and b is d = gcd(a, b).
-24 -3 18 39 60 81 102 123 144 165
-39 -18 3 24 45 66 87 108 129 150
-54 -33 -12 9 30 51 72 93 114 135
-69 -48 -27 -6 15 36 57 78 99 120
-84 -63 -42 -21 0 21 42 63 84 105
-99 -78 -57 -36 -15 6 27 48 69 90
-114 -93 -72 -51 -30 -9 12 33 54 75
-129 -108 -87 -66 -45 -24 -3 18 39 60
-144 -123 -102 -81 -60 -39 -18 3 24 45
Figure 3.3: The Z-module generated by 21 and 15. The number 21x + 15y
is printed by the point (x, y).
Proof: Suppose that d is the smallest positive number in the Z-

module {xa+yb x, y ∈ Z}. There are particular values of x and y
(which we will distinguish with over-lines) such that d = xa + yb.
Now, it is easy to see that if c is any common divisor of a and b
then c | d, so what remains to be proved is that d itself is a divisor
of both a and b. Consider dividing d into a. By the division
algorithm there are uniquely determined numbers q and r such
that a = qd+r with 0 ≤ r < d. We will show that r = 0. Suppose,
to the contrary, that r is positive. Note that we can write r as
r = a − qd = a − q(xa + yb) = (1 − qx)a − qyb. The last equality
shows that r is in the Z-module under consideration, and so, since
d is the smallest positive integer in this Z-module it follows that
r ≥ d which contradicts the previously noted fact that r < d.
Thus, r = 0 and so it follows that d | a. An entirely analogous
argument can be used to show that d | b which completes the proof
that d = gcd(a, b).
Q.E.D.
Exercises — 3.6
1. Show that there is a perfect square that is the sum of two perfect
squares.
2. Show that there is a perfect cube that is the sum of three perfect cubes.
3. Show that the WOP doesn’t hold in the integers. (This is an existence
proof, you show that there is a subset of Z that doesn’t have a smallest
element.)
4. Show that the WOP doesn’t hold in Q+ .
5. In the proof of Theorem 3.6.4 we weaseled out of showing that d | b.

Fill in that part of the proof.
6. Give a proof of the unique existence of q and r in the division algorithm.
7. A digraph is a drawing containing a collection of points that are con-

nected by arrows. The game known as scissors-paper-rock can be rep-
resented by a digraph that is balanced (each point has the same number
of arrows going out as going in). Show that there is a balanced digraph
having 5 points.
scissors
cuts
paper
smashes
covers
rock
Chapter 4
Sets
No more turkey, but I’d like some more of the bread it ate. –Hank Ketcham
4.1 Basic notions of set theory

In modern mathematics there is an area called Category theory1 which stud-
ies the relationships between different areas of mathematics. More precisely,
the founders of category theory noticed that essentially the same theorems
and proofs could be found in many different mathematical fields – with only
the names of the structures involved changed. In this sort of situation one
can make what is known as a categorical argument in which one proves the
desired result in the abstract, without reference to the details of any par-
ticular field. In effect this allows one to prove many theorems at once – all
you need to convert an abstract categorical proof into a concrete one rele-
vant to a particular area is a sort of key or lexicon to provide the correct
names for things. Now, category theory probably shouldn’t really be stud-
ied until you have a background that includes enough different fields that
1
The classic text by Saunders Mac Lane [11] is still considered one of the best intro-
ductions to Category theory.
159
160 CHAPTER 4. SETS
you can make sense of their categorical correspondences. Also, there are a
good many mathematicians who deride category theory as “abstract non-
sense” and the whole field has fallen out of favor in recent years. But, as
someone interested in developing a facility with proofs, you should be on
the lookout for categorical correspondences. If you ever hear yourself utter
a sentence like “Well, the proof of that goes just like the proof of the (insert
weird technical-sounding name here) theorem.” you are probably noticing a
categorical correspondence.
Okay, so category theory is currently somewhat unpopular, won’t be of
much use to you until much later in your mathematical career (if at all), and
doesn’t really save that much effort (Why not just do two or three different
proofs instead of learning a whole new field so we can combine them into
one?) And yet, category theory is being mentioned here at the beginning of
the chapter on sets. Why?
We are about to see our first example of a categorical correspondence.
Logic and Set theory are different aspects of the same thing. To describe a
set people usually quote Kurt Gödel – “A set is a Many that allows itself to
be thought of as a One.” (Note how the attempt at defining what is really an
elemental, undefinable concept ends up sounding rather mystical.) A more
practical approach is to think of a set as the collection of things that make
some open sentence true.2
Recall that in Logic the atomic concepts were “true”, “false”, “sentence”
and “statement.” In Set theory, they are “set”, “element” and “member-
ship.” These concepts (more or less) correspond to one another. In most
books, a set is denoted either using the letter M (which stands for the Ger-
man word “menge”) or early alphabet capital roman letters – A, B, C, et
2
This may sound less metaphysical, but this statement is also faulty because it defines
“set” in terms of “collection” – which will of course be defined elsewhere as “the sort of
things of which sets are one example.”
4.1. BASIC NOTIONS OF SET THEORY 161
cetera. Here, we will often emphasize the connection between sets and open
sentences in Logic by using a subscript notation. The set that corresponds to
the open sentence P (x) will be denoted SP , we call SP the truth set of P (x).
SP = {x P (x)}
On the other hand, when we have a set given in the absence of any open
sentence, we’ll be happy to use the early alphabet, capital roman letters
convention – or frankly, any other letters we feel like! Whenever we have a
set A given, it is easy to state a logical open sentence that would correspond
to it. The membership question: MA (x) = “Is x in the set A?” Or, more
succinctly, MA (x) = “x ∈ A”. Thus the atomic concept “true” from Logic
corresponds to the answer “yes” to the membership question in Set theory
(and of course “false” corresponds to “no”).
There are many interesting foundational issues that are going to be sidestepped
in our current development of Set theory. For instance, recall that in Logic
we always worked inside some “universe of discourse.” As a consequence of
the approach we are taking now, all of our set theoretic work will be done
within some unknown “universal” set. Attempts at specifying (a priori ) a
universal set for doing mathematics within are doomed to failure. In the
early days of the twentieth century they attempted to at least get Set theory
itself on a firm footing by defining the universal set to be “the set of all sets”
– an innocuous sounding idea that had funny consequences (we’ll investigate
this in Section 4.5).
In Logic we had “sentences” and “statements,” the latter were distin-
guished as having definite truth values. The corresponding thing in Set
theory is that sets have the property that we can always tell whether a given
object is or is not in them. If it ever becomes necessary to talk about “sets”
where we’re not really sure what’s in them we’ll use the term collection.
You should think of a set as being an unordered collection of things, thus
162 CHAPTER 4. SETS
{popover, 1, froggy} and {1, froggy, popover} are two ways to represent the
same set. Also, a set either contains, or doesn’t contain, a given element. It
doesn’t make sense to have an element in a set multiple times. By conven-
tion, if an element is listed more than once when a set is listed we ignore
the repetitions. So, the sets {1, 1} and {1} are really the same thing. If the
notion of a set containing multiple instances of its elements is needed there is
a concept known as a multiset that is studied in Combinatorics. In a multi-
set, each element is preceded by a so-called repetition number which may be
the special symbol ∞ (indicating an unlimited number of repetitions). The
multiset concept is useful when studying puzzles like “How many ways can
the letters of MISSISSIPPI be rearranged?” because the letters in MISSIS-
SIPPI can be expressed as the multiset {1 · M, 4 · I, 2 · P, 4 · S}. With the
exception of the following exercise, in the remainder of this chapter we will
only be concerned with sets, never multisets.
Exercise. (Not for the timid!) How many ways can the letters of MISSIS-
SIPPI be arranged?
If a computer scientist were seeking a data structure to implement the

notion of “set,” he’d want a sorted list where repetitions of an entry were
somehow disallowed. We’ve already noted that a set should be thought of as
an unordered collection, and yet it’s been asserted that a sorted list would
be the right vehicle for representing a set on a computer. Why? One reason
is that we’d like to be able to tell (quickly) whether two sets are the same or
not. If the elements have been presorted it’s easier.
Consider the difficulty in deciding whether the following two sets are
equal.
S1 = {♠, 1, e, π, ♦, A, Ω, h, ⊕, }
S2 = {A, 1, , π, e, s, ⊕, ♠, Ω, ♦}
If instead we compare them after they’ve been sorted, the job is much easier.
S1 = {1, A, ♦, e, , h, Ω, ⊕, π, ♠}
S2 = {1, A, ♦, e, , Ω, ⊕, π, s, ♠}
This business about ordered versus unordered comes up fairly often so

it’s worth investing a few moments to figure out how it works. If a collection
of things that is inherently unordered is handed to us we generally put them
in an order that is pleasing to us. Consider receiving five cards from the
dealer in a card game, or extracting seven letters from the bag in a game
of Scrabble. If, on the other hand, we receive a collection where order is
important we certainly may not rearrange them. Imagine someone receiving
the telephone number of an attractive other but writing it down with the
digits sorted in increasing order!
Exercise. Consider a universe consisting of just the first 5 natural numbers

U = {1, 2, 3, 4, 5}. How many different sets having 4 elements are there
in this universe? How many different ordered collections of 4 elements are
there?
The last exercise suggests an interesting question. If you have a universal

set of some fixed (finite) size, how many different sets are there? Obviously
you can’t have any more elements in a set than are in your universe. What’s
the smallest possible size for a set? Many people would answer 1 – which
isn’t unreasonable! – after all a set is supposed to be a collection of things,
and is it really possible to have a collection with nothing in it? The standard
answer is 0 however, mostly because it makes a certain counting formula
work out nicely. A set with one element is known as a singleton set (note
the use of the indefinite article). A set with no elements is known as the
empty set (note the definite article). There are as many singletons as there
164 CHAPTER 4. SETS
are elements in your universe. They aren’t the same though, for example
1 6= {1}. There is only one empty set and it is denoted ∅ – irrespective of
the universe we are working in.
Let’s have a look at a small example, suppose we have a universal set with
3 elements, without loss of generality, {1, 2, 3}. It’s possible to construct a
set, whose elements are all the possible sets in this universe. This set is
known as the power set of the universal set. Indeed, we can construct the
power set of any set A and we denote it with the symbol P(A). Returning
to our example we have

P({1, 2, 3}) = ∅,
{1}, {2}, {3},
{1, 2}, {1, 3}, {2, 3},

{1, 2, 3} .
Exercise. Find the power sets P({1, 2}) and P({1, 2, 3, 4}). Conjecture a
formula for the number of elements (these are, of course, sets) in P({1, 2, . . . n}).
Hint: If your conjectured formula is correct you should see why these sets are
named as they are.
Exercises — 4.1
1. What is the power set of ∅? Hint: if you got the last exercise in the
chapter you’d know that this power set has 20 = 1 element.
2. Try iterating the power set operator. What is P(P(∅))? What is

P(P(P(∅)))?
3. What, in Logic, corresponds the the notion ∅ in Set theory?
4. What, in Set theory, corresponds to the notion t (a tautology) in Logic?
5. What is the truth set of the proposition P (x) = “3 divides x and 2

divides x.”
6. Find a logical open sentence such that {0, 1, 4, 9, . . .} is its truth set.
7. How many singleton sets are there in the power set of {1, 2, 3, . . . n}?
166 CHAPTER 4. SETS
4.2 Containment
There are two notions of being “inside” a set. A thing may be an element
of a set, or may be contained as a subset. Distinguishing these two notions
of inclusion is essential. One difficulty that sometimes complicates things is
that a set may contain other sets as elements. For instance, as we saw in the
previous section, the elements of a power set are themselves sets.
A set A is a subset of another set B if all of A’s elements are also in B.
The terminology superset is used to refer to B in this situation, as in “The set
of all real-valued functions in one real variable is a superset of the polynomial
functions.” The subset/superset relationship is indicated with a symbol that
should be thought of as a stylized version of the less-than-or-equal sign, when
A is a subset of B we write A ⊆ B. We say that A is a proper subset of B if
B has some elements that aren’t in A, and in this situation we write A ⊂ B
or if we really want to emphasize the fact that the sets are not equal we can
write A ( B.
As we’ve seen earlier, the symbol ∈ is used between an element of a set
and the set that it’s in. The following exercise is intended to clarify the
distinction between ∈ and ⊆.

Exercise. Let A = 1, 2, {1}, {a, b} . Which of the following are true?
i) {a, b} ⊆ A. vi) {1} ⊆ A.
ii) {a, b} ∈ A. vii) {1} ∈ A.
iii) a ∈ A. viii) {2} ∈ A.
iv) 1 ∈ A. ix) {2} ⊆ A.
v) 1 ⊆ A. x) {{1}} ⊆ A.
Another perspective that may help clear up the distinction between ∈ and
⊆ is to consider what they correspond to in Logic. The “element of” symbol
∈ is used to construct open sentences that embody the membership question
– thus it corresponds to single sentences in Logic. The “set containment”
4.2. CONTAINMENT 167
symbol ⊆ goes between two sets and so whatever it corresponds to in Logic

should be something that can appropriately be inserted between two sen-
tences. Let’s run through a short example to figure out what that might be.
To keep things simple we’ll work inside the universal set U = {1, 2, 3, . . . 50}.
Let T be the subset of U consisting of those numbers that are divisible by
10, and let F be those that are divisible by 5.
T = {10, 20, 30, 40, 50}
F = {5, 10, 15, 20, 25, 30, 35, 40, 45, 50}
Hopefully it is clear that ⊆ can be inserted between these two sets like
so: T ⊆ F . On the other hand we can re-express the sets T and F using
set-builder notation in order to see clearly what their membership questions
are.
T = {x ∈ U 10 | x}
F = {x ∈ U 5 | x}
What logical operator fits nicely between 10 | x and 5 | x? Well, of course,

it’s the implication arrow. It’s easy to verify that 10 | x =⇒ 5 | x, and it’s
equally easy to note that the other direction doesn’t work, 5 | x ; 10 | x —
for instance, 5 goes evenly into 15, but 10 doesn’t.
The general statement is: if A and B are sets, and MA (x) and MB (x) are
their respective membership questions, then A ⊆ B corresponds precisely to
∀x ∈ U, MA (x) =⇒ MB (x).
Now to many people (me included!) this looks funny at first, ⊆ in Set
theory corresponds to =⇒ in Logic. It seems like both of these symbols
are arrows of a sort – but they point in opposite directions! Personally, I
resolve the apparent discrepancy by thinking about the “strength” of logical
predicates. One predicate is stronger than another if it puts more conditions
168 CHAPTER 4. SETS
on the elements that would make it true. For example, “x is doubly-even”

is stronger than “x is (merely) even.” Now, the stronger statement implies
the weaker (assuming of course that they are stronger and weaker versions
of the same idea). If a number is doubly-even (i.e. divisible by 4) then it
is certainly even – but the converse is certainly not true, 6 is even but not
doubly-even. Think of all this in terms of sets now. Which set contains the
other, the set of doubly-even numbers or the set of even numbers? Clearly
the set that corresponds to more stringent membership criteria is smaller
than the set that corresponds to less restrictive criteria, thus the set defined
by a weak membership criterion contains the one having a stronger criterion.
If we are asked to prove that one set is contained in another as a subset,
A ⊆ B, there are two ways to proceed. We may either argue by thinking
about elements, or (although this amounts to the same thing) we can show
that A’s membership criterion implies B’s membership criterion.
Exercise. Consider S, the set of perfect squares and F , the set of perfect
fourth powers. Which is contained in the other? Can you prove it?
We’ll end this section with a fairly elementary proof – mainly just to
illustrate how one should proceed in proving that one set is contained in
another.
Let D represent the set of all integers that are divisible by 9,
D = {x ∈ Z ∃k ∈ Z, x = 9k}.
Let C represent the set of all integers that are divisible by 3,
C = {x ∈ Z ∃k ∈ Z, x = 3k}.
The set D is contained in C. Let’s prove it!
Proof: Suppose that x is an arbitrary element of D. From the

definition of D it follows that there is an integer k such that
4.2. CONTAINMENT 169
x = 9k. We want to show that x ∈ C, but since x = 9k it is easy

to see that x = 3(3k) which shows (since 3k is clearly an integer)
that x is in D.
Q.E.D.
170 CHAPTER 4. SETS
Exercises — 4.2
1. Insert either ∈ or ⊆ in the blanks in the following sentences (in order

to produce true sentences).
i) 1 {3, 2, 1, {a, b}} iii) {a, b} {3, 2, 1, {a, b}}

ii) {a} {a, {a, b}} iv) {{a, b}} {a, {a, b}}
2. Suppose that p is a prime, for each n in Z+ , define the set Pn = {x ∈

Z+ pn | x}. Conjecture and prove a statement about the containments
between these sets.
3. Provide a counterexample to dispel the notion that a subset must have

fewer elements than its superset.
4. We have seen that A ⊆ B corresponds to MA =⇒ MB . What

corresponds to the contrapositive statement?
5. Determine two sets A and B such that both of the sentences A ∈ B

and A ⊆ B are true.
6. Prove that the set of perfect fourth powers is contained in the set of
perfect squares.
4.3. SET OPERATIONS 171
4.3 Set operations

In this section we’ll continue to develop the correspondence between Logic
and Set theory.
The logical connectors ∧ and ∨ correspond to the set-theoretic notions
of union (∪) and intersection (∩). The symbols are designed to provide a
mnemonic for the correspondence; the Set theory symbols are just rounded
versions of those from Logic.
Explicitly, if P (x) and Q(x) are open sentences, then the union of the
corresponding truth sets SP and SQ is defined by
SP ∪ SQ = {x ∈ U P (x) ∨ Q(x)}.
Exercise. Suppose two sets A and B are given. Re-express the previous
definition of “union” using their membership criteria, MA (x) = “x ∈ A” and
MB (x) = “x ∈ B.”
The union of more than two sets can be expressed using a big union
symbol. For example, consider the family of real intervals defined by In =
(n, n + 1].3 There’s an interval for every integer n. Also, every real number
is in one of these intervals. The previous sentence can be expressed as
[
R = In .
n∈Z
The intersection of two sets is conceptualized as “what they have in com-

mon” but the precise definition is found by considering conjunctions,
A ∩ B = {x ∈ U x ∈ A ∧ x ∈ B}.
3
The elements of In can also be distinguished as the solution sets of the inequalities
n < x ≤ n + 1.
172 CHAPTER 4. SETS
Exercise. With reference to two open sentences P (x) and Q(x), define the
intersection of their truth sets, SP ∩ SQ .
There is also a “big” version of the intersection symbol. Using the same
family of intervals as before,
\
∅ = In .
n∈Z
Of course the intersection of any distinct pair of these intervals is empty

so the statement above isn’t particularly strong.
Negation in Logic corresponds to complementation in Set theory. The
complement of a set A is usually denoted by A (although some prefer a
superscript c – as in Ac ), this is the set of all things that aren’t in A. In
thinking about complementation one quickly sees why the importance of
working within a well-defined universal set is stressed. Consider the set of all
math textbooks. Obviously the complement of this set would contain texts
in English, Engineering and Evolution – but that statement is implicitly
assuming that the universe of discourse is “textbooks.” It’s equally valid to
say that a very long sequence of zeros and ones, a luscious red strawberry,
√
and the number π are not math textbooks and so these things are all
elements of the complement of the set of all math textbooks. What is really
a concern for us is the issue of whether or not the complement of a set is
well-defined, that is, can we tell for sure whether a given item is or is not
in the complement of a set. This question is decideable exactly when the
membership question for the original set is decideable. Many people think
that the main reason for working within a fixed universal set is that we
then have well-defined complements. The real reason that we accept this
restriction is to ensure that both membership criteria, MA (x) and MA (x),
are decideable open sentences. As an example of the sort of strangeness that
can crop up, consider that during the time that I, as the author of this book,
was writing the last paragraph, this text was nothing more than a very long
sequence of zeros and ones in the memory of my computer. . .
Every rule that we learned in Chapter 2 (see Table 2.2) has a set-theoretic
equivalent. These set-theoretic versions are expressed using equalities (i.e.
the symbol = in between two sets) which is actually a little bit funny if you
think about it. We normally use = to mean that two numbers or variables
have the same numerical magnitude, as in 122 = 144, we are doing some-
thing alltogether different when we use that symbol between two sets, as in
√ √ √
{1, 2, 3} = { 1, 4, 9}, but people seem to be used to this so there’s no
sense in quibbling.
Exercise. Develop a useful definition for set equality. In other words, come
up with a (quantified) logical statement that means the same thing as “A =
B” for two arbitrary sets A and B.
Exercise. What symbol in Logic should go between the membership criteria

MA (x) and MB (x) if A and B are equal sets?
In Table 4.1 the rules governing the interactions between the set theoretic
operations are collected.
We are now in a position somewhat similar to when we jumped from
proving logical assertions with truth tables to doing two-column proofs. We
have two different approaches for showing that two sets are equal. We can
do a so-called “element chasing” proof (to show A = B, assume x ∈ A and
prove x ∈ B and then vice versa). Or, we can construct a proof using the
basic set equalities given in Table 4.1. Often the latter can take the form of
a two-column proof.
174 CHAPTER 4. SETS
Intersection Union
version version
Commutative
A∩B =B∩A A∪B =B∪A
laws
Associative A ∩ (B ∩ C) A ∪ (B ∪ C)
laws = (A ∩ B) ∩ C = (A ∪ B) ∪ C
Distributive A ∩ (B ∪ C) = A ∪ (B ∩ C) =
laws (A ∩ B) ∪ (A ∩ C) (A ∪ B) ∩ (A ∪ C)
DeMorgan’s A∩B A∪B

laws = A∪B = A∩B
Complementarity A∩A = ∅ A∪A = U
Identity
A∩U =A A∪∅=A
laws
Domination A∩∅=∅ A∪U =U
Idempotence A∩A=A A∪A=A
Absorption A ∩ (A ∪ B) = A A ∪ (A ∩ B) = A
Table 4.1: Basic set theoretic equalities.

There are some notions within Set theory that don’t have any clear par-
allels in Logic. One of these is essentially a generalization of the concept of
“complements.” If you think of the set A as being the difference between
the universal set U and the set A you are on the right track. The difference
between two sets is written A \ B (sadly, sometimes this is denoted using the
ordinary subtraction symbol A − B) and is defined by
A \ B = A ∩ B.
The difference, A \ B, consists of those elements of A that aren’t in B. In

some developments of Set theory, the difference of sets is defined first and
then complementation is defined by A = U \ A.
The difference of sets (like the difference of real numbers) is not a commu-
tative operation. In other words A \ B 6= B \ A (in general). It is possible to
define an operation that acts somewhat like the difference, but that is com-
mutative. The symmetric difference of two sets is denoted using a triangle
(really a capital Greek delta)
A4B = (A \ B) ∪ (B \ A).
Exercise. Show that A4B = (A ∪ B) \ (A ∩ B).
Come on! You read right past that exercise without even pausing!
What? You say you did try it and it was too hard?
Okay, just for you (and this time only) I’ve prepared an aid to help you
through. . .
On the next page is a two-column proof of the result you need to prove,
but the lines of the proof are all scrambled. Make a copy and cut out all the
pieces and then glue them together into a valid proof.
So, no more excuses, just do it!
176 CHAPTER 4. SETS
= (A ∩ B) ∪ (B ∩ A) identity law
= (A ∪ B) ∩ (A ∩ B) def. of relative difference
(A ∪ B) \ (A ∩ B) Given
= ((A ∩ A) ∪ (A ∩ B)) ∪ ((B ∩ A) ∪ (B ∩ B)) distributive law
= (A \ B) ∪ (B \ A) def. of relative difference
= (A ∩ (A ∩ B)) ∪ (B ∩ (A ∩ B)) distributive law
= A4B def. of symmetric difference
= (A ∩ (A ∪ B) ∪ (B ∩ (A ∪ B)) DeMorgan’s law
= (∅ ∪ (A ∩ B)) ∪ ((B ∩ A) ∪ ∅) complementarity

Exercises — 4.3
1. Let A = {1, 2, {1, 2}, b} and let B = {a, b, {1, 2}}. Find the following:
(a) A ∩ B
(b) A ∪ B
(c) A \ B
(d) B \ A
(e) A4B
2. In a standard deck of playing cards one can distinguish sets based on

face-value and/or suit. Let A, 2, . . . 9, 10, J, Q and K represent the sets
of cards having the various face-values. Also, let ♥, ♠, ♣ and ♦ be the
sets of cards having the possible suits. Find the following
(a) A ∩ ♥
(b) A ∪ ♥
(c) J ∩ (♠ ∩ ♥)
(d) K ∩ ♥
(e) A ∩ K
(f) A ∪ K
3. Foreach positive integer n, we’ll define an interval In by
In = [−n, 1/n).
Find the union and intersection of all the intervals in this infinite family.
[
In =
n∈N
178 CHAPTER 4. SETS
\
In =
n∈N
4. There is a set X such that, for all sets A, we have X4A = A. What
is X?
5. There is a set Y such that, for all sets A, we have Y 4A = A. What is

Y?
6. In proving a set-theoretic identity, we are basically showing that two

sets are equal. One reasonable way to proceed is to show that each is
contained in the other. Prove that A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C) by
showing that A ∩ (B ∪ C) ⊆ (A ∩ B) ∪ (A ∩ C) and (A ∩ B) ∪ (A ∩ C) ⊆
A ∩ (B ∪ C).
7. Prove the set-theoretic versions of DeMorgan’s laws using the technique

discussed in the previous problem.
4.4. VENN DIAGRAMS 179
4.4 Venn diagrams
Hopefully, you’ve seen Venn diagrams before, but possibly you haven’t thought
deeply about them. Venn diagrams take advantage of an obvious but impor-
tant property of closed curves drawn in the plane. They divide the points
in the plane into two sets, those that are inside the curve and those that
are outside! (Forget for a moment about the points that are on the curve.)
This seemingly obvious statement is known as the Jordan curve theorem,
and actually requires some details. A Jordan curve is the sort of curve you
might draw if you are required to end where you began and you are required
not to cross-over any portion of the curve that has already been drawn. In
technical terms such a curve is called continuous, simple and closed. The
Jordan curve theorem is one of those statements that hardly seems like it
needs a proof, but nevertheless, the proof of this statement is probably the
best-remembered work of the famous French mathematician Camille Jordan.
The prototypical Venn diagram is the picture that looks something like
the view through a set of binoculars.
A B
180 CHAPTER 4. SETS
In a Venn diagram the universe of discourse is normally drawn as a rect-

angular region inside of which all the action occurs. Each set in a Venn
diagram is depicted by drawing a simple closed curve – typically a circle, but
not necessarily! For instance, if you want to draw a Venn diagram that shows
all the possible intersections among four sets, you’ll find it’s impossible with
(only) circles.
C D
Exercise. Verify that the diagram above has regions representing all 16 pos-
sible intersections of 4 sets.
There is a certain “zen” to Venn diagrams that must be internalized, but

once you have done so they can be used to think very effectively about the
relationships between sets. The main deal is that the points inside of one
of the simple closed curves are not necessarily in the set – only some of the
points inside a simple closed curve are in the set, and we don’t know precisely
where they are! The various simple closed curves in a Venn diagram divide
the universe up into a bunch of regions. It might be best to think of these
regions as fenced-in areas in which the elements of a set mill about, much
like domesticated animals in their pens. One of our main tools in working
with Venn diagrams is to deduce that certain of these regions don’t contain
any elements – we then mark that region with the emptyset symbol (∅).
Here is a small example of a finite universe.
Mr. Ed Donald Duck

Black Beauty
Snowball
Shadowfax
Ren
Heckle
Silver
Tweety Bird
Misty
Wile E. Coyote
Secretariat
And here is the same universe with some Jordan curves used to encircle two
subsets.
182 CHAPTER 4. SETS
Mr. Ed Donald Duck

Black Beauty
Snowball
Shadowfax
Ren
Heckle
Silver
Tweety Bird
Misty
Wile E. Coyote
H
Secretariat C
This picture might lead us to think that the set of cartoon characters and
the set of horses are disjoint, so we thought it would be nice to add one more
element to our universe in order to dispel that notion.
Mr. Ed Donald Duck

Black Beauty
Snowball
Shadowfax
Night Mare Ren
Heckle
Silver
Tweety Bird
Misty
Wile E. Coyote
H
Secretariat C
Suppose we have two sets A and B and we’re interested in proving that
B ⊆ A. The job is done if we can show that all of B’s elements are actually in
the eye-shaped region that represents the intersection A ∩ B. It’s equivalent
if we can show that the region marked with ∅ in the following diagram is
actually empty.
A B
Let’s put all this together. The inclusion B ⊆ A corresponds to the

logical sentence MB =⇒ MA . We know that implications are equivalent
to OR statements, so MB =⇒ MA ∼ = ¬MB ∨ MA . The notion that the
region we’ve indicated above is empty is written as A ∩ B = ∅, in logical
terms this is ¬MA ∧ MB ∼ = c. Finally, we apply DeMorgan’s law and a
commutation to get ¬MB ∨ MA ∼ = t. You should take note of the convention
that when you see a logical sentence just written on the page (as is the case
with MB =⇒ MA in the first sentence of this paragraph) what’s being
asserted is that the sentence is universally true. Thus, writing MB =⇒ MA
is the same thing as writing MB =⇒ MA ∼ = t.
One can use information that is known a priori when drawing a Venn
diagram. For instance if two sets are known to be disjoint, or if one is known
to be contained in the other, we can draw Venn diagrams like the following.
184 CHAPTER 4. SETS
A B
However, both of these situations can also be dealt with by working with
Venn diagrams in which the sets are in general position – which in this
situation means that every possible intersection is shown – and then marking
any empty regions with ∅.
Exercise. On a Venn diagram for two sets in general position, indicate the
empty regions when
a) The sets are disjoint.
b) A is contained in B.
A B
There is a connection, perhaps obvious, between the regions we see in a

Venn diagram with sets in general position and the recognizers we studied in
the section on digital logic circuits. In fact both of these topics have to do
with disjunctive normal form. In a Venn diagram with k sets, we are seeing
the universe of discourse broken up into the union of 2k regions each of which
corresponds to an intersection of either one of the sets or its complement.
An arbitrary expression involving set-theoretic symbols and these k sets is
true in certain of these 2k regions and false in the others. We have put the
arbitrary expression in disjunctive normal form when we express it as a union
of the intersections that describe those regions.
186 CHAPTER 4. SETS
A B
A∩B∩C A∩B∩C
A∩B∩C
A∩B∩C
A∩B∩C A∩B∩C
A∩B∩C
A∩B∩C
C
Exercises — 4.4
1. Venn diagrams are usually made using simple closed curves with no
further restrictions. Try creating Venn diagrams for 3, 4 and 5 sets (in
general position) using rectangular simple closed curves.
2. We call a curve rectilinear if it is made of line segments that meet

at right angles. Use rectilinear simple closed curves to create a Venn
diagram for 5 sets.
3. Argue as to why rectilinear curves will suffice to build any Venn dia-
gram.
4. Find the disjunctive normal form of A ∩ (B ∪ C).
5. Find the disjunctive normal form of (A4B)4C
6. The prototypes for the modus ponens and modus tollens argument
forms are the following:
All men are mortal. All men are mortal.
Socrates is a man. Zeus is not mortal.
and
Therefore Socrates is Therefore Zeus is not a
mortal. man.
Illustrate these arguments using Venn diagrams.
7. Use Venn diagrams to convince yourself of the validity of the following

containment statement
(A ∩ B) ∪ (C ∩ D) ⊆ (A ∪ C) ∩ (B ∪ D).
Now prove it!

188 CHAPTER 4. SETS
4.5 Russell’s Paradox

There is no Nobel prize category for mathematics.4 Alfred Nobel’s will
called for the awarding of annual prizes in physics, chemistry, physiology
or medicine, literature, and peace. Later, the “Bank of Sweden Prize in Eco-
nomic Sciences in Memory of Alfred Nobel” was created and certainly several
mathematicians have won what is improperly known as the Nobel prize in
Economics. But, there is no Nobel prize in Mathematics per se. There is an
interesting urban myth that purports to explain this lapse: Alfred Nobel’s
wife either left him for, or had an affair with a mathematician — so Nobel,
the inventor of dynamite and an immensely wealthy and powerful man, when
he decided to endow a set of annual prizes for “those who, during the pre-
ceding year, shall have conferred the greatest benefit on mankind” pointedly
left out mathematicians.
One major flaw in this theory is that Nobel was never married.
In all likelyhood, Nobel simply didn’t view mathematics as a field which
provides benefits for mankind — at least not directly. The broadest division
within mathematics is between the “pure” and “applied” branches. Just
precisely where the dividing line between these spheres lies is a matter of
opinion, but it can be argued that it is so far to one side that one may as
well call an applied mathematician a physicist (or chemist, or biologist, or
economist, or . . . ). One thing is clear, Nobel believed to a certain extent in
the utilitarian ethos. The value of a thing (or a person) is determined by
how useful it is (or they are), which makes it interesting that one of the few
mathematicians to win a Nobel prize was Bertrand Russell (the 1950 prize
in Literature “in recognition of his varied and significant writings in which
4
There are prizes considered eqivalent to the Nobel in stature – the Fields Medal,
awarded every four years by the International Mathematical Union to up to four mathe-
matical researchers under the age of forty, and the Abel Prize, awarded annually by the
King of Norway.
4.5. RUSSELL’S PARADOX 189
he champions humanitarian ideals and freedom of thought”).

Bertrand Russell was one of the twentieth century’s most colorful intel-
lectuals. He helped revolutionize the foundations of mathematics, but was
perhaps better known as a philosopher. It’s hard to conceive of anyone who
would characterize Russell as an applied mathematician!
Russell was an ardent anti-war and anti-nuclear activist. He achieved
a status (shared with Albert Einstein, but very few others) as an eminent
scientist who was also a powerful moral authority. Russell’s mathematical
work was of a very abstruse foundational sort; he was concerned with the
idea of reducing all mathematical thought to Logic and Set theory.
In the beginning of our investigations into Set theory we mentioned that
the notion of a “set of all sets” leads to something paradoxical. Now we’re
ready to look more closely into that remark and hopefully gain an under-
standing of Russell’s paradox.
By this point you should be okay with the notion of a set that contains
other sets, but would it be okay for a set to contain itself ? That is, would it
make sense to have a set defined by
A = {1, 2, A}.
The set A has three elements, 1, 2 and itself. So we could write
A = {1, 2, {1, 2, A}},
and then
A = {1, 2, {1, 2, {1, 2, A}}},
and then
A = {1, 2, {1, 2, {1, 2, {1, 2, A}}}},

190 CHAPTER 4. SETS
et cetera.
This obviously seems like a problem. Indeed, often paradoxes seem to be
caused by self-reference of this sort. Consider The sentence in this box is false.
So a reasonable alternative is to “do” math among the sets that don’t exhibit
this particular pathology.
Thus, inside the set of all sets we are singling out a particular subset that
consists of sets which don’t contain themselves.
S = {A A is a set ∧ A ∈
/ A}
Now within the universal set we’re working in (the set of all sets) there
are only two possibilities: a given set is either in S or it is in its complement
S. Russell’s paradox comes about when we try to decide which of these
alternatives pertains to S itself, the problem is that each alternative leads us
to the other!
If we assume that S ∈ S, then it must be the case that S satisfies the
membership criterion for S. Thus, S ∈
/ S.
On the other hand, if we assume that S ∈
/ S, then we see that S does
indeed satisfy the membership criterion for S. Thus S ∈ S.
Russell himself developed a workaround for the paradox which bears his
name. Together with Alfred North Whitehead he published a 3 volume
work entitled Principia Mathematica 5 [17]. In the Principia, Whitehead and
Russell develop a system known as type theory which sets forth principles for
avoiding problems like Russell’s paradox. Basically, a set and its elements
are of different “types” and so the notion of a set being contained in itself
(as an element) is disallowed.
5
Isaac Newton also published a 3 volume work which is often cited by this same title,
Philosophiae Naturalis Principia Mathematica.
4.5. RUSSELL’S PARADOX 191
Exercises — 4.5
1. Verify that (A =⇒ ¬A) ∧ (¬A =⇒ A) is a logical contradiction

in two ways: by filling out a truth table and using the laws of logical
equivalence.
2. One way out of Russell’s paradox is to declare that the collection of sets
that don’t contain themselves as elements is not a set itself. Explain
how this circumvents the paradox.
3. On the TV show Star Trek (this is the original 1960’s incarnation with
William Shatner playing Captain Kirk, and Leonard Nimoy playing
Mr. Spock), the lead characters manage to foil an evil robot who has
taken over their spaceship by catching it in a paradox very similar to
the form of Russell’s paradox. Captain Kirk and Mr. Spock both
approach the robot and Kirk says “You know, everything Mr. Spock
says is truthful.” Mr. Spock catches on quickly and makes a statement
that causes the robot to become increasingly agitated – eventually its
head begins smoking and sparking and the crew is able to “transport”
it off of the ship just prior to its explosion. (See the power of a good
paradox?)
What did Mr. Spock say to the robot?
192 CHAPTER 4. SETS
Chapter 5
Proof techniques II —
Induction
Who was the guy who first looked at a cow and said, ”I think I’ll drink
whatever comes out of these things when I squeeze ’em!”? –Bill Watterson
5.1 The principle of mathematical induction

The Principle of Mathematical Induction (PMI) may be the least intuitive
proof method available to us. Indeed, at first, PMI may feel somewhat like
grabbing yourself by the seat of your pants and lifting yourself into the air.
Despite the indisputable fact that proofs by PMI often feel like magic, we
need to convince you of the validity of this proof technique. It is one of the
most important tools in your mathematical kit!
The simplest argument in favor of the validity of PMI is simply that it
is axiomatic. This may seem somewhat unsatisfying, but the axioms for
the natural number system, known as the Peano axioms, include one that
justifies PMI. The Peano axioms will not be treated thoroughly in this book,
but here they are:
193
194 CHAPTER 5. PROOF TECHNIQUES II — INDUCTION
i) There is a least element of N that we denote by 0.
ii) Every natural number a has a successor denoted by s(a). (Intuitively,

think of s(a) = a + 1.)
iii) There is no natural number whose successor is 0. (In other words, -1

isn’t in N.)
iv) Distinct natural numbers have distinct successors. (a 6= b =⇒ s(a) 6=

s(b))
v) If a subset of the natural numbers contains 0 and also has the property
that whenever a ∈ S it follows that s(a) ∈ S, then the subset S is
actually equal to N.
The last axiom is the one that justifies PMI. Basically, if 0 is in a subset,
and the subset has this property about successors1 , then 1 must be in it. But
if 1 is in it, then 1’s successor (2) must be in it. And so on . . .
The subset ends up having every natural number in it.
Exercise. Verify that the following symbolic formulation has the same con-
tent as the version of the 5th Peano axiom given above.
∀S ⊆ N (0 ∈ S) ∧ (∀a ∈ N, a ∈ S =⇒ s(a) ∈ S) =⇒ S = N
On August 16th 2003, Ma Lihua of Beijing, China earned her place in

the record books by single-handedly setting up an arrangement of dominoes
standing on end (actually, the setup took 7 weeks and was almost ruined by
some cockroaches in the Singapore Expo Hall) and toppling them. After the
first domino was tipped over it took about six minutes before 303,621 out of
the 303,628 dominoes had fallen. (One has to wonder what kept those other
7 dominoes upright . . . )
1
Whenever a number is in it, the number’s successor must be in it.
5.1. THE PRINCIPLE OF MATHEMATICAL INDUCTION 195
This is the model one should keep in mind when thinking about PMI:
domino toppling. In setting up a line of dominoes, what do we need to do in
order to ensure that they will all fall when the toppling begins? Every domino
must be placed so that it will hit and topple its successor. This is exactly
analogous to (a ∈ S =⇒ s(a) ∈ S). (Think of S having the membership
criterion, x ∈ S = “x will have fallen when the toppling is over.”) The other
thing that has to happen (barring the action of cockroaches) is for someone
to knock over the first domino. This is analogous to 0 ∈ S.
Rather than continuing to talk about subsets of the naturals, it will be
convenient to recast our discussion in terms of infinite families of logical state-
ments. If we have a sequence of statements, (one for each natural number)
P0 , P1 , P2 , P3 , . . . we can prove them all to be true using PMI. We have to
do two things. First – and this is usually the easy part – we must show that
P0 is true (i.e. the first domino will get knocked over). Second, we must
show, for every possible value of k, Pk =⇒ Pk+1 (i.e. each domino will
knock down its successor). These two parts of an inductive proof are known,
respectively, as the basis and the inductive step.
An outline for a proof using PMI:
Theorem ∀n ∈ N, Pn
Proof: (By induction)

Basis:
.. (Here we must show
.
that P0 is true.)
Inductive step:
(Here we must show

..
. that ∀k, Pk =⇒ Pk+1
is true.)
Q.E.D.
Soon we’ll do an actual example of an inductive proof, but first we have

to say something REALLY IMPORTANT about such proofs. Pay attention!
This is REALLY IMPORTANT ! When doing the second part of an inductive
proof (the inductive step), you are proving a UCS, and if you recall how
that’s done, you start by assuming the antecedent is true. But the particular
UCS we’ll be dealing with is ∀k, Pk =⇒ Pk+1 . That means that in the
course of proving ∀n, Pn we have to assume ∀k, Pk . Now this sounds very
much like the error known as “circular reasoning,” especially as many authors
don’t even use different letters (n versus k in our outline) to distinguish the
two statements. (And, quite honestly, we only introduced the variable k to
assuage a certain lingering guilt re circular reasoning.) The sentence ∀n, Pn
is what we’re trying to prove. The sentence ∀k, Pk is known as the inductive
hypothesis. Think about it this way: If we were doing an entirely separate

proof of ∀n, Pn =⇒ Pn+1 , it would certainly be fair to use the inductive
hypothesis, and once that proof was done, it would be okay to quote that
result in an inductive proof of ∀n, Pn . Thus we can compartmentalize our
way out of the difficulty!
Okay, so on to an example. In Section 4.1 we discovered a formula relating
the sizes of a set A and its power set P(A). If |A| = n then |P(A)| = 2n .
What we’ve got here is an infinite family of logical sentences, one for each
value of n in the natural numbers,
|A| = 0 =⇒ |P(A)| = 20 ,
|A| = 1 =⇒ |P(A)| = 21 ,
|A| = 2 =⇒ |P(A)| = 22 ,
|A| = 3 =⇒ |P(A)| = 23 ,
et cetera.
This is exactly the sort of situation in which we use induction.
Theorem 5.1.1. For all finite sets A, |A| = n =⇒ |P(A)| = 2n .
Proof: Let n = |A| and proceed by induction on n.
Basis: Suppose A is a finite set and |A| = 0, it follows that

A = ∅. The power set of ∅ is {∅} which is a set having 1 element.
Note that 20 = 1.
Inductive step: Suppose that A is a finite set with |A| = k + 1.

Choose some particular element of A, say a, and note that we can
divide the subsets of A (i.e. elements of P(A)) into two categories,
those that contain a and those that don’t.
Let S1 = {X ∈ P(A) a ∈ X} and let S2 = {X ∈ P(A) a ∈

/ X}.
We have created two sets that contain all the elements of P(A),
and which are disjoint from one another. In symbolic form, S1 ∪
S2 = P(A) and S1 ∩ S2 = ∅. It follows that |P(A)| = |S1 | + |S2 |.
Notice that S2 is actually the power set of the k-element set A \

{a}. By the inductive hypothesis, |S2 | = 2k . Also, notice that
each set in S1 corresponds uniquely to a set in S2 if we just remove
the element a from it. This shows that |S1 | = |S2 |. Putting this
all together we get that |P(A)| = 2k + 2k = 2(2k ) = 2k+1 .
Q.E.D.
We close this section with a few pieces of advice.
• Statements that can be proved inductively don’t always start out with
P0 . Sometimes P1 is the first statement in an infinite family. Sometimes
its P5 . Don’t get hung up about something that could be handled by
renumbering things.
• In your final write-up you only need to prove the initial case (whatever
it may be) for the basis, but it is a good idea to try the first several
cases while you are in the “draft” stage. This can provide insights into
how to prove the inductive step, and it may also help you avoid a classic
error in which the inductive approach fails essentially just because there
is a gap between two of the earlier dominoes.2
• It is a good idea to write down somewhere just what it is that needs to

be proved in the inductive step — just don’t make it look like you’re
assuming what needs to be shown. For instance in the proof above
2
See exercise 2, the classic fallacious proof that all horses are the same color.
it might have been nice to start the inductive step with a comment
along the following lines, “What we need to show is that under the
assumption that any set of size k has a power set of size 2k , it follows
that a set of size k + 1 will have a power set of size 2k+1 .”
Exercises — 5.1
1. Consider the sequence of number that are 1 greater than a multiple of

4. (Such numbers are of the form 4j + 1.)
1, 5, 9, 13, 17, 21, 25, 29, . . .
The sum of the first several numbers in this sequence can be expressed
as a polynomial.
n
X
4j + 1 = 2n2 + 3n + 1
j=0
Complete the following table in order to provide evidence that the

formula above is correct.
Pn
n j=0 4j + 1 2n2 + 3n + 1
0 1 1
2
1 1+5=6 2·1 +3·1+1=6
2 1+5+9=
3
4
2. What is wrong with the following inductive proof of “all horses are the
same color.”?
Theorem Let H be a set of n horses, all horses in H are the same
color.
Proof: We proceed by induction on n.

Basis: Suppose H is a set containing 1 horse. Clearly this
horse is the same color as itself.
Inductive step: Given a set of k + 1 horses H we can con-

struct two sets of k horses. Suppose H = {h1 , h2 , h3 , . . . hk+1 }.
Define Ha = {h1 , h2 , h3 , . . . hk } (i.e. Ha contains just the first
k horses) and Hb = {h2 , h3 , h4 , . . . hk+1 } (i.e. Hb contains the
last k horses). By the inductive hypothesis both these sets
contain horses that are “all the same color.” Also, all the
horses from h2 to hk are in both sets so both Ha and Hb con-
tain only horses of this (same) color. Finally, we conclude
that all the horses in H are the same color.
Q.E.D.
3. For each of the following theorems, write the statement that must be
proved for the basis – then prove it, if you can!
(a) The sum of the first n positive integers is (n2 + n)/2.
(b) The sum of the first n (positive) odd numbers is n2 .
(c) If n coins are flipped, the probability that all of them are “heads”
is 1/2n
(d) Every 2n ×2n chessboard – with one square removed – can be tiled
perfectly3 by L-shaped trominoes. (A tromino is like a domino
but made up of 3 little squares. There are two kinds, straight
and L-shaped . This problem is only concerned

with the L-shaped trominoes.)
3
Here, “perfectly tiled” means that every trominoe covers 3 squares of the chessboard
(nothing hangs over the edge) and that every square of the chessboard is covered by some
trominoe.
4. Suppose that the rules of the game for PMI were changed so that one
did the following:
• Basis. Prove that P (0) is true.
• Inductive step. Prove that for all k, Pk implies Pk+2
Explain why this would not constitute a valid proof that Pn holds for
all natural numbers n. How could we change the basis in this outline
to obtain a valid proof?
5.2 Formulas for sums and products

Gauss, when only a child, found a formula for summing the first 100 natural
numbers (or so the story goes. . . ). This formula, and his clever method for
justifying it, can be easily generalized to the sum of the first n naturals.
While learning calculus, notably during the study of Riemann sums, one
encounters other summation formulas. For example, in approximating the
integral of the function f (x) = x2 from 0 to 100 one needs the sum of the
first 100 squares. For this reason, somewhere in almost every calculus book
one will find the following formulas collected:
n
X n(n + 1)
j=
j=1
2
n
X n(n + 1)(2n + 1)
j2 =
j=1
6
n
X n2 (n + 1)2
j3 = .
j=1
4
5.2. FORMULAS FOR SUMS AND PRODUCTS 203
A really industrious author might also include the sum of the fourth pow-
ers. Jacob Bernoulli (a truly industrious individual) got excited enough to
find formulas for the sums of the first ten powers of the naturals. Actually,
Bernoulli went much further. His work on sums of powers lead to the def-
inition of what are now known as Bernoulli numbers and let him calculate
P1000 10
j=1 j in about seven minutes – long before the advent of calculators! In
[16, p. 320], Bernoulli is quoted:
With the help of this table it took me less than half of a quarter
of an hour to find that the tenth powers of the first 1000 numbers
being added together will yield the sum
91, 409, 924, 241, 424, 243, 424, 241, 924, 242, 500.
To the beginning calculus student, the beauty of the above relationships

may be somewhat dimmed by the memorization challenge that they repre-
sent. It is fortunate then, that the right-hand side of the third formula is
just the square of the right-hand side of the first formula. And of course, the
right-hand side of the first formula is something that can be deduced by a six
year old child (provided that he is a super-genius!) This happy coincidence
leaves us to apply most of our rote memorization energy to formula number
two, because the first and third formulas are related by the following rather
bizarre-looking equation,
n n
!2
X X
j3 = j .
j=1 j=1
The sum of the cubes of the first n numbers is the square of their sum.
For completeness we should include the following formula which should
be thought of as the sum of the zeroth powers of the first n naturals.
n
X
1=n
j=1
Exercise. Use the above formulas to approximate the integral

Z 10
x3 − 2x + 3dx
x=0
Our challenge today is not to merely memorize these formulas but to

prove their validity. We’ll use PMI.
Before we start in on a proof, it’s important to figure out where we’re
trying to go. In proving the formula that Gauss discovered by induction
we need to show that the k + 1–th version of the formula holds, assuming
that the k–th version does. Before proceeding on to read the proof do the
following
Exercise. Write down the k + 1–th version of the formula for the sum of the
first n naturals. (You have to replace every n with a k + 1.)
Theorem 5.2.1.
n
X n(n + 1)
∀n ∈ N, j=
j=1
2
Basis: Notice that when n = 0 the sum on the left-hand side has
no terms in it! This is known as an empty sum, and by definition,
an empty sum’s value is 0. Also, when n = 0 the formula on the
right-hand side becomes (0 · 1)/2 and this is 0 as well.4
4
If you’d prefer to avoid the “empty sum” argument, you can choose to use n = 1 as
the basis case. The theorem should be restated so the universe of discourse is positive
naturals.
Inductive step: Consider the sum on the left-hand side of the

k + 1–th version of our formula.
k+1
X
j
j=1
We can separate out the last term of this sum.
k
X
= (k + 1) + j
j=1
Next, we can use the inductive hypothesis to replace the sum (the
part that goes from 1 to k) with a formula.
k(k + 1)
= (k + 1) +
2
From here on out it’s just algebra . . .
2(k + 1) k(k + 1)
= +
2 2
2(k + 1) + k(k + 1)
=
2
(k + 1) · (k + 2)
= .
2
Q.E.D.
Notice how the inductive step in this proof works. We start by writing
down the left-hand side of Pk+1 , we pull out the last term so we’ve got the left-
hand side of Pk (plus something else), then we apply the inductive hypothesis
and do some algebra until we arrive at the right-hand side of Pk+1 . Overall,
we’ve just transformed the left-hand side of the statement we wish to prove
into its right-hand side.
There is another way to organize the inductive steps in proofs like these
that works by manipulating entire equalities (rather than just one side or the
other of them).
Inductive step (alternate): By the inductive hypothesis, we

can write
k
X k(k + 1)
j= .
j=1
2
Adding (k + 1) to both side of this yields
k+1
X k(k + 1)
j = (k + 1) + .
j=1
2
Next, we can simplify the right-hand side of this to obtain
k+1
X (k + 1)(k + 2)
j= .
j=1
2
Q.E.D.
Oftentimes one can save considerable effort in an inductive proof by cre-

atively using the factored form during intermediate steps. On the other hand,
sometimes it is easier to just simplify everything completely, and also, com-
pletely simplify the expression on the right-hand side of P (k + 1) and then
verify that the two things are equal. This is basically just another take on
the technique of “working backwards from the conclusion.” Just remember
that in writing-up your proof you need to make it look as if you reasoned di-
rectly from the premises to the conclusion. We’ll illustrate what we’ve been
discussing in this paragraph while proving the formula for the sum of the
squares of the first n naturals.
Theorem 5.2.2.
n
X n(n + 1)(2n + 1)
∀n ∈ N, j2 =
j=1
6

Basis: When n = 1 the sum has only one term, 12 = 1. On the
1(1 + 1)(2 · 1 + 1)
other hand, the formula is = 1. Since these
6
are equal, the basis is proved.
Inductive step:
Before proceeding with the inductive step, in this box, we

will figure out what the right-hand side of our theorem
looks like when n is replaced with k + 1:
(k + 1)((k + 1) + 1)(2(k + 1) + 1)
6
(k + 1)(k + 2)(2k + 3)
=
6
(k 2 + 3k + 2)(2k + 3)
=
6
2k 3 + 9k 2 + 13k + 6
= .
6
By the inductive hypothesis,
k
X k(k + 1)(2k + 1)
j2 = .
j=1
6
Adding (k + 1)2 to both sides of this equation gives
k
2
X k(k + 1)(2k + 1)
(k + 1) + j2 = + (k + 1)2 .
j=1
6
Thus,
k+1
X k(k + 1)(2k + 1) 6(k + 1)2
j2 = + .
j=1
6 6
Therefore,
k+1
X (k 2 + k)(2k + 1) 6(k 2 + 2k + 1)
j2 = +
j=1
6 6
(2k 3 + 3k 2 + k) + (6k 2 + 12k + 6)
=
6
2k 3 + 9k 2 + 13k + 6
=
6
(k 2 + 3k + 2)(2k + 3)
=
6
(k + 1)(k + 2)(2k + 3)
=
6
(k + 1)((k + 1) + 1)(2(k + 1) + 1)
= .
6
This proves the inductive step, so the result is true.
Q.E.D.
Notice how the last four lines of the proof are the same as those in the
box above containing our scratch work? (Except in the reverse order.)
We’ll end this section by demonstrating one more use of this technique.
This time we’ll look at a formula for a product rather than a sum.
Theorem 5.2.3.
n
Y 1 n+1
∀n ≥ 2 ∈ Z, 1− 2 = .
j=2
j 2n
Before preceeding with the proof let’s look at an example (although this
has nothing to do with proving anything, it’s really not a bad idea – it can
keep you from wasting a lot of time trying to prove something that isn’t
actually true!) When n = 4 the product is

1 1 1
1− 2 · 1− 2 · 1− 2 .
2 3 4
This simplifies to

1 1 1 3 8 15 360
1− · 1− · 1− = · · = .
4 9 16 4 9 16 576
The formula on the right-hand side is
4+1 5
= .
2·4 8
Well! These two expressions are clearly not equal to one another. . . What?
You say they are? Just give me a second with my calculator. . .
Alright then. I guess we can’t dodge doing the proof. . .
Proof: (Using mathematical induction on n.)
Basis: When n = 2 the product has only one term, 1 − 1/22 =

2+1
3/4. On the other hand, the formula is = 3/4. Since these
2·2
are equal, the basis is proved.
Inductive step:
Let k be a particular but arbitrarily chosen integer such that

k
Y 1 k+1
1− 2 = .
j=2
j 2k
Multplying5 both sides by the k + 1-th term of the product gives
k
1 Y 1 k+1 1
1− · 1− 2 = · 1− .
(k + 1)2 j=2
j 2k (k + 1)2
Thus
k+1
Y
1 k+1 1
1− 2 = · 1−
j=2
j 2k (k + 1)2
k+1 (k + 1)
= −
2k 2k(k + 1)2
k+1 (1)
= −
2k 2k(k + 1)
(k + 1)2 − 1
=
2k(k + 1)
k 2 + 2k
=
2k(k + 1)
k(k + 2)
=
2k(k + 1)
5
Really, the only reason I’m doing this silly proof is to point out to you that when
you’re doing the inductive step in a proof of a formula for a product, you don’t add to
both sides anymore, you multiply. You see that, right? Well, consider yourself to have
been pointed out to or . . . oh, whatever.
k+2
= .
2(k + 1)
Q.E.D.
Exercises — 5.2
1. Write an inductive proof of the formula for the sum of the first n cubes.
2. Find a formula for the sum of the first n fourth powers.
3. The sum of the first n natural numbers is sometimes called the n-th
triangular number Tn . Triangular numbers are so-named because one
can represent them with triangular shaped arrangements of dots.
The first several triangular numbers are 1, 3, 6, 10, 15, et cetera.

Determine
! a formula for the sum of the first n triangular numbers
Xn
Ti and prove it using PMI.
i=1
4. Consider the alternating sum of squares:
1
1 − 4 = −3
1−4+9=6
1 − 4 + 9 − 16 = −10
et cetera
Guess a general formula for ni=1 (−1)i−1 i2 , and prove it using PMI.
P
5. Prove the following formula for a product.
n
Y 1 1
1− =
i=2
i n
6. The Fibonacci numbers are a sequence of integers defined by the rule

that a number in the sequence is the sum of the two that precede it.
Fn+2 = Fn + Fn+1
The first two Fibonacci numbers (actually the zeroth and the first) are
both 1.
Thus, the first several Fibonacci numbers are
F0 = 1, F1 = 1, F2 = 2, F3 = 3, F4 = 5, F5 = 8, F6 = 13, F7 = 21, et cetera
Use mathematical induction to prove the following formula involving

Fibonacci numbers.
n
X
(Fi )2 = Fn · Fn+1
i=0
5.3 Divisibility statements and other proofs

using PMI
There is a very famous result known as Fermat’s Little Theorem. This would
probably be abbreviated FLT except for two things. In science fiction FLT
means “faster than light travel” and there is another theorem due to Fermat
that goes by the initials FLT: Fermat’s Last Theorem. Fermat’s last theorem
states that equations of the form an + bn = cn , where n is a positive natural
number, only have integer solutions that are trivial (like 03 + 13 = 13 ) when
n is greater than 2. When n is 1, there are lots of integer solutions. When
n is 2, there are still plenty of integer solutions – these are the so-called
Pythagorean triples, for example 3,4 & 5 or 5,12 & 13. It is somewhat unfair
that this statement is known as Fermat’s last theorem since he didn’t prove
it (or at least we can’t be sure that he proved it). Five years after his death,
Fermat’s son published a translated6 version of Diophantus’s Arithmetica
containing his father’s notations. One of those notations – near the place
where Diophantus was discussing the equation x2 + y 2 = z 2 and its solution
in whole numbers – was the statement of what is now known as Fermat’s last
theorem as well as the following claim:
Cuius rei demonstrationem mirabilem sane detexi hanc marginis

exiguitas non caperet.
In English:
I have discovered a truly remarkable proof of this that the margin

of this page is too small to contain.
Between 1670 and 1994 a lot of famous mathematicians worked on FLT

but never found the “demonstrationem mirabilem.” Finally in 1994, Andrew
6
The translation from Greek into Latin was done by Claude Bachet.
5.3. OTHER PROOFS USING PMI 215
Wiles of Princeton announced a proof of FLT, but in Wiles’s own words, his
is “a twentieth century proof” it can’t be the proof Fermat had in mind.
These days most people believe that Fermat was mistaken. Probably he
thought a proof technique that works for small values of n could be gen-
eralized. It remains a tantalizing question, can a proof of FLT using only
methods available in the 17th century be accomplished?
Part of the reason that so many people spent so much effort on FLT over
the centuries is that Fermat had an excellent record as regards being correct
about his theorems and proofs. The result known as Fermat’s little theorem
is an example of a theorem and proof that Fermat got right. It is probably
known as his “little” theorem because its statement is very short, but it is
actually a fairly deep result.
Theorem 5.3.1 (Fermat’s Little Theorem). For every prime number p, and
for all integers x, the p-th power of x and x itself are congruent mod p.
Symbolically:
xp ≡ x (mod p)
A slight restatement of Fermat’s little theorem is that p is always a divisor

of xp − x (assuming p is a prime and x is an integer). Math professors enjoy
using their knowledge of Fermat’s little theorem to cook up divisibility results
that can be proved using mathematical induction. For example, consider the
following:
∀n ∈ N, 3 | (n3 + 2n + 6).
This is really just the p = 3 case of Fermat’s little theorem with a little
camouflage added (n3 + 2n + 6) = (n3 − n) + 3(n + 2)). But let’s have a look
at proving this statment using PMI.
Theorem 5.3.2. ∀n ∈ N, 3 | (n3 + 2n + 6)

Proof: (By mathematical induction)

Basis: Clearly 3 | 6.
Inductive step:
(We need to show that 3 | (k 3 + 2k + 6) =⇒ 3 | ((k + 1)3 + 2(k +
1) + 6.)
Consider the quantity (k + 1)3 + 2(k + 1) + 6.
(k + 1)3 + 2(k + 1) + 6
= (k 3 + 3k 2 + 3k + 1) + (2k + 2) + 6
= (k 3 + 2k + 6) + 3k 2 + 3k + 3
= (k 3 + 2k + 6) + 3(k 2 + k + 1)
By the inductive hypothesis, 3 is a divisor of k 3 + 2k + 6 so there

is an integer m such that k 3 + 2k + 6 = 3m. Thus,
(k + 1)3 + 2(k + 1) + 6
= 3m + 3(k 2 + k + 1)
= 3(m + k 2 + k + 1).
This equation shows that 3 is a divisor of (k + 1)3 + 2(k + 1) + 6,

which is the desired conclusion.
Q.E.D.
Exercise. Devise an inductive proof of the statement, ∀n ∈ Z, 5 | x5 +4x−10.
There is one other subtle trick for devising statements to be proved by

PMI that you should know about. An example should suffice to make it
clear. Notice that 7 is equivalent to 1 (mod 6), it follows that any power of
7 is also 1 (mod 6). So, if we subtract 1 from some power of 7 we will have
a number that is divisible by 6.
The proof (by PMI) of a statement like this requires another subtle little
trick. Somewhere along the way in the proof you’ll need the identity 7 = 6+1.
Theorem 5.3.3.
∀n ∈ N, 6 | 7n − 1
Proof: (By PMI)

Basis: Note that 70 − 1 is 0 and also that 6 | 0.
Inductive step:
(We need to show that if 6 | 7k − 1 then 6 | 7k+1 − 1.)
Consider the quantity 7k+1 − 1.
7k+1 − 1 = 7 · 7k − 1
= (6 + 1) · 7k − 1
= 6 · 7k + 1 · 7k − 1
= 6(7k ) + (7k − 1)
By the inductive hypothesis, 6 | 7k − 1 so there is an integer m

such that 7k − 1 = 6m. It follows that
7k+1 − 1 = 6(7k ) + 6m.
So, clearly, 6 is a divisor of 7k+1 − 1.
Q.E.D.
Mathematical induction can often be used to prove inequalities. There

are quite a few examples of families of statements where there is an inequality
for every natural number. Often such statements seem to be obviously true
and yet devising a proof can be illusive. If such is the case, try using PMI.
One hint: it is fairly typical that the inductive step in a PMI proof of an
inequality will involve reasoning that isn’t particularly sharp. Just remember
that if you have an inequality and you make the big side even bigger, the
resulting statement is certainly still true!
Consider the sequences 2n and n!.
n 0 1 2 3
n
2 1 2 4 8
n! 1 1 2 6
As the table illustrates, for small values of n, 2n > n!. But from n = 4
onward the inequality is reversed.
Theorem 5.3.4.
∀n ≥ 4 ∈ N, 2n < n!
Proof: (By mathematical induction)
Basis: When n = 4 we have 24 < 4!, which is certainly true

(16 < 24).
Inductive step: Suppose that k is a natural number with k > 4,

and that 2k < k!. Multiply the left hand side of this inequality
by 2 and the right hand side by k + 17 to get
2 · 2k < (k + 1) · k!.
7
It might be smoother to justify this step by first proving the lemma that ∀a, b, c, d ∈
+
R , a < b ∧ c < d =⇒ ac < bd.
So
2k+1 < (k + 1)!.
Q.E.D.
The observant Calculus student will certainly be aware of the fact that,
assymptotically, exponential functions grow faster than polynomial functions.
That is, if you have a base b which is greater than 1, the function bx is eventu-
ally larger than any polynomial p(x). This may seem a bit hard to believe if
b = 1.001 and p(x) = 500x10 . The graph of y = 1.001x is practically indistin-
guishable from the line y = 1 (at first), whereas the graph of y = 500x10 has
already reached the astronomical value of five trillion (5, 000, 000, 000, 000)
when x is just 10. Nevertheless, the exponential will eventually outstrip the
polynomial. We can use the methods of this section to get started on proving
the fact mentioned above. Consider the two sequences n2 and 2n .
n 0 1 2 3 4 5 6
n2 0 1 4 9 16 25 36
2n 1 2 4 8 16 32 64
If we think of a “race” between the sequences n2 and 2n , notice that 2n

starts out with the lead. The two sequences are tied when n = 2. Briefly, n2
goes into the lead but they are tied again when n = 4. After that it would
appear that 2n recaptures the lead for good. Of course we’re making a rather
broad presumption – is it really true that n2 never catches up with 2n again?
Well, if we’re right then the following theorem should be proveable:
Theorem 5.3.5. For all natural numbers n, if n ≥ 4 then n2 ≤ 2n .

Proof:
Basis: When n = 4 we have 42 ≤ 24 , which is true since both
numbers are 16.
Inductive step: (In the inductive step we assume that k 2 ≤ 2k
and then show that (k + 1)2 ≤ 2k+1 .)
The inductive hypothesis tells us that
k 2 ≤ 2k .
If we add 2k + 1 to the left-hand side of this inequality and 2k to

the right-hand side we will produce the desired inequality. Thus
our proof will follow provided that we know that 2k + 1 ≤ 2k .
Indeed, it is sufficient to show that 2k + 1 ≤ k 2 since we already
know (by the inductive hypothesis) that k 2 ≤ 2k .
So the result remains in doubt unless you can complete the exer-
cise that follows. . .
Q.E.D.???
Exercise. Prove the lemma: For all n ∈ N, if n ≥ 4 then 2n + 1 ≤ n2 .

Exercises — 5.3
Give inductive proofs of the following
1. ∀x ∈ Z, 3 | x3 − x
2. ∀x ∈ Z, 3 | x3 + 5x
3. ∀x ∈ Z, 11 | x11 + 10x
4. ∀n ∈ N, 3 | 4n − 1
5. ∀n ∈ N, 7 | 8n + 6
6. ∀n ≥ 3 ∈ N, 3n2 + 3n + 1 < 2n3
7. ∀n > 3 ∈ N, n3 < 3n
5.4 The strong form of mathematical induc-

tion
The strong form of mathematical induction (a.k.a. the principle of complete
induction, PCI; also a.k.a. course-of-values induction) is so-called because
the hypotheses one uses are stronger. Instead of showing that Pk =⇒ Pk+1
in the inductive step, we get to assume that all the statements numbered
smaller than Pk+1 are true. To make life slightly easier we’ll renumber things
a little. The statement that needs to be proved is
∀k(P0 ∧ P1 ∧ . . . ∧ Pk−1 ) =⇒ Pk .
An outline of a strong inductive proof is:
Theorem ∀n ∈ N, Pn
Proof: (By complete induction)

Basis:
(Technically, a PCI
proof doesn’t require a
..
. basis. We recommend
that you show that P0
is true anyway.)
Inductive step:
(Here we V must show

.. k−1
. that ∀k, i=0 Pi =⇒
Pk is true.)
Q.E.D.
5.4. THE STRONG FORM OF MATHEMATICAL INDUCTION 223
It’s fairly common that we won’t truly need all of the statements from P0
to Pk−1 to be true, but just one of them (and we don’t know a priori which
one). The following is a classic result; the proof that all numbers greater
than 1 have prime factors.
Theorem 5.4.1. For all natural numbers n, n > 1 implies n has a prime
factor.
Proof: (By strong induction) Consider an arbitrary natural num-

ber n > 1. If n is prime then n clearly has a prime factor (itself),
so suppose that n is not prime. By definition, a composite natu-
ral number can be factored, so n = a · b for some pair of natural
numbers a and b which are both greater than 1. Since a and b are
factors of n both greater than 1, it follows that a < n (it is also
true that b < n but we don’t need that . . . ). The inductive hy-
pothesis can now be applied to deduce that a has a prime factor
p. Since p | a and a | n, by transitivity p | n. Thus n has a prime
factor.
Q.E.D.
Exercises — 5.4
Give inductive proofs of the following
1. A “postage stamp problem” is a problem that (typically) asks us to

determine what total postage values can be produced using two sorts
of stamps. Suppose that you have 3c stamps and 7c stamps, show
(using strong induction) that any postage value 12c or higher can be
achieved. That is,
∀n ∈ N, n ≥ 12 =⇒ ∃x, y ∈ N, n = 3x + 7y.
√
1+ 5
2. The polynomial equation x2 = x + 1 has two solutions, α = 2
and
√
1− 5
β= 2
. Show that the Fibonacci number Fn is less than or equal to
αn for all n ≥ 0.
Chapter 6
Relations and functions
If evolution really works, how come mothers only have two hands? –Milton
Berle
6.1 Relations
A relation in mathematics is a symbol that can be placed between two num-
bers (or variables) to create a logical statement (or open sentence). The main
point here is that the insertion of a relation symbol between two numbers
creates a statement whose value is either true or false. For example, we have
previously seen the divisibility symbol (|) and noted the common error of
mistaking it for the division symbol (/); one of these tells us to perform an
arithmetic operation, the other asks us whether if such an operation were
performed there would be a remainder. There are many other symbols that
we have seen which have this characteristic, the most important is probably
=, but there are lots: 6=, <, ≤, >, ≥ all work this way – if we place them
between two numbers we get a Boolean thing, it’s either true or false. If, in-
stead of numbers, we think of placing sets on either side of a relation symbol,
then =, ⊆ and ⊇ are valid relation symbols. If we think of placing logical
225
226 CHAPTER 6. RELATIONS AND FUNCTIONS
expressions on either side of a relation then, honestly, any of the logical sym-
bols is a relation, but we normally think of ∧ and ∨ as operators and give
things like ≡, =⇒ and ⇐⇒ the status of relations.
In the examples we’ve looked at the things on either side of a relation are
of the same type. This is usually, but not always, the case. The prevalence
of relations with the same kind of things being compared has even lead to
the aphorism “Don’t compare apples and oranges.” Think about the symbol
∈ for a moment. As we’ve seen previously, it isn’t usually appropriate to put
sets on either side of this, we might have numbers or other objects on the left
and sets on the right. Let’s look at a small example. Let A = {1, 2, 3, a, b}
and let B = {{1, 2, a}, {1, 3, 5, 7, . . .}, {1}}. The “element of” relation, ∈, is
a relation from A to B.
2 {1, 2, a}
3 {1, 3, 5, 7, . . .}
a {1}
Figure 6.1: The “element of” relation is an example of a relation that goes
from one set to a different set.
A diagram such as we have given in Figure 6.1 seems like a very natural
thing. Such pictures certainly give us an easy visual tool for thinking about
6.1. RELATIONS 227
relations. But we should point out certain hidden assumptions. First, they’ll
only work if we are dealing with finite sets, or sets like the odd numbers in
our example (sets that are infinite but could in principle be listed). Second,
by drawing the two sets separately, it seems that we are assuming they are
not only different, but disjoint. The sets not only need not be disjoint, but
often (most of the time!) we have relations that go from a set to itself so
the sets in a picture like this may be identical. In Figure 6.2 we illustrate
the divisibility relation on the set of all divisors of 6 — this is an example in
which the sets on either side of the relation are the same. Notice the linguistic
distinction, we can talk about either “a relation from A to B” (when there
are really two different sets) or “a relation on A” (when there is only one).
1 1
2 2
3 3
6 6
Figure 6.2: The “divides” relation is an example of a relation that goes from
a set to itself. In this example we say that we have a relation on the set of
divisors of 6.
Purists will note that it is really inappropriate to represent the same set
in two different places in a Venn diagram. The diagram in Figure 6.2 should
really look like this:
Indeed, this representation is definitely preferable, although it may be

more crowded. A picture such as this is known as the directed graph (a.k.a.
digraph) of the relation.
Recall that when we were discussing sets we said the best way to describe
a set is simply to list all of its elements. Well, what is the best way to describe
a relation? In the same spirit, it would seem we should explicitly list all the
things that make the relation true. But it takes a pair of things, one to go
on the left side and one to go on the right, to make a relation true (or for
that matter false!). Also it should be evident that order is important in this
context, for example 2 < 3 is true but 3 < 2 isn’t. The identity of a relation
is so intimately tied up with the set of ordered pairs that make it true, that
when dealing with abstract relations we define them as sets of ordered pairs.
Given two sets, A and B, the Cartesian product of A and B is the set of
all ordered pairs (a, b) where a is in A and b is in B. We denote the Cartesian
product using the symbol ×.
6.1. RELATIONS 229
A × B = {(a, b) a ∈ A ∧ b ∈ B}
From here on out in your mathematical career you’ll need to take note of
the context that the symbol × appears in. If it appears between numbers go
ahead and multiply, but if it appears between sets you’re doing something
different – forming the Cartesian product.
The familiar x–y plane, is often called the Cartesian plane. This is done
for two reasons. Rene Descartes, the famous mathematician and philosopher,
was the first to consider coordinatizing the plane and thus is responsible for
our current understanding of the relationship between geometry and algebra.
Rene Descartes’ name is also memorialized in the definition of the Cartesian
product of sets, and the plane is nothing more than the product R × R.
Indeed, the plane provided the very first example of the concept that was
later generalized to the Cartesian product of sets.
Exercise. Suppose A = {1, 2, 3} and B = {a, b, c}. Is (a, 1) in the Cartesian

product A × B? List all elements of A × B.
In the abstract, we can define a relation as any subset of an appropriate

Cartesian product. So an abstract relation R from a set A to a set B is just
some subset of A × B. Similarly, a relation R on a set S is defined by a subset
of S × S. This definition looks a little bit strange when we apply it to an
actual (concrete) relation that we already know about. Consider the relation
“less than.” To describe “less than” as a subset of a Cartesian product we
must write
< = {(x, y) ∈ R × R y − x ∈ R+ }.
This looks funny.

Also, if we have defined some relation R ⊆ A × B, then in order to say
that a particular pair, (a, b), of things make the relation true we have to write
aRb.
This looks funny too.

Despite the strange appearances, these examples do express the correct
way to deal with relations.
Let’s do a completely made-up example. Suppose A is the set {a, e, i, o, u}
and B is the set {r, s, t, l, n} and we define a relation from A to B by
R = {(a, s), (a, t), (a, n), (e, t), (e, l), (e, n), (i, s), (i, t), (o, r), (o, n), (u, s)}.
Then, for example, because (e, t) ∈ R we can write eRt. We indicate the
negation of the concept that two elements are related by drawing a slash
through the name of the relation, for example the notation 6= is certainly
familiar to you, as is ≮ (although in this latter case we would normally write
≥ instead). We can denote the fact that (a, l) is not a pair that makes the
relation true by writing a6 Rl.
We should mention another way of visualizing relations. When we are
dealing with a relation on R, the relation is actually a subset of R × R, that
means we can view the relation as a subset of the x–y plane. In other words,
we can graph it. The graph of the “<” relation is given in Figure 6.3.
A relation on any set that is a subset of R can likewise be graphed. The
graph of the “|” relation is given in Figure 6.4.
Eventually, we will get around to defining functions as relations that have
a certain nice property. For the moment, we’ll just note that some of the op-
erations that you are used to using with functions also apply with relations.
When one function “undoes” what another function “does” we say the func-
tions are inverses. For example, the function f (x) = 2x (i.e. doubling) and
the function g(x) = x/2 (halving) are inverse functions because, no matter
what number we start with, if we double it and then halve that result, we
6.1. RELATIONS 231
Figure 6.3: The “less than” relation can be viewed as a subset of R × R, i.e.
it can be graphed.
Figure 6.4: The divisibility relation can be graphed. Only those points (as
indicated) with integer coordinates are in the graph.
end up with the original number. The inverse of a relation R is written R−1
and it consists of the reversals of the pairs in R,
R−1 = {(b, a) (a, b) ∈ R}.
This can also be expressed by writing
bR−1 a ⇐⇒ aRb.
The process of “doing one function and then doing another” is known as
√
functional composition. For instance, if f (x) = 2x + 1 and g(x) = x, then
we can compose them (in two different orders) to obtain either f (g(x)) =
√ √
2 x + 1 or g(f (x)) = 2x + 1. When composing functions there is an
“intermediate result” that you get by applying the first function to your
input, and then you calculate the second function’s value at the intermediate
result. (For example, in calculating g(f (4)) we get the intermediate result
f (4) = 9 and then we go on to calculate g(9) = 3.)
The definition of the composite of two relations focuses very much on this
idea of the intermediate result. Suppose R is a relation from A to B and S
is a relation from B to C then the composite S ◦ R is given by
S ◦ R = {(a, c) ∃b ∈ B, (a, b) ∈ R ∧ (b, c) ∈ S}.
In this definition, b is the “intermediate result,” if there is no such b that

serves to connect a to c then (a, c) won’t be in the composite. Also, notice
that this is the composition R first, then S, but it is written as S ◦ R – watch
out for this! The compositions of relations should be read from right to
left. This convention makes sense when you consider functional composition,
f (g(x)) means g first, then f so if we use the “little circle” notation for the
composition of relations we have f ◦ g(x) = f (g(x)) which is nice because the
6.1. RELATIONS 233
symbols f and g appear in the same order. But beware! there are atavists
out there who write their compositions the other way around.
You should probably have a diagram like the following in mind while
thinking about the composition of relations. Here, we have the set A =
{1, 2, 3, 4}, the set B is {a, b, c, d} and C = {w, x, y, z}. The relation R goes
from A to B and consists of the following set of pairs,
R = {(1, a), (1, c), (2, d), (3, c), (3, d)}.
And
S = {(a, y), (b, w), (b, x), (b, z)}.

R S
1 a w
2 b x
3 c y
4 d z
Exercise. Notice that the composition R ◦ S is impossible (or, more properly,

it is empty). Why?
What is the (only) pair in the composition S ◦ R ?
Exercises — 6.1
1. The lexicographic order, <lex , is a relation on the set of all words,

where x <lex y means that x would come before y in the dictionary.
Consider just the three letter words like “iff”, “fig”, “the”, et cetera.
Come up with a usable definition for x1 x2 x3 <lex y1 y2 y3 .
2. What is the graph of “=” in R × R?
3. The inverse of a relation R is denoted R−1 . It contains exactly the

same ordered pairs as R but with the order switched. (So technically,
they aren’t exactly the same ordered pairs . . . )
R−1 = {(b, a) (a, b) ∈ R}
Define a relation S on R × R by S = {(x, y) y = sin x}. What is S−1 ?

Draw a single graph containing S and S−1 .
4. The “socks and shoes” rule is a very silly little mnemonic for remem-
bering how to invert a composition. If we think of undoing the process
of putting on our socks and shoes (that’s socks first, then shoes) we
have to first remove our shoes, then take off our socks.
The socks and shoes rule is valid for relations as well.
Prove that (S ◦ R)−1 = R−1 ◦ S−1 .
6.2. PROPERTIES OF RELATIONS 235
6.2 Properties of relations

There are two special classes of relations that we will study in the next
two sections, equivalence relations and ordering relations. The prototype
for an equivalence relation is the ordinary notion of numerical equality, =.
The prototypical ordering relation is ≤. Each of these has certain salient
properties that are the root causes of their importance. In this section we
will study a compendium of properties that a relation may or may not have.
A relation that has three of the properties we’ll discuss:
1. reflexivity
2. symmetry
3. transitivity
is said to be an equivalence relation; it will in some ways resemble =.

A relation that has another set of three properties:
1. reflexivity
2. anti-symmetry
3. transitivity
is called an ordering relation; it will resemble ≤.

Additionally, there is a property known as irreflexivity that many rela-
tions have.
There are a total of 5 properties that we have named, and we will discuss
them all more thoroughly. But first, we’ll state the formal definitions. Take
note that these properties are all stated for a relation that goes from a set
to itself, indeed, most of them wouldn’t even make sense if we tried to define
them for a relation from a set to a different set.
A relation R on a set S is reflexive iff

∀a ∈ S, aRa
“Everything is related to itself.”
A relation R on a set S is irreflexive iff
∀a ∈ S, a6 Ra
“Nothing is related to itself.”
A relation R on a set S is symmetric iff
∀a, b ∈ S, aRb =⇒ bRa
“No one-way streets.”
A relation R on a set S is anti-symmetric iff
∀a, b ∈ S, aRb ∧ bRa =⇒ a=b
“Only one-way streets.”
A relation R on a set S is transitive iff
∀a, b, c ∈ S, aRb ∧ bRc =⇒ aRc
“Whenever there’s a roundabout route, there’s a direct route.”
Table 6.1: Properties that relations may (or may not) have.
The digraph of a relation that is reflexive will have little loops at every
vertex. The digraph of a relation that is irreflexive will contain no loops
at all. Hopefully it is clear that these concepts represent extreme opposite
possibilities – they are not however negations of one another.
Exercise. Find the logical denial of the property that says a relation is re-
flexive
¬(∀a ∈ S, aRa).
How does this differ from the defining property for “irreflexive”?
If a relation R is defined on some subset S of the reals, then it can be

graphed in the Euclidean plane. Reflexivity for R can be interpreted in terms
of the line L defined by the equation y = x. Every point in (S × S) ∩ L must
be in R. A similar statement can be made concerning the irreflexive property.
If a relation R is irreflexive its graph completely avoids the line y = x.
Note that the reflexive and irreflexive properties are defined with a single
quantified variable. Symmetry and anti-symmetry require two universally
quantified variables for their definitions.
A relation R on a set S is symmetric iff
∀a, b ∈ S, aRb =⇒ bRa.
This can be interpreted in terms of digraphs as follows: If a connection from

a to b exists in the digraph of R, then there must also be a connection from
b to a. In Table 6.1 this is interpreted as “no one-way streets” and while
that’s not quite what it says, that is the effect of this definition. Since if
a connection exists in one direction, there must also be a connection in the
other direction, it follows that we will never see a one-way connection.
Because most of the properties we are studying are defined using condi-
tional statements it is often the case that a relation has a property for vacuous
reasons. When the “if” part doesn’t happen, there’s no need for its corre-
sponding “then” part to happen either – the conditional is still true. In the
context of our discussion on the symmetry property of a relation this means
that the following digraph is the digraph of a symmetric relation (although
it is neither reflexive nor irreflexive).
c
a
d
b
Anti-symmetry is described as meaning “only one-way streets” but the

definition is given as:
A relation R on a set S is anti-symmetric iff

∀a, b ∈ S, aRb ∧ bRa =⇒ a = b.
It may be hard at first to understand why the definition we use for anti-
symmetry is the one above. If one wanted to insure that there were never
two-way connections between elements of the set it might seem easier to
define anti-symmetry as follows:
(Alternate definition) A relation R on a set S is anti-symmetric

iff
∀a, b ∈ S, aRb =⇒ b6 Ra.
This definition may seem more straight-forward, but it turns out the
original definition is easier to use in proofs. We need to convince ourselves
that the (first) definition really accomplishes what we want. Namely, if a
relation R satisfies the property that ∀a, b ∈ S, aRb ∧ bRa =⇒ a=
b, then there will not actually be any pair of elements that are related in
both orders. One way to think about it is this: suppose that a and b are
distinct elements of S and that both aRb and bRa are true. The property now
guarantees that a = b which contradicts the notion that a and b are distinct.
This is a miniature proof by contradiction; if you assume there are a pair of
distinct elements that are related in both orders you get a contradiction, so
there aren’t!
A funny thing about the anti-symmetry property is this: When it is true
of a relation it is always vacuously true! The property is engineered in such a
way that when it is true, it forces that the statement in its antecedent never
really happens.
Transitivity is an extremely useful property as witnessed by the fact that
both equivalence relations and ordering relations must have this property.
When speaking of the transitive property of equality we say “Two things that
are equal to a third, are equal to each other.” When dealing with ordering
we may encounter statements like the following. “Since ’Aardvark’ precedes
’Bulwark’ in the dictionary, and since ’Bulwark’ precedes ’Catastrophe’, it is
plainly true that ’Aardvark’ comes before ’Catastrophe’ in the dictionary.”
Again, the definition of transitivity involves a conditional. Also, transi-
tivity may be viewed as the most complicated of the properties we’ve been
studying; it takes three universally quantified variables to state the property.
A relation R on a set S is transitive iff

∀a, b, c ∈ S, aRb ∧ bRc =⇒ aRc
We paraphrased transitivity as “Whenever there’s a roundabout route,

there’s a direct route.” In particular, what the definition says is that if
there’s a connection from a to b and from b to c (the roundabout route from
a to c) then there must be a connection from a to c (the direct route).
You’ll really need to watch out for relations that are transitive for vacuous
reasons. So long as one never has three elements a, b and c with aRb and bRc
the statement that defines transitivity is automatically true.
A very useful way of thinking about these various properties that relations
may have is in terms of what doesn’t happen when a relation has them. Before
we proceed, it is important that you do the following
Exercise. Find logical negations for the formal properties defining each of
the five properties.
If a relation R is reflexive we will never see a node that doesn’t have a

loop.
c
a
d
b
If a relation R is irreflexive we will never see a node that does have a loop!
c
a
d
b
If a relation R is symmetric we will never see a pair of nodes that are

connected in one direction only.
c
a
b d
If a relation R is anti-symmetric we will never see a pair of nodes that are

connected in both directions.
c
a
d
b
If a relation R is transitive the thing we will never see is a bit harder to

describe. There will never be a pair of arrows meeting head to tail without
there also being an arrow going from the tail of the first to the head of the
second.
c
a
b d
Exercises — 6.2
1. Consider the relation S defined by S = {(x, y) x is smarter than y}.

Is S symmetric or anti-symmetric? Explain.
2. Consider the relation A defined by A = {(x, y) x has the same astrological sign as y}.
Is A symmetric or anti-symmetric? Explain.
3. Explain why both of the relations just described (in problems 1 and 2)
have the transitive property.
4. For each of the five properties, name a relation that has it and a relation
that doesn’t.
6.3 Equivalence relations

The main idea of an equivalence relation is that it is something like equal-
ity, but not quite. Usually there is some property that we can name, so
that equivalent things share that property. For example Albert Einstein and
Adolf Eichmann were two entirely different human beings, if you consider
all the different criteria that one can use to distinguish human beings there
is little they have in common. But, if the only thing one was interested
in was a person’s initials, one would have to say that Einstein and Eich-
mann were equivalent. Future examples of equivalence relations will be less
frivolous. . . But first, the formal definition:
Definition. A relation R on a set S is an equivalence relation iff R is re-

flexive, symmetric and transitive.
Probably the most important equivalence relation we’ve seen to date is

“equivalence mod m” which we will denote using the symbol ≡m . This
relation may even be more interesting than actual equality! The reason for
this seemingly odd statement is that “equivalence mod m” gives us non-
trivial equivalence classes. Equivalence classes are one of the most potent
ideas in modern mathematics and it’s essential that you understand them,
so we’ll start with an example. Consider equivalence mod 5. What other
numbers is (say) 11 equivalent to? There are many! Any number that leaves
the same remainder as 11 when we divide it by 5. This collection is called
the equivalence class of 11 and is usually denoted using an overline — 11,
another notation that is often seen for the set of things equivalent to 11 is
11/ ≡5 .
11 = {. . . , −9, −4, 1, 6, 11, 16, . . .}
It’s easy to see that we will get the exact same set if we choose any other
6.3. EQUIVALENCE RELATIONS 245
element of the equivalence class (in place of 11), which leads us to an infinite
list of set equalities,
1 = 6 = 11 = . . .
And similarly,
2 = 7 = 12 = . . .
In fact, there are really just 5 different sets that form the equivalence classes
mod 5: 0, 1, 2, 3, and 4. (Note: we have followed the usual convention of
using the smallest possible non-negative integers as the representatives for
our equivalence classes.)
What we’ve been discussing here is one of the first examples of a quo-
tient structure. We start with the integers and “mod out” by an equivalence
relation. In doing so, we “move to the quotient” which means (in this in-
stance) that we go from Z to a much simpler set having only five elements:
{0, 1, 2, 3, 4}. In moving to the quotient we will generally lose a lot of in-
formation, but greatly highlight some particular feature – in this example,
properties related to divisibility by 5.
Given some equivalence relation R defined on a set S the set of equiva-
lence classes of S under R is denoted S/R (which is read “S mod R”). This
use of the slash – normally reserved for division – shouldn’t cause any con-
fusion since those aren’t numbers on either side of the slash but rather a set
and a relation. This notation may also clarify why some people denote the
equivalence classes above by 0/ ≡5 , 1/ ≡5 , 2/ ≡5 , 3/ ≡5 and 4/ ≡5 .
The set of equivalence classes forms a partition of the set S.
Definition. A partition P of a set S is a set of sets such that
[
S= X and ∀X, Y ∈ P, X 6= Y =⇒ X ∩ Y = ∅.
X∈P
In words, if you take the union of all the pieces of the partition you’ll
get the set S, and any pair of sets from the partition that aren’t identical
are disjoint. Partitions are an inherently useful way of looking at things,
although in the real world there are often problems (sets we thought were
disjoint turn out to have elements in common, or we discover something that
doesn’t fit into any of the pieces of our partition), in mathematics we usually
find that partitions do just what we would want them to do. Partitions divide
some set up into a number of convenient pieces in such a way that we’re
guaranteed that every element of the set is in one of the pieces and also so
that none of the pieces overlap. Partitions are a useful way of dissecting sets,
and equivalence relations (via their equivalence classes) give us an easy way
of creating partitions – usually with some additional structure to boot! The
properties that make a relation an equivalence relation (reflexivity, symmetry
and transitivity) are designed to ensure that equivalence classes exist and do
provide us with the desired partition. For the beginning proof writer this all
may seem very complicated, but take heart! Most of the work has already
been done for you by those who created the general theory of equivalence
relations and quotient structures. All you have to do (usually) is prove
that a given relation is an equivalence relation by verifying that it is indeed
reflexive, symmetric and transitive. Let’s have a look at another example.
In Number Theory, the square-free part of an integer is what remains
after we divide-out the largest perfect square that divides it. (This is also
known as the radical of an integer.) The following table gives the square-free
part, sf (n), for the first several values of n.
n 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
sf (n) 1 2 3 1 5 6 7 2 1 10 11 3 13 14 15 1 17 2 19 5
It’s easy to compute the square-free part of an integer if you know its
prime factorization – just reduce all the exponents mod 2. For example1
1
This is the size of largest sporadic finite simple group, known as “the Monster.”
808017424794512875886459904961710757005754368000000000
= 246 · 320 · 59 · 76 · 112 · 133 · 17 · 19 · 23 · 29 · 31 · 41 · 47 · 59 · 71
the square-free part of this number is
5 · 13 · 17 · 19 · 23 · 29 · 31 · 41 · 47 · 59 · 71
= 3504253225343845
which, while it is still quite a large number, is certainly a good bit smaller
than the original!
We will define an equivalence relation S on the set of natural numbers by
using the square-free part:
∀x, y ∈ N, xSy ⇐⇒ sf (x) = sf (y)
In other words, two natural numbers will be S-related if they have the
same square-free parts.
Exercise. What is 1/S?
Before we proceed to the proof that S is an equivalence relation we’d

like you to be cognizant of a bigger picture as you read. Each of the three
parts of the proof will have a similar structure. We will show that S has
one of the three properties by using the fact that = has that property. In
more advanced work this entire proof could be omitted or replaced by the
phrase “S inherits reflexivity, symmetry and transitivity from equality, and
is therefore an equivalence relation.” (Nice trick isn’t it? But before you’re
allowed to use it you have to show that you can do it the hard way . . . )
Theorem 6.3.1. The relation S defined by
∀x, y ∈ N, xSy ⇐⇒ sf (x) = sf (y)
is an equivalence relation on N.
Proof: We must show that S is reflexive, symmetric and transi-

tive.
reflexive — (Here we must show that ∀x ∈ N, xSx.) Let x be

an arbitrary natural number. Since sf (x) = sf (x) (this is the
reflexive property of =) it follows from the definition of S that
xSx.
symmetric — (Here we must show that ∀x, y ∈ N, xSy =⇒

ySx.) Let x and y be arbitrary natural numbers, and further
suppose that xSy. Since xSy, it follows from the definition of S
that sf (x) = sf (y), obviously then sf (y) = sf (x) (this is the
symmetric property of =) and so ySx.
transitive — (Here we must show that ∀x, y, z ∈ N, xSy ∧

ySz =⇒ xSz.) Let x, y and z be arbitrary natural numbers,
and further suppose that both xSy and ySz. From the definition
of S we deduce that sf (x) = sf (y) and sf (y) = sf (z). Clearly,
sf (x) = sf (z) (this deduction comes from the transitive property
of =), so xSz.
Q.E.D.
We’ll end this section with an example of an equivalence relation that

doesn’t “inherit” the three properties from equality.
A graph is a mathematical structure consisting of two sets, a set V of

points (a.k.a. vertices) and a set2 E of edges. The elements of E may be
either ordered or unordered pairs from V . If E consists of ordered pairs
we have a directed graph or digraph – the diagrams we have been using to
visualize relations! If E consists of unordered pairs then we are dealing with
an undirected graph. Since the undirected case is actually the more usual,
if the word “graph” appears without a modifier it is assumed that we are
talking about an undirected graph.
The previous paragraph gives a relatively precise definition of a graph in
terms of sets, however the real way to think of graphs is in terms of diagrams
where a set of dots are connected by paths. (The paths will, of course, need to
have arrows on them in digraphs.) Below are a few examples of the diagrams
that are used to represent graphs.
Two graphs are said to be isomorphic if they represent the same connec-
tions. There must first of all be a one-to-one correspondence between the
vertices of the two graphs, and further, a pair of vertices in one graph are
connected by some number of edges if and only if the corresponding vertices
in the other graph are connected by the same number of edges.
2
Technically, E is a so-called multiset in many instances – there may be several edges
that connect the same pair of vertices.
Exercise. The four examples of graphs above actually are two pairs of iso-
morphic graphs. Which pairs are isomorphic?
This word “isomorphic” has a nice etymology. It means “same shape.”

Two graphs are isomorphic if they have the same shape. We don’t have
the tools right now to do a formal proof (in fact we need to look at some
further prerequisites before we can really precisely define isomorphism), but
isomorphism of graphs is an equivalence relation. Let’s at least verify this
informally.
Reflexivity Is a graph isomorphic to itself? That is, does a graph have
the “same shape” as itself? Clearly!
Symmetry If graph A is isomorphic to graph B, is it also the case that
graph B is isomorphic to graph A? I.e. if A has the “same shape” as B,
doesn’t B have the same shape as A? Of course!
Transitivity Well . . . the answer here is going to be “Naturally!” but
let’s wait to delve into this issue when we have a usable formal definition for
graph isomorphism. The question at this stage should be clear though: If A
is isomorphic to B and B is isomorphic to C, then isn’t A isomorphic to C?
Exercises — 6.3
1. Consider the relation A defined by A = {(x, y) x has the same astrological sign as y}.
Show that A is an equivalence relation. What equivalence class under
A do you belong to?
2. Define a relation on the integers by xy ⇐⇒ x2 = y 2 . Show

that is an equivalence relation. List the equivalence classes x/ for
0 ≤ x ≤ 5.
3. The two diagrams below both show a famous graph known as the Pe-
tersen graph. The picture on the left is the usual representation which
emphasizes its five-fold symmetry. The picture on the right highlights
the fact that the Petersen graph also has a three-fold symmetry. Label
the right-hand diagram using the same letters (A through J) in order
to show that these two representations are truly isomorphic.
F
E B
J G
I H
D C
4. Define a relation Q on the set of all pairs of integers by (a, b)Q(c, d) ⇐⇒

ad = bc. Show that Q is an equivalence relation.
5. The relation Q defined in the previous problem partitions the set of all
pairs of integers into an interesting set of equivalence classes. Explain
why
Q = (Z × Z)/Q.
Ultimately, this is the “right” definition of the set of rational numbers!

6.4. ORDERING RELATIONS 253
6.4 Ordering relations

The prototype for ordering relations is ≤. Although a case could be made
for using < as the prototypical ordering relation. These two relations differ
in one important sense: ≤ is reflexive and < is irreflexive. Various authors,
having made different choices as to which of these is the more prototypical,
have defined ordering relations in slightly different ways. The majority view
seems to be that an ordering relation is reflexive (which means that ordering
relations are modeled after ≤). We would really like to take the contrary
position – we always root for the underdog – but one of our favorite ordering
relation (divisibility) is reflexive and it would be eliminated if we made the
other choice3 . So. . .
Definition. A relation R on a set S is an ordering relation iff R is reflexive,

anti-symmetric and transitive.
Now, we’ve used ≤ to decide what properties an ordering relation should

have, but we should point out that most ordering relations don’t do nearly
as good a job as ≤ does. The ≤ relation imposes what is known as a total
order on the sets that it acts on (you should note that it can’t be used to
compare complex numbers, but it can be placed between reals or any of the
sets of numbers that are contained in R.) Most ordering relations only create
what is known as a partial order on the sets they act on. In a total ordering
(a.k.a. a linear ordering) every pair of elements can be compared and we
can use the ordering relation to decide which order they go in. In a partial
ordering there may be elements that are incomparable.
Definition. If x and y are elements of a set S and R is an ordering relation

on S then we say x and y are comparable if xRy ∨ yRx.
3
If you insist on making the other choice, you will have a “strict ordering relation”
a.k.a. an “irreflexive ordering relation”
Definition. If x and y are elements of a set S and R is an ordering relation

on S then we say x and y are incomparable if neither xRy nor yRx is true.
Consider the set S = {1, 2, 3, 4, 6, 12}. If we look at the relation ≤ on

this set we get the following digraph.
1 2 3 4 6 12
On the other hand, perhaps you noticed these numbers are the divisors of
12. The divisibility relation will give us our first example of a partial order.
2 4
1 12
3 6
Exercise. Which elements in the above partial order are incomparable?

A set together with an ordering relation creates a mathematical structure

known as a partially ordered set. Since that is a bit of a mouthful, the
abbreviated form poset is actually heard more commonly. If one wishes to
refer to a poset it is necessary to identify both the set and the ordering
relation. Thus, if S is a set and R is an ordering relation, we write (S, R) to
denote the corresponding poset.
The digraphs given above for two posets having the same underlying set
provide an existence proof – the same set may have different orders imposed
upon it. They also highlight another issue – these digraphs for ordering
relations get pretty crowded! Hasse diagrams for posets (named after the
famous German mathematician Helmut Hasse) are a way of displaying all
the information in a poset’s digraph, but much more succinctly. There are
features of a Hasse diagram that correspond to each of the properties that
an ordering relation must have.
Since ordering relations are always reflexive, there will always be loops at
every vertex in the digraph. In a Hasse diagram we leave out the loops.
Since ordering relations are anti-symmetric, every edge in the digraph
will go in one direction or the other. In a Hasse diagram we arrange the
vertices so that that direction is upward – that way we can leave out all the
arrowheads without loosing any information.
The final simplification that we make in creating a Hasse diagram for a
poset has to do with the transitivity property – we leave out any connections
that could be deduced because of transitivity.
Hasse diagrams for the two orderings that we’ve been discussing are shown
in Figure 6.5
Often there is some feature of the elements of the set being ordered that
allows us to arrange a Hasse diagram in “ranks.” For example, consider
P({1, 2, 3}), the set of all subsets of a three element set – this set can be
partially ordered using the ⊆ relation. (Technically, we should verify that
12
12
6
4 6
2 3
2
1
1
Figure 6.5: Hasse diagrams of the set {1, 2, 3, 4, 6, 12} totally ordered by ≤
and partially ordered by |.
this relation is reflexive, anti-symmetric and transitive before proceeding, but

by now you know why subset containment is denoted using a rounded version
of ≤.) Subsets of the same size can’t possibly be included one in the other
unless they happen to be equal! This allows us to draw the Hasse diagram
for this set with the nodes arranged in four rows. (See Figure 6.6.)
{1, 2, 3}
{1, 2} {1, 3} {2, 3}
{1} {2} {3}
Figure 6.6: Hasse diagram for the power set of {1, 2, 3} partially ordered by
set containment.
Exercise. Try drawing a Hasse diagram for the partially ordered set
(P({1, 2, 3, 4}), ⊆).
Posets like (P({1, 2, 3}), ⊆) that can be layed out in ranks are known as
graded posets. Things in a graded poset that have the same rank are always
incomparable.
Definition. A graded poset is a triple (S, R, ρ), where S is a set, R is an

ordering relation, and ρ is a function from S to Z.
In the example we’ve been considering (the graded poset of subsets of a

set partially ordered by set inclusion), the grading function ρ takes a subset
to its size. That is, ρ(A) = |A|. Another nice example of a graded poset is the
set of divisors of some number partially ordered by the divisibility relation
(|). In this case the grading function takes a number to its total degree – the
sum of all the exponents appearing in its prime factorization. In Figure 6.7
we show the poset of divisors of 72 and indicate the grading.
72 = 23 · 32 5
24 = 23 · 31 36 = 22 · 32 4
8 = 23 12 = 22 · 31 18 = 21 · 32 3
4 = 22 6 = 21 · 31 9 = 32 2
21 31 1
1 0
Figure 6.7: Hasse diagram for the divisors of 72, partially ordered by divisi-
bility. This is a graded poset.
We will end this section by giving a small collection of terminology rele-

vant to partially ordered sets.
A chain in a poset is a subset of the elements, all of which are comparable.
If you restrict your attention to a chain within a poset, you will be looking
at a total order. An antichain in a poset is a subset of the elements, none of
which are comparable. Thus, for example, a subset of elements having the
same rank (in a graded poset) is an antichain. Chains and antichains are said
to be maximal if it is not possible to add further elements to them (whilst
maintaining the properties that make them chains and/or antichains). An
element x, that appears above another element y – and connected to it –
in a Hasse diagram is said to cover it. In this situation you may also say
that x is an immediate successor of y. A maximal element is an element
that is not covered by any other element. Similarly, a minimal element is
an element that is not a cover of any other element. If a chain is maximal,
it follows that it must contain both a maximal and a minimal element (with
respect to the surrounding poset). The collection of all maximal elements
forms an antichain, as does (separately) the collection of all minimal elements.
Finally, we have the notions of greatest element (a.k.a. top) and least element
(a.k.a. bottom) – the greatest element is greater than every other element in
the poset, the least element is smaller than every other element. Please be
careful to distinguish these concepts from maximal and minimal elements –
a greatest element is automatically maximal, and a least element is always
minimal, but it is possible to have a poset with no greatest element that
nevertheless has one or more maximal elements, and it is possible to have a
poset with no least element that has one or more minimal elements.
In the poset of divisors of 72, {2, 6, 12, 24} is a chain. Since it would be
possible to add both 1 and 72 to this chain and still have a chain, this chain
is not maximal. (But, of course, {1, 2, 6, 12, 24, 72} is.) On the other hand,
{8, 12, 18} is an antichain (indeed, this is a maximal antichain). This poset
has both a top and a bottom – 1 is the least element and 72 is the greatest
element. Notice that the elements which cover 1 (the least element) are the
prime divisors of 72.
Exercises — 6.4
1. In population ecology there is a partial order “predates” which basically

means that one organism feeds upon another. Strictly speaking this
relation is not transitive; however, if we take the point of view that
when a wolf eats a sheep, it is also eating some of the grass that the
sheep has fed upon, we see that in a certain sense it is transitive.
A chain in this partial order is called a “food chain” and so-called
apex predators are said to “sit atop the food chain”. Thus “apex
predator” is a term for a maximal element in this poset. When poisons
such as mercury and PCBs are introduced into an ecosystem, they
tend to collect disproportionately in the apex predators – which is why
pregnant women and young children should not eat sharks or tuna but
sardines are fine.
Below is a small example of an ecology partially ordered by “predates”
Fox Alligator
Cow
Duck Robin
Goose
Grass Worms
Find the largest antichain in this poset.

2. Referring to the poset given in exercise 1, match the following.
1. An (non-maximal) an-
tichain a. Grass
2. A maximal antichain b. Goose
3. A maximal element c. Fox
4. A (non-maximal) chain d. {Grass, Duck}
5. A maximal chain e. There isn’t one!
6. A cover for “Worms” f. {Fox, Alligator, Cow}
7. A least element g. {Cow, Duck, Goose}
8. A minimal element h. {Worms, Robin, Fox}
3. The graph of the edges of a cube is one in an infinite sequence of

graphs. These graphs are defined recursively by “Make two copies of
the previous graph then join corresponding nodes in the two copies
with edges.” The 0-dimensional ‘cube’ is just a single point. The
1-dimensional cube is a single edge with a node at either end. The
2-dimensional cube is actually a square and the 3-dimensional cube is
what we usually mean when we say “cube.”
Make a careful drawing of a hypercube – which is the name of the graph

that follows the ordinary cube in this sequence.
4. Label the nodes of a hypercube with the divisors of 210 in order to

produce a Hasse diagram of the poset determined by the divisibility
relation.
5. Label the nodes of a hypercube with the subsets of {a, b, c, d} in order

to produce a Hasse diagram of the poset determined by the subset
containment relation.
6.5. FUNCTIONS 263
6.5 Functions
The concept of a function is one of the most useful abstractions in mathemat-
ics. In fact it is an abstraction that can be further abstracted! For instance
an operator is an entity which takes functions as inputs and produces func-
tions as outputs, thus an operator is to functions as functions themselves
are to numbers. There are many operators that you have certainly encoun-
tered already – just not by that name. One of the most famous operators is
“differentiation,” when you take the derivative of some function, the answer
you obtain is another function. If two different people are given the same
differentiation problem and they come up with different answers, we know
that at least one of them has made a mistake! Similarly, if two calculations
of the value of a function are made for the same input, they must match.
The property we are discussing used to be captured by saying that a
function needs to be “well-defined.” The old school definition of a function
was:
Definition. A function f is a well-defined rule, that, given any input value

x produces a unique output4 value f (x).
A more modern definition of a function is the following.
Definition. A function is a binary relation which does not contain distinct

pairs having the same initial element.
When we think of a function as a special type of binary relation, the pairs

that are “in” the function have the form (x, f (x)), that is, they consist of an
input and the corresponding output.
We have gotten relatively used to relations “on” a set, but recall that the
more general situation is that a binary relation is a subset of A × B. In this
4
The use of the notation f (x) to indicate the output of function f associated with input
x was instituted by Leonard Euler, and so it is known as Euler notation.
setting, if the relation is actually a function f , we say that f is a function

from A to B. Now, quite often there are input values that simply don’t work
for a given function (for instance the well-known “you can’t take the square
root of a negative” rule). Also, it is often the case that certain outputs just
can’t happen. So, when dealing with a function as a relation contained in
A × B there are actually four sets that are of interest – the sets A and B
(of course) but also some sets that we’ll denote by A0 and B 0 . The set A0
consists of those elements of A that actually appear as the first coordinate
of a pair in the relation f . The set B 0 consists of those elements of B that
actually appear as the second coordinate of a pair in the relation f . A generic
example of how these four sets might look is given in Figure 6.8.
a v B
A b w
B0
0
A c x = f (c)
d y = f (d)
e z = f (e)
Figure 6.8: The sets related to an arbitrary function.
Sadly, only three of the sets we have just discussed are known to the
mathematical world. The set we have denoted A0 is called the domain of
the function f . The set we have denoted B 0 is known as the range of the
function f . The set we have denoted B is called the codomain of the function
6.5. FUNCTIONS 265
f . The set we have been calling A does not have a name. In fact, the formal
definition of the term “function” has been rigged so that there is no difference
between the sets A and A0 . This seems a shame, if you think of range and
domain as being primary, doesn’t it seem odd that we have a way to refer
to a superset of the range (i.e. the codomain) but no way of referring to a
superset of the domain?
Nevertheless, this is just the way it is . . . There is only one set on the
input side – the domain of our function.
The domain of any relation is expressed by writing Dom(R). Which is
defined as follows.
Definition. If R is a relation from A to B then Dom(R) is a subset of A

defined by
Dom(R) = {a ∈ A ∃b ∈ B, (a, b) ∈ R}
We should point out that the notation just given for the domain of a
relation R, (Dom(R)) has analogs for the other sets that are involved with
a relation. We write Cod(R) to refer the the codomain of the relation, and
Rng(R) to refer to the range.
Since we are now thinking of functions as special classes of relations, it
follows that a function is just a set of ordered pairs. This means that the
identity of a function is tied up, not just with a formula that gives the output
for a given input, but also what values can used for those inputs. Thus the
function f (x) = 2x defined on R is a completely different animal from the
function f (x) = 2x defined on N. If you really want to specify a function
precisely you must give its domain as well as a formula for it. Usually, one
does this by writing a formula, then a semicolon, then the domain. (E.g.
f (x) = x2 ; x ≥ 0.)
Okay, so, finally, we are prepared to give the real definition of a function.
Definition. If A and B are sets, then f is a function from A to B (which is

expressed symbolically by f : A −→ B), if and only if f is a subset of A × B,
Dom(f ) = A and ((a, b) ∈ f ∧ (a, c) ∈ f =⇒ b = c.
Recapping, a function must have its domain equal to the set A where its
inputs come from. This is sometimes expressed by saying that a function is
defined on its domain. A function’s range and codomain may be different
however. In the event that the range and codomain are the same (Cod(R) =
Rng(R)) we have a rather special situation and the function is graced by the
appellation “surjection.” The term “onto” is also commonly used to describe
a surjective function.
Exercise. There is an expression in mathematics, “Every function is onto

its range.” that really doesn’t say very much. Why not?
If one has elements x and y, of the domain and codomain, (respectively)

and y = f (x)5 then one may say that “y is the image of x” or that “x is a
preimage of y.” Take careful note of the articles used in these phrases – we
say “y is the image of x” but “x is a preimage of y.” This is because y is
uniquely determined by x, but not vice versa. For example, since the squares
of 2 and −2 are both 4, if we consider the function f (x) = x2 , the image of
(say) 2 is 4, but a preimage for 4 could be either 2 or −2.
It would be pleasant if there were a nice way to refer to the preimage of
some element, y, of the range. One notation that you have probably seen
before is “f −1 (y).” There is a major difficulty with writing down such a
thing. By writing “f −1 ” you are making a rather vast presumption – that
there actually is a function that serves as an inverse for f . Usually, there is
not.
One can define an inverse for any relation, the inverse is formed by simply
exchanging the elements in the ordered pairs that make up R.
5
Or, equivalently, (x, y) ∈ f .
6.5. FUNCTIONS 267
Definition. The inverse relation of a relation R is denoted R−1 and
R−1 = {(y, x) (x, y) ∈ R}.
In terms of graphs, the inverse and the original relation are related by
being reflections in the line y = x. It is possible for one, both, or neither of
these to be functions. The canonical example to keep in mind is probably
f (x) = x2 and its inverse.
The graph that we obtain by reflecting y = f (x) = x2 in the line y = x

doesn’t pass the vertical line test and so it is the graph of (merely) a relation
√
– not of a function. The function g(x) = x that we all know and love is not
truly the inverse of f (x). In fact this function is defined to make a specific
(and natural) choice – it returns the positive square root of a number. But
this leads to a subtle problem; if we start with a negative number (say −3)
and square it we get a positive number (9) and if we then come along and
take the square root we get another positive number (3). This is problematic
since we didn’t end up where we started which is what ought to happen if
we apply a function followed by its inverse.
We’ll try to handle the general situation in a bit, but for the moment
let’s consider the nice case: when the inverse of a function is also a function.
When exactly does this happen? Well, we have just seen that the inverse
of a function doesn’t necessarily pass the vertical line test, and it turns out
that that is the predominant issue. So, under what circumstances does the
inverse pass the vertical line test? When the original function passes the so-
called horizontal line test (every horizontal line intersects the graph at most
once). Thinking again about f (x) = x2 , there are some horizontal lines that
miss the graph entirely, but all horizontal lines of the form y = c where c
is positive will intersect the graph twice. There are many functions that do
pass the horizontal line test, for instance, consider f (x) = x3 . Such functions
are known as injections, this is the same thing as saying a function is “one-to-
one.” Injective functions can be inverted – the domain of the inverse function
of f will only be the range, Rng(f ), which as we have seen may fall short of
the being the entire codomain, since Rng(f ) ⊆ Cod(f ).
Let’s first define injections in a way that is divorced from thinking about
their graphs.
Definition. A function f (x) is an injection iff for all pairs of inputs x1 and
x2 , if f (x1 ) = f (x2 ) then x1 = x2 .
This is another of those defining properties that is designed so that when

it is true it is vacuously true. An injective function never takes two distinct
inputs to the same output. Perhaps the cleanest way to think about injective
functions is in terms of preimages – when a function is injective, preimages are
unique. Actually, this is a good time to mention something about surjective
functions and preimages – if a function is surjective, every element of the
codomain has a preimage. So, if a function has both of these properties it
means that every element of the codomain has one (and only one) preimage.
A function that is both injective and surjective (one-to-one and onto)
is known as a bijection. Bijections are tremendously important in mathe-
matics since they provide a way of perfectly matching up the elements of
two sets. You will probably spend a good bit of time in the future devising
6.5. FUNCTIONS 269
maps between sets and then proving that they are bijections, so we will start
practicing that skill now. . .
Ordinarily, we will show that a function is a bijection by proving sepa-
rately that it is both a surjection and an injection.
To show that a function is surjective we need to show that it is possible
to find a preimage for every element of the codomain. If we happen to know
what the inverse function is, then it is easy to find a preimage for an arbi-
trary element. In terms of the taxonomy for proofs that was introduced in
Chapter 3, we are talking about a constructive proof of an existential state-
ment. A function f is surjective iff ∀y ∈ Cod(f ), ∃x ∈ Dom(f ), y = f (x), so
to prove surjectivity is to find the x that “works” for an arbitrary y. If this
is done by literally naming x, we have proved the statement constructively.
To show that a function is an injection, we traditionally prove that the
property used in the definition of an injective function is true. Namely, we
suppose that x1 and x2 are distinct elements of Dom(f ) and that f (x1 ) =
f (x2 ) and then we show that actually x1 = x2 . This is in the spirit of a proof
by contradiction – if there were actually distinct elements that get mapped to
the same value then f would not be injective, but by deducing that x1 = x2
we are contradicting that presumption and so, are showing that f is indeed
an injection.
Let’s start by looking at a very simple example, f (x) = 2x − 1; x ∈ N.
Clearly this function is not a surjection if we are thinking that Cod(f ) = N
since the outputs are always odd. Let O = {1, 3, 5, 7, . . .} be the set of odd
naturals.
Theorem 6.5.1. The function f : N −→ O defined by f (x) = 2x − 1 is a

bijection from N to O.
Proof: First we will show that f is surjective. Consider an

arbitrary element y of the set O. Since y ∈ O it follows that y
is both positive and odd. Thus there is an integer k, such that

y = 2k + 1, but also y > 0. From this it follows that 2k + 1 > 0
and so k > −1/2. Since k is also an integer, this last inequality
implies that k ∈ Znoneg . (Recall that Znoneg = {0, 1, 2, 3, . . .}.) We
can easily verify that a preimage for y is k + 1, since f (k + 1) =
2(k + 1) − 1 = 2k + 2 − 1 = 2k + 1 = y.
Next we show that f is injective. Suppose that there are two input
values, x1 and x2 such that f (x1 ) = f (x2 ). Then 2x1 −1 = 2x2 −1
and simple algebra leads to x1 = x2 .
Q.E.D.
For a slightly more complicated example consider the function from N to

Z defined by
(
x/2 if x is even
f (x) =
−(x − 1)/2 if x is odd
This function does quite a handy little job, it matches up the natural
numbers and the integers in pairs. Every even natural gets matched with a
positive integer and every odd natural (except 1) gets matched with a neg-
ative integer (1 gets paired with 0). This function is really doing something
remarkable – common sense would seem to indicate that the integers must
be a larger set than the naturals (after all N is completely contained inside
of Z), but the function f defined above serves to show that these two sets
are exactly the same size!
Theorem 6.5.2. The function f defined above is bijective.
Proof: First we will show that f is surjective.

It suffices to find a preimage for an arbitrary element of Z. Sup-
pose that y is a particular but arbitrarily chosen integer. There
are two cases to consider: y ≤ 0 and y > 0.
6.5. FUNCTIONS 271
If y > 0 then x = 2y is a preimage for y. This follows easily since

x = 2y is obviously even and so x’s image will be defined by the
first case in the definition of f . Thus f (x) = f (2y) = (2y)/2 = y.
If y ≤ 0 then x = 1 − 2y is a preimage for y. Clearly, 1 − 2y is

odd whenever y is an integer, thus this value for x will fall into
the second case in the definition of f . So, f (x) = f (1 − 2y) =
−((1 − 2y) − 1)/2 = −(−2y)/2 = y.
Since the cases y > 0 and y ≤ 0 are exhaustive (that is, every y
in Z falls into one or the other of these cases), and we have found
a preimage for y in both cases, it follows that f is surjective.
Next, we will show that f is injective.
Suppose that x1 and x2 are elements of N and that f (x1 ) = f (x2 ).

Consider the following three cases: x1 and x2 are both even, both
odd, or have opposite parity.
If x1 and x2 are both even, then by the definition of f we have

f (x1 ) = x1 /2 and f (x2 ) = x2 /2 and since these functional values
are equal, we have x1 /2 = x2 /2. Doubling both sides of this leads
to x1 = x2 .
If x1 and x2 are both odd, then by the definition of f we have

f (x1 ) = −(x1 − 1)/2 and f (x2 ) = −(x2 − 1)/2 and since these
functional values are equal, we have −(x1 − 1)/2 = −(x2 − 1)/2.
A bit more algebra (doubling, negating and adding one to both
sides) leads to x1 = x2 .
If x1 and x2 have opposite parity, we will assume w.l.o.g. that

x1 is even and x2 is odd. The equality f (x1 ) = f (x2 ) becomes
x1 /2 = −(x2 − 1)/2. Note that x1 ≥ 2 so f (x1 ) = x1 /2 ≥ 1.
Also, note that x2 ≥ 1 so
x2 − 1 ≥ 0
(x2 − 1)/2 ≥ 0
−(x2 − 1)/2 ≤ 0
f (x2 ) ≤ 0
therefore we have a contradiction since it is impossible for the two

values f (x1 ) and f (x2 ) to be equal while f (x1 ) ≥ 1 and f (x2 ) ≤ 0.
Since the last case under consideration leads to a contradiction,
it follows that x1 and x2 never have opposite parities, and so the
first two cases are exhaustive – in both of those cases we reached
the desired conclusion that x1 = x2 so it follows that f is injective.
Q.E.D.
We’ll conclude this section by mentioning that the ideas of “image” and
“preimage” can be extended to sets. If S is a subset of Dom(f ) then the
image of S under f is denoted f (S) and
f (S) = {y ∃x ∈ Dom(f ), x ∈ S ∧ y = f (x)}.
Similarly, if T is a subset of of Rng(f ) we can define something akin to

the preimage. The inverse image of the set T under the function f is denoted
f −1 (T ) and
f −1 (T ) = {x ∃y ∈ Cod(f ), y ∈ T ∧ y = f (x)}.
Essentially, we have extended the function f so that it goes between the

power sets of its codomain and range! This new notion gives us some elegant
ways of restating what it means to be surjective and injective.
6.5. FUNCTIONS 273
A function f is surjective iff f (Dom(f )) = Cod(f ).

A function f is injective iff the inverse images of singletons are always
singletons. That is,
∀y ∈ Rng(f ), ∃x ∈ Dom(f ), f −1 ({y}) = {x}.

Exercises — 6.5
1. Find a bijection from the set of odd squares, {1, 9, 25, 49, . . .}, to the
non-negative integers, Znoneg = {0, 1, 2, 3, . . .}. Prove that the function
you just determined is both injective and surjective. Find the inverse
function of the bijection above.
2. The natural logarithm function ln(x) is defined by a definite integral

with the variable x in the upper limit.
Z x
1
ln(x) = dt.
t=1 t
From this definition we can deduce that ln(x) is strictly increasing on

its entire domain, (0, ∞). Why is this true?
We can use the above definition with x = 2 to find the value of ln(2) ≈
.693. We will also take as given the following rule (which is valid for
all logarithmic functions).
ln(ab ) = b ln(a)
Use the above information to show that there is neither an upper bound
nor a lower bound for the values of the natural logarithm. These facts
together with the information that ln is strictly increasing show that
Rng(ln) = R.
3. Georg Cantor developed a systematic way of listing the rational num-

bers. By “listing” a set one is actually developing a bijection from N to
that set. The method known as “Cantor’s Snake” creates a bijection
from the naturals to the non-negative rationals. First we create an
infinite table whose rows are indexed by positive integers and whose
6.5. FUNCTIONS 275
columns are indexed by non-negative integers – the entries in this table

are rational numbers of the form “column index” / “row index.” We
then follow a snake-like path that zig-zags across this table – whenever
we encounter a rational number that we haven’t seen before (in lower
terms) we write it down. This is indicated in the diagram below by
circling the entries.
Effectively this gives us a function f which produces the rational num-
ber that would be found in a given position in this list. For example
f (1) = 0/1, f (2) = 1/1 and f (5) = 1/3.
What is f (26)? What is f (30)? What is f −1 (3/4)? What is f −1 (6/7)?
0 1 2 3 4 5 6 7 8
1 0/1 1/1 2/1 3/1 4/1 5/1 6/1 7/1 8/1
2 0/2 1/2 2/2 3/2 4/2 5/2 6/2 7/2 8/2
3 0/3 1/3 2/3 3/3 4/3 5/3 6/3 7/3 8/3
4 0/4 1/4 2/4 3/4 4/4 5/4 6/4 7/4 8/4
5 0/5 1/5 2/5 3/5 4/5 5/5 6/5 7/5 8/5
6 0/6 1/6 2/6 3/6 4/6 5/6 6/6 7/6 8/6
7 0/7 1/7 2/7 3/7 4/7 5/7 6/7 7/7 8/7
8 0/8 1/8 2/8 3/8 4/8 5/8 6/8 7/8 8/8

6.6 Special functions

There are a great many functions that fail the horizontal line test which we
nevertheless seem to have inverse functions for. For example, x2 fails HLT
√
but x is a pretty reasonable inverse for it – one just needs to be careful
about the “plus or minus” issue. Also, sin x fails HLT pretty badly; any
horizontal line y = c with −1 ≤ c ≤ 1 will hit sin x infinitely many times.
But look! Right here on my calculator is a button labeled “sin−1 .”6 This
apparent contradiction can be resolved using the notion of restriction.
Definition. Given a function f and a subset D of its domain, the restriction

of f to D is denoted f D
and
f D
= {(x, y) x ∈ D ∧ (x, y) ∈ f }.
The way we typically use restriction is to eliminate any regions in Dom(f )

that cause f to fail to be one-to-one. That is, we choose a subset D ⊆ Dom(f )
so that f D
is an injection. This allows us to invert the restricted version of
f . There can be problems in doing this, but if we are careful about how we
choose D, these problems are usually resolvable.
Exercise. Suppose f is a function that is not one-to-one, and D is a subset

of Dom(f ) such that f D
is one-to-one. The restricted function f D
has an
inverse which we will denote by g. Note that g is a function from Rng(f D
)
to D. Which of the following is always true:
f (g(x)) = x or g(f (x)) = x?

6
It might be labeled “asin” instead. The old-style way to refer to the inverse of a trig.
function was arc-whatever. So the inverse of sine was arcsine, the inverse of tangent was
arctangent.
6.6. SPECIAL FUNCTIONS 277
Technically, when we do the process outlined above (choose a domain D

so that the restriction f D
is invertible, and find that inverse) the function
we get is a right inverse for f .
Let’s take a closer look at the inverse sine function. This should help us
to really understand the “right inverse” concept.
A glance at the graph of y = sin x will certainly convince us that this
function is not injective, but the portion of the graph shown in bold below
passes the horizontal line test.
If we restrict the domain of the sine function to the closed interval [−π/2, π/2],
we have an invertible function. The inverse of this restricted function is the
function we know as sin−1 (x) or arcsin(x). The domain and range of sin−1 (x)
are (respectively) the intervals [−1, 1] and [−π/2, π/2].
Notice that if we choose a number x in the range −1 ≤ x ≤ 1 and apply
the inverse sine function to it, we will get a number between −π/2 and π/2
– i.e. a number we can interpret as an angle in radian measure. If we then

proceed to calculate the sine of this angle, we will get back our original
number x.
On the other hand, if we choose an angle first, then take the sine of it to
get a number in [−1, 1] and then take the inverse sine of that, we will only
end up with the same angle we started with if we chose the angle so that it
lay in the interval [−π/2, π/2].
Exercise. We get a right inverse for the cosine function by restricting it to

the interval [0, π]. What are the domain and range of cos−1 ?
The winding map is a function that goes from R to the unit circle in the
x–y plane, defined by
W (t) = (cos t, sin t).
One can think of this map as literally winding the infinitely long real line
around and around the circle. Obviously, this is not an injection – there are
an infinite number of values of t that get mapped to (for instance) the point
(1, 0), t can be any integer multiple of 2π.
Exercise. What is the set W −1 ({(0, 1)}) ?
If we restrict W to the half-open interval [0, 2π) the restricted function

W [0,2π)
is an injection. The inverse function is not easy to write down, but
it is possible to express (in terms of the inverse functions of sine and cosine)
if we consider the four cases determined by what quadrant a point on the
unit circle may lie in.
Exercise. Suppose (x, y) represents a point on the unit circle. If (x, y) hap-
pens to lie on one of the coordinate axes we have
W −1 ((1, 0)) = 0
W −1 ((0, 1)) = π/2
W −1 ((−1, 0)) = π
W −1 ((0, −1)) = 3π/2.
If neither x nor y is zero, there are four cases to consider. Write an

expression for W −1 ((x, y)) using the cases (i) x > 0 ∧ y > 0, (ii) x <
0 ∧ y > 0, (iii) x < 0 ∧ y < 0 and (iv) x > 0 ∧ y < 0.
This last example that we have done (the winding map) was unusual in
that the outputs were ordered pairs. In thinking of this map as a relation
(that is, as a set of ordered pairs) we have an ordered pair in which the second
element is an ordered pair! Just for fun, here is another way of expressing
the winding map:
W = {(t, (cos t, sin t)) t ∈ R}
When dealing with very complicated expressions involving ordered pairs,

or more generally, ordered n-tuples, it is useful to have a way to refer suc-
cinctly to the pieces of a tuple.
Let’s start by considering the set P = R × R — i.e. P is the x–y plane.
There are two functions, whose domain is P that “pick out” the x, and/or
y coordinate. These functions are called π1 and π2 , π1 is the projection onto
the first coordinate and π2 is the projection onto the second coordinate.7
7
Don’t think of the usual π ≈ 3.14159 when looking at π1 and π2 . These functions
are named as they are because π is the Greek letter corresponding to ‘p’ which stands for
“projection.”
Definition. The function π1 : R × R −→ R known as projection onto the

first coordinate is defined by
π1 ((x, y)) = x.
The definition of π2 is entirely analogous.

You should note that these projection functions are very bad as far as
being one-to-one is concerned. For instance, the preimage of 1 under the
map π1 consists of all the points on the vertical line x = 1. That’s a lot of
preimages! These guys are so far from being one-to-one that it seems im-
possible to think of an appropriate restriction that would become invertible.
Nevertheless, there is a function that provides a right inverse for both π1 and
π2 . Now, these projection maps go from R × R to R so an inverse needs to
be a map from R to R × R. What is a reasonable way to produce a pair of
real numbers if we have a single real number in hand? There are actually
many ways one could proceed, but one reasonable choice is to create a pair
where the input number appears in both coordinates. This is the so-called
diagonal map, d : R × R −→ R, defined by d(a) = (a, a).
Exercise. Which of the following is always true,
d(π1 ((x, y)) = (x, y) or π1 (d(x)) = x?
There are a few other functions that it will be convenient to introduce at

this stage. All of them are aspects of the characteristic function of a subset,
so we’ll start with that.
Whenever we have a subset/superset relationship, S ⊆ D, it is possible
to define a function whose codomain is {0, 1} which performs a very useful
task – if an input x is in the set S the function will indicate this by returning
1, otherwise it will return 0. The function which has this behavior is known
as 1S , and is called the characteristic function of the subset S (There are
those who use the term indicator function of S for 1S .) By definition, D is

the domain of this function.
1S : D −→ {0, 1}
(
1 if x ∈ S
1S (x) =
0 otherwise
Exercise. If you have the characteristic function of a subset S, how can you
create the characteristic function of its complement, S.
A characteristic function may be thought of as an embodiment of a mem-

bership criterion. The logical open sentence “x ∈ S” being true is the same
thing as the equation “1S (x) = 1.” There is a notation, growing in popu-
larity, that does the same thing for an arbitrary open sentence. The Iverson
bracket notation uses the shorthand [P (x)] to represent a function that sends
any x that makes P (x) true to 1, and any inputs that make P (x) false will
get sent to 0.
(
1 if P (x)
[P (x)] =
0 otherwise
The Iverson brackets can be particularly useful in expressing and simpli-

fying sums. For example, we can write 24
P
i=1 [2 | i] to find the number of even
natural numbers less than 25. Similarly, we can write 24
P
i=1 [3 | i] to find the
number of natural numbers less than 25 that are divisible by 3.
Exercise. What does the following formula count?
24
X
[2 | i] + [3 | i] − [6 | i]
i=1
There is a much more venerable notation known as the Kronecker delta

that can be thought of as a special case of the idea inherent in Iverson
brackets. We write δij as a shorthand for a function that takes two inputs,
δ(i, j) is 1 if and only if i and j are equal.
(
1 if i = j
δij =
0 otherwise
The corresponding Iverson bracket would simply be [i = j].

We’ll end this section with a function that will be especially important
in Chapter 8. If we have an arbitrary subset of the natural numbers, we
can associate it with an infinite string of 0’s and 1’s. By sticking a decimal
point in front of such a thing, we get binary notation for a real number in
the interval [0, 1]. There is a subtle problem that we’ll deal with when we
study this function in more detail in Chapter 8 — some real numbers can
be expressed in two different ways in base 2. For example, 1/2 can either be
written as .1 or as .01 (where, as usual, the overline indicates a pattern that
repeats forever). For the moment, we are talking about defining a function
φ whose domain is P(N) and whose codomain is the set of all infinite binary
strings. Let us denote these binary expansions by .b1 b2 b3 b4 . . .. Suppose
A is a subset of N, then the binary expansion associated with A will be
determined by bi = 1A (i). (Alternatively, we can use the Iverson bracket
notation: bi = [i ∈ A].)
The function φ defined in the last paragraph turns out to be a bijection –
given a subset we get a unique binary expansion, and given binary expansion
we get (using φ−1 ) a unique subset of N.
A few examples will probably help to clarify this function’s workings.
Consider the set {1, 2, 3} ⊆ N, the binary expansion that this corresponds to
will have 1’s in the first three positions after the decimal – φ({1, 2, 3}) = .111
this is the number written .875 in decimal. The infinite repeating binary
number .01 is the base-2 representation of 1/3, from this it is easy to see that
.01 is the image of the set of odd naturals, {1, 3, 5, . . .}.
Exercise. Find the binary representation for the real number which is the
image of the set of even numbers under φ.
Exercise. Find the binary representation for the real number which is the
image of the set of triangular numbers under φ. (Recall that the triangular
numbers are T = {1, 3, 6, 10, 15, . . .}.)
Exercises — 6.6
1. The n-th triangular number, denoted T (n), is given by the formula

T (n) = (n2 + n)/2. If we regard this formula as a function from R to
R, it fails the horizontal line test and so it is not invertible. Find a
suitable restriction so that T is invertible.
2. The usual algebraic procedure for inverting T (x) = (x2 +x)/2 fails. Use
your knowledge of the geometry of functions and their inverses to find
a formula for the inverse. (Hint: it may be instructive to first invert
the simpler formula S(x) = x2 /2 — this will get you the right vertical
scaling factor.)
3. What is π2 (W (t))?
4. Find a right inverse for f (x) = |x|.
5. In three-dimensional space we have projection functions that go onto

the three coordinate axes (π1 , π2 and π3 ) and we also have projections
onto coordinate planes. For example, π12 : R × R × R −→ R × R,
defined by
π12 ((x, y, z)) = (x, y)
is the projection onto the x–y coordinate plane.
The triple of functions (cos t, sin t, t) is a parametric expression for a

helix. Let H = {(cos t, sin t, t) t ∈ R} be the set of all points on the
helix. What is the set π12 (H) ? What are the sets π13 (H) and π23 (H)?
6. Consider the set {1, 2, 3, . . . , 10}. Express the characteristic function

of the subset S = {1, 2, 3} as a set of ordered pairs.
7. If S and T are subsets of a set D, what is the product of their charac-

teristic functions 1S · 1T ?
8. Evaluate the sum
10
X 1
· [i is prime].
i=1
i
Chapter 7
Proof techniques III —

Combinatorics
Tragedy is when I cut my finger. Comedy is when you fall into an open sewer
and die. –Mel Brooks
7.1 Counting
Many results in mathematics are answers to “How many . . . ” questions.
“How many subsets does a finite set have?”
“How many handshakes will transpire when n people first meet?”
“How many functions are there from a set of size n to a set of size m?”
The title of this section, “Counting,” is not intended to evoke the usual
process of counting sheep, or counting change. What we want is to be able
to count some collection in principle so that we will be able to discover a
formula for its size.
There are two principles that will be indispensible in counting things.
These principles are simple, yet powerful, and they have been named in the
most unimaginative way possible. The “multiplication rule” which tells us
287
288 CHAPTER 7. PROOF TECHNIQUES III — COMBINATORICS
when we should multiply, and the “addition rule” which tells us when we
should add.
Before we describe these principles in detail, we’ll have a look at a simpler
problem which is most easily described by an example: How many integers
are there in the list (7, 8, 9, . . . 44)? We could certainly write down all the
integers from 7 to 44 (inclusive) and then count them – although this wouldn’t
be the best plan if the numbers 7 and 44 were replaced with (say) 7, 045, 356
and 22, 355, 201. A method that does lead to a generalized ability to count
the elements of a finite sequence arises if we think carefully about what
exactly a finite sequence is.
Definition. A sequence from a set S is a function from N to S.
Definition. A finite sequence from a set S is a function from {0, 1, 2, . . . , n}

to S, where n is some particular (finite) integer.
Now it is easy to see that there are n+1 elements in the set {0, 1, 2, . . . , n}
so counting the elements of a finite sequence will be easy if we can determine
the function involved and figure out what n is by inverting it (n is an inverse
image for the last element in a listing of the sequence).
In the example that we started with, the function is f (x) = x + 7. We
can sum up the process that allows us to count the sequence by saying “there
is a one-to-one correspondence between the lists
(7, 8, 9, . . . , 44)
and
(0, 1, 2, . . . , 37)
and the later has 38 entries.”

7.1. COUNTING 289
More generally, if there is a list of consecutive numbers beginning with k

and ending with n, there will be n − k + 1 entries in the list. Lists of consecu-
tive integers represent a relatively simple type of finite sequence. Usually we
would have some slightly more interesting function that we’d need to invert.
The following exercise involves inverting the function (x + 5)2 .
Exercise. How many integers are in the list (25, 36, 49, . . . , 10000) ?
We will have a lot more practice with counting the elements of sequences
in the exercises at the end of this section, let’s continue on our tour of count-
ing by having a look at the addition rule.
The addition rule says that it is appropriate to add if we can partition a
collection into disjoint pieces. In other words, if a set S is the union of two
or more subsets and these subsets are mutually disjoint, we can find the size
of S by adding the sizes of the subsets.
In the game Yahtzee, one rolls 5 dice and (optionally) performs a second
roll of some or all of the dice. The object is to achieve several final con-
figurations that are modelled after the hands in Poker. In particular, one
configuration, known as a “full house,” is achieved by having two of one
number and three of another. (Colloquially, we say “three-of-a-kind plus a
pair is a full house.”)
Now, we could use Yahtzee “hands” to provide us with a whole collection
of counting problems once we have our basic counting principles, but for
the moment we just want to make a simple (and obvious) point about “full
houses” – the pair is either smaller or larger than the three-of-a-kind. This
means we can partition the set of all possible full houses into two disjoint
sets – the full houses consisting of a small pair and a larger three-of-a-kind
and those where the pair is larger than the three-of-a-kind. If we can find
some way of counting these two cases seperately, then the total number of
full houses will be the sum of these numbers.
Figure 7.1: In Yahtzee, a full house may consist of a pair and a larger three-
of-a-kind, or vice versa.
The multplication rule gives us a way of counting things by thinking

about how we might construct them. The numbers that are multiplied are
the number of choices we have in the construction process. Surprisingly
often, the number of choices we can make in a given stage of constructing
some configuration is independent of the choices that have gone before – if
this is not the case the multiplication rule may not apply.
If some object can be constructed in k stages, and if in the first stage we
have n1 choices as to how to proceed, in the second stage we have n2 choices,
et cetera. Then the total number of such objects is the product n1 n2 · · · nk .
A permutation of an n-set (w.l.o.g. {1, 2, . . . , n}) is an ordered n-tuple
where each entry is a distinct element of the n-set. Generally, a permutation
may be regarded as a bijection from an n-set to itself. Our first use of
the multiplication rule will be to count the total number of permutations of
{1, 2, 3, . . . , n}.
Let’s start by counting the permutations of {1, 2, 3}. A permutation will
be a 3-tuple containing the numbers 1, 2 and 3 in some order. We will think
about building such a thing in three stages. First, we must select a number to
go in the first position – there are 3 choices. Having made that choice, there
will only be two possibilities for the number in the second position. Finally
7.1. COUNTING 291
there is just one number remaining to put in the third position1 . Thus there
are 3 · 2 · 1 = 6 permutations of a 3 element set.
The general rule is that there are n! permutations of {1, 2, . . . , n}.
There are times when configurations that are like permutations (in that
they are ordered and have no duplicates) but don’t consist of all n numbers
are useful.
Definition. A k-permutation from an n-set is an ordered selection of k dis-

tinct elements from a set of size n.
There are certain natural limitations on the value of k, for instance k can’t
be negative – although (arguably) k can be 0, it makes more sense to think
of k being at least 1. Also, if k exceeds n we won’t be able to find any k-
permutations, since it will be impossible to meet the “distinct” requirement.
If k and n are equal, there is no difference between a k-permutation and an
ordinary permutation. Therefore, we ordinarily restrict k to lie in the range
0 < k < n.
The notation P (n, k) is used for the total number of k-permutations of
a set of size n. For example, P (4, 2) is 12, since there are twelve different
ordered pairs having distinct entries where the entries come from {1, 2, 3, 4}.
Exercise. Write down all twelve 2-permutations of the 4-set {1, 2, 3, 4}.
Counting k-permutations using the multiplication rule is easy. We build

a k-permutation in k stages. In stage 1, we pick the first element in the
permutation – there are n possible choices. In stage 2, we pick the second
element – there are now only n − 1 choices since we may not repeat the first
entry. We keep going like this until we’ve picked k entries. The number
P (n, k) is the product of k numbers beginning with n and descending down
1
People may say you have “no choice” in this last situation, but what they mean is
that you have only one choice.
to n − k + 1. To verify that n − k + 1 is really the right lower limit, check

that there are indeed k entries in the sequence
(n, n − 1, n − 2, . . . n − k + 1).
This verification may be easier if we rewrite the sequence as
(n − 0, n − 1, n − 2, . . . n − (k − 1)).
Let’s have a look at another small example – P (8, 4). There will be 8
choices for the first entry in a 4-tuple, 7 choices for the second entry, 6 choices
for the third entry and 5 choices for the last entry. (Note that 5 = 8 − 4 + 1.)
Thus P (8, 4) = 8 · 7 · 6 · 5 = 1680.
Finally, we should take note that it is relatively easy to express P (n, k)
using factorials. If we divide a number factorialized by some smaller number
factorialized, we will get a descending product just like those above.
Exercise. What factorial would we divide 8! by in order to get P (8, 4)?
n!
The general rule is that P (n, k) = (n−k)!
.
If we were playing a card game in which we were dealt 5 cards from a deck
of 52, we would receive our cards in the form of P (52, 5) = 52·51·50·49·48 =
311875200 ordered 5-tuples. Normally, we don’t really care about what order
the cards came to us in. In a card game one ordinarily begins sorting the
cards so as to see what hand one has – this is a sure sign that the order the
cards were dealt is actually immaterial. How many different orders can five
cards be put in? The answer to this question is 5! = 120 since what we are
discussing is nothing more than a permutation of a set of size 5. Thus, if we
say that there are 311,875,200 different possible hands in 5-card poker, we
are overcounting things by quite a bit! Any given hand will appear 120 times
in that tabulation, which means the right value is 311875200/120 = 2598960.
7.1. COUNTING 293
Okay, so there around 2.6 million different hands in 5-card poker. Unless
you plan to become a gambler this isn’t really that useful of a piece of in-
formation – but if you generalize what we’ve done in the paragraph above,
you’ll have found a way to count unordered collections of a given size taken
from a set.
A k-combination from an n-set is an unordered selection, without rep-
etitions, of k things out of n. This is the exact same thing as a subset of
size k of a set of size n, and the number of
such
things is denoted by several
n
different notations – C(n, k), nCk and among them2 . We can come
k
up with a formula for C(n, k) by a slightly roundabout argument. Suppose
we think of counting the k-permutations of n things using the multiplicaton
rule in a different way then we have previously. We’ll build a k-permutation
in two stages. First we’ll choose k symbols to put into our permutation –
which can be done in C(n, k) ways. And second, we’ll put those k sym-
bols into a particular order – which can be accomplished in k! ways. Thus
n!
P (n, k) = C(n, k) · k!. Since we already know that P (n, k) = (n−k)!
, we can
substitute and solve to obtain
n!
C(n, k) = .
k! · (n − k)!
It is possible to partition many counting problems into 4 “types” based
on the answers to two questions:
Is order important in the configurations being counted?
Are we allowed to have repeated elements in a configuration?
Suppose that we are in the general situation of selecting k things out of
a set of size n. It should be possible to write formulas involving n and k in
the four cells of the following table.
n n
2

Watch out for the notation, it is easy to confuse it with the fraction
k k . They
are not the same — the fraction bar is supposed to be missing in nk .

Does order matter?

Yes No
Are repeats okay?
No
Yes
Ordered with repetition

Selecting a PIN number3 for your bank account is a good example of
the kind of problem that is dealt with in the lower left part of the table.
Obviously, the order in which you key-in the digits of your PIN is important.
If one’s number is 1356, it won’t do to put in 6531! Also there is no reason
that we couldn’t have repeated digits in a PIN. (Although someone who
chooses a PIN like 3333 is taking a bit of a security risk! A bad guy looking
over your shoulder may easily discern what your PIN is.) A PIN is an
ordered selection of 4 things out of 10, where repetition is allowed. There are
104 possible PINs. We can determine this by thinking of the multiplication
principle – there are 10 choices for the first digit of our PIN, since repetition
is okay there are still 10 choices for our second digit, then (still) 10 choices
for the third digit as well as the fourth digit.
In general, when selecting k things out of n possibilities, where order
counts and repetition is allowed, there are nk possible selections.
Ordered without repetition
3
The phrase “PIN number” is redundant. The ‘N’ in PIN stands for “number.” Any-
way, a PIN is a four digit (secret) number used to help ensure that automated banking
(such as withdrawing your life’s savings) is only done by an authorized individual.
7.1. COUNTING 295
Suppose that one wishes to come up with a password for a computer

account. There are 52 letters (both upper and lower case) 10 numerals and
32 symbols and punctuation marks – for a total of 94 different characters
that may be used. Some system administrators can be very paranoid about
passwords that might be guessable – for instance no password that appears
in a dictionary should ever be used on a system where security is a concern.
Suppose that your system administrator will reject any password that has
repeated symbols, and that passwords must have 8 characters. How many
passwords are possible?
This is an instance of a counting problem where we are selecting 8 things
out of a set of size 94 – clearly order is important and the system adminis-
trator’s restriction means that we may not have repeats. The multiplication
rule tells us that there are 94 · 93 · 92 · 91 · 90 · 89 · 88 · 87 = 4488223369069440
different passwords. And in the general case (selecting k things out of a set of
size n, without repetition, and with order counting) there will be n!/(n − k)!
possibilities. This is the number we have denoted previously by P (n, k).
Unordered without repetition
This is also a case that we’ve considered previously. If we are chosing k
things out of n and order is unimportant and there can be no repetitions, then
n!
what we are describing is a k-subset of the n-set. There are C(n, k) = k!(n−k)!
distinct subsets. Here, we’ll give an example that doesn’t sound like we’re
talking about counting subsets of a particular size (although we really are!)
How many different sequences of 6 strictly increasing numbers can we
choose from {1, 2, 3, . . . 20}?
Obviously, listing all such sequences would be an arduous task. We
might start with (1, 2, 3, 4, 5, 6) and try to proceed in some orderly fashion
to (15, 16, 17, 18, 19, 20), but unfortunately there are 38,760 such sequences
so unless we enlist the aid of a computer we are unlikely to finish this job
in a reasonable time. The number we’ve just given (38,760) is C(20, 6) and
so it would seem that we’re claiming that this problem is really unordered
selection without repetition of 6 things out of 20. Well, actually, some parts
of this are clearly right – we are selecting 6 things from a set of size 20, and
because our sequences are supposed to be strictly increasing there will be no
repetitions – but, a strictly increasing sequence is clearly ordered and the
formula we are using is for unordered collections.
By specifying a particular ordering (strictly increasing) on the sequences
we are counting above, we are actually removing the importance of order.
Put another way: if order really mattered, the symbols 1 through 6 could
be put into 720 differenct orders – but we only want to count one of those
possibilities. Put another other way: there is a one-to-one correspondence
between a 6-subset of {1, 2, 3, . . . 20} and a strictly increasing sequence. Just
make sure the subset is written in increasing order!
Okay, at this point we have filled-in three out of the four cells in our table.
Does order matter?
Yes No
No
Are repeats okay?
n! n!
P (n, k) = (n−k)!
C(n, k) = k!(n−k)!
Yes
nk
7.1. COUNTING 297
What kinds of things are we counting in the lower right part of the table?
Unordered selections of k things out of n possibilities where there may (or
may not!) be repetitions. The game Yahtzee provides a nice example of
this type of configuration. When we roll 5 dice, we do not do so one-at-
a-time, rather, we roll them as a group – the dice are indistinguishable so
there is no way to order our set of 5 outcomes. In fact, it would be quite
reasonable to, after one’s roll, arrange the die in (say) increasing order. We’ll
repeat a bit of advice that was given previously: if one is free to rearrange a
configuration to suit one’s needs, that is a clue that order is not important
in the configurations under consideration. Finally, are repetitions allowed?
The outcomes in Yahtzee are 5 numbers from the set {1, 2, 3, 4, 5, 6}, and
while it is possible to have no repetitions, that is a pretty special outcome!
In general, the same number can appear on two, or several, or even all 5 of
the die4 .
So, how many different outcomes are there when one rolls five dice? To
answer this question it will be helpful to think about how we might express
such an outcome. Since order is unimportant, we can choose to put the
numbers that appear on the individual die in whatever order we like. We
may as well place them in increasing order. There will be 5 numbers and
each number is between 1 and 6. We can list the outcomes systematically by
starting with an all-ones Yahtzee:
(1,1,1,1,1) (1,1,1,1,2) (1,1,1,1,3) (1,1,1,1,4) (1,1,1,1,5) (1,1,1,1,6)

(1,1,1,2,2) (1,1,1,2,3) (1,1,1,2,4) (1,1,1,2,5) (1,1,1,2,6) (1,1,1,3,3)
(1,1,1,3,4) (1,1,1,3,5) (1,1,1,3,6) (1,1,1,4,4) (1,1,1,4,5) (1,1,1,4,6)
(1,1,1,5,5) (1,1,1,5,6) (1,1,1,6,6) (1,1,2,2,2) (1,1,2,2,3) (1,1,2,2,4)
(1,1,2,2,5) (1,1,2,2,6) (1,1,2,3,3) (1,1,2,3,4) (1,1,2,3,5) (1,1,2,3,6)
(1,1,2,4,4) (1,1,2,4,5) (1,1,2,4,6) (1,1,2,5,5) (1,1,2,5,6) (1,1,2,6,6)
4
When this happens you are supposed to jump in the air and yell “Yahtzee!”
(1,1,3,3,3) (1,1,3,3,4) (1,1,3,3,5) (1,1,3,3,6) (1,1,3,4,4) (1,1,3,4,5)

(1,1,3,4,6) (1,1,3,5,5) (1,1,3,5,6) (1,1,3,6,6) (1,1,4,4,4) (1,1,4,4,5)
(1,1,4,4,6) (1,1,4,5,5) (1,1,4,5,6) (1,1,4,6,6) (1,1,5,5,5) (1,1,5,5,6)
(1,1,5,6,6) (1,1,6,6,6) (1,2,2,2,2) (1,2,2,2,3) (1,2,2,2,4) (1,2,2,2,5)
(1,2,2,2,6) (1,2,2,3,3) (1,2,2,3,4) (1,2,2,3,5) (1,2,2,3,6) (1,2,2,4,4)
(1,2,2,4,5) (1,2,2,4,6) (1,2,2,5,5) (1,2,2,5,6) (1,2,2,6,6) (1,2,3,3,3)
(1,2,3,3,4) (1,2,3,3,5) (1,2,3,3,6) (1,2,3,4,4) (1,2,3,4,5) (1,2,3,4,6)
(1,2,3,5,5) (1,2,3,5,6) (1,2,3,6,6) (1,2,4,4,4) (1,2,4,4,5) (1,2,4,4,6)
(1,2,4,5,5) (1,2,4,5,6) (1,2,4,6,6) (1,2,5,5,5) (1,2,5,5,6) (1,2,5,6,6)
(1,2,6,6,6) (1,3,3,3,3) (1,3,3,3,4) (1,3,3,3,5) (1,3,3,3,6) (1,3,3,4,4)
(1,3,3,4,5) (1,3,3,4,6) (1,3,3,5,5) (1,3,3,5,6) (1,3,3,6,6) (1,3,4,4,4)
(1,3,4,4,5) (1,3,4,4,6) (1,3,4,5,5) (1,3,4,5,6) (1,3,4,6,6) (1,3,5,5,5)
(1,3,5,5,6) (1,3,5,6,6) (1,3,6,6,6) (1,4,4,4,4) (1,4,4,4,5) (1,4,4,4,6)
(1,4,4,5,5) (1,4,4,5,6) (1,4,4,6,6) (1,4,5,5,5) (1,4,5,5,6) (1,4,5,6,6)
(1,4,6,6,6) (1,5,5,5,5) (1,5,5,5,6) (1,5,5,6,6) (1,5,6,6,6) (1,6,6,6,6)
(2,2,2,2,2) (2,2,2,2,3) (2,2,2,2,4) (2,2,2,2,5) (2,2,2,2,6) (2,2,2,3,3)
(2,2,2,3,4) (2,2,2,3,5) (2,2,2,3,6) (2,2,2,4,4) (2,2,2,4,5) (2,2,2,4,6)
(2,2,2,5,5) (2,2,2,5,6) (2,2,2,6,6) (2,2,3,3,3) (2,2,3,3,4) (2,2,3,3,5)
(2,2,3,3,6) (2,2,3,4,4) (2,2,3,4,5) (2,2,3,4,6) (2,2,3,5,5) (2,2,3,5,6)
(2,2,3,6,6) (2,2,4,4,4) (2,2,4,4,5) (2,2,4,4,6) (2,2,4,5,5) (2,2,4,5,6)
(2,2,4,6,6) (2,2,5,5,5) (2,2,5,5,6) (2,2,5,6,6) (2,2,6,6,6) (2,3,3,3,3)
(2,3,3,3,4) (2,3,3,3,5) (2,3,3,3,6) (2,3,3,4,4) (2,3,3,4,5) (2,3,3,4,6)
(2,3,3,5,5) (2,3,3,5,6) (2,3,3,6,6) (2,3,4,4,4) (2,3,4,4,5) (2,3,4,4,6)
(2,3,4,5,5) (2,3,4,5,6) (2,3,4,6,6) (2,3,5,5,5) (2,3,5,5,6) (2,3,5,6,6)
(2,3,6,6,6) (2,4,4,4,4) (2,4,4,4,5) (2,4,4,4,6) (2,4,4,5,5) (2,4,4,5,6)
(2,4,4,6,6) (2,4,5,5,5) (2,4,5,5,6) (2,4,5,6,6) (2,4,6,6,6) (2,5,5,5,5)
(2,5,5,5,6) (2,5,5,6,6) (2,5,6,6,6) (2,6,6,6,6) (3,3,3,3,3) (3,3,3,3,4)
(3,3,3,3,5) (3,3,3,3,6) (3,3,3,4,4) (3,3,3,4,5) (3,3,3,4,6) (3,3,3,5,5)
(3,3,3,5,6) (3,3,3,6,6) (3,3,4,4,4) (3,3,4,4,5) (3,3,4,4,6) (3,3,4,5,5)
7.1. COUNTING 299
(3,3,4,5,6) (3,3,4,6,6) (3,3,5,5,5) (3,3,5,5,6) (3,3,5,6,6) (3,3,6,6,6)

(3,4,4,4,4) (3,4,4,4,5) (3,4,4,4,6) (3,4,4,5,5) (3,4,4,5,6) (3,4,4,6,6)
(3,4,5,5,5) (3,4,5,5,6) (3,4,5,6,6) (3,4,6,6,6) (3,5,5,5,5) (3,5,5,5,6)
(3,5,5,6,6) (3,5,6,6,6) (3,6,6,6,6) (4,4,4,4,4) (4,4,4,4,5) (4,4,4,4,6)
(4,4,4,5,5) (4,4,4,5,6) (4,4,4,6,6) (4,4,5,5,5) (4,4,5,5,6) (4,4,5,6,6)
(4,4,6,6,6) (4,5,5,5,5) (4,5,5,5,6) (4,5,5,6,6) (4,5,6,6,6) (4,6,6,6,6)
(5,5,5,5,5) (5,5,5,5,6) (5,5,5,6,6) (5,5,6,6,6) (5,6,6,6,6) (6,6,6,6,6)
Whew . . . err, I mean, Yahtzee!

You can describe a generic element of the above listing by saying “It
starts with some number of 1’s (which may be zero), then there are some 2’s
(again, it might be that there are zero 2’s), then some (possibly none) 3’s,
then some 4’s (or maybe not), then some 5’s (I think you probably get the
idea) and finally some 6’s (sorry for all the parenthetical remarks).”
We could, of course, actually count the outcomes as listed above (there are
252) but that would be pretty dull – and it wouldn’t get us any closer to solv-
ing such problems in general. To count things like Yahtzee rolls it will turn
out that we can count something related but much simpler – blank-comma
arrangements. For the Yahtzee problem we count arrangements of 5 blanks
and 5 commas. That is, things like , ,, , , and ,,,,,
and ,,, ,,. These arrangements of blanks and commas corre-
spond uniquely to Yahtzee rolls – the commas serve to separate different
numerical values and the blanks are where we would write-in the 5 outcomes
on the die.
Convince yourself that there really is a one-to-one correspondence be-
tween Yahtzee outcomes and arrangements of 5 blanks and 5 commas by
doing the following
Exercise. What Yahtzee rolls correspond to the following blank-comma ar-

rangements?
, , , , , , ,,,, ,,,,,
What blank-comma arrangements correspond to the following Yahtzee out-
comes?
{2, 3, 4, 5, 6} {3, 3, 3, 3, 4} {5, 5, 6, 6, 6}
It may seem at first that this blank-comma thing is okay, but that we’re
still no closer to answering the question we started with. It may seem that
way until you realize how easy it is to count these blank-comma arrange-
ments! You see, there are 10 symbols in one of these blank-comma arrange-
ments and if we choose positions for (say) the commas, the blanks will have to
go into the other positions – thus every 5-subset of {1, 2, 3, 4, 5, 6, 7, 8, 9, 10}
gives us a blank-comma arrangement and every one of them gives us a Yahtzee
outcome. That is why there are C(10, 5) = 252 outcomes listed in the giant
tabulation above.
In general, when we are selecting k things from a set of size n (with repeti-
tion and without order) we will need to consider blank-comma arrangements
having k blanks and n−1 commas. As an aid to memory, consider that when
you actually write-out the elements of a set it takes one fewer commas than
there are elements – for example {1, 2, 3, 4} has 4 elements but we only need
3 commas to separate them. The general answer to our problem is either
C(k + n − 1, k) or C(k + n − 1, n − 1), depending on whether you want to
think about selecting positions for the k blanks or for the n − 1 commas.
It turns out that these binomial coefficients are equal so there’s no problem
with the apparent ambiguity.
So, finally, our table of counting formulas is complete. We’ll produce it
here one more time and, while we’re at it, ditch the C(n, k) notation in favor
of the more usual “binomial coefficient” notation nk .

7.1. COUNTING 301
Does order matter?
Yes No
No
Are repeats okay?
n! n n!

Yes P (n, k) = (n−k)! k
= k!(n−k)!
n+k−1

nk k
Exercises — 7.1
1. Determine the number of entries in the following sequences.
(a) (999, 1000, 1001, . . . 2006)
(b) (13, 15, 17, . . . 199)
(c) (13, 19, 25, . . . 601)
(d) (5, 10, 17, 26, 37, . . . 122)
(e) (27, 64, 125, 216, . . . 8000)
(f) (7, 11, 19, 35, 67, . . . 131075)
2. How many “full houses” are there in Yahtzee? (A full house is a pair
together with a three-of-a-kind.)
3. In how many ways can you get “two pairs” in Yahtzee?

n+k−1 n+k−1
4. Prove that the binomial coefficients and are
k n−1
equal.
5. The “Cryptographer’s alphabet” is used to supply small examples in

coding and cryptography. It consists of the first 6 letters, {a, b, c, d, e, f }.
How many “words” of length up to 6 can be made with this alphabet?
(A word need not actually be a word in English, for example both “fed”
and “dfe” would be words in the sense we are using the term.)
6. How many “words” are there of length 4, with distinct letters from the
Cryptographer’s alphabet, in which the letters appear in increasing
order alphabetically? (“Acef” would be one such word, but “cafe”
would not.)
7.1. COUNTING 303
7. How many “words” are there of length 4 from the Cryptographer’s

alphabet, with repeated letters allowed, in which the letters appear in
non-decreasing order alphabetically?
8. How many subsets does a finite set have?
9. How many handshakes will transpire when n people first meet?
10. How many functions are there from a set of size n to a set of size m?
11. How many relations are there from a set of size n to a set of size m?
7.2 Parity and Counting arguments
This section is concerned with two very powerful elements of the proof-
making arsenal: “Parity” is a way of referring to the result of an even/odd
calculation; Counting arguments most often take the form of counting some
collection in two different ways – and then comparing those results. These
techniques have little to do with one another, but when they are applicable
they tend to produce really elegant little arguments.
In (very) early computers and business machines, paper cards were used
to store information. A so-called “punch card” or “Hollerith card” was used
to store binary information by means of holes punched into it. Paper tape
was also used in a similar fashion. A typical paper tape format would involve
8 positions in rows across the tape that might or might not be punched,
often a column of smaller holes would appear as well which did not store
information but were used to drive the tape through the reading mechanism
on a sprocket. Tapes and cards could be “read” either by small sets of
electrical contacts which would touch through a punched hole or be kept
separate if the position wasn’t punched, or by using a photo-detector to sense
whether light could pass through the hole or not. The mechanisms for reading
and writing on these paper media were amazingly accurate, and allowed early
data processing machines to use just a couple of large file cabinets to store
what now fits in a jump drive one can wear on a necklace. (About 10 or 12
cabinets could hold a gigabyte of data).
Paper media was ideally suited to storing binary information, but of
course most of the real data people needed to store and process would be
alphanumeric5 . There were several encoding schemes that served to trans-
late between the character sets that people commonly used and the binary
5
“Alphanumeric” is a somewhat antiquated term that refers to information containing
both alphabetic characters and numeric characters – as well as punctuation marks, etc.
7.2. PARITY AND COUNTING ARGUMENTS 305
numerals that could be stored on paper. One of these schemes still survives
today – ASCII. The American Standard Code for Information Interchange
uses 7-bit binary numerals to represent characters, so it contains 128 different
symbols. This is more than enough to represent both upper- and lower-case
letters, the 10 numerals, and the punctuation marks – many of the remaining
spots in the ASCII code were used to contain so-called “control characters”
that were associated with functionality that appeared on old-fashioned tele-
type equipment – things like “ring the bell,” “move the carriage backwards
one space,” “move the carriage to the next line,” etc. These control charac-
ters are why modern keyboards still have a modifier key labelled “Ctrl” on
them. The following listing gives the decimal and binary numerals from 0 to
127 and the ASCII characters associated with them – the non-printing and
control characters have a 2 or 3 letter mnemonic designation.
0 0000 0000 NUL 64 0100 0000 @
1 0000 0001 SOH 65 0100 0001 A
2 0000 0010 STX 66 0100 0010 B
3 0000 0011 ETX 67 0100 0011 C
4 0000 0100 EOT 68 0100 0100 D
5 0000 0101 ENQ 69 0100 0101 E
6 0000 0110 ACK 70 0100 0110 F
7 0000 0111 BEL 71 0100 0111 G
8 0000 1000 BS 72 0100 1000 H
9 0000 1001 TAB 73 0100 1001 I
10 0000 1010 LF 74 0100 1010 J
11 0000 1011 VT 75 0100 1011 K
12 0000 1100 FF 76 0100 1100 L
13 0000 1101 CR 77 0100 1101 M
14 0000 1110 SO 78 0100 1110 N
15 0000 1111 SI 79 0100 1111 O
16 0001 0000 DLE 80 0101 0000 P
17 0001 0001 DC1 81 0101 0001 Q
18 0001 0010 DC2 82 0101 0010 R
19 0001 0011 DC3 83 0101 0011 S
20 0001 0100 DC4 84 0101 0100 T
21 0001 0101 NAK 85 0101 0101 U
22 0001 0110 SYN 86 0101 0110 V
23 0001 0111 ETB 87 0101 0111 W

24 0001 1000 CAN 88 0101 1000 X
25 0001 1001 EM 89 0101 1001 Y
26 0001 1010 SUB 90 0101 1010 Z
27 0001 1011 ESC 91 0101 1011 [
28 0001 1100 FS 92 0101 1100 \
29 0001 1101 GS 93 0101 1101 ]
30 0001 1110 RS 94 0101 1110 ^
31 0001 1111 US 95 0101 1111 _
32 0010 0000 96 0110 0000 ‘
33 0010 0001 ! 97 0110 0001 a
34 0010 0010 " 98 0110 0010 b
35 0010 0011 # 99 0110 0011 c
36 0010 0100 $ 100 0110 0100 d
37 0010 0101 % 101 0110 0101 e
38 0010 0110 & 102 0110 0110 f
39 0010 0111 ’ 103 0110 0111 g
40 0010 1000 ( 104 0110 1000 h
41 0010 1001 ) 105 0110 1001 i
42 0010 1010 * 106 0110 1010 j
43 0010 1011 + 107 0110 1011 k
44 0010 1100 , 108 0110 1100 l
45 0010 1101 - 109 0110 1101 m
46 0010 1110 . 110 0110 1110 n
47 0010 1111 / 111 0110 1111 o
48 0011 0000 0 112 0111 0000 p
49 0011 0001 1 113 0111 0001 q
50 0011 0010 2 114 0111 0010 r
51 0011 0011 3 115 0111 0011 s
52 0011 0100 4 116 0111 0100 t
53 0011 0101 5 117 0111 0101 u
54 0011 0110 6 118 0111 0110 v
55 0011 0111 7 119 0111 0111 w
56 0011 1000 8 120 0111 1000 x
57 0011 1001 9 121 0111 1001 y
58 0011 1010 : 122 0111 1010 z
59 0011 1011 ; 123 0111 1011 {
60 0011 1100 < 124 0111 1100 |
61 0011 1101 = 125 0111 1101 }
62 0011 1110 > 126 0111 1110 ~
63 0011 1111 ? 127 0111 1111 DEL
Now it only takes 7 bits to encode the 128 possible values in the ASCII
system, which can easily be verified by noticing that the left-most bits in all
of the binary representations above are 0. Most computers use 8 bit words
or “bytes” as their basic units of information, and the fact that the ASCII
code only requires 7 bits lead someone to think up a use for that additional
bit. It became a “parity check bit.” If the seven bits of an ASCII encoding
have an odd number of 1’s, the parity check bit is set to 1 — otherwise, it is
set to 0. The result of this is that, subsequently, all of the 8 bit words that
encode ASCII data will have an even number of 1’s. This is an example of a
so-called error detecting code known as the “even code” or the “parity check
code.” If data is sent over a noisy telecommunications channel, or is stored
in failable computer memory, there is some small but calculable probability
that there will be a “bit error.” For instance, one computer might send
10000111 (which is the ASCII code that says “ring the bell”) but another
machine across the network might receive 10100111 (the 3rd bit from the left
has been received in error) now if we are only looking at the rightmost seven
bits we will think that the ASCII code for a single quote has been received,
but if we note that this piece of received data has an odd number of ones
we’ll realize that something is amiss. There are other more advanced coding
schemes that will let us not only detect an error, but (within limits) correct
it as well! This rather amazing feat is what makes wireless telephony (not
to mention communications with deep space probes — whoops! I mentioned
it) work.
The concept of parity can be used in many settings to prove some fairly
remarkable results.
In Section 6.3 we introduced the idea of a graph. This notion was first
used by Leonhard Euler to solve a recreational math problem posed by the
citizens of Königsberg, Prussia (this is the city now known as Kaliningrad,
Russia.) Königsberg was situated at a place where two branches of the
Pregel river6 come together – there is also a large island situated near this
confluence. By Euler’s time, the city of Königsberg covered this island as well
as the north and south banks of the river and also the promontory where the
branches came together. A network of seven bridges had been constructed
to connect all these land masses. The townsfolk are alleged to have become
enthralled by the question of whether it was possible to leave one’s home
and take a walk through town which crossed each of the bridges exactly once
and, finally, return to one’s home.
Figure 7.2: A simplified map of Königsberg, Prussia circa 1736.
Euler settled the question (it can’t be done) be converting the map of
Königsberg into a graph and then making some simple observations about
the parities of the nodes in this graph. The degree of a node in a graph is the
number of edges that are incident with it (if a so-called “loop edge” is present
it adds two to the node’s degree). The “parity of a node” is shorthand for
6
Today, this river is known as the Pregolya.
the “parity of the degree of the node.”
Figure 7.3: Euler’s solution of the “seven bridges of Königsberg problem”

involved representing the town as an undirected graph.
The graph of Königsberg has 4 nodes: one of degree 5 and three of degree
3. All the nodes are odd. Would it be possible to either modify Königsberg
or come up with an entirely new graph having some even nodes? Well,
the answer to that is easy – just tear down one of the bridges, and two of
the nodes will have their degree changed by one; they’ll both become even.
Notice that, by removing one edge, we effected the parity of two nodes. Is
it possible to create a graph with four nodes in which just one of them is
even? More generally, given any short list of natural numbers, is it possible
to draw a graph whose degrees are the listed numbers?
Exercise. Try drawing graphs having the following lists of vertex degrees.
(In some cases it will be impossible. . . )
- {1, 1, 2, 3, 3}
- {1, 2, 3, 5}
- {1, 2, 3, 4}
- {4, 4, 4, 4, 5}
- {3, 3, 3, 3}
- {3, 3, 3, 3, 3}
When it is possible to create a graph with a specified list of vertex degrees,

it is usually easy to do. Of course, when it’s impossible you struggle a bit. . .
To help get things rolling (just in case you haven’t really done the exercise)
I’ll give a hint – for the first list it is possible to draw a graph, for the second
it is not. Can you distinguish the pattern? What makes one list of vertex
degrees reasonable and another not?
Exercise. (If you didn’t do the last exercise, stop being such a lame-o and
try it now. BTW, if you did do it, good for you! You can either join with
me now in sneering at all those people who are scurrying back to do the last
one, or try the following:)
Figure out a way to distinguish a sequence of numbers that can be the
degree sequence of some graph from the sequences that cannot be.
Okay, now if you’re reading this sentence you should know that every
other list of vertex degrees above is impossible, you should have graphs drawn
in the margin here for the 1st, 3rd and 5th degree sequences, and you may
have discovered some version of the following
Theorem 7.2.1. In an undirected graph, the number of vertices having an

odd degree is even.
A slightly pithier statement is: All graphs have an even number of odd
nodes.
We’ll leave the proof of this theorem to the exercises but most of the work
is done in proving the following equivalent result.
Theorem 7.2.2. In an undirected graph the sum of the degrees of the vertices
is even.
Proof: The sum of the degrees of all the vertices in a graph G,
X
deg(v),
v∈V (G)
counts every edge of G exactly twice.
Thus,
X
deg(v) = 2 · |E(G)|.
v∈V (G)
In particular we see that this sum is even.
Q.E.D.
The question of whether a graph having a given list of vertex degrees can
exist comes down to an elegant little argument using both of the techniques
in the title of this section. We count the edge set of the graph in two ways –
once in the usual fashion and once by summing the vertex degrees; we also
note that since this latter count is actually a double count we can bring in
the concept of parity.
Another perfectly lovely argument involving parity arises in questions
concerning whether or not it is possible to tile a pruned chessboard with
dominoes. We’ve seen dominoes before in Section 5.1 and we’re just hoping
you’ve run across chessboards before. Usually a chessboard is 8 × 8, but we
would like to adopt a more liberal interpretation that a chessboard can be
any rectangular grid of squares we might choose.7 Suppose that we have a

supply of dominoes that are of just the right size that if they are laid on a
chessboard they perfectly cover two adjacent squares. Our first question is
quite simple. Is it possible to perfectly tile an m × n chessboard with such
dominoes?
First let’s specify the question a bit more. By “perfectly tiling” a chess-
board we mean that every domino lies (fully) on the board, covering precisely
two squares, and that every square of the board is covered by a domino.
The answer is straightforward. If at least one of m or n is even it can
be done. A necessary condition is that the number of squares be even (since
every domino covers two squares) and so, if both m and n are odd we will
be out of luck.
A “pruned board” is obtained by either literally removing some of the
squares or perhaps by marking them as being off limits in some way. When
we ask questions about perfect tilings of pruned chessboards things get more
interesting and the notion of parity can be used in several ways.
Here are two tiling problems regarding square chessboards:
1. An even-sided square board (e.g. an ordinary 8 × 8 board) with diag-

onally opposite corners pruned.
2. An odd-sided board with one square pruned.
Both of these situations satisfy the basic necessary condition that the
number of squares on the board must be even. You may be able to deter-
mine another “parity” approach to these tiling problems by attempting the
following
7
The game known as “draughts” in the UK and “checkers” in the US is played on an
8 × 8 board, but (for example) international draughts is played on a 10 × 10 board and
Canadian checkers is played on a 12 × 12 board.
Exercise. Below are two five-by-five chessboards each having a single square
pruned. One can be tiled by dominoes and the other cannot. Which is which?
The pattern of black and white squares on a chessboard is an example of

a sort of artificial parity, if we number the squares of the board appropriately
then the odd squares will be white and the even squares will be black. We
are used to chessboards having this alternating black/white pattern on them,
but nothing about these tiling problems required that structure8 If we were
used to monochromatic chessboards, we might never solve the previous two
problems – unless of course we invented the coloring scheme in order to solve
them. An odd-by-odd chessboard has more squares of one color than of the
other. An odd-by-odd chessboard needs to have a square pruned in order
for it to be possible for it to be tiled by dominoes – but if the wrong colored
square is pruned it will still be impossible. Each domino covers two squares
– one of each color! (So the pruned board must have the same number of
white squares as black.)
We’ll close this section with another example of the technique of counting
in two ways.
8
Nothing about chess requires this structure either, but it does let us do some error
checking. For instance, bishops always end up on the same color they left from and knights
always switch colors as they move.
A magic square of order n is a square n × n array containing the numbers

1, 2, 3, . . . , n2 . The numbers must be arranged in such a way that every row
and every column sum to the same number – this value is known as the magic
sum.
For example, the following is an order 3 magic square.
1 6 8
5 7 3
9 2 4
The definition of a magic square requires that the rows and columns sum
to the same number but says nothing about what that number must be.
It is conceivable that we could produce magic squares (of the same order)
having different magic sums. This is conceivable, but in fact the magic sum
is determined completely by n.
n3 + n
Theorem 7.2.3. A magic square of order n has a magic sum equal to .
2
Proof: We count the total of the entries in the magic square in

two ways. The sum of all the entries in the magic square is
S = 1 + 2 + 3 + . . . + n2 .
Pk
Using the formula for the sum of the first k naturals ( i=1 i=
k2 +k
2
) and evaluating at n2 gives
n4 + n2
S= .
2
On the other hand, if the magic sum is M , then each of the n

rows has numbers in it which sum to M so
S = nM.
By equating these different expressions for S and solving for M ,

we prove the desired result:
n4 + n2
nM = ,
2
therefore
n3 + n
M= .
2
Q.E.D.
Exercises — 7.2
1. A walking tour of Königsberg such as is described in this section, or

more generally, a circuit through an arbitrary graph that crosses each
edge precisely once and begins and ends at the same node is known
as an Eulerian circuit. An Eulerian path also crosses every edge of a
graph exactly once but it begins and ends at distinct nodes. For each
of the following graphs determine whether an Eulerian circuit or path
is possible, and if so, draw it.
2. Complete the proof of the fact that “Every graph has an even number
of odd nodes.”
3. Provide an argument as to why an 8 × 8 chessboard with two squares

pruned from diagonally opposite corners cannot be tiled with dominoes.
4. Prove that, if n is odd, any n × n chessboard with a square the same

color as one of its corners pruned can be tiled by dominoes.
5. The five tetrominoes (familiar to players of the video game Tetris) are
relatives of dominoes made up of four small squares.
All together these five tetrominoes contain 20 squares so it is conceiv-

able that they could be used to tile a 4 × 5 chessboard. Prove that this
is actually impossible.
6. State necessary and sufficient conditions for the existance of an Eulerian

circuit in a graph.
7. State necessary and sufficient conditions for the existance of an Eulerian

path in a graph.
8. Construct magic squares of order 4 and 5.
9. A magic hexagon of order 2 would consist of filling-in the numbers from

1 to 7 in the hexagonal array below. The magic condition means that
each of the 9 “lines” of adjacent hexagons would have the same sum.
Is this possible?
10. Is there a magic hexagon of order 3?

7.3. THE PIGEONHOLE PRINCIPLE 319
7.3 The pigeonhole principle

The word “pigeonhole” can refer to a hole in which a pigeon roosts (i.e.
pretty much what it sounds like) or a series of roughly square recesses in a
desk in which one could sort correspondence (see Figure 7.4).
Whether you prefer to think of roosting birds or letters being sorted, the
first and easiest version of the pigeonhole principle is that if you have more
“things” than you have “containers” there must be a container holding at
least two things.
If we have 6 pigeons who are trying to roost in a coop with 5 pigeonholes,
two birds will have to share.
If we have 7 letters to sort and there are 6 pigeonholes in our desk, we
will have to put two letters in the same compartment.
The “things” and the “containers” don’t necessarily have to be interpreted
in the strict sense that the “things” go into the “containers.” For instance,
a nice application of the pigeonhole principle is that if there are at least 13
people present in a room, some pair of people will have been born in the
same month. In this example the things are the people and the containers
are the months of the year.
The abstract way to phrase the pigeonhole principle is:
Theorem 7.3.1. If f is a function such that |Dom(f )| > |Rng(f )| then f is

not injective.
The proof of this statement is an easy example of proof by contradiction

so we’ll include it here.
Proof: Suppose to the contrary that f is a function with |Dom(f )| >

|Rng(f )| and that f is injective. Of course f is onto its range, so
since we are presuming that f is injective it follows that f is a
pigeonholes
Figure 7.4: Pigeonholes in an old-fashioned roll top desk could be used to

sort letters.
bijection between Dom(f ) and Rng(f ). Therefore (since f pro-

vides a one-to-one correspondence) |Dom(f )| = |Rng(f )|. This
clearly contradicts the statement that |Dom(f )| > |Rng(f )|.
Q.E.D.
For a statement with an almost trivial proof the pigeonhole principle is

very powerful. We can use it to prove a host of existential results – some are
fairly silly, others very deep. Here are a few examples:
There are two people (who are not bald) in New York City having exactly
the same number of hairs on their heads.
There are two books in (insert your favorite library) that have the same
number of pages.
Given n married couples (so 2n people) if we choose n + 1 people we will
be forced to choose both members of some couple.
Suppose we select n + 1 numbers from the set {1, 2, 3, . . . , 2n}, we will be
forced to have chosen two numbers such that one is divisible by the other.
We can come up with stronger forms of the pigeonhole principle by con-

sidering pigeonholes with capacities. Suppose we have 6 pigeonholes in a
desk, each of which can hold 10 letters. What number of letters will guar-
antee that one of the pigeonholes is full? The largest number of letters we
could have without having 10 in some pigeonhole is 9 · 6 = 54, so if there are
55 letters we must have 10 letters in some pigeonhole.
More generally, if we have n containers, each capable of holding m objects,
than if there are n · (m − 1) + 1 objects placed in the containers, we will be
assured that one of the containers is at capacity.
The ordinary pigeonhole principle is the special case m = 2 of this
stronger version.
There is an even stronger version, which ordinarily is known as the “strong

form of the pigeonhole principle.” In the strong form, we have pigeonholes
with an assortment of capacities.
Theorem 7.3.2. If there are n containers having capacities m1 , m2 , m3 , . . . , mn ,

and there are 1 + ni=1 (mi − 1) objects placed in them, then for some i, con-
P
tainer i has (at least) mi objects in it.
Proof: If no container holds its full capacity, then the largest

the total of the objects could be is ni=1 (mi − 1).
P
Q.E.D.
Exercises — 7.3
1. The statement that there are two non-bald New Yorkers with the same
number of hairs on their heads requires some careful estimates to justify
it. Please justify it.
2. A mathematician, who always rises earlier than her spouse, has de-
veloped a scheme – using the pigeonhole principle – to ensure that
she always has a matching pair of socks. She keeps only blue socks,
green socks and black socks in her sock drawer – 10 of each. So as
not to wake her husband she must select some number of socks from
her drawer in the early morning dark and take them with her to the
adjacent bathroom where she dresses. What number of socks does she
choose?
3. If we select 1001 numbers from the set {1, 2, 3, . . . , 2000} it is certain

that there will be two numbers selected such that one divides the other.
We can prove this fact by noting that every number in the given set
can be expressed in the form 2k · m where m is an odd number and
using the pigeonhole principle. Write-up this proof.
4. Given any set of 53 integers, show that there are two of them having
the property that either their sum or their difference is evenly divisible
by 103.
5. Prove that if 10 points are placed inside a square of side length 3, there
√
will be 2 points within 2 of one another.
6. Prove that if 10 points are placed inside an equilateral triangle of side

length 3, there will be 2 points within 1 of one another.
7. Prove that in a simple graph (an undirected graph with no loops or

parallel edges) having n nodes, there must be two nodes having the
same degree.
7.4. THE ALGEBRA OF COMBINATIONS 325
7.4 The algebra of combinations

Earlier in this chapter we determined the number of k-subsets of a set of size
n. These numbers, denoted by C(n, k) = nCk = nk and determined by the

n!
formula k!(n−k)!
are known as binomial coefficients. It seems likely that you
will have already seen the arrangement of these binomial coefficients into a
triangular array – known as Pascal’s triangle, but if not. . .
1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
1 6 15 20 15 6 1
et cetera.
The thing that makes this triangle so nice and that leads to the strange
name “binomial coefficients” for the number of k-combinations of a n-sets is
that you can use the triangle to (very quickly) compute powers of binomials.
A binomial is a polynomial with two terms. Things like (x + y), (x + 1)
and (x7 + x3 ) all count as binomials but to keep things simple just think
about (x + y). If you need to compute a large power of (x + y) you can just
multiply it out, for example, think of finding the 6th power of (x + y).
We can use the F.O.I.L rule to find (x + y)2 = x2 + 2xy + y 2 . Then
(x + y)3 = (x + y) · (x + y)2 = (x + y) · (x2 + 2xy + y 2 ).
You can compute that last product either by using the distributive law
or the table method:
x2 +2xy +y 2
x
+y
Either way, the answer should be (x + y)3 = x3 + 3x2 y + 3xy 2 + y 3 .

Finally the sixth power is the square of the cube thus
(x + y)6 = (x + y)3 · (x + y)3

= (x3 + 3x2 y + 3xy 2 + y 3 ) · (x3 + 3x2 y + 3xy 2 + y 3 )
For this product I wouldn’t even think about the distributive law, I’d
jump to the table method right away:
x3 +3x2 y +3xy 2 +y 3
x3
+3x2 y
+3xy 2
+y 3
In the end you should obtain
x6 + 6x5 y + 15x4 y 2 + 20x3 y 3 + 15x2 y 4 + 6xy 5 + y 6 .
Now all of this is a lot of work and it’s really much easier to notice the
form of the answer: The exponent on x starts at 6 and descends with each
successive term down to 0. The exponent on y starts at 0 and ascends to 6.
The coefficients in the answer are the numbers in the sixth row of Pascal’s
triangle.
Finally, the form of Pascal’s triangle makes it really easy to extend. A
number in the interior of the triangle is always the sum of the two above
it (on either side). Numbers that aren’t in the interior of the triangle are
always 1.
We showed rows 0 through 6 above. Rows 7 and 8 are
1 7 21 35 35 21 7 1
1 8 28 56 70 56 28 8 1.
With this information in hand, it becomes nothing more than a matter

of copying down the answer to compute
(x + y)8 = x8 + 8x7 y + 28x6 y 2 + 56x5 y 3 + 70x4 y 4 + 56x3 y 5 + 28x2 y 6 + 8xy 7 + y 8 .
Exercise. Given the method using Pascal’s triangle for computing (x + y)n
we can use substitution to determine more general binomial powers.
Find (x4 + x2 )5 .
All of the above hinges on the fact that one can compute a binomial
coefficient by summing the two that appear to either side and above it in
Pascal’s triangle. This fact is the fundamental relationship between binomial
coefficients – it is usually called Pascal’s formula.
Theorem 7.4.1. For all natural numbers n and k with 0 < k ≤ n,

n n−1 n−1
= + .
k k k−1
We are going to prove it twice.
Proof: (The first proof is a combinatorial argument.)

There are nk subsets of size k of the set N = {1, 2, 3, . . . , n}. We

will partition these k-subsets into two disjoint cases: those that
contain the final number, n, and those that do not.
Let
A = {S ⊆ N |S| = k ∧ n ∈
/ S}
and, let
B = {S ⊆ N |S| = k ∧ n ∈ S}.
Since the number n is either in a k-subset or it isn’t, these sets

are disjoint and exhaustive. So the addition rule tells us that

n
= |A| + |B|.
k
The set A is really just the set of all k-subsets of the (n − 1)-set
{1, 2, 3, . . . , n − 1}, so |A| = n−1

k
.
Any of the sets in B can be obtained by adjoining the element n to

a k −1 subset of the (n−1)-set {1, 2, 3, . . . , n−1}, so |B| = n−1

k−1
.
Substituting gives us the desired result.

Q.E.D.
Proof: (The second proof is algebraic in nature.)
Consider the sum

n−1 n−1
+ .
k k−1
Applying the formula we deduced in Section 7.1 we get

n−1 n−1
+
k k−1
(n − 1)! (n − 1)!
= +
k!(n − 1 − k)! (k − 1)!((n − 1) − (k − 1))!
(n − 1)! (n − 1)!
= +
k!(n − k − 1)! (k − 1)!(n − k)!
A common denomenator for these fractions is k!(n − k)!. (We

will have to multiply the top and bottom of the first fraction by
(n − k) and the top and bottom of the second fraction by k.)
(n − k)(n − 1)! k(n − 1)!

= +
k!(n − k)(n − k − 1)! k(k − 1)!(n − k)!
(n − k)(n − 1)! k(n − 1)!

= +
k!(n − k)! k!(n − k)!
(n − k)(n − 1)! + k(n − 1)!

=
k!(n − k)!
(n − k + k)(n − 1)!
=
k!(n − k)!
(n)(n − 1)!
=
k!(n − k)!
n!
= .
k!(n − k)!
n

We recognize the final expression as the definition of k
, so we
have proved that

n−1 n−1 n
+ = .
k k−1 k
Q.E.D.
There are quite a few other identities concerning binomial coefficients

that can also be proved in (at least) two ways. We will provide one or two
other examples and leave the rest to you in the exercises for this section.

n n−1
k· =n· .
k k−1
Let’s try a purely algebraic approach first.
Proof:
Using the formula for the value of a binomial coefficient we get

n n!
k· =k· .
k k!(n − k)!
We can do some cancellation to obtain

n n!
k· = .
k (k − 1)!(n − k)!
Finally we factor-out an n to obtain

n (n − 1)!
k· =n· ,
k (k − 1)!(n − k)!
since (n − k) is the same thing as ((n − 1) − (k − 1)) we have

n (n − 1)! n−1
k· =n· =n·
k (k − 1)!((n − 1) − (k − 1))! k−1
Q.E.D.
A combinatorial argument usually involves counting something in two

ways. What could that something be? Well, if you see a product in some
formula you should try to imagine what the multiplication rule would say in
that particular circumstance.
Proof: Consider the collection of all subsets of size k taken from

N = {1, 2, 3, . . . , n} in which one of the elements has been marked
to distinguish it from the others in some way.9
We can count this collection in two ways using the multiplication

rule.
n

Firstly, we could select a k-subset in k
ways and then from
among the k elements of the subset we could select one to be
marked. By this analysis there are nk · k elements in our collec-

tion.
Secondly, we could select an element from the n-set which will be

the “marked” element of our subset, and then choose the addi-
tional k − 1 elements from the remaining n − 1 elements of the
n-set. By this analysis there are n· n−1

k−1
elements in the collection
we have been discussing.
Thus,

n n−1
k· =n·
k k−1
Q.E.D.
9
For example, a committee of k individuals one of whom has been chosen as chairper-
son, is an example of the kind of entity we are discussing.
The final result that we’ll talk about actually has (at least) three proofs.
One of which suffers from the fault that it is “like swatting a fly with a sledge
hammer.”
The result concerns the sum of all the numbers in some row of Pascal’s
triangle.
n
X n
= 2n
k=0
k
Our sledge hammer is a powerful result known as the binomial theorem

which is a formalized statement of the material we began this section with.
Theorem 7.4.4 (The Binomial Theorem). For all natural numbers n, and
real numbers x and y,
n
n
X n n−k k
(x + y) = x y .
k=0
k
We won’t be proving this result just now. But, the following proof is a
proof of the previous theorem using this more powerful result.
Proof: Substitute x = y = 1 in the binomial theorem.

Q.E.D.
Our second proof will be combinatorial. Let us re-iterate that a combi-

natorial proof usually consists of counting some collection in two different
ways. The formula we have in this example contains a sum, so we should
search for a collection of things that can be counted using the addition rule.
Proof: The set of all subsets of N = {1, 2, 3, . . . , n}, which we

denote by P(N ), can be partitioned into n + 1 sets based on the
sizes of the subsets,
P(N ) = S0 ∪ S1 ∪ S2 ∪ . . . ∪ Sn ,
where Sk = {S S ⊆ N ∧ |S| = k} for 0 ≤ k ≤ n. Since no subset

of N can appear in two different parts of the partition (a subset’s
size is unique) and every subset of N appears in one of the parts
of the partition (the sizes of subsets are all in the range from 0
to n). The addition principle tells us that
|P(N )| = |S0 | + |S1 | + ∪|S2 | + . . . + |Sn |.
We have previously proved that |P(N )| = 2n and we know that

|Sk | = nk so it follows that

n n n n n
2 = + + + ... + .
0 1 2 n
Q.E.D.
Exercises — 7.4
1. Use the binomial theorem (with x = 1000 and y = 1) to calculate

10016 .
2. Find (2x + 3)5 .
3. Find (x2 + y 2 )6 .
4. The following diagram contains a 3-dimensional analog of Pascal’s tri-

angle that we might call “Pascal’s tetrahedron.” What would the next
layer look like?
1 1
1
2 1
1
2
2 1
3 1
3
1 3
6
3 3
3 1
5. The student government at Lagrange High consists of 24 members cho-

sen from amongst the general student body of 210. Additionally, there
is a steering committee of 5 members chosen from amongst those in
student government. Use the multiplication rule to determine two dif-
ferent formulas for the total number of possible governance structures.
6. Prove the identity

n k n n−r
· = ·
k r r k−r
combinatorially.
7. Prove the binomial theorem.
n
n
X n
∀n ∈ N, ∀x, y ∈ R, (x + y) = xn−k y k
k=0
k
Chapter 8
Cardinality
The very existence of flame-throwers proves that some time, somewhere,

someone said to themselves, “You know, I want to set those people over there
on fire, but I’m just not close enough to get the job done.” –George Carlin
8.1 Equivalent sets

We have seen several interesting examples of equivalence relations already,
and in this section we will explore one more: we’ll say two sets are equivalent
if they have the same number of elements. Usually, an equivalence relation
has the effect that it highlights one characteristic of the objects being studied,
while ignoring all the others. Equivalence of sets brings the issue of size (a.k.a.
cardinality) into sharp focus while, at the same time, it forgets all about the
many other features of sets. Sets that are equivalent (under the relation we
are discussing) are sometimes said to be equinumerous 1 .
A couple of examples may be in order.
1
Perversely, there are also those who use the term equipollent to indicate that sets are
the same size. This term actually applies to logical statements that are deducible from
one another.
337
338 CHAPTER 8. CARDINALITY
• If A = {1, 2, 3} and B = {a, b, c} then A and B are equivalent.
• Since the empty set is unique – ; is the only set having 0 elements – it
follows that there are no other sets equivalent to it.
• Every singleton set2 is equivalent to every other singleton set.
Hopefully these examples are relatively self-evident. Unfortunately, that

very self-evidence may tend to make you think that this notion of equivalence
isn’t all that interesting — nothing could be further from the truth! The
notion of equivalence of sets becomes really interesting when we study infinite
sets. Once we have the right definition in hand we will be able to prove
some truly amazing results. For instance, the sets N and Q turn out to be
equivalent. Since the naturals are wholly contained in the rationals this is
(to say the least) counter-intuitive! Coming up with the “right” definition
for this concept is crucial.
We could make the following:
Definition. (Well . . . not quite.) For all sets A and B, we say A and B
are equivalent, and write A ≡ B iff |A| = |B|.
The problem with this definition is that it is circular. We’re trying to

come up with an equivalence relation so that the equivalence classes will
represent the various cardinalities of sets (i.e. their sizes) and we define the
relation in terms of cardinalities. We won’t get anything new from this.
Georg Cantor was the first person to develop the modern notion of the
equivalence of sets. His early work used the notion implicitly, but when he
finally developed the concept of one-to-one correspondences in an explicit
way he was able to prove some amazing facts. The phrase “one-to-one cor-
respondence” has a fairly impressive ring to it, but one can discover what it
means by just thinking carefully about what it means to count something.
2
Recall that a singleton set is a set having just one element.
8.1. EQUIVALENT SETS 339
Consider the solmization syllables used for the notes of the major scale in
music; they form the set {do, re, mi, fa, so, la, ti}. What are we doing when
we count this set (and presumably come up with a total of 7 notes)? We first
point at ‘do’ while saying ‘one,’ then point at ‘re’ while saying ‘two,’ et cetera.
In a technical sense we are creating a one-to-one correspondence between the
set containing the seven syllable and the special set {1, 2, 3, 4, 5, 6, 7}. You
should notice that this one-to-one correspondence is by no means unique. For
instance we could have counted the syllables in reverse — a descending scale,
or in some funny order – a little melody using each note once. The fact that
there are seven syllables in the solmization of the major scale is equivalent
to saying that there exists a one-to-one correspondence between the syllables
and the special set {1, 2, 3, 4, 5, 6, 7}. Saying “there exists” in this situation
may seem a bit weak since in fact there are 7! = 5040 correspondences, but
“there exists” is what we really want here. What exactly is a one-to-one
correspondence? Well, we’ve actually seen such things before – a one-to-one
correspondence is really just a bijective function between two sets. We’re
finally ready to write a definition that Georg Cantor would approve of.
Definition. For all sets A and B, we say A and B are equivalent, and write
A ≡ B iff there exists a one-to-one (and onto) function f , with Dom(f ) = A
and Rng(f ) = B.
Somewhat more succinctly, one can just say the sets are equivalent iff
there is a bijection between them.
We are going to ask you to prove that the above definition defines an
equivalence relation in the exercises for this section. In order to give you a
bit of a jump start on that proof we’ll outline what the proof that the relation
is symmetric should look like.
To show that the relation is symmetric we must assume that A

and B are sets with A ≡ B and show that this implies that
B ≡ A. According to the definition above this means that we’ll

need to locate a function (that is one-to-one) from B to A. On
the other hand, since it is given that A ≡ B, the definition tells
us that there actually is an injective function, f , from A to B.
The inverse function f −1 would do exactly what we’d like (namely
form a map from B to A assuming that we can show that f −1
has the right properties. We need to know that f −1 is a function
(remember that in general the inverse of a function is only a
relation) and that it is one-to-one. That f −1 is a function is a
consequence of the fact that f is one-to-one. That f −1 is one-to-
one is a consequence of the fact that f is a function.
The above is just a sketch of a proof. In the exercise you’ll need to fill
in the rest of the details as well as provide similar arguments for reflexivity
and transitivity. For each possible finite cardinality k, there are many, many
sets having that cardinality, but there is one set that stands out as the most
basic – the set of numbers from 1 to k. For each cardinality k > 0, we use
the symbol Nk to indicate this set:
Nk = {1, 2, 3, . . . , k}.
The finite cardinalities are the equivalence classes (under the relation of
set equivalence) containing the empty set and the sets Nk . Of course there
are also infinite sets! The prototype for an infinite set would have to be
N. The long-standing tradition is to use the symbol ℵ0 3 for the cardinality
of sets having the same size as N, alternatively, such sets are known as
“countable.” One could make a pretty good argument that it is the finite sets
that are actually countable! After all it would literally take forever to count
3
The Hebrew letter (capital) aleph with a subscript zero – usually pronounced “aleph
naught.”
8.1. EQUIVALENT SETS 341
the natural numbers! We have to presume that the people who instituted
this terminology meant for “countable” to mean “countable, in principle”
or “countable if you’re willing to let me keep counting forever” or maybe
“countable if you can keep counting faster and faster and are capable of
ignoring the speed of light limitations on how fast your lips can move.” Worse
yet, the term “countable” has come to be used for sets whose cardinalities are
either finite or the size of the naturals. If we want to refer specifically to the
infinite sort of countable set most mathematicians use the term denumerable
(although this is not universal) or countably infinite. Finally, there are sets
whose cardinalities are bigger than the naturals. In other words, there are
sets such that no one-to-one correspondence with N is possible. We don’t
mean that people have looked for one-to-one correspondences between such
sets and N and haven’t been able to find them – we literally mean that it
can’t be done; and it is has been proved that it can’t be done! Sets having
cardinalities that are this ridiculously huge are known as uncountable.
Exercises — 8.1
1. Name four sets in the equivalence class of {1, 2, 3}.
2. Prove that set equivalence is an equivalence relation.
3. Construct a Venn diagram showing the relationships between the sets of

sets which are finite, infinite, countable, denumerable and uncountable.
4. Place the sets N, R, Q, Z, Z × Z, C, N2007 and ∅; somewhere on the

Venn diagram above. (Note to students (and graders): there are no
wrong answers to this question, the point is to see what your intuition
about these sets says at this point.)
8.2. EXAMPLES OF SET EQUIVALENCE 343
8.2 Examples of set equivalence
There is an ancient conundrum about what happens when an irresistible force

meets an immovable object. In a similar spirit there are sometimes heated
debates among young children concerning which super-hero will win a fight.
Can Wolverine take Batman? What about the Incredible Hulk versus the
Thing? Certainly Superman is at the top of the heap in this ordering. Or is
he? Would the man of steel even engage in a fight with a female super-hero,
say Wonder Woman? (Remember the 1950’s sensibilities of Clark Kent’s
alter ego.)
To many people the current topic will seem about as sensible as the school-
yard discussions just alluded to. We are concerned with knowing whether one
infinite set is bigger than another, or are they the same size. There are gen-
erally three reasons that people disdain to consider such questions. The first
is that, like super-heros, infinite sets are just products of the imagination.
The second is that there can be no difference because “infinite is infinite” –
once you get to the size we call infinity, you can’t add something to that to
get to a bigger infinity. The third is that the answers to questions like this
are not going to earn me big piles of money so “who cares?”
Point one is actually pretty valid. Physicists have determined that we
appear to inhabit a universe of finite scope, containing a finite number of
subatomic particles, so in reality there can be no infinite sets. Nevertheless,
the axioms we use to study many fields in mathematics guarantee that the
objects of consideration are indeed infinite in number. Infinity appears as
a concept even when we know it can’t appear in actuality. Point two, the
“there’s only one size of infinity” argument is wrong. We’ll see an informal
argument showing that there are at least two sizes of infinity, and a more
formal theorem that shows there is actually an infinite hierarchy of infinities
in Section 8.3
Point three, “who cares?” is in some sense the toughest of all to deal with.
Hopefully you’ll enjoy the clever arguments to come for their own intrinsic
beauty. But, if you can figure a way to make big piles of money using this
stuff that would be nice too.
Let’s get started.
Which set is bigger – the natural numbers, N or the set, Enoneg , of non-
negative even numbers? Both are clearly infinite, so the “infinity is infinity”
camp might be lead to the correct conclusion through invalid reasoning. On
the other hand, the even numbers are contained in the natural numbers so
there’s a pretty compelling case for saying the evens are somehow smaller
than the naturals. The mathematically rigorous way to show that these
sets have the same cardinality is by displaying a one-to-one correspondence.
Given an even number how can we produce a natural to pair it with? And,
given a natural how can we produce an even number to pair with it? The
map f : N −→ Enoneg defined by f (x) = 2x is clearly a function, and just
about as clearly, injective4 . Is the map f also a surjection? In other words, is
every non-negative even number the image of some natural under f ? Given
some non-negative even number e we need to be able to come up with an
x such that f (x) = e. Well, since e is an even number, by the definition of
“even” we know that there is an integer k such that e = 2k and since e is
either zero or positive it follows that k must also be either 0 or positive. It
turns out that k is actually the x we are searching for. Put more succinctly,
every non-negative even number 2k has a preimage, k, under the map f . So
f maps N surjectively onto Enoneg . Now the sets we’ve just considered,
N = {0, 1, 2, 3, 4, 5, 6, . . .}
and
4
If x and y are different numbers that map to the same value, then f(x) = f(y) so 2x =
2y. But we can cancel the 2’s and derive that x = y, which is a contradiction.
Enoneg = {0, 2, 4, 6, 8, 10, 12, . . .}
both have the feature that they can be listed – at least in principle. There
is a first element, followed by a second element, followed by a third element,
et cetera, in each set. The next set we’ll look at, Z, can’t be listed so easily.
To list the integers we need to let the dot-dot-dots go both forward (towards
positive infinity) and backwards (towards negative infinity),
Z = {. . . , −3, −2, −1, 0, 1, 2, 3, . . .}.
To show that the integers are actually equinumerous with the natural num-
bers (which is what we’re about to do – and by the way, isn’t that pretty
remarkable?) we need, essentially, to figure out a way to list the integers in
a singly infinite list. Using the symbol ± we can arrange for a singly infinite
listing, and if you think about what the symbol ± means you’ll probably
come up with
Z = {0, 1, −1, 2, −2, 3, −3, . . .}.
This singly infinite listing of the integers does the job we’re after in a sense
– it displays a one-to-one correspondence with N. In fact any singly infinite
listing can be thought of as displaying a one-to-one correspondence with N
– the first entry (or should we say zeroth entry?) in the list is corresponded
with 0, the second entry is corresponded with 1, and so on.
0 1 2 3 4 5 6 7 ...
l l l l l l l
0 1 −1 2 −2 3 −3 4 ...
To make all of this precise we need to be able to explicitly give the one-
to-one correspondence. It isn’t enough to have a picture of it – we need a
formula. Notice that the negative integers are all paired with even naturals
and the positive integers are all paired with odd naturals. This observation
leads us to a piecewise definition for a function that gives the bijection we
seek

−x/2 if x is even
f (x) = .
(x + 1)/2 if x is odd
By the way, notice that since 0 is even it falls into the first case, and
fortunately that formula gives the “right” value.
Exercise. The inverse function, f −1 , must also be defined piecewise, but

based on whether the input is positive or negative. Define the inverse func-
tion.
The examples we’ve done so far have shown that the integers, the natural
numbers and the even naturals all have the same cardinality. This is the
first infinite cardinal number, known as ℵ0 . In a certain sense we could view
both of the equivalences we’ve shown as demonstrating that 2 · ∞ = ∞. Our
next example will lend credence to the rule: ∞ · ∞ = ∞. The Cartesian
product of two finite sets (the set of all ordered pairs with entries from the
sets in question) has cardinality equal to the product of the cardinalities of
the sets. What do you suppose will happen if we let the sets be infinite?
For instance, what is the cardinality of N × N? Consider this: the subset of
ordered pairs that start with a 0 can be thought of as a copy of N sitting
inside this Cartesian product. In fact the subset of ordered pairs starting
with any particular number gives another copy of N inside N × N. There are
infinitely many copies of N sitting inside of N × N! This just really ought to
get us to a larger cardinality. The surprising result that it doesn’t involves an
idea sometimes known as “Cantor’s Snake” – a trick that allows us to list the
elements of N × N in a singly infinite list5 . You can visualize the set N × N
as the points having integer coordinates in the first quadrant (together with
5
Cantor’s snake was originally created to show that Qnoneg and N are equinumerous.
(0, 6) (1, 6) (2, 6) (3, 6) (4, 6) (5, 6) (6, 6)
(0, 5) (1, 5) (2, 5) (3, 5) (4, 5) (5, 5) (6, 5)
(0, 4) (1, 4) (2, 4) (3, 4) (4, 4) (5, 4) (6, 4)
(0, 3) (1, 3) (2, 3) (3, 3) (4, 3) (5, 3) (6, 3)
(0, 2) (1, 2) (2, 2) (3, 2) (4, 2) (5, 2) (6, 2)
(0, 1) (1, 1) (2, 1) (3, 1) (4, 1) (5, 1) (6, 1)
(0, 0) (1, 0) (2, 0) (3, 0) (4, 0) (5, 0) (6, 0)
Figure 8.1: Cantor’s snake winds through the set N × N encountering its
elements one after the other.
the origin and the positive x and y axes). This set of points and the path
through them known as Cantor’s snake is shown in Figure 8.1.
This function was introduced in the exercises for Section 6.5. The version we are presenting
here avoids certain complications.
The diagram in Figure 8.1 gives a visual form of the one-to-one correspon-
dence we seek. In tabular form we would have something like the following.
0 1 2 3 4 5 6 7 8 ...
l l l l l l l l l
(0, 0) (0, 1) (1, 0) (0, 2) (1, 1) (2, 0) (0, 3) (1, 2) (2, 1) . . .
We need to produce a formula. In truth, we should really produce two
formulas. One that takes an ordered pair (x, y) and produces a number n.
Another that takes a number n and produces an ordered pair (x, y) The
number n tells us where the pair (x, y) lies in our infinite listing. There is a
problem though: the second formula (that gives the map from N to N × N)
is really hard to write down – it’s easier to describe the map algorithmically.
A simple observation will help us to deduce the various formulas. The or-
dered pairs along the y-axis (those of the form (0, something)) correspond to
triangular numbers. In fact the pair (0, n) will correspond to the n-th trian-
gular number, T (n) = (n2 + n)/2. The ordered pairs along the descending
slanted line starting from (0, n) all have the feature that the sum of their
coordinates is n (because as the x-coordinate is increasing, the y-coordinate
is decreasing). So, given an ordered pair (x, y), the number corresponding
to the position at the upper end of the slanted line it is on (which will have
coordinates (0, x+y)) will be T (x+y), and the pair (x, y) occurs in the listing
exactly x positions after (0, x + y). Thus, the function f : N × N −→ N is
given by
(x + y)2 + (x + y)
f (x, y) = x + T (x + y) = x + .
2
To go the other direction – that is, to take a position in the listing and derive
an ordered pair – we need to figure out where a given number lies relative
to the triangular numbers. For instance, try to figure out what (x, y) pair
position number 13 will correspond with. Well, the next smaller triangular
number is 10 which is T (4), so 13 will be the number of an ordered pair along

the descending line whose y-intercept is 4. In fact, 13 will be paired with an
ordered pair having a 3 in the x-coordinate (since 13 is 3 larger than 10) so
it follows that f −1 (13) = (3, 1).
Of course we need to generalize this procedure. One of the hardest parts
of finding that generalization is finding the number 4 in the above example
(when we just happen to notice that T (4) = 10 ). What we’re really doing
there is inverting the function T (n). Finding an inverse for T (n) = (n2 +n)/2
was the essence of one of the exercises in Section 6.6. The parabola y =
(x2 + x)/2 has roots at 0 and −1 and is scaled by a factor of 1/2 relative to
the “standard” parabola y = x2 . Its vertex is at (−1/2, −1/8). The graph
of the inverse relation is, of course, obtained by reflecting through the line
y = x and by considering scaling and horizontal/ vertical translations we can
deduce a formula for a function that gives a right inverse for T ,
p
T −1 (x) = 2x + 1/4 − 1/2.
p
So, given n, a position in the listing, we calculate A = b 2n + 1/4−1/2c.
The x-coordinate of our ordered pair is n−T (A) and the y-coordinate is A−x.
It is not pretty, but the above discussion can be translated into a formula for
f −1 .

−1 (p p
f (n) = 2n + 1/4 − 1/2c2 + b 2n + 1/4 − 1/2c2,
n−
b
p p !
p b 2n + 1/4 − 1/2c2 + b 2n + 1/4 − 1/2c
b 2n + 1/4 − 1/2c − n + .
2
When restricted to the appropriate sets (f ’s domain is restricted to N × N

and f −1 ’s domain is restricted to N), these functions are two-sided inverses
for one another. That fact is sufficient to prove that f is bijective. So
far we have shown that the sets Enoneg , N, Z and N × N all have the same
cardinality — ℵ0 . We plan to provide an argument that there actually are

other infinite cardinals in the next section. Before leaving the present topic
(examples of set equivalence) we’d like to present another nice technique for
deriving the bijective correspondences we use to show that sets are equivalent
– geometric constructions. Consider the set of points on the line segment
[0, 1]. Now consider the set of points on the line segment [0, 2]. This second
line segment, being twice as long as the first, must have a lot more points on
it. Right?
Well, perhaps you’re getting used to this sort of thing. . . The interval [0, 1]
is a subset of the interval [0, 2], but since both represent infinite sets of points
it’s possible they actually have the same cardinality. We can prove that this is
so using a geometric technique. We position the line segments appropriately
and then use projection from a carefully chosen point to develop a bijection.
Imagine both intervals as lying on the x-axis in the x-y plane. Shift the
smaller interval up one unit so that it lies on the line y = 1. Now, use
projection from the point (0, 2), to visualize the correspondence see Figure 8.2
By considering appropriate projections we can prove that any two ar-
bitrary intervals (say [a, b] and [c, d]) have the same cardinalities! It also
isn’t all that hard to derive a formula for a bijective function between two
intervals.
(x − a)(d − c)
f (x) = c +
(b − a)
There are other geometric constructions which we can use to show that
there are the same number of points in a variety of entities. For example,
consider the upper half of the unit circle (Remember the unit circle from
Trig? All points (x, y) satisfying x2 + y 2 = 1.) This is a semi-circle having a
radius of 1, so the arclength of said semi-circle is π. It isn’t hard to imagine
that this semi-circular arc contains the same number of points as an interval
of length π, and we’ve already argued that all intervals contain the same
Figure 8.2: Projection from a point can be used to show that intervals of
different lengths contain the same number of points.
number of points. . . But, a nice example of geometric projection — vertical

projection (a.k.a. π1 ) — can be used to show that (for example) the interval
(−1, 1) and the portion of the unit circle lying in the upper half-plane are
equinumerous.
Once the bijection is understood geometrically it is fairly simple to provide
formulas. To go from the semi-circle to the interval, we just forget about the
y-coordinate:
f (x, y) = x.
To go in the other direction we need to recompute the missing y-value:
√
f −1 (x) = (x, 1 − x2 ).
Now we’re ready to put some of these ideas together in order to prove
Figure 8.3: Vertical projection provides a bijective correspondence between

an interval and a semi-circle.
something really quite remarkable. It may be okay to say that line segments
of different lengths are equinumerous – although ones intuition still balks
at the idea that a line a mile long only has the same number of points on
it as a line an inch long (or, if you prefer, make that a centimeter versus
a kilometer). Would you believe that the entire line – that is the infinitely
extended line – has no more points on it than a tiny little segment? You
should be ready to prove this one yourself.
Exercise. Find a point such that projection from that point determines a
one-to-one correspondence between the portion of the unit circle in the upper
half plane and the line y = 1.
In the exercises from Section 8.1 you were supposed to show that set
equivalence is an equivalence relation. Part of that proof should have been
showing that the relation is transitive, and that really just comes down to
showing that the composition of two bijections is itself a bijection. If you
didn’t make it through that exercise give it another try now, but whether
or not you can finish that proof it should be evident what that transitivity
means to us in the current situation. Any pair of line segments are the same
size – a line segment (i.e. an interval) and a semi-circle are the same size –
the semi-circle and an infinite line are the same size – transitivity tells us that
an infinitely extended line has the same number of points as (for example)
the interval (0, 1).
Exercises — 8.2
1. Prove that positive numbers of the form 3k + 1 are equinumerous with

positive numbers of the form 4k + 2.
(x − a)(d − c)
2. Prove that f (x) = c + provides a bijection from the
(b − a)
interval [a, b] to the interval [c, d].
3. Prove that any two circles are equinumerous (as sets of points).
4. Determine a formula for the bijection from (−1, 1) to the line y = 1

determined by vertical projection onto the upper half of the unit circle,
followed by projection from the point (0, 0).
5. It is possible to generalize the argument that shows a line segment is

equivalent to a line to higher dimensions. In two dimensions we would
show that the unit disk (the interior of the unit circle) is equinumerous
with the entire plane R × R. In three dimensions we would show that
the unit ball (the interior of the unit sphere) is equinumerous with the
entire space R3 = R × R × R. Here we would like you to prove the
two-dimensional case.
Gnomonic projection is a style of map rendering in which a portion of
a sphere is projected onto a plane that is tangent to the sphere. The
sphere’s center is used as the point to project from. Combine vertical
projection from the unit disk in the x–y plane to the upper half of the
unit sphere x2 + y 2 + z 2 = 1, with gnomonic projection from the unit
sphere to the plane z = 1, to deduce a bijection between the unit disk
and the (infinite) plane.
8.3. CANTOR’S THEOREM 355
8.3 Cantor’s theorem
Many people believe that the result known as Cantor’s theorem says that
the real numbers, R, have a greater cardinality than the natural numbers, N.
That isn’t quite right. In fact Cantor’s theorem is a much broader statement,
one of whose consequences is that |R| > |N|. Before we go on to discuss
Cantor’s theorem in full generality, we’ll first explore it, essentially, in this
simplified form. Once we know that |R| 6= |N|, we’ll be in a position to
explore a lot of interesting issues relative to the infinite. In particular, this
result means that there are at least two cardinal numbers that are infinite –
thus the “infinity is infinity” idea will be discredited. Once we have the full
power of Cantor’s theorem, we’ll see just how completely wrong that concept
is.
To show that some pair of sets are not equivalent it is necessary to show
that there cannot be a one-to-one correspondence between them. Ordinarily,
one would try to argue by contradiction in such a situation. That is what
we’ll need to do to show that the reals and the naturals are not equinumer-
ous. We’ll presume that they are in fact the same size and try to reach a
contradiction.
What exactly does the assumption that R and N are equivalent mean?
It means there is a one-to-one correspondence, that is, a bijective function
from R to N. In a nutshell, it means that it is possible to list all the real
numbers in a singly-infinite list. Now, it is certainly possible to make an
infinite list of real numbers (since N ⊆ R, by listing the naturals themselves
we are making an infinite list of reals!). The problem is that we would need
to be sure that every real number is on the list somewhere. In fact, since
we’ve used a geometric argument to show that the interval (0, 1) and the set
R are equinumerous, it will be sufficient to presume that there is an infinite
list containing all the numbers in the interval (0, 1).
Exercise. Notice that, for example, π − 3 is a real number in (0, 1). Make
a list of 10 real numbers in the interval (0, 1). Make sure that at least 5 of
them are not rational.
In the previous exercise, you’ve started the job, but we need to presume
that it is truly possible to complete this job. That is, we must presume that
there really is an infinite list containing every real number in the interval
(0, 1).
Once we have an infinite list containing every real number in the inter-
val (0, 1) we have to face up to a second issue. What does it really mean
to list a particular real number? For instance if e − 2 is in the seventh
position on our list, is it OK to write “e − 2” there or should we write
“0.7182818284590452354. . . ”? Clearly it would be simpler to write “e − 2”
but it isn’t necessarily possible to do something of that kind for every real
number – on the other hand, writing down the decimal expansion is a prob-
lem too; in a certain sense, “most” real numbers in (0, 1) have infinitely long
decimal expansions. There is also another problem with decimal expansions;
they aren’t unique. For example, there is really no difference between the
finite expansion 0.5 and the infinitely long expansion 0.49.
Rather than writing something like “e−2” or “0.7182818284590452354. . . ”,
we are going to in fact write “.1011011111100001010100010110001010001010
. . . ” In other words, we are going to write the base-2 expansions of the real
numbers in our list. Now, the issue of non-uniqueness is still there in binary,
and in fact if we were to stay in base-10 it would be possible to plug a certain
gap in our argument – but the binary version of this argument has some es-
pecially nice features. Every binary (or for that matter decimal) expansion
corresponds to a unique real number, but it doesn’t work out so well the
other way around — there are sometimes two different binary expansions
that correspond to the same real number. There is a lovely fact that we
are not going to prove (you may get to see this result proved in a course in
Real Analysis) that points up the problem. Whenever two different binary
expansions represent the same real number, one of them is a terminating ex-
pansion (it ends in infinitely many 0’s) and the other is an infinite expansion
(it ends in infinitely many 1’s). We won’t prove this fact, but the gist of the
argument is a proof by contradiction — you may be able to get the point by
studying Figure 8.4. (Try to see how it would be possible to find a number
in between two binary expansions that didn’t end in all-zeros and all-ones.)
.0 .1
.00 .01 .10 .11
.000 .001 .010 .011 .100 .101 .110 .111
Figure 8.4: The base-2 expansions of reals in the interval [0, 1] are the leaves
of an infinite tree.
So, instead of showing that the set of reals in (0, 1) can’t be put in one-to-
one correspondence with N, what we’re really going to do is show that their
binary expansions can’t be put in one-to-one correspondence with N. Since
there are an infinite number of reals that have two different binary expansions
this doesn’t really do the job as advertised at the beginning of this section.
(Perhaps you are getting used to our wily ways by now — yes, this does mean
that we’re going to ask you to do the real proof in the exercises.) The set of
binary numerals, {0, 1}, is an instance of a mathematical structure known as

a field; basically, that means that it’s possible to add, subtract, multiply and
divide (but not divide by 0) with them. We are only mentioning this fact so
that you’ll understand why the set {0, 1} is often referred to as F2 . We’re
only mentioning that fact so that you’ll understand why we call the set of
all possible binary expansion F∞
2 . Finally, we’re only mentioning that fact
so that we’ll have a succinct way of expressing this set. Now we can write
“F∞
2 ” rather than “the set of all possible infinitely-long binary sequences.”
Suppose we had a listing of all the elements of F∞

2 . We would have an
infinite list of things, each of which is itself an infinite list of 0’s and 1’s.
So what? We need to proceed from here to find a contradiction.
This argument that we’ve been edging towards is known as Cantor’s diag-
onalization argument. The reason for this name is that our listing of binary
representations looks like an enormous table of binary digits and the contra-
diction is deduced by looking at the diagonal of this infinite-by-infinite table.
The diagonal is itself an infinitely long binary string — in other words, the
diagonal can be thought of as a binary expansion itself. If we take the com-
plement of the diagonal, (switch every 0 to a 1 and vice versa) we will also
have a thing that can be regarded as a binary expansion and this binary
expansion can’t be one of the ones on the list! This bit-flipped version of
the diagonal is different from the first binary expansion in the first position,
it is different from the second binary expansion in the second position, it is
different from the third binary expansion in the third position, and so on.
The very presumption that we could list all of the elements of F∞
2 allows us
to construct an element of F∞
2 that could not be on the list!
This argument has been generalized many times, so this is the first in a
class of things known as diagonal arguments. Diagonal arguments have been
used to settle several important mathematical questions. There is a valid
diagonal argument that even does what we’d originally set out to do: prove
that N and R are not equinumerous. Strangely, the argument can’t be made
to work in binary, and since you’re going to be asked to write it up in the
exercises, we want to point out one of the potential pitfalls. If we were to
use a diagonal argument to show that (0, 1) isn’t countable, we would start
by assuming that every element of (0, 1) was written down in a list. For
most real numbers in (0, 1) we could write out their binary representation
uniquely, but for some we would have to make a choice: should we write down
the representation that terminates, or the one that ends in infinitely-many
1’s? Suppose we choose to use the terminating representations, then none
of the infinite binary strings that end with all 1’s will be on the list. It’s
possible that the thing we get when we complement the diagonal is one of
these (unlisted) binary strings so we don’t necessarily have a contradiction.
If we make the other choice – use the infinite binary representation when we
have a choice – there is a similar problem. You may think that our use of
binary representations for real numbers was foolish in light of the failure of
the argument to “go through” in binary. Especially since, as we’ve alluded to,
it can be made to work in decimal. The reason for our apparent stubbornness
is that these infinite binary strings do something else that’s very nice. An
infinitely long binary sequence can be thought of as the indicator function of
a subset of N. For example, .001101010001 is the indicator of {2, 3, 5, 7, 11}.
Exercise. Complete the table.
binary expansion subset of N

.1 {0}
.0111
{2, 4, 6}
.01
{3k + 1 k ∈ N}
The set, F∞
2 , we’ve been working with is in one-to-one correspondence
with the power set of the natural numbers, P(N). When viewed in this
light, the proof we did above showed that the power set of N has an infinite
cardinality strictly greater than that of N itself. In other words, P(N) is
uncountable.
What Cantor’s theorem says is that this always works. If A is any set,
and P(A) is its power set then |A| < |P(A)|. In a way, this more general
theorem is easier to prove than the specific case we just handled.
Theorem 8.3.1 (Cantor). For all sets A, A is not equivalent to P(A).
Proof: Suppose that there is a set A that can be placed in one-to-

one correspondence with its power set. Then there is a bijective
function f : A −→ P(A). We will deduce a contradiction by
constructing a subset of A (i.e. a member of P(A)) that cannot
be in the range of f .
Let S = {x ∈ A x ∈
/ f (x)}.
If S is in the range of f , there is a preimage y such that S = f (y).
But, if such a y exists then the membership question, y ∈ S, must
either be true or false. If y ∈ S, then because S = f (y), and S
consists of those elements that are not in their images, it follows
that y ∈
/ S. On the other hand, if y ∈
/ S then y ∈
/ f (y) so (by the
definition of S) it follows that y ∈ S. Either possibility leads to
the other, which is a contradiction.
Q.E.D.
Cantor’s theorem guarantees that there is an infinite hierarchy of infinite

cardinal numbers. Let’s put it another way. People have sought a construc-
tion that, given an infinite set, could be used to create a strictly larger set.
For instance, the Cartesian product works like this if our sets are finite —
A × A is strictly bigger than A when A is a finite set. But, as we’ve already

seen, this is not necessarily so if A is infinite (remember the “snake” argu-
ment that N and N × N are equivalent). The real import of Cantor’s theorem
is that taking the power set of a set does create a set of larger cardinality.
So we get an infinite tower of infinite cardinalities, starting with ℵ0 = |N|,
by successively taking power sets.
ℵ0 = |N| < |P(N)| < |P(P(N))| < |P(P(P(N)))| < . . .

Exercises — 8.3
1. Determine a substitution rule – a consistent way of replacing one digit

with another along the diagonal so that a diagonalization proof showing
that the interval (0, 1) is uncountable will work in decimal. Write up
the proof.
2. Can a diagonalization proof showing that the interval (0, 1) is uncount-

able be made workable in base-3 (ternary) notation?
3. In the proof of Cantor’s theorem we construct a set S that cannot

be in the image of a presumed bijection from A to P(A). Suppose
A = {1, 2, 3} and f determines the following correspondences: 1 ←→ ∅,
2 ←→ {1, 3} and 3 ←→ {1, 2, 3}. What is S?
4. An argument very similar to the one embodied in the proof of Can-

tor’s theorem is found in the Barber’s paradox. This paradox was
originally introduced in the popular press in order to give laypeople an
understanding of Cantor’s theorem and Russell’s paradox. It sounds
somewhat sexist to modern ears. (For example, it is presumed without
comment that the Barber is male.)
In a small town there is a Barber who shaves those men (and

only those men) who do not shave themselves. Who shaves
the Barber?
Explain the similarity to the proof of Cantor’s theorem.
5. Cantor’s theorem, applied to the set of all sets leads to an interesting

paradox. The power set of the set of all sets is a collection of sets, so
it must be contained in the set of all sets. Discuss the paradox and
determine a way of resolving it.
6. Verify that the final deduction in the proof of Cantor’s theorem, “(y ∈
S =⇒ y ∈
/ S) ∧ (y ∈
/ S =⇒ y ∈ S),” is truly a contradiction.
8.4 Dominance
We’ve said a lot about the equivalence relation determined by Cantor’s defini-
tion of set equivalence. We’ve also, occasionally, written things like |A| < |B|,
without being particularly clear about what that means. It’s now time to
come clean. There is actually a (perhaps) more fundamental notion used
for comparing set sizes than equivalence — dominance. Dominance is an
ordering relation on the class of all sets. One should probably really define
dominance first and then define set equivalence in terms of it. We haven’t
followed that plan for (at least) two reasons. First, many people may want
to skip this section — the results of this section depend on the difficult
Cantor-Bernstein-Schröder theorem6 . Second, we will later take the view
that dominance should really be considered to be an ordering relation on the
set of all cardinal numbers – i.e. the equivalence classes of the set equiva-
lence relation – not on the collection of all sets. From that perspective, set
equivalence really needs to be defined before dominance.
One set is said to dominate another if there is a function from the latter
into the former. More formally, we have the following
Definition. If A and B are sets, we say “A dominates B” and write |A| >
|B| iff there is an injective function f with domain B and codomain A.
It is easy to see that this relation is reflexive and transitive. The Cantor-
Bernstein-Schröder theorem proves that it is also anti-symmetric — which
means dominance is an ordering relation. Be advised that there is an abuse
of terminology here that one must be careful about — what are the domain
and range of the “dominance” relation? The definition would lead us to
6
This theorem has been known for many years as the Schröder-Bernstein theorem, but,
lately, has had Cantor’s name added as well. Since Cantor proved the result before the
other gentlemen this is fitting. It is also known as the Cantor-Bernstein theorem (leaving
out Schröder) which doesn’t seem very nice.
8.4. DOMINANCE 365
think that sets are the things that go on either side of the “dominance”
relation, but the notation is a bit more honest, “|A| > |B|” indicates that
the things really being compared are the cardinal numbers of sets (not the
sets themselves). Thus anti-symmetry for this relation is
(|A| > |B|) ∧ (|B| > |A|) =⇒ (|A| = |B|).
In other words, if A dominates B and vice versa, then A and B are

equivalent sets — a strict interpretation of anti-symmetry for this relation
might lead to the conclusion that A and B are actually the same set, which
is clearly an absurdity.
Naturally, we want to prove the Cantor-Bernstein-Schröder theorem (which
we’re going to start calling the C-B-S theorem for brevity), but first it’ll be
instructive to look at some of its consequences. Once we have the C-B-S
theorem we get a very useful shortcut for proving set equivalences. Given
sets A and B, if we can find injective functions going between them in both
directions, we’ll know that they’re equivalent. So, for example, we can use
C-B-S to prove that the set of all infinite binary strings and the set of reals in
(0, 1) really are equinumerous. (In case you had some remaining doubt. . . )
It is easy to dream up an injective function from (0, 1) to F∞
2 : just send a
real number to its binary expansion, and if there are two, make a consistent
choice — let’s say we’ll take the non-terminating expansion.
There is a cute thought-experiment called Hilbert’s Hotel that will lead
us to a technique for developing an injective function in the other direction.
Hilbert’s Hotel has ℵ0 rooms. If any countable collection of guests show
up there will be enough rooms for everyone. Suppose you arrive at Hilbert’s
hotel one dark and stormy evening and the “No Vacancy” light is on — there
are already a denumerable number of guests there — every room is full. The
clerk sees you dejectedly considering your options, trying to think of another
hotel that might still have rooms when, clearly, a very large convention is
in town. He rushes out and says “My friend, have no fear! Even though we
have no vacancies, there is always room for one more at our establishment.”
He goes into the office and makes the following announcement on the PA
system. “Ladies and Gentlemen, in order to accommodate an incoming guest,
please vacate your room and move to the room numbered one higher. Thank
you.” There is an infinite amount of grumbling, but shortly you find yourself
occupying room number 1.
To develop an injection from F∞

2 to (0, 1) we’ll use “room number 1” to
separate the binary expansions that represent the same real number. Move
all the digits of a binary expansion down by one, and make the first digit
0 for (say) the terminating expansions and 1 for the non-terminating ones.
Now consider these expansions as real numbers — all the expansions that
previously coincided are now separated into the intervals (0, 1/2) and (1/2, 1).
Notice how funny this map is, there are now many, many, (infinitely-many)
real numbers with no preimages. For instance, only a subset of the rational
numbers in (0, 1/2) have preimages. Nevertheless, the map is injective, so C-
B-S tells us that F∞
2 and (0, 1) are equivalent. There are quite a few different
proofs of the C-B-S theorem. The one Cantor himself wrote relies on the
axiom of choice. The axiom of choice was somewhat controversial when it was
introduced, but these days most mathematicians will use it without qualms.
What it says (essentially) is that it is possible to make an infinite number of
choices. More precisely, it says that if we have an infinite set consisting of
non-empty sets, it is possible to select an element out of each set. If there
is a definable rule for picking such an element (as is the case, for example,
when we selected the nonterminating decimal expansion whenever there was
a choice in defining the injection from (0, 1) to F∞
2 ) the axiom of choice isn’t
needed. The usual axioms for set theory were developed by Zermelo and
Frankel, so you may hear people speak of the ZF axioms. If, in addition,
we want to specifically allow the axiom of choice, we are in the ZFC axiom
8.4. DOMINANCE 367
system. If it’s possible to construct a proof for a given theorem without using
the axiom of choice, almost everyone would agree that that is preferable. On
the other hand, a proof of the C-B-S theorem, which necessarily must be able
to deal with uncountably infinite sets, will have to depend on some sort of
notion that will allow us to deal with huge infinities.
The proof we will present here7 is attributed to Julius König. König
was a contemporary of Cantor’s who was (initially) very much respected by
him. Cantor came to dislike König after the latter presented a well-publicized
(and ultimately wrong) lecture claiming the continuum hypothesis was false.
Apparently the continuum hypothesis was one of Cantor’s favorite ideas,
because he seems to have construed König’s lecture as a personal attack.
Anyway. . .
König’s proof of C-B-S doesn’t use the axiom of choice, but it does have
its own strangeness: a function that is not necessarily computable — that is,
a function for which (for certain inputs) it may not be possible to compute
an output in a finite amount of time! Except for this oddity, König’s proof
is probably the easiest to understand of all the proofs of C-B-S. Before we
get too far into the proof it is essential that we understand the basic setup.
The Cantor-Bernstein-Schröder theorem states that whenever A and B are
sets and there are injective functions f : A −→ B and g : B −→ A, then it
follows that A and B are equivalent. Saying A and B are equivalent means
that we can find a bijective function between them. So, to prove C-B-S, we
hypothesize the two injections and somehow we must construct the bijection.
Figure 8.5 has a presumption in it — that A and B are countable — which
need not be the case. Nevertheless, it gives us a good picture to work from.
The basic hypotheses, that A and B are sets and we have two functions,
one from A into B and another from B into A, are shown. We will have to
build our bijective function in a piecewise manner. If there is a non-empty
7
We first encountered this proof in a Wikipedia article[3].
A B
a1 b1
a2 b2
a3 b3
a4 b4
a5 b5
a6 b6
a7 b7
a8 b8
a9 b9
a10 b10
a11 b11
a12 b12
Figure 8.5: Hypotheses for proving the Cantor-Bernstein-Schröder theorem:

two sets with injective functions going both ways.
8.4. DOMINANCE 369
intersection between A and B, we can use the identity function for that part
of the domain of our bijection. So, without loss of generality, we can presume
that A and B are disjoint. We can use the functions f and g to create infinite
sequences, which alternate back and forth between A and B, containing any
particular element. Suppose a ∈ A is an arbitrary element. Since f is defined
on all of A, we can compute f (a). Now since f (a) is an element of B, and g
is defined on all of B, we can compute g(f (a)), and so on. Thus, we get the
infinite sequence
a, f (a), g(f (a)), f (g(f (a))), . . .
If the element a also happens to be the image of something under g (this

may or may not be so — since g isn’t necessarily onto) then we can also
extend this sequence to the left. Indeed, it may be possible to extend the
sequence infinitely far to the left, or, this process may stop when one of f −1
or g −1 fails to be defined.
. . . g −1 (f −1 (g −1 (a))), f −1 (g −1 (a)), g −1 (a), a, f (a), g(f (a)), f (g(f (a))), . . .
Now, every element of the disjoint union of A and B is in one of these

sequences. Also, it is easy to see that these sequences are either disjoint
or identical. Taking these two facts together it follows that these sequences
form a partition of A ∪ B. We’ll define a bijection φ : A −→ B by deciding
what it must do on these sequences. There are four possibilities for how the
sequences we’ve just defined can play out. In extending them to the left, we
may run into a place where one of the inverse functions needed isn’t defined
— or not. We say a sequence is an A-stopper, if, in extending to the left, we
end up on an element of A that has no preimage under g (see Figure 8.6).
Similarly, we can define a B-stopper. If the inverse functions are always
defined within a given sequence there are also two possibilities; the sequence
may be finite (and so it must be cyclic in nature) or the sequence may be
truly infinite.
Finally, here is a definition for φ.
g −1 (x) if x is in a B-stopper

φ(x) =
f (x) otherwise
Notice that if a sequence is either cyclic or infinite it doesn’t matter
whether we use f or g −1 since both will be defined for all elements of such
sequences. Also, certainly f will work if we are in an A-stopper. The function
we’ve just created is perfectly well-defined, but it may take arbitrarily long
to determine whether we have an element of a B-stopper, as opposed to an
element of an infinite sequence. We cannot determine whether we’re in an
infinite versus a finite sequence in a prescribed finite number of steps.
8.4. DOMINANCE 371
A B
a1 b1
a2 b2
a3 b3
a4 b4
a5 b5
a6 b6
a7 b7
a8 b8
a9 b9
a10 b10
a11 b11
a12 b12
Figure 8.6: An A-stopper is an infinite sequence that terminates to the left

in A.
Exercises — 8.4
1. How could the clerk at the Hilbert Hotel accommodate a countable

number of new guests?
2. Let F be the collection of all real-valued functions defined on the real

line. Find an injection from R to F . Do you think it is possible to find
an injection going the other way? In other words, do you think that F
and R are equivalent? Explain.
3. Fill in the details of the proof that dominance is an ordering relation.

(You may simply cite the C-B-S theorem in proving anti-symmetry.)
a
4. We can inject Q into Z by sending ± to ±2a 3b . Use this and an-
b
other obvious injection to (in light of the C-B-S theorem) reaffirm the
equivalence of these sets.
8.5. CH AND GCH 373
8.5 The continuum hypothesis and the gen-

eralized continuum hypothesis
The word “continuum” in the title of this section is used to indicate sets of
points that have a certain continuity property. For example, in a real interval
it is possible to move from one point to another, in a smooth fashion, without
ever leaving the interval. In a range of rational numbers this is not possible,
because there are irrational values in between every pair of rationals. There
are many sets that behave as a continuum – the intervals (a, b) or [a, b], the
entire real line R, the x-y plane R × R, a volume in 3-dimensional space (or
for that matter the entire space R3 ). It turns out that all of these sets have
the same size.
The cardinality of the continuum, denote c, is the cardinality of all of the
sets above.
In the previous section we mentioned the continuum hypothesis and how
angry Cantor became when someone (König) tried to prove it was false. In
this section we’ll delve a little deeper into what the continuum hypothesis
says and even take a look at CH’s big brother, GCH. Before doing so, it
seems like a good idea to look into the equivalences we’ve asserted about all
those sets above which (if you trust us) have the cardinality c.
We’ve already seen that an interval is equivalent to the entire real line
but the notion that the entire infinite Cartesian plane has no more points
in it than an interval one inch long defies our intuition. Our conception
of dimensionality leads us to think that things of higher dimension must be
larger than those of lower dimension. This preconception is false as we can see
by demonstrating that a 1×1 square can be put in one-to-one correspondence
with the unit interval. Let S = {(x, y) 0 < x < 1 ∧ 0 < y < 1} and let I
be the open unit interval (0, 1). We can use the Cantor-Bernstein-Schroeder
theorem to show that S and I are equinumerous – we just need to find
injections from I to S and vice versa. Given an element r in I we can map

it injectively to the point (r, r) in S. To go in the other direction, consider a
point (a, b) in S and write out the decimal expansions of a and b:
a = 0.a1 a2 a3 a4 a5 . . .
b = 0.b1 b2 b3 b4 b5 . . .
From these decimal expansions, we can create the decimal expansion of

a number in I by interleaving the digits of a and b. Let
s = 0.a1 b1 a2 b2 a3 b3 . . .
be the image of (a, b). If two different points get mapped to the same value
s then both points have x and y coordinates that agree in every position of
their decimal expansion (so they must really be equal). It is a little bit
harder to create a bijective function from S to I (and thus to show the
equivalence directly, without appealing to C-B-S). The problem is that, once
again, we need to deal with the non-uniqueness of decimal representations
of real numbers. If we make the choice that, whenever there is a choice to
be made, we will use the non-terminating decimal expansions for our real
numbers there will be elements of I not in the image of the map determined
by interleaving digits (for example 0.15401050902060503 is the interleaving
of the digits after the decimal point in π = 3.141592653 . . . and 1/2 = 0.5,
this is clearly an element of I but it cant be in the image of our map since 1/2
should be represented by 0.49 according to our convention. If we try other
conventions for dealing with the non-uniqueness it is possible to find other
examples that show simple interleaving will not be surjective. A slightly
more subtle approach is required.
Presume that all decimal expansions are non-terminating (as we can,
WLOG) and use the following approach: Write out the decimal expansion
of the coordinates of a point (a, b) in S. Form the digits into blocks with as
8.5. CH AND GCH 375
many 0s as possible followed by a non-zero digit. Finally, interleave these

blocks.
For example if
a = 0.124520047019902 . . .
and
b = 0.004015648000031 . . .
we would separate the digits into blocks as follows:
a = 0.1 2 4 5 2 004 7 01 9 9 02 . . .
and
b = 0.004 01 5 6 4 8 00003 1 . . .
and the number formed by interleaving them would be
s = 0.10042014556240048 . . .
We’ve shown that the unit square, S, and the unit interval, I, have the
same cardinality. These arguments can be extended to show that all of R×R
also has this cardinality (c).
So now let’s turn to the continuum hypothesis.
We mentioned earlier in this chapter that the cardinality of N is denoted
ℵ0 . The fact that that capital letter aleph is wearing a subscript ought to
make you wonder what other aleph-sub-something-or-others there are out
there. What is ℵ1 ? What about ℵ2 ? Cantor presumed that there was a
sequence of cardinal numbers (which is itself, of course, infinite) that give
all of the possible infinities. The smallest infinite set that anyone seems
to be able to imagine is N, so Cantor called that cardinality ℵ0 . What

ever the “next” infinite cardinal is, is called ℵ1 . It’s conceivable that there
actually isn’t a “next” infinite cardinal after ℵ0 — it might be the case that
the collection of infinite cardinal numbers isn’t well-ordered! In any case, if
there is a “next” infinite cardinal, what is it? Cantor’s theorem shows that
there is a way to build some infinite cardinal bigger than ℵ0 — just apply
the power set construction. The continuum hypothesis just says that this
bigger cardinality that we get by applying the power set construction is that
“next” cardinality we’ve been talking about.
To re-iterate, we’ve shown that the power set of N is equivalent to the
interval (0, 1) which is one of the sets whose cardinality is c. So the continuum
hypothesis, the thing that got Georg Cantor so very heated up, comes down
to asserting that
ℵ1 = c.
There really should be a big question mark over that. A really big ques-
tion mark. It turns out that the continuum hypothesis lives in a really weird
world. . . To this day, no one has the least notion of whether it is true or false.
But wait! That’s not all! The real weirdness is that it would appear to be
impossible to decide. Well, that’s not so bad – after all, we talked about
undecidable sentences way back in the beginning of Chapter 2. Okay, so
here’s the ultimate weirdness. It has been proved that one can’t prove the
continuum hypothesis. It has also been proved that one can’t disprove the
continuum hypothesis.
Having reached this stage in a book about proving things I hope that the
last two sentences in the previous paragraph caused some thought along the
lines of “well, ok, with respect to what axioms?” to run through your head.
So, if you did think something along those lines pat yourself on the back.
And if you didn’t then recognize that you need to start thinking that way
8.5. CH AND GCH 377
— things are proved or disproved only in a relative way, it depends what

axioms you allow yourself to work with. The usual axioms for mathematics
are called ZFC; the Zermelo-Frankel set theory axioms together with the
axiom of choice. The “ultimate weirdness” we’ve been describing about the
continuum hypothesis is a result due to a gentleman named Paul Cohen that
says “CH is independent of ZFC.” More pedantically – it is impossible to
either prove or disprove the continuum hypothesis within the framework of
the ZFC axiom system.
It would be really nice to end this text by mentioning Paul Cohen, but
there is one last thing we’d like to accomplish — explain what GCH means.
So here goes.
The generalized continuum hypothesis says that the power set construc-
tion is basically the only way to get from one infinite cardinality to the next.
In other words GCH says that not only does P(N) have the cardinality known
as ℵ1 , but every other aleph number can be realized by applying the power set
construction a bunch of times. Some people would express this symbolically
by writing
∀n ∈ N, ℵn+1 = 2ℵn .
I’d really rather not bring this text to a close with that monstrosity so
instead I think I’ll just say
Paul Cohen.
Hah! I did it! I ended the book by sayi. . . Hunh? Oh.
Paul Cohen.
Chapter 9
Proof techniques IV — Magic
If you can keep your head when all about you are losing theirs, it’s just
possible you haven’t grasped the situation. –Jean Kerr
The famous mathematician Paul Erdös is said to have believed that God
has a Book in which all the really elegant proofs are written. The greatest
praise that a collaborator1 could receive from Erdös was that they had dis-
covered a “Book proof.” It is not easy or straightforward for a mere mortal
to come up with a Book proof but notice that, since the Book is inaccessible
to the living, all the Book proofs of which we are aware were constructed by
ordinary human beings. In other words, it’s not impossible!
The title of this final chapter is intended to be whimsical – there is no real
magic involved in any of the arguments that we’ll look at. Nevertheless, if you
reflect a bit on the mental processes that must have gone into the development
of these elegant proofs, perhaps you’ll agree that there is something magical
there. At a minimum we hope that you’ll agree that they are beautiful –
they are proofs from the Book2 .
1
The collaborators of Paul Erdös were legion. His collaborators, and their collaborators,
and their collaborators, etc. are organized into a tree structure according to their so-called
Erdös number, see [5].
2
There is a lovely book entitled “Proofs from the Book” [2] that has a nice collection
379
380 CHAPTER 9. PROOF TECHNIQUES IV — MAGIC
Full disclosure: Several of the topics in this section were unknown to

the author until he visited the excellent mathematics website maintained by
Alexander Bogomolny at
http://www.cut-the-knot.org/
of Book proofs.
9.1. MORLEY’S MIRACLE 381
9.1 Morley’s miracle
Probably you have heard of the impossibility of trisecting an angle. (Hold on

for a quick rant about the importance of understanding your hypotheses. . . )
What’s actually true is that you can’t trisect a generic angle if you accept
the restriction of using the old-fashioned tools of Euclidean geometry: the
compass and straight-edge. There are a lot of constructions that can’t be
done using just a straight-edge and compass – angle trisection, duplication
of a cube3 , squaring a circle, constructing a regular heptagon, et cetera.
If you allow yourself to use a ruler – i.e. a straight-edge with marks on
it (indeed you really only need two marks a unit distance apart) then angle
trisection can be done via what is known as a Neusis construction. Never-
theless, because of the central place of Euclid’s Elements in mathematical
training throughout the centuries, and thereby, a very strong predilection
towards that which is possible via compass and straight-edge alone, it is per-
haps not surprising that a perfectly beautiful result that involved trisecting
angles went undiscovered until 1899, when Frank Morley stated his Trisector
Theorem. There is much more to this result than we will state here – so much
more that the name “Morley’s Miracle” that has been given to the Trisector
theorem is truly justified – but even the simple, initial part of this beautiful
theory is arguably miraculous! To learn more about Morley’s theorem and
its extension see [8].
3
Duplicating the cube is also known as the Delian problem – the problem comes from
a pronouncement by the oracle of Apollo at Delos that a plague afflicting the Athenians
would be lifted if they built an altar to Apollo that was twice as big as the existing altar.
The existing altar was a cube, one meter on a side, so they carefully built a two meter cube
– but the plague raged on. Apparently what Apollo wanted was a cube that had double
√
the volume of the present altar – it’s side length would have to be 3 2 ≈ 1.25992 and since
this was Greece and it was around 430 B.C. and there were no electronic calculators, they
were basically just screwed.
So, let’s state the theorem!

Start with an arbitrary triangle 4ABC. Trisect each of its angles to
obtain a diagram something like that in Figure 9.1.
A B
Figure 9.1: The setup for Morley’s Miracle – start with an arbitrary triangle
and trisect each of its angles.
The six angle trisectors that we’ve just drawn intersect one another in
quite a few points.
Exercise. You could literally count the number of intersection points between
the angle trisectors on the diagram, but you should also be able to count them
(perhaps we should say “double-count them”) combinatorially. Give it a try!
Among the points of intersection of the angle trisectors there are three
that we will single out – the intersections of adjacent trisectors. In Figure 9.2
the intersection of adjacent trisectors are indicated, additionally, we have
connected them together to form a small triangle in the center of our original
triangle.
A B
Figure 9.2: A triangle is formed whose vertices are the intersections of the
adjacent trisectors of the angles of 4ABC.
Are you ready for the miraculous part? Okay, here goes!
Theorem 9.1.1. The points of intersection of the adjacent trisectors in an

arbitrary triangle 4ABC form the vertices of an equilateral triangle.
In other words, that little blue triangle in Figure 9.2 that kind of looks
like it might be equilateral actually does have all three sides equal to one
another. Furthermore, it doesn’t matter what triangle we start with, if we
do the construction above we’ll get a perfect 60◦ − 60◦ − 60◦ triangle in the
middle!
Sources differ, but it is not clear whether Morley ever proved his theorem.
The first valid proof (according to R. K. Guy in [8] was published in 1909 by
M. Satyanarayana [15]. There are now many other proofs known, for instance
the cut-the-knot website (http://www.cut-the-knot.org/) exposits no less
than nine different proofs. The proof by Satyanarayana used trigonometry.
The proof we’ll look at here is arguably the shortest ever produced and it is
due to John Conway. It is definitely a “Book proof”!

Let us suppose that an arbitrary triangle 4ABC is given. We want to
show that the triangle whose vertices are the intersections of the adjacent tri-
sectors is equilateral – this triangle will be referred to as the Morley triangle.
Let’s also denote by A, B and C the measures of the angles of 4ABC. (This
is what is generally known as an “abuse of notation” – we are intentionally
confounding the vertices (A, B and C) of the triangle with the measure of
the angles at those vertices.) It turns out that it is fairly hard to reason
from our knowledge of what the angles A, B and C are to deduce that the
Morley triangle is equilateral. How does the following plan sound: suppose
we construct a triangle, that definitely does have an equilateral Morley tri-
angle, whose angles also happen to be A, B and C. Such a triangle would be
similar4 to the original triangle 4ABC – if we follow the similarity transform
from the constructed triangle back to 4ABC we will see that their Morley
triangles must coincide; thus if one is equilateral so is the other!
One of the features of Conway’s proof that leads to its great succinctness
and beauty is his introduction of some very nice notation. Since we are
dealing with angle trisectors, let a, b and c be angles such that 3a = A,
3b = B and 3c = C. Furthermore, let a superscript star denote the angle
that is π/3 (or 60◦ if you prefer) greater than a given angle. So, for example,
a? = a + π/3
and
a?? = a + 2π/3.
4
In Geometry, two objects are said to be similar if one can be made to exactly coincide
with the other after a series of rigid translations, rotations and scalings. In other words,
they have the same shape if you allow for differences in scale and are allowed to slide them
around and spin them about as needed.
Now, notice that the sum a + b + c must be π/3. This is an immediate

consequence of A + B + C = π which is true for any triangle in the plane. It
follows that by distributing two stars amongst the three numbers a, b and c
we will come up with three quantities which sum to π. In other words, there
are Euclidean triangles having the following triples as their vertex angles:
(a, b, c?? ) (a, b? , c? )

(a, b?? , c) (a? , b? , c)
(a?? , b, c) (a? , b, c? )
Exercise. What would a triangle whose vertex angles are (0? , 0? , 0? ) be?
In a nutshell, Conway’s proof consists of starting with an equilateral tri-

angle of unit side length, adding appropriately scaled versions of the six
triangles above and ending up with a figure (having an equilateral Morley
triangle) similar to 4ABC. The generic picture is given in Figure 9.3. Be-
fore we can really count this argument as a proof, we need to say a bit more
about what the phrase “appropriately scaled” means. In order to appropri-
ately scale the triangles (the small acute ones) that appear green in Figure 9.3
we have a relatively easy job – just scale them so that the side opposite the
trisected angle has length one; that way they will join perfectly with the
central equilateral triangle.
The triangles (these are the larger obtuse ones) that appear purple in 9.3
are a bit more puzzling. Ostensibly, we have two different jobs to accomplish
– we must scale them so that both of the edges that they will share with green
triangles have the correct lengths. How do we know that this won’t require
two different scaling factors? Conway also developed an elegant argument
that handles this question as well. Consider the purple triangle at the bottom
of the diagram in Figure 9.3 – it has vertex angles (a, b, c?? ). It is possible
to construct triangles similar (via reflections) to the adjacent green triangles
c c c
b?? a? b ?
a??
c? c?
b? a? b
a
??
c b
a
a b
Figure 9.3: Conway’s proof involves putting these pieces together to obtain
a triangle (with an equilateral Morley triangle) that is similar to 4ABC.
(a, b? , c? ) and (a? , b, c? ) inside of triangle (a, b, c?? ). To do this just construct
two lines that go through the top vertex (where the angle c?? is) that cut the
opposite edge at the angle c? in the two possible senses – these two lines will
coincide if it should happen that c? is precisely π/2 but generally there will
be two and it is evident that the two line segments formed have the same
length. We scale the purple triangle so that this common length will be 1.
See Figure 9.4.
Exercise. If it should happen that c? = π/2, what can we say about C?
Of course the other two obtuse triangles can be handled in a similar way.
b ? a?
a c? c? b
Figure 9.4: The scaling factor for the obtuse triangles in Conway’s puzzle
proof is determined so that the segments constructed in there midsts have
unit length.
Exercises — 9.1
1. What value should we get if we sum all of the angles that appear around
one of the interior vertices in the finished diagram? Verify that all three
have the correct sum.
c c c
b?? a? b? a??
c? c?
b ? a?
c?? b
a
a b
a b
2. In this section we talked about similarity. Two figures in the plane

are similar if it is possible to turn one into the other by a sequence of
mappings: a translation, a rotation and a scaling.
Geometric similarity is an equivalence relation. To fix our notation, let
T (x, y) represent a generic translation, R(x, y) a rotation and S(x, y)
a scaling – thus a generic similarity is a function from R2 to R2 that
can be written in the form S(R(T (x, y).
Discuss the three properties an equivalence relation (reflexivity, sym-
metry and transitivity) in terms of geometric similarity.
9.2. FIVE STEPS INTO THE VOID 389
9.2 Five steps into the void

In this section we’ll talk about another Book proof also due to John Conway.
This proof serves as an introduction to a really powerful general technique –
the idea of an invariant. An invariant is some sort of quantity that one can
calculate that itself doesn’t change as other things are changed. Of course
different situations have different invariant quantities.
The setup here is simple and relatively intuitive. We have a bunch of
checkers on a checkerboard – in fact we have an infinite number of checkers,
but not filling up the whole board, they completely fill an infinite half-plane
which we could take to be the set
S = {(x, y) x ∈ Z ∧ y ∈ Z ∧ y ≤ 0}.
See Figure 9.5.

Think of these checkers as an army and the upper half-plane is “enemy
territory.” Our goal is to move one of our soldiers into enemy territory as far
as possible. The problem is that our “soldiers” move the way checkers do,
by jumping over another man (who is then removed from the board). It’s
clear that we can get someone into enemy territory – just take someone in
the second row and jump a guy in the first row. It is also easy enough to
see that it is possible to get a man two steps into enemy territory – we could
bring two adjacent men a single step into enemy territory, have one of them
jump the other and then a man from the front rank can jump over him.
Exercise. The strategy just stated uses 4 men (in the sense that they are
removed from the board – 5 if you count the one who ends up two steps into
enemy territory as well). Find a strategy for moving someone two steps into
enemy territory that is more efficient – that is, involves fewer jumps.
Exercise. Determine the most efficient way to get a man three steps into
Figure 9.5: An infinite number of checkers occupying the integer lattice points
such that y ≤ 0.
enemy territory. An actual checkers board and pieces (or some coins, or
rocks) might come in handy.
We’ll count the man who ends up some number of steps above the x-axis,
as well as all the pieces who get jumped and removed from the board as a
measure of the efficiency of a strategy. If you did the last exercise correctly
you should have found that eight men are the minimum required to get 3
steps into enemy territory. So far, the number of men required to get a given
distance into enemy territory seems to always be a power of 2.
# of steps # of men
1 2
2 4
3 8
As a picture is sometimes literally worth one thousand words, we include

here 3 figures illustrating the moves necessary to put a scout 1, 2 and 3 steps
into the void.
In order to show that 8 men are sufficient to get a scout 3 steps into
enemy territory, we show that it is possible to reproduce the configuration
that can place a man two steps in – shifted up by one unit.
You may be surprised to learn that the pattern of 8 men which are needed
to get someone three steps into the void can be re-created – shifted up by
one unit – using just 12 men. This means that we can get a man 4 steps into
enemy territory using 12 + 8 = 20 men. You were expecting 16 weren’t you?
(I know I was!)
Exercise. Determine how to get a marker 4 steps into the void.
The real surprise is that it is simply impossible to get a man five steps
into enemy territory. So the sequence we’ve been looking at actually goes
2, 4, 8, 20, ∞.
Figure 9.6: One man is sacrificed in order to move a scout one step into
enemy territory.
2
1
Figure 9.7: Three man are sacrificed in order to move a scout two steps into
enemy territory.
1 2 3 4
Figure 9.8: Eight men are needed to get a scout 3 steps into the void.
The proof of this surprising result works by using a fairly simple, but
clever, strategy. We assign a numerical value to a set of men that is dependent
on their positions – then we show that this value never increases when we
make “checker jumping” moves – finally we note that the value assigned to
a man in position (0, 5) is greater than the value of the entire original set of
men (that is, with all the positions in the lower half-plane occupied). This is
a pretty nice strategy, but how exactly are we going to assign these numerical
values?
A man’s value is related to his distance from the point (0, 5) in what is
often called “the taxicab metric.” We don’t use the straight-line distance,
but rather determine the number of blocks we will have to drive in the north-
south direction and in the east-west direction and add them together. The
value of a set of men is the sum of their individual values. Since we need to
deal with the value of the set of men that completely fills the lower half-plane,
we are going to have to have most of these values be pretty tiny! To put it
in a more mature and dignified manner: the infinite sum of the values of the
men in our army must be convergent.
8 7 6 5 6 7 8
9 8 7 6 7 8 9
10 9 8 7 8 9 10
11 10 9 8 9 10 11
Figure 9.9: The taxicab distance to (0, 5).

Exercises — 9.2
1.
9.3 Monge’s circle theorem

9.3. MONGE’S CIRCLE THEOREM 399
Exercises — 9.3
1.
9.4 Pythagorean triples

9.4. PYTHAGOREAN TRIPLES 401
Exercises — 9.4
1.
Bibliography
[1] R. E. Greenwood A. M. Gleason and L. M. Kelly. The William Lowell

Putnam Mathematical Competition Problems & Solutions: 1938-1964.
The Mathematical Association of America, Reissued 2003.
[2] Martin Aigner and Gunter M. Ziegler. Proofs from THE BOOK.
Springer-Verlag, 2nd edition, 2001.
[3] Wikipedia contributors. Cantor-bernstein-schroeder theorem.

Wikipedia, the free encyclopedia. http://en.wikipedia.org/wiki/Cantor-
Bernstein-Schroeder theorem.
[4] Wikipedia contributors. Christian Goldbach. Wikipedia, the free ency-

clopedia. http://en.wikipedia.org/wiki/Goldbach.
[5] Wikipedia contributors. Erdos number. Wikipedia, the free encyclope-

dia. http://en.wikipedia.org/wiki/Erdos number.
[6] Wikipedia contributors. The four color theorem. Wikipedia, the free
encyclopedia. http://en.wikipedia.org/wiki/Four color theorem.
[7] Leonard F. Klosinski Gerald L. Alexanderson and Loren C. Larson. The

William Lowell Putnam Mathematical Competitions Problems & Solu-
tions: 1965 - 1984. The Mathematical Association of America, Reissued
2003.
403
404 BIBLIOGRAPHY
[8] Richard K. Guy. The lighthouse theorem, Morley & Malfatti – a budget
of paradoxes. American Mathematical Monthly, 2007.
[9] Bjorn Poonen Kiran S. Kedlaya and Ravi Vakil. The William Lowell
Putnam Mathematical Competition 1985-2000: Problems Solutions, and
Commentary. The Mathematical Association of America, 2002.
[10] C. W. H. Lam. The search for a finite projective plane of order 10.
http://www.www.cecm.sfu.ca/organics/papers/lam/paper/html/paper.html.
[11] Saunders MacLane. Categories for the Working Mathematician.

Springer-Verlag, 2nd edition, 1998.
[12] John J. O’Connor and Edmund F. Robertson. http://www-

history.mcs.st-andrews.ac.uk/history/index.html. The MacTutor His-
tory of Mathematics archive.
[13] Stanislaw Radziszowski. Small ramsey numbers.

http://www.combinatorics.org/Surveys/ds1.pdf.
[14] Gian-Carlo Rota. Indiscrete Thoughts. Birkhäuser, 1997.
[15] M. Satyanarayana. none given. Math. Quest. Educ. Times (New Series),
1909.
[16] D. J. Struik. A Source Book in Mathematics, 1200-1800. Princeton

University Press, 1986.
[17] Alfred North Whitehead and Bertrand Russell. Principia Mathematica.

Cambridge University Press, 1910.
Index
absorption, 84, 85 Boole, George, 61

addition rule, 289 bottom, in a poset, 259
algorithm, 35 bound variables, 93
alphanumeric, 304
Cantor’s Snake, 346
and gates, 64
Cartesian product, 228
antecedent, 70
Category theory, 159
anti-symmetry, 235
ceiling function, 26
antichain, 258
chain, 258
Appel, Kenneth, 145
characteristic function, 280
arithmetic mean, 129
circular reasoning, 89
arithmetic sequence, 129
Cohen, Paul, 377
arithmetic-geometric mean inequality,
combination, 293
129
commutative law, 62, 77, 85
ASCII, 305
complement, 172
associative law, 62, 78, 85
complementarity law, 85
atomic concepts, 57
complementarity laws, 83
begging the question, 89 complex numbers, 5
biconditional, 72 component-wise operations, 6
bijection, 268 composite, 13
binary relation, 51 composition, of functions, 232
binomial, 325 composition, of relations, 232
binomial coefficients, 28 compound sentence, 59
Blaise Pascal, 29 conditional statement, 70
405
406 INDEX
congruence, 27 distributive law, 62, 80, 85

conjunction, 59 divisibility, 25
conjunctive simplification, 102 division algorithm, 38
consequent, 70 domain, 54
constructive dilemma, 104 domination law, 85
contradiction, 83 doubly-even, 32
contrapositive, 73 duck, flaming, 112
converse, 72
converse error, 111 empty set, 163
countably infinite, 341 empty sum, 204
counterexample, 139 equinumerous, 337
cover, in a poset, 259 equivalence class, 244
Crocodile’s dilemma, 103 equivalence relation, 244

Eratosthenes of Cyrene, 12
deduction, 100 Erdos number, 379
deductive argument, 101 Erdos, Paul, 379
degree, 308 error detecting code, 307
DeMorgan’s law, 85 Euclidean algorithm, 41
DeMorgan’s laws, 82 Euler, Leonhard, 95, 146, 307
denumerable, 341 Eulerian circuit, 316
Descartes, Rene, 229 Eulerian path, 316
destructive dilemma, 104, 110 evenness, 128
diagonal map, 280
exclusive or, 62
difference (of sets), 175
existential quantification, 19
digraph, 158, 228, 249
direct proofs, 123 factorials, 30
directed graph, 228 Fermat numbers, 94
disjunction, 60 Fermat’s last theorem, 214
disjunctive normal form, 66, 185 Fermat’s little theorem, 214
disjunctive syllogism, 103 Fermat, Pierre de, 94
INDEX 407
Fibonacci numbers, 213 identity law, 85

finite sequence, 288 identity laws, 83
floor function, 26 iff, 72
flowchart, 36 image, of a set, 272
form (of an argument), 111 imaginary part, 6
four color theorem, 144 inclusive or, 63
Incompleteness Theorem, 58
Gödel, Kurt, 58, 160
indicator function, 281
general position, 184
indirect proof, 133
generalizing from the generic partic-
induction, 57, 193
ular, 123
inductive argument, 100
geometric mean, 129
inductive hypothesis, 196
geometric sequence, 129
infinitude of the primes, 133
Goldbach’s conjecture, 18, 146
injection, 268
Goldbach, Christian, 146
integers, 2
graded poset, 257
intersection, 171
graph, 249
inverse, 72
graph isomorphism, 249
inverse error, 111
greatest common divisor, 39
inverse image, of a set, 272
greatest common divisor, gcd, 154
inverse relation, 234, 267
greatest element, in a poset, 259
inverse, of a relation, 232
Haken, Wolfgang, 145 invertible function, 55
Hasse diagrams, 255 Iverson bracket, 281
Hasse, Helmut, 255
Hollerith card, 304 Jordan curve, 179
hypercube, 261 Jordan curve theorem, 179
hypothetical syllogism, 103 Jordan, Camille, 179
idempotence, 85 Königsberg, 146, 307

idempotent, 84 Kaliningrad, 307
408 INDEX
Kronecker delta, 282 octal representation, 32

Kronecker, Leopold, 1 open sentence, 92
operator, 263
laws of logical equivalence, 77 or gates, 64
least common multiple, 39 ordering relation, 253
least element, in a poset, 259
lemmas, 47 parallel connection, 62
parity check code, 307
lexicographic order, 234
partial order, 253
Lihua, Ma, 194
partially ordered set, 255
logic gates, 64
partition, 245
logical equivalence, 76
Pascal’s triangle, 29, 325
Luxembourg, 145
Peano axioms, 193
magic square, 314 Peirce arrow, 69
maximal element, in a poset, 259 permutation, 290
minimal element, in a poset, 259 Petersen graph, 251
modulus, of a complex number, 7 pigeonhole, 319
modus ponens, 102 pigeonhole principl, 319
modus tollens, 102 pigeonhole principle, strong form, 322
Morley triangle, 384 PIN, 294
multiplication rule, 290 place notation, 23
multiset, 162 polynomial multiplication, 120
poset, 255
NAND, 69 power set, 164
natural numbers, 1 predicate variable, 60
negation, 60 Pregel, Pregolya, 308
Newton, Isaac, 4 premise, 101
noneg, 7 prime factorization, 14
NOR, 69 prime numbers, 12
not gates, 64 projection, 280
INDEX 409
projective plane of order 10, 144 restriction, of a function, 276

proof by cases, 144 right inverse, 277
proof by contradiction, 133 rules of inference, 102, 104, 106
proof by contraposition, 97, 134 rules of replacement, 85, 104
proof by exhaustion, 144
Scheffer stroke, 69
proper subset, 166
sentence, 58
Properties of relations, 236
sequence, 288
pseudocode, 36
series connection, 62
punch card, 304
set theoretic equalities, 174
Pythagoras, 4
set-builder notation, 2
Pythagorean triple, 138
sieve of Eratosthenes, 13
quantification, 92 singleton set, 163
quod erat demonstrandum, 127 Sophie Germain prime, 99
quotient structure, 245 soundness (of an argument), 109
square-free part, of an integer, 246
radical, of an integer, 246
statement, 58
Ramsey number, 144
subset, 166
range, 54
successor, 259
Rational approximation, 49
superset, 166
rationals, 2
syllogism, 103
real part, 6
symmetric difference, 175
reals, 4
symmetry, 235
recognizers, 66
reductio ad absurdam, 46 tautology, 83
reflexivity, 235 ternary relation, 51
relations, 50 tetromino, 317
relative primality, 45 TFAE, 12
repeated division algorithm, 32 top, in a poset, 259
repetition number, 162 total order, 253
410 INDEX
transistor, 62
transitivity, 235
triangular numbers, 212
trichotomy, 7
trichotomy property, 150
truth table, 60
Twin Prime conjecture, 18
two-column proof, 88
uncountable, 341
union, 171
unique existence, 99, 154
universal conditional statement, 118
universal quantification, 19
universal set, 161
universe of discourse, 19, 161, 180
vacuous truth, 71
valid argument form, 109
vampire number, 150, 153
Venn diagram, 179
weak Goldbach conjecture, 146

weasels, ice, 117
well-ordering principle, 152, 158
William Lowell Putnam Mathematics
Competition, 147
winding map, 278
Yahtzee, 289
Z-module, 155

A Gentle Introduction To The Art of Mathematics

Uploaded by

Copyright:

Available Formats

A Gentle Introduction To The Art of Mathematics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Gentle Introduction To The Art of Mathematics

Uploaded by

Copyright:

Available Formats

A Gentle Introduction to the Art of

Southern Connecticut State University

1 Introduction and notation 1

2 Logic and quantifiers 57

2.6 Deductive reasoning and Argument forms . . . . . . . . . . . . 100

3 Proof techniques I 117

5 Proof techniques II — Induction 193

6 Relations and functions 225

6.6 Special functions . . . . . . . . . . . . . . . . . . . . . . . . . 276

7 Proof techniques III — Combinatorics 287

9 Proof techniques IV — Magic 379

1.1 The sieve of Eratosthenes. . . . . . . . . . . . . . . . . . . . . 13

2.1 A schematic representation of a transistor. . . . . . . . . . . . 62

3.1 A four-color map. . . . . . . . . . . . . . . . . . . . . . . . . . 145

6.1 An example of a relation. . . . . . . . . . . . . . . . . . . . . . 226

6.7 Hasse diagram of divisors of 72. . . . . . . . . . . . . . . . . . 258

7.1 Full houses in Yahtzee. . . . . . . . . . . . . . . . . . . . . . . 290

8.1 Cantor’s snake. . . . . . . . . . . . . . . . . . . . . . . . . . . 347

9.1 The setup for Morley’s Miracle. . . . . . . . . . . . . . . . . . 382

2.1 Converse, inverse and contrapositive. . . . . . . . . . . . . . . 73

3.1 The definitions of elementary number theory restated. . . . . . 122

4.1 Basic set theoretic equalities. . . . . . . . . . . . . . . . . . . . 174

6.1 Properties of relations. . . . . . . . . . . . . . . . . . . . . . . 236

1. Mathematics is alive! Math is not just something to be studied from

2. Math is not scary! There is an incredibly terse and compact language

4. You have two cerebral hemispheres – use them both! In perhaps no

onymous with left-brain function. In truth, doing mathematics is often

Also, there are amusing quotations at the start of every chapter.

and language common to all Mathematicians – although there are certainly

Introduction and notation

1.1 Basic sets

This is a good time to start building a “glossary” – a translation lexicon

The rational numbers are defined to be the set of all

a∈Z and b∈Z

a is an element of the integers and b is an element of the integers

and b is nonzero. (the final curly brace is silent)

It is quite apparent that the mathematical notation represents a huge

C = {a + bi a ∈ R and b ∈ R and i2 = −1}

Translating this bit of mathematics into English we get:

The complex numbers are defined to be the set of all

expressions of the form a plus b times i such that

a∈R and b∈R

a is an element of the reals and b is an element of the reals

and i has the property that its square is negative one.

We sometimes denote a complex number using a single variable (by con-

Complex numbers are added and multiplied as if they were binomials

Presumably, we could also use “nonpos” as a superscript to indicate non-

2. Write the set Z of integers using a singly infinite listing.

3. Give a description of the set of rational numbers whose decimal ex-

6. Give an argument as to why the product of two rational numbers is

7. Perform the following computations with complex numbers

(a) (4 + 3i) − (3 + 2i)

(d) (2 − 3i) · (3 − 2i)

8. The conjugate of a complex number is denoted with a superscript star,

1.2 Definitions: Prime numbers

had better agree with one another!

Definition. A prime number is a positive integer, greater than 1, whose only

Definition. A prime is a quantity p such that whenever p is a factor of some

Exercise. The number 1 is not considered to be a prime. Does 1 satisfy the