Classnotes Discrete Math
Uploaded by
bszw5eys5vClassnotes Discrete Math
Uploaded by
bszw5eys5vCLASS NOTES FOR DISCRETE MATHEMATICS
NOTE ADDED 14 June 2008
These class notes were used for fifteen years in a discrete math class taught at
Case Western Reserve University until I retired in 1999. I am making them available
as
a resource to anyone who wishes to use them. They may be copied and distributed for
educational use, provided that the recipients are charged only the copying costs.
If I were revising these notes today I would make some sizeable changes. The
most important would be to reformulate the definition of division on page 4 to
require that
the divisor be nonzero. The result would change the statement “0 divides 0” from
true to
false, and would affect the answers to a number of exercises.
I will be glad to receive comments and suggestions at [email protected].
Interested readers may wish to look at my other books and websites concerned with
teaching:
The Abstract Math website
Astounding Math Stories
The Handbook of Mathematical Discourse
Charles Wells
Charles (at) abstractmath.org
DISCRETE
MATHEMATICS
Charles Wells
June 22, 1999
Supported in part by the Fund for the Improvement of Post-Secondary Education
(Grant
GCO8730463)
Charles Wells
Department of Mathematics
Case Western Reserve University
10900 Euclid Avenue
Cleveland, OH 44106-7058, USA
Email: [email protected]
Home Page: http://www.cwru.edu/artsci/math/wells/home.html
Copyright ©1999 by Charles Frederick Wells
Contents
39
Functions
56
40
The graph of a function
61
How to read these notes
41
Some important types of functions
63
2
Integers
42
Anonymous notation for functions
64
Definitions and proofs in mathematics
43
Predicates determine functions
65
Division
44
Sets of functions
66
5
More about proofs
45
Binary operations
67
Primes
10
46
Fixes
68
Rational numbers
11
47
More about binary operations
69
Real numbers
12
48
Associativity
70
Decimal representation of real numbers
12
49
Commutativity
71
10
Decimal representation of rational num-
50
Identities
72
bers
14
51
Relations
73
11
Propositions
15
52
Relations on a single set
75
12
Predicates
16
53
Relations and functions
75
13
Universally true
19
54
Operations on relations
77
14
Logical Connectives
21
55
Reflexive relations
77
15
Rules of Inference
24
56
Symmetric relations
78
16
Sets
25
57
Antisymmetric relations
79
17
List notation for sets
26
58
Transitive relations
80
18
Setbuilder notation
27
59
Irreflexive relations
81
19
Variations on setbuilder notation
29
60
Quotient and remainder
82
20
Sets of real numbers
31
61
Trunc and Floor
86
21
A specification for sets
32
62
Unique factorization for integers
87
22
The empty set
33
63
The GCD
88
23
Singleton sets
34
64
Properties of the GCD
90
24
Russell’s Paradox
35
65
Euclid’s Algorithm
92
25
Implication
35
66
Bases for representing integers
93
26
Vacuous truth
37
67
Algorithms and bases
97
27
How implications are worded
38
68
Computing integers to different bases
99
28
Modus Ponens
40
69
The DeMorgan Laws
102
29
Equivalence
40
70
Propositional forms
104
30
Statements related to an implication
42
71
Tautologies
105
31
Subsets and inclusion
43
72
Contradictions
107
32
The powerset of a set
46
73
Lists of tautologies
107
33
Union and intersection
47
74
The tautology theorem
110
34
The universal set and complements
48
75
Quantifiers
112
35
Ordered pairs
49
76
Variables and quantifiers
114
36
Tuples
50
77
Order of quantifiers
115
37
Cartesian Products
52
78
Negating quantifiers
116
38
Extensions of predicates
79
Reading and writing quantified state-
with more than one variable
55
ments
117
IV
80
Proving implications: the Direct Method 119
120 The quotient of a function
184
81
Proving implications: the Contrapositive
121 The fundamental bijection theorem
186
Method
120
122 Elementary facts about finite sets
and
82
Fallacies connected with implication
121
functions
187
83
Proving equivalences
122
123 The Pigeonhole Principle
189
84
Multiple equivalences
123
124 Recurrence relations in counting
189
85
Uniqueness theorems
124
125 The number of subsets of a set
190
86
Proof by Contradiction
125
126 Composition of relations
195
87
Bezout’s Lemma
127
127 Closures
197
88
A constructive proof of Bezout’s Lemma
128
128 Closures as intersections
198
89
The image of a function
131
129 Equivalence relations
200
90
The image of a subset of the domain
132
130 Congruence
201
91
Inverse images
132
131 The kernel equivalence of a function
203
92
Surjectivity
133
132 Equivalence relations and partitions
204
93
Injectivity
134
133 Partitions give equivalence relations
205
94
Bijectivity
136
134 Orderings
206
95
Permutations
137
135 Total orderings
208
96
Restrictions and extensions
137
136 Preorders
209
97
Tuples as functions
138
137 Hasse diagrams
210
98
Functional composition
140
138 Lexical ordering
211
99
Idempotent functions
143
139 Canonical ordering
212
100
Commutative diagrams
144
140 Upper and lower bounds
212
101
Inverses of functions
146
141 Suprema
213
102
Notation for sums and products
150
142 Lattices
215
103
Mathematical induction
151
143 Algebraic properties of lattices
216
104
Least counterexamples
154
144 Directed graphs
218
105
Recursive definition of functions
157
145 Miscellaneous topics about digraphs
220
106
Inductive and recursive
159
146 Simple digraphs
221
107
Functions with more than one starting
147 Isomorphisms
223
point
160
148 The adjacency matrix of a digraph
224
108
Functions of several variables
163
149 Paths and circuits
225
109
Lists
164
150 Matrix addition and multiplication
227
110
Strings
167
151 Directed walks and matrices
228
111
Formal languages
169
152 Undirected graphs
230
112
Families of sets
171
153 Special types of graphs
233
113
Finite sets
173
154 Subgraphs
234
114
Multiplication of Choices
174
155 Isomorphisms
234
115
Counting with set operations
176
156 Connectivity in graphs
236
116
The Principle of Inclusion and Exclusion 178
157 Special types of circuits
237
117
Partitions
180
158 Planar graphs
239
118
Counting with partitions
182
159 Graph coloring
241
119
The class function
183
Answers to Selected Exercises
243
Bibliography
253
Index of Symbols
260
Index
254
About these notes
These class notes are for MATH 304, Fall semester, 1999. Previous versions are not
usable because the text has been rewritten.
It would be a good idea to leaf through this copy to see that all the pages are
there and correctly printed.
Labeled paragraphs This text is written in an innovative style intended to make
the logical status of each part of the text as clear as possible. Each part is
marked
with labels such as “Theorem” , “Remark” , “Example” , and so on that describe
the intent of that part of the text. These descriptions are discussed in more
detail
in Chapter 1.
Exercises The key to learning the mathematics presented in these notes is in doing
all the exercises. Many of them are answered in the back; when that is so, the text
gives you the page the answer is on. You should certainly attempt every exercise
that has an answer and as many of the others that you have time for.
Exercises marked “(discussion)” may be open-ended or there may be disagree-
ment as to the answer. Exercises marked “(Mathematica)” either require Mathema-
tica or will be much easier to do using Mathematica. A few problems that require
knowledge of first-year calculus are marked “(calculus)”.
Indexes On each page there is a computer-generated index of the words that occur
on that page that are defined or discussed somewhere in the text. In addition,
there
is a complete computer-generated index on page 254. In some cases the complete
index has entries for later pages where significant additional information is given
for
the word.
There is also an index of symbols (page 260).
Bibliography The bibliography is on page 253. References to books in the bib-
liography are written like this: [Hofstadter, 1979]. Suggestions for other books to
include would be welcome.
Acknowledgments A grant from the Fund for the Improvement of Post- Secondary
Education supported the development of these class notes. A grant from the Con-
solidated Natural Gas Corporation supported the development of the Mathematica
package dmfuncs.m and the concomitant revisions to these notes.
I would like to thank Michael Barr, Richard Charnigo, Otornar Hajek, Ernest
Leach, Marshall Leitman and Arthur Obrock for finding mistakes and making many
helpful suggestions.
I would appreciate being notified of any errors or ambiguities. You may contact
me at [email protected].
Charles Wells
1
1. How to read these notes proposition 15
specification 2
This text introduces you to the subject matter of discrete mathematics; it includes
a theorem 2
substantial portion of the basic language of mathematics used by all
mathematicians,
as well as many topics that have turned out to be useful in computer science.
In addition, this text constitutes a brief introduction to mathematical reasoning.
This may very well be the first mathematics course in which you are expected
to produce a substantial amount of correct mathematical reasoning as well as to
compute answers to problems.
Most important concepts can be visualized in more than one way, and it is vital
to be able to conceive of these ideas in some of the ways that mathematicians and
computer scientists conceive of them. There is discussion in the text about most of
the concepts to help you in doing this. The problem is that this type of discussion
in general cannot be cited in proofs ; the steps of a proof are allowed to depend
only
on definitions, and previously proved theorems. That is why the text has labels
that
distinguish the logical status of each part.
What follows is a brief glossary that describes many of the types of prose that
occur in this book.
1.1 Glossary
Corollary A corollary to a theorem P is another theorem that follows easily
from P .
Definition Provides a definition of one or more concepts. Every statement to be
proved should be rewritten to eliminate terms that have definitions. This is
discussed
in detail in Chapter 3.
Not all concepts are defined in this text. Basic ideas such as integers and real
numbers are described but not defined; we depend on your familiarity with them
from earlier courses. We give a specification for some of these.
Example An example of a concept is a mathematical object that fits the definition
of the concept. Thus in Definition 4.1, we define “divides” for integers, and then
Example 4.1.1 we observe that 3 and 6 form an example of “divides” (3 divides
6 ).
For study purposes it is worthwhile to verify that each example does fit the
definition. This is usually easy.
A few examples are actually non-examples: mathematical objects that you might
think are examples of the concept but in fact are not.
Fact A fact is a precise statement about mathematics that is correct. A fact is
a theorem, but one that is easy to verify and not necessarily very important. The
statements marked “fact” in this text are usually immediately obvious from the
definitions.
This usage is peculiar to these notes. Many texts would mark what we call facts
as “propositions”, but here the word “proposition” is used in a slightly different
way.
corollary 1
fact 1
lemma 2
proof 4
theorem 2
usage 2
warning 2
Lemma A lemma is a theorem that is regarded as a tool to be used in proving
other theorems rather than as interesting in its own right. In fact, some theorems
are traditionally called lemmas that in fact are now perceived as quite important.
Method A paragraph marked “Method” provides a method for calculating some
object or for determining the truth of a certain type of statement.
Proof A mathematical proof of a statement is a sequence of closely reasoned claims
about mathematical objects (numbers, sets, functions and so on) with each claim
depending on the given assumptions of the statement to be proved, on known def-
initions and previously proved theorems (including lemmas, corollaries and facts),
and on the previous statements in the proof.
Proofs are discussed in more detail in Chapters 3,5, and in a sequence of chapters
beginning with Chapter 80. Particular proof techniques are described in smaller
sections throughout the text.
“Show” is another word for “prove”. (Not all math texts use the word “show”
in this way.)
Remark A remark is a statement that provides some additional information about
a concept. It may describe how to think about the concept, point out some aspects
that follow (or don’t follow!) from the definition that the reader on first reading
might miss, or give further information about the concept.
Note: As of this revision (June 22, 1999) there are some statements called
“remark” that perhaps should be called “fact”, “usage” or “warning”. The author
would appreciate being told of any mislabeled statement.
Specification A specification of a mathematical concept describes some basic
properties of the concept but does not pin down the concept in terms of other
concepts the way a definition does.
Theorem A theorem is a precise statement about mathematics that has been
proved (proved somewhere — not always in this text). Theorems may be quoted as
reasons in a proof, unless of course the statement to be proved is the theorem
being
quoted!
Corollaries, lemmas and facts are all theorems. Statements marked “Theorem”
are so marked because they are important. Particularly important theorems are
enclosed in a box.
Usage A paragraph marked “Usage” describes the way some terminology or sym-
bolism is used in mathematical practice. Sometimes usage varies from text to text
(example: Section 2.2.1) and in many cases, the usage of a term or symbol in mathe-
matical texts is different, often in subtle ways, from its usage in other texts
(example:
Section 14.1.2).
Warning A paragraph marked “Warning” tells you about a situation that has
often (in my experience) misled students.
2. Integers
2.1 Specification: integer
An integer is any whole number. An integer can be zero, greater than
zero or less than zero.
2.1.1 Remark Note that this is not a formal definition; it is assumed that you
are familiar with the integers and their basic properties.
2.1.2 Example —3, 0, 55 and one million are integers.
definition 4
integer 3
natural number 3
negative 3
nonnegative integer 3
nonnegative 3
positive integer 3
positive 3
specification 2
theorem 2
usage 2
2.2 Definition: Properties of integers
For any integer n:
a) n is positive if n > 0 .
b) n is negative if n < 0.
c) n is nonnegative if n > 0.
d) An integer n is a natural number if n is nonnegative.
2.2.1 Usage
a) A few authors define zero to be both positive and negative, but that is not
common mathematical practice in the USA.
b) In pure mathematics the phrase “natural number” historically meant positive
integer, but the meaning “nonnegative integer” used in this book has become
more common in recent years.
The following theorem records some familiar facts.
2.3 Theorem
If m and n are integers, then so are m + n, m — n and mn . If m and
n are not both zero and n is nonnegative, then m n is also an integer.
2.3.1 Remarks
a) In this text, 0° is undefined.
b) Observe that m n may not be an integer if n is negative.
2.3.2 Exercise Describe precisely all integers m and n for which m n is an integer.
Note that Theorem 2.3 does not quite answer this question!
boldface 4
definition 4
divide 4
integer 3
negative integer 3
nonnegative integer 3
positive integer 3
3. Definitions and proofs in mathematics
Each Definition in this text gives the word or phrase being defined in boldface.
Each definition gives a precise description of what is required for an object to
fit
that definition. The only way one can verify for sure that a statement about a
defined object is correct is to give a proof that it is correct based on the
definition
or on previous facts proved using the definition.
Definition 2.2 gives a precise meaning to the words “positive”, “negative”, “non-
negative” and “natural number”. Any question about whether a given integer is
positive or negative or is a natural number must be answered by checking this defi-
nition.
Referring to the definition in trying to understand a concept is the first of many
methods which are used throughout the book. We will give such methods formal
status, like this:
3.1.1 Method
To prove that a statement involving a concept is true, begin by using
the definition of the concept to rewrite the statement.
3.1.2 Example The statement “0 is positive” is false. This claim can be justified
by rewriting the statement using Definition 2.2: “0 > 0”. Since this last statement
is false, 0 is not positive.
3.1.3 Remark The preceding example illustrates the use of Method 3.1.1: I jus-
tified the claim that “0 is positive” is false by using the definition of
“positive”.
3.1.4 Example It also follows from Definition 2.2 that 0 is not negative (because
the statement 0 < 0 is false), but it is nonnegative (because the statement 0 > 0
is
true) .
3.1.5 Exercise Is —(—3) positive? (Answer on page 243.)
4. Division
4.1 Definition: division
An integer n divides an integer rn if there is an integer q for which
m = qn. The symbol for “divides” is a vertical line: n\m means n
divides m.
4.1.1 Example Because 6 = 2 x 3, it is true that 3 | 6. It is also true that —3 |
6,
since 6 = (—2) x (—3) , but it is not true that 4 | 14 since there is no integer q
for
which 14 = 4 q. There is of course a fraction q = 14/4 for which 14 = 4 q, but 14/4
is not an integer.
5
definition 4
divide 4
divisor 5
even 5
existential state-
ment 5, 113
factor 5
integer 3
odd 5
usage 2
4.1.6 Exercise Find all the factors of 7, 24, 26 and 111.
4.1.7 Remarks
a) Warning: Don’t confuse the vertical line a verb meaning “divides”, with
the slanting line “/” used in fractions. The expression “3 | 6” is a sentence, but
the expression “6/3” is the name of a number, and does not form a complete
sentence in itself.
b) Warning: Definition 4.1 of “divides” requires that the numbers involved be
integers. So it doesn’t make sense in general to talk about one real number
dividing another. It is tempting, for example, to say that 2 divides 27 t, but
according to the definition given here, that statement is meaningless.
c) Definition 4.1 does not say that there is only one integer q for which m = qn.
However, it is true that if n is nonzero then there is only one such q , because
then q = m/n. On the other hand, for example 0 = 5- 0 = 42-0 so 0 | 0 and
there is more than one q proving that fact.
d) Definition 4.1 says that m\n if an integer q exists that satisfies a certain
property. A statement that asserts the existence of an object with a property
is called an existential statement. Such statements are discussed in more
detail on page 113.
4.1.8 Example According to the definition, 0 divides itself, since 0 = 0x0. On
the other hand, 0 divides no other integer, since if m/0, then there is no integer
q for which m = q x 0 .
4.1.2 Exercise Does 13 | 52 ? (Answer on page 243.)
4.1.3 Exercise Does — 37 | 111 ?
4.1.4 Usage If n divides m, one also says that n is a factor of m or that n is
a divisor of m.
4.1.5 Worked Exercise Find all the factors of 0, 1, 10 and 30.
Answer Number Factors
0 every integer
1 - 1 , 1
10 -1, -2, -5, -10, 1, 2, 5, 10
30 -1, -2, -3, -5, -6, -10, -15, -30, 1, 2, 3, 5, 6, 10, 15, 30
4.1.9 Usage Many authors add the requirement that n/0 to Definition 4.1,
which has the effect of making the statement 0 | 0 meaningless.
4.1.10 Exercise Find all the integers m for which m | 2. (Answer on page 243.)
4.2 Definition: even and odd
An integer n is even if 2 | n . An odd integer is an integer that is not
even.
4.2.1 Example —12 is even, because —12 = (—6) x 2, and so 2 | —12.
definition 4
divide 4
division 4
integer 3
proof 4
theorem 2
5. More about proofs
We will state and prove some simple theorems about division as an illustration of
some techniques of proof (Methods 5.1.2 and 5.3.3 below.)
5.1 Theorem
Every integer divides itself.
Proof Let m be any integer. We must prove that m\ m. By Definition 4.1, that
means we must find an integer q for which m = gm. By first grade arithmetic, we
can use q= 1 .
5.1.1 How to write a proof (1) In the preceding proof, we start with what is
given (an arbitrary integer to ) , we write down what must be proved (that m\m),
we apply the definition (so we must find an integer q for which to = qrn ) , and we
then write down how to accomplish our goal (which is one step in this simple proof
- let q= 1 ) .
We will continue this discussion in Section 5.3.7.
The proof of Theorem 5.1 also illustrates a method:
5.1.2 Method: Universal Generalization
To prove a statement of the form “Every x with property P has property
Q" , begin by assuming you have an x with property P and prove without
assuming anything special about x (other than its given properties) that
it has property Q.
5.1.3 Example Theorem 5.1 asked us to prove that every integer divides itself.
Property P is that of being an integer and property Q is that of dividing itself.
So we began the proof by assuming m is an integer. (Note that we chose a name,
to, for the integer. Sometimes the theorem to be proved gives you a name; see
for example Theorem 5.4 on page 8.) The proof then proceeds without assuming
anything special about to. It would have been wrong, for example, to say
something
like “Assume to. = 5” because then you would have proved the theorem only for
5.
5.2 Theorem
Every integer divides 0.
Proof Let to be an integer (Method 5.1.2!). By Definition 4.1, we must find an
integer q for which 0 = qm. By first grade arithmetic, we can use q = 0.
5.2.1 Remark Theorem 5.2 may have surprised you. You can even find texts in
which the integer q in the definition of division is required to be unique. For
those
texts, it is false that every integer divides 0.
This illustrates two important points:
a) The definition of a mathematical concept determines the truth of every
state-
ment about that concept. Your intuition and experience don’t count in deter-
mining the mathematical truth of a statement. Of course they do count in
being able to do mathematics effectively!
b) There is no agency that standardizes mathematical terminology. (There are
such agencies for physics and chemistry.)
5.3 Theorem
1 divides every integer.
divide 4
factor 5
integer 3
proof 4
theorem 2
Proof Let m be any integer. By Definition 4.1, we must find an integer q for
which m = q T . By first grade arithmetic, we can use q = m.
5.3.1 Exercise Prove that if m\n and a and b are nonnegative integers such
that a <b, then m a \ n b .
5.3.2 Worked Exercise Prove that 42 is a factor of itself.
Proof Theorem 5.1 says that every integer is a factor of itself. Since 42 is an
integer, it is a factor of itself.
This worked exercise uses another proof method:
5.3.3 Method: Universal Instantiation
If a theorem says that a certain statement is true of every object of a
certain type, and c is an object of that type, then the statement is true
of c.
5.3.4 Example In Example 5.3.2, the theorem was Theorem 5.1, the type of
object was “integer”, and c was 42.
5.3.5 Remark Make sure you understand the difference between Method 5.1.2
and Method 5.3.3.
5.3.6 Worked Exercise Prove that 0 is even.
Answer Bu definition of even, we must show that 2 | 0. By Theorem 5.15.2, every
integer divides 0. Hence 2 divides 0 (Method 5.3.3).
5.3.7 How to write a proof (2) Worked Exercise 5.3.8 below illustrates a more
complicated proof. In writing a proof you should normally include all these
steps:
PS.l Write down what is given , and translate it according to the definitions
of the
terms involved in the statement of what is given. This translation may involve
naming some of the mathematical objects mentioned in the statement to be
proved.
PS. 2 Write down what is to be proved, and translate it according to the
definitions
of the terms involved.
PS. 3 Carry out some reasoning that, beginning with what is given, deduces what
is
to be proved.
The third step can be quite long. In some very simple proofs, steps PS-1 and
PS-2
may be trivial. For example, Theorem 5.3 is a statement about every integer. So
for
step PS-1, one merely names an arbitrary integer: “Let m be any integer.” Even,
here, however, we have named what we will be talking about.
Another very important aspect of proofs is that the logical status of every
state-
ment should be clear. Each statement is either:
divide 4
integer 3
nonnegative integer 3
positive integer 3
proof 4
theorem 2
universal instantia-
tion 7
usage 2
a) Given by the hypothesis of the theorem.
b) A statement of what one would like to prove (a goal). Complicated proofs
will
have intermediate goals on the way to the final goal.
c) A statement that has been deduced from preceding known statements. For
each of these, a reason must be given, for example “Universal Instantiation”
or “high school algebra” .
5.3.8 Worked Exercise Prove that any two nonnegative integers which divide
each other are the same.
Answer First, we follow PS-1 and write down what we are given and translate it
according to the definition of the words involved (“divides” in this case):
Assume
we are given integers m and n. Suppose m\n and n\m. By Definition 4.1, the
first statement means that for some q, n = qm .. The second statement means
that
for some q' , m = q'n. Now we have written and translated what we are given.
PS-2: We must prove that m = n. (This translates the phrase “are the same”
using the names we have given the integers.)
PS-3: We put these statements that we have assumed together by simple
algebra:
rn = q'n = q'qm . Now we have two cases: either m = 0 or m/ 0.
a) If m = 0 , then n = qm = g x 0 = 0 , so m = n.
b) If m / 0, then also n / 0, since m = q'n. Then the fact that m = q'n =
q'qm
means that we can cancel the m (because it is nonzero!) to get qq' = 1 .
This
means either q = q' = 1 , so m = n, or q = q' = — 1 , so m = —n. But the
latter
case is impossible since m and n are both positive. So the only possibility
that is left is that m = n.
We give another illustration of writing a proof by rewriting what is given
and what
is to be proved using the definitions by proving this proposition:
5.4 Theorem
For all integers k, m and n, if k\m and k\n then k\m + n.
Proof What we are given is that k \ m and k\n. If we rewrite these
statements
using Definition 4.1, we get that there are integers q and q' for which m =
qk and
n = q'k. What we want to show, rewritten using the definition, is that
there is an
integer q" for which m + n = q"k . Putting the hypotheses together gives
m + n = qk + q'k = (q + q')k
so we can set q" = q + q' to prove the theorem.
5.4.1 Usage In the preceding paragraph, I follow common mathematical
practice
in putting primes on a variable like q or r in order to indicate another
variable q'
of the same type. This prime has nothing to do with the concept of
derivative used
in the calculus.
5.4.2 Existential Bigamy In the proof of Theorem 5.4, we were given that
k \ m
and k\n. By using the definition of division, we concluded that there are
integers
q and q' for which m = qk and n = q'k. It is a common mistake called
existential
bigamy to conclude that there is one integer q for which m = qk and n = qk.
Consider that the phrase “Thurza is married” by definition means that there
is
a person P to whom Thurza is married. If you made the mistake just
described
you would assume that if Amy and Thurza were both married, then they would
be
married to the same person. That is why it is called “existential bigamy” .
Mrs. Thurza Golightly White was the author’s great great grandmother, and
Mrs. Amy
Golightly Walker was her sister. They were very definitely married to
different people.
5.5 Exercise set
In problems 5.5.1 through 5.5.5, you are asked to prove certain statements
about
integers and division. Your proofs should involve only integers — no
fractions should
appear. This will help insure that your proof is based on the definition of
division
and not on facts about division you learned in high school. As I mentioned
before,
you may use algebraic facts you learned in high school, such as that fact
that for
any integers, a(b + c) = ab + ac.
5.5.1 Exercise Prove that 37 | 333. (Answer on page 243.)
5.5.2 Exercise Prove that if n > 0, then any nonnegative integer less than
n
which is divisible by n must be 0. (Answer on page 243.)
5.5.3 Exercise Prove that if k is an integer which every integer divides,
then
k = 0.
5.5.4 Exercise Prove that if k is an integer which divides every integer,
then
k = 1 or k = — 1 .
5.5.5 Exercise Prove that if k \ m and m \ n then k\n.
5.6 Factors in Mathematica
The DmFuncs package contains the function DividesQ [k,n] . It returns True
if k | n and False otherwise. For example, DividesQ [3 , 12] returns True
but
DividesQ [5 , 12] returns False.
You can get a list of all the positive factors of n by typing
AllFactors [n] .
Thus AllFactors [12] returns {1,2,3,4,6,12]-. As always, lists in
Mathematica
are enclosed in braces.
divide 4
division 4
existential bigamy 9
factor 5
integer 3
nonnegative integer 3
5.6.1 Remark AllFactors returns only the positive factors of an
integer. In this
text, however, the phrase “all factors” includes all the positive and
all the negative
factors.
10
composite integer 10
composite 10, 140
definition 4
even 5
factor 5
integer 3
odd 5
positive integer 3
prime 10
6. Primes
Prime numbers are those, roughly speaking, which don’t have nontrivial
factors.
Here is the formal definition:
6.1 Definition: prime number
A positive integer n is a prime if and only if it is greater than 1 and
its only positive factors are 1 and n. Numbers bigger than 1 which are
not primes are called composite numbers.
6.1.1 Example The first few primes are 2,3,5,7,11,13,17,... .
6.1.2 Example 0 and 1 are not primes.
6.1.3 Worked Exercise Let k be a positive integer. Prove that 4k-\-2 is
not a
prime.
Answer 4&i + 2 = 2(2k+ 1) Thus it has factors 1, 2, 2k + 1 and 4k + 2.
We know
that 2 ^ 4k + 2 because k is positive. Therefore 4k + 2 has other
positive factors
besides 1 and 4k + 2, so 4k + 2 is not prime.
6.1.4 Exercise Prove that any even number bigger than 2 is composite.
6.1.5 Exercise Which of these integers are prime and which are
composite? Fac-
tor the composite ones: 91, 98, 108, 111. (Answer on page 243.)
6.1.6 Exercise Which of these integers are prime and which are
composite? Fac-
tor the composite ones: 1111, 5567, 5569.
6.1.7 Exercise Prove that the sum of two odd primes cannot be a prime.
6.2 Primes in Mathematica
The command PrimeQ determines if an integer is prime (it is guaranteed
to work for
n < 2.5 X 10 10 ). Thus PrimeQ [41] will return True and PrimeQ [111]
will return
False.
The command Prime [n] gives the nth prime in order. For example. Prime
[1]
gives 2, Prime [2] gives 3, and Prime [100] gives 541.
6.2.1 Exercise (Mathematica) Find all the factors of your student
number.
11
7. Rational numbers
7.1 Definition: rational number
A rational number is a number representable as a fraction m/n, where
m and n are integers and n 0 .
7.1.1 Example The numbers 3/4 and —11/5 are rational. 6 is rational
because
6 = 6/1. And .33 is rational because .33 = 33/100.
7.2 Theorem
Any integer is rational.
definition 4
divide 4
divisor 5
fact 1
integer 3
lowest terms 11
proof 4
rational number 11
rational 11
representation 15
theorem 2
Proof The integer n is the same as the fraction n/1.
7.2.1 Remark The representation of a rational number as a fraction is
not unique.
For example,
3 _ 6 _ -9
4 “ 8 “ —12
7.2.2 Fact Two representations m/n and r/s give the same rational
number if
and only if ms = nr .
7.3 Definition: lowest terms
Let m/n be the representation of a rational number with m fiO and
n > 0. The representation is in lowest terms if there is no integer
d > 1 for which d \ m and d\n.
7.3.1 Example 3/4 is in lowest terms but 6/8 is not, because 6 and 8
have 2 as
a common divisor.
7.3.2 Exercise Is ^ in lowest terms?
7.4 Theorem
The representation in lowest terms described in Definition 7.3 exists
for
every rational number and is unique.
Proof Left for you to do (Problems 64.2.5 and 63.4.1).
7.4.1 Warning You can’t ask if a rational number is in lowest terms,
only if its
representation as a fraction of integers is in lowest terms.
7.5 Operations on rational numbers
Rational numbers are added, multiplied, and divided according to the
familiar rules
for operating with fractions. Thus for rational numbers a/b and c/d, we
have
a c ac
b X d bd
and
ad + be
bd
(7.1)
7.5.1 Exercise If a/b and c/d are representations of rational
numbers in lowest
terms, must their sum ( ad + bc) /bd and their product ac/bd be in
lowest terms?
(Answer on page 243.)
12
decimal expansion 12
decimal representa-
tion 12
decimal 12, 93
digit 93
integer 3
rational 11
real number 12
specification 2
usage 2
8. Real numbers
8.1 Specification: real number
A real number is a number which can be represented as a directed
distance on a straight line. A real number r is positive if r > 0
and
negative if r < 0 .
8.1.1 Remark Specification 8.1 is informal, but it’s all you are
going to get, since
a formal definition is quite involved.
8.1.2 Example Any integer or rational number is a real number, and
so are num-
bers such as 7 r and \/2. We will see a proof in Section 86
that \/2 is not rational,
which shows that there are real numbers that are not rational.
8.1.3 Usage The symbol \f\ denotes 2. It does not denote —2. In
general, for
a positive real number x , the notation \J x denotes the positive
square root of x ,
which is precisely the unique positive real number r with the
property that r 2 = x.
The unique negative number s such that s 2 = x is denoted by —\fx.
This usage may conflict with usage you saw in high school, but it
is standard in
college-level and higher mathematics.
8.1.4 Exercise For what real numbers x is it true that \J (— x') 2
= xl
8.2 Infinity
In calculus you may have used the symbols oo and — oo in connection
with limits.
By convention, oo is bigger than any real number and — oo is less
than any real
number. However, they are not themselves real numbers. There is no
largest real
number and there is no smallest real number.
9. Decimal representation of real numbers
A real number always has a decimal representation, possibly with an
unending
sequence of digits in the representation. For example, as you know,
the first few
decimal places of 7r are 3.14159 .... As a general rule, you don’t
expect to know the
exact value of a real number, but only an approximation to it by
knowing its first
few decimal places. Note that 22/7 is not i r, although it is close
to it.
9.1.1 Usage The decimal representation is also called the decimal
expansion.
9.1.2 Approximations Mathematicians on the one hand and scientists
and engi-
neers on the other tend to treat expressions such as “3.14159” in
two different ways.
The mathematician will think of it as a precisely given number,
namely , so
in particular it represents a rational number. The scientist or
engineer will treat it
as the known part of the decimal representation of a real number.
From their point
of view, one knows 3.14159 to six significant figures. This book
always takes the
mathematician’s point of view.
13
Mathematicians referring to an approximation may use an ellipsis
(three dots),
as in “ 7 r is approximately 3.14159...”.
The decimal representations of two different real numbers must be
different. How-
ever, two different decimal representations can, in certain
circumstances, represent
the same real number. This is specified precisely by the following
rule:
9.2 Theorem
If m = do.did 2 d, 3 . . . and n = eo-eie 2 e 3 . . . , where all
the di and et are
decimal digits, and for some integer k > 0 the following four
statements
are all correct, then m = n:
DR.l di = e* for 0 <i <k;
DR. 2 dk = + 1 ;
DR.3 di = 0 for all i > k; and
DR.4 e* = 9 for all i> k.
Moreover, if the decimal representations of m and n are not
identical but
do not follow this pattern for some k , then m ^n.
decimal 12, 93
digit 93
integer 3
real number 12
string 93, 167
theorem 2
usage 2
9.2.1 Usage We use a line over a string of digits to indicate
that they are repeated
infinitely often.
9.2.2 Example 4.9 = 5 (here k = 0 in Theorem 9.2) and 1.459 =
1.46 (here k =
2 ).
9.2.3 Remarks
a) As it stands, Theorem 9.2 applies only to real numbers
between 0 and 10,
but that was only to avoid cumbersome notation. By multiplying
or dividing
by the appropriate power of 10, you can apply it to any real
number. For
example, 499.9 = 500, since Theorem 9.2 applies to those
numbers divided by
100 .
b) The proofs of Theorems 9.2 and 10.1 (below) are based on the
theory of
geometric series (and are easy if you are familiar with that
subject) but that
belongs to continuous mathematics rather than discrete
mathematics and will
not be pursued here.
9.2.4 Exercise Which of these pairs of real numbers are equal?
a) 1.414, y/2.
b) 473,472.999.
c) 4.09, 4.1.
(Answer on page 243.)
9.2.5 Exercise Which of these pairs of real numbers are equal?
a) 53.9, 53.0.
b) 39/13, 2.9.
c) 5698/11259 and .506084.
14
decimal 12,93 9.2.6 Exercise If possible, give two different
decimal representations of each num-
digit 93 ber. If not possible, explain why not.
lowest terms 11
rational 11
real number 12
theorem 2
a) f .
b) f •
c) 105.3.
10. Decimal representation of rational numbers
The decimal representation of a rational number m/n is
obtainable by dividing n
into m using long division. Thus 9/5 = 1.8 and 1/3 = 0.333...
A decimal representation which is all 0’s after a certain point
has to be the
decimal representation of a rational number. For example, 1.853
is the rational
number 1853/10 3 . On the other hand, the example of 1/3 shows
that the decimal
representation of a rational number can go on forever.
The following fact is useful: If the decimal representation of
a number n starts
repeating in blocks after a certain point, then n is rational.
For example, 1/7 =
0.142857 with the block 142857 repeated forever.
The following theorem says exactly which rational number is
represented by a
decimal representation with a repeating block of consecutive
digits:
10.1 Theorem
If n = O.bbb. . . , where b is a block of k consecutive digits,
then n =
b/(10 k — 1).
10.1.1 Example 0.13 is 13/99. As another example, the theorem
says that 0.3
is 3/9, which of course is correct.
10.1.2 Exercise Give the exact rational value in lowest terms
of 5.1, 4.36, and
4.136. (Answer on page 243.)
10.1.3 Remark Theorem 10.1 says that if the decimal
representation of a real
number repeats in blocks then the number is rational, and
moreover it tells you how
to calculate it. Actually, the reverse is true, too: the
decimal representation of a
rational number must repeat in blocks after a certain point.
You can see why this is true by thinking about the process of
long division:
Suppose you have gone far enough that you have used up all the
digits in the
dividend (so all further digits are zero). Then, if you get a
certain remainder in the
quotient twice, the process necessarily repeats the second time
what it did the first
time.
15
10.2 Representations in general
It is important to distinguish between a mathematical object
such as a number and
its representation, for example its decimal representation or
(in the case of a rational
number) its representation as a fraction of integers. Thus 9/5,
27/15 and 1.8 all
represent the same number which is in fact a rational number.
We will return to
this idea several times, for example in Section 17.1.3 and in
Section 66.8.
10.3 Types of numbers in Mathematica
Mathematica knows about integers, rational numbers and real
numbers. It treats a
number with no decimal point as an integer, and an explicit
fraction, for example
6/ 14, as a rational number. If the number has a decimal point,
it is always regarded
as real number.
IntegerQ [n] returns True if n is represented as an integer in
the sense just
described. Thus IntegerQ [3] returns True, but IntegerQ [3 . 0]
returns False.
Mathematica will store a number given as the fraction of two
integers as a ratio-
nal number in lowest terms. For example, if you type 6/14, you
will get 3/7 as the
answer. It will return the sum, product, difference and
quotient of rational numbers
as rational numbers, too. Try typing 3/7+5/6 or (3/7)/ (5/6),
for example.
The function that gives you the decimal representation of a
number is N. For
example, N[3/7] gives 0.4285714285714286. You may give a second
input to N
that gives the number of decimal digits that you want. Thus N
[3/7, 20] gives
decimal 12, 93
digit 93
integer 3
lowest terms 11
positive integer 3
predicate 16
proposition 15
rational 11
real number 12
specification 2
usage 2
0.42857142857142857143
You can invoke N by typing //N after an expression, too. For
example, instead
of typing N [3/7+5/4] , you can type 3/7 + 5/4 //N.
11. Propositions
Sentences in English can express emotion, state facts, ask
questions, and so on. A
sentence in a computer language may state a fact or give a
command. In this section
we are concerned with sentences that are either true or false.
11.1 Specification: proposition
A proposition is a statement which is either true or false.
11.1.1 Example Let P be the proposition “4 > 2”, and Q the
proposition “25 <
—2”. Both statements are meaningful; P is true and Q is false.
11.1.2 Example In Example 3.1.2, page 4, we showed that 0 is
not positive by
using the definition of positive to see that 0 is positive if
the proposition 0 > 0 is
true. Since it is not true, 0 is not positive.
11.1.3 Example The statement x > 4 is not a proposition, since
we don’t know
what x is. It is an example of a predicate.
11.1.4 Usage In many textbooks on logic a proposition is called
a sentence.
algebraic expres-
sion 16
instance 16
integer 3
predicate 16
proposition 15
relational symbols
specification 2
usage 2
16
11.1.5 Remark Textbooks on logic define propositions (and
predicates, the sub-
ject of the next chapter) rather than merely specifying them as
we have done. The
definition is usually by an recursive process and can be fairly
complicated. In order
to prove theorems about logic, it is necessary to do this. This
text explains some of
the basic ideas about logic but does not prove theorems in
logic.
11.2 Propositions in Mathematica
A statement such as 2 < 3 is a proposition in Mathematica; if
you type it in, it will
return True. The symbol for equals is == rather than “ = ”, so
for example 2 == 3
returns False.
12. Predicates
12.1 Specification: predicate
A predicate is a meaningful statement containing variables that
becomes true or false when appropriate values are substituted
for the
variables. The proposition obtained by substituting values for
each of
the variables in a predicate is called an instance of the
predicate.
12.1.1 Usage In other texts, a predicate may be called a
“formula” or an “open
sentence” .
12.1.2 Example If x is a variable of type integer, the
statement “25 < x” is
a predicate. If you substitute an integer for x, the statement
becomes true or
false depending on the integer. If you substitute 44 for x you
get the proposition
“25 < 44” , which is true; if you substitute 5 for x, you get
the proposition “25 < 5” ,
which is false.
12.1.3 Usage We will regard a proposition as a predicate with
no variables. In
other words, every proposition is a predicate.
12.1.4 Algebraic expressions and predicates An algebraic
expression is
an arrangement of symbols such as
x 2 --+4 y (12.1)
It consists of variables {x and y in this case) and operation
symbols. The expression
must be correctly formed according to the rules of algebra.
A predicate is analogous to an algebraic expression, except
that it also con-
tains symbols such as “<” and “=” (called relational symbols)
that make the
expression denote a statement instead of a number.
12.1.5 Example The expression
x 2 b4 y>x + y (12.2)
is a predicate.
17
integer 3
predicate 16
proposition 15
real number 12
On the other hand, if data of the correct type are substituted
into a predicate
the result is not a number but a statement which is true or
false, in other words a
proposition.
12.2.2 Example If you substitute x = 3 into the predicate x 2 <
4 you get the
proposition 9 < 4, which is false. The substitution x = 1 gives
1 < 4, which is true.
12.2.3 Example Substituting x = 2 and y = 3 into the expression
(12.2) gives
the proposition 13 > 5, which is true.
12.2.4 Exercise Find a pair of numbers x and y that when
substituted in 12.2
give a false statement.
12.2 Substitution
When numbers are substituted for the variables in an algebraic
expression, the result
is a number.
12.2.1 Example Setting x = 2 and y = 3 in the expression (12.1)
gives the num-
ber 13.
12.2.5 Example Expressions can be substituted into other
expressions as well.
For example one can substitute xy for x in the expression
(12.2) to get
22 6 .
xy b4 y>xy + y
xy
In doing such substitution you must take into account the rules
concerning how
algebra is written; for example to substitute x + y for x and y
+ z for y in (12.1)
you must judiciously add parentheses:
{x + y) 2
b 4 (y + z) > x + y + y + z
x + y
And the laws of algebra sometimes disallow a substitution; for
example you cannot
substitute 0 for x in 12.2.
12.2.6 Exercise Write the result of substituting x for both x
and for y in 12.2.
(Answer on page 243.)
12.3 Types
In this book, variables are normally assumed to be of a
particular type; for example
the variable x mentioned in Example 12.1.2 is of type integer.
We do not always
specify the type of variables; in that case, you can assume
that the variable can
be replaced by any data that makes the predicate make sense.
For example, in the
predicate x < 25, x can be any number for which “<” makes sense
— thus any
real number number, but not a complex number. This informal
practice would have
to be tightened up for a correct formal treatment of
predicates; the intent here is
to provide an informal introduction to the subject in which
predicates are used the
way they are normally used in common mathematical practice.
18
divide 4
integer variable
predicate 16
proposition 15
real variable 18
substitution 17
usage 2
12.3.1 Usage A real variable is a variable of type real. An
integer variable is
18 a variable of type integer. Don’t forget that both integer
variables and real variables
are allowed to have negative values.
12.3.2 Worked Exercise Let x be a variable of type real. Find a
value of x
that makes the statement “ x > 1 and x < 2” true, and another
that makes it false.
Do the same for the case that x is an integer variable.
Answer Any real number between 1 and 2 makes “x > 1 and x < 2”
true, for
example x = \ or x = \[2 . The values x = 0 , x = 1 , x = — 1 ,
and x = 42 all make
it false.
No integer value of x makes the statement true; it is false for
every integer.
12.4 Exercise set
Let m be an integer variable. For each predicate in problems
12.4.1 through 12.4.5,
give (if it is possible) a value of m for which it is true and
another value for which
it is false.
12.4.1
m 4
. (Answer on page 243.)
12.4.2
m =
m. (Answer on page 243.
12.4.3
m =
m + 1 .
12.4.4
m =
2 m.
12.4.5
9
m =
- m.
12.5 Naming predicates
We will name predicates with letters in much the same way that
we use letters to
denote numbers in algebra. It is allowed, but not required , to
show the variable(s)
in parentheses. For example, we can say: let P{x) denote the
predicate “25 < x ” .
Then P(42) would denote the proposition “25<42”, which is true;
but P(— 2)
would be false. P(42) is obtained from P{x) by substitution.
We can also say, “Let P denote the predicate 25 < x” without
the x being
exhibited. This is useful when we want to refer to an arbitrary
predicate without
specifying how many variables it has.
Predicates can have more than one variable. For example, let
Q(x,y ) be “x <
y" . Then Q(25,42) denotes the proposition obtained by
substituting 25 for x and
42 for y. Q( 25,42) is true; on the other hand, Q( 25,-2) is
false, and Q(25,y) is a
predicate, neither true nor false.
12.5.1 Worked Exercise Let m and n be integer variables. Let
P(n) denote the
predicate n < 42 and Q(m,n ) the predicate n \ (m + n ) . Which
of these predicates
is true when 42 is substituted for m and 4 is substituted for n
?
Answer P(4) is 4 < 42, which is true, and <3(42,4) is 4 | 46,
which is false.
19
12.5.2 Exercise If Q(x) is the predicate x 2 <4, what are Q(—l)
and Q(x— 1)?
(Answer on page 243.)
12.5.3 Exercise Let P(x,y,z ) be the predicate xy<x + z + 1.
Write out each
of these predicates.
a) P(l,2,3).
b) P(l,3,2).
c) P(x,x,y)
d) P(x,x + y,y + z) .
(Answer on page 243.)
12.5.4 Exercise Let P be the predicate of Exercise 12.5.3.
Write out P(x, x,x)
and P(x,x — l,x + 1) and for each predicate give a value of x
for which it is true
and another value for which it is false.
12.5.5 Warning You may have seen notation such as “ f(x) ” to
denote a function.
Thus if f(x) is the function whose value at x is 2x + 5, then /
( 3) = 11. We
will consider functions formally in Chapter 39. Here we only
want to call your
attention to a difference between that notation and the
notation for predicates: If
f(x) = 2x + 5, then “/(x)” is an expression. It is the name of
something. On the
other hand, if P(x) denotes the predicate “25 < x”, then P(x)
is a statement - a
complete sentence with a subject and a verb. It makes sense to
say, “If a = 42 , then
P(a)" , for that is equivalent to saying, “If a = 42, then 25 <
a”. It does not make
sense to say, “If a = 42, then /(a)”, which would be “If a =
42, then 2a + 5”. Of
course, it is meaningful to say “If a = 42, then /(a) = 89”.
12.6 Predicates in Mathematica
A statement such as 2 < x is a predicate. If x has not been
given a value, if you type
2 < x you will merely get 2 < x back, since Mathematica doesn’t
know whether it
is true or false.
definition 4
law 19
predicate 16
real number 12
type (of a vari-
able) 17
universally true 19
usage 2
13. Universally true
13.1 Definition: universally true predicate
A predicate containing a variable of some type that is true for
any value
of that type is called universally true.
13.1.1 Example If x is a real number variable, the predicate “x
2 — 1 = (x +
l)(x — 1)” is true for any real number x. In this example the
variable of the defini-
tion is x, its type is “real”, and so any value of that type
means any real number.
In particular, 42 is a real number so we know that 42 2 — 1 =
(42 + 1)(42 — 1)
13.1.2 Usage In some contexts, a universally true predicate is
called a law. When
a universally true predicate involves equality, it is called an
identity.
20
definition 4
predicate 16
quantifier 20, 113
real number 12
type (of a vari-
able) 17
usage 2
13.1.3 Example The predicate “x 2 — 1 = (x + 1) (x — 1) ” is an
identity. An exam-
ple of a universally true predicate which is not an identity is
“x + 3 > x ” (again, x
is real number).
13.1.4 Remark If P(x) is a predicate and c is some particular
value for x for
which -P(c) is false, then P(x) is not universally true. For
example, x > 4 is not
universally true because 3 > 4 is false (in this case, c = 3).
This is discussed further
in Chapter 75.
13.2 Definition: V
We will use the notation (Vx) to denote that the predicate
following it
is true of all x of a given type.
13.2.1 Example (Vx)(x + 3 > x) means that for every x, x + 3>x.
13.2.2 Worked Exercise Let x be a real variable. Which is true?
(a) (Vx)(x >
x). (b) (\/x)(x > x) . (c) (Vx)(x/0).
Answer (a) is false, (b) is true and (c) is false.
13.2.3 Remark In Exercise 13.2.2, it would be wrong to say that
the answer to
(c) is “almost always true” or to put any other qualification
on it. Any universal
statement is either true or false, period.
13.2.4 Example The statement “i/O” is true for x = 3 and false
for x = 0, but
the statement (Vx)(x / 0) is just plain false.
13.2.5 Exercise Let x be a real variable. Which is true? (a)
(Vx)(x x) . (b)
(Vx)((Vy)(x^y)). (c) (Vx) ((Vy)(x > y )) .
13.2.6 Usage The symbol “V” is called a quantifier We take a
more detailed
look at quantifiers in Chapter 75.
13.2.7 Exercise Which of these statements are true? n is an
integer and x a
real number.
a) (Vn)(n + 3 > n) .
b) (Vx) (x + 3 > x) .
c) (Vn)(3n>n).
d) (Vn)(3n + 1 > n) .
e) (Vx)(3x>x).
(Answer on page 243.)
21
14. Logical Connectives
Predicates can be combined into compound predicates using
combining words called
logical connectives. In this section, we consider “and”, “or”
and “not”.
14.1 Definition: “and”
If P and Q are predicates, then P f\Q (“P and Q”) is also a
predicate,
and it is true precisely when both P and Q are true.
14.1.1 Worked Exercise Let n be an integer variable and let
P(n) be the pred-
icate (n > 3 and n is even) . State whether P( 2) , P(6) and P(
7) are true.
Answer P( 2) is false, P( 6) is true and P(7) is false.
14.1.2 Usage
a) A predicate of the form “FAQ” is called a conjunction.
b) Another notation for PaQ is “PQ”. In Mathematica, “PaQ” is
written
P && Q.
14.2 Definition: “or”
PV Q (“P or Q”) is a predicate which is true when at least one
of P
and Q is true.
14.2.1 Usage
a) A compound predicate of the form P V Q is called a
disjunction.
b) Often “P + Q” is used for “PVQ”. In Mathematica, it is
written P | | Q.
14.2.2 Example If P is “4 > 2” and Q is “25 < —2”, then “PaQ”
is false but
“PVQ” is true.
14.2.3 Exercise For each predicate P(n) given, state whether
these propositions
are true: P( 2), P( 6), P( 7).
a) (n > 3 or n is even)
b) (n | 6 or 6 | n)
c) n is prime or (n | 6)
(Answer on page 243.)
14.2.4 Exercise For each predicate give (if possible) an
integer n for which the
predicate is true and another integer for which it is false.
a) (n + 1 = n) V (n = 5) .
b) (n > 7) V (n < 4) .
c) (n > 7) A (n < 4) .
d) (n < 7) V (n > 4) .
(Answer on page 243.)
14.2.5 Exercise Which of the predicates in Problem 14.2.4 are
universally true
for integers? (Answer on page 243.)
and 21, 22
conjunction 21
definition 4
disjunction 21
divide 4
even 5
integer 3
predicate 16
prime 10
proposition 15
usage 2
22
definition 4
even 5
fact 1
integer 3
negation 22
or 21, 22
positive integer 3
predicate 16
truth table 22
usage 2
14.3 Truth tables
The definitions of the symbols ‘A’ and ‘V’ can be
summarized in truth tables:
P Q PAQ
T T T
T F F
FT F
F F F
P Q PVQ
T T T
T F T
FT T
F F F
14.3.1 Remark As the table shows, the definition of ‘V’
requires that PVQ be
true if either or both of P and Q are true; in other words,
this is “or” in the sense
of “and/or”. This meaning of “or” is called “inclusive or”.
14.3.2 Usage In computer science, “1” is often used for
“true” and “0” for
“false”.
14.4 Definition: “xor”
If P and Q are predicates, the compound predicate PXORQ is
true if
exactly one of P and Q is true.
14.4.1 Fact The truth table of XOR is
P Q P XOR Q
T T F
T F T
FT T
F F F
14.4.2 Usage
a) XOR in Mathematica is Xor. PXORQ may be written either P
~Xor~ Q or
Xor [P,Q] .
b) In mathematical writing, “or” normally denotes the
inclusive or, so that a
statement like, “Either a number is bigger than 2 or it is
smaller than 4”
is considered correct. The writer might take pity on the
reader and add the
phrase, “or both”, but she is not obliged to.
14.4.3 Worked Exercise Which of the following sentences say
the same thing?
In each sentence, n is an integer.
a) Either n is even or it is positive.
b) n is even or positive or both.
c) n is both even and positive.
Answer (a) and (b) say the same thing, (c) is not true of
7, for example, but (a)
and (b) are true of 7.
14.5 Definition: “not”
The symbol ‘->P’ denotes the negation of the predicate P .
14.5.1 Example For real numbers x and y, —>(x < y ) means
the same thing as
x>y.
23
14.5.2 Fact Negation has the very simple truth table
P -nP
~T F
F T
14.5.3 Usage
a) Other notations for —>P are P and ~ P .
b) The symbol in Mathematica for “not” is ! , the
exclamation point. —> P is
written !P.
c) The symbol ‘ -i ’ always applies to the first predicate
after it only. Thus in the
expression —>P V Q , only P is negated. To negate the whole
expression P V Q
you have to write (P V Q) ” .
divide 4
fact 1
integer 3
negation 22
predicate 16
truth table 22
usage 2
14.5.4 Warning Negating a predicate is not (usually) the
same thing as stating
its opposite. If P is the statement “3 > 2”, then -> P is
“3 is not greater than 2”,
rather than “3 < 2”. Of course, —>P can be reworded as “3 <
2”.
14.5.5 Example Writing the negation of a statement in
English can be surpris-
ingly subtle. For example, consider the (false) statement
that 2 divides every inte-
ger. The negation of this statement is true; one way of
wording it is that there is
some integer which is not divisible by 2. In particular,
the statement, “All integers
are not divisible by 2” is not the negation of the
statement that 2 divides every
integer.
We will look at this sort of problem more closely in
Section 77.
14.6 Truth Tables in Mathematica
The dmfuncs.m package has a command TruthTable that
produces the truth table
of a given Mathematica logical expression. For example, if
you define the expression
e = a && (b | | ! c)
then TruthTable [e] produces
a b c a Sz&z (b || ! c)
T T T T
T T F T
TFT F
T F F T
F T T F
F T F F
F F T F
F F F F
24
and 21, 22
definition 4
logical connective 21
or 21, 22
propositional vari-
able 104
rule of inference 24
usage 2
15. Rules of Inference
15.1 Definition: rule of inference
Let P±, P- 2 , ■■■P n and Q be predicates. An expression of
the form
Pi,,,. ,P n |— Q
is a rule of inference. Such a rule of inference is valid
if whenever Pi ,
P 2 . . . and P n are all true then Q must be true as well.
15.1.1 Example If you are in the middle of proving
something and you discover
that P A Q is true, then you are entitled to conclude that
(for example) P is true,
if that will help you proceed with your proof. Hence
PAQ \- P (15.1)
is a valid rule of inference.
That is not true for ‘V’, for example: If PVQ is true,
you know that at least
one of P and Q are true, but you don’t know which one.
Thus the purported rule
of inference “PVQ |— P” is invalid.
15.1.2 Usage The symbol ‘h’ is called the “turnstile”.
In this context, it can be
read “yields”.
15.1.3 Example The basic rules of inference for “or”
are
P |— PVQ and Q |— PVQ (15.2)
These say that if you know P , you know PvQ, and if you
know Q , you know
PVQ.
15.1.4 Example Another rule of inference for “and” is
P,Q f- PAQ (15.3)
15.1.5 Exercise Give at least two nontrivial rules of
inference for XOR. The rules
should involve only propositional variables and XOR and
other logical connectives.
15.1.6 Exercise Same instructions as for Exercise
15.1.5 for each of the connec-
tives defined by these truth tables:
P Q P*Q
T T F
T F F
FT T
F F F
(a)
P Q P NAND Q
T T F
T F T
FT T
F F T
(b)
P Q P NOR Q
T T F
T F F
FT F
F F T
(c)
25
15.2 Definitions and Theorems give rules of inference
What Method 3.1.1 (page 4) says informally can be
stated more formally this way:
Every definition gives a rule of inference.
Similarly, any Theorem gives a rule of inference.
15.2.1 Example The rule of inference corresponding to
Definition 4.1, page 4, is
that for m, n and q integers,
m = qn |— n \ m
One point which is important in this example is that it
must be clear in the rule
of inference what the types of the variables are. In
this case, we required that the
variables be of type integer. Although 14 = (7/2) x 4,
you cannot conclude that
4 | 14, because 7/2 is not an integer.
divide 4
integer 3
natural number 3
nonnegative integer 3
positive integer 3
positive 3
rational 11
real number 12
rule of inference 24
truth table 22
usage 2
15.2.2 Worked Exercise State Theorem 5.4, page 8, as a
rule of inference.
Answer k\m,k\n |— k\m + n.
15.2.3 Exercise (discussion) What is the truth table
for the English word
“but”?
16 . Sets
The concept of set, introduced in the late nineteenth
century by Georg Cantor, has
had such clarifying power that it occurs everywhere in
mathematics. Informally, a
set is a collection of items. An example is the set of
all integers, which is traditionally
denoted Z.
We give a formal specification for sets in 21.1.
16.1.1 Example Any data type determines a set — the set
of all data of that
type. Thus there is a set of integers, a set of natural
numbers, a set of letters of the
English alphabet, and so on.
16.1.2 Usage The items which constitute a set are
called the elements or mem-
bers of the set.
16.2 Standard notations
The following notation for sets of numbers will be used
throughout the book.
a) N is the set of all nonnegative integers
b) N + is the set of all positive integers.
c) Z is the set of all integers.
d) Q is the set of all rational numbers.
e) R is the set of all real numbers.
f) R + is the set of all nonnegative real numbers.
g) R ++ is the set of all positive real numbers.
16.2.1 Usage Most authors adhere to the notation of the
preceding table, but
some use N for N + or I for Z .
26
definition 4
integer 3
set 25, 32
type (of a vari-
able) 17
16.3 Definition: “e”
If x is a member of the set A, one writes “x € A if it
is not a member
of A, “xgA”.
16.3.1 Example 4eZ, — 5 G Z, but 4/3 0 Z.
16.4 Sets, types and quantifiers
When using the symbol V, as in Section 13.1, the type
of the variable can be
exhibited explicitly with a colon followed by the name
of a set, as is done in Pas-
cal and other computer languages. Thus to make it clear
that x is an integer, one
could write (Vx:Z )P{x).
16.4.1 Worked Exercise Which of these statements is
true?
a) (Vx:Z)x > 0
b) (Vx:N)x > 0
Answer Part (a) says that every integer is nonnegative.
That is false; for example,
—3 is negative. On the other hand, part (b) is true.
17. List notation for sets
There are two common methods for defining sets: list
notation, discussed here, and
setbuilder notation, discussed in the next chapter.
17.1 Definition: list notation
A set with a small number of members may be denoted by
listing them
inside curly brackets.
17.1.1 Example The set {2,5,6} contains the numbers 2,
5 and 6 as elements,
and no others. So 2 e {2,5,6} but 7 ^ {2,5,6}.
17.1.2 Remark
a) In list notation, the order in which the elements
are given is irrelevant: {2,5,6}
and {5,2,6} are the same set.
b) Repetitions don’t matter, either: {2,5,6}, {2, 2, 5,
6} and {2, 5, 5, 5, 6, 6} are
all the same set. Note that {2, 5, 5, 6, 6} has three
elements.
17.1.3 Remark The preceding remarks indicate that the
symbols {2,5,6} and
{2, 2, 5, 6} are different representations of the same
set. We discussed different rep-
resentations of numbers in Section 10.2. Many
mathematical objects have more
than one representation.
27
17.1.4 Exercise How many elements does the set {1,1, 2,
2, 3,1} have? (Answer
on page 243.)
17.2 Sets in Mathematica
In Mathematica, an expression such as
12,2,5,6}
denotes a list rather than a set. (Lists are treated in
detail in Chapter 109.) Both
order and repetition matter. In particular, 12,2,5,6}
is not the same as 12,5,6}
and neither are the same as 12,6,5} .
A convenient way to list the first n integers is Table
[k , Ik , 1 , n}] . For example,
Table [k, Ik, 1,10}] returns 11,2,3,4,5,6,7,8,9,10}.
comprehension 27,
29
defining condition 27
definition 4
integer 3
predicate 16
setbuilder nota-
tion 27
set 25, 32
type (of a vari-
able) 17
usage 2
17.3 Sets as elements of sets
A consequence of Specification 21.1 is that a set,
being a “single entity”, can be
an element of another set. Furthermore, if it is, its
elements are not necessarily
elements of that other set.
17.3.1 Example Let A = {{1,2}, {3}, 2, 6}. It has four
elements, two of which
are sets.
Observe that 1 € {1,2} and {1,2} 6 A, but the number 1
is not an element
of A. The set {1,2} is distinct from its elements , so
that even though one of its
elements is 1, the set {1,2} itself is not 1. On the
other hand, 2 is an element
of A because it is explicitly listed as such.
17.3.2 Exercise Give an example of a set that has {1,2}
as an element and 2 as
an element but which does not have 1 as an element.
18. Setbuilder notation
18.1 Definition: setbuilder notation
A set may be denoted by the expression {x \ P(x)} ,
where P is a pred-
icate. This denotes the set of all elements of the type
x for which the
predicate P(x) is true. Such notation is called
setbuilder notation.
The predicate P is called the defining condition for
the set, and the
set {x | -P(x)} is called the extension of the
predicate P .
18.1.1 Usage
a) Sometimes a colon is used instead of ‘ | ’ in the
setbuilder notation.
b) The fact that one can define sets using setbuilder
notation is called compre-
hension. See 18.1.11.
18.1.2 Example The set {n | n is an integer and 1 < n <
6} denotes the set
{ 2 , 3 , 4 , 5 }.
28
and 21, 22
extension (of a
predicate) 27
integer 3
predicate 16
prime 10
real number 12
set 25, 32
subset 43
type (of a vari-
able) 17
usage 2
18.1.3 Example The set S = {n \ n is an integer and n
is prime} is the set of all
primes.
18.1.4 Worked Exercise List the elements of these sets,
where n is of type
integer.
a) {n | n 2 = 1} .
b) {n | n divides 12} .
c) {?r | 1 < n < 3} .
Answer a) {-1,1}. b) {1,2, 3, 4, 6, 12, -1,-2, -3, -4,
-6, -12}. c) {2}.
18.1.5 Exercise How many elements do each of the
following sets have? In each
case, x is real.
a) {2, 1,1,1} c) {x | x 2 — 1 = 0}
b) { 1 , 2 , — 1 , x/4, | — 1 1 } d) {x | x 2 + 1 = 0}
(Answer on page 243.)
18.1.6 Example The extension of the predicate
is the set {3,4} .
(x e Z) A (x < 5) A (x > 2)
18.1.7 Example The extension of a predicate whose main
verb is “equals” is what
one would normally call the solution set of the
equation. Thus the extension of the
predicate x 2 = 4 is {—2,2}.
18.1.8 Exercise Write predicates whose extensions are
the sets in exercise 18.1.5
(a) and (b). Use a real variable x.
18.1.9 Exercise Give these sets in list notation, where
n is of type integer.
a) {n | n > 1 and n < 4} .
b) {n | n is a factor of 3} .
18.1.10 Usage In some texts, a predicate is defined to
be what we have called its
extension here: in those texts, a predicate P(x) is a
subset (see Chapter 31) of the
set of elements of type x. In such texts, “(x = 2) V (x
= —2)” would be regarded as
the same predicate as “x 2 = 4”.
29
18.1.11 Method: Comprehension
Let P(x) be a predicate and let A = {x | P(a;)} . Then
if you know that
a G A , it is correct to conclude that P(a) . Moreover,
if P(a) , then you
know that a£i.
18.1.12 Remark The Method of Comprehension means that
the elements of
{x | P(x)} are exactly all those x that make P(x) true.
If A = {x | P(x)}, then
every x for which P{x) is an element of A, and nothing
else is.
This means that in the answer to Worked Exercise
18.1.4, the only correct
answer to part (b) is {1,2,3,4,6,12,— 1,-2,— 3,— 4,— 6,
— 12}. For example, the
set {1,2, 3, 4, 6, —3, —4, —6, —12} would not be a
correct answer because it does not
include every integer that makes the statement “n
divides 12” true (it does not
contain —2, for example).
infinite 174
integer 3
predicate 16
real number 12
rule of inference 24
setbuilder nota-
tion 27
set 25, 32
type (of a vari-
able) 17
unit interval 29
usage 2
18.1.13 Rules of inference for sets It follows that we
have two rules of infer-
ence: If P(x) is a predicate, then for any item a of
the same type as x ,
and
P(a) |- a e {x -P(x)}
(18.1)
a e {x P(x)} |- P(a)
(18.2)
18.1.14 Example
The set
I = {x \ x is real and 0 < x < 1}
(18.3)
which has among its elements 0, 1/4, 7r/4, 1, and an
infinite number of other
numbers. I is fairly standard notation for this set —
it is called the unit interval.
18.1.15 Usage Notation such as “a < x < 6” means a < x
and x < b. So the
statement “0 < x < 1” in the preceding example means “0
< x” and “x < 1”. Note
that it follows from this that 5 < x < 3 means (5 < x)
A (x < 3) — there are no
numbers x satisfying that predicate. It does not means
“(5 < x) V (x < 3)”!
18.1.16 Exercise What is required to show that a ^ {x \
P(x)}? (Answer on
page 243.)
19. Variations on setbuilder notation
Frequently an expression is used left of the vertical
line in setbuilder notation,
instead of a single variable.
19.1 Typing the variable
One can use an expression on the left side of
setbuilder notation to indicate the type
of the variable.
30
and 21, 22
integer 3
predicate 16
rational 11
real number 12
set 25, 32
unit interval 29
19.1.1 Example The unit interval I could be defined as
I = {x 6 R | 0 < x < 1}
making it clear that it is a set of real numbers rather
than, say rational numbers.
19.2 Other expressions on the left side
Other kinds of expressions occur before the vertical
line in setbuilder notation as
well.
19.2.1 Example The set {n 2 | n£Z} consists of all the
squares of integers; in
other words its elements are 0, 1,4,9, 16, ....
19.2.2 Example Let A ={1,3,6}. Then
{n - 2 | n G A} = {—1,1,4}
19.2.3 Remark The notation introduced in the preceding
examples is another
way of putting an additional condition on elements of
the set. Most such defini-
tions can be reworded by introducing an extra variable.
For example, the set in
Example 19.2.1 could be rewritten as
{n 2 | n € Z} = {k \ (k = n 2 ) A (n G Z)}
and the set in Example 19.2.2 as
{n — 2 | n G A} = {m | (m = n-2)A(n£ A)}
19.2.4 Warning Care must be taken in reading such
expressions: for example,
the integer 9 is an element of the set {n 2 | n£ZAn/3},
because although 9 = 3 2 ,
it is also true that 9 = (— 3) 2 , and —3 is an integer
not ruled out by the predicate
on the right side of the definition.
19.2.5 Exercise Which of these equations are true?
a) R + = {x 2 | x G R}
b) N = {x 2 | x G N}
c) R = {x 3 | x G R}
(Answer on page 243.)
19.2.6 Exercise List the elements of these sets.
a) {n — 1 G Z | n divides 12}
b) {n 2 G N | n divides 12}
c) {n 2 G Z | n divides 12}
(Answer on page 243.)
19.2.7 Exercise List the elements of these sets, where
x and y oare of type real:
a) {x + y | y = 1 — x}.
b) {3.x | x 2 = 1} .
31
19.2.8 Exercise How many elements does the set
have?
,, | x = — , -,—2,2}
x 2 1 2 2
19.3 More about sets in Mathematica
The Table notation described in 17.2 can use the
variations described in 19. For
example, Table [k"2 , {k, 1 , 5}] returns
{1,4,9,16,25}.
Defining a set by setbuilder notation in Mathematica is
accomplished using the
command Select. Select [list , criterion] lists all the
elements of the list that
meet the criterion. For example, Select [{2, 5, 6,7,8},
PrimeQ] returns {2,5,7}.
The criterion must be a Mathematica command that
returns True or False for each
element of the list. The criterion can be such a
command you defined yourself; it
does not have to be built in.
closed interval 31
definition 4
open interval 31
real number 12
setbuilder nota-
tion 27
set 25, 32
usage 2
19.3.1 Exercise (Mathematica) Explain the result you
get when you type
Select [{2 ,4, Pi , 5 . 0 , 6 . 0} , Integer Q]
in Mathematica.
20. Sets of real numbers
Now we use the setbuilder notation to define a notation
for intervals of real numbers.
20.1 Definition: interval
An open interval
(a . . b) = {x G R | a < x < b}
(20.1)
for any specific real numbers a and b. A closed
interval includes its
endpoints, so is of the form
[a . . b] = {x G R | a < x < 6}
(20.2)
20.1.1 Example The interval I defined in (18.3),
page 29, is [0 . . 1] .
20.1.2 Usage The more common notation for these
sets uses a comma instead of
two dots, but that causes confusion with the
notation for ordered pair which will be
introduced later.
20.1.3 Exercise Which of these are the same set? x
is real.
a) {0,1,— 1} d) {x | x 3 = — x}
b) {x | x = — x} e) [— 1..1]
c) {x | x 3 = x} f) (— 1..1)
(Answer on page 243.)
32
real number 12
setbuilder nota-
tion 27
set 25, 32
specification 2
20.2 Bound and free variables
The variable in setbuilder notation, such as the x
in Equation (18.3), is bound, in
the sense that you cannot substitute anything for
it. The “dummy variable” x in
an integral such as j ^ f(x) dx is bound in the
same sense. On the other hand, the
a and b in Equation (20.2) are free variables: by
substituting real numbers for a
and b you get specific sets such as [0 . . 2] or [—
5.. 3]. Free variables which occur in
a definition in this way are also called parameters
of the definition.
21. A specification for sets
We said that Method 18.1.11 “determines the set {x
| P(x)} precisely.” Actually,
what the method does is explain how the notation
determines the elements of the set
precisely. But that is the basic fact about sets: a
set is determined by its elements.
Indeed, the following specification contains
everything about what a set is that
you need to know (for the purposes of reading this
book!).
21.1 Specification: set
A set is a single entity distinct from, but
completely determined by, its
elements (if there are any).
21.1.1 Remarks
a) This is a specification, rather than a
definition. It tells you the operative
properties of a set rather than giving a definition
in terms of previously known
objects.
Thus a set is a single abstract thing (entity) like
a number or a point, even
though it may have many elements. It is not the
same thing as its elements,
although it is determined by them.
b) In most circumstances which arise in mathematics
or computer science, a kind
of converse to Specification 21.1 holds: any
collection of elements forms a set.
However, this is not true universally. (See Section
24.)
21.2 Consequences of the specification for sets
A consequence of Specification 21.1 is the
observation in Section 17.1 that, in using
the list notation, the order in which you list the
elements of a set is irrelevant.
Another consequence is the following method.
21.2.1 Method
For any sets A and B , A = B means that
a) Every element of A is an element of B and
b) Every element of B is an element of A.
33
21.2.2 Example For x real,
{x | x 2 = 1} = {x | (x = 1) V (x = —1)}
We will prove this using Method 21.2.1. Let
A = {x | x 2 = 1} and B = {x \ (x = 1) V (x = —1)}
Suppose x 6 A. Then x 2 = 1 by 18.2. Then x 2 — 1 =
0, so (x — l)(x + 1) = 0, so
x = 1 or x = —1. Hence x 6 B by 18.1. On the other
hand, if i£ 5, then x = 1
or x = — 1 , so x 2 = 1 , so
definition 4
empty set 33
extension (of a
predicate) 27
interval 31
or 21, 22
predicate 16
real number 12
set 25, 32
usage 2
21.2.3 Remark The two statements, “x 2 = 1” and “(x
= 1) V (x = —1)” are dif-
ferent statements which nevertheless say the same
thing. On the other hand, the
descriptions {x | x 2 = 1} and {x | (x = 1) V (x =
— 1)} denote the same set; in other
words, the predicates “x 2 = 1” and “(x = 1) V (x =
—1)” have the same extension.
This illustrates that the defining property for a
particular set can be stated in var-
ious equivalent ways, but what the set is is
determined precisely by its elements.
22. The empty set
22.1 Definition: empty set
The empty set is the unique set with no elements at
all. It is denoted
{} or (more commonly) 0.
22.1.1 Remark The existence and uniqueness of the
empty set follows directly
from Specification 21.1.
22.1.2 Example {x £ R | x 2 < 0} = 0.
22.1.3 Example The interval notation “[a.. 6]”
introduced in 20.1 defines the
empty set if a> b. For example, [3. .2] =0.
22.1.4 Example Since the empty set is a set, it can
be an element of another set.
Consider this: although “0” and “{}” both denote
the empty set, {0} is not the
empty set; it is a set whose only element is the
empty set.
22.1.5 Usage This symbol “0” should not be confused
with the Greek letter phi,
written <p , nor with the way the number zero is
sometimes written by older printing
terminals for computers.
22.1.6 Exercise Which of these sets is the empty
set?
a) {0}.
b) {0,0}-
c) {x G Z | x 2 < 0} .
d) {x G Z | x 2 = 2} .
(Answer on page 243.)
34
definition 4
divisor 5
empty set 33
integer 3
positive integer 3
set 25, 32
singleton set 34
singleton 34
23. Singleton sets
23.1 Definition: singleton
A set containing exactly one element is called a
singleton set.
23.1.1 Example {3} is the set whose only element is
3.
23.1.2 Example {0} is the set whose only element is
the empty set.
23.1.3 Remark Because a set is distinct from its
elements, a set with exactly one
element is not the same thing as the element. Thus
{3} is a set, not a number,
whereas 3 is a number, not a set. Similarly, the
President is not the same as the
Presidency, although the President is the only
holder of that office.
23.1.4 Example [3.. 3] is a singleton set, but (3..
3) is the empty set.
23.1.5 Exercise Which of these describe (i) the
empty set (ii) a singleton?
a) {1,-1} e) {x € R + | x < 1}
b) {x £ N | x < 1} f) {x € R | x 2 — 1 = 0}
c) {x G R | x 2 = 0} g) {x e R j x 3 + x = 0}
d) {x £ R | x 2 < 0}
(Answer on page 243.)
23.1.6 Exercise For each positive integer n, let D
n be the set of positive divisors
of n.
a) For which integers n is D n a singleton?
b) Which integers k are elements of D n for every
positive integer n?
(Answer on page 243.)
23.1.7 Exercise Simplify these descriptions of sets
as much as possible, where n
is of type integer.
a) {n | 1 < n < 2} .
b) {n j |n| < 2} .
c) {?r | for all integers m, n < m } .
35
24. Russell’s Paradox
The setbuilder notation has a bug: for some
predicates P(x) , the notation
{x | P(x)} does not define a set. An example is the
predicate “x is a set”. In
that case, if {x | x is a set} were a set, it would
be the set of all sets. However,
there is no such thing as the set of all sets. This
can be proved using the theory of
infinite cardinals, but will not be done here.
We now give another example of a definition {x |
P(x)} which does not give a
set, and we will prove that it does not give a set.
It is historically the first such
example and is due to Bertrand Russell. He took
P(x) to be “x is a set and x is
not an element of itself.” This gives the
expression “{x | x ^ x}”.
We now prove that that expression does not denote a
set. Suppose S =
{x | x^x} is a set. There are two possibilities:
(i) S € S. Then by definition
of S , S is not an element of itself i.e. , S S .
(This follows from the rule of infer-
ence (18.1) on page 29.) (ii) S ^ S . In this case,
since S is not an element of S
and S is the set of all sets which are not elements
of themselves, it follows from
Rule (18.1) that S G S . Both cases are impossible,
so there is no such set as S .
This is an example of a proof by contradiction,
which we will study in detail in
Section 86, page 125.
As a result of the phenomenon that the setbuilder
notation can’t be depended on
to give a set, set theory as a mathematical science
(as opposed to a useful language)
had to be developed on more abstract grounds
instead of in the naive way described
in this book. The most widely-accepted approach is
via Zermelo-Frankel set theory,
which unfortunately is complicated and not very
natural in comparison with the
way mathematicians actually use sets.
Luckily, for most practitioners of mathematics or
computer science, this diffi-
culty with the setbuilder notation does not usually
arise. In most applications, the
notation “{x | P(x)}” has x varying over a specific
type whose instances (unlike
the type “set”) are already known to constitute a
set (e.g., x is real — the real num-
bers form a set). In that case, any meaningful
predicate defines a set {x | P(x)} of
elements of that type.
For more about Russell’s Paradox, see [Wilder,
1965], starting on page 57.
and 21, 22
implication 35, 36
or 21, 22
predicate 16
real number 12
rule of inference 24
Russell’s Paradox 35
setbuilder nota-
tion 27
set 25, 32
type (of a vari-
able) 17
24.0.8 Exercise (discussion) In considering
Russell’s Paradox, perhaps you
tried unsuccessfully to think of a set which is an
element of itself. In fact, most
axiomatizations of set theory rule out the
possibility of a set being an element of
itself. Does doing this destroy Russell’s example?
What does it say about the
collection of all sets?
25. Implication
In Chapter 14, we described certain operations such
as “and” and “or” which com-
bine predicates to form compound predicates. There
is another logical connective
which denotes the relationship between two
predicates in a sentence of the form
“If P, then Q” , or “P implies Q” . Such a
statement is called an implication.
36
antecedent 36 Implications are at the very heart of
mathematical reasoning. Mathematical proofs
conclusion 36 typically consist of chains of
implications.
conditional sen-
tence 36
25.1 Definition: implication
consequent 36, 121
definition 4
hypothesis 36
For predicates P and Q , the implication P => Q is
a predicate defined
by the truth table
implication 35, 36
P Q
P => Q
logical connective 21
T T
T
material condi-
T F
tional 36
predicate 16
truth table 22
F T
F F
type (of a vari-
able) 17
In the implication P Q , P is the hypothesis or
antecedent and Q
is the conclusion or consequent.
usage 2
25.1.1 Example Implication is the logical
connective used in translating state-
ments such as “If m> 5 and 5 > n, then m > n” into
logical notation. This state-
ment could be reworded as, “m > 5 and 5 > n implies
that rn > n.” If we take
P(m,n) to be “(m > 5) A (5 > n)” and Q(m,n) to be
“m > n”, then the statement
“If m > 5 and 5 >n, then m > n” is “P(m,n) =>
Q(m,n)” .
25.1.2 Usage The implication connective is also
called the material condi-
tional, and P => Q is also written P D Q . An
implication, that is, a sentence
of the form P => Q, is also called a conditional
sentence.
25.1.3 Remarks
a) Definition 25.1 gives a technical meaning to the
word “implication”. It also
has a meaning in ordinary English. Don’t confuse
the two. The technical
meaning makes the word “implication” the name of a
type of statement.
b) Warning: The truth table for implication has
surprising consequences which
can cause difficulties in reading technical
articles. The first line of the truth
table says that if P and Q are both true then P =>
Q is true. In Exam-
ple 25.1.1, we have “7 > 5 and 5 > 3 implies 7 > 3”
which you would surely
agree is true.
However, the first line of the truth table also
means that other statements
such as “If 2 > 1 then 3 x 5 = 15” are true. You
may find this odd, since the
fact that 3 x 5 = 15 doesn’t seem to have anything
to do with the fact that
2 > 1 . Still, it fits with the truth table.
Certainly you wouldn’t want the fact
that P and Q are both true to be grounds for P =4>
Q being false.
25.1.4 Exercise Which of these statements are true
for all integers ml
a) m > 7 =4> m > 5 .
b) m > 5 =4- m > 7 .
c) m 2 =4 => m = 2.
(Answer on page 243.)
37
conclusion 36
definition 4
divide 4
fourtunate 37
hypothesis 36
implication 35, 36
integer 3
natural number 3
odd 5
predicate 16
proposition 15
truth table 22
vacuously true 37
26.1.2 Example Both these statements are vacuously
true:
a) If 4 is odd then 3 = 3.
b) If 4 is odd then 3/3.
26.1.3 Remarks Although this situation may be
disturbing when you first see it,
making either statement in Example 26.1.2 false
would result in even more peculiar
situations. For example, if you made P => Q false
when P and Q are both false,
you would then have to say that the statement
discussed previously,
“For any integers m and n, if m > 5 and 5 > n then
rn > n,”
is not always true (substitute 3 for rn and 4 for n
and you get both P and Q
false). This would surely be an unsatisfactory
state of affairs.
Most of the time in mathematical writing the
implications which are actually
stated involve predicates containing variables, and
the assertion is typically that the
implication is true for all instances of the
predicates. Implications involving propo-
sitions occur only implicitly in the process of
checking instances of the predicates.
That is why a statement such as, “If 3 > 5 and 5 >
4, then 3 > 4” seems awkward
and unfamiliar.
26.1.4 Example Vacuous truth can cause surprises in
connection with certain
concepts which are defined by using implication.
Let’s look at a made-up example
here: to say that a natural number n is fourtunate
(the spelling is intentional)
means that if 2 divides n then 4 divides n. Thus
clearly 4, 8, 12 are all fourtunate.
But so are 3 and 5. They are vacuously fourtunate!
26.1.5 Exercise For each implication, give (if
possible) an integer n for which it
is true and another for which it is false.
a) (n > 7) => (n < 4) d) (n = lVn = 3) 4 (n is odd)
b) (n > 7) =>- (n > 4) e) (n = 1 A n = 3) =>- (n is
odd)
c) (n > 7) =>- (n > 9) f) (n = 1 V n = 3) =>- n = 3
(Answer on page 243.)
26. Vacuous truth
The last two lines of the truth table for
implication mean that if the hypothesis of
an implication is false, the implication is
automatically true.
26.1 Definition: vacuously true
In the case that P => Q is true because P is false,
the implication
P => Q is said to be vacuously true.
26.1.1 Remark The word “vacuous” refers to the fact
that in that case the impli-
cation says nothing interesting about either the
hypothesis or the conclusion. In
particular, the implication may be true, yet the
conclusion may be false (because of
the last line of the truth table).
38
implication 35, 36 26.1.6 Exercise If possible,
give examples of predicates P and Q for which each
logical connective 21 of these is (i) true and (ii)
false,
predicate 16 a ) p =>. (p =>. Q)
b) Q => (P => Q)
c) (P =*► Q) =► P
d) (P => Q) =>■ Q
27. How implications are worded
Implication causes more trouble in reading
mathematical prose than all the other
logical connectives put together. An implication
may be worded in various ways; it
takes some practice to get used to understanding
all of them as implications.
The five most common ways of wording P Q are
WI.l If P, then Q.
WI.2 P only if Q.
WI.3 P implies Q.
WI.4 P is a sufficient condition for Q.
WI.5 Q is a necessary condition for P.
27.1.1 Example For all x G Z,
a) If x > 3 , then x >2.
b) x > 3 only if x > 2 .
c) x > 3 implies x > 2 .
d) That x > 3 is sufficient for x > 2.
e) That x > 2 is necessary for x > 3 .
all mean the same thing.
27.1.2 Remarks
a) Watch out particularly for Example 27.1.1(b): it
is easy to read this statement
backward when it occurs in the middle of a
mathematical argument. Perhaps
the meaning of (b) can be clarified by expanding
the wording to read: “x can
be greater than 3 only if x > 2 .”
Note that sentences of the form “P only if Q” about
ordinary everyday
things generally do not mean the same thing as “If
P then Q ” ; that is because
in such situations there are considerations of time
and causation that do not
come up with mathematical objects. Consider “If it
rains, I will carry an
umbrella” and “It will rain only if I carry an
umbrella” .
b) Grammatically, Example 27.1.1(c) is quite
different from the first two. For
example, (a) is a statement about x. whereas (c) is
a statement about state-
ments about x. However, the information they
communicate is the same.
See 27.3 below.
39
27.1.3 Exercise You have been given four cards each
with an integer on one side
and a colored dot on the other. The cards are laid
out on a table in such a way
that a 3, a 4, a red dot and a blue dot are
showing. You are told that, if any of
the cards has an even integer on one side, it has a
red dot on the other. What is
the smallest number of cards you must turn over to
verify this claim? Which ones
should be turned over? Explain your answer.
27.2 Universally true implications
Implications which are universally true are
sometimes stated using the word “every”
or “all”. For example, the implication, “If x > 3,
then x > 2”, could be stated this
way: “Every integer greater than 3 is greater than
2” or “All integers greater than
3 are greater than 2” . You can recognize such a
statement as an implication if what
comes after the word modified by “every” or “all”
can be reworded as a predicate
(“greater than 3” in this case).
even 5
implication 35, 36
integer 3
positive real num-
ber 12
predicate 16
proposition 15
real number 12
rule of inference 24
27.2.1 Exercise Which of the following sentences
say the same thing?
a) If a real number is positive, it has a square
root.
b) If a real number has a square root, it is
positive.
c) A real number is positive only if it has a
square root.
d) Every positive real number has a square root.
e) For a real number to be positive, it is
necessary that it have a square root.
f) For a real number to be positive, it is
sufficient that it have a square root.
(Answer on page 243.)
27.2.2 Exercise Suppose you have been told that the
statement P =>• Q is false.
What do you know about PI About Q‘l
27.3 Implications and rules of inference
Suppose P and Q are any predicates. If P =$■ Q ,
then the rule of inference P |— Q
is valid, and conversely if P |— Q is valid, then P
=>■ Q must be true. This is stated
formally as a theorem in texts on logic, but that
requires that one give a formal
definition of what propositions and predicates are.
We will take it as known here.
27.3.1 Example It is a familiar fact about real
numbers that for all x and y,
(x > y) =>• (x > y — 1) . This can be stated as the
rule of inference x > y x > y — 1.
40
biconditional 40
conclusion 36
definition 4
divide 4
equivalence 40
equivalent 40
hypothesis 36
implication 35, 36
predicate 16
rule of inference 24
truth table 22
28. Modus Ponens
The truth table for implication may be summed up by
saying:
An implication is true unless the hypothesis is
true and the conclusion is false.
This fits with the major use of implications in
reasoning: if you know that the
implication is true and you know that its
hypothesis is true, then you know its
conclusion is true. This fact is called “modus
ponens”, and is the most important
rule of inference of all:
28.1 Definition: modus ponens
Modus ponens is the rule of inference
(P, P => Q) \- Q
which is valid for all predicates P and Q .
(28.1)
28.1.1 Remark That modus ponens is valid is a
consequence of the truth table
for implication (Definition 25.1). If P is true
that means that one of the first two
lines of the truth table holds. If P =4> Q is
true, one of lines 1, 3 or 4 must hold.
The only possibility, then, is line 1, which
says that Q is true.
28.2 Uses of modus ponens
A theorem (call it Theorem T) in a mathematical
text generally takes the form of
an implication: “If [hypotheses H \,. . . , H n
] are true, then [conclusion].” It will then
typically be applied in the proof of some
subsequent theorem using modus ponens.
In the application, the author will verify that
the hypotheses H i, . . . , H n of Theorem
T are true, and then will be able to assert
that the conclusion is true.
28.2.1 Example As a baby example of this, we
prove that 3 | 6 using Theorem 5.1
and Theorem 5.4. By Theorem 5.1, 3 | 3. The
hypotheses of Theorem 5.4 are that
k | m and k\n. Using k = m = n = 3 this becomes
3 | 3 and 3 | 3, which is true.
Therefore the conclusion 3 | 3 + 3 must be true
by Theorem 5.4. Since 3 + 3 = 6 we
have that 3 I 6 .
29. Equivalence
29.1 Definition: equivalence
Two predicates P and Q are equivalent, written
P <=> Q . if for any
instance, both P and Q are true or else both P
and Q are false. The
statement P Q is called an equivalence or a
biconditional.
41
29.1.1 Fact The truth table for equivalence is
PaaQ
T
This is the same as (P =4- Q) A (Q =>• P) .
equivalent 40
fact 1
implication 35, 36
logical connective 21
or 21, 22
predicate 16
theorem 2
truth table 22
usage 2
29.1.2 Usage The usual way of saying that P 44
Q is, “P if and only if Q” , or
“P is equivalent to Q.” The notation “iff” is
sometimes used as an abbreviation
for “if and only if” .
29.1.3 Example x > 3 if and only if both
x>3 and
29.1.4 Warning The statement “P44(2” does
not say that P is true.
29.2 Theorem
Two expressions involving predicates and
logical connectives are equiva-
lent if they have the same truth table.
29.2.1 Example P =4- Q is equivalent to —>
P V Q , as you can see by constructing
the truth tables. This can be understood as
saying that P =$■ Q is true if and only
if either P is false or Q is true.
29.2.2 Worked Exercise Construct a
truth table that shows that (PVQ) AP
is equivalent to (P A R) V (Q A R) .
Answer
P Q R P\JQ
T T T T
T T F T
TFT T
T F F T
F T T T
F T F T
F F T F
F F F F
(PVQ)AP PAR QAR
T T T
F F F
T T F
F F F
T FT
F F F
F F F
F F F
(PAP) V(QAP)
29.2.3 Exercise Construct truth tables
showing that the following three state-
ments are equivalent:
a) P =>■ Q
b) ~^P\J Q
c) -.(PA -.<2)
29.2.4 Exercise Write English sentences
in the form of the three sentences in
Exercise 29.2.3 that are equivalent to
(x>2) =4- (x> 2)
contrapositive 42
converse 42
decimal expansion
decimal 12, 93
definition 4
equivalent 40
implication 35, 36
rational 11
real number 12
theorem 2
truth table 22
42
30. Statements related to an
implication
12
30.1
is
If x > 2 , then x > 3
The first is true for all real numbers
x, whereas there are real numbers for which
the seconxd one is false: An
implication does not say the same thing as its converse.
(If it’s a cow, it eats grass, but if
it eats grass, it need not be a cow.)
30.1.2 Example In Chapter 10, we
pointed out that if the decimal expansion of
a real number r is all 0’s after a
certain point, then r is rational. The converse of
this statement is that if a real number
r is rational, then its decimal expansion is all
0’s after a certain point. This is
false , as the decimal expansion of r = 1/3 shows.
The following Theorem says more about
an implication and its converse:
30.2 Theorem
If P =>• Q and its converse are both
true, then P Q .
30.2.1 Exercise Prove Theorem 30.2
using truth tables and Theorem 29.2.
30.3 Definition: contrapositive
The contrapositive of an implication P
Q is the implication
-i Q -i P . (Note the reversal.)
30.3.1 Example The contrapositive of
If x > 3 , then x > 2
is (after a little translation)
If x < 2 , then x < 3
These two statements are equivalent.
This is an instance of a general rule:
30.4 Theorem
An implication and its contrapositive
are equivalent.
30.4.1 Exercise Prove Theorem 30.4
using truth tables.
30.4.2 Remark To say, “If it’s a cow,
it eats grass,” is logically the same as
saying, “If it doesn’t eat grass, it
isn’t a cow.” Of course, the emphasis is different,
but the two statements communicate the
same facts. In other words,
(P Q) (-■ Q ->P)
Make sure you verify this by truth
tables. The fact that a statement and its contra-
positive say the same thing causes many
students an enormous amount of trouble
in reading mathematical proofs.
30.1 Definition: converse
The converse of an implication P =>■ Q
is Q =4> P.
.1 Example The converse of
If x > 3 , then x > 2
43
30.4.3 Example Let’s look again at this
(true) statement (see Section 10,
page 14):
If the decimal expansion of a real
number r has all 0’s after a certain
point, it is rational.
The contrapositive of this statement is
that if r is not rational, then its decimal
expansion does not have all 0’s after
any point. In other words, no matter how
far out you go in the decimal expansion
of a real number that is not rational,
you can find a nonzero entry further
out. This statement is true because it is the
contrapositive of a true statement.
30.4.4 Remark Stating the
contrapositive of a statement P => Q requires form-
ing the statement —>Q => —> P, which
requires negating each of the statements P
and Q . The preceding example shows
that this involves subtleties, some of which
we consider in Section 77.
30.4.5 Exercise Write the
contrapositive and converse of “If 3 | n then n is
prime”. Which is true? (Answer on page
243.)
30.4.6 Exercise Write the converse and
the contrapositive of each statement in
Exercise 26.1.5 without using
contrapositive 42
converse 42
decimal 12, 93
definition 4
divide 4
extension (of a
predicate) 27
implication 35, 36
include 43
integer 3
predicate 16
prime 10
rational 11
real number 12
rule of inference 24
set 25, 32
subset 43
theorem 2
type (of a vari-
able) 17
usage 2
31. Subsets and inclusion
Every integer is a rational number (see
Chapter 7). This means that the sets Z and
Q have a special relationship to each
other: every element of Z is an element of Q .
This is the relationship captured by
the following definition:
31.1 Definition: inclusion
For all sets A and B , AC B if and only
if xC A => xC B is true for
all x.
31.1.1 Usage The statement A C B is
read “A is included in B v or “A is a
subset of B" .
31.1.2 Example Z C Q, QCR and ICR.
Definition 31.1 gives an immediate rule
of inference and a method:
31.1.3 Method
To show that A C B , prove that every
element of A is an element of B .
31.1.4 Remark If P{x) is a predicate
whose only variable is x and x is of type S
for some set S, then the extension of
P(x), namely {x | P(x)}, is a subset of S.
Some useful consequences of
Definition 31.1 are included in the following theo-
rem.
44
definition 4
equivalent 40
hypothesis 36
implication 35, 36
include 43
proof 4
properly included
set 25, 32
vacuous 37
31.2 Theorem
a) For any set A, AC A.
b ) For any set A, 0 C A.
c) For any sets A and B , A = B if
and only if A C B and B C A .
Proof Using Definition 31.1, the
statement AC A translates to the statement
x&A =y x&A, which is trivially
true. The statement 0 C A is equivalent to the
statement x £ 0 =t- x^A, which is
vacuously true for any x whatever (the hypoth-
esis is always false). We leave the
third statement to you.
31.2.1 Exercise Prove part (c) of
Theorem 31.2.
31.3 Definition: strict inclusion
If A C B but Af^B then every
element of A is in B but there is at
least one element of B not in A .
This is symbolized by A B , and is
read “A is properly included in £”.
31.3.1 Warning Don’t confuse the
statement U A(^ B ” with “-i(AC£)”: the
latter means that there is an
element of A not in B .
31.3.2 Exercise Prove that for all
sets A and B, (A B) =y — 1 ( B C A).
31.4 Inclusion and elementhood
The statement “4 C 5” must be
carefully distinguished from the statement
31.4.1 Example Consider these sets:
A = {1,2,3}
£ = { 1 , 2 , { 1 , 2 , 3 }}
C = {1,2, 3, {1,2, 3}}
A and B have three elements each
and C has four. A £ B because A occurs in
the list which defines B . However,
A is not included in B since 3 G A but 3 ^ B .
On the other hand, AgC and AC C
both.
31.4.2 Exercise Answer each of (i)
through (iii) for the sets X and Y as defined:
(i) age,
(ii) X C Y , and
(iii) X = Y.
a) X={ 1 ,3} , Y={1,3,5}.
b) X={ 1 ,2} , Y= { 1 , { 1 , 2 } }
.
c) X={1,2}, Y={2,1,1}.
d) X={1,2,{1,3}}, Y={1,3,{1,2}}.
e) X={1,2,{1,3}}, Y={ 1 ,{ 1 ,2},{1
,3}} .
(Answer on page 243.)
45
definition 4
include 43
nontrivial subset 45
proper subset 45
set 25, 32
subset 43
usage 2
31.5.1 Usage
a) The word “contain” is ambiguous
as mathematicians usually use it. If iGl,
one often says “A contains x” , and
if B C A, one often says “ A contains B"\
One thing that keeps the
terminological situation from being worse than it
is is that most of the time in
practice either none of the elements of a set are
sets or all of them are. In fact,
sets such as B and C in Example 31.4.1 which
have both sets and numbers as
elements almost never occur in mathematical
writing except as examples in texts
such as this which are intended to bring
out the difference between “element
of” and “included in”!
Nevertheless, when this book uses
the word “contain” in one of these senses,
one of the phrases “as an element”
or “as a subset” is always added.
b) The original notation for “A C
B" was “AcB”. In recent years authors of
high school and college texts have
begun using the symbol ‘ C ’ by analogy
with '<’. However, the symbol ‘C’
is still the one used most by research
mathematicians. Some authors have
used it to mean ‘C^\ but that is an
entirely terrible idea considering
that ‘ C ’ originally meant and is still widely
used to mean ‘C’. This text avoids
the symbol ‘c’ altogether.
31.5.2 Exercise Explain why each
statement is true for all sets A and B , or give
an example showing it is false for
some sets A and B :
a) 0 «E A
b) If A C 0 , then A = 0 .
c) If A = B , then A C B .
d) If 0 € A then A ^ 0.
e) If A £ B and B £ C , then AcC .
f) If AC B and B C C , then ACC.
g) If A C/_ B and B C , then Ac^C .
h) If A B and B ^ C , then 4/C.
31.5.3 Exercise Given two sets S
and T, how do you show that S is not a subset
of T? (Answer on page 244.)
31.4.3 Remark The fact that A C A
for any set A means that any set is a subset
of itself. This may not be what you
expected the word “subset” to mean. This leads
to the following definition:
31.5 Definition: proper
A proper subset of a set A is a set
B with the property that B C A
and B ^ A. A nontrivial subset of A
is a set B with the property
that B C A and B ^ 0 .
46
definition 4
empty set 33
fact 1
include 43
powerset 46
rule of inference
set builder nota-
tion 27
set 25, 32
subset 43
32. The powerset of a set
32.1 Definition: powerset
If A is any set, the set of all
subsets of A is called the powerset of A
and is denoted VA.
32.1.1 Remark Using setbuilder
notation, VA = {X \ X C A}.
32.1.2 Example The powerset of
{1,2} is {0,{1},{2},{1,2}}, and the powerset
of {1} is {0,{1}}.
32.1.3 Fact The definition of
powerset gives two rules of inference:
BCA |- B € VA
and
B £ VA \- BCA
(32.1)
(32.2)
32.1.4 Example The empty set is an
element of the powerset of every set, since
it is a subset of every set.
32.1.5 Warning The empty set is not
an element of every set; for example, it is
not an element of {1,2}.
32.1.6 Exercise How many elements
do each of the following sets have?
a) {1,2, 3, {1,2, 3}}
b) 0
c) {0}
d) {0,{0}}
(Answer on page 244.)
32.1.7 Exercise Write the powerset
of {5,6,7}. (Answer on page 244.)
32.1.8 Exercise State whether each
item in the first column is an element of each
set in the second column.
a)
a)
b)
b)
c)
7 r
c)
{1,3,7}
d)
{1,3}
d)
{xGR x
e)
{3,tt}
e)
V(Z)
f)
f)
g)
g)
{Z,R}
(Answer on page 244.)
47
33. Union and intersection
33.1 Definition: union
For any sets A and B , the union A
U B of A and B is defined by
Au B = {x | x&A VxG-B} (33.1)
33.2 Definition: intersection
For any sets A and B , intersection
.4 f] B is defined by
AnB = {x I xeA AxeB} (33.2)
definition 4
disjoint 47
extension (of a
predicate) 27
intersection 47
logical connective 21
powerset 46
predicate 16
set 25, 32
union 47
33.2.1 Example Let A = {1,2} and B
= {2,3,4}. Then 4U B = {1,2, 3, 4} and
4nB = {2}. If C = {3,4,5} , then
AnC = 0.
33.2.2 Exercise What are {1,2, 3} U
{2, 3, 4, 5} and {1,2, 3} n{2,3,4, 5}? (Answer
on page 244.)
33.2.3 Exercise What are NUZ and
NnZ? (Answer on page 244.)
33.2.4 Remark Union and
intersection mirror the logical connectives ‘V’ and
‘A’ of section 14. The connection
is by means of the extensions of the predicates
involved. The extension of PV Q is
the union of the extensions of P and of Q , and
the extension of P A Q is the
intersection of the extensions of P and of Q .
33.2.5 Example Let S' be a set of
poker chips, each of which is a single color,
either red, green or blue. Let R,
G, B be respectively the sets of red, green and
blue chips. Then PUP is the set of
chips which are either red or blue; the ‘U’
symbol mirrors the “or” . And R n B
= 0 , since it is false that a chip can be both
red and blue.
33.2.6 Warning Although union
corresponds with “V”, the set RU B of the
preceding example could also be
described as “the set of red chips and blue chips” !
33.2.7 Exercise Prove that for any
sets A and B , An B C AU B . (Answer on
page 244.)
33.2.8 Exercise Prove that for any
sets A and B, AnB C A and A C AU B .
33.3 Definition: disjoint
If A and B are sets and An B = 0
then A and B are said to be disjoint.
33.3.1 Exercise Name three
different subsets of Z that are disjoint from N.
(Answer on page 244.)
33.3.2 Exercise If A and B are
disjoint, must V(A) and V(B) be disjoint?
48
complement 48
definition 4
fact 1
set difference 48
set of all sets 35
set 25, 32
subset 43
type (of a vari-
able) 17
universal set 48
usage 2
34. The universal set and
complements
Since we cannot talk about the set
of all sets, there is no universal way to mir-
ror TRUE as a set. However, in many
situations, all elements are of a particular
type. For example, all the elements
in Example 33.2.5 are chips. The set of all ele-
ments of that type constitutes a
single set containing as subsets all the sets under
consideration. Such a set is called
a universal set, and is customarily denoted U .
Given a universal set, we can
define an operation corresponding to ‘ — i ’, as in the
following defintion.
34.1 Definition: complement
If A is a set, A c is the set of
all elements in U but not in 4. A c is
called the complement of A (note
the spelling).
34.1.1 Usage A c may be denoted A
or A 1 in other texts.
34.1.2 Example The complement of N
in Z is the set of all negative integers.
34.2 Definition: set difference
Let A and B be any two sets. The
set difference A — B is the set
defined by
A — B = {x \ x&AAx^B} (34.1)
34.2.1 Example Let A = {1,2,3} and
B = {3,4,5} ; then A — B = {1,2} .
34.2.2 Exercise What is Z — N? What
is N — Z? (Answer on page 244.)
34.2.3 Fact If there is a universal
set U . then A c = U — A.
34.2.4 Usage A — B is written A\B
in many texts..
34.2.5 Exercise Let A = {1,2,3}, H
= {2,3,4,5} and (7 = 11,7,8}. Write out
the elements of the following sets:
a)
AUB
f)
B-C
b)
AnB
g)
dn(BuC)
c)
BUC
h)
Bu(AnC)
d)
Bnc
i)
Bll(A-C)
e)
A-B
(Answer on page 244.)
49
34.2.6 Exercise State whether each
item in the first column is an element of each
set in the second column. A =
{1,3,7}, B = {1,2, 3, 4, 5}, and the universal set is
Z.
1 ) 1
2) 4
3) 7
4) -2
5) 0
6) {2,4,5}
7) {1,3}
1) AUB
2) AnB
3) A-B
4) A- Z
5) B c
6) VA
7) V(AnB)
(Answer on page 244.)
equivalent 40
first coordinate 49
include 43
integer 3
powerset 46
real number 12
second coordinate 49
set 25, 32
specification 2
type (of a vari-
able) 17
universal set 48
34.2.7 Exercise Explain why the
following statements are true for all sets A
and B or give examples showing they
are false for some A and B .
a) V{A)nV(B)=V(AnB)
b) V{A)uV{B)=V{AuB)
c) V(A)-V(B)=V(A-B)
34.2.8 Exercise Show that for any
sets A and B included in a universal set IA .
if A U B = U and 4nB = 0, then B =
A c .
35. Ordered pairs
In analytic geometry, one
specifies points in the plane by ordered pairs of real
numbers, for example (3,5).
(Most books use round parentheses instead of pointy
ones.) This is not the same as
the two-element set {3,5}, because in the ordered
pair the order matters: (3,5)
is not the same as (5,3).
In the ordered pair (3,5), 3 is
the first coordinate and 5 is the second
coordinate. Sometimes, the two
coordinates are the same: for example, (4,4) has
first and second coordinates
both equal to 4 .
An ordered pair in general need
not have its first and second coordinates of the
same type. For example, one
might consider ordered pairs whose first coordinate is
an integer and whose second
coordinate is a letter of the alphabet, such as (5, ‘a’)
and (—3 ,‘d’).
The following specification
gives the operational properties of ordered pairs:
35.1 Specification: ordered
pair
An ordered pair (x,y) is a
mathematical object distinct from x and y
which is completely determined
by the fact that its first coordinate is x
and its second coordinate is y.
35.1.1 Remark Specification
35.1 implies that ordered pairs are the same if and
only if their coordinates are
the same:
((x,y) = (, x f ,y ')) 44 (x =
x' A y = y')
Thus we have a method:
50
coordinate 49
definition 4
integer 3
ordered pair 49
ordered triple 50
specification 2
tuple 50, 139, 140
union 47
usage 2
35.1.2 Method
To prove two ordered pairs
(x,y) and (. x',y ') are the same, prove that
x = x' and y = y' .
35.1.3 Exercise Which of these
pairs of ordered pairs are equal to each other?
a) (2,3), (3,2).
b) (3 , V / 4), (3,2).
c) (2,>/4), (V4,2).
(Answer on page 244.)
35.1.4 Exercise (discussion) In
texts on the foundations of mathematics, an
ordered pair (a, b) is often
defined to be the set j{a,6},{a}j. Prove that at least
when a and b are numbers this
definition satisfies Specification 35.1 (with a suitable
definition of “coordinate”).
36. Tuples
In order to generalize the idea
of ordered pair to allow more than two coordinates,
we need some notation.
36.1 Definition: n
Let n be an integer, n > 1 .
Then n is defined to be the set
{i G N | 1 < i < n}
36.1.1 Example 3 = {1,2,3}.
36.1.2 Exercise Let m and n be
positive integers. What is mfln? What is
mUn? (Answer on page 244.)
A tuple is a generalization of
the concept of ordered pair. A tuple satisfies this
specification:
36.2 Specification: tuple
A tuple of length n, or n
-tuple, is a mathematical object which
T.l has an ith entry for each i
6 n, and
T.2 is distinct from its
entries, and
T.3 is completely determined by
specifying the ith entry for every
i G n.
36.2.1 Example An ordered pair
is the same thing as a 2 -tuple.
36.2.2 Usage
a) A 3 -tuple is usually called
an ordered triple.
b) The usual way of denoting a
tuple is by listing its entries in order inside angle
brackets.
51
36.2.3 Example (1,3,3,— 2) is a
tuple of integers. It has length 4. The integer
3 occurs as an entry in this 4-
tuple twice, for i = 2 and i = 3.
36.2.4 Usage Tuples and their
coordinates are often named according to a sub-
scripting convention, by which
one refers to the ith entry by subscripting i to the
name of the tuple. For example,
let a = (1,3,3, —2} ; then ci2 = 3 and 04 = —2. One
often makes this convention
clear by saying, “Let a = (ai)i en be an n-tuple.”
Many authors would use curly
brackets here: “{aj}i en .” Nevertheless, a is not
a set.
coordinate 49
empty set 33
equivalent 40
integer 3
null tuple 51
set 25, 32
theorem 2
tuple 50, 139, 140
usage 2
36.2.5 Usage Many computer
scientists refer to a tuple as a “vector”. Although
this usage is widespread, it is
not desirable; in mathematics, a vector is a geometric
object which can be represented
as a tuple, but is not itself a tuple.
It follows from Specification
36.2 that two n-tuples are equal if and only if they
have the same entries.
Formally:
36.3 Theorem: How to tell if
tuples are equal
Let a and b be n-tuples. Then
a = 6 44 (Vi:n)(ai = 6*)
36.3.1 Exercise Which of these
pairs of tuples are equal?
a) (3,3), (3,3,3).
b) (2,3), (2,3,3).
c) (2,3,2), (3,2,2).
(Answer on page 244.)
36.4 Special tuples
For formal completeness, one
also has the concept of the null tuple (or empty
tuple) () , which has length 0
and no entries, and a 1 -tuple, which has length 1 and
one entry.
The index set for the null
tuple is the empty set. There is only one null tuple.
In the context of formal
language theory the unique null tuple is often denoted “A”
(capital lambda) or sometimes
“e” (small epsilon). We will use the notation A here.
36.4.1 Exercise For each tuple,
give the integer n for which it is an n-tuple and
also give its second entry.
a)
b)
c)
(Answer on page 244.)
(3,4,0) d) (((2, (1,5)), 7), 9)
((3,4), (1,5)) e) (3, {1,2})
(3, (5, (2,1))) f) (N,Z,Q,R)
52
Cartesian product
coordinate 49
definition 4
diagonal 52
factor 5
ordered pair 49
real number 12
set 25, 32
subset 43
theorem 2
tuple 50, 139, 140
52 37. Cartesian Products
37.1 Definition: Cartesian
product of two sets
Let A and B be sets. Ax B is
the set of all ordered pairs whose first
coordinate is an element of A
and whose second coordinate is an element
of B. Ax B is called the
Cartesian product of A and B (in that
order) .
37.1.1 Example if A = {1,2} and
B = {2,3,4} , then
AxB= {(1,2), (1,3), (1,4), (2,
2), (2, 3), (2, 4)}
and
B x A = | (2, 1), (2,2), (3,
1), (3,2), (4, 1), (4,2) j
37.1.2 Exercise Write out the
elements of {1,2} x {a, b} . (Answer on page 244.)
37.2 Theorem
If A is any set, then Ax0 = 0xA
= 0.
37.2.1 Exercise Prove Theorem
37.2.
37.2.2 Example R x R is often
called the “real plane” , since it consists of all
ordered pairs of real numbers,
and each ordered pair represents a point in the plane
once a coordinate system is
given. Graphs of straight lines and curves are subsets
of R x R. For example, the x-
axis is j(x,0) | x G r| and the parabola y = x 2
is | (x,y) | x € R A y = x 2
j , which could be written |(x,x 2 ) | x £ R j (recall the
discussion in Section 19.2).
37.3 Definition: diagonal
For any set A, the subset |
(a,a) | o £ a| of A x A of all pairs whose
two coordinates are the same is
called the diagonal of A , denoted A a ■
37.3.1 Worked Exercise Write
out the diagonal of {1,2} x {1,2}.
Answer {(1, 1), (2,2)} .
37.3.2 Example The diagonal Ar
of R x R is the 45-degree line from lower left
to upper right. It is the graph
of the equation y = x.
37.4 Cartesian products in
general
The notion of Cartesian product
can be generalized to more than two factors using
the idea of tuple.
53
37.5 Definition: Cartesian
product
Let Ai,A 2 ,... ,A n be sets —
in other words, let (Aj)j 6n be an n-tuple
of sets. Then A± x A 2 x • • •
x A n is the set
|(oi,a 2 ,... ,a n ) | (V*:n)
(aj G A { ) j (37.1)
of all n-tuples whose ith
coordinate lies in A{ .
37.5.1 Example The set R x Z x
R has triples as elements; it contains as an
element the ordered triple (n,—
2,3), but not, for example, (— 2, 7r, 3) .
37.5.2 Warning Observe that
RxNxR, (RxN)xR and R x (N x R) are three
different sets; in fact, any
two of them are disjoint. Of course, in an obvious sense
they all represent the same
data.
Cartesian product 52
coordinate 49
definition 4
disjoint 47
ordered triple 50
powerset 46
proper subset 45
relation 73
set 25, 32
subset 43
tuple 50, 139, 140
union 47
37.5.3 Example Consider the set
D = {(m,n) | m divides n}
where m and n are of type
integer. Thus (3,6), (—3,6) and (5,0) are elements of
D but not (3,5). D is not a
Cartesian product, although it is a (proper) subset of
the cartesian product Z x Z .
The point is that a pair in Ax B can have any element
of A as its first coordinate
and any element of B as its second coordinate, regardless
of what the first coordinate
is. In D what the second coordinate is depends on what
the first coordinate is.
A set such as D is a relation,
a concept discussed later.
37.6 Exercise set
Exercises 37.6.1 through 37.6.6
give “facts” which may or may not be correct for
all sets A, B and C. State
whether each “fact” is true for all sets A, B and C,
or false for some sets A, B or
C , and for those that are not true for all sets, give
examples of sets for which they
are false.
37.6.1 AxA = A. (Answer on page
244.)
37.6.2 Ax B = B x A.
37.6.3 Au{BxC) = {AuB)x{AuC).
37.6.4 An{BxC) = {AnB)x{AnC).
37.6.5 Ax(BxC) = (AxB)xC.
37.6.6 V{AxB)=V{A)xV{B).
37.7 Exercise set
The statements in problems
37.7.1 through 37.7.3 are true for all sets A, B and C ,
except that in some cases some
of the sets A , B and C have to be nonempty if the
statement is to be true for all
other sets named. Amend the statement in each case
so that it is true.
54
Cartesian powers 54
Cartesian product 52
Cartesian square 54
implication 35, 36
include 43
powerset 46
set 25, 32
singleton 34
tuple 50, 139, 140
union 47
37.7.1 For all sets A, B and
C , A x C = B x C =t- A = B . (Answer on page
244.)
37.7.2 For all sets A and B , A
x B = B x A => A = B .
37.7.3 For all sets A, B and
C , 4C B =t- (A x C) C (B x C) .
37.8 Cartesian product in
Mathematica
The dmfuncs .m package contains
the command CartesianProduct, which gives the
Cartesian product of a sequence
of sets. For example,
produces
CartesianProduct [{1 , 2} ,
{a,b, c} , {x,y}]
{{1, a, x>, {1, a, y>, {I, b,
x>, {1, b, y>, {1, c, x>, {1, c, y>,
{2, a, x>, {2, a, y>, {2, b,
x>, {2, b, y>, {2, c, x>, {2, c, y»
37.8.1 Exercise (Mathematica)
The command CartesianProduct mentioned
in 37.8 works on any lists, not
just sets (see 17.2, page 27). Write a precise descrip-
tion of the result given when
CartesianProduct is applied to a sequence of lists,
some of which contain repeated
entries.
37.9 Exponential notation
If all the sets in a Cartesian
product are the same, exponential notation is also used.
Thus A 2 = Ax A, A 3 = Ax Ax A,
and in general
A n = Ax Ax ■■■ x A
(n times). These are called
Cartesian powers of A ; in particular, A 2 is the
Cartesian square of A. This
notation is extended to 0 and 1 by setting A° = {()}
(the singleton set containing
the null tuple as an element) and A 1 = A.
37.9.1 Exercise
set:
Let
A = {1,2} and B = {3,4,5}.
Write all the elements of each
a)
A 0
f)
BxA
b)
A 1
g)
Ax AxB
c)
A 2
h)
Ax (AxB)
d)
A 3
i)
(AxB)uA
e)
AxB
j)
(Ax B)nA
(Answer on page 244.)
37.9.2 Exercise For each pair
of numbers ( m,n ) € {1,2, 3, 4, 5,6,7} x {1,2, 3, 4, 5, 6, 7},
state whether item m in the
first column is an element of the set in item n of the
second column. A = {1,3,7}, B =
{1,2, 3,4,5} .
55
1.
(3,5)
1. Ax A
Cartesian product 52
2.
(3,3)
2. A 3
diagonal 52
3.
(3,3,5)
3. Ax B
extension (of a
4.
{(3,5), (7,5)}
4. B x A
predicate) 27
intersection 47
5.
({3, 7}, {3, 5})
5. V(A x B)
powerset 46
6.
6. VA x VB
predicate 16
7.
(1,7,7)
7. B 2
real number 12
(Answer on page 244.)
set 25, 32
subset 43
38. Extensions of predicates
with more than one variable
In section 18.1, page 27, we
discussed the extension of a predicate containing one
variable; the extension is a
subset of the type set of the variable. For example, the
extension of u x < 5” (x real)
is the subset {x \ x < 5} of R.
38.1.1 Remark A predicate can
contain several occurrences of one variable. If
it contains no occurrences of
other variables, it is still said to contain one variable.
For example, “(x < 5) A (x >
1)” contains two occurrences of one variable, namely
x . On the other hand, “ (x — y
< 5) A (x + y > 1) ” contains two variables, x and y .
A predicate with more than one
variable, such as u x < y” ( x,y real), has an
extension which is a subset of
a Cartesian product of its variable types.
38.1.2 Example The extension of
“x < y” in R x R is
{{x,y) | x < y}
which is a subset of R x R .
38.1.3 Example The extension of
the predicate “x = x” in R is the subset R of
R, whereas the extension of the
predicate “x = y” in R x R is Ar , the diagonal
subset of R x R .
38.1.4 Worked Exercise Write
out the extension of the predicate “has the same
prime divisors as” in {2,3,4,
6} 2 .
Answer { (2, 2) , (3, 3) , (2,
4) , (4, 2) , (4, 4) , (6, 6) } .
38.1.5 Remark The Cartesian
product in which the extension of a predicate lies
is not uniquely determined. For
example, the predicate “x < y” has an extension
in the set R x R x R, namely
the subset j (x,y,z) \ x < y} . In this case, there is
no condition on the variable
z . There is a good reason for allowing this situation.
For example, the predicate “y <
z” also has an extension in R x R x R, namely
| (x,y,z) | y < zj. Looking at
it this way allows us to say that the extension of
“x<yAy<z” in RxRxR is the
intersection of the extensions of “x < y” and
“y < z”.
56
Cartesian product 52
character 93
codomain 56
coordinate 49
domain 56
extension (of a
predicate) 27
integer 3
predicate 16
real number 12
set 25, 32
specification 2
string 93, 167
tuple 50, 139, 140
type (of a vari-
able) 17
value 56, 57
By the way, we could have
regarded R x R x R as the set of tuples
\{y,z,x) | x,y,z £ r|
Then the extension of “x < y"
would be j (y,z,x) \ x < yj . Because of this sort of
thing, it pays to be careful to
describe exactly what set the extension of a predicate
lies in.
38.2 Exercise set
In Problems 38.2.1 through
38.2.4, describe explicitly the extensions of the predi-
cates in the given set; x, y
and z are real and n is an integer. Associate x,y,z
with coordinates in a tuple in
alphabetical order.
38.2.1 x > n, in R x N. (Answer
on page 244.)
38.2.2 x + y = x + l,inRxR.
(Answer on page 244.)
38.2.3 y = 1 , in R. (Answer on
page 244.)
38.2.4 x + y = z, in RxRxRxR.
(Answer on page 244.)
39. Functions
39.1 The concept of function
In analytic geometry or
calculus class you may have studied a real-valued function
such as G(x) = x 2 + 2x + 5 .
This function takes as input a real number and gives
a real number as value; for
example, the statement that G( 3) = 20 means that an
input of 3 gives an output of
20. It also means that the point (3,20) is on the
graph of the equation y =
G(x) .
The concept of function to be
studied here is more general than that example in
several ways. In the first
place, a function F can have one type of input and another
type of output. An example is
the function which gives the number of characters in
an English word: the input is a
string of characters, say ‘cat’, and the output is its
length, 3 in this case. Also,
the most general sort of function need not be given by
a formula the way G is. For
example, a price list is a function with input the name
of an item and output the price
of the item. The relationship between the name
and the price is rarely given
by a formula.
Here is the specification:
39.2 Specification: function
A function F is a mathematical
object which determines and is com-
pletely determined by the
following data:
F.l F has a domain, which is a
set and is denoted domF.
F.2 F has a codomain, which is
also a set and is denoted codF.
F.3 For each element a £ domF,
F has a value at a. This value is
completely determined by a and
F and must be an element of
codF. It is denoted by F(a) .
57
39.2.1 Warning This
specification for function is both complicated and subtle
and has conceptual traps. One
of the complications is that the concept of function
given here carries more
information with it than what is usually given in calculus
books. One of the traps is that
you may tend to think of a function in terms of a
formula for computing it,
whereas a major aspect of our specification is that what
a function is is independent of
how you compute it.
39.2.2 Usage The standard
notation F:A-^B communicates the information
that F is a function with
domain A and codomain B. A and B are sets; A
consists of exactly those data
which you can use as input to (you can “plug into” )
the function F , and every
value of F must lie in B . (Not every element of B has
to be a value.)
In the expression u F(a)” , a
is called the argument or independent variable
or input to F and F(a) is the
value or dependent variable or output. The
operation of finding F(a) given
F and a is called evaluation or application.
If F(a) = b, one may say U F
takes a to 6” or “a goes to b under F" .
It follows from the definition
that we have the following rule of inference:
F:A->B, aeA (- F(a) <E B (39.1)
39.2.3 Warning You should
distinguish between F , which is the name of the
function, and F(a), which is
the value of F at the input value a. Nevertheless, a
function is often referred to
as F(x) — a notation which has the value of telling
you what the notation for the
input variable is.
39.2.4 Usage Functions are also
called mappings, although in some texts the
word “mapping” is given a
special meaning.
39.3 Examples of functions
We give some simple examples of
functions to illustrate the basic idea, and then
after some more discussion and
terminology we will give more substantial examples.
39.3.1 Example The first
example is the function G : R — > R defined by G(x) =
x 2 + 2x + 5 , which was
discussed previously. Referring to it as G : R — > R specifies
that the domain and codomain
are both R.
39.3.2 Usage As is often the
case in analytic geometry and calculus texts, we
did not formally specify the
domain and codomain of G when we defined it at the
beginning of this section. In
such texts, the domain is often defined implicitly as the
set of all real numbers for
which the defining formula makes sense. For example, a
text might set S(x) = l/x,
leaving you to see that the domain is R — {0} . Normally
the codomain is not mentioned
at all; it may usually be assumed to be R. In this
text, on the other hand, every
function will always have an explicit domain and
codomain.
39.3.3 Example Here is an
example of a function with a finite domain and codo-
rnain. Let A = {1,2,3} and B =
{2,4, 5, 6}. Let F:A-^B be defined by requiring
that F(l) = 4 and F( 2) = F( 3)
= 5 .
application 57
argument 57
codomain 56
dependent vari-
able 57
domain 56
evaluation 57
finite 173
function 56
independent vari-
able 57
input 57
output 57
real number 12
rule of inference 24
set 25, 32
usage 2
value 56, 57
58
codomain 56
divisor 5
domain 56
finite 173
function 56
powerset 46
prime 10
set 25, 32
39.3.4 Example Let S be some
set of English words, for example the set of
words in a given dictionary.
Then the length of a word is a function L : S — > N . For
example, L(‘caf’) = 3 and L( l
abbadabbadoo ’) = 12. (This function in Mathematica
is StringLength. For example,
StringLength["cat"] returns 3.)
39.3.5 Example Let F: N — )• N
be defined by requiring that F( 0) = 0 and for
n > 0, F(n) is the rath prime
in order. Thus F(l) = 2, F( 2) = 3, F( 3) = 5, and
.F(IOO) = 541. (This function
is given by the word Prime in Mathematica.)
39.3.6 Remarks The preceding
examples illustrate several points:
a) You don’t have to give a
formula to give a function. In the case of Exam-
ple 39.3.3, we defined F by
giving its value explicitly at every element of the
domain. Of course, this is
possible only when the domain is a small finite set.
b) There must be a value F(x)
for every element x of the domain, but not every
element of the codomain has to
be a value of the function. Thus 4 is not a
value of the function in
39.3.5.
c) Different elements of the
domain can have the same value (different inputs
can give the same output). This
happens in Example 39.3.1 too; thus G(l) =
G(— 3) = 8.
39.3.7 Exercise Let A = {1, 2,
3} . Let F : A -> VTA be defined by requiring that
F(n) = {B G VA | n G B} . What
are F( 1) and F( 2)? (Answer on page 244.)
39.3.8 Exercise Let A be as in
the preceding exercise, and define G : A — > WA
by G(n) = {B G VA \ n ^ B} .
What are G(l) and G(2)?
39.3.9 Exercise Let F : Z — »
VZ be defined by requiring that F(n) be the set of
divisors of n. What are F( 0) ,
F(l), F( 6) and F(12)?
39.3.10 Exercise Give an
example of a function F : R — > R with the property
r is an integer if and only if
F(r) is not an integer
39.3.11 Exercise Let S be a set
and G:S -^VS a function. Let the subset A
of S be defined by
A = {x | x £ G(x)}
Show that there is no element x
G S for which G(x) = A.
39.4 Functions in Mathematica
In Mathematica, the name of the
function is followed by the input in square brackets.
For example, sinx is entered as
Sin[x] .
You can define your own
functions in Mathematica. The function G(x) = x 2 +
2x + 5 of Example 39.3.1 can be
defined by typing
g[x_] := x~2 + 2 x + 5 (39.2)
Then if you typed g [3] ,
Mathematica would return 20, and if you typed g [t] ,
Mathematica would return 5 + 2
t + t‘2. Comments:
59
a) All built-in Mathematica
functions, such as Sin, start with a capital letter.
It is customary for the user to
use lowercase names so as to avoid overwriting
the Mathematica definition of
some function. (You could define Sin to be
anything you want, but that
would be undesirable.) Thus there would be no
error message if you typed G
instead of g in (39.2), but it is not the Done
Thing.
b) On the left side of a
definition, the variable must be followed by an underline.
This is how Mathematica
distinguishes between a function and the value of a
function.
c) One normally writes := for
the equals sign in making a definition. There are
occasions when the ordinary
equals sign may be used, but a rule of thumb for
definitions is to use : =.
A function that is defined by
giving individual values instead of a formula can
be defined in Mathematica by
doing just that. For example, the function F in
Example 39.3.3 can be defined
by entering
codomain 56
domain 56
equivalent 40
function 56
theorem 2
F [1] := 4; F [2] := 5; F
[3] := 5 (39.3)
(When commands are strung
together with semicolons in this way, Mathematica
answers with the last value, 5
in this case. These commands could have been
entered on separate lines.)
Mathematica does not keep track
of the domain and codomain of the function.
If you try to evaluate the
function at an input for which it has not been defined,
you get back what you typed.
For example, if you had entered only the commands
in (39.3) and then typed F
[6] , Mathematica would return F [6] .
39.5 More about the input to a
function
Let’s look at the function G of
Example 39.3.1 again. We can calculate that G( 3) =
20. Since 1 + 2 = 3, it follows
that (7(1 + 2) = 20. Since \/9 = 3, it follows that
G(y/ 9) = 20. It is the element
x (here 3) of the domain (here R) that is being given
as input to G, not the name of
the element. It doesn’t matter how you represent
3, the value of G at 3 is still
20.
This is summed up by the
following theorem:
39.6 Theorem: The Substitution
Property
Let F : A—t B be any function,
and suppose that a G A and a' £ A. If
a = a' , then F(a) = F(a') .
The last sentence of
Specification 39.2 can be stated more precisely this way:
39.7 Theorem: How to tell if
functions are equal
If Fj : Aj — >■ Bj , (i = 1,2),
are two functions, then
(Fi = F 2 ) ^
60
codomain 56
domain 56
equivalence 40
function 56
39.7.2 Exercise Suppose F : A—$
B and G : B . What do you have to do to
prove that F / G?
39.7.3 Warning Since Formula
(39.4) is an equivalence, this means that the func-
tion S : R — > R for which S(x)
= x 2 is not the same as the function T : R — > R + for
which T (x) = x 2 . They have
the same domain and the same value at every element
of the domain, but they do not
have the same codomain. This distinction is often
not made in the literature. In
some theoretical contexts it is vital to make it, but
in others (for example
calculus) it makes no difference and is therefore quite rightly
ignored. In this text we are
purposely making all the distinctions made in a sizeable
fraction of the research
literature.
39.8 The abstract idea of
function
As was noted previously, the
specification given for functions says nothing about
the formula for the function.
The function G in Example 39.3.1 was defined by the
formula G(x) = x 2 + 2x + 5 ,
but the definition of the function called F in Exam-
ple 39.3.3 never mentioned a
formula.
Until late in the nineteenth
century, functions were usually thought of as defined
by formulas. However, problems
arose in the theory of Fourier analysis which made
mathematicians require a more
general notion of function. The definition of func-
tion given here is the modern
version of that more general concept.lt replaces the
algorithmic and dynamic idea of
a function as a way of computing an output value
given an input value by the
static, abstract concept of a function as having a domain,
a codomain, and a value lying
in the codomain for each element of the domain. Of
course, often a definition by
formula will give a function in this modern sense. How-
ever, there is no requirement
that a function be given by a formula.
The modern concept of function
has been obtained from the formula-based idea
by abstracting basic properties
the old concept had and using them as the basis of
the new definition. This
process of definition by abstracting properties is a major
tool in mathematics, and you
will see more examples of it later in the book (see
Chapter 51, for example).
The concept of function as a
formula never disappeared entirely, but was studied
mostly by logicians who
generalized it to the study of function-as-algorithm. (This
is an oversimplification of
history.) Of course, the study of algorithms is one of
the central topics of modern
computer science, so the notion of function-as-formula
(more generally, function-as-
algorithm) has achieved a new importance in recent
years.
Nevertheless, computer science
needs the abstract definition of function given
here. Functions such as sin may
be (and quite often are) programmed to look up
their values in a table instead
of calculating them by a formula, an arrangement
which gains speed at the
expense of using more memory.
39.7.1 Method
To show that two functions are
the same you have to show they have the
same domain, the same codomain
and for each element of the domain
they have the same value.
61
Cartesian product 52
coordinate 49
definition 4
domain 56
fact 1
function 56
graph (of a func-
tion) 61
implication 35, 36
ordered pair 49
single- valued 61
subset 43
usage 2
40.1.2 Example For the function
G of Example 39.3.1, the graph
r(G) = j(x,G(x)) | x € R j = |
(x,y) \ x G R A y = x 2 + 2x + 5|
which is a subset of R x R.
T(G) is of course what is usually called the graph of G
in analytic geometry — in this
case it is a parabola.
40.1.3 Example The graph of the
function F of Example 39.3.3 is
{(1,4), (2,5), (3,5)}
40.2 Properties of the graph of
a function
Using the notion of graph,
Specification 39.2 can be reworded as requiring the
following statements to be true
about a function F : A-^ B:
GS.l domF is exactly the set of
first coordinates of the graph, and
GS.2 For every aei, there is
exactly one element b of B such that (a, b) 6 F(F) .
40.2.1 Fact GS-2 implies that,
for all agi and b E B ,
^(a,6) € r(F) A (a,b') G T(F)^
=>• b = b ' (40.1)
40.2.2 Usage The requirement of
formula (40.1) is sometimes described by saying
that functions have to be
single- valued.
40.2.3 Warning Do not confuse
the property of being single- valued with the Sub-
stitution Property of Theorem
39.6, which in terms of the graph can be stated this
way: For all a & A and b £ B ,
40. The graph of a function
40.1 Definition: graph of a
function
The graph of a function F : A—$
B is the set
|(a,F(a)) | a G A j
of ordered pairs whose first
coordinates are all the elements of A with
the second coordinate in each
case being the value of F at the first
coordinate. The graph of F is
denoted by T(F)
40.1.1 Fact T(F) is necessarily
a subset of Ax B .
a,b) €T(F) A a = a') =>
(a',b)eT
(40.2)
62
Cartesian product 52
codomain 56
coordinate 49
functional prop-
erty 62
functional 62
function 56
graph (of a func-
tion) 61
implication 35, 36
include 43
opposite 62, 77, 220
ordered pair 49
usage 2
40.2.4 Remark When a function
goes from R to R the way the function G of
Example 39.3.1 does, its graph
can be drawn, and then the single- valued property
implies that a vertical line
will cross the graph only once. In general, you can’t draw
the graph of a function (for
example, the length function defined on words, as in
Example 39.3.4).
40.2.5 Remark Not every set of
ordered pairs can be the graph of a function. A
set P of ordered pairs is said
to be functional or to have the functional property
if
({a,b)£P A (a.i'Je?)
b = b'
(40.3)
Of course, Formula (40.1)
above says that the graph of a function is functional.
Conversely, if a set P of
ordered pairs is functional, then there are sets A and B
and a function F : B for
which T (F) = P . F is constructed this way:
FC.l A must be the set of
first coordinates of pairs in P .
FC.2 B can be any set
containing as elements all the second coordinates of pairs
in P.
FC.3 For each a £ A, define
F(a) = b, where (a, b) is the ordered pair in P with
a as first coordinate:
there is only one such by the functional property.
40.2.6 Exercise For A =
{1,2, 3, 4}, B = {3, 4,5,6} , which of these sets of ordered
pairs is the graph of a
function from A to B?
a) {(1,3), (2, 3), (3, 4),
(4, 6)}.
b) {(1,3), <2, 3), <4, 5),
<4, 6)}.
c) {(1,3), (2, 3), (4, 6)}.
d) {(1,3), (2, 4), (3, 5),
(4, 6)}.
(Answer on page 245.)
40.2.7 Exercise If P C A x
B , then the opposite of P is the set P op = {(6, a) \
(a,b) £ P | . Give examples
of:
a) a function F : A—t B for
which (T(F)) op is the graph of a function.
b) a function G : A—> B for
which (r(G)) op is not the graph of a function.
40.2.8 Exercise Create a
Mathematica command InGraphQ with the property
that the expression
InGraphQ [F , {x , y] } returns True if (. x,y ) &T(F) and False
otherwise.
40.2.9 Usage
a) In mathematical texts in
complex function theory, and in older texts in general,
functions are not always
assumed single- valued.
b) As you can see, part FD.
2 requires T(F) to have the functional property. In
texts which do not require
that a function’s codomain be specified, a function
is often defined simply as
a set of ordered pairs with the functional property.
63
40.3 Explicit definitions
of function
In many texts, the concept
of function is defined explicitly (as opposed to being
given a specification) by
some such definition as this: A function F is an ordered
triple (A,B,T(F)} for which
FD.l A and B are sets and
T(F) C A x B , and
FD.2 If a € A, then there
is exactly one ordered pair in T(F) whose first coordinate
is a.
41. Some important types of
functions
41.1.1 Identity function
For any set A, the identity function id a'-A-^t A is
the function that takes an
element to itself; in other words, for every element a £ A,
idyi (a) = a- Thus its
graph is the diagonal of Ax A (see 37.3).
41.1.2 Warning Do not
confuse the identity function with the concept of identity
for a predicate of Section
13.1.2, or with the concept of identity for a binary operation
of Section 45.
binary operation 67
codomain 56
constant function 63
coordinate 49
diagonal 52
domain 56
empty function 63
empty set 33
function 56
graph (of a func-
tion) 61
identity function 63
identity 72
include 43
inclusion function 63
ordered pair 49
ordered triple 50
take 57
tuple 50, 139, 140
41.1.3 Inclusion
function If A C B , then there is an inclusion function
inc :A—t B which takes
every element in A to itself regarded as an element of B .
In other words, inc(a)
= a for every element a £ A. Observe that the graph of inc
is the same as the
graph of idyi and they have the same domain, so that the only
difference between them
is what is considered the codomain ( A for idyi , B for the
inclusion of A in B ) .
41.1.4 Constant
function If A and B are nonempty sets and b is a specific
element of B , then the
constant function Cf,: A — > B is the function that takes
every element of A to
6; that is, Cb(a) = b for all a £ A. A constant function from
R to R has a horizontal
line as its graph.
41.1.5 Empty function
If A is any set, there is exactly one function E : 0 — > A.
Such a function is an
empty function. Its graph is empty, and it has no values.
“An identity function
does nothing. An empty function has nothing to do.”
41.1.6 Coordinate
function If A and B are sets, there are two coordinate
functions (or
projection functions) pi : A x B — > A and p 2 : A x B -> B . The
function p t takes an
element to its i th coordinate ( i = 1,2). Thus for a £ A and
b 6 B , pi (a, b) = a
and p 2 {a,b) = b. More generally, for any Cartesian product
ntiA there are n
coordinate functions; the ith one takes a tuple (ai,... ,a n ) to
cii .
41.1.7 Binary
operations The operation of adding two real numbers gives a
function
+ : R X R ->• R
which is an example of
a binary operation, treated in detail in Chapter 45.
64
anonymous nota-
tion 64
constant function 63
function 56
graph (of a func-
tion) 61
identity function 63
identity 72
inclusion function 63
lambda notation 64
41.1.8 Exercise For
each function F:A-^B, give F( 2) and F( 4).
a) A = B = R, F is the
identity function.
b) A = B = R, F = C 42
(the constant function).
c) A = R+,B = R, F is
the inclusion function.
(Answer on page 245.)
41.1.9 Exercise Give
the graphs of these functions. A = {1,2,3}, B = {2,3}.
a) id, 4 .
b) The inclusion of B
into A.
c) The inclusion of B
into Z .
d) C 3 :A-+B.
e) pi : Ax B A.
(Answer on page 245.)
42. Anonymous notation
for functions
The curly-brackets
notation for sets has the advantage that it allows you to refer to
a set without giving it
a name. For example, you can say, “{1,2,3} has three ele-
ments,” instead of,
“The set A whose elements are 1 , 2 and 3 has three elements.”
This is useful when you
only want to refer to it once or twice. A notation which
describes without
naming is called anonymous notation.
The notation we have
introduced for functions does not have that advantage.
When the two versions
of the squaring function were discussed, it was necessary to
call them S and T in
order to say anything about them.
42.1 Lambda notation
Two types of anonymous
notation for functions are used in mathematics. The older
one is called lambda
notation and is used mostly in logic and computer science.
To illustrate, the
squaring function would be described as “the function Xx.x 2 ”.
The format is: A, then
a letter which is the independent variable, then a period,
then a formula in terms
of the independent variable which gives the value of the
function. In the A-
notation, the variable is bound and so can be changed without
changing the function:
Xx.x 2 and A t.t 2 denote the same function.
42.1.1 Example The
function defined in Example 39.3.1 is Xx.(x 2 + 2x + 5) .
42.1.2 Example On a set
A, the identity function idyi is Xx.x.
65
42.2 Barred arrow
notation
The other type of
anonymous notation is the barred arrow notation, which has
in recent years gained
wide acceptance in pure mathematics and appears in some
texts on computer
science, too. In this notation, the squaring function would be
called the function x H
> x 2 : R — > R, and the function in Example 39.3.1 could be
written x +->• x 2 + 2x
+ 5 .
The barred arrow tells
you what an element of the domain is changed to by the
function. The straight
arrow goes from domain to codomain, the barred arrow from
element of the domain
to element of the codomain.
42.2.1 Example On a set
A , the identity function id^ is x H > x : A — > A .
42.2.2 Other notations
One would write Function [x,x"2] or #"2& in Mathe-
rnatica for x H > x 2
or Xx.x 2 . The # sign stands for the variable and the & sign at
the end indicates that
this is a function rather than an expression to evaluate. More
complicated examples
require parentheses; for example, x H > x 2 + 2x + 5 becomes
(#"2+2 #+5) &.
42.2.3 Exercise Write
the following functions using A notation and using barred
arrow notation. A and B
are any sets.
a) F : R — > R given by
F(x) = x 3 .
b) p\\ Ax B — >• A.
c) Addition on R.
(Answer on page 245.)
anonymous nota-
tion 64
barred arrow nota-
tion 65
characteristic func-
tion 65
codomain 56
constant function 63
definition 4
domain 56
even 5
extension (of a
predicate) 27
fact 1
function 56
identity function 63
identity 72
integer 3
lambda notation 64
predicate 16
subset 43
43. Predicates
determine functions
43.1 Definition:
characteristic function
Let A be a set. Any
subset B of A determines a characteristic
function : A -> {TRUE,
FALSE} defined by requiring that Xb ( x ) =
TRUE if x € B and Xb( x
) = FALSE if x i B.
43.1.1 Example If A =
{1,2, 3, 4} and B = {1,4} then Xb(1) = TRUE and
Xs(2) = FALSE.
43.1.2 Fact xjf is th e
constant function which is always FALSE, and Xa i s
constant TRUE.
43.1.3 Predicates as
characteristic functions Since the extension of a predi-
cate is a subset of its
data type, the truth value of a predicate is the characteristic
function of its
extension. For example, the statement “n is even” (about integers)
is TRUE if n is even
and FALSE otherwise, so that the value of the characteristic
function of the subset
E of Z consisting of the even integers is the truth value of
the predicate “n is
even”.
66
Cartesian product 52
characteristic func-
tion 65
constant function 63
definition 4
extension (of a
predicate) 27
function 56
graph (of a func-
tion) 61
integer 3
odd 5
predicate 16
subset 43
43.1.5 Exercise Suppose
that a predicate P regarded as the characteristic func-
tion of its extension
is a constant function. What can you say about PI
44. Sets of functions
As mathematical
entities, functions can be elements of sets; in fact the discovery of
function spaces, in
which functions are regarded as points in a space, was one of the
great advances of
mathematics.
44.1 Definition: D J ^
If A and B are sets,
the set of all functions F : A — >■ B is denoted B ^ .
44.1.1 Warning The
notation B ^ refers to the functions from A to 11, from the
exponent to the base.
It is easy to read this notation backward.
44.1.2 Remark Remark
97.1.5, page 139, and Theorem 122.3, page 188, explain
why the notation B ^ is
used.
44.1.3 Example The
function G of Example 39.3.1 is an element of the set R ,
and the function of
Example 39.3.3 is an element of the set
{2,4,5,6}{ 1,2,3 }
44.1.4 Example The
function + : R x R — > R is an element of R^ x ^ .
Predicates with more
than one variable similarly correspond to characteristic
functions of subsets of
Cartesian products. Thus the truth value of the statement
“ rn < n” (about
integers) is the characteristic function of the subset
| (m,n) | m < j
of Z x Z .
43.1.4 Exercise Give
the graphs of these functions. A = {1,2,3}, B = { 2,3}.
a) Xb : A ->•{ TRUE,
FALSE}.
b) The predicate “n is
odd” where n is an element of A, regarded as a function
to {TRUE, FALSE}.
c) + : B x B Z.
(Answer on page 245.)
67
44.1.5 Exercise Let A =
{1,2, 3,4,5} . For each item in the first column, state
which of the items in
the second column it is an element of.
a) id R 1) R R
b) the inclusion of A
in Z 2)
c) (1,2,1) 3) Rx Z x R
d) .t e- )• x 2 : R — >
R 4) (R+) R
(Answer on page 245.)
45. Binary operations
45.1 Definition: binary
operation
For any set S, a
function S x S — > S is called a binary operation on
5 .
binary operation 67
Cartesian product 52
codomain 56
complement 48
definition 4
divide 4
domain 56
function 56
identity 72
inclusion function 63
intersection 47
powerset 46
real number 12
right band 67
take 57
unary operation 67
45.1.1 Remark The
domain of a binary operation is the Cartesian square of
its codomain. Thus a
binary operation on a set S is an element of the function
set x ^ . i n
particular, a function G : Ax B — >• C is a binary operation only if
A = B = C.
45.1.2 Example The
function that takes (1,2) to 1, and (1,1), (2,1) and (2,2)
all to 2 is a binary
operation on the set {1,2}.
45.1.3 Example The
usual operations of addition, subtraction and multiplication
are binary operations
on the set R of real numbers. Thus addition is the function
(x, y) i-4 x + y : R x
R — > R
45.1.4 Example Division
is a function from R x (R — {0}) to R, and so does not
fit our definition of
binary operation. Restricted to the nonzero reals, however, it
is a function from R —
{0} x R — {0} to R — {0} (this says if you divide a nonzero
number by another one,
you get a nonzero number), and so is a binary operation
on R — {0} .
45.1.5 Example For any
set A, union and intersection are binary operations on
VA. This means that
each of union and intersection is a function from VA x VA
to VA (not from Ax A to
A).
45.1.6 Example For any
set A, define the binary operation P on A by requiring
that aPb = b for all a
and b in A . P is called the right band on A .
45.1.7 Unary operations
In the context of abstract algebra, a function from a
set A to A is called a
unary operation on A by analogy with the concept of
binary operation.
45.1.8 Example Taking
the complement of a set is a unary operation on a pow-
erset.
68
argument 57
binary operation 67
function 56
infix notation 68
negative integer 3
Polish notation 68
postfix notation 68
prefix notation 68
reverse Polish nota-
tion 68
take 57
45.1.9 Example The
function — : Z — >■ Z (similarly for R) that takes a number
r to its negative — r
is a unary operation on R. This is distinct from the binary
operation of
subtraction (m,n) i-4 m — n.
46. Fixes
46.1 Prefix notation
I have normally written
the name of the function to the left of the argument (input
value), thus: E(x).
This is called prefix notation for functions and is familiar
from analytic geometry
and calculus texts.
46.1.1 Parentheses
around the argument Trigonometric functions like sinx
are also written in
prefix notation, but it is customary to omit parentheses around
the argument. (Pascal
and many other computer languages require the parentheses,
however, and
Mathematica requires square brackets). Many mathematical writers
omit the parentheses in
other situations too, writing “Ex” instead of “E(x)”. It is
important not to
confuse evaluation written like this with multiplication.
46.2 Infix notation
Many common binary
operations are normally written between their two arguments,
“a + 6” instead of “+
(a,6)”. This is called infix notation and naturally applies
only to functions with
two arguments.
46.2.1 Example The
expression 3 — (5 + 2) is in infix notation. In prefix notation,
the same expression is
—(3, +(5,2)).
46.3 Postfix notation
Some authors write
functions on the right, for example “xE” or “(x)E” instead
of “E(x)”. This is
called postfix notation. This has real advantages which will
become apparent when we
look at composition in Chapter 98.
46.4 Polish notation
When binary operations
are written in either prefix or postfix notation, parentheses
are not necessary to
resolve ambiguities. In infix notation, for example, parentheses
are necessary to
distinguish between “a + 6*c” (which is the same as “a+ (b*c) n )
and “(a + 6)*c”. In
prefix notation, “a + 6*c” can be written “+ a * b c” and
u (a + b)*c ,, can be
written “* + a b c”. Note the use of spaces to separate
the items. This is
particularly important when multidigit constants are used: for
example 35 22 + in
postfix notation returns 57.
Writing functions of
two or more arguments using prefix notation and no paren-
theses is called Polish
notation after the eminent Polish logician Jan Lukasiewicz,
who invented the
notation in the 1920’s. Writing functions on the right which are
normally infixed,
without parentheses, is naturally called reverse Polish nota-
tion.
69
Most computer languages
use prefix and infix notation similar to that of ordinary
algebra. The language
Lisp uses prefix notation (with parentheses) and the various
dialects of Forth
characteristically use reverse Polish notation (no parentheses).
Either prefix or
postfix notation in a computer language makes it easier to write an
interpreter or compiler
for the language.
46.4.1 Example The
expression of Example 46.2.1 in prefix notation without
using parentheses is —
3 + 5 2 . In postfix notation it is 3 5 2 + — .
46.4.2 Example a + b +
c in reverse Polish notation can be written either as a b +
c+orasa6c + +.
46.4.3 Exercise Write
(35 + 22) (6 + 5) in reverse Polish notation. Use * for
multiplication. (Answer
on page 245.)
46.4.4 Exercise Write b
2 — 4aci in reverse Polish notation. Use * for multiplica-
tion and don’t use
exponents.
Fix notation in
Mathematica Mathematica gives the user control over whether
a function is written
in infix notation or not. For example, we remarked in Sec-
tion 14.4 that in
Mathematica one writes Xor[p,q] for the expression pXORg.
However, by putting
tildes before and after the name of a function in Mathematica,
you can use it as an
infix; thus you can write p ~Xor~ q instead of for Xor [p,q] .
A function F can be
used in postfix form by prefixing it with //. For example,
one can write Sqrt [2]
or 2 // Sqrt.
47. More about binary
operations
47.1 Notation
In discussing binary
operations in general, we will refer to an operation A on a
set A ; thus A : A x A
— > A. This operation will be used in infix notation, the way
addition and
multiplication are normally written, so that we write aAb for A (a,b) .
Using an unfamiliar
symbol like ‘A’ avoids the sneaky way familiar symbols like
“+” cause you to fall
into habits acquired by long practice in algebra (for example,
assuming commutativity)
that may not be appropriate for a given situation.
47.1.1 Warning Don’t
confuse A , representing a binary operation, with the diag-
onal A^ C Ax A defined
in Definition 37.3, page 52.
47.1.2 Multiplication
tables We will sometimes give a binary operation A on a
small finite set by
means of a multiplication table: For example, here is a binary
operation on the set
{a,b,c}.
b
c
binary operation 67
Cartesian product 52
diagonal 52
finite 173
function 56
include 43
infix notation 68
multiplication
table 69
postfix notation 68
prefix notation 68
70
associative 70
binary operation
definition 4
function 56
intersection 47
multiplication
table 69
postfix notation
powerset 46
prefix notation 68
real number 12
right band 67
union 47
67
The value of x Ay is in
the row marked x and the column marked y . This means
for example that a A b
= c and c A a = a .
47.1.3 Example The
binary operation of Example 45.1.2 is
68
2
2
47.1.4 Exercise
{1,2,3}.
Give the
multiplication table for the right band on the set
47.1.5 Exercise
Give the multiplication table for the operation of union on the
powerset of
{1,2,3}.
48. Associativity
48.1 Definition:
associative
A binary operation
A is associative if for any elements x, y, z of A,
x A (y Az) = (x Ay)
Az (48.1)
48.1.1 Remark In
ordinary functional notation (prefix notation), the definition
of associative says
A(x,A (y,z) = A(A (x,y,),z)) . In postfix notation: x y A z A =
x y z A A.
48.1.2 Example The
usual operations of addition and multiplication are asso-
ciative, but not
subtraction; for example, 3 — (5 — 7) / (3 — 5) — 7. The opera-
tion given in
47.1.2 is not associative; for example, (a A a) A c = b A c = a, but
a A (a Ac) = a Aa =
b.
48.1.3 Example For
any nonempty set X , union and intersection are associative
binary operations
in VX .
48.1.4 Example For
real numbers r and s , let max : II x II — > R and min : R x
R — > R be the
functions defined by: max(r, s ) is the larger of r and s and min(r, s)
the smaller. If r =
s then max(r, s) = min(r, s) = r = s. These are both associative
binary operations
on the set R of real numbers.
48.1.5 Exercise
Prove that for any set S , union is an associative binary operation
on VS. (Answer on
page 245.)
48.1.6 Exercise
Prove that for any set S, intersection is an associative binary
operation on VS .
48.1.7 Exercise
Show that the right band operation on any set A is associative.
71
48.1.8 Exercise
Find a binary operation A on some set A with the property
that, for some
element a € A,
(a A a) A a) / a A
(a A a)
48.1.9 Exercise Is
the binary operation A given by this table associative? Give
reasons for your
answer.
a
a
associative 70
binary operation 67
commutative 71
definition 4
General Associative
Law 71
max 70
subset 43
48.1.10 Exercise
Prove that max: R x R — > R is associative (see Example 48.1.4).
48.2 The general
associative law
If A is an
associative operation on A, then it is associative in a more general sense,
in that it
satisfies the General Associative Law: Any two meaningful products
involving A and
ai,a2,--,a n (names of elements of A) in that order denote the
same element of A.
48.2.1 Example If a
A (bA c) = (aAb)Ac, then all five meaningful ways of writ-
ing the product of
four elements are the same:
a A (b A (c A d)) =
aA((bAc)Ad) = (aAb)A(cAd)
=
((aAb)Ac)Ad=(aA(bAc))Ad
49. Commutativity
49.1 Definition:
commutative
A binary operation
A on a set A is commutative if for all x,y £ A,
x Ay = y Ax.
49.1.1 Example The
operations of addition and multiplication, but not subtrac-
tion, are
commutative operations on R.
49.1.2 Example The
binary operations mentioned in Examples 48.1.2, 48.1.3 and
48.1.4 are
commutative.
49.1.3 Exercise Let
C be a set. Define the binary operation A for all subsets A
and B of C by
AAB = (A(J B) — (An
B)
a) Show that A is
commutative.
b) Show that AAB =
(A — B) U (B — A) .
72
associative 70
binary operation
commutative 71
definition 4
even 5
identity function
identity 72
integer 3
max 70
powerset 46
right band 67
unity 72
49.1.4 The General
Commutative Law There is a general commutative law
67 analogous to the
general associative law: It says that if A is commutative and
associative, then
the names a\,...,a n in an expression a\ A a 2 A ... A a n can be
rearranged in any
way without changing the value of the expression. We will not
prove that law
here.
Do
50. Identities
50.1 Definition:
identity
If A is a binary
operation on a set A , an element e is a unity or
identity for A if
for all x G A,
xAe = eAx = x
(50.1)
50.1.1 Warning
Don’t confuse the concept of identity for a binary operation with
the concept of an
identity function in 41.1.1, page 63. These are two different ideas,
but there is a
relationship between them (see 98.2.3, page 141).
50.1.2 Example The
binary operation of Example 45.1.2 has no identity.
50.1.3 Example The
number 1 is an identity for the binary operation of multi-
plication on R ,
and 0 is an identity for + .
50.1.4 Exercise
Which of these binary operations (i) is associative, (ii) is com-
mutative, (iii) has
an identity?
A
a
c
c
( 1 ) ( 2 )
(Answer on page
245.)
50.1.5 Exercise
Show that the right band operation on a set with more than one
element does not
have an identity.
50.1.6 Example The
binary operation of multiplication on the set of even integers
is associative and
commutative, but it has no identity.
50.1.7 Exercise Let
S be any set. What is the identity element for the binary
operation of union
on VS 1 (Answer on page 245.)
50.1.8 Exercise Let
S be any set. What is the identity element for the binary
operation of
intersection on V SI
50.1.9 Exercise
Does max :RxR-^R have an identity? What about max : R + x
R + — > R + defined
the same way?
The basic fact
about identities is:
73
50.2 Theorem:
Uniqueness theorem for identities
If A is a binary
operation on a set A with identity e, then e is the only
identity for A .
Proof This follows
immediately from Definition 50.1: if e and / are both identi-
ties, then e = eAf
because / is an identity, and eAf = f because e is an identity.
50.2.1 Exercise
Give a rule of inference that allows one to conclude that a certain
object is an
identity for a binary operation A .
50.2.2 Exercise
(hard) Find all the binary operations on the set {a, b} , and
state whether each
one is associative, is commutative, and has an identity element.
51. Relations
The mathematical
concept of relation is an abstraction of the properties of relations
such as “=” and “<”
in much the same way as the modern concept of function was
abstracted from the
concrete functions considered in freshman calculus, as described
in Section 39.8.
associative 70
binary operation 67
Cartesian product
52
commutative 71
definition 4
equivalence 40
equivalent 40
fact 1
function 56
identity 72
predicate 16
proof 4
relation 73
rule of inference
24
subset 43
theorem 2
type (of a vari-
able) 17
usage 2
51.1 Definition:
binary relation
A binary relation a
from a set A to a set B is a subset of A x B . If
(a, b) G a , then
one writes a ab.
51.1.1 Remark Any
subset of Ax B for any sets A and B is a binary relation
from A to B.
51.1.2 Fact
Definition 51.1 gives the following equivalence, which describes two
different ways of
writing the same thing:
a ab 44 (a, 6) G a
(51.1)
51.1.3 Usage A
relation corresponds to a predicate with two variables, one of
type A and the
other of type B : the predicate is true if a ab (that is, if (a, b) G a)
and false
otherwise. Logic texts often define a relation to be a predicate of this type,
but the point of
view taken here (that a relation is a set of ordered pairs) is most
common in
mathematics and computer science.
51.1.4 Example Let
A = {1,2, 3, 6} and B = {1,2, 3, 4, 5}. Then
a= | <2, 2), (1,5),
(1,3), (2,5), <2, 1) j
is a binary
relation from A to B . For this definition, we know 1 cc 5 and 2 a 1 but
it is not true that
1 a 2 .
74
Cartesian product
coordinate func-
tion 63
definition 4
digraph 74, 218
divide 4
empty relation 74
finite 173
function 56
include 43
ordered pair 49
powerset 46
relation 73
subset 43
total relation 74
52 51.1.5 Exercise
Write all ordered pairs in the relation from A to B:
a) A = {1,2,3} , B
= {1,3,5} . a is “/”.
b) A = {2, 3, 5,
7}, B = {1,2,3,4,5,6,7,8,9,10}, a is “divides”.
c) A = {0,1, 2, 3},
B = {1,2,3}, a is “divides”.
(Answer on page
245.)
51.2 Picturing
relations
A relation on a
small finite set can be exhibited by drawing dots representing the
elements of A and B
and an arrow from x to y if and only if x ay . Here is the
relation in Example
51.1.4 exhibited in this way:
1 2
v v
3 5
Such a picture is
called the digraph representing the relation. Digraphs are studied
in depth in
Chapters 144 and 151.
51.2.1 Example Two
not very interesting binary relations from A to B are the
empty relation 0 C
Ax B and the total relation Ax B . If E denotes the
empty relation,
then aE6 is false for any a £ A and b G B , and if T denotes the
total relation, aT6
is true for any a € A and b £ B .
51.2.2 Example In a
university, the pairs of the form (student, class) where the
student is
registered for the class form a relation from the set of students to the set
of classes.
51.3 Definition:
Rcl(A,i?)
The set of all
relations from A to B is denoted by Rel(A,R).
51.3.1 Remark By
Definition 51.1, Rel(A,R) is the same thing as the powerset
V(A x B) ; the only
difference is in point of view.
51.4 The
projections from a relation
A relation a from A
to B is a subset of Ax B by definition, so there are func-
tions Pi : a — >■ A
, P2 : a — >■ B , which are the restrictions of the coordinate func-
tioncoordinate
(projection) functions (see 41.1.6, page 63) from Ax B to A and to
B.
51.4.1 Example If a
is defined as in Example 51.1.4, then : a — > {1,2, 3, 6}
and p2 : a ► {1,2,
3, 4, 5} . In particular, p“((l,5)) = 1 .
75
52. Relations on a
single set
52.1 Definition:
relation on a set
If ct is a relation
from A to A for some set A, then a is a subset of
Ax A. In that case,
a is called a relation on A.
52.1.1 Example “>”
is a relation on R; one element of it is (5,3).
52.1.2 Example A
particular relation that any set A has on it is the diagonal
A a ; A a = {(a, a)
\ a E A}. A a is simply the equals relation on A. Don’t confuse
this with the use
of A to denote an arbitrary binary operation as in Chapter 45.
52.1.3 Exercise Let
A = {1, 2, 3,4} . Write out all the ordered pairs in the relation
R on A if
a) aRb o a < b
b) aRb a = b
c) aR6 6 = 3.
d) aR6 a and 6 are
both odd.
(Answer on page
245.)
Cartesian product
52
codomain 56
definition 4
diagonal 52
domain 56
equivalent 40
functional prop-
erty 62
functional relation
75
function 56
graph (of a func-
tion) 61
implication 35, 36
odd 5
ordered pair 49
relation on 75
relation 73
subset 43
usage 2
53. Relations and
functions
53.1 Functional
relations
The graph T(F) of a
function F : A — > B is a binary relation from A to B . It
relates a E A to 6
£ B precisely when 6 = F(a ) . Of course, not any relation can be
the graph of a
function: to be the graph of a function, a binary relation a from A
to B must have the
functional property described in 40.2:
(a a b and a a b')
=4> b = b' (53.1)
A relation
satisfying Equation (53.1) is called a functional relation.
This requirement
can fail because there are ordered pairs (a, 6) and (a,b r ) in a
with 6 yf b' . Even
if it is satisfied, a may not be the graph of a function from A to
B , since there may
be elements a E A for which there is no ordered pair (a, 6) E a .
However, a
functional relation in Rel(A,L>) is always the graph of a function whose
domain is some
subset of A .
53.1.1 Usage For
some authors a function is simply a functional relation. For
them, the domain
and codomain are not part of the definition.
53.1.2 Exercise
Which of these are functional relations?
a) {(1,3), (2,3),
(3,4)}.
b) {(1,1), (1,2),
(2, 3)}.
c) {(x,y/x) | igR}.
d) {(y/x,x) | iGR}.
e) {(y/x,x) j x G R
+ } .
(Answer on page
245.)
76
definition 4
empty set 33
equivalent 40
function 56
integer 3
ordered pair 49
powerset 46
relation 73
singleton 34
subset 43
As we have seen,
the concept of relation from A to B is a generalization of the
concept of function
from A to B . In general, for a given a & A there may be no
ordered pairs ( a,b
) E a or there may be more than one. Another way of saying this
is that for a given
element a E A, there is a set {b E B \ (a,b) E a}. For a to be
a function from A
to B , each such set must be a singleton. In general, a relation
associates a
(possibly empty) subset of B to each element of A.
53.2 Definition:
relation as function to powerset
If a is a relation
from A to B, let a*:A-^-VB denote the function
defined by a* (a) =
{6 E B \ (a, b) E a} .
53.2.1 Remark
Definition 53.2 gives us a process that constructs a function from
A to the powerset
of B for each relation from A to B . For any a E A and b E B ,
b E a* (a) 44 aab
53.2.2 Example For
the relation a of Example 51.1.4, we have a*(l) = {3,5},
a*(2) = {1,2,5} and
a*( 3) =0.
53.2.3 Exercise
Write the function a* : A -^VB corresponding to the relation in
Problem 51.1.5(a).
(Answer on page 245.)
Conversely, if we
have a function F : A— >VB , we can construct a relation:
53.3 Definition:
relation induced by a function to a pow-
erset
Given F : A — >
VB , the relation ap from A to B is defined by aapb if
and only if b E
F(a) .
53.3.1 Remark In
the preceding definition, it makes sense to talk about b E F(a ) ,
because F(a) is a
subset of B.
53.3.2 Example Let
F : {1,2,3} — »• "P({1,2,3}) be defined by F(1) = {1,2},
F{ 2) = {2} and
F(3) = 0. Then a F = {( 1,1), (1,2), (2,2)}.
53.3.3 Exercise A
function F : Z — y PZ has F{ 1) = {3,4}, F{ 2) = {1,3,4},
F(— 666) = {0}, and
F(n) = 0 for all other integers n. List the ordered pairs
in the
corresponding relation a.p on Z. (Answer on page 245.)
53.3.4 Exercise Let
F be the function of Problem 39.3.9, page 58. List the
ordered pairs in ay
that have 6 as first element.
77
54. Operations on
relations
54.1 Union and
intersection
Since relations
from A to B are subsets of A x B , all the usual set operations such
as union and
intersection can be performed on them.
54.1.1 Example On
R, the union of Ar and “<” is (of course!) “< ”, and the
intersection of “<”
and “>” is Ar. These statements translate into the obviously
true statements
r<s»(r<sVr = s)
and
(r<sAr>s)<S>r = s
definition 4
equivalence 40
equivalent 40
fact 1
function 56
include 43
intersection 47
near 77
opposite 62, 77,
220
powerset 46
reflexive 77
relation 73
subset 43
union 47
54.2 Definition:
opposite
The opposite of a
relation a £ Rel(A,.E>) is the relation a op E Rel(.B, A)
defined by a op = j
(6, a) | (a, b) E aj.
54.2.1 Fact This
definition gives an equivalence
ba op a aab
It follows that a i
— > a op : Rel(A,H) — >• Rel(H, A) is a function.
54.2.2 Example On R
the opposite of “>” is “<” and the opposite of “<” is
Of course, for any
set A, the opposite of Aq is Aq .
55. Reflexive
relations
55.1 Definition:
reflexive
Let a be a binary
relation on 1. a is reflexive if a a a for every
element a E A.
55.1.1 Example Aq
is reflexive on any set A, and the relation “< ” is reflexive
on R.
55.1.2 Example On
the powerset of any set the relation “C” is reflexive.
55.1.3 Example The
relation “<” is not reflexive on R, and neither is the rela-
tion S “is the
sister of” on the set W of all women, since no one is the sister of
herself.
55.1.4 Example
Another important type of reflexive relation are the relations
like xM y \x — y\ <
0.1 , defined on R. “A7” stands for “near”. The choice of 0.1
as a criterion for
nearness is not important; what is important is that it is a fixed
number.
The relations S and
M will be used several times below in examples.
78
definition 4
divide 4
equivalent 40
fact 1
implication 35, 36
nearness relation
77
reflexive 77
relation 73
sister relation 77
symmetric 78, 232
vacuous 37
55.1.5 Fact Let a
be a relation on a set A. Then a is reflexive if and only if
Aq C a.
55.1.6 Remark The
statement that a relation a defined on a set A is reflexive
depends on both a
and A. For example, the relation
{( 1 , 1 ), ( 1 , 2
), ( 2 , 2 )}
is reflexive on
{1,2} but not on {1,2,3}.
55.1.7 Warning It
is wrong to say that the relation a of 55.1.6 is “reflexive at 1
but not at 3”.
Reflexivity and irreflexivity are properties of the relation and the
set it is defined
on, not of particular elements of the set on which the relation is
defined. This
comment also applies to the other properties of relations discussed in
this section.
55.1.8 Fact The
digraph of a reflexive relation must have a loop on every node.
55.1.9 Exercise
Which of these relations is reflexive?
a) a = {(1, 1),
(2,2), (3,3)} on {1,2,3}.
b) a = {(1,1), (2,
2), (3, 3)} on N.
c) “divides” on Z.
d) a on R defined
by xay x 2 = y 2 .
(Answer on page
245.)
56. Symmetric
relations
56.1 Definition:
symmetric
A relation a on a
set A is symmetric if for all a, b £ A,
a ab => baa
56.1.1 Example The
equals relation on any set is symmetric, and so is the near-
ness relation J\f
(see Example 55.1.4). The sister relation S (Example 55.1.2) is
not symmetric on
the set of all people, but its restriction to the set of all women is
symmetric.
56.1.2 Warning It
is important to understand the precise meaning of the defini-
tion of symmetric.
It is given in the form of an implication: a ab =t- baa. Thus
(a) it could be
vacuously true (the empty relation is symmetric!) and (b) it does
not assert that a
ab for any particular elements a and b : that a is symmetric does
not mean (a ab) A
(6 a a) .
56.1.3 Remark The
digraph of a symmetric relation has the property that
between two
distinct nodes there must either be two arrows, one going each way, or
no arrow at all.
79
56.1.4 Exercise
Which of these relations is symmetric?
a) a = {<1,2), <2,
3), <1, 3), <2, 1), <4, 1)} on {1,2, 3, 4}.
b) a = {<1, 1), <2,
2), <3,3)} on N.
c) The empty
relation on N.
d) “is the brother
of” on the set of all people.
(Answer on page
245.)
56.1.5 Exercise
Show that if a relation a on a set A is not symmetric, then A
has at least two
distinct elements.
antisymmetric 79
definition 4
divide 4
implication 35, 36
include 43
nearness relation
77
negation 22
powerset 46
relation 73
symmetric 78, 232
vacuous 37
57. Antisymmetric
relations
57.1 Definition:
antisymmetric
A relation a on a
set A is antisymmetric if for all a,b G A,
(a a b Aba a) =4- a
= b
57.1.1 Warning
Antisymmetry is not the negation of symmetry; there are rela-
tions such as A
which are both symmetric and antisymmetric and others such as
“divides” on Z
which are neither symmetric nor antisymmetric.
57.1.2 Exercise
Prove that on any set A, Aa is antisymmetric.
57.1.3 Exercise
Prove that on Z, “divides” is neither symmetric nor antisym-
metric.
57.1.4 Remark The
digraph of an antisymmetric relation may not have arrows
going both ways
between two distinct nodes.
57.1.5 Example
Antisymmetry is typical of many order relations: for example,
the relations “<”
and “<” on R are antisymmetric. Orderings are covered in
Chapter 134.
57.1.6 Example The
inclusion relation on the powerset of a set is antisymmetric.
This says that for
any sets A and B , A C B and B C A together imply A = B .
57.1.7 Example The
relation “<” is vacuously antisymmetric, and on any set
S , A s is both
symmetric and antisymmetric.
57.1.8 Example The
nearness relation J\f is not antisymmetric; for example,
0.25A70.3 and
0.3A70.25, but 0.25/0.3.
80
antisymmetric 79
definition 4
equivalent 40
implication 35, 36
include 43
nearness relation
77
relation 73
sister relation 77
symmetric 78, 232
transitive 80, 227
vacuous 37
57.1.9 Exercise
Which of these relations is antisymmetric?
a) a = {<1,2), <2,
3), <3, 1), <2, 2}} on N.
b) “divides” on N.
c) > on R.
d) “is the brother
of” on the set of all people.
(Answer on page
245.)
57.1.10 Exercise
Show that if a relation a on a set A is not antisymmetric, then
A has at least two
distinct elements.
57.1.11 Exercise
Let a be a relation on a set A. Prove that a is antisymmetric
if and only if afla
op C Aq. (Another problem like this is Problem 84.2.5, page 124.)
58. Transitive
relations
58.1
Definition: transitive
A relation a on
A is transitive if for all elements a, b and c of A,
(a a b Abac) =>
aac
58.1.1 Example
All the relations Aq , “<”, “<” and “C” are obviously tran-
sitive. That
equals is transitive is equivalent to the statement from high-school
geometry that
two things equal to the same thing are equal to each other.
58.1.2 Example
The sister relation S is not transitive, not even on the set of
all women. Thus
Agatha may be Bertha’s sister, whence Bertha is Agatha’s sister,
but Agatha is
not her own sister. This illustrates the general principle that when a
definition uses
different letters to denote things, they don’t have to denote different
things. In the
definition of transitivity, a, b and c may be but don’t have to be
different.
58.1.3 Example
Nearness relations are not transitive.
58.1.4 Example
Let A be the set {{1,2}, {3}, 2, 6, {{1,3}, {1,2}}} The relation
“G” on A is not
transitive, since 2 £ {1,2} and {1,2} £ {{1,3}, {1,2}}, but 2 ^
{{1,3}, {1,2}}.
58.1.5 Warning
Transitivity is defined by an implication and can be vacuously
true. In fact,
all the properties so far have been defined by implications except
reflexivity.
And indeed the empty relation is symmetric, antisymmetric and transi-
tive!
81
58.1.6 Remark
The digraph of a transitive relation must have the property that
every “path of
length two” , such as
must be
completed to a triangle, like this:
antisymmetric
79
definition 4
equivalent 40
irreflexive 81
negation 22
reflexive 77
relation 73
symmetric 78,
232
transitive 80,
227
Paths are
covered formally in Section 149.
58.1.7 Exercise
Give an example of a nonempty, symmetric, transitive relation
on the set
{1,2} that is not reflexive.
58.1.8 Exercise
State and prove a theorem similar to Problem 56.1.5 for non-
transitive
relations.
58.1.9 Exercise
Let the relation R be defined on the set {x £ R | 0 < x < 1} by
xRy ■&3t(x + t
= y and 0 < t < 1)
Is R
transitive?
58.1.10
Exercise (hard) If possible, give examples of relations on the set {1,2, 3}
which have
every possible combination of the properties reflexive, symmetric, anti-
symmetric and
transitive and their negations. (HINT: There are 14 possible com-
binations and
two impossible ones.)
59. Irreflexive
relations
59.1
Definition: irreflexive
A relation a is
irreflexive if a a a is false for every a € A.
59.1.1 Example
The “<” relation on R is irreflexive.
59.1.2 Warning
Irreflexive is not the negation of reflexive: a relation might be
neither
reflexive nor irreflexive, such as for example the relation
a = {(1,1),
(1,2), (2, 2)}
on {1,2,3}.
82
antisymmetric
79
definition 4
divide 4
div 82
equivalent 40
integer 3
irreflexive 81
mod 82, 204
positive 3
reflexive 77
relation 73
remainder 83
symmetric 78,
232
transitive 80,
227
(d) (e) (/)
(Answer on page
245.)
59.1.4 Exercise
List the properties (reflexive, symmetric, antisymmetric, transi-
tive, and
irreflexive) of each relation.
a) “not equals”
on R.
b) x a y 44 x 2
= y 2 on R
c) x a y ^ x =
—y on R
d) x a y 44 x <
y 2 on R
e) “divides” on
N
f) “leaves the
same remainder when divided by 3” on N
g) {(1,1),
(2,3), (3,2), (3,4)} on {1,2, 3, 4}
(Answer on page
245.)
59.1.5 Exercise
Let /? be an irreflexive, antisymmetric relation on a set S. Show
that at most
one of the statements u ai3b ” and “6/3a” holds for any pair of elements
a,b of S.
60. Quotient
and remainder
Let m and n be
positive integers with n / 0 . If you divide n into m you get a
quotient and a
remainder; for example, if you divide 4 into 14 you get a quotient 3
and a remainder
2 . We will write the quotient when m is divided by n as m div n
and the
remainder as mnrodn, so that 14 div 4 = 3 and 14 mod 4 = 2. The basis
for the formal
definition given below is the property that 14 = 3x4 + 2.
The
following formal definition allows m and n to be negative as well as positive.
This has
surprising consequences discussed in Section 61.3.
59.1.3
Exercise List the properties (reflexive, symmetric, antisymmetric, transi-
tive, and
irreflexive) of the relations given by each picture.
83
60.1
Definition: quotient and remainder
Let m and n
be integers. Then q = mdivn and r = mmodn if and
only if q
and r are integers that satisfy both the following equations:
Q.l m = qn
+ r, and
Q.2 0 < r <
\n\ .
If q =
mdivn, then q is the quotient (of integers) when m is divided
by n . If r
= m mod n , then r is the remainder when m is divided by
n.
definition
4
div 82
integer 3
mod 82, 204
quotient
(of inte-
gers) 83
remainder
83
60.1.1
Remarks
a) It
follows from the definition that the equation
m =
(mdivn)n + (mmodn) (60.1)
is always
true for n / 0 .
b) On the
other hand, if n = 0, Q.2 cannot be true no matter what r is. In other
words,
“mdivO” and “mmodO” are not defined for any integer m.
60.1.2
Exercise Find the quotient (of integers) and remainder when m is divided
by n:
a) m = 2 ,
n = 4 .
b) m = 0, n
= 4.
c) m = 24,
n = 12.
d) rn = 37,
n = 12.
(Answer on
page 245.)
60.1.3
Warning For q to be ?ndivn and r to be mmodn, both Q.l and Q.2
must be
true. For example, 14 = 2 x 4 + 6 (so Q.l is satisfied with q = 2 and r = 6),
but 14 mod
4 / 6 because Q.2 is not satisfied.
60.1.4
Exercise Suppose that a and b leave the same remainder when divided
by m. Show
that a — b is divisible by m. (Answer on page 245.)
60.1.5
Exercise Suppose that a — b is divisible by m. Show that a and b leave
the same
remainder when divided by m.
60.1.6
Exercise Suppose that adivm = 6divm. Show that \a — b\<\m\.
60.1.7
Exercise Is the converse of Exercise 60.1.6 true? That is, if \a — b\ < \m \ ,
must it be
true that arnodm = b mod rri ?
The
following theorem is what mathematicians call an “existence and uniqueness”
theorem for
quotient and remainder.
84
divide 4
div 82
function 56
integer 3
mod 82, 204
negative
integer 3
nonnegative
integer 3
proof 4
quotient
(of inte-
gers) 83
remainder
83
theorem 2
60.2
Theorem: Existence and Uniqueness Theorem for
quotient
and remainder
For given
integers rn and n with n / 0 , there is exactly one pair of
integers q
and r satisfying the requirements of Definition 60.1.
60 . 2.1
Remark This theorem says that when jj/ 0 there is a quotient and a
remainder,
i.e. , there is a pair of numbers q and r satisfying Q.l and Q.2, and that
there is is
only one such pair.
60 . 2.2
Worked Exercise Suppose that m = 3n + 5 and n > 7 . What is mdivn?
Answer m
divn = 3. The fact that m = 3n + 5 and n > 7 (hence n > 5) means
that q = 3
and r = 5 satisfy the requirements of Definition 60.1.
60 . 2.3
Exercise Suppose a, b, m . and n are integers with m and n nonnegative
such that m
= (a + l)n + b + 2 and m div n = a. Show that b is negative. (Answer
on page
245.)
60 . 2.4
Exercise Suppose n > 0, 0 < s < n and n\s. Show that s = 0. (Answer
on page
246.)
There is a
connection between these ideas and the idea of “divides” from Defini-
tion 4.1
(page 4):
60.3
Theorem
If n 7 ^ 0
and rn mod n = 0 , then n\m.
Proof If
mmodn = 0, then by Q.l, m = (mdivn)n, so by Definition 4.1 (using
mdivn for
q), n divides m.
60.4 Mod
and div in Mathematica
To compute
mdivn in Mathematica, you type Quotient [m,n] , and to compute
mmodn, you
type Mod[m,n], You can if you wish place either of these function
names
between the inputs surrounded with tildes: m "Quotient" n is the same as
Quotient
[m,n] , and m "Mod" n is the same as Mod[m,n] .
60.5 Proof
of uniqueness
We will
prove that the quotient and remainder exist in Section 104.3.2, page 156.
It is
worthwhile to see the proof that the quotient and remainder are unique, since
it shows
how it is forced by Definition 60.1.
Suppose m =
qn + r = q'n + r' and both pairs (q,r) and ( q',r '} satisfy Q.2. We
must show
that the two ordered pairs are the same, that is, that q = q’ and r = r'.
By Q.2 we
have 0 < r < \n\ and 0 < r' < |n|. Since r and r' are between 0
and \n\ on
the number line, the distance between them, which is |r — r'\ , must also
be less
than n. A little algebra shows that
\r — r'\ =
| q' — q | |n|
It then
follows from Definition 4.1, page 4, that | r — r'\ is divisible by |n| . But a
non-
negative
integer less than |n| which is divisible by \n\ must be 0 (Exercise 60.2.4).
85
So r = r
' . Since qn + r = q'n + r ' , it must be that q = q ' , too. So there can be only
one pair of
numbers q and r satisfying Q.l and Q.2.
This proof
uses the following method.
60.5.1
Method
To prove
that an object that satisfies a certain condition is unique,
assume
there are two objects A and A! that satisfy the condition and
show that A
= A' .
characterize 85
div 82
integer 3
mod 82, 204
quotient
(of inte-
gers) 83
remainder
83
well-
defined 85
60.5.2
Exercise Use Definition 60.1 and Theorem 60.2 to prove that when 37 is
divided by
5, the quotient is 7 and the remainder is 2. (Answer on page 246.)
60.5.3
Exercise Use Definition 60.1 and Theorem 60.2 to prove that 115 div 37 =
3.
60.5.4
Exercise Suppose that fn = 36q + 40 . What is m mod 36? (Answer on
page 246.)
60.5.5
Exercise Prove that if q, m and n are integers and 0 <m — qn< \n\,
then q = m
div n .
60.5.6
Exercise Show that if a and b are positive integers and amod4 = 6mod
4 = 3, then
ab mod 4 = 1.
60.5.7
Exercise Prove that for any integer c, c 2 mod3 is either 0 or 1.
60.6 More
about definitions
Observe
that Definition 60.1 defines “mdivn” and “mrnodn” without telling you
how to
compute them. Normally, you would calculate them using long division, but
the
uniqueness Theorem 60.2 tells you that if you can find them some other way
you know
you have the right ones. A mathematician would say that Theorem 60.2
ensures
that the quotient (of integers) m divn and the remainder mrnodn are well-
defined, or
that Definition 60.1 and Theorem 60.2 work together to characterize
the
quotient and remainder.
It is
typical of definitions in abstract mathematics that they characterize a con-
cept
without telling you how to compute it. The technique of separating the two
ideas,
“what is it?” and “how do you compute it?”, is fundamental in mathematics.
86
decimal 12,
93
definition
4
digit 93
div 82
fact 1
floor 86
greatest
integer 86
integer 3
mod 82, 204
quotient
(of inte-
gers) 83
real number
12
rule of
inference 24
trunc 86
usage 2
61. Trunc
and Floor
Many
computer languages have one or both of two operators trunc and floor which
are related
to div and are confusingly similar. Both are applied to real numbers.
61.1
Definition: floor
Floor (r),
or the greatest integer in r, is the largest integer n with
the
property n < r .
61.1.1
Example floor(3.1415) = 3, floor(7/8) =0, and floor(— 4.3) = —5.
61.1.2
Usage Floor(r) is denoted by |_rj in modern texts, or by [r] in older ones.
61.1.3
Exercise State a rule of inference for floor(?’) . (Answer on page 246.)
61.2
Definition: trunc
Trunc (r)
is obtained from r by expressing r in decimal notation and
dropping
all digits after the decimal point.
61.2.1 Fact
The function trunc satisfies the equation
trunc(r)
floor(r) r
> 0 or r an integer
floor(r) +
1 r < 0 and not an integer
61.2.2
Example trunc(— 4.3) = —4, but floor(— 4.3) = —5. On the other hand,
trunc(— 4)
= floor(— 4) = —4, and if r is any positive real number, trunc(r) =
floor(r) .
61.2.3
Exercise Find trunc(x') and floor(x) for
a) x = 7/5.
b) x =
-7/5.
c) x = — 7.
d) x = —
6.7.
(Answer on
page 246.)
61.3
Quotients and remainders for negative integers
61.3.1
Example According to Definition 60.1, — 17div5 = — 4 and — 17mod5 =
3, because
—17 = (—4) -5 + 3 and 0 < 3 < 5. In other words, the quotient is
floor(—
17/5) , but not trunc(— 17/5) .
87
61.3.2
Usage A computer language which has an integer division (typically called
div or
“/” ) which gives this answer for the quotient is said to have floored division.
Mathenratica has floored division.
Other
possibilities include allowing the remainder in Definition 60.1 to be nega-
tive when m
is negative. This results in the quotient being trunc instead of floor,
and, when
implemented in a computer language, is called centered division. That
is how many
implementations of Pascal behave. When n is negative the situation
also allows
several possibilities (depending on whether m is negative or not).
In this
book, integer division means floored division, so that it conforms to
Definition
60.1.
62. Unique
factorization for integers
62.1 The
Fundamental Theorem of Arithmetic
It is a
fact, called The Fundamental Theorem of Arithmetic, that a given
positive
integer m> 1 has a unique factorization into a product of positive primes.
Thus 12 = 2
x 2 x 3, 111 = 3 x 37, and so on. The factorization of a prime is that
prime
itself: thus the prime factorization of 5 is 5. The Fundamental Theorem of
Arithmetic
is proved in a series of problems in Chapter 103 as an illustration of the
proof
techniques discussed there.
The
factorization into primes is unique in the sense that different prime factor-
izations
differ only in the order they are written.
Here is the
formal statement:
62.2
Theorem
Let m be an
integer greater than 1 . Then for some integer n > 1 there
is a unique
list of primes p\,p 2 ,... ,p n and a unique list of integers
h,k
2 ,... ,k n such that
FT.l pi <
pi + i for 1 < i < n.
FT. 2 m =
p\ x p 2 2 ■ ■ -Pn 1 ■
62.2.1
Example
12 = 2x2x3
= 2x3x2 = 3x2x2
Theorem
62.2 specifically gives 12 = 2 2 x 3 1 . Here n = 2, p\ = 2, p 2 = 3, k\ = 2
and k 2 =
l.
62.2.2
Exercise Give the prime factorizations of 30, 35, 36, 37 and 38.
(Answer on
page 246.)
62.3
Definition: exponent of a prime in an integer
The largest
power of a prime p which divides a positive integer n is the
exponent of
p in n and is denoted e p (n) .
62.3.1
Example The exponent of 2 in 24 is 3; in other words, e2(24) = 3. You
can check
that e 37 (lll) = 1 and e 37 ( 110 ) = 0.
centered
division 87
definition
4
divide 4
div 82
exponent 87
floored
division 87
floor 86
Fundamental
Theo-
rem of
Arith-
metic 87
integer 3
negative
integer 3
positive
integer 3
prime 10
quotient
(of inte-
gers) 83
remainder
83
theorem 2
trunc 86
usage 2
88
coordinate
49
definition
4
divide 4
divisor 5
exponent 87
GCD 88
greatest
common
divisor 88
integer 3
least
common multi-
ple 88
nonnegative
integer 3
positive
integer 3
prime 10
theorem 2
62.3.2
Exercise Find the exponent of each of the primes 3, 7 and 37 in the
integers
98, 99, 100, 111, 1332, and 1369. (Answer on page 246.)
The fact
that the prime factorization is unique implies the following theorem:
62.4
Theorem
Let m and n
be positive integers. If m\n and p is a prime, then
e p (m) <e
p (n). Conversely, if for every prime p, e p (m) <e p (n), then
m\n.
62.5 Prime
factorization in Mathematica
Factorlnteger is the Mathematica command for finding the factors of an integer.
The answer
is given as a list of pairs; the first coordinate in each pair is a prime and
the second
coordinate is the exponent of the prime in the number being factored.
Thus if you
type Factorlnteger [360] , the answer will be {{2, 3}, {3, 2 }, {5, 1}},
meaning
that 360 = 2 3 • 3 2 • 5 .
62.5.1
Exercise Factor all the two-digit positive integers that begin with 9.
(Answer on
page 246.)
62.5.2
Exercise Show that for every positive integer k, there is an integer n that
has exactly
k positive divisors.
62.5.3
Exercise (hard) Prove Theorem 62.4.
62.5.4
Exercise (discussion) Type Factorlnteger [6/7] in Mathematica.
Explain the
answer you get. Should the name “Factorlnteger” be changed to
some other
phrase?
63. The GCD
63.1
Definition: greatest common divisor
The
greatest common divisor or GCD of two nonnegative integers
m and n is
0 if m = n = 0 ; otherwise the GCD is the largest number
which
divides both of them.
63.2
Definition: least common multiple
The least
common multiple (LCM) of two nonnegative integers m
and n is 0
if either m or n is 0; otherwise it is the smallest positive
integer
which both m . and n divide.
63.2.1
Example It follows from the definition that GCD(0,0) = 0, GCD(0,4) =
GCD(4,0)
=4, GCD(16,24) = 8, and GCD(5,6) = 1. Similarly, LCM(0,0) = 0,
LCM(1, 1) =
1, LCM(8,12) = 24 and LCM(5,6) =30.
89
63.2.2
Exercise Find GCD(12,12), GCD(12,13), GCD(12,14), GCD(12,24),
and also
find the LCM’s of the same pairs of numbers. (Answer on page 246.)
63.2.3
Exercise Compute GCD(48,72) and LCM(48,72).
63.2.4
Exercise If m and n are positive integers and d = GCD(m,n) , must
GCD (m/d,n)
= 1? Explain your answer. (Answer on page 246.)
63.2.5
Exercise Let A = {1, 2, 3,4} . Write out all the ordered pairs in the relation
a on A
where a is defined by: aab 44 GCD (a, 6) = 1. (Answer on page 246.)
63.2.6
Exercise Let a be the relation on Z defined by aab 44 GCD(a,6) = 1.
Determine
which of these properties a satisfies: Reflexive, symmetric, transitive,
antisymmetric.
63.2.7
Usage Some texts call the GCD the Greatest Common Factor (GCF).
63.2.8
Remark In general, GCD(0,m) = GCD(m,0) = m for any nonnegative
integer m.
Note that Definition 63.1 defined GCD(0,0) as a special case. This
is
necessary because every integer divides 0, so there is no largest integer that
divides 0.
This awkward detail occurs because our definition is in a certain sense
not the
best definition. (See Corollary 64.2 below.)
63.3
Definition: relatively prime
If GCD(m,n)
= 1, then m and n are relatively prime.
63.3.1
Example 5 and 6 are relatively prime, but 74 and 111 are not relatively
prime since
their GCD is 37.
63.3.2
Exercise Show that for any integer n, n and n+1 are relatively prime.
(Answer on
page 246.)
63.3.3
Exercise
a) Show
that if n + 1 distinct integers are chosen from the set {1,2, .. . ,2n} , then
two of them
are relatively prime.
b) Show
that there is a way to choose n integers from {1,2,... ,2n} so that no
two
different ones are relatively prime.
63.3.4
Warning The property “relatively prime” concerns two integers. It makes
no sense to
speak of a single integer as being “relatively prime” .
63.4
Definition: lowest terms
A rational
number m/n is in lowest terms (see Definition 7.3, page 11)
if m and n
are relatively prime.
63.4.1
Exercise Prove that if m/n and r/s are rational numbers represented in
lowest
terms and m/n = r/s, then \m\ = \r\ and |n| = |s| .
definition 4
divide
4
equivalent 40
GCD 88
integer
3
lowest
terms 11
nonnegative integer 3
ordered
pair 49
positive integer 3
relation 73
relatively prime 89
usage 2
90
Cartesian product 52
commutative 71
corollary 1
divide
4
exponent 87
Fundamental Theo-
rem of
Arith-
metic
87
GCD 88
integer
3
lowest
terms 11
nonnegative integer 3
positive integer 3
prime
10
relatively prime 89
theorem
2
64.
Properties of the GCD
If in >
1 and n > 1 , and you know the prime factorizations of both of them, the
GCD and
LCM can be calculated using the following theorem, in which e p (m)
denotes
the exponent of p in m (Definition 62.3), min(r, s) denotes the smaller of r
and s
and max(r, s) the larger.
64.1
Theorem
Let p
be a prime and m and n positive integers. Then
e p
(GCD(m,n)) = min(e p (m),e p (n))
and
e p
(LCM(m,n)) = max(e p (m),e p (n))
64.1.1
Example 60 = 2 2 x 3 x 5 and 72 = 2 3 x 3 2 . Their GCD is 12 = 2 2 x 3, in
which 2
occurs min(2,3) times, 3 occurs min(l,2) times, and 5 occurs min(l,0)
times.
Their LCM is 360 = 2 3 x 3 2 x 5.
64.2
Corollary
Let m
and n he nonnegative integers. GCD(m,n) is the unique non-
negative integer with these properties:
a)
GCD(m,n) divides both m and n.
b) Any
integer e which divides both m and n must divide
GCD
(m,n) .
64.2.1
Remark The property of GCD given in this corollary is often taken as the
definition of GCD. Note that no special consideration has to be given to the case
m = n =
0 .
64.2.2
Exercise Prove Corollary 64.2. (This corollary can be proved without
using
the Fundamental Theorem of Arithmetic. See Exercise 88.3.8, page 130.)
(Answer
on page 246.)
64.2.3
Exercise Use Theorems 62.4 and 64.1 to prove these facts about the GCD
and the
LCM:
a)
GCD(m,n)LCM(m,ri) = mn for any positive integers m and n.
b) If m
and n are relatively prime, then LCM(m,n) =mn.
64.2.4
Exercise Prove that if d = GCD (m,n), then m/d and n/d are relatively
prime.
(Answer on page 246.)
64.2.5
Exercise Prove that every rational number has a representation in lowest
terms.
64.2.6
Exercise Prove that GCD is commutative: for all integers m and n,
GCD(m,
n) = GCD(n, m) .
91
64.2.7
Exercise Prove that GCD is associative:
GCD((GCD (k,m),n) = GCD(fc,GCD(m,n))
Hint:
Use Theorem 64.1 and the fact that the smallest of the numbers x, y and z
is
min(x,min(y,z)) = min(min(a;,y),,z) = va.m(x,y,z)
64.2.8
Exercise (Mathematica)
a) Use
Mathematica to determine which ordered pairs ( a,b ) of integers, with
a € {1,
. . . , 10} , b £ {1, . . . , 10} , have the property that the sequence a + b,2a +
b, . . . , 10a + b contains a prime.
b) Let
(C) be the statement:
associative 70
commutative 71
definition 4
divide
4
function 56
GCD 88
integer
3
ordered
pair 49
predicate 16
prime
10
There
is an integer k > 0 for which ak + b is prime.
(The
integer k does not have to be less than or equal to 10.) Based on
the
results, formulate a predicate P(a,b) such that the condition (C) implies
P(a,b).
The predicate P should not mention k.
c)
Prove that (C) implies P(a,b).
Note:
Define a function by typing t [a_ , b_] : = Table [a k + b , {k , 1 , 10}] (notice
the
spacing and the underlines). Then if you type, for example, t[3,5], you will
get
{8,11,14,17,20,23,26,29, 32,35}. If L is a list, Select [L,PrimeQ] produces
a list
of primes occurring in L.
64.3
Extensions of the definition of GCD
GCD is
often defined for all integers, so that GCD(m,n) is GCD(|m| , |n|) . For
example, GCD(— 6,4) = GCD(6,— 4) = GCD(— 6,— 4) = 2. With this extended
definition, GCD is an associative and commutative binary operation on Z (Sec-
tion
143.2.1). Associativity means it is unambiguous to talk about the GCD of
more
than two integers. In fact, we can define that directly:
64.4
Definition: generalized GCD
Let
ni , ri 2 ,...,nfc be integers. Then GCD(rai, . . . T rik) is the largest
integer
that divides all the numbers |ni| , |ri 2 1 , ■ • ■ ,\rik\-
64.4.1
Example GCD(4,6, -8, 12) = 2.
64.4.2
Remarks
a)
Similar remarks can be made about the LCM.
b)
These functions are implemented in Mathematica using the same names. For
example, GCD [4 , 6 , -8 , 12] returns 2.
92
divide
4
div 82
Euclidean algo-
rithm
92
GCD 88
integer
3
nonnegative integer 3
proof 4
remainder 83
theorem
2
65.
Euclid’s Algorithm
Theorem
64.1 is fine for finding the GCD or LCM of two numbers when you know
their
prime factorization. Unfortunately, the known algorithms for finding the prime
factorization are slow for large numbers. There is another, more efficient method
for
finding the GCD of two numbers which does not require knowledge of the prime
factorization. It is based on this theorem:
65.1 Theorem: Euclid’s Algorithm
For
all nonnegative integers m and n :
EA.l GCD(m,0)=m and GCD(0,n)=n.
EA.2 Let r be the remainder when m is divided by n. Then
GCD(m,n) = GCD(n,r)
Proof Both parts of Theorem 65.1 follow from Definition 6.1, page 10. EA.l
follows because every integer divides 0 (Theorem 5.1(2)), so that if m/ 0, then
largest integer dividing m and 0 is the same as the largest integer dividing m,
which of course is m.
To
prove EA.2, suppose d is an integer that divides both m and n. Since
r =
m — qn . where q = m div n , it follows from Theorem 5.4, page 8, that d divides
r.
Thus d divides both n and r.
Now
suppose e divides both n and r . Since m = qn + r, it follows that e divides
m.
Thus e divides both m and n.
In
the preceding two paragraphs, I have shown that m and n have the same
common divisors as n and r. It follows that m and n have the same greatest
common divisor as n and r, in other words GCD(m,n) = GCD(n,?’).
65.1.1 How to compute the GCD Theorem 65.1 provides a computational
process for determining the GCD . This process is the Euclidean algorithm. The
process always terminates because every time EA.2 is used, the integers involved
are
replaced by smaller ones (because of Definition 60.1(Q.2), page 83) until one of
them becomes 0 and EA.l applies.
65.1.2 Example
GCD(164,48) = GCD(48,20) = GCD(20,8) = GCD(8,4) = GCD(4,0) = 4
65.2 Pascal program for Euclid’s algorithm
A
fragment of a Pascal program implementing the Euclidean algorithm is given
formally in Program 65.1.
93
{M>0 , N>0 , K=M, L=N}
while N <> 0 do
begin
rem : = M mod N ;
M := N;
N := rem;
end;
{M=GCD(K,L)}
decimal 12, 93
integer 3
positive integer 3
specification 2
string 93, 167
Program 65.1: Pascal Program for GCD
66. Bases for representing integers
66.1 Characters and strings
The number of states in the United States of America is an integer. In the usual
notation, that integer is written ‘50’.
In this section, we discuss other, related ways of expressing integers which are
useful in applications to computer science. In doing this it is important to
distinguish
between numerals like ‘5’ and ‘0’ and the integers they represent. In particular,
the
sequence of numerals ‘50’ represents the integer which is the number of states in
the
USA, but it is not the same thing as that integer.
Numerals, as well as letters of the alphabet and punctuation marks, are char-
acters. Characters are a type of data, distinct from integers or other numerical
types. In order to distinguish between a character like ‘5’ and the number 5 we put
characters which we are discussing in single quotes. Pascal has a data type CHAR
of which numerals and letters of the alphabet are subtypes. Single quotes are used
in Pascal as we use them.
66.2 Specification: string
A sequence of characters, such as ‘50’ or ‘cat’, is also a type of data
called a string.
66.2.1 Remarks
a) Strings will be discussed from a theoretical point of view in Chapter 109.
b) In this book we put strings in single quotes when we discuss them. Thus ‘cat’
is a string of characters whereas “cat” is an English word (and a cat is an
animal!).
66.3 Bases
The decimal notation we usually use expresses an integer as a string formed of the
numerals ‘O’, ‘1’, ... ,‘9’. These numerals are the decimal digits. The word
“digit”
is often used for the integers they represent, as well. The notation is based on
the
fact that any positive integer can be expressed as a sum of numbers, each of which
is the value of a digit times a power of ten. Thus
258 = 2 x 10 2 + 5 x 10 1 + 8 x 10° .
94
base 94
decimal 12, 93
definition 4
digit 93
integer 3
least significant
digit 94
more significant 94
most significant
digit 94
nonnegative integer 3
octal notation 94
radix 94
The expression ‘258’ gives you the digits multiplying each power of 10 in decreas-
ing order, the rightmost numeral giving the digit which multiplies 1 = 10° .
Any integer greater than 1 can be used instead of 10 in an analogous way to
express integers. The integer which is used is the base or radix of the notation.
In
octal notation, for example, the base is 8, and the octal digits are ‘O’,
‘1’, ... , ‘7’.
For example,
258 = 4 x 8 2 + 0 x 8 1 + 2 x 8°
so the number represented by ‘258’ in decimal notation is represented in octal
notation by ‘402’ .
Here is the general definition for the representation of an integer in base b.
66.4 Definition: base
If n and b are nonnegative integers and b > 1 , then the expression
dm d m — 1 d m — 2 •••dido’ (66.1)
represents n in base b notation if for each i , d* is a symbol (base- b
digit) representing the integer n . L ,
n = n m b m + n m -ib m ~ l 4 f n 0 b° (66.2)
and for all i ,
0 < rii < b — 1
(66.3)
66.4.1 Remarks
a) We will say more about the symbols d* below. For bases b < 10 these symbols
are normally the usual decimal digits,
do = ‘O’, di = ‘l’,...,d 9 = ‘9’
as illustrated in the preceding discussion.
b) Efficient ways of determining the base- b representation of some integer are
discussed in Chapter 68. Note that you can do the exercises in this section
without knowing how to find the base- 6 representation of an integer — all you
need to know is its definition.
66.4.2 Notation When necessary, we will use the base as a subscript to make
it clear which base is being used. Thus 258io = 402s , meaning that the number
represented by ‘258’ in base 10 is represented by ‘402’ in base 8.
66.5 Definition: significance
The digit d,; is more significant than dj if i> j. Thus, if a number
n is represented by ‘d m d m _i . . .dido’, then do is the least significant
digit and, if d m does not denote 0, it is the most significant digit.
66.5.1 Example The least significant digit in 258io is 8 and the most significant
is 2.
95
66.5.2 Remark For a given b and n, the following theorem says that the rep-
resentation given by definition 66.4 is unique, except for the choice of the
symbols
representing the rii . We will take this theorem as known.
66.6 Theorem
If n and b are positive integers with b > 1 , then there is only one
sequence no,ni,... ,n m of integers for which n m / 0 and formulas (66.2)
and (66.3) are true.
66.6.1 Worked Exercise Prove that the base 4 representation of 365 is 11231.
Answer 365 = 1 • 4 4 + 1 • 4 3 + 2 • 4 2 + 3 • 4 1 + 1 • 4° , and 1,2,3 are all
less than 4 , so
the result follows from Theorem 66.6.
Note that in this answer we merely showed that 11231 fit the definition. That
is all that is necessary. Of course, if you are not given the digits as you were in
this
problem, you need some way of calculating them. We will describe ways of doing
that in Chapter 68.
66.6.2 Exercise Prove that the base 8 representation of 365 is 555.
66.6.3 Exercise Prove that if an integer n is represented by • • • d\ in
base b, then ‘c? TO d m _i • • • g?iO’ represent bn in base b notation. (Answer on
page
246.)
66.6.4 Exercise Suppose b is an integer greater than 1 and suppose n is an
integer such that the base b representation of n is 352. Prove using only the
definition of representation to base b that the base b representation of b 2 n + 1
is
35201.
66.7 Specific bases
66.7.1 Base 2 The digits for base 2 are ‘O’ and ‘1’ and are called bits. Base 2
notation is called binary notation.
66.7.2 Bases larger than 10 For bases b< 10, the usual numerals are used,
as mentioned before. A problem arises for bases bigger than 10: you need single
symbols for the integers 10, 11, .... Standard practice is to use the letters of
the
alphabet (lowercase here, uppercase in many texts): ‘a’ denotes 10, ‘b’ denotes 11,
and so on. This allows bases up through 36.
66.7.3 Base 16 Base 16 (giving hexadecimal notation) is very commonly used
in computing. For example, 95io is 5fi@, and 266io is hexadecimal lOaig (read
this “one zero a”, not “ten a”!) In texts in which decimal and nondecimal bases are
mixed, the numbers expressed nondecimally are often preceded or followed by some
symbol; for example, many authors write $10a or HlOa to indicate 266io expressed
hexadecimally.
alphabet 93, 167
base 94
binary notation 95
decimal 12, 93
digit 93
hexadecimal nota-
tion 95
hexadecimal 95
integer 3
positive integer 3
theorem 2
96
base 94
decimal 12, 93
digit 93
integer 3
least significant
digit 94
nonnegative integer 3
positive integer 3
prime 10
realizations 96
66.8 About representations
(This continues the discussion of representations in Section 10.2 and Remark
17.1.3.)
It is important to distinguish between the (abstract) integer and any
representation
of it. The number of states in the U.S.A is represented as ‘50’ in decimal
notation,
as ‘110010’ in binary, and as a pattern of electrical charges in in a computer.
These
are all representations or realizations of the abstract integer. (The word
“realiza-
tion” here has a technical meaning, roughly made real or made concrete.) All the
representations are matters of convention, in other words, are based on agreement
rather than intrinsic properties. Moreover, no one representation is more fundamen-
tal or correct than another, although one may be more familiar or more convenient
than another.
There is also a distinction to be made between properties of an integer and
properties of the representation of an integer. For example, being prime is a
property
of the integer; whether it is written in decimal or binary is irrelevant. Whether
its
least significant digit is 0, on the other hand, is a property of the
representation:
the number of states in the USA written in base 10 ends in ‘O’, but in base 3 it
ends
in ‘2’.
66.8.1 Exercise Suppose b is an integer greater than 1, a is an integer dividing
b, and n is an integer. When n is written in base b, how do you tell from the
digits
of n whether n is divisible by a? Prove that your answer is correct.
66.8.2 Exercise Would Theorem 66.6 still be true if the requirement that 0 <
rii < b — 1 for all i were replaced by the requirement that the n t be nonnegative?
66.8.3 Exercise (Mathematica) A positive integer is a repunit if all its decimal
digits are 1 .
a) Use Mathematica to determine which of the repunits up to a billion are divis-
ible by 3 .
b) Based on the results of part (a), formulate a conjecture as to which repunits
are divisible by 3 . The conjecture should apply to all repunits, not just those
less than a billion.
c) Prove the conjecture.
66.8.4 Exercise (discussion) Some computer languages (FORTH is an example)
have a built-in integer variable BASE. Whatever integer you set BASE to will be
used
as the base for all numbers output. How would you discover the current value of
BASE in such a language? (Assume you print the value of a variable X by writing
PRINT (X)).
97
67. Algorithms and bases
Among the first algorithms of any complexity that most people learn as children are
the algorithms for adding, subtracting, multiplying and dividing integers written
in
decimal notation. In medieval times, the word “algorithm” referred specifically to
these processes.
67.1 Addition
The usual algorithm for addition you learned in grade school works for numbers in
other bases than 10 as well. The only difference is that you have to use a
different
addition table for the digits.
67.1.1 Example To add 95a and b87 in hexadecimal you write them one above
the other:
95a
+b87
14el
Here is a detailed description of how this is done, all in base 16.
• Calculate a+7 = 11m, with a carry of 1 since 11m > 10m- (Pronounce 10m
as “one-zero”, not “ten”, since it denotes sixteen, and similarly for 11m which
denotes seventeen. By the way, the easiest way to figure out what a + 7 is is to
count on your fingers!)
• Then add 5 and 8 and get d (not 13!) and the carry makes e. e < 10m so there
is no carry.
• Finally, 9 + b = 14m-
So the answer is 14e lm • The whole process is carried out in hexadecimal without
any conversion to decimal notation.
67.1.2 Addition in binary The addition table for binary notation is especially
simple: 0 + 0 = 0 without carry, l + 0 = 0+ l = l without carry, and 1 + 1 = 0 with
carry.
67.2 Multiplication
The multiplication algorithm similarly carries over to other bases. Normally in a
multiplication like
346 (multiplicand)
X527 (multiplier)
2422
6920 (partial products)
173000
182342 (product)
you produce successive partial products, and then you add them. The partial prod-
uct resulting from multiplying by the ith digit of the multiplier is
base 94
decimal 12, 93
digit 93
hexadecimal nota-
tion 95
integer 3
digit x multiplicand x 10*
98
base 94
digit 93
hexadecimal nota
tion 95
67.2.2 Example With trailing zeroes suppressed:
1101
X1101
1101
1101
1101
10101001
67.2.3 Exercise Perform these additions and multiplications in binary.
a) 110001 b) 1011101 c) 10011 d) 11100
+ 101111 +1110101 XIOlOl X11001
(Answer on page 246.)
67.2.4 Exercise Perform these additions in hexadecimal:
a) 9ae b) 389 c) feed
+b77 +777 +dad
(Answer on page 246.)
67.2.5 Exercise Show that in adding two numbers in base b, the carry is never
more than 1 , and in multiplying in base b , the carry is never more than b — 2.
67.2.6 Exercise (discussion) Because subtracting two numbers using pencil and
paper is essentially a solitary endeavor, most people are not aware that there are
two different algorithms taught in different public school systems. Most American
states’ school systems teach one algorithm (Georgia used to be an exception), and
many European countries teach another one. Ask friends from different parts of the
world to subtract 365 from 723 while you watch, explaining each step, and see if
you detect anyone doing it differently from the way you do it.
(Most people are taught in grade school to suppress the zeroes to the right of the
multiplying digit.)
67.2.1 Binary multiplication Multiplication in binary has a drastic simplifica-
tion. In binary notation, the only digits are 0, which causes a missing line, and
1,
which involves only shifting the top number. So multiplying one number by another
in binary consists merely of shifting the first number once for each 1 in the
second
number and adding.
99
68. Computing integers to different bases
68.1 Representing an integer
68.1.1 Remark Given a nonnegative integer n and a base b, the most significant
nonzero digit of n when it is represented in base b is the quotient when n is
divided
by the largest power of b less than n. For example, in base 10, the most
significant
digit of 568 is 5, and indeed 5 = 568 div 100 (100 is the largest power of 10 less
than
568). Furthermore, 68 is the remainder when 568 is divided by 500.
This observation provides a way of computing the base- b representation of an
integer.
base 94
digit 93
div 82
integer 3
mod 82, 204
most significant
digit 94
nonnegative integer 3
quotient (of inte-
gers) 83
remainder 83
68.1.2 Method
Suppose the representation for n to base b is l d m d m - \ ■
represents the integer n t in base b. Then
■ ■ do , where di
d m = n div b m
and
d m - 1 = (n - d m b m ) div b m ~ l
In general, for all i = 0, 1, . . . ,m — 1 ,
di = (n i+ 1 - d i+1 b l+1 ) div b l
(68.1)
where
n m = n
(68.2)
and for i = 0,l,...,m — 1,
rii = n i+ 1 - di + ±b l+1
(68.3)
68.1.3 Example The ‘6’ in 568 is
(568 - 5 • 100) div 10
(here m = 2: note that the 5 in 568 is since we start counting on the right at 0).
68.1.4 Remark Observe that (68.1) can be written
di = (n mod b l+l ) div b l (68.4)
which is correct for all i = 0,1,... ,m. The way (68.1) is written shows that the
computation of nmodfr*" 1 " 1 uses the previously-calculated digit d l+ \ .
68.1.5 Example We illustrate this process by determining the representation of
775 to base 8. Note that 512 = 8 3 :
a) 775 div 512 = 1.
b) 775-1 x 512 = 263.
c) 263div64 = 4.
d) 263-4 x 64 = 7.
100
base 94
digit 93
div 82
integer 3
mod 82, 204
octal notation 94
string 93, 167
e) 7 div 8 = 0.
f) 7 — 0 x 8 = 7.
g) 7 div 1 = 7.
And 775 in octal is indeed 1407.
68.2 The algorithm in Pascal
The algorithm just described is expressed in Pascal in Program 68.1. This algorithm
is perhaps the most efficient for pencil-and-paper computation. As given, it only
works as written for bases up to and including 10; to have it print out ‘a’ for 11,
‘b’
for 12 and so on would require modifying the “write (place)” statement.
var N, base, count, power, limit, place: integer;
(* Requires B > 0 and base > 1 *)
begin
power := 1; limit := N div base;
(♦calculate the highest power of the base less than N*)
while power <= limit do
begin
power := power*base
end;
while power > 1 do
begin
place := N div power; write (place) ;
n := n-place*power ; power := power div base
end
end
Program 68.1: Program for Base Conversion
68.3 Another base conversion algorithm
Another algorithm, which computes the digits backwards, stores them in an array,
and then prints them out in the correct order, is given in Program 68.2. It is more
efficient because it is unnecessary to calculate the highest power of the base less
than N first. This program starts with the observation that the least significant
digit in a number n expressed in base b notation is nrnodfo. The other digits in
the representation of n represent (n — (nrnod b))/b. For example, 568modl0 = 8,
and the number represented by the other digits, 56, is (568 — 8)/10.
In the program in Program 68.2, count and u are auxiliary variables of type
integer. The size longest of the array D has to be known in advance, so there
is a bound on the size of integer this program can compute, in contrast to the
previous algorithm. It is instructive to carry out the operations of the program in
Program 68.2 by hand to see how it works.
68.4 Comments on the notation for integers
Suppose n is written . . . do’ i n base b. Then the exact significance of
d m , namely the power b m that its value n m is multiplied by in Equation (66.2)
of
Definition 66.4 (page 94), depends on the length of the string of digits
representing
101
var count, u, N, base: integer;
var D: array [0.. longest] of integer;
begin
count : = 0 ; u : = N ;
while u<>0 do
begin
D [count] := u mod base;
u := (u-D [count]) div base;
count := count+1
end;
while countoO do
begin
count := count- 1;
write D [count]
end
end;
base 94
digit 93
hexadecimal nota-
tion 95
integer 3
octal notation 94
Program 68.2: Faster Program for Base Conversion
n (the length is m + 1 because the count starts at 0). If you read the digits from
left
to right, as is usual in English, you have to read to the end before you know what
m is. On the other hand, the significance of the right digit do is known without
knowing the length m. In particular, the program in Program 68.1 has to read to
the end of the representation to know the power b m to start with.
The fact that the significance of a digit is determined by its distance from the
right is the reason a column of integers you want to add is always lined up with
the
right side straight. In contrast to this, the sentences on a typewritten page are
lined
up with the left margin straight.
There is a good reason for this state of affairs: this notation was invented by
Arab mathematicians, and Arabic is written from right to left.
68.4.1 Exercise Represent the numbers 100, 111, 127 and 128 in binary, octal,
hexadecimal and base 36. (Answer on page 246.)
68.4.2 Exercise Represent the numbers 3501, 29398 and 602346 in hexadecimal
and base 36.
68.5 Exercise set
Exercises 68.5.1 through 68.5.4 are designed to give a proof of Formula (68.4),
page 99, so they should be carried out without using facts about how numbers
are represented in base b. In these exercises, all the variables are of type
integer.
68.5.1 Exercise Let b > 1 . Prove that if for all i > 0, 0 < d{ < b, then
d m b m + d m -ib m ~ l + --- + d 1 b + d 0 < b m+1
68.5.2 Exercise Let 6>1 and n>0. Let n = d m b m + \-d\b + do with 0<
di <b for i = 0, 1 , . . . , m . Prove that for any i > 0 ,
n — b'\d m b" 1 ' + d m —ib" 1 ' 1 + • • • + di] + di—ib‘ 1 + • • • + d\b + do
102
conjunction 21
defining condition 27
definition 4
DeMorgan Law 102
div 82
equivalent 40
mod 82, 204
proposition 15
rule of inference 24
unit interval 29
and 0 < dj-ib 1 1 H b d\b + do < b l . (Hint: Use Exercise 68.5.1.)
68.5.3 Exercise Let b > 1 and n > 0 and let n = d m b m -\ \-d\b + do with 0 <
di < b for * = 0,1,... , m . Prove that for any i > 0 ,
d m b m ~ l + d m -ib m ~ l ~ l -\ b di = n div b l
and
dib 1 H b d\b + do = nrnod b l+1
68.5.4 Exercise Prove Equation (68.4), page 99.
69. The DeMorgan Laws
Consider what happens when you negate a conjunction. The statement — i(P A Q )
means that it is false that P and Q are both true; thus one of them must be false.
In other words, either —>P is true or — i Q is true. This is one of the two
DeMorgan
Laws:
69.1 Definition: DeMorgan Laws
The DeMorgan Laws are:
DM.1 ~^{PAQ)
DM.2 — i(P V Q) 44 — '-P A — 'Q .
These laws are true no matter what propositions P and Q are.
69.1.1 Remark The DeMorgan Laws give rules of inference
—i(P AQ) \— -iP V-iQ and —>P\/—^Q \— — >(P AQ) (69.1)
and
~ i(PVQ) |— — <P A~ <Q and — <P A— <Q |— —> (P\/Q ) (69.2)
69.1.2 Example The negation of (x + y = 10) A (x < 7) is (x + y / 10) V-i(x < 7) .
Of course, ~^{x < 7) is the same as x > 7.
69.2 Using the DeMorgan Laws in proofs
The unit interval I = {x | 0 < x < 1} , which means that x G I if and only if 0 < x
and x < 1 . Therefore to prove that some number a is not in I , you must prove the
negation of the defining condition, namely that it is not true that 0 < x and x < 1
.
By the DeMorgan Laws, this means you must prove
-■(0 < x) V -i(x < 1)
which is the same as proving that (0 > x) V {x > 1) .
103
69.2.1 Warning When proving that a conjunction is false, it is easy to forget the
DeMorgan Laws and try to prove that both negatives are true. In the preceding
example, this would require showing that both 0 > x and x > 1 , which is obviously
impossible.
In contrast, if you must prove that a disjunction P\/ Q is false, you must show
that both P and Q are false. An error here is even more insidious, because if you
are tempted to prove that only one of P and Q is false, you often can do that
without noticing that you have not done everything required.
69.2.2 Example Consider the statement, “A positive integer is either even or it
is prime”. This statement is false. To show it is false, you must find a positive
integer such as 9 which is both odd and nonprime.
and 21, 22
conjunction 21
DeMorgan Law 102
even 5
integer 3
odd 5
positive integer 3
predicate 16
prime 10
real number 12
union 47
69.2.3 Method
To prove that P V Q is false, prove that ->P A~*Q is true. To prove that
P A Q is false, prove that ~*P V -i Q is true.
69.2.4 Example Given two sets A and B , how does one show that 4/ B? By
Method 21.2.1 on page 32, A = B means that every element of A is an element of
B and every element of B is an element of A. By DeMorgan, to prove A^ B you
must show that one of those two statements is false: you must show either that
there is an element of A that is not an element of B or that there is an element of
B that is not an element of A. You needn’t show both, and indeed you often can’t
show both. For example, {1,2} ^ {1,2,3}, yet every element of the first one is an
element of the second one.
69.2.5 Worked Exercise Let A and B be sets. How do you prove x <£. AU B?
How do you prove x ^ An B?
Answer To prove that x (ji Au B , you must prove both that x (j A and that
x B . This follows from the DeMorgan Law and the definition of union. To prove
x (f An B , you need only show x ^ A or x ^ B .
69.3 Exercise set
Reword the predicates in Exercises 69.3.1 through 69.3.3 so that they do not begin
with “-i”. x is real.
69.3.1 -i(x < 10) A(x> 12) . (Answer on page 246.)
69.3.2 -i(x < 10) A (x < 12) . (Answer on page 247.)
69.3.3 — 1 ( — i(a: > 5) A -<(x < 6)) .
104
DeMorgan Law 102
logical connective 21
predicate 16
propositional
form 104
propositional vari-
able 104
proposition 15
70. Propositional forms
The letters P and Q in the DeMorgan Laws are called propositional variables.
They are like variables in algebra except that you substitute propositions or pred-
icates for them instead of numbers. Don’t confuse propositional variables with the
variables which occur in predicates such as “x < y n . The variables in predicates
are
of the type of whatever you are talking about, presumably numbers in the case of
“x < y ” . Propositional variables are of type “proposition”: they vary over
proposi-
tions in the same way that x and y in the statement “x < y” vary over numbers.
70.1.1 Worked Exercise Write the result of substituting x = 7 for P and
for Q in the expression —>P V (P A Q) .
Answer x / 7 V (x = 7 Ax / 5) .
70.2 Variables in Pascal
Pascal does not have variables or expressions of type proposition. It does have
Boolean variables, which have TRUE and FALSE as their only possible values.
An expression such as ‘A < Y' has numerical variables, and a Boolean value
TRUE or FALSE, so it might correctly be described as a proposition (assuming
the program has already given values to X and Y). However, if B is a Boolean
variable, an assignment statement of the form B: = X <Y sets B equal to the truth
value of the statement ‘A < Y ’ at that point on the program; B is not set equal to
the proposition l X <Y\ If A and Y are later changed, changing the truth value
of ‘A <Y\ the value of B will not automatically be changed.
70.2.1 Example The following program prints TRUE. Here B is type BOOLEAN
and X is of type INTEGER:
X := 3;
B := X < 5;
X : = 7;
PRINT (B) ;
70.3 Propositional forms
Meaningful expressions made up of propositional variables and logical connectives
are called propositional forms or propositional expressions. The expressions
in DM.1 and DM.2 are examples of propositional forms. Two simpler ones are
P V —>P (70.1)
and
-i P V Q
(70.2)
70.3.1 Substituting in propositional forms If you substitute propositions for
each of the variables in a propositional form you get a proposition.
You may also substitute predicates for the propositional variables in a proposi-
tional form and the result will be a predicate.
105
70.3.2 Example If you substitute the proposition “3 < 5” in formula (70.1) you
get (after a little rewording) “3 < 5 or 3 > 5” which is a proposition (a true one,
in
fact).
If you substitute x < 5 for P in formula (70.1) you get “x < 5 or x > 5”, which
is true for any real number x. This is not surprising because formula (70.1) is a
tautology (discussed later).
If you substitute x < 5 for P and x / 6 for Q in ->P V Q you get “x > 5 or
x^6”, which is true for some x and false for others.
70.3.3 Remarks
a) This would be a good time to reread Section 12.1.4. Propositional forms are
a third type of expression beside algebraic expressions and predicates. In an
algebraic expression the variables are some type of number and the output
when you substitute the correct type of data for the variables is a number.
In a predicate the output is a proposition: a statement that is either true or
false. And now in propositional forms the variables are propositions and when
you substitute a proposition for each propositional variable the output is a
proposition.
b) We have not given a formal definition of “meaningful expression”. This is
done in texts on formal logic using definitions which essentially constitute a
context-free grammar.
algebraic expres-
sion 16
definition 4
DeMorgan Law 102
equivalent 40
expression 16
fact 1
predicate 16
propositional
form 104
propositional vari-
able 104
proposition 15
tautology 105
71. Tautologies
71.1 Discussion
Each DeMorgan Law is the assertion that a certain propositional form is true no
matter what propositions are plugged in for the variables. For example, the first
DeMorgan Law is
—'(P A Q) 44- —iP V —'Q
No matter which predicates we let P and Q be in this statement, the result is a
true statement.
71.1.1 Example let P be the statement x < 5 and Q be x = 42. Then the first
DeMorgan Law implies that
1 ((x < 5) A (x = 42)) 44 ((x > 5) V (x / 42))
is a true statement.
71.2 Definition: tautology
A propositional form which is true for all possible substitutions of propo-
sitional variables is called a tautology.
71.2.1 Fact The truth table for a tautology S has all T’s in the column under S .
106
equivalence 40
equivalent 40
implication 35, 36
law of the excluded
middle 106
predicate 16
propositional vari-
able 104
proposition 15
real number 12
tautology 105
truth table 22
71.2.2 Example Both DeMorgan laws are tautologies, and so is the formula (70.1),
which is called The law of the excluded middle. Both lines of its truth table
have T.
P -.P PV-.P
~T F T
FT T
71.2.3 Warning Don’t confuse tautologies with predicates all of whose instances
are true. A tautology is an expression containing propositional variables which
is true no matter which propositions are substituted for the variables. Expres-
sion (70.2) is not a tautology, but some instances of it, for example “not x > 5
or x > 3” are predicates which are true for all values (of the correct type) of the
variables.
71.2.4 Example Formula (70.2) (page 104) is not a tautology. For example,
let P be “4 > 3” and Q be “4 > 5”, where x ranges over real numbers; then
Formula (70.2) becomes the proposition “(not 4 > 3) or 4 > 5”, i.e. , “4 < 3 or 4 >
5”, which is false.
71.2.5 Exercise Show that FV Q 44 -i(->P /\—>Q) is a tautology. (Answer on
page 247.)
71.2.6 Exercise Show that the following are tautologies.
a) P A Q 44 — 1 ( — 'P V — 'Q)
b) (P A —iP) => Q
c) P =4- (Q V -i Q)
d) Pv(P =s> Q)
e) ((PAQ) => P) 44 [p =► (Q => R)j
f) P A (Q V R) => PV(QAP)
71.2.7 Remark Many laws of logic are equivalences like the DeMorgan laws. By
Theorem 29.2, an equivalence between two expressions is a tautology if the truth
tables for the two expressions are identical. Thus the truth tables for -i(PaQ) and
— i P V -i Q are identical:
P Q PAQ
T T T
T F F
FT F
F F F
-i(PAQ) -i P -i Q
F F I 7
T FT
T T F
T T T
—>P V —i Q
71.2.8 Example You can check using this method that -> P\/Q (i.e., For-
mula (70.2)) is equivalent to P =4- Q.
71.2.9 Exercise Prove by using Theorem 29.2 that the propositional forms P =>
Q, -'P V Q and -i(PA-iQ) are all equivalent. (Answer on page 247.)
71.2.10 Exercise Prove that (P =>• Q) Q is equivalent to PVQ.
107
72. Contradictions
72.1 Definition: contradiction
A propositional form is a contradiction if it is false for all possible
substitutions of propositional variables.
72.1.1 Fact The truth table for a contradiction has all F’s.
72.1.2 Example The most elementary example of a contradiction is “PA-iP”.
72.1.3 Exercise Show that the following are contradictions.
a) ~<(P V —<P) ■
b) n(PV(P =}► Q)).
c) Q A -i (P => Q ) .
72.1.4 Exercise If possible, give an example of a propositional form involving
“ =>• ” that is neither a tautology nor a contradiction.
associative 70
commutative 71
complement 48
contradiction 107
definition 4
fact 1
idempotent 143
implication 35, 36
intersection 47
predicate 16
propositional calcu-
lus 107
propositional vari-
able 104
proposition 15
transitive 80, 227
truth table 22
universal set 48
73. Lists of tautologies
Tables 72.1 and 72.2 give lists of tautologies. Table 72.1 is a list of tautologies
involving “and” , “or” and “not” . Because union, intersection and complementation
for sets are defined in terms of “and”, “or” and “not”, the tautologies correspond
to universally true statements about sets, which are given alongside the
tautologies.
Table 72.2 is a list of tautologies involving implication. Because of the modus
ponens rule, the major role implication plays in logic is to provide successive
steps
in proofs. These laws can be proved using truth tables or be deriving them from
the laws in Table 72.1 and the first law in Table 72.2, which allows you to define
‘ =>■ ’ in terms of ‘ — > ’ and ‘V’. It is an excellent exercise to try to
understand why
the tautologies in both lists are true, either directly or by using truth tables.
73.1 The propositional calculus
The laws in Tables 72.1 and 72.2 allow a sort of computation with propositions in
the
way that the rules of ordinary algebra allow computation with numbers, such as the
distributive law for multiplication over addition which says that 3(x + 5) = 3x +
15.
This system of computation is called the propositional calculus, a phrase which
uses the word “calculus” in its older meaning “computational system”. (What is
called “calculus” in school used to be taught in two parts called the “differential
calculus” and the “integral calculus”.)
Recall that every predicate becomes a proposition (called an “instance” of the
predicate) when constants are substituted for all its variables. Thus when
predicates
are substituted for the propositional variables in these laws, they become
predicates
which are true in every instance.
108
equivalent 40
(consistency)
J J
t t
u c = $
Q) C = U
(unity)
P AT AA P
PVFO P
Anu = A
AuQ) = A
(nullity)
PAFaa f
P\J T aaT
II H
si 15
C D
(idempotence)
PAPaaP
P\j P aaP
II II
C D
^1 ^1
(commutativity)
P AQ AA Q AP
PV QaaQm P
AnB=BnA
AUB = BUA
(associativity)
pa(Qar)
aa(paQ)ar
PV(QVP)
4A (PVQ) VP
An(BrC)
= (AnB)nC
AU{BUC)
= {AUB)UC
(distributivity)
P A (Q V P)
4A(PAQ)V(PAP)
pv(Qap)
4A (P V Q) A (P V P)
An{BuC)
= {AnB)u{AnC)
Au{BnC)
= ( AuB)n{Auc )
(complement)
E-i tin
$ $
CL, eg
r r
> <
eg eg
it* ^
D C
it* it*
o o
II II
s s;
(double negation)
^P4VP
(A C ) C = A
(absorption)
PA(PVQ)4AP
PV(PAQ)4AP
it* it*
C D
D C
II II
it* it*
(DeMorgan)
-i(P V Q) 4A -iP A -iQ
— '(P A Q) 4A — 'P V — '<5
(AUB) C = (A c )n(B c )
{AnB) c ={A c UB c )
Table 72.1: Boolean Laws
109
(‘ => ’-elimination)
(transitivity)
(modus ponens)
(modus tollens)
(inclusion)
(simplification)
(cases)
(everything implies true)
(false implies everything)
(P =4- Q) 44 (- 1 PV Q)
(( P =4 Q) A (Q =4 R)) =4 (P =4 P)
(PA(P =4 Q)) =4 Q
(~>QA(P =4- Q)) =4- -iP
P => (PvQ)
(PAQ) => P
(n?A(PVQ)) =4 Q
Q =>• (P =>• Q)
iP => (P => Q)
Table 72.2: Laws of Implication
equivalent 40
implication 35, 36
logical connective 21
truth table 22
73.1.1 Example When you substitute x > 7 for P and x = 5 for Q i n the second
absorption law PV(PAQ)<t>P you get, in words, “Either r >7 or both x > 7
and x = 5” is the same thing as saying “x > 7”. This statement is certainly true:
it is true by its form, not because of anything to do with the individual
statements
“x > 7” and “x = 5” .
73.1.2 Exercise Define the logical connective NAND by requiring that PNAND
Q be true provided at least one of P and Q is false.
a) Give the truth table for NAND .
b) Write a statement equivalent to “PNAND Q” using only ‘A’, ‘V’, ‘-T, ‘P’,
‘ Q ’ and parentheses.
c) Give statements equivalent to P”, “PAQ” and “PVQ” using only ‘P’,
‘ Q ‘NAND’, parentheses and spaces.
73.1.3 Exercise Do the same as Problem 73.1.2 for the connective NOR, where
PNORQ is true only if both P and Q are false.
73.1.4 Exercise Show how to define implication in terms of each of the connec-
tives NAND and NOR of exercises 73.1.2 and 73.1.3.
73.1.5 Exercise Let denote the operation XOR discussed in Chapter 11.
Prove the following laws:
a) P * Q Q * P .
b) P*(Q*P)4A(P*Q)*P.
c) PA(Q*P)4A(PAQ)*(PAP).
73.1.6 Exercise (Mathematica)
a) Show that there are 16 possible truth tables for a Boolean expression with
two variables.
110
distributive law 110
equivalent 40
implication 35, 36
logical connective 21
modus ponens 40
proof 4
propositional
form 104
proposition 15
rule of inference 24
tautology 105
theorem 2
truth table 22
b) Produce Boolean expressions with and “ =>■ ” as the only logical connec-
tives that give each of the possible truth tables. Both variables must appear
in each expression. Include a printout of Mathematica commands that verify
that each expression gives the table claimed.
(Enter p =>• q as p ~ Implies' q.)
73.1.7 Exercise (hard) A distributive law involving binary operations ‘A’
and ‘ V ’ is a tautology of the form
PV{QAR) 4A (PVQ)A(PVP)
Let be defined as in Problem 73.1.5. Give examples showing that of the four
possible distributive laws combining with ‘A’ or ‘V’, the only correct one is that
in Problem 73.1.5(c).
74. The tautology theorem
In Section 28, we discussed the rule of inference called “modus ponens”:
P, P => Q b Q
This rule is closely related to the tautology also called modus ponens in section
71:
(pa(P =► Q)') => Q
This tautology is a propositional form which is true for any proposition P and Q .
This is a special case of the general fact that, roughly speaking, any implication
involving propositional forms which is a tautology is equivalent to a rule of
inference:
74.1 Theorem: The Tautology Theorem
Suppose that F\ , . . . , F n and G are propositional forms. Then
F u ...,F n b G
is a valid rule of inference if and only if
(F\ A ... A F n ) =f- G
is a tautology.
(74.1)
(74.2)
Proof If the rule of inference (74.1) is correct, then whenever all the
propositions
Fi, . . . , F n are true, G must be true, too. Then if F\ A ■ ■ ■ A F n is true,
then every
one of F{,. . . ,F n is true, so G must be true. This means that (74.2) must be a
tautology, for the only way it could be false is if Fi A ■ ■ ■ A F n is true and G
is false.
(This is because any implication P =^> Q is equivalent to -i(PA-iQ).)
On the other hand, if (74.2) is a tautology, then whenever T\, . . . , F n are all
true,
then F\ A • • ■ A F n is true, so that G has to be true, too. That means that
(74.1) is
a valid rule of inference.
Ill
74.1.1 Example The preceding theorem applies to modus ponens: Take F\ to be
equivalent 40
the formula P, F -2 to be “P =4- Q ” , and G to be Q. Since ( PA(P =4- Q) ] =4- Q |
m P^ ca ^ on ^5,
V v / logical connective 21
is a tautology, the validity of the rule of inference called modus ponens follows
by modus ponens 40
the Tautology Theorem from the tautology called modus ponens. propositional
form 104
74.1.2 Remark Not all rules of inference come from tautologies - only those rule of
inference 24
involving propositional forms. We have already seen examples of rules of inference
Tautology Theo-
not involving propositional forms in 18.1.11, page 29. rem ^
tautology 105
74.1.3 Warning The Tautology Theorem does not say that “|— ” is the same
thing as “ . “ [— ” is not a logical connective and cannot be used in formulas the
way “ =4- ” can be. For example you may write P A (P =4- Q) but not P A (P |- Q) .
“ |— ” may be used only in rules of inference.
74.2 Exercise set
For problems 74.2.1 to 74.2.6, state whether the given rule is a valid rule of
inference.
74.2.1 -i P, PVQ |- Q (Answer on page 247.)
74.2.2 -iQ, P =4- (QAR) \- -i P (Answer on page 247.)
74.2.3 -i P, (PAQ) =4- R \- -i R (Answer on page 247.)
74.2.4 nPAQ, Q |- -.P
74.2.5 {PvQ) ^ R, P ^ R
74.2.6 (P A Q) =4- R, ~ i R |— — 'PA - <Q
74.2.7 Exercise Show that the statement (P = 4 - Q) = 4 > Q is not a tautology by
giving an example of statements P and Q for which it is false. (Answer on page
247.)
74.2.8 Exercise Show that the following statements are not tautologies by giving
examples of statements P and Q for which they are false.
a) (P 44 Q) => P
b) f(P => Q) => r) 44 (p => {Q =► R))
74.2.9 Exercise Use the Tautology Theorem to prove that the following rules of
inference are valid:
a) Q \- P =4 Q
b) P, Q b PAQ
c) PAQ h P
d) ^P (- P =4 Q
e) nQ,P4()h -P
112
counterexample 112
definition 4
implication 35, 36
real number 12
universal quanti-
fier 112
75. Quantifiers
75.1 Definition: universal quantifier
Let Q(x) be a predicate. The statement (Mx)Q(x) is true if and only if
Q(x) is true for every value of the variable x. The symbol V is called
the universal quantifier.
75.1.1 Example Let P(x) be the statement (x > 5) =>• (x > 3) . P(x) is uni-
versally true, that is, it is true for every real number x. Therefore, the
expression
(Vx)-P(x) is true.
We defined V in 13.2; now we will go into more detail.
75.1.2 Showing the types of the variables A short way of saying that x is
of type real and that (Mx)Q(x) is to write (Vx:R)Q(a:) , read “for all x of type R,
Q(x)" or “for all real numbers x. Q(x)” .
75.1.3 Example The statement (Vn:Z)((n > 5) is false because “n > 5” is false
for n = 3 (and for an infinite number of other values of n).
75.1.4 Example The statement (Vn:Z)((n > 5) V (n < 5)) is false because the
statement “(n > 5) V (n < 5)” is false when n = 5. Note that in contrast to Exam-
ple 75.1.3, n = 5 is the only value for which the statement “(n > 5) V (n < 5)” is
false.
A statement like (\/x)Q(x) is true if Q(x) is true no matter what is substituted
for
x (so long as it is of the correct type). If there is even one x for which Q(x) is
false, then (\/x)Q(x) is false. A value of x with this property is important
enought
to have a name:
75.2 Definition: counterexample
Let Q(x) denote a predicate. An instance of x for which Q(x) is false
is called a counterexample to the statement (Vx)Q( x) . If there is a
counterexample to the statement (Vx)Q(x) , then that statement is false.
75.2.1 Example (Vx:N)((x < 5) V (x > 6)) is true, but (V.x:R)((a: < 5) V (x > 6))
is false (counterexample: ).
75.2.2 Example A counterexample to the statement (Vn:Z)((n > 5) is 3; in fact
there are an infinite number of counterexamples to this statement. In contrast, the
statement (Vn:Z)((n > 5) V (n < 5)) has exactly one counterexample.
75.2.3 Exercise Find a universal statement about integers that has exactly 42
counterexamples .
75.2.4 Exercise Find a universal statement about real numbers that has exactly
42 counterexamples.
113
75.3 Definition: existential quantifier
Let Q(x) be a predicate. The statement (3x)Q(x) means there is some
value of x for which the predicate Q{x ) is true. The symbol 3 is called
an existential quantifier, and a statement of the form ( 3x)Q(x ) is
called an existential statement. A value c for which Q(c) is true is
called a witness to the statement ( 3x)Q(x ).
75.3.1 Remark One may indicate the type of the variable in an existential state-
ment in the same way as in a universal statement.
75.3.2 Example Let x be a real variable and let Q{x ) be the predicate x > 50.
This is certainly not true for all integers x . Q(40) is false, for example.
However,
Q(62) is true. Thus there are some integers x for which Q(x) is true. Therefore
(zh:R)<5(x) is true, and 62 is a witness.
75.3.3 Exercise Find an existential statement about real numbers with exactly
42 witnesses.
counterexample 112
definition 4
even 5
existential quanti-
fier 113
existential state-
ment 5, 113
implication 35, 36
infinite 174
integer 3
natural number 3
predicate calcu-
lus 113
predicate 16
prime 10
propositional calcu-
lus 107
usage 2
witness 113
75.3.4 Exercise In the following sentences, the variables are always natural num-
bers. P(n) means n is a prime, E{n) means n is even. State which are true and
which are false. Give reasons for your answers.
a) (3n) (E(n) A P(n ))
b) (Vn)(L(n)VP(n))
c) (3n)(E(n) =4- P(n))
d) (\/n)(E(n) =4- P(n))
(Answer on page 247.)
75.3.5 Exercise Which of these statements are true for all possible one-variable
predicates P(x) and Q(x)? Give counterexamples for those which are not always
true.
a) (\/x)(P(x) A Q(x)) =4- (V.x)P(x) A (\/x)Q(x)
b) (Vx)P(x) A (\/x)Q(x) =t- (\/x)(P(x) A Q(x))
c) (3x)(P(x) A Q(x)) =4- (3x)P(x) A (3x)Q(x)
d) (3x)P(x) A (3x)Q(x) =4- (3x)(P(x) A Q(x))
(Answer on page 247.)
75.3.6 Exercise Do the same as for Problem 75.3.5 with ‘V’ in the statements in
place of ‘A’.
75.3.7 Exercise Do the same as for Problem 75.3.5 with ‘ =t- ’ in the statements
in place of ‘A’.
75.3.8 Usage The symbols V and 3 are called quantifiers. The use of quantifiers
makes an extension of the propositional calculus called the predicate calculus
which allows one to say things about an infinite number of instances in a way that
the propositional calculus does not.
114
divide 4
GCD 88
implication 35, 36
integer 3
predicate 16
proposition 15
76. Variables and quantifiers
If a predicate P{x) has only one variable x in it, then using any quantifier in
front
of P{x) with respect to that variable turns the statement into one which is either
true or false — in other words, into a proposition.
76.1.1 Example If we let P{n) be the statement (n > 4) A (n < 6) , for n rang-
ing over the integers, then (3 n)P(n), since P( 5) is true (5 is a witness).
However,
(Vn)P(n) is false, because for example P( 6) is false (6 is a counterexample). Both
statements (3 n)P(n) and (Vn)P(n) are propositions; propositions, unlike predi-
cates, are statements which are definitely true or false.
76.1.2 Predicates with more than one variable When a predicate has more
than one variable, complications ensue. Let P(x,y) be the predicate (x > 5) V (5 >
y) . Let Q(y) be the predicate (Vx:N )P(x,y). Then Q(y) is the statement: “For
every integer x , x > 5 or 5 > y.” This is still not a proposition. It contains one
variable y . for which you can substitute an integer. It makes no sense to
substitute
an integer for x in Q(y) (what would “For all 14, 14 > 5 or 5 > y” mean?) which
is why x is not shown in the expression “Q(y) n .
76.1.3 Bound and free A variable which is controlled by a quantifier in an
expression is bound in the sense of 20.2. A logical expression in which all vari-
ables are bound is a proposition which is either true or false. If there are one or
more free variables, it is not a proposition, but it is still a predicate.
76.1.4 Exercise Let P(x,y ) be the predicate
(x = y) V (x > 5)
If possible, find a counterexample to (Vy)P(14,y) and find a witness to
(3x)P(x,3) .
(Answer on page 247.)
76.1.5 Exercise Let Q(m,n) be each of the following statements. Determine in
each case if (Vm:N)Q(m, 12) and (3n:Z)Q(3,ro) are true and give a counterexample
or witness when appropriate.
a) m\n.
b) GCD(m,n) = l.
c) (m | n) =>■ (m | 2 n) .
d) ( m | n) =4- (mn = 12) .
115
Archimedean prop-
erty 115
implication 35, 36
integer 3
proof 4
real number 12
rule of inference 24
theorem 2
trunc 86
In other words, “For any real number x there is an integer n bigger than x.”
Proof If you are given a real number x, then trunc(x) + 1 is an integer bigger
than x.
77.1.2 Example On the other hand, the statement
(3n:N)(Vx:R)(x < n) (77.2)
is false. It says there is an integer which is bigger than any real number. That is
certainly not true: if you think 456,789 is bigger than any real number, then I
reply,
“It is not bigger than 456,790”. In general, for any integer n, n + 1 is bigger —
and
of course it is a real number, like any integer.
As these examples illustrate, in general, (\/x)(3y)P(x,y) does not mean the same
as (3y)(\/x)P(x,y) , although of course for particular statements both might be
true.
On the other hand, two occurrences of the same quantifier in a row can be
interchanged:
77.2 Theorem
For any statement P(x,y),
(Vx)(Vy)P(x,y) |- (Vy)(Vx)P(x,y)
(77.3)
and
(\/y)(\/x)P(x,y) |- (\/x)(Vy)P(x,y)
(77.4)
and similarly
(3x)(3y)P(x,y) \- (3y)(3x)P(x,y)
(77.5)
and
{3y)(3x)P(x,y) |- (3x)(3y)P(x,y)
(77.6)
77.2.1 Exercise Are these statements true or false? Explain your answers. All
variables are real.
a) (Vx)(3y)(x>y).
b) (3x)(Vy)(x>y)
c) (3x)(3y)((x>y) => (x = y)).
(Answer on page 247.)
77. Order of quantifiers
Many important mathematical principles are statements with several quantified
variables. The ordering of the quantifiers matters. The subtleties involved can
be confusing.
77.1.1 Example The following statement is the Archimedean property of the
real numbers.
(Vx:R)(3n:N)(x < n) (77. 1)
116
counterexample 112
divide 4
equivalence 40
equivalent 40
implication 35, 36
integer 3
negation 22
positive integer 3
predicate 16
prime 10
proof 4
proposition 15
real number 12
theorem 2
77.2.2 Exercise Are these statements true or false? Explain your answers. All
variables are of type integer.
a) (Vm)(3n)(m | n) .
b) (3m)(Vn)(m | n) .
c) (Vm)(3n) ((m | n) =4- (m\mn)).
d) (3m)(Vn) ((m | n) =4- (m|?nn)).
77.2.3 Exercise Are these statements true or false? Give counterexamples if they
are false. In these statements, p and q are primes and m and n are positive
integers.
a) (Vp) (Vm) (Vn) ((p | m =4> p \ n) =4 m \ n)
b) (Vm)(Vn)(m|n =4 (3p)(p | m Ap | n))
77.2.4 Exercise (hard) Are these equivalences true for all predicates P and Q1
Assume that the only variable in P is x and the only variables in Q are x and y.
Give reasons for your answer.
a) (Vx)(3y)(P(z) =4 Q(x,y)) 44 (Vx) (P(x) (3 y)Q(x,y))
b) (3 x)(\/y)(P(x) => Q(x,y)) <&(3x){P(x) => (Vy)Q{x,y))
78. Negating quantifiers
Negating quantifiers must be handled with care, too:
78.1 Theorem: Moving “not” past a quantifier
For any predicate P ,
Q.l -((3 x)P{x)) 44 (Vx)(-iP(s))
Q.2 -.((Vs)P(x)) 44 (3z)(-.P(z)).
Proof We give the argument for Q.l; the argument for Q.2 is similar.
For (3 x:A)P(x) to be false requires that P(x) be false for every x of type A;
in other words, that —>P(x) be true for every x of type A. For example, if P(x) is
the predicate (x > 5) A (x < 3) , then (Bx:R)P(x) is false. In other words, the
rule
Q.l is valid.
78.1.1 Remark Finding the negation of a proposition with several quantifiers can
be done mechanically by applying the rules (Q.l) and (Q.2) over and over.
78.1.2 Example The negation of the Archimedean property can take any of the
following equivalent forms:
a) -i((Vx:R)(3?r:N)(x < n))
b) (3x:R)-i((3ro:N)(x < n))
c) (3x:R)(Vn:N)(x > n)
The last version is easiest to read, and clearly false — there is no real number
bigger than any integer. It is usually true that the easiest form to understand is
the
one with the ‘ -i ’ as “far in as possible” .
117
78.1.3 Worked Exercise Express the negation of (Vx)(x < 7) without using a
word or symbol meaning “not”.
Answer (3.x) (x > 7) .
78.1.4 Exercise Express the negation of (3x)(x < 7) without using a word or
symbol meaning “not” .
78.1.5 Exercise Write a statement in symbolic form equivalent to the negation
of
(Vx)(P(x) => Q(x))
equivalent 40
implication 35, 36
negation 22
nonnegative integer 3
predicate calcu-
lus 113
predicate 16
real number 12
without using the ‘V’ symbol.
78.1.6 Exercise Write a statement in symbolic form equivalent to the negation
of the expression “(3x)(P(x) =$■ -> Q(x))" without using ‘3’, ‘ =$■ ’ or ‘-i’.
79. Reading and writing quantified statements
An annoying fact about the predicate calculus is that even when you get pretty
good at disentangling complicated logical statements, you may still have trouble
reading mathematical proofs. One reason for this may be unfanriliarity with certain
techniques of proof, some of which are discussed in the next chapter. Another is
the variety of ways a statement in logic can be written in English prose. You have
already seen the many ways an implication can be written (Section 27).
Much more about reading mathematical writing may be found in the author’s
works [Wells, 1995], [Bagchi and Wells, 1998b], [Bagchi and Wells, 1998a], and
[Wells, 1998].
79.1.1 Example The true statement, for real numbers,
(Vx)(x>0 => (3y)(y 2 = x)) (79.1)
could be written in a math text in any of the following ways:
a) If x > 0 , then there is a y for which y 2 = x.
b) For any x > 0 , there is some y such that y 2 = x .
c) If x is nonnegative, then it is the square of some real number.
d) Any nonnegative real number is the square of another one.
e) A nonnegative real number has a square root.
Or it could be set off this way
x > 0 =4- (3 y)(y 2 =x) (x)
with the (x) on the far right side denoting “Vx”. Sometimes (x) is used instead of
Vx next to the predicate, too:
(x)(x > 0 =>- (3 y)(y 2 = x))
118
implication 35, 36
integer 3
predicate 16
quantifier 20, 113
real number 12
79.1.2 Warning The words “any”, “all” and “every” have rather delicate rules of
usage, as well. Sometimes they are interchangeable and sometimes not. The Archi-
medean axiom could be stated, “For every real x there is an integer n > x,” or “For
any real x there is an integer n > x.” But it would be misleading, although perhaps
not strictly wrong, to say, “For all real numbers x there is an integer n > x ,”
which
could be misread as claiming that there is one integer n that works for all x.
79.1.3 Warning Observe that the statements in (a), (c) and (e) have no obvious
English word corresponding to the quantifier. This usage there is somewhat similar
to the use of the word “dog” in a sentence such as, “A wolf mates for life”,
meaning
every wolf mates for life.
Students sometimes respond to a question such as, “Prove that an integer divis-
ible by 4 is even” with an answer such as, “The integer 12 is divisible by 4 and it
is even”. However, the question means, “Prove that every integer divisible by 4 is
even.” This blunder is the result of not understanding the way a universal
quantifier
can be signaled by the indefinite article.
79.1.4 Example Consider the well-known remark, “All that glitters is not gold.”
This statement means
-i(Vx) (GLITTER(x) => GOLD(x))
rather than
(Vx) (GLITTER(x) => ^GOLD(x))
In other words, it means, “Not all that glitters is gold.” (We do not say the
statement is incorrect English or correct English with a different meaning; we only
give it as an illustration of the subtleties involved in translating from English
to
logic.)
79.1.5 Worked Exercise Write these statements in logical notation. Make up
suitable names for the predicates.
a) All people are mortal.
b) Some people are not mortal.
c) All people are not mortal.
Answer (a) (Vx) (Person(x) => Mortal(x))
(b) (3x) (Person(x) A -Mortal(x))
(c) (Vx) (Person(x) =4- -iMortal(x))
79.1.6 Exercise Write these statements in logical notation.
a) Everybody likes somebody.
b) Everybody doesn’t like something.
c) Nobody likes everything.
d) You can fool all of the people some of the time and some of the people all of
the time, but you can’t fool all of the people all of the time.
79.1.7 Exercise Write the statement in GS.2, page 61, using quantifiers.
119
80. Proving implications: the Direct Method
Because so many mathematical theorems are implications, it is worthwhile consid-
ering the ways in which an implication can be proved. We consider two common
approaches in this chapter.
80.1 The direct method
If you can deduce Q from P , then P =>- Q must be true. That is because the only
line of the truth table for ‘ => ’ (Table 25.1) which has an ‘F’ is the line for
which
P is true and Q is false, which cannot happen if you can deduce Q from P . This
gives:
80.1.1 Method: Direct Method
To prove P =>■ Q , assume P is true and deduce Q .
direct method 119
divide 4
Fundamental Theo-
rem of Arith-
metic 87
hypothesis 36
implication 35, 36
integer 3
positive integer 3
prime 10
proof 4
theorem 2
truth table 22
80.1.2 Remark Normally, in proving Q, you would use other facts at your dis-
posal as well as the assumption that P is true. As an illustration of the direct
method, we prove the following theorem.
80.2 Theorem
If a positive integer is divisible by 2 then 2 occurs in its prime factor-
ization.
Proof Let n be divisible by 2. (Thus we assume the hypothesis is true.) Then 2
divides n, so that by definition of division n = 2m for some integer m. Let
m = pf 1 x ... x p^f
be the prime factorization of m. Then
n = 2 xpf 1 x ... x p
is a factorization of n into primes (since 2 is a prime), so is the prime
factorization
of n because the prime factorization is unique by the Fundamental Theorem of
Arithmetic.
80.2.1 Coming up with proofs In a more complicated situation, you might
have to prove P => P \ , P\ =A P 2 , . . . , Pk =>• Q in a series of deductions.
Normally, although your final proof would be written up in that order, you
would not think up the proof by thinking up P\,p 2 ,. . . in order. What happens
usually is that you think of statements which imply Q, statements which imply
them (backing up), and at the same time you think of statements which P implies,
statements which they imply (going forward), and so on, until your chain meets in
the middle (if you are lucky) . Thinking up a proof is thus a creative act rather
than
the cut-and-dried one of grinding out conclusions from hypotheses.
80.2.2 Exercise Prove by the direct method that for any integer n, if n is even
so is n 2 .
120
conclusion 36
contrapositive 42
direct method 119
divide 4
equivalent 40
even 5
hypothesis 36
implication 35, 36
integer 3
odd 5
positive integer 3
prime 10
proof 4
theorem 2
universal generaliza-
tion 6
81. Proving implications: the Contrapositive Method
It is very common to use the contrapositive to prove an implication. Since “P =4>
Q”
is equivalent to Q => -cP” , you can prove " P =$■ Q” by using the direct method
to prove u -iQ =>• -> P” . In detail:
81.0.3 Method: Contrapositive Method
(The contrapositive method) To prove P Q , assume Q is false and
deduce that P is false.
81.0. 4 Warning This method is typically used in math texts without mentioning
that the contrapositive is being used. You have to realize that yourself.
81.0. 5 Example The proof of the following theorem is an illustration of the use
of the contrapositive, written the way it might be written in a math text. Recall
that an integer k is even if 2 | k .
81.1 Theorem
For all positive integers n, if n 2 is even, so is n.
Proof Let n be odd. Then 2 does not occur in the prime factorization of n. But
the prime factorization of n 2 merely repeats each prime occurring in the
factoriza-
tion of n, so no new primes occur. So 2 does not occur in the factorization of n 2
either, so by Theorem 80.2, n 2 is odd. This proves the theorem.
81.1.1 Remarks
a) If you didn’t think of proving the contrapositive, you might be dumbfounded
when you saw that a proof of a theorem which says “if n 2 is even then n is
even” begins with, “Let n be odd...” The contrapositive of the statement to
be proved is, “If n is odd, then n 2 is odd.” The proof of the contrapositive
proceeds like any direct-method proof, by assuming the hypothesis ( n is odd) .
b) The contrapositive of Theorem 80.2 is used in the proof of Theorem 81.1. That
theorem says that if n is even, then its prime factorization contains 2. Here
we are using it in its contrapositive form: if 2 does not occur in the prime
factorization of n, then n is not even, i.e., n is odd. Again, the proof does
not mention the fact that it is using Theorem 80.2 in the contrapositive form.
c) Theorem 81.1, like most theorems in mathematics, is a universally quanti-
fied implication, so using universal generalization we showed that if n is an
arbitrary positive integer satisfying the hypothesis, then it must satisfy the
conclusion. In such a proof, we are not allowed to make any special assump-
tions about n except that it satisfies the hypothesis. On the other hand,
if we suspected that the theorem were false, we could prove that it is false
merely by finding a single positive integer n satisfying the hypothesis but not
the conclusion. (Consider the statement, “If n is prime, then it is odd.”)
This phenomenon has been known to give students the impression that prov-
ing statements is much harder than disproving them, which somehow doesn’t
seem fair.
121
81.1.2 Exercise Prove by the contrapositive method that if n 2 is odd then so
is n.
81.2 Exercise set
Exercises 81.2.1 through 81.2.3 provide other methods of proof.
81.2.1 Exercise Prove that
(P A -iQ) -i(P =4- Q) (81.1)
is a tautology. Thus to prove that an implication is false, you must show that
its hypothesis is true and its conclusion is false. In particular, the negation of
an
implication is not an implication.
81.2.2 Exercise Prove that the rule
affirming the hypoth-
esis 121
conclusion 36
definition 4
denying the conse-
quent 121
divisor 5
equivalent 40
fallacy 121
hypothesis 36
implication 35, 36
negation 22
prime 10
rule of inference 24
tautology 105
=*. Q |_ P\/ Q
(81.2)
is a valid inference rule. (A proof using this rule would typically begin the proof
of
PVQ by saying, “Assume -i P ..." and then proceed to deduce Q.)
81.2.3 Exercise Prove that the rule
P^Q,Q^R^P^R
is a valid inference rule. (This allows proofs to be strung together.)
81.2.4 Exercise (hard) Use the methods of this chapter to prove that n is prime
if and only if n > 1 and there is no divisor k of n satisfying 1 < k < \/n.
82. Fallacies connected with implication
82.1 Definition: fallacy
An argument which does not use correct rules of inference is called a
fallacy.
82.1.1 Example Two very common fallacies concerning implications are
F.l assuming that from P =>- Q and Q you can derive P (“A cow eats grass. This
animal eats grass, so it must be a cow.”) and
F.2 assuming that from P => Q and —> P that you can derive —>Q (“A cow eats
grass. This animal is not a cow, so it won’t eat grass.”)
82.1.2 Remark You will sometimes hear these fallacies used in political argu-
ments. F.l is called affirming the hypothesis and F.2 is called denying the
consequent.
82.1.3 Remark Fallacious arguments involve an incorrect use of logic, although
both the hypothesis and the conclusion might accidentally be correct. Fallacious
arguments should be distinguished from correct arguments based on faulty assump-
tions.
122
conclusion 36
contrapositive 42
equivalence 40
equivalent 40
even 5
hypothesis 36
implication 35, 36
integer 3
odd 5
positive integer 3
prime 10
82.1.4 Example The statement, “A prime number bigger than 2 is odd. 5 is odd,
so 5 is prime” is fallacious, even though the conclusion is true. (The hypothesis
is
true, too!). It is an example of affirming the hypothesis.
82.1.5 Example The statement “An odd number is prime, 15 is odd, so 15 is
prime” is not fallacious- it is a logically correct argument based on an incorrect
hypothesis (“garbage in, garbage out”).
82.1.6 Example The argument, “Any prime is odd, 16 is even, so 16 is not a
prime” is a logically correct argument with a correct conclusion, but the
hypothesis,
“Any prime is odd”, is false. The latter is a case of “getting the right answer for
the wrong reason,” which is a frequent source of friction between students and math
teachers.
82.2 Exercise set
In Problems 82.2.1 through 82.2.5, some arguments are valid and some are
fallacious.
Some of the valid ones have false hypotheses and some do not. (The hypothesis is in
square brackets.) State the method of proof used in those that are valid and
explain
the fallacy in the others. The variable n is of positive integer type.
82.2.1 [n > 5 only if n> 3]. Since 17 > 5, it must be that 17 > 3. (Answer on
page 247.)
82.2.2 [n > 5 only if n > 3]. Since 4 > 3, it must be that 4 > 5. (Answer on page
247.)
82.2.3 [If n is odd, then n / 2], 6 is not odd, so 6 = 2. (Answer on page 247.)
82.2.4 [n is odd only if it is prime]. 17 is odd, so 17 is a prime. (Answer on page
247.)
82.2.5 [If n is even and n > 2, then n is not prime]. 15 is odd, so 15 is prime.
(Answer on page 247.)
83. Proving equivalences
83.1.1 Method
An equivalence “P 44 Q” is proved by proving both P => Q
and Q =4- P.
83.1.2 Remark Remember the slogan: To prove an equivalence you must prove
two implications.
83.1.3 Remark Quite commonly the actual proof proves (for example) P => Q
and -i P =4- -iQ (the contrapositive of Q =4- P), so the proof has two parts: the
first part begins, “Assume P” , and the second part begins, “Assume —>P...”
123
83.1.4 Example Here is an example of a theorem with such a proof. The proof
avoids the use of the Fundamental Theorem of Arithmetic, which would make it
easier, so as to provide a reasonable example of the discussion in the preceding
paragraph.
83.2 Theorem
For any integer n, 2 | n if and only if 4 | n 2 .
Proof If 2 | n then by definition there is an integer k for which n = 2k. Then
n 2 = 4 k 2 , so n 2 is divisible by 4.
Now suppose 2 does not divide n , so that n is odd. That means that n = 2k + 1
for some integer k . Then n 2 = 4k 2 + Ak + 1 = 2(2k 2 + 2k) + 1 which is odd, so
is
not divisible by 2, much less by 4.
contrapositive
method 120
direct method 119
divide 4
equivalent 40
even 5
Fundamental Theo-
rem of Arith-
metic 87
implication 35, 36
integer 3
odd 5
positive integer 3
proof 4
theorem 2
83.2.1 Remark The preceding proof is written the way such proofs commonly
appear in number theory texts: no overt statement is made concerning the structure
of the proof. You have to deduce the structure by the way it proceeds. In this
proof,
P is the statement “2 | n” and Q is the statement “4 | n 2 ”. To prove P 44 Q , the
proof proceeds to prove first (before the phrase “Now suppose”) that P =>■ Q by
the direct method, and then to prove that Q => P by the contrapositive method,
that is, by proving —>P => — i Q by the direct method.
83.2.2 Exercise Prove that for all integers m and n, m + n is even if and only
if rn — n is even.
83.2.3 Exercise Let cc be a relation on a set A. Prove that a is reflexive if and
only if C a .
83.2.4 Exercise Let a be a relation on a set A. Prove that a is antisymmetric
if and only if
a n a op C Aq
84. Multiple equivalences
Some theorems are in the form of assertions that three or more statements are
equivalent.
This theorem provides an example:
84.2 Theorem
The following are equivalent for a positive integer n:
D.l n is divisible by 4.
D.2 n/2 is an even integer.
D.3 n/4 is an integer.
124
conclusion 36
div 82
equivalent 40
implication 35, 36
include 43
integer 3
mod 82, 204
nonnegative integer 3
positive integer 3
quotient (of inte-
gers) 83
relation 73
remainder 83
rule of inference 24
symmetric 78, 232
84.2.1 Remark In proving such a theorem, it is only necessary to prove three
implications, not six, provided the three are chosen correctly. For example, it
would
be sufficient to prove P => Q , Q => R and R => P . Then for example Q => P
follows from Q =>■ R and R =>■ P. (See Problem 84.2.3).
84.2.2 Warning Theorem 84.2 does not say that n is divisible by 4. It says that
if one of the statements is true, the other two must be true also (so if one is
false
the other two must be false). It therefore says
(P 44 Q) A (Q 44 R) A (P 44 R)
for certain statements P, Q and R. That is the same as asserting six implications,
P=^Q,Q=^P,P=^R,R=^P,Q=^R, and R =>■ Q .
84.2.3 Exercise Write out careful proofs of Theorem 84.2 in two ways:
a) (D.l) => (D.2), (D.2) =$■ (D.3), and (D.3) =4- (D.l), and
b) (D.l) => (D.3), (D.3) => (D.2), and (D.2) =* (D.l).
84.2.4 Exercise Prove that the following three statements are equivalent for any
sets A and B:
a) AC B
b) AU B = B
c) AnB = A
84.2.5 Exercise Let a be a relation on a set A. Prove that the following three
statements are equivalent.
a) a is symmetric.
b) a C a op .
c) a = a op .
85. Uniqueness theorems
In a particular system such as the positive integers, any uniqueness theorem gives
a
rule of inference. Such a rule only applies to the data type for which the
uniqueness
theorem is stated.
85.1.1 Example Theorem 60.2 says that the quotient and remainder are uniquely
determined by Definition 60.1. This provides a rule of inference for nonnegative
integers:
m = qn + r, 0 < r < n (— (q = mdivn) A (r = mmodn) (85.1)
85.1.2 Remark The conclusion of this rule of inference can be worded this way:
q is the quotient and r is the remainder when rn is divided by n. For example,
because to = 50, n = 12, q = 4 and r = 2 satisfy Rule (85.1), 4 = 50 div 12 and
2 = 50 mod 12. You do not have to do a long division to verify that; it follows
from
Rule (85.1).
125
85.1.3 Exercise Prove that if 0 <m — qn<n, then q = mdivn. (Answer on
page 247.)
85.1.4 Exercise Use Rule (85.1) to prove that if r = mmodn and r' = m/rnodn,
then (m + m^modn is either r + r' or r + r 7 — n.
85.1.5 Exercise Use Rule (85.1) to prove that if r = mmodn, then n\m — r.
85.1.6 A rule for GCD’s For positive integers m and n, the greatest common
divisor GCD(m,?r) is the largest integer dividing both m and n; this definition was
also given in Chapter 60. This obviously determines the GCD uniquely — there
cannot be two largest integers which divide both m and n. This can be translated
into a rule of inference:
conclusion 36
divide 4
div 82
exponent 87
GCD 88
implication 35, 36
integer 3
mod 82, 204
positive integer 3
prime 10
rule of inference 24
tautology 105
theorem 2
(Ve) (^[d | m) A (d \ n) A ((e | m A e | n) =4- e<d)^j (— d = GCD(m,n) (85.2)
85.1.7 Example Let’s use the rule just given to prove Theorem 64.1. We must
prove that the number d which is the product of all the numbers p mm ( e p( m )> e
p( n ))
for all primes p which divide m or n or both is GCD(m,n) .
First, d | m and d | n, since the exponent of any prime p in d, which is
min(e p (m),e p (n))
is obviously less than or equal to e p (m) and to e p (n) , so Theorem 62.4
applies. Thus
we have verified two of the three hypotheses of Rule (85.2). As for the third,
suppose
e | m and e\n. Then e p (e) < e p (m ) and e p (e) < e p (n) , so e p (e) < min(e p
(m),e p (n)) ,
so e | d. But if e\d, then e < d, so the third part of Rule (85.2) is correct.
Hence
the conclusion that d = GCD (m,n) must be true.
85.1.8 Exercise State and prove a rule like Rule (85.2) for LCM(m,?i).
86. Proof by Contradiction
Another hard-to-understand method of proof is proof by contradiction, one form of
which is expressed by this rule of inference:
86.1 Theorem
-iQ, P => Q (— -i-P
( 86 . 1 )
86.1.1 Remarks
a) Theorem 86.1 follows from the tautology
(-.QA(P Q )) -i P
b) This rule says that to prove -i P it suffices to prove ~>Q and that P =f- Q .
126
decimal 12, 93
divide 4
even 5
factor 5
finite 173
Fundamental Theo-
rem of Arith-
metic 87
implication 35, 36
infinite 174
integer 3
odd 5
prime 10
proof by contradic-
tion 126
proof 4
rational 11
real number 12
reductio ad absur-
dum 126
remainder 83
rule of inference 24
theorem 2
usage 2
86.1.2 Usage A proof using the inference rule of Theorem 86.1 is called proof
by contradiction, or reductio ad absurdum (“r.a.a”).
86.1.3 Remarks
a) In practice it frequently happens that Q is obviously false so that the work
goes into proving P =t- Q . Thus a proof of —> P by contradiction might begin,
“Suppose P is true ... ” !
b) Authors typically don’t tell the reader they are doing a proof by contradiction.
It is generally true that mathematical authors are very careful to tell the reader
which previous or known theorems his proof depends on, but says nothing at
all about the rule of inference or method of proof being used.
As an illustration of proof by contradiction, we will prove this famous theorem:
86.2 Theorem
V2 is not rational.
86.2.1 Remarks
a) The discovery of this theorem by an unknown person in Pythagoras’ religious
colony in ancient Italy caused quite a scandal, because the “fact” that any
real number could be expressed as a fraction of integers was one of the beliefs
of their religion (another was that beans were holy).
b) Theorem 86.2 is a remarkable statement: it says that there is no fraction m/n
for which (m/n) 2 = 2. Although y/2 is approximately equal to 1414/1000, it
is not exactly equal to any fraction of integers whatever. The fact that y/2
has a nonterminating decimal expansion does not of course prove this, since
plenty of fractions (e.g., 1/3) have nonterminating decimal expansions.
How on earth do you prove an impossibility statement like that? After all,
you can’t go through the integers checking every fraction m/n. It is that sort
of situation that demands a proof by contradiction.
Proof Here is the proof, using the Fundamental Theorem of Arithmetic. Suppose
y/2 is rational, so that for some integers m and n, 2 = (m/n) 2 . Then 2 n 2 = m
2 .
Every prime factor in the square of an integer must occur an even number of times.
Thus e 2 (m 2 ) is even and ^(n 2 ) is even. But e2(2n 2 ) = l + e 2 (n 2 ), so
e2(2n 2 ) is
odd, a contradiction.
86.2.2 Remark In fact, n (and many other numbers used in calculus) is not
rational either, but the proof is harder.
86.2.3 Worked Exercise Use the Fundamental Theorem of Arithmetic to prove
that there are an infinite number of primes.
Answer This will be a proof by contradiction. Suppose there is a finite number
of primes: suppose that p\,p 2 ,... ,Pk are all the primes. Let m = p\ -p 2 ■ ■ -pk
+ 1 •
Then the remainder when m is divided by any prime is 1 . Since no prime divides
m, it cannot have a prime factorization, contradicting the Fundamental Theorem
of Arithmetic.
127
86.2.4 Exercise Use proof by contradiction to prove that if p is a prime and
p > 2, then p is odd. (Answer on page 247.)
86.2.5 Exercise Prove that for all rational numbers x, ( x 2 < 2) ( x 2 < 2) .
86.2.6 Exercise Give an example of a pair of distinct irrational numbers r and
s with the property that r + s is rational.
86.2.7 Exercise Use proof by contradiction to prove that if r and s are real
numbers and r is rational and s is not rational, then r + s is not rational.
86.2.8 Exercise Use proof by contradiction to prove that for any integer k > 1
and prime p, the k th root of p is not rational.
86.2.9 Exercise (hard) Use Problem 86.2.8 to prove that the kth. root of a
positive integer is either an integer or is not rational.
definition 4
equivalent 40
even 5
Fundamental Theo-
rem of Arith-
metic 87
integer 3
integral linear combi-
nation 127
mod 82, 204
odd 5
positive integer 3
prime 10
proof by contradic-
tion 126
rational 11
86.2.10 Exercise (hard) Show that there are infinitely many primes p such that
p mod 4 = 3. Hint: Use proof by contradiction. Assume there are only finitely
many such primes, and consider the number m which is the product of all of them.
Consider two cases, rn mod 4 = 1 and mmod4 = 3, and ask what primes can divide
771 + 2 or 777 + 4. Use problem 60.5.6, page 85 and other similar facts. Note that
the
similar statement about p mod 4 = 1 is also true but much harder to prove.
87. Bezout’s Lemma
The Fundamental Theorem of Arithmetic, that every integer greater than one has a
unique factorization as a product of primes, was stated without proof in Chapter
62.
It actually follows from certain facts about the GCD by a fairly complicated proof
by
contradiction. This proof is based on Theorem 87.2 below, a theorem which is worth
knowing for its own sake. The proof of the Fundamental Theorem is completed in
Problems 104.4.1 through 104.4.4.
87.1 Definition: integral linear combination
If 777 and 77 are integers, an integral linear combination of 777 . and n
is an integer d which is expressible in the form d = am + bn , where a
and b are integers.
87.1.1 Example 2 is an integral linear combination of 10 and 14, since
3x10-2x14 = 2
However, 1 is not an integral linear combination of 10 and 14, since any integral
linear combination of 10 and 14 must clearly be even.
87.1.2 Remark Note that in the definition of integral linear combination, the
expression d = am . + bn does not determine a and b uniquely for a given m and n .
128
divide 4
Euclidean algo-
rithm 92
Fundamental Theo-
rem of Arith-
metic 87
GCD 88
integer 3
integral linear combi-
nation 127
intersection 47
mod 82, 204
positive integer 3
theorem 2
87.1.3 Example 3x10 — 2x14 = 2 and — 4 x 10 + 3 x 14 = 2. (See Exer-
cise 88.3.7.)
87.1.4 Exercise Show that if d\m and d\n then d divides any integral linear
combination of m and n.
87.2 Theorem: Bezout’s Lemma
If m and n are positive integers, then GCD(m,n) is the smallest positive
integral linear combination of m and n .
87.2.1 Remark Bezout’s Lemma should not be confused with Bezout’s Theorem,
which is a much more substantial mathematical result concerning intersections of
surfaces defined by polynomial equations.
87.2.2 Example GCD(10,14) = 2, and 2 is an integral linear combination of 10
and 14 (2 = 3 • 10 + (— 2) • 14) but 1 is not, so 2 is the smallest positive
integral
linear combination of 10 and 14.
87.2.3 Proof of Bezout’s Lemma We prove this without using the Fundamental
Theorem of Arithmetic, since the lemma will be used later to prove the Fundamental
Theorem.
Let e be the smallest positive integral linear combination of m and n . Suppose
e = am J rbn. Let d = GCD(m,n).
First, we show that d < e. We know that d \ m and d\n, so there are integers h
and k for which m = dh and n = dk. Then e = am + bn = adh + bdk = d(ah + bk )
is divisible by d . It follows that d<e.
Now we show that e | m and e | n . Let m = eq + r with 0 < r < e . Then
r = m — eq = m — ( am. + bn)q = (1 — aq)m — bqn
so r is an integral linear combination of m and n. Since e is the smallest positive
integral linear combination of m and n and r < e , this means r = 0 , so e | m . A
similar argument shows that e | n .
It follows that e is a common divisor of m and n and d is the greatest common
divisor ; hence e <d. Combined with the previous result that d < e, we see that
d = e, as required.
88. A constructive proof of Bezout’s Lemma
The preceding proof of Bezout’s Lemma does not tell us how to calculate the
integers
a and b for which am + bn = GCD (m,n). For example, see how fast you can find
integers a and b for which 13a + 216= 1. (See Exercise 107.3.4.)
We now give a modification of the Euclidean algorithm which constructs integers
a and 6 for which GCD(m,n) = am + bn. The Euclidean algorithm is given as
program 65.1, page 93, based on Theorem 65.1, which says that for any integers rn
and n, GCD(m,n) = GCD(n,mnrodn) . Program 65.1 starts with M and N and
129
repeatedly replaces N by M mod N and M by N . The last value of N before it
becomes 0 is the GCD. This lemma shows how being an integral linear combination
is preserved by that process:
88.1 Lemma
Let m and n be positive integers.
B.l The integers m and n are integral linear combinations of m and
n.
B.2 If u and v are integral linear combinations of m and n and
then umodv is also an integral linear combination of m and n.
div 82
Euclidean algo-
rithm 92
GCD 88
integer 3
integral linear combi-
nation 127
lemma 2
mod 82, 204
positive integer 3
proof 4
Proof B.l is trivial: m=lxm + 0xn and n = 0xm+lxn. As for B.2, suppose
u = wm + xn and v = yrri + zn . Let u = qv + r with 0 < r < u , so r = u mod v .
Then
r = u — qv = wm + xn — q(ym + zn) = (w — qy)m + (x — qz)n
so r is an integral linear combination of m and n, too.
88.2 A method for calculating the Bezout coefficients
We now describe a method for calculating the Bezout coefficients based on Lemma
88.1.
Given positive integers m and n with d = GCD(m,n) , we calculate integers a and
b for which am + bn = d as follows: Make a table with columns labeled u, v, iv
and w = am + bn .
1. Put u = m, v = n, w = mrnodn in the first row, and in the last column put
the equation w = m — (mdivn)n. Note that this equation expresses mrnodn
in the form am + bn (here a = 1 and b = — m divn).
2. Make each succeeding row v! , v' , w' , w' = a' m + b'n by setting u! =v (the
entry under v in the preceding row) , v' = w and w' = v mod w , and solving
for a' and b' by using the equation w' = v! — (u' divv')v' and the equations in
the preceding rows. Note that the entry in the last column always expresses
w in terms of the original m and n, not in terms of the u and v in that row.
3. Continue this process until the entry under w is GCD(?n,n) (this always
happens because the first three columns in the process constitute the Euclidean
algorithm) .
88.2.1 Example The following table shows the calculation of integers a and b
for which 100a + 366 = 4 .
u v w
100 36 28 28 = 100-2-36 Note that 100 div 36 = 2
36 28 8 8 = 36- 28 = 36- (100-2 -36) = 3-36- 100
28 8 4 4 = 28- 3 -8 = 100- 2 -36-3(3-36- 100) =4- 100- 11 -36
so that a = 4 , 6 = — 11.
130
constructive 130
divide 4
Fundamental Theo-
rem of Arith-
metic 87
GCD 88
infinite 174
integer 3
integral linear combi-
nation 127
nonconstructive 130
relatively prime 89
rule of inference 24
88.3 Constructive and nonconstructive
The two proofs we have given for Theorem 87.2 illustrate a common phenomenon
in mathematics. The first proof is nonconstructive; it shows that the requisite
integers a and 6 exist but does not tell you how to get them. The second proof is
constructive; it is more complicated but gives an explicit way of constructing a
and b.
88.3.1 Exercise Express a as an integral linear combination of b and c, or explain
why this cannot be done.
12
16
4
12
16
26
30
26
30
26
30
51
100
(Answer on page 247.)
88.3.2 Exercise Express 1 as an integral linear combination of 13 and 21.
88.3.3 Exercise (M. Leitman) Suppose a, b, m and n are integers. Prove that
if m and n are relatively prime and am + bn = e, then there are integers a' and b'
for which a'm + b'n = e + 1 . (Answer on page 247.)
88.3.4 Exercise Prove without using the Fundamental Theorem of Arithmetic
that if GCD(m,n) = 1 and m \ nr then m\r . (Use Bezout’s Lemma, page 128.)
88.3.5 Exercise Suppose that a, b and c are positive integers for which c =
12a — 86. Show that GCD(a,6) < |.
88.3.6 Exercise Prove that the following rule of inference is valid (use Bezout’s
Lemma, page 128).
e\m, e\n (— e | GCD(m,n)
(It follows that the statement “e < d” in Rule (85.2) can be replaced by “e | d” .)
88.3.7 Exercise (hard) Prove that if d is an integral linear combination of m
and n then there are an infinite number of different pairs of integers a and 6 for
which d = am + bn .
88.3.8 Exercise Use Bezout’s Lemma (page 128) to prove Corollary 64.2 on
page 90 without using the Fundamental Theorem of Arithmetic.
131
89. The image of a function
If F : A — >■ B is a function, it can easily happen that not every element of B is
a value of F . For example, the function x >— > x 2 : R — > R takes only
nonnegative
values.
89.1 Definition: image of a function
The image of F : A-^ B is the set of all values of F , in other words
the set {b £ B \ (3a: A)(F(a) = b)} . The image of F is also denoted
Irn(F) .
codomain 56
definition 4
equivalence 40
equivalent 40
fact 1
function 56
image 131
include 43
real number 12
take 57
usage 2
89.1.1 Fact This definition gives the equivalence:
(3a)(F(a) = b) 44 feelrnF
89.1.2 Fact For any function F, Im(F)CcodF.
89.1.3 Usage Many authors use the word “range” for the image, but others use
“range” for the codomain.
89.1.4 Example The image of the squaring function x i — > x 2 : R — > R is the set
of
nonnegative real numbers.
89.1.5 Example Let the function F : {1,2,3} — > {2, 4, 5, 6} be defined by F( 1) =
4
and F( 2) = F( 3) = 5. Then F has image {4,5}.
89.1.6 Remark The image of a function can be difficult to determine if it is
given by a formula; for example it requires a certain amount of analytic geometry
(or calculus) to determine that the image of the function G(x ) = x 2 + 2x + 5 is
the
set of real numbers > 4 , and determining the image of more complicated functions
can be very difficult indeed.
89.1.7 Exercise Find the image of the function n+ 1 : N — >• N. (Answer on
page 247.)
89.1.8 Exercise Find the image of the function m — > n — 1 : Z — > Z.
89.1.9 Exercise Find the image of the function x>— > x 2 — 1 : R — >-R.
89.1.10 Exercise Find the image of the function x>— )-x 2 + a: + l:R— >-R.
132
definition 4
function 56
image function 132
image 131
include 43
interval 31
inverse image 132
powerset 46
under 57, 132
90. The image of a subset of the domain
The word “image” is used in a more general way which actually makes the image a
function itself.
90.1 Definition: Image of a subset
Let F : A-^ B is a function, and suppose CCA. Then F{C) denotes
the set {F(x) \ x € C} , and is called the image of C under F . The
map C H > F(C) defines a function from VA to VB called the image
function of F .
90.1.1 Remark In particular, F(A) is what we called Inr(T 1 ) in Chapter 89.
90.1.2 Example If F : {1,2,3} — > {2, 4, 5, 6} is defined as in 89.1.5 by F(l) = 4
and F( 2) = F( 3) = 5, then F({1,2}) = {4,5} and F($) = 0. Thus the image of
{1,2} under F is {4,5}.
90.1.3 Warning The image function is not usually distinguished from F in nota-
tion. A few texts use F* : VA — > VB , and so would write F(x) for x G A but F*(C)
for a subset CCA. In this text, as in almost all mathematics texts, we simply write
F(C). Context usually disambiguates this notation (but there are exceptions!).
90.1.4 Exercise Describe a function where our notation F(C) is ambiguous.
90.1.5 Exercise Let F be defined as in Example 90.1.2. What are F({2,3}) and
F({3})? (Answer on page 247.)
90.1.6 Exercise Let F : R — > R be defined by F(x) = x 2 + 1 . What is F((3 . .
4)) ?
What is FQ-1..1])?
90.1.7 Exercise Let F be defined as in Example 90.1.2. How many ordered pairs
are in the graph of the image function of FI
91. Inverse images
91.1 Definition: Inverse image
Let F : A—> B be a function. For any subset C C B , the set
{acA | F(a) € C }
is called the inverse image of C under F . also written 1 ?_1 (C').
91.1.1 Example Let F : {1,2,3} — > {2, 4, 5, 6} be defined (as in Example 89.1.5)
by F{ 1) = 4 and F{ 2) = F( 3) = 5. Then F-^^,6}) = {1}, E“ 1 ({5}) = {2,3},
and F~ 1 ({2,6}) =0.
133
91.1.2 Example For the function F : R — >■ R defined by F(x) = x 2 + 1 ,
F“ 1 ([2..3]) = [l..\/2]U[-v/2..-l]
and
F- 1 ([0..1]) = {0}
91.1.3 Inverse image as function Like the image function, this inverse image
function can also be defined as a function F :VB — > V A (note the reversal), where
F~ 1 (D) = {xeA | F(x)eD}
for any D C B . F~ 1 is sometimes denoted F* .
91.1.4 Usage It is quite common to write F~ 1 (x) instead of _F _1 ({x}).
91.1.5 Example For the function of Example 91.1.2, F _1 (3) = {— \/2, V2)} .
codomain 56
definition 4
fact 1
function 56
graph (of a func-
tion) 61
image 131
include 43
inverse image 132
onto 133
powerset 46
real number 12
surjection 133
surjective 133
union 47
usage 2
91.1.6 Exercise Let F : R -> R be defined by F(x) = x 2 + 1 . What is F 1 ({1,2})?
What is F- 1 ^!.^))?
91.1.7 Exercise For any function F : A-+ B , what is F 1 (0)? What is F 1 (B)7
92. Surjectivity
92.1 Definition: surjective
Let F : A—> B be a function. F is said to be surjective if and only if
Im (F) = B.
92.1.1 Fact F : A—t B is surjective if and only if for every element element b 6 B
there is an element a G A for which F(a) = b.
92.1.2 Usage If F is surjective, it is said to be a surjection or to be onto.
92.1.3 Warning Whether a function is surjective or not depends on the codomain
you specify for it.
92.1.4 Example For the two functions S : R — > R and T : R — > R + of 39.7.3, with
S(x) = T(x) = x 2 , S is not surjective but T is. To say that T is surjective is to
say that every nonnegative real number has a square root. Authors who do not
normally specify codomains have to say, “T is surjective onto R + .”
92.1.5 Example A function F : R — > R is surjective if every horizontal line
crosses
its graph.
134
contrapositive 42
converse 42
coordinate func-
tion 63
definition 4
fact 1
function 56
identity function 63
identity 72
image 131
implication 35, 36
inclusion function 63
injection 134
injective 134
one to one 134
powerset 46
reflexive 77
relation 73
surjective 133
take 57
usage 2
92.1.6 Exercise How do you prove that a function F : A->- B is not surjective?
92.1.7 Exercise Let a be a relation on A.
a) Show that if a is reflexive, then the coordinate functions pf : a — > A and
P 2 '-u->A are surjective.
b) Show that the converse of (a) need not be true.
92.1.8 Exercise (hard) Show that there for any set S , no function from S to
VS is surjective. Do not assume S is finite.
Extended hint: If F : S — >■ VS is a function, consider the subset
{x | x is not an element of E(x)}
No argument that says anything like “the powerset of a set has more elements than
the
set” can possibly work for this problem, and therefore such arguments will not be
given
even part credit. The reason is that we have developed none of the theory of what
it means
to talk about the number of elements of an infinite set, and in any case this
problem is a
basic theorem of that theory.
Let’s be more specific: One such invalid argument is that the function that takes x
to
{cc} is an injective function from S to VS , and it clearly leaves out the empty
set (and
many others) so VS has “more elements” than S . This is an invalid argument.
Consider
the function from N to N that takes n to 42n. This is injective and leaves out lots
of
integers, so does N have more elements than itself?? (In any case you can come up
with
other functions from N to N that don’t leave out elements.)
93. Injectivity
93.1 Definition: injective
F : A—t B is injective if and only if different inputs give different out-
puts, in other words if a yf a! => F(a) / F(a') for all a, a' £ A.
93.1.1 Fact To say F : A — >• B is injective is equivalent to saying that F(a) =
F(a') =4- a = a' for all a, a' € A (the contrapositive of the definition).
93.1.2 Usage An injective function is called an injection or is said to be one to
one.
93.1.3 Example The squaring function S : R — > R is not injective since S(x) =
S(—x) for every x £ R. The cubing function x x 3 : R — > R of course is injective,
and so is any identity function or inclusion function on any set.
93.1.4 Exercise In this problem, A = {1,2, 3, 4} and B = {2,3,4}. For each of
these functions, state whether the function is injective, whether it is surjective,
and
give its image explicitly.
a) F:A-tB, T(F) = {(1,4), (2,4), (3,2), (4,3)} .
b) F : A-* B , T(F) = j (1,3), (2,2), (3,2), (4,3) j .
c) id, 4 .
135
d) The inclusion of B into A.
e) The inclusion of B into Z.
f) C 3 : A — >• B (the constant function).
g) {TRUE, FALSE}.
h) pi : A x B — > A.
i) + : B x B ^ Z.
j) The predicate “n is even” regarded as a characteristic function with domain
A.
(Answer on page 247.)
93.1.5 Exercise Same instructions as for Exercise 93.1.4
a) x 3x — 4 : R — >■ R .
b) x 1 — y x 3 : R — )■ R.
c) F = Xx.(x 2 + 1) : R -» R.
d) ii->2-s 2 :R-^R.
(Answer on page 248.)
93.1.6 Exercise Let F : A—> B be a function of the type indicated. Give a precise
description of all the sets A and B for which F is injective, and a precise
description
of all the sets A and B for which F is surjective.
a) An identity function.
b) An inclusion function.
c) A constant function.
d) An empty function.
e) A coordinate function.
93.1.7 Exercise How do you prove that a function F : A B is not injective?
(Answer on page 248.)
93.1.8 Exercise Prove that the function (m,n) >->• 2 m 3 n — 1 : N x N — >■ N is
injec-
tive.
93.1.9 Exercise Give an example of a function F : R — > R with the property that
F is not injective but TjN is injective.
93.1.10 Exercise (calculus)
a) Show that if a cubic polynomial function x *->• ax 3 + bx 2 + cx + d is not
injec-
tive, then b 2 — 3 ac > 0. (The “3” is not a misprint.)
b) Show that the converse of the statement in (a) is not true.
c) Think of a more sophisticated condition involving a, b, c and d that is true
if and only if the function is injective.
characteristic func-
tion 65
constant function 63
coordinate func-
tion 63
empty function 63
even 5
function 56
identity function 63
inclusion function 63
injective 134
lambda notation 64
predicate 16
surjective 133
136
bijection 136
bijective 136
Cartesian product 52
coordinate func-
tion 63
definition 4
functional relation 75
function 56
graph (of a func-
tion) 61
identity 72
injective 134
one to one correspon-
dence 136
positive real num-
ber 12
relation 73
restriction 137
subset 43
surjective 133
usage 2
94. Bijectivity
94.1 Definition: bijective
A function which is both injective and surjective is bijective.
94.1.1 Remark A bijection F : A — > B matches up the elements of A and B
each element of A corresponds to exactly one element of B and each element of B
corresponds to exactly one element of A.
94.1.2 Usage A bijective function is called a bijection and is said to be a one
to one correspondence.
94.1.3 Example For any set A, id^:A — > A is bijective. Another example is the
function F : {1,2,3} — > {2,3,4} defined by F( 1) = 3, F( 2) = 2, F( 3) = 4.
94.1.4 Exercise Show that the function G:N — >■ Z defined by
is a bijection.
G{n)
— | n even
^±1 n odd
94.1.5 Exercise Show how to construct bijections (3 as follows for any sets A, B
and C.
a) (3: AxB^-BxA.
b) (3: {Ax B)xC -> Ax{BxC).
c) f3 : {1} x A -» A.
94.1.6 Exercise Let a be a relation from A to B.
a) Prove that a is functional if and only if the first coordinate function pf is
injective. (See Section 51.4.)
b) Prove that a is the graph of a function from A to B if and only if the first
coordinate function is bijective.
94.1.7 Exercise Give an example of a function F: R — R ++ for which F is bijec-
tive. (R ++ is the set of positive real numbers.)
94.1.8 Exercise (hard) Give an example of a function F : R — - > R + for which F
is bijective. (R + is the set of nonnegative real numbers.)
94.1.9 Exercise (hard) Let F : A — >■ B be a function. Prove that the restriction
to T{F) of the first coordinate function from Ax B is a bijection.
94.1.10 Exercise (hard) Prove that a subset C of AxB is the graph of a
function from A to B if and only if the restriction to C of the first coordinate
function is a bijection.
137
94.1.11 Exercise (hard) Let /3:Kel(A,B) — >■ (' PB)^ be the function which takes
a relation a to the function a* : A -> VB defined by a* (a) = {6 e B \ aab} (see
Definition 53.2). Show that (3 is a bijection. (This function is studied further in
Problem 100.1.8, page 145, and in Problem 101.5.10, page 150.)
94.1.12 Exercise (hard) Let A, B and C be sets. In this exercise we define a
particular function (3 from the set B ^ x C ^ to the set ( B x (7) , so that (3
as input a pair of functions (f,g) , with / : A—t B and g : A — >• C , and outputs
a
function f3(f,g) from A to B x C . Here is the definition of (3: for all a & A,
/3{f,g)(a) = ( f(a),g(a ))
bijection 136
definition 4
function 56
identity 72
include 43
permutation 137
powerset 46
relation 73
restriction 137
take 57
usage 2
Prove that (3 is a bijection.
95. Permutations
95.1 Definition: permutation
A permutation of a set A is a bijection f3: A—t A.
95.1.1 Example The fact just noted that id/i is a bijection says that id^ is a
(not very interesting) permutation of A for any set A.
95.1.2 Example The function F : {1,2,3} — >■ {1,2,3} that takes 1 to 2, 2 to 1
and 3 to 3 is a permutation of {1,2,3}.
95.1.3 Usage Many books define a permutation to be a list exhibiting a rear-
rangement of the set {1,2, .. . ,n} for some n. If the ith entry in the list is a*
that
indicates that the permutation takes i to a* .
95.1.4 Example The permutation of Example 95.1.2 would be given in the list
notation as (2,1,3) .
95.1.5 Worked Exercise List all the permutations of {1,2, 3, 4} that take 1 to
3 and 2 to 4.
Answer (3,4, 1,2) and (3,4,2, 1),
95.1.6 Exercise List all six permutations of {1,2,3}.
96. Restrictions and extensions
96.1 Definition: restriction
Suppose F :A—> B is a function and A' C A. The restriction of F
to A' is a function denoted F\A! : A' — » B, whose value ( F\A')(a ) for
a € A' is F(a) .
96.1.1 Remark Note that the codomain of the restriction is the codomain of the
function.
138
codomain 56
constant function 63
coordinate 49
definition 4
domain 56
function 56
graph (of a func-
tion) 61
identity 72
inclusion function 63
injective 134
integer 3
lambda notation 64
positive integer 3
predicate 16
restriction 137
subset 43
surjective 133
tuple 50, 139, 140
usage 2
96.1.2 Example Let F : {1,2,3} — > {2, 4, 5, 6} be defined by F(l) = 4 and F(2) =
F( 3) = 5, as before. Then F restricted to {2,3} has graph |(2,5), (3,5)| and
F|{1,3} has graph |(1,4), (3,5)| . Observe that F|{2,3} is a constant function and
-F|{1,3} is injective, whereas F is neither constant nor injective.
96.2 Definition: extension of a function
Let F : A — > B and let C be a set containing A as a subset. Any function
G'.C^rB for which G\ A = F is called an extension of F to C.
96.2.1 Remark Note that both “restriction” and “extension” have to do with the
domain.
96.2.2 Example Let F : {1,2,3} — > {2, 4, 5, 6} be defined by F(l) = 4 and F(2) =
F( 3) = 5, as before. Then F has four extensions F\ , F 2 , T3, and F 4 , to {1,2,
3, 7 },
defined by F\(7) = 2 , F 2 (7) = 4 ,Fs(7) = 5 and F 4 (7) = 6. (Of course in all
cases
Fi(n) is the same as F(n) for n = 1,2,3).
96.2.3 Example The absolute value function ABS :R — > R is an extension of the
inclusion of R + into R, and id^ is a different extension of the same function.
96.2.4 Usage The meaning just given of “extension” is a different usage of the
word from the meaning used in Definition 18.1 of the set of data items for which a
predicate is true.
You may wonder how the word “extension” got two such different meanings.
The answer is that the concept of extension of a predicate was named by logicians,
whereas the concept of extension of a function was named by mathematicians.
96.2.5 Exercise For each of these functions from R to R, state whether the
function is injective or surjective, and state whether its restriction to R + = {r£
R | r > 0} is injective or surjective.
a) xH’i 2 .
b) \x.x-\-l.
c) Ax.l — x.
(Answer on page 248.)
97. Tuples as functions
Let n be a positive integer, and let
n = {1,2,... ,n}
An n-tuple
a — (ai , • • • , a n )
in A n associates to each element i of n an element at of A. This determines a
function i i-y a t with domain n and codomain A. Conversely, any such function
determines an n-tuple in A n by setting its coordinate at i to be its value at i .
139
When a £ A\ x A 2 x • • • x A n , so that different components are in different
sets,
this way of looking at n-tuples is more complicated. Every coordinate ai is an
element of the union C = A\ U A -2 U • • • U A n , so that a can be thought of as a
function from n — >• C. In this case, however, not every such function is a tuple
in
A\ x A 2 x • • • x A n : we must impose the additional requirement that a* E A t .
We sum all this up in an alternative definition of tuple:
97.1 Definition: tuple as function
A tuple in rnuA is a function
a : n -> Ai U A 2 U • • • U A n
with the property that for each i , a(i) E Ai .
Cartesian product 52
coordinate 49
decimal 12, 93
definition 4
digit 93
domain 56
function 56
graph (of a func-
tion) 61
set 25, 32
string 93, 167
tuple 50, 139, 140
union 47
97.1.1 Example The tuple (2,1,3) is the function 1 1 — > 2 , 2 1 — >• 1 , 3 1 — >3
(compare
Section 95.1.3).
97.1.2 Example The tuple (5, 5, 5, 5) is the constant function C§ : {1,2, 3, 4} — >
Z.
97.1.3 Exercise Write the domain and the graph of these tuples regarded as
functions on the index set.
a) (2, 5, -1,3, 6).
b) <7T, 5, 7T — 1, x/2) .
c) ((3, 5), (8, -7), (5, 5)).
(Answer on page 248.)
97.1.4 Example A simple database might have records each of which consists of
the name of a student, the student’s student number, and the number of classes the
student takes. Such a record would be a triple (w, x, n ) , where w is an element
of the
set A* of strings of English letters and spaces (this notation is introduced
formally
in Definitions 109.2 and 110.1), x is an element of the set D* of strings of
decimal
digits, and n E N. This triple corresponds to a function F : {1,2,3} A* x D* x N
with the property that F( 1) E A* , F( 2) € D* and F( 3) E N.
Modeling detabases this way is the principle behind relational database theory.
97.1.5 Remark In the case that all the Ai are the same, so that a E A n , we
now have the situation that A n (the set of functions from n to A, where n =
{1,2,... , n}) and A n (the set of n-tuples in A ) are essentially the same thing.
That is the origin of the notation B ^ .
97.2 Tuples with other index sets
The discussion above suggests that by regarding a tuple as a function set, we can
use any set as index set.
97.2.1 Example In computer science it is often convenient to start a list at 0
instead of at 1 , giving a tuple (ao,ai, . . . ,a n ) . This is then a tuple
indexed by the
set {0,1,... ,n} for some n (so it has n+1 entries!).
140
composite (of func-
tions) 140
composite 10, 140
definition 4
domain 56
family of elements
of 140
field names 140
functional composi-
tion 140
function 56
indexed by 140
infinite 174
integer 3
set 25, 32
tuple 50, 139, 140
97.2.2 Example An infinite sequence of integers is indexed by N + , so it is an
element of .
97.2.3 Example This is another look at Example 97.1.4. The point of view
that a triple (Jones, 1235551212,4) is a function with domain {1,2,3} has an arbi-
trary nature: it doesn’t matter that the name is first, the student number sec-
ond and the number of classes third. What matters is that Jones is the name,
1235551212 is the student number and 4 is the number of classes. Thus it would
be conceptually better to regard the triple as a function whose domain is the set
{Name,StudentNumber,NumberOfClasses}, with the property that /(Name) G A* ,
-F(StudentNumber) £ D* and T/NumberOfClasses) GN. This eliminates the spu-
rious ordering of data imposed by using the set {1,2,3} as domain.
In this context, the elements of a set such as
{Name,StudentNumber,NumberOfClasses}
are called the field names of the database.
97.3 Definition: function as tuple
A function T : S — > A is is also called an A-tuple or a family of ele-
ments of A indexed by S .
97.3.1 Exercise Write each of these functions as tuples.
a) F : {1,2, 3,4, 5} — >■ R, T(E) = { (2,5), (1,5), (3,3), (5,-1), (4, 17)} .
b) F : {1,2, 3,4,5} — > R, F(n) = (n + l)ir .
c) x i — y : {1,2, 3, 4, 5, 6} — y R .
(Answer on page 248.)
98. Functional composition
98.1 Definition: composition of functions
If F : A — > B and G : B — » C , then G ° F : A — > C is the function defined
for all a G A by (G° F) (a) = G(xxF(a )) . G° F is the composite of F
and G, and the operation is called functional composition.
98.1.1 How to think about composition The composite of two functions is
obtained by feeding the output of one into the input of the other. Suppose F: A—> B
and G : B — >■ C are functions. If a is any element of A, then F(a ) is an element
of
B , and so G(F(a)) is an element of C . Thus applying F . then G, gives a function
from A to C , and that is the composite G° F : A — > C .
98.1.2 Remarks
a) You may be familiar with the idea of functional composition in connection
with the chain rule in calculus.
141
b) Our definition of G°F requires that the codomain of F be the domain of G.
Actually, the expression G(F(a)) makes sense even if codF is only included
in dornG, and many authors allow the composite G°F to be formed in that
case, too. We will not follow that practice here.
98.1.3 Example If A = {1,2, 3,4} , B = {3,4, 5, 6}, C = {1,3, 5, 7}, F is defined
by F(l) = F( 3) =5, F( 2) = 3 and F(4) = 6, and G is defined by G(3) = 7, G( 4) =
5 , G(5) = 1 and G(6) = 3 , then G ° F takes 1 1, 2 H > 7 , 3 e- )• 1 and 4 1 — > 3
.
98.1.4 Warning Applying the function G°F to an element of A involves apply-
ing F , then G — in other words, the notation “G°F” is read from right to left.
Functional composition is associative when it is defined:
98.2 Theorem
If F : A — >■ B , G : B — * C and H : C — > D are all functions, then
H°(G°F ) and ( H°G)°F are both defined and
H°(G°F) = (H°G)°F
Proof Let a 6 A. Then by applying Definition 98.1 twice,
(ffo(Gof)J(a) =H((G°F)(a)'j = H(G(F(a )))
and similarly
^(H °G) (a) = (H°G)(F(a)) = H(G(F(a)))
so H°(G°F) = {H°G)°F.
98.2.1 Warning Commutativity is a different story. If F :A—> B and G:B — » C ,
G ° F is defined, but F ° G is not defined unless A = C . If A = C , then G°F:A^C
and F°G :C — )■ A, so normally F°G G° F . Commutativity may fail even when
A = B = C: For example, let S = x >-)• x 2 : R — > R and T = ii->a: + l:R->R. Then
for any ilR, S{T(x )) = (x + l) 2 and T(S(x )) =x 2 + l,so S°T / T° S .
Pondering the following examples of functional composition may be helpful in
understanding the idea of composition.
98.2.2 Example Let SQ = x i->- x 2 : R — > R + and SQRT = i4 y Ax : R + — >• R.
(yT denotes the nonnegative square root of x.) Let ABS denote the absolute
value function from R to R. Then the following are true.
(i) SQRT°SQ = ABS:R-> R.
(ii) (SQRT°SQ)|R+ = id R+ .
(iii) SQ°SQRT = id R + .
98.2.3 Example If F : A—} B is any function, then
(i) F ° id^ = F and
(ii) id b ° F = F.
This is analogous to the property that an identity element for a binary operation
has (see 50.1), but in fact composition of functions is not a binary operation
since
it is not defined for all pairs of functions.
associative 70
binary operation 67
codomain 56
commutative 71
composite (of func-
tions) 140
composition (of
functions) 140
domain 56
function 56
identity 72
include 43
proof 4
take 57
theorem 2
142
bijective 136
codomain 56
composition (of
functions) 140
domain 56
function 56
graph (of a func-
tion) 61
include 43
inclusion function
injective 134
surjective 133
98.2.4 Example If A C B and BCC, and i:A^-B and j : B — >■ C are the
corresponding inclusion functions, then j°i is the inclusion of A into C .
98.2.5 Example If F : A B and CCA with inclusion map i:C—>A, then
F\C = F°i. In other words, restriction is composition with inclusion.
98.2.6 Exercise Describe explicitly (give the domain and codomain and either
the graph or a formula) the composite G°F if
a) F : {1,2, 3, 4} — »• {3, 4, 5, 6}, with 1 ^ 3, 2 i— >4, 3e-)-6, and 4 i — >- 5 ,
and G :
{3, 4, 5, 6} —>■ {1,3, 5, 7, 9} with 3^1, 4^7, 5^7 and 6^3.
b) F : x !->• x 3 : R — > R, G : x e-x 2x : R — > R.
c) F : x e-x 2x : R — > R. G : x e-t x 3 : R -> R,
d) F = inclusion : N — >■ R , G : x (x/2) : R — > R.
e) F = p\ : R X R — >• R, G:rH’(3,i):R-tRxR.
(Answer on page 248.)
98.2.7 Exercise Let F : A—t B , G : B —$■ C . Show the following facts:
a) If F and G are both injective, so is G°F .
b) If F and G are both surjective, so is G°F .
c) If F and G are both bijective, so is G° F .
d) If G°F is surjective, so is G.
e) If G°F is injective, so is F .
98.2.8 Exercise Give examples of functions F and G for which G°F is defined
and
a) F is injective but G°F is not.
b) G is surjective but G°F is not.
c) G°F is injective but G is not.
d) G°F is surjective but F is not.
98.2.9 Exercise (hard) Let A, B and C be sets.
a) Prove that if F : A — > B is a function and C is nonempty, then G i— >• F°G :
C 1 C 1
is a function which is injective if and only if F is injective, and
surjective if and only if F is surjective.
b) Prove that if H : B — >■ C is a function and A has more than one element, then
TD
G e - > (G°H) : 4 U — » A D is a function which is injective if and only if H is
surjective, and surjective if and only if H is injective.
143
99. Idempotent functions
99.1 Definition: Idempotent function
A function F : A — >■ A is idempotent if F° F = F .
99.1.1 How to think about idempotent functions F is idempotent if doing
F twice is the same as doing it once: If you do F , then do it again, the second
time
nothing happens.
99.1.2 Example The function (i,|/) 4 (i,0):RxR->RxR is idempotent.
Note the close connection between this function and the first coordinate function
p\ : R x R y R .
Cartesian product 52
coordinate func-
tion 63
definition 4
fixed point 143
function 56
idempotent 143
identity 72
image 131
injective 134
surjective 133
theorem 2
usage 2
99.1.3 Example Let S be a set of files that contains a sorted version of every
file in the set. Then “sort” is a function that takes each file in the set to a
possibly
different file. Sorting a file that is already sorted does not change it (that is
true
of many sorting functions found on computers, but not all). Thus sorting and then
sorting again is the same as sorting once, so sorting is idempotent.
99.1.4 Usage Following Example 99.1.2, the word “projection” is used in some
branches of mathematics to mean “idempotent function”. In other brances, “pro-
jection” means “coordinate function” .)
99.1.5 Exercise Let A = {1,2,3}. Give an example of an idempotent function
F : A — > A that is not id 4 . (Answer on page 248.)
99.1.6 Exercise Show that if F : A — > A is injective and idempotent, then F =
id .4 •
99.2 Definition: Fixed point
Let F : A — > A be any function. An element x £ A is a fixed point of
F if F{x) = x.
This is the fundamental theorem on idempotent functions:
99.3 Theorem
A function F : A — > A is idempotent if and only if every element of ImF
is a fixed point of F .
99.3.1 Exercise Prove Theorem 99.3.
99.3.2 Exercise Use Theorem 99.3 to show that if F : A — > A is surjective and
idempotent, then F = id^.
144
codomain 56
commutative
gram 144
definition 4
domain 56
function 56
identity 72
dia-
100. Commutative diagrams
F : A — >■ B and G : B — >• C can be illustrated by this diagram:
(100.1)
If the two ways of evaluating functions along paths from A to C in this diagram
give the same result, then, by definition of composition, H = G°F .
100.1 Definition: commutative diagram
A diagram with the property that any two paths between the same
two nodes compose to give the same function is called a commutative
diagram.
100.1.1 Example To say that the following diagram commutes is to say that
H°F = K°G\ in other words, that for all a € A, H ^F(a)^ = K ^G(a)^ .
F
*■ B
(100.2)
100.1.2 Remarks
a) Commutative diagrams exhibit more of the data involved in a statement such
as U H°F = K°G ” than the statement itself shows (in particular it shows
what the domains and codomains are), and moreover they exhibit it in a
geometric way which is easily grasped.
b) Warning: The concept of commutativity of diagrams and the idea of the
commutative law for operations such as addition are distinct and not very
closely related ideas, in spite of their similar names.
100.1.3 Example Example 98.2.3 on page 141 says that for any function F , this
diagram commutes:
(100.3)
145
100.1.4 Example Theorem 98.2 says that if both triangles in this diagram com-
mute,
■>■ B
(100.4)
then the whole diagram commutes. Thus the associative law for for functional
composition becomes a statement that commutative triangles can be pasted together
in a certain way.
100.1.5 Exercise Draw commutative diagrams expressing these facts:
a) The square of the square root of a nonnegative real number is the number
itself.
b) The positive square root of the square of a real number is the absolute value
of the number.
(Answer on page 248.)
associative 70
commutative dia-
gram 144
composition (of
functions) 140
equivalent 40
function 56
idempotent 143
identity 72
positive real num-
ber 12
powerset 46
real number 12
relation 73
take 57
100.1.6 Exercise Draw commutative diagrams which express each of the follow-
ing facts. No one arrow should be labeled with a composite of two functions -
draw a separate arrow for each function.
a) Addition, as a binary operation on Z, is commutative.
b) Addition as in (a) is associative.
100.1.7 Exercise Prove that if F : A — > A is an idempotent function, then there
is a set B and functions G : A — > B and H : B — > A such that both the following
diagrams commute:
100.1.8 Exercise (hard) Let /? be defined as in Problem 94.1.11, page 137. Let
F : A' A and G : B' -+ B . Let
H:Rel{A,B)-^Rel{A , ,B')
be the function which takes a to the relation a! defined by a' a' b' 44 F(a') a
G(b') .
Let /
K : (VB) A — >• (VB') A
be the function which takes r : A — >■ VB to the function r' : A! -^r VB' defined
by
r'(a') = G~ 1 (r(F(a))
146
commutative 71
composition (of
functions) 140
definition 4
fact 1
function 56
identity 72
inverse function 146
invertible 146
left inverse 146
powerset 46
right inverse 146
usage 2
Show that the following diagram commutes.
Rel(A,B)
Re\(A',B')
(VB)
(VB
'i\A!
101. Inverses of functions
(100.5)
The number 1/2 is the “multiplicative inverse” of the number 2 because their prod-
uct is 1 . A similar relationship can hold between functions, but because
functional
composition is not normally commutative, one has to specify which way the com-
posite is taken.
101.1 Definition: inverse of a function
If F : A — >• B and G : B — »• A, then G is a left inverse to F , and F is
a right inverse to G. if
G°F = id a (101.1)
If G is both a left and a right inverse to F, in other words if both
Equation (101.1) and
FoG = id B (101.2)
hold, then G is an inverse to F.
101.1.1 Usage A function that has an inverse is said to be invertible.
101.1.2 Fact It follows from the definition that if G is an inverse to F , then F
is an inverse to G.
101.1.3 Fact The definition of inverse function can be expressed in other ways
equivalent to Definition 101.1.
a) G is the inverse of F if and only if for all a £ A, G(F(a )) = a and for all
b G B , F(G(b)) = b. ( Both equations must hold.)
b) G is the inverse of F if and only if the following diagrams commute:
(101.3)
147
101.1.4 Example Let F : {1,3,5} — > {2,3,4} be the function that takes 1 to 3,
3 to 4 and 5 to 2. Then the function G: {2,3,4} — > {1,3,5} that takes 2 to 5, 3
to 1 and 4 to 3 is the inverse of F . (And F is the inverse of G.)
101.1.5 Example Example 98.2.2(3) above says that the squaring function is a
left inverse to the square root function: squaring the positive square root gives
you
what you started with. It is not the inverse, however: taking the square root of
the
square won’t give you the number you started with if it is negative. On the other
hand, the cubing function is the inverse of the cube root function.
A function can have more than one left inverse (Problem 101.2.6) but not more than
one inverse:
floor 86
function 56
graph (of a func-
tion) 61
identity 72
inverse function 146
positive real num-
ber 12
proof 4
rule of inference 24
take 57
theorem 2
usage 2
101.2 Theorem: Uniqueness Theorem for Inverses
If F : A—> B has an inverse G : B — >• A, then G is the only inverse to
F.
Proof The proof uses the definition of inverse, Theorem 98.2 and Example 98.2.3:
If H : B — >■ A is another inverse of F , then
H = Hoid B = H°(F°G) = {H°F)°G = id A °G = G
101.2.1 Usage The fact that if a function has an inverse, it has only one, means
that we can give the inverse a name: The inverse of F, if it exists, is denoted F^
1 .
101.2.2 Remark The uniqueness theorem also means we have a rule of inference:
Given F : A — > B and G : B — > A ,
G°F = id A , FoG = id B b G = F~ 1 (101.4)
101.2.3 Exercise Which of the following functions have inverses? If it has one,
give the inverse (by describing its graph or by a formula).
a) F : {1,2, 3, 4} — > {3, 4, 5, 6} , with 1 1 — y 3 , 2 i — >- 4 , 3 >-)• 6, and 4
1 — 5 .
b) G : {1,2, 3, 4} — > {3, 4, 5, 6, 7}, with 1 1 — >- 3 , 2 i->- 4, 3 i— >6, and 4
i — >■ 5 .
c) H : {1,2, 3,4} — >■ {3, 4, 5, 6} with 1 1 — 3 , 2 i — >• 5, 3 i-4 6, and 4 i — >
5.
d) m->2n:N->N.
e) n i-4 n+ 1 : N — » N.
f) 1 : Z Z.
g) n !->• (1/n) : N — {0} — > R.
h) K : {1,2, 3,4,5} — >■ {1,2,3} with K(n) = floor((n+ l)/2) .
(Answer on page 248.)
101.2.4 Exercise Which of the functions in Exercise 101.2.3 have (a) left inverses,
(b) right inverses? (Answer on page 248.)
101.2.5 Exercise Show that if a function G has an inverse F , then it has only
one left inverse and that is F . (Answer on page 248.)
148
composition (of
functions) 140
function 56
identity 72
inclusion function 63
infinite 174
inverse function 146
proof 4
theorem 2
101.2.6 Exercise Let I :R + — »• R denote the inclusion function. Show that I has
infinitely many left inverses.
The inverse of the composite of two functions which have inverses is the composite
of the inverses, only in the reverse order:
101.3 Theorem: The Shoe-Sock Theorem
If F : A—t B and G: B — > C both have inverses, then
(G°F)- 1 = F- 1 °G~ 1
101.3.1 Remark The name comes from the fact that the inverse of putting on
your socks and then putting on your shoes is taking off your shoes and then taking
off your socks.
Proof To prove the Shoe-Sock Theorem, we will prove that
(F- 1 °G~ 1 )o(G°F) =id A
and
(G°F)°(F~ 1 °G~ 1 ) = idc
(101.5)
(101.6)
and then apply Rule (101.4), page 147. To prove Equation (101.5), we note that
the following diagram commutes: the left and right triangles are the diagrams in
Figure (101.3), and the middle triangle is the left triangle in Figure (100.3).
(101.7)
Equation (101.6) is proved similarly.
Another fact with a similar proof (left as an exercise) is:
101.4 Theorem
If F has an inverse, then (F -1 ) -1 = F.
101.4.1 Remark Another way of saying this is that a function is the inverse of
its own inverse.
101.4.2 Exercise Prove Theorem 101.4.
A final fact about inverses is the very important:
149
101.5 Theorem: Characterization of invertible functions
A function F : A — >■ B has an inverse if and only if it is a bijection.
101.5.1 Remark The importance of Theorem 101.5 lies in the fact that having
an inverse is defined in terms of functional composition but being a bijection is
defined in terms of application of the function to an element of its domain. Any
time a mathematical fact connects two such differently-described ideas it is
probably
useful.
Proof I will go through the proof in some detail since it ties together several
of the ideas of this chapter. We have to prove an equivalence, which means two
implications.
First we show that if F has an inverse then it is a bijection. Suppose F has
an inverse. We must show that it is injective and surjective. To show that it is
injective, suppose F(a) = F(a'). Then
a = F~ 1 (F(a)) = F~ 1 (F(a')) = a
(The first and last equations follow from 101.1.3a and the middle equation from the
substitution property, Theorem 39.6.) So F is injective.
To show F is surjective, let b £ B . We must find an element a 6 A for which
F(a) = b. The element is F _1 (6), since F(F~ 1 (b)) = b. Thus F is bijective.
Now we must show that if F is bijective, then it has an inverse. Suppose F
is bijective. We must define a function G : B — > A which is the inverse of F . Let
b 6 B . Then, since F is surjective, there is an element a £ A for which F(a) = b.
Since F is injective there is exactly one such a. Let G(b) = a. Since F(a) = b,
that says that G(F(a)) = a, which is half of what we need to show to prove (using
Definition 101.1) that G = F _1 . The other thing needed is that F(G(b)) = b. But
by definition of G, G(b) is the element a for which F(a) = b, so F(G(b)) = b. That
finishes the proof.
101.5.2 Remarks
a) The second part of the proof says this: If F(a) = b , then F _1 (6) = a, and if
F~ l (b) = a, then F(a) = b.
b) You might experiment with proving the contrapositives of the two implications
in the preceding proof; some people find them easier to understand.
101.5.3 Exercise Write a formula for the inverse of each of these bijections.
a) in’! 2 : R + R + .
b) iH’S-LRH’R.
c) x i — y 2x : R — y R.
d) x i — y (1/x) : R — y R .
(Answer on page 248.)
101.5.4 Exercise Prove that a function has a left inverse if and only if it is
injec-
tive.
101.5.5 Exercise Prove that a function has a right inverse if and only if it is
surjective.
bijection 136
bijective 136
contrapositive 42
domain 56
equivalence 40
function 56
implication 35, 36
injective 134
inverse function 146
proof 4
surjective 133
theorem 2
150
Cartesian product 52
definition 4
dummy variable 150
expression 16
function 56
GCD 88
injective 134
integer 3
inverse function 146
left cancellable 150
powerset 46
relation 73
surjective 133
take 57
101.5.6 Exercise Give a right inverse of the function GCD :N + x N + — )• N + .
(You
are being asked to give the right inverse explicitly, not merely show it exists.)
101.5.7 Exercise Show that GCD:N + x N + — > N + does not have a left inverse.
101.5.8 Exercise (hard) A function F:A-^B is left cancellable if whenever
G : D — >■ A and H : D — >• A are functions for which F °G = F° H , then G must be
the same as H . Right cancellable is defined analogously. Prove that a function
is left cancellable if and only if it is injective and right cancellable if and
only if it
is surjective.
101.5.9 Exercise (hard) Let F:A—>B be a function and suppose A has more
than one element. Show that if F has exactly one left inverse then the left inverse
is also a right inverse (hence F has an inverse).
101.5.10 Exercise (hard) Let A and B be sets. Let /3 : Rel(A,R) — > (VB)^
defined in Problem 94.1.11, page 137. Let 7 : ( VB )^ — >• Rel {A,B) be the
function
(defined in Definition 53.3, page 76) that takes a function F : A — > VB to the
relation
ap defined by
aa,pb if and only if b G F(a)
Prove that 7 is the inverse of (3 (hence j3 is the inverse of 7 ).
102. Notation for sums and products
In this section we introduce a notation for sums and products that may be familiar
to you from calculus. This will be used in studying induction in Chapter 103.
102.1 Definition: sum and product of a sequence
Let 01 , 02 , . . . ,a n be a sequence of numbers (not necessarily integers).
The expression Y17= 1 a * d en °tes the sum ai + 02 H I -a n of the num-
bers in the sequence and the expression nf=i a * denotes the product
ai 02 • • • a n of the numbers in the sequence.
102.1.1 Example £* =1 * = 1 + 2 + 3 + 4 = 10 and 1}*=!* = 1 * 2 x 3 x 4 = 24.
102.1.2 Example 2A: — 1 = 1 + 3 + 5 + 7 + 9 = 25 . The sum 2i — 1 also
gives 25 — the i is a dummy variable just like the x in J 3 5 x 2 dx , which has
the
same value as /!? t 2 dt . On the other hand, ]C 5 =I 2fc — l = 10fc — 5.
102.1.3 Exercise What is J^ = 1 fc 2 ? What is f[fc=i (Answer on page 248.)
102.1.4 Example For b any fixed number, y )- 1 b = b+b+b+b = 4 b and
Hl =1 b = b-b-b-b = b i .
102.1.5 Remark The numbering of the sequence does not have to start at 1.
Thus a sequence 03 , 04 ,... ,012 would have sum X^= 3 a *-
151
102.1.6 Exercise What are Y%= i 2 ^ Ylt= o 2 (& + l) an d J2k= o 2 ^+l? Two of
direct method 119
them are the same. Explain why. hypothesis 36
implication 35, 36
102.1.7 Sums and products in Mathematica To compute infinite 174
integer 3
odd 5
positive integer 3
range expression 151
in Mathematica, you type the expression Sum [2 k-1 , {k, 1 ,5}] . Similarly, the
rule of inference 24
expression Product [k , {k , 1 , 6}] calculates 1-2-3-4-5-6.
The expression {k,l,5} is a range expression; range expressions are used
in many Mathematica commands. The range expression {x,a,b}- means that the
variable x ranges from the value of a to the value of b.
E 2 *- 1
k = 1
103. Mathematical induction
The positive integers contain some fascinating patterns. For example,
1 = 1, 1+3+5+7 = 16,
1+3 = 4, 1+3+5+7+9 = 25,
1+3+5 = 9, 1+3+5+7+9+11 = 36,
In general it appears that the sum of the first n odd positive integers is n 2 .
This
is a statement Q(n) about an infinite number of positive integers n.
The subject of this section is an inference rule allowing the proof of such state-
ments. Before the rule is stated, we will reformulate Q{n) and see why it is true.
Using summation notation, Q(n) is the statement
J2( 2k ~l ) = n 2
k = 1
(You should check that 2k — l is indeed the fcth odd positive integer.) Clearly
Q( 1)
is true: it says 1 = 1.
I will now prove that for any positive integer n, Q(n) =+ Q(n + 1), using the
direct method. The direct method requires us to assume the hypothesis is true, so
suppose we knew that Q(n) is true, that is that the sum of the first n odd integers
is n 2 . Then the sum of the first n + 1 odd integers is
(the sum of the first n odd integers) + (the n+lst odd integer)
We know the left term is n 2 because we are assuming Q(n) , and the right term is
2n + 1 . Hence the sum of the first n + 1 odd integers is n 2 + 2n + 1 . But n 2 +
2 n +
1 = (n+ l) 2 , in other words the sum of the first n + 1 odd integers is (n + l)
2 , so
that Q(n + 1) is true. This proves that Q(n) =+ Q(n + 1).
Now we know these two things:
a) Q(l).
b) For any n, Q(n) +> Q(n + 1).
152
basis step 152
contrapositive
method 120
direct method 119
divide 4
implication 35, 36
induction hypothe-
sis 152
induction step 152
induction 152
inductive proof 152
integer 3
negative integer 3
nonnegative integer 3
positive integer 3
rule of inference 24
theorem 2
usage 2
Using these facts, you should be able to convince yourself that Q(n) is true for
any
positive integer, since Q( 1) is true, and the implication Q(n) => Q(n + 1) allows
you to see that Q( 2) is true, Q(3) is true, . . . , jacking the proof up, so to
speak, until
you get to any positive integer. You need to know both Q(l) and the implication
Q(n) => Q{n + 1) for all ji to make this work.
This approach is the basis for the following rule of inference:
103.2 Theorem: The principle of mathematical induction
For any statement about the positive integers, this rule of inference is
valid:
P( 1), (Vn:N + ) (P(n) => P(n + 1)) b (Vn:N+)P(n)
103.2.1 Usage A proof using the principle of mathematical induction is called
an inductive proof. The proof that P(l) is true is the basis step and that
P(n) =>- P(n+1) is the induction step.
103.2.2 Remarks
a) The induction step is sometimes stated as P(n — 1) => P(n) , which must hold
for all integers > 1 , but that is only a change in notation.
b) The proof of the induction step, which is an implication, may be carried out
by the direct method as was done above, or by the contrapositive method.
If it is carried out by the direct method, one assumes that P(n ) is true and
deduces P(n + 1). In doing this, P(n) is called the induction hypothesis.
c) The principle of mathematical induction gives you a scheme for proving a
statement about all positive integers. You still have to be clever somewhere in
the proof. In the example just given, algebraic cleverness was required in the
induction step.
103.3 Other starting points for proofs by induction
We have formulated mathematical induction as a scheme for proving a statement
about all positive integers. One can similarly prove statements about all
nonnegative
integers by starting the induction at 0 instead of at 1 (see Example 103.3.1
below).
In that case you must prove P(0) and
(Vn:N) (P(n) =► P(n + 1))
Indeed, a proof by mathematical induction can be started at any integer, positive
or negative. For example, if you prove P(— 47) and P(n) => P(n + 1) for n > —47,
then P(n) is true for all n > —47.
One could also go down instead of up, but we won’t do that in this text.
103.3.1 Example Let’s prove that for all nonnegative integers n, 3 | n 3 + 2n.
Basis step: We must show 3 | 0 3 + 0, which is obvious.
153
Induction step: assume 3|n 3 + 2n. (This is the induction hypothesis.) Then
n 3 + 2n = 3k for some integer k . Then
(n+ l) 3 + 2(n + 1) = n 3 + 3n 2 + 3n + 1 + 2n + 2
= n 3 + 2n + 3n 2 + 3n + 3 (103.1)
= 3(fc + n 2 + n + 1)
so is divisible by 3 as required.
103.3.2 Remark The statement 3 | n 3 + 2 n is true of negative integers, too. Once
you know it for positive integers, the proof for negative integers is easy:
substitute
— n for n in the statement and do a little algebra. This trick often works for
proving
things about all integers. However, the principle of mathematical induction by
itself
is not a valid method of proof for proving statements about all integers.
103.3.3 Example A statement about the value of a sum or product can often be
proved by induction. Let us prove that
n 1
^2 k = 2 n ( n+1 )
k = 1
Proof Basis step: A; = 1 = | • 1 • 2 , as required.
Induction step:
Efci i k = ra+l + EL i k
= n + 1 + \n{n + 1) (by the induction hypothesis)
= {\n + l)(n + 1)
= ^(n + 1) (n + 2) (by algebra)
This proof uses a basic trick: separate out the term in the sum (or product) of
highest index, in this case n + 1 . Then the rest of the sum can be evaluated using
the induction hypothesis.
103.3.4 Remark In all proofs by induction you should label the basis step, the
induction step and the induction hypothesis. If you find yourself writing “and so
on...” or “continuing in this way...” or anything like that, you are not doing an
inductive proof.
103.4 Exercise set
Prove the statements in Exercises 103.4.1 through 103.4.8 by induction.
103.4.1 ££ =1 = n ” r (Answer on page 248.)
103.4.2
U ( —
£(-i)‘* = l T±i)
k= 1 2
(Answer on page 249.)
103.4.3 J2k= i k i k + 1) = \n{n + l)(n + 2).
basis step 152
divide 4
even 5
induction hypothe-
sis 152
induction step 152
induction 152
integer 3
negative integer 3
odd 5
positive integer 3
proof 4
(n even)
(n odd)
154
counterexample 112
even 5
induction 152
integer 3
nonnegative integer 3
positive integer 3
theorem 2
universal quanti-
fier 112
usage 2
well-ordered 154
103.4.4 ELi^ = |n(n+l)(2ra + l).
103.4.5 ELi 2 k = 2 n+1 - 2.
103.4.6 ELi k2k = (« - l)2 n+1 + 2.
103.4.7 ELi^ 3 = •
103.4.8 ELiC- 1 )^ 2 = ( ^Y~ri{n+ 1).
103.4.9 Exercise Prove the following inequalities by induction.
a) 2 n > 2n + 1 for n > 3.
b) 2 n > n 2 for n > 4.
104. Least counterexamples
Proof by induction as described in Chapter 103 is based on a very basic fact about
the positive integers that has wider applications. Suppose P(n) is a statement
about positive integers, and suppose the statement (Vn:N + )P(?r) is false. Then
there is a counterexample m, a positive integer for which P(m) is false. Among all
such counterexamples, there is a smallest one:
104.1 Theorem: The Principle of the Least
Counterexample
Every false statement of the form (Vn:N + )P(?r) about the positive inte-
gers has a smallest counterexample.
104.1.1 Usage This property of the positive integers is often referred to by saying
the positive integers are well-ordered.
104.1.2 Remark Of course, one can replace the positive integers by the nonneg-
ative integers, or by the set of all integers greater than a particular one, in the
statement of Theorem 104.1.
The existence of least counterexamples is characteristic of such sets; for most
other data types, least counterexamples need not exist. For example, the statement,
“All integers are even” is a false universally quantified statement about the
integers
which has many counterexamples, but no smallest one.
155
104.2 Least counterexample and induction
The principle of mathematical induction, in other words Theorem 103.2, can be
proved using the principle of the least counterexample. The proof is by contradic-
tion.
Suppose that the hypotheses of the theorem are true: P(l) , and
(Vn:N + )^P(n) =► P(n+ 1)
Suppose that (Vn:N + )P(n) is false. Then there is a least counterexample m, so
P(k) is true if k < m but P(m) is false. Now we have two cases.
(i) m = 1. Then P( 1) is false — but this contradicts one of the hypotheses of
the theorem.
(ii) m > 1 . In this case, P{m ) is false, since m is a counterexample to the
statement (Vn:N + )P(n). Since m is the least counterexample, the statement
P(m — 1) is true. It follows from the truth table for implication that the
statement P(m— 1) => P(m) is false. But that means the hypothesis
counterexample 112
equivalent 40
Fundamental Theo-
rem of Arith-
metic 87
implication 35, 36
induction hypothe-
sis 152
induction 152
integer 3
least counterexam-
ple 154
proof by contradic-
tion 126
strong induction 155
(Vn:N + ) f P(n) => P(n + 1)
is false since n = m — 1 provides a counterexample.
So in either case, one of the hypotheses of Theorem 103.2 must be false. There-
fore there can be no least counterexample, so by Theorem 104.1 there can be no
counterexample. Hence (Vn:N + )P(n) is true.
The principle of mathematical induction and the principle of the least counterex-
ample are actually equivalent.
104.2.1 Exercise Use the principle of mathematical induction (Theorem 103.2)
to prove Theorem 104.1.
104.3 Strong induction
The principle of the least counterexample is useful in its own right for proving
things.
For example, it is used in Problems 104.3.3 and 104.4.4 to prove the Fundamental
Theorem of Arithmetic.
The principle allows you to assume as a kind of induction hypothesis that P(k)
is true for all k <n, not just for n — 1. This is stated formally in Exercise
104.3.1
below. It is handy for proving things about factoring integers, since the prime
factorization of an integer n has little to do with the factorization of n — 1 .
This
more general approach is often called strong induction, and another statement of
it is in Problem 104.3.1.
In this book, proofs using this technique are usually presented as direct appli-
cations of the least counterexample principle.
156
divide 4
finite 173
Fundamental Theo-
rem of Arith-
metic 87
GCD 88
implication 35, 36
integer 3
least counterexam-
ple 154
nonnegative integer 3
positive integer 3
prime 10
Principle of Strong
Induction 156
proof by contradic-
tion 126
quotient (of inte-
gers) 83
remainder 83
rule of inference 24
104.3.1 Exercise Use the principle of the least counterexample to prove the fol-
lowing rule of inference for positive integers n. This rule is called the Principle
of
Strong Induction.
(Vn:N + ) ^(Vm:N + ) ^ m < n =>• P(m)^ =4- P(n)^ (— (Vn:N + )P(n)
104.3.2 Example: Existence of quotient and remainder We will use the
Principle of the Least Counterexample to prove the existence half of Theorem 60.2,
page 84. That is, we will prove that for given integers m and n with n / 0, there
are integers q and r satisfying
Q.l m = qn + r, and
Q.2 0 < r < \n\ .
That there is only one such pair of integers was proved on page 84.
We will give the proof for m > 0 and n > 0 and leave the other cases to you
(Exercise 104.3.4). Let S be the set of all nonnegative integers of the form m —
xn.
S is nonempty (any negative x makes m — xn positive, but there may also be
positive x that do so). Let m — qn be the smallest element of S. Let r = m — qn.
Then qn + r = qn + m — qn = m, so Q.l is true. Since m — qn G N by assumption,
we know that r, which is m — qn, is nonnegative, which is half of Q.2. As for
the other half, suppose for the purpose of proof by contradiction that r > n. Then
m — qn>n, that is, m > n + qn = (q + l)n. But then m — (q + l)n is nonnegative,
and it is certainly smaller than m — qn, contradicting our choice of m — qn as the
least element of S .
104.3.3 Exercise (hard) Show that if m is any integer greater than 1 , then there
is a finite list of primes pi,... ,Pk and integers ei, . . . ,e& for which m = n!
°=iPi’ • Use
the principle of the least counterexample. Do not use the Fundamental Theorem of
Arithmetic. Note that if m is prime, then this holds for k = 1, p\ = m, and e\ = 1.
104.3.4 Exercise Complete the proof that the quotient (of integers) and remain-
der exist (see 104.3.2).
104.4 Proof of the Fundamental Theorem of Arithmetic
Exercises 104.4.1 through 104.4.4, together with Exercise 104.3.3, lead up to a
proof of the Fundamental Theorem of Arithmetic. Thus the Fundamental Theorem
should not be used in the proofs of those problems.
104.4.1 Exercise Show that if p is a prime and m an integer not divisible by p,
then GCD(p,7b) = 1. (Answer on page 249.)
104.4.2 Exercise Show that if p is a prime and m and n are integers for which
p | mn but p does not divide m, then p\n. (Hint: Use Problem 104.4.1 and Bezout’s
Lemma, page 128.) (Answer on page 249.)
Suppose p is prime, p\mn, but p does not divide m. Then GCD (p,m) = 1,
so there are integers a and b for which ap + bm = 1 . There is also an integer k
for which mn = pk . Putting these facts together, n = anp + bmn = anp + bkp =
( an + bk)p , so n is divisible by p.
157
104.4.3 Exercise Use Problem 104.4.2 to show that if p is a prime and mi , . . . ,
to*,
are positive integers for which p | JXf=i m i , then for some i , p | rn % .
104.4.4 Exercise (hard) Show that if p\ < p 2 < ■ ■ ■ < Pk in the prime factor-
ization m = YIa-iPT in Exercise 104.3.3, then the factorization is unique. (Hint:
Assume m is the least positive integer which has two such factorizations and use
Problem 104.4.3 to obtain a prime which occurs in both factorizations. Then divide
by that prime to obtain a smaller integer with two factorizations.)
divide 4
function 56
integer 3
iterative 157
positive integer 3
prime 10
recursive 157
105. Recursive definition of functions
Many functions are defined in such a way that the value at one input is defined in
terms of other values of the function. Such a definition is called recursive.
105.1.1 Example One way of defining the function F : N — > N for which F(n) =
2 n would be to say
{ F(n + 1) = 2 • F(n) (105.1)
for all n € N .
Programs 105.1 and 105.2 give Pascal functions which calculate 2 n .
FUNCTION TWOREC (N: INTEGER) : INTEGER;
BEGIN
IF N=0 THEN TWOREC := 1
ELSE TWOREC := 2*TW0REC (N-l)
END;
Program 105.1: Program for 2 n
Program 105.1 simply copies Definition 105.1. Since the function TWOREC calls
itself during its execution, this program is also said to be recursive. Program
105.2
is a translation of Definition 105.1 which avoids the function calling itself.
Since it
is implemented by a loop it is called an iterative implementation of the function.
105.1.2 Remark Many common algorithms are easily to define recursively, so the
study of recursively-defined functions and how to implement them is a major part of
computer science. Very often, the iterative implementation like Program 105.2 is to
be preferred to the recursive one, but in complicated situations it is not always
easy
to transform the recursive definition into an iterative one. In some applications,
for
example in writing programs to parse expressions, the recursively written program
may be the preferred method for writing the first attempt, since the iterative
version
can be much harder to understand and debug.
158
divide 4
factorial function 158
function 56
inductive defini-
tion 159
integer 3
FUNCTION TW0IT(N : INTEGER) : INTEGER; VAR COUNT: INTEGER;
BEGIN
COUNT := 0; POWER := 1;
(♦POWER is a integer variable declared in
the program containing this procedure.*)
WHILE COUNT <N DO
BEGIN
POWER := 2*P0WER;
COUNT := COUNT+1
END
TWO IT := POWER
END;
Program 105.2: Another program for 2 n
105.1.3 Exercise Find the values for n = 1 through 5 of the functions defined as
follows:
a) F(0) = — 3, F(n + 1) = (n + l)F(n)
b) F(l) = 1 , F(n) = n 2 + F(n - 1)
. f 0 if 3 1 n
c) F(n ) = <
( 1 + F(n + 1) otherwise
d) F(0) = 1, F(l) = 3, F{n) = F(n — 1) + F(n — 2)
e) F(1)=0, F( 2) = 1, F{n) = (n — 1) ^F(n — 1) + F(n — 2)^
(Answer on page 249.)
105.1.4 Example For a fixed sequence {dfc} fcgN + ,
F(n) = ^a k
k = 1
is a function from N + to N + which has a natural definition by induction:
f Sfc=l a k — a \
l Sfcil a k = a n+ 1 + ^2k= 1 a k
(105.2)
105.1.5 Example The product has a similar definition:
f n fe =i a k =
i re;i«»=on + i'(nhi«b
(105.3)
105.1.6 The factorial function A particularly important function which can be
defined by induction is the factorial function. Its value at n is denoted n! and it
is defined this way:
I 0 ! = 1
\ (n + 1)! = (n+ 1) • n\
(105.4)
159
Thus for n > 0 , n\ = H£ =1 k ; you can prove this by induction because both n\ and
the product are defined by induction (Exercise 105.2.1). The factorial function
will
be used in various combinatorial formulas in later sections.
105.2 Proofs involving inductively defined functions
Defining a function by induction makes it convenient to prove things about it by
induction. For example, let us use induction to prove that n\ > 2" for n > 3 . We
start the induction at n = 4. Then 4! = 24 and 2 4 = 16, so the statement is true
for n = 4. For the induction step, suppose n ! > 2 n and n > 4. It is necessary to
prove that (n + 1)! > 2 n+1 . Both these functions are defined by induction, so we
can apply their definitions and the induction hypothesis to get
(n + 1)! = (n + 1) • n\ > 2 • n\ > 2 • T = 2 n+1
as required.
defined by induc-
tion 159
domain 56
function 56
induction hypothe-
sis 152
induction step 152
induction 152
inductive defini-
tion 159
integer 3
ninety-one func-
tion 159
positive integer 3
recursive defini-
tion 157
105.2.1 Exercise Prove directly from the inductive definition of n\ that n ! =
nLi ^ f° r all positive integers n. (Answer on page 249.)
105.2.2 Exercise Prove that for all integers n > 0, 2 n < 2(n!) .
105.2.3 Exercise Find constants C and D for which for all integers n > 0 , 3 n <
C(nl) and 4 n < D(nl). Prove your answers are correct.
106. Inductive and recursive
Definition 105.1 gives the value at n in terms of the value of the function at a
smaller
integer. In general, a function F : N — >■ N is defined by induction if certain
initial
values -F(0),-F(1),. . . ,F(k) are defined and for each n£N, F{n- 1-1) is defined
in
terms of some or all of the preceding values F(0),F(1), . . . ,F(n). Thus inductive
definition is a special case of recursive definition. In a more formal treatment of
this
subject, the phrase “in terms of” would have to be precisely defined.
Recursive definitions which are not inductive may involve domains other than N
which have no natural ordering (so that “in terms of smaller values” makes no
sense)
or functions on N which involve definition in terms of both larger and smaller
values.
The general concept of recursion is fundamental to much of theoretical computer
science. It is a common theme uniting the different threads in [Hofstadter, 1979].
106.1.1 Example The ninety-one function F: N — >• N is defined by:
F(n) =
(n < 100)
(n > 100)
(106.1)
This is a well defined function. It has the property that F{n) = 91 if n < 100 and
F(n) = n — 10 if n > 100.
160
Collatz function
definition 4
even 5
Fibonacci func-
tion 160
function 56
odd 5
160 106.1.2 Example The Collatz function T :N + — > N + defined by:
( 1 (if n = 1)
T(n) = < T(|) (if n is even)
\ T(3n+1) (if n is odd and n > 1)
Looking at the formula, there is no reason to believe that the computation wouldn’t
loop forever for some value of the input, but no one has ever been able to discover
such a value or to prove that such a value does not exist. (Every input that has
ever
been computed does in fact given an answer, namely 1 .) In other words, although
we called it “the Collatz function”, we don’t actually know that it is a function!
Note that if you change the ‘3n+ 1’ to ‘3n — 1’ in the third line, then T( 5) is
not
defined. There is much more about this in [Guy, 1981], Problem E-16, page 120 and
in [Lagarias, 1985].
106.1.3 Exercise (hard) Prove that the ninety-one function defined by Equa-
tion (106.1) on page 159 satisfies F(n ) = 91 if n < 100 and F(n) = n— 10 if n >
100.
107. Functions with more than one starting point
The Fibonacci function is an example of a function defined in terms of two previous
values (hence requiring two initial conditions):
107.1 Definition: Fibonacci function
The Fibonacci function F : N — > N is defined by
r f(o)=o
{ F( 1) = 1
[ F(n) = F(n — 1) + F(n — 2)
(107.1)
107.1.1 Remarks
a) The Fibonacci function is called “Fibonacci” after Leonardo di Pisa, who
described it in 1220 AD. He was the son (Fi, short for Figlio) of Bonaccio.
b) The Fibonacci function has traditionally been described as the formula for the
number of pairs of rabbits you have after n months under these assumptions:
initially you have just one pair of rabbits, and every month each pair of rabbits
over one month old have a pair of children, one male and one female. And
none of them die.
Suppose you buy (trap?) the first pair of rabbits at the beginning of
month 1. Then -F(O) =0 and F( 1) = 1. At the nth month, F must satisfy
the equation
F(n) = F(n — 1) + F(n — 2) (n > 2)
since the F{n— 1) rabbits you had one month ago are still around and you
have a new pair for each of the F(n — 2) pairs born two or more months ago.
161
This explanation bears no relation to reality since rabbits take six months,
not one, to mature sexually, and they do not reliably produce one male and
one female each gestation period.
107.1.2 Example The Perrin function is defined with three starting points:
P(0) = 3
P(1) = 0
P( 2) = 2
P(n) = P(n — 2) + P(n — 3)
(107.2)
For integers larger than 1 up to a fairly large number, this function has the
property
n | P(n) n is prime.
The smallest integer > 1 for which this is false is apparently 271,441 , which is
521 2 ,
but I have not been able to check this.
A number n for which n \ P(n) is called a Perrin pseudoprime.
107.2 Recurrence relations
Since the Fibonacci function has domain N, it is the same as an infinite sequence
(see
Example 97.2.2). The values F(0),F(1),F(2),... are often called the Fibonacci
numbers. When expressed in sequence notation, the definition becomes
divide 4
domain 56
equivalent 40
Fibonacci func-
tion 160
Fibonacci num-
bers 161
induction 152
inductive defini-
tion 159
infinite 174
integer 3
odd 5
Perrin function 161
Perrin pseudo-
prime 161
recurrence rela-
tion 161
recurrence 161
/o = 0
/l = l
fn = fn— 1 T fn—2
(107.3)
Fibonacci function is called a recurrence relation or simply a recurrence.
Sometimes, but not always, a function defined by a recurrence relation can
be given a nonin ductive definition by a formula. Finding such a “closed form”
definition is called solving the recurrence relation. We have already solved
some recurrence relations. For example, the statement that the sum of the first
n odd integers is n 2 can be reworded to say that the solution to the recurrence
relation
{ :::u + i+ s „ <io7 - 4 >
is s n = n 2 .
If you can guess a solution to a recurrence relation, you can often prove it is
correct by induction. Problem 107.3.11 gives a closed solution to the Fibonacci
recurrence. Note that it would generally be better to calculate Fibonacci numbers
for small n using the recurrence relation rather that the complicated formula given
in Problem 107.3.11.
107.3 Exercise set
Exercises 107.3.2 through 107.3.11 refer to the Fibonacci sequence.
162
divide 4
div 82
even 5
GCD 88
integer 3
mod 82, 204
nonnegative integer 3
positive integer 3
107.3.1 Exercise Prove that for all nonnegative integers n, f%+i ~ fnfn +2 =
(— l) n . (Answer on page 249.)
107.3.2 Exercise Prove that for all nonnegative integers n, f n is even if and only
if 3 | n.
107.3.3 Exercise Prove that for all [positive integers n, / n +idiv/ n = l and
,/n+l mod fn = f n —l .
107.3.4 Exercise Prove that for all nonnegative integers n,
fnfn + 3-fn+lfn + 2 = (-l) n+1
107.3.5 Exercise Prove that for all nonnegative integers n, GCD(/ n +i,/ n ) = 1.
(Hint: You can use Exercise 107.3.3, or you can look at Exercise 107.3.4 and
meditate
upon Bezout.)
107.3.6 Exercise Prove by induction that
SLl/fc = fnfn+1
107.3.7 Exercise Give a proof by induction on n that for all n > 0,
fn + 2 > (|)"
(You can also prove this using Problem 107.3.11 below.)
107.3.8 Exercise Show that for all n > 0, /^ +1 — ,/ n ,/n+i — fn = ±1 •
107.3.9 Exercise (hard) (Matijasevich) Prove that if x and y are nonnegative
integers such that y 2 — xy — x 2 = ±1 , then for some nonnegative integer n , x =
f n
and y = f n +i ■ (Be careful: You are not being asked to show that (f n ,f n + i)
is a
solution of the equation for each n — that is what the Problem 107.3.8 asks for.
You are being asked to show that no other pair of integers is a solution.)
107.3.10 Exercise (hard) (Matijasevich) Show that for all nonnegative integers
m and n , if | f n , then f m \n.
107.3.11 Exercise (hard) Prove that for all nonnegative integers n,
f n = (l/Vb)(r n -s n )
where r and s are the two roots of the equation x 2 — x — 1 = 0 and r > s.
107.3.12 Exercise Let a function F:N — > N be defined by
'F(O) = 0
< .F(l) = 1
-F(ro) = 5F(n — 1) — 6F(n — 2) (n > 1)
Prove by induction that for all n > 0 , F(n) = 3 n — 2 n .
163
107.3.13 Exercise Define a function F:N— >-N by
f F(0) = F(l) = 1
\ F(n) = 2F(n — 1) + F(n — 2) (n > 1).
Show
a) F(n) is always odd.
b) -F(4fc + 2) is divisible by 3 for any integer k > 0.
function 56
integer 3
natural number 3
odd 5
successor func-
tion 163
take 57
107.3.14 Exercise (hard) (Myerson and van der Poorten [1995]) Define a func-
tion G: N^N by G(l) = G(3) = G(5) =0, G(0) = G(4) = 8, G(2) = 9, and G(n +
6) = 6G(n + 4) — 12G(n + 2) + 8G(n) for n> 5. Show that G(n) = 0 if n is odd
and
G{n) = (n- 8) 2 -2^
otherwise.
107.3.15 Exercise (Myerson and van der Poorten [1995]) Define a function G:
N — >■ Z by G(0) = 0, G(l) = 1 , G(2) = — 1 , and
G(n) = — G(n - 1) + G(n - 2) + G(n - 3)
for n > 2 . Show that
G'(n)
— ^ n even
^ odd
(Compare Exercise 94.1.4, page 136.)
108. Functions of several variables
Functions F : N 2 — > N can be defined by induction, too. One technique is to
define
a function of two variables for all values of one variable by induction on the
other
variable.
108.1.1 Example Multiplication in N, which is a function N 2 — > N, can be
defined by
f m- 0 = 0
\ m ■ (n + 1) = m ■ n + m
(108.1)
This defines m ■ n for each m € N by induction on n . The definition shows how to
define multiplication in terms of adding one.
108.1.2 Exercise The successor function s : N — > N is the function which takes
each natural number to the next one: s(n) = n + 1 . Show how to define addition
inductively in terms of the successor function.
108.1.3 Exercise Show how to define the operation (m,n) i->- m n inductively in
terms of the successor function and multiplication (defined inductively in Exam-
ple 108.1.1).
164
definition 4
empty list 164
GCD 88
head 164
nonempty list 164
recursive defini-
tion 157
recursive 157
tail 164
tuple 50, 139, 140
108.1.4 Example Theorem 65.1, page 92, gives a recursive definition of the GCD
function. It translates directly into the Pascal function in Program 108.1.
FUNCTION GCD (M,N: INTEGER) ;
BEGIN
IF M=0 THEN GCD : = N
ELSE
IF N=0 THEN GCD : = M
ELSE
GCD := GCD(N,M MOD N)
END;
Program 108.1: Program to compute the GCD
108.1.5 Exercise Define the function A : N x N — >■ N by
' A(0,y) = 1
< - 4 ( 1 ) 0 ) = 2
A(x, 0) = x + 2 for x > 2
A(x,y) = A(A(x-l,y),y-l)
a) Prove by induction that A(x, 1) = 2x for all x > 1 .
b) Prove by induction that A{x,2) = 2 X for all x > 0.
c) Prove by induction that A(x,3) = 2 j4 ( a:_1 ’ 3 ) for all x > 0.
d) Calculate A(4,4).
109. Lists
Informally, a list of elements of a set A consists of elements of A arranged from
first
to last, with order and repetition mattering. We will write them using the same
notation that we use for tuples. Thus (1,4, 3, 3, 2) is a list of elements of N. It
is
not the same list as (1,4, 3, 2) or as (4, 1,3, 3, 2) . A particular list is the
empty list,
denoted () .
We could have said that a list of elements of A is just a tuple of elements of A.
However, the specification for lists is different from that for tuples, so our
formal
treatment will start from scratch. The definition is recursive.
109.1 Definition: list
For any set A, a list of elements of A is either the empty list () or
a nonempty list. A nonempty list of elements of A has a head, which
is an element of A, and a tail, which is a list of elements of A. The list
with head a and tail (&i,... , &&) is denoted (a,bi ,... , bjf) . The list with
head a and empty tail is denoted (a ) . Every list of elements of A is
constructed by repeated application of this definition starting with the
empty list.
165
109.1.1 Remark The head of a nonempty list is not a list, but the tail is a list.
The empty list does not have a head or a tail.
109.1.2 Example () , (5), (2,1,1,— 3) and (3,3,3) are all lists of elements of Z
(lists of integers). The head of (2,1,1,— 3) is 2 and the tail is (1,1,— 3). The
head
of (5) is 5 and the tail is () .
109.2 Definition: set of lists
The set of all lists of elements of A is denoted A * . The set of all
nonempty lists of elements of A is denoted A + .
cons 165
definition 4
empty list 164
integer 3
length (of a list) 165
list constructor
function 165
list 164
recursive 157
union 47
109.2.1 Example Let A be the English alphabet. Then the lists (), (a, a, 6)
and (c,a,t,c,h) are all elements of A*. The list (2,2) is an element of N*, and
(c,a,t,c,h, 2,2) is an element of (dUN)* but not of A* or of N* .
109.2.2 Lists in Mathematica A list such as (1,5, 3, 6) in Mathematica is writ-
ten {1,5, 3, 6}.
109.3 The list constructor
Most concepts connected with lists are defined recursively using Definition 109.1.
To make this easy, we introduce the list constructor function cons :SxS*^S +
(note carefully the domain and codomain of this function), which is defined by
requiring
cons(a, <61,62, - - - ,b n )) = (a,b 1 ,b 2 ,... ,b n ) (109.1)
Thus cons (c,(a,t,c,h)) = ( c,a,t,c,h ) and cons(o, ()) = (a).
109.4 Definition: length of a list
The length (of a list) of a list L of elements of S is denoted \L\ and
is defined by
LL.l |()|=0.
LL.2 |cons(a,L)| = 1 + \L\ .
109.4.1 Example |(c,a,t)| = 3, because, by repeatedly applying Rule (109.1),
page 165, and LL.l and LL.2, we have
|<c,a,t)|
|cons(c, (a,t))|
|cons(c, cons(a, (t)))\
|cons(c,cons(a,cons(f, ())))|
1 + |cons(a,cons(f, ()))|
1 + 1 + | cons (f, ())|
1 + 1 + 1 + IOI
1 + 1 + 1 + 0 = 3
166
cons 165
definition 4
induction hypothe-
sis 152
induction 152
length (of a list) 165
list 164
proof 4
recursive 157
theorem 2
tuple 50, 139, 140
109.4.2 Remark It can be proved by induction on the length of a list that a list
of length k satisfies the specification for a fc-tuple (Definition 36.2, page 50).
Nev-
ertheless, the recursive definition of list given above has provides a useful
alternative
approach to the idea which simplifies much of the theory of lists.
109.5 Concatenation
Informally, the concatenate of two lists is obtained by writing the entries of one
and then the other in a single list. Concatenation is denoted by juxtaposition;
thus
<1,4, 4)<2, 3) = <1,4, 4, 2, 3) and <3,2,2)<) = (3,2,2) .
Again, we give a formal definition by induction.
109.6 Definition: concatenate of lists
The concatenate LN of two lists L and N is defined recursively as
follows:
CL.l ()N = N
CL. 2 cons(a, L)N = cons(a, LN) .
109.6.1 Example
(c,a,t)(c,h)
cons(c, ( a,t))(c,h )
cons(c, (a,t)(c,h))
cons (c, cons (a, ( t))(c,h ))
cons(c,cons(a, ( t){c,h )))
cons(c,cons(a,cons(f, (})(c,h)))
cons(c,cons(a,cons(f, ()(c, h))))
cons(c,cons(a,cons(f, (c, h))))
cons(c,cons(a, ( t,c,h .)))
cons(c, (a,t,c,h))
( c,a,t,c,h )
109.6.2 Remark Definition 109.6 implies that, for example, ( )(c,a,t ) = (c,a,t).
We would expect that (c,a,t)() = (c, a,f) as well. This can be proved by induction:
109.7 Theorem
For any list L, L{) = L.
Proof If L has length 0, that is, if L = () , then L() = ()() = () by CL.l. Other-
wise, assume the theorem is true for lists of length k and let L have length k +
1 .
Then L = cons (a, L') for some element a and list L' of length k, and
L() = cons (a,L r )() = cons (a,L r Q) = cons (a, L r ) = L
by CL. 2 and the induction hypothesis.
167
109.8 Theorem
Concatenation is associative. Precisely, for any lists L, M and N,
(. LM)N = L(MN ).
Proof This is also proved by induction on the length of L. If L = (} , then
(. LM)N = (( )M)N = M N = () (MN) by CL.l applied twice. Now assume that
L = cons(a,L / ) and that (L'M)N = L\MN). Then
(LM)N = (cons (a,L')M)N
= cons (a,L'M)N
by CL. 2
= cons (a,(L'M)N)
by CL. 2
= cons (a,L'(MN))
induction hypothesis
= cons (a,L')(MN)
by CL. 2
= L(MN)
109.8.1 Exercise Prove by induction that the length of the concatenate of two
lists is the sum of the lengths of the lists. Use Definitions 109.4 and 109.6
explicitly.
alphabet 93, 167
associative 70
character 93
cons 165
definition 4
digit 93
induction hypothe-
sis 152
induction 152
inductive defini-
tion 159
list 164
proof 4
real number 12
string 93, 167
theorem 2
tuple 50, 139, 140
usage 2
109.8.2 Exercise Give an inductive definition of the last entry of a list. (Answer
on page 249.)
109.8.3 Exercise Give an inductive definition of the maximum of a nonempty list
of real numbers. It should satisfy max(l,3, 17,2) = 17 and max(5) = 5, for example.
109.8.4 Exercise Give an inductive definition of the sum of the entries of a list
of real numbers. It should satisfy SUM(3,4,2,3) = 12 and SUM(42) = 42. The sum
of the empty list should be zero.
109.8.5 Exercise (hard) Prove that a list of length k satisfies the specification
for a tuple of length k (Definition 36.2, page 50).
110. Strings
110.1 Definition: string
A string is a list of characters in some alphabet.
110.1.1 Example (c,a,t) is a string in the English alphabet.
110.1.2 Usage It is customary to denote such a string by writing the characters
down next to each other and enclosing them in quotes. We will use single quotes.
Thus ‘cat’ is another notation for the string (c,a,t). We specifically regard ‘ cat
’
and ( c,a,t ) as the same mathematical object written using two different
notations.
110.1.3 Remarks
a) Note carefully that ‘cat’ is a string, “cat” is an English word, and a cat is a
mammal! Similarly, ‘52’ is a string and 52 is a number.
b) The alphabet can be any set of characters. For example ‘0101’ is a string in
the alphabet of binary digits.
168
concatenate (of
lists) 166
cons 165
even 5
induction 152
inductive defini-
tion 159
odd 5
string 93, 167
110.2 Concatenation of strings
In string notation, concatenation is simply juxtaposition: to say that the concate-
nate of ‘cat’ and ‘ch' is ‘ catch ’ , we write
l cat'‘ch' = ‘ catch '
Strings are often denoted by lowercase letters, particularly those late in the
alphabet. For example, let w = ‘cat' and x = ‘doggie'. Then wx = ‘catdoggie ' ,
ww = ‘ catcat ’ and xw = ‘doggiecat' . It is very important to distinguish w, which
here is the name of a string, from ‘w' which is a string of length one.
110.3 The empty string
The empty string could be denoted ”, but this makes it hard to read, so we will
follow common practice and use a symbol to denote the empty string. In this text,
the symbol will be A. Other texts use e or 0.
110.3.1 Example A. 1 abba' = ‘abba' = ‘abba' A, and AA = A.
110.3.2 Remark Note carefully that ‘cat' is a string, but that “A” is the name
of a string.
110.4 Exponential notation for concatenation
To designate a string concatenated with itself several times an exponential
notation
is used. If w is a string, w n is the concatenate of the string w with itself n
times.
110.4.1 Example Let w = ‘0110’. Then it follows that
iv 2 = ‘01100110’ and w 3 = ‘011001100110’
Note in particular that ‘O’ 3 = ‘000’ and ‘ 1 ,2 ‘0’ 4 = ‘110000’. We always take w
1 = w
and w° = A.
110.4.2 Exercise Find the concatenate wx if
a) w = ‘Oil’, x = ‘1010’ d) w = x = A.
b) w = A, x = ‘Oil’ e)
c) w = ‘011’, x = A. f)
w = ‘Oil’, x = w 2 .
x = ‘011’, w = x 2 .
(Answer on page 249.)
110.4.3 Exercise Let A = {a, 6} and let E be the set of strings in A* of even
length. Give an inductive definition of E . (Answer on page 249.)
110.4.4 Exercise Give an inductive definition of the set of strings in {a,b} of
odd length.
110.4.5 Exercise Give an inductive definition of the k th entry of a string. It
should exist for strings of length k or greater but not for strings of length less
than
k. Follow the pattern of the answer to Exercise 109.8.2, using cons.
110.4.6 Exercise Give an inductive definition of w n for an arbitrary string w.
The induction should be on n.
169
111. Formal languages
111.1 Definition: language
A language is a set of strings in some finite alphabet A.
111.1.1 Usage
a) In the research literature, this concept of language is often call “formal lan-
guage” .
b) If L is a language consisting of strings in A* for some finite alphabet A , then
one says that L is a “language in A" . This is common terminology but may
be slightly confusing since in fact the elements of L are not elements of A,
they are elements of A* .
111.1.2 Remark The definition says that a language is a subset of A* . Note that
the language may be infinite although the alphabet is finite.
111.1.3 Example The empty language is the set 0. No strings are elements of
the empty language.
111.1.4 Example Another example is the language {A} whose only element is
the empty string. It is important to distinguish this from the empty language 0.
111.1.5 Example Another uninteresting language is the language A* , containing
as elements every string in the alphabet A.
111.1.6 Example The set {‘01 ’,‘011’, ‘1’} is a language in {0,1}.
111.1.7 Example The set of strings in {0,1}* with 1 in the second place is a
language. Note that ‘0110’ is in the language but ‘1’ and ‘100’ are not in the
language.
111.1.8 Example If n is a positive integer, then A n denotes the set of strings
in the alphabet A of length n. Thus if A = {0,1}, then A 2 = {‘00’, ‘01’, ‘10’,
‘11’} .
We take A 0 = {A}. Note that A 1 is the set of strings of length 1 in A, and so is
not the same thing as A.
111.1.9 Example The set L of strings in {0,1}* which read the same forward
and backward is a language. For example, ‘0110’ G L, but ‘10010’ ^ L. Such strings
are called palindromes.
111.2 Theorem
For any alphabet A,
A* = A 0 U A 1 U • • • U A n U • • • (111.1)
the union of the infinite sequence of languages A 0 , A 1 ,... ,A n ,... .
Proof This follows from the fact that every string in A* has some length n.
alphabet 93, 167
definition 4
empty language 169
empty string 168
finite 173
infinite 174
integer 3
language 169
positive integer 3
proof 4
string 93, 167
subset 43
theorem 2
union 47
usage 2
170
alphabet 93, 167
definition 4
empty string 168
induction 152
inductive defini-
tion 159
infinite 174
integer 3
string 93, 167
111.2.1 Remark An element of A* is a string of finite length. A* contains as
elements no infinite sequences of elements of A, although Equation (111.1)
expresses
it as the union of an infinite sequence of sets. This follows from the definition
of
“union”: to be in A* according to 111.1, an element has to be in A n for some
integer n , so has to be a string of length n for some n .
111.3 Inductive definition of languages
A language can sometimes be given an inductive definition paralleling the
definition
of A* given previously.
111.3.1 Example Let L be the set of strings in {0,1} of the form 0 fc l fc , for k
= 1 ,
2, .... In other words, L consists of A, ‘01’, ‘0011’, ‘000111’, ‘00001111’, and so
on. Then L can be defined by induction this way:
L.l The empty string A is a string in L.
L.2 If wei, then ‘0’u/l’ £ L.
L.3 Every string in L is given by one of the preceding rules.
111.3.2 Example The set P of palindromes can be defined this way:
111.4 Definition: the set of palindromes
Let A be a set.
PAL.l The empty string A is a string in P .
PAL. 2 If a £ A, then ‘a’ is a string in P.
PAL. 3 If re is a string in P and a £ A, then awa is a string in P .
PAL. 4 Every string in P is given by one of the preceding rules.
111.4.1 Remark Thus to show that ‘a66a’ is a palindrome, we say that A is a
palindrome by PAL.l, so ‘66’ (which is ‘6A6’) is a palindrome by PAL. 3, so ‘a66a’,
which is ‘a’‘66’‘a’, is a palindrome by PAL. 3.
111.4.2 Exercise Give inductive definitions of the following languages in the
alphabet {a, 6}:
a) The set of strings containing no o’ s.
b) The set of strings containing exactly one a.
c) The set of strings containing exactly two a’s.
171
112. Families of sets
112.1 Definition: family of sets
A tuple whose coordinates are sets is called a family of sets.
112.1.1 Usage A variant of this concept is to consider a set whose elements are
sets. For some authors, a family of sets is a set of sets instead of a tuple of
sets.
112.1.2 Example Let A\ = {1,2,3}, A 2 = {2,3,4,5} and A 3 = {3, 4, 5, 7}. Then
(Ai , A 2 , A 3 ) is a family of sets, and so is (Ai,{4,5,6},0) .
112.2 Definition: union and intersection
of a family of sets
Let S = (Aj)i £n be an n-tuple of sets Ai,A 2 ,... ,A n Then
ur=i A i = i x I 3*( X € Ai)} (112.1)
nr=i A i = i x \ x ^ A i)} (112.2)
Archimedean prop-
erty 115
coordinate 49
definition 4
family of sets 171
infinite 174
real number 12
subset 43
tuple 50, 139, 140
union 47
usage 2
112.2.1 Example Let A\ = {1,2,3}, A 2 = {2,3,4,5} and A 3 = {3, 4, 5, 7}. Then
UL^ = 11,2,3,4,5,7} and ft=iA = {3}.
112.2.2 Example This notation is frequently used for infinite sets. As an exam-
ple, recall that ( a..b ) denotes the subset {r £ R | a < r < b} of the reals. Then
if
T = {(— n..n) | n £ N + }, then = (— 1..1), and, by the Archimedean property,
IJJF = R. This is often written in the notation of infinite sequences:
OO OO
|^J (— n..n) = R and P| (— n. .n) = (— 1 . . 1)
n=l n — 1
112.2.3 Warning The symbol (jf=i Ai denotes A 1 UA 2 UA 3 . In contrast, the
symbol denotes the union of all the sets A* for each positive integer i ,
specifically not including anything denoted A^. Since “ 00 ” is not an integer, A^
(if such a thing has been defined) is not included in the union.
Thus “ U )= 1 A i ” goes up to 3 and includes 3, but does not include “ 00 ”.
There is notation analogous to that of Definition 112.2 for a set of sets (in
contrast
to a tuple of sets).
112.3
If T is
Definition: union and intersection
of a set of sets
a set whose elements are sets, then
[fiT = {x (3A £ F)(x £ A)}
(112.3)
and
n^={x| (VA £ JF)(x £ A)}
(112.4)
172
empty set 33
equivalent 40
family of sets 171
hypothesis 36
implication 35, 36
intersection 47
powerset 46
subset 43
union 47
vacuous 37
112.3.1 Example Let J 7 = {{1,2, 3}, {2, 3}, {3, 4}}. Then = {1,2, 3, 4} and
rv !••*}•
112.3.2 Exercise Give an explicit description of these sets.
a) U“i(-*--* + 2 )
b) U“i(-l/i"V0
c) n“i("l/i-l + (V0)
d)
e)
112.4 Intersection and union over the empty set
If T is a family of subsets of a set B, then we can reword the definition of the
intersection of the sets in J 7 as follows: it is the set T defined by the
property:
(V5)(5e.A ^ xeS)^x&T
If J 7 is empty, the hypothesis is vacuously true, so x G T for every x £ B ; in
other
words, T = B . Thus we define the intersection of the empty set of subsets of a set
B to be B itself. Note that this definition is relative to a set containing as
subsets
all the sets in J 7 , in contrast to the intersection of families of sets in
general as
defined in the preceding section.
The union U of a family of sets T of subsets of B can be described by the
property:
(35) (S^J 7 /\x^S)<^x^U
(note the placement of the parentheses). If J 7 is empty, then there is no 5 6 J
7 , so
we define the union of an empty family of sets to be the empty set.
112.4.1 Warning In discussing sets of sets, remember that if J 7 is a set of sets,
an element of T is a set. It is a mistake to think of the words “element” and “set”
as contrasting with each other. An element of a set may or may not be a set itself.
Also, any set 5 is an element of some other set, for example of {5} .
112.4.2 Exercise Give an explicit description of IJiF and |"| J- for each of these
families of subsets of R:
a) J 7 = {{2, 4}, {1,3, 4}, {2, 5}}.
b) J 7 = {(—3. .3), (—2 . .2), (— 1 . . 1)} .
c) -T 7 = {(— 1 • • 1), (1 • • 2), (2 . . 3)} .
(Answer on page 249.)
112.4.3 Exercise What are {JVA and f) VA for any set A?
173
113. Finite sets
We begin by giving a mathematical definition of the idea that a set has n elements.
No doubt you have no trouble understanding a statement such as “S has 5 elements”
without a formal definition; however, giving a formal meaning to such statements
allows us to prove theorems about the number of elements of a set that have turned
out to have many applications.
In this definition we use the set n = {?' £ N | 1 < i < ra} (Definition 36.1,
page 50).
bijection 136
cardinality 173
definition 4
divisor 5
empty set 33
finite set 173
finite 173
integer 3
nonnegative integer 3
positive integer 3
113.1 Definition: number of elements
of a finite set
Let n be a nonnegative integer. The statement, “A set S has n ele-
ments” means there is a bijection F : n — > S .
113.1.1 Example A set has 5 elements if there is a bijection from {1,2, 3, 4, 5}
to the set. Thus the formal definition captures the usual meaning of number of
elements: if a set has 5 elements, the process of counting them — “This is the
first
element, this is the second element, ...” — in effect constructs a bijection from n
to the set.
113.1.2 Exercise Give an explicit proof that the set of positive divisors of 8 has
4 elements. (Answer on page 249.)
113.2 Definition: finite
A finite set is a set with n elements, where n is some nonnegative
integer.
113.2.1 Example The empty set is finite, since it has 0 elements, and the set
{1,3, 5, 7, 9} is finite because it has 5 elements.
113.3 Definition: cardinality
The number of elements of a finite set is the cardinality of the set. For
any finite set A, the cardinality of A is denoted |A| .
113.3.1 Example |0|=O and | {1 , 3}| = 2 .
113.3.2 Exercise Show that if A is a finite set and /3: B — )• A is a bijection
then
B is finite.
113.3.3 Exercise Show that a subset of a finite set is finite. Make sure you use
the definition of finite in your proof.
174
bijection 136
countably infi-
nite 174
definition 4
finite 173
independent 174
infinite 174
integer 3
nonnegative integer 3
positive integer 3
113.4 Infinite sets
A set which is not finite is infinite. Sets such as N, Z and R are infinite. Since
“infinite” merely means “not finite”, to say that R (for example) is infinite means
just that there is no nonnegative integer n for which the statement “R has n
elements” is true. This is certainly correct in the case of R, since if you claim
(for
example) that R has 42 elements, all I have to do is add up the absolute values of
those 42 numbers to get a number which is bigger than all of them, so is a 43rd
element.
We do not go into the extensive theory of infinite sets in this book, but it is
important to understand the difference between “finite” and “infinite” since many
theorems, such as the ones in this section, concern only finite sets.
113.4.1 Warning It is tempting when faced with proving a theorem about pos-
sibly infinite sets to talk about one set having “more elements than another” .
Such
arguments are often fallacious. For example: “There cannot possibly be an injec-
tive function from N x N to N since N x N has more elements than N.” But there
are such functions: see Exercises 93.1.8 and 113.5.3. Compare the extended hint to
Exercise 92.1.8.
113.5 Exercise set
A set S is countably infinite if there is a bijection /3:N — > S . Problems 113.5.1
through 113.5.4 explore this property.
113.5.1 Exercise Show that the set N + of positive integers is countably infinite.
(Answer on page 249.)
113.5.2 Exercise Show that Z is countably infinite.
113.5.3 Exercise Show that N x N is countably infinite.
113.5.4 Exercise (hard) Show that Q is countably infinite.
114. Multiplication of Choices
The principle of multiplication of choices, stated below, is behind the sort of
rea-
soning illustrated in the following argument: You are at a restaurant whose menu
has three columns, A, B and C. To have a complete meal, you order one of the three
items in column A, one of the five items in column B, and one of the three items in
column C. You can therefore choose 45 = 3 x 5 x 3 different meals.
114.1 Definition: independent tasks
Suppose that there are k tasks Ti, T 2 , . . . , T) 0 which must be done in
order, and, for each i = 1,2, . . . , k, there are n t ways of doing task 7) .
Suppose furthermore that doing T t in any particular way does not change
the number nj ways of doing any later task Tj . Then we say that the
tasks are independent of each other.
175
114.2 Theorem: The Principle of
Multiplication of Choices
Suppose there are k independent tasks Ti (i = 1, . . . ,k) and suppose that
for each i there are nt ways of doing T t (i = 1, . . . , k) . Then there are
nil n i = nin 2 ■ ■ - Tifc ways of doing the tasks Ti, . . . ,T k in order.
Proof We prove Theorem 114.2 by induction on k, starting at 1.
If you have one task Ti which can be done in ri\ different ways, Theorem 114.2
says you can do Ti in n!=i n * = n i different ways, which of course is true.
Now suppose the theorem is true for k tasks. Assume you have k + 1 tasks
Ti,... ,T k ,T k+ i, and for each i there are n* ways of doing task T % . Let m be
the
total number of ways of doing the tasks Ti, . . . ,T k in order. Suppose you have
done
them in one of the m ways. Then you can do T k+ \ in any of ni+i ways. Thus
for each of the m ways of doing the first k tasks, you have ni + \ ways of doing
the
(k + l)st; therefore, there are altogether n l+ \ + n ,; + 1 H hnj+i (sum of m
terms)
ways of doing the k + 1 tasks. This means that there are m x m+ 1 ways to do
Ti,... ,T k+ 1 in order.
By induction hypothesis, m = n!=i n i > so the number of ways of doing the tasks
Ti,... ,T k+ 1 is
n k+ 1 • (J^n*)
1=1
which by 105.1.5 is nf=i lra *> as required.
114.2.1 Worked Exercise How many three-digit integers (in decimal notation)
are there whose second digit is not 5?
Answer Writing such a sequence of digits can be perceived as carrying out three
tasks in a row:
T.l Write any digit except 0.
T.2 Write any digit except 5.
T.3 Write any digit.
There are 9 ways to do T.l, 9 ways to do T.2, and 10 ways to do T.3, so
according to Theorem 114.2, there are 810 ways to do T.l, T.2, T.3 in order.
114.2.2 Worked Exercise Find the number of strings of length n in {a,b,c}*
that contain exactly one a.
Answer This requires us to look at the problem in a slightly different way from
Worked Exercise 114.2.1. To construct a string of length in {a,b,c}* with exactly
one a requires us to
a) Choose which of n possible locations to put the one and only a (n ways to
do this).
b) For each of the n — 1 other locations, choose whether to put a b or c there (2
choices for each location, 2 n_1 choices altogether).
It follows that there are n-2” _1 such strings.
decimal 12, 93
digit 93
induction hypothe-
sis 152
induction 152
integer 3
proof 4
theorem 2
176
digit 93
even 5
finite 173
include 43
integer 3
powerset 46
string 93, 167
theorem 2
114.2.3 Exercise Find the number of 5-digit integers with ‘3’ in the middle place.
(Answer on page 249.)
114.2.4 Exercise Find the number of even 5-digit integers. (Answer on page
249.)
114.3 Exercise set
In exercises 114.3.2 through 114.3.5, A = {a,b,c}.
114.3.1 Exercise Find the number of strings of length n in {a,b,c}* with no
a’s. (Answer on page 249.)
114.3.2 Exercise Find a formula F{n) for the number of strings in A* of length
n, for each n £ N. (Answer on page 249.)
114.3.3 Exercise Find a formula G{n ) for the number of strings in A* of length
n which begin and end with a. (Answer on page 249.)
114.3.4 Exercise Find a formula H{n ) for the number of strings in A* of length
n which do not begin or end with c.
114.3.5 Exercise Find a formula for the number of strings in A* of length n> 2
which have a ‘a’ in the third place.
114.3.6 Exercise In the USA a local telephone number consists of a string of 7
digits, the first two of which cannot be 0 or 1 . How many possible local telephone
numbers are there?
115. Counting with set operations
Almost every operation associated with set theory has a corresponding combinatorial
principle or counting technique applicable to finite sets associated with it. Some
of
these are obvious, others are more subtle. The first example has to do with
inclusion:
115.1 Theorem
If A and B are finite sets and A C B , then \ A\ < \B\ .
(We told you some of the principles were obvious!)
115.1.1 Exercise Show that if A and B are finite then |Anl?| < |A|.
There is a principle for powersets, too.
177
115.2 Theorem
If a set A has n elements then VA has 2 n elements.
Proof The easiest proof of this theorem uses the Principle of Multiplication of
Choices (Theorem 114.2). If A has n elements and you want to describe a subset of
A, you may go through the n elements of A one by one and say whether each one
is in the subset. There are two choices (yes or no) for each element and n
elements,
so the Principle of Multiplication of Choices says that you can make 2 n choices
altogether.
115.2.1 Remark As is the case with any counting technique based on the Prin-
ciple of Multiplication of Choices, it is also possible to prove Theorem 115.2 by
a direct argument using induction. (Recall that the Principle of Multiplication of
Choices was proved by induction.)
Cartesian product 52
even 5
induction 152
Multiplication of
Choices 175
odd 5
powerset 46
proof 4
singleton 34
subset 43
theorem 2
115.2.2 Worked Exercise How many subsets with an even number of elements
does a set with n elements have? Explain your answer.
Answer A set S with n elements has 2 n ~ 1 subsets with an even number of ele-
ments. Proof: To give a subset A of S, for each element of S except the last one
you must choose whether that element is in A. That requires 2 n ~ 1 independent
choices. You have no choice concerning the last element: if at that point the
subset
has an odd number of elements so far, you must include the last one, and if it has
an even number so far, you must not include the last one.
115.2.3 Exercise Let S be an ra-element set. How many elements do the follow-
ing sets have?
a) The set of nonempty subsets of S .
b) The set of singleton subsets of S .
c) The set of subsets of the powerset of S .
(Answer on page 249.)
The following theorem can be proved using Multiplication of Choices.
115.3 Theorem
If A and B are finite, then \A x B\ = \A\ \B
115.3.1 Exercise If A has m elements and B has n elements, how many ele-
ments do each of these sets have?
a) Ax A
b) V(AxA)
c) V(A x B)
115.3.2 Exercise Prove Theorem 115.1.
115.3.3 Exercise Prove Theorem 115.3.
178
family of sets
finite 173
function 56
proof 4
subset 43
theorem 2
union 47
171 115.3.4 Exercise Suppose A has m elements and B has n elements.
a) Prove that MAX(m,n) < |TU B\ < m + n.
b) Prove that 0 < \A(^B\ < MIN(m,n) .
c) Prove that the symbols ‘ < ’ in (a) and (b) cannot be replaced by ‘ < ’ .
115.3.5 Exercise Let A be a finite set and F:A—$B a function. Prove that
\T(F)\ = \A\.
116. The Principle of Inclusion and Exclusion
116.1 Theorem
Let A and B be finite sets. Then
\AUB\ = \A\ + \B\-\AHB\ (116.1)
Proof This follows from the fact that the expression |T| + \B\ counts the elements
which are in both sets twice, so to get the correct count for \AlJB\, you have to
subtract \AnB\.
116.1.1 Remark More generally, if C and D are also finite sets, then
\AUBUC\ = \A\ + \B\ + \C\ -\AnB\-\Anc\-\Bnc\ + \AnBnc\
and
\AUBUCUD\ = \A\ + \B\ + \C\ + \D\
- \AnB\ - |AnC| - \ACD\
-\Bnc\-\BnD\-\cnD\
+ \AnBnC\ + \AnBnD\
+ \AnCnD\ + \BnCnD\
-\AnBnCnD\
(116.2)
(116.3)
116.1.2 The general principle Equations (116.1)-(116.3) are special cases of a
general principle which requires some notation to state properly. Let T be a family
of n distinct finite sets. For each k = 1,2,. . . ,n, let Fk be the set of fc-
element
subsets of F . For example, if F = {A,B,C,D} , then
F 3 = {{A,B,C},{A,B,D},{A,C,D},{B,C,D}}
Then we have:
179
116.2 Theorem: The Principle of
Inclusion and Exclusion
Using the notation of Section 116.1.2,
W = £1*1- £ |nG|
+ £ |nG| - ...
(116.4)
XeT G&X 2
G&T . 3
-(-P E i n °i +
Ge^fe
•••-(-I) n |n^|
116.2.1 Remarks
a) The first sum is over the elements of T (which are themselves sets), whereas
the others are over intersections of subfamilies G of J- , with a plus sign for
subfamilies with an odd number of elements and a minus sign for those with
an even number of elements.
b) You should check that Equations (116.1)— (116.3) are special cases of this Prin-
ciple.
c) The Principle of Inclusion and Exclusion will not be proved here, but you
should be able to see with no trouble why it is true for families of three or
four sets.
116.2.2 Example The Principle of Inclusion and Exclusion is stated as an equa-
tion, so you can solve for one of its terms if you know all the others.
For example, suppose there was a party with 9 people, including 5 Norwegians.
There was only one man at the party who was neither a vegetarian nor a Norwe-
gian. All the vegetarians were Norwegians and two of the women were Norwegians.
Exactly one woman was a vegetarian. How many women were at the party?
To solve this, let W be the set of women, N the set of Norwegians, and V
the set of vegetarians. The party had 9 people, and only one was not in W U
N U V, so |ITUlVUH| = 8. We are given that \N\ = 5. Since 2 of the women
were Norwegians, |WnIV| =2, and since one woman was a vegetarian and every
vegetarian was a Norwegian, we know \W n V\ = | W n IVn V| = 1 and also |H| =
\NHV\.
Thus in the sum
\WUNUV\ = \W\ + \N\ + \V\-
\wnN\-\wnv\-\Nnv\ + \wnNnv\
we have
8=\w\ + 5 + \v\-2-\wnv\-\Nnv\ + \wnNnv\
or since |H| = \N D V \ ,
8 = \ W\ + 5 — 2 = \ W\ + 3
even 5
inclusion and exclu-
sion 179
intersection 47
odd 5
theorem 2
so that there were 5 women at the party.
180
definition 4
fact 1
family of sets 171
finite 173
implication 35, 36
pairwise disjoint 180
partition 180
subset 43
union 47
usage 2
116.2.3 Exercise You have a collection of American pennies. Three of them are
zinc pennies and eight of them were minted before 1932. What do you have to know
to determine the total number of pennies? Explain your answer! (Answer on page
249.)
116.2.4 Exercise A, B and C are finite sets with the following properties: A U
BuC has 10 elements; B has twice as many elements as A; C has 5 elements; B
and C are disjoint; and there is just one element in A that is also in B . Show
that
A has at least 2 elements.
116.2.5 Exercise Suppose that A, B and C are finite sets with the following
properties:
(i) B has one more element than A.
(ii) C has one more element than B .
(iii) AnB is twice as big as An C .
(iv) B and C have no elements in common.
Prove that |AU BL)C\ is divisible by 3.
116.2.6 Exercise Cornwall Computernut has 5 computers with hard disk drives
and one without. Of these, several have speech synthesizers, including the one
with-
out hard disk. Several have Pascal, including all those with synthesizers. Exactly
3
of the computers with hard disk have Pascal. How many have Pascal?
117. Partitions
117.1 Definition: partition
If C is a set, a family n of nonempty subsets of C is called a partition
of C if
PAR.l C = un, and
PAR. 2 For all A, B e n , A^B => AnB = 0 .
117.1.1 Usage The elements of the partition n are called the blocks of n. If
igC, the block of n that has x as an element is denoted [x]n> or just [x] if the
partition is clear from context.
117.1.2 Fact P.2 says the blocks of n (remember that these subsets of C) are
pairwise disjoint: if they are different, they can’t overlap.
117.1.3 Fact P.l and P.2 together are equivalent to saying that every element of
C is in exactly one block of n.
117.1.4 Example Here are three partitions of the set {1,2, 3, 4, 5}:
a) n 1 = {{l,2},{3,4},{5}}.
b) H 2 = {{1},{2},{3},{4},{5}}.
c) n 3 = {{1,2, 3, 4, 5}}.
181
117.1.5 Example The set {{1,2},{3,4},{5},0} is not a partition of any set
because it contains the empty set as an element.
117.1.6 Example Let S be any nonempty set and A any proper nontrivial subset
of A. Then {S', S' — A} is a partition of A with two blocks.
117.1.7 Example The empty set has a unique partition which is also the empty
set. It has no blocks.
117.1.8 Exercise Why does Example 117.1.6 have to require that A bea proper
nontrivial subset of A?
block 180
empty set 33
finite 173
infinite 174
nontrivial subset 45
partition 180
proper subset 45
tuple 50, 139, 140
union 47
usage 2
117.1.9 Worked Exercise Let S be a nonempty finite set with n elements. How
many partititions of S with exactly two blocks are there?
Answer There are 2 n — 1 nonempty subsets of S and, except for S itself, each one
induces a two-block partition as in Example 117.1.6. This does not mean that there
are 2 n — 2 two-block partitions because that would count each two-block partition
twice (a subset and its complement each induce the same two-block partition). So
the correct answer is that there are
l(2 n -2) = 2 n - 1 -l
two-block partitions.
117.1.10 Warning One of the commonest mistakes made by people just begin-
ning to learn counting is to come up with a seemingly reasonable technique which
unfortunately counts some things more than once.
117.1.11 Exercise Find a formula for the number of partitions with exactly three
blocks of an n -element set.
117.1.12 Usage A partition with a finite number of blocks (even though the
blocks might be infinite sets) is commonly written as a tuple, e.g., n = (Aj)j en .
Even so, if n' is another partition which is the same as n except for ordering,
they
are regarded as the same partition even though they are different tuples. We will
follow that practice here.
117.1.13 Exercise Which of the following are partitions of S' = {1,2, 3, 4, 5}?
Here, A = {1,2}, B = {3,4,5}, U7 = {3} , L> = {4,5}.
a)
b)
c)
d)
{A,B} e) {S}
{A,B,C} f) {{4 |zGS}
{A,C,D} g) {C,S-C}
{A,H,0} h) {AuC,D}
(Answer on page 250.)
182
block 180
family of sets 171
finite 173
floored division 87
inclusion and exclu-
sion 179
infinite 174
integer 3
negative integer 3
partition 180
positive integer 3
remainder 83
theorem 2
117.2 Partition of Z by remainders
Any poesitive integer n induces a very important partition of the set Z of
integers.
This partition is denoted Z/n. The blocks of Z/n are the n sets
C r = {m £ Z | m leaves a remainder of r when divided by n}
for 0 < r < n. For negative m floored division must be used. (Observe that the
notation “ C r ” requires you to depend on context to know what n is.) Thus Z/n =
{C r | 0 < r < n}.
117.2.1 Remark It is important to understand that Z/n is a finite set, even
though each block is an infinite set.
117.2.2 Example If n = 3, Z/3 has three blocks. One of them is C\ , which is
the set of integers which leave a remainder of 1 when divided by 3. Thus 1, —2
and 16 are in C\ . Cq is the set of integers divisible by 3. Thus Z/3 = {Cq,Ci,C
2 } ■
117.3 Exercise set
In problems 117.3.1 through 117.3.5, provide an example of a partition II of Z with
the given property.
117.3.1 II has at least one block with exactly three elements. (Answer on page
250.)
117.3.2 {1,2} and {3} are blocks of II.
117.3.3 II has at least one finite block and at least one infinite block.
117.3.4 II has an infinite number of finite blocks.
117.3.5 II has an infinite number of infinite blocks.
118. Counting with partitions
P.2 in Definition 117.1 implies that, in the statement of the Principle of
Inclusion
and Exclusion, the sums over families with more than one element disappear. This
gives the following theorem, which is obvious anyway:
118.1 Theorem
If II = (Aj)j en is a partition of a finite set C , then \C\ = £” =1 \Ai\ .
This Theorem together with the phenomenon of Example 117.1.6 gives a method:
183
118.1.1 Method
To count the number of elements of a subset A of a set S, count the
number of elements of S and subtract the number of elements of the
complement S — A.
118.1.2 Worked Exercise How many strings of length n in {a,b,c}* are there
that have more than one a?
Answer We will use Method 118.1.1. We know from Exercises 114.2.2 and 114.3.1
that there are 2 n strings with no a and n ■ 2 n ~ 1 strings with one a. Since
there are
3 n strings of length n in {a, b, c}* , the answer is 3 n — 2 n — n ■ 2 n ~ 1 .
block 180
class function 183
definition 4
partition 180
surjective 133
take 57
usage 2
118.1.3 Exercise How many strings of length n in {a, b}* are there that have
more than one a?
118.1.4 Exercise How many strings of length n in {a, b}* are there that satisfy
the following requirement: If it has an a in it, it has at least two.
118.1.5 Exercise How many strings of length n in {a,b,c}* are there that have
exactly two different letters in them (so each one is either all a’s and 6’s, all
a’s
and c’s, or all b 1 s and c’s.)?
118.1.6 Exercise In the USA the identifying name of a radio station consist of
strings of letters of length 3 or 4, beginning with K or W. Upper and lower case
are not distinguished. How many legal identifying names are there?
119. The class function
119.1 Definition: the class function
If n is a partition of a set A, then the class function clsn : A — > n
takes an element a of A to the block of n that has it as an element.
119.1.1 Example If A = {1,2, 3, 4, 5} and H = {{1,2},{3,4,5}}, then clsn (3) =
{3,4,5}.
119.1.2 Usage A common notation for the class function is [] : A — > n; in Exam-
ple 119.1.1, one would write [3] = {3,4,5}.
119.1.3 Example In Example 117.1.4, [2]n 1 = {l,2}.
119.1.4 Warning Note that in Example 119.1.1, [3] = [4] = [5] , but [2] / [3] . In
mathematics, the fact that two different names are used does not mean they name
different things. (This point was made before, in Example 58.1.2.)
119.1.5 Example If n is the partition Z/3, then [2] = [5] = [—1] = C' 2 , and [3] =
C 0 .
119.1.6 Exercise Prove that for any set S with partition n, the class function
els : S — > n is surjective.
184
block 180
definition 4
family of sets 171
floored division 87
function 56
image 131
integer 3
list 164
mod 82, 204
negative integer 3
partition 180
quotient set (of a
function) 184
remainder 83
take 57
theorem 2
120. The quotient of a function
We mentioned the partition Z/n = {C r | 0 < r < n} in section 117.2. It is a
special
case of a construction which works for any function:
120.1 Theorem
Let F : A—¥ B be a function. Then the family of sets
{F _1 (6) | belmF}
is a partition of A .
120.2 Definition: quotient set
The set {T 1_1 (6) | b G ImT 1 } is denoted A/F and is called the quotient
set of F .
120.2.1 Example Consider the function F : {1,2,3} — > {2, 4, 5, 6} defined by
F(l) = 4 and F( 2) = F( 3) = 5. Its quotient set (of a function) is {{1},{2,3}}.
120.2.2 Example The quotient set (of a function) of the squaring function S :
R — > R defined by S(x) = x 2 is
R/S = {{r, — ?'} | r G R}
Every block of R/S has two elements with the exception of the block {0}. The
notation “{{r,— r} | r G R}” for R/S 1 lists {0} as {0,-0}, but that is the same
set as |0j. Note that every set except |0| is listed twice in the expression
“{{r,-r} | r G R}” .
120.2.3 Example Let’s look at the remainder function R n (k ) = fcmodn for a
fixed integer n. This function takes an integer k to its remainder when divided by
n.
(As earlier, we use floored division for negative k). For a particular remainder r,
the set of integers which leave a remainder of r when divided by n is the set we
called C r earlier in the section. Thus the quotient set of R n is the set we
called
Z /n.
120.3 Proof of Theorem 120.1
We must show that the blocks of A/F are nonempty and that every element of A
is in exactly one block of A/F .
That the blocks are nonempty follows the fact that A/F consists of those 1 ?_1 (6)
for which b G ImF; if b G ImF, then there is some a G A with F(a) = b, which
implies that a G F~ 1 (6), so that F _1 (6) is nonempty. Since a G F~ 1 (F(a )) ,
every
element of A is in at least one block. If a G F~ 1 (b) also, then F(a) = b by
definition,
so F~ 1 (F(a)) = F _1 (6) , so no element is in more than one block.
120.3.1 Exercise For a function F : S —> T , define a condition on the quotient
set S/F which is true if and only if F is injective. (Answer on page 250.)
185
120.3.2 Exercise Give examples of two functions F : N — > N and G : N — >■ N with
the property that F is surjective, G is not surjective and F and G have the same
quotient set. (Thus, in contrast to Exercise 120.3.1, there is no condition on the
quotient set of a function that forces the function to be surjective.)
120.4 Exercise set
In Problems 120.4.1 through 120.4.5, provide an example of a function F : R — > R
for which R /F has the given property.
120.4.1 R /F has at least one block with exactly three elements. (Answer on
page 250.)
120.4.2 R/F has exactly three blocks.
120.4.3 R/F is finite.
120.4.4 Every block of R/F is finite.
120.4.5 Every block of R/F has exactly two elements.
120.4.6 Exercise Suppose F : A — >• B is a function, and x and y are distinct
elements of B . Suppose also that |A| = 7, |£>| = 4, ImF = B — {y} , and that the
function F | (A — F~ 1 (x) s ) is injective.
a) How many elements does A/F have?
b) How many elements are there in each block of A/ FI
120.4.7 Exercise (hard) Let A be a set, n a partition of A and B a subset of
A. Define the set n|R of subsets of B by
U\B = {CnB I Gen and CnR/0}
a) Prove that n|L> is a partition of B .
b) Give an example to show that the set {CnB \ C 6 n} need not be a partition
of B .
120.4.8 Exercise (hard) Let A be a set, n a partition of A, and $ a partition
of n . For any block C € 4> , let Be be the union of all the blocks Bg n for which
B G C . Show that {Be \ C £ <h} is a partition of A. (For many people, this
exercise
will be an excellent example of a common phenomenon in conceptual mathematics:
It seems incomprehensible at first, but when you finally figure out what the
notation
means, you see that it is obviously true.)
block 180
finite 173
function 56
image 131
injective 134
partition 180
subset 43
186
bijection 136
block 180
codomain 56
function 56
image 131
injective 134
surjective 133
theorem 2
121. The fundamental bijection theorem
The following theorem forms a theoretical basis for very important constructions in
abstract mathematics:
121.1 Theorem: The Fundamental Bijection Theorem
for functions
Let F : A-$ B be a function, and define @f to be the function F~ l {b) >-)•
b. Then (3p is a bijection fip : A/F ->ImF.
121.1.1 Example For the function F : {1,2,3} — > {2, 4, 5, 6} defined by F(l)=4
and F( 2) = F( 3) = 5, we have /3f({1}) = 4 and ^({2,3}) = 5.
121.1.2 Remark The input to the bijection is a set, namely a block of A/F , and
the output is an element of the codomain of F . The statement that /3p({2,3}) = 5
means that when you plug {2,3} into (5f ( not when you plug 2 in or 3 in!) you
get 5.
121.2 Proof of Theorem 121.1
It is easy to see that (5f really is a bijection. If b £ Irn F , then there is some
element
a £ A for which F(a) = b, so 1 ?_1 (6) is nonempty and hence an element of A/F .
Then /3 f(F -1 (&)) = b so (3f is surjective.
Proving injectivity reduces to showing that if F _1 (&) T 7, 1 (c) , then bfic.
If F _1 (6) F-\c), then there is some element a £ A for which a £ 1 ? ^ 1 (6) but
a $/ F~ x {c) (or vice versa). The statement a £ 1 ?_1 (6) means that F(a) = b, and
the statement a ^ F _1 (c) means that F(a) c. Thus bfic, as required.
121.2.1 Exercise Let A = {1,2, 3,4,5} . For each function F : A R given below,
write out all the values of the bijection fip '■ A/F — > ImP given by Theorem
121.1.
a) F(l) = F( 3) = F( 5) = 4, F( 4) = 6, F( 2) = 0.
b) F(n) = 3 for all n £ A .
c) F(n) = n for all n £ A.
d) F(n) = n 2 for all n £ A.
e) F(n) = 7i 3 — 3n 2 + 2n — 5 for all n £ A.
(Answer on page 250.)
187
122. Elementary facts about finite sets and functions
This chapter contains miscellaneous results, mostly easy, concerning finite sets
and
functions between them. The facts about finite sets A and B in the following
theorem are not difficult to see using examples. We give part of the proof and
leave
the rest to you.
122.1 Theorem
Let A and B be finite sets. Then:
a) |T| = \B\ if and only if there is a bijection (3: A—t B .
b) |T| < \B\ if and only if there is an injective function F : A—> B .
c) If B is nonempty, |T| > \B\ if and only if there is a surjective
function G : T — > B .
bijection 136
bijective 136
composition (of
functions) 140
finite 173
function 56
image 131
injective 134
proof 4
quotient set (of a
function) 184
subset 43
surjective 133
theorem 2
Proof By Definition 113.1, if A and B both have n elements then there are
bijections (3 : n — > A and f3':n—^B. Then, using Theorem 101.5, page 149, Theo-
rem 101.3, page 148 and Exercise 98.2.7 of Chapter 98, (3 ' ° (3~ l : A -» B is a
bijection.
To finish the proof of (a), we must show that if there is a bijection (3: T — >■ B
then
A and B have the same number of elements. This is left as an exercise.
We also leave (b) as an exercise, and prove half of (c). Suppose A has m
elements and B has n elements with m>n> 0 . Then there are bijections f3 : m — >
A and f3' : n — >■ B . Let us define a function F : m — > n by: F(k) = k if k < n,
and F(k) = n if k > n. F is surjective, because if 1 < i < n, then F(i) = i . Then
(3' °F° (3~ l : A—> B is the composite of a bijection, a surjection and a
bijection, so
is a surjection by Exercise 98.2.7 of Chapter 98.
122.1.1 Exercise Complete the proof of Theorem 122.1.
122.1.2 Exercise Use the principles of counting for finite sets that we have intro-
duced to prove that if II is a partition of a finite set A , then |II| < 14 -
Here is another useful theorem:
122.2 Theorem
If A and B are finite sets and \A\ = \B\ , then a function F : A—> B is
injective if and only if it is surjective.
Proof Let F : T — >■ B be injective. Then ImL, being a subset of B , has no more
than \B\ elements by Theorem 115.1. Since F is injective, ImF has at least |T|
elements by Theorem 122.1(a). Since |T| = |L>|, it follows that ImL has exactly
\B\ elements, so Im F = B. Hence F is surjective.
Conversely, if F is not injective, then the quotient A/F has fewer elements
than A. The fundamental bijection theorem (Theorem 121.1) says that then ImF
has fewer elements than A, so it has fewer elements than B since |T| = |L>|. That
means ImF B , so F is not surjective.
188
alphabet 93, 167
bijection 136
decimal 12, 93
digit 93
finite 173
function 56
include 43
infinite 174
injective 134
integer 3
Multiplication of
Choices 175
powerset 46
proof 4
shift function 188
surjective 133
theorem 2
122.2.1 Warning Observe that if |A| = \B\, then Theorem 122.1(a) says there
is an injection from A to B and Theorem 122.1(b) says that there is a surjection
from A to B . But Theorems 122.1(a) and (b) do not say that the injection and
the surjection have to he the same function, so it would be a fallacy to deduce
Theorem 122.2 from those two facts.
122.2.2 Warning Theorem 122.2 allows you to determine whether a function
from a finite set to itself is a bijection by testing either injectivity or
surjectivity —
you don’t have to test both. However, you have to test both for infinite sets. For
example, the shift function m- )-n+l:N— >N is injective but not surjective (0
is not a value) and 0 i — 0 , nun - 1 for n > 0 defines a function N — > N which is
surjective but not injective, since 0 and 1 both have value 0.
Here is a counting principle for function sets:
122.3 Theorem
If A = n and \B
| = m, then there are m n functions from A to B . In
other words,
b a
II
Proof To construct an element of B , that is, a function from A to B, you have
to say what F(a) is for each element of A. For each a you have m choices for
B(a) since F(a) has to be an element of B and B has m elements. There are
n elements a of A for each of which you have to make these choices, so by the
Principle of Multiplication of Choices there are m n possibilities altogether.
122.3.1 Exercise How many ways are there of assigning a letter of the alphabet
to each decimal digit, allowing the same letter to be assigned to different digits?
(Answer on page 250.)
122.3.2 Exercise
a) Show by quoting principles enunciated here that if A and B are finite, A C B
and Af^B, then there is no bijection from A to B .
b) Show that the statement in (a) can be false if A and B are infinite.
122.3.3 Exercise Let F{n) be the number of functions from VS to S , where S
is a set with n elements, and let G(n ) be the number of functions from S to its
powerset. For which integers n is F(n ) = G(n)l
189
123. The Pigeonhole Principle
In its contrapositive form. Theorem 122.1(b) says the following:
123.1 Theorem
For any finite sets A and B , if |A| > \B\ , then no function from A to
B is injective.
123.1.1 Example If you have a set A of pigeons and a set B of pigeonholes,
\A\ > \B\, and you put each pigeon in a pigeonhole (thereby giving a function from
A to B ) , then at least one pigeonhole has to have two pigeons in it (the function
is
not injective). For this reason, Theorem 123.1 is called the Pigeonhole Principle.
block 180
contrapositive 42
function 56
injective 134
partition 180
Pigeonhole Princi-
ple 189
recurrence 161
subset 43
theorem 2
123.1.2 Example An obvious example of the use of the Pigeonhole Principle is
that in any room containing 367 people, two of them must have the same birthday.
Note that the Pigeonhole Principle gives you no way to find out who they are.
123.1.3 Worked Exercise Let S = {n:N | 1 < n < 10}. Show that any subset
T of S with more than 5 elements contains two numbers that add up to 11.
Answer The following are all the two-elenrent subsets of S whose elements add
up to 11: {1,10}, {2,9}, {3,8}, {4,7}, {5,6}. They form a partition of S with
five blocks. Every element of T is in one of these subsets, and since T has more
than five elements, by the Pigeonhole Principle two different elements must be in
the same block of the partition.
123.1.4 Exercise Let S be as in Worked Exercise 123.1.3. Show that if T C S
and \T\ > 4 then there are two different elements of T that have the same remainder
when divides by 3.
123.1.5 Exercise Let A = {n : N | 1 < n < 12} . Find the least integer n so that
the following statement is true: If T C A and |T| > n, then T contains two distinct
elements whose product is 12.
124. Recurrence relations in counting
Many counting formulas can be derived as recurrence relations. In many cases, you
can then find a closed formula which evaluates the recurrence relation, but even
if you cannot do that, the recurrence relation gives you a way of evaluating the
formula for successive values of n.
124.1 Theorem
If A has n elements, then there are n\ different permutations of A .
To prove this, it is useful to prove something more general.
190
bijection 136
definition 4
even 5
odd 5
proof 4
recurrence 161
string 93, 167
subset 43
124.2 Theorem
The number of bijections between two n-element sets is n\ .
Proof Let P(n) be the number of bijections between two n-element sets. Then
P(0) = P(l) = 1. Let A and B be two sets with n+ 1 elements. Let a G A. Then
in constructing a bijection from A to B we have n + 1 choices for the value of the
bijection at a. If we choose b £ B , then what is left is a bijection from A — {a}
to
B — {b} . These are both n-element sets, so there are P(ji) of these, by definition
of P(n) . Hence
P(n + 1) = (n + 1) • P(n)
This is the recurrence relation which (with P(0) = 1) defines n! (see Section
105.1.6,
page 158), so P{n ) =n!.
Here is another example of using recurrences in counting:
124.2.1 Worked Exercise Derive a formula or recurrence relation for the number
of strings of length n in {0,1}* with an even number of l’s.
Answer Let Fiji) be the number of such strings. Obviously F( 0) = T\l) = 1.
There are Fiji) strings of length n with an even number of ones and 2 n — Fiji)
with an odd number of ones. (Note that there is no justification at this point for
assuming that the number of strings of length n with an even number of ones and
the number with an odd number of ones are the same.) You can adjoin a 0 to a
string of the first type and a 1 to a string of the second type to get a string of
length
n + 1 with an even number of ones. Thus Fiji + 1) = Fiji) + 2 n — Fiji) = 2 n .
This
is a case of a recurrence relation that solves itself!
124.2.2 Exercise Derive a formula or recurrence relation for the number of ways
to arrange n people around a circular table. (All that matters is who sits on each
person’s left and who sits on his or her right.)
124.2.3 Exercise Derive a formula or recurrence relation for the amount of money
in a savings account after n years if the interest rate is i% compounded annually
and you start with $100.
125. The number of subsets of a set
125.1 Definition: binomial coefficient
C(n,k) denotes the number of A; -element subsets of an n-element set.
125.1.1 Example (7(4,0) = 1 (there is exactly one subset with no elements in a
set with 4 elements), (7(4,1) = 4 (there are four singleton subsets of a four-
element
set) and (7(4,2) =6 (count them).
We can deduce some immediate consequences of the definition:
191
125.2 Theorem
For all n> 0 and k> 0 ,
a) C(n, 0) = 1 .
b) C(n,n) = 1.
c) C(n,k) = C(n,n— k) .
d) C(n,k) = 0 ifk>n.
Proof
a) There is exactly one empty subset of any set, so C(n, 0) = 1 for any n.
b) An n-element set clearly has exactly one subset with n elements, namely
itself.
c) This follows from the fact that for a particular k there is a bijection between
k element subsets of an n-element set and their complements, which of course
are (n — k) -element subsets.
d) Obvious.
C(n,k ) is called a binomial coefficient because of the formula in the following
theorem. C(n,k) is also written
125.3 Theorem
For all real x and y and all nonnegative integers n and k,
(x + y) n = J2C(n,k)x n - k y k (125.1)
k = 0
I won’t give a formal proof, but just sketch the idea.
(x + y) n = (x + y)(x + y)---(x + y) (125.2)
where (x + y) occurs n times in the expression on the right. In the expanded
version
of {x + y) n , each term occurs by selecting an x or a y in each factor of the
right side
of Equation (125.2) and multiplying them together (try this on {x + y) (x + y) (x +
y) ). You get one occurrence of x n ~ k y k by choosing a subset of k factors (out
of the
n that occur) and using y from those factors and x from the n — k other factors.
There are C(n,k) ways to do this, so that Equation (125.1) follows.
125.4 Recurrence relation for C(n,k )
We can get a recurrence relation for C(n,k) which will allow us to calculate it.
Suppose, for a fixed k, we want to know C(n + 1 ,k) , the number of k -element
subsets of an n + 1 -element set A . Let a £ A . Then we can get each subset of A
that has a in it exactly once by adjoining a to a ( k — l)-element subset of A —
{a} ,
so there are C(n,k — 1) A; -element subsets of A that have a as an element. On the
other hand, every A’ -element subset of A that does not contain a as an element is
a A -element subset of A — {a}, so there are C(n,k) of them.
Every subset of A either has a as an element or not, so we have the following
theorem:
binomial coeffi-
cient 191
empty set 33
proof 4
recurrence 161
subset 43
theorem 2
192
basis step 152
recurrence rela-
tion 161
theorem 2
125.5 Theorem
For all n> 0 and k > 0 ,
(7(71,0) = 1
C(n,k ) = 0 if k > n (125.3)
C(n + l,k) = C(n,k — 1) + C(n,k) otherwise.
125.5.1 Example
(7(4,2) = (7(3,1) + (7(3, 2)
= (7(2,0) + (7(2,1) + (7(2,1) + (7(2, 2)
= 1 + 2 -(7(2,1) + (7(2,2)
= 1 + 2((7(1,0) + (7(1, 1)) + (7(1, 1) + (7(1,2)
= 1 + 2(1 + (7(0, 0) + (7(0, 1)) + (7(0, 0) + (7(0, 1)
= l + 2- 2 + l = 6
(125.4)
(125.5)
125.5.2 Example The recurrence relation for C(n,k) can be used to give an
inductive proof of Theorem 125.3.
The basis step is to prove that
(x + y)° = ^(7(O,£0*-y
k = 0
The sum on the right has only one term, namely (7(0, 0)x°y° , which is 1, as is the
expression on the left.
Inductive step: Assume
(x + y) n = Y,C(n,k)x n - k y k
k = 0
We must prove
n + 1
(* + y) n+1 = J2 C ( n + 1 ,k)x n+1 ~ k y k
k = 0
We now make a calculation. In this calculation it is convenient to define C(n, — 1)
193
to be 0.
(■ x + y) n+1 = ( x + y)(x + y) n
= (■ x + y ) E C(n,k)x n k y k by induction hypothesis
k= 0
n n
= x^C(n, k)x n ~ k y k + y^2 C(n , k)x n ~ k y k
k= 0 k= 0
n n
= ^ <?(«, k)x n+1 ~ k y k + C(n, k)x n ~ k y k+1
k= 0 k= 0
(now change k to k — 1 in the second term)
n n+l
= J2 C ( n ’ k)x n+1 ~ k y k + ^2 <?+ k - 1 ) x ”-( fc - 1 )y fc
fc=0 fc=l
n+l
= fe) + C(n,k — l])x n+l ~ k y k
k= 0
n+l
= y^C(n + l,fc)g w+1 ~ fc y fc by Theorem 125.5
fc=0
Note that I changed the limits on the sum in the next to last line of this proof,
using
the facts that C(n,n + 1) = 0 and C(n, — 1) = 0.
There is a sense in which this proof forces you to believe Theorem 125.3, but
the earlier proof (on page 191) explains why it is true. Mathematicians sometimes
call a proof like the earlier one a conceptual proof.
The following theorem gives an explicit formula for the binomial coefficient.
125.6 Theorem
For 0 <k <n,
77 1
c{n ’ k) = W^W. <125 ' 6)
The proof is omitted.
125.6.1 Worked Exercise Find the number of strings of length n in {a,b,c}*
that contain exactly two a’s.
Answer Now that we have the function C(n,r) we can solve this using the idea of
Worked Exercise 114.2.2. To construct such a string, we must choose two locations
in the string where the two a’s will be. There are C(n, 2) ways of doing this. Then
there are two choices (b or c) for each of the other locations, so the answer is
C(n,2) -2 n ^ 2 , which by Theorem 125.6 is
n ( n ~ 1) on- 2
2
conceptual proof 193
theorem 2
194
binomial coeffi-
cient 191
identity (predi-
cate) 19
recurrence rela-
tion 161
recurrence 161
string 93, 167
125.6.2 Proving identities for the binomial coefficient An enormous num-
ber of identities are known for the binomial coefficient. We consider one here to
illustrate how one goes about proving such identities. The identity is
^2C(n,k) 2 = C(2n,n) (125.7)
k = 0
This can be proved using the recurrence relation of Theorem 125.5, but the
proof is rather tedious. I quail with terror at the idea of using the formula in
Theorem 125.6 to prove this theorem.
It is much easier to use Definition 125.1. C(2n,n) is the number of ways of
choosing n balls from a set of 2 n balls. Now suppose that we have 2 n balls and
n of them are red and n of them are green. Then an alternative way of looking at
the task of choosing n balls from this set is that we must choose k red balls and
n — k green balls for some integer k such that 0 < k < n . For a particular k there
are C(n,k)C(n,n — k) ways of doing this. By Theorem 125.2(c), this is the same as
C{n,k) 2 . Altogether this alternative method of choosing a n -element subset gives
5 Z C (n,k ) 2
k = 0
possibilities.
125.6.3 Remark Like most concepts in mathematics, C{n,k ) has a conceptual
definition, namely Definition 125.1, and a method of calculating it, in this case
two
of them: Theorems 125.5 and 125.6. It is generally good advice to try the
conceptual
approach first.
In this case there is a second conceptual description, as coefficients in a polyno-
mial (Formula 125.1), and in fact that formula allows a faily easy second proof of
Formula (125.7).
125.6.4 Exercise Prove that Ylk= oC(n,k) = 2 n .
125.6.5 Exercise Prove that ^fe=o( — 7) k C(n,k) = 0 for n > 0.
125.6.6 Exercise Prove two ways that for all n > 4,
71 — 2
C(n,3) = — -C(n,2)
a) Prove it by using the definition of C(n,k).
b) Prove it using formula (125.6).
125.6.7 Exercise Prove Theorem 125.6. (It can be done by induction, but is a
bit complicated.)
125.6.8 Exercise Prove Formula (125.7) using Formula (125.1).
125.6.9 Exercise Let F(n,k ) be the number of strings of length n in {a,b,c}*
with exactly k b’ s. Find a formula or recurrence relation for F(n,k).
195
125.6.10 Exercise Derive a formula or recurrence relation for the number of
strings of length n in {a, b}* with the same number of a’s as b 's.
125.6.11 Exercise (hard) Find a recurrence relation for the number of partitions
of an n-element set that have exactly k blocks.
125.6.12 Exercise (hard) Prove formula (125.6).
126 . Composition of relations
126.1 Definition: composition of relations
Let a be a relation from A to B and (3 be a relation from B to C .
The composite a° (3 is a relation from A to C, defined this way: For
all a € A and c € C ,
a(a ° (3)c 36 £ B (o a b A b /3 c)
126.1.1 Example Let A = {1,2, 3, 4, 5}, B = {3,5, 7, 9} and C = { 1,2, 3,4, 5, 6},
with
a= | <1) 3), (1,5), (2,7), (3,5), (3,9), (5,7) j
and
P = {<3, 1), (3,2), (3,3), (7,4), (9,4), (9,5), (9,6)}
Then
ao/3= {(1,1), (1,2), (1,3), (2, 4), (3, 4), (3, 5), (3, 6), (5, 4)}
126.1.2 Usage As you can see, although functions are composed from right to
left, relations are composed from left to right. It is not hard to see that if F :
A — > B
and G : B — >■ C are functions, then
T(G°F) = r(F)°T(G)
126.1.3 Exercise Let A = {2,3, 4, 5}, B = {6, 7, 8, 9}, C = {a,b, c,d, e} , and a€
Rel(A,L>), p € Rel(R,C) be defined as follows. Give the ordered pairs in a°(3.
a) a is “divides”, /3 is |(6,a), (6,c), (7, 6), (9,d)| .
b) a is “divides”, (3 is |(7,a), (7,6), (7,c)| .
c) a ={(2,7), (2, 8), (3, 7), (3, 9), (4, 8), (4, 9)} and
P= |(6,a),(6,6),(7,c),(8,c),(9,c),(9,d),(9,e)}
(Answer on page 250.)
block 180
definition 4
divide 4
equivalent 40
function 56
ordered pair 49
partition 180
recurrence 161
relation 73
string 93, 167
theorem 2
usage 2
196
associative 70
composite (of rela-
tions) 195
composition pow-
ers 196
definition 4
functional relation
include 43
interpolative 196
proof 4
relation 73
transitive 80, 227
126.2 Theorem
Composition of relations is associative: if a G Rel(A,R), f3 G Rel(S,C),
and 7 G Rel(C, D) , then
(a ° (3) ° 7 = a ° ((3 ° 7 ) G Rel(A, D )
75 Proof Left as Problem 126.3.4.
126.3 Definition: composition powers
The composition powers of a relation a on a set A are o° = Aa (the
equals relation), a 1 = a, of = a° a, and in general a n = a 0 a” -1 .
126.3.1 Exercise For each relation R in Exercise 52.1.3, page 75, determine
whether 1R 2 3, 1R 3 3, and 3R 2 1. (Answer on page 250.)
126.3.2 Exercise Prove that if F:A—*B and G : B — >• C are functions, then
T(G°F) = T(F)°T(G).
126.3.3 Exercise Let A = {1,2, 3, 4}.
a) Construct a nonempty relation a on A for which a 2 is empty.
b) Construct a relation a f A x A on A for which a 2 = A x A.
126.3.4 Exercise Prove that composition of relations is associative.
126.3.5 Exercise Show that the composite of functional relations is a functional
relation.
126.3.6 Exercise Let a be a relation on a set A. Prove that a is transitive if
and only if a 0 a C a .
126.3.7 Exercise A relation a on a set A is interpolative if a C a°a. Show
that <, as a relation on R, is interpolative, but as a relation on Z, it is not
interpolative.
197
127. Closures
Given any relation a on S, and any property P that a relation can have there
may be a “smallest” relation with property P containing a as a subset. It may not
exist, but if it does, it is called the P-closure of a. Here is the formal
definition.
127.1 Definition: closure
A relation f3 on A is the P-closure of a if
C.l (3 has property P.
C.2 a.C/3.
C.3 If 7 has property P and a C 7 , then /? C 7 .
127.1.1 How to think of closures ft is the “smallest” (in the sense of inclusion)
relation with property P containing a as a subset.
127.1.2 Fact The reflexive, symmetric, and transitive closures of relations always
exist. We will look at each of these in turn. The antisymmetric closure of a
relation
need not exist (Problem 128.2.5).
127.2 Theorem
The reflexive closure of a relation a is a U A g . It is denoted by a R .
Proof To prove this formally you must show that it fits Definition 127.1; that is,
that
RC.l aUAs is reflexive,
RC.2 a C a U As, and
RC.3 if 7 is a reflexive relation and a C 7 , then aU As C 7 .
RC.l and RC.2 are obvious. As for RC.3, suppose that a C 7 and 7 is reflexive.
If i(aUAs)z then either x a z or x = z (that is, rAsz). In the first case ary z
because a C 7 , and in the second case, ary z because 7 is reflexive. Thus
x(aU Ag)y =$■ ary y
so a U As C 7 , as required.
127.2.1 Exercise What is the reflexive closure of the relation “<” on R?
(Answer on page 250.)
127.3 Theorem
The symmetric closure of a relation a is
a s = a U a op
127.3.1 Exercise What is the symmetric closure of “<” on R? (Answer on
page 250.)
127.3.2 Exercise What is the symmetric closure of “<” on R?
127.3.3 Exercise Give an example of a relation whose symmetric closure has
exactly three elements.
definition 4
fact 1
implication 35, 36
include 43
P-closure 197
proof 4
reflexive 77
relation 73
subset 43
symmetric 78, 232
theorem 2
union 47
198
family of sets 171
include 43
integer 3
intersection 47
ordered pair 49
positive integer 3
proof 4
relation 73
theorem 2
transitive 80, 227
union 47
127.3.4 Exercise Show that the symmetric closure of a relation a is a Ua op .
(Answer on page 250.)
The most important type of closure in practice is the transitive closure:
127.4 Theorem
Let a be a relation on a set S . The transitive closure a T of a is
U^ =1 a fc , where a k = a° a ° ...° a (k times), the composition power.
Proof Let f3 = U 7 £ =1 a k . Any member of a family of sets is enclosed in the
union
of the family, so a C (3 . This verifies C.2 of Definition 127.1. As for C.3,
suppose
7 is transitive and a C 7 . Then a k C 7 (Exercise 127.4.2), so (3 C 7 because any
ordered pair in (3 is in at least one of the sets a k .
Finally, we must show that (3 is transitive. Suppose x(3z and z(3y. Then for
some integers k and m, xa k z and za m y. Then it is easy to see that xa k+m y, so
x(3y as required.
127.4.1 Exercise What is the transitive closure of the relation a on Z defined
by xay if and only if y = x + 1 ?
127.4.2 Exercise Suppose 7 is transitive and a C 7 . Show that a k C 7 for all
positive integers k.
a U A g is the only reflexive closure of a . That is why we could use the notation
a R — it means only one thing. It is always true that if a relation has a P-
closure,
it has only one:
127.5 Theorem
Let P be a property of relations, and suppose (3 and (3 ' are P-closures
of a relation a on a set S . Then (3 = f3' .
Proof By C.2 of Definition 127.1, a C. (3 and a C (3' . Then by C.3, f3 C f3' and
f3' C (3 . Thus (3 = (3' .
128. Closures as intersections
The following set-theoretic description of P-closures is useful. It does not make
the P-closure easy to calculate, but it does give a conceptual description useful
for
proving properties of closures.
199
128.1 Definition: intersection-closed
A property P of relations on a set A is intersect ion- closed if:
IC.l A x A has property P.
IC.2 For any set S of relations on A, all of which have property P, the
intersection of all the relations in S also has property P.
128.1.1 Remark The set Ax A can be regarded as the intersection of the empty
family of relations on A. The reasoning is this: In the case of relations, each
relation
on A is a subset of Ax A, and by Section 112.4 the intersection of the empty family
of relations on A is Ax A. From this point of view, IC.l is unnecessary.
definition 4
empty set 33
family of sets 171
include 43
intersection-
closed 199
intersection 47
proof 4
relation 73
subset 43
theorem 2
128.2 Theorem
Let P be an intersection-closed, property of relations. Then for any rela-
tion a, the P-closure of a exists and is the intersection of the set of all
P-closed relations containing a as a subset.
Proof Let (3 be the intersection of all the P-closed relations containing a as a
sub-
set. We must verify C.l, C.2 and C.3. (3 has property P because P is intersection-
closed. a C (3 because a C Ax A and Ax A has property P, and [3 is the inter-
section of all the relations with property P that contain a as a subset. Finally,
the
intersection of a family of sets is included in any member of the family.
128.2.1 Exercise Prove that for any property P, if a has property P then the
P-closure of a is a itself.
128.2.2 Exercise Show that the following hold for any relation a:
a) a RS = a SR .
b) a R1 = a TR .
128.2.3 Exercise
a) Prove that for any relation a , a TS C a ST .
b) Give an example of a relation a for which a TS a ST .
128.2.4 Exercise Let P be the property of a relation [3 that either 1/32 or
2/31. On the set S = {1,2}, let a = {(1,1)}. Let /3 = {(1,1), (1,2)} and 7 =
{(1, 1), (2, 1)} . Then /3 and 7 both include a and both have property P. On the
other hand, a does not have property P. Does this contradict Theorem 127.5?
128.2.5 Exercise Show that a relation need not have an “antisymmetric closure” .
200
definition 4
equivalence rela-
tion 200
equivalence 40
equivalent 40
even 5
natural number 3
nearness relation 77
odd 5
partition 180
predicate 16
proposition 15
reflexive 77
relation 73
symmetric 78, 232
transitive 80, 227
union 47
129. Equivalence relations
If an object a is like an object b in some specified way, then b is like a in that
respect. And surely a is like itself — in every respect! Thus if you want to give
an abstract definition of a type of relation intended to capture the idea of being
alike in some respect, two of the properties you could require are reflexivity and
symmetry. Relations with those two properties are studied in the literature (the
nearness relation J\f in Section 55.1.4 is such a relation), but here we are going
to
require the additional property of transitivity, which roughly speaking forces the
objects to fall into discrete types, making a partition of the set of objects being
studied.
129.1 Definition: equivalence relation
An equivalence relation on a set S' is a reflexive, symmetric, transitive
relation on S .
129.1.1 Remark This is an abstract definition — you don’t have to have some
property or mode of similarity in mind to define an equivalence relation.
129.1.2 Example Let A = {1,2, 3, 4,5,6} . Here is an equivalence relation a on
the set A:
a = «n, n) \ n e A} U {(2, 5) , (5, 2) , (3, 4) , (4,3) , (3, 6) , (6, 3) , (4,
6) , (6, 4)} (129. 1)
129.1.3 Example The relation “equals” on any set is an equivalence relation.
129.1.4 Example The relation “has the same parity as” on the set N of natural
numbers is an equivalence relation. Two numbers have the same parity if they are
both even or both odd.
129.1.5 Example The relation of being in the same suit on a deck of cards is an
equivalence relation.
129.1.6 Example Both the congruence relation and the similarity relation on the
set of triangles are equivalence relations.
129.1.7 Example The relation called equivalence on the set of propositions or
the set of predicates is an equivalence relation. (This example requires that the
set
of propositions or predicates be precisely defined, which is done in formal
treatments
of logic but which has not been done in this text.)
129.2 Exercise set
In questions 129.2.1 through 129.2.9, let E be the relation defined in the question
on Z. Is L an equivalence relation? Explain your answer.
129.2.1 mEn ■&m<n (Answer on page 250.)
129.2.2 mEn 44 m 2 = n (Answer on page 250.)
201
129.2.3 mEw 44 m = n+ 1 V n = m + 1 (Answer on page 250.)
129.2.4 mEn 44 2 | m — n V 3 | m — n (Answer on page 250.)
129.2.5 mE?r 44 fn 2 = n 2
129.2.6 mEn 44 m \ n A n \ m
129.2.7 mEn 44 \m — ro| < 6.
129.2.8 mEn 44 12 | (m — n+ 1) .
129.2.9 mEn 44 (6 | (m — n) and 8 | (m — n)) .
129.3 Exercise set
In questions 129.3.1 through 129.3.6, let E be the relation defined in the question
on R. Is i? an equivalence relation?
congruent (mod
k) 201
definition 4
divide 4
equivalence rela-
tion 200
equivalent 40
floor 86
integer 3
modulus of congru-
ence 201
positive integer 3
relation 73
remainder 83
union 47
usage 2
129.3.1 rEs 44 r/s = 1 (Answer on page 250.)
129.3.2 rEs 44 floor(?’) = floor(s) . (Answer on page 250.)
129.3.3 rEs 44[r = sV(0<r<lA0<s<l)] (Answer on page 250.)
129.3.4 rEs 44 r + s = 1 .
129.3.5 rE.s 44 r — s G N.
129.3.6 rEs 44 r — s G Z
129.3.7 Exercise If E and F are equivalence relations on a set S, are EP\F
and EU F always equivalence relations?
130. Congruence
130.1 Definition: congruence (mod k )
Let k be a fixed positive integer. Two integers rn and n are congruent
(mod k), written “m = n (mod k ) ” , if k divides m — n, in other words,
if there is an integer q for which m — n = qk.
130.1.1 Example 9 = 3 (mod 6) , —5 = 16 (mod 7), 146 = —22 (mod 12).
130.1.2 Usage
a) In the phrase “m = n (mod k ) n , k is called the modulus of congruence.
b) The syntax for “mod” here is different from that of the operator “MOD” used
in Pascal and other languages. In Pascal, “MOD” is a binary operator like
“ + ”; when used between two variables, as in the phrase “M MOD K” , it causes
the calculation of the remainder when M is divided by K. Thus “5 MOD 3”,
for example, is an expression (not a statement) having value 2. The phrase
“5 = 2 (mod 3)”, on the other hand, is a sentence that is either true or false.
202
divide 4
equivalence rela-
tion 200
hypothesis 36
integer 3
mod 82, 204
positive integer 3
proof 4
quotient (of inte-
gers) 83
remainder 83
theorem 2
transitive 80, 227
130 . 1.3 Exercise List all the positive integers < 100 that are congruent to 3 mod
24. (Answer on page 250.)
130 . 1.4 Exercise List all the positive integers < 100 that are congruent to —3
mod 24.
130 . 1.5 Remark Recall that the remainder when m is divided by k is the unique
integer r with 0 < r < \k\ for which there is an integer q such that m = qk + r.
Then
we can prove:
130.2 Theorem
Two positive integers m and n are congruent mod k if and only if m
and n leave the same remainder when divided by k .
Proof If in = qk + r and n = q'k + r (same r ) , then m — n=(q — q')k, so k divides
m — n. Then by definition m = n (mod k ) .
Conversely, if m = n (mod k ) , let r be the remainder when m is divided by
k and r' the remainder when n is divided by k. Then there are quotients q and
q' for which m = qk + r and n = q'k + r 1 . Then r — r' = (m — qk) — (n — q'k) =
m — n+ ( q' — q)k. Since m — n is divisible by k , this means r — r' is divisible
by k .
Since r and r' are both between 0 and k (not including k ) , this means r = r' , as
required.
130.3 Theorem
Congruence ( modk ) is an equivalence relation.
Proof Here is the proof that it is transitive; the rest is left to you. Suppose
that
m = n (mod k) and n = p (mod k ) . Then m leaves the same remainder as n when
divided by k, and n leaves the same remainder as p when divided by k. Since
remainders are unique, m leaves the same remainder as p when divided by k, so,
by Theorem 130.2 m = p (mod k ) .
Congruence has an important special property connected with addition and multi-
plication that has given it extensive applications in computer science:
130.4 Theorem
If m = m' (modk) and n = n' (modk) then m + n = m' + n' (modk)
and mn = m!n! (mod k) .
Proof The hypothesis translates into the statement
k | m — in and k\n — n
Then (in + n) — (m' + n') = m — m' + n — n! is the sum of two numbers divisible
by k, so is divisible by k. Hence m + n = m' + n' (mod A;). Also mn — m'n l =
mn — mn' + mn' — m'n' = m(n — n') + n'(m — m') , again the sum of two numbers
divisible by k, so that mn = m'n' (mod k) .
203
130.4.1 Remark The consequence of Theorem 130.4 is that if you have an expres-
sion involving integers, addition and multiplication, you can freely substitute
inte-
gers congruent to the integers you replace and the expression will evaluate to an
integer that, although it may be different, will be congruent (mod k) to the
original
value.
130.4.2 Example As an example, what is 5 8 congruent to (mod 16)? The arith-
metic is much simplified if you reduce each time you multiply by 5 :
5=5 (mod 16)
5 2 = 25 = 9 (mod 16)
5 3 = 5-9 = 45 = 13 (mod 16) (130.1)
5 4 = 5-13 = 65 = 1 (mod 16)
5 8 = (5 4 ) 2 = l 2 = 1 (mod 16)
definition 4
divide 4
domain 56
equivalence rela-
tion 200
equivalent 40
fact 1
function 56
integer 3
kernel equiva-
lence 203
relation 73
remainder func-
tion 203
remainder 83
130.4.3 Remark This ability to compute powers fast is the basis of an important
technique in cryptography.
130.4.4 Exercise Compute:
a) 5 12 (mod 4)
b) 5 12 (mod 10)
c) 5 12 (mod 16)
(Answer on page 250.)
130.4.5 Exercise Prove that if s\t, then
ms = ns (mod t) 44 rri = n (mod t/s)
131. The kernel equivalence of a function
If F : A—> B is a function, it induces an equivalence relation K(F ) on its domain
A by identifying elements that go to the same thing in B . Formally:
131.1 Definition: kernel equivalence
If F : A — >■ B is a function, the kernel equivalence of P on A, denoted
K ( F ) , is defined by
aK(F)a! 44 F(a) = F(a')
131.1.1 Fact It is easy to see that the kernel equivalence of a function is an
equivalence relation.
131.1.2 Example The congruence relations described in the preceding section
are kernel equivalences. Let A: be a fixed integer > 2. The remainder function
F : Z — > Z is defined by F(n) = n (mod k) , the remainder when n is divided by k.
Theorem 130.2, reworded, says exactly that the relation of congruence (mod k ) is
the kernel equivalence of the remainder function.
204
block 180
definition 4
division 4
empty set 33
equivalence class 204
equivalence rela-
tion 200
fact 1
include 43
mod 82, 204
partition 180
proof 4
quotient set (of
an equivalence
relation) 204
remainder 83
subset 43
symmetric 78, 232
theorem 2
transitive 80, 227
131.1.3 Exercise Give an example of a function F: N — >• N with the property
that 3/1 (F)5 but ~‘(3K(F)6. (Answer on page 250.)
132. Equivalence relations and partitions
132.0.4 Discussion If an equivalence relation E is given on a set S , the elements
of S can be collected together into subsets, with two elements in the same subset
if they are related by E. This collection of subsets of S is a set denoted S/E ,
the
quotient set of S by E. Here is the formal definition of S/E:
132.1 Definition: quotient set of
an equivalence relation
Let E be an equivalence relation on a set S . For each x G S , the equiva-
lence class of x mod E, denoted [x]e, is the subset {y G S \ yEx} of
S. The quotient set (of an equivalence relation) S/E of E is the
set {[x\e | x £ S}.
132.1.1 Example The quotient set of the equivalence relation a defined in 129.1
above is {{1}, {2, 5}, {3, 4, 6}}, which is a partition.
132.1.2 Example The quotient set of congruence (mod 6) is the partition of Z
by remainders upon division by 6 . The quotient set is always a partition:
132.2 Theorem
If S is a set and E is an equivalence relation on S , then the quotient
set S/E is a partition of S .
Proof To see why S/E is a partition, we have to see why
a) every element of S is in an equivalence class in S/E,
b) no element of S is in two equivalence classes in S/E, and
c) S/E does not contain the empty set as an element.
(This just spells out the definition of partition.)
Part (a) is easy: if x G S then, by reflexivity, xEx , so [x\e-
Part (c) is similar: by definition of S/E, an element of S/E is an equivalence
class [x\e for some iGS; since x G [x\e , [x\e is not empty.
As for (b) , x G [x] e ; if also x G [y\E for some y G S , then we have to show
that
[v\e = [x\e- To do this, we have to show two things:
(i) [v\e Q [x\ E , and
(h) \x\eQ\v\e-
For (i), let z G [v]e- Then zEy by definition. Since x G \v\ei xEy. By symmetry
and transitivity, zEx, so z € [x\e ■ Hence [y]E C [x\e-
For (ii), let z G [x\e- Then zEx . Since x G [v]e, xEy. So by transitivity, zEy .
Hence z G \v\ei as required.
132.2.1 Fact The equivalence class [x\e is a block of the partition S/E.
205
132.2.2 Worked Exercise Let S = {1,2, 3,4,5} . Find S/E if
E = A5 U {(1,3), (3, 1), (3,4), (4,3), (1,4), (4, 1)}
Answer
{{1,3, 4}, {2}, {5}}
132.2.3 Exercise Let S = {1,2, 3, 4,5,6} . Find S/E if
E = As U {<1,3), (3, 1), <3, 4), <4, 3), <1 , 4) , <4, 1), <2, 5), <5, 2)}
132.2.4 Exercise Let S = {1,2, 3, 4, 5}. Find two different equivalence relations
E and E' with the property that the subset {1,2} is an element of both S/E and
S/E'. (Answer on page 250.)
132.2.5 Exercise Give an example of an equivalence relation E on the set R
with the property that
{x £ R | 0 < x < 1}
block 180
equivalence rela-
tion 200
equivalent 40
identifies 205
partition 180
proof 4
quotient set (of
an equivalence
relation) 204
reflexive 77
relation 73
symmetric 78, 232
theorem 2
transitive 80, 227
union 47
is one of the equivalence classes of E.
132.2.6 Exercise Let S = {1,2, 3, 4, 5}. Find two different equivalence relations
E and E' on S with the property that S/EnS/E' = {{1,5}, {3}}
132.2.7 How to think of equivalence relations If E is an equivalence relation
on S , the quotient set S/E is often thought of as obtained by merging equivalent
elements of S . One often says that one identifies equivalent elements. Here,
“iden-
tify” means “make identical” rather than “discover the identity of” .
Mathematicians
informally will say we glue equivalent elements together.
133. Partitions give equivalence relations
For a partition n of a set S , we will use the notation [x]n or just [x] if the
context
makes clear which partition is being used, to denote the (unique) block of n that
has x as an element. Given a partition n, you get an equivalence relation En by
the definition:
xE n y (x £ [y]n)
(133.1)
133.1 Theorem
If n is a partition of a set S , then the relation En defined by (133.1)
is an equivalence relation.
Proof To see that xEnx requires x £ [x], which is true by definition of [x] . Hence
En is reflexive. If xEn y then x £ [y] . That means [x] = \y\ , since by definition
of
partition an element is in only one block. Since y £ [y\ by definition and [x] = [y
\ ,
we know that y £ [x] , so yEnx . Hence En is symmetric. Note that we now know
that xEn y if and only if x and y are in the same block of n . To prove
transitivity,
206
antisymmetric 79
bijection 136
block 180
definition 4
domain 56
equivalence rela-
tion 200
function 56
include 43
inverse function 146
irreflexive 81
ordering 206
partition 180
quotient set (of a
function) 184
quotient set (of
an equivalence
relation) 204
reflexive 77
relation 73
strict ordering 206
subset 43
transitive 80, 227
weak ordering 206
suppose xEn y and yE-^z. Then x and y are in the same block, and y and z are
in the same block, so [x] = [y\ = [z ] . This means xEnz , so En is transitive.
133.2 The fundamental theorem on equivalence relations
We gave two constructions in the preceding sections. Given an equivalence relation
E , in Definition 132.1 we constructed a partition S/E, and given a partition II,
in
Section 133 we constructed an equivalence relation En .
If we let 7 tS denote the set of partitions of S (this is standard notation) and
E(S)
denote the set of equivalence relations on S (there is no standard notation for
this),
we now have functions E e-)- S/E : E(S) — > ttS and II >-)• En : nS — > E(S ) ,
where En
is defined in formula (133.1) above. The basic fact about these constructions is
that
these two functions are bijections and each is the inverse of the other. This fact
is
the “fundamental theorem on equivalence relations.”
In other words, if you have an equivalence relation E, construct the quotient
set S/E, which is a partition, and then construct the equivalence relation E s / e
corresponding to that partition, you get the equivalence relation E you started
with. And if you have a partition II of S , construct the corresponding equivalence
relation En, and then construct the quotient set S'/En of E, you get the partition
II back again. The proof of the fundamental theorem involves the same sort of
arguments given earlier, and is left as a problem.
133.2.1 Exercise Prove the fundamental theorem on equivalence relations.
133.2.2 Exercise Prove that any partition of a set A is the quotient of some
function with domain A.
134. Orderings
An ordering is a special sort of relation that is the mathematical formulation of
the concept of comparison or priority. It includes as special cases the relation “
< ”
between numbers and the relation of inclusion between subsets of a set. Here is the
formal definition:
134.1 Definition: ordering
A relation a on a set A is an ordering if it is antisymmetric and
transitive. If it is also reflexive, it is a weak ordering, and if it is also
irreflexive, it is a strict ordering.
134.1.1 Example The relation “<” on a set of numbers is a weak ordering, and
“<” is a strict ordering.
134.1.2 Example An example of an ordering a on a set S that is neither weak
nor strict is the relation
{(1,1), (1,2), (2,3), (1,3)}
on the set {1,2,3}. It is not reflexive because 2 is not related to itself, but it
is not
irreflexive because 1 is related to itself.
207
134.1.3 Remark Essentially all the orderings considered in this text are either
weak orderings or strict orderings, but the more general concept is occasionally
useful.
134.2 Definition: ordered set
If a is an ordering on A, then (A, a) is an ordered set. If a is a weak
ordering, (A, a) is a poset.
134.2.1 Example (R, <) and (R, >) are posets, and so is (' VA,C. ) for any set
A. The set of all relations on a set S is ordered by inclusion; it is the poset
(V(SxS),Q.
134.2.2 Usage In many texts, a weak ordering is called a partial ordering, and
“poset” is short for “partially ordered set” .
antisymmetric 79
definition 4
divide 4
include 43
integer 3
ordered set 207
partial ordering 207
poset 207
positive integer 3
powerset 46
reflexive 77
relation 73
theorem 2
transitive 80, 227
usage 2
134.2.3 Example Not only are “<” and “<” orderings on R, but so are “>”
and “>”.
134.2.4 Example The relation m \ n on N is a weak ordering; thus (N, |) is a
poset. Reflexivity is the obvious fact that n \ n for any n £ N, transitivity
requires
proving that if m\n and n \ p then rn \ p . and antisymmetry is the almost obvious
fact that if m\n and n \ rn then m = n.
I will prove antisymmetry and leave the others to you. By definition, m | n
means that n = hm for some positive integer h . Likewise n \ m means that m = kn
for some positive integer k . Thus m = kn = khm . If m / 0 you can cancel m and
get kh = 1 . Since k and h are positive integers, that means k = h = 1 . Hence
m = n. As for the case m = 0, the fact that n = hm means n = 0, so m = n again.
134.2.5 Example If you have a collection T of tasks, there is a natural ordering
of T defined this way: t a u if task t must be done before task u can be started.
This is obviously transitive. If a were not antisymmetric, that would say there are
two different tasks t and u, each of which had to be done before the other, so that
it
is in fact impossible to perform the set of tasks. Thus for any reasonable
collection
f of tasks, (T, a) is antisymmetric as well as transitive and therefore an
ordering.
134.3 Theorem
Let a be an ordering. Then a op (see Section 54-2, page 11) is also an
ordering. Moreover, a op is strict if a is strict and weak if a is weak.
134.3.1 How to think of orderings If a is an ordering on a set S and a a b,
one says that “a is smaller than 6” . This phraseology has to be used with caution
—
one would not use it, for example, for the relation “>” on R. More subtle problems
with this terminology arise with other orderings. For example, in the poset (N, |),
3 is smaller than 6 but 3 is not smaller than 5 . Nor, for that matter, is 5
smaller
than 3. You have to be very clear that “smaller” here is not the usual relation “<”
on N.
208
definition 4
divide 4
include 43
linear ordering 208
powerset 46
reflexive 77
relation 73
strict total order-
ing 208
theorem 2
total ordering 208
transitive 80, 227
trichotomy 208
usage 2
The following Theorem, whose proof is left to you, shows that a relationship analo-
gous to that between “<” and “<” holds for all orderings.
134.4 Theorem
For any ordering a on a set S , a — As is a strict ordering of S and
the reflexive closure a R is a weak ordering.
135. Total orderings
135.1 Definition: total ordering
An ordering a on a set A with the property that for any pair of elements
a, b E A , either a a b or b a a , is a total ordering.
135.1.1 Usage A total ordering is also called a linear ordering.
135.1.2 Example The relations “<” and “>” are total orderings on R, as well
as other sets of numbers.
135.1.3 Example The ordered set (N, |) is not totally ordered: as we observed
previously, 3 and 5 are not related to (do not divide) each other.
135.1.4 Example If A has more than one element, then (VA, C) is not a totally
ordered set.
135.2 Theorem
A total ordering is reflexive, in other words is a weak ordering.
135.2.1 Exercise Prove Theorem 135.2.
135.2.2 Usage In most writing in pure mathematics, a total ordering is a type of
strict ordering, defined axiomatically in Definition 135.3 below. We call it
“strict
total ordering” here.
135.3 Definition: strict total ordering
A relation cc on a set S' is a strict total ordering if it is transitive
and satisfies trichotomy: For all a, 6 £ S , exactly one of the following
statements hold:
(i) a a b
(ii) baa
(iii) a = b.
135.3.1 Remark This definition has the consequence that a strict total ordering
is not a total ordering in the sense of Definition 135.1. However, it is
straightforward
to prove that if a is a strict total ordering then a R is a total ordering in the
sense
of Definition 135.1.
209
The relation “divides” on Z is not an ordering because it is not antisymmetric.
For example, 6 | — 6 and — 6 | 6 but 67 ^— 6 . “Divides” is, however, reflexive and
transitive on Z .
135.3.2 Exercise Let a be a relation on a set A. Prove that if a is a strict total
ordering in the sense of Definition 135.3, then a is a strict ordering. (Answer on
page 250.)
135.3.3 Exercise Let ct be a relation on a set A.
a) Assume that a is a strict total ordering in the sense of Definition 135.3. Prove
that a R is a total ordering in the sense of Definition 135.1.
b) Prove that if a is a total ordering then a — A a is a strict total ordering.
135.3.4 Exercise How many total orderings of an n -element set are there? Prove
your answer correct.
135.3.5 Exercise For any natural number n, let D(n ) denote the set of positive
divisors of N. Thus D( 6 ) = {1,2, 3, 6 }. Show that (D(n),\) is totally ordered if
and only if n is a power of a prime.
136. Preorders
136.1 Definition: preordering
A reflexive, transitive relation a on a set A is called a preorder or
preordering on A, and (A, a) is a preordered set.
antisymmetric 79
definition 4
divide 4
divisor 5
equivalence rela-
tion 200
equivalent 40
function 56
natural number 3
positive integer 3
preordered set 209
preordering 209
preorder 209
prime 10
quotient set (of a
function) 184
reflexive closure 197
reflexive 77
relation 73
strict ordering 206
strict total order-
ing 208
total ordering 208
transitive 80, 227
usage 2
136.1.1 Usage Sometimes “quasi-ordering” is used for “preordering”, but that
word is used with other meanings, too.
136.1.2 Remark Every preorder can be converted into a partial order by a pro-
cess resembling the construction of the quotient of a function. This process is
explored in exercises below.
136.1.3 Exercise (hard) Let a be a preorder on a set S .
a) Prove that the relation E defined by
xEy 44 {xay A yax)
is an equivalence relation.
b) Define a relation A on S/E by
[x]A[y] 44 xay
Prove that A is well-defined, that is, that if [x\ = [V] , [y] = \y'], and [x]A[y],
then [x']A[y'].
c) Prove that A is an ordering
210
divide 4
division 4
divisor 5
Hasse diagram 210
include 43
ordering 206
poset 207
positive integer 3
relation 73
subset 43
total ordering 208
transitive 80, 227
weak ordering 206
137. Hasse diagrams
Exhibiting an ordering using a digraph as in Section 51.2 tends to be messy-
looking
because transitivity causes lots of arrows to exist. Orderings are normally
illustrated
using a different sort of picture called a Hasse diagram. The elements of the set
are represented as dots, as before, and the diagram is drawn so that when there is
a rising line from a to b, then a a b. (“Rising” means toward the top of the page.)
The rising line from a to b does not have to go directly from a to b, but may pass
through other nodes; this makes use of the fact that the relation is transitive.
Note
that the diagram does not show whether a node is related to itself. In this text,
Hasse diagrams are used only for weak orderings.
(137.1)
137.1.1 Example The two Hasse diagrams in Figure 137.1 show the inclusion
relation on the set of subsets of {1,2,3} and the relation of division on the set
{1,2, 3, 4, 5, 6}.
137.1.2 Remark Note that b can be higher on the page than a without it being
true that a a b — there must be a rising line from a to b to make a a b. For
example, in the right diagram, 5 is not less than 6.
137.1.3 Exercise Draw the Hasse diagram of the indicated poset (A, a):
a) A = {1,2, 3, 4, 5} ,
a = {(1,1), <2, 2), <3, 3), <4, 4), <5, 5), (1,2), (2, 3), (1,3), (5, 4), (4, 3),
(5, 3)}
b) A = {0,{1},{2},{1,2},{2,3}}, a is inclusion.
c) A =set of positive divisors of 20, a is divisibility.
d) A =set of positive divisors of 25, a is divisibility.
(Answer on page 250.)
137.1.4 Exercise Which of the posets in Exercise 137.1.3 are total orderings?
(Answer on page 251.)
137.1.5 Exercise Draw the Hasse diagram for the relation “divides” on:
1. The set of positive divisors of 12.
2. The set {n G N | 1 < n < 12} .
211
138. Lexical ordering
A finite totally ordered set A used as an alphabet induces a total order on the
strings in A* called the lexical order on A* . When A is the English alphabet,
the result is the familiar alphabetical ordering of strings.
To define lexical ordering, we need a preliminary idea.
138.1 Definition: initial segment
A string u is an initial segment of a string w if w = ux for some string
x in A* .
alphabet 93, 167
definition 4
finite 173
infinite 174
initial segment 211
lexical ordering 211
lexical order 211
string 93, 167
total ordering 208
138.1.1 Example ‘ ab’ is an initial segment of ‘abbac’.
138.1.2 Example Any string is an initial segment of itself (since A £ A*) .
138.1.3 Example A is an initial segment of any string.
138.2 Definition: lexical order
Let (A, a) be a finite totally ordered set. Then the lexical order or
lexical ordering A on A* is defined as follows: wXx if either
LE.l w is an initial segment of x, or
LE.2 If i is the first position where w and x differ, then Wi a x* .
138.2.1 Example If A is the English alphabet with the usual ordering, ‘car’
comes before ‘card’ in alphabetical ordering because ‘car’ is an initial segment of
‘card’, and ‘car’ comes before ‘cat’ because the first place where ‘car’ and ‘cat’
differ
is the third place, and ‘r’ comes before ‘t’.
138.2.2 Example If A is nonempty, A* is an infinite set. Consider the lexical
ordering on {0,1}*, where {0,1} is ordered so that 0 comes first. The first few
elements of {0,1}* are A, ‘O’, ‘00’, ‘000’, ‘0000’, ‘00000’, ... Thus if you go
through
the strings in order, there are strings such as ‘1’ that you can’t get to in a
finite
amount of time: there are an infinite number of strings in {0,1}* before ‘1’.
138.2.3 Exercise Prove that the lexical ordering on {0, 1}* (with 0 < 1 ) is a
total
ordering.
212
alphabet 93, 167
base 94
canonical order-
ing 212
definition 4
fact 1
finite 173
include 43
integer 3
lexical ordering 211
string 93, 167
total ordering 208
upper bound 212
139. Canonical ordering
The canonical ordering, defined below, is often used on infinite sets of strings to
remedy the problem described in Example 138.2.2. It is the most commonly used
ordering on {0,1}*.
139.1 Definition: canonical ordering
The canonical ordering on {0,1}*, usually denoted “<”, is defined
this way: w < x if
a) w is shorter than x (|w| < |x|) or
b) |ir| = | a: | and the integer represented by w in binary notation
is less than or equal to the integer represented by x in binary
notation.
139.1.1 Example 1110 comes before 00001 because it is shorter, and 0011 comes
before 0101 because 0011 is 3 in binary and 0101 is 5.
139.1.2 Example In the canonical ordering of {0,1}*, the first few strings are
A, 0, 00, 01, 10, 11, 000, 001, 010, 011, 100, ...
139.1.3 Fact The canonical ordering is linear and, unlike the lexical ordering,
there are only a finite number of strings between any two strings.
139.1.4 Remark This idea can obviously be extended to strings in the alphabet
{0, 1, . . . ,n} where n is a small integer (use base n + 1).
139.1.5 Exercise List the elements of the set
A = {00,01,110, 111,0101,0111,10101,10111,01111}
in the lexical ordering and in the canonical ordering. (Answer on page 251.)
139.1.6 Exercise Prove that the canonical ordering on {0, 1}* is a total ordering,
and that there are only a finite number of strings between any two given strings.
140. Upper and lower bounds
140.1 Definition: upper bound
If (A, a) is a poset and RCA, an element a € A is an upper bound
of B in (A, a) if baa for every 6 g5.
140.1.1 Remark Note that the upper bound a of Definition 140.1 need not be
in B.
140.1.2 Example In the right poset in Figure 137.1, 6 is an upper bound (in fact
the only one) of {1,2,3} and the set {1,2, 3, 4} has no upper bound.
213
140.1.3 Example {1,2, 3, 4} has many upper bounds in the poset (N, |) , for
example 12, 24 and 144.
140.1.4 Remark A lower bound of a subset is defined in the analogous way: a
is a lower bound of B if a a b for all 6 G B .
140.2 Definition: maximum
Let A be a poset and B a subset of A. The maximum of B (plural
“maxima” ) is an element rn of B with the property that for all b G B ,
6am.
140.2.1 Fact The maximum of B, if it exists, is clearly an upper bound of B;
unlike an upper bound, however, it must actually be in B. More is true:
140.3 Theorem
The maximum of a subset B of a poset A, if it exists, is unique.
definition 4
divide 4
fact 1
include 43
least upper
bound 213
lower bound 213
maximum 213
minimum 213
proof 4
rule of inference 24
subset 43
supremum 213
theorem 2
upper bound 212
Proof If m and m! were both maxima of B, then both would be elements of B
and so it would have to be the case that m a m' and m f a m . Then antisymmetry
forces m. = m'.
140.3.1 Remark The minimum of B is an element n of B with nab for all
6 E B . A similar proof shows that a subset B has at most one minimum. Note that
the minimum of B in A is the minimum of B in the opposite poset of A.
140.3.2 Exercise Find all the maxima and minima of the posets in Exer-
cise 137.1.3 of Chapter 134. (Answer on page 251.)
140.3.3 Exercise What are the maxima and minima, if any, of (N, | ) ? Of (N —
{0}, |) ? Of (N — {0,1},|)7 (Answer on page 251.)
141. Suprema
The two ideas of upper bound and minimum combine to form a concept that is more
important than either of them.
141.1 Definition: supremum
Let A be a poset with subset B . An element rn E A is a supremum
of B, or least upper bound of B, if it is the minimum of the set of
upper bounds of B .
141.1.1 Fact The supremum m must be unique if it exists, and it may or may
not be in B. Because of its uniqueness, we denote the supremum of B as supB.
141.1.2 Reformulation of the definition It is worth spelling out the definition
of supremum: If B C A and m £ A , then m . is the supremum of B if m is an upper
bound of B and m a a for every other upper bound a of B. This gives rise to a
rule of inference.
214
definition 4
divide 4
division 4
fact 1
implication 35, 36
infhnum 214
interval 31
join 214
meet 214
ordering 206
positive integer 3
powerset 46
prime 10
rule of inference 24
subset 43
supremum 213
theorem 2
141.2 Theorem
If (A, a) is a poset and B C A, then
(Vb:B)(b am), (\/a:A) ((\/b:B)(b a a) =>■ m a a) (- m = sup.B
141.2.1 Fact Note that m is the “least” upper bound in the sense of the ordering
a: if a is an upper bound of B , then m a a. Specifically, no upper bound can be
unrelated to m.
141.2.2 Example The supremum of {{1},{1,2},{3}} in the set of all subsets of
{1,2,3} is {1,2,3} itself (See Figure 137.1).
141.2.3 Example The supremum in (R, <) of the open interval (0 . . 1) is 1 , which
is also the supremum of the closed interval [0 . . 1] .
141.2.4 Example The set
S' = {xGQ|0<x and x 2 < 2} = {x € Q | 0 < x < y/2}
has no supremum in (Q, <) . That is because if it had a supremum m G Q , m would
have to be its supremum in R, too, but the supremum in R is \/2, which is not
in Q.
141.3 Definition: infimum
The infimum of B, or infR, if it exists, is the unique element n for
which
a) nab for all b £ B , and
b) if a a 6 for all b e B , then a an.
141.3.1 Example In the set {1,2, 3, 4, 5, 6} ordered by division, the supremum of
the subset {2,5} does not exist, and the infimum is 1.
141.3.2 Exercise Find the suprema and infima, if they exist, of the subset S of
the poset [T, a) :
a) S = {3,4,5} , T = N, a is “<”.
b) S = {3,4,5}, T = N, a is “divides”.
c) S is the set of all positive primes, T = N, and a is “<”.
d) S is the set of all positive primes, T = N, a is “divides”.
e) S = {{1,2}, {2, 3}}, T = V{ 1,2,3}, a is inclusion.
(Answer on page 251.)
141.3.3 Least upper bounds of two elements There is a special notation for
suprema and infima of subsets of two elements. If (A, a) is a poset and a,b € A,
then the supremum of {a,b} is denoted a V b and called the join of a and b, and
the infimum is denoted a/\b and called the meet of a and b. Using this notation,
Rule (141.2) then gives this rule of inference:
aac, b a c, ((Vd)(a a d and bad) =>■ c a d) |— c = a V 6
There is a similar rule for a A b .
215
141.3.4 Exercise (hard) Let ( T,a ) be a poset, and suppose A C S CT .
a) Show that if m is the supremum of A in S' and n is the supremum of A in
T , then n <m.
b) Show that if n is the supremum of A in T and n £ S, then n is the supremum
of A in S.
c) Give an example where the situation in (a) holds and m^n.
141.3.5 Exercise (hard) Show that if a and ft are real numbers and
J = {t£Q\a<t<b}
then the supremum of J in Q, if it exists, is 6, so that b is rational. (Hint: Let
n
be the supremum of J in Q. Use Problem 141.3.4 to show that ft < n. Now assume
b < n and use the Archimedean property to get an integer k for which 1 /(n — b ) <
k ,
so that b < n — (l/k) <n and n — (l/k) is rational.)
142. Lattices
142.1 Definition: lattice
A poset (A, a) with the property that for any two elements a and b,
a/\b and a Vft always exist, is called a lattice. If a Aft always exists, but
not necessarily aVft, then (A, a) is called a lower semilattice, and if
oVft always exists but not necessarily a Aft, it is an upper semilattice.
Archimedean prop-
erty 115
definition 4
include 43
integer 3
join 214
lattice 215
lower semilattice 215
max 70
meet 214
minimum 213
min 70
powerset 46
rational 11
real number 12
subset 43
supremum 213
total ordering 208
union 47
unit interval 29
upper semilattice 215
weak ordering 206
142.1.1 Remark Some texts require that a lattice have a minimum and a mini-
mum, as well.
142.1.2 Example The following are Hasse diagrams of lattices. Note that, for
example, in (d) , xA z = b, xV z = t, and x V y = x .
(142.1)
142.1.3 Example In the unit interval I = {r£R|0<r'<l}, the meet r As and
the join r Vs with respect to the usual weak ordering < always exist, and in fact
rAs = min(r, s) and r V s = max(r, s) . Thus (I,<) is a lattice. More generally,
any
total ordering is a lattice (Exercise 142.1.11).
142.1.4 Example Let A be a set and B and C subsets of A. Then in (VA, C),
B A C and BV C always exist and moreover B AC = B nC and B V C = B U C .
Thus (VA, C) is a lattice. (See Exercise 142.1.7.)
216
divide 4
divisor 5
finite 173
GCD 88
include 43
infimum 214
integer 3
lattice 215
lower semilattice 215
minimum 213
natural number 3
positive integer 3
powerset 46
proof 4
relation 73
subset 43
supremum 213
theorem 2
upper semilattice 215
142.1.5 Example Let m and n be natural numbers. Then in (N, |) , m An and
mV n always exist, and moreover m An = GCD(m,ro) and mVn = LCM(m,n).
Thus (N, | ) is a lattice. This follows immediately from Corollary 64.2, page 90.
142.1.6 Exercise Which of these posets are lattices?
a) (N, <) .
b) (Z,<).
c) (R, <)•
d) ( A , |) , where A is the set of positive divisors of 25.
e) ( A , |) , where A is the set of positive divisors of 30.
f) (A, I) , where A = {1,2, 3, 4, 5, 6}.
(Answer on page 251.)
142.1.7 Exercise Prove that for any set A, (PA,C) is a lattice. (Answer on
page 251.)
142.1.8 Exercise Give an example of a lattice in which for some elements a, b
and c, a A (bV c) ^ (a Ab) V (a A c) .
142.1.9 Exercise Show that in the lattice (N — {0}, |) , every subset has an infi-
mum and every finite subset has a supremum, but not every subset has a supremum.
142.1.10 Exercise Let n be a positive integer. Show that the set of positive
divisors of n with “divides” as the relation is a lattice.
142.1.11 Exercise Prove that if ( L,a ) is a lattice, then a is a total ordering if
and only if x V y is the minimum of x and y and x A y is the minimum of x and y .
143. Algebraic properties of lattices
The following theorem gives algebraic properties of meet and join.
143.1 Theorem
If (A, a) is an upper semilattice, then for all a,b,c€ A,
a) aVa = a (idempotence).
b) aVb = bVa (commutativity).
c) aV (bV c) = (aV b)V c (associativity).
Similarly, if (A, a) is a lower semilattice, then for all a, b, c € A ,
a) a Aa = a.
b) aAb=bAa
c) a A (b A c) = (a A b) A c.
Proof We will prove the associativity of A and leave the rest as an exercise. This
proof involves applying the definition of infimum repeatedly to prove that each
side
of the equation is the infimum of the set {a,b,c}, and using the uniqueness of the
infimum. I will show that a A (6 Ac) =inf{a,6, c} and leave the other side to you.
The definition of infimum tells us that all the following are true:
217
( 1 ) b Acab
(2) bAcac
(3) a A (b Ac) a a
( 4 ) aA(bAc)abAc.
Putting (1), (2) and (4) together and using transitivity gives that
(5) aA(bAc)ab
( 6 ) a A(b Ac) ac
(3), (5) and (6) tells us that
( 7 ) a A (b Ac) a inf{a, b , c} .
On the other hand, by definition
(8) inf{a,6,c} a 6
(9) inf {a,b,c}ac
so
( 10 ) inf{a,6,c} a bAc.
Also
( 11 ) inf- {a, b, c} a a
so by (10) and (11),
( 12 ) inf {a, b,c} a a A (b Ac) .
Now (7), (12) and antisymmetry give us the desired result.
143.1.1 Exercise Complete the proof of Theorem 143.1.
143.1.2 Exercise Prove that in a lattice, xay^Ax = xAy^Ay = x\/y.
143.2 The Axiomatic Method
The proof that A and V are associative is rather long, although conceptually not
difficult. The value is that having done it once, we know it is true for every
situation
in which A and V occur.
143.2.1 Example We now know immediately, by examples 142.1.3 through
142.1.5, that max and min, intersection, union, and GCD and LCM are all
idempotent, commutative and associative. It is not hard to prove these directly
(although the proof for GCD and LCM is not trivial), but once we know Theo-
rem 143.1 and the corresponding fact for sups, the associativity doesn’t need
proof.
143.2.2 The idea is that we have extracted salient properties of union,
intersection,
GCD and LCM and made them into axioms', then any theorem derived from those
axioms is true in all the cases all at once. This is an example of the axiomatic
method in mathematics. The axiomatic method is largely responsible for the power
of modern mathematics.
associative 70
axiomatic
method 217
commutative 71
equivalent 40
GCD 88
idempotent 143
intersection 47
max 70
min 70
transitive 80, 227
218
arrow 218
definition 4
digraph 74, 218
directed graph 218
finite 173
function 56
graph 230
infinite 174
node 218, 230
source 218
target 218
144. Directed graphs
144.1 About graphs in general
A graph is a mathematical construction that is used to encode information about
connections between things. There are two main types of graphs, the kind called
“undirected graph” in which only the connection between two things matters, and
the kind called “directed graph” or “digraph” in which the direction of the connec-
tion matters. Each of these main types occurs in numerous subvarieties, only some
of which are commonly used in computer science.
The terminology for different kinds of graphs in the literature is notoriously
varied; it is probably true that if two graph theory books by different authors use
the same terminology, one of the authors was the graduate student of the other one.
The terminology in this text is similar to the usage in many (but not all) computer
science books, but is quite different from that in books written by
combinatorialists
or graph theorists.
In this book, “graph” means undirected graph and “digraph” means directed
graph. All graphs here are finite; although the definitions work for infinite
graphs,
many of the theorems are not true as stated for the infinite case.
144.1.1 Digraphs Informally, a digraph is a bunch of dots called nodes with
arrows going from some nodes to others. Here are two examples.
Here is a more precise definition:
x < ^ y
I b
(144.1)
144.2 Definition: directed graph
A directed graph or digraph G consists of two finite sets Go and G\
and two functions source :Gi — > Go and target \G\ — > Go-
The elements of Go are called the nodes or vertices (singular: vertex)
of G and the elements of Gi are the arrows or directed edges of G.
If an arrow a has source x and target y we write a:x — »• y in the same
way we write functions.
144.2.1 Drawing digraphs A digraph (Go,Gi,s,i) is conventionally drawn
using dots or labels for the nodes, and an (actual) arrow going from node x to node
y for each arrow a (element of Gi) with source x and target y.
144.2.2 Exercise Draw the following digraphs:
a) The graph with nodes {A,B,C, D} and exactly one arrow from each node
to A.
b) G = (Go,Gi,s,t) where Go = {1,2,3}, Gi = {a,b,c, d,e}, s(a) = s(e) = l,
s(b) = s(c ) = s(d) = 2 , t(a) = 2 , t(b) = t(c) = 1 , and t(d) = t(e) = 3 .
219
(Answer on page 251.)
144.2.3 Exercise Draw the graph Go = {2,3,4,5,6,7,8,9,10}, with n arrows
going from r to s if and only if r n \ s and r n+1 does not divide s.
144.3 Definition: abstract description
The information about a digraph given by the definition, that is the
sets Go, Gi and the source and target functions, is called the abstract
description of the digraph.
144.3.1 Remark We will frequently encode the abstract description for a digraph
as an ordered quadruple: thus “G is the digraph (Go,Gi,s,f)” means Go is the set
of nodes, Gi the set of arrows, and s and t are the source and target functions.
144.3.2 Example The abstract description of the digraph on the left of Fig-
ure (144.1) has Go = {x,y,z,w} , G\ = {a,b,c,d,e,u} ,
source(a) = source (6) = source (ci) = target (c) = x
target (a) = target (6) = target (e) = source (it) = target (it) = y
and source(c) = source(e) = z.
144.4 Graphs and abstraction
A digraph is defined here in an abstract way, not as a picture. The interplay
between
the abstract definitions and the pictures is analogous to that between the formula
of a function such as f(x) = x 2 + 1 and its graph (a parabola) in analytic
geometry.
The pictures are more suggestive and comprehensible than the abstract definition,
but it is difficult to prove things using pictures because it is hard to be sure
you
have the most general case. It may also be difficult or wasteful (or both) to store
pictures directly in the computer. The abstract treatment is both more rigorous
and more amenable to computation.
144.5 Digraphs in applications
144.5.1 Example Digraphs provide a natural way to encode data about certain
kinds of complex systems. The flow chart of a program, for example, is a digraph.
The commutative diagrams of sets and functions in Chapter 98 are examples of
labeled digraphs. However, the information concerning the composites of the func-
tions is additional information not encoded by the description of the diagrams as a
digraph.
144.5.2 Example Digraphs are the natural way to model the sequencing of a
collection of tasks that must be performed to accomplish a goal. Each node is a
task and there is an arrow from task a to task b if task a must be completed before
task b can be started. For example, the task of computing log(x 2 + y 3 ) can be
arrow 218
commutative dia-
gram 144
composite (of func-
tions) 140
definition 4
digraph 74, 218
divide 4
function 56
graph 230
labeling 221
node 218, 230
source 218
target 218
220
arrow 218
definition 4
digraph 74, 218
function 56
graph 230
indegree 220
loop 220
node 218, 230
opposite 62, 77, 220
outdegree 220
source 218
modeled this way:
start
calculate x 2
add calculate log
calculate y 3
(144.2)
This graph shows, for example, that if you had two people or two processors to
perform the squaring you could speed up the computation. Digraphs arising in this
way often have a weight function on the arrows.
144.5.3 Exercise Draw the digraph modeling the computation of the truth value
of the equation
2 , 2 2
x + xy = x — y
145. Miscellaneous topics about digraphs
145.1 Definition: loop
An arrow a from a node to itself, in other words a : x — > x for some node
x, is called a loop.
145.1.1 Example u is a loop in the left digraph in Figure (144.1).
145.2 Definition: indegree and outdegree
The number of arrows that have a node as source is called the outdegree
of the node, and the number of arrows that have the node as target is
the indegree.
145.2.1 Example The node y in the left graph of Figure (144.1) has indegree 4
and outdegree 1 .
145.3 Definition: opposite of a graph
The opposite of a digraph G is the digraph with the same nodes and
all the arrows reversed. It is called G op . Thus if G = (Go,Gi,s,t ) , then
G°v = {G 0 ,G 1 ,t,s).
221
145.3.1 Example
The digraphs below are opposites of each other.
A »- B
C D
A* B
C D
145.4 Labeling
A labeling of the nodes of a digraph G is a function L : Gq — >■ S , where S is a
set.
If x is a node, its label is L(x ) . Similarly a function L : G\ — )■ S would label
the
arrows. As an example, the digraph below shows the cost of traveling by rail in a
(mythical) mountainous country between three cities A, B, and C . (The fare for
going to a higher elevation is more than for going to a lower one.)
arrow 218
definition 4
digraph 74, 218
function 56
injective 134
integer 3
labeling 221
node 218, 230
real number 12
weight function 221
(145.1)
The nodes are labeled by {A, B,C} and the arrows are labeled by integers repre-
senting cost. Here the labeling is a function F : G\ — >■ Z. A function labeling
arrows
by integers or real numbers is commonly called a weight function on the arrows.
You can see that the labeling of the nodes is injective but the labeling of the
arrows
is not. When the labeling of the nodes is injective, there is usually no harm in
tak-
ing the attitude that the labels are actually the nodes; a similar remark applies
to
an injective labeling of the arrows.
146. Simple digraphs
146.1 Definition: simple digraph
A digraph is simple if for two distinct arrows a and b, either source(a) ^
source(fe) or target (a) ^ target (6) . In other words, only one arrow can
go from a node to another node. (However, one arrow is allowed each
way.)
146.1.1 Example The left graph in Figure (144.1), page 218, is not a simple
digraph, whereas the right one is.
146.1.2 Exercise What is the largest number of arrows a simple digraph with n
nodes can have?
999
arrow 218
Cartesian product 52
coordinate func-
tion 63
coordinate 49
definition 4
digraph 74, 218
fact 1
include 43
node 218, 230
relational descrip-
tion 222
simple digraph 221
source 218
subset 43
target 218
146.1.3 Variation in terminology In many books the word “digraph” is used
only for simple digraphs; those that allow more than one arrow from a node to a
node are called “multigraphs” or “multidigraphs” .
A simple digraph can be given a much simpler (!) abstract description (of a graph).
Since there can be at most one arrow from a node to another one, all you have to do
to describe the digraph is to give the set Go of nodes and the subset A of Go x Go
of ordered pairs of those nodes that have an arrow going from the first node to the
second one. This is summed up in the following definition.
146.2 Definition: relational description
The relational description of a simple digraph G is (Go, A), where
A C Go x Go is the set of ordered pairs
{(m,n) | There is an arrow from rn to n}
146.2.1 Remark We saw this correspondence between simple digraphs and rela-
tions from the opposite point of view in 51.2.
146.2.2 Example In the case of the right graph in Figure (144.1), which is simple,
G 0 is {x,y,z} and A is {(x,y), (y,x), (x, z)j .
146.2.3 Exercise Which of the digraphs in Exercise 144.2.2 are simple? Give the
relational description of each one that is. (Answer on page 251.)
146.2.4 Exercise Give the relational description of the graph (147.1), page 223.
146.2.5 Fact The relational description can be converted to the original definition
of digraph by calling a pair (x,y) in A an arrow from x to y; thus the source is
the first coordinate and the target is the second.
To sum up:
(i) If (Go,Gi,s,t) is the abstract description (of a graph) of a simple digraph,
you get the relational description (G, A) of the same graph by taking G = Go
and
A = {(x,y) G G 0 x G 0 | (3s)(s :x -» y) in Gi}
(ii) If ( G,A ) is the relational description of a simple digraph, the abstract
descrip-
tion (of a graph) of the same graph is defined to be (Go,Gi,s,f), where
Gq = G , Gi = A , s = pi (the first coordinate function) and t = p 2 ■
223
147. Isomorphisms
The two digraphs below are abstractly identical in a sense that can be made
precise.
The idea is that node a in the left digraph plays the same role as node 2 in the
right digraph, and similarly b and 1 match up and c and 3 match up. “Playing
the same role” means precisely that if you match node x in one digraph to node m
in another, and similarly node y to n, then the arrows from x to y must match
up with the arrows from m to n. (You should check these two digraphs to see that
this happens).
c
This is made precise this way:
147.1 Definition: isomorphism
Let G = (Go,G\,s,t) and G' = {G^G'^s' ,t') be digraphs. An iso-
morphism from G to G’ is a pair of bijections /3q : Gq — > G' 0 and
Pi: G i — >• G\ with the property that a:x — >• y in G if and only if
Pi{a) : Po{x) Po(y) in G' .
147.1.1 Remark Since there is rarely any problem with ambiguity, the subscripts
may be omitted from Pq and Pi .
147.1.2 Example In Figure 147.1 there is an isomorphism P from the left figure
to the right figure defined by
/3(a) = 2 P(f)=v
P{b) = 1 P(g) = u
P(c) = 3 P(h) = w
P(k) = x
The inverse of this isomorphism (meaning (/J^ 1 ,/?^ 1 )) is also an isomorphism;
in
fact the inverse of any digraph isomorphism is also an isomorphism.
147.1.3 Remark It is easily possible for two digraphs to be isomorphic in more
than one way. This happens in Figure 147.1, for example.
147.1.4 Exercise (hard) Show that two digraphs are isomorphic if and only if
there is an ordering of their nodes for which their adjacency matrices are
identical.
147.1.5 Exercise Draw both (nonisonrorphic) simple digraphs that have only one
node, and all ten (nonisonrorphic) simple digraphs that have two nodes.
147.1.6 Exercise Let G = (Go,Gi,s,t) and G' = (G^G^s' ,t') be digraphs.
Prove that Pq\Gq^- G’q and Pi : Gi — > G' x constitute an isomorphism if and only
if
P and P' are bijections and s' ° P\ = Po°s and t! ° P\ = Po°t.
bijection 136
definition 4
digraph 74, 218
inverse function 146
isomorphism 223,
235
node 218, 230
224
adjacency
matrix 224, 232
automorphism 224
Cartesian product 52
definition 4
digraph 74, 218
identity function 63
integer 3
node 218, 230
nonnegative integer 3
positive integer 3
147.1.7 Exercise Let /? = ((3® : Go -4- G' 0 , : G\ — > G' x ) be a digraph
isomorph-
ism from G= (Go,Gi,s,t) to G' = {G' 0 ,G \ , s' ,t') . Show that (3~ l , i.e. , (Pq
1 ,^ 1 ),
is a digraph isomorphism from G' to G .
147.2 Definition: automorphism
An isomorphism P : G — > G of a digraph with itself is called an auto-
morphism.
147.2.1 Example For any digraph, the identity function is an automorphism.
The digraphs in Figure 147.1 each have two automorphisms, the identity and one
other.
147.2.2 Exercise Find the automorphisms of the digraphs in exercise 144.2.2.
(Answer on page 251.)
147.2.3 Exercise (hard) Let G be a digraph with exactly n automorphisms, and
let G' be a digraph isomorphic to G . Show that there are exactly n isomorphisms
from G to G'.
147.2.4 Exercise (hard) For any positive integer n, show how to construct a
digraph with exactly n automorphisms.
148. The adjacency matrix of a digraph
A convenient way for representing a digraph G in a computer program is by means
of its adjacency matrix.
148.1 Definition: adjacency matrix
The adjacency matrix of a digraph G is a matrix of nonnegative
integers whose entries are indexed by Go x Go and whose entry in the
location indexed by the pair of nodes ( x , y) is the number of arrows from
x to y .
148.1.1 Example For the left digraph in Figure 144.1 the adjacency matrix is
x y z w
x 0 2 1 0
2/0100
z 1 1 0 0
0 0 0 0
148.1.2 Remark The adjacency matrix depends on the way the nodes are
ordered; thus if you permute the nodes you get a different adjacency matrix for
the same graph. Note that the adjacency matrix does not contain the information
concerning the names of the arrows.
225
definition 4
digraph 74, 218
directed walk 225
divide 4
equivalent 40
node 218, 230
prime 10
tuple 50, 139, 140
148.1.4 Exercise Give the relational description of the digraph in Exercise
148.1.3.
(Answer on page 251.)
148.1.5 Uses of the adjacency matrix You can use the adjacency matrix of a
graph to determine properties of the graph:
(i) It is simple if no entry in the adjacency matrix is greater than 1 .
(ii) It has no loops if the entries down the main diagonal (the one from upper left
to lower right) are all 0.
(iii) The outdegree of a node is the sum over its row and the indegree is the sum
over its column.
The adjacency matrix will be used in the next section to calculate which nodes
can be reached from a given node.
148.1.6 Exercise Give the adjacency matrices of the digraphs in Figure 147.1.
(Answer on page 251.)
148.1.7 Exercise Draw this digraph and give its adjacency matrix: The nodes
are the numbers 1,2,3,4,6,12 and there is an arrow from a to b if and only if a and
b have the same prime factors (in other words, for all primes p. p \ a 44 p \ b).
148.1.3 Exercise Draw the graph with this adjacency matrix:
12 3 4
10 111
2 0 0 1 1
3 0 10 1
4 10 0 0
(Answer on page 251.)
149. Paths and circuits
149.1 Definition: directed walk
A directed walk of length k from a node p to a node q in a digraph
is a tuple (a\,. .. ,aj~) of arrows for which
P.l source(ai) = p;
P.2 target(afc) = q ; and
P.3 if k > 1 , then for each i = 1 , ... ,k — 1 , source(aj + i) = target(aj) .
149.1.1 Remarks
a) By definition, the length of a directed walk is the number of arrows it goes
through. If it goes through an arrow twice, the arrow is counted twice. A
directed walk of length n will thus make n + 1 visits to nodes, counting the
start and finish nodes, and the same node may be visited more than once.
b) We allow the empty walk () from any node to itself.
226
definition 4
digraph 74, 218
directed circuit 226
directed path 226
function 56
node 218, 230
recursive 157
simple directed
path 226
149.1.2 Example All these refer to digraph (149.1) below.
a) The walk (u) on the left digraph is of length one and touches the node y
twice.
b) The empty walk () from y to y is also a walk (of length 0); it is not the same
as (u) .
c) The walk (c, d, e) goes from z to y and touches z twice.
d) The walk (c, d, c, d) goes from z to z and touches each of x and z twice.
e) (e,a,d) is not a directed walk because an arrow goes the wrong way.
149.2 Definition: directed path
A directed path is a directed walk in which the arrows a\, . . . , are
all different.
149.2.1 Example In the digraph (149.1):
a) (c,a,u) is a directed path of length 3 from z to y.
b) (d, c, a) is a directed path of length 3 from x to y .
c) (e) is a directed path of length 1 from z to y .
d) (d, c, d, e ) is a directed walk that is not a directed path.
149.3 Definition: directed circuit
A directed circuit is a directed path from a node to itself.
149.3.1 Remark A directed circuit must be a path, not merely a walk.
149.3.2 Example In the digraph (149.1), the only directed circuits are the three
empty paths, (c,d), ( d,c ) and (u ) . (Thus a loop is a directed circuit.)
149.4 Definition: simple directed path
A simple directed path is a directed path not containing any directed
circuits, so that you never hit a node twice.
149.4.1 Example The only simple directed paths from z to y in the digraph (149.1)
are (c,a), (c, 6), and (e) .
149.4.2 Example Programs in many languages such as Pascal are made up of
procedures or functions that call on each other. It is often useful to draw a
digraph
in which the nodes are the procedures and functions and there is an arrow from P
to Q if Q is called when P is run. A loop in such a digraph indicates a procedure
or function that calls itself recursively. Larger circuits indicate indirect
recursion.
227
149.4.3 Exercise Find all the simple directed paths from 1 to 3 in the digraph
G = (Go, G\,s,t), where Go = {1,2,3}, G\ = {a, 6, c, d, e} , s(a) = s(e) = l, s(b)
=
s(c) = s(d ) = 2, t(a) = 2, t(b ) = t(c) = 1, and t(d) = t(e) = 3. (This is the
same as
the digraph in Exercise 144.2.2(b).) (Answer on page 251.)
149.4.4 Exercise A digraph is transitive if whenever there are arrows x — > y
and y — » z, there must be an arrow x—>z. Show that a digraph is transitive if and
only if whenever there is a walk from x to y there is an arrow x — >• y .
150. Matrix addition and multiplication
The adjacency matrix of a digraph can be used to compute directed walks from one
node to another. This involves the concepts of matrix addition and multiplication,
which are described briefly here.
associative 70
Cartesian product 52
commutative 71
definition 4
digraph 74, 218
fact 1
function 56
integer 3
node 218, 230
positive integer 3
scalar product 227
transitive 80, 227
tuple 50, 139, 140
usage 2
150.1 Definition: scalar product
Let V and W be two n-tuples of real numbers. The scalar product
V ■ W is the sum E” =1 VjtR .
150.1.1 Example ( 3 , 5 , - 1 , 0 ) • ( 1 , 2 , 3 , 4 ) = 10 .
150.1.2 Usage The scalar product is also called the “dot” product. You may be
familiar with its geometrical meaning when the tuples represent vectors.
150.1.3 Remark The scalar product is only defined for two tuples of the same
length. For each positive integer n, it is a function R n x R n — > R.
150.2 Definition: product of matrices
Let A be a k x m matrix with real entries, and B an mx n matrix
with real entries; specifically, A has the same number of columns as B
has rows. Then the product AB of the matrices is the k x n matrix
whose (i,j ) th entry is the scalar product of the zth row of A and the
j th column of B . In other words,
(AB)ij = Y% =l A ik B kj ( 150 . 1 )
150.2.1 Example
/ 1 3 0
\2 2 2
2 10 5
3 - 21-1
5 110
11 -5 3 2 \
20 0 4 8 )
( 150 . 2 )
150.2.2 Fact Matrix multiplication is associative, when it is defined; in other
words, for a k x m matrix A, an m x n matrix B and an nxp matrix C , ( AB)C =
A(BC) . Matrix multiplication is not, however, commutative. There are nxn
matrices A and B for which AB ^ BA. (Note that if AB and BA are both
defined, then A and B must be square matrices.)
228
associative 70
commutative 71
definition 4
digraph 74, 218
induction hypothe-
sis 152
induction 152
integer 3
node 218, 230
proof 4
theorem 2
150.2.3 Exercise Give examples of 2 x 2 matrices showing that matrix multipli-
cation is not commutative.
150.2.4 Exercise Show that matrix multiplication is associative when it is
defined.
150.3 Definition: sum of matrices
Let M and N be m x n matrices. Then the sum M + N is defined by
requiring that (M + N)ij = M t j + Nij .
150.3.1 Remark Two matrices can be added if and only if they have the same
dimensions.
150.3.2 Example
(150.3)
150.4 Powers of matrices
In the following, we will use powers of square matrices with integer coefficients.
If
M is a square m x m matrix, M n denotes M multiplied by itself n — 1 times. This
is best defined by induction: M° = I , M 1 = M, and M n = M n ~ l ■ M . It follows
from this and Definition 150.2 that
(M'% = E f =1 {M n - l ) ik M kj (150.4)
151. Directed walks and matrices
151.1 Theorem
If G = (Go,Gi,s,t) is a digraph with adjacency matrix M, then the
number of directed walks of length k from node p to node q is the (p, q) th
entry of M k .
Proof This fact can be proved by induction on k. It is clear for k = 1, since a
directed walk of length 1 is just an arrow, and the ( p , q) th entry in M 1 = M is
the
number of arrows from p to q by definition.
Suppose it is true that for all nodes p and q, the (p,q) th entry of M k is the
number of directed walks of length k from p to q . A directed walk of length k + 1
from p to q is a directed walk of length k from p to some node r followed by
an arrow (directed walk of length 1) from r to q. By the induction hypothesis,
there are ( M k ) pr directed walks of length k from p to r, and there are M rq
arrows
from r to q . Hence the number of directed walks of length k + 1 from p to q that
consist of a directed walk of length k from p to r followed by an arrow from r to
q is ( M k ) pr x M rq . The total number of directed walks of length k + 1 from p
to q
must be obtained by adding up this number ( M k ) pr x M rq for each node r of the
229
Or — 3 Clr — 2 — 1 ®r+s— 1 ®r+s+ 2 ®r+s+3
corollary 1
digraph 74, 218
node 218, 230
nonnegative integer 3
proof 4
reachable 229
Figure 151.1: Walk with loop.
digraph; in other words, if there are n nodes in the digraph, the total number of
walks is
E? =1 |(MVK,M rg j (151.1)
That sum, by formula (150.1), is the (p,q ) th entry of Af fc+1 , which is M k x M,
and that is what we had to prove.
151.2 Reachability
Let p and q be nodes of a digraph G. One says that q is reachable from p if
there is at least one directed walk of some length (possibly zero) from p to q.
Since a directed walk of length k touches k + 1 nodes, it follows from the pigeon-
hole principle that a directed walk of length n or more in a digraph G with n nodes
must touch some node twice. Suppose such a walk (gi,... ,«*) touches a node x
twice; say arrow a r has source x and arrow a r + s (with s > 0) has target x. Then
the directed walk (a r , . . . , a r + s ) can be eliminated from the walk, as in
Figure 151.1,
giving
(ttl , . . . , Q>r—\ , Gr+s+1 > • ■ ■ j (151.2)
from p to q. (Note: if r = 1 or r + s = k, the walk (151.2) has to be modified in
an obvious way.)
Clearly, by successively eliminating circuits, one can replace the walk by a path
(not just a walk) of length < n. This leads to:
151.3 Corollary
Let G be a digraph as in Theorem 151.1 with n nodes and matrix M .
Then q is reachable from p if and only if the ( p , q) th entry of the matrix
K = I + M + M 2 + . . . + Af n_1 is nonzero.
Proof If there is a directed walk from p to q, then the argument before the state-
ment of the Corollary shows that there must be one of length n — 1 or less. This
means that one of the matrices M , A/ 2 , ... , M n ~ l has a nonzero (p,q ) th
entry.
Since all the entries in these matrices are nonnegative, this means that the (p,q)
th
230
definition 4
digraph 74, 218
even 5
finite 173
function 56
graph 230
implication 35, 36
reachability
matrix 230
subset 43
transitive
(digraph) 227
entry of K is nonzero. Conversely, if that entry is nonzero it must be because the
(p,q) th entry in M* for some i is nonzero.
151.3.1 Exercise Use matrix multiplication to find all the directed walks of
length 1,2,3 and 4 that go from 1 to 3 in these digraphs:
(151.3)
(Answer on page 251.)
151.4 Definition: reachability matrix
The matrix
K = I + M + M 2 + ... + M n ~ l
is called the reachability matrix for the digraph G.
151.4.1 Exercise Calculate the reachability matrices for the digraphs in Fig-
ure 144.1, page 218. (Answer on page 251.)
151.4.2 Exercise Let G be the digraph whose set of nodes is {1,2, 3, 4}, with an
arrow from a to b if and only if a is even and b is 2 or 3. Find the reachability
matrix of G by counting paths and by direct addition and multiplication of
matrices.
(You may use Mathenratica for the latter.)
151.4.3 Exercise Let D be a digraph with adjacency matrix M . Show that D
is transitive (as defined in the preceding problem) if and only if
( M 2 )ij / 0 =>- Mij / 0
for all pairs (i,j) .
152. Undirected graphs
In Chapters 144 through 151, we considered digraphs that consisted of nodes and
arrows between some of the nodes. The graphs considered in this section have nodes
with edges between them, but the edges have no direction assigned to them.
152.1 Definition: graph
A graph G consists of two finite sets Go and Gi together with a func-
tion r from G\ to the set of two-elenrent subsets of Go • The elements
of Go are called nodes or xvertices of G and the elements of G± are
called edges.
231
(a)
(b)
definition 4
edge 230
graph 230
injective 134
node 218, 230
simple graph 231
subset 43
Table 152.1: Some graphs
152.2 Definition: simple graph
G is a simple graph if T is injective, so that there is no more than one
edge connecting two nodes.
152.2.1 Exercise Which of the graphs in Table (152.1) are simple? (Answer on
page 251.)
152.2.2 Remark We will sometimes use the word “multigraph” to emphasize
that we are talking about a graph that is not necessarily simple.
152.2.3 Drawing graphs One draws a graph by using dots for the nodes, and
drawing a line between nodes p and q for each edge e for which T(e) = {p, q} . In
common with most of the literature on the subject, our graphs do not have loops:
the requirement that T have values in the set of two-element subsets rules out the
possibility of loops.
232
adjacency
matrix 224, 232
adjacent 232
definition 4
fact 1
graph 230
incident 232
symmetric 78, 232
152.2.4 Example The figure below shows two graphs; the one on the right is
simple.
In the left graph the set of nodes is {q,r,s,t}, the set of edges is {a,b,
u,v,x,w,y} ,
and, for example, T(a) = {r,t} .
152.3 Definition: incidence
If e is an edge in a graph and T(e) = {p, q} then e is said to be incident
on p (and on q ). Two nodes connected by an edge in a simple graph
are adjacent. If n edges connect two nodes the nodes are said to be
adjacent with multiplicity n.
152.4 Definition: adjacency matrix
The adjacency matrix of a graph is the square matrix A whose rows
and columns are indexed by the set of nodes, with A(p,q) =the number
of edges between p and q .
152.4.1 Fact It follows from the definition that for any (multi)graph with adja-
cency matrix A,
(i) for any node p, A(p,p ) = 0;
(ii) for any nodes p and q, A(p,q) = A(q,p) (this says A is symmetric); and
(iii) if the graph is simple, A has only 0’s and l’s as entries.
152.4.2 Remark Because of 152.4. l(i) and (ii), all the information about the
graph is contained in the triangular matrix consisting of the entries A(p,q) with
p < q.
152.4.3 Example The adjacency matrix of the left graph in Figure (152.1) is
0
l
s
0
233
152.5 Definition: degree
The degree of node is the number of edges incident on that node.
152.5.1 Example The degree of the node c in the right graph in Figure (152.1),
page 232, is 3, and the degree of d is 1.
152.5.2 Fact The degree of a node is the sum over the row (and also over the
column) of the adjacency matrix corresponding to that node.
152.5.3 Exercise Show that the sum of the degrees of the nodes of a graph is
twice the number of edges.
153. Special types of graphs
Two special kinds of graphs that will be referred to later are given in the
following
definitions.
153.1 Definition: complete graph on n nodes
A complete graph on n nodes is a simple graph with n nodes, each
pair of which are adjacent. Such a graph is denoted K n .
153.1.1 Example I\ 4 is shown in diagram (153.1) below.
153.1.2 Exercise Give a formula for the number of edges of I\ n for n > 0.
153.2 Definition: bipartite graph
A bipartite graph G is a graph whose nodes are the union of two dis-
joint nonempty subsets A and B , called its moieties, with the property
that every edge of G connects a node of A to a node of B .
153.2.1 Fact It follows from Definition 153.2 that no two nodes of A are adjacent,
and similarly for B .
153.3 Definition: complete bipartite graph
A bipartite graph G with moieties A and B is a complete bipartite
graph if every node of A is connected to every node of B . A complete
bipartite graph for which A has m elements and B has n elements with
m<n is denoted K m ^ n .
153.3.1 Example The right graph in the following figure is A 3 , 4 .
iu
adjacency
matrix 224, 232
bipartite graph 233
complete bipartite
graph 233
complete graph on n
nodes 233
definition 4
degree 233
edge 230
fact 1
graph 230
moiety 233
node 218, 230
subset 43
(153.1)
234
definition 4
digraph 74, 218
fact 1
full subgraph 234
full 234
function 56
graph 230
restriction 137
simple graph 231
subgraph 234
subset 43
usage 2
153.3.2 Exercise Which of the graphs in Table (152.1), page 231 are complete
graphs? (Answer on page 251.)
153.3.3 Exercise Which of the graphs in Table (152.1), page 231 are bipartite
graphs? Which are complete bipartite graphs? (Answer on page 251.)
1 53.3.4 Exercise Give a formula for the number of edges of the complete bipartite
graph K rnn .
154. Subgraphs
154.1 Definition: subgraph
A subgraph of a graph G is a graph G' whose nodes Gq are a subset
of the nodes Go of G, and for which every edge of G' is an edge of G
between nodes of G' . If every edge of G that connects nodes of G' is
an edge of G ' , then G' is a full subgraph of G .
154.1.1 Usage For some authors, “subgraph” means what we call a full subgraph.
154.1.2 Fact If G' is a subgraph of G, the edge function T 7 for G' is the restric-
tion to G o of the edge function T of G .
154.1.3 Example The following graph is a non-full subgraph of the left graph in
Figure (152.1), page 232.
r
(154.1)
154.1.4 Exercise Show that if K n is a subgraph of a simple graph G, then it is
a full subgraph. Is the same true of K rnn ?
155. Isomorphisms
155.0.5 Remark Isomorphism of graphs is analogous to isomorphism of digraphs:
it captures the idea that two graphs are the same in their connectivity — there is
a
way of matching up the nodes so that the edges match up too.
235
155.1 Definition: isomorphism
Let G and H be simple graphs. A function (3 : Gq — > Hq is an iso-
morphism from G to H if it is a bijection with the property that p
and q are adjacent in G if and only if /3(p) and /3(q) are adjacent in H.
G and H are isomorphic if there is an isomorphism from G to H .
155.1.1 Usage In electrical engineering, isomorphic graphs are said to have the
“same topology” .
155.1.2 Example In general there may be more than one isomorphism between
G and H . The graphs below are isomorphic. Altogether, there are 12 isomorphisms
between them.
adjacent 232
bijection 136
complete bipartite
graph 233
complete graph 233
definition 4
full subgraph 234
function 56
graph 230
identity function 63
integer 3
isomorphic 235
isomorphism 223,
235
moiety 233
node 218, 230
usage 2
155.1.3 Example The left graph below is not isomorphic to the right graph. The
identity map is a bijection on the nodes, and if nodes are adjacent in the left
graph,
they are adjacent in the right graph, but there are nodes in the right graph that
are adjacent there but not in the left graph. The definition of isomorphism
requires
that p and q be adjacent if and only if j3(p) and /3(q) are adjacent.
(155.2)
155.1.4 Exercise Group the graphs in Table (152.1), page 231 according to which
are isomorphic to each other. (Answer on page 251.)
155.1.5 Exercise In Table (152.1), page 231, show that (b) is isomorphic to a
full subgraph of (c), and to a nonfull subgraph of (c). (Answer on page 251.)
155.1.6 Exercise
a) Prove that any two complete graphs on n nodes are isomorphic.
b) Prove that if n < m, then a complete graph on n nodes is isomorphic to a full
subgraph of a complete graph on m nodes.
c) Prove that for fixed integers m and n, two complete bipartite graphs, each of
which has one moiety with m nodes and the other moiety with n nodes, are
isomorphic.
236
circuit 236
connected compo-
nent 236
connected 236
cycle 236
definition 4
digraph 74, 218
edge 230
fact 1
graph 230
isomorphic 235
isomorphism 223,
235
length 236
list 164
node 218, 230
path 236
simple graph 231
simple path 236
theorem 2
walk 236
155.1.7 Exercise
a) Give a definition of isomorphism for multigraphs.
b) Prove that a graph isomorphic to a simple graph (using your definition of
isomorphic) is simple.
c) Prove that for simple graphs your definition of isomorphism is the same as
Definition 147.1.
156. Connectivity in graphs
We talk about walks, paths and circuits in graphs in much the same way as for
digraphs.
156.1 Definition: walk
A walk from node p to node q in a graph is a sequence
(no,ei,rai,e 2 ,... ,rik- i,ek,rik)
of alternating nodes and edges for which no = p ■ nk = q, and e* is inci-
dent on rii - 1 and n t for i = 1, 2, . . . , k . The length of such a walk is k ,
which is the number of edges occurring in the list (counting repetitions),
or one less than the number of nodes ocurring in the list.
156.2 Definition: path
A path in a graph is a walk in which no edges are repeated. A simple
path is a path in which no nodes are repeated.
156.3 Definition: circuit
A circuit is a path (not a walk) from a node to itself, and a cycle is
a circuit in which no nodes are repeated except that the beginning and
end are the same.
156.3.1 Fact It is easy to see (eliminate circuits) that if there is a walk between
two nodes then there is a simple path between them.
156.4 Definition: connected
A graph is connected if there is a path (hence a simple path) between
any two nodes. If p is a node in a graph, let C (p) denote the set
consisting of p and of all nodes q for which there is a path between p
and q. The sets C (p) are called the connected components of the
graph G.
156.4.1 Fact Part (a) of the theorem below implies that two nodes in a graph are
joined by a path if and only if they are in the same connected component. A graph
is therefore connected if and only if it has just one connected component.
237
156.5 Theorem
Let G be a graph.
a) Let p be a node in G . For any two nodes q and r in C (p) there
is a path from q to r .
b) If q G C(p) then C(p) = C(q) .
c) The set {C(p) \ p £ Go} is a partition of G .
Proof For (a), if p = q or p = r there is a path from q to r by definition of C (p)
.
Otherwise, just connect the path from p to q to the path from p to r. The result
might only be a walk, but by eliminating circuits, you get a path. That proves (a).
If q £ C(p) , (a) implies there is a path from p to r if and only if there is a
path
from g to r, so (b) follows. Finally, any node p is an element of C(p) ; this and
(b)
implies that every node is in exactly one set C (p) , so the sets C (p) form a
partition
of the nodes. That proves (c).
circuit 236
connected graph 236
cycle 236
definition 4
diameter 237
distance 237
edge 230
Eulerian circuit 237
graph 230
node 218, 230
partition 180
path 236
proof 4
simple path 236
theorem 2
walk 236
156.6 Definition: distance
The distance between two nodes p and q in a connected graph is the
length of the shortest simple path between p and q.
156.6.1 Example In the right graph of Figure (152.1), the distance between
nodes d and / is 3 . There are of course simple paths of length 4 and 5 between
nodes d and / , but the shortest one has length 3 .
156.7 Definition: diameter
The diameter of a connected graph is the maximum distance between
any two nodes in the graph.
156.7.1 Example The diameter of the graph just mentioned is 3.
157. Special types of circuits
157.1 Definition: Eulerian circuit
An Eulerian circuit is a circuit in a graph which contains each edge
exactly once. It need not be a cycle; in other words, nodes may be
repeated, but not edges.
A graph need not have an Eulerian circuit. For example, the graph in Fig-
ure (152.1) has no Eulerian circuit. There is a simple criterion for whether a
graph
has an Eulerian circuit:
238
circuit 236
connected graph 236
connected 236
converse 42
definition 4
degree 233
edge 230
Eulerian circuit 237
even 5
fact 1
finite 173
graph 230
Hamiltonian cir-
cuit 238
incident 232
integer 3
node 218, 230
proof 4
157.2 Theorem
A connected graph G has an Eulerian circuit if and only if the degree of
every node is even.
Proof Suppose G has an Eulerian circuit. As you go around the circuit, you
have to hit every edge exactly once. Every time you go through a node, you must
therefore leave by a different edge from the one you entered. So for each node p.
you can divide the edges incident to p into two groups: those you enter p on and
those you leave p on. Since you enter and leave p the same number of times, these
two groups of edges must have the same number of elements. Thus the number of
edges incident on p is even.
Now for the converse: suppose every node of G has even degree. To construct
an Eulerian circuit, pick a node p. If that is the only node in G you are finished.
Otherwise, there is an edge on p. Travel along that edge to some node q and mark
the edge so you won’t use it again. Because there are an even number of edges
incident on q . there is an unmarked edge. Leave on the edge and repeat the process
until you arrive at p again.
This process will produce a circuit containing p. No edge can be repeated
because you are marking the ones you use, and because of finiteness you have to
return to p sometime. However, the circuit may not pass over every edge. If it does
not, there is an unmarked edge e incident on some node q already in your circuit,
because G is connected. Start with that node and that edge and repeat the process,
continuing until you return to q. This will give another circuit containing q. Note
that the second circuit may hit nodes of the first circuit, but there will always
be
an unmarked edge to leave on because each node in the first circuit has even degree
and an even number of marked edges. You now can put these two circuits together
into a big circuit — go around the first circuit starting at p until you get to q,
go
around the second circuit until you return to q, and then continue around the first
circuit until you get back to p. If you still don’t hit all the edges, you can
repeat
this process a second time, and so on until all the edges are used up. The result
will
be an Eulerian circuit.
This problem was first solved by Leonhard Euler, who was asked whether it was
possible to walk around the city of Konigsberg (then in Prussia, now in Russia and
called Kaliningrad) in such a way that you could traverse each of its seven bridges
exactly once. The arrangement of bridges in Euler’s time is represented by the left
graph in Figure (152.1), page 232 (each edge represents a bridge), which clearly
has
no Eulerian circuit since in fact none of its nodes has even degree.
157.2.1 Exercise For which integers n does K n have an Eulerian circuit?
157.2.2 Exercise For which integers m and n does K m , n have Eulerian circuit?
157.3 Definition: Hamiltonian circuit
A Hamiltonian circuit in a graph is a circuit which hits each node
exactly once.
157.3.1 Fact Such a graph must be connected (why?).
239
157.3.2 Remark Our main purpose in mentioning Hamiltonian circuits is to con-
trast their theory with that of Eulerian circuits: there is no known simple
criterion
to determine whether a graph has a Hamiltonian circuit or not. The problem is com-
putationally difficult in general, although for special classes of graphs the
question
can be answered more easily (Problems 157.4.5 and 157.3.3).
157.3.3 Exercise For which integers m and n does K rnn have a Hamiltonian
circuit?
157.4 Exercise set
Exercises 157.4.1 through 157.4.3 concern the graphs in Table 152.1, page 231.
definition 4
diameter 237
edge 230
embedded in the
plane 239
Eulerian circuit 237
graph 230
Hamiltonian cir-
cuit 238
integer 3
planar 239
157.4.1 Give the diameter of each graph. (Answer on page 251.)
157.4.2 Which of the graphs has an Eulerian circuit? (Answer on page 252.)
157.4.3 Which of the graphs has a Hamiltonian circuit? (Answer on page 252.)
157.4.4 Give examples of:
a) A graph which has an Eulerian circuit but not a Hamiltonian circuit.
b) A graph which has a Hamiltonian circuit but not an Eulerian circuit.
157.4.5 For which integers n does K n have a Hamiltonian circuit?
158. Planar graphs
158.1 Definition: Planar
A graph is embedded in the plane if it is drawn in such a way that
no two edges cross. It is planar if can can be embedded in the plane.
158.1.1 Example Graphs can be used to represent electric circuits. It is desirable
in a printed circuit that no two lines (edges of the graph) cross each other. This
is
exactly the statement that the graph is embedded in the plane.
158.1.2 Example The left graph in Figure (155.1), page 235, can be embedded
in the plane as the right graph in the same figure.
158.1.3 Warning The fact that a graph is drawn with edges crossing does not
mean it is not planar. For example, K 4 is planar, in spite of the way it is drawn
in
Figure (153.1), page 233.
240
complete bipartite
graph 233
complete graph 233
definition 4
edge 230
embedded in the
plane 239
graph 230
node 218, 230
planar 239
subdivision 240
subgraph 234
theorem 2
utility graph 240
158.1.4 Exercise Which graphs on page 231, are planar? (Answer on page 252.)
158.1.5 Example Not all graphs can be embedded in the plane. For example,
the complete graph on 5 vertices (left graph below) cannot be embedded in the
plane. Another such graph is the utility graph, the right graph below (which is
the complete bipartite graph 143 , 3 ). It arises if you have three houses a, b and
c
that must each be connected to the water, sewer and gas plants (w, s and g ). If it
is drawn in the plane, edges must cross.
(158.1)
k 5
A',3,3
There is an easy-to-use criterion to determine whether a graph is planar. It
requires a new concept:
158.2 Definition: subdivision
A subdivision of a graph is obtained by repeatedly applying the fol-
lowing process zero or more times: take an edge e connecting two nodes
x and y and replace it by a new node z and two edges e! and ef' with
e' connecting x and z and e" connecting y and z.
158.2.1 Example The graph H below is a subdivision of G ; it is obtained by
subdividing three times. Note that a graph is always a subdivision of itself.
(158.2)
158.3 Theorem: Kuratowski’s Theorem
A graph is not planar if and only if it contains as a subgraph either a
subdivision of K$ or a subdivision of the utility graph
158.3.1 Remark This theorem has a fairly technical proof that will not be given
here. Note that it turns a property that it would appear difficult to verify into
one
that is fairly easy to verify.
241
159. Graph coloring
Some very difficult questions arise concerning labeling the node of a simple graph.
159.1 Definition: coloring
A coloring of a simple graph G is a labeling L : G$ — >■ S (where S is
some finite set) with the property that if nodes p and q are adjacent,
then L(p) ^ L(q ) . In this context the elements of S are called colors.
159.1.1 Remark This terminology arises from the problem of coloring a map
of countries in such a way that countries with a common border are colored with
different colors. In the (very large) literature on coloring problems, two states
or
countries that have only a point on their borders in common, such as Arizona and
Colorado in the U.S.A., are regarded as not bordering each other. The common
border must have a nonzero length.
characteristic func-
tion 65
chromatic num-
ber 241
coloring 241
color 241
definition 4
edge 230
finite 173
graph 230
labeling 221
node 218, 230
odd 5
simple graph 231
159.1.2 Example The state of Kentucky in the U.S.A., and the seven states
bordering it, require four colors to color them in such a way that adjoining states
do not have the same color. This is turned into a problem of graph theory by
drawing a graph with one node for each state and an edge between two nodes if the
corresponding states border each other:
MO TN
(159.1)
159.2 Definition: chromatic number
The smallest number of colors needed to color a simple graph G is called
the chromatic number of G, denoted x(G).
Warning: Note that we have already used x f° r the characteristic function of
a subset of a set.
159.2.1 Example The chromatic number of the graph in Figure (159.1) is 4, and
the chromatic number of the right graph in Figure (152.1) is 3.
159.2.2 Exercise Show that a graph with at least one edge is bipartite if and
only if its chromatic number is 2 .
159.2.3 Exercise Show that a graph has chromatic number 2 if and only if it has
no cycles of odd length.
242
bipartite graph 233
chromatic num-
ber 241
coloring 241
color 241
complete graph 233
Four Color Theo-
rem 242
graph 230
Kempe graph 242
Kuratowski’s Theo-
rem 240
moiety 233
node 218, 230
planar 239
subgraph 234
subset 43
159.2.4 Remark It is in general a nontrivial question to determine the chromatic
number of a graph. However, some things can be said.
a) The complete graph on n nodes has chromatic number n, since every node is
adjacent to every other one.
b) A bipartite graph has chromatic number 2 (if it has any edges): since none
of the nodes in one of the moieties are adjacent to each other, they can all
be colored the same color, and the nodes in the other moiety can be colored
another color.
c) It is known that any planar graph has chromatic number < 4. This fact is
called the Four Color Theorem and is difficult to prove.
159.2.5 Example As an indication of the problems involved in proving the Four
Color Theorem, observe that the graph of states in Figure (159.1) has chromatic
number 4, although it does not contain the complete graph K 4 as a subgraph.
In other words, although there is no four-element subset of the states involved in
Figure (159.1) that all border each other (thus turning into a copy of K 4 in
(159.1)),
it nevertheless takes four colors to color the whole graph. It follows that you
can’t
use Kuratowski’s Theorem to prove the Four Color Theorem: the fact that no
planar graph contains K 5 as a subgraph does not rule out the possibility that a
planar graph needs five colors to color it.
159.2.6 Exercise Give an example of a graph with chromatic number 3 that does
not contain a subgraph isomorphic to K 3 .
159.2.7 Exercise Find a place in the world with four political subdivisions that
all border each other. (There are no four states in the U.S.A. like this, although
you
will observe that North Carolina, South Carolina, Georgia and the Atlantic Ocean
all “border” each other.)
159.2.8 Exercise A Kempe graph is a graph with n+1 nodes, consisting of
n nodes in a cycle and another node connected to each node in the cycle, and no
other edges. Figure (159.1), page 241, is a Kempe graph.
a) Show that a Kempe graph is planar.
b) Find the chromatic number of a Kempe graph. (It will depend on n.)
159.2.9 Garbage routes The effort to prove the Four Color Theorem resulted in
the discovery of fast coloring algorithms and of a lot of detailed information
about
graph coloring. This has other applications besides coloring maps. For example,
consider the following problem: A city is divided into a number of garbage pickup
routes. Some of the routes overlap, because businesses must be picked up more
often than residences and therefore are assigned to two or more routes. What is
the best way to distribute the routes among the five working days of the week, with
each route traveled once a week?
If each route is regarded as a node, with two routes adjacent if they overlap,
the result is a graph. A scheduling of the routes that avoids scheduling
overlapping
routes on the same day is a five-coloring of this graph. An efficient way of
coloring
the graph would be a start towards finding a good schedule. Note that this problem
has nothing to do with planarity or the Four Color Theorem.
243
Answers to Selected Exercises
3 . 1.5 Yes, because —(—3) = 3 and 3 > 0, so by
Definition 2.2, —(—3) is positive.
4 . 1.2 Yes, because 52 = 4-13.
4 . 1.10 -2, -1, 1, 2.
5 . 5.1 333 = 9 x 37 and 9 is an integer, so 37 | 333
by Definition 4.1.
5 . 5.2 Suppose 0 < k < n and suppose k is divisi-
ble by n. By Definition 4.1, there is an integer q for
which k = qn. Since k and n are nonnegative, so is
q . Since k = qn < n , dividing through the inequal-
ity by n (which is positive) gives q < 1 . Since q is
nonnegative, it must be 0 . Since k = qn, k = 0 as
well.
6 . 1.5 91 = 7x13; 98 = 2 x 7 2 ; 108 = 2 2 x 3 3 ;
111 = 3 x 37 ; 211 is prime
7 . 5.1 No. For example,
112 ,232
- + - = - and - - = -
9 . 2.4 Only the pair in (c) are equal.
10 . 1.2 5.1 = 46/9; 4.36 = 48/11; 4.136 = 91/22.
12 . 2.6 x 2 — | + 4x > 2x .
12 . 4.1 to = 2 makes it true and to = 8 makes it
false.
12 . 4.2 Any to makes it true. No value of to
makes it false.
12 . 5.2 Q(-l) is 1 < 4 and Q(x — 1) is (x — l) 2 <
4.
12 . 5.3 a. 2<5. b. 3<4. c. x 2 < x + y + 1 . d.
x{x + y ) <x + y + z + l.
13 . 2.7 (a) and (b) are true, and the others are
false. It is wrong to say that (c) is “sometimes
true” or “usually true” . The statement that 3 • 0 > 0
is false, so the statement (Va;:N)(3a; > x) is simply
false, with no qualification.
14 . 2.3
2 6
T T
T T
T T
14 . 2.4 a) True: n = 5. False: Any n other than
5.
b) True: n = 8, for example, or n = 0. False: n =
4, 5, 6, 7 are the only ones.
c) True: Impossible. False: any n.
d) True: Any n.
14 . 2.5 Only (d).
17 . 1.4 3.
18 . 1.5 a) 2. b) 3. c) 2. d) 0. (For (b), see
Remark 8.1.3.)
18 . 1.16 You must show that P(a) is false.
19 . 2.5 (a) and (c) are true and (b) is false.
19 . 2.6 -13, -7, -5, -4, -3, -2, 0,1, 2, 3, 5, 11. b)
1,4,9,16,36,144. c) Same as (b).
20 . 1.3 (a) and (c) are the same, and so are (b)
and (d).
22 . 1.6 Only (d) is the empty set.
23 . 1.5 d is the empty set and b, c and g are single-
tons.
23 . 1.6 (a) Di is the only singleton, (n) 1 is the
only integer which is an element of D n for every
positive integer n.
25 . 1.4 Item (a) is true for all integers to but
(b) and (c) are false. For example, (b) is false for
to = 6 (then the hypothesis is true and the conclu-
sion is false, and that is the line in the truth table
that makes the implication false), and (c) is false for
to = — 2 .
26 . 1.5 a) True: n = 6, for example (this is vacu-
ously true). False: n= 8.
b) True: any n. False: not possible.
c) True: n = 10. False: n = 8.
d) True: any n. False: not possible.
e) True: any n (always vacuously true). False: not
possible.
f) True: Any n except 1 . False: n = 1 .
27 . 2.1 (a), (c), (d) and (e) say the same thing,
and (b) and (f) say the same thing.
30 . 4.5 The contrapositive is “If n is not prime,
then 3 does not divide n”, which is not true for
some integers n. The converse is “If n is prime,
then 3 | n”, which is also false for some n.
244
=
a)
b)
c)
d)
N
N
e)
31.5.3 You must show that there is an element
x G S that is not an element of T . This is because
of Definition 31.1, which defines AC B to mean the
implication x £ A => x £ B , and the only way that
implication can be false is for the hypothesis to be
true and the conclusion false.
32.1.6
a: 4.
b: 0
. c
: 1. d:
2.
32.1.7
{0,{5},{6},
{7}, {5
',6}
,{6,
7}, {5
',7}, {5,6,7}}
32.1.8
cl
e
b
Y
N
N
N
N
g
N
33.2.2
{1,
2,3}
U {2,3,4,
5}
= {1,2,3
,4,5}
and
{1,2,3} n
{2,3
,4,5}
{2,3}
33.2.3
NU
Z = Z
and NflZ
= N
33.2.7
By
Definition 31.
1,
we :
must
show
that
if x € An B ,
then
GiU
B.
By
Definition
33.2
(of intersection), x
cAnB
implies
that
x £ A
and
x G B . By Definition 33.1 (of union), if x G A, then
x G AuB.
33.3.1 There are of course an infinite number
of answers. Some correct answers are: The set of
all negative integers, the set of all negative even
integers, {— 1,— 2,— 3}, {—42}, and the empty set
(which is disjoint from every set).
34.2.2 Z — N is the set of all negative integers.
N — Z = 0.
34.2.5 (a) 1,2, 3, 4, 5; (b) 2,3; (c) 1,2, 3, 4, 5, 7, 8;
(d) none; (e) 1; (f) 2, 3, 4, 5; (g) 1,2,3; (h) 1,2, 3, 4, 5;
(i) 2, 3, 4, 5.
34.2.6 1) 1 and 2. 2) 1. 3) 1, 3 and 5. 4) 5. 5) 6
and 7. 6) None. 7) 6 and 7.
35.1.3 The pairs in (a) are different; the pairs in
(b) and (c) are equal.
36.1.2 m fin is k , where k is the minimum of m
and n , and m U n is l , where l is the maximum of
to and n.
36.3.1 None of them are equal.
36.4.1 1. a) 3,4. b) 2, (1,5). c) 2, (5, (2,1)).
d) 2,9. e) 2, {1,2}. f) 4, Z.
37.1.2 (l,o), (1,6), (2,o), (2, 6).
37.6.1 This is false for any nonempty set A
because the elements of A x A are pairs of elements
of A, and an ordered pair is distinct from its coor-
dinates (see 35.1). (The last statement implies that
in fact for nonempty A, Ax A and A have no ele-
ments in common.) The statement A x A = A is
true if 4 = 0.
37.7.1 “For all sets A and B and all nonempty
sets C , . . . ”
37.9.1
(a) A
(b) 1,2
(c) (1,1), (1,2), (2,1), (2, 2)
(d) (1,1,1), (1,1, 2), (1,2,1), (1,2, 2),
( 2 , 1 , 1 ), ( 2 , 1 , 2 ), ( 2 , 2 , 1 ), ( 2 , 2 , 2 )
(e) (1,3), (1,4), (1,5), (2, 3), (2, 4), (2, 5)
(f) (3,1), (3, 2), (4,1), (4, 2), (5,1), (5, 2)
(g) (1,1, 3), (1,1, 4), (1,1, 5), (1,2, 3), (1,2, 4), (1,2, 5),
(2, 1,3), (2, 1,4), (2, 1,5), (2, 2, 3), (2, 2, 4), (2, 2, 5)
(h) (1,(1, 3)), (1,(1, 4)), (1,(1, 5)), (1,(2, 3)),
(1,(2, 4)), (1,(2, 5)), (2, (1,3)), (2, (1,4)),
(2, (1,5)), (2, (2, 3)), (2, (2, 4)), (2, (2, 5))
(i) (1,3), (1,4), (1,5), (2, 3), (2, 4), (2, 5), 1,2
(j) 0
37.9.2
4
5
N
Y
N
N
6
N
N
38.2.1 j(a:,n) | a: > ?rj C R x N.
38.2.2 {(x,y) \ x G R ,y = l} = |(x,l) | x G
R j C R x R
38.2.3 {1} CR,
245
38.2.4 ^(x,y,z,w) \ x + y = z j CRxRxRxR.
39.3.7 F(l) = {{1}, {1,2}, {1,3}, {1,2, 3}} and
F(2) = {{2},{1,2},{2,3},{1,2,3}}.
40.2.6 (a) and (d) only.
41.1.8 F( 2) F(4)
a) 2 4
b) 42 42
c) 2 4
41.1.9 a) (2,2), (3,3)
b) (2,2), (3,3)
c) (2,2), (3,3)
d) (1,3), (2,3), (3,3)
e) ((1, 2) ,1)),((1,3),1), ((2, 2) , 2)),
((2, 3), 2) , ((3, 2), 3)) , ((3, 3) ,3)
42.2.3 a) \x.x 3 \ x i-+ x 3 : R — ► R b) A (a, 6). a;
(a, 6) i->- a: Ax B -+ A. c) A (a,b).a + b\ (a, 6) i — >•
a + b R. x R. — y R.
43.1.4 a) (1, FALSE), (2, TRUE), (3, TRUE)
b) (1,TRUE), (2, FALSE), (3, TRUE)
c) ((2, 2), 4), ((2, 3), 5), ((3, 2), 5), ((3, 3), 6)
44.1.5 a) (1) only, b) (2) only, c) (3) only, d) (1)
only. Note that (4) is not an answer to (d) because
the function is given as having codomain R. Of
course there is a function x H > x 2 : R — > R + with the
same graph but it is technically a different function.
For many purposes, this is merely a technicality, but
there are places in mathematics where the distinc-
tion is quite important.
46.4.3 35 22 + 6 5 +*.
48.1.5 We must show, for all subsets A, B and
C of S, that dU(5UC) = (4UB)U(7. We will
do this using Method 21.2.1, page 32. Suppose that
x £ iU(BUC) . Then by (33.1), page 47, either x £
A or x £ BUC . If x £ A, then x £ AuB , so x £
(A\J B)U C by using the definition of union twice.
If x £ B U C , then either x £ B or x £C . If x £ B ,
then x £ AU B , so x £ (AU B) U C . If x £ C , then
again by definition of union, x £ ( A U B) U C . So we
have verified that in every case,
x £ AU(BUC) => x£(AUB)\JC
so that by Definition 31.1, page 43, AU(BUC) C
(A UB)UC. A similarly tedious argument shows
that (A U B) U C C A U (B U C) . Therefore by
Method 21.2.1, A U (B U C) = (A U B) U C.
50 . 1.4 (1) is associative, not commutative, and
does not have an identity. (2) is not associative
(because (aA6)Ac = abut a A (b Ac) = b), is com-
mutative, and does not have an identity.
50 . 1.7 The empty set, since for any subset A of
S, AU0 = 0UA = A.
51 . 1.5
a) (1,3), (1,5), (2,1), (2, 3), (2, 5), (3,1), (3, 5)
b) (2, 2), (2, 4), (2, 6), (2, 8), (2, 10), (3, 3),
(3, 6), (3, 9), (5, 5), (5, 10), (7, 7)
c) (1,1), (1,2), (1,3), (2, 2), (3, 3)
52 . 1.3
a) (1,2), (1,3), (1,4), (2, 3), (2, 4), (3, 4)
b) (1,1), (2, 2), (3, 3), (4, 4). (This is A a .)
c) (1,3), (2, 3), (3, 3), (4, 3).
d) (1,1), (3, 3), (1,3), (3,1).
53 . 1.2 (a), (c) and (e) are functional relations.
53 . 2.3 1 ^ {3,5} , 2 h+ {1,3,5}, 3 h>{1,5}.
53 . 3.3 {(1,3), (1,4), (2,1), (2, 3), (2, 4), (-666,0)}
55 . 1.9 (b) is not reflexive, the others are reflexive.
56 . 1.4 (b) and (c) are symmetric, (a) and (d) are
not.
57 . 1.9 (a), (b) and (c) are antisymmetric; (d) is
not. Note that (c) is vacuously antisymmetric.
59 . 1.3
ref
sym i
N
b
cl
Y
f
59 . 1.4
ref
sym
c
N
cl
N
N
concerning (d):
2 <
3 2 , 3 <
60 . 1.2
a: q -
= 0, r
= 2. b:
r = 0. d:
q = 3
, r =
1.
ant
trs
irr
Y
N
ant trs irr Note
N N Y
N Y N
N N N
N N N
Y Y N
N Y N
N N N
2 2 , 8 < 3 2 .
g = 0, r = 0. c: q = 2,
246
60.1.4 Suppose a=qm + r and b=q'm + r.
Then a — b = qm — q'm = (q — q')m so it is divisi-
ble by to.
60.2.3 Since m div n — a, m = an + r for some
integer r such that 0 < r < n. We are given
that m = an + n + b+2, so r = n + 6 + 2 . Hence
n + b + 2 < n , so that 6 + 2 < 0 , so 6 < 0 .
60.2.4 Since n\s, s = qn for some integer q . q
is not less than 0 since n and s are nonnegative. It
is not greater than 0 since then qn>n>s but we
are given s = qn. So q must be 0 , so that s is 0
too.
60.5.2 By Definition 60.1, we must show that
37 = 7 • 5 + 2 and that 0 < 2 < 5 . Both are simple
arithmetic. It follows from Theorem 60.2 that the
quotient is 7 and the remainder 2 as claimed. (Yes,
you knew this in fourth grade. The point here is
that it follows from the definitions and theorems we
have.)
60.5.4 4, because to = 36g + 40 = 36(g + 1) + 4
and 0 < 4 < 36 .
61.1.3 n < r < n + 1 | — n = floor(?’) , where n is
of type integer.
61 . 2.3 a: trunc(7/5) = floor(7/5)
= 1.
b: trunc(— 7/5) = — 1;
floor (— 7/5) =
= -2.
c: trunc(— 7) = floor (-
-7) = -7.
d: trunc(— 6.7) = — 6;
floor (— 6.7) =
-7.
62 . 2.2 30 = 2 1 x 3
1 x5 1 , 35 =
5 1 x7 1 , 36
2 2 x 3 2 , 37 = 37 1 , 38 =
= 2 1 x 19 1 .
62 . 3.2
prime 98 99
100 111 1332
1369
3 0 2
0 1
7 2 0
0 0
0
37 0 0
0 1
62 . 5.1
90 =
2 1 x 3 2 x 5 1
91 =
7 1 x 13 1
92 =
2 2 x 23 1
93 =
3 1 x 31 1
94 =
2 1 x 47 1
95 =
5 1 x 19 1
96 =
2 5 x 3 1
97 =
97 1
98 =
2 1 x 7 2
99 =
3 2 x ll 1
63 . 2.2
PAIR
GCD
LCM
12, 12
12
12
12, 13
156
12, 14
84
12, 24
12
24
63 . 2.4
False:
for
example GCD (4, 2) =
GCD(2,2)
= 2. If
you said
“TRUE” you may have
fallen into the trap of saying “the GCD of to and n
is the product of the primes that m and n have in
common,” which is incorrect.
63 . 2.5 (1,1), (1,2), (1,3), (1,4), (2,1), (2, 3),
(3,1), (3,2), (3,4), (4,1), (4,3)
63 . 3.2 If d divides both n and n+ 1 it must
divide their difference, which is 1 . Hence the largest
integer dividing n and n + 1 is 1 .
64 . 2.2 Suppose e | to and e\n. Let p be any
prime. Then e p (e) must be less than or equal to
e p (m) and also less than or equal to e p (n). Thus
it is less than or equal to e p (d), which by Theo-
rem 64.1 is the minimum of e p ( to) and e p (n) . This
is true for every prime p, so in the prime factoriza-
tion of e, every prime occurs no more often than it
does in d, so by Theorem 62.4, e\d.
64 . 2.4 Let p be any prime. By Theorem 64.1,
e p (d) = min (e p (m),e p (n)) . Observe that e p (m/d) =
e p (m) — e p (d) and e p {n/d) = e p (n) — e p (d) . We
know that e p (d) = min(e p (m),e p (n)) , so one of the
numbers e p (w) — e p (d) and e p (n) — e p (d) is zero.
That means p does not divide both in and n.
Since p was assumed to be any prime, this means
no prime divides both m and n. Therefore,
GCD (m/d,n/d) = 1, as required.
66 . 6.3 By Definition 66.4,
n = d m b m -\ 1- dib 1 + dob 0
so
bn = d m b m+1 -\ 1- d\b 2 + d 0 b 1 + 06°
which means that bn is represented by
dm.dm—1 * * ' d\Q .
67 . 2.3 a) 1100000. b) 11010010. c) 110001111.
d) 1010111100.
67 . 2.4 a) 1525. b) bOO. c) 10c9a.
68 . 4.1
247
DEC
OCT
HEX
BASE
36
BINARY
100
144
64
2s
1100100
111
157
6f
33
1101111
127
177
7f
3j
1111111
128
200
80
3k
10000000
69.3.1 ( x > 10) V (x < 12) . Of course, this is true
of all real numbers.
69.3.2 (x > 10) V (x > 12) . Of course, this is the
same as saying x >10.
71.2.5 Here are the truth tables:
P Q
PVQ =P
~^Q
—iP A —iQ
-1 ( _, P A —iQ)
T T
T F
F
T
F T
F F
T
F
The third and seventh columns are the same.
71.2.9
P Q -
P V Q P => Q
_i Q PA-iQ -i(PA-.Q)
T T
F F
T F
F
F
T T
F T
F F
F F
T F
T
The fourth, fifth and eighth columns are the
same.
74.2.1 Valid.
74.2.2 Valid.
74.2.3 Invalid.
74.2.7 Let P be 3 >5 and Q be 4 >6. Then
P => Q is true because both hypothesis and conclu-
sion are false; on the other hand, Q is false. Since
the hypothesis of (P => Q) => Q is therefore true
and the conclusion false, the statement is false.
75.3.4 a: True. Witness: 2. b: False. Coun-
terexample: 9. c: True. Witness: 2. d: False.
Counterexample: 3.
75.3.5 (a) True, (b) True, (c) True, (d) False;
a counterexample is given by taking P to be x > 7
and Q to be x <7 .
76.1.4 There are no counterexamples to
(Vy)P(14,j/) since it is the statement
(Vy) ((14 = y) V (14 > 5))
which is true because “14 > 5” is true.
The number 3 and any number greater than 5
is a witness to (3x)P(x,3).
77.2.1 (a) means that for every real number the
statement (3 y)(x > y) is true. A witness for that
statement is x — 1 , so the statement is true, (b)
means that there is a real number greater than any
real number, which is false, (c) is true. Witness: Let
x = y = 3. Then the statement becomes ((3 > 3) =>
(3 = 3)), which is (vacuously) true.
82.2.1 valid: direct method.
82.2.2 invalid: fallacy of affirming the hypothesis.
82.2.3 invalid: fallacy of denying the conse-
quence.
82.2.4 valid with false hypothesis.
82.2.5 invalid: fallacy of denying the conse-
quence.
85.1.3 This follows from Rule (85.1), page 124,
going from top to bottom. To use it, we must ver-
ify the two hypotheses of the rule with r = to — qn.
The first is qn + r = qn+ (to — qn) = m , as required.
The other, 0 < r < n, is immediate. Therefore the
conclusion, part of which states that q = in div n ,
must be true.
86.2.4 This is a proof by contradiction. Suppose
p > 2 and p is not odd. Then p is even, so it is
divisible by 2 . Therefore p is divisible by a num-
ber other that p and 1 (namely 2, which is not p
because p> 2 ) . This contradiction to the definition
of prime (Definition 6.1, page 10) shows that the
claim is correct.
88.3.1 a b c
12
16
Impossible, since
GCD( 12,16) =4.
12
16
4= 16-12.
26
30
2 = 7x 26-6 x 30.
26
30
4 = 14 x 26 - 12 x 30
26
30
-2 = 6 x 30 -7x 26.
51
100
1 = 25 x 100-49 x 51
88.3.3 If to and n are relatively prime, then
by Theorem 87.2 there are integers a " and b" for
which a"m + b"n = 1 . Then (a + a”)m+ (b+b")n =
am +bn + a"m + b"n = e + 1 . Note: If you reasoned
as follows: “Because a and b are relatively prime
and am. + bn = e , it follows that e = 1 by Theo-
rem 87.2,” then you are guilty of the fallacy of affirm-
ing the hypothesis (page 121).
89.1.7 The set of positive integers.
248
90.1.5 E({2,3} = {5} and F({3}) is also {5}.
inj?
surj?
image
a)
B
b)
{2,3}
c)
d)
e)
N
B
f)
{3}
g)
{TRUE, FALSE}
h )
i)
N
{4, 5, 6, 7, 8}
j)
{TRUE, FALSE}
inj?
surj?
image
a)
b)
Y
c)
{r € R r > 1}
d)
{r € R r < 2}
93.1.7 You must show that there are two differ-
ent elements a and a' of A for which F(a) = F (a') .
That is because the definition of injective is the
implication
a' => F(a) ^ F(a')
and the negation of that implication is the statement
a^a'A —i(F(a) ^ F(a'))
in other words
7^ a'
A F(a) =
F(a’)
domain
R+
inj'
? surj?
inj?
surj?
a)
b)
c)
Y
N
If the answers in the last column puzzle you, remem-
ber that the codomain of the restriction of a function
is the same as the codomain of the function.
97.1.3 a) Domain: {1,2, 3, 4, 5}.
Graph: { (1 , 2) , (2, 5) , (3, - 1) , (4, 3) , (5, 6) } .
b) Domain: {1,2, 3, 4}.
Graph: |(1 ,tt), (2, 5), (3,tt - 1) , (4, v / 2}|.
c) Domain: {1,2,3}.
Graph: { (1, (3, 5)), (2, (8, -7)), (3, (5,5))} .
97.3.1 a) (5,5,3, 17, — 1) . b) (27r,37r,47r,57r,67r) .
c) (1,4,9,16,25,36).
98.2.6
a) G° F : {1,2, 3, 4} -A {1,3, 5, 7, 9}, graph
{(1,1), (2, 7), (3, 3), (4, 7)}.
b) G° F : R — x R , (G° F)(x) = 2x 3 .
c) G° F : R — x R , (G°F)(x) = 8x 3 .
d) n t-x (n/2) : N — X R.
e) (i,|/)i->(3,i):RxR->RxR.
99.1.5 li-xl, 2 i— x 2, 3 i — >- 2 .
100.1.5
X I X Jx
R+ *-R +
101.2.3 Only (a) and (f) have inverses. For
(a) the inverse is F~ Y : {3, 4, 5, 6} — X {1,2, 3, 4} with
graph {(3,1), (4, 2), (6, 3), (5, 4)}. For (f) it is n i-x
n — 1 : Z — x Z .
101.2.4 All except (c) and (h) have left inverses,
(a), (f) and (h) have right inverses.
101.2.5 If L is a left inverse of G\A— >B,
then for any x in the domain of G , L = L° ids =
L°(G°F) = (L°G)°F = id A °F = F.
101.5.3
a) a; i— x y/x .
b) x i— x x + 1 .
c) xi->x/2.
d) This one is its own inverse.
102.1.3 i fc 2 = 55 and f{^ = i k 2 = 14,400.
249
103 . 4.1 Basis: Ylk=i k(k+ 1) = \ • Eduction step:
n+l .. j n 1
' b(h +- 1 i f rj 4- 1 i (n -I- 9i '
^ jfe(jfe + l)
(n + l)(n + 2) 4-^fc(fc + l)
1 n
(n + l)(n + 2) n + l
1 + n(n + 2)
(n + 1) (n + 2)
n 2 + 2?r + 1
(n + 1) (n + 2)
(n + l) 2
(n + 1) (n + 2)
n + l
n + 2
as required.
103 . 4.2 Induction step: If n is even,
n+l n
E(-l) fcfc = -(»+!) + £(-!)**
k = 1
fe=l
= -(n + l) + -
= ^ (n — 2n — 2)
—n — 2
-(n + l + 1)
as required, and if n is odd,
n+l n
E(-l)+ = (n + l) + E(-l) fe fc
k— 1 fc=l
. n + l
= n + l —
n + l
105 . 2.1 1! = 1, and (n + l)! = (n+l)n! which by
the induction hypothesis is
{n + mi^k = JT k t\k
as required.
107 . 3.1 For n = 1 , this is l 2 — 0 = (— l) 2 . For the
induction step, suppose / 2 -/„_i/„+i = (-l) n+1 .
Then
/n+l - fnfn+2 = fn+1 ~ fn{fn + /n+l)
= fn+1 fn fnfn+1
fn+1 — fnfn + 1 .fn—lfn+1
fn + fn—lfn+1
The first three terms are /„+i(/ n +i - /„ - f n -i) ,
which is 0 by definition of the Fibonacci recurrence.
By the induction hypothesis, the last two terms are
(— 1)(— l) n+1 = (— l) n+2 as required.
109 . 8.2 The last entry of (a) is a, and the last
entry of cons(a,L) is the last entry of L.
110 . 4.2 (a) ‘0111010’. (b) ‘Oil’, (c) ‘Oil’,
(d) A. (e) ‘011011011’. (f) ‘011011011’.
110 . 4.3
EV.l The empty string A is a string in E .
EV.2 If w is a string in E then ‘aw’ , ‘ awb’ , ‘ bwa ’
and 'b'wb' are all strings in E.
EV.3 Every string in E is given by one of the pre-
ceding rules.
112 . 4.2
a) = {1,2, 3, 4, 5}, f+ = 0-
b) U^=(-3.-3), f+=(~ 1-1).
c) = (—1.. 3) — {1,2}, n^ = 0-
again as required.
104 . 4.1 Suppose d is a positive integer and d\p
and d\m. The only divisors of p are 1 and p . If
d = p, then p does not divide m . So the only possi-
bility is that d = 1 . Thus 1 is the largest divisor of
p and m, so GCD(p,m) = 1.
2
3
a)
-3
-6
-18
-72
-360
b)
14
30
55
c)
d)
11
18
e)
1
2
44
113 . 1.2 The set of positive divisors of 8 is
{1,2, 4, 8}. Let the bijection /3 required by Defi-
nition 113.1 be defined by: /3( 1) = 1, /3(2) = 2,
/?(3) =4, and /3(4) = 8.
113 . 5.1 x e+ x + 1 : N -+ N + is a bijection.
114 . 2.3 9 -10 -10 -10 = 9000.
114 . 2.4 9 -10 -10- 10 -5 = 45,000.
114 . 3.1 2 n — 1 .
114 . 3.2 F(n) = 3" .
114 . 3.3 G(0) = 0, G(l) = 1, and for n > 2,
G(n) = 3” -2 .
104 . 4.2
105 . 1.3
250
115 . 2.3 (a) 2” — 1 . (b) n. (c) 2 ^ .
116 . 2.3 Let Z the set of zinc pennies, B the set
of pennies minted before 1932, and A the set of pen-
nies that are neither zinc nor minted before 1932.
Let P be your whole collection. Then
\P\ = \Z\ + \B\ + \A\-\ZnB\-\ZnA\
-\AnB\ + \znAnB\
Since
\znA\ = \AnB\ = \AnBnz\ = o
we have
\P\ = 3 + 8+ |A| - \ZnB\
so you need to know the number of pennies that are
neither zinc nor minted before 1932 and the number
of zinc pennies minted before 1932. (In fact, all zinc
pennies were minted in 1943.)
117 . 1.13 All are partitions except (b) and (d).
Even though every element of S is an element of
exactly one set in (d), (d) is not a partition because
it contains the empty set as an element.
117 . 3.1 Let A = {1,3,5} and let 11= {A,Z-A}.
120 . 3.1 Every block of S/F must be a singleton.
120 . 4.1 Let F(l) = F( 2) = F(ir) = 42 and
F(x) = 41 for all other real numbers x.
121 . 2.1
a) /3 _f({1,3,5}) = 4; /? F ({4})=6; 0 F ({ 2})=0.
b) 0 f {A) = 3.
c) Pp({n}) = n for n£ A.
d) 0 F ({n}) = n 2 for n£A. (Observe that for
(c) and (d), A/F is the same set.)
e) 0 F ({1,2}) = -5; /3 F ({3}) = 1; /3 F ({4}) = 21;
/? f ({5}) = 55.
122 . 3.1 26 10 .
126 . 1.3
a) | (2,a), (2,c), (3, a), <3, c) , (3, rf) }
b) 0
c) | (2, c) , (3, c) , (3, d) , (3, e) , (4, c) , (4, d) , (4, e) j .
1 R 2 3?
1 R 3 3?
3R 2 1?
Y
N
Y
127 . 2.1
127 . 3.1 R x R — A : any two different real num-
bers are related.
127 . 3.4 By Definition 127.1, we must show that
C.l aU a op is symmetric.
C.2 a C aU a op .
C.3 If 7 is symmetric and a C 7, then aUa op C 7.
To prove C.l, suppose x(aU a op )y . Then xay and
xa op y, so ya op x and y(a op ) op x, that is, yax. So
y(aLia op )x. Hence aUa op is symmetric.
C.2 follows because for any sets S and T, S C
SUT. As for C.3, suppose 7 is symmetric and
a C 7 . Suppose xa op y . Then yax , so y^x because
a C 7. Since 7 is symmetric, xjy. Thus a^Cy.
We already know that aCy, so it follows that
a U a° p C 7 as required.
129 . 2.1 No; not symmetric.
129 . 2.2 No; not symmetric or transitive.
129 . 2.3 No. Not reflexive or transitive.
129 . 2.4 No. Not transitive.
129 . 3.1 No, not symmetric or transitive.
129 . 3.2 Yes. [0] E = [0..1) and [3] E = [3..4) .
129 . 3.3 Yes. [0] E = [0..1] and [3] E = {3}.
130 . 1.3 3, 27, 51, 75, 99.
130 . 4.4 a) 1. b) 5. c) 1.
131 . 1.3 F(6) = 0, F(n) = 1 otherwise. (There
are many answers.)
132 . 2.4 Here are all the possible values of E and
E ' :
E S/E
As U {(1,2), (2,1)} {{1,2}, {3}, {4}, {5}}
As U {(1, 2), (2, 1), (3,4), (4,3)} {{1,2}, {3, 4}, {5}}
A s U { (1, 2) , (2, 1) , (3, 5), (5, 3)} {{1,2}, {3, 5}, {4}}
As U {(1, 2), (2, 1), (4,5), (5,4)} {{1,2}, {3}, {4, 5}}
As U {(1,2), (2,1), (3,4), (3,5),
(5,4), (5,3), (4,5), (5,4)} {{1,2}, {3, 4, 5}}
135 . 3.2 We must show that a is antisymmetric,
transitive, and irreflexive. If aab and baa , this
contradicts the requirement that exactly one of the
statements in 135.3 holds unless a = b. Thus aab
and baa imply a=b , so a is antisymmetric, a
is transitive by assumption. Finally, for any a £ A,
a = a, so that rules out a a a, so a is irreflexive.
251
137 . 1.3
20
{1,2} {2,3}
10
25
/ \
/ 1
2 4
{1} {2}
2"
5
/
1 5
\ /
. /
(a)
(b)
(c)
(d)
137 . 1.4
Only (d).
139 . 1.5
Lexical ordering: 00,
01,
0101,
0111,
01111 , 10101 , 10111 , 110 , 111 .
Canonical ordering: 00, 01, 110, 111, 0101, 0111,
01111 , 10101 , 10111 .
140 . 3.2 ( a) max= 3, no min.
(b) no max, min= 0.
(c) max= 20, min= 1.
(d) max= 25, min= 1.
140 . 3.3 (a) max= 0, min= 1 (b) no max, min= 1
(c) no max, no min.
141 . 3.2 (a) sup = 5, ninf= 3
(b) sup= 60, inf= 1
(c) no sup, inf= 2
(d) sup= 0, inf= 1
(e) sup= {1,2,3}, inf= {2}
142 . 1.6 All except (f).
142 . 1.7 We will show that the infimum of any
two elements is the intersection. The proof for the
supremum is similar. By Theorem 141.2, we must
show
144 . 2.2
BCD
%/ \d
• ' a
(b)
3
146 . 2.3 (a) is simple. The relational definition of
(a) is:
G 0 = {A,B,C,D}
Gi = {(A,A),(B,A),(C,A),(D,A)}
147 . 2.2 There are six automorphisms of (a), rep-
resenting every possible way of permuting the set
{B,C, D} . There are two automorphisms of (b) (the
identity and the one that switches b and c.
148 . 1.3
1 2
3 ^4
148 . 1.4
Go = {1,2, 3, 4}
Gi. = {(1,2), (1,3), <1, 4), <2, 3), (2,4), (3,2), (3,4), (4,1)}
148 . 1.6
(left) a b c
a 0 1 1
6 10 1
c 0 0 0
(right) 12 3
10 11
2 10 1
3 0 0 0
149 . 4.3 (e) and (a,d) . Note that a path of length
n or more in a digraph with n nodes cannot be sim-
ple.
(i) If B C A and CCA, then B n C C B and
unc C G.
(ii) If B C A, CCA, DCB and D C G, then
DCBHC.
To see (i), suppose x £ BflC. By Definition 33.2,
page 47, x £ B and x £ C . Then by Definition 31.1,
B fl G C B and B fl G C G. For (ii), suppose x € D .
Then by assumption, x G B and x £ C . Then by
Definition 33.2, x £ B fl G. Hence D C B fl G.
151 . 3.1
(a) 1 of length 1 , 1 of length 2 , 2 of length 3 , 2
of length 4 .
(b) 1 each of length 1 and 2, 2 of length 3 and 4
of length 4 .
151 . 4.1 (a) x y z w (b) 2 1 1
a: 2 10 2 0 121
2/0400 001
^ 2 8 2 0
w 0 0 0 1
252
152 . 2.1 All but c and e.
153 . 3.2 b and d.
153 . 3.3 a and f are bipartite, a is complete bipar-
tite.
157 . 4.1 b and cl have diameter 1, f has diameter
3, the others have diameter 2.
157 . 4.2 a and b.
155 . 1.4 No pair of the graphs are isomorphic.
155 . 1.5 Map (b) to the triangle with horizontal
bottom edge (full) and to one of the triangles with
horizontal top edge (nonfull).
157 . 4.3 All of them.
158 . 1.4 All of them!
Bibliography
At the end of each entry, the pages on which that entry is cited are listed in
parentheses.
Bagchi, A. and C. Wells (1998). “On the communication of mathematical reasoning”.
PRIMUS , volume 8,
pages 15-27. Also available by web browser from URL:
http://www.cwru.edu/artsci/math/wells/
pub/papers . html. (117)
Bagchi, A. and C. Wells (1998). “Varieties of mathematical prose”. PRIMUS, volume
8, pages 116-136.
Also available by web browser from URL:
http://www.cwru.edu/artsci/math/wells/pub/papers.
html. (117)
Ebbinglraus, H. D., J. Flum, and W. Thomas (1984). Mathematical Logic. Springer-
Verlag.
Graham, R. L., D. E. Knuth, and O. Patashnik (1989). Concrete Mathematics. Addison-
Wesley.
Guy, R. (1981). Unsolved Problems in Number Theory. Springer- Verlag. (160)
Hofstadter, D. (1979). Godel, Escher, Bach: An Eternal Golden Braid. Basic Books,
(vi, 159)
Knuth, D. E. (1971). The Art. of Computer Programming, Volume 2. Addison- Wesley.
Lagarias, J. (1985). “The 3x +1 problem and its generalizations”. American
Mathematical Monthly ,
volume 92. (160)
Myerson, G. and A. J. van der Poorten (1995). “Some problems concerning recurrence
sequences”. American
Mathematical Monthly , volume 102. (163)
Raymond, E. S. (1991). The New Hacker’s Dictionary. The MIT Press.
Riesel, H. (1985). Prime Numbers and Computer Methods for Factorization.
Birklrauser.
Rosen, K. H. (1992). Elementary Number Theory and its Applications, Third Edition.
Addison- Wesley.
Skiena, S. (1990). Implementing Discrete Mathematics. Addison- Wesley.
Wells, C. (1995). “Communicating mathematics: Useful ideas from computer science”.
American Mathe-
matical Monthly, volume 102, pages 397-408. Also available by web browser from URL:
http://www.
cwru. edu/ artsci/math/wells/pub/papers .html. (117)
Wells, C. (1998). “Handbook of mathematical discourse”. URL:
http://www.cwru.edu/artsci/math/
wells/pub/papers .html. (117)
Wilder, R. L. (1965). Introduction to the Foundations of Mathematics. Second
Edition. John Wiley and
Sons. (35)
Wilf, H. (1990). Generatingfunctionology. Academic Press.
253
Index
The page number(s) in boldface indicate where the definition or basic explanation
of the word is found.
The other page numbers refer to examples and further information about the word.
1 -tuple, 51
absolute value, 138
abstract description
examples, 219
abstract description (of a
graph), 219
abstraction, 60, 73, 200, 219
addition, 11, 66, 67, 69-71, 97,
107, 163, 202
addition (of matrices), 228
addition of rational numbers,
11
adjacency matrix, 224 , 232
examples, 224
adjacent, 232
adjacent with multiplicity n,
232
affirming the hypothesis, 121
algebraic expression, 16 , 105
algorithm, 97
algorithm for addition, 97
algorithm for multiplication,
97, 98
AllFactors, 9
alphabet, 93 , 167
and, 21 , 22 , 24, 102, 108
examples, 21
anonymous notation, 64
antecedent, 36
antisymmetric, 79
examples, 79
antisymmetric closure, 199
application, 57
Archimedean property, 115
argument, 57
arrow, 218
associative, 70 , 71
associativity (in lattices), 216
automorphism, 224
axiomatic method, 217
barred arrow notation, 65
base, 94
examples, 99
basis step, 152
Bezout’s Lemma, 128-130, 156,
162
biconditional, 40
bijection, 136 , 149, 186
bijective, 136 , 187
examples, 136
binary notation, 95 , 97, 98
binary operation, 67 , 69
examples, 67, 70, 91
binary relation
examples, 74
binomial coefficient, 191, 191 ,
192
examples, 190, 192
bipartite graph, 233
bit, 95
block, 180 , 182
boldface, 4
Boolean variable, 104
bound (variable), 32, 64, 114
calculus, 107
canonical ordering, 212
examples, 212
cardinality, 173
examples, 173
carry, 97, 98
Cartesian powers, 54
Cartesian product, 52, 52 , 54,
177
examples, 52, 53, 74
Cartesian square, 54
CartesianPoduct, 54
centered division, 87
character, 93
characteristic function, 65
examples, 65
characterize, 85
chromatic number, 241
circuit, 236
class function, 183
examples, 183
closed interval, 31
closure, 197
codomain, 56 , 131
Collatz function, 160
color, 241
coloring, 241
commutative, 71
examples, 71
commutative diagram, 144
examples, 145
commutativity (in lattices),
216
complement, 48 , 108
examples, 67
complete bipartite graph, 233
complete graph, 233
complete graph on n nodes,
233
component (of a graph), 236
composite, 10 , 140
examples, 10
composite (of functions), 140
examples, 141, 142
composite (of relations), 195
examples, 195
composite integer, 10
composition, 195
composition (of functions) , 140
composition powers, 196
Comprehension, 28
comprehension, 27 , 29
concatenate (of lists), 166 , 168
conceptual proof, 193
conclusion, 36
conditional sentence, 36
congruence, 200
congruent (mod fc), 201
examples, 201, 203
conjunction, 21 , 103
connected, 236
connected component, 236
connected graph, 236
cons, 165
consequent, 36 , 121
constant function, 63
constructive, 130
contain, 45
contradiction, 107
contrapositive, 42
examples, 43, 120
Contrapositive Method, 120
contrapositive method, 120
254
255
converse, 42
examples, 42
coordinate, 49 , 143
coordinate function, 63 , 74
examples, 74
corollary, 1
countably infinite, 174
counterexample, 112 , 154
cycle, 236
decimal, 12 , 93
decimal expansion, 12
decimal representation, 12 , 14,
15
defined by induction, 159
defining condition, 27
definition, 1, 4 , 25
examples, 15
degree, 233
DeMorgan Law, 102
DeMorgan law
examples, 103, 105
denying the consequent, 121
dependent variable, 57
diagonal, 52 , 69
diameter, 237
digit, 14, 93
digraph, 74 , 218 , 222
Direct Method, 119
direct method, 119
directed circuit, 226
directed edge, 218
directed graph, 218
directed path, 226
directed walk, 225
disjoint, 47
disjunction, 21 , 103
distance, 237
distributive law, 110
div, 82
divide, 4 , 6, 8, 207
examples, 4
divides
examples, 5
DividesQ, 9
division, 4 , 87
division (of real numbers), 67
divisor, 5
domain, 56
dummy variable, 150
edge, 230
element, 25, 172
embedded in the plane, 239
empty function, 63
empty language, 169
empty list, 164
empty relation, 74
empty set, 33 , 34, 46, 63, 108
empty string, 168, 168
empty tuple, 51
equivalence, 40 , 122, 123
examples, 123, 200
equivalence class, 204
equivalence relation, 200 , 206
examples, 200
equivalent, 40 , 41, 42, 109
examples, 41, 42, 106
Euclidean algorithm, 92
Eulerian circuit, 237
evaluation, 57
even, 5 , 200
examples, 5, 10
example, 1
existential bigamy, 9
existential quantifier, 113
examples, 113, 115
existential statement, 5 , 113
exponent, 87
examples, 87
exponential notation, 54
exponential notation for
strings, 168
expression, 16 , 105
extension (of a function), 138
examples, 138
extension (of a predicate), 27 ,
55
examples, 28, 55
fact, 1
factor, 5 , 9
factorial function, 158 , 159,
189
Factorlnteger, 88
factorization, 87
fallacy, 121
examples, 121, 122
family of elements of, 140
family of sets, 171
examples, 171
Fibonacci function, 160
Fibonacci numbers, 161
field names, 140
finite, 173, 173 , 182, 187
examples, 173
finite set, 173, 173 , 187
first coordinate, 49
first coordinate function, 63
fixed point, 143
floor, 86
examples, 86
floored division, 87
formal language, 169
formula, 16
Forth, 69
Four Color Theorem, 242
fourtunate, 37
free variable, 32
full, 234
full subgraph, 234
Function, 65
function, 56, 56 , 57, 59, 60, 62,
63, 68, 75, 131, 184, 186
examples, 57, 58, 61, 63, 67
function as algorithm, 60
function set, 66, 67, 188
examples, 66
functional, 62
functional composition, 140
functional property, 62 , 75
functional relation, 75
functions in Mathematica, 58
Fundamental Theorem of
Arithmetic, 87 , 127
GCD, 88, 90-92, 125, 128, 164
examples, 88, 90-92, 128
GCD, 91
General Associative Law, 71
graph, 230
graph (of a function), 61
examples, 138
greatest common divisor, 88
greatest integer, 86
Hamiltonian circuit, 238
Hasse diagram, 210
examples, 210
head, 164
hexadecimal, 95
hexadecimal notation, 95 , 97
hypothesis, 36
idempotence (in lattices), 216
idempotent, 143
identifies, 205
256
identity, 72
examples, 72
identity (for a binary opera-
tion), 72
identity (predicate), 19
examples, 20
identity function, 63, 64, 65,
72, 141
examples, 64, 137
image, 131
examples, 131
image function, 132
image of a subset, 132
implication, 35 , 36 , 37-39, 41,
42, 107, 109, 119
examples, 36-39, 117, 118
implies, 107, 109, 119
incident, 232
include, 43, 44, 45, 77, 176,
207, 208
examples, 43, 63, 79, 207
inclusion, 79
inclusion and exclusion, 179
examples, 179
inclusion function, 63 , 138, 142
inclusive or, 22
indegree, 220
examples, 220
independent, 174
independent variable, 57
indexed by, 140
induction, 152 , 159, 175, 192
examples, 152, 153
induction hypothesis, 152
induction step, 152
inductive definition, 159
examples, 157, 158, 161
inductive proof, 152
infimum, 214, 214
examples, 214
infinite, 174 , 182
infinity symbol, 12
infix notation, 68
initial segment, 211
examples, 211
injection, 134
injective, 134 , 187, 189
examples, 134, 138
injective function, 187
input, 57
instance, 16
integer, 3 , 15, 87, 93, 127
examples, 3
integer variable, 18
IntegerQ, 15
integral linear combination,
127 , 129
examples, 127-129
interpolative, 196
intersect
examples, 171, 172
intersection, 47 , 67, 77, 108,
199, 217
examples, 47, 55, 77, 172,
178
intersection-closed, 199
interval, 31
examples, 31, 33
inverse function, 146
examples, 147
inverse image, 132
invertible, 146 , 149
irreflexive, 81
examples, 81
isomorphic, 235
isomorphism, 223 , 235
examples, 223
iterative, 157
join, 214
examples, 215
Kempe graph, 242
kernel equivalence, 203
examples, 203
Kuratowski’s Theorem, 240
labeling, 221
lambda notation, 64
examples, 64
language, 169
examples, 169, 170
lattice, 215
examples, 215
law, 19 , 39
law of the excluded middle,
106
LCM, 88, 90
LCM, 91
least common multiple, 88
least counterexample, 154
least significant digit, 94
least upper bound, 213
left cancellable, 150
left inverse, 146
lemma, 2
length, 236
length (of a list), 165
examples, 165
lexical order, 211
lexical ordering, 211
examples, 211
linear ordering, 208
List, 69
list, 27, 164
examples, 165
list constructor function, 165
list notation
examples, 26
list notation (for sets), 26 , 32
logical connective, 21 , 35
loop, 220
examples, 220
lower bound, 213
lower semilattice, 215 , 216
lowest terms, 11
examples, 11
mapping, 57
material conditional, 36
Mathematica, vi, 9, 10, 15, 16,
19, 21-23, 27, 31, 54, 58,
59, 62, 65, 68, 69, 84, 87,
88, 91, 96, 109, 151, 165
mathematical induction, 152 ,
175
matrix addition, 228
matrix multiplication, 227
max, 70 , 167, 215
maximum, 70, 167, 213, 213
meet, 214
examples, 215
member, 25
method, 2
min, 70 , 215
minimum, 70, 213
Mod, 84
mod, 82 , 204
modulus of congruence, 201
modus ponens, 40 , 109, 110
moiety, 233
more significant, 94
examples, 94
most significant digit, 94
multidigraph, 222
multigraph, 222, 231
257
multiplication, 11, 67, 69-72,
97, 107, 163, 202
multiplication (of matrices),
227
multiplication algorithm, 97,
98
Multiplication of Choices, 175
multiplication of rational
numbers, 11
multiplication table, 69
N, 15
NAND, 109
natural number, 3
near, 77
nearness relation, 77 , 78-80,
200
negation, 22 , 23
examples, 23, 102, 116
negative, 3
negative integer, 3
negative real number, 12
ninety-one function, 159
node, 218 , 230
nonconstructive, 130
nonempty list, 164
nonnegative, 3
nonnegative integer, 3
nontrivial subset, 45
NOR, 109
not, 22, 102, 108
null tuple, 51 , 54
number of elements of a finite
set, 173
examples, 173
octal notation, 94
odd, 5, 200
one to one, 134
one to one correspondence, 136
onto, 133
open interval, 31
open sentence, 16
opposite, 62 , 77 , 220
examples, 77
or, 21 , 22, 22 , 24, 102, 108
examples, 21
ordered pair, 49, 49 , 50
ordered set, 207
ordered triple, 50
ordering, 206
examples, 206-208
outdegree, 220
examples, 220
output, 57
P-closure, 197
pairwise disjoint, 180
palindrome, 169
parameter, 32
partial ordering, 207
partition, 180 , 181-185, 195,
204, 206, 237
examples, 180-183
Pascal, 26, 68, 87, 92, 93, 100,
104, 157, 164, 180, 201,
226
path, 236
permutation, 137
examples, 137
Perrin function, 161
Perrin pseudoprime, 161
Pigeonhole Principle, 189
examples, 189
planar, 239
Polish notation, 68
poset, 207
examples, 207
positive, 3
positive integer, 3
positive real number, 12
postfix notation, 68
power (of matrices), 228
power set
examples, 207
powerset, 46 , 74, 76, 77, 132,
133, 177, 207
examples, 46, 67, 76
predicate, 16 , 73, 105
examples, 16, 19, 20
predicate calculus, 113
prefix notation, 68
preorder, 209
preordered set, 209
preordering, 209
Prime, 10, 58
prime, 10, 10 , 58, 87, 127
examples, 10
prime factorization, 87, 92
examples, 87
PrimeQ, 10
Principle of Inclusion and
Exclusion, 179
examples, 179
Principle of mathematical
induction, 152
Principle of Multiplication of
Choices, 175
Principle of Strong Induction,
156
Principle of the Least Coun-
terexample, 154
Product, 151
product, 150, 150 , 153
examples, 150, 158
product (of matrices), 227
product (of matrices), 228
projection, 63 , 74, 143
proof, 2, 4, 4
proof by contradiction, 126
proper subset, 45
properly included, 44
proposition, 15 , 17, 104
examples, 15
propositional calculus, 107
propositional expression, 104
propositional form, 104
propositional variable, 104
quantifier, 20, 20 , 113
examples, 112, 115, 116, 118
Quotient, 84
quotient, 84, 156
quotient (of integers), 83
quotient set (of a function),
184
examples, 184
quotient set (of an equivalence
relation), 204 , 206
examples, 204
rabbit, 160
radix, 94
range, 131
range expression, 151
rational, 11, 126
examples, 11, 13, 14
rational number, 11, 11 , 12,
14, 15
addition, 11
multiplication, 11
representation, 11
reachability matrix, 230
reachable, 229
real number, 12, 12 , 13-15, 22,
115
258
real variable, 18
realizations, 96
recurrence, 161
recurrence relation, 161 , 189
examples, 191
recursive, 157 , 164
recursive definition, 157 , 159
examples, 157, 159, 160, 163,
164, 170
reductio ad absurdum, 126
reflexive, 77
examples, 77
reflexive closure, 197, 197
relation, 73 , 74, 76, 77
examples, 74, 75, 195
relation on, 75
relational database, 139
relational description, 222
relational symbols, 16
relatively prime, 89
examples, 89
remainder, 83 , 84, 92, 156, 182,
184
examples, 184
remainder function, 203
remark, 2
representation, 15 , 96
representation (of a rational
number), 11
representation (of a set), 26
restriction, 137 , 142
examples, 138
reverse Polish notation, 68
right band, 67 , 70, 72
right cancellable, 150
right inverse, 146
rule of inference, 24 , 25, 39,
110
examples, 24, 25, 39, 40, 43,
46, 110, 125, 147, 152, 213
Russell’s Paradox, 35
scalar product, 227
scandalous theorem, 126
second coordinate, 49
second coordinate function, 63
Select, 31
semicolons in Mathematica, 59
sentence, 15
set, 25 , 32 , 35, 172, 174
examples, 25-28, 33, 34
set difference, 48
examples, 48
set of all sets, 35 , 48
set of functions
examples, 140
setbuilder notation, 27 , 29, 35
examples, 27-29, 33
sets of numbers, 25
sex, 161
shift function, 188
shoe-sock theorem, 148
show, 2
significant figures, 12
simple, 231
simple digraph, 221
simple directed path, 226
simple graph, 231
simple path, 236
single- valued, 61
singleton, 34
singleton set, 34
sister relation, 77 , 78, 80
solution set, 28
solve (a recurrence relation),
161
sorting, 143
source, 218
specification, 2
square root symbol, 12
statement, 19
strict ordering, 206
examples, 206
strict total ordering, 208
string, 93 , 167
examples, 167
StringLength, 58
strong induction, 155
subdivision, 240
subgraph, 234
subset, 43 , 45, 190
substitution, 17
subtraction, 67, 68, 70, 71
successor function, 163
Sum, 151
sum, 150, 150 , 153
examples, 150, 158
supremum, 213 , 214
examples, 214
surjection, 133
surjective, 133 , 187
examples, 133, 138
Swedish rock group, 170
symmetric, 78 , 124, 232
examples, 78
symmetric closure, 197
symmetric matrix, 232
Table, 27, 31
tail, 164
take, 57
target, 218
tautology, 105
examples, 106
Tautology Theorem, 110
terrible idea, 45
theorem, 2
total ordering, 208
examples, 208
total relation, 74
transitive, 80 , 196, 227
examples, 80
transitive (digraph), 227
transitive closure, 198
transitivity (of implication),
109
trichotomy, 208
trunc, 86, 86
examples, 86
truth table, 22
TruthTable, 23
tuple, 50, 50 , 52, 138, 139 ,
140
examples, 51, 139, 140
tuple as function, 138
turnstile, 24
type (of a variable), 17 , 25, 26,
29, 104
unary operation, 67
examples, 67
under, 57 , 132
union, 47 , 67, 77, 108, 217
examples, 47, 77, 169, 171,
172, 178, 233
unit interval, 29
unity, 72
Universal Generalization, 6
universal generalization, 6
Universal Instantiation, 7
universal instantiation, 7
universal quantifier, 112 , 154
examples, 112, 115, 118
universal set, 48 , 108
universally true, 19 , 39
examples, 19, 20
259
upper bound, 212
examples, 212
upper semilattice, 215 , 216
usage, 2
utility graph, 240
vacuous, 37
vacuously true, 37
examples, 37
valid (rule of inference), 24
value, 56 , 57
value (of a function), 56, 59, 60
variable, 8, 16, 17
vertex, 218 , 230
vertices, 218
walk, 236
warning, 2
weak ordering, 206
examples, 206
weight function, 221
well-defined, 85
well-ordered, 154
witness, 113
Xor, 22
xor, 22
yields, 24
Zermelo-Frankel set theory, 35
zero, 3-5, 33
260
Index of Symbols
23, 158
to | n 4
:= 59
-n 22, 102, 108
(A, a) 207
to mod n 83
X 65
NOR 109
(a..b) 171
n! 158, 189
Xi 65
M 77
(x)F 68
n (mod k) 203
= 201
N+ 25
/ 5
PAQ 21
cons 165
o 140, 195
// 69
P <=>Q 40
A 69
C 45
0 4, 5
P => Q 36
52, 77
V 21, 22, 23, 24,
() 51, 225
pvg 21
div 82
102, 108
{a, b) 49
pop 62
1 4
A 48
(&i)i£ n 51
K 63, 74
0 33, 63, 108
n 180
(xi,...,x n ) 51
R n 184
4A 40, 109
n 205
A' 48
5/P 204
3 113
n?=i 150
A-B 48
w n 168
(3 x:Q)(x) 113
nLi 158
A/F 184
xF 68
e p 87
PA 46
A\B 48
x i-a f(x) 65
E n 205
V 46
A c 108
[a., b] 31
floor 86
Q 25
AcB 26
[a] 183
V 20, 26, 112
R 25, 52
AC B 43
[r] 86
B A 66
Rel(A,P) 74
AnB 47
[x] 180, 205
r 6i
R+ 25
Ac B 45
[x] E 204
r(p). 6i
R++ 25
A(^B 44
[x] n 205
GCD 88, 128, 164
{x | P(x)} 27
Ax B 52
& 65
I 29
44
AuB 47
&& 21
kU 63
‘cat’ 93
flVfc 215
\A\ 173
=> 36, 109
rr=i 150
aAb 215
a 73
G 26, 80
EL i 158
A* 165, 211
a* 76
C 43
sup 213
A+ 165
cxf 76
oo 12
els 183
A c 48
a° p 195
n 47, 108
x 52
A n 54
a op 77, 124, 207
A 211
-> 57
B A 188
a" 196
A x.f(x) 64
trunc 86
C(n,k) 190
197
A 51, 168
U 47
C b 63
a s 197
A 64
U 108
C r 182
a T 198
LCM 88
U 48, 108
F(a) 57
(Vx:Z)P(x) 26
\A\ 187
h 24
F-.A-+B 57
(Vx)P(x) 112
< 206
V 215
F~ l 147
(Vx)Q(x) 112
[rj 86
A 215
F* 133
A 21, 22, 102, 108
< 206
Z 25
F~ l 132, 184
^ 65
max 70
Z/n 184
G° F 140
Pf 186
min 70
Z„ 182
i:A—>B 142
171
mod 82
{xi, — ,x n } 26
K(F) 203
mu ^ 171
N 25
| 27
m = n 201
IJJ 17 171
n 50, 173
II 21
to div n 83
Uti A i 171
NAND 109
You might also like
- Canadian MO 1969-2020, Solutions From 1994 - 2020No ratings yetCanadian MO 1969-2020, Solutions From 1994 - 2020245 pages
- Klappenecker A., Lee H. Discrete Structures 2025No ratings yetKlappenecker A., Lee H. Discrete Structures 2025429 pages
- Discrete Mathematics With Graph Theory by Edgar G. Goodaire and Michael M. Parmenter83% (36)Discrete Mathematics With Graph Theory by Edgar G. Goodaire and Michael M. Parmenter557 pages
- NMP Grade 8 2024-LS-G8-NMP - Mathematics - Q1 - W3 - D1No ratings yetNMP Grade 8 2024-LS-G8-NMP - Mathematics - Q1 - W3 - D111 pages
- GIAM. A Gentle Introduction To The Art of Mathematics - FieldsNo ratings yetGIAM. A Gentle Introduction To The Art of Mathematics - Fields436 pages
- First Set of Questions (1-10) - HISTORY OF MATHEMATICS100% (7)First Set of Questions (1-10) - HISTORY OF MATHEMATICS5 pages
- Gen Math11 - Q1 - Mod10 - Solving Real Life Problems Involving Rational - 08082020 182% (56)Gen Math11 - Q1 - Mod10 - Solving Real Life Problems Involving Rational - 08082020 129 pages
- A Gentle Introduction To The Art of Mathematics100% (2)A Gentle Introduction To The Art of Mathematics464 pages
- Edgar G. Goodaire, Michael M. Parmenter, Edgar G Goodaire, Michael M Parmenter-Discrete Mathematics With Graph Theory - Prentice Hall (2001)100% (5)Edgar G. Goodaire, Michael M. Parmenter, Edgar G Goodaire, Michael M Parmenter-Discrete Mathematics With Graph Theory - Prentice Hall (2001)557 pages
- Lesson2.1 - Integers and Absolute ValueNo ratings yetLesson2.1 - Integers and Absolute Value16 pages
- CBSE Sample Papers For Class 2 Maths - Mock Paper 20% (1)CBSE Sample Papers For Class 2 Maths - Mock Paper 28 pages
- Florida State University Course Notes MAD 2104 Discrete Mathematics INo ratings yetFlorida State University Course Notes MAD 2104 Discrete Mathematics I216 pages
- Discrete Mathematics: Thomas Goller January 2013No ratings yetDiscrete Mathematics: Thomas Goller January 201387 pages
- Discrete Mathematics With Graph Theory Goodaire&ParmenerNo ratings yetDiscrete Mathematics With Graph Theory Goodaire&Parmener557 pages
- Draft: Lecture Notes On Discrete MathematicsNo ratings yetDraft: Lecture Notes On Discrete Mathematics209 pages
- Introduction To Number Theory 2nd Edition Anthony Vazzana 2024 Scribd DownloadNo ratings yetIntroduction To Number Theory 2nd Edition Anthony Vazzana 2024 Scribd Download67 pages
- Building Blocks For Theoretical Computer ScienceNo ratings yetBuilding Blocks For Theoretical Computer Science268 pages
- Discrete Structures, Logic, and Computability: Student Study GuideNo ratings yetDiscrete Structures, Logic, and Computability: Student Study Guide161 pages
- Discrete Mathematics Discrete MathematicsNo ratings yetDiscrete Mathematics Discrete Mathematics90 pages
- Mathematics For Junior High Schools: September 2012No ratings yetMathematics For Junior High Schools: September 2012121 pages
- 1 638468101749646329 3143 01 Ms 6rp Afp tcm143700722No ratings yet1 638468101749646329 3143 01 Ms 6rp Afp tcm14370072221 pages
- CIS160 Mathematical Foundations of Computer Science Some NotesNo ratings yetCIS160 Mathematical Foundations of Computer Science Some Notes18 pages
- Lab Manual: Programming and Problem Solving LABNo ratings yetLab Manual: Programming and Problem Solving LAB20 pages
- Arithmetic Sequence and It'S Application in Real LifeNo ratings yetArithmetic Sequence and It'S Application in Real Life3 pages
- The Olympiad Corner No. 211 - Canadian Mathematical Society (PDFDrive)No ratings yetThe Olympiad Corner No. 211 - Canadian Mathematical Society (PDFDrive)14 pages
- Ntse Number Theory Divisibility Rule Assignment SolutionsNo ratings yetNtse Number Theory Divisibility Rule Assignment Solutions3 pages
- CS 61A Structure and Interpretation of Computer Programs Fall 2019No ratings yetCS 61A Structure and Interpretation of Computer Programs Fall 20196 pages
- Mathematical Handbook for Scientists and Engineers: Definitions, Theorems, and Formulas for Reference and ReviewFrom EverandMathematical Handbook for Scientists and Engineers: Definitions, Theorems, and Formulas for Reference and Review4/5 (4)
