MAS162

Foundations of Discrete Mathematics

UNIT NOTES
(TJD 2023)

Unit Coordinator:
Dr Philip Schrader
School Mathematics & Statistics, Chemistry and Physics

c Published by Murdoch University, Perth, Western Australia, 2022.

©

Revised by Philip Schrader 2022. Based on earlier versions prepared by various colleagues including Dr Amy Glen, A/Prof
Gerd Schröder-Turk, Dr Mark Lukas, and others.

This publication is copyright. Except as permitted by the Copyright Act, no part of it may in any form or by any electronic,
mechanical, photocopying, recording or any other means be reproduced, stored in a retrieval system or be broadcast or
transmitted without the prior written permission of the publisher.
PREVIOUS UNIT COORDINATORS

M163: Discrete Mathematics

1987 Dr Ken Harrison
1988 Dr Ken Harrison (First Semester)
Dr Ute Mueller (Second Semester)
1989 Dr Ken Harrison (First Semester)
Dr Ute Mueller (Second Semester)
1990 Dr Ute Mueller (First Semester)
Dr Ken Harrison (Second Semester)
1991 Dr Ken Harrison
1992 Dr Mark Short (First Semester)
Dr Ken Harrison (Second Semester)
1993 Dr Mark Short (First Semester)
Dr Geoff Yeo (Second Semester)
1994 Dr Mark Short (First Semester)
Dr Tim Stokes (Second Semester)
1995 Dr Tim Stokes
1996 Dr Ken Harrison
1997 Dr Ken Harrison (First Semester)
Dr Tim Stokes (Second Semester)

MAS167: Computational Mathematics (previously known as M163: Discrete Mathematics)

1998 Dr Mark Lukas (First Semester)

Dr Tim Stokes (Second Semester)
1999 Dr Mark Lukas
2000 Dr Mark Lukas (First Semester)
A/ Prof Ken Harrison (Second Semester)
2001 Dr Mark Lukas
2002 Dr Mark Lukas (First Semester)
A/ Prof Ken Harrison (Second Semester)
2003 A/ Prof Ken Harrison
2004 Dr Mark Lukas
2005 Dr Mark Lukas
2006 Dr Mark Lukas
2007 Dr Mark Lukas (First Semester)
Prof Walter Bloom (Second Semester)
2008 Dr Mark Lukas
2009 Ha Nguyen (First Semester)
Dr Mark Lukas (Second Semester)
2010 Dr Mark Lukas (First Semester)
Dr Amy Glen (Second Semester)
2011 Dr Amy Glen (First Semester)
Dr Mark Lukas (Second Semester)
2012 Dr Amy Glen (First Semester)
Dr Mark Lukas (Second Semester)
2013 Dr Amy Glen (First Semester)
Dr Duncan Farrow (Second Semester)

MAS162: Foundations of Discrete Mathematics (previously MAS167: Computational Mathematics)

 3

2014 Dr Duncan Farrow (First Semester)

Mrs Ha Nguyen (Second Semester)
2015 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2016 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2017 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2018 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2019 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2020 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
2021 Dr Amy Glen (First Semester)
Dr Gerd Schröder-Turk (Second Semester)
Contents

1 Recurrence Relations 1
1.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Introduction to MATLAB 11
2.1 Using MATLAB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Scalars, vectors and matrices . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Plotting in two dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Plotting in three dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 For loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.6 Selection statements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.7 User-defined functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3 Computer Arithmetic 44
3.1 Numbers and their representations . . . . . . . . . . . . . . . . . . . . . . 44
3.1.1 Base conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.2 Addition in other bases . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Scientific notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Computer storage of numbers . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.4 Subtractive cancellation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.5 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4 Counting & Discrete Probability 60

4.1 Basic Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.1 Sets & Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.1.2 Universal Set & Empty Set . . . . . . . . . . . . . . . . . . . . . . 62
4.1.3 Set Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.4 Venn Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.1.5 Algebra of Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.1.6 Finite Sets & Counting . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1.7 Inclusion-Exclusion Principle . . . . . . . . . . . . . . . . . . . . . . 68

i
ii CONTENTS

4.2 Basic Counting Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

4.2.1 Product & Sum Rules . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2.2 Factorials & Binomial Coefficients . . . . . . . . . . . . . . . . . . . 70
4.2.3 Pascal’s Triangle & Binomial Theorem . . . . . . . . . . . . . . . . 71
4.2.4 Permutations & Combinations . . . . . . . . . . . . . . . . . . . . . 72
4.2.5 Permutations Without Repetitions . . . . . . . . . . . . . . . . . . 73
4.2.6 Permutations With Repetitions . . . . . . . . . . . . . . . . . . . . 75
4.2.7 Ordered Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.8 Combinations Without Repetitions . . . . . . . . . . . . . . . . . . 76
4.2.9 Combinations With Repetitions . . . . . . . . . . . . . . . . . . . . 77
4.2.10 The Pigeonhole Principle . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3 Discrete Probability Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.3.1 Terminology & Notation . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3.2 Rules for Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.3 Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.4 Probability Distribution of a Discrete Random Variable . . . . . . . 86
4.3.5 Transforming a Discrete Random Variable . . . . . . . . . . . . . . 87
4.4 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5 Matrices 94
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.2 Matrix arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.1 Addition and subtraction . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2.2 Multiplication by scalars . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.3 Matrix multiplication . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.2.4 Matrix operations in MATLAB . . . . . . . . . . . . . . . . . . . . 99
5.3 Some special matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.4 Further matrix operations . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4.1 Inverses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4.2 Division . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.4.3 Powers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.4.4 Transposes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.4.5 Matrix functions in MATLAB . . . . . . . . . . . . . . . . . . . . . 105
5.5 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

6 Linear Transformations of the Plane 110

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.2 Special types of transformations . . . . . . . . . . . . . . . . . . . . . . . . 116
6.2.1 Dilations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
CONTENTS iii

6.2.2 Shears . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6.2.3 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.2.4 Reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.3 Degenerate transformations . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.4 Composition of transformations . . . . . . . . . . . . . . . . . . . . . . . . 125
6.5 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

7 Propositional Logic 131

7.1 Propositions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.2 The basic connectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
7.3 Truth tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.4 More connectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.5 Tautologies and contradictions . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.6 Logical equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.7 Logical implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.8 Rules of inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
7.9 Testing validity of inferences . . . . . . . . . . . . . . . . . . . . . . . . . . 156
7.10 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

8 Predicate Logic 166

8.1 Truth sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
8.2 Quantifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
8.3 Proofs and counterexamples . . . . . . . . . . . . . . . . . . . . . . . . . . 170
8.4 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

9 Mathematical Induction 173

9.1 Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

10 Boolean Algebra 182

10.1 Definition and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
10.2 Proofs in Boolean algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.3 Some more Boolean operations . . . . . . . . . . . . . . . . . . . . . . . . 192
10.4 Boolean expressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
10.5 Disjunctive normal form . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
10.6 Minimal sum of products form . . . . . . . . . . . . . . . . . . . . . . . . . 196
10.7 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
iv

11 Logic Networks 210

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
11.2 Adders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
11.3 Solutions to exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
Chapter 1

Recurrence Relations

1.1 Sequences
A sequence can be defined informally as a list of objects arranged in some definite order,
for example
1, 3, 5, 7, 9, 11, 13, . . .
The objects in a sequence are called the terms, and are usually denoted by symbols
such as t1 , t2 , t3 , and so on. We shall be dealing mainly with numerical sequences,
that is sequences whose terms are numbers, but there are other types of sequences. For
example, it makes sense to consider sequences of sets, sequences of triangles, or sequences
of functions.
The subscripts 1, 2, 3, . . . indicate the position of that particular term in the sequence.
Thus t1 is the first term, t2 is the second term, and so on. The symbol tn denotes the nth
term, commonly called the general term of the sequence. The sequence whose nth term
is tn is usually denoted by (tn ), or by t1 , t2 , t3 , . . . .

◮ Example 1.1 In the sequence 1, 3, 5, 7, . . . shown above we have t1 = 1, t2 = 3,

t3 = 5, and so on. A formula for the general term is tn = 2n − 1. ◭

Sequences may be specified in one of three ways. We illustrate each way with reference
to Example 1.1.

(a) One way is simply to write out the first few terms followed by three dots, for example,
1, 3, 5, 7, 9, . . . , but this method is only useful if these first few terms are sufficient
to establish the pattern of the sequence. Otherwise, they just look like a bunch of
puzzling numbers.

(b) Another method is to give a formula for the general term, for example, tn =
2n − 1. This explains how to calculate any term in the sequence. For example,

1
 2 Chapter 1. Recurrence Relations

t27 = 2 × 27 − 1 = 53. The only disadvantage of this method is that sometimes

a sequence with a simple pattern needs a complicated formula for tn . We will see
examples of this later in the chapter.

(c) A third method is to describe how to construct the list of terms, by starting with
the first one t1 , and working up the sequence progressively. This requires a formula
for calculating any particular term from the ones that come before it. For example,

t1 = 1
tn = tn−1 + 2 (for n ≥ 2).

This means that for each n ≥ 2, we compute tn by adding 2 to tn−1 . This formula
doesn’t work for t1 because there is no t0 ; that’s why we have to specify t1 explicitly.

The main purpose of this chapter is to study the third method of defining a sequence.
This method is an example of a recursive definition. A function that is defined recur-
sively is recognisable because its own definition refers back to earlier values of itself. For
example, the function f which computes the sequence of Example 1.1 can be recursively
defined by:
f (1) = 1;
f (n) = f (n − 1) + 2 (for n # 2).
Recursive functions are very common in computer programming, and can be very
complicated. However, in this unit we will usually only deal with sequences. Here is the
definition that we’ll use:

A sequence is said to be defined recursively if:

(B) the first one or first few terms are specified, and
(R) the remaining terms are defined in terms of previous ones.

It is important to realise that a recursive definition of a sequence has two parts, and
that neither part is sufficient on its own. The first part (B) is called the basis, and the
second part (R) is known as the recurrence relation or recurrence formula. The
basis describes how the sequence starts, and the recurrence relation describes how the
sequence continues once it is started. For this reason the specifications in the basis are
also known as the initial conditions of the recurrence relation.

◮ Example 1.2 In 1202 Leonardo of Pisa, also known simply as Fibonacci, introduced the
sequence
1, 1, 2, 3, 5, 8, 13, 21, 34, . . .
where each term is the sum of the previous two. This is now called the Fibonacci
sequence and has become very famous because it arises so often in mathematics and
 Section 1.1. Sequences 3

nature. A recursive definition of the sequence is:

(B) F1 = 1, F2 = 1;
(R) Fn = Fn−1 + Fn−2 (n # 3).

We can use (R) to calculate F3 , then F4 , then F5 and so on. In this way we could calculate
as many terms of the Fibonacci sequence as we want. Although we are not given a specific
formula for particular terms such as F1000 , we accept that we could calculate this, or any
other term, by using (B) and repeated use of (R). ◭

It is also useful to note that the recursive part of any recursive definition can be
expressed in more than one way. For example, the recursive part of the definition of the
Fibonacci sequence could be written as

Fn+1 = Fn + Fn−1 (n # 2).

It is important to understand that both this formula and the one given earlier say exactly
the same thing, namely that each term in the Fibonacci sequence (from the third term
onwards) is the sum of the two before it. All that differs is the value of n that we must
use to get a specific instance of this rule.
The recursive method of defining sequences often turns out to be more efficient than
direct methods. The next example shows this.

◮ Example 1.3 The factorial sequence (n!) can be defined

(a) directly by the formula n! = n × (n − 1) × · · · × 3 × 2 × 1, or
(b) recursively by 1! = 1, and n! = n × (n − 1)! for n # 2.
The first few values of n! can be calculated directly by multiplication:

1! = 1 = 1
2! = 1×2 = 2
3! = 1×2×3 = 6
4! = 1×2×3×4 = 24
5! = 1×2×3×4×5 = 120

However it soon becomes clear to anyone wanting to compile a list of values of n! (up to
n = 10, say) that the recursive definition is much quicker, because to go from (n−1)! to n!,
only one multiplication is required. Factorials will be further discussed in Section 4.2.2. ◭

We often wish to find formulas for the general terms of recursively defined sequences.
Sometimes this is easy to do, but not always. Some apparently simple recurrence formulas
define sequences which appear to be quite erratic and unpredictable.
 4 Chapter 1. Recurrence Relations

◮ Example 1.4 Consider the sequence (tn ) recursively defined by

t1 = 3; tn = tn−1 + 5 (n # 2).

According to the recursive formula for (tn ), the difference between successive terms in the
sequence t1 , t2 , t3 , . . . is always 5. This is an example of an arithmetic progression.
In fact the recursive definition of any arithmetic progression has the form

(B) t1 = a, and
(R) tn = tn−1 + d (n # 2),

where a is the first term and d is the common difference. As you probably learnt at school,
the formula for the general term of an arithmetic sequence is

tn = d × (n − 1) + a.

Arithmetic progressions arise in situations where a fixed amount is being added to a total
at successive stages, for example in bank balances where simple interest is being added
periodically. ◭

◮ Example 1.5 Consider the sequence (tn ) defined recursively by

t1 = 3; tn = 5tn−1 (n # 2).

The ratio between successive terms in the sequence t1 , t2 , t3 , . . . is always 5. This is an

example of a geometric progression. The recursive formula for any geometric sequence
has the form
(B) t1 = a, and
(R) tn = r · tn−1 n # 2
where a is the first term and r is the common ratio between successive terms. Once again,
there is a fairly easy formula for the general term. It is

tn = a · rn−1 (n # 1).

Geometric progressions are very common in mathematical models of the physical

world, and the reason for this is in the recursive definition. This is the mathematical
description of situations where the amount at any one time period is proportional to the
amount at the previous time period, for example in compound interest situations or simple
population models where the birth rate is proportional the current population size. ◭
 Section 1.1. Sequences 5

◮ Example 1.6 Consider the following recursively defined sequences:

1
s1 = 1; sn = 1 +
sn−1
t1 = 2; tn = (tn−1 )2
u1 = 1; un = 2un−1
v1 = 2; vn = (vn−1 )vn−1
1
w1 = 1; wn = wn−1 + 4
n
1! 59 "
x1 = 1; xn = xn−1 +
2 xn−1
Calculating the first few terms of the sequence s1 , s2 , s3 , . . . produces the numbers 1, 2,
3 5 8 13
, , , 8 , . . . , and we recognise the numerators and denominators in this sequence as
2 3 5
successive Fibonacci numbers. So it would seem that sn = Fn+1 /Fn for each n # 1. It
is possible to prove this using a method known as mathematical induction. This type of
proof is very recursive in nature, in fact very similar to recursive definitions of sequences.
You will meet mathematical induction in a later chapter.
Sequences (tn ), (un ) and (vn ) demonstrate that seemingly innocent recurrence formulas
can lead to very fast-growing sequences. Try to calculate v4 on any available calculator
and see what happens!
There are no simple formulas for the general terms in sequences (wn ) and (xn ), but
a little experimentation shows that the terms wn and xn both converge to limits quite
rapidly as n increases. These sequences have their uses in numerical work. As a hint to
what these might be, compute the terms w2 , w3 , w4 , . . . and x2 , x3 , x4 , . . . up to the stage
where the differences between wn and wn−1 , and between xn and xn−1 , are negligible. Now
calculate x2n and (90wn )1/4 , and study your answers.
To appreciate the usefulness of the recurrence formula for sequence (xn ), repeat this
experiment, first using a different number for x1 , the starting value, and then replacing 59
by a different positive number such as 289 or 5041. Don’t be concerned about trying to
understand why these sequences have the properties that you have just discovered. That
is not possible without a knowledge of calculus. Our purpose here is simply to meet a
variety of sequences, and to gain confidence in using recurrence formulas. ◭

So far we have considered sequences whose terms have positions labelled 1, 2, 3, . . . .

However it is also very common to have sequences that begin with a term whose starting
term is labelled with a 0. We designate this term t0 . Apart from this minor labelling
difference, such sequences are no different from the ones we have been considering.

◮ Example 1.7 Write the first three terms of the sequence defined by t0 = 3 and tn =
2tn−1 − 2 for n # 1.
Solution. t0 = 3, t1 = 2 · 3 − 2 = 4, and t2 = 2 · 4 − 2 = 6. ◭
 6 Chapter 1. Recurrence Relations

Sometimes it is useful to have a recurrence formula for a sequence even though you
know a formula for the general term. (Such a general formula could be complicated
whereas a recurrence formula might be simple and show up something important.) The
next examples show two methods for finding a recurrence formula.

◮ Example 1.8 Suppose we want a recurrence formula for the sequence (sn ) where sn =
2n2 − 5 for n # 0. We take the difference of successive terms:
! "
sn − sn−1 = 2n2 − 5 − 2(n − 1)2 − 5
= 2n2 − 5 − 2(n2 − 2n + 1) + 5
= 2n2 − 2n2 + 4n − 2
= 4n − 2

Therefore a recurrence relation is sn = sn−1 + 4n − 2. ◭

◮ Example 1.9 Suppose we want a recurrence formula for the sequence (sn ) where sn = n
n+1
for n # 1. We take the quotient of successive terms:
#
sn n n−1
=
sn−1 n+1 n
n n
= ·
n+1 n−1
n2
=
n2 − 1
n2
Therefore a recurrence relation is sn = n2 −1
sn−1 . ◭

Note that generally one of these methods provides a “nicer” answer than the other,
but either method produces a correct answer.
We shall soon be looking at ways a computer can be programmed to add a set of
numbers. If these numbers occur as terms of a sequence, then a recursive approach will
be most appropriate.
Suppose that we have a sequence of numbers (tn ), and we construct another sequence
(sn ) recursively as follows:

(B) s0 = 0, and
(R) sn = sn−1 + tn for n # 1.
It is easy to check using the recursive formula repeatedly that

s1 = s0 + t 1 = t1 ,
s2 = s1 + t 2 = t1 + t2 ,
s3 = s2 + t 3 = t1 + t2 + t3 ,
s4 = s3 + t 4 = t1 + t2 + t3 + t4 ,
s5 = s4 + t 5 = t1 + t2 + t3 + t4 + t5 ,
 Section 1.1. Sequences 7

and it is clear that the general rule is that sn is always the sum of the first n terms of the
sequence (tn ).

◮ Example 1.10 If (sn ) is defined recursively by s0 = 0 and sn = sn−1 + n, then sn is the

sum of the first n natural numbers, that is,

sn = 1 + 2 + 3 + · · · + n

for each n # 1. ◭

◮ Example 1.11 Consider the sequence (pn ) defined recursively in terms of a given sequence
(tn ) by the rule
(B) p0 = 1, and
(R) pn = pn−1 × tn for n # 1.
It is easy to check that each pn is the product of the first n terms of the sequence (tn ).
In particular, if tn = n for each n # 1, then (pn ) is just the sequence of factorials (n!). ◭

In many relevant applied problems, a sequence will be defined in words and from the
given information it is important to be able to formulate its recursive definition. To help
in this formulation, it is useful to draw a time-line showing when the given events occur.

◮ Example 1.12 A bank pays compound interest of 5% per year (calculated daily but com-
pounded yearly) on a certain type of account. Suppose that Joan opens such an account
with an initial amount of $1000 and deposits $100 at the end of each year thereafter (just
after the interest is compounded). Find a recursive definition for the amount An dollars
in the account at the beginning of year n, n # 1. Find the values of A1 , A2 and A3 .
A time-line representing the problem is given below.

deposit: 100 100 100

amount: A1 A2 A3 An−1 An
5% 5% 5%
❄ ❄
❄ ❄
❄ "❆ ❄ ❄
❄
" ✲
❆❆"
"
year: 1 2 n−1 n

Clearly, by definition, A1 = 1000. At the end of year 1, the bank will pay $0.05 × 1000
in interest and Joan will deposit $100, so the amount A2 at the beginning of year 2 will
be A2 = 1000 + (0.05)1000 + 100 = (1.05)1000 + 100 = 1150. Now, considering year n − 1
and year n, we find that (see the time-line)

An = An−1 + (0.05)An−1 + 100 = (1.05)An−1 + 100 for n # 2.

Together with A1 = 1000, this is the recursive definition for An . Using this definition, we
get A1 = 1000, A2 = 1150 and A3 = 1307.50. ◭
8 Chapter 1. Recurrence Relations

Exercises
Some of the questions for this section ask you to find a formula for the general term
of a sequence. You have not been taught any methods for doing this yet, so you
should play with these questions in any ways you can think of. They are designed
to build up your intuition about patterns of numbers. (The best way to see patterns
is to write out the first few terms of a sequence.)

1. Find a formula for the nth term tn of the sequences defined as follows:

(a) t1 = 1; tn = tn−1 + 1 (n # 2).

n−1
(b) s1 = 1; sn = sn−1 (n # 2).
n
(c) t1 = 1; tn = t1 + t2 + · · · + tn−1 (n # 2).
(d) s0 = 1; sn = −sn−1 (n # 1).
(e) t1 = 1; tn+1 = 2tn (n # 1).
(f ) t1 = 1; tn+1 = tn + 2n + 1 (n # 1).

2. Give a recursive definition of (tn ), if the formula for the general term tn is:

(a) tn = 3n (n # 1).
1 × 3 × 5 × · · · × (2n − 1)
(b) tn = (n # 1).
2 × 4 × 6 × · · · × (2n)
(c) tn = n2 (n # 0).
xn
(d) tn = (n # 1).
n!
3. Find a formula for the general term an , if

(a) a0 = 1 and an = −3an−1 for n # 1.

(b) a1 = 1 and an = an−1 + 3n2 − 3n + 1 for n # 2.

4. In a certain video game there is a spaceship in the middle of a screen, and a number
of inter-galactic space invaders spread across the rest of the screen. The player
tries to eliminate the invaders by shooting them with a laser gun attached to the
spaceship. However, at the end of every 5 second period, each of the remaining
invaders splits into two invaders; so if the player is not very good at shooting, the
screen is eventually filled with invaders.
Let In , for n # 1, denote the number of invaders on the screen at the beginning of
the nth time period, and let A denote the number of invaders on the screen at the
start of the game. (So I1 = A.)
Section 1.1. Sequences 9

(a) Draw a time-line representing the problem and write down a recurrence relation
for In if the player never eliminates any invader.
(b) How long does it take before there are over 1 000 000 invaders on the screen if
A = 3?

Now suppose that the player is able to eliminate invaders, and becomes more skillful
as the game progresses. In fact, in the nth period the player can eliminate as many
as 2n invaders.

(c) Draw a time-line and write down the recursive formula for In in this case.
(d) Find, by trial and error, the smallest value of A which leads to the screen
eventually being filled by invaders (i.e. over 1 000 000).
(e) Repeat parts (c) and (d) if in the nth period the player is able to eliminate
n2 invaders.

5. The recurrence formula

$
tn−1 /2 if tn−1 is even
tn =
3tn−1 + 1 if tn−1 is odd

leads to some interesting sequences of positive integers.

Suppose that t1 = A where A is a positive integer. It seems that no matter what
positive integer we choose for A, the sequence (tn ) eventually settles down to a
cyclical pattern . . . , 1, 4, 2, 1, 4, 2, . . . .

(a) For each A between 1 and 10, evaluate the terms in the sequence until you
reach the cyclical pattern.
(b) Find, by trial and error, the smallest value of A for which some term in the
sequence (tn ) is greater than 100.
(c) Can you find a value for A which gives a sequence (tn ) that does not eventually
collapse to . . . , 1, 4, 2, 1, 4, 2, . . .? There is a large reward for the first person
who finds such an A or proves that there is none! (This is a famous unsolved
problem that has been around since the late 1930s.)
10 Chapter 1. Recurrence Relations

1.2 Solutions to exercises

1. (a) tn = n (n # 1).
1
(b) sn = (n # 1).
n
(c) t1 = 1; tn = 2n−2 (n # 2).
(d) sn = (−1)n (n # 0).
(e) tn = 2n−1 (n # 1).
(f ) tn = n2 (n # 1).

2. (a) t1 = 3; tn = 3tn−1 (n # 2).

1 2n − 1
(b) t1 = ; tn = tn−1 (n # 2).
2 2n
(c) t0 = 0; tn = tn−1 + 2n − 1 (n # 1).
x
(d) t1 = x; tn = tn−1 (n # 2).
n
3. (a) an = (−3)n (n # 0).
(b) an = n3 (n # 1).

4. (a) In = 2In−1 (n # 2).

(b) I19 $ 106 < I20 . So it takes 19 × 5 = 95 seconds.
% ! "
&2 In−1 − 2(n − 1) if 2(n − 1) < In−1 ;
(c) In =
'0 if 2(n − 1) # In−1 .
(d) A = 8.
% ! "
&2 In−1 − (n − 1)2 if (n − 1)2 < In−1 ;
(e) (i) In =
'0 if (n − 1)2 # In−1 .
(ii) A = 12.

5. (a) 1, 4, 2, 1, . . .
2, 1, 4, 2, . . .
3, 10, 5, 16, 8, 4, 2, 1, 4, . . .
4, 2, 1, 4, . . .
5, 16, 8, 4, 2, 1, 4, . . .
6, 3, 10, 5, 16, 8, 4, 2, 1, 4, . . .
7, 22, 11, 34, 17, 52, 26, 13, 40, 20, 10, 5, 16, 8, 4, 2, 1, 4, . . .
8, 4, 2, 1, 4, . . .
9, 28, 14, 7, 22, 11, 34, 17, 52, 26, 13, 40, 20, 10, 5, 16, 8, 4, 2, 1, 4, . . .
10, 5, 16, 8, 4, 2, 1, 4, . . .
(b) A = 15: 15, 46, 23, 70, 35, 106, . . .
Chapter 2

Introduction to MATLAB
Objectives:

! To understand the basic features of the MATLAB package, in particular the repre-
sentation of scalars, vectors and matrices, operations between them, and elementary
MATLAB functions.

! To be proficient in using the two and three dimensional plotting facilities.

! To understand the logical and looping commands in MATLAB, and to be able to

write short MATLAB programs.

Read Chapters 1–5 of the textbook Introduction to MATLAB by Etter (2nd or 3rd
edition), excluding Section 4.1.5 concerning other types of two-dimensional plots, and do
a selection of the problems at the end of Chapters 2–5. (Keep in mind that if you have
the older 2nd edition of the textbook, the MATLAB user interface has changed since it
was published.) Also read Section 7.1 of the textbook, up to and including the command
ezplot. Section 3.6 on user-defined functions can be read after the other sections. Note
that in Section 3.3 titled “Trigonometric Functions”, the only functions that will be used
in this unit are sin(x) and cos(x). When you read about a new command, try it out in
MATLAB by typing in an example from the textbook. The textbook applies MATLAB in
many different applications, but it is not necessary to go through all of these applications.
In parallel to the textbook, also read and work through the following notes.

2.1 Using MATLAB

These Unit Notes refer to version 8 of MATLAB (R2012b and later), but earlier versions
(from 5–7) are suitable and can be used mostly in the same way as described here. Version
8 of MATLAB was released in 2012, introducing a fresh new MATLAB Desktop, which
features a “toolstrip” that replaces menus and toolbars (used in previous versions of
MATLAB) and includes a gallery of pre-built MATLAB “apps”. If you use a version of

11
12 Chapter 2. Introduction to MATLAB

MATLAB prior to the 8th (R2012a or earlier), please keep in mind such differences in
the user interface when reading these notes.
To start MATLAB, select and load the MATLAB application, as you would normally
open any application on your computer (e.g., double click the MATLAB icon). The de-
fault screen when opening MATLAB is made up of three windows: Current Folder (on
the left), Workspace (on the right) and the central Command Window (which is the most
important). The Current Folder window shows the chosen folder containing your MAT-
LAB files (if you’ve created and saved any such files – see below). By default, the current
folder will be the MATLAB subfolder within your Documents folder and this is where
we recommend saving all of your MATLAB files (as explained shortly). The Workspace
window (which used to be the Command History window in earlier versions of MATLAB)
shows the variables/output from the current session. From that window, you can view,
manipulate, save, or clear variables from the current session. The Command Window is
where MATLAB commands or MATLAB programs (called script M-files with the exten-
sion .m) are run and output is displayed. This window has the MATLAB prompt >> (or
EDU>> for the Student Edition) which means it is ready for you to enter a command.
If you want to execute just one or two MATLAB commands, it is simplest to enter
each command at the prompt in the Command Window (one at at time), followed by
the Enter or Return key each time. However, a MATLAB program (which consists of a
sequence of commands over several lines) should be entered and saved as a script M-file.
This allows you to modify the program easily by simply editing the file. All M-files must
have a .m extension.
To create a script M-file, select the New Script button on the MATLAB toolstrip
under the HOME tab (or alternatively, click on New (Script) from the toolstrip under the
EDITOR tab). This brings up an Editor window above the Command Window (if it wasn’t
already open) where you can type the commands (i.e., lines of code) that constitute your
MATLAB program. As shown in examples that follow, the first one or more lines of a
MATLAB program should be a short description of the purpose of the program and there
should also be adequate comments throughout the program. On each line, the description
and comments must be preceded by the % symbol. To save the program as a .m file, select
Save/Save As on the toolstrip under the EDITOR tab and give the program an appropriate
and legal filename, i.e., one that starts with a letter and contains only letters, numbers or
the underscore (_), without any spaces, and is not the same name as a built-in function or
command (e.g., not something like plot). The same rules apply to variable names within
a MATLAB program. The .m extension should be automatically added to your filename
(e.g. program1.m), but if not, you need to add the .m extension yourself before saving.
As mentioned above, we recommend that you save all of your MATLAB programs (script
M-files) in the MATLAB subfolder within your Documents folder and make sure that the
contents of this MATLAB folder is showing in the Current Folder window.
Section 2.2. Scalars, vectors and matrices 13

To execute a program (called program1.m, for example), first make sure that the M-
file is showing up in the list of files in the Current Folder window and then simply
type program1 (without the .m extension) at the prompt in the Command Window and
press Enter or Return on your keyboard. (Alternatively, with your MATLAB program
showing in the Editor window, click the Run button on the toolstrip under the EDITOR
tab.) Executing/running a program has the same effect as if all the commands (in your
program) had been typed in the Command Window. If you want to make any changes to
the program and the Editor window is still open, then simply make the changes, save the
file and again hit the Run button or enter the name of the program (e.g. program1) at the
prompt in the Command Window. (Note that, within the Command Window, you can
repeatedly click the up arrow key ↑ on your keyboard to scroll back through previously
typed commands or filenames.) To open an existing M-file, click the Open button on the
toolstrip and select the file. Alternatively, double click the M-file from the list of files in
the Current Folder window. For some more information about M-files, see the section
entitled “Script Files” in the MATLAB User Guide.
The online help facility in MATLAB is very useful. The command help provides a
help menu for all topics. However, if you know the name of the topic or command that you
want to check, it is quicker to type help commandname (e.g. help plot) at the prompt in
the Command Window. You can also obtain help on the help command itself by entering
help help at the prompt. The help facility can also be accessed via MATLAB’s Help
menu or the Help button on the toolstrip under the HOME tab.
To exit MATLAB, choose Quit MATLAB from the MATLAB drop-down menu. Alterna-
tively, type quit or exit at the prompt (in the Command Window) and then press Enter
or Return on your keyboard.

2.2 Scalars, vectors and matrices

MATLAB is a powerful technical computing package and environment incorporating
mathematical computation, graphic visualisation (plots) and a high level programming
language. It is especially suited to computations involving matrices. In fact, the name
MATLAB stands for Matrix Laboratory. It started as a public domain package in the
1970s and was then developed by the MathWorks company from the 1980s. In addition
to the core MATLAB package, there are several MATLAB toolboxes, i.e. libraries for
special purposes, for example, optimisation.
The basic data structure in MATLAB is a matrix, i.e. a rectangular array of numbers.
Scalars (i.e. numbers) and vectors are just special cases of matrices.
14 Chapter 2. Introduction to MATLAB

◮ Example
( 2.1+
1 2
) ,
C = * 3 4 - is a 3 × 2 matrix (3 rows, 2 columns).
5 6
. /
B = 7 8 9 9.5 is a 1 × 4 matrix (1 row, 4 columns), called a row vector.
A = [2.3] is a 1 × 1 matrix, which is equivalent to the scalar 2.3.
The elements of a matrix are identified by their row and column numbers in that
order. For example, in the first matrix C3,2 = 6.
The matrices above can be entered in MATLAB in explicit list form as follows.
(Note the % symbol is used before a comment.)

C=[1,2; 3,4; 5,6] % row by row separated by semicolons, or

C=[1 2; 3 4; 5 6] % the commas are replaced by spaces

B=[7,8,9,9.5] % or
B=[7 8 9 9.5] % the commas are replaced by spaces

A=2.3

Having entered a vector or matrix, an element can be found using round brack-
ets, for example

B(2) % gives element 8

C(3,2) % gives element 6

Two (or more) scalars, vectors or matrices (of appropriate size) can be concate-
nated easily by simply enclosing them in square brackets, for example

S=[6.5 B] % gives vector [6.5,7,8,9,9.5]

An element of a vector or matrix can be redefined easily, as in

S(2)=7.4 % gives new vector [6.5,7.4,8,9,9.5]

S(8)=2 % gives new vector [6.5,7.4,8,9,9.5,0,0,2]

Note the second command adds three elements to the vector, the 8th being the
one specified (2), and the 6th and 7th assigned the default value zero (since they
were not specified). ◭
Section 2.2. Scalars, vectors and matrices 15

Exercise With A, B and C entered as above, write down the output of each of the
MATLAB statements:

P=[0.7;0.8;0.9]
[P(1) B]
[C,P]
[A;C(2,2)]

Colon operator
The colon operator is a very important operator in MATLAB. It is used in several ways
– to define vectors with equal increments, to specify elements of a vector or matrix, and
in the for loop statement. These uses are illustrated in the following example.

◮ Example 2.2

y=5:8 % defines y to be [5,6,7,8] (default increment is 1)

time=0:0.5:2 % defines time=[0,0.5,1,1.5,2]
values=10:-1:6 % defines values=[10,9,8,7,6]

B=[7 8 9 9.5]
x=B(1:3) % defines x=[7 8 9]
C=[1 2; 3 4; 5 6]
C(2,:) % gives second row 3 4 (here : means any column position)
E=[1 2 3; 4 5 6; 7 8 9]
F=E(1:2,2:3) % defines F=top right 2 by 2 matrix

Here ( +
1 2 3 0 1
) , 2 3
E=* 4 5 6 - and F =
5 6
7 8 9
The following program uses a for loop to compute and list all the even numbers
from 2 to 20.

for n=1:10
x=2*n
end

The for loop will be discussed in more detail in Section 2.5. ◭

16 Chapter 2. Introduction to MATLAB

Transpose operator
The transpose D of an m × n matrix C is the n × m matrix in which the elements of the
ith row of D are the corresponding elements of the ith column of C. For example,
( +
1 2 0 1
) , 1 3 5
if C = * 3 4 - its transpose is D =
2 4 6
5 6

Note also that the columns of D are the corresponding rows of C. Clearly the transpose
of a row vector is just the corresponding column vector, and the transpose of a column
vector is the corresponding row vector. In MATLAB the transpose operator is denoted
by the prime symbol, as in

C=[1 2; 3 4; 5 6]
D=C’ % D is the transpose of C

◮ Example 2.3 The following program creates a table.

x=0:4; % semicolon suppresses output from statement

y=[5,10,12,10,5];
table=[x’,y’] % with a comma or space

It gives

table =
0 5
1 10
2 12
3 10
4 5

Input
Sometimes it is useful to allow the user of a program to enter the value of some variable
(possibly a scalar, vector or matrix) using the keyboard. This can be done with the input
command as follows:

n=input(’Enter value of n’)

x=input(’Enter elements of vector x in square brackets’)

The first statement displays the message

Enter value of n

in the Command Window and waits for a response. If you type say 25 and press
Enter/Return, then n has the value 25.
Section 2.2. Scalars, vectors and matrices 17

Similarly, the second statement displays the message

Enter elements of vector x in square brackets

in the Command Window and waits for a response. If you type [2,3,5.5] and press
Enter/Return, then x is defined to be the vector [2,3,5.5].
Output
The format of the numerical output in MATLAB can be specified using one of the following
format statements. In each case the output is rounded to the given number of decimal
places.

format short % default, 4 dec places and <=3 whole digits e.g. 15.2345
format long % 14 decimal places e.g. 15.23453170892113
format bank % 2 decimal places e.g. 15.23
format short e % 4 decimal places e.g. 1.5235e+001 (e+001 means x10^1)
format long e % 15 decimal places e.g. 1.523453170892113e+001
format compact % suppress blank lines
format loose % default

The value of a variable (scalar, vector or matrix) can be displayed using the disp
command. This command can also be used to display text strings as follows.

temp=35;
disp(’Temperature is’)
disp(temp)
disp(’degrees C’)

This gives

Temperature is
35
degrees C

If more control of the output format is desired, one can use the command fprintf
(but this is not essential in this unit). It is used as follows

temp=28.53;
fprintf(’Temperature is %4.1f degrees C’,temp)

which gives output

Temperature is 28.5 degrees C

18 Chapter 2. Introduction to MATLAB

where we have 4 places for the numerical output (including decimal point) with 1 decimal
place. For more information on fprintf see the MATLAB help facility.

Scalar and array operations

The basic operations of +, −, × and / between numbers (scalars) are entered in MATLAB
as +, -, * and /, respectively. The operation of taking a power (or index) is entered as ^.
The operations in MATLAB have a precedence order that is the same as is used when
writing expressions by hand: brackets, indices, multiplication and division left to right,
addition and subtraction left to right. For example, the MATLAB expression
(3+2)*(4^2-6/2*4)
is computed as 5 × (16 − 3 × 4) = 5 × 4 = 20. When entering an expression in MATLAB
it is essential to include round brackets where necessary. (One should also avoid putting
in too many unnecessary brackets since this can make the MATLAB expression difficult
to read.) For example, the expression
2 × 67 + 1
3×5.2
+ 4 × 23
4π
is entered in MATLAB as:
(2*6/7 + 1/(3*5.2) + 4*2^3)/(4*pi)
Note that the number π is entered as pi.
In MATLAB, computations with arrays can be done very efficiently, both in the ease
of coding and in the speed of the computation. Suppose that a and b are two variables
representing arrays of the same size, say a=[1, 2] and b=[3, 4]. Each of the basic
operations has a corresponding array operation that is carried out element-by-element.
The syntax for these array operations and their results are shown below:

a + b [1+3, 2+4] = [4, 6]

a - b [1-3, 2-4] = [-2, -2]
a .* b [1*3, 2*4] = [3, 8]
a ./ b [1/3, 2/4] = [1/3, 1/2]
a .^ b [1^3, 2^4] = [1, 16]
a .^ k [1^k, 2^k] % for any scalar k, e.g. k=2
a + k [1+k, 2+k] % for any scalar k, e.g. k=2
a - k [1-k, 2-k] % for any scalar k, e.g. k=2
k*a [k*1, k*2] % for any scalar k, e.g. k=2
a/k [1/k, 2/k] % for any non-zero scalar k, e.g. k=2

Note that some of the array operations are denoted by two symbols with . at the front
(which is essential). The reason for this is that the single symbols *, / and ^ have been
reserved for other operations between arrays or matrices; in particular, A*B means matrix
multiplication of matrices A and B .
Section 2.2. Scalars, vectors and matrices 19

A useful application of array operations is to generate tables of conversions, as shown

in the following example.

◮ Example 2.4 Generate a table of conversions of speeds from km/hr to miles/hr.

Take the range from 0 to 150 km/hr in increments of 10 km/hr. Recall that 1 mile
equals 1.61 km.

% Generate a table of conversions of speeds from km/hr to miles/hr

clear % clear all variables
speedkm=0:10:150; % define vector of speeds in km/hr
speedmi=speedkm/1.61; % convert to miles/hr using array operation
disp(’speed km/hr miles/hr’) % display heading
disp([speedkm’ speedmi’]) % display table (transpose is used to
% create columns)

Data analysis functions

The following MATLAB functions are useful for analysing data. Suppose x and y are
vectors of the same size and A is a matrix.

[m,n]=size(A) % defines m = no. of rows and n = no. of columns of A

length(x) % number of elements in x
max(x) % largest element of x
[xmax,k]=max(x) % defines xmax = largest element of x and
% k = first index where largest element occurs
% for example
x=[5,4,8,2,8,6];
[xmax,k]=max(x) % defines xmax = 8 and k = 3
max(x,y) % vector of the largest of corresponding elements of x and y
min(x) % smallest element of x
[xmin,k]=min(x) % defines xmin = smallest element of x and
% k = first index where smallest element occurs
min(x,y) % vector of the smallest of corresponding elements of x and y
sum(x) % sum of elements of x
prod(x) % product of elements of x
mean(x) % mean of elements of x
std(x) % standard deviation of elements of x
sort(x) % vector with elements of x in ascending order
20 Chapter 2. Introduction to MATLAB

2.3 Plotting in two dimensions

The array operations in MATLAB provide an efficient means of computing a vector of
function values and plotting the graph of the function, as shown below.

◮ Example 2.5 Find the vector y of function values of y(x) = x2 + 3x − 4 at values

of x between -5 and 2 in increments of 0.5, and plot the corresponding points.
Note the x values are the components of the vector
x=-5:0.5:2
Clearly, for each component xi of x, the corresponding component of y is x2i +3xi −4.
Therefore we define the vector y as
y=x.^2 + 3*x - 4
The first term x.^2 is the vector with each component equal to the square of the
corresponding component of x. The second term is the vector whose components
are simply 3 times those of the corresponding components of x. Although the
third term 4 is just a number, MATLAB interprets the expression correctly by
simply subtracting 4 element-by-element. To plot the points using circles, with
appropriate labels, the commands are:

plot(x,y,’o’)
xlabel(’x’)
ylabel(’y’)
title(’Function y(x)=x^2+3x-4’)

Execute all the above commands. ◭

The plot command can produce plots of both point type (using one of the plot symbols
’o’, ’+’ or ’*’) or line type (using the default of solid or one of ’--’ (dashed) or ’:’
(dotted)). The line type plots are produced simply by joining the points defined by the
vectors x and y in plot(x,y). Therefore, to create an accurate line graph of a function,
generally one needs to use a small increment for the x values, say 0.01.
The axis command is used to control the limits and scaling of a plot. It is included
after the plot command in the following possible ways.

axis([xmin,xmax,ymin,ymax]) % defines limits on x and y axes

axis equal % sets equal scaling on x and y axes
axis auto % default, return to automatic scaling

A grid can be added to a plot using grid on (after the plot command) and it can be
removed using grid off.
Section 2.3. Plotting in two dimensions 21

◮ Example 2.6 Write a MATLAB program to plot the graphs of the functions
f (x) = (x3 − 3x2 )(2x + 5) and g(x) = (x + 3)2
together using values of x between −3 and 3 in increments of 0.01. Set the y-axis
limits to be −50 and 50, and include a grid.
% Plot graphs of two functions
clear
x=-3:0.01:3;
f=(x.^3 - 3*x.^2).*(2*x + 5); % note element-by-element multiplication
% between the brackets
g=(x+3).^2;
plot(x,f,x,g,’--’)
xlabel(’x’)
ylabel(’y’)
title(’Graphs of f(x)=(x^3-3x^2)(2x+5) (solid) and g(x)=(x+3)^2 (dashed)’)
axis([-3,3,-50,50])
grid on
Execute this program. ◭
Another way to plot more than one graph in the same figure is to use the hold on
command, which holds the figure ready for subsequent plots. It is used as follows:
x=-3:0.01:3;
f=(x.^3 - 3*x.^2).*(2*x + 5);
plot(x,f)
hold on
g=(x+3).^2;
plot(x,g,’--’)
hold off
Exercise Write a MATLAB program to compute and plot the functions
2 − 3x
f (x) = 2 and g(x) = (x/5)4
4x + 5x
together for 1 $ x $ 3.
MATLAB has built-in functions to evaluate the elementary functions of mathemat-
√
ics; in particular sqrt(x) for x, sin(x) for sin(x), cos(x) for cos(x), exp(x) for the
exponential function ex , where e is the special number e = 2.71828 · · · , and log(x) for
the natural log function ln(x) (defined to be the number y that satisfies the equation
ey = x). Note that if x is a vector, then the built-in MATLAB function produces a vector
of function values. If the function is part of a larger expression, be careful to use the
correct array operations.
22 Chapter 2. Introduction to MATLAB

◮ Example 2.7 To generate the values of the function

f (x) = (sin(x))2 + 3x/(1 + ex )

for x between 0 and 3 in increments of 0.1, the appropriate statements are:

x=0:0.1:3;
f=(sin(x)).^2 + 3*x./(1+exp(x));

Exercise Write down the appropriate MATLAB statements to generate the values of
the function
g(x) = (3 + x cos(x))e2x
for x between 0 and 1 in increments of 0.1. Execute the statements in MATLAB.

◮ Example 2.8 The following MATLAB program generates the values of f (x) =
1 + sin(2πx) for x between 0 and 1 in increments of 0.01, plots the function and
sums its values.

% Plot and sum values of f(x)

clear
x=0:0.01:1;
f=1+sin(2*pi*x);
plot(x,f)
xlabel(’x’)
ylabel(’f’)
title(’Graph of f(x)=1+sin(2*pi*x)’)
pause % execution continues when any key is pressed
sumf=sum(f);
disp(’sum of f values = ’)
disp(sumf)

Execute this program in MATLAB. Can you explain why the sum of the values of
f equals 101, the number of elements in the vector x? ◭

Sometimes it is necessary to plot the graphs of two functions on different domains,

together in the same figure. This can be done easily with the plot command by using
different variables for the two vectors of x-coordinates, as shown in the next example.
Section 2.3. Plotting in two dimensions 23

√
◮ Example 2.9 Plot the graphs of f (x) = x/(x + 1) for 0 $ x $ 9 and g(x) =
x/2 + 1 for −5 $ x $ 1 together in the same figure. Note that we can’t plot both
√
functions over −5 $ x $ 9 because x is undefined for x < 0.

% Plot two functions on different domains

clear
x=0:0.01:9;
f=sqrt(x)./(x+1);
t=-5:1; % different variable
g=t/2 + 1;
plot(x,f,t,g)
xlabel(’x’)
ylabel(’y’)
title(’Graphs of f(x)=sqrt(x)/(x+1) and g(x)=x/2+1’)

Note that we can use the large increment 1 to plot g(x) because the graph is a
straight line (and MATLAB just joins the defined points by straight line segments).
Execute this program in MATLAB. ◭

The ezplot command is an easy way to plot an algebraically defined function. Note
that the function is entered as an algebraic expression, without using any array operations.
The interval domain of the independent variable is entered in square brackets. If the
interval is not present, the default domain is [−2π, 2π]. For example

ezplot(’exp(-x^2)’)
2
plots e−x on [−2π, 2π] while

ezplot(’exp(-x^2)’,[-1,1])
2
plots e−x on [−1, 1].
A useful plotting command is zoom. As its name suggests, it allows you to zoom in on
any part of a 2-D graph and see it enlarged in the plot window. The statement zoom on
turns on the zoom mode. Then, to zoom in about a particular point on the screen, move
the cursor to the point and click on the mouse button (left one for a PC). This expands
the plot by a factor of 2 centred on the point. The process can be repeated over and over
again to zoom in closer. You can also click and drag the mouse to zoom in on a particular
rectangular area. To zoom out by a factor of 2, click the right mouse button on a PC or
shift and click on an Apple Mac. The statement zoom out returns the plot to its initial
state and zoom off turns off the zoom mode. In MATLAB 5.3 and later versions, the
zoom (in or out) mode can also be activated using the menu bar of the figure window.
24 Chapter 2. Introduction to MATLAB

◮ Example 2.10 The following MATLAB program allows you to zoom in on the
graph of f (x) = x2 .

% Zoom in on a graph
x=-2:0.01:2;
f=x.^2;
plot(x,f)
xlabel(’x’)
ylabel(’f’)
title(’Graph of f(x)=x^2’)
zoom on

Execute the program and zoom in (and out) on a few different points of the graph.
Notice that as you zoom in, the section of graph looks more and more like a line
segment. In fact this is true for any point of the graph. This means that near
any point (x1 , y1 ) of the graph, the graph can be approximated well by a line
segment. If you have studied some calculus, you will know that this property holds
because the function f (x) = x2 is differentiable, that the line segment is part of the
tangent line at the point, and the slope of the line is the derivative f ′ (x1 ) = 2x1 .
The equation of the line is y = 2x1 (x − x1 ) + y1 . Choose a point on the graph (say
(x1 , y1 ) = (1, 1)) and plot this line together with the graph of f (x) = x2 . ◭

◮ Example 2.11 Use the following program to zoom in on the graphs of the two
functions f (x) = sin(x) and g(x) = x near the origin.

% Zoom in on two graphs

x=-2:0.01:2;
f=sin(x);
g=x;
plot(x,f,x,g,’--’)
xlabel(’x’)
ylabel(’f and g’)
title(’Graphs of f(x)=sin(x) and g(x)=x’)
zoom on

Notice that as you zoom in, the graphs become closer and closer. This means that
for x near 0, sin(x) is close to x, and the nearer x is to 0, the closer sin(x) is to
x. This is an important fact about the sin function. What function do you think
h(x) = 3 sin(x2 ) is close to near x = 0? (Answer: 3x2 ) Modify the program to
verify this graphically. ◭
Section 2.3. Plotting in two dimensions 25

Another useful command to interact with a 2-D graph window is ginput. This com-
mand records the coordinates of any desired position of the cursor, which can then be
printed or operated on in any way. In its simplest form, the command is used as:

[xin yin]=ginput

where the variable names (xin and yin) can be chosen as desired. Then, with each click
of the mouse button, the x and y coordinates of the cursor (cross) are saved in the column
vectors xin and yin. To stop this process, press the Enter or Return key. For more
information on the ginput command, look it up using the help facility.
A common mathematical problem is to solve two simultaneous equations in two vari-
ables, say x and y. A simple example is: solve y = 2x and y = 1 − x. This simple example
can be solved algebraically by writing

y = 2x = 1 − x

so 3x = 1 and therefore x = 1/3. Substituting this into the first (or second) equation
gives y = 2 × 1/3 = 2/3. Thus the solution is x = 1/3, y = 2/3. This solution can also be
estimated graphically by plotting the lines y = 2x and y = 1 − x, and finding (accurately)
their point of intersection.
For more complicated equations it is usually impossible to find the solution(s) al-
gebraically, but one can still approximate the solution(s) graphically. The MATLAB
commands plot, zoom and ginput are very useful for this purpose.

◮ Example 2.12 Find the positive solution of the two equations y = cos(x) and
y = x2 . This is equivalent to finding the intersection point of the graphs of the
functions cos(x) and x2 in the first quadrant (i.e., in the positive quadrant of the
X-Y plane). The following program plots the graphs and allows the user to zoom
in on the intersection point to approximate the solution.

% Solve two equations using plot, zoom and ginput

clear
x=0:0.01:1;
y1=cos(x);
y2=x.^2;
plot(x,y1,x,y2,’--’)
zoom on
pause % press Enter/Return in Command Window to continue
zoom off
[xint yint]=ginput % (xint, yint) is the intersection point
26 Chapter 2. Introduction to MATLAB

Execute the program and zoom in several times by clicking on the intersection
point in the plot window. Press Enter/Return to stop the zoom and ginput
procedures. Then the estimated coordinates of the intersection point, i.e. the
approximate solution values, will be displayed in the Command Window. If a
certain accuracy is required, it is important to use a sufficiently small increment
for the plots and to zoom in to the intersection point sufficiently many times. ◭

2.4 Plotting in three dimensions

MATLAB has a number of specialized commands for plotting in three dimensions. Run
the following MATLAB program to plot the surface defined by the function
1
f (x, y) =
1 + x2 + y 2

% Plot a surface in three dimensions

clear
x = -2:0.1:2;
y = -1:0.1:2;
[xgrid, ygrid] = meshgrid(x,y); % defines grid matrices
z = 1./(1 + xgrid.^2 + ygrid.^2);
mesh(xgrid,ygrid,z) % generates a mesh plot

Replace the last line by

surf(xgrid,ygrid,z) % generates a shaded mesh plot

and note the difference in the plot. Also on the last line use each of the commands
contour(x,y,z) and meshc(xgrid,ygrid,z) to plot contours of this function.
Exercise Use the MATLAB command meshc to plot the surface defined by

f (x, y) = (x3 − 3x)/(1 + y 2 ), −3 $ x $ 3, −3 $ y $ 3.

Also use the command contour to draw the contours of f . Note that you can specify the
number n of contour lines by adding this as an option in the form contour(x,y,f,n). If
you want to draw the contours corresponding to specific values of the function f , this can
be done using contour(x,y,f,vec), where vec is a vector containing the desired values.
Experiment with each of these options in the contour plot of the given function f . Use
ginput with the contour plot to estimate the positions of the local maximum and local
minimum of f . What do you think the exact positions are?
Section 2.5. For loop 27

2.5 For loop

In computing it is often necessary to repeat the execution of a set of statements, possibly
to update some variable(s). This can be achieved with a for loop, which in MATLAB in
its simplest form is
for n=p:q
... % some statement(s)
end
where n is the loop variable or index, which goes from p to q with increment 1. Note that
n is not a vector here; it is a scalar variable that takes on each of the integer values from
p to q, inclusive, in turn.
More generally, the for loop in MATLAB has the form
for x=vec
... % some statement(s)
end
where the loop variable x takes the value of each of the elements of the vector vec in
order from the first to the last.
An important use of a for loop is to compute a sequence of numbers that are defined
recursively.

◮ Example 2.13 Consider the sequence defined by

t1 = 3; tn = tn−1 + 2n for n # 2.

The following MATLAB program implements this recursive definition and com-
putes 10 terms of the sequence.
% Compute a sequence
clear
t=3 % initialize variable as first term
for n=2:10
t=t+2*n % update variable: new value = old value + 2*n
end
The statement t=t+2*n in this program may look odd at first sight (because it
is incorrect as an equation). However, such statements are common in computer
programming and have a logical meaning. For this statement, the result of the
computer addition of the values of t and 2*n is assigned to and stored in the
variable t, i.e. the old value of this variable is replaced by the new value. In this
way, at the end of each pass of the loop, the variable t has the value of the new
term of the sequence. Run this program and check the first few terms by hand. ◭
28 Chapter 2. Introduction to MATLAB

◮ Example 2.14 Consider Example 1.12 about the amounts An , n # 1, in a cer-

tain bank account, where An is defined recursively as: A1 = 1000 and An =
(1.05)An−1 + 100, n # 2. Write a MATLAB program to compute An for n =
1, . . . , 10, and also the total deposits Dn that Joan pays into the account up to the
beginning of year n for n = 1, . . . , 10. Display the values in three columns:
n An Dn

% Compute the amounts in a bank account and the total deposits

clear
format bank % two decimal places
A=1000; % initial amount in account
D=0; % initial deposits
disp(’ year n amount A_n tot deposits D_n’)
disp([1 A D])
for n=2:10
A=1.05*A+100;
D=D+100;
disp([n A D])
end

Execute this program. Extend the program to compute the total interest In paid
by the bank up to the beginning of year n for n = 1, . . . , 10. ◭

◮ Example 2.15 Write a MATLAB program to generate 30 terms of the Fibonacci

sequence defined by f1 = 1, f2 = 1 and fi = fi−1 + fi−2 for i # 3, and plot fi
against i. Also compute the ratio fi /fi−1 of successive terms and plot it against
i − 1. Since in this example we want to plot the the Fibonacci sequence, we will
define the terms of the sequence as elements of a vector f.

% Generate Fibonacci sequence and ratios of successive terms

clear
format long
n=input(’Enter number of terms: ’);
f=zeros(1,n); % initialize size of vector
ratio=zeros(1,n-1); % initialize size of vector
f(1)=1; f(2)=1;
ratio(1)=1;
for i=3:n
f(i)=f(i-1)+f(i-2);
ratio(i-1)=f(i)/f(i-1);
Section 2.5. For loop 29

end
disp(’Fibonacci sequence: ’)
disp(f)
plot(f,’o’)
xlabel(’i’)
ylabel(’f’)
title(’Fibonacci number f(i)’)
pause
plot(ratio,’o’)
xlabel(’i-1’)
ylabel(’ratio’)
title(’Ratios f(i)/f(i-1) of Fibonacci numbers’)

Execute this program and observe the behaviour of the ratios. It is known that
√
the ratios approach the golden ratio (1 + 5)/2 = 1.618 · · · as i approaches ∞.
Note that the purpose of the statements initializing f and ratio is to allocate the
correct final size to these vectors. Without these statements, MATLAB must resize
the vectors at each pass of the for loop, which is not as efficient. ◭

A common task arising in many applications is to find the sum of a set of numbers. If
the numbers can be generated easily as a vector v, the most efficient way of doing this in
MATLAB is to use the command sum(v) (as in Example 2.8). Sometimes it is difficult
or inconvenient to generate a vector of values and then a for loop can be used.

◮ Example 2.16 The following MATLAB program computes the sum of the series
1 1 1 1
S = 1+ + + + ··· +
1 1×2 1×2×3 1 × 2 × · · · × 20
% Sum a finite series using for loop
clear
format long
N=20;
s=1;
term=1; % first term
for n=1:N
term=term/n; % new term is defined recursively
s=s+term;
end
disp(’sum of series is’)
disp(s)
30 Chapter 2. Introduction to MATLAB

Note that this program implements two recursive definitions - one for the terms
of the series (computed in the variable term) and one to accumulate the terms in a
partial sum (computed in the variable s). Then at the completion of the for loop,
the variable s has the value of the whole sum.
Run this program to see the result. If you have studied the exponential function,
you may recognize the above series. The sum of the corresponding infinite series
is the number e = 2.71828182845905 · · · . ◭

In MATLAB long for loops should be avoided where possible. In particular, a for
loop should not be used to generate simple function values. It is much more efficient to
use the basic array operations in MATLAB as demonstrated in the following example.

◮ Example 2.17 The following two programs both compute the logarithms of the
integers from 1 to 800000 and time the total computation.

% Compute logarithms using array operation

clear
t0=clock; % set starting time
indv=1:800000; % indv is an index vector
y=log(indv);
elapse=etime(clock,t0); % elapsed computer time
disp(’elapsed time in seconds is’)
disp(elapse)

% Compute logarithms using for loop

clear
t0=clock; % set starting time
y=zeros(1,800000); % initialize size of y
for ind=1:800000 % ind is an index number
y(ind)=log(ind);
end
elapse=etime(clock,t0); % elapsed computer time
disp(’elapsed time in seconds is’)
disp(elapse)

Run the two programs and compare the elapsed times. Which one is more efficient
and by what multiple? Also run the second program without the statement that
initializes the size of y and compare the elapsed time. ◭
Section 2.5. For loop 31

Exercise Write two MATLAB programs to compute the sum of the finite geometric
series
S = 1 + r + r2 + · · · + rn
where the common ratio is r = 0.9 and n = 2000, one using array operations and the
other using a for loop. Use the etime command to time the two programs. Which one
is more efficient and by how much?

Consider the problem of finding a root z of an equation f (x) = 0, i.e. a value z such
that f (z) = 0. Graphically, z is the x-coordinate of a point at which the graph of f (x)
crosses the x-axis.
A well-known simple iterative method for root finding is the Newton-Raphson method.
Given the equation f (x) = 0 and an initial estimate x0 of the root z, this method generates
a sequence of estimates recursively by the formula

f (xn )
xn+1 = xn − , n # 0,
f ′ (xn )

where f ′ is the derivative of the function f . From the following diagram it is easy to see
how the method works. It simply uses the tangent line determined by the current point
xn to generate the next point xn+1 as the point where the line crosses the x-axis. Since
the slope of the tangent line at (xn , f (xn )) is f ′ (xn ), and so the equation of the line is

y = f (xn ) + f ′ (xn )(x − xn ),

the next point is defined by setting y = 0 and solving for x. This yields (check this) the
expression for xn+1 above.

z
0
y

xn xn+1

−2 y=f(x)

y=f(xn)+f′(xn) (x−xn)
−4

−6
−1 −0.5 0 0.5 1 1.5 2
x

Figure 2.1: Graphical interpretation of the Newton-Raphson method

32 Chapter 2. Introduction to MATLAB

It can be shown that if the starting estimate x0 is sufficiently close to the root z and
the slope of y = f (x) at z is not zero, then the iterates xn , n = 1, 2, . . . , of the Newton-
Raphson method converge quickly to z. In fact, after each step the number of correct
decimal places is approximately doubled. However, if the starting estimate is not close
enough to a root, the iterates may not even converge.

◮ Example 2.18 Write a MATLAB program using the Newton-Raphson method to

√ √
estimate 2. Clearly 2 is the positive solution of the equation f (x) = x2 − 2 =
0. The derivative of f is f ′ (x) = 2x, so the Newton-Raphson method gives the
recurrence formula
x2 − 2
xn+1 = xn − n , n # 0.
2xn
The following program implements this method.

% Find a root using Newton-Raphson method

clear
format long
nit=input(’Enter number of iterations: ’);
x0=input(’Enter starting estimate: ’);
x=x0;
for n=1:nit
x=x-(x^2-2)/(2*x);
end
disp(’Final estimate of root is: ’)
disp(x)
disp(’Square root of 2 is: ’)
disp(sqrt(2))

Run this program with 10 iterations and starting estimates of 1.4, 1, 0.1, 0.001
√
and 0.00001. In the last case the final estimate is a long way from 2. Explain
this using a graph of f (x) = x2 − 2 and the above geometric interpretation of the
Newton-Raphson method. ◭
Section 2.6. Selection statements 33

2.6 Selection statements

The basic selection statement is the if statement, which in MATLAB has the form, e.g.

if y > 0.5
... % some statement(s)
end

which means that the statement(s) are executed if and only if the value of y is > 0.5.

◮ Example 2.19 Work through the following MATLAB program. What do you
think the final value of count represents?

clear
count=0;
for x=-1:0.1:1
y=x^2;
if y > 0.5
count=count+1;
disp([count y])
end
end

Execute the program and check that it gives what you expect. ◭

Instead of > in y > 0.5 above, we could have any one of:
relational operator meaning
< less than
<= less than or equal
> greater than
>= greater than or equal
== equal
~= not equal
A statement of the form y > 0.5 is called a logical expression. Depending on the value
of the variable y, the expression y > 0.5 has value 0 (false) or 1 (true). For example the
statements

y=0.8;
y > 0.5

give the value 1.

34 Chapter 2. Introduction to MATLAB

Similarly, logical expressions involving arrays are treated element-by-element. For exam-
ple the statements

a=[2 4 6];
b=[3 5 1];
a < b

give the result: 1 1 0

Logical expressions can be combined in MATLAB using:
logical operator symbol
not ~
and &
or |
For example, the if statement in the program above could be replaced by
if y > 0.5 & y < 0.8
or by
if y < 0.2 | y > 0.5

Nested if
If statements can be nested as in the following.

◮ Example 2.20

clear
count=0;
for x=-1:0.1:1
y=x^2;
if y > 0.5
count=count+1;
disp([count y])
if x > 0.5
disp(’Both x and y are > 0.5’)
end
end
end

Execute the program and check that it gives what you expect. ◭
Section 2.6. Selection statements 35

else and elseif

When there are two cases to consider, one uses else, as demonstrated in the following
example.

◮ Example 2.21

% Define the absolute value of x

clear
x=input(’Enter value of x: ’);
if x >= 0
y=x
else
y=-x
end

Execute the program several times with different values of x and check that it gives
what you expect. ◭

When there are more than two cases to consider in levels, one uses elseif as follows.

◮ Example 2.22

% Classify Goldilocks’ porridge

clear
temp=input(’Enter temperature: ’);
if temp > 60
disp(’porridge is too hot’)
elseif temp > 30
disp(’porridge is just right’)
elseif temp > 0
disp(’porridge is too cold’)
else
disp(’porridge is frozen solid’)
end

Execute the program several times with different values of temp and check that it
gives what you expect. ◭

Logical function find

The function find is very useful to find the elements of a vector that satisfy some
condition. Note that find returns the index set of these elements.
36 Chapter 2. Introduction to MATLAB

◮ Example 2.23 Write a MATLAB program using the find command to find the
values of y = x2 for x = −1, −0.9, . . . , 1 that are greater than 0.5, and the number
of these values. Then redefine these values to equal 0.5 and plot the resulting
points.

% Find elements of a vector that satisfy a condition

clear
x=-1:0.1:1;
y=x.^2;
indset=find(y > 0.5) % index set of elements that are > 0.5
number=length(indset) % number of elements that are > 0.5
disp(y(indset)) % elements of y that are > 0.5
y(indset)=0.5; % redefine elements
plot(x,y,’o’)

Note that the length command returns the length of a vector. Run this program
and note that the value of number is the same as the final value of count in
Example 2.19. ◭

◮ Example 2.24 The recurrence formula

$
tn−1 /2 if tn−1 is even
tn =
3tn−1 + 1 if tn−1 is odd

has the interesting property that the terms appear to settle down to the simple
cyclical pattern 1,4,2,1,4,2,. . . no matter what the starting positive integer a = t1 .
(See the corresponding exercise at the end of Section 1.1.) It is unknown if this
is true in general and is a famous unsolved problem. The following MATLAB
program generates the terms for any starting integer. Read through the program
and make sure you understand the if-else statements, and the remainder function
rem. Note we have used the command break to jump out of the for loop when
the terms start cycling.

% Generate terms of interesting recurrence formula

format compact % display output in compact form
clear
nmax=1000; % maximum number of iterations
a=input(’Enter starting integer: ’);
t(1)=a;
for n=2:nmax
if rem(t(n-1),2)==0 % if t(n-1) is even
Section 2.6. Selection statements 37

t(n)=t(n-1)/2;
else
t(n)=3*t(n-1) + 1;
end
if t(n)==1 % if terms start cycling
break % jump out of for loop
end
end
disp(’Terms of recurrence formula: ’)
disp(t)

Execute the program a few times with different starting integers. Extend the
program to also display the number of iterations of the formula used before cycling
starts. (Hint: Try using the statement nt=length(t).) ◭

Exercise Write MATLAB programs to solve parts (d) and (e) of Exercise 4 from Section
1.1 (about the space invaders game).

◮ Example 2.25 A certain parcel delivery company has size limits on the parcels it
will accept. The parcel must be no more than 1 metre in length and no more than
0.3 cubic metres in volume. In a particular load, the parcels have the following
sizes:
length volume
0.8 0.15
1.1 0.35
0.75 0.25
1.2 0.2
0.9 0.4
1.15 0.2
0.6 0.15
The size data (numbers) are stored in this format in a file called parcels.dat. Write
a MATLAB program to classify the parcels and for each parcel, print out its length,
volume, and whether its length and volume are acceptable or too big. Also find
and print out: the number of parcels, the number of parcels with length acceptable,
the number of parcels with volume acceptable, the number of parcels with both
length and volume acceptable.
38 Chapter 2. Introduction to MATLAB

% Classify parcels using size rules

clear
load parcels.dat
len=parcels(:,1); % lengths of parcels
vol=parcels(:,2); % volumes of parcels
ntot=length(vol); % total number of parcels
lenok=0; volok=0; sizeok=0; % initialize counters
for n=1:ntot
if len(n)<=1
lenok=lenok+1;
if vol(n)<=0.3
volok=volok+1;
sizeok=sizeok+1;
fprintf(’%4.2f %4.2f length acceptable, volume acceptable \n’,...
len(n), vol(n))
else
fprintf(’%4.2f %4.2f length acceptable, volume too big \n’,...
len(n), vol(n))
end
else
if vol(n)<=0.3
volok=volok+1;
fprintf(’%4.2f %4.2f length too big , volume acceptable \n’,...
len(n), vol(n))
else
fprintf(’%4.2f %4.2f length too big , volume too big \n’,...
len(n), vol(n))
end
end
end
disp(’ ’), disp(’Number of parcels: ’), disp(ntot)
disp(’Number of parcels with length acceptable: ’), disp(lenok)
disp(’Number of parcels with volume acceptable: ’), disp(volok)
disp(’Number of parcels with both length and volume acceptable: ’)
disp(sizeok)

Run this program and check that it gives what you expect. Extend the program
using the find command to find the same numbers as above, i.e. the number of
parcels with length acceptable, the number of parcels with volume acceptable and
the number of parcels with both length and volume acceptable, and check that the
results are the same as found above. ◭
Section 2.7. User-defined functions 39

2.7 User-defined functions

In many situations one wants to define a function in MATLAB that is not a built-in
function. This is done using a function M-file. Note that a function M-file always starts
with the word function.

◮ Example 2.26 The following user-defined function (saved as the function M-file
areacirc.m) has input argument r equal to the radius or vector of radii of some
circles and output argument equal to the areas of these circles.

function A=areacirc(r)
% Compute areas of circles
A=pi*r.^2; % the variable name A is the same as in first line

The following MATLAB program (saved as a separate M-file, say progcirc.m) uses
this function M-file to evaluate the area of a circle of given radius and to plot the
area of a circle against its radius for values less than or equal to the given radius.

% Compute area of a circle of given radius

clear
rad=input(’Enter given radius: ’);
Area=areacirc(rad);
disp(’Area of circle of given radius is’)
disp(Area)
% Plot area of circle against its radius
rvec=0:0.01*rad:rad; % vector of radii
Avec=areacirc(rvec); % corresponding vector of areas
plot(rvec,Avec)
xlabel(’radius’)
ylabel(’area’)
title(’Graph of area of circle’)

Enter these two M-files and execute the program. ◭

Consider the problem of finding a zero of a function f (x) (or a root of the equation
f (x) = 0), i.e., a value z such that f (z) = 0. Graphically, the value of z is the x-coordinate
of a point at which the graph of f (x) crosses the x-axis.
MATLAB has a command called fzero that implements a sophisticated iterative
algorithm for finding the zero of a function near a given point. The algorithm has the
important properties that it is guaranteed to converge to a zero and it converges quickly
once the iterates become sufficiently close to the zero. Note that the computed zero is
not the exact zero but a very good estimate of it. The form of the command is
40 Chapter 2. Introduction to MATLAB

z=fzero(’func’,x0)

where func is either a built-in MATLAB function (e.g. sin) or a user-defined function,
and x0 is a rough estimate of the zero.

◮ Example 2.27 Find the smallest positive zero z of the function f (x) = x sin(x).
First get a rough estimate of z using a hand-drawn or MATLAB plot. This gives
z ≈ 3. Then create the function M-file called (say) fxsin.m containing

function y=fxsin(x)
y=x.*sin(x);

and a program M-file containing

% Find zero of a function

clear
format long
z=fzero(’fxsin’,3);
disp(’zero is: ’)
disp(z)

Run the program to see the result. The exact value of the zero is z = π. How
accurate is the computed value? ◭

In MATLAB 6 and on, one can also enter the expression for a simple function directly
in the fzero command as
z=fzero(’x*sin(x)’,0.8) (using the variable name x) or
z=fzero(inline(’t*sin(t)’),0.8) (using a variable name other than x).
This may be suitable for a simple expression, but in general one one should use a user-
defined function as above.

◮ Example 2.28 Consider Example 2.12 again, i.e. find the positive solution of
the equations y = cos(x) and y = x2 . If x and y satisfy these equations, then
y = x2 = cos(x), so x2 − cos(x) = 0. This means the x value of the solution is
the zero of f (x) = x2 − cos(x) for x > 0. A simple sketch shows that this zero is
near 0.8. To find the zero (and hence the solution of the equations) in MATLAB,
create a function M-file called func.m containing

function y=func(x)
y=x.^2 - cos(x);

and a program M-file containing

Section 2.7. User-defined functions 41

% Solve equations using fzero

clear
z=fzero(’func’,0.8);
y=z^2;
disp(’Solution of equations is [x y]: ’)
disp([z y])

Run this program to get the output

0.8241 0.6792

and compare the result with that obtained in Example 2.12. ◭

To find a local minimum point (a local trough in the graph) or local maximum point
(a local peak in the graph) of a function, one can plot the function in MATLAB and use
zoom and ginput (as before) to estimate the point. Clearly this also applies to a global
minimum point (deepest trough) and a global maximum point (highest peak).
Exercise Write a MATLAB program to plot the graph of y = x sin(6x) for x ∈ [0, 2π]
and use the zoom and ginput commands to find the coordinates of the global minimum
and global maximum points.
There is a special command fminbnd in MATLAB to accurately compute the location
of a local minimum of a function of one variable. The command is used as:

xmin=fminbnd(’funcm’,x1,x2)

where funcm is either a built-in MATLAB function (e.g. sin) or a user-defined function.
Then xmin has a value that locally minimizes the function in the interval between x1 and
x2. Note that xmin is only a local minimizer of the function in the interval [x1,x2]; it is not
necessarily a global minimizer in the interval. In other words, if the graph of the function
has several troughs in the interval, fminbnd will find the location of one of the troughs,
but not necessarily the deepest trough. To find a particular local minimum (perhaps the
global minimum), it is best to plot the function first and narrow down the interval so
it contains only the one desired local minimum. With this new interval entered in the
fminbnd command, the location of the desired minimum will be found. The minimum
value of the function is then found using

ymin=funcm(xmin)
42 Chapter 2. Introduction to MATLAB

◮ Example 2.29 The following MATLAB program computes a local minimum point
of a function contained in a specified subinterval.

% Compute a local minimum point of a function contained

% in a subinterval identified from the graph
clear
a=input(’Enter left hand endpoint of domain interval: ’);
b=input(’Enter right hand endpoint of domain interval: ’);
h=(b-a)/100;
x=a:h:b;
y=funcm(x);
plot(x,y)
xlabel(’x’)
ylabel(’y’)
pause % press any key to continue
x1=input(’Enter left hand endpoint of subinterval: ’);
x2=input(’Enter right hand endpoint of subinterval: ’);
xmin=fminbnd(’funcm’,x1,x2)
ymin=funcm(xmin)

We can use the program on the function f (x) = x sin(6x), which is defined in a
function M-file called funcm.m as follows:

function y=funcm(x)
y=x.*sin(6*x);

Run the program with the domain equal to [0, 2*pi]. Use the graph to identify
a small subinterval containing the global minimum (and no other local minima).
After entering this subinterval, you should obtain the values

xmin=6.0260
ymin=-6.0237

Run the program with both the domain interval and the subinterval equal to
[0, 2*pi]. Do you get the same results as before? ◭

In MATLAB 6 and on, it is possible to enter the expression for a simple function
directly in the fminbnd command, for example
y=fminbnd(’x^4-2*x’,-2,2) (using the variable name x) or
y=fminbnd(inline(’t^4-2*t’),-2,2) (using a variable name other than x).
This may be suitable for a simple expression, but in general one should use a user-defined
function as above.
Section 2.7. User-defined functions 43

The fminbnd command can also be used to compute a local maximum point of a
function f (x). To do this, let g(x) = −f (x) and compute a local minimizer xm of g(x)
using the fminbnd command with a function M-file defining g(x). Then xm also defines a
local maximum of f (x). To convince yourself of this, draw a simple graph of a function
f (x) having a local maximum at x =xm. Then draw the graph of g(x) = −f (x) by simply
reflecting the graph of f (x) about the x-axis. Notice that g(x) has a local minimum at
x =xm.
Exercise Use the program in Example 2.29 with a new function M-file to find the global
maximum point of f (x) = x sin(6x) for 0 $ x $ 2π.
Chapter 3

Computer Arithmetic

3.1 Numbers and their representations

The numbers most commonly used in scientific work are the so-called real numbers. Each
real number x has an infinite decimal expansion

x = (±)Dn . . . D1 D0 .d1 d2 d3 . . . ,

where each of the Dn , Dn−1 , . . . , D1 , D0 , d1 , d2 , d3 . . . is a decimal digit, (that is, one of

0, 1, 2, 3, 4, 5, 6, 7, 8, 9). When we write the number x in this way we are in fact expressing
the absolute value of x as an infinite sum involving powers of the base number 10. Thus

|x| = Dn × 10n + . . . + D1 × 101 + D0 × 100 + d1 × 10−1 + d2 × 10−2 + d3 × 10−3 + . . . .

(Recall that 100 = 1, 10−1 = 1/10, 10−2 = 1/102 = 1/100, and so on.)
Included in the set of real numbers are all whole or natural numbers 1, 2, 3, . . . , all
integers, 0, ±1, ±2, ±3, . . . , and all fractions or rational numbers of the form p/q where p
and q are integers with q ∕= 0. In the decimal expansion of any whole number, the digits
d1 , d2 , d3 to the right of the decimal point are all equal to 0, and are usually not included.
The decimal expansions of some fractions terminate in the same manner. All others are
repeating, and the repeating parts are often indicated by lines or dots.

◮ Example 3.1 The decimal expansions of the rational numbers 198/125 and 125/198
are
198/125 = 1.584 and 125/198 = .63131313131 . . . = .631

The set of real numbers also includes all irrational numbers. Their decimal expansions
neither terminate nor repeat.
√
◮ Example 3.2 The decimal expansions of π and 2 begin
√
π = 3.14159265 . . . and 2 = 1.41421356 . . . .
44
Section 3.1. Numbers and their representations 45

Binary numbers The fact that 10 is used as the base for the everyday number system
is most likely a product of human anatomy. If people had 6 fingers on each hand then
numbers would probably most commonly be expressed in the duodecimal system (which
uses 12 as the base). The binary system uses the number 2 as the base and only two
digits, 0 and 1 are needed in binary expansions of real numbers. It is very suitable for
mechanical or electronic devices with components for which there are just two natural
states, for example switches that are either on (1) or off (0), or capacitors that are either
charged (1) or not charged (0).
The binary expansion of a real number x has the form

x = (±)Bm . . . B1 B0 .b1 b2 b3 . . . ,

where each of the Bm , Bm−1 , . . . , B1 , B0 , b1 , b2 , b3 . . . is a binary digit or bit, (that is 0

or 1). As with decimal expansions this means that

|x| = Bm × 2m + . . . + B1 × 21 + B0 × 20 + b1 × 2−1 + b2 × 2−2 + b3 × 2−3 + . . . .

(Recall that 20 = 1, 2−1 = 1/2, 2−2 = 1/22 = 1/4, 2−3 = 1/23 = 1/8, and so on.)
The disadvantage with this system is that longer strings of digits are needed to represent
integers or fractions to the same accuracy. For example, the sixteen digit binary number
1001000011111101 has the five decimal digit expansion 37117, and the binary expansion
of π to twenty places, namely, 11.00100100001111111010, approximates π no better than
its decimal expansion to six places. In general, about 10 binary digits are required to
match every three decimal digits, and the reason for this is that

210 = 1024 ≈ 1000 = 103 .

This approximation is commonly used in engineering and science and is well worth re-
membering. In fact when computer scientists speak of kilobytes or megabytes, they are
usually talking of 210 = 1024 or 220 = 1048576 bytes rather than a thousand or a million
bytes. (In computer jargon a byte consists of a small number of bits.)

Octal and hexadecimal numbers To reduce the tedium and higher likelihood of
error associated with long sequences of bits, it is often convenient to use a system with a
base that is a power of 2. The numbers 8 and 16 are the bases of the popular octal and
hexadecimal systems, respectively.
The 8 octal digits are normally written as 0, 1, 2, 3, 4, 5, 6, and 7. The representation of
numbers in octal is more compact than it is in binary because every octal digit corresponds
to three binary digits according to:

octal 0 1 2 3 4 5 6 7
binary 000 001 010 011 100 101 110 111
46 Chapter 3. Computer Arithmetic

The fact that the octal digits are the same symbols that we use in decimal expansions
has the potential for confusion. To avoid this we write the base as a subscript.

◮ Example 3.3 (367.35)8 is an octal number whose decimal expansion can be ob-
tained as follows:

(367.35)8 = 3 × 82 + 6 × 81 + 7 × 80 + 3 × 8−1 + 5 × 8−2

= 192 + 48 + 7 + 3/8 + 5/64
= 247 29
64
= (247.453125)10 .

There are 16 hexadecimal digits and for these we need another 6 symbols apart from
the 10 decimal digits. For these it is usual to use A, B, C, D, E and F . The decimal
equivalents of these hexadecimal digits are given by

A = 10 B = 11 C = 12 D = 13 E = 14 F = 15.

The representation of numbers in hexadecimal is much more compact than it is in

binary because every hexadecimal digit corresponds to four binary digits according to:

0 1 2 3 4 5 6 7 8 9 A B C D E F
0000 0001 0010 0011 0100 0101 0110 0111 1000 1001 1010 1011 1100 1101 1110 1111

To make it clear that we are using hexadecimal representation we write the base 16
as a subscript.

◮ Example 3.4 (5BE.3C)16 is a hexadecimal number whose decimal expansion can

be obtained as follows:

(5BE.3C)16 = 5 × 162 + 11 × 16 + 14 × 160 + 3 × 16−1 + 12 × 16−2

= 1280 + 176 + 14 + 3/16 + 12/256
60
= 1470 256
= (1470.234375)10

3.1.1 Base conversions

Examples 3.3 and 3.4 show how numbers in bases other than 10 can be converted to dec-
imal, i.e., simply use the meaning of the representation. We also need ways of converting
from decimal to other bases and between bases other than 10.
Section 3.1. Numbers and their representations 47

Conversions from decimal There is a very simple method for doing such conversions
of decimal numbers to other bases. We illustrate it by considering conversions to binary.
Suppose that Bm Bm−1 . . . B1 B0 is the binary expansion of a whole number N . This
means that

N = Bm × 2m + Bm−1 × 2m−1 + . . . + B1 × 21 + B0 × 20
(3.1)
= 2(Bm × 2m−1 + Bm−1 × 2m−2 + . . . + B1 × 20 ) + B0

and so when we divide N by 2 the remainder is B0 . This gives us a means of calculating

the last binary digit B0 .
Equation 3.1 tells us that
N = 2 × N1 + B 0 ,
where the binary expansion of N1 is Bm Bm−1 Bm−2 . . . B1 . So we find the next binary
digit B1 by dividing the quotient N1 by 2 and keeping the remainder.
By doing this repeatedly until the quotient is 0, we find all of the binary digits, in the
order B0 , B1 , . . . , Bm .

◮ Example 3.5 To convert the decimal number 6023 to binary, divide repeatedly by
2 and note the remainders. Since

6023 = 2 × 3011 + 1
3011 = 2 × 1505 + 1
1505 = 2 × 752 + 1
752 = 2 × 376 + 0
376 = 2 × 188 + 0
188 = 2 × 94 + 0
94 = 2 × 47 + 0 ,
47 = 2 × 23 + 1
23 = 2 × 11 + 1
11 = 2×5 + 1
5 = 2×2 + 1
2 = 2×1 + 0
1 = 2×0 + 1

we see that (6023)10 = (1 011 110 000 111)2 . ◭

The same approach works for other bases. So, for example, to convert numbers to
hexadecimal, we divide repeatedly by 16, noting the remainders along the way, and to
convert to octal we repeatedly divide by 8.
48 Chapter 3. Computer Arithmetic

◮ Example 3.6 To convert the decimal number 6023 to octal, divide repeatedly by
8 and note the remainders. Since
6023 = 8 × 752 + 7
752 = 8 × 94 + 0
94 = 8 × 11 + 6 ,
11 = 8×1 + 3
1 = 8×0 + 1

we see that (6023)10 = (13607)8 . Notice how the digits in the octal expansion
can be obtained simply by grouping the binary digits in threes, starting from the
right. ◭

Conversions to decimal In principle we can convert to decimal from other bases in

the same way as above, that is, by repeated division by 10, where the arithmetic is done
in the original base. The problem with this is that people are not proficient at doing long
division in bases other than 10, so it is not a practical method.
It is much easier to simply use the meaning of the representation, that is to calculate
the powers of the base that we need and then add the appropriate multiples. All of the
arithmetic is done in decimal.

◮ Example 3.7 To convert the octal number 13607 to decimal, note that

(13607)8 = 1 × 84 + 3 × 83 + 6 × 82 + 7 × 80
= 1 × 4096 + 3 × 512 + 6 × 64 + 7 × 1
= 4096 + 1536 + 384 + 7
= 6023.

Binary, octal and hexadecimal conversions. Conversions between binary and octal
and between binary and hexadecimal are easy because we simply group the binary digits
in threes or fours respectively.

◮ Example 3.8

1. Binary to octal : (10110001101001)2 = (10 110 001 101 001)2 = (26151)8

2. Octal to binary : (34217)8 = (11 100 010 001 111)2 (The leading 0 is omitted)

3. Binary to hexadecimal (10110001101001)2 = (10 1100 0110 1001)2 = (2C69)16

4. Hexadecimal to binary (5E3B)16 = (101 1110 0011 1011)2 (The leading 0 is omitted)

5. Octal to hexadecimal (573)8 = (101 111 011)2 = (1 0111 1011)2 = (17B)16

6. Hexadecimal to octal (F 2C)16 = (1111 0010 1100)2 = (111 100 101 100)2 = (7454)8
Section 3.1. Numbers and their representations 49

3.1.2 Addition in other bases

If we are given two numbers x and y, both written in some base n, how do we find x + y in
base n? One way to do any kind of arithmetic is to convert everything to base 10, do the
arithmetic in base 10, and then convert the answer back into the desired base. However,
this is not the quickest way to proceed if the arithmetic only involves addition.
Let’s consider a situation that we are all familiar with, namely “time”. There are 60
seconds to a minute, and 60 minutes to an hour. Suppose we want to add the time 3
hours, 43 minutes, 16 seconds to the time 1 hour, 18 minutes, 49 seconds. First we add
the seconds to get 16 + 49 = 65 seconds. But 65 seconds is really 1 minute and 5 seconds.
So we “carry” the 1 minute to our calculation of minutes: 43 + 18 + 1 = 62 minutes.
But 62 minutes is really 1 hour and 2 minutes, so we carry the 1 hour: 3 + 1 + 1 = 5
hours. So the sum of the two times is 5 hours, 2 minutes, 5 seconds. This idea of carrying
digits is exactly the method we use for doing sums in any base. Before we see some
examples, make sure you realise that arithmetic with time is actually more complicated
than arithmetic in some base n, because time is expressed in a bizarre mix of bases: the
“ones” digits of the seconds and minutes are in base 10, while the “tens” digits of the
seconds and minutes are in base 6. (And if you look at the hours digits there is even more
bizarre behaviour!)
Now here are some examples to illustrate addition in any base.

◮ Example 3.9 Add (11101)2 and (1101)2 .

Solution: Here is the sum set out with carry digits indicated as boldface subscripts:
11 1 0 1
1 1 11 1 10 1
1 0 1 0 1 0
Explanation: In the 20 column, we add (1)2 and (1)2 to get (10)2 . So we write 0
under the horizontal line, and carry the 1 to the 21 column. The numbers in this
column now add to (1)2 , so we just write 1 as the next digit under the line. In the
22 column the sum is again (10)2 , so we write 0 under the line and carry the 1.
The 23 column then adds to (1)2 + (1)2 + (1)2 = (11)2 . So we write 1 under the
line and carry the 1. The 24 column then adds to (10)2 , so we write 0 under the
line and carry the 1. Since there is no 25 column, this last carried 1 is copied under
the line and we arrive at the answer to the sum. ◭

◮ Example 3.10 Add (32375)8 and (1644)8 .

Solution: Here is the sum set out with carry digits indicated as boldface subscripts:
3 2 3 7 5
11 16 14 4
3 4 2 4 1
50 Chapter 3. Computer Arithmetic

◮ Example 3.11 Add (AB2F1)16 and (5BC7)16 .

Solution: It is useful to draw a table at the side showing the decimal values of the
hexadecimal digits A,B,. . .,F. Here is the sum set out with carry digits indicated
as boldface subscripts:

A B 2 F 1
A B C D E F
1 5 1B C 7
10 11 12 13 14 15
B 0 E B 8

Note that with the aid of the table, the addition in each column can be done
mentally and the sum is then represented in hexadecimal, e.g. (F + C)16 is the
number twenty seven, which is 16 + 11 = (1B)16 , so we write B and carry the 1. ◭

It is natural to wonder about subtraction, multiplication and division in a general

base n. Unlike the case of addition, humans seem to perform these arithmetic operations
best by first converting to base 10, doing the arithmetic in that base, and then converting
back. For example, consider the octal subtraction (31)8 − (17)8 . The way we do this is:

(31)8 − (17)8 = 25 − 15 = 10 = (12)8 .

We perform multiplication similarly:

(7)16 × (B)16 = 7 × 11 = 77 = (4D)16 .

So the only new arithmetic you need to learn is addition in a general base.

Exercises
1. Make the following base conversions :

(a) (623651)8 to decimal,

(b) (7562)10 to binary,
(c) (3D43A)16 to binary,
(d) (3D43A)16 to octal,
(e) (10110111010)2 to octal,
(f ) (10110111010)2 to hexadecimal.

2. Compute the following sums :

(a) (11011001)2 + (10101010)2 ,

(b) (3241014)5 + (2414040)5 ,
Section 3.2. Scientific notation 51

(c) (A7E93A)16 + (B329F 56)16 ,

(d) (A21C2)16 + (10011110111001)2 .

3. Estimate the number of digits in :

(a) the binary expansion of 10345 ,

(b) the decimal expansion of 2239 ,
(c) the hexadecimal expansion of 2768 ,
(d) the decimal expansion of 1699 .

4. Write out a multiplication table which shows the products of all pairs of hexadecimal
digits.

3.2 Scientific notation

Engineers and scientists often prefer to express numbers in scientific or floating point
form. This form of representing numbers is especially useful for representing very large
or very small numbers.

◮ Example 3.12 The speed of light c in a vacuum is exactly 299, 792, 500 m/s. The
floating point form of c is
c = 0.2997925 × 109 ,
in which 0.2997925 is the mantissa and 9 is the exponent. ◭

Equally valid representations of the number c in scientific notation are

2.997925 × 108 , 29.97925 × 107 , and 0.02997925 × 1010 .

We move from one form to another by shifting the decimal point left or right and adjusting
the exponent accordingly. However we shall adopt the convention that the floating point
decimal representation of a non-zero real number x has the form

x = (±) 0.δ1 δ2 δ3 . . . × 10ε ,

where the exponent ε is an integer and where δ1 , δ2 , δ3 are decimal digits with δ1 ∕= 0. In
other words, the leading non-zero digit in the expansion is to occur immediately to the
right of the decimal point. This digit δ1 is often called the most significant digit, δ2 is
called the second-most significant digit and so on.
The (decimal) mantissa of any non-zero number x lies between 0.1 and 1, and so the
size or magnitude of x is reflected mainly by its exponent. If |x| is very large then the
exponent is large and positive, whereas if |x| is very small (i.e. x is very close to 0), then
the exponent is large and negative.
52 Chapter 3. Computer Arithmetic

◮ Example 3.13 Planck’s constant ! is a basic quantity in quantum physics. Its

value (in the appropriate units) is given by

! = 0.6626196 . . . × 10−33

The large but negative exponent indicates that this is an extremely small num-
ber. ◭

The same idea can be applied to other systems besides decimals. For example, the
floating point binary representation of a non-zero real number x has the form

x = (±) 0.β1 β2 β3 . . . × 2γ ,

where the binary exponent γ is an integer, and where β1 , β2 , β3 , . . . are binary digits with
β1 ∕= 0. Similarly, octal floating point representations of non-zero numbers have the form

x = (±) 0.µ1 µ2 µ3 . . . × 8ν ,

where the octal exponent ν is an integer, and where µ1 , µ2 , µ3 , . . . are octal digits with
µ1 ∕= 0.

3.3 Computer storage of numbers

There are infinitely many real numbers, and no computer can store more than finitely
many of them. A typical computer stores and manipulates real numbers internally in
floating point form, with a power of 2 such as 2 itself, 8 or 16 as the base. To make the
following discussion a little more concrete, we shall imagine that we are dealing with a
computer that stores each such number in one 32-bit word, that is, as a sequence of 32
binary digits. Of these, one bit is used for the sign of the number, another bit for the
sign of the exponent, six bits are for the exponent, and twenty four bits for the mantissa.
We shall also assume that the computer stores real numbers in their floating point octal
form; thus the 24 bit mantissa corresponds to 8 octal digits, and the 6 bit exponent to 2
octal digits.

◮ Example 3.14 In the computer described above, a 32 bit word such as

. /
0 1 001 010 100 001 011 110 000 011 100 010
represents the octal number

0.41360342 × 8−(12)8 .

The first 2 bits represent the signs of the number and its exponent respectively
(we assume that 0 ↔ + and 1 ↔ −), the next 6 bits (001 010) represent the octal
exponent (12)8 and the last 24 bits (100 001 011 110 000 011 100 010) represent the
mantissa. ◭
Section 3.3. Computer storage of numbers 53

Overflow and underflow errors: The largest positive number that can be stored in
the form described above is, in octal form,

xmax = 0.77777777 × 8(77)8 ≈ 8(77)8 ,

which in decimal form is

xmax ≈ 863 = 2189 ≈ 1057 .
Attempts to store numbers greater than this result in overflow errors, and computations
usually halt in these cases.
Similarly, the smallest (floating point octal) positive number that can be stored is, in
octal form,
xmin = 0.10000000 × 8−(77)8 = 8−(100)8 .
(Recall that both the mantissa and the exponent are octal numbers.) The decimal form
of xmin is
xmin = 8−64 = 2−192 ≈ 10−58 .
Underflow errors occur when positive numbers smaller than xmin arise. These cause some
computers to halt calculations and in others the offending number is replaced by 0.

Roundoff errors: Since our computer uses a mantissa with 8 octal digits, it cannot
store accurately any number with more than 8 significant digits in its floating point octal
form. Such numbers must be approximated by numbers having no more than 8 octal
digits in the mantissa. One common and easy way to obtain such an approximation is to
chop all but the first few significant digits (8 in this case).
An alternative to chopping is the so called symmetric rounding. When symmetric
rounding is used, the last digit kept is adjusted up by one if the next digit is at least half
of the base. So for example, if only 3 octal digits are being kept, 4.233 would be rounded
to 4.23, but 4.236 would be rounded to 4.24. Symmetric rounding is more familiar with
decimals. For example, if we are rounding to 3 significant decimal figures, 5.231 is rounded
to 5.23 but 5.237 is rounded to 5.24.
The errors associated with these approximations are called roundoff errors. The round-
off error when using symmetric rounding is, on average, half the roundoff error when using
chopping.
There are two ways of looking at the error involved in approximating one number by
another. If the number x∗ is used to approximate another number x, then the absolute
error in the approximation is defined by

εabs = |x∗ − x| .
54 Chapter 3. Computer Arithmetic

On the other hand, the relative error compares the magnitude of the error with the
magnitude of the numbers in question, and is defined by

|x∗ − x|
εrel = .
|x|

(Of course this only makes sense if x ∕= 0.)

For numbers close to 1, absolute and relative errors are much the same, but in general
the two are quite different.

◮ Example 3.15 If we keep just the first three significant digits in the decimal ex-
pansion of c, the speed of light (as in Example 3.12) then the absolute error is

εabs = 0.2997925 × 109 − 0.299 × 109 = 0.0007925 × 109 = 0.7925 × 106 ,

and the relative error is

0.7925 × 106
εrel = = 2.6435 × 10−3 .
0.2997925 × 109
On the other hand, if we keep just the first three significant digits in the decimal
expansion of Planck’s constant !, (as in Example 3.13) then the absolute error is

εabs = 0.6626196 . . . × 10−33 − 0.662 . . . × 10−33 = 0.0006196 × 10−33 = 0.6196 × 10−36 ,

and the relative error is

0.6196 × 10−36
εrel = = 0.93508 × 10−3 .
0.6626196 . . . × 10−33
In the first case the absolute error is quite large, and in the second it is extremely
small. However the relative error is approximately 10−3 in both.

Roundoff error and the size of the mantissa: The results in Example 3.15
illustrate the fundamental property that when we chop or round numbers to K significant
decimal digits, the relative roundoff error (in practically all cases) is approximately 10−K .
Similar results hold for other bases. For example, in the computer described at the
beginning of this section, where 24 binary digits are set aside for storing the 8 octal
digit mantissa of a non-zero number, the relative roundoff error will be (in most cases)
approximately 8−8 . Since 8−8 ≈ 10−7 , such a computer stores numbers correct to about
7 significant decimal digits.

IEEE floating point standard: Beginning in 1985, all computers used the ANSI/IEEE
(American National Standards Institute/Institute of Electrical and Electronic Engineers)
Standard 754-1985 for floating point arithmetic. This standard was superseded in 2008
Section 3.4. Subtractive cancellation 55

by the current version (IEEE 754-2008), which includes nearly all of the original IEEE
754-1985 standard as well as the IEEE Standard for Radix-Independent Floating-Point
Arithmetic (IEEE 854-1987). For the usual double precision representation, the IEEE
754-2008 standard has a 64-bit word and uses base 2. The Standard involves an extension
of the ideas just discussed, and although the details will not be discussed further here,
you can find more information about this in appropriate books and on the web.

3.4 Subtractive cancellation

When we multiply or divide two numbers expressed in scientific form, we essentially deal
with the mantissas and exponents separately. For example

c × ! = (0.2997925 × 109 ) × (0.6626196 . . . × 10−33 ) = 0.1986482 × 10−24 .

Notice that the mantissa of the product 0.1986482 is the product of the mantissas (0.2997925
and 0.6626196), and the exponents 9 and −33 are added to give the exponent of the prod-
uct. (Sometimes it is necessary to adjust the exponent down by one.)
If we chop both c and ! to 3 significant figures and multiply, we get

c × ! = (0.299 × 109 ) × (0.662 . . . × 10−33 ) = 0.197938 × 10−24 ,

which is still accurate to about 3 significant decimal digits.

This is typical of machine multiplication and division. In fact if the relative errors in
two non-zero numbers are small, then the relative error of their product or quotient is no
more than the sum of the relative errors. This tells us that when we multiply numbers
rounded to a certain number of significant figures, then the answer we get is accurate to
about the same number of significant figures.
We don’t lose much accuracy when we add positive numbers either. In fact if we add
two positive numbers rounded to a certain number of significant figures, then the answer
is also accurate to about the same number of significant figures. However, we can run
into trouble when we do subtraction.
If two numbers are very close, then their difference is much smaller in magnitude, and
so any errors that we make in estimating these numbers become much more significant.
This phenomenon is known as subtractive cancellation.

◮ Example 3.16 The derivative f ′ (x) of a function f at a point x can be approxi-

mated by calculating the average difference between the values of the function at
points near x. Precisely

f (x + h) − f (x)
f ′ (x) ≈ .
h
56 Chapter 3. Computer Arithmetic

This difference quotient must give a better and better approximation of f ′ (x) as
h → 0. However if the values of f (x) are known only to a certain number of
significant digits, then there is a definite limit to the accuracy to which we can
approximate f ′ (x).
√
Consider the case if f (x) = x, and we want to estimate f ′ (2), the derivative
at x = 2, using only 5 decimal figure accuracy. The following table shows the value
of the difference quotient, for various choices of h.
√ √ √ √ √ √
h 2+h 2 2 + h − 2 ( 2 + h − 2)/h
1 1.7321 1.4142 0.3179 0.3179
10−1 1.4491 1.4142 0.0349 0.349
−2
10 1.4177 1.4142 0.0035 0.35
−3
10 1.4146 1.4142 0.0004 0.4
10−4 1.4142 1.4142 0 0

None of these is a particularly good estimate of the exact answer, which we can show
√
using calculus is 1/(2 2) which is 0.35355, to 5 significant figures.

Exercises
5. Find the absolute and relative errors in the following calculations, given that the
numbers 7.351 and 19.163 are to be chopped to 3 significant digits:

(a) 7.35 + 19.1

(b) 7.35/19.1
(c) 7.35 × 19.1
(d) 19.1 − 7.35

6. The sum
220
1 1 1 1 1 1 1
= + + + + . . . + +
n=1
n2 1 4 9 16 361 400
can be calculated by adding the terms in decreasing order of magnitude, as indicated
here, or in increasing order of magnitude, as in

220
1 1 1 1 1 1
2
= + + ... + + .
n=1
n 400 361 9 4 1

(a) Carry out these summations, using only 2 significant figures at each stage of
the calculations?
(b) Which answer is more likely to be correct? Justify your claims.
Section 3.4. Subtractive cancellation 57

7. Consider a quadratic equation ax2 + bx + c = 0, with b > 0 and b2 − 4ac > 0.

The following MATLAB program computes the smaller (in magnitude) root of the
equation in two ways.

% Compute root of quadratic equation q(x)=ax^2+bx+c=0 in two ways

clear
format long
a=1; c=1;
b=input(’Enter b: ’);
r=(-b+sqrt(b^2-4*a*c))/(2*a)
qatr=a*r^2+b*r+c
s=-2*c/(b+sqrt(b^2-4*a*c))
qats=a*s^2+b*s+c

Execute the program with b = 5, 10, 100 and 1000. Why is the second computed
root more accurate than the first, especially for large b?

8. Consider the problem of computing the sum of the finite geometric series

S = 1 + r + r2 + · · · + rn ,

where the common ratio satisfies 0 < r < 1 and n is large. Is it better to compute the
sum in ascending or descending order to obtain the most accurate value? (Remember
there will be accumulated roundoff errors in the computed values.) Since we know
a formula for the exact value of the sum, we can just compute the sum in both ways
and compare the errors. This is implemented in the following MATLAB program,
with a range of r from 1/50 to 49/50 in increments of 1/50.

% Sum a geometric series in descending and ascending order

clear
format long
n=input(’Enter highest power: ’);
nr=49;
rr=zeros(1,nr); % initialize vector size
errd=zeros(1,nr); erra=zeros(1,nr); % initialize vector size
for j=1:nr
r=j/(nr+1);
rr(j)=r;
sumd=0;
suma=0;
58 Chapter 3. Computer Arithmetic

for i=0:n
sumd=sumd+r^i;
suma=suma+r^(n-i);
end
sumex=(1-r^(n+1))/(1-r); % formula for exact sum
errd(j)=sumd-sumex;
erra(j)=suma-sumex;
end
plot(rr,errd,rr,erra,’--’)
xlabel(’Common ratio r’)
ylabel(’Errors in finite sum of geometric series’)
title(’Sum order: descending (solid), ascending (-)’)

Execute the above program with n = 100. What do you observe about the errors
in the two computed sums? Execute the program again with n = 500.
Note that care must be taken in the use of the exact formula for sumex. If r is very
close to 1, subtractive cancellation in both the numerator and denominator will lead
to a large roundoff error in the computed value.

3.5 Solutions to exercises

1. (a) (206761)10
(b) (1110110001010)2
(c) (111101010000111010)2
(d) (752072)8
(e) (2672)8
(f ) (5BA)16

2. (a) (110000011)2
(b) (11210104)5
(c) (BDA8890)16
(d) (A497B)16

3. (a) Approximately 1150 binary digits

(b) Approximately 72 decimal digits
(c) 192 hexadecimal digits
Section 3.5. Solutions to exercises 59

(d) Approximately 119 decimal digits

5. (a) absolute error = 0.064, relative error ≈ 0.0024

(b) absolute error ≈ 0.0012, relative error ≈ 0.0032
(c) absolute error ≈ 0.48, relative error ≈ 0.0034
(d) absolute error = 0.062, relative error ≈ 0.0052

6. The value of the sum computed in ascending order is more accurate.

7. Note that for a = c = 1 and large b, subtractive cancellation occurs in the term
√
-b+sqrt(b^2-4*a*c) because b2 − 4ac is close to b. This makes the computed
root r less accurate than s.
Chapter 4

Counting & Discrete Probability

4.1 Basic Set Theory

Set theory is the branch of mathematics that studies sets, which are well-defined collections
of objects, called elements or members. Although any type of objects can be collected
into a set, set theory is applied most often to objects that are relevant to mathematics.
The modern study of set theory was initiated by Cantor and Dedekind in the 1870s.
Formalised using first-order logic, set theory is the most common foundational system for
mathematics and is a branch of mathematics in its own right, with an active research
community.
We will often use capital letters A, B, X, Y , . . . to denote sets, and lower case letters
a, b, x, y, . . . to denote elements of sets. We will sometimes use the terms “class”,
“collection”, or “family” to refer to a set.
We use the notation a ∈ S to specify that an object a belongs to a set S, and a, b ∈ S
to specify that both a and b belong to a set S. Here the symbol ∈ means “is an element
of”. We use the symbol ∕∈ to mean “is not an element of”.

4.1.1 Sets & Subsets

To specify a set, we either list its elements separated by commas, or state properties that
characterise the elements in the set. For example, the set S of primes less than 11 is
written as S = {2, 3, 5, 7} or S = {x | x is a prime, x < 11}. The latter set notation is
read “the set of all x such that x is a prime and x is less than 11.” Observe that 11 ∕∈ S.
Some sets of numbers occur in mathematics so frequently that it is convenient to set
aside special symbols to denote them. We shall adopt the following conventions.

60
Section 4.1. Basic Set Theory 61

N := { x | x is a natural number } = {0, 1, 2, . . .}

N+ := { x | x is a positive natural number } = {1, 2, 3, . . .}
Z := { x | x is an integer } = {. . . , −2, −1, 0, 1, 2, . . .}
Q := { x | x is a rational number } = the set of all fractions
R := { x | x is a real number }
R+ := { x | x is a positive real number }
I := {x | x ∈ R, x ∈ / Q} (irrational numbers)

(There is disagreement among mathematicians about whether or not the number zero
should belong to N, the so-called natural numbers. Consequently, some textbooks will
have 0 ∈ N while others will have 0 ∈/ N. Watch out for this when you are looking at
other textbooks.)

! Suppose that every element in a set A is also an element of a set B, i.e., a ∈ A

implies a ∈ B. Then A is said to be a subset of B. We also say that A is contained
in B or that B contains A. This relation is expressed as

A ⊆ B.

! Two sets are equal if they have exactly the same elements as each other, or equiva-
lently if each set is contained in the other. That is,

A=B if and only if B ⊆ A and B ⊆ A.

For example, the three sets A = {x | 3x2 + 4x + 1 = 0}, B = {−1, −1/3}, and
C = {−1, −1/3, −1} are equal.

! If A is not a subset of B, i.e., at least one element of A is not in B, then we write

A ∕⊆ B.

! If A ⊆ B, but A ∕= B, then we say that A is a proper subset of B and we write

A ⊂ B.

Theorem 4.1 Let A, B, and C be any sets. Then:

i) A ⊆ A;

ii) if A ⊆ B and B ⊆ A, then A = B;

iii) if A ⊆ B and B ⊆ C, then A ⊆ C.

62 Chapter 4. Counting & Discrete Probability

4.1.2 Universal Set & Empty Set

Unless otherwise stated, all sets are assumed to be contained in some fixed set U, called
the universal set. It is often clear from the context what we should take as our universal
set. For example, if we speak of the set of all positive integers that are prime numbers,
then we would assume that our universal set is the set of all positive integers, denoted by
N+ or Z+ .
Given a universal set U and a property P , there may not be any elements of U that
have property P . For example, the following set has no elements:

{x | x is an odd positive integer, x/2 ∈ Z}.

There is only one such set with no elements, which is called the empty set and is denote
by ∅. That is, if S and T are both empty sets then S = T = ∅, since S and T have
exactly the same elements (none). The empty set is regarded as a proper subset of every
other non-empty set. More formally, for any set A, we have ∅ ⊆ A ⊆ U. This statement
is best proved to be true by showing that it cannot be false. We observe that ∅ ⊆ A is
false only if there is an element in ∅ which is not in the set A. Since ∅ has no elements
at all, this is absurd. We must conclude that ∅ ⊆ A for every set A.
◮ Example 4.1 For any set A, the power set P(A) is defined to be the set of all
subsets of A. Some examples are below.

(a) P(∅) = {∅}, since ∅ is the only subset of ∅.

3 4
(b) P({a}) = ∅, {a} .
3 4
(c) P({a, b}) = ∅, {a}, {b}, {a, b} .
3 4
(d) P({a, b, c}) = ∅, {a}, {b}, {c}, {a, b}, {b, c}, {a, c}, {a, b, c} . ◭

Exercises
1. Which of the following statements are true, and which are false, given that
A = {p, q, r}?

(a) p ∈ A
(b) p ⊂ A
(c) {p} ∈ A
(d) {p} ⊂ A
(e) p ∈ {A}
(f ) {p} ⊂ {A}
Section 4.1. Basic Set Theory 63

2. Which of the following sets are equal: ∅, {0}, {∅}, and {∅, ∅}?

3. Suppose that X is the 4-element set {a, b, c, d}.

(a) Write down the subsets of X.

(b) How many subsets of X are there?
(c) How many subsets of X contain the element d?

4.1.3 Set Operations

Union: The union of two sets A and B, denoted by A ∪ B, is the set of all elements that
belong to A or to B (or to both), i.e.,

A ∪ B = {x | x ∈ A or x ∈ B}.
Example: R = Q ∪ I.

Intersection: The intersection of two sets A and B, denote by A ∩ B, is the set of all
elements that belong to both A and B, i.e.,

A ∩ B = {x | x ∈ A and x ∈ B}.
Disjoint Union: Two sets A and B are said to be disjoint (or non-intersecting) if A
and B have no elements in common, i.e., A and B are disjoint if A ∩ B = ∅, in
which case A ∪ B is called the disjoint union of A and B.
Example: The set of all real numbers R is the disjoint union of the rational numbers
(fractions) Q and the irrational numbers I.

Difference: The difference of two sets A and B, denoted by A \ B (or sometimes A − B),
is the set of elements that belong to A but do not belong to B, i.e.,

A \ B = {x | x ∈ A and x ∕∈ B}.

Note that if A = B, then A \ B = B \ A = ∅; otherwise A \ B is non-empty.

Example: I = R \ Q.

Symmetric Difference: The symmetric difference of two sets A and B, denoted by

A ⊕ B, consists of all those elements that belong to A or belong to B, but do not
belong to both A and B, i.e.,

A ⊕ B = (A ∪ B) \ (A ∩ B) or A ⊕ B = (A \ B) ∪ (B \ A).
Complement: The complement of a set A with respect to a given universal set U,
denoted by Ā (or AC ), is the set of elements that belong to U but which do not
belong to A. That is,

Ā = {x | x ∈ U, x ∕∈ A} = U \ A.
64 Chapter 4. Counting & Discrete Probability

Direct Product: The direct product (or Cartesian product) of two sets A and B, denoted
by A × B, is the set of all ordered pairs (a, b) where a ∈ A and b ∈ B, i.e.,

A × B = {(a, b) | a ∈ A, b ∈ B}.

For example, (x, y) × {1, 2, 3} = {(x, 1), (x, 2), (x, 3), (y, 1), (y, 2), (y, 3)}. Note that
the order of the elements in a pair matters; for instance, (x, 1) is not the same as
(1, x) in that (1, x) ∕∈ (x, y) × {1, 2, 3}.

◮ Example 4.2 If U = {1, 2, . . . , 10}, A = {1, 2, 4, 8}, and B = {1, 3, 5, 7, 9}, then
(a) A ∪ B = {1, 2, 3, 4, 5, 7, 8, 9}
(b) A ∩ B = {1}
(c) A ⊕ B = {2, 3, 4, 5, 7, 8, 9}
(d) A \ B = {2, 4, 8}
(e) B \ A = {3, 5, 7, 9}
(f ) Ā = {3, 5, 6, 7, 9, 10} ◭

4.1.4 Venn Diagrams

A Venn diagram is a diagrammatical representation of sets in which sets are represented
by enclosed areas in the plane. The universal set U is represented by the interior of a
rectangle, and the other sets are represented by circles lying within the rectangle and
obeying the following rules.

! If B ⊆ A, then the circle representing the set B will lie entirely within the circle
representing the set A.
U

B A

! If A and B are disjoint sets, then the circles representing these sets do not overlap.
U

B A
Section 4.1. Basic Set Theory 65

! For any two sets A and B, it is possible that some elements are in A but not in B,
some elements are in B but not in A, some elements are in both A and B, and some
elements are neither A nor B. This more general situation is depicted below.
U

A B

4.1.5 Algebra of Sets

Under the operations of union, intersection, and complement, any given sets A, B, and
C satisfy the following fundamental laws.

Idempotent Laws

! A∪A=A

! A∩A=A

Associative Laws

! (A ∪ B) ∪ C = A ∪ (B ∪ C)

! (A ∩ B) ∩ C = A ∩ (B ∩ C)

Commutative Laws

! A∪B =B∪A

! A∩B =B∩A

Distributive Laws

! A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C)

! A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C)

Identity Laws

! A∪∅=A

! A∪U=U

! A∩U=A

! A∩∅=∅
66 Chapter 4. Counting & Discrete Probability

Involution Law

! Ā¯ = A

Complement Laws

! A ∪ Ā = U

! A ∩ Ā = ∅

! Ū = ∅

! ∅
¯ =U

DeMorgan’s Laws

! (A ∪ B) = Ā ∩ B̄

! (A ∩ B) = Ā ∪ B̄

Exercises
4. Let U = {a, b, c, d, . . . , x, y, z} (the set of letters of the English alphabet) and let us
define the following four sets:
A = {a, j, k}
E = {b, c, d, e, g, p, t, v}
F = {a, b, c, d, e, f, g, h, i, j, k, l, m}
V = {a, e, i, o, u}

Find each of the following subsets of U:

(a) A ∪ E
(b) V ∩ F
(c) F̄
(d) E \ V
(e) A ⊕ V
(f ) A ∪ (E ∩ F )

5. Prove that for any sets A and B, we have A∩B ⊆ A ⊆ A∪B and A∩B ⊆ B ⊆ A∪B.

6. Given any two sets A and B with A ∩ B = ∅, what are the sets A \ B and B \ A?

7. Prove the Distributive Law : A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C).

Section 4.1. Basic Set Theory 67

8. Let A and B be any two sets.

(a) Prove that B \ A = B ∩ Ā.

(b) Using part (a) and the laws of algebra of sets, prove that

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

This shows that the LHS or the RHS can be used to define A ⊕ B.

9. Given any two sets A and B such that A ∩ B̄ = ∅, what are the sets A ∩ B and
A ∪ B?

10. Show that we can have:

(a) A ∩ B = A ∩ C with A ∕= C and B ∕= C

(b) A ∪ B = A ∪ C with A ∕= C and B ∕= C

11. Let A and B be any sets.

(a) What are the sets A × ∅ and ∅ × A?

(b) Prove or disprove the following statement: if A × B = B × A, then A = B.

4.1.6 Finite Sets & Counting

A set S is said to be finite if S is empty or if S contains exactly m elements for some
positive integer m; otherwise S is said to be infinite.
For example:

! The set of denominations of Australian banknotes, namely {5, 10, 20, 50, 100}, is
clearly finite.

! The set of prime numbers {2, 3, 5, 7, 11, 13, . . .} is infinite (proved by Euclid).

! The unit interval on the real line, namely [0, 1] = {x ∈ R | 0 ≤ x ≤ 1}, is an infinite
set.

A set S is said to be countable if S is finite or if the elements of S can be arranged as an

infinite sequence (i.e., if the elements of the set can be paired with the positive integers),
in which case S is said to be countably infinite; otherwise S is said to be uncountable.
For example:

! Two sets are said to be equivalent if there is a one-to-one correspondence between

their elements (i.e., if we can pair them up). Sets that are equivalent to the positive
integers are countable.
68 Chapter 4. Counting & Discrete Probability

! It is well-known that the unit interval [0,1] on the real line is uncountable.

! Are the prime numbers countable?

The number of elements in a set S, denoted by |S| (or card(S)), is called the cardinality
of S. Note that |∅| = 0 since the empty set contains no elements.

Proposition 4.2 Suppose A and B are disjoint finite sets. Then A ∪ B is finite and
|A ∪ B| = |A| + |B|.

That is to say, if S is the disjoint union of finite sets A and B, then |S| = |A| + |B|.
For any finite sets A and B, the set A is the disjoint union of A \ B and A ∩ B; hence
we deduce the following result from Proposition 4.2.

Corollary 4.3 Suppose A and B are finite sets. Then |A \ B| = |A| − |A ∩ B|.

The next result is also an immediate consequence of Proposition 4.2.

Corollary 4.4 Suppose A ⊆ U. Then |Ā| = |U| − |A|.

Exercises
12. Prove Proposition 4.2.

13. Verify Proposition 4.2 and Corollaries 4.3–4.4 for the sets U = {1, 2, 3, 4, 5, 6},
A = {2, 4, 6}, and B = {1, 3, 5}.

4.1.7 Inclusion-Exclusion Principle

The following theorem gives a formula for |A ∪ B| when A and B are not disjoint.

Theorem 4.5 (Inclusion-Exclusion Principle) Suppose A and B are finite sets. Then
the sets A ∪ B and A ∩ B are finite, and we have

|A ∪ B| = |A| + |B| − |A ∩ B|.

The name comes from the idea that the principle is based on over-generous inclusion,
followed by compensating with exclusion. That is, to find the number of elements in A or
B (or both), i.e., the number of elements in the union of A and B, we count the number
of elements in A and the number of elements in B (inclusion) and then we subtract
the number of elements that are in both A and B (exclusion) since these elements were
counted twice.
Using mathematical induction (see Chapter 9, later), one can generalise the preceding
theorem to any number of finite sets. For instance, in the case of three finite sets A, B,
and C, we have

|A ∪ B ∪ C| = |A| + |B| + |C| − |A ∩ B| − |A ∩ C| − |B ∩ C| + |A ∩ B ∩ C|. (4.1)

Section 4.2. Basic Counting Techniques 69

Exercises
14. Convince yourself that formula (4.1) is true.

15. For a group of students undergoing a Mathematics degree, the following facts were
observed.
Out of a total of 50 students, 30 were taking a course in Statistics, 18 were taking
a course in Computer Science, and 26 were taking a course in Engineering. More-
over, 9 students were taking courses in both Statistics and Computer Science, 16
students were taking courses in both Statistics and Engineering, 8 students were
taking courses in both Computer Science and Engineering, and 47 students were
taking courses in at least one of the three areas: Statistics, Computer Science, and
Engineering.
Find the number of students who were:

(a) not taking courses in any of the three areas (Statistics, Computer Science, and
Engineering);
(b) taking courses in all three of the areas.

4.2 Basic Counting Techniques

In the previous section, we considered several results concerning the number of elements
in a set. We will now learn some basic techniques for counting elements of sets, and more
generally, for counting the number of different ways of arranging objects under certain
conditions.

4.2.1 Product & Sum Rules

We have the following two basic counting principles.
Product Rule: If A and B are independent events such that A can occur in m different
ways and B can occur in n different ways, then the events A and B together can occur
in m · n different ways.

More generally, let E1 , E2 , . . . , Ek be independent events and suppose that each event
Ei can occur in ni ways. Then all of the events together can occur in n1 n2 · · · nk different
ways.
Sum Rule: If A and B are events such that A can occur in m different ways and B can
occur in n different ways, then there are m + n different ways in which either A or B (but
not both) can occur.
70 Chapter 4. Counting & Discrete Probability

More generally, let E1 , E2 , . . . , Ek be events such that Ei and Ej cannot occur

simultaneously for all i, j with i ∕= j, 1 ≤ i, j ≤ k, and suppose that each event Ei can
occur in ni different ways. Then there are n1 + · · · + nk different ways in which (exactly)
one of the events can occur.
Note: The word “and” suggests the product rule and the word “or” suggests the sum
rule. Also note that a so-called event is defined more precisely in the context of probability
theory in Section 4.3 later.

Exercises
16. How many ordered pairs of integers (x, y) are there such that 0 < |xy| ≤ 5?

17. A cycling club consists of 17 male members and 12 female members. The club
needs to decide on which of its members will compete in a forthcoming cycling
championship. Assuming all members are available and capable of competing in
any event at the championship, determine the number of ways that the club can
choose:

(a) one member to compete in a 240km race;

(b) two members to compete in a 50km time trial, one male and one female;
(c) two members to compete in a 120km race.

4.2.2 Factorials & Binomial Coefficients

Before we learn some more general techniques for counting patterns and arrangements,
we need to recall the definition of a factorial and define so-called binomial coefficients.
Factorial: The factorial of a positive integer n, denoted by n!, is the product of all
positive integers less than or equal to n; that is;

n! = 1 · 2 · · · (n − 1) · n = n · (n − 1) · · · 2 · 1.
For example, 3! = 3 · 2 · 1 = 6, 4! = 4 · 3 · 2 · 1 = 24, and 5! = 5 · 4! = 5(24) = 120.
Note: By convention 0! = 1. This is consistent with the combinatorial interpretation
of there being exactly one way to arrange 0 objects, i.e., only one permutation of zero
elements (namely, the empty set ∅) – see Corollary 4.9 later.
5 6
Binomial Coefficients: For any positive integers n, r with r ≤ n, the symbol nr , read
as “n choose r”, is defined by
7 8
n n! n(n − 1) · · · (n − r + 1)
= = .
r r!(n − r)! r(r − 1) · · · 2 · 1
Section 4.2. Basic Counting Techniques 71

By convention, 7 8 7 8
n n! 0 0!
= = 1 and = = 1.
0 0!n! 0 0!0!
5n 6
Fact: r
has exactly r factors in both the numerator and denominator.
Examples:
7 8 7 8 7 8
7 7·6·5 10 10 · 9 8 8·7·6·5·4
= = 35 = = 45 = = 56
3 3·2·1 2 2·1 5 5·4·3·2·1

An important identity is the following.

Proposition 4.6 For any non-negative integers n, r with r ≤ n, we have

7 8 7 8 7 8 7 8
n n n n
= , or equivalently = where a + b = n.
n−r r a b

Note: Proposition 4.6 is easily verified using the definition of a binomial coefficient. The
identity is also evident from the left/right symmetry of Pascal’s Triangle (discussed in
the next subsection).
5 6
To compute 11 9
in the same way as in the preceding examples, there will be 9 factors
in both the numerator and denominator. But 11 − 9 = 2, so by Proposition 4.6 we have
7 8 7 8
11 11 11 · 10
= = = 55.
9 2 2·1

4.2.3 Pascal’s Triangle & Binomial Theorem

As you might have already seen before, the binomial coefficients can be arranged in a
5 6
triangular array of numbers, called Pascal’s Triangle, in which nr is the r-th entry in the
n-th row.
5 06
5 16 0 5 16
5 26 0 5 26 1 5 26
5 36 0 5 36 1 5 36 2 5 36
5 46 0 5 46 1 5 46 2 5 46 3 5 46
5 56 0 5 56 1 5 56 2 5 56 3 5 56 4 5 56
0 1 2 3 4 5
.. .. .. .. .. ..
. . . . . .

This triangle has many interesting properties (Google it!). Of particular interest is
the fact that the numbers in Pascal’s Triangle satisfy the following two properties.
72 Chapter 4. Counting & Discrete Probability

! The first and last number in each row is 1.

! Every other number in the triangle can be obtained by adding the two numbers
directly above it.

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

Since the numbers in Pascal’s Triangle are binomial coefficients, we can state the above
properties more succinctly in the form of an identity, as follows.

Proposition 4.7 For any positive integers n and r with r ≤ n, we have

7 8 7 8 7 8
n+1 n n
= + .
r r−1 r
5 6
The numbers nr are called binomial coefficients since they appear in the expansion
of (a + b)n according to the following well-known theorem.
Binomial Theorem: For any positive integer n and variables a, b, we have
n 7 8
2 7 8 7 8 7 8 7 8
n n n−i i n n n n−1 n n−1 n n
(a + b) = a b = a + a b + ··· + ab + b .
i=0
i 0 1 n−1 n

This is a remarkably handy theorem that crops up all over the place.

Exercises
18. Expand and simplify (2x4 − y)3 .

19. Find the fifth term in the expansion of (3x − 4)12 .

4.2.4 Permutations & Combinations

Roughly speaking:

! a permutation is an ordered arrangement of objects selected from a given set;

! a combination is an unordered arrangement of objects selected from a given set.

Section 4.2. Basic Counting Techniques 73

Most counting problems we shall be dealing with can be classified into one of four
categories. We will explain each of these categories by means of an example.
Consider the set {a, b, c, d}. Suppose we wish to choose two letters from the given set
of four letters. Depending on our interpretation, we may obtain the following answers.
Permutations with repetitions: The order of listing the letters is important, and rep-
etition is allowed. In this case, there are 4 · 4 = 16 possible choices.
aa ab ac ad
ba bb bc bd
ca cb cc cd
da db dc dd

Permutations without repetitions: The order of listing the letters is important, and
repetition is not allowed. In this case there are 4 · 3 = 12 possible choices.
ab ac ad
ba bc bd
ca cb cd
da db dc

Combinations without repetitions: The order of listing the letters is not important,
and repetition is not allowed. In this case there are 4·3
2
= 6 possible choices.
ab ac ad
bc bd
cd

Combinations with repetitions: The order of listing the letters is not important, and
repetition is allowed. In this case there are 4·3
2
+ 4 = 10 possible choices.
aa ab ac ad
bb bc bd
cc cd
dd

4.2.5 Permutations Without Repetitions

! An arrangement of a set of n distinct objects in a given order is called a permutation
of the objects (taken all at once).

! An arrangement of r ≤ n of these objects in a given order is called an r-permutation

or a permutation of n objects taken r at a time.
74 Chapter 4. Counting & Discrete Probability

For example:

! There are 6 (= 3 · 2 · 1) permutations of the three letters a, b, c (taken all at once),

namely: abc, acb, bac, bca, cab, and cba.

! The number of 2-permutations of the four letters a, b, c, d is 12 (= 4 · 3), namely:

ab, ac, ad, ba, bc, bd, ca, cb, cd, da, db, dc (as we determined previously).

More generally, we have the following result.

Theorem 4.8 For any positive integers n, r with r ≤ n, the number of r-permutations
of n distinct objects, denoted by P (n, r), is given by

n!
P (n, r) = n(n − 1)(n − 2) · · · (n − r + 1) =
(n − r)!

Note: There are r factors in the product n(n − 1)(n − 2) · · · (n − r + 1).

In the special case of P (n, r) with r = n, we have P (n, n) = n!. That is:

Corollary 4.9 The number of permutations of n distinct objects is given by

n! = n · (n − 1) · · · 2 · 1.

Exercises
20. Find the number of 3-letter words using only the six letters a, b, c, d, e, f without
repetition.

21. Find n if P (n, 2) = 56.

22. Find the number of ways a judge can award first, second, and third places in a
contest with 18 contestants.

23. Find the number of ways that 9 people can arrange themselves:

(a) in a row of chairs;

(b) around a circular table.

24. How many ways can n objects be arranged in a circle?

Section 4.2. Basic Counting Techniques 75

4.2.6 Permutations With Repetitions

Often we would like to know the number of permutations of a multiset, i.e., of a set of
objects, some of which are alike.

Theorem 4.10 Suppose a set of n objects can be sorted into k types: n1 objects of type
1, n2 objects of type 2, . . . , nk objects of type k. Then the number of permutations of
these n (= n1 + n2 + · · · + nk ) objects is

n!
n1 !n2 ! · · · nk !

For example, let us find the number of 9-letter words that can be formed using the
letters of the word unusually. We seek the number of permutations of 9 letters of which
9!
3 are u’s and 2 are l’s and no other letters are repeated. This number is 3!2!1!1!1!1! = 30240,
by Theorem 4.10.
5 6
Note: The expression appearing in Theorem 4.10 is often denoted by n1 ,n2n,...,nk , called
a multinomial coefficient.

4.2.7 Ordered Samples

An ordered sample of size r drawn from a set S is a selection of r elements of S chosen

from S one element after the other.
Sampling with replacement: Each time an element is chosen from a set S, it is re-
placed before choosing the next element. So, if S contains n elements, then each
time we make a selection there are n ways to choose an element. Hence, by the
Product Rule, the number of such samples is

r
9 · n:;· · · n< = n .
n
r times

Sampling without replacement: Each time an element is chosen from a set S, it is

not replaced before choosing the next element. So, if S contains n elements and we
are drawing a sample of size r from S, then such a sample is an r-permutation; that
is, the number of such samples is
n!
P (n, r) = n(n − 1) · · · (n − r + 1) =
(n − r)!
76 Chapter 4. Counting & Discrete Probability

Exercises
25. Find the number of different “words” that can be formed from all the letters in each
of the following words:

(a) maths
(b) running
(c) committee

26. In a small box of Christmas lights, there are 4 identical red lights, 2 identical green
lights, and 3 identical yellow lights. The lights are to be placed in a row along a
window sill. Find the number of different arrangements of the 9 lights.

27. A box contains 12 Cadbury TM dairy milk chocolate bars. Find the number of ordered
samples of size 3 if:

(a) you eat each chocolate bar before selecting the next one;
(b) you are indecisive and put each chocolate bar back in the box before selecting
the next one.

4.2.8 Combinations Without Repetitions

Let S be a set containing n distinct elements. A combination of the elements of S taken r

at a time is any selection of r distinct elements of S where order does not matter. Such a
selection is a subset of S consisting of r elements, called an r-combination. The number
of r-combinations of n distinct objects is denoted by C(n, r) (or n Cr ).

Theorem 4.11 For any positive integers n, r with r ≤ n, the number of r-combinations
of n distinct objects is given by

P (n, r) n!
C(n, r) = =
r! r!(n − r)!

This is the number of r-permutations of n distinct objects divided by r! where r! is the

number of ways that r distinct objects can be permuted. The division by r! “corrects”
for order not mattering in combinations.
5 6 5 6
Recall that the binomial coefficient nr was defined to be r!(n−r)!
n!
. Hence C(n, r) = nr .
Section 4.2. Basic Counting Techniques 77

Exercises
28. A jar contains 8 red jellybeans and 6 purple jellybeans (yum!). Find the number of
ways 2 jellybeans can be taken from the jar if:

(a) the jellybeans can be of any colour;

(b) the jellybeans must be of the same colour.

29. In a Discrete Mathematics class, consisting of 12 students, the lecturer has prepared
3 different final examinations. Find the number of different ways in which the
lecturer can allocate 4 students to each of the three exam papers, assuming each
students sits only one exam.

4.2.9 Combinations With Repetitions

Let us consider a particular problem that I faced the other day. A certain supermarket
sells k = 4 different flavours of yoghurt: blueberry (b), plain (p), strawberry (s), and
vanilla (v). In how many ways can I buy r = 8 tubs of yoghurt?
Since order does not matter, this is an example of counting combinations with repe-
titions. Each combination can be listed with the b’s first, followed by p’s, then s’s, and
lastly v’s. For instance, we have combinations of the form:

bbppssvv, bbbpsvvv, ppppsvvv, bbbbbvvv, and so on . . .

How do we count all possible combinations?

Suppose we code the above combinations with only two symbols, say 0 and 1, by
letting 0 represent a tub of yoghurt and 1 represent a change from one flavour of yoghurt
to another. Then each combination will require r = 8 zeros (one for each tub of yoghurt)
and k − 1 = 3 ones, where the first one denotes the change from b to p, the second one
from p to s, and the third one from s to v. For example, the above combinations will be
coded as follows:

00100100100, 00010101000, 10000101000, 00000111000.

Now we can easily count these “codewords” since each codeword contains r + k − 1 = 11
digits where r = 8 are 0’s and k−1 = 3 are 1’s. Hence the number of possible combinations
is
11 · 10 · 9
C(11, 8) = C(11, 3) = = 165.
3·2·1
Too many different possibilities for me to worry about, so I just bought one tub of each
flavour of yoghurt!
Using similar reasoning as above, we can prove the following more general result.
78 Chapter 4. Counting & Discrete Probability

Theorem 4.12 Suppose there are k different kinds of objects. Then the number of
combinations of r such objects is
C(r + k − 1, r) = C(r + k − 1, k − 1).

Exercises
30. (a) In how many ways can 100 be written as a sum of four non-negative integers?
That is, find the number of non-negative integer solutions to the equation
x + y + w + z = 100.
(b) In how many ways can 100 be written as a sum of four positive integers?

31. (a) Prove the following theorem of De Moivre: For any positive integer n, the
number of positive integer solutions to the equation x1 + · · · + xr = n is
C(n − 1, r − 1).
(b) Using De Moivre’s theorem, prove that the number of non-negative integer
solutions to the equation x1 + · · · + xr = n is C(n + r − 1, r − 1).

4.2.10 The Pigeonhole Principle

One of the most widely used tools for proving that a certain kind of arrangement or
pattern exists is the pigeonhole principle (a.k.a. Dirichlet drawer principle or shoebox
principle), which essentially says the following: if there are “many” pigeons and “few”
pigeonholes, then there must be two or more pigeons occupying the same pigeonhole.
More formally, we have the following theorem.
Theorem 4.13 (Pigeonhole Principle) If n pigeonholes are occupied by n + 1 or more
pigeons, then at least one pigeonhole is occupied by more than one pigeon.
This principle can be applied to many problems where one wants to show that a given
situation can occur. For example:
! If there are 13 people in a room, at least two of them are sure to have a birthday in

the same month.

! In a selection of 677 people from a telephone directory, there will be at least two
people whose first names and last names begin with the same letters.
More generally, we observe that if n pigeonholes are occupied by kn+1 or more pigeons
for some positive integer k, then at least one pigeonhole is occupied by k + 1 or more
pigeons.
Theorem 4.14 (Generalised Pigeonhole Principle) If m pigeons are placed into k
pigeonholes, then at least one pigeonhole will contain more than ⌊ m−1
k
⌋ pigeons (where ⌊x⌋
denotes the greatest integer less than or equal to x).
Section 4.3. Discrete Probability Theory 79

Exercises
32. Prove that any set of m + 1 distinct elements taken from the set {1, 2, . . . , 2m}
contains two consecutive numbers.

33. Suppose five points are chosen from the interior of a square with each side of length 2.
√
Show that the distance between at least two of the points must be no more than 2.

34. Determine the minimum number of people needed to ensure that at least three of
them are born in the same month.

35. An employee’s time-clock shows that she worked 81 hours over a period of 10 days.
Show that on some pair of consecutive days, the employee worked at least 17 hours.

4.3 Discrete Probability Theory

Probability theory is the branch of mathematics which describes random behaviour. A
random phenomenon is one in which individual outcomes are unpredictable, but over
many independent repetitions a regular pattern (or distribution) emerges. The most
basic type of question is the following: What is the chance of a particular event (or set of
events) happening?
Examples
! If we toss a coin once, it is not known in advance whether we’ll get a Head or a Tail,
but we expect that the chance of getting a Head, for example, is 0.5 because over
many tosses the ratio
No. of Heads
No. of tosses
will settle down to around 0.5.

! On roulette wheel, there are 37 possible slots where the ball might come to rest.
We cannot predict the outcome each time, but we can record the distribution of the
number of times each slot occurs over a long period of time.
There are two ways to get probabilities.

! Long term relative frequencies – For example, in a large number of tosses of a

“fair coin” we observe that the probability of a Head gets close to 0.5.

! A model for probabilities – For example, we suppose that we have a “perfect

coin” or “perfect dice” (i.e., we suppose that all outcomes equally likely), and thus
conclude that the probability of each result is 12 (in the case of coin tosses) or 16 (in
the case of rolling a dice).
(Note that we can collect data to test such models.)
80 Chapter 4. Counting & Discrete Probability

There is a straightforward formula that we use for calculating basic probabilities.

Question: What is the probability p of a particular outcome occurring?
Answer:
Number of ways in which the particular outcome can occur
p=
Total number of all possible outcomes
Remarks:

! Note that the numerator $ denominator with the numerator non-negative and the
denominator positive; hence we have 0 $ p $ 1.

! p = 0 implies the outcome is impossible.

! p = 1 implies the outcome is 100% certain.

! p = 0.5 means a 50/50 chance.

Calculating p means counting outcomes – we draw on what we have learnt so far.

4.3.1 Terminology & Notation

! A sample space S is set of all possible outcomes of a random phenomenon.

! An event is a set of possible outcomes.

◮ An event is a subset of the sample space.

◮ Events are usually denoted by uppercase letters, such as A, B, C, etc.

We can represent the sample space S and the events of a random phenomenon using set
theory and Venn diagrams, as follows.

! Complementary Event: If A is an event, then Ac denotes the complementary

event and is the set of all outcomes in S not contained in A.

! Disjoint Events: Two events A and B that have no outcomes in common are said
to be disjoint events, i.e., A and B are disjoint events if they cannot both occur at
the same time.

Clearly A and Ac are disjoint events.

Section 4.3. Discrete Probability Theory 81

! Intersection of Events A ∩ B: The intersection of two events A and B, denoted

by A ∩ B, is the event consisting of all outcomes which are in both A and B – it is
the shaded area in the following figure.

! Union of Events A ∪ B: The union of two events A and B is the event consisting
of all outcomes which are in A or B or both – it is the shaded area in the following
figure.

Along with the terminology and diagrammatical representations that we have just
considered, we have some mathematical notation too.

! P (A) = probability that event A will occur.

! P (A ∩ B) = probability that both events A and B will occur.

! P (A ∪ B) = probability that either one of events A or B will occur.

Some intuitive ideas:

! On a single coin toss, if we believe that P (H) = p (where H is the event of a Head
occurring), then we presume that P (T ) = 1 − p (where T is the event of a Tail
occurring), because if it isn’t a Head it must be a Tail, i.e., T = H c .

! Since probabilities can be thought of as long term relative frequencies, all rules for
probabilities must work for proportions as well.

These ideas lead us to some rules that probabilities must follow.

82 Chapter 4. Counting & Discrete Probability

4.3.2 Rules for Probabilities

RULE 1: For any event A, the probability of A occurring lies between 0 and 1, i.e.,
0 $ P (A) $ 1.
RULE 2: The collection S of all possible outcomes has probability 1, i.e., P (S) = 1.
Remarks:

! The empty set ∅ is the set with no outcomes. Hence P (∅) = 0.

! If two events A and B are disjoint, then A ∩ B = ∅ and hence P (A ∩ B) = 0.

RULE 3: For any event A, the probability that A does not occur is P (Ac ) = 1 − P (A).
RULE 4: For any events A and B, we have

P (A ∪ B) = P (A) + P (B) − P (A ∩ B)

This rule follows from Theorem 4.5 (Inclusion-Exclusion Principle).

Note: P (A ∪ B) is the probability that event A occurs, or that event B occurs, or that
both A and B occur.

Conditional Probability: If a card is drawn from a deck and not replaced, the outcome
for the first card could influence the outcome for the second card. Thus, the probability
of getting an Ace on the second card depends on the result for the first card. So, knowing
the result for event A may change the probability of another event B occurring.
Notation: P (B occurs given that A has occurred) ≡ P (B | A)
RULE 5: Provided P (A) > 0, the conditional probability of B given A is

P (A ∩ B)
P (B | A) =
P (A)
Section 4.3. Discrete Probability Theory 83

Notes:
! Since we know A has occurred, this is the same as restricting the sample space to
just those outcomes which make up event A.

! It is always true that

P (A ∩ B) = P (A)P (B | A) = P (B)P (A | B)

◮ Example 4.3 Suppose two cards are drawn from a deck without replacement. Let
A be the event that the first card is an Ace and let B be the event that the second
card is an Ace. The probability that the two cards are both Aces is
P(both cards are Aces) = P (A ∩ B) = P (A)P (B | A)
where
P (A) = 4/52 and P (B | A) = 3/51
and hence
P (A ∩ B) = (4/52) × (3/51) ≈ 0.0045. ◭

Independent Events: Two events A and B are independent if knowledge about whether
A occurs does not alter the probability that B also occurs.
Hence, for independent events, we have
P (B | A) = P (B)
P (A | B) = P (A)
P (A ∩ B) = P (A)P (B)

◮ Example 4.4 Now suppose two cards are drawn from a deck with replacement.
Let A be the event that the first card is an Ace and let B be the event that the
second card is an Ace. Clearly the two events are independent and we have
P (A) = 4/52 and P (B | A) = P (B) = 4/52.
So the probability of both cards being Aces is
P (A ∩ B) = P (A)P (B) = (4/52) × (4/52) = (4/52)2 ≈ 0.0059. ◭

Note: Always use

P (A ∩ B) = P (A)P (B | A) = P (B)P (A | B)
unless A and B are known to be independent.
84 Chapter 4. Counting & Discrete Probability

Law of Total Probability: P (A) = P (A ∩ B) + P (A ∩ B c )

◮ Example 4.5 A certain brand of lawn seed contains 35% Couch and 65% Rye
Grass. When the seed is sown 90% of the Couch germinates and 83% of the Rye
Grass germinates.
For a randomly chosen seed, let C be the event that the seed is Couch, let R
be the event that it is Rye Grass and let G be the event that it germinates.

(a) Write the given information in probability notation.

We have the following probabilities from the given information.

P (C) = 0.35
P (R) = 0.65
P (G | C) = 0.90
P (G | R) = 0.83

(b) Are the events R and C disjoint?

Yes, R and C are disjoint events because the lawn seed cannot be of both types,
i.e., it is either Couch or Rye Grass, but cannot be both. In particular, we have
Rc = C (and C c = R), i.e., R and C are complementary events.

(c) Find P (C ∩ G) and P (R ∩ G) and hence find P (G), i.e., the probability that the
seed germinates.
Using Rule 5, we have

P (C ∩ G) = P (C)P (G | C) = 0.35 × 0.90 = 0.315

P (R ∩ G) = P (R)P (G | R) = 0.65 × 0.83 = 0.5395
Section 4.3. Discrete Probability Theory 85

Furthermore, by the Law of Total Probability, we have

P (G) = P (G ∩ R) + P (G ∩ Rc )

where Rc = C, so

P (G) = P (G ∩ R) + P (G ∩ C) = 0.5395 + 0.315 = 0.8545.

(d) Are the events R and G independent?

The events R and G are not independent because germination depends on the type
of lawn seed. In particular, we see that P (G | R) ∕= P (G) and P (R∩G) ∕= P (R)P (G)
so G and R are clearly not independent events.

(e) Explain the meaning of P (C | G) in context and calculate it.

P (C | G) is the probability that the lawn seed is Couch given that it germinates.
Using Rule 5, we have

P (G ∩ C) 0.315
P (C | G) = = ≈ 0.3686. ◭
P (G) 0.8545

Exercises
36. The American Red Cross says that 45% of the US population has Type O blood,
40% Type A, 11% Type B, and the rest Type AB.
For a randomly selected blood donor:

(a) What is the probability that this donor has Type AB blood?
(b) What is the probability that this donor has Type A or Type B blood?
(c) What is the probability that this donor does not have Type O blood?
(d) It is also true that 85% of the population has Rh+ blood, and that this is
independent of blood group. What is the probability that this donor has Type
O, Rh+ blood?
86 Chapter 4. Counting & Discrete Probability

4.3.3 Random Variables

! A random variable is a variable whose value is the numerical outcome of a random
phenomenon.

◮ We use capital letters, such as X, Y , to denote a random variable; it is what

we might get if we do an experiment.
◮ When we collect data, and get actual values, we denote these by the corre-
sponding lower case letters, such as x, y.

! A discrete random variable is random variable X that can take one of only a
countable (often finite) number of possible different values x1 , x2 , . . .

◮ A probability model assigns probabilities to each of these possible outcomes:

P (X = xi ) = pi

◮ The sum of the P (X = xi ) for all possible values of xi is 1.

◮ Probabilities are usually presented in a table, as shown below.

4.3.4 Probability Distribution of a Discrete Random Variable

To specify a probability model, we need to assign probabilities to each possible outcome.
More precisely, suppose a discrete random variable X has k possible (different) outcomes
x1 , x2 , . . . , xk with corresponding probabilities p1 , p2 , . . . , pk . Then the following table
represents the probability distribution of X.

Values x1 x2 ... xk−1 xk Sum

P (X = xi ) p1 p2 ... pk−1 pk 1

For a discrete random variable X, the mean of X is defined to be

k
2
µX = xi × pi = x1 p1 + x2 p2 + · · · + xk pk
i=1

where pi = P (X = xi ).

Note: µX is a parameter as it describes the distribution of a random variable; it is a

measure of location for the probability distribution of X.
Section 4.3. Discrete Probability Theory 87

◮ Example 4.6 Suppose that in a particular local council area the number of dogs
that can be kept by a householder is limited to a maximum of three. Let D denote
the number of dogs in a household.
Records show the following probability distribution for D.

d 0 1 2 3
P (D = d) 0.5 0.3 0.15 0.05

We have k = 4 possible outcomes, namely d1 = 0, d2 = 1, d3 = 2, and d4 = 3, and

hence the mean of D is
4
2
µD = di × P (D = di )
i=1

= 0 × P (D = 0) + 1 × P (D = 1) + 2 × P (D = 2) + 3 × P (D = 3)
= 0 × 0.5 + 1 × 0.3 + 2 × 0.15 + 3 × 0.05
= 0.75

So the average number of dogs per household in the local council area is 0.75. ◭

For the discrete random variable X, the standard deviation of X is defined to be

=
> k
>2
σX = ? pi × (xi − µX )2 .
i=1

Note: σX is a parameter and is a measure of spread for the probability distribution of X.

◮ Example 4.7 For the previous example, we have

=
> k
>2
σD = ? P (D = di ) × (di − µD )2
i=1
@
= 0.5 × (0 − 0.75)2 + 0.3 × (1 − 0.75)2 + 0.15 × (2 − 0.75)2 + 0.05 × (3 − 0.75)2
= 0.8874 ◭

4.3.5 Transforming a Discrete Random Variable

Linear Transformation

! Let X be a discrete random variable with mean µX and standard deviation σX .

! If Y = a + bX, where a and b are constants, then it can be shown that

µY = a + bµX and σY = |b|σX

88 Chapter 4. Counting & Discrete Probability

◮ Example 4.8 A technician charges a call-out fee of $60 plus an additional $2 for
each minute that he spends on a job. He has found that the time (T ) taken on a
job has a mean (µT ) of 18 minutes and a standard deviation (σT ) of 3.5 minutes.
Let C be the total cost, excluding parts, of the job.

(a) Write down the formula for C in terms of T , and hence the values of a and b.
We have
C = 60 + 2T
and so a = 60 and b = 2 in this case.

(b) Hence find the mean and standard deviation of the cost C.
We have
µC = 60 + 2µT = 60 + 2 × 18 = 96 minutes
and
σC = 2σT = 2 × 3.5 = 7 minutes. ◭

4.4 Solutions to exercises

1. (a) True
(b) False
(c) False
(d) True
(e) False
(f ) False

2. The first three sets are all different. The third and fourth are equal.

3. (a) ∅, {a}, {b}, {c}, {d}, {a, b}, {a, c}, {a, d}, {b, c}, {b, d}, {c, d}, {a, b, c},
{a, b, d}, {a, c, d}, {b, c, d}, {a, b, c, d}
(b) 24 = 16
(c) 8

4. (a) {a, j, k, b, c, d, e, g, p, t, v}
(b) {a, e, i}
(c) {n, o, p, q, r, s, t, u, v, w, x, y, z}
(d) {b, c, d, g, p, t, v}
Section 4.4. Solutions to exercises 89

(e) {j, k, e, i, o, u}
(f ) {a, j, k, b, c, d, e, g}

5. Every element in A ∩ B belongs to both A and B. Hence, for any x ∈ A ∩ B, we

have x ∈ A and x ∈ B. Therefore, A ∩ B ⊆ A and A ∩ B ⊆ B. Moreover, x ∈ A ∪ B
if x ∈ A or x ∈ B. Thus, every element of A belongs to A ∪ B (i.e., A ⊆ A ∪ B)
and also every element of B belongs to A ∪ B (i.e., B ⊆ A ∪ B).

6. Using a Venn Diagram, it is easy to see that if A ∩ B = ∅ then A \ B = A and

B \ A = B.

7. A ∩ (B ∪ C) = {x | x ∈ A, x ∈ B ∪ C}
= {x | x ∈ A, x ∈ B or x ∈ A, x ∈ C}
= {x | x ∈ (A ∩ B) or x ∈ (A ∩ C)}
= (A ∩ B) ∪ (A ∩ C)

8. (a) B \ A = {x | x ∈ B, x ∕∈ A}
= {x | x ∈ B, x ∈ Ā}
= {x | x ∈ B ∩ Ā}
=B∩A

(b) (A ∪ B) \ (A ∩ B) = (A ∪ B) ∩ (A ∩ B) (by part (a))

= (A ∪ B) ∩ (Ā ∪ B̄) (De Morgan’s Law)
= [(A ∪ B) ∩ Ā] ∪ [(A ∪ B) ∩ B̄] (distributive law)
= [(A ∩ Ā) ∪ (B ∩ Ā)] ∪ [(A ∩ B̄) ∪ (B ∩ B̄)] (distributive law)
= [∅ ∪ (B ∩ Ā)] ∪ [(A ∩ B̄) ∪ ∅] (complement law)
= (B ∩ Ā) ∪ (A ∩ B̄) (identity law)
= (A \ B) ∪ (B \ A)

9. Using a Venn diagram, it is easy to show that if A ∩ B̄ = ∅ then A ⊆ B, and hence

A ∩ B = A and A ∪ B = B.

10. (a) The sets A = {1, 2}, B = {2, 3}, C = {2, 4} satisfy A ∩ B = A ∩ C with A ∕= C
and B ∕= C.
(b) The sets A = {1, 2}, B = {2, 3}, C = {1, 3} satisfy A ∪ B = A ∪ C with A ∕= C
and B ∕= C.

11. (a) By definition, A × ∅ = {(x, y) | x ∈ A, y ∈ ∅}. But there does not exist any
element y ∈ ∅, so there cannot exist any pair (x, y) with x ∈ A and y ∈ ∅.
Therefore, A × ∅ contains no elements, i.e., A × ∅ = ∅. Similarly, ∅ × A = ∅.
90 Chapter 4. Counting & Discrete Probability

(b) It is not true in general that A × B = B × A implies A = B. To show this, let

A be any non-empty set and let B = ∅. Then A × B = ∅ = B × A (by part
(a)), but A ∕= B.

12. Proof of Proposition 4.2: Suppose A and B are disjoint finite sets. To count
the number of elements in A ∪ B, first count the number of elements in A, namely
|A, and then the only other elements of A ∪ B are those in B, but not in A. Since
A ∩ B = ∅, no element of B is in A, so there are |B| elements in B \ A. Hence
|A ∪ B| = |A| + |B| and this number is finite because both |A| and |B| are finite.

13. Suppose U = {1, 2, 3, 4, 5, 6}, A = {2, 4, 6}, B = {1, 3, 5}. Then |A ∪ B| = 6 =

|A| + |B| (verifying Proposition 4.2), |A \ B| = |A| = 3 = |A \ B| − |A ∩ B| (verifying
Corollary 4.3), and |Ā| = 3 = |U| − |A| (verifying Corollary 4.4).

14. Shade in a Venn diagram to convince yourself that formula (4.1) is true.

15. Let us define four sets:

U = M = {all students undergoing a Mathematics degree}

S = {all students taking a course in Statistics}
C = {all students taking a course in Computer Science}
E = {all students taking a course in Engineering}

The following information is given: |U| = |M | = 50, |S| = 30, |C| = 18, |S ∩ C| = 9,
|S ∩ E| = 16, |C ∩ E| = 8, |S ∪ C ∪ E| = 47.

(a) |S ∪ C ∪ E| = |U| − |S ∪ C ∪ E| = 50 − 47 = 3 (by Corollary 4.4).

(b) By formula (4.1), |S ∪ C ∪ E| = |S| + |C| + |E| − |S ∩ C| − |S ∩ E| − |C ∩ E| +
|S ∩ C ∩ E|. Therefore, from the given information, we have

47 = 30 + 18 + 26 − 9 − 16 − 8 + |S ∩ C ∩ E|

and hence |S ∩ C ∩ E| = 6.

16. 40 (Hint: Consider the sets Ei = {(x, y) ∈ Z2 : |xy| = i} for i = 1, 2, 3, 4, 5.)

17. (a) 17 + 12 = 29
(b) 17 · 12 = 204
(c) 29 · 28 = 812

18. Using the Binomial Theorem with a = 2x4 and b = −y, the (simplified) expansion
is 8x12 − 12x8 y + 6x4 y 2 − y 3 .
Section 4.4. Solutions to exercises 91

19. 831409920x8

20. 120

21. P (n, 2) = n(n − 1) = n2 − n = 56. Hence n2 − n − 56 = (n − 8)(n + 7) = 0. Since

n must be positive, the only answer is n = 8.

22. 18 · 17 · 16 = 4896

23. (a) P (9, 9) = 9!

(b) One person can sit at any place on the table. The other 8 people can arrange
themselves in 8! ways, so the answer is 1 · 8! = 8!. This is an example of
a circular permutation. In general, n objects can be arranged in a circle in
(n − 1)! ways.

24. See the answer directly above.

25. The problem concerns permutations with repetitions.

(a) There are five letters and no repetitions, so the answer is 5! = 120.
(b) There are 7 letters of which 3 are n and no other letter is repeated, so the
answer is 7!/3! = 840.
(c) There are 9 letters of which 2 are m, 2 are t, and 2 are e, so the answer is
9!/2!2!2!1!1!1! = 45, 360.

26. 9!/4!2!3! = 1260

27. (a) This is sampling without replacement, so the answer is P (12, 3) = 12!/9! =
1320.
(b) This is sampling with replacement, so the answer is 123 = 1728.

28. (a) C(14, 2) = (14 · 13)/(2 · 1) = 7 · 13 = 91

(b) There are C(8, 2) = 28 ways two choose 2 of the 8 red jellybeans and C(6, 2) =
15 ways to choose 2 of the 6 purple jellybeans, so by the sum rule, the total
number of ways is 28 + 15 = 43.

29. There are C(12, 4) = 495 ways to choose 4 of the 12 students to take the first exam,
and then C(8, 4) = 70 ways of choosing 4 of the remaining 8 students to take the
second exam. The remaining students take the third exam. Thus the answer is
495 · 70 = 34, 650.
92 Chapter 4. Counting & Discrete Probability

30. (a) We can view each of the solutions (say x = 25, y = 5, w = 20, z = 50) as a
combination of r = 100 objects consisting of 25 a’s, 5 b’s, 20 c’s, and 50 d’s,
where there are k = 4 kinds of objects. By Theorem 4.12, the number of non-
negative integers solutions is C(100 + 3, 3) = C(103, 3) = 103·102·101
3·2·1
= 176, 851.
(b) We wish to find the number of positive integer solutions to x + y + w + z = 100.
Consider writing 100 = 1 + 1 + · · · + 1 where there are 100 1’s and 99 + signs.
To decompose 100 into 4 summands, we need only choose 3 of the 99 + signs.
Hence, the number positive integer solutions to x + y + w + z = 100 is
7 8
99 99 · 98 · 97
C(99, 3) = = = 156, 849.
3 3·2·1

31. The proofs of parts (a) and (b) generalise what was done in the preceding exercise.

32. Let the following m pairs represent pigeonholes: (1, 2), (3, 4), . . . , (2m − 1, 2m). If
A is a set of m + 1 elements taken from the set {1, 2, . . . , 2m}, then for each element
j ∈ A, put j in pigeonhole (i, i + 1) if j = i or j = i + 1. Since A contains m + 1
elements, one of the pigeonholes (pairs) contains two distinct elements of A and
these two elements are consecutive.

33. Partition the square into 4 sub-squares of equal size, each of whose sides have
length 1. By the Pigeonhole Principle (Theorem 4.13), two of the points lie in
√
one of the sub-squares. The length of the diagonal of each square is 2 so the
√
distance between the two points is at most 2.

34. By the Generalised Pigeonhole Principle (Theorem 4.14), if the pigeonholes are the
12 months in a year (k = 12), then the minimum number m of people needed to
ensure that at least 3 of them are born in the same month satisfies ⌊(m−1)/12⌋ = 2.
Thus m = 2 × 12 + 1 = 25.

35. Let the 81 hours be pigeons and let the pigeonholes be the following 5 sets of
two consecutive days: {1, 2}, {3, 4}, {5, 6}, {7, 8}, {9, 10}. By the Generalised
Pigeonhole Principle (Theorem 4.14), the employee worked at least ⌊(81−1)/5⌋+1 =
16 + 1 = 17 hours.
Section 4.4. Solutions to exercises 93

36. (a) P (Type O) = 0.45, P (Type A) = 0.40, P (Type B ) = 0.11, P (Type AB ) =

1 − (0.45 + 0.4 + 0.11) = 0.04
(b) Using Rule 4 (Inclusion-Exclusion Principle):

P (Type A or Type B )= P (Type A) +P (Type B ) −P (Type A and Type B )

= 0.4 + 0.11 − 0 = 0.51

(c) P ((Type O)c ) = 1 − P (Type O) = 1 − 0.45 = 0.55 = P (Type A) +P (Type B )

+P (Type AB )
(d) Since blood type and Rhesus factor are independent, we have

P (Type O and Rh + ) = P (Type O)×P (Rh + ) = 0.45 × 0.85 = 0.3825.

Chapter 5

Matrices

5.1 Introduction
A matrix is a rectangular array of numbers. These numbers are called the terms,
elements, or entries of the matrix.

◮ Example 5.1 Four matrices are shown below.

A D
A D 1 A D A D
2 4 B2 E 3 5 −7 2 6
B E B E B E B E
A = C3 −6F , B = B E, C = C4 2 0 F, D = C−1 −5F .
C5 F
4 0 0 −1 6 0 3
3

We say that A is a 3 × 2 matrix, since it has 3 rows and 2 columns. Similarly, B

is a 4 × 1 matrix, C is a 3 × 3 matrix, and D is a 3 × 2 matrix. ◭

It is usual to use upper case letters to denote matrices and lower case letters to denote
their elements. We use subscripts to indicate the position of an element within a matrix.
Suppose T is an m × n matrix. We write
A D
t11 t12 t13 . . . t1n
B E
B t21 t22 t23 . . . t2n E
B E
T =B E
B t31 t32 t33 . . . t3n E ,
B .. .. .. . . . E
C . . . . .. F
tm1 tm2 tm3 . . . tmn

where tij is the element of T which appears in the ith row and the jth column. Instead of
writing down some of the elements of a matrix in their correct position as we have done
above, it is often convenient to use the following more compact form: T = [tij ]. These
equations are both meant to associate the elements tij with the matrix T .

94
Section 5.2. Matrix arithmetic 95

◮ Example 5.2 If [aij ] is the matrix A in Example 5.1, then a11 = 2, a12 = 4,
a21 = 3, a22 = −6, a31 = 4, and a32 = 0. ◭

Some types of matrices are given special names. A matrix which has the same number
of rows and columns is called a square matrix. Of the matrices in Example 5.1, only C
is square.
A matrix which has just a single row is usually called a row vector, and a matrix
with just one column is a column vector. The matrix B in Example 5.1 is a column
vector.

5.2 Matrix arithmetic

Before we can do any arithmetic we have to know what “equals” means: We say that the
matrices A = [aij ] and B = [bij ] are equal if and only if they have the same size, and
each element of A is equal to the corresponding element of B, that is

aij = bij , for each i and each j.

Thus two matrices are equal if and only if they are the same in every respect, that is, one
is a duplicate of the other.
In certain circumstances we can add, subtract, multiply and divide matrices. The rules
are similar in many respects to those for ordinary numbers, but there are some important
differences.

5.2.1 Addition and subtraction

If A and B have the same size, they can be added to form the sum A + B. The elements
of A + B are the sums of the corresponding elements in A and B. Thus A + B = [sij ],
where
sij = aij + bij , for each i and each j.
Similarly, if A and B have the same size, then the difference A − B is formed by
subtracting each element of B from the corresponding element of A. Thus A − B = [tij ],
where
tij = aij − bij , for each i and each j.

◮ Example 5.3 Of the matrices in Example 5.1, only A and D have the same size.
So it is possible to form the sum A + D, and the differences A − D and D − A.
We compute A + D and A − D below. The entries in D − A are the same as the
96 Chapter 5. Matrices

entries in A − D, except that the signs are reversed.

A D A D
2+2 4+6 4 10
B E B E
A + D = C3 − 1 −6 − 5F = C2 −11F
4+0 0+3 4 3
A D A D
2−2 4−6 0 −2
B E B E
A − D = C3 + 1 −6 + 5F = C4 −1F ◭
4−0 0−3 4 −3

5.2.2 Multiplication by scalars

If λ is any number and A is any matrix, then λA is the matrix obtained by multiplying
each element of A by λ. Thus λA = [λaij ]. Note that 2A = A + A, and more generally,
for any positive integer k,

kA = A + A + · · · + A (k terms).

If λ = −1 we obtain the negative of A, which we denote by −A. For any matrices A

and B of the same size, A − B = A + (−B). That is, subtracting any matrix is the same
as adding its negative.
Because addition, subtraction, and multiplication of matrices by scalars are done
element-by-element, they obey the same rules as ordinary numbers. These rules are
given in Table 5.1. In this table A, B and C are any matrices of the same size, and κ and
λ are any numbers.

A+B =B+A (commutative law)

A + (B + C) = (A + B) + C (associative law)
κ(A + B) = κA + κB (distributive law)
(κ + λ)A = κA + λA (distributive law)
(κλ)A = κ(λA) (associative law)

Table 5.1: Properties of matrix addition

5.2.3 Matrix multiplication

The product of two matrices needs to be defined carefully. We start with the product of
a row vector and a column vector. Suppose A is a row vector and B is a column vector
Section 5.2. Matrix arithmetic 97

with the same number of elements. Write

A D
b1
! " B E
B b2 E
A = a1 a2 . . . an and B = B E
B .. E .
C.F
bn

The product AB is the 1 × 1 matrix whose single element is obtained

! " by adding the
products of corresponding terms in A and B. To be precise, if AB = c , then

c = a1 b1 + a2 b2 + · · · + an bn .
! "
◮ Example 5.4 If A = 3 5 −1 4 , and B is the column vector given in Exam-
! " ! "
ple 5.1, then AB = 3 × 1 + 5 × 2 + (−1) × 5 + 4 × 3 = 20 . ◭

In order to understand how to form the product AB of more general matrices A and
B, it helps to think of A as a collection of row vectors and B as a collection of column
vectors. Suppose that A = [aij ], B = [bij ], A has size m × n and B has size p × q. The
product AB exists if and only if n = p, that is, the number of columns of A equals the
number of rows of B. If this is true, then AB is an m × q matrix, and AB = [cij ], where

cij = ai1 b1j + ai2 b2j + · · · + ain bnj .

According to this formula, the element cij in the ith row and the jth column of the product
AB is obtained by multiplying the ith row of A with the jth column of B. The condition
n = p ensures that these row and column vectors have the same number of elements. The
following diagram shows the location of the ith row of A, and the jth column of B. These
must be multiplied as row and column vectors to obtain cij .
A D A D
a11 a12 . . . a1n A D c11 ... ... . . . c1q
B .. .. .. .. E b11 . . . b1j . . . b1q B .. .. E
B . . . . EB E B . . E
B EB b21 . . . b2j . . . b2q E B E
B EB B .. .. E
B
B
ai1 ai2 . . . ain EB
EC .. .. .. .. .. E
E=B . cij . E
B .. .. .. .. E . . . . . F B
B .. ..
E
E
C . . . . F C . . F
bn1 . . . bnj . . . bnq
am1 am2 . . . amn cm1 . . . ... . . . cmq

It is important to realise that the order of the factors counts. If A and B are matrices,
it may be possible to form neither, just one, or both of the products AB and BA. To
emphasise the importance of the order of the factors, we say that in the product AB, A
premultiplies B, or that B postmultiplies A. When the product AB does exist we
say that A and B are conformable for multiplication (in that order).
98 Chapter 5. Matrices

◮ Example 5.5 Refer to the matrices in Example 5.1. The products which do exist
amongst these matrices are CC, which is 3 × 3, and CA and CD, which are both
3 × 2. All other combinations are not conformable. We compute CA below:
A DA D
3 5 −7 2 4
B EB E
CA = C4 2 0 F C3 −6F
0 −1 6 4 0
A D
3 × 2 + 5 × 3 + (−7) × 4 3 × 4 + 5 × (−6) + (−7) × 0
B E
= C 4×2+2×3+0×4 4 × 4 + 2 × (−6) + 0 × 0 F
0 × 2 + (−1) × 3 + 6 × 4 0 × 4 + (−1) × (−6) + 6 × 0
A D
6 + 15 − 28 12 − 30 + 0
B E
= C 8 + 6 + 0 16 − 12 + 0F
0 − 3 + 24 0+6+0
A D
−7 −18
B E
= C 14 4 F ◭
21 6

The rule for multiplying matrices may look peculiar at first, but the reason why matrix
multiplication is defined this way will become clearer in the next chapter.
As you know, multiplication of numbers is commutative, i.e., if A and B are numbers
then AB = BA. Matrix multiplication, however, does not satisfy this law, as the next
example shows.

◮ Example 5.6 Define the matrices A and B as follows:

G H G H
1 0 0 0
A := and B := .
0 0 2 1

Since each is square of the same size, both products exist:

G H G H
0 0 0 0
AB = and BA = .
0 0 2 0

Clearly, AB ∕= BA. ◭

If A and B are particular matrices for which AB = BA, then we say that A and
B commute. The above example shows that there is no commutative law for matrix
multiplication. However, there are some familiar laws that do apply, and these are shown
in Table 5.2 on the next page.
Section 5.2. Matrix arithmetic 99

A(BC) = (AB)C (associative law)

A(B + C) = AB + AC (distributive law)
(A + B)C = AC + BC (distributive law)
λ(AB) = (λA)B = A(λB)

Table 5.2: Properties of matrix multiplication

In the table, A, B, and C are any matrices conformable for multiplication in the orders
shown in the table and λ is any number.

5.2.4 Matrix operations in MATLAB

As we have already seen in Chapter 2, matrices (including vectors and scalars) are the
basic elements in the MATLAB computer package. One of the most important features
of MATLAB is that matrix computations can be performed easily and efficiently. In fact,
the name MATLAB stands for matrix laboratory.
Read Chapter 6 up to and including Section 6.2.4 in the textbook by Etter (2nd or 3rd
edition).

Exercises
G H G H
w 3−x 2 −1
1. If = , solve for w, x, y and z.
y2 3z 16 1
G H G H G H
2 −1 3 −2 1 3/2 3 0 11
2. Compute 7 −2 −3 .
−1 3 2 −5 1 1/2 1 −4 7
G H G H
1 −3 1 −1 5
3. Let A = and B = .
2 1 0 2 −3
Compute the products AB and BA, if they exist.
A D
G H −1 7 G H
8 −1 2 B E 2 1
4. Let A = , B = C 3 −2F, and C = .
2 0 −5 3 5
1 5
Verify that (AB)C = A(BC).

5. What can be said about the size of A if AA exists?

6. If A is a square matrix we write A2 to mean AA.

(a) Find counterexamples to the following assertions.

(i) (A + B)2 = A2 + 2AB + B 2 , for all n × n matrices A and B.
(ii) A2 − B 2 = (A + B)(A − B), for all n × n matrices A and B.
(b) What can be said about A and B if either of these equations is true?
100 Chapter 5. Matrices

7. With A, B and C as in Exercise 4, use MATLAB to find ABAB + 2C, if it exists.

G H G H G H G H
5 1 2 −1 3 2 −1 3
8. Let B = ,A= , A1 = and A2 = ,
6 2 1 4 2 1 4 2
where clearly A1 and A2 can be regarded as submatrices of A.

(a) Find BA and show that it is made up of the 2 × 2 matrix BA1 on the left and
the 2 × 1 matrix BA2 on the right.
(b) Verify the above result using MATLAB.
(c) Show that the result is true in general for any 2 × 2 matrix B and 2 × 3 matrix
A, which is partitioned into a 2 × 2 matrix A1 on the left and a 2 × 1 matrix
A2 on the right.

5.3 Some special matrices

The number 0 has the special property that a+0 = a for any number a. There are matrices
which have similar properties. A zero matrix is any matrix whose entries are all equal
to 0. If Z is a zero matrix and A is any matrix of the same size, then A + Z = A. Where
there is no confusion about size we shall simply write Z = 0. Example 5.6 demonstrates
another difference between multiplication of matrices and multiplication of numbers. In
that example we saw that A ∕= 0 and B ∕= 0 and yet AB = 0. This would not be possible
if A and B were numbers.
If T is a square matrix, then t11 , t22 , . . . , tnn are called its diagonal elements. If all
of the other elements of T are equal to 0, then we say that T is a diagonal matrix. A
diagonal matrix has the general form
A D
d1 0 0 . . . 0
B E
B 0 d2 0 . . . 0 E
B E
B 0 0 d3 . . . 0 E
B E
B .. .. .. . . .. E
C. . . . .F
0 0 0 . . . dn

where d1 , d2 , . . . , dn are the diagonal elements. To describe a diagonal matrix, we often use
a shorter form of notation in which only the diagonal elements are shown. The expression
diag(d1 , d2 , . . . , dn ) is another way of representing the diagonal matrix shown above.
A scalar matrix is a diagonal matrix whose diagonal elements are all equal. If
S = diag(λ, λ, . . . , λ) then S is a scalar matrix, and if A is any other matrix of the same
size, then SA = AS = λA. It is possible to show that the converse is also true: if S is a
square matrix which commutes with all other matrices of the same size, then S must be
a scalar matrix.
Section 5.3. Some special matrices 101

The number 1 has the special property that 1 · a = a · 1 = a for any number a.
There are matrices which have similar properties. Let In denote the n × n scalar matrix
diag(1, 1, . . . , 1). Then if A is any n × n matrix, In A = AIn = A. More generally, if A is
any m × n matrix, then
Im A = AIn = A.
We call In the identity (or unit) matrix of size n. Where there is no confusion about
size, we often ignore the subscript and write I for the identity matrix of the appropriate
size.
In MATLAB, some special matrices are defined as follows:

A = zeros(m,n) % m times n zero matrix

A = zeros(size(B)) % zero matrix A of same size as matrix B

A = ones(m,n) % m times n matrix with all entries equal to 1

A = ones(size(B)) % matrix of ones of same size as B

A = eye(n) % n times n identity matrix

A = diag(d) % diagonal matrix with vector d on diagonal

If B is a matrix, then v=diag(B) returns the vector v of diagonal elements of B.

◮ Example 5.7 Write a MATLAB program to define the n × n matrix A = [aij ],

where aij = i2 + j 2 and n is to be entered by the user. Also define the matrix B
with the same elements as A except with zeros on the diagonal.

% Define a special matrix of arbitrary size

clear
n=input(’Enter the size n of the square matrix: ’)
for i=1:n
for j=1:n
A(i,j) = i^2 + j^2;
end
end
A
% Define the corresponding matrix with zeros on diagonal
d=diag(A);
B = A - diag(d)

Enter and run this program with several different values of n. ◭

102 Chapter 5. Matrices

5.4 Further matrix operations

5.4.1 Inverses
We all know that for each non-zero number a there is a number b such that ab = 1, namely
b = a1 . This motivates the following definition:

Definition. If A is an m × n matrix and R is an n × m matrix such that AR = Im , then

we call R a right inverse of A. Similarly, if L is an n × m matrix such that LA = In ,
then we call L a left inverse of A.

Notice how we needed to be careful about our left and right because the commutative
law doesn’t hold for matrix multiplication.
G H
1 0 −1
◮ Example 5.8 Let A = . You can easily check that each of the fol-
1 1 0
lowing matrices are right inverses of A:
A D A D
0 1 1 0
B E B E
R1 = C 0 0F and R2 = C−1 1F .
−1 1 0 0

You can also check that A doesn’t have any left inverses by trying to solve the
equations arising from
A D A D
a b G H 1 0 0
B E 1 0 −1 B E
Cc dF = C0 1 0 F .
1 1 0
e f 0 0 1

You will find that there are no solutions. ◭

The above example shows that the story for matrix inverses is much more complicated
than it is for number inverses. We have a non-zero matrix that has more than one right
inverse, and yet no left inverses! Fortunately, the situation for square matrix inverses is
not so complicated, as the next two theorems show.

Theorem. Let A be a square matrix. If B is a left inverse of A, then it is also a right

inverse. Similarly, if B is a right inverse of A, then it is also a left inverse.

A proof of this theorem is beyond the scope of this course, so you will just have to
believe it for now. The theorem shows that for square matrices the concepts of left inverse
and right inverse are the same. So we don’t use these words for square matrices: If A
is a square matrix and B is a matrix such that AB = I or BA = I, then we call B an
inverse of A.
Section 5.4. Further matrix operations 103

Theorem. A square matrix either has no inverse or exactly one inverse.

Proof. Let A be a square matrix, and suppose that both B and C are inverses of A. To
prove the theorem we must show that B = C.

B = BI property of I
= B(AC) because C is an inverse of A
= (BA)C associative law
= IC because B is an inverse of A
= C property of I

This completes the proof. %

Since a square matrix has only one inverse (if it has one at all), we can give it a special
notation: we write the inverse of A as A−1 . Two simple rules of inversion are shown in
Table 5.3 below.

(A−1 )−1 = A
(AB)−1 = B −1 A−1

Table 5.3: Rules for inversion

Every number except 0 has an inverse, or reciprocal. The situation is not so straight-
forward for matrices. In many of the applications of matrix theory the problem of finding
inverses arises. For now, we will just learn the formula for 2 × 2 inverses.
G H
a b
Inverses of 2 × 2 matrices: Let A = . The determinant of A, written det(A),
c d
is defined by:
det(A) = ad − bc.
It can be shown that A has an inverse if and only if det(A) ∕= 0, in which case
G H
1 d −b
A−1 = .
det(A) −c a

You should memorise this formula.

◮ Example 5.9 Consider the two matrices

G H G H
1 2 1 2
A= and B = .
3 4 2 4
104 Chapter 5. Matrices

Clearly det(A) = 4 − 6 = −2 and det(B) = 4 − 4 = 0. So A has an inverse and B

does not. The inverse of A is
G H G H
1 4 −2 −2 1
A−1 = = ,
−2 −3 1 1.5 −0.5
which you can check by multiplying with A. (You will get I.) You can also check
that B doesn’t have an inverse by trying to find numbers p, q, r and s such that
G HG H G H
1 2 p q 1 0
= .
2 4 r s 0 1
You will find that there are no such numbers. ◭

The above example shows that there are square matrices which have no inverses. Such
matrices are called singular or non-invertible. A matrix which does have an inverse is
called non-singular or invertible.

5.4.2 Division
Dividing a number b by a non-zero number a is the same as multiplying b by a1 , that is,
a−1 . So we can think of dividing a matrix B by a non-singular matrix A as multiplying
B by the inverse A−1 . But remember that we cannot divide B by A unless A−1 exists.
Furthermore, we need to be careful about the order of multiplication. Premultiplying
and postmultiplying by A−1 do not necessarily give the same result. That is, BA−1 is
not necessarily equal to A−1 B. Because of these complications, we rarely speak about a
matrix operation called division. It is better to speak about pre-multiplication or post-
multiplication by an inverse.

5.4.3 Powers
The nth power of a square matrix A is obtained by taking the product of n copies of A.
That is,
An := A × A × · · · × A (n terms).
This makes sense when n is a positive integer. However we can extend the definition of
An to cover other integer values of n. First we define

A0 := I.

Then, provided A is non-singular, we define

A−n := (A−1 )n , for n = 1, 2, 3, . . .

With these definitions of powers the index law applies:

Ar × As = Ar+s , for any integers r and s.

Section 5.4. Further matrix operations 105

5.4.4 Transposes
Interchanging the rows and columns of an m × n matrix A gives an n × m matrix called
the transpose of A, which we denote by AT . We can think of AT as the reflection of A
across its main diagonal. Formally, if A = [aij ] and AT = [cij ], then aij = cji , for each i
and each j.
A D
G H 1 5
1 2 3 4 B 6E
B2 E
◮ Example 5.10 If A = , then AT = B E. ◭
5 6 7 8 C3 7F
4 8

Some rules for matrix transposition are shown in Table 5.4 (where λ is any number).

(AT )T = A
(A + B)T = AT + B T
(λA)T = λAT
(AB)T = B T AT
(AT )−1 = (A−1 )T

Table 5.4: Rules for transposition

If A = AT , then we say that A is symmetric. Clearly only square matrices can be

symmetric. Symmetric matrices occur frequently in applications, especially in the physical
sciences.

5.4.5 Matrix functions in MATLAB

Read Sections 6.2.5 and 6.2.6 in the textbook by Etter (2nd or 3rd edition).
An ill-conditioned matrix can be thought of as one that is nearly singular. Such
matrices pose a serious problem for numerical algorithms and packages like MATLAB
because the machine computations are done to a fixed (high) precision. Although an
ill-conditioned matrix is invertible in theory, the inverse cannot be computed accurately.
This is because for such matrices, the computer roundoff errors that arise in every step
are likely to be magnified substantially throughout the computational procedure. This
means that the computed inverse matrix will be an inaccurate approximation of the true
inverse matrix. A precise statement of this result and its rigorous justification are beyond
the scope of this unit. If the matrix A is mildly or moderately ill-conditioned, then the
MATLAB command inv(A) returns an inaccurate inverse matrix and possibly a warning
message that the matrix is close to singular. If the matrix A is severely ill-conditioned,
then inv(A) gives a warning message that the matrix is singular to working precision.
106 Chapter 5. Matrices

◮ Example 5.11 Consider the matrix

G H
−1 1
A= .
−1 1 + c

For any c ∕= 0, the matrix A is non-singular since det(A) = −(1 + c) + 1 = −c ∕= 0,

and the inverse is G H
−1 − 1/c 1/c
A−1 = .
−1/c 1/c
However, for c close to 0, the matrix is ill-conditioned. Write a MATLAB program
to print out A and the matrix A−1 defined above, and to compute the inverse of
A using inv(A) and print out the result. Allow the user to enter the value of c.

% Compute the inverse of an ill-conditioned matrix

clear
format long
c=input(’Enter the value of c: ’);
A=[-1 1; -1 1+c]
Exacinv=[-1-1/c 1/c; -1/c 1/c]
Compinv=inv(A)

Run this program with c equal to each of

0.5, 0.1, 10^-5, 10^-10, 10^-20 and 0.

Note that the inverse matrix computed by inv(A) becomes more and more inac-
curate as c decreases until for c = 10−20 , MATLAB prints the message

Warning: Matrix is singular to working precision.

Exercises
G H
0 1
9. If A = , show that A2 + I2 = 0.
−1 0
A D A D A D
0 0 −1 0 0 −1 0 0 0 0 0 −1
B0 0 0 −1EE B E B 1 0E
B B1 0 0 0E B0 0 E
10. Let A = B E, B = B E, and C = B E.
C1 0 0 0F C0 0 0 1F C0 −1 0 0F
0 1 0 0 0 0 −1 0 1 0 0 0

(a) Verify that A2 = B 2 = C 2 = ABC = −I4 .

(b) Use algebra to prove that AB = C.
(c) Use algebra to prove that −BA = C.
Section 5.4. Further matrix operations 107

G H G H
3 5 3 11
11. Suppose that A = and B = .
7 11 10 37

(a) Find A−1 and B −1 .

(b) Verify that (AB)−1 = B −1 A−1 .

12. Find a counterexample to the following assertion: “If A is singular and B is singular,
then A + B is singular.”
G H G H
2 1 0 3 −7
13. For the matrices A = and B = , verify that (AB)T = B T AT .
3 5 5 1 −2

14. Explain why the products AAT and ATA exist for any matrix A.

15. Prove that A commutes with B if and only if AT commutes with B T .

16. Find a counterexample to the following assertion: “If A is symmetric and B is

symmetric, then AB is symmetric.”

17. (a) Write a MATLAB program to define the n × n Hilbert matrix H = [hij ],
hij = 1/(i + j − 1), where n is entered by the user, and the diagonal matrix

D = diag(1, 3, . . . , 2n − 1).

(b) Extend the program to compute, for n = 5, the products DH and HD.
(c) Note that multiplying H by D on the left has the effect of multiplying each
row of H by the corresponding diagonal element of D, whereas multiplying
H by D on the right has the effect of multiplying each column of H by the
corresponding diagonal element of D. Show that this is true in general for DA
and AD, where A is an arbitrary square matrix and D is an arbitrary diagonal
matrix of the same size.

18. Write a MATLAB program using the inv command to compute the inverse of the
Hilbert matrix of size n defined above. Run the program with n equal to 5, 10 and
20. Is the Hilbert matrix ill-conditioned?
G H
1 2
19. (a) Let A = .
0 1/2
Write a MATLAB program to find the matrix powers A2 , A3 , . . ., An , where
n = 10. What do these matrices converge to as n → ∞?
A D
1 2 3 4
B0 1/2 1/3 1/4E
B E
(b) Repeat part (a) for the matrix A = B E.
C0 0 1/3 1/4F
0 0 0 1/4
108 Chapter 5. Matrices

5.5 Solutions to exercises

1. w = 2, x = 4, y = 4 and z = 0.
G H
9 −9 −15
2.
0 31 −8
G H
1 −7 14
3. AB = . BA does not exist.
2 0 7

5. A must be square.

6. (a) (i) Choose any pair of matrices which do not commute. For example, choose
A and B as in Example 5.6.
(ii) Same as (i).
(b) If either statement holds, then A and B commute.
G H
−391 −1358
7. ABAB + 2C = .
146 −345
G H
11 −1 17
8. BA = .
14 2 22

10. (b) AB = ABI = −AB(CC) = −(ABC)C = IC = C.

(c) −BA = −BAI = BA(ABC) = B(AA)BC = −BIBC = −BBC = IC = C.
G H G H
11 −5 37 −11
11. (a) A−1 = − 12 and B −1 = .
−7 3 −10 3
G H G H
1 0 0 0
12. A = and B = . Then det(A) = det(B) = 0 so A and B are singular,
0 0 0 1
but det(A + B) = 1, so A + B is non-singular.

14. If A is m × n, then AT is n × m, and hence A is conformable with AT (and AAT is

m × m) and AT is conformable with A (and ATA is n × n).

15. Proof.
AB = BA ⇔ (AB)T = (BA)T
⇔ B TAT = ATB T (see Table 5.4). %
G H G H G H
0 1 1 0 0 2
16. If A = and B = , then AB = . So A and B are symmetric
1 0 0 2 1 0
but AB is not.
Section 5.5. Solutions to exercises 109

17. % Define the Hilbert matrix H and a diagonal matrix D

% and compute the products DH and HD
clear
n=input(’Enter the size n of the matrix: ’)
for i=1:n
for j=1:n
H(i,j) = 1/(i+j-1);
end
end
H
D = diag(1:2:2*n-1)
DH = D*H
HD = H*D
Chapter 6

Linear Transformations of the Plane

6.1 Introduction
We are all familiar with the following diagram:
Y ✻

✛ ! ✲
(0, 0) X

Figure 6.1: The X–Y plane

The mathematical concept depicted in this diagram is usually called “the X–Y plane”,
“the Euclidean plane” or “real 2-space”, and is denoted by R2 . The notation reminds us
that the plane is really a set:

R2 = R × R = {(x, y) | x ∈ R, y ∈ R}.

Each point in the plane is an ordered pair (x, y), where x measures how far the point
is from the origin (0, 0) in the X, or horizontal, direction, and y measures how far the
point is from the origin in the Y , or vertical, direction. We have the convention that X is

110
Section 6.1. Introduction 111

positive to the right and negative to the left, and that Y is positive upwards and negative
downwards.

Definition. A linear transformation of the plane is a mapping

T : R2 → R2 , where T (x, y) = (ax + by, cx + dy),

for some fixed numbers a, b, c and d.

There is a strong link between linear transformations and matrices, which we now
develop. First, we represent points in the plane by matrices, as follows:
G H
x
The point (x, y) will be represented by the matrix .
y
Now let’s compute a certain matrix product:
G HG H G H
a b x ax + by
= .
c d y cx + dy
This shows that the linear transformation T can be thought of in terms of a matrix
multiplication. And it works the other way too: if we start with a matrix, we can define
a linear transformation based on it.

Definition. Let A be a 2 × 2 matrix of real numbers. The linear transformation

induced by A is the mapping
G H G H
x x
TA : R2 → R2 defined by TA ( ) := A .
y y

It is important to understand that a linear transformation TA is completely determined

by the images of the points (1, 0) and (0, 1). This follows since
G H G H
1 1
TA ( ) = A
0 0
G HG H
a b 1
=
c d 0
G H
a
=
c
which is simply the first column of A. Similarly,
G H G HG H
0 a b 0
TA ( ) =
1 c d 1
G H
b
=
d
112 Chapter 6. Linear Transformations of the Plane

which is the second column of A. This means that, given the matrix A, we can read off
the images of the points (1, 0) and (0, 1) from the columns of A. It also means that, if
we know the images of the points (1, 0) and (0, 1) under a linear transformation TA (for
instance if they are easily identified geometrically), then we can immediately write down
the matrix A. We will be using this fact in the next section.
G H
1 0
◮ Example 6.1 Let A = . Then
1 1
G H G H
x x
TA ( ) = A
y y
G HG H
1 0 x
=
1 1 y
G H
x
=
x+y

So TA is the mapping that sends (x, y) to (x, x+y) no matter what x and y are. For
example, (1, 2) is sent to (1, 1 + 2) = (1, 3). Note that some points aren’t shifted
at all by TA , for example, (0, 1) is sent to (0, 0 + 1) = (0, 1), namely itself! ◭

Like any mapping, TA is just shuffling everything around a bit. Some points stay where
they are, while others get moved to different points. However, unlike many possible maps
we could define from R2 to R2 , a matrix transformation is easy to visualise. That is, there
is a way of looking at TA that enables us to see exactly what kind of shuffling is going on.
This special way of looking at TA involves looking at what TA does to geometric figures.
The figure that we will be using mostly is depicted in Figure 6.2 on the next page. It is
a square in the first quadrant with one corner at the origin and of side length 1. We call
this a unit square and denote it by □1 . If we need to refer to the unit squares in the
other quadrants we will use the notation □i for the unit square in the ith quadrant. We
label the corners of □1 as O, P , Q and R, and we label their images under TA with a
dash, for example, we denote TA (P ) as P ′ .
Section 6.1. Introduction 113

Y ✻

2nd quadrant 1st quadrant

R(0, 1) ! ! Q(1, 1)

✛ ! ! ✲
O(0, 0) P (1, 0) X

3rd quadrant 4th quadrant

Figure 6.2: The unit square □1

◮ Example 6.2 Let’s see what happens to □1 under the mapping TA of Example 6.1:
(0, 0) is sent to (0, 0)
(1, 0) is sent to (1, 1)
(0, 1) is sent to (0, 1)
(1, 1) is sent to (1, 2)
So TA deforms □1 into the parallelogram depicted in Figure 6.3 below.
Y ✻

!
Q′
2
"
"
! !
"
R′ " P′
"
"

!"
"
✛ ✲
′ 1 X
O

Figure 6.3: TA (□1 )❄ (Example 6.2)

114 Chapter 6. Linear Transformations of the Plane

If you look at what TA does to □3 , you will begin to see the overall pattern:
Points with positive x-coordinate are dragged upwards, and points with negative
x-coordinate are dragged downwards, while points on the Y -axis are not shifted at
all. ◭

◮ Example 6.3 The following MATLAB program plots the four unit squares □1 , . . . , □4 ,
and their images under the linear transformation used in the examples above.

% Plot geometric effect of linear transformation of the plane

clear
clf
% Define x and y coordinates of four unit squares
sq1x=[0 0 1 1 0];
sq1y=[0 1 1 0 0];
sq2x=[0 0 -1 -1 0];
sq2y=[0 1 1 0 0];
sq3x=[0 0 -1 -1 0];
sq3y=[0 -1 -1 0 0];
sq4x=[0 0 1 1 0];
sq4y=[0 -1 -1 0 0];
% Plot the four squares
subplot(2,2,1), plot(sq2x,sq2y,’--’)
axis([-2,2,-2,2]), axis square
hold on
subplot(2,2,2), plot(sq1x,sq1y,’--’)
axis([-2,2,-2,2]), axis square
hold on
subplot(2,2,3), plot(sq3x,sq3y,’--’)
axis([-2,2,-2,2]), axis square
hold on
subplot(2,2,4), plot(sq4x,sq4y,’--’)
axis([-2,2,-2,2]), axis square
hold on
% Enter matrix A of linear transformation
A=[1 0; 1 1]
% Apply linear transformation to corners of squares
Asq1=A*[sq1x;sq1y];
Asq1x=Asq1(1,:);
Asq1y=Asq1(2,:);
Asq2=A*[sq2x;sq2y];
Section 6.1. Introduction 115

Asq2x=Asq2(1,:);
Asq2y=Asq2(2,:);
Asq3=A*[sq3x;sq3y];
Asq3x=Asq3(1,:);
Asq3y=Asq3(2,:);
Asq4=A*[sq4x;sq4y];
Asq4x=Asq4(1,:);
Asq4y=Asq4(2,:);
% Plot the images of the four squares
subplot(2,2,1), plot(Asq2x,Asq2y)
axis([-2,2,-2,2]), axis square
subplot(2,2,2), plot(Asq1x,Asq1y)
axis([-2,2,-2,2]), axis square
subplot(2,2,3), plot(Asq3x,Asq3y)
axis([-2,2,-2,2]), axis square
subplot(2,2,4), plot(Asq4x,Asq4y)
axis([-2,2,-2,2]), axis square

Read through the program and make sure you understand how it works. Run the
program and verify the description given above of the effect of the transforma-
tion. ◭

Every matrix transformation of R2 does something “geometric” to the plane like in

the last example. We can get a good idea of what happens to the whole plane by looking
at what happens to the unit squares. However, if we try to play with a mapping that is
not a matrix transformation, geometric figures can be destroyed and things are harder to
visualise.

◮ Example 6.4 Let’s have a look at the non-linear mapping f : R2 → R2 defined

by f (x, y) = (x2 , xy). Here’s what happens to □1 :

(0, 0) is sent to (0, 0)

(1, 0) is sent to (1, 0)
(1, 1) is sent to (1, 1)
(0, 1) is sent to (0, 0)
□1 has been changed into a triangle! Three corners of □1 didn’t move at all, but
the top left corner R(0, 1) has collapsed onto the bottom left corner O(0, 0). The
important observation here is that □1 was destroyed by f . It is no longer a four
sided figure. Linear transformations don’t do this kind of weird thing. (However,
as we will see later, there are some kinds of linear transformations that “squash”
□1 into a single line segment.) ◭
116 Chapter 6. Linear Transformations of the Plane

To finish this section, we introduce a little more notation: the symbol 6→ means
“is mapped to” or “is sent to” or “goes to”. So the statement “the function f , where
f (x, y) = (x + 1, y 2 )” can also be written as “the function f , where (x, y) 6→ (x + 1, y 2 )”.

Exercises
1. Find the images of the points on □1 under the linear transformations whose matrices
are given below and give geometrical descriptions of these transformations. Verify
your results by entering the matrices into the MATLAB program in Example 6.3.
G H
1 1
(a)
0 1
G H
2 0
(b)
0 2
G H
1 0
(c)
0 1

2. Show that every linear transformation maps the origin to itself.

3. Let A be a 2 × 2 matrix and let T be the linear transformation induced by A. How

is the image of the point (1, 1) related to the images of the points (1, 0) and (0, 1)?

4. (a) Find
G theHimage of □1 under the linear transformation whose matrix is A =
3 0
and find its area.
0 2
(b) Compare your result with det(A), the determinant of A.
G H G H G HG H
3 0 0 1 0 1 1 4
(c) Repeat parts (a) and (b) using , ,
0 −2 1 0 −1 0 0 1
G H
1 2
and in place of A.
2 4
(d) Use your discoveries to guess the general result.

6.2 Special types of transformations

One of the main applications of linear transformations of the plane is in computer graphics.
In a graphics display we need to be able to control the positions and movements of various
objects. To keep track of their positions we simply impose a coordinate system on the
screen. For example, the lower left corner of the screen might be designated (0, 0) and the
lower right corner (1024, 0), depending on how many pixels are on the screen. The way
Section 6.2. Special types of transformations 117

we control the movement of a point is to apply a linear transformation to it. Therefore

it is important to know how to accomplish basic movements such as stretching, shrinking
and rotation. These simple transformations are the main objects of study in this chapter.

6.2.1 Dilations
G H
α 0
A dilation is a linear transformation whose matrix is , where α and β are the
0 β
horizontal and vertical dilation factors, respectively. Note that both α and β must be
positive. Such a transformation has the effect of stretching or shrinking coordinates along
the two axes. Stretching occurs when the dilation factor is greater than 1, and shrinking
occurs when the dilation factor is less than 1. Nothing happens if the dilation factor
equals 1.
G H
1/3 0
◮ Example 6.5 Consider the transformation T whose matrix is . We
0 2
can tell from the defining matrix that the x-coordinates will be shrunk by one
third and the y-coordinates will be doubled. The effect of T on □1 is shown in
Figure 6.4 below.

Y ✻

2 ! !

✛ ! ! ✲
0 1 X
3

Figure 6.4: T (□1 ) (Example 6.5)

Enter the matrix of T into the MATLAB program given in Example 6.3 to see the
effect of T on all four unit squares. ◭
118 Chapter 6. Linear Transformations of the Plane

6.2.2 Shears
G H
1 α
A shear is a linear transformation whose matrix is (a horizontal shear) or
0 1
G H
1 0
(a vertical shear). The number α, which can be positive or negative, is called
α 1
the shear factor.
We have already seen a vertical shear in Example 6.2. This transformation had the
effect of dragging the 1st and 4th quadrants upwards and the 2nd and 3rd quadrants
downwards, while the Y -axis was left fixed. The shear factor determines by how much
these quadrants are dragged, with larger factors resulting in larger drags. The sign of the
shear factor determines the direction of the drag. A negative vertical shear means that
the 1st and 4th quadrants go down instead of up, and the 2nd and 3rd quadrants go up
instead of down.
In a positive horizontal shear the 1st and 2nd quadrants are dragged to the right, the
3rd and 4th quadrants are dragged to the left, and the X-axis is left fixed. In a negative
horizontal shear, these pairs of quadrants are dragged the other way.

6.2.3 Rotations
Among the most important linear transformations are those which cause points to rotate
around the origin. The angle of rotation measures how much rotation occurs. By
convention, we regard anticlockwise as the positive direction for rotation, and clockwise
as the negative direction. So a rotation of 30◦ means 30◦ anticlockwise. A clockwise
rotation of this amount would be called a rotation of −30◦ .
Theorem. The transformation which causes every point to rotate anticlockwise about
the origin through an angle of θ is the linear transformation whose matrix is
G H
cos θ − sin θ
.
sin θ cos θ
We won’t give a proof of this theorem. You will probably come to believe it once you
have seen some examples.
√
◮ Example 6.6 Let’s start with θ = 30◦ . Note that cos 30
G √
◦
= 3/2 and
H sin 30 =
◦

3/2 −1/2
1/2. So consider the transformation TA where A = √ . Let’s see
1/2 3/2
what TA does to □1 :
(0, 0) 6→ (0, 0)
√
(1, 0) 6 → ( 3/2, 1/2)
√ √
(1, 1) 6 → ( 12 ( 3 − 1), 12 (1 + 3))
√
(0, 1) 6 → (−1/2, 3/2)
Section 6.2. Special types of transformations 119

Looking at Figure 6.5 (below) we see that TA has caused □1 to rotate anticlockwise
by 30◦ .

Figure 6.5: TA (□1 ) (Example 6.6)

Enter the matrix A into the MATLAB program given in Example 6.3 to see
the effect of TA on all four unit squares. ◭

◮ Example 6.7 Now let’s try θ = 90◦ . In this case, it is easy to determine the matrix
of the transformation from the images of (1, 0) and (0, 1). Clearly, this rotation
mapsG(1, 0) to (0,
H 1) and maps (0, 1) to (−1, 0), so it is the transformation TA where
0 −1
A= . (Note this agrees with the result from the formula above since
1 0
cos 90◦ = 0 and sin 90◦ = 1.) Here’s the action of TA on □1 :

(0, 0) 6→ (0, 0)
(1, 0) 6 → (0, 1)
(1, 1) 6 → (−1, 1)
(0, 1) 6 → (−1, 0)

Looking at Figure 6.6 (on the next page) we see that TA (□1 ) occupies the same
place in the plane as □2 .
120 Chapter 6. Linear Transformations of the Plane

Y ✻

Q′ ! ! P′

✛ ! ! ✲
′ ′ X
R O

Figure 6.6: TA (□1 ) (Example 6.7)

We might well conclude from this that TA has rotated □1 by 90◦ , as expected.
However, □1 would also finish on top of □2 if TA were a reflection across the Y -axis
(to be defined in the next section). How can we tell which of these two transfor-
mations TA is? Or is it even another transformation? In ambiguous situations like
this, the way to determine the geometric effect of TA is to find the images of some
specific points, rather than a shape. For example, look at the corners of □1 . The
point P (1, 0) has been sent to (0, 1), and R(0, 1) has been sent to (−1, 0). This
certainly means that TA is not a reflection across the Y -axis, and should be enough
to convince you that TA is indeed a rotation of 90◦ . ◭

6.2.4 Reflections
The last special type of linear transformation that we will look at is the type that causes
points to be reflected across a line (which passes through the origin). The critical infor-
mation we need about the line is the angle θ it makes with the X-axis (measured from
the positive end of the X-axis travelling anticlockwise towards the line). If we start with
the equation y = mx of the line, we can compute θ using the formula tan θ = m. (And
the vertical line x = 0 yields the angle θ = 90◦ .)
Section 6.2. Special types of transformations 121

Theorem. The transformation which causes every point to reflect across the line hitting
the X-axis at an angle of θ is the linear transformation whose matrix is
G H
cos 2θ sin 2θ
.
sin 2θ − cos 2θ

Again, we won’t give a proof of this theorem.

◮ Example 6.8 Suppose that θ = 45◦ , so we are reflecting across the line y = x
(because tan 45◦ = 1). Clearly, the image of (1,G 0) is (0,H
1) and the image of (0, 1)
0 1
is (1, 0), so the matrix of this reflection is A = . (This also follows from
1 0
the general formula since cos 90◦ = 0 and sin 90◦ = 1.) It is clear from this matrix
that the point (x, y) is mapped to (y, x). If you now compute TA (□1 ) you will find
that it remains in the same place in the plane, but that the bottom right corner
P (1, 0) swaps positions with the top left corner R(0, 1) while the other two corners
stay fixed. Thus TA is indeed a reflection across the line y = x. Enter the matrix
A into the MATLAB program given in Example 6.3 to see the effect of TA on all
four unit squares. ◭
√
◮ Example 6.9 Now let’s try θ = 120◦ , so that we reflect across the line y = − 3x.
√
Note
G that cos 240◦ =H −1/2 and sin 240◦ = − 3/2, so the matrix we use is A =
√
−1/2 − 3/2
√ . Let’s see how TA affects □1 :
− 3/2 1/2

(0, 0) 6→ (0, 0)
√
(1, 0) 6 → (−1/2, − 3/2)
√ √
(1, 1) 6 → ( 12 (−1 − 3), 12 (− 3 + 1))
√
(0, 1) 6 → (− 3/2, 1/2)

Figure 6.7 (on the next page) shows that □1 has indeed been reflected across the
√
line y = − 3x. ◭
122 Chapter 6. Linear Transformations of the Plane

Figure 6.7: TA (□1 ) (Example 6.9)

Exercises
5. Write down the matrices which accomplish the following transformations. Check
your answers by entering the matrices into the MATLAB program given in Exam-
ple 6.3.

(a) A horizontal dilation of factor 3 and a vertical dilation of factor 1/2.

(b) A horizontal shear of factor 4.
(c) A vertical shear of factor −2.
(d) A rotation of −90◦ .
(e) A rotation of 180◦ .
(f ) A rotation of 45◦ .
(g) A reflection across the line y = −x.
(h) A reflection across the X-axis (i.e., the line y = 0).
(i) A reflection across the Y -axis (i.e., the line x = 0).
(j) A reflection across the line y = − √13 x.
Section 6.3. Degenerate transformations 123

6. Describe the effects of the linear transformations with the following matrices. Check
your answers by entering the matrices into the MATLAB program given in
Example 6.3.
G H
11 0
(a)
0 1/6
G H
1 −3
(b)
0 1
G H
1 0
(c)
5 1
G H
0 −1
(d)
1 0
G √ √ H
1/ 2 −1/ 2
(e) √ √
1/ 2 1/ 2
G H
−1 0
(f )
0 1
G H
0 1
(g)
1 0

6.3 Degenerate transformations

G H
1 2
Consider the transformation T with corresponding matrix A = . We have
2 4

(0, 0) 6→ (0, 0)
(1, 0) 6 → (1, 2)
(1, 1) 6 → (3, 6)
(0, 1) 6 → (2, 4)

If you don’t study the image points too carefully, everything looks normal. However, if
you plot these points you get what is shown in Figure 6.8 (on the next page), which is
not normal at all!
124 Chapter 6. Linear Transformations of the Plane

Y ✻

!
Q′
6
✁
✁
✁
✁
R′ ✁!
✁
4
✁
✁
✁
✁
2 !
✁
✁ P′
✁
✁
✁

!
✁
✛ ✁ ✲
O′ 1 2 3 X
❄

Figure 6.8: T (□1 )

What has happened here?

The square □1 has been squashed into a line segment! The segment is part of the line
y = 2x. If you check some more points you will find that T maps every point in the plane
to somewhere on the line y = 2x. You can also see this by entering the matrix A into the
MATLAB program given in Example 6.3.
T is an example of what we call a degenerate transformation. The definition of a
degenerate transformation is one which maps the plane R2 , which is 2-dimensional, to a
line, which is 1-dimensional, or even to a single point, which is 0-dimensional. (Though the
only matrix which maps the whole plane to a single point is the zero matrix.) Degenerate
transformations can be recognised geometrically because they squash each and every 2-
dimensional object to a line or line segment. So if □1 gets squashed, then you can conclude
immediately that the transformation is degenerate, that is, you don’t need to check any
other points—all points will be mapped to the line containing T (□1 ).
We have just learned a geometric way of recognising whether a transformation is
degenerate or not. However, there is also an arithmetic way to do this:

Theorem. A linear transformation is degenerate if and only if its matrix is singular, that
is, its determinant is zero.

This theorem gives us a method for recognising a degenerate transformation T very

quickly: compute the determinant of its matrix A; if det(A) = 0, then T is degenerate,
and if det(A) ∕= 0, then T is non-degenerate.
Section 6.4. Composition of transformations 125

Exercises
7. Which of the following matrices represent degenerate transformations?
G H
0 0
(a)
3 4
G H
0 1
(b)
2 3
G H
3 6
(c)
2 4
G H
43 5
(d)
215 25
G H
a b
8. Let be the matrix of a degenerate transformation. Show that if a ∕= 0,
c d
c
then the entire plane is mapped to the line y = x. Check this result by entering
a
some matrices of degenerate transformations into the MATLAB program given in
Example 6.3.

6.4 Composition of transformations

One linear transformation followed by another gives a third linear transformation. If we
know the matrices for the first two, how do we find the matrix for the third?
Let TA be the linear transformation with matrix A, and let TB be the linear trans-
formation with matrix B. The linear transformation “TA followed by TB ” is called the
composition or product of TA and TB and is written as TB ◦ TA or TB TA and defined
by TB TA (x, y) = TB (TA (x, y)). Note the order here, since TA is applied first to (x, y)
followed by TB . The order is important because “TB followed by TA ” might be a different
transformation to “TA followed by TB ”. We will see why this can happen in a moment.
First let’s find the matrix of TB TA :
G H G H
x x
(TB TA )( ) = TB (TA ( )) by definition of composition
y y
G H
x
= TB (A ) by definition of TA
y
G H
x
= B(A ) by definition of TB
y
G H
x
= (BA) by associative law for matrices.
y
126 Chapter 6. Linear Transformations of the Plane

This working yields the following nice result.

Theorem. The matrix of TB TA is BA.

This theorem also shows how TB TA and TA TB can be different linear transformations:
they will be the same if and only if A and B happen to commute.
G H
2 0
◮ Example 6.10 Let TS be the horizontal dilation with matrix S = , and
0 1
G H
0 −1
let TR be the 90◦ rotation with matrix R = . Is “stretch then rotate”
1 0
the same as “rotate then stretch”? The geometric way of finding out is to compare
TR TS (□1 ) with TS TR (□1 ). Instead, let’s do it the arithmetic way by comparing
RS and SR. You can check that these products are as follows:
G H G H
0 −1 0 −2
RS = and SR = .
2 0 1 0

So “stretch then rotate” is definitely different from “rotate then stretch”! ◭

◮ Example 6.11 Let TA be “rotation by 90◦ ” and let TB be “reflection across the
Y -axis”. Use matrix multiplication to find the geometric effect of “TA followed by
TB ”.
By drawing a diagram, it is clear that TA maps the basic point (1, 0) to (0, 1) and
it maps the other basic point (0, 1) to (−1, 0). Hence the corresponding G matrixHA
0 −1
has first column (0, 1)T and second column (−1, 0)T , and so it is A = .
1 0
Using another diagram, it is clear that TB maps the basic point (1, 0) to (−1, 0)
and it maps the other basic pointG (0, 1) H to (0, 1) (i.e. it is fixed). Hence the
−1 0
corresponding matrix B is B = . Now “TA followed by TB ” is TB TA ,
0 1
which has corresponding matrix
G HG H G H
−1 0 0 −1 0 1
BA = = .
0 1 1 0 1 0

From the first and second columns of BA, we know that TB TA maps (1, 0) to (0, 1)
and it maps (0, 1) to (1, 0). By drawing a diagram to represent this, it is clear that
TB TA must be “reflection across the line y = x”.

◮ Example 6.12 By composing two or more linear transformations of the plane, it is

possible to create interesting visual effects using computer graphics. The following
MATLAB program plots the transformed images of a well known curve under both
rotations and dilations combined.
Section 6.4. Composition of transformations 127

% Plot the images of a curve under rotations and dilations

clear
clf
% Define and plot the curve
t=0:2*pi/100:2*pi;
x=cos(t);
y=sin(3*t);
plot(x,y)
pause % press any key to continue
P=[x; y];
% Apply rotation and dilation transformations and plot images
for j=1:30
th(j)=2*pi*j/60;
Pnewj=th(j)*rotat(P,th(j));
% th(j) is angle of rotation, and horizontal and vertical dilation factor
plot(Pnewj(1,:),Pnewj(2,:))
axis([-4 4 -4 4])
axis square
pause(1) % pause one second
hold on
end

Here rotat.m is the function M-file containing:

function Pnew=rotat(P,theta)
% Rotate point(s) P (size 2 times n) by theta radians
% anticlockwise about origin
R=[cos(theta) -sin(theta); sin(theta) cos(theta)];
Pnew= R*P;

Note that the program and function M-file use radian measure for the angle of
rotation. This is the default measure for the argument of the sin and cos functions
in MATLAB (and other programming languages). Radian measure can be defined
as follows: a ray from the origin of angle θ with the positive X-axis defines a point
on the unit circle, and the radian measure of the angle is the signed (positive for
anticlockwise, negative for clockwise) length of the corresponding arc on the unit
circle from (1, 0). Therefore, 360◦ = 2π radians (circumference of the unit circle),
180◦ = π radians, 90◦ = π/2 radians, and in general d◦ = d × π/180 radians.
Enter the M-files above and run the program. What is the final dilation and
rotation of the basic curve x = cos t, y = sin 3t? ◭
128 Chapter 6. Linear Transformations of the Plane

Exercises
G H G H
1 a 1 b
9. Consider the horizontal shear matrices A = and B = . Arguing
0 1 0 1
geometrically, is there a difference between applying A then B, and applying B then
A? Check your answer by calculating AB and BA.

10. (a) Write down the matrices corresponding to clockwise and anticlockwise rota-
tions of 90◦ about the origin.
(b) What is the combined effect of one rotation followed by the other?
(c) Does the order of the rotations make a difference?
(d) Use matrix multiplication to check your answers.

11. (a) Write down the matrices corresponding to the transformations “rotate by 90◦ ”
and “reflect across Y -axis”.
(b) Use matrix multiplication to describe the products of these transformations,
taken in the two possible orders.
(c) Check your results by testing the effects of the products on □1 .
G H
a b
12. Show that a linear transformation whose matrix is of the form , where a
0 d
and d are positive, can be accomplished by a dilation followed by a shear.

13. Modify the MATLAB program in Example 6.12 to dilate and rotate the curve
defined by x = 2 + cos t, y = sin t, for 0 $ t $ 2π. Use steps of 2π/60 up to a final
dilation (in both x and y directions) of 2π and final rotation of 2π radians. Modify
the program again to rotate the curve in the opposite direction.

6.5 Solutions to exercises

1. (a) 1st and 2nd quadrants dragged to the right, 3rd and 4th quadrants dragged to
the left, X-axis left fixed.
(b) Everything expands by a factor of 2 both vertically and horizontally.
(c) Nothing happens! (This is called the identity transformation.)
G HG H G H
a b 0 0
2. = .
c d 0 0
G H G H G H G H
1 1 0 a+b
3. Since A =A +A = , the image of (1, 1) is the sum
1 0 1 c+d
of the image of (1, 0) and the image of (0, 1).
Section 6.5. Solutions to exercises 129

4. The area of TA (□1 ) equals the absolute value of det(A).

G H
3 0
5. (a)
0 1/2
G H
1 4
(b)
0 1
G H
1 0
(c)
−2 1
G H
0 1
(d)
−1 0
G H
−1 0
(e)
0 −1
G √ √ H
1/ 2 −1/ 2
(f ) √ √
1/ 2 1/ 2
G H
0 −1
(g) (θ = 135◦ )
−1 0
G H
1 0
(h) (θ = 0◦ )
0 −1
G H
−1 0
(i) (θ = 90◦ )
0 1
G √ H
1/2 − 3/2
(j) √ (θ = 150◦ )
− 3/2 −1/2

6. (a) A horizontal dilation of factor 11 and a vertical dilation of factor 1/6.

(b) A horizontal shear of factor −3.
(c) A vertical shear of factor 5.
(d) A rotation of 90◦ .
(e) A rotation of 45◦ .
(f ) A reflection across the Y -axis.
(g) A reflection across the line y = x.

7. (a), (c) and (d) are degenerate.

130 Chapter 6. Linear Transformations of the Plane

8. Since the transformation T is degenerate the determinant of its matrix is zero.

bc
Therefore ad − bc = 0, that is, d = (because a ∕= 0). Now
a
G HG H G H
a b x ax + by
=
c d y cx + dy
G H
ax + by
=
cx + bca y
G H
ax + by
= c
.
a
(ax + by)
c
So in T (R2 ) the Y -coordinate is always times the X-coordinate. Hence T (R2 ) is
a
c
the line y = x.
a
9. These transformations commute.
G H G H
0 1 0 −1
10. (a) Clockwise is and anticlockwise is .
−1 0 1 0
(b) Combined effect is “do nothing” (the identity transformation).
(c) No.
(d) These two matrices are inverses of each other so their product (in either order)
is the identity matrix.
G H G H
0 −1 −1 0
11. (a) R = is “rotate by 90◦ ” and M = is “reflect across
1 0 0 1
Y -axis”.
G H G H
0 1 0 −1
(b) M R = and RM = .
1 0 −1 0
(c) M R is “reflect across line y = x” and RM is “reflect across line y = −x”.
G H G HG H
a b 1 b/d a 0
12. First observe that = . Thus T can be accomplished
0 d 0 1 0 d
by first performing a horizontal dilation of (positive) factor a and a vertical dilation
of (positive) factor d and then performing a horizontal shear of factor b/d.
Chapter 7

Propositional Logic

7.1 Propositions
A proposition is a statement that can be classified as either true or false, but not both.
That is, it has a truth value (true or false) which is not ambiguous in any sense.

◮ Example 7.1 The following are propositions:

(a) Canberra is the capital of Australia.
(b) The Earth is flat.
(c) 2 + 3 = 5.
(d) If the Earth is flat, then 2 + 3 = 5.
Propositions (a), (c) and (d) are true, and (b) is false. Proposition (d) is a com-
pound proposition. It is a combination of the elementary or basic propositions
(b) and (c). ◭

Propositional logic (also called propositional calculus) is really a study of compound

propositions, and how truth values should be assigned to them. We shall see that a
compound proposition may be true even though some or all of its component parts are
false.

◮ Example 7.2 The following are not propositions:

(a) Is Canberra the capital of Australia?
(b) The Earth is beautiful.
(c) n + 3 = 5.
(d) Proposition (d) in the previous example is silly.
Question (a) is not considered to be a proposition, because for our purposes propo-
sitions must be statements. Statements (b) and (d) are not propositions, because
their truth values are not well defined unless we can all agree what “beautiful” and
“silly” mean. Statement (c) is not a proposition in the normal sense, because the
symbol n is not specified. ◭
131
132 Chapter 7. Propositional Logic

Statement (c) above is an example of an open proposition. In general an open

proposition over a set is a statement which contains one or more variables, and which
becomes an ordinary proposition when each of the variables is replaced by a specific
element of a set. This set is known as the index set or universal set.

◮ Example 7.3 The following are open propositions:

(a) n # 3.
(b) f is a continuous function.
(c) x + y = 5.
The variables in these open propositions are n, f , x, and y. The index sets have not
been specified, but would have to be chosen so that the open propositions become
ordinary propositions when the variables are replaced by specific elements. ◭

◮ Example 7.4 The following are open propositions over the set of positive integers:

(a) n is even.
(b) n is a power of 3.
(c) n2 − 8n + 12 = 0.
(d) 12 + 22 + · · · + n2 = 16 n(n + 1)(2n + 1). ◭
This chapter is devoted to the study of ordinary propositions.

7.2 The basic connectives

Propositions can be combined to form new ones, using logical operators known as connec-
tives. The propositions formed in this way are usually called compound propositions.
The truth value of a compound proposition depends on the truth values of the propositions
and the connectives from which it is formed. The three basic connectives are described
below.

Negation (¬). The simplest connective is negation. The negation of a proposition p is

“not p”, and is denoted by ¬p. If p is true then ¬p is false, and if p is false then ¬p is
true.

Conjunction (∧). The conjunction of two propositions p and q is the single proposition
“p and q”, and is denoted by p ∧ q. The truth value of p ∧ q reflects the normal meaning
of the word “and”; p ∧ q will be true when both p is true and q is true. In all other cases
p ∧ q will be false.

Disjunction (∨). The disjunction of two propositions p and q is the single proposition
“p or q”, and is denoted by p ∨ q. The truth value of p ∨ q reflects the meaning of the
inclusive or; p ∨ q will be true when p is true, or when q is true, or when both are true.
Section 7.2. The basic connectives 133

Note that in ordinary English, “or” has two meanings. Sometimes it means one or
other or both, and at other times it means one or other but not both. The former is known
as the inclusive or, and is the way “or” is meant to be interpreted in a sentence such
as: “The orchestra needs someone who can play the flute or the piccolo”. (Presumably
someone who can play both instruments would be acceptable.) In mathematics “or” is
usually taken to mean the inclusive or.
The other meaning of “or” is known as the exclusive or, and is the way “or” is meant
to be interpreted in a sentence such as: “That large, shining object in the sky is either
the sun or the moon”. (It couldn’t be both.) As we shall see later, a different symbol is
used to denote the exclusive or.
Recall from Section 2.6 that in MATLAB, the logical operator not is denoted by the
symbol: ~ Also recall that the logical operator and has the symbol: & and the logical
operator or has the symbol: |

◮ Example 7.5 Suppose that p and q are the following propositions:

p denotes “the moon is a balloon”,
and q denotes “today is my birthday”.
Then ¬p denotes “the moon is not a balloon”,
p∧q denotes “the moon is a balloon and today is my birthday”,
and p∨q denotes “the moon is a balloon or today is my birthday”. ◭

Exercises
1. Let i, w, e and r denote the following propositions.
i “Sarah is inside”
w “Sarah is watching TV”
e “Sarah is eating her dinner”
r “Sarah is riding her bicycle”

(a) Translate the following into English sentences.

(i) ¬i ∧ w
(ii) r ∧ (w ∨ ¬e)
(b) Translate the following into logical notation.
(i) Sarah is eating her dinner or she is riding her bicycle.
(ii) Sarah is neither inside nor is she riding her bicycle.
(iii) Sarah is inside, and she is eating her dinner while watching TV.
134 Chapter 7. Propositional Logic

7.3 Truth tables

The truth values of a compound proposition can be displayed in a table known as a
truth table. This gives the truth values of the compound proposition in terms of the
truth values of the basic propositions p1 , p2 , p3 , . . . from which it is composed. It helps
to regard p1 , p2 , p3 , . . . as variables and the compound proposition as a function of these
variables. The values of such functions and variables are the truth values, “true” and
“false”. The truth table of a compound proposition depending upon n variables has 2n
rows in it, one for each logical possibility, or combination of truth values for the variables.
Although “T” and “F” are commonly used to signify the truth values “true” and
“false”, we shall use “1” and “0” to denote “true” and “false”, respectively. This is the
notation used in MATLAB and it will also be more appropriate for our study of Boolean
algebra in Chapter 10. The truth tables of the propositions ¬p, p ∧ q and p ∨ q are given
in Table 7.1.

p q p∧q p q p∨q
p ¬p 0 0 0 0 0 0
0 1 0 1 0 0 1 1
1 0 1 0 0 1 0 1
1 1 1 1 1 1

Table 7.1: Truth tables for ¬p, p ∧ q, and p ∨ q

You should study the next example carefully.

◮ Example 7.6 Let p, q and r be propositions. Draw the truth table of the com-
pound proposition P := q ∧ ¬(¬p ∨ r).
Solution.
p q r q ∧ ¬ (¬ p ∨ r).
0 0 0 0 0 0 1 0 1 0
0 0 1 0 0 0 1 0 1 1
0 1 0 1 0 0 1 0 1 0
0 1 1 1 0 0 1 0 1 1
1 0 0 0 0 1 0 1 0 0
1 0 1 0 0 0 0 1 1 1
1 1 0 1 1 1 0 1 0 0
1 1 1 1 0 0 0 1 1 1
(1) ∗(7)∗ (6) (3) (2) (5) (4)
Explanation. We regard P as a function of the 3 variables p, q and r. Each of
the variables can take the values 0 and 1, and thus there are 23 = 8 possibilities.
Section 7.3. Truth tables 135

We list all these possibilities to the left of the vertical line in the table. Then we
fill in the columns to the right of the vertical line in the order indicated by the
numbers in parentheses in the last row:
(1) This is just the column of values of q copied from the left.
(2) This is the column of values of p copied from the left.
(3) This is ¬ of column (2), i.e., ¬p.
(4) This is r.
(5) This is (3) ∨ (4), i.e., ¬p ∨ r.
(6) This is ¬(5).
(7) This is (1) ∧ (6).
The final column (7) is the column of values of the proposition P . We write ∗(7)∗
to help highlight this fact. The column (7) shows that P is almost always a false
proposition. It is only true in the one case when p and q are true and r is false. ◭

This example shows how we can analyse a complicated compound proposition P by

drawing its truth table. However, this method is obviously inconvenient if there are many
basic propositions involved in P . If there are n such basic propositions, then the table
needs 2n rows. Thus the method is only practical by hand if n $ 4.
Using a computer, it is possible to construct much larger truth tables. The following
MATLAB program called truth.m constructs the truth table for one or more compound
propositions defined in terms of the basic variables (propositions) p(1), p(2), . . . . The
number num of basic variables, and the compound proposition(s) prop(1), prop(2), . . . ,
are defined in the function M-file called propos.m. Here the function M-file contains three
compound propositions related to Example 7.6:
prop(1) = ~p(1) | p(3); corresponding to ¬p ∨ r
prop(2) = ~(~p(1) | p(3)); corresponding to ¬(¬p ∨ r)
prop(3) = p(2) & ~(~p(1) | p(3)); corresponding to q ∧ ¬(¬p ∨ r)

% Program truth.m
% Construct the truth table for proposition(s) defined in propos.m
clear
format compact
global num % number of basic variables p(j)
% Find value of num defined in propos.m
propos(-1);
tot=2^num; % number of rows in truth table
% Print heading including a separator column
separt=888;
disp(’ Basic variables 888 Proposition value(s)’)
% Define truth values of basic variables p(j), j=1,...,num,
136 Chapter 7. Propositional Logic

% using binary representation of integers from 0 to tot-1

for i=1:tot
q=i-1;
for j=1:num
r=rem(q,2);
p(num+1-j)=r;
q=(q-r)/2;
end
% Compute truth values of proposition(s) defined in propos.m
prop=propos(p);
disp([p separt prop])
end

function prop=propos(p)
% Define proposition(s) for truth table constructed by truth.m
global num
if p<0
% Define number of basic variables p(j)
num=3;
else
% Enter the proposition(s)
prop(1) = ~p(1) | p(3);
prop(2) = ~(~p(1) | p(3));
prop(3) = p(2) & ~(~p(1) | p(3));
end

Enter and save the program and function M-files in MATLAB, and run the program
to obtain the following output.

Basic variables 888 Proposition value(s)

0 0 0 888 1 0 0
0 0 1 888 1 0 0
0 1 0 888 1 0 0
0 1 1 888 1 0 0
1 0 0 888 0 1 0
1 0 1 888 1 0 0
1 1 0 888 0 1 1
1 1 1 888 1 0 0

Note that the first three columns are the same as in the truth table given in Exam-
ple 7.6, and the last three columns are the same as those numbered (5), (6) and (7),
Section 7.4. More connectives 137

respectively. The column of “888” is used as a separator column.

Looking at the left hand side of the table, it is clear that each possible triple of truth
values for p, q and r, in the order given, can be thought of as the three binary digits in
the binary representation of the numbers 0, 1, . . . , 7. In fact, the program generates these
triples by computing the binary representations, using the algorithm given in Chapter 3,
i.e. by finding remainders after successive division by 2.
Note that instead of defining the propositions as above, it is easier (and better) to use:

prop(1) = ~p(1) | p(3);

prop(2) = ~prop(1);
prop(3) = p(2) & prop(2);

Exercises
2. Construct truth tables for the following compound propositions. Check your answers
using the MATLAB program truth.m together with a modified version of propos.m.

(a) ¬(p ∨ q)
(b) ¬(p ∧ q) ∨ r
(c) (p ∧ ¬q) ∨ q
! "
(d) ¬ (p ∧ q) ∨ r

7.4 More connectives

There are two other connectives for propositions which are important for our study of
logic.
Conditional (→). The conditional proposition, p → q, is defined to be true in all logical
possibilities, except the case when p is true and q is false. In this case p → q is false. The
truth table of p → q is shown in Table 7.2.

p q p→q
0 0 1
0 1 1
1 0 0
1 1 1

Table 7.2: Truth table of p → q

138 Chapter 7. Propositional Logic

There are several ways of expressing p → q in English:

“p implies q”,
“if p then q”,
“q if p”,
“p only if q”,
“p is sufficient for q”,
“q is necessary for p”.

The first two lines of the truth table for p → q may look strange, because they seem to
be saying that a false proposition implies anything. But really we are simply defining the
compound proposition p → q to be true if p is false. The following example may make
this seem more acceptable.

◮ Example 7.7 Consider the conditional proposition

“If I win the lottery, then I will give everybody a present.”

Surely I should not be regarded as a liar if I fail to give presents after not winning
the lottery. ◭

It is important to realise that the conditional proposition p → q does not by itself

imply the truth of either p or q. It merely says that if p is true, then q is true.
It is easy to check that p → q has the same truth table as ¬p ∨ q, and so from the
point of view of propositional logic, they have the same meaning. (We say that they are
logically equivalent, and we shall study logical equivalence in detail in Section 7.6).
So p → q is true if and only if p is false or q is true. If p is false we say that the
conditional p → q is vacuously true . This is because p → q true in this case, regardless
of the truth, or otherwise, of q. Similarly, we say that p → q is trivially true if q is true.
This is because the truth value of p is irrelevant in this case.

◮ Example 7.8 (a) The proposition “If 2 + 2 = 5, then the moon is a balloon” is
vacuously true, because “2 + 2 = 5” is false.
(b) The proposition “If 2 + 2 = 4, then Canberra is the capital of Australia” is
trivially true, because “Canberra is the capital of Australia” is true. ◭

The conditional proposition p → q is true in every logical possibility except the one
in which p is true and q is false. So in order to prove that p → q is true, it is enough
to prove that this possibility does not occur. That is, it is sufficient to prove that if p is
true, then q is true. Such a proof is called a direct proof.
Section 7.4. More connectives 139

◮ Example 7.9 Consider the following propositions:

p denotes “there are 23 401 words in this chapter, and

q denotes “there are more than 20 000 words in this chapter.

Then the truth of the conditional, p → q, could be established directly by

arguing as follows:
Suppose that p is true. Then there are 23 401 words in this chapter. Since
23 401 > 20 000, it follows that q is true.
Notice that by showing that p → q is true, we have not shown that either p
is true or that q is true. Presumably we would need to count the words in this
chapter in order to decide the truth values of p and q. ◭

It is important to realise that the conditionals p → q and q → p are not the same.
Table 7.3 shows that their truth tables are different. This table also shows two other
related conditionals ¬p → ¬q and ¬q → ¬p.

p q ¬p ¬q p → q q → p ¬p → ¬q ¬q → ¬p
0 0 1 1 1 1 1 1
0 1 1 0 1 0 0 1
1 0 0 1 0 1 1 0
1 1 0 0 1 1 1 1

Table 7.3: Four related conditionals

%
I
& the converse is q→p
For the conditional p → q, the inverse is ¬p → ¬q
I
'
the contrapositive is ¬q → ¬p
Notice that the truth tables of p → q and its contrapositive ¬q → ¬p are the same,
and the truth tables of the converse q → p and the inverse ¬p → ¬q are the same.

◮ Example 7.10 Suppose that p and q denote the following propositions:

p denotes “the fan is working”;
q denotes “there is a breeze”.
Then
p→q denotes “If the fan is working, then there is a breeze”.
¬q → ¬p denotes “If there is no breeze, then the fan is not working”.
q→p denotes “If there is a breeze, then the fan is working”.
¬p → ¬q denotes “If the fan is not working, then there is no breeze”.
Clearly the first two propositions have the same meaning, and the last two propo-
sitions have the same meaning. Note that if the fan is not working and there is
140 Chapter 7. Propositional Logic

a breeze, then the first two propositions would be true and the last two would be
false. ◭

The conditional connective can be represented in MATLAB by the relational operator:

<= (see Section 2.6). This is clear from Table 7.2, since when p and q are either 0 or 1,
the MATLAB statement p <= q and p → q have precisely the same truth values.

Biconditional (↔). The proposition p ↔ q is known as a biconditional, and is shorthand

for the compound proposition (p → q) ∧ (q → p). The truth table of p ↔ q is shown in
Table 7.4.

p q p↔q
0 0 1
0 1 0
1 0 0
1 1 1

Table 7.4: Truth table for p ↔ q

Observe that p ↔ q is true if both p and q are true or both p and q are false. That is,
p ↔ q is true if p and q have the same truth value, and false if they have different truth
values.
English language synonyms for p ↔ q are “p if and only if q” (this is often shortened
to “p iff q”), and “p is necessary and sufficient for q”. Compare these synonyms to those
given for p → q.
The biconditional connective is represented in MATLAB as the relational operator: ==
This is clear from Table 7.4 by considering the value of the MATLAB statement p == q
for each row.
There is one further logical connective, but it is not used very often:

Exclusive Or (⊕). The connective corresponding to the exclusive or is denoted by ⊕:

p ⊕ q is true when exactly one of p and q is true. The truth table of p ⊕ q is shown in
Table 7.5 below.
p q p⊕q
0 0 0
0 1 1
1 0 1
1 1 0

Table 7.5: Truth table for p ⊕ q

Section 7.4. More connectives 141

Notice that p ⊕ q and p ↔ q have opposite truth values: p ⊕ q is true if and only if
p ↔ q is false.
English language synonyms for p ⊕ q are “p or q, but not both”, and “exactly one of
p and q”.
The exclusive or connective is represented in MATLAB as the relational operator: ~=
This is clear from Table 7.5; p ⊕ q is true precisely when p and q have different truth
values.
The following table summarizes the connectives that have been introduced.

Connective Logical symbol MATLAB operator

Negation (not) ¬ ~
Conjunction (and) ∧ &
Disjunction (or) ∨ |
Conditional (implies) → <=
Biconditional (iff) ↔ ==
Exclusive Or ⊕ ~=

Exercises
3. Let i, w, e and r denote the propositions in Exercise 1.

(a) Translate the following into English sentences.

(i) r → ¬i
(ii) i → (w ∨ e)
(iii) (i ∨ r) ∧ (i → w)
! "
(iv) ¬r → i ∧ (w ∨ e)
(b) Translate the following into logical notation.
(i) If Sarah is not watching TV, then she is outside.
(ii) If Sarah is inside and eating her dinner, then she is not riding her bicycle.

4. Construct truth tables for the following compound propositions. Check your answers
using the MATLAB program truth.m together with a modified version of propos.m.

(a) ¬p → (p ∨ q)
(b) (p ↔ q) ↔ (¬p ↔ ¬q)
(c) (p ⊕ q) ⊕ r
(d) p ⊕ (q ⊕ r)
(e) (p ⊕ p) ⊕ p
142 Chapter 7. Propositional Logic

5. Determine the truth values of the following compound propositions.

(a) If 5 + 5 = 55, then 5 × 5 = 55.

(b) If 5 + 5 = 10, then 5 × 5 = 25.
(c) If 5 + 5 = 55, then 5 × 5 = 25.
(d) If 5 + 5 = 10, then 5 × 5 = 55.

6. Give the converses, inverses and contrapositives of the following compound propo-
sitions. Notice how the contrapositive has the same meaning as the original condi-
tional, and how the inverse has the same meaning as the converse.

(a) If Sarah is watching TV, then she is inside.

(b) I will pass the course if I work hard.
(c) If xy = 0, then x = 0.
(d) If A = B ∩ C, then A ⊆ C.

7.5 Tautologies and contradictions

We say a compound proposition is a tautology if it is true for all possible truth values
of the basic propositions from which it is composed. We recognise tautologies from their
truth tables: each value in the relevant column of the table is “1”. Tautologies are very
important for our understanding and use of logical reasoning.

◮ Example 7.11 The proposition (p → q) ∨ ¬q is a tautology, as can be seen from

its truth table:
p q (p → q) ∨ ¬ q
0 0 0 1 0 1 1 0
0 1 0 1 1 1 0 1
1 0 1 0 0 1 1 0
1 1 1 1 1 1 0 1
(1) (3) (2) ∗(6)∗ (5) (4)
Each value in column (6) is a “1”, which proves that the proposition is a tautol-
ogy. ◭

A compound proposition is a contradiction if it is false for all possible combinations

of truth values of its variables. Each value in the relevant column of the truth table of
a contradiction is a “0”, and a compound proposition P is a contradiction if and only if
¬P is a tautology.
Section 7.6. Logical equivalence 143

◮ Example 7.12 The proposition (p ∧ ¬q) ∧ (p → q) is a contradiction, as can be

seen from its truth table:

p q (p ∧ ¬ q) ∧ (p → q)
0 0 0 0 1 0 0 0 1 0
0 1 0 0 0 1 0 0 1 1
1 0 1 1 1 0 0 1 0 0
1 1 1 0 0 1 0 1 1 1
(1) (4) (3) (2) ∗(8)∗ (5) (7) (6)

Each value in column (8) is a “0”, which proves that the proposition is a contra-
diction. ◭

Exercises
7. Show that the following propositions are tautologies:

(a) p ∨ ¬p
(b) p → p
(c) (p → q) ↔ (¬q → ¬p)

8. Show that the following propositions are contradictions:

(a) p ∧ ¬p
(b) (p ∨ q) ∧ (¬p ∧ ¬q)
(c) (p ⊕ q) ∧ (p ↔ q)
(d) p ⊕ p

9. Show that for any two propositions p and q, (p → q) ∨ (q → p) is a tautology.

Comment: This illustrates the danger in assuming that any causal relationship is
implied by the conditional proposition p → q. Even though we say p “implies” q,
we do not mean that p “causes” q. Propositions p and q may be totally unrelated,
but because (p → q) ∨ (q → p) is a tautology, either p → q is true or q → p is true.

7.6 Logical equivalence

Two compound propositions P and Q are said to be logically equivalent if they have
the same truth values for all values of the variables from which they are composed. When
this happens we write P ⇔ Q. Since P ↔ Q is true precisely when the truth values of P
and Q agree, logical equivalence and the biconditional are related as follows:
144 Chapter 7. Propositional Logic

Logical equivalence and the biconditional:

P ⇔ Q means P ↔ Q is a tautology.

It is important to understand the difference between P ⇔ Q and P ↔ Q. Informally,

⇔ is like an equals sign for propositions, whereas ↔ is just a connective. Formally, the
expression P ⇔ Q is really a proposition about a proposition, namely that “P ↔ Q is a
tautology”, whereas the expression P ↔ Q is just a compound proposition that may or
may not be a tautology.
In order to show that a proposition is true, it is sufficient to show that a logically
equivalent one is true. This fact is often used to establish the truth of a conditional
proposition. Since (p → q) ↔ (¬q → ¬p) is a tautology, p → q and ¬q → ¬p are logically
equivalent. So if ¬q → ¬p can be shown to be true, then p → q is also true. This method
of proving a conditional to be true is called proof of the contrapositive.

◮ Example 7.13 Suppose that p and q are defined as follows:

√ √
p denotes “ 2 is irrational”, and q denotes “ 8 is irrational”.

To prove that p → q is true it is easier to use the contrapositive. So suppose

√ √
that ¬q is true. Then 8 is rational. So write 8 = m/n, where m and n are
√ √ √
positive integers. Then 2 = 8/2 = m/2n, and so 2 is rational, that is, ¬p is
true. Therefore ¬q → ¬p is true, and it follows that p → q is also true. ◭

Proofs of the contrapositive are preferred to direct proofs if ¬q is easier to deal with
√ √
than p. In the above example ¬q denotes “ 8 is rational”, whereas p denotes “ 2 is
irrational”. Since irrational numbers are defined only negatively, that is, as numbers that
√
are not rational, it is easier to start the proof with the assumption that 8 is rational,
√
rather than the assumption that 2 is irrational.
Some logical equivalences are listed in Table 7.6 below. In this table T stands for any
tautology and F stands for any contradiction. In each case we can show that P ⇔ Q by
showing (with a truth table) that P ↔ Q is a tautology. If the truth table for P ↔ Q
has a 0 in the final row, then P ⇔ Q is not valid, and the values of the basic propositions
corresponding to this row make up a counterexample.
Section 7.6. Logical equivalence 145

For any propositions p, q, and r,

1. (a) p ∨ q ⇔ q ∨ p commutativity
(b) p ∧ q ⇔ q ∧ p
2. (a) (p ∨ q) ∨ r ⇔ p ∨ (q ∨ r) associativity
(b) (p ∧ q) ∧ r ⇔ p ∧ (q ∧ r)
3. (a) p ∨ (q ∧ r) ⇔ (p ∨ q) ∧ (p ∨ r) distributivity
(b) p ∧ (q ∨ r) ⇔ (p ∧ q) ∨ (p ∧ r)
4. (a) p ∨ p ⇔ p idempotence
(b) p ∧ p ⇔ p
5. (a) p ∨ F ⇔ p identity
(b) p ∧ T ⇔ p
6. (a) p ∨ T ⇔ T nullity
(b) p ∧ F ⇔ F
7. (a) p ∨ ¬p ⇔ T excluded middle
(b) p ∧ ¬p ⇔ F
8. ¬¬p ⇔ p double negation
9. (a) ¬T ⇔ F
(b) ¬F ⇔ T
10. (a) ¬(p ∨ q) ⇔ ¬p ∧ ¬q De Morgan’s laws
(b) ¬(p ∧ q) ⇔ ¬p ∨ ¬q

Table 7.6: Logical equivalences

We can use Table 7.6 to simplify, or replace, long and complicated compound propo-
sitions by shorter, logically equivalent ones. These are usually easier to understand.

◮ Example 7.14 Show that p ∨ (p ∧ q) ⇔ p.

Solution.
! "One way of proving this equivalence is to compute the truth table of
p ∨ (p ∧ q) ↔ p and show that it is a tautology. This can be done either by hand
or with a computer. However it can also be done algebraically using our table of
logical equivalences. Here is an algebraic proof:

p ∨ (p ∧ q) ⇔ (p ∧ T) ∨ (p ∧ q) identity 5(b)
⇔ p ∧ (T ∨ q) distributivity 3(b)
⇔ p ∧ (q ∨ T) commutativity 1(a)
⇔ p∧T nullity 6(a)
⇔ p identity 5(b) ◭

There are other logical equivalences which involve the conditional and biconditional
connectives. These are shown in Table 7.7 below.
146 Chapter 7. Propositional Logic

For any propositions p, q, r and s,

11. p → q ⇔ ¬q → ¬p contrapositive
12. p → q ⇔ ¬p ∨ q implication
13. p → F ⇔ ¬p reductio ad absurdum
14. (p → r) ∧ (q → r) ⇔ (p ∨ q) → r
15. (p → q) ∧ (p → r) ⇔ p → (q ∧ r)
16. (p ∧ q) → r ⇔ p → (q → r) exportation
17. p ↔ q ⇔ (p → q) ∧ (q → p)

Table 7.7: More logical equivalences

◮ Example 7.15 Show that (p → q) ∧ (p → ¬q) ⇔ ¬p.

Proof. An algebraic proof goes like this :

(p → q) ∧ (p → ¬q) ⇔ p → (q ∧ ¬q) equivalence 15

⇔ p→F excluded middle 7(b)
⇔ ¬p reductio ad absurdum 13

Once you have become familiar with a few of these basic logical equivalences, proofs
of this type are very easy. However, it is important to remember that proofs via
truth tables
! are just as valid." So in this case it is enough to compute the truth
table of (p → q) ∧ (p → ¬q) ↔ ¬p and show that it is a tautology. ◭

Exercises
Many students are uncomfortable with the algebra of propositions and prefer to use
truth tables when proving logical equivalences. However, sometimes it is actually
quicker to work algebraically, especially if there is no computer handy and you have
to do all the calculations yourself. For example, an equivalence involving 4 basic
propositions needs 24 = 16 rows (and probably many columns) in its truth table,
whereas an algebraic proof might only take 6 short lines. So don’t be scared of the
algebra of propositions. It is not that hard, and the more you practice, the better
you will become. Also, you will find the following sections less difficult if you can
master the algebra of propositions.

10. Prove the following equivalences:

(a) p ∧ q ⇔ ¬(¬p ∨ ¬q)

(b) p ∨ q ⇔ ¬(¬p ∧ ¬q)
Section 7.7. Logical implication 147

11. Prove the following equivalences:

(a) p → T ⇔ T
(b) T → p ⇔ p
(c) p → F ⇔ ¬p
(d) F → p ⇔ T

12. Use truth tables to discover whether the following “equivalences” are valid or not:

(a) p → (q → r) ⇔ (p → q) → r.
(b) p ↔ (q ↔ r) ⇔ (p ↔ q) ↔ r.
(c) (p ∨ r) ∧ (q → r) ⇔ (p → q) → r.

13. Show that p ↔ q ⇔ (p ∧ q) ∨ (¬p ∧ ¬q).

7.7 Logical implication

◮ Example 7.16 “James is a policeman or a footballer. If he is a policeman, then
he has big feet. James doesn’t have big feet. Therefore he is a footballer.”
Does this argument make sense? That is, is it right to conclude from the given infor-
mation that James is a footballer? Or is there some trick that we have missed? ◭

This section is devoted to the study of arguments just like the one in the above example.
We will learn how to analyse such arguments and determine whether they are correct or
not. Propositional calculus is perfect for this study because an argument is really just
a sequence of propositions. All but the last proposition are considered to be true (for
whatever reason), and the last proposition is claimed to be also true, by deduction.
We say that a proposition P logically implies another Q, and we write P =⇒ Q,
if Q is true whenever P is true. So just like logical equivalence, we can express logical
implication in terms of a tautology.

Logical implication and the conditional:

P =⇒ Q means P → Q is a tautology.

Logical inference is a method by which we can derive new assertions or propositions

from old ones. It usually takes the form of a logical implication

(H1 ∧ H2 ∧ · · · ∧ Hn ) =⇒ C.
148 Chapter 7. Propositional Logic

This is traditionally written in the form:

H1
H2
..
.
Hn

∴ C

The symbol ∴ denotes “therefore”, or “it logically follows that”. Each of the propositions
H1 , H2 , . . . , Hn is called an hypothesis or premise, and the proposition C is called the
conclusion.
The inference is said to be valid if the conditional proposition

(H1 ∧ H2 ∧ · · · ∧ Hn ) → C

is a tautology, in which case we say that C is a logical consequence of the hypotheses

H1 , H2 , . . . , Hn . Otherwise the inference is said to be invalid or is called a fallacy.
We use logical inference not just in mathematics, but in any discipline in which logical
reasoning is used. However, it is not always used correctly. We shall use the words
“inference” and “argument” to denote a set of hypotheses, together with a suggested
conclusion, regardless of whether the conclusion is a logical consequence of the hypotheses.

◮ Example 7.17 Let us now analyse the argument in Example 7.16. The argument
can be written as:
James is a policeman or James is a footballer.
If James is a policeman, then James has big feet.
James doesn’t have big feet.

Therefore James is a footballer.

These propositions are clumsy to work with and so we assign symbols to them as
follows:
p denotes “James is a policeman”;
f denotes “James is a footballer”;
b denotes “James has big feet”.
We can now write the argument in symbolic form:

p∨f
p→b
¬b

∴ f
Section 7.7. Logical implication 149

To analyse the argument, we compute the truth table of the proposition

! "
(p ∨ f ) ∧ (p → b) ∧ ¬b → f.

(We won’t print the truth table here, but you may care to compute it yourself,
either by hand or using a computer.) From the truth table we find that this last
proposition is a tautology. Hence the argument is valid. That is, we can conclude
that James is a footballer from the information given. ◭

You should note that the inference in the previous example will always be valid, no
matter what p, f and b stand for. To analyse the argument, we threw away the context and
just dealt with the underlying logical structure. It is important in any field of knowledge
to be able to separate the underlying logical structure of an argument from the actual
context, because often the context makes the analysis harder rather than easier. (Just
think of the tricks that politicians use to sway their audiences!) Here are some examples
of inference that are context-free.

◮ Example 7.18 Consider the following inference.

p∨q
p→r

∴ q∨r

To decide whether or not this inference is valid, we examine the truth table of the
corresponding conditional:
! "
p q r (p ∨ q) ∧ (p → r) → (q ∨ r)
0 0 0 0 0 0 0 0 1 0 1 0 0 0
0 0 1 0 0 0 0 0 1 1 1 0 1 1
0 1 0 0 1 1 1 0 1 0 1 1 1 0
0 1 1 0 1 1 1 0 1 1 1 1 1 1
1 0 0 1 1 0 0 1 0 0 1 0 0 0
1 0 1 1 1 0 1 1 1 1 1 0 1 1
1 1 0 1 1 1 0 1 0 0 1 1 1 0
1 1 1 1 1 1 1 1 1 1 1 1 1 1
(1) (3) (2) (7) (4) (6) (5) ∗(11)∗ (8) (10) (9)

Column (3) is the truth column for the first hypothesis H1 := p ∨ q.

Column (6) is the truth column for the second hypothesis H2 := p → r.
Column (7) is the truth column for H1 ∧ H2 .
Column (10) is the truth column for the conclusion C := q ∨ r.
150 Chapter 7. Propositional Logic

Column (11) is the truth column for the inference (H1 ∧ H2 ) → C.

Since each row of column (11) is a 1, the conditional is a tautology and the inference
is valid. So q ∨ r is a valid conclusion from the hypotheses p ∨ q and p → r. ◭

◮ Example 7.19 Consider the following inference.

p → (q ∨ r)
¬q

∴ r

To decide whether or not this inference is valid, we examine the truth table of the
corresponding conditional:

((p → (q ∨ r)) ∧ ¬q) → r

This time we use the MATLAB program truth.m given in Section 7.3 to construct
the truth table. The appropriate function M-file propos.m is given below, together
with the output of the program.

function prop=propos(p)
% Define proposition(s) for truth table constructed by truth.m
global num
if p<0
% Define number of basic variables p(j)
num=3;
else
% Enter the proposition(s)
prop(1) = p(1) <= (p(2) | p(3));
prop(2) = ~p(2);
prop(3) = prop(1) & prop(2);
prop(4) = p(3);
prop(5) = prop(3) <= prop(4);
end

Basic variables 888 Proposition value(s)

0 0 0 888 1 1 1 0 0
0 0 1 888 1 1 1 1 1
0 1 0 888 1 0 0 0 1
0 1 1 888 1 0 0 1 1
1 0 0 888 0 1 0 0 1
1 0 1 888 1 1 1 1 1
1 1 0 888 1 0 0 0 1
1 1 1 888 1 0 0 1 1
Section 7.7. Logical implication 151

The first column on the right is the truth column for the first hypothesis H1 := p → (q∨r).
The second column on the right is the truth column for the second hypothesis H2 := ¬q.
The third column on the right is the truth column for H1 ∧ H2 .
The fourth column on the right is the truth column for the conclusion C := r.
The final column is the truth column for the inference (H1 ∧ H2 ) → C.
Since a 0 appears in the final column, the logical implication is not a tautology, and the
inference is invalid. So r is not a logical conclusion from the hypotheses p → (q ∨ r)
and ¬q. ◭

◮ Example 7.20 Consider the following famous example of inference.

If Socrates is a man, then Socrates is mortal.
Socrates is mortal.

Therefore Socrates is a man.

In symbolic form the inference is

p→q
q

∴ p
! "
This is invalid because (p → q)∧q → p is not a tautology, as can be demonstrated
by its truth table. (Look at the row in which p = 0, q = 1.) Yet we know that
the conclusion is true, so what has gone wrong?! What has happened is that we’ve
used outside knowledge to deduce that the conclusion is true. Look again at the
information given: Socrates could be somebody’s pet lizard for all we know! It is
not valid to conclude from the two hypotheses that Socrates is a man. The moral
of this example is: The fact that the conclusion is true does not make the inference
valid. The conclusion is true because of the meanings of the words “Socrates” and
“man”, not by inference. ◭

Exercises
14. Analyse the following arguments for logical validity.

(a) Today is either Monday or Tuesday.

But today is not Monday.
Therefore today is Tuesday.
(b) If today is Saturday, then tomorrow is Sunday.
Today is not Saturday.
Therefore tomorrow is not Sunday.
152 Chapter 7. Propositional Logic

7.8 Rules of inference

Some valid inferences tend to arise very frequently, and it is convenient to use them over
and over again instead of having to draw up a truth table each time we use them. A list
of logical implications appears in Table 7.8 below.

1. p =⇒ p ∨ q addition
2. p ∧ q =⇒ p simplification
3. p ∧ (p → q) =⇒ q modus ponens
4. ¬q ∧ (p → q) =⇒ ¬p modus tollens
5. (p ∨ q) ∧ ¬p =⇒ q disjunctive syllogism
6. (p → q) ∧ (q → r) =⇒ p → r hypothetical syllogism

Table 7.8: Logical implications

In each case we can show that P =⇒ Q by showing that P → Q is a tautology. To do

this it is not necessary to examine the entire truth table. Since P → Q is automatically
true whenever P is false, it is sufficient to show that Q is true whenever P is true.
You should not be intimidated by the complicated sounding names in the table. The
fact that some of them are in Latin, and others are derived from Greek, merely indicates
that people have been studying the rules of logic for many centuries. The rules themselves
become straightforward and “obvious”, when the symbols are replaced by simple verbal
propositions.

◮ Example 7.21 Consider the following famous example of inference.

If Socrates is a man, then Socrates is mortal.

Socrates is a man.

Therefore Socrates is mortal.

This inference is valid because the logical implication is an example of modus po-
nens (sometimes called the rule of detachment). To see this let p denote “Socrates
is a man”, and let q denote “Socrates is mortal”. With these symbols, the inference
is formulated as
p→q
p

∴ q
This is valid because, by Table 7.8(3), p∧(p → q) =⇒ q is a logical implication. ◭
Section 7.8. Rules of inference 153

◮ Example 7.22 The following inferences can be shown to be valid by using a form
of logical implication known as modus tollens (Table 7.8, rule 4).
(a) If the rains come, then the crops will grow.
The crops did not grow.
Therefore the rains did not come.
(b) The program will stop if there is no infinite loop.
The program did not stop.
Therefore the program has an infinite loop.
(c) The dog will bark if an intruder comes.
The dog did not bark.
Therefore no intruder came.
Each of these inferences is a logical implication of the form

(p → q) ∧ ¬q =⇒ ¬p. ◭

Now here are some more complicated examples of valid inferences. Don’t forget that
when proving them we are allowed to use the logical equivalences in Tables 7.6 and 7.7
to simplify intermediate propositions.

◮ Example 7.23 Write the following argument in symbolic form and prove that it
is logically valid:
“If the batteries are flat or if the connections are loose, then the flash doesn’t work.
If the flash doesn’t work, then the photo is too dark. The photo is not too dark.
Therefore the connections are not loose.”
Denote propositions as follows (using letters suggestive of the statement):
b: “The batteries are flat”;
c: “The connections are loose”;
f: “The flash doesn’t work”;
p: “The photo is too dark”.
The argument in symbolic form is:

(b ∨ c) → f
f →p
¬p

∴ ¬c
154 Chapter 7. Propositional Logic

Proof.
(f → p) ∧ ¬p H2 ∧ H3
! =⇒" ¬f . . . (1) modus tollens
(b ∨ c) → f ∧ ¬f H1 ∧ (1)
=⇒ ¬(b ∨ c) modus tollens
⇔ ¬b ∧ ¬c De Morgan
=⇒ ¬c C simplification
Explanation. We are given three hypotheses to work with. We may use any or
all of these as many times as we wish, together with the logical equivalences in
Table 7.6 and 7.7, and the rules of logical implication in Table 7.8, in order to
arrive at the conclusion.
First we apply modus tollens to the second and third hypotheses to deduce
a useful intermediate conclusion ¬f . We then use this conclusion with the first
hypothesis in another application of modus tollens, deducing ¬(b ∨ c). We replace
this with the logically equivalent proposition ¬b ∧ ¬c. Finally, we use the rule of
simplification to arrive at the desired conclusion. ◭

Obviously this method is much shorter than constructing by hand the truth table for
!! " "
(b ∨ c) → f ∧ (f → p) ∧ ¬p → ¬c.

However, using the rules of inference takes some getting used to, and some students are
frightened into always doing a truth table. It will be far more beneficial to you if you try
to master the rules of inference. And as always, practice makes perfect.

◮ Example 7.24 Prove that the following inference is valid:

(p ∨ q) → r
¬p → r
r → ¬s
s∨q

∴ r
Section 7.8. Rules of inference 155

Proof.
(¬p → r) ∧ (r → ¬s) H2 ∧ H3
=⇒ ¬p → ¬s hypothetical syllogism
⇔ s→p . . . (1) contrapositive
(s ∨ q) ∧ (s → p)! " ! " H4 ∧ (1)
⇔ s ∧ (s → p) ∨ q ∧ (s → p) distributivity
! "
=⇒ p ∨ q ∧ (s → p) modus ponens

! =⇒" p ∨ q . . . (2) simplification

(p ∨ q) → r ∧ (p ∨ q) H1 ∧ (2)
=⇒ r C modus ponens

Although all four hypotheses were used in the above proof, the last two are
in fact redundant, i.e. the conclusion follows logically from just the first two
hypotheses. Can you prove this? As a hint, try to get the intermediate conclusion
p → r. ◭

Exercises
15. Write the following argument in symbolic form and then prove that it is valid.
“If I study law, then I will make a lot of money. If I study archaeology, then I will
travel a lot. If I make a lot of money or travel a lot, then I will not be disappointed.
I am disappointed. Therefore I studied neither law nor archaeology.”

16. Prove that the following inference is valid:

q→p
¬q → r
¬p → ¬r

∴ p

Next is a new style of question. You are given hypotheses, but no conclusion!
You should study the hypotheses (in symbolic form), and try to apply the rules of
inference to them in order to deduce one or more simple conclusions.

17. What simple logical conclusions, if any, follow from the following sets of hypotheses?

(a) If the patient has a virus her temperature is above 40◦ C.

The patient’s temperature is 38.5◦ C.
(b) If interest rates rise inflation occurs.
Inflation occurs.
156 Chapter 7. Propositional Logic

(c) The neighbours have bought a new boat.

The neighbours are not rich.
To buy a new boat you need to be rich or you must borrow money.
(d) If there are typographical errors in these notes, then it is because either I can’t
spell correctly, or the word-processor has not been programmed properly.
If the word-processor has not been programmed properly, then the manufac-
turer will fix it.
There are typogriphical errors in these notes.
The manufacturer will not fix the word processor.
(e) (Adapted from a textbook on logic by the mathematician Lewis Carroll, author
of “Alice’s Adventures in Wonderland”.)
A kitten that loves fish can be taught.
Only kittens with whiskers have tails.
A green-eyed kitten cannot be taught.
Only a kitten with a tail will play with a gorilla.
A kitten that does not love fish does not have whiskers.

7.9 Testing validity of inferences

As we saw in section 7.7, a proposition C is a valid conclusion from the hypotheses H1 ,
H2 , . . . , Hn if and only if the compound proposition

P := (H1 ∧ H2 ∧ · · · ∧ Hn ) → C

is a tautology. So we can always determine the validity or otherwise of an inference simply

by computing a truth table. This is relatively easy to do if we have a computer available
with the appropriate software, but it can be quite long and tedious if we are forced to do
it by hand.
In the last section we saw how the rules of logical implication can be used to prove
that a given inference is valid, but they can’t be used directly to show that an argument
is invalid. For this we need another approach.
Look back at the inference in Example 7.24. Suppose you were not told in advance
that the inference was valid, but instead were asked to determine its validity. How would
you go about doing this without computing the truth table? The first approach you might
think of is to try to use the rules of logical implication to prove the inference. Well, this
will work fine if the inference happens to be valid (and if you manage to find a proof),
but will get you nowhere if the inference is invalid—you’ll just keep arriving at dead ends,
but will never know whether you are missing some clever step in the “proof”. So that
method doesn’t always work.
Section 7.9. Testing validity of inferences 157

However there is a fast way to discover the validity of an inference. It is a kind of

“backwards abbreviated truth table”, as you will see. You may find the explanation of
the method hard to understand at first. However, the method is really quite easy. You
should read the explanation, follow through all the examples carefully, and then read the
explanation again. The second reading should make more sense to you.
Remember that we are trying to discover whether the compound proposition

P := (H1 ∧ H2 ∧ · · · ∧ Hn ) → C

is a tautology or not. If P is a tautology, then it always has the value 1, no matter

what values we substitute in for its variables p1 , p2 , . . . . On the other hand, if P is
not a tautology, then there is an assignment of values to the variables p1 , p2 , . . . that
makes P have the value 0. The method we use is to try to find such an assignment of
values. If we succeed, then P is not a tautology and as a result of our working we have
a counterexample to show why. On the other hand, if we fail to find the assignment of
values, then we know that P must be a tautology, and so we can set about proving it.
Note that P is of the form H → C, and we want to know whether P can ever be false.
The critical observation is that p → q is always true except in the one case where p is
true and q is false. So P is not a tautology if and only if we are able to choose values for
the variables so that H is true and C is false. Next observe that H = H1 ∧ · · · ∧ Hn . If H
is to be true we need each Hi to be true as well. We repeat these important facts before
illustrating the procedure with examples.

The inference is invalid if and only if we are able to choose values for the variables
so that each hypothesis is true and yet the conclusion is false.
◮ Example 7.25 Test the following inference for validity. If it is valid, give a proof
using the rules of inference, and if it is invalid, give a counterexample.
¬p → q
(q ∨ s) → t
p∨s

∴ t
Solution. We try to falsify the inference by making each of the hypotheses true
but the conclusion false. This is indicated by the 1s and 0s in the following working.
1
¬p → q
1
(q ∨ s) → t
1
p∨s

0
∴ t
158 Chapter 7. Propositional Logic

This tells us that t = 0, so we assign this value to t throughout the inference:

1
¬p → q
1 0
(q ∨ s) →t
1
p∨s

0
∴ t

Now look at the information we have about the second hypothesis: ((q ∨ s) → 0) =
1. This forces q ∨ s = 0, because (1 → 0) ∕= 1.
1
¬p → q
0 1 0
(q ∨ s) →t
1
p∨s

0
∴ t

This forces q = s = 0, because if either q or s were 1, then q ∨ s would also be 1.

We assign these values throughout the inference.

1 0
¬p →q
000 1 0
(q ∨s) →t
10
p ∨s

0
∴ t

The third hypothesis says p ∨ 0 = 1. Thus p = 1. But does this conflict with
the first hypothesis? It says ¬p → 0 = 1, i.e., ¬p = 0, i.e., p = 1. So these two
hypotheses agree.
01 1 0
¬ p →q
000 1 0
(q ∨s) →t
110
p∨ s

0
∴ t
The working is now complete. We have successfully assigned values to the variables
so that each hypothesis is true and yet the conclusion is false. So the inference is
invalid, and a counterexample is p = 1, q = s = t = 0. ◭
Section 7.9. Testing validity of inferences 159

◮ Example 7.26 Test the following inference for validity. If it is valid, give a proof
using the rules of inference, and if it is invalid, give a counterexample.

(¬p ∨ q) → r
p→s
¬s ∨ q

∴ r

Solution. We try to falsify the inference by making each of the hypotheses true
but the conclusion false.
1
(¬p ∨ q) → r
1
p→s
1
¬s ∨ q

0
∴ r
This forces r = 0.
1 0
(¬p ∨ q) →r
1
p→s
1
¬s ∨ q

0
∴ r
The first hypothesis says ((¬p ∨ q) → 0) = 1. Thus ¬p ∨ q = 0, i.e., ¬p = q = 0,
i.e., p = 1 and q = 0.
0100 1 0
( ¬ p ∨ q ) →r
1 1
p→ s
10
¬s ∨q

0
∴ r
The second hypothesis says 1 → s = 1. Thus s = 1. But the third hypothesis
says ¬s ∨ 0 = 1, i.e., ¬s = 1, i.e., s = 0. So the second and third hypotheses give
conflicting information. Thus we have failed to find an assignment of values to the
variables so that each hypothesis is true and the conclusion is false, that is, there
is no counterexample to this inference. Hence this inference is valid. Now we give
160 Chapter 7. Propositional Logic

a proof using the rules of inference:

¬s ∨ q H3
⇔ s→q . . . (1) implication
(p → s) ∧ (s → q) H2 ∧ (1)
=⇒ p → q hypothetical syllogism
! ⇔ " ¬p ∨ q . . . (2) implication
(¬p ∨ q) → r ∧ (¬p ∨ q) H1 ∧ (2)
=⇒ r C modus ponens ◭
To finish off this section, here are some words of advice on how to approach and set
out your answers to questions like those in the above examples.
The first thing to realise is that you don’t have to draw up a new inference diagram
each time you deduce new information. Just use the one diagram and gradually add in
more and more 1s and 0s until it is complete. These notes are meant to teach you the
method and so obviously we had to show all the steps involved in arriving at the final
diagram.
If a given inference is relatively simple, it may be worthwhile to spend a few moments
staring at the inference to see if you can work out in your head a proof using the rules of
inference. If you do see a proof, then you can go straight into writing it (without using
the method above). In that case the inference is valid and you have a proof to show it.
For more complicated examples or if you can’t see a proof quickly, you should use the
above method to test the validity. If the method fails (by reaching conflicting information
as in Example 7.26), then the inference is valid. In this case, usually you are asked to
prove the inference using the rules of inference. This is because the proof reveals why
the inference is valid, i.e. what sequence of simple logical deductions leads from the
hypotheses to the conclusion.
If the method succeeds, then you have a ready-made counterexample. However, be-
cause someone else trying to follow your work doesn’t know the order in which you put
the 1s and 0s on your diagram, you should not finish your answer just by reporting the
counterexample values. The marker needs to be convinced that these values really con-
stitute a counterexample. So, you should substitute the values back into the components
of the inference to show that each hypothesis is true and the conclusion is false.
◮ Example 7.27 A perfect solution to Example 7.25 would look like this:
01 1 0
¬ p →q
000 1 0
(q ∨s) →t
110
p∨ s

0
∴ t
Section 7.10. Solutions to exercises 161

The inference is invalid.

Counterexample: p = 1, q = s = t = 0.
With these values the hypotheses are true but the conclusion is false:

H1 := ¬p → q = ¬1 → 0 = 0 → 0 = 1.
H2 := (q ∨ s) → t = (0 ∨ 0) → 0 = 0 → 0 = 1.
H3 := p ∨ s = 1 ∨ 0 = 1.
C := t = 0. ◭

Exercises
In each of the following questions, analyse the inference for logical validity. If
the inference is valid, give a proof using the rules of inference. Otherwise, give a
counterexample.

18.
q → ¬t
s↔r
¬q ∨ s

∴ t∨r

19.
p → (q → ¬w)
¬s → q
¬t
¬p ∨ t

∴ w→s

7.10 Solutions to exercises

1. (a) (i) Sarah is outside and she is watching TV.
(ii) Sarah is riding her bicycle and either watching TV or not eating her dinner.
(b) (i) e ∨ r
(ii) ¬i ∧ ¬r
(iii) i ∧ e ∧ w

3. (a) (i) If Sarah is riding her bicycle, then she is outside.

(ii) If Sarah is inside, then either she is watching TV or eating her dinner.
162 Chapter 7. Propositional Logic

(iii) Sarah is inside or riding her bicycle, and if she is inside then she is watching
TV.
(iv) If Sarah is not riding her bicycle, then she is inside and either watching
TV or eating her dinner.
(b) (i) ¬w → ¬i
(ii) (i ∧ e) → ¬r

4. (a)
p q ¬ p → (p ∨ q)
0 0 1 0 0 0 0 0
0 1 1 0 1 0 1 1
1 0 0 1 1 1 1 0
1 1 0 1 1 1 1 1
(2) (1) ∗(6)∗ (3) (5) (4)
(b)
p q (p ↔ q) ↔ (¬ p ↔ ¬ q)
0 0 0 1 0 1 1 0 1 1 0
0 1 0 0 1 1 1 0 0 0 1
1 0 1 0 0 1 0 1 0 1 0
1 1 1 1 1 1 0 1 1 0 1
(1) (3) (2) ∗(9)∗ (5) (4) (8) (7) (6)
(e)
p (p ⊕ p) ⊕ p
0 0 0 0 0 0
1 1 0 1 1 1
(1) (3) (2) ∗(5)∗ (4)

5. (a) True (vacuously)

(b) True (trivially)
(c) True (vacuously and trivially)
(d) False

6. (a) Converse: If Sarah is inside, then she is watching TV.

Inverse: If Sarah is not watching TV, then she is outside.
Contrapositive: If Sarah is outside, then she is not watching TV.
(b) Conditional (re-written): If I work hard, then I will pass the course.
Converse: If I passed the course, then I worked hard.
Inverse: If I don’t work hard, then I will fail the course.
Contrapositive: If I failed the course, then I didn’t work hard.
Section 7.10. Solutions to exercises 163

(c) Converse: If x = 0, then xy = 0.

Inverse: If xy ∕= 0, then x ∕= 0.
Contrapositive: If x ∕= 0, then xy ∕= 0.
(d) Converse: If A ⊆ C, then A = B ∩ C.
Inverse: If A ∕= B ∩ C, then A ∕⊆ C.
Contrapositive: If A ∕⊆ C, then A ∕= B ∩ C.

7. (c)
p q (p → q) ↔ (¬q → ¬p)
0 0 1 1 1 1 1
0 1 1 1 0 1 1
1 0 0 1 1 0 0
1 1 1 1 0 1 0
(1) ∗(5)∗ (2) (4) (3)

8. (c)
p q (p ∨ q) ∧ (¬p ∧ ¬q)
0 0 0 0 1 1 1
0 1 1 0 1 0 0
1 0 1 0 0 0 1
1 1 1 0 0 0 0
(1) ∗(5)∗ (2) (4) (3)

10. (a)
¬(¬p ∨ ¬q) ⇔ ¬¬p ∧ ¬¬q De Morgan
⇔ p∧q double negation %

11. (a)
p → T ⇔ ¬p ∨ T implication
⇔ T nullity %
(b)
T → p ⇔ ¬T ∨ p implication
⇔ F∨p
⇔ p identity %

12. (a) Invalid. There are two counterexamples: p = q = r = 0 (LHS = 1, RHS = 0)

and p = r = 0, q = 1 (LHS = 1, RHS = 0).
(b) Valid.
(c) Valid.

13. The simplest way to verify this equivalence is by a truth table.

164 Chapter 7. Propositional Logic

! "
14. (a) Valid. (m ∨ t) ∧ ¬m → t is a tautology.
! "
(b) Invalid! (a → u) ∧ ¬a → ¬u is not a tautology. The reason this argument is
invalid is because, from the information given, it could be that Sunday follows
another day besides Saturday. Remember that you need to separate the logic
from its context.

15. Denote propositions as follows:

l: “I study law”;
m: “I make a lot of money”;
a: “I study archaeology”;
t: “I travel a lot”;
d: “I am disappointed”.
The argument in symbolic form is:

l→m
a→t
(m ∨ t) → ¬d
d

∴ ¬l ∧ ¬a

Proof. ! "
(m ∨ t) → ¬d ∧ d H3 ∧ H4
=⇒ ¬(m ∨ t) modus tollens
⇔ ¬m ∧ ¬t . . . (1) De Morgan
! =⇒ ¬m
" . . . (2) simplification
(l → m) ∧ ¬m H1 ∧ (2)
=⇒ ¬l . . . (3) modus tollens
¬m ∧ ¬t (1)
! =⇒ " ¬t . . . (4) simplification
(a → t) ∧ ¬t H2 ∧ (4)
=⇒ ¬a . . . (5) modus tollens
¬l ∧ ¬a (3) ∧ (5)
⇔ ¬l ∧ ¬a C %
Section 7.10. Solutions to exercises 165

16.
¬p → ¬r H3
⇔ r→p . . . (1) implication
(¬q → r) ∧ (r → p) H2 ∧ (1)
=⇒ ¬q → p . . . (2) hypothetical syllogism
(q → p) ∧ (¬q → p) H1 ∧ (2)
⇔ (¬q ∨ p) ∧ (q ∨ p) implication
⇔ (¬q ∧ q) ∨ p distributivity
⇔ F∨p excluded middle
⇔ p C identity %

17. (a) The patient doesn’t have a virus.

(b) No conclusions.
(c) The neighbours have borrowed money.
(d) The word-processor has been programmed properly.
I can’t spell correctly.
(e) The hypotheses are:
f →l
¬w → ¬t
g → ¬l
¬t → ¬p
¬f → ¬w
There are many conclusions. The one needing the most logical steps is “A
green-eyed kitten will not play with a gorilla.”

18. Invalid. Counterexample: q = r = s = t = 0.

19. Invalid. Counterexample: p = s = t = 0, q = w = 1.

Chapter 8

Predicate Logic

8.1 Truth sets

In Chapter 7 we defined an open proposition on a set U as a statement which contains
one or more variables, and which becomes an ordinary proposition when the variables are
replaced by specific elements of U. The set U on which the open proposition is defined is
known as the index set or universal set. Open propositions are also known as predicates
and in the remainder of this chapter we study predicate logic.
If P is a predicate defined on U, then the set of all x in U for which P (x) is true is
denoted by {x | P (x)}. This set is called the truth set of P . Its complement, the set of
all x in U for which P (x) is false, is known as the falsity set of P . If P (x) is true for all
x in U, then we say that P is a tautology, and if P (x) is false for all x in U, we say that
P is a contradiction.

◮ Example 8.1 Consider the following predicates on R:

(a) P1 (x) denotes “x3 − 5x2 − 34x + 80 = 0”.
(b) P2 (x) denotes “x2 − 1 = (x + 1)(x − 1)”.
(c) P3 (x) denotes “x2 + 1 = 0”.
(d) P4 (x) denotes “x5 = x + 1”.
The truth sets of these predicates are, respectively:
(a) {−5, 2, 8}
(b) R
(c) ∅
(d) {1.167303 . . .}
Since the truth sets of P2 and P3 are the universal set and the empty set, respec-
tively, P2 is a tautology and P3 is a contradiction. Graphs of y = x5 and y = x + 1
drawn on the same axes show that the truth set of P4 consists of a single number.
The first seven significant digits of this number are shown above. ◭

166
Section 8.2. Quantifiers 167

8.2 Quantifiers
The logical connectives of propositional calculus allow us to deal with two, three, or any
finite combination of propositions. For example, the fact that {−5, 2, 8} is the truth set of
the predicate P1 in Example 8.1 could be summarised by the true compound proposition
P1 (−5) ∧ P1 (2) ∧ P1 (8). However, the fact that the predicate P2 (x) is true for all real
numbers x cannot be expressed in this way because there are infinitely many real numbers.
To express the truth set of P2 (x) as a proposition we need to introduce a symbol which
means “for all”.
The expression “for all” is known as the universal quantifier, and is denoted by ∀,
which is an inverted A. Other translations of ∀ are “for each” and “for every”. Another
common translation is “for any” but this can be confused with “for any one” so you
should avoid it. The universal quantifier is used with a predicate such as P (x) to build
the compound proposition ∀x P (x) which is read as “for all x, P (x)”. It is important to
realise that ∀x P (x) is a proposition which has a truth value:
$
true if P (x) is true for every x ∈ U;
∀x P (x) is
false otherwise.

The truth value of ∀x P (x) can also be expressed as follows:

∀x P (x) is true if and only if P is a tautology.

◮ Example 8.2 Of the predicates defined in Example 8.1 only P2 is a tautology. So

∀x P2 (x) is true, but ∀x P1 (x), ∀x P3 (x), and ∀x P4 (x) are all false. ◭

The other quantifier of predicate logic is the existential quantifier, which is denoted
by ∃, a reversed E. This stands for “there exists”, or “there is”. The expression ∃x P (x)
is a proposition which has a truth value:
$
true if P (x) is true for at least one x ∈ U;
∃x P (x) is
false otherwise.

An alternative formulation of the truth value of ∃x P (x) is:

∃x P (x) is true if and only if P is not a contradiction.

◮ Example 8.3 Of the predicates defined in Example 8.1, only P3 is a contradiction.

So ∃x P1 (x), ∃x P2 (x), and ∃x P4 (x) are true, but ∃x P3 (x) is false. ◭

In mathematics a proposition of the form ∃x P (x) is known as an existence theorem.

It asserts the existence of some mathematical object, such as a number or a function, which
has some particular property, without necessarily saying how to find it. For example,
168 Chapter 8. Predicate Logic

“x5 = x + 1” (proposition P4 (x) of Example 8.1) is a predicate which is false for most
values of x. However there is one value of x, approximately equal to 1.67303, for which
“x5 = x + 1” is true. For this reason ∃x (x5 = x + 1) is a true proposition.
If the universal set U is finite, say U = {x1 , x2 , x3 , . . . , xn }, then the quantifiers ∀ and
∃ behave like “large” conjunctions and disjunctions. In fact, if P (x) is any predicate over
U, then
∀x P (x) ⇔ P (x1 ) ∧ P (x2 ) ∧ · · · ∧ P (xn )
and ∃x P (x) ⇔ P (x1 ) ∨ P (x2 ) ∨ · · · ∨ P (xn ).
The quantifiers ∀ and ∃ can be combined with the connectives of the propositional
calculus.

◮ Example 8.4 Consider the following predicates on N:

P (n) denotes “n is even”, and Q(n) denotes “n2 + 3n $ 70”.
The predicate P (n) ∧ Q(n) is true when n = 0, 2, 4 and 6, and false for all
other values of n. Therefore ∃n (P (n) ∧ Q(n)) is true, but ∀n (P (n) ∧ Q(n)) is false.
Similarly, the predicate P (n) ∨ Q(n) is true for n = 0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12,
14, 16, 18, 20, . . . . So ∃n (P (n) ∨ Q(n)) is true, but ∀n (P (n) ∨ Q(n)) is false. ◭

◮ Example 8.5 There are many propositions and theorems in mathematics of the
form ∀x (P (x) → Q(x)). Some of these are listed below, and you will be able to
think of many more.
(a) If x2 − 2x + 1 = 0, then x = 1.
(b) If n is even, then n2 is even.
(c) If C is a circle, then its circumference divided by its diameter equals π. ◭
◮ Example 8.6 Let P (n) be the predicate on N defined by “n is even”. Naturally,
¬P (n) means “n is odd”. The following English translations show how ¬ is to be
interpreted when used with quantifiers.
∀n ¬P (n) Every n ∈ N is odd.
¬∀n P (n) Not every n ∈ N is even.
∃n ¬P (n) There is an n ∈ N that is odd.
¬∃n P (n) There is no n ∈ N that is even. ◭
In the previous example, look at the first and fourth sentences: they have the same
meaning. And look at the second and third sentences: they have the same meaning
too. This is no coincidence. Although there are four ways of combining one ¬ with one
quantifier, there are only two meanings.

(Generalised De Morgan’s Laws) For any predicate P with one variable,

(a) ∀x ¬P (x) ⇔ ¬∃x P (x)
and (b) ∃x ¬P (x) ⇔ ¬∀x P (x).
Section 8.2. Quantifiers 169

If the universal set is finite, say U = {x1 , x2 , . . . , xn }, then these logical equivalences
can be reformulated as:
! "
(a) ¬P (x1 ) ∧ ¬P (x2 ) ∧ · · · ∧ ¬P (xn ) ⇔ ¬ P (x1 ) ∨ P (x2 ) ∨ · · · ∨ P (xn ) ,
! "
(b) ¬P (x1 ) ∨ ¬P (x2 ) ∨ · · · ∨ ¬P (xn ) ⇔ ¬ P (x1 ) ∧ P (x2 ) ∧ · · · ∧ P (xn ) .

This is why these equivalences are known as the Generalised De Morgan’s laws.

Proof of the laws. (a) ∀x ¬P (x) is true if and only if ¬P is a tautology, that is, P
is a contradiction. Similarly, ¬∃x P (x) is true if and only if ∃x P (x) is false, that is,
P is a contradiction. Since both ∀x ¬P (x) and ¬∃x P (x) are logically equivalent to the
statement “P is a contradiction”, they are logically equivalent to each other.
(b) By a similar argument, we can show that ∃x ¬P (x) and ¬∀x P (x) are both logically
equivalent to the statement “P is not a tautology”, and are therefore logically equivalent
to each other. %

De Morgan’s laws allow us to simplify expressions that contain two or more negations.
For example,
¬∀x ¬P (x) ⇔ ¬¬∃x P (x) De Morgan
⇔ ∃x P (x) double negation

Exercises
1. Consider the following predicates over the set of all birds.
F (x) denotes “x can fly” W (x) denotes “x has wings”
N (x) denotes “x builds a nest” E(x) denotes “x lays eggs”

(a) Write the following propositions in symbolic form.

(i) All birds have wings.
(ii) Some bird cannot fly.
(iii) There is some bird which neither flys nor builds a nest.
(iv) Not every bird lays eggs.
(b) Write down English equivalents of the following.
! "
(i) ∃x W (x) ∧ ¬F (x) .
! "
(ii) ∀x N (x) ∧ E(x) .
! "
(iii) ¬∀x W (x) → F (x) .
! "
(iv) ∃x (F (x) ∧ ¬N (x)) ∨ (E(x) ∧ N (x)) .

2. Consider the following predicates defined on N+ .

170 Chapter 8. Predicate Logic

E(n) denotes “n is even”, and P (n) denotes “n is prime”.

(a) Translate the following into ordinary English.

(i) ∃n (P (n) ∧ E(n)).
(ii) ∀n (E(n) ∨ ¬P (n)).
(iii) ¬∀n (E(n) ∨ P (n)).
(b) Write the following propositions in symbolic form.
(i) If n is even, then 2n + 1 is prime.
(ii) Not every positive integer is either even or prime.

8.3 Proofs and counterexamples

It is very important in mathematics to be able to decide whether a given proposition
involving quantifiers is true or false. It is equally important to be able to justify your
belief to other people. This means providing a proof if the proposition is true, or a
counterexample if the proposition is false.
Before we see some examples, we introduce some extra notation. So far we have been
writing ∀x P (x) and it has been assumed that the universal set U has been mentioned
elsewhere. However, often the universal set is included in the proposition itself by writing
∀x ∈ U P (x). This means “For all x in U, P (x)”.

◮ Example 8.7 Prove or disprove the following propositions:

(a) ∀x ∈ N (x3 < 42)
(b) ∃x ∈ N (x3 < 42)
(c) ∀x ∈ N ¬(x3 < 42)
(d) ∃x ∈ N ¬(x3 < 42)
Solution. (a) This says that the cube of every natural number is less than 42,
which is clearly false. A counterexample is x = 4 because for this x the predicate
x3 < 42 is false.
(b) This says that there is a natural number whose cube is less than 42, which is
clearly true: x = 0 is such a number.
(c) This says that the cube of every natural number is not less than 42, which is
clearly false. A counterexample is x = 0 because for this x the predicate ¬(x3 < 42)
is false. Another way to approach this proposition is to use De Morgan. By his
laws, ∀x ∈ N ¬(x3 < 42) ⇔ ¬∃x ∈ N (x3 < 42). This second proposition is the
negation of (b), and so must have the opposite truth value, i.e., it must be false.
(d) This says that there is a natural number whose cube is not less than 42, which
is clearly true: x = 4 is such a number. Another way to approach this proposition
Section 8.3. Proofs and counterexamples 171

is to use De Morgan. By his laws, ∃x ∈ N ¬(x3 < 42) ⇔ ¬∀x ∈ N (x3 < 42). This
second proposition is the negation of (a), and so must have the opposite truth
value, i.e., it must be true. ◭

An important class of propositions in predicate logic consists of statements of the form

∀n ∈ N P (n). An example is that for every n ∈ N, 0 + 1 + 2 + · · · + n = n(n+1) 2
, which
is a true statement. In the next chapter we will discuss a general recursive method for
proving such propositions, called mathematical induction.
So far in this chapter we have only considered propositions involving one quantifier.
However there are many statements in mathematics (and in everyday language) in which
two or more quantifiers are used. An example is: ∃B ∈ R ∀n ∈ N+ (n2 /2n $ B), which
should be read as: there exists B ∈ R such that for all n ∈ N+ , n2 /2n $ B. This is a
true statement and means that the sequence n2 /2n , n = 1, 2, . . ., has an upper bound B.
Compared to statements with one quantifier, statements involving two or more quantifiers
can be quite a bit harder to understand and prove or disprove. We shall not study these
statements any further in this unit.

Exercises
3. Give a counterexample to the proposition “For each natural number n, 4n + 1 is
prime.”

4. Prove or disprove the following propositions. The universal set is N.

(a) ¬∃x (2x = 26).

(b) ∃x ¬(2x = 26).
(c) ¬∀x (2x = 26).
(d) ∀x ¬(2x = 26).
172 Chapter 8. Predicate Logic

8.4 Solutions to exercises

1. (a) (i) ∀x W (x)
(ii) ∃x ¬F (x)
! "
(iii) ∃x ¬F (x) ∧ ¬N (x)
(iv) ¬∀x E(x)
(b) (i) There is a bird that has wings but can’t fly.
(ii) Every bird builds a nest and lays eggs.
(iii) It is not true that every bird with wings can fly.
(iv) There is a bird that either can fly but doesn’t build a nest, or lays eggs
and builds a nest.

2. (a) (i) There is a number that is prime and even.

(ii) Every number is even or not prime.
(iii) Not every number is even or prime.
! "
(b) (i) ∀n E(n) → P (2n + 1)
! "
(ii) ¬∀n E(n) ∨ P (n)

3. n = 3

4. (a) False: ∃x (2x = 26) is obviously true (choose x = 13).

(b) True: Choose x = 0 to satisfy predicate.
(c) True: Equivalent to (b).
(d) False: Equivalent to (a).
Chapter 9

Mathematical Induction

9.1 Induction
Now we will combine some of the ideas from the previous chapters to learn a recursive
method for proving propositions of the form ∀n ∈ N P (n).

Principle of Mathematical Induction. Let P (n) be a predicate defined on N. Suppose

that:
(B) P (0) is !
true, and; "
(R) ∀n ∈ N P (n) → P (n + 1) is true.
Then ∀n ∈ N P (n) is true.

The principle of mathematical induction has two parts. The first part (B) is known
as the basis for induction, and second part (R) is known as the recursive or inductive
step.
Why should we accept that these two parts together constitute a proof that the propo-
sitions P (0), P (1), P (2), . . . are all true? The reason is that together they can be used to
show that any one of these propositions is true. For suppose that n is a natural number.
To prove that P (n) is true we would argue as follows:
The basis (B) tells us that P (0) is true.
The recursive
! step (R) with " n = 0 tells us that P (0) → P (1) is true.
P (0) ∧ P (0) → P (1) =⇒ P (1) by modus ponens.
Thus P (1) is true.
The recursive " n = 1 tells us that P (1) → P (2) is true.
! step (R) with
P (1) ∧ P (1) → P (2) =⇒ P (2) by modus ponens.
Thus P (2) is true.
And so on . . .
By repeated use of (R) and modus ponens we will eventually conclude that P (n) is true.

173
174 Chapter 9. Mathematical Induction

◮ Example 9.1 Prove that 0 + 1 + 2 + · · · + n = n(n+1)

2
for all natural numbers n.
Proof. Let P (n) be the predicate on N defined by “0 + 1 + 2 + · · · + n = n(n+1)2
”.
We will prove ∀n ∈ N P (n) by induction. The proof has two parts, the base step
and the inductive (or recursive) step.
BASE STEP: We must prove P (0). From the definition above, P (0) is the
proposition “0 = 0(0+1)
2
”, that is, “0 = 0”. Thus P (0) is true.
INDUCTIVE STEP: We must prove P (n) → P (n + 1) for every n ∈ N. To do
this we assume that P (n) is true and try to deduce P (n + 1).
So, assume that P (n) is true for some n ∈ N. This assumption is called the
inductive hypothesis. Using this assumption we must prove P (n + 1), that is,
we must prove that “0 + 1 + 2 + · · · + (n + 1) = (n+1)(n+2)
2
” is true. Here’s how we
do it:
0 + 1 + 2 + · · · + (n + 1) = (0 + 1 + 2 + · · · + n) + (n + 1)
= n(n+1)
2
+ (n + 1) by inductive hypothesis
5 6
= (n + 1) n2 + 1
= (n + 1) n+2
2
= (n+1)(n+2)
2

Thus we have shown that P (n + 1) is true. To be more precise, we have shown

that P (n) → P (n + 1) is true because we needed to use the inductive hypothesis
in our proof of P (n + 1). Notice that the proof doesn’t depend on any particular
value of n, that
! is, the proof works
" for every n. So what we have really proved is
that ∀n ∈ N P (n) → P (n + 1) is true. This completes the inductive step of the
argument.
Since we have proved both (B) and (R), the principle of mathematical induction
tells us that ∀n ∈ N P (n) is true. ◭
Proofs by mathematical induction may appear to be a little strange at first, because
they seem to be assuming the very thing that they are trying to prove. In the proof of
the recursive step in the above example, notice how P (n) is used to prove that P (n + 1)
is true. However, our objections are overcome once we realise that the truth of just the
conditional P (n) → P (n + 1) does not imply the truth of either P (n) or P (n + 1). The
truths of P (1), P (2), P (3), . . . are not established until we combine the recursive step
(R) with the basis (B). Here is another example.
◮ Example 9.2 If we add successive powers of 2, starting with 1 (which is 20 ), the
following pattern emerges:
1=1
1+2=3
1+2+4=7
1 + 2 + 4 + 8 = 15
Section 9.1. Induction 175

This leads naturally to the conjecture:

“For each n ∈ N, 1 + 2 + 4 + · · · + 2n = 2n+1 − 1.”

We prove this by induction as follows.

Proof. Let P (n) denote the predicate on N defined by “1 + 2 + 4 + · · · + 2n =
2n+1 − 1”.
BASE STEP: P (0) is the proposition “1 = 21 −1”, which is true. This completes
part (B) of the proof.
INDUCTIVE STEP: Assume that P (n) is true for some n ∈ N. We seek to
prove P (n + 1), which is “1 + 2 + 4 + · · · + 2n+1 = 2n+2 − 1”. We have

1 + 2 + 4 + · · · + 2n+1 = (1 + 2 + 4 + · · · + 2n ) + 2n+1
= (2n+1 − 1) + 2n+1 by inductive hypothesis
= 2(2n+1 ) − 1
= 2n+2 − 1 as required.
! "
So ∀n ∈ N P (n) → P (n + 1) is true. This completes part (R) of the proof.
Since both the basis (B) and the recursive step (R) have been proved, by the
principle of mathematical induction, ∀n ∈ N P (n) is true. ◭

Mathematical induction can be illustrated in a variety of non-mathematical ways.

Imagine a row of coins, all balanced on their edges and waiting to be knocked over. (These
are the propositions P (0), P (1), P (2), . . . waiting to be shown to be true.) Suppose also
that they are arranged so that if any one of the coins is knocked over, then it will also
knock over the coin in front of it, (this is the recursive step (R)). If the first coin is knocked
over (this is the basis (B)), it is not hard to imagine what will happen: eventually each
coin will be knocked over. So it is with mathematical induction. The recursive step (R)
and the base step (B) together are enough to prove that all of the propositions in the
sequence are true.
The analogy with falling coins suggests a way of slightly modifying the principle of
mathematical induction. What happens if, instead of knocking over the first coin in the
row, one of the coins in front of it is knocked over? Clearly, all the coins in front of the
one first knocked over will be eventually knocked over, but the coins behind it will remain
upright.

Modified Principle of Mathematical Induction. Let i be an integer, and define

I := {n ∈ Z | n # i}. Let P (n) be a predicate defined on some set of integers containing I.
Suppose that:
(B) P (i) is !true, and; "
(R) ∀n ∈ I P (n) → P (n + 1) is true.
Then ∀n ∈ I P (n) is true.
176 Chapter 9. Mathematical Induction

Note that instead of writing ∀n ∈ I P (n), where I is the set just defined, we can also
write ∀n # i P (n).

◮ Example 9.3 Exponential functions grow faster than power functions. By this we
mean that if a > 0 and if b > 0, then an > nb , for all “sufficiently large” natural
numbers n.
To illustrate what is meant by “for all sufficiently large natural numbers n”,
we shall consider the particular case in which a and b are both equal to 2. Values
of 2n and n2 for the first few natural numbers n appear below.

n 0 1 2 3 4 5 6 7 8
n
2 1 2 4 8 16 32 64 128 256
n2 0 1 4 9 16 25 36 49 64

This table makes it clear that “2n > n2 ” is not true for all numbers n. However
it does suggest that “2n > n2 ” is true for n # 5, that is, ∀n # 5 (2n > n2 ) is true.
We can prove this using the modified principle of induction.
Proof. Let P (n) be the predicate on N defined by “2n > n2 ”.
BASE STEP: Because 25 = 32 and 52 = 25, P (5) is true. This establishes (B).
INDUCTIVE STEP: Now suppose that P (n) is true for some n # 5. We have
to prove that “2n+1 > (n + 1)2 ” is true.

2n+1 = 2(2n )
> 2(n2 ) by inductive hypothesis
= n2 + n2
= n2 + n · n
# n2 + 5 · n because n # 5
= n2 + 2n + 3n
> n2 + 2n + 1 because n # 5 so 3n # 15 > 1
= (n + 1)2

Thus P (n + 1) is true. This establishes the recursive step (R).

So, by the modified principle of mathematical induction, ∀n # 5 (2n > n2 ) is
true. ◭

Our last example of induction shows how to prove that a formula for the general term
defined by a simple recurrence relation is correct.

◮ Example 9.4 Consider the recursive definition: t1 := 2 and tn := 3tn−1 + 5, and

the proposed formula for the general term of the sequence: tn = 12 (3n+1 − 5). Using
induction we will prove that this formula is correct for all n. (To be precise, we
use modified induction because the sequence starts at t1 rather than t0 .)
Section 9.1. Induction 177

Proof. Let P (n) be the predicate on N+ defined by “tn = 12 (3n+1 − 5)”.

BASE STEP: P (1) says “t1 = 12 (31+1 − 5)”, that is, “t1 = 2”, which agrees with
the definition we were given of t1 . Thus P (1) is true.
INDUCTIVE STEP: We assume that P (n) is true for some n ∈ N+ . Now we
have to prove P (n + 1), that is, “tn+1 = 12 (3n+2 − 5)”.

tn+1 = 3tn + 5 by definition

1 n+1
= 3 · 2 (3 − 5) + 5 by inductive hypothesis
1 n+1
= 2 (3 · 3 − 15) + 5
1 n+2
= 2 (3 − 15 + 10)
1 n+2
= 2 (3 − 5)

Thus P (n + 1) is true.
So, by induction, ∀n ∈ N+ P (n) is true. ◭

Exercises
Note that if a question asks you to prove something by induction, it could mean the
ordinary principle of induction or the modified principle.

1. Use mathematical induction to prove that each of the following statements is true
for all n ∈ N.

(a) 02 + 12 + 22 + · · · + n2 = n(n + 1)(2n + 1)/6.

(b) 03 + 13 + 23 + · · · + n3 = (1 + 2 + 3 + · · · + n)2 .

2. Let r be any real number different from 1. Use induction to prove the following
formula for all n ∈ N:
1 − rn+1
1 + r + r2 + · · · + rn = .
1−r

3. Use induction to prove that each of the following statements is true for all n ∈ N.

(a) 6 is a factor of n(n2 + 5).

(b) 6 is a factor of n(n + 1)(n + 2).
(c) 4 is a factor of 5n − 1.
(d) 5 is a factor of 8n − 3n .

4. Use modified induction to prove that each of the following statements is true for all
n ∈ N+ .

(a) 12 + 32 + 52 + · · · + (2n − 1)2 = n(2n − 1)(2n + 1)/3.

178 Chapter 9. Mathematical Induction

(b) 13 + 33 + 53 + · · · + (2n − 1)3 = n2 (2n2 − 1).

5. Use modified induction to prove the following:

(a) n2 # n + 2 for n # 2.
(b) n3 # n2 + 3 for n # 2.
(c) 3n # n3 for n # 5.

6. Below are recursive definitions of sequences, together with a suggested general for-
mula. Use induction to prove that the formula is correct.

(a) t1 := 1; tn := tn−1 + 2n − 1 (n # 2). Formula: tn = n2 .

(b) s0 := −1; sn := 2sn−1 − 3 (n # 1). Formula: sn = 3 − 2n+2 .
! "
(c) a1 := 3; an+1 := −an + 7 (n # 1). Formula: an = 12 (−1)n + 7 .

7. Guess a general formula which simplifies the product

1 1 1 1
pn := (1 − )(1 − )(1 − ) · · · (1 − 2 ) ,
4 9 16 n
and prove it by induction.

8. Consider the predicate P (n) on N+ defined by “n2 + n is even”.

(a) Show that for each positive integer n, P (n) → P (n + 1) is true.

(b) Show that for each positive integer n, ¬P (n) → ¬P (n + 1) is true.
(c) Is P (1) true or false?
(d) What conclusions can be drawn from (a), (b) and (c)?
(e) What is the moral of the exercise?

9. Consider the following open propositions defined on N+ :

P1 (n) denotes “6 is a factor of n3 + 5n”.
P2 (n) denotes “n = n + 1”
P3 (n) denotes “n2 + 3n is odd”
! "
(a) Show that for each of these open propositions ∀n P (n) → P (n + 1) is true.
(b) For which of these propositions is ∀n P (n) true? Justify your answer briefly.
Section 9.2. Solutions to exercises 179

9.2 Solutions to exercises

1. (a) Proof. Let P (n) be the predicate on N defined by “02 + 12 + 22 + · · · + n2 =
n(n + 1)(2n + 1)/6”.
BASE STEP: P (0) says “02 = 0(1)(1)/6”. Both the LHS and RHS are 0,
therefore P (0) is true.
INDUCTIVE STEP: Assume P (n) is true for some n ∈ N. We have to prove
P (n + 1) which is “02 + 12 + · · · + n2 + (n + 1)2 = (n + 1)(n + 2)(2n + 3)/6”.

02 + 12 + · · · + n2 + (n + 1)2 = n(n + 1)(2n + 1)/6 + (n + 1)2 by ind. hyp.

= (n(n + 1)(2n + 1) + 6(n + 1)(n + 1))/6
= ((n + 1)(n(2n + 1) + 6(n + 1)))/6
= (n + 1)(2n2 + 7n + 6)/6
= (n + 1)(n + 2)(2n + 3)/6 as required.

So ∀n ∈ N (P (n) → P (n + 1)) is true.

Since both the basis and the recursive step have been proved, by the principle
of mathematical induction, ∀ ∈ P (n) is true. %
(b) Hint. Assuming 03 + 13 + · · · n3 = (1 + 2 + · · · + n)2 is true for some n ≥ 0, we
want to show that 03 + 13 + · · · + n3 + (n + 1)3 = (1 + 2 + · · · + n + (n + 1))2 .
By our inductive hypothesis we know that 03 +13 +· · ·+n3 +(n+1)3 = (1+2+
· · · + n)2 + (n + 1)3 , so we now need to show that (1 + 2 + · · · + n)2 + (n + 1)3 =
(1 + 2 + · · · + n + (n + 1))2 .

(1 + 2 + · · · + n + (n + 1))2 = (1 + · · · + n)2 + 2(1 + · · · + n)(n + 1) + (n + 1)2

= (1 + 2 + · · · + n)2 + n(n + 1)(n + 1) + (n + 1)2
(see Example 7.1!)
= (1 + 2 + · · · + n)2 + (n + 1)(n(n + 1) + n + 1)
= (1 + 2 + · · · + n)2 + (n + 1)(n2 + 2n + 1)
= (1 + 2 + · · · + n)2 + (n + 1)3

2. Hint. Assuming 1 + r + · · · + rn = (1 − rn+1 )/(1 − r) is true for some n ∈ N, we

want to show that 1 + r + r2 + · · · + rn + rn+1 = (1 − rn+2 )/(1 − r).
1−rn+1
1 + r + r2 + · · · + rn + rn+1 = 1−r
+ rn+1 by ind. hyp.
1−r n+1 +(1−r)rn+1
= 1−r
1−rn+1 +rn+1 −r.rn+1
= 1−r
1−rn+2
= 1−r

3. (a) Proof. Let P (n) be the predicate on N defined by “6 is a factor of n(n2 + 5)”.
180 Chapter 9. Mathematical Induction

BASE STEP: P (0) says “6 is a factor of 0”, which is true. (Explanation: Every
n ∈ N+ is a factor of zero because n0 is a whole number.)
INDUCTIVE STEP: Assume P (n) is true for some n ∈ N, that is, assume that
6 is a factor of n(n2 + 5). This means that n(n2 + 5) = 6k for some whole
number k. We will use this k later in our working. Now we have to prove
P (n + 1), that is, “6 is a factor of (n + 1)((n + 1)2 + 5)”.

(n + 1)((n + 1)2 + 5) = (n + 1)(n2 + 2n + 1 + 5)

= (n + 1)((n2 + 5) + (2n + 1))
= n(n2 + 5) + n(2n + 1) + 1(n2 + 5) + 1(2n + 1)
= 6k + 2n2 + n + n2 + 5 + 2n + 1 by ind. hyp.
= 6k + 3n2 + 3n + 6
= 6(k + 1) + 3n(n + 1)

We must show that this last expression is divisible by 6. Since 6 is a factor of

6(k + 1), the problem is to show that 3n(n + 1) is divisible by 6. Now n and
n + 1 are consecutive integers, and so one of them is even. Therefore n(n + 1) is
even. So 3n(n + 1) is a number divisible by both 2 and 3: hence it is divisible
by 6. This shows that the expression above is indeed divisible by 6. So we
have proved P (n + 1).
By induction, we have proved ∀n ∈ N P (n). %
(c) Hint. Assuming 4 being a factor of 5n − 1 is true for some n ∈ N we want to
show that 4 is a factor of 5n+1 − 1.

5n+1 − 1 = 5.5n − 1
= (4 + 1)5n − 1
= 4.5n + 5n − 1
= 4.5n + (5n − 1)
= 4.5n + 4k by ind. hyp.
= 4(5n + k)

4. (a) Hint. Show that (n + 1)(2(n + 1) − 1)(2(n + 1) + 1)/3 and n(2n − 1)(2n +
1)/3 + (2(n + 1) − 1)2 both multiply out to the same thing.

5. (a) Hint. Assuming n2 ≥ n + 2 is true for some n ≥ 2, we must show that

(n + 1)2 ≥ (n + 1) + 2.

(n + 1)2 = n2 + 2n + 1
≥ (n + 2) + 2n + 1 by ind. hyp.
= 3n + 3
≥ n + 3 since n ≥ 2
Section 9.2. Solutions to exercises 181

(b) Hint. Assuming n3 ≥ n2 + 3 is true for some n ≥ 2, we want to show that

(n + 1)3 ≥ (n + 1)2 + 3.

(n + 1)3 = n3 + 3n2 + 3n + 1
≥ (n2 + 3) + 3n2 + 3n + 1 by ind. hyp.
= (n2 + 2n + 1) + 3n2 + n + 3
= (n + 1)2 + 3n2 + n + 3
≥ (n + 1)2 + 3 since n ≥ 2

(c) Hint. Assuming 3n ≥ n3 is true for some n ≥ 5, we must show that 3n+1 ≥
(n + 1)3 i.e. 3n+1 ≥ n3 + 3n2 + 3n + 1.

3n+1 = 3.3n
≥ 3n3 by ind. hyp.
= n3 + 2n3
> n3 + n3
≥ n3 + 5n2 since n ≥ 5
= n3 + 3n2 + n2 + n2
> n3 + 3n2 + 3n + 1

7. pn = n+1
2n
(n # 2).

8. (c) P (1) is true.

(d) From (a) we can conclude nothing. From (b) we can conclude nothing. From
(a) and (c) together, we can conclude ∀n ∈ N+ P (n), by induction.
(e) The moral of the exercise is that to prove something by induction you need
both the base step and the inductive step. The inductive step alone doesn’t
prove anything.

9. ∀n P1 (n) is true. The other two are false.

Chapter 10

Boolean Algebra

10.1 Definition and examples

Boolean algebra is an abstract form of the algebra of propositions, as described in Chap-
ter 7. Why do we use an abstract version rather than the real thing? The reason is
because the algebra of propositions has striking similarities with other kinds of algebras
which occur in mathematics, and they all have very similar properties. It is almost as
if the same text has been written in different languages; the words are different but the
underlying story is the same.
Our purpose here is to examine the story, without becoming bogged down in the
different languages in which it can be told. This process is known as abstraction, and is
very common in mathematics. At the most elementary level, we did this with counting
when we discovered that 2 + 3 = 5, whether the objects counted were people, coins, or
anything else; this general result about counting applies no matter what is being counted.
When we abstract, we choose properties common to the specific systems we are dealing
with, we set them out in “neutral” language and notation, and then we use these as
“axioms” of an abstract system. The trick of successful abstraction is to choose just
enough properties to capture the common features of the systems being abstracted without
losing too much generality. Anything we can prove in the abstract setting will apply in
each specific version. There is no need to repeat the same proofs over and over again in
the different “languages” of these versions.
The abstract system we shall study is called Boolean algebra, in honour of the
British mathematician George Boole (1815–1864). It is based on the similarities between
the algebras of sets and propositions.

Definition. A Boolean algebra is a set B, which has two special elements denoted by
0 and 1, together with three operations denoted by ′ , ·, and +, satisfying the properties
listed in Table 10.1.

182
Section 10.1. Definition and examples 183

For any elements x, y and z in a Boolean algebra:

1. (a) x+y =y+x commutativity
(b) xy = yx
2. (a) (x + y) + z = x + (y + z) associativity
(b) (xy)z = x(yz)
3. (a) x + yz = (x + y)(x + z) distributivity
(b) x(y + z) = xy + xz
4. (a) x+0=x identity
(b) x1 = x
5. (a) x + x′ = 1 complementation
(b) xx′ = 0

Table 10.1: The axioms of Boolean algebra

The elements of B are sometimes called terms. Terms of the Boolean algebra can be
combined using the operations ′ , ·, and + to produce new terms.
The operation ′ acts on one variable at a time: x′ (read as “x dash”) is called the
complement of x.
The operation + acts on two variables at a time: x + y is called the sum of x and y.
The operation · also acts on two variables at a time: x · y is called the product of x
and y. Normally we omit the · and just write xy for the product: x · y is used only when
we wish to emphasise the multiplication.
To show that our abstract Boolean algebras have practical applications, we give some
examples.

◮ Example 10.1 Let U be any set, and let P(U) be its power set, the set of all
subsets of U (see Example 4.1). The sets U and ∅ are the special elements 1 and
0, respectively. The operations that make P(U) a Boolean algebra are the usual
set operations of (complement), ∪ (union) and ∩ (intersection), corresponding
to ′ , + and ·, respectively. It is easy to check that the axioms of a Boolean algebra
are satisfied by P(U), with these operations and these special elements. ◭

◮ Example 10.2 Let F (210) denote the set of all positive integer factors of 210.
That is,

F (210) := {1, 2, 3, 5, 6, 7, 10, 14, 15, 21, 30, 35, 42, 70, 105, 210}.

We can make this finite set a Boolean algebra by defining operations , ∨, and ∧,
as follows.
184 Chapter 10. Boolean Algebra

For any two elements m and n of F (210),

m := 210/m;
m ∨ n := the lowest common multiple of m and n;
m ∧ n := the highest common factor (called the greatest common divisor) of m and n.

For example, 10 = 21, 6 ∨ 15 = 30, and 6 ∧ 15 = 3.

It can be verified that with , ∨ and ∧ taking the roles of ′ , + and ·, respec-
tively, F (210) is a Boolean algebra. The special elements 1 and 0 are 210 and 1,
respectively. ◭

We could define F (n), and the operations , ∨, and ∧ on F (n), for any positive
integer n. However F (n) is a Boolean algebra only for special positive integers n. We
shall look more closely at this in one of the problems at the end of the section.
The Boolean algebra F (210), and many others like it, are really just curiosities, with
little or no practical value. We shall return to more serious matters. The next example of
a Boolean algebra is very simple, but of great importance in logic and computer science.

◮ Example 10.3 The simplest of all Boolean algebras has just two elements, namely
0 and 1. We denote this Boolean algebra by B2 . The complements, products and
sums of the elements 0 and 1 are as shown in Table 10.2 below.

x x′ · 0 1 + 0 1
0 1 0 0 0 0 0 1
1 0 1 0 1 1 1 1

Table 10.2: Arithmetic tables for B2

Then it can be easily verified that B2 is a Boolean algebra. It is sometimes

known as the 0–1 Boolean algebra. ◭

The 0–1 Boolean algebra B2 is important in logic and computer science precisely
because it has only two elements. In logic, the elements 1 and 0 are used to represent
the states “true” and “false”, respectively, which we associate with propositions. The
operations ′ , · and + on B2 defined in Table 10.2 correspond precisely to the logical
connectives ¬, ∧ and ∨ between propositions. This is clear by comparing Table 10.2 with
the truth tables for ¬, ∧ and ∨ given in Table 7.1.
In computer science the elements of B2 can be treated as binary digits, which in turn
can be regarded as the basic units of information, or as digits of the binary number
system. The addition and multiplication tables for B2 have a close relationship with
binary arithmetic. In fact, the products in Table 10.2 correspond exactly to the products
Section 10.1. Definition and examples 185

of the binary digits 0 and 1. However the sums in Table 10.2 do not correspond exactly
to the sums of the binary digits 0 and 1—there are different results for the sum 1 + 1.
There are many other systems where the two elements of B2 are used to represent the
two possible states. Examples of such systems include electric circuits in which a current
may or may not be flowing, capacitors which may be charged or not charged, switches that
may be open or closed, and conducting coils which may be magnetised in one direction
or the other.

◮ Example 10.4 A Boolean expression is an algebraic expression obtained by

combining Boolean variables using Boolean operations. For example, E1 = xy + z ′ ,
E2 = x(y + z) and E3 = x′ + y ′ z are all examples of Boolean expressions in the
three Boolean variables x, y and z.
Boolean expressions can themselves be combined using Boolean operations. For
example E1 + E2 = xy + z ′ + x(y + z), and (E3 )′ = (x′ + y ′ z)′ . In fact it is possible
to show that the set of all Boolean expressions involving a given set of Boolean
variables is itself a Boolean algebra. (We won’t give the proof here).
In the simplest case, the Boolean variables x, y, z, . . . can take only 0 and 1 as
values. In that case the set of Boolean expressions can be interpreted in a number
of useful ways. For example, we can think of x, y, z, . . . as basic propositions. Then
the possible values of x, y, z, . . . correspond to the logical possibilities and Boolean
expressions correspond to compound propositions. So for example, E1 corresponds
to (x ∧ y) ∨ (¬z) and E2 corresponds to x ∧ (y ∨ z).
Alternatively, as we shall see in Chapter 11, we can think of x, y, z, . . . as the
“input states” in an electronic circuit and the Boolean expressions as the outputs
obtained by combining the inputs via devices known as “gates”. ◭

From now on we shall assume that all Boolean variables that we encounter can take
only the values 0 or 1. With this simplification we shall be able to use truth tables, as
well as the general axioms of Boolean algebra, to prove results such as the equality of
Boolean expressions.

Exercises
1. The set of real numbers has no zero-divisors. That is, if x and y are real numbers,
and if xy = 0, then x = 0 or y = 0. Show by counterexample that there is no such
general rule for Boolean algebras. That is, give an example of a Boolean algebra B,
and non-zero elements x and y in B, such that xy = 0.

2. The cancellation law also holds for real numbers. That is, if x, y and z are
real numbers, and if xz = yz, and if z ∕= 0, then x = y. Show, by giving a
counterexample, that the cancellation law fails in Boolean algebras.
186 Chapter 10. Boolean Algebra

3. Refer to Example 10.2.

(a) Write 210, and then each of the elements of F (210), as a product of prime
numbers.
(b) For each element m in F (210) describe how the prime factors of m and the
prime factors of m are related.
(c) Describe how the prime factors of m ∧ n and m ∨ n are related to the prime
factors of m and n.
(d) Describe the correspondence between F (210) and the Boolean algebra of sub-
sets of {2, 3, 5, 7}.

For any positive integer n, we can define the operations , ∧, and ∨ on F (n), the
set of all positive factors of n, as in Example 10.2.

(e) Show that F (66) is a Boolean algebra.

(f ) Show that F (12) is not a Boolean algebra.
(g) Write 66 and 12 as the product of prime numbers.
(h) Describe those positive numbers n for which F (n) is a Boolean algebra.

10.2 Proofs in Boolean algebra

The similarities of the axioms in Table 10.1 with corresponding laws for logical equiva-
lences (Table 7.6) are obvious, but there are fewer axioms. We try to keep the number of
axioms in any algebraic system to a minimum for two reasons.
Firstly, it achieves simplicity. Including as an axiom any property of the system that
can be deduced from the other axioms does not add anything, and only tends to obscure
the key features of the system.
Secondly, it makes easier the task of showing that any particular mathematical object
is an example of the algebraic system. In order to show that X is an example of an
algebraic system S, it is necessary to show that X satisfies each of the axioms of S.
Clearly this is less trouble if there are fewer axioms.
Each of the laws of logical equivalences in Table 7.6 has a Boolean algebra counterpart.
Those which are not axioms can be deduced from the axioms in Table 10.1, and are called
theorems. Some of these theorems are listed in Table 10.3 below.
Section 10.2. Proofs in Boolean algebra 187

For any elements x and y in a Boolean algebra:

6. (a) x+x=x idempotence
(b) x·x=x
7. (a) x+1=1 dominance
(b) x0 = 0
8. x′′ = x double complementation
9. (a) (x + y)′ = x′ y ′ De Morgan’s laws
(b) (xy)′ = x′ + y ′
10. (a) x + xy = x absorption
(b) x(x + y) = x

Table 10.3: Some basic properties of Boolean algebra

These theorems (except 10(a)–(b)) correspond to the parts of Table 7.6 that are not
already listed in Table 10.1. We will see proofs of some of these theorems in the following
examples.
You may have noticed a kind of symmetry in the axioms and properties in Tables 10.1
and 10.3. The following definition explains this so-called duality.

Definition. The dual of a Boolean expression is obtained by replacing each + with a ·,

each · with a +, each 1 with a 0, and each 0 with a 1.

For example, the dual of (x + y ′ ) · 1 is (x · y ′ ) + 0. Note that duality is not equality.

That is, if E is a Boolean expression and Ẽ is its dual, then usually E ∕= Ẽ. However,
if F is another Boolean expression and E = F , then Ẽ = F̃ . This fact (which we won’t
prove) is called the principle of duality.

Principle of duality. If two Boolean expressions can be shown to be equal, then the
dual expressions are also equal.

Note that the axioms and properties listed in Tables 10.1 and 10.3 are in dual pairs,
except for property 8, which is its own dual.
188 Chapter 10. Boolean Algebra

◮ Example 10.5 We shall establish the idempotence laws of Boolean algebra:

(a) x + x = x, and (b) x · x = x.
Proof. (a) x+x = (x + x) · 1 axiom 4(b)
= (x + x)(x + x′ ) axiom 5(a)
= x + xx′ axiom 3(a)
= x+0 axiom 5(b)
= x axiom 4(a)

(b) x·x = xx + 0 axiom 4(a)

= xx + xx′ axiom 5(b)
= x(x + x′ ) axiom 3(b)
= x·1 axiom 5(a)
= x axiom 4(b)

Notice that (a) and (b) are duals, and that each line in the proof of (b) is
the dual of the corresponding line in the proof of (a). This is typical of this type
of proof, and instead of writing dual proofs as we have done here, we can simply
appeal to the duality principle. ◭
◮ Example 10.6 We shall establish the following dominance laws.
(a) x + 1 = 1, and (b) x · 0 = 0.
Proof. (a) x+1 = (x + 1) · 1 axiom 4(b)
′
= (x + 1)(x + x ) axiom 5(a)
= x + 1x′ axiom 3(a)
= x + x′ 1 axiom 1(b)
′
= x+x axiom 4(b)
= 1 axiom 5(a)
Since (b) is the dual of (a) and (a) is true, the principle of duality tells us that
(b) is also true. ◭
We can use theorems already proved to prove new ones; it is not necessary to go back
to the axioms at every step.
◮ Example 10.7 We shall prove the following absorption laws:
(a) x + xy = x and (b) x(x + y) = x.
Proof. (a) x + xy = x1 + xy axiom 4(b)
= x(1 + y) axiom 3(b)
= x(y + 1) axiom 1(a)
= x1 dominance
= x axiom 4(b)
Since (b) is the dual of (a) and (a) is true, the principle of duality tells us that
(b) is also true. ◭
Section 10.2. Proofs in Boolean algebra 189

We now move on to proving some more complicated statements. First, though, a

word about the setting out of proofs. The axioms in Table 10.1 and basic properties in
Table 10.3 can be used at any time in a proof. You don’t have to remember the numbers
of these axioms and properties, but you should remember their names, and should write
the relevant name next to each line of working.

◮ Example 10.8 Prove that y + (x′ + y)′ = x + y.

Proof.
y + (x′ + y)′ = y + (x′′ · y ′ ) De Morgan
= y + xy ′ double complementation
= (y + x)(y + y ′ ) distributivity
= (y + x) · 1 complementation
= y+x identity
= x+y commutativity ◭

Because of the correspondence between Boolean algebra and the algebra of logical
propositions, the MATLAB programs truth.m and propos.m given in Section 7.3 can be
used to verify that two Boolean expressions are equal. To verify that y + (x′ + y)′ = x + y
from the previous example, we show that q ∨ ¬(¬p ∨ q) ↔ p ∨ q is a tautology. The
appropriate function M-file propos.m and the output of the program are given below.

function prop=propos(p)
% Define the proposition(s) for the truth table
global num
if p<0
% Define number of basic variables p(j)
num=2;
else
% Enter the proposition(s)
prop(1) = p(2) | ~(~p(1) | p(2));
prop(2) = p(1) | p(2);
prop(3) = prop(1) == prop(2);
end

Basic variables Proposition value(s)

0 0 888 0 0 1
0 1 888 1 1 1
1 0 888 1 1 1
1 1 888 1 1 1

Since all the entries in the last column are equal to 1, the result is verified.
190 Chapter 10. Boolean Algebra

◮ Example 10.9 Prove that (x+y = y) =⇒ (x′ +y = 1). This is a common type of
statement. It says “If the first equation is true, then so is the second”. Therefore,
in order to prove the second equation, we must use the first one somewhere in the
proof. At the point where we use it we write “given” to indicate that this is neither
an axiom nor a theorem, but that we were assuming it is true for this particular x
and y.
Proof.
x′ + y = x′ + (x + y) given
= (x′ + x) + y associativity
= (x + x′ ) + y commutativity
= 1+y complementation
= y+1 commutativity
= 1 dominance ◭

The above proof contains two steps that use the commutativity axiom, namely to get
from lines 2 to 3 and from 4 to 5. Steps like this are considered so trivial that they are
not normally written. So the proof would usually be presented as follows:

x′ + y = x′ + (x + y) given
= (x′ + x) + y associativity
= 1+y complementation
= 1 dominance

You may use this abbreviated style of proof in your assignments, tests and exams.
To verify the result in the previous example using MATLAB, we show that (p ∨ q ↔
q) → (¬p ∨ q ↔ T ) is a tautology. The appropriate function M-file propos.m, and the
output of the program truth.m are given below.

function prop=propos(p)
% Define the proposition(s) for the truth table
global num
if p<0
% Define number of basic variables p(j)
num=2;
else
% Enter the proposition(s)
prop(1) = (p(1) | p(2)) == p(2);
prop(2) = (~p(1) | p(2)) == 1;
prop(3) = prop(1) <= prop(2);
end
Section 10.2. Proofs in Boolean algebra 191

Basic variables Proposition value(s)

0 0 888 1 1 1
0 1 888 1 1 1
1 0 888 0 0 1
1 1 888 1 1 1

Since all the entries in the last column are equal to 1, the result is verified.

Exercises
4. Prove each of the following Boolean identities. Use MATLAB to verify the results.

(a) x + x′ y = x + y
(b) (x + y + z)(x′ + y ′ + z ′ ) = (xyz + x′ y ′ z ′ )′

5. Prove that in any Boolean algebra:

(a) (x + y)(x′ + y ′ )′ = xy
(b) x + x′ y + x′ y ′ = 1
(c) (x + y)(x + y ′ ) = x
(d) xy ′ z ′ + xyz + x′ yz + xyz ′ = yz + xz ′

6. We saw in Exercise 1 that the zero-divisors law for numbers doesn’t work in Boolean
algebra. However, there is a zero-divisors law for Boolean algebra: it is just a bit
more complicated than the one for numbers. Here it is: If xy = 0 and xy ′ = 0, then
x = 0. Prove this law.

7. There is also a cancellation law for Boolean algebra (see Exercise 2): If xz = yz and
xz ′ = yz ′ , then x = y. Prove this law.

8. Prove the following implications for elements x and y of a Boolean algebra.

(a) (xy = x) =⇒ (x + y = y).

(b) (x′ + y = 1) =⇒ (xy ′ = 0).
(c) (xy ′ = 0) =⇒ (xy = x).

9. Show that each element in a Boolean algebra has just one complement. That is,
show that if xy = 0 and x + y = 1, then y = x′ .

10. Which of the following propositions are always true in a Boolean algebra? Justify
your answer with either a proof or a counterexample. Check your answer using
MATLAB.
192 Chapter 10. Boolean Algebra

(a) If x(y + z ′ ) = x(y + w′ ), then z = w.

(b) If z = w, then x(y + z ′ ) = x(y + w′ ).

10.3 Some more Boolean operations

A Boolean algebra is defined to have the three operations ′ , + and · but we can build new
ones in terms of these three.

Symmetric difference. The first one we will look at is one that we’ve used before in
logic: the “symmetric difference” or “exclusive or” operation. In logic we defined ⊕ in
words, but we cannot do this in the abstract setting of Boolean algebra: we may only
refer to the three inherent operations. It turns out that the correct way to define the
exclusive or is:
x ⊕ y := xy ′ + x′ y.
You should check that this definition matches our idea of ⊕ from logic. That is, check on
a truth table that p ⊕ q = (p ∧ ¬q) ∨ (¬p ∧ q). Note that x ⊕ y is also written x XOR y.

NAND. The second operation we look at is the NAND operation, denoted by x ↑ y

or x NAND y, and defined by
x ↑ y := (xy)′ .
NAND is short for NOT AND: You should be able to see where this name comes from if
you write (xy)′ in the notation of propositional logic.

NOR. The third and final new operation we consider is the dual of ↑. It is the NOR
operation, denoted by x ↓ y or x NOR y, and defined by

x ↓ y := (x + y)′ .

NOR is short for NOT OR: If you write (x + y)′ in the notation of propositional logic
you will see where this name comes from.

NAND and NOR are examples of universal operations. This means that the three
basic operations of Boolean algebra can be expressed using only NAND or only NOR.
Here are the relevant expressions using only NAND:

x′ = x ↑ x;
x + y = (x ↑ x) ↑ (y ↑ y);
xy = (x ↑ y) ↑ (x ↑ y).
Section 10.4. Boolean expressions 193

And here are the relevant expressions using only NOR:

x′ = x ↓ x;
x + y = (x ↓ y) ↓ (x ↓ y);
xy = (x ↓ x) ↓ (y ↓ y).

Notice how these expressions are dual to those given for NAND.

Exercises
11. Show that for any elements x, y and z in a Boolean algebra B,

(a) x ⊕ y = y ⊕ x.
(b) (x ⊕ y) ⊕ z = x ⊕ (y ⊕ z).
(c) x + y = (x ⊕ y) + xy.
(d) x(y ⊕ z) = xy ⊕ xz.

12. Show that the operations ↑ and ↓ are related as follows:

! " ! "
(a) x ↑ y = (x ↓ x) ↓ (y ↓ y) ↓ (x ↓ x) ↓ (y ↓ y) .
! " ! "
(b) x ↓ y = (x ↑ x) ↑ (y ↑ y) ↑ (x ↑ x) ↑ (y ↑ y) .

10.4 Boolean expressions

Let B be a Boolean algebra. A function f which maps B n to B is called an n-variable
Boolean function. The “inputs”, or independent variables, x1 , x2 , . . . , xn take values
from B. The “output” f (x1 , x2 , . . . , xn ) also has a value from B and depends on the
inputs x1 , x2 , . . . , xn .
Some Boolean functions are easier to describe than others. However there is always
a formula for f (x1 , x2 , . . . , xn ) in terms of the variables x1 , x2 , . . . , xn and the Boolean
operations ′ , +, ·, ⊕, ↑ and ↓.

◮ Example 10.10 The following formulas define five 3-variable Boolean functions.
(a) f1 (x, y, z) := xy ′ + yz ′ + zx′ .
(b) f2 (x, y, z) := xyz ′ + xy ′ z + x′ yz + x′ y ′ z + x′ yz ′ + xy ′ z ′ .
(c) f3 (x, y, z) := (x + y + z)(x′ + y ′ + z ′ ).
(d) f4 (x, y, z) := x + y ′ .
(e) f5 (x, y, z) := 1. ◭
194 Chapter 10. Boolean Algebra

The expressions on the right hand side of these formulas can all be regarded as ex-
amples of 3-variable Boolean expressions. The first three involve all three variables, but
the last two don’t. However it does not normally matter whether we regard x + y ′ as
an expression in two variables or in more. If it does matter, then we need to indicate
carefully just how we are to regard such an expression. In the same manner, 0 and 1 can
be regarded as Boolean expressions in any number of variables.
Every Boolean expression determines a Boolean function and every Boolean function
can be defined by a Boolean expression. However the correspondence is not 1–1. There are
many different Boolean expressions for each Boolean function. We say that two Boolean
expressions are equivalent if one can be shown to equal the other by applying the laws
of Boolean algebra.

◮ Example 10.11 The Boolean expressions (x + y ′ z ′ )′ and x′ (y + z) are equivalent.

We can see this by applying some of the laws of Boolean algebra: From De Morgan’s
laws and double complementation it follows that

(x + y ′ z ′ )′ = x′ (y ′ z ′ )′ = x′ (y ′′ + z ′′ ) = x′ (y + z) ◭

The problem of deciding whether two Boolean expressions are equivalent is very im-
portant because of its application to logic and to the design of electronic circuits. In these
two cases the 0–1 Boolean algebra B2 is the underlying algebraic system. The decision
problem can be done systematically for B2 by using truth tables, but because the truth
table for a Boolean expression with n variables has 2n rows, this becomes very slow as n
increases, even for computers! No fast and systematic method for testing the equivalence
of Boolean expressions in many variables has yet been discovered.

10.5 Disjunctive normal form

The standard way of writing Boolean expressions is as sums of products. Each of the
expressions in Example 10.10, except for the expression in (c), is a sum of products. (The
expression in (c) is a product of sums.) Sums of products are quite easy to evaluate when
the variables are replaced by specific terms of the Boolean algebra.
However, there are many equivalent ways of writing a Boolean expression as a sum of
products. If we want uniqueness, then we need to be more specific. We shall study a way
to do this, but first we need some extra terminology.
A minterm in n-variables x1 , x2 , . . . , xn is a product u1 u2 · · · un , where, for each i,
ui is xi or x′i . We can regard minterms as the basic building blocks for Boolean functions.

◮ Example 10.12 Here are four minterms in 3 variables: xyz, x′ yz, xy ′ z ′ and x′ y ′ z ′ .
There are eight altogether. You shouldn’t have any trouble writing down the other
four. ◭
Section 10.5. Disjunctive normal form 195

Any Boolean expression that is the sum of different minterms in the same variables is
said to be a disjunctive normal form (DNF) or in minterm canonical form. It is
also convenient to regard 0 as a disjunctive normal form.
◮ Example 10.13 In Example 10.10 the expression in (b) is in DNF because each
summand is different and involves all three variables. The expressions in (a), (d)
and (e) are not in DNF because there are summands that do not involve all three
variables. The expression in (c) is not even a sum of products and so cannot be
in DNF. ◭
The following theorem (which we won’t prove in this unit) shows why the DNF is so
important.
Theorem 10.1 Every Boolean expression is equivalent to a disjunctive normal form,
with equivalent Boolean expressions having the same DNF.
The theorem gives us a way to test whether two Boolean expressions are equivalent
or not:
Method for testing equivalence. Convert each expression to its DNF. If the DNFs
are the same, then the original expressions are equivalent. If the DNFs are different, then
the original expressions are not equivalent.
The next example shows how we convert a Boolean expression to its DNF.
◮ Example 10.14 To convert xy ′ + yz ′ + x′ y to its DNF, we first introduce the third
variable into each summand, and then expand and simplify.
xy ′ + yz ′ + x′ y = xy ′ (z + z ′ ) + (x + x′ )yz ′ + x′ y(z + z ′ )
= xy ′ z + xy ′ z ′ + xyz ′ + x′ yz ′ + x′ yz + x′ yz ′
= xy ′ z + xy ′ z ′ + xyz ′ + x′ yz ′ + x′ yz (DNF) ◭
There are algorithms for converting a Boolean expression to its DNF, but we will not
study them here. We shall just give a further example that illustrates another aspect of
the procedure.
◮ Example 10.15 To convert xy ′ + (x + yz)′ to DNF, we first use De Morgan’s laws
to simplify the expression to a sum of products.
xy ′ + (x + yz)′ = xy ′ + x′ (yz)′
= xy ′ + x′ (y ′ + z ′ )
= xy ′ + x′ y ′ + x′ z ′
To complete the conversion to DNF, we introduce a third variable into each term:
xy ′ + x′ y ′ + x′ z ′ = xy ′ (z + z ′ ) + x′ y ′ (z + z ′ ) + x′ (y + y ′ )z ′
= xy ′ z + xy ′ z ′ + x′ y ′ z + x′ y ′ z ′ + x′ yz ′ + x′ y ′ z ′
= xy ′ z + xy ′ z ′ + x′ y ′ z + x′ y ′ z ′ + x′ yz ′ (DNF) ◭
196 Chapter 10. Boolean Algebra

Exercises
13. Obtain the DNF of each of the following:

(a) x + y.
(b) (x + y + z)(x′ + y ′ + z ′ ).
(c) (x + yz ′ )′ (y + zx′ )′ (z + xy ′ )′ .
(d) (x′ + y ′ z)x(y + z ′ ).

14. Which of the following Boolean expressions are equivalent? Check your answers
using MATLAB.

(a) xy + yz + zx.
(b) x.
(c) (x + y)(y + z)(z + x).
(d) (x + y)(x + y ′ ).
(e) x(y + z) + yz.
(f ) xy + (x′ + y)′ .

10.6 Minimal sum of products form

The disjunctive normal form is theoretically important because it can be used to show
whether or not two Boolean expressions are equivalent. However, the DNF of a Boolean
expression is often long and complicated. It is usually more important in practical appli-
cations to simplify Boolean expressions by finding equivalent expressions which use fewer
variables and operations.

◮ Example 10.16 The following pairs of Boolean expressions are equivalent:

(a) x′ y + (y + xz)′ and (x(y + z))′ .
(b) xy(z + (x + y)z ′ ) + (x′ + y)′ and x.
In each case the second expression involves fewer variables and operations, and is
therefore a simplification of the first. ◭

In the next chapter we will see how to build the logic network for a Boolean expression.
From this an electronic circuit can be constructed which models the behaviour of the
Boolean expression as a function. Boolean operations are represented in logic networks as
gates. Since logic networks with many gates are expensive to make and slow to operate,
it is useful to simplify the Boolean expressions as much as possible. One way of doing
this is to find equivalent Boolean expressions which are sums of products and which use
the least possible number of terms. We call such expressions minimal sums of products.
Section 10.6. Minimal sum of products form 197

◮ Example 10.17 If f (x, y, z) = xy ′ z + xyz ′ + x′ z + xy ′ z ′ + x′ yz ′ then it can be

shown that f is equivalent to g(x, y, z) = xy ′ + yz ′ + x′ z. Both f and g are sums
of products, but g has just 3 summands and involves 6 different terms (2 in each
summand). Furthermore there is no equivalent Boolean expression that has fewer
than 3 summands or that uses fewer than 6 terms. (We will be able to verify this
later.) For this reason we say that g is a minimal sum of products, or that g is a
minimal sum of products form of f . ◭

To say precisely what is meant by a minimal sum of products, we need more termi-
nology. A literal is any variable x, or its complement x′ , and a fundamental product
is any product of literals. The constants 0 and 1 can also be regarded as fundamental
products. As a matter of interest, fundamental products can be defined recursively as
follows.
(B) 0 and 1 are fundamental products, and every literal is a fundamental product;
(R) If P1 and P2 are fundamental products, then P1 P2 is a fundamental product.

◮ Example 10.18 (a) 1, x′ , xyz ′ , x′ y, wxyz are fundamental products.

(b) Any minterm is a fundamental product.
(c) x ⊕ y, xyz + x′ y, (xzyz)′ are not fundamental products. ◭

A Boolean expression is a sum of products if it is a sum of fundamental products.

Here is a recursive definition:
(B) Every fundamental product is a sum of products;
(R) If E1 and E2 are sums of products, then E1 + E2 is a sum of products.

◮ Example 10.19 (a) The Boolean expressions xy ′ + yz ′ + xz ′ and xy ′ + yz ′ + xyz ′

are sums of products.
(b) Any Boolean expression in DNF is a sum of products.
(c) x ⊕ y, (x + y)′ are not sums of products. ◭

A Boolean expression is a minimal sum of products (MSP) if

(i) it is a sum of products;
(ii) there is no equivalent sum of products with fewer summands, and;
(iii) there is no equivalent sum of products with the same number of summands and a
smaller total number of literals.
Recall that the DNF of an expression is unique. This is not so for the MSP form. For
example, the function f in Example 10.17 has two MSP forms. One of them is g, as given
in the example, and the other is h(x, y, z) = x′ y + y ′ z + xz ′ . So a Boolean expression may
have several different MSP forms.
One way of finding an MSP form of a Boolean expression in four variables or less is
to use a method introduced by Karnaugh in 1953. This can be thought of as a Boolean
198 Chapter 10. Boolean Algebra

algebra analogue of truth tables for propositions. Before we can explain the method itself
we need to introduce the framework, sort of like an empty truth table.
A Karnaugh table for a Boolean expression in n variables is a rectangular table with
n
2 entries, one for each of the possible minterms. The table is arranged so that entries
corresponding to any two minterms which differ in just one variable are adjacent (i.e.,
they share a common edge). For example, the minterms xy ′ z and x′ y ′ z differ in just the
x variable, whereas xyz and x′ yz ′ differ in both x and z.
A Karnaugh table for Boolean expressions in two variables is shown in Figure 10.1. It
is not usual to label the entries themselves, but the rows and columns instead. The label
of any entry in the table is the “product” of the corresponding row and column labels.

y y′
x
x′

Figure 10.1: Karnaugh table for expressions in two variables

A Karnaugh table for Boolean expressions in three variables is shown in Figure 10.2.
It is a 2 × 4 rectangular table with each square corresponding to one of the eight possible
minterms. The labels on the long side are the four possible minterms in two of the
variables, and the labels on the short side are the minterms in the remaining variable. It
does not matter which two variables we group together. Note that to satisfy all of the
adjacency conditions we imagine that the short sides at each end are joined together, as if
the rectangle is wrapped around into a cylinder. That is, we consider the first and fourth
columns to be adjacent, because their labels only differ by one literal.

yz yz ′ y′z′ y′z
x
x′

Figure 10.2: Karnaugh table for expressions in three variables

A Karnaugh table for Boolean expressions in four variables is shown in Figure 10.3.
It is a 4 × 4 rectangular table. The labels on one side are the minterms in two of the
variables, and the labels on the other side are the minterms in the remaining two variables.
Again, in order to satisfy the adjacency requirements, we must imagine that the top edge
is joined to the bottom edge, and that the left edge is joined to the right edge. It may
help to imagine that the table is drawn on a doughnut (or torus).

Now that we have the framework we can discuss Karnaugh’s method for finding an
Section 10.6. Minimal sum of products form 199

yz yz ′ y′z′ y′z
wx
wx′
w ′ x′
w′ x

Figure 10.3: Karnaugh table for expressions in four variables

MSP form of a Boolean expression. Each square or entry in a Karnaugh table corresponds
to a minterm. So we can regard a collection of squares in a Karnaugh table as a sum of
minterms. The Karnaugh map of a Boolean expression is the collection of squares in
a Karnaugh table corresponding to its disjunctive normal form. We mark each of these
squares with a + (or some other symbol) to highlight them. So there is one + symbol for
each minterm in the DNF.

◮ Example 10.20 The Karnaugh map of xy ′ z + xyz ′ + x′ yz + x′ y ′ z + xy ′ z ′ + x′ yz ′

is shown below.
yz yz ′ y ′ z ′ y ′ z
x + + +
x′ + + + ◭

However, you do not need to convert an expression to DNF before you can draw its
Karnaugh map. All you need is for the expression to be a sum of products. The procedure
for drawing the map from such an expression is illustrated in the next example.

◮ Example 10.21 Suppose we wish to draw the Karnaugh map of the expression
E = x+(y ↓ z). First we convert E to a sum of products: E = x+(y+z)′ = x+y ′ z ′ .
Here is the map—the explanation is given below it.

yz yz ′ y′z′ y′z
x + + + +
x′ +

We draw the map by referring to the expression x + y ′ z ′ summand by summand. The

first summand is x, so we put a + in every square of the table that has an x in its row
or column label, i.e. each square for which x is a factor of the corresponding minterm.
There are four such squares. The next summand is y ′ z ′ , so we put a + in every square
that has both y ′ and z ′ in its row or column label, i.e. each square for which y ′ z ′ is
a factor of the corresponding minterm. There are two such squares, but one of them
already has a + in it so we just leave it as it is. Thus the Karnaugh map of E has five
squares covered as shown above. ◭
200 Chapter 10. Boolean Algebra

Karnaugh maps are useful for finding MSP forms because the Karnaugh maps of fundamental
products are easy to recognise. As we saw in the previous example, each fundamental product
(i.e., summand in a sum of products form) is a rectangular subset of the Karnaugh table in
which the number of squares is a power of 2. So the Karnaugh map of an expression in sum of
products form is just the superposition of the Karnaugh maps of each of the summands.
The Karnaugh map of a fundamental product is called a fundamental rectangle, or an s-
rectangle if it contains s squares. Some fundamental rectangles are shown in Figures 10.4, 10.5
and 10.6 below. Note that the size of the rectangle for a fixed fundamental product depends
on how many variables there are. Also notice the “wrap around” properties of some of the
rectangles. This results from the adjacency conditions we mentioned earlier.

Figure 10.4: Some fundamental rectangles in two variables

dd

Figure 10.5: Some fundamental rectangles in three variables

Section 10.6. Minimal sum of products form 201

Figure 10.6: Some fundamental rectangles in four variables

The procedure for finding an MSP form of a Boolean expression is to cover its Kar-
naugh map with the fewest and the largest fundamental rectangles. We use the fewest
number of fundamental rectangles to minimise the number of summands, and we use large
rectangles to minimise the total number of literals.
We start with the largest fundamental rectangles and work down to the smallest. In
the case of a Boolean expression in four variables, for example, we try to cover the Kar-
naugh maps with 16-rectangles, then 8-rectangles, then 4-rectangles, then 2-rectangles,
and finally with 1-rectangles. The procedure is described in detail below.

Procedure for finding an MSP form for a Boolean expression in n variables.

1. Construct the Karnaugh map of the expression.

2. Set k = n, and regard all squares in the Karnaugh map as “uncovered”.

3. Determine which, if any, of the uncovered squares in the Karnaugh map can be
covered by 2k -rectangles which are subsets of the Karnaugh map.
202 Chapter 10. Boolean Algebra

4. Choose the fewest number of 2k -rectangles required to cover all such squares. (There
may be more than one way to do this.)

5. Decrease k by 1. Go to step 6.

6. Repeat steps 3, 4 and 5 until the entire Karnaugh map is covered by fundamental
rectangles. Then go to step 7.

7. Remove any “unnecessary rectangles” (i.e., rectangles that, if removed, do not un-
cover any marked squares in the Karnaugh map).

8. The minimal sum of products form is the sum of the fundamental products corre-
sponding to the chosen fundamental rectangles.

◮ Example 10.22 Find MSP forms of

(a) f1 = y + xy ′ z
(b) f2 = xyz + xyz ′ + xy ′ z + x′ y ′ z ′ + x′ y ′ z
(c) f3 = x + wx′ z + (x′ y ⊕ z)
Solution. (a) The Karnaugh map of f1 , together with chosen fundamental rect-
angles, is shown below.

The chosen fundamental rectangles correspond to fundamental products y and xz.

So an MSP form of f1 is y + xz.
(b) Two Karnaugh maps for f2 are shown below.

The Karnaugh map contains no 8-rectangles and no 4-rectangles, but can be cov-
ered by 2-rectangles. This can be done in two ways, which is why we have depicted
two maps. The squares corresponding to xyz ′ and x′ y ′ z ′ are contained in just one
2-rectangle each, namely the 2-rectangles corresponding to xy and x′ y ′ respectively.
Thus any MSP form of f2 incorporates xy and x′ y ′ which together cover xyz, xyz ′ ,
x′ y ′ z ′ and x′ y ′ z. To cover xy ′ z we then have the choice of using either y ′ z or xz.
So the two MSP forms for f2 are xy + x′ y ′ + y ′ z and xy + x′ y ′ + xz. Note that
xy + x′ y ′ + xy ′ z is also equivalent to f2 , and that the three corresponding funda-
mental rectangles cover the Karnaugh map without overlapping. However it is not
minimal as it involves a total of 7 literals, rather than 6.
Section 10.6. Minimal sum of products form 203

(c) We first convert f3 to a sum of products:

f3 = x + wx′ z + (x′ y ⊕ z)
= x + wx′ z + (x′ y)z ′ + (x′ y)′ z
= x + wx′ z + x′ yz ′ + (x′′ + y ′ )z
= x + wx′ z + x′ yz ′ + (x + y ′ )z
= x + wx′ z + x′ yz ′ + xz + y ′ z

Two Karnaugh maps for f3 are shown below.

The Karnaugh map of f3 can be covered in two ways by one 8-rectangle and
three 4-rectangles. The corresponding MSP forms are x + yz ′ + y ′ z + wz and
x + yz ′ + y ′ z + wy. ◭

The MSP form is usually a useful simplification of a Boolean expression. However it

is not always the very best. Consider, for example, the expression (w ⊕ x) ⊕ (y ⊕ z). Its
Karnaugh map is shown below.

The MSP form of (w ⊕ x) ⊕ (y ⊕ z) has 8 summands, whereas (w ⊕ x) ⊕ (y ⊕ z) itself

has just 3 ⊕ summands. Unfortunately, there is no known method for finding minimal
expressions that are allowed to involve ⊕, ↑ and/or ↓.

Connection with truth tables

As seen in Section 10.2, a Boolean expression or function (on B2 ) can be represented
using a truth table. Each row in the truth table shows the value (0 or 1) of the Boolean
expression for a particular set of values of the variables.
It is not hard to see that for each row in the truth table, there is precisely one minterm
that equals 1 for this set of values of variables, and moreover this minterm equals 0 for
every other assignment of values. For example, for x = 0, y = 0 and z = 0 (the first row
of the truth table), the only minterm that equals 1 is x′ y ′ z ′ (since with any of x, y or z
204 Chapter 10. Boolean Algebra

as a factor, the value would be 0), and moreover x′ y ′ z ′ = 0 for any other assignment of
values. Similarly, the only minterm that equals 1 for x = 0, y = 1 and z = 1 is x′ yz,
which can be identified easily by including as factors x′ (since x = 0), y (since y = 1) and
z (since z = 1). Furthermore, clearly x′ yz equals 0 for every other assignment of values.
Now, given the truth table of a Boolean expression E, consider the set of minterms
that equal 1 for the rows where E = 1. If we add these minterms, the resulting Boolean
sum is equivalent to the given Boolean expression E. This is because the sum has value:
0 + · · · + 0 = 0, when E = 0 (since each minterm equals 0), and value:
0 + · · · + 0 + 1 + 0 · · · + 0 = 1 (where the 1 corresponds to a single minterm), when E = 1.
Therefore this sum is precisely the disjunctive normal form of the Boolean expression E.
Thus the truth table of a Boolean expression can be used to write down the DNF of
the expression. Since the Karnaugh map also represents the DNF, there is a one-to-one
correspondence between each marked square in the Karnaugh map and a row in the truth
table for which the expression has value 1. This connection between truth tables and
Karnaugh maps will be used in the next chapter.

◮ Example 10.23 Draw the truth table for the Boolean expression f1 = y + xy ′ z.
You will find there are five rows for which f1 = 1, namely when x, y and z are:
[0 1 0], [0 1 1], [1 0 1], [1 1 0] and [1 1 1]. These correspond to the five minterms:
x′ yz ′ , x′ yz, xy ′ z, xyz ′ and xyz, respectively. Hence the DNF of f1 is

f1 = x′ yz ′ + x′ yz + xy ′ z + xyz ′ + xyz.

These five minterms are precisely the same as those represented in the Karnaugh
map of f1 in Example 10.22(a). ◭

Exercises
15. Find MSP forms for the Boolean functions with these Karnaugh maps.

(a)
yz yz ′ y′z′ y′z
x + +
x′ +

(b)
yz yz ′ y′z′ y′z
x + + +
x′ + + +
Section 10.7. Solutions to exercises 205

(c)
yz yz ′ y′z′ y′z
wx + + +
wx′ + + +
w ′ x′ + + + +
w′ x + + +

(d)
yz yz ′ y′z′ y′z
wx + +
wx′ + +
w ′ x′ + +
w′ x +

16. Draw the Karnaugh map of f := (wx′ y ′ + yz + w′ x)′ without doing any algebraic
simplification.

17. Find an MSP form for each of the following expressions.

(a) x′ (y + z) + xy ′ z
(b) (x + z ′ )(y + x′ )(y ′ + z)
(c) (w′ ⊕ y) + xyz ′ + wxz ′
(d) wx ↑ (xy ↓ yz)

10.7 Solutions to exercises

1. Let U := {1, 2} and let B be the Boolean algebra P(U). Put x := {1} and y := {2}.
Then x ∩ y = ∅ but x ∕= ∅ and y ∕= ∅.

2. Let U := {1, 2} and let B be the Boolean algebra P(U). Put x := {1, 2}, y := {2}
and z := {2}. Then x ∩ z = y ∩ z and z ∕= ∅ but x ∕= y.

3. (a) 210 = 2 · 3 · 5 · 7
(b) They have no prime factors in common, and together they have all the prime
factors of 210.
(c) The prime factors of m ∧ n are those that are common to both m and n. The
prime factors of m ∨ n are those that are common to m or n.
(d) ∧ is like “intersection of common primes” and ∨ is like “union of common
primes”.
206 Chapter 10. Boolean Algebra

(f ) Some of the axioms are not satisfied for F (12). For example, 2 ∨ 2 is supposed
to equal 12, by axiom 5(a), but in reality 2 ∨ 2 = 2 ∨ 6 = 6.
(g) 66 = 2 · 3 · 11 and 12 = 22 · 3.
(h) F (n) is a Boolean algebra if and only if n does not have any factor p2 where p
is a prime.

4. (a) Proof.
x + x′ y = (x + x′ )(x + y) distributivity
= 1(x + y) complementation
= x+y identity %

(b) Proof.

(xyz + x′ y ′ z ′ )′ = (xyz)′ (x′ y ′ z ′ )′ De Morgan

′ ′ ′ ′′ ′′ ′′
= (x + y + z )(x + y + z ) generalised De Morgan
= (x′ + y ′ + z ′ )(x + y + z) double complementation %

6. Proof.
0 = 0+0 identity
= xy + xy ′ given
= x(y + y ′ ) distributivity
= x1 complementation
= x identity %

7. Proof.
x = x1 identity
= x(z + z ′ ) complementation
= xz + xz ′ distributivity
= yz + yz ′ given
= y(z + z ′ ) distributivity
= y1 complementation
= y identity %

8. (c) Proof.
x = x1 identity
= x(y + y ′ ) complementation
= xy + xy ′ distributivity
= xy + 0 given
= xy identity %
Section 10.7. Solutions to exercises 207

9. Proof.
y = 1y identity
= (x + x′ )y complementation
= xy + x′ y distributivity
= 0 + x′ y given
= x′ x + x′ y complementation
= x′ (x + y) distributivity
= x′ 1 given
= x′ identity %

10. (a) False: Put U := {1, 2} and consider the Boolean algebra P(U ) with x :=
{1, 2}, y := {1, 2}, z := {1} and w := {2}. This is a counterexample to the
proposition.
(b) True: If z = w, then z ′ = w′ .

11. (d) Proof.

xy ⊕ xz = (xy)(xz)′ + (xy)′ (xz) definition

= xy(x′ + z ′ ) + (x′ + y ′ )xz De Morgan
= xyx′ + xyz ′ + x′ xz + y ′ xz distributivity
= xx′ y + xyz ′ + xx′ z + xy ′ z commutativity
= 0 + xyz ′ + 0 + xy ′ z complementation
= xyz ′ + xy ′ z identity
= x(yz ′ + y ′ z) distributivity
= x(y ⊕ z) definition %

13. (a) xy + xy ′ + x′ y.
(b) xyz ′ + xy ′ z ′ + xy ′ z + x′ yz + x′ yz ′ + x′ y ′ z.
(c) x′ y ′ z ′ .
(d) 0.

14. (a), (c) and (e) are equivalent, and (b), (d) and (f ) are equivalent.
15. (a)

The MSP form is xz + x′ y ′ z ′ .

208 Chapter 10. Boolean Algebra

(b)

The MSP form is z + xy + x′ y ′ .

(c)

There are two MSP forms. The one shown here is z ′ + xy + x′ y ′ + w′ x′ . The
other is z ′ + xy + x′ y ′ + w′ y.
(d)

The MSP form is wz + w′ x′ z ′ + xyz.

16. Recall that f + f ′ = 1. Thus the Karnaugh maps of f and f ′ are “opposite” or
“complementary” in the sense that they have no squares in common and together
cover every square. So if we already had the Karnaugh map of f ′ we could obtain
the map of f simply by taking the complement of the first map. Now note that f ′
is already in a sum of products form and so its map can be drawn without doing
any algebra. In the Karnaugh table below, the map of f ′ is marked with − symbols
and the map of f with + symbols.

yz yz ′ y′z′ y′z
wx − + + +
wx′ − + − −
w ′ x′ − + + +
w′ x − − − −

The lesson of this exercise is that if you have a sum of products form for f ′ , then
you don’t need to do any more algebra to get the Karnaugh map of f .
Section 10.7. Solutions to exercises 209

17. (a) x′ y + y ′ z
(b) xyz + x′ y ′ z ′
(c) wy + w′ y ′ + xz ′
(d) You should have been able to simplify the given expression to (wxy ′ + wxy ′ z ′ )′ .
The Karnaugh map is then the complement of the (easy to draw) map for this
expression. (See previous exercise if this puzzles you.) From the map, the MSP
form is w′ + x′ + y.
However, if you look carefully, (wxy ′ + wxy ′ z ′ )′ simplifies further, by the ab-
sorption law, to just (wxy ′ )′ . We can then apply the generalised De Morgan’s
law to get the MSP form immediately, without drawing the Karnaugh map at
all! (Unfortunately, though, we don’t know that this is an MSP form until we
draw the map and cover it as usual.)
Chapter 11

Logic Networks

11.1 Introduction
Digital computers are built mainly out of “binary” components, that is, components that
have only two possible states. Examples of binary components are electric switches and
light bulbs, which can be either “on” or “off”; electric circuits in which currents are
either flowing or not flowing; magnetic cores which can be magnetised either clockwise
or anticlockwise; and locations on a card which can be either punched or unpunched.
Binary components are used because they are cheaper to make, faster to operate, and
more reliable than non-binary components.
We can regard the various functional units in a computer as logic networks or
circuits which accept a collection of inputs and generate a collection of outputs. Each
input and output is “binary”, that is, capable of assuming only two distinct values which,
for convenience, we label 0 and 1. Let x1 , x2 , . . . , xm denote the binary inputs and let z1 ,
z2 , . . . , zn denote the binary outputs. Since the outputs are determined by the inputs,
we can think of a logic network as a representation of a transmission function f which
maps B2m to B2n (where B2 is the 0–1 Boolean algebra). This is shown in Figure 11.1.

x1 ✲
✲ z1
x2 ✲
✲ z2
· f ·
· ·
· ·
✲ zn
xm ✲

Figure 11.1: f (x1 , x2 , . . . , xm ) = (z1 , z2 , . . . , zn )

210
Section 11.1. Introduction 211

The transmission function can be specified by a truth table, which lists the values of
the outputs z1 , z2 , . . . , zn for each of the 2m combinations of the input values x1 , x2 , . . . ,
xm .
The simplest logic networks, which are the basic building blocks of all networks, are
called gates. These have just a single binary output, and correspond to simple Boolean
operations. The six most elementary gates and the special symbols used to represent them
are shown in Figure 11.2. In each of these symbols, the inputs are shown on the left, and
the Boolean expressions corresponding to the outputs are underneath. The truth tables
of these six gates are shown in Table 11.1. Note that a NOT gate is more commonly
called an inverter.

x x
x z z z
y y
Inverter: z = x’ AND gate: z = xy OR gate: z = x + y
x x x
z z z
y y y
NAND gate: z = (xy)’ NOR gate: z = (x + y)’ XOR gate: z = x + y

Figure 11.2: The six standard gates

x′ xy x + y x ↑ y x ↓ y x ⊕ y
x y NOT AND OR NAND NOR XOR
0 0 1 0 0 1 1 0
0 1 1 0 1 1 0 1
1 0 0 0 1 1 0 1
1 1 0 1 1 0 0 0

Table 11.1: Truth tables of the six standard gates

Gates can be combined to form more complicated networks. The output(s) of such
networks can be described with Boolean expressions.

◮ Example 11.1 The expression wx + yz contains three connectives, namely two

products and one sum. A circuit to implement this expression can be made from
two AND gates and an OR gate. The circuit is shown below in Figure 11.3.
212 Chapter 11. Logic Networks

w
x
y
z

Figure 11.3: Example 11.1

A signal line may be “split” so that the signal can be used as input to several gates.

◮ Example 11.2 The expression (x′ + y)(x + z ′ ) can be implemented by the circuit
shown below.

Figure 11.4: Example 11.2

It is possible to build gates with more than two inputs.

! "′
◮ Example 11.3 The expression (xy)′ (y ′ + r + t′ )(s ⊕ t′ ) contains a three-term
AND, and one of the terms is a three-term OR. A circuit for this expression is
shown below.
x
y

s
t

Figure 11.5: Example 11.3

Section 11.1. Introduction 213

Suppose we have an m-input network N1 with single output u (a Boolean expression),

and another m-input network N2 with single output v. If u and v are equivalent Boolean
expressions, then we say that N1 and N2 are equivalent networks.

◮ Example 11.4 Let’s do some algebra on the expression we had in Example 11.2:

(x′ + y)(x + z ′ ) = x′ z ′ + yx + yz ′
= (x ↓ z) + y(x + z ′ ).

This last expression can be implemented by the circuit below.

x

Figure 11.6: Example 11.4

Since the two expressions are equivalent, the two networks (Figures 11.4 and
11.6) are also equivalent. That is, each network computes the same Boolean func-
tion of its inputs. ◭

Since there may be many equivalent networks that compute a given Boolean function,
how should we choose the final design? There are two general principles that dictate this
choice: we want to the circuit to be as cheap as possible to manufacture, and as fast as
possible at performing its task.
Generally speaking, cost of manufacture is measured in number of gates. Therefore,
the fewer the gates the cheaper the circuit. (It isn’t really as simple as this because some
gates cost more than others, but we’ll ignore this fact.) On the other hand, the speed of
a circuit depends on so-called “levels of gating”. Each gate causes a slight delay in the
signal on the line, so the more gates a signal has to go through, the slower the circuit is.
For example, the circuit in Example 11.2 has three levels of gating because some of its
inputs (though not all) have to go through three gates. Now look at the equivalent circuit
in Example 11.4. This network also has five gates, but there are four levels of gating. So
these two networks cost the same to make, but the first is faster.
As a rule of thumb, the fewer the gates, the fewer the levels of gating. However,
this is not always the case. It can easily happen that network N1 is faster than network
N2 but N1 has more gates (they are just arranged in a better way). So sometimes the
cheapest circuit is not the fastest! In this unit we are not going to worry too much
214 Chapter 11. Logic Networks

about this practical problem. The main thing is that we learn how to take an expression,
manipulate it algebraically, and come up with a circuit that uses as few gates as possible.
In some cases this might not be the fastest, but we won’t worry about that. If you can
see that there are two or more circuits with the fewest gates, then you should choose one
with the fewest levels of gating. However, this generally takes too much time, so you
don’t have to do this is your assignments, tests and exams. You get full marks for just
using the fewest gates possible. So remember, whenever you are asked to design a logic
network you must use as few gates as possible, but don’t have to pay attention to the levels
of gating.
The process of circuit design is illustrated in the next few examples. Note that a key
step is obtaining the MSP form of the expression.

◮ Example 11.5 Design a circuit to compute xyz + xyz ′ + x′ yz ′ .

Solution. The Karnaugh map of this expression is

and so the MSP form is xy + yz ′ . In this expression there are only four operations,
a big improvement on the original seven gates. However, we can do even better
than four gates simply by writing the expression as y(x + z ′ ). Now there are only
three operations, and this is the best possible. So the most efficient network for
the expression is:

What happened in this example is very typical. The expression we begin with
is very inefficient, so we convert to MSP, which is a big improvement. However,
the MSP might be able to be manipulated further to reach an optimal design. ◭

Instead of being given a Boolean expression to design from, it is also very common
to be given a description in words of what the output of the circuit should be. In this
situation we have to convert the description to a Boolean expression.
Section 11.1. Introduction 215

◮ Example 11.6 Consider MAJ(x, y, z), the majority function on 3 variables, de-
fined by:
%
&0 if at most one of x, y and z has the value 1.
MAJ(x, y, z) =
'1 if two or three of x, y and z have the value 1

We will design a network to compute MAJ. To determine a Boolean expression for

MAJ (in fact the DNF) we use the connection at the end of Section 10.6 and draw
the truth table:
x y z MAJ
0 0 0 0
0 0 1 0
0 1 0 0
0 1 1 1 ← x′ yz
1 0 0 0
1 0 1 1 ← xy ′ z
1 1 0 1 ← xyz ′
1 1 1 1 ← xyz
So MAJ equals 1 at any of the four minterms listed and is 0 at the other minterms.
So to get a Boolean expression for MAJ we simply add together the minterms for
which it equals 1. Thus MAJ = x′ yz + xy ′ z + xyz ′ + xyz.
Next we draw the Karnaugh map for MAJ and get an MSP form:

So the MSP form is yz + xy + xz. This can be simulated with three AND gates
all feeding into a 3-input OR gate, totalling 4 gates and 2 levels. Alternatively, we
can re-write the MSP as xy + (x + y)z. This can also be simulated in 4 gates, but
needs 3 levels so is a little slower. So the optimal circuit is:
x

y
MAJ

z
216 Chapter 11. Logic Networks

◮ Example 11.7 Design a network with four inputs w, x, y and z and one output u
so that
u = 1 ⇔ (wx = 0 and y + z = 1).
The truth table for u is shown in Table 11.2.

w x y z wx = 0 y + z = 1 u
0 0 0 0 1 0 0
0 0 0 1 1 1 1
0 0 1 0 1 1 1
0 0 1 1 1 1 1
0 1 0 0 1 0 0
0 1 0 1 1 1 1
0 1 1 0 1 1 1
0 1 1 1 1 1 1
1 0 0 0 1 0 0
1 0 0 1 1 1 1
1 0 1 0 1 1 1
1 0 1 1 1 1 1
1 1 0 0 0 0 0
1 1 0 1 0 1 0
1 1 1 0 0 1 0
1 1 1 1 0 1 0

Table 11.2: Truth table for u

From the truth table we can draw the Karnaugh map of u:

yz yz ′ y′z′ y′z
wx
wx′ + + +
w ′ x′ + + +
w′ x + + +
The Karnaugh map of u′ is just the complement of the map of u, i.e. the unmarked
squares. It is much easier to cover the map of u′ than the map of u, so it seems that
the most efficient network for u might be obtained by considering the complement
of the MSP form of u′ . We have u′ = wx + y ′ z ′ (MSP) and so

u = (wx + y ′ z ′ )′
= wx ↓ y ′ z ′
= wx ↓ (y ↓ z)
Section 11.1. Introduction 217

This leads to the following circuit for u:

w
x

y
z

Exercises
Remember to use as few gates as possible when designing a network.

1. For each of the following Boolean expressions sketch a logic network using just NOT,
OR and AND gates.

(a) x(y + z ′ ) + (x + y)z.

(b) x ⊕ yz.
(c) x ↑ y.
(d) x ↓ y.

2. Sketch logic networks, using any of the six standard gates, in which there are three
binary inputs x, y, and z, and one binary output u, where

(a) u = 1 if and only if x = y = z.

(b) u = 1 if and only if xyz ′ = 0.

3. Sketch logic networks, using any of the six standard gates, in which there are four
inputs u, v, w and x, and one output z, where

(a) z = 1 if and only if u = v and w = x.

(b) z = 1 if and only if u = v = w = x.
(c) z = 1 if and only if at least two of the inputs are 1.
(d) z = 1 if and only if uv = 0 and wx = 0.
218 Chapter 11. Logic Networks

11.2 Adders
The most basic digital arithmetic function is the addition of two binary digits (bits). A
logic circuit that represents the addition of two bits is called a half-adder. A half-adder
has two inputs x and y, and two outputs, s and c, called the sum and carry digits. The
truth table for s and c is shown below:

x y c s
0 0 0 0
0 1 0 1
1 0 0 1
1 1 1 0

Note that this table is just the base-2 addition table of x and y. We see from the table
that c = xy and s = x ⊕ y. Thus a half-adder can be implemented as shown below:

x
c

s
y

Now suppose we want to add two 2-digit binary numbers X and Y whose binary
representations are x1 x0 and y1 y0 , respectively. As we learned in Chapter 3, first we must
add the digits x0 and y0 . This can be performed by a half-adder to give a sum digit s0 and
a carry digit c0 . Then we add the carry digit c0 to the digits x1 and y1 to give a new sum
digit s1 and a new carry digit c1 . Thus the binary representation of X + Y is c1 s1 s0 . Note
that the second stage of this computation cannot be performed by a half-adder because
we need to add three bits at once, namely x1 , y1 and c0 . A logic circuit that performs
this operation is called a full-adder. A full-adder has three inputs x, y and c, and two
outputs s and c̃. The input c represents the “old” carry digit, and the output c̃ represents
the “new” carry digit. The truth table for s and c̃ is shown below:

x y c c̃ s
0 0 0 0 0
0 0 1 0 1
0 1 0 0 1
0 1 1 1 0
1 0 0 0 1
1 0 1 1 0
1 1 0 1 0
1 1 1 1 1
Section 11.2. Adders 219

From this table we see that the DNFs of the outputs are

c̃ = x′ yc + xy ′ c + xyc′ + xyc and s = x′ y ′ c + x′ yc′ + xy ′ c′ + xyc.

These expressions can then be used to design a logic network for a full-adder. This is left
as an exercise. Now, finally, we have a circuit for adding two 2-digit binary numbers x1 x0
and y1 y0 :

x1 Full c1
y1 Adder s1
c
0
x0 Half
y Adder s0
0

Figure 11.7: Adding two 2-bit numbers

Now we are ready to consider the addition of larger binary numbers. Suppose that
X and Y are two natural numbers, and that xn . . . x1 x0 and yn . . . y1 y0 are their binary
representations. Let S = X + Y , and let sn+1 . . . s1 s0 be its binary representation. The
binary digits s0 , s1 , . . . , sn+1 are computed as follows:

s0 = x0 + y0 (half-adder)
s1 = x1 + y1 + c0 (full-adder)
s2 = x2 + y2 + c1 (full-adder)
.. ..
. .
sn = xn + yn + cn−1 (full-adder)
sn+1 = cn

So this is how computers perform addition! For example, a logic circuit which represents
the addition of two 3-bit numbers x2 x1 x0 and y2 y1 y0 is shown in Figure 11.8 below.

x2 Full s3
y2 Adder s2
c1
x1 Full
y1 Adder s1

c0
x0 Half
y0 Adder s0

Figure 11.8: Adding two 3-bit numbers

220 Chapter 11. Logic Networks

It should be pointed out that more sophisticated circuits have been designed which
perform binary addition and other arithmetic operations faster than the ones shown above.
Our examples are merely to show how complicated circuits can be built out of simple ones.

Exercises
4. In practice, AND gates and OR gates are actually built from NAND gates. Design
a half-adder that uses only NAND gates and inverters.

5. (a) Find MSP forms for the outputs of a full-adder.

(b) Try to find more efficient Boolean expressions for these outputs.
(c) Design a full-adder using as few gates as possible.

6. Show that the output c̃ of a full-adder can be expressed as c̃ = MAJ(x, y, c), where
MAJ is the three-input majority function defined in Example 11.6.

11.3 Solutions to exercises

2. (a) u = x′ y ′ z ′ + xyz (MSP)
x

y u

3. (a)
z = uvwx + uvw′ x′ + u′ v ′ wx + u′ v ′ w′ x′ MSP
= uv(wx + w′ x′ ) + u′ v ′ (wx + w′ x′ )
= (uv + u′ v ′ )(wx + w′ x′ )
= (u ⊕ v ′ )(w ⊕ x′ )

u
v
z
w
x
Section 11.3. Solutions to exercises 221

(c)
z = uv + wx + uw + ux + vw + vx MSP
= uv + wx + u(w + x) + v(w + x)
= uv + wx + (u + v)(w + x)

v
z
w

4. c = xy = (xy)′′ = (x ↑ y)′ .
s = x ⊕ y = xy ′ + x′ y = (x ↑ y ′ )′ + (x′ ↑ y)′ = (x ↑ y ′ ) ↑ (x′ ↑ y).

x
c

5. (a) c̃ = xy + xc + yc (MSP). The MSP form of s is the same as its DNF.

(b) The MSP form of c̃ is the most efficient. For s the most efficient expression is
x ⊕ y ⊕ c.