MA103: Introduction to Abstract Mathematics (LT) 

2018/19

MA103
Introduction to Abstract Mathematics
Second part, Analysis and Algebra
Amol Sasane
Revised by Jozef Skokan, Konrad Swanepoel, and Graham Brightwell

Copyright
c London School of Economics 2016

All rights reserved. No part of this work may be reproduced in any form, or by any
means, without permission in writing from the author. This material is not licensed for
resale.
Contents

1 Analysis 1
1.1 The real numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Intervals and absolute values . . . . . . . . . . . . . . . . . . . . . 6
1.2 Sequences and limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.1 Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.2 Limit of a convergent sequence . . . . . . . . . . . . . . . . . . . 10
1.2.3 The sequence (xn )n∈N . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.4 Bounded and monotone sequences . . . . . . . . . . . . . . . . . . 16
1.2.5 Algebra of limits . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.2.6 The Sandwich Theorem . . . . . . . . . . . . . . . . . . . . . . . 25
1.2.7 Subsequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.1 Definition of continuity . . . . . . . . . . . . . . . . . . . . . . . . 30
1.3.2 Continuous functions preserve convergent sequences . . . . . . . . 34
1.3.3 Restrictions and compositions of functions . . . . . . . . . . . . . 36
1.3.4 Intermediate value theorem . . . . . . . . . . . . . . . . . . . . . 37
1.3.5 Extreme value theorem . . . . . . . . . . . . . . . . . . . . . . . . 40
1.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2 Algebra 55
2.1 Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.1.1 Definition of a Group . . . . . . . . . . . . . . . . . . . . . . . . . 56
2.1.2 Proving theorems, laws of exponents and solving equations in groups 63
2.1.3 Abelian groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.1.4 Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.1.5 The order of an element of a group . . . . . . . . . . . . . . . . . 68
2.1.6 Homomorphisms and isomorphisms . . . . . . . . . . . . . . . . . 69
2.1.7 Cosets and Lagrange’s theorem . . . . . . . . . . . . . . . . . . . 72
2.1.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.2 Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
2.2.1 Definition of a vector space . . . . . . . . . . . . . . . . . . . . . 83
2.2.2 Subspaces and linear combinations . . . . . . . . . . . . . . . . . 87
2.2.3 Basis of a vector space . . . . . . . . . . . . . . . . . . . . . . . . 90

iii
Contents

2.2.4 Linear transformations . . . . . . . . . . . . . . . . . . . . . . . . 93

2.2.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

iv
Chapter 1
Analysis

Analysis is the theory behind real numbers, sequences, and functions. The word ‘theory’
is important. You might, for example, have a good idea of what we mean by a ‘limit’ of
a convergent sequence of numbers, or the notion of a ‘continuous’ function, but in this
part of the course we aim to formalize such notions.

1.1 The real numbers

The rational number system is inadequate for many purposes. For instance, there is no
rational number q such that q 2 = 2, and the set
n  o
S = q ∈ Q  q2 ≤ 2


does not have a largest element in Q. So we see that the rational number system has
“gaps”. The real number system R includes numbers to fill all of these gaps. Thus the set
n  o
T = x ∈ R  x2 ≤ 2


has a largest element. This is a consequence of a very important property of the real
numbers, called the least upper bound property. Before we state this property of R, we
need a few definitions.

Definitions. Let S be a subset of R.

1. An element u ∈ R is said to be an upper bound of S if, for all x ∈ S, x ≤ u. If S

has an upper bound, then we also say that S is bounded above.

2. An element l ∈ R is said to be a lower bound of S if, for all x ∈ S, l ≤ x. If S has

a lower bound, then we also say that S is bounded below.

3. The set S is said to be bounded if it is bounded above and bounded below.

1
1 Analysis

Examples.

1. The set S = {x ∈ R | 0 ≤ x < 1} is bounded. Any real number y satisfying y ≥ 1 (for

instance 1, π, or 100) is an upper bound of S, and any real number z satisfying z ≤ 0
(for instance 0, or −1) is a lower bound of S.

2. The set S = {n | n ∈ N} is bounded below; indeed, any real number x ≤ 1 serves as a

lower bound. It has no upper bound – we shall give a formal proof of this shortly –
and so it is not bounded above, and not bounded.

3. The set S = {(−1)n | n ∈ N} is bounded. Note that S = {−1, 1}. It is bounded above
by 1 and bounded below by −1.
More generally, any finite set S is bounded.
n  o
4. The set S = n1  n ∈ N is bounded. Any real number x satisfying 1 ≤ x is an upper


bound, and 0 is a lower bound.

5. The sets Z and R are neither bounded above nor bounded below.

6. The set T = {x ∈ R | x2 ≤ 2} is bounded. Any real number x satisfying x2 ≤ 2 also

satisfies x2 < 4. Therefore, x2 − 4 < 0, that is, (x − 2)(x + 2) < 0. It follows that
either x − 2 < 0 and x + 2 > 0,
or x − 2 > 0 and x + 2 < 0.
The second case is impossible. Thus the first case has to hold, that is −2 < x < 2. It
follows that T is bounded above by 2 and bounded below by −2.

7. The set ∅ is bounded. For instance, 1 is an upper bound for ∅, since the condition
“for every element x of ∅, x ≤ 1” is satisfied: there is certainly no element x of ∅
that doesn’t satisfy x ≤ 1. Indeed, every real number is an upper bound for ∅, and
similarly every real number is a lower bound for ∅.

We now introduce the notions of a least upper bound (also called supremum) and a
greatest lower bound (also called infimum) of a subset S of R.

Definitions. Let S be a subset of R.

1. An element u∗ ∈ R is said to be a least upper bound of S (or a supremum of S)

if
a) u∗ is an upper bound of S, and
b) for each upper bound u of S, it holds that u∗ ≤ u.

2. An element l∗ ∈ R is said to be a greatest lower bound of S (or an infimum

of S) if
a) l∗ is a lower bound of S, and
b) for each lower bound l of S, it holds that l ≤ l∗ .

2
 1.1 The real numbers

Example. Set S = {x ∈ R | x < 1}. Show that the supremum of S is 1.

Solution. Clearly 1 is an upper bound of S.

Now we show that if u is another upper bound, then 1 ≤ u. Suppose not, that is,
suppose that there is an upper bound u of S with u < 1. Then we have
u+1
u< < 1,
2
where both inequalities follow using u < 1. From the second inequality, it follows that
the number u+1 2
belongs to S. The first inequality above then shows that u is not an
upper bound for S, a contradiction. Hence 1 is a supremum.
Next we show that 1 is the only supremum. Indeed, if u∗ is another supremum, then
in particular u∗ is also an upper bound, and the above argument shows that 1 ≤ u∗ . But
1 < u∗ is not possible, since 1 is an upper bound for S, and as u∗ is a supremum, u∗
must be less than or equal to the upper bound 1. So it follows that u∗ = 1.
In the above example, there was a unique supremum of the set S. In fact, this is
always the case and we have the following result.

Theorem 1.1.1. If the least upper bound of a subset S of R exists, then it is unique.

Proof. The statement of the theorem means that the set S cannot have two different
least upper bounds. Suppose then that u∗ and u0∗ are two least upper bounds of S. Then
in particular u∗ and u0∗ are also upper bounds of S. Now since u∗ is a least upper bound
of S and u0∗ is an upper bound of S, it follows that

u∗ ≤ u0∗ . (1.1)

Furthermore, since u0∗ is a least upper bound of S and u∗ is an upper bound of S, it

follows that
u0∗ ≤ u∗ . (1.2)
From (1.1) and (1.2) we obtain u∗ = u0∗ .
Thus it makes sense to talk about the least upper bound, or the supremum, of a set.
Similarly, the infimum of a set S (if it exists) is also unique (exercise).

Definitions.

1. The least upper bound of a set S (if it exists) is denoted by sup S.

2. The greatest lower bound of a set S (if it exists) is denoted by inf S.

When the supremum and the infimum of a set belong to the set, we give them the
following familiar special names:

3
1 Analysis

Definitions.

1. If sup S ∈ S, then sup S is called a maximum of S, denoted by max S.

2. If inf S ∈ S, then inf S is called a minimum of S, denoted by min S.

Examples.

1. If S = {x ∈ R | 0 ≤ x < 1}, then sup S = 1 6∈ S and so max S does not exist. But
inf S = 0 ∈ S, and so min S = 0.

2. If S = N, then sup S does not exist, inf S = 1, max S does not exist, and min S = 1.

3. If S = {(−1)n | n ∈ N}, then sup S = 1, inf S = −1, max S = 1, min S = −1.

n  o
4. If S = n1  n ∈ N , then sup S = 1 and max S = 1. We show below (after Theorem


1.1.2) that inf S = 0. So min S does not exist.

5. For the sets Z and R, sup, inf, max, min do not exist.

6. For the set ∅, sup, inf, max, min do not exist.

In the above examples, we note that if S is non-empty and bounded above, then its
supremum exists. In fact this is a fundamental property of the real numbers, called the
least upper bound property of the real numbers, which we now state:

The Least Upper Bound Property.

If S is a subset of R that is non-empty and bounded above, then S has a least upper
bound.

In other words, for a subset S ⊂ R, if

1. S 6= ∅, and

2. S has an upper bound,

then sup S exists.

Examples.

1. Use the least upper bound property to show that there exists a number s ∈ R such
that s > 0 and s2 = 2.

Solution. Let S = {x ∈ R | x2 < 2}. We’ll show that S has a least upper bound
s = sup S and that this s is the real number we want: s > 0 and s2 = 2.
In order to apply the least upper bound property, we first have to show that S 6= ∅
and that S has an upper bound. This is easy: since 12 ≤ 2, 1 ∈ S and therefore S 6= ∅.
Also, since for any x ∈ S, x2 ≤ 2 < 4, it follows that x < 2, which shows that 2 is an
upper bound of S.

4
 1.1 The real numbers

By the least upper bound property, S has a least upper bound s = sup S. Since we
have already seen that 1 ∈ S, and we know that s is an upper bound of S, s ≥ 1 > 0. It
remains to prove that s2 = 2. We do this by showing that each of the two alternatives
s2 > 2 and s2 < 2 leads to a contradiction.
Suppose first that s2 > 2. In this case, our plan is to show that, for a suitably small
ε > 0, the real number s − ε < s is also an upper bound for S. In order to find a
suitable ε, calculate
(s − ε)2 = s2 − 2sε + ε2 > s2 − 2sε.
Since s2 > 2, the right-hand side will be greater than 2 if ε is chosen sufficiently small
2 2
that s2 − 2sε > 2, which can be rearranged as ε < s 2s−2 . Thus if we choose ε = s 4s−2 ,
2
for instance, then we have ε < s 2s−2 which gives (s − ε)2 > s2 − 2sε > 2 > x2 for all
x ∈ S. Thus s − ε > x for all x ∈ S, which means that s − ε is an upper bound of S
smaller than s. This is a contradiction to the choice of s as the least upper bound.
Suppose next that s2 < 2. In this case we show that by adding a suitably small ε > 0,
we obtain that s + ε ∈ S, which contradicts that s is an upper bound. To find a
suitable ε, calculate as follows:

(s + ε)2 = s2 + 2sε + ε2 < s2 + (2s + 1)ε if ε < 1.

(Note that in this case we couldn’t throw away the positive number ε2 , but instead
we bounded it using ε2 < ε, which holds as long as ε < 1.) The right-hand side
2−s2 2−s2
s2 + (2s + 1)ε is less than 2 if ε < 2s+1 . Choosing ε = min{ 2(2s+1) , 21 }, we have ε < 1
2−s 2
and ε < 2s+1 , which implies (s + ε)2 < s2 + (2s + 1)ε < 2. Thus s + ε ∈ S. Since
s + ε > s, s is not an upper bound of S, which is a contradiction.
The only remaining possibility is that s2 = 2. We have shown the existence of s ∈ R
such that s > 0 and s2 = 2.

2. Prove the ‘greatest lower bound property’: if S is a non-empty subset of R that is

bounded below, then S has a greatest lower bound.

Solution. See Exercise 6.

Formally, we treat R as a number system, satisfying all the usual rules of arithmetic,
together with a total order <. We also treat the least upper bound property as a
fundamental property of R. Given these principles, we now deduce some other familiar
properties of R.
The following theorem is called the Archimedean property of the real numbers.

Theorem 1.1.2 (Archimedean property). If x, y ∈ R and x > 0, then there exists an

n ∈ N such that y < nx.

Proof. Suppose that the conclusion is false, that is, that there does not exist an n ∈ N
such that y < nx. This means that for all n ∈ N, y ≥ nx. In other words, y is an upper
bound of the non-empty set S = {nx | n ∈ N}. By the least upper bound property of the

5
1 Analysis

reals, S has a least upper bound u∗ . Note that u∗ − x is not an upper bound of S, since
u∗ − x is smaller than u∗ (x is positive) and u∗ is the least upper bound. Hence there
exists an element mx ∈ S (with m ∈ N) such that u∗ − x < mx, that is, u∗ < (m + 1)x.
Then (m + 1)x is also an element of S with (m + 1)x > u∗ : this contradicts the fact that
u∗ is an upper bound of S.
As a consequence of the Archimedean property we are now able to prove that the set
N of natural numbers is not bounded above.
Examples.
1. Show that the set N is not bounded above.

Solution. Suppose that N is bounded above. Then N has an upper bound y ∈ R.

Since 1 ∈ N, 1 ≤ y, and in particular, y > 0. Let x = 1. By the Archimedean property
(Theorem 1.1.2), there exists an n ∈ N such that y < nx = n. This contradicts the
fact that y is an upper bound of N.
n  o
1
2. Set S = n

 n ∈ N . Show that inf S = 0.

Solution. We know that 0 is a lower bound of S. Suppose that l is a lower bound

of S such that l > 0. By the Archimedean property (with the real numbers x and y
taken as x = 1 (> 0) and y = 1l ), there exists n ∈ N such that 1l = y < nx = n · 1 = n,
and so n1 < l, contradicting the fact that l is a lower bound of S. Thus any lower
bound of S must be less than or equal to 0. Hence 0 is the infimum of S.

1.1.1 Intervals and absolute values

Definition. An interval (in R) is a non-empty subset S of R with the property: if
x, y ∈ S and x ≤ z ≤ y, then z ∈ S.
For instance, the set S = {x ∈ R | x < 2} is an interval: if x and y are both in S, and
x ≤ z ≤ y, then in particular z ≤ y < 2 so z ∈ S.
An interval may or may not have an upper bound: if it does have an upper bound, then
it has a supremum which may or may not be in the interval. Similarly, an interval may
or may not have a lower bound: if it does have a lower bound, then it has an infimum
which may or may not be in the interval. In the example above, S has an upper bound,
and the supremum is not in S, while S has no lower bound. There are thus three possible
forms for the “lower” end of an interval, and three possible forms for the “upper end”,
making nine forms in all. These are listed in the figure below, along with the notation
for each type of interval.
An interval of the form (−∞, b), (a, b) or (a, ∞) is called an open interval. An interval
of the form (−∞, b], [a, b] or [a, ∞) is called a closed interval.
Thus in the notation for intervals used in the Figure, a parenthesis ‘(’ or ‘)’ means
that the respective endpoint is not included, and a square bracket ‘[’ or ‘]’ means that
the endpoint is included. For example, [0, 1) is the set of all real numbers x such that

6
 1.1 The real numbers

(a, b) = {x ∈ R | a < x < b}

a b
[a, b] = {x ∈ R | a ≤ x ≤ b}
a b
(a, b] = {x ∈ R | a < x ≤ b}
a b
[a, b) = {x ∈ R | a ≤ x < b}
a b
(a, ∞) = {x ∈ R | a < x}
a
[a, ∞) = {x ∈ R | a ≤ x}
a
(−∞, b) = {x ∈ R | x < b}
b
(−∞, b] = {x ∈ R | x ≤ b}
b
(−∞, ∞) = R

0 ≤ x < 1. (Note that the use of the symbol ∞ in the notation for intervals is simply a
matter of convenience and is not be taken as suggesting that there is a real number ∞.)
We do not give any special name to intervals of the form [a, b) or (a, b].
In analysis, in order to talk about notions such as convergence and continuity, we will
need a notion of ‘closeness’ between real numbers. This is provided by the absolute value
| · |, and the distance between real numbers x and y is |x − y|. We give the definitions
below.
Definitions.
1. For a real number x, the absolute value |x| of x is defined as follows:

 x if x ≥ 0,
|x| =
−x if x < 0.

2. The distance between two real numbers x and y is the absolute value |x − y| of their
difference.
Note that |x| ≥ 0 for all real numbers x, and that |x| = 0 iff1 x = 0. Thus |1| = 1,
|0| = 0, | − 1| = 1, and the distance between the real numbers −1 and 1 is equal to
| − 1 − 1| = | − 2| = 2. The distance gives a notion of closeness of two points, which is
crucial in the formalization of the notions of analysis.
We can now specify regions comprising points close to a certain point x0 ∈ R in terms
of inequalities in absolute values, that is, by demanding that the distance of the points
of the region, to the point x0 , is less than a certain positive number δ, say δ = 0.01 or
δ = 0.0000001, and so on. See Exercise 12 and Figure 1.1.
1
The word “iff” is in common use in mathematics as an abbreviation for “if and only if”.

7
1 Analysis

I
z }| {

x0 − δ x0 x x0 + δ

Figure 1.1: The interval I = (x0 − δ, x0 + δ) = {x ∈ R | |x − x0 | < δ} is the set of all

points in R whose distance to the point x0 is strictly less than δ (> 0).

The following properties of the absolute value will be useful.

Theorem 1.1.3. If x, y are real numbers, then

|xy| = |x| |y| and (1.3)

|x + y| ≤ |x| + |y|. (1.4)

Proof. We prove (1.3) by exhausting all possible cases:

x = 0 or y = 0. Then |x| = 0 or |y| = 0, and so |x| |y| = 0. On the other hand, as

x = 0 or y = 0, it follows that xy = 0 and so |xy| = 0.

x > 0 and y > 0. Then |x| = x and |y| = y, and so |x| |y| = xy. On the other hand,
as x > 0 and y > 0, it follows that xy > 0 and so |xy| = xy.

x > 0 and y < 0. Then |x| = x and |y| = −y, and so |x| |y| = x(−y) = −xy. On the
other hand, as x > 0 and y < 0, it follows that xy < 0 and so |xy| = −xy.

x < 0 and y > 0. This follows from 1.1.1 above by interchanging x and y.

x < 0 and y < 0. Then |x| = −x and |y| = −y, and so |x| |y| = (−x)(−y) = xy. On
the other hand, as x < 0 and y < 0, it follows that xy > 0 and so |xy| = xy.

This proves (1.3).

Next we prove (1.4). First observe that from the definition of | · |, it follows that for any
real x ∈ R, |x| ≥ x: indeed if x ≥ 0, then |x| = x, while if x < 0, then −x > 0, and so
|x| = −x > 0 > x. From (1.3), we also have | − x| = | − 1 · x| = | − 1||x| = 1|x| = |x|, for
all x ∈ R, and so it follows that |x| = | − x| ≥ −x for all x ∈ R. We have the following
cases:

x + y ≥ 0. Then |x + y| = x + y. As |x| ≥ x and |y| ≥ y, we obtain |x| + |y| ≥ x + y =

|x + y|.

x + y < 0. Then |x + y| = −(x + y). Since |x| ≥ −x and |y| ≥ −y, it follows that
|x| + |y| ≥ −x + (−y) = −(x + y) = |x + y|.

This proves (1.4).

8
 1.2 Sequences and limits

1.2 Sequences and limits

1.2.1 Sequences
The notion of a sequence occurs in ordinary conversation. An example is the phrase “an
unfortunate sequence of events”. In this case, we envision one event causing another,
which in turn causes another event and so on. We can identify a first event, a second
event, etcetera.
A sequence of real numbers is a list

a1 , a 2 , a 3 , . . .

of real numbers, where there is the first number (namely a1 ), the second number (namely
a2 ), and so on. For example,
1 1
1, , , . . .
2 3
is a sequence of real numbers. The first number is 1, the second number is 12 and so on.
(There may not be a connection between the numbers appearing in a sequence.) If we
think of a1 as f (1), a2 as f (2), and so on, then it becomes clear that a sequence of real
numbers is a special type of function, namely one with domain N and co-domain R. This
leads to the following formal definition.

Definition. A sequence of real numbers is a function f : N → R.

Only the notation is somewhat unusual. Instead of writing f (n) for the value of f at a
natural number n, we write an . The entire sequence is then written in any one of the
following ways:
(an )n∈N , (an )∞
n=1 , (an )n≥1 , (an ).

In (an )∞
n=1 , the ∞ symbol indicates that the assignment process 1 7→ a1 , 2 7→ a2 , . . .
continues indefinitely. In these notes, we shall normally use the notation (an )n∈N . In
general, the terms of a sequence need not be real numbers, but in this guide we shall
only be dealing with sequences whose entries are real numbers, so we shall simply refer
to them as sequences from now on.
The nth term an of a sequence may be defined explicitly by a formula involving n, as
in the example given above:
1
an = , n ∈ N.
n
It might also sometimes be defined recursively. For example,
n
a1 = 1, an+1 = an for n ∈ N.
n+1
(Write down the first few terms of this sequence.)

9
1 Analysis

Examples.
 
1
1. n n∈N
is a sequence with the nth term given by n1 , for n ∈ N. This is the sequence

1 1
1, , , ....
2 3
 
2. 1 + n1 is a sequence with the nth term given by 1 + n1 , for n ∈ N. This is the
n∈N
sequence
3 4 5 6 7
2, , , , , , . . . .
2 3 4 5 6
    
3. (−1)n 1 + n1 is a sequence with the nth term given by (−1)n 1 + 1
n
, for n ∈ N.
n∈N
This is the sequence
3 4 5 6 7
−2, , − , , − , , . . . .
2 3 4 5 6
4. ((−1)n )n∈N is a sequence with the nth term given by (−1)n , for n ∈ N. This sequence
is simply
−1, 1, −1, 1, −1, 1, . . .
with the nth term equal to −1 if n is odd, and 1 if n is even.

5. (1)n∈N is a sequence with the nth term given by 1, for n ∈ N. This is the constant
sequence
1, 1, 1, . . . .

6. (n)n∈N is a sequence with the nth term given by n, for n ∈ N. This is the strictly
increasing sequence
1, 2, 3, . . . .
 
1
7. 11
+ 212 + 313 + · · · + n1n is a sequence with the nth term given by 1
11
+ 1
22
+ 1
33
+
n∈N
· · · + n1n , for n ∈ N. This is the sequence of ‘partial sums’
1 1 1 1 1 1
1
, 1 + 2, 1 + 2 + 3, ....
1 1 2 1 2 3

1.2.2 Limit of a convergent sequence

A
 sequence
 can be graphed. For instance, the first 7 points of the graph of the sequence
1
n
are displayed in Figure 1.2. This portion of the graph suggests that the terms of
n∈N  
the sequence n1 “tend toward 0” as n increases. This is consistent with the idea of
n∈N
convergence that you might have encountered before: a sequence (an )n∈N converges to
some real number L, if the terms an get “closer and closer” to L as n “increases without
bound”. Symbolically, this is represented using the notation

lim an = L,
n→∞

10
 1.2 Sequences and limits

1
0.5
0
0 1 2 3 4 5 6 7

 
1
Figure 1.2: First 7 points of the graph of the sequence n n∈N
.

where L denotes the limit of the sequence. If there is no such finite number L to which
the terms of the sequence get arbitrarily close, then the sequence is said to diverge.
The problem with this characterization is its imprecision. Exactly what does it mean for
the terms of a sequence to get “closer and closer”, or “as close as we like”, or “arbitrarily
close” to some number L? Even if we accept this apparent ambiguity, how would one use
the definition given in the preceding paragraph to prove theorems that involve sequences?
Since sequences are used throughout analysis, the concepts of their convergence and
divergence must be carefully defined.

2
1.5
1
0.5
0
0 1 2 3 4 5 6 7

 
1
Figure 1.3: First 7 points of the graph of the sequence 1 + n n∈N
.
 
For example, the terms of 1 + n1 get “closer and closer” to 0 (indeed the distance
n∈N
to 0 keeps decreasing),
 but
 its
limit is 1. See Figure 1.3.
n 1
The terms of (−1) 1 + n get “as close as we like” or “arbitrarily close” to 1,
n∈N
but the sequence has no limit. See Figure 1.4.

Definition. The sequence (an )n∈N is said to converge to L if for every real number
ε > 0, there exists an N ∈ N (possibly depending on ε) such that for all n > N ,

|an − L| < ε.

Then we say that (an )n∈N is convergent with limit L and write

lim an = L.
n→∞

11
1 Analysis

0
1 2 3 4 5 6 7 8
−1

−2

  
1
Figure 1.4: First eight points of the graph of the sequence (−1)n 1 + n
.
n∈N

lim an = L, then the sequence (an )n∈N is

If there does not exist a number L such that n→∞
said to be divergent.
Note that |an − L| < ε iff an ∈ (L − ε, L + ε). Hence pictorially, for a convergent
sequence with limit L, this definition means the following, as illustrated in Figure 1.5:
Pick any ε > 0, and consider the shaded strip of width ε around the horizontal line
passing through L. Then one can find an N ∈ N, large enough, such that all the terms
an of the sequence, for n > N , lie in the shaded strip.

L+ε

L−ε

... ...
N N +1 N +2 N +3

Figure 1.5: Convergence of a sequence with limit L.

Examples.
 
1
1. Use the definition of the limit of a sequence to show that n n∈N
is a convergent
sequence with limit 0.
Solution. Given ε > 0, we need to find N such that, for all n > N ,
1 1
 

|an − L| =  − 0 = < ε.
n n

12
 1.2 Sequences and limits

If we choose N ∈ N such that N > 1ε (such an N exists by the Archimedean property!),

then for n > N (⇔ n1 < N1 ), we have
1 1 1
 

|an − L| =  − 0 = < < ε.

n n N
1
Hence lim = 0.
n→∞ n
 
1
2. Use the definition of the limit of a sequence to show that 1 + 2n2 −n
is a convergent
n∈N
sequence with limit 1.
Solution. Given ε > 0, we need to find N such that for all n > N ,
1 1
 
 
|an − L| = 1 + − 1 = 2 < ε.

2n − n
2 2n − n
We do not need to be very precise in our choice of ε: if we note that, for all n ∈ N,
2n2 − n ≥ n, then we see that we may choose N ∈ N such that N > 1ε . So, for n > N ,
we have
1 1 1
|an − L| = 2 ≤ < < ε.
2n − n n N
1
 
lim 1 + 2
Hence n→∞ = 1.
2n − n
  
1
3. Use the definition to show that (−1)n 1 + n
is a divergent sequence.
n∈N
 
Solution. The idea is as follows. If L ≥ 0, then for each odd n, (−1)n 1 + n1 < −1,
so these terms are not “close to L” and L is not a limit of the sequence. Similarly,
if L < 0, then the even terms of the sequence are not close to L, and so again L is
not the limit of the sequence. We now proceed to make this informal argument into a
formal proof.
 
1
Let an = (−1)n 1 + n
for n ∈ N.
In order to prove that
(an )n∈N is divergent,
we have to show that
¬ [(an )n∈N is convergent] ,
that is,

¬ [∃L ∈ R such that ∀ε > 0 ∃N ∈ N such that ∀n > N, |an − L| < ε] ,

that is,

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

Let L ∈ R. Now we need to prove

∃ε > 0 such that ∀N ∈ N ∃n > N such that |an − L| ≥ ε.

13
1 Analysis

Fix ε = 1. (It is not obvious that ε = 1 will work, but it has been found by trial and
error.) Now we will show that

∀N ∈ N ∃n > N such that |an − L| ≥ ε.

So let N ∈ N be given. If L ≥ 0, then choose n to be any odd number > N . Then we

have
1 1 1
     
(−1)n 1 +
   
|an − L| = − L = −1 − − L = 1 + + L > 1 = ε.

n n n
If L < 0, then choose n to be any even number > N . Then we have
1 1 1
     
(−1)n 1 +
   
|an − L| = − L = 1 + − L = 1 + − L > 1 = ε.

n n n
Thus we have shown that for all L ∈ R, there exists a ε > 0 (namely ε = 1) such
that for all N ∈ N, there exists a n > N (namely any odd number > N if L ≥ 0,
and any even number > N if L < 0) such that |an − L| ≥ ε. Thus the sequence is
divergent.
The notation n→∞lim an suggests that the limit of a convergent sequence is unique. Indeed
this is the case, and we prove this below.
Theorem 1.2.1. A convergent sequence has a unique limit.
Proof. The statement means that both the following are true.
(i) A convergent sequence has a limit;
(ii) A convergent sequence cannot have two different limits.
Here, (i) is true by the definition of convergence, so we only have to prove (ii). Accordingly,
consider a convergent sequence (an )n∈N and suppose that it has two limits L1 and L2 ,
where L1 6= L2 . Let
|L1 − L2 |
ε= > 0,
2
where the positivity of the ε defined above follows from the fact that L1 6= L2 . Since L1
is a limit of the sequence (an )n∈N , ∃N1 ∈ N such that for all n > N1 ,

|an − L1 | < ε.

Since L2 is a limit of (an )n∈N , ∃N2 ∈ N such that for all n > N2 ,

|an − L2 | < ε.

Consequently for n > max{N1 , N2 },

2ε = |L1 − L2 | = |L1 − an + an − L2 |
≤ |L1 − an | + |an − L2 | = |an − L1 | + |an − L2 |
< ε + ε = 2ε,

a contradiction. Therefore the sequence (an )n∈N does not have two different limits, as
required.

14
 1.2 Sequences and limits

1.2.3 The sequence (xn )n∈N .

In this section, we prove that, whenever |x| < 1, the sequence (xn )n∈N is convergent, with
limit 0. This is a basic, and perhaps “obvious” result, but we need to establish it “from
first principles”.
We proceed by first proving a useful result known as Bernoulli’s Inequality.
Theorem 1.2.2. For all real x ≥ −1 and all n ∈ N,
(1 + x)n ≥ 1 + nx.
Proof. We prove this result by induction on n. Note first that, for n = 1, the inequality
states that 1 + x ≥ 1 + x, which is certainly true.
Suppose now that, for some n ∈ N, (1 + x)n ≥ 1 + nx. Now we have (1 + x)n+1 =
(1 + x)(1 + x)n ≥ (1 + x)(1 + nx) = 1 + nx + x + nx2 ≥ 1 + (n + 1)x. So the inequality
holds for n + 1. Hence, by induction, the inequality holds for all n ∈ N.
Now we use Bernoulli’s Inequality to show that, whenever |x| < 1, (xn )n∈N os convergent
with limit 0.
Theorem 1.2.3. Let x be any real number with −1 < x < 1. Then (xn )n∈N is a
convergent sequence with limit 0.
Proof. First note that, if x = 0, then xn = 0 for every n ∈ N, and so certainly
1
limn→∞ xn = 0. So we may assume that x 6= 0. In this case, we have that 1 < |x| , and so
1
h = |x| − 1 > 0. Now, by Bernoulli’s Inequality, Theorem 1.2.2, we have
1
= (1 + h)n ≥ 1 + nh > nh,
|x|n
and so
1
0 ≤ |x|n ≤ .
nh
1
Now we use the definition of convergence. Given any ε > 0, we take N ≥ εh
. Now, for
n > N,
1 1
|xn − 0| = |x|n ≤ < ≤ ε.
nh Nh
Hence limn→∞ xn = 0.
Example.
1
Use Bernoulli’s Inequality to show that lim 2 n = 1.
n→∞
1 1
Solution. 2 > 1 and so 2 n > 1 (for otherwise 2 = (2 n )n ≤ 1, a contradiction). Let
1
an := 2 n − 1 ≥ 0. Then 2 = (1 + an )n ≥ 1 + nan Hence
1
0 ≤ an ≤ .
n
Now, given any ε > 0, choose N ≥ 1/ε. Then, for n > N ,
1 1 1
|2 n − 1| = |an | = an ≤ < ≤ ε.
n N
√ √ √ √
Therefore the sequence 2, 2, 3 2, 4 2, 5 2, . . . is convergent with limit 1.

15
1 Analysis

1.2.4 Bounded and monotone sequences

It is cumbersome to check from the definition if a sequence is convergent or not. In
this section, we will study a condition under which we can conclude that a sequence
is convergent even without knowing its limit! We will prove that if a sequence is both
‘bounded’ as well as ‘monotone’, then it is always convergent.

Definition. A sequence (an )n∈N is said to be bounded if there exists a real number
M > 0 such that
for all n ∈ N, |an | ≤ M. (1.5)

Note that a sequence is bounded iff the set S = {an | n ∈ N} is bounded. (See
Exercise 15.)

Examples.

a) (1)n∈N is bounded, since |1| = 1 ≤ 1 for all n ∈ N.

   
1
b) n n∈N
is bounded, since  n1  = 1
n
≤ 1 for all n ∈ N.
   
1
c) 1 + is bounded, since 1 + n1  = 1 + 1
≤ 2 for all n ∈ N.
 
n n∈N n
     
1 1 1
d) (−1)n 1 + is bounded, since (−1)n 1 + =1+ ≤ 2 for all n ∈ N.
 
n n n

n∈N

e) Show that the sequence (an )n∈N defined by

1 1 1 1
an = 1
+ 2 + 3 + ··· + n, n ∈ N
1 2 3 n
is bounded.
Solution. We have to find an upper bound of |an | that is independent of n:

|an | = an
1 1 1 1
= 1 + 2 + 3 + ··· + n
1 2 3 n
1 1 1 1
< 1 + 2 + 3 + ··· + n
1 2 2  2
1 1 1 1 1 1 1
   
= 1+ 1− + 2 1− + · · · + n−1 1 −
1 2 2 2 2 2 2
1 1 1 1 1 1
= 1 + − 2 + 2 − 3 + − · · · + n−1 − n
2 2 2 2 2 2
1 1
=1+ − n
2 2
3
< .
2
(Write down a detailed proof using induction on n.) Thus all the terms are bounded
by 32 , and so the sequence is bounded.

16
 1.2 Sequences and limits

f) Show that the sequence (an )n∈N given by an = n for n ∈ N, is not bounded.
Solution. Given any M > 0, there exists an N ∈ N such that M < N (Archimedean
property with y = M and x = 1). Thus

¬[∃M > 0 such that for all n ∈ N, |an | = |n| = n ≤ M ],

and so (n)n∈N is not bounded.

   
The sequences (1)n∈N , n1 , 1 + n1 are all convergent, and we have shown
n∈N n∈N
above that these are also bounded. This is not a coincidence, and in the next theorem
we show that the set of all convergent sequences is contained in the set of all bounded
sequences. See Figure 1.6.

convergent bounded
sequences sequences

Figure 1.6: All convergent sequences are bounded.

Theorem 1.2.4. If a sequence is convergent, then it is bounded.

Proof. Let (an )n∈N be a convergent sequence with limit L. Set ε = 1. Then, using the
definition of convergence for this value of ε, we see that there exists N ∈ N such that,
for all n > N ,
|an − L| < 1.
Hence for all n > N ,

|an | = |an − L + L| ≤ |an − L| + |L| < 1 + |L|.

Let M = max{|a1 |, . . . , |aN |, 1 + |L|}. Then for all n ∈ N

|an | ≤ M

and so (an )n∈N is bounded.

The above theorem can be used to prove the divergence of sequences. Indeed, in
contrapositive form it asserts that unbounded sequences are not convergent. Thus, one
way to prove that a sequence is divergent, would be to try to prove that it is unbounded.
Example. Show that the sequence (n)n∈N is divergent.
Solution. We have seen above that (n)n∈N is unbounded. It follows from Theorem 1.2.4
that (n)n∈N is not convergent.

17
1 Analysis

Definitions. A sequence (an )n∈N is said to be

monotonically increasing (or simply increasing) ‘if for all n ∈ N, an ≤ an+1 ,

strictly increasing if for all n ∈ N, an < an+1 ,

monotonically decreasing (or simply decreasing) if for all n ∈ N, an ≥ an+1 ,

strictly decreasing if for all n ∈ N, an > an+1 ,

monotone if it is either monotonically increasing or monotonically decreasing.

Thus a sequence (an )n∈N is monotonically increasing if

a1 ≤ a2 ≤ a3 ≤ . . . ,

strictly increasing if
a1 < a2 < a3 < . . . ,
monotonically decreasing if
a1 ≥ a2 ≥ a3 ≥ . . . ,
and strictly decreasing if
a1 > a2 > a3 > . . . .

Examples.
monotonically strictly monotonically strictly
Sequence monotone?
increasing? increasing? decreasing? decreasing?
 
1
n n∈N No No Yes Yes Yes
 
1 + n1 No No Yes Yes Yes
n∈N
  
(−1) 1 + n1
n No No No No No
n∈N
(1)n∈N Yes No Yes No Yes
(n)n∈N Yes Yes No No Yes
 
1 1 1 1
11
+ 22
+ 33
+ ··· + nn n∈N Yes Yes No No Yes

The following theorem can be useful in showing that sequences converge when one
does not know the limit beforehand.

Theorem 1.2.5. If a sequence is monotone and bounded, then it is convergent.

Proof. Let (an )n∈N be a monotonically increasing sequence. Since (an )n∈N is bounded,
it follows that the set
S = {an | n ∈ N}
has an upper bound and so sup S exists. We show that in fact (an )n∈N converges to
sup S.

18
 1.2 Sequences and limits

Let ε > 0 be given. Since sup S − ε < sup S, it follows that sup S − ε is not an upper
bound for S and so ∃aN ∈ S such that sup S − ε < aN , that is

sup S − aN < ε.

Since (an )n∈N is a monotonically increasing sequence, for n > N , we have aN ≤ an . Since
sup S is an upper bound for S, an ≤ sup S and so |an − sup S| = sup S − an . Thus for
n > N we obtain

|an − sup S| = sup S − an ≤ sup S − aN < ε.

Hence (an )n∈N is a convergent sequence with limit sup S, as required.

On the other hand, if (an )n∈N is a monotonically decreasing sequence, then clearly
(−an )n∈N is a monotonically increasing sequence. Furthermore if (an )n∈N is bounded,
then (−an )n∈N is bounded as well (| − an | = |an | ≤ M ). Hence by the case considered
above, it follows that (−an )n∈N is a convergent sequence with limit

sup {−an | n ∈ N} = − inf {an | n ∈ N} = − inf S,

(see Exercise 6). So given ε > 0, ∃N ∈ N such that for all n > N , | − an − (− inf S)| < ε,
that is, |an − inf S| < ε. Thus (an )n∈N is convergent with limit inf S.

Examples.

1. We have shown that the sequence (an )n∈N defined by

1 1 1 1
an = 1
+ 2 + 3 + ··· + n, n ∈ N
1 2 3 n
is monotone (indeed, it is strictly increasing since
1
an+1 − an = >0
(n + 1)(n+1)

for all n ∈ N) and bounded (see Example e on page 16). Thus it follows from Theorem
1.2.5 that this sequence2 is convergent.

2
Although it is known that this sequence is convergent to some limit L ∈ R, it is so far not even
known if the limit L is rational or irrational, and this is still an open problem in mathematics! Also
∞ Z 1
X 1 1
associated with this sequence is the interesting identity n
= x
dx, the proof of which is
n=1
n 0 x
beyond the scope of this course.

19
1 Analysis

2. The following table gives a summary of the valid implications, and gives counterexam-
ples to implications which are not true.
Question Answer Reason/Counterexample

Is every convergent sequence bounded? Yes Theorem 1.2.4

Is every bounded sequence convergent? No ((−1)n )n∈N is bounded,
but not convergent.
(−1)n
 
Is every convergent sequence monotone? No n is convergent, but not
n∈N
1 1
monotone: −1 < 2 and 2 > − 13 .
Is every monotone sequence convergent? No (n)n∈N is not convergent.
Is every bounded AND monotone Yes Theorem 1.2.5
sequence convergent?

1.2.5 Algebra of limits

In this section we will learn that if we ‘algebraically’ combine the terms of convergent
sequences, then the new sequence which is obtained is again convergent, and moreover the
limit of this sequence is the same algebraic combination of the limits. In this manner we
can sometimes prove the convergence of complicated sequences by breaking them down
and writing them as an algebraic combination of simple sequences. Thus, we conveniently
apply arithmetic rules to compute the limits of sequences if the terms are the sum,
product, quotient of terms of simpler sequences with a known limit. For instance, using
the formal definition of a limit, one can show that the sequence (an )n∈N defined by

4n2 + 9
an =
3n2 + 7n + 11
converges to 43 . However, it is simpler to observe that
 
9
n2 4 + n2 4 + n92
an = = ,
3 + n7 + n112
 
7 11
n2 3 + n
+ n2

where the terms n92 , n7 , n112 all have limit 0, and by a repeated application of Theorem
1.2.6 given below, we obtain that
9 9
 
lim 4 + 2 lim 4 + lim 2 4+0 4
lim an =
n→∞ n  n→∞ n→∞ n
7 11 = 7 11 = 3 + 0 + 0 = 3 .

n→∞
lim 3 + + 2 lim 3 + n→∞
lim + n→∞
lim 2
n→∞ n n n→∞ n n

20
 1.2 Sequences and limits

Theorem 1.2.6. If (an )n∈N and (bn )n∈N are convergent sequences, then the following
hold:
1. For all α ∈ R, (αan )n∈N is a convergent sequence and n→∞
lim αan = α n→∞
lim an .
 
 
2. (|an |)n∈N is a convergent sequence and lim |an | =  lim an .
n→∞ n→∞

3. (an + bn )n∈N is a convergent sequence and n→∞

lim (an + bn ) = n→∞
lim an + n→∞
lim bn .
  
4. (an bn )n∈N is a convergent sequence and lim an bn = lim an lim bn .
n→∞ n→∞ n→∞

 k
5. For all k ∈ N, (akn )n∈N is a convergent sequence and lim ak = lim an .
n→∞ n n→∞

1
 
6. If for all n ∈ N, bn 6= 0 and lim bn 6= 0, then is convergent and moreover,
n→∞ bn n∈N
1 1
lim = .
n→∞ b
n lim
n→∞
b n

7. For all k ∈ N, (an+k )n∈N is convergent and lim an+k = lim an .

n→∞ n→∞
√  √ q
8. If for all n ∈ N, an ≥ 0, then an is convergent and lim an = lim an .
n∈N n→∞ n→∞

Proof. Let (an )n∈N and (bn )n∈N converge to La and Lb , respectively.
1. Let ε > 0 be given. By the definition of limn→∞ an = La , there exists an N ∈ N such
that for all n > N ,
ε
|an − La | < .
|α| + 1
(We added 1 to |α| in the denominator to deal with the possibility that α = 0.) Then
ε
|αan − αLa | = |α| |an − La | ≤ |α| < ε,
|α| + 1
and (again by the definition of a convergent sequence) (αan )n∈N is convergent with
limit αLa , that is,
lim αan = αLa = α lim an .
n→∞ n→∞

2. Given ε > 0, let N ∈ N be such that for all n > N ,

|an − La | < ε.

Then (by Exercise 14) we have for all n > N :

||an | − |La || ≤ |an − La | < ε.

Hence (|an |)n∈N is convergent with limit |La |, that is,

 
 
lim |an | = |La | =  lim an  .
n→∞ n→∞

21
1 Analysis

3. Given ε > 0, let N1 ∈ N be such that for all n > N1 ,

ε
|an − La | < .
2
Let N2 ∈ N be such that for all n > N2 ,
ε
|bn − Lb | < .
2
Then for all n > N := max{N1 , N2 }, we have by the triangle inequality:
ε ε
|an + bn − (La + Lb )| = |an − La + bn − Lb | ≤ |an − La | + |bn − Lb | < + = ε.
2 2
Hence (an + bn )n∈N is convergent with limit La + Lb , that is,
lim (an + bn ) = La + Lb = lim an + lim bn .
n→∞ n→∞ n→∞

4. Note that
|an bn − La Lb | = |an bn − La bn + La bn − La Lb |
≤ |an bn − La bn | + |La bn − La Lb |
= |an − La | |bn | + |La | |bn − Lb |. (1.6)
Given ε > 0, we need to find N such that for all n > N ,
|an bn − La Lb | < ε.
This can be achieved by finding an N such that each of the summands in (1.6) is less
than 2ε for n > N . This can be done as follows.
Step 1. Since (bn )n∈N is convergent, by Theorem 1.2.4 it follows that it is bounded:
∃M > 0 such that for all n ∈ N, |bn | ≤ M . Let N1 ∈ N be such that, for all n > N1 ,
ε
|an − La | < .
2M
Step 2. Let N2 ∈ N be such that for all n > N2 ,
ε
|bn − Lb | < .
2|La | + 1
(Note that it is possible that La = 0.) Thus, for n > N := max{N1 , N2 }, we have
|an bn − La Lb | ≤ |an − La | |bn | + |La | |bn − Lb |
ε ε
< M + |La |
2M 2|La | + 1
ε ε
< +
2 2
= ε.
It follows that (an bn )n∈N is a convergent sequence with limit La Lb , that is,
  
lim an bn = La Lb = lim an lim bn .
n→∞ n→∞ n→∞

22
 1.2 Sequences and limits

5. This can be shown by using induction on k and from part 4 above. It is trivially true
with k = 1. Suppose that it holds for some k, then (akn )n∈N is convergent and
 k
lim ak = lim an .
n→∞ n n→∞

Hence by part 4 above applied to the sequences (an )n∈N and (akn )n∈N , we obtain that
the sequence (an · akn )n∈N is convergent and
     k  k+1
lim an akn = lim an lim akn = lim an lim an = lim an .
n→∞ n→∞ n→∞ n→∞ n→∞ n→∞

Thus (ak+1
n )n∈N is convergent and
 k+1
lim ak+1 = lim an .
n→∞ n n→∞

6. Let N1 ∈ N be such that, for all n > N1 ,

|Lb |
|bn − Lb | < .
2
Thus, for all n > N1 ,
|Lb |
|Lb | − |bn | ≤ ||Lb | − |bn || ≤ |bn − Lb | < ,
2
|Lb |
and so |bn | ≥ 2
. Let ε > 0, and let N2 ∈ N be such that for all n > N2 ,
ε|Lb |2
|bn − Lb | < .
2
Hence for n > N := max{N1 , N2 }, we have
1 1  |bn − Lb | ε|Lb |2 2 1
 

 −  = < = ε.

bn Lb |bn | |Lb | 2 |Lb | |Lb |
1 1 1 1
 
So is convergent and n→∞
lim = = .
bn n∈N bn Lb lim bn
n→∞

7. Exercise 38.
8. Since an ≥ 0 for each n, we have L ≥ 0 (see Exercise 24). The case L = 0 is left as an
exercise.
In the case L > 0, we use a technique called “rationalising the numerator”: note that
√ √
√ √ √ √ an + L a −L
an − L = ( an − L) √ √ =√ n √ .
an + L an + L
√
Now, given ε > 0, choose N ∈ N so that, for n > N , |an − L| < ε L. Then we have,
for n > N , √
√ √ |an − L| ε L
| an − L| = √ √ ≤ √ = ε.
an + L L
√ √
Thus ( an )n∈N is convergent, with limit L,

23
1 Analysis

Examples. 1. Determine whether the following sequence is convergent and find its limit.

!
n2 − 24n3 + 3n4 − 12
1 + 7n + 21n4 n∈N

1
Solution. By Example 1 on page 12 we know that lim = 0. We now use Theo-
n→∞ n
rem 1.2.6, we obtain
1 24
n2 − 24n3 + 3n4 − 12 n4 n2
− n
+ 3 − n124
lim = lim · 1
n→∞ 1 + 7n + 21n4 n→∞ n4
n4
+ n73 + 21
 2  4
lim 1
n→∞ n
lim n1 + 3 − 12 n→∞
− 24 n→∞ lim n1
=  4  3
1 1
lim + 7 lim n + 21
n→∞ n n→∞

02 − 24 · 0 + 3 − 12 · 04
=
04 + 7 · 03 + 21
3 1
= = .
21 7

2. Determine whether the following sequence is convergent and find its limit.
 n
2 + 3n + 1


3n+1 + 3 n∈N

Solution. Divide top and bottom by the fastest growing term appearing, which is
3n , and use that limn→∞ xn = 0 for |x| < 1:

2n + 3n + 1 (2/3)n + 1 + (1/3)n
lim =
n→∞ 3n+1 + 3 3 + 3(1/3)n
limn→∞ (2/3)n + limn→∞ 1 + limn→∞ (1/3)n
=
limn→∞ 3 + 3 limn→∞ (1/3)n
0+1+0 1
= = .
3+0 3

Remark. It follows from Theorem 1.2.6.3 that if we have three convergent sequences
(an )n∈N , (bn )n∈N , (cn )n∈N , then their sum (an + bn + cn )n∈N is also convergent with limit

lim (an + bn + cn ) = n→∞

n→∞
lim an + n→∞
lim bn + n→∞
lim cn .

24
 1.2 Sequences and limits

This is also true for the sum of four convergent sequences, the sum of five convergent
sequences, and by an easy induction proof, the sum of any number of convergent sequences.
However, the following reasoning is clearly incorrect (because it would show that 1 = 0):
1 1 1
1= + + ··· +
|n n {z n}
n terms

and therefore,

1 1 1
 
1 = n→∞
lim 1 = n→∞
lim + + ··· +
n n n
! 1 1 1
= lim + lim + · · · + lim
n→∞ n n→∞ n n→∞ n
| {z }
n limits
= |0 + 0 +{z· · · + 0}
n terms
= 0.

The incorrect step is marked with a ‘!’. The problem here is that the number of terms is
not a fixed value, but depends on n, and therefore Theorem 1.2.6.3 cannot be applied.

1.2.6 The Sandwich Theorem

Another theorem that is useful in proving that sequences are convergent and in determining
their limits is the so-called Sandwich theorem. Roughly speaking, it says that if a sequence
is sandwiched between two convergent limits with the same limit, then the sandwiched
sequence is also convergent with the same limit.
Theorem 1.2.7 (Sandwich theorem). Let (an )n∈N , (bn )n∈N be convergent sequences with
the same limit, that is,
lim an = lim bn .
n→∞ n→∞

If (cn )n∈N is a third sequence such that

for all n ∈ N, an ≤ cn ≤ bn ,

then (cn )n∈N is also convergent with the same limit, that is,

lim an = n→∞
n→∞
lim cn = n→∞
lim bn .

Proof. Let L denote the common limit of (an )n∈N and (bn )n∈N :

lim an = L = lim bn .
n→∞ n→∞

Given ε > 0, let N1 ∈ N be such that for all n > N1 , |an − L| < ε. Hence for n > N1 ,

L − an ≤ |L − an | = |an − L| < ε,

25
1 Analysis

and so L − an < ε, that is,

L − ε < an .
Still for the same given ε > 0, let N2 ∈ N be such that for all n > N2 , |bn − L| < ε. So
for n > N2 , bn − L < ε, that is,
bn < L + ε.
Thus for n > N := max{N1 , N2 }, we have
L − ε < an ≤ cn ≤ bn < L + ε,
and so L − ε < cn < L + ε. Consequently, cn − L < ε and −(cn − L) < ε, and so
|cn − L| < ε.
This proves that (cn )n∈N is convergent with limit L.

Examples.
n
1. Use the Sandwich theorem to show that lim = 0.
n→∞ 10n
Solution. It can be shown by induction that for all n ∈ N, n2 < 10n .
Consequently, we have
n n 1
0≤ n
≤ 2 = .
10 n n
1
Since n→∞
lim 0 = 0 = n→∞
lim , from the Sandwich theorem it follows that the sequence
  n
n
10n is convergent and
n∈N
n
lim n = 0.
n→∞ 10
1 2 3 4
Thus the sequence 10 , 100 , 1000 , 10000 , . . . is convergent with limit 0.
1
2. Use the Sandwich theorem to show that for any a, b ∈ R, lim (|a|n + |b|n ) n =
n→∞
max{|a|, |b|}.

Solution. Clearly,
(max{|a|, |b|})n ≤ |a|n + |b|n ≤ (max{|a|, |b|})n + (max{|a|, |b|})n
and so
1 1
max{|a|, |b|} ≤ (|a|n + |b|n ) n ≤ 2 n max{|a|, |b|}.
1
lim (|a|n + |b|n ) n = max{|a|, |b|}.
So using the Sandwich theorem, it follows that n→∞
1
In particular, with a = 24 and b = 2005, we have that lim (24n + 2005n ) n = 2005,
n→∞
that is, the sequence
2029, 2005.1436, 2005.001146260873, . . .
is convergent with limit 2005.

26
 1.2 Sequences and limits

n n n
 
3. Show that n→∞
lim 2
+ 2 + ··· + 2 = 1.
n +1 n +2 n +n
n n
Solution. There are n terms in the sum: the smallest is n2 +n
and the largest is n2 +1
.
Thus, for all n ∈ N, we have

n2 n n n n2
≤ + + · · · + ≤ ,
n2 + n n2 + 1 n2 + 2 n2 + n n2 + 1
and since
n2 n2
lim = 1 = lim 2 .
n→∞ n2 + n n→∞ n + 1

it follows from the Sandwich theorem that

n n n
 
lim 2
+ 2 + ··· + 2 = 1.
n→∞ n +1 n +2 n +n

1.2.7 Subsequences
In this section we prove an important result in analysis, known as the Bolzano-Weierstrass
theorem, which says that every bounded sequence has a convergent ‘subsequence’. We
begin this section by defining what we mean by a subsequence of a sequence.

Definition. Let (an )n∈N be a sequence and let (nk )k∈N be a strictly increasing sequence
of natural numbers. Then (ank )k∈N is called a subsequence of (an )n∈N .

Examples.
1 1 1 1 1
         
1. , , and are all subsequences of .
2n n∈N n2 n∈N n! n∈N nn n∈N n n∈N

2. Let pn be the nth prime number. (Thus p1 = 2, p2 = 3, p3 

= 5, p4 = 7, etc.) Then the
1 1
sequence (an )n∈N defined by an = is a subsequence of .
pn n n∈N
3. The sequence
1 1 1
, 1, , , . . .
2 3 4
1
 
is not a subsequence of .
n n∈N

4. The sequence ((−1)2n )n∈N , that is, the constant sequence

1, 1, 1, . . .

and the sequence ((−1)2n−1 )n∈N , that is, the constant sequence

−1, −1, −1, . . .

are both subsequences of ((−1)n )n∈N .

27
1 Analysis

Theorem 1.2.8. If (an )n∈N is a convergent sequence with limit L, then any subsequence
of (an )n∈N is also convergent with the limit L.

Proof. Let (ank )k∈N be a subsequence of (an )n∈N . Given ε > 0, let N ∈ N be such that
for all n > N , |an − L| < ε. Since the sequence n1 < n2 < n3 < . . . is not bounded
above, it follows that there exists a K ∈ N such that nK > N . Then for all k > K,
nk > nK > N . Hence for k > K, |ank − L| < ε, and so (ank )k∈N is convergent with limit
L.

Examples.
       
1 1 1 1
1. 2n n∈N
, n2 n∈N
, n! n∈N
and nn n∈N
are convergent sequences with limit 0.

2. The sequence ((−1)n )n∈N is divergent since the subsequence 1, 1, 1, . . . has limit 1,
while the subsequence −1, −1, −1, . . . has limit −1.

The following theorem is a very important step in the proof of the Bolzano-Weierstrass
theorem. It gives us a fact about arbitrary sequences of real numbers.

Theorem 1.2.9. Every sequence has a monotone subsequence.

Since we do not assume anything about the sequence, other than that it is a sequence
of real numbers, there is virtually no obvious way to start the proof. Nevertheless, even
though it is very difficult to discover such a proof, it is not so difficult to understand it,
once found.
Before doing the formal proof, we first illustrate the idea behind it. Let (an )n∈N be the
given sequence. Imagine that an is the height of a hotel with number n, which is followed
by hotel n + 1, and so on, along an infinite line, where the sea is in the far distance. A
hotel is said to have the seaview property if it is higher than all hotels following it.
See Figure 1.7.

...

...
1 2 3 4 5 6

Figure 1.7: The seaview property.

Now there are only two possibilities:

First possibility. There are infinitely many hotels with the seaview property. Then their
heights form a strictly decreasing subsequence.

28
 1.2 Sequences and limits

Second possibility. There is only a finite number of hotels with the seaview property.
Then after the last hotel with the seaview property, one can start with any hotel
and then always find one that is at least as high, which is taken as the next hotel,
and then finding yet another that is at least as high as that one, and so on. The
heights of these hotels form a monotonically increasing subsequence.

Proof of Theorem 1.2.9. Let

S = {m ∈ N | for all n > m, an < am } ,

that is, S is the set of all natural numbers m such that an < am for all n > m. Then we
have the following two cases.

S is infinite. Arrange the elements of S in strictly increasing order: n1 < n2 < n3 < . . . .
Then (ank )k∈N is a strictly decreasing subsequence of (an )n∈N .

S is finite. If S is empty, then define n1 = 1, and otherwise let n1 = max S + 1. Then

since n1 is larger than any element of S, n1 ∈ / S, which means that there exists
an n2 > n1 such that an2 ≥ an1 . Again, n2 is larger than any element of S, thus
n2 ∈/ S, which means there exists an n3 > n2 such that an3 ≥ an2 . Continuing
in this way, for any k ∈ N we choose an nk+1 > nk such that ank+1 ≥ ank . Then
(ank )k∈N is a monotonically increasing subsequence of (an )n∈N .

The Bolzano-Weierstrass theorem is now a simple consequence of the above theorem.

Theorem 1.2.10. (Bolzano3 -Weierstrass4 theorem.) Every bounded sequence has a

convergent subsequence.

Proof. Let (an )n∈N be a bounded sequence. Then there exists M > 0 such that for all
n ∈ N, |an | ≤ M . From Theorem 1.2.9 above, it follows that the sequence (an )n∈N has
a monotone subsequence (ank )k∈N . Then, clearly, for all k ∈ N, |ank | ≤ M and so the
sequence (ank )k∈N is also bounded. Since (ank )k∈N is monotone and bounded, it follows
from Theorem 1.2.5 that it is convergent.
√
Example. Consider the sequence (an )n∈N of fractional parts of integral multiples of 2,
defined by
√ √
an = n 2 − bn 2c, for n ∈ N,

where, for x ∈ R, bxc is the floor function of x, as defined in Exercise 9.

3
Bernhard Bolzano (1781–1848)
4
Karl Weierstrass (1815–1897)

29
1 Analysis

The terms of the sequence (an )n∈N are as follows:

√
2 = 1.414213 . . .a1 = 0.414213 . . .
√
2 2 = 2.828427 . . .a2 = 0.828427 . . .
√
3 2 = 4.242640 . . .a3 = 0.242640 . . .
√
4 2 = 5.656854 . . .a4 = 0.656854 . . .
√
5 2 = 7.071067 . . .a5 = 0.071067 . . .
√
6 2 = 8.485281 . . .a6 = 0.485281 . . .
..
.

The sequence (an )n∈N is bounded: indeed, 0 ≤ an < 1 for every n ∈ N. So by the
Bolzano-Weierstrass theorem this sequence has a convergent subsequence5 .

1.3 Continuity
A function f : R → R is a rule of correspondence that assigns to each real number a
unique real number. The functions we are most familiar with are actually unusually
well-behaved, and most functions are impossible to describe fully, let alone to work with.
Many bizarre functions make their appearance in analysis, and in order to avoid falling
into pitfalls with simplistic thinking based on our experience with “nice” functions, we
need our definitions and the hypotheses (assumptions) of theorems to be stated carefully
and clearly.
Within the huge collection of functions, there is an important subset: the continuous
functions. Continuous functions play a prominent role in analysis since they include all
the most familiar and useful functions, and because the all continuos functions share
many useful properties.
In this section we give the formal definition of a continuous function and prove two of
the most important properties of continuous functions: the Extreme value theorem and
the Intermediate value theorem.

1.3.1 Definition of continuity

In everyday speech, a ‘continuous’ process is one that proceeds without gaps of interrup-
tions or sudden changes. What does it mean for a function f : R → R to be continuous?
The common informal definition of this concept states that a function f is continuous if
one can sketch its graph without lifting the pencil. In other words, the graph of f has no
breaks in it. If a break does occur in the graph, then this break will occur at some point.
Thus (based on this visual view of continuity), we first give the formal definition of the
continuity of a function at a point. Next, if a function is continuous at each point, then
it will be called continuous.
5
In fact, it can be shown that these fractional parts an are “dense” in (0, 1). Thus given any number
L ∈ (0, 1), there exists a subsequence of the sequence (an )n∈N above that converges to L.

30
 1.3 Continuity

If a function has a break at a point, say c, then even if points x are close to c, the
points f (x) do not get close to f (c). See Figure 1.8.
This motivates the following definition of continuity, which guarantees that if a function
is continuous at a point c, then we can make f (x) as close as we like to f (c), by choosing
x sufficiently close to c. See Figure 1.9.
Definitions.
1. Let I be an interval in R and let c ∈ I. A function f : I → R is continuous at c
if for every ε > 0, there exists a δ > 0 such that for all x ∈ I satisfying |x − c| < δ,
|f (x) − f (c)| < ε.

2. If f is not continuous at c, we say that f is discontinuous at c.

3. A function f : I → R is continuous on I (or just continuous if I is clear from

the context) if for every c ∈ I, f is continuous at c.

4. If f is not continuous on I, we say that f is discontinuous on I (or just discon-

tinuous if I is clear from the context).
Examples.
1. Show that the function f : R → R given by f (x) = x for all x ∈ R is continuous.

Solution. To show that f is continuous, we have to show that it is continuous at c for

each c ∈ R. Accordingly, we start by fixing some c ∈ R. To show that f is continuous
at c, we have to show that, for all ε > 0, there is a δ > 0 such that for all x ∈ R, if
|x − c| < δ, then |f (x) − f (c)| < ε. Suppose then that we are given some ε > 0: we
now need to fnd a suitable δ > 0. A very important point is that δ should be
independent of x, but may (and usually will) depend on ε and c. In this case, the
following choice of δ will work: δ = ε. Then, if x ∈ R and |x − c| < δ, we have:

|f (x) − f (c)| = |x − c| < δ = ε.

This proves that f is continuous at c. Since the choice of c ∈ R was arbitrary, it

follows that f is continuous on R.

f (c)

Figure 1.8: A function with a break at c. If x lies to the left of c, then f (x) is not close
to f (c), no matter how close x comes to c.

31
1 Analysis

f (c) + ε
f (x)
f (c)

f (c) − ε

c−δ c x c+δ x

Figure 1.9: The definition of the continuity of a function at point c. If the function is
continuous at c, then given any ε > 0 (which determines a strip around the
line y = f (c) of width 2ε), there exists a δ > 0 (which determines an interval
of width 2δ around the point c) such that whenever x lies in this interval (so
that x satisfies c − δ < x < c + δ, that is, |x − c| < δ), then f (x) satisfies
f (c) − ε < f (x) < f (c) + ε, that is, |f (x) − f (c)| < ε.

2. Show that the function f : R → R given by f (x) = 2x + 1 for all x ∈ R is continuous.

Solution. Fix c ∈ R. Given ε > 0, we again need to fnd a suitable δ > 0. Here we
set δ = ε/2. Then, if x ∈ R and |x − c| < δ, we have:

|f (x) − f (c)| = |(2x + 1) − (2c + 1)| = 2|x − c| < 2δ = ε.

This proves that f is continuous at c. Since the choice of c ∈ R was arbitrary, it

follows that f is continuous on R.

3. Show that the function f : R → R given by f (x) = 1 for all x ∈ R is continuous.

Solution. Let c ∈ R = (−∞, ∞). We have to prove that f is continuous at c. Let

ε > 0 be given. In this case, any positive choice of δ will work; for instance, let δ = 1.
Then if x ∈ R and |x − c| < δ = 1, we have:

|f (x) − f (c)| = |1 − 1| = |0| = 0 < ε.

So f is continuous at c. Since the choice of c ∈ R was arbitrary, it follows that f is

continuous on R.

32
 1.3 Continuity

4. Show that the function f : R → R given by


0 if x = 0,
f (x) =
1 if x ∈ R \ {0},

is discontinuous at 0 and continuous at all c ∈ R \ {0}.

0 x

Figure 1.10: A function continuous everywhere except at 0.

Solution. Suppose that f is continuous at 0. Then for any given ε > 0 there exists
a δ > 0 such that whenever |x − 0| < δ, |f (x) − f (0)| < ε. In particular, setting
ε = 12 , we see that there is a δ > 0 such that for all x ∈ R, if |x| = |x − 0|
 
< δ, then
|f (x) − f (0)| = |f (x) − 0| = |f (x)| < ε = 2 . Take x = 2 ∈ R: then |x| =  2  = 2δ < δ,
1 δ δ
  
δ  1
but |f (x)| = f = |1| = 1 > = ε, which is a contradiction. So f is not

2 2

continuous at 0.
Next we show that for all c ∈ R \ {0}, f is continuous at c. Let ε > 0 be given. Take
δ = |c|
2
> 0. Then if x ∈ R and |x − c| < δ, we have

|c|
|c| − |x| ≤ ||c| − |x|| ≤ |c − x| = |x − c| < δ =
2
and so
|c|
|x| > > 0.
2
Thus x 6= 0 and so f (x) = 1. Hence if x ∈ R and |x − c| < δ, we obtain

|f (x) − f (c)| = |1 − 1| = |0| = 0 < ε.

Consequently f is continuous at c.

1
5. Show that the function f : (0, ∞) → R given by f (x) = x
for all x ∈ R is continuous.
n o
c εc2
Solution. Let c ∈ (0, ∞). Given ε > 0, let δ = min ,
2 2
(> 0). Then if x ∈ (0, 1)
and |x − c| < δ, we have
c c
c − x ≤ |c − x| < δ ≤ , and so x > > 0.
2 2

33
1 Analysis

Consequently, if x ∈ (0, ∞) and |x − c| < δ,

1 1  |c − x| 1 1 2 1 2δ
 
 −  = = |x − c| · · < δ · · = 2 ≤ ε.

x c xc x c c c c
So f is continuous at c. Since the choice of c ∈ (0, ∞) was arbitrary, it follows that f
is continuous on (0, ∞).
Note: In practice, the choice of δ is the last part of the proof to be filled in. As
we go through the rest of the proof, we see the need to have (a) δ ≤ |c|/2, and then
(b) 2δ
c2
≤ ε. Only after we have collected all the conditions do we make the choice of
δ.

1.3.2 Continuous functions preserve convergent sequences

We now give an alternative characterisation of continuity.

Theorem 1.3.1. Let I be an interval in R and let c ∈ I. Suppose that f : I → R is a

function. Then f is continuous at c iff

for every convergent sequence (xn )n∈N contained in I with limit c,

(1.7)
(f (xn ))n∈N is convergent and lim f (xn ) = f (c).
n→∞

Proof. “Only if” (=⇒) direction: Assume that f is continuous at c ∈ I and let
(xn )n∈N be a convergent sequence contained in I with limit c. We have to show that
(f (xn ))n∈N converges to f (c).
Since f is continuous at c ∈ I, given any ε > 0, ∃δ > 0 such that for all x ∈ I satisfying
|x − c| < δ, |f (x) − f (c)| < ε. Since (xn )n∈N is convergent with limit c, ∃N ∈ N such
that for all n > N , |xn − c| < δ.
Consequently for all n > N , |f (xn ) − f (c)| < ε. Thus (f (xn ))n∈N is convergent with
limit f (c).

“If” (⇐=) direction: Suppose that (1.7) holds. We have to show that f is continuous
at c. We prove this by contradiction. Assume that f is not continuous at c, that is,

¬ [∀ε > 0 ∃δ > 0 such that ∀x ∈ I such that |x − c| < δ, |f (x) − f (c)| < ε,]

or equivalently,

∃ε > 0 such that ∀δ > 0 ∃x ∈ I such that |x − c| < δ but |f (x) − f (c)| ≥ ε.

By choosing δ = n1 , we find xn ∈ I such that |xn − c| < δ = n1 and |f (xn ) − f (c)| ≥ ε.

Claim 1: The sequence (xn )n∈N is contained in I and is convergent with limit c.
Indeed, we have for all n ∈ N, xn ∈ I. Furthermore, given any ζ > 0, we can find
N ∈ N such that ζ1 < N (Archimedean property), that is, N1 < ζ. Hence for n > N ,
|xn − c| < N1 < ζ. So (xn )n∈N is convergent with limit c.

34
 1.3 Continuity

Claim 2: The sequence (f (xn ))n∈N does not converge to f (c).

Indeed for all n ∈ N, we have |f (xn ) − f (c)| ≥ ε. Thus for instance 2ε > 0, but it is not
possible to find a large enough N ∈ N such that for all n > N , |f (xn ) − f (c)| < 2ε (for if
this were possible, then we would arrive at the contradiction ε ≤ |f (xn ) − f (c)| < 2ε ).
Claims 1 and 2 show that (1.7) does not hold, a contradiction. Hence f is continuous
at c.

The above theorem allows us to easily prove the following useful Theorem 1.3.2. Before
stating it, we introduce some convenient notation.
Definitions. Let I be an interval in R. Given functions f : I → R and g : I → R, we
define the following:
1. If α ∈ R, then we define the function αf : I → R by (αf )(x) = α · f (x), x ∈ I.
2. We define the absolute value of f to be the function |f | : I → R given by |f |(x) =
|f (x)|, x ∈ I.
3. The sum of f and g is the function f +g : I → R defined by (f +g)(x) = f (x)+g(x),
x ∈ I.
4. The product of f and g is the function f g : I → R defined by (f g)(x) = f (x)g(x),
x ∈ I.
5. If k ∈ N, then we define the kth power of f , to be the function f k : I → R given by
f k (x) = (f (x))k , x ∈ I.
 
1 1 1
6. If for all x ∈ I, g(x) 6= 0, then we define the function g
: I → R by g
(x) = g(x)
,
x ∈ I.
Theorem 1.3.2. Let I be an interval in R and let c ∈ I. Suppose that f : I → R and
g : I → R are continuous at c. Then:
1. For all α ∈ R, αf is continuous at c.
2. |f | is continuous at c.
3. f + g is continuous at c.
4. f g is continuous at c.
5. For all k ∈ N, f k is continuous at c.
1
6. If for all x ∈ I, g(x) 6= 0, then g
is continuous at c.
Proof. Suppose that (xn )n∈N is a convergent sequence contained in I, with limit c. Since
f and g are continuous at c, from Theorem 1.3.1, it follows that (f (xn ))n∈N and (g(xn ))n∈N
are convergent with limits f (c) and g(c), respectively. Each one of the statements now
follows easily from Theorem 1.2.6 and a second application of Theorem 1.3.1. For
example, consider statement number 4 (the other cases may be proved as exercises). By
Theorem 1.2.6, (f (xn )g(xn ))n∈N is convergent with limit f (c)g(c), that is, ((f g)(xn ))n∈N
is convergent with limit (f g)(c). Thus by Theorem 1.3.1, f g is continuous at c.

35
1 Analysis

Example. Since f : R → R given by f (x) = x for x ∈ R is continuous (see Example 1

on p31), it follows that for all k ∈ N, xk is continuous. Thus given arbitrary scalars
a0 , a1 , . . . , aN in R, it follows that the functions a0 · 1, a1 · x, . . . , aN · xN are continuous.
Consequently the polynomial function p : R → R defined by
p(x) = a0 + a1 x + · · · + aN xN , x ∈ R
is continuous.

1.3.3 Restrictions and compositions of functions

Definition. If f : I → R is a function on an interval I, and J is an interval contained
in I (that is, J ⊆ I), then the restriction of f to J , is the function f  : J → R

J
defined by 
f  (x) = f (x), x ∈ R.

J

The following theorem implies (completely as expected) that the restriction of a

continuous function is continuous.
Theorem 1.3.3. Let I and J be intervals such that J ⊆ I and let c ∈ J. If f is
continuous at c, then f  is continuous at c.

J

Proof. Exercise 71.

The converse of the above theorem is not true unless J is an open interval (see
Exercise 71).
Definition. If f : I → R and g : J → R are functions such that for all x ∈ I, f (x) ∈ J,
then the composition of g with f , is the function g ◦ f : I → R defined by
(g ◦ f )(x) = g(f (x)), x ∈ R.
The following theorem implies that the composition of continuous functions is continu-
ous.
Theorem 1.3.4. If f : I → R is continuous at c ∈ I and g : J → R is continuous at
f (c) (with f (x) ∈ J for all x ∈ I), then g ◦ f is continuous at c.
Proof. Take ε > 0. Then, as g is continuous at f (c), there is some η > 0 such that
y ∈ J and |y − f (c)| < η imply that |g(y) − g(f (c))| < ε. As also f is continuous at c,
there exists δ > 0 such that x ∈ I and |x − c| < δ imply |f (x) − f (c)| < η.
Now we assemble what we know. Suppose that x ∈ I and |x − c| < δ. Then
f (x) ∈ J and |f (x) − f (c)| < η, from which we deduce that |g(f (x)) − g(f (c))| < ε, i.e.,
|(g ◦ f )(x) − (g ◦ f )(c)| < ε. We conclude that g ◦ f is continuous at c.
Example. Show that the function f : (1, ∞) → R defined by
s
x2 − 1
f (x) = , x > 1,
x2 + 1
is continuous on (1, ∞).

36
 1.3 Continuity

√
Solution. Let g : (0, ∞) → R be defined by g(x) = x. By Exercise 68, g is continuous.
Let h : R → (0, ∞) be defined by

x2 − 1
h(x) = , x ∈ R.
x2 + 1
We have already seen that all polynomial functions are continuous. Therefore, p1 , p2 : R →
R defined by p1 (x) = x2 − 1 and p2 (x) = x2 + 1 are continuous. Since p2 (x) 6= 0 for all
1
x ∈ R, it follows by Theorem 1.3.2(6) that is continuous. Then by Theorem 1.3.2(4),
p2
1
p1 · = h is continuous. Note that the restriction of h to (1, ∞), h|(1,∞) assumes only
p2
positive values: if x ∈ (1, ∞) then x > 1, therefore x2 − 1 > 0, and since x2 + 1 > 0
anyway, h(x) > 0. This shows that the composition g ◦ h|(1,∞) = f is properly defined.
By Theorem 1.3.3, h|(1,∞) is continuous. Finally, by Theorem 1.3.4, g ◦ h|(1,∞) = f is
continuous at each c > 1, and therefore, continuous.

1.3.4 Intermediate value theorem

We now prove one of the most fundamental (and obvious!) theorems on continuous
functions: a continuous function cannot “hop over” intermediate values. For instance, if
the height of a mountain is 1976 meters above sea level, then given any number between
0 and 1976, say 399, there must exist a point on the mountain that is exactly 399 meters
above sea level.
y
f (b)

f (a)

a c x
b

The Intermediate value theorem was first proved by Bernhard Bolzano in 1817.
Theorem 1.3.5. (Intermediate value theorem). If f : [a, b] → R is continuous and y
is such that f (a) ≤ y ≤ f (b) or f (b) ≤ y ≤ f (a), then there exists c ∈ [a, b] such that
f (c) = y.
Proof. Suppose first that f (a) ≤ y ≤ f (b). The case f (b) ≤ y ≤ f (a) is similar, and
will be discussed at the end of the proof below.
Let S = {x ∈ [a, b] | f (x) ≤ y}. We want to prove that
1. sup S exists, and

2. if we set c = sup S, then c ∈ [a, b] and f (c) = y.

37
1 Analysis

Since f (a) ≤ y, it follows that a ∈ S, so S 6= ∅. Secondly, S is clearly bounded above by

b. Therefore, by the l.u.b. property, sup S exists.
Write c = sup S. Since a ∈ S, a ≤ c. Since b is an upper bound of S, c ≤ b. Therefore,
c ∈ [a, b], and it remains to prove that f (c) = y. We do this in two steps: first we show
that f (c) ≤ y and then that f (c) ≥ y.
Proof that f (c) ≤ y: For each n ∈ N, c − n1 is not an upper bound of S, since it is less
than the least upper bound c. Therefore, there exists xn ∈ S such that c − n1 < xn ≤ c.
By the Sandwich theorem, n→∞ lim xn = c. By Theorem 1.3.1, n→∞ lim f (xn ) = f (c). Since
xn ∈ S, f (xn ) ≤ y. Therefore, f (c) = lim f (xn ) ≤ lim y = y (see Exercise 44.)
n→∞ n→∞
Proof that f (c) ≥ y: The sequence (xn ) defined by
b−c
xn = c +
n
lies in [a, b], and n→∞
lim xn = c. By Theorem 1.3.1, n→∞lim f (xn ) = f (c). Since xn > c and
c is an upper bound of S, xn ∈ / S, and therefore, f (xn ) > y. As before, it follows that
f (c) = lim f (xn ) ≥ lim y = y.
n→∞ n→∞

We have shown that f (c) ≤ y and f (c) ≥ y. In conclusion, we have found c ∈ [a, b]
such that f (c) = y.
Suppose now that y ∈ R is such that f (b) ≤ y ≤ f (a). To prove the theorem in this
case, it is easy to modify the above proof. Alternatively, we could use what we have just
proved, as follows. Note that we have

(−f )(a) ≤ −y ≤ (−f )(b),

so we may apply the above proof to the function −f (which is continuous by The-
orem 1.3.2(1) with α = −1). Then we obtain the existence of c ∈ [a, b] such that
(−f )(c) = −y, that is, f (c) = y.

Examples.
1. Show that every polynomial of odd degree with real coefficients has at least one real
root.

Solution. Suppose that p is a polynomial with degree 2m + 1, m ∈ N ∪ {0}. Let

p(x) = a0 + a1 x + · · · + a2m x2m + a2m+1 x2m+1 ,

where a2m+1 6= 0. If we divide p by its leading coefficient a2m+1 then the resulting
polynomial will still have the same roots (if any). We may therefore assume without
loss of generality that a2m+1 = 1.
In order to show that p has a real root, we will choose a large enough n ∈ N such
that p(n) > 0 and p(−n) < 0 and restrict our attention to an interval [−n, n]. Then
appealing to the Intermediate Value Theorem, we can conclude that p must vanish at
some point in this interval, that is, for some real c ∈ [−n, n], p(c) = 0.

38
 1.3 Continuity

It remains to find an n ∈ N such that p(n) > 0 > p(−n). We expect such an n to be
large in general. Intuitively, the term in the polynomial that dominates when x is
large is the term with highest exponent x2m+1 . We now compare p(x) with x2m+1 by
considering the quotient:

p(x) a0 a1 a2m
2m+1
= 2m+1 + 2m + · · · + +1 (1.8)
x x x x
(recall we assumed a2m+1 = 1). If we make the substitution x = n and let n → ∞, we
obtain
p(n) a0 a1 a2m
 
lim = lim + 2m + · · · + + 1 = 1.
n→∞ n2m+1 n→∞ n2m+1 n n
By choosing ε = 1/2 in the definition of a limit, we obtain that there exists N1 ∈ N
such that for all n > N1 ,  
 p(n)  1
 2m+1 − 1 < ε = ,
 
n  2
and since |A| ≥ −A for any A ∈ R, it follows that

p(n) 1
1− 2m+1
< ,
n 2
n2m+1
or p(n) > 2
> 0 as long as n > N1 .
To obtain the inequality p(−n) < 0 for suitably large n, we next make the substitution
x = −n in (1.8) and take the limit as n → ∞ to obtain
!
p(−n) a0 a1 a2m
lim = lim + + ··· + + 1 = 1.
n→∞ (−n)2m+1 n→∞ (−n)2m+1 (−n) 2m −n

Again applying the definition of a limit with ε = 1/2, we obtain as before that there
exists an N2 ∈ N such that for all n > N2 ,

p(−n) 1
1− < .
(−n)2m+1 2

Multiplying by n2m+1 we obtain (careful with the minus!)

n2m+1
n2m+1 + p(−n) < ,
2
2m+1
or p(−n) < − n 2
< 0 as long as n > N2 .
Thus if we choose n = 1 + max{N1 , N2 }, we obtain both inequalities p(n) > 0 and
p(−n) < 0, which finishes the proof.
1
For example, the polynomial p(x) = x3 − x2014 + 399 has a real root in [−1, 1]: indeed
1 1
p(1) = 399 > 0 and p(−1) = −2 + 399 < 0 and so ∃c ∈ [−1, 1] such that p(c) = 0.

39
1 Analysis

2. Show that at any given time, there exists a pair of diametrically opposite points on
the equator which have the same temperature.

Solution. Let T (Θ) denote the surface temperature at the point at longitude Θ. See
Figure 1.11. (Note that Θ(0) = Θ(2π).) Assuming that Θ is a continuous function of

Figure 1.11: The point on the equator with longitude Θ.

Θ, it follows that the function f : [0, π] → R defined by f (Θ) = T (Θ) − T (Θ + π) is

continuous as well. If f (0) = 0, then it follows that the temperatures at 0 and 180◦
longitude are the same. If f (0) 6= 0, then since f (0) and f (π) = −f (0) have opposite
signs, by the intermediate value theorem, it follows that f must vanish at some point,
and so the claim follows.

1.3.5 Extreme value theorem

The next theorem states that a continuous function on an interval [a, b] attains its
maximum and minimum values.

d
a c x
b

Theorem 1.3.6 (Extreme value theorem). Let [a, b] be any closed and bounded interval
and let f : [a, b] → R be a continuous function. Then there exists c ∈ [a, b] such that

f (c) = sup {f (x) | x ∈ [a, b]} , (1.9)

and there exists d ∈ [a, b] such that

f (d) = inf {f (x) | x ∈ [a, b]} . (1.10)

40
 1.3 Continuity

Since c, d ∈ [a, b] in the above theorem, the supremum and infimum in (1.9) and (1.10)
are in fact the maximum and minimum, respectively. This proof is a beautiful application
of the Bolzano-Weierstrass theorem.
Proof. We prove the first half of the theorem, leaving the second half as an exercise.
Let S = {f (x) | x ∈ [a, b]}. We have to prove that there exists c ∈ [a, b] such that
f (c) = sup S. The plan of the proof is as follows.

1. First we show that sup S exists.

2. Then we show that sup S ∈ S.

Then, by the definition of S, we may conclude that sup S = f (c) for some c ∈ [a, b].
To show that sup S exists, we use the l.u.b. property of R. First, S 6= ∅, since for
example f (a) ∈ S. Secondly, S is bounded above, as shown by the following proof by
contradiction.
Suppose S is not bounded above. For each n ∈ N, choose an f (xn ) for some xn ∈ [a, b]
such that f (xn ) > n. Then (xn ) is a bounded sequence. By the Bolzano-Weierstrass
Theorem, (xn ) has a convergent subsequence (xnk )k∈N . Let its limit be c = lim xnk . Since
k→∞
a ≤ xn ≤ b and limits preserve inequalities, a ≤ c ≤ b. By Theorem 1.3.1, (f (xnk )k∈N ) is
convergent (we assumed that f is continuous). By Theorem 1.2.4, (f (xnk )k∈N ) is bounded.
On the other hand, for each k ∈ N, f (xnk ) > nk ≥ k (since 1 ≤ n1 < n2 < n3 < . . . ).
Therefore, (f (xnk )k∈N ) is unbounded, a contradiction.
Next we show that sup S ∈ S. Recall that S = {f (x) | x ∈ [a, b]}. Write M = sup S.
We have to prove that M ∈ S. Since M is the least upper bound of S, it follows that
for all n ∈ N, M − n1 is not an upper bound of S. Therefore, there exists f (xn ) ∈ S
(with xn ∈ [a, b]) such that M − n1 < f (xn ). The sequence (xn ) is bounded. As before,
by the Bolzano-Weierstrass Theorem it has a convergent subsequence (xnk )k∈N with limit
c ∈ [a, b], say. We next show that this c satisfies f (c) = M = sup S.
By Theorem 1.3.1, (f (xnk )k∈N is convergent with limit f (c). Since c ∈ [a, b], f (c) ∈ S.
Finally, because
1 1
M− < f (xnk ) ≤ M and lim M − = M,
nk k→∞ nk
the Sandwich Theorem shows that f (c) = lim f (xnk ) = M .
k→∞
We have found c ∈ [a, b] such that f (c) = M = sup S. This finishes the proof of the
first half of the theorem.
The proof that there exists d ∈ [a, b] such that f (d) = inf S is left as an exercise
(Exercise 80).

41
 1 Analysis

Example. Show that the statement of Theorem 1.3.6 does not hold if [a, b] is replaced
by (a, b).
Solution. The function f : (0, 1) → R given by f (x) = x is continuous on (0, 1). If

S = {f (x) | x ∈ (0, 1)} ,

then S = {x | x ∈ (0, 1)} = (0, 1). Thus sup S = 1 ∈ / S and inf S = 0 ∈

/ S, that is, the
values 0 and 1 are not attained by the function f . We conclude that the statement of
Theorem 1.3.6 does not hold if [a, b] is replaced by (a, b).

1.4 Exercises
1. Fill in the following table for each of the sets S = A, B, . . . , K below.

An A If sup S exists, If inf S exists,

Is S
upper lower sup S inf S then is sup S then is inf S max S min S
bounded?
bound bound in S? in S?

a) A = {x ∈ R | 0 ≤ x < 1} g) G = {x ∈ R | x2 + 2x ≤ 1}
b) B = {x ∈ R | 0 ≤ x ≤ 1} h) H = {0, 2, 5, 2016}
c) C = {x ∈ R | x ∈ N and xprime} n   o
1
i) I = (−1)n 1 + n∈N
n  o 
1  n
d) D = n ∈ Z \ {0}

n

j) J = {x2 | x ∈ R}
n  o
e) E = − n1  n ∈ N

n  o
x2
f) F = {x ∈ R | x > x2 } k) K = x∈R

1+x2


2. Determine whether the following statements are TRUE or FALSE. Give reasons for
your answers in each case: either a (brief) proof or a counterexample. (For instance, if
you think statement (d) is false, you should give an example of a bounded non-empty
subset of R that does not have a maximum, and explain (briefly) why your set has
the required properties.)
a) Let S be a subset of R such that, for each x ∈ S, there exists u ∈ R with x ≤ u.
Then S is bounded above.
b) If u is an upper bound of a subset S of R, and u0 < u, then u0 is not an upper
bound of S.
c) If u∗ is the supremum of a subset S of R, and ε is any positive real number, then
u∗ − ε is not an upper bound of S.
d) Every bounded non-empty subset of R has a maximum.
e) Every bounded subset of R has a supremum.

42
 1.4 Exercises

6
f) A non-empty subset of R is bounded above iff it has a supremum.
g) For every set that has a supremum, the supremum belongs to the set.
h) For every non-empty bounded set S, if inf S < x < sup S, then x ∈ S.
i) For S a non-empty subset of R, {x2 | x ∈ S} is bounded iff S is bounded.

3. For any non-empty bounded set S, prove that inf S ≤ sup S, and that the equality
holds iff S is a singleton set (that is a set with exactly one element).

4. Let A and B be non-empty subsets of R that are bounded above and such that A ⊆ B.
Prove that sup A ≤ sup B.

5. Let A and B be non-empty subsets of R that are bounded above and define

A + B = {x + y | x ∈ A and y ∈ B} .

a) Show that sup A + sup B is an upper bound for A + B.

Deduce that sup(A + B) exists and that sup(A + B) ≤ sup A + sup B.
b) For any real number ε > 0, show that (sup A − ε) + (sup B − ε) is not an upper
bound for A + B.
Deduce that sup(A + B) ≥ sup A + sup B − 2ε, for every ε > 0.
c) Show that sup(A + B) = sup A + sup B.

6. Let S be a non-empty subset of real numbers which is bounded below. Let −S denote
the set of all real numbers −x, where x belongs to S.
a) Show that −S is bounded above, and therefore has a supremum.
b) Prove that inf S exists, and that inf S = − sup(−S).
Note: in this exercise, we deduce the “greatest lower bound principle” from the least
upper bound principle. Accordingly, you may not assume the greatest lower bound
principle in this exercise. (However, once we have established this result, we can use
the greatest lower bound principle henceforth.)

7. In this exercise it is shown that for any positive real number r there exists a positive
real number s such that s2 = r. In other words, any positive real number has a
positive square root.
Given r ∈ R with r > 0, let S = {x ∈ R | x2 < r}.
a) Prove that S 6= ∅.
b) Prove that S is bounded above.
Hint: Distinguish between the cases r > 1 and r ≤ 1.
c) Conclude that S has a least upper bound s. Show that s > 0.
6
If you think the statement is TRUE, you have two things to prove: (i) if a non-empty set S is bounded
above, then it has a supremum; (ii) if a non-empty set S has a supremum, then it is bounded above.
If you think the statement is FALSE, you should provide a counterexample to one of (i) and (ii).

43
 1 Analysis

d) Show that if s2 > r then there exists ε > 0 such that s − ε is an upper bound of S.
e) Show that if s2 < r then there exists ε > 0 such that s + ε ∈ S.
f) Conclude that s2 = r by explaining why each of the conclusions in 7d) and 7e)
gives a contradiction.
n  o
8. Let S be a non-empty set of positive real numbers, and define S −1 = 1
x∈S .

x


a) Show that, if inf S = 0, then S −1 is not bounded above.

b) Show that, if inf S > 0, then S −1 is bounded above and sup S −1 = 1
inf S
.

9. In this exercise we define the floor and ceiling functions. For any x ∈ R define the
set Sx = {n ∈ Z | n ≤ x}.
a) Show that, for any x ∈ R, the set Sx is non-empty and bounded above.
Hint: To show that Sx 6= ∅, you will need the Archimedean property.
b) Show that sup Sx exists and sup Sx ∈ Sx for any x ∈ R. Explain why we obtain as
a consequence that the following gives a proper definition of the floor function
b·c : R → Z
bxc = max {n ∈ Z | n ≤ x} , x ∈ R.

c) Show that x − 1 < bxc ≤ x, for every x ∈ R.

d) Adapt 9a) and 9b) to explain why the following gives a proper definition of the
ceiling function d·e : R → Z

dxe = min {n ∈ Z | n ≥ x} , x ∈ R.
√
e) Show that b k 2 + kc = k for all k ∈ N.

10. Let S be a subset of R that does not have a maximum. Show that, for every x ∈ S,
there is y ∈ S with y > x.
Hint: Deal separately with the two cases: (i) S has a supremum; (ii) S has no
supremum.

11. Determine whether the following statements are TRUE or FALSE. Give reasons (brief
proofs or counterexamples) for your answers in each case.
a) If a ≤ b then (−∞, a) ⊆ (−∞, b).
b) For real numbers a and b, a ≤ b iff (b, ∞) ⊆ (a, ∞).
c) (a, b) ⊆ (c, d) iff c ≤ a and b ≤ d.
d) (a, b) ⊆ [c, d] iff c < a and b < d.
e) For any x ∈ R, −|x| ≤ x ≤ |x|.
f) For any x ∈ R, −x ≤ |x| ≤ x.
g) For any x, y ∈ R, if x < y then |x| ≤ |y|.
h) For any x, y ∈ R, if x2 ≤ y 2 then |x| ≤ |y|.

44
 1.4 Exercises

i) If I and J are intervals, then I ∩ J is an interval.

j) If I and J are intervals with 0 ∈ I ∩ J, then I ∪ J is an interval.
12. Let x0 ∈ R and δ > 0. Prove that (x0 − δ, x0 + δ) = {x ∈ R | |x − x0 | < δ}.
13. Prove that the distance satisfies the following properties:
D1 (Positive definiteness) For all x, y ∈ R, |x − y| ≥ 0. If |x − y| = 0 then x = y.
D2 (Symmetry) For all x, y ∈ R, |x − y| = |y − x|.
D3 (Triangle inequality) For all x, y, z ∈ R, |x − z| ≤ |x − y| + |y − z|.
14. Prove that if x, y are real numbers, then ||x| − |y|| ≤ |x − y|.
15. Let S be a subset of R.
a) Show that, if S is bounded, then there is some M ∈ R such that |x| ≤ M for all
x ∈ S.
b) Show also the converse: if there is some M ∈ R such that |x| ≤ M for all x ∈ S,
then S is bounded.
Thus we have an alternative definition of what it means for a set S to be bounded:
this definition is often more convenient to use.
16. Let a, b, c be real numbers such that a < c < b. Show that there exists a t > 0 such
that (c − t, c + t) ⊆ (a, b).
17. Prove, directly from the definition, that the constant sequence (1)n∈N is convergent.
 
18. Prove, directly from the definition, that the sequence √2 is convergent, with
n n∈N
limit 0.
 
3n−1
19. Prove, directly from the definition, that the sequence n+2 n∈N
is convergent, and
find its limit.
20. a) Let (an )n∈N be a convergent sequence with limit L, and let M be some real number
with M 6= L. Show that the set {n ∈ N | an = M } is bounded above.
b) Prove that the sequence ((−1)n )n∈N is divergent.
21. Use the definition of limit to prove directly that 1 is not a limit of the sequence
(1/n)n∈N .
22. In each of the cases listed below, give an example of a divergent sequence (an )n∈N that
satisfies the given conditions.
a) For every ε > 0, there exists an N such that, for infinitely many n > N , |an −1| < ε.
b) There exists an ε > 0 and an N ∈ N such that for all n > N , |an − 1| < ε.
23. Let S be a non-empty subset of R that is bounded above. Show that there exists a
sequence (an )n∈N contained in S (that is, an ∈ S for all n ∈ N) which is convergent
with limit equal to sup S.

45
 1 Analysis

24. Let (an )n∈N be a sequence such that for all n ∈ N, an ≥ 0. Prove that if (an )n∈N is
convergent with limit L, then L ≥ 0.

25. A sequence (an )n∈N is said to be a Cauchy sequence if for every ε > 0, there exists
an N ∈ N such that for all n, m > N , |an − am | < ε.
Show that every convergent sequence is Cauchy.
Hint: |an − am | = |an − L + L − am | ≤ |an − L| + |am − L|.

26. In the proof of Theorem 1.2.2, where did we use the assumption that x ≥ −1? Is the
inequality true for x = −2? Is it true for x = −3?
√
27. For a real number x such that |x| < 1, show that the sequence (xn n)n∈N is convergent,
with limit 0.

28. a) Use Bernoulli’s Inequality to show that, for all n ∈ N,

!2
1 √ 2 1 3
1 ≤ n < (1 +
n n) ≤ 1 + √
n ≤1+ √ .
n n

1
b) Prove that (n n )n∈N is convergent with limit 1.
n+a
 
29. Let a and b be positive real numbers. Show that the sequence is monotone.
n+b n∈N

30. Let (an )n∈N be a sequence defined by

2n + 1
a1 = 1 and an = an−1 for n ≥ 2.
3n
Prove that (an )n∈N is convergent.

31. Use Bernoulli’s Inequality (Theorem 1.2.2) to show that, for x a real number with
x > 1, the sequence (xn )n∈N is divergent.

32. Suppose that the sequence (an )n∈N converges to 0 and the sequence (bn )n∈N is bounded.
Use the definition to prove that lim an bn = 0.
n→∞

33. For each n ∈ N, let Hn = 1 + 12 + 13 + · · · + n1 . The sequence (Hn )n∈N is the sequence
of harmonic numbers.
a) Show that for any m, n ∈ N with m > n,
m−n
Hm − Hn > .
m
1 1 1
(Hint: note that Hm − Hn = n+1
+ n+2
+ ··· + m
, a sum of m − n terms.)
1
b) Deduce that H2k+1 − H2k > 2
for any k ∈ N.
3 4 k+2
c) Show that H2 = 2
, H4 > 2
, H8 > 52 , H16 > 26 , H32 > 27 , and in general, H2k > 2
for all k ≥ 2.

46
 1.4 Exercises

d) Let M > 0 be given. Show that there exists n ∈ N such that Hn > M .
e) Show that (Hn )n∈N is divergent.

34. a) (∗) Let (an )n∈N be a convergent sequence with limit L. Prove that the sequence
(sn )n∈N , where
a1 + · · · + an
sn = , n ∈ N,
n
is also convergent with limit L.
Hint: use the fact that (an )n∈N is bounded.
b) Give an example of a sequence (an )n∈N such that (sn )n∈N is convergent but (an )n∈N
is divergent.

35. a) Given a bounded sequence (an )n∈N , define

`k = inf {an | n ≥ k} , k ∈ N.

Show that the sequence (`n )n∈N is bounded above and increasing, and conclude
that it is convergent. (Its limit is denoted by lim inf an .)
n→∞

b) Find lim inf (2 + (−1)n − 1/n).

n→∞

36. Recall the definition of a Cauchy sequence from Exercise 25. Prove that every Cauchy
sequence is bounded. (Hint: follow the proof of Theorem 1.2.2.)

37. Determine whether the following sequences are convergent and, if so, find their limits:
√ !
n+ n

1/2n
√ 3 n
√ ; 1 + n + n( ) .
3n2 − 1 n∈N 4 n∈N

38. Let (an )n∈N be a convergent sequence with limit L, and let k ∈ N. Show that (an+k )n∈N
is also convergent with limit L.

39. Recall the convergent sequence (an )n∈N from Exercise 30 defined by

2n + 1
a1 = 1 and an = an−1 for n ≥ 2.
3n
Determine its limit.
Hint: Use the previous exercise.

40. a) Use Theorem 1.2.6 and induction to show that for any fixed k ∈ N,

1 1 1 1 1
 
lim + + + ··· + + = 0.
n→∞ n+1 n+2 n+3 n+k−1 n+k
Indicate step-by-step which parts of Theorem 1.2.6 you use.

47
 1 Analysis

b) Find the mistake in the following calculation:

1 1 1 1 1
 
lim + + + ··· + +
n→∞ n + 1 n+2 n+3 2n − 1 2n
1 1 1 1 1
= n→∞
lim + lim + lim + · · · + lim + lim
n+1 n→∞ n+2 n→∞ n+3 n→∞ 2n − 1 n→∞ 2n
= 0 + 0 + 0 + ··· + 0 + 0
= 0.

Remark: By Exercise 33a),

1 1 1 1 1 1
+ + + ··· + + = H2n − Hn > ,
n+1 n+2 n+3 2n − 1 2n 2
and therefore the limit, if it exists, cannot equal 0.7

41. Suppose that the sequence (an )n∈N is bounded. Prove that the sequence (cn )n∈N
defined by
a3 + 5n
cn = n2
an + n
is convergent, and find its limit.

42. Let (an )n∈N be a convergent sequence with limit 0 and suppose that an ≥ 0 for all
√
n ∈ N. Prove that the sequence ( an )n∈N is also convergent, with limit 0.
√ 
43. Show that n2 + n − n is a convergent sequence and find its limit.
n∈N
Hint: ‘Rationalise the numerator’ (as in the proof of Theorem 1.2.6.(8)).

44. Prove that if (an )n∈N and (bn )n∈N are convergent sequences such that for all n ∈ N,
an ≤ bn , then
lim an ≤ lim bn .
n→∞ n→∞

Hint: Use Exercise 24.

45. Show that, for each k ∈ N and each real number x with |x| < 1, limn→∞ xn nk = 0.
√
Hint: Write the elements of the sequence as (y n n)2k , for some suitably chosen y,
and use Exercise 27 as well as the Algebra of Limits.
 
46. Prove that the sequence (3 + (−1)n )1/n is convergent, with limit 1.
n∈N

47. Recall the floor function defined in Exercise 9: bxc = max{n ∈ Z | n ≤ x}.
bnxc
Let x be any real number. Set an = , for each n ∈ N. Show that the sequence
n
(an )n∈N is convergent, with limit x.
 
7 1 1 1 1 1
In fact, lim + + + ··· + + = ln 2, the proof of which is outside the
n→∞ n + 1 n+2 n+3 2n − 1 2n
scope of this course.

48
 1.4 Exercises

Is the sequence (an )n∈N always increasing?

Notice that the an are all rational numbers, so we have found a sequence of rational
numbers converging to the arbitrary real number x.
!
n! n!
48. Prove that the sequence is convergent and n→∞
lim = 0.
nn n∈N
nn
n! 1 2 n 1 1
Hint: Observe that 0 ≤ n
= · · ··· · ≤ · 1 · ··· · 1 = .
n n n n n n
2n
X 1
49. Use the Sandwich Theorem to find n→∞
lim n 2
.
i=n+1 i
1 1 1 1 1 1 1
Hint: − = ≤ 2 ≤ = − . Compare with Exercise 40b.
i i+1 i(i + 1) i i(i − 1) i−1 i
50. Prove that for all k ∈ N, the sequence
1 k + 2 k + 3 k + · · · + nk
!

nk+2 n∈N

is convergent and
1k + 2k + 3k + · · · + nk
lim = 0.
n→∞ nk+2
1
51. Suppose that (an )n∈N is a sequence with n
≤ an ≤ n for each n ∈ N. Show that
limn→∞ a1/n
n = 1.
Hint: See Exercise 28.
52. a) Show that a monotonically increasing sequence always has a constant subsequence
or a strictly increasing subsequence.
b) Show that any sequence of real numbers always has either a constant subsequence
or a strictly increasing subsequence or a strictly decreasing subsequence.
53. a) For x ∈ R, show that x(1 − x) ≤ 14 .
b) Define the sequence (an )n∈N recursively by a1 = 1/3 and an+1 = 4an (1 − an ) for
each n ∈ N. Write down a2 , a3 and a4 . Show that (an )n∈N has a convergent
subsequence.
54. a) Let L ∈ R and let (an )n∈N be a sequence of real numbers that does not converge to
L (that is, it is either divergent or its limit is not equal to L). Use the definition of
convergence to L to show that for some ε > 0, (an )n∈N has a subsequence (ank )k∈N
such that ank ∈ / (L − ε, L + ε) for all k ∈ N.
b) Use 54a) and the Bolzano-Weierstrass theorem to show that given any L ∈ R, any
bounded, divergent sequence has a convergent subsequence with limit not equal
to L.
c) Use 54b) to show that any bounded, divergent sequence has two subsequences
which converge to different limits.

49
 1 Analysis

d) Give an example of a bounded divergent sequence such that any convergent

subsequence converges to either 1 or −1. (This shows that we cannot hope for
more than two different limits for subsequences in c).)
55. Recall the definition of a Cauchy sequence from Exercise 25, where we had already
seen that every convergent sequence is Cauchy. Use the Bolzano-Weierstrass theorem
to prove the converse: if a sequence is Cauchy, then it is convergent.
Hint: Proceed as follows. Let (an )n∈N be a Cauchy sequence. From Exercise 36, it
follows that (an )n∈N is bounded. By the Bolzano-Weierstrass theorem, it follows that
(an )n∈N has a convergent subsequence, say (ank )k∈N with limit L. Prove (using the
fact that (an )n∈N is Cauchy), that then (an )n∈N is itself convergent with limit L.
56. Prove that the following functions are continuous:
a) f : R → R given by f (x) = 2 − 3x.
b) f : R → R given by f (x) = |x|.
57. Prove, using
√ the definition of continuity, that the function f : [0, ∞) → R given by
f (x) = x is continuous at 4.
√
Hint: First show that |x − 4| ≥ 2| x − 2| for every x ∈ [0, ∞).
58. Prove that the function f : R → R given by

x if x is irrational,
f (x) = 
0 if x is rational,
is continuous at 0.
59. Let the function f : R → R be given by f (x) = x2 .
a) Prove that f is continuous at 0.
b) Prove that f is continuous at 1.
c) (∗) Suppose that c is a nonzero real number. Prove that f is continuous at c.
(In Exercise 64, we will give a simpler proof of the fact that f is continuous on R.)
60. Let f : R → R be a function that satisfies f (x1 + x2 ) = f (x1 ) + f (x2 ) for all x1 , x2 ∈ R.
a) Show that f (0) = 0.
b) Suppose that f is continuous at 0. Prove that f is continuous on R.
Hint: Since f is continuous at 0, given ε > 0, ∃δ > 0 such that for all x ∈ R
satisfying |x| < δ, |f (x)| < ε (why?). Show that given any other point c ∈ R, the
function f is continuous at c by showing that the same δ works (for this ε).
c) Give an example of such a function.
61. Suppose that f : R → R and there exists an M > 0 such that for all x ∈ R,
|f (x)| ≤ M |x|. Prove that f is continuous at 0.
Hint: First find f (0).

50
 1.4 Exercises

62. Let f : R → R be defined by

(
0 if x is rational,
f (x) =
1 if x is irrational.

Prove that for every c ∈ R, f is discontinuous at c.

Hint: Use the fact that there are irrational numbers arbitrarily close to any rational
number and rational numbers arbitrarily close to any irrational number.

63. Let f : R → R be a continuous function. Prove that if for some c ∈ R, f (c) > 0, then
there exists a δ > 0 such that for all x ∈ (c − δ, c + δ), f (x) > 0.

64. Recall Exercise 59. The function f : R → R is given by f (x) = x2 . Using the
characterization of continuity provided in Theorem 1.3.1, prove that f is continuous
on R.

65. Prove that if f : R → R is continuous and f (x) = 0 whenever x is rational, then

f (x) = 0 for all x ∈ R. (Compare this to Exercises 58 and 62.)
Hint: Given any real number c, there exists a sequence of rational numbers (qn )n∈N
that converges to c: see Exercise 47.

66. Let f : R → R be a function that preserves divergent sequences, that is, for every
divergent sequence (xn )n∈N , (f (xn ))n∈N is divergent as well. Prove that f is one-to-one.
Hint: Let x1 , x2 be distinct real numbers, and consider the sequence x1 , x2 , x1 , x2 , . . . .

67. Write out a complete proof of statement 6 of Theorem 1.3.2.

√
68. Show, using Theorem 1.2.6, that the function f : [0, ∞) → R given by f (x) = x, is
continuous on [0, ∞).

69. Show that the rational function f : R → R defined by

x2
f (x) = , x ∈ R,
1 + x2
is continuous on R.

70. Prove Theorem 1.3.3 using the definition of continuity.

71. a) Let J = (a, b) be an open interval contained in another interval I. Let f : I → R

be a function. Let c ∈ J and assume that f  is continuous at c. Prove that f is

J
continuous at c.
Hint: You’ll need Exercise 16.
b) Give an example
 of intervals J and I with J ⊆ I, c ∈ J and a function f : I → R
such that f  is continuous at c, but f is not continuous at c.

J
(This shows that it is necessary to assume that J is an open interval in 71a).)
Hint: The floor function.

51
 1 Analysis
q
72. a) Suppose g : I → R is continuous, and g(x) ≥ 0 for x ∈ I. Let f (x) = g(x). Show
that f is continuous on I.
b) Prove that the function f : R → R defined by
v s
u
u 1 + 2x + x2
f (x) = 1+ , x ∈ R,
t
2 − 2x + x2
is continuous.
73. Suppose that f : [0, 1] → R is a continuous function such that for all x ∈ [0, 1],
0 ≤ f (x) ≤ 1. Prove that there exists at least one c ∈ [0, 1] such that f (c) = c.
Hint: Consider the continuous function g(x) = f (x) − x, and use the intermediate
value theorem.
74. At 8:00 a.m. on Saturday, a hiker begins walking up the side of a mountain to his
weekend campsite. On Sunday morning at 8:00 a.m., he walks back down the mountain
along the same trail. It takes him one hour to walk up, but only half an hour to walk
down. At some point on his way down, he realizes that he was at the same spot at
exactly the same time on Saturday. Prove that he is right.
Hint: Let u(t) and d(t) be the position functions for the walks up and down, and
apply the intermediate value theorem to f (t) = u(t) − d(t).
75. Show that the polynomial function p(x) = 2x3 − 5x2 − 10x + 5 has a real root in the
interval [−1, 2].
76. Set p(x) = x8 + bx3 − 5, where b is a real number. Show that p(x) has a root in [0, 1]
iff b ≥ 4.
77. Let f : R → R be continuous. If S := {f (x) | x ∈ R} is neither bounded above nor
bounded below, prove that S = R.
Hint: If y ∈ R, then since S is neither bounded above nor bounded below, there exist
x0 , x1 ∈ R such that f (x0 ) < y < f (x1 ).
78. Let f : [0, 1] → R be an arbitrary continuous function. Show that there exists a
c ∈ [0, 1] such that
f (c) − f (1) = (f (0) − f (1))c.

Hint: Find an appropriate function g on which the Intermediate Value theorem can
be applied.
79. a) Show that, given any continuous function f : R → R, there exists x0 ∈ [0, 1] and
m ∈ Z such that f (x0 ) = mx0 . In other words, the graph of f intersects some line
y = mx with integer slope, at some point x0 in [0, 1].
Hint: If f (0) = 0, take x0 = 0 and any m ∈ Z. If f (0) > 0, then choose N ∈ N
satisfying N > f (1), and apply the intermediate value theorem to the continuous
function g(x) = f (x) − N x on the interval [0, 1]. If f (0) < 0, then proceed in a
similar manner.

52
 1.4 Exercises

b) Prove that there does not exist a continuous function f : R → R such that f
assumes rational values at irrational numbers, and irrational values at rational
numbers, that is,
f (Q) ⊆ R \ Q and f (R \ Q) ⊆ Q.

Hint: Show that, if f is such a function, then there do not exist m ∈ Z and
x0 ∈ R such that f (x0 ) = mx0 .

80. Complete the proof of Theorem 1.3.6 by proving that if f : [a, b] → R is continuous,
then there exists d ∈ [a, b] such that f (d) = inf {f (x) | x ∈ [a, b]}.

81. Show that the statement of Theorem 1.3.6 does not hold if [a, b] is replaced by [a, b).

82. Give examples of functions f1 , f2 , f3 : [0, 1] → R, continuous on (0, 1], with the following
properties.
a) f1 is not continuous at 0, and {f (x) | x ∈ [0, 1]} is not bounded above;
b) f2 is not continuous at 0, and {f (x) | x ∈ [0, 1]} is bounded above, but does not
have a maximum;
c) f3 is not continuous at 0, and {f (x) | x ∈ [0, 1]} has a maximum.

83. Let f : [a, b] → R be an arbitrary continuous function. Let S = {f (x) | a ≤ x ≤ b}.

Show that if S contains more than one element, then S is an interval of the form [c, d].
Hint: First apply the Extreme Value theorem, then the Intermediate Value theorem.

84. The function f : R → R is given by f (x) = 3 + 4x − x2 − x4 . Show that there exists

c ∈ R such that f (c) ≥ f (x) for all x ∈ R.
Hint: Show that f (x) ≤ 0 if |x| > 2.

85. A function f : R → R is a periodic function if there exists T > 0 such that for all
x ∈ R, f (x + T ) = f (x). If f : R → R is continuous and periodic, then prove that f
is bounded, that is, the set S = {f (x) | x ∈ R} is bounded.

86. Let f : [a, b] → R be continuous on [a, b], and define f∗ as follows:

(
f (a) if x = a,
f∗ (x) =
max {f (y) | y ∈ [a, x]} if x ∈ (a, b].

a) Show that f∗ is a well-defined function.

b) Prove that f∗ is continuous on [a, b].
c) If f : [−1, 1] → R is given by f (x) = x − x2 , then find f∗ .

53
1 Analysis

87. Sample Exam Question. 2007 Q5.

(a) What does it mean to say that a sequence (an )n≥1 is convergent? Use this
definition to show that if (an )n≥1 is convergent, then (an+1 )n≥1 is also a convergent
sequence and n→∞lim an = n→∞
lim an+1 .
Let b be a real number with 2 < b < 3. We define a sequence (bn )n≥1 by

b1 = b and bn+1 = b2n − 4bn + 6 for every n ∈ N.

(b) Show that 2 < bn < 3 for every n ∈ N.

(c) Prove that (bn )n≥1 is a monotone sequence.
(d) Explain why n→∞
lim bn exists and find its value.
(e) Let S = {bn | n ∈ N}. Find sup S, inf S, max S, min S. Justify your answers.

88. Sample Exam Question. 2009 Q5.

(a) Let I be an interval in R and let c ∈ I.
What does it mean to say that a function f : I → R is continuous at c ? What
does it mean to say that a function f : I → R is continuous on I?
(b) Suppose that f : R → R, g : R → R and h : R → R are functions such that
(1) f and g are continuous on R,
(2) f (0) = g(0), and
(3) for every x ∈ R, f (x) ≤ h(x) ≤ g(x).
Show that h is continuous at 0.
(c) State the Intermediate Value Theorem.
Suppose that the function f : [0, 2] → R is continuous on the interval [0, 2]
and f (0) = f (2) ≥ f (1). Prove that there exist numbers a, b ∈ [0, 2] such that
|a − b| = 1 and f (a) = f (b).
Hint: Consider the function g : [0, 1] → R given by g(x) = f (x + 1) − f (x).

54
Chapter 2
Algebra

We have already seen many different “algebraic structures”, that is, different sets with
algebraic operations defined on them. For example, we know of the natural numbers
N on which addition (+) and multiplication (·) are defined, the integers Z on which
subtraction (−) is additionally defined, and the rational numbers Q on which division (|)
is defined in addition to the previous operations. What distinguishes the real numbers R
from Q is not so much an algebraic property as an order property (the l.u.b. property).
On the other hand, the difference between R and the complex numbers C is that certain
equations with no real solutions, such as x2 + 1 = 0, have solutions in C. Thus C has an
algebraic property not shared by R.
When we consider the operations on the above sets, we see for example that addition
and multiplication are commutative:

x + y = y + x and xy = yx

and associative
(x + y) + z = x + (y + z) and (xy)z = x(yz).
Subtraction and division do not share these properties. On the other hand, they are not
really operations in their own right, but are derived from addition and multiplication
along with the notion of inverses. Also, addition and multiplication have identity elements
(0 and 1, respectively).
We have also seen other algebraic systems, such as matrices, on which addition and
multiplication have been defined. We know that matrix addition is commutative and
associative (A + B = B + A and AB = BA), and matrix multiplication is associative:

(AB)C = A(BC),

but matrix multiplication is definitely not commutative:

AB 6= BA in general.

Again, there are identity elements (the zero matrix O for addition and the identity matrix
I for multiplication).

55
2 Algebra

There are many more different algebraic structures in mathematics. Abstract algebra
is the study of general properties of these algebraic structures. In this part of the course,
we study two important abstract algebraic structures: groups and vector spaces. We
begin with a discussion about groups, one of the simplest algebraic structures.

2.1 Groups
In this section, we study one of the most basic algebraic objects, namely a group. A
group is a set on which an operation or law of composition is defined, such that certain
properties hold. The precise definition is given in the next subsection.

2.1.1 Definition of a Group

By a law of composition on a set S, we mean a rule for combining pairs a, b of S to get
another element, say c, of S. We denote the set of all ordered pairs of elements from a
set S by S × S, that is, S × S = {(a, b) | a, b ∈ S}.
Definition. A law of composition on a set S is a function f : S × S → S.
The functional notation c = f (a, b) is not very convenient for what is going to follow,
and so instead, the element obtained by applying the law of composition to a pair (a, b)
is usually denoted using a notation resembling that used for addition or multiplication:

c = a ∗ b, or ab, or a ◦ b, or a + b and so on,

with a fixed choice being made for the particular law in question. In this notation we
then have that a ∗ b is a law of composition on S iff a ∗ b ∈ S for any a, b ∈ S. For later
reference, we state the property of ∗ being a law of composition (which any particular ∗
may either have or not have), as an axiom.
G0 (Law of Composition) For all a, b ∈ G, a ∗ b ∈ G.
Examples.
1. The addition of integers is a law of composition on Z. Indeed, the sum of two integers
is yet another integer, and addition is the function from the set Z × Z to the set Z
that assigns a + b to the pair (a, b), denoted by (a, b) 7→ a + b.
2. The multiplication of real numbers is a law of composition on R.
3. If a, b are rational numbers, then let a ∗ b = a + b − ab. The function from Q × Q to
Q given by (a, b) 7→ a ∗ b is a law of composition on Q.
√
4. If a, b are real numbers, then define a ∗ b = a2 √+ b2 . Then (a, b) 7→ a ∗ b is not a law
of composition on Q, since 1 ∈ Q, but 1 ∗ 1 = 2 6∈ Q. However, (a, b) 7→ a ∗ b is a
law of composition on R.
5. Let n ∈ N, and let S denote the set of all matrices of size n × n with real entries.
Then matrix multiplication is a law of composition on S.

56
 2.1 Groups

6. Let n ∈ N, and let GL(n, R) denote the set of all invertible matrices of size n × n
with real entries. Then matrix multiplication is a law of composition on GL(n, R).
Indeed, if A, B ∈ GL(n, R), then the matrix AB is again a matrix of size n × n with
real entries, and moreover, since A and B are invertible, it follows that AB is also
invertible.

7. Let a, b be real numbers such that a < b. Let C[a, b] denote the set of all continuous
functions on the interval [a, b]. Let addition of functions be defined as follows: if f, g
belong to C[a, b], then

(f + g)(x) = f (x) + g(x), x ∈ [a, b].

Then addition of functions is a law of composition on C[a, b], since the sum of
continuous functions is again continuous; see Theorem 1.3.2.

8. If a, b ∈ N, then let
a
a∗b= .
b
∗ is not a law of composition on N, since 1 ∗ 2 = 12 6∈ N. The function (a, b) 7→ ab is
also not a law of composition on Q, since (1, 0) is not mapped to any rational number
by the function.

On a set there may be many different laws of compositions that can be defined. Some
laws of compositions are nicer than others, that is, they possess some desirable properties.
A group is a set G together with a law of composition on G that has three such desirable
properties, and we give the definition below.

Definition. A group is a set G together with a law of composition (a, b) 7→ a∗b : G×G →
G, which has the following properties:

G1 (Associativity axiom) For all a, b, c ∈ G, (a ∗ b) ∗ c = a ∗ (b ∗ c).

G2 (Identity axiom) There exists an element e ∈ G such that for all a ∈ G, a ∗ e = a =

e ∗ a. Such an element e is called an identity element of the group G.

G3 (Inverse axiom) For every a ∈ G, there exists an element a−1 ∈ G such that
a ∗ a−1 = e = a−1 ∗ a. Such an element a−1 is called an inverse of the element a
in the group G.

G1, G2, G3 are called the group axioms. Sometimes we use the notation (G, ∗) for the
group.

Remark. Note that hidden in the definition of a group is the axiom G0 stating that ∗ is
actually a law of composition on G. Hence when checking that a certain set G is a group
with respect to a certain operation ∗ that combines pairs of elements from G, we have
to check first of all that the axiom G0 is satisfied as well: for every a, b ∈ G, a ∗ b
belongs to G.

57
2 Algebra

Examples.

1. Z with addition is a group, as may be seen by checking each of the axioms G0, G1,
G2 and G3:
G0. For all a, b ∈ Z, a + b ∈ Z.
G1. For all a, b, c ∈ Z, (a + b) + c = a + (b + c).
G2. 0 serves as an identity element: for all a ∈ Z, a + 0 = a = 0 + a.
G3. If a ∈ Z, then −a ∈ Z and a + (−a) = 0 = −a + a.

2. R with multiplication is not a group. Indeed, although the multiplication of real

numbers satisfies G0, G1 and G2, the group axiom G3 does not hold:
G0. For all a, b ∈ R, ab ∈ R.
G1. For all a, b, c ∈ R, (ab)c = a(bc).
G2. If e is an identity element, then we must have ae = a = ea for all a ∈ R, and in
particular, with a = 1, we should have 1e = 1, and so, e = 1. And so if e is an
identity element, then it must necessarily be equal to 1.
We then check that 1 serves as an identity element: for all a ∈ R, a1 = a = 1a.
G3. Does not hold, since 0 ∈ R, but for any potential inverse b ∈ R we find that
0b = b0 = 0 6= 1 = e, a contradiction. So there is no inverse of the element 0 ∈ R.
However, the set R∗ = R \ {0} of non-zero real numbers with multiplication forms a
group:
G0. For all a, b ∈ R, if in addition a, b 6= 0, then ab ∈ R as well as ab 6= 0.
G1. For all a, b, c ∈ R∗ , (ab)c = a(bc), since it already holds for all a, b, c ∈ R.
G2. 1 ∈ R∗ is the identity element, since for all a ∈ R∗ , a1 = a = 1a. (Note that 1 is
still in the smaller set R∗ .)
G3. Let a ∈ R∗ . To find out if a has an inverse x, we have to solve the equations
ax = xa = 1. It is clear that x = a1 (which exists, since a 6= 0).
Thus we have for any a ∈ R∗ that a−1 = 1
a
serves as an inverse, since a · 1
a
=
1
a
· a = 1.
Similarly, the set (0, ∞) of positive real numbers with multiplication is also a group,

3. For n ∈ N, let GL(n, R) denote the set of all invertible matrices of size n × n with
real entries. Then GL(n, R) is group with matrix multiplication:
G0. For all A, B ∈ GL(n, R), since A and B are invertible n × n matrices, det A 6= 0
and det B = 6 0. Therefore, det(AB) = (det A)(det B) 6= 0, which gives that AB
is also an invertible n × n matrix. That is, AB ∈ GL(n, R).
G1. For all A, B, C ∈ GL(n, R), (AB)C = A(BC), since matrix multiplication is
associative.

58
 2.1 Groups

G2. The identity matrix

 
1 0 ... 0

 0 1 ... 0 

In =  .... . . .. 

 . . . . 

0 0 ... 1

serves as an identity element. In is an invertible matrix of size n × n with real

entries, and so it belongs to GL(n, R), and moreover, for all A ∈ GL(n, R),
AIn = A = In A.
G3. If A ∈ GL(n, R), then A is an invertible matrix, and so there exists a matrix
A−1 such that AA−1 = In = A−1 A. The matrix A−1 is thus in GL(n, R), and
serves as an inverse of A.
This group is called the general linear group.

4. Let a, b be real numbers such that a < b. Then C[a, b] is a group with addition of
functions:
G0. For all f, g ∈ C[a, b], f + g ∈ C[a, b] since the sum of continuous functions is
continuous.
G1. For all f, g, h ∈ C[a, b], and any x ∈ [a, b] we have

(f + (g + h))(x) = f (x) + (g + h)(x)

= f (x) + (g(x) + h(x))
= (f (x) + g(x)) + h(x)
(since addition is associative in R!)
= (f + g)(x) + h(x)
= ((f + g) + h)(x).

Hence f + (g + h) = (f + g) + h.
G2. The constant function 0 defined by 0(x) = 0 for all x ∈ [a, b], serves as an identity
element. 0 is a continuous function on [a, b] and so 0 ∈ C[a, b]. Moreover, for all
f ∈ C[a, b], we have for all x ∈ [a, b]:

(f + 0)(x) = f (x) + 0(x)

= f (x) + 0
= f (x) (0 is an identity element for addition in R)
= 0 + f (x)
= 0(x) + f (x)
= (0 + f )(x).

Hence f + 0 = f = 0 + f .

59
2 Algebra

G3. If f ∈ C[a, b], then define −f by (−f )(x) = −f (x), for x ∈ [a, b]. Given
f ∈ C[a, b], we have for all x ∈ [a, b]:

(f + (−f ))(x) = f (x) + (−f )(x)

= f (x) + (−f (x))
= 0 = 0(x) = 0
= −f (x) + f (x)
= (−f )(x) + f (x)
= (−f + f )(x).

Hence f + (−f ) = 0 = −f + f .

5. For any set S, its power set P(S) is defined to be the set consisting of all subsets of
S:
P(S) = {A | A ⊆ S} .
Define the following law of composition on P(A), called the symmetric difference
operation:
A 4 B = (A \ B) ∪ (B \ A), A, B ∈ P(S).

A B

Then P(S) with 4 is a group:

G0. For any A, B ∈ P(S) we have A, B ⊆ S, and therefore, A \ B, B \ A ⊆ S and
A 4 B = (A \ B) ∪ (B \ A) ⊆ S. Thus A 4 B ∈ P(S).
G1. For any A, B, C ∈ P(S) we have to check that A 4 (B 4 C) = (A 4 B) 4 C.
It is easily checked that both expressions correspond to the grey part in the
following Venn diagram:

A B

G2. Since A 4 ∅ = A = ∅ 4 A for all A ∈ P(S), the empty set ∅ serves as an identity
element.
G3. Since A 4 A = ∅, each element is its own inverse: A−1 = A for all A ∈ P(S).

60
 2.1 Groups

6. Given n ∈ N, let Sn be the set of all bijections from the set {1, 2, . . . , n} to itself,
and let ◦ denote the composition of functions, so α ◦ β means the function with
(α ◦ β)(x) = α(β(x)). Then (Sn , ◦) is a group, called the symmetric group on n
symbols.
G0. You have seen that the composition of two bijections is a bijection, so indeed if
α and β are in Sn then α ◦ β is also in Sn .
G1. If α, β, γ are three elements of Sn , and x ∈ {1, . . . , n}, then
   
(α ◦ β) ◦ γ (x) = (α ◦ β)(γ(x)) = α(β(γ(x))) = α((β ◦ γ)(x)) = α ◦ (β ◦ γ) (x),

and so (α ◦ β) ◦ γ = α ◦ (β ◦ γ) for all α, β, γ ∈ Sn .

G2. The identity function ι, with ι(x) = x for each x, is a bijection, and evidently
ι ◦ α = α = α ◦ ι for each α ∈ Sn .
G3. If α ∈ Sn is a bijection, then the inverse function α−1 is also a bijection, and
serves as an inverse to α in (Sn , ◦).

Definitions. A group (G, ∗) is said to be a finite group if the set G has finite cardinality.
The order of a finite group (G, ∗) is the cardinality of G. A group is said to be an
infinite group if it is not finite.

Examples.

1. The set {−1, 1} with multiplication is a finite group of order 2.

2. The set Z with addition is an infinite group.

3. The set P(S) with the symmetric difference operation 4 is a finite group if S is a
finite set, and an infinite group if S is an infinite set.

A finite group can be completely described by writing its group table.

Definition. The group table of a finite group is a table that displays the law of
composition as follows: the elements of the group are listed in the first row and the
first column. Conventionally, the two lists have the group elements in the same order,
with the identity element first. Given a, b ∈ G, the element a ∗ b is entered in the row
corresponding to a and the column corresponding to b, as shown below.
∗ ··· b ···
..
.
a a∗b
..
.
We clarify this further by considering a few examples.

61
2 Algebra

Examples.
1. The finite group {−1, 1} with multiplication can be described by the group table given
below.
· 1 −1
1 1 −1
−1 −1 1

The table completely describes the law of composition: 1 · 1 = 1, 1 · (−1) = −1,

−1 · 1 = −1 and −1 · (−1) = 1.
2. The group S3 of bijections from the set {1, 2, 3} to itself has 3! = 6 elements, listed as
follows:
x 1 2 3 x 1 2 3 x 1 2 3
ι(x) 1 2 3 α(x) 1 3 2 β(x) 3 2 1
x 1 2 3 x 1 2 3 x 1 2 3
γ(x) 2 1 3 δ(x) 2 3 1 ε(x) 3 1 2

We can now form the group table for (S3 , ◦). For example, since
α ◦ β(1) = α(β(1)) = α(3) = 2,
α ◦ β(2) = α(β(2)) = α(2) = 3,
α ◦ β(3) = α(β(3)) = α(1) = 1,
we obtain that α ◦ β = δ. In a similar way, the whole group table may be calculated:
◦ ι α β γ δ ε
ι ι α β γ δ ε
α α ι δ ε β γ
β β ε ι δ γ α
γ γ δ ε ι α β
δ δ γ α β ε ι
ε ε β γ α ι δ

3. The finite group P({1, 2}) with the symmetric difference operation 4 has four elements:
P({1, 2}) = {∅, {1}, {2}, {1, 2}}
and has the following group table:
4 ∅ {1} {2} {1, 2}
∅ ∅ {1} {2} {1, 2}
{1} {1} ∅ {1, 2} {2}
{2} {2} {1, 2} ∅ {1}
{1, 2} {1, 2} {2} {1} ∅
Again the table completely describes the law of composition. For example,
{1, 2} 4 {2} = ({1, 2} \ {2}) ∪ ({2} \ {1, 2}) = {1} ∪ ∅ = {1}.

62
 2.1 Groups

2.1.2 Proving theorems, laws of exponents and solving equations in

groups
We now prove a few elementary theorems concerning groups.

Theorem 2.1.1. There is a unique identity element in a group.

Proof. Let e and e0 be identity elements in (G, ∗). Since e ∈ G and e0 is an identity, we
obtain
e = e ∗ e0 .
Moreover, since e0 ∈ G and e is an identity, we also have

e ∗ e0 = e0 .

Consequently, e = e0 .

Theorem 2.1.2. Let (G, ∗) be a group and let a ∈ G. Then a has a unique inverse.

Proof. Let the group have the identity e. If a1 and a2 are inverses of a, then we have

a1 = a1 ∗ e (since a1 ∈ G and e is the identity)

= a1 ∗ (a ∗ a2 ) (since a2 is an inverse of a)
= (a1 ∗ a) ∗ a2 (associativity)
= e ∗ a2 (since a1 is an inverse of a)
= a2 (since a2 ∈ G and e is the identity).

Examples.

1. Let a and b be elements of a group (G, ∗). Find all x ∈ G such that:

a ∗ x = b.

Solution. Left-multiply both sides of the equation by a−1 (which exists by G3 and
satisfies a ∗ a−1 = e = a−1 ∗ a where e is the identity element of G):

a−1 ∗ (a ∗ x) = a−1 ∗ b.
Then by G1, (a−1 ∗ a) ∗ x = a−1 ∗ b,
by G3, e ∗ x = a−1 ∗ b,
and finally by G2, x = a−1 ∗ b.

We must now check that x = a−1 ∗ b is a solution to the equation a ∗ x = b, i.e., that
a ∗ (a−1 ∗ b) = b. This follows via a very similar calculation.

63
2 Algebra

2. Use G1 to show that (a ◦ b) ◦ (c ◦ d) = (a ◦ (b ◦ c)) ◦ d for any elements a, b, c, d of a

group (G, ◦).

Solution. Note in each step how the associativity axiom (G1) is applied:

(a ◦ b) ◦ (c ◦ d) = a ◦ (b ◦ (c ◦ d))
= a ◦ ((b ◦ c) ◦ d)
= (a ◦ (b ◦ c)) ◦ d.

In the previous example we saw that by the associativity axiom, two ways of using
parentheses on a ◦ b ◦ c ◦ d resulted in equal expressions. It can be shown that in
general, for any elements a1 , a2 , . . . , an of a group (G, ∗), any two ways of parenthesising
a1 ∗ a2 ∗ · · · ∗ an will result in equal elements in the group. We therefore in general do
not show the parentheses. Instead we write a ◦ b ◦ c ◦ d or a1 ∗ a2 ∗ · · · ∗ an , or even just
abcd or a1 a2 . . . an , when it is clear which law of composition is meant.

Definitions. Let G be a group and let a ∈ G. We define

a0 = e and an = an−1 ∗ a for n ∈ N.

Moreover, if n ∈ N, define
a−n = (an )−1 .

It can be shown that the usual laws of exponents hold: for all m, n ∈ Z,

am ∗ an = am+n and (am )n = amn .

Remark. It is not necessarily true that (a ∗ b)n = an ∗ bn .

Example. Let b be an element of a group G with identity element e. Find all solutions
x to the following simultaneous equations:

x2 = b and x5 = e.

Solution. Square the first equation: b2 = (x2 )2 = x4 . Left-multiply by x: xb2 = xx4 =

x5 = e by the second equation. Right multiply by b−2 : xb2 b−2 = eb−2 . Simplifying,
xb2−2 = xb0 = xe = x and eb−2 = b−2 . Therefore, x = b−2 .
So far, we have shown that, if there is a solution x to the equations, then that solution
is x = b−2 . We must now ask whether this is indeed a solution: substituting x = b−2
into the original equations yields b−4 = b and b−10 = e: the first of these is equivalent
to b5 = e, and if this is the case then also b−10 = e. The conclusion is that, if b5 = e,
then x = b−2 = b3 is the unique solution to the equations, whereas if b5 6= e, then the
equations have no solutions.

64
 2.1 Groups

2.1.3 Abelian groups

The groups in Examples 1, 2, 4 and 5 on page 58 above also satisfy the following axiom:
G4 (Commutativity axiom) For all a, b ∈ G, a ∗ b = b ∗ a.
On the other hand, the group in Example 3 does not satisfy G4 for any n ≥ 2. For
example, when n = 2:
" #" # " # " # " #" #
1 1 1 0 2 1 1 1 1 0 1 1
= 6= = .
0 1 1 1 1 1 1 2 1 1 0 1

This gives rise to the following natural definition.

Definition. A group (G, ∗) is said to be an abelian group1 if its law of composition is
commutative, that is, for all a, b ∈ G, a ∗ b = b ∗ a.
Examples.
1. The group (Z, +) of integers with addition is an abelian group.

2. The group (R∗ , ×) of non-zero real numbers with multiplication is an abelian group.

3. Let m, n ∈ N. The set of matrices Rm×n of size m × n with entries in R with matrix
addition is an abelian group.

4. Let n ∈ N. The set GL(n, R) with matrix multiplication is a group, but it is not an
abelian group if n ≥ 2.

5. For any S, the group (P(S), 4) is an abelian group.

6. The symmetric group S3 is not an abelian group. For instance, referring to the group
table given earlier, we see that α ◦ β = δ, but β ◦ α = ε 6= δ.
We normally use additive notation when referring with abelian groups. We then
use a symbol such as ‘+’, ‘⊕’, ‘
’, etc., to denote the law of composition,

use a symbol such as 0, o, etc., to denote the identity element,

and denote the inverse of an element x by −x.

Furthermore, instead of writing am for a + a + · · · + a, we use the more natural-looking
notation m · a. As before, we extend it also to m = 0 by writing 0 · a = 0 (where the 0
on the right-hand side is the identity element of the group) and to m < 0, where we still
write m · a instead of am . In particular, in this notation we then have −x = (−1) · x.
This notation is nothing more than a new notation for am in the case of abelian
groups where additive notation is used. The laws of exponents look as follows in additive
notation:
m · a + n · a = (m + n) · a and m · (n · a) = (mn) · a.
1
after the Norwegian mathematician Niels Henrik Abel (1802–1829).

65
2 Algebra

2.1.4 Subgroups
Definition. A subset H of a group G, with operation ∗, is called a subgroup of G if it
has the following properties:

H1 (Closure) If a, b ∈ H, then a ∗ b ∈ H.

H2 (Identity) e ∈ H.

H3 (Inverses) If a ∈ H, then a−1 ∈ H.

The three properties of a subgroup H ensure that (H, ∗) is itself a group, as we now
explain.

H1. This condition tells us that the law of composition ∗ on the group G can be used
to define a law of composition on H, namely, the function from H × H to H
given by (a, b) 7→ a ∗ b, which is called the induced law of composition. Since ∗ is
associative, it follows that the induced law of composition is associative as well: for
all a, b, c ∈ H, a ∗ (b ∗ c) = (a ∗ b) ∗ c.

H2. This shows that H, with the induced law of composition, has an identity element.
Indeed the identity element from G (which also belongs to H) serves as the identity
element also in H: for every a ∈ H, a ∗ e = a = e ∗ a.

H3. Finally this shows that every element in H possesses an inverse element in H. Of
course, since G is a group, we already knew that a possesses an inverse element
a−1 ∈ G. But now H3 says that this inverse element is in H. Thus for all a ∈ H,
∃a−1 ∈ H such that a ∗ a−1 = e = a−1 ∗ a.

Thus the conditions H1, H2, H3 imply that the subset H, with the induced law of
composition, is a group. Thus, a subgroup is itself a group which sits in a larger group.

Examples.

1. The subset of even integers {2m | m ∈ Z} is a subgroup of the group of integers Z

with addition. Indeed, the sum of even numbers is even (and so H1 holds), 0 is even
(and so H2 holds), and finally, given the even number 2m, −2m = 2(−m) is even as
well (and so H3 holds).

2. The group of integers with addition (Z, +) is a subgroup of the group of rational
numbers with addition (Q, +), which in turn is a subgroup of the group of real numbers
with addition (R, +).

3. If G is a group with identity e, then {e} and G are both subgroups of G.

66
 2.1 Groups

4. The subset of 2 × 2 symmetric matrices with real entries, namely

(" # )
a b 
a, b, d ∈ R

b d



is a subgroup of the set of all 2 × 2 matrices with real entries together with matrix
addition. Indeed, given any two symmetric matrices
" # " #
a b a0 b 0
and ,
b d b0 d0
their sum " # " # " #
a b a0 b 0 a + a0 b + b 0
+ =
b d b0 d 0 b + b0 d + d0
is also symmetric, and so H1 holds. Clearly the identity element
" #
0 0
0 0
is symmetric, and so H2 also holds. Finally, the inverse (with respect to matrix
addition) of any symmetric matrix
" #
a b
,
b d
is the element " #
−a −b
−b −d
which is also symmetric, and so H3 holds.
5. The subset of upper triangular invertible matrices,
(" # )
a b 
U T (2, R) = 
a, b, d ∈ R and ad 6= 0
0 d



is a subgroup of the group GL(2, R) with matrix multiplication. Indeed, if

" # " #
a b a0 b 0
, ∈ U T (2, R),
0 d 0 d0
then " #" # " #
a b a0 b 0 aa0 ab0 + bd0
= ∈ U T (2, R),
0 d 0 d0 0 dd0
and so H1 holds. Clearly " #
1 0
∈ U T (2, R),
0 1
and so H2 also holds. Finally,
" # " #−1 " #
1 b
a b a b − ad
if ∈ U T (2, R), then = a
1 ∈ U T (2, R).
0 d 0 d 0 d

67
2 Algebra

2.1.5 The order of an element of a group

Definitions. Let G be a group and suppose that a ∈ G.

1. If there exists an m ∈ N such that am = e, then a is said to have finite order.

2. If a has finite order, then the order of a, denoted by ord(a), is

ord(a) = min {m ∈ N | am = e} .

3. If a does not have finite order, then a is said to have infinite order.

Examples.

1. The element −1 has order 2 in the group of non-zero real numbers with multiplication.

2. The element 2 has infinite order in the group of integers with addition.
 
0 0 1
3. The element  1 0 0  is an element of order 3 in the group GL(3, R).
 

0 1 0

We now prove that given any element from a group, the set of its powers forms a
subgroup of the group.

Theorem 2.1.3. Suppose that (G, ∗) is a group and a ∈ G. Let hai = {an | n ∈ Z}.
Then:

1. hai is a subgroup of G. Moreover, hai is the smallest subgroup of G containing a.

2. If a is an element with finite order m, then hai = {e, a, a2 , a3 , . . . , am−1 }.

Proof.

1. We prove that H1, H2, H3 hold.

H1. Given m, n ∈ Z, clearly m + n ∈ Z and so am ∗ an = am+n ∈ hai.
H2. e = a0 ∈ hai.
H3. For any m ∈ Z, −m ∈ Z and so (am )−1 = a−1·m = a−m ∈ hai.
So hai is a subgroup of G.
Conversely, if H is any subgroup of G containing the element a, then:
(a) a0 = e ∈ H by H2.
(b) Since a ∈ H, so is a ∗ a = a2 by H1, and similarly a3 , a4 , . . . are all in H.
(c) By H3, all the inverses a−n = (an )−1 are also in H.
So H includes all the elements an , for n ∈ Z, in other words hai ⊆ H.

68
 2.1 Groups

2. Clearly {e, a, a2 , a3 , . . . , am−1 } ⊂ hai. Conversely, if n ∈ Z, then there exist integers q

and r, such that 0 ≤ r ≤ m − 1, and n = q · m + r. So

an = aq·m+r = aq·m ∗ ar = (am )q ∗ ar = (e)q ∗ ar

= e ∗ ar = ar ∈ {e, a, a2 , a3 , . . . , am−1 }.

Example. The element [1] in the group Z5 with addition modulo 5 has finite order, since

[1] ⊕ [1] ⊕ [1] ⊕ [1] ⊕ [1] = [0],

and the subgroup h[1]i = {[0], [1], [2], [3], [4]} is in fact the whole group.
The above example motivates the following definition.
Definition. A group G is said to be a cyclic group if there exists an element a ∈ G
such that G = hai. Such an element a is then called a generator of the group G.

2.1.6 Homomorphisms and isomorphisms

The study of algebra is not just about structures with algebraic operations on them, but
also about functions between such structures, that in an obvious sense “respect” the
algebraic operations. The following definition illustrates this idea in the case of groups.
Definition. Let (G, ∗) and (G0 , ∗0 ) be groups. A homomorphism ϕ : G → G0 is a
function such that
for all a, b ∈ G, ϕ(a ∗ b) = ϕ(a) ∗0 ϕ(b).
Examples.
1. Let G be R with addition, and G0 the set of positive reals with multiplication. The
exponential function from G to G0 , given by x 7→ 2x , is a homomorphism. Indeed, we
have 2x+y = 2x 2y for all real x, y.

2. Let G be the group GL(2, R) with matrix multiplication, and G0 be the group of non-
zero real numbers with multiplication. Then the determinant function det : G → G0 is a
homomorphism, since for all 2×2 real matrices A, B we have det(AB) = det(A) det(B).

3. Let G be the group C[0, 1] of continuous functions on the interval

  [0, 1] with addition,
0
and G be the group R with addition. Then the function f 7→ f 12 is a homomorphism.
 
1
Indeed, if f, g are continuous functions on the interval [0, 1], then (f + g) 2
=
   
1 1
f 2
+g 2
, by the definition of f + g.

4. If G is a group, then the identity map ι : G → G defined by ι(a) = a for all a ∈ G,

and the trivial map ζ : G → G defined by ζ(a) = e for all a ∈ G are homomorphisms.
Thus a homomorphism is a function between two groups that respects the law of
composition. In the next theorem we show that it preserves the identity element and the
inverses of elements as well.

69
2 Algebra

Theorem 2.1.4. Let G be a group with identity e and G0 be a group with identity e0 . If
ϕ : G → G0 is a homomorphism, then:

1. ϕ(e) = e0 .

2. If a ∈ G, then (ϕ(a))−1 = ϕ(a−1 ).

Proof. We have
e0 ∗0 ϕ(e) = ϕ(e) = ϕ(e ∗ e) = ϕ(e) ∗0 ϕ(e),
and so cancelling ϕ(e) on both sides, we obtain e0 = ϕ(e). Next,

ϕ(a−1 ) ∗0 ϕ(a) = ϕ(a−1 ∗ a) = ϕ(e) = e0

and similarly,
e0 = ϕ(e) = ϕ(a ∗ a−1 ) = ϕ(a) ∗0 ϕ(a−1 ).
Thus ϕ(a−1 ) ∗0 ϕ(a) = e0 = ϕ(a) ∗0 ϕ(a−1 ), and by the uniqueness of the inverse of ϕ(a)
in G0 , we obtain (ϕ(a))−1 = ϕ(a−1 ).
Every group homomorphism ϕ determines two important subgroups: its image and its
kernel.

Definitions. Let G, G0 be groups and let ϕ : G → G0 be a group homomorphism.

1. The kernel of ϕ is the set ker(ϕ) = {a ∈ G | ϕ(a) = e0 }.

2. The image of ϕ is the set im(ϕ) = {a0 ∈ G0 | ∃a ∈ G such that ϕ(a) = a0 }.

Using Theorem 2.1.4, we now prove the following result.

Theorem 2.1.5. Let G, G0 be groups and let ϕ : G → G0 be a group homomorphism.

Then:

1. ker(ϕ) is a subgroup of G.

2. im(ϕ) is a subgroup of G0 .

Proof. If a, b belong to ker(ϕ), then ϕ(a ∗ b) = ϕ(a) ∗0 ϕ(b) = e0 ∗0 e0 = e0 , and so H1

holds. Moreover, as ϕ(e) = e0 , e ∈ ker(ϕ) and so H2 holds too. Finally, H3 holds, since if
a ∈ ker(ϕ), then ϕ(a−1 ) = (ϕ(a))−1 = (e0 )−1 = e0 , and so a−1 ∈ ker(ϕ). Hence ker(ϕ) is
a subgroup of G.
We now also check that im(ϕ) is a subgroup of G0 . If a0 , b0 belong to im(ϕ), then there
exist elements a, b in G such that ϕ(a) = a0 and ϕ(b) = b0 . Consequently, ϕ(a ∗ b) =
ϕ(a) ∗0 ϕ(b) = a0 ∗0 b0 , and so there exists an element in G, namely a ∗ b, such that
ϕ(a ∗ b) = a0 ∗0 b0 , that is, a0 ∗0 b0 ∈ im(ϕ). Thus H1 holds. Since ϕ(e) = e0 , it follows that
e0 ∈ im(ϕ). Finally, if a0 ∈ im(ϕ), then there exists an a ∈ G such that ϕ(a) = a0 , and so
a0−1 = (ϕ(a))−1 = ϕ(a−1 ). Hence a0−1 ∈ im(ϕ), and H3 holds. So im(ϕ) is a subgroup
of G0 .

70
 2.1 Groups

Examples.
1. Let G be R with addition, and G0 be the set of positive reals with multiplication. The
exponential function from G to G0 , given by x 7→ 2x , is a homomorphism with kernel
the trivial subgroup comprising the element 0, and the image is the whole group of
positive reals with multiplication, since it can be shown that given any y > 0, there
exists a unique real number (called the logarithm of y to the base 2, denoted by log2 y)
such that y = 2log2 y .
2. Let G be the group GL(2, R) with matrix multiplication, and G0 be the group of non-
zero real numbers with multiplication. Then the determinant function det : G → G0
is a homomorphism. Its kernel is the set of all invertible matrices with determinant
equal to 1, and we denote this subgroup by SL(2, R), and it is called the special
linear group:
SL(2, R) = {A ∈ GL(2, R) | det(A) = 1} .
The image of this homomorphism is the whole group of non-zero reals: indeed, given
any real number a not equal to zero, we have that
" #
1 0
A := ∈ GL(2, R),
0 a

and det(A) = 1 · a − 0 · 0 = a.
3. Let G be the group C[0, 1] of continuous functions on the interval
  [0, 1] with addition,
0
and G be the group R with addition. Then the function f 7→ f 12 is a homomorphism,
and its kernel is the set of all continuous functions on the interval [0, 1] that have
a root at 12 (for instance the straight line f (x) = x − 12 belongs to the kernel). The
image is the set of all real numbers, since given  any a ∈ R, the constant function
f (x) = a for all x ∈ [0, 1] is continuous, and f 12 = a.
In Example 1 above, the homomorphism between the two groups was also bijective.
We give a special name to such homomorphisms.
Definition. Let G, G0 be groups. A homomorphism ϕ : G → G0 is said to be an iso-
morphism if it is bijective.
Examples.
1. Let G be R with addition, and G0 be the set of positive reals with multiplication. The
exponential function from G to G0 , given by x 7→ 2x , is an isomorphism.
2. If G is a group, then the identity function ι : G → G defined by ι(a) = a for all a ∈ G
is an isomorphism.
" #
1 x
3. Let G be the subgroup of GL(2, R) comprising all matrices of the form ,
0 1
where x ∈" R. Let# G0 be the group R with addition. Then the function from G to G0 ,
1 x
given by 7→ x, is an isomorphism.
0 1

71
2 Algebra

Isomorphisms are important because their existence between two groups means that
the two groups are essentially the “same”, in the sense that as far as algebraic properties
go, there is no real difference between them.

Definition. Two groups G and G0 are called isomorphic if there exists an isomorphism
ϕ : G → G0 .

2.1.7 Cosets and Lagrange’s theorem

The main purpose of this final section on group theory is to establish perhaps the most
important result on the topic, Lagrange’s Theorem. This states that, if H is a subgroup
of a finite group, then the order of H divides the order of G. One consequence, very
useful in itself, is that the order of any element of a finite group divides the order of the
group.
We start by developing a key tool in the proof of Lagrange’s Theorem.

Definition. Given a subgroup H of a group G, define the relation R on G by

a R b if b = a ∗ h for some h ∈ H. (2.1)

Then R is an equivalence relation:

E1 (Reflexivity). For all a ∈ G, a R a since e ∈ H and a = a ∗ e.

E2 (Symmetry). For all a, b in G, if a R b, then there exists a h ∈ H such that b = a ∗ h

and so b ∗ h−1 = a. Since H is a subgroup, and h ∈ H, it follows that h−1 ∈ H.
Thus b R a.

E3 (Transitivity). For all a, b, c in G, if a R b and b R c, then there exist elements h1 , h2

in H such that b = a ∗ h1 and c = b ∗ h2 . Hence we obtain c = b ∗ h2 = (a ∗ h1 ) ∗ h2 =
a ∗ (h1 ∗ h2 ). Since h1 , h2 ∈ H and H is a subgroup, it follows that h1 ∗ h2 ∈ H,
and so a R c.

Definition. Let H be a subgroup of a group G, and let R be the equivalence relation

given by (2.1). If a ∈ G, then the equivalence class of a, namely the set

{x ∈ G | a R x} = {x ∈ G | ∃h ∈ H such that x = a ∗ h} = {a ∗ h | h ∈ H} ,

is called a left coset of H, and is denoted by a ∗ H.

We know that the equivalence classes of an equivalence relation partition the set. Recall
that by a partition of a set S, we mean a subdivision of the set S into nonoverlapping
subsets:

72
 2.1 Groups

G b∗H

a ∗ H = a0 ∗ H
(a R a0 )

e∗H =H

Hence we obtain the following result:

Corollary 2.1.6. Let H be a subgroup of a group G. Then the left cosets of H partition
the group G.

Remarks.

1. The notation a ∗ H denotes a certain subset of G. As with any equivalence relation,

different notations may represent the same subset. In fact, we know that a ∗ H is the
unique left coset containing a, and so

a ∗ H = b ∗ H iff a R b.

Corollary 2.1.6 says that if a ∗ H and b ∗ H have an element in common then they are
equal.

2. Indeed, the following are all equivalent:

a R b; b R a; a ∗ H = b ∗ H; a ∈ b ∗ H; b ∈ a ∗ H.

3. One can also define the relation R0 on G by a R0 b if b = h ∗ a for some h ∈ H. The

associated equivalence classes are called right cosets.

Examples.

1. Consider the group (Z6 ⊕) of integers modulo 6, with addition modulo 6, and let H be
the subgroup h[2]i = {[0], [2], [4]}. The left cosets, which we now denote by a ⊕ H are

[0] ⊕ H = [2] ⊕ H = [4] ⊕ H = {[0], [2], [4]}, and

[1] ⊕ H = [3] ⊕ H = 5 ⊕ H = {[1], [3], [5]}.

Note that the cosets {[0], [2], [4]} and {[1], [3], [5]} form a partition of G:

G = {[0], [1], [2], [3], [4], [5]} = {[0], [2], [4]} ∪ {[1], [3], [5]},
and {[0], [2], [4]} ∩ {[1], [3], [5]} = ∅.

73
2 Algebra

2. Consider the group (Z, +), and let H be the subgroup of even numbers {2m | m ∈ Z}.
The left cosets, which we now denote by a + H are

· · · = −4 + H = −2 + H = 0 + H
= {2m | m ∈ Z} = {. . . , −4, −2, 0, 2, 4, . . . }
= 2 + H = 4 + H = 6 + H = ···,

and

· · · = −3 + H = −1 + H = 1 + H
= {2m + 1 | m ∈ Z} = {. . . , −3, −1, 1, 3, . . . }
= 3 + H = 5 + H = 7 + H = ···.

Note that the cosets {. . . , −4, −2, 0, 2, 4, . . . } and {. . . , −3, −1, 1, 3, . . . } do indeed
partition the set of all integers.

Note that in Example 1 above, there are only two distinct cosets, and

|G| = 6 = 3 · 2 = |H| · (number of cosets of H).

In particular the order of H (namely 3) divides the order of G (namely 6). This is not a
coincidence. We now prove an important result concerning the order of a group G and
the number of cosets of a subgroup H, due to Lagrange2 .
Theorem 2.1.7 (Lagrange’s Theorem). Let H be a subgroup of a finite group G. Then
the order of H divides the order of G.
Proof. We claim that, for any a ∈ G, the function φ from the subgroup H to the coset
a ∗ H, given by φ(h) = a ∗ h, for h ∈ H, is a bijection. Indeed, it is a surjection by
definition of the coset a ∗ H; to see that it is an injection, we note that φ(h) = φ(h0 ) ⇒
a ∗ h = a ∗ h0 ⇒ h = h0 .
Consequently, each coset a ∗ H has the same number of elements as H does.
Since G is the union of the cosets of H, and since these cosets do not overlap, we
obtain the counting formula

|G| = |H| · (number of cosets of H).

In particular, |H| divides |G|.

Corollary 2.1.8. Let G be a finite group and let a ∈ G. Then the order3 of a divides
the order of G. Moreover, a|G| = e.
Proof. Let a have order m. Recall that hai is a subgroup of G, so Theorem 2.1.3
implies that |hai| = m divides |G|, and so |G| = m · k for some k ∈ N. Thus also
a|G| = am·k = (am )k = ek = e.
2
Joseph Louis Lagrange (1736–1813)
3
By Exercise 24, every element in a finite group has a finite order.

74
 2.1 Groups

The following theorem characterises all groups whose order is a prime number.

Corollary 2.1.9. If G is a group with prime order p, then G is cyclic, and G = hai for
every a ∈ G \ {e}.

Proof. If a =6 e, then a has order > 1, say m. Since m divides p, and p is prime, it
follows that m = p. As G itself has order p, it now follows that G = hai, and so G is
cyclic.

2.1.8 Exercises
1. For each of the following definitions for a ∗ b and a given set, determine which of the
axioms G0, G1, G2, G3 are satisfied by a ∗ b. In which cases do we obtain a group?
a+b i) a + b − ab on the set R.
a) on the set Z.
ab
q j) max{a, b} on the set N.
b) |ab| on the set Q.
k) min{a, b} on the set Z.
c) a − b on the set Z.
l) a + b on R.
d) |a − b| on the set N ∪ {0}.
e) −ab on the set Q. m) a + b on [0, ∞).

f) −ab on the set Q \ {0}. n) ab on {2n | n ∈ Z}.

g) a + b + 1 on the set R. o) A ∪ B on P(S) where S is any set.
h) a + 2b + 4 on the set R. p) A ∩ B on P(S) where S is any set.

2. Consider the set

(" # )
a b 
2 2
S= 
a, b ∈ R and a + b 6= 0 .
−b a



Show that S with the operation of matrix multiplication forms a group.

3. a) Show that the set Z6 of integers modulo 6, with addition modulo 6 is a group.
Write down its group table.
b) Is Z6 with multiplication modulo 6 also a group? If, instead, we consider the set
Z∗6 of non-zero integers modulo 6, then is Z∗6 a group with multiplication modulo
6?
c) Let m be an integer such that m ≥ 2, and let Z∗m denote the set of non-zero
integers modulo m. Prove that Z∗m is a group with multiplication modulo m iff m
is a prime number.
Hint: If m is a prime number, then any r ∈ {1, . . . , m − 1} is coprime to m, and
so there exist integers s and t such that sm + tr = 1.

75
2 Algebra

4. Let (G, ∗) be a group.

a) Show that if a, b, c ∈ G are such that a ∗ b = a ∗ c, then b = c.
b) Show that if a, b ∈ G, then the equation a ∗ x = b has a unique solution. Explain
why this implies that every row of a group table (of a finite group) contains each
element of the group exactly once.
c) Show that if a, b ∈ G, then (a ∗ b)−1 = b−1 ∗ a−1 .

5. Let (G, ∗) be a group and a ∈ G.

a) Show that, for all n ∈ N, an = a ∗ an−1 . Use induction.
b) Show that, for all n ∈ Z, an = an−1 ∗ a.
c) Show that, for any m ∈ Z and n ∈ N, am ∗ an = am+n . Use induction on n, with
m fixed.
d) Show that, for any m ∈ Z and n ∈ N, am ∗ a−n = am−n . Use the previous part.
e) Deduce that am ∗ an = am+n for any m, n ∈ Z.

6. Let G be a group with identity element e, and let a and b be elements of G. For each
of the following pairs of equations, find all solutions x, and determine conditions on a
and b for there to be any solutions to the equations.
a) x2 a = bx and ax = xa.
b) ax2 = b and x3 = e.
c) (xax)3 = bx and x2 a = (xa)−1 .
Note that, in this and some later exercises, the symbol for the group operation is
omitted – so ab is used instead of a ∗ b.

7. Let a and b be elements of a group (G, ∗). Show that (a ∗ b)2 = a2 ∗ b2 iff a ∗ b = b ∗ a.
Give an example of a group (G, ∗) and two elements a, b ∈ G such that (a∗b)2 6= a2 ∗b2 .

8. Determine if the following statements are TRUE or FALSE: Give reasons for your
answers in each case.
In any group G with identity element e,
a) for any x ∈ G, if x2 = e then x = e.
b) for any x ∈ G, if x2 = x then x = e.
c) for any x ∈ G there exists y ∈ G such that x = y 2 .
d) for any x, y ∈ G there exists z ∈ G such that y = xz.

9. Prove the following statements for any group G with identity element e:
a) For all a, b, c ∈ G, if abc = e then cab = e and bca = e.
b) Suppose that a = a−1 , b = b−1 and c = c−1 (that is, each of a, b and c is its own
inverse). For all a, b, c ∈ G, if ab = c then bc = a and ca = b.
c) For all a ∈ G, a = a−1 iff a2 = e.

76
 2.1 Groups

10. a) Show that in any group, the inverse of the inverse of any element is itself.
b) (∗) Let (G, ∗) be a finite group with identity element e. Show that if the order
of G is even, then there exists an element x ∈ G such that x 6= e and x−1 = x.
(That is, in a group of even order there always exists an element other than the
identity which is its own inverse.)
Hint: Use 10a) to show that there is an even number of elements that are not
their own inverses.

11. Show that the general linear group GL(3, R) with matrix multiplication is not an
abelian group.

12. Prove that a group G is abelian iff (ab)−1 = a−1 b−1 for all a, b ∈ G.

13. Show that S2 is abelian. Show that Sn is non-abelian for all n ≥ 3.

14. Determine if the following statements are TRUE or FALSE. Give reasons for your
answers in each case.
a) The nonnegative integers form a subgroup of Z with addition.
b) The odd integers form a subgroup of Z with addition.
c) The set {4k | k ∈ Z} is a subgroup of Z with addition.
d) If G is abelian and H is a subgroup of G, then H is abelian.
e) Every group has an abelian subgroup.
f) If H and K are subgroups of G and H ⊆ K, then H is a subgroup of K.

15. Does there exist an infinite group with a finite subgroup?

n  o
16. Show that the set H = ab  a, b odd integers is a subgroup of the group (Q∗ , ×) of


non-zero rational numbers with multiplication.

17. For n ∈ N, show that the set H = {α ∈ Sn | α(1) = 1} is a subgroup of (Sn , ◦).

18. Show that the set H = {x ∈ R | sin x ∈ Q and cos x ∈ Q} is a subgroup of the group
(R, +) of real numbers with addition.
You may use any properties of the sine and cosine functions that you know (or can
look up).

19. Let S and T be sets such that S ⊆ T . Note that then P(S) ⊆ P(T ), that is, every
subset of S is also a subset of T . Show that P(S) is a subgroup of the group (P(T ), 4).

20. a) Let (G, +) be an abelian group. Show that the set H = {x ∈ G | x = −x} (that
is, the set of all elements of G that are inverses of themselves) is a subgroup of G.
b) Show that the set H = {θ ∈ S3 | θ = θ−1 } is not a subgroup of (S3 , ◦).

77
 2 Algebra

21. a) Show that, if H and K are subgroups of a group G, then H ∩ K is also a subgroup
of G.
b) Consider the group of integers Z with addition. Suppose that H is the subgroup of
Z comprising multiples of 4, and let K be the subgroup of Z comprising multiples
of 6. What is H ∩ K?
c) If H and K are subgroups of a group G, is H ∪ K necessarily a subgroup of G? If
true, give a proof. If false, provide a counterexample.
22. Let C[0, 1] denote the group comprising the set of continuous functions on the interval
[0, 1] with addition of functions defined pointwise: if f, g belong to C[0, 1], then for
all x ∈ [0, 1], (f + g)(x) = f (x) + g(x). Prove that each of the following subsets of
C[0, 1] are subgroups of C[0, 1].
n    o
1
a) H1 = f ∈ C[0, 1]  f =0 .

2

b) The set H2 of all polynomial functions, that is: the set of all functions p : [0, 1] → R
such that there exists an n ∈ N ∪ {0} and real numbers a0 , a1 , a2 , . . . , an such that
p(x) = a0 + a1 x + a2 x2 + · · · + an xn , for all x ∈ [0, 1].
c) H3 = {f ∈ C[0, 1] | ∀x ∈ [0, 1] f (1 − x) = −f (x)}.
23. Let G be a group. The centre of G is the set Z(G) = {z ∈ G | ∀a ∈ G, z ∗ a = a ∗ z}.
a) Show that Z(G) is not empty.
b) If G is abelian, then determine Z(G).
c) Show that Z(G) is a subgroup of G.
d) (∗) If G is the group GL(2, R) with matrix multiplication, then determine Z(G).
" # " #
1 1 1 0
Hint: Consider A = ,B = . Which elements Z of GL(2, R)
0 1 1 1
satisfy both ZA = AZ and ZB = BZ?
24. If G is a finite group, then show that every element in the group has finite order,
which is at most equal to |G|.
Hint: If a ∈ G, then consider the set S = {e, a, a2 , a3 , . . . , a|G| }, and use the pigeon-
hole principle.
" #
1 1
25. Determine the order of in the group GL(2, R).
−1 0
26. Determine the orders of all the elements of the following groups:
a) the group Z6 with addition modulo 6.
b) the symmetric group S3 .
c) the group (P(S), 4).
27. Let H be a subgroup of a group (G, ∗), and let h be an element of H. Is it necessarily
true that the order of h in (H, ∗) is the same as the order of h in (G, ∗)? Justify your
answer, by means of a proof or a counterexample.

78
 2.1 Groups

28. Prove that in any group (G, ∗), and for any a, b in G, the orders of a ∗ b and b ∗ a are
the same. (Note that it is not given that (G, ∗) is abelian.)

29. Is the group of integers with addition cyclic? What is a generator of this group? Is it
unique?

30. Consider the group (Z, +), and let a be a positive integer. What are the elements of
the subgroup hai?
Show that if a and b are two positive integers, then the smallest subgroup of (Z, +)
containing both a and b is hdi, where d = gcd(a, b).

31. Show that any cyclic group is abelian. Give an example of an abelian group that is
not cyclic.

32. Define φ : (R, +) → (C∗ , ×) by φ(x) = eix = cos x + i sin x. Show that φ is a
homomorphism, and find its kernel and image. You may use any properties you know
about the function x 7→ eix .

33. a) Let G, G0 , G00 be groups and let ϕ : G → G0 and ψ : G0 → G00 be group homomor-
phisms. Prove that the composition ψ ◦ϕ : G → G00 defined by (ψ ◦ϕ)(a) = ψ(ϕ(a)),
a ∈ G is a group homomorphism.
b) Describe the kernel of ψ ◦ ϕ.

34. Let ϕ : G → G0 be a homomorphism from a group G with identity element e to a

group G0 . Show that ϕ is an injection iff ker(ϕ) = {e}.

35. Let G be a group and let a be an element of G. Prove that the function from Z to
hai, given by m 7→ am , is a homomorphism from the group (Z, +) to G. What is the
image of this homomorphism? What is the kernel of the homomorphism: (a) if a has
finite order m, (b) if a has infinite order.

36. Show that the function ϕ(n) = 2n is an isomorphism from (Z, +) to its subgroup of
even integers {2k | k ∈ Z}.

37. Let G be the group defined on R with law of composition a ∗ b = a + b + 1. Show that
G and (R, +) are isomorphic.
Hint: Try ϕ(x) = x + 1.

38. Let G be the group defined on R \ {1} with law of composition a ∗ b = a + b − ab.
Show that G is isomorphic to the group R \ {0} with multiplication.
Hint: Try ϕ(x) = 1 − x.

39. A subgroup N of a group G is called a normal subgroup if for every a ∈ N and

every b ∈ G, then b ∗ a ∗ b−1 ∈ N . Prove that if G, G0 are groups and ϕ : G → G0 is a
homomorphism, then ker(ϕ) is a normal subgroup of G.

40. Show that the function from G to G given by a 7→ a−1 is an isomorphism iff G is
abelian.

79
 2 Algebra

41. Let G, G0 be groups and let ϕ : G → G0 be an isomorphism. Prove that the inverse
function ϕ−1 : G0 → G is also an isomorphism.

42. a) Let G and G0 be isomorphic groups with isomorphism ϕ : G → G0 .

i) Show that for any a ∈ G, a has the same order as ϕ(a) in G0 .
ii) Show that if G is finite, then G0 is also finite and |G| = |G0 |.
iii) Show that if G is abelian, then G0 is also abelian.
b) Use 42a) to show that (Z6 , ⊕) and S3 are not isomorphic.
c) Use 42a) to show that (Z4 , ⊕) and P({1, 2}) are not isomorphic.

43. Let (G, ∗) be a group. Define a new law of composition ◦ on G by a ◦ b = b ∗ a. (Thus

if (G, ∗) is abelian, ◦ is exactly the same as ∗, but we are not assuming that (G, ∗) is
abelian.)
a) Show that (G, ◦) is a group.
b) Show that the groups (G, ∗) and (G, ◦) are isomorphic.
Hint: Try a 7→ a−1 for an isomorphism.

44. Determine if the following statements are TRUE or FALSE. Give reasons for your
answers in each case.
a) If H is a subgroup of G and a, b ∈ G are such that a 6= b, then (a ∗ H) ∩ (b ∗ H) = ∅.
b) If H is a subgroup of G and a ∈ G is such that a ∗ H has 4 elements, then H has
4 elements.
c) If H is a subgroup of a finite group G, then for any a ∈ G, the left coset a ∗ H has
the same number of elements as the right coset H ∗ a.
d) All groups of order 7 are isomorphic to (Z7 , +).
e) All groups of order 6 are isomorphic to (Z6 , +).
f) The group S4 does not have an element of order 13.
g) The group S4 does not have an element of order 24.
h) (∗) The group S4 does not have an element of order 12.

45. a) Verify that (" # )

x y 
G= x, y ∈ R and x > 0

0 1



is a group with matrix multiplication.

b) Show that (" # )
x 0 
H= x ∈ R and x > 0

0 1



is a subgroup of G.

80
 2.1 Groups
" #
x y
c) An element of G can be represented by the point (x, y) in the right
0 1
half-plane {(x, y) | x, y ∈ R and x > 0}. Draw the partitions of the right half-plane
into left and into right cosets of H.

46. Let (" # )

a b 
G= a, b ∈ Z5 and a 6= 0 .

0 a−1



(a) Show that G is closed under matrix multiplication, with the addition and multi-
plication carried out in Z5 .
" #
a b
(b) Find the inverse of the matrix in (G, ⊗), where ⊗ denotes matrix
0 a−1
multiplication as in (a).
You may assume without further proof that (G, ⊗) is a group.
(c) What is the order of G?
(d) Let H be the subset of G consisting of the diagonal matrices. Show that H is a
subgroup of G.
" #
2 1
(e) Let m = . Find all the elements of the left coset m ⊕ H.
0 3

47. Let H and K be subgroups of a group G of orders 3 and 5 respectively. Prove that
H ∩ K = {e}.

48. a) Let p be a prime, and let (Z∗p , ⊗) be the group of non-zero integers modulo p, with
multiplication modulo p. Show that, if a is an integer such that a is not divisible
by p, then [a]p−1 = [1].
b) Prove Fermat’s little theorem: for any integer a, ap ≡ a (mod p).
Hint: If a ∈ Z is not divisible by p, then [a] 6= [0], and so by part (48a) above,
[a]p−1 = [1]. Hence p | (ap−1 − 1), and so p | (ap − a).
c) Show that 7 divides 22225555 + 55552222 .
Hint: Note that 2222 = 7 · 317 + 3, so that in Z7 , [22225555 ] = [3]5555 . Now use
the fact that [3]6 = [1] to conclude that [22225555 ] = [35 ]. Now simplify [55552222 ]
in a similar fashion.

81
2 Algebra

49. Sample Exam Question. 2012 Q7.

(a) Let G = R \ {0}, and for each a, b ∈ G define a ∗ b = −ab.
(i) Show that (G, ∗) is an Abelian group.
(ii) Define the function ϕ : (G, ∗) → (G, ∗) by
x
ϕ(x) = .
|x|

Show that ϕ is a homomorphism. Determine ker(ϕ) and im(ϕ).

(iii) Give an example of an infinite subgroup of (G, ∗) not equal to G, and an
example of a finite subgroup of (G, ∗) of order greater than 1.
(b) Let G be a group, and a an element of G.
(i) State the definition of the order of a.
(ii) Show that, if the order of a is m2 , then the order of am is m.
(iii) Suppose that the order of G is 25. Use (ii) and the theorem of Lagrange to
show that G contains an element of order 5.

82
 2.2 Vector spaces

2.2 Vector spaces

In this section we introduce our second important example of an algebraic structure,
called a vector space. Roughly speaking it is a set of elements, called “vectors”. Any two
vectors can be “added”, resulting in a new vector, and any vector can be multiplied by
an element from R so as to give a new vector. The precise definition is given in the next
subsection.

2.2.1 Definition of a vector space

Definition. A vector space V is a set together with two functions, + : V × V → V ,
called vector addition, and · : R × V → V , called scalar multiplication, such that
(V, +) is an abelian group, and the following hold:

V1. For all v ∈ V , 1 · v = v.

V2. For all α, β ∈ R and all v ∈ V , α · (β · v) = (αβ) · v.

V3. For all α, β ∈ R and all v ∈ V , (α + β) · v = α · v + β · v.

V4. For all α ∈ R and all v1 , v2 ∈ V , α · (v1 + v2 ) = α · v1 + α · v2 .

V3 and V4 are called the distributive laws. The elements of a vector space are called
vectors.

We observe that since (V, +) is an abelian group, a vector space also has the following
properties built into its definition: (note the use of additive notation)

G1. For all v1 , v2 , v3 ∈ V , v1 + (v2 + v3 ) = (v1 + v2 ) + v3 .

G2. There exists an element o ∈ V (called the4 zero vector) such that for all v ∈ V ,
v + o = v = o + v.

G3. For every v ∈ V , there exists a unique5 element in V , denoted by −v, such that
v + (−v) = o = −v + v.

A. For all v1 , v2 ∈ V , v1 + v2 = v2 + v1 .

Examples.

1. Let m, n ∈ N, and consider the set Rm×n of m × n matrices having real entries. It
is easy to check that Rm×n , with matrix addition, is an abelian group: we omit the
verification of G1, G2, G3 and A. Define scalar multiplication as follows: if α ∈ R and
 
a11 . . . a1n
 . ..  m×n
A =  ..

∈R
.  ,
am1 . . . amn
4
Since there is a unique identity element in a group, the zero vector is unique!
5
In a group, every element has a unique inverse!

83
2 Algebra

then
 
αa11 . . . αa1n
α · A =  ... .. 

.
.  (2.2)

αam1 . . . αamn

Then α · A ∈ Rm×n , and moreover V1, V2, V3, V4 are satisfied:

V1. If A ∈ Rm×n , then clearly
   
a11 . . . a1n 1a11 . . . 1a1n
 . .
1 · A = 1 ·  .. ..  =  ... .. 
 
  
. 

am1 . . . amn 1am1 . . . 1amn
 
a11 . . . a1n
 . .. 
=  ..

 = A.
. 
am1 . . . amn

V2. For all α, β ∈ R and all A ∈ Rm×n ,

    
a11 . . . a1n βa11 . . . βa1n
 . . .. .. 
β ·  .. .. 
  
α · (β · A) = α ·  
 = α · 
 
. . 

am1 . . . amn βam1 . . . βamn
   
α(βa11 ) . . . α(βa1n ) (αβ)a11 . . . (αβ)a1n

= .
.. .
..
 
= .. .. 
   . . 

α(βam1 ) . . . α(βamn ) (αβ)am1 . . . (αβ)amn
 
a11 . . . a1n
 . .. 
= (αβ) · 
 .
.  = (αβ) · A.
. 
am1 . . . amn

V3. For all α, β ∈ R and all A ∈ Rm×n ,

 
a11 . . . a1n
 . .. 
(α + β) · A = (α + β) ·  ..

. 

am1 . . . amn
 
(α + β)a11 . . . (α + β)a1n

= .. .. 

. . 

(α + β)am1 . . . (α + β)amn
 
αa11 + βa11 . . . αa1n + βa1n

= .. .. 

. . 

αam1 + βam1 . . . αamn + βamn
   
αa11 . . . αa1n βa11 . . . βa1n
=  ..
 . .
..  +  ...
  .. 
  
. 

αam1 . . . αamn βam1 . . . βamn

84
 2.2 Vector spaces
   
a11 . . . a1n a11 . . . a1n
 . .  . .. 
. .  + β ·  ..
.

=α· .
  
. 

am1 . . . amn am1 . . . amn
= α · A + β · A.

V4. For all α ∈ R and all A, B ∈ Rm×n ,

   
a11 . . . a1n b11 . . . b1n
 . .   . ..
 . . + .

α · (A + B) = α · 
 . .   . . 

am1 . . . amn bm1 . . . bmn
 
a11 + b11 . . . a1n + b1n

=α· .. .. 

. . 

am1 + bm1 . . . amn + bmn
 
α(a11 + b11 ) . . . α(a1n + b1n )

= .. .. 

. . 

α(am1 + bm1 ) . . . α(amn + bmn )
 
αa11 + αb11 . . . αa1n + αb1n

= .. .. 

. . 

αam1 + αbm1 . . . αamn + αbmn
   
αa11 . . . αa1n αb11 . . . αb1n
=  ..
 . .
..
  .. .. 
+ . .
   

αam1 . . . αamn αbm1 . . . αbmn
   
a11 . . . a1n b11 . . . b1n
 . .. .. ..
= α ·  ..
  
+α·

.   . . 

am1 . . . amn bm1 . . . bmn
= α · A + α · B.

Hence Rm×n is a vector space with matrix addition and with scalar multiplication
defined by (2.2). If n = 1, then we denote the vector space of column vectors Rm×1
by Rm .
2. Let a, b be real numbers with a < b. The set C[a, b] of continuous functions on the
interval [a, b] with addition of functions is an abelian group. Let scalar multiplication
be defined as follows:
if α ∈ R and f ∈ C[a, b], then (α · f )(x) = αf (x), x ∈ [a, b]. (2.3)
(We say that scalar multiplication, and addition, are defined “pointwise”.) Then
α · f ∈ C[a, b], and moreover V1, V2, V3, V4 are satisfied:
V1. Let f ∈ C[a, b]. For all x ∈ [a, b], we have
(1 · f )(x) = 1f (x) = f (x),
and so 1 · f = f .

85
2 Algebra

V2. Let α, β ∈ R and f ∈ C[a, b]. For all x ∈ [a, b], we have

(α · (β · f ))(x) = α(β · f )(x)

= α(βf (x))
= (αβ)f (x)
= ((αβ) · f )(x),

and so (α · (β · f ) = (αβ) · f .
V3. Let α, β ∈ R and f ∈ C[a, b]. For all x ∈ [a, b], we have

((α + β) · f )(x) = (α + β)f (x)

= αf (x) + βf (x)
= (α · f )(x) + (β · f )(x)
= (α · f + β · f )(x),

and so (α + β) · f = α · f + β · f .
V4. Let α ∈ R and f, g ∈ C[a, b]. For all x ∈ [a, b], we have

(α · (f + g))(x) = α(f + g)(x)

= α(f (x) + g(x))
= αf (x) + αg(x)
= (α · f )(x) + (α · g)(x)
= (α · f + α · g)(x),

and so α · (f + g) = α · f + α · g.
Hence C[a, b] with addition and scalar multiplication is a vector space, called the
vector space of continuous functions on [a, b].

We now prove a few elementary properties of vector spaces.

Theorem 2.2.1. Let V be a vector space. Then the following hold:

1. For all v ∈ V , 0 · v = o.

2. For all α ∈ R, α · o = o.

3. If v ∈ V , then (−1) · v = −v.

Proof. 1. To see this, we use the distributive law to write

0 · v + 0 · v = (0 + 0) · v = 0 · v.

Now add −(0 · v) on both sides, to obtain 0 · v = o.

2. Similarly, α · o + α · o = α · (o + o) = α · o, and hence α · o = o.

86
 2.2 Vector spaces

3. Finally, we have

v + (−1) · v = 1 · v + (−1) · v (since 1 · v = v)

= (1 + (−1)) · v (distributive law)
= 0 · v (since 1 + (−1) = 0)
= o (since 0 · v = o),

and so v + (−1) · v = o = (−1) · v + v. Hence (−1) · v is an inverse of v. But the

inverse of an element from a group is unique, and so (−1) · v = −v.

2.2.2 Subspaces and linear combinations

Definition. Let V be a vector space. A subset U is called a subspace of V if

S1. o ∈ U .

S2. If v1 , v2 ∈ U , then v1 + v2 ∈ U .

S3. If v ∈ U and α ∈ R, then α · v ∈ U .

Subspaces of a vector space are analogous to subgroups of a group in the sense that
a subspace of a vector space is itself a vector space with the same addition and scalar
multiplication as with V (this is easy to check). So a subspace is really a smaller vector
space sitting inside a larger vector space.

Examples.

1. If V is a vector space, then the subset U comprising only the zero vector, namely
U = {o}, is a subspace of V .
Also, the entire vector space, that is U = V , is a subspace of V .
If a subspace U of V is neither {o} nor V , then it is called a proper subspace of V .

2. Consider the vector space R2×2 with matrix addition and scalar multiplication defined
by (2.2). Then the set of upper triangular matrices
(" # )
a b 
U1 = a, b, d ∈ R

0 d



is a subspace of R2×2 .
Also, the set of symmetric matrices
(" # )
a b 
U2 = a, b, d ∈ R

b d



is a subspace of R2×2 .

87
2 Algebra

3. Let a, b ∈ R and a < b. Consider the set of all polynomial functions

∃n ∈ N ∪ {0} and
  
  
  
a0 , a1 , a2 , . . . , an ∈ R such that

  

P [a, b] = p : [a, b] → R 
.
p(x) = a0 + a1 x + a2 x2 + · · · + an xn


  

  
for all x ∈ [a, b]
  

Then U = P [a, b] is a subspace of the vector space C[a, b] with addition of functions
and scalar multiplication defined by (2.3).
Definitions. Let V be a vector space.
1. If v1 , . . . , vn are vectors in V and α1 , . . . , αn belong to R, then the vector α1 · v1 +
· · · + αn · vn is called a linear combination of the vectors v1 , . . . , vn .

2. Let S be a non-empty subset of a vector space V . The span of S, denoted by lin(S),

is defined as the set of all possible linear combinations6 of vectors from S:

lin(S) = {α1 · v1 + · · · + αn · vn | n ∈ N, v1 , . . . , vn ∈ S, α1 , . . . , αn ∈ R} .

We also define the span of the empty set as lin(∅) = {o}.

Examples.
1. Let m ∈ N. Any vector in the vector space Rm is a linear combination of the vectors
     
1 0 0

 0 


 1 


 0 

e1 =  .. , e2 =  .. , . . . , em =  .. .
. . .
     
     
0 0 1

Hence lin({e1 , e2 . . . , em }) = Rm .

2. Let a, b ∈ R with a < b. Any polynomial p on the interval [a, b] is a linear combination
of the functions from the set S = {1, x, x2 , . . . }. Hence lin(S) = P [a, b] in the vector
space C[a, b].
The span of a set of vectors turns out to be a special subspace of the vector space.
Theorem 2.2.2. Let V be a vector space and S be a subset of V . Then lin(S) is a
subspace of V , and moreover lin(S) is the smallest subspace of V that contains S.
Proof. The result is true in the case where S = ∅, as lin(∅) = {o} is the smallest
subspace of V .
Suppose that S is non-empty, and take any v ∈ S. Then o = 0 · v ∈ lin(S). If
u, v ∈ lin(S), then we know that

u = α1 · u1 + · · · + αn · un and v = β1 · v1 + · · · + βm · vm
6
Note that although S might be infinite, a linear combination, by definition, is always a linear
combination of a finite set of vectors from S.

88
 2.2 Vector spaces

for some vectors u1 , . . . , un , v1 , . . . , vm ∈ S and scalars α1 , . . . , αn , β1 , . . . , βm ∈ R. Con-

sequently,

u + v = α1 · u1 + · · · + αn · un + β1 · v1 + · · · + βm · vm ∈ lin(S).

Finally, if v ∈ lin(S), then we know that v = α1 · v1 + · · · + αn · vn for some v1 , . . . , vn ∈ S

and scalars α1 , . . . , αn ∈ R, and so for any β ∈ R, we have

β · v = β · (α1 · v1 + · · · + αn · vn )
= β · (α1 · v1 ) + · · · + β · (αn · vn )
= (βα1 ) · v1 + · · · + (βαn ) · vn ∈ lin(S).

So lin(S) satisfies S1,S2,S3, and so it is a subspace of V . Moreover, if v ∈ S, then

v = 1 · v ∈ lin(S), and so S ⊂ lin(S). Thus lin(S) contains S.
If U is another subspace that contains the vectors from S, then from S2 and S3 it
follows that all linear combinations of vectors from S belong to U , and so lin(S) ⊂ U .
Hence lin(S) is the smallest subspace of V containing S.
Definitions. Let V be a vector space, and suppose that v1 , . . . , vn are vectors that belong
to V .
1. The vectors v1 , . . . , vn are called linearly dependent if the following condition holds:
∃α1 , . . . , αn ∈ R such that7 (α1 , α2 , . . . , αn ) 6= (0, 0, . . . , 0) and α1 · v1 + · · · +
αn · vn = o.
An arbitrary subset S of vectors from V is said to be a linearly dependent set if
there exists a non-empty finite set of dependent vectors from S.
2. The vectors v1 , . . . , vn are called linearly independent if they are not linearly
dependent, that is,
∀α1 , . . . , αn ∈ R, if α1 · v1 + · · · + αn · vn = o, then α1 = · · · = αn = 0.
An arbitrary subset S of vectors from V is said to be a linearly independent set if
every non-empty finite set of vectors from S is an independent set of vectors.
Examples.
1. Let V be a vector space. Then any finite set of vectors from V containing the zero
vector is linearly dependent. Indeed if v1 , . . . , vn ∈ V and vk = o, then

0 · v1 + · · · + 0 · vk−1 + 1 · vk + 0 · vk+1 + · · · + 0 · vn = o.

2. The vectors      
1 0 0

 0 


 1 


 0 

e1 =  .. , e2 =  .. , . . . , em =  .. 
. . .
     
     
0 0 1
7
Note that (α1 , α2 , . . . , αn ) 6= (0, 0, . . . , 0) iff not all of α1 , . . . , αn are equal to 0, that is, iff at least
one of them is non-zero.

89
2 Algebra

are linearly independent in Rm . Indeed if α1 · e1 + e2 + · · · + αm · em = o, then

     
  1 0 0  
α1  .  0
 . 

0  
1 
 ..   . 
 .  = α1 · 
 =  ..  ,
   
 .  ..  + α2 · ..  + · · · + αm ·    
. . 0 
    
αm 0
    
0 0 1

and so α1 = · · · = αm = 0.

3. Let a, b ∈ R with a < b. The functions 1, x on the interval [a, b] are linearly independent.
Indeed, if for all x ∈ [a, b], α · 1 + β · x = 0(x), then in particular, we have

α + βa = 0 and α + βb = 0,

and since a 6= b, it follows that α = β = 0.

4. It follows from the definition that the empty set is not linearly dependent, since it
does not contain a non-empty subset. Thus the empty set is a linearly independent
subset of any vector space.

2.2.3 Basis of a vector space

Definition. Let V be a vector space. Then a set B of vectors is said to be a basis of V
if

B1. lin(B) = V , and

B2. B is linearly independent.

Examples.

1. The vectors      
1 0 0

 0 


 1 


 0 

e1 =  .. , e2 =  .. , . . . , em =  .. 
. . .
     
     
0 0 1
form a basis of Rm .

2. The empty set ∅ is a basis of the subspace {o} of any vector space.

Theorem 2.2.3. Let V be a vector space and B be a basis of V such that B has n
elements. Then for any linearly independent set S of m ≤ n vectors there exists a subset
B 0 ⊂ B of n − m vectors such that S ∪ B 0 is also a basis of V .

90
 2.2 Vector spaces

Proof. We use induction on m ≥ 0.

The case m = 0 is simple: S = ∅ and for B 0 we take the whole of B. Then
S ∪ B 0 = ∅ ∪ B = B, a basis of V .
Now assume as induction hypothesis that m ∈ N is fixed, then for any linearly
independent set S1 of m − 1 ≤ n vectors there exists a subset B10 ⊂ B of n − (m − 1)
vectors such that S1 ∪ B10 is a basis of V .
We have to prove that for any linearly independent set S of m ≤ n vectors there exists
a subset B 0 ⊂ B of n − m vectors such that S ∪ B 0 is also a basis of V .
Write S = {v1 , . . . , vm }. To apply the induction hypothesis we choose S1 = {v1 , . . . , vm−1 } =
S \ {vm }. Then S1 contains m − 1 vectors, and because m ≤ n, also m − 1 ≤ n − 1 < n, so
we may apply the induction hypothesis to obtain a subset B10 ⊂ B of n−(m−1) = n−m+1
vectors such that S1 ∪ B10 is a basis of V . Write B10 = {u1 , . . . , un−m+1 }.
We now want to remove a vector from B10 to obtain the required B 0 , that is, such that
S ∪ B 0 will end up as a basis. Since

S1 ∪ B10 = {v1 , v2 , . . . , vm−1 , u1 , u2 , . . . , un−m+1 }

is a basis, there exist numbers

α1 , α2 , . . . , αm−1 , β1 , β2 , . . . , βn−m+1 ∈ R

such that

vm = α1 v1 + α2 v2 + · · · + αm−1 vm−1 + β1 u1 + β2 u2 + · · · + βn−m+1 un−m+1 . (2.4)

Suppose that β1 = β2 = · · · = βn−m+1 = 0. Then vm = α1 v1 + α2 v2 + · · · + αm−1 vm−1 ,

which can be rewritten as

α1 v1 + α2 v2 + · · · + αm−1 vm−1 − vm = o,

which implies that S = {v1 , v2 , . . . , vm } is linearly dependent, a contradiction.

Therefore, some βj 6= 0. We may then write uj as a linear combination of v1 , v2 , . . . , vm−1
and the remaining ui , i 6= j, as follows:

βj uj = −α1 v1 − · · · − αm−1 vm−1 + vm

− β1 u1 − · · · − βj−1 uj−1 − βj+1 uj+1 − · · · − βn−m+1 un−m+1 ,

which gives
α1 αm−1 1 β1
uj = − v1 − · · · − vm−1 + vm − u1 − . . .
βj βj βj βj
(2.5)
βj−1 βj+1 βn−m+1
− uj−1 − uj+1 − · · · − un−m+1 .
βj βj βj

Now remove uj from B10 to obtain B 0 = B10 \ {uj }. Then B 0 contains n − m vectors. It
remains to prove that S ∪ B 0 is a basis of V . For this we have to show that

91
2 Algebra

1. lin(S ∪ B 0 ) = V , and

2. S ∪ B 0 is linearly independent.

First we show that lin(S ∪ B 0 ) = V . Let v ∈ V . Since S1 ∪ B10 is a basis, there exist
λ1 , . . . , λm−1 , µ1 , . . . , µn−m+1 ∈ R such that

v = λ1 v1 + · · · + λm−1 vm−1 + µ1 u1 + · · · + µn−m+1 un−m+1 .

Use (2.5) to eliminate uj in the right-hand side. It is not necessary to calculate this
explicitly; it is sufficient to know that this will give v as a linear combination of
v1 , . . . , vm , u1 , . . . , un−m+1 except for uj , and then we may conclude that v ∈ lin(S ∪ B 0 ).
Thus we have shown that V ⊂ lin(S ∪ B 0 ). The opposite inclusion lin(S ∪ B 0 ) ⊂ V is
obvious. Thus lin(S ∪ B 0 ) = V .
Next we show that S ∪ B 0 is a linearly independent set. To do this, write o as a linear
combination of the vectors in S ∪ B 0 :
o = γ1 v1 + · · · + γm vm + δ1 u1 + · · · + δj−1 uj−1
(2.6)
+ δj+1 uj+1 + · · · + δn−m+1 un−m+1 .

We may eliminate vm by using (2.4):

o = γ1 v1 + · · · + γm (α1 v1 + · · · + αm−1 vm−1 + β1 u1 + · · · + βn−m+1 un−m+1 )

+ δ1 u1 + · · · + δj−1 uj−1 + δj+1 uj+1 + · · · + δn−m+1 un−m+1
= (γ1 + γm α1 )v1 + (γ2 + γm α2 )v2 + · · · + (γm−1 + γm αm−1 )vm−1
+ (γm β1 + δ1 )u1 + (γm β2 + δ)2)u2 + · · · + (γm βj−1 + δj−1 )uj−1
+ γm βj uj + (γm βj+1 + δj+1 )uj+1 + · · · + δn−m+1 un−m+1 .

Since {v1 , . . . , vm−1 , u1 , . . . , un−m+1 } = S1 ∪ B10 is a basis, all the coefficients in the latter
linear combination equal 0:

γ1 + γm α1 = 0, γ2 + γm α2 = 0, ... γm−1 + γm αm−1 = 0,

γm β1 + δ1 = 0, γm β2 + δ2 = 0, ... γm βj−1 + δj−1 = 0,
γm βj = 0,
γm βj+1 + δj+1 = 0, ... γm βn−m+1 + δn−m+1 = 0.

Since βj 6= 0 it follows that γm = 0. It then follows that all

γ1 = · · · = γm−1 = 0 = δ1 = · · · = δj−1 = δj+1 = · · · = δn−m+1 .

Thus all the coefficients in (2.6) are equal to 0, which shows that S ∪ B 0 is linearly
independent.
It follows that S ∪ B 0 is a basis, and the induction step is complete.

Corollary 2.2.4. Let V be a vector space and B be a basis of V such that B has n
elements. Then any linearly independent set S of n vectors is also a basis of V .

92
 2.2 Vector spaces

Proof. This corollary follows from Theorem 2.2.3 by taking m = n. Then B 0 = ∅

because it has n − m = n − n = 0 elements and S ∪ B 0 = S is a basis of V .
Given a vector space, there are of course many bases. However, the next result says
that the cardinality of the basis is unique for any given vector space.

Corollary 2.2.5. If a vector space V has a basis with n elements, then every basis of V
has the same number of elements.

Proof. Suppose that B is a basis of V with n elements and suppose that B 0 is another
basis of V .
Let B 0 have more than n elements. Then take any n distinct elements v1 , . . . , vn from B 0 .
From the previous corollary, it follows that these span V , and so if v ∈ B 0 \{v1 , . . . , vn }, it
can be written as a linear combination of {v1 , . . . , vn }, which contradicts the independence
of B 0 .
If B 0 has fewer than n elements, then by interchanging the roles of B and B 0 and
proceeding as above, we once again arrive at a contradiction.
The above result motivates the following natural definitions.

Definitions. Let V be a vector space.

1. If there exists a basis B of V such that B has n elements, then n is called the
dimension of V , and it is denoted by dim(V ).

2. If a vector space has a basis with a finite number of elements, then it is called a finite
dimensional vector space.

3. If a vector space is not finite dimensional, then it is called an infinite dimensional

vector space.

Examples.

1. It follows from Example 1 on p. 90 that dim(Rm ) = m.

2. It follows from Example 2 on p. 90 that dim({o}) = 0.

2.2.4 Linear transformations

Definition. Let U, V be two vector spaces. A function T : U → V is called a linear
transformation if it satisfies the following two properties:

L1. For all u1 , u2 ∈ U , T (u1 + u2 ) = T (u1 ) + T (u2 ).

L2. For all u ∈ U , and all α ∈ R, T (α · u) = α · T (u).

Just as group homomorphisms are functions that respect group operations, linear
transformations are functions that respect vector space operations.

93
2 Algebra

Examples.
1. Let m, n ∈ N. Let  
a11 . . . a1n
 . ..  m×n
 ..
A= ∈R
.  .
am1 . . . amn
Then the function TA : Rn → Rm defined by
 n
X

    a1k xk
x1 a11 x1 + · · · + a1n xn
 
 k=1 
 .  ..

..

.
 
TA  = = 
(2.7)
 .   .  
 . 

n
xn am1 x1 + · · · + amn xn  X 

amk xk 
k=1
 
x1
 ..  n n
for all  ∈ R , is a linear transformation from the vector space R to the vector
. 


xn
space Rm . Indeed, we have
     
x1 y1 x1 + y1
 .   ..  ..
 ..  + 
  
TA   = TA 
  
. 
  . 

xn yn xn + yn
 n
X

 a1k (xk + yk ) 
 k=1 
..
 
=
 

 . 

 Xn 

amk (xk + yk ) 
k=1

 n
X
  n
X n
X

 (a1k xk + a1k yk )   a1k xk + a1k yk 
 k=1   k=1 k=1

.. ..
   
= 
 .

 = 
 .


   
n
 X  n
 X n
X 

(amk xk + amk yk )  
amk xk + amk yk 
k=1 k=1 k=1

Xn   n
X

 a1k xk   a1k yk 
 k=1   k=1 
.. ..
   
= +
   

 .   . 

 Xn   Xn 

amk xk  
amk yk 
k=1 k=1
   
x1 y1
 .  .. 
.

= TA 
 .  + TA 
  ,
. 
xn yn

94
 2.2 Vector spaces

for all    
x1 y1
 .   . 
 .  ,  .  ∈ Rn ,
 .   . 
xn yn
and so L1 holds. Moreover,
 n
X

     a1k αxk 
x1 αx1

 k=1 
 .   .  
..

 .  = TA  .  = 

TA α ·

  .   .  
 . 

n
xn αxn  X 

amk αxk 
k=1
 n
X
  n
X

α a1k xk  a1k xk  
x1
  
k=1
   k=1 
.. ..  . 
   
= = α· = α · TA  .. 
,
   

 . 
  . 


n n
 X   X  xn

α amk xk  
amk xk 
k=1 k=1

for all α ∈ R and all vectors  

x1
 . 
 .  ∈ Rn ,
 . 
xn
and so L2 holds as well. Hence TA is a linear transformation.

2. The function T : C[0, 1] → R given by

1
 
Tf = f for all f ∈ C[0, 1],
2
is a linear transformation from the vector space C[0, 1] to the vector space R. Indeed,
1 1 1
     
T (f + g) = (f + g) =f +g = T (f ) + T (g)
2 2 2
for all f, g ∈ C[0, 1], and so L1 holds. Furthermore,
1 1
   
T (α · f ) = (α · f ) = αf = αT (f )
2 2
for all α ∈ R and all f ∈ C[0, 1], and so L2 holds too. Thus T is a linear transformation.

Just as in the case of homomorphisms between groups, there are two important subsets
associated with a linear transformation between vector spaces, whose definitions are
given below.

95
2 Algebra

Definitions. Let U, V be vector spaces and T : U → V a linear transformation.

1. The kernel of T is defined to be the set ker(T ) = {u ∈ U | T (u) = oV }, where oV

denotes the zero vector of V .

2. The image of T is defined to be the set im(T ) = {v ∈ V | ∃u ∈ U such that T (u) = v}.

Examples.

1. Let A ∈ Rm×n , and let TA : Rn → Rm be the linear transformation defined by (2.7).

The kernel of the linear transformation is the set of all vectors
 
x1
 .  n
x =  .. 

∈R
xn

such that the system of linear equations

a11 x1 + · · · + a1n xn = 0
..
.
am1 x1 + · · · + amn xn = 0

is simultaneously satisfied.
The range of TA is the set of all vectors
 
y1
 .  m
y =  .. 

∈R
ym

such that there exists a vector

 
x1
 .  n
 ..  ∈ R
x= 
xn

such that

a11 x1 + · · · + a1n xn = y1
..
.
am1 x1 + · · · + amn xn = ym .
 
2. The function T : C[0, 1] → R given by T f = f 12 for all f ∈ C[0, 1], is a linear
transformation from the vector space C[0, 1] to the vector space R, with kernel equal
to the set of continuous functions on the interval [0, 1] that vanish at the point 12 . The
range of T is the whole vector space R.

96
 2.2 Vector spaces

Analogous to Theorem 2.1.5 for homomorphisms between groups, we now prove the
following result.

Theorem 2.2.6. Let U, V be vector spaces and T : U → V a linear transformation. Then

1. ker(T ) is a subspace of U .

2. im(T ) is a subspace of V .

Proof. In each of the cases, we check that S1, S2, S3 hold.

Let oU , oV denote the zero vectors in the vector spaces U, V , respectively. It is easy to
check that ker(T ) is a subspace of U . Indeed, as T (oU ) = T (oU + oU ) = T (oU ) + T (oU ),
it follows that T (oU ) = oV , and so oU ∈ ker(T ). Thus S1 holds. If u1 , u2 belong to
ker(T ), then T (u1 + u2 ) = T (u1 ) + T (u2 ) = oV + oV = oV , and so S2 holds as well.
Finally, S3 holds, since if α ∈ R and u ∈ ker(T ), then T (α · u) = α · T (u) = α · oV = oV .
Hence ker(T ) is a subspace of U .
We now also check that im(T ) is a subspace of V . Since T (oU ) = oV , it follows that
oV ∈ im(T ), and so S1 holds. If v1 , v2 belong to im(T ), then there exist elements u1 , u2
in U such that T (u1 ) = v1 and T (u2 ) = v2 . Consequently, T (u1 +u2 ) = T (u1 )+T (u2 ) =
v1 +v2 , and so there exists an element in U , namely u1 +u2 , such that T (u1 +u2 ) = v1 +v2 ,
that is, v1 + v2 ∈ im(T ). Thus S2 holds. Finally, if α ∈ R and v ∈ im(T ), then there
exists a u ∈ U such that T (u) = v, and so α·v = α·T (u) = T (α·u). Hence α·v ∈ im(T ),
and S3 holds. So im(T ) is a subspace of V .

Definition. Let U, V be vector spaces and T : U → V a linear transformation. Then

1. the nullity of T is defined to be the dimension of the kernel of T :

nullity(T ) = dim(ker(T )),

2. and the rank of T is defined to be the dimension of the image of T :

rank(T ) = dim(im(T )).

Our final result shows a connection between the nullity and the rank of a linear transfor-
mation.

Theorem 2.2.7. Let U be a finite dimensional vector space, V be a vector space, and
let T : U → V be a linear transformation. Then:

nullity(T ) + rank(T ) = dim(U ).

Proof. Let U be a vector space with dimension n and basis B, V be a vector space,
and T : U → V be a linear transformation. By the previous theorem, ker(T ) is a
subspace of U , thus it is finite dimensional (see Exercise 19), with dimension k ≤ n. Let
S = {u1 , u2 , . . . , uk } be a basis of ker(T ). By Theorem 2.2.3, there exists a subset B 0 of
B with n − k elements such that S ∪ B is a basis of U . Denote the elements of B 0 by

97
2 Algebra

v1 , . . . , vn−k . We are going to prove that W := {T (v1 ), . . . , T (vn−k )} is a basis of im(T )

and, hence,

dim(ker(T )) + dim(im(T )) = k + (n − k) = n = dim(U ).

First, we verify B2. Let α1 , . . . , αn−k ∈ R be real numbers such that

α1 · T (v1 ) + · · · + αn−k · T (vn−k ) = oV .

Using L1 and L2 we obtain

oV = α1 · T (v1 ) + · · · + αn−k · T (vn−k ) = T (α1 · v1 + · · · + αn−k · vn−k ).

In other words, α1 · v1 + · · · + αn−k · vn−k ∈ ker(T ). Since S is a basis of ker(T ), there

are real numbers β1 , . . . , βk ∈ R so that

α1 · v1 + · · · + αn−k · vn−k = β1 · u1 + · · · + βk · uk .

But u1 , . . . , uk , v1 , . . . , vn−k are linearly independent vectors because S ∪ B 0 is a basis of

U , so we must have α1 = · · · = αn−k = β1 = · · · = βk = 0. Hence, W is a set of linearly
independent vectors.
To see B1, let v be any vector in im(T ). So there is u ∈ U such that T (u) = v. Since
S ∪ B 0 is a basis of U , there exist α1 , . . . , αn−k , β1 , . . . , βk ∈ R such that

u = β1 · u1 + · · · + βk · uk + α1 · v1 + · · · + αn−k · vn−k .

From this we obtain that

v = T (u)
= T (β1 · u1 + · · · + βk · uk + α1 · v1 + · · · + αn−k · vn−k )
= β1 · T (u1 ) + · · · + βk · T (uk ) + α1 · T (v1 ) + · · · + αn−k · T (vn−k ).

Since S ⊂ ker(T ), we have T (u1 ) = · · · = T (uk ) = oV , therefore,

v = β1 · T (u1 ) + · · · + βk · T (uk ) + α1 · T (v1 ) + · · · + αn−k · T (vn−k )

= β1 · oV + · · · + βk · oV + α1 · T (v1 ) + · · · + αn−k · T (vn−k )
= α1 · T (v1 ) + · · · + αn−k · T (vn−k ) ∈ lin W.

Hence, im(T ) = lin W . (Why?)

2.2.5 Exercises
1. a) Is the set of invertible 2 × 2 matrices having real entries with matrix addition and
with scalar multiplication defined by (2.2) a vector space?
b) Is the set of invertible 2 × 2 matrices having real entries with matrix multiplication
and with scalar multiplication defined by (2.2) a vector space?

98
 2.2 Vector spaces

2. Let V be a vector space. Prove, directly from the axioms, that if α ∈ R and v ∈ V
are such that α · v = o, then either α = 0 or v = o.
Hint: If α 6= 0, how can you manipulate the equation α · v = o?

3. Let U and V be vector spaces, and define

U × V = {(u, v) | u ∈ U, v ∈ V } .

Show that U × V , with addition defined by (u1 , v1 ) + (u2 , v2 ) = (u1 + u2 , v1 + v2 )

and scalar multiplication by α · (u, v) = (α · u, α · v), is a vector space.

4. Consider the set R∞ of all sequences with addition defined as follows:

if (an )n∈N , (bn )n∈N ∈ R∞ , then (an )n∈N + (bn )n∈N = (an + bn )n∈N , (2.8)

and scalar multiplication defined as follows:

if α ∈ R and (an )n∈N ∈ R∞ , then α · (an )n∈N = (αan )n∈N . (2.9)

Prove that R∞ is a vector space with the above addition and scalar multiplication.

5. Show that a subspace of a vector space is itself a vector space.

6. Determine if the following statements are TRUE or FALSE. Give reasons for your
answers in each case.
a) The union of two subspaces of a vector space V is a subspace of V .
b) The intersection of two subspaces of a vector space V is a subspace of V .
     
1  1
 0 
c)  2  ∈ lin  3  ,  1  in the vector space R3 .
       

3 
4 1 
     
1 1 0
d) The vectors  0  ,  2  ,  3  are linearly independent in the vector space R3 .
     

1 1 0
e) If v1 , v2 , v3 , v4 are linearly independent vectors, then v1 , v2 , v3 are linearly inde-
pendent.
f) If v1 , v2 , v3 , v4 are linearly dependent vectors, then v1 , v2 , v3 are linearly dependent.
g) If v1 , v2 are linearly independent, and v1 , v2 , v3 are linearly dependent, then v3 is
a linear combination of v1 and v2 .
h) If v1 , v2 are linearly dependent, then there is a real number α such that v1 = αv2 .

7. Let X and Y be two subspaces of a vector space V . Define the set X + Y to be

X + Y = {x + y | x ∈ X, y ∈ Y } .

Show that X + Y is a subspace of V .

99
 2 Algebra

8. Consider the space C[0, π] of continuous functions on [0, π], and the four functions
f1 , f2 , f3 , f4 , where f (x) = 1 (i.e., f1 is the constant function 1), f2 (x) = cos x,
f3 (x) = cos 2x and f4 (x) = cos2 x.
a) Show that {f1 , f2 , f3 , f4 } is linearly dependent. You may use any properties of the
trigonometric functions that you know.
b) Show that {f1 , f2 , f3 } is linearly independent.
Hint: If α1 · f1 + α2 · f2 + α3 · f3 = o, then α1 f1 (x) + α2 f2 (x) + α3 f3 (x) = 0 for
all x ∈ [0, 1], and so in particular for x = 0, x = π/2 and x = π.

9. Let v1 , v2 , v3 , v4 be linearly independent vectors in a vector space V . Let V1 =

lin({v1 , v2 , v3 }) and V2 = lin({v2 , v3 , v4 }). Find V1 ∩ V2 , and justify your answer in
detail.

10. a) Prove that if t1 and t2 are distinct real numbers in R, then

(" # " #)!
1 1
lin , = R2
t1 t2

in the vector space R2 .

b) Prove that R2 is not the union of a finite number of proper subspaces.
("Consider
Hint: # the
) intersection of any proper subspace with the infinite subset
1 
S= t∈R .
t 

11. Let R∞ be the vector space of all sequences, with addition and scalar multiplication
defined by (2.8) and (2.9), respectively. We define the following subsets of R∞ :
a) `∞ is the set of all bounded sequences.
b) c is the set of all convergent sequences.
c) c0 is the set of all convergent sequences with limit 0.
d) c00 = {(an )n∈N ∈ R∞ | ∃N ∈ N such that ∀n > N, an = 0}, is the set of all se-
quences that are eventually zero.
Prove that c00 ⊂ c0 ⊆ c ⊆ `∞ ⊆ R∞ , and that each is a subspace of the next one.

12. Consider the vector space C[0, 1] with addition of functions and scalar multiplication
defined by (2.3). Let S(y1 , y2 ) = {f ∈ C[0, 1] | f (0) = y1 and f (1) = y2 }. Show that
S(y1 , y2 ) is a subspace of C[0, 1] iff y1 = 0 = y2 .

100
 2.2 Vector spaces

13. Consider the vector space C[−1, 1] with addition of functions and scalar multiplication
defined by (2.3).
a) Let Se be the set of all even functions in C[−1, 1]:

Se = {f ∈ C[−1, 1] | ∀x ∈ [−1, 1] f (x) = f (−x)} .

Show that Se is a subspace of C[−1, 1].

b) Is the set So of odd functions, defined as follows, also a subspace of C[−1, 1]?

So = {f ∈ C[−1, 1] | ∀x ∈ [−1, 1] f (−x) = −f (x)} .

14. Consider the vector space C[0, 1], with the usual addition and scalar multiplication.
Show that the set S = {1, x, x2 , x3 , . . . } of functions in C[0, 1] is linearly independent.
What is lin(S)?
Hint: In order to show that a function f in C[0, 1] is not equal to the zero function,
one approach is to try and show that there is some δ > 0 such that f (x) > 0 for all x
with 0 < x < δ.

15. Prove or disprove that      


 1 1 1 
B= 
 0 , 1 , 1 
   

 

0 0 1 
is a basis for R3 .

16. Prove that if B is a basis of a finite-dimensional vector space V , then every element
v ∈ V can be written as a unique linear combination of the vectors from B.

17. Let V be a vector space, and let U be a subspace of V of dimension n. Let v be a

vector in V such that v ∈
/ U . Let W = {u + kv | u ∈ U, k ∈ R}. Show that W is a
subspace of V of dimension n + 1.

18. Is the following statement true or false? Justify your answer, by means of a proof or a
counterexample.
Let V be a vector space of dimension 3. Let u, v, w be vectors in V such that neither
v nor w is in lin({u}). Then {u, v, w} is a basis of V .

19. a) Suppose that S is a linearly independent set of vectors in a vector space V such
that lin(S) 6= V . Show that there is a vector v in V such that S ∪ {v} is linearly
independent.
b) Suppose U is an infinite-dimensional vector space. Show that, for each n ∈ N, U
has an independent set with n elements.
c) Suppose that V is a finite-dimensional vector space and that U is a subspace of V .
Show that U does not contain an independent set with more than dim(V ) elements.
Deduce that U is finite-dimensional, and dim(U ) ≤ dim(V ).
Show also that, if dim(U ) = dim(V ), then U = V .

101
 2 Algebra

20. Show that C[0, 1] is infinite dimensional.

Hint: See Exercise 14.

21. a) For k ∈ N, let ek denote the sequence with the k th term equal to 1, and all other
terms equal to zero:

1 if n = k,
ek = (an )n∈N , where an =
0 6 k,
if n =

and let B = {ek | k ∈ N}. Prove that B is a basis for the vector space c00 ,
comprising all sequences that are eventually zero.
b) Is B = {ek | k ∈ N} also a basis for the vector space R∞ ?
Hint: Consider the constant sequence (1)n∈N .

22. Find the kernel and image of the linear transformations TA : R2 → R2 , where A ∈ R2×2
is given by:
" #
1 1
a) A = .
0 1
" #
1 1
b) A = .
0 0
" #
0 0
c) A = .
0 0
" #
x
Each vector in R2 represents a point in the plane. In each of the cases, draw a
y
picture of the subspaces ker(TA ) and im(TA ) in the plane.

23. Consider the vector space R2 with matrix addition and the usual scalar multiplication
defined by (2.2). Define the function T : R2 → R2 as follows:
 " #
0
if x2 6= 0,



" # " # 
 x2
x1 x1


if ∈ R2 , then T = "
x2 x2 #

 x1
if x2 = 0.




x2

Show that T satisfies L2, but not L1, and hence it is not a linear transformation.

24. Let P [0, 1] denote the space of polynomial functions on [0, 1], with pointwise addition
and scalar multiplication. For a function p ∈ P [0, 1], with p(x) = a0 + a1 x + a2 x2 +
· · · + an xn , define p∗ (x) = a1 + a2 x + · · · + an xn−1 .
Show that the function T : P [0, 1] → P [0, 1] given by T (p) = p∗ is a linear transfor-
mation.

102
 2.2 Vector spaces

25. Let c denote the vector space of all convergent sequences. Consider the function
T : c → R given by
T ((an )n∈N ) = lim an .
n→∞

a) Prove that T is a linear transformation from the vector space c to the vector
space R.
b) What is the kernel of T ?
c) Show that im(T ) = R.

26. Recall the definitions of the subspaces of even functions Se and odd functions So of
C[−1, 1] in Exercise 13.
a) Define, for each function f ∈ C[−1, 1], the function fe : [−1, 1] → R by

f (x) + f (−x)
fe (x) = , x ∈ [−1, 1].
2
Show that fe ∈ Se for each f ∈ C[−1, 1].
b) Define the function T : C[−1, 1] → C[−1, 1] by T (f ) = fe . Show that T is a
linear transformation.
c) Find ker(T ) and im(T ).

27. In this exercise, the associativity of matrix multiplication A(BC) = (AB)C is proved
using linear transformations.
a) Show that for any linear transformation T : Rn → Rm there is a unique m × n
matrix A such that T (x) = Ax for all x ∈ Rn .

Hint: If {e1 , e2 , . . . , en } is the standard basis of Rn , then explain why the only
possibility for A is
.. .. ..
 
 . . ··· . 
A=
 T (e1 ) T (e2 ) · · · T (en ) .


.. .. .. 
. . ··· .

b) Let T1 : Rn → Rm and T2 : Rk → Rn be linear transformations. Denote the matrices

corresponding to T1 and T2 by A1 and A2 , respectively.
i) Show that the composition T1 ◦ T2 : Rk → Rm , which is defined (in the usual
way) by (T1 ◦ T2 )(v) = T1 (T2 (v)) for all v ∈ Rk , is also a linear transformation.
ii) Show that the matrix that corresponds to T1 ◦ T2 is the product A1 A2 .

103
2 Algebra

c) Let T1 : Rn → Rm , T2 : Rk → Rn and T3 : Rp → Rk be linear transformations.

i) Show that (T1 ◦ T2 ) ◦ T3 and T1 ◦ (T2 ◦ T3 ) are the same transformations.
ii) Denote the matrices corresponding to T1 , T2 and T3 by A1 , A2 , A3 , respectively.
Show that the matrix corresponding to (T1 ◦ T2 ) ◦ T3 is (A1 A2 )A3 and the
matrix corresponding to T1 ◦ (T2 ◦ T3 ) is A1 (A2 A3 ).
Conclude that (A1 A2 )A3 = A1 (A2 A3 ) for any m × n matrix A1 , n × k matrix
A2 and k × p matrix A3 .
d) Suppose that U , V and W are vector spaces, and that T : U → V and S : V → W
are linear transformations. Consider the composition R : U → W given by
R(u) = S(T (u)).
(a) Show that R is a linear transformation.
(b) Show that im(R) ⊆ im(S) and that ker(T ) ⊆ ker(S). Deduce that r(R) ≤
min(r(S), r(T )).
e) Sample Exam Question. 2010 Q8.
(a) Define the following terms: linear transformation, kernel and image of a linear
transformation.
(b) Define what it means for vectors v1 , v2 , . . . vk to be linearly independent.
Let V and W be two vector spaces, T : V → W a linear transformation, and
v1 , v2 , . . . , vk vectors in V . Suppose that T (v1 ), T (v2 ), . . . , T (vk ) are linearly
independent vectors in W .
Show that v1 , v2 , . . . , vk are linearly independent in V .
(c) Consider a vector space V with basis {v1 , v2 }. Let T : V → V be a linear
transformation such that

T (v1 ) = v1 − v2 ; T (v1 + v2 ) = v2 − v1 . (∗)

Explain why there is a linear transformation that satisfies (∗), and show that
there is only one such linear transformation.
Find ker(T ) and im(T ).

104
Written by:
Professor Amol Sasane

Revised by:
Professor Jozef Skokan, Professor Konrad
Swanepoel and Professor Graham
Brightwell