You are on page 1of 39

London School of Economics

MA203
Real Analysis: MA103 revision material
Lecture Notes

Written by Martin Anthony,


Department of Mathematics, London School of Economics.
MA203 Real Analysis: MA103 revision material

Contents

1 Supremum and infimum 1


1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The triangle inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Properties of real numbers: supremum and infimum . . . . . . . . . . . . . . . . . . . 1
1.4 Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Sample examination questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Comments on selected activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.7 Sketch answers to or comments on selected questions . . . . . . . . . . . . . . . . . . 5

2 Sequences and limits 7


2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Examples of sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Sequences and limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Sequences: formal definition . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.2 The definition of a limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Some standard results on and properties of limits . . . . . . . . . . . . . . . . . . . . 11
2.5 Algebra of limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.6 Some useful limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.7 The sandwich theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.8 Subsequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.9 Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.10 Sample examination questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.11 Comments on selected activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.12 Sketch answers to or comments on selected questions . . . . . . . . . . . . . . . . . . 22

3 Limits of functions and continuity 28


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2 Limit of a function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.1 Definition of limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Algebra of limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.4 More on limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Continuity and sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Continuous functions on closed intervals . . . . . . . . . . . . . . . . . . . . . 32
3.5 The Intermediate Value Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.6 Learning outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.7 Sample examination questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.8 Comments on selected activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.9 Sketch answers to or comments on selected questions . . . . . . . . . . . . . . . . . . 36
Chapter 1
Supremum and infimum

1.1 Introduction

In this short chapter we explore some properties of sets of real numbers. Whereas bounded sets of
integers will always have a least and a greatest member (that is, a minimum and a maximum), this is
not true for bounded sets of real numbers. Instead, the corresponding notions are infimum and
supremum.

1.2 The triangle inequality

We start with a useful observation about real numbers. For any real number x, the absolute value of
x, denoted by |x|, is x itself if x ≥ 0 and −x if x < 0. For instance, |5| = 5 and | − 3.5| = 3.5.

The basic triangle inequality is very simple, but extremely useful. It states that for any two real
numbers x, y, we have
|x + y| ≤ |x| + |y|.

Learning activity 1.1

Prove the triangle inequality. (Just consider each of four cases: x > 0, y < 0, etc.)

This has many useful consequences, which we shall use often. In particular, for any three real numbers
x, y, z, we have
|x − y| ≤ |x − z| + |y − z|
(simply because |x − y| = |(x − z) + (z − y)| ≤ |x − z| + |z − y|).

Sometimes useful is the fact that for any two real numbers x, y,

|x − y| ≥ ||x| − |y||.

1.3 Properties of real numbers: supremum and infimum

Recall that for real numbers a, b, we define different types of interval as follows:

[a, b] = {x ∈ R | a ≤ x ≤ b}

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

1
MA203 Real Analysis: MA103 revision material

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

We shall sometimes use the symbol R+ to denote the set of all positive real numbers.

It is very important (especially in what follows) that you understand that ∞ is not a number; it is
merely a symbol. You cannot perform arithmetic with it. It is quite simply nonsense to write 1/∞ = 0
and so on.

Some of these intervals are ‘bounded above’, some ‘bounded below’ and some ‘bounded above and
below’. The formal definitions of boundedness are:

Definition 1.1 S ⊆ R is bounded above if and only if there is M ∈ R such that


for all x ∈ S, x ≤ M.
In this case, M is called an upper bound of (or ‘on’ or ‘for’) S.

Definition 1.2 S ⊆ R is bounded below if and only if there is m ∈ R such that


for all x ∈ S, x ≥ m.
In this case, m is called a lower bound of (or ‘on’ or ‘for’) S.

A set is said simply to be bounded if it is bounded above and bounded below.

Definition 1.3 An element x ∈ S is the maximum of S if it is an upper bound of S (similarly, we can


define the minimum of S).

A set that is bounded above need not have a maximum; a set has a maximum element if and only if it
has an upper bound that belongs to the set. Similarly, a set that is bounded below need not have a
minimum; a set has a minimum element if and only if it has a lower bound that belongs to the set.

Every bounded subset of the integers has a maximum and a minimum. The corresponding notions for
subsets of the real line are supremum and infimum, defined below.

We assume the following property of real numbers (as an axiom).

The continuum property: Every non-empty set of real numbers that is bounded above has a least
upper bound, and every non-empty set of real numbers that is bounded below has a greatest upper
bound.

Definition 1.4 For a non-empty subset S of R, if S is bounded above, the least upper bound of S
(which exists, by the continuum property) is called the supremum of S, and is denoted sup S. If S is
bounded below, the greatest lower bound is called the infimum of S and is denoted inf S.

Note that the supremum and infimum of a set S (when they exist; i.e. when S is bounded
above/below) need not belong to S.

Example Consider the real interval S = (1, 2). This is bounded above: for example, 2 is an upper
bound, since every member of S is at most 2. In fact, any number greater than or equal to 2 is an
upper bound, so the set of upper bounds is [2, ∞). It is clear that the least upper bound is therefore
2. That is, sup S = 2. But notice that the set S has no maximum. For, suppose that M is a
maximum. Then (two things!): (i) M ∈ S and, (ii) for all x ∈ S, x ≤ M . Now, because M ∈ S, we
must have M < 2. But now consider the number (M + 2)/2, half-way between M and 2. This is still
less than 2, so it belongs to S. Thus it is not true that x ≤ M for all x ∈ S, because this doesn’t
hold when we take x = (M + 2)/2. So S has a supremum but no maximum.

Example On the other hand, the interval S = (1, 2] has a maximum: its maximum (and its
supremum) is 2.

2
Chapter 1. Supremum and infimum

Example Consider the set S = {1/n | n ∈ N} = {1, 1/2, 1/3, . . .}. This is clearly bounded below, by
0 for example. In fact, inf S = 0. To see this, we observe (as we already have) that 0 is a lower bound
on S. To prove it is the greatest lower bound, we need to show that no number τ > 0 is a lower
bound on S. So let’s suppose that τ is any positive number. To prove it is not a lower bound on S we
have to show that some member of S is less than τ . Well, this means we need to find a number of
the form 1/n such that 1/n < τ . But it’s clear that this inequality is equivalent to n > 1/τ . So if we
take any natural number n that is greater than 1/τ , then 1/n ∈ S and 1/n < τ . So we are done: any
τ > 0 can’t be a lower bound, and 0 is, so 0 is the greatest lower bound, the infimum.

Learning activity 1.2

Make sure you understand the preceding three examples by reading them again and trying to
reproduce the arguments given.

An alternative description of the supremum is: σ = sup S if and only if σ is an upper bound for S and
if for any σ ′ < σ, there is some x ∈ S with x > σ ′ . (A similar condition can be given for infimum.)

Learning activity 1.3

Write down the corresponding description of infimum.

A further characterisation (which follows from the one just given by taking σ ′ = σ − ǫ) is as follows.

Theorem 1.5 For S ⊆ R, S non-empty, σ = sup S if and only if σ is an upper bound for S and

∀ǫ > 0, ∃y ∈ S with y > σ − ǫ.

Learning activity 1.4

Convince yourself that this characterisation of the supremum is equivalent to the one given in the
definition.

There is also a similar characterisation of the infimum.

Theorem 1.6 For S ⊆ R, S non-empty, τ = inf S if and only if τ is a lower bound for S and

∀ǫ > 0, ∃x ∈ S with x < τ + ǫ.

1.4 Learning outcomes

At the end of this chapter and the relevant reading, you should be able to:

understand and use the triangle inequality


know what is meant by a set of real numbers being bounded above, bounded below, or bounded
explain exactly what is meant by the supremum and infimum of a non-empty set and be able to
determine these.

3
MA203 Real Analysis: MA103 revision material

1.5 Sample examination questions

Question 1.1 The distance between two real numbers x and y is defined as the absolute value
|x − y| of their difference. This is a non-negative real number, and satisfies |x − y| = |y − x| and
|x + y| ≤ |x| + |y|. Use these properties to show the following “triangle inequality”

|x − z| ≤ |x − y| + |y − z|

and the inequality


|x − y| ≥ ||x| − |y||.

Question 1.2 For each of the following sets, find the sup, inf, max and min whenever these exist.

[−1, ∞),

{x ∈ R | x2 − 2x − 1 < 0},
{x2 − 2x − 1 | x ∈ R}.

Question 1.3 Suppose A is a bounded subset of R. Let B be the set defined by

B = {b | b = 2a + 3, a ∈ A}.

Prove that
sup B = 2 sup A + 3.
[Hint: let σA = sup A and start by proving that 2σA + 3 is an upper bound on B.]

Question 1.4 A bounded subset A of R has the property that for all x, y in A, |x − y| < 1. Prove by
a contradiction argument that
(sup A − inf A) ≤ 1.

Question 1.5 Let S be a non-empty set of reals and let τ be a lower bound of S. Prove that

τ = inf S ⇐⇒ ∀ǫ > 0 ∃y ∈ S y < τ + ǫ.

1.6 Comments on selected activities

Learning activity 1.4 What we need to prove is that if S is a non-empty set of reals and σ is an
upper bound of S, then

σ = sup S ⇐⇒ ∀ǫ > 0 ∃y ∈ S y > σ − ǫ.

The proof is almost there once it is understood what the condition on the right-hand side means. It
says ‘for all positive ǫ there is an element y of S so that y > σ − ǫ’. So let us prove that this
condition holds if σ is a supremum, i.e. show the implication ‘=⇒’ of the equivalence that we want to
prove. Suppose otherwise; that is, σ is a supremum but ‘∀ǫ > 0 ∃y ∈ S y > σ − ǫ’ is false, that is,
there is some ǫ > 0 so that ∃y ∈ S y > σ − ǫ’ is false, that is, for all y in S we have that ‘y > σ − ǫ’
is false, that is, for all y in S we have y ≤ σ − ǫ. But this means that σ − ǫ is an upper bound for S,
which is less than σ since ǫ > 0, contradicting our assumption that σ is the least upper bound. So
‘=⇒’ indeed holds.

Conversely, assume that σ is an upper bound for S and that for all positive ǫ there is an element y of
S so that y > σ − ǫ. Then σ must be the least upper bound of S since if there was an upper bound
σ ′ of S with σ ′ < σ, taking ǫ = σ − σ ′ (which is > 0) would give an ǫ > 0 with ∀y ∈ S y ≤ σ − ǫ,
contradicting the assumption. Hence σ = sup S (since σ is an upper bound of S by assumption),
showing the implication ‘⇐=’ as well.

4
Chapter 1. Supremum and infimum

In fact, this theorem is not that difficult, but we have argued very carefully using the logic of
quantifiers and the negations of quantified formulae to get you used to that. Remember that
‘∀y P (y)’ means ‘for all y property P (y) holds’, and that its negation ‘¬∀y P (y)’ is equivalent to
‘∃y ¬P (y)’, meaning that P (y) is not true for all y if it is false for at least one y (a rather obvious
statement). Similarly, the statement ‘∃y P (y)’ means ‘there is a y so that P (y) holds’, and its
negation ‘there is no y such that P (y) holds’, that is, ‘¬∃y P (y)’, is equivalent to ‘∀y ¬P (y)’ and
means that for all y, P (y) is false (again, rather obviously). When quantifiers start to pile up, it is
useful to go through these negation processes step by step, as illustrated in the preceding proof.

1.7 Sketch answers to or comments on selected questions

Question 1.1 For the first inequality, note that, since |a + b| ≤ |a| + |b|, we have

|x − z| = |(x − y) + (y − z)| ≤ |x − y| + |y − z|.

For the second inequality, we need to prove

−|x − y| ≤ |x| − |y| ≤ |x − y|.

So there are two inequalities to prove. The leftmost follows from the triangle inequality, as follows:

|y| = |(y − x) + x| ≤ |y − x| + |x|,

so
−|x − y| = −|y − x| ≤ |x| − |y|.
For the second, we observe that

|x| = |(x − y) + y| ≤ |x − y| + |y|.

Question 1.2 For the first set, min=inf=−1, √ and there


√ is no sup and no max. For the second, we

note that the
√ set is the open interval (1 − 2, 1 + 2). There is no max or min, but sup = 1 + 2,
inf = 1 − 2. The third set equals [−2, ∞). No max or sup, but inf=min=−2.

Question 1.3 There are several ways of proving the result. It is given that A is bounded, so
σA = sup A exists and, a ≤ σA for all a ∈ A. So, ∀a, 2a + 3 ≤ 2σA + 3. So B is bounded above by
2σA + 3. Since B is bounded above, σB = sup B exists and, since 2σA + 3 is an upper bound, we
must have
σB ≤ 2σA + 3.
We need to show =, rather than ≤. We could now proceed to prove it by contradiction, supposing

that σB < 2σA + 3 and showing that, as a consequence of this supposition, σA = (σB − 3)/2 is an
upper bound on A, which is less than σA —a contradiction. Almost equivalently, we could simply
observe that
A = {(b − 3)/2 | b ∈ B}
so, arguing as at the beginning,
σA ≤ (σB − 3)/2;
that is,
2σA + 3 ≤ σB .
Either way, we get what we need to complete the proof.

Question 1.4 This is tricky. A diagram based on the real line might help. Suppose, for a
contradiction, that σ = sup A, τ = inf A, and σ − τ > 1. Suppose that σ − τ = 1 + r. Then, for any
s > 0, by the definition of sup and inf, there are x, y ∈ A with x < τ + s and y > σ − s. Let’s choose
s small enough that this will imply that y > x and also that y − x > 1 (giving our contradiction). To
make y > x, we need s < (σ − τ )/2. We then have

|x − y| = y − x = (σ − s) − (τ + s) = (σ − τ ) − 2s,

5
MA203 Real Analysis: MA103 revision material

and the right-hand side will be at least 1 if s < r/2.

In other words, we can formally argue as follows: Suppose σ − τ > 1, and that σ − τ = 1 + r where
r > 0. Let 0 < s < min(r/2, (σ − τ )/2). Then, by the properties of sup and inf, there exist y, x ∈ A
with y > σ − s and x < τ + s. We then have y > x (because s < (σ − τ )/2) and, because s < r/2,
we have

|x − y| = y − x > (σ − s) − (τ + s) = (σ − τ ) − 2s > (1 + r) − 2s > (1 + r) − r = 1.

So |x − y| ≥ 1, a contradiction to the given information.

Question 1.5 The proof is almost there once it is understood what the condition on the right-hand
side means. It says “for all positive ǫ there is an element y of S so that y < z + ǫ”. So let us prove
that this condition holds if z is an infimum, i.e., show the implication “=⇒” of the equivalence that
we want to prove. Suppose otherwise, that is, z is an infimum but “∀ǫ > 0 ∃y ∈ S y < z + ǫ” is
false, that is, there is some ǫ > 0 so that ∃y ∈ S y < z + ǫ” is false, that is, for all y in S we have
that “y < z + ǫ” is false, that is, for all y in S we have y ≥ z + ǫ. But this means that z + ǫ is a lower
bound for S, which is bigger than z since ǫ > 0, contradicting our assumption that z is the greatest
lower bound. So “=⇒” indeed holds.

Conversely, assume that for all positive ǫ there is an element y of S so that y < z + ǫ. Then z must
be the greatest lower bound of S since if there was a lower bound w of S with w > z, taking
ǫ = w − z would give an ǫ > 0 with the following property:

∀y ∈ S z + ǫ ≤ y,

contradicting the assumption. Hence z = inf S (since z is a lower bound of S by assumption),


showing the implication “⇐=” as well.

6
Chapter 2
Sequences and limits

2.1 Introduction

The single most important idea in the Analysis part of this subject is the notion of the limit of a
sequence, and this is the topic of this chapter. This idea can sometimes appear daunting, and many
students struggle with it when they first meet it. But it is of such importance in this subject that you
are urged to grapple with it until it makes sense to you.

2.2 Examples of sequences

A sequence is an infinite ordered list a1 , a2 , a3 , . . . of real numbers. An example is given by an = 1/n


for all natural numbers (positive integers) n; that is, the sequence

1 1 1 1
1, , , , , . . . .
2 3 4 5
What happens to the numbers 1/n as n gets larger and larger? Clearly they ‘approach’ 0. But how do
we formalise this? Before considering limits of sequences in a formal way, we consider another
important sequence of numbers (which will show us that things are not always so simple as they are
for the sequence just considered).

For a fixed x > 0, let the numbers a1 , a2 , a3 , . . . be defined by

x2 x2 x3
a1 = 1, a2 = 1 + x, a3 = 1 + x + , a4 = 1 + x + +
2 2 6
and, in general,
n−1
x2 xn−1 X xi
an = 1 + x + + ··· + = .
2 (n − 1)! i=0
i!

Although that sequence (obviously) increases with each step, it turns out that, as n increases, the
numbers approach a fixed number. That is, the sequence has a limit, which we will denote by
limn→∞ an . This is far from clear. The value of that limit depends on x and defines a very important
function in mathematics, which has its own name, exp(x). That is,
n−1
X xi
exp(x) = lim ,
n→∞
i=0
i!

which we write as

X xi
exp(x) = .
i=0
i!

(Be aware that the infinite sum notation used here is just a shorthand for the limiting value of the
finite sum of the an ; we don’t really ‘add up’ an infinite sequence of numbers. You will find out more
about such matters if you take Advanced Real Analysis, in which series are studied.) Considered as
a function of x, this is the so-called exponential function, a function of central importance in analysis
and calculus.

7
MA203 Real Analysis: MA103 revision material

2.3 Sequences and limits

2.3.1 Sequences: formal definition

Let N = {1, 2, 3, . . .} be the set of natural numbers (positive integers). Formally, a sequence is a
function from N to R. Only the notation is unusual: Instead of writing f (n) for the value of the
function f , say, for the natural number n, that value is written an . The entire sequence is then
written in one of the following ways, all legitimate:

(an )n∈N , (an )∞


n=1 , (an )n≥1 , (an ),

where the last expression (an ), the one we will use most often, is an abbreviation that should only be
used when it is clear that the index n runs through the natural numbers.

The nth element (or nth term) an of a sequence may be defined explicitly by a formula involving n,
as in the two examples given above. Alternatively, a sequence might be defined recursively (or
inductively). For example, we might have a1 = 1 and an+1 = an /2 for n ≥ 1.

2.3.2 The definition of a limit

As an example, let us calculate the first few terms of the sequence given by

an 3
a1 = 1, an+1 = + (n ≥ 1).
2 2an

We have
a1 = 1, a2 = 2, a3 = 1.75, a4 = 1.73214, a5 = 1.7320508, . . . .

Learning activity 2.1

Check these calculations of a2 , . . . , a5 .

Note that the terms get closer√and closer to 1.73205 . . . as n increases. In fact, the number√the terms
approach can be√shown to be 3 = 1.7320508057 . . .. We say that the sequence has limit 3, or that
it converges to 3.

Let’s take things easy to start with. We intuitively know what we mean by saying that the sequence
(1/n) tends to the limit 0 as n tends to infinity. In symbols, we may write 1/n → 0 as n → ∞. Well, I
say we know what it means, but do we, really? Before we go further, try the following learning activity.

Learning activity 2.2

Try to write down, precisely, what it means to say that a sequence (an ) of numbers tends to 0 as
n → ∞. It’s not that easy.

You might suggest something like ‘the terms get closer and closer to 0’, but although this is indeed a
property of the sequence (1/n), it is not a sufficient definition of tending towards 0. For, consider
an = 1/2 + 1/n. This does not tend to 0. (It tends to 1/2, as you can begin to see by calculating

8
Chapter 2. Sequences and limits

some of the numbers.) Although it is the case that the terms get closer and closer to 0. (The point
is, they don’t get ‘close enough’.) Consider now the sequence bn given as follows

1 1 1 1 1 1 1
, 1, , , , , , , . . . .
2 3 2 4 3 5 4
So, b2k = 1/k and b2k−1 = 1/(k + 1). This sequence tends to the limit 0, but it is not the case that
the terms get closer and closer to 0, because we have, for any k, b2k > b2k−1 .

Anyway, all this is just to try to convince you that we really do need a formal, precise, definition of
what it means for a sequence to tend to 0 if we are to have a notion of limit that works well. And
here it is. . ..

Definition 2.1 The sequence (an ) is said to tend to 0 (as n tends to infinity) if for all positive real
numbers ǫ there is a number N = N (ǫ) such that for any natural number n > N , the distance of the
element an from 0 is at most ǫ:

∀ǫ > 0 ∃N ∀n > N |an | < ǫ. (∗)

If this holds, we write


an → 0 as n→∞
and also
lim an = 0.
n→∞

The definition may also be understood as saying that for any ǫ > 0, there is N such that if n > N
then the ‘size’ |an | of the nth term is smaller than ǫ. Another interpretation is: given any ‘error
margin’ ǫ, the terms of the sequence are ultimately equal to 0 within that error margin.

It should be understood in (*) that ǫ is an arbitrary positive real number, whereas n is a natural
number. Also, N depends on ǫ, something I have stressed by writing N = N (ǫ).

As an example, consider the sequence (an ) defined by an = 1/n for n ∈ N. The limit of this sequence
is zero, which can be seen, formally, as follows: To show (*) holds, consider any ǫ > 0. Then we look
for an N such that whenever n > N , then |an | < ǫ, that is, 1/n < ǫ or, equivalently, 1/n < ǫ, that is,
n > 1/ǫ. So a suitable value for N in that case is N = 1/ǫ (which may not be an integer, but this
poses no problem since n in (*) is always an integer).

In general, (*) states that no matter how close one wants to be to the limit 0 (that is, no matter how
small ǫ is chosen), the elements an are eventually all closer to 0 than distance ǫ. ‘Eventually all’
means ‘all except for a finite number’, where these finite exceptions must be among the elements an
with 1 ≤ n ≤ N . In general, and fairly obviously, the smaller ǫ is, the larger N must be. We see this
in the above example, in which N = N (ǫ) = 1/ǫ.

What about a sequence tending to some limit other than 0? The following definition shows how we
can easily adapt the one just given.

Definition 2.2 The sequence (an ) is said to have the limit a (as n tends to infinity) if and only if
an − a tends to 0.

Explicitly, we therefore have:

Definition 2.3 The sequence (an ) is said to have the limit a (as n tends to infinity) if for all positive
real numbers ǫ there is a number N such that for any natural number n > N , the distance of the
element an from a is at most ǫ:

∀ǫ > 0 ∃N = N (ǫ) ∀n > N |an − a| < ǫ. (∗∗)

If this holds, one also says that the sequence tends to a as n tends to infinity, written as

an → a as n → ∞

9
MA203 Real Analysis: MA103 revision material

and also as
a = lim an .
n→∞
A sequence that has a limit is called convergent, otherwise divergent.

In general, (**) states that no matter how close one wants to be to the limit a (that is, no matter
how small ǫ is chosen), the elements an are eventually all closer to a than distance ǫ.

Definition (**) is crucial and should be understood thoroughly. Note that the following conditions are
equivalent:
|an − a| < ǫ,
a − ǫ < an < a + ǫ,
an ∈ (a − ǫ, a + ǫ) .
The interval (a − ǫ, a + ǫ) is also called the ǫ-neighbourhood of a.

Note also, for example, that the reference to N in (**) cannot be omitted: The (rather useless)
statement ∀ǫ > 0 ∃n |an − a| < ǫ would be trivially true if a was one of the elements an of the
sequence. Furthermore, it is also not sufficient that infinitely many elements of the sequence are close
to a: The sequence (an ) defined by an = (−1)n is given by −1, 1, −1, 1, −1, . . . and alternates
between 1 and −1. Either of these two numbers has an infinite number of elements of the sequence
close to it, but not eventually all of them. (Take ǫ = 1 in (**) where there is no N with the property
stated there.) This is an example of a divergent sequence.

We can give a diagrammatic representation of convergence of a sequence. Suppose we have a


sequence (an ) with limit L. Then, as illustrated in Figure 2.1, we can think of convergence as follows.
Pick any ǫ > 0, and consider the shaded strip of width ǫ around the horizontal line passing through L.
Then one can find some N ∈ N, large enough, such that all the terms an of the sequence, for n > N
lie in the shaded strip.

L+ǫ
L
L−ǫ

L+ǫ
L
L−ǫ

N N +1 N +2 N +3

Figure 2.1: Convergence of a sequence with limit L.

Certain sequences diverge because their members become arbitrarily large. For these sequences, a
useful definition can be given.

Definition 2.4 The sequence (an ) tends to infinity, written an → ∞ as n → ∞, if


∀K ∃N ∀n > N an > K.

10
Chapter 2. Sequences and limits

This is also written as


lim an = ∞ .
n→∞

Similarly, the sequence tends to minus infinity if (−an ) tends to infinity.

Note: If a sequence tends to infinity, it is divergent (not convergent). Note also that a sequence may
diverge because it tends to ∞ or −∞, or because, while remaining bounded like an = (−1)n , it
simply fails to tend towards any fixed real number. Do not confuse divergence with tending to infinity
or minus infinity: divergence is more general than this. The key thing is to be clear that convergence
means tending towards a (finite) real number.

As I mentioned, the definition of a limit is the most important thing in this part of the subject. It’s
even more important than any of the Theorems. Definitions matter. We need them because we need
to know precisely what we mean by saying, for example, that a sequence converges to a number. You
cannot even properly begin to prove things about the limits of sequences until you know the formal
definition. This should be clear: how can you prove that a sequence converges to 0 if you don’t even
know what ‘converges to 0’ means? Try to make sure, now, that you understand this definition. Its
meaning will become clearer as we work further with it. I know it might be difficult: this is known to
be one of the hardest conceptual hurdles in university mathematics. Bryant (on the back cover of his
book) claims that ‘A first course in Analysis at college is always regarded as one of the hardest in the
curriculum’. Anybody who has taught – or even studied – university-level mathematics is very aware
that it can initially be difficult to handle the concept of limit. So, it might be difficult at first, but it is
nonetheless important that you do your best to grapple with this important concept. Once the idea
‘clicks’ into place you will wonder why you ever had any problems with it.

2.4 Some standard results on and properties of limits

The notation limn→∞ an suggests that the limit of a sequence is unique. Indeed, this is the case:

Theorem 2.5 A sequence has at most one limit.

Proof. Consider a sequence (an ) and assume, to the contrary, that it has two limits a and b with
a 6= b. We will arrive at a contradiction by choosing ǫ in the definition of limit to be small enough.
Intuitively, if ǫ is less than half of the distance between a and b, then an element that is less than ǫ
away from a must be farther than ǫ away from b. So it can’t be the case that eventually all elements
of the sequence can be that close to both a and b.

To formalize this, let ǫ = |a − b|/2, which is positive since a 6= b. Then for some N1 and all n > N1 ,
since a is a limit of the sequence, |an − a| < ǫ, and similarly for some N2 and all n > N2 , since b is a
limit of the sequence, |an − b| < ǫ. But then for n > N = max{N1 , N2 },

2ǫ = |a − b| ≤ |a − an | + |an − b| = |an − a| + |an − b| < ǫ + ǫ = 2ǫ,

which is a contradiction. Hence it is not possible that the sequence has more than one limit.

A slightly different way to argue this, which is essentially the same as that just given, is as follows. As
above, let ǫ > 0 be any given positive number. Then, with N1 and N2 as above, we deduce that for
any n > N = max{N1 , N2 }, |a − b| < 2ǫ. Now, this must be true for any ǫ > 0. The only way that a
non-negative number X can satisfy X < 2ǫ for all ǫ > 0 is if X = 0. (For, if X > 0, then the
inequality X < 2ǫ will not hold if ǫ = X/2.)

Using the formal definition of limit

We now look at a few examples in which the formal definition of limit is explicitly used.

11
MA203 Real Analysis: MA103 revision material


Example Suppose that xn = 1/( n + 1). We will prove that xn → 0. Now, this means that we need
to show
√ that, for √
any ǫ > 0, there is N = N√(ǫ) such that if n > N , then xn < ǫ. Since
1/( n + 1) < 1/ n, it suffices to have 1/ n < ǫ, which means n > N = 1/ǫ2 .

Note that we do not need to find the smallest such N (ǫ): it is enough to find any suitable one. What
this means is√that it is fine to use the inequalities we used above, and we do not need to solve the
equation 1/( n + 1) = ǫ. (This can, however, be done, in this case. In other, more complicated
examples, it would be a nightmare.)

Example Suppose that xn = 1 − n12 . We prove that xn → 1. This means we must show that for any
We have |xn − 1| = 1/n2 . This is
ǫ > 0, there is N = N (ǫ) such that if√n > N , then |xn − 1| < ǫ. √
2
< ǫ provided n > ǫ; that is, n > 1/ ǫ. So a suitable N (ǫ) is 1/ ǫ

Example Let (xn ) be the sequence defined by


2n − 3
xn = .
n+3
We prove xn → 2. What does it mean to say xn → 2? It means that if somebody gives us any positive
ǫ, we have to produce a number N (ǫ) with the property that if n > N (ǫ) then |xn − 2| < ǫ. Now,

2n − 3 2n − 3 − 2(n + 3) −9 9
n + 3 − 2 = = n + 3 = n + 3.

n+3
We therefore want to know that 9/(n + 3) < ǫ for all n greater than some number N (ǫ). We could
solve this inequality for n to obtain n > 9/ǫ − 3. So a suitable N (ǫ) is 9/ǫ − 3. But we don’t need to
be so careful. Just approximate (in the right way, by which is meant bounding above): since
9/(n + 3) < 9/n, it suffices to have 9/n < ǫ; that is, n > N (ǫ) = 9/ǫ. This is a bit easier.

Example Let’s prove directly that


4n2 − n + 3
an = → 2.
2n2 − n + 1
The difference between an and 2 is
2
4n − n + 3
|an − 2| = 2 − 2
2n − n + 1
2
4n − n + 3 − 4n2 + 2n − 2

=
2n2 − n + 1
n+1
= 2
.
2n − n + 1
Now, we want to find N (ǫ) such that this will be < ǫ for all n > N (ǫ). Now, since the inequality
(n + 1)/(2n2 − n + 1) < ǫ is tricky, we approximate, by bounding above by some simple bn . Now,
we’d like to bound the numerator above, and get the terms on the numerator to be the same degree
as each other. Well, n + 1 ≤ n + n = 2n. What about the denominator? We can throw away the +1,
but we cannot just throw away the −n, because we need to lower bound the denominator. Well,
n ≤ n2 , so 2n2 − n + 1 ≥ 2n2 − n2 + 1 ≥ n2 , so we have (for n ≥ 2)
2n 2
|an − 2| ≤ 2
=
n n
and this is < ǫ provided n > N (ǫ) = 2/ǫ.

Fortunately, there are easier ways of proving limits, but these examples are all good, clean healthy fun
– and good practice in understanding the definition. Moreover, if an examination question asks you to
use the definition of limit, it is expected that you answer in this sort of way.

Bounded sequences

A sequence is said to be bounded if all its elements belong to an interval of the form [a, b]. (That is,
if the set of terms of the sequence is a bounded set of real numbers.) Then, a is any lower bound and

12
Chapter 2. Sequences and limits

b any upper bound of the set of elements of the sequence. The case that only lower or upper bounds
exist is also of interest.

Definition 2.6 Let (an ) be a sequence and S = {an | n ∈ N}. Then the sequence (an ) is said to be
bounded below if S has a lower bound, bounded above if S has an upper bound, and bounded if it is
bounded above and below.

Theorem 2.7 Any convergent sequence is bounded.

Proof. Let (an ) be a convergent sequence with limit a. Then the sequence is bounded since for
sufficiently large n, the elements an of the sequence are close to a, whereas there is only a finite
number of them that are not close to a (and those finitely many elements are also bounded). To
make this precise, we use the definition of a limit with ǫ = 1, so that for some N and all n > N we
have a − 1 < an < a + 1 (any other positive number ǫ instead of 1 would do as well). Then we define
as lower and upper bounds L and U

L = min ({an | 1 ≤ n ≤ N } ∪ {a − 1}) ,

U = max ({an | 1 ≤ n ≤ N } ∪ {a + 1}) ,


where these are minima and maxima of finite sets and therefore exist. Then clearly, L ≤ an ≤ U holds
for all n in N, so the sequence is indeed bounded.

This result says that convergent sequences are bounded, but there’s no reason at all why a bounded
sequence must converge. For example, the sequence given by xn = (−1)n is certainly bounded, but
does not converge.

Learning activity 2.3

Convince yourself that xn = (−1)n is bounded, but does not converge.

Monotonic sequences

The next definition concerns sequences a1 , a2 , a3 , . . . that satisfy

a1 ≤ a2 ≤ a3 ≤ · · ·

or
a1 ≥ a2 ≥ a3 ≥ · · ·
which are called increasing and decreasing, respectively. Note that some elements in such a sequence
can be equal to their predecessors, so the sequence does not have to go strictly up or down.

Definition 2.8 A sequence (an ) is increasing if an ≤ an+1 for all n ∈ N, decreasing if an ≥ an+1 for
all n ∈ N, and monotonic if it is increasing or decreasing.

The following result is very useful.

Theorem 2.9 An increasing sequence that is bounded above has a limit.

Proof. Let (an ) be an increasing sequence that is bounded above and consider the set of its elements
S = {an | n ∈ N}. There is a natural candidate for the limit of (an ), namely σ = sup S, which exists
since S is non-empty and has an upper bound by assumption. Because σ = sup S, if ǫ > 0 there is
some y ∈ S such that y > σ − ǫ. But, because S is simply the set of an , this means there is some
N ∈ N such that aN > σ − ǫ. Now, the sequence is increasing, so for all n > N , we must have

13
MA203 Real Analysis: MA103 revision material

an ≥ aN > σ − ǫ. Also, since σ is an upper bound on S, an ≤ σ for all n. Thus, for all n > N ,
σ − ǫ < an ≤ σ and, in particular, |an − σ| < ǫ. This shows that an → σ as n → ∞.

We could also, alternatively, prove the result by contradiction. Suppose, to the contrary, that σ is not
the limit of (an ). Then
∃ǫ > 0 ∀N ∃n > N |an − σ| ≥ ǫ.
So consider such an ǫ > 0 and take any N ∈ N, so that, according to this condition, there is some
n > N with |an − σ| ≥ ǫ. Since the sequence is increasing and since σ is an upper bound for S, this
means
σ − aN ≥ σ − an = |an − σ| ≥ ǫ.
Because N is an arbitrary natural number here, aN ≤ σ − ǫ for all N ∈ N. But this means that σ − ǫ
is an upper bound for S that is strictly less than the supremum σ of that set, and this is a
contradiction. So σ is indeed the limit of the sequence.

We have not only shown that an increasing sequence that is bounded above has a limit, but that this
limit is the supremum of its elements. Similarly, one can see the following:

Theorem 2.10 A decreasing sequence that is bounded below converges to the infimum of its
elements.

2.5 Algebra of limits

When computing the limits of a sequence, it is useful to apply arithmetic rules if the terms are the
sum, product, etc., of terms of sequences that have a known limit behaviour. For example, one can
prove using the formal definition of a limit that the sequence (an ) defined by

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

4 + 9/n2
an =
3 + 7/n + 11/n2

where the terms 9/n2 , 7/n and 11/n2 all have limit zero and can be replaced by their limit to obtain
that
lim 4 + 9/n2

n→∞
lim an =
lim 3 + 7/n + 11/n2

n→∞
n→∞
lim (4) + lim (9/n2 )
n→∞ n→∞
=
lim (3) + lim (7/n) + lim (11/n2 )
n→∞ n→∞ n→∞
4+0 4
= = .
3+0+0 3
Such an ‘algebra of limits’ is possible because of the following observation.

Theorem 2.11 Let (an ) and (bn ) be convergent sequences with limits a and b, respectively. Let C
be a real number and let k be a positive integer. Then as n → ∞,

(a) Can → Ca,


(b) |an | → |a|,
(c) an + bn → a + b,
(d) an bn → ab,
(e) akn → ak ,
(f) if bn 6= 0 for all n and b 6= 0, then 1/bn → 1/b as n → ∞.

14
Chapter 2. Sequences and limits

Proof. We prove (d) as an example to illustrate the proof method in one of the more difficult cases.
Note that (d) implies (e) by induction on k, taking (bn ) to be the sequence defined by bn = ak−1
n . To
show that an bn → ab as n → ∞, consider any ǫ > 0. The goal is to show that |an bn − ab| < ǫ for all
sufficiently large n. Here we need a trick in order to exploit that an → a and bn → b as n → ∞,
which is all we know at this point. The trick is to insert the term −abn + abn , which is zero, into the
sum to obtain
|an bn − ab| = |an bn − abn + abn − ab|
≤ |an bn − abn | + |abn − ab|
= |(an − a) bn | + |a (bn − b)|
= |an − a| · |bn | + |a| · |bn − b|.
We can make this less than ǫ if each of the two terms on the right-hand side is less than ǫ/2. Since
the sequence (bn ) converges, it is bounded, as proved above, so that for some real number B we have
|bn | < B for all n, where B > 0 (we will need this shortly since we divide by B). Then

|an − a| · |bn | < |an − a| · B,

so in order to have the left-hand side smaller than ǫ/2 it suffices to have
ǫ
|an − a| < ,
2B
which holds for n > N1 , say, since (an ) converges to a, so that then
ǫ
|an − a| · |bn | < |an − a| · B < .
2
Similarly, we obtain
|a| · |bn − b| < ǫ/2
immediately if a = 0, otherwise for all n > N2 so that
ǫ
|bn − b| < ,
2 · |a|

where N2 exists since (bn ) converges to b. Combining these observations, we let N = max{N1 , N2 }
so that for any n > N , |an bn − ab| < ǫ, as desired. This ‘reverse construction’ of N1 and N2 using a
ǫ
modified term like 2B instead of ǫ in the definition of limit is very typical in convergence proofs.

Be careful not to misinterpret (e). This works only for a fixed number k. So, for example, since
1 + 1/n → 1 as n → ∞, we have (1 + 1/n)k → 1k = 1 for any k ∈ N. However, (1 + 1/n)n → e as
n → ∞.

Learning activity 2.4

Prove (c) of the Theorem above by using the fact that

|(an + bn ) − (a + b)| = |(an − a) + (bn − b)| ≤ |an − a| + |bn − b|.

2.6 Some useful limits

The following result is useful.

Theorem 2.12 If |a| < 1, then limn→∞ an = 0.

15
MA203 Real Analysis: MA103 revision material

Proof. Suppose first that 0 < a < 1, and let xn = an . Then each term in the sequence (xn ) is
positive. Also, xn+1 = an+1 = axn < xn , so (xn ) is a decreasing sequence. Since xn ≥ 0 for all 0,
the sequence is bounded below. So the sequence is bounded and monotonic and hence it must
converge to some limit L. Now (convince yourself!), if xn → L then xn+1 → L too. But xn+1 = aL,
so by the Heredity results,
lim xn+1 = lim(axn ) = a lim xn = aL.
Hence L = aL, so L(1 − a) = 0. But since 1 − a 6= 0, this means L = 0. To sum up, then,
xn = an → 0 as n → ∞. If −1 < a < 0 then for any n, |an − 0| = |a|n . But 0 < |a| < 1, so we know
from what we’ve just shown that |a|n → 0 as n → ∞. Thus |an − 0| → 0. Hence an → 0. (This last
step can be seen as an application of a general result, namely that if |xn | → 0 then xn → 0. It should
not be thought, however, that if |xn | → L for general L then xn → L; that simply isn’t true, as can
be seen by considering the sequence xn = (−1)n .)

In this proof we used the simple observation that if xn → L then xn+1 → L. This is often useful in
determining what the limit of a recursively defined sequence must be (if it converges).

Learning activity 2.5

Prove that xn → L if and only if xn+1 → L.

Example Consider the example we met earlier:


an 3
a1 = 1, an+1 = + (n ≥ 1).
2 2an
If this sequence converges, to L, then an → L and an+1 → L too. But
an 3 L 3
an+1 = + → + ,
2 2an 2 2L
so we must have
L 3
L= + ,
2 2L
which may be written as
L2 3
L2 = + , L2 = 3,
2 2
√ √
so L = 3 or − 3. Since (as can easily be shown) an ≥ 0 for all n, we have L ≥ 0 and therefore √ the
limit (if it converges, as in fact it does, though to show this requires extra work), then the limit is 3.

Another useful result is the following. This is often described as saying that ‘nth powers dominate
powers of n’.

Theorem 2.13 Given any a > 1, and any k ∈ N, an /nk → ∞ as n → ∞. Also, given any b with
0 < b < 1 and any k ∈ N, bn nk → 0 as n → ∞.

2.7 The sandwich theorem

A useful result is the sandwich theorem (also known as the squeeze theorem).

Theorem 2.14 (Sandwich Theorem) Let (an ), (bn ), (cn ) be sequences such that an ≤ bn ≤ cn for
all n ≥ 1 and
lim an = L = lim cn .
n→∞ n→∞
Then
lim bn = L.
n→∞

16
Chapter 2. Sequences and limits

Proof. Because an → L and cn → L, given any ǫ > 0, there are N1 , N2 such that for n > N1 ,
|an − L| < ǫ and, for n > N2 , |cn − L| < ǫ. If we set N = max{N1 , N2 }, the larger of N1 and N2 ,
then for n > N we have n > N1 and n > N2 , so both the above conditions hold: |an − L| < ǫ and
|cn − L| < ǫ. These may be written as

L − ǫ < an < L + ǫ, L − ǫ < cn < L + ǫ,

so for n > N ,
L − ǫ < an ≤ bn ≤ cn < L + ǫ,

which says |bn − L| < ǫ. Thus bn → L as n → ∞.

As an example, consider the sequence (an )n∈N where

1 1 1
an = + + ... + 2 .
n2 + 1 n2 + 2 n +n

1 1 1
Clearly an ≤ n · = 1 ≤ . Furthermore, an ≥ 0. Thus, since
n2 + 1 n+ n
n

1
lim 0 = 0 = lim ,
n→∞ n→∞ n
we have by the sandwich theorem that
 
1 1 1
lim 2
+ 2 + ... + 2 = 0.
n→∞ n +1 n +2 n +n

Another application of the sandwich theorem is to prove that xn tends to zero if |x| < 1. One
approach to proving this goes as follows: Let ǫ > 0. We are looking for a number N so that n > N
implies |xn | < ǫ, which is equivalent to |x|n < ǫ. Here we can without loss of generality assume
x > 0, since otherwise we can replace x by −x and if x = 0 the claim holds trivially. Now observe the
following equivalences, where the logarithm is taken for any basis and log x < 0 since x < 1, so the
inequality reverses when dividing both sides by log x:

xn < ǫ
log(xn ) < log ǫ
n log x < log ǫ
log ǫ
n> .
log x

That is, if N = log ǫ/ log x, then n > N implies xn < ǫ as desired. The problem with this derivation
of N is that it uses the definition of the logarithm, which if done formally requires a number of other
concepts in analysis that are not yet proven at this point.

In short, we want a more elementary proof that xn converges to zero if 0 < x < 1. To achieve this,
observe that, because 0 < x < 1, x can be written as 1/(1 + h) for some h > 0. By induction on n,
one can easily prove
(1 + h)n ≥ 1 + hn

which implies
1 1 1
xn = ≤ < .
(1 + h)n 1 + hn hn
Then taking an = 0, bn = xn and cn = 1/hn and applying the sandwich theorem shows that xn → 0
since cn → 0 as n → ∞, as desired.

This is not the only way to show that xn → 0 if 0 < x < 1. We’ve seen another way above.

17
MA203 Real Analysis: MA103 revision material

2.8 Subsequences

Consider the sequence


1 1 1 1 1
1, , , , , , . . .
2 3 4 5 6
If we were to cross out every other term our sequence would become
1 1 1
, , ,...
2 4 6
The resulting sequence is called a subsequence of the original sequence. The first term of the new
sequence is the second term of the original sequence; the second term of the new sequence is the
fourth term of the original sequence, and so on. Thus the elements of the new subsequence can be
found by looking at the appropriate position in the original sequence. More precisely, the nth term of
the subsequence is simply the (2n)th term of the original sequence. The formal definition of a
subsequence is as follows.

Definition 2.15 Let (an )n∈N be a sequence and consider strictly increasing natural numbers
k1 , k2 , k3 , . . ., that is, k1 < k2 < k3 < k4 < · · ·. Then the sequence (akn )n∈N is called a subsequence
of the sequence (an )n∈N .

In the above example, we have obtained a subsequence given by the indices k1 = 2, k2 = 4, etc., in
general kn = 2n. The indices kn tell us precisely which terms to keep from the original sequence.

An alternative definition of a subsequence changes the set of indices N to a subset S of the natural
numbers. Namely, let S be an infinite subset of N. Then a subsequence of (an )n∈N is simply given by

(am )m∈S

with the understanding that the index m runs through the elements of S in increasing order. Namely,
here m = kn and the index set S is the set

S = {k1 , k2 , k3 , . . .} = {kn | n ∈ N} .

The set S can always be written in this way because it is infinite.

Theorem 2.16 Let (an ) be a sequence which tends to a limit a. Then any subsequence also tends to
the limit a.

Proof. The proof is almost immediate: Let an → a as n → ∞, and let (akn ) be a subsequence. Then
by the definition of limits,
∀ǫ > 0 ∃N ∀n > N |an − a| < ǫ.
Now, because the sequence k1 , k2 , . . . is increasing, we have kn ≥ n for all n. So, for all n > N , since
kn > N , we have |akn − a| < ǫ. This means that the subsequence (ank ) converges to a.

Here are two examples. The sequences (1/2n )n∈N and (1/n!)n∈N are subsequences of (1/n)n∈N .
Secondly, ((−1)2n )n∈N and ((−1)2n−1 )n∈N are subsequences of ((−1)n )n∈N . The latter are two
constant sequences with constant terms 1 and −1, respectively. From this we can see that ((−1)n )
has no limit by using the result that each subsequence of a convergent sequence has the same limit.
This is because ((−1)2n ) tends to 1 as n → ∞, whereas ((−1)2n−1 ) tends to −1.

Theorem 2.17 Every sequence has a monotonic subsequence.

Proof. We first give a nice illustration of this proof taken from Bryant’s book. Assume that (an ) is
the given sequence, and 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 at infinity (to the right) there is the sea. A hotel is said
to have the seaview property if it is higher than all hotels following it. That is, a hotel without
seaview is followed sooner or later by a hotel of at least that height. See Figure 2.2.

18
Chapter 2. Sequences and limits

...

1 2 3 4 5 6 ...

Figure 2.2: The seaview property.

Now there are only two possibilities: Either there are infinitely many hotels with seaview. Then they
form a decreasing (in fact strictly decreasing) subsequence. Or there is only a finite number of hotels
with seaview, so that after the last hotel with seaview, one can start with any hotel and then always
find one later that is at least as high, which is taken as the next hotel, then considering yet another
that is at least as high as that one, and so on. Then the subsequence of hotels generated in this way
is increasing, although not necessarily strictly.

Formally, let S = {m ∈ N | ∀n > m am > an } (which is the set of numbers of hotels with seaview).
Consider the following two possible cases.

Case 1. S is infinite. Then clearly, (am )m∈S is a decreasing subsequence of (an )n∈N .

Case 2. S is finite. Then if S is empty, let k1 = 1, otherwise let k1 = max(S) + 1. This means that
for all n ≥ k1 , the sequence element an does not belong to S (hotel n does not have seaview). Then
define inductively for n = 1, 2, 3, . . .

kn+1 = min{m ∈ N | m > kn and am ≥ akn },

which is possible since the set on the right-hand side is not empty (otherwise kn would belong to S).
In other words, kn+1 (which we intend to be the next index in the subsequence) is the smallest
number greater than kn so that akn+1 ≥ akn . This implies that k1 , k2 , k3 , . . . are the indices of a
subsequence of (an )n∈N so that ak1 ≤ ak2 ≤ ak3 ≤ · · ·. That is, (akn )n∈N is an increasing
subsequence of (an )n∈N .

We obtain from this as a corollary the following famous result.

Theorem 2.18 (The Bolzano-Weierstrass Theorem) Every bounded sequence of real numbers has
a convergent subsequence.

Proof. Let (an ) be a bounded sequence. By Theorem 2.17, it has a monotonic subsequence (akn ).
This subsequence is then also bounded and we have seen (Theorem 2.9) that bounded monotonic
sequences are convergent.

2.9 Learning outcomes

At the end of this chapter and the relevant reading, you should be able to:

explain what is meant by a sequence


work with sequences defined by a formula or defined recursively
explain precisely what it means to say that a sequence converges or tends to infinity or minus
infinity
determine limits and prove results using the formal definition of the limit of a sequence
explain what is meant by a sequence being bounded, bounded above or bounded below
comprehend the links between boundedness and convergence
prove and use the fact that bounded monotonic sequences converge

19
MA203 Real Analysis: MA103 revision material

calculate limits of sequences using the algebra of limits


prove and use the Sandwich Theorem
know precisely what is meant by a subsequence of a sequence
know that any bounded sequence has a convergent subsequence

2.10 Sample examination questions

Question 2.1 Suppose that (xn ) is a sequence of real numbers which converges to a limit L > 0.
Show that there is N ∈ N such that for all n ≥ N , xn > 0.

Question 2.2 Let (an ) be the sequence defined by

4n2 + 9
an = .
3n2 + 7n + 11
Using the formal definition of the limit of a sequence, show explicitly that

4
an → as n → ∞.
3
[Like the previous exercise, this means you must produce N (ǫ).]

Question 2.3 Let (an )n∈N be a sequence, and let (bn )n∈N be the sequence defined by bn = |an | for
n ∈ N. Which of the following two statements implies the other?
(a) (an ) converges.
(b) (bn ) converges.

Question 2.4 Prove that a sequence which is decreasing and bounded below converges.

Question 2.5 Let (xn ) be a sequence of non-negative real numbers (i.e., xn ≥ 0 for each n).
Suppose that the sequence converges to x. Prove that x ≥ 0. [Note, however, that a sequence of
positive terms need not have a positive limit; for example, (1/n) is a sequence of positive terms
converging to 0.]

Question 2.6 Suppose the sequence (xn ) of positive terms converges to 0. Prove that if yn = 1/xn
then the sequence (yn ) tends to infinity.

Let (xn ) be a sequence of positive terms such that the sequence (1/xn ) tends to infinity. Show that
(xn ) converges to 0.

Question 2.7 Find the limits as n → ∞ of:

4n − 5 5(32n ) − 1 1 + 2 + · · · + n n2 − 1
, , , 3 .
22n − 7 4(9n ) + 7 n2 n +1

Question 2.8 A sequence (xn ) is defined as follows. Let x1 be any positive real number and, for
n ≥ 1, let
x2 + K
xn+1 = n ,
2xn
where K is a fixed positive number. √
(i) By using the inequality a2 + b2 ≥ 2ab for real numbers a, b (or otherwise), prove that xn ≥ K
for n ≥ 2.
(ii) Prove that xn+1 ≤ xn for all n ≥ 2.
(iii) Deduce that (xn ) converges and find its limit.

20
Chapter 2. Sequences and limits

Question 2.9 Find limn→∞ xn in the following cases:

2n3 + 1
xn = ,
3n3 + n + 2
n (n + 1) 2n
xn = + + ··· + 2.
n2 n2 n
Question 2.10 Prove by induction that if n ∈ N and x ≥ −1 then

(1 + x)n ≥ 1 + nx.

Taking x = 1/ n in this, show that
2
√ 2/n

1
1 ≤ n1/n < 1 + n ≤ 1+ √ .
n
Deduce that
lim n1/n = 1.
n→∞

Question 2.11 Using the fact that for a > 0, a1/n → 1 as n → ∞, and a sandwiching argument, find
1/n
lim (1n + 2n + 3n + 4n ) .
n→∞

Question 2.12 Find  


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

Question 2.13 For the positive sequence (an ), it is given that there exist numbers N and α such
that α < 1 and
an+1
0< < α ∀n > N.
an
Prove that (an ) is eventually decreasing and that an → 0 as n → ∞. Give an example to show that if
we relax this condition and only have
an+1
0< < 1 ∀n > N,
an
then we cannot conclude that an → 0.

Question 2.14 Discuss the behaviour as n → ∞ of the following:


 n  n
3 1 2n 2n3 + 1 3
(n + n) , , ,
2 n3 + n n+1 4

22n + n √ √
, n+1− n.
n3 3n + 1

Question 2.15√ Let (an ) be a sequence of non-negative numbers. Prove that if an → L as n → ∞



then an → L as n → ∞. [This is not covered by the ‘algebra of limits’ results (—why not?):
prove the result explicitly using the formal definition of a limit.]

2.11 Comments on selected activities

Learning activity 2.4 In a proof like this, you need to know what information you have, precisely,
and you need to know where you want to get to, precisely. That all seems very obvious, but it’s
crucial. What we know is that an → a and bn → b. What that really means is the following two
statements: for all ǫ1 > 0, there is N1 such that |an − a| < ǫ1 for all n > N1 ; and, for all ǫ2 > 0,
there is N2 such that |bn − b| < ǫ2 for all n > N2 . (Why didn’t I just write plain old ǫ and N in each?
Well, I want to keep things as general as possible at the moment until I see what choices of ǫ I’ll want

21
MA203 Real Analysis: MA103 revision material

to use.) What we want to establish is that an + bn → a + b, which means we must show that for any
ǫ > 0, there is some N such that |(an + bn ) − (a + b)| < ǫ for all n > N . Now, if I want to do this
given the information I know, I have to relate |(an + bn ) − (a + b)| to the two quantities |an − a| and
|bn − b|. From the clue given, we have

|(an + bn ) − (a + b)| = |(an − a) + (bn − b)| ≤ |an − a| + |bn − b|.

How can I ensure that the right-hand side will be less than ǫ? Well, I can do this if I can guarantee
that |an − a| and |bn − b| are both < ǫ/2. So take ǫ1 = ǫ2 = ǫ/2. There there are N1 , N2 such that
|an − a| < ǫ/2 for n > N1 and |bn − b| < ǫ/2 for n > N2 . How big should N be so that both of these
hold for all n > N ? Well, take N = max(N1 , N2 ), the larger of N1 and N2 . Then we’ll have, for all
n > N,
ǫ ǫ
|(an + bn ) − (a + b)| = |(an − a) + (bn − b)| ≤ |an − a| + |bn − b| < + = ǫ,
2 2
which is exactly what we need.

That was a bit long-winded, just to show you the thought process. A perfectly acceptable way to
write a proof would be as follows:

Let ǫ > 0. Since an → a, there is N1 such that |an − a| < ǫ/2 for all n > N1 and, since bn → b,
there is N2 so that |bn − b| < ǫ/2 for all n > N2 . Then, for all n > N = max(N1 , N2 ), we have
ǫ ǫ
|(an + bn ) − (a + b)| = |(an − a) + (bn − b)| ≤ |an − a| + |bn − b| < + = ǫ.
2 2
It follows that an + bn → a + b.

Learning activity 2.5 Suppose first that xn+1 → L as n → ∞. Then, given ǫ > 0 there is N such
that n > N implies |xn+1 − L| < ǫ. This means that if n > N + 1 then |xn − L| < ǫ, and this shows
that xn → L. For the other part, if xn → L then for ǫ > 0 there is N so that n > N implies
|xn − L| < ǫ. Then it’s clearly true (because n + 1 > n – think about it!) that if n > N , we also have
|xn+1 − L| < ǫ.

2.12 Sketch answers to or comments on selected questions

Question 2.1 Taking ǫ = L/2 > 0 in the definition of a limit, there is N so that for n ≥ N ,
|xn − L| < L/2. In particular, for n ≥ N , xn − L > −L/2 and so xn > L/2 > 0.

Question 2.2 We have

4n2 + 9

4
|an − 4/3| = −
3n2 + 7n + 11 3
12n2 + 27 − 12n2 − 28n − 44

=
3(3n2 + 7n + 11)

28n + 17
= 2
9n + 21n + 33

28n + 17n

9n2
45n

9n2
5
= .
n
This is < ǫ for n > N (ǫ) = 5/ǫ.

If you do not make the approximations indicated (or something like them) then this becomes much
more difficult. Note the way the inequalities work: to be sure that |an − 4/3| < ǫ, we bound |an − 4/3|

22
Chapter 2. Sequences and limits

from above by some simpler sequence bn , and we then solve bn < ǫ. And, we bound |an − 4/3| by
above by upper bounding its numerator and lower bounding its denominator. Note also that we do
this in such a way that the resulting bn has all terms in its numerator of the same degree, so that
solving bn < ǫ becomes easy. Without approximations, you end up having to solve the inequality

28n + 17 < ǫ(9n2 + 21n + 33)

for n, and this is a messy quadratic.

Here’s another (slightly different) example. Let’s prove directly that

4n2 − n + 3
an = → 2.
2n2 − n + 1
The difference between an and 2 is
2 2
4n − n + 3 − 4n2 + 2n − 2

4n − n + 3 n+1
|an − 2| = 2 − 2 = =
2n2 − n + 1 .
2n − n + 1 2n2 − n + 1

Now, we want to find N (ǫ) such that this will be < ǫ for all n > N (ǫ). Now, since the inequality
(n + 1)/(2n2 − n + 1) < ǫ is tricky, we approximate, by bounding above by some simple bn . Now,
we’d like to bound the numerator above, and get the terms on the numerator to be the same degree
as each other. Well, n + 1 ≤ n + n = 2n. What about the denominator? We can throw away the +1,
but we cannot just throw away the −n, because we need to lower bound the denominator. Well,
n ≤ n2 , so 2n2 − n + 1 ≥ 2n2 − n2 + 1 ≥ n2 , so we have (for n ≥ 2)

2n 2
|an − 2| ≤ =
n2 n
and this is < ǫ provided n > N (ǫ) = 2/ǫ.

Fortunately, there are easier ways of proving limits, but this is all good, clean healthy fun – and good
practice in understanding the definition.

Question 2.3 (a) implies (b). For, suppose that an → L. Then for any ǫ > 0 there is N such that
n > N implies |an − L| < ǫ.

[Right, now for some ‘scratching around’: How do we prove that (bn ) converges. Well, first, we’d like
some idea of what it converges too. But this is easy: if the an can be made as close as we like to L
then, the fact that bn = |an | suggests that bn should converge, if it does converge, to |L|. So we’re
going to try to show that |bn − |L|| can be made as small as we like. Back to the formalities...]

We have
|bn − |L|| = ||an | − |L||.
For any numbers x and y, we have
||x| − |y|| ≤ |x − y|,
so
|bn − |L|| ≤ |an − L|.
For n > N , this is < ǫ. So we have shown that for all ǫ > 0, there is N such that |bn − |L|| < ǫ for
n > N . That is, bn → |L|.

(b) does not imply (a). To see this, we just need a counterexample. Consider the sequence (an ) given
by an = (−1)n . Then (bn ) is the sequence with every term equal to 1 and it is therefore, trivially,
convergent. But (an ) does not converge.

Question 2.4 Suppose the sequence is (xn ) and let τ = inf{xn | n ∈ N} (which exists, by
boundedness). Let ǫ > 0. Since τ + ǫ cannot be a lower bound on the set of xn , there’s N so that
xN < τ + ǫ. But the sequence is decreasing, so for n > N , τ ≤ xn ≤ xN < τ + ǫ and hence
|xn − τ | < ǫ. This shows the sequence converges to τ .

23
MA203 Real Analysis: MA103 revision material

Question 2.5 Proof by contradiction. We show x can’t be negative. Suppose it was. Then
−x/2 > 0, so taking ǫ = −x/2 in the definition of limit, there is N so that for n > N ,

|xn − x| < −x/2, i.e., 3x/2 < xn < x/2.

But xn < x/2 says xn < 0 for n > N —a contradiction. Draw a picture!

Question 2.6 Let K > 0. Taking ǫ = 1/K, there is N so that n > N implies |xn − 0| < 1/K. Since
xn > 0 this means 0 < xn < 1/K for n > N . So, for such n, yn = 1/xn > K. So yn → ∞. For the
second part, let ǫ > 0 and let K = 1/ǫ. Then, because 1/xn → ∞, ∃N such that for n > N ,
1/xn > K. But then, 0 < xn < 1/K = ǫ, so |xn | < ǫ. So xn → 0.

Question 2.7 We use the algebra of limits.


4n − 5 4n − 5 1 − 5/4n 1
2n
= n
= n
→ = 1.
2 −7 4 −7 1 − 7/4 1

5(32n ) − 1 5 − 1/9n 5
= → .
4(9n ) + 7 4 + 7/9n 4
For the next one, you might be tempted to write the n term as
1 2 n
xn = + 2 + ··· + 2,
n2 n n
and observe that
1 n
n× 2
≤ xn ≤ n × 2 ;
n n
that is,
1
≤ xn ≤ 1.
n
This is true, but is not of much use, because the sequence on the left tends to 0, which is not the same
as the limit on the right (which is 1). So although this bounds xn , it does not sandwich it between
two sequences that converge to the same limit. So the Sandwich theorem is no use here. Instead:

1 + 2 + ··· + n n(n + 1)/2 n2 /2 + n/2 1 + 1/(2n) 1


2
= 2
= 2
= → = 1.
n n n 1 1
(For the numerator, we can use the formula for the sum of an arithmetic progression.) Finally,

n2 − 1 1/n − 1/n3 0
3
= 3
→ = 0.
n +1 1 + 1/n 1
Be careful how you argue here. Don’t say something like

n2 − 1 n2 1
3
≃ 3 = → 0,
n +1 n n
where ≃ means ‘approximately equal to’. This isn’t precise enough. Also, don’t write

n2 − 1 n2 1
3
→ 3 = → 0.
n +1 n n
This has no meaning: sequences tend to numbers, not other sequences. (I know what you mean when
you write these things, but they are too vague.)

Question 2.8 We have, for n ≥ 2,


√ √
x2n−1 + k x2n−1 + ( k)2 2xn−1 k √
xn = = ≥ = k.
2xn−1 2xn−1 2xn−1

Alternatively, we can see that, since xn+1 = (x2n + K)/(2xn ), 2xn xn+1 = x2n + K. But we also know
that 2xn xn+1 ≤ x2n + x2n+1 . So,
x2n + x2n+1 ≥ x2n + K,

24
Chapter 2. Sequences and limits


so x2n+1 ≥ K for all n. Since each xn is clearly positive, this implies xn+1 ≥ K for all n, and hence

xn ≥ K for all n ≥ 2.

We now show the sequence is decreasing. Again, there are various ways in which this can be done.
For n ≥ 1,
x2 + k k − x2n
xn+1 − xn = n − xn = ≤ 0,
2xn 2xn

where we have used the fact that xn ≥ k. So xn+1 ≤ xn . Alternatively, we can note that

x2n + K x2 + x2n
xn+1 = ≤ n = xn .
2xn 2xn


So the sequence is bounded below (by k) and it is decreasing. These two facts mean that it
converges. Suppose xn → L. Then xn+1 → L too. But

x2n + k L2 + k
xn+1 = → ,
2xn 2L

and so√L = (2L2 + k)/2L, or L2 = k. Since xn is positive for all n, L ≥ 0, so L = k, and we have
xn → k.

Question 2.9 First,


2n3 + 1 2 + 1/n3 2
= → .
3n3 + n + 2 3 + 1/n2 + 2/n3 3
For the second, although it looks like it’s worth trying, the Sandwich Theorem doesn’t help much.
Instead, we use

n + (n + 1) + · · · + 2n = (n + 1)(n + 2n)/2 = 3n2 /2 + 3n/2,

so
3n2 /2 + 3n/2 3/2 + 3/(2n)
xn = = → 3/2.
n2 1
(Straightforward Sandwich theorem application only tells us that xn lies between 1 and 2.)

Question 2.10 When n = 1 both sides are 1 + x, so it holds. Suppose it’s true for n = k. Then
(1 + x)k ≥ 1 + kx and

(1 + x)k+1 = (1 + x)(1 + x)k ≥ (1 + x)(1 + kx) = 1 + (k + 1)x + kx2 ≥ 1 + (k + 1)x,



and so it’s true for n = k + 1. With x = 1/ n we get
√ √
(1 + n/ n)n ≥ 1 + n/ n,

which is equivalent to √ √
(1 + n) ≥ (1 + n)1/n ,
from which we get (on squaring),
√ √ √
(1 + n)2 ≥ (1 + n)2/n > ( n)2/n = n1/n .

Obviously, since n ≥ 1, n1/n ≥ 1. The chain of inequalities stated in the question is now established.
The term on the far left is the constant 1 and the term on the far right tends to (1 + 0)2 = 1. By the
Sandwich theorem, n1/n → 1 too.

Question 2.11 We have, just using some crude bounding of the quantity inside the parentheses,

(4n )1/n ≤ xn = (1n + 2n + 3n + 4n )1/n ≤ (4n + 4n + 4n + 4n )1/n .

That is,
4 ≤ xn ≤ 4(41/n ).
41/n → 1, so both sides tend to 4. Hence xn → 4 as n → ∞.

25
MA203 Real Analysis: MA103 revision material

Question 2.12 Let xn denote the quantity we need to find the limit of. Then
1 1
n× √ ≤ xn ≤ n × √ .
n2 + n n2 + 1
Now,
n 1 1
√ =p → = 1,
n2+n 1 + 1/n 1
and
n 1 1
√ =p → = 1.
n2 +1 1+ 1/n2 1
By the Sandwich theorem, xn → 1.

Question 2.13 For n > N , an+1 < αan < an , so the sequence is decreasing after N . Since an > 0,
the sequence is bounded below (by 0). So it converges, to some limit L ≥ 0. (At this stage, this is all
we know about L. We have to prove that L = 0.) If the limit is L then we must have also an+1 → L.
If L 6= 0, then an+1 /an → L/L = 1, which is contrary to an+1 /an < α < 1. So L = 0. We may also
argue, alternatively, that since an+1 < αan for n > N , then L ≤ αL, from which it again must follow
that L = 0 since α < 1.

Question 2.14 First,

(n3 + n)(1/2)2 = n3 (1/2)2 + n(1/2)n → 0 + 0 = 0.

The second example is the reciprocal of the first, so it tends to ∞. For the third, we have
 n  n  n
2n3 + 1 3 3n3 3 2 3
0< < = 3n → 0,
n+1 4 n 4 4

so the term tends to 0. Next,


22n + n 22 n (4/3)n
3 n
> 3 n = → ∞,
n 3 +1 2n 3 2n3
so the term tends to ∞. The last one is different. The clever thing is to realise that
√ √ √ √
( n + 1 − n)( n + 1 + n) = (n + 1) − n = 1,

so
√ √ 1 1
0< n+1− n= √ √ < √ → 0,
n+1+ n 2 n
so the expression tends to 0.

Question 2.15 We use the fact that


√ √ √ √
( an − L)( an + L) = an − L.

First, we deal with the case L = 0. If an → 0 then, given ǫ > 0, there is N so that for n > N ,
√ √
|an − 0| < ǫ2 . So, for n > N , 0 ≤ an < ǫ, and so an → 0. Suppose L 6= 0. Then (taking
ǫ = L/2), there is N1 so that for n > N1 , an > L/2. Further, given ǫ, since an → L, there is N2 so
that |an − L| < ǫ for n > N2 . So, for n > N = max(N1 , N2 ),

√ √ |an − L| ǫ
| an − L| √ √ <p √ .
an + L L/2 + L
√ √
The denominator is just a constant, so this proves an → L.

Here is a different proof, this time based on the fact that


√ √ √
x+y ≤ x+ y

for x and y non-negative (to see this, square both sides).

26
Chapter 2. Sequences and limits

Choose ǫ > 0, and take N so that, for n > N , |an − L| < ǫ2 , or in other words L − ǫ2 < an < L + ǫ2 .
Then we have √ p √ p √
L − ǫ ≤ L − ǫ2 < an < L + ǫ2 ≤ L + ǫ,
√ √
which implies that | an − L| < ǫ.
Again, this proof doesn’t quite work if L = 0 (why not?), and we need to treat that case separately,
as above.
Make sure you see where each of the inequalities in the string above comes from. Of course, when
constructing the proof, you work from both ends, so the last thing to fill in is that you need to start
with ǫ2 .

Do you see why this result doesn’t follow from the ‘Algebra of Limits’ theorem? That tells us that if
an → L then for any fixed positive integer k, akn → Lk , but it tells us nothing about non-integer
powers such as 1/2.

27
MA203 Real Analysis: MA103 revision material

Chapter 3
Limits of functions and continuity

3.1 Introduction

In the previous chapter we studied limits of sequences. In this chapter, we look at limits of functions.
We also look at a key property that functions might have, namely continuity.

3.2 Limit of a function

3.2.1 Definition of limit

Definition 3.1 Let f : R → R be a function. We say that L is the limit of f (x) as x approaches a,
denoted by limx→a f (x) = L (or f (x) → L as x → a) if for each ǫ > 0, there exists δ > 0 such that
0 < |x − a| < δ =⇒ |f (x) − L| < ǫ.

The definition states that if someone gives us any arbitrarily small ǫ, then there is some
neighbourhood of a, (a − δ, a + δ), such that any x in this neighbourhood – other than a itself – will
have f (x) in the ǫ-neighbourhood (L − ǫ, L + ǫ) of L. We start by defining what we mean by the
limit of a function f (x) as the argument x approaches some value a ∈ R. Note that we are not
concerned in this definition about the value of the function at the point a (indeed we don’t even care
if f is defined at a): the notion of limit applies only to the behaviour of the function f in a
neighbourhood near to a, and not at a itself.

Figure 3.1 indicates what f (x) → L as x → a means graphically.

L+ǫ
L
f (x)
L+ǫ

a−δ x a a+δ

Figure 3.1: A function with limit L as x tends to a. Given any ǫ > 0 (which determines a strip around
the line y = L of width 2ǫ), there exists a δ > 0 (which determines an interval around the point a of
width 2δ such that whenever x lies in this interval (but x 6= a), (so that x satisfies 0 < |x − a| < δ),
then f (x) satisfies L − ǫ < f (x) < L + ǫ, that is, |f (x) − L| < ǫ.

We can also make the following definition.

Definition 3.2 Let f : R → R be a function. We say that f (x) tends to infinity as x → a (denoted
by f (x) → ∞ as x → a) if for each K, there exists δ > 0 such that
0 < |x − a| < δ =⇒ f (x) > K.

28
Chapter 3. Limits of functions and continuity

Learning activity 3.1

Write down the definition of f (x) → −∞ as x → a.

3.2.2 Examples

Example Suppose that f (x) = 3x − 1. We can then show directly from the definition of a limit that
limx→1 f (x) = 2. To prove this, we must argue that for any ǫ we can bound the value
|f (x) − 2| = |(3x − 1) − 2| < ǫ in some neighbourhood of 1. But we easily see that

|f (x) − 2| = 3|x − 1|.

In other words, the distance of f (x) from 2 is three times the distance of x from 1. Thus if we choose
x to be within distance ǫ/3 from 1, then f (x) is within distance ǫ of 2.

A formal proof would go as follows. Let ǫ > 0. Then, let δ = ǫ/3. For any x s.t. 0 < |x − 1| < δ we
have:
ǫ
|f (x) − 2| = 3|x − 1| < 3 · < ǫ.
3
Therefore f (x) → 2 as x → 1.

Example As a second example, consider f (x) = x2 + x. We show that limx→2 f (x) = 6. To see this,
suppose that ǫ > 0 is given. If x is such that 0 < |x − 2| < δ, then

|f (x) − 6| = |x2 + x − 6| = |x + 3||x − 2| < |x + 3|δ.

Now, if |x − 2| < δ then |x + 3| ≤ 5 + |x − 2| < 5 + δ, by the triangle inequality. So to make


|f (x) − 6| less than ǫ, it suffices to have (5 + δ)δ ≤ ǫ. There is no need to solve a quadratic equation
in δ. If we just put δ = min{1, ǫ/6}, the smaller of 1 and ǫ/6, then we have 5 + δ ≤ 6 and δ ≤ ǫ/6, so

(5 + δ)δ ≤ 6(ǫ/6) = ǫ,

and we’re done.

Another way of arguing this is as follows. We have

|f (x) − 6| = |x + 3||x − 2|.

If we assume that |x − 2| < 1 then 1 < x < 3 and so 4 < x + 3 < 6 and so |x + 3| < 6. It follows that
|f (x) − 6| < 6|x − 2|. So to have |f (x) − 6| < ǫ, it will suffice to have both |x − 2| < 1 and
|x − 2| < ǫ/6, which will hold if |x − 2| < δ = min{1, ǫ/5}. (By the way, there is nothing special
about the choice of 1 in assuming |x − 2| < 1 here: we could equally well have assumed, for example,
that |x − 2| < 2.)

Learning activity 3.2

Prove that f (x) = x2 − 2 → 2 as x → 2.

3.2.3 Algebra of limits

Let f, g : R → R be two functions and c be any real number, then we derive other functions by
applying an algebra on the set of functions. For example, a new function (f + g) is obtained by

29
MA203 Real Analysis: MA103 revision material

defining for each x, (f + g)(x) = f (x) + g(x). We say that (f + g) is derived point-wise since the
value of (f + g) at x is defined by the normal arithmetic sum of the two real numbers f (x) and g(x).
Similarly, we may define the functions |f |, (cf ), (f − g), (f + g), (f · g) and (f /g), provided g(x) 6= 0.

Theorem 3.3 Let f, g : R → R be two functions and c be any real number. Suppose that
limx→a f (x) = L and limx→a g(x) = M . Then

1. limx→a (cf )(x) = cL


2. limx→a (|f |)(x) = |L|
3. limx→a (f + g)(x) = L + M
4. limx→a (f − g)(x) = L − M
5. limx→a f (x)g(x) = LM
6. limx→a (f /g)(x) = L/M provided g(x) 6= 0 for each x in some neighbourhood of a.

Note that in this theorem, I have used f (x)g(x) rather than the more usual f g in order to avoid
confusion with the composition of the functions.

3.2.4 More on limits

Sometimes, we may have a situation where the limit from one side is different from that from the
other. We adapt the definition as follows.

Definition 3.4 Let f : R → R be a function. We say that L is the limit of f (x) as x approaches a
from the left (or from below), denoted by limx→a− f (x) = L if for each ǫ > 0, there exists δ > 0 such
that
a − δ < x < a ⇒ |f (x) − L| < ǫ.

A similar definition applies to limits from the right (or from above), denoted limx→a+ f (x) = L. For
example, suppose the graph of a function looked like the following. Then the left-hand and right-hand
limits at a are L, M respectively.

M u

-
a

So far we’ve discussed limits of a function as x approaches some value a. But we can also make the
following definition (similar to the definition of the limit of a sequence).

Definition 3.5 Let f : R → R be a function. We say that L is the limit of f (x) as x approaches ∞,
denoted by limx→∞ f (x) = L if for each ǫ > 0, there exists M > 0 such that
x ≥ M ⇒ |f (x) − L| < ǫ.

30
Chapter 3. Limits of functions and continuity

1
Example Consider f (x) = . Then f (x) → 0 as x → ∞. For,
x2 +1
1
|f (x)| ≤ <ǫ
x2
p
if x > M = 1/ǫ.

3.3 Continuity

Loosely speaking, a function is continuous if its graph can be drawn without lifting pen from paper.
For a formal definition of this notion, we first define continuity at a point.

Definition 3.6 A function is continuous at the point a if f (a) is defined and is equal to limx→a f (x).

Definition 3.7 A function is continuous if it is continuous at each point a.

Definition 3.8 A function is continuous on the interval [a, b] if it is continuous at each point in (a, b)
and (i) f (a) = limx→a+ f (x) and (ii) f (b) = limx→b− f (x).

As an example of a discontinuous function (that is, one that is not continuous), consider the ‘sign’
function
1 if x > 0,
(
f (x) = 0 if x = 0,
−1 if x < 0.
This function f (x) makes a ‘jump’ when x = 0. This represents a discontinuity since when we
approach zero from the left (that is, with negative values for x), the function always has value −1, no
matter how close we are to zero, whereas at zero it has a different value. That is, the function values
when approaching zero ‘tend to’ something other than the value of the function. (The same happens
when zero is approached from the right, but already a strange behaviour like this when coming from
one direction is enough to make the function discontinuous.)

Figure 3.2 gives a diagrammatic representation of continuity.

f (a) + ǫ
f (a)
f (x)
f (a) + ǫ

a−δ x a a+δ

Figure 3.2: The definition of the continuity of a function at point a. If the function is continuous at
a, then given any ǫ > 0 (which determines a strip around the line y = f (a) of width 2ǫ), there exists
a δ > 0 (which determines an interval around a of width 2δ) such that whenever x lies in this interval
(so that x satisfies a − δ < x < a + δ, that is, |x − a| < δ), f (x) satisfies f (a) − ǫ < f (x) < f (a) + ǫ,
that is, |f (x) − f (a)| < ǫ.

It follows from the results on the algebra of limits that:

Theorem 3.9 Let f, g : R → R be functions which are continuous at a ∈ R and c be any real
number. Then |f |, (cf ), (f − g), (f + g), (f (x)g(x)) are all continuous at a, and (f /g) is continuous
provided g(x) 6= 0 for any x in some neighbourhood of a.

31
MA203 Real Analysis: MA103 revision material

Pk
As a corollary we see that any polynomial p(x) = i=0 ai xi is continuous. This can be proved by
induction as follows. Clearly any constant function is continuous by the above results. Moreover, the
function f (x) = x is continuous. Thus by the product rule, so is f (x) = x2 , and by induction
f (x) = xk is continuous for any finite k. Hence so is any function of the form f (x) = ai xi . Finally,
by repeated application of the summation rule, the polynomial p(x) is deduced to be continuous.

Recall that if f, g are functions, then the composition function f ◦ g is defined by


(f ◦ g)(x) = f (g(x)) for each x.

Theorem 3.10 If g is a function which is continuous at a, and f is a function which is continuous at


g(a). Then (f ◦ g) is continuous at a.

Proof. Let’s be clear about what we want to prove. To show that the composite function is
continuous at a, we must show that for any ǫ > 0, there is δ > 0 such that |x − a| < δ implies
|f (g(x)) − f (g(a))| < ǫ. By the continuity of f at g(a), we know that for any given ǫ > 0 there is
some δ1 > 0 such that if |y − g(a)| < δ1 then |f (y) − f (g(a))| < ǫ. So if we can find δ > 0 such that
|x − a| < δ implies that (taking y = g(x)) |g(x) − g(a)| < δ1 , then we’ll have that |x − a| < δ implies
|f (g(x)) − f (g(a))| < ǫ. But we can certainly find such a δ, simply from the definition of g being
continuous at a. (Note the use of the notation δ1 here, so that we do not confuse ‘intermediate
δ-values’ with the δ we need to finally produce.)

3.4 Continuity and sequences

We now give an alternative definition of continuity which ties in the concept of limits for sequences.

Theorem 3.11 A function f is continuous at a if and only if for each sequence (xn ) such that
limn→∞ xn = a we have limn→∞ f (xn ) = f (a).

Proof. Let (*) be the statement that for any sequence (xn ) such that limn→∞ xn = a,
limn→∞ f (xn ) = f (a).

Suppose first that f is continuous at a. We prove that this implies (*). Let (xn ) be a sequence of
reals converging to a. We want to show that f (xn ) → f (a) as n → ∞, that is,

∀ǫ > 0 ∃N ∀n ≥ N |f (xn ) − f (a)| < ǫ. (∗∗)

To prove this, let ǫ > 0. Choose, according to the definition of continuity, a δ > 0 so that for all x,
whenever |x − a| < δ, then |f (x) − f (a)| < ǫ. Since xn → a as n → ∞, there is an N so that n ≥ N
implies |xn − a| < δ, which in turn implies |f (xn ) − f (a)| < ǫ. This shows (**) as desired.

Conversely, assume that property (*) holds. In order to show continuity, we assume, to the contrary,
that the function is discontinuous at a. This means that there is an ǫ > 0 so that for all δ > 0 there is
an x with |x − a| < δ but |f (x) − f (a)| ≥ ǫ. In particular, for every natural number n, letting
δ = 1/n, there is a real number x, call it xn , with |xn − a| < 1/n but |f (xn ) − f (a)| ≥ ǫ. But then
clearly xn → a as n → ∞, but we do not have f (xn ) → f (a) as n → ∞, a contradiction to (*).

We find this a very useful alternative for the sake of manipulation of limits since it states that for any
sequence xn → a, the limit of the f (xn )’s is simply f applied to the limit of the xn ’s:
 
lim f (xn ) = f lim xn .
n→∞ n→∞

3.4.1 Continuous functions on closed intervals

Definition 3.12 For a subset X of the domain of a function f , we say that f is bounded on X if
there exists M such that |f (x)| ≤ M for each x ∈ M .

32
Chapter 3. Limits of functions and continuity

Definition 3.13 We define the supremum (or maximum) of f on X as sup{f (x) | x ∈ X} (or
max{f (x) | x ∈ X} if it exists).

The following result is very important and useful. It is often known as the Extreme Value Theorem.

Theorem 3.14 Let f be continuous on [a, b]. Then f is bounded on [a, b] and it achieves its
maximum; that is, the supremum is equal to the maximum.

Proof. Suppose first that f is unbounded above. For each n ∈ N, let xn be a point in [a, b] such that
f (xn ) > n. The sequence (xn ) is bounded, so has a convergent subsequence (xkn ), tending to some
limit c (by Theorem 2.18). Necessarily c ∈ [a, b]. Since f is continuous at c, f (xkn ) → f (c) as
n → ∞. But this contradicts the construction of the sequence (xn ), since f (xkn ) > n → ∞. So f is
bounded above. Let M = sup{f (x) | x ∈ [a, b]}. For each n ∈ N, let xn be a point in [a, b] such that
f (xn ) > M − n1 . Again take a convergent subsequence (xkn ) of (xn ), tending to some limit
c ∈ [a, b]. Arguing as before, we see f (c) = M .

The same result holds with max/sup replaced by min/inf.

3.5 The Intermediate Value Theorem

In this section we prove one of the most fundamental (and obvious!) theorems in real analysis. One
useful property of continuous functions f lies in the fact that they have solutions x to equations of
the form f (x) = C for any given C where such a solution might reasonably be expected, namely if f
takes values below and above C. In other words, a continuous function cannot “hop over”
intermediate values as it moves from one value to another. This central property of continuous
functions is known as the “intermediate value theorem”.

Theorem 3.15 (The Intermediate Value Theorem) Let f be a continuous function on [a, b] and
let K be such that f (a) < K < f (b). Then for some c ∈ (a, b), f (c) = K.

Figure 3.3 helps us understand this theorem.

f (b)
K

f (a)

a c b

Figure 3.3: The Intermediate value theorem: the graph of f (x), passing from y-coordinate f (a) to
y-coordinate f (b) as x passes from a to b, must pass through all y-values in between f (a) and f (b).

We prove a special case of the result, from which the full Theorem follows.

Theorem 3.16 Let f be a continuous function on [a, b] such that f (a) < 0 and f (b) > 0. Then for
some c ∈ (a, b), f (c) = 0.

Proof. We construct a sequence of intervals [an , bn ] such that

1. f (an ) < 0, f (bn ) > 0 for each n


2. [an+1 , bn+1 ] ⊆ [an , bn ] for each n.

33
MA203 Real Analysis: MA103 revision material

We start by letting [a1 , b1 ] = [a, b]. Then for each n ≥ 1, we define [an+1 , bn+1 ] as follows. Let
cn = (an + bn )/2, be the midpoint of the previous interval. If f (cn ) = 0, then the theorem is proved
and so we need not continue constructing intervals!

Otherwise, if f (cn ) < 0, we define an+1 = cn and bn+1 = bn . And if f (cn ) > 0, we define bn+1 = cn
and an+1 = an . Note that the condition 1. is satisfied by choosing our intervals in this manner.

Moreover, note that the (n + 1)st interval is half the size of the nth interval and so
bn+1 − an+1 ≤ (b1 − a1 )/2n . It follows that
lim (bn − an ) = 0. (3.1)
n→∞

Finally, note that (an ) is increasing and bounded above (by b1 ) and so it has a limit; similarly (bn ) is
decreasing and bounded below and so has a limit. Thus by (3.1) (and algebra of limits) these limits
are equal to, say, c. Thus by continuity (using Theorem 3.11),
f (c) = lim f (bn ) ≥ 0,
n→∞

where the last inequality follows from the fact that each f (bn ) ≥ 0 (in fact > 0). Similarly,
f (c) = lim f (an ) ≤ 0.
n→∞

Thus f (c) must be equal to zero, and the proof is complete.

Learning activity 3.3

Show that Theorem 3.16 implies Theorem 3.15. [Hint: consider g defined by g(x) = f (x) − K.]

3.6 Learning outcomes

At the end of this chapter and the relevant reading, you should be able to:

explain what is meant formally by the limit of a function at a point, or as x tends to ∞ or −∞,
and by one-sided limits
determine limits and prove results using the formal definitions of the limit of a function
calculate limits of functions using the algebra of limits
explain what is meant by continuity
prove functions are continuous or discontinuous using the formal definition and the algebra of
limits
explain and be able to use the interaction between continuous functions and convergent sequences
use the fact that a continuous function has a maximum and a minimum value on a closed
bounded interval
state and use the Intermediate Value Theorem

3.7 Sample examination questions

Question 3.1 Prove, from the formal definition of a limit, that f defined on R \ {2} by
5
f (x) =
(x − 2)2
tends to infinity as x → 2.

34
Chapter 3. Limits of functions and continuity

Question 3.2 Evaluate the following limit:


x3 + 5x + 7
lim .
x→1 x4 + 6x2 + 8

Question 3.3 Let ⌊x⌋ denote the largest integer n such that n ≤ x. Determine
lim (x − ⌊x⌋) and lim (x − ⌊x⌋).
x→1+ x→1−

Question 3.4 Let f : R → R be defined by


(x − 1)2 if x < 1;
(
f (x) = 1 if x = 1;
3x + 2 if x > 1.
Use the formal definitions of limits to prove that f (x) → 0 as x → 1− and f (x) → 5 as x → 1+. Is
the function

continuous
continuous on the right
continuous on the left

at the point 1?

Question 3.5 The function f is defined on R by


0, if x is rational;
n
f (x) =
1 if x is irrational.
Prove that f is discontinuous at every point of R. [Hint: use the fact that there are irrational
numbers arbitrarily close to any rational number and that there are rational numbers arbitrarily close
to any irrational number.]

Question 3.6 Suppose the real function f is continuous at c and f (c) > 0. Prove, directly from the
definition of continuity, that there is δ > 0 such that f (x) is positive for x in the interval (c − δ, c + δ).

Question 3.7 Suppose the real function f is continuous, positive and unbounded on R and that
inf{f (x) | x ∈ R} = 0. Use the Intermediate Value Theorem to prove that the range of f is (0, ∞),
the set of all positive real numbers. [It might be obvious, but give a watertight proof. Explicitly, prove
that for any y > 0 there is some c ∈ R such that y = f (c).]

Question 3.8 Let T (θ), for 0 ≤ θ ≤ 2π, be the surface temperature at the point at θ degrees
longitude on the equator. (Note that T (0) = T (2π).) Assuming T is a continuous function of θ,
prove that at any given time, there are two points on the equator which have the same temperature
and are diametrically opposite. [Regrettably, field trips are not available. Instead, consider the
function f (θ) = T (θ) − T (θ + π), and use the Intermediate Value Theorem.]

Question 3.9 Suppose that the real function f is continuous on the closed interval [a, b] and that f
maps [a, b] into [a, b]. By considering the function h(x) = f (x) − x, show that there is c ∈ [a, b] with
f (c) = c.

Suppose the real function g is continuous on R and that g maps [a, b] into [d, e] and maps [d, e] into
[a, b], where
a < b, d < e.
By considering the function
k(x) = g(g(x)),
prove that there are p, q ∈ R such that
g(p) = q, g(q) = p.
Hence show that there is c ∈ R such that g(c) = c.

35
MA203 Real Analysis: MA103 revision material

3.8 Comments on selected activities

Learning activity 3.1 We say that f (x) tends to minus infinity as x → a (denoted by f (x) → −∞
as x → a) if for each K, there exists δ > 0 such that

0 < |x − a| < δ =⇒ f (x) < K.

Learning activity 3.2 We have

|f (x) − 2| = |(x2 − 2) − 2| = |x2 − 4| = |(x + 2)(x − 2)| = |x + 2||x − 2|.

Suppose that |x − 2| < 1. Then 1 < x < 3 and so 3 < x + 2 < 5, so that |x + 2| < 5. It follows that,
then, |f (x) − 2| < 5|x − 2|. If, also, |x − 2| < ǫ/5, then we will have |f (x) − 2| < ǫ. So if both
|x − 2| < 1 and |x − 2| < ǫ/5, we have |f (x) − 2| < ǫ. Therefore, if |x − 2| < δ = min{1, ǫ/5}, we
have |f (x) − 2| < ǫ.

Learning activity 3.3 The function g(x) = f (x) − K will be such that 0 is between g(a) and g(b) if
K is between f (a) and f (b). Also, continuity of f on the interval [a, b] implies continuity of g there
too. So the version of the theorem we’ve proved shows there is some c ∈ (a, b) with g(c) = 0. But
g(c) = f (c) − K, so this means f (c) = K, as required.

3.9 Sketch answers to or comments on selected questions

Question 3.1 Let K > 0. We want to find δ so that if 0 < |x − 2| < δ then f (x) > K. If
|x − 2| < δ then
5 5
f (x) = > 2,
(x − 2)2 δ
p
so it suffices to have 5/δ 2 ≥ K, so we may take δ = 5/K.

Question 3.2
x3 + 5x + 7 13 + 5(1) + 7 13
lim 4 2
= 4 = .
x→1 x + 6x + 8 1 + 6(1)2 + 8 15

Question 3.3 If 1 < x < 2 then x − ⌊x⌋ = x − 1, so

lim (x − ⌊x⌋) = lim (x − 1) = 0.


x→1+ x→1+

If 0 < x < 1 then x − ⌊x⌋ = x − 0 = x, so

lim (x − ⌊x⌋) = lim (x) = 1.


x→1− x→1−


Question√3.4 For x < 1, |f (x) − 0| = (x − 1)2 . This is < ǫ if 0 < |x − 1| < δ where δ = ǫ. So, if
x ∈ (1 − ǫ, 1) then |f (x) − 0| < ǫ. Thus, the left-limit at 1 is 0. For x > 1,
|f (x) − 5| = |3x + 2 − 5| = 3|x − 1|. This is < ǫ if |x − 1| < δ = ǫ/3. So if x ∈ (1, 1 + ǫ/3) then
|f (x) − 5| < ǫ, and hence the right-limit is 5. The function is not continuous at 1, nor is it continuous
on the left or on the right. (The function value at 1 is not equal to either the left or the right limit.)

Question 3.5 We have to show that for each x ∈ R, the function is not continuous at x. There are
two cases: x rational and x irrational. Suppose first that x is rational. Then, for any δ > 0 there are
irrational numbers y such that |y − x| < δ, and we would have |f (y) − f (x)| = |1 − 0| = 1. If f were
continuous at the point x then (taking ǫ = 1) there should be δ such that |y − x| < δ implies
|f (y) − f (x)| < 1. But our observations have just shown this is impossible. The case of x irrational is
analogous.

36
Chapter 3. Limits of functions and continuity

Question 3.6 Taking ǫ = f (c)/2 > 0 in the formal definition of continuity at c, there is some δ such
that if x ∈ (c − δ, c + δ) then |f (x) − f (c)| < f (c)/2. Thus, for x ∈ (c − δ, c + δ),
f (x) > f (c) − f (c)/2 = f (c)/2 > 0, so f is positive on that interval.

Question 3.7 Let y ∈ (0, ∞). We show that there is some c ∈ R such that f (c) = y. This shows
that the range is the whole of (0, ∞). (The fact that it is no larger follows from the given fact that f
is positive.) Now, f (R) = inf{f (x) | x ∈ R} = 0, so, since y > 0, there must be some y1 ∈ f (R)
with y1 < y. This means there is some x1 ∈ R such that y1 = f (x1 ) < y. Similarly, because f is
unbounded, which means f (R) is unbounded, there must be some y2 ∈ f (R) with y2 > y and there
will be some x2 ∈ R such that y2 = f (x2 ) > y. Then y lies between f (x1 ) and f (x2 ) and, since f is
continuous, the Intermediate Value Theorem shows that there is some c between x1 and x2 with
f (c) = y.

Question 3.8 Note that

f (π) = T (π) − T (2π) = T (π) − T (0) = −(T (0) − T (π)) = f (0),

so either f (0) = f (π) = 0, in which case the diametrically opposite points at longitude 0 and π have
the same temperature, or f (0) and f (π) lie either side of 0, in which case the Intermediate Value
Theorem establishes that there is c such that f (c) = 0—that is, T (c) = T (c + π).

Question 3.9 The function h satisfies h(a) = f (a) − a ≥ 0, since f (a) ∈ [a, b] implies f (a) ≥ a.
Similarly, h(b) ≤ 0. So a < b and h(a) ≥ 0 ≥ h(b). Since f is continuous, as is the function x 7→ x, it
follows that h is continuous. By the Intermediate Value Theorem, ∃c ∈ [a, b] with h(c) = 0; that is,
f (c) = c. Consider g now. The function k given by k(x) = g(g(x)) maps [a, b] into [a, b], so, by the
first part, there is some c ∈ [a, b] such that k(c) = c; that is, g(g(c)) = c. Let p = c and q = g(c).
Then g(p) = q (obviously), and g(q) = g(g(c)) = c = p. For the last part, consider the function
h(x) = g(x) − x. If p = q then g(c) = c and we are done. Otherwise,

h(q) = g(q) − q = p − g(p) = −h(p),

so 0 lies between h(p) and h(q). By the Intermediate Value Theorem (and the continuity of h) there
is some c between p and q such that h(c) = 0; that is, g(c) = c.

37

You might also like