Analysis II: MA137

Keith Ball
Chapter 1. Continuous Functions

Continuity

Given two sets A and B a function f from A to B assigns an element

of B to each element of A. Thus, every function comes “equipped,”
with two sets: A, the set of points where the function is defined, and
B, a set to which the values of the function belong.

We can draw attention to these sets by writing

f : A → B.
 f : A → B.

The set A is called the domain of f ; B is called its codomain. For

each element x of the domain we write f (x) for the place to which x
is sent: the image of x under f .
An interval of the real line is a subset of R with the property that
if x < y < z and x and z both belong to the subset then so does y.
So an interval contains all the points between its ends but the ends
themselves may or may not be included.

Examples are

{x : a ≤ x ≤ b} = [a, b] a closed interval

{x : a < x < b} = (a, b) an open interval
{x : a ≤ x < b} = [a, b) a half-open interval
{x : a ≤ x} = [a, ∞) a half-infinite interval
For this course we will be considering real valued functions whose
domains are intervals of the real line. We want to say what it means
for such a function to be continuous.

At each point of the domain there are essentially three types of problem
that can occur. In each of the cases below, the function is discontin-
uous at 0 (but continuous elsewhere).
 0 0

The first two are fairly straightforward. In the first case the function
has a value at zero that doesn’t “match” the rest of the function. In
the second case, the function jumps in value as we pass through 0.
 0

In the third situation f (x) oscillates infinitely often over a wide range
as x approaches 0.

We shall define what it means for a function to be continuous at a

point in such a way as to rule out these problems. Then we say that a
function is continuous on an interval if it is continuous at each point
of the interval.
Definition (Continuity). A function f : I → R defined on an interval
I containing the number c is said to be continuous at c if for every
ε > 0, there is a δ > 0 so that if x ∈ I and |x − c| < δ then

|f (x) − f (c)| < ε.

The function is continuous on I if it is continuous at each point of I.

 f(c)+ϵ
f(c)
f(c)-ϵ

c-δ c c+δ

The picture shows the situation. The number ε specifies a band around
the value f (c). You want to be able to guarantee that f (x) falls into
that band as long as x is close enough to c.
Example (Continuity of x 7→ x). The function x 7→ x is continuous at
every point of the real line.

Proof If f (x) = x then we can always take δ = ε since if |x − c| < ε then

|f (x) − f (c)| = |x − c| < ε.

As an exercise, you can check that constant functions are continuous.

Example (A discontinuity). The function f given by
(
1 if x > 0
f (x) =
0 if x ≤ 0
is discontinuous at 0.

Proof We need to find a band around f (0) = 0 which we cannot

guarantee to land in, merely by starting near 0. A band of width
ε = 1/2 will do because if x is positive, however close to zero, f (x) = 1
and this is not within 1/2 of f (0).
We will do one more example which will also follow from our later
arguments but will help to fix the ideas: the function x 7→ x2.

The aim will be to show that if x − c is small then so is x2 − c2. Now

x2 − c2 = (x − c)(x + c) so if |x − c| < δ then |x2 − c2| < δ|x + c| which
looks good because the factor δ is small. But the second factor |x + c|
might be large (if c = 1, 000, 000 say).

However, if x is close to c then x + c is close to 2c which may be large

but is a fixed number. So we shall cook up δ to compensate: roughly
we want δ = ε/(2|c|). However we need to be a bit careful because
x + c isn’t actually equal to 2c.
Example (Continuity of x 7→ x2). The function x 7→ x2 is continuous
at every point of the real line.

Proof Let f (x) = x2, let c ∈ R and ε > 0. We want to guarantee that

|x2 − c2| < ε

by choosing x close enough to c. We know that |x2 − c2| = |x − c|.|x + c|.

If we choose |x − c| < 1 then |x| cannot be larger than |c| + 1 and so

|x + c| cannot be larger than 2|c| + 1.
Then if we choose
ε
|x − c| <
2|c| + 1
we will get
ε
|x2 − c2| = |x − c|.|x + c| < (2|c| + 1) = ε
2|c| + 1
which is what we wanted.

We end up needing two conditions on |x − c|: namely |x − c| < 1

ε . But that is not a problem. If we choose δ to
and |x − c| < 2|c|+1
ε
be the smaller of 1 and 2|c|+1 the two conditions will be satisfied
simultaneously as long as |x − c| < δ.
We wish to check the continuity of functions such as polynomials, more
complicated than x 7→ x2. To do this we want a machine which allows
us to build continuous functions from simpler ones.

We want to know that when you add or multiply continuous functions

the result is still continuous and also when you compose two continuous
functions.

There are a number of ways to do this but in view of what you already
know about limits there is a particularly simple approach which depends
upon rewriting the continuity property in terms of limits of sequences.
Theorem (Sequential continuity). Let f : I → R be defined on the
interval I and suppose that c ∈ I. Then f is continuous at c if and
only if for every sequence (xn) of points in I which converges to c,

f (xn) → f (c) as n → ∞.

Proof Suppose first that f is continuous at c and xn → c. Then, given

ε > 0 we can find δ > 0 so that if |x − c| < δ then |f (x) − f (c)| < ε.

Now choose N so that if n > N , |xn − c| < δ.

Then if n > N we have |f (xn) − f (c)| < ε. So f (xn) → f (c).

On the other hand suppose f is not continuous at c and choose ε > 0
with the property that whatever δ we pick there is a point x within δ
of c for which

|f (x) − f (c)| ≥ ε.

Now build a sequence as follows. For each n choose a point xn with

|xn − c| < 1/n but |f (xn) − f (c)| ≥ ε.

Then xn → c as n → ∞ but f (xn) does not converge to f (c).

The sequential continuity theorem immediately gives us the algebra of
continuous functions.

Theorem (Algebra of continuous functions). Let f, g : I → R be

defined on the interval I and continuous at c ∈ I. Then

1. f + g is continuous at c

2. f.g is continuous at c

3. if g(c) 6= 0 then f /g is continuous at c.

Proof The proofs of all 3 are essentially the same. We shall do the
first.

We wish to show that if xn → c then (f + g)(xn) = f (xn) + g(xn) →

f (c) + g(c) as n → ∞.

But we know that f (xn) → f (c) and g(xn) → g(c) so we can apply
the properties of limits of sequences to conclude that f (xn) + g(xn) →
f (c) + g(c).

From this we can immediately conclude that polynomials are contin-

uous and that rational functions are continuous except where the de-
nominator is zero.
Corollary (Continuity of polynomials and rational functions). If p
is a polynomial then p is continuous at every point of R. If r = p/q
is a ratio of two polynomials then it is continuous at every point of R
where q 6= 0.

Proof Exercise.
To complete the continuity machine we need to know that we can
compose continuous functions. The statement looks a bit complicated
but the idea is as simple as for the algebraic properties.

Theorem (Composition of continuous functions). Let f : I → R be

defined on the interval I, g : J → I be defined on the interval J. If g is
continuous at c and f is continuous at g(c) then the composition f ◦ g
is continuous at c.

Proof Let (xn) be a sequence in J converging to c. Then g(xn) → g(c)

in I and hence f (g(xn)) → f (g(c)).
The Intermediate Value Theorem

This section is devoted to a principle which has many uses and which
you may have met at school. Roughly it says that if a continuous
function crosses from below zero to above zero then there must be a
point where it is equal to zero: it cannot jump across zero.

Our intuition about continuous functions is that we can draw them

without taking pen from paper and therefore it should be “obvious”
that a continuous function can’t skip a value.

However the proof is not quite so simple because we have defined

continuity at each point individually. We therefore have to find the
point where the function is supposed to take the correct value and
then use continuity at that point.
We have to find the point where the function is supposed to take the
correct value and then use continuity at that point.

When we are trying to demonstrate the existence of a particular real

number we usually have to invoke the completeness principle. The
completeness principle is our way to finger a real number.

Theorem (Intermediate Value Theorem). Let f : [a, b] → R be con-

tinuous and suppose that u lies between f (a) and f (b). Then there is
a point c between a and b where f (c) = u.
Theorem (Intermediate Value Theorem). Let f : [a, b] → R be con-
tinuous and suppose that u lies between f (a) and f (b). Then there is
a point c between a and b where f (c) = u.

Proof Assume that f (a) < u < f (b) and let A be the set

{x ∈ [a, b] : f (x) ≤ u}.

This set is non-empty since it contains a and is bounded above by b.

Let s be its least upper bound.

The aim is to show that f (s) = u.

Assume that f (a) < u < f (b) and let A be the set

{x ∈ [a, b] : f (x) ≤ u}.

This set is non-empty since it contains a and is bounded above by b.

Let s be its least upper bound. The aim is to show that f (s) = u.

a s b
Suppose that we have f (s) < u.

u
ϵ
f(s)

s-δ s s+δ
Suppose f (s) < u.

Note that that s 6= b because we know f (b) > u. If we put  = u − f (s)

then for some δ > 0

|f (x) − f (s)| < 

as long as |x − s| < δ.

So in particular if x = s + δ/2 then f (x) < f (s) +  = u.

This means that s + δ/2 ∈ A contradicting the fact that s is an upper

bound for A.
Suppose on the other hand that f (s) > u.

f(s)
ϵ
u

s-δ s s+δ
Suppose f (s) > u.

Note that s 6= a because f (a) < u. If we put  = f (s) − u then for some
δ>0

|f (x) − f (s)| < 

as long as |x − s| < δ.

Therefore if s − δ < x ≤ s, f (x) > f (s) −  = u and hence x is not in A.

This means that s − δ is an upper bound for A contradicting the fact

that s is the least upper bound.
Note that there were two things we had to check about the point s
that we found: that f (s) was not too small and that it was not too
large.

We used the two different properties of the least upper bound respec-
tively for these two checks. This is a common feature of arguments
using least upper bounds.
The IVT has many uses. The most obvious is that it guarantees the
existence of solutions to equations.

You already saw that each positive number has a positive square root
(found as a limit). We can now deduce this very simply from the IVT.

Corollary (The Existence of Square Roots). Every positive real

number has an unique positive square root.
Corollary (The Existence of Square Roots). Every positive real
number has an unique positive square root.

Proof Let R be a positive number and consider the function f : x 7→ x2.

Notice that f (0) < R and that for example

f (R + 1) = R2 + 2R + 1 > R.

So there is a number c between 0 and R + 1 where f (c) = R. Since

c2 = R we know that c is a positive square root of R.

√
To see that R is unique we observe that if 0 < x < y then x2 < y 2
and so two distinct positive numbers cannot have the same square.
Example.

There is a solution of the equation x3 + x − 1 = 0 between 0 and 1.

Because if f (x) = x3 + x − 1 then f (0) = −1 < 0 while f (1) = 1 > 0.

The existence of the square root is one example of a more general

principle that follows from the IVT.

Whenever we have a continuous, strictly increasing function we can

√
find an inverse for it: just as y 7→ y is the inverse of the function
x 7→ x2 on [0, ∞).
An obvious question that we didn’t address in the case of the square
root is whether the inverse function is also continuous.

In order to make this precise we define the range of a function f : A → B

to be the set of values that f takes:

{f (x) : x ∈ A}.

Corollary (Continuous image of an interval). If f : I → R is contin-

uous on the interval I then its range is an interval.

Proof If x and y are in the range then by the IVT so is any point
between x and y. So the range is an interval.
Theorem (Existence of inverses). Let f : [a, b] → R be continuous
and strictly increasing. Then f has an inverse defined on its range and
f −1 is continuous.

The same holds if f is defined on an open interval or on the whole of

R. The proofs are essentially the same.

The picture says it all. If you want to end up within ε of x = g(y) then
make sure you start between f (x − ε) and f (x + ε).
f(x+ϵ)
y
f(x-ϵ)

x-ϵ x x+ϵ
Proof Let f (a) = c and f (b) = d. Since f is increasing all its values lie
between c and d. So the range of f is exactly the interval [c, d].

For each y there is an unique number x with f (x) = y because f is

strictly increasing, so we may define the inverse g = f −1 by setting
g(y) = x in each case.

Clearly g is strictly increasing.

The final aim is to show that g is continuous. Suppose y lies in the

open interval (c, d), y = f (x) and ε > 0. We have f (x−ε) < y < f (x+ε).

Hence there is some δ > 0 with the property that

f (x − ε) < y − δ < y < y + δ < f (x + ε).

Hence there is some δ > 0 with the property that

f (x − ε) < y − δ < y < y + δ < f (x + ε).

For any z between y − δ and y + δ the fact that g is increasing ensures

that

x − ε < g(z) < x + ε

and so |g(z) − g(y)| < ε.

If y is one of the end points then the argument is the same but with
one of f (x − ε) and f (x + ε) replaced by the corresponding end point
of [c, d].
Corollary (Existence of roots). For each positive x and natural num-
ber n there is an unique positive nth root x1/n. The map

x 7→ x1/n

is continuous.

Example (A composition). Let f : R → R be given by

q
f (x) = 1 + x2 .

Then f is continuous on R.
Boundedness and Attainment of Bounds

A continuous function on an open interval such as (0, 1) may take

arbitrarily large values. For example the function x 7→ 1/x approaches
∞ as x approaches 0.

This can’t happen if the function is continuous on the closed interval

[0, 1] since the function is supposed to approach f (0) as x approaches
0.

It turns out that indeed a continuous function on a closed interval must

be bounded.

Theorem (Boundedness of continuous functions). Let f : [a, b] → R

be continuous. Then f is bounded.
The proof uses a method that used to be called rather appealingly
“condensation of singularities”.

The idea is that if the function is unbounded we can find a sequence

of points where it gets bigger and bigger.

Using Bolzano-Weierstrass we find a subsequence of these points which

converges.

So a sequence of bad points converges to a point which we can con-

clude is so bad that it can’t exist.
Theorem (Boundedness of continuous functions). Let f : [a, b] → R
be continuous. Then f is bounded.

Proof Suppose f is unbounded. For each n choose xn ∈ [a, b] where

|f (xn)| ≥ n.

Now choose a subsequence xnk which converges to x say. Since the

interval [a, b] is closed we must have x ∈ [a, b].

Then we know that f (xnk ) → f (x) but this can’t happen since the
values f (xnk ) are becoming arbitrarily large.
In your schoolwork you were often asked to find the maximum value of
a function on some interval but probably never addressed the question
of whether a nice function (a continuous one say) really must have a
maximum.

Obviously the previous example of 1/x shows that on an open interval

a function need not have a maximum but this is also true even if the
function is bounded.

The function x 7→ x has no maximum on (0, 1). Its least upper bound
is 1 but there is no point in the interval where the function is equal to
1. Again, this situation can’t happen on closed intervals.
Theorem (Attainment of bounds). Let f : [a, b] → R be continuous.
Then f has a maximum and a minimum attained in the interval.

Proof Let M be the least upper bound of the set

{f (x) : x ∈ [a, b]}.

If there is no point in the interval where f (c) = M then the function

g given by g(x) = M − f (x) is strictly positive and continuous on the
interval.

Hence by the algebra of continuous functions its reciprocal is continu-

ous and therefore bounded. Let’s say
1
≤R
M − f (x)
for all x ∈ [a, b].
Then 1/R ≤ M −f (x) and hence f (x) ≤ M −1/R for all x in the interval.

This shows that M − 1/R is an upper bound for the set of values and
contradicts the fact that M is the least upper bound.

A similar argument works for the minimum.

This theorem will form the basis of our proof of the so-called Mean
Value Theorem which plays a key role in many applications of the
derivative.
Chapter 2. Power Series I

A power series is a series of the form

∞
an xn
X

n=0
or more generally
∞
an (x − x0)n
X

n=0

in which we centre the series at a number x0 rather than at 0. We

will develop the theory just for the first type: everything transfers
immediately to the more general case.
We think of a power series as a kind of infinite polynomial in which the
(an) are coefficients. The series may or may not converge depending
upon the value of x.

Where it does converge it defines a function of x and we shall see that

the function will automatically be continuous and differentiable.

Many of the most important functions in mathematics can be written

as power series: in particular the exponential function.
Basic properties of power series

P∞
We want to know the values of x for which a power series 0 an xn
converges.

Of course, the answer will depend upon what the an are but there is a
fundamental feature of the answer that is common to all power series.

The set of values where a power series converges will be an interval

(possibly infinite) of the real line.

antn converges and we choose x to be

P
The idea is quite simple: if
smaller than t in size then the terms anxn will be much smaller than
the terms antn so we should expect an xn to converge as well.
P
We would like to prove this just by quoting the Comparison Test for
an tn might not
P
series but that doesn’t quite work because the series
converge absolutely.

However we can get around this problem because anxn is not just
smaller than antn but is very much smaller when n is large.

P∞
Theorem (Radius of convergence I). Let 0 anxn be a power series
an tn convergent. Then
P
with
∞
anxn
X

0
converges absolutely for all x with |x| < |t|.
 an tn converges we know that an tn → 0 as n → ∞ and
P
Proof Since
so the sequence is bounded. There is some M for which |antn| < M
for all n.

Now
N N  n
x
|anxn| = n
X X
|ant |  
0 0 t
N  n
X x
≤ M  
0 t
 
∞  n
X x
≤ M  
0 t
 

1
= M
1 − |x|/|t|

P∞ n
Hence 0 |an x | < ∞ and the series converges absolutely.
 P∞ n
Theorem (Radius of convergence II). Let 0 an x be a power series.
One of the following holds.

• The series converges only if x = 0.

• The series converges for all real numbers x.

• There is a positive number R with the property that the series

converges if |x| < R and diverges if |x| > R.

In the third case the number R is called the radius of convergence.

In the first case we say that the radius of convergence is 0 and in the
second that it is ∞.
Proof Obviously the series does converge if x = 0. If the first option
does not hold than it converges for some other values of x.

The set of x for which it converges might be unbounded: it might

contain arbitrarily large numbers. In that case the power series will
converge absolutely for all real x and defines a function on R.

Otherwise the set might contain some numbers other than 0 but be
bounded. In that case we let

an tn converges}.
X
R = sup{|t| :

Then by our previous theorem the series converges absolutely whenever

|x| < R. It does not converge if |x| > R by definition of R.
As an immediate corollary we get an observation that will be useful
later.

P∞ n
Corollary (Absolute series). Let 0 an x be a powers series with ra-
P∞
dius of convergence R. Then 0 |an|xn also has radius of convergence
R.

P∞ n
Example (The geometric series I). The series 0 x has radius of
convergence R = 1.

Proof We know that the series converges if and only if x ∈ (−1, 1).
 P∞ n n
Example (The geometric series II). If p is real the series 0 p x
has radius of convergence R = 1/|p|.

Proof We know that the series converges if and only if px ∈ (−1, 1)

which is the same as saying that x ∈ (−1/|p|, 1/|p|).

P∞ xn
Example (The log series). The series 0 n has radius of conver-
gence R = 1.

Proof From homework we know that the series converges if and only
if x ∈ [−1, 1).
 P∞ xn
Example (The log series). The series 0 n has radius of conver-
gence R = 1.

Notice that although we have put in the extra factor 1/n in front of
xn we have not changed the radius of convergence.

The point is that if we increase x beyond 1, its powers xn increase

exponentially fast and this swamps the effect of the factor 1/n.

However, the extra factor of 1/n does make the series converge at
x = −1 but not absolutely.
 P∞
Example. The series 0 nxn has radius of convergence R = 1.

Proof From homework we know that the series converges if and only
if x ∈ (−1, 1).

For most interesting power series we can find the radius of convergence
using the ratio test.
As a taste let us look at another example:

Example. The series 1 + 0x + x2 + 0x3 + x4 + · · · has radius of con-

vergence R = 1.

Proof The series is

1 + x2 + x4 + x6 + · · ·

which is a geometric series with ratio x2. So it converges if and only

if x2 ∈ (−1, 1).
Example. The series 1 + x2 + x4 + · · · has radius of convergence R = 1.

Another way to to do it which is a much more flexible method is as

follows.

If x > 1 then the terms do not tend to zero so the series diverges.

If 0 < x < 1 then

1 + x2 + x4 + · · · < 1 + x + x2 + x3 + x4 + · · ·

which we already know converges.

This means that the series converges if 0 < x < 1 so the radius of
convergence is 1.
We remarked that it is possible to consider power series centred at
points other then 0.

For example
1 1 1 1
= =
1−x 2 − (x + 1) 2 1 − (x + 1)/2

x+1 2 (x + 1)2
!
1 x+1 1 x+1
 
= 1+ + + ··· = + + + ··· .
2 2 2 2 4 8

 
 x+1 
This converges if  2  < 1: in other words if x differs from −1 by at
most 2.

The series is centred at −1 and has radius of convergence 2.

As was remarked several times, many of the most important functions
in mathematics are given by power series so we want to know that such
functions have nice properties: that they are continuous for example.

P∞ n with radius of
Theorem (Continuity of power series). Let 0 an x
convergence R. Then the function
∞
anxn
X
x 7→
0
is continuous on the interval (−R, R).

We already saw some examples in which the power series converges

at one or both ends of the interval [−R, R]. If so it makes sense to
ask whether the function is continuous on this larger set. Perhaps
surprisingly this is a bit more subtle than the statement above.
There are some quite slick proofs of continuity of power series. We will
take a very down to earth approach. Inside the radius of convergence
a power series can be approximated by a polynomial: and polynomials
are continuous.

Proof Suppose −R < x < R. We want to show that the function is

continuous at x.

Choose a number T with |x| < T < R. |an| T n

P
Then the series
converges, so for each ε > 0 there is some number N for which
∞
|an| T n < ε/3.
X

N +1
Now if |y − x| < T − |x| we will have |y| < T as well as |x| < T .

Hence
∞ ∞
|an| |x|n < ε/3 |an| |y|n < ε/3.
X X
and
N +1 N +1

PN n is a polynomial in y and polynomials are

The partial sum 0 an y
continuous so there is some δ0 > 0 with the property that if |y − x| < δ0
 
N N 
X n n
X 

 an y − an x  < ε/3.
 0 0 
Therefore if we choose δ to be the smaller of δ0 and T − |x| then if
|y − x| < δ we get

       
∞ ∞   ∞   N N   ∞ 
X n− n ≤  n +  n− n +  n 
X   X   X X   X

 an y an x   an y   an y an x   an x 
 0 0   N +1   0 0   N +1 

 
∞ N N  ∞
|an| |y|n +  an y n − an xn + |an| |x|n < ε.
X  X X  X
≤
N +1  0 0  N +1