MA137: Analysis II

Richard Lissaman

Please e-mail [email protected] with details of any errors you might

notice on the notes here.

1 Functions
We shall study functions f :X →R where X ⊆R i.e real-valued functions of a
real variable.

1.1 Denitions
The domain, X, of f will usually be an interval, i.e. a set of the type

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

The diagram below shows notation for some of the intervals above.

However X could be any subset of R. Some other examples are the set of
rational numbers Q, the set of non-zero real numbers R\{0} and the set {1, 2, 3}.

1
1.2 Denitions and examples
1.2.1 Denition
Let X ⊆ R. A real valued function f with domain X is a rule that associates to
each x ∈ X a real number f (x). We write f : X → R.

A function is often given by a mathematical formula but it doesn't have to be.

The set {f (x) | x ∈ X} is called the image of f.

1.2.2 Denition
Let X⊆R and let f :X→R be a function. The graph of f is

G(f ) = {(x, f (x)) ∈ X × R | x ∈ X}.

The graph of f :X→R given by f (x) = x2 is shown below.

Sometimes it is easy to draw the graph of a given function. Sometimes it is

very dicult and sometimes it is impossible.

1.2.3 Examples
i. Constant functions
The simplest function is a constant function.

Let c ∈ R. Then f : R → R dened by f (x) = c ∀x ∈ R is the constant function

with value c.

Its graph is a horizontal straight line.

2
 ii. Linear functions
Let a∈R and dene f : R → R by f (x) = bx. The graph of f is a straight
line through the origin. If b = 0 then f is the contant function with value 0.

iii. Polynomials
Let a0 , a1 , ..., an be real numbers. Dene f :R→R by

f (x) = an xn + an−1 xn−1 + ... + a1 x + a0 .

f (x) = 0 has at most n distinct solutions. We will see later that these play
an important role in many functions in this course. Here are some examples
of graphs of polynomials.

iv. Sine
To begin with we will dene f : R → R, f (x) = sin(x) as follows. Consider
the point the point A on the unit circle centred at the origin, O, where the
angle AOB is θ (θ is measured in radians).

3
 The y -coordinate of A is sin(θ). This is then extended by repetition to
negative values and values greater than or equal to 2π according to the
property sin(x + 2π) = sin(x). This gives a graph of f (x) = sin(x) as
follows.

v. Cosine
f : R → R, f (x) = cos(x) is dened similarly. Consider the point the point
A on the unit circle centred at the origin, O, where the angle AOB is θ (θ is
measured in radians).

4
 The x-coordinate of A is cos(θ). This is then extended by repetition to
negative values and values greater than or equal to 2π according to the
property cos(x + 2π) = cos(x).
This gives a graph of f (x) = cos(x) as follows.

 
1
vi. sin
x
The expression above is not dened at 0 and so for a function f :R→R we
need to dene it there separately, e.g.
 
1
sin if x 6= 0

Let f :X→R be dened by f (x) = x .
0 if x=0


5
  
1
vii. x sin
x
  
1
x sin if x 6= 0

Let f :R→R be dened by f (x) = x .
0 if x=0


viii. The characteristic function of Q


1 if x∈Q
Let f :R→R be dened by f (x) = .
0 if x 6∈ Q
It's not possible to draw a graph of this function. Even within the width of
a pencil or a pixel the graph takes the value of both 0 and 1.

ix. The Thomae function

6
 Let f : (0, 1) → R be dened by

 1 p
if x ∈ Q x = p, q ∈ Z, q > 0, hcf(p, q) = 1.
f (x) = q q .
 0 if x 6∈ Q

Again, it's not possible to draw a graph of this function.

What do we mean when we say that two function are equal?

1.2.4 Denition
Let X⊆R and let f : X → R be a function, and let W ⊆ R and let g : W → R
be a function. To say f = g means X = W and f (x) = g(x) ∀x ∈ X .

So both the domain and the `rule' must be the same. For example f : [0, 7] → R
given by f (x) = sin x is a dierent function from g:R→R given by g(x) = sin x
because they have dierent domains.

The function f : R → R given by f (x) = x2 is equal to the function g : R → R

given by f (x) = |x|2 , because they have the same domain and x2 = |x|2 ∀x ∈ R.

1.3 New functions from old ones

Operations that can be performed on real numbers can often be performed on
real-valued functions; the arithmetic of functions is dened as follows.

1.3.1 Denition
Let X ⊆ R and let f : X → R and g : X → R be functions. Then f + g : X → R,
f − g : X → R, f g : X → R are dened by

i. (f + g)(x) = f (x) + g(x) ∀ ∈ X

ii. (f − g)(x) = f (x) − g(x) ∀ ∈ X

iii. (f g)(x) = f (x)g(x) ∀ ∈ X ,

 
f f f (x)
iv. If g(x) 6= 0 ∀x ∈ X then :X→R is dened by (x) = ∀ ∈ X.
g g g(x)

It is usually hard to draw graphs other than in very easy cases like when g is
a constant function.

7
1.3.2 Examples
i. Iff : R → R is dened by f (x) = sin(x) and g : R → R is dened by
sin(10x)
g(x) = then the graphs of f and g are shown in the diagram below.
10

sin(10x)
Then f + g : R → R is given by (f + g)(x) = sin(x) + and its graph
10
is as below.

ii. Iff : R → R is dened by f (x) = sin(x) and g : R → R is dened by

g(x) = sin(10x) then the graphs of f and g are shown in the diagram below.

8
 Then fg : R → R is given by (f g)(x) = sin(x) sin(10x) and its graph is as
below.

1.3.3 Remarks
Note that f − g = f + (−1)g . Here we are interpreting −1 to be the constant
function h(x) = −1, ∀x ∈ X .
Note that f + g = g + f , gf = f g and f (g + h) = f g + f h.

These denitions can be extended to any nite number of functions. If fi : X → R,

1≤i≤n are given we can dene

(f1 + f2 + ... + fn ) : X → R by (f1 + f2 + ... + fn )(x) = f1 (x) + f2 (x) + ... + fn (x)

and

(f1 f2 ...fn ) : X → R by (f1 f2 ...fn )(x) = f1 (x)f2 (x)...fn (x).

9
We can also dene |f |.

1.3.4 Denition
If X ⊆ R and f : X → R is a function then the function |f | : X → R is dened
by |f |(x) = |f (x)| ∀x ∈ X .

The graph of |f | is obtained from the graph of f by reecting the part of the
graph of f below the x-axis to above the x-axis.

For example, here is the graph of f :X→R given by f (x) = | sin(x)|

1.3.5 Denition
Let W ⊆ X ⊆ R and let f : X → R be a function. The restriction of f to W,
f |W : W → R is dened by f |W (x) = f (x) ∀w ∈ W .
Another way to consider complicated functions in terms of simpler ones is
through composition.

1.3.6 Denition
Let X ⊆ R and Y ⊆ R and let g : X → R and f : Y → R be functions. If
{g(x) | x ∈ X} ⊆ Y we can dene (f ◦ g)(x) = f (g(x)) ∀x ∈ X .

1.3.7 Warning
fg means product. f ◦g means composition.

Always gf = f g . However it may be that f ◦g is dened and g◦f is not dened.

Even when f ◦ g and g ◦ f are both dened they can be dierent functions.

10
2 Continuity
We would like to prove that large classes of functions have certain natural and
useful properties.

Consider the temperature along a metal bar of length b, f : [0, b] :→ R. We ex-

pect that if x is near to c then the temperature at x is near to the temperature at c.

We don't expect the graph of f at c to have a jump in it. This is the case
for many functions occuring theoretically in pure mathematics and in `real world'
applications.

2.1 Denition and examples

We shall consider functions f : X → R, where X ⊆ R, and consider what it means
for f to be continuous at a point c ∈ X .

2.1.1 Denition
Let X ⊆ R and let f : X → R be a function. Let c ∈ X . We say f is continuous at c
if ∀ > 0, ∃δ > 0 such that whenever x ∈ X and |x − c| < δ then |f (x) − f (c)| < .

We often write this as

∀ > 0, ∃δ > 0 s.t. x ∈ X and x ∈ (c − δ, c + δ) =⇒ f (x) ∈ (f (c) − , f (c) + )

or

∀ > 0, ∃δ > 0 s.t. x ∈ X ∩ (c − δ, c + δ) =⇒ f (x) ∈ (f (c) − , f (c) + )

2.1.2 Remark
To check that a function, f, is continuous at a point c in its domain we have to
imagine that >0 has been given. The task then is to nd a corresponding δ>0
so that the property above holds (usually this will have to be speciced in terms
of  because we have to show that we can do it for any ).

Given an >0

11
 you need to show that there is a δ>0 such that if x is within a distance δ of
c then its image. f (x), is within a distance  of f (c).

To show that a function, f, is not continuous at a point c, we have to show

that the condition fails for some  > 0. In other words that there is an >0 such
that for all δ>0 the property above does not hold. More formally:

2.1.3 Remark
f : X → R is not continuous at c ∈ X if ∃ > 0 such that ∀δ > 0, ∃x ∈ X with
|x − c| < δ and |f (x) − f (c)| ≥ .

2.1.4 Examples
i. A constant function is continuous at every point in its domain. Let f :R→R
be given by f (x) = d ∀x ∈ R. Then f (x) − f (c) = 0 and so we can choose δ

12
 to be anything we like.

Formally :
Let  > 0 be given. Let δ = 1. Then |x − c| < δ =⇒ |f (x) − f (c)| =
|d − d| = 0 < .

ii. Let a ∈ R. Then f : R → R given by f (x) = ax is continuous at every

c ∈ R.
If a=0 then f (x) = 0 and so f is continuous at every c by i.

Let a 6= 0. Let c ∈ R.
We have to nd δ>0 so that |x − c| < δ =⇒ |f (x) − f (c)| < .
Note that |f (x) − f (c)| = |ax − ac| = |a||x − c|.

Let δ= .
|a|

Formally :
 
Let  > 0 be given. Let δ = . Then |x − c| < δ = =⇒ |a||x − c| <
|a| |a|

|a| , i.e. |f (x) − f (c)| = |ax − ac| < .
|a|

0 if x < 0
iii. Let f : R → R be dened by f (x) = .
1 if x ≥ 0
Then f is not continuous at 0 because if x < 0 then f (x) = 0 and so
|f (x) − f (0)| = 1.

Formally :
1 δ δ
Let  = . Let δ > 0. Let x=− . Then |x − 0| = <δ and
2 2 2
1
|f (x) − f (0)| = 1 > = .
2

 0   if x=0
iv. Let f :R→R be dened by f (x) = 1 .
 sin if x 6= 0
x
 
1
Then f is not continuous at 0. This is because f π .
+ 2nπ
2

13
 Formally :
1 1
Let  = . Let δ > 0. Choose n so that π < δ (this is possible
2 + 2nπ
2
1 1
because the sequence π tends to zero). Let x = π . Then
+ 2nπ + 2nπ
2 2
|x − 0| =< δ and

  
 
 1  = 1 > 1 = .

|f (x) − f (0)| = f  π
 + 2nπ 
 2
2

v. sin is continuous at 0.
We want to nd δ>0 so that |x − 0| < δ =⇒ |f (x) − f (0)| = | sin(x) −
sin(0)| < .
In other words if |x| < δ then | sin(x)| < .
π
However | sin x| ≤ |x| when |x| < (in fact this is true for all x ∈ R, the
2
π
condition |x| < is just to focus attention on the particular case in the
2
diagram below).

π
δ = min (, ).
so we take
2
Formally :

14
  π π
Let  > 0 be given. δ = min ,
Let . If |x − 0| < δ , |x| < and so
2 2
| sin(x)| < |x| (see diagram above). Also |x| < . Therefore | sin(x)−sin(0)| =
| sin(x)| ≤ |x| < .

We often refer to the following result.

2.1.5 Theorem
Let c∈X and let f :X→R be continuous at c.

If f (c) > 0 then ∃δ > 0 such that if x∈X and |x − c| < δ then f (x) > 0.

Similarly, if f (c) < 0 then ∃δ > 0 such that if x ∈ X and |x − c| < δ then f (x) < 0.

Proof

Suppose f (c) > 0. In the denition of continuity at c, take  = f (c).

Then ∃δ > 0 such that x∈X and |x − c| < δ implies |f (x) − f (c)| < f (c).

But |f (x) − f (c)| < f (c) means −f (c) < f (x) − f (c) < f (c) and so 0 < f (x).

Hence x∈X and |x − c| < δ imply 0 < f (x).

Proof is similar in the case f (c) < 0.

2.1.6 Theorem
Let c∈X⊆R and let f :X→R be continuous at c. Then

i. |f | : X → R is continuous at c.

ii. if c∈W ⊆X then f |W : W → R is continuous at c.

Proof

i. Let  > 0. We have to nd δ >0 such that x∈X and |x − c| < δ =⇒
||f |(x) − |f |(c)| < .

But ||f |(x) − |f |(c)| = ||f (x)| − |f (c)|| ≤ |f (x) − f (c)|.

Since f is continuous at c ∃δ > 0 such that x ∈ X and |x − c| < δ =⇒

15
 |f (x) − f (c)| < .

Therefore x∈X and |x − c| < δ =⇒ ||f |(x) − |f |x)| < .

So |f | is continuous at c.

ii. Let  > 0. ∃δ > 0 such that x∈X and |x − c| < δ =⇒ |f (x) − f (c)| < .

Hence x∈W and |x−c| < δ =⇒ x ∈ X and |x−c| < δ =⇒ |f (x)−f (x)| <
.

Therefore f |W : W → R is continuous at c.

So, for example h(x) : [0, 2π] → R given by h(x) = | sin(x)| is continuous at all
c ∈ [0, 2π].

2.2 Continuity and natural operations on functions

We have seen the very easy proof that f : X → R given by f (x) = x is continuous
at every c∈R (take δ = ).

We shall now see how to use this to deduce that any polynomial is continuous
at every c∈R (it would be hard to nd δ for a given  and c in this case).

We do this by considering how continuity ts in with the operations of functions

that we discussed in section 1.3.

2.2.1 Theorem
Let f : X → R and g : X → R and let c ∈ X. Suppose both f and g are
continuous at c. Then

i. f +g is continuous at c

ii. fg is continuous at c
f
iii. if g(x) 6= 0 for all x∈X then is continuous at c.
g
Proof

16
i. Let  > 0. We have to nd δ>0 such that if |x − c| < δ then |(f + g)(x) −
(f + g)(c)| < .

What we know is that |f (x) − f (c)| is small and |g(x) − g(c)| is small when
|x − c| is suciently small.

So we try to write |(f + g)(x) − (f + g)(c)| in terms of |f (x) − f (c)|and

|g(x) − g(c)|.

We have (f + g)(x) − (f + g)(c) = f (x) + g(x) − f (c) − g(c) = f (x) −

f (c) + g(x) − g(c)

So |(f + g)(x) − (f + g)(c)| = |f (x) + g(x) − f (c) − g(c)| ≤ |f (x) − f (c)| +

|g(x) − g(c)| by the triangle inequality.

This is the relationship that we need because we can make the RHS small.

By the continuity of f at c, ∃δ1 > 0 such that if |x − c| < δ1 then |f (x) −


f (c)| < .
2
By the continuity of g at c, ∃δ2 > 0 such that if |x−c| < δ2 then |g(x)−g(c)| <

.
2
Let δ = min(δ1 , δ2 ).

Then if x∈X and |x − c| < δ , |x − c| < δ1 and |x − c| < δ2 so

 
|(f +g)(x)−(f +g)(c)| = |f (x)+g(x)−f (c)−g(c)| ≤ |f (x)−f (c)|+|g(x)−g(c)| ≤ + = .
2 2
So f +g is continuous at c.
ii. Let  > 0. We have to nd δ>0 such that if |x − c| < δ then |(f g)(x) −
(f g)(c)| < .

Again we try to write |(f g)(x) − (f g)(c)| in terms of |f (x) − f (c)| and
|g(x) − g(c)|.

(f g)(x) − (f g)(c)
= f (x)g(x) − f (c)g(c)
= f (x)g(x) − f (x)g(c) + f (x)g(c) − f (c)g(c)
= f (x)(g(x) − g(c)) + g(c)(f (x) − f (c))

17
 So |(f g)(x)−(f g)(c)| ≤ |f (x)||g(x)−g(c)|+|g(c)||f (x)−f (c)| by the triangle
inequality.

Since f is continuous at c and 1 > 0, ∃δ1 > 0 such that |x − c| < δ1 =⇒

|f (x) − f (c)| < 1.

So, if |x − c| < δ1 then ||f (x)| − |f (c)|| ≤ |f (x) − f (c)| < 1 and |f (x)| <
|f (c)| + 1.

Since g is continuous at c and > 0, there exists δ2 such that if
2(|f (c)| + 1)

|x − c| < δ2 then |g(x) − g(c)| < .
2(|f (c)| + 1)

Since f is continuous at c and > 0, there exists δ3 such that if
2(|g(c)| + 1)

|x − c| < δ3 then |f (x) − f (c)| < .
2(|g(c)| + 1)
Let δ = min(δ1 , δ2 , δ3 ).

Then if x ∈ X and |x − c| < δ , |x − c| < δ1 , |x − c| < δ2 and |x − c| < δ3


so |f (x)| < |f (c)| + 1, |g(x) − g(c)| < and |f (x) − f (c)| <
2(|f (c)| + 1)

. Therefore
2(|g(c)| + 1)

|(f g)(x) − (f g)(c)| ≤ |f (x)||g(x) − g(c)| + |g(c)||f (x) − f (c)|

   
< (|f (c)| + 1) + |g(c)| < + = .
2(|f (c)| + 1) 2(|g(c)| + 1 2 2
Hence fg is continuous at c.

1
iii. By ii. it suces to show that is continuous at c.
g
Let  > 0. We have to nd δ > 0 such that if |x − c| < δ then

    
 1 1 

 g (x) − (c) < .
g 

18
     
 1 1 
We express
 g (x) − g (c) is terms of |g(x) − g(c)| which we know
 

is small if |x − c| is suciently small:

   
1 1 1 1 g(c) − g(x)
(x) − (c) = − = .
g g g(x) g(c) g(x)g(c)

    
 1 1  |g(c) − g(x)|
So  (x) − (c)= .
 g g  |g(x)||g(c)|
We need to replace the term |g(x)| by something `constant' (i.e. independent
of x, this can be something that depends on , c and g ).

|g(c)|
Since > 0 and g is continuous at c there exists δ1 > 0 such that if
2
|g(c)|
|x − c| < δ1 then |g(x) − g(c)| < .
2
|g(c)|
This means that, if |x − c| < δ1 , ||g(x)| − |g(c)|| < |g(x) − g(c)| < and
2
so
|g(c)| |g(c)|
|g(x)| > |g(c)| − = .
2 2
1 2
Therefore, if |x − c| < δ1 , < .
|g(x)| |g(c)|
|g(c)|2
Since > 0, ∃δ2 > 0 such that if |x − c| < δ2 then |g(x) − g(c)| <
2
|g(c)|2
.
2
Let δ = min(δ1 , δ2 ).

If|x − c| < δ then |x − c| < δ1 and |x − c| < δ2 . This means that both
1 2 |g(c)|2
< and |g(x) − g(c)| <
|g(x)| |g(c)| 2
Therefore |x − c| < δ then

2|g(c)|2
    
 1 1  |g(c) − g(x)| 2|g(c) − g(x)|
 g (x) − g (c) = |g(x)||g(c)| < < = .
 
|g(c)|2 2|g(c)|2

19
2.2.2 Corollary
Let f : X → R and g : X → R both be continuous at c. Let a ∈ R. Then af and
f − g are continuous at c.
Proof Let h : X → R be given by h(x) = a for all x ∈ X . Then h is continuous
at c by 2.1.4(i). Therefore hf = af is continuous at c.

Since g is continuous at c then −g is continuous at c by the above. Therefore

f + (−g) = f − g is continuous at c by 2.2.1(i).

2.2.3 Remarks
If g : X → R is a function, let W = x ∈ X | g(x) 6= 0. Then we can dene
1 1
h:W →R by h(x) = for all x ∈ W. In other words, h= .
g(x) g|W
If g(c) 6= 0 (i.e. c ∈ W) and g is continuous at c, then g|W is continuous at
c, by 2.1.6(ii). Therefore h is continuous at c by 2.2.1(iii).

Note that by 2.1.5 if c ∈ W and g is continuous at c then ∃δ > 0 so that

X ∩ (c − δ, c + δ) ⊆ W .

For example we know g:R→R given by g(x) = x is continuous at all c ∈ R by

1
2.1.4(ii) with a = 1. In this case W = R/0 so h : R/0 → R given by h(x) = is
x
continuous at c for all c 6= 0.

Given function f : X → R and g : X → R, the function max{f, g} : X → R

given by max{f, g}(x) = max{f (x), g(x)} is continuous at all c ∈ X .

f (x) + g(x) + |f (x) − g(x)|

This is because for all x ∈ X , max{f (x), g(x)} =
2
(see assignment 1) and so the result follows by 2.2.1(i), this remark, and 2.1.6(i).

min{f, g} is continuous at all c∈X similarly.

By induction of 2.2.1(i) and 2.2.1(ii) if fi : X → R, 1 ≤ i ≤ n are given and are

all continuous at c ∈ X then (f1 + f2 + ... + fn ) : X → R and (f1 f2 ...fn ) : X → R
by (f1 f2 ...fn )(x) = f1 (x)f2 (x)...fn (x) are both continuous at c.

20
2.2.4 Applications
i. Let a0 , a1 , ..., an be real numbers. The polynomial function p:R→R given
by
p(x) = an xn + an−1 xn−1 + ... + a1 x + a0 .
is continuous at c for all c ∈ R.

To see this let c ∈ R. It would be hard to apply the −δ denition here!

By 2.1.4(i) a constant function is continuous at c and by example 2.1.4(ii)

f (x) = x is continuous at c.

For a xed j≤n consider hj : R → R given by hj (x) = aj xj .

Since hj (x) = aj xx...x 2.2.3 asserts that hj is continuous at c.

Then by 2.2.3 again, p = hn + hn−1 + hn−2 + .... + h1 + h0 is continuous

at c.

ii. Let p:R→R and q:R→R both be polynomial functions. Let W = {x ∈

p(x)
R|q(x) 6= 0}. Then h:W →R given by h(x) = is continuous at all
q(x)
c∈W by 2.2.4(i), 2.2.3 and 2.2.1(ii).

We'll now consider continuity when composing functions.

2.2.5 Theorem
Let f :X→R and g : Y → R be functions. Suppose that {f (x)|x ∈ X} ⊆ Y so
that the function g ◦ f : X → R is dened.

Suppose that f is continuous at c and g is continuous at f (c), then g◦f is con-

tinuous at c.
Proof
Let ε > 0. Since g is continous at f (c) there exists δ1 > 0 such that if y ∈ Y and
|y − f (c)| < δ then |g(y) − g(f (c))| < , i.e. |g(y) − g ◦ f : X → R(c)| < .

Since f is continous at c there exists δ > 0 such that if x ∈ X and |x − c| < δ then
|f (x) − f (c)| < .

So, if x ∈ X and |x − c| < δ then |f (x) − f (c)| < δ1 and so |g ◦ f (x) − g ◦ f (c)| =
|g(f (x)) − g(f (c))| < .

21
Therefore g◦f is continuous at c.

So, checking for continuity for g◦f can often be reduced to checking it for g
and f individually.

By induction we can prove that if f1 , f2 , ...fn are functions so that fn ◦fn−1 ◦...◦f1 :
X → R is dened and f1 is continuous at c ∈ X , f2 is continuous at f1 (c), and
each fj is continous at fj−1 ◦ ... ◦ f1 (c) then fn ◦ fn−1 ◦ ... ◦ f1 is continuous at c.

2.2.6 Remark
We can show that g:R→R given by g(x) = sin(x) is continous at all c∈R as
follows.
For any x, c ∈ R,
   
x−c x+c
sin(x) − sin(c) = 2 sin cos
2 2
.
Therefore
     
 x − c   x + c 
|sin(x) − sin(c)| = 2 sin

 cos
2 2 

and   
 x − c 
|sin(x) − sin(c)| ≤ 2 sin

2 
.
Let c ∈ R. Let  > 0.

Since sin is continuous at 0, there exists δ > 0 such that if |x| < δ then | sin(x)| =

| sin(x) − sin(0)| < .
2     
x − c  x − c 
If |x − c| < δ then
 2  < δ and so |sin(x) − sin(c)| ≤ 2 sin  <
  
2

2 = . Therefore sin is continous at c.
2

2.2.7 Examples
 
1
i. Dene h : R/{0} → R by h(x) = sin .
x
Then h is continuous at all c 6= 0 as follows.

22
 1
Let f : R/{0} → R be dened by f (x) = . By 2.2.3, f is continous
x
at all c 6= 0.

By 2.2.6, g:X→R dened by g(x) = sin(x) is continous at all d ∈ R.

Therefore, by theorem 2.2.5, h = g ◦ f : R \ {0} is continous at all c 6= 0.

2.3 Continuity in terms of sequences

Let f be a function and let c ∈ X . If (xn ) is a sequence of points in X with
lim xn = c then does the sequence (f (xn )) converge and, if so, is lim f (xn ) =
n→∞ n→∞
f (c)?
The answers are yes and yes:

2.3.1 Theorem (Sequential Continuity)

Let c ∈ X and let f : X → R be a function. The following statements are
equivalent

1. f is continuous at c

2. for every sequence (xn ) with xn ∈ X for all n≥1 and lim xn = c we have
n→∞
lim f (xn ) = f (c).
n→∞

Proof We have to show that (i) =⇒ (ii) and that (ii) =⇒ (i).

(i) =⇒ (ii)

Assume (i) is true, i.e. f is continuous at c. We want to prove (ii) so let (xn )
be a sequence with xn ∈ X for all n ≥ 1 and lim xn = c.
n→∞

We want to show that lim f (xn ) = f (c), i.e. given  > 0 there exists N ∈ N
n→∞
such that if n>N then |f (xn ) − f (c)| < .

Let  > 0. Since f is continuous at c there exists δ such that x∈X and |x − c| < δ
implies |f (x) − f (c)| < .

Since lim xn = c and δ > 0, there exists N ∈ N such that if n > N then
n→∞
|xn − c| < δ . Therefore n > N =⇒ |xn − c| < δ =⇒ |f (xn ) − f (c)| <  and

23
lim f (xn ) = f (c). Hence (ii) is true.
n→∞

(ii) =⇒ (i)

Now we assume (ii) is true and we prove (i). We shall do this by showing that if
(i) is false then (ii) is false.

Assume (i) is false, i.e. f is not continuous at c. Since f is not continuous at

c ∃0 such that ∀δ > 0, ∃x ∈ X with |x − c| < δ but |f (x) − f (c)| ≥ 0 .
1 1
For every n ∈ N take δ = and we get xn ∈ X with |xn − c| < but
n n
|f (xn ) − f (c)| ≥ 0 . Hence lim xn = c but (f (xn )) does not tend to f (c) as
n→∞
n → ∞. Therefore (ii) is not true. We have proved that if (i) is false then (ii) is
false. Therefore if (ii) is true then (i) is true.

2.3.2 Remark
Sometime 2.3.1 above is taken as the denition of continuity at c. It is referred to
as sequential continuity.
2.3.1 also gives a good way of saying what it means for f to be not continuous
at c.

f is not continuous at c if and only if there exists a sequence (xn ) with xn ∈ X

∀n ≥ 1 and lim xn = c but the sequence (f (xn )) does not tend to f (c) as n → ∞.
n→∞ 
 0   if x = 0
For example, consider f : R → R be dened by f (x) = 1 .
 sin if x 6= 0
x
1
Then f is not continuous at 0 since if an = π then (an ) tends to 0 as n
+ 2nπ
2
tends to ∞ but f (an ) = 1 for all n ∈ N and so f (an ) tends to 1 as n tends to ∞
but f (0) = 0.

2.4 Open and closed sets

So what are open and closed sets anyway?

2.4.1 Denition
X ⊆ R is said to be open if for all x ∈ X there exists δ > 0 such that (x−δ, x+δ) ⊆
X.

24
2.4.2 Remark
For any a, b ∈ R with a < b, (a, b) is open. If x is such that a<x<b then take δ
to be min{b − x, x − a} in the above.

For any a ∈ R, (a, ∞) and (−∞, a) are open. In the rst case, if x > a, take
δ to be x − a, in the second case, if x < a, take δ = a − x.

It's clear that any union of open sets is open (take an element in the union of
some open sets, it's in one of those open sets, there is a δ that works for that open
set, the same δ works for the union).

2.4.3 Denition
X⊆R is said to be closed if R\X is open.

2.4.4 Remark
For any a, b ∈ R with a < b, [a, b] is closed. This is because R\X = (−∞, a)∪(b, ∞)
is open.

2.4.5 Remark
There is a more useful and equivalent denition that `X contains all its limit
points'. This is very familiar in the context of closed intervals, i.e. if xn ∈ [a, b]
and (xn ) → l then l ∈ [a, b].

2.5 Functions that are continuous at every point of a closed

bounded interval [a,b]
We know what it means for a function to be continuous at a point.

2.5.1 Denition
Let X ⊆ R and let f : X → R be a function. We say f is continuous on X if f is
continuous at every point c of X . This means that ∀c ∈ X , ∀ > 0, ∃δ > 0 (which
can depend on f , c and ) such that x ∈ X and |x−c| < δ implies |f (x)−f (c)| < .

We shall proved some results about functions continuous on a closed interval [a, b].
The main ingredient in the proofs will be the completeness property.

25
2.5.2 Denition
Let X⊆R and the f :X→R be a function.

We say f is bounded above if ∃K ∈ R such that f (x) ≤ K for all x ∈ X.

In other words if sup{f (x) | x ∈ X} < ∞.

We say f bounded below if ∃M ∈ R

is such that f (x) ≥ M for all x ∈ X.
In other words if inf{f (x) | x ∈ X} > −∞.

We say f is bounded if it is bounded above and bounded below.

2.5.3 Remarks
i. f is bounded below if and only if −f is bounded above (because β ≤ f (x) if
and only if −f (x) ≤ −β ).

ii. f is bounded if and only if ∃M > 0 with |f (x)| ≤ M for all x ∈ X.

We come to the rst of three properties of functions continuous on a closed

bounded interval.

1
Note that f : (0, 1] → R given by f (x) = is continous on (0, 1] but not bounded
x
above. This cannot happen if the domain is a closed bounded interval:

2.5.4 Theorem
Let f : [a, b] → R be continuous on [a, b]. Then f is bounded.

Proof

We rst show that f is bounded above. Let C = {x ∈ [a, b] | f |[a,x] is bounded above}.
We want to show that b ∈ C.

Note that a ∈ C so C is non-empty. Also b is an upper bound for C. By the

completeness property of R, sup C exists.

Let d = sup C , so that c≤d for all c∈C and for all >0 there exists c∈C such
that c > d − .

We shall rst show that d(= sup C) = b (note that this is not enough to show

26
that b ∈ C, there will be more to do).

Clearly d ≤ b because b is an upper bound for C. Suppose d < b (for a con-

tradiction).

Since f is continuous at d, taking  = 1 is the denition of continuity gives δ1 > 0

such that if x ∈ [a, b] and |x − d| < δ1 then f (d) − 1 < f (x) < f (d) + 1.

min{b − d, δ1 }
Now let d+ = d + . Then d+ ∈ C as follows:
2
Since d = sup C , there exists d− ∈ C with d − δ1 < d− .

Therefore f |[a,d− ] is bounded above, say f (x) ≤ M for all x ∈ [a, d− ].

But if x ∈ [d− , d+ ] ⊆ (d − δ1 , d + δ1 ) then f (x) < f (d) + 1.

So, for all x ∈ [a, d+ ], f (x) ≤ max{M, f (d) + 1}.

So f |[a,d+ ] is bounded above and d+ ∈ C . But d+ > d = sup C .

This contradiction means that d cannot be strictly less than b, therefore d = b.

Note that this is not enough to conclude that b ∈ C (which gives the result)
a subset of R which is bounded above does not necessarily contain its supremum.

We will show that f |[a,b] is bounded using the fact (now established) that b = sup C .

Since f is continuous at b, ∃δ2 > 0 such that x ∈ [a, b] and |x − b| < δ2 =⇒

f (b) − 1 < f (x) < f (b) + 1 (again, we've taken  = 1 in the denition).

Since b = sup C , ∃d∗ ∈ C with b − δ2 < d∗ .

Since d∗ ∈ C , f |[a,d∗ ] is bounded above, i.e. ∃L > 0 such that f (x) ≤ L ∀x ∈ [a, d∗ ].

Therefore, for all x ∈ [a, b], f (x) ≤ max{L, f (b) + 1} and f is bounded above
on [a, b].

We have shown that any function which is continuous on [a, b] is bounded above.
This applies to −f which is continous on [a, b], so ∃K > 0 such that −f (x) ≤ K
for all x ∈ [a, b].

27
This means that f (x) ≥ −M for all x ∈ [a, b] and so f is bounded below. There-
fore f is bounded.

Now we know that, if f is continous on [a, b] then {f (x) | x ∈ [a, b]} is bounded
above and sup{f (x) | x ∈ [a, b]} exists.

Is this supremum actually the value of f at some point in [a, b]. In other words is
the supremum a maximum?

This isn't true on a non-closed interval because f : [0, 1) → R given by f (x) = x

does not take the value sup{f (x) | x ∈ [0, 1)} = 1.

2.5.5 Theorem (The Extreme Value Theorem)

Let f : [a, b] → R be continuous on [a, b]. Then there exists x+ ∈ [a, b] such that
f (x) ≤ f (x+ ) for all x ∈ [a, b] and there exists x− ∈ [a, b] such that f (x− ) ≤ f (x)
for all x ∈ [a, b].

Another way of stating this conclusion is f (x+ ) = sup{f (x) | x ∈ [a, b]} and
f (x− ) = inf{f (x) | x ∈ [a, b]}.

The points x+ , x− may not be unique. For example, for a constant functionon
[a, b] we can take both x+ and x− to be any point point in [a, b].

Proof

We rst show x+ exists.

Let M = sup{f (x) | x ∈ [a, b]} by theorem 2.5.4. So f (x) ≤ M for all x ∈ [a, b].

We want to nd x+ ∈ [a, b] with f (x+ ) = M . If no such x+ exists then f (x) < M
1
for all x ∈ [a, b] and we can dene a function g : [a, b] → R by g(x) =
M − f (x)
for all x ∈ [a, b].

Clearly g(x) > 0 for all x ∈ [a, b]. Also g is continuous on [a, b] by theorem
2.2.1.

By theorem 2.5.4, g is bounded above so there exists L∈R so that 0 < g(x) ≤ L

28
for all x ∈ [a, b].
1
Hence ≤L for all x ∈ [a, b].
M − f (x)
1
So ≤ M − f (x) for all x ∈ [a, b].
L
1
And f (x) ≤ M − for all x ∈ [a, b].
L
1
Therefore M − is an upper bound for {f (x) | x ∈ [a, b]} contradicting the
L
fact that M is the least upper bound.

This contradiction arose from assuming that f (x) < M for all x ∈ [a, b], so this is
false, and ∃x+ ∈ [a, b] with f (x+ ) = M .

We have shown that for every continuous function f : [a, b] → R, if M =

sup{f (x) | x ∈ [a, b]} then there exists x+ ∈ [a, b] such that f (x+ ) = M .

This also applies to −f , so ∃x− ∈ [a, b] with −f (x− ) = sup{−f (x) | x ∈ [a, b]} =
− inf{f (x) | x ∈ [a, b]}.

So f (x− ) = inf{f (x) | x ∈ [a, b]}.

The following lemma is a special case of the Intermediate Value Theorem, and
we shall easily be able to deduce the general case from it. The proof of the lemma
uses the completeness property of R.

2.5.6 Lemma
Let h : [a, b] → R be continuous on [a, b]. Suppose h(a) < 0 and h(b) > 0. Then
there exists u with a < u < b such that h(u) = 0.

Proof

Let A = {x ∈ [a, b] | h(x) < 0}. We know a ∈ A. (By theorem 2.1.5 we know
[a, a + δ1 ) ⊆ A for some δ1 > 0.)

Since A is non-empty and is bounded above (by b) the completeness property

of R asserts that u = sup A exists.

Clearly a≤u≤b and so h is continuous at u.

29
We shall show h(u) = 0. There are three possibilities h(u) = 0, h(u) < 0 or
h(u) > 0.

Suppose h(u) < 0. Note that this means that u 6= b and so u < b.

By theorem 2.1.5, ∃δ1 > 0 so that if x ∈ [a, b] ∩ (u − δ1 , u + δ1 ) then h(x) < 0. Since
min (b − u, δ1 )
u < b, this means that u1 = u + > u and h(u1 ) < 0. So u1 ∈ A and
2
u1 > u.

This contradicts the fact the u is an upper bound of A. This contradiction arose
from assuming h(u) < 0 and so this is not the case.

Now suppose h(u) > 0. Then u 6= a and so u > a. By theorem 2.1.5, ∃δ2 > 0 such
that ifx ∈ [a, b] ∩ (u − δ2 , u + δ2 ) then h(x) > 0. Let  = min (δ2 , u − a). Then
 > 0. Since u = sup A, there is a point z ∈ A with u −  < z ≤ u and this point
has h(z) < 0.

This contradiction came from assuming h(u) > 0, so this is not the case.

Therefore h(u) = 0. It also follows that u ∈ (a, b) i.e. that u 6= a and u 6= b.

We can now deduce.

2.5.7 Theorem (The Intermediate Value Theorem)

Let f : [a, b] → R be continuous on [a, b]. Suppose f (a) 6= f (b). Let v be any
number strictly between f (a) and f (b).

Then there exists some x ∈ (a, b) such that f (x) = v .

Proof

i. Case f (a) < f (b)

Let v be such that f (a) < v < f (b).

Dene h : [a, b] → R by h(x) = f (x) − v for all x ∈ [a, b].

30
 By theorem 2.2.1 h is continuous on [a, b]. Also h(a) = f (a) − v < 0 and
h(b) = f (b) − v > 0. By lemma 2.5.6 there exists x ∈ (a, b) such that
h(x) = 0, i.e. f (x) = v .
ii. Case f (a) > f (b)

Let v be such that f (a) > v > f (b).

Dene h : [a, b] → R by h(x) = v − f (x) for all x ∈ [a, b].

By theorem 2.2.1 h is continuous on [a, b]. Also h(a) = v − f (a) < 0 and
h(a) = v − f (a) > 0. By lemma 2.5.6 there exists x ∈ (a, b) such that
h(x) = 0, i.e. f (x) = v .

2.5.8 Remark
We can summarise these results in one statement:

If f : [a, b] → R is continuous on [a, b] then the image of f , {f (x) | x ∈ [a, b]}, is

the closed interval [A, B] where A = inf{f (x) | x ∈ [a, b]} and B = sup{f (x) | x ∈
[a, b]}.

We nish this section with some applications of the IVT.

2.5.9 Theorem
Let f : [a, b] → R be continuous on [a, b]. Suppose {f (x) | x ∈ [a, b]} ⊆ Q. Then
f is a constant function.

Proof

Suppose f is not constant. Then we can nd x1 < x2 with f (x1 ) 6= f (x2 ).

Let v be an irrational number strictly between f (x1 ) and f (x2 ). By the IVT
applied to f |[x1 ,x2 ] , which is continuous by 2.1.6, there exists x ∈ [x1 , x2 ] with
f (x) = v .

This contradicts the assumption that {f (x) | x ∈ [a, b]} ⊆ Q. This contradic-
tion arose from assuming that f is not constant. Hence f is constant.

31
2.5.10 Theorem - Fixed Point Theorem
Let f : [a, b] → R be a function continuous on [a, b]. Suppose {f (x) | x ∈ [a, b]} ⊆
[a, b]. Then there is some x ∈ [a, b] such that f (x) = x.

Proof

We have f (a) ≥ a and f (b) ≤ b If f (a) = a or f (b) = b we are nished.

So suppose f (a) > a and f (b) < b. Let h(x) = x − f (x) for all x ∈ [a, b].

Then h : [a, b] → R is continuous on [a, b] and h(a) < 0 and h(b) > 0.

By lemma 2.5.6 there exists x ∈ (a, b) with h(x) = 0, i.e. f (x) = x.

2.6 Continuity of inverse functions

2.6.1 Denitions
Suppose that X ⊆R and Y ⊆R and f :X →R is a function with {f (x) | x ∈
X} ⊆ Y .

Then we say g : Y → R is the inverse function of f if {g(y) | y ∈ Y } ⊆ X

and, ∀x ∈ X , g ◦ f (x) = x and ∀y ∈ Y , f ◦ g(y) = y .

A function f : X → R is said to be injective if f (x1 ) = f (x2 ) =⇒ x1 = x2

for all x1 , x2 ∈ X .

A function f : X → R is said to be increasing if when x1 , x2 ∈ X and x1 < x2

then f (x1 ) ≤ f (x2 ) and strictly increasing if when x1 , x2 ∈ X and x1 < x2 then
f (x1 ) < f (x2 ).

Decreasing and strictly decreasing are denied similarly.

2.6.2 Lemma
Let f : [a, b] → R be continuous on [a, b] and injective. Then f is either strictly
increasing or strictly decreasing.

Proof

32
Since f is injective, f (a) 6= f (b).

i. Case f (a) < f (b).

Let a < x < b. Then f (a) < f (x) < f (b).

If not, either f (x) < f (a) or f (x) > f (b) (can't have f (a) = f (x) or
f (b) = f (x) since f is injective.

In the rst case, f (x) < f (a) < f (b) and so, by the IVT applied to f |[x,b] ,
∃x∗ ∈ (x, b) such that f (x∗ ) = f (a).

But then x∗ 6= a, contradicting f being injective.

In the second case, f (x) > f (b).

So we have f (a) < f (b) < f (x) and, by the IVT applied to f |[a,x] there
∗∗
exists x ∈ (a, x) such that f (x∗∗ ) = f (b).

But x∗∗ 6= b contradicting f being injective.

Now suppose x1 , x2 ∈ R with a ≤ x1 < x2 ≤ b. We will show that

f (x1 ) < f (x2 ) (which means that f is strictly increasing).

Suppose not, for a contradiction. Then f (x1 ) > f (x2 ) (since f (x1 ) 6= f (x2 )
by the injectivity of f) and f (x2 ) < f (x1 ) < f (b)).

By applying the IVT to f |[x2 ,b] there exists d ∈ (x2 , b) such that f (d) = f (x1 ).

Since d 6= x1 this contradicts f being injective.

ii. Case f (b) < f (a).

Then −f (a) < −f (b) and so −f : [a, b] → R is strictly increasing by case

i. Therefore if a ≤ x1 < x2 ≤ b then −f (x1 ) < −f (x2 ). This means that
f (x1 ) > f (x2 ) and f is strictly decreasing.

33
2.6.3 Theorem
Let f : [a, b] → R be continuous on[a, b] and injective. Then f −1 : [A, B] → R
is where A = f (a) and B = f (b) when f (a) < f (b) and A = f (b) and B = f (a)
when f (a) > f (b) is continuous.

Proof

Suppose f (a) < f (b). By lemma 2.5.2, f is strictly increasing and so {f (x) | x ∈
[a, b]} = [f (a), f (b)].

We shall rst show that f −1 : [f (a), f (b)] → R is strictly increasing.

Let f (a) ≤ y1 < y2 ≤ f (b). Let x1 = f −1 (y1 ) and x2 = f −1 (y2 ).

Suppose x1 ≥ x2 . Then y1 = f (x1 ) ≥ f (x2 ) = y2 , a contradiction. Therefore

f −1 (y1 ) = x1 < x2 =f −1
(y2 ) and f −1 is strictly increasing.

Letc ∈ (f (a), f (b). Suppose d = f −1 (c), so that f (d) = c. Let  > 0 be given. Let
 = min{, d−a, b−d}. Since f is strictly increase f (d−∗ ) < f (d) = c < f (d+∗ ).
∗

Let δ = min{c − f (d − ∗ ), f (d + ∗ ) − c. Then, since f −1 is strictly increas-

∗ −1
ing, if |x − c| < δ then f (d − ∗) < x < f (d +  ) which implies f (c) −  ≤
d − ∗ < f −1 (x) < d + ∗ ≤ f −1 (c) +  and so |f −1 (x) − f −1 (c)| < .

Suppose c = f (a). Let >0 be given.

Let δ = f (a + ) − c. If |x − c| < δ then a ≤ f −1 (x) < a +  (if not f −1 (x) ≥ a + 

and x ≥ f (a + ) giving x ≥ c + δ which contradicts |x − c| < δ ).

Suppose c = f (b). Let >0 be given.

Let δ = c − f (b − ). If |x − c| < δ then b −  < f −1 (x) ≤ b (if not f −1 (x) ≤ b − 

and x ≤ f (b − ) giving x ≤ c − δ which contradicts |x − c| < δ ).

Therefore f −1 is continous at c for all c ∈ [f (a), f (b)].

The proof in the case f (b) < f (a) is similar (or use strategy of applying the
above to −f ).

34
3 Limits
We now want to dene limits of functions, i.e. to attach some meaning to the
statement
lim f (x) = M.
x→c

3.1 Denition and some basic properties

Let X ⊆ R and let f : X → R be a function. Let c ∈ R be so that ∃r > 0 such
that (c − r, r) ∪ (c, c + r) ⊆ X (c may or may not be in X ). Let M ∈ R.

We say that f (x) tends to M as x tends to c if ∀ > 0, ∃δ > 0 such that whenever
x∈X and 0 < |x − c| < δ then |f (x) − M | < .

We write f (x) → M as x → c or lim f (x) = M . If we just write this it to

x→c
be considered implicit that the domain of f satises the condition  ∃r >0 such
that (c − r, r) ∪ (c, c + r) ⊆ X  .

It's very important to note that this denition does not require c to be in X,
i.e. forf to be dened at c. In fact, if c ∈ X , so that f (c) is dened, the value of
f (c) does not concern this denition because of the left hand strict inequality in
0 < |x − c| < δ .

This is very useful; it allows us to consider the limit as x tends to 0 of an ex-

pression like
sin(x)
x
sin(0)
where is undened. We will see shortly that this limit is 1. More importantly
0
it will allow us to dene the derivative of f at c as

f (x) − f (c)
lim
x→c x−c
an expression which is again undened when x = c.

3.1.1 Theorem
Limits are unique. In other words if f (x) → M as x→c and f (x) → L as x→c
then M = L.

Proof

35
 |M − L|
Suppose M 6= L. Let = > 0.
2
Since lim f (x) = M and lim f (x) = L, ∃δ1 > 0 such that if x ∈ X and 0 < |x−c| <
x→c x→c
δ1 then |f (x)− M| <  and ∃δ2 > 0 such that if x∈X and 0 < |x − c| < δ2 then
|f (x) − L| < .

Let x∗ ∈ X be such that 0 < |x − c| < min (δ1 , δ2 ). Then |M − f (x∗ )| =

|f (x ) − M | <  and |f (x∗ ) − L| < .
∗

Therefore |M − L| = |M − f (x∗ ) + f (x∗ ) − L| ≤ |M − f (x∗ )| + |f (x∗ ) − L| <

 +  = |M − L|. This contradiction occurs if M 6= L. Therefore M = L.

3.1.2 Theorem
Let f :X→R be a function and suppose lim f (x) = M > 0. Then there exists a
x→c
δ>0 such that if x∈X and 0 < |x − c| < δ then f (x) > 0.

Proof

Let  = M . Then ∃δ > 0 such that if x ∈ X and 0 < |x − c| < δ then

|f (x) − M | <  = M .

Then 0 = M − M < f (x) < M + M = 2M . In particular f (x) > 0.

3.1.3 Remark
The denitions look very similar, so what is the relationship between limits and
continuity?
If f : X → R is a function and c∈X is so that (c − r, c + r) ⊆ R for some
r>0 then f is continuous at c if and only if lim f (x) = f (c).
x→c

3.1.4 Theorem
Let f :X →R be a function. Let c, M ∈ R. The following two statements are
equivalent:

i. lim f (x) = M ,
x→c

36
 ii. for every sequence (xn ) with xn ∈ X and xn 6= c for all n ∈ N and lim xn = c
n→∞
we have lim f (xn ) = M .
n→∞

Proof

In the proof of 2.3.1 replace f (c) by M and include reference to the fact that
xn 6= c for all n ∈ N.

3.1.5 Theorem
Let f :X→R and g:X→R be functions and let c ∈ R. Suppose lim f (x) = L
x→c
and lim g(x) = M . Then
x→c

i. lim(f + g)(x) = L + M
x→c

ii. lim(f g)(x) = LM

x→c
 
f L
iii. if M 6= 0 and g(x) 6= 0 for all x∈X then lim (x) = .
x→c g M
Proof

i. Let (xn ) be a sequence with xn 6= c, xn ∈ X for all n∈N and lim xn = c.

n→∞

Then by 3.1.4 lim f (xn ) = L and lim g(xn ) = M .

n→c n→c

By results from Analysis I (Theorem 3.22), lim (f + g)(xn ) = lim f (xn ) +

n→∞ n→∞
g(xn ) = lim f (xn ) + lim g(xn ) = M + L.
n→∞ n→∞

Therefore by 3.1.4 lim(f + g)(x) = L + M .

x→c

ii. Let (xn ) be a sequence with xn 6= c, xn ∈ X for all n∈N and lim xn = c.
n→∞

Then by 3.1.5 lim f (xn ) = L and lim g(xn ) = M .

n→∞ n→∞

By results from Analysis I (Theorem 3.22),

lim (f g)(xn ) = lim f (xn )g(xn ) = lim f (xn ) lim g(xn ) = M L.

n→∞ n→∞ n→∞ n→c

Therefore by 3.1.5 lim(f g)(x) = LM .

x→c

37
 iii. Let (xn ) be a sequence with xn 6= c, xn ∈ X for all n∈N and lim xn = c.
n→∞

Then by 3.1.5 lim f (xn ) = Llim g(xn ) = M .

and
n→∞ n→∞
 
f f (xn )
By results from Analysis I (Theorem 3.22), lim (xn ) = lim =
n→∞ g n→∞ g(xn )
lim f (xn ) M
 
f L
n→∞
= . Therefore by 3.1.5 lim (x) = .
lim g(xn ) L x→∞ g M
n→c

sin(x)
The next theorem will help us consider and x tends to zero.
x

3.1.6 Theorem
Let f : X → R, g : X → R and h : X → R be functions and let c ∈ R. Sup-
pose lim g(x) = lim h(x) = M and g(x) ≤ f (x) ≤ h(x) for all x ∈ X . Then
x→c x→c
lim f (x) = M .
x→c

Proof

Let  > 0. Since lim g(x) = lim h(x) = M , ∃δ1 > 0 and ∃δ2 > 0 such that if
x→c x→c
x ∈ X and 0 < |x − c| < δ1 then |g(x) − M | < and if x∈X and 0 < |x − c| < δ2
then |h(x) − M | < .

Let δ = min{δ1 , δ2 }.

Suppose x∈X and 0 < |x − c| < δ , then |g(x) − M | <  and |h(x) − M | < .

Therefore, if 0 < |x − c| < δ then M −  < g(x) ≤ f (x) ≤ h(x) < M + 

and |f (x) − M | < . Therefore lim f (x) = M .
x→c

Let's now consider the behaviour of limits under composition of functions.

Let f : X → R and g : Y → R be functions with lim f (x) = M and lim g(x) = L.

x→c x→M
Do we have lim g ◦ f (x) = L?
x→c

Not in general. The problem is that f (x) could equal M near to c but lim g(x)
x→M
does not involve M.

Here is an example of the problem.

38
Let f :R→R be given by f (x) = 2 for all x ∈ R.

5 if x=2
Let g:R→R be dened by g(x) =
7 if x 6= 2.
Then lim f (x) = 2 and lim g(x) = 7 but g◦f : R → R is the constant func-
x→0 x→2
tion g ◦ f (x) = 5 and so lim g ◦ f (x) = 5.
x→2

However we do have:

3.1.7 Theorem
Let f :X→R and g:Y →R be functions with {f (x) | x ∈ X} ⊆ Y , lim f (x) =
x→c
M and lim g(x) = L. If either
x→M

i. ∃t > 0 such that if 0 < |x − c| < t then f (x) 6= M ,

ii. M ∈Y and g is continuous at M (i.e. L = g(M )).

Then lim g ◦ f (x) = L.

x→c
Proof

Let  > 0. We have to nd δ > 0 so that x ∈ X and 0 < |x − c| < δ implies
|(g ◦ f )(x) − L| < .

i. ∃δ1 > 0 such that y∈Y and 0 < |y − M | < δ1 implies |g(y) − K| < .

∃δ2 > 0 such that x∈X and 0 < |x − c| < δ implies |f (x) − M | < δ1 .

Let δ = min{δ2 , t}. Then x ∈ X and 0 < |x−c| < δ implies 0 < |f (x)−M | <
δ1 which implies |(g ◦ f )(x) − L| = |g(f (x)) − L| < .

ii. ∃δ1 > 0 such that y ∈ Y and |y−M | < δ1 implies |g(y)−L| = |g(y)−g(M )| <
.

∃δ > 0 such that x∈X and 0 < |x − c| < δ implies |f (x) − M | < δ1 .

Therefore x ∈ X and 0 < |x − c| < δ implies |f (x) − M | < δ1 which

implies |(g ◦ f )(x) − L| = |g(f (x)) − L| < .

39
3.1.8 Examples
i. By remark 3.1.3, if f : X → R is continuous at c∈X and ∃r > 0 such that
(c − r, c + r) ⊆ R then lim f (x) = f (c).
x→c

Hence lim x3 = c3 , lim sin(x) = sin(c) and lim cos(x) = cos(c).

x→c x→c x→c


3 if x=0
ii. Let f :R→R be dened by f (x) = .
1 if x 6= 0

Then lim f (x) = 1 as follows.

x→0

Let  > 0. We have to choose δ > 0 such that 0 < |x| < δ implies
|f (x) − 1| < . But if x 6= 0 then f (x) = 1 and |f (x) − 1| = 0, so choose δ
to be any positive number.

Notice that if we redene the function at 0 to be lim f (x) = 1 to get

x→0
f¯ : R → R dened by


f (x) if x=0
f¯(x) =
1 if x 6= 0
then f¯ is continuous at 0.

In general if lim f (x) = M ∈ R and we dene f¯ : X ∪ {c} → R by

x→c

f (x) if x 6= c
f¯(x) =
M if x=c

then f¯ is continuous at c.

0 if x<0
iii. Let f :R→R be dened by f (x) = .
1 if x≥0
Then there is no M ∈R such that f (x) → M as x → 0.

To see this, suppose M exists so that f (x) → M as x → 0.

1
Then, taking = in the denition, there ∃δ > 0 such that 0 < |x| < δ
4

40
 1
implies |f (x) − M | < .
4
δ 1
Taking x = − this gives |0 − M | = |M − 0| < which implies that
2 4
1
M< .
4
δ 1 3
Taking x = this gives |1 − M | = |M − 1| < which implies that M> .
2 4 4
This contradiction shows that no such M exists.

sin(x)
iv. lim = 1.
x→0 x

In the graph above 0 is the centre of a circle with radius 1. A and B are on
π
the circle. Angle AOC is x radians where 0 < x < and angles ODA and
2
OBC are right angles.

Note that triangles AOD and COB are similar and this gives

length of AD length of CB
=
length of OD length of OB
or

sin(x) length of CB
=
cos(x) 1

41
 Therefore
length of CB = tan(x).

It's clear that

Area of triangle AOB < Area of sector AOB < Area of triangle COB.

This gives

1 x 1
× 1 × sin(x) < × π × 12 < × 1 × tan(x).
2 2π 2
π
Multiplying by 2 gives sin(x) < x < tan(x) if 0<x< .
2
It follows that
sin(x) sin(x)
cos(x) = ≤ ≤1
tan(x) x
π
for any x with 0<x< .
2
sin(x)
All these functions, cos(x), and 1, are even functions; they are un-
x
changed upon replacing x by −x and therefore

sin(x) sin(x)
cos(x) = ≤ ≤1
tan(x) x

π π
for any x 6= 0 with − <x< .
2 2
sin(x)
By example (i) lim cos(x) = 1 and lim 1 = 1 so by theorem 3.1.6 lim =
x→0 x→0 x→0 x
1.
Note that this gives rise to the approximation sin(x) ≈ x for small x.
x
v. Since f : R → R given by f (x) = satises f (x) 6= 0 for all x 6= 0 and
2
lim f (x) = 0, by theorem 3.1.7,
x→0
x
sin
lim 2 = 1.
x→0 x
2

42
 1 − cos(x)
vi. lim = 0.
x→0 x
Note that it's clear that lim (1 − cos(x)) = 0 (by the continuity of cos)
x→0
and so this result tells us that 1 − cos(x) gets small `more quickly' than x as
x tends to zero.

x
2 2
We have cos(2t) = 1 − 2 sin (t) and so cos(x) = 1 − 2 sin .
2
Therefore
x x
2 2
1 − cos(x) 2 sin sin
lim = lim 2 = lim 2
x→0 x x→0 x x→0 x
2
x
x sin
= lim sin lim 2 = 0 × 1 = 0.
x→0 2 x→0 x
2

3.2 One sided limits

We can consider limits from below and from above. For example we may want
sin(x)
to consider √ which is only dened for x > 0 and ask if it has a limit as x → 0
x
from above.

3.2.1 Denition
Let f :X→R be a function

i. let c∈R be so that (c, c + r) ⊆ X for some r > 0 (c may or may not belong
to X ).

Let M ∈ R.

We say that f (x) tends to M as x tends to c from above if ∀ > 0, ∃δ > 0

such that if x ∈ X and c < x < c + δ then |f (x) − M | < .

We then write lim f (x) = M .

x→c+

ii. let c∈R be so that (c − r, c) ⊆ X for some r > 0 (c may or may not belong
to X ).

43
 Let M ∈ R.

We say that f (x) tends to M as x tends to c from below if ∀ > 0, ∃δ > 0

such that if x ∈ X and c − δ < x < c then |f (x) − M | < .

We then write lim f (x) = M .

x→c−

3.2.2 Remark
It's clear that lim f (x) = M ⇐⇒ lim+ f (x) = M and lim f (x) = M .
x→c x→c x→c−

Supposef : (a, b) → R is a function and c ∈ (a, b). Then f is continuous at

c ⇐⇒ lim f (x) = f (c) ⇐⇒ lim+ f (x) = f (c) and lim− f (x) = f (c).
x→c x→c x→c

So f is not continous at c if and only if one of the following hold

i. lim f (x) exists but is not equal to f (c). This is sometimes called a `remove-
x→c
able discontinuity'. For example:


x2 if x 6= 2
Let f :R→R be dened by f (x) =
5 if x=2

ii. lim f (x) and lim+ f (x) both exist but are not equal. This is sometimes
x→c− x→c
called a `jump discontinuity'. For example:


3 if x<2
Let f :R→R be dened by f (x) =
5 if x≥2

iii. At least one of lim f (x) or lim f (x) doesn't exist. For example:
x→c− x→c+
For example:

 0   if x≤0
Let f :R→R be dened by f (x) = 1
 sin if x>0
x

We get similar results for one-sided limits to those in the previous section so
that

sin(x) sin(x) √ sin(x) √

lim+ √ = lim+ x = lim+ lim+ x = 1 × 0 = 0.
x→0 x x→0 x x→0 x x→0

44
3.3 Limits to and at innity
1
If we consider f : R\{0} → R given by f (x) = , we would like to say f (x) → ∞
|x|
as x → 0.

3.3.1 Denition - limits to innity

Let f : X → R be a function and let c be so that (c − r, c) ∪ (c, c + r) ⊆ X for
some r > 0 (c may or may not belong to X ).

i. We say that f (x) tends to ∞ as x tends to c if ∀M ∈ R, ∃δ > 0 such that if

x∈X and 0 < |x − c| < δ then f (x) > M .

We write lim f (x) = ∞ or f (x) → ∞ as x → c.

x→c

ii. We say that f (x) tends to −∞ as x tends to c if ∀K ∈ R, ∃δ > 0 such that

if x ∈ X and 0 < |x − c| < δ then f (x) < K .

We write lim f (x) = −∞ or f (x) → −∞ as x → c.

x→c

3.3.2 Examples
1
i. →∞ as x→0 as follows.
|x|
1
Let M > 0. We have to nd δ>0 such that 0 < |x| < δ =⇒ > M.
|x|
1 1 1
But > M ⇐⇒ |x| < so take δ= .
|x| M M
1
ii. − → −∞ as x → 0.
|x|
1
iii. f : R \ {0} → R given by f (x) = has no limit as x → 0.
x
There are one sided versions of the above.

3.3.3 Denition - limits at innity

Let f :X→R be a function. Let M ∈ R.

45
 i. Suppose that (r, ∞) ⊆ X for some r.

We say that f (x) tends to M as x tends to ∞ if ∀ > 0, ∃K such that

if x∈X and x > K then |f (x) − M | < .

We write lim f (x) = M or f (x) → M as x → ∞.

x→∞

ii. Suppose that (−∞, r) ⊆ X for some r.

We say that f (x) tends to M as x tends to −∞ if ∀ > 0, ∃L such that

if x ∈ X and x < L then |f (x) − M | < .

We write lim f (x) = M or f (x) → M as x → −∞.

x→−∞

3.3.4 Remark
We can also dene limits that are both too and at innity such as lim f (x) = ∞,
x→∞
lim f (x) = ∞, lim f (x) = −∞, lim f (x) = −∞. It's left as an exercise to
x→−∞ x→∞ x→−∞
write these down and their sequential equivalents.

46
4 Dierentiation
We shall consider functions f : X → R and consider dierentiability at a point
c ∈ X with the property that (c − r, c + r) ⊆ X for some r > 0. Mostly X will be
an open interval (a, b) or a set of the form R \ {0}.

We want to say f is dierentiable at c if we can draw a unique tangent to the

graph of f at c.

f (x) − f (c)
We want the slope to tend to a limit as x tends to c.
x−c

4.1 Denition and examples

4.1.1 Denition
Let f :X→R be a function and let c∈X be so that (c − r, c + r) ⊆ X for some
r > 0.

f (x) − f (c)
We say f is dierentiable at c if, as a function of x, tends to a real
x−c
number M as x tends to c.

When this happens we call M the derivative of f at c and denote it by f 0 (c).

Hence
f (x) − f (c)
f 0 (c) = lim .
x→c x−c

47
4.1.2 Remarks
i. To avoid having to state the technical condition regarding (c − r, c + r) we
will mostly take the domain of functions in results in this chapter to be the
open interval (a, b).
dy
ii. Some authors write y = f (x) and then they write instead of f 0.
dx
iii. Using the denition of limit we get that f is dierentiable at c if and only if
there exists M ∈R with the property that for
 all  > 0, thereexists δ > 0
 f (x) − f (c) 
such that x∈X and 0 < |x − c| < δ implies 
 x−c − M  < .

We then write f 0 (c) instead of M.

f (x) − f (c) f (c + h) − f (c)
iv. We could write f 0 (c) = lim as f 0 (c) = lim by
x→c x−c h→0 h
putting x = c + h.

4.1.3 Examples
i. Let f :R→R be given by f (x) = d for all x ∈ R.

f (x) − f (c) d−d

Let c ∈ R. If x 6= c then = =0→0 as x → c.
x−c x−c
So f 0 (c) = 0 for all c ∈ R.
ii. Let f :R→R be given by f (x) = ax for all x ∈ R.

f (x) − f (c) ax − ac
Let c ∈ R. If x 6= c then = =a→a as x → c.
x−c x−c
So f 0 (c) = a for all c ∈ R.

iii. Let f :R→R be given by f (x) = x2 .

f (x) − f (c) x 2 − c2
Let c ∈ R. If x 6= c then = = x + c → 2c as x → c.
x−c x−c
So f 0 (c) = 2c for all c ∈ R.
iv. Let f :R→R given by f (x) = |x|.

f (x) − f (0) |x| 1 if x>0
If x 6= 0 then = =
x−0 x −1 if x < 0.

48
 f (x) − f (0)
Hence lim does not exist.
x→0 x−0
So f is not dierentiable at 0.

v. sin(x).

sin(x) − sin(c)
Let c ∈ R and x 6= c. We need to consider . We'll make
 x − c 
A+B A−B
use of the identity sin(A) − sin(B) = 2 cos sin .
2 2
By this identity we have, if x 6= c,

sin(x) − sin(c)
=
x−c
   
x+c x−c
2 cos sin
2 2
=
x−c
 
x−c
  sin
x+c 2
cos x−c
2
2
Now
sin(x) − sin(c)
lim
x→c x−c
 
x−c
  sin
x+c 2
lim cos x − c
x→c 2
2
 
x−c
  sin
x+c 2
= lim cos lim x − c
x→c 2 x→c
2
= cos(c) × 1 = cos(c).

This is by theorem 3.1.5 (arithmetic of limits), 3.1.7 (composition of limits)

and 3.1.8 (examples).

49
4.1.4 Theorem
Let f : (a, b) → R be dierentiable at c ∈ (a, b). Then f is continuous at c.

Proof

Since f is dierentiable at c, there exists δ1 > 0 such that x ∈ (a, b) and 0 <
|x − c| < δ1 implies
 
 f (x) − f (c) 0  < 1 =⇒ |f (x) − f (c)| < 1 + |f 0 (c)|.


 x−c − f (c) |x − c|

In other words x ∈ (a, b) and 0 < |x − c| < δ1 implies |f (x) − f (c)| < (1 +
0
|f (c)|)|x − c|.  

Let  > 0 be given. Then let δ = min δ1 , . If x ∈ (c − r, c + r)
1 + |f 0 (c)|
0
and 0 < |x − c| < δ then |f (x) − f (c)| < (1 + |f (c)|)|x − c| < .

Therefore if x ∈ (a, b) and |x−c| < δ then |f (x)−f (c)| <  (since |f (x)−f (c)| =
0 when x = c).

Note that the converse is not true since f : R → R given by f (x) = |x| is
continuous at 0 but not dierentiable at 0.

4.2 How to calculate the derivative of complicated functions

from those of simpler functions
4.2.1 Theorem
Let f : (a, b) → R and g : (a, b) → R be dierentiable at c ∈ (a, b). Then

i. f + g : (a, b) → R is dierentiable at c and (f + g)0 (c) = f 0 (c) + g 0 (c).

ii. f g : (a, b) → R is dierentiable at c and (f g)0 (c) = f (c)g 0 (c) + g(c)f 0 (c).
f
iii. if g(x) 6= 0 for all x ∈ (a, b) and g 0 (c) 6= 0 then : (a, b) → R is dierentiable
 0 g
f g(c)f 0 (c) − f (c)g 0 (c)
at c and (x) = .
g g(c)2

Proof

50
 i. if x 6= c then

(f + g)(x) − (f + g)(c) f (x) − f (c) (g(x) − g(c)

= +
x−c x−c x−c
→ f 0 (c) + g 0 (c) as x→c
by theorem 3.1.5 i.

ii. if x 6= c then

(f g)(x) − (f g)(c) f (x)g(x) − f (x)g(c) + f (x)g(c) − f (c)g(c)

=
x−c x−c
f (x)(g(x) − g(c)) g(c)(f (x) − f (c))
= + → f (c)g 0 (c) + g(c)f 0 (c) as x→c
x−c x−c
by theorem 3.1.5 i. and ii. and by remark 3.1.3 due to the continuity of f
at c.
1
iii. rst consider the derivative of at c, if x 6= c then
g
 
1 1
(x) − (c)  
g g g(c) − g(x) g(x) − g(c) 1 1
= =−
x−c (x − c)g(x)g(c) x−c g(x) g(c)

g 0 (c)
→− as x→c
(g(c))2
by theorem 3.1.5 ii. and by remark 3.1.3 due to the continuity of g at c.
1
So is dierentiable at c and
g
 0
1 g 0 (c)
(c) = − .
g (g(c))2

f 1
The result follows by ii) since =f × and so
g g
 0
f g 0 (c) 0 1 g(c)f 0 (c) − f (c)g 0 (c)
(c) = f (c) × − + f (c) × =
g (g(c))2 g(c) g(c)2
.

51
4.2.2 Remark
If g:X →R c and g(c) 6= 0 then, since g is continuous at c,
is dierentiable at
by theorem 2.1.5 there exists r > 0 such that x ∈ (c−r, c+r) implies that g(x) 6= 0.

Then g|(c−r,c+r) is dierentiable at c and has derivative g 0 (c) there.

f |(c−r,c+r)
So if f : X → R is dierentiable at c then is dierentiable at c and
g|(c−r,c+r)
0 0
g(c)f (c) − f (c)g (c)
has derivative at c equal to .
g(c)2

4.2.3 Corollary
If, for 1 ≤ i ≤ n, each function fi : (a, b) :→ R is dierentiable at c ∈ (a, b) then

i. f1 + f2 + ... + fn : (a, b) → R is dierentiable at c and

(f1 + f2 + ... + fn )0 (c) = f10 (c) + f20 (c) + ... + fn0 (c)

ii. f1 f2 ...fn : (a, b) → R is dierentiable at c and

n
X
0
(f1 f2 ...fn ) (c) = f1 (c)...fi−1 (c)fi0 (c)fi+1 (c)...fn (c).
i=1

Proof

By induction on 4.2.1 i. and ii.

4.2.4 Examples
i. Let m∈N and dene f :R→R by f (x) = xm .

Then f is dierentiable at every c ∈ R and f 0 (c) = mcmm−1 , by corollary

4.2.3 9 (taking each fi R → R to be fi (x) = x for 1 ≤ i ≤ m.

Note also that if x 6= c then

x m − cm
= xm−1 +cxm−2 +c2 xm−3 +...+cm−1 → cm−1 +cm−1 +cm−1 +...cm−1 = mcm−1
x−c
by 2.2.4 (i) (continuity of polynomials).

52
 ii. Let an , an−1 , ..., a1 , a0 ∈ R and dene f :R→R by

f (x) = an xn + an−1 xn−1 + ... + a1 x + a0 .

Then f is dierentiable at every c∈R and

f 0 (c) = an nxn−1 + an−1 (n − 1)xn−2 + ... + a1

example i. and corollary 4.2.3 above.

1
iii. Let m∈N and dene f : R \ {0} → R by f (x) = .
xm
m
Then f is dierentiable at each c 6= 0 and f 0 (c) = − m+1 by (i).
c
m
This is by remark 4.2.2 and the example above since if g(x) = x for all
0 m−1
1 0 g (c) −mc
x ∈ R \ {0} then f = and for c 6= 0, f (c) = − = − =
g g(c)2 c2m
m
− .
cm+1
What is the derivative of cos(7x2 + x3 )? How does dierentiability behave for
composition.

4.2.5 Theorem (The Chain Rule)

Let f : (a, b) → R, g : (d, e) → R be such that {f (x)|x ∈ (a, b)} ⊆ (d, e) so that
g ◦ f : (a, b) → R is dened.

If f is dierentiable at c and g is dierentiable at f (c) then g◦f is dieren-

tiable at c and (g ◦ f )0 (c) = g 0 (f (c))f 0 (c).

Proof

g(f (x)) − g(f (c)) f (x) − f (c)
g(f (x)) − g(f (c))  . if f (x) 6= f (c)
For x 6= c, = f (x) − f (c) x−c
x−c 
0 if f (x) = f (c).

 g(f (x)) − g(f (c))
if f (x) 6= f (c)
Dene Φ : (a, b) → R by Φ(x) = f (x) − f (c)
 0
g (f (c)) if f (x) = f (c).
 
g(f (x)) − g(f (x)) f (x) − f (c)
Then = Φ(x). if x 6= c.
x−c x−c
So the chain rule follows if we show lim Φ(x) = g 0 (f (c)).
x→c

53
Let  > 0. There exists δ1 > 0 such that y ∈ (d, e) and 0 < |y − f (c)| <
 g(y) − g(f (c))
− g 0 (f (c)) < . Call this *.

δ1 =⇒ 
y − f (c)
Since f is continuous at c (theorem 4.1.3), there exists δ2 > 0 such that x ∈ (a, b)
and |x − c| < δ2 =⇒ |f (x) − f (c)| < δ1 .

Therefore x ∈ (a, b) and 0 < |x − c| < δ2 =⇒ |Φ(x) − g 0 (f (c))| < .

This is true if f (x) 6= f (c) by * and true if f (x) = f (c) too because then
0
Φ(x) = g (f (c)).

4.2.6 Examples
i. cos(x).
 π
Since cos(x) = sin x +
2
π
Let h:R→R be h(x) = cos(x). Let f :R→R be given by f (x) = x +
2
and let g:R→R be given by g(x) = sin(x).
Therefore by the chain rule
 = g π◦ f .
h Also f 0 (c) = 1, and g 0 (c) = cos(c), so
 π
g 0 (f (c)) = g 0 x + = cos x + .
2 2
0 0 0 π
0

By the chain rule cos (c) = h (c) = g (f (c))f (c) = cos c + = − sin(c).
2
ii. Let h : R → R be h(x) = sin(x2 ). Let f : R → R be given by f (x) = x2 and
let g : R → R be given by g(x) = sin(x).
Then h = g ◦ f , f 0 (c) = 2c and g 0 (c) = cos(c), so g 0 (f (c)) = g 0 (c2 ) = cos(c2 ).
Therefore by the chain rule h0 (c) = 2c cos(c2 ).

4.3 The derivative function and higher derivatives

Again, we consider domains such as (a, b) and R \ {0}.

4.3.1 Denition
Let X⊆R be so that for all c∈X there exists r>0 with (c − r, c + r) ⊆ X .

If f : X → R is dierentiable at every point c ∈ X then we say f is dieren-

0
tiable on X and the derivative of f is the function f : X → R whose value at

54
c∈X is f 0 (c).

If, in addition, f0 is dierentiable on X (f 0 )0 : X → R is

then its derivative
00 (2)
dened, and is called the second derivative of f . We denote it by either f or f .
We say f is twice dierentiable on X .

Similarly if the (n − 1)th derivative of f exists and is dierentiable on X we call

its derivative the n
th derivative of f and denote if by f (n) : X → R. We say f is
n-times dierentiable on X .

4.3.2 Examples
i. If m∈N andf : R → R is dened by f (x) = xm then f is n-times dieren-
tiable on R for all n ∈ N.

If 1 ≤ n ≤ m then f (n) (x) = m(m − 1)...(m − n + 1)xm−n and if n > m,

(n)
f (x) = 0 for n > m.
ii. Let f :R→R be given by
  
2 1
x sin if x 6= 0

f (x) = x
0 if x = 0.


Then, if c 6= 0,
         
0 1 2 1 1 1 1
f (c) = 2c sin + c − 2 cos = 2c sin − cos .
c c c c c
 
1
x2 sin  
0 x 1
And f (0) = lim = lim x sin = 0.
x→0 x x→0 x
Therefore f0 : R → R is given by

 0   if x = 0
0
 
f (x) = 1 1
 2x sin − cos if x 6= 0.
x x
 
0 0 1
Note that f is not continuous at 0 since f = 1 for any n ∈ N.
(2n + 1)π
This means that f is once dierentiable on R (f 0 exists with the same do-
main as f) but not twice dierentiable on R (because f 0 is not continuous
at 0 and so can't be dierentiable at 0).

55
 iii. Let f :R→R be given by

x2 if x≥0
f (x) =
−x2 if x < 0.

Then f0 : R → R exists as follows:

If c>0 then f 0 (c) = 2c. If c<0 then f 0 (c) = −2c.


f (x) − f (0) x if x>0
Suppose c=0 and x 6= 0. Then = = |x|.
x−0 −x if x < 0.
Since |x| → 0 as x → 0, f 0 (0) = 0. Hence f 0 (x) = 2|x| for all x ∈ R.

So f0 is not dierentiable at 0. If it were then so would be g : R → R

given by g(x) = |x| by 4.2.1. Therefore f is not twice dierentiable on R.

4.4 Local maxima and minima and the Mean Value Theo-
rem
One use of the derivative is in nding local maxima and minima. This helps with
sketching graphs of functions and optimization problems.

4.4.1 Denition
Let f :X→R and let c ∈ X.

We say f has a local maximum at c if there exists δ>0 such that if x∈X
and x ∈ (c − δ, c + δ) then f (x) ≤ f (c).

We say f has a local minimum at c if there exists δ > 0 such that if x ∈ X and
x ∈ (c − δ, c + δ) then f (x) ≥ f (c).

4.4.2 Remarks
i. In proofs that follow, when X = (a, b) we'll assume without loss of generality
that any δ>0 is small enough so that (c − δ, c + δ) ⊆ (a, b).

ii. f has a local minimum at c if and only if −f has a local maximum at c.

iii. A function can have several local maximum on
  X. For example f : (0, 1) → R
1
given by f (x) = x sin has innitely many. Indeed, every point is both
x
a local max and min for a constant function.

56
 We'll see in the next few results that 3.1.2 has deep consquences.

4.4.3 Theorem
Let f : (a, b) → R and let c ∈ (a, b). If f has a local maximum, or local minimum,
0
at c and if f is dierentiable at c then f (c) = 0.

Proof

Suppse there is a local maximum at c.

Then there exists δ>0 such that x ∈ (c − δ, c + δ) implies f (x) ≤ f (c).

f (x) − f (c)
This means that if c−δ < x < c then ≥ 0 and if c < x < c+δ
x−c
f (x) − f (c)
then ≤ 0.
x−c
Suppose f 0 (c) < 0. By theorem 3.1.2, there exists δ1 > 0 such that if 0 < |x − c| <
f (x) − f (c)
δ1 then < 0.
x−c
min{δ, δ1 } f (x1 ) − f (c) f (x1 ) − f (c)
But then if x1 = c − we have ≥ 0 and < 0,
2 x1 − c x1 − c
a contradiction.

Suppose f 0 (c) > 0. By theorem 3.1.2, there exists δ2 > 0 such that if 0 < |x − c| <
f (x) − f (c)
δ2 then > 0.
x−c
min{δ, δ2 } f (x2 ) − f (c) f (x2 ) − f (c)
But then if x2 = c + we have ≤ 0 and > 0,
2 x2 − c x2 − c
a contradiction.

Hence f 0 (c) = 0.

If c
is a local minimum then apply the above proof to −f to get (−f )0 (c) = 0. It
0
follow then that f (c) = 0.

57
4.4.4 Denition
If f :X →R and f is dierentiable at c then c is called a critical point of f if
0
f (c) = 0.

4.4.5 Remark
Theorem 4.4.3 says a local max or min where a function is dierentiable is a critical
point.

However not all critical points are a local max or a local min.

For example f : R → R given by f (x) = x3 has critical point 0 which isn't a

local max or a local min.

4.4.6 Theorem - Rolle's theorem

Suppose h : [a, b] → R is continuous on [a, b] and dierentiable on (a, b). Suppose
h(a) = h(b).

Then there exists some c ∈ (a, b) with h0 (c) = 0.

Proof

If h is constant then, for any c ∈ (a, b) and x ∈ (a, b) with x 6= c,

h(x) − h(c)
h0 (c) = lim = 0.
x→c x−c

Suppose h is not constant.

Case (i): There exists d ∈ (a, b) such that h(d) > h(a) = h(b).

Then, by the EVT, there exists c ∈ [a, b] such that h(c) ≥ h(x) for all x ∈ [a, b].

So h(c) ≥ h(d) > h(a) = h(b) and therefore c 6= a and c 6= b and c ∈ (a, b).

By theorem 4.4.3, since h is dierentiable at c and c is a local maximum, h0 (c) = 0.

Case (ii): There exists d ∈ (a, b) such that h(d) < h(a) = h(b).

Then, by the EVT, there exists c ∈ [a, b] such that h(c) ≤ h(x) for all x ∈ [a, b].

58
So h(c) ≤ h(d) < h(a) = h(b) and therefore c 6= a and c 6= b and c ∈ (a, b).

By theorem 4.4.3, since h is dierentiable at c and c is a local minimum, h0 (c) = 0.

4.4.7 Theorem - Mean Value Theorem

Suppose f : [a, b] → R is continuous on [a, b] and dierentiable on (a, b). Then
f (b) − f (a)
there exists some c ∈ (a, b) with f 0 (c) = .
b−a
Proof

f (b) − f (a)
Let h : [a, b] → R be given by h(x) = f (x) − f (a) − (x − a).
b−a
Then h(a) = h(b) = 0, h is continuous on [a, b] and dierentiable on (a, b).

Therefore by 4.4.6 there exists c ∈ (a, b) such that h0 (c) = 0.

f (b) − f (a) f (b) − f (a)

h0 (x) = f 0 (x) − and so 0 = h0 (c) = f 0 (c) − giving
b−a b−a
f (b) − f (a)
f 0 (c) = .
b−a

4.4.8 Corollary
Iff : (a, b) → R is dierentiable and f 0 (x) = 0 for all x ∈ (a, b) then f is constant
on (a, b).

Proof

Let x1 , x2 ∈ (a, b) with x1 < x2 .

By the MVT applied to f |[x1 ,x2 ] there exists c ∈ (x1 , x2 ) such that f |0[x1 ,x2 ] (c) =
f (x2 ) − f (x1 )
.
x2 − x1
But f |0[x1 ,x2 ] (c) = f 0 (c) = 0 and so f (x2 ) − f (x1 ) = 0 or f (x1 ) = f (x2 ).

Since x1 and x2 were arbitrary f is constant on (a, b).

59
4.4.9 Corollary
If f : (a, b) → R is dierentiable then

i. if f 0 (x) > 0 for all x ∈ (a, b) then f is strictly increasing on (a, b)

ii. if f 0 (x) < 0 for all x ∈ (a, b) then f is strictly decreasing on (a, b).

Proof

i. Let x1 , x2 ∈ (a, b) with x1 < x2 .

By the MVT applied to f |[x1 ,x2 ] there exists c ∈ (x1 , x2 ) such that f |0[x1 ,x2 ] (c) =
f (x2 ) − f (x1 )
.
x1 − x2
But f |0[x1 ,x2 ] (c) = f 0 (c) > 0. Since x2 > x1 , f (x2 ) − f (x1 ) > 0 or f (x1 ) >
f (x2 ).

Therefore f is strictly increasing on (a, b).

ii. Similar.

4.4.10 Theorem
Let f : (a, b) → R be twice dierentiable on (a, b). Let c ∈ (a, b) be a critical point
for f (i.e. f 0 (c) = 0). Then

i. If f (2) (c) > 0 then c is a local minimum for f

ii. If f (2) (c) < 0 then c is a local maximum for f.

f 0 (x) − f 0 (c) f 0 (x)

Proof We have f 00 (c) = lim = lim (since f 0 (c) = 0).
x→c x−c x→c x − c

i. Suppose f (2) (c) > 0. By theorem 3.1.2 there exists δ > 0 such that if
f 0 (x)
0 < |x − c| < δ then > 0.
x−c
0
This means that if x ∈ (c − δ, c) then f (x) < 0 and by corollary 4.4.9 f is
strictly decreasing on (c − δ, c).

Similarly if x ∈ (c, c + δ) then f 0 (x) > 0 and f is strictly increasing on

(c, c + δ).

60
 c−x
Take x ∈ (c − δ, c). Consider the sequence an = c − . Then x <
n
an ∈ (c − δ, c) for all n ≥ 2 and lim an = c. Since f is decreasing on
n→∞
(c − δ, c), f (x) > f (an )lim f (an ) = f (c) (continuity of
and f at c). Hence
n→∞
(by example 25 from Analysis I) f (x) ≥ f (c).

x−c
Take x ∈ (c, c + δ). Consider the sequence bn = c + . Then x > an ∈
n
(c, c + δ) for all n ≥ 2 and lim bn = c Since f is increasing on (c, c + δ),
n→∞
f (x) > f (bn ) and lim f (bn ) = f (c) (continuity of f at c). Hence (by example
n→∞
25 from Analysis I) f (x) ≥ f (c).

ii. Apply [i.] to −f .

4.4.11 Remark
Sometimes it happens that f 0 (c) = 0 and f 00 (c) = 0. We can then use either of the
following methods to see if c is a local maximum or a local minimum or neither.

i. Find the sign of f 0 (x) for x near to c so we can see where f is increasing and
descreasing.

ii. See Taylor's theorem.

4.5 Inverse Function Theorem

4.5.1 Theorem - Dierentiability of inverse function
Let f : [a, b] → R be continuous on [a, b] and injective. Then f −1 : [A, B] → R
is where A = f (a) and B = f (b) when f (a) < f (b) and A = f (b) and B = f (a)
when f (a) > f (b) is continuous on [A, B].
Let c ∈ (a, b). Then

i. if f is dierentiable at c and f 0 (c) = 0 then f −1 is not dierentiable at f (c).

ii. if f is dierentiable at c then f −1 is dierentiable at f (c) and (f −1 )0 (f (c)) =

1
.
0
f (c)

Proof

Note that the rst part of the statement is just 2.6.3.

61
 i. Suppose f −1 is dierentiable at c. Then, by the chain rule (4.2.5), since
−1
f ◦ f : [a, b] → R is the identity function we have

1 = (f −1 )0 (f (c))f 0 (c) = (f −1 )0 (f (c)) × 0 = 0

a contradiction.

ii. We know that f −1 is continuous at f (c) by 2.6.3. Note also that f (c) ∈ (A, B)
since c ∈ (a, b) and f is injective.

f −1 (y) − f −1 (f (c)) f −1 (y) − c

Note that, if y 6= f (c) and y ∈ (A, B) then = .
y − f (c) f (f −1 (y)) − f (c)
Since f is dierentiable at c and f 0 (c) 6= 0 we have, by 3.1.5:

x−c 1
lim = 0 .
x→c f (x) − f (c) f (c)

This means that for all

 >  0 there exists δ>0 such that if 0 < |x − c| < δ
 x−c 1 
then
 f (x) − f (c) − f 0 (c)  < .
 

Then, by the continuity and injectivity of f −1 at f (c) there exists δ1 >

0 such that if 0 < |y − f (c)| < δ1 then 0 < |f −1 (y) − c| < δ and so
 f −1 (y) − c 1  −1
 −
 f (f −1 (y)) − f (c) f 0 (c)  < . Therefore f is dierentiable at f (c) and

1
(f −1 )0 (f (c)) = 0 .
f (c)

4.6 Applications of dierentiation to limits

4.6.1 Cauchy Mean Value Theorem
Let f : [a, b] → R and g : [a, b] → R be continuous on [a, b] and dierentiable on
(a, b). Then there exists w ∈ (a, b) such that

f 0 (c)(g(b) − g(a)) = g 0 (c)(f (b) − f (a)).

Note that the MVT is the case g(x) = x.

Proof

If g(b) = g(a) the results is true if we choose c ∈ (a, b) with g 0 (c) = 0 by Rolle's

62
theorem.

If g(b) 6= g(a) we can dene h : [a, b] → R by

f (b) − f (a)
h(x) = f (x) − f (a) − (g(x) − g(a)).
g(b) − g(a)
Then h(a) = 0 andh(b) = 0 (check). Also h is continuous on [a, b] and dier-
0
entiable on (a, b) so by Rolle's theorem there exists c ∈ (a, b) with h (c) = 0.

f (b) − f (a) 0
In other words f 0 (c) = g (c).
g(b) − g(a)

4.6.2 L'Hopital's rule

Let c∈R and let X = (c − r, c + r) or X = (c − r, c) ∪ (c, c + r).

Let f : X → R, g : R → R be functions with lim f (x) = 0 and lim g(x) = 0.

x→c x→c

Suppose f and g are both dierentiable on (c − r, c) ∪ (c, c + r) and that g 0 (x) 6= 0

for all x ∈ (c − r, c) ∪ (c, c + r).

f 0 (x)
Then g(x) 6= 0 for all x ∈ (c − r, c) ∪ (c, c + r) and if lim = M ∈ R
x→c g 0 (x)
f (x)
then lim = M.
x→c g(x)

Proof

Dene fˆ : (c − r, c + r) → R by

f (x) if x 6= c
fˆ(x) =
0 if x = c.

Then fˆ is continuous at c. Similarly if ĝ : (c − r, c + r) → R is dened by


g(x) if x 6= c
ĝ(x) =
0 if x = c.

Then ĝ is continuous at c.

Let x ∈ (c − r, c). Then g(x) 6= 0 because, by the MVT applied to ĝ|[x,c] we

0
have g(x) = ĝ(c) − ĝ(x) = g (z)(c − x) 6= 0 for some z ∈ (x, c).

63
Applying the Cauchy MVT (4.5.1) to fˆ|[x,c] and ĝ|[x,c] we get wx ∈ (x, c) such
that
f 0 (wx ) fˆ(c) − fˆ(x) f (x)
0
= = .
g (wx ) ĝ(c) − ĝ(x) g(x)

Note that |wx − c| < |x − c|.

Let x ∈ (c, c + r). Then, as above, g(x) 6= 0 and there exists wx ∈ (c, x) with
0
f (wx ) f (x)
= . Again |wx − c| < |x − c|.
0
g (wx ) g(x)
f 0 (wx ) f (x)
So if 0 < |x−c| < r, there exists wx with 0 < |wx −c| < |x−c| and 0
= .
g (wx ) g(x)
f 0 (y)
But lim = M so that if  > 0 there exists δ>0 such that 0 < |y − c| < δ
y→∞ g 0 (y)
 0 
 f (y) 
implies that 
 g 0 (y) − M  < .

 
 f (x) 
Then 0 < |x − c| < δ implies 0 < |wx − c| < δ implies 
 g(x) − M  < .

f (x)
Hence lim = M.
x→∞ g(x)

4.6.3 Examples
sin(x)
i. lim = 1.
x→0 x
 
−π π
We can now see this by taking X= , , f :X→R as f (x) = sin(x)
2 2
and g:X→R as g(x) = x in the above.

The conditions are satised, lim f (x) = 0 and lim g(x) = 0, and f 0 (x) =
x→0 x→0
cos(x), g 0 (x) = 1.

sin(x) cos(x)
So L'Hopital's rule says that lim = lim = 1.
x→0 x x→0 1
1 − cos(x)
ii. lim = 0.
x→0 x
 
−π π
Take X = , , f :X →R as f (x) = 1 − cos(x) and g :X →R as
2 2
g(x) = x in the above.

64
 The conditions are satised, lim f (x) = 0 and lim g(x) = 0, and f 0 (x) =
x→0 x→0
sin(x), g 0 (x) = 1.

1 − cos(x) sin(x)
So L'Hopital's rule says that lim = lim = 0.
x→0 x x→0 1
x3 − 8
iii. Exercise for lectures: lim = 12.
x→2 x − 2

(2x − π)2
iv. Exercise for lectures: limπ
x→ 2 1 − sin(x)

4.7 Taylor's Theorem

For a function f : [a, b] → R denition 4.1.1 applies only to points c ∈ (a, b). To
discuss dierentiablity of f at a and b we have to consider one-sided limits.

4.7.1 Denitions
Let f : [a, b] → R be a function.

f (x) − f (a)
i. We say that f is dierentiable at a is tends to a real number
x−a
M as x tends to a from above. In other words if there exists M ∈R with
f (x) − f (a)
lim = M . We call M the derivative of f at a and denote it by
x→a+ x−a
f 0 (a).
f (x) − f (b)
ii. We say that f is dierentiable at b is tends to a real number
x−b
K as x tends to b from below. In other words if there exists K ∈ R with
f (x) − f (b)
lim = K . We call K the derivative of f at b and denote it by
x→b+ x−b
f 0 (b).

iii. If f : [a, b] → R is dierentiable at every point c ∈ [a, b] then we say f is

0
dierentiable on [a, b] and the derivative of f is the function f : [a, b] → R
whose value at c ∈ [a, b] is f 0 (c).

iv. If, in addition, f0 [a, b] then its derivative (f 0 )0 : [a, b] → R

is dierentiable on
is dened, and is called the second derivative of f . We denote it by either
f 00 or f (2) . We say that f is twice dierentiable on [a, b].

65
 v. Similarly if the (n − 1)th derivative of f exists and is dierentiable on [a, b]
we call its derivative the n
th derivative of f and denote if by f (n) : [a, b] → R.
We say the f is n-times dierentiable on [a, b].

vi. We say that f is innitely dierentiable on [a, b] if f is n-times dierentiable

on [a, b] for all n ∈ N.
Note that equivalent denitions can be made for a domain of the form [a, b)
and of the form (a, b].
Let I be an interval (closed, open or half open half closed). f : I → R is said
to be in the class C 0 (I) if f is continuous on I .
f :I → R is said to be in the class C 1 (I) if f is dierentiable on I and f 0 is
continuous on I.
f :I→R is said to be in the class C 2 (I) if f is twice dierentiable on I and
00
f is continuous on I.
Note also that C 0 (I) ⊇ C 1 (I) ⊇ C 2 (I) ⊇ ..... If f ∈ C n [I] for all n = 0, 1, 2, ..
then we say f ∈ C ∞ [I].

4.7.2 Remark
If f : [a, b] → R is dierentiable at a then it is continuous at a and similarly for b.
The proof is the same as theorem 4.1.4.

4.7.3 Theorem - Taylor's theorem

Let n ≥ 1. Suppose that f ∈ C n ([a, b]) and that f (n) : [a, b] → R is dierentiable
on (a, b).
Then

i. there exists c ∈ (a, b) such that

f 00 (a)
f (b) = f (a) + f 0 (a)(b − a) + (b − a)2 +
2!
f (n) (a) f (n+1) (c)
... + (b − a)n + (b − a)n+1 .
n! (n + 1)!
f (n+1) (c)
Here Rn = (b − a)n+1 is called the Lagrange form of the remainder.
(n + 1)!
ii. the same is true in the above with a and b swapped, i.e. there exists a
(possibly dierent) c ∈ (a, b) such that

0 f 00 (b)
f (a) = f (b) + f (b)(a − b) + (a − b)2 +
2!
66
 f (n) (b) f (n+1) (c)
... + (a − b)n + (a − b)n+1 .
n! (n + 1)!
f (n+1) (c)
Here Rn = (a − b)n+1 .
(n + 1)!
iii. there exists d ∈ (a, b) such that

f 00 (a)
f (b) = f (a) + f 0 (a)(b − a) + (b − a)2 +
2!
f (n) (a) f (n+1) (d)
... + (b − a)n + (b − d)n (b − a).
n! n!
(n+1)
f (d)
Here Rn = (b−d)n (b−a) is called the Cauchy form of the remainder.
n!
iv. the same is true in the above with a and b swapped, i.e. there exists a
(possibly dierent) d ∈ (a, b) such that

f 00 (b)
f (a) = f (b) + f 0 (b)(a − b) + (a − b)2 +
2!
f (n) (b) f (n+1) (d)
... + (a − b)n + (a − d)n (a − b).
n! n!
f (n+1) (d)
Here Rn = (a − d)n (a − b).
n!
Proof
i. Let γ∈R be such that

f 00 (a)
f (b) = f (a) + f 0 (a)(b − a) + (b − a)2 +
2!
f (n) (a)
... + (b − a)n + γ(b − a)n+1 .
n!
Let h : [a, b] → R be dened by

f 00 (x)
h(x) = f (b) − [f (x) + f 0 (x)(b − x) + (b − x)2 +
2!
f (n) (x)
... + (b − x)n + γ(b − x)n+1 ].
n!
Then h(a) = h(b) = 0, h is continuous on [a, b] and dierentiable on (a, b).
0
Therefore there exists c ∈ (a, b) such that h (c) = 0 by Rolle's theorem.

f (n+1) (c)
It can be checked that this implies that γ= .
(n + 1)!

67
ii. Let g : [−b, −a] be dened by g(x) = f (−x) for all x ∈ [−b, −a].

By i. there exists c ∈ (−b, −a) such that

0 g 00 (−b)
g(−a) = g(−b) + g (−b)(−a − (−b)) + (−a − (−b))2 +
2!
g (n) (a) g (n+1) (c)
... + (−a − (−b))n + (−a − (−b))n+1 .
n! (n + 1)!

But g(−x) = f (x) and g (i) (−x) = −1i f (i) (x) for any x ∈ [a, b].

Using this in the above gives

f 00 (b)
f (a) = f (b) + f 0 (b)(a − b) + (a − b)2 +
2!
f (n) (b) f (n+1) (−c)
... + (a − b)n + (a − b)n+1 .
n! (n + 1)!
iii. Let h : [a, b] → R be dened by

f 00 (x)
h(x) = f (b) − [f (x) + f 0 (x)(b − x) + (b − x)2 +
2!
f (n) (x)
... + (b − x)n ].
n!
Then h is continuous on [a, b] and dierentiable on (a, b). Note that the
remainder R is h(a) and h(b) = 0. By the Mean Value Theorem there exists
c ∈ (a, b) such that

h(b) − h(a)
h0 (c) = .
b−a
Therefore h0 (c)(a − b) = h(a) = R. Now

f (3) (x)
h0 (x) = −f 0 (x) + f 0 (x) − f 00 (x)(b − x) + f 00 (x)(b − x) − (b − x)2 + ...
2!
f (n+1) (x) f (n+1) (x)
− (b − x)n = − (b − x)n .
n! n!
f (n+1) (c)
This gives R = h(a) = h0 (c)(a − b) = − (b − c)(n) (a − b) as required.
n!
iv. Similar to ii.

68
4.7.4 Corollary
Let I be an interval (open, closed or half open-half closed) and assume f : I → R
n (n)
is in C (I) and f : I → R is dierentiable on the open interval determined by
I . Let x0 ∈ I . Let Rn (x0 , x) be dened for x ∈ I by
n
X f (j) (x0 )
f (x) = (x − x0 )j + Rn (x0 , x).
j=0
j!
n f (j) (x )
0
(x − x0 )j
P
Pn,x0 (x) : I → R given by Pn,x0 (x) = is called the Taylor
j=0 j!
polynomial of f at x0 of degree n.
Then

i. there is a value c between x0 and x such that

f (n+1) (c)
Rn (x0 , x) = (x − x0 )n+1 .
(n + 1)!

ii. there is a value d between x0 and x such that

f (n+1) (d)
Rn (x0 , x) = (x − x0 )(x − d)n
n!
Proof
If x = x0 take c = x0 in part i. and d = x0 in part ii.

i. If x > x0 then the results follows by Taylor's theorem 4.7.4.i with x0 = a

and x = b.

Ifx < x0 then the results follows by Taylor's theorem 4.7.4.ii with x0 = b
and x = a.

ii. If x > x0 then the results follows by Taylor's theorem 4.7.4.iii with x0 = a
and x = b.

Ifx < x0 then the results follows by Taylor's theorem 4.7.4.iv with x0 = b
and x = a.

4.7.5 Example
1. Let x0 = 0 and x1 = 0.1 and f :R→R be given by f (x) = sin(x).
Then by the corollary 4.7.4

f 00 (0)
0
sin(0.1) = f (0.1) = f (0) + f (0)(0.1 − 0) + (0.1 − 0)2
2!
69
 f 00 (0) f 00 (c)
+ (0.1 − 0)3 + (0.1 − 0)4
3! 4!
sin(0)
= sin(0) + cos(0) × 0.1 − × 0.01
2!
cos(0) sin(c)
− × 0.001 + × 0.0001
3! 4!

1 sin(c)
= 0.1 − × 0.001 + × 0.0001.
3! 4!
0.001 sin(c)
This can be interpreted as sin(0.1) ≈ 0.1 − with an error of ×
6 4!
0.0001.
0.001
A calculator givessin(0.1) as 0.09983341664 and 0.1− as 0.09983333333.
6
0.0001
Clearly the size of the error is bounded by . A calculator gives this
4!
value as 0.00000416666.

2. Exercise for lectures: Use Taylor's theorem with n = 2 to approximate

cos(0.1) and give a bound for the size of the error.

4.7.6 Corollary
Suppose f : (a, b) → R is innitely dierentiable on (a, b) and let x0 be a critical
(n)
point. Suppose n ≥ 1 is the smallest natural number for which f (x0 ) 6= 0.

i. Ifn is even then f (n) (x0 ) < 0 implies x0 is a local maximum and f (n) (x0 ) > 0
implies x0 is a local minimum.

ii. If n is odd then x0 is neither a local maximum nor a local minimum.

Proof

i. Suppose n is even and f (n) (x0 ) < 0. By the continuity of f (n) at x0 there
(n)
exists δ > 0 such that if |c − x0 | < δ then f (c) < 0 Let Pn−1,x0 (x) be the
Taylor polynomial of f of degree n − 1 at c. Then by corollary 4.7.4

f (x) = Pn−1,x0 + Rn−1 (x0 , x)

where there is a value c between x0 and x such that

f (n) (c)
Rn−1 (x0 , x) = (x − x0 )n .
n!

70
 Note that written in full this is

n−1 (j)
X f (x0 ) f (n) (c)
f (x) = (x − x0 )j + (x − x0 )n .
j=0
j! n!

Since f 0 (x0 ) = f 00 (x0 ) = ... = f (n−1) (x0 ) = 0 this gives

f (n) (c)
f (x) = f (x0 ) + (x − x0 )n .
n!

Since |c − x0 | < |x − x0 |, if |x − x0 | < δ then f (n) (c) < 0. Since n is even

if x 6= x0 then (x − x0 )n > 0. Therefore f (x) ≤ f (x0 ) and x0 is a local
maximum.

The case when f (n) (x0 ) > 0 is similar.

ii. Suppose f (n) (x0 ) < 0. Following the same steps as above there is a value c
between x and x0 such that

f (n) (c)
f (x) = f (x0 ) + (x − x0 )n
n!
and a δ>0 such that if |c − x0 | < δ then f (n) (c) < 0.
Since |c − x0 | < |x − x0 |, |x − x0 | < δ implies that f (n) (c) < 0. Since n is odd
if x > x0 then (x − x0 )n > 0 and if x < x0 then (x − x0 )n < 0. Therefore x0
is neither a local maximum or a local minimum.

4.7.7 Corollary
Let p : R → R be a polynomial function given by p(x) = an xn + an−1 xn−1 + ... +
a1 x + a0 . Then p is completely determined by its rst n derivatives at any point
x0 ∈ R .
Proof Note that p(n+1) (x) = 0 for all x ∈ R. Corollary 4.7.4 says that
n
X p(j) (x0 )
p(x) = (x − x0 )j + Rn (x0 , x)
j=0
j!

p(n+1) (c)
where there is a value c between x such that Rn (x0 , x) = (x − x0 )n+1 .
(n + 1)!
Therefore Rn (x0 , x) = 0 and so

0 p00 (x0 ) 2 p(n)

p(x) = p(x0 ) + p (x0 )(x − x0 ) + (x − x0 ) + ... + (x0 )(x − x0 )n .
2! n!
71
4.7.8 Example
Let p:R→R be given by p(x) = x2 + 3x + 2. Clearly p(3) (x) = 0 for all x ∈ R.

The Taylor polynomial of p at 1 of degree 2, denoted P2,1 is

p00 (1)
P2,1 (x) = p(1) + p0 (1)(x − 1) + (x − 1)2 .
2!

p(1) = 6
p0 (x) = 2x + 3 so p0 (1) = 5
p00 (x) = 2 so p00 (1) = 2.
Substituting into the above gives

2
P2,1 (x) = 6 + 5(x − 1) + (x − 1)2 = 6 + 5x − 5 + x2 − 2x + 1 = x2 + 3x + 2.
2

72
5 Functions dened by power series
In this chapter we'll be considering expressions of the form

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

∞ ∞
an x n an (x − x0 )n ).
P P
or ( or
n=1 n=1

Results about series from Analysis I will be important.

5.1 Preliminaries about series

5.1.1 Denitions and terminology
∞
P
Recall that, for a series bn
n=0

n
P
i. if the sequence of partial sums tn = b0 + b1 + ... + bn = bi converges to
i=0
∞
P ∞
P
a real number b then we say bn converges and write bn = b (5.1 in
n=0 n=0
Analysis I notes).

∞
P ∞
P
ii. if (tn ) → ∞ then we say bn diverges to innity and write bn = ∞ (5.1
n=0 n=0
in Analysis I notes).

∞
P ∞
P
iii. if (tn ) → −∞ then we say bn diverges to minus innity and write bn =
n=0 n=0
−∞ (5.1 in Analysis I notes).

∞
P
iv. if none of the above happen then we say bn diverges (5.1 in Analysis I
n=0
∞
P
notes). Note that when we say that bn diverges we could be in any of
n=0
cases ii, iii or iv.

∞
P ∞
P
v. bn |bn | converges. A series that is
is said to converge absolutely if
n=0 n=0
absolutely convergent is convergent (5.20 in Analysis I notes).

73
5.1.2 Results we'll need
∞
P ∞
P ∞
P
i. if an and bn converge then an + b n converges and
n=0 n=0 n=0

∞
X ∞
X ∞
X
an + b n = an + bn
n=0 n=0 n=0

(5.10 in Analysis I notes).

∞
P ∞
P ∞
P ∞
P
ii. if an converges λ ∈ R then λan converges and λan = λ an (5.10
n=0 n=0 n=0 n=0
in Analysis I notes).
 
∞
P ∞
P P∞ 
iii. if bn is absolutely convergent then bn is convergent and
n=0  ≤
b n
 
n=0 n=0
∞
P
|bn | (5.21 in Analysis I notes).
n=0

∞
P
iv. if bn converges then (bn ) → 0 as n→∞ (5.15 in Analysis I notes).
n=0

∞
P ∞
P
v. Suppose 0 ≤ an ≤ b n for all n and that bn converges. Then an
n=0 n=0
converges and
∞
X ∞
X
an ≤ bn .
n=0 n=0

This is the comparison test (5.17 in the Analysis I notes).

∞
P ∞
P ∞
P
vi. By the above, if an and bn converge absolutely then an + b n con-
n=0 n=0 n=0
∞
P ∞
P ∞
P
verges absolutely and |an + bn | ≤ |an | + |bn |.
n=0 n=0 n=0
 
∞
P  bn+1 
vii. Given a series bn with bn 6= 0 for n = 0, 1, 2, ..., if
 bn  → L ∈ R
  then
n=0
∞
P
bn converges absolutely if 0 ≤ L < 1 and diverges if L > 1 (including if
n=0
 
 bn+1 
 bn  → ∞). This is the ratio test (5.23 in the Analysis I notes).
 

viii. Bernoulli's inequality which says that (1+x)n ≥ 1+nx if x ≥ −1 and n ∈ N.

This is 3.36 in the Analysis I notes.

74
5.2 Denition and examples
5.2.1 Denition
∞
an x n
P
A power series is an expression of the form or an expression of the form
n=0
∞
an (x − x0 )n
P
where x, x0 ∈R and an ∈ R for all n≥0 (although sometime our
n=0
summations will start at n = 1). By convention x0 = 1 for allx ∈ R, 0n = 0 for
∞
an x n .
P
all n≥0 and 0! = 1. We will concentrate on power series of the form
n=0

5.2.2 Examples
1 P∞ xn
i. Taking an = gives the power series . Given any xed x∈R the
n! n=0 n!
P∞ xn
series converges by the ratio test.
n=0 n!

1 P∞ xn
ii. Taking an = gives the power series . This converges for x ∈ R with
n n=1 n!
|x| < 1 and diverges for x∈R with |x| > 1. It converges when x = −1 and
diverges when x = 1.
∞
xn .
P
iii. Taking an = 1 gives the power series This converges if x ∈ (−1, 1)
n=0
and diverges for all other x∈R (by the ratio test for |x| > 1 and by direct
check for x=1 and x = −1).
∞
n!xn .
P
iv. Taking an = n! gives the power series This converges if x=0 but
n=0
diverges for all other values of x (again by the ratio test).

5.2.3 Lemma
∞
an y n
P
Suppose that for some 0 6= y ∈ R, converges. If x∈R with |x| < |y| then
n=0
∞
an x n
P
converges absolutely.
n=0

Proof
∞
an y n , (an y n ) → 0 (|an y n |) → 0
P
Since by 5.1.2vi as n→∞ and so as n → ∞.
n=0

Therefore (|an y n |) is bounded above and there exists 0 < K ∈ R such that

75
|an y n | < K for all n.

Then
K
|an | <
|y n |
for all n and  n
n K|xn | x
|an x | < = K  
|y n | y
for all n.
∞ ∞
|an xn | an x n
P P
Therefore, by the comparison test converges, i.e. converges
n=0 n=0
absolutely.

5.2.4 Theorem - existence of radius of convergence

∞
an x n
P
For a given power series one of the following three possibilities holds:
n=0

i. the power series converges only when x = 0.

ii. there exists R > 0 such that the power series converges (absolutely) if |x| < R
and diverges if |x| > R.

iii. the power series converges (absolutely) for all x ∈ R.

The value R in ii. is said to be the radius of convergence of the power series.

In case i. we say the radius of convergence is 0 and in case iii. we say it is

innity.

(Note that, in ii., we can't say anything about convergence when |x| = R.)

Proof
∞
an x n
P
Consider the set A = {|x| |x ∈ R and converges absolutely}.
n=0

0∈A and so A is not empty.

Suppose A is not bounded above. Let x ∈ R.

|x| is not an upper bound for A
Then
∞
an y n
P
and so there exists y ∈ R with |y| > x such that |y| ∈ A. This means that
n=0

76
 ∞
an x n
P
converges (absolutely) and so by 5.2.3 converges absolutely. Since x∈R
n=0
was arbitrary we are in case iii.

Suppose A is bounded. Let L = sup A. If L = 0 then A = {0} and we are in case i.

Suppose L>0 (clearly we can't have L < 0). Let x∈R with |x| < L.

Then there exists y∈R with |y| > |x| and |y| ∈ A.
∞ ∞
an y n an x n
P P
This means that converges (absolutely) and so by lemma con-
n=0 n=0
verges absolutely.

∞
an x n
P
Now suppose x ∈ R with |x| > L. Then must diverge. Otherwise,
n=0
|x| + L ∞
an c n
P
by 5.2.3, with c= the series converges absolutely and c∈A with
2 n=0
c > L = sup A.

Hence we are in case ii with R = L.

5.2.5 Remark
Note that in case ii. above we can dene a function f : (−R, R) → R by f (x) =
∞
an xn and in case iii above we can dene a function f : R → R by
P
f (x) =
n=0
∞
an x n .
P
n=0

5.3 Dierentiability of functions dened by power series

We'll now show that functions dened by power series are innitely dierentiable.

We'll do this in two stages.

∞ ∞
an x n nan xn−1
P P
Firstly we'll show that the power series and have the same
n=0 n=1
radius of convergence.

This means that they both dene functions on X

X = (−R, R) (where
where
the radius of convergence R is a positive real number) or X = R (where the radius
of convergence is ∞).

77
 ∞
an x n
P
We'll then have f : X → R given by f (x) = and g : X → R given
n=0
∞
nan xn−1 .
P
by g(x) =
n=1

We'll then go on to show that f is dierentiable at all c∈X and f 0 (c) = g(c).

We'll need the following lemma for this.

5.3.1 Lemma
0<λ<1
If then lim nλn = 0.
n→∞
Proof

Bernoulli's inequality says (1 + x)n ≥ 1 + nx for x ≥ −1, n ∈ N.

n (1 + x)n
Therefore (1 + x) > nx and so if x>0 then n< .
x
It follows, since 0 < λ < 1, that

[λ(1 + x)]n
nλn <
x
for all x>0 and n ∈ N.
1 1
Choose x > 0 λ(1 + x) < 1 i.e. 1 + x <
so that or 0 < x < − 1. For
  λ λ
1 1
example take x = −1 .
2 λ
[λ(1 + x)]n [λ(1 + x)]n
n
Then 0 < nλ < for all n ∈ N and lim = 0 so lim nλn =
x n→∞ x n→∞
0.

5.3.2 Theorem
∞ ∞
an x n nan xn−1
P P
The power series and have the same radius of convergence.
n=0 n=1

Proof
∞
an x n R0
P
Let R be the radius of convergence of and let be the radius of conver-
n=0
∞
nan xn−1 .
P
gence of
n=1

78
 ∞
0 6= R0 ∈ R. x1 ∈ R |x1 | < R0 n|an ||x1 |n−1
P
Suppose with so that converges.
n=0
∞ ∞
n|an ||x1 |n = |x1 |
n|an ||x1 |n−1 converges so, since |an ||x1 |n ≤ n|an ||x1 |n
P P
Then
n=0 n=0
∞
|an ||x1 |n converges by the comparison test 5.1.2vi.
P
for all n ≥ 1 we have that
n=0
Therefore |x1 | ≤ R.

So x1 ∈ R and |x1 | < R0 implies |x1 | < R and we must have that R0 ≤ R.
0
This also holds if R = 0.

The above argument also shows that if R0 = ∞ then R = ∞.

Now suppose 0 6= R ∈ R. Let x2 ∈ R with |x2 | < R.

∞
n|an ||x2 |n−1
P
We want to show that converges.
n=1

Pick some y∈R with |x2 | < y < R. Then

 n−1
n  x2 
|nan xn−1
2 | = |a n y n
|   .
yy
 n−1  n−1
 x2   x2 
Since n 
y
 →0 and n→∞ there exists K>0 such that n   ≤K for
y
all n ∈ N .

K ∞
|nan xn−1 |an y n | n|an ||x2 |n−1
P
Therefore 2 | ≤ and converges by the compari-
y n=1
son test.

Therefore |x2 | < R implies |x2 | ≤ R. This means that R ≤ R0 This also holds for
R = 0.

The same argument also shows that if R=∞ then R0 = ∞.

So we have

1. if either of R or R0 are equal to ∞ then both of them are, and so R = R0 .

2. if neither of R or R0 are ∞ then both R ≤ R0 and R0 ≤ R.
Hence R = R0 .

79
5.3.3 Remark
Note that the values of x for which these two power series converge need not be
∞ xn ∞
xn−1 does not.
P P
the same, converges when x = −1 but
n=1 n n=1
∞ ∞ ∞
an x n , nan xn−1 and n(n − 1)an xn−2
P P P
Note also that 5.3.2 also implies that
n=0 n=0 n=0
all have the same radius of convergence.

5.3.4 Theorem
∞
an xn
P
Let have radius of convergence R and suppose R 6= 0. The sum func-
n=0
∞
an x n
P
tion s : (−R, R) → R given by s(x) = is dierentiable on (−R, R) and
n=0
∞
s0 (x) = nan xn−1 .
P
n=1

Proof

Let c ∈ (−R, R). We will show s is dierentiable at c and calculate s0 (c).

We have to consider

∞
s(x) − s(c) X
− nan cn−1
x−c n=1

for x near to c.
∞
(n − 1)nan rn−2
P
Choose r so that |c| < r < R so that by 5.3.2. Consider x
n=2
with 0 < |x − c| < r − |c| so that |x| < x 6= c. For each n ≥ 1
r and

n n x−c if n=1
x −c =
(x − c)(xn−1 + xn−2 c + xn−3 c2 + ... + xcn−2 + cn−1 ) if n≥2

This means that

x n − cn
− ncn−1
x−c

0 if n=1
=
(x − c)(xn−2 + 2xn−3 c + 3xn−4 c2 + ... + (n − 2)xcn−3 + (n − 1)cn−2 if n≥2

80
 Multiplying by an and summing (terms for a0 cancel)

 ∞

 s(x) − s(c) X 
− nan cn−1 
 
 x−c


n=1

≤ |x − c| |a2 | + r(1 + 2)|a3 | + r2 (1 + 2 + 3)|a4 | + r3 (1 + 2 + 3 + 4)|a5 | + ....

 

since |x| < r and |c| < r

       
2×3 2 3×4 3 4×5
≤ |x − c| |a2 | + r |a3 | + r |a4 | + r |a5 | + .... .
2 2 2

Let

    ∞
2×3 2 3×4 X n(n − 1)
M = |a2 | + r |a3 | + r |a4 | + ... = |an |rn−2 .
2 2 n=2
2
n  o
If >0 then 0 < |x − c| < min r − |c|, implies that
M
 ∞

 s(x) − s(c) X 
n−1 
− nan c  < .

 x−c


n=1

∞
s0 (c) = nan cn−1 .
P
Hence s is dierentiable at c and
n=1

5.3.5 Corollary
∞
an x n
P
Let be a power series with radius of convergence R 6= 0. Suppose 0<
n=0
∞
an x n .
P
r<R and let f : (−r, r) → R be given by f (x) = Then f is innitely
n=0
∞
(−r, r) and for any k ∈ N, f (k) (x) = n(n−1)...(n−k+1)an xn−k
P
dierentiable on
n=k
f (n) (0)
for all x ∈ (−r, r). We have an = . Therefore if f (x) is the sum of a power
n!
series on some interval (−r, r) then this power series is unique and this power series
∞ f (n) (0)
xn .
P
is
n=0 n!

81
5.4 Power series with arbitrary centre
∞
an (x − x0 )n
P
Let x0 ∈ R. The series is a power series centred at x0 .
n=0

By putting y = x − x0 and using our previous results we get a radius of con-

vergence R so that

∞
an (x − x0 )n
P
i. if R = 0, converges only for x = x0
n=0

∞
an (x − x0 )n
P
ii. if R = ∞, converges (absolutely) for all x ∈ R.
n=0

∞
an (x − x0 )n
P
iii. if R ∈ R with R > 0 then |x − x0 | < R =⇒ converges
n=0
∞
an (x − x0 )n
P
absolutely and |x − x0 | > R =⇒ diverges.
n=0

∞
an x n
P
The sum function s(x) = is dierentiable on (x0 − R, x0 + R) and
n=0
∞
s0 (x) = nan (x − x0 )n−1
P
for all x ∈ (x0 − R, x0 + R). Suppose 0 < r < R
n=1
∞
an (x − x0 )n .
P
and let f : (x0 − r, x0 + r) → R be given by f (x) = Then
n=0
f is innitely dierentiable on (x0 − r, x0 + r) and for any k ∈ N, f (k) (x) =
∞
n(n − 1)...(n − k + 1)an (x − x0 )n−k for all x ∈ (x0 − r, x0 + r). We have
P
n=k
f (n) (x0 )
an = . Therefore f (x) is the sum of only one power centred at x0 and this
n!
∞ f (n) (x )
0
(x − x0 )n .
P
unique power series is
n=0 n!
We give such power series a special name.

5.4.1 Denition
Let x0 ∈ R. Let r>0 and let f : (x0 − r, x0 + r) → R
f : R → R be innitely or
∞ f (n) (x )
0
(x − x0 )n
P
dierentiable (in C infty) on (x0 − r, x0 + r) or R. The series
n=0 n!
is called the Taylor series of f centred at x0 .

82
Here f (0) = f .

For which x does the Taylor series converge? When it converges at x is the sum
f (x)? We concentrate on the case x0 = 0.

5.4.2 Theorem
Let f : (−r, r) → R be innitely dierentiable (in C infty). Dene Rn : (−r, r) →
R by
f 00 (0) 2 f (n)(x0 ) n
 
0
Rn (x) = f (x) − f (0) + f (0)x + x + ... x .
2! n!
Note that corollary 4.6.4 gives two expressions for Rn (x).
∞ f (j) (0)
xj
P
Then f (x) = if and only if Rn (x) → 0 and n → ∞.
j=0 j!
Proof
∞ f (j) (0) n f (j) (0)
xj . Then xj → f (x) as n → ∞. There-
P P
Suppose f (x) =
j=0 j! j=0 j!
n f (j) (0)
xj → 0 as n → ∞. In other words Rn (x) → 0 as n → ∞.
P
fore f (x) −
j=0 j!
n f (j) (0)
xj → 0 as n → ∞
P
Suppose Rn (x) → 0 and n → ∞. Then f (x) −
j=0 j!
n f (j) (0) ∞ f (j) (0)
xj → f (x) as n → ∞. In other words f (x) = xj .
P P
and
j=0 j! j=0 j!

5.4.3 Example
i. Let f : R → R be given by f (x) = sin(x). Then f (2n) (0) = 0 and f (2n+1) (0) =
∞ x2n+1
(−1)n so the Taylor series of f is (−1)n
P
.
n=0 (2n + 1)!
f (n+1) (c) n+1
Let x ∈ R. Then Rn (x) = x for some c between 0 and x.
(n + 1)!
|x|n+1
Therefore |Rn (x)| = → 0 as n → ∞.
(n + 1)!
∞ x2n+1
(−1)n
P
Therefore sin(x) = for all x ∈ R.
n=0 (2n + 1)!
ii. Let f :R→R be given by f (x) = cos(x). This is similar.

83
5.5 The exponential function
5.5.1 Denition
∞
x2 x3 X xn
Dene exp : R → R by exp(x) = 1 + x + + + .... = .
2! 3! n=0
n!
The ratio test shows that the radius of convergence of this power series is in-
nity.

∞ nxn−1
exp0 =
P
Then exp is dierentiable on R by 5.3.4 and = exp(x). Also
n=1 n!
exp(0) = 1.

5.5.2 Theorem
There is only one function f : R → R which is dierentiable on R and satises
f 0 (x) = f (x) for all x ∈ R and f (0) = 1.

Proof
First note that f (x) 6= 0 for all x ∈ R because, dening F : R → R by
F (x) = f (x)f (−x) gives that F 0 (x) = f 0 (x)f (−x) − f (x)f 0 (−x) = f (x)f (−x) −
f (x)f (−x) = 0.

Hence F is a constant function by 4.4.8.

Therefore f (x)f (−x) = f (0)f (−0) = 1 for all x ∈ R.

1
Note that, this means that f (x) 6= 0 (and f (−x) = ) for all x ∈ R.
f (x)
Now let g:R→R g 0 = g and g(0) = 1.
also satisfy

   
f f f (x)
By the above we can dene : R → R by (x) = for all x ∈ R.
g g g(x)
 0
g f (x)g 0 (x) − g(x)f 0 (x)
Then (x) = = 0 for all x ∈ R and so is constant
f [f (x)]2
by 4.4.8. Therefore
g(x) g(0)
= =1 for all x ∈ R.
f (x) f (0)
So g = f.

84
Note that in the rst part of the proof we showed that exp(x) 6= 0 and exp(−x) =
1
for all x ∈ R.
exp(x)

5.5.3 Theorem
i. exp : R → R is innitely dierentiable and exp(n) (x) = exp(x) for all x ∈ R,
for all n ≥ 1;

ii. exp(x) > 0 for all x∈R and exp is strictly increasing;

iii. exp(x + y) = exp(x) exp(y) for all x, y ∈ R.

Proof
i. By induction on n. It is true when n = 1.

If exp
is dierentiable n-times and exp(n) = exp then, since exp is dier-
(n)
entiable exp is dierentiable.

So exp(n+1) exists and exp(n+1) = exp0 = exp.

Hence the result is true by induction.

ii. From the proof of 5.5.2 we know that exp(x) 6= 0 for all x ∈ R.

Since exp(0) = 1 and since exp is continuous, by the IVT exp(x) > 0 for all
x∈R (otherwise there would be c∈R such that exp(c) = 0.

Therefore exp0 (x) > 0 for all x∈R and exp is strictly increasing by 4.4.9.

exp(x + y)
iii. Fix y∈R and consider g:R→R given by g(x) = for all x ∈ R.
exp(y)
exp(x + y)
Then, by the chain rule g 0 (x) = = g(x) and g(0) = 1.
exp(y)
Hence by 5.5.2 g = exp i.e. g(x) = exp(x) for all x ∈ R, that is exp(x + y) =
exp(x) exp(y).

5.5.4 Corollary
exp(nx) = (exp(x))n for all n ∈ N, x ∈ R.

Proof

85
By induction on 5.5.4 iii. It is true when n = 1. Suppose exp(nx) = (exp(x))n .

Then exp((n + 1)x) = exp(nx + x) = exp(nx) exp(x) = (exp(x))n exp(x) =

(exp(x))n+1 .

5.5.5 Remark
5.5.4 means that if we let e = exp(1) then exp(n) = en for all n ∈ N.
1
Since exp(−n) = = e−n we get that exp(n) = e n
for all n ∈ Z.
exp(n)
If p, q ∈ Z and q 6= 0, then by 5.5.3 again

 q  
p pq
exp = exp = exp(p) = ep .
q q
 
p p
Therefore exp = eq .
q
Because of this we can write ex instead ofexp(x) for all x ∈ R on the under-
standing that this denition agrees with exp(x) when x is rational and denes ex
as exp(x) when x is irrational.

By 5.5.3 iii ex+y = ex × ey .

Since exp is strictly increasing we have e0 < e1 and so 1 < e. Therefore e−n → 0
as n → ∞ and en → ∞ as n → ∞. Hence, by the IVT {ex | x ∈ R} = (0, ∞).

The graph is as follows.

86
5.6 The natural logarithm
We know that exp : R → R is strictly increasing and {ex | x ∈ R} = (0, ∞).

5.6.1 Denition
Let ln : (0, ∞) → R be the inverse function of exp, i.e. ln = exp−1 .

Hence ln(exp(x)) = x for all x∈R and exp(ln(y)) = y for all y ∈ (0, ∞).

5.6.2 Theorem
The function ln : (0, ∞) → R has the following properties:

i. ln(1) = 0
1
ii. ln is strictly increasing and dierentiable on (0, ∞) and ln0 (y) = for all
y
y>0

iii. ln(yz) = ln(y) + ln(z) for all y, z ∈ (0, ∞)

87
  
1
iv. ln = − ln(y) for all y>0
y
v. ln(y c ) = c ln(y) for all y>0 and c ∈ Q.

Proof

i. Since exp(1) = 0, ln(1) = 0

ii. Suppose 0 < x < y . If ln(x) ≥ ln(y) then, by applying exp (which is strictly
increasing) we get x ≥ y . Therefore ln(x) < ln(y) and ln is strictly increasing.

For each n ∈ N we can apply 4.5.1 to exp |[−n,n] to get that ln |[e−n ,en ] is
strictly increasing and continuous on [e−n , en ] and, since exp0 (x) 6= 0 for all
x∈ R, ln |[e−n ,en ] is dierentiable on (e−n , en ).

Because this is true for all n ∈ N we have that ln is dierentiable (and

therefore continuous) on (0, ∞).
1 1 1
Also by 4.5.1 and ln0 (y) = = = for all y > 0.
exp0 (ln(y)) exp(ln(y)) y
iii. See assignment 9.

iv. See assignment 9.

v. See assignment 9.

5.7 Real number powers and the binomial series

5.7.1 Denition
If a>0 and x∈R dene ax by ax = ex ln(a) . Note that when x ∈ Q, ax agrees
with the usual denition by 5.6.2 v. This is because, if p, q ∈ Z and q 6= 0,
p p
ln(a)
p
ln(a q )
√
q
a =e
q q =e = ap .

5.7.2 Theorem
Let a > 0, x, y ∈ R. Then

i. ax > 0

ii. ax+y = ax ay

iii. a0 = 1

88
 iv. (ax )y = axy .

Proof
Exercise (assignment 9).

5.7.3 Theorem
i. Let a > 0 and dene f : R → R by f (x) = ax . Then f is dierentiable on R
0 x
and f (x) = (ln(a))a .

ii. Let b ∈ R and dene g : (0, ∞) → (0, ∞) by g(x) = xb . Then g is dieren-

0 b−1
tiable on (0, ∞) and g (x) = bx .

iii. Dene h : (0, ∞) → (0, ∞) by h(x) = xx . Then h is dierentiable on (0, ∞)

0 x
and h (x) = x (ln(x) + 1).

Proof

i. f (x) = ex ln(a) and so f is dierentiable at every x∈R by the chain rule and
f 0 (x) = ln(a)ex ln(a) = (ln(a))ax .

ii. g(x) = eb ln(x) and sog is dierentiable at every x > 0 by the chain rule and
b
g 0 (x) = eb ln(x) = xb x−1 b = bxb−1 by 5.6.4ii.
x
iii. h(x) = ex ln(x) so h is dierentiable at every x > 0 by the chain rule and
h0 (x) = ex ln(x) (ln(x) + 1) = xx (ln(x) + 1).

89
5.7.4 Remark
We know that for all u, v ∈ R and n∈N
n  
X n k n−k
n
(u + v) = u v
k=0
k

n

n!
where
k
= .
k!(n − k)!
n
n

k
x ∈ R, n ∈ N, (1 + x)n =
P
And if
k
x . Now let a ∈ R and consider
k=0
f : (−1, ∞) → R dened by f (x) = (1 + x)a , i.e. f (x) = ea ln(1+x) .

We calculate its Taylor series about 0.

f 0 (x) = a(1+x)a−1 , f 00 (x) = a(a−1)(1+x)a−2 , f 000 = a(a−1)(a−2)xa−3 and so on.

This gives f (0) = 1, f 0 (0) = a, f 00 (0) = a(a − 1), f 000 (0) = a(a − 1)(a − 2)
and so the Taylor series of f about 0 is

a(a − 1) 2 a(a − 1)(a − 2) 3 a(a − 1).....(a − n + 1) n

1 + ax + x + x + ... + x + ...
2! 3! n!

∞ a(a − 1)...(a − n + 1)
a a
P
n

which we write at
n
x where
n
= for n > 0 and
n=0 n!
a


0
= 1.
0 0



Note that the case a=0 gives
0
=1 and
n
=0 for all n ≥ 1.
a


Also, if a∈N then
n
=0 for all n ≥ a + 1.

5.8 Theorem
∞
a

n
f (x) = (1 + x)a ,
P
The Taylor series g(x) = n
x for f : (−1, ∞) → R given by
n=0
where 0 6= a ∈ R and a 6∈ N has radius of convergence 1 and f (x) = g(x) when
|x| < 1.

Proof
a


Note that if a 6= 0 and a 6∈ N then
n
6= 0 for all n ≥ 0.

90
The ratio test gives
 
a


xn+1 
 

n+1
 a(a − 1)...(a − n) n! 

=   |x|
a


xn  (n + 1)! a(a − 1)..(a − n + 1) 
 
 n
   
a − n  a n 
=
  |x| = 
 − |x| → |x|
n + 1 n + 1 n + 1
as n → ∞.

Therefore the radius of convergence is 1.

∞
a


xn .
P
So g : (−1, 1) → R is dened by g(x) = n
n=0
∞
a
g 0 (x) =


xn−1 .
P
By theorem 5.3.4 n n
n=1

Since     
a a a
(n + 1) +n =a
n+1 n n
we get (1 + x)g 0 (x) = ag(x).
g
Now consider : (−1, 1) → R.
f
 0
g f (x)g 0 (x) − g(x)f 0 (x)
(x) =
f [f (x)]2

(1 + x)a g 0 (x) − g(x)a(1 + x)a−1 (1 + x)g 0 (x) − ag(x)

= = = 0.
(1 + x)2a (1 + x)a+1
 
g(x) g(0) a
Therefore is constant with value = = 1. So g(x) = (1 + x)a for all
f (x) f (0) 0
x ∈ (−1, 1).

5.9 A function which isn't the sum of its Taylor series

5.9.1 Theorem
For every n∈N

i. lim x−n ex = ∞
x→∞

91
 ii. lim xn e−x = 0
x→∞

1
iii. lim x−n e− x2 = 0.
x→0

Proof
x2 x3 xn+1
i. ex = 1 + x + + + ... + + ... and so
2! 3! (n + 1)!
x−n+2 x x
x−n ex = x−n + x−n+1 + + ... + + ... ≥ .
2! (n + 1)! (n + 1)!
x
Therefore lim x−n ex = ∞ since lim = ∞.
x→∞ x→∞ (n + 1)!

ii. As above, we also have, if x > 0,

1 (n + 1)!
0 < xn e−x = ≤ →0
x−n ex x
as x → ∞.
Therefore lim xn e−x = 0.
x→∞

1 1 1 1 1
iii. If x 6= 0 then e x2 = 1 + 2
+ 4
+ 6
+ ... + + ....
x 2!x 3!x (n + 1)!x2(n+1)
1 1
Therefore e x2 > for all n ∈ N.
(n + 1)!x2(n+1)
If x>0 this gives
1
0 < e− x2 < (n + 1)!x2(n+1)
or
1
0 < x−n e− x2 < (n + 1)!xn+2 .
1
Since (n + 1)!xn+2 → 0 as x → 0 lim x−n e− x2 = 0.
x→0

5.9.2 An innitely dierentiable function which is only the sum of its

Taylor series at 0 at 0
Consider the function f :R→R dened by


0 if x=0
f (x) = 1
e− x2 if x 6= 0

92
 We shall show that f is innitely dierentiable on R and f (n) (0) = 0 for all
n ≥ 1.
∞ f (n) (0)
xn is 0 + 0 + ....
P
Hence the Taylor series at 0, so converges to 0 for
n=0 n!
x∈R
all and is only equal to f (x) when x = 0.
Proof

We'll start by showing that for n ≥ 1, f (n) (0) = 0 and for x 6= 0

1
ha a3n−2 an+2 i
3n
f (n) (x) = e− x2 3n + 3n−2 + ... + n+2
x x x
for some real numbers a3n , a3n−2 , ...an+2 .

We'll do this by induction on n.

2 − 12
For n = 1, if x 6= 0 then f 0 (x) = e x .
x3
1

0 e− x2
And f (0) = lim =0 by 5.9.1iv. So the result is true when n = 1.
x→0 x

Assuming the result is true for n=k we prove the result for n = k + 1.

If x 6= 0 then

2 − 12 h a3k a3k−2 ak+2 i

f (k+1) (x) = e x + + ... +
x3 x3k x3k−2 xk+2
 
− 1 −3ka3k (3k − 2)a3k−2 (k + 2)ak+2
+e x2 − − ... −
x3k+1 x3k−1 xk+3
 
− 1 2a3k 2a3k−2 − 3ka3k (k + 2)ak+2
=e x2 + + ... − .
x3(k+1) x3k+1 xk+3

93
 (k+1) f (k) (x) − 0
Therefore f (0) = lim =0 by 5.9.1iv.
x→0 x
x2 00 x3
Hence the Taylor series of f at 0, f (0) + xf 0 (0) + f (0) + f 000 (0) + ... is
2! 3!
0 + 0 + ..... So the sum is 0 for all x∈R and is only equal to f (x) when x = 0.

94