Uniform Continuity

P. Sam Johnson

July 13, 2020

P. Sam Johnson Uniform Continuity 1

Overview

The following are well-known concepts in analysis.

uniform continuity ;
Lipschitz ;
contraction.

These are three more concepts of global continuity, all of them stronger
than continuity and each one stronger than the one before it.

In the lecture, we discuss uniform continuity, Lipschitz and contraction.

Michael-147

P. Sam Johnson Uniform Continuity 2

 Part - 1

P. Sam Johnson Uniform Continuity 3

Continuity

Let us recall the definition of continuity between metric spaces.

Let (X , d) and (Y , ρ) be metric spaces.

Definition 1.
A function f : X → Y is said to be continuous at a point c ∈ X if for
every ε > 0, there exists δ > 0 such that x ∈ X ,

d(x, c) < δ implies ρ(f (x), f (c)) < ε.

Note that the choice of δ depends on ε and on the point c in X . If this

point c ∈ X is changed and the same ε > 0 is given, generally δ > 0 can
be different from the original one.

Let A be a non-empty subset of X . If f is continuous at every point in the

set A, then we say that f is continuous on A.

Karunakaran-5-25
P. Sam Johnson Uniform Continuity 4
Examples

The set of real numbers R equipped with the metric of absolute distance
d(x, y ) = |x − y | defines the standard metric space of real numbers R.
We now see some examples of continuous functions f : A ⊆ R → R.
Example 2.
To show that f (x) = 3x + 1 is continuous at an arbitrary point c ∈ R, we
must argue that |f (x) − f (c)| can be made arbitrarily small for values of x
near c. Now, |f (x) − f (c)| = |(3x + 1) − (3c + 1)| = 3|x − c|, so, given
ε > 0, we choose δ = ε/3. Then, |x − c| < δ implies
ε
|f (x) − f (c)| = 3|x − c| < 3 = ε.
3
Of particular importance for this discussion is the fact that the choice of
δ is the same regardless of which point c ∈ R we are considering.

Stephen-116
P. Sam Johnson Uniform Continuity 5
Graph of f

Graph of f (x) = 3x + 1
y

10

x
−4 −2 2 4

−10

P. Sam Johnson Uniform Continuity 6

Examples
Example 3.
Let’s contrast the earlier example with what happens when we prove
g (x) = x 2 is continuous on R.
Given c ∈ R, we have

|g (x) − g (c)| = |x 2 − c 2 | = |x − c| |x + c|.

We need to get a bound for |x + c| that does not depend on x.

We notice that if |x − c| < 1, say, then c − 1 < x < c + 1, so |x| < |c| + 1
and hence |x + c| ≤ |x| + |c| < 2|c| + 1.
Thus we have
|g (x) − g (c)| < |x − c| (2|c| + 1)
provided |x − c| < 1.

Stephen-116

P. Sam Johnson Uniform Continuity 7

Examples
Example 3 (contd...).
Let ε > 0 be given.
ε
To arrange for |x − c| (2|c| + 1) < ε, it suffices to have |x − c| < 2|c|+1
and also |x − c| < 1. So we put
n ε o
δ(ε, c) = min 1, .
2|c| + 1

The work above shows that |x − c| < δ implies |g (x) − g (c)| < ε, as
desired.
Now, there is nothing deficient about this argument, but it is important to
notice that the algorithm for choosing δ depends on the value of c.

Stephen-116

P. Sam Johnson Uniform Continuity 8

 n o
1 1
Let us take ε = 1. Then δ = min 1, 2|c|+1 = 2|c|+1 .

g (x) for x ≥ 0
y

30
f (c3 )

20

f (c2 )
10

f (c1 )
x
c1 c2 c3
1 2 3 4 5 6 7 Stephen-116

P. Sam Johnson Uniform Continuity 9

Observations
Example 3 (contd...).
From the graph of the function, we may observe that if |x| is very large,
small increments in x produce large differences in the values of g taken at
these points.

So we guess that the function g does not produce a common δ > 0 on R.

Stephen-116, Bartle-137

P. Sam Johnson Uniform Continuity 10

Observations
Example 3 (contd...).
The statement
n ε o
δ(ε, c) = min 1, (1)
2|c| + 1

means that larger values of c are going to require smaller values of δ.

When c is increasing, the steepness of the graph of g (x) tells that a

much smaller δ is required. When c = 1, the choice δ = 1/3 is sufficient.
But when c = 10, a much smaller δ is required.

We note that the value of δ(ε, u) given in (1) certainly depends on the
point c ∈ R. If we wish to consider all c ∈ R, formula (1) does not lead to
one value δ(ε) > 0 that will “ work simultaneously” for all c ∈ R, since
inf{δ(ε, c) : c ∈ R} = 0.
Stephen-116, Bartle-137

P. Sam Johnson Uniform Continuity 11

Another Example
Example 4.
1
We now show that h(x) = x is continuous on (0, ∞).
25
y

20

15

10

x
0.5 1 1.5 2 2.5 3

P. Sam Johnson Uniform Continuity 12

Another Example

Given c ∈ (0, ∞), we have

1 1 |x − c|
|h(x) − h(c)| = − = [ since x > 0, c > 0 ].
x c x.c
1
We need to get a bound for x.c that does not depend on x.

We notice that if |x − c| < c2 , say, then − c2 < x − c < c2 , so c

2 <x < 3c
2
and hence x1 < c2 .

Thus we have
|h(x) − h(c)| < (2/c 2 ) |x − c|
provided |x − c| < c2 .

P. Sam Johnson Uniform Continuity 13

Another Example

Let ε > 0 be given. To arrange for (2/c 2 ) |x − c| < ε, it suffices to have

2
|x − c| < εc2 and also |x − c| < c2 . So we put
n c c 2ε o
δ(ε, c) = min , . (1)
2 2
2
Thus if |x − c| < δ(ε, c), then |h(x) − h(c)| < ( c22 )( c2ε ) = ε.

We have seen that the selection of δ(ε, c) by the formula (1) “works” in
the sense that it enables us to give a value of δ that will ensure that
|h(x) − h(c)| < ε when |x − c| < δ and x, c ∈ (0, ∞).

We note that the value of δ(ε, u) given in (1) certainly depends on the
point c ∈ (0, ∞). If we wish to consider all c ∈ (0, ∞), formula (1) does
not lead to one value δ(ε) > 0 that will “work” simultaneously for all
c > 0, since inf{δ(ε, c) : c > 0} = 0.

P. Sam Johnson Uniform Continuity 14

There are other selections that can be made for δ.

There are other selections that can be made for δ. We can start with
|x − c| < c3 ; then − c3 < x − c < c3 , so 2c 4c 1 3
3 < x < 3 and hence x < 2c .

Thus we have
|h(x) − h(c)| < (3/2c 2 ) |x − c|
provided |x − c| < c3 .

Therefore we could take

n c 2c 2 ε o
δ1 (ε, c) = min , .
3 3
However, we still have inf{δ1 (ε, c) : c > 0} = 0.

In fact, there is no way of choosing one value of δ that will “work” for all
c > 0 for the function h(x) = 1/x, as we shall see.

Bartle-137
P. Sam Johnson Uniform Continuity 15
Motivation

As c → 0, we see that h(c) goes to ‘infinity’. Thus, it is intuitively clear

that if we want to control the value of h(x) for x near to c, nearer c to 0,
smaller value for δ is required.

In other words, if we want to assert that h(x) is within ε-distance of h(c),

then we may have to restrict x to smaller and smaller open intervals
around c as c goes nearer and nearer to 0.

We have analyzed the behaviour of δ required to prove the continuity of f

at c varying in (0, ∞). It is observed that given ε > 0, we cannot make a
single choice of δ which will work for all c in (0, ∞), or, even in (0, 1).

Kumaresan-75

P. Sam Johnson Uniform Continuity 16

Motivation

Motivated by Examples 3 and 4, it turns out to be very useful to know

when the δ in the definition (of continuity) can be chosen to depend only
on ε > 0 and A, so that δ does not depend on the particular point c.

We shall give a special name for functions satisfying the above condition
by giving the following definition. Such functions are said to be uniformly
continuous on A.
Definition 5.
A function f : X → Y is said to be uniformly continuous on A ⊆ X if
for every ε > 0, there exists δ > 0 such that x, y ∈ A,

d(x, y ) < δ implies ρ(f (x), f (y )) < ε.

Karunakaran-5-25

P. Sam Johnson Uniform Continuity 17

Uniform Continuity

Note that while the continuity of f at a point is discussed, we are

concerned only with the values (behaviour) of the function near the point
under consideration.

But when we wish to say that a function is uniformly continuous on its

domain, we need to know the values of f on the entire domain. Hence the
continuity is known as a local concept whereas the uniform continuity is
known as a global concept.

Uniform continuity is defined on a set; unlike continuity, it has no local

counterpart. The best way to understand uniform continuity is to look at
some examples and see how the behaviour of the function on the entire
domain plays a role.

Kumaresan-76, Michael-147

P. Sam Johnson Uniform Continuity 18

Uniform Continuity

Uniform continuity is a property concerning a function and a set [on which

it is defined]. It makes no sense to speak of a function being uniformly
continuous at a point.

Uniform continuity is always discussed in reference to a particular

domain.

The criterion for uniform continuity is very like the epsilon-delta criterion
for continuity, but, in this case, for each ε > 0, there is a δ > 0 that serves
the purpose of the definition right across the subset.

Kumaresan-76, Ross-134

P. Sam Johnson Uniform Continuity 19

Caution !
Example 6.
√
Consider f (x) = x defined on A = {x ∈ R : x ≥ 0}. We give an ε-δ
proof for the fact that f is continuous on A.

Let ε > 0. We need to argue that |f (x) − f (c)| can be made less than ε
for all values of x in some δ neighborhood around c.
√
If c = 0, this reduces to the statement x < ε, which happens as long as
√
x < ε2 . Thus, if we choose δ = ε2 , we see that x < ε, which happens as
long as x < ε2 .

Thus, if we choose δ = ε2 , we see that |x − 0| < δ implies |f (x) − 0| < ε.

Stephen-112

P. Sam Johnson Uniform Continuity 20

Caution !
Example 6 (contd...).
√ √
For a point c ∈ A different from zero, we need to estimate | x − c|.
This time, write
√ √ 
√ √ √ √ x+ c |x − c| |x − c|
| x − c| = | x − c| √ √ =√ √ ≤ √ .
x+ c x+ c c
√
In order to make this quantity less than ε, it suffices to pick δ = ε c.
√ √ √
Then, |x − c| < δ implies | x − c| < ε√cc = ε, as desired.

The above example is included to keep us from jumping to the erroneous

conclusion that functions that are having δ depending of the point c
chosen in the set A, are not uniformly continuous. We shall prove later
√
that f (x) = x is uniformly continuous on A = {x ∈ R : x ≥ 0}.

Stephen-112
P. Sam Johnson Uniform Continuity 21
Uniform Continuity
Theorem 7.
Let X and Y be metric spaces. Let f : X → Y and A ⊆ X .
(i) If f is uniformly continuous on A, then f is continuous on A.
(ii) If f is uniformly continuous on X , then f is uniformly continuous on
A.

Uniform continuity is a strictly stronger property.

We shall see several examples of continuous functions which are not

necessarily uniformly continuous. Continuous functions that are not
uniformly continuous abound.

Michael-147

P. Sam Johnson Uniform Continuity 22

Uniform Continuity

The diameter of A is the extended real number defined by

(
sup{d(x, y ) : x, y ∈ A} if this quantity is finite
diam(A) :=
+∞ otherwise.

In the finite case, this is, by the definition of the supremum, the smallest
real number D such that any two points of A are at most a distance D
apart.
Exercise 8.
Let f : X → Y be a map of metric spaces. Then f is uniformly continuous
on A if and only if for every ε > 0, there is a δ > 0 such that whenever
E ⊆ A with diam(E ) < δ, we have diam(f (E )) < ε.

P. Sam Johnson Uniform Continuity 23

Criterion for Non-Uniform Continuity

It is useful to formulate a condition equivalent to saying that f is not

uniformly continuous on A. We give such criterion in the next result.

Theorem 9 (criterion for non-uniform continuity).

Let X and Y be metric spaces. Let f : X → Y and A be a non-empty
subset of X . The following statements are equivalent :
1. f is not uniformly continuous on A ;
2. there exists an ε0 > 0 such that for every δ > 0, no matter how small,
it is possible to find x, y ∈ A (depending on δ) with d(x, y ) < δ but
ρ(f (x), f (y )) ≥ ε0 ;

P. Sam Johnson Uniform Continuity 24

Non-Uniform Continuity - A Sequential Characterization
Theorem 10 (sequential criterion for non-uniform continuity).
Let X and Y be metric spaces and f : X → Y . Then f is not uniformly
continuous on A if and only if there exist an ε0 > 0 and two sequences
(xn ) and (yn ) in A satisfying

d(xn , yn ) → 0 but ρ(f (xn ), f (yn )) ≥ ε0 for all n ∈ N.

Proof : Apply Theorem 9 for δ = 1/n, for every n ∈ N to get the

sequences (xn ) and (yn ) in A.

P. Sam Johnson Uniform Continuity 25

Uniform Continuity - A Sequential Characterization
Theorem 11 (sequential criterion for uniform continuity).
Let X and Y be metric spaces and f : X → Y . Then f is uniformly
continuous on A ⊆ X iff for every sequences (xn ) and (yn ) in A satisfying
d(xn , yn ) → 0, we have ρ(f (xn ), f (yn )) → 0.

Proof. ⇒) : Suppose f : X → Y is uniformly continuous on A ⊆ X . Let

(xn ) and (yn ) be sequences in A such that d(xn , yn ) → 0. Let ε > 0 be
given.

Since f is uniformly continuous on A, there exists δ > 0 such that x, y ∈ A

with
d(x, y ) < δ implies ρ(f (x), f (y )) < ε.
As d(xn , yn ) → 0, there exists a natural number N0 ∈ N be such that
d(xn , yn ) < δ for all n ≥ N0 . Then, it follows that ρ(f (xn ), f (yn )) < ε for
all n ≥ N0 .
P. Sam Johnson Uniform Continuity 26
Proof (contd...)

⇐) : Suppose f is not uniformly continuous. By Theorem 9 (criterion for

non-uniform continuity), there exists ε0 > 0 such that for every δ > 0,
there exist x, y ∈ A (depending on δ) such that

d(x, y ) < δ, but ρ(f (x), f (y )) ≥ ε0 .

In particular, for every n ∈ N, there exist xn , yn ∈ A such that

d(xn , yn ) < n1 , but ρ(f (xn ), f (yn )) ≥ ε0 .

Thus, we have proved that, if f is not uniformly continuous, then there

exist sequences (xn ) and (yn ) in A satisfying

d(xn , yn ) → 0, but ρ(f (xn ), f (yn )) 6→ 0.

This completes the proof.

P. Sam Johnson Uniform Continuity 27

Example - Uniformly Continuous
Example 12.
We have noticed (in Example 2) that the function f (x) = 3x + 1 is
continuous on R.

Since
|f (x) − f (y )| = |(3x + 1) − (3y + 1)| = 3|x − y |,
so, given ε > 0, we choose δ = ε/3. Then, |x − c| < δ implies
ε
|f (x) − f (c)| = 3|x − c| < 3 = ε.
3
Hence the function f (x) = 3x + 1 is uniformly continuous on R.

P. Sam Johnson Uniform Continuity 28

Example - Not Uniformly Continuous
Example 13.
We have observed (in Example 3) that the function g (x) = x 2 could not
be uniformly continuous on R because larger values of x require smaller
and smaller values of δ.

By applying Theorem 10, for ε0 = 2, xn = n and yn = n + 1/n, the

function g (x) = x 2 is not uniformly continuous on R because

1
|xn − yn | = →0
n
and
 1 2 1
|g (xn ) − g (yn )| = n2 − n + = 2 + 2 ≥ 2 for all n ∈ N.
n n

Stephen-118

P. Sam Johnson Uniform Continuity 29

Example - Uniformly Continuous
Example 14.
However, that g (x) is uniformly continuous on the bounded set [−10, 10].
Notice that if we restrict our attention to the domain [−10, 10] then
|x + y | ≤ 20 for all x and y .

Given ε > 0, we can now choose δ = ε/20, and verify that if

x, y ∈ [−10, 10] satisfy |x − y | < δ, then
ε
|g (x) − g (y )| = |x 2 − y 2 | = |x − y kx + y | < 20 = ε.
20

In fact, it is not difficult to see how to modify this argument to show that
g (x) is uniformly continuous on any bounded subset A of R.

Stephen-118

P. Sam Johnson Uniform Continuity 30

Example - Uniformly Continuous
Example 15.
We have observed (in Example 4) that the function h(x) = x1 is continuous
1 1
on (0, ∞). By applying Theorem 10, for ε0 = 1, xn = √n+2 and yn = n+2 ,
1
the function h(x) = x is not uniformly continuous on (0, ∞) because

1 1
|xn − yn | = | √ − |→0
n+2 n+2
and √
|h(xn ) − h(yn )| = n + 2 − (n + 2) ≥ 1 for all n ∈ N.

However, that h(x) is uniformly continuous on any set of the form [a, ∞),
where a is a positive constant.

P. Sam Johnson Uniform Continuity 31

Examples - Uniformly Continuous
Example 15 (contd...).
Let ε > 0. We need to show that there exists δ > 0 such that
x ≥ a, y ≥ a and |x − y | < δ imply |h(x) − h(y )| < ε. (1)

We have
1 1 |x − y | |x − y |
|h(x) − h(y )| = − = ≤ .
x y |x| |y | a2
So we set δ = εa2 . It is now straightforward to verify (1).

P. Sam Johnson Uniform Continuity 32

Example - Uniformly Continuous
Example 16.
We now show that h(x) = x12 is uniformly continuous on any set of the
form [a, ∞) where a is a positive constant.

Let ε > 0. We need to show that there exists δ > 0 such that
x ≥ a, y ≥ a and |x − y | < δ imply |h(x) − h(y )| < ε. (1)
We have
(y − x)(y + x)
h(x) − h(y ) =
x 2y 2
y +x
If we can show that x 2y 2
is bounded on [a, ∞) by a constant M, then we
will take δ = Mε .

Ross-134

P. Sam Johnson Uniform Continuity 33

Examples - Uniformly Continuous
Example 16 (contd...).
But we have
y +x 1 1 1 1 2
2 2
= 2 + 2 ≤ 3 + 3 = 3,
x y x y xy a a a
3
so we set δ = εa2 . It is now straightforward to verify (1).
In fact, x ≥ a, y ≥ a and |x − y | < δ imply

|y − x| · |y + x| 2δ
|h(x) − h(y )| = 2 2
< 3 = ε.
x y a

We have shown that h(x) = x12 is uniformly continuous on [a, ∞) (where

a > 0 is fixed) since δ depends only on ε and the set [a, ∞).

Kumaresan-76

P. Sam Johnson Uniform Continuity 34

Using ε-δ, we prove non-uniform continuity of g (x) = x 2
on the set [0, ∞).

Using the ε-δ definition (criterion for non-uniform continuity), we now

prove that the function g (x) = x 2 (in Example 3) is not uniformly
continuous on the set [0, ∞).

We can choose ε0 = 1 and then for any δ > 0, we have

|g (x + δ) − g (x)| = (x + δ)2 − x 2 = 2xδ + δ 2

and we can choose x > 1/(2δ) so that 2xδ > 1.

Hence |g (x + δ) − g (x)| ≥ 1.

So there is no δ that works for every x in the infinite interval.

Thus g (x) = x 2 is not uniformly continuous on [0, ∞).

P. Sam Johnson Uniform Continuity 35

 1
Using ε-δ, we prove non-uniform continuity of h(x) = x2
on (0, ∞) or even on the set (0, 1).

Using the ε-δ definition (criterion for non-uniform continuity), we now

prove that the function h(x) = x12 (in Example 4) is not uniformly
continuous on the set (0, ∞) or even on the set (0, 1).

We will prove this by directly violating the definition of uniform continuity

for ε0 = 1. That is, for each δ > 0 there exist x, y in (0, 1) such that

|x − y | < δ and yet |h(x) − h(y )| ≥ 1. (1)

δ
To show (1) it suffices to take y = x + 2 and arrange for

δ

h(x) − h x + ≥ 1. (2)
2
The motivation for this calculation is to go from two unknowns, x and y ,
in (1) to one unknown, x, in (2). Ross-135

P. Sam Johnson Uniform Continuity 36

 1
Using ε-δ, we prove non-uniform continuity of h(x) = x2
on (0, ∞) or even on the set (0, 1).

(2) is equivalent to
δ δ
(x + 2 − x)(x + 2 + x) δ(2x + 2δ )
1≤ = . (3)
x 2 (x + 2δ )2 2x 2 (x + 2δ )2

It suffices to prove (1) for δ < 21 . To obtain (3), let us try x = δ. Then
5δ 2
δ(2δ + 2δ ) 2 5 5 20
δ 2
= 9δ 4
= 2
≥ 1 2 = > 1.
2δ 2 (δ + 2)
9δ 9( 2 ) 9
2

To summarize, we have shown that if 0 < δ < 12 , then

|h(δ) − h(δ + 2δ )| > 1, so (1) holds with x = δ and y = δ + 2δ .
1
Thus h(x) = x2
is not uniformly continuous on (0, 1).

Michael-148
P. Sam Johnson Uniform Continuity 37
Example of uniformly continuous functions

We shall now give some examples of uniformly continuous functions,

Example 17.
k
Consider f : [a, b] → R defined by f (x) = x−2 where a, b, k are positive
constants with a > 2. We now show that f is uniformly continuous on
[a, b]. Let ε > 0 be given. Now

k k k(y − x)
f (x) − f (y ) = − =
x −2 y −2 (x − 2)(y − 2)

Since a > 2, a = 2 + η for some η > 0 and if x, y ∈ [a, b], then

2
|x − 2| = x − 2 ≥ η and y − 2 ≥ η. Choose δ < ηk ε . Thus if |x − y | < δ
and x, y ∈ [a, b], then

k|x − y | kδ
|f (x) − f (y )| = ≤ 2 < ε.
(x − 2)(y − 2) η
Karunakaran 5-26
P. Sam Johnson Uniform Continuity 38
Example of uniformly continuous functions
Example 18.
Let f : (0, 6) → R defined by f (x) = x 2 + 2x − 5. We now show that f is
ε
uniformly continuous on (0, 6). For given ε > 0, choose δ = 14 . Now for
any x, y ∈ (0, 6) with |x − y | < δ, we have

|f (x) − f (y )| = |x − y ||x + y + 2| < 14δ < ε

(since 0 ≤ x, y < 6, we have |x + y + 2| < 14).

Karunakaran 5-26

P. Sam Johnson Uniform Continuity 39

Continuous function that is not uniformly continuous
Example 19.
The exponential function x 7→ e x defined on R is not uniformly continuous
on R. Indeed, there is no δ > 0 that guarantees

|e x − e y | < 1 for all x, y ∈ R with |x − y | < δ.

Specifically, for all x, y ∈ R with x < y , we have

(e y − e x )/(y − x) > e x .

Then, for any γ ∈ (0, 1), pick x = − ln γ and y = γ − ln γ, so that

e y − e x > (y − x)e x = 1.

Michael-148

P. Sam Johnson Uniform Continuity 40

Continuous functions that are not uniformly continuous
abound.
Exercises 20.
Prove the following statements :
1. No polynomial function of degree greater than 1 is uniformly
continuous on R.
2. The logarithmic function is not uniformly continuous on (0, ∞).

P. Sam Johnson Uniform Continuity 41

Example - Continuous on Bounded Domain - Not
Uniformly Continuous

The following example is included to keep us from jumping to the

erroneous conclusion that functions that are continuous on bounded
domains are necessarily uniformly continuous.
Example 21.
Consider h : (0, 1] → R defined by h(x) = x12 . This function is continuous
on (0, 1], but not uniformly continuous on (0, 1]. We apply Theorem 10.
To see this, take ε0 = 2 and consider the sequences ( √1n ) and ( n1 ).

We see that √1 − 1
→ 0, but h( √1n ) − h( n1 ) = |n − n2 | ≥ 1, for all
n n
n ∈ N.

P. Sam Johnson Uniform Continuity 42

Example - Continuous Bounded Function - Not Uniformly
Continuous
Example 22.
f (x) = sin( x1 ) is a bounded continuous function on (0, 1) but it is not
uniformly continuous on (0, 1).

sin( x1 )

0.5 1 1.5 2 2.5 3 x

The function f (x) := sin( x1 ) (for x > 0) changes its values rapidly near
“zero”.

Stephen-117
P. Sam Johnson Uniform Continuity 43
Example - Not Uniformly Continuous
Example 15 (contd...).
The problem arises near zero, where the increasingly rapid oscillations take
domain values that are quite close together to range values a distance 2
apart. To illustrate Theorem 10, take ε0 = 2 and set
1 1
xn = and yn = .
π/2 + 2nπ 3π/2 + 2nπ
Because each of these sequences tends to zero, we have |xn − yn | → 0, and
a short calculation reveals |f (xn ) − f (yn )| = 2 for all n ∈ N.

The above example reveals that bounded continuous functions can fail to
be uniformly continuous if they oscilate arbitrarily quickly.

Stephen-117

P. Sam Johnson Uniform Continuity 44

Example - Not Uniformly Continuous
Exercise 23.
g (x) = cos( x1 ) is a bounded continuous function on (0, 1) but it is not
uniformly continuous on (0, 1).

cos( x1 )

0.5 1 1.5 2 2.5 3 x

P. Sam Johnson Uniform Continuity 45

Continuous functions that are not uniformly continuous
abound.

Every uniformly continuous function on A is continuous on A. We have

seen some examples of functions which are not uniformly continuous.

Continuous function defined on a bounded set which is not uniformly

continuous. (Example 21);
Continuous bounded function which is not uniformly continuous.
(Example 22 and 23).

P. Sam Johnson Uniform Continuity 46

Continuity and uniform continuity are the same on
compact sets.

The importance of uniform continuity lies not so much in knowing which

functions are uniformly continuous as in knowing on which sets continuity
of a given function is uniform. The most useful theorem in this regard is
that every continuous function is uniformly continuous on all compact
subsets of its domain.

We shall prove that continuous on compact domains are necessarily

uniformly continuous. So, when the domain is compact, uniform continuity
and continuity are the same on the domain.

Michael-148

P. Sam Johnson Uniform Continuity 47

Compact Sets
Definition 24.
A subset K of a metric space (X , d) is said to be compact if every
sequence in K has a convergent subsequence that converges to a limit in
K.

Definition 25 (Equivalent definition).

By an open cover of a set A in a metric space X we mean a collection
{Gα } of open subsets of X such that E ⊆ Uα Gα .
A subset K of a metric space (X , d) is said to be compact if every open
cover of K contains a finite subcover.

Every finite set is compact.

Compact subsets of metric spaces are closed.
Closed subsets of compact sets are compact.
The Heine-Borel theorem states that a subset of Rn (with the usual
topology) is compact iff it is closed and bounded.

P. Sam Johnson Uniform Continuity 48

Eevery continuous function is uniformly continuous on all
compact subsets of its domain.
Theorem 26 (Uniform Continuity Theorem).
Let X and Y be metric spaces and let K be a compact subset of X . If
f : X → Y is continuous on K , then f is uniformly continuous on K .

Proof : Suppose f is not uniformly continuous on K . Then, by Theorem

10, there exists an ε0 > 0 and two sequences (xn ) and (yn ) in K satisfying
d(xn , yn ) → 0 but ρ(f (xn ), f (yn )) ≥ ε0 for all n ∈ N. (?).
Since K is compact, (xn ) and (yn ) have convergent subsequences, say (x̃n )
and (ỹn ), converging to x and y in K , respectively. That is,
d(x̃n , x) → 0 and d(ỹn , y ) → 0.

Also we have d(x̃n , ỹn ) → 0.

Ross-161
P. Sam Johnson Uniform Continuity 49
Proof (contd...)
By ε/3-argument, for any ε > 0, there are natural numbers N1 , N2 , N3
such that
ε ε ε
d(x̃n , x) < , for all n ≥ N1 ; d(ỹn , y ) < , for all n ≥ N2 ; d(x̃n , ỹn ) < , for all n ≥ N3 .
3 3 3

Let N0 = max{N1 , N2 , N3 }. Thus

ε ε ε
d(x, y ) ≤ d(x, xn ) + d(xn , yn ) + d(yn , y ) < + + = ε, for all n ≥ N0 .
3 3 3

Since the left-hand side is independent of n, d(x, y ) < ε, for every ε > 0.
Thus x = y .

Since both (x̃n ) and (ỹn ) converge to x and f is continous on K ,

ρ(f (x̃n ), f (ỹn )) → 0, this is a contradiction to (?). Ross-161

P. Sam Johnson Uniform Continuity 50

Another Proof

Proof : Let ε > 0 be given.

Since f is continuous on K , for every u ∈ K , there exists δu > 0 such that

x ∈ K and
ε
d(x, u) < δu implies ρ(f (x), f (u)) < . (1)
2

Since {BX u, δ2u : u ∈ K } is an open cover of K and K is compact, there

exist u1 , . . . , un in K such that

n δui
K ⊆ Ui=1 BX (ui , ). (2)
2
Let
1 n o
δ= min δu1 , δu2 , . . . , δun .
2
Ross-161

P. Sam Johnson Uniform Continuity 51

Another Proof (contd...)

Now, let x, y ∈ K be such that d(x, y ) < δ.

δ
Since x ∈ K , by (2), x ∈ BX (ui , 2ui ) for some i ∈ {1, 2, . . . , n}, so
δui
d(x, ui ) < 2 . Then we have

δui δu δu
d(y , ui ) ≤ d(y , x) + d(x, ui ) < δ + < i + i = δui .
2 2 2
Therefore, applying (1) twice, we have
ε ε
ρ(f (x), f (ui )) < and ρ(f (y ), f (ui )) < .
2 2
Hence,
ρ(f (x), f (y )) ≤ ρ(f (x), f (ui )) + ρ(f (ui ), f (y )) < ε
as desired. This completes the proof. Ross-161

An alternatre proof is sketched in the book by Rudin (page-99).

P. Sam Johnson Uniform Continuity 52
Compactness is essential.
Theorem 27.
Let E be a subset of R. If E is bounded but not closed, then there exists a
continuous function on E which is not uniformly continuous on E .

Proof : Since E is bounded but not closed, there exists a limit point a of
E which is not a point of E . Consider
1
f (x) = for x ∈ E . (1)
x −a
Obviously, f is continuous on E (being a quotient of two continuous
functions with denominator non-zero).

Since a is a limit point of E , given any real number M > 0 there exists
1
x0 ∈ E such that |x0 − a| < M or that |f (x0 )| > M. This shows that f is
unbounded on E .

Rudin-92, Karunakaran-5-27
P. Sam Johnson Uniform Continuity 53
Proof (contd...)

To see that (1) is not uniformly continuous, let ε > 0 and δ > 0 be
arbitrary, and choose a point x ∈ E such that |x − x0 | < δ. Taking y close
enough to x0 , we can then make the difference |f (x) − f (y )| greater than
ε, although |x − y | < δ. Since this is true for every δ > 0, f is not
uniformly continuous on E .

Indeed, using the definitions, for each ε > 0 and each δ > 0 we can choose
points p, q ∈ S such that
1
|p − a| < δ and |q − a| < min{δ − |p − a|, ε + }.
|p − a|

P. Sam Johnson Uniform Continuity 54

Proof (contd...)

Now

|p − q| ≤ |p − a| + |q − a| < δ

but
1 1
|f (p) − f (q)| = −
p−a q−a
p−q |p − a| − |q − a|
= ≥
(p − a)(q − a) |(p − a)(q − a)|
1 1
= −
|q − a| |p − a|
1 1
≥ε+ − =ε
|p − a| |p − a|

Hence f is not uniformly continuous. Rudin-92, Karunakaran-5-28

P. Sam Johnson Uniform Continuity 55

Boundedness is essential.

We may also note that if E is unbounded, it is possible to have a

continuous function, which is uniformly continuous.

Theorem 27 would be false if boundedness were omitted from the

hypothesis. There are unbounded subsets of R on which every function is
uniformly continuous. The following example illustrates that boundedness
of E alone is not sufficient to guarantee the existence of non-uniform
continuous function.
Example 28.
Let Z be the set of all integers. Then every function f : Z → R is not only
continuous but is uniformly continuous on Z.
Take any ε > 0. Then choose δ < 1. Then we prove
|x − y | < δ =⇒ |f (x) − f (y )| < ε. But |x − y | < δ means, there is
only one point in (y − δ, y + δ), so f (x) = f (y ) and hence the result
follows. Rudin-92, Karunakaran-5-29

P. Sam Johnson Uniform Continuity 56

Sine and Cosine Functions
Example 29.
The sine and cosine functions are uniformly continuous on the whole of R.
To see this, Theorem 26 ensures that they are both uniformly continuous
on the interval [−π, π].

Then 2π-periodicity of the functions clinches the matter: let ε > 0 and let
δ > 0 be such that, for all a, b ∈ [−π, π] with |b − a| < δ, we have
| sin b − sin a| < ε.

Then for all x, y ∈ R, there exist a, b ∈ [−π, π] with |a − b| ≤ |x − y | and

sin x = sin a and sin y = sin b, from which the result follows easily.

Michael-149

P. Sam Johnson Uniform Continuity 57

Exponential Function and All Polynomial Functions
Example 30.
In view of Theorem 26, the following functions are uniformly continuous
on the indicated sets:
x 73 on [−13, 13] ;
√
x on [0, 400] ;
x 17 sin(e x ) − e 4x cos 2x on [−8π, 8π] ;
1
x6
on [ 14 , 44], and so on.

Exercise 31.
The exponential function and all polynomial functions, being continuous
on R, are uniformly continuous on every bounded subset of R.

Michael-149, Ross-137

P. Sam Johnson Uniform Continuity 58

Tangent Function
Example 32.
The tangent function is continuous but not uniformly continuous on the
bounded interval (−π/2, π/2) of R.

Theorem 26 does not apply here because the function is not continuous on
any superset of (−π/2, π/2) which is compact.

For r ∈ (0, π/2), the tangent function is uniformly continuous on the

interval (−(π/2) + r , (π/2) − r ) because it is continuous on the closed
interval [−(π/2) + r , (π/2) − r ].

Michael-150

P. Sam Johnson Uniform Continuity 59

Continuous Functions
Example 33.
Suppose X and Y are metric spaces and f : X → Y . If X is a discrete
metric space (4.3.7), then f is continuous irrespective of the metric on Y .
This is so because every subset of a discrete metric space is open, so, for
each subset V of Y , f −1 (V ) is necessarily open in X . In particular, if N is
endowed with its usual metric inherited from R, or with the discrete
metric, or indeed with the inverse metric (m, n) 7→ |m−1 − n−1 |, then
every function from N into a metric space is continuous. In other words,
all sequences are continuous functions provided N is endowed with a
suitable metric.
There are metric spaces with metrics that may differ from the discrete
metric yet generate the same topology, namely the power set. Such spaces
are collectively called discrete metric spaces. Michael-131

P. Sam Johnson Uniform Continuity 60

Discrete Metric Spaces
Definition 34.
A metric space (X , d) is called a discrete metric space if its subsets are
open (and therefore also closed) in X .

Example 35.
Every finite metric space is a discrete space. N with its usual metric
inherited from R is a discrete metric space. N with the metric
(m, n) 7→ |m−1 − n−1 | is a discrete metric space

Exercise 36.
Show that Q is a countable metric space that is not a discrete metric
space.

P. Sam Johnson Uniform Continuity 61

Uniformly Continuous Functions

Suppose (X , d) and (Y , ρ) are metric spaces and f : X → Y . If d is the

discrete metric, then f is uniformly continuous irrespective of the metric
on Y . Specifically, for each ε > 0, we can choose δ = 1 so that whenever
x, y ∈ X with d(x, y ) < 1. We get that x = y , hence
0 = ρ(f (x), f (y )) < ε.

It is important in this example that the metric is discrete and the space
not merely a discrete space. There are continuous functions on discrete
metric spaces that are not uniformly continuous. The subspace
S = {1/n : n ∈ N} of R is a discrete metric space because each of its
singleton sets is both open and closed in S. The function 1/n 7→ n is
continuous, as it must be (by Example 33), because S is discrete, but it is
not uniformly continuous. Specifically, if δ > 0 and m, n ∈ N with
m > n > 2/δ, then |1/m − 1/n| < δ but |m − n| ≥ 1.
Michael-150

P. Sam Johnson Uniform Continuity 62

Uniformly Continuous Functions
Exercise 37.
Suppose (X , d) and (Y , ρ) are metric spaces, f , g : X → Y . If f and g are
uniformly continuous on A ⊆ X , then f + g is uniformly continuous on A.

Exercise 38 (Composition of uniformly continuous functions

are uniformly continuous).
Suppose (X , d), (Y , ρ) and (Z , γ) are metric spaces, f : X → Y and
g : Y → Z . If f and g are uniformly continuous on X and f (X ),
respectively, then g ◦ f is uniformly continuous on X .

Michael-157

P. Sam Johnson Uniform Continuity 63

Uniformly Continuous Functions
Exercise 39 (Product of uniformly continuous functions).
1. Show that if f and g are uniformly continuous on A ⊆ R and if they
are both bounded on A, then their product fg is uniformly continuous
on A.
[Hint : If M is a bound for both f and g on A, show that
|f (x)g (x) − f (u)g (u)| ≤ M|f (x) − f (u)| + M|g (x) − g (u)| for all
x, u ∈ A.]
2. If f (x) := x and g (x) := sin x, show that both f and g are uniformly
continuous on R, but that their product fg is not uniformly
continuous on R.

Bartle-144

P. Sam Johnson Uniform Continuity 64

Uniformly Continuous Functions
Exercises 40.
1. Show that f : [1, ∞) → R given by f (x) = 1/x is uniformly
continuous on [1, ∞).
2. Show that the function f : [1, ∞) → R given by
f (x) = 1/x n , for n ∈ N, is uniformly continuous on [1, ∞).
3. Show that f : (0, 1) → R given by g (x) = 1/x cannot be uniformly
continuous on (0, 1), using
(a) ε-δ criterion for non-uniform continuity ;
(b) sequential criterion for non-uniform continuity.

Kumaresan-77

P. Sam Johnson Uniform Continuity 65

Uniformly Continuous Functions
Exercises 41.
1. Show that the function f (x) = 1/x is uniformly continuous on the set
[a, ∞), where a is a positive constant.
[Hint : Since 1/x − 1/u = (u − x)/xu, it follows that
[1/x − 1/u] ≤ (1/a2 )|x − u| for x, u ∈ [a, ∞).]
2. Show that the function f (x) := 1/(1 + x 2 ) for x ∈ R is uniformly
continuous on R.
3. Prove that if f and g are each uniformly continuous on R, then the
composite function f ◦ g is uniformly continuous on R.
[Hint : Given ε > 0 there exists δf > 0 such that |y − v | < δf implies
|f (y ) − f (v )| < ε. Now choose δg > 0 so that |x − u| < δg implies
|g (x) − g (u)| < δf .]
4. If f is uniformly continuous on A ⊆ R, and |f (x)| ≥ k > 0 for all
x ∈ A, show that 1/f is uniformly continuous on A.

Bartle-144
P. Sam Johnson Uniform Continuity 66
Uniformly Continuous Functions
Exercises 42.
1. Prove that if f is uniformly continuous on a bounded subset A of R,
then f is bounded on A.
2. Show that the function f (x) := 1/x 2 is uniformly continuous on
A := [1, ∞), but that it is not uniformly continuous on B := (0, ∞).
3. Show that if f is continuous on [0, ∞) and uniformly continuous on
[a, ∞) for some positive constant a, then f is uniformly continuous on
√
[0, ∞). Using the result, one can show that f (x) = x is uniformly
continuous on [0, ∞).
4. Assume that g is defined on an open interval (a, c) and it is known to
be uniformly continuous on (a, b] and [b, c), where a < b < c. Prove
that g is uniformly continuous on (a, c).

P. Sam Johnson Uniform Continuity 67

Uniformly Continuous Functions
Exercises 43.
1. Let A ⊆ R and suppose that f : A → R has the following property:
for each ε > 0 there exists a function gε : A → R such that gε is
uniformly continuous on A and |f (x) − gε (x)| < ε for all x ∈ A.
Prove that f is uniformly continuous on A.
2. A function f : R → R is said to be periodic on R if there exists a
number p > 0 such that f (x + p) = f (x) for all x ∈ R. Prove that a
continuous periodic function on R is bounded and uniformly
continuous on R.
[Hint : Since f is bounded on [0, p], it follows that it is bounded on
R. Since f is continuous on J := [−1, p + 1], it is uniformly
continuous on J. Now show that this implies that f is uniformly
continuous on R.]

Bartle-144

P. Sam Johnson Uniform Continuity 68

 Part - 2

P. Sam Johnson Uniform Continuity 69

Conservation by Uniformly Continuous Functions

Uniformly continuous functions have some very nice conserving

properties. They map totally bounded sets onto totally bounded sets and
Cauchy sequences onto Cauchy sequences.

On the other hand, uniformly continuous functions need not preserve

boundedness.

Michael-150

P. Sam Johnson Uniform Continuity 70

Nice Properties of Uniformly Continuous Functions

We now discuss some nice properties of uniformly continuous functions.

Theorem 44.
Let f : (X , d) → (Y , ρ) be a map of metric spaces. If f is uniformly
continuous on A ⊆ X and (xn ) is a Cauchy sequence in A, then (f (xn )) is
a Cauchy sequence in Y .
Proof : Let (xn ) be a Cauchy sequence in A and let ε > 0. Since f is
uniformly continuous on A, there exists δ > 0 so that

x, y ∈ A and d(x, y ) < δ imply ρ(f (x), f (y )) < ε. (1)

Since (xn ) is a Cauchy sequence, there exists N0 ∈ N so that

d(xn , xm ) < δ for all m, n > N0 .

From (1) we see that m, n > N0 implies ρ(f (xn ), f (xm )) < ε. This
proves that (f (xn )) is a Cauchy sequence in Y .
P. Sam Johnson Uniform Continuity 71
Nice Properties of Uniformly Continuous Functions

The preceding result gives us an alternative way of seeing that f (x) := 1/x
is not uniformly continuous on (0, 1).

We note that the sequence given by xn := 1/n in (0, 1) is a Cauchy

sequence, but the image sequence, where f (xn ) = n, is not a Cauchy
sequence.

We shall now use the relationship between the concepts of the limit of a
function and that of a sequence. This relationship will be very useful in
proving the existence of limit of a function in many cases.
Theorem 45.
Let f : E → R where E ⊆ R be a function. Then
limx→a f (x) = ` (−∞ ≤ a ≤ ∞, −∞ ≤ ` ≤ ∞) if and only if for every
sequence {xn } in S\{a} with xn → a as n → ∞, we have f (xn ) → ` as
n → ∞.
P. Sam Johnson Uniform Continuity 72
Nice Properties of Uniformly Continuous Functions
Theorem 46.
Let f : E ⊆ R → R be uniformly continuous on E . If x0 is a limit point of
E , then f has a limit at x0 .

Proof : Let x0 be a limit point of S and let {xn } be any sequence in

S\{x0 } converging to x0 . In view of Theorem 45, it is sufficient to prove
that the sequence {f (xn )} is Cauchy (note that every Cauchy sequence of
real numbers is convergent) and that all these sequences converge to the
same limit. Since {xn } converges, it is Cauchy in E . As f is uniformly
continuous on E , {f (xn )} is Cauchy.

Further, if xn → x0 and yn → x0 , then the sequence {x1 , y1 , x2 , y2 . . .} also

converges to x0 and hence the sequence {f (x1 ), f (y1 ), f (x2 ), f (y2 ) . . .}
also converges. It now follows that {f (xn )} and {f (yn )} converge to the
same limit.

P. Sam Johnson Uniform Continuity 73

Nice Properties of Uniformly Continuous Functions
Example 47.
Observe that Theorem 46 gives a necessary condition for uniform
continuity but it is not a sufficient condition. For example, consider
g : R → R where g (x) = x 2 for all x ∈ R. Clearly g is continuous on R
and g has a limit at every limit point of R (note that R contains all its
limit points) but g is not uniformly continuous.
Karunakaran-5-29

P. Sam Johnson Uniform Continuity 74

Nice Properties of Uniformly Continuous Functions
Definition 48.
A function f : X → Y of metric spaces is said to be Cauchy continuous
if f carries every Cauchy sequence in X to a Cauchy sequence in Y .
Exercise 49.
Show that every Cauchy continuous function f : X → Y is continuous.
[Hint : For otherwise, we can find a point x0 , an ε0 > 0, and a sequence
(xn ) such that d(xn , x0 ) → 0 but ρ(f (xn ), f (x0 )) ≥ ε0 for all n. Take
yn = x0 if n is odd and take yn = xn if n is even. Then (yn ) is a Cauchy
sequence, but (f (yn )) is not.]
A Cauchy continuous function may not be uniformly continuous. For
example, g (x) = x 2 on R.
A continuous function may not be Cauchy continuous. For example,
f (x) = 1/x on (0, 1).
uniformly continuous ( Cauchy continuous ( continuous.
P. Sam Johnson Uniform Continuity 75
Totally Bounded
Definition 50.
Let (X , d) be a metric space and ε be an arbitrary positive number. Then
a subset A ⊆ X is said to be an ε-net for X if given any x ∈ X , there
exists a point y ∈ A such that d(x, y ) < ε.
A finite ε-net for X is an ε-net of X consisting of finite number of
elements of X .

Definition 51.
A subset A ⊆ X is said to be totally bounded if for every ε > 0, there
exists a finite ε-net of A.

Every totally bounded set is bounded.

A is totally bounded iff every sequence in A contains a Cauchy
subsequence. (Cauchy criterion for total boundedness)
Every compact metric space is totally bounded.
A metric space is compact iff it is complete and totally bounded.
P. Sam Johnson Uniform Continuity 76
A Nice Property of Uniformly Continuous
Theorem 52.
Suppose (X , d) and (Y , ρ) are metric spaces and f : X → Y is uniformly
continuous on X . Then f maps every totally bounded subset of X to a
totally bounded subset of Y .

Proof : Suppose S is a totally bounded subset of X . Suppose (yn ) is any

sequence in f (S). For each n ∈ N, the subset S ∩ f −1 ({yn }) of X is
non-empty.

We choose a sequence (xn ) with xn ∈ S ∩ f −1 ({yn }) for each n ∈ N. Then

f (xn ) = yn for each n ∈ N. By the Cauchy criterion for total boundedness
of S, (xn ) has a Cauchy subsequence (xnm ). Then, by what we have just
proved, (f (xnm )), that is, (ynm ), is a Cauchy subsequence of (yn ). Since
(yn ) is an arbitrary sequence in f (S), f (S) satisfies the Cauchy criterion
for total boundedness and so is totally bounded.

P. Sam Johnson Uniform Continuity 77

Example

Not all continuous functions are sending totally bounded sets to totally
bounded sets.
Example 53.
The function x 7→ 1/x defined on (0, 1) is continuous.
But it maps the totally bounded subset (0, 1) of R to the closed
unbounded subset (1, ∞) of R.

Example 54.
The tangent function maps the totally bounded interval (−π/2, π/2) of R
to the unbounded interval (−∞, ∞). It is not, as we already know,
uniformly continuous on (−π/2, π/2).

Michael-151

P. Sam Johnson Uniform Continuity 78

A Nice Property of Uniformly Continuous
Theorem 55.
If f is a continuous function from a compact metric space (X , d) into a
metric space (Y , ρ), then the range f (X ) of f is also compact.

As every uniformly continuous function on A is continuous on A, we have

the following property.

Third property :

Suppose (X , d) and (Y , ρ) are metric spaces and f : X → Y is uniformly

continuous on X . Then f maps every compact subset of X to a compact
subset of Y .

P. Sam Johnson Uniform Continuity 79

Question

Do any or all of the above three properties characterize uniform

continuity?

The answer is no.

For example, the exponential function satisfies all of them. It satisfies the
first property because all Cauchy sequences are convergent in R and
continuous functions map convergent sequences onto sequences that are
convergent and therefore Cauchy. It satisfies the second property because,
being continuous on R, it is uniformly continuous on every bounded subset
of R. Such sets are totally bounded and so are mapped by the function to
totally bounded subsets of R. It satisfies the third property because it is
uniformly continuous on closed bounded subsets of R.

Michael-152

P. Sam Johnson Uniform Continuity 80

Question

Do uniformly continuous functions map all bounded sets onto

bounded sets?

They do in familiar situations where the domain and codomain are subsets
of Rn , but that is because boundedness and total boundedness are the
same thing in those spaces.

It is not always so.

Consider the identity function from N to N, where the domain is given the
discrete metric and the codomain the usual metric. The identity function
is uniformly continuous because the metric on its domain is the discrete
metric, but the domain is a bounded space and the range is not.

Michael-152

P. Sam Johnson Uniform Continuity 81

Uniform Continuity on Subsets of the Cantor Set

Subsets of the Cantor set are totally bounded since they are bounded
subsets of R. So every uniformly continuous image of a subset of the
Cantor set is also totally bounded. But this is actually a
characterization of totally bounded metric spaces.
Theorem 56.
Suppose (X , d) is a non-empty metric space. Then X is totally bounded
if, and only if, there exists a bijective uniformly continuous function from a
subset of the Cantor set K onto X .

Proof : Suppose X is totally bounded. For each m ∈ N, choose a finite

collection Bm of open balls of radius 1/m that covers X . All these balls
together form a countable collection.

Michael-153

P. Sam Johnson Uniform Continuity 82

Proof (contd...)

By enumerating all the members of each Bm in turn, we form a sequence

(Un ) of open balls in which, for each m ∈ N, the balls of Bm precede those
of Bm+1 . 1 Then (Un ) has the property that diam(Un ) → 0 as n → ∞.

For eachPx ∈ X , let αn (x) = 2 if x ∈ Un and αn (x) = 0 otherwise, and set

g (x) = ∞ n
n=1 αn (x)/3 . Then g (x) ∈ K . Note that there is an infinite
number of values of n for which αn (x) = 2 because each Bm is a cover for
X . It follows that g is injective because, for x, z ∈ X with x 6= z, we have
αn (z) = 0 whenever both αn (x) = 2 and d(x, z) > diam(Un ).

Let φ = g −1 . Then φ is a bijective map from the subset g (X ) of K onto

X . We want to show that φ is uniformly continuous.

Michael-153

1
Duplications are possible because radii are not well-defined, but this does not affect
the argument.
P. Sam Johnson Uniform Continuity 83
Proof (contd...)

Let ε > 0 be given. Let p ∈ N be such that p > 2/ε. Then every member
of Bp has diameter less than ε. Let k ∈ N be the largest subscript
assigned to a member of Bp in the enumeration (Un ) of the covering balls.

Suppose a and b are arbitrary members of g (X ) that satisfy

|a − b| < 1/3k . Let φ(a) = x and φ(b) = z. Then g (x) = a and
g (z) = b, so that αn (x) = αn (z) for all n ∈ Nk -in other words, for all
n ∈ Nk , x ∈ Un if, and only if, z ∈ Un . Since Bp covers X and all
members of Bp occur in the first k terms of (Un ), there exists q ∈ Nk with
z ∈ Uq ∈ Bp , whence also x ∈ Uq .

Then d(φ(a), φ(b)) = d(x, z) ≤ diam(Uq ) ≤ 2/p < ε. Since ε > 0 is

arbitrary, the uniform continuity of φ follows. This proves the forward
implication; the proof of the backward one is stated before the present
result.

Michael-153

P. Sam Johnson Uniform Continuity 84

Uniform Continuity on Subsets of the Cantor Set

We deduce from Theorem 56 that all totally bounded metric spaces are
relatively small. Since they are all in one-to-one correspondence with a
subset of R, none has cardinality greater than R.

Michael-154

P. Sam Johnson Uniform Continuity 85

Extensions of Functions

The next theorem involves extensions of functions. We say that a function

g is an extension of a function f if dom(f ) ⊆ dom(g ) and f (x) = g (x)
for all x ∈ dom(f ).

Example 57.
Let f (x) = x sin( x1 ) for x ∈ (0, π1 ]. The function defined by
(
x sin( x1 ) for 0 < x ≤ π1
g (x) =
0 for x = 0

is an extension of f . Note that dom(f ) = (0, π1 ] and dom(g ) = [0, π1 ]. In

this case, g is a continuous extension of f .

Ross-139

P. Sam Johnson Uniform Continuity 86

Graph of f (x) = x sin( x1 ) for x ∈ (0, π1 ].

x sin( x1 )

x
0.25 0.5

P. Sam Johnson Uniform Continuity 87

Extensions of Functions
Example 58.
Let f (x) = sin( x1 ) for x ∈ (0, π1 ]. The function f can be extended to a
function g with domain [0, π1 ]. in many ways, but g will not be continuous.

Ross-141

P. Sam Johnson Uniform Continuity 88

Graph of f (x) = sin( x1 ) for x ∈ (0, π1 ].

sin(1/x)

x
0.5 1 1.5

P. Sam Johnson Uniform Continuity 89

Uniform Extension Theorem
Theorem 59 (Uniform Extension Theorem).
Let X and Y be metric spaces and D be a dense subset of X . If
f : (D, d) → (Y , ρ) is uniformly continuous and if Y is complete, then
there exists a uniformly continuous function g : X → Y such that
g (x) = f (x) for all x ∈ D.

Steps involved:
1. For x ∈ X , take a sequence (xn ) in D such that xn → x.
2. Observe that f (xn ) is a Cauchy sequence in Y .
3. Write g (x) = limn→∞ f (xn ).
4. Observe that if (xn0 ) in D with xn0 → x, then
limn→∞ f (xn0 ) = limn→∞ f (xn ).
5. Observe that g : X → Y is well-defined.
6. Observe that g (x) = f (x) for every x ∈ D.
7. Show that g is uniformly continuous on X .
P. Sam Johnson Uniform Continuity 90
Proof of the theorem

Let x ∈ X . Then by density of D in X , there exists a sequence (xn ) in D

such that xn → x. Note that (xn ) is Cauchy (since every convergent
sequence is Cauchy). Since f is uniformly continous, the sequence (f (xn ))
is Cauchy in Y . Since Y is complete, there exists y ∈ Y such that
f (xn ) → y .

We set g (x) = y . We need to show that g (x) is well-defined in the

sense that if (xn0 ) in D converges to x and if f (xn0 ) → y 0 , then y = y 0 .

Suppose there is a sequence (xn0 ) in D converging to x and f (xn0 )

converges to some y 0 in X . We need to show that y = y 0 .

Since f is uniformly continuous on D and d(xn , xn0 ) → 0 as n → ∞,

ρ(f (xn ), f (xn0 )) → 0 as n → ∞.

P. Sam Johnson Uniform Continuity 91

Proof (contd...)
Let ε > 0 be given. Since each of the sequences ρ(f (xn ), y ), ρ(f (xn0 ), y 0 )
and ρ(f (xn ), f (xn0 )) is converging 0 as n → ∞, there are natural numbers
N1 , N2 , N3 such that
ε
ρ(f (xn ), y ) < , for all n ≥ N1 (1)
3
ε
ρ(f (xn0 ), y 0 ) < , for all n ≥ N2 (2)
3
ε
ρ(f (xn ), f (xn0 )) < , for all n ≥ N3 . (3)
3

Let N0 = max{N1 , N2 , N3 }.

Then by the triangle inequality we have for all n ≥ N0

ε ε ε
ρ(y , y 0 ) ≤ ρ(y , f (xn )) + ρ(f (xn ), f (xn0 )) + ρ(f (xn0 ), y 0 ) < + + = ε.
3 3 3
Since LHS of the above inequality is independent of n, we have
ρ(y , y 0 ) < ε, for every ε > 0. Thus y = y 0 .

P. Sam Johnson Uniform Continuity 92

Proof (contd...)

We proved that the map g is well-defined. Note that g (x) = f (x) for
x ∈ D, as we may take the constant sequence (xn := x) convergent to x.

Next prove that g is uniformly continuous on X .

Let ε > 0 be given. Since f is uniformly continuous on D, there exists

δ > 0 such that for all a, b ∈ D with d(a, b) < δ, we have
ρ(f (a), f (b)) < ε. We have to show the samething happens for any two
points in X . Let x, y ∈ X with d(x, y ) < δ.

Since D is dense in X , there are sequences (xn ) and (yn ) in D such that
xn → x and yn → y . In particular, (xn ) and (yn ) are Cauchy. Since f is
uniformly continuous on D, (f (xn )) and (f (yn )) are Cauchy sequences in
Y , converging to some u and v respectively.

P. Sam Johnson Uniform Continuity 93

Proof (contd...)

Hence by the above arguments, we have g (x) = u, g (y ) = v .

We need to show that ρ(g (x), g (y )) = ρ(u, v ) < ε. We use ε/3-argument.

Now,
ρ(u, v ) ≤ ρ(u, f (xn )) + ρ(f (xn )), f (yn )) + ρ(f (yn ), v )
< ε for all n ≥ max{N1 , N2 }.
Please note that we got natural numbers N1 and N2 from the convergence
of sequences ρ(u, f (xn )) and ρ(f (yn ), v ) to 0 respectively (we apply
ε/3-argument) ; also ρ(f (xn )), f (yn )) < ε/3 (applying uniform continuity
of f on D).

LHS is independent of n. Hence

ρ(g (x), g (y )) < ε.
This completes the proof.

Karunakaran 5-43, 5-47

P. Sam Johnson Uniform Continuity 94
Application of the uniform extension theorem

An instructive application of the theorem is the extension of the meaning

of ar for r ∈ Q to ax for any x ∈ R for any fixed a > 0.

Starting from the existence of nth roots (Theorem A), one assigns a
meaning to am/n where m, n ∈ Z and n 6= 0. One also verifies the laws of
exponents hold. We shall assume these results in the proof of the theorem
below.

Theorem A : [Existence of nth roots] Let α ∈ [0, ∞) and n ∈ N.

Then there exists a unique x ∈ [0, ∞) such that x n = α.
Theorem 60.
Fix a positive a ∈ R and N ∈ N. Then the function r 7−→ ar from
Q ∩ [−N, N] to R is uniformly continuous. Hence it extends to a
continuous function from [−N, N] to R. This function is denoted by ax for
x ∈ [−N, N].

P. Sam Johnson Uniform Continuity 95

Uniform Continuity

Proof : Let x, x + h ∈ Q ∩ [−N, N]. We estimate

|ax+h − ax | = ax |ah − 1| ≤ aN |ah − 1|.

If we show that ah → 1 as h → 0 in Q, then we are through. (Why? How

does the uniform continuity follow?) This follows from the fact that
a1/n → 1 as n → ∞ (How?) We also used the fact that x 7−→ ax is an
increasing function on Q (Where ?) Prove this.

Remark 61.
Since ax defined on various [−N, N] coincide on their common domain, it
follows that we have a function x 7−→ ax for all x ∈ R.

Kumaresan-78

P. Sam Johnson Uniform Continuity 96

Continuous Extension Theorem

We have seen examples of functions that are continuous but not uniformly
continuous on open intervals; for example, the function f (x) = 1/x on the
interval (0, 1).

On the other hand, by the uniform continuity theorem, a function that is

continuous on a closed bounded interval is always uniformly continuous.
So the question arises: Under what conditions is a function uniformly
continuous on a bounded open interval ?

The answer reveals the strength of uniform continuity, for it will be

shown that a function on (a, b) is uniformly continuous iff it can be
defined at the end points to produce a function that is continuous on the
closed interval.

Bartle-140

P. Sam Johnson Uniform Continuity 97

Continuous Extension Theorem
Theorem 62 (Continuous Extension Theorem).
A function f : (a, b) → R is uniformly continuous on the interval (a, b) if
and only if it can be defined at the endpoints a and b such that the
extended function is continuous on [a, b].

Proof : (⇐) This direction is trivial.

(⇒) Suppose f is uniformly continuous on (a, b). We shall show how to
extend f to a; the argument for b is similar. This is done by showing that
lim f (x) = L exists, and this is accomplished by using the sequential
x→c
criterion for limits. If (xn ) is a sequence in (a, b) with lim(xn ) = a, then it
is a Cauchy sequence, and by the preceding theorem, the sequence (f (xn ))
is also a Cauchy sequence, and so is convergent. Thus the limit
lim(f (xn )) = L exists.

Bartle-140

P. Sam Johnson Uniform Continuity 98

Proof (contd...)

If (un ) is any other sequence in (a, b) that converges to a, then

lim(un − xn ) = a − a = 0, so by the uniform continuity of f we have

lim(f (un )) = lim(f (un ) − f (xn )) + lim(f (xn ))

= 0 + L = L.

Since we get the same value L for every sequence converging to a, we infer
from the sequential criterion for limits that f has limit L at a. If we define
f (a) := L, then f is continuous at a.

The same argument applies to b, so we conclude that f has a continuous

extension to the interval [a, b].

Bartle-140

P. Sam Johnson Uniform Continuity 99

Uniform Extension Theorem
Example 63.
Since the limit of f (x) := sin(1/x) at 0 does not exist, we infer from the
Continuous Extension Theorem that the function is not uniformly
continuous on (0, b] for any b > 0.

On the other hand, since lim x sin(1/x) = 0 exists, the function

x→0
g (x) := x sin(1/x) is uniformly continuous on (0, b] for all b > 0.

Bartle-140

P. Sam Johnson Uniform Continuity 100

Equicontinuity

We have discussed that in the case of uniform continuity of f , given ε > 0,

the choice of δ > 0 depends only on ε, not on the point.

In the following definition, we have a family F of real functions defined on

a common set E ⊆ R in which the choice of δ > 0 does not depend on
either the point or the funciton f ∈ F. In particular, every function f in
the family F is uniformly continuous.
Definition 64.
A family F of real functions defined on a common set E ⊆ R is said to be
equicontinuous on E if given ε > 0 there exists a δ > 0 such that
|f (x) − f (y )| < ε for all x, y ∈ E with |x − y | < δ and for all f ∈ F.

We shall later discuss equicontinuity.

Karunakaran-9-22

P. Sam Johnson Uniform Continuity 101

Uniform Limit
Exercise 65.
Show that the uniform limit of a sequence of uniformly continuous
functions on R is uniformly continuous on R.

Karunakaran-9-39

P. Sam Johnson Uniform Continuity 102

Uniform Continuity for Functions of Two Variables

The notion of uniform continuity for functions of one variable can be easily
extended to functions of two variables. Let D be a subset of R2 . A
function f : D → R is said to be uniformly continuous on D if for any
sequences ((xn , yn )) and ((un , vn )) in D such that |(xn , yn ) − (un , vn )| → 0,
we have |f (xn , yn ) − f (un , vn )| → 0. Note that |.| denotes Euclidean
distance in R2 .

Specializing one of the two sequences to a constant sequence, we readily

see that a uniformly continuous function is continuous. As in the case of
functions of one variable, the converse is true if the domain is closed and
bounded.

P. Sam Johnson Uniform Continuity 103

Uniform Continuity for Functions of Two Variables
Proposition 66.
Let D ⊆ R2 be a closed and bounded set. Then every continuous function
on D is uniformly continuous on D.
Proof : Suppose f : D → R is continuous but not uniformly continuous
on D. Then there are sequences ((xn , yn )) and ((un , vn )) in D such that
|(xn , yn ) − (un , vn )| → 0, but |f (xn , yn ) − f (un , vn )| 9 0, The latter implies
that there are ε > 0 and positive integers n1 < n2 < · · · such that
|f (xnk , ynk ) − f (unk − vnk | ≥ ε for all k ∈ N. Now, by the
Bolzano-Weierstarass Theorem, ((xnk − ynk )) has a convergent sub
sequence, say ((xnkj , ynkj )). If (xnkj , ynkj ) → (x0 , y0 ), then
(xnkj , ynkj ) → (x0 , y0 ), because |(xn , yn ) − (un , vn )| → 0. Since f is
continuous on D, we see that
|f (xnkj , ynkj ) − f (unkj , vnkj )| → |f (x0 , y0 ) − f (x0 , y0 )| = 0. But this is a
contradiction, since |f (xnkj , ynkj ) − f (unkj , vnkj )| ≥ ε for all j ∈ N. Sudhir 61-63

P. Sam Johnson Uniform Continuity 104

Uniform Continuity for Functions of Two Variables
Example 67.
(i) Consider f : R2 → R defined by f (x, y ) := x + y . Then it is clear that
f is uniformly continuous on R2 .
(ii) If D ⊆ R2 and f : D → R is uniformly continuous, then for every
fixed (x0 , y0 ) ∈ D, the functions φ : D1 → R and ψ : D2 → R (of one
variable) defined by

φ(x) := f (x, y0 ) for x ∈ D1 and ψ(y ) := f (x0 , y ) for y ∈ D2

are uniformly continuous. This follows from the definition of uniform

continuity by specializing one of the coordinates in the two sequences
to a constant sequence.

Sudhir 61-63

P. Sam Johnson Uniform Continuity 105

Uniform Continuity for Functions of Two Variables
Example 68.
Consider D ⊆ R2 and f : D → R given by
1
D := {(x, y ) ∈ R2 : x, y ∈ [0, 1] and (x, y ) 6= (0, 0)} and f (x, y ) := x+y .
Then f is continuous on D but not uniformly continuous on D. To see the
latter, consider the sequences ((xn , yn )) and ((un , vn )) in D given by
(xn , yn ) := (1/n, 0) and (un , vn ) := (1/(n + 1), 0) for n ∈ N. We have
|(xn , yn ) − (un , vn )| = 1/n(n + 1) → 0, but
|f (xn , yn ) − f (un , vn )| = |n − (n + 1)| = 1 9 0. Alternatively, we could use
(ii) above and the fact that φ : (0, 1] → R defined by
φ(x) = f (x, 0) = 1/x is not uniformly continuous on (0, 1]. It may be
noted here that the domain of f is bounded but not closed.

Sudhir 61-63

P. Sam Johnson Uniform Continuity 106

Uniform Continuity for Functions of Two Variables
Example 69.
Consider f : R2 → R defined by f (x, y ) : x 2 + y 2 . Then f is continuous on
R2 , but not uniformly continuous on R2 . To see the latter, consider the
sequences ((xn , yn )) and ((un , vn )) in D given by (xn , yn ) := (n, 0) and
(un , vn ) := (n − (1/n), 0) for n ∈ N. We have
|xn , yn ) − (un , vn )| = 1/n → 0, but
|f (xn , yn ) − f (un , vn )| = |n2 − [n2 − 2 + (1/n2 )]) = 2 − (1/n2 ) 9 0.
Alternatively, we could use (ii) above and the fact that φ : R → R defined
by φ(x) = f (x, 0) = x 2 is not uniformly continuous on R. It may be noted
here that the domain of f is closed but not bounded. On the other hand,
the restriction of f to any bounded subset of R2 is uniformly continuous.

Sudhir 61-63

P. Sam Johnson Uniform Continuity 107

Uniform Continuity for Functions of Two Variables

A criterion for the uniform continuity of a function of two variables that

does not involve convergence of sequences can be given as follows.
Proposition 70.
Let D ⊆ R2 . Consider a function f : D → R. Then f is uniformly
continuous on D if and only if it satisfies the following ε − δ condition: For
every ε > 0, there is δ > 0 such that
(x, y , (u, v ) ∈ D and|(x, y ) − (u, v )| < δ ⇒ |f (x, y ) − f (u, v )| < ε.

Proof : Assume that f is uniformly continuous on D. Suppose the ε − δ

condition does not hold. Then there id ε > 0 such that for any δ > 0, we
can find (x, y ), (u, v ) ∈ D for which |(x, y ) − (u, v )| < δ, but
|f (x, y ) − f (u, v )| ≥ ε. Sudhir 61-63

P. Sam Johnson Uniform Continuity 108

Proof (contd...)

Considering δ := 1/n for n ∈ N, we obtain sequences ((xn , yn )) and

((un , vn )) in D such that |(xn , yn ) − (un , vn )| < n1 and
|f (xn , yn ) − f (un , vn )| ≥ ε for all n ∈ N. Consequently,
|(xn , yn ) − (un , vn )| → 0, but |f (xn , yn ) − f (un , vn )| 9 0. this contradicts
the assumption that f is uniformly continuous on D.

Conversely, assume that the ε − δ condition is satisfied. Suppose ((xn , yn ))

and ((un , vn )) are any sequences in D such that |(xn , yn ) − (un , vn )| → 0.
Let ε > 0 be given. Then there is δ > 0 such that if (x, y ), (u, v ) ∈ D
satisfy |(x, y ) − (u, v )| < δ, then |f (x, y ) − f (u, v ) < ε. Now, for this
δ > 0, we can find n0 ∈ N such that |(xn , yn ) − (un , vn )| < δ for all n ≥ n0 .
Consequently, |f (xn , yn ) − f (un , vn )| < ε for all n ≥ n0 . Thus
|f (xn , yn ) − f (un , vn )| → 0. This proves the uniform continuity of f on D.

Sudhir 61-63

P. Sam Johnson Uniform Continuity 109

Lipschitz Function

The type of global continuity that we habitually encounter amongst linear

maps between normed linear spaces (in Functional Analysis) is Lipschitz
continuity. It is stronger than uniform continuity and has the advantage
that it preserves boundedness.
Definition 71.
Let X and Y be metric spaces. A function f : X → Y is said to be
Lipschitz function if there exists k > 0 such that

ρ(f (x), f (y )) ≤ k d(x, y ) ∀x, y ∈ X . (1)

The number k above is called a Lipschitz constant. Every real number

larger than k is also a Lipschitz constant for f .
If f : X → Y is a Lipschitz function function with Lipschitz constant
k < 1, then f is called a contraction.

Michael-154
P. Sam Johnson Uniform Continuity 110
Isometry
Definition 72.
Let (X , d) and (Y , ρ) be metric spaces and f : X → Y . The function f is
said to be isometry if d(x, y ) = ρ(f (x), f (y )) for all x, y , ∈ X .

1. Isometry is necessarily injective and its inverse is also an isometry.

2. As metric spaces, X and f (X ) are indistinguishable; f (X ) is merely
a relabelling of X .
3. Every metric space has a completion and any two completions are
isometric to each other.

Example 73.
Isometries behave much better than other Lipschitz functions, as they
obey an equality rather than an inequality.

Michael-155
P. Sam Johnson Uniform Continuity 111
Lipschitz Function

Let f be a real-valued function whose domain I is an interval.

The condition (1) that a function f : I → R on an interval I is a Lipschitz

function can be interpreted geometrically as follows. If we write the
condition as
f (x) − f (u)
≤k for all x, y ∈ I, x 6= u,
x −u
then the quantity inside the absolute values is the slope of a line segment
joining the points (x, f (x)) and (u, f (u)). Thus a function f satisfies a
Lipschitz condition if and only if the slopes of all line segments joining two
points on the graph of y = f (x) over I are bounded by some number k.

Bartle-139, Stephen-120

P. Sam Johnson Uniform Continuity 112

Lipschitz Function

Every Lipschitz function function is uniformly continuous.

Theorem 74.
Suppose (X , d) and (Y , ρ) are metric spaces, A is a subset of X and
f : X → Y.
(i) If f is a Lipschitz function on A with Lipschitz constant k > 0, then f
is uniformly continuous on A and δ in the definition of uniform
continuity can be taken to be ε/k.
(ii) If f is a Lipschitz function on X with Lipschitz constant k > 0, then
f is a Lipschitz function on A with Lipschitz constant k.

Proof : For every x, y ∈ X , we have

d(x, y ) < ε/k ⇒ ρ(f (x), f (y )) ≤ k d(x, y ) < ε,

which proves (i). The proof of (ii) is obvious. Michael-154

P. Sam Johnson Uniform Continuity 113

Lipschitz Function
Example 75.
If f (x) := x 2 on A := [0, b], where b > 0, then

|f (x) − f (y )| = |x + y ||x − y | ≤ 2b |x − y |

for all x, y in [0, b]. Thus f satisfies the Lipschitz condition with k = 2b
on A, and therefore f is uniformly continuous on A.
Note that f does not satisfy the Lipschitz condition (1) on the interval
[0, ∞).

P. Sam Johnson Uniform Continuity 114

Lipschitz Function
Exercise 76 (Compositions of Lipschitz functions are
Lipschitz).
Suppose (X , d), (Y , ρ) and (Z , γ) are metric spaces, f : X → Y and
g : Y → Z . If f and g are Lipschitz functions with Lipschitz constants k
and ` on X and f (X ), respectively, then g ◦ f is a Lipschitz function on X
with Lipschitz constant k`.

Michael-158

P. Sam Johnson Uniform Continuity 115

Lipschitz Function

Not every uniformly continuous function is a Lipschitz function.

Example 77.
√
Let g (x) := x for x in the closed bounded interval I := [0, 2]. Since g is
continuous on I , it follows from the Uniform Continuity Theorem that g is
uniformly continuous on I .
√
If |g (x) − g (0)| ≤ k |x − 0| for all x ∈ [0, 1], then x ≤ k x for x ∈ [0, 1].
But if xn := 1/n2 , then k must satisfy n ≤ k for all n ∈ N, which is
impossible.
Hence, there is no number k > 0 such that |g (x)| ≤ k |x| for all x ∈ I .
Therefore, g is not a Lipschitz function on I .

Bartle-139

P. Sam Johnson Uniform Continuity 116

Lipschitz Function

The Uniform Continuity Theorem and Theorem 74 (every Lipschitz

function is uniformly continuous) can sometimes be combined to establish
the uniform continuity of a function on a set.
Example 78.
√
We consider g (x) := x on the set A := [0, ∞). The uniform continuity
of g on the interval I := [0, 2] follows from the Uniform Continuity
Theorem. If J := [1, ∞), then if both x, y are in J, we have
√ √ |x − y | 1
|g (x) − g (y )| = | x − y | = √ √ ≤ |x − y |.
x+ y 2

Thus g is a Lipschitz function on J with constant k = 12 , and hence by

Theorem 74, g is uniformly continuous on [1, ∞). Since A = I ∪ J, it
follows that g is uniformly continuous on A by taking
δ(ε) := inf{1, δI (ε), δJ (ε)}. Bartle-139

P. Sam Johnson Uniform Continuity 117

Lipschitz Function
Theorem 79.
Suppose (X , d) and (Y , ρ) are metric spaces, S is a bounded subset of X
and f : X → Y is a Lipschitz function. Then f (S) is bounded in Y .

Proof : Let k > 0 be a Lipschitz constant for f and suppose x, y ∈ S.

Then ρ(f (x), f (y )) ≤ k d(x, y ) ≤ k diam(S), so that

diam(f (S)) ≤ k diam(S).

Michael-154

P. Sam Johnson Uniform Continuity 118

Continuity on Unions

A function f is continuous if, and only if, it is continuous at every point of

its domain; the same applies to restrictions of f . Let us suppose that the
domain of f is split up into several constituent parts and that the
restriction of f to each of those parts is continuous; in other words, each
restriction of f is continuous at every point of the appropriate constituent
part.

Does it follow that f is continuous at every point of its domain and is

therefore a continuous function? It is important to know that it does not,
even if the constituent parts are closed in the domain and mutually disjoint
(8.3.7). However, 8.7.1 gives a sufficient condition for the truth of the
implication.

P. Sam Johnson Uniform Continuity 119

Continuity on Unions
Theorem 80.
Suppose (X , d) and (Y , ρ) are metric spaces. Suppose C is non-empty
collection
S of mutually disjoint non-empty subspaces of X and
f : CS → Y . Suppose that, for each A ∈ C, we have S
A ∩ Cl( (C {A})) = φ and f |A continuous. Then f is continuous on C.

S
Proof : Suppose a ∈ C, and let A ∈ C be such that a ∈ A. If C = {A},
the result is trival, so we suppose otherwise. Let ε > 0. Since f |A is
continuous, it is continuous at a, so there exists γ > 0 such that, for all
x ∈ A for which d(x, a) < γ, we have ρ(f (x), f (a)) < ε.

Michael-138

P. Sam Johnson Uniform Continuity 120

Continuity on Unions
S S
Since A ∩ Cl( (C\{A})) = φ, 3.6.10 given η = dist(a, (C\{A})) 6= 0,
and,because A is notSthe only member of C, η > 0. Let δ = min{γ, η}.
Then, for each x ∈ C with d(x, a) < δ, we have x ∈ A, so that
e(f (x), f (a)) < ε.

Because
S ε > 0 is arbitrary, f is continuous at a; and because
S a is arbitrary
in C, f satisfies
S the local criterion for continuity on C, so f is
continuous on C. Michael-138

P. Sam Johnson Uniform Continuity 121

Continuity on Unions

Question 8.7.2

In 8.3.7 and 8.7.1, we considered only disjoint subsets of a domain. Let us

look now at overlapping parts of a domain. Suppose X and Y are metric
spaces and A and B are subsets X with A ∩ B 6= φ. Suppose
f : A ∪ B → Y has continuous restrictions to A and B. Does f have to be
continuous? The answer is no. Consider the real function defined on C by

0,
 if < ≥ 0 and =z ≥ 0;
z 7→ |<(z)|, if <z < 0 and =z ≥ 0;

|=(z)|, if =z < 0.


This function is continuous on {z ∈ C|=z ≥ 0} and is also continuous on

{z ∈ C|<z ≥ 0 or =z < 0}, but, being discontinuous at every point of the
negative part of the real line, is not continuous on their union C.

Michael-138
P. Sam Johnson Uniform Continuity 122
Continuity of Mappings into Product Spaces

Every finite product of metric spaces comes equipped with natural

projections onto the coordinate spaces (1.6). These projections are
continuous provided only that the product is endowed with a product
metric. Moreover, a function that maps into the product is continuous if,
and only if, its compositions with the natural projections are all continuous.
Theorem 81.
Suppose n ∈ N and, for each i ∈ Nn , (Xi , τi ) is a space. Endow
P = Πni=1 Xi with a product metric. Then, for each j ∈ Nn , the natural
projection πj : P → Zj is continuous.

Proof : Suppose j ∈ Nn and V is open in Xj . Then

πj−1 (V ) = {x ∈ P|xj ∈ V }, which can be expressed as Πni=1 Ui , where
Uj = V and Ui = Xi for all i ∈ Nn {j}. This is certainly a member of the
product topology (4.5.2) because V is open in Xj and Xi is open in Xi for
all i ∈ Nn \{j}. Therefore πj is continuous.
P. Sam Johnson Uniform Continuity 123
Continuity of Mappings into Product Spaces
Theorem 82.
Suppose n ∈ N and, for each i ∈ Nn , (Xi , τi ) is a metric space. Endow
P = Πni=1 Xi with a product metric. Suppose Z is a metric space and
f : Z → P. Then f is continuous if, and only if, π ◦ f is continuous for all
i ∈ Nn .
A function that is uniformly continuous on a number of disjoint closed sets
may well not be uniformly continuous on their union even if the condition
of Theorem 8.7.1 is satisfied; for a sufficient condition for uniform
continuity, we confine ourselves to finite unions. Lipschitz continuity has
even less stability.
Theorem 83.
Suppose (X , d) and (Y , ρ) are metric spaces and C is a finite collection of
non-empty subsets
S of X such that dist(A, B) > 0 for all A, B ∈ C.
Suppose f : C → Y has uniformly continuous restriction
S to each
member of C. Then f is uniformly continuous on C.
P. Sam Johnson Uniform Continuity 124
Proof of the theorem

Let ε > 0 be given. For each A ∈ C, f |A is uniformly continuous by

hypothesis. Let δA be such that, for all

u, v ∈ A, d(u, v ) < δA ⇒ ρ(f (u), f (v )) < ε.

The set
{dist(A, B)|A, B ∈ C} ∪ {δA |A ∈ C}
is a finite subset of the set of all positive real numbers and so has a
minimum member δ > 0.
S
Then, for all u, v ∈ C with d(u, v ) < δ, there exists A ∈ C such that
u, v ∈ A and then, since δ ≤ γA , we have ρ(f (u), f (v )) < ε.

Michael-158

P. Sam Johnson Uniform Continuity 125

Uniform and Lipschitz Continuity on Unions
Example 84.
The condition dist(A, B) > 0 in Theorem 83 cannot in general be
weakened to A ∩ B = ∅. Consider the two subsets
A = {(x, 1/x)|x ∈ R\{0}} and B = {(x, 0)|x ∈ R} of R2 with the
Euclidean metric.
Both are closed in R2 . Define f to be 1 on A and 0 on B. Then f is
uniformly continuous on each of A and B but not on A ∪ B. Specifically,
let δ > 0. Then, for x > 1/δ, the distance from (x, 1/x) to (x, 0) is less
than δ and the distance between their images under f is 1.

Michael-158

P. Sam Johnson Uniform Continuity 126

Uniform and Lipschitz Continuity on Unions
Example 85.
No theorem like Theorem 83 is possible for Lipschitz functions. Consider
the function f : N → N given by 2n − 1 7→ 2n − 1 and 2n 7→ 4n for each
n ∈ N, where N is endowed, as domain and codomain, with its usual
metric.
The restriction of f to the odd natural numbers is the identity function,
which is Lipschitz with Lipschitz constant 1; and the restriction of f to the
even natural numbers is the doubling function, which is Lipschitz with
Lipschitz constant 2. These two sets are a distance 1 apart.
But f is not Lipschitz because for each k > 0 and n ∈ N with n ≥ k/2, we
have f (2n) − f (2n − 1) = 2n + 1 > k(2n − (2n − 1)).

Michael-158

P. Sam Johnson Uniform Continuity 127

Uniform and Lipschitz Continuity on Unions
Exercise 86.
Find a function that is uniformly continuous on an infinite number of
closed sets, every two of which are of distance at least 1 from each other,
but is not uniformly continuous on their union.

Michael-162

P. Sam Johnson Uniform Continuity 128

Continuity of Mappings into Product Spaces

Every finite product of metric spaces comes equipped with natural

projections onto the coordinate spaces. These projections are continuous
provided only that the product is endowed with a product metric.
Moreover, a function that maps into the product is continuous iff its
compositions with the natural projections are also continuous.
Theorem 87.

Q n ∈ N and for each i ∈ Nn , (Xi , τi ) is a metric space. Endow

Suppose
P = ni=1 Xi with a product metric. Then, for each j ∈ N, the natural
projection πj : P → Xj is continuous.

Theorem 88.

Qn n ∈ N and for each i ∈ Nn , (Xi , τi ) is a metric space. Endow

Suppose
P = i=1 Xi with a product metric. Suppose Z is a metric space and
f : Z → P. Then f is continuous iff πi ◦ f is continuous, for all i ∈ Nn .
P. Sam Johnson Uniform Continuity 129
Uniform and Lipschitz Continuity on Products

Not every product metric ensures the uniform continuity of the natural
projections. We need to make restrictions in order to get theorems similar
to Theorem 87 and Theorem 88.
Theorem 89.
Suppose n ∈ N and, for each i ∈ Nn , (Xi , τi ) is a metric space. Endow
P = Πni=1 Xi with a conserving metric e. Then all the natural projections
πi : P → Xi are Lipschitz maps with Lipschitz constant 1.

Proof : Suppose a, b ∈ P. Then, for each

i ∈ Nn , τi (πi (a), πi (b)) ≤ ρ(a, b) because e is a conserving metric.

Michael-159

P. Sam Johnson Uniform Continuity 130

Uniform and Lipschitz Continuity on Products
Theorem 90.

Qn n ∈ N and, for each i ∈ Nn , (Xi , τi ) is a metric space. Endow

Suppose
P = i=1 Xi with a conserving metric e. Suppose (Z , m) is a metric space
and f : Z → P. Then:
(i) f is uniformly continuous if, and only if, πi ◦ f is uniformly continuous
for all i ∈ Nn .
(ii) f is a Lipschitz function if, and only if, and only if, πi ◦ f is a
Lipschitz function for all i ∈ Nn .

Michael-159

P. Sam Johnson Uniform Continuity 131

Proof of the theorem

The forward implications are immediate consequences of Theorem 89,

Exercise 38 and Exercise 76. For the backward implication in (i), suppose
πi ◦ f is uniformly continuous for each i ∈ Nn .

Let ε > 0 and, for each i ∈ Nn , let γi be such that, for each a, b ∈ Z , we
have m(a, b) < γ + i ⇒ τi (πi (f (a)), πi (f (b)))
P < ε/n. Let
δ = min{γi |i ∈ Nn }. Then m(a, b) < δ ⇒ ni=1 τi (πi (f (a)), πi (f (b))) < ε
and, because ρ is a conserving metric, we have also
m(a, b) < δ ⇒ ρ(f (a), f (b)) < ε, as required.

Michael-159

P. Sam Johnson Uniform Continuity 132

Proof (contd...)

For the backward implication in (ii), suppose that, for each i ∈ Nn , πi ◦ f is

a Lipschitz function with Lipschitz constant `i .

Let k = ni=1 `i . Then, for eac a, b ∈ Z , we have

P
τi (πi (f (a)), πi (f (b))) ≤ `i m(a, b), whence
P n
i=1 τi (πi (f (a)), πi (f (b))) ≤ km(a, b).

Then, because e is a conserving metric, we get e(f (a), f (b)) ≤ km(a, b),
as required.

Michael-159

P. Sam Johnson Uniform Continuity 133

Solved Exercises
Exercises 91.
Which of the following continuous functions are uniformly continuous on
the specified set? Justify your answers. Use any theorems you wish.

(a) f (x) = (d) f (x) = x 3 onR,

x 17 sin x − e x cos 3x 3 on[0, π], (e) f (x) = 1
on(0, ],
x3
(b) f (x) = x 3 on[0, 1], (f) f (x) = sin x12 on(0, 1],
( c) f (x) = x 3 on(0, 1),
(g) f (x) = x 2 sin x1 on(0, 1].

Hints: To decide (a) and (b), use Theorem 19.2. Parts (c), (e ), (f) and
(g) can be settled using Theorem 19.5. Theorem 19.4 can also be used to
decide (e) and (f); compare Example 6. One needs to restore to the
definition to handle (d).

P. Sam Johnson Uniform Continuity 134

Solved Exercises
Exercises 92.
Prove that each of the following functions is uniformly continuous on the
indicated set by directly verifying the ε − δproperty in Definition 19.1.
x
(a) f (x) = x+1 on[0, 2],
5x
(b) f (x) = 2x−1 on[1, ∞).

P. Sam Johnson Uniform Continuity 135

Solution

(a) Discussion. Let ε > 0. We want

x y x −y
| − | < ε or | |<ε
x +1 y +1 (x + 1)(y + 1)

For |x − y | small, x, y , ∈ [0, 2]. since x + 1 ≥ 1 and y + 1 ≥ 1 for

x, y ∈ [0, 2], it suffices to get |x − y | < ε. So we let δ = ε.
Formal Proof : Let ε > 0 and let δ = ε. Then x, y ∈ [0, 2] and
|x − y | < δ = ε imply
x −y
|f (x) − f (y )| = | | ≤ |x − y | < ε.
(x + 1)(y + 1)
(b) (a) Discussion. Let ε > 0. We want
5y −5x
|g (x) − g (y )| = | (2x−1)(2y −1) | < ε for |x − y | small, x ≥ 1, y ≥ 1. For
x, y ≥ 1, 2x − 1 ≥ 1 and 2y − 1 ≥ 1, so it suffices to get
|5y − 5x| < ε. So let δ = 5ε . You should write out the formula proof.
P. Sam Johnson Uniform Continuity 136
Solved Exercises
Exercises 93.
which of the following continuous functions is uniformly continuous on the
specified set? Justify your answers, using appropriate theorems or Exercise
19.4(a).

(a) tan x on[0, π4 ], (d) 1

x−3 on(0, 3),
(b) tan x on[0, π2 ), (e) 1
x−3 on(3, ∞),
1 1
(c ) x sin2 xon(0, π], (f) x−3 on(4, ∞).

(a) ) tan x is uniformly continuous on [0, π4 ] by Theorem 19.2.

(b) tan x is not uniformly continuous on [0, π2 ) by Exercise 19.4(a), since the function is not
bounded on that set.
(c) Let h̃ be as in Example 9. Then(sin x)h̃(x) s a continuous extension of ( x1 ) sin2 x on [0, π].
Apply Theorem 19.5.
1
(e) x−3 is not uniformly continuous on (3, 4) by Exercise 19.4(a), so it is not uniformly
continuous on (3, ∞) either.
1
(e) (f) Remark. It is easy to give an ε − δ proof that x−3 is uniformly continuous on (4, ∞).
It is even easier to apply Theorem 19..4
P. Sam Johnson Uniform Continuity 137
Solved Exercises
Exercises 94.
(a) Let f be a continuous function on [0, ∞). Prove that if f is uniformly
continuous on [k, ∞) for some k, then f is uniformly continuous on
[0, ∞).
√
(b) Use (a) and Exercise 19.6(b) to prove that x is uniformly
continuous on [0, ∞).

(a) We are given that f is uniformly continuous on [k, ∞), and f is uniformly continuous on
[0, k + 1] by Theorem 19.2. Let ε > 0. There exist δ1 and δ2 so that

|x − y | < δ1 , x, y ∈ [k, ∞) imply |f (x) − f (y )| < ε, (1)

|x − y | < δ2 , x, y ∈ [0, k + 1] imply |f (x) − f (y )| < ε.(2)

Let δ = min{1, δ1 , δ2 } and show that

|x − y | < δ, x, y ∈ [0, ∞) imply |f (x) − f (y )| < ε.

P. Sam Johnson Uniform Continuity 138

Solved Exercises
Exercises 95.
Let f (x) = x sin( x1 ) for x 6= 0 and f (x) = 0.
(a) Observe that f is continuous on R; see Exercises 17.3(f ) and
417.9(c).
(b) why is f uniformly continuous on any bounded subset of R?
(c ) Is f uniformly continuous on R?

(c ) This is tricky, but it turns out that f is uniformly continuous on R. A

simple modification of Exercise 19.7(a) shows that it suffices to show that
f is uniformly continuous on [1, ∞) and (−∞, −1]. This can be done
using Theorem 19.6. Note that we cannot apply Theorem 19.6 on R
because f is not differentiable at x = 0; also f 0 is not bounded near x = 0.

P. Sam Johnson Uniform Continuity 139

Solved Exercises
Exercises 96.
Accept the fact that the function h̃ in Example 9 is continuous on R;
prove that it is uniformly continuous on R.

As in the solution to Exercise 19.9(c), it suffices to show that h̃ is

uniformly continuous on [1.∞) and (−∞, −1]. Apply Theorem 19.6.

P. Sam Johnson Uniform Continuity 140

 Part - 3

P. Sam Johnson Uniform Continuity 141

Applications of Uniform Continuity

One of the important applications of uniform continuity concerns

the integrability of continuous functions on closed intervals.

We proved the uniform continuity theorem (Theorem 26) which states

that any continuous function from a compact subset K of a metric space
X into a metric space Y is uniformly continuous on K .

Theorem 26 is important in the context of Riemann integration, where to

show that every continuous function f : [a, b] → R is Riemann integrable,
what we essentially use is the uniform continuity of f .

Ross-137

P. Sam Johnson Uniform Continuity 142

Riemann Integrable Functions

Let f be a bounded function on [a, b] and P be a partition of [a, b]. Let

Mi and mi be the supremum and infimum, respectively, of f (x) on the
subinterval [xi−1 , xi ]. The upper sum of f corresponding to the partition P
is
n
X
U(f , P) = Mi (xi − xi−1 )
i=1

and the lower sum of f corresponding to the partition P is

n
X
L(f , P) = mi (xi − xi−1 ).
i=1

Since Mi ≥ mi for each i, we have U(f , P) ≥ L(f , P). Moreover,

sup L(f , P) over all partitions P is less than or equal to inf U(f , P) – they
are called the lower and upper Riemann integrals of f over [a, b]
respectively.

P. Sam Johnson Uniform Continuity 143

Riemann Integrable Functions

If the upper and lower Riemann integrals are equal, then the common
value is called the Riemann integral of f over [a, b] and we say that f is
Riemann integrable on [a, b]. To show that the integral exists, it is
sufficient to find, for any ε > 0, a partition P such that

U(f , P) − L(f , P) < ε.

Theorem 97.
If f (x) is a ( uniformly) continuous function on the closed, bounded
interval [a, b], then f is Riemann integrable on [a, b].

Note that the above result is true for piecewise-continuous functions as

well. Most functions that commonly occur in applications are continuous
or piecewise-continuous functions. Piecewise-continuous function is the
one having not more than a finite number of jump discontinuities on [a, b].
P. Sam Johnson Uniform Continuity 144
Proof

Let ε > 0 be given.

Since f is uniformly continuous on [a, b], there exists δ > 0 be such that
|f (x) − f (y )| < ε/(b − a) whenever x, y ∈ [a, b] and |x − y | < δ. Now
choose N0 so that (b − a)/N0 < δ, and let P be the regular partition of
[a, b] into N0 subintervals. Note that the regular partitions have the
property that each partition interval is exactly the same size.
ε
We will have xi − xi−1 < δ for all i, so Mi − mi < b−a for all i. Then

n N0
X b−a X ε b−a
U(f , P) − L(f , P) = (Mi − mi ) < = ε.
N0 b − a N0
i=1 i=1

So U(f , P) − L(f , P) < ε, providing the integrability of f .

P. Sam Johnson Uniform Continuity 145

Exercises

Let (X , d) be a metric space. For E ⊆ X and x ∈ X . let

d(x, E ) := inf{d(x, y ) : y ∈ E }.

Exercise 98.
Let (X , d) be a metric space. For x0 ∈ X , the map x 7−→ d(x, x0 ) is
Lipschitz function form X to R (with usual metric), with Lipschitz
constant k = 1. [Note that for x, y ∈ X ,
d(x, x0 ) − d(y , x0 ) ≤ d(x, y ), d(y , x0 ) − d(x, x0 ) ≤ d(x, y ). Hence,
|d(x, x0 ) − d(y , x0 )| ≤ d(x, y ).]

Exercise 99.
Let (xn ) be a sequence in a metric space (X , d). Show that the function
x 7−→ inf{d(x, xn ) : n ∈ N} is uniformly continuous on X . Kumaresan-77

P. Sam Johnson Uniform Continuity 146

Another useful criterion that implies uniform continuity

Here is another useful criterion that implies uniform continuity.

Theorem 100.
Let f be a continuous function on an interval I [I may be bounded or
unbounded ]. Let I ◦ be the interval obtained by removing from I any
endpoints that happen to be in I . If f is differentiable on I ◦ and if f 0 is
bounded on I ◦ , then f is Lipschitz function on I .

Proof : For this proof we need the Mean Value Theorem. Let M be a
bound for f 0 on I so that |f 0 (x)| ≤ M for all x. Consider a, b ∈ I where
a < b. By the Mean Value Theorem, there exists x ∈ (a, b) such that
f 0 (x) = f (b)−f
b−a
(a)
, so

|f (b) − f (a)| = |f 0 (x)| · |b − a| ≤ M|b − a|.

This proves the Lipschitz continuity of f on I . Ross-142

P. Sam Johnson Uniform Continuity 147

Another useful criterion that implies uniform continuity
Corollary 101.
Let f : R → R be differentiable with |f 0 (x)| ≤ M. Then f is Lipschitz.

Example 102.
Let a > 0 and consider f (x) = x12 . Since f 0 (x) = − x23 we have |f 0 (x)| ≤ 2
a3
on [a, ∞). Hence f is Lipschitz function on [a, ∞) by Theorem 100.

Ross-144, Kumaresan-78, Michael-156,157

P. Sam Johnson Uniform Continuity 148

Local Continuity

Suppose (X , d) and (Y , ρ) are non-empty metric spaces, z ∈ X and

f : X → Y . Rephrased, the epsilon-delta ball criterion for continuity of f
at z is that, for each ε > 0, the set
Sz,ε = {δ > 0 | f (b[z; δ)) ⊆ b[f (z); ε)}
is not empty.

Once the existence of some δ in Sz,ε is established, it is clear that

(0, δ] ⊆ Sz,ε . This means that every positive number smaller than δ would
satisfy the requirement for continuity just as well as δ itself.

Is there a maximum value that δ can have? In other words, is

sup Sz,ε ∈ Sz,ε ? The supremum can be infinite, in which case the answer
to the question is no. But, if x ∈ X and d(x, z) < sup Sz,ε , then
d(x, z) < δ for some δ ∈ Sz,ε (B.6.6), so that e(f (x), f (z)) < ε. It follows
that if the supremum is finite, it is a member of Sz,ε .

Michael-127
P. Sam Johnson Uniform Continuity 149
Local Continuity

A further question now arises. The function ε 7→ sup Sz,ε is certainly

decreasing in that if µ ∈ (0, ε), then Sz,µ ⊆ Sz,ε , so that
sup Sz,µ ≤ sup Sz,ε .

Can we say anything about the ratio of ε to sup Sz,ε as ε tends to 0? The
question is put partly out of curiosity and partly because we think that, in
some very special cases when f is a real differentiable function, there
ought to be a relationship between the derivative and this ratio. We give
some examples to illustrate that, even for real functions, all sorts of things
can happen to the ratio.

If f is constant, then sup Sz,ε is infinite for all ε and all z ∈ X .

If f is not constant but is constant for all x sufficiently close to z,
then sup Sz,ε is larger than some fixed positive number for all ε, so
that ε/sup Sz,ε → 0 as ε → 0.

Michael-128
P. Sam Johnson Uniform Continuity 150
Local Continuity

If f is an isometry, then, saving exceptional cases, sup Sz,ε = ε for all

ε, so that ε/ sup Sz,ε → 1 as ε → 0.
√
If X = [−1, 1] and y = R and f (x) = 1 − x 2 for all x ∈ [−1, 1],
then ε/ sup S1,ε → ∞ as ε → 0. Note also that |f 0 (x)| → ∞ as x → 1
in[−1, 1].
If f is the
 modified step function defined on [0, 1]by
2n +1 2n −1 2 1
 2 x − 2n , if n ∈ N and 2n +1 ≤ x ≤ 2n−1 ,

f (x) = 21n , if n ∈ N and 21n < x < 2n2+1

0, if x = 0,


then f is continuous at every point of [0,1] and differentiable at all points

other than those where x = 0 or x = 2/(2n + 1) or x = 1/2n for some
n ∈ N.

Michael-128

P. Sam Johnson Uniform Continuity 151

Local Continuity

The values of ε/ sup S0,ε range between 1/2 and 1 as ε → 0.

However, for each n ∈ N and ε ∈ [0, 1/2n+1 ], we have

ε/ sup S2−n , ε = 2n + 1/2.

Michael-128
P. Sam Johnson Uniform Continuity 152
Differentiable Lipschitz Functions

Let us recall the ratio ε/δ that we discussed for continuous functions. For
each ε > 0, a uniformly continuous function admits a corresponding δ > 0,
applicable now across the whole of the domain, that enables the function
to satisfy the condition for uniform continuity. But we know that as
smaller and smaller values are taken for ε, there is no guarantee that
admissible values of δ follow a regular pattern.

For a Lipschitz function f , however, the ratio ε/δ need never exceed any
Lipschitz constant for f . And our comment about differentiable functions
is justified by the following theorem.

Michael-156

P. Sam Johnson Uniform Continuity 153

Differentiable Lipschitz Functions
Theorem 103.
Suppose I is a non-degenerate interval of R and f : I → R is differentiable
on I . then f is a Lipchitz function on I if, and only if, f 0 is bounded on I .

Proof : Suppose first that k > 0 and that |f 0 (x)| ≤ k for all x ∈ I .
Suppose a, b ∈ I and a 6= b. By the Mean Value Theorem, there exists
c ∈ I with c between a and b such that f (b) − f (a) = (b − a)f 0 (c). This
yields |f (b) − f (a) ≤ k|b − a|. So f is a Lipschitz function with Lipchitz
constant k.

For the converse, suppose that f 0 is not bounded on I and let r > 0 be
arbitrary. Then there exist a, b ∈ I such that (f (b) − f (a))/(b − a) > r ,
whence |f (b) − f (a)|.r |b − a|. So, since r is arbitrary positive real, f is not
Lipschitz. Michael-157

P. Sam Johnson Uniform Continuity 154

Differentiable Lipschitz Functions
Example 104.
√
The function x 7→ 1 − x 2 defined on [0, 1] is differentiable on the interval
[0, 1); in fact, the derivative is continuous.
But the derivative is bounded only on intervals [0, α] for α ∈ (0, 1); it is
not bounded on [0, 1). So this function is Lipschitz on every interval [0, α]
with alpha ∈ (0, 1) but not Lipschitz on [0, 1). It is, however, uniformly
continuous on [0, 1] simply because it is continuous on this closed bounded
interval.

Example 105.
The function x 2 has bounded derivative 2x on every bounded interval of
R, so that, although x 2 is not even uniformly continuous on R, it is
Lipschitz on every bounded interval of R.

Michael-157

P. Sam Johnson Uniform Continuity 155

Linear Maps Between Normed Spaces

Linear maps between normed linear spaces have an extraordinary property

that makes their continuity very much easier to handle than that of other
maps: continuity at any one point of the domain implies Lipschitz
continuity throughout the domain.
Theorem 106.
Let T : X → Y be a linear map between normed spaces X and Y . Then
T is continuous iff it is Lipschitz on X .

Kumaresan-78, Michael-155

P. Sam Johnson Uniform Continuity 156

Banach Contraction Theorem
Theorem 107.
Let (X , d) be a complete metric space. If f : X → X is a contraction with
contraction constant k (0 < k < 1), then f has a unique fixed point p,
and for any x0 in X the sequence (f n (x0 )) converges to p.

In fact,
kn
d(f n (x0 ), p) ≤ d(x0 , f (x0 )). (?)
1−k

Proof : Given that f is a contraction on X with contraction constant

k ∈ (0, 1). We have d(f (x), f (y )) ≤ k d(x, y ) for all x, y ∈ X .

Let x0 ∈ X and let (xn )n≥1 be the sequence defined iteratively by

xn+1 = f (xn ) for n = 0, 1, 2, . . . . We shall prove that (xn )n≥1 is a Cauchy
sequence.

P. Sam Johnson Uniform Continuity 157

Existence

For ` = 1, 2, . . . , we have

d(x`+1 , x` ) = d(f (x` ), f (x`−1 )) ≤ k d(x` , x`−1 ).

Repeated application of the above inequality gives

d(x`+1 , x` ) ≤ k d(x` , x`−1 )

≤ k 2 d(x`−1 , x`−2 ) ≤ · · · ≤ k ` d(x1 , x0 ).

Now, let n, p be positive integers. By the triangle inequality,

d(xn+p , xn ) ≤ d(xn+p , xn+p−1 ) + d(xn+p−1 , xn+p−2 ) + · · · d(xn+1 , xn )

≤ k n+p−1 + k n+p−2 + · · · + k n d(x1 , x0 )
 

≤ k n k p−1 + k p−2 + · · · + 1 d(x1 , x0 )

 
h 1 − kp i
≤ kn d(x1 , x0 ).
1−k
P. Sam Johnson Uniform Continuity 158
Existence

For large values of n and p, we have k n and k p become smaller near 0

since limn→∞ k n = 0. It follows that (xn ) is a Cauchy sequence in X ,
which is complete. Let p = limn→∞ xn .

Since f is contraction, it is continuous. It follows that

f (p) = f ( lim xn ) = lim f (xn ) = lim xn+1 = p.

n→∞ n→∞ n→∞

Thus, p is a fixed point of f .

Uniqueness : Note that we cannot have more than one fixed point.

If there are two fixed points, say p, q, we would have

d(f (p), f (q)) = d(p, q). But we know that d(f (p), f (q)) ≤ k d(p, q)
where k < 1 since f is a contraction.

P. Sam Johnson Uniform Continuity 159

Importanace of the Inequality (?)

The importanace of the inequality (?) (given in Banach contraction

theorem) is as follows : Suppose we are willing to accept an “error” of ε,
i.e., instead of the actual fixed point p of f we will be statisfied with a
point p 0 of X satisfying d(p, p 0 ) < ε, and suppose also that we start our
iteration at some point x0 in X .

Then from the inequality it is easy to specify an integer N so that

p 0 = f N (x0 ) will be a satisfactory answer. Since we want d(f N (x0 ), p) < ε,
kN
we just have to pick N so large that 1−k d(x0 , f (x0 )) < ε.

Now the quantity D = d(x0 , f (x0 )) is something that we can compute

after the first iteration and we can then compute how large N has to be by
taking the log of the inequality (?) and solving for N (remembering that
log(k) is negative).

P. Sam Johnson Uniform Continuity 160

The result is the stopping rule.

If D = d(x0 , f (x0 )) and

log(ε) + log(1 − k) − log D

N>
log k

then d(f N (x0 ), p) < ε.

Exercise 108.
Prove the inequality (?).

P. Sam Johnson Uniform Continuity 161

Uniformly Continuous Functions
Exercises 109.
1. Suppose X is a metric space and C is a non-empty closed subset of
X . Show that x 7→ dist(x, C ) is uniformly continuous on X .
2. Show that the function f : x 7→ x 2 is uniformly continuous on the set
S = U{[n, n + n−2 ]|n ∈ N}.
3. Show that not every product metric ensures the uniform continuity of
the natural projections.
4. Give an example to show that the image of an open set under a
uniformly continuous map need not be open. Is the same true for
Lipschitz maps? For contractions? For isometric maps?
5. Suppose (X , d) and (Y , ρ) are metric spaces and f : X → Y .
Suppose there exists k > 0 such that e(f (a), f (b)) ≥ kd(a, b) for all
a, b ∈ X . Show that f is injective and that f −1 is a Lipschitz function.
Show also that f is an open mapping if f (X ) is open in Y . Michael-163

P. Sam Johnson Uniform Continuity 162

Uniformly Continuous Functions
Exercises 110.
1. Suppose X is a non-empty set, (Y , ρ) is a metric space and
S ⊆ B(X , Y ). For each x ∈ X , let x̂ denote the function f 7→ f (x)
defined on S (see 9.4.6). Show that {x̂|x ∈ X } is a bounded subset
of e(S, Y ) if, and only if, S is bounded in B(X , Y ).
2. Suppose X and Y are normed linear spaces and f : X → Y is a linear
map. Show that if f is a bounded function, then f = 0.
3. Suppose (X , d) and (Y , ρ) are metric spaces and (fn ) is a sequence of
uniformly continuous functions from X to Y that converges uniformly
to a function g : X → Y . Show that g is uniformly continuous.

Michael-163

P. Sam Johnson Uniform Continuity 163

Uniformly Continuous Functions
Exercises 111.
1. Suppose (X , d) and (Y , ρ) are metric spaces and (fn ) is a sequence of
functions from X to Y that are Lipschitz with Lipschitz constant
k > 0. Suppose that (fn ) converges uniformly to g : X → Y . Show
that g is Lipschitz with Lipschitz constant k.
2. Show that the function x 7→ x + x −1 defined on [1, ∞) is a
contraction that is not strong.

Michael-163

P. Sam Johnson Uniform Continuity 164

References

1. V. Karunakaran, Real Analysis, Pearson, India.

2. Kenneth A. Ross, Elementary Analysis : The Theory of Calculus,
Springer, India.
3. S. Kumaresan, Topology of Metric Spaces, Narosa, India.
4. Robert G. Bartle and Donald R. Sherbert, Introduction to Real
Analysis, Wiley, India.
5. Micheal O Searcoid, Metric Spaces, Springer, India.
6. Stephen Abbott, Understanding Analysis, Springer, India.
7. Walter Rudin, Principles of Mathematical Analysis, McGraw-Hill,
India.
8. Sudhir R. Ghorpade and Balmohan V. Limaye, A Course in
Multivariable Calculus and Analysis, Springer, India.

P. Sam Johnson Uniform Continuity 165