(19) Uniform Continuity

Recall the epsilon-delta definition of f being continuous at x0 : For every  > 0, there is
δ > 0 such that |f (x) − f (x0 )| <  whenever |x − x0 | < δ. Now δ depends on , but it also
depends on c. So if  is particularly small, we might need a smaller δ, but if x0 were a
certain number, we also might need a smaller δ. But sometimes, for every , you can find
δ that works, no matter what x0 is. This is known as uniform continuity.

Definition. We say f is uniformly continuous on an set S if for every  > 0 there is δ > 0
such that |f (x) − f (y)| <  whenever x, y ∈ S and |x − y| < δ.

Example. Consider f (x) = x2 . We can show f is continuous at x0 = 2.

Let  > 0. Suppose |x − 2| < 1. Then |x + 2| < 5. So |x2 − 4| = |x − 2||x + 2| ≤ 5|x − 2|.
So if we let δ = min{1, /5}, we see that |x2 − 4| < . So if  is less than 1 then δ = /5.

Now we consider the continuity of f at x0 = 5. This time, |x − 5| < 1 implies |x + 5| < 11.
So we let δ = min{1, /11}. Then |x2 − 25| < |x − 5||x + 5| < 11|x − 5| < 11/11 = .

So given a small , we needed δ = /3 at x0 = 2 but at x0 = 5, we needed δ = /11.

So δ depends on x0 as well as .

Example. We can show directly that f (x) = x2 is uniformly continuous on I = [0, 5].

Let  > 0. We need δ > 0 such that |x2 − y 2 | <  whenever |x − y| < δ and x, y ∈ [0, 5].

Turns out δ = /10 works. Now if x, y ∈ [0, 5] then |x + y| ≤ |x| + |y| ≤ 5 + 5 = 10.

Suppose |x − y| < δ. Then |x − y| < /10 and |x2 − y 2 | = |x − y||x + y| ≤ 10|x − y| <
10/10 = .

Example. But f (x) = x2 is not uniformly continuous on [0, ∞).

The negation of the definition of uniform continuity is: There exists an  > 0 such that for
every δ > 0 we have x, y ∈ I such that |x − y| < δ but |f (x) − f (y)| ≥ .

For f (x) = x2 on [0, ∞), we can let  = 1. Then no matter how small δ is, we can find x
and y in [0, ∞) such that |x − y| < δ but |x2 − y 2 | ≥  = 1. If we can show this, we will
have shown f is not uniformly continuous on [0, ∞).

Now |x2 − y 2 | = |x − y||x + y|. So we look for x and y that are large but such that
|x − y| < δ. The basic idea is that the graph of f (x) = x2 gets very steep for large x, so
we can find two x’s close to each other but whose y-coords on the curve are more than 1

1
unit apart.

1 1
Let x = δ
and y = δ
+ δ2 . Then |x − y| = δ
2
and
 2  2
2 2 1 δ 1 1 δ2 1 δ2
|x − y | = + − = 2 +1+ − 2 =1+ > 1 = .
δ 2 δ δ 4 δ 4

Here is a vivid way of thinking about uniform continuity. Suppose we have a metal rod,
and suppose f (x) is the temperature at location x on the metal rod. Imagine we have a
special device with two probes that can measure the temperature at two locations at the
same time. Suppose the separation of the two probes is controlled by a knob, so we can
set the two probes to be a given distance δ apart. Suppose the device has a dial on it that
indicates the difference between the temperatures of the two probes. Uniform continuity is
this: Suppose we want the temperature difference between the probes to never be greater
than a given small number . Then we can adjust the δ knob so that the device never
displays a temperature difference greater than epsilon, no matter where on the metal rod
the device is placed.

The following theorem provides some intuition as to what uniform continuity means.

Theorem. Suppose f is continuous on an open interval I (possibly an unbounded interval

such as I = (a, ∞)). Suppose there is a number M such that |f 0 (x)| < M for all x ∈ I.
Then f is uniformly continuous.

Proof. Let  > 0. We seek δ > 0 such that |f (x) − f (y)| <  whenever |x − y| < δ. Now
by the Mean Value Theorem for derivatives, for every x, y ∈ I there is c ∈ I such that
f (x) − f (y)
= f 0 (c).
x−y
So it happens that
|f (x) − f (y)|
= |f 0 (c)| < M
|x − y|
and therefore for all x, y ∈ I we have |f (x) − f (y)| < M|x − y|. So we let δ = /M . So if
|x − y| < δ then |x − y| < /M . So finally |f (x) − f (y)| < M |x − y| < M (/M ) =  and
we are done.

However, it is possible for a function to be uniformly continuous

√ even if its derivative is
unbounded
√ on an interval. A simple example is f (x) = x on [0, 1]. (We have f 0 (x) =
1/(2 x) so limx→0+ f 0 (x) = ∞.) This function is uniformly continuous on this interval.
In fact:

Theorem. Suppose f is continuous on [a, b]. Then f is uniformly continuous on this

interval.

2
Proof. This is another application of the Bolzano-Weierstrass Theorem.

Suppose f is continuous on [a, b] but f is not uniformly continuous on [a, b]. Then there is
a number  > 0 such that for every δ > 0 there are x, y ∈ [a, b] such that |x − y| < δ but
|f (x) − (y)| ≥ .

So let δ = 1/n. Then there are xn , yn in [a, b] such that |xn − yn | < 1/n and
|f (xn ) − f (yn )| ≥ .

Now the sequence {xn } is bounded (being confined to [a, b]), so by the Bolzano-Weierstrass
Theorem, there is a subsequence {xnk } that converges to some c ∈ [a, b].

Since f is continuous, f (xnk ) → f (c). Since |xnk − ynk | < 1/nk ≤ 1/k, we can conclude
that ynk → c as well, so f (ynk ) → f (c).

But since |f (xnk ) − f (ynk )| ≥  for all k, that can’t be. So we have a contradiction, so our
assumption that f is not uniformly continuous cannot be. So f is uniformly continuous
on [a, b], so we are done.

Example. A primary application of uniform continuity is to prove the following theorem:

Rb
Suppose f is continuous on the interval [a, b]. Then the Riemann integral a f (x) dx exists.

One scheme for the Riemann integral is as follows. Let n ∈ N and choose points

a = x0 < x1 < x2 < · · · < xn−2 < xn−1 < xn = b.

The set P = {x0 , x1 , . . . , xn } is called a partition of the interval [a, b]. Define ∆xi =
xi − xi−1 for 1 ≤ i ≤ n. Now define for 1 ≤ i ≤ n the numbers

Mi = max{f (x) : xi−1 ≤ x ≤ xi }

and
mi = min{f (x) : xi−1 ≤ x ≤ xi }.
(These numbers exist by the Extreme Value Theorem.) For this partition P , we define the
upper sum and the lower sum of f by
n
X n
X
UP (f ) = Mi ∆xi and LP (f ) = mi ∆xi
i=1 i=1

respectively.

Now define the upper integral of f over [a, b] by

Z b
f = inf{UP (f ) : P is a partition of [a, b]}
a

3
and define the lower integral by
Z b
f = sup{LP (f ) : P is a partition of [a, b]}.
a

These exist because the sets are bounded (by the extreme value theorem).
Rb Rb
We now say that f is Riemann integrable over [a, b] if a f = a f . We call the common
Rb Rb
value a f or a f (x) dx. A simple criterion for integrability is the following: Suppose for
every  > 0 there is a partition P of [a, b] such that UP (f ) − LP (f ) < . (This is because
Z b Z b
f− f ≤ UP (f ) − LP (f ) < ,
a a

which, since  > 0 is arbitrary, forces the upper and lower integrals to be equal.)

We can now actually prove the theorem:

Proof. Let  > 0. We seek a partition such that UP (f ) − LP (f ) < . Now f is continuous
on [a, b] so f is uniformly continuous on that interval. Therefore, there is δ > 0 such
that x, y ∈ [a, b] and |x − y| < δ implies |f (x) − f (y)| < /(b − a). Now pick n so that
(b − a)/n < δ. For 1 ≤ i ≤ n let ∆xi = ∆x = (b − a)/n (so all the subintervals of
the partition have equal length). Now we define the partition itself for 0 ≤ i ≤ n to be
xi = a + i∆x.

Now mi = f (ui ) and Mi = f (vi ) for some ui , vi ∈ [xi−1 , xi ], by the extreme value theorem.
Therefore,
Xn X n
UP (f ) − LP (f ) = Mi ∆xi − mi ∆xi
i=1 i=1
Xn


= Mi − mi ∆x
i=1
Xn

b − a
= f (vi ) − f (ui )
i=1
n
Xn
 b−a
< =
i=1
b−a n

and we are done.