0% found this document useful (0 votes)
16 views

Uniform Convergence-chapter4

Uploaded by

Tutul Malakar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
16 views

Uniform Convergence-chapter4

Uploaded by

Tutul Malakar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Chapter 4

Uniform Convergence of Series

4.1 Preamble We say the series of functions




fn (x)
1

converges pointwise on an interval I if the sequence


N

sN (x) = fn (x)
1

of partial sum functions converges pointwise on I. Which is to say that there exists a
sum function s(x) such that
sN (x) → s(x)
pointwise on I. We say


fn (x)
1

is uniformly convergent on I if
sN (x) → s(x)
uniformly on I.
For example, suppose fn (x) = xn . Then
N
 x − xN+1
sN (x) = xn = (x = 1)
1
1−x
=N (x = 1)

which converges to
x
s(x) =
1−x
pointwise over the interval I = (−1, 1). However,

MN = sup |sN (x) − s(x)|


|x|<1
 N+1 
x 
= sup  

|x|<1 1 − x
=∞

for all N so sN (x) → s(x) uniformly on I. Therefore the series




xn
1

25
is pointwise convergent but not uniformly convergent on |x| < 1.
Theorems analogous to those of Chapter 2 also hold for uniformly convergent series.
For example, if all the terms of a uniformly convergent series are continuous functions then
so is the sum function. Also it is legitimate to integrate the series term by term, and to
differentiate it term by term provided the terms are all differentiable and the differentiated
series is uniformly convergent. See 4.7 for details.
We shall be mainly concerned in this Chapter with tests for uniform convergence of
series. We shall give applications to Power Series and to Trigonometric Series in Chapters
5 and 6.
The most important test we shall consider is the Weierstrass M-Test. This test has
the merit of reducing the question of whether a series of functions converges uniformly to
that of whether a series of positive constants converges. And of course elementary analysis
provides us with a large number of tests for resolving this problem. The Weierstrass M-Test
is a test for uniform absolute convergence, meaning

  
fn (x)
1

converges uniformly. There are analogues of Dirichlet’s and Abel’s tests available for testing
for uniform conditional convergence. We also consider tests for non-uniform convergence.
These include the Uniform Non-Null Test and the Uniform Cauchy Criterion. Also non-
continuity of the sum function in the case where the terms are continuous.
The logical order of presentation requires us to cover the Uniform Cauchy Criterion
first since the proof of the Weierstrass M-Test depends on it. We would not wish to
discourage the reader from taking the tests in a different order if he or she prefers.

Exercises Is the series




xn
1

uniformly convergent over (i) −1 < x ≤ 0? (ii) 0 ≤ x ≤ 1/2?

4.2 Uniform Cauchy Criterion In accordance with the tradition laid down in Chap-
ter 3 we shall prove the Uniform Cauchy Criterion in the context of sequences first and
then proceed to series.

Theorem The sequence of functions fn (x) is uniformly convergent on the interval I


if and only if given  > 0 there exists N such that

|fm (x) − fn (x)| < 

for all m, n > N and all x ∈ I.

Proof Suppose fn (x) → f (x) uniformly on I. Then given  > 0 we can choose N
such that
|fm (x) − f (x)| < /2

26
for all n > N and all x ∈ I. Therefore for any m, n > N and x ∈ I we have

|fm (x) − fn (x)| ≤ |fm (x) − f (x)| + |f (x) − fn (x)|


 
< +
2 2
= .

On the other hand suppose fn (x) satisfies the Uniform Cauchy Criterion. Then for
each fixed x ∈ I the numerical sequence fn (x) satisfies the Ordinary Cauchy Criterion so
there exists a limit which we can call f (x). In other words there exists a function f (x)
defined on I such that fn (x) → f (x) pointwise on I. We show fn (x) → f (x) uniformly on
I. In fact, for any given  > 0 we can choose N such that

|fm (x) − fn (x)| < 

for all m, n > N and all x ∈ I. Fix n > N and x ∈ I and let m → ∞. We get

|f (x) − fn (x)| ≤ 

for all n > N and all x ∈ I. Q.E.D.

Corollary The series




fn (x)
1

is uniformly convergent on the interval I if and only if given  > 0 there exists N such that
 Q 
 
 
 fn (x) < 
 
P

for all P, Q > N and all x ∈ I.

Proof Apply the theorem to the sequence

N

sN (x) = fn (x)
1

of partial sums. Q.E.D.

4.3 Uniform Absolute Convergence We say




fn (x)
1

27
is uniformly absolutely convergent on the interval I if


|fn (x)|
1

is uniformly convergent on I.

Theorem Uniform absolute convergence implies uniform convergence.

Proof Suppose


|fn (x)|
1

is uniformly convergent on I and that  > 0 is given. Then there exists N such that
Q

|fn (x)| < 
P

for all P, Q > N and all x ∈ I. But therefore also


 Q 
   Q
 
 f n (x)  ≤ |fn (x)| < 
 
P P

for all P, Q > N and all x ∈ I. Hence




fn (x)
1

is uniformly convergent on I. Q.E.D.

4.4 The Weierstrass M-Test This is easily the most useful of the tests for uniform
convergence and is the easiest to apply. We present it in the form of a theorem.

Theorem The series




fn (x)
1

is uniformly absolutely convergent over the interval I if the series




Mn
1

converges where
Mn = sup |fn (x)|.
x∈I

28
Proof Suppose  > 0 is given. Then by the ordinary Cauchy Criterion there exists N
such that
Q

Mn < 
P

for all P, Q > N . Therefore also


Q Q
 
|fn (x)| ≤ Mn < 
P P

for all P, Q > N and all x ∈ I. Hence




fn (x)
1

is uniformly absolutely convergent over I by the Uniform Cauchy Criterion. Q.E.D.

Example The series




xn
1

is almost uniformly absolutely convergent over |x| < 1.

Proof If I is the interval |x| ≤ 1 −  then

Mn = sup |xn | = (1 − )n


x∈I

and ∞

(1 − )n
1

converges. Therefore by the Weierstrass M-Test the series




xn
1

converges uniformly absolutely on I. Q.E.D.

Exercises Use the Weierstrass M-Test to show the following series are uniformly
absolutely convergent over 0 ≤ x ≤ 1.

 ∞

xn
x2 e−nx
1
n2 1

29
4.5 Tests for Non-Uniform Convergence Consider the example fn (x) = xe−nx .
The series

 ∞

fn (x) = xe−nx
1 1

is pointwise convergent over x ≥ 0. Applying the Weierstrass M-Test we have

Mn = sup xe−nx = 1/ne


x≥0

attained at x = 1/n. But the series



 ∞
 1
Mn =
1 1
ne

diverges. Can we deduce that the series




xe−nx
1

is not uniformly convergent over x ≥ 0?


Unfortunately we can’t. The Weierstrass M-Test only tells us what happens if the
series


Mn
1

converges. It says nothing about what happens when




Mn
1

diverges. To settle the question of the uniform convergence of




xe−nx
1

over x ≥ 0 we have to turn to other tests.


Option 1 We can use the Uniform Cauchy Criterion. To show non-uniform conver-
gence of
∞
fn (x)
1

over the interval I we have to show the Cauchy Criterion fails. This is equivalent to
showing that
Q
fn (x)
P

30
fails to → 0 uniformly over I as P, Q both → ∞. It is sufficient to show

2N

fn (x) → 0
N+1

uniformly over I as N → ∞.
In the case fn (x) = xe−nx over x ≥ 0 we have

2N

MN = sup xe−nx
x≥0
N+1

≥ value at x = 1/N
2N
 1 −n/N
= e
N
N+1
2N
 1 −2N/N
≥ e
N
N+1
2
= 1/e
→ 0

as N → ∞. Therefore
2N

xe−nx → 0
N+1

uniformly over x ≥ 0 and hence




xe−nx
1

is not uniformly convergent over x ≥ 0.

Exercises Use the above method to show non-uniform convergence of the following
series over 0 ≤ x ≤ 1.

 ∞

−nx2
xe xn (1 − x)
1 1

Option 2 is of limited application, though very effective when it can be applied. It


is actually a special case of Option 1, but we shall present it as a theorem in its own right.

Theorem If the series




fn (x)
1

is uniformly convergent over the interval I then its nth term fn (x) → 0 uniformly over I.

31
Proof One can either observe that
N

fN (x) = fn (x)
N

and use the Cauchy Criterion, or argue directly as follows. Suppose

N

sN (x) = fn (x) → s(x)
1

uniformly over I. Then

fn (x) = sn (x) − sn−1 (x) (n ≥ 2)


→0−0
=0

uniformly over I. Q.E.D.

To show non-uniform convergence of a series




fn (x)
1

over an interval I one simply has to show fn (x) → 0 uniformly over I. For example,


xn
1

can be shown to be non-uniformly convergent over |x| < 1 by observing that xn → 0


uniformly over |x| < 1.
Exercises Show the following series are not uniformly convergent over x ≥ 0 by the
above method.
∞ ∞  n
−nx x
e sin nx
1 1
x+1

Option 3 is also of limited application but for a different reason. It depends on the
following theorem.

Theorem If fn (x) is continuous on the interval I for all n and if




fn (x)
1

32
converges uniformly on I then the sum


s(x) = fn (x)
1

is also continuous on I.

Proof The N th partial sum

N

sN (x) = fn (x)
1

is continuous on I from elementary analysis. Hence the result follows from 2.2. Q.E.D.

It follows from this theorem that if the terms of a pointwise convergent series are
all continuous but the sum isn’t continuous then the convergence cannot be uniform.
Unfortunately it is only on very rare occasions that it is possible to obtain a closed form
for the sum of a series.

One example where there is a closed form available for the sum of the series is

 x
xe−nx = (x > 0),
1
ex −1
= 0 (x = 0).

Observe that by l’Hôpital’s rule

x 1
lim = lim x = 1
x→0 ex −1 x→0 e

so the sum is discontinuous at x = 0. Hence we have another proof of the non-uniform


convergence of


xe−nx
1

over x ≥ 0.
Exercises Sum the following series and comment on their uniform convergence over
x ≥ 0.
∞ ∞
−nx2
xe e−nx sin nx
1 1

4.6 Tests for Uniform Conditional Convergence In this section we prove the
uniform analogues of Dirichlet and Abel’s tests. (See 3.4.)

33
Uniform Dirichlet Test The series


an (x)bn (x)
1

converges uniformly over the interval I if


(i) there exists M such that
 
N 
 
 an (x) ≤ M
 
1

for all N ≥ 1 and all x ∈ I,


(ii) bn (x) ≥ bn+1 (x) for all n ≥ 1 and all x ∈ I,
(iii) bn (x) → 0 as n → ∞ uniformly over I.

Proof Write
N

sN (x) = an (x).
1

Then for any P, Q ≥ 1 and any x ∈ I we have


Q Q
   
an (x)bn (x) = sn (x) − sn−1 (x) bn (x)
P P
Q−1
  
= −sP −1 (x)bP (x) + sn (x) bn (x) − bn+1 (x) + sQ (x)bQ (x)
P

and therefore
 Q 
  Q−1
 
 
 a n (x)b n (x)  ≤ M b P (x) + M bn (x) − bn+1 (x) + M bQ (x)
 
P P
= 2M bP (x).

Now suppose  > 0 is given. Then we can choose N such that

|bn (x)| ≤ /2M

for all n > N and all x ∈ I. Therefore


Q 
 
 
 an (x)bn (x) < 
 
P

for all P, Q > N and all x ∈ I. Hence




an (x)bn (x)
1

34
converges uniformly over I by the Uniform Cauchy Criterion. Q.E.D.

As an example of the application of the Uniform Dirichlet Test we show



 sin nx
1
n

is almost uniformly convergent over 0 < x < 2π. We take an (x) = sin nx and bn (x) = 1/n.
If I is the interval  ≤ x ≤ 2π −  then
N 
 
  1 1
 sin nx ≤ ≤
  sin x/2 sin /2
1

for all N ≥ 1 and all x ∈ I. (See 3.4.) Clearly bn (x) ≥ bn+1 (x) for all n ≥ 1 and all
x ∈ I. Also bn (x) → 0 uniformly over I trivially since bn (x) is independent of x. So the
conditions of Dirichlet’s Test for Uniform Convergence are fulfilled and hence

 sin nx
1
n

converges uniformly over  ≤ x ≤ 2π − .

Abel’s Test for Uniform Convergence The series




an (x)bn (x)
1

converges uniformly over the interval I if


(i) the series
∞
an (x)
1

converges uniformly over I,


(ii) bn (x) ≥ bn+1 (x) for all n ≥ 1 and all x ∈ I,
(iii) there exists M such that
|bn (x)| ≤ M
for all n ≥ 1 and all x ∈ I.

Proof We cannot use Dirichlet’s test (c.f. 3.4) since whilst

lim bn (x)
n→∞

certainly exists pointwise over I there is no guarantee that bn (x) → b(x) uniformly over I.

35
Suppose  > 0 is given. Then we can choose N such that
 Q 
 
 
 an (x) < /M
 
P

for all P, Q > N and all x ∈ I. If we write, for N ≥ P ,


N

σN (x) = an (x)
P

then we have
Q Q
   
an (x)bn (x) = σP (x)bP (x) + σn (x) − σn−1 (x) bn (x)
P P +1
.
Q−1
  
= σn (x) bn (x) − bn+1 (x) + σQ (x)bQ (x)
P

Therefore Q 
  Q−1
  
  
 an (x)bn (x) ≤ bn (x) − bn+1 (x) + bQ (x)
  M M
P P

= bP (x)
M
≤
for all P, Q > N and all x ∈ I. Hence


an (x)bn (x)
1

converges uniformly over I. Q.E.D.

As an application of Abel’s Test for Uniform Convergence we show



 xn
(−1)n
1
n

converges uniformly over 0 ≤ x ≤ 1. In fact, we take an (x) = (−1)n /n and bn (x) = xn .


The series

 (−1)n
1
n
converges therefore trivially uniformly over any interval since the nth term is independent
of x. Also bn (x) ≥ bn+1 (x) and 0 ≤ bn (x) ≤ 1 for all n ≥ 1 and all 0 ≤ x ≤ 1. So the
conditions of Abel’s Test are satisfied.

36
Exercises 1. Show

 xn
(−1)n
1
n

is uniformly convergent over 0 ≤ x ≤ 1 by taking an (x) = (−1)n xn and bn (x) = 1/n and
using Dirichlet’s Test for Uniform Convergence.
2. Discuss the uniform convergence of the series

 xn sin nx
1
n

over 0 ≤ x ≤ 1.

4.7 Analytic Consequences of Uniform Convergence We proved in 4.5 that the


sum of a uniformly convergent series of continuous functions is continuous. It now remains
to consider the consequences of uniform convergence for differentiation and integration of
series.

Differentiation Term by Term If fn (x) is a sequence of differentiable functions on


an interval I and if
(i) the series
∞
fn (x)
1

is pointwise convergent on I,
(ii) the series


fn (x)
1

is uniformly convergent on I, then




s(x) = fn (x)
1

is differentiable on I and its derivative is





s (x) = fn (x).
1

Proof If we write
N

sN (x) = fn (x)
1

37
then sN (x) is differentiable on I and

N

sN (x) = fn (x).
1

Also sN (x) → s(x) pointwise on I and the sequence sN (x) is uniformly convergent on I.
Hence the result follows from 2.4. Q.E.D.

Integration Term by Term If fn (x) is a sequence of continuous functions on the


interval [a, b] such that the series


fn (x)
1

converges uniformly on [a, b] then the sum



s(x) = fn (x)
1

is continuous on I and  b ∞ 
 b
s(x)dx = fn (x)dx.
a 1 a

Proof We showed s(x) is continuous in Section 4.5. If we write

N

sN (x) = fn (x)
1

then we have sN (x) → s(x) uniformly on [a, b] and therefore


 b  b
sN (x)dx → s(x)dx
a a

by Section 4.4. But


 b N 
 b
sN (x)dx = fn (x)dx.
a 1 a

Hence
∞ 
 b  b
fn (x)dx = s(x)dx
1 a a

as required. Q.E.D.

38
4.8 Miscellaneous Exercises 1. Discuss the uniform convergence of the following
series over all real x.
∞ ∞
x x
1+n x 2 2 1 + n 3 x2
1 1

 ∞

sinn x cos x (−1)n sinn x cos x
1 1

2. Discuss the continuity of the sums of the following series.



 ∞

sin nx cos nx
√ √
1
n 1
n n


 ∞

n + 1 sin nx sin nx
√ log n √
1
n+2 n 1
n

3. Prove the following assertions about the series




(−1)n xn (1 − x).
1

(i) It is pointwise absolutely convergent over 0 ≤ x ≤ 1.


(ii) It is uniformly convergent over 0 ≤ x ≤ 1.
(iii) It is not uniformly absolutely convergent over 0 ≤ x ≤ 1.
4. Let fn (x) be the isosceles triangle with base [1/(n + 1), 1/n] and height 1/n,
otherwise zero. Show that the series


fn (x)
1

converges uniformly over 0 ≤ x ≤ 1.


Show however that if
Mn = max fn (x)
0≤x≤1

then


Mn
1

diverges.
Why does this not contradict the Weierstrass M-Test?

39

You might also like