Series of Functions

Before we study the series of functions, as in the series of real numbers, we recall here first some
definitions and properties concerning sequences of functions.
Definition 1. A sequence of functions {fn } on X ⊆ R is a function whose domain is the set N
and whose range is contained in the set

{fn (x) | fn : X → R, n ∈ N}.

The convergence of the sequence of functions can be classified as pointwise or uniform. We have
the following definitions.
Definition 2. Let fn : X → R. We say that the sequence {fn } converges pointwise to the function
f : X → R , denoted

lim fn (x) = f (x) or fn (x) → f (x) for x ∈ X,

n→∞

if and only if for each x ∈ X and  > 0, there exists N (x, ) ∈ N such that for every n ≥ N ,

|fn (x) − f (x)| < .

Definition 3. Let fn : X → R. We say that the sequence {fn } converges uniformly to the function
f : X → R, denoted by
fn ⇒ f or fn → f (uniformly ) on X,
if and only if for every  > 0, there exists N () ∈ N such that for every n ≥ N ,

|fn (x) − f (x)| < , ∀x ∈ X.

A property relating the two types of convergences is given by the next theorem.
Theorem 1. Uniform convergence implies pointwise convergence of the sequence of functions.
Next, one can show the uniform convergence of the sequence by applying either of the next two
theorems below.
Theorem 2. Let fn be bounded on X for every n. The sequence {fn } converges uniformly to a
function f on X if and only if
kfn − f kx → 0,
where
kf kX = sup{|f (x)|}.
X

Theorem 3 (Cauchy Criterion for Uniform Convergence). Let fn be a bounded function on X for
every n. The sequence {fn } converges uniformly to a bounded function f on X ⇐⇒ ∀ > 0, there
exists N ∈ N such that for all m, n ≥ N

kfn − fm kX < .

The following theorems provide the properties of the limit function depending on the charac-
teristics of the terms of the sequence.
Theorem 4. Let {fn } be a sequence of continuous functions on A s.t. fn ⇒ f on A. Then f is
continuous on A.
Theorem 5. Let {fn } be a sequence of functions integrable on I = [a, b]. If fn ⇒ f on I then,

1
 Z b Z b
f is integrable on I and lim fn = lim fn .
a a

Theorem 6. Let {fn } be a sequence of functions differentiable on I = [a, b] s.t.

(1) ∃ x0 ∈ I s.t. {fn (x0 )} converges,

(2) fn0 ⇒ g (for some g) on I,

then, ∃f differentiable on I s.t. fn ⇒ f on I and f 0 = g on I.

Let us now study the series of functions which are especially useful in obtaining approximations
to a given function and in obtaining new functions from the known ones. This includes the prop-
erties of the limit function that are associated with the uniform convergence of the series in this
section, among others. We start by the following definition.

Definition 4. Let {fn } be a sequence of functions defined on D ⊂ R with values in R.

P
1. The sequence of partial sums, {sn } of the infinite series fn is defined for x in D by
.
s1 (x) = f1 (x)
.
s2 (x) = f1 (x) + f2 (x)
..
.
.
sn+1 (x) = sn (x) + fn+1 (x)
..
. .

2. If the sequence {sn } of functions

P converges (pointwise) on D to a function f , we say that the
infinite series of functions fn converges to f (pointwise) on D. We shall often write

X ∞
X ∞
X
fn , fn , or (fn ),
n=1 n=1

to denote either the limit function or the series, when it exists.

P P
3. If the series |fn (x)| converges for each x ∈ D, we say that fn is absolutely convergent
on D.
P
4. If the sequence {sn } of functions is uniformly convergent on D to f , we say that fn is
uniformly convergent on D, or that it converges to f uniformly on D.
P
Theorem 7. If the series fn converges uniformly to f , then the series converges pointwise to
f.

Some fundamental questions arising are:

1. Is it possible to interchange the integration or differentiation and summation without affecting

the result?
P
2. If sn is the n-th partial sum of the series fn with sn converging to a function f , does the
limit function exhibits the same properties as those of fn ?

The following theorems hand us the analogues of Theorems 4-6. These give the properties of
the limit function and answers the questions raised above.

2
 P
Theorem P 8. If fn is continuous on D ⊂ R to R for each n ∈ N and if fn converges to f
uniformly ( fn ⇒ f ), then f is continuous on D.

Proof:
Since fn is continuous for all n ∈ N, then the partial sums sn of the series are also continuous.
Moreover, the series uniformly converges to some function f means that

sn ⇒ f.

The continuity of f then, follows by invoking Theorem 4. 

Remark 0.1. The continuity of the function fn which assures us the continuity of the partial sums
sn , does not necessarily imply that the limit function f is also continuous.

Example 1. Consider the series

∞ ∞
X X x2
fn (x) = .
(1 + nx2 )(1 + (n − 1)x2 )
n=1 n=1

By partial fractions, we have

x2 1 1
fn (x) = 2 2
= 2
− . (1)
(1 + nx )(1 + (n − 1)x ) 1 + (n − 1)x 1 + nx2

With this formula, considering the partial sums sn gives us a telescoping sum and so

1 nx2
sn = 1 − = .
1 + nx2 1 + nx2
We take note that both fn and sn are continuous at x = 0.
Now, as we take the limit as n → ∞ we get limit function

1, x 6= 0,
f (x) =
0, x = 0,

which implies that sn does not uniformly converges to f (x) = 0. Observe that f is not continuous
at 0 since
lim f (x) = 1 6= 0 = f (0).
x→0

Theorem 9.PSuppose that the real-valued function fn , n ∈ N, are integrable on the interval J =
[a, b]. If the fn ⇒ f on J, then f is integrable and
Z b XZ b
f= fn .
a a

This means, in particular, that the series on the right converges.

Proof: P
Let fn be integrable for all n ∈ N and that fn ⇒ f . This means that the partial sums sn of
the series are also integrable and that
sn ⇒ f.
Thus, f is also integrable by Theorem 5. Moreover, one has
XZ b Z b  Z b Z b
fn = lim sn = lim sn = f,
a a a a

which proves the theorem. 

3
Remark 0.2. For non-uniform convergence, interchanging the order of infinite summation and
integration may fail.

Example 2. Take the series

∞
X
xn−1 (1 − x)(n2 x − (n − 1)2 ), x ∈ (−1, 1],
n=1

and so,
sn = n2 xn (1 − x).

(i) When x 6= 1, we get

f (x) = lim sn = 0, for x ∈ (−1, 1).
n→∞

(ii) If x = 1, sn = 0 so that lim sn = 0.

n→∞

Thus, we have f (x) = 0 but that the convergence of sn to this limit is not uniform.
Now,
Z 1X ∞ Z 1
fn (x)dx = f (x)dx = 0.
0 n=1 0

However,
n Z
X 1 Z n
1X
fi (x)dx = fi (x)dx
i=1 0 0 i=1
Z 1
= sn (x)dx
0
Z 1
2 xn − xn+1 dx


=n
0
n2
= ,
(n + 1)(n + 2)
which converges to 1 as n → ∞ so that
∞ Z
X 1
fn = 1.
n=1 0

Therefore, we have
Z ∞
1X ∞ Z
X 1
0= fn (x)dx 6= fn = 1,
0 n=1 n=1 0

that is, we cannot interchange infinite summation and integration.

Theorem 10. For each n ∈ N, let Pfn be a real-valued function on J = [a, b] that has a derivative
fn0 on J. Suppose
P 0 that the series fn converges for at least one point of J and that the series of
derivatives fn converges uniformly on J. Thus, there exists Pa real-valued function f on J s.t.
fn ⇒ f on J. Moreover, f has a derivative on J and f 0 = fn0 .
P

Proof:
The proof of this theorem follows from its analogue Theorem 6 and is left as an exercise. 

Remark 0.3. As in the previous two remarks, we cannot interchange infinite summation and
integration without uniform convergence.

4
Example 3. Consider the series
∞
X
sin(x)fn (x),
n=1

where fn is given by (1). Let gn = sin(x)fn (x) and note that gn is differentiable. In view of example
1, one has
n sin(x)x2
sn (x) = ,
1 + nx2
and so we have 
sin(x), x 6= 0,
f (x) = lim sn (x) =
0, x = 0.
X∞
= gn (x).
n=1
As one observes, this convergence is not uniform since the convergence to 0 depends on the values
of x.
Now, one can show that gn0 (0) = 0 and thus,
∞
X
gn0 (0) = 0.
n=1

∞
X
However, f 0 (0) = 1 when x 6= 0 and f 0 (0) = 0 when x = 0 which is not the same as gn0 (0).
n=1
Hence, one cannot interchange differentiation and infite summation without uniform convergence.

Like in the sequence of functions, we also have here a criterion for uniform convergence.

Theorem 11 (Cauchy Criterion for Uniform Convergence

P (CCUC)). Let {fn } be a sequence of
functions on D ⊂ R to R . The infinite series fn is uniformly convergent on D iff ∀ε >
0, ∃ M (ε) ∈ N s.t. if m > n ≥ M (ε), then

|fn+1 (x) + · · · + fm (x)| < ε for all x ∈ D.

Proof: Exercise.

Aside from the criterion, an alternative way in showing or determining the uniform convergence
of a series of functions is given by the next test.

Theorem 12 (Weierstrass M-Test). Let {M n } be a sequence of positive

P Preal numbers s.t. |fn (x)| ≤
Mn for all x ∈ D and n ∈ N. If the series Mn is convergent, then fn is uniformly converges
on D.

Proof:
If m > n, we have

|fn+1 (x) + · · · + fm (x)| ≤ Mn+1 + · · · + Mm for x ∈ D.

Applying the CCSeries of real numbers and CCUC gives the conclusion of the Test. 

Remark: One can also conclude from the test the absolute convergence of the series.

Let us now have several examples to demonstrate the properties presented in this section.
∞
X nx2
Example 4. Show that is uniformly convergent on any interval [0, b].
n3 + x3
n=1
Solution:

5
 Let x ∈ [0, b]. Since n3 ≤ n3 + x3 for all n ∈ N and x ∈ [0, b], we have
1 1
≤ 3. (2)
n3 + x3 n

Also, x2 ≤ b2 implies that nx2 ≤ nb2 so that with (2), we have

nx2 nb2 b2
≤ = .
n3 + x3 n3 n2
∞ ∞
X 1 X b2
Now, we note that is a convergent p−series which makes also convergent. Taking
n2 n2
n=1 n=1
∞
b2 X nx2
Mn = 2 in the Weierstrass M-Test, proves the uniform convergence of on the interval
n n3 + x3
n=1
[0, b].
∞
X
Example 5. Investigate the pointwise or uniform convergence of the series fn where
n=1

1 1 1
(a) fn = (b) fn = (c) fn =
xn +1 n2 (x + 1)2 x2 + n2

where x ∈ [0, ∞). If the uniform convergence only happens on a subset of this interval,find this
subset.
Solutions:

(a) We have a number of cases.

(a.1) If x ∈ [0, 1], then lim fn 6= 0 from the previous section. Thus, the series diverges by the
simple test for divergence.
(a.2) When x > 1, then by the Density Theorem, there exists a real number a such that
x > a > 1. This yields xn > an which implies that

xn + 1 > an + 1 > an ,

or that is,
1 1 1
n
> n > n .
a a +1 x +1
∞
1 X 1
Since a > 1, one has 1 > > 0 so that the series converges. By choosing
a an
n=1
∞
1 X 1
Mn = n in the Weierstrass M-Test, the uniform convergence of the series
a xn + 1
n=1
for x > 1 is assured.

(b) For x ∈ [0, ∞), we have n2 ≤ n2 (1 + x)2 so that

1 1
≤ 2.
n2 (1 + x)2 n
∞ ∞
X 1 X 1
Since 2
converges, the uniform convergence of the series follows from the
n n (1 + x)2
2
n=1 n=1
1
Weierstrass M-Test with Mn = 2 .
n

6
 (c) Note that n2 ≤ x2 + n2 for every x ∈ [0, ∞) and n ∈ N. This yields
1 1
≤ 2.
x2 +n 2 n
1
The uniform convergence of the series then, follows from Weierstrass M-Test with Mn = .
n2

Example 5. Show that for all x ∈ (−1, 1),

∞
x2 x3 X xn
log(1 − x) = −x − − − ··· = − . (3)
2 3 n
n=1

Solution:
∞
X
Let a ∈ R be such that 0 ≤ |x| < a < 1 and consider the geometric series xn for |x| ≤ a.
n=0
This gives
∞
1 − xn X 1
sn (x) = and so xn = lim sn = = s(x).
1−x 1−x
n=0

Now, when |x| < a, then x < a and so −a < −x. This implies that 1 − a < 1 − x and that
xn an
< .
1−x 1−a
Thus, in [−a, a] one has
ksn − sk = sup{|sn − s|}
1 − xn
 
1
= sup −
 1− n
x 1−x
x
= sup
1−x
an
= (exercise),
1−a
which converges to 0 as n → ∞, because 0 < a < 1. In other words, sn uniformly converges to s
on [−a, a]. Therefore, in view of Theorem 9, for x ∈ [−a, a] one gets
Z ∞
xX Z x ∞ Z x Z x Z x Z x
n dt X
n
t dt = = t dt = dt + tdt + t2 dt · · · .
0 n=1 0 1−t
n=1 0 0 0 0

That is,
x2 x3
− log(1 − x) = x + + + ··· ,
2 3
which shows (3).
P P
Example 6. Show that if an is absolutely convergent, then an sin(nx) is also absolutely and
uniformly convergent.
Proof: P
Since an is absolutely convergent, then it is convergent. Now,

|an sin(nx)| ≤ |an |, (4)

P P
for all x since | sin(nx)| ≤ 1. Thus,Psince |an | is convergent, then |an sin(nx)| is also convergent
by the Comparison Test. Hence, an sin(nx) is absolutely convergent.
P
Moreover, by taking Mn = |an |, the uniform convergence of an sin(nx) follows from the
Weierstrass M-Test.