MATH 3033 Real Analysis

Sequences and Series of Functions

Dr. Albert Ku

1 Convergence of Functions
Definition 1.1. Let U ⊆ Rm and fn : U → R for n ∈ N. We say the sequence of functions {fn } converges
pointwise on U to the limit function f : U → R if for each x ∈ U , {fn (x)} converges and

f (x) = lim fn (x)

n→∞

Equivalently, for any ε > 0 and x ∈ U , there is an integer N , which may depend on x as well as ε, such that

|fn (x) − f (x)| < ε if n > N

Definition 1.2. Let U ⊆ Rm and fn : U → R for n ∈ N. We say the sequence of functions {fn } converges
uniformly on U to the limit function f : U → R if for any ε > 0, there is an integer N such that

|fn (x) − f (x)| < ε for all x ∈ U if n > N

In fact, the above condition can also be stated as follows:

∥fn − f ∥U = sup {|fn (x) − f (x)| | x ∈ U } < ε if n > N

(In general, for any g : U → R, ∥g∥U = sup {|g(x)| | x ∈ U } is called the sup-norm of g on U .)

Remarks 1.1.

1. If {fn } convereges uniformly on U to f , then it converges pointwise on U to f . However, the converse

is not true.
2. The definition of pointwise convergence of functions implies for each x0 , fn (x0 ) approximates f (x0 )
for sufficiently large n. It does not imply that any particular fn is a good approximation of f over the
entire domain.

1
Example 1.1.
x
1. Let fn (x) = for x ∈ R. Then {fn } converges uniformly to the limit f = 0 on R because for
1 + n2 x2
any x ∈ R,
|x| 1 2n|x| 1
|fn (x) − f (x)| = = · ≤ .
1 + n2 x2 2n 1 + n2 x2 2n
In this example, {fn } is a good approximation of f for sufficiently large n.

nx
2. Let gn (x) = and g(x) = 0 for x ∈ R. Then for any x ̸= 0,
1 + n2 x2
n|x| n|x| 1
|gn (x) − g(x)| = 2 2
≤ 2 2 = → 0 as n → ∞,
1+n x n x n|x|

and since gn (0) = 0 for all n, gn
(x) converges to the limit g(x) = 0 for all x ∈ R.
However, for any n ∈ N, gn n1 = 21 . Thus {gn } does not converge uniformly to g on R i.e. gn does
not approximate g over the entire real line.

More generally, we have the following test for uniform convergence of sequence of functions:
Theorem 1.1. (L-Test) Let {fn } be a sequence of functions on U ⊆ Rm such that it converges pointwise
to f on U . If there exists a sequence of real numbers {Ln } such that for any n ∈ N,

|fn (x) − f (x)| ≤ Ln for all x ∈ U

and lim Ln = 0, then {fn } converges uniformly to f on U .

n→∞

2
We will see that many nice properties of functions can be preserved in the limit of a uniformly convergent
sequence of functions. The following theorem shows that continuity is one such property:
Theorem 1.2. If {fn } converges uniformly to f on U and fn is continuous on U for all n ∈ N, f is continuous
on U .

Proof. Take any x0 ∈ U . For any n ∈ N and x ∈ U ,

|f (x) − f (x0 )| ≤ |f (x) − fn (x)| + |fn (x) − fn (x0 )| + |fn (x0 ) − f (x0 )|
ε
Suppose ε > 0. Since {fn } converges uniformly to f on U , we can choose n so that |fn (x) − f (x)| < 3 for
all x ∈ U . Therefore, we have
2ε
|f (x) − f (x0 )| < |fn (x) − fn (x0 )| + , x∈U
3
ε
Since fn is continuous at x0 , there exist δ > 0 such that |fn (x) − fn (x0 )| < 3 whenever |x − x0 | < δ. Apply
this to the above, we have
ε 2ε
|f (x) − f (x0 )| < + = ε whenever |x − x0 | < δ
3 3
Therefore, f is continuous at x0 .

Remarks 1.2. In the above theorem, the uniform convergence cannot be replaced by pointwise convergence.
The following exercise is an example of a sequence of continuous functions that converges pointwise to a
non-continuous function.

Exercise 1.1. For any n ∈ N, define fn : [0, 1] → R by fn (x) = xn for x ∈ [0, 1]. Show that {fn } converges
pointwise to the following function f on [0, 1]:
(
0 when 0 ≤ x < 1
f (x) =
1 when x = 1

(By the above theorem, {fn } does not converge uniformly on [0, 1].)

3
Theorem 1.3. (Cauchy critierion) The sequence of functions {fn }, defined on U , converges uniformly
on U if and only if for every ε > 0, there exists an integer N such that n, m ≥ N , x ∈ U implies

|fn (x) − fm (x)| < ε

Proof.
(⇒) Suppose {fn } converges uniformly on U , and let f be the limit function. Then there exists N ∈ N such
that n ≥ N , x ∈ U implies
ε
|fn (x) − f (x)| <
2
Then we have
ε ε
|fn (x) − fm (x)| ≤ |fn (x) − f (x)| + |f (x) − fm (x)| < + = ε
2 2
if n, m ≥ N and x ∈ U .

(⇐) Suppose the Cauchy criterion holds. Then for any x ∈ U , the sequence {fn (x)} converges, to a limit
which we may call f (x). Thus {fn } converges pointwise to f on U . We have to prove that the convergence
is uniform.
Let ε > 0, and choose N such that |fn (x)−fm (x)| < 2ε for all n, m ≥ N and x ∈ U . Also, since fn (x) → f (x)
as n → ∞, there exists a sufficiently large m (which might depend on x) such that m > N and
ε
|fm (x) − f (x)| <
2
Then for any n ≥ N and x ∈ U , we have
ε ε
|fn (x) − f (x)| ≤ |fn (x) − fm (x)| + |fm (x) − f (x)| < + =ε
2 2
Consequently, {fn } converges to f uniformly on U .

4
 ∞
X
n
Definition 1.3. If {fn } is a sequence of functions defined on U ⊆ R , then fn is said to be an infinite
n=0
∞
X
series of functions on U . The partial sums of fn are defined by
n=0

n
X
Fn = fk
k=0

∞
X
If {Fn } converges pointwise to a function F on U , we say that fn converges pointwise to the sum F
n=0
on U , and write
∞
X
F (x) = fn (x)
n=0
∞
X
If {Fn } converges uniformly to a function F on U , we say that fn converges uniformly to the sum F
n=0
on U .
∞
X
Example 1.2. The functions fk (x) = xk , k = 0, 1, 2, . . . define the infinite series xk on R. Its nth partial
k=0
sum is
 1 − xn+1

2 n , x ̸= 1
Fn (x) = 1 + x + x + · · · + x = 1−x
n + 1, x=1


It is easy to see that {Fn } converges pointwise to

1
F (x) = , if |x| < 1
1−x
Hence, we write
∞
X 1
xk = , −1 < x < 1
1−x
k=0

For any fixed n, since the difference

xn+1
F (x) − Fn (x) =
1−x
can be made arbitrarily large by taking x close to 1, and so the convergence is not uniform on (−1, 1).
However, the series does converge uniformly on any interval [−r, r] with 0 < r < 1, since for any x ∈ [−r, r],

rn+1
|Fn (x) − F (x)| ≤ → 0 as n → ∞.
1−r

5
The following are two tests for uniform convergence of series:
∞
X
Theorem 1.4. (Cauchy criterion for uniform convergence of series) A series fn converges uni-
n=0
formly on U if and only if for each ε > 0, there exists N ∈ N such that for any x ∈ U and m ≥ n > N ,

|fn (x) + fn+1 (x) + · · · + fm (x)| < ε

Proof. Follows from applying Cauchy criterion on the sequence of partial sum {Fn }.

Theorem 1.5. (Weierstrass M-test) Suppose |fn (x)| ≤ Mn for all x ∈ U i.e. fn is bounded by Mn .
∞
X ∞
X
Then, if Mn converges, fn converges uniformly on U .
n=0 n=0

∞
X
Proof. Since the partial sum of Mn converges, it is a Cauchy sequence, which means that for any ε > 0,
n=0
there exists N ∈ N such that if m ≥ n > N ,

Mn + Mn+1 + · · · + Mm < ε,

which implies that for any x ∈ U ,

|fn (x) + fn+1 (x) + · · · + fm (x)| ≤ |fn (x)| + |fn+1 (x)| + · · · + |fm (x)| ≤ Mn + Mn+1 + · · · + Mm < ε
∞
X
By Theorem 1.4, fn converges uniformly on U .
n=0

∞ ∞
X 1 X sin(nx) 1
Example 1.3. The series 2 + n2
and 2
converges uniformly on R by considering Mn = 2 .
n=0
x n=1
n n

6
 ∞
X ∞
X
Definition 1.4. fn is said to converge absolutely pointwise/uniformly on U if |fn | converges
n=0 n=0
pointwise/uniformly on U .

∞
X
Exercise 1.2. If a series fn satisfies the Weierstrass M-test, then it converges absolutely uniformly.
n=0

∞
X 1
Exercise 1.3. Consider the series x
.
n=1
n

(a) Given any p > 1, prove that the series converges uniformly on [p, ∞).
∞
X 1
(b) Define ζ(x) = for any x > 1. Prove that ζ(x) is continuous on (1, ∞).
n=1
nx

7
2 Power Series
In this section, we will study a very special kind of series of functions:
∞
X
Definition 2.1. Let a ∈ R and {an } be a sequence of real numbers. Then the series an (x − a)n is called
n=0
a power series. a is called the center of the power series.

A very natural question about power series is to determine the set of values of x on which a given power
series converges pointwise (or uniformly). To answer this question, we first need the following definitions:

Definition 2.2. Given {an } a sequence of real numbers. The limit superior of {an } is the limit of
Mn = sup{am | m ≥ n} as n → ∞:

lim sup an = lim sup{am | m ≥ n} = lim Mn

n→∞ n→∞ n→∞

Similarly, the limit inferior of {an } is the limit of mn = inf{am | m ≥ n} as n → ∞:

lim inf an = lim inf{am | m ≥ n} = lim mn

n→∞ n→∞ n→∞

It is always true that lim inf an ≤ lim sup an .

n→∞ n→∞

Remarks 2.1.
1. If {an } is unbounded above, then Mk = ∞ for all k ∈ N and lim sup an = lim Mk = ∞.
n→∞ k→∞

2. If {an } is unbounded below, then mk = −∞ for all k ∈ N and lim inf an = lim mk = −∞.
n→∞ k→∞

3. Suppose {an } is bounded. By definition, {Mn } is a decreasing sequence and is bounded below. There-
fore, lim sup an = lim Mn exists. Similarly, {mn } is an increasing sequence and is bounded above.
n→∞ n→∞
Therefore, lim inf an = lim mn exists.
n→∞ n→∞

4. For any sequence {an }, we have mn ≤ an ≤ Mn for all n ∈ N. If lim sup an = lim inf an = L, then
n→∞ h→∞
lim an = L by sandwich theorem.
n→∞

8
Theorem 2.1. Let H = lim sup an and h = lim inf an for a bounded sequence {an }. For any ε > 0,
n→∞ n→∞

(a) there are infinitely many n ∈ N such that an ∈ (H − ε, H + ε), and there are at most finitely many
n ∈ N such that an > H + ε;
(b) there are infinitely many n ∈ N such that an ∈ (h − ε, h + ε), and there are at most finitely many n ∈ N
such that an < h − ε.
Proof. To prove part (a), we first prove that there are infinitely many n ∈ N such that an > H − ε for
any ε > 0. Suppose not, then there is an ε0 > 0 such that there are only finitely many n ∈ N such
that an > H − ε0 . This implies that there exists N ∈ N such that an ≤ H − ε0 for all n > N . Hence
Mn = sup{am | m ≥ n} ≤ H − ε0 for all n > N . Thus, H = lim Mn = H ≤ H − ε0 , a contradiction.
n→∞

We next prove that there are only finitely many n ∈ N such that an > H + ε for any ε > 0. For any given
ε > 0, since lim Mn = H, there exists N ∈ N such that Mn < H + ε for n > N . Since an ≤ Mn for any n,
n→∞
an < H + ε for n > N .

The proof of part (b) is similar.

Corollary 2.1. Let H = lim sup an and h = lim inf an for a bounded sequence {an }.
n→∞ n→∞

(a) There is a subsequence converging to h and a subsequence converging to H.

(b) If L = {l ∈ R| l is a limit of some convergent subsequence of {an }}, then h and H are respectively the
smallest and the largest element of L.
Proof.

(a) Since for each k ∈ N there are infinitely many n ∈ N such that
an ∈ H − k1 , H + k1 , we can pick

successively for each k ∈ N an nk such that ank ∈ H − k1 , H + k1 and nk > nk−1 . The subsequence
{ank } clearly converges to H.
Similarly, we can construct a subsequence converging to h.
(b) Let l be the limit of some subsequence {ank }. It suffices to prove that h ≤ l ≤ H. For any ε > 0,
there exists K ∈ N such that ank > l − ε for all k > K. But there are only finitely many nk such that
ank > H + ε. Hence l − ε ≤ H + ε. This implies l ≤ H + 2ε for any ε > 0, i.e. l ≤ H.
Similarly, we can prove that h ≤ l.

9
Remarks 2.2.
1. Bolzano-Weierstrass Theorem (on R) follows immediately from part (a) of the above corollary.
2. By part (b), we have lim sup an = sup L and lim inf an = inf L. Therefore, we have the following nice
n→∞ n→∞
properties:
ˆ For any c ∈ R such that c > 0, we have

lim sup(can ) = sup(cL) = c sup L = c lim sup an

n→∞ n→∞

lim inf (can ) = inf(cL) = c inf L = c lim inf an

n→∞ n→∞

ˆ More generally, if {cn } is a sequence such that lim cn = c > 0, then we have
n→∞

lim sup(cn an ) = c lim sup an

n→∞ n→∞

lim inf (cn an ) = c lim inf an

n→∞ n→∞

10
This is the main result we will use in this section is as follows:
Theorem 2.2. (Strong Form of Root Test) Suppose {an } is a sequence in R.
∞
1
X
1. lim sup |an | n < 1 ⇒ an converges absolutely.
n→∞
n=1

∞
1
X
2. lim sup |an | n > 1 ⇒ an diverges.
n→∞
n=1
1
Proof. To prove the strong form of root test, we let M = lim sup |an | n . If M < 1, then choose r such that
n n→∞ o
1
M < r < 1. Since M = lim Mk , where Mk = sup |an | | n ≥ k , there exists K ∈ N such that MK < r.
n
k→∞
1
This implies that |an | n < r, or equivalently, |an | < rn , for any n ≥ K. Hence, we have
∞
X ∞
X
|an | ≤ rn < ∞
n=K n=K

∞
X
which means that |an | converges.
n=1
1
If M > 1, then there exists a subsequence of |ank | nk that converges to M , which means that there exist
1
K ∈ N such that for all k > K, |ank | nk > 1. This implies that |ank | > 1 for all k > K and hence lim an ̸= 0.
n→∞
∞
X
By the test for divergence, an diverges.
n=1

Remarks 2.3. Sometimes it is more convenient to use ratio test. In fact, if an ̸= 0 for all n ∈ N, we have
the following inequalites:

an+1 1 1 an+1
lim inf ≤ lim inf |an | n ≤ lim sup |an | n ≤ lim sup
n→∞ an n→∞ n→∞ n→∞ an

The above inequalities imply the following strong form of ratio test: If {an } is a sequence such that
an ̸= 0 for all n ∈ N. Then
∞
an+1 X
1. lim sup <1⇒ an converges absolutely.
n→∞ an n=1

∞
an+1 X
2. lim inf >1⇒ an diverges.
n→∞ an n=1

an+1 1
Moreover, if lim exists, then it is equal to lim |an | n . (Proof omitted)
n→∞ an n→∞

11
The following is the main theorem about the convergence of power series:
∞
X
Theorem 2.3. For any power series an (x − a)n , there exists ρ such that 0 ≤ ρ ≤ +∞ and
n=0

(a) if 0 < r < ρ, the series converges absolutely uniformly on [a − r, a + r],

(b) if |x − a| > ρ, the series diverges at x.

Proof. Let ρ be defined by
1 1 1 1
ˆ Case 1: = lim sup |an | n if {|an | n } is bounded and lim sup |an | n ̸= 0,
ρ n→∞ n→∞
1 1
ˆ Case 2: ρ = ∞ if {|an | n } is bounded and lim sup |an | n = 0, and
n→∞
1
ˆ Case 3: ρ = 0 if {|an | } is not bounded above.
n

Obviously, we have
|an (x − a)n | ≤ |an |rn
for |x − a| ≤ r. Therefore, by Weierstrass M-test, the power series converges absolutely uniformly on
∞
X
[a − r, a + r] if |an |rn converges.
n=0
∞
1
X
By the strong form of root test, |an |rn converges if lim sup(|an |rn ) n < 1, which is equivalent to
n→∞
n=0

1
r lim sup |an | n < 1
n→∞

It clearly holds if r < ρ for Case 1 and 2. This proves part (a) of the theorem.

1 |x − a|
For Case 1 and 3, if |x − a| > ρ, then |x − a| lim sup |an | n = > 1. Then we have
n→∞ ρ
1 1
lim sup |an (x − a)n | n = |x − a| lim sup |an | n > 1
n→∞ n→∞

By the strong form of root test, the power series diverges.

12
Definition 2.3. The value of ρ (necessarily unique) in Theorem 2.3 is called the radius of convergence
∞
X
of power series an (x − a)n .
n=0

Remarks 2.4.
1. No general statement can be made concerning convergence of a power series at x = a ± ρ i.e. some
series may converge at x = a ± ρ while some other series may diverge.
2. In general, a power series does not converge uniformly on (a − ρ, a + ρ). For example, ρ = 1 for the
P∞ 1
series n=0 xn and the series converges to for |x| < 1, diverges for |x| ≥ 1, and the series does
1−x
not converge uniformly on (−1, 1).

∞
X
Theorem 2.4. If f (x) = an (x − a)n has radius of convergence ρ > 0, then f is differentiable for
n=0
|x − a| < ρ with
∞
X
f ′ (x) = nan (x − a)n−1 .
n=1

The power series on the right also has radius of convergence ρ.

∞
X
Proof. Here we only prove that the power series nan (x − a)n−1 has radius of convergence ρ. The proof
n=1
of the convergence to f ′ (x) will be deferred to the next section.

∞
X
Let ρ1 be the radius of convergence of the power series nan (x − a)n−1 . Then we use the argument similar
n=1
to the proof of Therorem 2.3. Let r > 0. On [a − r, a + r], we have

|nan (x − a)n−1 | ≤ |nan |rn−1

∞
X
Then by Weierstrass M-test, the power series converges uniformly on [a − r, a + r] if |nan |rn−1 converges.
n=1
∞
1
X
By the strong form of root test, |nan |rn−1 converges if lim sup(|nan |rn−1 ) n < 1, which is equivalent to
n→∞
n=1

n−1 1 1 1
lim sup(r n n n |an | n ) = r lim sup |an | n < 1
n→∞ n→∞

Therefore, ρ1 = ρ.

13
 ∞
X
Corollary 2.2. If f (x) = an (x − a)n has radius of convergence ρ > 0, then f is infinitely differentiable
n=0
for |x − a| < ρ with
∞
X
f (k) (x) = n(n − 1) · · · (n − k + 1)an (x − a)n−k .
n=k
(k)
for any k ∈ N. In particular, f (a) = k!ak .
Proof. Apply Theorem 2.4 k times.

∞
X
Given a power series an (x − a)n with radius of convergence ρ > 0, we can define f (x) to be the power
n=0
series and by Corollary 2.2, f is infinitely differentiable on (a − ρ, a + ρ) with f (k) (a) = k!ak . It is natural to
consider the following question: Suppose we are given a function f that is infinitely differentiable on an open
interval containing a, can it be expressed as a power series centered at a? This is the motivation behind the
following definition:
Definition 2.4. Let a ∈ R, ρ > 0 and f : (a − ρ, a + ρ) → R be a function that is infinitely differentiable.
The Taylor series of f centered at a is
∞
X f (n) (a)
(x − a)n
n=0
n!

The partial sum of the Taylor series up to the k th term is called the Taylor polynomial of f of degree k
centered at a:
k
X f (n) (a)
(x − a)n
n=0
n!

14
It is easy to see that if an infinitely differentiable function f can be expressed as a power series, then the
power series must be the Taylor series of f . However, the following example shows that there exists an
infinite differentiable function that is not equal to its Taylor series.
Example 2.1. Let f : R → R be defined by
( 2
e−1/x when x ̸= 0
f (x) =
0 when x = 0

It can be shown that for any k ∈ N, f (k) (0) = 0. Therefore, the Taylor series of f centered at 0 is 0, which
is certainly not equal to f .

If an infinitely differentiable function equals its Taylor series, it is called an analytic function.

The following theorem expresses the error term explicitly when a function is approximated by its Taylor
polynomial:
Theorem 2.5. Let f : (a − ρ, a + ρ) → R be a function such that its (k + 1)th derivative is continuous at
a, where k is a non-negative integer. Then for any x ∈ (a − ρ, a + ρ),
k
X f (n) (a)
f (x) = (x − a)n + Rk+1 (x)
n=0
n!

where
f (k+1) (c)
(Mean-value form) Rk+1 (x) = (x − a)k+1 , for some c between a and x
(k + 1)!
1 x (k+1)
Z
(Integral form) Rk+1 (x) = f (t)(x − t)k dt
k! a
Proof. The mean-value form was already proved in MATH 2033. Here we will only prove the integral form.
For each x, we apply integration by parts repeatedly to obtain the following:
Z x
f (x) = f (a) + f ′ (t)dt
a
x
Z x
′
= f (a) + f (t)(t − x) + f ′′ (t)(x − t)dt
a a
Z x
′
= f (a) + f (a)(x − a) + f ′′ (t)(x − t)dt
a
Z x
′ ′′ (x − t)2 x (x − t)2
= f (a) + f (a)(x − a) − f (t) + f ′′′ (t) dt
2! a a 2!
x
f ′′ (a) (x − t)2
Z
= f (a) + f ′ (a)(x − a) + (x − a)2 + f ′′′ (t) dt
2! a 2!
···
k
f (n) (a) 1 x (k+1)
X Z
= (x − a)k + f (t)(x − t)k dt.
n=0
n! k! a

15
The following corollary gives a sufficient condition for an infinitely differentiable function to be analytic:
Corollary 2.3. Let f : (a − ρ, a + ρ) → R be an infinitely differentiable function. Moreover, if there exists
r ∈ (0, ρ) such that
ˆ for any n ∈ N, there exists Mn > 0 such that |f (n) (x)| ≤ Mn for all x ∈ [a − r, a + r], and
Mn r n
ˆ → 0 as n → ∞,
n!
then the Taylor series of f centered at a has radius of convergence at least r and converges uniformly to f
on [a − r, a + r].
Proof. Let fk (x) be the k th Taylor polynomial of f centered at a. Then by Theorem 2.5, for x ∈ [a − r, a + r],

f (k+1) (c)
|fk (x) − f (x)| ≤ (x − a)k+1 for some c between x and a
(k + 1)!
Mk+1 rk+1
≤
(k + 1)!

Mk+1 rk+1
By assumption, → 0 as n → ∞. Hence, the Taylor series converges uniformly to f on [a−r, a+r].
(k + 1)!
Since the Taylor series is a power series, its radius of convergence must be at least r.

Example 2.2. Special functions like ex , sin x, cos x are analytic functions i.e. they are equal to their own
Taylor series because they satisfy the conditions in Corollary 2.3.
∞
X xn
ex =
n=0
n!
∞
X (−1)n x2n+1
sin x =
n=0
(2n + 1)!
∞
X (−1)n x2n
cos x =
n=0
(2n)!

All the above Taylor series have radius of convergence ∞.

16
Another important result is the generalization of Binomial Theorem:
Theorem 2.6. (Binomial Theorem) For any a ∈ R, x ∈ (−1, 1),
∞
a(a − 1) 2 X a(a − 1) · · · (a − k + 1)
(1 + x)a = 1 + ax + x + ··· = 1 + xk
2 k!
k=1

Proof. Let f (x) = (1 + x)a . Then f (k+1) (x) = a(a − 1) · · · (a − k)(1 + x)a−k−1 . By Theorem 2.5, we have

1 x (k+1)
Z
|Rk+1 (x)| = f (t)(x − t)k dt
k! 0
k
a(a − 1) · · · (a − k) x x − t
Z 
= (1 + t)a−1 dt
k! 0 1 + t

x−t
Claim: ≤ |x| for any t in the closed interval with 0 and x as endpoints.
1+t
x−t −1 − x
Let g(t) = . g(0) = x and g ′ (t) = < 0. Hence g is decreasing and we have
1+t (1 + t)2

g(t) ≤ g(0) = x if x > 0

g(t) ≥ g(0) = x if x < 0

x−t
Therefore, we get ≤ |x| for any t in the closed interval with 0 and x as endpoints and we have
1+t
Z x
a(a − 1) · · · (a − k) k
|Rk+1 (x)| ≤ |x| (1 + t)a−1 dt = bk+1
k! 0

In order to show that lim |Rk+1 | = 0, we consider the following:

k→∞

∞
bk+1 |a − k| X
lim = lim |x| = |x| < 1 ⇒ bk converges by ratio test
k→∞ bk k→∞ k
k=1

⇒ lim bk = 0 by term test ⇒ lim |Rk+1 (x)| = 0.

k→∞ k→∞

17
3 Uniform Convergence and Differentiation
Suppose {fn } is a sequence of differentiable functions such that it converges uniformly f . Then it is natural
to ask the following questions:
ˆ Is f differentiable?

ˆ If f is differentiable, will {fn′ } converge uniformly to f ′ ?

Let’s consider the following examples.

Example 3.1. For any n ∈ N, we define
sin(nx)
fn (x) = √
n
Then it can easily be shown that {fn } converges uniformly on R to f = 0. However, we have
√
fn′ (x) = n cos(nx)
√
so that {fn′ } does not converge pointwise to f ′ = 0. For instance, fn′ (0) = n → ∞ as n → ∞. But
f ′ (0) = 0.

Example 3.2. For any n ∈ N, we define

1 1


 −x − if − 1 ≤ x ≤ −

 2n n
n 2 1 1

fn (x) = x if − < x <
 2 n n
 x− 1 1


if ≤x≤1

2n n
{fn } is a sequence of differentiable functions on [−1, 1] and it can be shown that {fn } converges uniformly
on [−1, 1] to f (x) = |x|. But clearly f (x) = |x| is not differentiable at x = 0.

18
Therefore, uniform convergence of differentiable functions does not guarantee the differentiability of the limit
function. We need much stronger conditions. The following is the main theorem of this section:
Theorem 3.1. Suppose {fn } is a sequence of functions, differentiable on [a, b] and such that {fn (x0 )}
converges for some point x0 ∈ [a, b]. If {fn′ } converges uniformly on [a, b], then {fn } converges uniformly on
[a, b] to a function f , and for any x ∈ [a, b],

f ′ (x) = lim fn′ (x)

n→∞

Proof. First, we need to prove that {fn } converges uniformly on [a, b]. Let ε > 0. Since {fn (x0 )} converges
and {fn′ } converges uniformly on [a, b], there exists N ∈ N such that for m, n ≥ N ,
ε
|fn (x0 ) − fm (x0 )| <
2
and
ε
|fn′ (t) − fm
′
(t)| < (1)
2(b − a)
for any t ∈ [a, b]. Using mean value inequality on the function fn − fm and the above inequality, we have
ε ε
|fn (x) − fm (x) − (fn (t) − fm (t))| ≤ |x − t| ≤ (2)
2(b − a) 2

for any x, t ∈ [a, b]. Combining the inequalities, for any x ∈ [a, b],

|fn (x) − fm (x)| ≤ |fn (x) − fm (x) − (fn (x0 ) − fm (x0 ))| + |fn (x0 ) − fm (x0 )|
ε ε
< + =ε
2 2
By Cauchy criterion, {fn } converges uniformly on [a, b].

Now, we define f (x) = lim fn (x) for any x ∈ [a, b] i.e. {fn } converges uniformly to f on [a, b]. We need to
n→∞
show that for any x ∈ [a, b],
f ′ (x) = lim fn′ (x)
n→∞

Fix any x ∈ [a, b]. Define 

 fn (t) − fn (x)
if t ̸= x
ϕn (t) = t−x
f ′ (x) if t = x
n

By definition, ϕn (t) is continuous on [a, b]. Then by the inequalities (1) and (2) again, for t ∈ [a, b],
ε
|ϕn (t) − ϕm (t)| ≤
2(b − a)

Therefore, {ϕn } converges uniformly on [a, b] and moreover, the limit function ϕ is continuous on [a, b] by
f (t) − f (x)
Theorem 1.2. Obviously, for t ̸= x, {ϕn (t)} converges pointwise to ϕ(t) = . By the continuity of
t−x
the limit function, {ϕn (x)} converges to ϕ(x) = lim ϕ(t). Hence, we have
t→x

f (t) − f (x)
f ′ (x) = lim = ϕ(x) = lim ϕn (x) = lim fn′ (x)
t→x t−x n→∞ n→∞

19
Remarks 3.1. We are now ready to complete the proof of Theorem 2.4: Consider the power series f (x) =
∞
X n
X
ak (x − a)k with radius of convergence ρ > 0. Let fn (x) = ak (x − a)k . Obviously, for 0 < r < ρ, fn is
k=0 k=0
differentiable on [a − r, a + r] and we already proved that
∞
X
kak (x − a)k−1
k=1

is also a power series with radius of convergence ρ, which implies that {fn′ } converges uniformly on [a−r, a+r].
By Theorem 3.1,
X∞
′
f (x) = kak (x − a)k−1
k=1

for any x ∈ [a − r, a + r]. Since r can be freely chosen between 0 and ρ, the above equality holds for any
x ∈ (a − ρ, a + ρ).

20
4 Uniform Convergence and Integration
In this section, we first define rigorously the definite integral of a real-valued function that we learned before.
Definition 4.1. Let f : [a, b] → R be a bounded real-valued function. A partition P of [a, b] is a finite set
of points t0 , t1 , . . . , tn such that
a = t0 < t1 < · · · < tn = b
For j = 1, 2, . . . , n, define
Mj = sup f ([tj−1 , tj ]), mj = inf f ([tj−1 , tj ]),
and define
n
X n
X
L(f, P ) = mj (tj − tj−1 ) U (f, P ) = Mj (tj − tj−1 )
j=1 j=1

Then L(f, P ) is called the lower Riemann sum for f corresponding to P and U (f, P ) is the upper Rie-
mann sum for f corresponding to P .

Clearly, we have U (f, P ) ≥ L(f, P ).

If we write L(f ) = sup{L(f, P )| P is a partition}, U (f ) = inf{U (f, P )| P is a partition}, then it can be
shown that for any partition P and P ′ of [a, b],

L(f, P ) ≤ L(f ) ≤ U (f ) ≤ U (f, P ′ ).

Definition 4.2. L(f ) is called the lower Riemann integral of f on [a, b]; U (f ) is called the upper
Riemann integral of f on [a, b]. If L(f ) = U (f ), then f is said to be Riemman integrable on [a, b], and
Z b Z b
the common value of L(f ) and U (f ) is written f (x) dx or f , and is called the Riemann integral of
a a
f on [a, b].
Remarks 4.1. If f is bounded on [a, b], then there exist m and M such that m ≤ f (x) ≤ M for all x ∈ [a, b].
It follows easily that
m(b − a) ≤ L(f ) ≤ U (f ) ≤ M (b − a).
If f is Riemann integrable, then it follows that
Z b
m(b − a) ≤ f ≤ M (b − a).
a

21
The following theorem gives a condition for the Riemann integrability of a function:
Theorem 4.1. If f is bounded on [a, b], then f is Riemann integrable on [a, b] if and only if, given any
ε > 0, there is a partition P of [a, b] such that

U (f, P ) − L(f, P ) < ε. (3)

Remarks 4.2.
1. The following two general classes of functions are Riemann integrable:
ˆ Continuous functions on [a, b]
ˆ Monotonic functions on [a, b]

2. Define f : [0, 1] → R by (
1 if x is rational
f (x) =
0 if x is irrational
Then it can easily be shown that for any partition P of [0, 1], U (f, P ) = 1 and L(f, P ) = 0. Therefore,
we have
U (f ) − L(f ) = 1
and by Theorem 4.1, f is not Riemann integrable
3. In the next chapter, we will introduce Lebesgue Integral, which is more general than Riemann
integral. For example, f is not Riemann integrable but is “Lebesgue integrable”.

22
Now we are ready to prove the main theorem of this section:
Theorem 4.2. Let {fn } be a sequence of Riemann integrable functions defined on [a, b] such that it converges
uniformly to f on [a, b]. Then f is also Riemann integrable and
Z b Z b
lim fn = f
n→∞ a a

Proof. For any ε > 0, since {fn } converges uniformly to f on [a, b], there exists N ∈ N such that for n ≥ N ,
ε
|fn (x) − f (x)| < for all x ∈ [a, b] (4)
3(b − a)

Also, as fN is Riemann integrable, there exists a partition P of [a, b] such that

ε
U (fN , P ) − L(fN , P ) <
3
By (4), we have
ε ε
L(fN − , P ) < L(f, P ) ≤ U (f, P ) < U (fN + ,P)
3(b − a) 3(b − a)
Combining the inequalities, we have
ε ε
U (f, P ) − L(f, P ) < U (fN + , P ) − L(fN − ,P)
3(b − a) 3(b − a)
2ε
= U (fN , P ) − L(fN , P ) + (b − a)
3(b − a)
ε 2ε
< + =ε
3 3
Therefore, f is Riemann integrable. Also, for n ≥ N ,
Z b Z b Z b Z b
ε ε
fn − f ≤ |fn − f | < = <ε
a a a a 3(b − a) 3
(Z )
b Z b
which implies that fn converges to f.
a a

23
Example 4.1. It can easily be shown that for each r ∈ (0, 1), the geometric series
∞
X 1
(−1)n x2n converges uniformly to on [−r, r].
n=0
1 + x2

Therefore, by Theorem 4.2, we can integrate the power series term by term and obtain the following:
∞
X (−1)n x2n+1
converges uniformly to arctan x on [−r, r]
n=0
2n + 1

In particular, for any x ∈ (−1, 1),

∞
X (−1)n x2n+1
arctan(x) =
n=0
2n + 1

Question: Does the above equality holds when x = 1?

24
5 Abel’s Limit Theorem
The following are some useful properties about uniform convergence:

Theorem 5.1. Suppose {fn } converges uniformly to f on U .

1. (Subset Property) If V ⊆ U , {fn } converges uniformly to f on V .

2. (Union Property) If {fn } also converges uniformly to f on W , then it converges uniformly to f on

U ∪ W.
3. (Bounded Multiplier Property) If g(x) is a bounded function on U , {gfn } converges uniformly to
gf on U .

4. (Substitution Property) If h : U0 → U , then {fn ◦ h} converges uniformly to f ◦ h on U0 .

Proof. Exercise

Are there specials cases where pointwise convergence of nice functions implies uniform convergence? When
the domain is a closed and bounded interval, we have the following theorem:

∞
X
Theorem 5.2. (Abel’s limit theorem) Suppose the power series ak (x − a)k converges pointwise to
k=0
S(x) on a closed and bounded interval [u, v], then it converges uniformly to S(x) on [u, v]. Moreover, we
have
X∞ X∞
lim+ S(x) = S(u) = ak (u − a)k and lim− S(x) = S(v) = ak (v − a)k
x→u x→v
k=0 k=0

n
X
Proof. We first consider the special case when a = 0 and [u, v] = [0, 1]. Let sn = ak . Since lim sn =
n→∞
k=0
∞
X
ak = S(1), {sn } is Cauchy and hence for any ε > 0, there exists K ∈ N such that for j > n ≥ K,
k=0

j
X ε
|sj − sn | = ak <
2
k=n+1

25
 j
X
Define Aj = ak for j > n and An = 0. Then for any x ∈ [0, 1] and m > n ≥ K, we have
k=n+1

m
X m
X m
X m
X
a k xk = (Ak − Ak−1 )xk = Ak x k − Ak−1 xk
k=n+1 k=n+1 k=n+1 k=n+1
m
X m−1
X m−1
X
= Ak x k − Ak xk+1 = Am xm + Ak (xk − xk+1 )
k=n+1 k=n k=n+1
m−1
X
≤ |Am |xm + |Ak |(xk − xk+1 )
k=n+1
m−1
ε m ε X ε n+1
≤ x + (xk − xk+1 ) = x <ε
2 2 2
k=n+1

∞
X
By Cauchy criterion for uniform convergence, ak xk converges uniformly to S(x) on [0, 1]. Moreover, since
k=0
the partial sum is continuous on [0, 1], S(x) is continuous on [0, 1] and we have
∞
X
lim S(x) = S(0) = a0 and lim S(x) = S(1) = ak
x→0+ x→1−
k=0

For the general case, we have 

[c, v]
 if c ≤ u
[u, v] ⊆ [u, c] ∪ [c, v] if u < c < v

[u, c] if v ≤ c


Then by subset and union properties in Theorem 5.1, it suffices to consider the cases [c, v] and [u, c].

x−c
For [c, v], we consider the substitution t = h(x) = and bk = ak (v − c)k . Then we have
v−c
∞
X ∞
X
ak (x − c)k = bk tk
k=0 k=0

and x ∈ [c, v] ⇔ t ∈ [0, 1]. By substitution property in Theorem 5.1 and the special case we have already
∞
X
proved, ak (x − c)k converges uniformly on [c, v].
k=0

c−x
For [u, c], we consider the substitution t = h(x) = and bk = ak (c − u)k . Then we have
c−u
∞
X ∞
X
ak (x − c)k = bk tk
k=0 k=0

and x ∈ [u, c] ⇔ t ∈ [0, 1]. By substitution property in Theorem 5.1 and the special case we have already
∞
X
proved, ak (x − c)k converges uniformly on [c, v].
k=0

26
Remarks 5.1.
1. Abel’s limit theorem is false for other types of intervals. For example,
∞
X 1
xk = on [0, 1)
1−x
k=0

1
However, the series does not converge uniformly to on [0, 1).
1−x
2. In previous section, for any x ∈ (−1, 1),
∞
X (−1)n x2n+1
arctan(x) =
n=0
2n + 1

∞
X (−1)n
Moreover, when x = 1, converges by alternating series test. Hence the power series con-
n=0
2n + 1
verges pointwise on [0, 1] and by Abel’s limit theorem,
∞
π X (−1)n
= lim− arctan(x) =
4 x→1
n=0
2n + 1

∞
X (−1)k−1
Exercise 5.1. Prove that = ln 2.
k
k=1
1
(Hint: Consider the Taylor series of and integrate.)
1+x

27