Chapter 2

Sequences and Series of Functions

2.1 Sequences of functions

2.1.1 Pointwise Convergence and Uniform Convergence

Let A ⊂ R, and for every n ∈ N: fn : A → R, f : A → R.

Definition 2.1.1. We say that the sequence of functions (fn )n converges pointwise to
f on A if and only if for every fixed x ∈ A, the numerical sequence (fn (x))n converges
to f (x), and we write
∀x ∈ A : lim fn (x) = f (x)
n→+∞
or
fn −→ f, n → +∞.
A
In other words

∀x ∈ A, ∀ε > 0, ∃N = N (ε, x) ∈ N : ∀n ≥ N ⇒ |fn (x) − f (x)| < ε

Example 2.1.2. Let fn : R → R be defined by

fn (x) = xn , n ∈ N.

We have

• For |x| < 1, lim fn (x) = lim xn = 0.

n→+∞ n→+∞
• For |x| > 1, lim fn (x) = lim xn = ∞.
n→+∞ n→+∞
• For x = −1, we have fn (−1) = (−1)n doesn’t have a limit.
• For x = 1, we have fn (1) = 1 → 1, n → +∞.

So, the sequence (fn )n converges pointwise to the function f defined by


0, |x| < 1
f (x) =
1, x = 1.

43
Example 2.1.3. Let fn : R → R be defined by

nx3
fn (x) = , n ∈ N.
1 + nx3
• For x = 0, fn (0) = 0, hence limn→+∞ fn (0) = 0.
nx3
• Pour x ̸= 0, lim fn (x) = lim 1+nx 3 = 1.
n→+∞ n→+∞

So, the sequence (fn )n converges pointwise to the function f defined by

(
1, x ∈ R∗ ,
f (x) =
0, x = 0.

In the above definition, the integer N generally depends on both ε and x. This will some-
times lead us to write N (ε, x) instead of just N . By requiring that N be independent
of x, we achieve a more restrictive mode of convergence called uniform convergence.

Definition 2.1.4. We say that the sequence of functions (fn )n converges uniformly on
A to the function f if and only if

∀ε > 0, ∃N = N (ε) ∈ N, ∀n ≥ N, ∀x ∈ A : |fn (x) − f (x)| < ε.

Remark. If A is an interval in R, the uniform convergence of the sequence (fn ) to f

means that for every ε > 0, there exists a rank N such that the graph of fn is contained
within the region of the xOy-plane defined by x ∈ A and y ∈ [f (x) − ε, f (x) + ε]. In
other words, there is a rank from which the graph of fn lies between the graphs of f − ε
and f + ε.

We provide an equivalent and more practical definition of uniform convergence here.

Definition 2.1.5. We say that the sequence of functions (fn ) converges uniformly on
A to the function f if and only if

dn = sup |fn (x) − f (x)| −→ 0, n → +∞.

x∈A

Note that the sequence dn ≤ +∞, n ∈ N.

Theorem 2.1.6. If (fn ) converges uniformly to f on A, then (fn )n converges pointwise

to f on A.

Proof. Just observe that, for all x ∈ A,

|fn (x) − f (x)| ≤ sup |fn (x) − f (x)| → 0

x∈A

as n → +∞.

The inverse of Theorem 2.1.6 is not true, as shown by the following example.

44
Example 2.1.7. Let
fn (x) = xn , A = [0, 1], n ∈ N.

For a fixed x ∈ [0, 1), lim fn (x) = 0, and for x = 1, we have lim fn (1) = 1. The
n→+∞ n→+∞
sequence (fn )n converges pointwise to the function f defined by
(
0, x ∈ [0, 1[
f (x) =
1, x = 1

Let’s study uniform convergence. We have

dn = sup |fn (x) − f (x)| = sup |xn − 0| = 1 ↛ 0, n → +∞

0≤x≤1 0≤x<1

Therefore, the convergence is not uniform on A.

Let’s study uniform convergence on

Ba = [0, a], 0 < a < 1.

The sequence {xn , x ∈ [0, a], n ∈ N} converges uniformly on Ba because...

dn = sup |fn (x) − f (x)| = sup |xn − 0| = an → 0, n → +∞.

0≤x≤a 0≤x≤a

Remark. In the previous example, we observe that the pointwise limit of continuous
functions on A = [0, 1] is not a continuous function.

The following is a direct consequence of Theorem 2.1.6.

Corollary 2.1.8. For (fn )n not being convergent uniformly to f on A, it is sufficient

that there exists a sequence (xn )n of points in A such that (fn (xn ) − f (xn ))n does not
tend to zero.

Example 2.1.9. Let’s consider the sequence of functions defined on R by

sin nx
fn (x) = .
1 + n2 x2
For any fixed x in R, the numerical sequence (fn (x)) converges to 0. Therefore, the
sequence (fn ) converges pointwise to the zero function on R. Now, for n ≥ 1, we have
π 1
fn = ↛ 0 lorsque n → +∞.
2n 1 + π 2 /4

Therefore, (fn )n does not converge uniformly to 0.

Definition 2.1.5 of uniform convergence assumes that the limit function f of the sequence
(fn )n is known. When f is unknown, one can utilize the highly significant criterion as
follows.

45
Theorem 2.1.10. Let fn : A → R n ∈ N be a sequence of functions. For the sequence
(fn )n to converge uniformly on A, it is necessary and sufficient that:

∀ε > 0, ∃N ∈ N, ∀m ≥ N, ∀n ≥ N, ∀x ∈ A : |fm (x) − fn (x)| < ε.

Proof. Suppose that (fn )n converges uniformly to a function f on A, and let ε > 0.
Then, there exists N ∈ N such that
ε
∀n ≥ N, ∀x ∈ A : |fn (x) − f (x)| < .
2
So, ∀m ≥ N, ∀n ≥ N, ∀x ∈ A

|fm (x) − fn (x)| = |(fm (x) − f (x)) − (fn (x) − f (x))|

ε ε
≤ |fm (x) − f (x)| + |fn (x) − f (x)| < + = ε.
2 2
If the stated condition is met, the sequence (fn (x))n is Cauchy for all x ∈ A. Therefore

∃f (x) ∈ R : lim fn (x) = f (x), x ∈ A.

n→+∞

Let ε > 0, and let N ∈ N be chosen according to the condition of the theorem. For all
x∈A
∀m ≥ N, ∀n ≥ N : |fm (x) − fn (x)| < ε.
So, by letting m approach infinity, we obtain

∀n ≥ N, ∀x ∈ A : |fn (x) − f (x)| < ε,

which shows that the sequence (fn )n converges uniformly to f on A.

Corollary 2.1.11. For the sequence (fn )n not to converge uniformly on A, it is nec-
essary and sufficient that

∃ε > 0, ∀N ∈ N : ∃n ≥ N, ∃p ∈ N∗ , ∃x ∈ A et |fn+p (x) − fn (x)| ≥ ε.

Example 2.1.12. Study the uniform convergence of the sequence

x
fn (x) = sin , n ∈ N
n
on A = R. Since for all x ∈ R, we have lim nx = 0, therefore
x→+∞
x x
lim fn (x) = lim sin = lim = 0.
n→+∞ n→+∞ n n→+∞ n
So, the sequence (fn )n converges pointwise to the function f (x) = 0, ∀x ∈ R. However,
the convergence is not uniform over R. Indeed, in the difference

fn+p (xn ) − fn (xn )

we consider xn = nπ
2 ∈ R, p = 2n (n ∈ N∗ ) and let ε = 14 . Then
xn xn π π 1 1
|f3n (xn ) − fn (xn )| = sin − sin = sin − sin = > = ε,
3n n 6 2 2 4
for all n ≥ N , which shows that the sequence (fn )n does not satisfy the Cauchy criterion
for uniform convergence on A = R.

46
2.1.2 Uniform Convergence and Continuity

Theorem 2.1.13. Let the sequence (fn )n uniformly converges to f on A. If the func-
tions fn are continuous at x0 ∈ A, then f is continuous at x0 .

Proof. Let ε > 0, there exists an N ∈ N such that

ε
∀n ≥ N, ∀x ∈ A : |fn (x) − f (x)| < .
3
Let’s fix an integer n ≥ N . Since fn is continuous at x0 , there exists δ > 0 such that
ε
∀x ∈ A : |x − x0 | < δ ⇒ |fn (x) − fn (x0 )| < .
3
From the above two inequalities, we have for all x ∈ A that satisfies |x − x0 | < δ

|f (x) − f (x0 )| ≤ |f (x) − fn (x)| + |fn (x) − fn (x0 )| + |fn (x0 ) − f (x0 )|
ε ε ε
< + + = ε,
3 3 3
which proves the theorem.

Example 2.1.14. The sequence of functions defined by

fn (x) = enx , n ∈ N, x≤0

converges pointwise on R− to the function f defined by

(
0, x < 0,
f (x) =
1, x = 0.

The functions fn are continuous, and their limit f is not continuous at x = 0. Therefore,
the convergence of (fn )n is not uniform on R− .

We should point out here the existence of sequences (fn ) of continuous functions that
converge pointwise but not uniformly to a limit function f that is continuous.

Example 2.1.15. Let

x
fn (x) = sin , n ∈ N, x ∈ R.
n
The functions fn are continuous, the sequence (fn )n converges pointwise to the contin-
uous function f (x) = 0, x ∈ R. However, the convergence is not uniform because.

2.1.3 Uniform Convergence and Integration

Theorem 2.1.16. Let (fn )n be a sequence of integrable functions on A = [a, b] that

converges uniformly to f . Then

(1) f is Riemann integrable on A (i.e., f ∈ R(A)).

47
(2) We have
Z b Z b Z b
lim fn (x)dx = lim fn (x)dx = f (x)dx
n→+∞ a a n→+∞ a

Proof. Let ε > 0 be given. There exists n ∈ N such that

ε ε
∀x ∈ [a, b] : fn (x) − < f (x) < fn (x) +
4(b − a) 4(b − a)
as (fn )n converges uniformly to f on A.
Since the function fn is integrable, there exists a partition d = {x0 , x1 , . . . , xk } of [a, b]
such that
k−1
X ε
(Mj (fn ) − mj (fn )) ∆xj < ,
2
j=0

where
Mj (fn ) = sup fn ; mj (fn ) = inf fn , ∆xj = xj+1 − xj .
[xj ,xj+1 ] [xj ,xj+1 ]

From the first inequality we have for all 0 ≤ j ≤ k − 1 :

ε ε
mj (fn ) − ≤ mj (f ) ≤ Mj (f ) ≤ Mj (fn ) + .
4(b − a) 4(b − a)
This, together with the second inequality result in
k−1 k−1  
X X ε ε
(Mj (f ) − mj (f )) ∆xj ≤ Mj (fn ) + − mj (fn ) + ∆xj
4(b − a) 4(b − a)
j=0 j=0
k−1 k−1
X ε X
= (Mj (fn ) − mj (fn )) ∆xj + ∆xj
2(b − a)
j=0 j=0

ε ε
< + (b − a) = ε.
2 2(b − a)
So, f ∈ R(A).
To prove the second assertion, we have
Z b Z b Z b
fn (x)dx − f (x)dx ≤ |fn (x) − f (x)| dx
a a a
≤ (b − a) sup |fn (x) − f (x)| −→ 0, n → +∞
x∈[a,b]

and we are done!

This result allows, in some examples, to demonstrate the lack of uniform convergence.
Therefore, if a sequence (fn )n of integrable functions on A = [a, b] converges to an
integrable function on A = [a, b] and
Z b Z b
lim fn (x)dx ̸= f (x)dx.
n→+∞ a a

Then, the convergence is not uniform.

48
Example 2.1.17. Let

fn (x) = n2 x(1 − x)n , n ∈ N, 0 ≤ x ≤ 1.

The sequence (fn )n converges pointwise to the function f defined by f (x) = 0, x ∈ [0, 1].
So Z 1
f (x)dx = 0.
0
On the other hand, we have
Z 1 Z 1 Z 0
2 n 2
fn (x)dx = n x(1 − x) dx = n (1 − t)tn (−dt), (1 − x = t)
0 0 1
Z 1
n2
 
2 n n+1

2 1 1
=n t −t dt = n − = −→ 1, n → +∞,
0 n+1 n+2 (n + 1)(n + 2)

which means that (fn )n is not coverging uniformly.

Corollary 2.1.18. Let (fn ) be a sequence of integrable functions on the compact in-
terval [a, b], converging uniformly to a function f on [a, b]. Denote
Z x Z x
Fn (x) = fn (t)dt, F (x) = f (t)dt.
a a

Then (Fn )n converges uniformly to F .

Proof. For x ∈ [a, b], we have

Z x Z x
|Fn (x) − F (x)| = fn (t)dt − f (t)dt
a a
Z x
≤ |fn (t) − f (t)| dt
a
Z b
≤ |fn (t) − f (t)| dt
a
≤ (b − a) sup |fn (t) − f (t)| → 0, n → +∞.
t∈[a,b]

The following theorem improves the previous result.

Theorem 2.1.19. Let (fn ) be a sequence of integrable functions on the compact interval
[a, b], converging pointwise to a function f on [a, b]. If the functions fn are all bounded
by M , and if the convergence of (fn ) to f is uniform on every compact interval [α, β] ⊂
(a, b), then f is integrable on [a, b], and we have
Z b Z b
lim fn (x)dx = f (x)dx.
n→+∞ a a

49
Proof. The limit function f is integrable on [a, b]. Indeed, it is bounded because as
we take the limit as n approaches infinity, we have |f (x)| ≤ M for all x ∈ [a, b].
Furthermore, f is integrable on any compact interval contained within (a, b) as the
uniform limit of integrable functions.
Given the number ε > 0, let’s choose
ε ε
α<a+ , β >b− such that a < α < β < b.
6M 6M
For all n ∈ N, we have
Z α
ε
(fn (x) − f (x)) dx ≤ 2M (α − a) ≤
a 3

and Z b
ε
(fn (x) − f (x)) dx ≤ 2M (b − β) ≤ .
β 3
The numbers α and β fixed, the uniform convergence of (fn ) to f on [α, β] implies the
existence of an integer N such that for all n > N , we have
Z β
ε
(fn (x) − f (x)) dx ≤ ,
α 3

So, for any integer n > N :

Z b
(fn (x) − f (x)) dx ≤ ε
a

So, we have demonstrated that

Z b
lim (fn (x) − f (x)) dx = 0.
n→+∞ a

Example 2.1.20. Let (fn ) be the sequence defined on [0, π/2] by fn (x) = sinn x.
This sequence converges uniformly to the zero function on any interval [0, a], for any
0 < a < π/2, because on such an interval we have |sinn x| ≤ sinn a with 0 < a < 1.
Furthermore, the functions fn are bounded by 1. Therefore, thanks to the previous
theorem, we have
Z π/2
lim sinn xdx = 0.
n→+∞ 0

2.1.4 Uniform Convergence and Differentiation

Consider the sequence (fn )n≥1 of functions defined on R by

1 1/2
 
2
fn (x) = x + 2
n

50
This sequence converges pointwise to the function f : x 7→ |x|. For each n ≥ 1, fn is
differentiable (and even of class C ∞ ) on R. Furthermore, we have:
r
1 1 1
|x| ≤ x2 + 2 ≤ |x| + , then |fn (x) − f (x)| ≤ ,
n n n
which proves that the convergence of (fn ) to f is uniform on R. However, the function
f is not differentiable at the origin!
In other words, the limit of a sequence (even uniformly convergent) of differentiable
functions is not necessarily differentiable. In this subsection, we will establish a sufficient
(but not necessary) condition for the limit of a sequence (fn ) of differentiable functions
to be differentiable.

Theorem 2.1.21. Let (fn )n be a sequence of functions of class C 1 on A = [a, b] satis-

fying the following conditions:

(i) The sequence (fn′ )n converges uniformly to g on [a, b].

(ii) There exists a point x0 ∈ [a, b] such that the sequence (fn (x0 ))n∈N is convergent.

Then the sequence (fn )n converges uniformly to a function f of class C 1 , and we have
f ′ = g. In other words,  ′
′
lim fn (x) = lim f (x) .
n→+∞ n→+∞

Proof. According to the Newton-Leibniz formula

Z x
fn (x) = fn (x0 ) + fn′ (t)dt
x0

By virtue of the previous Corollary and assumption (i)

Z x Z x  Z x
′ ′
lim fn (t)dt = lim fn (t) dt = g(t)dt
n→+∞ x x0 n→+∞ x0
0

uniformly on [a, b]. Therefore, the sequence (fn )n converges uniformly to f defined by
Z x
f (x) = lim fn (x0 ) + g(t)dt.
n→+∞ x0

Since g is continuous, the function f is continuously differentiable, and we have f ′ =

g.

Example 2.1.22. Consider the sequence of functions

ln 1 + nx2


fn (x) = , x ∈ [1, 2].
2n
Here fn (1) converges to 0 , and the sequence of derivatives
2nx x
fn′ (x) = 2
= , x ∈ [1, 2]
2n (1 + nx ) 1 + nx2

51
converges pointwise to g(x) = 0, x ∈ [0, 2].
The convergence of fn′ to g is uniform on [1, 2] because |x| ≤ 2 and 1 + nx2 ≥ n implies

x |x| 2
2
−0 = 2
≤
1 + nx 1 + nx n

Then by the above Theorem, the sequence (fn ) converges uniformly (something not
easily proved directly) and the limit function f = lim fn is differentiable with f ′ = g.
Here f (x) = 0 for x ∈ [1, 2] whose derivative is 0 .

Example 2.1.23. Let a > 0 be a fixed number, and denote by (fn )n the sequence of
functions defined by
fn (x) = e−nx , n ∈ N, x ≥ a

For x ≥ a, we have
fn′ (x) = −ne−nx ≤ ne−na .

Since the sequence (ne−na )n tends to zero, this shows the uniform convergence of (fn′ )n
to zero. We have
lim fn′ (x) = lim f ′ (x) = 0
n→+∞ n→+∞

as lim fn (x) = f (x) = 0.

n→+∞

2.2 Series of functions

As with numerical series, from a sequence of functions (fn )n , one can construct a
sequence (Sn )n where
Sn = f0 + . . . + fn .
P
This gives what is called a series of functions, denoted as fn .

Definition 2.2.1. Given (fn )n a sequence of functions defined on a set A and taking
real values. We call the series of functions the expression:
+∞
X
f0 (x) + f1 (x) + f2 (x) + · · · + fn (x) + · · · = fn (x), x ∈ A.
n=0

Define
n
X
Sn (x) = fk (x).
k=0
P
The function Sn is referred to as the n-th partial sum of the series fn .

52
2.2.1 Pointwise convergence
P
Definition 2.2.2. We say that the series of functions n≥0 fn converges pointwise on
A if and only if the sequence (Sn )n of partial sums converges pointwise on A. In this
P
case, the limit function S of (Sn )n is called the sum function of the series fn , and it
is denoted as
+∞
X
S= fn .
n=0

Example 2.2.3. Study the pointwise convergence of the series of functions

+∞
X x2
, x ∈ R.
(1 + x2 )n
n=0

We have two cases:

• If x ̸= 0, we have
 
1
n
X x2 1− (1+x2 )n+1
Sn (x) = = x2 1
(1 +
k=0 x2 )k 1 − 1+x 2
 
2

1
= 1+x 1− .
(1 + x2 )n+1
1 1
Since 1+x2
< 1, then lim n+1 = 0, and so
n→+∞ (1+x2 )

lim Sn (x) = 1 + x2 .
n→+∞

• If x = 0, then Sn (0) = 0 and lim Sn (0) = 0.

n→+∞

So the proposed series converges pointwise on R, and its sum

+∞
(
X x2 1 + x2 , x ∈ R∗ ,
S(x) = =
(1 + x2 )n 0, x = 0.
n=0
P+∞
Definition 2.2.4. Given : A → R a series of functions, we call the set
n=0 un , un

+∞
( )
X
D= x∈A: un (x) converges
n=0
P+∞
The domain of pointwise convergence of the series n=0 un is called D. The set D can
be empty or coincide with A.
P+∞
Example 2.2.5. Study the pointwise convergence of the series of functions n=0 un ,
where
un : R → R
2 x2
x 7→ x2 e−n , n ∈ N.
Let x be fixed.

53
 • If x ∈ R∗ , then
p
n
1 2
lim x2 e−n2 x2 = lim x2 n
e−nx = 0 < 1,
n→+∞ n→+∞
P+∞
which shows that the series n=0 un is absolutely
 convergent (we can also remark
2 2 2 −n2 x2
that n un (x) = n x e −→ 0, n → +∞ .
P+∞
• If x = 0, the series n=0 un (0) = 0 converges.

Then +∞
P
n=0 un converges pointwise on D = R.

Example 2.2.6. Determine the domain of pointwise convergence D of the series of

functions +∞
P
n=1 un , where
√
nx
ne
un (x) = (−1) , n ∈ N∗ , x ∈ R.
n
• If x > 0 fixed, √
e nx
|un (x)| = −→ +∞, n → +∞
n
and the series diverges.
(−1)n P+∞ (−1)n
• If x = 0, un (0) = n and the series n=1 n is conditionally convergent.
√
• If x > 0 is fixed, then n2 |un (x)| = ne nx −→ 0, n → +∞, which shows the
convergence of the series.

So the domain of convergence is D = R− =] − ∞, 0].

P
Definition 2.2.7. The n-th order remainder of a simply convergent series fn is
called the function Rn , defined by
+∞
X
∀x ∈ A, Rn (x) = fk (x)
k=n+1

Remark. For all x ∈ A, we have then S(x) = Sn (x) + Rn (x).

2.2.2 Absolute convergence

Definition 2.2.8. We say that +∞

P
n=0 un converges absolutely on A if and only if the
P+∞
series n=0 |un | is convergent on A.
Example 2.2.9. The series
+∞
X (−1)n x2 cos nx
,x ∈ R
(1 + x2 )n
n=0
is absolutely convergent on R as
(−1)n x2 cos nx x2
∀x ∈ R : n ≤
(1 + x2 ) (1 + x2 )n
P+∞ x2
and the series x ∈ R is convergent.
n=0 (1+x2 )n ,
P
Theorem 2.2.10. If the series fn converges absolutely on A, then it converges
pointwise on A.

54
2.2.3 Uniform convergence
P
Definition 2.2.11. We say that the series fn converges uniformly on A if the se-
quence of functions (Sn )n converges uniformly on A.

Example 2.2.12. Consider the series of functions

X  xn−1 xn

− , x ∈ [−1, 1].
n n+1
n≥1

Let’s calculate the sequence of partial sums Sn (x):

n  k−1
xk x x2
  
X x x
Sn (x) = − = 1− + −
k k+1 2 2 3
k=1
 n−1
xn xn

x
+ ··· + − =1− .
n n+1 n+1

xn 1
Since ∀x ∈ [−1, 1] : n+1 ≤ n+1 → 0, n → +∞, then

xn
 
lim Sn (x) = S(x) = lim 1− =1
n→+∞ n→+∞ n+1

Let’s study the uniform convergence of the sequence (Sn )n on [−1, 1]. Since

xn
 
sup |Sn (x) − S(x)| = sup 1− −1
x∈[−1,1] x∈[−1,1] n+1
|x|n 1
= sup ≤ −→ 0, n → +∞,
x∈[−1,1] n + 1 n+1

then the series

+∞  n−1
xn

X x
−
n n+1
n=1

converges uniformly on [−1, 1].

Definition 2.2.13. We say that the series

+∞
X
fn (x) = S(x), x ∈ A
n=0

is uniformly convergent on A if and only if

dn = sup |Sn (x) − S(x)| = sup |Rn (x)| −→ 0, n → +∞,

x∈A x∈A

This shows that the sequence (Rn )n of remainders converges uniformly to zero on A.

Example 2.2.14. The series

+∞
X 1
xn = , x ∈] − 1, 1[
1−x
n=0

55
is uniformly convergent on Ba = [−a, a] (where 0 < a < 1) because
+∞
X xn+1
dn = sup |Rn (x)| = sup xk = sup
x∈[−a,a] x∈[−a,a] k=n+1 x∈[−a,a] 1−x
|x|n+1 an+1
= sup ≤ −→ 0, n → +∞.
x∈[−a,a] 1 − x 1−a

However, on the interval ] − 1, 1[, the convergence is not uniform, as

|x|n+1
dn = sup |Rn (x)| = sup = +∞.
x∈]−1,1[ x∈]−1,1[ 1 − x
P
Theorem 2.2.15. If the series fn converges uniformly on A, then the sequence of
functions (fn ) uniformly converges to zero (where zero represents the zero function) in
A.

2.2.4 Uniform Convergence Cauchy Criterion for a Series of Functions

P
Theorem 2.2.16. For the series of functions fn to converge uniformly on A, it is
n≥0
necessary and sufficient that

∀ε > 0, ∃N ∈ N : ∀n ≥ N, ∀p ∈ N, ∀x ∈ A : |fn+1 (x) + · · · + fn+p (x)| < ε.

Proof. This is nothing but the uniform Cauchy criterion applied to the sequence of
partial sums (Sn ).

Example 2.2.17. The series

+∞
X sin(nx)
, (0 < α ≤ 1)
nα
n=1

is not uniformly convergent on [−π, π]. Indeed, for the sum

2n
X sin(kx)
kα
k=n+1

k
we consider x = 1/2n. Then, for all 0 < 2n ≤ 1 < π2 , we have
 
k 2 k
sin ≥ .
2n π 2n
Hence,
2n 2n   2n
X sin(kx) X 1 k X 1 2 k
α
= α
sin ≥
k k 2n k α π 2n
k=n+1 k=n+1 k=n+1
2n 2n
1 X 1 1 X 1 1
= = k 1−α ≥ · n(1 + n)1−α ≥ .
nπ k α−1 nπ nπ π
k=n+1 k=n+1
P+∞ sin(nx)
This shows that the series < α ≤ 1) does not satisfy the Cauchy
n=1 nα , (0
criterion for uniform convergence on [−π, π].

56
2.2.5 Uniform and Absolute Convergence
P∞
Theorem 2.2.18. If a series n=1 |fn | converges uniformly on A, then the series
P∞
n=1 fn converges uniformly and absolutely on A.

P∞
Proof. If a series n=1 |fn | converges uniformly on A, the absolute convergence of
P∞
n=1 fn at every fixed point of A follows directly from the properties of series of
numbers. To show the uniform convergence we use the Cauchy criterion:
m
X m
X
fk (x) ≤ |fk (x)| < ε
k=n+1 k=n+1

for all x ∈ X at once and for ∀m > n > N , where N is from the Cauchy criterion for
the series ∞
P
n=1 |fn |.

2.2.6 Normal convergence

Definition 2.2.19. Let +∞

P
n=0 fn be a series of functions, where fn : A → R. We say
P
that this series is normally convergent on A if and only if the numerical series ∥fn ∥∞ ,
where
∥fn ∥∞ = sup |fn (x)| ,
x∈A

is convergent.

Example 2.2.20. The series

+∞
X 1
3 ,x ∈ R
n=1 n + x2
2

is normally convergent on R, since

1 1
sup |fn (x)| = sup 3 = 3
x∈R x∈R n + x2
2 n2
P+∞ 1
and the series n=1 3 converges.
n2
P
Theorem 2.2.21. The series fn is normally convergent on A if and only if there
exists a positive real sequence (αn )n satisfying the following two conditions:

• ∀n ∈ N, ∀x ∈ A, |fn (x)| ≤ αn ;
• The series
P
αn converges.
P
Proof. If the series fn is normally convergent, then the sequence with general term
αn = ∥fn ∥∞ satisfies the two conditions of the previous proposition.
Conversely, if the two conditions of the theorem are satisfied, then each fn is bounded,
P
and the series ∥fn ∥∞ converges as a series of positively bounded terms, being domi-
nated by the sum of a convergent series.

57
Example 2.2.22. Consider the series
+∞
X sin nx
, x ∈ R.
n!
n=0

We have
sin nx 1
∀n ∈ N, ∀x ∈ R : ≤ ,
n! n!
and since +∞
P 1 P+∞ sin nx
n=0 n! converges (according to the ratio test), then n=0 n! is normally
convergent over R.
P
Theorem 2.2.23. If the series fn converges normally on A, then it converges ab-
solutely and uniformly on A, and we have
+∞
X +∞
X
fn ≤ ∥fn ∥∞ .
n=0 ∞ n=0
P
Proof. If the series fn converges normally, there exists a positive real sequence (αn )
such that
∀n ∈ N, ∀x ∈ A,
|fn (x)| ≤ αn .
P P
This, together with the convergence of the series αn . The series fn (x) therefore
P
converges absolutely on A. Hence, the absolute convergence of fn on A.
Now, we have
+∞
X +∞
X +∞
X
dn = sup fk (x) ≤ sup |fk (x)| ≤ αk → 0, n → +∞.
x∈A k=n+1 x∈A k=n+1 k=n+1
P+∞
This shows that the series n=0 un is uniformly convergent on A.

2.2.7 Dirihlet’s and Abel’s tests for series of functions

P∞
here we prove the two tests for uniform convergence of series n=1 fn (x) with the
general term in the form fn (x) = an (x)bn (x). These results are similar to Dirichlet’s
and Abel’s tests for series of numbers and we adapt the line of reasoning used in the
proofs of those tests to the context of uniform convergence. Before we proceed to our
results, let’s recall the following Abel’s Summation by Parts Formula

Lemma 2.2.24. Let ∞

P
n=1 fn (x) be a series of functions with the general term in the
form fn (x) = an (x)bn (x) and let Bn (x) = nk=1 bk (x), all the functions defined on a
P

set A ⊂ R. Then for any m > n ≥ 1 and for each x ∈ X the following formula is true
m
X m−1
X
ak bk = Bk (ak − ak+1 ) + Bm am − Bn an+1
k=n+1 k=n+1
P∞
Theorem 2.2.25 (Dirihlet-Abel’s test 1). Let n=1 fn be a series of functions defined
on a set A ⊂ R with fn (x) = an (x)bn (x). If

58
(1) ∃M such that
n
X
|Bn (x)| = bk (x) ≤ M,
k=1
for all n and for all x ∈ A,
(2) The sequence (an )n is monotone at every fixed x ∈ A,
(3) The sequence (an )n converges uniformly to 0 on A.
P∞
Then the series n=1 an bn is uniformly convergent on A.

Proof. Let us show that the series ∞

P
n=1 fn satisfies the conditions of the Cauchy cri-
terion for uniform convergence. To do this, take m > n and use Abel’s summation
formula to evaluate the sums S ≡ m
P Pm
k=n+1 fk = k=n+1 ak · bk , in the form

m
X m−1
X
|S| = ak bk = Bk (ak − ak+1 ) + Bm am − Bn an+1
k=n+1 k=n+1
m−1
X
≤ |Bk | · |ak − ak+1 | + |Bm | · |am | + |Bn | · |an+1 | ≡ S1 .
k=n+1

By the condition (1) of the Theorem, the partial sums Bn are uniformly bounded, that
is, there exists M > 0 such that |Bn | ≤ M for ∀n ∈ N and for all x ∈ A. Then, we can
proceed with the evaluation of S1 in the following way:
m−1
!
X
S1 ≤ M |ak − ak+1 | + |am | + |an+1 | ≡ S2 .
k=n+1

Since the sequence an is monotone at each x ∈ A, all the differences ak − ak+1 have the
same sign, and consequently, the sum of the absolute values is equal to the absolute
value of the sum:
m−1
!
X
S2 = M (ak − ak+1 ) + |am | + |an+1 |
k=n+1

= M (|an+1 − an+2 + an+2 − an+3 + . . . + am−1 − am | + |am | + |an+1 |)

= M (|an+1 − am | + |am | + |an+1 |) ≤ 2M (|an+1 | + |am |) .

A
Now we use the condition of the uniform convergence of an : an ⇒ 0. By definition,
n→∞
ε
this means that for ∀ε > 0 (take for convenience ε1 = 4M ) there exists Nε such that
ε
for ∀n > Nε and simultaneously for all x ∈ A it follows that |an | < 4M . Choosing for
the evaluation of the sum S the indices ∀m > n > Nε , we have
m  ε
X ε 
|S| = ak bk ≤ 2M (|an+1 | + |am |) < 2M + = ε.
4M 4M
k=n+1

Hence, according to the Cauchy criterion for uniform convergence, the series
P∞
n=1 an (x)bn (x) converges uniformly on A.

59
 P∞
Theorem 2.2.26 (Dirihlet-Abel’s test 2). Let n=1 fn be a series of functions defined
on a set A ⊂ R with fn (x) = an (x)bn (x). If
P∞
(1) the series n=1 bn converges uniformly on A,
(2) The sequence (an )n is monotone for every fixed x ∈ A,
(3) ∃M such that |an (x)| ≤ M, for all n and for all x ∈ A,
P∞
Then the series n=1 an bn is uniformly convergent on A.

Proof. Let us show that the series ∞

P
n=1 an bn satisfies the conditions of the Cauchy
criterion for uniform convergence. First we use Abel’s summation by parts formula and
the telescopic representation an+1 = m−1
P
k=n+1 (ak − ak+1 ) + am in order to express the
Pm Pm
sum S ≡ k=n+1 uk = k=n+1 ak bk , m > n in the following manner
m
X m−1
X
S= ak bk = Bk (ak − ak+1 ) + Bm am − Bn an+1
k=n+1 k=n+1
m−1 m−1
!
X X
= Bk (ak − ak+1 ) + Bm am − Bn (ak − ak+1 ) + am
k=n+1 k=n+1
m−1
X
= (Bk − Bn ) (ak − ak+1 ) + (Bm − Bn ) am .
k=n+1

From this formula we get the following evaluation:

m−1
X
|S| ≤ |Bk − Bn | · |ak − ak+1 | + |Bm − Bn | · |am | ≡ S1 .
k=n+1

Since the series ∞

P
n=1 bn converges uniformly on A, it satisfies the Cauchy criterion for
uniform convergence, that is, ∀ε1 > 0 there exists Nε1 such that for ∀m > n > Nε1 and
simultaneously for all x ∈ A we have
m
X
bk = |Bm − Bn | < ε1 .
k=n+1

Applying this property in the evaluation of S1 , we obtain

m−1
!
X
S1 < ε1 |ak − ak+1 | + |am | ≡ S2 .
k=n+1

Since the sequence an is monotone at each x ∈ A, all the differences ak − ak+1 have the
same sign, and consequently, the sum of the absolute values is equal to the absolute
value of the sum:
m−1
!
X
S2 = ε1 (ak − ak+1 ) + |am |
k=n+1

= ε1 (|an+1 − an+2 + an+2 − an+3 + . . . + am−1 − am | + |am |)

= ε1 (|an+1 − am | + |am |) ≤ ε1 (|an+1 | + 2 |am |) .

60
Finally, we employ the condition of the uniform boundedness of an : |an | ≤ M , for
ε
∀n ∈ N and for all x ∈ A, and specify ε1 = 3M to arrive at the inequality
m
X ε ε
|S| = ak bk < (|an+1 | + 2 |am |) ≤ · 3M = ε.
3M 3M
k=n+1

This shows that the series ∞

P
n=1 an bn satisfies the conditions of the Cauchy criterion
for uniform convergence, and consequently, this series converges uniformly on A.

Corollary 2.2.27 (Generalized Leibniz’s Test). If a sequence of function (an )n con-

verges uniformly to 0 on A, and for any fixed x ∈ A the sequence (an )n is non-negative
and decreasing (that is, 0 ≤ an+1 (x) ≤ an (x) for ∀n ∈ N and ∀x ∈ A), then the series
P∞ n
n=1 (−1) an (x) converges uniformly on A.

Let us solve some examples with application of Dirichlet’s and Abel’s tests.

Example 2.2.28. Consider the series of functions

∞
X xn
(−1)n , A = [0, 1]
n
n=1

First, let us apply Dirichlet’s test. For this purpose, choose an = n1 and bn (x) =
(−1)n xn . Then, the series ∞
P P∞ n n
n=1 bn (x) = n=1 (−1) x has the partial sums bounded
uniformly on [0, 1] :
n
X 1 − (−x)n x
|Bn (x)| = (−1)k xk = −x ≤2 < 2, ∀n, ∀x ∈ [0, 1]
1+x 1+x
k=1

The sequence an = n1 is monotone and uniformly convergent to 0 (since it does not

depend on x ). Thus, all the conditions of Dirichlet’s test are satisfied, and consequently,
the original series converges uniformly on [0, 1].
n
To use Abel’s test, we choose an = xn and bn (x) = (−1) n . Then, the alternating
P∞ P∞ (−1)n
harmonic series n=1 bn (x) = n=1 n converges uniformly on [0, 1] (since bn (x)
does not depend on x ). The sequence an = xn is decreasing for every fixed x ∈ [0, 1]
and uniformly bounded by 1 on [0, 1]. Thus, all the conditions of Abel’s test hold, and
consequently, the original series converges uniformly on [0, 1].

Example 2.2.29. The series

+∞
X (−1)n
√
n=1
2 n + sin 3x

converges uniformly over R according to the Leibniz rule. Indeed, if we define un (x) =
√ 1 for n ∈ N and x ∈ R, then we have:
2 n+sin 3x

• ∀n ∈ N, ∀x ∈ R : un (x) = √ 1
2 n+sin 3x
>0

61
 • ∀x ∈ R fixed:
1 1
un+1 (x) = √ ≤ un (x) = √ ,n ∈ N
2 n + 1 + sin 3x 2 n + sin 3x

and so (un (x))n is decreasing.

• The sequence (un )n converges uniformly to zero over R because

1 1
|un (x)| = √ ≤ √ → 0, n → +∞.
2 n + sin 3x 2 n−1

2.2.8 Uniform Convergence and Continuity

P+∞
Theorem 2.2.30. Let n=0 fn , fn : A → R be a series of functions satisfying the
following conditions:

(1) ∀n ∈ N : fn ∈ C(A);
(2) The series +∞
P
n=0 fn (x) = S(x) converges uniformly over the set A.

Then S ∈ C(A)

Proof. It suffices to show the continuity of S at a point x0 ∈ A. Therefore, it is sufficient

to apply the theorem (Sequences of continuous functions) to the sequence (Sn )n of
partial sums Sn (x) = nk=0 fk (x).
P

Remark. Under the same assumptions as in Theorem 2.2.30, we can thus write:
+∞
X +∞
X +∞
X
lim fn (x) = lim fn (x) = fn (x0 ) .
x→x0 x→x0
n=0 n=0 n=0

In addition,

e−n|x|
Example 2.2.31. The function x 7→ +∞
P
n=1 n2 is continuous over R. Indeed, it is
the sum of a series of continuous functions that converges normally (hence uniformly)
over R since
e−n|x| 1
∀x ∈ R, 2
≤ 2.
n n

Example 2.2.32. The series

+∞
X 1
x2 + (n + 1)2
n=0
1
is uniformly convergent on R and lim 2 2 = 0. Hence
x→∞ x +(n+1)

+∞ +∞
X 1 X 1
lim S(x) = lim 2 2
= lim 2 = 0.
x→∞ x→∞ x + (n + 1) x→∞ x + (n + 1)2
n=0 n=0

62
Example 2.2.33. The series
+∞
X x
(1 + x2 )n
n=0

converges pointwise over R, and its sum

+∞
(
1+x2
X x x ,x ∈ R∗ ,
S(x) = =
(1 + x2 )n 0, x = 0,
n=0

x
but not uniformly over R because each un (x) = (1+x 2 )n (where n ∈ N) is continuous at

0, while the sum S(x) is discontinuous at this point since

1 + x2
lim S(x) = lim = ∞ and S(0) = 0
x→0 x→0 x

2.2.9 Term-by-Term Integration

P+∞
Theorem 2.2.34. Let n=0 fn , fn : [a, b] → R be a series of functions satisfying the
following conditions:

(1) ∀n ∈ N : fn ∈ R([a, b])

(2) The series +∞
P
n=0 fn = S converges uniformly over [a, b].

Then, S ∈ R([a, b]), and we have:

+∞
Z bX +∞ Z
X b
fn (x)dx = fn (x)dx.
a n=0 n=0 a

Proof. We apply the theorem of sequences of integrable functions to the sequence of

partial sums.

Example 2.2.35. Show that

+∞
1 x n −t
X Z
∀x ∈ R : t e dt = x.
n! 0
n=0
P+∞
Let x ∈ R be fixed. Consider the series of functions n=0 un , where

tn e−t
un : [0, x] → R, t 7→ .
n!
Since
tn e−t |x|n e−|x|
sup |un (t)| = sup ≤
t∈[0,x] t∈[0,x] n! n!
|x|n e−|x|
and the series +∞ converges, then the series +∞
P P
n=0 n! n=0 un (t) converges normally,
thus uniformly over [0, x]. According to Theorem 2.2.34, we can write:
+∞ +∞ n
!
1 x n −t
Z Z x X Z x
X t −t
t e dt = e dt = et e−t dt = x.
n! 0 0 n! 0
n=0 n=0

63
Example 2.2.36. Show that
+∞
# #
X (−1)n+1
ln(1 + x) = xn , x ∈ − 1, 1 .
n
n=1

If |t| < 1, then

+∞
1 X
= (−1)n tn .
1+t
n=0

This series converges uniformly over [−α, α] (where 0 < α < 1) because

∀n ∈ N, ∀t ∈ [−α, α] : |(−1)n tn | ≤ αn

and the series +∞ n

P
n=0 α converges. Therefore, we can integrate the above series term
by term from 0 to x, where |x| < 1. Thus, for |x| < 1, we have:
+∞ +∞
Z x Z x X ! Z x
dt n n
X
n
= (−1) t dt = (−1) tn dt
0 1+t 0 n=0 0 n=0
+∞ +∞
X xn+1 X (−1)n+1 n
= (−1)n = x ,
n+1 n
and so
+∞
#
X (−1)n+1 n
ln(1 + x) = x , x ∈ − 1, 1[.
n
n=1

By using the continuity of the function ln(1 + x) at x = 1,

lim ln(1 + x) = ln 2
x→1

and the uniform convergence of the series

+∞
X (−1)n+1
xn
n
n=1

on [0, 1], we obtain

+∞ +∞
X (−1)n+1 X (−1)n+1
lim xn = = ln 2.
x→1 n n
n=1 n=1

Thus
+∞
# #
X (−1)n+1 n
ln(1 + x) = x , x ∈ − 1, 1 .
n
n=1

2.2.10 Derivation term by term

P+∞
Theorem 2.2.37. Let n=0 fn , fn : [a, b] → R be a series of functions satisfying the
following conditions:
P+∞
(1) ∃x0 ∈ [a, b] for which the series n=0 fn (x0 ) converges.

64
(2) ∀n ∈ N : fn ∈ C 1 ([a, b])
(3) The series +∞ ′
P
n=0 fn (x) converges uniformly on [a, b].

Then the series

+∞
X
fn (x), x ∈ [a, b]
n=0

converges uniformly on [a, b], and its sum S ∈ C 1 ([a, b]), and we have
+∞
! +∞
d X X
fn (x) = fn′ (x), x ∈ [a, b].
dx
n=0 n=0

Proof. We apply the theorem of sequences of differentiable functions to the sequence

of partial sums.

Example 2.2.38. Define

un :] 0, +∞[ → R
(−1)n x
x 7→ .
n+x
Study the differentiability of the sum S(x) = +∞
P
n=1 un (x) on ]0, +∞[. Assume

+∞
X (−1)n
S(x) = x
n+x
n=1

P+∞ (−1)n
According to Leibniz’s rule, the series S1 (x) = n=1 n+x converges uniformly for
∗ (−1)n
every x ∈ ] 0, +∞ [ . For all n ∈ N , gn (x) = n+x is of class C 1 on ]0, +∞[, and

(−1)n+1

∀x ∈]0, +∞ : gn′ (x) =
(n + x)2

We have ∀n ∈ N∗ , ∀x ∈] 0, +∞[ :

(−1)n+1 1 1
gn′ (x) = = ≤ 2
(n + x)2 (n + x)2 n

As +∞
P 1 P+∞ ′
n=1 n2 converges, we deduce that n=1 gn (x) converges normally, hence uniformly
on ]0, +∞[. According to Theorem 2.2.11, we conclude that S1 is of class C 1 on ]0, +∞[
and consequently S(x) = xS1 (x) is also of class C 1 on ]0, +∞[. Moreover,
+∞
"
X (−1)n n
∀x ∈]0, +∞ : S ′ (x) = S1 (x) + xS1′ (x) =
(n + x)2
n=1

65
Chapter 3

Power Series

Power series are one of the most useful type of series in analysis. For example, we can
use them to define transcendental functions such as the exponential and trigonometric
functions (and many other less familiar functions).

3.1 Introduction

A power series (centered at 0 ) is a series of the form

∞
X
an xn = a0 + a1 x + a2 x2 + · · · + an xn + . . .
n=0
where the an are some coefficients. If all but finitely many of the an are zero, then
the power series is a polynomial function, but if infinitely many of the an are nonzero,
then we need to consider the convergence of the power series. Power series work just
as well for complex numbers as real numbers, and are in fact best viewed from that
perspective, but we restrict our attention here to real-valued power series.
Definition 3.1.1. Let (an )∞ n=0 be a sequence of real numbers and c ∈ R. The power
series centered at c with coefficients an is the series
∞
X
an (x − c)n . (3.1.1)
n=0

Here are some power series centered at 0:

∞
X
xn = 1 + x + x2 + x3 + x4 + . . .
n=0
∞
X 1 n 1 1 1
x = 1 + x + x2 + x3 + x4 + . . .
n! 2 6 24
n=0
X∞
(n!)xn = 1 + x + 2x2 + 6x3 + 24x4 + . . .
n=0
∞
n
X
(−1)n x2 = x − x2 + x4 − x8 + . . .
n=0

66
and here is a power series centered at 1
∞
X (−1)n+1 1 1 1
(x − 1)n = (x − 1) − (x − 1)2 + (x − 1)3 − (x − 1)4 + . . .
n 2 3 4
n=1

We will first study power series of the form

+∞
X
an xn .
n=0

Note that any series (3.1.1) can be brought to this form through the change of variables
x − x0 → y → x.

Example 3.1.2. The series

+∞
X
(−1)n xn
n=0
is obviously a power series. The same applies to the series.
+∞
X
(−1)n (x − x0 )n
n=0

This series is convergent if and only if |x − x0 | < 1, and its sum is

1
S(x) = , |x − x0 | < 1.
1 + x − x0
Theorem 3.1.3 (Abel). If the power series
+∞
X
an xn
n=0

converges at a point x0 ̸= 0, then it converges absolutely at every point x such that

|x| < |x0 |. Conversely, If +∞ n
P
̸ 0, then it is divergent at
n=0 an x diverges at a point x0 =
every point x such that |x| > |x0 |.
P+∞ n
Proof. Since the series n=0 an x0 converges, we have

lim an xn0 = 0.
n→+∞

Therefore, there exists M > 0 such that

∀n ∈ N : |an xn0 | ≤ M

Then, for every x such that |x| < |x0 |, we have

n n
xn x x
|an xn | = an xn0 n = |an xn0 | ≤M ,
x0 x0 x0
Hence,
+∞ +∞ n
X
n
X x
|an x | ≤ M < +∞
x0
n=0 n=0

67
 x
as < 1, which completes the proof of the first assertion.
x0

Assume now that the series +∞ n

P
n=0 an x diverges at some point x0 , and suppose that
there exist at least x1 such that |x1 | > |x0 | for which the series +∞ n
P
n=0 an x1 converges,
P+∞
this implies that the series n=0 an xn converges for all x satisfying |x| < |x1 |, which is
a contradiction.

Remark. It seems that we can replace the condition ”converges at a point x0 ̸= 0”

in Abel’s lemma with ”the sequence (an xn0 )n is bounded”.

P+∞ n
Theorem 3.1.4. If the power series n=0 an x converges on the interval ] − a, a[
(a > 0), then it converges normally on every interval [−r, r], where r < a.

Proof. Let’s fix R, with r < R < a. There exists M > 0, such that

∀n ∈ N : |an Rn | ≤ M.

As in the proof of Abel’s Lemma, we obtain for every x ∈ [−r, r]

xn x n x n  r n
|an xn | = an Rn n
= |an Rn | ≤M ≤M ,
R R R R

and we are done.

3.2 Radius of convergence

Before we introduce the concept of the radius of convergence, we present the following
result.

Theorem 3.2.1. Let

+∞
X
an xn (3.2.1)
n=0

be a power series. Then, there exists one and only one number R such that

• If |x| < R, the series (3.2.1) converges absolutely,

• If |x| > R, the series (3.2.1) diverges.

Proof. Let E be the set of non-negative real numbers r ≥ 0 such that the sequence
(|an | rn )n is bounded. The set E is not empty since it contains at least 0. If E is not
upper-bounded, we set R = +∞. If E is upper-bounded, we set R = sup E, which is the
least upper bound of the set E (the existence of sup E is guaranteed by the supremum
axiom).

68
 • Suppose |x| < R. Then there exists r0 ∈ E such that |x| < r0 < R. The sequence
(|an | r0n ) is bounded. If, in Abel’s Lemma, we take x0 = r0 , we see that the series
an xn converges absolutely.
P

• Let’s suppose |x| > R. Then |x| is not in the set E. Therefore, the sequence (an xn )
an xn diverges because its
P
is unbounded. We deduce that the numerical series
general term does not tend towards 0.

Let’s prove the uniqueness of R. If there existed R and R′ such that 0 ≤ R < R′ ≤ +∞,
both satisfying properties 1) and 2), then for a chosen number r such that R < r < R′ ,
an rn would have to both converge and diverge, which is absurd.
P
the series

The following definition therefore makes sense for every power series.

Definition 3.2.2. The element R in R+ = R+ ∪ +∞ defined by

R = sup {r ∈ R+ , the sequence (an rn ) is bounded }

an xn . The set {x ∈ R; −R <

P
is called the radius of convergence of the power series
an xn .
P
x < R} is the interval of convergence of the power series

More generally, we can define the radius of convergence as follows.

Definition 3.2.3. If the power series

∞
X
an (x − c)n
n=0

converges for |x − c| < R and diverges for |x − c| > R, then 0 ≤ R ≤ ∞ is called the
radius of convergence of the power series.

Theorem 3.2.1 does not say what happens at the endpoints x = c ± R, and in general
the power series may converge or diverge there. We refer to the set of all points where
the power series converges as its interval of convergence, which is one of

(c − R, c + R), (c − R, c + R], [c − R, c + R), [c − R, c + R]

In addition, Theorem 3.2.1 does not give an explicit expression for the radius of con-
vergence of a power series in terms of its coefficients. The ratio test gives a simple, but
useful, way to compute the radius of convergence, although it doesn’t apply to every
power series.

Theorem 3.2.4 (Ratio test). Suppose that an ̸= 0 for all sufficiently large n and
the limit
an
R = lim
n→∞ an+1

exists or diverges to infinity. Then the power series

∞
X
an (x − c)n
n=0

has radius of convergence R.

69
Proof. Let
an+1 (x − c)n+1 an+1
r = lim = |x − c| lim .
n→∞ an (x − c)n n→∞ an
By the ratio test, the power series converges if 0 ≤ r < 1, or |x − c| < R, and diverges
if 1 < r ≤ ∞, or |x − c| > R, which proves the result.

Example 3.2.5. Consider the series

+∞
X
(nx)n
n=1

According to the rule of d’Alembert, we have an = nn

an nn 1 1
= =
,
an+1 (n + 1) n+1 n + 1 1 + n1 n

So,
an 1
lim = 0 · = 0.
n→+∞ an+1 e
Therefore, R = 0 and ∆ = {0}.

Remark. If the sequence with the general term an+1 an has no limit in R+ , the rule of
D’Alembert is inapplicable. This is, for example, the case for the power series
X
(2 + (−1)n ) xn .

The rule of D’Alembert is also inapplicable for series of the type

2
X X X
an x2n , an x2n+1 , an xn . . .

Example 3.2.6. Consider the series

+∞
X (x − 2)2n−1
.
n3
n=1

If x = 2, the series converges trivially. For a fixed x ∈ R\2, we define

(x − 2)2n−1
un (x) = ,
n3
we have
un+1 (x) (x − 2)2n+1 n3 2 n3
= = (x − 2) → (x − 2)2 , n → +∞
un (x) (x − 2)2n−1 (n + 1)3 (n + 1)3

So the series converges absolutely if |x − 2| < 1 and diverges if |x − 2| > 1, we conclude

that R = 1. We have
|x − 2| < 1 ⇔ 1 < x < 3.

Let’s examine the series for x = 1, x = 3 :

70
For x = 1 , we obtain the series
+∞
X −1
n3
n=1

which is convergent.
For x = 3, we obtain the convergent series
+∞
X 1
.
n3
n=1

Therefore, the interval of convergence is ∆ = [1, 3].

The root test gives an expression for the radius of convergence of a general power series.

Theorem 3.2.7 (Cauchy-Hadamard). The radius of convergence R of the power series

∞
X
an (x − c)n
n=0

is given by 
 0, ρ = +∞

R = ρ1 , 0 < ρ < +∞

+∞, ρ = 0


where
p
n
ρ = lim |an |,
n→+∞

Proof. Let
r = lim sup |an (x − c)n |1/n = |x − c| lim sup |an |1/n
n→∞ n→∞

By the root test, the series converges if 0 ≤ r < 1, or |x − c| < R, and diverges if
1 < r ≤ ∞, or |x − c| > R, which proves the result.

Example 3.2.8. Consider the series

+∞
X xn
(ln n)5n
n=2

We have
1 √ 1 √
an = 5n
⇒ n an = 5
⇒ lim n an = 0
(ln n) (ln n) n→+∞

Therefore, R = +∞, and the series is absolutely convergent for every x ∈ R.

Example 3.2.9. Consider the series

+∞
X (3 − (−1)n )n
xn .
ln n
n=2

71
To calculate the radius of convergence, we will use the Cauchy-Hadamard rule. We have
n )n
an = (3−(−1)
ln n , and
p
n 3 − (−1)n √ 2 √ 4
|an | = √ n
; 2k √
a2k = 2k ; 2k+1 a2k+1 = 2k+1
p .
ln n ln 2k ln(2k + 1)
Hence
p
n
lim |an | = 4.
n→+∞

Hence R = 14 . Therefore, the series converges absolutely if |x| < 1

4 and diverges if
|x| > 14 . Let’s examine the series for x = − 41 , x = 14 .
For x = 41 , we have
(
 n 1 1
(3 − (−1)n )n 1 n 2k · , n = 2k
 
1
un = an = = 2 1 ln(2k)
4 ln n 4 ln(2k+1) , n = 2k + 1

If we assume
( (
1 1
0, n = 2k, 22k
· ln(2k) , n = 2k,
cn = 1 ∗
, dn =
ln(2k+1) , n = 2k + 1 (k ∈ N ) 0, n = 2k + 1 (k ∈ N∗ ) ,

then un = cn + dn (n ≥ 2). Since dn ≥ 0 for n ≥ 2, we have un = cn + dn ≥ cn ≥ 0 for

n ≥ 2. As the series +∞
P P+∞ 1
n=2 cn = k=1 ln(2k+1) diverges, we conclude the divergence of
P+∞
the series n=2 un .
For x = − 14 , we consider the series +∞
P n
P+∞
n=2 (−1) un = (dn − cn ). Since the series
P+∞ P+∞ 1 P+∞ n=2 P+∞ 1 1
n=2 cn = k=1 ln(2k+1) diverges, and the series n=2 dn = k=1 22k · ln(2k) converges,
P+∞
the series n=2 (−1)n un diverges. Therefore, the series diverges for x = ± 14 , and the
interval of convergence is ∆ =] − 41 , 14 [.

Theorem 3.2.10. Let ] − R, R[ be the interval of convergence of the power series

+∞
X
an xn .
n=0

Then the series converges normally (hence absolutely and uniformly) in each interval
[a, b] ⊂] − R, R[.

Proof. It is immediate according to Theorem 3.1.4.

3.3 Comparison of convergence radii

an xn and bn xn are two power series with respective conver-

P P
In this paragraph,
gence radii Ra and Rb . The following two results immediately follow from the Cauchy-
Hadamard formula. Nevertheless, we provide a proof based on the definition of the
convergence radius.

Theorem 3.3.1. If for every n ∈ N, we have |an | ≤ |bn |, then Ra ≥ Rb .

72
Proof. Follows from the definition of the convergence radius. Indeed, let r ∈ R+ such
that the sequence (bn rn ) is bounded. Since |an rn | ≤ |bn rn |, the sequence (an rn ) is also
bounded. Thus, we have the inequality: Ra ≥ Rb .

Example 3.3.2. For every n ∈ N, we have e−1 ≤ ecos n ≤ e. Since the power series
P −1 n P n P cos n n
e x and ex both have a convergence radius equal to 1, the series e z
also has a convergence radius equal to 1.

Corollary 3.3.3. if an = O (bn ), then Ra ≥ Rb .

Proof. By definition, there exists M > 0 and a rank N such that for all n ≥ N ,
M bn xn has a convergence radius of Rb . The previous
P
|an | ≤ M |bn |. The power series
theorem allows us to conclude that indeed Ra ≥ Rb .

Corollary 3.3.4. If an ∼ bn , then Ra = Rb .

n→+∞

Proof. If an ∼ bn , then an = O (bn ) and bn = O (an ). Therefore, according to the

previous corollary, we have Ra ≥ Rb and Rb ≥ Ra . Hence, the announced equality.

an xn be a power series such that there exists a non-zero ra-

P
Corollary 3.3.5. Let
tional function F with, for every n ∈ N, an = F (n). Then the convergence radius of
an xn is equal to 1.
P

Proof. There exist two non-zero polynomials P and Q such that F = P/Q. Denoting
λX α (resp. µX β ) as the highest-degree term of P (resp. Q), we have λµ ̸= 0, and

λ α−β
|an | ∼ n .
n→+∞ µ

Since α−β
(n + 1)α−β

1
= 1+ −→ 1
nα−β n n→+∞

the radius of convergence of the power series

+∞
X λ α−β n
n x
µ
n=1

an xn is also equal to 1
P
is equal to 1 according to the ratio test. Therefore, that of
according to the previous proposition.

3.4 Operations on Power Series

an xn and bn xn are two power series, we call

P P
Definition 3.4.1. If

• Sum series, the power series (an + bn ) xn ,

P

• Product series by λ ∈ C, the power series (λan ) xn or λ an xn .

P P

73
 • Product series (also called Cauchy product), the power series cn xn defined by
P

n
X
∀n ∈ N, cn = ak bn−k
k=0

an xn and bn xn be two power series with respective conver-

P P
Theorem 3.4.2. Let
gence radii Ra and Rb . Then,

(1) The convergence radius R of the sum series satisfies R ≥ min (Ra , Rb ), with equal-
ity if Ra ̸= Rb . Moreover, if |x| < min (Ra , Rb ), then
+∞
X +∞
X +∞
X
n n
an x + bn x = (an + bn ) xn
n=0 n=0 n=0

an xn by λ ̸= 0 does not change the convergence radius. Moreover,

P
(2) multiplying
if |x| < Ra , then
+∞
X +∞
X
n
λ an x = (λan ) xn
n=0 n=0

(3) the convergence radius R of the product series satisfies R ≥ min (Ra , Rb ). More-
over, if |x| < min (Ra , Rb ), then
+∞
! +∞ ! +∞
X X X
n n
an x bn x = cn xn
n=0 n=0 n=0

where
n
X
∀n ∈ N, cn = ak bn−k .
k=0

Proof. (1) If |z| < Ra and |z| < Rb , both power series converge absolutely. The sum
power series also converges, as the triangle inequality yields

|(an + bn ) z n | ≤ |an z n | + |bn z n |

Therefore, we have R ≥ min (Ra , Rb ). If Ra ̸= Rb , for example, if Ra < Rb , then for

bn z n converges, and an z n diverges. In
P P
all z such that Ra < |z| < Rb , the series
this case, there cannot be convergence, and consequently, absolute convergence of the
sum series. In this scenario, we indeed have R = min (Ra , Rb ). The final formula follows
from the one giving the sum of two convergent series.
(2) The proof is straightforward (note that if multiplied by 0, we obtain the zero power
series, with a convergence radius equal to +∞).
(3) For |z| < Ra and |z| < Rb , both power series converge absolutely. Their product
series also converges according to Theorem 5.3 of Chapter 2. Therefore, we have R ≥
min (Ra , Rb ). The formula for the product of two power series comes from the same
Theorem 5.3.

74
Remark. For the convergence radius R of the sum series and the product series, it is
P n P n
possible to have R > max (Ra , Rb ). For instance, the series z and −z both have
a convergence radius of 1, but the sum series is the zero series, hence with a convergence
an xn and bn xn
P P
radius equal to +∞. Similarly, consider the series
( (
2 if n = 0 −1 if n = 0
an = n
and bn =
2 if n ≥ 1 1 if n ≥ 1
For n ≥ 1, we have
1
|an |1/n = 2 and |bn |1/n = 1, hence Ra = and Rb = 1.
2
cn z n is defined by
P
The product series
n−1
X n−1
X
2k − 2n = 2 + 2 2n−1 − 1 − 2n


cn = a0 bn + ak bn−k = 2 +
k=1 k=1
so cn = 0 for all n ≥ 0. Therefore, the product series is the zero series, and its conver-
gence radius is equal to +∞.

3.4.1 Differentiation of power series

an z n , the power series

P
Definition 3.4.3. We call the derivative power series of
(n + 1)an+1 z n .
P

an z n be a power series with a convergence radius of R. Then,

P
Theorem 3.4.4. Let
the convergence radius R′ of its derivative power series is equal to R.

Proof. Let R and R′ denote the convergence radii of the power series an z n and
P

(n + 1)an+1 z n respectively.
P

If |z| > R, then the sequence (an z n ) is unbounded. Consequently, the sequence
((n + 1)an+1 z n ) is also unbounded, implying |z| ≥ R′ . Therefore, R ≥ R′ .
If |z| > R′ , let’s choose a number ρ such that |z| > ρ > R′ . The sequence
ρ |z| n
 
n n−1
|an z | = n |an | ρ ×
n ρ
is the product of two sequences with positive terms, one of which is unbounded (since
ρ > R′ ), and the other tends to +∞ (as |z| > ρ). Consequently, the sequence (|an z n |)
is unbounded, implying |z| ≥ R. Therefore, R′ ≥ R.

Remark. By applying the previous result to a primitive power series, that is,
P an n+1 P an n+1
an z n , we observe that
P
n+1 z instead of n+1 z has the same convergence
n
P
radius as an z .

an z n be a power series, and F ̸≡ 0 be a rational function. It is clear

P
Remark. Let
F (n)an z n
P
that the proof of the previous proposition can be adapted to show that
an z n .
P
has the same convergence radius as

75
3.5 Properties of the sum of a power series

3.5.1 Continuity of the Sum Function

an xn be a power series, R its radius of convergence, and S its

P
Theorem 3.5.1. Let
sum. Then, the function S is continuous on the the interval of convergence ] − R, R[.

Proof. The series is uniformly convergent on every interval [a, b] ⊂] − R, R[. Hence, the
sum of the power series is continuous within the interval of convergence ] − R, R[.

Example 3.5.2. The ratio test easily shows that the series (z n /n!) has a convergence
P
P n
radius equal to +∞. Therefore, its sum function z 7→ (z /n!) is continuous on R.

Corollary 3.5.3. Let R > 0 be the convergence radius of the power series
+∞
X
an xn .
n=0

If this series converges for x = R, then its sum is left-continuous at this point, and
+∞
X +∞
X
n
lim an x = an Rn
x→R−
n=0 n=0

Proof. We have
+∞
X +∞
X  x n
an xn = an R n · .
R
n=0 n=0

is uniformly convergent on the interval [0, R] because the series +∞ n

P
n=0 an R converges,
x n


and the sequence vn = R , x ∈ [0, R] is decreasing and bounded.
 x n+1  x n  x n
∀n ∈ N, ∀x ∈ [0, R], vn+1 = ≤ = vn , vn = ≤1
R R R
P+∞ n
Therefore, the sum of the power series n=0 an x is continuous on the interval [0, R].

3.5.2 Integration of the Sum Function

an xn be a power series with a convergence radius of R > 0. If

P
Theorem 3.5.4. Let
[a, b] is a segment included in the convergence interval ] − R, R[, then
Z b +∞
X Z b
S(x)dx = an xn dx.
a n=0 a

In particular,
Z x +∞
X an n+1
∀x ∈] − R, R[: S(t)dt = x .
0 n+1
n=0

Proof. Trivial!

76
Remark. On the convergence interval, one can integrate a power series infinitely many
times, and the resulting series maintains the same convergence radius.

Example 3.5.5. Calculate the sum of the series

+∞
X (−1)n
.
2n + 1
n=0

We have
+∞ +∞ +∞ Z x
X (−1)n X (−1)n x2n+1 X
= lim = lim (−1)n t2n dt
2n + 1 x→1− 2n + 1 x→1− 0
n=0 n=0 n=0

+∞
!
Z x Z x
X
n 2n 1 π
= lim (−1) t dt = lim 2
dt = lim arctan x = .
x→1− 0 x→1− 0 1+t x→1 − 4
n=0

3.5.3 Differentiability of the Sum Function

Theorem 3.5.6. Let +∞ n

P
n=0 an x is a power series with radius R > 0 and S as its sum,
then S ∈ C 1 (] − R, R[), and
+∞
X
∀x ∈] − R, R[: S ′ (x) = nan xn−1 .
n=1

Proof. The functions fn : x 7→ an xn are of class C 1 , the series an xn converges

P
P ′
pointwise on ] − R, R[, and the series of derivatives fn converges uniformly on any
compact subset of ] − R, R[, according to Theorem 3.1. The theorem 5.1 from Chapter
4 allows us to conclude that S ′ is the sum of the series
P ′
fn .

Corollary 3.5.7. Under the assumptions of the above theorem, the sum function S is
of class C ∞ on ] − R, R[, and
+∞
(k)
X k!
∀k ∈ N, ∀x ∈] − R, R[, S (x) = an xn−k .
(n − k)!
n=k

In particular, for any k ∈ N,

S (k) (0)
ak = .
k!
Example 3.5.8. We have seen that the geometric series xn has a convergence radius
P

of 1, and
+∞
X 1
∀x ∈] − 1, 1[, xn = .
1−x
n=0

By differentiation, for all x ∈] − 1, 1[,

  +∞ +∞
1 d 1 X X
2
= = nxn−1 = (n + 1)xn
(1 − x) dx 1−x
n=1 n=0

77
By immediate induction on k, we deduce that
+∞
!
1 dk 1 dk
 
1 1 X
= = xn
(1 − x)k+1 k! dxk 1−x k! dxk
n=0
+∞ +∞
1 X X (n + k)! n
= (n + k) . . . (n + 1)xn = x
k! k!n!
n=0 n=0

Corollary 3.5.9. Suppose two power series

∞
X
f (x) = ak (x − c)k
0

and
∞
X
g(x) = bk (x − c)k
0

agree on some interval centered at c, that is f (x) = g(x) for |x − c| < ρ and some
positive ρ. Then ak = bk for all k = 0, 1, 2, . . ..

Proof. Proof It follows immediately from the fact

f (k) (c) g (k) (c)

ak = = = bk
k! k!
for all k = 0, 1, 2, . . ..

Example 3.5.10. The series for the exponential function

∞
x
X xk
e =
k!
k=0

reveals one of the key facts about the exponential function, namely that it is its own
derivative. Note simply that
∞ ∞ ∞
d x d X xk X d xk X xk−1
e = = = = ex .
dx dx k! dx k! k − 1!
k=0 k=0 k=1

3.6 Taylor Series and Maclaurin series

In this Section we examine how functions may be expressed in terms of power series.
This is an extremely useful way of expressing a function since (as we shall see) we
can then replace ’complicated’ functions in terms of ’simple’ polynomials. The only
requirement (of any significance) is that the ’complicated’ function should be smooth;
this means that at a point of interest, it must be possible to differentiate the function
as often as we please.

78
Definition 3.6.1. Let f be a function of class C ∞ in a neighborhood of x0 . The power
series
+∞ (n)
X f (x0 )
(x − x0 )n
n!
n=0

is called the Taylor series of f in the neighborhood of x0 . If x0 = 0, this series is also

called the Maclaurin series.

We have seen that the sum f of the power series

+∞
X
an (x − x0 )n
n=0

is of class C ∞ within the interval of convergence, and moreover

f (n) (x0 )
∀n ∈ N : an = .
n!
We can thus state
P+∞
Theorem 3.6.2. Every power series n=0 an (x − x0 )n is the Taylor series of its sum
in the neighborhood of x0 .

The question is: under what conditions can we express the function f as a Taylor series?
In other words, what are the conditions under which we can write
∞
X f (k) (x0 )
f (x) = (x − x0 )k
k!
k=0

According to the foregoing, a necessary condition for f to be expandable as Taylor

series is (
f ∈ C∞
f (k) (x0 )
ak = k!

However, the converse is not generally true; a function f possessing derivatives of every
order at x0 is not necessarily equal to the corresponding power series
∞
X f (k) (x0 )
(x − x0 )k
k!
k=0

The following example shows that the Taylor series for f (x) converges, but its sum
differs from f (x) at every point, except the central one. For example, one can consider
the function ( 2
e−1/x , x ̸= 0,
f (x) =
0, x = 0.
This function is of class C ∞ in R. Indeed, for x ̸= 0, we have
 
2 2 6 4 2
f ′ (x) = 3 e−1/x , f ′′ (x) = − 4 + 6 e−1/x , . . .
x x x

79
By induction, we have
 
1 2
f (n)
(x) = Pn e−1/x (n = 1, 2, . . .),
x
where Pn x1 is a polynomial of degree 3n. Let’s show that all derivatives of this function

exist at x = 0, and they are all zero. We have

2
e−1/x
 
′ f (x) − f (0) y 1
f (0) = lim = lim = lim y2 = 0 y=
x−→0 x−0 x−→0 x y−→∞ e x
So, f ′ (0) = 0. We now demonstrate by induction; suppose that f (n−1) (0) = 0 is true
and show that f (n) (0) = 0. We have
f (n−1) (x) − f (n−1) (0)
f (n) (0) = lim
x−→0 x−0
1

−1/x2  
Pn−1 x e yPn−1 (y) 1
= lim = lim =0 y= .
x−→0 x y−→∞ ey2 x
Thus, f (x) is infinitely differentiable. Considering the central point x0 at the origin, we
obtain the following Taylor series of f (x) :
∞ ∞ ∞
X X f (n) (0) X
cn xn = xn = 0 · xn = 0 + 0 + 0 + . . . = 0
n!
n=0 n=0 n=0
This series converges at every real point and the sum of the series is zero at ∀x ∈
(−∞, +∞), that is, the Taylor series constructed for the function f (x) converges to the
zero function that has nothing to do with f (x).
Definition 3.6.3. A function f whose Taylor series converges to f in a neighborhood
of c is said to be analytic at c. In addition, f is analytic on an open interval I if it is
analytic at each point of I.

As f : I → R is of class C ∞ on I, with x0 ∈ I, we have, according to the Taylor formula,

for every x ∈ I,
n
X f (k) (x0 )
f (x) = (x − x0 )k + Rn (x),
k!
k=0
where
f (n+1) (c)
Rn (x) = (x − x0 )n+1
(n + 1)!
is the remainder of the Taylor formula, and c = x0 + θ (x − x0 ), where 0 < θ < 1.

Remark. One can also use the integral form of the remainder in the Taylor expansion,
that is,
1 x
Z
Rn (x) = (x − t)n f (n+1) (t)dt
n! x0

Thus, a necessary and sufficient condition for f to be expandable as a power series is

lim Rn (x) = 0, for all x ∈ I.

n→∞
The verification of this condition is generally delicate. However, we have a sufficient
condition.

80
Theorem 3.6.4. If f (x) is infinitely differentiable on an interval (−R, R), R > 0 and
its derivatives satisfy the evaluations
n!
f (n) (x) ≤ M , ∀x ∈ (−R, R), ∀n ∈ N, (3.6.1)
Rn
where M is a constant, then f (x) is analytic on (−R, R).

Proof. It suffices to show that lim Rn (x) = 0. Using the remainder (at a = 0 ) in the
n→∞
f (n+1) (c) n+1
Lagrange form Rn (x) = (n+1)! x , we obtain for every fixed x in (−R, R) :
n+1
f (n+1) (c) n+1

|x|
|Rn (x)| = x ≤M .
(n + 1)! R
 n+1
Since |x|
R < 1, ∀x ∈ (−R, R), it follows that
|x|
R → 0, which implies that
n→∞
Rn (x) → 0, ∀x ∈ (−R, R).
n→∞

Remark. As usual, inequalities (3.6.1) can be imposed starting from some index n-
the violation of these inequalities for a finite number of derivatives does not affect the
resulting convergence of Rn .

3.6.1 Discussion

In practice, weaker versions of Theorem3.6.4 are frequently used, which impose stronger
restrictions on derivative. One of these formulations uses the condition

f (n) (x) ≤ M q n , ∀x ∈ (−R, R), ∀n ∈ N (3.6.2)

where q is a constant, and another employs the uniform boundedness of derivatives

f (n) (x) ≤ M, ∀x ∈ (−R, R), ∀n ∈ N. (3.6.3)

It is easy to show that the last restriction guarantees inequality (3.6.2), which in turn
guarantees condition (3.6.1) of Theorem 3.6.4. Indeed, without a loss of generality, we
can consider the constant q > 1 in (3.6.2) (if (3.6.2) holds for q ≤ 1, it is also valid
for q > 1 ). Then M < M q n , that is, (3.6.2) is more restrictive than (3.6.3). At the
next step, we compare q n and Rn!n . Notice that for any two constants q and R the ratio
qn (qR)n
n! = n! approaches 0 . To show this, we choose a fixed index N = [qR] and use the
Rn
representation n = N + k (n > N ) to obtain
k
(qR)n (qR)N +k (qR)N (qR)k (qR)N

qR
= = · < · → 0
n! (N + k)! N! (N + 1) · . . . · (N + k) N! N +1 k→∞

Example 3.6.5. Let f (x) = sin x. Since

 π 
∀n ∈ N : f (n) (x) = sin x + n ≤ 1
2

81
The hypothesis of the previous theorem is satisfied; thus, we have

x3 x5 x2n+1
sin x = x − + − · · · + (−1)n + ··· , x ∈ R.
3! 5! (2n + 1)!

Similarly, for f (x) = cos x

 π 
∀n ∈ N : f (n) (x) = cos x + n ≤ 1
2
Hence,
x2 x4 x2n
cos x = 1 − + − · · · + (−1)n + ··· , x ∈ R.
2! 4! (2n)!
Example 3.6.6. Let f (x) = ex . Applying the Taylor-Lagrange formula, we obtain:

x x2 xn
f (x) = 1 + + + ··· + + Rn (x)
1! 2! n!
where
eθx
Rn (x) = xn+1 , 0 < θ < 1
(n + 1)!
for any fixed x ∈ R

e|x|
|Rn (x)| ≤ |x|n+1 → 0, n → +∞
(n + 1)!

Hence,
x x2 xn
∀x ∈ R : ex = 1 + + + ··· + + ···
1! 2! n!
Similarly,
x x2 xn
∀x ∈ R : e−x = 1 − + + · · · + (−1)n + ···
1! 2! n!
and as a conclusion
ex − e−x x3 x5 x2n+1
∀x ∈ R : sinh x = =x+ + − ··· + + ··· ,
2 3! 5! (2n + 1)!
ex + e−x x2 x4 x2n
∀x ∈ R : cosh x = =1+ + + ··· + + ···
2 2! 4! (2n)!

82