Lectures 4: Sequence-series and sequence spaces

Benoit F. Sehba
December 19, 2024

1 Sequence of series
1.1 Definition and basic examples
Definition 1.1. A sequence-series or sequence of series is any sequence (an )n∈N
where each an is the sum of a convergent series.
P∞ j+1
Example 1. 1- an = j=1 (−1)j+n
P∞ 1
2- bn = j=1 (j+n) 2.

are sequence-series.
In general, by series-sequence, we mean sequence (an )n∈N defined by
X
an = ajn
j∈Z

where each series is convergent.

P∞ 1 P∞ 1
Remarque 1. The sequences an = j=1 j+n and bn = (−1)n j=1 j+n are
∞
X 1
not sequence-series since does not converge for any n ∈ N.
j=1
j + n

1.2 Convergence theorems

Let us consider again the example
∞
X 1
an =
j=1
j2 +n

We use that an is a p-series-like to see that it is a sequence-series. We will be

interested in knowing if (an )n is convergent and even ask for its limit when it
exists.
Warning: When considering for example the series
∞
X 1
bn = ,
j=1
j+n

1
one might be tempted to write
∞
X 1
lim bn = lim
x→∞
j=1
x→∞ j+n
=0

This is not only wrong for this example but wrong in general. We have to
give proper conditions under which the limit can be taken inside the sum sign.

1.2.1 Sequence-Series of positive terms

We put ourselves in the case where
∞
X
an = ajn
j=1

and

∀j, n ∈ N.
ajn > 0
P∞ 
Proposition 1.2. Let an = j=1 ajn be a sequence-series. Suppose that
n∈B
the following conditions are satisfied:
i for each j ∈ N fixed, the sequence (ajn )n∈N is increasing;
ii there exists a sequence of positive terms (bj )j∈N such that for n fixed

ajn ≤ bj ∀j ∈ N.

Assume that,
∞
X
bj <∞.
j=1

Then the sequence-series (an )n∈N is convergent.

Proof. We have by hypothesis that for each j fixed,

ajn ≤ aj(n+1) .
It follows that
∞
X ∞
X
an = ajn ≤ aj(n+1) = an+1 .
j=1 j=1

That is the sequence (an )n∈N is increasing.

2
 Also, by hypothesis, it holds that
∞
X ∞
X
an = ajn ≤ bj < ∞
j=1 j=1

That is (an )n∈N is bounded above. Thus (an )n∈N is convergent.

Proposition 1.3.
 Suppose thatfor each j fixed, the sequence (ajn )n∈N is de-
P∞
creasing. Then an = j=1 ajn is convergent.
n∈N

Proof. (an )n∈N is bounded below by 0. From the hypothesis, we obtain as above
that (an )n∈N is decreasing. Hence (an )n∈N is convergent.
Example 2.
∞
X 1
an =
j=1
j2 + n
1 1
Then ajn = j 2 +n . For j fixed, n 7→ j 2 +n is decreasing. Hence (an )n∈N is
convergent.
Theorem 1.4 (Monotone Convergence Theorm). Suppose that ajn > 0 and
that for each fixed j, n → ajn is increasing. Then
∞
X ∞
X
lim ajn = lim ajn .
n→∞ n→∞
j=1 j=1

Proof. Note that

∞
X N
X
ajn = lim ajn
N →∞
j=1 j=1
N
X
= sup ajn
N j=1

∞
X
But as an = ajn , we have that (an )n∈N is increasing since by hypothesis,
j=1
(ajn )n∈N is increasing. Hence

lim an = sup an
n→∞ n≥0

3
It follows that
∞
X
lim an = lim ajn
n→∞ n→∞
j=1
N
X
= sup sup ajn
n≥0 N ≥0 j=1

N
X
= sup sup ajn
N ≥0 n≥0 j=1

N
X
= sup sup ajn
N ≥0 j=1 n≥0

N
X
= sup lim ajn
N ≥0 j=1 n→∞

N
X
= lim ajn .
n→∞
j=1

Example 3. Consider the expression

 n X n  
1 n 1
1+ = .
n j=0
j nj

Put
ajn = nj n1j if j ≤ n


ajn = 0 if j > n.
Observe that for j ≤ n,
   
1 n(n − 1) · · · (n − j + 1) 1 1 j−1
ajn = = 1− ··· 1 − .
j! nj j! n n

Hence for each j fixed, the sequence n 7→ ajn is increasing. It follows from the
Monotone Convergence Theorem that
 n ∞ ∞
1 X X 1
lim 1+ = lim ajn = = e.
n→∞ n j=0
n→∞
j=0
j!

4
1.2.2 Some general convergence results.
Theorem 1.5 (Fatou’s Lemma). Suppose that ajn ≥ 0. Then,
∞
X ∞
X
lim inf ajn ≤ lim inf ajn
n→∞ n→∞
j=1 j=1

Proof
Let bjk = inf ajn . Then (bjk )k∈N is increasing. Thus
n≥k

lim bjk = sup bjk = supk inf ajn = lim inf ajn .
k→∞ k n≥k n→∞

As
bjk ≤ ajn for n ≥ k,
we have
∞
X ∞
X
bjk ≤ ajn , n ≥ k.
j=1 j=1

Hence
∞
X ∞
X
bjk ≤ lim inf ajn .
n→∞
j=1 j=1

Using the Monotone Convergence Theorem, we obtain that

∞
X ∞
X
lim inf ajn = lim bjk
n→∞ k→∞
j=1 j=1
∞
X
= lim bjk (we have applied the MCT)
k→∞
j=1
∞
X
≤ lim inf ajn .
n→∞
j=1

Theorem 1.6 (Dominated Convergence Theorem). Suppose that

(i) lim ajn = bj for each j ∈ N;
n→∞

(ii) there exists a positive sequence (λj )j∈N such that

|ajn | ≤ λj ∀j ∈ N

and
∞
X
λj < ∞.
j=1

5
Then
∞
X
lim |ajn − bj | = 0
n→∞
j=1

and
∞
X ∞
X
lim ajn = bj .
n→∞
j=1 j=1

Proof. Note that |ajn | ≤ λj ∀j ∈ N implies that λj + ajn > 0 and λj − ajn > 0.
Let us apply Fatou’s lemma to these two sequences. We first obtain
∞
X ∞
X ∞
X ∞
X ∞
X ∞
X
lim inf ajn + λj = lim inf (ajn + λj ) ≤ lim inf (ajn + λj ) = lim inf ajn + λj .
n→∞ n→∞ n→∞ n→∞
j=1 j=1 j=1 j=1 j=1 j=1

∞
X
Hence as λj < ∞, we conclude that
j=1

∞
X ∞
X
lim inf ajn ≤ lim inf ajn . (1)
n→∞ n→∞
j=1 j=1

We next apply Fatou’s lemma to the positive sequence λj − ajn to obtain

 
∞
X ∞
X ∞
X ∞
X ∞
X ∞
X
λj + lim inf (−ajn )+ = lim inf (λj − ajn ) ≤ lim inf (λj − ajn ) = λj +lim inf − ajn  .
n→∞ n→∞ n→∞ n→∞
j=1 j=1 j=1 j=1 j=1 j=1

Hence
∞
X ∞
X
− lim sup ajn ≤ − lim sup ajn ,
n→∞ n→∞
j=1 j=1

i.e
∞
X ∞
X
lim sup ajn ≤ lim sup ajn . (2)
n→∞ n→∞
j=1 j=1

From (1) and (2), we deduce that

∞
X ∞
X ∞
X ∞
X
lim sup ajn ≤ lim sup ajn = lim inf ajn ≤ lim inf ajn .
n→∞ n→∞ n→∞ n→∞
j=1 j=1 j=1 j=1

∞
X ∞
X
Since lim inf ajn ≤ lim sup ajn , and lim sup ajn = lim ajn = lim inf ajn ,
n→∞ n→∞ n→∞ n→∞ n→∞
j=1 j=1
we conclude that
∞
X ∞
X
lim ajn = lim ajn .
n→∞ n→∞
j=1 j=1

6
 Example of applications.

Example 4. Consider

∞
X 1
an =
j=1
(j + n)2

Then

1
ajn =
(j + n)2

we have

lim ajn = 0; (3)

n→∞
1
ajn ≤ = bj (4)
j2
∞ ∞
X X 1 π2
bj = = < ∞. (5)
j=1 j=1
j2 6

Hence by the Dominated convergence Theorem.

∞ ∞
X 1 X 1
lim 2
= lim
n→∞ (j + n) n→∞ (j + n)2
j=1 j=1

=0

Example 5. Put
∞
X n
an =
j=1
nj 2 + 1

so that
n
ajn = >0
nj 2 +1

7
n → ajn is increasing. Hence by the Monotone convergence Theorem,
∞
X 1
lim an = lim
n→∞
j=1
n→∞ (j + n)2
∞
X 1
= 2
j=1
j
π2
=
6
Remark that one can also use the Dominated Convergence Theorem for the
last example.

2 Sequence spaces.
Definition 2.1. Let 0 < p < ∞. we denote by ℓp (Z), the space of all sequences
(an )n∈Z such that
 1/p
X
∥(an )n ∥ℓp =  |aj |p  < ∞.
j∈Z

For p = ∞, the space ℓ∞ (N) consists of all sequences (an )n∈N such that

∥(an )n ∥ℓ∞ := sup |aj | < ∞.

j∈N

Remark:
Z can be replaced by N. In which case the space is denoted ℓ∞ (N).
Remarque 2. In the above definition, if N is replaced by Z, we have the space
ℓp (Z).
Lemma 2.2. Let 0 < p < ∞, Then the following holds.

i If 1 ≤ p ≤ ∞, then for any a ≥ 0 and b ≥ 0,

ap + bp ≤ (a + b)p ≤ 2p (ap + bp ).

ii If 0 < p ≤ 1, then for any a ≥ 0 and b ≥ 0

(a + b)p ≤ ap + bp ≤ 2(a + b)p

Proof. Consider the function

f (x) = (x + λ)p − xp − λp , λ > 0.

Then for x ≥ 0,

f ′ (x) = p(x + λ)p−1 − pxp−1 = p[(x + λ)p−1 − xp−1 ].

8
 clearly if p > 1, then f ′ (x) > 0 and as f (0) = 0, we conclude that f (x) ≥ 0.
That is
(x + λ)p ≥ xp + λp for p ≤ 1.
If p < 1, then f ′ (x) ≤ 0, Hence as f(0)=0, we obtain that

(x + λ)p ≤ xp + λp for 0 < p < 1.

Suppose that a < b. Then

(a + b)p < 2p bp < 2p (ap + bp ).

Also,
ap + bb ≤ 2bp ≤ 2(a + b)p .

Corollary 2.3. Let 0 < p ≤ 1. ℓ∞ (Z) and ℓp (Z) are linear spaces.
Proof. Exercise.

Definition 2.4. Let p ∈ [1, ∞). The conjugate exponent of p is the extended
number q such that
1 1
+ = 1.
p q
It is clear that if p = 1, q = ∞.
Exercise

1. Check that
q p
1+ = q; p− = 1; q(p − 1) = p; that q ∈ [1, ∞].
p q

2. Prove that if p ∈ [1, 2], then q ∈ [2, ∞].

p
Lemma 2.5 (Young’s Inequality). Let p ∈ (1, ∞) and q = p−1 its conjugate
exponent. Then
ap bq
ab ≤ + . f or all a, b ≥ 0
p q
with equality if and only if ap = bq .
p
Proof. fix b > 0 and consider the function g(a) = ab − ap , a ≥ 0.
Then lim g(a) = −∞; and g(0) = 0. We have g′(a) = b − ap−1
a→∞

g′(a) = 0 ⇔ b − ap−1 = 0
1
⇔ a = b p−1 .

9
We have
1
1 1 (b p−1 )p
g(a) = g(b p−1 ) = b p−1 (b) −
p
p
p b p−1
= bp−1 −
p
 
p 1
= b p−1 1−
p
 
1
= bq
q
q
b
= .
q
bq bq
Hence the maximum value of g is . That is for any a ≥ 0, g(a) ≤ or
q q
equivalently,
ap bq ap bq
ab − ≤ or ab ≤ + ,
p q p q
1
with equality if and only if a = b p−1 ⇔ ap = bq .
Theorem 2.6 (Hölder Inequality). Let p, q ∈ [1, ∞] be conjugate exponents. If
(an )n ∈ ℓp (Z) and (bn )n ∈ ℓq (Z), then the sequence (an bn )n belongs to ℓ1 (Z).
Moreover,
∥(an bn )n ∥ℓ1 ≤ ∥(an )n ∥ℓp ∥(bn )n ∥ℓq .
Proof. If an = 0 ∀n ∈ N or bn = ∀n ∈ N, the result is obvious. We then
assume that ∥(an )n ∥ℓp ̸= 0 and ∥(bn )n ∥ℓq ̸= 0.
By the Young’s inequality,
∥(an bn )n ∥ℓ1 X |an bn |
=
∥(an )n ∥ℓp ∥(bn )n ∥ℓq ∥(an )n ∥ℓp ∥(bn )n ∥ℓq
n∈Z
X  1  |an | p 1  |bn | q 
≤ +
p ∥(an )n ∥ℓp q ∥(bn )n ∥ℓq
n∈Z
X |an |p X |bn |q
= p +
p∥(an )n ∥ℓp q∥(bn )n ∥qℓq
n∈Z n∈Z
1 1
= ∥(an )n ∥pℓp + ∥(bn )n ∥qℓq
p∥(an )n ∥pℓp q∥(bn )n ∥qℓq
1 1
= +
p q
= 1.
Thus,
∥(an bn )n ∥ℓ1 ≤ ∥(an )n ∥ℓp ∥(bn )n ∥ℓq .
The proof is complete.

10
Theorem 2.7 (The Minkowski’s Inequality). If 1 ≤ p ≤ ∞ and (an )n , (bn )n ∈
ℓp (Z), then

∥(an )n + (bn )n ∥ℓp = ∥(an + bn )n ∥ℓp ≤ ∥(an )n ∥ℓp + ∥(bn )n ∥ℓp .

Proof. Case 1:
If p = 1, we obtain
X
∥(an + bn )n ∥ℓ1 = |an + bn |
n∈Z
X
≤ (|an | + |bn |)
n∈Z
X X
= |an |+ |bn |
n∈Z n∈Z
= ∥(an )n ∥ℓ1 + ∥(bn )n ∥ℓ1 .

Case 2:
If p = ∞, we easily obtain

∥(an + bn )n ∥ℓ∞ = sup |an + bn |

n∈Z
≤ sup(|an | + |bn |)
n∈Z
= sup |an | + sup |bn |
n∈Z n∈Z
= ∥(an )n ∥ℓ∞ + ∥(bn )n ∥ℓ∞ .

Case 3:
If p ∈ (1, ∞) and (an )n , (bn )n ∈ ℓp (Z), we have that

|an + bn |p = |an + bn ||an + bn |p−1

which implies that

|an + bn |p ≤ (|an | + |bn |)|an + bn |p−1 .

Hence
X X
|an + bn |p ≤ (|an | + |bn |)|an + bn |p−1
n∈Z n∈Z
X X
= |an ||an + bn |p−1 + |bn ||an + bn |p−1 . (6)
n∈Z n∈Z

Thus using Hölder Inequality (Theorem 2.6), we obtain

! p1 ! q1 ∞
! p1 ! q1
X X X X X
|an + bn |p ≤ |an |p |an + bn |q(p−1) + |bn |p |an + bn |q(p−1) .
n∈Z n∈Z n∈Z n=1 n∈Z

11
 1 1
p and q are conjugate and so can be written as 1 − and p = q(p − 1). Thus
q p
 ! p1 ! p1  !1− p1
X X X X
|an + bn |p ≤  |an | p
+ |bn |p  |an + bn |p
.
n∈Z n∈Z n∈Z n∈Z

! p1 −1
X
p
Multiplying both members by |an + bn | , we obtain
n∈Z

! p1 ! p1 ! p1
X X X
p p p
|an + bn | ≤ |an | + |bn | .
n∈Z n∈Z n∈Z

Which can be written equivalently as

||(an + bn )n ||ℓp ≤ ||(an )n ||ℓp + ||(bn )n ||ℓp .

The proof is complete.

Corollary 2.8. Let 1 ≤ p ≤ ∞. Then ℓp (N) is a linear normed space.

Theorem 2.9. Let q ≥ 1, and let (ajn )j∈Z,n∈N be a sequence of complex num-
bers. Suppose that there exist a sequence of positive number (bj )j∈Z ∈ ℓq (Z) and
a sequence of complex numbers (λj )j∈Z such that
(i) |ajn | ≤ bj for all j ∈ Z and , n ∈ N;

(ii) lim ajn = λj for all j ∈ Z.

n−→∞

Then the sequence (λj )j∈Z and for each n ∈ N, the sequence (ajn )j∈Z belong to
ℓq (Z). Moreover,
 1/q
X
lim  |ajn − λj |q  =0 (7)
n−→∞
j∈Z

and  1/q  1/q

X X
lim  |ajn |q  = |λj |q  . (8)
n−→∞
j∈Z j∈Z

Proof. This is just an application of the Dominated Convergence Theorem.

12