Vol.

#, 1–20 ©2024
http://doi.org/will_be_added_later

Bases criteria on topological vector spaces

1
Nacib Gurgel Albuquerque and Evandio Demétrio
2
Júnior
1
Universidade Federal da Paraíba, Departamento de Matemática, João Pessoa,
PB, CEP 58051-900, Brasil
2
Universidade Federal da Paraíba, Departamento de Matemática, João Pessoa,
PB, CEP 58051-900, Brasil

Dedicated to Professor Renato Tribuzy

on the occasion of his 75th birthday

Abstract. This work briefly overviews basis and basic sequence cri-
teria on topological vector spaces. While this theory in normed spaces
is well established, studying these sequences in more general spaces
is rare in the literature. Our main goal is to present the Banach-
Grumblum-Nikolskii criteria for basic sequences on general environ-
ments that are not necessarily locally convex, such as F -spaces and
quasi-Banach spaces.
Keywords: topological vector spaces, locally convex spaces, F -
spaces, basis, basic sequence.
2020 Mathematics Subject Classification: 46A16, 46A35,
46A08.

1 Introduction
Determining the precise form of a Hamel basis in an arbitrary infinite-
dimensional vector space is a rough problem in general. The Hamel basis
email: [email protected]

1
2 N. G. Albuquerque, E. Demétrio Jr.

theory does not provide satisfactory information on the topology of the

space. For instance, let X be a Banach space, let (ei )i∈I be an algebraic
basis in X, and also let (xn )∞ n=1 be a sequence in X converging to x. Each
xn and x can be represented as finite components xn = i αin ei and x =
P
n ∞
P
i αi ei . We would like that each sequence (αi )n=1 converges to α for each
i, but this generally does not occur. In other words, the projection maps
are not usually continuous. In order to solve this and similar challenges,
particularly in many crucial spaces with linear topologies, the following
alternative notion of basis is introduced.

Definition 1.1. A sequence (xn )∞ n=1 in a topological vector space X is

called basis if for each element x ∈ X there exists a unique sequence of
P∞
scalars (an )∞
n=1 such that x =
∞
n=1 an xn . When the sequence (xn )n=1 is
a basis for span{xn : n ∈ N}, it is called a basic sequence.

The uniqueness of the representation allows us to consider, for each

n ∈ N, the following linear map
 
X∞
x∗n : X −→ K, x∗n  aj xj  = an
j=1

which are called coefficients functional. Furthermore, we can define the

projections associated with the basis (xn )∞
n=1 as follow:
 
∞
X Xn
Sn : X −→ X, Sn  aj xj =
 aj xj .
j=1 j=1

J. Schauder (see [9]) introduced the concept of basis. The original

definition dealt with Banach spaces, and the continuity of the coefficient
functionals was required. However, this condition was proved unneces-
sary since, in Banach spaces, the coefficients functional associated with
a basis are always continuous. For topological spaces, the situation is
more delicate. The existence of a base does not guarantee the continuity
of the coefficient functionals. Thus, a sequence (xn )∞
n=1 is said to be a
∗
Schauder basis when each xn is continuous. Shauder basis was revealed as
 Bases and basic sequence criteria on TVS 3

a seminal concept in the theory of Banach spaces. Most classical separable

Banach spaces and their Schauder basis were investigated. Furthermore,
the existence of such a basis provides essential information about the linear
structure of the space and its closed subspaces. On the other hand, the
presumed existence of a Shauder basis for every separable Banach space
was a long-standing open problem, negatively resolved by P. Enflo in [4].
It was then proven that for p ̸= 2, 1 ≤ p < +∞, the space Lp [0, 1] had a
subspace without a basis.
Despite this, no criterion exists to determine whether a space has
a basis. On the other hand, one can investigate its subspaces and in-
quire whether, in every infinite-dimensional Banach space, there exists
an infinite-dimensional subspace with a Schauder basis. From this per-
spective, this problem is intrinsically related to basic sequences, as it is
equivalent to the following statement: every infinite-dimensional Banach
space contains an infinite basic sequence?. In 1958, C. Bessaga and A.
Pelczyński positively answered this question in [1].

Theorem 1.2. (Bessaga-Pelczyński) Every infinite-dimensional Banach

space contains an basic sequence.

In the more general setting of locally convex separable topological

vector spaces or F -spaces (a complete metric space with a translation-
invariant metric), some results from the theory of basis in Banach spaces
remain unchanged. One of them is that every basis is a Schauder basis,
meaning that the coefficient functionals associated are continuous. Fur-
thermore, if (xn )∞n=1 is a basis in a topological vector space, then the
Pn
set { i=1 ri xi : ri ∈ Q, n ∈ N} is dense in X, implying that X is sep-
arable. Since there are Banach spaces without a basis, it is natural to
consider the existence of non-normed spaces without a basis. Indeed,
for 0 < p < 1, Lp [0, 1] is an F -space when equipped with the metric
d(f, g) = ∥f − g∥p = [0,1] |f − g|p dµ. The space Lp [0, 1] is separable and
R

not normed. Certainly, Day’s Theorem (see [3]) is the most well-known re-
sult about these spaces, which states that L∗p [0, 1] = 0. Therefore, if there
4 N. G. Albuquerque, E. Demétrio Jr.

were a basis in the F -space Lp [0, 1], it would automatically be a Schauder

basis, implying the existence of a non-zero continuous linear functional.
Consequently, we conclude that Lp [0, 1] does not have a basis. More gen-
erally,

Theorem 1.3. if X is a separable F -space with X ∗ = {0}, then X has

no basis.

By this, we observe that basic sequences are inherent in every Banach

space. Given the significance of the theory of basic sequences, a natural
question arises: under what conditions a sequence in a topological vector
space is a basis or a basic sequence? Some conditions for a sequence to
constitute a basis or a basic sequence are well-established, for instance,
on Banach spaces this is known as the Banach-Grumblum criterion. The
following is a variant of this result on Banach sapces, with a comprehensive
proof available in [2].

Theorem 1.4. A sequence (xn )∞ n=1 of non-zero vectors in a Banach space

X is basic if and only if, there exist a constant K ≥ 1 such that for every
sequence of scalars (an )∞
n=1 ,

m
X n
X
ai xi ≤ K ai xi , whenever n ≥ m. (1.1)
i=1 i=1

The main objective of this article is to investigate the Banach-Grumblum

criteria on specific classes of topological vector spaces that are not neces-
sarily endowed with a norm. More precisely, we investigate criteria for
basic sequences in both locally convex spaces and F -spaces. In 1966, J.
R. Retheford and C. W. McArthur initially characterized basic sequences
in locally convex spaces (see, e.g., [11]). In the context of F -spaces, which
may not necessarily be locally convex, P.K. Kamthan and M. Gupta es-
tablished an analogous version of Theorem 2.6, and we present this variant
in the subsequent section.
 Bases and basic sequence criteria on TVS 5

2 Notation and main results

We include basic results and terminology, along with the statement of
the main results. We will provide proofs for a few results presented here.
For more details on the topics covered in this section, we suggest references
[7], [6], [10] and [8] .

Definition 2.1. A topological vector space X is said to be locally convex

when the origin admits a (topologial) basis of convex neighborhoods.

The following result brings us a characterization of these spaces.

Theorem 2.2. A topological vector space X is locally convex if and only

if its topology is generated by a family of seminorms.

Definition 2.3. A barrel in locally convex space X is a subset convex,

balanced, absorvent and closed. A locally convex space X is barrelled if
each barrel in X is a neighborhoods of the origin.

Barrelled spaces are an important class of locally convex spaces. We

will now focus on the Banach-Steinhaus Theorem; however, before that,
we need the following definition.

Definition 2.4. Let X, Y be topological vector spaces, and F a family

of linear mappings from X to Y . We say that F is equicontinuous if, for
every neighborhood of the origin W in Y , there exists a neighborhood of
the origin U in X such that f (U ) ⊆ Y for every f ∈ F. We also say that
F is pointwise bounded when, for every x ∈ X, the set

F(x) = {f (x) : f ∈ F}

is bounded in Y .

Theorem 2.5. (Banach-Steinhaus). Let X and Y be barrelled spaces, and

F a family of continuous linear mappings in L(X, Y ). If F is pointwise
bounded, then F is equicontinuous.
6 N. G. Albuquerque, E. Demétrio Jr.

Definition 2.6. A sequence (xn )∞ in a topological vector space X is

Pn=1
∞
said to be ω−independent when i=1 ai xi = 0 implies that ai = 0 for all
i ∈ N.

Theorem 2.7. (see [11]) Let (X, τ ) be a complete topological vector space.
Consider (xn )∞
n=1 as an ω-independent sequence and
∞
( )
X
XC = x ∈ X : x = an xn .
n=1

We will denote by V the collection of neighborhoods of the balanced and

closed origin in X. Define the sets V ∗ = {V ∗ : V ∈ V} where V ∗ =
{x ∈ XC : ni=1 ai xi ∈ V for all n ∈ N where x = ∞
P P
n=1 an xn }. Then,

1. (XC , τ ∗ ) is a complete topological vector space, where τ ∗ has V ∗ as

a neighborhood base of the origin.

2. If X is a locally convex space and V is the collection of all closed,

convex, and balanced neighborhoods of the origin, then (XC , τ ∗ ) is a
complete locally convex space.

3. Let pV be the Minkowski functional for each V ∈ V, and pV ∗ be the

Minkowski functional of V ∗ . If x = ∞
P
i=1 ai xi
n
!
X
pV ∗ (x) = sup pV ai xi .
n∈N i=1

The topology of a Banach space is generated by a single norm, which,

in a way, implies the independence of the constant K in Theorem 1.4. On
the other hand, the topology of a locally convex space is generated by a
family of seminorms. Thus, we will replace the norm in inequality 1.1 with
seminorms, indirectly resulting in the dependence of the constant K on the
seminorm. Here, we introduce an initial version of the Banach-Grumblum
criterion tailored for complete locally convex spaces.

Theorem 2.8. Let (X, τ ) be a complete locally convex space, and (xn )∞
n=1
be a non-zero sequence in X, and X = span{xn : n ∈ N}. If Γ is the
 Bases and basic sequence criteria on TVS 7

collection of all continuous seminorms generating the topology τ , and for

each ρ ∈ Γ there exist σ ∈ Γ and K > 0 such that
n m
! !
X X
ρ ai xi ≤ Kσ ai xi , (2.1)
i=1 i=1

for any m ≥ n ∈ N and scalars a1 , a2 , . . . am . Then (xn )∞n=1 is a basis in

X.

Proof. Let XC = {x ∈ X : x = ∞
P P∞
i=1 ai xi }. If i=1 ai xi = 0, then for
each fixed n ∈ N and ρ ∈ Γ, there exists σ ∈ Γ and K > 0 such that
j
!
X
ρ(an xn ) ≤ Kσ ai xi for all j ≥ n,
i=1

letting j −→ ∞ implies that ρ(an xn ) = 0 for all ρ ∈ Γ. Since X is a

Hausdorff space, it follows that an xn = 0, and thus an = 0 for every
n ∈ N, showing that the sequence (xn )∞ n=1 is ω-independent.
For each ρ ∈ Γ, let Vρ = {x ∈ X : ρ(x) ≤ 1}. As X is locally convex,
we have that ρ is the Minkowski functional of Vρ , i.e., ρ = pVρ . Let V be the
family of all positive scalar multiplications by sets of the form Vρ , ρ ∈ Γ:

V = {εVρ : ε > 0 and ρ ∈ Γ}.

V is a neighborhood base of the origin in the topology τ . Therefore, from

Theorem 2.7, we conclude that (XC , τ ∗ ) is a complete locally convex space,
where V ∗ is a neighborhood base of the origin in τ ∗ . We will now show
that XC is closed in (X, τ ).
Let (yα )α∈A be a net in XC converging to x ∈ X. For each α ∈ A,
we have yα = ∞ α
P
i=1 ai xi . From Theorem 2.7 and inequality 2.1, it follows
that for each ρ ∈ Γ, there exist σ ∈ Γ and k > 0 such that
n
!
X
pVρ∗ (yα ) = sup pVρ aαi xi ≤ kσ(yα ).
n∈N i=1

The sequence (yα )α∈A is Cauchy in (XC , τ ∗ ), so there exists x0 ∈

XC such that yα −→ x0 in τ ∗ . As V ∗ ⊆ V for every V ∈ V, i.e., the
8 N. G. Albuquerque, E. Demétrio Jr.

topology τ ∗ is stronger than the topology τ on XC . Therefore, yα −→ x0

in the topology τ , and as X is a Hausdorff space, we have x = x0 ∈ XC .
Thus, XC is closed in (X, τ ), and as span{xn : n ∈ N} ⊆ XC , we have
X = span{xn : n ∈ N} ⊆ XC , implying X = XC . Since (xn )∞ n=1 is an
P∞
ω-independent sequence, the representation x = i=1 ai xi is unique, and
therefore (xn )∞
n=1 is a basis in X.

This result remains somewhat incomplete, as we have only established

one aspect of the criterion. Consequently, if (xn )∞
n=1 constitutes a basis in
a complete locally convex space, the inequality (2.1) may not necessarily
hold. In order to provide a comprehensive theorem, we must extend our
consideration to barrelled spaces.

Theorem 2.9. Let (X, τ ) is barrelled space complete and (xn )∞ n=1 be a
sequence in X such that xn ̸= 0 for all n ∈ N and X = span{xn : n ∈ N}.
Let Γ be the collection of all continuous seminorms that generate the topol-
ogy τ . Then (xn )∞n=1 is a base in X if, and only if, for every ρ ∈ Γ there
are σ ∈ Γ and K = K(ρ) > 0 such that
n m
! !
X X
ρ ai xi ≤ Kσ ai xi , (2.2)
i=1 i=1

for any m ≥ n ∈ N and all scalars a1 , a2 , · · · , am .

The proof of this initial criterion for determining a basis will be pre-
sented in the last section, where we will use the Banach-Steinhaus Theorem
in barrelled spaces to demonstrate the equicontinuity of the projections as-
sociated with a basis in a locally convex space.
Now we will seek on the search for a criterion similar to Theorem 1.4 for
spaces that do not have local convexity. An F -space is a complete metriz-
able topological vector space with a metric that is translation-invariant.
If X is an F -space, there exists a metric d on X such that the function
∥x∥ = d(x, 0) is an F -norm, satisfying:

1. ∥x∥ ≥ 0 for all x ∈ X, and ∥x∥ = 0 ⇔ x = 0.

 Bases and basic sequence criteria on TVS 9

2. ∥αx∥ ≤ ∥x∥ for all x ∈ X and α ∈ K with |α| ≤ 1.

x
3. lim = 0 for each x ∈ X;
n→∞ n
4. ∥x + y∥ ≤ ∥x∥ + ∥y∥ for all x, y ∈ X.

5. The metric d(x, y) = ∥x − y∥ is complete.

We denote by (X, ∥ · ∥) the pair, where X is an F -space, and ∥ · ∥ is an F -

norm on X satisfying the above conditions. Two F -norms are considered
equivalent if the topologies generated by them are equivalent.

Theorem 2.10. (Open Mapping) Let X and Y be F -spaces, and T :

X −→ Y be a continuous linear bijection. Then, T is an isomorphism.

Theorem 2.11. Let (X, ∥·∥) be an F -space, and (xn )∞

n=1 be a basis. Then,

n
X
∥x∥1 = sup ∥Sn (x)∥ = sup an xj ,
n∈N n∈N j=1

P∞
where x = n=1 an xn , defines an F -norm on X equivalent to ∥ · ∥.

Proof. For each n ∈ N, we have

n
X
aj xj ≤ ∥x∥1 ,
j=1

hence,
∥x∥ ≤ ∥x∥1 for all x ∈ X. (2.3)

The conditions (1), (2) and (4) of F -norm follow below. We will verify
condition (3). For this, let x = ∞
P
j=1 aj xj ∈ X. Given ε > 0, there exists
N ∈ N such that
L
X ε
ak xk < ,
2
k=M

whenever L > M > N . In particular, we obtain

10 N. G. Albuquerque, E. Demétrio Jr.

∞
X ε
ak xk ≤ .
2
k=N +1 1
P∞
Defining y = k=N +1 ak xk , for each n ∈ N we have

∞
1 X ak
y= xk .
n n
k=N +1

Since ∥ · ∥1 satisfies condition (2), we have

∞ ∞
X ak X ε
xk ≤ ak xk ≤ ,
n 2
k=N +1 1 k=N +1 1

hence,
N
x X ak ε
< xk + , for each n ∈ N.
n 1 n 2
k=1 1
x
We know that limn→∞ n = 0, hence each M ≤ N and n ∈ N
sufficiently large, we have

M M
X ak X ε
xk = n−1 ak xk < .
n 2
k=1 k=1
and consequently

x
< ε, for n large enough.
n 1
Thus, ∥ · ∥1 is an F -norm on X. Now let’s see that X is complete.
Consider (un )∞n=1 as a Cauchy sequence in (X, ∥ · ∥1 ) by the inequality
(2.3). The sequence (un )∞n=1 is also a Cauchy sequence in (X, ∥ · ∥) and,
hence, converges; let’s denote ∥un − u∥ −→ 0 for some u ∈ X. Moreover,
for each k ∈ N, we have

∥Sk (un )∥ ≤ ∥un ∥ for all n ∈ N.

Thus, the sequence (Sk (un ))∞

n=1 is Cauchy. Restricting the coefficient func-
tionals xk to the finite-dimensional subspace Sk (X), each x∗k is continuous
∗
 Bases and basic sequence criteria on TVS 11

as it is linear. Therefore, (e∗k (Sk (un )))∞

n=1 is a Cauchy sequence in K and,
in particular, convergent. Let’s denote

lim x∗k (Sk (un )) = αk for each k ∈ N,

n→∞

and there also exists yk ∈ X such that

lim Sk (un ) = yk . (2.4)

n→∞

From this, it follows that if j ≤ k, then

 
x∗j (xk ) = x∗j lim Sk (un ) = lim x∗j (Sk (un )) = αj
n→∞ n→∞

and since x∗j (Sk (un )) = 0 when j > k, then

k
X
lim Sk (un ) = yk = αi xi .
n→∞
i=1

Combining this equality with inequality 2.4, we obtain

k
X
sup Sk (un ) − αi xi −→ 0 as n −→ ∞.
k∈N i=1

P ∞
k
converges in ∥·∥ to u = ∞
P
The sequence i=1 α i x i i=1 αi xi , with
k=1
yk = Sk (u), or in other words,

lim Sk (un ) = Sk (u)

n→∞

and finally we obtain,

k
X
sup Sk (un ) − αi xi = sup ∥Sk (un − u)∥ = ∥un − u∥1 −→ 0.
k∈N i=1 k∈N

Therefore, the sequence (un )∞

n=1 is convergent, hence (X, ∥ · ∥1 ) is an
F -space. Finally, note that the identity map id : (X, ∥ · ∥1 ) −→ (X, ∥ · ∥)
is continuous, as ∥x∥1 ≥ ∥x∥ for every x ∈ X. It follows from the Open
Mapping Theorem that id is an isomorphism, so the F -norms ∥ · ∥ and
∥ · ∥1 are equivalent on X.
12 N. G. Albuquerque, E. Demétrio Jr.

Theorem 2.11 is an important tool for studying bases in these spaces,

since the F −norm ∥ · ∥1 has the following property:

l
X ∞
X
aj xj ≤ ∥x∥1 for every x = aj xj ∈ X and 0 ≤ k ≤ l. (2.5)
j=k j=1
1

Next, we state a variant of the Banach-Steinhaus Theorem for F -

spaces.

Theorem 2.12. . Let X be an F -space, Y be a topological vector space,

and F ⊆ L(X, Y ) be a family of functions. If F is pointwise bounded, then
F is equicontinuous.

Proposition 2.13. Let X be an F -space. If (xn )∞ n=1 is a basis in X, then

∞
the family of partial sums (Sn )n=1 is equicontinuous.

Proof. Consider X equipped with the F -norm,

∥x∥1 = sup ∥Sn (x)∥.

n∈N

For each x ∈ X, the sequence (Sn (x))∞

n=1 is bounded, since by the inequal-
ity (2.5), we have

∥Sn (x)∥1 ≤ ∥x∥1 for every n ∈ N.

As X is an F -space, and each Sn is linear, it follows from the Banach-

Steinhaus Theorem that (Sn )∞ n=1 is an equicontinuous family of linear
operators. Since the F -norms are equivalent, it follows that the family
(Sn )∞
n=1 is equicontinuous with respect to the original F -norm.

Corollary 2.14. Every basis in an F -space is a Schauder basis.

In the context of F -spaces, P. K. Kamthan and M. Gupta proved an

analogous version of Theorema 1.4, which we present next.
 Bases and basic sequence criteria on TVS 13

Theorem 2.15. (Banach-Grumblum-Nikolskii Criteria). Let (X, ∥ · ∥) be

an F -space containing a nonzero sequence (xn )∞
n=1 with span{xn : n ∈ N} =
X, where ∥ · ∥ is the F -norm on X. If there is a constant K ≥ 1 such that
m
X n
X
ai xi ≤ K ai xi whenever n ≥ m, (2.6)
i=1 i=1

then (xn )∞
n=1 is an Schauder basis.

When ∥·∥ is a norm, we return to the familiar setting of Banach spaces.

Theorem 1.4 is commonly known as the Banach-Grumblum criterion in the
context of Banach spaces. On the other hand, in the case of F -spaces, the
mathematician Nikolskii is also credited for significant contributions.
Finally, we will present our latest version of the criterion, on quasi-
normed spaces (see Theorem 2.20). Before that, let us recall some basic
facts.

Definition 2.16. Let X be a vector space. A quasi-norm on X is a

function ∥ · ∥ : X −→ R that satisfies:

1. ∥x∥ ≥ 0 for all x ∈ X, and ∥x∥ = 0 ⇔ x = 0;

2. ∥αx∥ = ∥α∥∥x∥ for all x ∈ X and α ∈ K;

3. There exists K ≥ 1 such that ∥x + y∥ ≤ K (∥x∥ + ∥y∥) for all x, y ∈

X.

In this case, we say that the pair (X, ∥ · ∥) is a quasi-normed space. If

there is no ambiguity about the quasi-norm in the space, we will simply
say that X is a quasi-normed space. When K = 1, we obtain a norm. The
constant K is often called the modulus of concavity. The quasi-norm is
called a p-norm (0 < p < 1) when it satisfies:

∥x + y∥p ≤ ∥x∥p + ∥y∥p , x, y ∈ X.

A quasi-norm on X defines a metrizable vector topology that has, as a

base of neighborhoods of the origin, the sets Un = {x ∈ X : ∥x∥ < 1/n}.
14 N. G. Albuquerque, E. Demétrio Jr.

When X is complete, we say that (X, ∥ · ∥) is a quasi-Banach space. If the

quasi-norm is a p-norm, we simply say that X is a p-Banach space. Every
quasi-norm on X induces a p-norm and therefore a translation-invariant
metric topology given by d(x, y) = ∥x−y∥p , where ∥·∥ is the p-norm on X.
This result is a consequence of the Aoki-Rolewicz Theorem (see [6]), which
states that every quasi-normed space is p-normed for some 0 < p < 1.

Theorem 2.17. (Aoki-Rolewicz) Let (X, ∥ · ∥) be a quasi-normed space.

Then, there exists 0 < p < 1 and a p-norm ∥ · ∥1 on X, equivalent to the
quasi-norm, such that where C is a positive constant.

The p-norm on X is defined by


n
!1 n

 X p X 
∥x∥1 = inf ∥xi ∥p :x= xi .
 
i=1 i=1

To avoid unnecessary complications, when we write (X, ∥ · ∥), we con-

sider the quasi-norm ∥ · ∥ as the p-norm that induces the topology on X
through the metric d(x, y) = ∥x − y∥p .

Theorem 2.18. Let X, Y be quasi-normed spaces. For each T ∈ L(X, Y ),

we define
∥T (x)∥
∥T ∥ = sup .
x̸=0 ∥x∥

Then, (L(X, Y ), ∥·∥) is a quasi-normed space. Moreover, if Y is quasi-

Banach, then L(X, Y ) is also quasi-Banach.

Proposition 2.19. Let (xn )∞n=1 be a Schauder basis in the quasi-Banach

∞
space X, and (Sn )n=1 be the canonical projections relative to this basis.
Then supn∈N ∥Sn ∥ < ∞.

The number supn∈N ∥Sn ∥ < ∞ is called the basis constant of the se-
quence (xn )∞
n=1 and denoted by K(xn )∞
n=1
. Next we provide the Banach-
Grumblum-Nikolskii Criteria on quasi-Banach spaces.
 Bases and basic sequence criteria on TVS 15

Theorem 2.20. A sequence (xn )∞ n=1 of non-zero vectors in a quasi-Banach

space X is a Schauder basis if and only if there exists a constant K ≥ 1
such that for every sequence of scalars (an )∞
n=1 ,

m
X n
X
ai xi ≤ K ai xi whenever n ≥ m.
i=1 i=1

3 Proof of the main results

Proof of Theorem 2.9. It is sufficient to verify that if the sequence (xn )∞
n=1
is a basis, then inequality (2.2) is satisfied. Consider the collection

V = {εVρ : ε > 0 and ρ ∈ Γ},

where Vρ = {x ∈ X : ρ(x) ≤ 1} forms a base of neighborhoods of the

origin in X. Using the same notation as in Theorem 2.7, we will verify
that each set (εVρ )∗ is a barrel in X. Let’s initially see that (Vρ )∗ is
balanced and convex. Let x = ∞ ∗
P
j=1 ai xi ∈ (Vρ ) and α ∈ K with |α| ≤ 1.
As Vρ is balanced since ρ is a seminorm, it follows that α nj=1 aj xj ∈
P

Vρ for each n ∈ N, thus αx ∈ (Vρ )∗ .

Convexity of (Vρ )∗ follows identically. Thus, (εVρ )∗ is balanced and
convex for each ε > 0 and ρ ∈ Γ. By item (1) of Theorem 2.7, the sets
(εVρ )∗ form a neighborhood base of the origin for a topology in X, so they
are absorbing.
We will now verify that (εVρ )∗ is closed. Initially, consider the partial
P∞
sum functions Sn : X −→ X. For each x = i=1 ai xi and ρ ∈ Γ, the
Pn ∞
sequence ρ ( i=1 ai xi )n=1 is bounded. Thus, if Vρ is a neighborhood of
the origin, taking ε = supn∈N ρ ( ni=1 ai xi ), we have
P

( n )
X
{Sn (x) : n ∈ N} = ai xi : n ∈ N ⊆ εVρ .
i=1

Thus, the family F = {Sn }∞

n=1 is pointwise bounded, and by the Banach-
Steinhaus Theorem, F is equicontinuous. In particular, each function Sn
16 N. G. Albuquerque, E. Demétrio Jr.

is continuous. Therefore, ρ ◦ Sn : X −→ R is a continuous seminorm for

each n ∈ N, so (εVρ ) is closed because
∞ n
( )
X X
(εVρ )∗ = x= ai xi ∈ X : ai xi ∈ εVρ for every n ∈ N
i=1 i=1
∞
\
= (ρ ◦ Sn )−1 (−∞, ε].
n=1

So, for each ε > 0 and ρ ∈ Γ, the set (εVρ )∗ is a barrel in X and thus a
τ -neighborhood of the origin since X is barrelled. If p ∈ Γ, as (Vρ )∗ is a
neighborhood of the origin, there exist ε = ε(ρ) > 0 and σ ∈ Γ such that

Sn (εVσ ) ⊆ (Vρ )∗ for every n ∈ N.

Let m be such that n ≤ m and a1 , · · · , am ∈ K. Considering the vector

m
εa x
P i i
X
x= ,
σ( mi=1 ai xi )
i=1

then σ(x) ≤ ε, so x ∈ εVσ . As Sm (εVσ ) ⊆ (Vρ )∗ , we obtain

n
!
X εai xi
ρ ≤ 1 for every n ≤ m,
σ( m
P
i=1 i=1 ai xi )

and finally, we have

n m
! !
X X
ρ ai xi ≤ Kσ ai xi ,
i=1 i=1

with
1
K = K(ρ) = > 0.
ε

Proof of Theorem 2.15. Consider Y = span{xn : n ∈ N}, and for each

n ∈ N define the function,
k n
!
X X
Tn : Y −→ X, Tn ai xi = ai xi ,
i=1 i=1
 Bases and basic sequence criteria on TVS 17

putting, if necessary, ak+1 = · · · an = 0. By (2.6), we have

k n k
! !
X X X
Tn ai xi = ai xi ≤ K ai xi ,
i=1 i=1 i=1

for all x = ki=1 ai xi ∈ Y . Therefore, the linear operator Tn is continuous.

P

As Y = X is an F -space, each Tn admits a linear and continuous

extension Sn : X −→ X. If x ∈ X, then there exists a sequence (yk )∞ k=1 in
Pnk k
Y with yk −→ x, where yk = i=1 ai xi . It follows that limk→∞ Sn (yk ) =
Sn (x) for each n ∈ N. Again, by inequality (2.6), we obtain

∥Sn (yk )∥ ≤ K∥yk ∥.

Moreover,

∥Sn (x)∥ ≤ K∥x∥, for every n ∈ N and x ∈ X. (3.1)

Pm
Arbitrarily choosing x ∈ X and ε > 0, there exists y = i=1 ai xi ∈ Y
such that ∥x − y∥ < ε. Thus, for every n ≥ m, we have

∥Sn (x) − x∥ ≤ ∥Sn (x) − Sn (y)∥ + ∥x − y∥ + ∥Sn (y) − y∥

< (K + 1)ε,

therefore,
lim Sn (x) = x.
n→∞
Finally, let’s define fi : X −→ K by fi (x)xi = (Si − Si−1 )(x). As each
Pn is continuous, fi is a continuous linear functional. Furthermore,
n
X n
X
fi (x)xi = (Sn − Sn−1 )(x) = Sn (x).
i=1 i=1

Hence,
n
X ∞
X
x = lim Sn (x) = lim fi (x)xi = fi (x)xi .
n→∞ n→∞
i=1 i=1
Through inequality 2.6 we obtain the uniqueness of the representation.
18 N. G. Albuquerque, E. Demétrio Jr.

Proof of theorem 2.20. Being X an quasi-Banach, then X is complete with

respect to the translation-invariant metric induced by the p-norm d(x, y) =
∥x − y∥p . In particular, X is an F -space. Thus, if inequality 2.20 holds,
then by Theorem 2.15, the sequence (xn )∞ n=1 is basic.
For the second part, suppose that the sequence (xn )∞
n=1 is basic, and
∞
let (an )n=1 be a sequence of scalars. If n ≥ m, then
m n m m
!
X X X X
ai xi = Sm ai xi ≤ ∥Sn ∥ ai xi = K(xn )∞
n=1
ai xi .
i=1 i=1 i=1 i=1

References
[1] C. Bessaga, A. Pelczyński, On basis an unconditional conver-
gence of series in Banach spaces, Studia Math, 17 (1958), 151-164.

[2] B. Beauzamy, Introduction to Banach Spaces And Their Geometry,

68, North-Hollanda, (1982).

[3] M. Day, The spaces Lp with 0 < p < 1, Math. Soc. 46, (1940),
816-823.

[4] P. Enflo, A counter-example to the approximation problem in Ba-

nach spaces, Acta Math, 130 (1973), 309-317.

[5] P. Kamthan, M. Gupta, Theory of bases and cones, Pitman Pub-

lishing, (1985).

[6] N. P. Kalton, J. Roberts, An F-space sampler, Cambridge Uni-

versity Press, (1984) .

[7] G. Rano, T. Bag, Bounded linear operators in quasi-normed linear

spaces, Journal of the Egyptian Mathematical Socyety, 23 (2015),
303-308.
 Bases and basic sequence criteria on TVS 19

[8] H. H. Schaefer, Topological vector spaces, Graduate texts in math-

ematics, Springer, (1971).

[9] J. Schauder, Zur Theorie stetiger Abbildungen in Funktionalrau-

men, Mathematische Zeitschrift 26 (1927), 47-65.

[10] J. Shapiro, Extension of linear functionals on F -spaces with bases,

Duke Math. 37 (1970), 639-645.

[11] J. R. Retheford, C. W. Mcarthur, Some remarks on bases in

linear topological spaces, Mathematische Annalen, 164, (1966), 38-41.