To appear in Topologı́a y sus aplicaciones 7 (2019)

CHAPTER 1

Paracompactness of a normed space with the weak

topology

Salvador Sánchez-Perales1 , Slaviša V. Djordjević2

Sonia Venancio1
1
Universidad Tecnológica de la Mixteca, Huajuapan de León, Oax. Mexico
2
Benemérita Universidad Autónoma de Puebla, Puebla Pue. Mexico

1. Introduction
Topological properties of a normed space X in the weak topology are of the
importance and have been intensively studied from many years (see [2], [3], [4], [5],
[7]). Corson, [2], was the first to develop a systematic study of certain topological
properties of the weak topology on Banach spaces. One of Corson’s results shows
that X is paracompact with the weak topology if and only if X is Lindelöf with the
weak topology. In this chapter we observed that if X is a reflexive normed space or
a separable space under the strong topology, then X is paracompact with the weak
topology.

2. Preliminaries
Throughout this chapter we assume that X is a normed space over the field F,
where F is the set of the real numbers R or the set of the complex numbers C.
2.1. Bounded linear functionals. A linear functional f : X → F is bounded,
if there exists a constant M ≥ 0 such that
|f (x)| ≤ M kxk
for all x ∈ X. The set of all bounded linear functionals from X into F is denoted
by B(X, F). This set with the usual operations of addition and multiplication by a
scalar is a vector space over F. Moreover B(X, F) is a normed space with the norm
kf k = inf{M ≥ 0 : |f (x)| ≤ M kxk for all x ∈ X}.
It is not difficult to prove that
|f (x)| ≤ kf kkxk
for all x ∈ X. With this inequality equivalent expressions of kf k are obtained:
|f (x)|
kf k = sup ; kf k = sup |f (x)|; kf k = sup |f (x)|.
x6=0 kxk kxk≤1 kxk=1

The next basic properties establish a relation between the concepts of bounded
linear functional and continuous linear functional.
1
2 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

Proposition 2.1. Let f : X → F be a linear functional. The following state-

ments are equivalent:
(1) f is continuous.
(2) f is continuous at some x0 ∈ X.
(3) There exist x0 ∈ X and r > 0 such that f (B(x0 , r)) is bounded.
(4) f is bounded.
The dual space of X is the space B(X, F) which is denoted by X ∗ . The bidual
of X is the space X ∗∗ = (X ∗ )∗ . Next we stablish well known results about the
space X ∗ which are consequence of the Hahn-Banach theorem. See [6]. For the
sake of completeness, we give its proofs.
Proposition 2.2. Let M be a proper closed linear subspace of X. If x0 ∈ X \M
then there exists f ∈ X ∗ such that f (x0 ) = 1, f (M ) = {0} and kf k = d(x01,M ) .

Proof: Let Y = gen({x0 }) + M and define g : Y → F as g(λx0 + m) = λ,

where λ ∈ F and m ∈ M . It is clear that g is a linear functional, g(x0 ) = 1 and
g(M ) = {0}. Observe that
1 1 1 1
= = = sup −1 mk
d(x0 , M ) inf kx0 − mk inf kx0 + mk λ6=0 kx0 + λ
m∈M m∈M m∈M

|λ| |g(λx0 + m)|

= sup = sup .
λ6=0 kλx0 + mk λ6=0 kλx0 + mk
m∈M m∈M

1
Therefore, g is bounded and kgk = d(x0 ,M ) .
Now, by the Hahn-Banach theo-
rem, there exists a bounded linear functional f : X → F such that f |Y = g and
kf k = kgk, so f (x0 ) = 1, f (M ) = {0} and kf k = d(x01,M ) . 

Theorem 2.3. The space X ∗ separates points of X, i.e. if x, y ∈ Y with x 6= y

then there exists f ∈ X ∗ such that f (x) 6= f (y).
x−y
Proof: Let x, y ∈ X with x 6= y. We set z = kx−yk , then by Proposition 2.2,
∗
there exists f ∈ X such that f (z) = 1. This implies that
f (x) − f (y) = f (x − y) = kx − yk =
6 0.


Theorem 2.4. If X ∗ is separable then X is separable.

Proof: Suppose that F = C. Let (ϕn )n∈N be a dense subset in X ∗ and ε ∈ (0, 31 ).
Since kϕn k − ε < sup |ϕn (x)| for all n ∈ N, it follows that for each n ∈ N there
kxk=1
exists xn ∈ X with kxn k = 1 such that
(1) |ϕn (xn )| > kϕn k − ε.
 n

P
We set D = (qk + isk )xk : qk , sk ∈ Q, n ∈ N . Then D is a closed linear sub-
k=1
space of X. Suppose that D 6= X. By Proposition 2.2 there exists f ∈ X ∗ such
that f (D) = {0} and kf k = 1. Let m ∈ N be such that kϕm − f k < ε. Note that
|ϕm xm | = |(f − ϕm )xm | ≤ kf − ϕm kkxm k < ε.
 2. PRELIMINARIES 3

Thus by (1), kϕm k < 2ε. This implies that

1 − 2ε < kf k − kϕm k ≤ kf − ϕm k < ε
1
and so 3 < ε which is a contradiction. Therefore D = X. 

Definition 2.5. For each x ∈ X let Jx : X ∗ → F be defined by

(2) Jx (f ) = f (x).
∗∗
The linear map J : X → X , defined by
(3) J(x) = Jx ,
is called the natural embedding of X into its bidual. The space X is called reflexive
if J is surjective.
Theorem 2.6. The natural embedding J is an isometric isomorphism1 of X
onto R(J) ⊆ X ∗∗ .
Proof: It is clear that J is a linear transformation. Let x ∈ X \ {0}, then
kJ(x)k = kJx k = sup |Jx (f )| = sup |f (x)| ≤ kxk.
kf k≤1 kf k≤1

By Proposition 2.2, there exists f ∈ X such that f 0 ( kxk

0 ∗ x
) = 1 and kf 0 k =
1 0 0
d( kxkx
,{0}) . This implies that f (x) = kxk and kf k = 1. Thus

sup |f (x)| ≥ |f 0 (x)| = kxk.

kf k≤1

Therefore kJ(x)k = kxk. 

Theorem 2.7 (The Riesz Representation Theorem). Let H be a Hilbert space

over the field F. If f ∈ H ∗ (resp. g : H → F is a bounded conjugate-linear
functional2) then there exists a unique vector zf ∈ H (resp. wg ∈ H) such that
f (x) = hx, zf i (resp. g(x) = hwg , xi)
for all x ∈ H.
Proof: Let g : X → F be a conjugate-linear functional. Without loss of generality
we may assume that g 6= 0. Then N (g)⊥ 6= {0}. Let z ∈ N (g)⊥ \ {0}, since
N (g) ∩ N (g)⊥ = {0} it follows that g(z) = 0. Now take an arbitrary x ∈ H , then
!
g(x) g(x)
g x− z = g(x) − g(z) = 0.
g(z) g(z)
g(x)
Thus x − g(z) z∈ N (g), therefore
* + * +
g(x) g(x) g(x)
0= z, x − z = hz, xi − z, z = hz, xi − hz, zi .
g(z) g(z) g(z)

1Let X, Y be normed spaces over the field F. A linear transformation T : X → Y is an

isometric isomorphism if T is surjective and
kT xk = kxk
for all x ∈ X.
2g : X → F is a conjugate-linear functional, if g(αx + y) = αg(x) + g(y) for all α ∈ F and
x, y ∈ X.
4 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

This implies that  

g(z)
g(x) = z, x .
kzk2


2.2. Paracompactness.
Definition 2.8. Let E be a topological space and let V, U be families of
subsets of E.
(1) We say that V is a refinement of U, if V is a covering of E and for each
V ∈ V there exists U (V ) ∈ U such that V ⊆ U (V ).
(2) V is locally finite, if for each x ∈ E there exists an open neighborhood Vx
of x that intersects only finitely many elements of V.
Definition 2.9. A topological space E is paracompact if every open covering
of E has a locally finite open refinement.
Lemma 2.10 (Michael, [8]). Let E be a regular space. Then E is paracompact
if and only if each open covering of E has a locally finite refinement.
Proof: Assume that every open covering of E has a locally finite refinement. Let
U be an open covering of E. Choose A to be a locally finite refinement of U. For
each x ∈ E let Vx be an open neighborhood of x that intersects only finitely many
elements of A. By regularity of E, for each x ∈ E and y ∈ Vx , there exists an open
set Gxy such that y ∈ Gxy ⊆ Gxy ⊆ Vx . Then {Gxy : x ∈ E, y ∈ Vx } is an open
covering of E. Again by hypothesis, there exists W a locally finite refinement of
{Gxy : x ∈ E, y ∈ Vx }
For every A ∈ A, let S(A) = X \ ∪{F : F ∈ ClW and F ∩ A = ∅}. We show
that S = {S(A) : A ∈ A} is a locally finite open covering of E. Since W is locally
finite, it follows that ClW is locally finite, so each S(A) is an open set. It is clear
that S is a covering of E, because A ⊆ S(A) for all A ∈ A and this is a covering
of E. Observe that if W ∈ W then there exist x ∈ E and y ∈ Vx such that
W ⊆ Gxy , which implies that W ⊆ Gxy ⊆ Vx . Therefore each element of ClW
intersects only finitely many elements of A. Now, since F ∩ A 6= ∅ if and only if
F ∩ S(A) 6= ∅, for all F ∈ ClW and A ∈ A, it follows that each element of ClW
intersects only finitely many elements of S. Let x ∈ E then there exists V an open
neighborhood of x such that {F ∈ ClW : F ∩ V 6= ∅} is a finite family. We set
n
{F1 , . . . , Fn } = {F ∈ ClW : F ∩ V 6= ∅}. Then ∪ Fi intersects only finitely many
i=1
n
elements of S. Therefore, since V ⊆ ∪ Fi , we have that V intersects only finitely
i=1
many elements of S. Thus S is locally finite.
Finally, for each A ∈ A we take U (A) ∈ U such that A ⊆ U (A). Then
{U (A) ∩ S(A) : A ∈ A} is an open locally finite refinement of U. 

Lemma 2.11. Every countable open covering of a topological space E has a

locally finite refinement.
Proof: Let U = {Un : n ∈ N} be an open covering of E. We set A1 = U1 and for
n−1
each n > 1, An = Un \ ∪ Ui . Then {An : n ∈ N} is a locally finite refinement of
i=1
U. Indeed, let x ∈ E then there exists m ∈ N such that x ∈ Um . This implies that
 3. REGULARITY AND HAUSDORF AXIOMS FOR THE WEAK TOPOLOGY 5

 
n−1
{n ∈ N : x ∈ Un } 6= ∅ and for each n > m, Um ∩ An = Um ∩ Un \ ∪ Ui = ∅.
i=1
Thus Um intersects only finitely many elements of {An : n ∈ N} and x ∈ An0 where
n0 = min{n ∈ N : x ∈ Un }. Finally, it is clear that An ⊆ Un for all n ∈ N. 

Definition 2.12. A topological space E is Lindelöf if every open covering of

E has a countable subcovering.
Theorem 2.13. If E is a regular Lindelöf space then E is paracompact.
Proof: Let U be an open covering of E. Since E is Lindelöf, there exists V
a countable subcovering of U. By Lemma 2.11, there exists W a locally finite
refinement of V. Therefore W is a locally finite refinement of U. Thus by Lemma
2.10, E is paracompact. 

3. Regularity and Hausdorf axioms for the weak topology

Let us begin this section remembering that the strong topology of X, denoted
by τN (X), is the topology induced by the norm of X. This topology is generated
by the basis {B(x, ε) : x ∈ X and ε > 0}.
Definition 3.1. The weak topology of X, denoted by σ(X, X ∗ ), is the smallest
topology that makes all bounded linear functionals x∗ : X → F continuous.
From Proposition 2.1 and definition above we have that
σ(X, X ∗ ) ⊆ τN (X).
The weak topology is generated by the basis
( n )
\ −1
∗ ∗ ∗
B= xi (Ui ) : xi ∈ X , Ui is open in F, n ∈ N .
i=1
For each x0 ∈ X, ε > 0 and x∗1 , x∗2 , . . . , x∗n ∈ X ∗ , we set
Vε;x∗1 ,x∗2 ,...,x∗n (x0 ) = {x ∈ X : |x∗i (x − x0 )| < ε, for all i = 1, . . . , n}.
Remark 3.2. Let x0 ∈ X. For given ε > 0 and x∗1 , x∗2 , . . . , x∗n ∈ X ∗ , it follows
that
n
\
Vε;x∗1 ,x∗2 ,...,x∗n (x0 ) = (x∗i )−1 (B(x∗i (x0 ), ε)) .
i=1
Thus Vε;x∗1 ,x∗2 ,...,x∗n (x0 ) is a basic element of σ(X, X ∗ ).
Proposition 3.3. If x0 ∈ X then the family
A = {Vε;x∗1 ,x∗2 ,...,x∗n (x0 ) : ε > 0, x∗1 , x∗2 , . . . , x∗n ∈ X ∗ , n ∈ N}
form a basis of neighborhoods of x0 for σ(X, X ∗ ).
Proof: Let U be a neighborhood of x0 for σ(X, X ∗ ). There exits a basic element
\n
W in σ(X, X ∗ ) such that x0 ∈ W and W ⊆ U . Express W as W = (x∗i )−1 (Ui ),
i=1
where x∗1 , . . . , x∗n ∈ X ∗ and U1 , . . . , Un are open sets in F. For each i = 1, . . . , n,
there exists εi > 0 such that B(x∗i (x0 ), εi ) ⊆ Ui . We set ε = min{εi : i = 1, . . . , n},
then
n
\ n
\
(x∗i )−1 (B(x∗i (x0 ), ε)) ⊆ (x∗i )−1 (Ui ) = W.
i=1 i=1
6 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

Thus,
n
\
Vε;x∗1 ,x∗2 ,...,x∗n (x0 ) = (x∗i )−1 (B(x∗i (x0 ), ε)) ⊆ W ⊆ U.
i=1

Theorem 3.4. The topology σ(X ∗ , X) is Hausdorff.

Proof: Let x1 , x2 ∈ X be such that x1 6= x2 . Since X ∗ separates the points of X,
see Theorem 2.3, there exists y ∗ ∈ X ∗ such that y ∗ (x1 − x2 ) 6= 0. Consider ε > 0
∗
such that 0 < ε < |y (x12−x2 )| . We claim that Vε,y∗ (x1 ) ∩ Vε,y∗ (x2 ) = ∅. Assume to
the contrary that there exists x0 ∈ Vε,y∗ (x1 ) ∩ Vε,y∗ (x2 ). This implies that
|y ∗ (x2 − x1 )| = |y ∗ ((x0 − x1 ) − (x0 − x2 ))|
≤ |y ∗ (x0 − x1 )| + |y ∗ (x0 − x2 )|
< 2ε
< |y ∗ (x1 − x2 )|,
which is a contradiction. 
σ(X,X ∗ )
In the following theorem we use the notation D to represent the closure
of a set D with respect to the weak topology of X.
Theorem 3.5. The topology σ(X, X ∗ ) is regular.
Proof: Let x0 ∈ X and V be a neighborhood of x0 for σ(X, X ∗ ). There exist ε > 0
σ(X,X ∗ )
and x∗1 , . . . , x∗n ∈ X ∗ such that Vε,x∗1 ,...,x∗n (x0 ) ⊆ V . Let x ∈ V 2ε ,x∗1 ,...,x∗n (x0 ) .
Then V 2ε ,x∗1 ,...,x∗n (x) ∩ V 2ε ,x∗1 ,...,x∗n (x0 ) 6= ∅. Take s ∈ V 2ε ,x∗1 ,...,x∗n (x) ∩ V 2ε ,x∗1 ,...,x∗n (x0 ),
so for each i = 1 . . . , n,
ε
|x∗i (s − x)| <
2
and
ε
|x∗i (s − x0 )| < .
2
Therefore |x∗i (x − x0 )| ≤ |x∗i (x − s)| + |x∗i (s − x0 )| < 2ε + 2ε = ε, for all i = 1, . . . , n.
This implies that x ∈ Vε,x∗1 ,...,x∗n (x0 ). Consequently,
σ(X,X ∗ )
V 2ε ,x∗1 ,...,x∗n (x0 ) ⊆ V.


4. Compactness of a closed ball under the weak topology

It is known that when X is of infinite dimension, a closed ball in X is not a
compact subset with strong topology. This section shows that a closed ball in X is
always compact with the weak topology.
Now, we introduce a topology for the dual space X ∗ of X.
Definition 4.1. The ?−weak topology of X ∗ , denoted by σ(X ∗ , X), is the
smallest topology that makes all functional {Jx }x∈X continuous, where each jx is
defined as in (2).
 4. COMPACTNESS OF A CLOSED BALL UNDER THE WEAK TOPOLOGY 7

Proposition 4.2. The ?−weak topology of X ∗ is contained in the weak topology

of X ∗ , i.e. σ(X ∗ , X) ⊆ σ(X ∗ , X ∗∗ ). Moreover, if X is reflexive then
σ(X ∗ , X) = σ(X ∗ , X ∗∗ ).
Proof: Since the weak topology of X ∗ makes continuous all elements of X ∗∗ , it
follows that Jx is continuous with this topology for all x ∈ X. Thus, σ(X ∗ , X) ⊆
σ(X ∗ , X ∗∗ ).
Now, suppose that X is reflexive. Take f ∈ X ∗∗ , then by the surjectivity
of the natural embedding J of X into its bidual, there exists x ∈ X such that
Jx = J(x) = f . Therefore f is continuous with the ?−weak topology. Consequently,
σ(X ∗ , X ∗∗ ) ⊆ σ(X ∗ , X). 
The space FX is the collection of all functions from X into the scalar field F.
We denote by η the product topology of FX . That is, η is the smallest topology
that makes the projections {πx }x∈X continuous, where each πx : FX → F is defined
as
(4) πx (f ) = f (x).
Since X ∗ ⊆ FX we can equip X ∗ with the product topology η, as illustrated in
the following proposition.
Proposition 4.3. If ηX ∗ is the topology induced by η on X ∗ then
σ(X ∗ , X) = ηX ∗ .
Proof: We show first that ηX ∗ makes each Jx continuous. Let x ∈ X and D be
an open set of F. By equality (4) it follows that Jx = πx |X ∗ . Then
Jx−1 (D) = (πx |X ∗ )−1 (D) = πx−1 (D) ∩ X ∗
is open with respect to ηX ∗ . Therefore Jx is continuous with the topology ηX ∗ .
Now, let τ be another topology of X ∗ which makes each Jx continuous. We
claim that β = {U ∪ E : U ∈ τ and E ∈ P(FX \ X ∗ )} is a topology of FX that
makes the projections {πx }x∈X continuous. It is clear that ∅, FX ∈ β. Let {Aα }α∈I
be a family of sets in β. For each α ∈ I, there exist Uα ∈ τ and Eα ∈ P(FX \ X ∗ )
such that Aα = Uα ∪ Eα . Then,
∪ Aα = ∪ (Uα ∪ Eα )
α∈I α∈I
   
= ∪ Uα ∪ ∪ Eα .
α∈I α∈I

This implies that ∪ Aα ∈ β.

α∈I
Let {V1 , V2 , . . . , Vn } be a finite collection of sets in β. For every i = 1, . . . , n,
there exist Ui ∈ τ and Ei ∈ P(FX \ X ∗ ) such that Vi = Ui ∪ Ei . Then,
n n
∩ Vi = ∩ (Ui ∪ Ei )
i=1 i=1
 n   n 
= ∩ Ui ∪ ∩ Ei ,
i=1 i=1
n
thus ∩ Vi ∈ β. Therefore β is a topology of FX .
i=1
8 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

Take x ∈ X and G an open set in F. Since

πx−1 (G) = (πx−1 (G) ∩ X ∗ ) ∪ (πx−1 (G) ∩ (KX \ X ∗ ))
= (πx |X ∗ )−1 (G) ∪ (πx−1 (G) ∩ (KX \ X ∗ ))
= Jx−1 (G) ∪ (πx−1 (G) ∩ (KX \ X ∗ )),
Jx−1 (G) ∈ τ and πx−1 (G) ∩ (FX \ X ∗ ) ∈ P(FX \ X ∗ ), it follows that πx−1 (G) ∈ β,
which implies that πx is continuous with the topology β.
Consequently
η ⊆ β.
Let W ∈ ηX ∗ , then W = V ∩ X ∗ for some V ∈ η. There exist U ∈ τ and
E ∈ P (FX \ X ∗ ) such that V = U ∪ E. Therefore
W = V ∩ X∗ = (U ∪ E) ∩ X ∗
= (U ∩ X ∗ ) ∪ (E ∩ X ∗ )
= U ∈ τ.
Thus ηX ∗ ⊆ τ and hence ηX ∗ is the smallest topology that makes all functionals
Jx continuous. Consequently σ(X ∗ , X) = ηX ∗ . 

Theorem 4.4 (Alaoglu, [1]). The unit closed ball BX ∗ = {x∗ ∈ X ∗ : kx∗ k ≤ 1}
is a compact set with the ?−weak topology of X ∗ .
Proof: First note that BX ∗ ⊆ X ∗ ⊆ FX , so in order to prove that BX ∗ is
compact with the ?−weak topology, we show that BX ∗ is a compact set in FX with
the topology η.
Consider the set Y
I= B[0, kxk].
x∈X
For each x ∈ X, the closed unit ball B[0, kxk] is a compac set in F. Then by
Tychonoff’s theorem it follows that I is a compact set in FX with the topology η.
Observe that for every x∗ ∈ BX ∗ and each x ∈ X it follows that
|x∗ (x)| ≤ kx∗ kkxk ≤ kxk.
Thus
BX ∗ ⊆ I.
We prove that BX ∗ is a closed set in FX with the topology η. For each x, y ∈ X
and λ ∈ F, define the functions ϕx,y : FX → F and ϕx,λ : FX → F as
ϕx,y (ξ) = πx (ξ) + πy (ξ) − πx+y (ξ) and ϕx,λ (ξ) = λπx (ξ) − πλx (ξ).
Since every projection πz : FX → F, z ∈ X, is continuous with the topology η, it
follows that ϕx,y and ϕx,λ are continuous with respect to the topology η. Let ε ≥ 1.
We claim that
     
−1
(5) BX ∗ = ∩ ϕ−1x,y ({0}) ∩ ∩ ψx,λ ({0}) ∩ ∩ Jx −1
(B[0, ε]) .
x,y∈X x∈X,λ∈K kxk≤1
∗
Indeed, if x ∈ BX ∗ then
ϕx,y (x∗ ) = πx (x∗ ) + πy (x∗ ) − πx+y (x∗ ) = x∗ (x) + x∗ (y) − x∗ (x + y) = 0,
ϕx,λ (x∗ ) = λπx (x∗ ) − πλx (x∗ ) = λx∗ (x) − x∗ (λx) = 0
 4. COMPACTNESS OF A CLOSED BALL UNDER THE WEAK TOPOLOGY 9

and
|Jx(x∗ )| = |x∗ (x)| ≤ sup |x∗ (x)| = kx∗ k ≤ 1 ≤ ε.
kxk≤1
∗
Therefore x is an element of the intersection in (5). To prove the reverse
inclusion, let ξ ∈ FX such that
(6) πx (ξ) + πy (ξ) − πx+y (ξ) =ϕx,y (ξ) = 0,
(7) λπx (ξ) − πλx (ξ) =ϕx,λ (ξ) = 0
and
(8) |ξ(z)| =|Jz(ξ)| ≤ ε,
for all x, y, z ∈ X with kzk ≤ 1 and λ ∈ F. The items (6) and (7) imply that
ξ(x + y) = ξ(x) + ξ(y) and ξ(λx) = λξ(x) for all x, y ∈ X and λ ∈ F, so ξ is a linear
functional, and by (8), ξ is continuous with the strong topology of X ∗ . Therefore
ξ ∈ BX ∗ . Thus the equality in (5) holds.
Since, ϕx,y and ϕx,λ are continuous with respect to η, for all x, y ∈ X and
−1
λ ∈ F, it follows that ∩ ϕ−1 x,y ({0}) and ∩ ψx,λ ({0}) are closed subsets of
x,y∈X x∈X,λ∈F
FX with respect to η. On the other hand, by (4), Jx = πx |X ∗ , for all x ∈ X.
This implies that each Jx is continuous with respect to ηX ∗ . Therefore for every
x ∈ X with kxk ≤ 1, Jx−1 (B[0, ε]) is a closed subset of X ∗ with respect to ηX ∗ ,
and so ∩ Jx−1 (B[0, ε]) is a closed subset of FX with respect to η, because
kxk≤1
X ∗ = π0−1 ({0}) is a closed subset of FX with respect to η. Then, BX ∗ is a closed
subset of FX with respect to η and hence BX ∗ is a compact set of FX with the
topology η.
Consequently BX ∗ is a compact set in X ∗ with the topology ηX ∗ and so by
Proposition 4.3, BX ∗ is a compact set with respect to the topology σ(X ∗ , X). 

Proposition 4.5. Let J be the natural embedding of X into its biudal. If X

is reflexive then
J : (X, σ(X, X ∗ )) → (X ∗∗ , σ(X ∗∗ , X ∗ ))
is an homeomorphism.
Proof: Let x ∈ X and W be a neighborhood of J(x) for σ(X ∗∗ , X ∗ ). Then
there exist x∗1 , . . . , x∗n ∈ X ∗ and ε > 0 such that Vε;x∗1 ,...,x∗n (J(x)) ⊆ W . Note that
Vε;x∗1 ,...,x∗n (x) is a basic in σ(X, X ∗ ). We prove that J(Vε;x∗1 ,...,x∗n (x)) ⊆ W . Let
z ∈ Vε;x∗1 ,...,x∗n (x), this implies that
|(J(z) − J(x))(x∗i )| = |x∗i (z) − x∗i (x)|
=|x∗i (z − x)| < ε,
for all i = 1, . . . , n. Therefore J(z) ∈ Vε;x∗1 ,...,x∗n (J(x)) and so J(z) ∈ W . Then
J is continuous at x, and since the choice of x was arbitrary, we have that J is
continuous.
On the other hand, since J is an isometric isomorphism, see Theorem 2.6, it
follows that J is injective and since X is reflexive, it follows that J is surjective.
Therefore J is invertible. We prove that J −1 : (X ∗∗ , σ(X ∗∗ , X ∗ )) → (X, σ(X, X ∗ ))
is continuous. Let x∗∗ 0 ∈ X
∗∗
and let V ∈ σ(X, X ∗ ) be such that J −1 (x∗∗ 0 ) ∈ V.
There exist x0 ∈ X, ε > 0 and x∗1 , . . . , x∗n ∈ X ∗ such that J(x0 ) = x∗∗ 0 and
10 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

Vε;x∗1 ,...,x∗n (x0 ) ⊆ V. We claim that Vε;x∗1 ,...,x∗n (x∗∗ ∗∗ ∗

0 ) is a basic in σ(X , X ) such that
−1 ∗∗ −1 ∗∗
J (Vε;x∗1 ,...,x∗n (x0 )) ⊆ V . In fact, let s ∈ J (Vε;x∗1 ,...,x∗n (x0 )), then there exists
w∗∗ ∈ X ∗∗ such that s = J −1 (w∗∗ ) and |(w∗∗ − x∗∗ ∗
0 )(xi )| < ε for all i = 1, . . . , n.
This implies that
|x∗i (s − x0 )| =|x∗i (s) − x∗i (x0 )|
=|Js (x∗i ) − Jx0 (x∗i )|
=|(J(s) − J(x0 ))(x∗i )|
=|(w∗∗ − x∗∗ ∗
0 )(xi )| < ε,

for all i = 1, . . . , n. Therefore, s ∈ Vε;x∗1 ,...,x∗n (x0 ) and so s ∈ V . 

Corollary 4.6. If X is reflexive then for every r > 0,

r
BX = {x ∈ X : kxk ≤ r}
is a compact set with the weak topology of X.
Proof: We first show that for each α > 0 and U ∈ σ(X ∗∗ , X ∗ ) it follows that αU ∈
σ(X ∗∗ , X ∗ ). Let y ∗∗ ∈ αU , then y ∗∗ = αu∗∗ for some u∗∗ ∈ U . There exist ε > 0
and x∗1 , . . . , x∗n ∈ X ∗ such that Vε;x∗1 ,...,x∗n (u∗∗ ) ⊆ U . If z ∗∗ ∈ Vαε;x∗1 ,...,x∗n (αu∗∗ )
then
 ∗∗ 
z 1
− u∗∗ (x∗i ) = |(z ∗∗ − αu∗∗ ) (x∗i )|
α α
1
< αε = ε,
α
∗∗ ∗∗
for all i = 1, . . . , n, and so zα ∈ U . Therefore z ∗∗ = α zα ∈ αU . Consequently,
∗∗
Vαε;x∗1 ,...,x∗n (y ) ⊆ αU .
∗∗
We show now that BX r
∗∗ = {x ∈ X ∗∗ : kx∗∗ k ≤ r} is a compact set with
the ?−weak topology of X . Let {Uα }α∈I be a family in σ(X ∗∗ , X ∗ ) such that
∗∗

BXr
∗∗ ⊆ ∪ Uα . Then BX ∗∗ = {x∗∗ ∈ X ∗∗ : kx∗∗ k ≤ 1} ⊆ ∪ 1r Uα . By Theorem
α∈I α∈I
n n
4.4, there exist α1 , . . . , αn ∈ I such that BX ∗∗ ⊆ ∪ 1r Uαi . Therefore BX
r
∗∗ ⊆ ∪ Uαi
i=1 i=1
r
and so BX ∗∗ is compact.
r
Finally, by Proposition 4.5, BX = J −1 (BX
r
∗∗ ) is a compact set with respect to
∗
the topology σ(X, X ). 

5. Paracompactness under the weak topology

In this section some conditions are given for a normed space to be paracompact
with the weak topology.
Theorem 5.1. If one of the following conditions hold:
(i) X is separable with the strong topology,
(ii) X is reflexive,
then (X, σ(X, X ∗ )) is a Lindelöf space.
Proof: Let {Uα }α∈I be a covering of X with Uα ∈ σ(X, X ∗ ) for all α ∈ I.
(i) Since X is a separable metric space with the metric generated by the norm, it
follows tha it is second countable. Let {Bn }n∈N be a basis of X. For each n ∈ N, let
Sn = {α ∈ I : Bn ⊆ Uα }. If Sn 6= ∅ then by the well ordering principle we consider
 5. PARACOMPACTNESS UNDER THE WEAK TOPOLOGY 11

Uαn = min{Uα : α ∈ Sn } for some order in {Uα : α ∈ Sn }. Let P = {n ∈ N : Sn 6=

∅}. We show that {Uαn }n∈P is a countable subcovering of {Uα }α∈I . Let x ∈ X,
then there exists α ∈ I such that x ∈ Uα . This implies that Vε,x∗1 ,...,x∗m (x) ⊆ Uα for
m
\
some ε > 0 and x∗1 , . . . , x∗m ∈ X ∗ . Since Vε;x∗1 ,x∗2 ,...,x∗m (x) = (x∗i )−1 (B(x∗i (f ), ε)),
i=1
it follows that Vε;x∗1 ,x∗2 ,...,x∗m (x) is an open set with the strong topology that contains
x. Therefore there exists n ∈ N such that f ∈ Bn ⊆ Vε;x∗1 ,x∗2 ,...,x∗m (x). Consequently,
α ∈ Sn and so Sn 6= ∅. Thus n ∈ P and x ∈ Bn ⊆ Uαn .
n
(ii) For each n ∈ N, let BX = {x ∈ X : kxk ≤ n}. Then
n
X = ∪ BX .
n∈N
n
By Corollary 4.6, each BX is a compact set with the weak topology of X, so
n n
for every n ∈ N, there exist α1 , . . . , αN n
∈ I such that
Nn
n
BX ⊆ ∪ Uαni .
i=1

This implies that  

Nn
X= ∪ ∪ Uαni .
n∈N i=1

Therefore, ∪ {Uαni : 1 = 1, . . . , Nn } is a countable subcovering of U. 

n∈N

Finally, we show that a reflexive normed space X is paracompact with the weak
topology.
Corollary 5.2. If one of the following conditions hold:
(i) X is separable with the strong topology,
(ii) X is reflexive,
then (X, σ(X, X ∗ )) is a paracompact space.
Proof: By Theorem 5.1, (X, σ(X, X ∗ )) is a Lindelöf space and by Theorem 3.5,
(X, σ(X, X ∗ )) is regular. Thus by Theorem 2.13, (X, σ(X, X ∗ )) is paracompact.


Example 5.3. The following normed spaces are paracompact with the weak
topology:
(i) Every Hilbert space H. Indeed, we show that H is reflexive. For each
z ∈ H let Rz : H → F be defined as Rz (x) = hx, zi. It is clear that Rz
is a linear functional. Moreover, from Schwarz inequality it follows that
|Rz (x)| = |hx, zi| ≤ kxkkzk for all x ∈ H, so Rz ∈ H ∗ . Let y ∗∗ ∈ H ∗∗
and define g : H → F by g(z) = y ∗∗ (Rz ). Then g is a conjugate-linear
functional and
|g(z)| = |y ∗∗ (Rz )| ≤ ky ∗∗ kkRz k
= ky ∗∗ k sup |Rz (x)|
kxk≤1

≤ ky ∗∗ k sup kxkkzk
kxk≤1
∗∗
≤ ky kkzk,
12 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

for all z ∈ H. Therefore g is a bounded. By Theorem 2.7, there exists

w ∈ H such that g(x) = hw, xi for all x ∈ H. Let f ∈ X ∗ , again by
Theorem 2.7, there exists zf ∈ H such that f (x) = hx, zf i for all x ∈ H.
Thus
Jw (f ) = f (w) = hw, zf i = g(zf ) = y ∗∗ (Rzf ) = y ∗∗ (f ).
Therefore J(w) = y ∗∗ .
(ii) The space
C([a, b]) = {f : [a, b] → F : f is continuous}
with the norm
kf k∞ = sup |f (x)|.
x∈[a,b]

We prove that C([a, b]) is separable. Without loss of generality we suppose

that F = R. For every n ∈ N, let Qn ([a, b]) = {q0 +q1 x+q2 x2 +· · ·+qn xn :
qi ∈ Q for all i = 1, . . . , n}. We set Q = ∪ Qn ([a, b]). It is clear that Q
n∈N
is a countable set. Let f ∈ C([a, b]) and ε > 0, then by the Weierstrass
Approximation Theorem, there exists a polynomial p(x) = a0 + a1 x +
a2 x2 + · · · + an xn such that
ε
kp − f k∞ < .
2
Take M > 0 such that |xk | ≤ M for all x ∈ [a, b] and k = 0, . . . , n.
By density of Q, it follows that there exists qk ∈ Q such that |qk − ak | <
ε 2 n
2M (n+1) for all k = 0, . . . , n. We set q(x) = q0 + q1 x + q2 x + . . . , qn x ,
then q ∈ Q and
kq − f k∞ ≤ kq − pk∞ + kp − f k∞
ε
< sup |q(x) − p(x)| +
x∈[a,b] 2
n
X ε
≤M |qk − ak | + < ε.
2
k=0

Therefore Q = C([a, b]).

(iii) The space
( )
X
n p
`p (N) = (xn ) ∈ K : |xn | < ∞ , p ∈ [1, ∞),
n∈N

with the norm

! p1
X
k(xn )kp = |xn |p .
n∈N
This space is separable. Indeed, suppose that K = R and let S =
{(q1 , q2 , . . . , qn , 0, . . . ) : qi ∈ Q for all i = 1, . . . , n, n ∈ N}, (xn )n∈N ∈
`p (N) and ε > 0. There exists n0 ∈ N such that
∞
X  ε p
|xi |p < .
i=n0 +1
2
 5. PARACOMPACTNESS UNDER THE WEAK TOPOLOGY 13

Moreover, there exist q1 , . . . , qn0 ∈ Q such that

Xn0  ε p
|qi − xi |p < .
i=1
2

Therefore, if y = (q1 , . . . , qn0 , 0, . . . ) and x = (xn )n∈N then ky − xkp < ε.

Thus S = `p (N).
(iv) The space
n
Lp ([a, b]) = f : [a, b] → F : f is Lebesgue medible and

Z b o
|f (x)|p dm(x) < ∞ , p ∈ [1, ∞),
a
with the semi-norm
! p1
Z b
kf kp = |f (x)|p dm(x) .
a

We show that this space is separble. By the proof of (ii), C([a, b]) is a
separable space, so there exists a countable set Q such that Q = C([a, b]).
Let f ∈ Lp ([a, b]) and ε > 0. By the density of C([a, b]) in Lp ([a, b]) (see
[9, Theorem 3.14]), there exists g ∈ C([a, b]) such that
ε
kg − f kp < .
2
Also, there exists q ∈ Q such that
ε
sup |q(x) − g(x)| < .
x∈[a,b] 2(b − a)1/p
Therefore
kq − f kp ≤kq − gkp + kg − f kp
ε
<kq − gkp +
2
Z b ! p1
p ε
= |q(x) − g(x)| dm(x) +
a 2
! 1
Z b p
ε ε
≤ dm(x) + = ε.
2(b − a)1/p a 2

Thus Lp ([a, b]) is separable.

Remark 5.4. Reflexivity of a normed space X is not a necessary condition
for paracompactness of X. Indeed, consider X = C([a, b]) the space of continuous
functions, defined in Example 5.3 (ii). This space is paracompact but is not reflex-
ive. We show this last assertion. For each x ∈ [a, b] consider δx : X → R defined
as δx (f ) = f (x). It is clear that δx is a linear functional, moreover it is bounded,
because
|δx (f )| = |f (x)| ≤ kf k∞ ,
14 1. PARACOMPACTNESS OF A NORMED SPACE WITH THE WEAK TOPOLOGY

for all f ∈ X. Therefore {δx }x∈[a,b] ⊆ X ∗ . Let x, y ∈ [a, b] with x 6= y. Without

loss of generality we may assume that x < y. Since
x 1
6= 0,
y 1
it follows that there exist α, β ∈ R such that αx + β = 0 and αy + β = 1. Let
f : [a, b] → R defined by

0
 if s ∈ [a, x)
f (s) = αs + β if s ∈ [x, y]

1 if s ∈ (y, b].


Then f (x) = 0, f (y) = 1, f ∈ X and kf k∞ = 1. This implies that

|δy (f ) − δx (f )|
kδy − δx k ≥ = 1.
kf k
Consequently, {B(δx , 12 )}x∈[a,b] is a family of pairwise disjoint open subsets of
X . Therefore, X ∗ is not second countable. Now, if X is reflexive then J : X → X ∗∗
∗

is an isometric isomorphism of X onto X ∗∗ . From Example 5.3 (ii), X is a separable

space, then X ∗∗ is a separable space, and so by Theorem 2.4, X ∗ is a separable
metric space. Therefore, X ∗ is second countable, which is a contradiction.
Acknowledgement. This work was supported by the CONACYT.
 Bibliography

[1] L. Alaoglu, Weak topologies of normed linear spaces, Ann. of Math. 41(1940), 252-267.
[2] H. H. Corson, The weak topology of a Banach space, Trans. Amer. Math. Soc. 101(1961),
207-230.
[3] H. H. Corson and J. Lindenstrauss, On function spaces which are Lindeldf spaces, Trans.
Amer. Math. Soc. 121(1966), 476-491.
[4] G. A. Edgar, R. F. Wheller, Topological properties of Banach spaces, Pacific J. Math.
115(1984), 317-350.
[5] S. Gabriyelyan, J. Kakol, W. Kubiś, W. Marciszewski, Networks for the weak topology of
Banach and Fréchet spaces, J. Math. Anal. Appl. 432(2015), 1183-1199.
[6] C. S. Kubrusly, Elements of operator theory, Birkhäuser, Boston, 2001.
[7] W. Marciszewski, Lindelf property in function spaces and a related selection theorem, Proc.
Amer. Math. Soc. 101(1987), 545-550.
[8] E. Michel, A note on paracompact spaces, Proc. Amer. Math. Soc. 4(1953), 831-838.
[9] W. Rudin, Real and complex analysis, McGraw-Hill, New York, 1987.

E-mail addresses:
[email protected] (Salvador Sánchez-Perales),
[email protected] (Slavisa V. Djordjević),
[email protected] (Sonia Venancio).

15