

()


@
Applied General Topology

c© Universidad Politécnica de Valencia

Volume 12, no. 2, 2011

pp. 143-162

Core compactness and diagonality in spaces of
open sets

Francis Jordan and Frédéric Mynard

Abstract

We investigate when the space OX of open subsets of a topological

space X endowed with the Scott topology is core compact. Such con-

ditions turn out to be related to infraconsonance of X, which in turn is

characterized in terms of coincidence of the Scott topology of OX ×OX
with the product of the Scott topologies of OX at (X, X). On the
other hand, we characterize diagonality of OX endowed with the Scott

convergence and show that this space can be diagonal without being

pretopological. New examples are provided to clarify the relationship

between pretopologicity, topologicity and diagonality of this important

convergence space.

2010 MSC: 54A20, 54A10, 54B20, 54D45.

Keywords: Scott convergence, Scott topology, upper Kuratowski conver-
gence, upper Kuratowski topology, core compact, diagonal con-

vergence, pretopology, consonance, infraconsonance.

1. Introduction

Definitions and notations concerning convergence structures follow [3] and
are gathered as an appendix at the end of these notes (1). In particular, if X
and Y are two convergence spaces, the continuous convergence [X, Y ] on the set
C(X, Y ) of continuous maps from X to Y is the coarsest convergence making
the evaluation jointly continuous. This is the canonical function space structure
in the cartesian closed category of convergence spaces and continuous maps.
This paper is concerned with certain properties of this canonical convergence
on functions valued into the Sierpiński space:

1Terms and notations that are not defined in the text can be found in the appendix.


144 F. Jordan and F. Mynard

Let $0 and $1 denote two versions of the Sierpiński space on {0, 1}: {0} is the
only non-trivial open subset of $0, and {1} is the only non-trivial open subset
of $1. Let 1A : X → {0, 1} denote the indicator function of a subset A of a
convergence space X defined by 1A(x) = 1 if and only if x ∈ A. With those
conventions, A is an open subset of X if and only if 1A : X → $1 is continuous
and closed if and only if 1A : X → $0 is continuous. Therefore, C(X, $1) can be
identified with the set OX of open subsets of X, and C(X, $0) can be identified
with the set CX of closed subsets of X.
If X is a topological space, the continuous convergences [X, $1] and [X, $0]

turn out to be familiar convergences, on OX and CX respectively (see, e.g., [5]):

U ∈ lim[X,$1] F ⇐⇒ U ⊆
⋃

F ∈F

int(
⋂

O∈F

O)(1.1)

C ∈ lim[X,$0] F ⇐⇒
⋂

F ∈F

cl(
⋃

A∈F

A) ⊆ C.(1.2)

Both are instances of Scott convergence (in the sense of, for instance, [11]),
i.e.,

(1.3) x ∈ lim F ⇐⇒
∨

F ∈F

∧
F ≥ x,

in the complete lattices (OX, ⊆) and (CX, ⊇) respectively. However, (1.2) is
usually called upper Kuratowski convergence. The topological modification
T [X, $0] is called upper Kuratowski topology. The topological modification
T [X, $1] is the Scott topology, whose open sets are exactly compact families:
families A of open subsets of X that are closed under open supersets and
satisfy ⋃

i∈I

Oi ∈ A =⇒ ∃F ∈ [I]
<∞ :

⋃

i∈F

Oi ∈ A

for any collection {Oi : i ∈ I} of open subsets of X, where [I]
<∞ denotes the

set of finite subsets of I. For instance, if K is a compact subset of X then

O(K) := {O ∈ OX : K ⊆ O}

is a compact family. The family of all sets O(K) where K ranges over compact
subsets of X is a basis for a topology on OX = C(X, $1). We write Ck(X, $1)
for the corresponding topological space.
Of course, complementation c : OX → CX defined by c(U) = X \ U is an

homeomorphism between [X, $1] and [X, $0] and we only need study one of
these two convergences. We choose to focus on OX. Hence, from now on, $
means $1 and we only formulate results on OX but they have counterparts
for the upper Kuratowski convergence and upper Kuratowski topology on CX,
which the interested reader can easily write out (See e.g., [8] for a study of
[X, $0] and T [X, $0] on CX).
It is clear from the definitions that

[X, $] ≥ T [X, $] ≥ Ck(X, $).


Core compactness and diagonality in spaces of open sets 145

A convergence space X is called T -dual if [X, $] = T [X, $] and consonant [5]
if T [X, $] = Ck(X, $).
It is easily seen that

∩ : [X, $] × [X, $] → [X, $]

(U, V ) 7→ U ∩ V

is continuous for any convergence space X.
To simplify the discussion, let us momentarily assume that X is a completely

regular topological space. In this case, X is T -dual if and only if X is locally
compact. Moreover,

Ck(X, $) = [KX, $],

where KX is the locally compact modification of X (2) [21, Proposition 4.3].
Therefore intersection is also jointly continuous for Ck(X, $). Additionally,
infraconsonance in the sense of [9] can then be characterized in similar terms:
X is infraconsonant if ∩ : T [X, $] × T [X, $] → T [X, $] is continuous. Thus

T -dual =⇒ consonant =⇒infraconsonant.

The problem of characterizing T -dual topological spaces has long been settled
(e.g., [13], [23]): a topological space X is T -dual if and only if it is core compact.
Recall that a topological space X is core compact if for every x and O ∈ O(x),
there is U ∈ O(x) such that every open cover of O has a finite subfamily that
covers U.
In the case of a general convergence space X, the situation is more compli-

cated. It is known (e.g., [24], [7]) that the following are equivalent:

∀Y, T (X × Y ) ≤ X × T Y ;(1.4)

∀Y = T Y, [X, Y ] = T [X, Y ]

T (X × [X, $]) ≤ X × T [X, $];

X is T -dual.

Moreover, it was shown in [7] that

(1.5) X is core compact =⇒ X is T -dual =⇒ X is T -core compact,

where a convergence space is called core compact if whenever x ∈ lim F, there
is G ≤ F with x ∈ lim G and for every G ∈ G there is G′ ∈ G such that G′ is
compact at G; and a convergence space is called T -core compact if whenever
x ∈ lim F and U ∈ OX(x), there is F ∈ F that is compact at U.
The three notions clearly coincide if X is topological. However, so far, it was

not known whether they do in general. At the end of the paper, we provide an
example (Example 5.8) of a T -dual convergence that is not core-compact.
Section 2 examines when [X, $] and T [X, $] are T -dual. The latter question,

while natural in itself, is motivated by its connection (established in Section
3) with the (now recently solved [18]) problem [9, Problem 1.2] of finding a

2with the abuse of notation that [KX, $] is identified with the convergence it induces on
the subset C(X, $) of C(KX, $).


146 F. Jordan and F. Mynard

completely regular infraconsonant topological space that is not consonant. We
obtain that X is infraconsonant whenever T [X, $] is T -dual, and we prove more
generally that X is infraconsonant if and only if the Scott topology on OX ×OX
for the product order coincides with the product of the Scott topologies at the
point (X, X) (Theorem 4.2).
Infraconsonance was introduced while studying the Isbell topology on the

set of real-valued continuous functions over a topological space. In fact a com-
pletely regular space X is infraconsonant if and only if the Isbell topology
on the set of real-valued continuous functions on X is a group topology [6,
Corollary 4.6]. On the other hand, the fact that the Scott topology on the
product does not coincide in general with the product of the Scott topologies
has been at the origin of a number of errors, as pointed out for instance in
[11, p.197]. Therefore, Theorem 4.2 provides new motivations to investigate
infraconsonance.
In [7], it is shown that a convergence space X is T -core compact if and only

if [X, $] is pretopological. Therefore, if X is topological, [X, $] is topological
whenever it is pretopological. As topologies are exactly the diagonal (3) pre-
topologies, it raises the question of whether [X, $] is diagonal whenever X is
topological. In Section 5, diagonality of [X, $] is characterized in terms of a vari-
ant of core-compactness that do not need to coincide with core-compactness.
As a result [X, $] does not need to be diagonal even if X is topological.

2. core-compactness of OX

For a general convergence space X, the underlying set of [X, $] can still
be identified with the collection OX of open subsets of X (or T X ), but the
characterization (1.1) of convergence in [X, $] needs to be modified. A family
S of subsets of a convergence space Y is a cover of A ⊆ Y if every filter on Y
converging to a point of A contains an element of the family S. Then we have:

U ∈ lim[X,$] F ⇐⇒ {
⋂

O∈F

O : F ∈ F} is a cover of U.

The space [[X, $], $] has as underlying set the set of Scott-open subsets of
OX, that is, if X is topological, the set κ(X) of openly isotone compact families
on X. Note that the family

{U∼ := {A ∈ κ(X) : U ∈ A} : U ∈ OX}

forms a subbase for a topology on κ(X), called Stone topology. It is the analog
on κ(X) of the standard topology on the set βX of ultrafilters on X.
As observed in [10, Proposition 5.2], when X is topological, the convergence

[X, $] is based in filters of the form

(2.1) O♮(P) := {O(P) : P ∈ P},

where P is an ideal subbase of open subsets of X, that is, such that there
is P ∈ P with

⋃
Q∈P0

Q ⊆ P whenever P0 is a finite subfamily of P. More

3in the sense of e.g., [4]. See Definition 5.1.


Core compactness and diagonality in spaces of open sets 147

precisely, for every filter F on [X, $] with U ∈ lim[X,$] F there is an open

cover P of U that forms an ideal subbase, such that U ∈ lim[X,$] O
♮(P) and

O♮(P) ≤ F.
Note also that

(2.2) A ⊆ B =⇒ A ∈ lim[[X,$],$]{B}
↑,

for every A and B in κ(X). In particular if O is [[X, $], $]-open, A ∈ O and
A ⊆ B ∈ κ(X) then B ∈ O.
It was observed in [12], as a consequence of a general theory, that if X is

topological, then so is [[X, $], $]. We provide here an independent proof, that
shows that [X, $] is then a core compact convergence.

Proposition 2.1. If X is topological, then [X, $] is core compact, so that
[[X, $], $] is topological. More precisely, it is homeomorphic to κ(X) with the
Stone topology.

Proof. Let U ∈ lim[X,$] O
♮(P) for an ideal subbase P of open subsets of X.

Then for each P ∈ P, the set O(P) is a compact subset of [X, $] because

P ∈ lim[X,$] O(P). Indeed, P = int
(⋂

O∈O(P ) O
)
.

U∼ is [[X, $], $]-open for each U ∈ OX. Indeed, if A ∈ U
∼ ∩ lim[[X,$],$] F

then
{⋂

B∈F
B : F ∈ F

}
is a cover of A (in the sense of convergence) so that

there is F ∈ F with
⋂

B∈F
B ∈ {U}↑ because U ∈ lim[X,$]{U}

↑ ∩ A. In other
words, F ⊆ U∼, so that U∼ ∈ F.
Conversely, if O is [[X, $], $]-open and A ∈ O, there is U ∈ A such that

U∼ ⊆ O. Otherwise, for each U ∈ A, there is B ∈ κ(X) with U ∈ B and B /∈ O.

In that case, Û := {B ∈ κ(X) : U ∈ B, B /∈ O} 6= ∅ for all U ∈ A. Note
also that in view of (2.2), BU ∩ BV ∈ Û ∩ V̂ whenever BU ∈ Û and BV ∈ V̂ .

Therefore
{⋂

i∈I Ûi : Ui ∈ A : card I < ∞
}

is a filter-base generating a filter

F. This filter converges to A in [[X, $], $]. To show that, we need to see that{⋂
B∈Û

B : U ∈ A
}
is a cover of A for [X, $]. In view of the form (2.1) of a base

for [X, $], it is enough to show that if U0 ∈ A and P is an ideal subbase of open
subsets of X covering U0, then there is A ∈ A with

⋂
B∈Â

B ∈ O♮(P). Because
U0 ⊆

⋃
P ∈P

P and A is a compact family, there is a finite subfamily P0 of P
such that

⋃
P ∈P0

P ∈ A. Since P is an ideal subbase, there is P ∈ P ∩A. Then
O(P) ⊆

⋂
B∈P̂

B, which concludes the proof that A ∈ lim[[X,$],$] F. On the
other hand, O /∈ F, which contradicts the fact that O is open for [[X, $], $]. �

Remark 2.2. If X is a non topological convergence space, then by [8, Corollary
16.3], the open subsets of [X, $] are the rigidly compact families: families A of
open subsets of X, closed under open supersets, such that adhξ H#A whenever
H is a filter such that for every H ∈ H there is a closed subset B of H with
B ∈ A#. Hence the underlying set of [[X, $], $] is no longer κ(X) but the larger
set of rigidly compact families on X. We will see below (Proposition 2.3) that
the convergence [[X, $], $] fails to be topological in this case. We do not know


148 F. Jordan and F. Mynard

whether T [[X, $], $] can be expressed as an analog of the Stone topology on the
set of rigidly compact families on X.

In order to investigate when T [X, $] is core compact, we will need notions and
results from [7]. The concrete endofunctor EpiT of the category of convergence
spaces (and continuous maps) is defined (on objects) by

EpiT X = i
−[T [X, $], $]

where i : X → [[X, $], $] is defined by i(x)(f) = f(x). In view of [7, Theorem
3.1]

(2.3) W ≥ EpiT X ⇐⇒ T [X, $] ≥ [W, $],

where X ≥ W have the same underlying set. In particular, X is T -dual if and
only if X ≥ EpiT X. A convergence space X is called epitopological if i : X →
[[X, $], $] is initial (in the category Conv of convergence spaces and continuous
maps). Epitopologies form a reflective subcategory Epi of Conv and the
(concrete) reflector is given (on objects) by Epi X = i−[[X, $], $]. Because
[Epi X, $] = [X, $], it is enough to consider epitopologies in the study of dual
convergences. Observe that a topological space is epitopological. Note that
if [X, $] is T -dual, then Epi X = X is topological. Therefore, in contrast to
Proposition 2.1, [X, $] is not T -dual if X is not topological.

Proposition 2.3. Let X be an epitopological space. Then X is topological if
and only if [X, $] is T -dual.

Note also that Epi X ≤ EpiT X and that EpiT ◦ Epi = EpiT , so that EpiT re-
stricts to an expansive endofunctor of Epi. By iterating this functor, we obtain
the coreflector on T -dual epitopologies. More precisely, if F is an expansive
concrete endofunctor of C, we define the transfinite sequence of functors F α

by F 1 = F and F αX = F
(∨

β<α
F βX

)
. For each epitopological space X,

there is an ordinal α(X) such that

Epi
α(X)
T X = Epi

α(X)+1
T X := DT X.

Proposition 2.4. The class of T -dual epitopologies is concretely coreflective
in Epi and the coreflector is DT .

While this proposition easily follows from general results in [7] or [20], and
Galois connections, we provide a self-contained proof.

Proof. The class of T -dual convergences is closed under infima because
[
∧

i∈I

Xi, Z

]
=
∨

i∈I

[Xi, Z].

Indeed, if each Xi is T -dual, then
[
∧

i∈I

Xi, $

]
=
∨

i∈I

[Xi, $] =
∨

i∈I

T [Xi, $] ≤ T

(
∨

i∈I

[Xi, $]

)
= T

([
∧

i∈I

Xi, $

])
,


Core compactness and diagonality in spaces of open sets 149

and
∧

i∈I
Xi is T -dual. The functor EpiT is expansive on Epi and therefore,

so is DT . Moreover, DT X is T -dual for each epitopological space X because

[DT X, $] = [Epi
α(X)+1
T X, $] ≤ T [Epi

α(X)
T X, $] = T [DT X, $].

Therefore, for each epitopological space X, there exists the coarsest T -dual
convergence X finer than X. By definition X ≤ X ≤ DT X. Then [X, $] ≤
[X, $] and [X, $] is topological, so that [X, $] ≤ T [X, $]. But EpiT X is the
coarsest convergence with this property. Therefore EpiT X ≤ X = EpiT X and
DT X ≤ X. �

Proposition 2.5. If X is a core compact topological space, then [X, $] is also
a core compact topological space.

Proof. [X, $] = T [X, $] because X is core compact, and [X, $] is T -dual by
Proposition 2.1, because X is topological. Therefore T [X, $] is a core compact
topology. �

However, if X is a non-topological T -dual convergence space (4), then [X, $] =
T [X, $] is not core compact, by Proposition 2.3. In other words, we have:

Proposition 2.6. If [X, $] is topological then X is topological if and only if
[X, $] is core compact.

In particular, DT X is topological if and only if [DT X, $] is core compact.

Theorem 2.7. If X ≥ T DT X then T [X, $] is core compact if and only if X
is a core compact topological space.

Proof. We already know that if X is a core compact topological space then
[X, $] = T [X, $] and that [X, $] is core compact by Proposition 2.1. Conversely,
if T [X, $] is core compact then [T [X, $], $] is topological, so that EpiT X is
topological. Under our assumptions,

X ≥ T DT X ≥ T EpiT X = EpiT X,

hence by (2.3), X is T -dual. Therefore [X, $] = T [X, $] is core compact and,
in view of Proposition 2.3, X is topological, and T -dual, hence a core compact
topological space. �

Remark 2.8. Note that, at least among Hausdorff topological spaces, Theorem
2.7 generalizes [19, Corollary 3.6] that states that if X is first countable, then
X is core compact if and only if T [X, $] is core compact. Indeed, the locally
compact coreflection KX of a Hausdorff topological space is T -dual so that
DT X ≤ KX. Moreover, [1] characterizes a number of topological properties in
terms of functorial inequalities of the form

X ≥ JE(X),

4Such convergences exist: take for a instance a non-locally compact Hausdorff regular
topological k-space. Then X = TKhX but X < KhX so that KhX is non-topological.


150 F. Jordan and F. Mynard

where J is a concrete reflector and E a concrete coreflector in the category of
convergence spaces. For instance, it is observed that a (Hausdorff) topological
space X is a k-space if and only if

(2.4) X ≥ T K X,

so that (2.4) can be taken as a definition of a k-convergence. Hence if X
is a Hausdorff topological k-space (in particular a first-countable space) then
X ≥ T DT X.
On the other hand, in view of the results of [1], if f : X → Y is a quotient

map (in the topological sense) and X is core compact (so that X = DT X) then
Y ≥ T DT Y .
We will see in the next section that similarly, if X is a consonant topological

space, then T [X, $] is core compact if and only if X is locally compact.

Problem 2.9. Are there completely regular non locally compact topological
spaces X such that T [X, $] is core compact?

Of course, in view of Remark 2.8, such a space cannot be a k-space or con-
sonant.

3. Core compact dual, Consonance, and infraconsonance

A topological space is consonant if T [X, $] = Ck(X, $), that is, if every Scott
open subset A of OX is compactly generated, that is, there are compact subsets
(Ki)i∈I of X such that A =

⋃
i∈I

O(Ki) [5]. A space is infraconsonant [9] if
for every Scott open subset A of OX there is a Scott open set C such that
C ∨ C ⊆ A, where C ∨ C := {C ∩ D : C, D ∈ C}.
The notion’s importance stems from Theorem 3.1 below. If the set C(X, Y )

of continuous functions from X to Y is equipped with the Isbell topology (5),
we denote it Cκ(X, Y ), while Ck(X, Y ) denotes C(X, Y ) endowed with the
compact-open topology. Note that Cκ(X, $) = T [X, $].

Theorem 3.1 ([6]). Let X be a completely regular topological space. The
following are equivalent:

(1) X is infraconsonant;
(2) addition is jointly continuous at the zero function in Cκ(X, R);
(3) Cκ(X, R) is a topological vector space;
(4) ∩ : T [X, $] × T [X, $] → T [X, $] is jointly continuous.

On the other hand, if X is consonant then Cκ(X, R) = Ck(X, R) so that
consonance provides an obvious sufficient condition for Cκ(X, R) to be a topo-
logical vector space.

5whose sub-basic open sets are given by

[A, U] := {f ∈ C(X, Y ) : ∃A ∈ A, f(A) ⊆ U} ,

where A ranges over openly isotone compact families on X and U ranges over open subsets
of Y .


Core compactness and diagonality in spaces of open sets 151

Hence Theorem 3.1 becomes truly interesting if completely regular examples
of infraconsonant non consonant spaces can be provided [9, Problem 1.2]. The
first author recently obtained the first example of this kind [18]. The following
results show that a space answering positively Problem 2.9 would necessarily be
infraconsonant and non-consonant and might provide an avenue to construct
new examples.

Theorem 3.2. If X is topological and T [X, $] is core compact then X is in-
fraconsonant.

Proof. [9, Lemma 3.3] shows the equivalence between (1) and (4) in Theorem
3.1, and that the implication (4)=⇒(1) does not require any separation. There-
fore, it is enough to show that ∩ : T [X, $] × T [X, $] → T [X, $] is continuous.
Since X is topological, [X, $] is T -dual by Proposition 2.1. In view of (1.4)

T ([X, $] × [X, $]) ≤ [X, $] × T [X, $]

so that T ([X, $] × [X, $]) ≤ T ([X, $] × T [X, $]). If T [X, $] is core compact,
hence T -dual then T ([X, $] × T [X, $]) ≤ T [X, $] × T [X, $] so that

(3.1) T ([X, $] × [X, $]) ≤ T [X, $] × T [X, $].

Therefore the continuity of ∩ : [X, $] × [X, $] → [X, $] implies that of ∩ :
T ([X, $] × [X, $]) → T [X, $] because T is a functor, and in view of (3.1), that
of ∩ : T [X, $] × T [X, $] → T [X, $]. �

Recall that a basis for the topology of Ck(X, $) is given by sets of the form
O(K) where K ranges over compact subsets of X.

Theorem 3.3. Let X be a topological space. If Ck(X, $) is core compact then
X is locally compact.

Proof. If X is not locally compact, then Ck(X, $) � [X, $] (e.g., [23, 2.19]) so
that there is U0 ∈ OX with U0 /∈ lim[X,$] Nk(U0). Therefore, there is x0 ∈ U0
such that x0 /∈ int(

⋂
V ∈O(K) V ) whenever K is a compact subset of X with

K ⊆ U0. In other words, for each such K and for each U ∈ O(x0) there is
VU ∈ O(K) and xU ∈ U \ VU . Then Ck(X, $) is not core compact at U0.
Indeed, there is U0 ∈ O(x0) such that for every compact set K with K ⊆ U0,
the k-open set O(K) is not relatively compact in O(x0). To see that, consider
the cover S := {O(xU ) : U ∈ O(x0)} of O(x0). No finite subfamily of S
covers O(K) because for any finite choice of U1, . . . , Un in O(x0), we have
W := ∩i=ni=1 VUi ∈ O(K) but W /∈ ∪

i=n
i=1 O(xUi). �

Note that a Hausdorff topological space X is locally compact if and only if it
is core compact, and that the Scott open filter topology on OX then coincides
with Ck(X, $) (e.g., [11, Lemma II.1.19]). Hence Theorem 3.3 could also be
deduced (for the Hausdorff case) from [19, Corollary 3.6].

Corollary 3.4. If X is a consonant topological space such that T [X, $] is core
compact, then X is locally compact.


152 F. Jordan and F. Mynard

4. Scott topology of the product versus product of Scott
topologies

We now turn to a new characterization of infraconsonance, which motivates
further the systematic investigation of the notion.
Recall that in a complete lattice (L, ≤) the Scott convergence is given by

(1.3), and the Scott topology is its topological modification. A subset A of L is
Scott-open if and only if it is upper-closed and satisfies

∨
D ∈ A =⇒ ∃d ∈ D ∩ A,

for every directed supset D of L (e.g., [11]). A product of complete lattices is
a complete lattice for the coordinatewise order, and we can therefore consider
the Scott topology on the product for the coordinatewise order, and compare
it with the product of the Scott topologies.

Proposition 4.1. T ([X, $]2) is the Scott topology on OX × OX.

Theorem 4.2. A space X is infraconsonant if and only if the product T [X, $]×
T [X, $] of the Scott topologies and the Scott topology T ([X, $] × [X, $]) on the
product coincide at (X, X).

Lemma 4.3. A subset S of OX × OX is [X, $]
2-open if and only if

(1) S = S↑, that is, if (U, V ) ∈ S and U ⊆ U′, V ⊆ V ′ then (U′, V ′) ∈ S;
(2) S is coordinatewise compact, that is,

(
⋃

i∈I

Oi,
⋃

j∈J

Vj) ∈ S =⇒ ∃I0 ∈ [I]
<ω, J0 ∈ [J]

<ω : (
⋃

i∈I0

Oi,
⋃

j∈J0

Vj) ∈ S

Proof. Assume S is [X, $]2-open and let (U, V ) ∈ S and U ⊆ U′, V ⊆ V ′.
Then (U, V ) ∈ lim[X,$]2{(U

′, V ′)}↑ so that (U′, V ′) ∈ S. Assume now that
(
⋃

i∈I
Oi,
⋃

j∈J
Vj) ∈ S. Then {O(

⋃
i∈F

Oi) : F ∈ [I]
<∞} is a filter-base for a

filter γ on OX such that
⋃

i∈I
Oi ∈ lim[X,$] γ and {O(

⋃
j∈D

Vj) : D ∈ [J]
<∞} is

a filter-base for a filter η on OX such that
⋃

j∈J
Vj ∈ lim[X,$] η. Hence S ∈ γ×η

because S is [X, $]2-open. Therefore, there are finite subsets I0 of I and J0 of
J such that O(

⋃
i∈I0

Oi) × O(
⋃

j∈J0
Vj) ⊆ S, so that (

⋃
i∈I0

Oi,
⋃

j∈J0
Vj) ∈ S.

Conversely, assume that S satisfies the two conditions of the Lemma and
(U, V ) ∈ S∩lim[X,$]2(γ×η). Since U ⊆

⋃
G∈γ

int(
⋂
G) and V ⊆

⋃
H∈η

int(
⋂

H∈H
H),

we have, by the first condition, that

(
⋃

G∈γ

int(
⋂

G∈G

G),
⋃

H∈η

int(
⋂

H∈H

H)) ∈ S.

By the second condition, there are G1, . . . , Gk ∈ γ and H1, . . . , Hn ∈ η such
that

(

k⋃

i=1

int(
⋂

G∈Gi

G),

n⋃

j=1

int(
⋂

H∈Hj

H)) ∈ S.


Core compactness and diagonality in spaces of open sets 153

Therefore (int(
⋂

G∈
⋂

k
i=1

Gi
G), int(

⋂
H∈

⋂
n
j=1

Hj
H)) ∈ S so that

(

k⋂

i=1

Gi,
n⋂

j=1

Hj) ⊆ S,

and S ∈ γ × η. �

Proof of Proposition 4.1. In view of Lemma 4.3, every [X, $]2-open subset of
OX ×OX is Scott open. Conversely, consider a Scott open subset S of OX ×OX.
We only have to check that S satisfies the second condition in Lemma 4.3. Let

(
⋃

i∈I
Oi,
⋃

j∈J
Vj) ∈ S. The set D := {

(⋃
i∈I0

Oi,
⋃

j∈J0
Vj

)
: I0 ∈ [I]

<ω, J0 ∈

[J]<ω} is a directed subset of OX ×OX (for the coordinatewise inclusion order)
whose supremum is (

⋃
i∈I

Oi,
⋃

j∈J
Vj). As S is Scott-open, there are finite

subsets I0 of I and J0 of J such that
(⋃

i∈I0
Oi,
⋃

j∈J0
Vj

)
∈ S. �

Lemma 4.4. If A ∈ κ(X) then SA := {(U, V ) ∈ OX × OX : U ∩ V ∈ A}
↑ is

[X, $]2-open.

Proof. Let (
⋃

i∈I
Oi,
⋃

j∈J
Vj) ∈ SA. Then

(
⋃

i∈I

Oi) ∩ (
⋃

j∈J

Vj) =
⋃

(i,j)∈I×J

Oi ∩ Vj ∈ A.

By compactness of A, there is a finite subset I0 of I and a finite subset J0 of J
such that

⋃
(i,j)∈I0×J0

Oi ∩ Vj ∈ A, so that (
⋃

i∈I0
Oi,
⋃

j∈J0
Vj) ∈ SA. In view

of Lemma 4.3, SA is [X, $]
2-open. �

Lemma 4.5. If S is [X, $]2-open, then

↓ S := OX({U ∪ V : (U, V ) ∈ S})

is a compact family on X.

Proof. If U ∪ V ⊆
⋃

i∈I
Oi for some (U, V ) ∈ S then

(⋃
i∈I

Oi,
⋃

i∈I
Oi
)
∈

S so that, in view of Lemma 4.3, there is a finite subset I0 of I such that(⋃
i∈I0

Oi,
⋃

i∈I0
Oi
)
∈ S. Hence

⋃
i∈I0

Oi ∈↓ S. �

Proof of Theorem 4.2. Suppose that X is infraconsonant. Note that (T [X, $])2 ≤
T ([X, $]2) is always true, so that we only have to prove the reverse inequality
at (X, X). Consider an [X, $]2-open neighborhood S of (X, X). By Lemma
4.5, the family ↓ S is compact. By infraconsonance, there is C ∈ κ(X) with
C ∨ C ⊆↓ S. Note that

C × C ⊆ S,

because if (C1, C2) ∈ C × C then C1 ∩ C2 ∈↓ S so that C1 ∩ C2 ⊇ U ∪ V for
some (U, V ) ∈ S, and therefore (C1, C2) ∈ S.
Conversely, assume that N[X,$]2(X, X) = NT [X,$]2(X, X) and let A ∈ κ(X).

By Lemma 4.4, SA ∈ N[X,$]2(X, X) so that SA ∈ NT [X,$]2(X, X). In other
words, there are families B and C in κ(X) such that B × C ⊆ SA. In particular


154 F. Jordan and F. Mynard

D := B ∩ C belongs to κ(X) and satisfies D × D ⊆ SA. By definition of SA, we
have that D ∨ D ⊆ A and X is infraconsonant. �

5. Topologicity, pretopologicity and diagonality of [X, $]

A selection for a convergence space X is a map S[·] : X → FX such that
x ∈ limX S[x] for all x ∈ X.

Definition 5.1. A convergence space X is diagonal if for every selection S[·]
and every filter F with x0 ∈ limX F the filter

(5.1) S[F] :=
⋃

F ∈F

⋂

x∈F

S[x]

converges to x0. If this property only holds when F is additionally principal,
we say that X is F0-diagonal.

Of course, every topology is diagonal. In fact a convergence is topological if
and only if it is both pretopological and diagonal (e.g., [4]).
In order to compare our condition for diagonality of [X, $] with core-compactness,

we first rephrase the latter.

Lemma 5.2. A topological space is core compact if and only if for every x ∈ X,
every U ∈ O(x) and every family H of filters on X, we have

(5.2) ∀H ∈ H : adh H ∩ U = ∅ =⇒ x /∈ adh
∧

H∈H

H.

Proof. If X is core compact, then there is V ∈ O(x) which is relatively compact
in U. If adh H∩U = ∅, then U ⊆

⋃
H∈H

(cl H)c so that, by relative compactness
of V in U there is, for each H ∈ H, a set HH ∈ H with V ∩ cl HH = ∅. Then⋃

H∈H
HH ∈

∧
H∈H

H but
⋃

H∈H
HH ∩ V = ∅ so that x /∈ adh

∧
H∈H

H.
Conversely, if (5.2) is true, consider the family H := {H ∈ FX : adh H ∩ U =

∅}. In view of (5.2), x /∈ adh
∧

H∈H
H so that there is V ∈ O(x) such that

V /∈
(∧

H∈H
H
)#

. Now V is relatively compact in U because any filter than
meshes with V cannot be in H and has therefore an adherence point in U. �

Recall that [X, $] = P [X, $] if and only if X is T -core compact, and that,
if X is topological, [X, $] is topological whenever it is pretopological. While
the latest is well-known, and follows for instance from the results of [7], it
seems difficult to find an elementary argument in the literature, which is why
we include the following proposition, which also illustrates the usefulness of
Lemma 5.2.

Proposition 5.3. If X is topological and [X, $] is pretopological, then [X, $] is
topological.

Proof. We will show that under these assumptions, X satisfies (5.2). Let x ∈ X
and U ∈ O(x). Let H be a family of filters satisfying the hypothesis of (5.2).
Let H ∈ H. Consider the filter base H∗ := {O(X \ cl(H)): H ∈ H} on [X, $].
Since adh(H)∩U = ∅, it follows that U ∈ lim H∗. Since [X, $] is pretopological,


Core compactness and diagonality in spaces of open sets 155

U ∈ lim
∧

H∈H
H∗. In particular, there exist, for each H ∈ H, a HH ∈ H such

that

x ∈ int(
⋂ ⋃

H∈H

O(X \ cl HH)) = int(
⋂

H∈H

(X \ cl HH))

= int(X \ (
⋃

H∈H

cl HH))

⊆ X \ cl(
⋃

H∈H

HH).

Thus, x /∈ adh(
∧

H∈H
H). �

In other words, if [X, $] is pretopological it is also diagonal, provided that
X is topological. We will see that even if X is topological, [X, $] is not always
diagonal. Moreover it can be diagonal without being pretopological (examples
5.5 and 5.7).
We call a topological space injectively core compact if for every x ∈ X and

U ∈ O(x) the conclusion of (5.2) holds for every family H of filters such that
there is an injection θ : H → O(U) satisfying adh H∩θ(H) = ∅ for each H ∈ H.
As such a family H clearly satisfies the premise of (5.2), every core compact
space is in particular injectively core compact.

Theorem 5.4. Let X be a topological space. The following are equivalent:

(1) X is injectively core compact;
(2) [X, $] is diagonal;
(3) [X, $] is F0-diagonal.

Proof. (1)=⇒(2): Let S[�] : OX → FOX be a selection for [X, $] and let U ∈
lim[X,$] F. If x ∈ U, there is F ∈ F such that x ∈ int

(⋂
O∈F

O
)
:= V . Note

that F ⊆ O(V ). For each O ∈ F , consider the filter HO on X generated
by {clX

(⋃
W∈S

W c
)
: S ∈ S[O]}. Because O ∈ lim[X,$] S[O], we have that

adhX HO ∩ O = ∅. Choose G ⊆ F so that {HO : O ∈ G} = {HO : O ∈ F}
and HO 6= HP for every two distinct O, P ∈ G. Because X is injectively
core compact and H := {HO : O ∈ G} satisfies the required condition (with
θ(HO) = O ), we conclude that x /∈ adhX

∧
O∈G

HO. By the way we chose
G, we have

∧
O∈G

HO =
∧

O∈F
HO. So, x /∈ adhX

∧
O∈F

HO. In other words,
there is an H ∈

∧
O∈F

HO such that x /∈ clX H, that is, x ∈ intX H
c. Therefore,

for each O ∈ F there is SO ∈ S[O] such that

x ∈ int(
⋂

O∈F

int(
⋂

W∈SO

W)) ⊆ int(
⋂

W∈
⋃

O∈F
SO

W).

In other words, there is F ∈ F and M ∈
∧

O∈F
S[O] such that x ∈

intX
(⋂

W∈M
W
)
, that is, U ∈ lim[X,$] S[F].

(2)=⇒(3) is clear. (3)=⇒(1): Suppose X is not injectively core compact.
Then there is x ∈ X, U ∈ O(x) and a family H of filters on X with an
injective map θ : H → O(U) such that θ(H) ∩ adhX H = ∅ for each H ∈ H
but x ∈ adhX

∧
H∈H

H. Define a relation ∼ on H by H1 ∼ H2 provided that


156 F. Jordan and F. Mynard

the collections {cl(H): H ∈ H1} and {cl(H): H ∈ H2} both generate the same
filter. Clearly, ∼ is an equivalence relation. Let H∗ ⊆ H be such that H∗

contains exactly one element of each equivalence class of ∼. For each H ∈ H∗

let H∗ be the filter with base {cl(H): H ∈ H}. Let J = {H∗ : H ∈ H∗}.
Define θ∗ : J → O(U) so that θ∗(J ) = θ(H), where H ∈ H∗ is such that

J = H∗. It is easily checked that θ∗ is injective. Since adh(H∗) = adh(H)
for every H ∈ H∗, we have θ∗(J ) ∩ adh(J ) = ∅. It is also easy to check that

x ∈ adh
(∧

J ∈J
J
)
.

For each J ∈ J, the filter J̃ generated on OX by the filter-base {OX(X \ J) :
J ∈ J } converges to θ∗(J ). Consider now the subset θ∗(J) of O(U) ⊆ OX and
the selection S[�] : OX → FOX defined by S[θ(J )] = J̃ for each J ∈ J and
S[O] = {O}↑ for O /∈ θ∗(J). This is indeed a well-defined selection because θ∗

is injective.
Notice that U ∈ lim[X,$] θ

∗(J) because θ∗(J) ⊆ O(U). Let L ∈ S[θ∗(J)]. We
may pick from each J ∈ J a closed set JJ ∈ J such that

⋃
J ∈J

Ox(X\JJ ) ⊆ L.
Let V be an open neighborhood of x. Since x ∈ adhX

∧
J ∈J

J and
⋃

J ∈J
JJ ∈∧

J ∈J J , there is an J0 ∈ J such that V ∩ JJ0 6= ∅. Since V 6⊆ X \ JJ0 and
X\JJ0 ∈ OX(X\JJ0), V 6⊆

⋂
OX(X\JJ0). Since OX(X\JJ0) ⊆ L, V 6⊆

⋂
L.

Since V was an arbitrary neighborhood of x, x 6∈ int(
⋂
L). Thus, U /∈ S[θ∗(J)].

Therefore, [X, $] is not F0-diagonal. �

A cardinal number κ is regular if a union of less than κ-many sets of car-
dinality less than κ has cardinality less than κ. A strong limit cardinal κ is a
cardinal for which card(2A) < κ whenever card(A) < κ. A strongly inaccessible
cardinal is a regular strong limit cardinal. Uncountable strongly inaccessible
cardinals cannot be proved to exist within ZFC, though their existence is not
known to be inconsistent with ZFC. Let us denote by (*) the assumption that
such a cardinal exist.

Example 5.5 (A Hausdorff space X such that [X, $] is diagonal but not pre-
topological under (*)). Assume that κ is a (uncountable) strong limit cardinal.
Let X be the subspace of κ ∪ {κ} endowed with the order topology, obtained by
removing all the limit ordinals but κ. Since X is a non locally compact Haus-
dorff topological space, [X, $] is not pretopological. To show that X is injectively
core compact, we only need to consider x = κ and U ∈ O(κ) in the definition,
because κ is the only non-isolated point of X. Let H be a family of filters on
X admitting an injective map θ : H → O(U) such that adh H ∩ θ(H) = ∅ for
each H ∈ H. For each H ∈ H there is HH ∈ H such that κ /∈ cl(HH) so that
card(HH) < κ. Since U is a neighborhood of κ, there is a β < κ such that
{ξ ∈ X : β ≤ ξ} ⊆ U. Since V \ U ⊆ {ξ ∈ X : ξ < β} for every V ∈ O(U), we
have cardH ≤ card O(U) ≤ 2β. Since κ is a strong limit cardinal, cardH < κ.
Since κ is regular, card

⋃
H∈H

HH < κ so that κ /∈ adh
∧

H∈H
H.


Core compactness and diagonality in spaces of open sets 157

We do not know if the existence of large cardinals is necessary for the con-
struction of a Hausdorff space X such that [X, $] is diagonal and not pretopo-
logical, but, as the next proposition shows, such a space cannot be too small.
Let c denote the cardinality of the real numbers.

Proposition 5.6. Let X be a Hausdorff topological space. If X is a non locally
compact space of character not exceeding c, then [X, $] is not diagonal.

Proof. Let p ∈ X be such that X is not locally compact at p. Since X is not
compact, there is a neighborhood U of p such that X \ U is infinite. Since
X is Hausdorff, there exists a countably infinite A ⊆ X \ U and mutually
disjoint open sets {Wa : a ∈ A} such that a ∈ Wa for every a ∈ A. It follows
that the collection {U ∪

⋃
a∈E

Wa : E ⊆ A} is a collection of c-many distinct
elements of O(U). Since the character of X is at most c, there is a neighborhood
base B at p with at most c-many elements. Since X is not locally compact
at p, there is for each B ∈ B a filter HB on B such that adh(HB) = ∅.
Let H = {HB : B ∈ B}. Since card B ≤ card O(U), there is an injection
θ : H → O(U). Clearly, adh(HB) ∩ θ(HB) = ∅ for every B ∈ B. However,
p ∈ adh

(∧
B∈B

HB
)
. Hence, X is not injectively core compact at p. Thus,

[X, $] is not diagonal. �

On the other hand, we can construct in ZFC a T0 space X such that [X, $]
is diagonal and not pretopological.

Example 5.7 (A T0 space X such that [X, $] is diagonal but not pretopo-
logical). Let Z stand for integers and c+ be the cardinal successor of c. Let
∞ be a point that is not in c+ × Z and X = {∞} ∪ (c+ × Z). For each
(α, n) ∈ c+ × Z define Sα,n = {(β, k): α ≤ β and n ≤ k}. For each α ∈ c

+, let
Tα = {(β, k): α ≤ β and k ∈ Z} ∪ {∞}. Topologize X by declaring all sets of
the form Tα and Sα,n to be sub-basic open sets.
We show that X is not core compact at ∞. Let U be a neighborhood of ∞.

There is an α such that Tα ⊆ U. Notice that Tα+1 ∪ {S0,n : n ∈ Z} is a cover
of X but no finite subcollection covers Tα. Thus, X is not core compact at ∞.
In particular, [X, $] is not pretopological.
Let (α, n) ∈ X \ {∞}. Let U be an open neighborhood of (α, n). Since

(α, n) ∈ U it follows from the way we chose our sub-base that Sα,n ⊆ U. Since
(α, n) has a minimal open neighborhood, X is core compact at (α, n).
Let V be an open neighborhood of ∞. There is an α such that Tα ⊆ V . Let

U ⊆ X be an open superset of V . For every n ∈ Z, U ∩ (c+ × {n}) 6= ∅.
For each n ∈ Z define αn = min{β : (β, n) ∈ U}. Notice that {β : αn ≤
β} × {n} = U ∩ (c+ × {n}) and αn ≤ α. Since each open superset of V will
determine a unique sequence (αn)n∈Z, it follows that the open supersets of V
can injectively be mapped into the countable sequences on {β : β ≤ α} × Z.
Since {β : β ≤ α} × Z has cardinality at most c, {β : β ≤ α} × Z has at most
c-many countable sequences. Thus, V has at most c-many supersets.
Let V be an open neighborhood of ∞, H be a collection of filters, and θ : H →

OX(V ) be an injection such that adh(H)∩θ(H) = ∅ for every H ∈ H. Since V


158 F. Jordan and F. Mynard

has at most c-many open supersets, card H ≤ c. Let H ∈ H. Since ∞ /∈ adh H,
there is an αH ∈ c

+ such that adh(H) ∩ TαH = ∅. Let α = (supH∈H αH) + 1 <
c
+. It is easy to check that, adh

(∧
H∈H

H
)
∩ Tα = ∅. Thus, X is injectively

core compact at ∞.
Since X is injectively core compact at each point, [X, $] is diagonal, by The-

orem 5.4.

Example 5.8 (A T -dual convergence space that is not core compact). Con-
sider a partition {An : n ∈ ω} of the set ω

∗ of free ultrafilters on ω satisfying
the condition that for every infinite subset S of ω and every n ∈ ω, there is
U ∈ An with S ∈ U. Let M := {mn : n ∈ ω} be disjoint from ω and let
X := ω ∪ M. Define on X the finest convergence in which lim{mn}

↑ = M for
all n ∈ ω, and each free ultrafilter U on ω converges to mn (and mn only),
where n is defined by U ∈ An.

Claim. X is not core compact.

Proof. Let mn ∈ M and U ∈ An. Pick S ⊆ ω, S ∈ U, and k 6= n. For every
U ∈ U there is W ∈ Ak such that U ∈ W. But lim W = {mk} is disjoint from
S. �

Claim. X is T -core compact, and therefore [X, $] is pretopological.

Proof. For each mn ∈ M, the set M is included in every open set containing
mn because mn ∈

⋂
k∈ω

lim{mk}
↑. If U is a non-trivial convergent ultrafilter in

X then lim U = {mn} for some n ∈ ω. For any S ∈ U, S ∩ ω is infinite and any
free ultrafilter W on S ∩ ω belongs to one of the element Ak of the partition,
so that lim W = {mk} intersects M, and therefore any open set containing
mn. �

Claim. [X, $] is diagonal.

Proof. Let S[�] : OX → FOX be a selection for [X, $] and let U ∈ lim[X,$] F.
Now, {

⋂
F : F ∈ F} is a (convergence) cover of U.

Let x ∈ U and D be a filter on X such that x ∈ lim D. There is an F ∈ F
and a D ∈ D such that D ⊆

⋂
F := V .

Assume x ∈ ω, in which case D = {x}↑. In particular, x ∈ O for every
O ∈ F . For every O ∈ F there is a TO ∈ S[O] such that x ∈

⋂
TO. Now,

x ∈
⋂⋂

O∈F TO ∈ S[F ]. So,
⋂⋂

O∈F TO ∈ {x}
↑ = D.

Assume x ∈ M. In this case, M ∩ O 6= ∅ for all O ∈ F and, by definition
of the convergence on X, M ⊆ O for all O ∈ F . Since O ∈ lim[X,$] S[O] and
M ⊆ O, there is S ∈ S[O] such that x ∈

⋂
S, and, since each element of

S is open, M ⊆
⋂
S. If there is no S ∈ S[O] such that O ⊆

⋂
S then the

filter H generated by {(O ∩ ω) \
⋂

S : S ∈ S[O]} is non degenerate. Notice
that it is not free, for otherwise there would be an n ∈ ω and U ∈ An with
U ≥ H. But mn ∈ lim U ∩ O, and there would be S ∈ S[O] such that

⋂
S ∈ U,

which is not possible. Therefore, there is y ∈
⋂

S∈S[O] ((O ∩ ω) \
⋂
S) which


Core compactness and diagonality in spaces of open sets 159

contradicts O ∈ lim[X,$] S[O]. Hence, there is SO ∈ S[O] such that O ⊆
⋂
SO.

Now, D ⊆
⋂
F ⊆

⋂
O∈F

⋂
SO. In particular,

⋂
O∈F

⋂
SO ∈ D.

Thus, {
⋂
J : J ∈ S[F]} is a cover of U, and [X, $] is diagonal. �

Therefore [X, $] is pretopological and diagonal, hence topological, and X is
T -dual.

6. Appendix: convergence spaces

A family A of subsets of a set X is called isotone if B ∈ A whenever A ∈ A
and A ⊆ B. We denote by A↑ the smallest isotone family containing A, that
is, the collection of subsets of X that contain an element of A. If A and B are
two families of subsets of X we say that B is finer than A, in symbols A ≤ B,
if for every A ∈ A there is B ∈ B such that B ⊆ A. Of course, if A and B
are isotone, then A ≤ B ⇐⇒ A ⊆ B. This defines a partial order on isotone
families, in particular on the set FX of filters on X. Every family (Fα)α∈I of
filters on X admits an infimum

∧

α∈I

Fα :=
⋂

α∈I

Fα =

{
⋃

α∈I

Fα : Fα ∈ Fα

}↑
.

On the other hand the supremum even of a pair of filters may fail to exist.
We call grill of A the collection A# := {H ⊆ X : ∀A ∈ A, H ∩ A 6= ∅}.
It is easy to see that A = A## if and only if A is isotone. In particular
F = F## ⊆ F# if F is a filter. We say that two families A and B of subsets of
X mesh, in symbols A#B, if A ⊆ B#, equivalently if B# ⊆ A. The supremum
of two filters F and G exists if and only if they mesh, in which case F ∨ G =

{F ∩ G : F ∈ F, G ∈ G}
↑
. An infinite family (Fα)α∈I of filters has a supremum∨

α∈I
Fα if pairwise suprema exist and for every α, β ∈ I there is γ ∈ I with

Fγ ≥ Fα ∨ Fβ.
A convergence ξ on a set X is a relation between X and the set FX of filters

on X, denoted x ∈ limξ F whenever x and F are in relation, satisfying that
x ∈ limξ{x}

↑ for every x ∈ X, and limξ F ⊆ limξ G whenever F ≤ G. The pair
(X, ξ) is called a convergence space. A function f : (X, ξ) → (Y, σ) between
two convergence space is continuous if

x ∈ limξ F =⇒ f(x) ∈ limσ f(F),

where f(F) is the filter {f(F) : F ∈ F}↑ on Y . If ξ and τ are two convergences
on the same set X, we say that ξ is finer than τ, in symbols ξ ≥ τ, if limξ F ⊆
limτ F for every F ∈ FX. This defines a partial order on the set of convergence
structures on X, which defines a complete lattice for which supremum ∨i∈Iξi
and infimum ∧i∈Iξi of a family {ξi : i ∈ I} of convergences are defined by

lim∨i∈Iξi F =
⋂

i∈I

limξi F,

lim∧i∈Iξi F =
⋃

i∈I

limξi F.


160 F. Jordan and F. Mynard

Every topology can be identified with a convergence, in which x ∈ lim F
if F ≥ N(x), where N(x) is the neighborhood filter of x for this topology.
A convergence obtained this way is called topological. Moreover, a function
f : X → Y between two topological spaces is continuous in the usual topological
sense if and only if it is continuous in the sense of convergence. On the other
hand, every convergence determines a topology in the following way: A subset
C of a convergence space (X, ξ) is closed if limξ F ⊆ C for every filter F on X
with C ∈ F. A subset O is open if its complement is closed, that is, if O ∈ F
whenever limξ F ∩ O 6= ∅. The collection of open subsets for a convergence
ξ is a topology T ξ on X, called topological modification of ξ. The topology
T ξ is the finest topological convergence coarser than ξ. If f : (X, ξ) → (Y, τ)
is continuous, so is f : (X, T ξ) → (Y, T τ). In other words, T is a concrete
endofunctor of the category Conv of convergence spaces and continuous maps.
Continuity induces canonical notions of subspace convergence, product con-

vergence, and quotient convergence. Namely, if f : X → Y and Y carries a
convergence τ, there is the coarsest convergence on X making f continuous
(to (Y, τ)). It is denoted f−τ and called initial convergence for f and τ. For
instance if S ⊆ X and (X, ξ) is a convergence space, the induced convergence
by ξ on S is by definition i−ξ where i is the inclusion map of S into X. Sim-
ilarly, if {(Xi, ξi) : i ∈ I} is a family of convergence space, then the product
convergence Πi∈Iξi on the cartesian product Πi∈IXi is the coarsest conver-
gence making each projection pj : Πi∈IXi → Xj continuous. In other words,
Πi∈Iξi = ∨i∈Ip

−
i ξi. In the case of a product of two factors (X, ξ) and (Y, τ),

we write ξ × τ for the product convergence on X × Y .
Dually, if f : X → Y and (X, ξ) is a convergence space, there is the finest

convergence on Y making f continuous (from (X, ξ)). It is denoted fξ and
called final convergence for f and ξ. If f : (X, ξ) → Y is a surjection, the
associated quotient convergence on Y is fξ. Note that if ξ is a topology, the
quotient topology is not fξ but T fξ.
The functor T is a reflector. In other words, the subcategory Top of Conv

formed by topological spaces and continuous maps is closed under initial con-
structions. Note however that the functor T does not commute with initial
constructions. In particular T ξ × T τ ≤ T (ξ × τ) but the reverse inequal-
ity is generally not true. Similarly, if i : S → (X, ξ) is an inclusion map,
i−(T ξ) ≤ T (i−ξ) but the reverse inequality may not hold. A convergence ξ
is pretopological or a pretopology if limξ

∧
α∈I

Fα =
⋂

α∈I
limξ Fα. Of course,

every topology is a pretopology, but not conversely. For any convergence ξ
there is the finest pretopology Pξ coarser than ξ. Moreover, x ∈ limP ξ F if and
only if F ≥ Vξ(x) where Vξ(x) :=

∧
x∈limξ F

F is called vicinity filter of x. The

subcategory PrTop of Conv formed by pretopological spaces and continuous
maps is reflective (closed under initial constructions). Moreover, in contrast
with topologies, the reflector P commutes with subspaces. However, like T, it
does not commute with products.


Core compactness and diagonality in spaces of open sets 161

The adherence adhξ F of a filter F on a convergence space (X, ξ) is by defi-
nition

adhξ F :=
⋃

H#F

limξ H =
⋃

U∈U(F)

limξ U,

where UX denotes the set of ultrafilters on X and U(F) denotes the set of
ultrafilters on X finer than the filter F. We write adhξ A for adhξ{A}

↑. Note
that in a convergence space X, adhξ may not be idempotent on subsets of
X. In fact a pretopology is a topology if and only if adh is idempotent on
subsets. We reserve the notations cl and int to topological closure and interior
operators.
A family A of subsets of X is compact at a family B for ξ if

F#A =⇒ adhξ F#B.

We call a family compact if it is compact at itself. In particular, a subset A of
X is compact if {A} is compact, and compact at B ⊆ X if {A} is compact at
{B}.
Given a class D of filters, a convergence is called based in D or D-based if for

every convergent filter F, say x ∈ lim F, there is a filter D ∈ D with D ≤ F and
x ∈ lim D. A convergence is called locally compact if every convergent filter
contains a compact set, and hereditarily locally compact if it is based in filters
with a filter-base composed of compact sets. For every convergence, there is the
coarsest locally compact convergence Kξ that is finer than ξ and the coarsest
hereditarily locally compact convergence Khξ that is finer than ξ. Both K and
Kh are concrete endofunctors of Conv that are also coreflectors.
If A ⊆ X and (X, ξ) is a convergence space, then O(A) denotes the collection

of open subsets of X that contain A and if A is a family of subsets of X then
O(A) :=

⋃
A∈A

O(A). A family is called openly isotone if A = O(A). Note
that in a topological space X, an openly isotone family A of open subsets of
X is compact if and only if, whenever

⋃
i∈I

Oi ∈ A and each Oi is open, there
is a finite subset J of I such that

⋃
i∈J

Oi ∈ A.
If (X, ξ) and (Y, σ) are two convergence spaces, C(X, Y ) or C(ξ, σ) denote

the set of continuous maps from X to Y . The coarsest convergence on C(X, Y )
making the evaluation map e : X × C(X, Y ) → Y , e(x, f) = f(x), jointly
continuous is called continuous convergence and denoted [X, Y ] or [ξ, σ]. Ex-
plicitly,

f ∈ lim[X,Y ] F ⇐⇒ ∀x∈X∀G∈FX:x∈limξ G f(x) ∈ limσ e (G × F) .

References

[1] S. Dolecki, Convergence-theoretic methods in quotient quest, Topology Appl. 73 (1996),
1–21.

[2] S. Dolecki, Properties transfer between topologies on function spaces, hyperspaces and
underlying spaces, Mathematica Pannonica 19, no. 2 (2008), 243–262.

[3] S. Dolecki, An initiation into convergence theory, in Beyond Topology, Contemporary
Mathematics 486, Mynard and Pearl (eds), AMS, 2009, 115–161.

[4] S. Dolecki and G. Greco. Topologically maximal pretopologies, Studia Math. 77 (1984),
no. 3, 265–281.


162 F. Jordan and F. Mynard

[5] S. Dolecki, G. H. Greco, and A. Lechicki, When do the upper Kuratowski topology
(homeomorphically, Scott topology) and the cocompact topology coincide?, Trans. Amer.
Math. Soc. 347 (1995), 2869–2884.

[6] S. Dolecki, F. Jordan and F. Mynard, Group topologies coarser than the Isbell topology,
Topology Appl., to appear.

[7] S. Dolecki and F. Mynard, Convergence-theoretic mechanisms behind product theorems,
Topology Appl. 104 (2000), 67–99.

[8] S. Dolecki and F. Mynard, Hyperconvergences, Appl. Gen. Topol. 4, no. 2 (2003), 391–
419.

[9] S. Dolecki and F. Mynard, When is the Isbell topology a group topology?, Topology Appl.
157, no. 8 (2010), 1370–1378.

[10] S. Dolecki and F. Mynard, A unified theory of function spaces and hyperspaces: local
properties, submitted.

[11] G. Gierz, K. Hofmann, K. Keimel, J. Lawson, M. Mislove and D. S. Scott, Continuous
lattices and domains, Encyclopedia of mathematics and its applications 93, Cambridge
University Press, 2003.

[12] R. Heckmann, A non-topological view of dcpos as convergence spaces, Theoret. Comput.
Sci. 305, no. 1-3 (2003), 159–186.

[13] K. H. Hofmann and J. D. Lawson, The spectral theory of distributive continuous lattices,

Trans. Amer. Math. Soc. 246 (1978), 285–309.
[14] J. R. Isbell, Function spaces and adjoints, Math. Scandinavica 36 (1975), 317–339.
[15] J. R. Isbell, Meet-continuous lattices, Symposia Mathematica 16 (1975), 41–54.
[16] F. Jordan, Coincidence of function space topologies, Topology Appl. 157, no. 2 (2010),

336–351.
[17] F. Jordan, Coincidence of the Isbell and fine Isbell topologies, Topology Appl. 157, no.

15 (2010), 2392–2401.
[18] F. Jordan, An infraconsonant nonconsonant completely regular space, Topology Appl.,

to appear.
[19] X. Lin and S. Yujin, On core compactness of semilattices and frames with the Scott open

filter topology, Questions Answers Gen. Topology 15 (1997), 189–194.
[20] F. Mynard, Coreflectively modified continuous duality applied to classical product theo-

rems, Appl. Gen. Topol. 2, no. 2 (2001), 119–154.
[21] F. Mynard, First-countability, sequentiality and tightness of the upper Kuratowski con-

vergence, Rocky Mountain J. Math. 33, no. 3 (2003), 1011–1038.
[22] Elliot Pearl (ed.), Open problems in topology II, Elsevier, 2007.
[23] F. Schwarz, Powers and exponential objects in initially structured categories and appli-

cation to categories of limits spaces, Quaest. Math. 6 (1983), 227–254.
[24] F. Schwarz, Product compatible reflectors and exponentiability, Categorical topology

(Toledo, Ohio, 1983), 505–522, Sigma Ser. Pure Math., 5, Heldermann, Berlin, 1984.

(Received December 2010 – Accepted June 2011)

Francis Jordan (fejord@hotmail.com)
Queensborough Community College, Queens, NY, USA

Frédéric Mynard (fmynard@georgiasouthern.edu)
Georgia Southern University, Statesboro, GA30460, USA


	Core compactness and diagonality in spaces of open sets. By F. Jordan and F. Mynard