DowPorterStWAGT.dvi


@
Applied General Topology

c© Universidad Politécnica de Valencia

Volume 5, No. 2, 2004

pp. 243-264

Spaces whose Pseudocompact Subspaces are
Closed Subsets

Alan Dow, Jack R. Porter, R.M. Stephenson, Jr., and

R. Grant Woods∗

Dedicated to Professor W. Wistar Comfort on the occasion of his seventieth

birthday

Abstract. Every first countable pseudocompact Tychonoff space X
has the property that every pseudocompact subspace of X is a closed
subset of X (denoted herein by “FCC”). We study the property FCC
and several closely related ones, and focus on the behavior of extension
and other spaces which have one or more of these properties. Char-
acterization, embedding and product theorems are obtained, and some
examples are given which provide results such as the following. There
exists a separable Moore space which has no regular, FCC extension
space. There exists a compact Hausdorff Fréchet space which is not
FCC. There exists a compact Hausdorff Fréchet space X such that X,
but not X2, is FCC.

2000 AMS Classification: Primary 54D20; Secondary 54D35, 54B10, 54D55
or 54G20

Keywords: feebly compact, pseudocompact, Fréchet, sequential, product,
extension

∗The first author gratefully acknowledges partial research support from the National Sci-
ence Foundation, Grant No. 2975010131. The third and fourth authors gratefully acknowl-
edge partial research support from the University of Kansas and the sabbatical leave programs
of their respective institutions, and in the case of the fourth author, from the Natural Sciences
and Engineering Research Council of Canada.



244 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

1. Introduction and terminology

For topological spaces X and Y , C(X, Y ) will denote the family of continuous
functions from X into Y , C(X) will denote C(X, R), and C∗(X) will denote
the family of bounded functions in C(X). A space X is called pseudocompact
provided that C(X) = C∗(X). This definition was first given for Tychonoff
spaces, i.e., completely regular T1-spaces, by E. Hewitt [10].

For terms not defined here, see [5], [6] or [15]. Except where noted otherwise,
no separation axioms are assumed.

Some properties of interest that are closely related to pseudocompactness
are listed in Theorem 1.1.

Theorem 1.1. Let X be a space. Then each statement below implies the next
one, and all of properties (B1)–(B6) are equivalent.

(A) The space X is pseudocompact and completely regular.
(B1) Every locally finite family of open sets of X is finite.
(B2) Every pairwise disjoint locally finite family of open sets of X is finite.
(B3) Every sequence of nonempty open subsets of X has a cluster point in

X.
(B4) If U = {Un : n ∈ N} is a sequence of nonempty open subsets of X such

that Ui ∩ Uj = ∅ whenever i 6= j, then U has a cluster point in X.
(B5) Every countable open filter base on X has an adherent point.
(B6) Every countable open cover of X has a finite subcollection whose union

is dense in X.
(C) X is pseudocompact.

We recall that the adherence of a filter base F on a space X is the intersection
of the closures of the members of F, and by a cluster point of a sequence
{Un : n ∈ N} of subsets of a space X is meant a point p ∈ X such that
for every neighborhood V of p, V ∩ Un 6= ∅ for infinitely many integers n.
A sequence denoted {Un : n ∈ N} will be referred to as a pairwise disjoint
sequence provided that Ui ∩ Uj = ∅ whenever i 6= j .

Proofs or references to proofs of the different results in Theorem 1.1 can
be found in [1], [7], [8], [15] or [26]. These properties have been found useful
by a number of authors, especially (B2), which has been referred to in [26] as
feebly compact and attributed to S. Mardešić and P. Papić, and (B1), which
was called lightly compact in [1]. I. Glicksberg [8] noted that every pseudocom-
pact completely regular space satisfies (B2), and every space satisfying (B2) is
pseudocompact.

One immediate corollary to Theorem 1.1 that will be used below is the
following.

Corollary 1.2 ([8]). Let X be a topological space.

(a) The union of finitely many feebly compact subspaces of X is feebly
compact.

(b) If X is feebly compact and U is any open subset of X, then U is a
feebly compact subspace of X.



Spaces whose Pseudocompact Subspaces are Closed Subsets 245

(c) If D is a feebly compact subspace of X and D ⊆ G ⊆ D, then G is
feebly compact.

Definition 1.3. We shall call a topological space X feebly compact closed
(“FCC”) provided that X is feebly compact and every feebly compact subspace
of X is a closed subset of X.

Definition 1.4. A space X will be called sequentially feebly compact provided
that for every sequence {Un : n ∈ N} of nonempty open subsets of X there exist
a point p ∈ X and a strictly increasing sequence {ni : i ∈ N} in N such that
for every neighborhood V of p, V ∩ Uni 6= ∅ for all but finitely many i ∈ N.

2. The properties FCC and sequentially feebly compact

The property FCC has been studied previously (but not named or labeled)
by several authors. It was proved in [23] that every first countable feebly com-
pact Hausdorff space, and hence every first countable pseudocompact Tychonoff
space, is FCC. Then a proof was given in [14] that if a feebly compact space
X is E1, i.e., if every point x of X is an intersection of countably many closed
neighborhoods of x, then X is FCC. The concept has been used in the study
of maximal feeble compactness. By a maximal feebly compact space is meant a
feebly compact space (X, T ) such that for every feebly compact topology U on
X, if T ⊆ U then U = T . Using the result of D. Cameron [3], that an FCC,
submaximal space (i.e., a space in which every dense set is open) is maximal
feebly compact, and a result of A.B. Raha [17], the authors proved in [16] that
a topological space is maximal feebly compact space if and only if it is FCC
and submaximal. Using the latter, a number of examples of maximal feebly
compact spaces are given in [16], e.g., the well-known Isbell-Mrówka space Ψ
described in [6, 5I] and in the proof below of Theorem 2.12. The property FCC
was also considered in the article [9], where the relationship between count-
ably compact regular spaces which are FCC and those which are Fréchet was
studied.

The next lemma provides conditions each of which implies or is implied by,
or under suitable restrictions is equivalent to, the property FCC. Let us recall
that a space X is called semiregular provided that that the regular open sets
(i.e., sets having the form int(cl(A)), where A is an open subset of X) form a
base for the topology on X.

Lemma 2.1. Let X be a topological space, and consider the conditions below.

(F1) Every feebly compact subspace of X is a closed subset of X.
(F2) For every feebly compact subspace S of X, dense subset D of S, and

point p ∈ S\D, there exists a pairwise disjoint sequence K = {Kn : n ∈
N} of nonempty open subsets of D such that for every neighborhood V
of p in S, V ⊇ Kn for all but finitely many n ∈ N.

(F3) Every feebly compact subspace of X with dense interior is a closed
subset of X.



246 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

(F4) For every open subset S of X and point p ∈ S\S, there exists a pairwise
disjoint sequence K = {Kn : n ∈ N} of nonempty open subsets of S
such that K has no cluster point in X \{p}, and for every neighborhood
V of p in S, V ⊇ Kn for all but finitely many n ∈ N.

Then the following hold.

(a) Property (F1) implies (F2) and (F3), and if X is a Hausdorff space
then (F2) implies (F1).

(b) Property (F4) implies (F3), and if X is feebly compact then (F3) implies
(F4).

(c) If S is semiregular then in each of the statements (F2) and (F4), the
containment “V ⊇ Kn” may be replaced by “V ⊇ Kn.”

(d) If X is Fréchet, Hausdorff and scattered, then it has property (F1).
(e) If X is Fréchet and Hausdorff and has a dense set of isolated points,

then it has property (F3).

Proof. We prove (b). The proof of (a) is similar.
(F4) implies (F3). Suppose (F3) is false. Then there exist an open subset S

of X, a feebly compact subspace F of X with S ⊆ F ⊂ S, and a point p ∈ F \F .
It would follow that S = F and thus p ∈ S \ S. By Corollary 1.2 (c), the feeble
compactness of F , and the relation F ⊆ S \ {p} ⊆ F , the subspace S \ {p}
would be feebly compact. By Theorem 1.1, every sequence K = {Kn : n ∈ N}
of nonempty open subsets of X such that each Kn ⊂ S would have a cluster
point in S \ {p}. Therefore, (F4) would not hold.

Suppose X is feebly compact and (F3) holds. Let S and p be as in the
hypothesis of (F4). It follows from (F3) and the characterizations in Theorem
1.1 that there exists a pairwise disjoint sequence W of nonempty open sets
of the space S \ {p} such that W has no cluster point in S \ {p}. Define
K = {Kn : n ∈ N}, where for each n ∈ N, Kn = Wn ∩ S. Since S is dense in
S \ {p} and open in X, it follows from the properties of W that K is a pairwise
disjoint sequence of nonempty open subsets of X, as well as of S, which has
no cluster point in X \ {p}. By the feeble compactness of X, K must have
a cluster point, so p is the unique cluster point of K in X. If there were an
infinite subset J of N and a neighborhood V of p in S such that Kj \ V 6= ∅
for every j ∈ J, then {Kj \ V : j ∈ J} would be an infinite locally finite family
of open subsets of X, in contradiction of Theorem 1.1.

Statement (c) is obvious. Let us prove (d). The proof of (e) is similar.
Suppose Y ⊆ X is feebly compact and p ∈ Y . Let I be the set of isolated
points of the space Y . Then clY I = Y since X is scattered, and thus p ∈ I.
As X is Fréchet, there is a sequence {xn : n ∈ N} in I which converges to
p. Then {{xn} : n ∈ N} is a sequence of nonempty open sets of the feebly
compact Hausdorff space Y which has only p as a cluster point. Hence p ∈ Y .
Therefore, Y is a closed subset of X. �



Spaces whose Pseudocompact Subspaces are Closed Subsets 247

Theorem 2.2. Let X be a topological space. Then the following hold.

(a) If X is a feebly compact space which is either (i) E1, or (ii) compact
Hausdorff and either hereditarily metacompact or hereditarily realcom-
pact, or (iii) Fréchet, Hausdorff and scattered, then it is FCC.

(b) If X is FCC, then it is a feebly compact T1-space and has the properties
(F1)–(F4).

(c) If X is a countably compact FCC space, then it is (i) (Y. Tanaka)
Fréchet and (ii) sequentially compact.

(d) If X is feebly compact and either (i) has property (F3) or (ii) is a
sequential space, then X is sequentially feebly compact. In particular,
FCC implies sequentially feebly compact.

(e) If X is feebly compact, Fréchet and Hausdorff and has a dense set of
isolated points, then it has properties (F3)–(F4).

(f) If X is sequentially feebly compact, then it is feebly compact.

Proof. As noted above, part (i) of (a) is obtained in [14]. Since by results of
E. Hewitt [10] and S. Watson [28], every realcompact and every metacompact
pseudocompact Tychonoff space is compact, one obtains (ii) of (a). Statement
(iii) follows from Lemma 2.1 (d).

Obviously (b) holds. In [9] a proof was given that a statement like (c) (i)
holds for regular spaces, and the author of [9] attributed the result to Y. Tanaka.
Here is a similar proof that does not require regularity of the space X: Suppose
A ⊂ X and x ∈ A\A. Since A\{x} is not feebly compact and has A as a dense
subset, there exists a pairwise disjoint sequence U = {Un : n ∈ N} of nonempty
open subsets of A which has no cluster point in A \ {x}. Choose xn ∈ Un
for each n ∈ N. Then the set C = {xn : n ∈ N} is a discrete subspace of the
countably compact T1-space C = C ∪{x}, and consequently, the sequence {xn}
in A must converge to x. The statement (c) (ii) follows from the easily verified
fact that every countably compact T1 Fréchet space is sequentially compact.

We prove (d). Let U = {Un : n ∈ N} be a sequence of nonempty open
subsets of the space X. We wish to show that there exist a point p ∈ X and a
strictly increasing sequence {ni : i ∈ N} in N such that for every neighborhood
V of p, V ∩ Uni 6= ∅ for all but finitely many i ∈ N (or equivalently, there exist
a point p ∈ X and an infinite subset J of N such that V ∩ Uj 6= ∅ for all but
finitely many j ∈ J).

Suppose first that the hypothesis of (d) (i) holds. Let us consider two cases.
Case 1: suppose there are an infinite subset J of N and a point p ∈ X such

that p ∈ Uj for every j ∈ J. Then p and J have the required properties.
Case 2: suppose that Case 1 does not hold. Since X is feebly compact,

the sequence U has a cluster point p. There exists k ∈ N such that for every
integer n > k, the point p /∈ Un. Define S =

⋃
n≥k+1

Un, and for each i ≥

k + 1, let Si =
⋃i

n=k+1
Un. Note that p ∈ S \ S and Si ⊆ S \ {p} for every

i ≥ k + 1. As every feebly compact subspace of X with dense interior is
closed, it follows from Lemma 2.1 that there exists a pairwise disjoint sequence
K = {Kn : n ∈ N} of nonempty open subsets of X such that each Kn ⊆ S, K



248 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

has no cluster point in X \ {p}, and for every neighborhood V of p, V ⊇ Kn
for all but finitely many n ∈ N. Since each Si is feebly compact (by Corollary
1.2), then for each i ≥ k + 1, Si ∩ Kn 6= ∅ for at most finitely many n ∈ N. By
mathematical induction one can find strictly increasing sequences {mi : i ∈ N}
and {ti : i ∈ N} in N such that for each i ∈ N:

Kmi ∩ Uk+ti 6= ∅; and if i > 1, then Km ∩ Sti−1+k = ∅ for every m ≥ mi.

Define ni = k +ti for each i ∈ N. Then the sequence {ni : i ∈ N} and the point
p have the properties required in the definition of sequentially feebly compact.

Next, we assume the hypothesis of (d) (ii) holds. Consider again the two
cases named above. The proof in Case 1 proceeds as above.

Assume Case 2 holds. Then as in Case 2 above, there are a cluster point q
of U and and k ∈ N so that for every integer n > k, the point q /∈ Un. Then
the set T =

⋃
n≥k+1

Un is not a closed set since q ∈ T \ T . Because X is a
sequential space, it follows that there exists a sequence {xn : n ∈ N} in T which
converges to a point p ∈ X \ T . For each integer n ≥ k + 1, note that since
p /∈ Un then xm ∈ Un for at most finitely many m ∈ N. Thus, there are strictly
increasing sequences {mi : i ∈ N} and {ti : i ∈ N} such that for each i ∈ N, one
has xmi ∈ cl(Uk+ti ). Therefore, the sequence {ni = k + ti : i ∈ N} and point p
satisfy the definition of sequentially feebly compact.

Statement (e) follows from Lemma 2.1, and statement (f) follows from the
characterizations in Theorem 1.1 and the appropriate definitions. �

The next result will be used in §5.

Corollary 2.3. Let X be a feebly compact space which has property (F3).
Suppose U = {Un : n ∈ N} is a sequence of nonempty open subsets of X such
that Um ∩ Un = ∅ whenever m 6= n. Then there are a point p in X, an infinite
subset J of N, and a sequence of nonempty open sets P = {Pn : n ∈ J} such
that Pn ⊆ Un for each n ∈ J, and for every neighborhood O of p, O contains
Pn for all but finitely many n ∈ J.

Proof. This follows from the proof of Case 2 of statement (d) (i) in Theorem
2.2. �

Here are some examples illustrating that these properties are distinct.

Example 2.4. Let X be the one-point compactification of some uncountable
discrete space. Then X is a scattered, Fréchet, compact Hausdorff, and hence
FCC, space (by Theorem 2.2 (a) (iii)) which is not first countable (or E1).

Example 2.5. There exists a space X which is a countable, compact, maximal
feebly compact, and hence FCC, space which is not Hausdorff: in [16, 2.12] a
proof is given that a certain countable, non-Hausdorff, maximal compact space
due to V.K. Balachandran is also maximal feebly compact.

Example 2.6. Let X be any feebly compact Hausdorff space which contains
a non-isolated P-point p. Then X cannot have property (F3): if U = {Un : n ∈
N} were any pairwise disjoint sequence of nonempty open subsets of X \ {p}



Spaces whose Pseudocompact Subspaces are Closed Subsets 249

and one chose, for each n ∈ N, a nonempty open set Vn ⊆ Un with p /∈ Vn, then
the sequence {Vn : n ∈ N} would have a cluster point in X \ {p}, and hence U
would also, i.e., X \ {p} would be feebly compact. The next two spaces are of
this type.

Example 2.7. Let X = ω1 + 1, the set of ordinals ≤ ω1, with the order
topology. The space X is a sequentially compact, and hence sequentially feebly
compact, compact Hausdorff space which does not have property (F3).

Example 2.8. Let T be the Tychonoff plank, T = (ω1+1)×(ω0+1)\{(ω1, ω0)}.
Then T is a locally compact Hausdorff space that does not have property (F3)
and is not sequentially compact [5], [6]. Since T has a dense, sequentially
compact subspace, namely T \ {(ω1, α) : α < ω0}, it follows easily that T is
sequentially feebly compact.

Example 2.9. Let βN be the Stone-Čech compactification of N, where N has
the discrete topology, and let X be any dense, pseudocompact subspace of βN.
Then X is a feebly compact Tychonoff space that is not sequentially feebly
compact, and hence is not FCC: it is well-known that no nontrivial sequence
in βN is convergent [6], and so for any infinite subset J of N and sequence
U = {{j} : j ∈ J}, there would exist no point p ∈ X and infinite subset K
of J with every neighborhood of p containing all but finitely many of the sets
{{k} : k ∈ K}.

Example 2.10. Let X = N ∪ {−∞, ∞}, where a subset T of X is defined to
be open iff T ⊆ N or X \ T is finite. Then X is a first countable, scattered,
compact T1-space satisfying property (F2), but none of (F1), (F3) and (F4).

The properties first countable, E1, Fréchet and sequential are well-known
to be closely related to one another. We shall give some examples illustrating
further similarities and, in some cases, differences between these properties and
the properties FCC and sequentially feebly compact. One such is the following
familiar space.

Example 2.11. Let (X, S) be [0,1], with its usual topology, and let T be the
topology on X generated by S and the family of co-countable subsets of X.
Then (X, T ) is an FCC space which is E1, but not a sequential space. The latter
follows from the fact that no infinite subset of (X, T ) is countably compact. It
is also known that for every set T ∈ T , clT T = clST , and consequently every
open filter base on (X, T ) has an adherent point. Thus (X, T ) has the stated
properties.

Example 2.11 also illustrates the observation that for any FCC space (X, S),
if T is any feebly compact topology on X such that S ⊆ T , then (X, T ) is also
FCC. More generally, see Theorem 3.2 (d) below.

A previously defined family of spaces related to FCC spaces was studied in
[11], where M. Ismail and P. Nyikos called a space X C-closed if every countably
compact subspace of X is a closed subset of X. They proved that (a) a se-
quential Hausdorff space is C-closed, and (b) a sequentially compact, C-closed



250 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

Hausdorff space is a sequential space. In their statement (b), if one replaces “se-
quentially compact, C-closed Hausdorff” by “countably compact FCC,” then
(as noted above in 2.2 (c)), one can replace their conclusion by “Fréchet and
sequentially compact.” The next result shows that in (a), even for feebly com-
pact symmetrizable spaces, one cannot replace “C-closed” by “FCC.” (A space
(X, T ) is called symmetrizable in the sense of A.V. Arhangel′skĭı if there exists
a symmetric d on X which induces T , where by a symmetric on X one means
a function d : X × X → [0, ∞) which vanishes exactly on the diagonal and
satisfies the symmetric property, d(x, y) = d(y, x) for all x, y ∈ X.)

Before stating the result, let us first recall that an almost disjoint (“AD”)
family P on a set X is a collection P ⊆ [X]ω such that P ∩P ′ is finite whenever
P , P ′ are distinct members of P. An AD family M on X (such that M ⊆ Q ⊆
[X]ω) is called a maximal almost disjoint family (“MAD” family) (respectively,
maximal almost disjoint subfamily of Q) provided that M is properly contained
in no AD family on X (respectively, no AD subfamily of Q).

Theorem 2.12. There exists a symmetrizable (therefore, sequential), scat-
tered, C-closed Hausdorff space (X, T ) which contains a non-isolated point p
such that X \ {p} is first countable, locally compact, feebly compact and zero-
dimensional, and hence X is sequentially feebly compact and C-closed but not
FCC.

Proof. Let Ψ be the Isbell-Mrówka space described in [6, 5I]: Let M be an
infinite MAD family on N and Ψ = N∪M, where a subset U of Ψ is defined to
be open provided that for any set M ∈ M, if M ∈ U then there is a finite subset
F of M such that {M} ∪ M \ F ⊆ U. The space Ψ is then a first countable
pseudocompact locally compact Hausdorff space that is not countably compact
[6].

List in a 1-1 manner as {Mn : n ∈ N} the members of an infinite subset
I of M, choose a point p /∈ Ψ, and define X = Ψ ∪ {p}. Next, define d :
X × X → [0, ∞) as follows: d(p, Mn) = d(Mn, p) = 1/n for each Mn ∈ I and
d(p, y) = d(y, p) = 1 for each y ∈ N ∪ (M \ I); for each n ∈ N and y ∈ Ψ \ {n},
d(n, y) = d(y, n) = 1/n whenever n ∈ y ∈ M, and d(n, y) = d(y, n) = 1
whenever either y ∈ M with n /∈ y or y ∈ N \ {n}; and d(x, x) = 0 for all
x ∈ X. Let T be the topology induced on X by d, i.e., define T to be the
collection of all subsets T of X such that for each point t ∈ T there exists ǫ > 0
such that T contains the “ball” {x ∈ X : d(t, x) < ǫ}.

It is straightforward to show that d is a symmetric for the space (X, T ), and
(X, T ) has the stated properties. Furthermore, it is known and not difficult to
prove that every symmetrizable space is sequential. �

Example 2.13. If one lets X be as in the proof above, but weakens the
topology T on X by choosing the topology S for which (X, S) is the one-point
compactification of Ψ, then it was noted in [12] that Eric van Douwen and
Peter Nyikos had noticed previously the resulting compact Hausdorff space
(X, S) was sequential but not Fréchet. Like (X, T ), the space (X, S) is not
FCC, and since Ψ ∈ S, these spaces do not even have the property (F3). Since



Spaces whose Pseudocompact Subspaces are Closed Subsets 251

every symmetrizable compact Hausdorff space is known to be metrizable, and
every scattered Fréchet feebly compact space is FCC, then S 6= T and (X, T )
is not Fréchet either. While (X, T ) fails to be countably compact, the space
(X, S) is known to be sequentially compact and C-closed.

It is natural to ask if the word “scattered” can be removed from the state-
ment in Theorem 2.2 (a) (iii). In [9] a very nice proof was given that, assuming
[MA], there exists a compact Hausdorff Fréchet space which is not FCC.

The next result, which does not require any special axioms beyond ZFC,
shows that there is a compact Hausdorff Fréchet space X which is not FCC.
Its proof is an elaboration on one due to Reznichenko that was outlined in 3.6
of [12]. The authors are grateful to Peter Nyikos for calling Reznichenko’s space
to our attention. In addition, we shall show that X can be used to construct a
compact Hausdorff and Fréchet space A(X) which is not FCC, but which has
property (F3) and also has a dense set of isolated points.

Theorem 2.14. There is a compact Hausdorff Fréchet space which is not FCC.

Proof. Let κ denote the cardinality of continuum.
We first define a compact 0-dimensional Fréchet topology on T = κ≤ω, i.e.,

T consists of the functions into κ which have domain either a nonnegative
integer or the entire set of nonnegative integers.

For any t ∈ κ<ω and α ∈ κ, we will let tα denote the function obtained by
extending the domain of t by one and setting the final value to α. For n ∈ ω
and t : n → κ, we occasionally denote t by 〈t0, . . . , tn−1〉.

Recall that T forms a tree when ordered by simple inclusion, i.e., for s, t ∈ T ,
s ⊆ t if dom(s) ≤ dom(t) and s = t ↾ dom(s). Now T is endowed with the
following topology. Simply for each s ∈ κ<ω, the set

[s] = {t ∈ T : s ⊆ t}

is defined to be clopen. Thus a neighborhood basis for s ∈ κ<ω is the family

{[s] \
⋃

i<n

[sαi] : n ∈ ω and α0 < α1 < · · · < αn−1 < κ} .

Furthermore, for f ∈ T ∩ κω, the family {[f ↾ n] : n ∈ ω} forms a neigh-
borhood base at f, and hence each such f is a point of countable character in
T .

We leave as an exercise that T is compact, and thus for each s ∈ κ<ω, the
clopen set [s] is compact. One can note that {[sα] : α ∈ κ}, is a pairwise-disjoint
family of clopen sets, and [s] is the one-point compactification of

⋃
{[sα] : α ∈

κ}. It follows easily then that T is a Fréchet space.
Next, for each n ∈ ω, let Tn denote the clopen set [〈n〉], i.e., all functions

t ∈ T such that t(0) = n. We will construct a compactification, X, of
⋃

n
Tn.

Our base space for X will be
⋃

n
Tn ∪ κ ∪ {∞}. We will define a locally

compact topology on
⋃

n
Tn ∪ κ in which

⋃
n
Tn with the above topology is

open and dense, and X will just be the one-point compactification. We will



252 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

work to ensure that X \ {∞} is feebly compact, thus ensuring that X is not
FCC.

For each α ∈ κ, we will select a sequence 〈tαn : n ∈ ω〉 such that for each n,
tαn is a member of Tn ∩ κ

<ω and α is in its range. The neighborhood basis for
α will be

{U(α, n) = {α} ∪
⋃

n<k

[tαk ] : n ∈ ω},

and hence α will be the point at infinity in the one-point compactification of
the union of the sequence of clopen sets {[tαk ] : k ∈ ω}.

In order to ensure that the space is Hausdorff we will make sure that for
β < α < κ, there will be an n such that U(β, n) ∩ U(α, n) is empty. This is
equivalent to requiring that for each m larger than this n, [tβm] and [t

α
m] are

disjoint, i.e., tβm and t
α
m are incomparable members of T (or Tm).

The sequences are chosen by induction on κ. In order to ensure that X\{∞}
is feebly compact, it suffices for us to require that for every infinite set I ⊂ ω
and every sequence {sn : n ∈ I} such that sn ∈ Tn ∩(κ

<ω) for n ∈ I, there is an
α ∈ κ such that α is a cluster point of the sequence of clopen sets {[sn] : n ∈ I}.
To do so, let {{sαn : n ∈ Iα} : α ∈ κ} enumerate the family of all such sequences.

The selection of the sequence {t0n : n ∈ ω} is handled the same as that for
any α. That is, assume that α < κ and that for each β < α we have chosen
the sequence {tβn : n ∈ ω} as described above (so that t

β
n ∈ Tn and β is in

the range of tβn). We therefore have defined, as above, a topology on the space
Xα =

⋃
n
Tn ∪ {β : β < α} with the neighborhood base {U(β, n) : n ∈ ω} for

each β < α.
Fix any γ < κ so large that for each β < α and n ∈ ω, γ is not in the

range of tβn. Observe that the sequence {[〈n, γ〉] : n ∈ ω} is a discrete sequence
of clopen sets in the space Xα. In fact, for each β < α, U(β, 0) is disjoint
from each member of the sequence. If the sequence {[sαn] : n ∈ Iα} already
has a cluster point in the space Xα, we define t

α
n to be 〈n, γ, α〉 for each n.

Otherwise, the sequence {[sαn] : n ∈ Iα} is also discrete (hence each U(β, 0)
meets only finitely many of these sets), and we define tαn to be s

α
n α for each

n ∈ Iα and set t
α
n = 〈n, γ, α〉 for n /∈ Iα. It follows easily that for each β < α,

there is an n such that U(β, n) ∩ U(α, 0) is empty.
This completes our construction of the space. It should be clear that the

space Xκ =
⋃
{Xα : α < κ} is locally compact, Hausdorff and feebly compact.

Furthermore, Xκ is easily seen to be Fréchet, for Xκ is first countable at each
α ∈ κ, and its open subspace Xκ \ κ =

⋃
n
Tn is also a subspace of the Fréchet

space T and hence is Fréchet. To finish the proof, we verify that the one-point
compactification X = Xκ ∪ {∞} of Xκ is Fréchet.

Assume that Y ⊆ Xκ does not have compact closure in Xκ. We wish to
show that there is a sequence {yn : n ∈ ω} ⊆ Y which converges to ∞. Since
{α : α ∈ κ} is a closed discrete subset of Xκ, we may assume that Y is contained
in

⋃
n
Tn. Two cases are considered.

Case 1: Suppose Y has only finitely many limit points in κ. Then one may
intersect Y with a neighborhood of ∞ which does not have any of those points



Spaces whose Pseudocompact Subspaces are Closed Subsets 253

in its closure, and obtain a subset Y ′ of Y which still has ∞ in its closure
but has no limit points in κ. Any sequence {yn : n ∈ ω} ⊆ Y

′ such that
yn ∈ Y

′ \
⋃

k<n
Tk will have no cluster point in Xκ and hence will converge to

∞.
Case 2: Suppose there is an infinite countable set A ⊂ κ which is contained

in the closure of Y . Since Xκ is Fréchet and A is countable, there is a countable
Y ′ ⊆ Y whose closure contains A. It is easily seen that one can also assume
Y ′ ∩ Tn is finite for each n. Since each member of Y is really a function with
range contained in κ, we can let B denote the set of all β ∈ κ such that β is
in the range of some member y of Y ′. If γ is any member of κ \ B, we claim
that U(γ, 0) ∩ Y ′ is empty. Indeed, if y ∈ Y ′ ∩ U(γ, 0), then there is an n such
that y ∈ [tγn], from which it follows that t

γ
n ⊆ y. However, γ is in the range of

tγn but not in the range of y, a contradiction. It follows now that the closure of
Y ′ in Xκ is a countable non-compact set and therefore is not feebly compact.
Thus its dense subspace Y ′ contains a sequence having no cluster point in Xκ,
and that sequence must converge to ∞. �

Before presenting the last example of this section, a lemma is needed.

Lemma 2.15. If X is Fréchet and Hausdorff, then the Alexandroff double,
A(X), is Fréchet and Hausdorff, and A(X) has property (F3).

Proof. It suffices to prove that the double of a Fréchet space is Fréchet (which
is probably well-known), for by Lemma 2.1 (e) we already know that a Fréchet
Hausdorff space with a dense set of isolated points has property (F3), and it
is well-known that the double of a Hausdorff space is Hausdorff. But let us
check. Let A(X) be the usual X × {0, 1} with X × {0} open and discrete,
and neighborhood base for (x, 1) be the usual (U × {0, 1}) \ {(x, 0)} for open
U ⊆ X containing x. If (x, 1) is in the closure of A ⊆ A(X) then clearly there is
a subsequence of A converging to (x, 1) if (x, 1) is in the closure of A∩(X ×{1})
since this subspace is homeomorphic to X. But just as easily we see that if
A ⊆ X ×{0}, then there is some A′ ⊆ X such that A = A′ ×{0} and x is a limit
of A′. Any subsequence of A′ which converges to x will yield a corresponding
subsequence of A which converges to (x, 1). �

Theorem 2.16. There is a compact Hausdorff Fréchet space which is not FCC,
but which has a dense set of isolated points and hence has property (F3).

Proof. Just take the space X as constructed in 2.14 and apply 2.15 on A(X).
�

3. Subspaces and images

We consider next whether these properties are or can be inherited by sub-
spaces, or preserved or reflected by continuous maps.

Some inheritance results hold that are analogous to the well-known ones
concerning feeble, countable or sequential compactness. These are stated next.



254 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

Theorem 3.1. Let X be a topological space.

(a) If X is FCC and A is a feebly compact subspace of X, then A is FCC.
(b) If X is FCC (respectively, sequentially feebly compact) and A is an open

subset of X, then A is FCC (respectively, sequentially feebly compact).
(c) If X has a dense, sequentially feebly compact subspace, then X is se-

quentially feebly compact.

The Tychonoff plank witnesses that sequential feeble compactness, as well
as feeble compactness, is not closed hereditary. If every countable closed subset
of a space X is pseudocompact (and the space X is T1), then every countably
infinite subset of X has a limit point (and X is countably compact).

Before stating some mapping theorems, we recall that a mapping f : X → Y
is called Z-closed (closed) provided that for every zero-set F (closed subset F)
of X, f(F) is a closed subset of Y . In case a mapping f : X → Y satisfies f(F)
is a proper closed subset of Y for every proper closed subset F of X, then f is
called irreducible. A closed mapping f : X → Y is called perfect (quasi-perfect)
provided that for every point y ∈ Y , the fiber f−1(y) is compact (countably
compact).

Some known theorems are these ([5], [20]): (a) if f ∈ C(X, Y ) and X is
feebly compact (respectively, sequentially compact, countably compact) then
so is f(X); (b) if f : X → Y is a closed mapping, and Y and each f−1(y),
y ∈ Y , are countably compact, then X is countably compact; and (c) if f is a
Z-closed open mapping of a Tychonoff space X onto a Tychonoff space Y , and
if Y and each f−1(y), y ∈ Y , are pseudocompact, then so is X. The following
also hold.

Theorem 3.2. Let X and Y be topological spaces and f : X → Y a mapping.

(a) If f ∈ C(X, Y ) and X is sequentially feebly compact, then f(X) is
sequentially feebly compact.

(b) If f is an open and Z-closed mapping of X onto Y , and Y and each
fiber f−1(y), y ∈ Y , are pseudocompact, then X is pseudocompact.

(c) If f is an irreducible quasi-perfect mapping of X onto Y , and Y is
feebly compact, then X is feebly compact.

(d) If f is a perfect mapping of X onto Y , and every feebly compact sub-
space of X is a closed subset of X, then every feebly compact subspace
of Y is a closed subset of Y .

(e) If X is FCC and f ∈ C(X, Y ) is a perfect mapping of X onto Y , then
Y is FCC.

(f) If X is feebly compact and semiregular, Y is FCC, and f ∈ C(X, Y ) is
a bijection, then f is a homeomorphism.

Proof. The proof of (a) is immediate. To verify (b), one can use a Urysohn’s
lemma-type of argument similar to the proof that establishes (b) for Tychonoff
spaces.

(c). Suppose that the hypotheses in (c) hold. By Theorem 1.1, it suffices for
us to prove that every countable open cover U of X has a finite subcollection



Spaces whose Pseudocompact Subspaces are Closed Subsets 255

whose union is dense in X. Let U be a countable open cover of X such that for
every finite subcollection G of U,

⋃
G ∈ U. Then, by the countable compactness

of each fiber f−1(y), it follows that for each y ∈ Y , there exists U ∈ U such that
f−1(y) ⊆ U. Hence {Y \ f(X \ U) : U ∈ U} is a countable open cover of Y . By
the feeble compactness of Y , there is a finite subcollection {Ui : i = 1, . . . , n}

such that Y =
⋃n

i=1 Y \ f(X \ Ui). Let W =
⋃n

i=1 f
−1(Y \ f(X \ Ui)). Since

f is closed, f(W) ⊇ f(W). Because f is onto, f(W) =
⋃n

i=1
Y \ f(X \ Ui).

Thus f(W) =
⋃n

i=1
Y \ f(X \ Ui) = Y . As f is irreducible, the latter implies

X = W . Since each f−1(Y \ f(X \ Ui)) ⊆ Ui, then X =
⋃n

i=1
Ui.

(d). Suppose the hypothesis of (d) holds and S is a feebly compact subspace
of Y . Then f|f−1(S) : f−1(S) → S is a perfect surjection (e.g., see [15, 1.8
(f) (2)]). By Zorn’s lemma [15, 6.5 (c)], there is a closed subset A of the space
f−1(S) such that f|A : A → S is a perfect irreducible surjection. Thus by (c)
above, A is feebly compact. Therefore, A is a closed subset of X, and since f
is a closed mapping, then S = f(A) must be a closed subset of Y .

(e). This is an immediate consequence of (d) and the fact that feeble com-
pactness is preserved by continuous mappings.

(f). Suppose the hypothesis holds and F is any closed subset of the space X.
By the semiregularity of X, the family R of regular closed subsets of X which
contain F satisfies F =

⋂
R. Each set R ∈ R is a feebly compact subspace of

X by Corollary 1.2 (b). Thus the continuous image f(R) of each such R under
the mapping f is a feebly compact, hence closed, subset of the space Y . As f
is one-to-one, f(F) =

⋂
{f(R) : R ∈ R}, and hence f(F) is a closed subset of

Y . Therefore, f is a homeomorphism. �

One interesting corollary to 3.2 (f) is the following.

Corollary 3.3. Let S and T be topologies on a set X such that (X, T ) is a
feebly compact semiregular space, (X, S) is FCC, and S ⊆ T . Then T = S.

The next example illustrates that the condition “irreducible” cannot be re-
moved from the hypothesis of (c) in Theorem 3.2.

Example 3.4. Let Ψ be the Isbell-Mrówka space described in the proof of
Theorem 2.12. Let N− be the set of the negative integers, with the discrete
topology, and let X be the discrete union of Ψ and N−. List in a 1-1 manner as
{Mn : n ∈ N} the members of an infinite subset of M, and define f : X → Ψ
by the rule: f(x) = x if x ∈ Ψ, and f(x) = M−x if x ∈ N

−. Then X is not
pseudocompact, Ψ is FCC, and f : X → Ψ is a closed, continuous map of X
onto Ψ, each of whose fibers is finite.

Example 3.5. This example shows that if f : X → Y is a perfect continuous
surjection and Y is FCC, then X need not be FCC. Moreover, it shows that if
Y is a first countable compact Hausdorff (hence sequentially compact) space,
and f is an irreducible, perfect continuous surjection, then X need not be
sequentially feebly compact. Let X = βN and Y = N∞ be the one-point
compactification of N, where N has the discrete topology. Let f be the Čech



256 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

mapping in C(βN, N∞) which extends the identity mapping on N. Then Y
has the stated properties, and as noted in Example 2.9, βN is not sequentially
feebly compact. It follows from 6.11 of [6] that f(βN \ N) = {∞}, and hence f
is irreducible, as well as perfect.

4. Product spaces

In [8] and [20] it was shown that a number of the properties considered
there are well behaved in the formation of feebly compact product spaces,
namely: sequentially compact; feebly compact and first countable; and feebly
compact and locally compact. We shall show that the property sequentially
feebly compact likewise is well behaved in the formation of feebly compact
product spaces. As was done in [8] and [20] for the product theorem proofs
presented in those articles, in several proof outlines below there will be no
loss of generality for us to assume that the sets in Theorem 1.1 (B3) are the
standard basic open sets for the product topology.

Theorem 4.1. The property sequentially feebly compact is productive.

Proof. Let X =
∏

a∈A
Xa be a product of sequentially feebly compact spaces.

Let U = {Un : n ∈ N} be a sequence of nonempty basic open sets of X.
The set R = {a ∈ A : pra(Un) 6= Xa, for some n ∈ N} of restricted coordi-

nates of the members of U is countable. Choose the smallest ordinal number
λ such that λ = |R|. For notational simplicity, and with no loss of generality,
we may assume that R = {α : α < λ} and R 6= ∅. Since each factor space is
sequentially feebly compact, we may use mathematical induction and choose
points {pα ∈ Xα : α < λ} and infinite subsets {Iα ⊆ N : α < λ} such that:
whenever α < γ < λ then Iα ⊇ Iγ; and for each α < λ and neighborhood V
of pα, V ∩ prα(Un) 6= ∅ for all but finitely many n ∈ Iα. For each a ∈ A \ R,
choose one point pa ∈ Xa.

Next, we define an infinite set I ⊆ N as follows: if λ is finite, set I = Iλ−1;
if λ = ω0, then select an infinite subset I of N such that for each α < λ, I \ Iα
is finite. In either case, let {ni : i ∈ N} be a strictly increasing mapping of N
onto I.

Then for every neighborhood V of p in X, V ∩ Uni 6= ∅ for all but finitely
many i ∈ N. Therefore, X is sequentially feebly compact. �

The next theorem extends an analogous theorem obtained by A.H. Stone
for first countable spaces—see [20]. First a lemma is given.

Lemma 4.2. Suppose that X is sequentially feebly compact and Y is feebly
compact. Then X × Y is feebly compact.

Proof. Let U = {Un × Wn : n ∈ N} be a sequence of nonempty open sets in
X ×Y . Since X is sequentially feebly compact, there exist p ∈ X and a strictly
increasing sequence {ni : i ∈ N} in N such that for every neighborhood V of p,
V ∩ Uni 6= ∅ for all but finitely many i ∈ N. Since Y is feebly compact, the
sequence {Wni : i ∈ N} has a cluster point q ∈ Y . Then (p, q) is a cluster point
of {Uni × Wni : i ∈ N} and hence also of U. �



Spaces whose Pseudocompact Subspaces are Closed Subsets 257

Theorem 4.3. Every product of feebly compact spaces, all but one of which
are sequentially feebly compact, is feebly compact.

Since every FCC space is sequentially feebly compact, and every pseudo-
compact completely regular space is feebly compact, there are applications of
the preceding product results to FCC spaces and pseudocompact completely
regular spaces.

Because each factor of a product space is a continuous image of the product
space and is homeomorphic to a subspace of the product space, the next result
follows from Theorem 3.2 (a).

Theorem 4.4. If a product space X is FCC (respectively, sequentially feebly
compact), then so is every factor space of X.

We consider next what can be said about products of FCC spaces. One
obvious consequence of Theorem 2.2 (a) is the following.

Theorem 4.5. Let X =
∏

a∈A
Xa be a product of FCC spaces.

(a) If A is countable and for each a ∈ A, Xa is an E1-space, then X is
FCC.

(b) If A is finite, X is Fréchet, and for each a ∈ A, Xa is Hausdorff and
scattered, then X is FCC.

A simple example, however, shows that in general X =
∏

a∈A
Xa is never

FCC if |A| ≥ ℵ1.

Example 4.6. Let A be any set with |A| ≥ ℵ1, D = {0, 1} have the discrete
topology, and X be the product space DA. Let C be the Corson Σ-subspace of
X based at 0, i.e., let C = {x ∈ X : xa = 0 for all but countably many a ∈ A}.
Then X is a product of first countable compact Hausdorff spaces. But it is
straightforward to show (and known) that C is a countably compact, proper
dense subspace of X. Therefore, X is not FCC.

Corollary 4.7. Let X =
∏

a∈A
Xa, where |A| ≥ ℵ1, and for each a ∈ A,

|Xa| ≥ 2. Then X is not FCC.

Example 4.8. Let A, D and X be as in Example 4.6, but require that |A| ≥
2ℵ0. Then the product space X is sequentially feebly compact by Theorem 4.1.
As noted above, it is not FCC. In [20] a proof was given that X fails to be
sequentially compact.

While Example 4.6 does not answer the question as to whether or not the
property FCC is countably productive, one can use theorems of V.I. Malykhin
and a theorem of P. Simon (see [21]) to do so. It is shown below that there
exist two FCC spaces whose product is not FCC, and furthermore each of those
spaces can be chosen to be compact, Fréchet, scattered and Hausdorff. In order
to develop this answer, some terminology and notation to be used are given
next.

Let P be an AD family on N. As in [21], we shall define F(P) to be the
space previously introduced by S.P. Franklin, the set N ∪ P ∪ {∞}, where



258 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

∞ /∈ N ∪ P, topologized as follows: each n ∈ N is isolated; a basic open
neighborhood of a point P ∈ P is {P} ∪ C, where C is any cofinite subset of
P ; and F(P) is the one-point compactification of N ∪ P. The family P will
be called nowhere infinitely MAD provided that for every infinite subset X of
N, the set {X ∩ P : X ∩ P is infinite and P ∈ P} fails to be an infinite MAD
family on X. Note that a MAD family (say on N) is always a MAD family
when restricted to a cofinite subset of N.

In [21] Simon attributed to Malykhin two results which can be stated as
follows. The space F(P) is Fréchet iff P is nowhere infinitely MAD. If P =
P1 ∪ P2 is AD on N and P1 ∩ P2 = ∅, then the product space F(P1) × F(P2)
is not Fréchet iff P is not nowhere infinitely MAD. Then Simon proved that
there exists an infinite MAD family P on N having a partition P = P0 ∪ P1
such that for each i, i = 0, 1, Pi is nowhere infinitely MAD. We shall call such
a partitioned family a Simon MAD family. Simon used his theorem to prove
that the product of two Fréchet compact Hausdorff spaces need not be Fréchet.
By Tanaka’s theorem (Theorem 2.2 (c) above), such a product would then not
be FCC. Moreover, one can prove the following.

Theorem 4.9. Let P = P0 ∪ P1 be a Simon MAD family. Then the compact
Hausdorff Fréchet spaces F(Pi), i = 0, 1, are scattered and hence are FCC
spaces, but their product F(P0) × F(P1) does not have property (F3).

Proof. Obviously any Franklin space is scattered, so each F(Pi) is FCC by
Theorem 2.2 (a) (iii). The proof that F(P0) × F(P1) does not have property
(F3) is similar to the proof that this space is not Fréchet. Let D = {(n, n) :
n ∈ N}. We show that (i) the point p = (∞0, ∞1) is in D and (ii) the set
F = D \ {p} is feebly compact.

(i). Let U0 × U1 be a basic open neighborhood of p. For each i, i = 0, 1,
F(Pi)\Ui is compact, and so there exist a finite subset Ni of the closed discrete
subset Pi of the space F(Pi) \ {∞i} and finite subsets Ci and Ni of N such
that F(Pi) \ Ui ⊆ Ni ∪ ((

⋃
Ni) \ Ni) ∪ Ci. Choose any set P ∈ P \ (N0 ∪ N1),

which we may do since P is infinite. Because P is AD, some cofinite subset C
of P satisfies C ∩(Ni ∪((

⋃
Ni)\ Ni)∪Ci) = ∅ for i = 0, 1. Thus C ⊆ U0 ∩U1,

and so for any n ∈ C, (n, n) ∈ U0 × U1. Therefore D ∩ (U0 × U1) 6= ∅.
(ii). Since each point of D is isolated and D is dense in F , it suffices to prove

that every infinite subset of D has a limit point in (F(P0)×F(P1))\{p}. Let I
be an infinite subset of D. Define X = {n ∈ N : (n, n) ∈ I}. Since P is MAD,
there exists P ∈ P such that X ∩ P is infinite. Then every neighborhood of
the point P contains all but finitely many integers in X ∩ P . Either P ∈ P0
or P ∈ P1. Suppose P ∈ P0. Since F(P1) is compact, the infinite set X ∩ P
has a limit point y ∈ F(P1). Thus the point (P, y) is a limit point of I. (In
fact, since P /∈ P1, one can show that y = ∞1, and any 1-1 listing of the
members of {(n, n) ∈ I : n ∈ X ∩ P} defines a sequence in D which converges
to (P, ∞1).) �



Spaces whose Pseudocompact Subspaces are Closed Subsets 259

Corollary 4.10. There exists a compact Hausdorff, scattered, Fréchet, and
hence FCC, space X whose product with itself, X2, does not have property
(F3).

Proof. Let X be the discrete union of the spaces F(P0) and F(P1) in Theorem
4.9. �

5. Extension spaces

A space E is called an extension space of a space X if X is a dense subspace
of E. We wish to examine necessary and sufficient conditions that a space
X have an FCC or sequentially feebly compact extension space E, where E
may be required to have other properties, such as complete regularity. Some
embedding theorems will be given, and it will be shown that there exist Moore
spaces, neither of which has a regular FCC extension space, and one of which
is separable and has no regular sequentially feebly compact T1-extension space.
A feebly compact extension space E of a space X is sometimes called a feeble
compactification of X.

The following links some of these concepts and maximal and minimal P-
spaces (for various properties P) and shows that: as far as the underlying
sets are concerned, an FCC extension space of a space X is a minimal feeble
compactification of X; and from the point of view of topological properties, a
semiregular FCC extension space E of a space X is minimal with respect to
being an FCC extension space of X and is maximal with respect to being a
semiregular feeble compactification of X.

Theorem 5.1. Let (X, U) be a space, and suppose (E, S) and (G, T ) are feeble
compactifications of (X, U) such that (E, S) is FCC, where X ⊆ G ⊆ E and
S|G ⊆ T . Then the following hold.

(a) G = E.
(b) If (G, T ) is semiregular then T = S.

Proof. For statement (a), note that because (G, S|G) is a continuous image of
(G, T ), the space (G, S|G) is feebly compact, and hence G is a closed subset
of (E, S). But X ⊆ G, and so G is a dense subset of (E, S). Thus G = E.
Statement (b) follows from (a) and Corollary 3.3. �

Before stating the next result, some notation and terminology is needed.
Recall that a filter base F on a space is called free iff the adherence of F is
empty. Given a space X and a family M of free open filter bases on X, we
shall denote by XM the set X ∪ M, topologized as follows: a set G ⊆ XM
is defined to be open iff (i) G ∩ X is open in X, and (ii) if F ∈ G ∩ M then
G ∩ X contains some member of F. For a space Y, we shall denote by sY the
semiregularization of Y , the semiregular space whose points are the same as
those of Y , and whose topology has as a base the regular open subsets of Y
(for a derivation of properties of sY , see [15]).



260 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

Theorem 5.2. Let X be a topological space and M a maximal family of count-
able, free, open filter bases on X such that whenever F and G are distinct mem-
bers of M then there exist F ∈ F and G ∈ G such that F ∩ G = ∅. Then the
following hold.

(a) X is a dense open subspace of XM, XM is a feeble compactification
of X, and each point of XM \ X is of countable character in XM and
is an E1-point of XM, i.e., is an intersection of countably many of its
closed neighborhoods.

(b) If X ⊆ G ⊆ XM is feebly compact then G = XM.
(c) The space X is sequential (Fréchet, first countable, Hausdorff) iff XM

has the same property, and in case X is sequential (first countable and
Hausdorff), then XM is sequentially feebly compact (first countable and
Hausdorff, and hence FCC).

(d) X is semiregular and first countable iff sXM is a semiregular first
countable extension space of X, and in this case, sXM is FCC if X is
Hausdorff.

(e) If X is scattered (has a dense set of isolated points), Fréchet and Haus-
dorff, then XM is FCC (has property (F3)).

The proof of 5.2 is straightforward, and some of its statements are similar
to known results: (a), (b) and (d) are extensions of results in [23]; the last
statement in (c) follows from 2.2 (d) and the preceding statements in (c), each
of which is a special case of a known result.

If one seeks conditions on a space X that it have an FCC extension space
having other desirable properties, such as regular, completely regular, zero-
dimensional or Moore, some results obtained previously that have applications
to these questions are the next four theorems.

Theorem 5.3.

(a) ([24]) Every locally feebly compact, first countable zero-dimensional
T1-space has a feebly compact, first countable zero-dimensional T1-
extension space.

(b) ([18]) Every locally compact Moore space which is zero-dimensional at
each point of a countable dense subset has a locally compact, feebly
compact Moore extension space.

Theorem 5.4 ([27]).

(a) Every locally pseudocompact (locally compact), first countable Tychonoff
space X has a pseudocompact, first countable Tychonoff extension space
E such that E is locally compact if X is.

(b) Every metrizable space has a pseudocompact first countable Tychonoff
extension space.

Theorem 5.5 ([13]). Let X be a separable, locally pseudocompact Tychonoff
(locally compact) Moore space. Then X can be embedded densely in a pseudo-
compact Tychonoff (locally compact) Moore space.



Spaces whose Pseudocompact Subspaces are Closed Subsets 261

Theorem 5.6 ([22]).

(a) Every locally feebly compact regular T1-space X can be embedded as an
open dense subspace in a feebly compact regular T1-space Y which is
first countable at every point of Y \ X.

(b) Every separable, locally feebly compact (locally pseudocompact, Tychonoff)
Moore space embeds as an open dense set in a feebly compact (pseudo-
compact Tychonoff) Moore space.

Theorems 5.5 and 5.6 answered questions raised in [17], [18] and [24] (While
made available to others, Theorem 5.5 and its proof have not been published yet
by P. Nyikos.) Simon and Tironi’s very nice Theorem 5.6 (a) implies that every
first countable, locally feebly compact, regular T1-space has a feebly compact,
first countable, regular T1, and hence FCC, extension space.

Remark 5.7. For converse properties, we make some observations. If a space
X has an FCC extension space E, then X must have property (F1), and if X is
locally feebly compact then it is an open subset of E. If X is an open subspace
of some feebly compact regular space (some space having property (F3)), then
X is locally feebly compact (has property (F3)). As noted in [25] (see also
[19]), every dense subspace of a feebly compact Moore space is separable and
has a dense metrizable subspace.

An example due to T. Terada and J. Terasawa (which was a modification of
an example due to E. van Douwen and T.C. Przymusiński) was given in [27] to
prove that there is a first countable zero-dimensional Čech-complete T1-space
which has no first countable, feebly compact, regular T1-extension space. In [25]
the Terada-Terasawa example was modified and used to obtain the following
result: There is a first countable, feebly compact zero-dimensional T1-space
which has no Urysohn, feebly compact, sequential extension space. (Recall
that a space X is said to be Urysohn provided that every pair of distinct points
of X can be separated by disjoint closed neighborhoods of those points.) We
show next that this example can be used to establish the following.

Theorem 5.8. There exists a zero-dimensional separable Moore space Y such
that Y has no Urysohn, sequentially feebly compact extension space, and hence
Y has no Urysohn, FCC extension space.

Proof. We refer the reader to the proof on page 24 of [25]. Let Y be the space
described there, and assume X is any sequentially feebly compact extension
space of Y . One can replace the fourth–sixth sentences of the last paragraph
on that page by the sentence: “Since X is sequentially feebly compact, the
sequence U = {{(n, i)} : i ∈ In} has associated with it an infinite subset Jn of
In and a cluster point xn of U such that every neighborhood V of xn contains
all but finitely many of the sets in {{(n, i)} : i ∈ Jn}.” Then one can use the
rest of that proof to show that X cannot be a Urysohn space. �

In Theorem 4.1 of [2] Murray Bell showed that if C is the Cantor set, F [C] is
the set of all finite subsets of C, and T is the Pixley-Roy topology on F [C], then



262 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

(F [C], T ) has no first countable pseudocompact Tychonoff extension space.
We show next how to modify his proof and strengthen his theorem. Due to
the complexity of his proof and for the convenience of the reader, we give
a self-contained extension of it, rather than just a fragmentary presentation
of the changes needed. First let us recall that if U is the algebra of clopen
subsets of C, and for each G ∈ F [C] and U ∈ U with G ⊆ U one defines
[G, U] = {H ∈ F [C] : G ⊆ H ⊆ U}, then T has as a base the family C =
{[G, U] : G ∈ F [C], G ⊆ U and U ∈ U}, and each member of C is a clopen
subset of (F [C], T ). It is known ([4]) that (F [C], T ) is a zero-dimensional, ccc,
non-separable Moore space.

Theorem 5.9. The Pixley-Roy space (F [C], T ) has no regular T1 feeble com-
pactification that has property (F3), and hence (F [C], T ) has no extension space
that is FCC and regular.

Proof. Assume that (F [C], T ) is a dense subspace of an FCC, regular space X.
We will construct by induction, a decreasing sequence of Cantor sets {Kn : n ∈
N} and a decreasing sequence of clopen sets {Bn : n ∈ N} with these properties
for each n ∈ N:

(1) Bn ⊇ Kn and diam(Bn) <
1

n
; and

(2) there are a point sn ∈
⋂
{clX[{p}, B

n] : p ∈ Kn} and a sequence {Gnk :
k ∈ N} of points of F [C] converging to sn with the property that Kn

⋂
(
⋃
{Gnk :

k ∈ N}) = ∅.
Assuming the existence of {Kn, Bn, sn, Gnk : n, k ∈ N} with properties

(1) and (2), a contradiction follows quickly. Let p ∈
⋂
{Kn : n ∈ N} and

R = intXclX([{p}, B
1]). Note that R ∩ F [C] = [{p}, B1] and {p} ∈ R. Since

{clX[{p}, B
n] : n ∈ N} is a neighborhood base for {p} in X, there is some n ∈ N

such that clX[{p}, B
n] ⊆ R. By (2), sn ∈ R and there is some k ∈ N such that

Gnk ∈ R. But p ∈ G
n
k since G

n
k ∈ R ∩ F [C] = [{p}, B

1], a contradiction as
Kn ∩ Gnk = ∅ by (2).

The inductive step goes as follows (the first step is similar). Choose a Cantor
set K ⊆ Kn and a clopen Bn+1 with K ⊆ Bn+1 and the diameter of Bn+1

less than 1/(n + 1). Choose a sequence 〈Fk : k < ω〉 of finite subsets of K
that strictly increase up to a dense subset of K. Define Vk = [Fk, B

n+1] \
[Fk+1, B

n+1] for each k < ω. Since {Vk : k < ω} is a pairwise disjoint family
of nonempty open subsets of the dense subset F [C] of the T3-space X, there
exists a sequence U = {Uk : k < ω} of nonempty open subsets of X whose
closures in X are pairwise disjoint, and which satisfy Uk ∩ F [C] ⊆ Vk for each
k < ω. It follows from our Corollary 2.3 that there exist a point sn+1 of X,
an infinite subset J of ω, and a sequence P = {Pk : k ∈ J} of nonempty open
subsets of F [C] such that Pk ⊆ Uk for each k ∈ J, and every neighborhood
of sn+1 contains all but finitely many sets in P. For each k ∈ J, choose
[Gn+1

k
, Wk] ⊆ Pk. Since the Fk’s increase to a dense subset of K and each

Fk is contained in the clopen set Wk, there exists an infinite A ⊆ J such
that K ∩

⋂
k∈A

Wk contains a Cantor set K
′. Since

⋃
k∈A

Gn+1
k

is countable

there exists a Cantor set Kn+1 ⊆ K′ with Kn+1 ∩
⋃

k∈A
Gn+1

k
= ∅. The



Spaces whose Pseudocompact Subspaces are Closed Subsets 263

sequence {Gn+1
k

: k ∈ A} converges to sn+1. If O is any neighborhood of sn+1

then for some k ∈ A, Pk ⊆ O. Consider any p ∈ K
n+1. One has p ∈ Wk and

Gn+1
k

⊆ Bn+1. Hence [{p}, Bn+1]∩[Gn+1
k

, Wk] 6= ∅ and so [{p}, B
n+1]∩O 6= ∅.

Thus sn+1 ∈ clX([{p}, B
n+1]). �

We conclude by asking a question similar to one raised several years ago
by M.V. Matveev. Let us call a feeble compactification E of a space X a
minimal feeble compactification of X provided that for every point p ∈ E \ X,
the space E \ {p} fails to be feebly compact. In each of Theorems 5.2–5.6,
the feeble compactifications obtained for the given space are minimal feeble
compactifications.

Problem 5.10. If P denotes one of the properties Urysohn, regular T1, or
Tychonoff, does every P-space have a P, minimal feeble compactification?

Acknowledgements. Finally, we thank the referee for very good and help-
ful suggestions, all of which have been followed.

References

[1] R.W. Bagley, E.H. Connell and J.D. McKnight, Jr., “On properties characterizing
pseudo-compact spaces”, Proc. Amer. Math. Soc. 9 (1958), 500–506.

[2] M.G. Bell, “First countable pseudocompactifications”, Topology Appl. 21 (1985), 159–
166.

[3] D. Cameron, “A class of maximal topologies”, Pacific J. Math. 70 (1977), 101–104.
[4] Eric K. van Douwen, “The Pixley-Roy topology on spaces of subsets”, Set-theoretic

Topology (Papers, Inst. Medicine and Math., Ohio Univ., Athens, Ohio, 1975–1976),
111–134.

[5] Ryszard Engelking, General Topology, Hermann Verlag Berlin, 1989.
[6] L. Gillman and M. Jerison, Rings of continuous functions, Van Nostrand, New York,

1960.
[7] I. Glicksberg, “The representation of functionals by integrals”, Duke Math. J. 19 (1952),

253–261.
[8] I. Glicksberg, “Stone-Čech compactifications of products”, Trans. Amer. Math. Soc. 90

(1959), 369–382.
[9] Zhou Hao-Xuan, “A conjecture on compact Fréchet spaces”, Proc. Amer. Math. Soc.

89 (1983), 326–328.
[10] E. Hewitt, “Rings of real-valued continuous functions, I”, Trans. Amer. Math. Soc. 64

(1948), 54–99.
[11] Mohammed Ismail and Peter Nyikos, “On spaces in which countably compact sets are

closed, and hereditary properties”, Topology Appl. 11 (1980), 281–292.
[12] Peter J. Nyikos, “Classes of compact sequential spaces”, Set Theory and its Applications,

Lecture Notes in Mathematics, No. 1401, edited by J. Steprans and S. Watson, Springer-
Verlag (1987), 135–159.

[13] Peter J. Nyikos, “Some Moore pseudocompactifications”, Seminar notes at the University
of South Carolina (1991).



264 A. Dow, J. R. Porter, R. M. Stephenson, Jr., and R. Grant Woods

[14] Jack R. Porter, “Minimal first countable spaces”, Bull. Austral. Math. Soc. 3 (1970),
55–64.

[15] Jack R. Porter and R. Grant Woods, Extensions and Absolutes of Hausdorff spaces,
Springer-Verlag, 1987.

[16] Jack R. Porter, R.M. Stephenson, Jr., and R. Grant Woods , “Maximal feebly compact
spaces”, Topology Appl. 52 (1993), 203–219.

[17] A. B. Raha, “Maximal topologies”, J. Austral. Math. Soc. 15 (1973), 279–290.
[18] G. M. Reed, “On chain conditions in Moore spaces”, Gen. Topology Appl. 4 (1974),

255–267.
[19] G. M. Reed, “On subspaces of separable first countable T2-spaces”, Fund. Math. 91

(1976), 189–202.
[20] C. T. Scarborough and A. H. Stone,“Products of nearly compact spaces”, Trans. Amer.

Math. Soc. 124 (1966), 131-147.
[21] Petr Simon, “A compact Fréchet space whose square is not Fréchet”, Comment. Math.

Univ. Carolinae 21 (1980), 749–753.
[22] Petr Simon and Gino Tironi, “First countable extensions of regular spaces”, Proc. Amer.

Math. Soc. to appear.
[23] R. M. Stephenson, Jr., “Minimal first countable topologies”, Trans. Amer. Math. Soc.

128 (1969), 115–127.

[24] R. M. Stephenson, Jr., “Minimal first countable Hausdorff spaces”, Pacific J. Math. 36
(1971), 819–825.

[25] R. M. Stephenson, Jr., “Moore-closed and first countable feebly compact extension
spaces”, Topology Appl. 27 (1987), 11–28.

[26] A. H. Stone, “Hereditarily compact spaces”, Amer. J. Math. 82 (1960), 900–914.
[27] Toshiji Terada and Jun Terasawa,“Maximal extensions of first-countable spaces”, Proc.

Amer. Math. Soc. 85 (1982), 95–99.
[28] S. Watson,“Pseudocompact metacompact spaces are compact”, Proc. Amer. Math. Soc.

81 (1981), 151–152.

Received July 2003

Accepted December 2003

Alan Dow (adow@uncc.edu)
Department of Mathematics, University of North Carolina, Charlotte, NC
28223, U.S.A.

Jack R. Porter (porter@math.ku.edu)
Department of Mathematics, University of Kansas, Lawrence, KS 66045, U.S.A.

R. M. Stephenson, Jr. (stephenson@math.sc.edu)
Department of Mathematics, University of South Carolina, Columbia, SC 29208,
U.S.A.

R. Grant Woods (rgwoods@cc.umanitoba.ca)
Department of Mathematics, University of Manitoba, Winnipeg, MB Canada
R3T 2N2