Applied General Topology

@ c Universidad Politécnica de Valencia

Volume 5, No. 2, 2004

pp. 231-242

Which topologies can have immediate

successors in the lattice of T 1-topologies?

Ofelia T. Alas and Richard G. Wilson∗

Abstract. We give a new characterization of those topologies which

have an immediate successor or cover in the lattice of T1 -topologies on
a set and show that certain classes of compact and countably compact
topologies do not have covers.

Keywords: lattice of T1 -topologies, maximal point, KC-space, cover of a

topology, upper topology, lower topology
2000 AMS Classification: Primary 54A10; Secondary 06A06

1. Introduction and Preliminary Results

The lattice L1 (X) of T1 -topologies on a set X has been studied, with differing
emphasis, by many authors. Most articles on the subject have considered com-
plementation in the lattice and properties of mutually complementary spaces
(see for example, [12, 13, 15]. Here we consider a facet of the order structure
of the lattice L1 (X), specifically the problem of when a jump can occur in the
order; that is to say, when there exist topologies τ and τ + on a set X such that
whenever µ is a topology on X such that τ ⊆ µ ⊆ τ + then µ = τ or µ = τ + .
The existence of jumps in L1 (X) has been studied in [10] and [16] where the
immediate successor τ + was said to be a cover of (or simply to cover) τ . We
prefer order-theoretic terminology and then call τ a lower topology for X; the
topology τ + will then be termed an upper topology. That such topologies exist
was noted in both of the previously mentioned papers - the best known example
is the topology of the space Σ of Problem 4M of [7], called an ultratopology in
[10] and [16], which is an anti-atom of L1 (ω) (for lattice-theoretic terminology
we refer the reader to the survey paper [9]). However, topologies which are
dense-in-themselves can also be lower, a Tychonoff example is the maximal
space constructed in [4]. A characterization of lower topologies was given in
∗Research supported by Consejo Nacional de Ciencia y Tecnologı́a (México), grant 38164-
E and Fundação de Amparo a Pesquisa do Estado de São Paulo (Brasil)
232 O. T. Alas and R. G. Wilson

Theorem 1 of [10] in terms of equivalence classes of locally equal sets. Here we

give a somewhat different characterization of lower topologies which we find
much simpler and easier to work with, but we note that Lemma 1.2 and Corol-
lary 1.3 can, with a little effort, be derived from Section 2 of [10]. Corollary
1 to Theorem 10 in [16] states that no first countable Hausdorff topology on
X has a cover in L1 (X), thus generalizing Example 2 of [10] where this same
result was proved for the space of real numbers with the usual metric topology.
However, these papers do not address further the problem: Which topologies
can have, and which classes of topologies do not have, covers in L1 (X)? This
is the problem we study below.
All spaces considered below are (at least) T1 and undefined topological no-
tation and terminology can be found in [6]. The closure (respectively, interior)
of a set A in a topological space (X, τ ) will be denoted by clτ (A) (respectively,
intτ (A)) or simply by cl(A) (respectively int(A)) if no confusion is possible. If
S is a family of sets, then we use the notation < S > to denote the topol-
ogy generated by S as a subbase and the symbol is used to denote proper
containment.
We call a point p a maximal point of a topological space (X, τ ) if it is not
isolated and whenever U is open in X and p ∈ clτ (U ), then U ∪ {p} ∈ τ . It
is clear that if p is a maximal point of X and p is an accumulation point of
A ⊆ X, then p is a maximal point of A ∪ {p}. Using the well-known fact (see
for example Problem 12G of [18]) that an open filter F is an ultrafilter if and
only if whenever U ∈ τ , then U ∈ F or X \ clτ (U ) ∈ F , it is an easy exercise
to check that p is a maximal point of X if and only if the trace of the open
neighbourhood system at p on X \ {p} is an open ultrafilter in the open sets of
this latter space. We formulate this fact as:
Proposition 1.1. A point p ∈ X is a maximal point if and only if the trace of
the open neighbourhood filter at p on the subspace X \ {p} is an open ultrafilter.
We note that a maximal point is also a 1-point in the sense of [17] but not
vice versa. It is also true that in a space which is maximal in the sense of [4],
that is to say, is maximal with respect to being dense-in-itself, every point is
a maximal point, but again the converse is false (the cofinite topology on a
countable set is the requisite counterexample). Our first task is to give a new
characterization of lower topologies in terms of maximal points.
Lemma 1.2. Let (X, τ ) be a T1 -space; if τ has an immediate successor, which
we denote by τ + , then there is p ∈ X and U ∈ τ such that U ∪ {p} 6∈ τ and
τ + =< τ ∪ {U ∪ {p}} >
Proof. First note that if τ σ, then we can find V ∈ σ \ τ and hence the
topology τ ∗ generated by τ ∪ {V } is such that τ τ ∗ ⊆ σ. Thus if τ + exists
it must have the form < τ ∪ {V } > for some V 6∈ τ .
We now claim that the set V defined in the previous paragraph must have
the form U ∪ {p} for some p ∈ X \ U and U ∈ τ ; again let σ =< τ ∪ {V } >. To
prove our claim, let U = intτ (V ); then if U = V , it follows that V ∈ τ and hence
 Topologies with immediate successors 233

σ = τ . Thus we suppose U V . If |V \ U | ≥ 2, then we can choose distinct

points p, q ∈ V \U and since p 6∈ intτ (V ) it is clear that V \{q} 6∈ τ . We consider
the topology τ ∗ =< τ ∪ {V \ {q}} >; since X is T1 and V \ {q} = V ∩ (X \ {q}),
it follows that V \ {q} ∈ σ. However, since V 6∈ τ ∗ we have that τ τ ∗ σ,
showing that σ is not an immediate successor of τ . Thus |V \ U | = 1 and we
are done. 

To simplify the notation somewhat, we denote the topology < τ ∪{U ∪{p}} >
by τU(p) or simply by τU whenever p is understood to be fixed.
Corollary 1.3. A topology τ on X is a lower topology if and only if there is
p ∈ X and U ∈ τ such that whenever V ∈ τ is such that τU = τU∩V then either
τV = τU or τV = τ .
Proof. Suppose that τ is a lower topology, whose immediate successor we again
denote by τ + . By Lemma 1.2, there exist p ∈ X and U ∈ τ such that τ + =
τU(p) (= τU ). But then, if V ∈ τ is such that τU∩V = τV , then τ ⊆ τV and
since V ∪ {p} = V ∪ [(U ∩ V ) ∪ {p}] it follows that τV ⊆ τU∩V = τU and hence
τV = τU or τV = τ .
Conversely, suppose that τ is not a lower topology and let p ∈ X be fixed.
If U ∈ τ , then there is some topology σ on X such that τ σ τU . It is clear
that there is then some V ∈ τ such that σ = τV and then since V ∪ {p} ∈ τU ,
it follows that (U ∩ V ) ∪ {p} ∈ τU and hence τU∩V = τU , a contradiction. 

Let D be a directed set and f : D → X a net which is finally in U ∪ V .

If f is cofinally in both U and V and DU = {α ∈ D : f (α) ∈ U } and DV =
{α ∈ D : f (α) ∈ V } then both DU and DV are directed sets and if the nets
fU = f |DU and fV = f |DV both converge to the same point p, then f also
converges to p. We will make use of this simple result in the next theorem.
Theorem 1.4. A topology τ is a lower topology on X if and only if (X, τ ) has
a closed subspace with a maximal point.
Proof. For the necessity, we suppose that no closed subspace of (X, τ ) has a
maximal point. Then for all W ∈ τ and p ∈ X \ W , p is not a maximal point of
X \ W . However, if τ were a lower topology, then there would exist U ∈ τ and
p ∈ X such that τ + = τU(p) 6= τ . For the rest of this paragraph, we consider p
fixed and write τU(p) = τU . Since p is not a maximal point of X \ U , there is
some open set V ′ ∈ τ |(X \ U ) such that V ′ ∪ {p} 6∈ τ |(X \ U ) and p ∈ clτ (V ′ );
then p ∈ clτ [(X \ U ) \ (V ′ ∪ {p})]. Let V ∈ τ be such that V ∩ (X \ U ) = V ′ .
We claim that τ τU∪V τU . That τ ⊆ τU∪V ⊆ τU is clear and to see that
(ı) τ 6= τU∪V , note that p 6∈ clτU ∪V [(X \ U ) \ (V ∪ {p})].
(ıı) τU∪V 6= τU , note that p is not isolated in τU∪V |(X \ U ).
For the converse, suppose that p is a maximal point of the closed subspace
C ⊆ X and let X \ C = U ∈ τ . Let V ∈ τ be such that τU∩V = τU , and hence
τU = τU∩V ⊇ τV ⊇ τ ; then by Corollary 1.3, it suffices to show that τV = τ or
τV = τU . There are two cases to consider.
234 O. T. Alas and R. G. Wilson

(1) If p 6∈ clτ (V ∩ C), then a net f that converges to p in the topology τV is

finally in each τ -open set containing p and
(a) Finally in V ∪ {p}, and
(b) Finally outside of V ∩ C = V \ U since C \ clτ (V ∩ C) is an open set in C
which contains p.
Thus the net f is finally in (V ∩ U ) ∪ {p} and in each τ -neighbourhood of p
and hence converges to p in τU∩V = τU . Since τU and τV coincide on X \ {p},
it follows that τU = τV .
(2) If p ∈ clτ (V ∩C), then since p is a maximal point of C, (V ∩C)∪{p} ∈ τ |C
and hence there is W ∈ τ such that p ∈ W and W ∩ C = (V ∩ C) ∪ {p},
that is, (V \ U ) ∪ {p} = W \ U . But then, p ∈ W ⊆ (U ∪ V ) ∪ {p} and
hence (U ∪ V ) ∪ {p} ∈ τ . Since τ and τU∪V coincide on X \ {p}, it follows
that τ = τU∪V . Thus a net f which converges to p in τ is finally in each
τ -neighbourhood of p, (including (U ∪ V ) ∪ {p}) and then either,
(c) It is finally in U ∪ {p} and hence converges in τU and then also in τV since
τV ⊆ τU , or
(d) It is finally in V ∪ {p} and hence converges in τV , or
(e) It is cofinally in both U ∪ {p} and V ∪ {p}. Using the notation introduced
in the paragraph prior to this theorem, the net fU converges to p in τU and
hence also in τV and the net fV converges to p in τV . Thus the net f converges
to p in τV and so τ = τV . 

On the other hand, τ can be a lower topology on X but have no maximal

point. Let p be a free ultrafilter on ω and let Σ = ω ∪ {p} be the space of
Problem 4M of [7]. Let (X, τ ) be the quotient space obtained by identifying
the point p ∈ Σ with the unique accumulation point of a convergent sequence.
That τ is a lower topology follows from Theorem 1.4 and the fact that p is a
maximal point of the closed subspace Σ, but the space (X, τ ) has no 1-point.

Corollary 1.5. If a topology σ on a set X has the property that for some
infinite closed subspace D ⊆ X, σ|D is a lower (respectively, upper) topology
on D, then σ is a lower (respectively upper) topology on X.

At this point, having mentioned upper topologies for the first time, the
reader might wonder why we have fixed our attention so exclusively on lower
topologies since to every lower topology there corresponds at least one upper
topology. However, as we shall see, unlike lower topologies, upper topologies
are abundant. Recall from [11] that a space is hyperconnected if it contains no
disjoint non-empty open sets, or equivalently, if every non-empty open set is
dense. Clearly no Hausdorff space with at least two points is hyperconnected.
If σ is the cofinite topology on a set X, then (X, σ) is hyperconnected, but
other examples of hyperconnected spaces are easy to construct. However, we
have the following simple result which we feel sure must be known.

Lemma 1.6. A T1 -topology σ on a set X is the cofinite topology if and only if

every infinite closed subspace of (X, σ) is hyperconnected.
 Topologies with immediate successors 235

Proof. The necessity is clear. For the sufficiency, suppose that σ is not the
cofinite topology on X; then there is some infinite closed proper subset C X.
Let p ∈ X \ C, then the infinite closed subspace C ∪ {p} is not hyperconnected.


Proposition 1.7. A topology σ is not an upper topology on an infinite set X

if and only if σ is the cofinite topology.
Proof. The sufficiency is clear, since the cofinite topology is the minimal T1 -
topology on a set.
For the necessity, suppose first that there is a non-empty regular closed set
C in (X, σ) whose complement is infinite and let p ∈ intσ (C). Let U be an
open ultrafilter on the infinite open subspace X \ C such that ∩U = ∅ and
define a topology τ on X by
V ∈ τ if and only if V ∈ σ and whenever p ∈ V ,
then there exists U ∈ U such that U ⊆ V .
Clearly σ|(X \ {p}) = τ |(X \ {p}) and p is a maximal point (in the topology
τ ) of the τ -closed set (X \ intσ (C)) ∪ {p} and hence it follows from Theorem 1.4
that τ is a lower topology. Furthermore, if we denote by W the set intσ (C)\{p},
then W ∈ τ and we see from the proof of Theorem 1.4 that σ = τ + = τW (p) .
Thus if σ is not an upper topology, then the closure of every non-empty
element of σ has finite complement. Now suppose that ∅ 6= U ∈ σ; if the finite
open set X \ clσ (U ) 6= ∅ for some U ∈ σ, each point q ∈ X \ clσ (U ) is isolated
and so {q} is a regular closed set with infinite complement, a contradiction.
Thus clσ (U ) = X, that is U is dense in X and so X is hyperconnected.
However, from Corollary 1.5, we have that if σ is not an upper topology on
X, then it is not an upper topology on any infinite closed subset of X. Hence
every infinite closed subspace of X is hyperconnected. The result now follows
from Lemma 1.6. 

It is not hard to see that in Proposition 1.7, if (X, σ) is Hausdorff but not
H-closed, then τ can be chosen to be Hausdorff as well.

2. Topologies which cannot be lower

It is now clear that if P is a topological property which is inherited by closed
subspaces, then the class of topologies with property P can contain a lower
topology if and only if there is a member of the class with a maximal point. We
now consider the problem of which classes of topologies do not contain a lower
topology. As mentioned earlier, the best known examples of lower topologies,
Σ and van Douwen’s maximal space are far from being compact. That this is
not a coincidence is shown in the next theorem. Recall that a topological space
is a KC-space if all compact subsets are closed. Hausdorff spaces are KC and
KC-spaces are necessarily T1 . Our main results below show that KC-spaces
with “nice” covering and convergence properties cannot be lower topologies.
However, these results fail in case the KC separation axiom is weakened to T1 .
236 O. T. Alas and R. G. Wilson

Theorem 2.1. A compact KC-space cannot have a maximal point.

Proof. Let (X, τ ) be a compact KC-space; we assume to begin with that X is
dense-in-itself. Suppose to the contrary that p is a maximal point of X and let
K = ∩{cl(V ) : p ∈ V ∈ τ }.
Clearly, K is compact and p ∈ K. Let λ be the minimal cardinal such that
K = ∩{cl(Vα ) : α ∈ λ, p ∈ Vα ∈ τ } and let Fβ = ∩{cl(Vα ) : α ≤ β}.
Proceeding as in Lemma 2.3 of [1], we can find a discrete subset D in X
such that p 6∈ D and cl(D) ∩ Fα 6= ∅ for each α ∈ λ. Since X is compact, it
follows that cl(D) ∩ K 6= ∅.
There are now two possibilities:
(ı) K = {p} and hence p ∈ cl(D). Since X has no isolated points it follows that
cl(D) is nowhere dense and hence U = X \ cl(D) is a dense open set. Since p
is maximal, U ∪ {p} is open, which contradicts the fact that p ∈ cl(D).
(ıı) If K \ {p} 6= ∅, then choose q ∈ K \ {p}. If F is an open covering of
X \ {p}, then there is U ∈ F such that q ∈ U and since p and q do not have
disjoint neighbourhoods, p ∈ cl(U ). Since p is maximal, U ∪ {p} is open and
hence F ∪ {U ∪ {p}} is an open covering of the compact space X, which then
has a finite subcovering. Since U ∈ F , this clearly induces a finite subcovering
of F of X \ {p}, showing that this latter space is compact. Since X is KC, p
must be an isolated point of X, a contradiction.
Now suppose that X is an arbitrary compact KC-space, p ∈ X and let D be
the set of isolated points of X. If p 6∈ cl(D), then there is some neighbourhood
U of p which is dense-in-itself and hence p is a maximal point of cl(U ) which is
a dense-in-itself compact KC-space. That p is not maximal now follows from
the previous argument.
If p ∈ cl(D), then if p were a maximal point of X it would also be a maximal
point of cl(D) and hence without loss of generality we assume that X = cl(D).
Since p is maximal, it follows that D ∪ {p} is open. Let K = ∩{cl(V ) : V is an
open neighbourhood of p}.
If K \ {p} =6 ∅, then as in (ıı) above, p is an isolated point of X, a contra-
diction.
If on the other hand K = {p}, then p can be separated from every point
q ∈ X by disjoint open sets. Since p is an isolated point of X \ D, it follows
that (X \ D) \ {p} is compact and a standard argument now shows that p
and (X \ D) \ {p} can be separated by open sets. Thus there is some closed,
hence compact neighbourhood V of p which misses (X \ D) \ {p}. Thus V is
a compact KC-space whose only accumulation point is a maximal point. This
is clearly seen to be impossible and we are done. 
Corollary 2.2. A compact KC-topology is not a lower topology.
In contrast with Corollary 2.2, the topology τ of the one-point compactifi-
cation X of the space Σ is a lower topology and the cofinite topology on ω is
even a first countable compact T1 -topology in which every point is a maximal
point.
 Topologies with immediate successors 237

If (X, τ ) is a Hausdorff space, then

T for each p ∈ X, set ψc (p, X) = min{|U| :
U ⊆ τ , p ∈ U for all U ∈ U and {cl(U ) : U ∈ U} = {p}}; as in [8], we then
define the closed pseudocharacter of X, ψc (X) = sup{ψc (p, X) : p ∈ X}.
Recall that if λ is a cardinal, then a space X is said to be initially λ-compact
if every open cover of size at most λ has a finite subcover. It is a trivial exercise
to show that if X is initially λ-compact, then every decreasing sequence of
length ≤ λ of closed sets in X has non-empty intersection. Furthermore, an
initially λ-compact Hausdorff space X with ψc (X) ≤ λ is regular (see Theorems
2.2 and 3.4 of [14]). The next theorem has a proof very similar to that of
Theorem 2.1, we mention only the minor points of difference.
Theorem 2.3. An initially λ-compact Hausdorff space X with ψc (X) ≤ λ
cannot have a maximal point.
Proof. Again we begin by assuming that X is dense-in-itself, p is a maximal
point of X and κ ≤ λ is the minimal cardinal such that there exists a family
{Vα : α ∈ κ} of open sets such that {p} = ∩{cl(Vα ) : α ∈ κ}. Using exactly
the same notation as that of Theorem 2.1 and applying Lemma 2.3 of [1],
we can find a discrete set D ⊆ X \ {p} such that cl(D) ∩ Fα 6= ∅ for each
α ∈ κ. WeT note that since X is initially λ-compact, it is κ-compact and
hence ∅ 6= {cl(D) ∩ Fα : α ∈ κ} ⊆ {p} and so p ∈ cl(D). The rest of the
proof proceeds exactly as in Theorem 2.1, after noting that X is, in fact, a
T3 -space. 

A space X is said to be weakly discretely generated if whenever A X is

not closed, then there is some discrete subset D ⊆ A such that cl(D) \ A 6= ∅.
The space X is discretely generated if whenever x ∈ cl(A) there is a discrete
subspace D ⊆ A such that x ∈ cl(D); being discretely generated is a hereditary
property. Clearly, a discretely generated space is weakly discretely generated
and it was shown in Proposition 3.1 of [5] that every compact Hausdorff space is
weakly discretely generated. A similar proof applying Lemma 2.3 of [1] can be
used to show that compact KC-spaces are weakly discretely generated. It was
further shown in [2] (and implicitly earlier in [3]) that a regular countably com-
pact space with countable tightness is discretely generated. Our next lemma
generalizes this last result and has a similar proof. We say that a space X is
locally countably compact if each point p ∈ X has a local base of countably
compact neighbourhoods. A countably compact T3 -space is locally countably
compact, but this is not the case for Hausdorff spaces as we shall see later. It
is easy to see that being locally countably compact is both open and closed
hereditary.
Lemma 2.4. A locally countably compact Hausdorff space X with countable
tightness is discretely generated.
Proof. Let A ⊆ X be a set which is not closed and y ∈ cl(A) \ A. Since X
has countable tightness, we may assume that A = {an : n ∈ ω} is countable.
Our aim is to construct a discrete subset D ⊆ A such that y ∈ cl(D). Since
238 O. T. Alas and R. G. Wilson

X is Hausdorff, for each n ∈ ω there is an open neighbourhood Un of y such

that
T an 6∈ cl(Un ). We assume that Un+1 ⊆ Un for each n ∈ ω and clearly
{cl(Un ) : n ∈ ω} ∩ A = ∅. Now let S be the set of all cluster points of sets
of the form {zn : n ∈ ω} where zn ∈ Un ∩ A; we claim that y ∈ cl(S). To see
this, let W be a countably compact neighbourhood of y and for each n ∈ ω,
pick zn ∈ W ∩ Un ∩ A. Each accumulation point of {zn : n ∈ ω} lies in S,
and since W is countably compact at least one of them also lies in W ; hence
W ∩ S 6= ∅. Again since X has countable tightness, there is a countable subset
{sn : n ∈ ω} ⊆ S such that y ∈ cl({sn : n ∈ ω}) and for each n ∈ ω we choose a
sequence {znk : n ∈ ω} ⊆ A which witnesses the fact that sk ∈ S, that is to say,
znk ∈ Un ∩ A and sk ∈ cl({znk : n ∈ ω}). Now let D = {znk : k, n ∈ ω and k ≤ n}.
It is clear that sk ∈ cl(D) for each k ∈ ω and hence y ∈ cl(D). To see that D
is discrete, fix k, n ∈ ω; then there is some m ≥ n such that znk 6∈ cl(Um ) and
since all but finitely many elements of D lie in Um , znk is not an accumulation
point of D. 
Theorem 2.5. A locally countably compact Hausdorff topology of countable
tightness is not a lower topology.
Proof. Let (X, τ ) be a locally countably compact Hausdorff space and suppose
that τ has an immediate successor which we denote by σ. Then there is some
U ∈ τ and p ∈ X such that σ =< τ ∪ {U ∪ {p}} >. Furthermore, since the
property of being locally countably compact is inherited by closed subspaces,
we may assume that p is a maximal point of X. Let S = X \ U and note
that p is an accumulation point of the closed subspace S, for otherwise we have
U ∪ {p} ∈ τ and hence σ = τ . Furthermore, by the previous lemma, S is
discretely generated. There are two possibilities to consider.
1) The point p is an accumulation point of some dense-in-itself closed subset
C of S. Since X is discretely generated, there is some discrete subset D in C \
{p} such that p ∈ clτ (D). Furthermore, since C is dense-in-itself, T = clτ (D) is
nowhere dense in C, hence also in S, and so p ∈ clτ (S \ T ). Now define τ # =<
τ ∪ {(X \ T ) ∪ {p} >; since p ∈ clτ (T \ {p}) it follows immediately that T \ {p}
is not closed in τ , but is closed in τ # and hence τ τ # . Furthermore, since
(X \T )∪{p} = ((X \S)∪{p})∪(X \T ) it follows that (X \T )∪{p} ∈ σ, showing
that τ # ⊆ σ. Finally, since each τ # -neighbourhood of p meets S \ T , but the
σ-neighbourhood U ∪ {p} of p misses S \ T , we have p ∈ clτ # (S \ T ) \ clσ (S \ T )
and it then follows that τ # 6= σ, contradicting the supposition that σ is the
immediate successor of τ .
2) Suppose now that p is not in the closure of any dense-in-itself subset of S.
If S is not scattered, then let G be the family of all dense-in-themselves subsets
of S ordered by set inclusion. Clearly the union of any chain of elements of G is
dense-in-itself, and hence G has a maximal element G say, which clearly must
be closed and p 6∈ G. Since S is closed in X, it is locally countably compact and
so there is a countably compact neighbourhood W of p in S missing G. Then
p is an accumulation point of the countably compact scattered space W ⊆ S.
Suppose that the scattering length of W is κ and for each α < κ, let Wα denote
 Topologies with immediate successors 239

the set of points of scattering order α; since p is not an isolated point of W ,

p ∈ Wα for some α > 0. There are two subcases to consider:
a) If p ∈ Wα where α > 1, then let T = W \ W0 . It is clear that T is a closed
subspace of W and hence T = clτ (T ) ∩ W , p is an accumulation point of T and
W \ T is dense in W . Now define τ # =< τ ∪ {(X \ clτ (T )) ∪ {p} >. Obviously
clτ (T ) \ {p} is closed in τ # but not in τ and hence τ τ # . Furthermore,
(X \ clτ (T )) ∪ {p} = (X \ S) ∪ {p} ∪ (X \ clτ (T )) ∈ τ and so τ # ⊆ σ. To show
that σ 6= τ # we will show that p ∈ clτ # (S \T )\clσ (S \T ). That p 6∈ clσ (S \T ) is
clear since U ∪ {p} ∈ σ is a neighbourhood of p and (S \ T ) ∩ (U ∪ {p}) = ∅. On
the other hand, an open τ # -neighbourhood of p is of the form (V \ clτ (T ))∪{p}
where p ∈ V ∈ τ and a short calculation now shows that (V \ clτ (T )) ∪ {p} ∩
(S \ T ) = V \ (U ∪ clτ (T )). This latter set is nonempty since if V ⊆ U ∪ clτ (T ),
then V ∩ W ⊆ clτ (T ) ∩ W = T which is a contradiction, since each open set in
W meets W0 . Thus we have again shown that σ is not the immediate successor
of τ .
b) If p ∈ W1 , then W0 ∪ W1 is an open neighbourhood of p in W and since W
is a neighbourhood of p in S, we have p ∈ intτ |S (W0 ∪ W1 ) and this latter set,
being an open subset of S, is locally countably compact. Thus there is some
countably compact neighbourhood V of p in which p is the only accumulation
point. Clearly V is compact and it then follows from Theorem 2.1 that p is not
a maximal point of V , a contradiction. 

Question 2.6. Can a countably compact Hausdorff topology with countable

tightness (or a countably compact regular topology) be a lower topology?

We note however that a countably compact Hausdorff topology may be a

lower topology and need not be locally countably compact. Let p ∈ βω \
ω and let V be the filter of open neighbourhoods of p. Denoting by τ the
usual topology of βω, we define σ =< τ ∪ {(U ∩ ω) ∪ {p} : U ∈ V} >. It is
straightforward to check that p is a maximal point of the countably compact
Urysohn space (βω, σ). This latter space is even H-closed (see 3.12.5 of [6])
and its semiregularization is (βω, τ ).
Our next theorem generalizes Corollary 1 to Theorem 10 of [16].

Theorem 2.7. A sequential KC-space is not a lower topology.

Proof. Suppose (X, τ ) is a sequential KC-space and τ σ. Then there is some

set V ⊆ X which is closed in (X, σ) but not closed in (X, τ ). Since this latter
space is sequential, there is a sequence S = {bn }n∈ω in V which converges to
b 6∈ V . Since (X, τ ) is KC, S ∪ {b} is τ -closed and hence σ-closed and so S is σ-
closed and discrete. Let µ =< τ ∪ {X \ {b2n : n ∈ ω}} >. Then τ µ σ since
{b2n+1 }n∈ω converges in (X, µ) but not in (X, σ), while {b2n }n∈ω converges in
(X, τ ) but not in (X, µ). 

Corollary 2.8. Between any two distinct comparable sequential KC-topologies

on a set X, there are an infinite number of topologies.
240 O. T. Alas and R. G. Wilson

We note in passing that all we have used in Theorem 2.7 is that sequences
have unique limits. On the other hand, as with Corollary 2.2, both the theorem
and its corollary are false if KC is replaced by the T1 separation axiom: If τ
is the cofinite topology on a countably infinite set X, then (X, τ ) is second
countable, but each point p ∈ X is maximal and (X, τ ) is a lower topology. It
is also easy to see that any successor topology to τ is also second countable.
However, the next theorem shows that the cofinite topology is crucial in the
construction of a first countable lower topology.
Theorem 2.9. A sequential T1 -space with a maximal point contains an infinite
subspace whose relative topology is cofinite.
Proof. Suppose (X, τ ) is a sequential T1 -space with a maximal point p. Since
p is not isolated, we can choose a sequence of distinct points S = {xn }n∈ω ⊆
X \ {p} converging to p and hence p is an accumulation point of the subspace
S ∪ {p}. If the set of isolated points I of (S, τ |S) is infinite, then p ∈ cl(I)
and if I = {dn : n ∈ ω} is an enumeration of I, then p ∈ cl({d2n : n ∈
ω}) ∩ cl({d2n+1 : n ∈ ω}) showing that p is not a 1-point, thus is not maximal
in I ∪ {p}, hence not maximal in X, which contradicts our hypothesis. Thus
I is finite and by replacing S with S \ I we may, without loss of generality,
assume that S has no isolated points. Suppose that U ∈ τ |S; since S has no
isolated points, U is infinite and hence p ∈ cl(U ). Since p is a maximal point
of S ∪ {p}, it follows that U ∪ {p} ∈ τ |(S ∪ {p}). If D = {n : xn ∈ S \ U } is
infinite then {xn : n ∈ D} is a subsequence of S which does not converge to
p, again a contradiction. Thus S \ U is finite, showing that S has the cofinite
topology. 
Corollary 2.10. If τ is a first countable lower T1 -topology on X then there is
an infinite subset S ⊆ X such that τ |S is the cofinite topology.
As a partial converse to Theorem 2.9, it follows from Theorem 1.4, that
if a space has an infinite closed subspace with the cofinite topology, then its
topology is lower. However, a first countable T1 -space may have an infinite
subset with the cofinite topology and still not be lower as the following example
shows:
Let µ denote the usual metric topology on the set of reals R and define a
new topology σ on R as follows:
U ∈ σ if and only if U ∈ µ and there is some ǫ > 0
such that U ⊇ (n − ǫ, n + ǫ) for all but finitely many n ∈ N.
Clearly σ is a first countable T1 topology on R and σ| N is the cofinite
topology on N. Further note that if A is a bounded set in R, then σ| A = µ| A.
To show that (R, σ) is not a lower topology, it suffices to show that no closed
subspace has a maximal point. Let C be a closed subset of (R, σ) and p ∈ C;
if C ∩ N is finite, then for some m ∈ N, p ∈ A = C ∩ (−m, m) and σ| A = µ| A.
Thus by Theorem 2.7, p is not a maximal point of A; now a little thought
shows that p is an accumulation point of A if and only if it is an accumulation
point of C and it then follows that p is not a maximal point of C. If C ∩ N is
 Topologies with immediate successors 241

infinite, then C is dense in S

(R, σ) and hence C = R. However, if p ∈ R, then
we define U = (p, p + 21 ) ∪ {(n − 12 , n + 21 ) : n ∈ N \ (p − 1, p + 1)}. That σ
is not a lower topology now follows since U ∈ σ, p ∈ clσ (U ), but U ∪ {p} 6∈ σ
showing that p is not a maximal point of (R, σ).
Recall that a space X is radial if whenever A ⊆ X and x ∈ cl(A) there is a
well-ordered net in A converging to x. A space X is pseudoradial if whenever
A X is not closed there is a well-ordered net in A which converges to a point
of X \ A. Every first countable space is radial and both sequential and radial
spaces are pseudoradial; clearly, being radial is a hereditary property.
Theorem 2.11. A radial Hausdorff space cannot have a maximal point.
Proof. Suppose that (X, τ ) is a radial Hausdorff space; by Theorem 4.3 of
[1], X is discretely generated. Suppose that p ∈ X is not isolated. Since X is
discretely generated, there is some discrete set D ⊆ X \{p} such that p ∈ cl(D).
Since X is radial, there is a well-ordered net {dα }α∈κ in D converging to p.
Let P = {dα : α ∈ κ}, S = {dα : α = γ + 2n for some limit ordinal γ ∈ κ and
n ∈ ω} and T = {dα : α = γ + 2n + 1 for some limit ordinal γ ∈ κ and n ∈ ω}.
Clearly both S and T are disjoint open subsets of the subspace P ∪ {p} and
p ∈ cl(S) ∩ cl(T ) showing that p is not a 1-point, thus is not maximal in the
space P ∪ {p} and hence is not maximal in X. 

Theorem 2.11 cannot be extended to KC-spaces since a set of size ω1 with

the cocountable topology is a radial KC-space, each point of which is maximal.
(This space even has the stronger property that each point has a nested local
base.) The question then arises:
Question 2.12. Can a pseudoradial T2 -space have a maximal point?
Preceding Theorem 2.7, we gave an example of an H-closed space with a
maximal point. Thus the following question arises:
Question 2.13. Can a minimal Hausdorff topology be a lower topology?

References
[1] O. T. Alas, V. V. Tkachuk and R. G. Wilson, Closures of discrete sets often reflect global
properties, Topology Proceedings 25 (2000), 27-44.
[2] A. Bella, Few remarks on spaces which are generated by discrete sets, to appear.
[3] A. Bella and V. I. Malykhin, F-points in countably compact spaces, Applied General
Topology 2 no. 1 (2001), 33-37.
[4] E. van Douwen, Applications of maximal topologies, Topology and Its Applications 51
(1993), 125-139.
[5] A. Dow, M. G. Tkachenko, V. V. Tkachuk and R. G. Wilson, Topologies generated by
discrete subspaces, Glasnik Mat. Ser. III, 37 (57) no. 1 (2002), 187-210.
[6] R. Engelking, General Topology, (Heldermann Verlag, Berlin, 1989).
[7] L. Gillman and M. Jerison, Rings of Continuous Functions, (Van Nostrand, Princeton,
NJ, 1960).
[8] I. Juhász, Cardinal Functions in Topology - Ten Years Later, Mathematical Centre
Tracts 123, (Amsterdam, 1980).
242 O. T. Alas and R. G. Wilson

[9] R. E. Larson and S. Andima, The lattice of topologies; A survey, Rocky Mountain J. of
Math. 5 no. 2 (1975), 177-198.
[10] R. E. Larson and W. J. Thron, Covering relations in the lattice of T1 -topologies, Trans-
actions of the American Mathematical Society 168 (1972), 101-111.
[11] L. A. Steen and J. A. Seebach, Counterexamples in Topology, (Springer Verlag, New
York, 1978).
[12] A. K. Steiner, Complementation in the lattice of T1 -topologies, Proceedings of the Amer-
ican Mathematical Society 17 (1966), 884-885.
[13] E. F. Steiner and A. K. Steiner, Topologies with T1 -complements, Fundamenta Mathe-
maticae 61 (1967), 23-28.
[14] R. M. Stephenson, Initially κ-compact and related spaces, in Handbook of Set-theoretic
Topology, K. Kunen and J. E. Vaughan, (eds.), (North Holland, Amsterdam, 1984).
[15] M. G. Tkachenko, V. V. Tkachuk, R. G. Wilson and I. V. Yaschenko, No submaximal
topology on a countable set is T1 -complementary, Proceedings of the American Mathe-
matical Society 128 no. 1 (1999), 287-297.
[16] R. Valent and R. E. Larson, Basic intervals in the lattice of topologies, Duke Math. J.
39 (1972), 401-411.
[17] R. C. Walker, The Stone-Čech Compactification, (Springer-Verlag, New York, 1974).
[18] S. Willard, General Topology, (Addison Wesley, Reading, Mass., 1970).

Received May 2003

Accepted November 2003

O. T. Alas (alas@ime.usp.br)
Instituto de Matemática e Estatı́stica, Universidade de São Paulo, Caixa Postal
66281, 05311-970 São Paulo, Brasil.

R. G. Wilson (rgw@xanum.uam.mx)
Departamento de Matemáticas, Universidad Autónoma Metropolitana, Unidad
Iztapalapa, Avenida San Rafael Atlixco, #186, Apartado Postal 55-532, 09340,
México, D.F., México.