AlasWilsonAGT.dvi


@
Applied General Topology

c© Universidad Politécnica de Valencia

Volume 8, No. 2, 2007

pp. 273-281

On complete accumulation points of discrete
subsets

Ofelia T. Alas and Richard G. Wilson∗

Abstract. We introduce a class of spaces in which every discrete
subset has a complete accumulation point. Properties of this class

are obtained and consistent examples are given to show that this class

differs from the class of countably compact and the class of compact

spaces. A number of questions are posed.

2000 AMS Classification: Primary 54A25, 54A35, 54D99

Keywords: Discrete subset, complete accumulation point, compact space,
countably compact space, discretely complete space, US-spaceWrite here some
important concepts used in your paper.

1. Introduction

A classical theorem of General Topology states that a Hausdorff space is
compact if and only if each infinite subset has a complete accumulation point
(for details, we refer the reader to [3], 3.12.1). Additionally, it was shown in
[14] that a Hausdorff space is compact if and only if the closure of every discrete
subspace is compact. On comparing these results, it is natural to ask whether
one can characterize compactness in terms of complete accumulation points of
discrete sets:

Question 1.1. Is it true that if every discrete subspace of a Hausdorff space
X has a complete accumulation point in X, then X is compact?

Shortly we shall see that the answer to this question is consistently, No. We
make the following definition.

Definition 1.2. A T1-space X is said to be discretely complete if every infinite
discrete subspace has a complete accumulation point in X.

∗Research supported by Consejo Nacional de Ciencia y Tecnoloǵıa (México), grant 38164-
E and Fundação de Amparo a Pesquisa do Estado de São Paulo (Brasil)



274 O. T. Alas and R. G. Wilson

Clearly every compact topological space is discretely complete and each dis-
cretely complete space is countably compact. Furthermore, since an accumula-
tion point of a countable subset of a T1-space is a complete accumulation point
of that set, it follows that:

Remark 1.3. A countably compact T1-space with countable spread is dis-
cretely complete; in particular, each hereditarily separable countably compact
T1-space is discretely complete.

We do not know if there exists a model of ZF C in which every discretely
complete Hausdorff space is compact. Since a discretely complete space is
countably compact, and it is well-known that a countably compact, linearly
Lindelöf space is compact, it is immediate that a completely discrete, linearly
Lindelöf space is compact. Furthermore, a space is linearly Lindelöf if and only
if every uncountable subset of regular cardinality has a complete accumulation
point. So a non-compact T1-space which is discretely complete must contain
some (non-discrete) subset of uncountable regular cardinality which has no
complete accumulation point.

In [10], Ostaszewski constructed a non-compact, perfectly normal, heredi-
tarily separable, countably compact Tychonoff space assuming CH + ♣(≡ ♦);
it follows immediately from Remark 1.3 that this is an example of a discretely
complete non-compact space.

In [4], Fedorčuk constructed using ♦, a hereditarily separable compact Haus-
dorff space in which every infinite closed set has cardinality 2c. In the sequel,
this space is used to show that a product of two discretely complete spaces
need not be countably compact.

Both of the examples we have just mentioned are S-spaces (regular, hered-
itarily separable but not hereditarily Lindelöf) and it is well-known that the
existence of an S-space is independent of ZF C. We also note that the exis-
tence of a discretely complete, non-compact regular space of countable spread
is independent of ZF C since it was shown in [2] that under the Proper Forcing
Axiom each regular countably compact space of countable spread is compact.
Furthermore, Peter Nyikos has informed us that the same result holds in the
case of countably compact Hausdorff spaces. However, it seems not to be known
whether or not there exists in ZF C a compact Hausdorff space X such that
|X| > 2s(X) and as we shall show later, the existence of such a space gives rise
to a non-compact discretely complete Hausdorff space.

All spaces in the sequel are assumed to be T1. All notation and terminology
not specifically defined below can be found in [3], [7] and [11].

2. The Results

The examples of Fedorčuk [4] and Ostaszewski [10] cited above in Section
1, both have countable spread, but a simple construction allows us to produce
discretely complete, non-compact Tychonoff spaces of arbitrary spread.



On complete accumulation points of discrete subsets 275

Lemma 2.1. If in a model of ZF C there is a discretely complete, non-compact
Tychonoff space of countable spread, then for each cardinal κ there is a dis-
cretely complete, non-compact Tychonoff space of spread κ.

Proof. Let X be a discretely complete non-compact Tychonoff space of count-
able spread and let Z = (κ × X) ∪ {∞} where κ has the discrete topology,
κ × X has the product topology and basic neighbourhoods of ∞ are of the
form {∞} ∪ [(κ \ F ) × X], where F ⊆ κ is finite. We denote the projection
from κ × X → κ by π. Clearly s(Z) = κ and if D ⊆ Z is an infinite discrete
set, there are two possibilities:

1) If π(D) is infinite, then ∞ is a complete accumulation point of D, or
2) If π(D) = {α0, . . . , αn} is finite, then D is countable and for some j ∈

{0, . . . , n}, D ∩ ({αj} × X) is countably infinite and hence D has a complete
accumulation point in {αj} × X. �

The class of discretely complete spaces, like that of compact space is closed
under taking continuous images and closed subsets.

Lemma 2.2. The continuous image of a discretely complete space is discretely
complete.

Proof. Suppose X is discretely complete and f : X → Y is continuous and
surjective. Let D = {dα : α ∈ κ} be a discrete subset of Y ; then if for each
α ∈ κ we pick xα ∈ f

−1[{dα}] it is clear that {xα : α ∈ κ} is a discrete subset of
X and hence must have a complete accumulation point, p say. Then if q = f (p)
and V is an open neighbourhood of q, it follows that |{α : xα ∈ f

−1[V ]}| = κ
and hence |{α : dα ∈ V }| = κ, showing that q is a complete accumulation point
of D. �

The proof of the following trivial result is left to the reader.

Lemma 2.3. A closed subspace of a discretely complete space is discretely
complete

However, unlike the class of compact spaces, the class of discretely complete
spaces is not closed under the taking of products.

Example 2.4. It is consistent with ZFC that there exist two discretely com-
plete Tychonoff spaces whose product is not countably compact.

Proof. Let X denote the (above mentioned) compact space constructed in [4]
under ♦. Let S be a fixed countably infinite subset of X; note that since X has
no non-trivial convergent sequences, S is not compact and hence not countably
compact. We enumerate the infinite discrete subsets of S as D = {Dα : α < c}.
Since |cl(D0)| = 2

c, we may choose distinct points a00, b
0
0 ∈ cl(D0) \ S. If for

some α < c, we have chosen sets A0,α = {a
0
β : β < α} and B0,α = {b

0
β : β < α},

such that A0,α ∩ B0,α = ∅ and a
0
β , b

0
β ∈ cl(Dβ ) \ S for each β < α, then

again, since |cl(Dα)| = 2
c and |S ∪ A0,α ∪ B0,α| ≤ c, we can choose distinct

points a0α, b
0
α ∈ cl(Dα) \ (S ∪ A0,α ∪ B0,α). Let K1 = S ∪ {a

0
α : α < c} and

L1 = S ∪ {b
0
α : α < c}.



276 O. T. Alas and R. G. Wilson

Clearly K1 and L1 have cardinality c and hence |[K1]
ω| = |[L1]

ω| = c.
Enumerate the countably infinite discrete subsets of K1 and L1 as {D1,α : α <
c} and {E1,α : α < c} respectively. Since, for each α < c, the sets cl(D1,α)
and cl(E1,α) have cardinality 2

c we can repeat the process described in the
previous paragraph so as to obtain, for each α < c, sets A1,α = {a

1
β : β < c}

and B1,α = {b
1
β : β < c}, such that A1,α∩B1,α = ∅ and a

1
β ∈ cl(D1,β )\(K1∪L1)

and b1β ∈ cl(E1,β ) \ (K1 ∪ L1); finally. let K2 = K1 ∪ A1,α and L2 = L1 ∪ B1,α.

Having defined sets Kβ and Lβ for each β < α < ω1 = c, if α is a limit
ordinal, then define Kα =

⋃
{Kβ : β < α}, and Lα =

⋃
{Lβ : β < α}; if α is

a successor ordinal then repeat the process of the previous paragraph. Thus
we define Kα and Lα of X for each α < ω1 = c. Clearly S ⊆ Kα ⊆ Kγ and
S ⊆ Lα ⊆ Lγ whenever α < γ < c and Kα ∩ Lα = S. Let K =

⋃
{Kα : α < c}

and L =
⋃
{Lα : α < c}. If T is a countable subset of K (respectively, L), then

there is some α < c such that T ⊆ Kα (respectively, T ⊆ Lα) and hence T has
an accumulation point in Kα+1 (respectively, Lα+1). Thus both K and L are
countably compact. Since X is hereditarily separable, it follows that both K
and L have countable spread and are not compact since they are proper dense
subsets of cl(S). It follows from Remark 1.3 that both K and L are discretely
complete. However, {(s, s) : s ∈ S} ∼= S is a closed subspace of K × L which is
not countably compact. Hence K × L is not countably compact. �

As our next result shows, even the product of a compact Hausdorff space
and a discretely complete space need not be discretely complete (although it
will certainly be countably compact).

Theorem 2.5. If X is a discretely complete, but non-compact T1-space, then
there is a compact Hausdorff space Y such that X×Y is not discretely complete.

Proof. Since X is not compact, there is a subset A = {aα : α ∈ κ} ⊆ X
which has no complete accumulation point. Since X is discretely complete,
it is countably compact, and hence κ must be uncountable. Let Y be the
Alexandroff compactification of the discrete space D(κ) of cardinality κ and let
{dα : α ∈ κ} be an enumeration of D(κ). We consider the set C = {(aα, dα) :
α ∈ κ} ⊆ X × Y . Since {dα} is open for each α ∈ κ, it follows that C is
discrete and if X × Y were discretely complete, then C would have a complete
accumulation point, say p = (x0, y0). Thus if U is a neighbourhood of x0 and
V is a neighbourhood of y0, then |{α : (aα, dα) ∈ U × V }| = κ. Thus for each
neighbourhood U of x0, |{α : aα ∈ U}| = κ, showing that x0 is a complete
accumulation point of A, a contradiction. �

Corollary 2.6. The property of being discretely complete is not an inverse
invariant of perfect mappings.

In contrast to Theorem 2.5, we have the following results. Recall that a
space is initially κ-compact if every open cover of size at most κ has a finite
subcover. The following two lemmas are immediate consequences of Theorems
2.2 and 5.2 of [12].



On complete accumulation points of discrete subsets 277

Lemma 2.7. A space is initially κ-compact if and only if every infinite subset
of cardinality at most κ has a complete accumulation point.

Lemma 2.8. If X is compact and Y is initially κ-compact, then X × Y is
initially κ-compact.

Theorem 2.9. If X is a compact space of weight κ and Y is a discretely
complete space which is initially κ-compact, then X × Y is discretely complete.

Proof. Note that Lemma 2.8 implies that X×Y is initially κ-compact. Suppose
that D = {(xα, yα) : α ∈ λ} is an infinite discrete subset of X × Y . There are
three cases to be considered.

1) If λ ≤ κ, then since X × Y is initially κ-compact, it follows from Lemma 2.7
that D has a complete accumulation point in X × Y .

2) If cof (λ) > κ, then fix a base B of X of size κ and for each B ∈ B, define
IB = {α ∈ λ : there is an open neighbourhood

Wα of yα such that (B × Wα) ∩ D = {(xα, yα)}}.

Since cof (λ) > κ, there is some B ∈ B such that |IB| = λ. The set YB =
{yα : α ∈ IB} is discrete in Y and hence has a complete accumulation point
q ∈ Y . But then, if for each x ∈ X, (x, q) is not a complete accumulation point
of D, then for each x ∈ X we can find open neighbourhoods Ux of x and Vx
of q such that |(Ux × Vx) ∩ D| < |D|. The open cover {Ux : x ∈ X} of X
has a finite subcover {Ux1, . . . , Uxn} and if we let V =

⋂
{Vxk : 1 ≤ k ≤ n}, it

follows that |(X ×V )∩D| < |D| which contradicts the fact that q is a complete
accumulation point of YB.

3) If λ > κ ≥ cof (λ), then we can find regular cardinals {λα : α ∈ cof (λ)} such
that κ < λα < λ and sup{λα : α ∈ cof (λ)} = λ. Now write D =

⋃
{Dα : α ∈

cof (λ)} where |Dα| = λα. By 2), each of the discrete sets Dα has a complete
accumulation point (pα, qα) ∈ X × Y , and since this latter space is initially κ-
compact, it again follows from Lemma 2.7 that the set {(pα, qα) : α ∈ cof (λ)}
has a complete accumulation point (p, q) ∈ X × Y . Now any neighbourhood
V of (p, q) contains cof (λ)-many points (pα, qα) and hence λα-many points of
Dα for cof (λ)-many α. It follows that (p, q) is a complete accumulation point
of D. �

Corollary 2.10. If X is a compact metrizable space and Y is discretely com-
plete, then X × Y is discretely complete.

Proof. Since X is metrizable, w(X) = ω. The space Y is discretely complete,
hence countably compact, that is to say, initially ω-compact; the result now
follows from the theorem. �

We now show that a construction very similar to that used in Example 2.4
can in fact be carried out on any compact Hausdorff space in which |X| >
2s(X) in order to construct a non-compact discretely complete space. The
construction is reminiscent of the classical construction of a countably compact,
non-compact dense subspace of βω of size c (see [5], 9.15).



278 O. T. Alas and R. G. Wilson

Theorem 2.11. In any model of ZFC in which there exists a compact Haus-
dorff space X with |X| > 2s(X), there exists a non-compact discretely complete
Tychonoff space.

Proof. Suppose κ = s(X); by Theorem 2.17 of [7], there is a dense subspace
E0 ⊆ X of cardinality at most 2

κ. We enumerate the discrete subsets of E0 as
{D0α : α ∈ λ0}, where λ0 ≤ |E0|

s(X) ≤ (2κ)κ = 2κ < |X|. Since X is compact,
each subset D0α has a complete accumulation point x

0
α ∈ X; let E1 = E0 ∪{x

0
α :

α ∈ λ0}. Clearly |E1| ≤ max{λ0, 2
κ} < |X|. Having constructed subsets Eα

for each α < β < κ+, with the property that |Eα| ≤ 2
κ < |X| and such that

Eµ ⊆ Eν whenever µ < ν, we construct Eβ as follows:
If β is a limit ordinal then Eβ =

⋃
{Eα : α ∈ β}.

If β = γ + 1, then since |Eγ| ≤ 2
κ, we can enumerate the discrete subsets of

Eγ as {D
γ
α : α ∈ λγ} where λγ ≤ 2

κ < |X|. Again, since X is compact, each
of the sets Dγα has a complete accumulation point in X and for each α ∈ λγ
we choose one such, xγα. Let Eβ = Eγ ∪ {x

γ
α : α ∈ λγ}. It is immediate that

|Eβ| ≤ 2
κ < |X|.

Now let E =
⋃
{Eα : α < κ

+}; clearly |E| ≤ κ+.2κ = 2κ and so E  X.
Furthermore, if S ⊆ E is discrete, then |S| ≤ κ and hence there is some α ∈ κ+

such that S ⊆ Eα and so S has a complete accumulation point in Eα+1 and
hence in E as well. �

We note in passing that by applying the technique of Gryzlov [6], the space
E in the previous theorem can even be made to be normal. Before stating a
generalization of this result we need some definitions.

Recall that a space X is a KC-space if every compact subspace of X is
closed, X is an SC-space (see for example, [1]) if every convergent sequence
together with its limit forms a closed subset of X and X is a US-space (see,
[8]) if every convergent sequence in X has a unique limit. It is easy to see that
KC ⇒ SC ⇒ U S ⇒ T1.

A similar technique to that used in Theorem 2.11 can be used to prove the
following result. We leave the details to the reader.

Theorem 2.12. In any model of ZFC in which there exists a compact KC-
space X with |X| > d(X)s(X), there exists a non-compact discretely complete
KC-space.

As mentioned in the Introduction, we do not know if there is a ZF C example
of a discretely complete, non-compact Hausdorff space; however, discretely
complete, non-compact U S-spaces exist in ZF C.

Example 2.13. There is in ZFC a non-compact U S-space which is discretely
complete.



On complete accumulation points of discrete subsets 279

Proof. Our aim is to define a topology σ on the set ω1 such that (ω1, σ) is a
non-compact, discretely complete U S-space; we begin by defining a topology
τ (used in [13]) generated by the following sub-base:

{{β : β < α} : α ∈ ω1} ∪ {C : ω1 \ C is finite}.

Clearly τ is a T1-topology which is weaker than the order topology on ω1
and hence (ω1, τ ) is countably compact but not Lindelöf, since the open cover
{{β : β < α} : α ∈ ω1} has no countable subcover. Furthermore, if A ⊂ ω1
has order type (induced by the order on ω1) greater than or equal to ω + 1,
then A is not discrete, and hence every discrete subset of (ω1, τ ) is countable.
That this space is discretely complete but not Lindelöf and hence not compact
is now a consequence of the remarks following Definition 1.2.

However, in (ω1, τ ), every injective sequence converges to an uncountable
number of points. To obtain a U S-space we use a construction used for compact
spaces by Künzi and van der Zypen, [8].

Let A = {Aα : α ∈ I} be a maximal almost disjoint (MAD) family of
injective sequences in ω1, where Aα = {x

n
α : n ∈ ω} and for each α ∈ I, choose

a limit ℓα 6∈ Aα. Denote the set {x
n
α : n ≥ m} ∪ {ℓα} by A

m
α and let σ be

the topology generated by the subbase τ ∪ {X \ Amα : m ∈ ω, α ∈ I}. We
claim that (ω1, σ) is a U S-space which is discretely complete and since it is
not Lindelöf, it is not compact. In order to show that the space is discretely
complete, we first show that its spread is countable. To this end, suppose that
B is an uncountable subset of ω1 and we write B =

⋃
{Bα : α ∈ ω1} where the

sets Bα are mutually disjoint and countably infinite. Consider the countably
infinite set B0 ≡ Bβ0 . Since A is a MAD family, there must exist Aα0 ∈ A
such that B0 ∩ Aα0 is infinite. Let b0 = sup(Aα0 ∪ B0 ∪ {ℓα0}) < ω1, clearly
B ∩ (b0, ω1) is uncountable. Let β1 = min{β ∈ ω1 : |Bβ ∩ (b0, ω1)| = ω}.
Again, since A is a MAD family, there is some Aα1 ∈ A such that Aα1 ∩
Bβ1 ∩ (b0, ω1) is infinite; let b1 = sup(Aα1 ∪ Bβ1 ∪ {ℓα1}). Continuing this
process, we obtain a family {Bβn : n ∈ ω} of countably infinite, mutually
disjoint subsets of B, each of which intersects an element of A in an infinite
set. Let S =

⋃
{Aαn ∩ Bβn ∩ (bn−1, ω1) : n ∈ ω} and let b ∈ B ∩ (sup(S), ω1).

We claim that b ∈ clσ(S), thus showing that B is not discrete. Suppose to
the contrary that b 6∈ clσ(S); then there is some basic closed set containing
S but not b. Since all τ -closed sets contain a cofinal interval of ω1, it follows
that there must be some finite subset of A, say {Aγ1 , Aγ1 , . . . , Aγn} such that
b 6∈

⋃
{Aγm : 1 ≤ m ≤ n} ∪ {ℓγm : 1 ≤ m ≤ n} ⊇ S. Clearly then there is some

j ∈ {1, . . . , n} and some k ∈ ω such that Aγj ∩ Aαk ∩ Bβk is infinite, which
since there are only finitely many possible such j but infinitely many such k,
contradicts the fact that A is an almost disjoint family. Thus b ∈ cl(S) and
to show that (ω1, σ) is discretely complete, it now suffices to show that it is
countably compact. However, if T = {tn : n ∈ ω} is a countably infinite subset
of ω1, then there is some Aλ ∈ A such that Aλ ∩ T is infinite. It is immediate
that ℓλ ∈ cl(T ).

Finally, the proof that (ω1, σ) is a U S-space follows exactly as in [8]. �



280 O. T. Alas and R. G. Wilson

Recall that a space is weakly Lindelöf if every open cover has a countable
dense subsystem.

The spaces βω and ω∗ = βω \ ω are the source of many examples and
counterexamples, thus the following is of interest.

Theorem 2.14. [CH] For each p ∈ ω∗, ω∗ \ {p} is not discretely complete.

Proof. In the proof of Corollary 1.5.4 of [9] it is shown (in ZF C) that ω∗ \ {p}
is not weakly Lindelöf. Hence there is an open cover U of ω∗ \ {p} with no
countable dense subsystem. Now, using CH, we may enumerate U as

{Uα : α < ω1}.

Choose x0 ∈ U0 and let α0 = 0. Suppose now that for some λ < ω1, we have
chosen points xγ , indices αγ ∈ ω1 and elements Uαγ ∈ U for each γ < λ. Note
that since αγ is a countable ordinal for all γ < λ, it follows that cl(

⋃
{Uξ : ξ <

αγ , γ < λ}) 6= ω
∗ \ {p} and hence we can define

αλ = min{β < ω1 : Uβ \ cl(
⋃

{Uξ : ξ < αγ , γ < λ}) 6= ∅}

and choose xλ ∈ Uαλ \ cl(
⋃
{Uξ : ξ < αγ , γ < λ}).

Note that this construction ensures that xλ 6∈ cl{xγ : γ < λ} and xγ 6∈
cl(Uαλ ) for each γ > λ and so {xα : α ∈ ω1} is discrete. Furthermore, the
discrete subset {xα : α ∈ ω1} has no complete accumulation point in ω

∗ \ {p},
since each of the open sets Uα contains only countably many points of the set
{xα : α ∈ ω1}. �

3. Open Questions

Below, we repeat the principal open questions regarding discretely complete
spaces.

Question 3.1. Is it consistent with ZFC that every Tychonoff discretely com-
plete space is compact?

Question 3.2. Is there in ZFC, an SC (or even a KC or Hausdorff ) example
of a discretely complete space which is not compact?

References

[1] O. T. Alas and R. G. Wilson, Minimal properties between T1 and T2, Houston J. Math.
32, no. 2 (2006), 493–504.

[2] Z. Balogh, A. Dow, D. H. Fremlin and P. J. Nyikos, Countable tightness and proper
forcing, Bulletin A.M.S. (N. S.) 19, no. 1 (1988), 295–298.

[3] R. Engelking, General Topology, Heldermann Verlag, Berlin 1989.
[4] V. Fedorčuk, On the cardinality of hereditarily separable compact Hausdorff spaces, Soviet

Math. Doklady 16 (1975), 651–655.
[5] L. Gillman and M. Jerison, Rings of Continuous Functions, Van Nostrand, Princeton,

1960.



On complete accumulation points of discrete subsets 281

[6] A. Gryzlov, Cardinal invariants and compactifications, Comm. Math. Univ. Carolinae
35, no. 1 (1994), 403–408.

[7] I. Juhasz, Cardinal Funcions in Topology - Ten years later, Mathematical Centre Tracts
123, Mathematisch Centrum, Amsterdam, 1980.

[8] H.-P. Künzi and D. van der Zypen, Maximal (sequentially) compact topologies, preprint,
http : //arxiv.org/PS cache/math/pdf/0306/0306082.pdf

[9] J. van Mill, An introduction to βω, in Handbook of Set-theoretic Topology, North
Holland, Amsterdam, 1984.

[10] A. J. Ostaszewski, On countably compact perfectly normal spaces, J. London Math. Soc.
14 (1976), 501–516.

[11] M. E. Rudin, Lectures on Set Theoretic Topology, CBMS, Regional Conference Series
in Mathematics, No. 23, A. M. S., Providence, 1975.

[12] R. M. Stephenson, Initially κ-compact and related spaces, in Handbook of Set-theoretic
Topology, North Holland, Amsterdam, 1984.

[13] A. Tamariz and R. G. Wilson, Example of a T1 topological space without a Noetherian
base, Proceedings A.M.S. 104, no. 1 (1988), 310–312.

[14] V. V. Tkachuk, Spaces that are projective with respect to classes of mappings, Trans.
Moscow Math. Soc. 50 (1988), 139–156.

Received May 2006

Accepted March 2007

O. T. Alas (alas@ime.usp.br)
Instituto de Matemática e Estat́ıstica, 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