Professional Documents
Culture Documents
ANDREA MEDINI
For all undefined topological notions, see [3]. Recall that a subset of
a topological space is clopen if it is closed and open.
Definition 1. A space X is CLP-compact if every cover of X consisting
of clopen sets has a finite subcover.
Using methods of Steprāns and Šostak (see Section 4 in [4]), we will
construct a non-CLP-compact product P of 2c Hausdorff spaces such
that every finite subproduct of P is CLP-compact. More precisely,
using an idea from [1], we will make sure that no infinite subproduct of
P is CLP-compact; see also Example 8.28 in [6]. Our example answers
the first half of Question 6.3 in [4]; see also Question 7.1 in [2].
Given a family F of non-empty closed subsets of N∗ , define the topo-
logical space X(F ) = N ∪ F as follows.
• The singleton {n} is open for every n ∈ N.
• The open neighborhoods of K ∈ F are those in the form {K} ∪
A, where A ⊆ N is such that K ⊆ A∗ .
Q
Our example will be in the Q form P = α∈2c X(Fα ). Given S ⊆ 2c , we
will use the notation PS = α∈S X(Fα ) for the subproduct.
Proposition 2 (Steprāns and Šostak). Assume that F consists of pair-
wise disjoint sets. Then X(F ) is a Hausdorff space.
Proof. Use the fact that disjoint closed sets in N∗ can be separated by
a clopen set.
The following condition will guarantee the CLP-compactness of each
X(Fα). For the proof of Proposition 3, see Lemma 4.1 in [4].
o(F ) For every infinite and coinfinite A ⊆ N there exists K ∈ F such
that K ∩ A∗ 6= ∅ and K ∩ (N \ A)∗ 6= ∅.
Date: May 13, 2010.
1
2 ANDREA MEDINI
Proposition 6. Assume that for every F ∈ [2c ]<ω and every clopen
F
set U ⊆ PF , either U ⊆ SN or TNF ⊆ U for some N ∈ ω. Then PF is
CLP-compact for every F ∈ [2c ]<ω .
Proof. We wil use induction on n = |F |. The case n = 1 is obvi-
ous. So assume that PF is CLP-compact for every F ∈ [2c ]≤n and
let F ∈ [2c ]n+1 . By Proposition 5, it is enough to prove that PF is
CLP-rectangular.
First we will show that if U ⊆ PF is a clopen set and x ∈ U has
at least one coordinate in N, say coordinate α, then there exists a
clopen rectangle R ⊆ U such that x ∈ R. It is easy to check that the
cross-section
V = {y ∈ PF \{α} : y ∪ {(α, xα )} ∈ U}
is clopen in PF \{α} . Therefore, by the inductive hypothesis and Propo-
sition 5, there exists a clopen rectangle Q ⊆ V such that x\{(α, xα )} ∈
A NON-CLP-COMPACT PRODUCT SPACE 3
Q. Now just let R = Q × {xα }{α} . This observation takes care of the
F
case U ⊆ SN .
On the other hand, if U ⊆ PF is a clopen set such that TNF ⊆ U, then
for every x ∈ U with no coordinate in N the desired clopen rectangle
is R = TNF itself.
We will also need the following definitions.
[
SFN = SN
F
∩ (NF ) = (NF \{α} × N {α} ).
α∈F
K(f, g) There exist Kα0 , . . . , Kαn−1 with Kαi ∈ Fαi for each i ∈ n such
that f ∩(Aα0 ×· · ·×Aαn−1 ) 6= ∅ and g ∩(Aα0 ×· · ·×Aαn−1 ) 6= ∅
whenever Aα0 , . . . , Aαn−1 ⊆ N are such that Kαi ⊆ A∗αi for every
i ∈ n.
Proposition 9. Fix F ∈ [2c ]n with n ≥ 2. Assume that K(f, g) holds
for every F -diagonal pair (f, g) of functions. If U ⊆ PF is a clopen
set, then either U ∩ (NF ) ⊆ SFN or TFN ⊆ U ∩ (NF ) for some N ∈ ω.
Proof. Assume, in order to get a contradiction, that U ∩(NF ) * SFN and
TFN * U ∩ (NF ) for every N ∈ ω. Then it is possible to construct (in ω
steps) diagonal subsets D and E of NF \{αn−1 } together with functions
f : D −→ N and g : E −→ N such that d ∪ {(αn−1, f (d))} ∈ U
whenever d ∈ D and e ∪ {(αn−1 , g(e))} ∈ / U whenever e ∈ E.
Now let Kα0 , . . . , Kαn−1 be as given by K(f, g). Define x ∈ PF by
setting xα = Kα for every α ∈ F . It is easy to see that x ∈ U ∩(PF \ U),
which contradicts the fact that U is clopen.
Theorem 10. Assume that the family {Fα : α ∈ 2c } satisfies the
following conditions.
• For every F ∈ [2c ]n with n ≥ 2, the condition K(f, g) holds for
every F -diagonal pair (f, g) of functions.
• The condition o(Fα ) holds for every α ∈ 2c .
Then PF is CLP-compact for every F ∈ [2c ]<ω .
Proof. We will show that the hypothesis of Proposition 6 holds. If
|F | = 1, this follows from Proposition 3; if |F | ≥ 2, this follows from
Proposition 9, Lemma 7 and Lemma 8.
2. Infinite subproducts
In this section we will show how to ensure that PC is non-CLP-
compact whenever C ∈ [2c ]ω . Consider the following condition. Recall
that a family L is linked if K ∩ L 6= ∅ whenever K, L ∈ L.
® For every DQ ∈ [2c ]ω , the family L = {xα : α ∈ D} is not linked
for any x ∈ α∈D Fα .
Theorem 11. Assume that condition ® holds. Fix C ∈ [2c ]ω . Then
PC can be written as the disjoint union of infinitely many of its non-
empty clopen subsets.
Proof. Let C = {αk : k ∈ ω} be an enumeration without repetitions.
For each n ∈ ω, define
Un = {x ∈ PC : xαk = n whenever 0 ≤ k ≤ n}.
A NON-CLP-COMPACT PRODUCT SPACE 5
3. The construction
The next theorem guarantees the existence of our example: finite
subproducts will be CLP-compact by Theorem 10 and infinite sub-
products will be non-CLP-compact by Theorem 11.
Theorem 12. There exists a family {Fα : α ∈ 2c } satisfying the fol-
lowing conditions.
• Each Fα consists of pairwise disjoint subsets of N∗ of size 2.
• The condition o(Fα ) holds for every α ∈ 2c .
• For every F ∈ [2c ]n with n ≥ 2, the condition K(f, g) holds for
every F -diagonal pair (f, g) of functions.
• The condition ® holds.
Proof. Enumerate as {(fη , gη ) : η ∈ 2c } all pairs of diagonal functions,
say fη : Dη −→ N and gη : Eη −→ N, where Dη and Eη are diagonal
subsets of NF \{αn−1 } for some F = F (η) ∈ [2c ]n with n ≥ 2, say F =
{α0 , . . . , αn−1 } with α0 < · · · < αn−1 .
We will construct S Fα for α ∈ 2c by transfinite recursion: in the end
we will set Fα = ξ∈2c Fαξ for each α. Start with Fα0 = ∅ for each α.
By induction, we will make sure that the following requirements are
satisfied.
(1) Fαη ⊆ Fαµ whenever η ≤ µ ∈ 2c .
6 ANDREA MEDINI
S S
(2) | α∈2c Fαη | < 2c for every η ∈ 2c .
(3) We will take care of o(Fη ) at stage ξ = η + 1: that is, each
witness K is already in Fηη+1 .
(4) We will take care of K(fη , gη ) at stage ξ = η + 1: that is, the
witnesses Kαi are already in Fαη+1 .
S i