1 Introduction
The main motivation for investigation of regular first-countable countably compact spaces is the following problem of Nyikos which is listed among 20 central problems in Set-theoretic Topology by Hrušak and Moore [Reference Hrušák and Moore16].
Problem 1.1 (Nyikos)
Does ZFC imply the existence of a regular separable first-countable countably compact non-compact space?
Following [Reference Bardyla and Zdomskyy3], a regular separable first-countable countably compact space is called a Nyikos space. Consistent examples of noncompact Nyikos spaces were constructed by Franklin and Rajagopalan [Reference Franklin and Rajagopalan14] (under
$\omega _1=\mathfrak t$
) and by Ostaszewski [Reference Ostaszewski25] (under
$\mathfrak b=\mathfrak c$
). Weiss [Reference Weiss35] showed that under MA every perfectly normal Nyikos space is compact. Proper Forcing Axiom or, simply, PFA is a stronger version of Martin’s Axiom. Basic information about PFA can be found in [Reference Moore20]. For fruitful applications of PFA in Topology see [Reference Dow10–Reference Dow and Tall12, Reference Tall30, Reference Viale and Weiss34]. The following result was proved by Nyikos and Zdomskyy [Reference Nyikos and Zdomskyy24].
Theorem 1.2 (PFA)
Every normal Nyikos space is compact.
A space X is called
$\mathbb R$
-rigid if each continuous real-valued function on X is constant. For more about rigid spaces see [Reference Bardyla and Osipov2, Reference Iliadis and Tzannes17, Reference Tzannes31–Reference Tzannes33]. Another problem related to Nyikos spaces appeared in [Reference Pearl29, Problem C65].
Problem 1.3 (Tzannes)
Does there exist a regular (separable, first countable) countably compact
$\mathbb R$
-rigid space?
Taking into account the properties in brackets, Tzannes problem can be considered as an ultimate version of Nyikos problem, as
$\mathbb R$
-rigid spaces, being not Tychonoff, are not locally compact.
Bardyla and Osipov [Reference Bardyla and Osipov2] constructed a ZFC example of a regular countably compact
$\mathbb R$
-rigid space. Bardyla and Zdomskyy [Reference Bardyla and Zdomskyy3] obtained the following answer to Problem 1.3 under the consistent assumption (
): “
$\omega _1=\mathfrak t<\mathfrak b=\mathfrak c$
and there exists a
$P_{\mathfrak c}$
-point in
$\beta (\omega )$
”.
Theorem 1.4 (Bardyla, Zdomskyy)
There exists an
$\mathbb R$
-rigid Nyikos space.
The proof of Theorem 1.4 had two steps. During the first one we constructed a regular separable first-countable
$\mathbb R$
-rigid space X of cardinality
$<\mathfrak c$
. The second step was to embed densely the space X into a first-countable countably compact space Y. Note that Y will be automatically an
$\mathbb R$
-rigid Nyikos space. This was done using the following result proved in [Reference Bardyla and Zdomskyy3] under the assumption (
$\heartsuit $
): “
$\mathfrak b=\mathfrak c$
and there exists a
$P_{\mathfrak c}$
-point in
$\beta (\omega )$
”.
Theorem 1.5 (Bardyla, Zdomskyy)
$(\heartsuit )$
Each regular first-countable space of cardinality
$< \mathfrak c$
embeds densely into a regular first-countable countably compact space.
Similar problems concerning embedding of topological spaces into first-countable compact-like spaces are known in General Topology. For instance, the most relevant to this paper is the following one posed in [Reference Stephenson27].
Problem 1.6 (Stephenson)
Does every locally feebly compact first-countable regular space embed densely into a feebly compact first-countable regular space?
Problem 1.6 was solved affirmatively by Simon and Tironi [Reference Simon and Tironi26]. In case of Tychonoff spaces the following result was obtained in [Reference Terada and Terasawa28].
Theorem 1.7 (Terada, Terasawa)
Each Tychonoff first-countable locally pseudocompact space embeds densely into a Tychonoff first-countable pseudocompact space.
The mentioned results motivate the following general question which we address in this paper.
Question 1.8. When does a space X embed into a regular first-countable compact-like space?
2 Preliminaries
By
$\mathbb N$
we denote the set of positive integers, i.e.,
$\mathbb N=\omega \setminus \{0\}$
. The cardinality of the continuum is denoted by
$\mathfrak c$
. The set of all infinite subsets of a countable set A is denoted by
$[A]^{\omega }$
. The set of all finite subsets of a set A is denoted by
$[A]^{<\omega }$
. A family
$\mathcal A\subset [\omega ]^\omega $
is called almost disjoint if
$|A\cap B|<\omega $
for each
$A,B\in \mathcal A$
. A family
$\mathcal S\subseteq [\omega ]^{\omega }$
is called splitting if for any
$A\in [\omega ]^\omega $
there exists
$S\in \mathcal S$
which splits A, i.e.,
$|A\cap S|=|A\cap (\omega \setminus S)|=\omega $
. For two subsets A and B of
$\omega $
we write
$A\subseteq ^* B$
or
$B\supseteq ^* A$
if
$|A\setminus B|<\omega $
. A family
$\mathcal T\subset [\omega ]^\omega $
is called a maximal tower if it is well-ordered by the relation
$\supseteq ^*$
and there is no set
$A\in [\omega ]^\omega $
such that
$A\subseteq ^* T$
for all
$T\in \mathcal T$
. For any functions
$f,g\in \omega ^\omega $
we write
$f\leq ^*g$
if
$f(n)\leq g(n)$
for all but finitely many
$n\in \omega $
. A family
$\Phi \subseteq \omega ^\omega $
is called unbounded if there exists no
$g\in \omega ^\omega $
such that
$f\leq ^* g$
for all
$f\in \Phi $
. We write
$f\equiv c$
if f is a constant function with value c.
We shall use the following cardinal characteristics of the continuum:
-
•
$\mathfrak s =\min \{|\mathcal S|: \mathcal S\subset [\omega ]^{\omega }$ is a splitting family
$\}$ ;
-
•
$\mathfrak t =\min \{|\mathcal T|: \mathcal T\subset [\omega ]^{\omega }$ is a maximal tower
$\}$ ;
-
•
$\mathfrak b =\min \{|\Phi |: \Phi \subset \omega ^{\omega }$ is an unbounded family
$\}$ .
It is well-known that
$\omega _1\leq \mathfrak t\leq \min \{\mathfrak s,\mathfrak b\}$
.
An ultrafilter
$\mathcal F$
on
$\omega $
is called a
$P_{\mathfrak c}$
point if
$\mathcal F$
has a base which forms a maximal tower of cardinality
$\mathfrak c$
. The existence of such an ultrafilter is consistent with ZFC and, in particular, follows from
$\mathfrak t=\mathfrak c$
.
Let
$X,Y$
be topological spaces. A map
$f:X\rightarrow Y$
is called an embedding of X into Y if the map
$f': X\rightarrow f(X)$
, obtained by restricting the range of f, is a homeomorphism. The weight of a topological space X is the smallest cardinal
$w(X)$
such that X has a base of cardinality
$w(X)$
.
A space X is called
-
• countably compact if each closed discrete subset of X is finite;
-
• pseudocompact if X is Tychonoff and each continuous real-valued function is bounded;
-
• feebly compact if every locally finite family of open subsets of X is finite.
It is known that countable compactness implies feeble compactness. In the case of Tychonoff spaces pseudocompactness is equivalent to feeble compactness.
The Pixley–Roy hyperspace
$\operatorname {PR}(X)$
of a space X is the set
$[X]^{<\omega }$
of all finite subsets of X endowed with the topology generated by the base consisting of the sets

where
$F\in [X]^{<\omega }$
and U is open in X. It can be easily checked that
$\operatorname {PR}(X)$
is Hausdorff and zero-dimensional, if X is Hausdorff. Moreover, if X is first-countable, then so is
$\operatorname {PR}(X)$
.
The notions used but not defined in this paper are standard and can be found in [Reference Bukovský7, Reference van Douwen, Kunen and Vaughan8, Reference Engelking13, Reference Kunen19].
3 Main results
The majority of the results in this section are equivalences, each being based on an “embedding theorem” combined with a “non-embedding theorem”, which are proved in Sections 4 and 5, respectively.
Theorem 3.1. The following assertions are equivalent:
-
(1)
$\omega _1=\mathfrak c$ .
-
(2) Every first-countable Tychonoff space of weight
$<\mathfrak c$ embeds into a Hausdorff first-countable compact space.
-
(3) Each separable first-countable locally compact normal space of cardinality
$<\mathfrak c$ embeds into a Hausdorff first-countable compact space.
Theorem 3.2. The following assertions are equivalent.
-
(1)
$\mathfrak b = \mathfrak c$ .
-
(2) Every Hausdorff locally compact first-countable space of weight
$< \mathfrak c$ can be densely embedded in a Hausdorff first-countable locally compact countably compact space.
-
(3) Every Hausdorff locally compact first-countable space of cardinality
$< \mathfrak c$ can be densely embedded in a Hausdorff first-countable countably compact space.
Observe that
$\mathfrak b=\mathfrak s=\mathfrak c$
follows from (
$\heartsuit $
). So, the following result generalizes Theorem 1.5.
Theorem 3.3. The following assertions are equivalent.
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every regular first-countable space of weight
$< \mathfrak c$ can be densely embedded in a regular first-countable countably compact space.
-
(3) Every regular first-countable space of cardinality
$< \mathfrak c$ can be densely embedded in a regular first-countable countably compact space.
Theorem 3.3 has the following zero-dimensional counterpart.
Theorem 3.4. The following assertions are equivalent.
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every Hausdorff zero-dimensional first-countable space of weight
$< \mathfrak c$ can be densely embedded in a Hausdorff zero-dimensional first-countable countably compact space.
-
(3) Every Hausdorff zero-dimensional first-countable space of cardinality
$< \mathfrak c$ can be densely embedded in a Hausdorff zero-dimensional first-countable countably compact space.
Combining Theorem 3.3 with the results from [Reference Bardyla and Zdomskyy3] we get the following.
Theorem 3.5. The following assertions are equivalent:
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every regular separable first-countable non-normal space of weight
$< \mathfrak c$ embeds into an
$\mathbb R$ -rigid Nyikos space.
-
(3) Every regular separable first-countable non-normal space of cardinality
$< \mathfrak c$ embeds into an
$\mathbb R$ -rigid Nyikos space.
Theorem 3.5 allows us to prove the following analog of Theorem 1.4 using a milder assumption.
Theorem 3.6 (
$\omega _1<\mathfrak b=\mathfrak s=\mathfrak c$
)
There exists an
$\mathbb R$
-rigid Nyikos space.
Theorems 1.2 and 3.5 imply the following corollary which shows a profound contrast in the behavior of normal and non-normal Nyikos spaces under PFA.
Corollary 3.7 (PFA)
The following assertions hold:
-
(1) Each normal Nyikos space is compact.
-
(2) Each regular separable first-countable space of weight
$<\mathfrak c$ embeds into an
$\mathbb R$ -rigid Nyikos space.
The following consistency result complements Theorem 1.5.
Theorem 3.8 (
$\heartsuit $
)
Every Tychonoff first-countable space of weight
$< \mathfrak c$
can be densely embedded into a Tychonoff first-countable countably compact space.
However, the following question remains open:
Question 3.9. Can the assumption
$(\heartsuit )$
be weakened to
$\mathfrak b=\mathfrak s=\mathfrak c$
in Theorem 3.8?
Turning to embeddings into pseudocompact spaces we obtain the following characterization.
Theorem 3.10. The following assertions are equivalent:
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every first-countable zero-dimensional Hausdorff space of weight
$<\mathfrak c$ embeds densely into a first-countable zero-dimensional pseudocompact space.
-
(3) Every first-countable zero-dimensional Hausdorff space of cardinality
$<\mathfrak c$ embeds densely into a first-countable zero-dimensional pseudocompact space.
Recall that a subspace A of the Cantor space is called a
$\lambda $
-set if each countable subset of A is
$G_{\delta }$
. As a by product we obtain the following characterization of
$\lambda $
-subsets of the Cantor space.
Theorem 3.11. A subspace X of the Cantor space is a
$\lambda $
-set if and only if the Pixley–Roy hyperspace
$\operatorname {PR}(X)$
embeds densely into a first-countable pseudocompact space.
4 Embedding theorems
We start with embeddings into countably compact spaces.
As we mentioned in the previous section, the assumption
$\mathfrak {s}=\mathfrak {b}=\mathfrak c$
is formally weaker than (
$\heartsuit $
). Indeed, fix a
$P_{\mathfrak c}$
-point p and a splitting family
$\mathcal C$
. To derive a contradiction, assume that
$|\mathcal C|<\mathfrak c$
. Consider the family
$\mathcal D=\mathcal C\cap p$
and find an arbitrary pseudointersection
$E\in p$
of the family
$\mathcal D\cup \{\omega \setminus C: C\in \mathcal C\setminus p\}$
. It is easy to check that there exists no
$C\in \mathcal C$
which splits E. This contradiction implies that
$\mathfrak s=\mathfrak c$
. On the other hand there exists a model of ZFC which satisfies
$\mathfrak {s}=\mathfrak b=\mathfrak c$
, but contains no
$P_{\mathfrak c}$
-points, see [Reference Brendle and Shelah6, Theorem 8]. Hence the next theorem generalizes Theorem 1.5.
Proposition 4.1 (
$\mathfrak {s} =\mathfrak {b}=\mathfrak c$
)
Let X be a regular first-countable space of weight
$\kappa <\mathfrak {c}$
. Then X can be densely embedded into a regular first-countable countably compact space.
Proof. Without loss of generality we can assume that the underlying set of X is disjoint with
$\mathfrak c$
. The first countability of X implies that
$|X|\leq \mathfrak c^{\omega }=\mathfrak c$
. If X is countably compact, then there is nothing to prove. Otherwise, let

Fix any bijection
$h: \mathcal {D}\cup [\mathfrak {c}]^{\omega }\rightarrow \mathfrak {c}$
such that
$h(a)\geq \sup (a)$
for any
$a\in [\mathfrak {c}]^{\omega }$
. It is easy to see that such a bijection exists. Next, for every
$\alpha \leq \mathfrak {c}$
we shall recursively construct a topology
$\tau _{\alpha }$
on
$X_{\alpha }\subseteq X\cup \alpha $
. For the sake of brevity we denote the space
$(X_{\alpha },\tau _{\alpha })$
by
$Y_{\alpha }$
. At the end, we will show that the space
$Y_{\mathfrak {c}}$
is regular first-countable countably compact and contains X as a dense subspace.
Let
$X_0=X$
. Since X is first-countable and regular there exists a base
$\mathcal B_0=\bigcup _{x\in X}B_{0}^x$
of the topology on X, where for each
$x\in X$
, the collection
$\mathcal B_{0}^x=\{U_{n,0}^x:n\in \omega \}$
is an open neighborhood base at x. With no loss of generality we can assume that
$|\mathcal B_{0}|<\mathfrak c$
;
$U_{0,0}^x=X$
for each
$x\in X$
; and
$\overline {U_{n+1,0}^x}\subset U_{n,0}^x$
for every
$n\in \omega $
and
$x\in X$
.
Assume that for each
$\alpha <\xi $
regular first-countable spaces
$Y_{\alpha }$
are already constructed by defining a base
$\mathcal B_{\alpha }=\bigcup _{x\in X_{\alpha }}\mathcal B_{\alpha }^x$
of the the topology
$\tau _{\alpha }$
, where for each
$x\in X_{\alpha }$
, the collection
$\mathcal B_{\alpha }^x=\{U_{n,\alpha }^x:n\in \omega \}$
is an open neighborhood base at x. Additionally assume that
$X_{\alpha }\subseteq X_{\beta }$
for any
$\alpha \in \beta $
and the family
$\mathcal B_{\alpha }$
satisfies the following conditions:
-
(1)
$|\mathcal B_{\alpha }|<\mathfrak c$ ;
-
(2)
$U_{0,\alpha }^x=X_{\alpha }$ for each
$x\in X_{\alpha }$ ;
-
(3)
$\operatorname {cl}_{Y_{\alpha }}(U_{n+1,\alpha }^x)\subset U_{n,\alpha }^x$ for every
$n\in \omega $ and
$x\in X_{\alpha }$ ;
-
(4) for every
$\alpha <\beta <\xi $ ,
$n\in \omega $ and
$x\in X_{\alpha }$ we have that
$U_{n,\alpha }^x=U_{n,\beta }^x\cap X_\alpha $ .
There are three cases to consider:
-
1)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is not an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
2)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
3)
$\xi $ is a limit ordinal.
1) Let
$X_{\xi }=X_{\gamma }$
and
$\mathcal B_{\xi }=\mathcal B_{\gamma }$
.
2) Put
$X_{\xi }=X_{\gamma }\cup \{\gamma \}$
. Let
$h^{-1}(\gamma )=\{z_n\}_{n\in \omega }$
. For each
$U_{n,\gamma }^x\in \mathcal B_{\gamma }$
consider the set

Since
$|\mathcal B_{\gamma }|<\mathfrak c=\mathfrak s$
we get that the family
$\{A_{U_{n,\gamma }^x}:U_{n,\gamma }^x\in \mathcal B_{\gamma }\}$
is not splitting, i.e., there exists a subset
$A\subset \omega $
such that for each
$U_{n,\gamma }^x\in \mathcal B_{\gamma }$
the set
$d_{\gamma }=\{z_n:n\in A\}$
is either almost contained in
$U_{n,\gamma }^x\setminus U_{n+1,\gamma }^x$
or almost disjoint with
$U_{n,\gamma }^x\setminus U_{n+1,\gamma }^x$
. Since the set
$\{z_n:n\in \omega \}$
is closed in
$Y_{\gamma }$
for each
$x\in X_{\gamma }$
there exists
$n\in \omega $
such that
$|U_{n,\gamma }^x\cap \{z_k:k\in \omega \}|\leq 1$
. Since
$U^x_{0,\gamma }=X_{\gamma }$
for all
$x\in X_{\gamma }$
, we obtain that for each
$x\in X_{\gamma }$
there exists a unique
$m(x)\in \omega $
such that
$d_{\gamma }\subset ^* U_{m(x),\gamma }^x\setminus U_{m(x)+1,\gamma }^x$
.
For each
$x\in X_\gamma $
let
$\mathcal E(x)$
be the pair
$(U^x_{m(x),\gamma }, U^x_{m(x)+2,\gamma })\in \mathcal B_\gamma {\times } \mathcal B_\gamma $
. Define a function
$f_{\mathcal E(x)}\in \omega ^A$
as follows: if
$z_n\in U^x_{m(x),\gamma }\setminus \operatorname {cl}_{Y_{\gamma }}(U^x_{m(x)+2,\gamma })$
, then

and
$f_{\mathcal E(x)}(n)=0$
, otherwise. Since
$|\mathcal B_{\gamma }|<\mathfrak c=\mathfrak b$
there exists a function
$f\in \omega ^A$
such that
$f\geq ^* f_{\mathcal E(x)}$
for every
$U^x_{m(x),\gamma }\in \mathcal B_{\gamma }$
. With no loss of generality we can additionally assume that
$U_{f(n),\gamma }^{z_n}\cap U_{f(m),\gamma }^{z_m}=\emptyset $
for each distinct
$n,m\in A$
. It is easy to see that the sequence
$\{U_{f(n),\gamma }^{z_n}:n\in A\}$
is locally finite in
$Y_{\gamma }$
.
Next we define an open neighborhood base at the point
$\gamma \in X_{\xi }$
: Put

for all
$k\in \mathbb N$
.
Now we are going to define an open neighborhood base
$\mathcal B_{\xi }^x$
at each point
$x\in X_{\gamma }$
. Let
$U_{0,\xi }^x=X_{\xi }$
. For each
$n\in \mathbb N$
let
$U_{n,\xi }^x=U_{n, \gamma }^x$
if
$n\geq m(x)+1$
and
$U_{n,\xi }^x=U_{n,\gamma }^x\cup \{\gamma \}$
if
$n\leq m(x)$
. It is easy to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: x\in X_{\xi },n\in \omega \}$
forms a base of a topology
$\tau _{\xi }$
, and for each
$x\in X_{\xi }$
the family
$\mathcal B_{\xi }^{x}=\{U_{n,\xi }^x:n\in \omega \}$
forms an open neighborhood base at x in
$Y_{\xi }$
.
At this point it is a tedious routine to check that the family
$\mathcal B_{\xi }$
satisfies conditions (1)–(4).
3) Let
$X_{\xi }=\bigcup _{\alpha \in \xi }X_{\alpha }$
. For each
$x\in X_{\xi }$
let
$\theta _x=\min \{\alpha :x\in X_{\alpha }\}$
, and for every
$n\in \omega $
let us put

It is quite straightforward to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: n\in \omega , x\in X_{\xi }\}$
forms a base of a topology
$\tau _{\xi }$
, and the family
$\mathcal B_{\xi }$
satisfies conditions (1), (2), and (4). In order to check the validity of condition (3) we need the following auxiliary claim.
Claim 4.2. For any
$\gamma \in \xi $
,
$n,m\in \omega $
and distinct points
$y_0, y_1\in Y_{\gamma }$
, if
$U^{y_0}_{n,\gamma }\cap U^{y_1}_{m,\gamma }=\emptyset $
, then
$U^{y_0}_{n,\xi }\cap U^{y_1}_{m,\xi }=\emptyset $
.
Proof. Seeking a contradiction, assume that
$U^{y_0}_{n,\xi }\cap U^{y_1}_{m,\xi }\neq \emptyset $
, but
$U^{y_0}_{n,\gamma }\cap U^{y_1}_{m,\gamma }=\emptyset $
for some
$\gamma \in \xi $
. It is easy to see that
$U^{y_0}_{n,\xi }\cap U^{y_1}_{m,\xi }\subset \xi \setminus \gamma $
. Let
$\delta =\min U^{y_0}_{n,\xi }\cap U^{y_1}_{m,\xi }$
. It follows that
$U^{y_0}_{n,\delta }\cap U^{y_1}_{m,\delta }=\emptyset $
and
$U^{y_0}_{n,\delta +1}\cap U^{y_1}_{m,\delta +1}=\{\delta \}$
. Then the set
$d_{\delta }$
(see case 2 above) is closed and discrete in
$Y_{\delta }$
. Since
$\delta \in U^{y_0}_{n,\delta +1}\cap U^{y_1}_{m,\delta +1}$
, the definition of
$U^{y_0}_{n,\delta +1}$
and
$U^{y_1}_{m,\delta +1}$
implies that
$d_{\delta }\subset ^*U^{y_0}_{n,\delta }\cap U^{y_1}_{m,\delta }=\emptyset $
, which contradicts our assumption.
Fix any
$x\in X_{\xi }$
,
$n\in \omega $
and
$z\in \overline {U_{n+1,\xi }^x}$
. There exists an ordinal
$\gamma \in \xi $
such that
$x,z\in X_{\gamma }$
. We claim that
$z\in \operatorname {cl}_{Y_{\gamma }}(U_{n+1,\gamma }^x)$
. Indeed, otherwise, there exists
$m\in \omega $
such that
$U^{z}_{m,\gamma }\cap U_{n+1,\gamma }^x=\emptyset $
. The above claim implies that
$U^{z}_{m,\xi }\cap U_{n+1,\xi }^x=\emptyset $
, which contradicts the choice of z. Thus
$z\in \operatorname {cl}_{Y_{\gamma }}(U_{n+1,\gamma }^x)\subset U_{n,\gamma }^x\subset U^x_{n,\xi }$
, which establishes condition (3).
By the construction, the space
$Y_{\mathfrak {c}}$
is regular, first-countable and contains X as a dense subspace. Let A be any countable subset of
$Y_{\mathfrak {c}}$
. If the set
$B=A\cap \mathfrak {c}$
is infinite, then consider
$h(B)\in \mathfrak {c}$
. By the construction, either B has an accumulation point in
$Y_{h(B)}$
or
$h(B)$
is an accumulation point of B in
$Y_{h(B)+1}$
. In both cases B has an accumulation point in
$Y_{\mathfrak c}$
. If
$A\subset ^* X$
, then either it has an accumulation point in X, or A is closed and discrete in X. In the latter case either A has an accumulation point in
$Y_{h(A)}$
or
$h(A)$
is an accumulation point of A in
$Y_{h(A)+1}$
. Thus
$Y_{\mathfrak c}$
is countably compact.
Proposition 4.3 (
$\mathfrak {s} =\mathfrak {b}=\mathfrak c$
)
Let X be a Hausdorff zero-dimensional first-countable space of weight
$\kappa <\mathfrak {c}$
. Then X embeds densely into a Hausdorff zero-dimensional first-countable countably compact space.
Proof. The proof of this theorem is very similar to one of Proposition 4.1, so we give only a sketch of it. Let
$\mathcal {D}$
and h be such as in the proof of Proposition 4.1.
Let
$X_0=X$
. Since X is first-countable and zero-dimensional there exists a base
$\mathcal B_0=\bigcup _{x\in X}\mathcal B_{0}^x$
of the topology on X consisting of clopen sets, where for each
$x\in X$
, the collection
$\mathcal B_{0}^x=\{U_{n,0}^x:n\in \omega \}$
is a nested open neighborhood base at x. With no loss of generality we can assume that
$|\mathcal B_{0}|<\mathfrak c$
and
$U_{0,0}^x=X$
for each
$x\in X$
.
Assume that for each
$\alpha <\xi $
Hausdorff zero-dimensional first-countable spaces
$Y_{\alpha }$
are already constructed by defining a base
$\mathcal B_{\alpha }=\bigcup _{x\in X_{\alpha }}\mathcal B_{\alpha }^x$
of the topology
$\tau _{\alpha }$
on a set
$X_{\alpha }\subseteq X\cup \alpha $
, where for each
$x\in X_{\alpha }$
, the collection
$\mathcal B_{\alpha }^x=\{U_{n,\alpha }^x:n\in \omega \}$
is a nested open neighborhood base at x consisting of clopen sets. Additionally assume that
$X_{\alpha }\subseteq X_{\beta }$
for any
$\alpha \in \beta $
and the family
$\mathcal B_{\alpha }$
satisfies the following conditions:
-
(1)
$|\mathcal B_{\alpha }|<\mathfrak c$ ;
-
(2)
$U_{0,\alpha }^x=X_{\alpha }$ for each
$x\in X_{\alpha }$ ;
-
(3) for every
$\alpha <\beta <\xi $ ,
$n\in \omega $ and
$x\in X_{\alpha }$ we have that
$U_{n,\alpha }^x=U_{n,\beta }^x\cap X_\alpha $ .
There are three cases to consider:
-
1)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is not an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
2)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
3)
$\xi $ is a limit ordinal.
1) Let
$X_{\xi }=X_{\gamma }$
and
$\mathcal B_{\xi }=\mathcal B_{\gamma }$
.
2) Put
$X_{\xi }=X_{\gamma }\cup \{\gamma \}$
. Let
$h^{-1}(\gamma )=\{z_n\}_{n\in \omega }$
. Similarly as in the proof of Theorem 4.1, one can find a subset
$A\subset \omega $
such that for each
$U_{n,\gamma }^x\in \mathcal B_{\gamma }$
the set
$d_{\gamma }=\{z_n:n\in A\}$
is either almost contained in
$U_{n,\gamma }^x\setminus U_{n+1,\gamma }^x$
or almost disjoint with
$U_{n,\gamma }^x\setminus U_{n+1,\gamma }^x$
. Thus, for each
$x\in X_{\gamma }$
there exists a unique
$m(x)\in \omega $
such that
$d_{\gamma }\subset ^* U_{m(x),\gamma }^x\setminus U_{m(x)+1,\gamma }^x$
.
For each
$x\in X_\gamma $
let us denote by
$\mathcal E(x)$
the pair
$(U^x_{m(x),\gamma }, U^x_{m(x)+1,\gamma })\in \mathcal B_\gamma {\times }\mathcal B_\gamma $
. Define a function
$f_{\mathcal E(x)}\in \omega ^A$
as follows: if
$z_n\in U^x_{m(x),\gamma }\setminus U^x_{m(x)+1,\gamma }$
, then

and
$f_{\mathcal E(x)}(n)=0$
, otherwise. Note that
$f_{\mathcal E(x)}$
is well-defined, as the set
$U^x_{m(x),\gamma }\setminus U_{m(x)+1,\gamma }^x$
is open for every
$x\in X_{\gamma }$
.Since
$|\mathcal B_{\gamma }|<\mathfrak c=\mathfrak b$
we get that there exists a function
$f\in \omega ^A$
such that
$f\geq ^* f_{\mathcal E(x)}$
for every
$\mathcal E(x)\in \mathcal B_{\gamma }{\times }\mathcal B_\gamma $
. We can additionally assume that
$U_{f(n),\gamma }^{z_n}\cap U_{f(m),\gamma }^{z_m}=\emptyset $
for each distinct
$n,m\in A$
. It is easy to see that the family
$\{U_{f(n),\gamma }^{z_n}:n\in A\}$
is locally finite in
$Y_{\gamma }$
.
Next we define an open neighborhood base at the point
$\gamma \in X_{\xi }$
: Put

for all
$k\in \mathbb N$
. Since the family
$\{U_{f(n),\gamma }^{z_n}:n\in A\}$
is locally finite, the set
$U_{k,\xi }^{\gamma }$
is clopen for each
$k\in \omega $
.
Now we are going to define an open neighborhood base
$\mathcal B_{\xi }^x$
at each point
$x\in X_{\gamma }$
. Let
$U_{0,\xi }^x=X_{\xi }$
for every
$x\in X_{\gamma }$
. For each
$x\in X_{\gamma }$
and
$n\in \mathbb N$
let
$U_{n,\xi }^x=U_{n, \gamma }^x$
if
$n\geq m(x)+1$
and
$U_{n,\xi }^x=U_{n,\gamma }^x\cup \{\gamma \}$
if
$n\leq m(x)$
. It is easy to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: x\in X_{\xi },n\in \omega \}$
forms a base of a topology
$\tau _{\xi }$
, and for each
$x\in X_{\xi }$
the family
$\mathcal B_{\xi }^{x}=\{U_{n,\xi }^x:n\in \omega \}$
forms a nested open neighborhood base at x in
$Y_{\xi }$
. Moreover, the space
$Y_{\xi }$
is Hausdorff and zero-dimensional.
At this point it is a tedious routine to check that the family
$\mathcal B_{\xi }$
satisfies conditions (1)–(3).
3) Let
$X_{\xi }=\bigcup _{\alpha \in \xi }X_{\alpha }$
. For each
$x\in X_{\xi }$
let
$\theta _x=\min \{\alpha :x\in X_{\alpha }\}$
. For each
$x\in X_{\xi }$
and
$n\in \omega $
put

It is straightforward to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: n\in \omega , x\in X_{\xi }\}$
forms a base of a topology
$\tau _{\xi }$
, satisfies conditions (1)–(3) and for each
$x\in X_{\xi }$
the family
$\mathcal B_{\xi }^{x}=\{U_{n,\xi }^x:n\in \omega \}$
forms an open neighborhood base at x in
$Y_{\xi }$
. Hausdorffness of the space
$Y_{\xi }$
follows from the next claim which can be proved similarly as Claim 4.2.
Claim 4.4. For any
$\gamma \in \xi $
,
$n,m\in \omega $
and distinct points
$y_0, y_1\in Y_{\gamma }$
, if
$U^{y_0}_{n,\gamma }\cap U^{y_1}_{m,\gamma }=\emptyset $
, then
$U^{y_0}_{n,\xi }\cap U^{y_1}_{m,\xi }=\emptyset $
.
It remains to check that the space
$Y_{\xi }$
is zero-dimensional which is done in the following claim.
Claim 4.5. For any
$x\in X_\xi $
and
$n\in \omega $
the set
$U^{x}_{n,\xi }$
is closed.
Proof. Fix any
$y\in X_{\xi }\setminus U^{x}_{n,\xi }$
and find
$\gamma <\xi $
such that
$x,y\in X_{\gamma }$
. It follows that
$y\notin U^{x}_{n,\gamma }$
. By the inductive assumption the set
$U^{x}_{n,\gamma }$
is closed, implying the existence of
$m\in \omega $
such that
$U^{y}_{m,\gamma }\cap U^{x}_{n,\gamma }=\emptyset $
. Then the previous claim implies that
$U^{y}_{m,\xi }\cap U^{x}_{n,\xi }=\emptyset $
, witnessing that the set
$U^{x}_{n,\xi }$
is closed.
Similarly as in the proof of Theorem 4.1 it can be checked that the space
$Y_{\mathfrak {c}}$
is Hausdorff, zero-dimensional, first-countable, countably compact and contains X as a dense subspace.
Let D be a subset of a set S. An expansion of D is a family
$\{A_d : d \in D\}$
of subsets of S such that
$d \in A_d$
and
$A_d \cap D = \{d\}$
for all
$d \in D$
. A family U of subsets of a space X is called discrete if every point of X has an open neighborhood intersecting at most one element of U. A topological space X satisfies (or has) Property D if every countable closed discrete subspace
$D \subset X$
has an expansion to a discrete family of open sets. A space X is called pseudonormal if, whenever
$F_0$
and
$F_1$
are disjoint closed sets, one of which is countable, there are disjoint open sets
$U_0$
and
$U_1$
containing
$F_0$
and
$F_1$
respectively. We will also need the following facts proved in [Reference Nyikos21, Proposition 3.6] and [Reference Nyikos21, Theorem 3.7], respectively.
Proposition 4.6 (Nyikos)
Every pseudonormal space X has property D.
Proposition 4.7 (Nyikos)
Every first-countable regular space of Lindelöf number
$<\mathfrak b$
is pseudonormal.
Propositions 4.6 and 4.7 imply the following useful corollary.
Corollary 4.8. Every first-countable regular space of weight
$<\mathfrak b$
has property D.
The proof of the following result uses techniques developed in [Reference Abraham, Gorelic and Juhász1, Reference Nyikos21, Reference Nyikos22, Reference Ostaszewski25].
Proposition 4.9 (
$\mathfrak b=\mathfrak c$
)
Each first-countable Hausdorff locally compact space of weight
$<\mathfrak c$
embeds densely into a first-countable Hausdorff locally compact countably compact space.
Proof. Fix any Hausdorff locally compact first-countable space X of weight
$< \mathfrak c=\mathfrak b$
. Without loss of generality we can assume that the underlying set of X is disjoint with
$[0,1){\times }\mathfrak c$
. First countability of X implies that
$|X|\leq \mathfrak c^{\omega }=\mathfrak c$
. If X is countably compact, then there is nothing to prove. Otherwise, let

Fix any bijection
$h:\mathfrak c{\times } \mathfrak c\rightarrow \mathfrak c$
such that
$h((a,b))\geq a$
.
Next, for every
$\alpha \leq \mathfrak {c}$
we shall recursively construct a topology
$\tau _{\alpha }$
on
$X_{\alpha }$
which will be a union of X and some pairwise disjoint family of copies of the half interval
$[0,1)$
. For the sake of brevity we denote the space
$(X_{\alpha },\tau _{\alpha })$
by
$Y_{\alpha }$
. At the end, we will show that the space
$Y_{\mathfrak {c}}$
has the desired properties.
Let
$X_0=X$
and fix any bijection
$g_0: \mathcal {D}_0\rightarrow \{0\}{\times }\mathfrak {c}$
. Note that by the definition of h,
$h^{-1}(0)\in \{0\}{\times }\mathfrak c$
, so
$g_0^{-1}(h^{-1}(0))$
is well-defined (and belongs to
$\in \mathcal D_0$
). By Corollary 4.8, the space X has property D. It follows that the closed discrete set
$\{z_n:n\in \omega \}=g_0^{-1}(h^{-1}(0))$
can be expanded to a discrete family
$\{U_n:n\in \omega \}$
of open sets such that
$z_n\in U_n$
. Since X is locally compact, we lose no generality assuming that
$\operatorname {cl}_{X}(U_n)$
is compact for every
$n\in \omega $
. For each
$n\in \omega $
fix a continuous function
$f_n$
such that
$f(z_n)=0$
and
$f(X\setminus U_n)=1$
. Let
$X_1=X_0\cup ([0,1){\times }\{1\})$
, and
$\tau _1$
be a topology on
$X_1$
which satisfies the following conditions:
-
(1)
$X_0$ is an open subspace of
$X_1$ ;
-
(2) the sets
$$\begin{align*}W_n(0,1)=[0,1/n){\times}\{1\}\cup \bigcup_{m\geq n}f_m^{-1}([0,1/n)),\end{align*}$$
$n\in \mathbb N$ form an open neighborhood base at the point
$(0,1)\in X_1$ ;
-
(3) For
$0<x<1$ the sets
$$\begin{align*}W_n(x,1)=(x-1/n,x+1/n){\times}\{1\}\cup\bigcup_{m\geq n}f_m^{-1}((x-1/n,x+1/n)),\end{align*}$$
$n\in \mathbb N$ satisfies
$(x-1/n,x+1/n)\subseteq (0,1)$ , form an open neighborhood base at the point
$(x,1)\in X_1$ .
It is easy to check that
$Y_1$
is a first-countable Hausdorff space of weight
$<\mathfrak b$
. Observe that
$(0,1)$
is an accumulation point of the set
$\{z_n:n\in \omega \}=g_0^{-1}(h^{-1}(0))$
in
$Y_1$
. Clearly,
$Y_1$
is locally compact at each point
$x\in X_0$
. Let us show that
$Y_1$
is locally compact at
$(0,1)$
. We claim that the set
$\operatorname {cl}_{Y_1}(W_2(0,1))$
is compact. Since the family
$\{U_n:n\in \omega \}$
is discrete and the sets
$U_n$
,
$n\in \omega $
have compact closure in X, it is easy to see that the set

is
$\sigma $
-compact. At this point it is enough to show that
$\operatorname {cl}_{Y_1}(W_2(0,1))$
is countably compact. Fix any infinite subset
$A=\{a_n:n\in \omega \}\subset \operatorname {cl}_{Y_1}(W_2(0,1))$
. If for some
$n\in \omega $
the set
$A\cap \operatorname {cl}_X(U_n)$
is infinite or the set
$A\cap ([0,1/2){\times }\{1\})$
is infinite, then A has an accumulation point in
$Y_1$
. So, let us assume that all the aforementioned intersections are finite. Shrinking the set A if necessary, we can assume that
$A\subset X$
and
$|A\cap U_n|\leq 1$
for all
$n\in \omega $
. For each
$n\in \omega $
let
$m(n)$
be such that
$a_n\in U_{m(n)}$
. Put
$y_n=f_{m(n)}(a_n)$
. Since
$\{y_n:n\in \omega \}\subset [0,1/2)$
and the interval
$[0,1/2]$
is compact there exists
$z\in [0,1/2]$
and a subsequence
$\{y_{n_k}:k\in \omega \}$
which converges to z. It is easy to see that
$(z,1)$
is an accumulation point of the set A in
$Y_1$
. Hence the
$\sigma $
-compact set
$\operatorname {cl}_{Y_1}(W_2(0,1))$
is countably compact and thus compact. Similarly one can check that
$Y_1$
is locally compact at
$(x,1)$
for each
$0<x<1$
. Let

Fix any bijection
$g_1: \mathcal {D}_1\rightarrow \{1\}{\times }\mathfrak {c}$
.
Assume that locally compact Hausdorff spaces
$Y_\alpha =(X_{\alpha },\tau _{\alpha })$
with weight
$<\mathfrak b$
are constructed for all
$\alpha \in \xi $
such that
$Y_\alpha $
is an open dense subspace of
$Y_{\eta }$
for all
$\alpha \in \eta \in \xi $
. Additionally assume that if
$X_{\alpha +1}\setminus X_{\alpha }\neq \emptyset $
, then
$X_{\alpha +1}\setminus X_{\alpha }=[0,1){\times }\{\alpha +1\}$
and for each
$\alpha \in \xi $
it is defined a bijection
$g_{\alpha }: \mathcal {D}_{\alpha }\rightarrow \{\alpha \}{\times }\mathfrak {c}$
, where

If
$\xi $
is a limit ordinal, then put
$X_{\xi }=\bigcup _{\alpha \in \xi }X_{\alpha }$
and
$\tau _{\xi }=\bigcup _{\alpha \in \xi }\tau _{\alpha }$
. Clearly,
$Y_\xi $
is first-countable, Hausdorff, locally compact with weight less than
$\mathfrak b$
. Fix any bijection
$g_{\xi }: \mathcal D_\xi \rightarrow \{\xi \}{\times }\mathfrak c$
.
Assume that
$\xi =\gamma +1$
for some
$\gamma \in \mathfrak {c}$
. Consider the pair
$h^{-1}(\gamma )=(a,b)\in \mathfrak c{\times }\mathfrak c$
. If the set
$g_a^{-1}(h^{-1}(\gamma ))\in \mathcal D_a$
already has an accumulation point in
$Y_{\gamma }$
, then put
$Y_{\xi }=Y_{\gamma }$
. If the set
$g_a^{-1}(h^{-1}(\gamma ))$
is closed and discrete in
$Y_{\gamma }$
, then we construct a Hausdorff locally compact first-countable space
$Y_{\xi }$
of weight
$<\mathfrak b$
similarly as we constructed
$Y_1$
. In particular,
$X_{\xi }=X_{\gamma }\cup [0,1){\times }\{\xi \}$
, and
$(0,\xi )$
is an accumulation point of the set
$g_a^{-1}(h^{-1}(\gamma ))$
.
Obviously,
$Y_{\mathfrak c}$
is a Hausdorff locally compact first-countable space. To derive a contradiction, assume that
$Y_{\mathfrak c}$
is not countably compact. Then
$Y_{\mathfrak c}$
contains an infinite countable discrete closed subset A. Since the cardinal
$\mathfrak b=\mathfrak c$
is regular, there exists an ordinal
$\xi \in \mathfrak c$
such that
$A\subset X_{\xi }$
. Obviously, A is an infinite closed discrete subset of
$Y_{\beta }$
for each
$\beta \geq \xi $
. Then
$g_{\xi }(A)=(\xi ,\delta )$
for some
$\delta \in \mathfrak c$
. Let
$\mu \in \mathfrak c$
be such that
$\xi \leq \mu $
and
$h((\xi ,\delta ))=\mu $
. By the construction, the point
$(0,\mu +1)\in X_{\mu +1}$
is an accumulation point of A in
$Y_{\mathfrak c}$
, which implies the desired contradiction. Hence the space
$Y_{\mathfrak c}$
is countably compact.
Next we turn to embeddings into pseudocompact spaces. We adopt the notations from [Reference Bell5].
Let
$S=\{S_n:n\in \omega \}$
and
$T=\{T_n:n\in \omega \}$
be sequences of subsets of a set X. We say that S refines T if there exists an injection
$f\in \omega ^\omega $
such that
$S_n\subseteq T_{f(n)}$
for all
$n\in \omega $
. We write that the sequence S is eventually contained in a subset
$A\subset X$
if there exists
$k\in \omega $
such that
$\bigcup _{n\geq k}S_n\subset A$
. The sequences S and T are called eventually disjoint if there exists
$k\in \omega $
such that
$\bigcup _{n\geq k}S_n\cap \bigcup _{n\geq k}T_n=\emptyset $
. A family
$\mathcal R$
of sequences of subsets of X is called eventually disjoint if any distinct elements
$S,T\in \mathcal R$
are eventually disjoint.
Let X be a zero-dimensional space and
$\mathcal B$
be a base of X consisting of clopen sets. By
$\mathbf {D}(\mathcal B)$
we denote the set of all locally finite sequences of elements of
$\mathcal B$
. A sequence
$S=\{S_n:n\in \omega \}\in \mathbf {D}(\mathcal B)$
is called
$\mathcal B$
-exact if for each
$B\in \mathcal B$
which intersects infinitely many
$S_n$
there exists
$k\in \omega $
such that
$\bigcup _{n\geq k}S_n\subset B$
. Let
$\mathbf {E}(\mathcal B)=\{S\in \mathbf {D}(\mathcal B): S$
is
$\mathcal B$
-exact
$\}$
.
For any eventually disjoint family
$\mathcal R\subset \mathbf {E}(\mathcal B)$
we define a topology
$\tau $
on the set
$X\cup \mathcal R$
as follows: for each
$B\in \mathcal B$
let
$B^*=B\cup \{R\in \mathcal R: R$
is eventually contained in
$B\}$
. For every
$R=\{R_n:n\in \omega \}\in \mathcal R$
and
$n\in \omega $
let
$U_n(R)=\{R\}\cup \bigcup _{m\geq n}R_m^*$
. The topology
$\tau $
is generated by the base
$\mathcal B^*=\{B^*:B\in \mathcal B\}\cup \{U_n(R):R\in \mathcal R,n\in \omega \}$
. By
$X(\mathcal R)$
we denote the space
$(X\cup \mathcal R,\tau )$
. It is easy to check that the space
$X(\mathcal R)$
is Hausdorff and zero-dimensional.
The following result is proved by Bell in [Reference Bell5, Claim 3.2].
Proposition 4.10. Let X be a first-countable zero-dimensional space and
$\mathcal B$
be a base of X consisting of clopen sets. Suppose that for every
$T\in \mathbf {D}(\mathcal B)$
there exists
$S\in \mathbf {E}(\mathcal B)$
such that S refines T. Then for every maximal eventually disjoint subfamily
$\mathcal R$
of
$\mathbf {E}(\mathcal B)$
the space
$X(\mathcal R)$
is zero-dimensional, first countable, and pseudocompact.
Proposition 4.11. Each first-countable zero-dimensional Hausdorff space X of weight
$<\min \{\mathfrak {s},\mathfrak {b}\}$
embeds densely into a first-countable zero-dimensional pseudocompact space.
Proof. Fix a base
$\mathcal B$
of X of size
$<\min \{\mathfrak {s},\mathfrak {b}\}$
consisting of clopen sets. We lose no generality assuming that
$\mathcal B=\{U_n^x:x\in X, n\in \omega \}$
, where for every
$x\in X$
the family
$\{U_n^x:n\in \omega \}$
forms a nested open neighborhood base at x. By Proposition 4.10 it is enough to check that for every
$T\in \mathbf {D}(\mathcal B)$
there exists
$S\in \mathbf {E}(\mathcal B)$
such that S refines T. Consider an arbitrary sequence
$T=\{T_n:n\in \omega \}\in \mathbf {D}(\mathcal B)$
and for each
$n\in \omega $
fix a point
$x_n\in T_n$
. To each
$B\in \mathcal B$
we assign a function
$f_B\in \omega ^\omega $
defined as follows: if
$x_n\in B$
, then put
$f_B(n)=\min \{k\in \omega : U_k^{x_n}\subset B\cap T_n\}$
; otherwise put
$f_B(n)=\min \{k\in \omega : U_k^{x_n}\subset T_n$
and
$U_k^{x_n}\cap B=\emptyset \}$
. Since
$|\mathcal B|<\mathfrak b$
, there exists a function
$f\in \omega ^\omega $
such that
$f\geq ^* f_B$
for every
$B\in \mathcal B$
. Consider the refinement
$\{U^{x_n}_{f(n)}:n\in \omega \}$
of T. Clearly, for each
$B\in \mathcal B$
there exists
$k\in \omega $
such that for every
$n\geq k$
either
$U^{x_n}_{f(n)}\cap B=\emptyset $
or
$U^{x_n}_{f(n)}\subset B$
. So, to each
$B\in \mathcal B$
we can assign a set
$C_B=\{n\in \omega : U^{x_n}_{f(n)}\subset B\}\subset \omega $
. Since
$|\mathcal B|<\mathfrak s$
we get that the family
$\{C_B:B\in \mathcal B\}$
is not splitting. Hence there exists an infinite subset A of
$\omega $
possessing the following property: for every
$B\in \mathcal B$
there exists
$k\in \omega $
such that either
$U^{x_n}_{f(n)}\subset B$
for each
$n\in A\setminus k$
or
$U^{x_n}_{f(n)}\cap B=\emptyset $
for each
$n\in A\setminus k$
. Hence the sequence
$S=\{U^{x_n}_{f(n)}:n\in A\}$
is a
$\mathcal B$
-exact refinement of T.
5 Non-embedding theorems
To each almost disjoint family
$\mathcal A\subset [\omega ]^\omega $
corresponds a Mrowka space
$\psi (\mathcal A)=(\mathcal A\cup \omega ,\tau )$
, where the topology
$\tau $
is defined as follows: points of
$\omega $
are isolated and for each
$A\in \mathcal A$
the family
$\{\{A\}\cup (A\setminus n):n\in \omega \}$
is an open neighborhood base at A in
$\tau $
. We refer the reader to [Reference Hrušák15] for basic properties of Mrowka spaces.
Proposition 5.1. There exists an almost disjoint family
$\mathcal A$
of cardinality
$\mathfrak b$
such that the corresponding Mrowka space
$\psi (\mathcal A)$
doesn’t embed into first-countable Hausdorff countably compact spaces.
Proof. Fix an unbounded subset
$\{b_{\alpha }:\alpha \in \mathfrak b\}\subset \omega ^\omega $
which satisfies the following conditions:
-
(1) the function
$b_{\alpha }$ is increasing for every
$\alpha \in \mathfrak b$ ;
-
(2)
$b_\alpha <^* b_{\beta }$ whenever
$\alpha \in \beta $ .
For
$i\in \omega $
set
$v_i=\{(i,n):n\in \omega \}$
. Let
. Condition (2) implies that
$\mathcal A$
is an almost disjoint family of subsets of
$\omega {\times }\omega $
. To derive a contradiction, assume that there exists a first-countable Hausdorff countably compact space X which contains the Mrowka space
$\psi (\mathcal A)$
as a subspace. Find an infinite subset C of
$\omega $
such that the sequence
$\{v_{n}:n\in C\}$
converges to a point
$x\in X$
. Fix an open neighborhood base
$\{U_n:n\in \omega \}$
at x. Without loss of generality we can assume that
$\{v_k:k\in C\setminus n\}\subset U_n$
for all
$n\in \omega $
. Observe that for each
$n\in \omega $
there exists a function
$f_n\in \omega ^C$
such that
$\{(k,m): k\in C\setminus n, m\geq f_n(k)\}\subset U_n$
. Since the functions
$b_{\alpha }$
,
$\alpha \in \mathfrak b$
are increasing, the family
$\{b_\alpha {\restriction }_C :\alpha \in \mathfrak b\}$
is unbounded. Then for each
$n\in \omega $
there exists
$\alpha _n\in \mathfrak b$
such that
$f_n\not \geq ^* b_{\alpha _n}{\restriction }_C$
. It follows that for each
$n\in \omega $
and
$\xi \geq \alpha _n$
,

Thus for
$\mu = \sup \{\alpha _n:n\in \omega \}$
the set
$b_{\mu }\cap \{(k,m): k\in C\setminus n, m\geq f_n(k)\}$
is infinite for all
$n\in \omega $
. Hence
$b_{\mu }\in \overline {U_n}$
for all
$n\in \omega $
, which contradicts the Hausdorffness of X.
The following result is essentially based on the work of van Douwen and Przymusiński [Reference van Douwen and Przymusinski9]. Throughout its proof
$\mathbb N=\omega \setminus \{0\}$
.
Proposition 5.2. There exists a separable first-countable normal Lindelöf space of cardinality
$\mathfrak s$
which cannot be densely embedded into a first-countable regular feebly compact space.
Proof. By
$\mathbb Q$
we denote the set of rational numbers from the interval
$[0,1]$
. Enumerate the set of all subsets of
$\omega $
as
$\{A_x:x\in [0,1]\}$
, and for each
$x\in [0,1]$
and
$m\in \mathbb N$
choose
$q_x(m)\in \mathbb Q$
such that
$0<|x-q_x(m)|<1/m$
and put
$Q_x=\{q_x(m):m\in \mathbb N\}$
. Let X be the set
$[0,1]{\times }\{0\}\cup \mathbb Q{\times }\mathbb {\mathbb N}$
endowed with a topology
$\tau $
satisfying the following conditions:
-
•
$\mathbb Q{\times }\mathbb {\mathbb N}$ is the set of isolated points;
-
• if
$x\in \mathbb Q$ , then basic neighborhoods of a point
$(x,0)\in [0,1]{\times }\{0\}$ have the form
$$\begin{align*}B_{m}(x)=\{(a,b)\in X: |x-a|<1/m\}\setminus (Q_x{\times}A_x\cup \{x\}{\times}\mathbb N) \qquad \mbox{for} \quad m\in\mathbb N;\end{align*}$$
-
• if
$x\in [0,1]\setminus \mathbb Q$ , then basic neighborhoods of a point
$(x,0)\in [0,1]{\times }\{0\}$ have the form
$$\begin{align*}B_{m}(x)=\{(a,b)\in X: |x-a|<1/m\}\setminus (Q_x{\times}A_x) \qquad \mbox{for} \quad m\in\mathbb N.\end{align*}$$
Clearly, the spaceFootnote 1 X is regular, separable and first-countable.
Fix an arbitrary splitting family
$\mathcal S\subset [\omega ]^\omega $
of cardinality
$\mathfrak s$
and consider the subspace

of X. It is clear that Y is regular and first-countable. Taking into account that the subspace
$([0,1]{\times }\{0\})\cap Y$
of Y is second-countable, we deduce that Y is separable, Lindelöf and subsequently normal.
Let us show that Y doesn’t embed densely into a first-countable regular feebly compact space. Assuming the contrary, fix any first-countable regular feebly compact space Z which contains Y as a dense subspace. The following claim is crucial.
Claim 5.3. For each
$B\in [\mathbb N]^{\omega }$
there exists
$q\in \mathbb Q$
such that
$\{q\}{\times }B$
is not a convergent sequence in Z.
Proof. Fix an arbitrary infinite subset
$B\subseteq \mathbb N$
and find
$S\in \mathcal S$
that splits B. Let
$x\in [0,1]$
be such that
$A_x=S$
. Fix an arbitrary open neighborhood W of
$(x,0)$
in Z such that
$W\cap Y\subseteq B_2(x)$
. By the regularity of Z there exists an open neighborhood V of
$(x,0)$
in Z such that
$\operatorname {cl}_Z(V)\subseteq W$
. Find
$m\in \mathbb N$
such that
$B_m(x)\subset V$
. Assume that the set
$\{q_{x}(m)\}{\times } B$
converges to some
$b\in Z$
. Then
$b\in \operatorname {cl}_Z(V)\subseteq W$
. Recall that
$W\cap Y\subseteq B_2(x)$
and
$B_2(x)\cap (\{q_x(m)\}{\times }\mathbb N)= \{q_x(m)\}{\times }(\mathbb N\setminus S)$
. Since the set
$B\cap S$
is infinite, the open set W contains no open neighborhood of b which contradicts the choice of W.
Fix any enumeration
$\mathbb Q=\{q_n:n\in \omega \}$
. Since the set
is open and discrete in a feebly compact space Z, there exists an accumulation point
$z_0$
of
$Q_0$
. Since Z is first-countable, there exists
$A_0\subseteq \mathbb N$
such that the sequence
$\{(q_0,n):n\in A_0\}$
converges to
$z_0$
. Assume that for each
$i\leq n$
we constructed a subset
$A_i$
of
$\mathbb N$
satisfying:
-
(1)
$A_i\subset A_j$ for every
$i\leq j$ ;
-
(2) the sequence
$\{(q_i,n):n\in A_i\}$ converges to some
$z_i\in Z$ .
Consider the discrete open set . Since Z is first-countable and feebly compact, there exists an infinite subset
$A_{n+1}\subset A_n$
such that the sequence
$\{(q_{n+1},n):n\in A_{n+1}\}$
converges to some point
$z_{n+1}\in Z$
.
After completing the induction we obtain a decreasing chain
$\{A_i:i\in \omega \}$
of infinite subsets of
$\mathbb N$
and a sequence of points
$\{z_i:i\in \omega \}$
such that
$z_i$
is the limit of the sequence
$\{(q_i,n):n\in A_i\}$
. Let A be any pseudointersection of the family
$\{A_i:i\in \omega \}$
, i.e.,
$A\subseteq ^* A_i$
for all
$i\in \omega $
. It is easy to see, that the sequence
$\{(q_i,n):n\in A\}$
converges to
$z_i$
for each
$i\in \omega $
. But this contradicts Claim 5.3.
The following trivial fact will be useful in the next proposition.
Lemma 5.4. Suppose that
$A\cup B$
is a
$G_{\delta }$
subset of a Hausdorff space X. If B is countable, then A is
$G_{\delta }$
.
Proof. Let
$B\setminus A=\{b_n:n\in \omega \}$
and
$A\cup B=\bigcap _{n\in \omega }U_n$
, where
$U_n$
is open for each
$n\in \omega $
. Then the set
is open for each
$n\in \omega $
, and
$\bigcap _{n\in \omega } V_n=A$
, i.e., A is
$G_{\delta }$
.
For each
$i\in \omega $
let
$\omega ^{\leq i}=\bigcup _{n\leq i}\omega ^n$
and
$\omega ^{<i}=\bigcup _{n<i}\omega ^n$
. Clearly, the sets
$\omega ^{\leq i}$
and
$\omega ^{<i}$
ordered by the inclusion are subtrees of the complete
$\omega $
-ary tree. Recall that a subspace A of the Cantor space is called a
$\lambda $
-set if each countable subset of A is
$G_{\delta }$
. For the definition of a Pixley–Roy hyperspace see Section 2. The proof of the next result uses the ideas of Bell [Reference Bell5].
Proposition 5.5. Let X be a subspace of the Cantor space which is not a
$\lambda $
-set. Then the Pixley–Roy hyperspace
$\operatorname {PR}(X)$
is a zero-dimensional Tychonoff first-countable space that doesn’t embed densely into first-countable regular feebly-compact spaces.
Proof. Let A be a countable infinite subset of X which is not
$G_\delta $
. Seeking a contradiction, assume that
$\operatorname {PR}(X)$
embeds densely into a regular first-countable feebly compact space Z. Fix any countable dense subset
$A'$
of X. Lemma 5.4 implies that
$Q=A\cup A'$
is not
$G_{\delta }$
. Fix an increasing sequence
$\{F^{\emptyset }_k:k\in \omega \}$
of finite subsets of Q such that
$\bigcup _{k\in \omega }F^{\emptyset }_k=Q$
. Let
. Consider the family
$\{\operatorname {cl}_Z[F_k^{\emptyset }, X]: k\in \omega \}$
. Taking into account that
$\operatorname {PR}(X)$
is dense in Z and
$[F_k^{\emptyset }, B_{\emptyset }]$
is clopen in
$\operatorname {PR}(X)$
, it is easy to check that
$[F_k^{\emptyset }, B_{\emptyset }]\subset \operatorname {int}_Z(\operatorname {cl}_Z[F_k^{\emptyset }, B_{\emptyset }])$
for all
$k\in \omega $
. So, the family
$\{\operatorname {int}_Z(\operatorname {cl}_Z[F_k^{\emptyset }, B_{\emptyset }]):k\in \omega \}$
consists of open subsets of Z and is centered. By feeble compactness of Z, there exists

Fix a countable open neighborhood base
$\{O^{\emptyset }_k:k\in \omega \}$
at
$z_{\emptyset }$
. Since
$\operatorname {PR}(X)$
is dense in Z, for each
$k\in \omega $
there exists a basic clopen subset
$[G_k^{\emptyset }, W_k^{\emptyset }]$
of
$\operatorname {PR}(X)$
such that
$[G_k^{\emptyset }, W_k^{\emptyset }]\subseteq O_k^{\emptyset }\cap [F_k^{\emptyset }, B_{\emptyset }]$
. Clearly,
$F_{k}^{\emptyset }\subseteq G_{k}^{\emptyset }$
and the sequence
$\{G_{k}^{\emptyset }:k\in \omega \}\subset \operatorname {PR}(X)$
converges to
$z_{\emptyset }$
.
Suppose that for some
$i\geq 0$
we accomplished the following things:
-
(1) constructed a subtree
$\mathbb T_i$ of
$\omega ^{\leq i}$ ;
-
(2) constructed a family
$\{B_s:s\in \mathbb T_i\}$ of clopen subsets of X satisfying the following conditions:
-
(2.1) for every
$s\in \mathbb T_i\cap \omega ^{<i}$ ,
, where
;
-
(2.2)
$B_{s\hat {~}n}\cap B_{s\hat {~}m}=\emptyset $ for every
$s\in \mathbb T_i\cap \omega ^{<i}$ and distinct
$n,m\in \beta _s$ ;
-
(2.3) for each
$s\in \mathbb T_i$ the diameter of
$B_s$ doesn’t exceed
$1/|s|$ .
-
-
(3)
$\forall s\in \mathbb T_i$ fixed an increasing sequence
$\{F_k^s:k\in \omega \}$ of finite subsets of
such that
$\bigcup _{k\in \omega }F_k^s= Q_s$ ;
-
(4)
$\forall s\in \mathbb T_i$ fixed a point
$z_s\in \bigcap _{k\in \omega }\operatorname {cl}_Z[F_k^s,B_s]$ and a nested open neighborhood base
$\{O_k^s:k\in \omega \}$ at
$z_s$ ;
-
(5)
$\forall s\in \mathbb T_i$ ,
$\forall k\in \omega $ fixed a subset
$[G^s_k,W_k^s]\subset O_k^s\cap [F_k^s,B_s]$ such that the following condition holds:
-
(5.1)
$\forall s\in \omega ^{<i}\cap \mathbb T_i$ ,
$\forall n\in \beta _s$ ,
$\exists k(s,n)\geq |s|+1$ such that
$B_{s\hat {~}n}\subseteq \bigcap _{j\leq |s|}\times W_{k(s,n)}^{s{\restriction }j}$ .
-
Fix any
$s\in \mathbb T_i\cap \omega ^{i}$
. Since
$B_s$
is open in X and Q is dense, the set
is not empty. Note that since X could contain isolated points, it is possible that
$|Q_s|<\omega $
. So, let us fix an enumeration
$Q_s=\{q_n^s:n\in \alpha _s\}$
, where
$\alpha _s\in \omega +1$
. Observe that
$q^s_0\in Q\cap B_{s{\restriction }j}$
for all
$j\leq |s|$
. Since
$\{F_k^{s{\restriction }j}: k\in \omega \}$
is an increasing sequence of finite subsets of
$Q\cap B_{s{\restriction }j}$
, there exists
$n_j\in \omega $
such that
$q^s_0\in F_{n}^{s{\restriction }j}$
for all
$n\geq n_j$
. Put
$k(s,0)=\max \{n_j:j\leq |s|\}+|s|+1$
and note that the open set
$\bigcap _{j\leq |s|} W_{k(s,0)}^{s{\restriction }j}$
contains
$q^s_0$
. So fix an arbitrary clopen set
$B_{s\hat {~}0}\subseteq \bigcap _{j\leq |s|} W_{k(s,0)}^{s{\restriction }j}$
which contains
$q_0^s$
and has diameter
$\leq 1/(i+1)$
. By condition (5), we get that
$W_{k(s,0)}^s\subseteq B_{s}$
which implies that
$B_{s\hat {~}0}\subseteq B_s$
. If
$Q_s\subset B_{s\hat {~}0}$
, we put
$\beta _s=1$
and move to another
$s\in \mathbb T_i\cap \omega ^{i}$
. If
$Q_s\setminus B_{s\hat {~}0}\neq \emptyset $
, find minimal
$m\in \omega $
such that
$q_m^s\notin B_{s\hat {~}0}$
. Similarly as above there exists
$k(s,1)\geq |s|+1$
such that the open set
$\bigcap _{j\leq |s|} W_{k(s,1)}^{s{\restriction }j}$
contains
$q^s_m$
. Since the set
$B_{s\hat {~}0}$
is clopen and doesn’t contain
$q_m^s$
, we can find a clopen neighborhood
$B_{s\hat {~}1}$
of
$q_m^s$
of diameter
$\leq 1/(i+1)$
which is contained in
$(B_s\setminus B_{s\hat {~}0})\cap \bigcap _{j\leq |s|} W_{k(s,1)}^{s{\restriction }j}$
. If
$Q_s\subset B_{s\hat {~}0}\cup B_{s\hat {~}1}$
, we put
$\beta _s=2$
and move to another
$s\in \mathbb T_i\cap \omega ^i$
. Otherwise, we can find minimal
$m\in \omega $
such that
$q_m^s\in B_s\setminus (B_{s\hat {~}0}\cup B_{s\hat {~}1})$
and repeat the previous steps. Proceeding this way we shall construct a family
$\{B_{s\hat {~}k}:k\in \beta _s\in \omega +1\}$
of clopen subsets of X which satisfies conditions (2.1), (2.2), (2.3), and (5.1). Note that
$\beta _s\leq \alpha _s$
can be finite. Let
$\mathbb T^s_{i+1}=\mathbb T_i\cup \{s\hat {~}k:k\in \beta _s\}$
.
After making the above construction for all
$s\in \mathbb T_i\cap \omega ^{i}$
, we obtain a tree
$\mathbb T_{i+1}=\bigcup _{s\in \mathbb T_i\cap \omega ^{i}}\mathbb T_{i+1}^s$
and a family
$\{B_{s\hat {~}k}:k\in \beta _s \mbox { and } s\in \mathbb T_i\cap \omega ^{i}\}$
of clopen subsets of X which satisfies conditions (2.1), (2.2), (2.3), and (5.1).
For each
$s\in \mathbb T_{i+1}\cap \omega ^{i+1}$
fix an increasing sequence
$\{F_k^s:k\in \omega \}$
of finite subsets of
$Q_s$
such that
$\bigcup _{k\in \omega }F_k^s=Q_s$
. Analogous arguments to those we used to find
$z_\emptyset $
imply that for each
$s\in \mathbb T_{i+1}\cap \omega ^{i+1}$
there exists a point
$z_s\in \bigcap _{k\in \omega }\operatorname {cl}_Z[F_k^s,B_s]$
. By the first-countability of Z, we can fix a nested open neighborhood base
$\{O_k^s:k\in \omega \}$
at
$z_s$
. For each
$s\in \mathbb T_{i+1}\cap \omega ^{i+1}$
and
$k\in \omega $
fix a subset
$[G^s_k,W_k^s]\subset O_k^s\cap [F_k^s,B_s]$
. This finishes step
$i+1$
of the induction.
Upon completing the induction we obtain a tree
$\mathbb T=\bigcup _{i\in \omega }\mathbb T_i$
, and a family
$\{B_s:s\in \mathbb T\}$
of clopen subsets of X which satisfies conditions (2.1), (2.2), (2.3), and (5.1). Also, for each
$s\in \mathbb T$
we found an element
$z_s\in Z$
which satisfies condition (4) and constructed the families
$\{W_k^s: k\in \omega , s\in \mathbb T\}$
(consisting of open subsets of X) and
$\{G_k^s: k\in \omega , s\in \mathbb T\}$
(consisting of finite subsets of X), which satisfy conditions (5) and (5.1).
Let

By the construction, P is a
$G_{\delta }$
-set containing Q. It follows that
$|P|>\omega $
. Indeed, otherwise we would have that
$Q=P\cap \bigcap _{x\in P\setminus Q}(X\setminus \{x\})$
is a
$G_\delta $
subset of X, which is not the case. Thus, we can fix

It is easy to see that there exists
$t\in \omega ^\omega $
such that
$p_{\infty }\in B_{t{\restriction }i}$
for any
$i\in \omega $
. Note that since the diameter of
$B_{t{\restriction }i}$
doesn’t exceed
$1/i$
, we get that the family
$\{B_{t{\restriction }i}:i\in \omega \}$
forms an open neighborhood base at
$p_{\infty }$
in X. It follows that the family
$\{[\{p_{\infty }\},B_{t{\restriction }i}]:i\in \omega \}$
forms an open neighborhood base of
$\{p_{\infty }\}$
in
$\operatorname {PR}(X)$
. Put
$R=\operatorname {int}_Z(\operatorname {cl}_Z([\{p_{\infty }\}, X]))$
. One can check that
$R\cap \operatorname {PR}(X)=[\{p_{\infty }\}, X]$
. By the regularity of Z, there exists an open in Z neighborhood W of
$\{p_{\infty }\}$
such that
$\operatorname {cl}_Z(W)\subseteq R$
. Thus, there exists
$i\in \omega $
such that
$\operatorname {cl}_Z([\{p_{\infty }\}, B_{t{\restriction }i}])\subseteq R$
.
Claim 5.6.
$z_{t{\restriction }i}\in \operatorname {cl}_Z([\{p_{\infty }\}, B_{t{\restriction }i}])$
.
Proof. Pick any integer
$l>i$
. Using condition (5.1) for
$s=t{\restriction }l$
and
$n=t(l)$
we can find an integer
$k(t{\restriction }l,t(l))\geq l+1$
such that
$B_{t{\restriction }(l+1)}\subseteq W_{k(t{\restriction }l,t(l))}^{t{\restriction }i}$
. Then
$p_{\infty }\in W_k^{t{\restriction }i}$
for infinitely many
$k\in \omega $
. It follows that
$[\{p_{\infty }\}, B_{t{\restriction }i}]\cap [G_k^{t{\restriction }i}, W_k^{t{\restriction }i}]\neq \emptyset $
for infinitely many k. Since for each
$k\in \omega $
,
$[G_k^{t{\restriction }i},W_k^{t{\restriction }i}]\subset O_k^{t{\restriction }i}$
and the family
$\{O_k^{t{\restriction }i}: k\in \omega \}$
forms an open neighborhood base at
$z_{t{\restriction }i}$
, we deduce that
$z_{t{\restriction }i}\in \operatorname {cl}_Z([\{p_{\infty }\}, B_{t{\restriction }i}])$
.
Since
$\operatorname {cl}_Z([\{p_{\infty }\}, B_{t{\restriction }i}])\subseteq R$
, Claim 5.6 implies that
$z_{t{\restriction }i}\in R$
. Condition (5) implies that for each
$s\in \mathbb T$
the sequence
$\{G_k^s:k\in \omega \}\subseteq \operatorname {PR}(X)$
converges to
$z_s$
. Since the set R is an open neighborhood of
$z_{t{\restriction }i}$
, there exists
$k\in \omega $
such that
$G_k^{t{\restriction }i}\in R\cap \operatorname {PR}(X)=[\{p_{\infty }\}, X]$
, which is impossible, as
$p_\infty \notin \bigcup _{s\in \mathbb T}\bigcup _{k\in \omega }G_k^s$
(see Equation 1).
By [Reference Bukovský7, Corollary 8.51] there exists a subspace A of the Cantor space which is not a
$\lambda $
-set and has cardinality
$\mathfrak b$
. Note that
$|\operatorname {PR}(A)|=|A|=\mathfrak b$
. Hence Proposition 5.5 implies the following.
Corollary 5.7. There exists a Tychonoff zero-dimensional first-countable space X of cardinality
$\mathfrak b$
which doesn’t embed densely into regular first-countable feebly compact spaces.
6 Proofs of the main results
Proof of Theorem 3.1
We need to show that the following assertions are equivalent:
-
(1)
$\omega _1=\mathfrak c$ .
-
(2) Every first-countable Tychonoff space of weight
$<\mathfrak c$ embeds into a Hausdorff first-countable compact space.
-
(3) Each separable first-countable locally compact normal space of cardinality
$<\mathfrak c$ embeds into a Hausdorff first-countable compact space.
Proof. Implication (1)
$\Rightarrow $
(2) follows from the fact that Each Tychonoff second-countable space is a subspace of the Tychonoff cube
$[0,1]^{\omega }$
. Implication (2)
$\Rightarrow $
(3) follows from the fact that the weight of a first-countable space doesn’t exceed its cardinality. Reference [Reference Bardyla, Supina and Zdomskyy4, Proposition 2.11] implies the existence of a separable normal locally compact first-countable space X which contains a homeomorphic copy of the cardinal
$\omega _1$
endowed with the order topology. Since the only compactification of
$\omega _1$
is
$\omega _1{+}1$
, which is not first-countable, we deduce that X cannot be embedded into a Hausdorff first-countable compact space.
Proof of Theorem 3.2
We need to show that the following statements are equivalent.
-
(1)
$\mathfrak b = \mathfrak c$ .
-
(2) Every Hausdorff, locally compact, first-countable space of weight
$< \mathfrak c$ can be (densely) embedded in a Hausdorff first-countable locally compact countably compact space.
-
(3) Every Hausdorff, locally compact, first-countable space of cardinality
$< \mathfrak c$ can be (densely) embedded in a Hausdorff first-countable countably compact space.
Proof. The implication (1)
$\Rightarrow $
(2) follows from Proposition 4.9. Since the weight of an infinite first-countable space doesn’t exceed its cardinality, implication (2)
$\Rightarrow $
(3) follows. Taking into account that Mrowka spaces are first-countable and locally compact, implication (3)
$\Rightarrow $
(1) follows from Proposition 5.1.
Proof of Theorem 3.3
We need to show that the following assertions are equivalent.
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every regular first-countable space of weight
$< \mathfrak c$ can be densely embedded in a regular first-countable countably compact space.
-
(3) Every regular first-countable space of cardinality
$< \mathfrak c$ can be densely embedded in a regular first-countable countably compact space.
Proof. Implication (1)
$\Rightarrow $
(2) follows from Proposition 4.1. Since the weight of an infinite first-countable space doesn’t exceed its cardinality, implication (2)
$\Rightarrow $
(3) holds true.
(3)
$\Rightarrow $
(1). Proposition 5.1 implies that
$\mathfrak b=\mathfrak c$
. Proposition 5.2 implies that
$\mathfrak s=\mathfrak c$
.
Proof of Theorem 3.4
We need to show that the following assertions are equivalent.
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every Hausdorff zero-dimensional first-countable space of weight
$< \mathfrak c$ can be densely embedded into a Hausdorff zero-dimensional first-countable countably compact space.
-
(3) Every Hausdorff zero-dimensional first-countable space of cardinality
$< \mathfrak c$ can be densely embedded into a Hausdorff zero-dimensional first-countable countably compact space.
Proof. Implication (1)
$\Rightarrow $
(2) follows from Proposition 4.3. Since the weight of an infinite first-countable space doesn’t exceed its cardinality, implication (2)
$\Rightarrow $
(3) holds true.
(3)
$\Rightarrow $
(1) Proposition 5.1 implies that
$\mathfrak b=\mathfrak c$
. Recall that a Tychonoff first-countable space Y constructed in the proof of Proposition 5.2 has cardinality
$\mathfrak s$
. If
$\mathfrak s<\mathfrak c$
, then the space Y is zero-dimensional (as every Tychonoff space of cardinality
$<\mathfrak c$
is zero-dimensional), and Y doesn’t embed into regular first-countable countably compact spaces, which contradicts assertion (3). Hence
$\mathfrak s=\mathfrak c$
, i.e., assertion (1) holds.
In order to prove Theorem 3.5 we need the following auxiliary lemma.
Lemma 6.1. There exists a regular zero-dimensional separable first-countable non-normal space of cardinality
$\omega _1$
.
Proof. We shall use ideas of Nyikos and Vaughan [Reference Nyikos and Vaughan23]. Consider a Hausdorff
$(\omega _1,\omega _1)$
gap, i.e., a family
$\{A_{\alpha }:\alpha \in \omega _1\}\cup \{B_{\alpha }:\alpha \in \omega _1\}\subset [\omega ]^{\omega }$
which satisfies the following conditions:
-
•
$A_{\alpha }\subset ^* B_{\beta }$ for any
$\alpha ,\beta \in \omega _1$ ;
-
•
$A_{\alpha }\subset ^* A_{\beta }$ for any
$\alpha <\beta <\omega _1$ ;
-
•
$B_{\alpha }\subset ^* B_{\beta }$ for any
$\beta <\alpha <\omega _1$ ;
-
• there exists no
$C\in [\omega ]^\omega $ such that
$A_{\alpha }\subset ^* C\subset ^* B_{\alpha }$ for all
$\alpha \in \omega _1$ .
For a finite set F and ordinals
$\alpha <\beta <\omega _1$
let us introduce two sorts of sets:


Let N be the set
$\{A_{\alpha }:\alpha \in \omega _1\}\cup \{B_{\alpha }:\alpha \in \omega _1\}\cup \omega $
endowed with the topology
$\tau $
defined by the subbase:

It is straightforward to check that the space N is regular, separable, first-countable and has cardinality
$\omega _1$
. Also, the closed sets
$\{A_{\alpha }:\alpha \in \omega _1\}$
and
$\{B_{\alpha }:\alpha \in \omega _1\}$
cannot be separated by disjoint open sets. Hence the space N is not normal.
Proof of Theorem 3.5
We need to show the equivalence of the following assertions:
-
(1)
$\mathfrak b = \mathfrak s= \mathfrak c$ .
-
(2) Every regular separable first-countable non-normal space of weight
$< \mathfrak c$ embeds into a Nyikos
$\mathbb R$ -rigid space.
-
(3) Every regular separable first-countable non-normal space of cardinality
$< \mathfrak c$ embeds into a Nyikos
$\mathbb R$ -rigid space.
Proof. (1)
$\Rightarrow $
(2). Let X be a regular separable first countable non-normal space of weight
$<\mathfrak c$
. Applying Jones machine (see [Reference Bardyla and Zdomskyy3] or [Reference Jones18] for details) we embed X into a first-countable regular separable space
$J(X)$
of weight
$<\mathfrak c$
containing two points
$a,b$
which cannot be separated by any real-valued continuous function. Applying the modified van Douwen’s extension
$E_{a,b}(J(X))$
of
$J(X)$
(see [Reference Bardyla and Zdomskyy3, Section 5]) we embed
$J(X)$
into a regular separable first-countable
$\mathbb {R}$
-rigid space of weight
$<\mathfrak c$
. By Theorem 4.1 we get a regular first-countable countably compact space Z which contains the separable
$\mathbb R$
-rigid space
$E_{a,b}(J(X))$
as a dense subspace. Hence Z is an
$\mathbb {R}$
-rigid Nyikos space which contains a homeomorphic copy of X.
The implication (2)
$\Rightarrow $
(3) holds true, because the weight of an infinite first-countable space doesn’t exceed its cardinality.
(3)
$\Rightarrow $
(1). Let
$N_1$
be the topological sum of the space N constructed in Lemma 6.1 and the Mrowka space
$\psi (\mathcal A)$
constructed in Proposition 5.1. It is clear that
$N_1$
is a regular separable first-countable non-normal space of cardinality
$\mathfrak b$
which doesn’t embed into Nyikos spaces. This example implies that
$\mathfrak b=\mathfrak c$
.
Let
$N_2$
be the topological sum of the space N constructed in Lemma 6.1 and the space Y constructed in Proposition 5.2. It can be checked that
$N_2$
is a separable regular first-countable non-normal space of cardinality
$\mathfrak s$
which doesn’t embed into Nyikos spaces. Hence
$\mathfrak s=\mathfrak c$
.
Proof of Theorem 3.6
Assuming that
$\omega _1<\mathfrak b=\mathfrak s=\mathfrak c$
we need to construct an
$\mathbb R$
-rigid Nyikos space.
Proof of Corollary 3.7
We need to show that under PFA the following assertions hold:
-
(1) Every normal Nyikos space is compact.
-
(2) Every regular separable first-countable space of weight
$< \mathfrak c$ embeds into an
$\mathbb R$ -rigid Nyikos space.
Proof. Assertion (1) follows from Theorem 1.2. In order to prove assertion (2) fix any regular separable first-countable space X of weight
$< \mathfrak c$
. Let N be the space constructed in Lemma 6.1. Recall that PFA implies
$\omega _1<\mathfrak b=\mathfrak s=\mathfrak c$
. Hence the topological sum
$X\cup N$
is a regular separable first-countable non-normal space of weight
$<\mathfrak c$
. By Theorem 3.5(2),
$X\cup N$
embeds into an
$\mathbb R$
-rigid Nyikos space.
In order to prove Theorem 3.8 we shall need the following auxiliary lemma.
Lemma 6.2. Let X be a first-countable Tychonoff space of weight
$\kappa $
. Then for each
$x\in X$
there exists an open neighborhood base
$\mathcal B_x=\{U_n^x:n\in \omega \}$
at x such that
$U^x_{n+1}\subseteq U^x_n$
for any
$n\in \omega $
, and
$\mathcal B=\bigcup _{x\in X}\mathcal B_x$
is a base of X of size
$\kappa $
satisfying the following condition:
-
(i) for each
$x\in X$ and
$U_n^x\in \mathcal B_x$ there exists
$f_{U_n^x}\in C(X,[0,1])$ such that
$f_{U_n^x}{\restriction }_{X\setminus U_n^x}\equiv 0$ and
$f_{U_n^x}{\restriction }_{U_{n+1}^x}\equiv 1$ .
Proof. Let
$\mathcal D$
be a base of X of size
$\kappa $
. Put

Claim 6.3. For every
$x\in X$
and open neighborhood U of x there exists
$\langle D_0,D_1\rangle \in \mathcal Y$
such that
$x\in D_1\subset D_0\subset U$
.
Proof. Fix any
$x\in X$
and an open neighborhood U of x. Since
$\mathcal D$
is a base, there exists an open set
$D_0\in \mathcal D$
such that
$x\in \mathcal D_0\subseteq U$
. Then there exists a continuous function
$h\in C(X,[0,1])$
such that
$h{\restriction }_{X\setminus D_0}\equiv 0$
and
$h(x)=1$
. Since the set
$h^{-1}([1/2,1])$
contains x in its interior, there exists
$D_1\in \mathcal D$
such that
$D_1\subset h^{-1}([1/2,1])$
. Clearly,
. It is easy to see that
$f{\restriction }_{D_1}\equiv 1$
and
$f{\restriction }_{X\setminus D_0}\equiv 0$
, i.e.,
$\langle D_0,D_1\rangle \in \mathcal Y$
.
Fix any
$x\in X$
and an open neighborhood base
$\mathcal V=\{V_n:n\in \omega \}\subset \mathcal D$
of x such that
$V_{n+1}\subset V_n$
for all
$n\in \omega $
. Wlog we can assume that
$V_0=X$
. Let
$U_0^x=X$
and
$f_{U_0^x}$
be the constant map from X to
$[0,1]$
with value
$1$
. Claim 6.3 implies that there exist a pair
$\langle D_0^1,D_1^1\rangle \in \mathcal Y$
such that
$x\in D_0^1\subset D_1^1\subseteq V_1$
and a function
$f_{\langle D_0^1,D_1^1\rangle }\in C(X,[0,1])$
such that
$f_{\langle D_0^1,D_1^1\rangle }{\restriction }_{D_1^1}\equiv 1$
and
$f_{\langle D_0^1,D_1^1\rangle }{\restriction }_{X\setminus D_0^1}\equiv 0$
. Put
$U_1^x=D_0^1$
and
$f_{U_1^x}=f_{\langle D_0^1,D_1^1\rangle }$
. Using Claim 6.3 again we can find a pair
$\langle D_0^2,D_1^2\rangle \in \mathcal Y$
such that
$x\in D_0^2\subset D_1^2\subseteq V_2\cap D_1^1$
and a function
$f_{\langle D_0^2,D_1^2\rangle }\in C(X,[0,1])$
such that
$f_{\langle D_0^2,D_1^2\rangle }{\restriction }_{D_1^2}\equiv 1$
and
$f_{\langle D_0^2,D_1^2\rangle }{\restriction }_{X\setminus D_0^2}\equiv 0$
. Put
$U_2^x=D_0^2$
and
$f_{U_2^x}=f_{\langle D_0^2,D_1^2\rangle }$
. Proceeding this way, for each
$x\in X$
we obtain an open neighborhood base
$\{U_n^x:n\in \omega \}$
at x and a family
$\{f_{U_n^x}:U_n^x\in \mathcal B_x\}\subset C(X,[0,1])$
such that
$\mathcal B=\bigcup _{x\in X}\mathcal B_x$
satisfies condition (i). Since
$\mathcal B\subseteq \mathcal D$
we get that
$|\mathcal B|\leq \kappa $
.
Recall that by (
$\heartsuit $
) we denote the following consistent assumption: “
$\mathfrak b=\mathfrak c$
and there exists a
$P_{\mathfrak c}$
-point in
$\beta (\omega )$
”.
Proof of Theorem 3.8
We need to show that assuming (
$\heartsuit $
) every Tychonoff first-countable space X of weight
$\kappa <\mathfrak {c}$
embeds densely into a Tychonoff first-countable countably compact space.
Proof. Let X be a first-countable Tychonoff space such that
$w(X)<\mathfrak {c}$
. Without loss of generality we can assume that the underlying set of X is disjoint with
$\mathfrak c$
. The first countability of X implies that
$|X|\leq \mathfrak c^{\omega }=\mathfrak c$
. If X is countably compact, then there is nothing to prove. Otherwise, let

Fix any bijection
$h: \mathcal {D}\cup [\mathfrak {c}]^{\omega }\rightarrow \mathfrak {c}$
such that
$h(a)\geq \sup (a)$
for any
$a\in [\mathfrak {c}]^{\omega }$
. Next, for every
$\alpha \leq \mathfrak {c}$
we shall recursively construct a topology
$\tau _{\alpha }$
on
$X_{\alpha }\subseteq X\cup \alpha $
. For the sake of brevity we denote the space
$(X_{\alpha },\tau _{\alpha })$
by
$Y_{\alpha }$
. At the end, we will show that the space
$Y_{\mathfrak {c}}$
has the desired properties.
Let
$X_0=X$
. By Lemma 6.2, there exists a base
$\mathcal B_0=\bigcup _{x\in X}\mathcal B_{0}^x$
of size
$<\mathfrak c$
of the topology on X, where for each
$x\in X$
, the collection
$\mathcal B_{0}^x=\{U_{n,0}^x:n\in \omega \}$
is a nested open neighborhood base at x and satisfies the following:
$U_{0,0}^x=X$
; for every
$n>0$
and
$x\in X_0$
there exists a continuous function
$f_{(U_{n,0}^x, U_{n+1,0}^x)}:X\rightarrow [0,1]$
such that
$f_{(U_{n,0}^x, U_{n+1,0}^x)}{\restriction }_{X\setminus U_{n,0}^x}\equiv 0$
and
$f_{(U_{n,0}^x, U_{n+1,0}^x)}{\restriction }_{U^x_{n+1,0}}\equiv 1$
. We also set
$f_{(U_{0,0}^x, U_{1,0}^x)}\equiv 1$
.
Assume that for each
$\alpha <\xi $
the Tychonoff first-countable spaces
$Y_{\alpha }$
are already constructed by defining a base
$\mathcal B_{\alpha }=\bigcup _{x\in X_{\alpha }}\mathcal B_{\alpha }^x$
of the the topology
$\tau _{\alpha }$
on a set
$X_{\alpha }\subseteq X\cup \alpha $
, where for each
$x\in X_{\alpha }$
, the collection
$\mathcal B_{\alpha }^x=\{U_{n,\alpha }^x:n\in \omega \}$
is a nested open neighborhood base at x. Additionally assume that
$X_{\alpha }\subset X_{\beta }$
for any
$\alpha \in \beta $
, and we have in parallel constructed a family

where for each
$(U_{n,\alpha }^x, U_{n+1,\alpha }^x)$
the function
$f_{(U_{n,\alpha }^x, U_{n+1,\alpha }^x)}: Y_{\alpha }\rightarrow [0,1]$
is continuous and

In what follows, for brevity, we put
$E(x,n,\alpha )=(U_{n,\alpha }^x, U_{n+1,\alpha }^x)$
for all
$n\in \omega $
,
$\alpha <\xi $
and
$x\in X_0$
.
The families
$\mathcal B_{\alpha }$
and
$\Phi _{\alpha }$
are presumed to satisfy the following conditions for all
$\alpha <\xi $
:
-
(1)
$|\mathcal B_{\alpha }|<\mathfrak c$ ;
-
(2)
$U_{0,\alpha }^x=X_{\alpha }$ for each
$x\in X_{\alpha }$ ;
-
(3)
$f_{E(x,0,\alpha )}\equiv 1$ for any
$x\in X_{\alpha }$ ;
-
(4) for every
$x\in X_{\alpha }$ and
$n\in \omega $ we have
$f_{E(x,n,\alpha )}{\restriction }_{U_{n+1,\alpha }^x}\equiv 1$ and
$f_{E(x,n,\alpha )}{\restriction }_{X_\alpha \setminus U_{n,\alpha }^x} \equiv 0$ ;
-
(5) for every
$\beta <\alpha $ ,
$n\in \omega $ and
$x\in X_{\beta }$ we have
$U_{n,\beta }^x=U_{n,\alpha }^x\cap X_\beta $ and
$\operatorname {cl}_{Y_\alpha }(U_{n,\beta }^x)=\operatorname {cl}_{Y_\alpha }(U_{n,\alpha }^x)$ ;
-
(6) for every
$\beta <\alpha $ ,
$n\in \omega $ and
$x\in X_{\beta }$ we have
$f_{E(x,n,\alpha )}{\restriction }_{X_{\beta }}=f_{E(x,n,\beta )}$ .
There are three cases to consider:
-
1)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is not an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
2)
$\xi =\gamma +1$ for some
$\gamma \in \mathfrak {c}$ and
$h^{-1}(\gamma )\cap X_{\gamma }$ is an infinite closed discrete subset of
$Y_{\gamma }$ ;
-
3)
$\xi $ is a limit ordinal.
1) Let
$X_{\xi }=X_{\gamma }$
,
$\mathcal B_{\xi }=\mathcal B_{\gamma }$
and
$\Phi _\xi =\Phi _{\gamma }$
.
2) Put
$X_{\xi }=X_{\gamma }\cup \{\gamma \}$
. Let
$h^{-1}(\gamma )=\{z_i\}_{i\in \omega }$
. Fix any
$P_{\mathfrak {c}}$
-point p which exists by the assumption. Since the set
$h^{-1}(\gamma )$
is closed and discrete, for any
$x\in X_{\gamma }$
there exists a positive integer
$k(x)$
such that
$|U_{k(x),\gamma }^{x}\cap h^{-1}(\gamma )|\leq 1$
. Since
$U_{0,\gamma }^x=X_{\gamma }$
and the ultrafilter p is free, for any
$x\in X_{\gamma }$
there exists
$m(x)<k(x)$
such that

Note that the set
$F_x$
actually depends on the pair
$\big ( U_{m(x),\gamma }^x, U_{m(x)+1,\gamma }^x\big )=E(x,m(x),\gamma )$
.
Fix an arbitrary
$f\in \Phi _{\gamma }$
. By the compactness of
$[0,1]$
, the ultrafilter
$f(p)$
on
$[0,1]$
generated by the family
$\{f(P): P\in p\}$
converges to a point
$b\in [0,1]$
. That is for each
$n\in \mathbb N$

Since p is a P-point, there exists
$G_f\in p$
such that
$G_f\subset ^*P_{f,1/n} $
for every
$n\in \mathbb N$
. It is easy to see that the sequence
$\{f(z_i):i\in G_f\}$
converges to b.
Since
$|\mathcal B_\gamma \cup \Phi _{\gamma }|<\mathfrak {c}$
we obtain that
$|\{F_x:x\in X_\gamma \}\cup \{G_f: f\in \Phi _{\gamma }\}|<\mathfrak c,$
and hence there exists
$F\in p$
such that
$F\subset ^{*}F_x\cap G_f$
for any
$x\in X_{\gamma }$
and
$f\in \Phi _\gamma $
.
For any
$n\in \omega $
and
$x\in X_{\gamma }$
set

and note that
$b^x_n=0$
for
$n>m(x)$
.
Let
$d_{\gamma }=\{z_i: i\in F\}.$
Condition (4) implies that
$\operatorname {cl}_{Y_{\gamma }}(U^x_{n+1,\gamma })\subseteq U^x_{n,\gamma }$
for every
$x\in X_\gamma $
and
$n\in \omega $
. Then for any
$x\in X_{\gamma }$
we have

It follows that for every
$x\in X_{\gamma }$
there exists a function
$g_x\in \omega ^F$
such that for all but finitely many
$i\in F$
the following holds:

By the continuity, for any
$f\in \Phi _{\gamma }$
there is a function
$\pi _f\in \omega ^F$
such that

Let
$h_{\gamma }\in \omega ^{F}$
be a function such that
$h_{\gamma }\geq ^* g_{x}$
and
$h_{\gamma }\geq ^*\pi _f$
for each
$x\in X_{\gamma }$
and
$f\in \Phi _\gamma $
. Such a function
$h_{\gamma }$
exists because
$|\mathcal B_{\gamma }\cup \Phi _{\gamma }|<\mathfrak {b}=\mathfrak {c}$
. Without loss of generality we can additionally assume that
$U_{h_{\gamma }(i),\gamma }^{z_i}\cap U_{h_{\gamma }(j),\gamma }^{z_j}=\emptyset $
for each distinct
$i,j\in \omega $
.
We are in a position now to define the open neighborhood base
$\mathcal B^\gamma _\xi $
at the point
$\gamma $
in the space
$Y_{\xi }$
: put

for all
$n\in \mathbb N$
.
Next, we define
$\mathcal B_{\xi }^x$
for
$x\in X_{\gamma }$
. Let
$U_{0,\xi }^x=X_{\xi }$
for every
$x\in X_{\gamma }$
. For each
$n\in \mathbb N$
let
$U_{n,\xi }^x=U_{n, \gamma }^x$
if either
$n\geq m(x)+1$
or
$n=m(x)$
and
$b_n^x=0$
, otherwise, put
$U_{n,\xi }^x=U_{n,\gamma }^x\cup \{\gamma \}$
. It is easy to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: x\in X_{\xi },n\in \omega \}$
forms a base of a topology
$\tau _{\xi }$
, and for each
$x\in X_{\xi }$
the family
$\mathcal B_{\xi }^{x}=\{U_{n,\xi }^x:n\in \omega \}$
forms an open neighborhood base at x in
$Y_{\xi }$
. Clearly, the first equality in condition (5) is satisfied for every
$\beta <\alpha \leq \xi $
and
$x\in X_\beta $
. Fix an arbitrary
$x\in X_{\gamma }$
. The choice of
$g_x$
and the inequality
$g_x\leq ^*h_\gamma $
ensure the existence of large enough
$n\in \omega $
such that
$U^x_{m(x)+2,\xi }\cap U_{n,\xi }^\gamma =\emptyset $
. Thus
$Y_\xi $
is Hausdorff. The next claim proves the second equality in condition (5) for
$\alpha =\gamma $
, and therefore also for any
$\alpha <\xi $
.
Claim 6.4. In the space
$Y_\xi $
,
$U_{n,\gamma }^x$
is dense in
$U_{n,\xi }^x$
for all
$n\in \omega $
and
$x\in X_\gamma $
.
Proof. It suffices to consider the case
$U_{n,\xi }^x=U_{n,\gamma }^x\cup \{\gamma \}$
. Thus either
$n< m(x)$
or
$n=m(x)$
and
$b_n^x>0$
. In both of these cases
$z_i\in U_{n,\gamma }^x$
for almost all
$i\in F$
, and hence
$U_{n,\gamma }^x$
intersects all neighborhoods of
$\gamma $
in
$Y_\xi $
.
Similarly as before, for each
$n\in \omega $
and
$x\in X_\xi $
put
$E(x,n,\xi )=(U_{n,\xi }^x, U_{n+1,\xi }^x)$
. Now we shall define a family
$\Phi _\xi =\{f_{E(x,n,\xi )}:n\in \omega , x\in X_\xi \}$
such that conditions (3), (4) and (6) are satisfied when
$\alpha $
is replaced with
$\xi $
. Put
$f_{E(x,0,\xi )}\equiv 1$
for all
$x\in X_{\xi }$
. For each
$x\in X_{\gamma }$
and
$n\in \mathbb N$
set

Conditions
$(\circ _0)$
,
$(\circ _1)$
together with the choice of
$b^x_n$
as the limit of the sequence
$\{f_{E(x,n,\gamma )}(z_i):i\in F\}$
, ensure the continuity of functions
$f_{E(x,n,\xi )}$
for all
$n\in \omega $
and
$x\in X_\gamma $
. Finally, we set

Since the family
$\{U_{h_{\gamma }(i),\gamma }^{z_i}:i\in \omega \}$
is locally finite in
$Y_\gamma $
, it remains to check that
$f_{E(\gamma ,n,\xi )}$
is continuous at
$\gamma $
. For this it is enough to prove the first equality of condition (4) for
$x=\gamma $
and
$\xi =\alpha $
. This follows from the following equation:

The fact that functions from
$\Phi _\xi $
satisfy the first equation from (4) also for
$x\in X_{\gamma }$
follows from (5) established in Claim 6.4 and from the continuity of functions in
$\Phi _\xi $
. It is clear that any function from
$\Phi _\xi $
satisfies the second equation of condition (4). Finally, it is a tedious routine to check that the families
$\mathcal B_{\xi }$
and
$\Phi _{\xi }$
satisfy conditions (1), (2), (3), and (6), which concludes case 2.
3) Let
$X_{\xi }=\bigcup _{\alpha \in \xi }X_{\alpha }$
. For each
$x\in X_{\xi }$
let
$\theta _x=\min \{\alpha :x\in X_{\alpha }\}$
. For each
$x\in X_{\xi }$
and
$n\in \omega $
put

It is straightforward to check that the family
$\mathcal B_{\xi }=\{U_{n,\xi }^x: n\in \omega , x\in X_{\xi }\}$
forms a base of a topology
$\tau _{\xi }$
on
$X_\xi $
, and the families
$\Phi _{\xi }=\{f_{E(x,n,\xi )}: n\in \omega , x\in X_\xi \}$
and
$\mathcal B_{\xi }$
satisfy conditions (1)–(6). Thus, it remains to establish the continuity of functions in
$\Phi _\xi $
, which we do next.
Assume to the contrary that there exists
$f_{E(x,n,\xi )}\in \Phi _{\xi }$
which is not continuous at a point
$z\in X_{\xi }$
. Then there exists a sequence
$\{a_k\}_{k\in \omega }$
such that
$\lim _{k\in \omega }a_k=z$
but

For each
$k\in \omega $
let
$\mu _k\in \xi $
be such that
$\{a_{k},x,z\}\subset X_{\mu _k}$
. Since
$Y_{\xi }$
is first-countable,
$X_0$
is a dense subset of
$X_{\xi }$
and the functions
$f_{E(x,n,\mu _k)}$
,
$k\in \omega $
are continuous on
$Y_{\mu _k}$
, for each
$k\in \omega $
we can pick
$b_k\in X_0$
such that

For each
$k\in \omega $
, by condition (6),
$f_{E(x,n,\xi )}{\restriction }_{ X_{\mu _k}}=f_{E(x,n,\mu _k)}$
implying that

Since the function
$f_{E(x,n,\mu _0)}$
is continuous on
$Y_{\mu _0}$
condition (6) implies the following:

But then

where the latter equality follows from equation
$(\bullet )$
. The obtained contradiction ensures that all functions in
$\Phi _{\xi }$
are continuous, and thus concludes case 3.
By the construction, the space
$Y_{\mathfrak {c}}$
is Tychonoff, first-countable and contains X as a dense subspace. Let A be any countable subset of
$Y_{\mathfrak {c}}$
. If the set
$B=A\cap \mathfrak {c}$
is infinite, then consider
$h(B)\in \mathfrak {c}$
. By the construction, either B has an accumulation point in
$Y_{h(B)}$
or
$h(B)$
is an accumulation point of B in
$Y_{h(B)+1}$
. In both cases B has an accumulation point in
$Y_{\mathfrak c}$
. If
$A\subset ^* X$
, then either it has an accumulation point in X, or A is closed and discrete in X. In the latter case either A has an accumulation point in
$Y_{h(A)}$
or
$h(A)$
is an accumulation point of A in
$Y_{h(A)+1}$
. Thus
$Y_{\mathfrak c}$
is countably compact, which completes our proof.
Proof of Theorem 3.10
We need to show that the following assertions are equivalent:
-
(1)
$\mathfrak b=\mathfrak s=\mathfrak c$ .
-
(2) Every first-countable zero-dimensional Hausdorff space of weight
$<\mathfrak c$ embeds densely into a first-countable zero-dimensional pseudocompact space.
-
(3) Every first-countable zero-dimensional Hausdorff space of cardinality
$<\mathfrak c$ embeds densely into a first-countable zero-dimensional pseudocompact space.
Proof. The implication (1)
$\Rightarrow $
(2) follows from Proposition 4.11. Since the weight of an infinite first-countable space doesn’t exceed its cardinality, implication (2)
$\Rightarrow $
(3) holds.
(3)
$\Rightarrow $
(1). Corollary 5.7 implies that
$\mathfrak b=\mathfrak c$
. Assuming that
$\mathfrak s<\mathfrak c$
, we get that the normal first-countable space Y constructed in Proposition 5.2 has cardinality
$\mathfrak s<\mathfrak c$
and, thus, is zero-dimensional. Then, by the assumption, Y embeds densely into a first-countable zero-dimensional pseudocompact space. But this contradicts Proposition 5.2. Hence
$\mathfrak b=\mathfrak s=\mathfrak c$
.
Proof of Theorem 3.11
We need to show that a subspace X of the Cantor space is a
$\lambda $
-set if and only if the Pixley-Roy hyperspace
$\operatorname {PR}(X)$
embeds densely into a first-countable pseudocompact space.
Proof. The implication (
$\Rightarrow $
) follows from [Reference Bell5, Theorem 4.2].
The implication (
$\Leftarrow $
) follows from Proposition 5.5.
Acknowledgments
We would like to thank Prof. Alan Dow for an idea to prove Proposition 5.2.
Funding
The research of the first named author was funded in whole by the Austrian Science Fund FWF [10.55776/ESP399]. The research of the third named author was funded in whole by the Austrian Science Fund FWF [10.55776/I5930].