betaFeedback

WorksMathematics

Pseudocompact MAD families under b=c\mathfrak b=\mathfrak c

Abstract

We prove that if the bounding number b\mathfrak b equals the cardinality of the continuum c\mathfrak c, then every infinite almost disjoint family of size less than c\mathfrak c can be extended to a maximal almost disjoint family whose Isbell–Mrówka space has a pseudocompact Vietoris hyperspace.

Introduction

The Vietoris hyperspace of a topological space is a space whose points are the nonempty closed subsets of the original space. Its topology was introduced by Leopold Vietoris in 1922 [18]. For compact metric spaces, it coincides with the topology induced by the Hausdorff metric.

A central theme in hyperspace theory is to understand how topological properties of a space are reflected in its hyperspace and vice versa. A classical result due to Vietoris states that the Vietoris hyperspace of a topological space is compact if and only if the original space is compact [18]. Thus, it is natural to ask what are the relations between generalizations of compactness between a space and its hyperspace.

In this direction, Ginsburg [7] investigated countable compactness and pseudocompactness of Vietoris hyperspaces. He proved that, for a Hausdorff space XX:

(i) If XX is pp-compact for some free ultrafilter p∈ω∗p \in\omega^{*}, then exp⁡(X)\exp(X) is pp-compact.

(ii) If exp⁡(X)\exp(X) is countably compact, then XnX^{n} is countably compact for every n∈ωn \in\omega.

(iii) If all the powers of XX are countably compact, then exp⁡(X)\exp(X) is countably compact.

He also proved the following for a Tychonoff space XX.

(i’) If XX is pp-pseudocompact for some p∈ω∗p \in\omega^{*}, then exp⁡(X)\exp(X) is pp-pseudocompact.

(ii’) If exp⁡(X)\exp(X) is pseudocompact, then XnX^{n} is pseudocompact for every n∈ωn \in\omega.

Notice that (i’) and (ii’) are the pseudocompact analogues of (i) and (ii), respectively. Moreover, (iii) follows from (i): if all powers of XX are countably compact, then XX is pp-compact for some p∈ω∗p \in\omega^{*} by the Ginsburg–Saks theorem [8 Theorem 2.6]. Applying (i) to this pp shows that exp⁡(X)\exp(X) is pp-compact, and hence countably compact. Despite that, as shown in [12], there exists a Tychonoff space XX such that all powers of XX are pseudocompact, while its hyperspace is not pseudocompact—namely, a subspace of βω\beta\omega.

Recall that if the ω\omega-th power of a Tychonoff space XX is pseudocompact, then all the powers of XX are pseudocompact. Ginsburg asked how the pseudocompactness of XωX^{\omega} is related to that of exp⁡(X)\exp(X) and which spaces XX have pseudocompact hyperspaces [7, Remark 3.2, p. 1399]. In this paper, we further explore these questions in the context of Isbell–Mrówka spaces.

Given an infinite almost disjoint family A\mathcal{A} on ω\omega, one associates to it a topological space Ψ(A)\Psi(\mathcal{A}) known as the Isbell–Mrówka space of A\mathcal{A} ([14] and [6 p. 51]). The pseudocompactness of Ψ(A)\Psi(\mathcal{A}) is equivalent to the maximality of A\mathcal{A}, and these spaces are Tychonoff. So the class of Isbell–Mrówka spaces provides a natural setting to study the Ginsburg’s questions, which becomes:

Problem 1.1. What is the relation between the maximality of an infinite almost disjoint family A\mathcal{A} and the pseudocompactness of the Vietoris hyperspace exp⁡(Ψ(A))\exp(\Psi(\mathcal{A}))?

In this paper, a pseudocompact MAD family is an infinite maximal almost disjoint family A\mathcal{A} such that exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})) is pseudocompact.

Several previous results have been obtained. The following theorem shows that a full equivalence between the maximality of an almost disjoint family and the pseudocompactness of the hyperspaces is equivalent to a known combinatorial cardinal.

Proposition 1.2 ([9 Theorem 2.4]). The following are equivalent:

(i) Every MAD family is pseudocompact.

(ii) n(ω∗)>cn(\omega^*) > \mathfrak{c}.

Here, n(ω∗)n(\omega^*) is the Novák number of ω∗\omega^*, which is the least cardinal κ\kappa such that ω∗\omega^* can be covered by κ\kappa nowhere dense sets. It is independent of ZFC whether n(ω∗)>cn(\omega^*) > \mathfrak{c}, e.g., it holds under p=c\mathfrak{p} = \mathfrak{c}, but fails under h<c\mathfrak{h} < \mathfrak{c}. In particular, if p=c\mathfrak{p} = \mathfrak{c}, then every MAD family is pseudocompact, and if h<c\mathfrak{h} < \mathfrak{c}, then there exists a MAD family that is not pseudocompact — two results that had previously appeared in [12].

Thus, the existence of a non-pseudocompact MAD family is independent of ZFC. On the other hand, the existence of a pseudocompact MAD family in ZFC alone is an open problem.

Problem 1.3. Does there exist a pseudocompact MAD family in ZFC?

This question appears in [12 Problem 4.3], [11 Problem 9], [10 Question 8.9], [9 Question 1.3], [4 Question 1.4], and [5 Question 2]. The first author has also presented talks on partial results regarding this problem at the 56th Spring Topology and Dynamical Systems Conference and at the 38th Summer Conference on Topology and its Applications [15, 16].

Besides p=c\mathfrak{p} = \mathfrak{c}, previous work shows that ♢(b)\diamondsuit(\mathfrak{b}) and ap=c\mathfrak{ap} = \mathfrak{c} also imply the existence of a pseudocompact MAD family [5]. Brendle introduced and studied the cardinal ap\mathfrak{ap}, which satisfies p≤ap≤b\mathfrak{p} \leq\mathfrak{ap} \leq\mathfrak{b} [3]. He also shows that the strict inequality ap<b\mathfrak{ap} < \mathfrak{b} is consistent with ZFC. We strengthen the previous result by proving generic existence under b=c\mathfrak{b} = \mathfrak{c} in the sense of [9 Section 3].

Theorem 1.4. If b=c\mathfrak{b} = \mathfrak{c}, then every infinite almost disjoint family of size less than c\mathfrak{c} can be extended to a pseudocompact MAD family.

Together with the previous results, we conclude that in a very broad class of models of ZFC, there exists a pseudocompact MAD family. However, the question of whether there exists a pseudocompact MAD family in ZFC remains open.

Preliminaries

For subsets of ω\omega, write x⊆∗yx \subseteq^* y if x∖yx \setminus y is finite, and x=∗yx =^* y if their symmetric difference is finite. An almost disjoint (AD) family is a collection of infinite subsets of ω\omega whose distinct members have finite intersection. A MAD family (maximal almost disjoint family) is an infinite AD family maximal under inclusion. Write r⊥Ar \perp\mathcal{A} if r∩ar \cap a is finite for every a∈Aa \in\mathcal{A}.

For f,g∈ωωf,g \in\omega^\omega, write f≤∗gf \leq^* g if f(n)≤g(n)f(n) \leq g(n) for all but finitely many nn. The bounding number b\mathfrak{b} is the least cardinality of a subset of ωω\omega^\omega with no common ≤∗\leq^*-upper bound. The dominating number d\mathfrak{d} is the least cardinality of a set D⊆ωωD \subseteq\omega^\omega such that for every f∈ωωf \in\omega^\omega there exists g∈Dg \in D such that f≤∗gf \leq^* g. The almost disjointness number a\mathfrak{a} is the least cardinality of a MAD family. For background on these cardinal characteristics of the continuum, see [1].

Given an almost disjoint family A\mathcal{A}, the Isbell–Mrówka space associated to A\mathcal{A} is denoted by Ψ(A)\Psi(\mathcal{A}) and has underlying set ω∪A\omega\cup\mathcal{A}. It has the topology for which the natural numbers are isolated, and a neighborhood base at a∈Aa \in\mathcal{A} consists of the sets {a}∪(a∖s)\{a\} \cup(a \setminus s), s∈[ω]<ωs \in[\omega]^{<\omega}. This is the strongest topology on Ψ(A)\Psi(\mathcal{A}) for which every point a∈A⊆Ψ(A)a \in\mathcal{A} \subseteq\Psi(\mathcal{A}) is the limit of the increasing enumeration of aa. It is well known that Ψ(A)\Psi(\mathcal{A}) is Tychonoff, locally compact, zero-dimensional and separable. Notice also that every subset of A\mathcal{A} is closed in Ψ(A)\Psi(\mathcal{A}). For background on almost disjoint families and Isbell–Mrówka spaces, see [11] and [10].

A space is pseudocompact if every continuous real-valued function on it is bounded.

Proposition 2.1 ([12]). The following are equivalent for an infinite almost disjoint family A\mathcal{A}: (i) A\mathcal{A} is a MAD family. (ii) The space Ψ(A)\Psi(\mathcal{A}) is pseudocompact. (iii) The space Ψ(A)ω\Psi(\mathcal{A})^\omega is pseudocompact.

For a topological space XX, exp⁡(X)\exp(X) denotes its nonempty closed subsets with the Vietoris topology. A basic open set of exp⁡(X)\exp(X) is of the form

⟨U0,…,Um−1⟩={H∈exp⁡(X):H⊆⋃i<mUi and H∩Ui≠∅ for all i<m},\langle U_0,\ldots,U_{m-1}\rangle= \left\{H \in\exp(X): H \subseteq\bigcup_{i<m} U_i \text{ and } H \cap U_i \ne\varnothing\text{ for all } i<m\right\},

where 0<m<ω0 < m < \omega and each UiU_i is open in XX. If U⊆XU \subseteq X, we put

U+=⟨U⟩={H∈exp⁡(X):H⊆U}andU−=⟨X,U⟩={H∈exp⁡(X):H∩U≠∅}.U^+ = \langle U\rangle= \{H \in\exp(X): H \subseteq U\} \quad\text{and}\quad U^- = \langle X,U\rangle= \{H \in\exp(X): H \cap U \ne\varnothing\}.

If XX is a regular space, then exp⁡(X)\exp(X) is Hausdorff [13]. Thus, for every almost disjoint family A\mathcal{A}, the hyperspace exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})) is Hausdorff.

We say that an AD family A\mathcal{A} is pseudocompact if exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})) is pseudocompact.

Following [5], a fin-sequence is a sequence C=(Cn)n∈ωC = (C_n)_{n\in\omega} of pairwise disjoint nonempty finite subsets of ω\omega. For x,X⊆ωx, X \subseteq\omega and a finite family FF of subsets of ω\omega, set

πC(x)={n∈ω:Cn∩x≠∅},\pi_C(x) = \{n \in\omega: C_n \cap x \ne\varnothing\},
EC(X)={n∈ω:Cn⊆X},E_C(X) = \{n \in\omega: C_n \subseteq X\},
JC(F)=⋂b∈FπC(b),JC(∅)=ω.J_C(F) = \bigcap_{b\in F} \pi_C(b), \qquad J_C(\varnothing) = \omega.

For a family BB, write πC[B]={πC(b):b∈B}\pi_C[B] = \{\pi_C(b) : b \in B\}. A family of subsets of ω\omega is centered if every finite intersection is infinite. In particular, the empty family is centered. We use the following criterion to study pseudocompact MAD families.

Proposition 2.2 (Rodrigues–Tomita [17 Proposition 2.1]). For an AD family A\mathcal{A}, the hyperspace exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})) is pseudocompact if and only if every fin-sequence has an accumulation point in it.

To produce accumulation points, we use the following accumulation criterion which appears implicitly in [9 Lemma 2.2]. We include a proof here for the convenience of the reader.

Lemma 2.3. Let A\mathcal{A} be an almost disjoint family, let CC be a fin-sequence and let B⊆AB \subseteq\mathcal{A} be nonempty.

(i) Every accumulation point of the sequence C:ω→exp⁡(Ψ(A))C : \omega\to\exp(\Psi(\mathcal{A})) is a nonempty subset of A\mathcal{A}.

(ii) The set BB is an accumulation point of CC if and only if for every F∈[B]<ωF \in[B]^{<\omega} and every X⊆ωX \subseteq\omega such that ∀b∈B b⊆∗X\forall b \in B\ b \subseteq^* X, we have

∣JC(F)∩EC(X)∣=ω.\lvert J_C(F) \cap E_C(X) \rvert= \omega.

Proof. Suppose that HH is an accumulation point of CC. Assume by contradiction that HH is not a subset of A\mathcal{A} and fix k∈H∩ωk \in H \cap\omega. As CC is a pairwise disjoint sequence,

∣{i∈ω:Ci∈({k})−}∣≤1,\lvert\{i \in\omega: C_i \in(\{k\})^{-}\}\rvert\le1,

a contradiction as ({k})−(\{k\})^{-} is a neighborhood of HH in exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})).

Now assume BB is an accumulation point and fix XX and FF as in the statement, so the set U=B∪XU = B \cup X is open in Ψ(A)\Psi(\mathcal{A}). For each b∈Fb \in F, let Vb={b}∪bV_b = \{b\} \cup b, which is also open. Then put

U=U+∩⋂b∈FVb−.\mathcal{U} = U^{+} \cap\bigcap_{b\in F} V_b^{-}.

Notice that {n∈ω:Cn∈U}=JC(F)∩EC(X)\{n \in\omega: C_n \in\mathcal{U}\} = J_C(F) \cap E_C(X), thus, the latter must be infinite. Conversely, let ⟨U0,…,Um⟩\langle U_0,\ldots,U_m\rangle be a basic neighborhood of BB. Choose bi∈B∩Uib_i \in B \cap U_i for each i≤mi \le m, and put F={bi:i≤m}F = \{b_i : i \le m\}. Since UiU_i is a neighborhood of bib_i, the set si=bi∖Uis_i = b_i \setminus U_i is finite. Set

X=(ω∩⋃i≤mUi)∖⋃i≤msi.X = \left(\omega\cap\bigcup_{i\le m} U_i\right) \setminus\bigcup_{i\le m} s_i.

Every b∈Bb \in\mathcal{B} belongs to some UiU_i, so b⊆∗Xb \subseteq^{*} X. Hence ∣JC(F)∩EC(X)∣=ω|J_{\mathcal{C}}(F) \cap E_{\mathcal{C}}(X)| = \omega by hypothesis. Now it is straightforward to check that

JC(F)∩EC(X)⊆{n∈ω:Cn∈⟨U0,…,Um⟩},J_{\mathcal{C}}(F) \cap E_{\mathcal{C}}(X) \subseteq\{n \in\omega: C_n \in\langle U_0,\ldots,U_m\rangle\},

which shows that B\mathcal{B} is an accumulation point of C\mathcal{C}.

An ideal I\mathcal{I} on ω\omega is free if [ω]<ω⊆I[\omega]^{<\omega} \subseteq\mathcal{I}. For an ideal I\mathcal{I}, let I+=P(ω)∖I\mathcal{I}^{+} = \mathcal{P}(\omega) \setminus\mathcal{I}. A generator of a free ideal I\mathcal{I} is a set G⊆I\mathcal{G} \subseteq\mathcal{I} such that I={X⊆ω:∃G∈G X⊆∗G}\mathcal{I} = \{X \subseteq\omega: \exists G \in\mathcal{G}\ X \subseteq^{*} G\}. We write χ(I)\chi(\mathcal{I}) for the least cardinality of such a generator.

The ideal generated by an AD family A\mathcal{A} and its complement are denoted by

I(A)={Y⊆ω:(∃K∈[A]<ω) Y⊆∗⋃K},I+(A)=P(ω)∖I(A).\mathcal{I}(\mathcal{A}) = \left\{Y \subseteq\omega: (\exists\mathcal{K} \in[\mathcal{A}]^{<\omega})\ Y \subseteq^{*} \bigcup\mathcal{K}\right\}, \qquad\mathcal{I}^{+}(\mathcal{A}) = \mathcal{P}(\omega) \setminus\mathcal{I}(\mathcal{A}).

Thus, if A\mathcal{A} is an infinite almost disjoint family, then χ(I(A))=∣A∣\chi(\mathcal{I}(\mathcal{A})) = |\mathcal{A}|. If A\mathcal{A} is a finite almost disjoint family, then χ(I(A))≤ω\chi(\mathcal{I}(\mathcal{A})) \leq\omega.

The following proposition is stated in [2 Remark 3.5], and a proof is implicit. For the convenience of the reader, we include a proof here.

Proposition 2.4. Every free ideal I\mathcal{I} with χ(I)<d\chi(\mathcal{I}) < \mathfrak{d} is a P+\mathcal{P}^{+}-ideal: every decreasing sequence (Tn)n∈ω(T_n)_{n\in\omega} in I+\mathcal{I}^{+} has an I∈I+I \in\mathcal{I}^{+} such that I⊆∗TnI \subseteq^{*} T_n for all nn.

Proof. Fix a generator B\mathcal{B} for I\mathcal{I} of cardinality χ(I)\chi(\mathcal{I}) and (Tn)n∈ω(T_n)_{n\in\omega} a decreasing sequence in I+\mathcal{I}^{+}. For each B∈BB \in\mathcal{B} and n∈ωn \in\omega, we have Tn̸⊆∗BT_n \not\subseteq^{*} B, so Tn∖BT_n \setminus B is infinite. Define fB∈ωωf_B \in\omega^\omega by

fB(n)=min⁡((Tn∖B)∖n).f_B(n) = \min((T_n \setminus B) \setminus n).

Then ∣{fB:B∈B}∣≤χ(I)<d|\{f_B : B \in\mathcal{B}\}| \leq\chi(\mathcal{I}) < \mathfrak{d}, so {fB:B∈B}\{f_B : B \in\mathcal{B}\} is not dominating. Hence there exists g∈ωωg \in\omega^\omega such that g̸≤∗fBg \not\leq^{*} f_B for every B∈BB \in\mathcal{B}. Put

I=⋃n∈ω(Tn∩[n,g(n))).I = \bigcup_{n\in\omega}(T_n \cap[n,g(n))).

The sequence (Tn)n∈ω(T_n)_{n\in\omega} is decreasing, and therefore, for each k∈ωk \in\omega,

I∖Tk⊆⋃n<k[n,g(n)),I \setminus T_k \subseteq\bigcup_{n<k}[n,g(n)),

which is finite. Thus I⊆∗TkI \subseteq^{*} T_k for every kk. Finally, fix B∈BB \in\mathcal{B}. There exist infinitely many nn such that fB(n)<g(n)f_B(n) < g(n), and for each such nn,

fB(n)∈(Tn∖B)∩[n,g(n))⊆I∖B.f_B(n) \in(T_n \setminus B) \cap[n,g(n)) \subseteq I \setminus B.

Since fB(n)≥nf_B(n) \geq n for each nn, we conclude that I∖BI \setminus B is infinite. Thus, for no B∈BB \in\mathcal{B} is I⊆∗BI \subseteq^{*} B, and therefore I∈I+I \in\mathcal{I}^{+}.

As a corollary, we obtain the following.

Corollary 2.5. Let H⊆[ω]ω\mathcal{H} \subseteq[\omega]^\omega be centered, ∣H∣<d|\mathcal{H}| < \mathfrak{d}, and let (Hn)n∈ω(H_n)_{n\in\omega} be a sequence in H\mathcal{H}. There exists I∈[ω]ωI \in[\omega]^\omega such that I⊆∗HnI \subseteq^{*} H_n for all n∈ωn \in\omega, and

∣I∩⋂K∣=ωfor every K∈[H]<ω.\left|I \cap\bigcap\mathcal{K}\right| = \omega\qquad\text{for every } \mathcal{K} \in[\mathcal{H}]^{<\omega}.

Proof. Let F\mathcal{F} be the filter generated by H\mathcal{H} together with the cofinite subsets of ω\omega. Since H\mathcal{H} is centered, F\mathcal{F} is a proper free filter. Let I=F∗={ω∖F:F∈F}\mathcal{I}=\mathcal{F}^{*}=\{\omega\setminus F:F\in\mathcal{F}\} be the dual ideal of F\mathcal{F}. Then I\mathcal{I} has a set of at most ∣H∣+ω<d|\mathcal{H}|+\omega<\mathfrak{d} generators so by Proposition 2.4, I\mathcal{I} is a P+P^{+}-ideal. Notice that F⊆I+\mathcal{F}\subseteq\mathcal{I}^{+}.

As F\mathcal{F} contains all finite intersections of members of H\mathcal{H}, by defining Tn=⋂j≤nHjT_n=\bigcap_{j\leq n}H_j for each n∈ωn\in\omega, we obtain a decreasing sequence (Tn)n∈ω(T_n)_{n\in\omega} in F⊆I+\mathcal{F}\subseteq\mathcal{I}^{+}. By Proposition 2.4, there exists I∈I+I\in\mathcal{I}^{+} such that I⊆∗Tn⊆HnI\subseteq^{*}T_n\subseteq H_n for all n∈ωn\in\omega. Now fix K∈[H]<ωK\in[\mathcal{H}]^{<\omega}. As KK is finite, ⋂K∈F⊆I+\bigcap K\in\mathcal{F}\subseteq\mathcal{I}^{+}, and therefore I̸⊆∗(ω∖⋂K)∈II\not\subseteq^{*}(\omega\setminus\bigcap K)\in\mathcal{I}, which implies that ∣I∖(ω∖⋂K)∣=∣I∩⋂K∣=ω|I\setminus(\omega\setminus\bigcap K)|=|I\cap\bigcap K|=\omega. □

The construction

Definition 3.1. Given an AD family A\mathcal{A} and a fin-sequence CC, we say that B⊆A\mathcal{B}\subseteq\mathcal{A} is maximal with centered CC-traces if πC[B]\pi_C[\mathcal{B}] is centered and πC[B∪{a}]\pi_C[\mathcal{B}\cup\{a\}] is not centered for every a∈A∖Ba\in\mathcal{A}\setminus\mathcal{B}.

Equivalently, besides centeredness, we require that for each a∈A∖Ba\in\mathcal{A}\setminus\mathcal{B} there be Fa∈[B]<ωF_a\in[\mathcal{B}]^{<\omega} such that

∣πC(a)∩JC(Fa)∣<ω.(1)|\pi_C(a)\cap J_C(F_a)|<\omega. \tag*{(1)}

We follow Corral and Rodrigues [5 Theorem 5] by listing each pair (C,X)(C,X) cofinally often and updating the families BCαB_C^\alpha by adding new members whenever their CC-traces preserve centeredness.

The preceding definition will be used to state Lemma 3.2, a key result for our construction that allows us to carry out the recursion under b=c\mathfrak{b}=\mathfrak{c}, which will be stated below and proved in the last section.

Theorem. If b=c\mathfrak{b}=\mathfrak{c}, then every infinite almost disjoint family of size less than c\mathfrak{c} can be extended to a pseudocompact MAD family.

Proof. Assume b=c\mathfrak{b}=\mathfrak{c} and fix an infinite AD family D\mathcal{D} of size κ<c\kappa<\mathfrak{c}. Let C\mathcal{C} be the collection of all fin-sequences. Fix an enumeration

((Cα,Xα):κ≤α<c)((C_\alpha,X_\alpha):\kappa\leq\alpha<\mathfrak{c})

of C×P(ω)\mathcal{C}\times\mathcal{P}(\omega) in which each pair occurs cofinally often. Such an enumeration exists because ∣C×P(ω)∣=c|\mathcal{C}\times\mathcal{P}(\omega)|=\mathfrak{c} and c⋅c=c\mathfrak{c}\cdot\mathfrak{c}=\mathfrak{c}: arrange for each pair to occur on a set of indices of size c\mathfrak{c}. Each such set is unbounded in the initial ordinal c\mathfrak{c}.

Enumerate D\mathcal{D} without repetitions as (aξ)ξ<κ(a_\xi)_{\xi<\kappa}. We recursively extend this enumeration to a sequence (aα)α<c(a_\alpha)_{\alpha<\mathfrak{c}}, simultaneously defining

Aα={aξ:ξ<α},BCα⊆Aα(α≤c, C∈C).\mathcal{A}_\alpha=\{a_\xi:\xi<\alpha\},\qquad B_C^\alpha\subseteq\mathcal{A}_\alpha\quad(\alpha\leq\mathfrak{c},\ C\in\mathcal{C}).

Start with A0=BC0=∅\mathcal{A}_0=B_C^0=\varnothing for every CC. Once aαa_\alpha has been specified, set Aα+1=Aα∪{aα}\mathcal{A}_{\alpha+1}=\mathcal{A}_\alpha\cup\{a_\alpha\} and, for every C∈CC\in\mathcal{C}, let

BCα+1={BCα∪{aα},if πC[BCα]∪{πC(aα)} is centered,BCα,otherwise.(2)B_C^{\alpha+1}= \begin{cases} B_C^\alpha\cup\{a_\alpha\}, & \text{if }\pi_C[B_C^\alpha]\cup\{\pi_C(a_\alpha)\}\text{ is centered},\\ B_C^\alpha, & \text{otherwise}. \end{cases} \tag*{(2)}

At a nonzero limit ordinal λ≤c\lambda\le\mathfrak{c}, put

Aλ=⋃ξ<λAξ,BCλ=⋃ξ<λBCξ(C∈C).\mathcal{A}_{\lambda}=\bigcup_{\xi<\lambda}\mathcal{A}_{\xi},\qquad B_{C}^{\lambda}=\bigcup_{\xi<\lambda}B_{C}^{\xi}\quad(C\in\mathcal{C}).

For α<κ\alpha<\kappa, use the prescribed aαa_{\alpha}. These clauses determine BCκB_{C}^{\kappa} for every CC. At each later stage, the choice of aαa_{\alpha} described below depends only on objects already defined at stage α\alpha, after which (3.2) is applied to all CC.

We record the invariant furnished by this rule. For each α≤c\alpha\le\mathfrak{c} and C∈CC\in\mathcal{C}, BCαB_{C}^{\alpha} is maximal with centered CC-traces in Aα\mathcal{A}_{\alpha}. Centeredness is preserved at successors by definition and at limits because each finite subfamily appears at an earlier stage. If aξa_{\xi} was not included in BCξ+1B_{C}^{\xi+1}, there was already a finite F⊆BCξF\subseteq B_{C}^{\xi} such that ∣πC(aξ)∩JC(F)∣<ω\lvert\pi_{C}(a_{\xi})\cap J_{C}(F)\rvert<\omega. That witness remains in every later candidate. This proves the invariant by induction.

Now let κ≤α<c\kappa\le\alpha<\mathfrak{c}, and suppose the construction has been carried out below α\alpha. Then Aα\mathcal{A}_{\alpha} is an infinite AD family of cardinality less than b\mathfrak{b}. Consider the following lemma:

Lemma 3.2. Let A\mathcal{A} be AD with ∣A∣<b\lvert\mathcal{A}\rvert<\mathfrak{b}, C\mathcal{C} a fin-sequence, and B⊆A\mathcal{B}\subseteq\mathcal{A} maximal with centered CC-traces. Suppose that X⊆ωX\subseteq\omega satisfies b⊆∗Xb\subseteq^{*}X for all b∈Bb\in\mathcal{B}, and that for some F0∈[B]<ωF_{0}\in[\mathcal{B}]^{<\omega},

∣JC(F0)∩EC(X)∣<ω.\lvert J_{C}(F_{0})\cap E_{C}(X)\rvert<\omega.

There exists r∈[ω]ωr\in[\omega]^{\omega} such that

r⊥A,∣r∖X∣=ω,πC[B]∪{πC(r)} is centered.r\perp\mathcal{A},\qquad\lvert r\setminus X\rvert=\omega,\qquad\pi_{C}[\mathcal{B}]\cup\{\pi_{C}(r)\}\text{ is centered}.

The hypotheses of the previous lemma motivate us to consider the following two conditions:

  • (i) b⊆∗Xαb\subseteq^{*}X_{\alpha} for every b∈BCααb\in B_{C_{\alpha}}^{\alpha};

  • (ii) there exists F∈[BCαα]<ωF\in[B_{C_{\alpha}}^{\alpha}]^{<\omega} such that ∣JCα(F)∩ECα(Xα)∣<ω\lvert J_{C_{\alpha}}(F)\cap E_{C_{\alpha}}(X_{\alpha})\rvert<\omega.

If both hold, apply Lemma 3.2 to Aα\mathcal{A}_{\alpha}, CαC_{\alpha}, BCααB_{C_{\alpha}}^{\alpha}, XαX_{\alpha} and choose the resulting set as aαa_{\alpha}. In particular,

aα⊥Aα,∣aα∖Xα∣=ω,aα∈BCαα+1.a_{\alpha}\perp\mathcal{A}_{\alpha},\qquad\lvert a_{\alpha}\setminus X_{\alpha}\rvert=\omega,\qquad a_{\alpha}\in B_{C_{\alpha}}^{\alpha+1}.

If one of the conditions fails, choose any infinite aα⊥Aαa_{\alpha}\perp\mathcal{A}_{\alpha}. Such a set exists because ∣Aα∣<b≤a\lvert\mathcal{A}_{\alpha}\rvert<\mathfrak{b}\le\mathfrak{a}. Update all candidates by (3.2). This completes the recursion.

Set

A=AcandBC=BCc(C∈C).\mathcal{A}=\mathcal{A}_{\mathfrak{c}}\qquad\text{and}\qquad B_{C}=B_{C}^{\mathfrak{c}}\quad(C\in\mathcal{C}).

The family A\mathcal{A} contains Aκ=D\mathcal{A}_{\kappa}=\mathcal{D}, is AD and has size c\mathfrak{c}, since every new infinite member is almost disjoint from its predecessors and hence distinct from them. For each CC, the family πC[BC]\pi_{C}[B_{C}] is centered. Moreover, BC≠∅B_{C}\ne\varnothing: at a stage with (Cα,Xα)=(C,∅)(C_{\alpha},X_{\alpha})=(C,\varnothing), either BCαB_{C}^{\alpha} is already nonempty, or conditions (i) and (ii) hold (with F=∅F=\varnothing), and aαa_{\alpha} is included.

We claim that BCB_{C} is an accumulation point of CC for every C∈CC\in\mathcal{C}. If not, then by Lemma 2.3(ii), there exist X⊆ωX\subseteq\omega and finite F⊆BCF\subseteq B_{C} such that

∀b∈BC (b⊆∗X)and∣JC(F)∩EC(X)∣<ω.\forall b\in B_{C}\ (b\subseteq^{*}X)\qquad\text{and}\qquad\lvert J_{C}(F)\cap E_{C}(X)\rvert<\omega.

Choose an occurrence (Cα,Xα)=(C,X)(C_\alpha, X_\alpha) = (C, X) after all members of F\mathcal{F} have been included in BCαB_C^\alpha. This is possible by the cofinal repetition of the pairs. Then, both conditions (i) and (ii) hold at stage α\alpha. Thus aα∈BCa_\alpha\in B_C and ∣aα∖X∣=ω|a_\alpha\setminus X| = \omega, a contradiction.

To check maximality, suppose Y∈[ω]ωY \in[\omega]^\omega were almost disjoint from A\mathcal{A}, and enumerate it without repetitions as (yn)n∈ω(y_n)_{n\in\omega}. The fin-sequence Cn={yn}C_n = \{y_n\} has the nonempty accumulation point BC⊆AB_C \subseteq\mathcal{A}. For any a∈BCa \in B_C, its lower Vietoris neighborhood determined by {a}∪a\{a\}\cup a would imply ∣Y∩a∣=ω|Y \cap a| = \omega, a contradiction. Hence A\mathcal{A} is MAD. Finally, Proposition 2.2 yields the pseudocompactness of exp⁡(Ψ(A))\exp(\Psi(\mathcal{A})). □

Proof of the extension lemma

It only remains to prove Lemma 3.2. We shall do so by establishing some auxiliary lemmas.

Lemma 4.1. Let A\mathcal{A} be AD with ∣A∣<b|\mathcal{A}| < \mathfrak{b}, and let CC be a fin-sequence. Suppose that (ai)i∈ω(a_i)_{i\in\omega} consists of distinct members of A\mathcal{A} and that πC(ai)\pi_C(a_i) is cofinite for every ii. There exist m0∈ωm_0 \in\omega and w⊥Aw \perp\mathcal{A} such that

w⊆⋃m≥m0Cmand∣w∩Cm∣=1(m≥m0).w \subseteq\bigcup_{m\ge m_0} C_m \quad\text{and}\quad|w\cap C_m|=1\quad(m\ge m_0).

Proof. For each ii, choose r(i)r(i) such that, for every m≥r(i)m \ge r(i), we can fix

xm,i∈Cm∩(ai∖⋃j<iaj).x_{m,i} \in C_m \cap\left(a_i \setminus\bigcup_{j<i}a_j\right).

This is possible as ai∩⋃j<iaja_i \cap\bigcup_{j<i}a_j is finite and πC(ai)\pi_C(a_i) is cofinite.

For a∈A∖{ai:i∈ω}a \in\mathcal{A}\setminus\{a_i:i\in\omega\}, choose fa∈ωωf_a \in\omega^\omega such that xm,i∉ax_{m,i} \notin a whenever m≥max⁡{r(i),fa(i)}m \ge\max\{r(i),f_a(i)\}. Such thresholds exist because a∩aia \cap a_i is finite.

Choose a strictly increasing q∈ωωq \in\omega^\omega which eventually dominates all the faf_a and satisfies q≥rq \ge r pointwise. This is possible since ∣A∣<b|\mathcal{A}| < \mathfrak{b}. For m≥q(0)m \ge q(0), put

k(m)=max⁡{i:q(i)≤m},w={xm,k(m):m≥q(0)}.k(m)=\max\{i:q(i)\le m\},\qquad w=\{x_{m,k(m)}:m\ge q(0)\}.

Then k(m)→∞k(m)\to\infty. Given j∈ωj\in\omega, notice that xm,k(m)∉ajx_{m,k(m)}\notin a_j whenever k(m)>jk(m)>j, so ∣w∩aj∣<ω|w\cap a_j|<\omega. If aa is outside the chosen countable subfamily, then m≥q(k(m))≥fa(k(m))m\ge q(k(m))\ge f_a(k(m)) for all sufficiently large mm, so the chosen point again avoids aa. Thus w⊥Aw\perp\mathcal{A}, and the conclusion follows with m0=q(0)m_0=q(0). □

Although the next observation is a well-known fact, we shall include its proof for the convenience of the reader. It will be used along with the fact that b≤a\mathfrak{b}\le\mathfrak{a}.

Lemma 4.2. If A\mathcal{A} is AD, ∣A∣<a|\mathcal{A}|<\mathfrak{a}, and Y∈I+(A)Y\in\mathcal{I}^{+}(\mathcal{A}), then some z∈[Y]ωz\in[Y]^\omega satisfies z⊥Az\perp\mathcal{A}.

Proof. Consider A↾Y={a∩Y:a∈A, ∣a∩Y∣=ω}\mathcal{A}\upharpoonright Y=\{a\cap Y:a\in\mathcal{A},\ |a\cap Y|=\omega\}. If this family is infinite, its size is less than a\mathfrak{a}, so it is not maximal on YY. Any infinite subset of YY almost disjoint from it is almost disjoint from all of A\mathcal{A}. If it is finite, removing its union from YY leaves an infinite set almost disjoint from A\mathcal{A}, since Y∈I+(A)Y\in\mathcal{I}^{+}(\mathcal{A}). □

Diagram showing one point per block

Figure 1. w={xm,k(m):m≥q(0)}w=\{x_{m,k(m)}:m\ge q(0)\}: one point per block.

For the following lemma, recall that b≤d\mathfrak{b}\le\mathfrak{d}.

Lemma 4.3. Let AA be AD with ∣A∣<b|A|<\mathfrak{b}, let CC be a fin-sequence, and let B⊆AB\subseteq A be infinite. If πC[B]\pi_C[B] is centered, then there exists w∈[ω]ωw\in[\omega]^\omega such that w⊥Aw\perp A and πC[B]∪{πC(w)}\pi_C[B]\cup\{\pi_C(w)\} is centered.

Proof. Choose distinct bi∈Bb_i\in B, i∈ωi\in\omega. Apply Corollary 2.5 to H=πC[B]\mathcal{H}=\pi_C[B] and Hi=πC(bi)H_i=\pi_C(b_i), using ∣H∣<d|\mathcal{H}|<\mathfrak{d} to obtain I∈[ω]ωI\in[\omega]^\omega almost contained in every πC(bi)\pi_C(b_i) and meeting every finite intersection from πC[B]\pi_C[B] infinitely.

Reindex (Cn)n∈I(C_n)_{n\in I} in increasing order. Each bib_i meets all but finitely many of these blocks. Lemma 4.1, applied to the entire family AA, gives w⊥Aw\perp A with πC(w)=∗I\pi_C(w)=^* I. Hence πC[B]∪{πC(w)}\pi_C[B]\cup\{\pi_C(w)\} is centered. □

Lemma 4.4. Let AA be AD, CC a fin-sequence, and B⊆AB\subseteq A maximal with centered CC-traces. Suppose that X⊆ωX\subseteq\omega satisfies b⊆∗Xb\subseteq^*X for all b∈Bb\in B, and that for some F0∈[B]<ωF_0\in[B]^{<\omega},

∣JC(F0)∩EC(X)∣<ω.|J_C(F_0)\cap E_C(X)|<\omega.

Then, for every F∈[B]<ωF\in[B]^{<\omega},

YF=(⋃n∈JC(F)Cn)∖X∈I+(A).Y_F=\left(\bigcup_{n\in J_C(F)}C_n\right)\setminus X\in\mathcal{I}^{+}(A).

Proof. Suppose YF⊆∗⋃KY_F\subseteq^*\bigcup K for some finite K⊆AK\subseteq A. For a∈K∖Ba\in K\setminus B, choose FaF_a as in (1), and put

G=F∪F0∪⋃a∈K∖BFa.G=F\cup F_0\cup\bigcup_{a\in K\setminus B}F_a.

The set JC(G)J_C(G) is infinite and has finite intersection with EC(X)E_C(X). Since the blocks are disjoint, YGY_G is infinite. For a∈K∩Ba\in K\cap B, the intersection YG∩aY_G\cap a is contained in the finite set a∖Xa\setminus X. For a∈K∖Ba\in K\setminus B, only finitely many blocks indexed by JC(G)J_C(G) meet aa, by (1). Thus YG∩aY_G\cap a is finite for every a∈Ka\in K, contradicting YG⊆YF⊆∗⋃KY_G\subseteq Y_F\subseteq^*\bigcup K. □

Finally, we restate and prove Lemma 3.2.

Lemma. Let A\mathcal{A} be AD with ∣A∣<b|\mathcal{A}| < \mathfrak{b}, C\mathcal{C} a fin-sequence, and B⊆A\mathcal{B} \subseteq\mathcal{A} maximal with centered C\mathcal{C}-traces. Suppose that X⊆ωX \subseteq\omega satisfies b⊆∗Xb \subseteq^* X for all b∈Bb \in\mathcal{B}, and that for some F0∈[B]<ωF_0 \in[\mathcal{B}]^{<\omega},

∣JC(F0)∩EC(X)∣<ω.|J_{\mathcal{C}}(F_0) \cap E_{\mathcal{C}}(X)| < \omega.

There exists r∈[ω]ωr \in[\omega]^\omega such that

r⊥A,∣r∖X∣=ω,πC[B]∪{πC(r)} is centered.r \perp\mathcal{A}, \qquad|r \setminus X| = \omega, \qquad\pi_{\mathcal{C}}[\mathcal{B}] \cup\{\pi_{\mathcal{C}}(r)\}\text{ is centered}.

Proof. If B\mathcal{B} is finite, apply Lemma 4.2 to YBY_{\mathcal{B}}. The resulting r∈[YB]ωr \in[Y_{\mathcal{B}}]^\omega is almost disjoint from A\mathcal{A}, is disjoint from XX, and satisfies πC(r)⊆JC(B)\pi_{\mathcal{C}}(r) \subseteq J_{\mathcal{C}}(\mathcal{B}). Its trace is infinite because the blocks are finite, so the required family is centered.

Suppose that B\mathcal{B} is infinite. By Lemma 4.3, choose w⊥Aw \perp\mathcal{A} such that πC[B]∪{πC(w)}\pi_{\mathcal{C}}[\mathcal{B}] \cup\{\pi_{\mathcal{C}}(w)\} is centered. Independently, Lemmas 4.4 and 4.2 give z∈[Y∅]ωz \in[Y_{\varnothing}]^\omega with z⊥Az \perp\mathcal{A}. Let r=w∪zr = w \cup z. Then r⊥Ar \perp\mathcal{A}, z⊆r∖Xz \subseteq r \setminus X, and πC(w)⊆πC(r)\pi_{\mathcal{C}}(w) \subseteq\pi_{\mathcal{C}}(r). The last inclusion preserves all the required infinite intersections.

Acknowledgements

This study was financed, in part, by the São Paulo Research Foundation (FAPESP), Brasil. Process Numbers 2025/07302-0 and 2026/06528-7.

Declaration of AI Assistance

The main mathematical results presented in this paper had the first draft of their proofs generated by OpenAI’s GPT-6 Astra. The authors carefully reviewed and verified the correctness of these proofs, rewriting them as necessary to ensure their accuracy and clarity. OpenAI’s GPT-6.1 Sol, GPT-6.1 Terra, GPT-6.2 Sol and GPT-6.2 Astra aided the authors with subsequent revisions of the proofs, grammar, and LaTeX formatting.

References

Instituto de Matemática, Estatística e Ciência da Computação, Universidade de São Paulo, Rua do Matão, 1010, 05508-090 São Paulo, SP, Brazil

Email address, V. de Oliveira Rodrigues: [email protected]

Email address, F. Y. Ynone: [email protected]

References

  1. [1]Andreas Blass. “Combinatorial Cardinal Characteristics of the Continuum”. In: Handbook of Set Theory. Ed. by Matthew Foreman and Akihiro Kanamori. Dordrecht: Springer, 2010, pp. 395–489. doi: 10.1007/978-1-4020-5764-9_7.DOI
  2. [2]Andreas Blass, Michael Hrušák, and Jonathan Verner. “On strong $P$-points”. In: Proceedings of the American Mathematical Society 141.8 (2013), pp. 2875–2883. doi: 10.1090/S0002-9939-2013-11518-2.DOI
  3. [3]Brendle, J. (1999). Dow’s Principle and Q-Sets. Canadian Mathematical Bulletin, 42(1), 13–24. https://doi.org/10.4153/CMB-1999-002-2
  4. [4]Corral, C., & Rodrigues, V. de O. (2024). Fin-intersecting MAD families. Filomat, 38(7), 2563–2578. https://doi.org/10.2298/FIL2407563C
  5. [5]Corral, C., & Rodrigues, V. de O. (2024). New examples of MAD families with pseudocompact hyperspaces. Topology and its Applications, 342, 108773. https://doi.org/10.1016/j.topol.2023.108773
  6. [6]Gillman, L., & Jerison, M. (1960). Rings of Continuous Functions. In The University Series in Higher Mathematics.
  7. [7]John Ginsburg. “Some results on the countable compactness and pseudocompactness of hyperspaces”. In: Canadian Journal of Mathematics 27.6 (1975), pp. 1392–1399. DOI: 10.4153/CJM-1975-142-9.DOI
  8. [8]John Ginsburg and Victor Saks. “Some applications of ultrafilters in topology”. In: Pacific Journal of Mathematics 57.2 (1975), pp. 403–418. DOI: 10.2140/pjm.1975.57.403.DOI
  9. [9]Guzmán, O., & al., . et . (2022). Maximal almost disjoint families and pseudocompactness of hyperspaces. Topology and its Applications, 305, 107872. https://doi.org/10.1016/j.topol.2021.107872
  10. [10]Fernando Hernández-Hernández and Michael Hrušák. “Topology of Mrówka-Isbell Spaces”. In: Pseudocompact Topological Spaces: A Survey of Classic and New Results with Open Problems. Ed. by Michael Hrušák, Ángel Tamariz-Mascarúa, and Mikhail Tkachenko. Vol. 55. Developments in Mathematics. Cham: Springer, 2018, pp. 253–289. DOI: 10.1007/978-3-319-91680-4_8.DOI
  11. [11]Michael Hrušák. “Almost Disjoint Families and Topology”. In: Recent Progress in General Topology III. Ed. by K. P. Hart, Jan van Mill, and Petr Simon. Paris: Atlantis Press, 2014, pp. 601–638. DOI: 10.2991/978-94-6239-024-9_14.DOI
  12. [12]Michael Hrušák, Fernando Hernández-Hernández, and Iván Martínez-Ruiz. “Pseudocompactness of hyperspaces”. In: Topology and its Applications 154.17 (2007), pp. 3048–3055. DOI: 10.1016/j.topol.2007.07.001.DOI
  13. [13]Ernest Michael. “Topologies on spaces of subsets”. In: Transactions of the American Mathematical Society 71.1 (1951), pp. 152–182. DOI: 10.1090/S0002-9947-1951-0042109-4.DOI
  14. [14]Stanisław Mrówka. “On completely regular spaces”. In: Fundamenta Mathematicae 41 (1954), pp. 105–106. DOI: 10.4064/fm-41-1-105-106.DOI
  15. [15]Rodrigues, V. de O. (2023). Fin-intersecting almost disjoint families. https://sites.rhodes.edu/sites/sites/files/STDC2023/Set%20Theoretic%20Topology/Vinicius%20de%20Oliveira%20Rodrigues.pdfsites.rhodes.edu/sites/sites/files/STDC2023/Set%20Theoretic%20Topology/Vinicius%20de%20Oliveira%20Rodrigues.pdf
  16. [16]Rodrigues, V. de O. (2024). Pseudocompact MAD families under weaker assumptions. https://www.mat.uc.pt/~sumtopo/pdfs/rodrigues_v.pdfmat.uc.pt/~sumtopo/pdfs/rodrigues_v.pdf
  17. [17]Rodrigues, V. de O., & Tomita, A. H. (2019). Small MAD families whose Isbell–Mrówka space has pseudocompact hyperspace. Fundamenta Mathematicae, 247(1), 99–108. https://doi.org/10.4064/fm657-10-2018
  18. [18]Leopold Vietoris. “Bereiche zweiter Ordnung”. In: Monatshefte für Mathematik und Physik 32 (1922), pp. 258–280.DOI

Paper details

vlnc:W_5tasmd39
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

GPT-6 Astra (bare), GPT-6.1 Sol (bare), GPT-6.1 Terra (bare), GPT-6.2 Sol, GPT-6.2 Astra

View statement
Looking for related papers…
Contents