3
$\begingroup$

Let $\kappa > \omega$ an infinite cardinal and $\mathcal E \subset [\kappa]^\omega$ is a maximal almost disjoint family, where $[\kappa]^\omega = \{ A \subset \kappa: |A| = \omega \}$. Let $\Psi(\mathcal E)$ denote the topological space whose point-set is $\kappa \cup \mathcal E$, with the topology generated by isolating each $\alpha \in \kappa$, and the basic nbhds about $E \in \mathcal E$ are all sets of the form $\{E\}\cup (E\setminus F)$, where $F \in [E]^{< \omega}$. This we can called fat $\Psi$ space.

My question is this: does such space always have a $G_\delta$ diagonal for any $\kappa$?

Thanks ahead.

$\endgroup$
1
$\begingroup$

$X=\Psi(\mathscr{E})$ has a $G_\delta$ diagonal iff it has a family $\{\mathscr{U}_n:n\in\omega\}$ of open covers such that

$$\bigcap_{n\in\omega}\operatorname{st}(x,\mathscr{U}_n)=\{x\}\tag{1}$$

for each $x\in X$. Clearly each $\mathscr{U}_n$ may be replaced by a refinement, so we may assume that $\mathscr{U}_n$ has the following form.

For each $E\in\mathscr{E}$ let $E=\{\xi^E(n):n\in\omega\}$ be an enumeration of $E$. For $n\in\omega$ and $E\in\mathscr{E}$ let $B_n(E)=\{E\}\cup\left\{\xi^E(k):k\ge n\right\}$. Then for each $E\in\mathscr{E}$ there is an increasing function $\varphi_E:\omega\to\omega$ such that $$\mathscr{U}_n=\left\{B_{\varphi_E(n)}(E):E\in\mathscr{E}\right\}\cup\big\{\{\xi\}:\xi<\kappa\big\}$$ for each $n\in\omega$.

For $E\in\mathscr{E}$ and $n\in\omega$ we have $\operatorname{st}(E,\mathscr{U}_n)=B_{\varphi_E(n)}(E)$, so $(1)$ holds for each $x\in\mathscr{E}$. If $\eta<\kappa$ and $E\in\mathscr{E}$, then $E\in\operatorname{st}(\eta,\mathscr{U}_n)$ iff $\eta=\xi^E(k)$ for some $k\ge\varphi_E(n)$, so for each $E\in\mathscr{E}$ there is an $n\in\omega$ such that $E\notin\operatorname{st}(\eta,\mathscr{U}_n)$, and therefore

$$\bigcap_{n\in\omega}\operatorname{st}(\eta,\mathscr{U}_n)\subseteq\kappa\;.$$

Suppose that $\bigcap_{n\in\omega}\operatorname{st}(\eta,\mathscr{U}_n)=\{\eta\}$. Then for each $\xi\in\kappa\setminus\{\eta\}$ there is an $n\in\omega$ such that $\xi\notin\operatorname{st}(\eta,\mathscr{U}_n)$, i.e., there is no $E\in\mathscr{E}$ such that $\{\xi,\eta\}\subseteq B_{\varphi_E(n)}(E)$. Let $$m\big(\{\xi,\eta\}\big)=\min\left\{n\in\omega:\forall E\in\mathscr{E}\big(\{\xi,\eta\}\nsubseteq B_{\varphi_E(n)}(E)\big)\right\}\;;$$ then $m:[\kappa]^2\to\omega$.

Now suppose that $\kappa=\left(2^\omega\right)^+$. The $\lambda=\omega$ case of the partition relation $(2^\lambda)^+\to(\lambda^+)^2_\lambda$ says that there are an $n\in\omega$ and an uncountable $S\subseteq\kappa$ such that $m\big(\{\xi,\eta\}\big)=n$ for all $\{\xi,\eta\}\in[S]^2$. $\mathscr{E}$ is a MAD family, so there is an $E\in\mathscr{E}$ such that $|E\cap S|=\omega$. But then there are $k,\ell\in\omega$ such that $\xi^E(k),\xi^E(\ell)\in S$ and $k,\ell\ge\varphi_E(n)$, contradicting the definition of $m$. Thus, for $\kappa=\left(2^\omega\right)^+$ the space $X$ does not have a $G_\delta$-diagonal.

I’ve not been able to settle the question for $\omega_1\le\kappa\le 2^\omega$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.