I was wondering, is it true that a Extremally Disconnected Hausdorff Space is Regular? Let $A$ be an open set. I must find a open set $V$ such that $\bar V \subset A$. Since $\bar V$ will be open I think we can do it somehow but I'm not managing to construct it.

Recall that $X$ is extremally disconnected iff the closure of every open set is open.


Not necessarily.

$p \subseteq \mathcal{P}(\mathbb{N})$ is an ultrafilter on $\mathbb{N}$ if

  • for all $A \subseteq \mathbb{N}$ either $A \in p$ or $\mathbb{N} \setminus A \in p$ (but not both);
  • whenever $A \subseteq B \subseteq \mathbb{N}$ are such that $A \in p$, then $B \in p$;
  • whenever $A , B \in p$, then $A \cap B \in p$.

We say that $p$ is nonprincipal if no finite set belongs to $p$.

A basic fact about ultrafilters is that if $\mathcal{A} \subseteq \mathcal{P}(\mathbb{N})$ is nonempty and the intersection of any finite number of elements of $\mathcal{A}$ is nonempty, then there is an ultrafilter $p$ on $\mathbb{N}$ including $\mathcal{A}$ as a subset. It follows that given any infinite $A \subseteq \mathbb{N}$ there is a nonprincipal ultrafilter on $\mathbb{N}$ containing $A$.

Let $\mathcal{U}$ denote the collection of all nonprincipal ultrafilters on $\mathbb{N}$, and define $X = \mathbb{N} \cup \mathcal{U}$. We topologise $X$ as follows:

  • each $n \in \mathbb{N}$ is isolated;
  • given $p \in \mathcal{U}$, the basic open neighbourhoods of $p$ are of the form $A \cup \{ p \}$ for $A \in p$.

Claim. This space is Hausdorff but not regular.

sketch. Hausdorffness is fairly straightforward, you just have to handle the cases. For example, if $p , q \in \mathcal{U}$ are distinct, then there is some $A \subseteq \mathbb{N}$ which is in $p$ but not $q$. But then $\mathbb{N} \setminus A \in q$ (but not in $p$), and so $A \cup \{ p \}$, $( \mathbb{N} \setminus A ) \cup \{ q \}$ are disjoint open neighbourhoods of $p$ , $q$, respectively.

To show that it is not regular, it suffices to show that if $A \cup \{ p \}$ is a basic open neighbourhood of $p \in \mathcal{U}$, then there is a $q \in \overline{ A \cup \{ p \}} \cap \mathcal{U}$ which is distinct from $p$. For this, partition $A$ into two infinite subsets as $B \cup C$. Then one of these, say $B$, does not belong to $p$. Letting $q \in \mathcal{U}$ contain $B$, clearly $q \neq p$. Note that if $D \in q$, then $D \cap B \neq \varnothing$, and so $(D \cup \{ q \} ) \cap ( A \cup \{ p \} ) \neq \varnothing$. It follows that $q \in \overline{ A \cup \{ p \} }$.

Claim. This space is extremally disconnected.

sketch. Let $U \subseteq X$ be open. Note that $\overline{U} \setminus U \subseteq \mathcal{U}$ (since the points in $\mathbb{N}$ are isolated). Suppose that $p \in \overline{U} \setminus U$. Letting $A = U \cap \mathbb{N}$, it follows that $B \cap A \neq \varnothing$ for all $B \in p$. But this implies that $A \in p$ (since otherwise $\mathbb{N} \setminus A \in p$, but this is disjoint from $A$). And so $A \cup \{ p \}$ is an open neighbourhood of $p$ included in $\overline{U}$.


Vinicius, see if this link answers your question: http://www.ams.org/journals/proc/1967-018-02/S0002-9939-1967-0210066-0/S0002-9939-1967-0210066-0.pdf


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.