I'm reading the exposition of the Stone-Čech compactification for locales in Johnstone's book Stone Spaces. In Chapter IV Paragraph 2.2, Johnstone constructs the Stone-Čech compactification of a locale $A$ as the locale $C(A)$ of completely regular ideals in $A$, and in Paragraph 2.3 he shows that the points of $C(A)$ may be identified with the maximal completely regular filters in $A$. There are two points here where I was unable to follow the reasoning. Unfortunately, they're a bit technical.

Let's start with the definitions. Fix a locale $A$.

  • We say that $a$ is well inside $b$, written $a \eqslantless b$, if there exists $c$ such that $a\land c = 0_A$ and $b\lor c = 1_A$. Equivalently, we could just require that $b \lor \lnot a = 1_A$.
  • We say that $a$ is really inside $b$, written $a \,\unicode{x02A9B}\, b$ if there is a scale between $a$ and $b$, i.e. if there exist $\{c_q\mid q\in \mathbb{Q}\cap [0,1]\}$ such that $a\leq c_0$, $c_1\leq b$, and $c_p \eqslantless c_q$ for all $p<q$.
  • We say that $A$ is completely regular if $a = \bigvee_A\{b\mid b\,\unicode{x02A9B}\,a\}$ for all $a$.
  • We say that an ideal $I\subseteq A$ is completely regular if for all $a\in I$ there exists $b\in I$ with $a\,\unicode{x02A9B}\,b$.
  • We say that a filter $F\subseteq A$ is completely regular if for all $a\in F$ there exists $b\in F$ with $b\,\unicode{x02A9B}\,a$.
  • $C(A)$ is the locale of completely regular ideals of $A$. $C(A)$ is a subframe of $\mathrm{Idl}(A)$, so the meet of finitely many completely regular ideals is their intersection and join of some completely regular ideals is the ideal generated by the union. Further, $C(A)$ is compact and completely regular.

Question 1: Lemma 2.3 states that the poset of completely regular filters of $A$ is isomorphic to the poset of completely regular filters of $C(A)$. In the last step, Johnstone writes, "A similar argument shows $\overline{f}_*\overline{f}^*(G) = G$ for any completely regular filter $G$ of $C(A)$."

It doesn't matter what the functions $\overline{f}_*$, $\overline{f}^*$, $f_*$, and $f^*$ are: What matters is that for $J\in C(A)$, $f_*f^*(J) = \{a\in A\mid a \,\unicode{x02A9B}\, \bigvee_A J\}$, and $\overline{f}_*\overline{f}^*(G)$ is the filter of $C(A)$ generated by $f_*f^*(J)$ for all $J\in G$. Since $J$ is completely regular, $J\subseteq f_*f^*(J)$ for all $J$, so $f_*f^*(J) \in G$, and it is clear that $\overline{f}_*\overline{f}^*(G) \subseteq G$.

To prove the converse, we would like to take any completely regular ideal $I\in G$ and find some $J\in G$ such that $f_*f^*(J) \subseteq I$. Since $G$ is completely regular, there is some $J\in G$ with $J\,\unicode{x02A9B}\,I$, so it would be reasonable to try to prove that $J\,\unicode{x02A9B}\,I$ implies $f_*f^*(J)\subseteq I$. But I don't see how to do this.

Question 2: In the second half of Proposition 2.3, $A$ is a compact completely regular locale. We are given a maximal completely regular filter $F$, which we would like to show is prime. If $a\lor b\in F$ and $a\notin F$, we construct the filter $G = \{c\in A\mid c\lor b\in F\}$. It's easy to see that $G$ is a filter and $F\subsetneq G$. But Johnstone writes that $G$ "is easily seen to be completely regular".

So suppose $c\in G$. We need to find some $e\,\unicode{x02A9B}\,c$ with $e\in G$. Well, $c\lor b\in F$, and since $F$ is completely regular there is some $d\in F$ with $d\,\unicode{x02A9B}\,c\lor b$. Since $c\land (c\lor b) = c$, it would be natural to hope that $e = c\land d$ works. But it is not true in general that $x \,\unicode{x02A9B}\,y$ implies $z\land x \,\unicode{x02A9B}\, z\land y$.

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.