4
$\begingroup$

Metric spaces have the following nice property: Every closed set is a countable intersection of open sets.

What other spaces have this property? Are there some nice known sufficient or necessary conditions on a topological space, for this property?

(e.g. do locally compact Hausdorff spaces satisfy it? )


The proof for metric spaces is very easy:

Let $A\subseteq X$ be closed. For all $n\in \mathbb N$ define $$U_n=\bigcup _{a\in A} B(a,\frac{1}{n}).$$ Each $U_n$ is open, and $A=\bigcap _{n\in \mathbb N} U_n$.

$\endgroup$
  • 1
    $\begingroup$ A topological space in which every closed set is a $G_\delta$-set (a countable intersection of open sets) is called a perfect space (not to be confused with a perfect set, a closed set with no isolated points); a normal space with this property is called a perfectly normal space. $\endgroup$ – bof Nov 9 '18 at 5:42
4
$\begingroup$

Such a space is called a $G_\delta$ space. A $G_\delta$ normal space is called "perfectly normal" and perfect normality is equivalent to normality plus every closed set being the vanishing set of some real-valued continuous function. Beyond metric spaces, another notable class of examples is that all CW-complexes are perfectly normal.

Not all compact Hausdorff spaces are $G_\delta$ spaces. For instance, the ordinal $\omega_1+1$ is compact Hausdorff with the order topology but the singleton $\{\omega_1\}$ is not a $G_\delta$ set.

$\endgroup$
  • $\begingroup$ What about $\beta\Bbb N$? $\endgroup$ – Asaf Karagila Nov 9 '18 at 8:05
3
$\begingroup$

As said above, these are $G_{\delta}$ sets.

For a normal space $X$ An equivalent description of these:

$A \subset X$ is a $G_{\delta}$ set iff $\exists f:[0,1] \to X$, continuous, s.t $f(a) = 0$ on $A$, and else $f(x) > 0$.

do locally compact Hausdorff spaces satisfy it?

No, as Eric Wofsey showed above, but it is true that every locally compact Hausdorff space is a Tychonoff space ($T_{3\frac{1}{2}}$)

$\endgroup$
  • $\begingroup$ Thanks; How do you prove that a $G_{\delta}$ set in a normal space is the zero level set of a non-negative continuous function? (I guess this is related to Urysohn's lemma somehow). $\endgroup$ – Asaf Shachar Nov 10 '18 at 7:00
  • $\begingroup$ here's a link (33.4): web.math.ku.dk/~moller/e02/3gt/opg/S33.pdf $\endgroup$ – Mariah Nov 11 '18 at 20:38
0
$\begingroup$

A notion that is a stronger version of this is called "stratifiable": a space $X$ is called stratifiable, when for every closed subset $F$ of $X$ we have a (WLOG decreasing) sequence of open subsets $U(F,n), n \in \mathbb{N}$ such that for all $F,G$ closed, if $F \subseteq G$ then for all $n$, $U(F,n) \subseteq U(G,n)$ and for all closed $F$ we have $F = \bigcap_{n \in \mathbb{N}} U(F,n)$. So every closed set is a $G_\delta$ in a monotonous way.

Metric spaces are a prime example of such spaces, as we can take $U(F,n) = \{x: d(x,F) < \frac{1}{n}\}$ in that case.

"Just" having every closed set a $G_\delta$ is called "perfect" and if the space is also $T_4$ it's then called $T_6$ or "perfectly normal". It implies that $X$ is hereditarily normal (every subspace is normal). Stratifiable also implies a strong form of normality, namely "monotonically normal".

$\endgroup$

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.