3
$\begingroup$

Given a compact metrisable topological space $X$, we write $\mathcal{N}(X)$ for the set of non-empty closed nowhere dense subsets of $X$, which is a Polish space under the topology induced by the Hausdorff distance.

Does there exist a compact metrisable topological space $X$ and a Borel probability measure $\nu$ on $\mathcal{N}(X)$ such that for all $p \in X$, $\,\nu(K : p \in K)>0$?

(If anyone has a reference for this, that would be particularly useful.)

Remark: Intuitively, I expect that the Baire category theorem will somehow imply that the answer is no. To prove that the answer is no, it would be sufficient to prove the following assertion:

Conjecture. Let $X$ be a compact metrisable topological space, let $Y$ be a Polish space, and let $G$ be a closed subset of $X \times Y$. Suppose that for every countable set $S \subset Y$ there is a dense set $D \subset X$ such that for every $p \in D$ and $y \in S$, $(p,y) \not\in G$. Then for every Borel probability measure $\nu$ on $Y$ there exists $p \in X$ such that $\,\nu(y \in Y : (p,y) \in G)=0$.

To see this: Suppose the conjecture is true. Let $X$ be a compact metrisable space, take $Y:=\mathcal{N}(X)$, and take $G:=\{(p,K): p \in K\}$. For every countable $S \subset \mathcal{N}(X)$, the Baire category theorem gives that $\,\bigcup S\,$ has empty interior; so set $D:=X \setminus \bigcup S$. Then $D$ is dense and for any $p \in D$ and $K \in S$, $p \not\in K$ and so $(p,K) \not\in G$. Hence for every probability measure $\nu$ on $\mathcal{N}(X)$ there exists $p \in X$ such that $\nu(K \in \mathcal{N}(X):p \in K)=0$.

$\endgroup$
5
  • 4
    $\begingroup$ Won't a random Cantor set of a.s. positive measure have this property? $\endgroup$ Dec 11, 2015 at 16:27
  • 4
    $\begingroup$ In particular, a random translate of a given positive measure Cantor set on the circle. $\endgroup$ Dec 11, 2015 at 16:38
  • $\begingroup$ Choosing $\nu$ to be the Dirac measure at $\{X\}$ gives a trivial "yes". What are the implicit assumptions on $\nu$? $\endgroup$
    – YCor
    Dec 28, 2015 at 13:27
  • $\begingroup$ @YCor: It is stated in the question that $\mathcal{N}(X)$ is the set of non-empty closed nowhere dense subsets of $X$. $\endgroup$ Dec 28, 2015 at 15:45
  • $\begingroup$ Thanks, I missed this one (it's just "empty interior" here). $\endgroup$
    – YCor
    Dec 28, 2015 at 16:11

1 Answer 1

1
$\begingroup$

Yes. As in the comments: take $X=\mathbb{S}^1$; and let $\nu$ be the law of the random set constructed by taking a positive-Lebesgue-measure Cantor set $K \subset \mathbb{S}^1$ and rotating $K$ through a random angle selected according to the uniform distribution.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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