7
$\begingroup$

This may be a naive question, but I don't see an immediate argument.

Question: Does there exist a sequence $\{C_m\}_{m=1}^\infty$ of Borel subsets of $[0,1]$ with positive Lebesgue measure $|C_m|>0$, such that every closed subset $C\subset[0,1]$ with positive Lebesgue measure $|C|>0$ contains at least one $C_m\subset C$?

It feels too strong to be true, but I need to know for sure. I was thinking of patterns like dyadic partitions of fat Cantor sets etc., but can't arrive at something useful.

Thank you.

$\endgroup$
2
  • $\begingroup$ may I ask why are you calling this property "representing"? $\endgroup$
    – erz
    Commented Dec 31, 2023 at 20:57
  • $\begingroup$ In the sense that a set of positive measure is "represented" if a positive measure subset of it is in the collection. Nothing deep. $\endgroup$
    – Bedovlat
    Commented Jan 9 at 18:23

2 Answers 2

11
$\begingroup$

Without the assumption that $C$ be closed the answer is no.

Indeed, suppose such $C_m$'s exist. Then without loss of generality $|C_m|\le1/3^m$ for all $m$. Let now $C:=\bigcap_{m=1}^\infty ([0,1]\setminus C_m)=[0,1]\setminus\bigcup_{m=1}^\infty C_m$.

Then $|C|\ge1/2>0$, but $C_m\not\subseteq C$ for any $m$.


Using the hint by Aleksei Kulikov (why didn't I think of that? :-)), one can modify the above construction as follows, to get an unqualified no:

Suppose such $C_m$'s exist. Then without loss of generality $|C_m|\le1/3^m$ for all $m$. Let now $B:=\bigcap_{m=1}^\infty ([0,1]\setminus C_m)=[0,1]\setminus\bigcup_{m=1}^\infty C_m$.

Then $|B|\ge1/2>0$ and $C_m\not\subseteq B$ for any $m$. Let finally $C$ be a closed subset of $B$ with $|C|>0$; such a set $C$ exists by the regularity of the Lebesgue measure. Then $C_m\not\subseteq C$ for any $m$. $\quad\Box$

$\endgroup$
6
  • 3
    $\begingroup$ Doesn't any set of positive measure contain a closed subset of positive measure, so the first sentence about closeness is irrelevant, no? So the answer is still no. $\endgroup$ Commented Dec 31, 2023 at 15:47
  • 2
    $\begingroup$ Even more elementary: Pick a point $x_m$ from each $C_m$ and enclose it by an open interval of length $3^{-m}$. Take the union of these intervals, and then the relative complement. $\endgroup$ Commented Dec 31, 2023 at 16:02
  • 1
    $\begingroup$ @AlekseiKulikov : Thank you for your comment. (It seems that my previous thank-you comments were deleted. I am trying again.) $\endgroup$ Commented Dec 31, 2023 at 17:09
  • 3
    $\begingroup$ @MichaelGreinecker : Thank you for your comment. (It seems that my previous thank-you comments were deleted. I am trying again.) $\endgroup$ Commented Dec 31, 2023 at 17:09
  • 1
    $\begingroup$ Are expressions of gratitude and acknowledgments of helpful efforts by other users considered bad manners here? $\endgroup$ Commented Dec 31, 2023 at 17:12
14
$\begingroup$

Suppose $\{C_m\}_{m=1}^\infty$ is a sequence of nonempty subsets of $[0,1]$. Choose $x_m\in C_m$ and let $X=\{x_m:m\in\mathbb N\}$. The set $X$, being countable, has measure zero, so $[0,1]\setminus X$ has positive measure and therefore contains a closed set $C$ of positive measure. Then $C_m\not\subseteq C$ since $x_m\notin C$.

$\endgroup$
1
  • $\begingroup$ Great observation. It is a pity I can't accept two answers. $\endgroup$
    – Bedovlat
    Commented Jan 9 at 18:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .