3
$\begingroup$

Suppose that $A\subseteq R^{m}$, let $B(R^{m})$ be the set of all Borel subsets of $R^{m}$. We say that $B\subseteq R^{m}$ is a Borel envelope of $A$ if $B\in B(R^{m})$ and for every $L^{m}$ measurable set (here $L^{m}$ means Lebesgue outer measure) $F$ in $R^{m}$ one has $L^{m}(A\cap F)=L^{m}(B\cap F)$. Then I want to show the following: Let $B\in B(R^{m})$ and $A\subseteqq B$. Prove that $B$ is a Borel envelope of A iff $L^{m}(S)=0$ whenever $S$ is an $L^{m}$-measurable subset of $B \setminus A$. I want to use the fact that there is a Borel set $A'$ such that $L^{m}(A)=L^{m}(A')$ but I don't know what to do next.

$\endgroup$
1
  • 1
    $\begingroup$ When cross-posting a question from MSE, it is highly recommended to give the link to the corresponding MSE question (math.stackexchange.com/questions/4316103) and also back refer this question at the MSE question. Furthermore, usually one first waits several days to see whether an answer will appear in the MSE. $\endgroup$
    – Z. M
    Nov 25, 2021 at 22:30

1 Answer 1

2
$\begingroup$

$\newcommand{\de}{\delta}\newcommand\R{\mathbb R}\newcommand{\Z}{\mathbb{Z}}$To simplify the writing, let us first work locally, say with subsets of $(0,1]^m$, instead of $\R^m$, so that to avoid infinite values of $L^m$. (Going back to $\R^m$ is then straightforward: using Carathéodory's criterion, for any set $A\subseteq\R^m$ we "collect" its "finite pieces" by the formula $L^m(A)=\sum_{z\in\Z^m}L^m(A\cap(z+(0,1]^m))$.)

Then it is easy to show (see Lemma 1 at the end of this answer) that the definition of a Borel envelope simplifies as follows: a Borel set $B$ is a Borel envelope of $A$ if $B\supseteq A$ and $L^{m}(A)=L^{m}(B)$.

We want to show the following: Suppose that $B$ is Borel and $A\subseteq B$; then $B$ is a Borel envelope of $A$ iff $L^{m}(S)=0$ whenever $S$ is an $L^{m}$-measurable subset of $B\setminus A$.

The only-if part: Suppose that $B$ is a Borel envelope of $A$. To obtain a contradiction, suppose that $L^{m}(S)>0$ for some $L^{m}$-measurable $S\subseteq B\setminus A$. Then $S\subseteq B$ and for the $L^{m}$-measurable set $S_1:=B\setminus S$ we have $A\subseteq S_1$ and hence $L^{m}(A)\le L^{m}(S_1)=L^{m}(B)-L^{m}(S)<L^{m}(B)$, which contradicts the condition $L^{m}(A)=L^{m}(B)$, as desired.

The if part: Suppose that $L^{m}(S)=0$ whenever $S$ is an $L^{m}$-measurable subset of $B\setminus A$. To obtain a contradiction, suppose that $B$ is not a Borel envelope of $A$ -- that is, $L^{m}(A)<L^{m}(B)$. Note that $L^{m}(A)=\inf\{L^{m}(C)\colon C\supseteq A, C\text{ Borel}\}$. So, $L^{m}(B)>L^{m}(C)$ for some Borel set $C$ such that $C\supseteq A$, whence $L^{m}(B)> L^{m}(C\cap B)$ and $L^{m}(B\setminus C)=L^{m}(B)-L^{m}(C\cap B)>0$, whereas the set $S:=B\setminus C$ is $L^m$-measurable and contained in $B\setminus A$ -- a desired contradiction.

To complete the proof, we only to need to state and prove

Lemma 1: A Borel set $B\subseteq(0,1]^m$ is a Borel envelope of a set $A\subseteq(0,1]^m$ iff $B\supseteq A$ and $L^{m}(A)=L^{m}(B)$.

Proof of Lemma 1: The only-if part of Lemma 1 is trivial: just take, for instance, $F=(0,1]^m$ or $F=\R^m$.

Let us now consider the if part of Lemma 1. Here we suppose that $A\subseteq B\subseteq(0,1]^m$, $B$ is Borel, and $L^{m}(A)=L^{m}(B)$. To obtain a contradiction, suppose that $B$ is not a Borel envelope of $A$. Then $L^{m}(A\cap F)\ne L^{m}(B\cap F)$ and hence $L^{m}(A\cap F)<L^{m}(B\cap F)$ for some $L^{m}$-measurable set $F\subseteq\R^m$. Also, obviously, $L^{m}(A\cap F^c)\le L^{m}(B\cap F^c)$, where $F^c:=\R^m\setminus F$. Adding the latter two inequalities and using Carathéodory's criterion again (together with the upper bound $L^{m}(A)\le L^{m}((0,1]^m)\le1<\infty$), we get $L^{m}(A)=L^{m}(A\cap F)+L^{m}(A\cap F^c)<L^{m}(B\cap F)+L^{m}(B\cap F^c) =L^{m}(B)$, which does contradict the condition $L^{m}(A)=L^{m}(B)$.

This completes the proof of Lemma 1 and thus the entire proof of the desired characterization of the Borel envelope.

$\endgroup$

Your Answer

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

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