5
$\begingroup$

Let $X$ and $Y$ be compact metric spaces and let $\mathcal B_X$ and $\mathcal B_Y$ be their respective Borel $\sigma$-algebras.

Let $\mu$ be a Borel probability measure on $X$ and let $\mathcal B^*_X$ be the $\mu$-completion of $\mathcal B_X$.

If $E$ belongs to the product $\sigma$-algebra $\mathcal B^*_X\otimes\mathcal B_Y$, does there exist a $\mu$-null set $A$ in $\mathcal B_X$ such that $E\cup(A\times Y)\in\mathcal B_X\otimes\mathcal B_Y$?

$\endgroup$
2
$\begingroup$

Let $\mathscr{C}$ be the collection of $E\in\mathcal{B}_X^*\otimes \mathcal{B}_Y$ for which:

  1. There exists $A\subseteq X$ null with $E\cup (A\times Y)\in\mathcal{B}_X\otimes\mathcal{B}_Y$.
  2. There exists $A\subseteq X$ null with $E\setminus (A\times Y)\in\mathcal{B}_X\otimes\mathcal{B}_Y$.

Then it suffices to show that $\mathscr{C}$ is a $\sigma$-algebra containing the rectangles $A\times B$, $A\in\mathcal{B}_X^*$, $B\in\mathcal{B}_Y$.

For rectangles, use the fact that we can approximate, both on the inside and the outside, elements of $\mathcal{B}_X^*$ by elements of $\mathcal{B}_X$.

For countable unions, property 1. is easy to deal with. For 2., suppose $E_n\in\mathscr{C}$, and for each $n$ choose $A_n$ satisfying $2$. Then $\bigcup_m A_m$ is null in $\mathcal{B}_X$ and $$\bigg(\bigcup_n E_n\bigg)\setminus\bigg(\bigg(\bigcup_m A_m\bigg)\times Y\bigg)=\bigcup_n\bigg(E_n\setminus \bigg(\bigcup_m \big(A_m\times Y\big)\bigg)\bigg)$$

For a given $n$, $E_n\setminus\big(\bigcup_m (A_m\times Y)\big)=\big(E_n\setminus(A_n\times Y)\big)\setminus\big(\bigcup_m(A_m\times Y)\big)$, which belongs to $\mathcal{B}_X\otimes\mathcal{B}_Y$, so their union also belongs to it.

For complements, given $E\in\mathscr{C}$, use property 1. for $E$ to obtain property 2. for $E^c$ and vice-versa, so $E^c\in\mathscr{C}$.

$\endgroup$
  • $\begingroup$ Thanks for this. Am I correct that it remains to show that $\mathcal C$ is closed under countable intersections? $\endgroup$ – mo15 May 22 '16 at 9:19
  • 1
    $\begingroup$ @mo15 This is not necessary, in general, since any class of sets which is closed under countable unions and complements is closed under countable intersections (because $\bigcap_n A_n=\left(\bigcup_n A_n^c\right)^c$), but you should mention this, if your definition of $\sigma$-algebra has this condition as well. $\endgroup$ – Luiz Cordeiro May 24 '16 at 6:59
  • $\begingroup$ You should probably add that the null sets $A $ in properties 1. and 2. should be Borel. Up to that: nice answer! $\endgroup$ – PhoemueX May 24 '16 at 7:10
  • $\begingroup$ @Luiz. I see. In this case the answer seems perfectly fine to me. $\endgroup$ – mo15 May 25 '16 at 14:39
  • $\begingroup$ @PhoemueX. I do not see why the null sets should be added to properties 1 and 2. $\endgroup$ – mo15 May 25 '16 at 14:40

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.