I've been looking through some posts about how projections of sets from product $\sigma$-algebras of two Borel $\sigma$-algebras onto their component spaces are not necessarily measurable which apparently leads into the rabbit hole of descriptive set theory.

My problem, though, is a little different and I thought perhaps someone had an idea of how to tackle it seeing as how I am not really knowledgeable when it comes to descriptive set theory.

What I have is a kind of Polish space $\mathcal{X}$ (complete, separable metric space), a countable (or even finite) set $\mathcal{A}$ (being polish by virtue of its discrete metric if I am not mistaken) and their respective Borel $\sigma$-algebras $\mathcal{B}({\mathcal{X}})$ and $\mathcal{B}(\mathcal{A})=2^{\mathcal{A}}$ since $\mathcal{A}$ is countable.

What I am wondering now is, if I look at some set from their product $S \in \mathcal{B}({\mathcal{X}}) \otimes \mathcal{B}(\mathcal{A})$, does the setup now give us that $$ \pi_{\mathcal{X}}(S) := \{ x \in \mathcal{X} | \, \exists\, a \in \mathcal{A}: \, (x,a) \in S \} $$

What I have tried to use is that since $\mathcal{B}(\mathcal{A})=2^{\mathcal{A}}$ and countable, we can write $S$ as

$$ S = \bigcup_{a \in \mathcal{A}} S_a \times \{a \} $$ where $ S_a = \{ x \in \mathcal{X}|\, (x,a) \in S \} $ hoping I could somehow conclude that $ S_a \in \mathcal{B}(\mathcal{X}) $ as intuitively, pairing Borel sets with countably (or even finitely) many point sets and using that as a generator for $ \mathcal{B}({\mathcal{X}}) \otimes \mathcal{B}(\mathcal{A}) $ shouldn't all of a sudden provide us with sets whose projections are non-Borel.

However, I cannot seem to properly prove this statement and perhaps it is false to begin with. I thought someone here could give some insights.



Your argument goes through. The set $S_a = \{ x \in \mathcal{X}|\, (x,a) \in S \}$ is known as the $a$-section, and it is generally measurable for the product $\sigma$-algebra (the smallest $\sigma$-algebra generated by rectangles).

The proof that sections are measurable is a crucial step in the proof of Fubini's theorem, and that is where you can probably find a proof in most textbooks that treat measure theory. You can also find a short proof here.

  • $\begingroup$ Right, I totally overlooked that! Thanks so much for your help :) $\endgroup$ – Sunshower Jul 2 '17 at 11:07

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.