Borel Canonization of Analytic Sets with Borel Sections

Abstract

Given an analytic equivalence relation, we tend to wonder whether it is Borel. When it is non Borel, there is always the hope it will be Borel on a "large" set -- nonmeager or of positive measure. That has led Kanovei, Sabok and Zapletal to ask whether every proper $\sigma$ ideal satisfies the following property: given $E$ an analytic equivalence relation with Borel classes, there exists a set $B$ which is Borel and $I$-positive such that $E{\upharpoonright }_B$ is Borel. We propose a related problem -- does every proper $\sigma$ ideal satisfy: given $A$ an analytic subset of the plane with Borel sections, there exists a set $B$ which is Borel and $I$-positive such that $A\cap (B\times {\omega }^{\omega })$ is Borel. We answer positively when a measurable cardinal exists, and negatively in $L$, where no proper $\sigma$ ideal has that property. Assuming ${\omega }_1$ is inaccessible to the reals but not Mahlo in $L$, we construct a ccc $\sigma$ ideal $I$ not having this property -- in fact, forcing with $I$ adds a non Borel section to a certain analytic set with Borel sections, and a non Borel class to a certain analytic equivalence relation with Borel classes. Various counterexamples are given for the case of a ${\mathbf{\Delta }}_2^1$ equivalence relation as well as for the case of an improper ideal.