2

Let $(X, \Sigma)$ be a measurable space and $f:X\to X$ be measurable bijection. When can we say that $f^{-1}$ is measurable? I'm aware that full answer may be hard to produce, so anything more than I posted will do.


I was assigned task in which a measurable and invertible function appeared, but it didn't specify if it was invertible in the category of sets, or measurable spaces. It got me wandering, when those are the same.


I know that if $\Sigma$ was family of Borel sets of topology, in which $f$ was homeorphism, then it would be true. But this is quite strong condition, as there are $\sigma$-algebras which are not generated by any topology. Also even when we assume, that $\Sigma$ was generated by some topology, finding one in which $f$ was homeomorphism seems to be non-trivial task (and perhaps not always doable as well).

Broader approach (from which first one emerged) may be to say, that $f^{-1}$ is measurable with respect to $\sigma$-algebra generated by $f$: $$ A\in \Sigma \implies f(B) \in \Sigma $$ Where $B = f^{-1}(A)$ is general form of set in $\sigma(f)$. Now we have, that if $\sigma(f) = \Sigma$, then $f^{-1}$ is measurable. But I don't know, how to easily check if a function generates whole $\Sigma$.

Esgeriath
  • 2,181

1 Answers1

1

If $(X, \mathcal X)$ and $(Y, \mathcal Y)$ are standard Borel spaces, and if $f : X \to Y$ is bijective and Borel measurable, then $f^{-1}$ is also Borel measurable.

This is an immediate consequence of corollary 15.2 on p.89 of "Classical Descriptive Set Theory" by Alexander Kechris (Springer, 1995).

Alex M.
  • 35,207
  • That's a nice result! Back when I asked this question I was dealing with standard probability space though, which more often than not aren't standard Borel spaces. But this definitely answers the question as stated. – Esgeriath Jan 11 '24 at 16:01