It is known that a measurable bijection $f \colon [0,1] \to [0,1]$ has a measurable inverse. (Here, all measurability is simply with respect to the Borel $\sigma$-algebra of $[0,1]$.)

Now fix an arbitrary measurable space $(\Omega,\mathcal{F})$, and let $(f_\omega)_{\omega \in \Omega}$ be a family of bijections $f_\omega \colon [0,1] \to [0,1]$ such that the map $(\omega,x) \mapsto f_\omega(x)$ is $(\mathcal{F}\otimes\mathcal{B}([0,1]),\mathcal{B}([0,1]))$-measurable. Is it necessarily the case that the map $(\omega,x) \mapsto f_\omega^{-1}(x)$ is $(\mathcal{F}\otimes\mathcal{B}([0,1]),\mathcal{B}([0,1]))$-measurable?

  • 3
    $\begingroup$ It's certainly true if $(\Omega, \mathcal{F})$ is standard Borel. Not sure if that's of any use to you. $\endgroup$ Oct 16, 2019 at 4:05
  • 2
    $\begingroup$ Thank you; yes, for that case it should presumably follow from the non-random case by considering the map $(\omega,x) \mapsto (\omega,f_\omega(x))$. But I really want not to assume that $(\Omega,\mathcal{F})$ is standard. $\endgroup$ Oct 16, 2019 at 11:35
  • 1
    $\begingroup$ If this is true, I would try to reconstruct the proof that injective Borel images of Borel sets are Borel (which involves the Lusin Separation Thm), using that the $f_\omega$ are “uniformly measurable” (i.e., the complexity of $f_\omega^{-1}(X)$ with varying $\omega$ is bounded because of the measurability of $(\omega,x) \mapsto f_\omega(x)$). But it's just a hunch. $\endgroup$ Oct 31, 2019 at 13:47


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy