Let $(X,S)$ be a measurable space and $f: X \to \mathbb{R}$ be $S-$measurable. Show that if $\psi:X \to \mathbb{R}$ is any $S_{0}-$measurable function, then there exists a Borel measurable function $\phi: \mathbb{R} \to \mathbb{R}$ such that $\psi=\phi \circ f$.

My try:


If $\psi$ is a simple $S_0-$measurable function , then $\psi=\sum_{i=1}^{n}a_n\chi_{f^{-1}(E)}$ for some positive integer $n$, $a_i \in \mathbb{R}$ for each $i$ and $E_i \in B_{\mathbb{R}}$. Thus $\psi=(\sum_{i=1}^na_i\chi_{E_{i}})\circ f$.


If $\psi$ is a non-negative $S_0-$measurable function, then there exists a increasing sequence of functions ${S_n}$ (simple measurable) such that $\psi(x)=\lim_{n \to \infty}S_{n}(x) \forall x \in X$. Since each $S_n(x)$ is a simple $S_0-$measurable function there exists a borel measurable function $\phi_n : \mathbb{R} \to \mathbb{R}$ such that $S_n=\phi_{n} \circ f$. This is where I am stuck. I am not sure if the limit of $\phi_n$ will exist or not and if it does the whether it is borel measurable or not. Moreover if the limit is $\phi(x)$, then can I conclude that $\psi(x)=\phi \circ f (x)$??

This will lead to the final conclusion as for any $\psi S_0-$measurable function $\psi^{+}$ and $\psi^{-}$ are non-negative $S_0$ measurable and I can conclude from the case-2.

Thanks for the help!!

  • $\begingroup$ Recently came about the same problem. I had the very same progress and question haha. Anyway, are you sure about the expression of $ψ$ in case one? $\endgroup$ – Marko Karbevski Oct 19 '15 at 19:34

If $\psi$ is a non-negative $S_0$-measurable function, choose a monotone sequence $(\psi_n)$ of simple $S_0$-measurable functions with $\psi_n \to \psi$. By case 1, there are Borel functions $\phi_n$ with $\psi_n = \phi_n \circ f$ for each $n$. On the range of $f$, $(\phi_n)$ is monotone, as for $y \in f(X)$ there is some $x \in X$ with $y = f(x)$, giving $$ \phi_n(y) = \phi_n f(x) = \psi_n(x) \le \psi_{n+1}(x) = \phi_{n+1}(y) $$ By changing $\phi_n$ outside $f(X)$ if necessary, e. g. replace $\phi_n$ with $\phi_n \cdot 1_{f[X]}$, we can assume that $(\phi_n)$ is monotone. Hence, there is some $\phi \colon X \to [0,\infty]$ with $\phi_n \to \phi$. As a limit of $S_0$-measurable functions, $\phi$ is $S_0$-measurable, moreover for $x \in X$: $$ \psi(x) = \lim_n \psi_n(x) = \lim_n \phi_n\bigl(f(x)\bigr) = \phi\bigl( f(x)\bigr) = (\phi \circ f)(x) $$

  • $\begingroup$ Don't you think we should compose rather than multiply? $\endgroup$ – tattwamasi amrutam Apr 2 '15 at 8:55
  • $\begingroup$ By $\phi_n f$ I mean the composition, but I can add the $\circ$, if you want ... $\endgroup$ – martini Apr 2 '15 at 9:00
  • $\begingroup$ Done :-) ${}{}$ $\endgroup$ – martini Apr 2 '15 at 9:00

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.