A local analysis textbook I have used has the following exercise:

Let $X$ be a finite-dimensional, $Y$ a separable Banach-space, $f\colon X\rightarrowtail Y$ any function. Show that $f'$ is Borel.

I have no idea how to approach this. I looked at the proof of Rademacher, but even that only gives Lebesgue-measurability as far as I can see, and this has no assumptions on the function at all. Any help would be appreciated.

  • $\begingroup$ Out of curiosity (I'm not sure if this statement is true or not): Have you actual need for this statement or is it out of interest that you ask? $\endgroup$ – t.b. May 11 '11 at 23:30
  • 2
    $\begingroup$ Can you do the case $X = Y = \mathbb R$ ? $\endgroup$ – GEdgar May 11 '11 at 23:59
  • $\begingroup$ I do not need it for an application now, it is more out of interest. $\endgroup$ – scineram May 12 '11 at 8:48
  • $\begingroup$ Even if $X=Y=\mathbb{R}$ I cannot approach this. I do not know what possible method could apply to functions like $id^2_\mathbb{R}\cdot\chi_A$ for any $A$. Should I bring this mathoverflow? $\endgroup$ – scineram May 12 '11 at 8:54
  • 1
    $\begingroup$ Define $f_n(x):=\frac{f(x + 1/n) - f(x)}{1/n}$... then... $\endgroup$ – Jonas Teuwen May 12 '11 at 9:22

It seems my professor whose book I originally met this problem in published the proof himself, generalizing the previous result of Federer for continuous functions. I was really surprised how much simpler it was than I expected.

In fact, I noticed the proof only requires that $L(X;Y)$ is separable, hence so are $X'$ and $Y$, since they embed into it. Given these, either of them being finite-dimensional is sufficient. It also raises the question whether $L(X;Y)$ can be separable when both spaces are infinite-dimensional. I read somewhere that this is an open problem when $X=Y$.


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.