0

Let $f:[a, b] \rightarrow \mathbb{R}$ and define $g:[a, b] \to \mathbb R$ by $$ g(x) := \begin{cases} f^{\prime}(x) & \text{if } f \text { Fréchet differentiable at } x, \\ 0 & \text {otherwise }. \end{cases} \quad \forall x \in (a, b), $$ and $g(a)=g(b)=0$.

I'm trying to make sense of integrating the derivative (defined a.e.) by proving below result.

Theorem If $f$ is Borel measurable, then so is $g$.

Could you have a check on my below attempt?


Proof Let $D := \{x \in (a, b) : f \text{ Fréchet differentiable at }x \}$. Then $D$ is a Borel subset of $\mathbb R$. This can be proved by extending $f$ to the whole $\mathbb R$ (for example, we set $f$ to be $0$ outside $[a, b]$), and then applying Section 3.5 of the book Fréchet Differentiability of Lipschitz Functions and Porous Sets in Banach Spaces by Lindenstrauss/Preiss/Tišer, i.e.,

"In this short section we show that for an arbitrary map between Banach spaces the set of its points of Fréchet differentiability is Borel; in fact it has type $F_{\sigma \delta}$.

Let $$ g_n(x) := \begin{cases} \frac{f\left(x+\frac{1}{n}\right)-f(x)}{\frac{1}{n}} & \text{if } x \in D, \\ 0 & \text{otherwise}. \end{cases} \quad \forall n\in \mathbb N, x \in [a, b]. $$

It follows from $f$ and $D$ are Borel measurable that $g_n$ is Borel measurable. Moreover, $g_n \to g$ everywhere. The everywhere limit of a sequence of Borel measurable functions is again Borel measurable. So $g$ is Borel measurable. This completes the proof.

Analyst
  • 5,637
  • Why is $D$ Borel measurable? (I know it is but you don't give an argument for that) – Mittens Dec 28 '22 at 15:33
  • @OliverDíaz I have just found a proof here. – Analyst Dec 28 '22 at 15:50
  • that proof is for $f$ continuous. For general $f$ (measurable or not) things are more subtle. – Mittens Dec 29 '22 at 02:23
  • @OliverDíaz It seems one of the answer works for Borel measurable function. – Analyst Dec 29 '22 at 10:06
  • 1
    Oh! I didn't not see that solution. The is a little problem which is easy to fix: $f(a)$ and $f(b)$ may not be defined, so to extend $f$ (defined on $(a,b)$) one can choose to set $f=0$ outside $(a, b)$ (after all one is only interesting in points in $(a, b)$. – Mittens Dec 29 '22 at 17:03

0 Answers0