2

Let $h: \Bbb{R}\to \Bbb{R}$ be a Lebesgue measurable function with $\int |h| < \infty$ and define $$h^*: \Bbb{R}\to [0,\infty]: b \mapsto \sup_{t > 0} \frac{1}{2t}\int_{b-2t}^{b+2t}|h|$$

Show that $h^*$ is Borel-measurable.

Attempt: We show $\{x\in \Bbb{R}: h^*(x) > c \}$ is an open subset of $\Bbb{R}$ for all $c\in \Bbb{R}$.

If $h^*(b) > c$, then there is $\epsilon > 0$ such that $$\frac{1}{2\epsilon}\int_{b-\epsilon}^{b+\epsilon}|h| > c$$ Choose $\delta > 0$ such that $$\frac{1}{2(\epsilon+\delta)}\int_{b-\epsilon}^{b+\epsilon}|h| > c$$

If $|z-b|< \delta$, then $[z-(\epsilon+\delta), z+(\epsilon+\delta)] \supseteq [b-\epsilon, b+\epsilon]$ so $$h^*(z) \geq \frac{1}{2(\epsilon+\delta)}\int_{z-(\epsilon+\delta)}^{z+\epsilon+\delta}|h| \geq \frac{1}{2(\epsilon+\delta)}\int_{ b-\epsilon}^{b+\epsilon}|h| > c$$ so $b$ is an interior point of $\{h^* > c\}$, as desired.

Is the above correct?

1 Answers1

2

From my pov your proof is ok and straight forward. Depending of what you already know $h^*$ is just a composition of measurable functions hence measurable itself. To see this consider the following measurable functions:

$$f(b) = \int_{a}^b |h| d\lambda$$ is absolutely continuous hence measurable for each $a \in \Bbb R$ by the general fundamental theorem of analysis.

For measurable $g_t$ the supremum $$\sup_{t > 0} g_t(b)$$ is measurable as well and with $$g_t(b) = \frac{f(b+2t) - f(b-2t)}{2t}$$ we have that all $g_t(b)$ are measurable being the composition of measurable functions and we get $$h^*(b) = \sup_{t>0}g_t(b)$$ so all together that $h^*$ is measurable.

Gono
  • 5,598
  • Thanks! This is definitely an approach I should keep in mind for similar problems. –  Aug 14 '20 at 11:13
  • 2
    @user745578 Definitely do so. But do not forget your approach over it. If you consider the analogous function $$h^{\ast}(x) = \sup_{r > 0} \frac{1}{\lambda(B_r)} \int_{B_r(x)} \lvert h(t)\rvert,d\lambda$$ when the domain of $h$ is $\mathbb{R}^n$ with $n > 1$, then your approach needs only notational modifications, while the modification of the approach given by Gono is not as straightforward. Generally, for some problems both approaches work (more or less) equally well, for some one approach works significantly better than the other, and for some neither approach works well. – Daniel Fischer Aug 14 '20 at 12:07