Exercise 6 from "Measure Theory and Integration" Micheal E.Taylor, a maximal inequality.

70 Views Asked by At

I'm wrapping up Michael E. Taylor's Measure Theory book and currently tackling the exercises in Chapter 11. I'm encountering difficulties with Exercise 6, specifically at a certain point where I'm stuck. I have at my disposal the standard covering lemmas and classical maximal inequalities.

Let $\Phi: (0, \infty) \rightarrow \mathbb{R}^+$ be a monotone decreasing $C^1$ function with the property that $-s^{n} \cdot \Phi'(s) \in L^1(\mathbb{R}^+)$. Then define $$\varphi_{r}(x) = r^{-n}\cdot \Phi(\frac{\vert x \vert}{r}),$$ and finally set $$M_{\varphi}f(x)= \sup_{r>0} \int_{\mathbb{R}^{n}} \left\vert f(y) \right\vert \cdot \phi_{r}(x-y) \ dy.$$ Now prove that there exists $g \in L^{1}(\mathbb{R}^{+})$ such that $$ \left\vert M_{\varphi}f(x) \right\vert \leq \Vert g \Vert_{L^{1}} \cdot Mf(x)$$ for every measurable $f.$

In my attempt, I proceeded as follows: $$\begin{align*} \int_{\mathbb{R}^{n}} \vert f(y) \vert \cdot r^{-n} \cdot \Phi(\frac{\vert x -y \vert}{r}) \ dy = \int_{\mathbb{R}^{n}} \vert f(x-r\ s) \vert \cdot \Phi(\vert s \vert) \ ds \\ = \int_{0}^{+\infty} \ d\rho \ \int_{S^{n-1}} \rho^{n-1} \ \Phi(\rho) \ \vert f(x-r \ \rho \ \omega) \vert \ d\omega \\ = \int_{0}^{+\infty} \rho^{n-1} \ \Phi(\rho) \ \int_{S^{n-1}} \vert f(x-r \ \rho \ \omega) \vert \ d\omega \ d \rho \\ \int_{0}^{+\infty} \Phi(\rho)\ \frac{d}{d\rho}\left( \int_{0}^{\rho}s^{n-1} \ ds \ \int_{S^{n-1}} \vert f(x-r \ s \omega) \vert \ d\omega \right) \ d \rho \\ = \int_{0}^{+\infty} \Phi(\rho)\ \frac{1}{r^{n}} \frac{d}{d\rho} \left( \int_{0}^{r \ \rho}s^{n-1} \ ds \ \int_{S^{n-1}} \vert f(x- \ s \omega) \vert \ d\omega \right) \ d \rho \\ = \int_{0}^{+\infty} \Phi(\rho)\ \frac{1}{r^{n}} \frac{d}{d\rho} \left( \int_{B_{r\ \rho}(x)} \vert f(y) \vert \ d\omega \right) \ d \rho \\ \text{now there is an integration by parts I cannot fully justify}\\ = \int_{0}^{+\infty} -\Phi'(\rho)\ \frac{1}{r^{n}} \left( \int_{B_{r\ \rho}(x)} \vert f(y) \vert \ d\omega \right) \ d \rho. \end{align*}$$ If I pass to the supremum on the last and first term of my chain of inequalities, I obtain $$ \begin{align*} \left\vert M_{\varphi}f(x) \right\vert \leq \sup_{r>0} \int_{\mathbb{R}^{n}} \vert f(y) \vert \cdot r^{-n} \cdot \Phi(\frac{\vert x -y \vert}{r}) \\ \leq \sup_{r>0} \int_{0}^{+\infty} -\Phi'(\rho)\ \rho^{n} \ \frac{1}{(\rho \ r)^{n}} \left( \int_{B_{r\ \rho}(x)} \vert f(y) \vert \ d\omega \right) \ d \rho \leq \int_{0}^{+\infty} -\Phi'(\rho)\ \rho^{n} \ \sup_{r>0} \left( \frac{1}{(\rho \ r)^{n}} \left( \int_{B_{r\ \rho}(x)} \vert f(y) \vert \ d\omega \right) \right) \ d\rho = Mf(x) \cdot \int_{0}^{+\infty} -\Phi'(\rho)\ \rho^{n} \ d \rho. \end{align*}$$ And we conclude by setting $g(\rho) := -\Phi'(\rho) \ \rho^{n}.$

The only passage not fully justified is the integration by parts. Why can I ignore the boundary term coming from the integration by parts? Perhaps, I can prove the inequality for a compactly supported $\Phi$ and arrive at a general $\Phi$ via approximation.