Extremal problems related to maximal dyadic-like operators (Q982605)

The dyadic maximal operator on \(\mathbb{R}^{n}\) is defined by \[ \mathcal{M}_{d}\phi (x)=\sup \left\{ \frac{1}{\left| Q\right| } \int_{Q}\left| \phi (u)\right| du:x\in Q, Q\subseteq \mathbb{R}^{n} \text{ is a dyadic cube}\right\} \] for every \(\phi \in L_{loc}^{1}(\mathbb{R}^{n}),\) where the dyadic cubes are those formed by the grids \(2^{-N}\mathbb{Z}^{n}\) for \(N=0,1,2,\ldots \). It is known that \[ \left\| \mathcal{M}_{d}\phi \right\| _{p}\leq \frac{p}{p-1}\left\| \phi \right\| _{p}.\tag{1} \] One way of studying such maximal operators is the computation of the so-called Bellman functions related to them, which reflect certain deeper properties of these maximal operators. Such functions related to inequality (1) were precisely evaluated by \textit{A. D. Melas} [Adv. Math. 192, No. 2, 310--340 (2005; Zbl 1084.42016)] and \textit{A. D. Melas} and the author [Trans. Am. Math. Soc. 362, No. 3, 1571--1597 (2010; Zbl 1194.42026)]. In fact, if we define for any \(p>1,\) \[ B_{p}(f,F)=\sup \left\{ \frac{1}{\left| Q\right| }\int_{Q}\left( \mathcal{M}_{d}\phi \right) ^{p}du:\frac{1}{\left| Q\right| } \int_{Q}\phi =f,\frac{1}{\left| Q\right| }\int_{Q}\phi ^{p}=F\right\}, \tag{2} \] where \(Q\) is a fixed dyadic cube, \(\phi \) is a non-negative in \(L^{p}(Q)\) and \(F,f\) satisfy \(0\leq f\leq F^{1/p}.\) It is proved by \textit{A. D. Melas} [loc. cit.] that \( B_{p}(f,F)=F\omega _{p}(\frac{f^{p}}{F}),\) where \(\omega _{p}:[0,1]\rightarrow [ 1,\frac{p}{p-1}]\) is the inverse of the function \(H_{p}(z)=-(p-1)z^{p}+pz^{p-1}.\) Using this result, it is possible to compute more complicated functions related to maximal dyadic-like operators. One may think of (2) as an extremum problem which reflects the deeper structures of dyadic maximal operators. Certain other extremum problems arise in this spirit. The main purpose of the paper under review is to study some extremum problems related to maximal dyadic-like operators. More precisely, the main purpose is to compute the following: \[ B(f,F)=\sup \left\{ \left\| \mathcal{M}_{\mathcal{T}}\phi \right\| _{p,\infty }:\phi \geq 0,\int_{X}\phi d\mu =f,\left\| \phi \right\| _{p,\infty }=F\right\} ,0<f\leq F, \] and \[ B_{1}(f,F)=\sup \left\{ \left\| \mathcal{M}_{\mathcal{T}}\phi \right\| _{p,\infty }:\phi \geq 0,\int_{X}\phi d\mu =f,\left\| \phi \right\| _{p,\infty }=F\right\} ,0<f\leq \frac{p}{p-1}F, \] where \((X,\mu )\) is a non-atomic probability measure space, \(\mathcal{T}\) is a tree on \(X,\left\| \phi \right\| _{p,\infty }=\sup \{\lambda \mu (\{\phi >\lambda \})^{1/p}:\lambda >0\}\) is the standard quasi-norm in \( L^{p,\infty }\) and \(\mathcal{M}_{\mathcal{T}}\) is the\ maximal operator associated to the tree \(\mathcal{T}\) defined by \[ \mathcal{M}_{_{\mathcal{T}}}\phi (x)=\sup \left\{ \frac{1}{\mu (I)} \int_{I}\left| \phi \right| d\mu :x\in I\subseteq \mathcal{T} \right\} \] for every \(x\in X\) and \(\phi \) a function on \(X\) with \(\phi \in L^{1}(X,\mu )\). The main results of this paper are the following: \textbf{(1) }\(B(f,F)=F\) if \(0<f\leq F\) and \textbf{(2) }\(B_{1}(f,F)=\frac{p}{p-1}F\) if \( 0<f\leq \frac{p}{p-1}F.\) We point out the computation of \(B(f,F)\) and \(B_{1}(f,F)\) were obtained respectively as an easy task from the computations of \(B(f,F,\lambda )\) and \( B_{1}(f,F,\lambda ).\) In fact, the author of the paper proved that \( B(f,F,\lambda )=\min (1,f/\lambda ,F^{p}/\lambda ^{p})\) and \( B_{1}(f,F,\lambda )=\min (1,f/\lambda ,\left( \frac{p}{p-1}\right) ^{p}F^{p}/\lambda ^{p}),\) where \[ B(f,F,\lambda )=\sup \left\{ \mu (\{\mathcal{M}_{_{\mathcal{T}}}\phi \geq \lambda \}):\phi \geq 0,\int_{X}\phi d\mu =f,\left\| \phi \right\| _{p,\infty }\leq F\right\} ,0<f\leq F,\lambda >0 \] and \[ \begin{multlined} B_{1}(f,F,\lambda )=\sup \left\{ \mu (\{\mathcal{M}_{_{\mathcal{T}}}\phi \geq \lambda \}):\phi \geq 0,\int_{X}\phi d\mu =f,\left\| \phi \right\| _{p,\infty }\leq F\right\} ,\\ 0<f\leq \frac{p}{p-1}F,\lambda >0. \end{multlined} \]