Atomic decomposition of Hardy spaces and characterization of \(BMO\) via Banach function spaces (Q2392032): Difference between revisions

The author investigates what happens when replacing one of the usual \(L^p\) spaces in the definition of \(BMO\) by a Banach function space \(X\). The condition for a function to belong to \(BMO\) is \[ \left(\frac1{| B| }\int_B | f(x) - f_B| ^p \, dx\right)^{1/p} \leq C, \] with \(C\) independent of \(B\), and the space is independent of \(p\). This is replaced by saying that \(f \in BMO_X\) if \[ \frac{\| f - f_B\| _X}{\| \chi_B\| _X} \leq C \] with \(C\) independent of \(B\). The associate space \(X^{\prime}\) plays the role for Banach function spaces that duals do for usual Banach spaces. The author also shows that, if the Hardy-Littlewood maximal function is bounded on \(X^{\prime}\), then \(BMO_X = BMO\). The author shows that, if instead the Hardy-Littlewood maximal function is bounded on \(X\), the associated Hardy space has an atomic decomposition with \(X\) atoms, and uses that result to give conditions on a variable \(p(x)\) in order that the space \(L^{p(x)}\) gives rise to the usual \(BMO\) when \(X = L^{p(\cdot)}\). In particular, when \(\mathrm{ess} \inf p(x) = p_{-} >1\), \(\mathrm{ess} \sup p(x) = p_{+} < \infty\), then \(L^{p(\cdot)}\) has the requisite property, and the associated Hardy space has an atomic decomposition and its dual space is the usual (non-variable) \(BMO\) space.