Mixed \(A_p\)-\(A_r\) inequalities for classical singular integrals and Littlewood-Paley operators (Q363216)

Establishing sharp bounds on the operator norm of singular integral operators (and their variants) acting on weighted \(L^p\) spaces is an important question in harmonic analysis. From a theoretic point of view, such estimates require a deep understanding of the nature of singular integrals (as averages of dyadic shifts, for instance). From the point of view of applications, sharp estimates are useful, for instance, in quasiconformal mappings theory, where they are linked to the regularity of solutions for certain partial differential equations (see the book [\textit{K. Astala, T. Iwaniec} and \textit{G. Martin}, Elliptic partial differential equations and quasiconformal mappings in the plane. Princeton Mathematical Series 48. Princeton, NJ: Princeton University Press (2009; Zbl 1182.30001)]). The most famous problem on this topic - the \(A_{2}\) conjecture - has recently been solved by \textit{T. Hytönen} [Ann. Math. (2) 175, No. 3, 1473--1506 (2012; Zbl 1250.42036)]). His result shows that \[ \begin{multlined} \|T\|_{B(L^{2}(\mathbb{R}^{n},w(x)dx))} \lesssim [w]_{A_{2}} \simeq \sup_{Q \; \text{dyadic cube}} A_{2}(w,Q) :=\\ \sup_{Q \; \text{dyadic cube}} (\frac{1}{|Q|}\int \limits _{Q} w(x)dx)(\frac{1}{|Q|}\int \limits _{Q} w(x)^{-1}dx),\end{multlined} \] for all Calderón-Zygmund operators \(T\). Given that this linear dependency on the \(A_{2}\) constant is optimal, one is tempted to believe that this result closes a beautiful chapter of modern harmonic analysis. However, new questions and results are rapidly emerging from the impressive set of methods developed by many authors in order to study the \(A_2\) conjecture. \textit{T. Hytönen} and \textit{C. Pérez}, for instance, showed in [Anal. PDE 6, No. 4, 777--818 (2013; Zbl 1283.42032)] that the \(A_{2}\) condition can be split, leading to \(A_{2}-A_{\infty}\) bounds of the form: \[ \|T\|_{B(L^{2}(\mathbb{R}^{n},w(x)dx))} \lesssim [w]_{A_{2}}^{\frac{1}{2}}\max([w]'_{A_{\infty}},[w^{-1}]'_{A_{\infty}})^{\frac{1}{2}}, \] for all Calderón-Zygmund operators \(T\). An interesting feature of these mixed bounds is that they do not extrapolate as easily as the \(A_{2}\) constant bounds appearing in the \(A_{2}\) conjecture (which give sharp \(L^p\) bounds through Rubio de Francia's extrapolation). Finding good \(L^p\) analogues has thus become the subject of extensive investigations. The paper under review is an important step in this direction. Building on his celebrated local mean oscillation formula from [Bull. Lond. Math. Soc. 42, No. 5, 843--856 (2010; Zbl 1203.42023)], the author proves that, for all \(p \geq 2\) and all \(r\geq p\), \[ \|T^{*}\|_{B(L^{p}(\mathbb{R}^{n},w(x)dx))} \lesssim \sup_{Q \; \text{dyadic cube}} A_{p}(w,Q)^{\frac{1}{p-1}}A_{r}(w,Q)^{1-\frac{1}{p-1}}, \] for a large class of maximal singular integral operators \(T^{*}\) (as well as a similar result for Littlewood-Paley square functions). The paper also shows that this new mixed condition is incomparable with other \(A_{p}-A_{\infty}\) conditions, and points out that this fact suggests that further improvements should still be possible.