On the resolvents of dyadic paraproducts (Q1344576): Difference between revisions

Let \(\omega(x)\) be a positive function on \(\mathbb{R}^n\). For each integer \(j\) we define a function \(\omega_j\) as follows: for \(Q\) a dyadic cube of side length \(2^{-j}\) and \(x\in Q\) set \(\omega_j(x)= |Q|\omega(x)/\int_Q\omega\). For a function \(f\) let \(\Delta_j f\) be the Haar decomposition of \(f\) at level \(j\). The author studies the boundedness of the operator which sends \(f\) to \(Tf(x)= \sum_j\omega_j(x) \Delta_jf(x)\) and related operators. It is shown that, assuming that \(\omega\) satisfies a doubling condition, \(T\) is bounded on \(L^p(\mathbb{R}^n)\) if and only if \(\omega\) satisfies a (dyadic) reverse Hölder condition and also iff the associated weighted maximal function and the square function operators are bounded. It is also shown that \(P\), a variant of \(T\), satisfies \(P= (I-\Pi_b)^{-1}\), where \(\Pi_b\) is the dyadic paraproduct with symbol \(b\) in dyadic BMO (whence the title of the paper). The relation between \(b\) and the weight function \(\omega\) is the interesting nonlinear relation between BMO functions and weights introduced in \textit{R. A. Fefferman}, \textit{C. E. Kenig} and \textit{J. Pipher} [Ann. Math., II. Ser. 134, No. 1, 65-124 (1991; Zbl 0770.35014)] and explored further in \textit{S. M. Buckley} [Mich. Math. J. 40, No. 1, 153-170 (1993; Zbl 0794.42011)]. Some of the proof works with the algebraic structure of the Haar functions. The analytical estimates center on the boundedness of a weighted dyadic square function acting on \(L^2\). That function is, in turn, controlled by a weighted maximal function. The analytical estimates for general \(L^p\) follow from those on weighted \(L^2\) by extrapolation.