On entropy bumps for Calderón-Zygmund operators (Q2353436): Difference between revisions

The entropy-bound method of \textit{S. Treil} and \textit{A. Volberg} [``Entropy conditions in two weight inequalities for singular integral operators'', Preprint (2014), \url{arXiv:1408.0385}] is extended to the \(L^p\) setting with short proofs. The general program is to prove two-weight inequalities for singular integrals and maximal functions with simply formulated weight conditions that are demonstrably close to being optimal. Three theorems are proved. The first deals with the maximal operator \(M_\sigma(f)=M(\sigma f)\). Define \(\lceil \sigma, w\rceil_{p,\epsilon}=\sup_Q \rho_{\sigma,\epsilon}(Q)\langle \sigma\rangle_Q^{p-1}\langle w\rangle_Q\), where \(\langle f\rangle_Q=(\int_Q f(x)\, dx)/|Q|\) and \(\rho_\sigma (Q)=\frac{1}{\sigma(Q)} \int_Q M(\sigma \chi_Q)\, dx\), \(\rho_{\sigma,\epsilon}(Q)=\rho_\sigma(Q)\epsilon(\rho_\sigma(Q))\), where \(\epsilon(t)\) is increasing on \([1,\infty)\) and the supremum taken over cubes in \(\mathbb{R}^d\). Also, let \(\lfloor \sigma, w\rfloor_{p,\epsilon}=\sup_{Q} \rho_{w,\epsilon}(Q)^{p-1}\langle \sigma\rangle_Q^{p-1}\langle w\rangle_Q\). Theorem 1 states that if \(\sigma\) and \(w\) are two absolutely continuous weights, \(1<p<\infty\) and \(\epsilon\) is monotone increasing on \((1,\infty)\) such that \(\int_1^\infty \frac{1}{\epsilon(t)}\,\frac{dt}{t}=1\), then \(M_\sigma\) is bounded from \(L^p(\sigma)\) to \(L^p(w)\) with norm bounded by a constant multiple of \(\lceil \sigma, w\rceil_{p,\epsilon}^{1/p}\). Theorem 2 states that any Calderón-Zygmund operator \(T\) with standard kernel \(K\) is bounded from \(L^p(\sigma)\) to \(L^p(w)\) with norm bounded by a constant (depending on the Calderón-Zygmund conditions) multiple of \(\lfloor \sigma, w\rfloor_{p,\epsilon}^{1/p}\). The case \(p=2\) was proved in by Treil and Volberg [loc. cit]. The condition on \(\epsilon\) is the same as in Theorem 1. Theorem 3 lets \(\epsilon_p\) and \(\epsilon_{p'}\) be two monotone increasing functions such that \(\int_1^\infty \frac{1}{\epsilon_p(t)^{1/p}}\frac{d t}{t}=1\) and similarly with \(p\) replaced by \(p'\). In this case any standard Calderón-Zygmund operator is bounded from \(L^p(\sigma)\) to \(L^p(w)\) with norm bounded by a constant (depending on the Calderón-Zygmund conditions) multiple of \(\lceil \sigma, w\rceil_{p,\epsilon_p}^{1/p} +\lceil \sigma, w\rceil_{p',\epsilon}^{1/{p'}}\). The integrability condition on \(\epsilon_p\) and \(\epsilon_{p'}\) here is stronger than in Theorem 2. The proof reduces to sparse operators based on [\textit{A. K. Lerner}, Int. Math. Res. Not. 2013, No. 14, 3159--3170 (2013; Zbl 1318.42018)] together with pigeon-hole and stopping time arguments and a certain \(A_p\)-\(A_\infty\) inequality.