A characterization of two weight norm inequalities for maximal singular integrals with one doubling measure (Q1951719)

Conditions are given for boundedness of certain maximal singular integrals on \(\mathbb{R}^n\) between two measure weighted \(L^p\)-spaces when the weight in the range is a doubling measure. A corresponding weak-type inequality is established without the doubling condition when an \(A_p\) condition is added. Let \(K(x,y)\) be a kernel function on \(\mathbb{R}^n\times \mathbb{R}^n\) such that \(|K(x,y)|\leq C|x-y|^{-n}\) and \(|K(x,y)-K(x',y)|\leq C \delta (|x'-x|/|x-y|)|x-y|^{-n}\) when \(|x'-x|\leq |x-y|/2\) and \(\delta\) is a Dini modulus (i.e. \(\delta\) is nondecreasing on \([0,1]\), \(\delta(0)=0\) and \(\int_0^1 \delta(s) ds/s <\infty\)). Let \(T_{\epsilon,R} f(x)\) be the truncation of the integral \(\int K(x,y) f(y)\, dy\) by multiplying the integrand by a smooth cutoff function that vanishes outside \(\epsilon/2<|x-y|<2R\) and equals one inside \(\epsilon<|x-y|<R\). One defines a maximal operator by \(T_\flat f(x)=\sup_{0<\epsilon<R<\infty} |T_{\epsilon,R} f(x)|\). Non-symmetric cutoffs with bounded eccentricity are also considered for \(x-y>\epsilon_1\) and \(y-x>\epsilon_2\): in dimension \(n=1\), \(T_\natural f(x)=\sup_{0<\epsilon_i<R<\infty; 1/4<\epsilon_1/\epsilon_2<4} |T_{{\epsilon},R} f(x)|\). In higher dimensions, one defines a corresponding sharp maximal operator \(T_\natural\) by taking products of smooth cutoffs in the separate variables. The maximal inequality is phrased as follows: Suppose that \(\sigma \) and \(\omega\) are positive, locally finite Borel measures on \(\mathbb{R}^n\) with \(\sigma\) doubling, let \(1<p<\infty\), and let \(T_\flat\) and \(T_\natural\) be as described. Then \(T_\natural\) satisfies \[ \int_{\mathbb{R}^n} T_\natural (f\sigma)(x)^p d\omega (x)\leq C\int_{\mathbb{R}^n} |f(x)|^p d\sigma(x) \] for all \(f\in L^p(\sigma)\) that are bounded with compact support in \(\mathbb{R}^n\), provided a certain maximal function inequality holds, the dual cube testing condition \[ \int_Q T_\natural (\chi_Q f \sigma)(x)\, d\omega(x) \leq C_2\|f\|_{L^p(\sigma)}\omega(Q)^{1/p'} \] holds for all \(f\in L^p(\sigma)\) and all cubes \(Q\), and the testing condition \[ \int_Q T_\natural (\chi_Q g \sigma)(x)^p\, d\omega(x) \leq C_1 \int_Q d\sigma(x) \] holds for all cubes \(Q\subset\mathbb{R}^n\) and all functions \(|g|\leq 1\). The maximal operator condition is that the maximal function \(\mathcal{M}\nu(x)=(\int_Q\nu)/|Q|\) satisfies \(\|(\mathcal{M}(\chi_Q\sigma))\chi_Q\|_{L^p(\omega)}\leq C\int_Q d\sigma\) along with the dual condition \(\|\mathcal{M}(g\omega)\|_{L^{p'}(\sigma)}\leq C\|g\|_{L^{p'}(\omega)}\) (these are \textit{E. T. Sawyer}'s conditions for two-weight maximal inequalities [Stud. Math. 75, 1--11 (1982; Zbl 0508.42023)]). The same characterization applies with \(T_\natural\) replaced with \(T_\flat\). When both \(\sigma\) and \(\omega\) are absolutely continuous, the \(L^p\) norm inequality holds for \(T\) if it holds for either \(T_\natural\) or \(T_\flat\), provided the maximal condition is satisfied. As noted, corresponding weak-type inequalities are established when neither weight is assumed doubling. General principles are laid out before diving into technical details. In particular, a schematic tree delineating different levels of decomposition of \(\int(T_\natural f\sigma)^p\omega\) and corresponding estimation techniques is provided.