On sharp aperture-weighted estimates for square functions (Q485262)

Let \(\psi\) be an integrable function satisfying \(\int_{\mathbb R^n}\psi(x)dx=0\), and \(|\psi(x)|\leq C(1+|x|)^{-n-\varepsilon}\), \(\int_{\mathbb R^n}|\psi(x+h)-\psi(x)|dx\leq C|h|^\varepsilon\) for some \(\varepsilon >0\). Define the Lusin type square function \(S_{\alpha,\psi}(f)\) by \[ S_{\alpha,\psi}(f)(x)=\Big(\int_{|y-x|<\alpha t}|f*\psi_t(y)|^2 \frac{dydt}{t^{n+1}}\Big)^{1/2} (\alpha>0). \] The author shows that for any \(1<p<\infty\) and \(\alpha\geq1\) it holds \[ \|S_{\alpha,\psi}(f)\|_{L^p(w)}\leq C \alpha^n[w]_{A_p}^{\max(1/2,1/(p-1))}, \] where \(A_p\) is the Muckenhoupt weight class and \([w]_{A_p}\) is the \(A_p\) characteristic defined by \[ [w]_{A_p}= \sup_{Q:\text{cubes}}\Big(|Q|^{-1}\int_Qwdx\Big)\Big(|Q|^{-1}\int_Qw^{(p-1)^{-1}}dx\Big)^{p-1}. \] On all classes \(A_p\) the result is sharp in \(\alpha\). For each fixed \(\alpha\geq1\) the dependence on \([w]_{A_p}\) is sharp, where the case \(\alpha=1\) was already shown by the author in [Adv. Math. 226, No. 5, 3912--3926 (2011; Zbl 1226.42010)]. The author explains the reason why one cannot get the result in this paper by using the proof method in the above paper.