On \(A_p\)-\(A_\infty \) type estimates for square functions (Q343655)

The authors study sharp weighted norm inequalities for the intrinsic square function \[ G_{\beta,\alpha}(f)(x):=\bigg(\int_{\Gamma_\beta(x)} (A_\alpha (f)(y,t))^2 \frac{dydt}{t^{n+1}}\bigg)^\frac{1}{2}, \] where \(\Gamma_\beta(x)\) is the cone in \(\mathbb R^{n+1} _+\) with vertex \(x\in\mathbb R^n\) and aperture \(\beta\), and \(A_\alpha (f)(y,t):=\sup_{\phi \in \mathcal C_\alpha} |f*\phi_t(y)|\). Here \(\mathcal C_\alpha\) is the family of \(\alpha\)-Hölder continuous functions, \(\alpha\in(0,1]\), supported in the unit ball of \(\mathbb R^n\) with mean 0. This type of ``mother'' square function, introduced by \textit{M. Wilson} in [Weighted Littlewood-Paley theory and exponential-square integrability. Berlin: Springer (2008; Zbl 1138.42011)], controls several other square functions of importance in harmonic analysis such as the usual Littlewood-Paley \(g\) function or the Lusin area integral. The authors prove two types of weight norm inequalities for this square function. The first is a two-weight \(A_p-A_\infty\) type estimate that reads \[ { \|G_{\beta,\alpha}(\cdot \sigma) \|_{L^p(\sigma)\to L^p(w)} }\lesssim \begin{cases} [w,\sigma]_{A_p} ^\frac{1}{p} [\sigma]_{A_\infty} ^\frac{1}{p},&\quad 1<p\leq 2, \\ [w,\sigma]_{A_p} ^\frac{1}{p} [\sigma]_{A_\infty} ^\frac{1}{p} \big([w]_{A_\infty} ^{\frac{1}{2}-\frac{1}{p}}+[\sigma]_{A_\infty} ^\frac{1}{p}\big),&\quad p>2.\end{cases} \] Here it is assumed that the pair of weights \((w,\sigma)\) satisfies the two-weight \(A_p\) condition \[ [w,\sigma]_{A_p}:= \sup_Q \langle w \rangle_Q \langle \sigma\rangle _Q ^{p-1}<\infty, \] and that both \(w,\sigma\in A_\infty\) (and \([v]_{A_\infty}\) is the \(A_\infty\)-constant of Fujii-Wilson, see for example the previous reference to Wilson.) If one specializes this result to the one-weight case then it includes the one-weight sharp \(A_p\)-bound of \textit{A. K. Lerner} from [Adv. Math. 226, No. 5, 3912--3926 (2011; Zbl 1226.42010)]. The second main result of the paper also addresses two-weight norm inequalities for the intrinsic square function. Here the mixed \(A_p\) assumptions and the \(A_\infty\) assumption on the weights \((w,\sigma)\) are replaced by appropriate ``one-supremum'' entropy bounds. To give an example, the authors obtain the estimate \[ \|G_{\beta,\alpha}(\cdot \sigma)\|_{L^p(\sigma)\to L^p(w)} \lesssim [w,\sigma]_{p,\epsilon}, \] where \([w,\sigma]_{p,\epsilon}\) is an appropriate entropy constant involving only one supremum. For comparison, in the case \(1<p\leq 2\) the entropy constant is given by \[ [w,\sigma]_{p,\epsilon}:=\sup_Q \langle w \rangle _Q ^\frac{1}{p} \langle \sigma \rangle _Q ^\frac{1}{p'} \rho_{\sigma,\epsilon} (Q)^\frac{1}{p}. \] Note that there is only one supremum in this definition. In the product of the first two terms one recognizes the \(1/p\)-th root of the two-weight \(A_p\) constant for the pair \((w,\sigma)\). The entropy term \(\rho_{\sigma,\epsilon}(Q)\) is a term that is roughly a log-bumped \([\sigma]_{A_\infty}\) when \(\sigma \in A_\infty\). However, it allows for bounds when the quantity \(\rho(Q)\), definining the \(A_\infty\)-constant of \(\sigma\) on cubes \(Q\), becomes arbitrarily large. Thus, this is a stronger theorem with a log-loss in the quantitative bounds, compared to the first main result of the paper. The proof of these results relies on sparse bounds for the square function, appropriate parallel corona decompositions to decompose the duality form of the sparse square function, and Carleson embeddings in terms of \(\epsilon\)-bumps.