Local \(Tb\) theorem with \(L^2\) testing conditions and general measures: square functions (Q1687970)

The paper under review establishes \(L^2\) square function bounds with respect to a measure \(\mu \) on \(\mathbb{R}^n\) that is \(m\)-dimensional, that is, \(\mu(B(x,r)\leq C r^m\) for all \(x\in\mathbb{R}^n\) and all \(r>0\). The results and methods parallel ones for Calderón-Zygmund operators (CZOs) established by the authors [Ann. Sci. Éc. Norm. Supér. (4) 49, No. 1, 57--86 (2016; Zbl 1345.42016)]. Conical and vertical square functions are defined as \[ Sf(x)=\biggl(\int\int_{\Gamma(x)}|\theta_t f(y)|^2\, d\mu(y)\frac{dt}{t^{m+1}} \biggr)^{1/2}\, \text{and}\, Vf(x)=\biggl(\int_0^\infty |\theta_t f(y)|^2\frac{dt}{t}\biggr)^{1/2} \] where \(\Gamma(x)\) is the cone based at \(x\): \(\{(y,t)\in\mathbb{R}^{n+1}_+:|x-y|\leq t\}\) . Here, \(\theta_tf(x)=\int_{\mathbb{R}^n} s_t(x,y)\, f(y)\, d\mu(y)\) with kernels \(s_t\) satisfying uniform molecular bounds \[ |s_t(x,y)|\leq C\frac{t^\alpha}{(t+|x-y|)^{m+\alpha}}\quad\text{and}\quad |s_t(x,y)-s_t(x,z)|\leq C\frac{|y-z|^\alpha}{(t+|x-y|)^{m+\alpha}} \] (regularity is not necessary in the \(x\)-coordinate). It is observed that if \(\widetilde{V}\) is the vertical square function with \(\tilde{s}_t(x,y)= s_t(x,y)\, ((\mu(B(x;t))/t^m)^{1/2}\) then \(\|\widetilde{V}(f)\|_{L^2(\mu)}=\|Sf\|_{L^2(\mu)}\) and \(\tilde{s}_t\) satisfies the same molecular bounds as \(s_t\), so it is enough to study \(L^2\)-boundedness of \(V\). One assumes that to each dyadic cube there is a function \(b_Q\) having unit average on \(Q\) such that \(\int |b_Q|^2\leq C\mu(Q)\) and such that \(\int\int_{\widehat{Q}} |\theta_t b_Q(x)|^2d\mu(x){dt\over t}\leq C\mu(Q)\) where \(\widehat{Q}=Q\times [0,\ell(Q)]\) is the Carleson box over \(Q\). The main theorem (1.3) states under these local \(Tb\)-type testing conditions, the vertical square function \(V\) is bounded on \(L^2(\mu)\). The analytic apparatus is similar to that applied in the authors' work on CZOs where corresponding bounds are established for Calderón-Zygmund operators. To a cube \(Q\in\mathcal{D}\), \(Q\subset Q^\ast\) with children denoted \(\text{ch}{Q}\), one associates the twisted martingale difference operator \[ \Delta_Q g=\sum_{Q'\in {\mathrm{ch}}Q} \Bigl(\frac{\langle g\rangle_{Q'}}{\langle b_{(Q')^a}\rangle_{Q'}} b_{(Q')^a}- \frac{\langle g\rangle_{Q}}{\langle b_{Q^a}\rangle_{Q}} b_{Q^a} \Bigr) {1}_{Q'}. \] where \(Q^a\) is the minimal cube containing \(Q\) in a certain set \(\mathcal{F}_{Q^\ast}\) of \textit{stopping cubes} defined in terms of the \(b_Q\) but here, in contrast to the authors' work on CZOs, also in terms of the function \(g\) on which a local \(L^2\)-bound (with constant independent of \(g\)) is sought. The main technical tool in proving Theorem 1.3 is Proposition 2.5 which states that \[ \sum_{Q\in\mathcal{D}, Q\subset Q^\ast} \|\Delta_Q g\|_{L^2(\mu)}^2\leq C\|g\|_{L^2(Q^\ast, \mu)}^2. \] with a bound independent of \(f\). Another key ingredient is Proposition 2.7 which states that \(g {1}_{Q^\ast}=\sum_{Q\in\mathcal{D}, Q\subset Q^\ast}\Delta_Q g+\langle g\rangle_{Q^\ast} b_{Q^\ast} {1}_{Q^\ast}\).