A discrete model for the bi-Carleson operator. (Q1811216)

The study of the Carleson operator reduces to the study of its linearized version \[ C(f)=\int_{\xi < N(x)} \widehat{f} e^{2\pi i x\xi} d\xi, \] where \(x\mapsto N(x)\) is arbitrary. In the same way, the study of the maximal variant of the bilinear Hilbert transform, \(B\), reduces to the study of the bi-Carleson operator, \[ T(f_1,f_2)=\int_{\xi_1<\xi_2<N(x)} \widehat{f}_1\widehat{f}_2 e^{2\pi i x(\xi_1+\xi_2)} d\xi_1d\xi_2. \] Both \(C\) and \(B\) have simpler Walsh model versions, \(C_{\text{walsh}}\) and \(B_{\text{walsh}}\), defined using the Walsh transform instead of the Fourier transfom. Just as \(C\) and \(B\), \(T\) has a Walsh model \(T_{\text{walsh},P,Q}\) asociated to every pair \(P\) and \(Q\) finite collections of quartiles (a quartile is a product of two dyadic intervals that has area four). The main theorem of the paper shows that \(T_{\text{walsh},P,Q}: L^{p_1} \times L^{p_2}\to L^{p'_3}\) as long as \(1<p_1,p_2\leq \infty\), \(1/p_1+1/p_2=1/p'_3\) and \(2/3<p'_3<\infty\), with uniform bounds in \(P\) and \(Q\). This theorem includes as particular cases the known \(L^p\) estimates for \(C_{\text{walsh}}\) and \(B_{\text{walsh}}\). The authors announce that the corresponding estimates for the Fourier model will be included in a forthcoming paper.