Endpoint bounds for the quartile operator (Q485176): Difference between revisions

The quartile operator was introduced by \textit{C. Thiele} [Trans. Am. Math. Soc. 352, No. 12, 5745--5766 (2000; Zbl 0976.42017)] as a discrete Walsh function model for the bilinear Hilbert transform capable of illustrating conceptual challenges in proving boundedness properties of the latter while also allowing for cleaner combinatorial arguments. In this work a restricted weak-type inequality is proved that implies boundedness of the quartile operator from \(L^{p_1}\times L^{p_2,2/3}\to L^{2/3,\infty}\) when \(1/{p_1}+1/{p_2}=3/2\). It is conjectured that the quartile operator is, in fact, bounded from \(L^{p_1}\times L^{p_2}\to L^{2/3,\infty}\). The quartile operator is defined as the sum \(V_{\mathbf{Q}}(f_1,f_2)(x)=\sum_{s\in\mathbf{Q}}\frac{1}{\sqrt{|I_S|}}\langle f_1,\, w_{s_1}\rangle\, \langle f_2,\, w_{s_2}\rangle \, w_{s_3}\) where \(\mathbf{Q}\) is the collection of quartiles and \(w_s\) is the Walsh packet \(w_s(x)=|I_s|^{-1/2}W_{n_s}((x-\ell(I_s))/|I_s|)\) where the Walsh functions \(W_n\) are defined by \(W_0=\mathbf{1}_{[0,1]}\) and \(W_n(x)=\prod_{k\geq 0} (\text{sign}(\sin(\pi 2^{k+1}x)))^{\epsilon_k}\) where \(n=\sum \epsilon_k 2^k\). A tile is a dyadic rectangle \(s=I_s\times \omega_s\subset \mathbb{R}\times\mathbb{R}_+\) of unit area, i.e. the time and sequency (Walsh frequency) intervals \(I_s\) and \(\omega_s\) satisfy \(|I_s||\omega_s|=1\). The associated Walsh packet has \(n_s=|I|_s \ell(\omega_s)\) where \(\ell(\omega)\) is the left endpoint of the frequency interval \(\omega\). Quartiles are dyadic rectangles of area four with subrectangles \(s_1,\dots s_4\) arranged in order of increasing sequency. The main result is the weak-type estimate \[ \lambda |\{x:|V_{\mathbf{Q}}(f_1,f_2)(x)\}>\lambda|^{3/2}\lesssim (p_1')\|f_1\|_{p_1}|F_2|^{1/p_2},\qquad |f_2|\leq\mathbf{1}_{F_2}\, . \] Walsh phase plane combinatorics are augmented with a John-Nirenberg type inequality for tree projections in terms of uniform \(L^2\) ``size'' estimates as well as a multi-frequency Calderón-Zygmund decomposition.