Distributional estimates for the bilinear Hilbert transform (Q2508667): Difference between revisions
From MaRDI portal
Latest revision as of 08:22, 30 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Distributional estimates for the bilinear Hilbert transform |
scientific article |
Statements
Distributional estimates for the bilinear Hilbert transform (English)
0 references
20 October 2006
0 references
The authors produce restricted weak-type estimates for the bilinear Hilbert transform that are analogous to \textit{R. A. Hunt}'s restricted weak-type estimates [On the convergence of Fourier series. Orthogonal Expansions continuous Analog., Proc. Conf. Southern Illinois Univ. Edwardsville 1967, 235--255 (1968; Zbl 0159.35701)] for the Carleson operator. The main theorem is as follows. Let \(2\leq q<\infty\) and \(\alpha\notin \{0,1\}\). Then there exist constants \(C,c\) depending on \(\alpha\) and \(q\) such that whenever \(E\) and \(F\) are measurable sets of finite measure, one has \[ | \{ H_\alpha (\chi_{E},\chi_F)>\lambda\}| \leq C \bigl(| E| ,| F| ^{1\over q}\bigr)^{q\over 1+q}G(\lambda) \] where \(G(\lambda)=\Bigl({1\over\lambda}\bigl(1+\log {1\over\lambda}\bigr)^2\Bigr)^{q\over 1+q} \) if \(\lambda<1\) and \(G(\lambda)=e^{-c\sqrt{\lambda}}\) if \(\lambda>1\). The roles of \(E\) and \(F\) are symmetric in terms of this estimate, although they are not completely symmetric in terms of the bilinear Hilbert transform \[ H_\alpha (f,g)(x)={1\over \pi}\lim_{\epsilon\downarrow 0} \int_{| t| \geq \epsilon} f(x-t) g(x+\alpha t){dt\over t}, \] in so far as \(H_\alpha\) depends on \(\alpha\). The particular case \(q=2\) yields a restricted weak-type endpoint estimate for the bilinear Hilbert transform. As in the work of \textit{M. T. Lacey} [Doc. Math., J. DMV Extra Vol. ICM Berlin 1998, Vol. II, 647--656 (1998; Zbl 0963.42007)] and \textit{M. T. Lacey} and \textit{C. Thiele} [Ann. Math. (2) 149, No. 2, 475--496 (1999; Zbl 0934.42012)], the authors reduce the problem to obtaining desired estimates for a certain model quartile (here called tri-tile) operator of the form \[ H_S (f,g)(x) = \sum_{s\in S} | I_s| ^{-1/2} \epsilon_s \langle f,\varphi_{s_1}\rangle \, \langle g,\varphi_{s_2}\rangle \,\varphi_{s_3} \] in which \(S\) is a collection of time-frequency rectangles \(s\) of fixed area and \(s_i=I_s\times \omega_{s_i}\) are three subrectangles, subject to certain combinatorial rules, such that the Fourier transform \(\widehat{\varphi}_{s_i}\) is supported in the frequency interval \(\omega_{s_i}\) and well-localized on the time interval \(I_s\). One can take advantage of \(f,g\) being characteristic functions in getting good estimates for the quantities \(| \langle H_s(\chi_E,\chi_F),\chi_K\rangle| \) with \(K\) a third measurable set.
0 references
bilinear Hilbert transform
0 references
distributional inequality
0 references
restricted weak-type inequality
0 references