Boundedness of bilinear multipliers whose symbols have a narrow support (Q351282): Difference between revisions

This paper is concerned with bilinear multipliers, which are defined as the operators \[ B_m(f,g)(x) := \frac{1}{(2\pi)^d} \int_{\mathbb{R}^{2d}} e^{ix(\xi+\eta)} m(\xi,\eta) \hat{f}(\xi) \hat{g}(\eta) d\xi d\eta, \] where \(m\) is a singular symbol. It provides a detailed discourse of such mutlipliers in the dimension \(d=1\) and when \(m\) is supported around a curve. More precisely, the authors take a smooth curve \(\Gamma\) in \(\mathbb{R}^2\) and for any \(\epsilon>0\) consider symbols \(m_\epsilon\) satisfying \(0\leq m_\epsilon\leq 1\) and having their support contained in the \(\epsilon\)-neighborhood of \(\Gamma\). The main question is: What is the quantitative dependence of the norms \(\|B_{m_\epsilon}\|_{\mathrm{L}^p\times\mathrm{L}^q\to\mathrm{L}^{r'}}\) on the parameter \(\epsilon\) and what conditions must be imposed on the curve \(\Gamma\), the symbol \(m_\epsilon\), and the exponents \(1\leq p,q,r\leq\infty\) in order to have nontrivial estimates? The authors point out the importance of the points on \(\Gamma\) where the tangent is parallel to one of the lines \(\xi=0\), \(\eta=0\), \(\xi+\eta=0\), and remark that the curvature of \(\Gamma\) plays a secondary role. They also consider either the standard symbol estimates, or those that favor the tangential direction. The exact conditions are quite technical to formulate, as they split into several cases with respect to \(\Gamma\) and many more with respect to the exponents. A detailed overview of the results can be found in the very informative introductory section of the paper. Furthermore, the paper discusses the connections between its main results and the following problems. The bilinear Bochner-Riesz problem is: If \(K\) is a compact domain with smooth boundary and \(m(\xi,\eta)=\chi_K(\xi,\eta)\,\mathrm{dist}((\xi,\eta),\partial K)^\kappa\), for which values of the parameter \(\kappa\) and the exponents \(p,q,r\) is \(B_m\) bounded from \(\mathrm{L}^p\times\mathrm{L}^q\) to \(\mathrm{L}^{r'}\)? The bilinear restriction-extension problem is: If \(\Gamma\) is a rectifiable curve and \(\sigma\) is the one-dimensional Hausdorff measure on it, for which exponents \(p,q,r\) is \(B_\sigma\) bounded from \(\mathrm{L}^p\times\mathrm{L}^q\) to \(\mathrm{L}^{r'}\)? Some motivation for the results also comes from dispersive PDEs and can be found in [\textit{F. Bernicot} and \textit{P. Germain}, Adv. Math. 225, No. 4, 1739--1785 (2010; Zbl 1204.42018)].