Twisted Hilbert transforms vs Kakeya sets of directions (Q901387)

The authors summarize the contents of this paper in the abstract of the paper as follows: Given a discrete group \(\mathrm{G}\) and an orthogonal action \(\gamma : \mathrm{G}\to \mathrm{O}(n)\) we study \(L_p\)-convergence of Fourier integrals which are frequency supported on the semidirect product \(\mathbb{R}^n \rtimes_\gamma \mathrm{G}\). Given a unit \(u \in \mathbb{R}^n\) and \(1< p \neq 2 < \infty\), our main result shows that the twisted (directional) Hilbert transform \(H_u \rtimes_\gamma\mathrm{id}_{\mathrm{G}}\) is \(L_p\)-bounded iff the orbit \(\mathcal{O}_\gamma(u)\) is finite. This is in sharp contrast with twisted Riesz transforms \(R_u \rtimes_\gamma\mathrm{id}_{\mathrm{G}}\), which are always bounded. Our result characterizes Fourier summability in \(L_p\) for this class of groups. We also extend de Leeuw's compactification theorem to this setting and obtain stronger estimates for functions with ``lacunary'' frequency support.