\(L^2\) bounds for a maximal directional Hilbert transform (Q2152580)

For a nonzero vector \(\vec{\omega} \in \mathbb{R}^{n+1}\), the directional Hilbert transform on \(\mathbb{R}^{n+1}\) in the direction \(\vec{\omega}\) is the operator \[ H_{v}f(x) = \text{p.v.} \frac{1}{\pi} \int_{\mathbb{R}} f(x-t \vec{\omega}) \frac{dt}{t}, \quad x \in \mathbb{R}^{n+1}, \] where \(\vec{\omega}\) is a vector of the form \(\vec{\omega} = <v, 1>\) with \(v \in \mathbb{R}^{n}\). Let \(\Omega\) be a set of points in \(\mathbb{R}^{n}\). Along with the directional Hilbert transform \(H_{v}\), the maximal directional Hilbert transform \(H_{\Omega}\) associated with the set of directions \(\{ <v, 1> : v \in \Omega \} \subseteq \mathbb{R}^{n+1}\) is defined by \[ H_{\Omega}f(x) := \sup_{v \in \Omega} |H_{v}f(x)|, \quad x \in \mathbb{R}^{n+1}. \] The main purpose of this paper is to obtain the \(L^2(\mathbb{R}^{n+1})\) operator norm of \(H_{\Omega}\) with \[ \| H_{\Omega}\|_{L^2 \rightarrow L^2 } \leq C_{\epsilon} \ N^{\frac{n-1}{2n} + \epsilon} \tag{1} \] that depends only on \(\Omega\) of cardinality \(N\) for \(n \geq 2\) and any \(\epsilon > 0\). Moreover, by using an almost-orthogonality principle for \(H_{\Omega}\), the authors establish the additional geometric assumptions on \(\Omega\) with a fixed cardinality \(N\) to improve the uniform bound in (1) as follows: Let \(d \in \mathbb{N}\), \(m, n \in \mathbb{Z}\) such that \(n \geq 2\) and \(0 \leq m \leq n\). Let \(\mathcal{V}(m,n,d)\) denote the collection of all algebraic varieties \(V\) in \(\mathbb{C}^n\) of dimension at most \(m\) and degree at most \(d\). Set \[ V(\mathbb{R}) := V \cap [ \mathbb{R}^n + i \{ \vec{0}\} ] = \{ x \in \mathbb{R}^n : x \in V \}. \] Then for every \(\epsilon > 0\), there are constants \(C_{\epsilon,m,d} > 0\) such that for any \(V \in \mathcal{V}(m,n,d)\) and any finite direction set \(\Omega \subset V(\mathbb{R})\) of cardinality \(N\), the inequality \[ \| H_{\Omega}\|_{L^2 \rightarrow L^2} \leq \begin{cases} d & \text{when} \quad m = 0 \\ C_{\epsilon,m,d} \ N^{\frac{m-1}{2m} + \epsilon} & \text{when} \quad 1 \leq m \leq n \end{cases} \tag{2} \] hold. It is pointed out that the \(n\)-dimensional complex Euclidean space \(\mathbb{C}^n\) is itself a variety of dimension \(n\) and degree \(1\), and thus (1) is a special case of (2) with \(m=n\) and \(d=1\) for \(V =\mathbb{C}^n\).