Uncertainty principles for Fourier multipliers (Q2007447)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Uncertainty principles for Fourier multipliers |
scientific article |
Statements
Uncertainty principles for Fourier multipliers (English)
0 references
14 October 2020
0 references
For \(f\in L^1(\mathbb T^d)\), let \(\mathcal F f\) on \(\mathbb Z^d\) be the Fourier transform of \(f\) defined by \[ \mathcal F f(k)= \int_{\mathbb T^d}f(x)e^{2\pi i xk} \, dx, \] where \(xk=x_1k_1+\dots +x_dk_d\), \(x=(x_1, \dots, x_d)\), \(k=(k_1, \dots, k_d)\). Let \(u\in L^2(\mathbb T^d)\) and \(2\leq q\leq \infty\). Then \(u\in \mathcal M_2^q\) (the set of \((2,q)\) Fourier multipliers) if the operator \(T_u\) is bounded from \(\ell^2(\mathbb Z^d)\) to \(\ell^q(\mathbb Z^d)\), where \[ T_u a =\mathcal F \left(u\mathcal F^{-1}a\right). \] Among other results of this article, Theorems 1 and 2 below are proved; also generalizations of them and applications are given. Theorem 1. Let \(d/2<s\leq d\). Suppose that \(w\in W^{s,2}(\mathbb T^d)\) (the Sobolev space) and \(w\) has a zero. \begin{itemize} \item[(1)] Further assume that \(s< d/2+1\), then \(1/w \notin \mathcal M_2^q\) for all \(q\) with \(2\leq q \leq d/(d-s)\). If \(q>d/(d-s)\), there exists \(w \in W^{s,2} (\mathbb T^d)\) such that \(1/w \in \mathcal M_2^q\). \item[(2)] If \(s=d/2 +1\), then \(1/w \notin \mathcal M_2^q\) for all \(q\) with \(2\leq q < 2d/(d-2)\). \end{itemize} As a fact related to Theorem 1 (1), we note the following. If \(0<s-d/2<1\) and \(|w(x)|\leq C|x|^{s-d/2}\) for \(|x|<1/2\), then \[ \int_{|x|<1/2}|w(x)|^{-2d/(2s-d)}\, dx \gtrsim \int_{|x|<1/2}|x|^{-(s-d/2)(2d/(2s-d))}\, dx =\int_{|x|<1/2}|x|^{-d}\, dx =\infty, \] which implies that \(1/w \notin L^{2q/(q-2)}(\mathbb T^d)\) when \(q=d/(d-s)\), since then \(2q/(q-2)=2d/(2s-d)\). This will be interesting in connection with Theorem 1 (1) if we recall that \(L^{2q/(q-2)}(\mathbb T^d) \subset \mathcal M_2^q\), which is a consequence of the Hausdorff-Young inequality. Let \(w\in L^1(\mathbb T^d)\) and \[ \Sigma(w)=\left\{x\in\mathbb T^d: \limsup_{\tau\to 0} \frac{1}{|I_\tau(x)|} \int_{I_\tau(x)} |w(y)|\, dy=0 \right\}, \] where \(I_\tau(x)\) denotes the cube with sidelength \(2\tau\) centered at \(x\). Theorem 2. Let \((d-\sigma)/2<s\leq 1\), \(w\in W^{s,2}(\mathbb T^d)\) and \(\mathcal H^\sigma(\Sigma(w))>0\) (the \(\sigma\)-dimensional Hausdorff measure). \begin{itemize} \item[(1)] If \(s<d-\sigma\), then \(1/w \notin \mathcal M_2^q\) for all \(q\) with \(2\leq q\leq d/(d-s-\sigma/2)\). \item[(2)] If \(s=d-\sigma\), then \(1/w\notin L^2(\mathbb T^d)\), and hence \(1/w\notin \mathcal M_2^q\) for all \(q\geq 2\). \end{itemize} In this article Theorems 1 and 2 are shown in more general forms and versions of matrix valued Fourier multipliers; also, their applications to time-frequency analysis for the Gabor system and shift invariant spaces are given. Let \(g\in L^2(\mathbb R)\) and \(G(g)=\{e^{2\pi imx}g(x-n)\}_{(m,n)\in \mathbb Z^2}\) be the Gabor system generated by \(g\). When \(G(g)\) is an exact \((C_q)\)-system for \(L^2(\mathbb R)\), sharp results of the Balian-Low type are shown. This is done by using the Zak transform from \(L^2(\mathbb R)\) to \(L^2(\mathbb T^2)\) defined by \[ Zg(x,y)= \sum_{k\in \mathbb Z} g(x-k)e^{2\pi iky}. \] It is shown that \(G(g)\) is an exact \((C_q)\)-system in \(L^2(\mathbb R)\) if and only if \(1/|Zg| \in \mathcal M_2^q(\mathbb T^2)\), where \(q\geq 2\) and \(g\in L^2(\mathbb R)\). Also, it is shown that if \(g\) and \(\hat g\) (the Fourier transform) belong to weighted \(L^2\) spaces with certain power weights, then \(Zg\) satisfies certain regularity estimates locally, which enables to apply a version of Theorem 1, where the local regularity estimates can replace the assumption in Theorem 1 stated using Sobolev spaces on the whole \(\mathbb T^2\), to conclude that \(1/|Zg| \notin \mathcal M_2^q(\mathbb T^2)\). These results impose certain constraints on the weighted square integrability of \(g\) and \(\hat g\) when \(G(g)\) is an exact \((C_q)\)-system in \(L^2(\mathbb R)\). Furthermore, analogues of the Balian-Low theorem for shift invariant spaces are considered by applying versions of Theorem 2 for matrix valued Fourier multipliers.
0 references
Fourier multiplier
0 references
Balian-Low theorem
0 references
uncertainty principle
0 references
0 references
0 references
0 references