Spectral functions for the vector-valued Fourier transform (Q1624151)

Summary: A scalar distribution function \(\sigma(s)\) is called a spectral function for the Fourier transform \(\hat{\varphi}(s) = \int_{\mathbb{R}} e^{i t s} \varphi(t) d t\) (with respect to an interval \(\mathcal{I} \subset \mathbb{R}\)) if for each function \(\varphi \in L^2(\mathbb{R})\) with support in \(\mathcal{I}\) the Parseval identity \(\int_{\mathbb{R}} |\hat{\varphi}(s)|^2 d\sigma(s) = \int_{\mathbb{R}} |\varphi|(t)^2 d t\) holds. We show that in the case \(\mathcal{I} = \mathbb{R}\) there exists a unique spectral function \(\sigma(s) = (1 / 2 \pi) s\), in which case the above Parseval identity turns into the classical one. On the contrary, in the case of a finite interval \(\mathcal{I} = (0, b)\), there exist infinitely many spectral functions (with respect to \(\mathcal{I}\)). We introduce also the concept of the matrix-valued spectral function \(\sigma(s)\) (with respect to a system of intervals \(\{\mathcal{I}_1, \mathcal{I}_2, \ldots, \mathcal{I}_n \}\)) for the vector-valued Fourier transform of a vector-function \(\varphi(t) = \{\varphi_1(t), \varphi_2(t), \ldots, \varphi_n(t) \} \in L^2(\mathcal{I}, \mathbb{C}^n)\), such that support of \(\varphi_j\) lies in \(\mathcal{I}_j\). The main result is a parametrization of all matrix (in particular scalar) spectral functions \(\sigma(s)\) for various systems of intervals \(\{\mathcal{I}_1, \mathcal{I}_2, \ldots, \mathcal{I}_n \}\).