A matrix characterization of boundary representations of positive matrices in the Hardy space

John E. Herr,Palle E.T. Jorgensen,Eric S. Weber
DOI: https://doi.org/10.48550/arXiv.1705.04198
2017-05-10
Abstract:Spectral measures give rise to a natural harmonic analysis on the unit disc via a boundary representation of a positive matrix arising from a spectrum of the measure. We consider in this paper the reverse: for a positive matrix in the Hardy space of the unit disc we consider which measures, if any, yield a boundary representation of the positive matrix. We introduce a potential characterization of those measures via a matrix identity and show that the characterization holds in several important special cases.
Functional Analysis,Complex Variables
What problem does this paper attempt to address?
The main problems that this paper attempts to solve are: for positive matrices in the Hardy space of the unit disk, determine which measures (if any) can produce the boundary representations of these positive matrices. Specifically, the authors explore whether, under a given measure, positive matrices in the Hardy space can reproduce themselves through the integral of their boundaries. This involves two main issues: 1. **Which positive matrices are included in the Hardy space** and can reproduce themselves through the boundary functions of a given measure. 2. **For a given positive matrix**, which measures can make it reproduce itself through the integral of the boundary function. To answer these questions, the authors introduce a potential matrix identity to characterize these measures and prove that this characterization holds in several important special cases. ### Specific Contents and Methods - **Hardy Space and Szegő Kernel**: - The paper first reviews the classical Hardy space \(H^2(D)\), where \(D\) is the unit disk. The norm of the Hardy space is defined as: \[ \|f\|_{H^2}^2=\sup_{0 < r < 1}\int_0^1|f(re^{2\pi ix})|^2dx < \infty \] - The Szegő kernel \(k_w(z)=\frac{1}{1 - wz}\) plays an important role in the Hardy space and satisfies: \[ f(z)=\langle f,k_z\rangle_{H^2}=\int_0^1f^*(e^{2\pi ix})k_z^*(e^{2\pi ix})dx \] - **Boundary Representations**: - The authors study the boundary behavior of functions in the Hardy space, especially the boundary representations under a given measure. They consider the following two problems: - Which positive matrices can reproduce themselves through the integral of the boundary function under a given measure? - For a given positive matrix, which measures can make it reproduce itself through the integral of the boundary function? - **Special Case of Diagonal Coefficient Matrices**: - For the diagonal coefficient matrix \(C\), the authors prove that when the diagonal elements of \(C\) are not all the same, there is no representative measure. In addition, they also study the specific diagonal matrices \(K_3\) and \(K_4\) and conclude that: - \(M(K_3)\) only contains the Lebesgue measure. - \(M(K_4)\) contains many absolutely continuous measures. - **Absolutely Continuous Measures**: - For the absolutely continuous measure \(\mu\), the authors prove that \(\mu\in M(K_C)\) if and only if the matrix equation \(C = CMC\) holds, where \(M = (\hat{\mu}(n - m))_{mn}\). ### Conclusion This paper characterizes the measures that can make positive matrices reproduce themselves through the integral of boundary functions by introducing the matrix identity \(C = CMC\) and verifies the validity of this characterization in the cases of diagonal matrices and absolutely continuous measures. This provides a new method for understanding the boundary representations of positive matrices in the Hardy space.