Optimal measures for Toeplitz matrices (Q556905): Difference between revisions

Let \(M\in M_n(\mathbb{C})\). The Schur norm \(\| M\| _S\) of \(M\) is defined as \[ \| M\| _S=\max\{\| M\circ X\|:\| X\| \leq1\}, \] where \(\circ\) is the Schur (or Hadamard) product and \(\| \cdot\| \) is the operator norm. If \(M\) is Toeplitz, i.e. \(M=M(c_{-n+1},\dots,c_{n-1})\), a measure \(\mu\) in \((-\pi,\pi]\) is said to represent \(M\) if its Fourier coefficients satisfy \(\hat\mu(k)=c_{-k}\) when \(| k| <n\). \textit{R. McEachin} [Linear Algebra Appl. 341, No. 1--3, 379--389 (2002; Zbl 1004.42013)] proved that when \(M\) is a real \(2\times2\) Toeplitz matrix, \(\| M\| _S\) coincides with infimum of the total variations of all representating measures \(\mu\). The authors extend McEachin's result to the complex case, and they use their result to derive formulas for the Schur norm of the Toeplitz \(2\times 2\) matrices. They also show that optimal measures (i.e. measures realizing the infimum) are supported in either one, two, or three points, and these cases are characterized in terms of inequalities between the coefficients of the matrix.