Matrix Fejér-Riesz theorem with gaps (Q5964096): Difference between revisions

The matrix Fejér-Riesz theorem is the following result: Let \(F(x)=\sum_{m=0}^{2N}F_mx^m\in M_n(\mathbb{C}[x])\) be an \(n\times n\) matrix polynomial which is positive semidefinite on \(\mathbb{R}\) (i.e. \(v^*F(x)v\geq 0\) for every \(x\in\mathbb{R}\) and every \(v\in\mathbb{C}^n\)). Then there exists a matrix polynomial \(G(x)=\sum_{m=0}^N G_mx^m\in M_n(\mathbb{C}[x])\) such that \(F(x)=G(x)^*G(x)\) where \(G(x)^*=\sum_{m=0}^N \overline{G_m}^Tx^m\). This result has been extended in the case \(n=1\) to arbitrary semialgebraic subsets of \(\mathbb{R}\) in [\textit{S. Kuhlmann} and \textit{M. Marshall}, Trans. Am. Math. Soc. 354, No. 11, 4285--4301 (2002; Zbl 1012.14019)]. The paper under review deals with the multivariate case on arbitrary semialgebraic subsets of \(\mathbb{R}\). For this an equivalent formulation of the above theorem of Fejér and Riesz in terms of \textit{matrix quadratic modules} and \textit{matrix preorderings} is used. In this setting, the author gives new results depending on the geometry of the semialgebraic subset of \(\mathbb{R}\) (like the number of unbounded intervals and the number of isolated points).