scientific article; zbMATH DE number 1389065
The real homogeneous polynomial \(f=f(t, x_1,\dots,x_m)\) of degree \(n\) in \(m+1\) indeterminates \(t, x_1,\dots, x_m\) is said to be hyperbolic with respect to \(\theta \in {\mathbb R}^{m+1}\) if \(f(\theta)\neq 0\) and if, for every fixed \(\xi \in{\mathbb R}^{m+1}\), the equations \(f(t\theta +\xi)=0\) has \(n\) real zeros (counting the multiplicity of each zero). There is no loss of generality in assuming that \(\theta =\theta_0=(1, 0,\dots, 0)\) and \(f(\theta_0)=1\). The polynomial \(f\) can be factorized with respect to the indeterminate \(t\): \[ f(t, x_1,\dots, x_m)= (t+\psi_1(x_1,\dots, x_m))(t+\psi_2(x_1,\dots, x_m)) \cdot\dots\cdot(t+\psi_n(x_1,\dots, x_m)), \] where the functions \(\psi_j\) are continuous on \({\mathbb R}^m\), positive-homogeneous and satisfy the inequalities \[ \psi_1(x_1,\dots, x_m) \geq \psi_2(x_1,\dots, x_m)\geq\dots\geq \psi_n(x_1,\dots, x_m). \] A theorem of Gårding states that the function \(\psi_1\) is convex on \({\mathbb R}^m\). The author gives the following generalization of this theorem: Suppose that \(f\) is a real homogeneous polynomial hyperbolic with respect to \(\theta_0\) and \(f(\theta_0)=1\). Suppose that \(c_1, c_2,\dots, c_n\) are real numbers satisfying \(c_1\geq c_2 \geq\dots \geq c_n\). Then the function \(\sum^n_{j=1} c_j\psi_j \) is convex on \({\mathbb R}^m\). The proof is based on an elementary theory of symmetric polynomials. An example of application of this theorem to the theory of generalized numerical ranges of matrices is also given.
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4933876)
