Positive semidefinite interval of matrix pencil and its applications to the generalized trust region subproblems (Q6087885)

Let \(\mathcal{S}^n\) denote the space of symmetric matrices in \(\mathbb{R}^{n \times n}\). Given \(A, B \in \mathcal{S}^n\), the authors study the set \(I_{\succeq}(A,B)\) of real numbers \(\mu\) for which \(A+\mu B\) is positive semidefinite: \[ I_{\succeq}(A,B)=\{ \mu \in \mathbb{R}\ |\ A+\mu B\succeq 0\}. \] The following cases are of interest: (1) If \(A\) and \(B\) are not simultaneously diagonalizable via congruence (SDC), then \(I_{\succeq}(A,B)\) is either empty or a singleton; (2) When \(A\) and \(B\) are SDC, the set \(I_{\succeq}(A,B)\), if nonempty, can be either a singleton or an interval. In the latter situation, in Section 2, the authors derive the interval \(I_{\succeq}(A,B)\), when at least one matrix is nonsingular, and both \(A\) and \(B\) are singular. Understanding the structure of \(I_{\succeq}(A,B)\) holds significance in tackling the \textit{hard-case} of the generalized trust-region subproblem (GTRS). In Section 3 the authors show how this knowledge allows one either a reduction of the problem to solving a linear equation system or to the \textit{easy-case} of a GTRS with a smaller size.