On the boundary of the numerical range of a matrix (Q972961): Difference between revisions

Every diagonal element of a~matrix~\(A\) is known to lie in the numerical range \(W(A)\) of~\(A\). A~characterization of real \(2\times 2\) matrices~\(A\) for which a~diagonal entry is a~boundary point of its numerical range is obtained, and is as follows: the diagonal entry~\(a_{11}\) is a~boundary point of the numerical range of~\(A\) if and only if \(a_{12}+a_{21}=0\), and \(a_{12}=a_{21}\) if \(a_{11}=a_{22}\). For general \(n\times n\) matrices, the following result is obtained: If~there exists an~\(i\) such that \(a_{ii}\in \text{bd} W(A)\), then \(a_{ij}+a_{ji} =0\) for all \(1\leq j\neq i \leq n\); the converse of this result being false. Also, given a~matrix~\(A=(a_{ij})\in M_n({\mathbb R})\), if \(x\) and~\(y\) are real orthonormal vectors such that \(x^{*}Ax\in \text{bd} W(A)\), then \(x^{*}Ay + y^{*}Ax = 0\).