grandes-ecoles 2013 QIV.A.4
Compute eigenvectors or eigenspaces
View
In this section we consider a circle $\mathcal{C}(\Omega, r)$ with center $\Omega$ and non-zero radius $r$, intersecting the $x$-axis. We denote by $L_1$ and $L_2$, with coordinates respectively $(\lambda_1, 0)$ and $(\lambda_2, 0)$, with $\lambda_1 < \lambda_2$, the two intersection points of $\mathcal{C}(\Omega, r)$ with the $x$-axis. Let $A = \left(\begin{array}{ll} a & b \\ c & d \end{array}\right)$ be a matrix whose eigenvalue circle equals $\mathcal{C}(\Omega, r)$. We keep the notations $E, F, G, H$ from III.D.
Show that the square of the cosine of the angle between two eigenvectors of $A$ associated with two distinct eigenvalues is determined by the circle $\mathcal{C}(\Omega, r)$, and does not depend on the choice of a matrix $A$ whose eigenvalue circle equals $\mathcal{C}(\Omega, r)$ (one may, if deemed useful, introduce the orthogonal projection of $\Omega$ onto the $x$-axis). What about if $A$ is symmetric?