6
$\begingroup$

Consider $Z = K - i \omega S - \omega^2 M$, where: $\omega$ is a positive real number; $K$ is a real, symmetric, and positive semi-definite matrix; $M$ is a real, symmetric, and positive definite matrix.

$S$ is a non-zero matrix.

Which property (or properties) $S$ should satisfy so that $Z$ is invertible? Or so that $(Z, M)$ has an eigendecomposition $Z = M \Phi \Lambda \Phi^T M$ (with $\Phi^T M \Phi = I$)?

Or which reference book discusses this class of problems?

Remark 1: Matrices like $Z$ arise in the discretization of the Helmholtz equation with absorbing boundary conditions or of the frequency response analysis with damping.

Remark 2: Many posts on the forum discuss the matrix $\left[ \begin{array}{cc} 1 & i \\ i & -1 \end{array} \right]$ as an example of non-diagonalizable complex symmetric matrix. Here I want to know which non-zero matrix $S$ would work.

EDIT: Generalization of proportional damping gives us a family of $S$ matrices that would allow the eigendecomposition: $S = M \sum_{j = J_1}^{J_2} s_j \left( M^{-1} K \right)^j$

$\endgroup$
4
  • $\begingroup$ Can $K$ or $M$ be zero? $\endgroup$
    – Leo
    Jul 30, 2019 at 20:44
  • $\begingroup$ @Leo $K$ or $M$ can not be zero. Eventually, we could consider: when does $W = K - i \omega S$ have an eigendecomposition $W = \Psi \Theta \Psi^T$ with $\Psi^T \Psi = I$ ? $\endgroup$
    – user7440
    Jul 30, 2019 at 21:27
  • $\begingroup$ We can reduce this to the case in which $M = I$ by noting that $$ M^{-1/2}ZM^{-1/2} = \tilde K - i \omega \tilde S - \omega^2 I $$ where $\tilde K = M^{-1/2} K M^{-1/2}$ is positive semidefinite. We can then further reduce your problem to the case where $\tilde K$ is diagonal. $\endgroup$
    – Ben Grossmann
    Jul 30, 2019 at 22:00
  • $\begingroup$ @Omnomnomnom I agree. $\endgroup$
    – user7440
    Jul 30, 2019 at 22:16

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.