Numerical solution of isospectral flows

From MaRDI portal





Isospectral flows are given by a differential equation \(L'= [B(L), L]\), where \(B(L)\) is a skew-symmetric matrix function and \([B, L]\) denotes the commutator of \(B\) and \(L\). They have the remarkable property that the eigenvalues of \(L(t)\) are independent of time. After reviewing some important special cases of isospectral flows, it is proved that the numerical flow of Runge-Kutta methods is in general not isospectral. The reason is that quadratic and cubic invariants cannot be conserved simultaneously. As an alternative approach it is proposed to apply methods, which conserve quadratic invariants, to Flaschka's reformulation \(U'= B(L)U\), \(L= UL_0U^T\). Numerical experiments with the modified implicit midpoint rule are reported.



Cites work


Cited in
(60)








This page was built for publication: Numerical solution of isospectral flows

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4372654)