A low cost Arnoldi method for large linear initial value problems

From MaRDI portal





The author develops a very efficient class of methods for solving large systems of linear ordinary differential equations of the form \[ y'= -Ay+ r(t)v \] by using the Arnoldi method to approximate the exact solution in a Krylov subspace. A method of order \(m-1\) has the form \[ y_{n+1}= k_n(\exp, -hA)y_n+ \sum^{n-1}_{j=0} r^{(j)}(t) h^{j+ 1}k_n(\phi_j, hA)v, \] where \[ \begin{gathered} k_n(f, nA)= V_n f(nH_n) V^T_n,\\ \phi_j(z)= (-1/z)^{j+1}\Biggl(e^{-z}- \sum^j_{k=0} {(-z)\over k!}^k\Biggr)\end{gathered} \] and \(H_n\) is the upper Heisenberg matrix \(H_n= V^T_n AV_n\).





Describes a project that uses

Uses Software






This page was built for publication: A low cost Arnoldi method for large linear initial value problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3155011)