Acceleration of implicit schemes for large systems of nonlinear ODEs (Q964098)

The authors study the acceleration of implicit schemes for large systems of nonlinear ordinary differential equations (ODEs). They describe the subspace of approximation and provide estimates for the error between the exact and the approximate solution of the implicit scheme. In view of the line search and trust region algorithm, a method of computing a good descent direction and a good initial guess for each linear system is given. The algorithmic aspect of the proposed approach and the numerical effectiveness on a very stiff problem is given. The generalized minimal residual (GMRES) Krylov subspace method is used in the discussions.