A new class of G(ϵ)-symplectic general linear methods

Abstract

A new class of G(ϵ)-symplectic general linear methods for numerical integration of Hamiltonian systems of differential equations is described. Order conditions for these methods are derived using Albrecht approach and the construction of G(ϵ)-symplectic method is described based on solving minimization problems with nonlinear inequality constrains. Examples of methods up to the order four are presented. Numerical experiments confirm that these methods achieve the expected order of accuracy and that they approximately preserve Hamiltonians of differential systems.