Symplectic Exponential Runge-Kutta Methods for Solving Nonlinear Hamiltonian Systems
Lijie Mei,Xinyuan Wu
DOI: https://doi.org/10.1016/j.jcp.2017.03.018
IF: 4.645
2017-01-01
Journal of Computational Physics
Abstract:Symplecticity is also an important property for exponential Runge–Kutta (ERK) methods in the sense of structure preservation once the underlying problem is a Hamiltonian system, though ERK methods provide a good performance of higher accuracy and better efficiency than classical Runge–Kutta (RK) methods in dealing with stiff problems: y′(t)=My+f(y). On account of this observation, the main theme of this paper is to derive and analyze the symplectic conditions for ERK methods. Using the fundamental analysis of geometric integrators, we first establish one class of sufficient conditions for symplectic ERK methods. It is shown that these conditions will reduce to the conventional ones when M→0, and this means that these conditions of symplecticity are extensions of the conventional ones in the literature. Furthermore, we also present a new class of structure-preserving ERK methods possessing the remarkable property of symplecticity. Meanwhile, the revised stiff order conditions are proposed and investigated in detail. Since the symplectic ERK methods are implicit and iterative solutions are required in practice, we also investigate the convergence of the corresponding fixed-point iterative procedure. Finally, the numerical experiments, including a nonlinear Schrödinger equation, a sine-Gordon equation, a nonlinear Klein–Gordon equation, and the well-known Fermi–Pasta–Ulam problem, are implemented in comparison with the corresponding symplectic RK methods and the prominent numerical results definitely coincide with the theories and conclusions made in this paper.