中文摘要：

对于线性Hamilton系统,辛差分方法可以保持系统的辛结构,有限元方法可以保证系统的辛性质并具有能量守恒特性。但辛差分方法和有限元方法时域上仍然存在相位误差,使得计算的精度不是很理想。提出极小化相位误差加权间断有限元辛方法（WDG-PF）,该方法是辛方法,同时,对Hamilton系统的求解具有极小的相位误差。数值显示该方法可以保证Hamilton系统的能量守恒性。WDG-PF方法解决了时间有限元方法（TFE）存在的相位漂移现象,同时指出间断有限元方法可以通过加权处理达到保辛要求。WDG-PF方法相较于针对相位误差设计的计算格式分数步对称辛算法（FSJS）、辛Runge-Kutta-Nystrom（RKN）格式以及辛分块Runge-Kutta（SPRK）等方法,WDG-PF显著地减少相位误差,和显著提高Hamilton系统能量精度的优点。相位误差和能量误差几乎达到计算机精度。同时单元内部具有超收敛现象。特别针对高低混频Hamilton系统,传统方法很难在固定的步长下同时实现对高频和低频信号的精确仿真,WDG-PF方法则可以在大步长下同时实现对低频信号和高频信号的高精度仿真。数值显示,WDG-PF方法切实有效。

英文摘要：

Symplectic finite difference method （FDM） can keep the symplectic structure, and finite element method （FEM） can keep the symplectic structure as well as energy conservation for linear Hamiltonian systems. However, symplectic FDM and FEM still have phase errors for the numerical solution, so, the com- putational accuracy is not very well in time domain analysis. Symplectic weighted discontinuous Galerkin method with minimal phase-lag （WDG-PF） is proposed for Hamiltonian systems. This method is symplectic and can highly decrease the phase error, compared to traditional method for Hamiltonian systems. Meanwhile, WDG-PF can keep the conservation of energy as well as the symplectic structure of Hamiltonian systems. WDG-PF can solve the phase-lag problem of continuous Galerkin method, and WDG is symplectic by the tech- nique of weight. Compared to symmetric symplectic （FSJS） algorithm, Runge-Kutta-Nystrom （SRKN） and symplectic partitioned Runge-Kutta （SPRK） methods which are aimed at increasing the accuracy of phase er- ror, WDG-PF ismuch more accurate and increase the energy accuracy of Hamiltonian systems, tremedously. The phase error and Hamiltonian function error almost achieve the accuracy of computer. WDG-PF has the ul- traconvergence point in each element. Especially, for the systems with high and low frequency signals, and seldom has a method can simulate the high and low frequency signals with a fixed time step, WDG-PF can effectively simulate the high and low frequency signals with large time step. The numerical experiments show its validity.