项目名称: 几何数值积分及其在常微分方程和偏微分方程中的应用
项目编号: No.11271357
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 孙雅娟
作者单位: 中国科学院数学与系统科学研究院
项目金额: 50万元
中文摘要: 许多来源于工程和物理中的问题,系统本身具有能量、动量等守恒特征。构造数值算法保持系统的这些守恒特征是解决问题的关键。 这类算法被称为几何数值积分或保结构算法,它遵循的基本准则是:数值算法应尽可能多的保持原系统的本质特征。 本项目主要针对常微分方程系统和偏微分方程系统,研究和构造保持系统相应特征的数值方法。 对哈密顿系统和保能量系统,研究时间有限元;建立时间有限元和保结构算法之间的关系;构造保能量和共轭辛的数值算法。对可积的常微分方程,研究可积离散和保持系统多个守恒律的数值算法。 对多辛哈密顿系统,结合系统的守恒特征研究保能量算法和多辛算法。 特别是对多辛算法, 研究高阶Runge-Kutta方法对偏微分方程的空间离散;研究半离散系统的误差和数值色散。对Maxwell方程研究分裂方法,根据Maxwell方程的结构特点对系统进行不同分裂,研究分裂组合方法的最优系数和稳定性。
中文关键词: 保结构算法;多辛哈密顿系统;守恒律;Vlasov-Maxwell 方程;
英文摘要: Many differential systems in celestial mechanics and molecular models have the conservative quantities. The qualitative properties of the numerical integrator become critical to the success of the numerical simulation. Geometric integrators are the numerical methods which can preserve the qualitative properties associated to the solutions of the original systems. In this project, we study the geometric numerical integrators in the systems of ODEs and PDEs. For Hamiltonian system and energy-preserving system, we study the finite element method in time (TFE) and construct the relationship between TFEs and geometric numerical integrators. For integrable differential equations, we study the integrable discretization and the numerical methods which can preserve several first integrals. Applying the (partitioned) Runge-Kutta method to the Hamiltonian PDEs in space, gives the (explicit)implicit ODEs. By considering the semidiscretized system,we study the order of spatial discretization and numerical dispersion. We also consider the application of geometric numerical integrators to Maxwell's equations. Via the geometric properties, we split the given system and study the composition method of high order. By comparison, we obtain the optimal coefficients for constructing the composition method,and analyze the stability
英文关键词: structure-preserving methods;multi-Hamiltonian system;conservation laws;Vlasov-Maxwell equations;