A phase-preserving and low-dispersive symplectic partitioned Runge-Kutta method for solving seismic wave equations

被引:25
|
作者
Ma, Xiao [1 ]
Yang, Dinghui [2 ]
机构
[1] Northwestern Polytech Univ, Dept Appl Math, Xian 710072, Shaanxi, Peoples R China
[2] Tsinghua Univ, Dept Math Sci, Beijing 100084, Peoples R China
基金
中国国家自然科学基金;
关键词
Numerical solutions; Numerical modelling; Wave propagation; ANALYTIC DISCRETE OPERATOR; REVERSE-TIME MIGRATION; FINITE-DIFFERENCE; DISCONTINUOUS-GALERKIN; OPTIMALLY ACCURATE; NYSTROM METHODS; SCHEME; ELEMENT; MEDIA; PROPAGATION;
D O I
10.1093/gji/ggx097
中图分类号
P3 [地球物理学]; P59 [地球化学];
学科分类号
0708 ; 070902 ;
摘要
The finite-difference method, which is an important numerical tool for solving seismic wave equations, is widely applied in simulation, wave-equation-based migration and inversion. As the seismic wave phase plays a critical role in forward simulation and inversion, it should be preserved during wavefield simulation. In this paper, we propose a type of phase-preserving stereomodelling method, which is simultaneously symplectic and low numerical dispersive. First, we propose three new time-marching schemes for solving wave equations that are optimal symplectic partitioned Runge-Kutta schemes with minimized phase errors. Relevant simulations on a harmonic oscillator show that even after 200 000 temporal iterations, our schemes can still avoid the phase drifting issue that appears in other symplectic schemes. We use these symplectic schemes as time integrators, and a numerically low dispersive operator called the stereomodelling discrete operator as a spatial discretization approach to solve seismic wave equations. Theoretical analysis on the stability conditions shows that the new methods are more stable than previous methods. We also investigate the numerical dispersion relations of the methods proposed in this study. To further investigate phase accuracy, we compare the numerical solutions generated by the proposed methods with analytic solutions. Several numerical experiments indicate that our proposed methods are efficient for various models and perform well with perfectly matched layer boundary conditions.
引用
收藏
页码:1534 / 1557
页数:24
相关论文
empty
未找到相关数据