A symplectic RKN scheme for solving elastic wave equations

Liu Shao-Lin,Li Xiao-Fan,Wang Wen-Shuai,Zhu Tong
DOI: https://doi.org/10.6038/cjg20150422
2015-01-01
Abstract:The construction of low dispersive and strong stable numerical schemes is critical for the investigation of wave propagation. Solving seismic wave equations in the time domain, the dispersion and stability are related to the temporal and spatial discretizations. Here we focus on the temporal discretization and develop a high-efficient time integration scheme for the elastic wave equation. Following the transformation of the elastic seismic wave equation into a Hamiltonian system, an explicit second-order symplectic Runge-Kutta-Nystrom (RKN) scheme is proposed for elastic wave modeling. The order conditions are obtained by the rooted trees theory. To obtain symmetric solutions, we assign a constraint condition to the symplectic coefficients. In order to minimize the error in temporal discretization, we propose a method that compares numerical angular frequency with real frequency, and obtain a constraint equation based on Taylor expansion. A symplectic RKN scheme with minimized numerical dispersion is developed by solving the order conditions associating with constraint equations. Through analyzing the time advance equation, we construct another constraint equation to allow stability limit to attain its maximal value, and then another symplectic RKN scheme with the same symplectic coefficients as the first scheme is developed. The optimization process for determining the coefficients of the RKN scheme is independent of spatial discretizations. For the ease of comparison, we simply choose finite difference method to approximate spatial derivatives. Generally, the numerical dispersion amount is lower than the conventional second-order schemes, and is slightly larger than the high-order schemes. The maximal Courant number for the new scheme is the largest compared to those for the conventional schemes. The largest time increment for the new scheme is nearly two times as large as those for the conventional second-order schemes, and is about 1.5 times as large as those for the high-order ones. In a practical simulation, all these schemes discussed in the paper consume approximately the same computer memory, but the numerical accuracy and computational speed are quite different. The numerical accuracy and complexity of our scheme indicate that it may be a good candidate for the balance of numerical accuracy and complexity. Both theoretical analyses and numerical experiments show that our scheme is superior to conventional symplectic schemes in some aspects including suppressing dispersion and increasing stability range. The optimization procedure proposed in the paper may be used to define the coefficients in other symplectic or non-symplectic schemes. Though striking numerical results are obtained by our scheme with a finite difference operator, a low-dispersive spatial operator is still needed to be combined with our scheme to form a more powerful tool for wave simulations.
What problem does this paper attempt to address?