A Robust Optimal Finite Difference Scheme for the Three‐Dimensional Helmholtz Equation

Abstract

We propose a robust optimal 27-point finite difference scheme for the Helmholtz equation in three-dimensional domain. In each direction, a special central difference scheme with 27 grid points is developed to approximate the second derivative operator. The 27 grid points are divided into four groups, and each group is involved in the difference scheme by the manner of weighted combination. As for the approximation of the zeroth-order term, we use the weighted average of all the 27 points, which are also divided into four groups. Finally, we obtain the optimal weights by minimizing the numerical dispersion with the least-square method. In comparison with the rotated difference scheme based on a staggered-grid method, the new scheme is simpler, more practical, and much more robust. It works efficiently even if the step sizes along different directions are not equal. However, rotated scheme fails in this situation. We also present the convergence analysis and dispersion analysis. Numerical examples demonstrate the effectiveness of the proposed scheme.

The Improved Element-Free Galerkin Method for 3D Helmholtz Equations

“…It is shown that the numerical results are in good agreement with the analytical ones. The third example [68] is…”

“…The CPU times of the IEFG method and the EFG method are 200.6 s and 208.1 s, respectively. The third example [68] is ∆u + k 2 u = (k 2 − 3π 2 ) cos(πx 1 ) sin(πx 2 ) sin(πx 3 ).…”

An Optimal Fourth-Order Finite Difference Scheme for the Helmholtz Equation Based on the Technique of Matched Interface Boundary

“…In order to suppress the "pollution effect" and reduce the numerical dispersion, many numerical methods have been proposed in the past decades (cf. [1,2,[15][16][17][18][19]).…”

“…In [6], an optimal 9point difference scheme is developed, which is consistent with the Helmholtz equation with perfectly matched layer (PML). To improve the robustness, the authors of [7,16] established robust optimal difference schemes which works efficiently even if the step sizes along different directions are not equal. To reduce the numerical error, higher-order finite difference schemes (cf.…”