Full text loading...
The finite difference scheme has been widely applied in the Laplace-Fourier (L-F) domain forward modeling. The average derivative method (ADM) of the finite-difference scheme is widely adopted, but it represents a special case of the general optimal scheme, which offers better accuracy and efficiency. However, the classical general optimal scheme suffers from inaccuracy at low wavenumbers and is not rigorously convergent. In this study, we propose an accuracy-constrained general optimal 25-point scheme for the scalar wave equation in L-F domain. Based on the optimized scheme, we calculate the accuracy-constrained general optimized 25-point difference coefficients for various spatial sampling intervals. Dispersion analysis shows that, to maintain relative error within 1%, the average derivative 25-point scheme requires 5 grid points per smallest wavelength and pseudo-wavelength, while the accuracy-constrained general optimal 25-point scheme only requires 4 grid points. To suppress the boundary reflection, the perfectly matched layer boundary is employed. Numerical results demonstrate that the accuracy-constrained general optimal 25-point scheme shows better accuracy than the average derivative 25-point scheme.