On accuracy conditions for the numerical computation of waves
A Bayliss,C.I Goldstein,E Turkel
DOI: https://doi.org/10.1016/0021-9991(85)90119-6
IF: 4.645
1985-07-01
Journal of Computational Physics
Abstract:The Helmholtz equation (Δ+K2n2)u=f with a variable index of refraction n and a suitable radiation condition at infinity serves as a model for a wide variety of wave propagation problems. Such problems can be solved numerically by first truncating the given unbounded domain, imposing a suitable outgoing radiation condition on an artificial boundary and then solving the resulting problem on the bounded domain by direct discretization (for example, using a finite element method). In practical applications, the mesh size h and the wave number K are not independent but are constrained by the accuracy of the desired computation. It will be shown that the number of points per wavelength, measured by (Kh)−1, is not sufficient to determine the accuracy of a given discretization. For example, the quantity K3h2 is shown to determine the accuracy in the L2 norm for a second-order discretization method applied to several propagation models.
physics, mathematical,computer science, interdisciplinary applications