Convergence of summation-by-parts finite difference methods for the wave equation (Q1704778)

It is rather well known that when a finite difference method is employed to solve a time dependent partial differential equation, the truncation error is frequently larger at a few grid points near a boundary or a grid interface than in the interior. Also, computationally, the observed convergence rate is higher than the order of the large truncation error. The authors study this phenomenon applied to the second order wave equation. The discretization is achieved by a finite difference operator satisfying a summation by parts property, while the boundary and grid interface conditions are imposed weakly by the simultaneous approximation term method. Again, it is well known that, if the semi-discretized wave equation satisfies the so-called determinant condition, then two orders are gained from the large truncation error localized at a few grid points. The authors show that an energy estimate does not automatically imply an optimal convergence rate; the determinant condition is not necessary for an optimal convergence rate; if there is an energy estimate, where however the determinant condition is not satisfied, then there can be an optimal gain of order \(2\) or a non-optimal gain of order \(1\), or only order half and, in certain cases, one may obtain a super-convergence. Numerical experiments support the derived theoretical accuracy analysis.