Convergence rates for dispersive approximation schemes to nonlinear Schrödinger equations (Q715667)

The authors investigate the convergence rates of numerical approximations for the Schrödinger equation on the real line. They distinguish between dispersive schemes, which fulfill the Strichartz estimation in a discrete sense, and non-dispersive schemes. It is proven in several estimations that the latter ones suffer from a loss of accuracy in the convergence rates, especially if the initial data is not sufficiently smooth. More precisely, the authors show that approximating the linear Schrödinger equation by a dispersive numerical scheme yields roughly the same error order (in a certain norm) as a classical consistency and stability analysis. This dispersive approximation of the linear case is then extended to the nonlinear equation. The resulting schemes are proven to exhibit a polynomial order of convergence, whereas the convergence rate of non-dispersive schemes significantly depends on the smoothness of the initial data. This is done using several technical results which are also verified in this paper.