Computational chaos may be due to a single local error (Q1208867)

This paper is concerned with the effect of (rounding or other computational) errors in the numerical solution of nonlinear ordinary differential equations by means of finite difference methods. It is well known that there are initial value problems and convergent finite difference methods such that the numerical solution really computed shows a completely different behavior from the true solution, in such a case the authors say that this is a ``diverting'' scheme for the initial value problem under consideration. In this paper, after defining what is meant by ``diverting'' finite difference approximations and some remarks on the occurrence and detection of this property, some examples are presented and commented. First, in some orbits of the restricted three-body problem with the canonical Runge-Kutta method of order four. Next, in the Lorentz equations with the same method. The paper ends with some remarks on the effect of diversions in the numerical determination of chaotic sets of solutions of ordinary differential equations.