Control of local error stabilizes integrations (Q1917834)

Consider explicit Runge-Kutta methods for initial value problems for a system of ordinary differential equations. At each step the error is estimated by \(\text{est}_n = \widehat y_{n+1} - y_{n+ 1}\), where \(\widehat y_{n+1}\) is obtained using a higher-order formula, and the stepsize is selected as necessary to keep the estimated error smaller than a given tolerance. This paper deals with the interrelation between the growth behavior of the error estimator and the stability behavior of the numerical method. New results are proved, which broaden and deepen the theory.