On error growth functions of Runge-Kutta methods (Q5961741)

The aim of this paper is to get more insight into the error growth function. Estimates of the form \(|y_1- \widehat y_1 |\leq \varphi(hv) |y_0- \widehat y_0 |\) are studied, with \(y_1, \widehat y_1\) the numerical solutions of a Runge-Kutta method applied to a stiff differential equation satisfying a one-sided Lipschitz condition (with constant \(v)\). An explicit formula for the optimal function \(\varphi(x)\) is given, and it is shown to be superexponential, i.e., \(\varphi(x_1) \varphi(x_2) \leq\varphi (x_1+ x_2)\) if \(x_1\) and \(x_2\) have the same sign. As a consequence, results on asymptotic stability of the numerical solution are obtained. Upper bounds for \(\varphi(x)\) are presented that can be easily computed from the coefficients of the method.