Linear and non-linear stability for general linear methods (Q1091771)

Stability in a numerical method prevents the growth of the approximate solution of a differential equation for which the solution is bounded. A variety of concepts of stability have been developed. For one-leg methods, two extremes, namely A-stability and algebraic stability together with a spectrum between these extremes are equivalent. For Runge-Kutta methods, A-stability is distinguished from the others. In this paper it is shown that general linear methods distinguish more of these concepts. In particular, A-stability is weaker than weak AN- stability which is weaker than strong AN-stability, while Euclidean AN- stability is equivalent to algebraic stability for such methods. Sufficient conditions for the equivalence of strong AN-stability to algebraic stability are promised in a later paper.