Discretizations of nonlinear differential equations using explicit finite order methods (Q1298494): Difference between revisions

The paper is devoted to the study of results of experimental computations of solutions of a certain class of ordinary differential equations dependent on a parameter. Computations made up to large values of the independent variable allow to observe asymptotic behaviour of the approximate solution, which often is different than asymptotic behaviour of the exact one. The authors are interested in application to this end of explicit methods of Runge-Kutta type. Observations of results of computations allow to formulate certain suggestions how to apply such explicit methods in order to avoid, or at least diminish, this asymptotical disaccordance. Remarks: The Reviewer's suggestion is to try some simple implicit method, as for instance so called ``trapesoidal rule'' \[ x_{k+1}=x_k+{h\over 2}(f_k+f_{k+1}) \] which is A-stable and of second order. This may be of a certain importance for computing on large distances. Moreover the right-hand-side of the ``trapesoidal rule'' is very similar to this of the original equation.