On spurious fixed points of Runge-Kutta methods (Q1357361): Difference between revisions

This paper deals with the existence and generation of spurious fixed points in the numerical integration of autonomous scalar differential equations \(u'=f(u)\) by means of Runge-Kutta methods. Since the numerical method can be written in the form \(u_{n+1}= u_n+ h\phi(u_n,h)\), where \(\phi\) is the increment function, it can be considered as a dynamical system parameterized by the time step \(h\). In this context, the author applies techniques from bifurcation theory [cf. \textit{S.-N. Chow} and \textit{J. K. Hale}, Methods of bifurcation theory (1982; Zbl 0487.47039)] and singularity theory [cf. \textit{M. Golubitsky} and \textit{D. G. Schaeffer}, Singularities and groups in bifurcation theory. Volume I (1985; Zbl 0607.35004)] to understand the generation and existence of spurious fixed points when some explicit and implicit Runge-Kutta methods are applied to the logistic equation. Thus, it is proved that all spurious fixed points of the logistic equation obtained by \textit{D. F. Griffiths}, \textit{P. K. Sweby} and \textit{H. C. Yee} [IMA J. Numer. Anal. 12, No. 3, 319-338 (1992; Zbl 0761.65056)] for some explicit Runge-Kutta methods bifurcate from the fixed points of the logistic differential equation. Further, the behaviour of some implicit methods (specially of SDIRK type) is studied by introducing the new concepts of B-, R- and BR-regularity related to the well-known concept of regularity given by \textit{A. Iserles} and \textit{S. P. Nørsett} [IMA. J. Numer. Anal. 10, No. 4, 463-488 (1990; Zbl 0712.65071)].