PARTITIONING STRATEGIES IN RUNGE-KUTTA TYPE METHODS

The numerical solution of ode's suffers from stability constraints, if there are solution components with different time constants. Two recommended approaches in software packages to handle these difficulties (i) automatic switching between stiff and nonstiff methods and (ii) the use of partitioned methods for a given splitting into stiff and nonstiff subsystems, are presented and investigated for Runge-Kutta type methods. A strategy for a dynamic partitioning in these methods is discussed. Numerical results illustrate the efficiency of partitioning.