ABSTRACT
The better codes based on Runge–Kutta and Adams methods are generally very effective, but they are inefficient when solving a class of problems called stiff, so inefficient that they are impractical for these problems. Constant step size implementations of the methods produce numerical results that oscillate and grow explosively. Implementations that vary the step size so as to control local errors can solve stiff problems, but they use step sizes that seem ridiculously small. This behavior is both puzzling and frustrating.
