The numerical method

Consider an adaptive Runge Kutta method of order 2 for the initial-value problem:

displaymath55

In our program, rk2adp, we use the second order modified Euler method

displaymath56

together with the third order error monitor (Kutta's method)

displaymath57

where tex2html_wrap_inline67 and tex2html_wrap_inline69 Notice that tex2html_wrap_inline71 and tex2html_wrap_inline73 are used in the modified Euler method and hence the error monitor requires only one additional function evaluation. It can be shown that the local error in the modified Euler method is approximately equal to

displaymath58

We accept the approximation given by modified Euler if tex2html_wrap_inline75 , where T denotes the final value and tol is a specified absolute error tolerance. Thus the criterion for an acceptable computed value is

displaymath59

If the computed value is unacceptable we recompute with a new stepsize given by tex2html_wrap_inline81 where

displaymath60

The Matlab file, rk2adp.m , in zipped form can be downloaded.


Next: Examples Previous: The Model Problem


J. Thomas King
Fri Aug 2 11:22:03 EDT 1996