[lammps-users] Runge-Kutta

Dear All,

Is there any feature in LAMMPS to use Runge-Kutta method for time
integration? Otherwise, is there any way to implement it?

Kind regards,
Mehdi

There is no RK in LAMMPS. You could try to implement it as a fix,
but the forces may be computed at incorrect times during the
timestep, since it is setup as a Verlet step. You could replace
verlet.cpp with your own rk.cpp, but that would be very non-trivial.

Steve

Dear All,

Is there any feature in LAMMPS to use Runge-Kutta method for time
integration? Otherwise, is there any way to implement it?

dear mehdi,

is there a good reason to use it?
it is generally regarded as highly inefficent
for MD since it requires many evaluations of
the forces per step.

cheers,
   axel.

Dear Axel,

Thank you for your interest in the subject. Really, there is a sort of
problem in which a high accuracy is needed. In those cases we should
either use very short time step or use higher order algorithm ( more
than the order of velocity Verlet algorithm). Also in some cases in
which the force field is a function of velocity (may be interested for
several problems in granular style), we could not use velocity Verlet
algorithm, and we have to use another time integrator.

I was thinking of implementing such an integrator in LAMMPS. Really,
it is not the case of Runge-Kutta, it could be another accurate
scheme.

I will be glad to receive more comments.

Kind regards,
Mehdi