Hi,
is there anyway to choose the algorithm for integration at each timestep in LAMMPS?
Regards.
Prithwish
Prithwish:
Why would you want to change the integrator after each timestep? From a methodological perspective, this makes absolutely no sense, as this means you’ll never collect enough statistics from a consistent ensemble, which means you learn nothing from the simulation.
Even if you’re doing some sort of hybrid MD-Monte Carlo method, I still don’t see how changing the integrator after every timestep helps.
–AEI
Dear A Ismail,
I didn't want to mean to change the integrator algorithm at each timestep. May be my language is not good enough to communicateit properly. I wanted to tell, can we choose the algorithm (e.g. velocity verlet etc.) for integration in LAMMPS? Or, is it fixed ?
Regards.
Prithwish
OK. Now your question makes sense: LAMMPS has velocity Verlet and RESPA algorithms available. See http://lammps.sandia.gov/doc/run_style.html for details.
–AEI
Everything in LAMMPS is really velocity-Verlet based (including rRESPA).
If you wanted to write fixes that did something different, like a
fix nve/euler or fix nve/runge/kutta, you could do it - but you'd
have to live with where the forces are called/computed during the timestep.
Steve