[lammps-users] PPPM Errors

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS Users

I was wondering if you could help me out with a problem that I cannot seem to get rid of. I am attempting to simulate a clay mineral which has a small polymer and a large amount of water in its interlayer. The whole system is typed in the ClayFF forcefield and has been well minimized using Accelrys Materials Studio. The aim is to run the simulation in th NPT ensemble for 1.5 ns after equilibrating from T = 0 K to 300 K over 250 ps (timestep 0.25 fs 100000 steps). So far I have been almost entirely unsuccessful in running this simulation due to the "Out of range atoms - cannot compute PPPM" error. This error occurs not during the equlibration (heating) phase but after 150000 timesteps of the final production run (timestep 0.5 fs). If I attempt to increase the length of time the equilibration (heating) phase runs for to 500 ps (timestep 0.25 fs 200000 steps) then the simulation fails on the first step of this equilibration phase…

I have made changes to the skin thickness and altered the frequency of the neighbour list update to every timestep, both of these steps have not improved the situation. However, I have been able to get a sensible run by excluding the equilibration (heating) phase and running the production phase with no heating steps. This seems odd to me as surely would leave the system more open to errors? Also I have been able to get the system to behave by using the kspace style ewald, my ultimate goal is to simulate large systems so this is not a time effective option. If I run failed jobs from their restart files for a short amount of time (100000 timesteps) with kspace style ewald then re-run the resulting restart file with kspace style pppm I seem to be able to get a stable simulation…

Does anyone have any ideas why I cannot simply run a kspace style pppm job from start to end with no errors…?

I have attached my input file and data files for this run

Thanks for your time

Rick Anderson

in.GDM_Jeff (2.27 KB)

GDM_test.lammps05 (397 KB)

2

Rick,

Looks to me like you’re getting wild volume fluctuations during your “heating up” phase. This is bad. To get rid of it, you could use a less aggressive pressure control during that phase (a larger damping constant), or simply do NVT during your heating phase. In my opinion, the heating phase is not necessary and you should get rid of it entirely.

Good luck,

Paul

3

This error is usually always a symptom that something else is flawed
in your simulation. B/c one or more atoms is moving a long distance
(more than half the skin distance in a single step) and cannot be mapped
to the PPPM grid by a processor. The trick is to figure out why that
is happening. If you instrument the error message and print out
the coords of the offending atom, maybe it will give you a clue. Viz
also helps. The message Paul sent is a good suggestion - big volume
fluctuations (in a single step) will cause it and they are unphysical.

Steve