[lammps-users] ERROR: Non-numeric pressure

Dear all,

I am using the following fix:

fix 5 all npt/cauchy temp {TempHi} {TempHi} (100.0 * dt) x 0.0 0.0 (1000.0 * dt) y 0.0 0.0 $(1000.0 * dt) couple none tchain 5 tloop 100 pchain 5 ploop 100 nreset 1000 alpha 0.001

and I get almost always the following error in the middle of my runs:

ERROR: Non-numeric pressure - simulation unstable (…/fix_npt_cauchy.cpp:1089)

Because of that error, I have to restart my simulations from a restart file to continue my simulations. Am I missing something here?

Thanks,

Reza.

Dear all,

I am using the following fix:

fix 5 all npt/cauchy temp {TempHi} {TempHi} (100.0 * dt) x 0.0 0.0 (1000.0 * dt) y 0.0 0.0 $(1000.0 * dt) couple none tchain 5 tloop 100 pchain 5 ploop 100 nreset 1000 alpha 0.001

I have one concern and one problem regarding the fix above:

  1. Is it okay to use “nreset 1000” for this fix? I mean that should I set it to zero or there is no harm in updating the reference in fx npt/cauchy?

  2. I get almost always the following error in the middle of my runs:

ERROR: Non-numeric pressure - simulation unstable (…/fix_npt_cauchy.cpp:1089)

Because of that error, I have to restart my simulations from a restart file to continue my simulations. I am very confident with my preparation stage, and I follow a very strict scheme to properly equilibrate and relax my samples. But in the production stage, when I want to mechanically loading my samples with full periodic conditions or having shrink-wrapped boundary conditions in one or two directions, I always get this error regardless of having orthogonal or triclinic box. I’m amusing EAM and MEAM potentials for metals, and they are well calibrated, I guess. Am I missing something here?

Thanks,

Reza.

Have you tried running the simplest possible NPT simulation? No chains, no Nreset, etc.
Does it work with those potentials and give good answers?

Once that works, then add one option at a time, e.g. Nreset, and see what happens.
Then you can also see what that keyword changes and whether it is needed for your model.

It’s generally a bad idea to start with a very complicated command or script and then hope to figure out: why doesn’t it work?
My advice is start simple and build up complexity one step at a time.

Steve