Sudden very large increase in temperature

LAMMPS LAMMPS Mailing List Mirror
1

Hi Everyone,

I am facing a problem in the NpT lammps simulations. What I am doing is using neural network potentials in the lammps simulations. After 4000 steps equilibration of the system with NVE and NVT (300K), in the NpT (300K; 1000 bar) part, the temperature suddenly gets very large and the simulation crushes:

Step PotEng KinEng TotEng Temp Press Density
. . . . . . .

. . . . . . .
4270 -237815.61 17.462146 -237798.15 308.43161 1062.1615 0.27194647
4280 -237815.59 17.292583 -237798.3 305.43662 1044.0842 0.27186027
4290 -237815.59 17.135646 -237798.46 302.66467 1018.7309 0.27177333
4300 -237815.61 16.993896 -237798.62 300.16096 975.45397 0.27168609
4310 -237815.64 16.847907 -237798.79 297.58238 954.57299 0.27159913
4320 -237992.12 263.40293 -237728.72 4652.4515 -45.235137 0.27152551
4330 -237979.28 473.96565 -237505.31 8371.5932 -1246.8783 0.2714544

4340 -227541.45 -nan -nan -nan -nan -nan

The simulation box is set according to the density of the system. Attached is the input file. It would be greatly appreciated if someone can help me to solve this problem. Many thanks.

Best,
Mostafa

md.lmp (2.04 KB)

2

these kind of things tend to happen, when something goes wrong with the forces or the timestep is too large for your system and the curvature of your potential energy surface.

if the error is easily reproducible, the first step is usually to write out a trajectory with frequent output around the time of the crash and closely examine that in a visualization program. that can give you an idea of what kind of process is happening. that in turn would them determine what to do to address the divergence.

axel.