Hi all,
I was wondering if someone had any insight as to why a simulation would be stable during a box/relax minimization and NVT ensemble, but then immediately crash during an NPT ensemble. The errors I get during the NPT run are typical of an unstable simulation: segmentation fault, and non-numeric pressure. But the structure and simulation are completely stable during NVT.
My only guess is that it is due to the reaxFF potential used, however, I know that others have had no issue using the same reactive force field for npt simulations of alpha quartz. I have attached an input and output file. If you have any insight please let me know.
Thanks,
Benjamin Geller
in.quartz.reax.eq (941 Bytes)
quartz.hex2.job.out (67.5 KB)
1 Like
Hi all,
I was wondering if someone had any insight as to why a simulation would be
stable during a box/relax minimization and NVT ensemble, but then
immediately crash during an NPT ensemble. The errors I get during the NPT
run are typical of an unstable simulation: segmentation fault, and
non-numeric pressure. But the structure and simulation are completely stable
during NVT.
My only guess is that it is due to the reaxFF potential used, however, I
know that others have had no issue using the same reactive force field for
npt simulations of alpha quartz. I have attached an input and output file.
If you have any insight please let me know.
another possible cause is a bad choice of the time relaxation constant
for pressure coupling for your system.
you have chosen 1fs with a time step of 0.1fs. that is way too short.
it is even a bad choice for the thermostat.
you want at least 10fs for Tdamp and 100fs for Pdamp (corresponding to
100 and 1000 time steps, respectively).
often a 10x or even 100x larger time constant than those may be
desirable (for long trajectories of well equilibrated systems)
the best choice depends on the prevalent vibrational states in your system.
axel.
1 Like
Thank you Axel, increasing the time relaxation constant fixed the issue.
Regards,
Ben