NVT simulations on recent versions of lammps

Please reply-all so that this stays in the loop, thanks.

Why don't you post the input script that reproduces the "some sort of
catastrophic error that the system either ejects atoms or collapses
them" and "simulation box explodes" under NVT ensemble?

Cheers,
Ray

sure thing

thanks

waterbox.data (126 KB)

waterbox.inp (3.49 KB)

Matthew,

Your input script is flawed in three places:

1. "fix 2 all nvt 0.0 300.0 100.0" This is not a correct
command. You should have seen an error complaining "Illegal fix
nvt/npt/nph command". I am not sure how you even ran this script.

2. A temperature of absolute 0 K is never a good idea in MD. Please
refer to the doc page and correct the above fix nvt command, then
change 0.0 to 0.01 (which is close enough to absolute zero). You will
see NVT works as it should.

3. "fix 3 all npt 300.0 300.0 100.0 xyz 0.0 0.0 1000.0" is
also incorrect.

After making changes, I think the script is running perfectly normal.

Cheers,
Ray

Ray,

As you might have guessed the input file was from the version of lammps ~2years ago. I seemed to have missed a major revision. I am going back and forth between the current and old versions to figure out the problem.
Why did this work:
fix 2 all nvt temp 0.0 300.0 100.0
in the old version, but now we need:
fix 2 all nvt temp 0.01 300.0 100.0

to correct for absolute zero.

I have also noticed that all of the example files show significant changes from the provided output. About 2/3 of them showed pressure changes on the order of 10%. Is this related? Even odder there was different thermochemistry on a job run with lammps distributed not two months apart.
I guess I am asking the general question of what is different.

thanks very much
matthew

Matthew,

Why did this work:
fix 2 all nvt temp 0.0 300.0 100.0
in the old version, but now we need:
fix 2 all nvt temp 0.01 300.0 100.0
to correct for absolute zero.

Hmm, I am not sure. But 0.0 is almost always a bad number for temperature.

I have also noticed that all of the example files show significant changes from the provided output. About 2/3 of them showed pressure changes on the order of 10%. Is this related? Even odder there was different thermochemistry on a job run with lammps distributed not two months apart.
I guess I am asking the general question of what is different.

I am not sure as well, for this depends on the modifications of the
pair_style, bond_style, angle_style and kspace_style that you are
using over the past two years. You will have to dig a lot deeper to
figure out where the differences originate.

Ray

Hi

I am not sure as well, for this depends on the modifications of the
pair_style, bond_style, angle_style and kspace_style that you are
using over the past two years. You will have to dig a lot deeper to
figure out where the differences originate.

The differences are probably not related to the forces but to the thermostat itself. At some point in the past the whole stack of thermo/barostats changed underneath in some subtle way to support some new stuff like this chain thermostating and some corrective adjustments. The current default values do NOT reproduce the original Nose Hoover thermostat, which was what LAMMPS did before.

I think the main thing you might need to change to get the old behaviour is providing: tchain 1 mtk no
But please read the documentation again. Ah and I don't mean to imply that it would be better to use the original Nose Hoover thermostat, there is a reason after all for the changes which were made.

Cheers
Christian

-------- Original-Nachricht --------