Meaning of zero pressure in LAMMPS's fix npt

LAMMPS LAMMPS General Discussion
1

I want to calculate the viscosity of silica at standard pressure and 2200 K through MD. In the equilibration process for the same I am using the following the following fix

fix NPT all npt temp 2200.0 2200.0 100.0 iso 1.0 1.0 1000.0

Am I on the right track ?

I have seen instances of people using

fix NPT all npt temp 2200.0 2200.0 100.0 iso 0.0 0.0 1000.0

What does this mean ? A clarification would be helpful. Thanks in advance.

2

For solids with low compressibility and at the system sizes available to MD simulations, the difference between 1 atm and 0 atm pressure is negligible. The instantaneous pressure will fluctuate significantly so maintaining an average pressure that is sufficiently statistically converged would take a crazy amount of computational effort for basically no gain. The systematic error in using a classical potential is much larger.

1 Like
3

Thank you for your reply.

Even though I expect my system to be at liquid state at 2200 K, you are correct. The fluctuations are indeed large and the system takes a long time to reach the said pressure.

I wanted to ask a couple of more questions

  1. Does setting 0 atm pressure mean that we are carrying out the simulation at vacuum ?

  2. Is there a better way to do the same, where in the computational effort involved maybe reduced ? I do not know the correct density of the systems at the required temperatures, thus I chose NPT instead of NVT.

4

No. With fix npt the effect of the target pressure is that the integrator will shrink the system if the instantaneous pressure is smaller and expand the system if it is larger. Thus a 0 target pressure would equilibrate to a state without an external pressure. As mentioned before, in this context the difference between 1 atm and 0 atm is negligibe.

Then you can do what experimentalists are doing, and fit the results from various points with fixed volume and controlled temperature to a suitable equation of state. Please take note that this will give you the (approximate) equation of state for the parameterization/model, which may be different from experimental data, if the parameters are not transferable to the state you are interested in.