Local Pressure in Lammps (normal and lateral)

Dear all,

I am using lammps-9Dec14. I am simulating methane molecules (modelled as single site LJ molecules) confined in a graphene slit-like channel of width 7.62nm at 300K. I want to calculate the local pressure profiles (normal and lateral) along the z direction (perpendicular to the graphene walls). I did this with aid of compute peratom command in Lammps and dump the results using following commands:

compute peratom methane stress/atom methT
dump 2 methane custom 500 dump_diag_stress.dat id type x y z fz c_peratom[1] c_peratom[2] c_peratom[3]

I post-proccessed the dump file results in slit bins of width 0.1 nm with x and y dimensions parallel to the graphene walls. For normal pressure, for example, I averaged the -1c_peratoms[3] and divide it by the the volume of each bin. However, the normal pressure turned out to be z-dependent which should be a constant value due to the mechanical equilibrium in z direction. And lateral pressure (0.5(pxx+pyy)) profile is half of what it should be. I looked into the pressure formulation of Lammps and it seems there is no error. I would appreciate it if anyone can help me figure out what’s wrong with pressure profiles from Lammps ( especially normal pressure which is not a constant)

I have attached the plots to this post comparing the pressure profiles to that of Gromacs software.

P.S. density profile of methane molecules are correct, which I guess there is no serious problem with the simulation.



Mohammad Hossein Motevaselian
PhD candidate
3216 Beckman Institute
405 North Mathews Avenue
Urbana, Illinois 61801

2Pl_lammps.eps (47.4 KB)

Pl_lammps.eps (59.9 KB)

Pn_lammps.eps (41.2 KB)

You can verify that compute stress/atom is

giving you something consistent with the pressure

tensor for the whole system, by using the commands

on the compute stress/atom doc page. If that is the

case, then you can verify that the sum of your binned

values (across the entire system) also give something

consistent with the global pressure for the entire volume.

If that’s the case then you’ll have to figure out why the

way you are plotting it does not match what you expect.

It’s just numbers at that point, not anything to do with

LAMMPS is calcullating.