Fluctuation in NPT ensemble and elasitc tensor

Dear LAMMPS developers,

I was trying to calculate the bulk modulus and stiffness tensor of a polymeric system
using volume and strain fluctuation in NPT ensemble. To check the validity of this
approach, I did a bunch of tests on FCC gold. To my surprise, I found that the results
are very sensitive to the pressure damping parameters used.

Using the Au_u3.eam potential provided by LAMMPS, I run several 10 nano-second
simulation of FCC gold at 300K and zero pressure. The number of atoms was 500 and
periodic boundary condition was applied. When the pressure damping parameter in
“fix npt" was set to be 1000 timesteps, the bulk modulus obtained was 449 GPa. When
the damping parameter was set to be 100 timesteps, the bulk modulus obtained was
1137 GPa. The results have “converged” in those simulation respectively, but apparently
the difference is too big to be reasonable. Also, they are very different from the “real”
bulk modulus of gold of this EAM potential, which is about 167 GPa.

I wonder if this is an intrinsic deficiency of NPT ensemble or something is wrong with
LAMMPS. I check the literature and some people were actually able to get reasonable
elastic properties using the fluctuation approach in NPT ensemble. Could you please
give some suggestions?

Thanks a lot!


Hi Daniel,

I suppose you used a 4x4x4 FCC supercell with 512, instead of 500,
atoms? If so, then you certainly can not avoid huge fluctuation in
stresses with such a small cell. You need a much larger supercell.