Calculation of the Müller-Plathe lattice TC on systems governed by the MEAM potential in LAMMPS: problems?

Dear users,

We are currently trying to calculate the lattice thermal conductivity of several metals (Ni, Cu, Pd, Ag, Pt and Au) using the non-equilibrium Müller-Plathe method in LAMMPS. We are considering big supercells (20x20x20 unit cells, 70k+ atoms) for each elemental metal separately. The interactions between atoms are mediated by the well-known MEAM potential. The heat flux is tallied using the fix thermal/conductivity command, and temperature profiles are recorded using the compute chunk/atom and fix ave/chunk commands. Please find attached the typical input run for reference.


include “system.in.init”
read_data “system.data”
include “system.in.settings”
include “system.in.charges”

neigh_modify every 1 delay 0 check yes

variable t equal 0.001
variable T equal 298.0
variable P equal 1.013
variable s equal 5
variable c equal 10000
variable d equal $s$c*
variable kB equal 8.617333262E−5

minimize 1.0e-6 1.0e-8 1000 100000

reset_timestep 0

timestep $t

thermo_style custom step etotal temp press lx ly lz density
thermo 1000

fix 1 all nvt temp $T $T 10.0
run 500000
unfix 1

fix 2 all npt temp $T $T 10.0 iso $P $P 100.0
run 2000000
unfix 2

reset_timestep 0

thermo_style custom step etotal epair ke temp press
thermo 1000

fix 3 all nvt temp $T $T 10.0
run 10000000
unfix 3

reset_timestep 0

fix 4 all nve
fix 5 all thermal/conductivity 100 z 50
compute ke all ke/atom
variable temp atom c_ke/(1.5${kB})*
compute layers all chunk/atom bin/1d z lower 0.02 units reduced
fix MP all ave/chunk $s $c $d layers v_temp file temp.profile ave one
thermo_style custom step etotal epair ke temp press f_5
thermo 1000
run 10000000
unfix MP
unfix 5
unfix 4


We have successfully run this for other systems (molecular liquids governed by OPLS-AA), but for MEAM metals we found that the temperature profiles are basically erratic noise. While the lattice thermal conductivity of metals should low, we do not expect it to be completely negligible.

Is anyone aware of existing problems with the application of the Müller-Plathe method together with the MEAM potential? Is there particular requirements for such a case? We have played with the Nevery and Nswap parameters in the fix thermal/conductivity command with no success.

Thank you in advance for your attention.

I don’t have much practical experience with thermal conductivity calculations, but I think it is rather unlikely that the unexpected behavior you report is due to the potential.

Since you are looking at a different material in a different phase compared to what you have investigated before, you need to expect that convergence behavior and magnitudes may be different. While I cannot give you practical advice on what are good settings, the approach to get a reference that you can build on would be to search the published literature for similar studies on similar materials (not necessarily using the same pair style even). You can then try to reproduce those studies and gain experience what choices and settings would be required to reproduce that data and then you may be able to translate that experience to your actual research project.