[lammps-users] Does anyone have experience in thermal conductivity simulation of graphene sheet?

Dear Steve and everyone,

Do you have any experience about the thermal conductivity simulation of graphene sheets?

Now I’m using Muller-Plathe method to simulate the thermal conductivity of the graphene sheet with the AIREBO potential, but my result is far away from the reported and experimental results. I simulated graphene sheets with several different sizes and found all of the thermal conductivity are around ~ 200W/K/m, which is 10~20 times smaller than the published result. I used the following input file to calculate the thermal conductivity:


dimension 3
boundary p p p

units metal

atom_style molecular
read_restart res_0.7ns # restart file after 0.5ns NVT simulation and 0.2ns NVE simulation at 300K

pair_style airebo 3.0 1 1
pair_coeff * * CH.airebo C

timestep 0.001
velocity all create 300 1234567 dist gaussian
fix 2 all nve #perform constant NVE integration to update position and velocity for atoms in the group each timestep

thermo 1000
fix flux all thermal/conductivity 50 z 24
compute ke all ke/atom
variable temp atom c_ke[]/0.0001292355
fix 3 all ave/spatial 10 100 1000 z lower 3.0 v_temp file tmp.profile units box

dump ani all xyz 5000 graphtraj.xyz
dump_modify ani format “C %14.6g %14.6g %14.6g”

thermo 1000
thermo_style custom step ke pe etotal temp press f_flux

run 4000000

figures.doc (240 KB)

Looking at the graph I see that you have clearly driven the system past linear response. I don't know if that is enough to explain the poor agreement with the experimental value, but you should redo it with a less frequent swap and see.

Quoting Dongshan Wei <[email protected]>:

Matt's advice is good. You're also getting a warning, which the
error page describes thusly:

Fix thermal/conductivity comes before fix ave/spatial
    The order of these 2 fixes in your input script is such that fix
thermal/conductivity comes first. If you are using fix ave/spatial to
measure the temperature profile induced by fix viscosity, then this
may cause a glitch in the profile since you are averaging immediately
after swaps have occurred. Flipping the order of the 2 fixes typically