compute heat/flux

Dear all,

I try to use the NEMD method to calculate the thermal conductivity of graphene (on x-y plane). I fixed the two ends of x direction, and applied periodic BC in y direction. I used “fix heat” to generate the heat flux along x direction. Besides recording the temperature profile, I also calculate the heat flux in the region between the two thermal reservoir using “compute heat/flux” command. Time step is 0.2 fs. Before sampling, I equilibrated system for 2 ns to reach steady state. I used another 2ns for sampling. But the calculated heat flux was different with the value I used in fix heat command. The other interesting thing is that the heat flux in y direction is non-zero, and actually it is comparable with that in x direction. Could anyone tell me what’s wrong with the simulation setup? or is there any other method the calculate the actual heat flux? Many thanks.

I also try the other direct method the calculate the thermal conductivity. I applied the temperature gradient and calculated the resulted heat flux. Periodic BCs were used in both x and y direction. I used nvt to thermostat the two ends of x direction at 290K, and themostat the middle region at 310 K. Because there is no constraint in z direction (monolayer graphene), the graphene will drift out of the simulation box after a while. Also there are bulk movements along x and y direction. How to cancel these kinds of movements without affecting the results ( e.g. temperature gradient and heat flux)? Thank you.


See section 4.20 of the manual for an overview of this

Re: drift of system thru box. There are various ways to
prevent that. You can look at fix recenter or fix momentum.
You can nail down an atom. You can attach a fix spring to
the entire system and loosely tether it to the center
of the box.