[lammps-users] who have experience to get good linear temperature gradient utilizing the command "fix thermal/conductivity"?

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,
Recently I use the command “fix thermal/conductivity” to build linear temperature gradient. At the beginning, I got good result (see the attachment fig 1). However, when I start to study the size effects, that is to say I enlarge the simulation box, no matter how I adjust the parameters: N and Nswap
(fix ID group-ID thermal/conductivity N edim Nbin Nswap), I didn’t get the better linear temperature gradient(see the attachment fig 2). Anyone who have experience utilizing this command, please give me some advice and suggestion, and that will be highly appreciated.
Sincerely,
Maosheng

1.jpg

2.jpg

2

Hi Maosheng. I haven't used this fix, but I have used the viscosity fix, which is very similar. In my experience, the curvature near the swap bins occurs when the system is being driven too fast. You should try swapping less frequently. This may mean running *very* long simulations to get a good measurement. Although, you shouldn't need to run as long as would be needed for a converged Green Kubo type measurement.
Matt

3

Dear Matt，
Thank you for your letters. And thank you very much for your advice. I will try it.
Merry Christmas!
Best,
Maosheng