[lammps-users] Thermal Conductivity Calculation usingGreen-Kubo Relations

Hi Steve,

There are several advantages of the Green-Kubo (GK) method compared to the non-equilibrium method (NEMD). 1) Because GK simulations are done with the system in equilibrium, there is no imposed driving force and so the system is always in the linear-response regime. 2) The GK method can compute the entire heat flux tensor with one simulation. NEMD would require a simulation run for each component of the heat flux vector. There the GK method lends itself much easier to the study of anisotropic effects. 3) A key advantage to the GK method is that it is much less sensitive to finite model size effects because a heat source and sink are not required. Hence the GK method can use much smaller models than the NEMD methods. For example, good results for diamond were obtained with only a 4000 atom model with a dimension of about 2.8 nm. The NEMD method would require a model comparable tot he mean free path length (174 nm).

Of course there are disadvantages of the GK method compared to the NEMD method but I have only discussed the positive attributes of the GK method.

A way to implement the GK in lammps would be beneficial to many researchers. I would be interested in obtaining a copy of any GK source code additions to lammps if anyone has been successful in implementing it.

Bill Evans

This is useful context. I understand GK better now.
Someone at Sandia is sending me a GK fix that
is supposed to work for any pair potential. I'll add it in the next couple
weeks, once I get a chance to look at it.

Thanks,
Steve

Hi Bill,
  I have working, validated code for computing the heat flux and a python script for computing the necessary autocorrelation and its integration. This was developed by Vikas Varshney, Phillip Howell and myself. I am cleaning it up and giving it to Steve P. in the next week or so.
   I can send you the cleaned up version and the correlation script at the same time I send it to Steve if you would like.
Reese

WJ Evans wrote:

Hi Reese,
I'm also interested in code for heat flux.
I would really appreciate if you can send it to me.
German Samolyuk

The compute heat/flux command authored by Reese, Philip, Vikas
was released today in a 6Jul09 patch.

Steve

Thank you!
German

Dear Steve,

This is good news.
Does the "compute heat/flux" support charged pairwise potentials such as
Coulombics interaction?

Regards,

Ali Rajabpour