[lammps-users] Choice of potential

Dear all,

While computing thermal conductivity of a SW (10,10) CNT of 50 nm length using Green-Kubo, I find the values to be dependent on potential functions used, which I guess is normal. Using AIREBO gives a value of about 300 W/mK, and using Tersoff gives about 520 W/mK, with all the other parameters of the simulation input code being same. Though both the values agree with some of the previously reported MD simulation results of CNTs, I am particularly confused about the appropriate choice of the potential function. Any suggestions would be helpful.

Also, is the Reverse NEMD technique a better choice for computing thermal properties of solids? But theoretically I would assume that all techniques (EMD, NEMD, RNEMD) should return the same result.


I don't have an answer about the potential you should use. I just wanted to make a comment about EMD, NEMD, RNEMD giving equivalent results. This is true in the limit of small perturbation for the NEMD and RNEMD. Even so for not so small perturbations, given the system is still within linear response.

Quoting Bala Ganesh <[email protected]>:

Dear Bala,

In my understanding, both approaches should give similar results (using the same potential) if the system thermal transport in diffusive in nature (if your characteristic phonon mean free path is smaller than system size). We performed simulations while ago on epoxy polymers (diffusive thermal transport) and got very good comparison using both approaches (Polymer, 50, 2009, 3378). Currently I am running few simulations on (8,8) CNTs using GK and MP method and I do see singificant different between the two results. It could be possible that I need to GK for much longer to get better statistics, which I am currently investigating.

Best Regards,