The problem about Al2O3's thermal conductivity using ReaxFF!

LAMMPS LAMMPS Mailing List Mirror
1

Dear Sir or Mad,

Thanks for your reading.

I do not know whether I can calculate the thermal conductivity using ReaxFF in LAMMPS.

I have a problem about the Al2O3’s thermal conductivity using ReaxFF. We can not calculate the thermal conductivity in C++ version, because " pair_style reax/c command cannot yet compute per-atom energy or stress". But, Fortran version don’t state this. So, I used fortran version reax to calculate the Al2O3’s thermal conductivity with EMD(Green-Kubo) method. But, I obtained the negative thermal conductivity, and I do not know the cause.

I state those:

  1. Before calculating the thermal conductivity, we must calculate the per-atom energy and stress, in order to calculate the heat current;
  2. The Al-O potention maybe not the cause: I obtained it from Adri. And I using other Al-O reax potention, from “potention” document, also obtained the negative thermal conducitity;
    3)Green-Kubo method maybe not the reason: I use it calculate the Si’s thermal conductivity, and the result is all right.

I guess two causes:

  1. The reax force field is not suit to calculate the thermal conductivity; (but I do not believe it)
  2. The per-atom energy or stress do not accuracy using LAMMPS of fortran version resulting in the problem.
    I want to eliminate the code reason, because C++ can not calculate the per-atom property, and maybe Fortran calculate is wrong.

Maybe there are other causes! Could give me some suggestions about this problem!

Best wishes,
Xiong

2

Aidan or others can comment. Your reasoning
is sound, but I don't think there is any problem
with the per-atom energy/stress that the Fortran ReaxFF
calculates (your (2) below). You can verify that it
is self-consistent, but summing the per-atom energy (or stress)
and comparing it to the total energy of the system. They
should be identical.

Steve

3

If the autocorrelation for Al2O3 is an oscillating one (like for example for amorphous silica) do the integration outside of lammps. I did get negative thermal conductivity for Silica using the integration in lammps and I think it is due to the error while integrating the oscillating autocorrelation.

Deena

4

I would suggest, that you start out by reproducing the thermal
conductivity for a previously studied system. Once you can do that, you
will be in a better position to assess the validity or not of the Al2O3
reuslts.

Aidan