Size effect in the GK Method for finding thermal conductivity

Dear All,

I’m trying to calculate the phonon thermal conductivity of Cu using the green kubo method. I did several runs to try to determine the optimum parameters for the statistical correlations. I was able to find the correct number for a system of about 108 atoms (by comparison to a reference) and I made sure that the HACF decays to zero. However, when I increase the cell size to 2000 or 4000 atoms the HACF does not converge at all. I tried increasing the number of runs and manipulating the correlation length as well but no luck.

Can anyone assist me with this issue? Any info about finite size effects or suitable correlation lengths.

Thank you

No one has answered your Q for a while. Possibly
Reese Jones has the expertise, CCing him.