I am trying to calculate the thermal conductivity of monolayer graphene using Green-Kubo method. I used periodic BC for the two in-plane direction x and y. Fixed BC was used in z direction. The time step was 0.2 fs, and simulation was done in NVE for 12 ns. Tersoff potential was used. I am wondering how to choose the correlation length. When correlation length is great than 500 ps, the integration of heat flux auto-correlation will drift and cannot converge. Do anyone has similar experience? Thank you.