Derek,

With the Green-Kubo method a single run is not very useful. Typically

you need to average over several independent runs to get a meaningful

result. Furthermore, because of the high thermal conductivity of carbon

nanostructures, you also need very long runs since the heat carriers

have long lifetimes. Nonetheless you should expect a 10% uncertainty in

your Green-Kubo calculated conductivities. If you don't mind a bit of

self-promotion, have a look at this preprint:

http://de.arxiv.org/abs/1303.1569

Muller-Plathe and the related "direct NEMD" method show very strong size

dependence for systems with high thermal conductivity. You also need to

try and stay in the low heat flux, low temperature gradient regime where

one can assume Fourier's law is valid.

Cheers,

-felipe.

Thank you very much for the reference. The problems you speak of with NEMD are true, and are the motivation of my desire to use the Green Kubo formalism. Another problem is that it is often very impractical to use NEMD method son carbon nanostructures without a degree of non-linearity in the thermal gradient near the reservoirs.

Your reference will be very helpful in narrowing down the reason my value is so low compared to an expected result.

Thank you very much for the reference. The problems you speak of with NEMD are true, and are the motivation of my desire to use the Green Kubo formalism. Another problem is that it is often very impractical to use NEMD method son carbon nanostructures without a degree of non-linearity in the thermal gradient near the reservoirs.

Above, related to the non linear gradient close to the heated areas, you don’t have to work so close to them.

Your measuring volume could be a small part of the system away from the heated/cooled regions. As long as you are calculating the right heat flux through such volume you should be ok. The material property does not depend on where you measure it.

Carlos