[gmx-users] sampling techniques when calculating diffusion coefficients using Green-Kubo

[Ying Chen]

Dear all,

I am trying to use Green-Kubo method to estimate diffusion coefficients.
For a simple box of pure water, MSD gives D~2.7E-9 m2/s, but velocity
autocorrelation function (VACF) method gives D~6.8E-9 m2/s, and this value
becomes smaller with smaller timestep dt, but it still doesn't reach ~3E-9
for dt~0.00002 ps and total time=2 ps.  I saw this page
http://www.pages.drexel.edu/~cfa22/msim/node29.html
mentioned that VACF may decay very slowly and this method becomes very
costly, "it is often desirable to use more sophisticated sampling
techniques when estimating transport coefficients using Green-Kubo-type
analyses". Does anyone know what these techniques are or where we can find
them? Thanks a lot!

Sincerely,
Ying