[gmx-users] Velocity autocorrelation

[Ignacio Fernández Galván]

Dear all,

I'm trying to estimate the correlation time for my system, for the purpose of selecting a number of uncorrelated configurations and setting a good value for nstxout.

My systems are usually small molecules in molecular liquids (water, methanol, acetonitrile, cyclohexane...). As a first approach, I calculate the velocity autocorrelation function for the pure liquid (or for a group containing only the solvent in a simulation of the full system).

I expect to find a function more or less like the one shown here: <http://www.compsoc.man.ac.uk/~lucky/Democritus/Theory/vaf.html>. But instead I get a line that rapidly oscillates around 0, with an amplitude that progressively vanishes. I think I understand this behaviour, as it would be due to intramolecular vibrations or librations, but I don't find samples of this in the literature.

My question is how can analyze this and extract some correlation time? How could I get some smooth function like the one shown in the above link? Am I doing something wrong? Is there another way to estimate a good value for nstxout?

Any help or pointers to where to find it would be appreciated.

Thank you,
Ignacio