[gmx-users] Velocity autocorrelation

[David van der Spoel]

On 2012-04-01 17:34, Ignacio Fernández Galván wrote:
> --- On Fri, 30/3/12, David van der Spoel<spoel at xray.bmc.uu.se>  wrote:
>
>> No need, I think it is just fine!
>> What your seeing is the average vacf per atom, and
>> apparently there are
>> correlations between the vibrations which cause the
>> oscillation. In
>> theory both VACF should be the same but they are relatively
>> close. A
>> longer simulation is always good of course.
>
> But how can I extract a meaningful relaxation time from there?

It depends what you want to compare to.

For e.g. rotational ACF you would use g_rotacf for a bond vector, and 
this can be compared to NMR. For dielectric relaxation you can use 
g_dipoles and compare to dielectric measurement.

 From a VACF you can compute the vibrational spectrum and compare to IR 
measurements (if you dare). This is now implemented directly in the 
development version of 4.6, but you can also do it manually in xmgrace.

Further, you can integrate VACF to get the diffusion constant, as you 
have done already. The relaxation time of a velocity averaged per atom 
can be analyze by fitting the vac to y = exp(-t/tau). This gives 5 fs in 
your graph.


>
> Maybe instead of individual atoms, I should analyze the VACF of the center of mass of each molecule, and the rotational ACF. The latter I know how to do it. Is there an "easy" way to calculate the former?
>
> Anyway, in this case it makes sense to divide the system in molecules, because that's what it is, but if it were say a protein, what would be the best way to analize the VACF and get some meaningful relaxation time?
>
> Thank you,
> Ignacio


-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se