[gmx-users] g_velacc again

[Nuno R. L. Ferreira]

David wrote:
> Have you looked at the graph in xmgrace? It is probably easiest to start
> there, use the integration function to see how ill-behaved the ACF is.
> It may just fluctuate wildly. Then use the non-linear fit to fit your
> ACF to a function of your choosing and integrate again. You can probably
> get the same result from gromacs if you use some more flags like
> -beginfit and -endfit

I've done the integration and the value is similar to the correlation time
given by g_velacc. So basically the Correlation time is the VACF integral?
But if it is, why does it come in ps? Should't be in nm2.s-1 or something
similar ?

Also, I've done the calculation with -normalize. After new try
with -nonormalize the value that comes out is:
COR: Correlation time (plain integral from 0.000 to 99.999 ps) = 0.986 ps
(by integration of Cvv(t) vs time graph, the value is similar (0.980).
Applying the formula given in last e-mail, and considering that velocities
come in nm.s-1 , I get a value of
Dt = 3.2 10-5 m2.s-1 (Dt exp. = 2.4 10-9 cm2.s-1).
So, I think that some units are wrong (a factor of 10-4).

Regards,
Nuno