GROMACS modification: No
I am trying to calculate the diffusion coefficient of SPCE water from a short MD simulation trajectory. Using the mean square deviation technique, I am able to get the value of D as 2.5991 (1e-5 cm^2/s) which seems to be in the correct ballpark.
However, when I try to derive the value using the velocity autocorrelation function, the value is quite different from that derived using MSD.
I ran a short simulation with the following parameters as the continuation of a 20 ns trajectory.
nsteps = 50000 ;
dt = 0.002
nstxout = 10
nstvout = 10
nstenergy = 10
nstlog = 10
and computed the velocity autocorrelation function using the flag -mol.
The resulting autocorrelation plot is shown here
Following the steps here: Diffusion Constant - Gromacs, (basically, integrating the area with gmx analyze and dividing by 3)
the calculated value of the diffusion constant comes to around: 5.97 (1e-5 cm^2/s).
I am hoping that someone could point what has changed in how gmx velacc computes the autocorrelation or integrate option of gmx analyze from older versions of gromacs. Or, if there is another way to compute the diffusion coefficient from velocity autocorrelation data.