AMBER Archive (1999)

Subject: solvent diffusion problem solved

From: Mathy Froeyen (Matheus.Froeyen_at_rega.kuleuven.ac.be)
Date: Wed May 26 1999 - 04:05:23 CDT

Hello Amber users,

My troubles with the calculation of the water diffusion coefficient from
the slope of the mean-square displacement correlation function is a
matter of including enough statistics.

First of all, there is no problem with my Amber PME trajectories.

The problem is in the simple program I wrote myself, which calculates
the correlation function.
I did not keep track of periodic conditions. I supposed this was not
necesesary, because I would only include waters moving close to the
solute in the center of the box, and at short times ( < 10 ps).

After a suggestion from Thomas Cheatham, I found that in my calculation,
because I do eliminate periodic conditions, I do not have enough
molecules in the calculation to get good statistics. That's why my
correlation curve moves up when I increase the radius around the solute
to include more bulk waters. The curve does not get worse but better! Of
course I could not make this radius too high because then I would get
serious problems with periodicity.

(The purpose of this radius is to investigate the influence of the
solute on the solvent diffusion, see for instance in the book Proteins,
Brooks, Karplus, & Pettitt 1988 p. 155).

I have to adapt my program to include periodicity to increase statistics
in my calculation (or I have to do longer MD simulations).

The high velocities for some molecules I found, I suppose they are
normal if you consider a Gaussian distribution around an average
velocity at 300K.

Thanks to all who responded,

Mathy Froeyen