Re: diffusion coefficients

From: Thomas Cheatham <>
Date: Mon, 14 Apr 2003 14:34:39 -0700 (PDT)

> ps. I use " diffusion :WAT 4 average filename" command

Professor Case answered your first question; use a time per frame of 0.2
ps rather than 4.0 as above.

> Diffusion coefficent can be obtained from the slope
> of the line. But I would like to calculate diffusion
> coefficent of several intersting solvent molecules.
> Howere there is no straight line in filename_r.xmgr
> figure but one curve with big fluctuation when I use
> method above to calculate. How should I obtain
> diffusion coefficient from the such filename_r.xmgr?

Given sufficient sampling, you will obtain a nice straight line from which
the diffision constant can be calculated. In practice, this is much
easier when averaging over multiple solvent or ion molecules (i.e. all
water or all Na+ ions) since there is more sampling and good statistics
can be obtained. However, when you look at a single molecule, these
values do not easily converge on a short time scale. Looking at a single
water likely will not give as accurate an estimate of the diffusion
constant as the average over all water molecules. Moreover this single
molecule estimate is more strongly biased by the local environment; some
individual "ions" or water or solvents may interact with a solute for part
of the simulation (and thereby slow down) and not represent bulk behavior.

To obtain good statistics on single molecule diffusion, longer simulations
and block averaging may be necessary. However, your diffusion example
*does* average over all water; if this is not linear, perhaps a
significant portion of the water is not diffusing properly (i.e. bound to
solute?) and/or you haven't run long enough. However, if I remember
correctly, bulk water diffusion estimates can be obtained from simulations
as short as 100 ps so I am guessing the former.

Good luck in tracking this down.

\ Thomas E. Cheatham, III (Assistant Professor) College of Pharmacy
| Departments of Medicinal Chemistry and of University of Utah
| Pharmaceutics and Pharmaceutical Chemistry 30 South 2000 East, Room 201
| & Center for High Performance Computing Salt Lake City, Utah 84112
| e-mail: phone: (801) 587-9652 FAX: (801) 585-9119
\ Offices: BPRP295A / INSCC 418
Received on Mon Apr 14 2003 - 22:53:01 PDT
Custom Search