[gmx-users] diffusion coefficient of oxygen

Justin A. Lemkul jalemkul at vt.edu
Wed Mar 10 13:35:41 CET 2010

Sunil Thapa wrote:
> Respected Experts
> I need your help
> In my study of diffusion of a oxygen molecule in 255 molecules of water, 
> I previously used SPC water model with ffgmx force field with cutoff L-J 
> and Coulomb interaction. I wanted to reproduce the experimental value 
> 2.4 unit for diffusion coefficient of oxygen in water at 298 K and 1 atm.
> To equilibrate the system to the experimental water density of 997 I 
> used Berendsen thermostat and the same barostat. The density of the 
> system was 980(+-) 10 kg/m3 which is not the experimental value. After 
> equilibration, I subjected the system to NVT ensemble md of 100 ns. I 
> got the msd for oxygen molecule and analyzing first 4 ns (which was a st 
> line part) of the plot, I got the diffusion coefficient of about 2.45 
> unit which is close to the experimental value. The question is how can 
> the diffusion coefficient be so close when density is not produced. Is 
> this a mere coincidence?

More likely a consequence of the water model itself.  Keep in mind that SPC 
water is not real water, it does not reproduce all properties of water 
accurately.  No water model does.  Look into the literature for the expected 
density of SPC, but 980 does sound close to what is expected.

> Thinking that I used TIP4P water model with the same number but 
> different force field ffG3a
> . I equilibrated the system to the same pressure and temperature with 
> the same algorithms. After equilibration, the correspondng density was 
> 991(+-) 15 kg/m3. Then I subjected it to the production run of 100ns. 
> Now what I see is that the MSD curve is a hill, initially increasing, 
> reaching maximum and again returning to the base. The diffusion 
> coefficient is 0.7 unit for 100 ns of trajectory.
> What would have happened? Is this due to the increased density? Your 
> precious knowledge on the subject matter would give me a sigh of relief.

You must be careful interpreting your results.  If you only have one molecule of 
your solute, then statistics will likely be very poor.  You probably need either 
more solute molecules within a given system or many replicates of the same 
single-solute system to gather any meaningful data.

The density of the system is a property dominated by the water model.  I am also 
unsure of the validity of G96 parameters (but you haven't said which parameter 
set, "ffG3a" is not real) when combined with TIP4P.  The Gromos force fields 
were parameterized with SPC; TIP4P is more often used with OPLS.  I don't know 
exactly what effect that will have, but you should probably at least be using a 
robust combination, or demonstrate somehow that your application of the force 
field/water model is correct.


> Looking forward to hearing from you
> neal


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list