[gmx-developers] calculating dielectric constant

[Axel Kohlmeyer]

On Tue, 14 Nov 2006, Qiao Baofu wrote:

QB> Hi all developers,
QB> 
QB> I have posted this problem on the gmx-user list. David gave me some
QB> suggestions, but the problem has not been solved completely.
QB> 
QB> I simulated an ionic liquid [mmimCl], 981 molecules for 1.5 ns, every 5 time
QB> steps saved.  In the .mdp file, I used PME:{ coulombtype  = pme; epsilon-r =
QB> 1}, the simulation temperatuer is 425K.

hmmm.... you may want to have a look at:
http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/files/talk-trieste2004-water.pdf
especially slides 18-21. sadly it does not contain any comments on using 
eps_RF=1 (which are in my PhD thesis, and -sorry- only in german). so 
i'll give a few comments on what i found out:

- epsilon_r _has_ to be set to infinity (i.e. 0 in gromacs) for any
  system where you may need to applay periodic boundary conditions
  on charged particles (e.g. charge groups in large molecules, or 
  whole molecules) or else the fluctuation formula does not work.

- epsilon_r = 1 dampens dielectric fluctuations, espilon_r = oo
  enhances them. thus for strucutural relaxation and reduced low
  frequency fluctuations epsilon_r = 1 makes a lot of sense.

- however, the dielectric constant is derived from exactly those
  fluctuations, so with epsilon_r = 1 you have a _huge_ error bar
  (please look at the fluctuation formula for that case and it
   should be evident).

- dielectric constant needs rather long trajectories, but does not
  need large systems as it is a _local_ property. for water you more
  or less need just two waters between the images and it works.
  if you look at the talk slides you see that the 64 water trajectory
  took about the same time to converge than all others (note time
  line is logarithmic!), but was way off until several nanoseconds
  into the run.

it may be useful to check, whether you can confirm this behavior
on your system.

best regards,
    axel.


QB> 
QB> Then I used  "g_dipoles -temp 425 -f -s -enx" to calculate the the
QB> dielectric constant, the result is: Epsilon = 826.496. I checked the
QB> epsilon.xvg, in which the result is converged to some extent with some
QB> fluctuations at the end.  I think the result is too big. Even though I have
QB> no exerimental data for [mmimCl] at hand, I have some reported data of some
QB> other ionic liquids, 10-20.
QB> 
QB> Is there anyone who can give some suggestions? Great thx!
QB> 
QB> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.