[gmx-developers] help dielectric constant

[David van der Spoel]

On 2014-10-25 04:32, 刘凤海 wrote:
> Dear Dr Spoel:
>
>        Thank you for your reply. I am very sorry I did not state
> clearly. I have bulit a cubic box with 1000 TIP3P or SPCE water
> molecules, two Na+ ions and two Cl- ions, after 100ps NVT or NPT running
> respectively. Then I do a 10ns simulation under NPT. Last I do a series
> of simulations base on previous 10ns simulation with electric field in x
> direction(stength 0.0003~3 V/nm, time 100 ps). I use g_dipoles to
> calculate dielectric constant.
>     EPSILONsub<0>=92.5039
>     EPSILONsub<0.0003>=87.3093
>     EPSILONsub<0.003>=112.201
>     EPSILONsub<0.03>=91.4558
>     EPSILONsub<0.3>=78.688
>     EPSILONsub<3>=81.9181
>     I also append mdp files to accessory. I want to what the mistake is.
> Thanks for your help Dr Spoel.
>
> Best wishes, yours sincerely Andy
>
>
First, you can not use g_dipoles to compute the dielectric constant of a 
charged system, since the dipole fluctuations are ill-defined in a 
periodic box with full charges.

Second, in order for the dielectric constant to converge for a system of 
neutral molecules you need many ns of simulation.

You probably want to have a look at g_current -nojump.


>
>


-- 
David van der Spoel, Ph.D., Professor of Biology
Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205.
spoel at xray.bmc.uu.se    http://folding.bmc.uu.se