[gmx-users] Re: Re: Re: Re: David, doubt about the definition of polarization.

[Yinghong]

David:


>  > David:
>  > 
>  >  >/ Yinghong wrote:
>  > />/
>  > />>/ David:
>  > />>/
>  > />>/  >/ Dear Dr. David:
>  > />>/ />/
>  > />>/ />/ According to the shell water model, I used this method to 
> another
>  > />>/ kind of molecule, which is composed of 6 atoms (e.g. benzene).
>  > />>/ Initially, I put a dummy and shell particle (a small mass is 
> given to
>  > />>/ shell, and doing a normal dynamcis) in the center of this 
> hexagon, in
>  > />>/ which shell particle is connected to dummy through your defined
>  > />>/ isotropic polarization method.
>  > />>/ />/
>  > />>/ />/ Theoretically, polarization can be looked as a spring-like
>  > />>/ connection with constant Kr = sqr(qS)/4*PHI*Epsilon*Alpha, and the
>  > />>/ distance between dummy and shell particle can be decided by rsd =
>  > />>/ 4*PHI*Epsilon*Alpha * E0 / qS. Is it right?
>  > />>/ />/
>  > />>/ />/ Now, in my simulation, I applied an external electric field 
> along
>  > />>/ Z direction, and the interactions (vdws + coulomb) between shell
>  > />>/ particle and all the other atoms are exclued. (Of course, here,
>  > What I /
>  >  >>/ did is only to make a test instead of a real case). Obviously, for
>  > />>/ dummy and shell particles, E0 is currently only referred to the
>  > />>/ external field, because local field is excluded.
>  > />>/ />/
>  > />>/ />/ Quantitively,  I set alpha = 0.3 nm^3, qS = 3.0e and E0 = 1.5
>  > />>/ V/nm, through "mdrun -debug", alpha and qS can be correctly output,
>  > />>/ and the calculated value for Kr = 4168 KJ/mol/nm^2 is also in the
>  > />>/ right way. After simulation, I used "g_dist" to check the distance
>  > />>/ between dummy and shell particle (rsd) under such electric 
> field. But
>  > />>/ the calculted value for rsd is only 10 percent of the theoretical
>  > />>/ value although I have tried for many times.
>  > />>/ />/
>  > />>/ />/ So, Could you tell me some possible errors in my defined model,
>  > />>/ and why rsd can not approach to the theoretical value? What is the
>  > />>/ principle for GMX to calculate this rsd?
>  > />>/ />/
>  > />>/ /
>  > />>/  > Isn't the problem nm vs. Ångström?
>  > />>/ I am very sure it is not that problem. Upon the parameters mentioned
>  > />>/ above, rsd should be ~0.1nm theoretically, but my calculation gave a
>  > />>/ value of only 0.01nm. So, any other suggestion?
>  > />>/
>  > />>/
>  > />/
>  > />/
>  > />/ We have
>  > />/
>  > />/ F = q E = k r or
>  > />/ r = q E / k
>  > />/ r = 0.00108 (eV/kj/mol) nm
>  > />/   = 0.1 nm
>  > /
>  >  > Just realized that I repeated your calculation and got the same 
> result.
>  >  > How about exclusions? Have you checked the tpr file for that?
>  > 
>  > Firstly, thanks for your calculating in person. Which result did you
>  > get, 0.1nm or 0.01nm?
>  >See above.
>  
>  > 
>  > In my simulation, I did not define the exclusions in top file. Instead,
>  > I defined two energy groups in mdp file: SHELL & Others. Here, I
>  > wanna check whether the movement of shell particle is only related to
>  > external field in the absence of any other non-bonded interations
>  > between shell and other atoms. So, I defined "energygrp_excl = SHELL 
>  > Others" in mdp files. Is that right?
>  >Maybe, but I'm not sure.
>  
> 
>  > 
>  > Besides, I checked the tpr file, which seems ok.
>  >Does this mean that all exclusions were there?
>  >Is there interaction energy between shell and others in the output edr 
> file?
>  
> Some other information should be useful for your help. Actually, I have 
> succeeded in defining such a shell model previously. In those 
> simulations, I set polarizability alpha = 0.05 nm^3,  qS = 0.6e and mSH 
> = 10 a.u., from which the movement of shell particles are linearly 
> proportional to the external field and agree with theoretical result. 
> But, When I change the values of alpha and qS to the current case (alpha 
> = 0.3 nm^3, and qS = 3e), I can not get a satisfied result any more. Do 
> you have any suggestion to this point?

> no, but try running an energy minimization and check the final coordinates.

What did you mean about checking the final coordinates?

Yes, you remind me, energy can not converge during EM process. Because, for some particular reason, during my simulation, I completely fixed the benzene atoms and make shell particles only move along the direction of electric field. So, as I think, the reason for no converging of energy is from no changes of energy during EM. Is it right?

Thanks again for your kind help, but I am actually lost in this point, which stopped my job from going on.


Xie Yinghong
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20051220/7c9dccc3/attachment.html>