[gmx-developers] External E Field PME

[Timothy DeWolf]

Hi,
>> An electric field can be applied directly on the charges, as it is done now.
What are these issues with the virial then?

>> There is some literature on this.
If you do have a moment/any ideas in your head of what a couple papers are, I would much appreciate the references.

Thanks,
Tim

-----Original Message-----
From: Berk Hess [mailto:berkhess at gmail.com] On Behalf Of Berk Hess
Sent: Saturday, December 13, 2014 2:17 AM
To: Timothy DeWolf
Cc: Discussion list for GROMACS development
Subject: Re: [gmx-developers] External E Field PME

Hi,

That sounds like an ancient comment.
An electric field can be applied directly on the charges, as it is done now. But note that there are unresolvable issues with the typical infinitely periodic simulation system and an external field while you usually want to mimic a small part of a large, non-periodic system. The boundary conditions at infinity are important here. There is some literature on this.
A proper solution is generating a field using confined ions.

Cheers,

BerkOn Dec 13, 2014 9:14 AM, Timothy DeWolf <tdewolf at uvic.ca> wrote:
>
> Hi!
>
>   I just signed up for this list, and I’d like to contribute to GROMACS.  I’m a Ph.D. student, with relevant background in physics, biochemistry, and engineering.
>
>   I’m interested in extending GROMACS’ abilities to impose user-defined electric fields, and specifically interested in spatially constant time-varying fields.  I looked at calc_f_el in sim_util.cpp, having read “Patching mdrun” on the GROMACS website, and see that it would be fairly easy to modify.  But I’m concerned about this “WARNING” in the comment lines above the function definition, which state that there “can be problems with the virial.”
>
>   The solution proposed there is to implement the “externally imposed” field directly in PME.  I’m not an expert yet, but I read the relevant section in Berendsen’s book and the ’95 smooth PME paper to get up to speed; I haven’t yet got to a detailed study of pme.c.
>
>   What is the best place to begin?  I suppose that a linearly increasing potential mapped to each charge along with the FT lattice sum terms in the PME routine(s) would do it?  May I please ask for any suggestions, ideas or hints (possibly including relevant spots in the code)?
>
>  
>
> Tim
>
>