[gmx-developers] modifing nonbonded interactions

[Nazish Hoda]

Thanks for the details. I really appreciate getting so much information from
you people.

I have developed a strategy to evaluate LJ parameters for the implicit
solvent simulations.
It involves a single empirical parameter, but mimics explicit water results
pretty well.

I also tried some runs with position dependent \epsilon, it is also working
well for 
one of the systems I have tried. I want to match 4 things, 3 rdfs and Rg.
Initial results are promising. 

Regarding the issue of the functional dependence of \epsilon, that is a bit
complicated, 
and I fully agree with Berk on that. At present, I am trying the simplest
possible form of \epsilon.

As of now, I am able to replicate pair formation properly. 
Next step, is to see how it works for complex formation.
I am not very hopeful about that, let see how it goes.
There I would need a more complicated function to describe \epsilon.

Nazish
 

-----Original Message-----
From: gmx-developers-bounces at gromacs.org
[mailto:gmx-developers-bounces at gromacs.org] On Behalf Of Berk Hess
Sent: Friday, October 24, 2008 3:45 AM
To: Discussion list for GROMACS development
Subject: Re: [gmx-developers] modifing nonbonded interactions

Hi,

To understand the approximations, lets have a look at the simplest case:
two single-atomic ions in water.

In general the dielectric permittivity of the medium around a single ion is:
(1) eps(r)
When you have two ions this I already don't see a reasonable approximation
to describe the permittivity, but lets say for arguments sake that it is 
still (1).
(If you read my papers you will see that even for high concentrations of 
NaCl(aq)
the permittivity still seems near to constant of the whole space, 
surprisingly this
even seems to hold for an Na-Cl contact pair!).

For two ions one can DEFINE the electrostatic interaction as:
(2) c(r)/r
Note that c(r) is NOT equal to 1/eps(r). To get c(r) from eps(r) you have
to solve the electrostatics problem with spatially varying permittivity.

For multi-atomic ions/molecules it is not clear how good approximation 
(1) is.
For systems with more than two molecules (2) will be a (maybe bad) 
approximation.
But using (2) in Gromacs with PME is possible, if c(r) is a constant 
eps1 after the cut-off.
Then you should use:
coulomb_type = pme-user
epsilon_r = eps1
and fill the Coulomb table with eps1*c(r)/r.

But as Mark said, the real work will be in parameterizing the "LJ" 
interactions,
which are no longer simply LJ, but contain all kinds of multi-body effects.

I think it would be much simpler and also not such a bad approximation
(especially compared to the many other approximations) to assume that the
permittivity is constant over space.

PS: Many people on this list seem to think that you do not need periodicity
with implicit solvent. This is only true when you have a single molecule
or a complex. In many cases, such as aggregation studies, one needs
periodic boundary conditions.

Berk


Mark Abraham wrote:
> Nazish Hoda wrote:
>>> PME requires periodicity, which you shouldn't need for implicit-water.
>> There are still charged molecules in the neighboring, next 
>> neighboring, etc
>> boxes.
>> How to account for them ?  
>
> If your system has a net charge, you have even more problems if you 
> require no solvent (so no counter-ions) and periodicity. PME doesn't 
> work for systems with net non-zero charge.
>
>>> If you're really wanting to do implicit-water simulation, you should 
>>> be looking at simulation packages other than GROMACS.
>>
>> Won't using a modified LJ parameters work ?
>
> Life is not that simple. These parameters are carefully optimized to 
> reproduce some (often vaguely) relevant data. It's inconceivable that 
> any form of 6,12 potential on solute atoms could account for the 
> electrostatic, frictional, thermostatic, dispersion and doubtless 
> other effects that explicit solvents have upon their solutes. See 
> http://wiki.gromacs.org/index.php/Parameterization
>
> Mark
> _______________________________________________
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the www 
> interface or send it to gmx-developers-request at gromacs.org.

_______________________________________________
gmx-developers mailing list
gmx-developers at gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-developers
Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-developers-request at gromacs.org.