[gmx-users] FEP softcore problematics

[David Mobley]

Maik,

You may be having the infamous "charge overlap" problem. In
particular, one additional reason not to change LJ interactions at the
same time as electrostatics is if you are inserting or disappearing
atoms.

Assume, for a moment, that you are disappearing a couple of atoms. The
soft core for LJ allows these disappearing atoms to begin overlapping
with other atoms before they are completely disappeared -- while they
also have some charge. Imagine, for example, you are disappearing an
oxygen with a negative charge. Now you find yourself in the position
of having a negatively charged site that can occasionally overlap with
positively charged atoms which you are NOT disappearing (especially
things like hydrogens on water molecules, since they have 0 LJ
parameters in most FF, but also possibly other atoms). Obviously,
putting a positive charge extremely close to a negative charge results
in huge forces. When this happens, the simulation will often explode
(which, in GROMACS, usually means LINCS errors).

This is really an additional reason to do the calculation in several
steps, in my opinion.

If you are desperate not to do it in two steps, there was a recent
paper (2006, I think) from a group in England that presented an
alternative method for handling the charge transformation that was
designed to smooth out the singularity encountered when two opposite
charges get really close to one another. The title had something to do
with "charged species". The idea was to allow both the electrostatic
and LJ transformations to be done at once, like you're trying to do.
If you can't find the reference and want it, let me know and I will
dig it up. I had thought at one point about implementing this into
GROMACS, but decided it wasn't worthwhile for me since the hard part
is the LJ transformation, anyway, and that's where I spend most of my
CPU time, so this would be a fairly minimal savings.

David

On 6/21/06, Maik Goette <mgoette at mpi-bpc.mpg.de> wrote:
> Hi all
>
> Because the fact of handling vdw and electrostatics with the same values
> was discussed before, I won't go into much detail about what Berk and
> David M. wrote.
> I changed the code to apply different alpha values for vdw and
> electrostatics. Now I am able to e.g. calculate vdw with an alpha=1
> (l-power linear or squared) and qq with alpha=0 (always linear, if alpha
> != 0).
> What I see are lincs errors after 40% of the sim-time in the upper case.
> Is it because of the fact, that vdw are treated soft and qq hard?
> With alpha=2 for qq it works, but still yields large fluctuations at
> lambda=0 & 1.
>
> Any suggestions?
>
> Regards
> --
> Maik Goette, Dipl. Biol.
> Max Planck Institute for Biophysical Chemistry
> Theoretical & computational biophysics department
> Am Fassberg 11
> 37077 Goettingen
> Germany
> Tel.  : ++49 551 201 2310
> Fax   : ++49 551 201 2302
> Email : mgoette[at]mpi-bpc.mpg.de
>          mgoette2[at]gwdg.de
> WWW   : http://www.mpibpc.gwdg.de/groups/grubmueller/
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>