[gmx-users] forcefield problems - advice needed

[Mark Abraham]

> hi guys,
>
> for quite a long time i am trying to 'create' a forcefield for aliphatic
> alcohols (right now, for ethanol) to use with a water potential we
> employ in our simulations.
>
> there is a problem - the water potential uses combination rule 2, so i
> thought i would just try to use FFGMX and OPLS and change it a bit. it
> didn't seem to work. i also tried the FFAMBER, that uses exactly this
> combination rule. unfortunately, it fails as well.
>
> i'm getting all sorts of errors - segfaults, lincs failures, ... in most
> cases, the whole simulation 'freezes' (kinetic energy of order of 1e-5,
> for example) and nothing happens. i am also quite frequently getting
> close contacts between atoms (within one molecule, or between two
> different molecules).
>
> it seems i am missing something in the forcefield, but i am not able to
> identify what is it... i would be very grateful if anyone of you could
> help me and shed more light in this.

To diagnose the source of this problem, you want to eliminate variables.
Start with a simple system, say methanol in vacuo, and see if you can do a
sensible energy minimization on that. Verify the various parts of the
potential energy by hand in a spreadsheet or something. Once you're happy
that your alcohol internal potentials are all good, then try an isolated
water-methanol complex and repeat. Gradually increase complexity and then
when something explodes you only have a few possible sources of problems.

> also a 'side question' - when using FFAMBER, there are dihedrals defined
> with all atom names specified (e.g., CT CT CT CT), and also dihedrals
> with wildcards (X  CT CT X). for a CT-CT-CT-CT dihedrals, both of these
> would be applicable - are they both used in the simulation, or does the
> latter (X CT CT X) override the former definition (CT CT CT CT)?

I would expect that a specific atom type would override a generic atom
type, although the manual seems to be silent on this issue. You can test
for this in your testing scheme above, of course!

Mark