[gmx-users] free energy and soft core

[Berk Hess]

>My molecule just has a few charges of which the interaction are excluded
>(Coulomb1-4=CoulombLR=CoulombSR = 0), so it's all about the LJ
>interactions within the molecule.
>I could reproduce the bug that you mentioned if rcoulomb<rvdw.
>This looks to me like another bug..?

It have found the bug.
It is present in 3.2.0 and 3.2.1, older Gromacs versions are ok.

When sc-alpha>0 and there are 1-4 interactions present, atom types
are permanently changed after the first md-step. After the first step
all normal (not 1-4) LJ interactions will be incorrect. In your case
the LJ energy at step 1 is zero.
This error will probably not happen unnoticed as the system will
behave very strangely with wrong LJ types and there is a jump
in dV/dlambda and the LJ energy after the first step.

This fix is simple:
move the bracket at line 1147 in src/gmxlib/fnbf.c to line 1142
and recompile.

Thank you very much for reporting this problem,

Berk.

_________________________________________________________________
Play online games with your friends with MSN Messenger 
http://messenger.msn.nl/