[gmx-users] [ nonbonded_params ] and opls-aa

[Paul Rowntree]

Greetings all;

I am trying to add an LJ interaction to an opls-aa ff; my way of extending the
ff is to include a small .itp file into the topology file, rather than
modifying the distribution ff files.  In this file I define my types etc, plus
the LJ parameters.

In the distribution ffoplsaa.itp file, the default combintaion rule is mode 3,
requiring sigma and epsilon values.  The printd manual indicates that this
convention will always be applied to the [nonbonded_params] inforamtion. 
However, when I do a gmxdump of the resulting .tpr file, I find a new function
type in the LJ list that has exactly the same numerical values as my sigma
epsilon pair, but the listing format implies they are being interpreted as C6
and C12 values, which would be MUCH more repulsive than intended.  The mdrun
results support this interpretation, and the run is clearly not behaving well. 
It seems as if grompp has ignored the combination rule setting in the [ default
] parameter block.

So the question is : should I be able to include values in sigma epsilon format
when I extend the opls-aa force field ?   I can recalculate the C6 and C12 of
course, but it takes away from the pure opls 'look and feel' of the topology.

I thank you in advance for any insight you can provide.

paul Rowntree / Universite de Sherbrooke