simtk.org

Dear all,

OpenMM's NonbondedForce class offers the possibility to set an
exception that overrides the normal force field parametrization
for the Coulomb and Lennard-Jones interaction between two given
particles.

I plan to do a Hamiltonian replica exchange simulation and
I thought about using this feature to prepare different force
fields, where the interaction between two groups of atoms is
weakened. Do you think, this is a good idea?

Now I wonder how the particle pair exceptions are treated
in the long-range part of Coulomb interaction, if I use PME.
Are these forces recalculated in direct space? If so, there
might be some loss of performance if I use too many exceptions?
Or is there some other trick involved?

Please help, I'm confused.

Thanks,
Fabian Paul

Hi Fabian,

When you use PME, you are calculating the full set of interactions in an infinite, periodic system. That is, every particle interacts with every one of the infinite periodic copies of every other particle (and, for that matter, of itself).

Nonbonded exceptions apply only to a single copy of each other atom. All of the infinite other periodic copies still interact with the normal Coulomb interaction based on the actual charges of the atoms.

You can see why this has to be the case by considering the main use of nonbonded exceptions: short range bonded pairs (1-2, 1-3, and 1-4 pairs in most force fields). Because these atoms are bonded to each other, their interaction isn't well described by Coulomb's law. So we omit (or weaken for the 1-4 pairs) the interaction, then add in explicit bonded energy terms that do a better job of describing it. But each atom is only bonded to a single copy of each other atom. It is not bonded to any of the other infinite periodic copies, so all of those interactions should be computed with Coulomb's law as usual.

Peter