LAMMPS experts,
Will there be a way in LAMMPS to specify a minimum for a cutoff rather than just the maximum? For instance, if I wanted to use Potential A for distances below a certain threshold but wanted to use Potential B for interatomic distances above that threshold? I understand that this can be accomplished by tabulating a potential - Though I am still curious if this is something that might be implemented into LAMMPS one day.
Best Regards,
Ben
LAMMPS experts,
Will there be a way in LAMMPS to specify a minimum for a cutoff rather than
just the maximum? For instance, if I wanted to use Potential A for distances
highly unlikely as this would require a massive rewrite of the code at
very little benefit and likely incurring some overhead for the common
case.
below a certain threshold but wanted to use Potential B for interatomic
distances above that threshold? I understand that this can be accomplished
by tabulating a potential - Though I am still curious if this is something
that might be implemented into LAMMPS one day.
the main benefit of having a "maximum" cutoff is to reduce having to
compute the number of interactions that contribute little to the final
force/energy. the number of interactions at a short cutoff are very
small in contrast and can be just skipped over. piecing together a
potential in the way you describe via hybrid/overlay is not a very
good idea, since it can be very challenging to get a proper clean
transition from potential A to potential B. a *much* better approach
would be to implement a combined potential that processes both (or
more) components in one sweep and can also guarantee that the forces
are always well defined and smooth to avoid artifacts from forces and
energies that are not continuous with the distance.
axel.
i should also add that the reasoning for using a cutoff in the first
place was that interactions are small at the point of the cutoff,
which is very likely not the case for atoms that are close, hence the
need for a continuous potential and thus an integrated solution (out
of which tabulation clearly is one).
axel.
Just to expand on Axel’s point: such an approach is already essentially what the lj/charmm potentials do, so you could look at those as an example of how to get a new version combining two other potentials started.
—AEI