Hyper local

Dear Lammps users

I try to run a hyperdynamics (HD) simulation on a molecular system including bonded atoms. In fix_hyper_local.cpp there is a comment saying : warn if molecular system, since near-neighbors may not appear in neigh list user should not be including bonded atoms as hyper “bonds”.

My question is : Since using "neigh_modify exclude" may give inconsistent coulombic energies, is there another way to exclude bonded atoms from hyper “bonds”?

Sincerely

Jehiel

You can also look at this the other way around. The special treatment of excluding bonded atoms happens only with the default special_bonds scaling value of exactly 0.0. so changing that to e.g.:

special_bonds lj/coul 1.0e-100 1.0e-100 1.0e-100

(or equivalent for other required settings) would not exclude those pairs of atoms from the neighbor list, but the contributed forces/energies will be far below the error margin of the model and thus equivalent to the situation with using exactly 0.0 0.0 0.0.

Axel.

Fix hyper/local takes a group as input and it will currently
not create hyper bonds (HD) between 2 atoms not in the group.
But it will between 2 atoms, one in the group and one not.

So if you create a group only for atoms not in molecules,
then you will not get HD bonds within the molecules. If
you also want to exclude HD bonds between a non-molecular
atom and a molecular atom, then we could add an option
to fix hyper/local to disable those bonds as well.

Are you saying (in your earlier emails) that those kind of
HD bonds are what is causing issues when HD runs?
I’ve actually not heard of anyone using HD for molecular
systems, so this is not something we thought about
when writing the code.

Steve

Dear Sirs,

Thank you for your replies. I have tried your suggestions but I still get the same issue. When I create a group including my target atom and permanently bonded atoms, these atoms effectively do not make HD bonds between them. However, when the permanently bonded atoms and other atoms not belonging to the created group are in a distance r < cutbond, fix hyper/local try to make a HD bond between them and the crash occurs. I think the solution would be, as suggested by Steve, to add an option that disables HD bonds between atoms in the group defined as input in fix hyper/local and permanently bonded atoms.

Thanks and regards.

Jehiel