How to turn off certain pairwise interactions.

Dear all,

I want to turn off some certain pairwise interactions because those interaction will crack my simulation, which is not what I want to explain in details right now. I want to get some ideas from you on how to turning of LJ potentials. And I have the following two ideas.

1) Can I use pair_ style hybrid? My model is basically lj/cut/coul/long, so can I set those interactions I do not want lj/cut and the parameters as epsilon and delta zero?
2) I did notice that there are two specific command regarding to this. The first one is neigh_modify. But according to the manual, it is incorrectly with the using of long-range coulombia force, which I have. And another one is special_bond. But I have not fully understand of the special_bond. It is used to those bonded atoms. But in my simulation, the pairwise atoms I want to turn off is not bonded at all. So how to use that as the manual instructed.

Best wishes,
Andy

Dear all,

I want to turn off some certain pairwise interactions because those interaction will crack my simulation, which is not what I want to explain in details right now. I want to get some ideas from you on how to turning of LJ potentials. And I have the following two ideas.

1) Can I use pair_ style hybrid? My model is basically lj/cut/coul/long, so can I set those interactions I do not want lj/cut and the parameters as epsilon and delta zero?

that won't do the task entirely. you still have the contributions from
kspace, which are not pairwise forces but computed as the interaction
of each atom with all periodic images of all the other atoms. using
neighborlist exclusions or pair style hybrid will only exclude the
real space part and thus the impact of the exclusion (and thus the
result of your simulation) dependent on the choice of coulomb cutoff
and kspace convergence.

2) I did notice that there are two specific command regarding to this. The first one is neigh_modify. But according to the manual, it is incorrectly with the using of long-range coulombia force, which I have. And another one is special_bond. But I have not fully understand of the special_bond. It is used to those bonded atoms. But in my simulation, the pairwise atoms I want to turn off is not bonded at all. So how to use that as the manual instructed.

special_bond only applies to non-bonded interactions that have a
direct bond. pick up an MD text book and look for where it explains
conventional (aka class 1) force fields and _exclusions_. given the
implementation of exclusions in LAMMPS, that would be your best shot
for a reasonably accurate exclusion of pairs of interactions, i.e.
define a bond with force constant zero for all pairs of interactions
that you want to remove. please note, however, that this has other
implications: both atoms in a bond have to be within the ghost
communication cutoff for the individual subdomain, when running in
parallel, so as your "fake" bond gets longer when the crack develops,
you may run into crashes unless your communication cutoff is large
enough. this also has the problem, that you have to specify each
excluded pair explicitly, and you'll can have all kinds of
discontinuities and and inconsistencies, if you continue the
simulation long enough and atoms that were not excludes can migrate
into the void(s) you create.

that all being said, it matters *very* much what it is that you want
to model and why you want to remove pairs of interactions where and
how. in general, those exclusions, e.g. in the bundled "crack"
example, are not so much used to actually model the cracking process,
but merely to introduce a nucleation point with the purpose to have
the crack happen where it is convenient to observe and easy to compare
in a consistent fashion. most people that are asking the kind of
questions that you are asking seem to be missing this point entirely.

axel.