Hello Lammps Users,
When setting up a peridynamics simulation, I would like to put a notch in the sample. What is the best way to instruct lammps to not form bonds between nodes? Would the “delete_bonds” command work?
Thanks,
Abby
No, delete_bonds will not work for peridynamics bonds. They
are not the same as other fixed-topology bonds in LAMMPS.
They are stored internally to the PD pair styles in a fix peri/neigh
command they invoke. That command prints out info
about the PD bonds it creates at the point in time a run starts.
I think however that process will respect any neigh_modify exclude
params you have set. So that should work to create a notch.
See the example/crack/in.crack script fora non-PD example of
doing this by labeling atoms on either side of a crack with
different types and using neigh_modify exclude.
Steve
Hi Steve,
Thanks for the suggestion.
One other question. I’m not too familiar with modifying the Lammps code but I was wondering how difficult it would be to add bond_create and bond_delete commands for the paridynamics code?
Thanks,
Abby
Hi Steve,
Thanks for the suggestion.
One other question. I'm not too familiar with modifying the Lammps code but
I was wondering how difficult it would be to add bond_create and bond_delete
commands for the paridynamics code?
if you don't know anything about LAMMPS internals, then i'd say it
would be difficult, as you'll have to understand how neighbor lists
are created, stored and manipulated (the bond-list for peridynamics is
constructed from a neighbor list).
as steve already mentioned, this should not be needed. for the same
reason i mentioned above, just using multiple atom types (but with
identical properties), you can easily tailor your neighbor list to not
contain neighbors between atoms you don't want to get bonded.
i don't see how using a bond_create equivalent would make any sense in
the context of peridynamics.
again, please keep in mind that what peridynamics calls bonds is
*very* different from what is managed by bond styles in LAMMPS, so
everything that is said about bonds elsewhere in LAMMPS does not
apply.
axel.
More generally, it would be nice to have more functionality
in LAMMPS for PD models. The ability to tailor bond creation
and deletion in different regions, of different types, bewteen
interfaces, etc would enable more advanced PD models.
There are also different flavors (linear, non-linear, etc) of
bonds in more general PD models. So a lot could be
done there. It would probably require new PD-specific commands.
Steve