Parallel run of a system with dihedral angle between two rigid bodies

Dear all,

I’m trying to use LAMMPS to simulate a system containing two types of rigid bodies, each composed of three particles.

Each one of the rigid bodies is connected to an opposite type rigid body, making a dimer, with a covalent bond (defined between two atoms from each rigid body). In addition, a dihedral potential (defined between four atoms of the dimer) controls how the two rigid bodies, within a dimer, orient with respect to each other.

I use “fix rigid” with harmonic bond and dihedral style potentials and I can run my system in serial mode without any issue. However when I switch to the parallel mode, the simulation stops with an error message, e.g.

ERROR on proc 4: Dihedral atoms 985 986 992 991 missing on proc 4 at step 213 (…/neigh_bond.cpp:540)

How can I fix this issue? Do I need to modify neighbour list properties?

Also, I get a same error with “fix rigid/small”.

Cheers,
Nima

Dear all,

I'm trying to use LAMMPS to simulate a system containing two types of rigid
bodies, each composed of three particles.

Each one of the rigid bodies is connected to an opposite type rigid body,
making a dimer, with a covalent bond (defined between two atoms from each
rigid body). In addition, a dihedral potential (defined between four atoms
of the dimer) controls how the two rigid bodies, within a dimer, orient with
respect to each other.

I use "fix rigid" with harmonic bond and dihedral style potentials and I can
run my system in serial mode without any issue. However when I switch to the
parallel mode, the simulation stops with an error message, e.g.

   ERROR on proc 4: Dihedral atoms 985 986 992 991 missing on proc 4 at step
213 (../neigh_bond.cpp:540)

How can I fix this issue? Do I need to modify neighbour list properties?

no. how long is the largest end-to-end distance between the atoms in
those specific dihedrals?
can it be longer than the cutoff for non-bonded interactions?

if yes, then you can address the issue by increasing the cutoff for
ghost atoms (i.e. the "thickness" of the layer of atoms communicated
from neighboring subdomains) to be large enough. both, the neighbor
list "master cutoff" and the "ghost atom cutoff" should be printed out
when a command is issued that requires a system and communication
initialization, e.g. a run command.

Also, I get a same error with "fix rigid/small".

it has nothing to do with fix rigid.

axel.

The comm_modify command has an option

to extend the cutoff further than the pair

style requires, if needed.

Steve