modeling a rod made of atoms with point-dipoles

Dear Lammps users and developers:

I wish to study the dynamics of a group of rigid rods in liquids. Each rod consists of five atoms A-B-C-D-E. Atom “C” has a point-dipole, which always points from atom “C” toward “D”. The dipole-dipole interactions between rods drives interesting dynamics. I have some ideas but also questions on how to model the problem, and wonder if you may give some advice.

My plan is to
– model the rod as a chain of five atoms
– assign a point dipole to atom “C”
– use lj/cut/dipole/cut for interactions
– use fix/rigid to update the rod

My questions are
– does the above scheme allow the torque on atom “C” (due to dipole-dipole forces) be used in computing the rotation of the rod?
– how can the dipole of atom C be updated when the rod rotates? It is clear that dipole moment can be updated (e.g., fix nve/sphere update dipole), but the manual does not say much about how the dipole moment is actually updated.

Thank you very much.


yes, what you describe should work. Note that you should

not update any atoms in the rod with fix nve/sphere or any

other integrator besides fix rigid (or better yet fix rigid/small,

or any of its variants).

Fix rigid (or small) should update the position of all the

atoms in each rod and the dipole orientation for particle C.

You can verify this by dumping the particle attributes

including the dipole direction and verifying it always points from C to D.

You can also use the relatively new compute rigid/local with dump local

to get the orientation of each rod.

These would be good to verify b/c fix rigid has code for updating

the dipoles of body particles, but it may not have been used/tested