[lammps-users] ghosts of deleted bonds

Hello all, I am having the familiar problem occurring when LAMMPS is run in parallel.

“ERROR on proc 3: Bond atoms 15110 15120 missing on proc 3 at step 21754”

Typically, the solution is to adjust the skin value, etc. However, my case is different. I am doing a simple simulation of a chain of beads, ie bead spring polymer type simulation. In addition to a bond to attach my beads into chains (bond type 1), I have assigned a bond type (type 2) to tether the ends of various chains for a pre-equilibration step. From my data.xxx file:

Bond Coeffs

1 500 1.0
2 50 1

After running a pre-equilibration, I use the following command in my input script to turn off the tethers:

delete_bonds all bond 2

Now, with the tethers turned off, I allow the simulation to run for a long time. Unfortunately, I get the error message about the Bond atoms missing… This means that despite the fact that I used delete_bonds to get rid of the type 2 bonds, processors are still trying to communicate information about these bonds. Of course, after the tethers are turned off, beads that were previously connected by tethers will wander far away from each other. This is exactly what I intend to happen. However, this diffusion of previously bonded atoms seems to be crashing the program.

(Incidently, I do get some output that the program has succesfully “deleted” the bonds:
Deleting bonds …
15360 total bonds, 14208 turned on, 1152 turned off

Can I truly delete these bonds, or will the neighbor searching always expect that the atoms of the deleted bonds must remain within a certain cutoff distance?


You want delete_bonds ... remove, and possibly special. The
doc page describes these options.