box image problem of fix bond/swap with periodic boundary

LAMMPS LAMMPS Mailing List Mirror
1

Dear All,
I am a new graduate student trying to simulate a linear polymeric melt system using LAMMPS. I use the bead spring model, and the trial system is 100_chain_length * 100_chain_number, the density is 0.85 at T=1.0.
I used the bond/swap fix to relax the system for several million timesteps. I wanted to calculate the end-to-end distance of the chains to monitor how the configuration of the system was changing with time and that is where I encountered a problem. For a given chain, that was involved in bond swapping, while the unscaled atom coordinates look fine in the most part, the box images don’t make sense since generating actual positions of beads using those two information results in unrealistic bond distance between bonded beads. I think, since bond swapping is done based on neighbor list (that is using unscaled positions of the beads) I wonder if the box images of the beads are properly taken care of after bond swapping happens between two chains. Any advice on how to solve the problem I’m having is highly appreciated.

Thanks,
Shichen Yuan

2

I wonder if the box images of the beads are properly taken care of after bond swapping happens between two chains

no, they are not. And it would not be easy to do, especially in parallel,
since a swap could affect
image flags of monomers very far from where the swap occurred. I'll add
a note to the doc page about this.

I think your best solution is to simply post-process what you want. If you
dump monomer coordinates, it is easy to walk the chain and figure out
when you cross the box.

Steve