lj/cut Giving Bond Atoms Missing at Proc 0 Error in A Polymer Model


I am currently using stable LAMMPS from March 3 2020 (just got it built!) with the MOLECULE and USER-CGDNA packages (necessary to run the in file). I’m running simulations using a STEP 1 soft potential + harmonic bonds, and then going into a STEP 2 lj/cut potential + fene bonds; in both cases using cosine/periodic angle style.

My problem is when I change to the lj/cut potential in STEP 2, it always immediately gives “bond atoms missing at proc 0 at step 0 (or step 1, never many steps)” error. It’s not immediately clear to me the possible source of the issue. For example, I looked at a lammpstrj file, and the two atoms in question in the current iteration of error are only 1.31 units apart, which is within the neighborhood size and the bond length. If I just run the soft potential, these issues don’t come up either. And it doesn’t have to do with the harmonic vs fene bond either it seems - if I run the simulation at both steps with fene bonds, the same pattern occurs. And looking at a visualization in VMD of the STEP 1 dynamics, everything looks as expected, and looking at the configuration right before STEP 2 starts, it looks fine overall.

The two problem atoms are always in a polymer together, and are always adjacent-in-sequence to each other (ie 24561, 24562). Since the issue comes up only when I introduce the lj/cut pair potentials (I have 4 types of atoms, and I define an lj pair potential between each type), I feel like I’m missing something (as opposed to being a bond potential problem).

I’m running a closed box with an lj wall potential as well. Messing with neighborhood bin size didn’t change anything, and neither did running STEP 1 longer (ie 20,000 iterations vs 2,000).

I’m wondering any thoughts on what the issue may be.

I’ve attached the involved parts of the in.chrom file being used. The data file loads in a polymer of about 35,000 beads and 800 “free-floating” atoms, following the parameters of this paper: https://www.biorxiv.org/content/10.1101/2020.05.12.091041v1.full.


in.chromnocomment (1.29 KB)

three things to experiment with are the height of the soft potential, the stiffness of the bond potential, and the duration of the “unoverlap” phase.
if you have overlapping bonds/polymers or need to mix things up a bit more, then perhaps using fix adapt to gradually increase the height of the soft potential over time, may be the ticket to get to a better state.

another option to consider would be to run a minimization after switching the pair style to remove any “hot spots”.


Thank you, I ran the fix adapt, and it didn’t fix the issue, but then I ran fix adapt in conjunction with minimization, and now it is working smoothly! I appreciate the help.


1 Like