ERROR on proc 3: Bond atoms 1888 1943 missing on proc 3 at step 3486 (src/ntopo_bond_all.cpp:61)

LAMMPS LAMMPS General Discussion
, ,
1

I am working on a system where a hydrogen transfer reaction occurs from a phenolic group to a radical, involving both forward and reverse reactions (two reactions). I am using the REACTION package with the REACTER protocol in LAMMPS to simulate this process.

Overview of my setup

  1. Prepared the system by performing atom typing, parameterization, and merging the reactant files into a single simulation input using LUNAR.
  2. Created a proper simulation box along with the pre- and post-reaction map files required by REACTER.
  3. Used the CVFF force field for these simulations.

Issue encountered

The problem I am facing is that, for the same molecule, the simulation fails inconsistently. Specifically:

I ran three simulations with different random velocity seeds and atomic positions (i.e., different realizations of the same system). Two simulations completed without any issues, but one failed with the following error:
ERROR on proc 3: Bond atoms 1888 1943 missing on proc 3 at step 3486 (src/ntopo_bond_all.cpp:61)

This randomness in error occurrence makes it challenging to identify the root cause.

Unfortunately I caannot attach anything since I am a new user. You can find the files here. Any insights on what might be causing this inconsistency, or suggestions on how to resolve it, would be greatly appreciated.

Sometimes I also got the following error:
ERROR on proc 38: Bond/react: Fix bond/react needs ghost atoms from further away (src/REACTION/fix_bond_react.cpp:1481)

2

@sahmed73 I cannot answer any questions related to fix bond/react (that is @jrgissing 's domain), but you should provide the following information to make it easier to provide assistance:

  • which LAMMPS version are you using exactly? what platform are you running on? and what is the exact command line (i.e. how many MPI processes are you using, any accelerators/threads and how many)?
  • I didn’t see any mention of equilibration. For how long have you been running equilibration (without fix bond/react)? and did you encounter any warnings or other issues in the process? How did you confirm that your equilibration was sufficient?
3

Your reaction templates are much larger than necessary. I highly suggest making the templates as small as possible and using ‘edge atoms’, which are atoms that connect to the rest of the molecule. This correction should solve the ghost atoms issue, and may also help the missing atoms. You will probably be able to decrease your stabilize_steps parameter.

4

Thanks @akohlmey

  • LAMMPS version & platform: 29Sep2021 (KOKKOS build) with Intel + OpenMPI 4.0.6 on our HPC cluster.
  • Command line & resources: mpirun -np 52 lmp_kk -k on -sf kk -in input.in (1 node, 52 MPI ranks, 1 OpenMP thread).
  • Equilibration: Performed NPT equilibration until density stabilized, then switched to NVT until potential energy plateaued (both without fix bond/react).
  • Warnings/issues: None observed during equilibration.
5

Thanks a lot for the suggestion! I will definitely give it a try soon and play around with the stabilization parameters you mentioned.

6

This is very old given the development progress of LAMMPS and also the REACTION package.
You are missing out on man improvements and bugfixes. We just released version 22 July 2025. That is almost 4 years.