[lammps-users] bond-order information written by reaxFF in Lammps

Dear Lammps Users
I have a question regarding the bond-order information written by reax forcefield in Lammps (using fix reax/bonds command in the Lammps input file). I am trying to model the thermal decomposition of phenol (I have 40 monomers in a periodic box built at a target density of 1.1g/cc). After the minimization, NVT and NPT equilibration I have an NVT production run. From this run I am taking the bond order information file generated by Lammps and I analyze it using my own tool to extract information regarding the number of various molecular fragments that are forming. To correctly perform this analysis, the atoms need to be in their default configuration (not packed back in the unit cell).
Do I need to use a special flag when running Lammps to ensure that the bond order analysis is performed on the default atom configuration? Or no matter what flag is being used in the LAMMPS input file, the fix reax/bonds command will by default use the not-unit-cell-packed representation? Rephrasing: is the bond order file taking into account the periodicity of the structure and if so, how exactly?

Thank you,

Balaji Shankar VenkatachariMechanical Engineering

Aidan will have to answer this Q.


Given a periodic cell, and given the position of one periodic image of each
atom, the potential energy of the system (per unit cell) should be uniquely
defined. This is true for any properly-constructed potential energy
function, including ReaxFF. To satisfy this requirement, the reaxFF bond
orders should also be uniquely defined. It is only the positions of the
atoms that are not uniquely defined, since there are actually an infinite
number of periodic images of each atom.

So, yes, LAMMPS does account for periodicity when computing bond-order. The
way that it does this is a general feature of LAMMPS, and is not specific to
ReaxFF or bond-order. Given a user-specified periodic cell, each atom is
first mapped to the periodic image in the cell, i.e. xlo <= x < xhi. Second,
given a maximum interaction distance, all necessary images of these atoms
are generated as "ghost atoms". In this way, bond-orders due to periodic
interactions are represented by local-ghost interactions.

The same scheme is used for parallel decomposition.

A very stringent test of correctness is to shift the periodic cell. The
effect on bond-order (in the first snapshot) should be down to machine
precision. Similarly, changing the processor decomposition should have a
similar effect.