Kokkos Reax Package Development Plan

Jim,

We are currently working on implementing the Extended-Lagrangian scheme for fix qeq in LAMMPS USER-REAXC, see the discussion here: https://github.com/lammps/lammps/issues/507. Once this is in the regular version of LAMMPS it could easily be added to the Kokkos version as well. Aidan, anything to add?

Stan

Jim,

Your request is noted. There are no fundamental barriers to add ACKS2 and eReaxFF to LAMMPS. However, it is a substantial amount of work and is not currently a high priority.

Aidan

Stan,

Thank for your response. Improving the performance of qeq is a worthy goal.

Unfortunately, the qeq method (and EEM) has a few drawbacks, the worst being its inability to restrain long-range charge-transfer, even between molecular fragments that are well separated, or by unphysically reducing the charge on species like OH-, Na+, etc. in a biomolecular environment.

ACKS2 significantly reduces these problems and is why the current ReaxFF method incorporates it as the charge equilibration methodology. Some descriptive overview can be found here:

https://www.nature.com/articles/npjcompumats201511

And the references cited therein go into more detail.

An independent, coded implementation of ACKS2 can be found here:

https://github.com/tovrstra/validation_eem_acks2_reaxff

I would suggest it to you for your inspection to see if it might facilitate ACKS2 implementation in the Kokkos package.

Thanks for your consideration.

Jim