[lammps-users] Reax vs Reax/c for Nanotube Reaction Simulation

Reax Developers,

I’m currently using the reax force field as implemented in the LAMMPS molecular dynamics package (Feb 11 2011). I’m simulating an all-carbon system similar to “Development of the ReaxFF Reactive Force Field for Describing Transition Metal Catalyzed Reactions, with Application to the Initial Stages of the Catalytic Formation of Carbon Nanotubes” by Neilson et al. and am using the parameter file included in the supporting info for that paper.

I am interested in verifying/comparing the two different ReaxFF implementations in LAMMPS, reax and reax/c. I have recreated most of Table 1 and Figure 1 from Neilson et al. with my results from the LAMMPS reax and reax/c codes included. I thought you might be interested in seeing how they compare, so I have attached the figure and table as well as a few other relavent figures I generated. Overall, they are very similar. Enjoy!

In running these codes there are a few things that caught my attention. Please feel free to comment:

  - For large carbon structures shown in Table 1 (nanotubes, graphene, C60 Buckyball) energies are EXACTLY the same between the fortran and C++ codes and are SIMILAR to the data published in Neilson et al. Excellent.

  - For small cyclic all-carbon molecules the fortran and C++ codes give EXACTLY the same energies. However these energies are, in general, HIGHER than those reported in Nielson et al. (This could be due to my molecules being at a local minimum)

  -For small acyclic all-carbon molecules the Fortran and C++ codes give DIFFERENT bond energies / bond orders.

  -For C2 the C++ code shows a lone-pair energy of ~14 kcal/atom. The lone pair energy should be 0 since the parameter P_lp2 is 0 for carbon. I see in the C++ code that the C2 correction energy is output in with the lone pair energy. I assume this is not true for the fortran code? I'm not as fluent with fortran as I am with C++.

  -What energy terms are included in “atom energy”? Undercoordination and overcoordination terms?

I can send you my input scripts, if you’re interested in seeing them, as well as my excel sheet which has additional figures. Thanks your work on ReaxFF!

Benjamin D. Jensen
PhD Student
Dept of Mechanical Engineering - Engineering Mechanics
Michigan Technological University
931 RL Smith
1400 Townsend Dr.
Houghton, MI 49931
email: [email protected]...

LAMMPS Reax Codes Comparison JENSEN.pdf (347 KB)

This looks like a nicel, careful comparison, both of the two versions in
LAMMPS and to the literature. Thanks for sending the data.

Aidan may want to comment on the details.

Steve

The problem that Ben identified with the C2 correction has been resolved by
a small change in pair reax/c, which Metin had already made in December, but
did not get in to the repository. I commited it yesterday. Here is Ben's
comment on the new version:

------ Forwarded Message

From: Benjamin Jensen <[email protected]...>
Date: Wed, 9 Mar 2011 13:41:02 -0700

:

I included the new reaxc_multi_body.cpp and now both
reax and reax/c report the exact same total & lone-pair energies for a C2
molecule. Also, the small differences in energies I was seeing for small
acyclic all-carbon molecules are gone and reax and reax/c are now reporting
exactly the same energies for those molecules as well.

: