[lammps-users] Discrepancy between Reax/F and Reax/C on a small system

Hi,
I have compared the potential energy obtained by Reax/F (original fortran version) with that obtained by Reax/C on a small Pt surface.
I found that there is a discrepancy of around 3 kcal/mol. This can be attributed to the vdW energy term, whereas all other energy terms are (nearly) identical.
This difference disappears when I make the system a cluster (i.e., I remove the in-plane periodicity).
The in-plane size of the system is smaller than the cutoff radius (I think it is set to 10 Ang), so not a good system. Yet, I’d expect the energy to be the same. It seems to me the discrepancy has to do with the periodic BCs.

I am attaching the log files.

Regards,
Paolo Valentini

log.lammps.REAXC (2.24 KB)

log.lammps.REAXF (2.24 KB)

Thanks - this is likely something Aidan and Metin
will want to look at. I've CCd them.

Steve

Hi Paolo,

Thanks for spotting this. I believe that pair style reax handles
interactions with multiple periodic images correctly; perhaps reax/c does
not. You can verify what the correct answer is by using the replicate
command to explicitly generate multiple periodic images. The energy per atom
should be unchanged. Let us know what you find.

Aidan

Hi Aidan,
thanks for your reply. I doubled the system in the x and y directions,
and indeed the difference in potential energy (to be precise, in the vdW energy term) is gone. Attached are the log files.

Paolo

log.lammps.REAXC_replicate (2.36 KB)

log.lammps.REAXF_replicate (2.36 KB)

Hi Paolo,

That clearly seems to be an issue with multiple periodic images. I will take a look at it later this week.

Thanks,
Metin

As I expected, the van der Waals energy/atom is unchanged for Reax/F (54.269
kcal/mol), but for Reax/C it is higher for the small system (54.384
kcal/mol). Metin, do you know if there is a way to fix this in Reax/C?