Hi, I am using drip potential combined with reaxff for the h-BN system using pair_style hybrid/overlap command
For drip, the interlayer interactions are between B-B, N-N & B-N.
In the documentation for drip, it is specified that pair_coeff command takes 4+N arguments, where N is atom types of lammps. So in my case, it is 2, i.e. B N.
However, when I use this I am getting the error as
ERROR: Potential file is missing an entry (…/pair_drip.cpp:341)
Last command: pair_coeff * * drip BN_real.drip B N
When I replace B N with B NULL, it stops by an error: segmentation fault.
I double-checked the potential file and sure that I provided all.
Please advise.
Thank you.
BN_real.drip (449 Bytes)
lmp_control (1012 Bytes)
ffield.reax2.BN (17.9 KB)
in.reaxff2 (669 Bytes)
2bnsheet_small.data (16.9 KB)
there are multiple problems here:
your data file is bogus.
its box dimension in z-direction is so that the two layers are forming a single extra dense layer
also, it contains explicit bonds, angles, dihedrals, and impropers. LAMMPS is warning you about it. didn’t you notice?
while you are not defining any potentials, the existence of a bond topology will lead to bogus calculations due to the excluded pairs that are triggered by the default special_bonds setting of lj/coul 0.0 0.0 0.0
all bonded interactions in ReaxFF are implicit and the topology (re-)computed on the fly so there MUST NOT be any explicit bonds computed at the same time.
you should have noticed that things are bad by running ReaxFF without hybrid/overlay and drip.
the error about the missing entry is due to the drip potential expecting not only a ‘B N’ entry but also an ‘N B’ entry.
whether that is intended or not, is something you need to discuss with the author of that potential. email address is in the source code and in the src/USER-MISC/README file.
the segfault is triggered in part due to the bogus data file, but also due to some obscure and hard to track down bug in the USER-REAXC code. if you try the KOKKOS version, it should work (if it wasn’t for a bug in the drip potential code which does not consider
that there may be a pair style hybrid/overlay/kk).
why even use ReaxFF on this kind of system in combination with drip??
if the ReaxFF parameterization was properly done, the inter-layer contribution should be properly included as well.
those inter-layer potentials (and we have a whole zoo of them in LAMMPS now) are designed more for the use with
potentials like tersoff or airebo, that are lacking those inter-layer interactions in their models.
axel.
Thank you for your reply. But when using a drip potential, the atom style should be full or molecular to indicate two different layers, that is why the atom style full is employed.
you should have noticed that things are bad by running ReaxFF without hybrid/overlay and drip.
I have done ReaxFF with atom style charge, It’s working fine, but I didn’t try with atom style full.
why even use ReaxFF on this kind of system in combination with drip??
if the ReaxFF parameterization was properly done, the inter-layer contribution should be properly included as well.
Since we took the parameters for ReaxFF from Dr. Adri van Duin and the drip parameters are taken from this paper https://arxiv.org/pdf/1910.12805.pdf
We thought ReaxFF and drip is a good start as we have parameters for sure.
I will try with the KOKKOS version and look at other potentials too.
Thank you.
Thank you for your reply. But when using a drip potential, the atom style should be full or molecular to indicate two different layers, that is why the atom style full is employed.
that is completely irrelevant. you can use atom style full, but do not have to have a bond topology.
also, you could add (only) molecule id to atom style charge using fix property/atom.
you should have noticed that things are bad by running ReaxFF without hybrid/overlay and drip.
I have done ReaxFF with atom style charge, It’s working fine, but I didn’t try with atom style full.
it runs, even with the bond topology enabled, but due to the overlapping layers and the excluded pairs, the energies and forces are bogus.
axel.
Dear adupa,
I see you are referencing our paper. I’ll send you a typical input file using tersoff and rebo for intralayer and drip for interlayer. We’ve updated the drip parameters since posting this arxiv so I’d recommend using the new ones. Please give me a day or two to ask my colleague for his most recent fits.
Regards,
Nicolas