bondch failed using reax/c in simulation using GCMC

Dear Lammps-users:

I have been doing simulations of atomic insertions in the GCMC ensemble using REAXFF parameters for materials. The pair style commands I have been using:

pair_style reax 6.0 0 0 1.0e-5
pair_coeff ** ffield.reax 1 2

However recently I am trying to use pair style hybrid to combine Van der Waal’s parameters for Carbon with Reaxff parameters for materials. As I can not use pair style reax with pair style hybrid, I am now using:

pair_style hybrid/overlay reax/c NULL checkqeq no safezone 1.6 mincap 100 lj/cut 10

pair_coeff ** reax/c ffield.reax A B NULL
pair_coeff 1*2 3 lj/cut 0.05 3.7367

pair_coeff 3 3 lj/cut 0.06 3.4147762

However as the simulation progresses, I am running into this error:

step1052521-bondchk failed: i=724 end(i)=20276 str(i+1)=20275

There seems to be a bad interaction between fix gcmc and the dynamic memory allocation in pair reaxc, at least under certain conditions. One workaround is to install the KOKKOS accelerator package, which contains a new improved implementation of ReaxFF, but with the same user interface as reax/c.


I recently tried with the KOKKOS package, but now I get the error message-
“Cannot yet use pair hybrid with Kokkos”

So I understand that at this moment there is no way to use pair style hybrid containing style reax/c with fix gcmc in a simulation, am I correct in this conclusion, or is there any other way I am overlooking?