NEB - ReaxFF convergence issue


I am trying to perform a NEB analysis of the CH4 + OH -> CH3 + H2O reaction, using the attached ReaxFF. I am using the lammps-7Aug19 version. It looks like I have to set intermediate replicas, or else the NEB will result in the energy barrier of the reaction CH4 -> CH3 + H, when inputting the initial and final coordinates.

However, when I specify “each” replica in the neb command I get no convergence, even for very large N1 and N2. I attach the intermediate replicas I used (file: coord_data_10.tar.gz), total number of 9, although I also tried with 4, 6 and 20 with no success. From the literature it seems that the transition path, LAMMPS tries to converge to, is close to the expected one, so I believe it is a convergence issue. Do you have any suggestions on how to improve the convergence?

Thank you very much for your help in advance,


The input script I use is:

units real

atom_style charge

atom_modify map array

read_data data.ch4

pair_style reax/c lmp_control

pair_coeff * * ffield.reax C H O

fix 1 all qeq/reax 1 0.0 10.0 1e-6 reax/c

neighbor 2 bin

neigh_modify every 1 delay 0 check no

thermo_style custom step temp press etotal pe ke

thermo 1

fix ensemble all nve

timestep 1
variable i equal part

dump dumpsim all atom 10 dump_data/dump.neb.$i

min_style quickmin

fix nudgf all neb 1.0 parallel ideal

neb 0.0 0.01 100000 100000 100 each coord_data_10/coord.initial.$i

coord_data_10.tar.gz (1.28 KB)

data.ch4 (632 Bytes)

ffield.reax (7.78 KB)

What do you mean exactly by “get no convergence”: how are the forces evolving? how is your system behaving? There can be lot of reason for what you observed: slow convergence, spring constant too stiff/soft, potential complexity, not relaxed initial/final structure, etc.
For example, adding spring to perpendicular nudging force (perp option) sometime improved/stabilized the convergence for complex potential energy landscape.