NEB convergence with ReaxFF

Hello everybody,

i am running NEB by using the final coordinates keyword. Both states are minimized but i cannot reach any convergence. I am using ReaxFF and tried different parameters. Even after 200k steps i still couldnt reach any convergence with 24 partitions on 2 proc.

The log file shows that my pathangle for the first replica is “nan’” and for the last one. What could be the issue here ? I am reading my minimized states with matlab to write the state file. The precision is set to 15 so that should not be the issue.

Attached you can find my input file, log file, data files and reaxFF file.

I am using lammps version of 24th december
ffield_reaxx_2_chon (21.5 KB)
log.lammps (1.5 MB) (3.0 KB) (159.5 KB) (74.6 KB)


i guess it should be “nan” for the first and last one since no angle can be calculated here.

Can anybody help me here ?

If you don’t get an answer and cannot find an explanation yourself, the next best thing is to look at the source code. I you look at the file src/REPLICAL/neb.cpp around line 656 to the end, you will see that for the first and the last replica, the code actually prints out the constant NAN.

Also, if you look at the description of the verbose output for the neb command, it says there:

The “path angle” (pathangle) for the replica i which is the angle between the 3N-length vectors (Ri-1 -
Ri) and (Ri+1 - Ri) (where Ri is the atomic coordinates of replica i). A “path angle” of 180 indicates that replicas i-1, i and i+1 are aligned.

There obviously cannot be a replica -1 or a replica N+1, so for the first and the last replica this property cannot be computed.

Thanks for the reply axel. I saw that too. Do you have an idea why my two states do not converge or how to improve that ? With less replicas the convergence is better (4 instead of 24).