NEB calculation don't run with ReaxFF force field

LAMMPS LAMMPS Mailing List Mirror
1

Hi, Steve/All,

Is there a possible bug when running NEB with reaxff force field?

I was trying to run NEB calculations with reax force field and the program stops immediately with “floating point exceptions” . Both NEB calculations with other potentials and MD simulations with reax runs fine so I think this only happens when they work together. I compiled lammps on two different machines with two different sets of compilers , one on linux cluster with gcc+ifort and the other on cray XT3 machine with Cray CC/ftn compilers and the error persists. I’ve changed NEB string constant from 0.5 to 500 and none of them worked. I’ve also tried both minstyle and none of them work either.

Below is the script I used:

dimension 3

boundary p p p

newton on

units real

atom_style charge

neighbor 3.0 bin

neigh_modify delay 1

atom_modify sort 0 0.0

pair_style reax

read_data silica.dat

pair_coeff * * ffield.reax 6 3 2

…(groupings and fixes)

variable u uloop 8 pad

dump 1 all atom 100 dump.$u

dump_modify 1 scale no

atom_modify sort 0 0.0

min_style quickmin

timestep 0.1

neb 0.0 0.0001 40000 20000 100 silica.final

Cheers,

Yun Liu

Research Assistant

Material Science Program

University of Wisconsin - Madison

Room 201A, 1509 University Avenue

Madison, WI 53706, U.S.A.

Office: 608-262-6798

2

Maybe Aidan wants to comment on this. Have you
tried running a single MD calculation with ReaxFF from
both of the endpoints of the NEB calculation?

Steve

3

Yes I have. Both points have been tested with reaxff and none of them showed
abnormal energy. Actually energy minimization alone also works (with the
same min_style). It is somehow only NEB that failed to work.

Cheers,

Yun Liu
Research Assistant
Material Science Program
University of Wisconsin - Madison
Room 201A, 1509 University Avenue
Madison, WI 53706, U.S.A.
Office: 608-262-6798

4

I would suggest first of all making sure you are using the latest version of
LAMMPS, and compile everything from scratch, including the reax library. If
the problem persists, try to reduce it down to the smallest possible size
w.r.t. atoms, replicas, lines of script and send it to us.