NEB not reporting Reaction Coordinate Values

LAMMPS LAMMPS Mailing List Mirror
1

Lammps Users,
I am running into NAN values for the reaction coordinates. I only seem to have this problem if the system converges before the maximum number of iterations specified in the neb command are met.

Any ideas would be appreciated.

INPUT:

variable N string “I1”
variable A string “I1.1000K”
variable c equal 3.5710425743336587
variable t equal 1000.0
variable u uloop 6

units metal
boundary s p p
atom_style atomic
lattice fcc $c

Read data and set up simulation

atom_modify map array sort 0 0.0
read_data min.0.data

EAM potential – Ni

pair_style eam/alloy
pair_coeff * * potentials/ni1.set Ni Ni Ni Ni Ni
mass * 58.710

neighbor 1.5 bin

group LB type 4
group RB type 5
group GB type 1 2
group GBL type 1
group GBR type 2

thermo 5
thermo_modify lost warn flush yes

Computes

compute CENTRO all centro/atom fcc
compute CNA all cna/atom 3.08

Prevent rigid body motion due to non-periodic x-boundary

fix lbforce LB setforce 0 NULL NULL
fix rbforce RB setforce 0 NULL NULL

dump gdump all atom 100 dump.$A.neb.$u
dump_modify gdump sort id
timestep 0.001

#neb required commands
fix nebatoms GB neb 0.001 #set active atoms
min_style fire

#neb run
#neb with intermediates given
variable i equal part
neb 1.0e-12 1.0e-12 14 14 7 each coord.$i.xyz

MASTER LOG FILE
LAMMPS (7 Jan 2014)
Running on 6 partitions of processors
Step MaxReplicaForce MaxAtomForce GradV0 GradV1 GradVc EBF EBR RDT RD1 PE1 RD2 PE2 … RDN PEN
0 1.1742183 0.85733174 0.1515361 1.0321598 1.1742183 0.09060015 0.024176807 0 0 -143622.28 nan -143622.22 nan -143622.19 nan -143622.21 nan -143622.23 nan -143622.22
7 1.1742172 0.8573311 0.15153589 1.0321588 1.1742172 0.090600026 0.024176778 0 0 -143622.28 nan -143622.22 nan -143622.19 nan -143622.21 nan -143622.23 nan -143622.22
Climbing replica = 3
Step MaxReplicaForce MaxAtomForce GradV0 GradV1 GradVc EBF EBR RDT RD1 PE1 RD2 PE2 … RDN PEN
7 1.1742172 0.8573311 0.15153589 1.0321588 1.1742172 0.090600026 0.024176778 0 0 -143622.28 nan -143622.22 nan -143622.19 nan -143622.21 nan -143622.23 nan -143622.22
14 1.1742161 0.85733045 0.15153568 1.0321578 1.1742161 0.090599901 0.024176749 0 0 -143622.28 nan -143622.22 nan -143622.19 nan -143622.21 nan -143622.23 nan -143622.22

2

you’re getting Nans on timestep 0, so it
doesn’t look like it has to do with convergence.

Are you certain your initial configs are all valid,
since you are using the “each” keyword? Do you
still get Nans if you only read in a final
configuration?

Steve

3

Steve,

I have run into the same problem.

Using my old LAMMPS executable with the “old” neb fix, I can get a simple system to converge just fine (i.e. finds the correct MEP with the correct reaction coord.).

Using the new LAMMPS (Dec 31) executable with the “new” neb fix, I am also observing nan values for the reaction coordinates. As a result of this, the NEB seems to collapse all the intermediate images onto the starting and final images. I only made the changes necessary for the final coordinates file (just added how many lines at the top) and I added the keyword “final” to the neb command when I ran the new executable.

Best regards,

Ross

4

If someone can post a simple example, I’ll take a look.

E.g. small # of atoms, small # of replicas.

Steve