NEB is diverging from Minimum energy path for ADP potential

Hello Respected all,

I am using fix NEB and command NEB to find the void migration energy barrier (MEB) in pure Fe, pure Ni & Ni3Fe for ADP,EAM(Bonny) potential.

Pure Fe
The MEB for Fe void in BCC Fe lattice is correctly calculated by subroutine for both potentials.

Pure Ni

EAM potential gave the result but for ADP potential, NEB dynamics did not converged and gave some unrealistic values.

Ni3Fe (Fe void Or Ni void)
Results for EAM potentials are reasonable.
For ADP potential run didn’t converged.

Nomenclature
vFe_mNi stands for : void at Fe site and Ni atom is migrating to the Fe void site

I need suggestions regarding how to control this divergence for MEP (minimum energy path) .

out.dat (1.14 KB)

in.final (90 Bytes)

in.neb.vNimNi (1020 Bytes)

in.neb.vFemNi (1.02 KB)

vFe_mNi.dat (9.37 KB)

vFe_mNi.final (68 Bytes)

Hello Respected all,

I am using fix NEB and command NEB to find the void migration energy
barrier (MEB) in pure Fe, pure Ni & Ni3Fe for ADP,EAM(Bonny) potential.

_Pure Fe_
The MEB for Fe void in BCC Fe lattice is correctly calculated by
subroutine for both potentials.
_
_
_Pure Ni_
EAM potential gave the result but for ADP potential, NEB dynamics did
not converged and gave some unrealistic values.
_
_
_Ni3Fe (Fe void Or Ni void)_
Results for EAM potentials are reasonable.
For ADP potential run didn't converged.

_Nomenclature_
vFe_mNi stands for : void at Fe site and Ni atom is migrating to the
Fe void site

I need suggestions regarding how to control this divergence for MEP
(minimum energy path) .

When the path is complex and/or the number of NEB intermediate images is
not high enough to well describe this path, the path can create acute
angles that could lead to instabilities.

This could happen, for example, with complex potential energy surface or
complex structure or complex mechanism, etc.

One of the solution could be to use the "perp" option of the "fix neb"
command, that provide additional force to the path, helping keeping it
smoother. It is all describe in the documentation, with the useful
referee to Maras et al.

https://lammps.sandia.gov/doc/fix_neb.html

Julien

Respected sir,
I used the prep with elastic constant 4.0 and parallel with elastic constant 10.0.
But still it did not converged for ADP.

Respected Sir,
I am trying void migration energy for Ni with ADP potential with 16 processors.
Parallel force spring constant value is 10 and perpendicular force spring constant value is 4.
N1 and N2 both are 100000.
only 1 particle is moving from its location to void co-ordinate in x-direction since its (110) oriented system.
It did not converged.

Then I changed the initial position of only moving particle to somewhere in the middle of the earlier initial and final position.
It is still not converging. What can be the probable reason for this Or how can I resolve this.

Seeking for the some guidance.
Thank you.

in.neb.Fevac (1.03 KB)

in.final (65 Bytes)

log.lammps.0 (21.1 KB)

out.dat (104 KB)

log.lammps (104 KB)

Respected Sir,
I am trying void migration energy for Ni with ADP potential with 16
processors.
Parallel force spring constant value is 10 and perpendicular force
spring constant value is 4.
N1 and N2 both are 100000.
only 1 particle is moving from its location to void co-ordinate in
x-direction since its (110) oriented system.
It did not converged.

Then I changed the initial position of only moving particle to
somewhere in the middle of the earlier initial and final position.
It is still not converging. What can be the probable reason for this
Or how can I resolve this.

I feel it is very difficult to answer with these information.

A curious aspect of your problem, is that the system converge with an
EAM potential, but not with and ADP potential.
Is you ADP potential correctly parametrized? Did you test the ADP with
"simple" minimization?

BTW, did you find some articles in the litterature that are performing
NEB calculations with the ADP potential?

Julien

Sir,
It converged for migration barrier for pure Fe vacancy using ADP.

Mishin (Developer of ADP for Ni-Fe system) himself has used NEB for Fe migration barrier but for Nickle void migration, he has not explicitly mentioned the method of calculating migration barrier energy.

On minimization, ADP gives value for configurational energy and lattice could start that matches well with literature.

Sir Dr. Julien ,

Continuing the above discussion I am attaching the
Basic input file and NEB output ‘out.dat’ file .
I am also attaching one excel sheet , showing results for various settings for spring constants and number of processors.

One observation which is concerning me that the reaction co-ordinate (st1-rc & st2-rc in excel sheet and RDT in out.dat) is sometimes gaining very high value.

What that exactly mean?
Or in some cases values of EBF/R are very high !!
Can’t figure out the reasons to resolve this.

Please see if this can be resolve.
Thankfully

out.dat (1.14 KB)

in.neb.vNimNi (750 Bytes)

NEB observation.xlsx (9.04 KB)

log.lammps.1 (6.5 KB)

screen.1 (5.83 KB)