Check the Correctness of NEB Calculations

[JJHuang]

Dear Community,

I recently undertook an NEB calculation for an FCC structure comprising 108 atoms and 5 distinct elements. I'm reaching out to seek insights and feedback on the correctness of my procedure, settings, and the results I obtained (VASP version 5.4.4).

Steps I followed:

1) Initial and Final POSCAR Files: I began by creating initial and final poscar files with one vacancy (107 atoms now), named POSCAR00 and POSCAR06. These denote a migration via the first nearest neighboring site.
2) Relaxation of POSCAR Files: I relaxed both POSCAR00 and POSCAR06. The key settings in my INCAR for this step were:
EDIFF = 0.0001
EDIFFG = -0.02
ENCUT = 450
IBRION = 2
ISIF = 2
#ICHARG = 1
ISMEAR = 1
SIGMA = 0.2
ISPIN = 2
3) CI-NEB Calculations with Relaxed Structures: Using the relaxed structures, CONTCAR00 and CONTCAR06, I created 5 images using nebmake.pl. For the CI-NEB calculations, I modified the INCAR settings as follows:
IBRION = 1
IMAGES = 5
SPRING = -5
LCLIMB = .TRUE.
ICHAIN = 0

Results:
From the neb.dat and nebefs.dat files, I observed the following::
neb.dat:
0 0.000000 0.000000 0.000000 0
1 0.524596 0.182031 -0.707022 1
2 1.049012 0.621648 -0.987728 2
3 1.572148 0.896909 0.006895 3
4 2.162715 0.633969 0.799269 4
5 2.751552 0.324707 0.595336 5
6 3.339337 0.434388 0.000000 6

nebfs.dat:
0 0.016613 -830.996100 0.000000
1 0.010667 -830.814100 0.182000
2 0.010935 -830.374500 0.621600
3 0.019744 -830.099200 0.896900
4 0.016421 -830.362100 0.634000
5 0.014184 -830.671400 0.324700
6 0.017808 -830.561700 0.434400

Upon analyzing, I discerned that the forces had converged. TI've attached the MEP.eps for reference, which depicts a local minimum between the two endpoints. This has raised questions regarding the accuracy of the results. Are there any evident missteps in the operations I executed?

While the observed local minimum in the MEP.eps might be characteristic of the system I'm simulating, what insights or implications does such a local minimum typically convey about the transition process?

I would be grateful if someone could review these steps and settings and provide feedback on whether they are appropriate for accurate NEB calculations. Any suggestions or corrections are welcome.

Thank you in advance for your time and expertise!
JJ

[graeme]

Relax image 05 (next to the final stat) and you should get a lower energy final state. Make sure that you don't have a constraint, such as ISYM, on in your final state calculation.

[JJHuang]

[quote=graeme post_id=25824 time=1697140681 user_id=2]
Relax image 05 (next to the final stat) and you should get a lower energy final state. Make sure that you don't have a constraint, such as ISYM, on in your final state calculation.
[/quote]

Hi graeme,

Thank you for replying!!! I relaxed the 05 image based on its CONTCAR from the previous neb calculation, the new nebefs.pl gives:

0 0.016613 -830.996100 0.000000
1 0.010667 -830.814100 0.182000
2 0.010935 -830.374500 0.621600
3 0.019744 -830.099200 0.896900
4 0.016421 -830.362100 0.634000
5 0.014184 -830.671400 0.419389
6 0.017808 -830.561700 0.434400

It increases the energy at the 05 image, and when I compare the relaxed cell of 05 image, it is very close to the image06 (the finial state). I guess t is due to no spring force there constraining the moving atom in 05 image.

In my neb INCAR, I set ISYM = 0; however, in the INCAR for cell relaxation (for POSCAR00 and POSCAR06), I did not set up ISYM (which should be 2 in my case). I tested that change the ISYM = 0 for cell relaxation, it does not change the final energy.

Best,
JJ