Nudged Elastic Band Set up with B3LYP

[gg33]

I have a 19-atom molecule (C,H,O only) in an 18 Angstrom cubic box and I'm trying to set up an NEB calculation. I did the following:

1) Relaxed the initial and final structures and called them POSCAR00 and POSCAR03
2) Made a parent directory containing the following files: KPOINTS, INCAR, POTCAR, POSCAR00, POSCAR03, and the vasp.sh script
3) Used the nebmake.pl script to generate 2 images so that I now have a POSCAR file in all of the following subdirectories: 00,01,02,03

My INCAR file for the NEB run with the B3LYP functional is:
SYSTEM = 2ACP
ISMEAR = 0 ! Gaussian smearing
NSW = 500 ! 500 ionic steps
LCLIMB = .TRUE.
LSCALAPACK= .FALSE.
IMAGES=2
SPRING = -5
ICHAIN=0
IBRION = 2 ! use the conjugate gradient algorithm
POTIM = 0.5
LHFCALC = .TRUE.
GGA = B3
AEXX = 0.2
AGGAX = 0.72
AGGAC = 0.81
ALDAC = 0.19
EDIFFG = -0.05

I'm wondering if this is set up correctly or if I am missing anything.

I've read that IBRION=2 usually fails to converge but it seems to be converging. I wasn't sure if I chose POTIM correctly or if it is necessary to include LSCALAPACK= .FALSE.

Thank you!

[graeme]

That could be ok but I would start with ~5 images to be safe and POTIM=0.1 with IBRION=3. You can switch to a different IBRION (e.g. 1) once the band has relaxed close to the minimum energy path.

SCALAPACK is optional and independent of the NEB calculation.

[gg33]

Could you please explain why I would want to start with more images? I thought less would make convergence easier.

Thanks for your reply!

[graeme]

You need a sufficient number of images to resolve the reaction path. For a linear path, only one image is needed, but in general five is a pretty safe place to start.

[gg33]

Thanks!

I also have one question in regards to convergence. My criteria for EDIFFG = -0.05 eV/A. If I understand correctly this is in reference to the CHAIN +TOTAL forces obtained by grep 'TOTAL' -A no.atoms OUTCAR

CHAIN + TOTAL (eV/Angst)
----------------------------------------------
0.02595 0.00667 -0.04955
-0.02838 0.00349 -0.03701
-0.02533 0.00341 -0.03680
-0.03096 0.00031 0.03256
-0.03429 0.00218 0.03985
0.04103 0.01190 -0.05209
-0.03306 0.00155 0.03835
0.02469 -0.00034 0.10074
0.02049 0.00458 0.09700
-0.04112 -0.00856 -0.05000
0.12643 -0.00365 -0.01414
-0.01342 0.00353 -0.02233
-0.01178 -0.00020 0.01944
0.05759 -0.00524 0.01095
0.02105 -0.00541 0.03180
0.02335 0.00397 -0.02538
0.00633 0.00047 0.06616
0.00548 -0.00609 -0.00162

If these forces do not drop to the convergence criteria within a reasonable amount of time would increasing POTIM to 0.5 and continuing to use IBRION=3 be a suitable approach to achieving convergence?