NEB basic questions

[hykim]

Hi sir,

I made the approx-MEP with 4-images generated by nebmake.pl script.
In my knowledge, if I use nebmake.pl then each image have equi-inter distances.
The distance between two specific images however was severely elongated after NEB run. After subsequent runs, I found there is the TS between these two images.
Why was the initial distance between two images elongated too much? Did I use too strong spring constance?

One more thing, I have another NEB run with 4-images, and it always finishes after 1st ionic relaxation step. Is this physically meaningful?

Many Thanks,
HY KIM

[graeme]

The images can become unequally spaced if you are using the climbing image. But if the saddle is between your widely spaced images, it sounds like something is wrong. Similarly for a linear band that converges right away - something is wrong. But I can't tell what's wrong without more information; if you send one or both of these run, I'll take a look at them.

[hykim]

Thank you for your reply.

I'm just doing NEB with as implemented form in VASP. I have know idea whether it is CI-NEB.

I found that for both cases, the tangential force and chain+total force are too high (somtimes 4~5 eV/Ang).
Yes, I'm feeling that something is wrong.

I know EDIFFG does not accouts for this. Could I set any criteria for these forces such as EDIFFG?

Following is my INCAR setting and some part of OUTCAR.

GGA=91
ISPIN=2 ! spin polarized calculation (2-yes 1-no)
NUPDOWN=1

PREC=Accurate ! precission:medium h
ENCUT=300.00 ! Plane wave cut-off energy
EDIFF=0.0001 ! stopping criterion for electr updt
EDIFFG=0.0001

ISMEAR=0 ! part.occupancies:-5 Blochl -4 tet -1 fermi 0-gaus
SIGMA=0.05 ! broadening in eV -4 tet -1 fermi 0-gaus
LREAL=AUTO ! Wavefunction projection in real space, automatic optimization of the projectors
NELMIN=4 ! Minimum number of minimizations to ensure accurate forces
NELM=120
IDIPOL=3 ! Dipole correction to energy along the Z direction
LMAXMIX=4 ! l-decomposed charge densities up to LMAXMIX are writ
ALGO=F ! algorithm:Use only 8(CG) or 48 (RMM-DIIS)

NSW=500 ! number of steps for ionic update
ISIF=1 ! relax atom positions
IBRION=1 ! ionic relaxation:0-MD 1-quasi new 2-CG
ISYM=0 ! symmetry:0-nonsym 1-usesym

IMAGES=4
SPRING=-5

-----------------------
energy of chain is (eV) 0.000000 for this image 0.000000
tangential force (eV/A) -3.127698
left and right image 0.648639 0.648639 A
TANGENT CHAIN-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
-0.02026 0.03698 0.00132 -0.063383 0.115652 0.004142
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
0.00000 0.00000 0.00000 0.000000 0.000000 0.000000
.
.
.
-----------------------------------------------------------------------------------
CHAIN + TOTAL (eV/Angst)
----------------------------------------------
-0.83914 0.31292 0.73985
0.00000 0.00000 0.00000
0.00000 0.00000 0.00000
0.00000 0.00000 0.00000
-0.02521 -0.02240 0.01466
0.33743 0.02946 -0.96081
-0.02797 0.00998 -0.18864
-0.41221 0.07330 0.11792
-1.29423 -0.87525 -1.78743
-4.56655 4.49707 2.29845
-0.17257 -1.96654 0.38647
2.61399 1.17989 -0.70233
0.29412 -0.66671 0.59236
-0.01940 -0.22209 0.18090
.
.
.
Many Thanks

HY KIM

[andri]

If you are using the elastic band method as implement by the VASP developers, you have answered your own question. This forum only provides assistance with issues related to code from http://theory.cm.utexas.edu/vtsttools . For other VASP related questions, turn to the official VASP discussion forum.

P.S.
Preferably you should start using the implementation provided by http://theory.cm.utexas.edu/vtsttools