NEB calculation with strage Relative energy

[juganta]

VTST Community,

In my first NEB calculation I observed very strange relative energy. The procedure I followed as mentioned in the VTST tools. I have optimized the IS and FS with standard VASP (the energy is almost same with VTST). In NEB calculation forces were converged for all the images, however, I got the data (nebefs.pl) as below:
Image Force Stress Volume Magnet Rel Energy
0 0.00000000 6.11124600 3526.99 0.00000000 0.00000000
1 0.00500100 9.78947900 3526.99 0.00000000 -138.22840000
2 0.01282100 8.71075500 3526.99 0.00000000 -138.07530000
3 0.02115400 8.56068000 3526.99 0.00000000 -136.92880000
4 0.01310700 8.31480200 3526.99 0.00000000 -137.26860000
5 0.02082400 8.92213100 3526.99 0.00000000 -137.34470000
6 0.00000000 0.00000000 3526.99 0.00000000 0.43010000

But the interesting observation is when I did Single Point calculation of that images I got different Relative energies, as follows:
0 0.0
1 1.2331
2 1.38191
3 2.18835
4 1.69337
5 1.67068
6 0.430029

My INCAR parameters are:
IVDW = 12 (DFT-D3 method of method with no damping)
LDAU = .TRUE. (Activate DFT+U)
LDATYPE= 2 (Dudarev; only U-J matters)
LDAUL = -1 -1 -1 2 (Orbitals for each species)
LDAUU = 0 0 0 0 5 (U for each species)
LDAUJ = 0 0 0 0 0 (J for each species)
LMAXMIX= 4 (Mixing cut-off; 4-d, 6-f)
# Global Parameters
ISTART = 1 (Read existing wavefunction; if there)
ISPIN = 1 (Non-Spin polarised DFT)
ALGO = VF
LREAL = AUTO (Projection operators: automatic)
ENCUT = 400 (Cut-off energy for plane wave basis set, in eV)
PREC = Normal (Precision level)
LWAVE = .FALSE. (Write WAVECAR or not)
LCHARG = .FALSE. (Write CHGCAR or not)
ADDGRID= .TRUE. (Increase grid; helps GGA convergence)
ISYM = 0
LASPH = T !use for LDA+U, hybrid or meta-GGA in f- and 3d-elements
# Budged Elastic Band (NEB)
LCLIMB = .TRUE.
LTANGENT = .TRUE. (following Graeme Suggestion with lower setup)
IMAGES = 5 (#of images without 2 endpoints,#of nodes=#of images/divisible,nodesOfGgroup/image)
NSW = 500 (number of ionic steps)
ISMEAR = 0 (gaussian smearing method)
SIGMA = 0.05 (please check the width of the smearing)
IBRION = 3 (do MD with a zero time step)
POTIM = 0 (Zero time step so that VASP does not move the ions)
SPRING = -5.0 (spring force (eV/A2) between images)
LCLIMB = .TRUE. (turn on the climbing image algorithm)
ICHAIN = 0 (Indicates which method to run. NEB (ICHAIN=0) is the default)
IOPT = 1 (LBFGS = Limited-memory Broyden-Fletcher-Goldfarb-Shanno)
# Relaxation
NELM = 200 (Max electronic SCF steps)
NELMIN = 6 (Min electronic SCF steps)
EDIFF = 1E-07 (SCF energy convergence; in eV)
ISIF = 2 (Stress/relaxation: 2-Ions, 3-Shape/Ions/V, 4-Shape/Ions)
EDIFFG = -0.025
# Surface dipole corrections
IDIPOL=3 ! 3 is for surfaces
LDIPOL= T


Would you please help me to find source of strange relative energies?

Thanks!

[graeme]

If you post a .tar.gz of your calculations, I can help debug.

[juganta]

Prof. Graeme,

Thanks for your reply. I have attached the compressed folder herewith.

[graeme]

Well, there is some mismatch with your endpoint calculations and your NEB calculations. Make sure to use the same version of vasp including the VTST code that you used for the band calculation for the endpoints. I'm not actually sure what the problem is, but just use the binary that you are using for the band to recalculate the endpoints. Also try to adjust your IALGO parameters so that the electronic structure converges along the band. In the band you can see that it reaches your limit of 400 electronic iteration without achieving electronic convergence. Again, I'm not actually sure why the endpoints differ in energy from the band by over 130 eV, but the fact that you are using a different version of vasp and one without the VTST code concerns me. Use the same version of vasp with VTST for both and if you find the same problem, please let me know.

[juganta]

Prof. Graeme, Thanks a lot for your suggestions. I ran the endpoints with the same version of VASP & VTST (as like NEB), however, I got the same energies as before. I also completed the NEB calculations with the same version of the VASP & VTST, the relative energy still now the same. Looks like different versions of VASP along with VTST do not affect the energies.

[juganta]

juganta wrote:
> Prof. Graeme, Thanks a lot for your suggestions. I ran the endpoints with
> the same version of VASP & VTST (as like NEB), however, I got the same
> energies as before. I also completed the NEB calculations with the same
> version of the VASP & VTST, the relative energy still now the same.
> Looks like different versions of VASP along with VTST do not affect the
> energies.

I just found that in the images OUTCARs LDAUU and LDAUJ are 0, although LDA+U is selected. The OUTCARs of the endpoints contains the U parameter 5 on Ti-3d. To check this I did the optimization of endpoints without U-parameter, surprisingly, I observed that the total energies are in the range of IMAGES total energies found in the NEB. I checked my INCAR files for both calculation and looks everything fine.

[graeme]

Ah perfect - yes, it is usually something simple like this. Well, good to hear that it's fixed.