My Community

Hello All,

I have recently performed a relatively simple electronic-structure calculation in VASP, and obtained a not-so-simple to interpret series of outputs.

Simulation Detail:

Structure - A single MnII(Ligand)6 (+2) molecule, without counter-ions, from experimental geometry in a near-perfect octahedron. Very large bounding-box to remove replica interaction. Gamma point, GVASP 5.2.11

INCAR -
[Spin-Polarized, symmetry-projected, charged HSE06 setup]
ISTART=0
ISMEAR=0
SIGMA=0.008
ISPIN=2
LHFCALC = .TRUE.
HFSCREEN = 0.2
ALGO = Damped ; TIME = 0.4
PRECFOCK= F
ICHARG=2
NELM=250
NELECT=107 (109 for no charge, -2 e for +2)
NUPDOWN=5
ENCUT=350
GGA=PE
NEDOS=1500
IDIPOL=4
LORBIT=11
MAGMOM=12*0 18*0 1*5 6*0

The calculation completes without issue. I turn to the PROCAR file to see the orbital character of the highest five occupied bands:
Bands 56/55 - twice degenerate d state - with highest orbital components 80% dxy, 60% dyz respectively.
Bands 54/53/52 - thrice degenerate d state, 50% dz2, 60% dx2, 40% dxz majority characters, respectively.

Already, the expected t2g/eg character is wrong, but I want to visualize.

I run a separate calculation to do the band-decomposed orbitals with the WAVECAR.

INCAR:

LPARD=TRUE
IBAND= 52 53 54 55 56
LSEPB= TRUE

The job finishes, and I am able to visualize these 5 bands. To my unpleasant surprise, they do not match the PROCAR descriptions at all. In fact, they are the "textbook" t2g/eg characters!

Does anyone have some advice as to why, for a specific band, the charge density does not at all match the description given by PROCAR?

I will be happy to furnish additional details if I was not clear enough here.

Thank you very much for your considerations.

Just to clarify:

The visualized orbitals from decomposed charge density job the were very clearly:

band 56 = dx2 (not dxy!)
band 55 = dz2 (not dyz!)

band 54 = dxy (not dz2!)
band 53 = dyz (not dx2!)
band 52 = dxz (consistent with PROCAR)
<span class='smallblacktext'>[ Edited Wed Jul 11 2012, 06:15PM ]</span>

Can anyone comment on this issue? It is creating some significant problems for me, and I have found two other threads asking this question which no one at all answers. Any advice from the admin?

When speaking about percentage of d orbitals in a band:
Bands 56/55 - twice degenerate d state - with highest orbital components 80% dxy, 60% dyz respectively.
Bands 54/53/52 - thrice degenerate d state, 50% dz2, 60% dx2, 40% dxz majority characters, respectively.
are these d orbitals coming from the transition metal?
Are you sure that highest five occupied bands are made of d bands of the TM?

Thank you very much for the reply!

Yes, I can confirm that these bands are from the d-Mn only (the Mn atom in the PROCAR matrix is the only center giving d character, and this character is the dominant percentage of the band contribution). I have also confirmed with non plane-wave methods at HSE06/LANL-2DZ that the highest 5 orbitals are in fact the d-Mn states with the same degeneracy (two and three degenerate).

It is simply that the character of the orbitals from PROCAR does not match at all the visual character from the decomposed density (see my second post).

This might be a decent time to ask: Are the projections in VASP rotationally invariant (I hope they are)? I can rotate the molecule in the cell to see if this has some effect on the PROCAR description, but this is my only guess right now for this discrepancy. Any thoughts on this?




<span class='smallblacktext'>[ Edited Wed Jul 11 2012, 06:42PM ]</span>

The density, band decomposed using IBAND =xyz, is not spin polarized. The final density is a mixture of alpha and beta contribution even in the case when beta levels are not occupied. Therefore the shape of the visualized alpha+beta otbital can differ from the orbital character of the corresponding level in PROCAR (alpha).

Thanks again for the reply. I understand this response in the case where alpha and beta levels are energetically near one another. However, my majority d states are energetically isolated from any beta levels. Should I extend this explanation to mixing beta density where beta levels do not exist (seems strange)? The second issue is again orthogonality: how is it possible that VASP will report a mixing of eg/t2g states in a (single) band? (!!)

I see follow-up questions don't get answered frequently, but it would go a very long way to helping myself and many others understand what is happening in this program.
<span class='smallblacktext'>[ Edited Thu Jul 19 2012, 04:19AM ]</span>

If the output is not spin-polarized, why do I get two blocks of data in my PARCHG files?

I assume this is the alpha+beta and alpha-beta blocks (same is in CHGCAR).

If I plot each of these blocks separately, there is no difference. Is this because, as I posted above, there is no beta density in the vicinity of the d-alpha states?

It appears that the alpha+beta issue does not account for my orbital discrepancy with PROCAR.

Can the admin please comment on this?

<span class='smallblacktext'>[ Edited Fri Jul 20 2012, 06:33AM ]</span>

The calculation is spin-polarized. However, there is no separate alpha and beta output of electron densities, but combined alpha+beta and alpha-beta.

Thank you again.

I appear to not be clear enough in my questions, so I will kindly ask again in a very straightforward way:

1. why does VASP not obey orbital orthogonality (mixing orthogonal orbitals in a single band)?
2. are the projections sensitive to molecular orientation in the cell? (I'm not seeing that they are, but I could be wrong.)
3. If:
a) projections are not sensitive to orientation and
b) alpha+beta and alpha-beta are the same (thus, no contributing beta levels), then:

how is it possible that my orbitals do not correspond to the PROCAR?

If these three points can be addressed, I will be able to continue using VASP with confidence.

Thank you.

No advice on these specifics?

The visualized orbitals from decomposed charge density job the were very clearly:









<span class='smallblacktext'>[ Edited Thu Jan 17 2013, 11:05AM ]</span>

Hello,
in some messages above you have said that you did analyze the character of the bands (the percentage of orbitals type).
In there a way to do it automatically with VASP ? Or in there any script around to do so ?
Thank you in advance !
Marco