[Wannier] band structure interpolation ends up with wiggly bands

[Aron Szabo]

Dear Yu,

I was also working on gold recently. As Nicola said, you should include 
the lowest two bands too. The other thing is that you should use the 
disentanglement method instead of just excluding the topmost bands, 
because they are entangled. When you exclude the n-th band it means that 
at every k-point the state with the n-th energy is excluded. It is not 
what you want when bands are crossing each other, because e.g. at the W 
point in your band structure you should keep the topmost blue line 
because that is the one "smoothly connected" to the lower energy states, 
although it has a higher energy at W than the other below it that you 
should exclude. The disentanglement method finds these optimally 
connected bands. The third thing is that you should try to set an sp3 
hybrid as an initial projection instead of the s orbital. I found that 
it converges better. The center of this Wannier function should move 
away from the gold atom, but if you choose an s orbital centered on the 
atom, it will stay there due to it's symmetry. So I would recommend 
these settings:

num_wann  = 12
num_bands = 16

dis_froz_max = 2
#dis_froz_min = -2
dis_win_max = 7

spinors = .true.
Begin Projections
Au: sp3-1;l=2
End Projections


You can also set a value for dis_froz_min, though it's not absolutely 
necessary. The default is just the lowest energy. States between 
dis_froz_min and dis_froz_max will be reproduced exactly by the 
Wannierization, but usually the larger interval you set, the larger the 
spread of the Wannier functions will be, due to the more constraint. 
Therefore you should set it only around those energies that you need to 
reproduce exactly, e.g: close to the Fermi energy if you want to 
calculate transport later.

There is just one strange thing I've found and don't know why it 
happens: if I create a 4-atomic cubic cell instead of the primitive 
1-atomic FCC cell, I get somewhat more compact Wannier functions with 
the exact same settings. Does anyone have a clue why is that? I've 
observed it in many cases that if I create a supercell with a different 
space group than the primitive cell (e.g. a rectangular cell made of two 
hexagonal unit cells of MoS2) the Wannierization might fail. And now 
it's the opposite, I get better results with a supercell.

Best,
Aron Szabo

On 02/12/2015 05:16 AM, Yu Zhang wrote:
> Dear wannier90 users and developers,
>
> I am using wannier90-1.2 and vasp5.3 to obtain the wannier wave 
> function of gold with spin-orbital coupling. I used d5s orbital as 
> target. However, it ends up with some wiggly bands as shown by the 
> attached figure. I have tried different projection (d5), and k-mesh, 
> but I still failed to get the smooth bands. Does anyone know what 
> causes the wiggly bands and how to get rid of them?
>
>
> My wannier90 input file is:
> ===========================================
> num_wann  = 12  ! set to NBANDS by VASP
> num_bands = 12
>
> exclude_bands : 1-2, 15-16
>
> spinors = .true.
> Begin Projections
> Au: l=0;l=2
> End Projections
>
> begin unit_cell_cart
>      0.0000000     2.0340000     2.0340000
>      2.0340000     0.0000000     2.0340000
>      2.0340000     2.0340000     0.0000000
> end unit_cell_cart
>
> begin atoms_cart
> Au       0.0000000     0.0000000     0.0000000
> end atoms_cart
>
> mp_grid =    21    21    21
>
> begin kpoints
>       0.000000000000      0.000000000000  0.000000000000
>       0.047619047619      0.000000000000  0.000000000000
>       0.095238095238      0.000000000000  0.000000000000
>       0.142857142857      0.000000000000  0.000000000000
>       0.190476190476      0.000000000000  0.000000000000
> .......
> ==================================================
>
> Thank you very much in advance!
>
> Best regards,
>
> Yu Zhang
>
>
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20150212/61dc844c/attachment.html>