[Wannier] k-mesh even/odd effect

[Elias Assmann]

Hi List,

I have noticed a funny even/odd effect from the k-mesh I use in a 
Wannier90 run.  (All meshes were N×N×N, and even/odd refers to the 
parity of N.)

The material was SrVO₃, where the bands around the Fermi level are 
(ascending in energy) O-p, V-t2g, V-eg.  I projected onto all these 14 
bands.

With an odd mesh, Wannier90 converges quickly and the results are 
unobjectionable.  But with an even mesh, the spread does not converge, 
and the interpolated band structure shows problems near R for the eg and 
some of the p bands (see plot 
<https://www.box.com/shared/yfvfo4vrmngda326j0ua>).

The problem happens if the eg bands are included, for t2g-only and p+t2g 
everything works fine.

Is this a known behavior?  I am using wien2wannier, so the problem could 
also originate there rather than in Wannier90.


	Elias


-- 
Elias Assmann (TU Wien)

Wien2Wannier: maximally localized Wannier functions
               from linearized augmented plane waves

<http://www.ifp.tuwien.ac.at/forschung/arbeitsgruppen/cms/software-download/wien2wannier/>