[Wannier] Supercell size, b vectors or something else

[Jonathan Yates]

Jeil,

 The difference between the 'good' and 'bad' results is due to the number of bands from which the optimal subspace is selected (i.e. nbnd in the nscf calculation). By chance I had 20, yours was 10.

There is a 'chemical' reason for this: First, notice that it is the pz WF which start with a large spread in the nbnd=10 case. Then focus on the bands at gamma. We need enough bands to reach the pi* state. For AA-stacked graphite 10 bands is sufficient. However, as we make a larger supercell 'free-electron' like vacuum states fall below the pi* state. The larger the supercell the more of these bands. At some point the pi* state does not lie within the lowest 10 bands. The MLWF routine does the best it can to form localised WF, but the symmetry is broken.

 So if you change nbnd to 20 in your runs you will get symmetric MLWF. For a really big cell it might need to be larger than this - if you do a calculation at gamma and gamma+(a bit) you can see where the pi* state is.

 Thanks for the interesting question!

Jonathan


On 7 Aug 2012, at 16:50, Jeil Jung wrote:

> Hi Nicola and Jonathan,
> Thanks for your suggestions. I tried first to reproduce Jonathan's results look great, unlike those that I am getting. 
> The results I get for c = 20 \AA are
> 
>  ------------------------------------------------------------------------------
>  Initial State
>   WF centre and spread    1  (  0.357549,  1.840705,  4.999300 )     0.60827442
>   WF centre and spread    2  (  0.357547,  0.619293,  4.999290 )     0.60825884
>   WF centre and spread    3  (  1.415324,  1.230001,  4.999302 )     0.60826060
>   WF centre and spread    4  (  0.710151,  1.230129,  4.952752 )     3.27122735
>   WF centre and spread    5  ( -0.000082,  0.000298,  4.905977 )     3.34246618
>   Sum of centres and spreads (  2.840489,  4.920426, 24.856621 )     8.43848738
>  
>       0     0.844E+01     0.0000000000        8.4384873837       1.19  <-- CONV
>         O_D=      2.0391342 O_OD=      2.8651740 O_TOT=      8.4384874 <-- SPRD
>  ------------------------------------------------------------------------------


-- 
Department of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK
tel: +44 (0)1865 612797                http://users.ox.ac.uk/~oums0549/