[Wannier] benzene example

[Jonathan Yates]

On Fri, 24 Oct 2008, ans341 at mail.usask.ca wrote:

>        xxx
>       xxxxx                 xxxxx
>   oo  xxxxx oo             xxxxxxx
>   ooo_xxxxx_ooo           xxxxxxxxx              _______
>   oo/ xxxxx \oo        xxo/oxxxxxo\oxx          /      o\    xxxx
>   oo  xxxxx  oo      xxxo/oxxxxxxxo\oxxx       /      oo \  xxxxxx
> 3x /    xxx    \  6x  xxo/oxxxxxxxxxo\oxx   6x /      ooo  \xxxxxxxx
>   \           /         \           /         \      ooo  /xxxxxxxx
>    \         /           \         /           \      oo /  xxxxxx
>     \_______/             \_______/             \______o/    xxxx
>
> Could someone provide me with the correct results of the example 12 so that I
> could try to reproduce them?

Anna,

  I like the ascii-art wannier functions! As far as I can tell you do have 
the 'correct' answer.
   See
Partly Occupied Wannier Functions
K. S. Thygesen, L. B. Hansen, and K. W. Jacobsen
http://link.aps.org/abstract/PRL/v94/e026405

  They are using a slightly difference formalism to construct their MLWF - 
but actually they turn out pretty much the same as those from w90.

  There is some discussion in the paper. Possibly you expected to see 
sigma and pi like orbitals. The fact that you don't is known in the 
chemistry literature, where Wannier functions in the limit of a finite 
system are called Foster-Boys orbitals. Chemists have other types of 
localised orbitals such as Edmiston-Ruedenberg which do preserve the 
pi-sigma distinction. I'm not an expert in this - but I though I could 
give you some terms to start a literature search with. I suspect others 
here know more about this.

> In addition, I would like to know if there is a way to use the PWscf wave
> functions obtained from a parallel run (say, with 16 processors) with less
> (say, 8) prosessors for pw2wannier.x? According to my short experience, it does
> not work without something extra.

See
  http://www.democritos.it/pipermail/wannier/2008-June/000151.html
and the response.
  If you change the number of processors you need to set wf_collect=.true.
for the nscf calculation.

  Yours
   Jonathan

-- 
Department of Materials
University of Oxford
Parks Road, Oxford, OX1 3PH, UK
http://www.tcm.phy.cam.ac.uk/~jry20/