[Wannier] example8

[Jonathan Yates]

On Mon, 21 Apr 2008, nafise rezaei wrote:

> I ran example8, but I wanted to consider that generated MLWF's are real or not.

Nafise,

  Sorry for the delay in replying (busy week..)

  It sounds like you are doing exactly the right steps. Indeed when I 
reproduce the calculation for iron I find wannier90 tells me the down MLWF 
are not real.
  Of course they should be real, so this is a coding bug.

I believe the error lies in pw2wannier90: it is writing the up spin u_nk 
in the files UNK00n.2 (when it should write the down spin u_nk). This 
explains why you get good spreads and interpolated band-structure, but the 
MLWF appear complex.

  with default pw2wannier90
      Wannier Function Num:    1       Maximum Im/Re Ratio =   19.686868
      Wannier Function Num:    4       Maximum Im/Re Ratio =   23.219154

line#1519 of pw2wannier90 reads
  call davcio (evc, nwordwfc, iunwfc, ikevc, -1 )
change this to
  call davcio (evc, nwordwfc, iunwfc, ik, -1 )

  and we get

      Wannier Function Num:    1       Maximum Im/Re Ratio =    0.000005
      Wannier Function Num:    4       Maximum Im/Re Ratio =    0.000005

  as expected.
We'll update the version of pw2wannier90 at www.wannier.org as soon as 
possible.

  Thanks for spotting this

  Yours
   Jonathan


btw. I'm going to take this opportunity to reproduce a quote from this 
month's Physics Today: Phillip Anderson has a article on bonding

  "There is a way within band theory of understanding the locality of most 
chemistry using Gregor Wannier's wonderful transformation from band 
wavefunctions to atomic-like ones, but this wasn't really understood until 
the late 1960's, and is still little used by band theorists and quantum 
chemists"
  [of course the last part is not so true these days!]




-- 
Dr Jonathan Yates         |    Theory of Condensed Matter Group
Corpus Christi College    |    Cavendish Laboratory
Cambridge, CB2 1RH, UK    |    Cambridge, CB3 OHE, UK
email jry20 at cam.ac.uk     |    Tel +44 (0)1223 337461