[Wannier] large Im/Re

[Jonathan Yates]

Jennifer,

 Your WF are 'nearly' real. I think the problem might be due to the positioning of the windows. You see from the band-structure of copper that if your frozen window is a little too high you force your WF to capture the character of some different bands. So in this case the inner window can only be a few eV above the Fermi level (if you wanted to capture higher bands, you might try 9 MWLF).
 It is helpful to compare the bands from PWSCF to those computed by W90 - that will guide you to a good choice for the windows.

 Copper is a well studied example - it was in the original disentanglement paper of Souza, Marzari and Vanderbilt. You will find input files for copper as example06 of the Wannier90 distribution. This will converge to give real MLWF in small number of steps.

Yours
  Jonathan



On 5 Aug 2013, at 19:31, Jennifer Wohlwend <j.wohlwend at live.com>
 wrote:

> Dear all,
> I'm using WANNIER90 (Release: 1.1.1) as a library with pwscf and epw.x for Copper. Although I'm using the same starting projections as the example (I've also tried starting with 'random') with  15 bands (7 WFs) and an 8x8x8 k-point mesh the Im/Re ratio is still quite high especially for the interstitial WF (6 and 7).  The spread is also somewhat larger than the example found here:http://www.tcm.phy.cam.ac.uk/~jry20/wannier/copper.html.
> The band structure reproduces well and the visualization of the (1-5) WF look like 3dx2-y2 , 3dz2 , 3dxy, 3dxz, and 3dyz orbitals whereas the 6 & 7 WF look like distorted s orbitals. I've included guiding_centres = T as well but I'm still getting imaginary WF. I have included the center/spread data, Im/Re ratio data, the .win file created by epw.x as well as the pwscf.in file for my system. Thank you in advance for any suggestions on how to remedy this situation.
> 



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