[Wannier] Atomic centres translations and lack of WFs localization

[Krzysztof Kosmider]

Dear all,

I calculate with VASP a nanotube (along z axis) made of 64 atoms
and then calculate 242 WFs by Bloch states projection over atomic
orbitals (p and d)  with radially oriented z-axis (x-azis along a nanotube).
(I did before successfully monolayer wannierization)

If I switch-on guiding centres (guiding_centres=.true.) and avoid
minimization procedure (num_iter=0), I obtain a basis of WFs with
a spread not greater than 2.0 A^2.
However, without guiding centres (guiding_centres=.false.) all
atomic centres are translated significantly (so that a sum of centres lies
on the  x=0.0 y=0.0 axis) and, most importantly, some of the WFs loose
localization (spread about 100 A^2). The same happens when I keep
guiding centres but allow minimization (even for num_iter=1).

With such a not maximally localized basis I obtain meaningful information
about electronics of  the system. I wonder, however, what is a reason of
those translations and why this causes a lack of WF's localization?
How to perform total spread minimization in my case?

Thank you in advance,
Krzysztof Ko?mider

INL - International Iberian Nanotechnology Laboratory
Av. Mestre Jos? Veiga s/n
4715-330 Braga
Portugal

email: kkusiu at gmail.com


Privileged and Confidential

E-mail does not guarantee the security of communications and INL does not accept any liability whatsoever that may arise as a result. This e-mail is intended only for the recipient and contains confidential information, subject to the professional duty of secrecy. Any disclosure, copying or unauthorized use of such information is contrary to law. If you receive this message in error, please notify us immediately and delete the original and any copies and attachments.

International Iberian Nanotechnology Laboratory <http://inl.int>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20131128/355ecfce/attachment.html>