[Wannier] Wannier functions' spreads increasing

[szaboa]

Dear Wannier90 experts,

I'm facing an issue I don't understand, so I would appreciate any help. 
My projections produce very well localized initial states (<2 A^2 for 
each state), but after one iteration many of the Wannier functions' 
spreads become huge (>1000). Those are states whose centers are also 
translated to a periodic replica of the unit cell after one step. I'm 
using guiding centres, but it does not help.
It is a supercell of a 2D heterostructure (435 atoms), with a hexagonal 
lattice, and a large out-of plane vacuum region. I'm using a 3x3x1 K 
point grid, but I've got the same problem with a 2x2x1 grid too, except 
that it was not the same set of Wannier functions that exhibited that 
weird behavior. In one case I've used 50 disentanglement steps, but it 
did not change anything compared to only 10 steps.
I'm attaching the win and wout files. I would be grateful for any hints 
on what might cause this problem or how to circumvent it.

Best regards,
Aron Szabo
IIS, ETH Zurich
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: wannier90.win
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20151112/eb8d2310/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: wannier90.wout
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20151112/eb8d2310/attachment-0001.ksh>