[Wannier] graphene +organometallic complex

[Stefan Badescu]

Dear colleagues,

I have a question about graphene. I am trying to disentangle and to 
wannierise the bands of a graphene 8x8 supercell (128 C atoms), with a 
complex of Cr+benzene adsorbed to it (Cr in the middle). Apparently the 
wannierisation has problems with the 3d and 4s orbitals of Cr that 
hybridize with the pi orbitals of graphene and benzene.

This is a gamma-point calculation with Q.Espresso v.4.2. I choose enough 
bands in the nscf step, in agreement with the works of YS  Lee et al. on 
NTs in PRL 97, 116801 (2006) and PRL 95, 076804 (2005). I can treat NTs 
and graphene with other kinds of defects following those papers.

I am choosing the manifold to include pz orbitals on the carbon atoms 
(to account for the pi manifolds), s orbitals midway between C atoms 
(for sp2 bonds) and between C-H of benzene, and 3s, 3p, 3d, 4s trial 
orbitals for Cr. I use the pbe-van potentials for all atoms. The lowest 
state is spin-unpolarized.

I cannot obtain convergence in wannierisation. I am using a frozen 
window that includes all the bands up to the Fermi level, or 
alternatively a bit larger (to include states above the Fermi level).  
The disentanglement seems to proceed well, with an average spread of ~1 
Angstrom/atom . The wannierisation does not work, most of the spreads 
are larger than 16, and some MLWFs move to the vacuum. I am aware about 
the latter problem from previous posts, but I am not sure whether a 
definitive solution was included in the latest version of wannier90.

I was wondering whether there are some special considerations for d 
orbitals. It seems like the hybrid d-pi orbitals are very close to the 
Fermi surface in my graphene+Cr+benzene system. I wonder whether this is 
the problem, since they are so localized. My understanding is that I 
cannot give up any of the C trial functions since the pi and sp2 bands 
depend on all of them, therefore I have to work with the large frozen 
window.

Thank you for any hint! Would it make sense to send my input files by 
email?

Sincerely,
--Stefan Badescu

PS. I came across the work of Weng HM at al. on (V+benzene)_n infinite 
chains (PRB 79, 235118, 2009). That problem is different because it has 
a periodicity in the z direction and benzene has a large energy gap. 
Nevertheless using for projections the atomic orbitals of V  seemed to 
work in that problem.