[Wannier] wannierization

Valerio Vitale vv284 at cam.ac.uk
Wed Oct 24 13:49:10 CEST 2018

Dear Ubaid,

The dis_win_max and dis_froz_max specify the energy window and inner (or 
frozer) window respectively, for the disentanglement procedure as 
explained in  Phys. Rev. B. 65 035109 (2001), but also in the 
comprehensive review Rev. Mod. Phys. 84 1419 (2012) . You do not need to 
specify these two parameters if you are only interested in the valence 
bands of an insulator or a semiconductor (like phosphorene).

However, if you are interested also in describing the conduction bands 
of such materials, then you need a disentanglement procedure as these 
bands will be most likely entangled with other bands higher in energy. 
In this cases, you want to specify at least dis_win_max (specifying 
dis_froz_max might be helpful in some cases, if you know what you are 
doing, but it is not mandatory for the disentanglement).

Intuitively, if N is the number of  Wannier functions sought (which for 
spin-unpolarized calculations it is also the number of interpolated 
bands) the disentanglement algorithm tries to find the N states, from a 
larger set in a given *energy window*, whose  "character" changes as 
little as possible from one *k *point to another (lying in a 
neighborhood of *k* in the Brillouin zone). This is achieved by 
minimising selfconsistently the "spillage" between the subspace spanned 
by the N states at *k *and the subspaces at *k*+*b *(i.e. in a 
neighborhood of *k*). Hence, the energy window, whose upper bound is 
given by dis_win_max, must contain at least N states at each *k* point. **

Unfortunately, there is no simple recipe for choosing the value of 
dis_win_max as it depends on how many bands you want to describe, the 
pseudopotential you used in your scf calculation (which defines the zero 
of your energy), and also on the contribution of states higher in 
energy. In this cases, the quality of the interpolated bands strongly 
depends also on the initial projections.

If you want to describe only the valence bands of phosphorene then you 
need 10 WFs (and 10 bands). No disentanglement is needed in this case 
(phosphorene has a finite band-gap). Good initial projections would be  
6 s-like functions centred on each P-P bond and 4 pz-like functions 
centred on each P atom.

If you also want to describe the first few low-lying conduction bands, I 
believe you want 16 WFs (and at least 20 bands). Good initial 
projections would be 4 sp3-like functions centred on each P atom. For 
dis_win_max you want to choose a value which is slightly larger than the 
minimum possible value, i.e. the one for which the window contains at 
least 16 states at each *k *point (it depends on your pseudopotential).

Kind regards,


On 24/10/2018 11:44, Ubaid Mohd wrote:
> Dear experts,
> I am calculating the band structure of phosphorene based materials and 
> there are two parameters that I dont quite understand: dis_win_max and 
> dis_froz_max. I got the band structure of graphene but still there is 
> something not right about it as I do not get a zero gap at the K 
> point. I looked at the graphite example and also read the users guide 
> but did not really get exactly what these parameters do. I know they 
> are very crucial to get a better result. Also I nned to know how to 
> choose right projections for the atoms. For P atom I am taking pz 
> projections, but bandstructure is not quit good. I would be very 
> thankful if someone could make things clearer for me.
> Regards
> /Mohammad Ubaid/
> /PhD Research Scholar/
> /Department of Physics/
> /Jamia Millia Islamia University/
> /New Delhi - 110025/
> _______________________________________________
> Wannier mailing list
> Wannier at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/wannier

Postdoctoral research associate
Imperial College London
Exhibition road
London SW7 2AZ
TCM Group, Cavendish Laboratory
19 JJ Thomson Avenue

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20181024/d2d53527/attachment.html>

More information about the Wannier mailing list