[Pw_forum] about wannier center

Andrea Ferretti ferretti.andrea at unimore.it
Sun Sep 2 16:27:25 CEST 2007


On Sat, 1 Sep 2007, Wei Zhou wrote:

> thank both of you for your helpful  discuss above, I just to learn to
> calcuation *maximally localized wannier function,  and  meet  some new
> question.i have done scf and nscf calculation without problem, but i donnot
> know how to choose the win_max  and froz_max ,is it accrossling to band
> calculation results? 

this is question which goes deep in the definition of the disentanglement 
procedure; first of all, refer to the original paper by
Souza, Marzari and Vanderbilt, PRB 65, 035109 (2001), if you didn't 

as a brief summary:
win_max and win_min are the input flags defining the bloch states 
entering the disentanglement procedure (all the eigenstates  
win_min < eps_nk < win_max  are taken in the starting bloch manyfold)

froz_min and froz_max  define a second (or inner) energy window, which 
should be contained in the previous one: all the eigenstates falling 
inside this frozen window will be directly taken into the wannier 
manyfold (subset of the bloch manyfold which will be used later on to 
localize maxim. loc. WFs)

note that the choice of these parameters may be quite sensible on the 
final results 

the largest the energy window (win_min, win_max) you take, the better 
invariant spread you obtain since you are enlarging the number of dregrees 
of freedom of the problem... of course, it will take longer

the smaller the frozen_window you take, the less constrained your 
minimization is, and the better minimization you get (but take care, 
the constrain is useful to take the lowest states as much similar to 
the origial ones (nscf calculation) as possible)

the basic ideas about how to chose these windows are stated in the 
original PRB paper above... some tricks can also be found in

> when I do disentangle calculation or wannier
> calculation ,it always appear  error result as below  
>  *** from PE    :     0
>  *** in routine : card_wannier_centers
>  *** error msg. : reading line I
>  *** error code :     1
>  *** aborting ***
> it seems that  wannier center is wrong  ,for my system is sodium , I take
> the s p orbits into accout, then I set the center as follow
>           atomic  1     0  1
>           atomic  1    -1  1
>           atomic  1    -1  2

there is a problem with the parameters you provided

the syntax is:

atomic   atomic_index     ang_mom_L_number   ang_mom_M number
the first integer is the index defining the atom you want to consider...
the second integer is the L angular momentum quantum number which should 
be >= 0  (and hence the error given by the code, since you set -1), the
third integer is M ( -L,..., +L)

to get an sp manyfold, use eg

>           atomic  1     0  0
>           atomic  1     1  0
>           atomic  1     1  1

> Is it right ? for wannier calculation ,I am not sure how to control the
> input date for "wannier_Na.in",would you help me .give  some important
> suggestion?
> another question is  when I  calcuate MLWF of  the  structure  with  54
> atoms  calculated from molecular dyanmics , the symmetry is P1 ,then how I
> should choose which atom  as the wannier center  so that the internal
> atomphere is represent

if I understand correctly, 
at the moment there is no way of imposing a certain symmetry of the final
subset of WFs... especially when you perform a disentanglement procedure.
typically the symmetry (or some lower symmetry, if the case) is only 
recovered during the minimization procedure of max.loc. WFs


Andrea Ferretti
National Research Center S3, CNR-INFM  ( http://s3.infm.it )
Dip. Fisica, Univ. di Modena e Reggio E. (Italy)
Tel: +39 059 2055301;  Fax: +39 059 374794;  Skype: andrea_ferretti
URL: http://www.nanoscience.unimo.it

Please, if possible, don't send me MS Word or PowerPoint attachments
Why? See:  http://www.gnu.org/philosophy/no-word-attachments.html

More information about the users mailing list