[Pw_forum] regarding Wannier transport code

Andrea Ferretti andrea.ferretti at unimore.it
Tue Jun 17 17:22:41 CEST 2008


Dear Stefano and T. Maitra,

>     I think there are WanT users (and developers) among the subscribers
> to PWscf forum and therefore your query may get some reply.
>     I just  wanted to  remind you  that you are asking  information
> about a code we have no control of, so maybe there is some other more
> effective mean to get help.
> stefano
>

WanT is a different project wrt Q-Espresso and at the moment is not 
distributted with it, while it is tightly 
interfaced to, and it is supposed to work with, Q-E.

Said that, I think that while questions about the Q-E/WanT interface may 
be proper to this forum, other questions purely related to 
WanT can be probably better suited for the WanT developers (myself, or 
anybody else from the page http://www.wannier-transport.org/contacts.html )...


>> T.Maitra at tnw.utwente.nl wrote:
>>
>>> 1. How does the programme 'kgrid.f90' which is supposed to calculate the k-points
>>> in the full Brillouin Zone take into account the lattice structure? I mean, for example
>>> the hexagonal lattice and simple cubic should have different set of k-points.

basically kgrid.f90 generates a kpt mesh expressed in terms of the 
reciprocal lattice vectors ("crystal" units), avoiding any symmetrization 
of the kpts...

such procedure is needed because the usage of symmetries is not 
currently implemented in want-2.1 (released version of the code),
while it is going on in the cvs version...

>>>
>>> 2. When I try to do a transport calculation with single k-point along the transport direction
>>> the code doesn't calculate R = (0,0,1) block of the Hamiltonian matrix. But with more than
>>> k-points along the transport direction it does it. Why is that so?
>>>

basically, the inequivalent R vectors (direct 
lattice vect) you can obtain are related to the monkhorst-pack grid you are using..
1kpt -> 1 R vect  =  (0 0 0)

this is because the Hamiltonian matrix elements on the localized basis are 
obtained as a discrete fourier transform:

H(R) = 1/Nk \sum_k H(k) e^-ikR

this is why we need to setup more that one kpt in the transport 
direction...



More information about the users mailing list