You don't have to take more than 1 k-point in z direction. I got also
confused by these examples.
A more detailed explanation can be found in
departments.icmab.es/leem/siesta/tlv14/slides13.pdf
page 5 and 17.

You have to use a high number of k-points in transport direction for the
electrodes, 1 k-point in transport direction for the device and also
TBTrans, because of the open boundary condition.
In the transverse direction however, you have to use the converged
k-point from the siesta runs in the electrode and device, and a much
higher number in the TBTrans run.

Best wishes,
Frank

On 11/12/2016 05:03 PM, Zara Nosh wrote:
> Dear Dr. Thomas Archer
> Thank you very much for your reply.
>
> I completely agree with you, but in all transiesta examples
> <http://dipc.ehu.es/frederiksen/tstutorial/index.php/Exercises_TS_School_2009>,
> They have taken high k-point in transport direction for device and
> also tbtran run. 
> As we have open boundary condition in z direction, I can't understand
> why we should take more than 1 k-point in z direction.
>
> Any comments are appreciate.
> Best
> Zara
>
> On Sun, Nov 6, 2016 at 9:32 PM, Thomas Archer <[email protected]
> <mailto:[email protected]>> wrote:
>
>     I have little knowledge of trans siesta but I work on smeagol, the
>     methodology should be the same.
>
>
>
>     K points are a little different than standard DFT, we typically
>     use several steps so avoid a massive self consistent cycle with a
>     dense K-point grid.
>
>
>     electrodes: These are constructed periodic along z so should be
>     constructed with k-points in the z direction. You want to get the
>     electrodes as accurate as possible since this is one of the least
>     expensive parts of the calculation you should use lots of K-points
>     along z. The xy plane should be dense enough to give the correct
>     electronic structure similar to a standard DFT
>     calculation.  (e.g.4x4x100)
>
>
>     SCF of the junction:  Since you break the symmetry along Z when
>     you construct the junction it makes no sense to use more than 1
>     k-point along z.   In the xy plane use a normal number  of
>     k-points (for siesta) and this should be kept consistent for the
>     electrodes and the junction. (e.g.4x4x1) 
>
>
>
>     Transmission: Here the xy plane you should have a very dense grid
>     as the transmission is very k-dependant, but typically does not
>     need to be done self consistently. (e.g. 100x100,1) 
>
>
>     Tom Archer
>
>
>      
>
>     ------------------------------------------------------------------------
>     *From:* [email protected] <mailto:[email protected]>
>     <[email protected] <mailto:[email protected]>> on behalf
>     of Zara Nosh <[email protected] <mailto:[email protected]>>
>     *Sent:* Sunday, November 6, 2016 5:20:35 PM
>     *To:* [email protected] <mailto:[email protected]>
>     *Subject:* [SIESTA-L] k-point sampling in transiesta
>      
>     Dear all,
>     For transiesta calculations, we have to use high k-point in the
>     transport direction (z) to simulate the semi-infinite electrodes.
>
>     Do we need high kz, just for  electrodes' calculations? or for
>     both electrodes and device?
>
>     Can we use normal kz (like usual siesta calculations) for device
>     part? For example if our system is large in z direction, about 60
>     Ang, can we use just 1 k point for kz? 
>
>     I really appreciate your help in advance.
>     With best regards,
>     -Zara
>     Tern uni
>
>

Responder a