[Wien] supercell

[Fecher, Gerhard]

There is no general answer for the Heusler compounds (as is also true for all other classes of compounds)
but I have some hundred calculations on my computer that show that 2000k points in the full BZ are not enough to have any reliable results for Heusler compounds.

Usually you should check for converegence, that is follow what happens with physical
quantities (magnetic moments, band gaps) when you increase the number of k points.
If you have this for the primitive cell, then you may use the suggestion of Laurence for the supercell.

As the supercell calculations may be much more timeconsumptive, you may start again with a somewhat lower amount of k-points for the supercell (just to see a preliminary result)
If you need the supercell for example for phonons than you should check again wether the forces are converged with respect to the number of k-points.
As Laurence mentioned, try to stick with the density of points rather than with a given number, as it may also depend on the shape and symmetry of youre supercell, that is it will differ for 8x1x1 compared to 2x2x2 cells.

By the way, you should test the behavior of the properties also with respect to the number of plane waves (Rmt * kmax) and in some cases this may be more critical than the number of k-points.
 
But finally the things always depend on the particular case that you are calculating (symmetry, metal or semiconductor, magnetic system, etc.)
and what you like to know from the calculation, for a bad defined density of states with arbitrary band gaps already 100 points in the full BZ
may be enough, however, for a good transport calculation you may need in some cases up to 64000 points in the full BZ.



Ciao
Gerhard


====================================
Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
________________________________________
Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at]" im Auftrag von "Laurence Marks [L-marks at northwestern.edu]
Gesendet: Freitag, 17. Februar 2012 20:16
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] supercell

What I always do is try and stick with a similar density of points in
reciprocal space. Hence if you used a grid of LxMxN for a 1x1x1 cell
then I would use (L/P)x(M/Q)x(N/R) for a PxQxR supercell.

Note: at the top of case.kgen you can see what the grid is, and this
is a better measure than "2000 k points" which is not what the code
really produces. You can also specify the grid.

On Fri, Feb 17, 2012 at 1:09 PM, Parker, David S. <parkerds at ornl.gov> wrote:
> Saba, generally if I am running a system that is not big, such as the Heuslers or supercells thereof, 2000 k points in the full BZ is enough. For bigger systems the number of k-points will be limited by your patience and access to parallel computing resources, but for a basic scf calc I very rarely use more than 2000 k points. I have recently run an mBJ calculation on a structure that is a variant of the half-Heusler. without problems.  There should be no problem running LSDA+so calcs as you ask, when I do these I often incorporate so from the start.  Best, David Parker
>
>
> On 2/17/12 1:54 PM, "Saba Sabeti" <raskolnikof6028 at yahoo.com> wrote:
>
> Dear all,
> Thanks to Mr Fecher for his attention to my previous question, which was solved just after posing the question.
> Now, I'm going to ask you all some other questions and I would be so thankful if you help me to solve them:
>
> 1. How many k-points are needed for a supercell calculation like AxB1-xCD when x=1/4-3/4, and while I use 5000 k-points for ACD and BCD?
>  2. Can I do LSDA+so calculations like ACD(a half-heusler) case,I mean:
> initialize+run scf
> save case_nrel
> initso
> run scf+so
> 3. And a calculation similar to which has come in userguide for Becke-Johnson?
> best regards
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien



--
Professor Laurence Marks
Department of Materials Science and Engineering
Northwestern University
www.numis.northwestern.edu 1-847-491-3996
"Research is to see what everybody else has seen, and to think what
nobody else has thought"
Albert Szent-Gyorgi
_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien