[Wien] How to find the best KPOINTS?

[Jihoon Park]

Dear Xavier,


I appreciate your kind help.
I have learned a lot from you, and been trying things from your suggestions.
Thank you again.


All my best,
Jihoon Park

On Sat, Oct 27, 2012 at 5:25 AM, Rocquefelte <xavier.rocquefelte at cnrs-imn.fr
> wrote:

>  Dear Jihoon Park,
>
> Yes there is a method, which is the one you are using.
> More precisely, it always depends on the property you are interested in.
> For instance, the kmesh for optical properties will be denser than for
> total energy (or a DOS).
> Thus, we should always check the convergency of the property we are
> calculating in terms of k-points, at least.
>
> For magnetic properties, you need a good k-mesh for sure. One solution
> could be to take the smallest unit cell (ferromagnetic order) with the
> full-symmetry of the crystal and to check the k-mesh convergency. Then you
> will know how many k-points you need in the FULL-Brillouin ZONE (FBZ) to
> properly describe the FM state .
>
> Based on this information, you can extrapolate for a larger cell.
> For instance, if you need 800 k-points in the FBZ in the crystallographic
> cell calculation (FM-state).
> For a supercell with a volume, which is 8 times bigger, you will need 100
> k-points.
>
> Such an approach is a pragmatic one and you always need to check the
> coherence of your assumptions.
> In particular, the k-mesh also depends on the magnetic order (to a smaller
> extent than volume of the cell).
>
> Finally, if your interest is to compare different magnetic orders, it
> means that you are interested in the convergency of energy differences.
> In many cases, "energy difference" converge faster than "total energy" ...
> It means that your convergence criteria could be less restrictive. But here
> also you have to do it with care.
>
> Best Regards
>
> Xavier
>
>
>
>
>
>
> Le 25/10/2012 18:03, Jihoon Park a écrit :
>
> Dear Users,
>
>
>  I am trying to calculate magnetic properties for a large unitcell (64
> atoms).
> For this kind of large unitcells, it take very long time to do the KPOINTS
> convergence test.
> Should I anyway "have to" do the KPOINTS convergence test or is there any
> other method to find a good KPOINTS?
>
>
>  All my best,
> Jihoon Park
>
>
> _______________________________________________
> Wien mailing listWien at zeus.theochem.tuwien.ac.athttp://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20121028/d8f24a1d/attachment.htm>