[Wien] Reg. Rkmax and K point

[Gerhard Fecher]

This means that the number you give for the k-points is not the one the program uses for the integration:
Assume a cubic system then 1000 k-points result in a 10x10x 10 mesh,
due to the symmetry operations, there are many k-pints in the cube that are equivalent,
and if one knows their multiplicity its much faster to calculate the sum of the equivalent densities than to calculate the density for each point again.

For example, for a fcc lattice only
47 of the 1000 in 10x10x10 will be used
or
72 if you give 2000 (12x12x12 mesh), or
104 if you use 3000 (14x14x14 mesh), and so on.
this is indeed particular for fcc, and in the k list you see how the number of those irreducible k-points depends on the symmetry of the lattice.

If you have non-cubic systems you may like to use different divisions like the one from (number)^1/3 x  (number)^1/3 x (number)^1/3.

Ciao
Gerhard




________________________________________
Von: wien-bounces at zeus.theochem.tuwien.ac.at [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Santhy Jaiker [k_santhy79 at yahoo.co.in]
Gesendet: Montag, 25. Februar 2008 14:03
An: A Mailing list for WIEN2k users
Betreff: Re: [Wien] Reg. Rkmax and K point

Dear Stefaan,
Thanks for your immediate reply.  But i am not able to understand the statement, "Roughly double the number of k-points at each step (doubled in the IBZ [=nr. of lines in case.klist], not in FBZ [=input for kgen])."  kindly elaborate the above sentence.

Regards,
Santhy Jaiker


Stefaan Cottenier <stefaan.cottenier at fys.kuleuven.be> wrote:

> For finding the suitable RKmax and K point for a transition metal or
> compound, i am running the SCF calculation with RKmax range 7 to 10.5
> with step of 0.5.
Your lower and higher ends are a bit high. Start at 5.5, and stop when
you see no further evolution (rarely beyond 8.5).
> Similarly, For k point with increment of 1000 or 2000
Also here you're taking too many steps. Roughly double the number of
k-points at each step (doubled in the IBZ [=nr. of lines in case.klist],
not in FBZ [=input for kgen]).
> upto 0.00001Ry of energy convergence. Simultaneously i am looking the
> other parameters like CTO, Fermi energy, DIS... that the energy is
> properly converges or not.
This criterium might be a bit strong for the sake of testing. Does
anything meaningful changes in observable quantities if you converge
only up to 0.000x?)

You are doing it the right way: first doing these tests before you go on
with the more interesting work. Congratulations! Many others carelessly
skip this part. But don't overshoot in the other direction...

Stefaan
> But it takes very long time, before going to volume optimization or
> mini. If there is any other shorter method is available, to find the
> suitable RKmax and K point.
>
> Thanks in advance.
>
> Regards,
> Santhy Jaiker
>
>
> ------------------------------------------------------------------------
> Why delete messages? Unlimited storage is just a click away.
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien


________________________________
Get the freedom to save as many mails as you wish. Click here to know how.<http://in.rd.yahoo.com/tagline_mail_5/*http://help.yahoo.com/l/in/yahoo/mail/yahoomail/tools/tools-08.html/>