[Wien] Bulk vs supercell Plasma frequency

[Peter Blaha]

In principle you are doing thinks now correct.
In reality, however, we use the tetrahedra-method in the integration of 
the BZ to calculate a joint-density-of-states. And for a metal !!! here 
there is a difference between k-meshes in the small or large cell. (it 
is related to the "back-folding" problem and in the supercell there are 
less tetrahedra and thus less possibilities to interpolate for the 
tetrahedron method than in the small cell).

In any case, you need to consider two things:

a) You have to decide to which accurcy a certain number should be 
calculated. Typically, a plasma frequency would be quoted with one or 
two digits after the comma (i.e. 5.9 or at most 5.94 eV).

b) Then you have to converge the numbers with respect to k-mesh (but 
also Rkmax !!)
Definitely you should use a ratio of 6:1 for the supercell, but 
eventually you need to increase the mesh until there are no changes 
within the desired accuracy. In your case, 90x15 is almost sufficient, 
but not quite if you want two digits accuracy.

c) in a cubic system, of course it is useless to quote two tensor 
components and one is sufficient.



Am 29.10.2015 um 10:35 schrieb Muhammad Sajjad:
> Dear Prof. Blaha
>
> I did some calculation for Bulk Na (bcc, space gp #229_Im-3m) wiht 
> k-mesh 30*30*30 and found the plasma frequency (in case.outputjoint) as
>  Plasma frequencies:
>
>        w_p_xx      w_p_zz    [eV]
>
>        5.9446      5.9446
>
> Now I constructed a supercell 1*1*6 (definitely symmetry reduced and 
> now sp.gp <http://sp.gp> is 1_P1), used k-mesh 30*30*5 and  found
> Plasma frequencies (much deviating from bulk values):
>
>        w_p_xx      w_p_zz    [eV]
>
>        5.76146      5.3446
>
> Then used k-mesh 90*90*15 and found
> Plasma frequencies (w_p_zz is deviating from bulk w_p_zz ):
>
>        w_p_xx      w_p_zz    [eV]
>
>        5.9485      5.8903
>
> I have read the previous post 
> (https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg09782.html) 
> but it is not speaking about w_p_zz. So when I plot epsilon data for 
> w_p_xx then it is matching with that of bulk but when I use w_p_zz  
> then it is definitely  away from bulk. Should I consider w_p_zz  or 
> use "number of choice = 1 in case.inop" and plot epsilon only with 
> w_p_xx ?
>
>
> Kind Regards
> Muhammad Sajjad
> Post Doctoral Fellow
> KAUST, KSA.
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20151029/e7311bd0/attachment.html>