[Wien] Band gap calculation

[Peter Blaha]

I would think that band gap calculations are the most easy thing to do, 
but I agree, it might be necessary to "think".
The problem could be that the default scf k-mesh, which could be shifted 
in certain symmetries, may NEVER yield the correct band gap. If you are 
interested in the gap, do NOT shift the k-mesh when running    x kgen 
(maybe 90 % of the compounds have their gap at Gamma)

Instructions for band gaps:

Initialization
x kgen (use unshifted mesh, since in most cases the gap is at Gamma)
scf cycle.

Check in scf file if there is a gap (:GAP)

Plot a band structure in various directions. Check the gap and in 
particular if the gap is at Gamma or at some other k-point; if it is 
direct or indirect. (Eigenvalues are listed in case.output1 in Ry, so 
you should be able to calculate the gap exactly once you know which 
k-points to check).

Compare your value with :GAP in the scf file.

If it agrees, ok.

If not: is the gap at Gamma and you have used a "shifted" scf k-mesh ???
Such a mesh does not include Gamma, and all "convergence tests of 
k-meshes" ??? will fail, since even with 100000 k-points, still the gap 
at Gamma is NOT included in the scf.


Am 23.12.2019 um 09:33 schrieb Gavin Abo:
> I don't recall seeing "test_conv-5.0.tar.gz" before.  I'm guessing that 
> is the newer version of the full convergence tests package at [1] 
> (unsupported at least by this group), which seems to not be accessible 
> to the public (and likely only accessible to and perhaps supported by 
> the private group [2]).
> 
> I'm guessing the algorithm that gets the gap value from the valence band 
> maximum (VBM) and conduction band minimum (CBM) is not accurate for your 
> calculation case.  Therefore, I would most likely go with the gap that 
> you get yourself from the raw data (case.spaghetti_ene [3] or [4]) of 
> the band structure plot from spaghetti.
> 
> If the differences are small like in [5], a percent difference [6] of 
> about 2% [= (1.54 eV - 1.51 eV)/{(1.54 eV + 1.51 eV)/2}*100%] seems 
> acceptable.  I believe greater than about 5% in when you should be 
> concerned [7].
> 
> [1] 
> http://wien2k-algerien1970.blogspot.com/2016/09/the-full-version-of-conversgence-tests.html 
> 
> [2] https://www.facebook.com/groups/wien2k.Algeria/
> [3] 
> https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg15255.html
> [4] 
> https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg15258.html
> [5] 
> https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg19405.html
> [6] 
> https://www.webassign.net/question_assets/ncsucalcphysmechl3/percent_error/manual.html 
> 
> [7] https://socratic.org/questions/what-percent-error-is-too-high
> 
> On 12/22/2019 11:36 PM, Peeyush kumar kamlesh wrote:
>> Dear Gavin Sir,
>> Thank you for your response.
>> I did k-point convergence test by a test package 
>> "test_conv-5.0.tar.gz". I dont know whether it is gives completely 
>> satisfactory results or not. If you have any other package for kpt 
>> convergence kindly let me know, so that i can do the right 
>> calculations. Also I used non shifted k-mash in the calculations of 
>> the band gap.
>>
>> Thank you so much!
>> Regards
>> Peeyush Kamlesh
> _______________________________________________
> 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

-- 
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at    WIEN2k: http://www.wien2k.at
WWW: 
http://www.imc.tuwien.ac.at/tc_blaha-------------------------------------------------------------------------