[Wien] Gauss-broadening"(G-b) in the DOS

[Tiem Leong Yoon]

We simulate a tantalate system using LDA+U. We use small k-points, 18,
k-points, and SCF is done in less than 20 cycles. Then it is followed by
LDA+U. After LDA+U, we do DOS  and bandstrtucture calculation separately.

DOS calculation is straight forward in Wien2k. For our system, the space
group is 13, and thus we refer to
http://www.cryst.ehu.es/cryst/get_kvec.html<https://mail2.usm.my/owa/redir.aspx?C=6b51c784aa744c6aa0f9b75ad7f9c13e&URL=http%3a%2f%2fwww.cryst.ehu.es%2fcryst%2fget_kvec.html>
to generate k-mesh, since we need to plot (Gamma, Z, Y, A, B, D, E) points
and we believe that the default and available simple_cubic, bcc, hcp and
fcc cannot generate the case.klist_band file we want.

The problem now is the band-gap value from DOS and bandstrcuture
calculation are the same if we do not do any "Gauss-broadening"(G-b) in the
DOS. It means, if we stick to the default G-b of 0.003, band-gaps from DOS
and Bandstrcuture calculation are basically the same. But if we change the
G-b from 0.003 to, say 0.008, the band gap is observed to change to smaller
value, >10% less than the value before.

Can someone give the reasoning why there is such a difference occured? By
the way, the reason why we adjust the value for G-b is because the graph
will look more resemblance to experimental data plot rather than "spiky"
features for smaller value of G-b

and how to read the values for band-gap? do we read from DOS plot or
bandstructure plot?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20120906/bbb99d6b/attachment.htm>