[Wien] Semi core band-ranges too large while finding best Rmt*Kmax

[Stefaan Cottenier]

> While trying to find the best Rmt*Kmax value for Ga2O3 following the
> procedure done by S. Cottenier in this document
> "http://www.wien2k.at/reg_user/textbooks/DFT_and_LAPW_2nd.pdf”, my
> simulation stops with the following error
>
>   LAPW2: semicore band-ranges too large
>
> when Rmt*Kmax is equal to 9.00 and using the command run_lapw -cc 0.000001.
>
> Looking into the .error files, I get
>
>   'LAPW2' - semicore band-ranges too large, ghostbands ?
>
> The .scf1 output gives me the following values for the energies

> :E0_0001: E( 0)=  -10.0138   E(BOTTOM)=  -10.065   E(TOP)=   -9.963  2

>   and the .scf2 output gives me this warning
>
> :WARN :BAN00001:   1  -13.912696  -10.003657  2.00000000


Something must be seriously wrong. I just verified this (assuming you 
are dealing with alpha-Ga2O3), and find

band       1   -9.9374299021   -9.9374142074      0.200000D+01

There are 4 of these bands (Ga-3s). The next set of bands is well 
separated from this (Ga-3p):

band       5   -6.3191843659   -6.3191504344      0.200000D+01

RMTs were 1.80 (Ga) and 1.63 (O).

Did you perhaps have 'eigenvalues below ...' errors? Did you inherit a 
value of Emin from a previous calculation, that was too high for Ga-3s? 
(I used -14 Ry, last line of case.in1, first value).

Mind also that there is probably no reason to include the Ga-3s as 
valence states (that requires -ecut -11). By -ecut -8 you have them as 
core states, and their band range will be zero by construction.

Stefaan