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

[Stefaan Cottenier]

> I’m sorry, I forgot to tell that I’m considering beta-Ga2O3 and not
> alpha-Ga2O3.
>
> In the .scf1 file I have a warning:
>
> :WAR   :             1 EIGENVALUES BELOW THE ENERGY  -14.00000
>
> I didn’t change the value of Emin in the .in1 file, so I used the
> default -14 Ry as you did.
> Should I lower it in order to make the warning disappear?

OK, I can confirm your observation for beta-Ga2O3. Starting from the 
cif, and using this sequence of commands:

cif2struct start.cif
cp start.struct beta-Ga2O3.struct
x sgroup
cp beta-Ga2O3.struct_sgroup beta-Ga2O3.struct
setrmt -r 5 beta-Ga2O3
cp beta-Ga2O3.struct_setrmt beta-Ga2O3.struct
init -b -ecut -11 -numk 500 -rkmax 6
run_lapw -cc 0.001
save_lapw -a rkm6

(change RKM from 6 to 9)

run_lapw -cc 0.001

it gives this error in the first iteration:

  LAPW1 END
LAPW2: semicore band-ranges too large

There are 'eigenvalues below...' errors in case.scf1. Emin has to be 
lowered -40 Ry (!) in order to get rid of those:

      K=   0.10000   0.12500   0.00000            1
       MATRIX SIZE  2425  WEIGHT= 2.00  PGR:
      EIGENVALUES ARE:
     -36.7224567  -29.1674472  -17.3058655  -14.2200370  -10.0319827
     -10.0319804  -10.0213123  -10.0213070   -6.4133083   -6.4131634
      -6.4131632   -6.4129289   -6.4087933   -6.4087569   -6.4033183
      -6.4033066   -6.4007536   -6.4007390   -6.3990601   -6.3983440

But even then the band range error is still there:

  Bandranges (emin - emax) and occupancy:

:WARN :BAN00001: 1 -39.990042 -24.394804 2.00000000


The problem persists if you start from a reinitialization (avoiding the 
previous convergence with smaller RKM, but having Emin=-40 right away):

        K=   0.05000   0.12500   0.05000            1
:RKM  : MATRIX SIZE  2425LOs: 102  RKM= 9.00  WEIGHT= 2.00  PGR:
        EIGENVALUES ARE:
:EIG00001:     -24.7321870  -17.8379385   -9.8935838   -9.8935771 
-9.8927672
:EIG00006:      -9.8927595   -7.7404445   -6.2892525   -6.2748230 
-6.2748075

There are still 4 Ga-3s states around -9.9 Ry. The two spurious deep 
states at -24 and -17 Ry are ghost bands. They probably appear because 
for the small spheres you have here (RMT-O = 1.55) a RKM of 9 is *very* 
large.

Quick fixes (no time for a full analysis...):

* either avoid the Ga-3s as valence at all (-ecut -8)

* don't go to that high RKM in your convergence testing. For a RMT=1.55, 
a RKM=9 corresponds to RKM=12.8 (!) for RMT=2.2, or RKM=14.5 for RMT=2.5 
. When you use RKM=6, you have already a very good basis set for these 
small RMT.

Stefaan