[Wien] QTL-B error for Zn K-edge ELNES

[Peter Blaha]

About core-hole calculations for XSPEC or TELNES:

1) You need to make a sizable supercell !!! Maybe start with 16 
atoms/cell, but 64 or more are even better. It is WRONG to make a 
core-hole calculation with just 2 Zn atoms/cell. Why: Consider reality: 
You excite only VERY few atoms in such a measurements (actually VERY few 
even means 1 out of 10^10 or less!!!), the majority of atoms remain in 
the ground state and can contribute to the screening. This is, what we 
want to simulate.

2) We need to have charge neutrality, i.e. the sum of nuclear charges 
and electronic charges must cancel. By removing 1 e- from one atom in 
the cell, you need to add this charge somewhere. 2 possibilities:
a) add it as valence electron (increase NE in case.in2; but don't forget 
to remove this e- AFTER the scf cycle and before calculation the spectra)
b) add a background charge in case.inm.
MSR1  -1.0  YES  (BROYD/PRATT, extra charge (+1 for additional e), norm)
As the comment says, you should set this value to +1, when you have 
added an e-. For core holes, we have REMOVED an electron, thus we must 
set this to   -1.0   !!!
You SHOULD ALWAYS CHECK your charge neutrality using
grep :NEC01 case.scf.

Now which method is better, a) or b) ???
In many cases there is NOT much difference, though in some cases there 
could be a difference.
Again, remember we want to simulate experiment as close as possible:
the Zn 1s electron will be excited (dipole rule) into Zn-4p electrons. 
The electronic structure of ZnO has mainly Zn-s and p character at the 
conduction band minimum, thus when we use method b), we actually add 
this e- into Zn-s+p states, which is "approximately" correct. However, 
since the Zn-4sp states are very delocalized states with little weight 
inside the Zn-sphere, I don't think it matters too much when you put the 
charge into the background.
An opposite example would be eg. CeO2, which has very localized Ce-4f 
states just above EF. When simulating O-K edges, it is NOT a good idea 
to add the e- into the valence, because it will go into LOCALIZED Ce-4f 
states and not contribute significantly to screen the core-hole on 
oxygen. The method b) would be better in this case. Of course, for Ce-M 
edges obviously method a) should be much better.

Hope this clarifies the unclear discussion on the mailing list.

Peter Blaha

-------------------------------
Subject: Re: [Wien] QTL-B error for Zn K-edge ELNES
From: Paul Fons <paul-fons at aist.go.jp>
Date: 08/27/2014 06:37 AM
To: A Mailing list for WIEN2k users <wien at zeus.theochem.tuwien.ac.at>
Hi Kevin,
     Are you sure you are correct for using "-1" in case.inm.  I can 
reproduce the example from the Wien2K course (MgO supercell) just fine 
when I use a "+1" in case.inm.  The documentation states that "1" adds 
another electron.  From what I understand this is to compensate for the 
electron that is removed when a core hole is generated by changing the 
occupancy in case.inc.  For the case of MgO.inc

  1 0.00  0  NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT
1,-1,2               ( N,KAPPA,OCCUP)
  1 0.00  0  NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT
1,-1,2               ( N,KAPPA,OCCUP)
  1 0.00  0  NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT
1,-1,2               ( N,KAPPA,OCCUP)
  1 0.00  0  NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT
1,-1,2               ( N,KAPPA,OCCUP)
  1 0.00  0  NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT
1,-1,2               ( N,KAPPA,OCCUP)
  0


  I changed the second line to 1,-1,1 to make a core hole.  I then 
compensated for the missing charge with case.inm to add a +1 background 
charge via

MSR1  1.0   YES  (BROYD/PRATT, extra charge (+1 for additional e), norm)
0.20            mixing FACTOR for BROYD/PRATT scheme
1.00  1.00      PW and CLM-scaling factors
9999  8         idum, HISTORY

After this, I run a SCF calculation.  When it is done, I then remove the 
charge from case.inm and carry out a XSPEC calculation.  This is what I 
learned in the Wien2K course a couple of years ago when the developers 
had a seminar here in Japan.  As I mentioned, the above procedure 
reproduces the MgO core hole present plot in the exercise notes.  I 
would be very much interested in hearing if I am doing something incorrect.

Cheers,
Paul




-- 

                                       P.Blaha
--------------------------------------------------------------------------
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    WWW: 
http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------