Dear Prof. Blaha,
Thank you very much for your reply!
This is very helpful.
Let me focus on my current approach, without considering PE or BSE, so
I can better explain what I am doing.
1. I create an excited state distribution (single particle picture)
by removing some electrons from valence bands to conduction bands,
just like TDDFT calculation. This is fed into lapw2.F to overwrite weigh.
For instance, around "FERMI level", my weight file looks like this.
Energy K-weight x occupuation
-7.006876394079 0.13707228701714D-02
-7.005370729925 0.13708743753079D-02
-7.005042197147 0.13709258712878D-02
-6.955685793309 0.13706667569952D-02
-6.954117252455 0.13704002422584D-02
-6.954115357204 0.13698089586212D-02
-5.936189328670 0.13710737741162D-02
-5.707575434173 0.13678085576708D-02
-4.187063164686 0.13624145475869D-02
-4.185285376071 0.13627677058236D-02
...
0.561216622892 0.13159990040964D-02
0.571698115465 0.12841250626557D-02
0.572675622534 0.12920402275963D-02
0.581408911740 0.12929368390139D-02
0.587122479196 0.12638067890570D-02 <---Nominal "Fermi level"
0.624650919328 0.11575060746309D-03
0.629635584357 0.58852070187352D-04
0.635765764218 0.76069983586581D-04
0.664550772055 0.79231428253519D-04
0.667904320058 0.41660338621999D-04
0.702333139421 0.95709602531662D-04
...
2.200294635520 0.25778331730391D-04
2.214731483047 0.47148654345984D-04
2.277512135430 0.18778395329644D-03
2.281816619070 0.18308654544186D-03
2. I carry out a self-consistent calculation with the above fixed
weight but with a lower charge convergence criterion -cc, with two choices.
(a) If I directly run with case.inm (first line),
MSR1 0.000 YES
Wien gives FERMI ERROR, since it finds few electrons than RNTOT.
(b) However, if I put the difference between NOS(1) and RNTOT
into case.inm like,
MSR1 0.600 YES
WIEN runs without FERMI ERROR.
So my questions are, (1) How does LAPW2 include the background charge
from case.inm which is supposed to be used in the mixing? (2) is
my approach (b) meaningful?
Thank you so much for your help in advance!
Best wishes,
Guoping
On Thu, Sep 17, 2020 at 6:52 AM Peter Blaha
wrote:
> Hi,
>
> It is not quite clear to me what you acutally want to do.
>
> When you do photoemission, why do you want to occupy conduction band
> states ?
>
> Core hole calculations are for core-photoemission the proper choice.
> For valence photoemission we usually neglect the hole in the valence
> bands - although this can be a crude approximation for more localized
> states.
>
> In optical spectroscopy one would do what you describe: transfer a
> valence electron into the conduction band.
> The best approach would be to use the Bethe-Salpeter approach (the BSE
> code is available on request for WIEN2k - see unsupported software).
> This treats excitonic effects quite accurately, but is also quite
> demanding (needs a large computer cluster).
>
> If I understand your approach correctly, you reduced the number of
> valence electrons and put the missing charge as background ??
> In this way you excite only electrons at the VBM !!!
> How many electrons would you remove ? If you remove 1 electron, you
> remove it in every unit cell of your infinite crystal !! This is not
> what happens in experiment.
> You could create a supercell (eg. 2x2x2) and excite 1 e, but again,
> usually this is "delocalized" and you will remove 1/8 of an electron in
> each cell of the whole crystal.
>
> PS: "Technically" it should still be possible to use a "2-window"
> calculation (previously used for semicore states) in WIEN2k. You can run
> lapw2 two (3) times, once with NE-1 electrons and once with NE+1
> electrons and set an appropriate EMIN in lapw2 (x lapw2 -sc -emin xxx),
> where xxx could be EF of the ground state). Mixer will then add the
> clmval and clmsc files together. But you always have the problem: you do
> this for the whole crystal).
>
> Am 15.09.2020 um 22:26 schrieb Guoping Zhang:
> > Dear Professor Blaha and Wien users,
> >
> > I am interested in states in conduction bands, but I am not interested
> > to get a Fermi energy (and I have no Fermi error in the ground state
> run).
> > I occupy some conduction states (removing some from valence bands) with
> > some electrons by changing the weight files. If I use the Tetrahedral
> > method, this leads to Fermi error in fermi.F at SOS. (TEMPS option in
> > case.in2 is not ideal for my case). So I calculate the electron
> > difference between NOS(1) and RNTOT in the NOS subroutine, and put this
> > difference into case.inm as a background charge. Wien2k runs without
> > an Fermi error message, but I wonder
> > (1) whether this is a sound approach.
> > (2) whether there is any other better method, besides creating a core
> > hole in case.inc file.
> > (3) whether this method can be used to treat photoemission, where
> > electrons are knocked out of the system.
