[Wien] [SPAM?] How to obtain ELNES simulation results consistent with experiments ?

[yfding]

This problem haunts me for nearly a year, and my questions are still unsolved after consulting a lot of literature and mailing list.


I tried to get Oxygen K edge of the 3*3*3 SrCrO3 supercell.



In my calculations, I first converted the SrCrO3 (space group P4/mmm) into a SrCrO3 3*3*3 supercell (space group P1) in CrystalMaker, then transformed the SrCrO3supercell.cif file into SrCrO3supercell.struct file under the Wien2k case directory, labelled one oxygen atom to make this atom unique, did initialization, and accepted all the queries from nn complaints. The initialization parameters were as follows:


Rmt: Sr:2.40 Cr:1.78 O:1.61
potential: GGA of PBE-96
separation energy: -6.0 Ry
RKmax: 7.0
Max L in WF: 10
number of k points in the whole BZ: 5*5*5
number of k points in the irreducible wedge of the BZ: 18


Then I edited SrCrO3.inc and removed one core electron from the 7th atom, i.e. 1st Oxygen atom 1s state. Next I changed 0.0 to -1.0 in SrCrO3.inm. Then run SCF in parallel (energy convergence 0.0001 Ry).


Then,
-edited case.innes :


SrCrO3 O K edge
7 (atom)
1, 0 (n, l core)
532 (E-Loss of 1st edge in eV)
200 (energy of the incident electrons in keV)
-5.0 20.0 0.1 (minimum energy, energy step, maximum energy)
0.37 0.75 (collection semiangle, convergence semiangle, both in mrad)
50 1 (NR, NT, defining the integration mesh in the detector plane)
0.7 (spectrometer broadening in eV)
OUTPUT
0 (full output)
DETECTOR POSITION
0.0 0.0 (thetax, thetay)
MODUS
energy (dscs wrt what?)
ORIENTATION SENSITIVE
0.0 0.0 0.0
SELECTION RULE
n (selection rule)
LSELECTION RULE
d
INITIALIZATION
y y (dos and xdos handling)
y y (handling of rotation matrices)
QGRID
U
END



- x lapw1
- x lapw2 -qtl
- x qtl -telnes
- x telnes3
- x broadening


As it is supercell, I calculated every peak of unequivalent O K edge and summed them. I'm not sure all the steps I have done are right. The result is different from the TEM EELS experiment (The O K edge EELS spectra were studied on an JEOL ARM200F Cs-corrected TEM with a FEG and a Gatan parallel electron-energy-loss spectroscope). The relative strength and position of the peaks are completely inconsistent.


It's not just this example, but I've had the same problem with all the other materials. I thought about all the possible reasons, and looked up lots of literature and mailing lists, but I couldn't get the calculation results closer to the experiment. I want to know how to obtain ELNES simulation results consistent with experiments ?


Any comment(s) would be highly appreciated. Thanks in advance!

[Laurence Marks]

There are two fundamental issues:
a) How to handle the dynamical diffraction of the swift electrons.
b) How to handle the core hole.

To my knowledge no DFT code handles a) correctly. (Some details are in Chpt
5 of "High-Resolution Transmission Electron Microscopy: and Associated
Techniques" Eds Buseck, Cowley and Eyring.) Particularly if you are on a
zone axis dynamical diffraction effects (often called channeling, but this
is not fully rigorous in my opinion) can be very strong. One way to
minimize the effect is to tilt off the zone axis to a pseudo-kinematical
condition. Of course, by doing this you lose imaging. You might (very
might) be able to reduce the issue by using a larger collection angle
making it similar to Precession (e.g. DOI 10.1016/j.ultramic.2012.11.007).
I am not sure if people have really explored this enough.

For the core hole, from what I have seen (I am not a full EELS expert)
often how much of a core hole one wants is an "adjustable parameter". At
least with Slater Transition Theory it should only be 1/2 a hole.

Not an easy problem, and I am not convinced that it has been fully solved
although I would be willing to be proved wrong.

Post by yfding
This problem haunts me for nearly a year, and my questions are still
unsolved after consulting a lot of literature and mailing list.
I tried to get Oxygen K edge of the 3*3*3 SrCrO3 supercell.
In my calculations, I first converted the SrCrO3 (space group P4/mmm) into
a SrCrO3 3*3*3 supercell (space group P1) in CrystalMaker, then transformed
the SrCrO3supercell.cif file into SrCrO3supercell.struct file under the
Wien2k case directory, labelled one oxygen atom to make this atom unique,
did initialization, and accepted all the queries from nn complaints. The
Rmt: Sr:2.40 Cr:1.78 O:1.61
potential: GGA of PBE-96
separation energy: -6.0 Ry
RKmax: 7.0
Max L in WF: 10
number of k points in the whole BZ: 5*5*5
number of k points in the irreducible wedge of the BZ: 18
Then I edited SrCrO3.inc and removed one core electron from the 7th atom,
i.e. 1st Oxygen atom 1s state. Next I changed 0.0 to -1.0 in SrCrO3.inm.
Then run SCF in parallel (energy convergence 0.0001 Ry).
Then,
SrCrO3 O K edge
7 (atom)
1, 0 (n, l core)
532 (E-Loss of 1st edge in eV)
200 (energy of the incident electrons in keV)
-5.0 20.0 0.1 (minimum energy, energy step, maximum energy)
0.37 0.75 (collection semiangle, convergence semiangle, both in mrad)
50 1 (NR, NT, defining the integration mesh in the detector plane)
0.7 (spectrometer broadening in eV)
OUTPUT
0 (full output)
DETECTOR POSITION
0.0 0.0 (thetax, thetay)
MODUS
energy (dscs wrt what?)
ORIENTATION SENSITIVE
0.0 0.0 0.0
SELECTION RULE
n (selection rule)
LSELECTION RULE
d
INITIALIZATION
y y (dos and xdos handling)
y y (handling of rotation matrices)
QGRID
U
END
- x lapw1
- x lapw2 -qtl
- x qtl -telnes
- x telnes3
- x broadening
As it is supercell, I calculated every peak of unequivalent O K edge and
summed them. I'm not sure all the steps I have done are right. The result
is different from the TEM EELS experiment (The O K edge EELS spectra were
studied on an JEOL ARM200F Cs-corrected TEM with a FEG and a Gatan parallel
electron-energy-loss spectroscope). The relative strength and position of
the peaks are completely inconsistent.
It's not just this example, but I've had the same problem with all the
other materials. I thought about all the possible reasons, and looked up
lots of literature and mailing lists, but I couldn't get the calculation
results closer to the experiment. I want to know how to obtain ELNES
simulation results consistent with experiments ?
Any comment(s) would be highly appreciated. Thanks in advance!

[丁一凡]

Respected Prof. Marks ,


Thank you so much for your reply.


'One way to minimize the effect is to tilt off the zone axis to a pseudo-kinematical condition', does this mean I need to change the angle a little bit? For example, change from

ORIENTATION SENSITIVE
0.0 0.0 0.0
to
ORIENTATION SENSITIVE
0.0 0.0 0.1
for minimizing dynamical diffraction effects. Am I correct?


Another question, when I set 1/2 core hole in the labelled oxygen atom 1s state in case.inc file,
i.e.
1 0.00 0 NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT 1,-1,1.5 ( N,KAPPA,OCCUP)


Then, do I need to set -0.5 in the case.inm file ?


With kind regards.


yfding

[丁一凡]

Respected Prof. Blaha ,


Thank you so much for your reply.


I did it because I read Kevin Jorissen's mail:
http://zeus.theochem.tuwien.ac.at/pipermail/wien/2006-August/007602.html



I probably never did the right thing. I'll do what you suggest.


With kind regards.


yfding

[Laurence Marks]

Post by 丁一凡
Respected Prof. Marks ,
Thank you so much for your reply.
'One way to minimize the effect is to tilt off the zone axis to a
pseudo-kinematical condition', does this mean I need to change the angle a
little bit? For example, change from
ORIENTATION SENSITIVE
0.0 0.0 0.0
to
ORIENTATION SENSITIVE
0.0 0.0 0.1
for minimizing dynamical diffraction effects. Am I correct?
No, you are not correct. You have to tilt off the zone axis in the

microscope. There is nothing you can do in the DFT.

Post by 丁一凡
Another question, when I set 1/2 core hole in the labelled oxygen atom 1s
state in case.inc file,
i.e.
1 0.00 0 NUMBER OF ORBITALS (EXCLUDING SPIN), SHIFT, IPRINT 1,-1,1.5 (
N,KAPPA,OCCUP)
Then, do I need to set -0.5 in the case.inm file ?
You would set 1,-1,1 ( N,KAPPA,OCCUP) in case.incup, and

copy the normal case.inc to case.incdn for a spin-polarized calculation.
For non-spin polarized you would put 1,-1,1.5 (
N,KAPPA,OCCUP)

Post by 丁一凡
With kind regards.
yfding

[丁一凡]

Respected Prof. Marks ,


Your help was greatly appreciated. I¡¯d like to express my gratitude.


yfding