EELS calculation, a correction method of peak/energy positions

[tomohisa_ogawa]

I have another technical question on the EELS calculation, especially on correction of peak positions (other word is correction of loss energy)

As pointed out in the XAS tutorial, “Note that the absolute values of the experimental peak positions is not captured due to fundamental limitations…”, we have to correct the loss energy when comparing it with experimental spectrum.

In the case of CASTEP, for instance, the beginning of a spectrum start from 0, where LUMO maybe baseline, and the overall spectrum is given as relative. Its correction is done based on the delta-SCF method, which is described in "https://www.edge.iis.u-tokyo.ac.jp/CAST ... nglish.htm, 5. Calculate Theoretical transition energy".

However, in the case of the VASP-EELS calculation, the beginning of the spectrum has already certain value (certain energy eigenvalue?), it’s not 0, certian correction seems to be applied. What correction methods are applied for the VASP-EELS spectrum? I couldn't find mention of the correction in the XAS tutorial.

I may have missed some descriptions, but I would appreciate your comments on this topic as well.

Best regards,
Tomohisa Ogawa

[martin.schlipf]

The spectrum is set relative to the energy of the core state. The lower end of the spectrum is 90% of the core level energy, the upper end of the spectrum is set so that all unoccupied states fit in the width. You should be able to tweak the range by setting OMEGAMAX and the number of points with CH_NEDOS. In an ideal world the energy of the core state would correspond to the experimental one and you could directly compare the two spectra. However, since DFT does not describe localized states very well, you need to shift the spectrum to match the experiment instead.

[tomohisa_ogawa]

Thank you for the reply.

My goal is actually to obtain an (stacked) averaged spectrum from specific local areas of an amorphous structure (for example, with 100 atoms in the unit cell). Due to the amorphous nature, the peak structure varies from one local area to another, making it impossible to align the spectrum based on peaks or other reference points.

For instance, in CASTEP, as indicated in the link above, the ΔSCF method allows for the calculation of energy for each atom when considering the ground state and the core-hole state. These energies can then be used to apply corrections to the EELS spectra calculated from each local area.

Of course, there is a certain degree of error in each calculation, so there may not be a perfect match with the experimental spectrum. Nevertheless, I have confirmed that there is a reasonable level of agreement using a different software.

Is a correction method like the ΔSCF method also possible in VASP?

Best regards,
Tomohisa

[martin.schlipf]

The implementation in VASP follows this paper. You can also read this paper for some more information about the method in the context of varying atomic environments. Furthermore you can also read this wiki entry for more information about the technique.

If you want to average over different atomic configurations the way to proceed is to compute separate VASP calculation, where you always select a different atom to be the one with the core excitation. Then as a postprocessing you can merge the different spectra.

Regarding to the onset of the peak there is better way than to manually in VASP. The reason for this is that the positions of the core levels are usually not well described by DFT so the peaks would not align with experiment.