Inconsistencies in XANES wiki documentation

[aohara]

Thank you for your help! Doing these things solved the problem and now I am getting similar results as the published and can move on to working with some other systems.

I did want to point out a minor discrepancy between the code default and what is listed on two different pages in the VASP wiki page for the CH_SIGMA value for consistency sake. I usually set this value myself to be safe (and set it quite small so I can do broadening in postprocessing as recommended). However, on the wiki page for CH_SIGMA it states the default is 1.0 whereas on the "SCH calculations" page it states that "The parameter CH_SIGMA sets the value of broadening in eV. By default this parameter is set to 0.0001." Lastly, in the actual source code of core_con_mat.F, the value defaults to 0.3 on line 44.

Thanks again,
Andrew

[ferenc_karsai]

Thank you very much for reporting the discrepancies in the wiki pages. I changed the default on both pages to 0.3.

[zhaoru_sun1]

Thank you very much for reporting the discrepancies in the wiki pages. I changed the default on both pages to 0.3.

Following the broadening problems metioned above, I have been confusing the broadening method used in XAS of VASP.
The tutorial said that

Lorentzian broadening (this is the only type of broadening built in for the calculation of XAS spectra in VASP).

When I read the code core_con_mat.F:

Code: Select all

! calculate delta function
                        CALL DELSTP(KPOINTS%ISMEAR,ENERGY_DIFF/CH_SIGMA,DELTAFUN,STEPFUN)
                        DELTAFUN = DELTAFUN/CH_SIGMA
! alternatively use Lorentzian
!                       DELTAFUN =AIMAG(1/(ENERGY_DIFF-(0.0_q,1.0_q)*CH_SIGMA))/PI

It seems that the parameter ISMEAR determines the broadening method, and the Lorientzian broadening is commented out.
According to the subroutine DELSTP in dos.F, if KPOINTS%ISMEAR = 0, it will return a Gaussian broadening:

Code: Select all

X = ENERGY_DIFF/CH_SIGMA
DELTAFUN = EXP(-(X*X))/SQRT(PI)
DELTAFUN = DELTAFUN/CH_SIGMA

So FWHM=2*SQRT( log(2) )*CH_SIGMA?
If using the Lorientzian which is commented out, the Lorientzian FWHM=2*CH_SIGMA.
Because there are few descriptions in VASPwiki, I think I have misunderstood the details.

Another question is about the pseudopotential and electronic Fermi occupancy you mentioned above. I am using VASP to simulate O K-edge XANES on Pt surface where oxygen is adsorbed. So will GW pseudopotencials and including Fermi occupancy improve simulation accuracy?

[ferenc_karsai]

Thank you very much for noticing the discrepancies in the wiki description.
We usually use Gaussian broadening and as you pointed out it is controlled by the ISMEAR=0 tag.
I've corrected the wiki description now.

I think you should always use the GW potentials if available. They simply have a better description of the excited states with a slightly higher computational cost.