[Wien] txspec

[Peter Blaha]

> I still do not understand.
>
> L2-edge spectrum is SUM of 2p_1/2 -> d (l+1) transition and 2p_1/2 -> s (l-1) transition. The same is also true for the L3-edge, SUM of 2p_3/2 -> d (l+1) transition and 2p_3/2 -> s (l-1) transition.

Correct.

> In 'txsepc.f' the l -> l+1 transition is calculated first and stored into 'X(I,1)'. Then the l -> l-1 transition is calculated and stored into 'X(I,2)'. These two transitions have nothing to do with the L2- or L3-edge. As I told you, both edges contain these two transitions. The sum of X(I,1) and X(I,2) is stored into new variable X(I) and another new variable is introduced, XINTER(I). XINTER(I)=0 untill I=JEMIN+ISPLIT-1 and XINTER(I)=X(I) for I=JEMIN+ISPLIT to I=JEMAX+ISPLIT. Later XOUT(I)=XINT1*X(I)+XINT2*XINTER(I) is performed for I=JEMIN,JEMAX+ISPLIT. FInally, energy, XOUT(I), X(I,1) and X(I,2) are printed out. So, I think the SPLIT can 'split' the 2p_1/2 -> d (l+1) transition and 2p_1/2 -> s (l-1) transition, not the L2- and L3-edges.
>
As you said:
The sum of X(I,1) and X(I,2) is stored into new variable X(I)
XINTER(I)=X(I)
XOUT(I)=XINT1*X(I)+XINT2*XINTER(I)

Thus in XOUT the sum of l+1 and l-1 transitions + the shifted spectra (stored
in XINTER is added (with weights XINT1/2 given in the input).

> Another question. How come the L2- and L3-edge spectra are identical to each other? Experimentally, they ARE different from each other. Theoretically, they should be different. Maybe the radial wave functions are not much different from each other, however, there are quite large difference in the occupation of states with different angular momentum (l) and its z-comp (m) quantum number. Therefore, it should make, at least, some difference in the dipole transition matrix elements.
>
In the (non-relativistic) dipole approximation m does not enter, but only l
determines the intensities.

> One more question. Is the program 'txspec' treated the m-selection rule properly? I mean, since the incident light is usually not polarized, we have to consider all three possibilitites, delta_m=+1,-1,0. The first two transitions correspond to the right and left cirularly polarized lights and they are very important to MCD and to calculate the magneto-optic effect in ordinary optical conductivity calculation.

No, txspec treat only l (and not m).
This was what I meant with my sloppy "relativistic selection rules".
With polarized light m does matter and you get different L2 and L3 spectra.

You can work out the proper selection rules yourself and use e.g. instead
of a p-DOS a px,y and a pz-DOS (depending on symmetry).....

You can also use telnes, which (I think) does have the proper m-selection rule.

Regards


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------