Re: Position of the Ef in LDOS

["John J. Rehr" <jjr@leonardo.phys.washington.edu>]

Dear Maurizio,

 Maurizio Mattesini asks about the position of the Fermi level
in e.g., CuS materials.
 
> after having plotted the L3-edge (XAS) for a M_{2}S system I've been trying to assign 
> the peaks on the spectra by looking at the local DOS. Everything looks fine except the position of the Fermi energy.
> First I've plotted the partial DOS and then I have scaled the states 
> with the value of the Fermi energy given at the top of the ldosNN.dat file.
> However, despite this the d-states (the states that are contributing most to the L-edge)
> still appear below the Ef, as if they were fully occupied. This is a bit strange since XAS should be very 

  We cannot confirm this behavior with FEFF8.2, so we think that there
is partly a misunderstanding of what the output from FEFF8 is and how
broadening affects the edges of Cu materials.

 With Maurizio's input file, the Fermi level comes out close to the top of
the Cu d band at -5.744 eV (from log1.dat, the d count is 9.59). This
is just what one expects.

  The XAS corresponds to transitions p-d and weakly p-s. It is *not*
the same as the dDOS, since the matrix elements are not constant
and the broadening at the edge includes some of the rapidly rising
tail of the occupied dDOS. This is especially true of Cu where the
dDOS is rapidly changing at the location of the Fermi energy.

  To see this, I suggest plotting xmu.dat col 2 and 4 and ldos00.dat
us col 1 and 4

 I should also mention that the DOS is made of only one 
> single peak (metal d-states) which has almost the same shape of the calculated XAS spectra.

  This is true but it seems to be a misundertstanding of what the
states are. In FEFF8.2 the d-band is mostly occupied and about 5 eV wide.
The XAS has a broad white line, but it comes mostly from the d-holes!
The similarity is accidental.

> To me there is no doubt that the XAS refers to these d-states.  
  On the contrary, in my view there is good reason to doubt this since
it is unphysical.  The plots just look similar, but they have no relation
to each other. A quantitative check shows that they differ in form.

> I have also checked the xmu.dat file and there I saw that the L-edge 
> start effectively from an energy around -10 eV, pointing therefore to the d-states
> located just below the Ef. 

  This comes from broadening. Without broadening, the edge would be a step
function. Broadening always brings in states below E_F and a broadened
exponential tail appears as a shift.
  
> Now, my question is: when plotting the DOS one has to account for the Ef value given 
> at the top of the ldosNN.dat file but it is evident that this is not enough to get the "real DOS".
> Is there any way to place the Ef in the DOS plot correctly?

  Yes - use FEFF's Ef and plot xmu.dat col 2 (E_k) and 4 (XAS)
and ldos00.dat us col 1 (E_k) and 4 (dDOS).

You can look at the background DOS by plotting rhoc00.dat col 1 and 4

  Please let me know if you have further questions/comments on this.

  Sincerely,
  John Rehr