Broadness of lDOS plot

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

Dear David,

>I have recently started running lDOS calculations on a simple rock salt 
>system, TiN. When I plot the rhocNN.dat data, I find myself with a plot 
>similar in general shape to experimental data but lacking in any fine 
>structure. I have reduced Eimag (resolution) to zero to avoid any 
>broadening from that card, but it has not made a difference. I've also ran 
>the calculation over a very short energy range to add more points to the 
>plot, and that also has not given me any more detail.

 Thanks for your msg regarding lDOS calculations in FEFF8.

 Unfortunately the use of the LDOS card was not very well documented in
the FEFF8.0 and 8.1 releases. However, we did attempt to fix this in
FEFF8.2 and its documentation which you can download from the FEFF WWW pages.
We also changed the name of the output files from rholnn.dat to ldosnn.dat
in an attempt to avoid confusion.

 In the section on the LDOS card, it now states that the output is written
to the files ldosnn.dat. Here nn refers to the ipot index as in several other
output files (see the section on input and output files).

 What FEFF does is to calculate the local density of states rho_{L,R}(E)
on each distinguishable atom in the system, which is labeled by the potential
index nn. In the Real Space Green's Function formalism, 

    rho_{L,R}(E) = rhoc{L,R}[1 + chi_{l,R}]  = (-1/pi) Im Tr_m G_{LR,LR}

where rhoc is the atomic contribution and chi is the fine structure due
to multiple scattering from the environment. The LDOS output files are

    ldosnn.dat  in FEFF8.2 or rholnn.dat in FEFF8.0 or 8.1.

Thus what is calculated by FEFF8 is the "local density of states" which is
discussed in many places in the literature. We also do this angular momentum
component by angular momentum component. To get the total DOS you have to
add up all the components. This is useful, e.g. for comparison to band structure
calculations. We have checked the LDOS calculated by FEFF8 against band
structure codes, and generally get good results for sufficiently large clusters.
Keep in mind, however, that ldos00.dat for the central atom is the density
of states with a core hole; to compare with band structure, you should compare
the ldos for the other sites.

The atomic contribution rhocNN.dat is also printed out but it shouldn't be
confused with the ldos.  Like the atomic background mu_0 in
the XAS mu=mu_0(1+chi), the quantities rhoc(E) are very smooth functions
of energy since they come from the first term in the Green's function
expansion G = Gc + Gc T Gc and don't include scattering.

We also improved the file formats and column labels in FEFF8.2 so the output is
more clear; this is the header to ldos00.dat, the lDOS of the central atom
for a typical run:

#  Fermi level (eV):  -9.257
#  Charge transfer :   -.177
#    Electron counts for each orbital momentum:
#       0      1.724
#       1      4.428
#       2       .025
#       3       .000
#  Number of atoms in cluster: 195
#  Lorentzian broadening with HWHH      .3000 eV
# -----------------------------------------------------------------------
#      e        sDOS           pDOS          dDOS          fDOS    @#

  I hope that this clarifies the situation.