Re: "Scandinavian FEFF question"

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

This is in response to Patric Lundqvist-Reis's msg to Bruce Ravel,
which has been excerpted below for this reply.

1) I reproduced the lack of convergence and the "WARNING:interstitial level
found above Fermi level" message with FEFF8.2 and the 9 atom input file
for EuO given below  (after interchanging the POTENTIALS and XANES cards).

2)  I cut the value of ca in half, i.e., from 0.2 to 0.1 [as suggested
by the second WARNING:], since this can help stabilize the convergence.
>>WARNING: fatal error in subroutine corval. Try
>   to reduce ca1 in SCF card. If does not help,
  SCF 3.1  1  30  0.1

Contrary to Patric's experience, then the code converged after 15 iterations
with # Mu=-1.056E+01 kf=1.529E+00 Vint=-1.190E+01 Rs_int= 2.371,
Thus Ving < Mu and the WARNINGS are gone! 

  I have two suggestions to avoid this kind of problem in the future
since it crops up often with f-electron materials.

1) Use feff8.20 for f-electron materials. Alex implemented by default
a frozen f atomic configuration for f-elements which greatly stabilizes
the SCF loop. [For anyone who wants to allow the f-count to float in the
SCF loop, an UNFREEZEF card can be used, as described in the FEFF8.2 doc].

2) Use a value of ca of 0.1 or 0.05 as needed and or run with a few
iterations before starting the Broyden procedure, i.e.,
  SCF 3.1  1  30  0.1 10
This is also discussed briefly in the doc.

 In future versions we will improve the wording of the original WARNING message
and include the value of Vint:

WARNING:Interstitial level Vint = xxxxx found above Fermi level;
        Results may be inaccurate near E_F.

  Please let us know if difficulties are still encountered for f-electron
materials with feff8.20.

  Cheers,
  John Rehr





>From: Patric Lundqvist-Reis <plr@ine.fzk.de>
>To: ravel@phys.washington.edu
>Subject: "Scandinavian FEFF question"
>Date: Thu, 01 Aug 2002 00:36:54 +0200
>
>I recently started to use feff8 to calculate XANES of simple molecular 
>clusters (such as hydrated metal ions), and I experienced that using the 
>FMS card would reproduce structure above the edge quite well, when 
>comparing to the experimental data (e.g. Eu L3 for Eu3+ aq). To improve the 
>intensity of the white line as well as the Fermi energy, I included the SCF 
>card. This card worked well a few times but most of the times the 
>calculations did not at all converge, and I receive messages like:
>WARNING:interstitial level found above Fermi level
>   Results may be unreliable. See manual for details

>The input file I used here was:
>
>   TITLE   aqua Eu3+ ion, as EuO8 (square antiprimatic geometry)
>   EDGE L3   1.0
>   CONTROL  1     1     1     1     1     1
>  *PRINT    0     0     0     0     0     0
>   SCF 3.1  1  30  0.2
>   FMS 4.0
>  *LDOS  -20  40  0.05
>   EXCHANGE 0 3. 0.1
>   POTENTIALS
>   XANES 6.0
>   *   ipot   z   label  lmax1  lmax2
>        0     63    Eu    -1       3
>        1      8    O     -1     3
>   ATOMS
>     0.0000000     0.0000000    0.0000000     0   Eu     0.000
>     1.3943009     1.3943009   -1.3943009     1   O      2.415
>    -1.3943009     1.3943009   -1.3943009     1   O      2.415
>    -1.3943009    -1.3943009   -1.3943009     1   O      2.415
>     1.3943009    -1.3943009   -1.3943009     1   O      2.415
>     1.9718392     0.0000000    1.3943009     1   O      2.415
>     0.0000000     1.9718392    1.3943009     1   O      2.415
>    -1.9718392     0.0000000    1.3943009     1   O      2.415
>     0.0000000    -1.9718392    1.3943009     1   O      2.415
>
>   END
>
>I also got a message at the end of the log file saying:
>>WARNING: fatal error in subroutine corval. Try
>   to reduce ca1 in SCF card. If does not help,
>SEND bug report to AUTHORS
>
>(_ but reducing the value of ca1 did not help)