XANES calculations are usually more challenging than EXAFS calculations. They usually take more time and require more experience from the user. Fortunately FEFF8 automates many steps in the procedure and includes important full multiple scattering terms and self-consistency. This improves on the the high order path expansion approach to XANES in FEFF7 which only allowed the maximum number of paths of amplitude larger than a plane wave criteria in PCRITERIA card. To account for the poor electron gas estimates of the Fermi level, CORRECTIONS cards were needed. Moreover, to obtain good results for the spectra, one had also to play with AFOLP, EXCHANGE and ION cards, which alter the way the scattering potential is constructed in somewhat uncontrolled ways.

One of the main advantages of FEFF8 (SCF card) is that it can make a more reliable self-consistent estimate of Fermi level (the CORRECTIONS card can still be used, since the error in Fermi level position is only a few eV) FEFF8 also accounts for charge transfer. The ION card should be used only to specify the total charge of a cluster. AFOLP in general leads to better results for XANES and is done by default. With FEFF7 we also had to use EXCHANGE 5 model for Pu hydrates, but with the self-consistent FEFF8 the standard EXCHANGE 0 works well. Thus the use of self-consistency leads to closer results for different exchange correlation models.

The use of the high order MS path expansion and PCRITERIA can lead to unreliable XANES calculations when the MS series converges poorly (for example, near the Fermi level). Thus the inclusion of FMS capabilities in FEFF8 is essential for calculations of LDOS and electronic densities and is often an improvement on path calculations of XANES. Actually we suggest that one use FMS exclusively for LDOS calculations and for testing convergence of the path expansion. This will cost CPU time, but will lead to more reliable results. The FMS calculations for a cluster of 87 atoms typically take more time and memory than the other 5 modules. The results can be somewhat better with larger clusters, but typically one achieves convergence with about 50-200 atoms and calculational time scales as a third power of number of atoms in a cluster and quickly becomes prohibitive. Below we present several sample input files for XANES calculations.

Thu Jul 1 14:31:17 PDT 1999