E.A. Stern, Physics Department, University of Washington

1. Presently, there is much information in the near edge data that is now not available, because of the present inadequate state of the theory. In most cases, the XAFS cannot be separated out, as the background is changing as rapidly or even more so than the XAFS near the edge. The near edge data also contain information about charge transfer and other modifications of the states near the Fermi level. A calculation of the embedded atom background with a similar accuracy currently available for EXAFS will be necessary to obtain the near edge XAFS signal with enough accuracy to obtain the full information. Also, the background itself has some of this information, such as charge transfer. In fact, it may not be possible or useful to separate out the background from the XANES in the vicinity of the edge, and the theory, in that case, should be capable of calculating both simultaneously with fitting physical parameters that can be varied.

2. Another need in XAFS is a better method to estimate errors in the fitting parameters than is presently available. In general, most XAFS publications presently have an inadequate error analysis. There are some groups that are carrying out an adequate job of error analysis, but even those can be improved. The main element requiring improvement is the determination of the number of effective independent data points (which depends on noise) and their contribution with the noise to the uncertainties in the fitting parameters. Currently, this is being done intuitively by limiting the k-range, by eliminating the data where the noise dominates the signal. It is clear that this is not the most precise way to optimize the information content of the XAFS, although it is roughly correct.