Hello Ifeffit friends, It is my understanding that FEFF6 or equivalently FEFF7 are considered the standard for Extended X-ray Absorption Fine Structure Analysis. In my experience, FEFF8 is a dangerous tool for EXAFS analysis as it has the ability to produce results that differ significantly from FEFF6 or FEFF7. In the review of EXAFS manuscripts where FEFF8 is the tool of choice, what parameters should be reported to give the reader some confidence that the authors have not tweaked the code to cover-up some errors in the model(s). A Big Thanks Shelly Shelly Kelly Bldg 203 RM E113 Skelly@anl.gov Argonne National Laboratory 630-252-7376 9700 S Cass Ave www.mesg.anl.gov Argonne, IL 60439
HI Shelly, I wouldn't say it's wrong to use Feff8 for EXAFS analysis, just rarely necessary. Feff8 was definitely designed for XANES, and improvements for EXAFS were not the main goal. But there are a few cases where Feff8 _is_ known to be better for EXAFS than Feff6. (I'll leave out Feff7 as it seems no one is using this much). Clusters with H is one, and transuranic elements is another. I've heard a few people claim that Feff8 gives "better E0's", which I think means that E0 refines to a value closer to 0 (that is, to where the it's arbitrarily assigned in the data processing). Since E0 has to be refined and has little physical significance, I don't find this to be a compelling advantage for Feff8. Early versions of Feff8 (pre-release? Including 8.00?? I'm not sure) were much worse for EXAFS.This did get fixed, but I'm not sure when (I have a hard time understanding Feff8 version numbers and can never really tell which get released and widely used: I've given up complaining and live wih the fact that "Feff8" means one snapshot of 8 years worth of work.). By Feff8.1 for sure, and maybe eralier, there were essentially no difference between Feff6, 7, and 8 for EXAFS. That was based on analysis of data on metal foils (Cu, Ag, Au), and could hardly be called thorough and possibly not even 'highly relevent to real-world problems', as it doesn't even test changes in exchange or overlap of metal-oxygen potentials. Better testing and comparisons would be very good. Anyway, I think it's fine to use Feff6, 7, or 8 for EXAFS. I don't think anything should change for these versions in the reporting of how a Feff calculation was done for EXAFS. Most EXAFS work using Feff give no details about the calculation, and I think that's mostly OK, as the Feff/Atoms/Artemis default parameters are generally quite reasonable. If the work discusses comparing different Feff calculations, then obviously more detail on what was changing would be needed. That's my opinion, anyway. --Matt
Hi Everyone, The newest release of FEFF8 is version 8.20 (according to the FEFF web site http://leonardo.phys.washington.edu/feff ). This version is also the most stable and fixes bugs in previous releases (8.10 , 8.00 and all experimental versions, which are designated with an x e.g., 8x39). In my view all earlier versions than 8.20 should be trashed. Matt is right that we haven't been consistent with version nomenclature, since 8.20 or 8.2 are often used interchangeably. Matt summarized well the differences between FEFF6 and FEFF8 in a previous message to this list. In my view, the key difference in FEFF8 for EXAFS is the availability of self-consistent scattering potentials and hence a better approximation of E0. In my opinion reducing the uncertainty in E0 should improve EXAFS analysis since this should also reduce uncertainties in distance determinations. However, the differences may not be that significant. Thus for EXAFS FEFF6 is often just as good. On the other hand, FEFF8 also yields electronic structure, e.g., charge counts, charge transfer, and densities of states, information which complements the structural information from EXAFS alone. A drawback in using FEFF8 is that getting the SCF loop to converge for complex materials can be tricky, and one of our challenges has been to develop a more automated procedure. Until that is available, we recommend always using the recommended defaults in FEFF8.2 in the SCF potential construction, unless one has good reasons for doing otherwise. We would be very interested to learn of cases where FEFF8.2 yields results which are inferior to those from FEFF6. J. Rehr
participants (3)
-
John J. Rehr
-
Kelly, Shelly D.
-
Matt Newville