Hi Juan, First, the delR question. You say that one of your structures is a "platinum structure." I assume that means fcc. In an fcc lattice, every atom is equivalent (maybe you've substituted some platinum with another metal, but nevertheless the lattice points are all equivalent). The only way in which the spacing between atoms in an fcc lattice can change and maintain the fcc structure is if the lattice constant changes. This means all path lengths must change by the same percentage (not the same number of angstroms). So you could do something like this: Guess delLattice = 0.0 Def delR = delLattice*reff (reff is a variable that uses the half-path-length from the model). Now, the more general question. If your fits don't look very good (e.g. have high r-factors) before even going to higher shells, there may be something going on with your system you didn't anticipate, such as oxidation. Often a first-shell only fit "works," in that the fitted spectrum is close to the data, but might not be as useful as it could be, due to high uncertainties or an inability to distinguish between relevant possible structures. In cases like that, going to more distant coordination shells often helps clarify the structure. On the other hand, if a single-shell fit already has trouble fitting the data, I'm not so sure that will fix itself with more scattering paths and more R-range. Other notes: --remember that the "R" in the Fourier transform is not itself an absorber-scatterer distance. (Uncorrected) paths tend to peak as much as a half-angstrom below the average absorber-scatterer distance. In addition, paths have width, as you can easily see in Artemis. This width stems from several effects, including the broadening of the distribution that is measured by sigma2 and truncation effects associated with the Fourier transform. Therefore, if you want to fit up to, for example, 3 angstroms, you should probably be including important paths up to at least 3.5 angstroms or so. (The default behavior of Artemis is to complain if you include paths far above the fitting range. This behavior can be changed with the preferences dialog, or you can just disregard that particular warning each time.) --Unfortunately, it is not always appropriate to disregard MS paths inside your fitting range. Fortunately, you often do OK by including the most prominent of them in some crude way, e.g. by giving them the same parameters you give to direct scattering paths. You can always try your fit with and without MS paths and see what happens; if you're on the right track you'll often find that the fit with MS paths is better, even if you weren't too careful about how to constrain them. Basically, MS paths usually represent a small but significant correction to the direct-only fit. The choice of parameters for those MS paths represents a correction to that correction; i.e. a second-order correction, and are therefore not crucial. As always with fitting, each case is different--otherwise someone would have written a program long ago that could just DO all our fits for us. Effects that are unimportant in one system may be important in another. Fortunately, it is easy to try a variety of fits on a system to make sure that we understand it. Likewise, we should always make sure that our results make physical sense. --Scott Calvin Sarah Lawrence College At 09:10 AM 11/10/2006, you wrote:
Thanks a lot Scott, I will do it. Sorry, but I do not understand what you mean with the sentence "in a physically reasonable way: delr might need to be scaled by reff". Could I use the same delR for all single scattering paths (even distant paths)? In principle, I am not going to consider MS paths.
I hope fits increasing R-range will improve very much fits in k-space because now they are not very good.