Hi Jenny,
1. For linear combination fitting, there are three indicators for the goodness of fitting: R-factor, chi-square and reduced chi-square. Could anyone tell me how they work?
This is actually documented in Athena, and the Users Guide. They are also defined in the Feffit documentation. In short, they are all scaled measures of Sum [(data-fit)^2] R-factor scales this my the data values, and chi-square scales by an estimate of the noise in the data. Reduced chi-square relates to chi-square through the usual statistical definition, in that it is chi-square / (number of free variables in the fit). Of course, chi-square requires one to know the uncertainty in the data -- generally we don't have a great estimate of this. I mean no offense of this, but if you're asking about these then you almost certainly haven't put in an estimate of the uncertainty. So chi-square is probably scaled incorrectly. On top of that, reduced chi-square needs to know the number of independent measurements. Normally one assumes each datum to be independent. This is arguable, but it we can make that assumption for now. But if chi-square is scaled poorly, so is reduced chi-square. If that's too vague, or I misunderstood the question, please ask again.
2. Since TEY is sensitive for the surface and FY for the bulk (and surface?), species detected by TEY should be also detected by FY, right?
Yes, but TEY samples a much smaller volume of material than FY, so the signal from the volume seen by TEY (that is, the surface) may be insignificant compared to the signal from the volume seen by FY.
3. How to calculate the maximum analysis depths for TEY and FY?
Google/Wikipedia might help here. The sampling depth for TEY is typically dominated by the mean-free-path for the Auger electrons, which is in the range of 20 - 50 Angstroms. Sampling depths for FY are typically set by the absorption length of fluorescence x-rays, which is in the range of 2 to 50 microns (yes, a much more variable range, depending on sample composition). In both cases, you'd need to calculate the depth that the x-ray beam penetrates the sample (depends strongly on matrix) too. In my experience, it's unusual for this to dominate the sampling depth, but it can be significant for FY in high-Z matrices. --Matt