On Monday 27 October 2003 03:10 pm, J. M. An wrote:
/ Dear IFEFFIT helper, />/ />/ I am a novice of IFEFFIT and would like to ask some questions about />/ fitting variables used in the program. />/ I've been trying to fit my FEFF 8 calculation result to an experimental />/ data for a bulk compound CdSe. />/ Essentially, the fitting variables, S02 and DW factor (the correlated />/ Debye model) for the first-shell, />/ are used for the fit. I can manage to fit quite nicely with those />/ variables, but the problem is that I want to use the values />/ of the resulting fitting variables as input for CdSe quantum dot />/ calculations. />/ />/ Normally, S02 and Debye keywords in FEFF are global paramaters, not />/ restricted to a single shell, so there seems to be />/ no way that I can put the fitting values obtained from the fit into new />/ FEFF calculations. />/ I was wondering if someone has experience to use bulk fitting values as />/ input in a new FEFF calculation. / I don't think I quite understand what you want to do, but here are a couple pieces of advice:
1. It is my strong preference never to introduce S02 or sigma^2 into the feff calculation when I am doing exafs analysis. It is much easier and much less confusing to use S02=1.0 and all sigma^2=0 in the feff calculation. That way, any values that you get for those parameters from a fit using ifeffit are absolute numbers and do not require that you remember what you did in feff in order to interpret the ifeffit result. 2. I do think that redoing the feff calculation might be useful at some time during the fit, but only if I have reason to believe that the *structural* parameters used in the feff calculation were in error relative to the real material. If the distance between atoms in the real material are quite different from what you used in the feff calculation, then you might worry that the phase shifts and scattering amplitudes are not as good as they might be. Given all that, I am not sure what you hope to accomplish by your new feff calculation. Hope that helps, B *********************************************************************** Thank you for the advice. The idea behind my research is (1) to extract information S02 and Sigma2 by fitting my feff result to experiment in crystal CdSe. (2) then to use this information to perform feff calculations for quantum dots CdSe of a certain shape. In other words, the info in (1) will be used as input to quantum dot calculations in (2). The reason is that no one knows how the surface terminates in the dot. To this end, I need to know reasonable fitting values of S02 and Sigma2 to be used in a feff input file. Suppose that I have obtained S02 and Sigma2 by doing (1) only for the first shell, leaving all other shells intact. (This is because in crystal CdSe, our calculated FEFF chi(k) using correlated Debye model differs from experiment for large k.) However, my problem is how we put these fitting values in (2), since SO2 and Sigma2 are global to all paths in Feff. Another complication may be that in (2) I am performing a set of calculations for different Se absorbing atomic sites. At the end, add all contribution from different Se sites to total chi(k), which is exactly what experimentalists measure. That is why I need to put resonable S02 and Sigma2 before I essentially perform quantum dot calculations. I will appreciate any further advice of yours on this, and let me know if this is not sufficient for you to understand what I intend to do. Joon Lawrence Berkeley National Lab
Dear Yoon, On Mon, 27 Oct 2003, J. M. An wrote:
The idea behind my research is (1) to extract information S02 and Sigma2 by fitting my feff result to experiment in crystal CdSe. (2) then to use this information to perform feff calculations for quantum dots CdSe of a certain shape.
In other words, the info in (1) will be used as input to quantum dot calculations in (2). The reason is that no one knows how the surface terminates in the dot. To this end, I need to know reasonable fitting values of S02 and Sigma2 to be used in a feff input file.
Suppose that I have obtained S02 and Sigma2 by doing (1) only for the first shell, leaving all other shells intact. (This is because in crystal CdSe, our calculated FEFF chi(k) using correlated Debye model differs from experiment for large k.) However, my problem is how we put these fitting values in (2), since SO2 and Sigma2 are global to all paths in Feff.
Another complication may be that in (2) I am performing a set of calculations for different Se absorbing atomic sites. At the end, add all contribution from different Se sites to total chi(k), which is exactly what experimentalists measure. That is why I need to put resonable S02 and Sigma2 before I essentially perform quantum dot calculations.
I will appreciate any further advice of yours on this, and let me know if this is not sufficient for you to understand what I intend to do.
If you are trying to forward-model EXAFS for particular configurations, than applying a best-fit S02 to the full calculation could make sense. Similarly, applying a best-fit Debye Temperature could be useful. It may not be perfectly applicable to quantum dots, but if it's all you've got, it's better than nothing. But if I might suggest another approach, I would replace Feff's primitive sum-of-paths with the sum-over-paths in Ifeffit and/or Feffit. In Ifeffit, you would simply define a set of paths, and use the ff2chi() command, In Feffit you can simply not specify any data to fit. With this approach, you can do a series of Feff calculations with the standard S02=1, sigma2=0 (these parameters aren't applied to the feffNNNN.dat files anyway), and still easily average over paths from different configurations, and can apply disorder and amplitude reduction factors as you like. --Matt
On Monday 27 October 2003 07:28 pm, J. M. An wrote:
In other words, the info in (1) will be used as input to quantum dot calculations in (2). The reason is that no one knows how the surface terminates in the dot. To this end, I need to know reasonable fitting values of S02 and Sigma2 to be used in a feff input file.
Matt's suggestion of using ifeffit, feffit, or Artemis to handle your forward modeling is a good one. It is certainly possible to run the fitting program without fitting -- that is, using it to make a sum of paths more elaborate than is possible using Feff alone. Not only can you assign S02 and sigma^2 on a path-by-path basis, but you can handle concepts like fractional occupancy (by weighting a path by a fraction). That is difficult to model using Feff alone and is probably important to modeling the termination of a nanoparticle. In Artemis, you can do a sum of paths without a fit by setting up your "fitting" model then selecting "Sum of paths for this data set" from the Data menu. B -- Bruce Ravel ----------------------------------- ravel@phys.washington.edu Code 6134, Building 3, Room 222 Naval Research Laboratory phone: (1) 202 767 5947 Washington DC 20375, USA fax: (1) 202 767 1697 NRL Synchrotron Radiation Consortium (NRL-SRC) Beamlines X11a, X11b, X23b National Synchrotron Light Source Brookhaven National Laboratory, Upton, NY 11973 My homepage: http://feff.phys.washington.edu/~ravel EXAFS software: http://feff.phys.washington.edu/~ravel/software/exafs/
participants (3)
-
Bruce Ravel -
J. M. An -
Matt Newville