Electron scattering factor scatteringfactortable.py

[João Batista Souza Junior]

Dear Simon and Pavol,

I have one question about the possibility of changing the scattering factor for Fe and O atoms to refine G(r) of magnetite nanoparticles obtained by electron PDF on Diffpy CMI. I have seen the code and the scatteringfactortable.py on Diffpy.SrReal, which if I understand correctly, leads to a Srreal_ext (diffpy.srreal.srreal_ext) that could be related to a SO file. Is there a possibility to change the atomic scattering (f0_waaskirf parametrization) for those two atoms and run Diffpy CMI to refine electron PDF G(r). I'm answering a manuscript submission, in which I have used PDFGui (X-rays scattering), but both referees ask about electron scattering factors instead of X-rays.

I would appreciate your help.

Best regards,

Dr. João Batista Souza Junior
Brazilian Nanotechnology National Laboratory - LNNano (CNPEM)

Tel: +55 (16) 98125-9261                                  

[Simon Billinge]

Thanks for the question João.

There are two issues with electron scattering factors that we need to keep clear in our minds in the realm of PDF and ePDF studies and I am not exactly sure the origin of your question so I will briefly mention both.

1) Data reduction to get the PDF from ePDF data.   In this case we divide the data by the average of the electron scattering factors just as we do in the case of x-rays.   This is done in ePDFsuite (and other ePDF data reduction programs, presumably).

2) Subsequent modeling of the data.  In this case, because we have divided by the average electron scattering factors we don't need to make any correction here.   In the case of x-rays, we simply scale the signal by the number of electrons in the atom (we usually use the atomic number, but could also take account of ionicity if we want...the differences are quite small so we often don't bother).  The question then arises, what should we use for the scaling when modeling ePDF data?  In my mind this is an unresolved question and could benefit from a close look by someone.  I expect it will be something close to the atomic number scaling we use in the case of x-rays because the potential that the electrons scatter off will scale with the nuclear charge.  Indeed, we seem to get quite good fits when we just use the atomic numbers in PDFgui (i.e., if a Na atom is on a site, in PDFgui, put Na there).  But the scaling may be somewhat different.  We are not scattering off electrons so the simple Z-scaling is not so clear.  The Q->0 asymptotic behavior of the electron scattering factors is also not defined, so we can't use that.

This may be a longer answer than you need, but I would start with using PDFgui as if you were fitting x-ray data (assuming you used the correct electron form factors to get the experimental PDFs) and see how good the fits are.  If you want to play around, you can replace atom types in PDFgui to get a non-linear scaling with atomic number, e.g., replace Na in PDFgui with another element that is slighlty lighter, for example, and see if you get a better fit.

As I said, it may be due to my ignorance, but I am not aware of this issue having been definitively sorted out, which needs to be done.  When there is a definitive solution we can provide a "electron" radio button in PDFgui so it gets done right.

S

--
You received this message because you are subscribed to the Google Groups "diffpy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to diffpy-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/diffpy-users/CALO4jSOz0FhBAdRQXviAb3Qkidohr9JqZ4_52VCx3%2BGEdKk7pw%40mail.gmail.com.

--

Simon Billinge

Professor, Columbia University

Physicist, Brookhaven National Laboratory