[GSAS-II] GSAS-II Model selection

[Von Dreele, Robert B.]

Hi,
I’d also add that the LS may shift size or mustrain to too small values if the start is too large (i.e. it over steps). Result is a size = 0.010 or mustrain very small; these are default minimums since negative values are “illegal” for mathematical reasons (as well as physical ones) & GSAS-II has reset them. If this happens to size, reset to 0.1 & repeat refinement. If mustrain is very small, reset to 100 & try again. They should then shift to the best value which is usually larger, but could be smaller. Especially when using the “generalized” mustrain model (of Popa & Stephens), always draw the mustrain surface for comparison to an equivalent isotropic model; the coefficients are not directly comparable due to the nature of the functions.
I need to emphasize again the need for high Q & high resolution visible diffraction peaks for successful refinement of size & mustrain with more than isotropic models. Also ask yourself about how the sample was made. If hand ground, the size will not usually be small enough to cause broadening, but there will be lots of mustrain that could be complex. If the sample is a precipitate or has been annealed, then the size & mustrain may be both “ideal” & cause little broadening at all. Nanomaterials (by definition) will have size broadening but unknown (& probably difficult to determine) mustrain because of the broad peaks. Always apply the “Occam’s Razor” principle & look for the simplest model that fits the data.
Bob

Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows

From: Toby, Brian H.<mailto:toby at anl.gov>
Sent: Sunday, January 22, 2023 10:39 AM
To: kasa belachew<mailto:kasabelachew at gmail.com>
Cc: Von Dreele, Robert B.<mailto:vondreele at anl.gov>; GSAS-II at aps.anl.gov<mailto:GSAS-II at aps.anl.gov>
Subject: Re: [GSAS-II] GSAS-II Model selection


On Jan 22, 2023, at 6:59 AM, kasa belachew via GSAS-II <gsas-ii at aps.anl.gov<mailto:gsas-ii at aps.anl.gov>> wrote:

It is so good, that, GSAS-II computes the crystallite sizes (  *10-6* m) in three different (isotropic, uniaxial and ellipsoidal) and Microstrains (isotropic, uniaxial and generalized) arrangement.
Q1# Can we use more than one model for the same material?

Are you asking, can one use (for example) isotropic and ellipsoidal crystallite size together? That is not possible and would not work if it were allowed, since they would be completely correlated.

If you are asking about mixing different models for crystallite size and microstrain? Yes, that definitely possible and it makes complete physical sense that a material with nearly isotropic crystallite sizes could have highly anisotropic microstrains (or the opposite). But, see next comment on most datasets being insufficient for doing this.


Q2# Why the value became different with different model of the same material?

I answered this before in my previous reply. This is because crystallite size and microstrain (and possibly background) tend to be quite correlated and most datasets can only be used for an accurate fit to one of the two broadening types. When you are overfitting there are many possible results that give about the same fit. In that case, the results are pretty arbitrary and changing one set of parameters causes the others to change. To fit both microstrain and crystallite broadening together, you need clearly resolved peaks at high Q, say >90 deg with Cu Kalpha. Even then it is only possible if both types of broadening are significantly larger than the instrumental broadening. Clearly high-resolution synchrotron and neutron measurements will be much better for this.

Brian


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.aps.anl.gov/pipermail/gsas-ii/attachments/20230122/467b4e82/attachment-0001.html>