aglie
commented
5 years ago Hi Jared,
Thank you for the very nice question. There are several things on needs to know to use size effect in Yell:
- Size effect happens in crystals which are substitutionally disordered and where atoms are allowed to move. It is expected to be accompanied with substitutional correlation and displacement correlations.
- This means that for the calculation one needs not to forget to define the zero-neighbour correlation for occupancy and for ADPCorrelation. The ADP correlation is tricky in practice because part of it comes from the TDS diffuse scattering (around Bragg peaks, usually removed when preparind diffuse scattering), so the ADPCorrelation for zero neighbour might be smaller than just the average ADPs. Nontheless some ADPCorrelations in the zero neighbour are a must, otherwise there will be plenty of negative intensities.
- Size-effect is not a perfect approximation - it introduces ADPCorrelation, as you correctly noted, so if you would want to interpret the values of the ADPCorrelations later, you should take it into account
- Size effect is an approximation (it does not completely preserve the average structure) so even if everything is done correctly there might still be some negative intensities, but thouse are usually not a problem.
There is an alternative way to describe ADP correlations which might avoid these problems: to describe your displaced position as a split-position and to use only the substitution correlations. This is easier from the maths point of view, but will introduce plenty more of variables to refine.
I hope this helps
We have been running into some issues with negative scattering intensity, and we cannot track down the source. We believe it has something to do with the implementation of the size effects.
Should size effects always be written with an accompanying ADP correlation with amplitude (se^2)/2?
The Yell reference is not entirely clear. It seems like the ADP correlation is intrinsically part of the SE term. Yet, the example regarding zero-neighbor correlations with size effects specifies that there must always be an ADP correlation of magnitude (Se^2)/2 accompanying the SE term. An earlier example doesn't include the ADP term, though.
In any case, when we do include the ADP term, the negative scattering gets a lot worse. If we only use it on the zero-neighbor correlation vector, the negative values are almost (but not completely) gone.