Avoiding Bad Data from Correlation

Austin

Hi all,

I have designed a .inp file with a focus on site occupancy refinement. I have noticed that both surface roughness (likely due to poor equipment) and scale are both highly correlated with site occupancy refinement of the material I am analyzing and significantly increase error of the data. I have been attempting to discover a way to automate the process of first refining scale/surface roughness with site occupancies fixed then refining site occupancies with scale/surface roughness fixed to no success. I figure that this would involve some type of custom macro or combination of prms, but I am unsure how to proceed at the moment. I would eventually like to incorporate this into a sequential refinement .inp file if possible. Any help would be greatly appreciated!

rowlesmr

Surface roughness, site occupancy, scale, displacement parameters, and probably a few other things, all impact on the intensity of the calculated pattern. You're going to get correlations just from the very nature of the regression.

What is the nature of your surface roughness? Does it fit the criteria for the roughness corrections you're using?

Austin

rowlesmr Thank you for the reply! I will admit that I have essentially learned everything related to TOPAS on my own through manuals and tutorials on YouTube, so I am not proficient by any means. I am currently using the Surface_Roughness_Pitschke_et_al macro refining the first parameter with the second fixed as it generally approaches zero under the xdd file. The powder being analyzed is a nanopowder with a relatively flat surface visually. The refined surface roughness factor is very small and has little effect on the outcome of the refinement to my knowledge. If I did comment the surface roughness macro out, how could I refine site occupancy and scale independently within the same file? If I were to refine the two parameters together, the error increases by an order of magnitude and the fit of the refinement degrades significantly.

rowlesmr

I wouldn't use surface roughness unless you had a very good reason.

You need to refine scale to get things to on a common intensity axis, so you can't really change that.

So it comes down to whether or not your data support the refinement of site occupancies. Another method is to do an elemental analysis and set the occupanies to those suggested from the elemental assay.