fitting experimental Background

remo

Hi

I have got a scan file from an "empty run", showing contribution to the background of Kapton windows, DAC, and whatnot. It's a nasty convoluted pattern that I failed to model. Is it possible to use the XY data of this scan file and scale it as is to the background during fitting of the actual PXRD patterns?

I suppose it has got something to do, if possible at all, with fit_obj and user_Y but I don't understand the use of them, not to speak of adapting it to my case.

Thanks

Remo

alancoelho

Hi Remo

user_y can be used to as a background function.

First save your background function to a file. Lets call it BKG. If the step size of the file is equal then place at the top of the file the following:

_x1_dx START STEP

Replace START and STEP with values corresponding to your data.

Also, see examples:

TEXT_EXAMPLES\USER_Y\USER_Y.INP
TEXT_EXAMPLES\QUANT\QUANT-7.INP

If the file is in XY format then place at the top of the file:

_xy

If its an XYE file then use:

_xye

To use BKG then place the following line in the INP file:

user_y bkg BKG

Here, the name of the file is BKG and the name given to the user_y object is bkg. The name bkg can then be used in all objects that can be a function of X. For a fit_obj then use:

fit_obj = bkg;

To scale bkg then simply add a scaling parameter; let call this parameter s:

prm s 1 min 1e-6
fit_obj = s bkg;

If you need to transform the x-axis of bkg then use fo_transform_X, see user_y.inp.

remo

excellent! Thank you very much, this works as I hoped for

djurdjijadzodan

Dear professor Coelho and/or Remo,

Thank you on the post.

I am struggling to successfully apply all your instructions (perhaps because I can't access both the TOPAS examples that you have referred to).
I have collected XRD of an empty sample holder which I used for collecting XRDs of the powders of interest. I have named the XRD of the empty holder as background.raw and included the following code lines at the beginning of the JEdit Script:

   xdd thexrdpatternofthepowderofinterest.raw  
          x_calculation_step = Yobs_dx_at(Xo); convolution_step 4     

      _x1_dx 10.0002 0.0148              
      user_y bkg background
      fit_obj = bkg;
      prm s 1 min 0.00001
          fit_obj = s bkg;

Question 1.
1a) With the code lines listed above I haven't managed even to run the script successfully.
According to the TOPAS.log, the problem lies under the "_x1_dx 10.0002 0.0148 ", but I can't figure out what exactly the problem is and how to fix it. I suspect that I have also missed to include some other trivial code lines?

1b) The "START" command under "_x1_dx START STEP" I have understood as the starting x value in the XRD pattern, is that correct?

Question 2. On the page 74 in the book [1] I have found a little bit different way to include the experimentally determined background, which seems to work for me, using the following code lines in my case:

  xdd thexrdpatternofthepowderofinterest.raw  
      x_calculation_step = Yobs_dx_at(Xo); convolution_step 4     

      user_y back_expt background.xy     (I have converted the data in .xy format in this case)
      prm back_scale 1
      fit_obj = back_scale * back_expt;
      Plot_Fit_Obj(back_expt)

2 a). Do the code lines under 2 automatically also subtract the experimentally determined background from the XRD pattern of the holder containing the powder of interest or do I need to include any additional code lines?

2 b). Is there any way to, after having completed the steps under 3a, manually correct the background in the regions of XRD pattern of interest (if needed) and if yes what code lines are to be used then?

Question 3.
Would subtracting background in the way as suggested in this post "eliminate" completely contribution of the sample holder to the XRDs of powders of interest even in case when the sample holders contribute to the XRDs in the form of "humps" (please see the XRD pattern labelled as "ZBH-32 empty" in the link [2])? My main goal is to determine as more accurately and precisely as possible the unit cell parameters of powders of interest and I am worried that the humps (that originate from the sample holders) in the XRDs of powders also shift the peak positions of the main phase due to which subtracting experimentally determined background might not address the issue completely?
Do I need to consider using some other sample holder that do not "produces" humps in the XRDs?
Or do I simply need to increase amount of powder to be put on the sample holder prior to collecting its' XRD in order to "cover" the hump(s) to as larger degree as possible (this idea I have gotten based on the content on the page 92 in [1], but it doesn't sound like a wise solution, rather as a potential way to "masking" the problem).

Thank you do much in advance.

Kind regards,
Djurdjija

[1] Dinnebier, Robert E., Andreas Leineweber, and John SO Evans. Rietveld refinement: practical powder diffraction pattern analysis using TOPAS. Walter de Gruyter GmbH & Co KG, 2018.

[2] http://texray-lab.com/2015/04/17/zero-background-sample-holders/

rowlesmr

2 a). Do the code lines under 2 automatically also subtract the experimentally determined background from the XRD pattern of the holder containing the powder of interest or do I need to include any additional code lines?

They don't subtract the background, but add it to the model which then calculates your pattern including that background function.

2 b). Is there any way to, after having completed the steps under 3a, manually correct the background in the regions of XRD pattern of interest (if needed) and if yes what code lines are to be used then?

What do you want to manually correct? you can set a user_y scale factor that depend on X. you could also probably r
restraining the bkg using the Bkg_at function. See the techref:

bkg @ 0.443519294` 0.0200324829` 0.0113774736`
penalty = 1000000 (Bkg_at(2036) - 0.43)^2;
penalty = 1000000 (Bkg_at(9000) - 0.50)^2;
penalty = 1000000 (Bkg_at(14600) - 0.50)^2;

Question 3.
Words

It's always better to not have to do a correction, so a specimen holder that doesn't introduce a background would be "better", but it also depends on how practicable that is.

Thank you do much in advance.

johnsoevans

I'd only add a couple of points:

If you're only interested in cell parameters you may not need to worry too much about the details of your background. If you've got a reasonably crystalline material then broad background humps may not have much influence. If you had a nano size sample you'd have to be much more careful. One pragmatic test might be to take a typical sample and mix in a large amount of amorphous material so you have a deliberately high background. What happens to your extracted cell parameters?

You have to be a bit careful with using an "empty sample holder" background fitting as the contribution from the sample holder can change when the sample is present. e.g. absorption by the sample may mean that different parts of the holder aren't "seen" when it's filled. When we have complex backgrounds from sample environment we'll often fit an experimental background plus a low order polynomial background for minor corrections.

djurdjijadzodan

Dear professors Evans and Matthew,

Thank you so much, I have eliminated almost all my concerns regarding the issue by reading your detailed explanations. I have only an additional question (I am sorry again a super trivial one) with several "subquestion".

Question 1. Regarding calculating transmission factor for the powder and radiation of interest.

a) Would calculating the transmission factor for the powder of interest (in my case the Prussian Blue Analogues) and used radiation (X-rays) be a reasonable approach if the goal was to determine the optimal thickness of the powder to be loaded on some flat sample holder for the diffractometers that operate in the reflection geometry (as could be a reasonable approach in case of diffractometers that operate in the transmission geometry)?

b) I know that the transmission factor should have the value as closer as possible to 1, is it the value that should be targeted in the calculations?

c) I also haven't managed to find in the literature (almost) anyone who mentions the values of linear absorption coef. of the Prussian Blue Analogue of interest to me. Hence, I have understood that I would need to find a way to calculate the coef. for the given composition of the compound and the radiation type myself?

I am asking the questions since I have observed different ratio of intensities of the Prussian Blue Analogues' peaks in their XRDs when loading the powders on the same or different flat sample holders (with or without cavities) which might not be a problem for me since I am mainly interested in the accurate and precise peak positions that should enable the accurate and precise determination of the unit cell parameters of the powders (and in case if I will need to do the Rietveld refinement, I assume I will manage to find some correction to be applied in order to address the attenuation of the X-rays afterwards), but I still want to learn the do things in the most professional way possible way.

Kind regards,
Djurdjija

johnsoevans

Are you confusing transmission and reflection requirements? For transmission capillary you want mu x r = 1 for highest intensity. There are various websites that will calculate mu x r for you (e.g. https://11bm.xray.aps.anl.gov/absorb/absorb.php), or you can do it from tabulated values. TOPAS also has macros to do this (e.g. https://topas.awh.durham.ac.uk/doku.php?id=capillary_absorption).

For reflection you ideally want an infinitely thick sample if you're doing Rietveld. P42 of the textbook has one way of estimating what this means for a given sample.

As you say, intensity changes may not matter if you only want cell parameters. However it ought not happen. It's unlikely to be absorption if changes are pronounced. Could it be preferred orientation? Or perhaps with some of the set ups the beam is larger than the sample dimensions at low angle?