silveira
Greetings everyone,
I am trying to refine some synchrotron data integrated from 30° cakes at two different orientations, 0° and 90° of the azimuthal ring. For each data, I am trying to obtain the instrumental contribution(LaB6 powder) at the respective orientation and same cake size.
The problem is: The refinement of the LaB6 is quite good, Rwp lower than 5 both two orientations, but when I refine the data, the results are very different, in my case, at 90° the refinement is very bad, rwp higher than 40 while at 0° the refinement is "quite good", around 7~10 rwp.
By investigating the LaB6 measurements I observed that at 0° the peaks have a larger broadness and lower intensity while at 90° the peaks are less broad and with higher intensity. Considering a standard material, I would expect a uniform intensity and peak broadness over the hole ring, however, this is not the case. My guess is that this effect is caused by the beam size which had a rectangle shape (200 µm height and 1000 µm width).
As a result, the instrumental parameters are quite different at 0° and 90°. Since this behavior is also observed in the experiment data, I thought that this would not be a problem, but the refinement is very problematic.
I would like to know your opinion about the convolutions used to describe the instrument. Should I introduce additional ones or remove some?
I don't fully understand the effect of all of them, but it seems that when I remove the hat convolutions which is related to sample tilt and receiving slit width, the refinement seems to fit very good, but I don't think it is a good idea to delete parameters without knowing exactly what they mean.
Below you can see the parameters for the instrument at each orientation, the parameters in red are the ones that I fixed and copied to the .INP file for the data refinement:
0° orientation
lam
ymin_on_ymax 0.001
la 1 lo 0.12703299019795109 lh @ 0.00812` lg @ 0.26837`
prm LC_my 0.00106` min 0 max 20
lor_fwhm = LC_my;
prm GC_my 0.00106` min 0 max 20
lor_fwhm = GC_my;
prm exp_2 -0.08356` val_on_continue = Val+ Val*Rand(-0.2, 0.2);
User_Defined_Dependence_Convolution(hat, 1/Cos(Th), @, 0.04258`)
User_Defined_Dependence_Convolution(lor_fwhm, 1/Cos(Th), @, 0.00112`)
User_Defined_Dependence_Convolution(gauss_fwhm, 1/Cos(Th), @, 0.00999`)
User_Defined_Dependence_Convolution(circles_conv, 1/Cos(Th), @, 0.00010`)
bring_2nd_peak_to_top
exp_conv_const = exp_2 ;
scale_top_peak stp 0.00442083665` min 1e-5 max 1e5
add_pop_1st_2nd_peak
------------------------------------------------------------------------------------------------
90° orientation
lam
ymin_on_ymax 0.001
la 1 lo 0.12703299019795109 lh @ 0.00000` lg @ 0.29796`
prm LC_my 0.00146` min 0 max 20
lor_fwhm = LC_my;
prm GC_my 0.00146` min 0 max 20
lor_fwhm = GC_my;
prm exp_2 -0.00584` val_on_continue = Val+ Val*Rand(-0.2, 0.2);
User_Defined_Dependence_Convolution(hat, 1/Cos(Th), @, 0.36534`)
User_Defined_Dependence_Convolution(lor_fwhm, 1/Cos(Th), @, 0.00057`)
User_Defined_Dependence_Convolution(gauss_fwhm, 1/Cos(Th), @, 0.01276`)
User_Defined_Dependence_Convolution(circles_conv, 1/Cos(Th), @, 0.00013`)
bring_2nd_peak_to_top
exp_conv_const = exp_2 ;
scale_top_peak stp 595.702057` min 1e-5 max 1e5
add_pop_1st_2nd_peak
-------------------------------------------------------------------------------------------------
In the forum and in the tutorials, it is usually referred to the use of the TCHZ model, which it seems to be different from what I am using, so that's why I creating this new topic. I hope it is not a repetitive question.
Best regards,
Antonio
johnsoevans
I've only got general advice here. The first thing is to talk with the instrument scientist. They're going to know their beamline inside-out and can probably give the best specific advice.
On peaks shapes in general there are two extreme viewpoints you could take. One is to take the view that you should understand all the physics of your instrument and detector and use a full fundamental parameters approach to describe the instrumental contribution to the peak shape. The other extreme is that you ignore the physics of the instrument completely and use an appropriate standard to describe it's contribution to the peak shape. In practice, people often use a "half way house" between these extreme views.
In the second case you might use an appropriate empirical peak shape description (e.g. a Caglioti relationship, TCHZ, Pearson, etc, etc) to fit the standard. For your sample you fix these terms, then use additional terms to describe the sample contribution. In your case you could use a different instrumental description for your different patterns. Alternatively, in TOPAS you can convolute different functions together with different dependences on theta to give the instrumental contribution. As long as you fix these when you analyse your samples it might not matter what convolutions you use. It looks like your input file is following that approach.
There's a lot of detail in the tech ref on peak shapes. There's also a topas-related chapter in the book (
http://topas.dur.ac.uk/topaswiki/doku.php?id=book). Some of the convolution ideas are outlined in tutorials (e.g.
https://topas.webspace.durham.ac.uk/tutorial_joyofconv/ and related).
Hope that's some help.
John