A source with a range of `Energy+-100eV` is used (E0=E0,dE=100 eV).
Energy takes the value of the scan that is done.
The scan (channel cut that changes angle and the Energy value of the source that changes) is done from 4.5 to 27 keV.
It is of importance that the source has a dE that is both small but big enough (explained [further below](#dE)).
The glitches found by McXtrace are shown in green.
The "theory glitches" are shown in yellow. They are found with a script that checks which hkl planes contribute to each glitch and assigns a deviation intensity by adding their respective `F^2` reflection values listed in the Si.lau file generated by cif2hkl.
The ssrl database glitches are shown in red.
The relative intensities are of interest. The intensities of the McXtrace and theory glitches found at the `6457` eV energy are made to be equal to an intensity deviation of `30.7` degrees which is the value of the ssrl database for that energy. In doing so, the relative intensities can be compared.
### Zoomed in
The relative intensities of the McXtrace and theory glitches are mostly the same.
Not always though as can bee seen for the glitches at the energies `6457` and `17084` eV. The theory glitches finds that the 17084 eV glitch is bigger than the 6457 eV glitch, whereas it is the opposite for McXtrace (and ssrl-db). See explanation for this [further below](#dE).
Overall, these relative intensities (both McXtrace and the theory glitches) differ (mostly) with the ssrl database. This is the problem.
## Filling up the glitches
### High enough Ncount
When there are not enough rays that hit the detectors (too low of an Ncount), a simple problem manifests itself (this is a computer problem not a physics problem).
When there is a glitch, if the ncount of rays of the primary plane of diffraction is not high enough, then it won't "cover up" the glitches. By covering up one means to reduce the intensity deviation of the glitches.
Therefore, this gives us glitches with higher intensity deviations than they should have.
(Example picture of this:
6 glitches with large source to 4 glitches with mono source (same as ssrl db))
There are many ways that lead to having bad statistics, i.e. not enough rays that hit the detectors:
- When using a large source (4 to 40 keV) with 1e7 rays, the statistics are not good enough. Not enough rays hit the detectors.
- When detuning 50% instead of 40% of the fwhm, less rays hit the detectors. Just the right amount of detuning to both eliminate harmonics, see the fundamental and most importantly have enough of an Ncount has to be done.
And there are ways to increase amount of rays hitting the detectors:
- Increasing the mosaic does help in increasing the Ncount, but the mosaic=1 seems to fit for the ssrl glitch db. This can be seen when looking at a particular glitch and measuring it's width for different mosaics. 1 is the best option, more than that and the McXtrace glitches become wider than their ssrl db counterparts. So the mosaic is left at 1.
- Using a (quasi)monochromatic source helps tremendously in increasing the Ncount.
- Using SPLIT on the two crystals.
### dE
A high enough Ncount is not the only reason of deviation intensity reduction of glitches. The explanation in the interpretation paragraph of the [first version](../v1#interpretation)'s doc is also a reason.
In short, the rays in the beam with a slightly different energy hitting the crystals at a slightly different theta will contribute to lowering the intensity deviation of the glitch.
This is only possible if there is both a theta and an energy divergence.
If one tries a scan with a very small dE, such as one for example, then the glitches won't be "filled in".
## Large source scan, interesting result
A source with a range of `4 to 40` keV is used (E0=22 keV,dE=18 keV).
The scan (channel cut that changes angle) is done from 4.5 to 5.1 keV.
The two glitches at the energies `4589` and `5020` ev that are found in the theory glitches are also found with McXtrace. This is expected.
But the interesting part is the intensity deviations at approximately: `4755` and `4929` are also found. And these are not found as glitches in theory. This is the interesting part where one does not expect any intensity dip, yet there are some found both in ssrl db and McXtrace.
The noise is quite large because the statistics when using a large source are much worse.
## Miscellaneous things tried
Calculating the `F^2` by using the atom positions: `F^2 = cos(h*x+k*y+l*z)^2+sin(h*x+k*y+l*z)^2`, where hkl are the plane indices and xyz the atoms positions. This did not yield good results.
Using mantid to calculate the `F^2`. The glitches seem similar (mostly) to the theory glitches. For higher energies, much less glitches, and/or much smaller. One confusing thing is there is no 111 plane in the hkl planes generated by mantid.
Powder diffractograms to see if the hkl intensities are the same as in the litterature for Silicon.
## Questions and improvements
### Improvements
- Take the lines with an `F^2` equal to 0 out of the Si.lau file.
### Questions
- Weird behaviour of SPLIT in trace.
- Even if an hkl plane has a reflection equal to 0,
i think im right in thinking that even if a hkl plane has 0 as his reflection, it can still get hit, therefore we can still see it when the Ncount is low even though it has an `F^2`=0. This is what happens when using a large source.
When using a small source, the Ncount is high enough and the glitch is "filled up".
