Meeting notes 29/11/2023

[astellhorn]

Presentations: 1) Annika: overview of polarization correction workflow and polarization components 2) Simon: overview on a typical pipeline in scipp

Discussions:

• To be inserted by Annika (see word file)

To do:

• Annika: send around ppt of polarization correction
• Annika/Hal/scipp-team: Review workflow
• Annika/instrument-scientists: speak about monitor correction and where to insert instrument-dependent stuff

[SimonHeybrock]

I am in particular unsure about how the time-dependence of the polarizer and analyzer transmission functions interact with the time-dependence of the monitors (and other normalization terms). Consider, e.g., SANS normalization (or the equivalent definition for $I(Q_x,Q_y)$:

Assume we have event-mode detector data and event-mode monitor data. We have the following options:

1. Handle polarization after computing $I(Q_x,Q_y)$. This would imply we need wavelength and time information. This could be done:
   1. in event mode (by normalizing in event mode[^1]) to handle time and wavelength dependence for the detectors, but I don't think it would work unless the monitors are unaffected by the polarization corrections, since the monitor also depends on time and wavelength.
   2. in histogram mode, computing $I(Q_x, Q_y, t, \lambda)$
2. Handle polarization before normalization. We can then independently treat detector data and monitor data before feeding them into the above equation.
   • Are there terms that actually cancel out here? May depend on relative position of monitors and polarizer/analyzer.

[^1]: This can cause trouble with error propagation

[SimonHeybrock]

Conclusion based on discussion:

1. $I$ from paper is $C/M$.
2. Apply polarization correction to that vector
3. Sum over time and wavelength afterwards?

Questions (for @SimonHeybrock):

• Can we do this in event mode?
• Do we need to worry about zeros in denominator?

[astellhorn]

First, here is the .pptx and .pdf file of my presentation, after including some corrections and clarifying the calculations and background measurements Annika - Overview presentation on pol analysis in general.pdf Annika - Overview presentation on pol analysis in general.pptx

Overview table workflow.docx

[astellhorn]

Second, now a comment on the beam monitor correction:

• In a paper from NIST ( doi:10.1107/S0021889812003445) I find that the 3He Direct Beam measurements should always be normalized to the beam monitor first, and I think this should be only the incident beam monitor before the sample, NOT the monitor after the sample position (be careful, this paper is written for a reactor source and for having a supermirror polarizer) - hence, here we should correct for M(lambda) first (?)
• In a paper from JPARC (arXiv:2005.14399v1 [physics.ins-det] 29 May 2020), they have characterized their 3He cell while including corrections from e.g. detector efficiency. Hence, maybe all corrections (especially detector efficiency and normalization to the incident beam monitor before sample) should be done before the Polarization corrections (?), but not including the wavelength to Q conversion ? --> Hence, this would be according to our last conclusion: Conclusion based on discussion:
  1. I from paper is C/M
  2. Apply polarization correction to that vector
  3. Sum over time and wavelength afterwards?

[astellhorn]

To think about:

• get 3He time decay: calculate 3He(C_DB(R,lambda)/M(lambda)) (to get transmission-functions as function of spin state T+,T-) after integration around direct beam

• to correct sample data: apply PA(C(R,lambda)/M(lambda)), before recalculating to Q ---> need to think about if this is applicable

• two issues: 1) is the incident beam monitor M before analyzer, but after polarizer ? --> answer from Mikhail and Sebastian: at the moment planned before polarizer! Mikhail: maybe need one after polarizer? would this one be just shifted then and we do not have one before anymore? 2) if bg with and without cells are same, it should not matter, but if not?

[astellhorn]

Simon, could you describe in detail again here how the data correction for unpolarized SANS is done, i.e., in which order exactly you typically do the different correction steps for monitor correction and solid angle integration etc? And why you think that if we do the sum over time and wavelength after PA correction, why you expect zeros in the denominator, i.e., which exact monitor is used (because I see it is a M(lambda) monitor, but no dependence on R).

[astellhorn]

SOrry for some previous mistakes. I have updated the pptx-file with the overview from tuesday again and also have uploaded a word file with a table on input data, input parameters, output parameters, for each part of the workflow. To comment on this:

• previously, we have made the workflow for each cell with assuming we need the information from both polarization channels through the cell, up and down. ---> after speaking with Hal, he mentioned it is sufficient to measure only one channel (either up OR down), as they mathematically should be the same and also experimentally this is a sufficient assumption. Hence, we dont need to split for the spin channels for the DREAM instrument, for later instruments that will have to be decided again
• I have changed some naming according to more typical convention, i.e.: Opacity O (instead of mu), and decay-time T(instead of Gamma)

[SimonHeybrock]

Simon, could you describe in detail again here how the data correction for unpolarized SANS is done, i.e., in which order exactly you typically do the different correction steps for monitor correction and solid angle integration etc? And why you think that if we do the sum over time and wavelength after PA correction, why you expect zeros in the denominator, i.e., which exact monitor is used (because I see it is a M(lambda) monitor, but no dependence on R).

• Description: https://scipp.github.io/ess/instruments/loki/sans2d_to_I_of_Q.html
• The implementation with Sciline is here (for Sans2d): https://scipp.github.io/esssans/examples/sans2d.html

[SimonHeybrock]

And why you think that if we do the sum over time and wavelength after PA correction, why you expect zeros in the denominator, i.e., which exact monitor is used (because I see it is a M(lambda) monitor, but no dependence on R).

In what I have linked above, we are using a time-integrated monitor. What I said earlier was probably wrong, I thought we would compute something like C(t)/M(t), then to the time-dependent polarization correction, and then sum. But obviously that cannot be the case, since $\sum_t C(t)/M(t) \ne \sum_t C(t) / \sum_t M(t)$.

So now I assume we simply compute $C(t) / M$, i.e., ignore time-dependence of the monitor? It is still unclear to me why that is fully correct, i.e., if we see low counts during some part of the polarizer "decay" but more in others, shouldn't that be taken into account? Or will that simply lead to a slightly suboptimal weighting of the different contributions? Will we actually compute a weighted sum in the end?

[astellhorn]

First I need to understand if you mean here the integration over time for one time_bin? I.e., ? --> we now thought that it would be fully correct to make the monitor corrections for each neutron pulse as stated above with single integrals (not on ) (this should be correct as the Monitor in DREAM is placed before any polarizing component, i.e., should not be impacted by the time decay of the He-cells, and in general any incident beam monitor SHOULD not be place behind the time-dependent 3He-cells) --> then to apply all other typical unpolarized SANS corrections for each neutron pulse --> then use these "unpolarized-SANS-corrected" C of the Direct Beam measurements with He-cell to get PA-correction factors for each neutron pulse --> then apply the correction-matrix on the "unpolarized-SANS-corrected" C of the sample measurements for each neutron pulse --> Then sum up over all neutron pulses

[astellhorn]

In general, does it matter if the "typical SANS corrections) (i.e., DB-function, solid-angle correction, ...) is done on time or on wavelength?

[astellhorn]

Also, does it matter if we do

or

?

[SimonHeybrock]

I don't think I can follow, maybe we should discuss this in person next week.

[astellhorn]

I also think thats the best! I think it is quite useful to meet twice, on 13th and on 19th, if that is still ok with you