/Background handling

Short overview from discussions/breakouts:

• Background subtraction is dependent on:
  • sample system/geometry/the observable physics.
• Each system has to be treated in a different way.
• A survey will help identifying the necessary / common steps among the community for specific classified systems. Potentially very helpful input for GISANS from a recent paper (Jung and Papadakis, J. Appl. Cryst. (2023). 56, 1330–1347) regarding strategy to simulate and fit GISAXS patterns.
• In case of multiple scattering: Include McStas as standard for GISANS analysis?

Notes:

• How do you know which type of (multiple coh/incoh.) scattering exists? And then which model for background to use accordingly? Are there approximations?
• always use multiple terms of the DWBA approximation? Where to draw the boundary?
• should we aim at always including a ray tracing approach (like McStas) to consider multiple scattering effects? (see union components McStas): https://mads-bertelsen.github.io/tutorial/Union_tutorial_1_processes_and_materials.html
• this all would strongly affect fitting of GISANS - we guess that this is one reason that no fitting of GISANS exists at the moment?
• typical approach of subtracting background doesnt work if it comes from "the sample itself, i.e., by multiple scattering in the sample"
• can this be "tested" how much the sample affects the background, are there test samples for comparison?
• why not the "typical approach": you subtract the instrument background (has to be well known), and all other "background" has to be from the sample and has to be simulated/fitted? this has then to be known for all wavelength bands and incident angles
• comparison to QENS where signals are always weak - how is that handled? Can what is known from there be taken over? QENS: Start with approximations for the model, and this then has to be refined. Also there the precise knowledge of the sample and the estimations of multiple scattering involved have to be taken into account!
• measure multiple states / dispersions / other observables to decrease ratio (unkown parameters)/(measured parameters)
• how to judge if the simulation (even if fitting perfectly to the data) is the physical correct one? For the question of which parameters are influencing the cross section mostly using bayesian fitting: see papers from Josh (for reflectometry): https://journals.iucr.org/j/issues/2021/04/00/ge5096/ge5096sup1.pdf
• X-ray community:
  • Very helpful the recent work (Jung and Papadakis, J. Appl. Cryst. (2023). 56, 1330–1347) regarding strategy to simulate and fit GISAXS patterns, potentially very helpful for GISANS.
  • Grazing incidence background subtraction is very difficult already in the X-ray techniques. State of the art for X-ray techniques is looking for areas around the strong peaks. This solution is not optimal. You can’t put down the sample and so the best situation is to do the same experiment at a liquid-liquid interface with motion. Everything is in the same place, you put only the liquids in the same cell, GISAXS image with everything except the molecule which will be placed at the interface. Then make a new sample with the same position of the interface with an added molecule. The background sample has to be as close to the real sample as possible. = To have more information, not to be taken as the recipe for handling background.
  • Surface scattering needs to be known to disentangle it from reflection.
• ToF aspects:
  • Consider also ToF Background!
• How to acquire the Background:
  • There is no general way to acquire data, many people do empirical background subtraction because there is no way to calculate it.
  • Cut and put a polynomial under a peak without knowing what the polynomial is supposed to represent.