/Using SANS software

Overview:

• Regarding BornAgain:
  • Need a set of prototype models (check what is on the BA webpage) - especially for inexperienced users
  • This should include a very detailed description about its physical limitations, directly implemented into BA sending error messages if overgone.
  • Further good features to implement: (i) intensity w.r.t. direct beam, w.r.t multiple scattering background, w.r.t. general expected background (according to experience + commissioning).

Detailed notes group 4:

• A set of prototype models for several example systems is really great piece of help for inexperienced users of BornAgain.
• However, it will be even more helpful if every model describes its physical limitations - in other words, a user should be warned if he is choosing a value of a parameter which breaks the validity of the theory behind this model (for example size of the particles, particle distribution over the surface etc). So that a user do not consider unrealistic models.
• On the top of this it will be nice if a user can estimate
  • intensity of GISANS signal in respect to the direct beam
  • in respect to the multiple scattering background calculated by BornAgain
  • in respect to a general background expected in the experiment (basing on the previous experience with similar sample environment, materials, guide halls etc).

• Sample environment should be continuously developed the knowledge is shared inside of the community.

Detailed notes group (?):

• Get inspiration from X-ray software:
  • Brookhaven, Ben Ocko
• SasView:
  • SASview people can read data in, put different resolution functions, co-fit X-ray and neutron data. For SANS it works but For GISANS not.
• BornAgain:
  • BornAgain does not have a good defined intermediate format. The challenge is to get it to do fitting rather than simulation. Peaks at the right position is simple, getting right intensities is more complicated. Pictures can look similar but not fit in reality. Fitting in 2D is much harder than fitting in 1D. Few people do it in general. Even reflectivity X-ray and neutron is much more advanced in combined view. But with GISANS it is much more difficult. X-ray industry e.g. XENOCS will sell a GI stage and a software package which will give q xyz components but the fitting will be reduced to cuts (qz at particular q xy) so it is a quasi 1d fitting.
  • Could collaborations like BA benefit from models defined by other software packages? E.g. Is there a form factor for this? Where can we find code for such form factors?
• Comparison BA and SasView:
  • E.g. the structure of SASview is easy if you want to put a new form factor. Very few people do it. In the last 6 months there have been more people doing that but in general this is one of the problems, you don’t have many people who writes the plugins. What people do more, they combine modules or their own purposes. They are not of great generality, something might be written in a paper and people elaborate the code themselves. One has to learn fro other collaborations.

Interaction between BA and SASview? ICNS conference. Code camp in Garsching for SASview. No-one from Joachim’s group was participating. There could be something to see how a collaboration works.

• Regarding data reduction / data formats (repeated there):
  • The package approach in BA might cause issues while SASview does not have them. Once the interface is defined it is hard to change it, same for the file format. One can design them to be extensible (NEXUS), to use new features. One has to think carefully what you insist people provide. For SAS data format, reflection format you need to provide certain information. The idea was that it can be extensible and if information is not provided it is simply skipped.
    • canSAS XML was developed for this. But people don’t use it in a right way, they don’t fill out the fields, data has incorrect format. It is not easy to solve these problems.
  • The challenge with BA is that we don’t understand what is the input data. Originally was built around a monochromatic instrument with cartesian coordinates around the scattered beam. Strictly speaking SASview works in q space and it doesn’t assume anything about the detector space.
  • One of the differences is how to make the software instrument independent. SASview: you provide reduced data where you have taken out all the instrument effects. The idea of q resolution without WL and angle, it can accommodate data from instruments with multiple detectors, you can concatenate them in one file.
  • Before software for SAS used to compare errors if you have an area detector there are many pixels that goes to the same bin. What is the variation pix to pix in there. Statistics over the individual pixels is compared with the variation. One needs to have more information to test that. When it comes to model fitting you don’t trouble the fitting program with too complicated procedure at the moment. The peak fitting for diffraction already causes people a lot of problems. Once you try to correct for more things you get more uncertainties.
  • With GISANS you re not obliged to throw away the original raw counts. Whether software can handle? The motivation for having a reduced dataset is to have data format that is transferable between softwares. But have you thrown too much at some point?

• Other:
  • BoToSim and BoToFit.
  • SpinW = software, neutron spinwave. It is Lucas Wilkins’ development in collaboration with another person.