Insert marketplacemodel bulk ferromagnets

[astellhorn]

I have added the marketplace model "SANS in bulk ferromagnets". It consists of the files "magnetic_functions.c", "micromagnetic_FF_3D.c", and "micromagnetic_FF_3D.py". In the issue "Insert model "SANS of bulk ferromagnets" from the model marketplace #2784", I have mentioned that it would be good to (i) add the parameter descriptions for the choosable parameters in the GUI (e.g. radius = "Radius of the core", thickness = "Thickness of shell"), and (ii) to add the description of the whole model (given in the .py file) to the sasview webpage. Unfortunately, for both I do not know how to do this - could you have a look on that?

[butlerpd]

I have not fully reviewed the docs or functionality yet. However some of the failing tests can be traced to the problems noted here.

Another question, perhaps for everyone, should we move magnetic_functions.c to /lib and make them part of the available library routines? That would probably require some documentation updates (regarding available library functions)?

[dehoni]

MacOs Failing not finding Python 3.7. It wasn't me!

[butlerpd]

we recently dropped support for testing on python 3.9 because mac runners didn't support it. We apparently did not update sasmodel runners. 3.7, 3.8, AND 3.9 should be removed. We should add 3.11. I'll open an issue for this.

[butlerpd]

It now loads and runs and produces reasonable curves as far as I can tell not being a magnetic person. I note the name attribute however is different from the python and c file names. The simple load seems to work but my recollection is that we still have places which use the file name instead of the name attribute? which is why they are always the same.

Is there are reason they are not the same?

Also, looking at the functions in magnetic_functions.c I really think these should be moved to the library as they are definitely more general than this model. Moreover it means that if people want to make a change and distribute on the marketplace, they won't need to carry the magnetic_functions.c around.

Note: the mac errors should go away once we merge the latest changes to the release branch

[butlerpd]

@butlerpd to talk to @pkienzle about how to handle

[krzywon]

I can't speak to the correctness of the model, but I do have a few comments in other parts. The limitation set by sasmodels, which only allows magnetic analysis on 2D data, is limiting you, which, I assume, is why you are recreating many of the built-in magnetic functions. Your approach might be a way to bridge the gap to allow more magnetic analysis in the future.

[pkienzle]

A more general question for something like a magnetic cylinder model which has orientation in both the nuclear and the magnetic scattering.

For non-magnetic models, kernel_iq.c rotates the model into the canonical orientation and calls Iqac for shapes with rotational symmetry or Iqabc for more general shapes. This removes the need for every model to implement the orientation transformation.

The underlying model here is a sphere so no need to reorient the shape. Instead it is using the Iqxy without any nuclear orientation parameters, but two magnetic orientation parameters.

If we try to implement magnetism on an oriented shape, then either we will need to move the rotation from (qx, qy, qz=0) back into Iqxy function, or we will need some way to indicate that the magnetic orientation parameters need to be rotated in the same way as the nuclear parameters so we can determine the projection into the scattering plane.

A related issue is dispersion in magnetic orientation. The main reason we converted to the Iqabc form was to handle orientational dispersion consistently regardless of the orientation of the shape. The solution was to apply two separate rotations, first applying the dispersion rotation, then rotating according to orientation. Actually, the inverse since we are rotating the measured (qx,qy,qz) into (qa, qb, qc) for the shape in canonical orientation.

With magnetic models implemented using Iqxy we will need to include dispersion parameters in the model parameter table and include the dispersion loop directly in the code.

[dehoni]

Here is a Mathematica notebook from the original paper that can produce testdata.

testdata_Michels_Honecker_2013.zip

[krzywon]

Based on the feedback from our call yesterday, the only outstanding issue here is the failing Ubuntu build. The unit tests for the model pass. I am working on the Ubuntu build in release_1.0.8_ubuntu_amd, so I think this can be merged. @pkienzle, can you give one final look, please?

Out of date.