Improve NCrystal-McStas integration

[tkittel]

Collecting here a few features that would improve the combined NCrystal-McStas capabilities or just make the integration more easy for users. Ideas were discussed in various chats with @willend and @mads-bertelsen. Opening just a single issue here for now, but dedicated issues for specific items can be added as sub-issues later (or as issues in the mccode project):

• Improve NCrystal-Union integration in classic .instr files, by making it possible for materials to also provide absorption values. This will do away for the need for invoking the currently utility script for generating the code.
• Add NCrystalFilter component (name to be finalised), which takes an NCrystal-cfgstr for an isotropic material and caches the cross section curve during initialisation. During TRACE it then simply attenuates the beam accordingly. This provides a way to use NCrystal x-sect curves in a context where one does not care about the scattered neutrons, and where multiple scattering is unimportant. We could even consider using the minimc to make it possible to use multiple scattering effects or more flexible definitions of what is scattered (i.e. only attenuate anything scattering more than 0.1degree). Notably this would work on GPU. This would for instance be useful for Al-windows or Be-filters.
• Make it possible for PowderN to accept NCrystal cfg-strings to define the material. PowderN would then get the reflection list via calls to NCrystal.
• (after the previous item is done) Make it possible for PowderN to run ncrystal_cif2ncmat on the fly, like it can run cif2hkl currently.

[mads-bertelsen]

In terms of making it easier to get NCrystal absorption into a material, I think it could make sense to have a dedicated Make_NCrystal_material, next to the existing NCrystal_process. Making a material directly would give a natural interface to the absorption. It is possible to add an optional input to processes that would be added on top of the NCrystal_process, but the default would be that the material just has the NCrystal_process and appropriate absorption.

So as I understand it this issue suggests something like:

a_process = NCrystal_process(cfg="")
extra_process = Phonon()

a_material = Make_material (process_string="a_process,extra_process", absorption_from="a_process")

And I am suggesting something like this instead:

extra_process = Phonon()

a_material = Make_NCrystal_material(cfg="", extra_processes="extra_process")

[tkittel]

Looks good to me @mads-bertelsen . I especially like the simple scenario:

a_material = Make_NCrystal_material(cfg="Al_sg225.ncmat;temp=400K")

Because you are then down to the same number of lines (1) as is used in OpenMC or Geant4 for NCrystal materials :-)

But when actually composing more advanced union materials with different processes, I kind of also like the more explicit composition, where one has to say from which material comes the absorption, etc.

I am sure we will discuss in more detail later, but good to think about!

[tkittel]

Update: mccode-dev/McCode#2371 was merged into McCode, so now the McStas PowderN component can accept config strings.

There is eventually still more to do on that item:

• Support multiphase materials
• Add the same capability to Single_crystal.comp
• Consider the helper function ncrystalpowdern_determine_sigma_inc (ncrystal_info_t ncobj) added in PowderN: consider if something similar could be added in ncrystal.h, to make such use-cases easier. Or alternatively, in a dedicated "library" snippet in McCode.