Split ThreeFlavorDecoherence

[JostMigenda]

Changes by @jpkneller to close #384. Should be merged after #351.

My review comments to follow in a bit …

[JostMigenda]

• Unit tests and integration tests are currently failing due to these changes
• With the third commit, we now have ThreeFlavorDecoherence (defined as TransformationChain(in_sn.ThreeFlavorDecoherence())) and Equilibrate in __init__.py. What (if anything) is the difference between them?
• What about TwoFlavorDecoherence? Do we have the same situation there and would it make sense to split that up into a “black box” transformation and a separate in_sn transformation as well?

[Sheshuk]

I like the idea of this splitting! But as a user I would find a class name Equilibrate misleading: it's a verb, while all other classes are nouns

[jpkneller]

Unit tests and integration tests are currently failing due to these changes

I missed some consequences of the changes. It looks like the dictionaries in snowglobes.py are having problems.

With the third commit, we now have ThreeFlavorDecoherence (defined as TransformationChain(in_sn.ThreeFlavorDecoherence())) and Equilibrate in __init__.py. What (if anything) is the difference between them?

It's subtle. Equilibrate != ThreeFlavorDecoherence if in_vacuum != None in a TransformationChain object. Both are useful.

What about TwoFlavorDecoherence? Do we have the same situation there and would it make sense to split that up into a “black box” transformation and a separate in_sn transformation as well?

I don't think a 'black box' version of TwoFlavorDecoherence adds anything useful. It is almost the same as the CompleteExchange prescription.

[jpkneller]

I like the idea of this splitting! But as a user I would find a class name Equilibrate misleading: it's a verb, while all other classes are nouns

Let's change it to Equilibration or FlavorEquilibration then.

[jpkneller]

I like the idea of this splitting! But as a user I would find a class name Equilibrate misleading: it's a verb, while all other classes are nouns

Let's change it to Equilibration or FlavorEquilibration then.

I changed i to FlavorEquilibration

[JostMigenda]

Fixed a merge conflict with the main branch and reverted an accidental change that caused SyntaxErrors across all tests. Now the failing integration tests are due to the same remaining notebook issues that are also present on the main branch right now; but the failing unit test is specific to this PR.

[JostMigenda]

With the third commit, we now have ThreeFlavorDecoherence (defined as TransformationChain(in_sn.ThreeFlavorDecoherence())) and Equilibrate in __init__.py. What (if anything) is the difference between them?

It's subtle. Equilibrate != ThreeFlavorDecoherence if in_vacuum != None in a TransformationChain object. Both are useful.

But the ThreeFlavorDecoherence defined at https://github.com/SNEWS2/snewpy/pull/385/files#diff-e680a3b29b6442c77439d1b901a0556c6b9d0893ffa8a04a84447b9759b9e1eeR74 does have the in_vacuum transformation set to None. So is there a difference between it and Equilibrate FlavorEquilibration?

(I understand they are different from the in_sn.ThreeFlavorDecoherence, which can be used as part of a TransformationChain alongside a nontrivial vacuum transformation.)

[JostMigenda]

This implementation seems to be just copied over from the flavor_transformation/__init__.py? Unfortunately, that doesn’t work. For one, the ThreeFlavor import from snewpy.flavor is missing in this file; but even if I add that, I get an error trying to use this:

import snewpy.flavor_transformation as ft
from snewpy.neutrino import MassHierarchy, MixingParameters

NMO = MixingParameters(MassHierarchy.NORMAL)
TFD = ft.ThreeFlavorDecoherence(NMO)  # <- TransformationChain that uses `in_sn.ThreeFlavorDecoherence`
TFD.P_ff(0,0)  # <- raises an AssertionError: Incompatible spaces <enum 'ThreeFlavor_MassBasis'>!=<enum 'ThreeFlavor'>

[JostMigenda]

Full traceback:

>>> TFD.P_ff(0,0)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
    TFD.P_ff(0,0)
    ~~~~~~~~^^^^^
  File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/snewpy/flavor_transformation/TransformationChain.py", line 60, in P_ff
    return self.transforms.in_earth.P_fm(t,E) @ \
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
           self.transforms.in_vacuum.P_mm(t,E) @ \
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
           self.transforms.in_sn.P_mf(t,E)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/snewpy/flavor.py", line 197, in __matmul__
    assert self.flavor_in==other.flavor_out, f"Incompatible spaces {self.flavor_in}!={other.flavor_out}"
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: Incompatible spaces <enum 'ThreeFlavor_MassBasis'>!=<enum 'ThreeFlavor'>

[jpkneller]

I think I fixed this.

[jpkneller]

You're right, we don't need FlavorEquilibration.

[JostMigenda]

Thanks, I think that addresses all remaining issues. I just added a small fix for the corresponding unit test.