Implement Neutrino decay in PISA (Prob3)

[anilak41]

• Anil Kumar (anil.kumar@desy.de)
• Sharmistha Chattopadhyay

Modify numba_osc_kernels.py

• Passed decay_flag and mat_decay to osc_probs_layers_kernel()
• Introduced get_H_decay()
• Modified get_transition_matrix()
• Modified get_transition_matrix_massbasis()
• Introduced get_dms_numerical() - seperate function to diagonalize non-hermitian hamiltonian
• Modified get_product() to handle general case

Modify numba_hostfunctions.py

• Modified propagate_array()
• Modified propagate_scalar()
• Modified get_transition_matrix_hostfunc()
• Modified get_transition_matrix_massbasis_hostfunc()
• Modified get_product_hostfunc()
• Introduced get_H_decay_hostfunc()
• Introduced get_dms_numerical_hostfunc()

Modify prob3.py

• Passed neutrino_decay (boolean to turn on/off decay)
• Passed decay_alpha3 parameter
• Modified propagate_array() to include decay_flag and decay_matrix

New - decay_params.py

[LeanderFischer]

Please pull the master branch again and resolve the conflicts.

[anilak41]

I have resolved the issues, and the numba_osc unit test is running fine on single and double precision on the local machine.

The modification of the numba_osc unit test has resulted in some new ref pkl files and these ref pkl files are not being recognized by Python 3.7 during the above check on git which resulted in the failure of the test and aborted the test for python 3.10 also.

Note that the double precision test on Python 3.10 was successful and it was checking for Python 3.7 when the test was aborted.

How can this issue be solved?

[anilak41]

The PISA installation dependencies show that at least Python 3.8 is needed, then why are we still checking these tests for Python 3.7?

[JanWeldert]

Looks ok to me. Have you checked if you get the same osc probabilities with your code setting neutrino decay to 0 and the old code?

[anilak41]

Yes.

• I have compared oscillation probabilities using old PISA and modified PISA for $\alpha_3 = 0$. They match exactly.
• I have also compared PISA with OscProb for a non-zero value of $\alpha_3$, and they also match nicely.

Please see the plots in the attached presentation.

Neutrino_decay_PISA_Benchmarking.pdf

[JanWeldert]

Alright, should be good to go then.