NLTE

[aoifeboyle]

New approach to NLTE. This time the beta_sobolevs from the previous iteration of the plasma are used to calculate the NLTE level populations, in a similar way to what was done in the old plasma. Seems to be working well, but I'm a bit confused about the atomic data, specifically whether the lines_idx in the NLTE data should be the same as lines.index in the atomic data. It seems like that should be the case for consistency, but it doesn't seem to be the case here, and it's creating some confusion when calculating the NLTE level populations. Here, lines_index is taken from the NLTE data:

[aoifeboyle]

@wkerzendorf So I think I've got a good system here with the NLTE now. Could you have a look and see what you think? It works on a similar principle to the NLTE in the old plasma. Quick summary:

So the property cycling issue consists of two property cycles. The NLTE calculation indirectly affects the beta_sobolev and electron_densities values, both of which are also inputs for it. In the old plasma, the ion_number_density convergence loop was performed making use of the NLTE calculation of the previous run. In the case of the new plasma, it's the reverse. The NLTE calculation is performed making use of the ion_number_density outputs of the previous iteration.

Old plasma calculation sequence:
NLTE (iteration 0) --> beta_sobolev, electron_densities (iteration 0) --> ion_number_density convergence (iteration 1) --> beta_sobolev, electron_densities (iteration 1) --> NLTE (iteration 1).
New plasma calculation sequence:
NLTE (iteration 0) --> beta_sobolev, electron_densities (iteration 0) --> ion_number_density convergence (iteration 0) --> beta_sobolev, electron_densities (iteration 1) --> NLTE (iteration 1).

I hope that makes some sense! If you're happy with this, I'd like to merge it although the coverage test has failed. I plan to open a new PR for adding the tests.

[wkerzendorf]

@aoifeboyle do they converge to the same values?

[aoifeboyle]

@wkerzendorf Yep! Better accuracy than I expected in fact.

[aoifeboyle]

@wkerzendorf Think I've found a better way of doing this now, without having to output and read in values from previous iterations from model.py. Created this new property class to store values from previous iterations. The values are set just prior to all of the properties being updated. I think it's a lot more straightforward! Let me know if you think I'm on the right track now and then I'll tidy it up/finish it off.

[aoifeboyle]

Hi @wkerzendorf. Just trying to get the basic structure here to work like you suggested earlier (lines 208-230). When I run this I get the error, 'Can't instantiate abstract class LevelBoltzmannFactorNLTE with abstract methods calculate.' I assume that's just because there's no explicit def calculate()? How do I make this work?

[wkerzendorf]

you define one:

def calculate(self, *args, **kwargs):
    pass

then overwrite it

[wkerzendorf]

that is smart!