Should effective cross sections be assigned at the gas level?

[daanboer]

Yesterday me and @janvdijk were discussing the test case presented in #235. The test case provides an effective cross section for N2(X), but provides populations for N2(X) and N2(A). LoKI-B then computes the elastic cross section from the effective by subtracting the inelastic cross sections, and assigns this cross section to both N2(X) and N2(A).

This is a little unintuitive. For example, what would happen if the effective cross section is attached to N2(A), or N2(X, v=0)? In my opinion, the only correct way to define an effective cross section is to define it on the level of the gas (or charge level), so e.g. e + N2 -> e + N2, or e + N2+ -> e + N2+. The resulting elastic cross section would then be assigned to all populated states of the corresponding gas. We could allow the user to "overwrite" elastic cross sections for specific states, but perhaps this is one of the cases were limiting the user is better.

@IST-Lisbon, @AntonioTejero what do you think?

@janvdijk IMO we can start the implementation, but instead of throwing an error we can issue a warning and move the effective cross section to the charge level?

[IST-Lisbon]

I agree that all this is a little unintuitive, starting with the very existence of "effective" cross sections in LXCat, as we discussed in our meetings with @daanboer, @janvdijk and @AntonioTejero .

To summarize my position in these discussions

1. LXCat 3.0 must reconcile the migration of effective cross sections with a detailed definition of the electronic structure of target states. In this respect, I no longer recognize the concepts of 'gas' or 'charge' levels as valid targets. Targets should always be defined by their electronic structure, e.g. N2(X) ou N2+(gnd).
2. Since effective cross sections are still out there, LoKI-B assumes them assigned only to electronic states of the neutral molecule. This decision is both (i) consistent with what has historically been defined in LXCat and (ii) practical — for example, I do not know how to construct an elastic cross section from an effective assigned simply to “N2”. Should we use rotational + vibrational populations for the ground state? And what about the electronic excited states? Don't we believe that, originally, N2 was meaning N2(X)?
3. We should avoid extending the concept to states where it was never used. Therefore: (i) N2(X,v=0), N2+(gnd) should not have effectives assigned; (ii) IST-Lisbon / LoKI-B assign effectives to the electronic ground-states of molecules, e.g. N2(X), O2(X), even if this was not clearly indicated in the original data by Phelps; (iii) Currently, LoKI-B accepts effectives assigned to any electronic state, other than the ground-state, e.g. N2(A). This is just for consistency. Personally, I have never seen this done, especially because effective cross sections were historically assigned to unspecified states..
4. I understand that the previous decisions limit the direct use of older datasets in LoKI-B, such as those of Phelps. Personally, I prefer this restriction over modifying the code to accommodate vague or poorly defined data, such as "Effective E + O2 → E + O2 "

@AntonioTejero, do you want to add anything ?