[newchem-cpp] problems related to dust grain destruction

[mabruzzo]

There are some issues related to dust grain destruction that must be addressed before newchem-cpp can be merged

• What does the dt parameter should represent in lookup_cool_rates1d_g? There is an inconsistency we need to resolve!
  • In solve_rate_cool we pass the full timestep that solve_chemistry is passed. In step_rate_newton_raphson, we just pass in the local subcycle timestep. This is only used for dust destruction.
  • I think the code only makes sense if we are modelling dust-sublimation as an instantaneous process (in which case the subcycle timestep makes more sense).
  • An extended discussion on this topic can be found here. Note: that discussion was written before I understood the difference between sublimation and sputtering (I also misread the code and mistakenly thought we were using gas temperature rather than dust temperature)
• If we are indeed modelling dust destruction, we need to document it! Furthermore, I don't think it is implemented properly:
  • In this scenario, we should be setting dust densities to zero before we compute the heating/cooling rates and other reaction rates (I think like H2 formation). Essentially I think the right way to implement the model is to operator split off the instantaneous destruction.
  • if we are near the dust temperature threshold, I suspect that the sharp discontinuity in the time derivatives of the dust density could make the newton-raphson scheme do some weird things (but I don't know for sure). In this case, operator splitting would avoid issues
  • If we decide that we don't want to model it the properly, we need to clearly document it (and should acknowledge the things that may go wrong)
• I'm definitely not an expert in dust chemistry (and I have very little intuition about dust temperature), but the fact that we don't model sputter implies that our dust model is HIGHLY incomplete
  • it seems improbable that sputtering is totally unimportant for all modeled grain species
  • I suspect that this logic was only tested in very particular scenarios and we don't know how the behavior generalizes in other contexts
  • It is plausible that the dust grain chemistry is only relevant for heating/cooling and the rest of the chemical network in very specific contexts, and that the current model produces mostly reasonable heating/cooling/chemistry in the other contexts, other than the amount of dust.
  • Based on my limited understanding, dust sputtering is important in T>2e4 K gas. So the lack of sputtering may produce inaccurate abundances in galactic outflows. (It's plausible that this could be somewhat important to chemistry/heating/cooling in gas that has cycled through a galactic fountains)
  • We need to be very clear about the limitations. Or, if we don't know what the limitations are, we need to make that extremely clear in the documentation (and discourage people from using this network unless they have expertise in this area)