cosmo hydro sims with AREPO with different models

[dnelson]

Some AREPO cosmological simulations will have details that differ significantly from our TNG "base case".

e.g. different units and/or different unit systems, different field names, different particle type number usage.

(This issue edited to reflect the discussion/direction below).

[arkordt]

I would like to support this issue and offer support for extending and generalizing scida in that regard. Here is a related issue that I encountered:

summary of the arepo setup

• cosmological simulation of a 1 (Mpc/h)^3 box
• these arepo modules produce output data
  • TRACER_MC as particle type 2
  • SGChem with radiative and chemical properties saved in particle type 0
  • supernova remnants maintained in particle type 3
  • SPRAI radiative transfer with sink particles as particle type 5
• basically, I am using an identical simulation setup as in Magg et al. 2022

After defining some additional units for quantities, opening the snapshot works fine. While following the scida documentation regarding visualization with holoviews and datashader, generating the dataframe including all fields does not work. Excerpt from the error report

...

File ~/miniforge3/envs/astro/lib/python3.10/site-packages/scida/fields.py:504, in FieldContainer.get_dataframe(self, fields)
    502     idim = int(k[i:])
    503     k = k.split(k[i:])[0]
--> 504 v = self[k]
    505 assert v.ndim <= 2  # cannot support more than 2 here...
    506 if idim is not None:

...

File ~/miniforge3/envs/astro/lib/python3.10/site-packages/scida/customs/arepo/dataset.py:1658, in Temperature(arrs, ureg, **kwargs)
   1655     m_p = m_p * ureg.g
   1656     k_B = k_B * ureg.erg / ureg.K
-> 1658 xe = arrs[\"ElectronAbundance\"]
   1659 u_internal = arrs[\"InternalEnergy\"]
   1661 mu = 4 / (1 + 3 * xh + 4 * xh * xe) * m_p

...

File ~/miniforge3/envs/astro/lib/python3.10/site-packages/scida/fields.py:661, in FieldContainer._getitem(self, key, force_derived, update_dict, evaluate_recipe)
    659     return field
    660 else:
--> 661     raise KeyError(\"Unknown field '%s'\" % key)

KeyError: \"Unknown field 'ElectronAbundance'\""

Reading through src/scida/customs/arepo/dataset.py, I think that the ArepoSnapshot class should be much more general, e.g.

• lines 90-97: other than PartType0 and PartType1, there should not be any other default alias
• lines 1640-1664: PartType0->ElectronAbundance does not exist but could be inferred from ChemicalAbundances array

[dnelson]

If you make a visualization of something simpler like Density, does it all work OK?

Either: (i) the current "Temperature" field should be made specific to TNG, and it should be clear that a user of a 'new' type of simulation needs to define how they calculate temperature, or else (ii) a "Temperature" field needs to be generally available, and somehow its details should simulation-specific?

[cbyrohl]

Apologies @arkordt , this update slipped my attention.

The first problem ("lines 90-97") is related to the default naming of aliases. Correct me if I am wrong, the problem right now is if you use aliases (like "tracer") rather than the underlying "PartType3", it will show the wrong particle type? Did you have a fix in mind? E.g. that these aliases should be configuration file specified unless some meta data is available?

Your latter problem ("lines 1640-1664") sounds like #185 which was fixed in #186 by only defining "Temperature" in Arepo Snapshots (for now) if its defining fields from above recipe are available. Could you try again? Thanks! [Right now this implies that there simply is no Temperature field available otherwise]

I would be very happy if you could help generalizing the interface. Otherwise, could you provide me with a snapshot to take a look myself?