Alternative dust continua

[drvdputt]

Problem: the standard Modified blackbody -- Blackbody(lambda; T) * (lambda**-2) -- shape for the dust continuum does not work for certain objects.

Idea: the lambda**-2 term represents an absorption coefficient of the dust. As a start, we could replace it by an empirical extinction curve. This is already in the works in the PAHFIT branch for a paper I'm writing.

I also want to highlight the shape of the continuum of the CAFE tool, shown in e.g. this paper https://arxiv.org/abs/2410.09020 (I'm writing this up now because this recent paper reminded me of it). The hot component has a silicate emission bump at 10 micron, which is partially reduced again by the silicate bump in their dust attenuation model. So there is a precedent for such a continuum shape being reasonable.

Implementation: To switch between the "default" MBB and alternative ones, we can use the "model" column that is already in the Features table. The backends could then to implement the available models, and provide an argument in the add_feature_dust_continuum() function of the Fitter interface.

[drvdputt]

I just realized this is a duplicate of #11, where some more info is given on what the problems are

[jdtsmith]

Historically I've shied away from silicate emission in the models, since emission and absorption tend to "fight" and you enter a regime where radiative transfer becomes the more appropriate tool. CAFE's continuum is interesting, but I know from users that it can be highly sensitive to starting point and input choice. Perhaps we could experiment with a few models for the opacity function for an MBB.

[karllark]

Definitely an issue with silicate emission/absorption fighting with each other. On the other had, silicate emission is definitely real, hence we should definitely have the option to include different emissivity relationships other than the simple powerlaw. Especially as the powerlaw is not like any dust emissivitiy law I'm familiar with for the NIR/MIR.

[jdtsmith]

Good thoughts. One idea to avoid the fight that's working well from M51 is to use H-recomb to form an independent estimate of tau_Si, fix that (or with a small range). Then you could allow a more complicated MBB emissivity to sit "on top" (or behind/mixed). We may need a bit more expressiveness in the science packs to accomplish this.

Absorption/Attenuation

• "Normal" attenuation curve (power-law + Si-O resonances), perhaps with a few sub-options, e.g. one with high crystalline fraction. Karl, we should discuss the current best "updated" curve to include. I know the 18µm resonance is challenging for full self-consistent UV-IR.
• Discrete ice absorption: H20/CO/CO2.

Continuum Emission

• Traditional $\nu^2$ MBBs
• "Fancy" MBBs with more realistic emissivity model (e.g. silicate emission). Can we get away with one emissivity law?
• Traditional T=5000K starlight
• Fancy template-based starlight (for e.g. <3µm)

Geometry

• screen
• mixed
• options for applying screen or mixed by a subset of the absorption components to a subset of the continuum components, and/or a subset of discrete (line/dust feature) emission. I can imagine how to compute this, but how to specify is a bit more challenging (and what defaults to use).

E.g. "attenuate all emission (mixed), apply CO/CO2/H2O absorption only to dust continuum (screen)".

[karllark]

I may be biased, but I would say that the R(V) dependent extinction curves in Gordon et al. (2023) are the best for the NIR/MIR. These are based on a sample of spectroscopic measurements in the NIR (Decleir et al. (2022) and MIR (Gordon et al. 2021). While the sample is not as large as would be ideal, it is the largest to date. Hence, these works as merged/summarized in Gordon et al. (2023) are our best empirical measurements of the full NIR/MIR curve, including the 10 and 20 micron silicate features. We will get updated curves in the NIR/MIR including the 10 micron silicate feature. We may get an update to the 20 micron silicate feature with JWST, but this is challenging due to the low throughput in MIRI/MRS channel 4 (main reason) and the fact that the 20 micron feature is really broad.

[karllark]

I like the idea of using the H-recombination lines. Has the usually complications as it is attenuation, coming from an extended source, etc. Regardless, maybe leave dust extinction free and compare the results with the H-recombination lines would be an interesting starting point/test. Could even be some science if we see a different in the attenuations between the ionized gas and the carbonaceous features. Really cool if we could see if different carbonaceous features from different physical regions in the PDRs (PAHs, aliphatics/HAC, fullerenes) suffer from different attenuations. Really leaning into the different attenuations for different components idea.

[jdtsmith]

My impression from work with @ThomasSYLai is that the 18-20µm resonance in the 2021/2023 curves are too wide/weak, in particular to credibly fit deeply embedded ULIRG sources, which often take a sharp turn up at 15-20µm. This could also be influenced by crystallinity. It's too bad we will struggle to nail this feature down with MIRI.

Definitely worth introducing a G23 curve into PAHFIT and seeing how we fare.

In terms of introducing an emissivity law, that gets more complicated as I understand it because, e.g. cold and hot silicates have different laws, hence strong shifts in the peak of absorption and emission.

I like the idea of using the H-recombination lines... leave dust extinction free and compare the results with the H-recombination lines would be an interesting starting point/test

Yes! There are tons of attenuation clues from HI, H2, even PAHs (see Lai+ 2024). This would be a really cool student or PD-side project. Karin also has a great data set for this. Maybe an "attenuation chat" is in order at some point soon.