Persistent ValueError when fitting spaxels without 6.2 micron PAH

[astrolojo]

Context: I am using CAFE to extract attenuation information in ~4500 MIRI/MRS spaxels. To minimize the runtime over such a large dataset, I’ve turned emission line fitting off so I can do my own Gaussian fitting separately, after subtracting off the best-fit continuum+PAH models. I should also note that some spaxels near the edge of my MRS Channel 1 field of view can have substantial gaps in spectral coverage, which I attribute to wavelength calibration and drizzling in the MRS pipeline. These gaps often include some or most of the interval between 5.5 and 6.5 micron.

Description: Because some edge spaxels are missing the 6.22 micron feature, CAFE ingests the spectra but fails to plot them, fit them, etc. Trying to do so raises the following error:

ValueError: Input spectrum does not contain any PAH band available to perform the PAH component scaling.

I understand from comments in the source code that cafe_helper.init_feats() “cannot deal with spectrum with gaps” and should raise an error if gaps are present. However, the ValueError persists even when the affected spectra have all NANs replaced with finite filler values. Also, the 7.6, 11.3, and 17.0 micron PAH features, which come later in the PAH inspection order, are always present without any gaps, so I would expect them to to be used instead for ref_pah_wave within cafe_helper.init_feats().

Attempted solutions & potential explanations: Setting a non-zero ref_pah_peak may require Gaussian line fitting to be turned on, since it depends on having a nonzero value for the variable flux_left (on, e.g., lines 307 and 309). I can confirm, however, that the ValueError persists for these choppy spaxels even when FitLINs = True.

I'm happy to give more information about my installation, other attempted hacks/solutions, etc as needed.

[tdiazsantos]

Hi astrolojo,

Thanks for your comment. We will look into how to make this scaling more robust but a quick solution to your problem would be to try not fit the 6.2um (or any other noisy) PAH feature, since CAFE will likely not be able to provide a good fit anyway. To do so, you can edit the "pah_template.txt" file within the "tables/" folder in your CAFE installation directory. There you can simply comment (#) or delete the PAH lines you don't want to fit, and only leave the features located in less noisy parts of the spectra, so CAFE will be able to make a better scaling with a well measured PAH in the pah_inspection_order list.

Let us know if that works.

Tanio