Unknown absorption in spectrum of DESI

Question

I give a expample in url https://www.notion.so/DESI-absorption-feature-12666eae7b3880328666de6e689c9a3a

Answer 1

Thank you for contacting us. The link provided does not contain any information about the spectrum other than an image, which does not have the unknown absorption feature highlighted. Could you please at least provide the DESI TARGETID and the RA, Dec of the object?

Comments

thanks a lot

Answer 2

If you are referring to the multiple downward spikes to flux=0, I'm guessing that those are masked values (e.g. due to cosmics) and would have MASK != 0 and/or IVAR == 0 at the wavelengths of those features and they should just be ignored. Agreed with BenjaminWeaver that we'd need TARGETID or TILEID+FIBER to investigate details more.

Comments

Thanks, yes I mean those spikes to flux=0, could you please tell how to find the underlying causes of these MASK?

---

spectra and coadd files have B/R/Z_MASK HDUs which match the B/R/Z_FLUX HDUs.  In these masks, MASK=0 is good, and non-zero values are a bitmask indicating the reason(s) that that spectrum/wavelength is masked.  The reasons for each bit is documented in the SPECMASK in https://desidatamodel.readthedocs.io/en/latest/bitmasks.html#specmask-bitmask .

e.g. if MASK[i,j] == 132, that means that spectrum i wavelength j has both bits 2 (COSMIC) and 7 (BAD2DFIT) set, since 2^2+2^7 = 132.

Caveat: that description is for access to the data *files*.  If you are accessing the data through the SPARCL interface (see https://github.com/desihub/tutorials/blob/main/01_getting_started/06_PublicDatabaseAccess.ipynb), then I'm not sure what mask information is propagated and we'd need to check with the SPARCL experts.