Integrated line fluxes - (pseudo)continuum subtraction

[jacopo-chevallard]

The fits run smoothly now but I discovered a potential issue with the procedure.

As far as I see, BEAGLE considers the values of the spectral indices as referring to line fluxes only. However, in my spectra I have no detection of the continuum, so the limits I derive are for continuum+line. The correct fit would require no pseudo-continuum subtracted by BEAGLE from the template. The problem is that the code returns an error if I do not put any pseudo-continuum window in the SPECTRAL INDICES CONFIGURATION. Is there any way to have BEAGLE not subtracting the continuum when computing the spectral index?

Originally posted by @MarcoCast79 in https://github.com/jacopo-chevallard/BEAGLE-general/issues/92#issuecomment-762836284

[jacopo-chevallard]

@MarcoCast79 you've raised also another issue in your messages on Slack, which I put in a new issue #96.

About this issue, I misunderstood you (and mixed up things) when I answered on Slack: we do not subtract the underlying absorption lines in Beagle, but we do subtract the continuum to obtain the line flux. There isn't currently an option to avoid doing this, but again we can think about a workaround (would be curious to know what @eclake thinks)

[eclake]

Hi @MarcoCast79 @jacopo-chevallard In general I think it doesn't matter when you only have line detections (not continuum) that you just define a continuum region far enough away from your line (i.e. take the instrument LSF into account) to give to BEAGLE. As @jacopo-chevallard says, we subtract the continuum to obtain the flux - in your case the continuum is effectively within the noise and you just can't estimate it but your uncertainties on derived line flux should account for that. There is a slight discrepancy that one has to consider if you're modelling the lines as Gaussian but expect there to be significant stellar absorption. Within BEAGLE we integrate the continuum-subtracted model over the wavelength range you specify which should be similar to you performing an integration on the data in a similar way. However, if you fit with a Gaussian (or function you think well represents your LSF) to a line that should be the summation of two Gaussians (the absorption profile and the emission profile) the model is no-longer wholly appropriate and might lead to some biases. I sometimes care about these biases at the level of wanting to understand stochasticity of star formation but in the case of noisy line fluxes it's probably fine as long as the uncertainties look representative. (sorry for the long wordy reply, I hope it's helpful)

[MarcoCast79]

Hi @eclake and @jacopo-chevallard thanks for your replies. Actually, my problem is that the upper limits I have are on the continuum+line so I would need BEAGLE not to subtract the continuum when using my limits in the fit. A possible workaround might be to give to BEAGLE a spectral window where the galaxy has zero flux in the continuum. What do you think? I see that if I give very short wavelengths (e.g. 60AA) the templates are effectively zero, so the code should not subtract anything and use the limits in practice as limits on continuum+flux. Concerning the shape of the line I don't think it is an issue because I have limits obtained from a direct integration on a given wavelength range, I'm not assuming Gaussian shapes.

[eclake]

ooo, I like your workaround @MarcoCast79 - I'll let @jacopo-chevallard if he sees any potential issues with that but I think it could be a good approach.

[jacopo-chevallard]

@MarcoCast79 I don't see any issue with that, the limits can be selected in an arbitrary way, they don't have to be forcedly around the line :)

[MarcoCast79]

great, thanks, I'll keep you posted on what I find