understanding onesource::substellar

[johannct]

Several remarks :

• name is bad : should be something like substract_stellar_component_in_IR
• bool substar should not be passed as an argument; the condition on it should be in photoz_lib
• I do not understand the comment only for lambda_rf>25000 when compared to the condition allFilters[k].lmean / (1 + consiz) < 25.
• when ab[k]-fluxMod<0, abIR[k] is set to sab[k]. Why? I thought that in LePHARE allowing below 0 fluctuations of fluxes was a thing. What is the statistical rationale behind setting abIR[k] to sab[k]?
• if I am not mistaken the last else condition means that sab[k] has to be negative, right? In which case busfir[k] and bscfir[k] should already be 0 no?

If I understand well, this code amounts to subtracting the predicted flux from interpolation of the best fit (interpolation in magnitude in the space of the templates, occurring in interp_lib). So what remains are IR emission lines and what else?
This is what gets fitted in fitIR so it means that the additional library for IR has templates for this component not coming from standard library. Can you write down what these are please, so that I can also put some explanation in the C API.

[olivierilbert]

here are some answers:

• we can change the naming. Fine for me. substract_stellar_component would be fine.
• the comment is wrong. It should be "only for lambda_rf<2500"
• In the case ab[k]-fluxMod<0, I think it would have been more logical to set abIR to 0. That was a choice of Stephane for his 2013 paper. I can ask him why. But you don't want to keep the negative value, because if the fit in optical of the stellar component of the SED is going wrong, you don't want to subtract it and get crazy value. It's highly model dependent.
• sab could be negative in case of upper limit, and the band is used in such case. I don't think it implies that budfir is set to 0 (i don't see why). So, I would keep it like this.

In this function, we substract the light coming from stellar population to the observed flux, in order to keep only the dust emission. Then, we can fit the dust emission (without the stellar component) using specific templates and get the LIR. To substract the stellar component, we fit templates like BC03 (without dust emission) in the UV-visible wavelength domain. The predicted flux at larger wavelength can be substracted from the observed ones, leaving only the observed dust emission (in reality, also the AGN but we don't include it yet).