[question] bandpass dependent Rp/Rs parameterization

[jpdeleon]

I am wondering if it is possible to implement in allesfitter a parameterization where Rp/Rs is bandpass-dependent.

[tdaylan]

Hi Jerome,

The current implementation allows nuisance (baseline and error) parameters to be modeled separately in each band in addition to band-common parameters for each companion (e.g., inclination, orbital period, epoch). However, it does not allow one to sample from the joint probability density function of radius ratios in each band in addition to the above parameters.

Nevertheless, you can i) first do a multi-band run with the same radius ratio across different bands, ii) then use the inferred band-common parameters for each companion (e.g., inclination, orbital period, epoch) as priors in multiple runs for each individual band, leaving the radius ratio free in each case.

Best, Tansu

[MNGuenther]

Hi @jpdeleon and @tdaylan,

I think this is actually do-able in one step. For example, you have a light curve from LCO in g' band, and one from LCO in i' band. You label those data files als LCO_g and LCO_i, meaning you treat them as separate instruments. Then you set up a joint fit as usual for two instruments. Meaning both get their own errors, their own baselines, and their own limb darkening.

Now, the only twist is that you'll want to scale b_rr according to bandpass For this, you can just fit a dilution parameter to one of the light curves. For example, LCO_g is fit without dilution to determine b_rr in the g' band, and LCO_i additionally gets a parameter called dil_LCO_i to scale for the bandpass effects in the i' band.

Example file attached: params_example_for_multi-color.csv.zip

Best Max

ps: if you want to go even further and have, for example, LCO_g, LCO_i, and SPECULOOS_i light curves, that is also already do-able. The second twist will then be to use the coupled_with functionality. However, to keep things simple for now, let's keep this for a separate question once it arises.

[tdaylan]

Hi Max,

True, thanks for the reminder. One thing you still want to watch out while using this parametrization, though, is that the the other parameter which should also depend on the bandpass (i.e., (Rs+Rp)/a aka rsuma) will still be constant across bands. This would only be problematic for cases such as if you are analyzing the spectrum of a relatively large planet whose atmosphere escapes readily at a few atmospheric scale heights or otherwise manifests large bandpass dependence of Rp due to scattering at short wavelengths.

Best, Tansu

[jpdeleon]

Hi @tdaylan and @MNGuenther,

Thanks for the quick responses. I think adding a dilution parameter can work as a proxy to scale Rp/Rs per band as Max suggested. But I also agree with the argument of Tansu that the result might be biased. For the meantime, I will try the suggestion of Max and compare the results of the model with dilution parameter to that where each light curve is run per individually with tight priors on passband-dependent parameters (if that makes sense).