Path of galaxies along the main sequence: recovering their most recent star formation history.

We use the ABC method to perform a model choice. 

For each galaxy: which is the most appropriate SFH to use between

• a tau-delayed SFH
• a tau-delayed SFH + flexibility (to model a rapid, recent, variation of SFH)

To do this, we model a catalogue of COSMOS-like galaxies using CIGALE and train and test the algorithm on it.

See the figures below.

We apply the ABC method on a subsample of COSMOS galaxies.

For this proof-of-concept study, we limit the analysis to a narrow range of redshift (0.5<z<1) and impose that galaxies are detected in all bands with a SNR>10.

These criteria are due to statistical limitations of the Approximate Bayesian Computation that still need to be developped to better fit the problematics and constraints of astrophysical data.

As a result: the majority of COSMOS galaxies have a probability p lower than 0.5 implying that the tau-delayed SFH is sufficient to model their SED.

However, there are galaxies for which a flexibility is needed, meaning that these sources must have undergone a rapid enhancement/decrease of their star formation activity.

Thanks to the ABC method and the work of Aufort et al. 2020, we now have a sample of galaxies for which there is a high probability (>90%) that they just underwent a rapid variation of SFH.

Since the ABC method does not tell us if this is a decrease or an enhancement of the star formation activity, we use CIGALE to estimate the physical properties of these sources, using the appropriate SFH.

In Ciesla et al. 2021, we focus on rapidly quenched sources. You can find two examples below.

Knowing the appropriate SFH to model the SED of our quenched galaxies sample, et after intensive tests on our ability to constrain the SFH free parameters (see Ciesla et al. 2016 for instance), we are able to put these galaxies on the SFR-M_* diagram as well as their previous position, before the rapid SFH variation.

Indeed, our simple SFH model provides us with the age when the variation occured as well as its amplitude. We can then model the SED at the time just before the variation happened.

On the SFR-M_* diagram below, we show the position of the COSMOS quenched galaxies as well as their position just before quenching.

We made an attempt to recover the IR luminosity (L_IR) of our galaxies before their quenching.

We used the constraints we obtained on their SFH and model the SED back before the quenching happened. Assumptions had to be made on the past attenuation of the galaxies. However, to control our results, we tested that the past properties of our quenched galaxies are consistent with the properties of galaxies with the same redshift and M_* that were not quenched and still on the main sequence.

From this, we obtained the L_IR before quenching that we compared to the L_IR now, as a function of the age of quenching (age_trunc) on the figure below.

The cyan square are local galaxies of the Herschel Reference Survey: we applied the same method on them as they serve as a benchmark for our study. Indeed, they are well-known objects undergoing ram-pressure stripping in the Virgo cluster, and for which we now the recent SFH and all properties.

The figure below shows the decrease of the L_IR as a function of the quenching age. COSMOS and HRS galaxies probe different timescale and are thus probably suffering from different quenching mechanisms.

Original study:

Ciesla L., Elbaz D., et al. 2018

Introduction of the Approximate Bayesian Computation and first application to real galaxies:

Aufort G., Ciesla. L, et al. 2020

Going back in time, recovering past properties of galaxies:

Ciesla L., Buat V., et al. 2021