Gravitational lensing is one of the most useful and spectacular consequences of General Relativity which can result in observing multiple lensed images of a distant quasar. It is the most accurate technique to weigh galaxies, study their dark matter content/mass profile over cosmic history, and one of the few techniques that allow the measurement of the Hubble constant H0 with <4% accuracy. At present only ~40 quadruply-imaged quasars are known (Ducourant et al. 2018), due to the small separation of components (<1"). The Gaia space mission offers for the first time an all-sky survey with an exceptional spatial resolution of ~ 0.2" up to magnitude G~20.7 mag. Its final Data Release (DR) is expected to contain, out of two billion sources, ~250 quadruply imaged quasars (quads) including compact (separations < 0.8") configurations and many more doubly-imaged quasars. This represents a ten-fold increase with respect to the actual number of known lenses.

Efficient and automated methods to detect lenses candidates are necessary to realize the full scientific potential of Gaia. To that end, our Gaia Gravitational lenses group (GraL) has already developed proof of concepts of supervised and unsupervised machine learning methods, suited to Gaia positions, fluxes (Delchambre et al. 2019) that can be complemented with ground-based images and time-series (Krone-Martins et al. 2019). Our prototypes have succeeded in discovering ~12 new quads (Stern et al 2020 in prep., Wertz et al. 2018) and 7 new doubles (Krone-Martins et al. 2019) in Gaia DR2. These prototype algorithms suffer from several drawbacks that limit their efficiency. In particular one method relies on a training set coming from an unrealistic simulation of the relative positions and magnitudes of the lensed images. A real distribution of deflecting galaxies in terms of morphologic type, scale length of the disc component, spatial distribution has to be considered. The colour information of the sources is also missing while it is mandatory to discard contaminants in the list of candidates (false positive). The several images of the background quasar must indeed have the same colour but can either suffer absorption by the lensing galaxy or can be blended due to instrumental or atmospheric effects.  

During the thesis, the student will perform realistic simulations of Gravitational Lenses by implementing a real distribution of QSOs within a cosmological hydrodynamic simulation. These simulations will be used to train the supervised machine learning algorithm for the blind detection of gravitational lenses in the Gaia data (DR3, 2021). The derived candidates with the highest probability will be selected for spectroscopic observations at ESO-VLT. A morpho-spectral component analysis will then be applied to the observations as a way to disentangle the quasar and galaxy spectra in order to estimate their redshift (eg. Joseph et al. 2016 and Joseph et al. 2019). This constitutes a first step towards a H0 determination.

The PhD student will be closely associated to the observing campaigns of the GraL group in major international facilities (ESO/NTT, ESO/VLT, Palomar, Keck) for ground-based validations of candidates and will actively participate to future proposal writing and data treatment.

It is mandatory that the applicant has a good knowledge in programming, is motivated by simulations and has serious interests in extragalactic Astronomy. The thesis will be conducted within an international environment related to the Gaia-ESA space mission, and the ability to work in group is mandatory.