Clusters of galaxies are the most massive entities in the universe.
Applying artificial intelligence to the cosmological analysis of X-ray cluster surveys allows us to tackle this problem from a totally new perspective. Only directly observable parametres are used (redshift, X-ray colour and flux) in a deep learning approach based on hydrodynamical simulations; this allows us to establish an implicit link between the X-ray parameters and the underlying dark matter distribution. From this, we can infer the cosmological parameters, without explicitly computing cluster masses and bypassing the empirical formalism of scaling relations between the X-ray properties and cluster masses.
The goal of the thesis is to apply this method (developed at DAP) to the XMM-XXL survey, which is, 24 years after the XMM launch, the only programme having assembled a cosmological cluster sample with controlled selection effects (~ 400 objects). The expected results will constitute a first in the history of observational cosmology.