Speaker
Description
The four year eROSITA survey revealed over one million quasars. The quasar phase in galaxy evolution is thought to be essential in transforming their host galaxy. However, the triggers of quasars and their impact is still uncertain. We aim to map the environmental conditions favorable to the growth of black holes, in particular its complex dependence on host galaxy stellar mass, star-formation, quasar luminosity, obscuration, cosmic time, etc. We simulate a eROSITA-like survey considering the X-ray survey depth, by combining the known galaxy population's stellar mass function with a parametric, probabilistic triggering probability dependent on stellar mass, luminosity, etc. From these physical parameters we simulate measured, noisy fluxes in 15 UV to MIR filters with the GRAHSP code. Our simulator overall matches observations, in terms of colors as a function of redshift, redshift distribution, and r-band flux distribution. Whether the actual observed fluxes lie within this simulated distribution requires a density estimator. The kernel density estimation models struggled with the dimensionality of the data. While the spline Normalizing Flow (NF) is more computationally expensive than Real NVP (Non-Volume Preserving), it does not underestimate the density in the flux distribution tails, and requires only a small training set of roughly seven thousand samples. We demonstrate the ability to learn astrophysical parameters by scanning over one parameter and evaluating the survey likelihood. In the future, optimizing the likelihood tunes the simulation to better match the observed massive surveys, inferring the link between quasars and their host. This will be essential as eROSITA is complemented by deep photometry surveys such as VRO/LSST and Euclid, while spectroscopic redshift completeness catches up.