The integrated effect of Black Hole feedback on galaxies: machine learning take on cosmological simulations and observations

Understanding the physical processes responsible for ceasing star formation in galaxies is one of the most important unresolved questions in the field of galaxy evolution. Decades of multi-wavelength observations delivered abundant evidence in favour of different mechanisms shutting down (or, ‘quenching’) star formation within galaxies, including e.g. galaxy interactions, virial shock heating, supernova explosions or AGN feedback. Among these, the influence of supermassive black holes has often been stipulated as the main driver behind galactic quiescence. However, until now, no study would unanimously identify AGN feedback as the key quenching mechanism among the highly inter-correlated set of galaxy scaling relations.

In this work we investigate how star formation is brought to a halt in local, massive, central galaxies by comparing Sloan Digital Sky Survey (SDSS) observations with three state-of-the-art hydrodynamical cosmological simulations – EAGLE, Illustris and IllustrisTNG. We address the complex nature of quenching by invoking machine learning techniques to determine which galactic property is the most predictive of quenching. We find that the supermassive black hole mass (MBH) is the most powerful parameter in determining whether a galaxy is star-forming or quenched – a statement which is true for all three implementations of AGN feedback in the simulations. Remarkably, this prediction is precisely confirmed in the SDSS observations, where we infer MBH from a variety of calibrations for ~230 000 local galaxies.

We further show that the observed correlations between star formation quenching and parameters like stellar mass, halo mass or black hole accretion rate result from the connection between these properties and MBH. Moreover, we find that the black hole accretion rate (and, by extension, AGN luminosity) is almost irrelevant for quiescence, explaining why observations of optically selected AGN struggle to find evidence for AGN feedback driven quenching in action. We then extend our study to redshift z=2, demonstrating that the uncontested importance of integrated AGN feedback for star formation quenching takes hold already at the Cosmic Noon. This Abstract is based on results presented in Piotrowska, J. M. et al. 2022 and Bluck, A. F. L., Piotrowska J. M. et al., submitted.