Cosmological simulations of the growth of supermassive black holes and feedback from active galactic nuclei: method and tests

We present a method that self-consistently tracks the growth of supermassive black holes (BHs) and the feedback from active galactic nuclei (AGN) in cosmological, hydrodynamical simulations. Our model is a substantially modified version of the one introduced by Springel, Di Matteo & Hernquist implemented in a significantly expanded version of the gadget III code, which contains new prescriptions for star formation, supernova feedback, radiative cooling and chemodynamics. We simulate the growth of BHs from an initial seed state via Eddington-limited accretion of the surrounding gas, and via mergers with other BHs. Because cosmological simulations at present lack both the resolution and the physics to model the multiphase interstellar medium, they tend to strongly underestimate the Bondi-Hoyle accretion rate. To allow low-mass BHs to grow, it is therefore necessary to increase the predicted Bondi-Hoyle rates in star-forming gas by large factors, either by explicitly multiplying the accretion rate by a numerical correction factor or by using an unresolved, subgrid model for the gas close to the BH. We explore the physical regimes where the use of such multiplicative factors is reasonable, and through this introduce a new prescription for gas accretion by BHs. Feedback from AGN is modelled by coupling a fraction of the rest-mass energy of the accreted gas thermally into the surrounding medium. We describe the implementation as well as the limitations of the model in detail and motivate all the changes relative to previous work. We demonstrate how general physical considerations can be used to choose many of the parameters of the model and demonstrate that the fiducial model reproduces observational constraints. We employ a large suite of cosmological simulations, in which the parameters of the BH model are varied away from their fiducial values, to investigate the robustness of the predictions for the cosmic star formation history and the redshift zero cosmic BH density, BH scaling relations and galaxy-specific star formation rates. We find that the freedom introduced by the need to increase the predicted accretion rates by hand, the standard procedure in the literature, is the most significant source of uncertainty. Our simulations demonstrate that supermassive BHs are able to regulate their growth by releasing a fixed amount of energy for a given halo mass, independent of the assumed efficiency of AGN feedback, which sets the normalization of the BH scaling relations. Regardless of whether BH seeds are initially placed above or below the BH scaling relations, they grow on to the same scaling relations. AGN feedback efficiently suppresses star formation in high-mass galaxies.