Numerical solutions for black hole feeding and feedback in cosmological simulations with OpenGADGET3

We investigate how different numerical implementations of black hole (BH) feeding and feedback in the OpenGADGET3 code affect the co-evolution of BHs and their host environment. For this purpose, we simulate the same galaxy group having a virial mass of ∼ 1 0 13 M ⊙ at z = 0 , multiple times at the same resolution but varying one aspect of the numerical implementation of BH feeding and feedback at the time. Specifically, we test (a) the impact of stochastic swallowing of gas particles, (b) a prescription for the evaporation of cold gas clouds, (c) alternative kernel-weighting schemes to distribute BH energy feedback, and (d) the interaction between BH feedback and supernova-driven wind particles. We find that neglecting stochastic swallowing leads to higher BH accretion rates and prolonged accretion phases, particularly in the presence of cold gas. Attempts to suppress this cold accretion episodes by evaporating gas clouds enhance the star formation rate of the central galaxy by up to two orders of magnitude at z ≲ 1 . Adopting a top-hat kernel in radio mode yields only minor differences, though occasionally it accelerates gas depletion in the BH surroundings. Finally, we show that the treatment of wind particles is critical: when decoupled from hydrodynamics, they can store AGN feedback energy and release it upon recoupling, strongly altering the gas dynamics. In general, the results of our analysis highlight the crucial importance that technical details of the implementation of the same AGN feedback model can lead to substantially different predictions of simulations on the resulting co-evolution of galaxies, diffuse hot baryons and population of massive BH.