Introducing RAFIKI: Refining AGN Feedback in Kinetic Implementations

TL;DR

RAFIKI addresses how dual-mode AGN feedback shapes galaxies, black holes, and the CGM by decoupling wind and jet mass-loading in SIMBA-C and removing the X-ray channel. By varying independent efficiencies $\epsilon_w$ and $\epsilon_j$ in a 50 $h^{-1}$ Mpc volume, the study finds that quenching massive galaxies requires high jet loading, while winds primarily regulate BH growth and intermediate-mass star formation, revealing significant degeneracies in the jet parameter space. The results indicate a need for large-scale CGM heating to reproduce the high-mass end of the GSMF, with winds affecting the BH–stellar scaling and CGM properties in more nuanced ways. RAFIKI thus provides a framework to disentangle AGN feedback uncertainties and to interpret upcoming CGM observations for tighter constraints on feedback models.

Abstract

Modern cosmological simulations have now matured to the point of reproducing the evolution of realistic galaxy populations across cosmic time. These simulations rely on feedback from active galactic nuclei (AGN) to quench massive galaxies, yet the details of this process remain poorly understood. To address this issue, we introduce RAFIKI (Refining AGN Feedback In Kinetic Implementations), a novel suite of simulations built upon SIMBA-C that vary the mass loading of AGN-driven winds. Unlike the fiducial SIMBA-C simulation, RAFIKI separates the efficiencies of the two kinetic feedback modes, enabling a detailed study of their impact on galaxies, black holes, and the circumgalactic medium. We explore a range of galaxy and baryon properties in the RAFIKI runs and find that even with enhanced mass loading, the lower-velocity, quasar-type mode cannot quench massive galaxies. However, it plays a significant role in regulating black hole growth and star formation in intermediate-mass galaxies. We also uncover degeneracies in the parameter space that highlight the limited current constraints on AGN feedback. RAFIKI provides a controlled framework to disentangle these degeneracies using current and upcoming observations.

Figures (10)

• Figure 1: Galaxy stellar mass functions at redshifts 4, 2, 1, and 0 (top to bottom). Each column represents a different value for $\epsilon_j$, with jet mass loadings at 1.125, 10, 20, and 30 $\%$ from left to right. In each panel, RAFIKI runs are compared against observational data in black. At $z$=4, observational points are from song2016, at $z$=2 and $z$=1 observational points are from tomczak14 and muzzin13 and at $z=0$ observations are from bernardi2017 and driver2022. Runs with $\epsilon_j$=0.0125 were unable to match the high end of the galaxy stellar mass function and were thus not continued past z=1. High $\epsilon_w$ runs overpredict galaxies at log($M_*/M_\odot$) $\approx$ 10 and underpredict log($M_*/M_\odot$) $\approx$ 11, so these runs were also not continued past z=1. Other parameter choices were all able to successfully reproduce the observed galaxy stellar mass function across cosmic time.
• Figure 2: Star formation rate density for all RAFIKI runs. From top to bottom, panels show runs with $\epsilon_j =$ 0.0125, 0.1, 0.2, and 0.3. Black points show observational data from madau_dickinson. Higher $\epsilon_j$ values result in a peak in SFRD at earlier times and a slight underprediction of SFRD at low redshift.
• Figure 3: Stellar mass versus star formation rate for four selected RAFIKI runs at $z=1$ spanning the parameter space, colored by the black hole-stellar mass ratio. Top row shows runs with $\epsilon_j = 0.0125$, bottom row has $\epsilon_j = 0.3$, left column has $\epsilon_w = 0.0125$, and right column has $\epsilon_w = 0.45$. The black lines show the best fit to observational data from popesso_2023. The RAFIKI runs consistently underpredict the SFR, which is a trend seen across cosmological simulations.
• Figure 4: Stellar mass against sSFR for RAFIKI runs at $z=0.$ The RAFIKI runs do not include the X-ray heating mode incorporated in SIMBA, which was shown to be important for quenching the most massive galaxies. Here we see that some runs might struggle to quench systems with log($M_*/M_\odot) \approx 12$, as there are some high-mass systems with high sSFR. However, higher $\epsilon_j$ values are more effective, which sSFRs 0.5 dex lower for $\epsilon_j$=0.3 compared to $\epsilon_j$=0.1.
• Figure 5: Stellar-halo mass ratio as a function of halo mass for central galaxies at $z$=4,2,1, and 0 (top to bottom). Each column represents a different value for $\epsilon_j$, with colors correspond to different values of $\epsilon_w$. Here we see how boosting the mass loading of the wind phase prevents the stellar growth of galaxies in halos with $M_{200c} \approx 10^{12.5}$, while leading to larger stellar mass galaxies in low-mass halos.