arXiv2025
Accretion onto supermassive black holes (BHs) can launch relativistic jets that inject energy and momentum into their surroundings. Understanding how such feedback shapes large-scale accretion is key to bridging observations from galactic scales (e.g., the Bondi radius, $r_{\rm B}$) down to event horizon scales ($r_{\rm g}$), spanning 5-6 orders of magnitude. We tackle this challenge by varying the spatial scale separation across 2-4 orders of magnitude and performing some of the longest contiguous 3D general relativistic magnetohydrodynamic (GRMHD) simulations to date ($t \lesssim 4\times10^6 r_{\rm g}/c$), of Bondi-like accretion of rotating, non-relativistic gas with weak vertical magnetic fields onto a rapidly spinning BH, achieving inflow equilibrium out to $r \gtrsim 10^3 r_{\rm g}$. We find that, regardless of scale separation or ambient gas rotation, all simulations reach a magnetically arrested disk (MAD) state where the BH becomes magnetically saturated. In this state, the mass inflow rate follows a universal radial scaling: $\dot{M}_{\rm in}(r) \sim r^s$ with $s = 0.66 \pm 0.03$. The MAD state self-regulates through jets, outflows, and magnetic flux eruptions that can disrupt coherent angular momentum inflow, giving rise to a rocking accretion disk (RAD) state. This RAD state features chaotically oriented inflows, weak intermittent jets, and a steeper inflow slope of $s = 0.87 \pm 0.05$. The MAD and RAD BH accretion rates become comparable at typical scale separations, $r_{\rm B}/ r_{\rm g} \gtrsim 10^5$. Weaker RAD outflows allow large-scale inflows to resume, restoring the MAD state and enabling a recurring MAD-RAD cycle. These cycles can last tens of Bondi timescales, $t_{\rm B} \sim 0.2\,\text{Myr} \times (r_{\rm B}/10^{5} r_{\rm g})^{3/2} \times (M_{\rm BH}/10^9M_\odot)$, potentially setting the duty cycle of jetted AGN outbursts, such as in M87*.