PI: Dr Tiago Costa Co-I Dr Ivan Almeida
AGN feedback operates across an extremely large range of spatial scales, from the event horizon of supermassive black holes (SMBHs) to clusters of galaxies, making it challenging to model within a single, physically consistent framework. A major di!culty lies in connecting the launching of relativistic jets and winds from the innermost accretion flow to their large-scale impact on the host galaxy. In this Science Highlight, we report the first stage of this programme: general relativistic magnetohydrodynamics (GRMHD) simulations of low-accretion-rate ( ˙ M → 0 . 01 , ˙ M Edd ) hot accretion flows.
We perform our simulations using H-AMR (Liska et al., 2017) on the DiRAC Data Intensive Service (Cambridge), using a total of 292 kGPU-h. H-AMR is a GPU-accelerated GRMHD code that evolves the equations of magnetised plasma in a fixed curved spacetime with metric g µω . We set the BH spin to a = 0 and a = 0 . 9375 in the two simulations. The initial setup consists of a BH of mass M located at the origin, surrounded by a magnetised torus following the equilibrium configuration of Fishbone & Moncrief (1976). The maximum density is normalised to ω 0 = 1. Our goal is to reach saturation of the magnetic flux at the event horizon and develop a magnetically arrested disc (MAD) state (Narayan et al., 2012).
The computational domain is defined in spherical polar coordinates ( r, ε, ϑ ). The radial grid is log- arithmically spaced, extending from 0 . 86 , r g to 10 4 , r g , where r g = GM/c 2 , being G the gravitational constant, c the speed of light, and M the black hole mass (a free parameter). The adopted resolution is 1152 ↑ 512 ↑ 256. Each simulation is evolved for t ↓ 3 ↑ 10 4 , t g , where t g = r g /c .
Our simulations (see Figure 1) show that the accretion flow naturally produces a multi-component outflow composed of
i) a collimated, low-density jet at approximately ε < 20 → and ε > 160 → (with ε = 0 → along the spin axis, perpendicular to the accretion-flow plane); this region is magnetically dominated and reaches velocities close to the speed of light;
ii) along the diagonals (approximately 20 → → ε → 60 → and 120 → → ε → 160 → ), gas is ejected in a non-collimated component that is denser than the jet and sub-relativistic, with characteristic velocities v w ↓ 300 km , s ↑ 1 ;
iii) an inflowing accretion flow at 60 → → ε → 120 → , which is the densest and least magnetised region, dominated by infalling gas.
We find that the wind carries the bulk of the mass outflow, whereas the jet dominates the energy output, reaching e!ciencies of ϖ jet = ˙ E jet / ˙ M BH c 2 = 2–4, while the wind has ϖ wind = 0 . 01–0 . 05. We also find strong angular stratification: mass is launched mainly at polar angles ε < 60 → , while the energy flux is highly concentrated near the poles. As expected, even at low accretion rates, RIAFs can produce energetically significant outflows capable of influencing their environment.
The radial and angular profiles of mass and energy fluxes derived from our simulations will provide a physically grounded characterisation of AGN outflows at sub-parsec scales. These GRMHD-based outflow prescriptions will serve as input for the second stage of this project, in which they will be injected into galaxy-scale simulations to assess how relativistic jets and hot winds interact with the interstellar medium.
This highlight summarises the main scientific progress enabled by DiRAC during the allocation period (January–December 2025). Analysis is ongoing, and a manuscript is in preparation.
Figure 1. ϑ -averaged maps of normalized log 10 ω ( ω max ( t = 0) = 1) with velocity field lines at three di”erent times: 0 , t g (left), 15000 , t g (centre), and 30000 , t g (right). The rotating BH is located at the centre of the system. As time progresses, the accretion flow becomes turbulent and develops outflows. Positive radial velocities indicate the formation of outflows. Credit. Ivan Almeida, DiRAC project DP370.
1
