AGN Feedback

What AGN feedback actually is

At the center of essentially every massive galaxy sits a supermassive black hole, with a mass between about 10⁶ and 10¹⁰ solar masses. Most of the time it is quiet. But when gas funnels down to its inner accretion disk, the black hole "switches on" as an active galactic nucleus (AGN) — a quasar, Seyfert, radio galaxy, or blazar depending on geometry and accretion rate. Accretion is the most efficient sustained energy source in the universe: roughly 10% of the infalling rest mass is converted to radiation, compared to just 0.7% for hydrogen fusion. A black hole growing at the Eddington limit can outshine its entire host galaxy of a hundred billion stars from a region smaller than the Solar System.

AGN feedback is what happens when a fraction of that energy couples back into the galaxy's gas instead of escaping to infinity. The black hole drives winds and jets that push, heat, and unbind the cold gas reservoir from which stars form. Because stars need cold, dense, collapsing gas, removing or reheating that gas shuts star formation down — a process called quenching. The result is a self-regulating loop: gas feeds the black hole, the black hole expels the gas, the fuel supply is cut, and accretion stalls until more gas arrives. This loop is now considered essential to galaxy evolution. It is the missing ingredient that lets theory match the observed galaxy population.

Why galaxy models can't live without it

The dark-matter halo mass function predicts far more massive halos than the number of massive, bright galaxies we actually see. In a naive model, gas in a large halo cools, sinks, and forms stars with high efficiency — and big galaxies would keep growing bluer and brighter indefinitely. Observations say the opposite: the galaxy luminosity function has a sharp exponential cutoff at the high-mass end, and the most massive galaxies are "red and dead" ellipticals dominated by old stars, with little ongoing star formation.

Stellar feedback — supernovae and stellar winds — solves the problem at the low-mass end, where shallow potential wells let modest energy blow gas out. But supernovae cannot unbind gas from the deep potential wells of the largest galaxies; there simply isn't enough energy, and the gas is too tightly bound. AGN feedback fills that gap. Its energy reservoir is enormous and concentrated, so it can act where supernovae fail. Modern cosmological simulations — Illustris/IllustrisTNG, EAGLE, SIMBA, Horizon-AGN — all require AGN feedback as a sub-grid prescription; turn it off and the simulated massive-galaxy end becomes hopelessly overgrown and too blue.

The two modes: quasar and radio

AGN feedback is not one mechanism but a duo, switched by the accretion rate relative to the Eddington limit.

Property	Quasar / radiative mode	Radio / kinetic (maintenance) mode
Accretion rate	High, near Eddington (ṁ ≳ 0.01)	Low, radiatively inefficient (ADAF)
Energy carrier	Radiation pressure → fast winds	Collimated relativistic jets
Speed	~1,000–10,000 km/s winds; UFOs up to ~0.3 c	Jets near c; bubbles rise sub-sonically
Where it acts	Galaxy-scale cold gas (ISM)	Hot halo / cluster atmosphere
Timescale	Episodic, ~10⁵–10⁷ yr bursts	Quasi-continuous over ~Gyr
Effect	Ejective: blows cold gas out, quenches fast	Preventive: reheats cooling gas, keeps it quenched
Typical host	Luminous quasars, Seyferts at high z	Brightest cluster galaxies, radio galaxies

Quasar mode dominates at high redshift (z ≈ 1–3, "cosmic noon"), when galaxies were gas-rich and black holes grew fastest. Radiation from the disk pushes on the surrounding gas through scattering and line absorption, launching wide-angle outflows. When the black hole reaches a critical mass, this wind can sweep the galaxy clear of star-forming gas in a single quasar episode lasting a few million years — a dramatic, one-time quenching event. This is the mode invoked to explain the abrupt shutdown of star formation in massive galaxies and the production of the first compact "red nuggets."

Radio mode dominates in the local, low-redshift universe inside galaxy clusters and groups. Here the central black hole accretes slowly and radiates inefficiently, but channels its power into jets that drill into the hot X-ray atmosphere and inflate buoyant bubbles. The bubbles do work on the gas, drive sound waves and turbulence, and offset radiative cooling. This is "maintenance" feedback: it doesn't expel much gas, it simply prevents the hot halo from cooling and refueling star formation — keeping already-quenched galaxies quenched for billions of years.

The energy budget that makes it work

The reason a black hole can dominate a galaxy a million times more massive is the staggering efficiency of accretion. The luminosity is L ≈ η ṁ c² with radiative efficiency η ≈ 0.1. Over its growth, a black hole of final mass M_BH liberates an energy E ≈ η M_BH c². For M_BH = 10⁸ M_⊙ this is about 2 × 10⁶¹ erg.

Compare that to the gravitational binding energy of the host bulge, roughly M_bulge σ², where σ is the stellar velocity dispersion (the spread of stellar orbital speeds). For a bulge of 10¹¹ M_⊙ and σ ≈ 200 km/s, the binding energy is only about 10⁵⁹–10⁶⁰ erg. The black hole's lifetime energy output therefore exceeds the binding energy of its entire host by a factor of tens to hundreds. Even if just a few percent of the AGN energy couples to the gas — and the coupling efficiency is the central uncertainty in the whole field — it is more than enough to unbind it. That single fact is why AGN feedback is plausible at all.

Quantity	Approximate value	Note
Accretion radiative efficiency η	~0.06–0.4	0.057 for non-spinning to ~0.42 for maximal Kerr
Lifetime energy, M_BH = 10⁸ M_⊙	~2 × 10⁶¹ erg	η M_BH c²
Bulge binding energy	~10⁵⁹–10⁶⁰ erg	M_bulge σ²
Assumed coupling efficiency	~0.5–5%	fraction of L reaching the gas
Molecular outflow rates	~10²–10³ M_⊙/yr	from CO / OH (e.g. Mrk 231)

Self-regulation and the M–σ relation

One of the most striking facts in extragalactic astronomy is the M–σ relation: black hole mass scales with the host bulge's stellar velocity dispersion as M_BH ∝ σ⁴ to σ⁵, with remarkably little scatter — about 0.3 dex. Since the black hole's gravity is utterly negligible at the scale of the whole bulge, this tight link demands a regulating mechanism, and AGN feedback supplies it.

The argument, due in its simplest form to Silk & Rees and to King, goes like this. The black hole grows by accretion, driving an energy- or momentum-conserving wind. As M_BH rises, the wind's push grows. The black hole stops growing once the wind can blow the surrounding gas out of the bulge faster than gravity can hold it — at that point the fuel is gone and accretion self-terminates. Setting the feedback strength equal to the gas's gravitational weight, a momentum-driven wind argument predicts M_BH ∝ σ⁴, while an energy-driven argument gives M_BH ∝ σ⁵. Both bracket the observed slope. The black hole, in effect, "weighs" its own galaxy and stops when it has done enough damage — a self-regulating thermostat written into the M–σ relation.

What we actually observe

• X-ray cavities. Chandra images of cluster cores (Perseus, MS 0735.6+7421, Hydra A) show giant bubbles blown into the hot gas by radio jets — dark voids ringed by bright shells. MS 0735's pair of cavities required roughly 10⁶¹ erg, the most energetic AGN outburst measured.
• Suppressed cooling flows. Cluster cores should radiate away their X-ray gas and form hundreds of stars per year. They don't — XMM-Newton spectroscopy showed the predicted cool gas is largely absent. Jet heating is the standard explanation.
• Molecular outflows. Massive cold outflows of 10²–10³ M_⊙/yr are seen in CO and OH (e.g. Markarian 231), carrying away more gas than is forming into stars — direct ejective feedback caught in the act.
• Ultra-fast outflows (UFOs). Broad, blueshifted X-ray absorption lines reveal disk winds at ~0.1–0.3 c launched from the inner accretion flow — the engine that drives the larger galaxy-scale outflows.
• Ionized winds & sound waves. Ripples in Perseus' X-ray gas are pressure waves from the AGN — a literal sound at about B-flat, fifty-seven octaves below middle C, dissipating jet energy into the halo.

Open questions

• Coupling efficiency. What fraction of AGN energy actually thermalizes in the gas, and how does it depend on geometry and gas phase? This single number controls every simulation's results and remains poorly constrained.
• Positive feedback. Jets and shocks can also compress gas and trigger bursts of star formation along their path. When does feedback quench versus ignite?
• The duty cycle. AGN flicker on and off; matching the bursty, episodic engine to the smooth, long-term quenching seen in galaxies is an active modeling challenge.
• High-redshift onset. JWST is finding luminous AGN and overmassive black holes earlier than expected, sharpening the question of when and how feedback first established the M–σ relation.