ADAF — Advection-Dominated Accretion Flow

A black hole that should be a beacon — and isn't

At the center of the Milky Way sits Sagittarius A*, a black hole of 4.3 million solar masses, wrapped in the winds of dozens of massive young stars. It is unambiguously being fed: gas from those stellar winds is gravitationally captured at a rate of roughly 10⁻⁵ to 10⁻⁶ solar masses per year. Drop that gas onto a black hole through the standard, textbook accretion disk and convert it to light at the canonical 10% efficiency, and Sgr A* should blaze at 10³⁹–10⁴⁰ erg/s — a beacon outshining a hundred million Suns, easily the brightest thing in the Galaxy.

Instead it is whisper-quiet. The measured bolometric luminosity is around 10³⁶ erg/s, only about 10⁻⁸ to 10⁻⁹ of its Eddington limit — roughly a billion times fainter than the thin-disk calculation predicts. The gas is going somewhere. The energy is being released by gravity. So where does the light go? The answer is one of the most important ideas in modern accretion physics: at low feeding rates, the flow reorganises into an advection-dominated accretion flow, an ADAF, that simply doesn't radiate. Most of the energy is carried — advected — straight across the event horizon, where it vanishes from the observable universe.

How an ADAF works

Accretion has two clocks. One is the inflow time — how long the gas takes to spiral from a given radius down to the black hole, set by viscosity transporting angular momentum outward. The other is the cooling time — how long the gas takes to radiate away the heat that viscous friction dumps into it. Which clock wins determines everything.

In a luminous system the gas is dense. Dense plasma radiates efficiently, so the cooling time is far shorter than the inflow time. Heat escapes as fast as it is generated, the gas stays cool (10⁴–10⁷ K), and it settles into a razor-thin, optically thick disk — the classic Shakura & Sunyaev disk of 1973. Each gram converts about 6–40% of its rest mass to radiation before crossing the horizon.

Now starve the system. Lower the accretion rate and the gas becomes tenuous. A thin plasma radiates poorly, because the relevant emission processes scale with density (often as density squared). The cooling time lengthens until it exceeds the inflow time. Now the gas cannot get rid of its heat before it falls in. The temperature climbs toward the virial temperature — the point where thermal energy is comparable to gravitational binding energy — and the gas, now pressure-supported, puffs up out of the plane into a thick, roughly spherical torus. The stored heat is no longer radiated; it is advected inward and swallowed. The flow has become advection-dominated.

The crossover happens at a critical accretion rate of about ṁ_crit ≈ 0.01 in Eddington units (with order-unity dependence on the viscosity parameter α). Above it: cool, thin, bright. Below it: hot, thick, dim. Sgr A*, fed at ~10⁻⁵–10⁻⁴ of Eddington, sits deep in ADAF territory.

The hot ion torus and its two temperatures

The defining microphysics of an ADAF is that it is a two-temperature plasma. In a dense thin disk, ions and electrons collide often enough to share energy and hold a single common temperature. In the dilute ADAF plasma, Coulomb collisions between the heavy ions and the light electrons become rare — too rare to equalise their temperatures within an inflow time. The two species fall out of equilibrium.

Viscous dissipation heats the ions preferentially. With nothing to drain their energy, the ions race toward the virial temperature, reaching T_ion ≈ 10¹² K near the horizon. This is the "hot ion torus" first sketched by Rees, Begelman, Blandford & Phinney in 1982. The electrons — which are the species that actually radiates, via synchrotron emission in the magnetic field and inverse-Compton scattering — equilibrate at a far lower T_e ≈ 10⁹–10¹¹ K, because the weak Coulomb coupling delivers them only a small share of the dissipated energy.

The consequence is decisive. The energy lives in the ions, which barely radiate; the electrons, which would radiate, are comparatively starved of energy. The flow's radiative efficiency collapses. The ions carry their enormous thermal reservoir down through the horizon, and it is gone — a black hole is the one object in nature that can hide energy this completely.

Worked example: the Sgr A* energy budget

Let us make the billion-fold faintness concrete. Start from the Eddington luminosity for a 4.3 × 10⁶ M_⊙ black hole:

L_Edd ≈ 1.26 × 10^38 (M / M_⊙) erg/s
      ≈ 1.26 × 10^38 × 4.3 × 10^6
      ≈ 5.4 × 10^44 erg/s

The Eddington accretion rate, assuming 10% efficiency (L_Edd = 0.1 ṁ_Edd c²):

ṁ_Edd = L_Edd / (0.1 c²)
      = 5.4 × 10^44 / (0.1 × 9 × 10^20)
      ≈ 6 × 10^24 g/s
      ≈ 0.1 M_⊙ / yr

The gas actually captured from stellar winds is roughly 10⁻⁵ M_⊙/yr, so in Eddington units:

ṁ ≈ 10^-5 / 0.1 ≈ 10^-4 ṁ_Edd

This is already a factor of 100 below the critical rate of ~0.01 ṁ_Edd — squarely in the ADAF regime. A radiatively efficient flow at this rate would still give L ≈ 0.1 ṁ c² ≈ 10⁻⁴ × 5.4 × 10⁴⁴ ≈ 5 × 10⁴⁰ erg/s. But the ADAF efficiency is not 0.1; it scales with the rate, ε ≈ 0.1 × (ṁ / ṁ_crit). Two suppressions stack:

L_ADAF ≈ 0.1 × (ṁ / ṁ_crit) × ṁ c²    ⟹   L ∝ ṁ²
With ṁ ≈ 10^-4 and ṁ_crit ≈ 10^-2:
efficiency suppression ≈ 10^-4 / 10^-2 = 10^-2
L_ADAF ≈ 10^-2 × 5 × 10^40 ≈ 5 × 10^38 erg/s   (order of magnitude)

Add further losses — much of the captured gas never reaches the horizon at all but is blown back out in winds (the ADIOS picture), reducing the rate near the hole by another one to two orders of magnitude — and the predicted output lands near the observed ~10³⁶ erg/s. The chain of reasoning explains a deficit of roughly nine orders of magnitude relative to the naive thin-disk beacon. No exotic physics required: just a flow that is too thin to radiate and a horizon that swallows the rest.

Derivation: why advection dominates and L ∝ ṁ²

The self-similar ADAF solution of Narayan & Yi (1994) starts from the height-integrated energy equation. The viscous heating rate per unit area, Q⁺, is balanced by two sinks: radiative cooling Q⁻_rad and advective cooling Q_adv (the rate at which entropy is carried inward):

Q⁺ = Q⁻_rad + Q_adv
Define the advection fraction  f = Q_adv / Q⁺ = 1 − Q⁻_rad / Q⁺

A thin disk has f → 0 (all the heat radiates). An ADAF has f → 1 (almost all the heat is advected). The radiative efficiency is ε = (1 − f) × ε_thin. The crossover is governed by comparing timescales: advection dominates when the cooling time t_cool exceeds the inflow time t_inflow. Because t_cool ∝ 1/density ∝ 1/ṁ while t_inflow is roughly independent of ṁ, advection takes over below a critical density — equivalently below ṁ_crit ≈ α² × (a few × 0.01) in Eddington units, where α is the Shakura–Sunyaev viscosity parameter.

The quadratic luminosity scaling follows directly. The emission processes (bremsstrahlung, synchrotron, Compton) scale with the square of density, while density scales with ṁ. So Q⁻_rad ∝ ṁ², and in the advection-dominated limit where Q⁺ ∝ ṁ is dominated by advection:

L ≈ Q⁻_rad ∝ ṁ²        (radiatively inefficient branch)
ε = L / (ṁ c²) ∝ ṁ      (efficiency falls with rate)

This is the signature that distinguishes an ADAF from a thin disk observationally: drop the fuel supply by a factor of ten and the luminosity drops by a factor of a hundred, not ten. It is also why the gap between a bright quasar (ṁ ~ 0.1–1) and a quiescent nucleus (ṁ ~ 10⁻⁵) is far wider in light than in fuel.

Variants and regimes: ADAF, RIAF, ADIOS, CDAF

The clean self-similar ADAF was the first model, but global simulations revealed that real hot flows lose much of their gas before it reaches the hole. The field broadened the language accordingly. The umbrella term today is RIAF — radiatively inefficient accretion flow — which contains the original ADAF and several refinements:

• ADAF (1994–95). The original two-temperature, advection-dominated, self-similar hot flow. Assumes the full accreted mass reaches the horizon.
• ADIOS — adiabatic inflow–outflow solution (Blandford & Begelman 1999). Because the hot gas is nearly unbound (its Bernoulli parameter is positive), strong winds carry mass, energy and angular momentum away at all radii. Only a small fraction of the Bondi-captured gas actually crosses the horizon — directly observed for Sgr A* by Chandra's resolution of the captured-versus-accreted rate.
• CDAF — convection-dominated accretion flow. The flow is convectively unstable; turbulent convection transports energy and angular momentum outward, further suppressing the inflow and the luminosity.
• LHAF / luminous hot accretion flow. Near ṁ_crit the flow is hot but still radiates appreciably, bridging the ADAF and thin-disk regimes; relevant to the brighter low-luminosity AGN.

All share the defining property: at low accretion rates the central object radiates far below the canonical 10% of rest-mass energy. They differ in how the energy escapes the radiative channel — swallowed, blown out, or convected.

Thin disk versus ADAF at a glance

Property	Thin disk (SSD)	ADAF / RIAF
Accretion rate (Eddington units)	~0.01 – 1	below ~0.01
Geometry	razor-thin, H/R ≪ 1	thick torus, H/R ~ 1
Optical depth	optically thick	optically thin
Gas temperature	10⁴ – 10⁷ K	ions ~10¹² K, electrons ~10⁹–10¹¹ K
Temperature structure	single-temperature	two-temperature (hot ion torus)
Radiative efficiency	~0.06 – 0.4 (≈0.1 typical)	≪ 0.1, scales as ε ∝ ṁ
Energy fate	radiated as light	advected into hole / blown out in winds
Luminosity scaling	L ∝ ṁ	L ∝ ṁ² (super-linear)
Prototype	luminous quasar, soft-state binary	Sgr A*, low-luminosity AGN, hard-state binary

Observational status and applications

• Sagittarius A*. The flagship. Decades of multiwavelength monitoring — radio, sub-mm, near-infrared flares, and the faint Chandra X-ray glow from the Bondi radius — match a RIAF/ADAF spectral model far better than a thin disk. The Event Horizon Telescope 2022 image of Sgr A* probes exactly the inner hot flow this model describes.
• M87*. The first black hole ever imaged (EHT 2019) is also a low-luminosity, RIAF-fed nucleus. Its powerful relativistic jet is thought to be launched from the magnetised hot flow threading the horizon.
• Low-luminosity AGN. The vast majority of nearby galactic nuclei — LINERs and the like — are not quasars but faint, hot-flow accretors. ADAF/RIAF models explain their spectra and their characteristic absence of the "big blue bump" that a thin disk would produce.
• X-ray binary state transitions. Stellar-mass black holes in binaries flip between a bright "soft" thin-disk state and a faint, hard, jet-dominated state on timescales of days to months as their accretion rate crosses ṁ_crit. The hard state is the textbook small-scale ADAF, watched in real time.
• Jet connection. Hot, thick, magnetically arrested flows are the natural launch pad for the Blandford–Znajek mechanism. The radiatively inefficient regime and powerful jets go hand in hand — energy that does not become light can become a jet instead.

Common pitfalls and misconceptions

• "Faint means barely accreting." Not necessarily. Sgr A* captures a real, measurable mass flux. The faintness is about inefficiency, not starvation alone — the luminosity is suppressed far below the already-low fuel rate by the ε ∝ ṁ scaling and by winds.
• "The energy is just lost." It is not violating conservation. The advected thermal and kinetic energy is carried across the event horizon and adds to the black hole's mass-energy. From outside, it is invisible — which is the whole point.
• "ADAFs are cold because they're faint." The opposite. An ADAF is dramatically hotter than a thin disk (ions at ~10¹² K). It is faint because it is too tenuous to radiate that heat, not because it is cool.
• "All low-rate accretion is an ADAF." The two-temperature ADAF is one solution among the RIAF family. Real flows include winds (ADIOS) and convection (CDAF); pure ADAF self-similar solutions are an idealisation that simulations refine.
• "The 0.01 Eddington threshold is exact." The critical rate ṁ_crit depends on the viscosity parameter α (roughly as α²) and on uncertain electron-heating physics. "About 1% of Eddington" is the order-of-magnitude crossover, not a sharp line.
• "Electrons set the temperature." The flow's pressure and thickness are set by the hot ions. The electrons are the radiators but carry little of the energy; conflating the two species is exactly the error that the two-temperature insight corrects.