Advection-Dominated Accretion Flows

Accretion is the most efficient power source in the universe short of matter–antimatter annihilation, and the textbook picture — Shakura and Sunyaev's cold, optically thick, geometrically thin disk — converts $6$ – $40\%$ of the infalling rest mass into radiation. Yet many accreting black holes are astonishingly dim. Sgr A, the four-million-solar-mass hole at the Galactic Center, captures gas from nearby stellar winds and ought to blaze like a quasar; instead it shines at $\sim 10^{-8}$ – $10^{-9}$ of its Eddington ceiling. The advection-dominated accretion flow (ADAF) resolves that paradox. When the accretion rate falls low enough the gas becomes so rarefied that it cannot radiate the viscously released energy within an inflow time: two-body cooling scales as $n^2$ while advective heat transport scales only as $n$ , so at low density radiation loses the race, and the dissipated energy is stored as thermal energy — mostly in the ions — and carried straight across the event horizon before it can become light. The result is a qualitatively different solution branch: hot, geometrically thick ( $H/R\sim 1$ ), optically thin, and two-temperature, with ions near the virial temperature ( $\sim 10^{12}$ K) and electrons much cooler. Because the gas is nearly virial, the flow is quasi-spherical, transonic, sub-Keplerian, and only weakly bound — a positive Bernoulli parameter that invites winds and jets. Developed by Narayan and Yi in 1994–95 on the two-temperature groundwork of Shapiro, Lightman and Eardley and of Ichimaru, the ADAF is the founding member of the radiatively inefficient accretion flow (RIAF) family, and it underlies the quiescence of Sgr A, the low/hard state of X-ray binaries, the low-luminosity AGN population, and the kinetic "radio-mode" feedback that shapes the most massive galaxies.

Key takeaways

An advection-dominated accretion flow (ADAF) is a hot, geometrically thick, optically thin solution in which the gas fails to radiate the viscously dissipated energy; instead it is stored as heat and advected across the horizon.
ADAFs appear at low rates, $\dot M\lesssim\dot M_{\rm crit}\sim\alpha^2\dot M_{\rm Edd}$ , where the electron–ion Coulomb coupling is too weak to move heat from the ions (which get the viscous energy) to the electrons (which radiate) — a two-temperature plasma with $T_i\sim 10^{12}$ K and $T_e\sim 10^{9}$ – $10^{11}$ K.
Because the energy is trapped, the efficiency scales as $\eta\propto\dot M$ , so $L\propto\dot M^2$ : ADAFs are extraordinarily underluminous — the paradigm for Sgr A* and low-luminosity AGN.
The flow is near-virial and pressure-supported ( $c_s\sim v_K$ , $H/R\sim 1$ ), transonic ( $|v_r|\sim\alpha v_K$ ), sub-Keplerian, and has a positive Bernoulli parameter, making it prone to winds (ADIOS) and jets.
ADAFs complement the cold, efficient Shakura–Sunyaev thin disk; the state switches branches as $\dot m$ crosses $\dot m_{\rm crit}\sim\alpha^2$ .
Key idea: ADAFs exist because radiative cooling ( $\propto n^2$ ) becomes slower than advective heat transport ( $\propto n$ ) at low density; the viscously dissipated energy is stored as ion thermal energy and advected across the horizon rather than radiated, giving a hot, thick, two-temperature flow with efficiency $\eta\propto\dot M$ instead of the constant $\eta$ of a thin disk.

Where this lives today

ADAF physics is now the default language for every accreting black hole not shining near its Eddington limit — which is most of them. The Event Horizon Telescope images of M87 (2019) and Sgr A (2022) are, physically, pictures of hot RIAF/ADAF plasma: the emission ring is synchrotron light from the two-temperature electrons in the very geometry Narayan and Yi predicted, and GRMHD models fit to the polarized images point to strongly magnetized, magnetically arrested (MAD) hot flows. In the low/hard state of X-ray binaries, spectral-timing analysis places a truncated thin disk feeding an inner ADAF/corona whose Comptonized power law and compact radio jet track the flow's weak binding. On the largest scales, because RIAFs deposit most of their energy kinetically, they power the "radio-mode" feedback that offsets cooling flows in massive ellipticals and clusters — a linchpin of galaxy-formation models and the $M$ – $\sigma$ relation. Sharp open questions remain — how much mass is lost to winds, how electrons are heated relative to ions in a collisionless plasma, how ADAFs reconnect to thin disks across the transition — now attacked with horizon-scale imaging, X-ray polarimetry (IXPE), and kinetic simulations.

The physical picture

Picture two clocks racing. The first is the cooling clock: how long the gas takes to radiate its heat, which speeds up as the gas gets denser ( $t_{\rm cool}\propto 1/n$ ). The second is the inflow clock: how long the gas takes to spiral in and cross the horizon, set by dynamics and nearly independent of density. A thin disk is dense, so cooling wins easily and the disk stays cold, flat, and efficient. Dilute the gas — lower the accretion rate — and the cooling clock slows until it loses the race; the heat then has no time to escape and simply rides inward with the flow, puffing the disk into a thick, nearly spherical, hot torus. The first figure sets the two branches — same gravity, radically different geometry and luminosity — side by side.

Figure

Two accretion branches selected by the rate Ṁ. At high rate a cold thin disk (H/R ≪ 1) radiates efficiently with η ≈ 0.06–0.4; at low rate a hot thick ADAF (H/R ∼ 1, T_i ∼ 10¹² K) advects its heat inward, giving L ∝ Ṁ² and launching winds because the gas is barely bound

The energy also goes to the wrong particles. Turbulent stresses heat the massive ions, but the light electrons do the radiating; in a rarefied ADAF the electron–ion energy-exchange time far exceeds the inflow time, so the species decouple, and the ions run away to nearly the virial temperature (a few $\times 10^{12}$ K) while the radiation-bled electrons settle an order of magnitude or two cooler and carry their heat down the drain unradiated. Treat gravitational energy as a budget: a thin disk spends almost all of it on light, whereas an ADAF pockets most of it as internal energy and swallows it — luminosity quadratic in accretion rate, which is what makes underfed black holes vanish, as the second figure plots directly.

Figure

The efficiency fork. Above ṁ_crit ≈ α² ≈ 0.01 the thin disk gives L/L_Edd = ṁ (slope 1); below it the ADAF branch gives L/L_Edd ∝ ṁ² (slope 2). Sgr A* at ṁ ≈ 10⁻⁵ lands at L/L_Edd ≈ 10⁻⁸ — about a thousand times fainter than a thin disk at the same rate would be

Finally, because ADAF gas is barely bound — near-virial, with positive Bernoulli energy — it sits like water at the lip of a basin: a small push sends it outward as a wind or up magnetic field lines as a jet. Radiative inefficiency and outflow/jet activity are two faces of the same weakly bound geometry, which is why the dimmest accreting black holes so often carry the most conspicuous jets.

The core equations

The whole edifice rests on one inequality: advection dominates when the gas cannot cool within an inflow time. It compares the cooling time — internal energy over the volumetric cooling rate — with the accretion time, radius over radial speed.

t_{\rm cool}\gtrsim t_{\rm acc}\quad\Longleftrightarrow\quad \text{advection dominates} \qquad (1)

where: $t_{\rm cool}$ — radiative cooling time, $\sim u/\Lambda$ with $u\sim nk_BT$ the internal energy density and $\Lambda\propto n^2$ the two-body cooling rate; $t_{\rm acc}$ — accretion (inflow) time, $\sim R/v_r$ . Since $t_{\rm cool}\propto 1/n$ while $t_{\rm acc}$ barely depends on $n$ , lowering $\dot M$ tips the balance to advection. Vertically integrated, the energy released by viscosity must go somewhere: into radiation, or into the enthalpy carried inward. The advected fraction $f$ labels where a flow sits between the two branches.

Q^{+}_{\rm visc} = Q^{-}_{\rm rad} + Q_{\rm adv},\qquad Q_{\rm adv}=f\,Q^{+}_{\rm visc} \qquad (2)

where: $Q^{+}_{\rm visc}$ — viscous heating rate per unit area; $Q^{-}_{\rm rad}$ — radiative cooling rate per unit area; $Q_{\rm adv}=\Sigma v_r T\,\mathrm{d}s/\mathrm{d}R$ — advective (entropy) transport per unit area; $f$ — advection parameter, $f\to 1$ in the pure ADAF limit and $f\to 0$ for a thin disk. The density is proportional to $\dot M$ , but the cooling goes as density squared. The defining energetic signature follows: the efficiency is no longer a constant of the spacetime but proportional to the accretion rate.

\eta \;\simeq\; \eta_0\,\frac{\dot M}{\dot M_{\rm crit}}\quad\Longrightarrow\quad L=\eta\dot M c^2 \propto \dot M^{2} \qquad (3)

where: $\eta$ — radiative efficiency; $\eta_0\approx 0.057$ – $0.42$ — the thin-disk efficiency (spin-dependent); $\dot M_{\rm crit}\sim\alpha^2\dot M_{\rm Edd}$ — the rate at which cooling and advection balance. Below $\dot M_{\rm crit}$ , halving $\dot M$ quarters $L$ . The trapped heat is nearly virial, so gas pressure supports the flow vertically as well as radially. That fixes the geometry: the sound speed rivals the Keplerian speed and the disk is as thick as it is wide.

\frac{k_B T_i}{\mu_i m_p}+\frac{k_B T_e}{\mu_e m_p}\;\sim\; v_K^2 \;\Rightarrow\; c_s\sim v_K,\quad \frac{H}{R}\sim 1 \qquad (4)

where: $T_i,T_e$ — ion and electron temperatures; $\mu_i,\mu_e$ — mean molecular weights per ion and per electron; $v_K=\sqrt{GM/R}$ — Keplerian speed; $c_s$ — isothermal sound speed; $H$ — vertical scale height. Thick geometry ( $H/R\sim1$ ) is a direct consequence of near-virial temperature. Two velocities distinguish an ADAF from a thin disk at a glance. The radial inflow is transonic — a large fraction of $v_K$ , not a thin disk's crawling drift — and the rotation is measurably sub-Keplerian because pressure gradients help hold the gas up.

v_r \simeq -\alpha\,\frac{c_s^2}{v_K}\sim -\alpha\,v_K,\qquad \Omega^2 \simeq (1-f\,g)\,\Omega_K^2 \qquad (5)

where: $\alpha$ — Shakura–Sunyaev viscosity parameter; $\Omega,\Omega_K$ — actual and Keplerian angular velocities; $g$ — an $O(1)$ factor encoding pressure support. With $c_s\sim v_K$ the inflow reaches $|v_r|\sim\alpha v_K$ , orders of magnitude above a thin disk. The ion temperature is set by the local virial condition — thermal energy comparable to binding energy. Evaluated at the Schwarzschild radius it is enormous, rising inward as $1/R$ .

T_i \simeq \frac{1}{2}\,\frac{G M m_p}{k_B R}\;\approx\; 2.7\times10^{12}\,\Big(\frac{R_s}{R}\Big)\ \mathrm{K} \qquad (6)

where: $R_s=2GM/c^2$ — Schwarzschild radius; $m_p$ — proton mass. At $R=R_s$ , $\tfrac12 GM m_p/(k_B R_s)=m_p c^2/(4k_B)=2.7\times10^{12}$ K — independent of black-hole mass, since both $T_i$ and $R_s$ scale with $M$ . The two-temperature state survives only because the ions cannot hand their heat to the electrons. At ADAF densities the Coulomb energy-exchange time between the species dwarfs the inflow time.

t_{ie} \sim \frac{m_p m_e c^3}{8\sqrt{2\pi}\,n_e\,e^4\,\ln\Lambda}\Big(\frac{k_B T_e}{m_e c^2}\Big)^{3/2}\;\gg\; t_{\rm acc} \qquad (7)

where: $t_{ie}$ — ion–electron Coulomb energy-exchange time; $n_e$ — electron number density; $\ln\Lambda\sim 20$ — the Coulomb logarithm; $e$ — electron charge. Low $n_e$ (hence low $\dot M$ ) is what makes $t_{ie}\gg t_{\rm acc}$ and decouples the temperatures. The trapped heat keeps the gas nearly unbound. The Bernoulli function — the conserved sum of kinetic, thermal, and gravitational specific energy along a streamline — comes out positive over much of the flow.

\mathrm{Be}\equiv \frac{1}{2}v^2 + \frac{\gamma}{\gamma-1}\frac{P}{\rho} - \frac{GM}{R} \;>\;0 \qquad (8)

where: $\mathrm{Be}$ — Bernoulli function; $\gamma$ — adiabatic index; $P/\rho$ — the specific-enthalpy contribution. $\mathrm{Be}>0$ means the gas can reach infinity with energy to spare — the seed of ADAF winds. That marginal binding is realized as mass loss. In the inflow–outflow generalization (ADIOS) a fraction of the gas is driven off at every radius, so the accretion rate is not constant but decreases inward.

\dot M(R) \propto R^{s},\quad 0<s\lesssim 1\qquad(\text{ADIOS}) \qquad (9)

where: $s$ — outflow index; $R$ — cylindrical radius. An observed $s>0$ (as inferred for Sgr A* and M87) is direct evidence that the Bondi-captured gas does not all reach the horizon.

Derivation ASelf-similar scalings of a hot advection-dominated flow

The goal is to extract the characteristic radial run of density, temperature, and velocity in an ADAF without solving the full global problem, by assuming the height-integrated conservation laws admit power-law self-similar solutions far from both boundaries. We impose mass, radial-momentum, and angular-momentum balance for a steady axisymmetric $\alpha$ -flow and demand that every term share a common radial scaling; requiring gravity, pressure, and inertia to fall off at the same rate fixes the exponents uniquely and reproduces the Narayan–Yi solution.

Step 1. Start from mass conservation, $\dot M=4\pi R H\rho\,|v_r|=\text{const}$ , with thickness tied to temperature by hydrostatic support, $H\sim (c_s/v_K)R$ . Why: the accretion rate is a global constant, so any radial run of $\rho$ must be compensated by $H$ and $v_r$ .

Step 2. Demand near-virial temperature, $c_s^2\propto R^{-1}$ — the only scaling for which the pressure, gravitational, and centrifugal terms in the radial momentum equation $v_r\,\mathrm{d}v_r/\mathrm{d}R-\Omega^2 R=-\rho^{-1}\mathrm{d}P/\mathrm{d}R-GM/R^2$ all vary as $R^{-2}$ . This fixes $T\propto R^{-1}$ and, with $v_K\propto R^{-1/2}$ , gives $c_s\propto R^{-1/2}$ and $H\propto R$ . Why: self-similarity requires no single term to dominate at large or small $R$ , and the virial profile enforces that.

Step 3. Read the radial velocity off the angular-momentum equation with an $\alpha$ -stress $\tau=\alpha\rho c_s^2$ : $|v_r|\sim\alpha\,c_s^2/v_K\propto R^{-1/2}$ . Since $c_s\sim v_K$ this is transonic, $|v_r|\sim\alpha v_K$ . Why: the viscous torque lets gas shed angular momentum, so the same $\alpha$ that heats the flow sets how fast it drains.

Step 4. Feed this back into mass conservation: $R\cdot R\cdot\rho\cdot R^{-1/2}=\text{const}$ forces $\rho\propto R^{-3/2}$ , and $\Omega\propto v_K/R\propto R^{-3/2}$ . Why: with $H$ and $v_r$ pinned, constant $\dot M$ leaves only the density profile free.

Step 5. The rotation is sub-Keplerian, $\Omega^2=\Omega_K^2[1-(c_s/v_K)^2 g(\alpha,f)]$ : pressure support subtracts an $O(1)$ fraction of gravity, an ADAF fingerprint absent in cold disks. Why: in a hot thick flow the radial pressure gradient is not negligible, so rotation need not balance gravity alone.

\boxed{\;\rho\propto R^{-3/2},\quad c_s^2\propto R^{-1},\quad |v_r|\propto R^{-1/2},\quad \Omega\propto R^{-3/2}\;}

Go deeper · Derivation B · 1 minRadiative inefficiency and the L ∝ Ṁ² scaling

Derivation BRadiative inefficiency and the L ∝ Ṁ² scaling

The goal is to show, from the density scaling alone, why an ADAF's luminosity falls off as the square of the accretion rate and its efficiency vanishes linearly at low rates — the single most important observational consequence of the whole picture. We express the density in terms of $\dot M$ , insert it into a two-body cooling rate, integrate over the fixed-geometry flow, and compare the luminosity with the rest-mass power $\dot M c^2$ to read off the efficiency.

Step 1. At fixed geometry and dynamics, $\dot M\sim 4\pi R H\rho\,|v_r|$ makes density linear in accretion rate, $\rho\propto\dot M$ . Why: the flow's shape and velocity field are set by gravity and $\alpha$ , so turning up $\dot M$ just scales the density in proportion.

Step 2. The dominant cooling channels — bremsstrahlung and the synchrotron self-Compton riding on it — are two-body processes, $Q^{-}_{\rm rad}\propto n_e^2\,\Lambda(T_e)\propto\rho^2$ . Why: emission needs two particles to meet, so the rate per volume goes as density squared — the crux of the argument.

Step 3. Integrate over the roughly fixed volume, $L\sim\int Q^{-}_{\rm rad}\,\mathrm{d}V\propto\rho^2\propto\dot M^2$ . Why: with geometry frozen, the integral carries the $\rho^2$ scaling straight through.

Step 4. Divide by the rest-mass power: $\eta=L/(\dot M c^2)\propto\dot M^2/\dot M=\dot M$ . Contrast the thin disk, where $\eta\approx\eta_0$ is fixed by the innermost orbit; here falling $\dot M$ drives $\eta\to0$ . Why: quiescent holes are dim twice over — little fuel, and each gram burned ever less efficiently.

\boxed{\;L\propto\dot M^2\quad\Longrightarrow\quad \eta=\frac{L}{\dot M c^2}\propto\dot M\;}

Go deeper · Derivation C · 1 minThe critical accretion rate that ends the ADAF branch

Derivation CThe critical accretion rate that ends the ADAF branch

The goal is to derive the ceiling $\dot m_{\rm crit}\sim\alpha^2$ above which the hot two-temperature solution cannot exist, by finding where radiative cooling first catches up with advective transport. We write both cooling and advection as functions of $\dot M$ , exploit the fact that cooling grows faster (density squared) than advection (density), and solve for the crossover, tracking how the viscosity $\alpha$ enters the viscous heating and inflow speed.

Step 1. Set the boundary where radiation just competes with advection, $Q^{-}_{\rm rad}\sim Q_{\rm adv}$ (i.e. $f\sim\tfrac12$ ). Why: for $Q^{-}_{\rm rad}\ll Q_{\rm adv}$ the flow is a clean ADAF and for $Q^{-}_{\rm rad}\gg Q_{\rm adv}$ it overcools; the crossover is where they match.

Step 2. From the previous derivation $Q^{-}_{\rm rad}\propto\dot M^2$ and $Q_{\rm adv}\propto\dot M$ , so their ratio grows linearly with rate. Why: quadratic cooling always beats linear advection at high enough $\dot M$ — the ADAF cannot survive arbitrary rates.

Step 3. Fold in the inflow: larger $\alpha$ shortens $t_{\rm acc}$ and favours advection, pushing the crossover up; bookkeeping the viscous heating and transonic $|v_r|\sim\alpha v_K$ gives $\dot M_{\rm crit}\propto\alpha^2\dot M_{\rm Edd}$ . Why: a stickier disk drains faster, so heat is advected before it can radiate even at higher rates — hence $\alpha^2$ .

Step 4. For $\dot m>\dot m_{\rm crit}$ the hot solution overcools and collapses to a Shakura–Sunyaev disk; with $\alpha\sim 0.1$ – $0.3$ , $\dot m_{\rm crit}\sim 10^{-2}$ – $10^{-1}$ . Why: this places the transition at a few percent of Eddington — exactly where X-ray binaries switch spectral states.

\boxed{\;\dot m_{\rm crit}\equiv\frac{\dot M_{\rm crit}}{\dot M_{\rm Edd}}\sim\alpha^2\approx 0.01\text{–}0.1\;}

Limiting cases

The accretion mode is selected by the dimensionless rate $\dot m=\dot M/\dot M_{\rm Edd}$ and the viscosity $\alpha$ , spanning a bright thin disk to a nearly luminosity-free free-fall.

Case	Result	Meaning
High rate, thin disk	$\dot m\gtrsim\alpha^2\ (\text{but }\dot m\lesssim 1)$	Dense, optically thick, cold Shakura–Sunyaev disk; radiatively efficient with nearly constant $\eta$ . The hot branch does not exist above $\dot m_{\rm crit}\sim\alpha^2$ .
Low rate, ADAF	$\dot m\lesssim\dot m_{\rm crit}\sim\alpha^2\approx 0.01\text{–}0.1$	Optically thin, two-temperature, geometrically thick hot flow; $\eta\propto\dot m$ so $L\propto\dot M^2$ . The regime of Sgr A*, quiescent X-ray binaries, and LLAGN.
Very low rate, radial limit	$v_r\to v_{\rm ff},\ L\to 0$	Approaches a quasi-spherical, adiabatic, near free-fall (Bondi-like) inflow; essentially all dissipated energy is advected — the strict advection-dominated ideal.
Temperature profile	$T_i(R)\simeq\tfrac12\,GM m_p/(k_B R)\sim 2.7\times10^{12}\,(R/R_s)^{-1}\,\mathrm{K}$	Ions near virial; pressure-supported with $c_s\sim v_K$ and $H/R\sim 1$ — the geometric fingerprint of an ADAF.

Common trap“Advection-dominated” means the flow does not radiate at all.

It does not. ADAFs radiate — through synchrotron, bremsstrahlung, and inverse Compton — and their spectra (the submillimeter bump of Sgr A, the hard X-ray power law of the low/hard state) are how we identify them. What is true is that the emitted power is a small fraction* of the dissipated power, so $L$ falls as $\dot M^2$ rather than $\dot M$ . A related trap: an ADAF is not just a "hot thin disk." Its large scale height, near-virial two-temperature plasma, transonic sub-Keplerian flow, and positive Bernoulli parameter make it a genuinely different, quasi-spherical branch that exists only below $\dot m_{\rm crit}$ ; push the rate above that ceiling and the hot solution overcools and collapses back onto the cold Shakura–Sunyaev disk.

Worked examples

1 · Why Sgr A* is a thousand times too faint (fully worked)

The Galactic-Center black hole has $M=4\times10^6\,M_\odot$ and captures gas at roughly the Bondi rate $\dot M_{\rm B}\sim 10^{-6}\,M_\odot\,\mathrm{yr}^{-1}$ . Take $\alpha=0.1$ and a fiducial efficiency $\eta_0=0.1$ . Predict its luminosity as (a) a thin disk and (b) an ADAF, and compare with the observed $L_{\rm bol}\sim 10^{36}\,\mathrm{erg\,s^{-1}}$ .

Step 1. Set the scales. $L_{\rm Edd}=1.26\times10^{38}(M/M_\odot)=\mathbf{5.0\times10^{44}\ erg\,s^{-1}}$ and $\dot M_{\rm Edd}=L_{\rm Edd}/(\eta_0 c^2)=5.6\times10^{24}\,\mathrm{g\,s^{-1}}=\mathbf{0.089\ M_\odot\,yr^{-1}}$ , so the captured gas is $\dot m=\dot M_{\rm B}/\dot M_{\rm Edd}=\mathbf{1.1\times10^{-5}}$ .

Step 2. Thin-disk prediction: $L=\eta_0\dot M c^2$ gives $L/L_{\rm Edd}=\dot m=1.1\times10^{-5}$ , i.e. $L_{\rm thin}=\mathbf{5.7\times10^{39}\ erg\,s^{-1}}$ — quasar-like, what a naive "black hole eats gas" estimate predicts.

Step 3. ADAF prediction: below the ceiling $\eta=\eta_0(\dot m/\dot m_{\rm crit})$ with $\dot m_{\rm crit}=\alpha^2=0.01$ , so $\eta=\mathbf{1.1\times10^{-4}}$ — a thousandfold below $\eta_0$ . Then $L/L_{\rm Edd}=(\eta/\eta_0)\dot m=1.3\times10^{-8}$ , i.e. $L_{\rm ADAF}=\mathbf{6.4\times10^{36}\ erg\,s^{-1}}$ .

Step 4. Compare. The ADAF lands within a factor of a few of the observed $L_{\rm bol}\sim10^{36}\,\mathrm{erg\,s^{-1}}$ ( $L/L_{\rm Edd}\sim2\times10^{-9}$ ), while the thin disk overpredicts by $\sim10^{3.5}$ ; the residual is closed by the inward decrease of $\dot M$ (ADIOS winds, $\dot M(R)\propto R^s$ ). The thin-disk model is $\sim 10^3$ times too bright; the ADAF, with $L\propto\dot M^2$ , is what makes Sgr A* invisible as a quasar.

2 · Near-virial temperature forces a thick disk (fully worked)

For the same black hole, evaluate the ion temperature at $R=10\,R_s$ and show the flow must be geometrically thick.

Step 1. Ion temperature. From eq. (6), $T_i\simeq 2.7\times10^{12}(R_s/R)=\mathbf{2.7\times10^{11}\ K}$ ; the radiation-drained electrons sit at $T_e\sim10^{10}$ – $10^{11}$ K — a genuine two-temperature plasma with $T_i/T_e\sim 3$ – $30$ . (Here $R_s=1.2\times10^{12}\,\mathrm{cm}\approx0.08\,\mathrm{AU}$ .)

Step 2. Speeds. At $R=10R_s$ , $v_K=c\sqrt{R_s/2R}=c/\sqrt{20}=\mathbf{6.7\times10^9\ cm\,s^{-1}}$ , while the ion thermal speed $v_{\rm th,i}=\sqrt{k_B T_i/m_p}=\mathbf{4.7\times10^9\ cm\,s^{-1}}$ , so $c_s\sim v_{\rm th,i}$ is comparable to $v_K$ .

Step 3. Geometry. Then $c_s/v_K\approx0.7$ , so hydrostatic support gives $H/R\sim c_s/v_K\approx\mathbf{0.7\sim 1}$ — a thick, quasi-spherical flow, exactly as eq. (4) demands, versus $c_s\ll v_K\Rightarrow H/R\ll1$ for a thin disk. Because the ions are within a factor of two of virial, the ADAF cannot help but be thick — temperature and geometry are one statement. (Equivalently the X-ray-binary state transition occurs where $L/L_{\rm Edd}\approx\dot m_{\rm crit}\sim\alpha^2$ — a few percent of Eddington, as observed.)

Physics of the portrait

ADAF theory is inseparable from one physicist who assembled the scattered two-temperature and ion-torus ideas of the 1970s–80s into a predictive, self-similar global solution — and then spent the next three decades testing it against the faintest black holes in the sky.

Origins · 1 minRamesh Narayan and how this was discovered

Ramesh Narayan

born 1950 · hot accretion flows and black-hole imaging

Ramesh Narayan, an Indian-American theoretical astrophysicist at Harvard, is the architect of the modern theory of radiatively inefficient accretion. The pieces existed before him — Shapiro, Lightman and Eardley's two-temperature hot plasma (1976, the unstable SLE solution), Ichimaru's advection idea (1977), and the "ion-supported torus" of Rees, Begelman, Blandford and Phinney (1982) — but what was missing was a self-consistent, predictive global solution. With his student Insu Yi, Narayan found the self-similar advection-dominated solution in 1994 ( $\rho\propto R^{-3/2}$ , $T\propto R^{-1}$ , transonic sub-Keplerian inflow), and in 1995 built the global two-temperature structure and its spectra. Applied at once to Sgr A, it explained why a four-million-solar-mass black hole could shine at $10^{-8}$ of Eddington — the paradox that had frustrated the field for two decades. He and collaborators then mapped the framework onto quiescent X-ray binaries, low-luminosity AGN, and low/hard-state transitions, and folded in outflows and convection. As a leader of the Event Horizon Telescope, whose 2019 image of M87 and 2022 image of Sgr A are of the very RIAF systems his theory describes, he closed the loop from a 1994 self-similar solution to a direct photograph of a hot accretion flow.

History

1976

— Shapiro, Lightman and Eardley propose a two-temperature hot accretion plasma (the SLE solution)

1977

— Ichimaru describes optically thin, advection-dominated hot accretion

1982

— Rees, Begelman, Blandford and Phinney propose the ion-supported torus

1994

— Narayan and Yi derive the self-similar advection-dominated solution

1995

— Narayan and Yi build the global two-temperature structure and apply it to Sgr A*

Going further

Texts. Frank, King & Raine, Accretion Power in Astrophysics, 3rd ed., Ch. 11 · Carroll & Ostlie, An Introduction to Modern Astrophysics, accretion chapter · Kato, Fukue & Mineshige, Black-Hole Accretion Disks, 2nd ed. · Shapiro & Teukolsky, Black Holes, White Dwarfs and Neutron Stars · Binney & Tremaine, Galactic Dynamics; Mo, van den Bosch & White, Galaxy Formation and Evolution (feedback context). The key results to keep in mind are: the self-similar exponents $\rho\propto R^{-3/2}$ , $c_s^2\propto R^{-1}$ , $|v_r|\propto R^{-1/2}$ ; the efficiency scaling $L\propto\dot M^2$ , $\eta\propto\dot M$ ; the critical rate $\dot m_{\rm crit}\sim\alpha^2$ ; the ion temperature $T_i\simeq\tfrac12 GM m_p/(k_B R)=2.7\times10^{12}(R_s/R)$ K, independent of $M$ ; and the Sgr A* numbers $\dot m\approx1.1\times10^{-5}$ , thin-disk $L/L_{\rm Edd}\approx10^{-5}$ , ADAF $\approx10^{-8}$ . Note that the $6\times10^{12}$ prefactor sometimes quoted for eq. (6) is inconsistent with the explicit $\tfrac12$ and has been corrected to $2.7\times10^{12}$ .

Prerequisites. Thin Accretion Disks and the Shakura–Sunyaev Model · Eddington Luminosity and Accretion Limits · Magnetorotational Instability and Accretion Turbulence. Leads to. Relativistic Jets and Launching Mechanisms · Supermassive Black Holes · AGN Feedback · Active Galactic Nuclei: The Unified Model.

Ask the tutor: Ask me to walk through why $L\propto\dot M^2$ makes Sgr A* invisible as a quasar, or to contrast the ADAF and thin-disk branches term by term.

Is an ADAF the same thing as a RIAF?

Almost. "Radiatively inefficient accretion flow" (RIAF) is the umbrella term for any hot, optically thin, geometrically thick flow whose efficiency is far below the thin-disk value. The ADAF is the founding and simplest member — the self-similar solution in which a fraction $f\to1$ of the dissipated energy is advected inward. The RIAF family also includes the physics the pure ADAF idealizes away: the ADIOS solution with mass-losing winds, the convection-dominated (CDAF) solution, and luminous hot flows near $\dot m_{\rm crit}$ . In casual use the terms are interchangeable; strictly, ADAF is the specific model and RIAF is the class.

Why do the ions get so much hotter than the electrons?

Two facts combine. First, turbulent dissipation acts on the bulk plasma, dominated by the massive ions, so most of the viscous heat lands on them. Second, in a thin disk Coulomb collisions would quickly share that heat — but an ADAF is so dilute that the ion–electron energy-exchange time $t_{ie}$ (eq. 7) far exceeds the infall time. The ions therefore run up to nearly the virial temperature, $\sim10^{12}$ K, while the electrons, drained by radiation, settle one to two orders of magnitude cooler, and the gas reaches the horizon before the two ever equilibrate.

If the luminosity is so low, how do we detect these flows?

By their distinctive non-thermal spectra rather than by brightness. The hot electrons emit synchrotron radiation, producing the characteristic submillimeter "bump" seen in Sgr A*; those photons are inverse-Compton up-scattered to X-rays, and bremsstrahlung adds a hard tail. The overall spectral energy distribution is broad and hard, quite unlike the blackbody-like "big blue bump" of a thin disk, and in X-ray binaries the corona shows up as a hard power law with a high-energy cutoff. The Event Horizon Telescope now resolves the emitting plasma directly, turning a spectral inference into an image.

Does all the captured gas actually reach the black hole?

Often not. Because an ADAF has a positive Bernoulli parameter (eq. 8), the gas is only weakly bound, and a substantial fraction can be driven off as a wind before reaching the horizon; in the ADIOS picture the accretion rate decreases inward, $\dot M(R)\propto R^s$ . This is not just theory: for both Sgr A and M87, the rate inferred at the horizon (from Faraday rotation and polarization) is far below the Bondi capture rate measured at large radius — direct evidence that most captured gas never makes it in, and why the simple $L\propto\dot M^2$ estimate slightly overpredicts Sgr A*.

What connects ADAFs to relativistic jets?

The same properties that make an ADAF radiatively inefficient make it good at launching jets: it is hot, thick ( $H/R\sim1$ ), strongly magnetized, and weakly bound, so ordered poloidal field can thread the inner region and extract rotational energy. In the magnetically arrested disk (MAD) limit, magnetic flux piles up at the hole until it resists inflow, maximizing the Blandford–Znajek power tapped from a spinning black hole. This is why the least luminous states — the low/hard state, low-luminosity AGN like M87 — so often carry the most powerful steady jets: the energy that is not radiated comes out kinetically.

Why does the accretion state switch abruptly between the branches?

Because the two solutions are distinct branches separated by thermal stability, not endpoints of a continuum. Below $\dot m_{\rm crit}\sim\alpha^2$ the hot advective solution is self-consistent. Push the rate up and cooling ( $\propto\dot M^2$ ) overtakes advection ( $\propto\dot M$ ); the gas radiates, cools, contracts vertically, raising its density and speeding cooling further — a runaway that collapses the flow onto the cold thin-disk branch. Because the collapse is a runaway it is fast and shows hysteresis, exactly what is seen in the hard-to-soft state transitions of black-hole binaries.