Bondi Accretion

The simplest way to feed a black hole

Strip away every complication — no rotation, no magnetic field, no relative motion, no disk — and ask the most basic question in accretion physics: if you drop a point mass into a static, uniform sea of gas, how fast does it swallow that gas? The answer, worked out by Hermann Bondi in 1952, is one of the cleanest results in astrophysics. The mass it captures per second is

Ṁ = 4π λ (GM)² ρ∞ / c_s³

where M is the accretor's mass, ρ∞ is the density of the ambient gas far away, c_s is that gas's sound speed, and λ is a pure number of order unity. The scaling is the whole story: Ṁ ∝ M² ρ / c_s³. Double the mass and the feeding rate quadruples. Double the density and it doubles. But raise the temperature so the sound speed doubles, and the rate drops by a factor of eight. Hot gas is slippery — it resists being captured.

This is the Bondi rate, and it is the reference value against which the entire field measures itself. When we say Sagittarius A* accretes "far below Bondi," or that a supermassive black hole's feedback "self-regulates near the Bondi rate," this single formula is the yardstick. It is the baseline a compact object would feed at if nothing got in the way.

How it works: from drift to free-fall through a sonic point

Picture a black hole at rest in a vast cloud of gas. Far away, the gas does not know the black hole is there: it sits in pressure equilibrium, thermal motions characterised by the sound speed c_s. As you move inward, gravity starts to compete with pressure. The crossover happens at the Bondi radius:

r_B = GM / c_s²

This is the radius at which the gravitational binding energy of a gas element, GM/r, equals its thermal energy, ~c_s². Equivalently, it is where the escape speed from the accretor matches the ambient sound speed. Outside r_B, pressure wins and the gas barely notices the accretor. Inside r_B, gravity wins and the gas is captured — it must fall in.

But the transition is not abrupt. Bondi's insight was that a steady, smooth inflow exists only if the gas accelerates through a single critical point — the sonic radius — where its inward speed equals the local sound speed. Outside the sonic point the flow is subsonic and quasi-hydrostatic, drifting inward slowly. At the sonic point it goes transonic. Inside it the flow is supersonic and in near free-fall, plunging toward the accretor. For an adiabatic gas with index γ, the sonic radius sits at

r_s = (5 − 3γ)/4 · r_B

so for the common monatomic case γ = 5/3 the sonic point is at r_s = r_B/2, and for an isothermal gas (γ = 1) it coincides with the Bondi radius. The requirement that the flow pass smoothly through this point is what pins down the eigenvalue λ and therefore the unique value of Ṁ. There are infinitely many mathematically allowed flows; only the transonic one is physical, connecting a static atmosphere at infinity to supersonic infall at the horizon.

Worked example: a 10-solar-mass black hole in the warm ISM

Let us put numbers on it. Take an isolated stellar-mass black hole, M = 10 M_⊙ = 2 × 10³¹ kg, drifting through the warm neutral interstellar medium: number density n ≈ 0.5 cm⁻³ (mass density ρ∞ ≈ 1 × 10⁻²⁴ g/cm³ = 10⁻²¹ kg/m³) at temperature T ≈ 8000 K, giving a sound speed c_s ≈ 10 km/s = 10⁴ m/s. First, the Bondi radius:

r_B = GM / c_s²
    = (6.67×10⁻¹¹)(2×10³¹) / (10⁴)²
    = (1.33×10²¹) / (10⁸)
    = 1.3×10¹³ m  ≈  90 AU

So gravity dominates within ~90 AU — about twice the size of the Solar System. Now the rate, using λ = 1/4 for γ = 5/3:

Ṁ = 4π λ (GM)² ρ∞ / c_s³
   = 4π (0.25) (1.33×10²¹)² (10⁻²¹) / (10⁴)³
   = (3.14) (1.78×10⁴²) (10⁻²¹) / (10¹²)
   = 5.6×10⁹ kg/s
   ≈ 9×10⁻⁸ M_⊙ / yr

About a hundred-millionth of a solar mass per year. If that gas reached the horizon and radiated at efficiency η ≈ 0.1, the luminosity would be L = η Ṁ c² ≈ 0.1 × (5.6×10⁹)(9×10¹⁶) ≈ 5×10²⁵ W ≈ 1.3×10⁵ L_⊙. In practice such isolated black holes are radiatively inefficient at these low rates and would be far dimmer — which is precisely why we have never unambiguously found one of the estimated 10⁸ isolated stellar-mass black holes wandering the Milky Way. The Bondi formula tells us they should be glowing faintly; the inefficiency of low-rate accretion tells us why they hide.

Quantitative analysis: where the formula comes from

The Bondi solution follows from two conservation laws for a steady, spherical, adiabatic flow. Mass conservation through every shell gives a constant rate:

Ṁ = 4π r² ρ(r) v(r) = const

The radial Euler equation (momentum) for steady flow is

v dv/dr + (1/ρ) dP/dr + GM/r² = 0

with a polytropic equation of state P = K ρ^γ so that c_s² = γP/ρ ∝ ρ^(γ−1). Combining these and differentiating, you obtain a relation in which the term multiplying dv/dr vanishes exactly when v = c_s. For the flow to remain smooth (finite gradients) at that radius, the right-hand side must vanish simultaneously — this is the critical-point condition that selects the sonic radius r_s = (5−3γ)/4 · r_B and the corresponding sonic sound speed c_s(r_s)² = 2c_s²/(5−3γ).

Tracing the transonic solution back out to infinity and evaluating Ṁ at the sonic point yields the closed form:

Ṁ = 4π λ_c (GM)² ρ∞ / c_s³

with   λ_c = (1/4) · [2/(5−3γ)]^((5−3γ)/(2(γ−1)))

This eigenvalue λ_c is a smooth function of γ alone. Two limits matter:

• γ = 5/3 (monatomic, adiabatic): λ_c = 1/4 = 0.25. The sonic point sits at r_B/2 and the formula is at its "stiffest."
• γ = 1 (isothermal): λ_c → e^(3/2)/4 ≈ 1.12. The gas can be compressed without heating, so accretion is more efficient and the sonic point reaches all the way to r_B.

Between these, λ_c rises monotonically from 0.25 to 1.12 — a factor of only ~4.5 — which is why we say λ is "of order unity." The remarkable feature of the Bondi result is how little the microphysics matters: across the entire physical range of γ, the captured rate changes by less than a factor of five, and the scaling Ṁ ∝ (GM)² ρ∞ / c_s³ is exact.

Variants and regimes

The pristine Bondi problem is an idealisation. The major generalisations relax its assumptions one at a time:

• Bondi-Hoyle-Lyttleton (moving accretor). When the accretor moves through the gas at speed v, the effective capture combines thermal and bulk motion: Ṁ ≈ 4π (GM)² ρ∞ / (c_s² + v²)^(3/2), with capture radius r_acc = 2GM/(c_s² + v²). In the static limit v → 0 this recovers Bondi; in the supersonic limit v ≫ c_s it becomes the Hoyle-Lyttleton (1939) result Ṁ ∝ (GM)² ρ / v³, with an accretion column forming in the wake behind the body.
• Accretion with angular momentum. Any net rotation in the gas defeats radial infall: a centrifugal barrier forms at the circularisation radius and the gas piles into a disk. The Bondi rate then describes only the supply captured at r_B; an inner disk and viscous transport govern what actually reaches the accretor.
• Radiatively inefficient flows (ADAF/RIAF). At rates well below Eddington, the inflowing gas cannot cool in a free-fall time; it stays hot, thick, and advects most of its energy inward or drives it out in winds. The horizon-scale rate falls far below the Bondi capture rate.
• Relativistic Bondi (Michel 1972). Replacing Newtonian gravity with the Schwarzschild metric gives the general-relativistic generalisation, important when r_B approaches the gravitational radius — relevant for very hot media or near-horizon scales.
• Bondi accretion with feedback. In galaxy and cluster simulations, the Bondi rate is the standard sub-grid prescription for fueling a central black hole; the energy it returns regulates the surrounding gas, closing a feedback loop that shapes galaxy evolution.

Observational status: the Sagittarius A* benchmark

The most-studied real-world application is the supermassive black hole at the centre of the Milky Way, Sagittarius A* (M ≈ 4 × 10⁶ M_⊙). Chandra X-ray imaging resolves the hot (~10⁷ K) gas at and beyond the Bondi radius, which for Sgr A* is r_B ≈ 0.04–0.1 pc (roughly 10⁵ Schwarzschild radii). Plugging the measured density and temperature into the Bondi formula gives a capture rate of order 10⁻⁵–10⁻⁶ M_⊙ per year.

Yet Sgr A* is astonishingly faint — about 10⁻⁹ of its Eddington luminosity. Reproducing that dimness requires the rate reaching the event horizon to be ~10⁻⁸ M_⊙ per year, a hundred to a thousand times smaller than the Bondi capture rate. The Event Horizon Telescope's polarised images and the resolved infrared flares both support a tenuous, magnetically arrested, radiatively inefficient inner flow. The resolution: most of the gas captured at the Bondi radius never reaches the horizon — it is expelled by convection and winds, or stored, in a flow whose density profile is far flatter than the ρ ∝ r⁻³/² that free-fall would imply. Sgr A* is the canonical case of sub-Bondi feeding, and it is the cleanest demonstration that the Bondi rate sets the outer supply, not the swallowed rate.

The same logic scales up. In giant elliptical galaxies and cluster cores, Chandra measurements of the hot atmosphere yield Bondi rates that, combined with feedback efficiency, plausibly power the radio jets that carve X-ray cavities — making Bondi accretion a key ingredient in the "maintenance mode" AGN feedback that keeps cluster cores from cooling catastrophically.

How the Bondi rate varies across systems

System	Mass M	Ambient n	Sound speed c_s	Bondi radius r_B	Ṁ_Bondi (M_⊙/yr)
Stellar BH in warm ISM	10 M_⊙	0.5 cm⁻³	~10 km/s	~90 AU	~9 × 10⁻⁸
Stellar BH in cold cloud	10 M_⊙	100 cm⁻³	~1 km/s	~9000 AU	~2 × 10⁻⁴
Neutron star in O-star wind (BHL)	1.4 M_⊙	~10⁶ cm⁻³	v ~ 1000 km/s	~0.02 AU	~10⁻⁸
Sagittarius A*	4 × 10⁶ M_⊙	~26 cm⁻³	~550 km/s	~0.04 pc	~10⁻⁵
M87* (cluster core)	6.5 × 10⁹ M_⊙	~0.1 cm⁻³	~700 km/s	~0.1 kpc	~0.1
Giant elliptical SMBH	10⁹ M_⊙	~0.1 cm⁻³	~600 km/s	~30 pc	~0.01

The table makes the M² ρ / c_s³ scaling vivid. The cold-cloud case beats the warm-ISM case by ~2000× for the same black hole, almost entirely because the sound speed is ten times lower (10³ in c_s³) and the density 200× higher. And the supermassive holes capture orders of magnitude more material despite sitting in tenuous, hot atmospheres, purely because of the M² term.

Common pitfalls and misconceptions

• Confusing the Bondi capture rate with the horizon rate. The formula gives the rate at which gas crosses inward through the Bondi radius. How much of that actually reaches the accretor depends on angular momentum, radiative efficiency, and outflows. For Sgr A* these differ by 100–1000×. Never quote Ṁ_Bondi as "what the black hole eats."
• Treating λ as a free parameter. λ is a fixed eigenvalue determined by γ alone, ranging only from 0.25 to ~1.12. It is not a tunable fudge factor; the transonic condition fixes it exactly.
• Forgetting that c_s³ dominates. Because the rate goes as c_s⁻³, modest temperature changes have outsized effects. A factor-of-2 heating cuts the rate by 8×. This is why feedback that heats the surrounding gas so effectively throttles accretion.
• Ignoring angular momentum entirely. Pure Bondi (spherical) flow is rare in nature; almost all real gas has enough angular momentum to form a disk inside r_B. Bondi is the supply boundary condition, not the full accretion story.
• Assuming Bondi implies high luminosity. The Bondi rate is a mass rate, not a luminosity. At low Ṁ/Ṁ_Edd the flow is radiatively inefficient, so even a "healthy" Bondi supply can correspond to a nearly invisible accretor.
• Using the static formula for a fast-moving accretor. A neutron star plowing through a 1000 km/s wind has v ≫ c_s; you must use the Bondi-Hoyle-Lyttleton form with (c_s² + v²)^(3/2), or you will overestimate the rate by orders of magnitude.