Bondi Accretion

Bondi accretion is the reference problem of accretion astrophysics: how fast does an isolated gravitating mass swallow gas from an ocean of it? Hermann Bondi answered this in 1952 with a model stripped to its essentials — the flow is steady in time, perfectly spherically symmetric, unmagnetized, non-self-gravitating, and carries zero net angular momentum. Far away the gas is a uniform reservoir of density $\rho_\infty$ and sound speed $c_{s,\infty}$ ; the accretor is a point mass $M$ that consumes whatever reaches the center. From this austere setup one number emerges, the Bondi rate, and it has become the fiducial estimate against which every real accretion flow is measured. The physics is a tug-of-war between gravity pulling gas in and pressure pushing it back. Deep in the potential well gas plunges faster than sound; far out it drifts subsonically, buoyed by its own pressure. Continuity and the steady Euler equation together force the flow to thread a single critical radius — the sonic point — where inflow speed equals the local sound speed. Demanding that the flow pass smoothly through that point is a regularity condition, and it does something remarkable: it collapses a one-parameter family of formally allowed flows down to one, selecting the accretion rate as an eigenvalue rather than a free choice. Two length scales organize the solution. The Bondi radius $r_B = GM/c_{s,\infty}^2$ marks where gravity overwhelms thermal pressure — gas inside it is captured. The sonic radius $r_s$ , a $\gamma$ -dependent fraction of $r_B$ , marks where the inflow goes supersonic. Because the answer scales as $\dot{M}\propto M^2\rho_\infty/c_{s,\infty}^3$ , the Bondi formula reaches everywhere in astrophysics: black-hole growth in hot halos, AGN fueling, feedback in cosmological simulations, and the glow of compact objects drifting through the interstellar medium.

Key takeaways

Bondi accretion is the steady, spherically symmetric, transonic inflow of gas from an infinite reservoir onto a gravitating point mass, driven by gravity against pressure with no angular momentum.
The mass accretion rate is fixed by requiring the flow to pass smoothly through a sonic point, giving $\dot{M}_B = 4\pi\lambda_s\,G^2M^2\rho_\infty/c_{s,\infty}^3$ .
The dimensionless eigenvalue $\lambda_s$ depends only on the adiabatic index $\gamma$ , running from $\lambda_s=1/4$ at $\gamma=5/3$ up to $e^{3/2}/4\approx1.12$ in the isothermal limit $\gamma\to1$ .
The Bondi radius $r_B=GM/c_{s,\infty}^2$ is the gravitational capture scale; the sonic radius $r_s=\tfrac{5-3\gamma}{4}\,r_B$ is where the inflow turns supersonic.
The rate is a ceiling, not a prediction: rotation, magnetic fields, turbulence, and radiative feedback all cut the real rate below the Bondi value.
For an accretor moving at speed $v_\infty$ through the gas, the Bondi–Hoyle–Lyttleton form replaces $c_{s,\infty}$ by the effective speed $(c_{s,\infty}^2+v_\infty^2)^{1/2}$ .
Key idea: steady spherical inflow must pass smoothly through a sonic point, and that single regularity condition turns continuity and Bernoulli into an eigenvalue problem whose eigenvalue is the accretion rate $\dot{M}_B\propto G^2M^2\rho_\infty/c_{s,\infty}^3$ .

The core insight

Steady spherical inflow must pass smoothly through a sonic point; that single regularity condition turns the continuity and Bernoulli equations into an eigenvalue problem whose eigenvalue is the accretion rate $\dot{M}_B\propto G^2M^2\rho_\infty/c_{s,\infty}^3$ .

Where this lives today

The Bondi formula is the workhorse boundary condition of modern accretion astrophysics. In cosmological simulations that cannot resolve an accretion disk — sub-parsec scales sit far below the grid — black-hole "sink particles" are grown with a Bondi (often Bondi–Hoyle) prescription on the resolvable $\rho$ , $c_s$ , and $M$ , sometimes with an ad hoc boost factor for unresolved cold clumps. This recipe underpins the AGN-feedback loops in EAGLE, IllustrisTNG, and their kin, and its systematic uncertainty is a live research topic. At the Galactic Center, Chandra's resolving of the Bondi radius of Sgr A* turned the formula into a measurement: the captured rate is $\sim10^{-5}\,M_\odot\,\mathrm{yr^{-1}}$ , yet horizon-scale polarization limits imply far less crosses the horizon — direct evidence for the strong outflows of radiatively inefficient accretion the Event Horizon Telescope now images. And for the Galaxy's hidden isolated neutron stars and stellar black holes, the Bondi–Hoyle–Lyttleton rate sets the faint luminosity that future X-ray surveys hope to catch — the same 1952 calculation, still the benchmark seventy years on.

The physical picture

Picture gas raining onto a point mass. Far away it barely feels the mass and drifts subsonically, held up by its own pressure; pressure signals travel outward freely because the gas moves slower than they do. Close in, gravity wins and the gas plunges faster than sound — pressure can no longer send word upstream. The switch from subsonic to supersonic happens at exactly one radius, the sonic point, and demanding smoothness there pins down the single acceptable flow out of the infinitely many the equations otherwise permit. The mathematics is a de Laval nozzle in reverse: the sonic point is the throat, and only one rate lets the flow accelerate through it without choking or stalling. The first figure shows the structure of the allowed solutions in the plane of Mach number versus radius. The two conservation laws admit a whole family of curves, all crossing at a single saddle — the sonic point. Only two curves pass through the saddle: the transonic accretion solution (slow at infinity, supersonic at the center) and its time-reverse, the transonic wind. Every other curve is subsonic everywhere or supersonic everywhere and never crosses $M=1$ . Nature, filling a reservoir onto a sink, picks the accretion diagonal.

Figure

Structure of Bondi solutions in the Mach-number–radius plane: the continuity and Euler equations cross at a single saddle, the sonic point r_s where the inflow speed equals the sound speed and u_s² = c_s² = GM/(2 r_s). Only the transonic accretion branch (blue) threads the saddle from subsonic at large r to supersonic at small r; the wind branch is its time-reverse, and the amber and green families never cross Mach 1

Why does the rate come out as it does? A bigger mass reaches farther: the capture cross-section scales as $r_B^2\propto M^2$ , so heavy accretors gather gas from a disproportionately larger volume. Hotter gas resists strongly: a larger $c_{s,\infty}$ shrinks $r_B$ and stiffens the pressure support, and because it enters as $c_{s,\infty}^{-3}$ the suppression is steep — which is why feedback heating so efficiently brakes accretion. Denser surroundings supply more, linearly in $\rho_\infty$ . The one piece dimensional analysis cannot give is the order-unity factor $\lambda_s$ ; that comes only from the transonic solution and depends on compressibility through $\gamma$ . The second figure shows that eigenvalue and the real numbers it takes.

Figure

The dimensionless eigenvalue λ_s as a function of adiabatic index γ, from the transonic condition λ_s = ¼ (2/(5−3γ)) raised to (5−3γ)/2(γ−1): it falls monotonically from λ_s = 1.12 at γ = 1 (isothermal) through 0.707 at γ = 4/3 and 0.500 at γ = 3/2 down to 0.25 at γ = 5/3 (monatomic gas), so a stiffer equation of state accretes more slowly

The core equations

The headline result is the Bondi rate itself: the steady mass per unit time captured by a point mass from a uniform reservoir. Everything below builds toward it or unpacks the eigenvalue $\lambda_s$ inside it.

\dot{M}_B = 4\pi\lambda_s\,\frac{G^2M^2\rho_\infty}{c_{s,\infty}^3} \qquad (1)

where: $\dot{M}_B$ — steady mass accretion rate; $\lambda_s$ — dimensionless transonic eigenvalue, set only by $\gamma$ ; $M$ — mass of the point accretor; $\rho_\infty$ — ambient gas density far away; $c_{s,\infty}$ — ambient adiabatic sound speed far away; $G$ — Newton's constant. Read it as capture cross-section ( $\propto M^2/c_{s,\infty}^4$ ) times supply ( $\rho_\infty c_{s,\infty}$ ), with the O(1) factor $\lambda_s$ from the sonic-point analysis. The rate rests on two conservation laws. The first is steady spherical continuity: the same mass flux crosses every shell, so area times density times speed is constant.

\dot{M} = 4\pi r^2\rho(r)\,|u(r)| = \text{const} \qquad (2)

where: $u(r)$ — radial velocity (negative inward, $|u|$ is the speed); $\rho(r)$ — density at radius $r$ ; $r$ — spherical radius. The constant is the accretion rate itself, evaluated at any shell. The second is the steady radial Euler equation — Newton's second law for a fluid element, balancing inertia against the pressure gradient and gravity.

u\frac{\mathrm{d}u}{\mathrm{d}r} = -\frac{1}{\rho}\frac{\mathrm{d}P}{\mathrm{d}r} - \frac{GM}{r^2} \qquad (3)

where: $P$ — gas pressure; $GM/r^2$ — inward gravitational acceleration of the point mass. The left side is the convective acceleration of steady flow; the right side is pressure buoyancy minus gravity. To close the system we relate pressure to density with a polytrope, which also defines the sound speed that does all the work.

P = K\rho^\gamma,\qquad c_s^2 = \frac{\mathrm{d}P}{\mathrm{d}\rho} = \gamma K\rho^{\gamma-1} \qquad (4)

where: $K$ — polytropic constant (an entropy label); $\gamma$ — adiabatic index; $c_s$ — local adiabatic sound speed. This turns the pressure gradient into $\tfrac{c_s^2}{\rho}\mathrm{d}\rho/\mathrm{d}r$ , reducing everything to $u$ , $\rho$ , and $c_s$ . Integrating Euler with the polytrope gives the Bernoulli integral — energy per unit mass conserved along a streamline, combining kinetic energy, enthalpy, and gravity.

\frac{1}{2}u^2 + \frac{c_s^2}{\gamma-1} - \frac{GM}{r} = \frac{c_{s,\infty}^2}{\gamma-1} \qquad (5)

where: $\tfrac12 u^2$ — specific kinetic energy; $c_s^2/(\gamma-1)$ — specific enthalpy of a polytrope; $-GM/r$ — specific potential energy. The right side is the value at infinity, where $u\to0$ and $r\to\infty$ . Regularity at the sonic point fixes both where the flow goes supersonic and how fast sound travels there.

r_s = \frac{5-3\gamma}{4}\,\frac{GM}{c_{s,\infty}^2},\qquad c_s(r_s)^2 = \frac{2}{5-3\gamma}\,c_{s,\infty}^2 \qquad (6)

where: $r_s$ — sonic (critical) radius where $|u|=c_s$ ; $c_s(r_s)$ — sound speed there, raised above ambient by adiabatic compression of the infalling gas. Both formulas require $\gamma<5/3$ for a finite positive $r_s$ ; the monatomic case is a delicate limit. Assembling the rate at the sonic point gives the eigenvalue as an explicit function of $\gamma$ .

\lambda_s = \frac{1}{4}\left(\frac{2}{5-3\gamma}\right)^{\!\frac{5-3\gamma}{2(\gamma-1)}} \qquad (7)

where: $\lambda_s$ — the transonic eigenvalue; it equals $1/4$ at $\gamma=5/3$ , $1/\sqrt{2}\approx0.707$ at $\gamma=4/3$ , $1/2$ at $\gamma=3/2$ , and tends to $e^{3/2}/4\approx1.12$ as $\gamma\to1$ . This one dimensionless number is the whole content of the transonic solution beyond the scaling. When the accretor moves through the gas, the same physics generalizes by adding the bulk speed in quadrature with the sound speed.

\dot{M}_{\mathrm{BHL}} = \frac{4\pi G^2M^2\rho_\infty}{\left(c_{s,\infty}^2 + v_\infty^2\right)^{3/2}} \qquad (8)

where: $v_\infty$ — bulk speed relative to the gas at infinity. For $v_\infty\ll c_{s,\infty}$ it reduces to the Bondi rate; for $v_\infty\gg c_{s,\infty}$ it becomes the Hoyle–Lyttleton rate $\propto v_\infty^{-3}$ with a downstream wake.

Derivation AThe transonic master equation and the sonic point

The goal is to combine steady continuity and the steady Euler equation into a single differential equation for the velocity gradient, and then to show that a smooth flow spanning subsonic to supersonic is possible only if it passes through a special radius where two factors vanish together. That radius is the sonic point, and the requirement that the flow be regular there — not that any boundary value be imposed — is what ultimately selects a unique accretion rate. We carry the pressure gradient entirely in terms of the sound speed so the gas's compressibility stays explicit throughout.

Step 1. Take the logarithm of continuity, $\ln(4\pi r^2\rho|u|)=\ln\dot{M}$ , and differentiate in $r$ : $\tfrac{2}{r}+\tfrac{1}{\rho}\tfrac{\mathrm{d}\rho}{\mathrm{d}r}+\tfrac{1}{u}\tfrac{\mathrm{d}u}{\mathrm{d}r}=0$ . Why: log-differentiating the conserved flux turns a product law into the sum of fractional gradients we need to eliminate the density.

Step 2. Rewrite Euler with the polytrope. Since $\mathrm{d}P/\mathrm{d}r=c_s^2\,\mathrm{d}\rho/\mathrm{d}r$ , equation (3) becomes $u\tfrac{\mathrm{d}u}{\mathrm{d}r}=-\tfrac{c_s^2}{\rho}\tfrac{\mathrm{d}\rho}{\mathrm{d}r}-\tfrac{GM}{r^2}$ . Why: replacing the pressure gradient with the sound speed removes $P$ and exposes $c_s$ as the quantity deciding whether pressure can resist the flow.

Step 3. Solve continuity for $\tfrac{1}{\rho}\tfrac{\mathrm{d}\rho}{\mathrm{d}r}=-\tfrac{2}{r}-\tfrac{1}{u}\tfrac{\mathrm{d}u}{\mathrm{d}r}$ and substitute into Step 2. Collecting the velocity-gradient terms gives the master equation $\left(u-\tfrac{c_s^2}{u}\right)\tfrac{\mathrm{d}u}{\mathrm{d}r}=\tfrac{2c_s^2}{r}-\tfrac{GM}{r^2}$ . Why: the whole problem in one line — the left bracket carries the sign of $u^2-c_s^2$ (sub- vs supersonic), the right side the contest between pressure support and gravity.

Step 4. Look at where the left bracket vanishes. Where $|u|=c_s$ , the coefficient of $\mathrm{d}u/\mathrm{d}r$ is zero; unless the right side vanishes there too, the gradient diverges and the solution is unphysical. A smooth transonic flow demands both sides vanish at the same radius: $u_s^2=c_s(r_s)^2$ and $\tfrac{2c_s(r_s)^2}{r_s}=\tfrac{GM}{r_s^2}$ . Why: this regularity condition is not a boundary condition we choose but a consistency requirement — the mathematical heart of why the rate is an eigenvalue.

Step 5. Rearrange the right-hand condition to $r_s=\dfrac{GM}{2c_s(r_s)^2}$ , tying the sonic radius to the sound speed there. Why: at $r_s$ the escape-speed-squared scale $2GM/r$ equals twice the sound-speed-squared — gravity has just won over pressure, the Bondi-radius idea at the critical point.

Go deeper · Derivation B · 2 minFrom the sonic point to the eigenvalue accretion rate

Derivation BFrom the sonic point to the eigenvalue accretion rate

The goal is to convert the two sonic-point conditions into an explicit accretion rate. We evaluate the Bernoulli integral at the sonic point to fix the sound speed there in terms of ambient conditions, then feed that back through continuity — using the polytrope for the sonic density — to assemble the rate and read off the eigenvalue $\lambda_s(\gamma)$ . The physical input is that energy is conserved along a streamline, so the enthalpy and kinetic energy the gas gains falling to $r_s$ come cleanly out of its potential energy.

Step 1. Insert the sonic conditions $u_s^2=c_s(r_s)^2$ and $GM/r_s=2c_s(r_s)^2$ into Bernoulli (5): $\tfrac12 c_s(r_s)^2+\tfrac{c_s(r_s)^2}{\gamma-1}-2c_s(r_s)^2=\tfrac{c_{s,\infty}^2}{\gamma-1}$ . Why: every term is now a multiple of $c_s(r_s)^2$ , so Bernoulli becomes one algebraic equation for the sonic sound speed.

Step 2. Collect the left side. The bracket is $\left(\tfrac12+\tfrac{1}{\gamma-1}-2\right)c_s(r_s)^2=\tfrac{5-3\gamma}{2(\gamma-1)}\,c_s(r_s)^2$ , so $\tfrac{5-3\gamma}{2(\gamma-1)}c_s(r_s)^2=\tfrac{c_{s,\infty}^2}{\gamma-1}$ , giving $c_s(r_s)^2=\dfrac{2}{5-3\gamma}\,c_{s,\infty}^2$ . Why: this is equation (6) — the infalling gas heats adiabatically, so its sound speed at the sonic point exceeds the ambient value, and the factor $(5-3\gamma)$ warns that $\gamma=5/3$ is singular.

Step 3. Get the sonic radius by substituting Step 2 into $r_s=GM/[2c_s(r_s)^2]$ : $r_s=\dfrac{5-3\gamma}{4}\dfrac{GM}{c_{s,\infty}^2}=\tfrac{5-3\gamma}{4}r_B$ . Why: the sonic point is a fixed fraction of the Bondi radius, ranging from $\tfrac12 r_B$ for isothermal gas down to zero for monatomic gas.

Step 4. Find the sonic density from the polytrope. Since $c_s^2\propto\rho^{\gamma-1}$ , $\rho_s/\rho_\infty=\left(c_s(r_s)/c_{s,\infty}\right)^{2/(\gamma-1)}=\left(\tfrac{2}{5-3\gamma}\right)^{1/(\gamma-1)}$ . Why: density piles up by the same adiabatic relation that raised the sound speed, and continuity needs it to compute the flux.

Step 5. Assemble $\dot{M}_B=4\pi r_s^2\,\rho_s\,c_s(r_s)$ from continuity at the sonic point, substituting $r_s$ , $\rho_s$ , and $c_s(r_s)$ . The powers of $\tfrac{2}{5-3\gamma}$ combine to give $\dot{M}_B=4\pi\lambda_s\,\dfrac{G^2M^2\rho_\infty}{c_{s,\infty}^3}$ with $\lambda_s=\tfrac14\left(\tfrac{2}{5-3\gamma}\right)^{(5-3\gamma)/[2(\gamma-1)]}$ . Why: equations (1) and (7) together — the rate is now fixed by ambient conditions and $\gamma$ alone, with no free boundary parameter, which is what "eigenvalue" means here.

Step 6. Check the endpoints. As $\gamma\to1$ , l'Hôpital on the exponent gives $\lambda_s\to e^{3/2}/4\approx1.12$ ; at $\gamma=5/3$ base and exponent conspire so $\lambda_s\to1/4$ even though $r_s\to0$ . Why: recovering Bondi's limiting values confirms the algebra and shows the monatomic case is a limit, not a substitution.

Go deeper · Derivation C · 1 minThe Bondi radius from energy balance

Derivation CThe Bondi radius from energy balance

The goal is to obtain the gravitational capture scale — the radius inside which the point mass, not gas pressure, controls the dynamics — by a simple energy argument, and to connect it to the sonic radius already derived. This scale is what makes the Bondi rate intuitive, since it is the effective size of the accretor's cross-section for gathering gas. We set the depth of the potential well against the gas's thermal energy and read off where they trade places.

Step 1. Compare the escape-speed-squared to the ambient sound-speed-squared: gravity dominates when $\tfrac{2GM}{r}\gtrsim c_{s,\infty}^2$ . Why: a gas parcel is gravitationally captured once its thermal energy per unit mass, of order $c_{s,\infty}^2$ , is smaller than the binding energy per unit mass it would need to escape, of order $GM/r$ .

Step 2. Solve for the boundary radius, $r\sim\dfrac{2GM}{c_{s,\infty}^2}$ . Inside it gas cannot pressure-support itself against infall; outside it is essentially free. Why: this is the accretor's sphere of influence, and the cross-section $\pi r^2\propto M^2$ is the geometric origin of the $M^2$ in the rate.

Step 3. Adopt the conventional $r_B\equiv\dfrac{GM}{c_{s,\infty}^2}$ , dropping the factor of two, and relate it to the sonic radius $r_s=\tfrac{5-3\gamma}{4}r_B$ . Why: this lets the sonic radius be quoted as a clean fraction of $r_B$ — one-half for isothermal, zero for monatomic gas — the form used throughout the literature.

Limiting cases

The eigenvalue $\lambda_s$ and the sonic radius depend entirely on the adiabatic index $\gamma$ ; the four rows below span the physically important equations of state.

Case	Result	Meaning
Isothermal ( $\gamma=1$ )	$\lambda_s=\tfrac14 e^{3/2}\approx1.12,\ \ r_s=\tfrac12\,GM/c_{s,\infty}^2$	Softest gas; largest rate, because pressure cannot build up to resist inflow. Sonic point at half the Bondi radius.
Radiation-dominated ( $\gamma=4/3$ )	$\lambda_s=1/\sqrt{2}\approx0.707,\ \ r_s=\tfrac14\,GM/c_{s,\infty}^2$	Relevant for radiation-pressure-supported gas; intermediate eigenvalue. Note this is $1/\sqrt2$ , not $1/2$ .
Diatomic / mid-stiff ( $\gamma=3/2$ )	$\lambda_s=1/2,\ \ r_s=\tfrac18\,GM/c_{s,\infty}^2$	The value $\lambda_s=1/2$ belongs here, at $\gamma=3/2$ — a common point of confusion with the $\gamma=4/3$ row.
Monatomic ideal gas ( $\gamma=5/3$ )	$\lambda_s=\tfrac14,\ \ r_s=0$	Stiffest adiabatic case; smallest $\lambda_s$ . The sonic radius formally shrinks to the origin and the flow is nearly free-fall, requiring a limiting treatment.

Common trapThe Bondi rate is a prediction of the real accretion rate

It is not — it is a ceiling. Equation (1) assumes zero angular momentum, no radiation pressure, no magnetic fields, and no turbulence, and every omitted ingredient lowers the real rate. Even a whisper of specific angular momentum forms a centrifugally supported disk that regulates inflow, and feedback drives outflows that choke the supply; observed rates onto real black holes run orders of magnitude below Bondi. Two further traps: the $\gamma=4/3$ eigenvalue is $1/\sqrt2\approx0.707$ , not $1/2$ (that is the $\gamma=3/2$ case), and at $\gamma=5/3$ the sonic-radius formula gives $r_s\to0$ , so the transonic solution must be taken as a careful limit, not by naive substitution.

Worked examples

1 · The Bondi rate for Sgr A* (fully worked)

The Galactic-Center black hole Sgr A\ has mass $M=4\times10^6\,M_\odot$ . Chandra finds hot gas near its Bondi radius with temperature $T\approx10^7\ \mathrm{K}$ and number density $n\approx100\ \mathrm{cm^{-3}}$ . Take a monatomic ionized plasma ( $\gamma=5/3$ , mean molecular weight $\mu=0.6$ ). Find the ambient sound speed, the Bondi radius, and the Bondi accretion rate, and compare it to the Eddington rate.*

Step 1. Ambient sound speed, $c_{s,\infty}=\sqrt{\gamma k_B T/(\mu m_H)}=\sqrt{(5/3)(1.38\times10^{-16})(10^7)/[(0.6)(1.67\times10^{-24})]}\approx4.8\times10^{7}\ \mathrm{cm\,s^{-1}}=\mathbf{479\ km\,s^{-1}}$ .

Step 2. Bondi radius $r_B=GM/c_{s,\infty}^2$ with $M=4\times10^6\times1.99\times10^{33}=7.96\times10^{39}\ \mathrm{g}$ : $r_B\approx2.3\times10^{17}\ \mathrm{cm}\approx\mathbf{0.075\ pc}$ — the scale Chandra actually resolves, a few arcseconds at the Galactic Center.

Step 3. Ambient mass density $\rho_\infty=\mu m_H n=(0.6)(1.67\times10^{-24})(100)\approx1.0\times10^{-22}\ \mathrm{g\,cm^{-3}}$ .

Step 4. Bondi rate. For $\gamma=5/3$ , $\lambda_s=1/4$ , so $\dot{M}_B=\pi G^2M^2\rho_\infty/c_{s,\infty}^3$ . Plugging in: $\dot{M}_B\approx8.1\times10^{20}\ \mathrm{g\,s^{-1}}$ . Converting with $M_\odot=1.99\times10^{33}\ \mathrm{g}$ and $1\ \mathrm{yr}=3.16\times10^7\ \mathrm{s}$ gives $\mathbf{\dot{M}_B\approx1.3\times10^{-5}\ M_\odot\,yr^{-1}}$ .

Step 5. Compare to Eddington. With $L_{\rm Edd}=1.26\times10^{38}(M/M_\odot)\approx5.0\times10^{44}\ \mathrm{erg\,s^{-1}}$ and efficiency $\eta=0.1$ , the Eddington rate is $\dot{M}_{\rm Edd}=L_{\rm Edd}/(\eta c^2)\approx5.6\times10^{24}\ \mathrm{g\,s^{-1}}\approx0.09\ M_\odot\,\mathrm{yr^{-1}}$ , so $\dot{M}_B/\dot{M}_{\rm Edd}\approx1.4\times10^{-4}$ . Sgr A\* is deeply sub-Eddington: even the generous Bondi estimate is ten-thousand times below Eddington, and the gas that actually reaches the horizon is smaller still — the signature of a radiatively inefficient, advection-dominated flow.

2 · A wandering stellar-mass black hole in the ISM (Bondi–Hoyle–Lyttleton)

A $10\,M_\odot$ black hole drifts at $v_\infty=30\ \mathrm{km\,s^{-1}}$ through warm neutral ISM with $T=10^4\ \mathrm{K}$ , $n=1\ \mathrm{cm^{-3}}$ ( $\mu=1$ , atomic hydrogen). Estimate its accretion rate and quantify how much the motion suppresses it.

Step 1. Warm-ISM sound speed: $c_{s,\infty}=\sqrt{(5/3)(1.38\times10^{-16})(10^4)/(1.67\times10^{-24})}\approx\mathbf{11.7\ km\,s^{-1}}$ , so the black hole moves supersonically ( $v_\infty/c_{s,\infty}\approx2.6$ ).

Step 2. Effective speed $v_{\rm eff}=\sqrt{11.7^2+30^2}\approx\mathbf{32.2\ km\,s^{-1}}$ ; density $\rho_\infty=m_H n=1.67\times10^{-24}\ \mathrm{g\,cm^{-3}}$ ; mass $M=1.99\times10^{34}\ \mathrm{g}$ .

Step 3. BHL rate (equation 8): $\dot{M}_{\rm BHL}=4\pi G^2M^2\rho_\infty/v_{\rm eff}^3\approx1.1\times10^{12}\ \mathrm{g\,s^{-1}}\approx\mathbf{1.8\times10^{-14}\ M_\odot\,yr^{-1}}$ — minuscule, which is why the Galaxy's isolated stellar-mass black holes are so hard to detect.

Step 4. Suppression by motion: the ratio to the stationary rate is $(v_{\rm eff}/c_{s,\infty})^3=(32.2/11.7)^3\approx\mathbf{21}$ . The 30 km/s drift cuts the accretion rate roughly twenty-fold versus an identical object at rest in the same gas — motion, like heat, brakes capture.

Physics of the portrait

Bondi accretion looks inevitable in hindsight and was anything but. The idea that a transonic flow could be selected by a regularity condition — nature picking a single eigenvalue rate from a continuum of candidates — was a genuinely new way of thinking about accretion, and it came from a physicist better known to the public for cosmology.

Origins · 1 minHermann Bondi and how this was discovered

Hermann Bondi

1919–2005 · spherical accretion and theoretical astrophysics

Hermann Bondi was an Austrian-British mathematician and cosmologist whose name attaches to problems across gravitation and fluid dynamics. Born in Vienna in 1919, he came to Cambridge and, interned briefly as an enemy alien during the war, fell in with Thomas Gold and Fred Hoyle — the trio who in 1948 proposed the steady-state theory of the universe. But his 1952 paper "On spherically symmetrical accretion" (MNRAS 112, 195) is the one this article celebrates: building on the moving-accretor work of Hoyle and Lyttleton (1939) and Bondi and Hoyle (1944), he solved the steady spherical inflow of a polytropic gas onto a point mass in full, showing that continuity plus Bernoulli force the flow through a sonic point and that regularity there fixes the accretion rate as an eigenvalue $\lambda_s(\gamma)$ . The result was so clean and general that it became the default estimate for accretion onto everything from protostars to supermassive black holes. Bondi went on to fundamental work in general relativity — Bondi mass, the Bondi–Metzner–Sachs symmetry group, gravitational-wave energy — and to a distinguished career in British science administration. He remained throughout a master of the physically transparent calculation, and Bondi accretion is his most-used monument.

History

1939

— Hoyle and Lyttleton compute gas capture by a star moving through interstellar matter

1944

— Bondi and Hoyle generalize the moving-accretor rate and its downstream wake

1952

— Bondi derives the steady transonic spherical accretion rate and eigenvalue λ_s(γ)

2003

— Chandra resolves the Bondi radius of Sgr A* and measures the hot-gas capture rate

2019

— the Event Horizon Telescope images the accretion-fed shadow of the black hole in M87

Going further

Texts. Bondi, H. (1952), MNRAS 112, 195, "On spherically symmetrical accretion" (the original) · Frank, King & Raine, Accretion Power in Astrophysics, 3rd ed., Ch. 2 (the standard derivation) · Shapiro & Teukolsky, Black Holes, White Dwarfs, and Neutron Stars, Ch. 14 · Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., accretion chapter · Shu, The Physics of Astrophysics, Vol. II: Gas Dynamics, on spherical accretion. Key anchor values: the sonic sound speed satisfies $c_s(r_s)^2=\tfrac{2}{5-3\gamma}c_{s,\infty}^2$ , and the eigenvalue $\lambda_s(\gamma)$ takes the values $\lambda_s(5/3)=1/4$ , $\lambda_s(4/3)=1/\sqrt2\approx0.707$ , $\lambda_s(3/2)=1/2$ , and $\lambda_s\to e^{3/2}/4\approx1.12$ as $\gamma\to1$ . For Sgr A\* the worked example gives $c_{s,\infty}\approx479\ \mathrm{km\,s^{-1}}$ , $r_B\approx0.075\ \mathrm{pc}$ , $\dot{M}_B\approx1.3\times10^{-5}\,M_\odot\,\mathrm{yr^{-1}}$ , and the BHL motion suppression factor is $(v_{\rm eff}/c_{s,\infty})^3\approx21$ .

Prerequisites. Fluid dynamics (Euler equation, continuity) · Polytropic equations of state · Newtonian gravity of a point mass · Sound speed and Mach number. Leads to. Neutron Stars · Pulsar Magnetospheres and Spin-Down · Magnetars.

Ask the tutor: Want to see how angular momentum or radiative feedback bends the Bondi solution away from its clean eigenvalue — and why real black holes fall short of the rate you just computed?

Why does the flow have to pass through a sonic point at all?

Because the two ends live on opposite sides of $M=1$ . Far out the gas drifts subsonically, held up by pressure; deep in the well it falls at nearly the free-fall speed, far above the local sound speed. A steady flow connecting these must cross $u=c_s$ somewhere. The master equation shows the velocity gradient stays finite at that crossing only where the gravity and pressure terms also balance — the sonic point. Crossing anywhere else would give an infinite gradient (a shock or a stall), so smoothness at the sonic point selects the physical solution.

What does it mean to call the accretion rate an "eigenvalue"?

Without an extra condition, continuity and Euler admit a one-parameter family of steady flows — you could imagine dialing the rate up or down. Almost all of those candidates are singular at the sonic radius: their velocity gradient blows up. Only one value of the rate lets the flow thread the sonic point smoothly. Just as a quantum boundary-value problem permits solutions only for discrete energies, this hydrodynamic problem permits a regular transonic solution only for one rate. That selected value is $\dot{M}_B$ , and its $\gamma$ -dependence is packaged in the eigenvalue $\lambda_s$ .

Why is the γ = 5/3 case singular, and how is it handled?

The sonic radius carries a factor $(5-3\gamma)$ that vanishes at $\gamma=5/3$ , formally sending $r_s\to0$ . Physically, a monatomic gas is the stiffest adiabatic case: it heats so strongly on compression that the sonic transition is pushed to the center and the flow is essentially free-fall throughout. You cannot simply substitute $\gamma=5/3$ into the sonic-radius formula. Instead you take the limit carefully — approaching $5/3$ from below — and find that although $r_s\to0$ , the eigenvalue converges cleanly to $\lambda_s=1/4$ , its smallest value. The accretion rate is perfectly well defined; only the intermediate quantity $r_s$ is degenerate.

Is the Bondi radius the same as the sonic radius?

No, though they are proportional. The Bondi radius $r_B=GM/c_{s,\infty}^2$ is an energy scale — inside it the potential well exceeds the gas's thermal energy, so gravity captures the gas. The sonic radius $r_s$ is a dynamical scale — where the accelerating inflow reaches the local sound speed. The transonic analysis relates them as $r_s=\tfrac{5-3\gamma}{4}r_B$ , so the sonic point sits at half $r_B$ for isothermal gas and shrinks to the center as the gas stiffens. Use $r_B$ (always finite) for the capture cross-section; $r_s$ is where the dynamics turn transonic.

Why is the real accretion rate onto black holes usually far below Bondi?

Bondi's model omits everything that regulates real accretion. First, angular momentum: gas in a galaxy or wind almost always carries some, and even a tiny amount forms a centrifugally supported disk that throttles the inflow long before it reaches the accretor. Second, feedback: accretion energy — radiation, jets, winds — heats and pushes back the surrounding gas, raising its sound speed and driving outflows. Third, magnetic fields and turbulence add pressure support the model ignores. So Bondi is a generous upper bound; Sgr A\* accretes far less at the horizon than its Bondi rate, and the gap is itself a diagnostic of the missing physics.

Does the Bondi–Hoyle–Lyttleton rate reduce to the Bondi rate?

Yes, smoothly. The BHL rate replaces the ambient sound speed with the effective speed $(c_{s,\infty}^2+v_\infty^2)^{1/2}$ , combining thermal and bulk motion in quadrature. When the accretor barely moves ( $v_\infty\ll c_{s,\infty}$ ) the effective speed is just $c_{s,\infty}$ and you recover the Bondi rate. When it moves fast ( $v_\infty\gg c_{s,\infty}$ ) the rate becomes the Hoyle–Lyttleton value $\propto v_\infty^{-3}$ , and the flow is no longer spherical — a bow shock and a dense downstream wake form, along which most of the captured gas actually reaches the accretor. The single interpolation formula spans both regimes and is what simulations use for moving sinks.