Circular Orbits and Their Stability

A planet tracing an ellipse around the Sun is, in Newtonian gravity, an eternal promise: pick any radius, supply the matching angular momentum, and a stable circular orbit is yours. Schwarzschild geometry breaks that promise. For a massive test particle on a timelike geodesic the two Killing vectors $\partial_t$ and $\partial_\phi$ hand us a conserved specific energy $E$ and angular momentum $L$ ; feeding these into $u^\mu u_\mu=-1$ collapses the geodesic problem onto a single radial equation that reads exactly like a nonrelativistic particle rolling in an effective potential $V_{\mathrm{eff}}(r)$ . Circular orbits are the flat spots of that potential — the valley floors and hilltops where the radius holds constant. The twist Newtonian mechanics never sees is a single extra term, $-ML^2/r^3$ , an attractive correction that grows faster than the centrifugal barrier it competes with, turning the infinite centrifugal wall into a finite hump. Above a threshold angular momentum every $L$ buys two circular orbits — an outer minimum that is stable and an inner maximum that is not — and as $L$ is lowered the two march toward each other and annihilate. Where they merge, at $r=6M$ with $L=2\sqrt3\,M$ , sits the innermost stable circular orbit, the ISCO: below it no stable circular motion exists. This one radius sets the inner edge of black-hole accretion disks, caps the efficiency of gravitational energy release near $5.7\%$ , and frames the final plunge of a merging binary. This article turns the effective-potential machinery into a census of circular orbits and their stability.

Key takeaways

Circular orbits are extrema of the effective potential $V_{\mathrm{eff}}(r)$ , where the radial velocity and its gradient both vanish; $V_{\mathrm{eff}}'=0$ locates them.
The sign of $V_{\mathrm{eff}}''$ decides stability: a minimum ( $V''>0$ ) is a stable orbit, a maximum ( $V''<0$ ) an unstable one. At a circular orbit $V_{\mathrm{eff}}''=M(r-6M)/[r^3(r-3M)]$ .
Above a critical angular momentum every $L$ gives two orbits — an outer stable minimum and an inner unstable maximum — which merge at the ISCO, $r=6M$ , when $L=2\sqrt3\,M$ .
No circular orbit of any kind exists inside $r=3M$ ; the unstable branch limits to the photon sphere at $3M$ , and the marginally bound orbit ( $E=1$ ) sits at $r=4M$ .
Falling from rest at infinity to the ISCO releases $1-E_{\mathrm{ISCO}}=1-2\sqrt2/3\approx5.72\%$ of the rest-mass energy — the maximum radiative efficiency of a non-spinning black hole.
Newtonian gravity admits a stable circular orbit at every radius; the relativistic $-ML^2/r^3$ term is what destabilizes close orbits and creates the ISCO in the first place.
Key idea: circular orbits are the extrema of $V_{\mathrm{eff}}(r)$ — $V_{\mathrm{eff}}'=0$ locates them, the sign of $V_{\mathrm{eff}}''$ decides stability — and the purely relativistic $-ML^2/r^3$ term turns the centrifugal barrier into a finite hump, producing an inner unstable branch and an innermost stable circular orbit at $r=6M$ .

The core insight

Circular orbits are extrema of $V_{\mathrm{eff}}(r)$ : $V_{\mathrm{eff}}'=0$ locates them and the sign of $V_{\mathrm{eff}}''$ decides stability. The relativistic $-ML^2/r^3$ term turns the would-be Newtonian centrifugal barrier into a finite hump, producing an inner unstable branch and an innermost stable circular orbit at $r=6M$ .

Where this lives today

The ISCO and the circular-orbit census are measured quantities at the astrophysical frontier. In X-ray binaries the disk's inner edge is fitted from the thermal continuum spectrum: the inner radius, identified with the ISCO, gives the black hole's spin, since the ISCO shrinks from $6M$ (non-spinning) toward $M$ for a maximally rotating Kerr hole. High-frequency quasi-periodic oscillations, seen at hundreds of hertz in black-hole and neutron-star systems, are modelled with the frequencies $\Omega$ and $\kappa$ near the ISCO — and the $\sim220\ \mathrm{Hz}$ scale of Worked Example 2 sits in the observed band. Gravitational-wave astronomy makes the ISCO audible: as a compact binary inspirals, the transition from adiabatic sweep to plunge occurs near the system's effective ISCO, imprinting the frequency at which the LIGO–Virgo–KAGRA chirp bends over before merger. And the Event Horizon Telescope's images of M87\ (2019) and Sagittarius A\ (2022) resolve the bright ring set by the unstable photon orbit at $\approx3M$ — the null-geodesic sibling of the massive unstable branch, the same $r_-\to3M$ limit of equation (5). Whether disks truncate exactly at the ISCO, and whether ISCO measurements can test general relativity in the strong field, remain live research programs.

The physical picture

Picture a marble rolling inside a bowl whose cross-section is $V_{\mathrm{eff}}(r)$ . A circular orbit is the marble at rest — at the bottom of the valley (stable) or balanced on a hilltop (unstable). This is no loose analogy: the radial equation is literally $\tfrac12\dot r^2 + V_{\mathrm{eff}}(r) = \tfrac12 E^2$ , the energy equation of a unit-mass particle in one dimension, with $\dot r$ the speed along the bowl. Where $\tfrac12 E^2$ just touches the curve and $V_{\mathrm{eff}}'=0$ , the radius is frozen and the motion is a perfect circle in the equatorial plane. The shape of that bowl is where relativity rewrites Newton. Newtonian gravity has $L^2/2r^2$ diverging faster than $-M/r$ , so the inner wall is infinitely high and there is one valley: a stable circular orbit at every radius. Schwarzschild adds $-ML^2/r^3$ , which climbs faster still as $r\to0$ and overturns the wall — the barrier peaks at finite height and rolls back over. That overhang is a second extremum, a maximum: an unstable circular orbit. The first figure shows the bowl with both features.

Figure

Schwarzschild effective potential V_eff(r) for specific angular momentum L² = 14M² in units G = c = M = 1: an unstable maximum sits at r₋ = 4.35M and a stable minimum at r₊ = 9.65M. The inner hump is the purely relativistic capture barrier, absent in Newtonian gravity, that the −ML²/r³ term creates

Now dial the angular momentum. Crank $L$ up and the stable valley slides outward while the unstable hilltop creeps inward toward $3M$ . Turn $L$ down and the two features glide together, the hump shrinking, until at $L=2\sqrt3\,M$ they meet at $r=6M$ and flatten into a horizontal inflection — a circular orbit stable and unstable at once, the ISCO. Push $L$ lower and both extrema vanish: the potential slides monotonically inward and every particle plunges. The second figure maps this — the radius of each circular orbit versus $L$ , one curve that folds back on itself at the ISCO.

Figure

Circular-orbit radii r±(L) in Schwarzschild geometry: for every L greater than 2√3 M two orbits exist — an outer stable branch r₊ and an inner unstable branch r₋. The two branches merge at the ISCO, r = 6M, when L = 2√3 M ≈ 3.46M; the unstable branch descends toward the photon sphere at r = 3M as L → ∞

The core equations

Everything begins with the radial equation of a timelike geodesic. The normalization $u^\mu u_\mu=-1$ , evaluated in the equatorial plane with the conserved $E$ and $L$ replacing $\dot t$ and $\dot\phi$ , leaves one first-order equation for $\dot r$ — the master equation of Schwarzschild orbital dynamics.

\Big(\frac{\mathrm{d}r}{\mathrm{d}\tau}\Big)^2 = E^2 - \Big(1-\frac{2M}{r}\Big)\Big(1+\frac{L^2}{r^2}\Big) \qquad (1)

where: $\tau$ — proper time; $E$ — conserved specific energy, $E=(1-2M/r)\,\mathrm{d}t/\mathrm{d}\tau$ ; $L$ — conserved specific angular momentum, $L=r^2\,\mathrm{d}\phi/\mathrm{d}\tau$ ; $M$ — geometric mass of the source ( $G=c=1$ , so $M$ has units of length). Read it as an energy balance: the right side is what is left for radial motion after the "potential" is paid. Grouping the terms as a one-dimensional energy equation makes the potential explicit. The factor $\tfrac12$ makes it match the Newtonian $\tfrac12 v^2 + V = E$ template, licensing the marble-in-a-bowl intuition.

\tfrac12\Big(\frac{\mathrm{d}r}{\mathrm{d}\tau}\Big)^2 + V_{\mathrm{eff}}(r) = \tfrac12 E^2,\qquad V_{\mathrm{eff}}(r)=\tfrac12\Big(1-\frac{2M}{r}\Big)\Big(1+\frac{L^2}{r^2}\Big) \qquad (2)

where: $V_{\mathrm{eff}}$ — effective potential per unit rest mass. The quantity playing the role of total energy is $\tfrac12 E^2$ , not $E$ — a bookkeeping point that matters when comparing energies. Multiplying out exposes the four terms that decide everything. Three are Newtonian; the fourth is the fingerprint of general relativity.

V_{\mathrm{eff}}(r)=\tfrac12 - \frac{M}{r} + \frac{L^2}{2r^2} - \frac{ML^2}{r^3} \qquad (3)

where: $\tfrac12$ — rest-energy offset; $-M/r$ — Newtonian attraction; $L^2/2r^2$ — centrifugal barrier; $-ML^2/r^3$ — the purely relativistic attractive term, the destabilizer with no Newtonian analogue, dominant at small $r$ . A circular orbit needs both $\dot r=0$ and $\ddot r=0$ , so $r$ must sit at an extremum of $V_{\mathrm{eff}}$ . Setting the derivative to zero and clearing powers of $r$ gives a clean quadratic.

V_{\mathrm{eff}}'(r)=0 \;\Longleftrightarrow\; Mr^2 - L^2 r + 3ML^2 = 0 \qquad (4)

where: $r$ — areal radius of a circular orbit. The quadratic has at most two positive roots — inner and outer orbits — whose existence hinges on its discriminant. Solving the quadratic gives the two radii — inner root the unstable maximum, outer root the stable minimum. The square root is real only when $L$ is large enough.

r_\pm = \frac{L^2 \pm \sqrt{L^4 - 12 M^2 L^2}}{2M} = \frac{L^2 \pm L\sqrt{L^2-12M^2}}{2M} \qquad (5)

where: $r_+$ — outer, stable orbit ( $+$ root); $r_-$ — inner, unstable orbit ( $-$ root). As $L\to\infty$ : $r_+\to\infty$ , $r_-\to3M$ . Stability is settled by the second derivative. Evaluated on the circular condition (using $L^2=Mr^2/(r-3M)$ to eliminate $L$ ), the messy expression collapses to a strikingly simple closed form whose sign is read off by inspection.

V_{\mathrm{eff}}''(r_c) = \frac{M\,(r_c-6M)}{r_c^{\,3}\,(r_c-3M)} \qquad (6)

where: $r_c$ — radius of the orbit tested. For $r_c>6M$ both factors are positive, $V''>0$ , a stable minimum; for $3M<r_c<6M$ the numerator flips sign, $V''<0$ , an unstable maximum. At $r_c=6M$ it vanishes — the marginal case. The threshold is where the discriminant in equation (5) hits zero: the roots coincide, minimum and maximum merge, and stability is lost — the ISCO.

L^2_{\mathrm{crit}} = 12M^2,\qquad L_{\mathrm{crit}}=2\sqrt{3}\,M,\qquad r_{\mathrm{ISCO}}=6M \qquad (7)

where: $L_{\mathrm{crit}}$ — smallest specific angular momentum admitting any circular orbit; $r_{\mathrm{ISCO}}=L_{\mathrm{crit}}^2/2M=6M$ . Eliminating $L$ between the circular condition and $\tfrac12 E^2 = V_{\mathrm{eff}}$ gives $E$ and $L$ as functions of radius alone. Both diverge as $r\to3M$ , the photon-sphere signature.

E^2(r)=\frac{(1-2M/r)^2}{1-3M/r},\qquad L^2(r) = \frac{M r^2}{r - 3M} \qquad (8)

where: the shared denominator $r-3M$ vanishes at the photon sphere, forcing $E,L\to\infty$ — no finite-energy massive particle can orbit at $3M$ . Evaluating the energy at the ISCO gives the most quoted number in accretion physics: the binding energy per unit rest mass, and hence the efficiency ceiling of a non-spinning black hole.

E_{\mathrm{ISCO}}=\frac{2\sqrt2}{3}\approx 0.9428,\qquad \eta = 1-E_{\mathrm{ISCO}}\approx 5.72\% \qquad (9)

where: $\eta$ — fraction of rest-mass energy released by matter spiralling quasi-statically from rest at infinity to the ISCO. Compare hydrogen fusion at $\approx0.7\%$ : gravity here is roughly eight times more efficient. Finally, two frequencies connect the stable orbits to observation. The orbital angular velocity in coordinate time obeys a relation identical in form to Kepler's third law; the radial epicyclic frequency carries the ISCO signature in a factor vanishing at $6M$ .

\Omega^2 = \frac{M}{r^3},\qquad \kappa^2 = \Omega^2\Big(1-\frac{6M}{r}\Big) \qquad (10)

where: $\Omega=\mathrm{d}\phi/\mathrm{d}t$ — orbital angular velocity (coordinate time); $\kappa$ — radial epicyclic frequency. That $\Omega^2=M/r^3$ holds exactly in Schwarzschild coordinates is a famous coincidence; $\kappa<\Omega$ makes bound orbits precess, and $\kappa\to0$ at $r=6M$ signals the ISCO.

Derivation AFrom geodesic normalization to the circular-orbit quadratic

The goal of this first derivation is to build the effective potential from the ground up and reduce the search for circular orbits to a quadratic. We start from the one algebraic constraint every massive particle obeys — its four-velocity has unit timelike norm — feed in the two conserved quantities the Schwarzschild Killing vectors supply, and reorganize into the one-dimensional energy form. Then the circular-orbit condition, that the radius sit at an extremum, turns the calculus into a polynomial we can solve by hand.

Step 1. Write the normalization $g_{\mu\nu}u^\mu u^\nu=-1$ for equatorial motion ( $\theta=\pi/2$ ), giving $-(1-2M/r)\dot t^2 + (1-2M/r)^{-1}\dot r^2 + r^2\dot\phi^2 = -1$ , with dots $\mathrm{d}/\mathrm{d}\tau$ . Why: the norm ties the time, radial, and angular rates together, so it is the natural place to eliminate everything except $r$ .

Step 2. Substitute the conserved quantities $\dot t = E/(1-2M/r)$ and $\dot\phi = L/r^2$ from the Killing vectors $\partial_t$ and $\partial_\phi$ . Why: replacing the rates by constants of motion removes two degrees of freedom, leaving a single equation in $r$ and $\dot r$ .

Step 3. Multiply through by $-(1-2M/r)$ to isolate $\dot r^2 = E^2 - (1-2M/r)(1+L^2/r^2)$ , then divide by two to read off $V_{\mathrm{eff}}=\tfrac12(1-2M/r)(1+L^2/r^2)$ . Why: this casts a curved-spacetime geodesic as a flat one-dimensional energy problem — exactly, not approximately.

Step 4. Expand $V_{\mathrm{eff}}=\tfrac12 - M/r + L^2/2r^2 - ML^2/r^3$ and impose $V_{\mathrm{eff}}'=0$ , i.e. $M/r^2 - L^2/r^3 + 3ML^2/r^4 = 0$ ; multiplying by $r^4$ gives $Mr^2 - L^2 r + 3ML^2 = 0$ . Why: an extremum is the only place a particle sits at constant radius, and clearing denominators turns that condition into a plain quadratic. *As a check: expanding $\tfrac12(1-2M/r)(1+L^2/r^2)$ directly reproduces equation (3), and multiplying $V_{\mathrm{eff}}'=0$ through by $r^4$ yields $Mr^2-L^2r+3ML^2=0$ .*

Go deeper · Derivation B · 1 minSolving for the orbit radii, the discriminant, and the ISCO

Derivation BSolving for the orbit radii, the discriminant, and the ISCO

Having reduced circular orbits to the quadratic $Mr^2 - L^2 r + 3ML^2 = 0$ , the aim here is to extract the two radii, determine when they exist, and locate the ISCO as the boundary case where the solutions collide. The key is the discriminant: its sign counts the real roots, and its vanishing is the physical event of the stable and unstable orbits merging into one marginally stable orbit. We confirm with the closed-form second derivative which branch is which.

Step 1. Apply the quadratic formula to $Mr^2 - L^2 r + 3ML^2 = 0$ : $r_\pm = (L^2 \pm \sqrt{L^4-12M^2L^2})/2M = (L^2 \pm L\sqrt{L^2-12M^2})/2M$ . Why: solving for $r$ makes both circular radii explicit functions of $L$ , which we need to track as $L$ varies.

Step 2. Require a nonnegative discriminant: $L^4-12M^2L^2\ge0\Rightarrow L^2\ge12M^2$ . For $L^2<12M^2$ the roots are complex and no circular orbit exists at any radius. Why: the discriminant is the gatekeeper, converting "when can a particle circle a black hole?" into the sharp inequality $L\ge2\sqrt3\,M$ .

Step 3. At threshold $L^2=12M^2$ the root vanishes and the two orbits coincide at $r = L^2/2M = 6M$ — the ISCO. Why: the merging of a maximum and a minimum is a fold, a horizontal inflection of marginal stability, fixed with no free parameters at $6M$ .

Step 4. Confirm the branches with equation (6): $V_{\mathrm{eff}}''(r_c)=M(r_c-6M)/[r_c^3(r_c-3M)]$ is positive for $r_+>6M$ (stable minimum), negative for $3M<r_-<6M$ (unstable maximum), and $r_-\to3M$ as $L\to\infty$ . Why: the sign of the closed form is the cleanest stability test, and it shows the inner branch is always unstable, sliding toward the photon sphere. *As a check: $L^2=12M^2$ gives $r=6M$ ; at $L^2=14M^2$ the roots are $r_\pm=(7\pm\sqrt7)M=9.646M,\,4.354M$ , with $V_{\mathrm{eff}}''(r_+)>0$ and $V_{\mathrm{eff}}''(r_-)<0$ , confirming the outer orbit is stable and the inner unstable.*

Go deeper · Derivation C · 1 minEnergy, angular momentum, and the binding energy at the ISCO

Derivation CEnergy, angular momentum, and the binding energy at the ISCO

This final derivation converts the geometry of circular orbits into the energetics astrophysics measures. The plan is to eliminate $L$ between the circular condition and the energy relation, producing $E(r)$ and $L(r)$ as functions of radius alone, then evaluate the energy at the ISCO to obtain the binding energy — the accretion efficiency. The same $E(r)$ relation, set to one, locates the marginally bound orbit separating capturable from escapable trajectories.

Step 1. Solve the circular quadratic for $L$ : $Mr^2 - L^2 r + 3ML^2 = 0\Rightarrow L^2(r-3M)=Mr^2\Rightarrow L^2 = Mr^2/(r-3M)$ . Why: expressing $L$ as a function of radius removes it as an independent variable, so energies tabulate against radius directly.

Step 2. On a circular orbit $\dot r=0$ , so $\tfrac12 E^2 = V_{\mathrm{eff}}$ gives $E^2 = (1-2M/r)(1+L^2/r^2)$ . Substitute $L^2/r^2 = M/(r-3M)$ and use $1+M/(r-3M)=(r-2M)/(r-3M)$ to get $E^2 = (1-2M/r)^2/(1-3M/r)$ . Why: this compresses the two-parameter orbit family into one curve $E(r)$ whose features carry all the physics.

Step 3. Evaluate at $r=6M$ : $E^2 = (1-1/3)^2/(1-1/2) = (2/3)^2/(1/2) = 8/9$ , so $E_{\mathrm{ISCO}}=2\sqrt2/3\approx0.9428$ and the binding energy is $1-E_{\mathrm{ISCO}}\approx5.72\%$ . Why: the ISCO is the last stable orbit, so its energy is the maximum a slowly-accreting particle radiates before plunging — the ceiling on Schwarzschild efficiency.

Step 4. Set $E=1$ : $(1-2M/r)^2 = 1-3M/r$ expands to $4M^2/r^2 = M/r$ , giving $r_{\mathrm{mb}}=4M$ , and $L^2=M(4M)^2/(4M-3M)=16M^2$ so $L_{\mathrm{mb}}=4M$ . Why: the marginally bound orbit divides bound ( $E<1$ ) from escapable ( $E\ge1$ ), pinning that boundary to the unstable branch at exactly $4M$ . *As a check: $E^2=8/9$ at $r=6M$ gives $E_{\mathrm{ISCO}}=2\sqrt2/3=0.94281$ and efficiency $0.05719$ ; setting $E=1$ in the energy relation yields $r=4M$ and $L=4M$ for the marginally bound orbit.*

Limiting cases

The orbit structure is controlled entirely by the specific angular momentum $L$ (in units of $M$ ). The four regimes trace the potential from the Newtonian limit to the strong-field boundary.

Case	Result	Meaning
Newtonian limit ( $r\gg M$ )	$V_{\mathrm{eff}}\to \tfrac12-\dfrac{M}{r}+\dfrac{L^2}{2r^2}$	The $-ML^2/r^3$ term is negligible; only the centrifugal barrier survives, giving a single stable circular orbit at every radius $r=L^2/M$ and no ISCO.
Critical angular momentum	$L=2\sqrt3\,M\;\Rightarrow\;r=6M$	Maximum and minimum merge into a horizontal inflection — the ISCO. For $L<2\sqrt3\,M$ there are no circular orbits at all and every particle plunges.
Marginally bound orbit	$E=1,\;L=4M,\;r=4M$	Separates bound orbits ( $E<1$ ) from those that can escape to infinity ( $E\ge1$ ); the inner unstable orbit at $E=1$ marks the boundary of gravitational capture.
Large- $L$ / photon limit	$r_-\to3M\;\;(L\to\infty)$	The unstable circular orbit approaches the photon sphere; no circular orbit of any kind, massive or null, exists inside $r=3M$ .

Common trapEvery extremum of V_eff is a stable minimum — the Newtonian reflex

The most common error, imported from Kepler, is to assume a circular orbit is automatically stable. In Newtonian gravity $V_{\mathrm{eff}}=-M/r+L^2/2r^2$ has one extremum, always a minimum, so the reflex is harmless there. In Schwarzschild it is wrong: for every $L>2\sqrt3\,M$ there are two circular orbits, and the inner one is an unstable maximum. Two further traps: a circular orbit requires $V_{\mathrm{eff}}'=0$ and $\tfrac12 E^2=V_{\mathrm{eff}}$ together (the energy condition alone gives only a turning point); and "bound" and "stable" are independent — the inner orbit can have $E<1$ yet sit on a hilltop, so for $L^2=14M^2$ both orbits are bound while only the outer is stable. Never read stability off the energy.

Worked examples

1 · Two orbits from one angular momentum (fully worked)

A test particle has $L^2=14M^2$ . Find the radii of its circular orbits, classify each as stable or unstable, and compute the specific energy of each. Units $M=1$ .

Step 1. Existence: $L^2=14>12$ , so the discriminant $L^2-12M^2=2M^2>0$ and two orbits exist. From equation (5): $r_\pm=(14\pm\sqrt{14}\cdot\sqrt{2})/2 = (14\pm\sqrt{28})/2 = (7\pm\sqrt7)\,M$ , i.e. $r_+=9.65M$ and $r_-=4.35M$ .

Step 2. Classify with equation (6). For $r_+=9.65M>6M$ the numerator $r_+-6M>0$ , so $V_{\mathrm{eff}}''>0$ : a stable minimum. For $r_-=4.35M$ in $3M<r_-<6M$ the numerator is negative, $V_{\mathrm{eff}}''<0$ : an unstable maximum.

Step 3. Energies from equation (8), $E^2=(1-2/r)^2/(1-3/r)$ . Outer: $E^2=(0.7927)^2/0.6890=0.9119$ , so $E_+=0.955$ . Inner: $E^2=(0.5407)^2/0.3110=0.9399$ , so $E_-=0.969$ .

Step 4. Both energies are below one, so both orbits are gravitationally bound — yet only the outer is dynamically stable. The inner is the textbook trap: bound but perched on a hilltop, so any radial nudge sends it plunging or flinging outward. Result: $r_+=9.65M$ (stable, $E=0.955$ ), $r_-=4.35M$ (unstable, $E=0.969$ ); both bound.

2 · The ISCO of a stellar-mass black hole (fully worked)

A non-spinning black hole of mass $M=10\,M_\odot$ accretes through a thin disk. Find the ISCO radius, the orbital frequency there, and the maximum radiative efficiency. Use $GM_\odot/c^2=1.477\ \mathrm{km}$ , $GM_\odot/c^3=4.925\times10^{-6}\ \mathrm{s}$ .

Step 1. ISCO radius. In geometric units $r_{\mathrm{ISCO}}=6M$ ; converting, $r_{\mathrm{ISCO}}=6\times(GM/c^2)=6\times1.477\ \mathrm{km}\times10 = \mathbf{88.6\ km}$ — smaller than a mid-sized city.

Step 2. Orbital frequency from equation (10): $\Omega^2=M/r^3$ at $r=6M$ gives $\Omega = 1/(6^{3/2}M) = 1/(14.70\,M)$ . With $M=10\times GM_\odot/c^3 = 4.925\times10^{-5}\ \mathrm{s}$ , $\Omega = 1381\ \mathrm{rad/s}$ , so $f=\Omega/2\pi=\mathbf{220\ Hz}$ — matter at the ISCO laps the hole 220 times a second.

Step 3. Radiative efficiency from equation (9): $\eta = 1-E_{\mathrm{ISCO}} = 1-2\sqrt2/3 = \mathbf{5.72\%}$ . Each kilogram spiralling to the ISCO can radiate up to $0.0572\,c^2\approx5.1\times10^{15}\ \mathrm{J}$ — roughly eight times the $\sim0.7\%$ yield of hydrogen fusion, which is why accreting black holes are the most efficient steady engines known.

Step 4. Check the epicyclic frequency: $\kappa^2=\Omega^2(1-6M/r)$ at $r=6M$ gives $\kappa=0$ — radial oscillations become infinitely slow at the ISCO, the orbit marginally stable. Result: $r_{\mathrm{ISCO}}=88.6$ km, $f\approx220$ Hz, $\eta\approx5.72\%$ , $\kappa\to0$ .

Physics of the portrait

The innermost stable circular orbit is the sharpest prediction of strong-field gravity that ordinary matter obeys — a hard radius at $6M$ , below which no stable orbit exists.

Origins · 1 mininnermost stable circular orbit and how this was discovered

innermost stable circular orbit

the marginal orbit r = 6M where V_eff stability fails · strong-field geodesic dynamics

The innermost stable circular orbit is a concept with no single inventor and a slow-burning history. Karl Schwarzschild wrote down the exterior vacuum metric in 1916, within weeks of Einstein's field equations and while serving on the Russian front, but did not chart its orbits; Johannes Droste, working independently in Leiden the same year, derived the same solution and took the first steps into its trajectories. The complete classification of timelike geodesics — bound, plunging, and circular orbits, and the recognition that stable circular motion terminates at a finite radius — was carried out by the Japanese astronomer Yusuke Hagihara in a monumental 1931 paper that stood as definitive for decades. For a long time the $6M$ radius was a mathematical curiosity, since no known object was compact enough to have an orbit at six times its mass radius. That changed in the early 1970s: Bardeen, Press and Teukolsky tabulated the circular-orbit structure in 1972, and Novikov and Thorne, alongside Shakura and Sunyaev, built the standard thin-disk model in 1973 with the ISCO as the disk's inner edge and $1-2\sqrt2/3$ as its efficiency ceiling. What began as an oddity in Hagihara's tables is now measured — in the thermal continuum of X-ray binaries, in quasi-periodic oscillations, and, since 2019, in the ring of light imaged by the Event Horizon Telescope.

History

1916

— Schwarzschild derives the exterior vacuum solution of Einstein's equations

1916

— Droste independently obtains the same metric and examines particle orbits

1931

— Hagihara completes the classification of timelike geodesics in Schwarzschild

1972

— Bardeen, Press and Teukolsky tabulate circular-orbit and ISCO properties

1973

— Novikov and Thorne place the accretion-disk inner edge at the ISCO

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §25.5 · Wald, General Relativity (1984), §6.3 · Carroll, Spacetime and Geometry (2004), §5.4 · Hartle, Gravity (2003), Ch. 9 · Schutz, A First Course in General Relativity (2009), §11.1. The key results in units $G=c=1$ are: $Mr^2-L^2r+3ML^2=0$ from $V_{\mathrm{eff}}'=0$ ; $L^2\ge12M^2$ giving $r_{\mathrm{ISCO}}=6M$ ; $E^2=(1-2M/r)^2/(1-3M/r)$ yielding $E_{\mathrm{ISCO}}=2\sqrt2/3\approx0.9428$ and efficiency $\approx5.72\%$ ; marginally bound orbit at $r=4M$ , $L=4M$ , $E=1$ ; and for $L^2=14M^2$ , $r_\pm=(7\pm\sqrt7)M$ . The closed form $V_{\mathrm{eff}}''(r_c)=M(r_c-6M)/[r_c^3(r_c-3M)]$ and the frequencies $\Omega^2=M/r^3$ , $\kappa^2=\Omega^2(1-6M/r)$ are standard results in the QPO literature.

Prerequisites. The Schwarzschild Metric · Killing Vectors and Conserved Quantities in Schwarzschild · Geodesic Equations in Schwarzschild Spacetime · Effective Potential for Massive Particle Orbits. Leads to. Innermost Stable Circular Orbit (ISCO) · Photon Orbits and the Photon Sphere · Perihelion Precession in Schwarzschild Geometry.

Ask the tutor: Ask me to walk you through why the ISCO sits at exactly $6M$ , or to work out how the picture shifts for a spinning Kerr black hole.

Why does the effective-potential picture work at all in curved spacetime?

Because the two Killing symmetries do all the heavy lifting. Time-translation and rotational symmetry give the conserved $E$ and $L$ , and the normalization $u^\mu u_\mu=-1$ is one more constraint. Together they reduce the four geodesic equations to a single first-order equation for $\dot r$ , with exactly the form $\tfrac12\dot r^2 + V_{\mathrm{eff}} = \tfrac12 E^2$ of a nonrelativistic particle in a potential. So the whole toolkit of one-dimensional mechanics — turning points, extrema, stability by second derivative — applies with full rigor. It is not an approximation but an exact rewriting of the geodesic problem, valid down to $r=2M$ .

Is the effective potential the same for photons?

No — the structure differs crucially. For null geodesics the normalization is $u^\mu u_\mu=0$ instead of $-1$ , so the "1" in $(1+L^2/r^2)$ becomes "0" and the potential loses its Newtonian $-M/r$ well. The photon potential $V=\tfrac12(1-2M/r)L^2/r^2$ has a single extremum, a maximum at $r=3M$ — the photon sphere — always unstable. So photons have no stable circular orbits, only the one unstable orbit at $3M$ . That is the radius the massive unstable branch approaches as $L\to\infty$ : infinitely energetic particles behave like photons, which is why $r_-\to3M$ in equation (5).

Can a stable circular orbit exist just outside the ISCO, say at r = 6.01M?

Yes. The ISCO is the innermost stable orbit, not the only one. Every radius $r>6M$ hosts a stable circular orbit, each with its own $L^2=Mr^2/(r-3M)$ and energy from equation (8); moving outward they grow more Newtonian and more robust. The special thing about $6M$ is that it is the boundary: there the restoring "spring constant" $V_{\mathrm{eff}}''$ passes through zero, so the orbit is marginally stable, and for any $r<6M$ it is negative — perturbations grow rather than oscillate. So $6.01M$ is stable, $5.99M$ is not.

Why is the inner circular orbit unstable even though it is bound?

Because "bound" and "stable" answer different questions. Bound is about energy: an orbit with $E<1$ lacks the energy to reach infinity. Stable is about dynamics: an orbit is stable only at a minimum of $V_{\mathrm{eff}}$ , where a radial nudge produces a restoring force and the particle oscillates. The inner orbit sits at a maximum — a hilltop — so the same nudge produces an anti-restoring force and the deviation grows. It can be bound (Worked Example 1 shows both orbits at $E<1$ ) while dynamically doomed. A marble on a hilltop inside a walled yard is trapped in the yard yet cannot stay on the hilltop; that is the inner orbit.

How does black-hole spin change this whole story?

Spin lowers the ISCO. The analysis here is for Schwarzschild — non-rotating — where the ISCO is locked at $6M$ and the efficiency at $5.72\%$ . For a Kerr black hole the ISCO depends on spin and orbit direction: prograde orbits shrink it from $6M$ toward $M$ as spin approaches maximal, while retrograde orbits push it out to $9M$ . Because binding energy grows as the orbit moves inward, the maximum efficiency rises from $5.7\%$ (Schwarzschild) to about $42\%$ (maximal Kerr). Measuring the ISCO radius from the disk spectrum is thus a main way to estimate black-hole spin, which is why continuum-fitting and iron-line methods target that inner edge.

What happens physically to a particle that crosses inside the ISCO?

It plunges. Just inside $6M$ there is no circular orbit to fall onto and no restoring force to hold a radius — the potential slopes inward, so a particle with the local orbital angular momentum accelerates toward the horizon on a near-radial path, crossing $r=2M$ in finite proper time. In an accretion disk this is what happens at the inner edge: viscous torques nudge gas across the ISCO, and inside it the gas spirals in almost ballistically, radiating little further. This is why the ISCO is the disk's inner edge and why the energy released up to the ISCO sets the efficiency — essentially all the useful radiation is extracted before the plunge, not during it.