The Innermost Stable Circular Orbit (ISCO)

The cliff edge of gravity

Drop a ball into a bowl and it rolls to the bottom and settles. Push it gently and it rocks back to the centre — the orbit is stable. Around a star or a planet, every circular orbit behaves like that bowl: the gravitational pull inward and the centrifugal tendency to fly outward balance at one radius, and small disturbances simply oscillate around it. This is why the planets have circled the Sun for four and a half billion years without spiralling in or wandering off.

Close to a black hole, the bowl runs out. General relativity adds a new attractive term to the gravitational pull that grows faster than the centrifugal push as you move inward. At a certain radius the bottom of the bowl flattens into a ledge, and just inside it the ground simply falls away. There is no longer a place to settle. A particle that strays inside that radius does not orbit, does not slow, does not get a second chance — it plunges to the horizon. That radius is the innermost stable circular orbit, or ISCO, and it is one of the cleanest fingerprints that general relativity leaves on the observable universe.

The physics: where the effective potential gives out

The cleanest way to see the ISCO is through the effective potential for a massive test particle in the Schwarzschild geometry. Writing radius in units of the gravitational radius r_g = GM/c² and specific angular momentum as ℓ = L/(M_particle c r_g), the effective potential per unit mass is

V_eff(r) = (1 − 2/r) (1 + ℓ²/r²)        [Schwarzschild, geometrised units]

Expanding it reveals the crucial difference from Newton:

V_eff(r) ≈ 1 − 2/r  +  ℓ²/r²  −  2ℓ²/r³
            (rest) (gravity) (centrifugal) (the GR term)

The Newtonian centrifugal barrier (+ℓ²/r²) always diverges to +∞ as r → 0, guaranteeing a stable orbit for any angular momentum. The relativistic correction (−2ℓ²/r³) falls off faster and eventually beats it. Circular orbits occur where dV_eff/dr = 0; a circular orbit is stable only where the potential is also a minimum, d²V_eff/dr² > 0. Setting both the first and second derivatives to zero simultaneously locates the inflection point — the marginally stable orbit. For Schwarzschild that condition solves exactly to

r_ISCO = 6 GM/c²   = 3 r_s   (Schwarzschild radius r_s = 2 GM/c²)

Outside 6 GM/c² there are two circular orbits for a given ℓ — a stable minimum and an unstable maximum. As you approach the ISCO these two merge. Inside it the potential has no extremum at all: it slopes monotonically down to the horizon, and circular motion is impossible. The specific angular momentum of the orbit reaches its minimum at the ISCO; you cannot make a particle orbit closer no matter how you tune its angular momentum, which is why "innermost" is literal.

Spin changes everything: the Kerr ISCO

Real astrophysical black holes rotate, often very fast. A rotating (Kerr) black hole drags spacetime around with it — frame dragging — so the ISCO splits into two values: a smaller one for matter orbiting in the same sense as the spin (prograde) and a larger one for matter orbiting against it (retrograde). The exact formula, derived by James Bardeen, William Press and Saul Teukolsky in 1972, uses the dimensionless spin a* = cJ/(GM²), which runs from −1 (maximal retrograde) through 0 (Schwarzschild) to +1 (maximal prograde):

Z1 = 1 + (1 − a*²)^(1/3) [ (1 + a*)^(1/3) + (1 − a*)^(1/3) ]
Z2 = √(3 a*² + Z1²)
r_ISCO / (GM/c²) = 3 + Z2 ∓ √[ (3 − Z1)(3 + Z1 + 2 Z2) ]     (− prograde, + retrograde)

Spin a*	Sense	ISCO (GM/c²)	Efficiency η	Comment
+1	Prograde (max)	1.000	0.423 (42 %)	ISCO coincides with horizon in B–L coords
+0.998	Prograde	1.237	0.321 (32 %)	Thorne spin limit from photon capture
+0.9	Prograde	2.321	0.156 (16 %)	Typical rapidly spinning AGN
0	None	6.000	0.057 (5.7 %)	Schwarzschild, the reference case
−0.9	Retrograde	8.717	0.039 (3.9 %)	Counter-rotating accretion
−1	Retrograde (max)	9.000	0.038 (3.8 %)	Largest possible ISCO

Three numbers are worth memorising: the ISCO is 6 GM/c² for no spin, drops to 1 GM/c² for a maximally prograde hole, and rises to 9 GM/c² for a maximally retrograde one. Because the ISCO sets how deep into the gravitational well matter falls before plunging, spin is the single biggest lever on a black hole's power output. Kip Thorne showed in 1974 that photons emitted by the disk and captured by the hole impose an effective spin ceiling of a* ≈ 0.998 for matter-fed accretion, so the practical efficiency limit is closer to 32 % than the theoretical 42 %.

The key numbers in real units

The gravitational radius converts the dimensionless formulas into kilometres: r_g = GM/c² ≈ 1.48 km × (M/M☉). So:

Object	Mass	r_g	Schwarzschild ISCO (6 r_g)	Orbital period there
Stellar black hole (Cygnus X-1)	21 M☉	31 km	≈ 186 km	≈ 9 ms
"Typical" stellar BH	10 M☉	14.8 km	≈ 89 km	≈ 4 ms
Intermediate-mass BH	10³ M☉	1 480 km	≈ 8 900 km	≈ 0.5 s
Sagittarius A*	4.3 × 10⁶ M☉	6.4 × 10⁶ km	≈ 3.8 × 10⁷ km	≈ 33 min
M87* (EHT target)	6.5 × 10⁹ M☉	9.6 × 10⁹ km	≈ 5.8 × 10¹⁰ km	≈ 34 days

For Sagittarius A* the ISCO lies at about 38 million kilometres, roughly a quarter of the Earth–Sun distance, yet a parcel of gas there is whipping around at close to half the speed of light. At the Schwarzschild ISCO the orbital speed measured locally by a static observer is exactly v = c/2 = 0.5 c (the coordinate quantity rΩ = √(M/r) is the smaller value 1/√6 ≈ 0.408 c). Combining the gravitational and special-relativistic effects, an orbiting clock runs slow by the factor dτ/dt = √(1 − 3GM/rc²) = √(1/2) ≈ 0.707 — clocks riding the ISCO tick at about 71 % of the rate of clocks at infinity. For a maximally spinning hole, matter at the prograde ISCO is moving and dilated far more extremely still.

From ISCO to luminosity: the efficiency it sets

The binding energy a particle has shed by the time it reaches the ISCO is the maximum energy an accretion disk can radiate per unit rest mass. The specific energy of a circular orbit at the ISCO, divided into the rest energy, gives the radiative efficiency η. For Schwarzschild,

E_ISCO/mc² = √(8/9) = 0.9428
η = 1 − √(8/9) = 0.0572     →   about 5.7 % of rest mass radiated

That is already eight times more efficient than hydrogen fusion (0.7 %). For a maximally prograde Kerr hole the orbiting matter retains only √(1/3) of its rest energy at the ISCO, so

E_ISCO/mc² = 1/√3 = 0.5774
η = 1 − 1/√3 = 0.4226       →   up to 42 % of rest mass radiated

This is why the ISCO matters far beyond geometry: it is the quantity that turns a black hole into the most efficient engine in the universe. Accretion onto a fast-spinning hole liberates more than fifty times the energy per gram that the Sun's core fusion does, which is how quasars outshine their entire host galaxies from a region no bigger than the solar system.

Worked example: the ISCO of Cygnus X-1

Cygnus X-1, the first widely accepted stellar black hole, has a mass of about 21 M☉ and a near-maximal measured spin of a* ≈ 0.95. Let us find its ISCO and the energy a gram of gas radiates falling to it.

Step 1 — gravitational radius.

r_g = GM/c² = 1.48 km × 21 = 31.1 km

Step 2 — ISCO radius from the Bardeen–Press–Teukolsky formula at a* = 0.95. Evaluating the Z-factors gives Z1 ≈ 1.745, Z2 ≈ 2.398, and the prograde root:

r_ISCO ≈ 1.94 GM/c²  ≈ 1.94 × 31.1 km  ≈ 60 km

So the disk's inner edge is at roughly 60 km — barely twice the gravitational radius, and only about a third of the 186 km the same 21 M☉ hole would have if it were non-spinning (6 × 31.1 km), illustrating how strongly spin pulls the edge inward.

Step 3 — radiative efficiency. Interpolating between the table values, a* = 0.95 gives η ≈ 0.19, i.e. about 19 % of rest mass.

Energy from 1 g falling to the ISCO:
E = η m c² = 0.19 × (10⁻³ kg) × (3 × 10⁸ m/s)²
  ≈ 1.7 × 10¹³ J  ≈ 4 kilotons of TNT, from a single gram

Cygnus X-1 captures a fraction of the dense wind from its blue-supergiant companion, and only a part of that captured material settles into the disk and accretes. At an accreted rate of a few × 10⁻⁹ M☉/yr and η ≈ 0.19 this sustains the observed steady X-ray luminosity of about 2 × 10³⁷ erg/s — bright enough to have flagged the source in the very first rocket-borne X-ray surveys of the 1960s.

How the ISCO is observed

The ISCO is not a thing you can photograph directly, but it leaves three measurable imprints, all of which let astronomers infer black hole spin:

• Continuum fitting. A thin accretion disk truncated at the ISCO radiates a thermal multi-temperature blackbody whose total luminosity and peak temperature depend on the inner-edge radius. With independent measurements of the hole's mass, distance and disk inclination, the disk's apparent inner radius solves for the ISCO, hence the spin. This method gave the spins of Cygnus X-1, GRS 1915+105 (a* > 0.98) and LMC X-3.
• Relativistic iron Kα reflection. Hard X-rays from the corona illuminate the inner disk and fluoresce a 6.4 keV iron line. General-relativistic effects — Doppler boosting, transverse redshift, and gravitational redshift that grows toward the hole — smear this narrow line into a broad, skewed profile with an extended red wing. The wing reaches lower energies the closer the emitting gas sits, so fitting it pins the inner radius to the ISCO. MCG−6-30-15, observed by ASCA in 1995 and later by XMM-Newton and NuSTAR, is the textbook case.
• Gravitational-wave ringdown. When two black holes merge, the remnant "rings down" in quasinormal modes whose frequencies encode its final mass and spin. LIGO and Virgo, since the first detection GW150914 in September 2015, routinely measure remnant spins of a* ≈ 0.6–0.9, independently fixing where the new ISCO settles.

The Event Horizon Telescope's 2019 image of M87* and 2022 image of Sagittarius A* show the photon ring and the disk's inner glow; the dark central "shadow" and the brightness asymmetry are sensitive to the ISCO and the hole's spin, though current resolution does not isolate the ISCO on its own.

Discovery and the people who found it

The ISCO is implicit in Karl Schwarzschild's 1916 solution of Einstein's field equations, but its physical meaning emerged only as relativists worked out the geodesics. The marginally stable orbit at 6 GM/c² for Schwarzschild was understood by the 1930s and laid out cleanly in the geodesic analyses of the mid-twentieth century. The decisive astrophysical step came with the Kerr metric: Roy Kerr published the rotating black hole solution in 1963, and in 1972 James Bardeen, William Press and Saul Teukolsky derived the closed-form spin-dependent ISCO and its energetics in their landmark paper "Rotating Black Holes: Locally Nonrotating Frames, Energy Extraction, and Scalar Synchrotron Radiation." Igor Novikov and Kip Thorne, also in 1973, built the general-relativistic thin-disk model that adopts the ISCO as the disk's zero-torque inner boundary — the framework still used to fit spins today. Thorne's 1974 photon-capture argument set the a* ≈ 0.998 spin ceiling. The ISCO moved from theory to measurement in the 1990s, when X-ray spectroscopy with ASCA, RXTE and later XMM-Newton, Chandra and NuSTAR began reading inner-disk radii off real accreting black holes.

Subtleties and common misconceptions

• The ISCO is not the event horizon. For Schwarzschild the horizon is at 2 GM/c² and the ISCO is three times farther out, at 6 GM/c². You can hover or escape from just inside the ISCO with a rocket; the horizon is the true point of no return. The ISCO only forbids stable free orbits, not all motion.
• It is not the photon sphere either. Light has its own innermost circular orbit, the photon sphere at 3 GM/c² (Schwarzschild). Massive particles can orbit between 3 and 6 GM/c², but only unstably. The ISCO at 6 GM/c² is where stability returns.
• "a* = 1 ISCO at 1 GM/c²" is a coordinate coincidence. For an extremal Kerr hole the ISCO, horizon and ergosphere edge all sit at r = 1 GM/c² in Boyer–Lindquist coordinates, but the proper radial distance between them is finite and nonzero — the coordinates collapse, not the geometry. Real holes also cannot reach exactly a* = 1.
• The inner edge is not perfectly sharp. The zero-torque assumption at the ISCO is an idealisation. Magnetohydrodynamic simulations show magnetic stresses can transport some angular momentum and torque across the ISCO, adding luminosity from the plunging region and blurring the edge by tens of percent. This is an active correction in precision spin measurements.
• Eccentric and inclined orbits have their own limits. The 6 GM/c² value is for equatorial circular orbits. The marginally bound orbit (4 GM/c² for Schwarzschild) and the marginally stable inclined orbits around Kerr holes differ; the "ISCO" specifically means the equatorial, marginally stable, circular case.