Circumplanetary Disk

A solar system within a solar system

Picture the early Solar System: a flattened disk of gas and dust spins around the young Sun, and over millions of years it condenses planets in concentric lanes — rock close in, ice farther out. Now zoom in on one growing gas giant. Around it, the same story replays in miniature. A second, much smaller disk wraps around the planet, spinning in its equatorial plane, sorting rock from ice by temperature, and growing little bodies in ordered lanes. That nested disk is the circumplanetary disk, and its lanes become the planet's moons.

The two-level structure is the whole point. The big disk — the protoplanetary disk — orbits the star and is hundreds of astronomical units across. The little disk orbits a single planet and is a fraction of an AU across. They are coupled: the planet has carved an annular gap in the parent disk, and gas leaks across that gap, falls toward the planet, and (because it carries angular momentum) cannot land directly. Instead it circularises into the circumplanetary disk, which then meters that gas onto the planet at its own pace. The disk is a gatekeeper, a flywheel, and a nursery all at once.

How the disk forms: gas falling through the Hill sphere

Once a forming planet exceeds roughly 10–20 Earth masses, it can hold onto a hydrogen-helium envelope and begins runaway gas accretion. The planet's gravitational reach is set by its Hill sphere, the region where its gravity dominates over the star's tide:

R_Hill = a (M_p / 3 M_star)^(1/3)

where a is the planet's orbital distance, M_p its mass, and M_star the stellar mass. Gas drawn from the parent disk streams in toward the planet from above and below the midplane and from the gap edges. Crucially, this gas is not falling radially — it shares the orbital shear of the parent disk, so it arrives with substantial angular momentum relative to the planet. It cannot reach the planet's surface directly. Instead the inflow shocks, loses vertical motion, and settles onto a rotationally supported disk in the planet's equatorial plane. Three-dimensional hydrodynamic simulations show the gas spirals down the polar regions and feeds the disk near its outer edge, then drifts inward.

Tidal truncation by the star limits how far out the disk can extend. A circumplanetary disk is gravitationally bound only out to about 0.3–0.4 of the Hill radius; beyond that, prograde orbits become unstable and the star strips the gas away. This is the same physics that keeps a planet's moons on stable orbits and is why all the regular satellites sit comfortably inside that boundary.

The gas-starved disk model

The benchmark theory is the "gas-starved" disk of Robin Canup and William Ward (2002). The key insight is that the circumplanetary disk is not a fixed reservoir that forms once and drains. It is continuously re-supplied by the inflow from the parent disk, and it processes that inflow viscously toward the planet. In steady state, the disk surface density is set by balancing the inflow rate against the viscous drainage rate — not by how much gas happens to be present at one instant.

This balance keeps the instantaneous disk mass low — far lower than the "minimum-mass subnebula" people once derived by simply adding up the satellites and their lost ice and scaling to solar composition. That older approach gave a disk so massive and hot that ice could never survive at Ganymede's and Callisto's orbits. The gas-starved model resolves the paradox: at any moment the disk holds only a small mass, perhaps 10⁻⁴ to 10⁻³ of the planet's mass, but it cycles many times that total through over the planet's growth. Satellites build up gradually from the trickle, and because the disk is low-density and optically thinner, it stays cold enough to retain ice in its outer parts.

A steady-state estimate of the disk's accretion rate onto the planet follows the same relation as any accretion disk:

Ṁ_p = 3π ν Σ        (steady-state mass flow through a viscous disk)
ν   = α c_s H        (Shakura–Sunyaev α-viscosity)

with α of order 10⁻³ to 10⁻², c_s the local sound speed, H the scale height, and Σ the surface density. The gas-starved model fixes Σ by requiring Ṁ_p to equal the supply from the parent disk. The result is a disk that is geometrically thin in its outer regions, warmer and puffier near the planet, and quietly conveys gas inward for the millions of years the planet spends growing.

Circumplanetary vs protoplanetary disk

The two disks share the physics of accretion but differ in nearly every number. The table below uses Jupiter and its parent solar nebula as the canonical example.

Property	Protoplanetary (circumstellar) disk	Circumplanetary disk
Central body	Young star (~1 M☉)	Giant planet (~10⁻³ M☉)
Radial extent	10–1000 AU	~0.01–0.1 AU (≲ ⅓ R_Hill)
Mass reservoir	0.001–0.1 M☉	~10⁻⁴–10⁻³ M_planet (instantaneous)
Temperature range	~10–1500 K	~50–1000+ K (warmer per unit radius)
Mass supply	Cloud collapse, then self-contained	Continuous inflow from parent disk
Lifetime	~3–10 Myr	As long as the planet accretes, then dissipates fast
Builds	Planets	Regular satellites (moons)
Snow line at	~2.7 AU (Sun)	Near Europa's orbit (~9–10 R_Jup)

The deepest difference is the supply. A protoplanetary disk is born with its gas and slowly loses it; a circumplanetary disk is a flow-through system whose mass at any instant tells you almost nothing about how much material it has handled. That single distinction is what rescued moon-formation theory from the overheated subnebula models of the 1980s.

Building moons: snow lines and satellitesimals

Inside the disk, the same condensation sequence that sorted the Solar System operates on a smaller scale. Close to the hot planet, only refractory rock and metal can condense; farther out, past the disk's own snow line, water ice freezes out and doubles or triples the available solid mass. Dust grains coagulate, settle to the midplane, and grow into kilometre-scale "satellitesimals," which then accrete into moons. Because solids are far more abundant beyond the snow line, the outer moons are icy and the inner ones rocky.

Jupiter's Galilean moons display this gradient precisely:

Moon	Orbit (R_Jup)	Density (g/cm³)	Composition
Io	5.9	3.53	Rock + iron, no ice
Europa	9.4	3.01	Rock + thin ice/ocean shell
Ganymede	15.0	1.94	~50% ice by mass
Callisto	26.3	1.83	~50% ice, undifferentiated

The decline in density from Io (rock) to Callisto (half ice) is the disk's snow line written into the moons. The inner three are locked in the Laplace resonance (orbital periods in a 1:2:4 ratio), a relic of how they migrated inward through the gas disk and got captured into resonance as the disk torqued their orbits — a miniature version of the migration that shaped the giant planets themselves.

Satellites can also migrate into the planet. Type I migration through the gas drives moons inward, and the survivors are those that piled up at the disk's inner cavity or were saved when the disk dissipated. Saturn's single large moon Titan, versus Jupiter's four, may reflect a different migration-and-merging history in a colder, lower-mass disk.

Setting the planet's spin and the disk's inner edge

One of the disk's quietest but most important jobs is regulating how fast the planet rotates. Gas reaching the planet from the disk's inner edge carries specific angular momentum of roughly the Keplerian value there. If all of it were retained, a Jupiter-mass planet would spin up to break-up in under a million years. Observed giant planets rotate well below break-up — Jupiter's day is 9.9 hours, about 40 percent of its centrifugal break-up rate.

The resolution involves the planet's magnetosphere and the disk together. A magnetised young giant truncates the disk near its corotation radius and channels accretion along field lines, exporting angular momentum back to the disk and to magnetised winds. This is exactly the "disk-locking" mechanism invoked to explain slowly rotating young stars, scaled down to planetary masses. The disk's inner edge therefore sits not at the planet's surface but near where the magnetic and ram pressures balance — and that radius helps set both the spin and the innermost stable moon orbit.

Catching one in the act: PDS 70 and ALMA

For decades circumplanetary disks were purely theoretical — too small and too embedded to see. That changed with PDS 70, a roughly 5-million-year-old T Tauri star 370 light-years away with a wide gap in its protoplanetary disk. Inside that gap sit two directly imaged, still-accreting gas giants, PDS 70 b and c, betrayed by their hydrogen-alpha accretion glow (the SPHERE/MagAO and Hubble detections) and by thermal infrared light (the Keck and VLT detections). They are the only confirmed protoplanets ever directly imaged.

In 2019, ALMA resolved a compact, dusty source of submillimetre continuum emission co-located with PDS 70c and clearly offset from the star — the first detection of a circumplanetary disk. The emission implied enough dust to build a Galilean-scale satellite system. Later ALMA and JWST observations have refined the picture, detecting gas tracers and constraining the disk's size to less than about an astronomical unit, set by the large Hill sphere of a several-Jupiter-mass planet at 34 AU. PDS 70 is now the textbook laboratory for watching a planet and its future moons form at the same time.

Where circumplanetary disks show up

• PDS 70 b and c. The only directly imaged accreting protoplanets; PDS 70c hosts the first resolved circumplanetary disk (ALMA 2019). The gold-standard observational case.
• Jupiter and the Galilean system. The best-preserved fossil. The flattened, prograde, density-stratified, resonance-locked moons are the signature of in-situ formation in a circumplanetary disk ~4.5 billion years ago.
• Saturn, Uranus, Neptune. Each large enough to have hosted a disk. Saturn's regular moons and its ring system, and Uranus's tilted but regular satellite plane, all encode their disks' geometry — Uranus's 98° tilt means its disk and moons tilted with the planet, evidence the moons formed before or during the giant impact that knocked it over.
• AB Aurigae b and other candidates. Several young systems show localised emission blobs that may be forming planets with disks; most remain ambiguous, underscoring how hard these are to confirm.
• Exomoon searches. The candidate exomoon signals around Kepler-1625b and Kepler-1708b, if real, would be moons that either formed in a circumplanetary disk or were captured — distinguishing the two is a key science goal for the next generation of instruments.

Common misconceptions and edge cases

• "It's just a tiny protoplanetary disk." Geometrically similar, but the supply physics is opposite. A protoplanetary disk is a closed reservoir that drains; a circumplanetary disk is an open, flow-through conveyor continuously fed from outside. Its instantaneous mass undercounts the material it processes by orders of magnitude.
• "The disk's mass equals the moons' mass." No. In the gas-starved model the disk holds far less than the final satellite mass at any instant; the moons accrete from a trickle over millions of years, not from a single full reservoir. This is why the old minimum-mass-subnebula estimates ran far too hot.
• "All moons formed in the disk." Only the regular moons — prograde, low-inclination, near-equatorial. Irregular moons (Jupiter's retrograde Carme/Ananke groups, Neptune's Triton on a retrograde orbit) were gravitationally captured later and have nothing to do with the disk. Triton's retrograde orbit is the smoking gun of capture.
• "Only gas giants have them." The disk requires a planet massive enough to open a gap and drive gas inflow — effectively the giant-planet regime. Terrestrial planets like Earth got their moon by a giant impact, not a disk, which is why our Moon's composition matches Earth's mantle rather than a condensation sequence.
• "The disk lasts as long as the protoplanetary disk." It lives only while the planet is fed. Once the parent disk's gas dissipates (a few to ten million years) the inflow stops, the circumplanetary disk drains in a comparatively short viscous time, and moon formation halts — fixing the satellite system in place.