Superbubbles: Clustered Supernovae Blowing a Cavity in a Galaxy

Why massive stars blow bubbles together, not separately

Massive stars are almost never solitary. Star formation turns a molecular cloud into a cluster of 10³–10⁵ solar masses packed inside a few parsecs, or a looser OB association of similar mass spread over tens of parsecs, and the initial mass function guarantees that a 10⁴-solar-mass cluster holds roughly 100 stars above 8 solar masses. Their lifetimes run from about 3 Myr at 100 solar masses to 40 Myr at 8, so the supernovae are spread over tens of millions of years but confined to a volume far smaller than the cavity they excavate. That is the whole trick: one explosion is a transient, a hundred explosions in one place are a continuous power source.

The energy arrives in two instalments. First the winds: an O star sheds 10⁻⁷–10⁻⁶ solar masses per year at 2,000–3,000 km/s, a mechanical luminosity L = ½Ṁv² near 10³⁶ erg/s each, so 100 O stars supply ~10³⁸ erg/s as soon as the cluster lights up. Then the supernovae: 100 events at 10⁵¹ erg over 40 Myr average to ~8 × 10³⁷ erg/s. The two are comparable, which is why a cluster is idealised as a constant source with L ≈ 10³⁸ erg/s.

The resulting structure, from Castor, McCray and Weaver (1975) and Weaver and colleagues (1977), has four zones. A freely expanding cluster wind hits a reverse shock and is thermalised to 10⁶–10⁷ K at n ≈ 0.001–0.01 cm⁻³, a gas whose cooling time exceeds 100 Myr, so it cannot cool. That is the piston, nearly uniform at P/k ≈ 10⁴–10⁵ K cm⁻³, several to a few tens of times the mean interstellar thermal pressure. Outside a contact discontinuity, where conduction evaporates shell gas inward and mass-loads the X-ray-emitting layer, the forward shock sweeps ambient gas into a thin radiatively cooled shell at 10⁴ K and below that holds essentially all the mass: 10⁵–10⁶ solar masses.

The three-fifths law, and the numbers behind it

If the interior is adiabatic and the shell is thin, the only quantities in the problem are the mechanical luminosity L, the ambient density ρ and the elapsed time t. Exactly one combination of them has units of length, so R = A (L t³/ρ)^(1/5), with A = (125/154π)^(1/5) ≈ 0.76 supplied by the full self-similar solution. Growth decelerates, since v = dR/dt = (3/5) R/t.

Put numbers in. With L = 10³⁸ erg/s and one hydrogen atom per cm³ (ρ ≈ 2.3 × 10⁻²⁴ g/cm³ including helium), at t = 10 Myr the radius is about 260 pc — some 850 light-years — expanding at 15 km/s. The scaled form is R ≈ 67 pc × (L₃₈/n₀)^(1/5) × t_Myr^(3/5): at 1 Myr the shell radius is 67 pc and it is moving at 40 km/s, and by 10 Myr it has quadrupled in size, slowed by a factor of 2.5, and swept up about 2 × 10⁶ solar masses.

The fifth root cuts both ways: a factor of 100 error in L changes R by only 2.5, so the prediction is robust but a measured radius pins L down only to an order of magnitude. Kinematics rescue the inverse problem, since v/R = 3/5t dates a bubble without knowing L at all.

The adiabatic energy budget is clean: five elevenths of the injected energy stays as thermal energy in the hot interior, about 20 percent becomes shell kinetic energy, the remaining third is radiated at the outer shock. A single supernova instead follows Sedov's R ∝ (E t²/ρ)^(1/5) and merges into the interstellar medium after a few hundred thousand years; continuous injection keeps a superbubble growing roughly a hundred times longer, until the expansion speed drops to the ambient turbulent speed of about 10 km/s after tens of Myr and a few hundred parsecs. The interior stays hot for another 10⁸ years.

Blowout: chimneys, fountains and galactic winds

A galactic disc is stratified: neutral hydrogen has a scale height of roughly 100–200 pc in the Milky Way, molecular gas only about 70 pc, so a growing bubble runs out of atmosphere vertically long before it does radially. Kompaneets solved the stratified-atmosphere blast problem in 1960, and the consequence is dramatic — beyond two or three scale heights the falling density ahead of the shell makes it accelerate. A dense shell accelerated by a light hot fluid is Rayleigh-Taylor unstable, so it fragments and the million-degree interior vents. Mac Low and McCray (1988) reduced the condition to one dimensionless number built from mechanical luminosity, ambient density and scale height, with blowout above roughly 100.

What follows depends on the galaxy. In a large spiral the vented gas cannot escape and follows the chimney picture of Norman and Ikeuchi (1989), feeding the galactic fountain proposed by Shapiro and Field in 1976: hot metal-rich gas rises into the halo, cools, and rains back as high-velocity clouds. In dwarfs and starbursts the escape speed is low and merged superbubbles drive a genuine wind, as in M82 at 3.5 Mpc. Because fresh ejecta is injected into the hot phase, blowout is a metal-selective valve — why low-mass galaxies sit low on the mass-metallicity relation.

Superbubbles also make stars, by the collect-and-collapse mechanism of Elmegreen and Lada (1977): Zucker and colleagues showed in Nature in 2022 that the seven nearest star-forming complexes, Taurus and Ophiuchus among them, lie on the Local Bubble's surface with velocities consistent with having been swept up by it. Clustering is also why superbubbles dominate feedback in galaxy formation — a supernova exploding into a 0.01 cm⁻³ cavity stays adiabatic far longer than one in 1 cm⁻³ gas, so much more of its 10⁵¹ erg survives as hot gas and momentum.

How astronomers actually detect and measure them

Neutral hydrogen. The 21 cm line traces the cold shell directly. Carl Heiles catalogued Galactic HI shells in 1979 and coined supershell for those needing more than about 3 × 10⁵² erg. In face-on galaxies, surveys such as THINGS with the Very Large Array (2008) show discs riddled with HI holes; the dwarf Holmberg II looks like Swiss cheese.

Optical kinematics. An expanding shell splits an emission line in two. Echelle spectroscopy of Hα in Large Magellanic Cloud superbubbles such as N11 and N70, pioneered by You-Hua Chu and later Sally Oey, gives expansion velocities of 20–100 km/s and hence dynamical ages.

X-rays and 3D maps. The hot interior radiates thermal bremsstrahlung and lines in the soft band: ROSAT's all-sky survey (1990–91) mapped the quarter-keV background, Chandra and XMM-Newton resolved individual Magellanic superbubbles, and eROSITA aboard Spektr-RG, launched in 2019, has mapped the Local Hot Bubble at kT ≈ 0.1 keV. The local cavity's wall, first traced by the absence of Na I and Ca II absorption toward nearby stars (Sfeir and colleagues 1999; Lallement and colleagues 2003; Welsh and colleagues 2010), is now a solid surface in Gaia-based 3D dust maps (Leike and Enßlin 2020; Edenhofer and colleagues 2024).

Radioactive fallout. Live iron-60, half-life 2.6 Myr, turns up in deep-ocean ferromanganese crusts, sediments and lunar regolith (Knie and colleagues 2004; Wallner and colleagues, Nature 2016), with pulses about 1.5–3 and 6.5–8.7 Myr ago: literal debris from the supernovae that shaped our own superbubble.

Gamma rays. Fermi-LAT discovered the Cygnus Cocoon in 2011 — freshly accelerated cosmic rays inside the superbubble around Cygnus OB2 at 1.4 kpc, later traced to about 100 TeV by HAWC — and H.E.S.S. detected TeV emission from 30 Doradus C in 2015. Superbubbles do not merely push gas; they accelerate particles.

The named ones: Local Bubble, Orion-Eridanus, 30 Doradus, W4

The Local Bubble. The Sun sits in a cavity roughly 1,000 light-years across, an effective radius near 165 pc, filled with gas at about 10⁶ K and 0.005 particles per cm³ — which is why the local interstellar medium is so transparent. Zucker and colleagues traced the motions of nearby molecular clouds back to a common origin about 14 million years ago in the Upper Centaurus-Lupus and Lower Centaurus-Crux subgroups of Scorpius-Centaurus, requiring roughly 15 supernovae; the shell still expands at about 6.7 km/s. The Sun entered around 5 million years ago and sits near the centre only by coincidence, and the cavity is open toward both Galactic poles: the Local Chimney.

Orion-Eridanus. The nearest superbubble still visibly powered by its cluster, spanning some 20 by 45 degrees of sky at 150–500 pc. Orion OB1 has been injecting energy for 10–12 Myr, of order 10⁵² erg. The Eridanus filaments trace the near side of the shell, only some 150–200 pc away, while Barnard's Loop wraps the far end around the Orion clouds near 400 pc; Reynolds and Ogden mapped its kinematics in 1979, Ochsendorf and colleagues modelled it in 2015.

30 Doradus and W4. In the Large Magellanic Cloud at 50 kpc, the R136 cluster packs 10⁴–10⁵ solar masses of stars into a few parsecs and drives the Tarantula Nebula's nested shells; its neighbour 30 Doradus C is a superbubble tens of parsecs across with a non-thermal X-ray rim and TeV emission. Closer to home, Normandeau, Taylor and Dewdney reported in Nature in 1996 that the cavity above Cas OB6, known as W4, opens into the halo: the textbook Galactic chimney.

Look-alikes, failure modes and open questions

Not a supernova remnant. A remnant is one instantaneous 10⁵¹ erg blast, tens of parsecs across and 10⁴–10⁵ years old, following Sedov's t^(2/5). A superbubble has a hundred times the energy, ten times the radius, a thousand times the age, and follows t^(3/5) because injection is continuous.

Not an HII region, and not the Fermi bubbles. A Strömgren sphere is photoionised gas at 10⁴ K driven by its own thermal pressure — a hundred times cooler and ten thousand times denser than a superbubble interior, and usually nested inside one, as the Orion Nebula sits inside Orion-Eridanus. The Fermi bubbles are 10 kpc lobes powered from the Galactic centre, a different engine at fifty times the scale. Nor is the Local Bubble the heliosphere (~10⁻³ pc across).

Where the model breaks. Add up a real superbubble's X-ray luminosity plus shell kinetic energy and you recover under a tenth of what its stars should have injected; the rest vents into the halo, is radiated in turbulent mixing layers at the contact discontinuity, or was never conducted because magnetic fields insulate the interface. Sizes misbehave too: Oey (1996) and Oey and García-Segura (2004) found Magellanic superbubbles whose radii and velocities disagree with the Weaver prediction for their stellar content, some too fast, some too slow. The law is a scaling guide, not a fit.

An instructive overcorrection. The quarter-keV X-ray background was long credited almost entirely to the Local Hot Bubble, until charge exchange with the solar wind inside our own heliosphere was found to supply roughly 40 percent of it (DXL sounding rocket; Galeazzi and colleagues, Nature 2014). The hot bubble survived, but smaller than advertised.

Still open. What fraction of the disc volume is hot gas — the 10 to 20 percent of some estimates, or the more than half implied by McKee and Ostriker's 1977 three-phase model? How much momentum does each clustered supernova really deposit, where simulations still differ by nearly an order of magnitude? And do superbubble interiors, with their many weak repeated shocks, produce most Galactic cosmic rays?

Cavity-blowing structures, from one star to a whole starburst

Structure	Energy source	Typical size	Radius–time behaviour
Stellar wind bubble (NGC 6888)	One Wolf-Rayet star, ~10⁵⁰ erg over ~10⁵ yr	~5–10 pc	R ∝ (L t³/ρ)^1/5, strongly decelerating
Supernova remnant (Cygnus Loop)	One explosion, ~10⁵¹ erg released at once	~10–50 pc	R ∝ (E t²/ρ)^1/5 (Sedov-Taylor)
Superbubble (Orion-Eridanus)	An OB association, ~10⁵²–10⁵³ erg over 10–40 Myr	~100–300 pc	R ∝ (L t³/ρ)^1/5 (Weaver)
Supershell or chimney (W4, Cas OB6)	Several clusters, ~10⁵³–10⁵⁴ erg, breaking out of the disc	~0.3–1 kpc	Re-accelerates up the density gradient (Kompaneets)
Starburst superwind (M82)	Hundreds of clusters in a nuclear starburst	~1–10 kpc outflow	Free-flowing wind (Chevalier-Clegg)