Galactic Wind

A galaxy that exhales

Picture a star-forming galaxy as a slow, smouldering fire. Cold gas falls in from the cosmic web, settles into a disk, and condenses into stars. The most massive of those stars burn out in a few million years and detonate as supernovae, each releasing about 10⁵¹ erg of energy. In a vigorously star-forming region, these explosions go off close enough together in space and time that their hot bubbles overlap and merge into a single, enormous over-pressured cavity of gas at 10⁶–10⁷ K. That hot gas has nowhere to go but up: it is far less dense than the disk around it, so it punches through the path of least resistance, perpendicular to the disk plane, and roars out into the halo. That outflow is a galactic wind.

The crucial point is that the wind doesn't just throw away energy — it throws away the fuel. The gas it carries off, or heats so it cannot cool, is the very material that would otherwise have collapsed into the next generation of stars. A galaxy with a strong wind is a galaxy that keeps burning its own kindling. This is why galactic winds sit at the centre of modern galaxy-evolution theory: they are the dominant way a galaxy regulates, and ultimately limits, its own growth.

What counts as a galactic wind

A galactic wind is a coherent, large-scale outflow of gas that leaves the main body of a galaxy, as opposed to local turbulence or a single supernova remnant. It is multiphase: the same outflow can contain a hot, fast, tenuous fluid (the X-ray-emitting "wind fluid" that carries most of the energy), a warm ionised phase (seen in Hα and in blueshifted absorption lines), a cool atomic and molecular phase (entrained clouds that ride the wind), and dust. These phases move at different speeds and the slower ones may not escape at all.

There are two principal launching mechanisms, and they define the two great families of winds:

• Stellar-feedback (starburst) winds. Powered by the combined energy and momentum of supernovae, massive-star winds, and radiation pressure on dust. These dominate in star-forming and starburst galaxies and produce the classic biconical "superwind" venting out of a disk.
• AGN-driven winds. Powered by an accreting supermassive black hole — through radiation pressure, hot winds off the accretion disk, and relativistic jets. These can reach far higher velocities and are essential to quenching the most massive galaxies, where supernovae alone cannot unbind the gas.

The mechanism: energy versus momentum

How a wind accelerates gas depends on whether it is energy-driven or momentum-driven. In an energy-driven wind, the supernovae heat a bubble whose thermal energy does work on the surrounding gas as it expands — this is efficient if radiative losses are small. In a momentum-driven wind, photons and the hot fluid impart momentum directly to clouds and dust grains, important when the gas is dense and radiates away its thermal energy quickly. Real winds are some blend of the two.

The canonical analytic picture is the Chevalier & Clegg (1985) model of a starburst superwind. Energy and mass are injected into a central volume at rates Ė and Ṁ. Inside the injection region the gas is subsonic; it passes through a sonic point at the edge and then flows out supersonically, asymptotically approaching a terminal velocity

v_∞ ≈ √(2 Ė / Ṁ)

The central temperature and the terminal velocity follow directly from the two injection rates:

T_central ≈ 0.4 (μ m_H / k_B) (Ė / Ṁ)
v_∞       ≈ √(2 Ė / Ṁ)   (km/s)

For a typical starburst with a thermalisation efficiency near unity, this gives a hot fluid at a few times 10⁷ K and a terminal velocity around 1000–3000 km/s — much faster than the few-hundred-km/s motions seen in cool absorbing gas, because the cool clouds are denser and accelerate less. The single most important derived quantity is not the velocity, though; it is the mass-loading factor.

Mass loading: the number that runs galaxy formation

The mass-loading factor is defined as the ratio of the gas mass-outflow rate to the star-formation rate:

η = Ṁ_out / SFR

It answers a deceptively simple question: for every solar mass of gas a galaxy turns into stars, how many solar masses does it blow out? If η = 3, the galaxy ejects three times as much gas as it locks into stars, so only a quarter of the gas processed through the disk ends up in stars. Cosmological simulations cannot reproduce the observed galaxy population without large, mass-dependent η, and the dependence is steep: theory and observation both point to η rising toward low-mass galaxies roughly as a power of the circular velocity,

η ∝ v_circ^(−1)     (momentum-driven scaling)
η ∝ v_circ^(−2)     (energy-driven scaling)

The physical reason is escape velocity. A dwarf galaxy with a circular velocity of 30 km/s can be largely emptied by one good starburst; a massive galaxy with v_circ ≳ 200 km/s holds its gas far more tightly. This is the lever that bends the steep dark-matter halo-mass function into the much flatter observed stellar-mass function — without strong feedback, simulations make far too many faint galaxies and far too many massive ones.

Galactic winds by the numbers

Driver / host	Velocity (cool gas)	Hot-fluid velocity	Mass loading η	Example
Milky-Way-type disk (fountain)	50 – 150 km/s	~500 km/s	~0.1 – 1	Milky Way nuclear outflow
Dwarf starburst	100 – 300 km/s	~1000 km/s	~3 – 10	NGC 1569, IC 10
Local starburst superwind	300 – 600 km/s	1000 – 2000 km/s	~1 – 4	M82, NGC 253
Ultra-luminous IR galaxy	500 – 1000 km/s	~3000 km/s	~1 – 5	Arp 220, Mrk 231
High-z star-forming galaxy	~200 – 800 km/s	—	~0.5 – 3	z ~ 2 "down-the-barrel" outflows
AGN ultra-fast outflow	—	0.1 – 0.3 c	up to ~1	PDS 456, APM 08279+5255

A few orders of magnitude are worth fixing in mind. A core-collapse supernova injects about 10⁵¹ erg. A galaxy forming stars at 10 M☉/yr produces roughly one such supernova every ~10 years (one per ~100 M☉ of stars formed, given the initial mass function), for a mechanical luminosity near 3 × 10⁴² erg/s. A genuine starburst like M82, forming stars at ~10 M☉/yr in a region only a few hundred parsecs across, channels a sizeable fraction of that into a wind that extends more than 10 kpc above the disk and carries on the order of a few M☉/yr of gas outward.

Worked example: can a dwarf galaxy keep its gas?

Take a dwarf galaxy with a circular velocity v_circ = 35 km/s, so its escape velocity from the inner halo is roughly v_esc ≈ 2.5 v_circ ≈ 90 km/s. Suppose a burst forms 10⁶ M☉ of stars. Using a standard initial mass function, about one supernova occurs per 100 M☉ of stars formed, giving N_SN ≈ 10⁴ supernovae, each with E_SN = 10⁵¹ erg:

E_total = N_SN × E_SN = 10⁴ × 10⁵¹ erg = 10⁵⁵ erg

Now ask how much gas that energy could unbind if even 10% of it (a typical thermalisation efficiency after radiative losses) couples to the gas as bulk kinetic energy at the escape velocity. Working in CGS, with v_esc = 90 km/s = 9 × 10⁶ cm/s:

E_avail   = 0.1 × E_total = 0.1 × 10⁵⁵ erg = 10⁵⁴ erg
M_ejected ≈ 2 E_avail / v_esc²
          = (2 × 10⁵⁴ erg) / (9 × 10⁶ cm/s)²
          = (2 × 10⁵⁴) / (8.1 × 10¹³)  g
          ≈ 2.5 × 10⁴⁰ g
          ≈ 1.2 × 10⁷ M☉      (M☉ = 2 × 10³³ g)

So a burst making 10⁶ M☉ of stars can eject on the order of 10⁷ M☉ of gas — a mass-loading factor of roughly ten. The dwarf cannot hold its gas; it is blown out. Repeat the calculation for a Milky-Way-mass galaxy with v_esc ≈ 500 km/s and v_esc² is ~30 times larger, so the same energy ejects ~30 times less mass — the wind is feeble and most gas stays bound. This single scaling, M_ejected ∝ v_esc⁻², is the heart of why feedback matters most for the smallest galaxies.

How we actually see winds

Galactic winds are observed across the electromagnetic spectrum, and combining the phases is the only way to get the full energy and mass budget:

• Blueshifted absorption lines. Looking "down the barrel" at a galaxy, foreground outflowing gas absorbs the galaxy's own light at a blueshift set by the wind speed. Resonance lines of Na I D, Mg II, Fe II, and Si IV trace cool and warm gas moving toward us at 100–1000 km/s. This is the workhorse for high-redshift galaxies.
• Extended Hα and [N II] filaments. Warm ionised gas in the wind glows in optical emission lines, tracing the biconical structure directly — the iconic red filaments of M82 are exactly this.
• X-ray emission. The hot 10⁶–10⁷ K wind fluid radiates soft X-rays; Chandra maps of M82 and NGC 253 show the diffuse hot outflow filling the cone.
• Molecular outflows. ALMA detects cold CO moving outward at hundreds of km/s, showing that even dense, star-forming molecular gas is swept up — most dramatically in AGN hosts like Mrk 231.
• X-ray ultra-fast outflows (UFOs). Blueshifted Fe K absorption in the X-ray spectra of quasars reveals AGN winds at 0.1–0.3 c, the fastest galactic-scale outflows known.

Famous galactic winds

• M82 (the Cigar Galaxy). At only 3.5 Mpc, M82 is the textbook nearby starburst superwind. A nuclear starburst forming stars at ~10 M☉/yr drives a biconical outflow visible in Hα filaments extending more than 10 kpc and in diffuse soft X-rays. It is the single most-imaged galactic wind in astronomy.
• NGC 253 (the Sculptor Galaxy). A southern-sky starburst at ~3.5 Mpc with a well-studied multiphase wind, including ALMA-detected molecular outflow streaming out of the disk.
• The Fermi Bubbles. Two enormous gamma-ray lobes extending ~50° (≈ 8 kpc) above and below the Galactic centre, discovered with the Fermi telescope in 2010 — fossil evidence that even the relatively quiet Milky Way drove a powerful outflow in its past, from a starburst or a phase of black-hole accretion.
• Mrk 231 and PDS 456. Nearby AGN hosts showing molecular outflows and ultra-fast X-ray winds respectively — the cleanest demonstrations that black-hole-driven winds can sweep up a galaxy's gas.
• High-redshift outflows. Down-the-barrel absorption studies of star-forming galaxies at z ≈ 2–3 (the cosmic-noon epoch) show that nearly every vigorously star-forming galaxy at that time drove a wind of several hundred km/s — winds were ubiquitous when the universe was building most of its stars.

Common misconceptions and edge cases

• "The wind blows the galaxy apart." No — the wind preferentially vents the diffuse, hot interstellar gas perpendicular to the disk, where the resistance is lowest. The dense disk and the stars themselves are essentially untouched. What the wind removes is the fuel, not the structure.
• "Wind speed equals escape; the gas is gone." Often it isn't. Much of the cool, slow gas in normal galaxies is launched below the escape velocity and falls back as a galactic fountain, recycling over tens of millions of years. Only the fast hot fluid, and the cool clouds in the lowest-mass galaxies, truly escape.
• "Mass loading is one number for a galaxy." η depends on which phase you measure and at what radius. Hot-gas η, cool-gas η, and the η measured at 1 kpc versus 10 kpc can differ by factors of several. Comparing simulations to data requires matching the phase and aperture.
• "Supernovae can quench any galaxy." They cannot quench the most massive ones. In a galaxy with v_esc ≳ 400 km/s, supernova energy is simply insufficient to unbind the gas, and AGN feedback is required. The crossover near a halo mass of ~10¹² M☉ is exactly where galaxy colours switch from blue to red.
• "The wind is a steady breeze." Winds are bursty. Because they are driven by star formation that is itself clumpy in space and time, outflows come in episodes; a galaxy can have a strong wind during a burst and almost none between bursts. The Fermi Bubbles are a fossil of one such past episode in the Milky Way.