Starburst Galaxy

What makes a galaxy a starburst

Every spiral galaxy is forming stars. The Milky Way produces roughly one new solar mass of stars per year, distributed quietly through molecular clouds in its arms. A starburst galaxy does the same job ten to a hundred times faster — and confines most of the action to a kiloparsec-scale region in its centre. The phrase "starburst" is reserved for a state in which a galaxy is converting its cold gas into stars so quickly that, at the current rate, the entire reservoir would be consumed in a time shorter than the time the galaxy has been around to consume it.

The precise quantitative criterion is the gas depletion time:

τ_dep ≡ M_gas / SFR

For a normal main-sequence galaxy like the Milky Way, τ_dep is roughly 1–2 Gyr — long compared to a galactic orbit but a sizable fraction of the 13.8 Gyr age of the universe. For a starburst, τ_dep drops to ~10–100 Myr. The galaxy is, in a real and quantifiable sense, burning through its lifetime supply of fuel in a single brief episode. Sustained over the age of the universe, the rate is impossible — there isn't enough hydrogen in the galaxy. So a starburst is inherently a transient, finite-duration event: a flare-up that will end either when the gas runs out or when feedback shuts it down.

The star-forming main sequence and where starbursts sit

Most star-forming galaxies fall along a relatively tight relation in stellar-mass vs. star-formation-rate space — the so-called galaxy main sequence. Its scatter is only about 0.3 dex over four decades in mass, from z = 0 to z = 4. Galaxies that lie a factor of four or more above this main sequence at their epoch are what observational papers usually classify as starbursts. Quenched galaxies — passively evolving ellipticals — sit a factor of ten or more below it. Roughly 90 % of cosmic star formation happens in normal main-sequence galaxies; the remaining 10–15 % happens in starbursts, which are a minority of the population but punch far above their weight in the infrared sky.

Three ways to ignite a starburst

What drives a galaxy off the main sequence? Three physical pathways dominate, all of them sharing a common proximate cause: a sudden compression of cold gas to high central density.

• Major mergers. Two gas-rich galaxies of comparable mass fall together. Tidal torques during the first close pass strip angular momentum from the gas, which loses about half of its specific L in a few rotation periods and pours into the central kiloparsec. The Antennae (NGC 4038/4039) is the textbook early-stage merger, with thirty-odd super-star clusters lit up across the contact region. Arp 220 and NGC 6240 are the textbook late-stage merger remnants, each hiding two supermassive black holes and a deeply embedded nuclear starburst behind 10²⁴ cm⁻² of obscuring gas.
• Tidal stripping and harassment. Gas-rich satellites being chewed up by a larger host, or galaxies undergoing repeated close encounters in a cluster, can have their disks shocked and torqued without a full coalescence. The shock-induced pressure rise compresses gas and triggers a milder starburst — what is observed in many blue-compact dwarfs and some cluster spirals.
• Cold-flow accretion. At high redshift, the cosmic-web filaments deliver narrow streams of dense cold gas directly into the centres of dark-matter haloes, by-passing the slow shock-heated virial-temperature mode that dominates locally. The cold-mode infall rate can match the SFR over Myr timescales and is the favoured mechanism for the cosmic-noon population of clumpy main-sequence and starburst galaxies seen by JWST and ALMA at z = 2–4.

The Schmidt-Kennicutt law

The empirical relationship between gas and star formation was first written down by Maarten Schmidt in 1959 and observationally refined by Rob Kennicutt in 1998 using a sample of 61 normal disks and 36 starburst galaxies. Spatially averaged over a galaxy or a kiloparsec-scale region, the relation is

Σ_SFR = (2.5 ± 0.7) × 10⁻⁴ × (Σ_gas / M☉ pc⁻²)^(1.4 ± 0.15) M☉ yr⁻¹ kpc⁻²

where Σ_gas is the surface density of total hydrogen (HI + H₂). The 1.4 power index has a clean physical motivation: if star formation in a vertically-supported gas disk happens on roughly a free-fall time, the rate per unit area scales as ρ_gas / t_ff ∝ Σ_gas × √(ρ_gas) ∝ Σ_gas^1.5, close to the observed slope. A "starburst" in this language is a galaxy that has driven Σ_gas to extreme values in a small region — concentrating the entire molecular reservoir into a central few hundred parsecs raises Σ_SFR by orders of magnitude because of the super-linear power.

A complementary version of the same law, the "molecular Schmidt-Kennicutt" relation, restricts attention to the molecular component Σ_H2 and gives a slope closer to 1.0 — consistent with the idea that essentially all star formation happens in molecular clouds and the apparent steepening of the total-gas relation reflects the increasing molecular fraction at high densities.

Local starbursts — the laboratory specimens

Galaxy	Distance	SFR (M☉/yr)	LIR (L☉)	Notes
M82 (Cigar Galaxy)	3.5 Mpc	~10	5 × 10¹⁰	Edge-on, biconical superwind, prototype nearby starburst
NGC 253 (Sculptor)	3.5 Mpc	~5	2 × 10¹⁰	Highly inclined, X-ray bright wind, OH megamaser
NGC 4038/4039 (Antennae)	22 Mpc	~20	1 × 10¹¹	Early-stage merger, hundreds of super-star clusters
Arp 220	78 Mpc	~200	1.5 × 10¹²	Late-stage merger, ULIRG, twin nuclei, Compton-thick
NGC 6240	110 Mpc	~100	7 × 10¹¹	Late merger, dual AGN, 'spider' morphology
Mrk 231	180 Mpc	~200	3 × 10¹²	Nearest QSO ULIRG, broad-absorption-line outflow
IRAS F10214+4724	z = 2.29	~700 (lens-corrected)	3 × 10¹³	First high-z ULIRG, gravitationally lensed

M82 is the workhorse. Three and a half megaparsecs away, edge-on, and only 4 % the mass of the Milky Way, it nonetheless cranks out about ten solar masses of new stars every year — and powers a biconical, X-ray–visible superwind that extends 10 kpc above and below the disk. Arp 220, almost a hundred megaparsecs further, is the local archetype of a late-stage major-merger ULIRG: it harbours two nuclei separated by 350 pc, each radiating 10¹² L☉ in the infrared, each buried behind a column so dense that even hard X-rays are absorbed.

Infrared bright: ULIRGs and the dust-reradiation engine

Newborn massive stars produce most of their light in the ultraviolet. In a starburst the gas density is so high that essentially every UV photon is absorbed by interstellar dust before escaping; the dust grains thermalise it and reradiate it as a smooth blackbody-like bump at 30–100 K, peaking in the far infrared. This is the energy budget behind the ULIRG class — galaxies with L_IR > 10¹² L☉ measured over 8–1000 μm — that IRAS discovered in 1983 and that Herschel mapped in detail in the 2010s.

The IR luminosity is, to a good approximation, the bolometric luminosity of the starburst, because the dust traps essentially all of the optical/UV output. The Kennicutt calibration of L_IR to SFR is

SFR (M☉/yr) ≈ 1.5 × 10⁻¹⁰ × (L_IR / L☉)   (Kroupa IMF, optically thick)

So a ULIRG at 10¹² L☉ is forming about 150 solar masses of stars per year; the most extreme HyLIRGs (L_IR > 10¹³ L☉) reach 1000 M☉/yr. Most of that is hidden from optical surveys entirely — JWST and ALMA recover it.

Submillimetre galaxies — the high-z analogue

At cosmic noon (z ≈ 2), the same dusty-starburst population reappears in selection surveys at 850 μm. The reason is a beautiful accident of the dust spectrum: a 30 K modified blackbody peaks rest-frame around 100 μm and falls off slowly on the Rayleigh-Jeans side. Redshift the SED by z = 1–8 and the 100-μm peak slides into the 250–1000 μm band. As the source moves out to higher z and the inverse-square law dims its flux, the K-correction works in the opposite direction — bringing the brighter spectral peak into the observing band — so the observed-frame submillimetre flux is nearly independent of redshift between z = 1 and z = 8. Submillimetre surveys (SCUBA, ALMA, Herschel-SPIRE) are therefore a high-z-blind window on the dusty starburst population. The SMGs they uncover are typically L_IR = 10¹²–10¹³ L☉, with SFR = 500–2000 M☉/yr, at z = 2–4. They are the highest-SFR systems in the universe — and they probably dominate the cosmic-noon contribution to today's stellar mass in massive ellipticals.

Galactic superwinds and metal enrichment

A starburst doesn't just make stars; it loses gas, dust, and metals to its surroundings through a galactic-scale wind. The mechanism is collective feedback. A young star cluster injects energy via stellar winds and Wolf-Rayet outflows for the first few Myr; after that, core-collapse supernovae take over. The combined mechanical luminosity is approximately

L_mech ≈ 7 × 10⁴¹ × (SFR / M☉ yr⁻¹) erg/s

In a confined nuclear starburst, this energy thermalises in a hot 10⁷–10⁸ K plasma whose pressure exceeds the gravitational weight of the overlying ISM. The plasma drives a biconical breakout along the path of least resistance — the minor axis of the disk — at terminal velocities of 300–1500 km/s. M82's wind is the classic example: a kpc-wide cone of soft X-rays observed by Chandra, with cool Hα filaments threading through it traced by HST, all in front of the cigar-shaped stellar disk.

The wind is not just hot gas. It is mass-loaded with entrained cool clouds, dust-loaded, and chemically loaded with the freshly synthesised metals from the starburst's massive stars. Stacking observations indicate that the mass-loading factor η = Ṁ_out / SFR is of order 1–10 for starbursts, comparable to or larger than the SFR itself. This means a starburst returns roughly as much gas to the circumgalactic medium as it converts into stars. Over Hubble time, this feedback is the leading explanation for (a) the metallicity of the circumgalactic and intergalactic media; (b) the mass-metallicity relation of galaxies; (c) the cutoff at the high-mass end of the galaxy stellar-mass function.

Cosmic noon — the peak of cosmic star formation

If you integrate the SFR over all galaxies as a function of cosmic time, you get the Madau-Dickinson cosmic star-formation history. Its shape is striking: a steep rise from the first stars at z > 10, a broad maximum at z ≈ 2 (cosmic noon), and a factor-of-ten decline from z = 1 to today. Cosmic noon is roughly 10 Gyr ago — the moment at which the universe was making stars at ten times today's rate. About half of all the stars ever born were born within the cosmic-noon window of z = 1 to z = 3.

ρ_SFR(z) ≈ 0.015 × (1+z)^2.7 / [1 + ((1+z)/2.9)^5.6]  M☉ yr⁻¹ Mpc⁻³
                                              (Madau & Dickinson 2014)

Cosmic noon is dominated by dusty IR-luminous galaxies — the high-z descendants of today's ULIRGs and the typical "starburst" of that era. ALMA, Herschel, and JWST have given us a near-complete inventory; the picture is one in which gas-rich main-sequence galaxies and brief merger-triggered starbursts together produced most of the universe's present-day stellar content.

Worked example: how long can M82 keep this up?

M82 has a current SFR of about 10 M☉/yr and a total H₂ + HI gas reservoir of roughly 8 × 10⁸ M☉ (mostly molecular, concentrated in the central 500 pc). The gas-depletion time is

τ_dep = M_gas / SFR = 8 × 10⁸ M☉ / 10 M☉/yr = 8 × 10⁷ yr ≈ 80 Myr

That is shorter than a single orbital period at the disk's outer edge and is two orders of magnitude shorter than the age of M82. M82 cannot keep going for long without resupply. Observationally, M82 is interacting with its companion M81 — tidal HI bridges are clearly seen — and the most likely explanation is that the current burst was triggered by a close pass roughly 100–200 Myr ago. The burst will end either when the central gas is exhausted, when the supernova-driven wind blows it out, or both — within the next ~100 Myr.

Quenching — what comes after the burst

A starburst is by construction a finite-duration episode. When it ends, the galaxy quenches: its sSFR drops by an order of magnitude or more, and it slides off the main sequence into the "green valley" and eventually onto the red sequence as a quiescent galaxy. Three principal pathways take a starburst into quiescence.

• Gas exhaustion. The starburst simply runs out of fuel. Combined with the consumption rate, the ejection by the wind makes this very efficient — the galaxy can deplete its central reservoir on a single τ_dep.
• Wind ejection. The starburst-driven superwind can drive a substantial fraction of the cold gas out of the galaxy. Once ejected to the circumgalactic medium, it may take Gyr (if ever) to cool and rain back.
• AGN feedback. The merger that triggered the starburst also feeds a central supermassive black hole. After a delay of 100–300 Myr, the AGN can reach a Quasi-Stellar Object phase, drive a faster (10⁴ km/s) wind that completes the job of unbinding the cold gas, and stamp the galaxy as red and dead. This is the "two-stage" merger-AGN coevolution sequence (Hopkins, Di Matteo, Springel) that produces present-day massive elliptical galaxies.

The signature of a recently quenched system is a "post-starburst" or E+A spectrum: weak emission lines (current SFR is low) but strong Balmer absorption (a population of A stars from the burst is still alive). Post-starburst galaxies are an observational link between active starbursts and quiescent ellipticals on Gyr timescales.

Common pitfalls

• Confusing AGN feedback with starburst feedback. Both are present in many ULIRGs and the line is not always clean. AGN winds are typically faster (10³–10⁴ km/s) and originate from the inner accretion disk; starburst winds are slower (10²–10³ km/s) and originate from the kpc-scale supernova-heated plasma. Diagnostics like soft X-ray morphology and emission-line ratios are used to disentangle them.
• Reading IR luminosity as bolometric without checking AGN contribution. A "ULIRG" of L_IR = 10¹² L☉ is not necessarily a starburst — it could be a heavily obscured Type-2 AGN. Mid-IR spectroscopy (Spitzer-IRS, JWST-MIRI) reveals AGN contributions through hot dust continua and high-ionisation lines, and is the standard way to budget the IR luminosity.
• Calling any high-SFR galaxy a "starburst". A main-sequence galaxy at cosmic noon can have an SFR of 100 M☉/yr — ten times the local Milky Way value — and yet not be a starburst, because its sSFR sits squarely on the main-sequence relation of that epoch. The definition is comparative: above the main sequence and short τ_dep.
• Inferring local Σ_SFR from disk-averaged Schmidt-Kennicutt. The disk-averaged relation has slope 1.4; resolved-mapping at sub-kpc scales gives different slopes (the molecular slope is closer to unity). Treating one as a substitute for the other will systematically over- or underpredict densities.
• Forgetting the IMF. SFR estimators (Hα, UV, IR, radio) all depend on an assumed initial mass function. Switching from a Salpeter to a Kroupa IMF changes inferred SFRs by about 0.2 dex. Compare published numbers only after harmonising.

Where starburst physics shows up

• Cosmic baryon budget. Starburst winds are the main mechanism for putting metals into the CGM and the IGM. The high-z Lyman-alpha forest and quasar absorption-line systems are direct probes of this enrichment.
• Mass-metallicity relation. Lower-mass galaxies are more efficient at losing wind-driven metals, leading to a clear correlation between stellar mass and gas-phase oxygen abundance — a probe of past starburst-driven outflows.
• Galactic archaeology. Globular cluster systems in elliptical galaxies preserve a population of "in-situ" clusters formed during the central merger-driven starburst that built the elliptical; abundance patterns match the chemical signatures predicted for short-duration high-SFR events.
• Gravitational-wave progenitors. A starburst produces hundreds of massive close binary systems per Myr, the progenitors of binary black-hole and binary neutron-star mergers detected by LIGO/Virgo. Star-formation history shapes the merger-rate density observed locally.
• Cosmic-ray production. The supernova rate in a starburst is two orders of magnitude higher than in a quiet galaxy; the resulting cosmic-ray luminosity and gamma-ray flux are detected from M82 and NGC 253 by Fermi-LAT and ground-based Cherenkov arrays.