The Green Valley

Two crowds and an empty road

Take a few hundred thousand galaxies, measure the colour of each one, and plot colour against brightness (or against stellar mass). You do not get a smooth smear. You get two crowds. On the blue side sits a loose, scattered swarm — the blue cloud — of galaxies busily forming stars. On the red side sits a tight, narrow ridge — the red sequence — of galaxies that stopped. Between them is a road that almost nobody is standing on: the green valley.

The valley is not empty. Galaxies are constantly walking across it, blue to red, as their star formation switches off. But they walk fast. The crossing takes only one or two billion years, against a cosmic backdrop that has been running for 13.8 billion. So at any one moment — and a galaxy survey is exactly one moment, frozen — you catch only the handful currently in transit. The under-density of the valley is the whole point: it is the statistical fingerprint of a process that happens quickly. The fewer galaxies you find in the valley, the faster the shutdown must be.

This bimodality — two peaks with a dip between — was made unmistakable by the Sloan Digital Sky Survey in the early 2000s (Strateva 2001, Baldry 2004) and sharpened by ultraviolet data from the GALEX satellite, which is far more sensitive to recent star formation than optical light. The phrase "green valley" entered the literature around 2007 (Wyder et al., Salim, Martin) as a deliberate pun: on a rainbow, the colour between blue and red is green. No galaxy is literally green.

What the colour is actually telling you

A galaxy's broadband colour is a thermometer for its recent star formation. Massive O and B stars are hot, blue, and short-lived: an O star burns out in a few million years, a B star in tens of millions. So a galaxy that has formed stars within the last ~100 Myr still hosts these blue beacons and looks blue. Cut off the star formation, and within a few hundred million years the hot stars die, leaving only the long-lived, cool, red low-mass stars (G, K, M dwarfs that live for billions of years). The galaxy reddens.

The single most useful quantity is the specific star formation rate, sSFR, the star formation rate divided by the stellar mass already present:

sSFR = SFR / M*        [units: yr^-1]

1 / sSFR = stellar-mass doubling time at the current rate

Star-forming galaxies obey a tight relation called the star-forming main sequence, SFR ∝ M*^β with β ≈ 0.7–1.0, clustering near sSFR ≈ 10⁻¹⁰ to 10⁻⁹ yr⁻¹. Quiescent galaxies sit a factor of 10–100 below this, at sSFR ≲ 10⁻¹¹ yr⁻¹. The green valley is the band in between, roughly:

Blue cloud:    sSFR ≈ 10^-10   to 10^-9    yr^-1
Green valley:  sSFR ≈ 10^-11.5 to 10^-10.5 yr^-1   (transition)
Red sequence:  sSFR ≲ 10^-11               yr^-1

Equivalently, a green-valley galaxy is one that has fallen roughly 0.5–1.5 dex below the main sequence but has not yet reached full quiescence. It is mid-fall.

How the valley is measured — and why UV matters

The cleanest way to separate the two populations is not optical colour but ultraviolet-to-optical colour, especially NUV − r (near-ultraviolet minus r-band). The NUV band is dominated by young stars, so NUV − r is extremely sensitive to even low levels of residual star formation — it stretches the colour axis and deepens the valley. In NUV − r, the blue cloud sits near NUV − r ≈ 1–2, the red sequence near NUV − r ≈ 5–6, and the valley spans roughly 3–4 mag.

There is a crucial complication: dust. A dusty, edge-on, star-forming spiral can be reddened by extinction until it masquerades as a valley galaxy in optical colour. Three remedies separate genuine transition galaxies from dust-reddened impostors:

• Use a colour–mass diagram, not colour–magnitude. Stellar mass is dust-robust (derived from near-infrared light), so plotting colour against M* tightens the populations.
• Use the UVJ or NUVrK diagram. Two colours together (e.g. U−V vs V−J) separate the dust-reddening vector from the age-reddening vector — they point in different directions — so quiescent and dusty-star-forming galaxies land in different regions.
• Measure sSFR directly from spectral lines (Hα), infrared dust emission, or SED fitting, sidestepping broadband colour entirely.

Done carefully, all of these recover the same conclusion: a real, physically meaningful minimum in the galaxy distribution, populated by genuinely transitioning systems.

Blue cloud vs green valley vs red sequence

Property	Blue cloud	Green valley	Red sequence
Dominant morphology	Spirals, irregulars	S0, early spirals, post-mergers	Ellipticals, lenticulars
NUV − r colour	~1–2	~3–4	~5–6
sSFR (yr⁻¹)	10⁻¹⁰–10⁻⁹	10⁻¹¹·⁵–10⁻¹⁰·⁵	≲ 10⁻¹¹
Stellar population age	≲ 1 Gyr (luminosity-weighted)	1–3 Gyr	≳ 5–10 Gyr
Locus shape	Broad, scattered cloud	Sparse bridge	Tight, narrow ridge
Cold gas fraction	High (10–50%)	Declining / depleted	Low (≲ few %)
Central black hole	Often low-luminosity	Elevated AGN fraction	Massive, often radio-mode
Number density	High peak	Few× below peaks	High peak

The single most diagnostic line in this table is the bottom one. The valley is a few times under-dense, not orders of magnitude — exactly what you expect for a transition that takes a non-negligible-but-short fraction of cosmic time.

Putting numbers on the crossing time

How fast do galaxies cross? You can read it two ways, and they agree.

Demographic argument. If the fraction of galaxies currently in the valley is f_GV and the typical crossing time is τ_cross, then in a population with a steady flux of galaxies quenching, the valley fraction is roughly the crossing time divided by the time since galaxies started leaving the blue cloud. Observed valley fractions of a few to ~10% over the last several billion years imply

τ_cross ≈ 1 – 2 Gyr   (optical colour crossing)
        ≪ t_Hubble = 13.8 Gyr

Stellar-population argument. Once star formation truncates, the integrated colour reddens on the timescale at which blue stars die and the light becomes dominated by the older population — a few hundred Myr to about 1 Gyr for the bluest part of the swing, longer to fully settle onto the red sequence. Slower, "feathered" declines (an exponential star formation history with e-folding time τ_SFH) stretch the crossing; abrupt truncations make it fast.

A useful framing is the delayed-then-rapid picture from cosmological studies (e.g. Schawinski et al. 2014): early-type and late-type galaxies cross the valley by different routes. Spirals fade slowly — gas is gradually exhausted, sSFR coasts down over a couple of Gyr — while ellipticals (often the products of mergers) are quenched abruptly and shoot across in a few hundred Myr. The valley is the same on the diagram, but two different vehicles are driving through it.

What turns the lights off

Reaching the valley means losing the ability to form stars, which means losing, heating, or stabilising the cold gas reservoir. The leading mechanisms, and the regimes where each dominates:

Mechanism	What it does	Where it dominates	Speed
AGN feedback (radio / maintenance mode)	Hot-halo heating prevents cooling; jets/winds expel gas	Massive galaxies, M* ≳ 10¹⁰·⁵ M☉	Slow, sustained
AGN feedback (quasar / wind mode)	Radiation-driven outflow blows out gas	Post-merger, gas-rich nuclei	Fast (≲ 1 Gyr)
Morphological quenching	Stellar bulge stabilises disk against fragmentation	Bulge-dominated, S0 galaxies	Gradual
Gas exhaustion + halted supply	Cold gas used up, no fresh inflow ("strangulation")	Isolated and satellite galaxies	~Few Gyr
Ram-pressure stripping	Hot intracluster gas sweeps cold gas out	Cluster/group satellites	Fast (≲ 1 Gyr)
Merger-driven starburst + blowout	Burst consumes gas, then feedback clears the rest	Major gas-rich mergers	Fast

A clean way to organise this is the distinction between mass quenching and environment quenching (Peng et al. 2010). Mass quenching scales with the galaxy's own stellar mass — internal processes (AGN, bulge) that grow with mass. Environment quenching scales with local galaxy density — external processes (stripping, strangulation) in groups and clusters. The two are separable in the data and act roughly independently, and both funnel galaxies into the same valley.

Why the central black hole keeps coming up

Massive quiescent galaxies almost universally host a supermassive black hole whose mass tracks the bulge via the M–σ relation, M_BH ≈ 1.5–3 × 10⁸ M☉ × (σ / 200 km s⁻¹)⁴·⁴. That correlation is a strong hint that black-hole growth and the cessation of star formation are linked. The current consensus is "maintenance mode": once a hot gaseous halo forms around a massive galaxy, mechanical energy from the black hole (radio jets inflating X-ray cavities, as seen in the Perseus and Virgo (M87) clusters) reheats the halo just enough to stop it cooling and refuelling the disk. Energetically the bookkeeping closes easily — the binding energy released by growing a 10⁸ M☉ black hole, ~10⁶¹ erg, exceeds the binding energy of the entire galaxy's gas. Even a percent of that, coupled to the gas, is more than enough to quench. The black hole does not need to expel the gas in one blast; it just has to keep the kettle from boiling.

Where you can point at it

• The Sloan bimodality. The defining dataset: ~10⁵ low-redshift galaxies showing two clean peaks in u − r with a valley between, the textbook colour–magnitude diagram (Baldry et al. 2004).
• The Milky Way. Our own Galaxy is a green-valley galaxy. Its star formation rate of ~1–2 M☉/yr at a stellar mass of ~6 × 10¹⁰ M☉ gives sSFR ≈ 2–3 × 10⁻¹¹ yr⁻¹ — squarely in the transition band. The Milky Way is slowly fading rather than vigorously forming stars; it likely entered the valley a few billion years ago.
• S0 (lenticular) galaxies. The disk-but-no-spiral-arms morphology is the archetypal valley resident — a faded spiral that has used up or lost its gas.
• Post-starburst (E+A / K+A) galaxies. Galaxies showing strong Balmer absorption (an A-star population formed ≲ 1 Gyr ago) but no current emission lines — caught just after an abrupt truncation, sprinting across the valley.
• Cluster infall regions. Spirals falling into the Virgo and Coma clusters are being ram-pressure stripped (the "jellyfish" galaxies with trailing gas tails) and reddening as they go — environment quenching captured in the act.

The valley deepens with cosmic time

Bimodality is not static. At high redshift, when the universe was young and gas-rich, almost all galaxies were forming stars: the red sequence was thin and the valley shallow. As cosmic time passed, the red sequence grew — its total stellar mass roughly doubling since z ≈ 1 (about 8 billion years ago) — fed by a steady stream of galaxies crossing the valley. This is "downsizing": the most massive galaxies quenched first and earliest, while lower-mass galaxies kept forming stars until more recently. The build-up of the red sequence is, quite literally, the accumulated traffic that has crossed the green valley over cosmic history.

Common misconceptions and edge cases

• "The valley is empty." It is a density minimum, only a few times below the peaks — not a void. Galaxies are there; they are just rare because they pass through quickly.
• "Galaxies are literally green." No broadband-measured galaxy peaks in green. The name is a position on the colour axis between blue and red. The integrated light of any real stellar population plus dust never produces a net green colour.
• "Optical colour alone defines it." Dust can redden a star-forming galaxy into the optical valley. Use NUV − r, a colour–mass diagram, or the UVJ two-colour diagram to remove dusty impostors before claiming a transition.
• "The valley is one-directional." Most traffic flows blue → red, but rejuvenation (a gas-rich minor merger, fresh cold accretion) can push a galaxy back toward blue. A few percent of red/green galaxies show such recent rejuvenation.
• "Crossing time is the same for everyone." Early types quench fast (≲ 1 Gyr, often merger-driven); late types fade slowly (~2 Gyr, gas-exhaustion-driven). The "delayed-then-rapid" picture means a single average crossing time hides two distinct routes.
• "Mass and environment do the same thing." They are statistically separable. Mass quenching depends on the galaxy's own M*; environment quenching depends on local density. A massive isolated galaxy and a low-mass cluster satellite can both end up red, by entirely different roads.