Mass–Metallicity Relation

The richest galaxies keep their gold

Line up galaxies by stellar mass and ask a single chemical question of each one — how much of its gas is made of elements heavier than helium? — and a remarkably clean pattern appears. The smallest dwarfs are metal-poor, their interstellar gas barely a few percent of the Sun's heavy-element content. The largest spirals and ellipticals are metal-rich, with gas approaching or exceeding solar abundance. The trend is monotonic across three full decades of stellar mass, and the scatter around it is tiny: about 0.1 dex. This is the mass–metallicity relation, or MZR, and it is one of the tightest and most informative scaling relations in all of galaxy evolution.

The deep reason is not that big galaxies make more metals per unit of star formation — to first order they all make the same metals from the same nuclear physics. The reason is that big galaxies keep their metals. Every burst of star formation is followed by supernovae that both manufacture heavy elements and blast metal-loaded gas outward in a galactic wind. Whether that enriched gas escapes the galaxy or rains back down to form new stars depends on one number: the depth of the gravitational potential well. Shallow dwarf potentials let the wind win, stripping the metals out into the circumgalactic medium. Deep massive potentials hold the enriched gas bound, recycling it. The mass–metallicity relation is the fingerprint of that competition, written in the chemistry of every galaxy we can measure.

How it works: produce, then either keep or lose

Three processes set a galaxy's gas-phase metallicity, usually written as the oxygen abundance 12+log(O/H), where the solar value is about 8.69.

• Production. Massive stars fuse hydrogen and helium into oxygen, neon, magnesium and silicon, and core-collapse supernovae disperse those metals into the interstellar medium within a few million years. The yield per unit stellar mass formed is set by stellar physics and the initial mass function, and is roughly universal.
• Dilution. Pristine or near-pristine gas accreting from the cosmic web and the circumgalactic medium dilutes the enriched gas, lowering the measured metallicity even as the absolute mass of metals climbs.
• Removal. Supernova energy and momentum, plus radiation pressure from young stars, drive galactic winds that carry metal-rich gas out of the galaxy. Because the freshly synthesized metals are concentrated in the hot, fast-moving phase of the wind, outflows preferentially remove the newest metals.

The removal term is the one that scales strongly with galaxy mass, and so it is the term that sculpts the relation. Its strength is parameterized by the mass-loading factor η, the ratio of the mass outflow rate to the star-formation rate. Energy-driven and momentum-driven wind models both predict that η falls steeply with the galaxy's circular velocity — roughly η ∝ v_circ⁻² for energy-driven winds. A dwarf with v_circ ≈ 50 km/s can have η ≈ 5–20, ejecting several times more gas than it converts into stars and exporting most of its freshly made oxygen. A massive galaxy with v_circ ≈ 200 km/s has η well below 1; it loses almost nothing. That single scaling, fed through a chemical-evolution model, reproduces the observed slope of the MZR with no fine-tuning.

Worked example: a dwarf versus a giant

Consider two galaxies forming stars at the same specific rate but with circular velocities differing by a factor of four: a dwarf at v_d = 50 km/s and a giant at v_g = 200 km/s. Take an energy-driven wind with η = η₀ (v/50 km/s)⁻², normalized so η = 8 at 50 km/s.

Dwarf  (v = 50 km/s):  η = 8 × (50/50)⁻²  = 8.0
Giant  (v = 200 km/s): η = 8 × (200/50)⁻² = 8 / 16 = 0.5

Now use the simplest analytic result for a galaxy in equilibrium between inflow, star formation and outflow — the "leaky-box" or gas-regulator effective yield. If a true nucleosynthetic yield y produces metals, but a fraction is carried away in the wind, the equilibrium metallicity is approximately

Z_eq ≈ y / (1 + η)

Plugging in the two galaxies:

Z_dwarf ≈ y / (1 + 8.0) = y / 9.0   →  0.11 y
Z_giant ≈ y / (1 + 0.5) = y / 1.5   →  0.67 y

Ratio  Z_giant / Z_dwarf ≈ 6×

The giant equilibrates at six times the metallicity of the dwarf despite using the same stellar physics. Converting to the observed logarithmic axis, a factor of six is 0.78 dex — close to the ≈ 1 dex span (12+log(O/H) ≈ 8.0 → 9.0) seen across the real SDSS relation from 10⁸·⁵ to 10¹⁰·⁵ M☉. The numbers are illustrative rather than precise, but they show that the wind-loading scaling alone, with no extra ingredients, lands in the right ballpark. This is why galactic outflows are considered the dominant driver of the mass–metallicity relation.

The relation in numbers

The canonical local measurement is Tremonti et al. (2004), who derived gas-phase oxygen abundances for ~53,000 star-forming galaxies in the Sloan Digital Sky Survey. The relation rises steeply at low mass and flattens above a turnover mass of about 10¹⁰·⁵ M☉, with a scatter of only ≈ 0.1 dex at fixed mass. Below is a representative tabulation across the mass range, including the well-studied Local Group dwarfs that anchor the low-mass end and the high-redshift offsets that define its cosmic evolution.

System / mass bin	log(M*/M☉)	12+log(O/H)	Z / Z☉ (approx)	Note
Leo P (dwarf irregular)	~ 5.7	7.17	~ 0.03	extreme metal-poor
SMC	~ 8.7	8.0	~ 0.2	nearby dwarf
LMC	~ 9.2	8.35	~ 0.5	nearby dwarf
SDSS bin (low)	8.5	~ 8.0	~ 0.2	Tremonti 2004
SDSS bin (mid)	9.5	~ 8.7	~ 1.0	near solar
SDSS bin (turnover)	10.5	~ 9.0	~ 2.0	relation flattens
Milky Way (ISM)	~ 10.7	~ 8.7	~ 1.0	solar neighborhood
z ≈ 2 (M* = 10¹⁰)	10.0	~ 8.4	~ 0.5	~0.3 dex below local
z ≈ 3.5 (M* = 10¹⁰)	10.0	~ 8.1	~ 0.3	up to ~0.6 dex below local

Two trends jump out. Reading down the local rows, metallicity climbs steadily with stellar mass and then saturates near the turnover — that is the relation itself. Comparing the bottom two rows against the SDSS bin at the same mass, the high-redshift galaxies sit well below the local relation — that is its evolution with cosmic time.

Variants and regimes

• Gas-phase MZR. The classic relation, measured from HII-region emission lines and expressed in 12+log(O/H). It reflects the present-day state of the interstellar medium and is the version that evolves most clearly with redshift.
• Stellar MZR. Measured from absorption features in integrated starlight (Gallazzi 2005 for SDSS; Kirby 2013 for Local Group dwarf spheroidals down to M* < 10⁴ M☉). It records the time-averaged enrichment over the galaxy's entire star-formation history and extends to far lower masses than emission-line work allows.
• Fundamental Metallicity Relation (FMR). Mannucci et al. (2010) showed the MZR is a projection of a tighter 3D surface in mass–metallicity–star-formation-rate space: at fixed mass, higher-SFR galaxies are more metal-poor. The FMR appears nearly invariant out to z ≈ 2.5, suggesting that the raw MZR's redshift evolution is largely driven by higher gas accretion and star-formation rates in the early universe.
• Resolved MZR. Integral-field surveys (MaNGA, CALIFA, SAMI) reveal a local relation between stellar-mass surface density and metallicity within individual galaxies — the global MZR partly emerges from a kpc-scale relation plus radial metallicity gradients.
• High-mass turnover and downsizing. Above ≈ 10¹⁰·⁵ M☉ the gas-phase relation flattens as winds become ineffective and enrichment saturates near the effective yield; this regime overlaps with quenching, where star formation itself shuts down.

Quantitative analysis: the gas-regulator model

The slope of the MZR follows from a simple equilibrium argument. Treat a galaxy as a reservoir of gas mass M_gas that is fed by inflow at rate Ṁ_in, drained by star formation at rate Ψ = SFR, and drained again by an outflow Ṁ_out = η Ψ. A fraction R of the mass formed into stars is promptly returned by dying stars, and each unit of star formation produces a mass y of new metals. The metal mass M_Z then obeys

dM_Z/dt = y·Ψ  −  Z·(1−R)·Ψ  −  Z_out·η·Ψ  +  Z_in·Ṁ_in

where Z = M_Z / M_gas is the gas metallicity. In the equilibrium ("steady-state") limit dM_Z/dt → 0, with metal-loaded winds (Z_out ≈ Z) and near-pristine inflow (Z_in ≈ 0), this rearranges to

Z_eq = y / [ (1 − R) + η ]

This is the central result. With (1 − R) of order 0.6 and η falling from ≈ 8 in dwarfs to ≈ 0.5 in giants, Z_eq rises by roughly an order of magnitude across the mass range — the observed MZR slope. Two limits are instructive. When η ≫ 1 (deep dwarf regime) Z_eq ≈ y/η ∝ v_circ², so metallicity tracks the potential-well depth directly. When η ≪ 1 (massive regime) Z_eq → y/(1−R) = the closed-box effective yield, a ceiling independent of mass — which is exactly the flattening seen above the turnover. The redshift evolution then enters through the gas fraction and accretion rate: at high z, galaxies sit further from equilibrium, accrete more pristine gas, and form stars faster, all of which push Z_eq down at fixed mass and reproduce the observed ≈ 0.3–0.6 dex offsets.

Observational status and reach

The local relation is among the most secure results in extragalactic astronomy, built on tens of thousands of SDSS spectra and confirmed by every subsequent large survey. The redshift evolution was established first by rest-frame optical spectroscopy with Keck and the VLT: Erb et al. (2006) at z ≈ 2 and the AMAZE/LSD programs (Maiolino et al. 2008) at z ≈ 3–3.5, finding offsets of 0.3 and up to 0.6 dex below local at fixed stellar mass. JWST has now extended gas-phase metallicity measurements to z > 8 using the same auroral-line and strong-line diagnostics redshifted into the near-infrared, generally finding galaxies a few tenths of a dex more metal-poor than even the z ≈ 3 relation, in line with younger systems that have processed less of their gas.

• Constraining feedback. The slope and turnover of the MZR are among the most demanding benchmarks for cosmological simulations (EAGLE, IllustrisTNG, FIRE); reproducing it forces the wind mass-loading to scale roughly as v_circ⁻² — a direct calibration of stellar feedback.
• Mapping the baryon cycle. Because the missing metals end up in the circumgalactic medium, the MZR and CGM metal budgets together close the cosmic baryon and metal accounting.
• A cosmic clock for chemistry. The evolving relation tracks when and how fast galaxies built up their heavy-element content, complementing the cosmic star-formation history.
• Anchoring the dwarf end. Local Group dwarfs, resolvable star by star, tie the low-mass stellar MZR to individual stellar abundances, testing the same wind physics down to M* < 10⁵ M☉.

Common pitfalls and misconceptions

• "Massive galaxies make more metals per star." They do not, to first order — the nuclear yields and IMF are roughly universal. They retain more of the metals they make. The MZR is a retention relation, not a production relation.
• Mixing metallicity calibrations. Different strong-line calibrations (R23, N2, O3N2, T_e direct) can differ by up to ~0.6 dex in absolute scale and even in the shape of the relation. Comparisons across studies must use a consistent calibration, or the redshift evolution can be mimicked or erased by calibration choice alone.
• Confusing gas-phase and stellar metallicity. The gas-phase MZR reflects today's interstellar medium; the stellar MZR reflects the integrated past. They are offset and respond to different timescales — neither is "the" metallicity.
• Treating the flattening as purely physical. Part of the high-mass turnover may be an artifact of strong-line indicators saturating at high O/H. The true asymptotic metallicity and turnover mass depend on the diagnostic used.
• Ignoring the SFR axis. Quoting an MZR without controlling for star-formation rate hides the FMR. Much of the apparent redshift evolution of the 2D relation is absorbed once the third dimension is included.
• Assuming the wind metallicity equals the ISM metallicity. Outflows are often more metal-rich than the ambient gas because they carry the freshly synthesized supernova ejecta preferentially, which strengthens the stripping of metals from dwarfs beyond a naive estimate.