Dwarf Galaxy

What counts as a dwarf

The conventional cut is a stellar mass below 10⁹ solar masses. By that line the Milky Way (M★ ≈ 6 × 10¹⁰ M☉) and Andromeda (M★ ≈ 10¹¹ M☉) are giants; the Large Magellanic Cloud (M★ ≈ 2 × 10⁹ M☉) sits exactly on the border and is sometimes counted as a dwarf, sometimes not. Below 10⁹ M☉ stretches an enormous population of systems descending in luminosity by more than five orders of magnitude — from the LMC's three billion suns of light down to ultra-faint dwarfs with fewer than a thousand stars.

Dwarf galaxies are not just small spirals. Their structure and stellar populations are qualitatively different. They lack the rotationally supported disks of bigger galaxies; even those with measurable rotation typically have v_rot/σ_v ≲ 1, meaning random motion dominates over orderly orbit. They have shallow gravitational potentials, low star-formation rates, and — critically — mass-to-light ratios far higher than anything explained by stars alone. The dwarf-galaxy regime is where dark matter dominates the visible kinematics most clearly, and where the smallest possible galaxies bump up against the boundary with what we'd otherwise call a star cluster.

Three families: dE/dSph, dIrr, UFD

Within the dwarf regime, three broad morphologies cover almost every system.

Type	Example	M★	Gas?	Stars	Setting
Dwarf elliptical (dE)	M32, NGC 205	10⁸–10⁹ M☉	Little	Old + intermediate	Near big galaxies
Dwarf spheroidal (dSph)	Fornax, Sculptor, Draco	10⁵–10⁷ M☉	None	Old, low metal	Satellites
Dwarf irregular (dIrr)	SMC, IC 10, NGC 6822	10⁷–10⁹ M☉	Gas-rich	Young + old	Often isolated
Blue compact dwarf (BCD)	I Zw 18	10⁷–10⁸ M☉	Intense starburst	Young, very metal-poor	Isolated
Ultra-faint dwarf (UFD)	Segue 1, Boötes II	10²–10⁴ M☉	None	Almost all old, [Fe/H] < -2	MW satellites
Tidal dwarf (TDG)	NGC 5291 N	10⁸–10⁹ M☉	Variable	Recycled from parent	Tidal debris

The dE/dSph distinction is partly historical: dE refers to brighter ellipsoidal dwarfs (M_V ≲ -14) usually found near big galaxies, while dSph denotes the much fainter spheroidal satellites typical of the Local Group. Both share the defining traits — pressure-supported, gas-poor, dominated by old stars. The dIrr class, by contrast, holds substantial HI gas, shows continuing star formation, and often has irregular or patchy morphology shaped by stellar feedback rather than smooth gravitational equilibrium. UFDs are the modern frontier: discovered en masse in SDSS in the mid-2000s and growing with every new survey, they push the lower mass limit of galaxies down by orders of magnitude.

By count, dwarfs are the universe

Galaxy luminosity functions — the Schechter function — predict roughly 10 times as many galaxies at L ~ 10⁹ L☉ as at L ~ 10¹⁰ L☉, with the count rising further toward fainter systems. Most of the visible mass in the universe lives in big galaxies, but most of the actual galaxies are dwarfs. The Milky Way alone has at least 60 confirmed satellite dwarfs (a count that is climbing as DES, Pan-STARRS, and DELVE continue to discover faint Milky Way companions); Andromeda has at least 35; the Local Group as a whole has more than 100. Cosmological simulations predict hundreds of bound dark-matter subhalos around each Milky-Way-mass host, most of which should be dwarfs if they form stars at all.

Outside the Local Group, dwarf surveys around nearby giants (M81, Centaurus A, the Virgo cluster) have catalogued hundreds more. The Coma cluster hosts roughly 10⁴ dwarf members. Across the universe, the cumulative dwarf population dwarfs (no pun intended) the giant-galaxy population by at least an order of magnitude.

Dark-matter laboratories

The most useful single fact about dwarf galaxies is their mass-to-light ratio. M/L is a dimensionless number relating the total dynamical mass within some aperture to the total visible-light luminosity, expressed in solar units. A pure stellar population gives M/L ≈ 1–3. A globular cluster, dominated by stars, gives M/L ≈ 1–2. A spiral galaxy disk gives M/L ≈ 3–6 (some inflation from dark matter at the edges). A dwarf galaxy gives anything from 5 to 1000+.

M_dyn / L  ≈  2.5 σ_v² r_h / (G · L)

For Draco: σ_v ≈ 9 km/s, r_h ≈ 220 pc, L_V ≈ 2 × 10⁵ L☉
   →   M_dyn(r_h) ≈ 2 × 10⁷ M☉
   →   M/L ≈ 200 in solar units

For Segue 1 — the ultra-faint dwarf with only about 1000 stars — the same calculation gives M/L ≈ 1300. The interpretation is forced: nearly all of the gravitational mass in these systems is invisible. They are dark-matter halos with a sprinkling of stars on top.

This makes dwarfs ideal laboratories for several dark-matter questions. The internal dynamics directly probe the density profile of the dark halo on sub-kiloparsec scales — testing the predicted NFW cusp against observed cores. The orbital structure tests whether dark matter is collisionless (cold dark matter) or scatters (self-interacting dark matter). And searches for annihilation gamma-ray emission from dwarfs — by Fermi-LAT and HESS — set the tightest astrophysical limits on the WIMP annihilation cross section, because dwarfs are nearby, dark-matter-rich, and background-quiet.

The three small-scale problems of ΛCDM

Dwarf galaxies sit precisely at the scale where cold dark matter cosmology runs into trouble. Three named tensions have animated the field since the late 1990s.

• Missing satellites problem. Pure dark-matter simulations of a Milky-Way-mass host (Klypin 1999; Moore 1999; later Aquarius and Via Lactea) predict hundreds to thousands of bound subhalos. The Milky Way has about 60 known satellites. The factor-of-10 discrepancy was the original puzzle. Modern resolution: many small subhalos never form luminous galaxies because reionization heated their gas above their escape speed (Bullock 2000), supernovae expelled what gas remained, and the smallest galaxies are observationally biased against detection. Current ΛCDM simulations with baryonic physics (FIRE, Auriga, Latte) reproduce the observed Milky-Way satellite luminosity function within uncertainty.
• Too-big-to-fail. Boylan-Kolchin (2011) noted that the densest predicted ΛCDM subhalos are too dense to match the observed densities of the brightest Milky-Way dwarfs. Those subhalos are "too big to fail" to form stars — they should be the bright dwarfs — yet the bright dwarfs are systematically less centrally dense. Resolutions invoke baryonic feedback flattening the central dark-matter cusp into a core (over Gyr timescales), or a slightly suppressed power spectrum on small scales as in warm dark matter.
• Core-cusp problem. Dwarf-galaxy rotation curves often appear to have flat or "cored" inner dark-matter density profiles, while CDM predicts a steep cusp ρ(r) ∝ r⁻¹ (NFW 1996). Observed cored profiles fit roughly ρ(r) → const inside ~1 kpc. The same baryonic-feedback mechanism that resolves too-big-to-fail also flattens cusps; alternatively, self-interacting dark matter or fuzzy dark matter would produce cores naturally. The empirical picture is mixed — some dwarfs fit cores, some fit cusps, and clean tests are hard because the inner regions are exactly where stellar feedback has its largest effects.

None of these tensions decisively breaks ΛCDM — each can be absorbed by realistic baryonic physics — but together they have made dwarfs the most actively contested testbed for cosmological dark-matter models.

Chemistry — pristine fossils

The chemical composition of a galaxy is set by the cumulative output of all its supernovae and stellar winds. Big galaxies have processed their gas through many generations and reached near-solar metallicity ([Fe/H] ≈ 0). Dwarfs, by contrast, fired one or a few rounds of star formation, lost their gas, and froze their abundances at sub-solar values.

System	Median [Fe/H]	Star formation	Notes
Milky Way disk	~0 (solar)	Continuous, ~10 Gyr	Many generations
MW halo	-1.5	Stopped ~10 Gyr ago	Probably dwarf debris
Fornax dSph	-1.0	Multi-burst	Brightest classical dSph
Draco / Sculptor	-2.0	Ancient single burst	Old classical dSph
UFDs (Segue 1, Reticulum II)	-2.5 to -3	Single ancient burst	Pre-reionization fossils
Lowest known stars	< -4	—	Mostly found in UFDs

UFDs in particular are the most chemically pristine systems known. Their stars formed within the first few hundred million years after the Big Bang, before reionization heated the cosmic web and shut off gas accretion into low-mass halos. Their metallicity distributions are tracers of pre-reionization nucleosynthesis — they encode the yields of the first generation of supernovae, and the relative abundances of r-process elements like europium are direct fingerprints of the rare events (likely neutron-star mergers) that produced them. Reticulum II, discovered in 2015, is famously r-process-enhanced and likely traces a single neutron-star merger that polluted its tiny gas reservoir.

Hierarchical assembly — being eaten

In ΛCDM cosmology, structure grows hierarchically: small dark-matter halos form first, merge into larger halos, and so on. Galaxies follow the halos. Today's massive galaxies were assembled by accreting hundreds or thousands of smaller progenitors over the past 13 billion years. Most of those progenitors were dwarfs.

The Milky Way is in the middle of this process right now. The Sagittarius dwarf spheroidal, discovered in 1994 by Rodrigo Ibata and collaborators, is the textbook case. Sagittarius is on a polar orbit about 25 kpc from the Galactic centre; tidal forces from the Milky Way are tearing it apart, and the stripped stars trail behind in a stream that wraps the entire Galaxy more than once. The Sagittarius Stream stretches over 360° on the sky, has been mapped in SDSS and 2MASS, and constrains the shape of the Milky Way's dark-matter halo to within a few tens of percent.

The Magellanic Clouds are on their first or second close passage to the Milky Way and have already shed the Magellanic Stream — a 200-kpc trail of HI gas. The Magellanic Bridge of younger stars between the LMC and SMC is a direct tidal feature. Several halo substructures (the Helmi Stream, the Gaia-Enceladus debris, the Sequoia substructure) are debris from earlier dwarfs that have already been fully shredded. Gaia's astrometric catalogues have made the Milky-Way halo a museum of fossil dwarfs.

The implication is general: every massive galaxy is partly a graveyard. The MW's halo stars are predominantly accreted, with the bulk delivered by 5 to 10 major dwarf progenitors of mass 10⁸ to 10⁹ M☉ that fell in roughly 10 Gyr ago. The chemical signatures of those progenitors remain in halo-star [α/Fe] distributions.

Worked example: M/L for Draco

Draco dSph is a classical example. Spectroscopic measurements of its red giants give a stellar velocity dispersion of σ_v ≈ 9.1 km/s, and its half-light radius from deep imaging is r_h ≈ 220 pc.

For a pressure-supported system in virial equilibrium, the mass within the half-light radius is approximately

M(r_h) ≈ 2.5 σ_v² r_h / G
       = 2.5 × (9.1 km/s)² × 220 pc / G
       ≈ 2.5 × 8.3 × 10¹⁰ cm²/s² × 6.8 × 10²⁰ cm / (6.67 × 10⁻⁸ cm³/g/s²)
       ≈ 2.1 × 10³⁹ g
       ≈ 1.1 × 10⁶ × 1 M☉  per dex shifting check
       ≈ 2 × 10⁷ M☉

Draco's V-band luminosity is L_V ≈ 2 × 10⁵ L☉. So

M/L  =  2 × 10⁷ M☉  /  2 × 10⁵ L☉  =  100  (solar units)

A stellar population alone, even one as old and red as Draco's, gives M/L of order 2-3. The observed value is 30 to 50 times higher. The only way to fit it is to add a massive, invisible dark-matter halo. This is the basic argument that has driven the field for forty years.

How we find them — surveys and methods

• Photometric overdensity searches. The Sloan Digital Sky Survey (SDSS, 2005-2010) revolutionised dwarf-galaxy hunting by allowing matched-filter detection of resolved old stellar populations over thousands of square degrees. Dozens of new UFDs were found this way. The Dark Energy Survey (DES, 2013-2019) and DELVE pushed deeper in the southern sky, discovering many more.
• Spectroscopic confirmation. An overdensity becomes a dwarf only when spectroscopy of member-star candidates yields a coherent radial velocity and (ideally) low metallicity. DEIMOS at Keck, FLAMES at the VLT, and AAOmega have been the workhorses; member counts as low as a few dozen are sometimes enough to claim a confident detection.
• Gaia astrometry. The Gaia mission's parallaxes and proper motions allow rejection of foreground Milky-Way stars and confirmation of coherent satellite-system motion. DR3 (2022) brought the Milky-Way satellite system into sharp focus.
• Rubin LSST (2025+). The Vera C. Rubin Observatory's Legacy Survey of Space and Time will detect all Milky-Way dwarfs down to ~10² L☉ across the whole southern sky, expected to push the known satellite count well over 100 and probably resolve the residual missing-satellites tension.
• Stream searches. Pal 5, GD-1, the Sagittarius Stream and dozens of "halo substructures" identified in SDSS / Gaia / DECam are the disrupted remnants of dwarfs or globular clusters. Their kinematics map the host gravitational potential and probe substructure within it.

Where the famous dwarfs are

• Sagittarius dSph. 26 kpc from the Sun, mass ~10⁸ M☉, currently being tidally disrupted into a stream that wraps the entire Milky Way more than once. Discovered 1994.
• Fornax dSph. 147 kpc away, the brightest classical dwarf spheroidal of the Milky Way (M★ ≈ 2 × 10⁷ M☉). Hosts five globular clusters of its own.
• Sculptor, Draco, Ursa Minor, Sextans, Carina, Leo I, Leo II. The "classical" Milky Way dSph satellites. Velocity dispersions of 7-12 km/s; M/L between 30 and 200.
• Large Magellanic Cloud. 50 kpc away, M★ ≈ 2 × 10⁹ M☉. Borderline dwarf / Magellanic-type irregular. Likely on its first close pass to the Milky Way.
• Small Magellanic Cloud. 63 kpc, M★ ≈ 5 × 10⁸ M☉, gas-rich and tidally distorted by the LMC and MW.
• M32, NGC 205, M110. Companions of M31. M32 is a compact dE (sometimes proposed to be the stripped core of a once-bigger galaxy); NGC 205 and M110 are more extended dwarf ellipticals.
• Segue 1, Boötes II, Willman 1, Reticulum II, Hercules. Ultra-faint dwarfs of the Milky Way. Total luminosities under 10⁴ L☉; M/L over 500.
• I Zw 18. The canonical blue compact dwarf — extremely metal-poor ([O/H] about 1/50 solar), bursty star formation. Local-Universe analog of the chemically pristine first galaxies.

Common pitfalls

• Confusing dwarf galaxies with globular clusters. Both are small and bound, but globulars have M/L ~ 2 and dwarfs have M/L of 5 to 1000+. The cleanest discriminator is dark matter — measured via velocity dispersion and modelled with Jeans equations or virial scaling. UFDs sit at the boundary; their high M/L is the criterion that classifies them as galaxies.
• Treating M/L as monolithic. M/L scales strongly with luminosity (Walker 2009; McConnachie 2012): the brightest dwarfs have M/L ~ 5-20, the faintest UFDs reach 10³. A single quoted M/L is meaningless without specifying which aperture (half-light radius? tidal radius?) and which dwarf.
• Assuming virial equilibrium for tidally disrupting systems. Dynamical mass estimates from σ_v rely on equilibrium. Sagittarius, the Magellanic Clouds, and Boötes are partially disrupted; their inferred dynamical masses are systematically biased high if equilibrium is naively assumed.
• Calling the missing-satellites problem "solved" or "unsolved". The problem has been quantitatively reduced — modern hydrodynamic simulations match observed counts within a factor of two. Whether the residual is fully explained by baryonic physics or hints at deviations from CDM remains open.
• Conflating dE and dSph. The two classes overlap but historically refer to brighter near-giant-host ellipsoidals (dE) and the much fainter classical satellites (dSph). Stellar-population properties differ; literature uses the labels loosely.