Ursa Major III/UNIONS I: Cluster or Dwarf?

• Ursa Major III/UNIONS I is an extremely faint Milky Way satellite exhibiting characteristics that blur the distinction between ultra-faint dwarf galaxies and star clusters.
• The discovery employed deep imaging and matched-filter techniques combined with spectroscopic follow-ups to assess its compactness, luminosity, and ambiguous dark-matter signature.
• Recent multi-epoch spectroscopy and advanced dynamical models favor a stellar-only interpretation, underscoring the need for further observations to resolve its true nature.

Ursa Major III/UNIONS 1, often abbreviated UMa3/U1, is an extremely faint, compact Milky Way satellite discovered in deep UNIONS imaging and initially characterized as the least luminous known satellite of the Galaxy. At a heliocentric distance of approximately $10$ kpc, with a half-light radius of $3 \pm 1$ pc and a total stellar mass of $16^{+6}_{-5}\,M_\odot$, it occupies an unusually compressed region of the size–luminosity plane where the empirical distinction between ultra-faint dwarf galaxies and star clusters becomes uncertain. Its scientific importance derives from that ambiguity: early equilibrium interpretations of its kinematics implied extreme dark-matter domination and exceptional annihilation $J$-factors, whereas later spectroscopic and collisional-dynamical analyses argued that a stellar-only interpretation remains viable and may now be preferred (Smith et al., 2023, Errani et al., 2023, Cerny et al., 2 Oct 2025).

1. Discovery, survey context, and nomenclature

Ursa Major III/UNIONS 1 was uncovered in the Ultraviolet Near Infrared Optical Northern Survey using a matched-filter search for old, metal-poor stellar overdensities in CFIS-$r$ and Pan-STARRS-$i$ imaging over $>3500\,\mathrm{deg}^2$. The search applied PARSEC isochrone filters over trial distances from 10 kpc to 1 Mpc, binned candidate stars into $0.5'\times0.5'$ pixels, smoothed the maps with Gaussian kernels of FWHM $1.2'$, $2.4'$, and $3 \pm 1$0, and identified the object as a $3 \pm 1$1 overdensity at a trial distance of $3 \pm 1$2 kpc (Smith et al., 2023).

The dual designation reflects the unresolved nature of the system. The discovery literature used the combined name “Ursa Major III/UNIONS 1,” while later work commonly shortened this to “UMa3/U1.” This suggests that the naming convention itself tracks the classification problem: “Ursa Major III” is typically used when the system is discussed as a dwarf galaxy, whereas “UNIONS 1” or “U1” appears in contexts where a star-cluster interpretation is entertained (Devlin et al., 30 Apr 2025).

The object immediately drew attention because it combined three properties that are rarely seen together: extreme faintness, extreme compactness, and close proximity. That combination makes it unusually informative for both chemodynamical classification and dark-matter phenomenology.

2. Stellar population and structural properties

Discovery-era photometric modeling described the system as an old, metal-poor stellar population at $3 \pm 1$3, corresponding to $3 \pm 1$4. The CMD was matched with $3 \pm 1$5 isochrones and $3 \pm 1$6, with the age quoted conservatively as $3 \pm 1$7 (Smith et al., 2023).

Property	Reported value	Source
Heliocentric distance	$3 \pm 1$8	(Smith et al., 2023)
Half-light radius	$3 \pm 1$9	(Smith et al., 2023)
Absolute magnitude	$16^{+6}_{-5}\,M_\odot$0	(Smith et al., 2023)
Stellar mass	$16^{+6}_{-5}\,M_\odot$1	(Smith et al., 2023)
Stars above $16^{+6}_{-5}\,M_\odot$2	$16^{+6}_{-5}\,M_\odot$3	(Smith et al., 2023)
Ellipticity	$16^{+6}_{-5}\,M_\odot$4	(Smith et al., 2023)
Position angle	$16^{+6}_{-5}\,M_\odot$5	(Smith et al., 2023)

The structural fit used an elliptical exponential surface-density profile plus constant background contamination. In that framework, the angular half-light radius is $16^{+6}_{-5}\,M_\odot$6, the effective surface brightness is $16^{+6}_{-5}\,M_\odot$7, and the total population extrapolated to $16^{+6}_{-5}\,M_\odot$8 is $16^{+6}_{-5}\,M_\odot$9 stars. A Monte Carlo population synthesis based on a Kroupa IMF and the observed number of stars brighter than $J$0 yielded the stellar mass estimate $J$1, implying $J$2 for an empirical $J$3 (Smith et al., 2023).

Subsequent dynamical and radio studies recast the stellar light profile for spherical modeling as

$J$4

with

$J$5

while the radio analysis also used $J$6 pc in defining the diffusion region and explicitly noted that the values differ slightly because they were drawn from different analyses. The consistent picture is that of an exceptionally compact system with a characteristic scale of only a few parsecs (Zhao et al., 2024, Zhang et al., 2024).

3. Kinematics, orbital context, and the initial dark-matter interpretation

Keck II/DEIMOS follow-up spectroscopy identified 11 radial-velocity members, 8 of which also had full Gaia astrometry and were co-moving in proper motion. Using a Gaussian likelihood for the line-of-sight velocities, the discovery analysis derived a systemic heliocentric velocity of $J$7 and an intrinsic velocity dispersion of

$J$8

However, the result was immediately recognized as fragile: excluding the largest velocity outlier reduced the dispersion to

$J$9

and removing an additional influential star rendered the dispersion unresolved (Smith et al., 2023).

If that km s$r$0-level dispersion were interpreted as equilibrium support, the implied dynamical mass would be extreme for such a tiny stellar system. The discovery paper, using the Wolf et al. estimator, reported

$r$1

for the full 11-star sample, while the one-outlier-removed case gave

$r$2

That parameter regime is ordinarily associated with ultra-faint dwarfs rather than star clusters (Smith et al., 2023, Rostami-Shirazi et al., 14 Aug 2025).

Errani et al. sharpened the argument by comparing those measurements with a purely stellar expectation. For an exponential stellar system with $r$3 and $r$4 pc, the projected virial theorem gives

$r$5

far below the discovery-era km s$r$6 estimates. The same work derived a pericenter of $r$7 kpc, an apocenter of $r$8 kpc, and a radial orbital period of $r$9 Gyr, and then used $i$0-body simulations to show that a self-gravitating stellar system with the observed present-day mass and size would be tidally destroyed within $i$1 Gyr, i.e. not much longer than a single orbit. On that basis, it argued that a substantial dark halo, of order $i$2, is required to stabilize the system and that such a halo would predict $i$3 of order $i$4 (Errani et al., 2023).

This produced the initial “darkest galaxy ever discovered?” framing. A plausible implication was that UMa3/U1 might represent an ultra-compact, ultra-faint dwarf at the extreme lower edge of the galaxy luminosity function rather than an unusually depleted star cluster.

4. Indirect-detection applications and the $i$5-factor literature

Because of its proximity and the possibility of an unusually concentrated dark halo, Ursa Major III rapidly became a major target in indirect dark-matter searches. The relevant astrophysical factor is

$i$6

with extensions to effective $i$7-factors in velocity-dependent annihilation scenarios (Zhao et al., 2024).

An early Fermi-LAT analysis of 15 years of data found no $i$8-ray excess from the source position. Adopting a fiducial $i$9, it concluded that, if the high $>3500\,\mathrm{deg}^2$0-factor were confirmed, standard thermal dark matter annihilation to $>3500\,\mathrm{deg}^2$1 would be ruled out up to $>3500\,\mathrm{deg}^2$2 TeV, with corresponding exclusions up to $>3500\,\mathrm{deg}^2$3 TeV for $>3500\,\mathrm{deg}^2$4 and $>3500\,\mathrm{deg}^2$5 TeV for $>3500\,\mathrm{deg}^2$6 (Crnogorčević et al., 2023).

A subsequent Jeans analysis used a spherical NFW halo, constant anisotropy, a circular exponential light profile, and the CLUMPY/GreAT unbinned likelihood on the 11-star kinematic sample. For an integration angle of $>3500\,\mathrm{deg}^2$7, it obtained

$>3500\,\mathrm{deg}^2$8

together with velocity-dependent effective factors

$>3500\,\mathrm{deg}^2$9

Under the 11-star interpretation, these values made Ursa Major III one of the most constraining individual targets for WIMP annihilation, particularly for Sommerfeld-enhanced models. The same paper also showed that excluding the largest velocity outlier reduced the median $0.5'\times0.5'$0-factor by several dex: $0.5'\times0.5'$1 without a tidal cut, or

$0.5'\times0.5'$2

when restricting to profiles with $0.5'\times0.5'$3. This identified member-star selection as the dominant systematic in the annihilation analysis (Zhao et al., 2024).

A radio study then treated Ursa Major III as a case study for synchrotron and inverse-Compton emission from annihilation-produced $0.5'\times0.5'$4, assuming an NFW halo and a 100 hour SKA observation. In its optimistic scenario $0.5'\times0.5'$5, the quoted sensitivity reached $0.5'\times0.5'$6 for $0.5'\times0.5'$7 and $0.5'\times0.5'$8 channels from several GeV to $0.5'\times0.5'$9 GeV, while emphasizing substantial uncertainty from magnetic field strength, diffusion coefficient, and the dark-matter density profile (Zhang et al., 2024).

Later spectroscopic revision materially altered that picture. Using updated kinematics, a 2025 analysis inferred

$1.2'$0

and recommended excluding UMaIII/U1 from indirect-detection samples, not because the reduced $1.2'$1-factor would be intrinsically uninteresting, but because the system no longer showed observational evidence for dark matter in the first place (Cerny et al., 2 Oct 2025).

5. Star-cluster and dark-star-cluster alternatives

The principal astrophysical counterargument to the dark-halo interpretation emerged from collisional $1.2'$2-body modeling. Devlin, Baumgardt, and Sweet modeled UMa3/U1 as a stellar system with stellar evolution, compact-remnant formation, two-body relaxation, a Galactic tidal field, and, in some runs, primordial binaries. They found that a dark-matter-free star cluster could survive for a substantial remaining lifetime of

$1.2'$3

primarily because of compact-remnant retention, mass segregation, and preferential loss of low-mass stars. In their primordial-binary runs with $1.2'$4, the luminosity-weighted present-day dispersion reached

$1.2'$5

whereas the single-star-only dispersion remained

$1.2'$6

The implied overestimate in virial mass, if binary motions are misinterpreted as equilibrium support, is

$1.2'$7

That work therefore argued that the observed high dispersion can be reproduced without dark matter and recommended that future observations focus on the present-day mass function, which in the cluster scenario should be strongly depleted in low-mass stars (Devlin et al., 30 Apr 2025).

An even more specific variant is the “dark star cluster” interpretation, in which the system is dominated not by a non-baryonic halo but by a centrally segregated black-hole subsystem. Direct $1.2'$8-body simulations identified a model with initial mass $1.2'$9, initial half-light radius $2.4'$0 pc, canonical IMF, and high black-hole retention that evolved to a UMa3/U1-like state with final $2.4'$1 pc, $2.4'$2, and age $2.4'$3 Gyr. In that scenario the cluster entered the dark-star-cluster phase around 4 Gyr ago, the luminous stars are expected to be depleted within the next 1 Gyr, and the central black-hole subsystem should disrupt over the subsequent Gyr (Rostami-Shirazi et al., 14 Aug 2025).

These cluster-based models do not establish that UMa3/U1 is definitely non-galactic, but they remove the uniqueness of the dark-matter inference. Thousand-level dynamical $2.4'$4 values are no longer diagnostic by themselves if binaries and remnant-dominated cores are allowed.

6. Revised spectroscopy, chemistry, rotation, and current status

A major reappraisal came from new spectroscopy covering 16 member stars. Higher-precision Keck/DEIMOS data, together with targeted follow-up of a suspected outlier, identified one confirmed spectroscopic binary and three additional binary candidates, and yielded a 95% confidence upper limit

$2.4'$5

The corresponding likelihood ratio favored a stellar-only dispersion of $2.4'$6 over the original $2.4'$7 value by $2.4'$8. The same study obtained the first metallicities for 12 member stars and found

$2.4'$9

with

$3 \pm 1$00

at the 95% credible level. Its conclusion was explicit: there is no observational evidence for dark matter in the system, and the chemodynamical properties are more consistent with a star cluster, although the dwarf-galaxy scenario is not fully ruled out (Cerny et al., 2 Oct 2025).

A subsequent Bayesian analysis examined whether unresolved rotation could explain part of the earlier kinematic signal. Fitting both non-rotating and rotating models to the available member-star samples, it found the non-rotating model preferred by a factor of $3 \pm 1$01–12. For the total population, the paper quoted a lower-bound rotational mass-to-light ratio of

$3 \pm 1$02

dropping to

$3 \pm 1$03

when one suspect star was removed, while removing two made the mass-to-light ratio unresolved. The authors concluded that UMa III/U1 remains ambiguous but is unlikely to be supported by rotational pressure (Adams et al., 20 Feb 2026).

The current state of the literature is therefore heterogeneous but substantially clarified. Discovery-era and early Jeans-based analyses showed that, under an 11-star high-dispersion interpretation, Ursa Major III would be an extraordinarily dark-matter-dominated “microgalaxy” and one of the strongest indirect-detection targets known (Smith et al., 2023, Zhao et al., 2024). Later collisional models demonstrated that compact remnants and binaries can reproduce both survival and elevated apparent dispersion in a dark-matter-free cluster (Devlin et al., 30 Apr 2025, Rostami-Shirazi et al., 14 Aug 2025). The most recent spectroscopy found no observational evidence for dark matter nor a large metallicity spread, shifting the balance of evidence toward a star-cluster classification (Cerny et al., 2 Oct 2025).

This suggests that UMa3/U1 is best regarded, at present, as a benchmark object at the cluster–galaxy boundary rather than as a settled member of either class. The decisive tests proposed across the literature are consistent: multi-epoch spectroscopy to map binaries, deeper photometry to measure the present-day mass function, improved abundance work to constrain any internal metallicity spread, and structural studies sensitive to tidal debris or remnant-dominated cores (Devlin et al., 30 Apr 2025, Cerny et al., 2 Oct 2025).