Pisces VII Ultra-Faint Dwarf

• Pisces VII is an ultra-faint dwarf galaxy characterized by its low luminosity, diffuse morphology, and a resolved horizontal branch confirming its satellite status to M33.
• Observations using DESI Legacy, TNG, Gemini, and HST with techniques like PSF photometry and MCMC fitting have determined its distance, structural parameters, and star-formation history.
• Its nearly quenched star formation, lack of neutral gas, and reionization-era fossil signature highlight its significance for the satellite census of M33 and the evolution of low-mass galaxies.

Pisces VII, also referred to as Triangulum III (Tri III), is an ultra-faint dwarf galaxy discovered in public imaging from the DESI Legacy Imaging Surveys in the outskirts of Messier 33 (M33). Its observational interpretation evolved from a candidate system that could either be an isolated ultra-faint or the second known satellite of M33 into a likely M33 satellite once the horizontal branch was resolved with deeper imaging. Subsequent analyses have constrained its distance, structural parameters, luminosity, neutral-gas content, and star-formation history, placing it among the faintest known galaxies in the Local Group and making it relevant to the satellite census of M33, the low-mass end of galaxy formation, and the effects of cosmic reionization on ultra-faint dwarfs (Martinez-Delgado et al., 2021, Collins et al., 2023, Jones et al., 1 Aug 2025).

1. Discovery and observational basis

Pisces VII was first noted by visual inspection of DESI Legacy Imaging Survey images around M33, outside the PAndAS footprint, using the Legacy Survey Sky Viewer. The detection was confirmed on SDSS DR9 and then followed up with deep imaging on the 3.58 m Telescopio Nazionale Galileo at La Palma using the DOLoRes focal-reducer spectro-photometer, which has a pixel scale of $0.252^{\prime\prime}\,\mathrm{px}^{-1}$. The survey material consisted of public images from the DESI Legacy Imaging Surveys, including DECaLS using DECam on the Blanco 4 m and complementary BASS/MzLS imaging, reaching surface-brightness limits $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$ (Martinez-Delgado et al., 2021).

The TNG follow-up obtained $41\times180\,\mathrm{s}$ in $g^\prime$ with median seeing $\simeq1.15^{\prime\prime}$ and $20\times180\,\mathrm{s}$ in $r^\prime$ with median seeing $\simeq0.85^{\prime\prime}$. Data reduction used standard IRAF preprocessing for bias subtraction and flat-fielding, followed by PSF photometry with DAOPHOT/ALLFRAME, including source detection, aperture and PSF fitting, image registration with DAOMASTER, stacked-image creation, and refined ALLFRAME runs. Morphology cuts based on sharpness and visual inspection were used to remove spurious detections. Photometric calibration used approximately 121 stars in common with Pan-STARRS1, with rms $\lesssim0.03$ mag, and extinction was corrected using Schlafly & Finkbeiner (2011) (Martinez-Delgado et al., 2021).

The deeper distance and structural re-analysis was based on Gemini/GMOS-N imaging totaling $4\times450\,\mathrm{s}$ in $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$0 and $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$1 in $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$2, which resolved the horizontal branch. Later Hubble Space Telescope imaging and Jansky Very Large Array observations extended the characterization from detection and morphology to star-formation history and neutral-gas limits (Collins et al., 2023, Jones et al., 1 Aug 2025).

2. Position, morphology, and structural characterization

In the discovery data, Pisces VII appeared as a round, diffuse over-density of unresolved light with central surface brightness $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$3, overlapping a small $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$4--$\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$5 clump of faint stars. Its J2000 center was measured as $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$6 and $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$7, with a projected separation from M33 of approximately $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$8. MCMC fits to the spatial distribution of RGB candidates gave an angular half-light radius $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$9, ellipticity $41\times180\,\mathrm{s}$0, position angle $41\times180\,\mathrm{s}$1, and number of member RGB stars $41\times180\,\mathrm{s}$2 (Martinez-Delgado et al., 2021).

The Gemini analysis remeasured the structure using an exponential profile,

$41\times180\,\mathrm{s}$3

with a uniform background $41\times180\,\mathrm{s}$4 added so that $41\times180\,\mathrm{s}$5, sampled with {\sc emcee}. This yielded $41\times180\,\mathrm{s}$6 arcmin, corresponding to $41\times180\,\mathrm{s}$7, ellipticity $41\times180\,\mathrm{s}$8 as an 84% upper limit, position angle $41\times180\,\mathrm{s}$9 degrees, and central coordinates $g^\prime$0, $g^\prime$1 (J2000) (Collins et al., 2023).

Taken together, the morphology is that of a compact but extremely low-surface-brightness stellar system in the ultra-faint dwarf regime rather than a globular cluster. The round appearance and low ellipticity are consistent across the discovery and Gemini-based remeasurement, while the deeper data revised the physical scale upward without changing its basic classification (Martinez-Delgado et al., 2021, Collins et al., 2023).

3. Distance determinations and satellite status

The initial distance estimate used the tip of the red giant branch (TRGB). The luminosity function was constructed from stars within $g^\prime$2, with a colour cut $g^\prime$3 and background subtraction from an annulus at $g^\prime$4--$g^\prime$5. Sobel edge detection identified a peak at $g^\prime$6 mag, roughly tracing the three brightest RGB candidates. A Bayesian broken-power-law fit following Tollerud et al. modeled the luminosity function as

$g^\prime$7

convolved with photometric errors and sampled with {\sc emcee}, giving $g^\prime$8 mag. With an adopted $g^\prime$9 mag, this implied $\simeq1.15^{\prime\prime}$0 mag and $\simeq1.15^{\prime\prime}$1. Because the TRGB was sparsely populated, this distance was only marginally consistent with both the M33 halo at $\simeq1.15^{\prime\prime}$2 and a truly isolated field ultra-faint at $\simeq1.15^{\prime\prime}$3, so membership to M33 could not be confirmed (Martinez-Delgado et al., 2021).

That ambiguity was reduced by the Gemini detection of a clear horizontal branch at $\simeq1.15^{\prime\prime}$4 mag. In that analysis, SDSS $\simeq1.15^{\prime\prime}$5 magnitudes were converted to Johnson $\simeq1.15^{\prime\prime}$6 using the colour relations of Jordi et al. (2006), and an MCMC fit to the $\simeq1.15^{\prime\prime}$7-band luminosity function around the HB used

$\simeq1.15^{\prime\prime}$8

Marginalization yielded $\simeq1.15^{\prime\prime}$9 mag. Adopting

$20\times180\,\mathrm{s}$0

with $20\times180\,\mathrm{s}$1 gave $20\times180\,\mathrm{s}$2 mag and therefore $20\times180\,\mathrm{s}$3 mag, corresponding to $20\times180\,\mathrm{s}$4. This placed Pisces VII at a three-dimensional separation of only $20\times180\,\mathrm{s}$5 from M33, compared with $20\times180\,\mathrm{s}$6 from M31, and thus made it almost certainly bound to M33 (Collins et al., 2023).

HST imaging later provided an independent HB-based determination by fitting the observed F606W magnitudes of HB stars to the fiducial HB locus of the metal-poor globular cluster M92, adopting $20\times180\,\mathrm{s}$7. The result, $20\times180\,\mathrm{s}$8 mag, corresponds to $20\times180\,\mathrm{s}$9. This is consistent with the Gemini value and places Pisces VII within the range $r^\prime$0--$r^\prime$1 derived for the three studied ultra-faint dwarfs. The convergence of the Gemini and HST distance scales strongly favors the interpretation of Pisces VII as M33’s second dwarf satellite galaxy rather than an isolated field system (Jones et al., 1 Aug 2025).

4. Luminosity, physical scale, and placement among ultra-faint dwarfs

The discovery analysis estimated the luminosity through Monte Carlo sampling of a 12 Gyr, $r^\prime$2 isochrone. Under the isolated-distance assumption, the galaxy would have $r^\prime$3 and $r^\prime$4; if placed at the distance of M33, it would have $r^\prime$5 and $r^\prime$6. Its physical half-light radius would be $r^\prime$7 at $r^\prime$8, or approximately $r^\prime$9 if at M33’s distance, with $\simeq0.85^{\prime\prime}$0 (Martinez-Delgado et al., 2021).

The deeper Gemini data remeasured these quantities using Monte Carlo sampling of a 12.5 Gyr, $\simeq0.85^{\prime\prime}$1 PARSEC luminosity function. In each trial, stars were drawn until $\simeq0.85^{\prime\prime}$2 fell within the CMD-selection box, their $\simeq0.85^{\prime\prime}$3 fluxes were summed, and the result was converted to $\simeq0.85^{\prime\prime}$4 over the distance and $\simeq0.85^{\prime\prime}$5 posterior. This yielded $\simeq0.85^{\prime\prime}$6, $\simeq0.85^{\prime\prime}$7, and $\simeq0.85^{\prime\prime}$8, corresponding to

$\simeq0.85^{\prime\prime}$9

and therefore $\lesssim0.03$0 (Collins et al., 2023).

In the size-luminosity plane, Pisces VII lies firmly among known ultra-faint dwarfs and is well separated from globular clusters. The discovery paper likewise placed it squarely among the Milky Way’s ultra-faint dwarfs, with $\lesssim0.03$1--$\lesssim0.03$2 and $\lesssim0.03$3, and noted similarity to And XXII at $\lesssim0.03$4. If the original isolated interpretation had held, it would have been the faintest known field dwarf detected to date; the later distance measurements instead anchor it in the M33 halo (Martinez-Delgado et al., 2021, Collins et al., 2023).

5. Stellar populations, neutral gas, and quenching history

The Gemini colour-magnitude diagram showed a metal-poor RGB with $\lesssim0.03$5 and $\lesssim0.03$6, together with a clearly resolved HB at $\lesssim0.03$7 mag. It also showed a blue plume of stars just fainter than the HB and a small population of putative core He-burning stars above the HB. BaSTI isochrones indicated that the blue plume is consistent with a $\lesssim0.03$8 population, while the bulk RGB and HB agree with an old $\lesssim0.03$9 population. The interpretation was explicitly conditional: if these stars are truly members of the galaxy, they imply star formation approximately $4\times450\,\mathrm{s}$0 after reionization and would require either retained or accreted cold gas, or a revision of how reionization feedback is understood in the faintest galaxies (Collins et al., 2023).

The later HST-based star-formation-history analysis used a new Python-based fitting code based on StarFISH. In that framework, the present-day Hess diagram was modeled as a linear combination of single-stellar-population Hess templates spanning age and metallicity, with SSPs generated from PARSEC, MIST, and BaSTI over $4\times450\,\mathrm{s}$1 to $4\times450\,\mathrm{s}$2 and $4\times450\,\mathrm{s}$3. Photometric completeness and uncertainty from $4\times450\,\mathrm{s}$4 artificial stars were folded into the templates, a Poisson maximum-likelihood fit solved for the weights $4\times450\,\mathrm{s}$5 of each age-metallicity bin under a weakly enforced metallicity evolution, and the recovered SFH was rebinned into broader intervals of $4\times450\,\mathrm{s}$6 dex. For Pisces VII, the PARSEC solution gave a star-formation rate of $4\times450\,\mathrm{s}$7 in the oldest $4\times450\,\mathrm{s}$8--$4\times450\,\mathrm{s}$9 bin and $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$00 in all younger bins. After applying a recycling fraction of 43%, the present-day stellar mass is $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$01 (Jones et al., 1 Aug 2025).

In the same analysis, cosmic reionization was taken to end at $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$02, corresponding to a lookback time of $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$03 for a $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$04CDM cosmology with $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$05 and $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$06. Pisces VII formed virtually all of its stars, $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$07, in the oldest bin and shows no significant star formation at later times; within the time resolution, its SFH is therefore consistent with complete quenching by, and indeed shortly after, reionization. This result differs in emphasis from the earlier blue-plume interpretation and suggests that any intermediate-age component remains contingent on confirming whether the candidate $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$08 stars are truly members (Collins et al., 2023, Jones et al., 1 Aug 2025).

Neutral-gas observations with the VLA detected no H I line. Assuming a top-hat profile of width $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$09 and a $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$10 threshold with $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$11 per $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$12 channel gives $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$13. Using

$\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$14

with $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$15 yields $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$16, or $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$17. More generally, the VLA observations place stringent upper limits on the H I content of all three studied systems, with all having $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$18. For Pisces VII, the extremely low H I fraction, $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$19, supports its categorization as a fossil of the reionization era (Jones et al., 1 Aug 2025).

6. Significance for the M33 satellite census and Local Group evolution

Pisces VII is significant because M33 has long lacked a substantial observed satellite population. Before its identification, only one potential satellite, Andromeda XXII/Tri I, had been detected around M33. If Pisces VII is included, the known count doubles, but the system still falls far short of the $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$20CDM expectation: simulations cited in the discovery work predict approximately $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$21--$\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$22 satellites with $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$23 around an M33-mass halo, including at least one with $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$24. This discrepancy was framed in terms of either a true dearth of low-mass subhalos around M33, possibly stripped by interactions with M31, or a population of satellites with surface brightness $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$25--$\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$26 that had remained below prior survey limits (Martinez-Delgado et al., 2021).

The later geometric placement of Pisces VII inside the M33 halo sharpened the dynamical implications. Its three-dimensional separation from M33 is $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$27, well inside the expected $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$28, while And XXII lies at $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$29 from M33 but remains kinematically ambiguous. The existence of two distant satellites at $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$30 was argued to favor a first-infall scenario for M33 into the Andromeda system: if M33 had undergone a close passage of less than $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$31 with M31 approximately $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$32 ago, most of its outer halo and satellites would have been heavily stripped. In that sense, Pisces VII is relevant not only to the census of low-mass galaxies, but also to M33’s orbital history within the Local Group (Collins et al., 2023).

The follow-up agenda remains correspondingly broad. Earlier work suggested Hubble Space Telescope imaging to reach the horizontal branch or main-sequence turn-off for a precise distance, and HB imaging has since delivered such a refinement. Remaining priorities include $\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$33--$\gtrsim 29\,\mathrm{mag\,arcsec^{-2}}$34 m-class spectroscopy of the brightest RGB stars to measure systemic velocity and metallicity, deep wide-field imaging in the outskirts of M33 to uncover fainter and more diffuse satellites, and space-based proper motions to test whether the satellites of M33 have been perturbed by M31. This suggests that Pisces VII will continue to function as a benchmark object for connecting ultra-faint dwarf demographics, reionization-era quenching, and Local Group assembly (Martinez-Delgado et al., 2021, Collins et al., 2023, Jones et al., 1 Aug 2025).