A Catalogue and Analysis of Ultra-Diffuse Galaxy Spectroscopic Properties

In order to help facilitate the future study of ultra-diffuse galaxies (UDGs) we compile a catalogue of their spectroscopic properties. Using it, we investigate some of the biases inherent in the current UDG sample that have been targeted for spectroscopy. In comparison to a larger sample of UDGs studied via their spectral energy distributions (SED), current spectroscopic targets are intrinsically brighter, have higher stellar mass, are larger, more globular cluster-rich, older, and have a wider spread in their metallicities. In particular, many spectroscopically studied UDGs have a significant fraction of their stellar mass contained within their globular cluster (GC) system. We also search for correlations between parameters in the catalogue. Of note is a correlation between alpha element abundance and metallicity as may be expected for a ‘failed galaxy’ scenario. However, the expected correlations of metallicity with age are not found and it is unclear if this is evidence against a ‘failed galaxy’ scenario or simply due to the low number statistics and the presence of outliers. Finally, we attempt to segment our catalogue into different classes using a machine learning K-means method. We find that the clustering is very weak and that it is currently not warranted to split the catalogue into multiple, distinct sub-populations. Our catalogue is available online and we aim to maintain it beyond the publication of this work.

Crucially the different proposed formation mechanisms are expected to leave different imprints in the stellar populations and dark matter halo properties of the resulting UDG.To provide a pair of contrasting examples: 1) a UDG forming via episodic stellar feedback is expected to have an extended star formation history, a dwarflike metallicity and a normal, dwarf-like dark matter halo (and thus lower velocity dispersion and globular cluster counts); while 2) a UDG forming at high redshift and quenching quickly is expected to have an old stellar population reflective of a single burst of star formation at high redshift, low metallicities reflective of the lack of time for chemical enrichment in the stellar population and a more massive dark matter halo (and thus higher velocity dispersion and globular cluster counts).Galaxies with properties resembling dwarf galaxies have been dubbed 'puffy dwarfs' in the literature due to their resemblance to the large-end tail of the dwarf half-light radius -luminosity relation (e.g., the UDGs forming via strong stellar feedback discussed above).Galaxies that have properties resembling a formation at high redshift and catastrophic quenching have been dubbed 'failed galaxies' in the literature (van Dokkum et al. 2015;Danieli et al. 2022;Forbes & Gannon 2024).Differentiating between these properties, and thus the corresponding formation scenario, may be accomplished through spectroscopy of the UDG's stellar body (e.g., Ferré-Mateu et al. 2023).
Alternatively, some of the properties (e.g., age/metallicity/star formation timescales) desired for elucidating UDG formation scenarios may be measured using spectral energy distribution (SED) fitting (e.g., Barbosa et al. 2020;Buzzo et al. 2022).This has the advantage of allowing larger samples of UDGs to be studied.Recent results from the SED fitting of UDGs have been able to separate them into two distinct classes using a K-means clustering analysis (Buzzo et al. 2024).Interestingly the mean properties of these classes were found to agree with the 'puffy dwarf'/'failed galaxy' examples given above.To date, no similar analysis has been performed on spectroscopic UDG samples.
Spectroscopy is extremely time intensive, requiring multiple hours on the world's largest optical telescopes (≥ 8m-class).As such, spectroscopic studies of UDG velocity dispersions and stellar populations tend to be limited to single objects and/or small samples (e.g., van Dokkum et al. 2017;Toloba et al. 2018;Gu et al. 2018;Alabi et al. 2018;Ferré-Mateu et al. 2018;Ruiz-Lara et al. 2018;Martín-Navarro et al. 2019;Emsellem et al. 2019;Danieli et al. 2019;van Dokkum et al. 2019b;Chilingarian et al. 2019;Müller et al. 2020;Gannon et al. 2020Gannon et al. , 2021Gannon et al. , 2022Gannon et al. , 2023;;Forbes et al. 2021).This has led to a UDG literature that requires significant effort to compile whenever a new object is studied and comparisons are wanted to previously published works.It has also led to a lack of understanding as to the selection biases of the current spectroscopic sample, which many UDG formation conclusions are based on.
In this work, we provide a compilation of current UDG spectroscopic properties in a single catalogue for easy access.In Section 2 we present the criteria for galaxies that have been included in our catalogue.In Section 3 we present the catalogue with individual galaxy notes.In Section 4 we provide a brief discussion of our sample in comparison to a large sample of UDGs Buzzo et al. (2024), investigate correlations in the sample and study its GC-richness.In Section 5 we provide some housekeeping details including referencing preferences and catalogue availability.We intend to keep the catalogue updated beyond the publication of this paper.Finally, a brief summary and conclusions are presented in Section 6.

INCLUSION CRITERION
In order to be included in this catalogue we require the galaxy to be both 1) a UDG and 2) have spectroscopically measured massweighted stellar ages and metallicities and/or a spectroscopically measured stellar/globular cluster (GC) velocity dispersion.
For the UDG definition, we wished to follow the original UDG definition ( ,0 > 24 mag arcsec −2 and  e > 1.5 kpc; van Dokkum et al. 2015) but derive it in the -band, to make it easier to search for UDGs in established catalogues such as those of McConnachie (2012) for the Local Group.We also convert from a central surface brightness ( ,0 ) to an average surface brightness within the half-light radius (⟨  ⟩ e ).We therefore take the original definition and apply the colour correction  =  − 0.3 along with an aperture correction of ⟨⟩ e =  0 +1.Our aperture correction is based on equations 7 and 9 in Graham & Driver (2005) for a galaxy of Sérsic index () slightly below 1, which is representative of a large population of UDGs in e.g., the Coma Cluster (Yagi et al. 2016).We, therefore, derive our UDG surface brightness criterion as: We make no changes to the half-light radius criterion from the original van Dokkum et al. (2015) definition, keeping a semi-major half-light radius  e > 1.5 kpc.
To be specific our final galaxy inclusion criteria for this catalogue are: (i) An average -band surface brightness within the half-light radius of ⟨  ⟩ e > 24.7 mag arcsec −2 .
(iii) Either a spectroscopically measured velocity dispersion and/or a mass-weighted stellar age and metallicity It is worth noting that different UDG definitions can bias the inferred different formation pathways (Van Nest et al. 2022) and the UDG definition itself may bias the sample to redder galaxies than one that searches for large-size outliers (Li et al. 2023).In addition, our choice of a mean surface brightness within the half-light radius may include a small percentage of higher-Sérsic index galaxies that a central surface brightness definition would exclude (see e.g., Greco et al. 2018a fig. 6).

CATALOGUE AND INDIVIDUAL GALAXY NOTES
We present the full catalogue in Appendix A, Table A1, Table A2 and Table A3 as well as online here1 .When the mean −band surface brightness within the half-light radius was unavailable it was calculated using the magnitude, half-light radius and equation 11 of Graham & Driver (2005).When magnitudes/surface brightnesses were only available in −band the magnitude has been transformed from -band using  =  − 0.3.Unless otherwise stated, when multiple measurements were available for the same property they were combined with weighting according to their uncertainties.Below we list individual notes for each UDG we have included in the catalogue.

Andromeda XIX
Andromeda XIX is a satellite of M31 and resides in the Local Group.Due to its extremely low surface brightness, it is unlikely similar analogues may be found outside of the Local Group.We note that Andromeda XIX is likely affected by tidal processes interacting with the nearby M31 (Collins et al. 2020(Collins et al. , 2022)).Any dynamical masses calculated with the data in the catalogue should be interpreted with caution.Due to the extremely diffuse nature of this object, the halflight radius, magnitude and surface brightness are highly uncertain.The listed stellar mass was calculated from the -band magnitude in Martin et al. (2016) assuming  ★ /  = 2.The data for this galaxy are taken from the works of Martin et al. (2016), Collins et al. (2020) and Gannon et al. (2021).From left to right, top to bottom these are: 1) Environment, where 1=cluster, 2=group and 3=field, 2) Distance to the UDG, 3) The -band absolute magnitude, 4) The average -band surface brightness within the half-light radius, 5) Total stellar mass, 6) 2D projected, semi-major half-light radius, 7) Axial ratio /, 8) Recessional velocity, 9) Stellar velocity dispersion, 10) GC system velocity dispersion, 11) Number of GCs, 12) Mass-weighted stellar age, 13) Mass-weighted stellar metallicity and 14) Stellar alpha abundance ([/Fe]).The catalogue data are plotted in blue.In orange we include results from the SED fitting of MATLAS Survey UDGs from the study of Buzzo et al. (2024).It is worth noting that for all of the SED sample, and the majority of the spectroscopic catalogue, the distance is assumed based on the environmental association.This assumption will affect several other panels that are dependent on the distance to derive physical units.In comparison to the larger SED sample, current spectroscopically studied UDGs tend to be intrinsically brighter, have higher stellar masses, are larger, more GC-rich, older and to have a wider spread in their metallicities.

Antlia II
Antlia II is a satellite of the Milky Way and resides in the Local Group.Due to its extremely low surface brightness, it is unlikely that similar analogues will be found outside of the local group.Dynamical modelling by Torrealba et al. (2019) suggests that a combination of tidal stripping and a cored dark matter profile can explain the properties of Antlia II.Due to the suggestion of tidal stripping, any dynamical mass calculated with the data should be treated with caution.The data for this galaxy are taken from the works of McConnachie (2012) and Torrealba et al. (2019).

DF44
DF44 is in the Coma cluster and has been one of the best-studied UDGs to date.It is one of only two UDGs that has had spatially resolved kinematic and stellar population gradients measured (the other being NGC 1052-DF2).This interest has mostly been the result of claims of a rich GC system associated with the galaxy van Dokkum et al. (2017) although there is currently some disagreement on the total GC numbers of DF44 in the literature (Saifollahi et al. 2021(Saifollahi et al. , 2022)).See Forbes & Gannon (2024) for a further discussion of these numbers.Following this work, we choose the van Dokkum et al. (2017) GC number.When quoting the  GC from van Dokkum et al. (2017) we use the number listed in their abstract (74±18) which is slightly different to that in Table 1.We have been advised this is the correct number (P. van Dokkum, private communication).While we classify DF44 as being in the Coma cluster, its phase space positioning suggests it may just be beginning to infall as part of a small group (van Dokkum et al. 2019b).As such, some authors have classified it with low-density UDGs when considering its formation (e.g., Ferré-Mateu et al. 2023).The radial velocity was derived using  r =  × ln (1 + ) from the redshift listed in footnote 6 of van Dokkum et al. (2017, 𝑧 =0.02132).The data for this galaxy are taken from the works of van Dokkum et al. (2016van Dokkum et al. ( , 2017van Dokkum et al. ( , 2019b)); Gannon et al. (2021); Villaume et al. (2022); Webb et al. (2022) and Saifollahi et al. (2022).

DF07
DF07 is in the Coma Cluster.The GC count is a combination of values by Lim et al. (2018, 39.1±23.8)and Saifollahi et al. (2022,  22 +5  −7 ).The data for this galaxy are taken from the works of van Dokkum et al. ( 2015

DF26
DF26 is a Coma cluster galaxy.This galaxy is also known as Y093 or Yagi 093.The magnitude was calculated from -band using  =  + 0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).Light-weighted ages and metallicities are available for this galaxy from Ruiz-Lara et al. (2018).The data for this galaxy are taken from the works of Yagi et al. (2016); Alabi et al. (2018); Lim et al. (2018) andFerré-Mateu et al. (2018).

DFX1
DFX1 is in the Coma Cluster.There is currently some disagreement on the total GC numbers of DF X1 in the literature (Saifollahi et al. 2021(Saifollahi et al. , 2022)).See further Forbes & Gannon (2024) for a discussion of these numbers.Following this work, we choose the van Dokkum et al. (2017) GC number.When quoting the  GC from van Dokkum et al. ( 2017) we use the number listed in their abstract which is slightly different from the number in Table 1.The radial velocity was derived using  R =  × ln (1 + ) from the redshift listed in section 2.1 of van Dokkum et al. (2017).Note that it is likely that the stellar velocity dispersion is also affected by the barycentric correction issue described in footnote 16 of van Dokkum et al. (2019b), however, the effect is likely small (P.van Dokkum, private communication).The data for this galaxy are taken from the works of van Dokkum et al.

DGSAT-I
DGSAT-I is listed as field although we note that it is located near the Pisces-Perseus supercluster and may potentially be a 'backsplash' galaxy (Martínez-Delgado et al. 2016;Papastergis et al. 2017;Benavides et al. 2021).The backsplash galaxy hypothesis has been disfavoured by Janssens et al. (2022) and thus we continue to list this galaxy as a field object.Note that some of the GCs counted are more luminous than expected given a traditional GC luminosity function (Janssens et al. 2022).The data for this galaxy are taken from the works of Martínez-Delgado et al. (2016); Martín-Navarro et al. (2019) and Janssens et al. (2022).

Hydra I UDG 11
Hydra I UDG 11 is in the Hydra I cluster.The magnitude was converted to  band using the listed  −  colour in Iodice et al. (2020) and then transformed to -band assuming  =  − 0.3.The data for this galaxy are taken from the works of Iodice et al. (2020) and Iodice et al. (2023).

J130026.26+272735.2
This UDG is in the Coma Cluster.The magnitude and surface brightness were calculated from -band using  =  + 0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).The data for this galaxy are taken from the work of Chilingarian et al. (2019).

NGC 1052-DF2
We classify NGC 1052-DF2 as being in the NGC 1052 group.However, there is the possibility that it is no longer bound to the NGC 1052 group as a result of its formation mechanism (e.g., Shen et al. 2021;van Dokkum et al. 2022).NGC 1052-DF2 is irregular for a galaxy in having both an extremely low measured velocity dispersion (van Dokkum et al. 2018a;Danieli et al. 2019) and an excess of bright GCs beyond what is expected given the established GC luminosity function for normal galaxies (van Dokkum et al. 2018b;Shen et al. 2021).The addition of a weak rotational component, as allowed by the data, may help alleviate the paucity of dark matter suggested by its velocity dispersion (Emsellem et al. 2019;Lewis et al. 2020;Montes et al. 2021).Furthermore, it may currently be undergoing a tidal interaction (Keim et al. 2022, although see Montes et al. 2021;Golini et al. 2024).We note that there existed some initial controversy over the distance to NGC 1052-DF2, whereby a smaller distance can solve much of the galaxy's irregular properties (see e.g., Trujillo et al. 2019;Monelli & Trujillo 2019).This controversy is now largely resolved by the deep HST imaging of Shen et al. (2021), with this distance being further updated in Appendix A of Shen et al. (2023).
We adopt the recessional velocity and velocity dispersion measurements reported from the Keck/KCWI data of Danieli et al. (2019) over those reported from the VLT/MUSE data of Emsellem et al. (2019) due to Keck/KCWI having the higher instrumental resolution.When quoting GC counts, we use the number of GCs measured by Shen et al. (2021) in the traditional GC luminosity function luminosity range, which excludes the brighter GC sub-population.We adopt the stellar population properties reported from VLT/MUSE data in

NGC 5846_UDG1
NGC 5846_UDG1 is in the NGC 5846 group.This galaxy is also known as MATLAS-2019 (Müller et al. 2020) and as NGC 5846-156 by Mahdavi et al. (2005).Here, we have adopted the velocity dispersion and redshift from Forbes et al. (2021) rather than those measured in Müller et al. (2020) due to the higher instrumental resolution in the data used by Forbes et al. (2021).We additionally adopt the distance/GC richness from Danieli et al. (2022) rather than that reported in Müller et al. (2021)

NGVSUDG-19
NGVSUDG-19 is in the Virgo cluster.The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

NGVSUDG-20
NGVSUDG-20 is in the Virgo cluster.The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

PUDG-R15
PUDG-R15 is in the Perseus cluster.The data for this galaxy are taken from the works of Gannon et al. (2022) and Ferré-Mateu et al. (2023).

PUDG-R16
PUDG-R16 is in the Perseus cluster.The data for this galaxy are taken from the work of Gannon et al. (2022).

PUDG-R84
PUDG-R84 is in the Perseus cluster.The data for this galaxy are taken from the works of Gannon et al. (2022) andFerré-Mateu et al. (2023).

PUDG-S74
PUDG-S74 is in the Perseus cluster.The data for this galaxy are taken from the works of Gannon et al. (2022) andFerré-Mateu et al. (2023).

Sagittarius dSph
The Sagittarius dSph is a satellite of the Milky Way in the Local Group and is known to be completely tidally disrupted around the Milky Way (Ibata et al. 2001).Any mass calculated with values listed in the catalogue should be treated with extreme caution due to the lack of equilibrium in the galaxy.The data for this galaxy are taken from the works of McConnachie (2012); Karachentsev et al. (2017) and Forbes et al. (2018).

UDG1137+16
UDG1137+16 is a satellite of the galaxy UGC 6594 in a group environment.It is also known as dw1137+16 by Müller et al. (2018).It has a disturbed morphology suggestive that it is undergoing stripping (Gannon et al. 2021).Any mass calculated with the values listed in the catalogue should be treated cautiously.  was transformed into -band using stated  −  colour (0.65) and  =  − 0.3.The data for this galaxy are taken from Gannon et al. (2021) and Ferré-Mateu et al. (2023).

VCC 1017
VCC 1017 is a Virgo cluster galaxy.The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

VCC 1052
VCC 1052 is a Virgo cluster galaxy.It has been noted to have a peculiar morphology with the possibility of spiral arms and/or tidal features (Lim et al. 2020).The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

VCC 1287
VCC 1287 is a Virgo cluster galaxy.Here the GC velocity dispersion is a combination of that measured by Beasley et al. (2016, 33

VCC 615
VCC 615 is a Virgo cluster galaxy.The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

VCC 811
VCC 811 is a Virgo cluster galaxy.The data for this galaxy are taken from the works of Lim et al. (2020) and Toloba et al. (2023).

VLSB-B
VLSB-B is a Virgo cluster galaxy.Note that many of the properties presented in the catalogue were updated in Toloba et al. (2023) from those listed in Toloba et al. (2018).The data for this galaxy are taken from the works of Toloba et al. (2018); Lim et al. (2020) and Toloba et al. (2023).

VLSB-D
VLSB-D is a Virgo cluster galaxy.It has an elongated structure and velocity gradient (Toloba et al. 2018) that suggests it is undergoing tidal stripping.Any dynamical mass derived with the properties listed must be treated with caution.Note that many of the properties presented in the catalogue were updated in Toloba et al. (2023) from those listed in Toloba et al. (2018).It is worth noting that while this galaxy has an estimated GC number of 13 ± 6.9, 14 GCs have been confirmed spectroscopically.The data for this galaxy are taken from the works of Toloba et al. (2018); Lim et al. (2020) and Toloba et al. (2023).

WLM
WLM is a galaxy on the outskirts of the Local Group.It is gas-rich and undergoing active star formation (Leaman et al. 2009).It also likely has a large rotation component in its dynamics (Leaman et al. 2009).The data for this galaxy are taken from McConnachie (2012) and Forbes et al. (2018).

Yagi 098
Yagi 098 is a Coma cluster galaxy.The magnitude was calculated from -band using  =  +0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).The data for this galaxy are taken from the works of Yagi et al. ( 2016

Yagi 275
Yagi 275 is a Coma cluster galaxy.The magnitude was calculated from -band using  =  + 0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).The data for this galaxy are taken from the works of Yagi et al. ( 2016

Yagi 276
Yagi 276 is a Coma cluster galaxy.The magnitude was calculated from -band using  =  +0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).The data for this galaxy are taken from the works of Yagi et al. (2016); Alabi et al. (2018) and Ferré-Mateu et al. (2018).

Yagi 358
Yagi 358 is a Coma cluster galaxy.The stellar mass was calculated from the absolute magnitude assuming  ★ /  = 2.The data for this galaxy are taken from the works of van Dokkum et al. ( 2017), Lim et al. (2018) and Gannon et al. (2023).

Yagi 418
Yagi 418 is a Coma cluster galaxy.The   was calculated from -band using  =  + 0.5 (based on Virgo dEs and Coma LSBs; van Zee et al. 2004;Alabi et al. 2020).Stellar population properties for this galaxy are presented in Ruiz-Lara et al. ( 2018) but here we prefer the Ferré-Mateu et al. (2023) age/metallicity values due to their being mass-weighted in contrast to the Ruiz-Lara et al. ( 2018) light-weighted values.We note that the ages are in good agreement between the two studies, as is expected for such intermediate-to-old stellar populations.The data for this galaxy are taken from the works of Yagi et al. (2016); Alabi et al. (2018) andFerré-Mateu et al. (2018).

Notable galaxies excluded from this catalogue
Here we discuss several notable galaxies and studies that we exclude from this catalogue: • While we include 2 galaxies from the study of Chilingarian et al. (2019) that meet our UDG definition the remaining 6 are too bright and/or small to meet our UDG criteria.As such, they are excluded from this sample.
• We exclude the galaxy PUDG-R24 from the study of Gannon et al. (2022) as it is too bright in surface brightness (⟨  ⟩ e ≈ 24.35 mag arcsec −2 ) to meet our definition.In Gannon et al. (2022) the galaxy was considered a UDG as it was expected to fade into the UDG regime in the next few Gyr.
• We exclude the galaxies OSG1 and OSG2 from Ruiz-Lara et al. ( 2018) due to their being light-weighted stellar population properties, rather than the mass-weighted properties presented herein.
• We exclude the stacked UDG stellar population properties from Rong et al. (2020) as it is both 1) not the results for a single galaxy and 2) includes in the stack many objects that are too bright to meet our UDG definition.It is worth noting that many of these objects do have similar stellar surface densities to the UDGs in our catalogue, it is their predominantly younger stellar populations that result in their being too bright for the surface brightness criterion (Rong et al. 2020).
• We exclude the two galaxies presented in Greco et al. (2018b) as: 1) the metallicities are lower limits and have not been measured and 2) the ages are not mean stellar ages but instead the age since the onset of star formation.We additionally note that the galaxy LSBG-285 presented by Greco et al. (2018b) is too small to meet our UDG definition.
• We exclude the UDGs presented in Trujillo et al. (2017) and Bellazzini et al. (2017) as only gas-phase metallicities and not stellar metallicities, are reported.We additionally note that both Bellazzini et al. (2017) galaxies are too bright to meet our UDG definition.
• We exclude the galaxy NGC 1052-DF4 (van Dokkum et al. 2019a) from our catalogue as it does not meet the surface brightness cut of our UDG definition.To be specific, using the surface brightness at the effective radius and Sérsic index for NGC 1052-DF4 reported in Cohen et al. (2018, 25.1 mag arcsec −2 and 0.79 respectively) and equation 9 of Graham & Driver (2005) we calculate an average surface brightness within the half-light radius of ⟨  ⟩ e ≈ 24.5 mag arcsec −2 which does not meet our definition.
It is also worth noting that many UDGs have measurements such as redshift and rotation available from their associated HI disk (e.g., Leisman et al. 2017;Spekkens & Karunakaran 2018;Mancera Piña et al. 2019, 2020, 2022;Karunakaran et al. 2020;Gault et al. 2021;Kong et al. 2022;O'Beirne et al. 2024).Our chosen criteria for this catalogue do not include these galaxies as we wish to focus on the galaxies' stellar population properties, and not that of their HI.We do note that much may be learned by comparing the two properties (e.g., Kado-Fong et al. 2022b,a) but that is beyond the scope of this work.

Catalogue Properties
In Figure 1 we present histograms of catalogue parameters.Where available, we include results from the SED fitting of field and group UDGs in the MATLAS survey from Buzzo et al. (2024).We picked this catalogue for comparison as it contains a greater number of UDGs (59) than our current work and as it has been used to argue for distinct formation pathways for UDGs through a K-means analysis.It is worth noting that the MATLAS survey primarily samples less The percentage of stellar mass in the GC system for UDGs in the catalogue.We calculate this property from the UDGs' GC counts using a mean GC-mass of 2×10 5 M ⊙ .Many of the spectroscopically studied UDGs have a significant percentage of their stellar mass contained within their GC system making it likely they are 'failed galaxy' UDGs.
dense field and group environments while the spectroscopic catalogue is heavily biased toward cluster environments.Moreover, the spectroscopic UDGs tend to be intrinsically brighter, have higher stellar masses, are larger, more GC-rich, older and have a wider spread in their metallicities.Spectroscopic UDGs being larger and brighter than those UDGs studied with SED fitting is likely a selection effect as it is a requirement of UDG spectroscopy for the target to be relatively bright to get meaningful results.Similar conclusions have also been drawn by Gannon et al. (2023).
Notably, non-UDGs that are more luminous and/or larger half-light radius galaxies tend to host richer GC systems (see e.g., Harris et al. 2017).On average the catalogue UDG sample presented here hosts more GCs than the SED sample of Buzzo et al as may be expected as they are also on average larger and brighter.Thus it is more likely that these UDGs have formed via the "failed galaxy" pathway that has been proposed by various authors (e.g., Peng & Lim 2016;Lim et al. 2018;Danieli et al. 2022;Forbes & Gannon 2024).
In Figure 2 we plot a histogram of the percentage of stellar mass in the GC system for UDGs in the catalogue.We calculate this percentage assuming a mean GC-mass of 2×10 5 M ⊙ from the stellar mass (M ★ ) and GC richness ( GC ) of the UDGs as: Note that for the UDGs NGC 1052-DF2 and DGSAT-I the approximation of a mean GC mass of 2×10 5 M ⊙ is likely too low given the overluminous star clusters known to be associated with these galaxies.The value included in the histogram will still provide a lower limit to the percentage of their stellar mass contained within their GC system.
There is an expectation that GCs will experience significant mass loss via tidal shocking, evaporation of stars bound to the GCs and the complete dissolution of the lowest mass GCs.It is commonly thought that GC systems may lose a significant fraction (> 75%) of their stellar mass after initial formation (Larsen et al. 2012;Reina-Campos et al. 2018).Accounting for these processes, many UDGs with  GC / ★ > 5% are consistent with having experienced little subsequent star formation post-GC formation (Danieli et al. 2022).
Due to the lack of star formation after the GC formation epoch, these may be interpreted as 'failed galaxy' UDGs, possibly consistent with being pure stellar halos (e.g., Peng & Lim 2016).

Catalogue Correlations
In Figure 3 we show the correlation matrix of the major properties included in the catalogue.We require each correlation to have 10 entries in the intersection of their parameters to calculate its coefficient.The vast majority of the properties are not correlated with coefficients between -0.5 and 0.5.Four correlations with |correlation coefficient| > 0.5 are found.We have checked and all these correlations remain if we exclude the two much fainter galaxies in the sample i.e., Andromeda XIX and Antlia II, for which analogues are likely not readily observable beyond the Local Group.The correlations found are: (i) Between   and ⟨  ⟩  .UDGs with higher luminosities also tend to exhibit higher fluxes.This is as expected.
(ii) Between the stellar mass ( ★ ) and   .Here the correlation coefficient is negative due to the nature of the magnitude system.UDGs that are more luminous also tend to exhibit higher stellar masses.This is as expected.
(iii) Between the stellar sigma ( ★ ) and the half-light radius ( e ).UDGs that have higher stellar sigma are dynamically hotter and tend to be larger.This is expected given the fundamental plane of elliptical galaxies and provides support for predicting UDG velocity dispersions via the fundamental plane (e.g., Zaritsky & Behroozi 2023;Zaritsky et al. 2023).
(iv) Between the alpha element abundance ([/Fe]) and the metallicity ([M/H]).UDGs that are more alpha-enhanced also tend to be lower in overall metallicity.A similar trend was found by Ferré-Mateu et al. (2023) from which much of our data is sourced.The leading line of reasoning to explain this trend is that observed UDGs cover a small stellar mass range.Thus, those that formed this stellar mass quickly in the early Universe will have elevated alpha abundances and low metallicities reflective of this early, fast formation.They will not experience significant subsequent star formation to change these metallicities as any significant subsequent star formation would cause them to not fulfil the UDG definition.
Under this line of reasoning, there is likely an expectation that there will also be a correlation between age and either alpha abundance/metallicity, which is not found in our catalogue.We show the [/Fe] -[M/H] correlation, along with the [M/H] -mean stellar age and [/Fe] -mean stellar age non-correlations in Figure 4.When looking at the centre panel, it is possible that a correlation is not found between age and metallicity due to the two galaxies at low age and metallicity.If these galaxies were removed the remaining galaxies would follow a standard age -metallicity relationship.Alternatively, the lack of trends may suggest the need for new formation pathways to be considered.e.g., the UDG DGSAT-I has both an elevated alpha abundance and signs of recent star formation (Martín-Navarro et al. 2019;Janssens et al. 2022) which does not fit our line of reasoning for a 'failed galaxy' UDG.

Catalogue UDG Populations
Finally, it was possible to split the Buzzo et al. (2024) UDG sample using the machine learning K-means method into two samples that resembled the expected properties for 'failed galaxy' UDGs and 'puffy dwarf' UDGs.We have attempted to perform a K-means analysis on the UDGs presented in this work to similarly split them into 'puffy dwarfs' and 'failed galaxies' but found that it was not applicable.We base this on measuring the silhouette score of the calculated K-means clusters as a function of the number of clusters found.The silhouette score is a measure of how similar an object is to its assigned cluster with values ranging from -1 to 1.In general, silhouette scores > 0.7 are required for a clustering to be considered 'strong'.When splitting into 2 clusters (i.e., the expectation of a 'puffy dwarf'/'failed galaxy' dichotomy) the clustering is at best very weak (i.e., silhouette score < 0.3).The addition of more K-means clusters does not solve this issue.We conclude that it is currently not warranted to segment the current spectroscopic data presented herein into separate, distinct UDG populations.We suggest this should be kept in mind when extrapolating the findings of current spectroscopic UDG studies more generally to the entire population.[M/H] for the catalogued galaxies.A correlation is found between these two parameters.Centre: [M/H] vs. mean stellar age for the catalogued galaxies.No correlation is found between these two parameters.Right: [/Fe] vs. mean stellar age for the catalogued galaxies.No correlation is found between these two parameters.The lack of an age -metallicity correlation is likely due to the presence of two outliers at low ages and metallicities that do not follow a standard age-metallicity relationship.

CATALOGUE ACCESS AND CITING
The catalogue described above has been made publicly available via the GitHub of the first author here.We include a QR code that will take the reader of this work to the catalogue in Figure 5.As part of the online catalogue a .bibLaTeX file is included which holds citations of all works that have contributed to this catalogue.It has been requested by community members via discussions at The Sunrise of Ultra-Diffuse Galaxies conference in Sesto, Italy, July 2023 that individual works contributing to this catalogue are cited when it is used.To facilitate this request a LaTeX input that should work with the provided .bibLaTeX file and the natbib package are included in the online catalogue.For reference, we include it below: We intend to continue to update the online version of the catalogue and reference list described herein as new UDG works are released.It is therefore advisable to include a date of retrieval when using these data.If we have missed data please contact the author for correspondence JSG (jonah.gannon@gmail.com)so that we may include it in this catalogue.

CONCLUSIONS
In this work, we have presented a literature compilation of UDG spectroscopic data along with the details to access it online.In comparison to the SED fitting of a larger UDG sample from the MATLAS survey we find the galaxies in our catalogue tend to be intrinsically brighter, have higher stellar mass, are larger, more GC-rich, older and have a wider spread in their metallicities.Spectroscopically studied UDGs also tend to be in denser, cluster environments while the SED sample is biased to groups and the field.These biases should be kept in mind when using UDG spectroscopic data to draw broad conclusions on the formation of the populations as a whole.
We show that many UDGs in this catalogue have a significant fraction of their stellar mass bound within their GC system.In current models for GC evolution, this may leave little room for star formation after the initial cluster formation epoch as much of their non-GC stellar mass can be explained as the product of GC dissolution/evaporation.
We investigate the correlations of major properties within the catalogue, finding the majority are uncorrelated.Of most interest is that alpha abundance and total metallicity are anti-correlated.UDGs that are more alpha-enhanced tend to have lower metallicity.This may be expected if some UDGs form fast and early when the Universe is less metal-enriched.Under this expectation, similar trends with age may be expected, but these are not found.We are currently unable to comment on whether this is related to the underlying formation pathways of UDGs or simply a result of outliers and low number statistics in the data.
Finally, we note that we are unable to reproduce the machine learning, K-means results of UDGs with SED fitting.The UDGs in our catalogue do not cluster strongly in K-space and do not cluster as distinctly as those studied in SED fitting.It is currently not warranted to separate the spectroscopically studied UDGs into multiple subpopulations.
Those wishing to use our catalogue may access it here or by scanning the QR code in Figure 5.We intend to keep this catalogue updated beyond the publication of this paper.A1.The first 8 columns of the full online catalogue.From left to right these are: 1) Primary Name, 2) Other names, 3) Environment where 1 = Cluster, 2 = Group and 3 = Field, 4) Distance noting that this is frequently assumed based on environmental association, 5) -band absolute magnitude, 6) the average -band surface brightness within the half-light radius, 7) Stellar mass, 8) Semi-major half-light radius and 9) Axial ratio, /.When values are not available they are listed as −999.The full table is available online here.

Figure 1 .
Figure1.Histograms of each of the UDG properties in the catalogue.From left to right, top to bottom these are: 1) Environment, where 1=cluster, 2=group and 3=field, 2) Distance to the UDG, 3) The -band absolute magnitude, 4) The average -band surface brightness within the half-light radius, 5) Total stellar mass, 6) 2D projected, semi-major half-light radius, 7) Axial ratio /, 8) Recessional velocity, 9) Stellar velocity dispersion, 10) GC system velocity dispersion, 11) Number of GCs, 12) Mass-weighted stellar age, 13) Mass-weighted stellar metallicity and 14) Stellar alpha abundance ([/Fe]).The catalogue data are plotted in blue.In orange we include results from the SED fitting of MATLAS Survey UDGs from the study ofBuzzo et al. (2024).It is worth noting that for all of the SED sample, and the majority of the spectroscopic catalogue, the distance is assumed based on the environmental association.This assumption will affect several other panels that are dependent on the distance to derive physical units.In comparison to the larger SED sample, current spectroscopically studied UDGs tend to be intrinsically brighter, have higher stellar masses, are larger, more GC-rich, older and to have a wider spread in their metallicities.
Fensch et al. (2019) over those reported from GTC/OSIRIS data in Ruiz-Lara et al. (2019) due to the larger field of view of VLT/MUSE being able to measure a more global value for the galaxy.Both values are in agreement.The data for this galaxy are taken from the works of van Dokkum et al. (2018a); Fensch et al. (2019); Danieli et al. (2019); Shen et al. (2021) and Shen et al. (2023).
due to the greater depth of the HST data.The data for this galaxy are taken from the works of Forbes et al. (2019); Müller et al. (2020, 2021); Forbes et al. (2021); Danieli et al. (2022) and Ferré-Mateu et al. (2023).

Figure 2 .
Figure2.The percentage of stellar mass in the GC system for UDGs in the catalogue.We calculate this property from the UDGs' GC counts using a mean GC-mass of 2×10 5 M ⊙ .Many of the spectroscopically studied UDGs have a significant percentage of their stellar mass contained within their GC system making it likely they are 'failed galaxy' UDGs.

Figure 3 .
Figure 3.A heatmap of the correlation matrix for the major properties in the catalogue.Correlations values are missing when they would rely on fewer than 10 datapoints for calculation.The majority of our properties are not correlated.For a full discussion of the interesting correlations found in the correlation matrix (i.e., those with |correlation coefficient| > 0.5), please refer to the text.

Figure 4 .
Figure 4. Left: [/Fe] vs.[M/H] for the catalogued galaxies.A correlation is found between these two parameters.Centre: [M/H] vs. mean stellar age for the catalogued galaxies.No correlation is found between these two parameters.Right: [/Fe] vs. mean stellar age for the catalogued galaxies.No correlation is found between these two parameters.The lack of an age -metallicity correlation is likely due to the presence of two outliers at low ages and metallicities that do not follow a standard age-metallicity relationship.

Figure 5 .
Figure 5.A QR code that you may scan to take you to the online catalogue.