A Photometric and Kinematic Analysis of UDG1137+16 (dw1137+16): Probing Ultra-Diffuse Galaxy Formation in a Group Environment
Jonah S. Gannon, Bililign T. Dullo, Duncan A. Forbes, R. Michael Rich, Javier Román, Warrick J. Couch, Jean P. Brodie, Anna Ferré-Mateu, Adebusola Alabi
MMNRAS , 1–15 (2020) Preprint 2 February 2021 Compiled using MNRAS L A TEX style file v3.0
A Photometric and Kinematic Analysis of UDG1137+16: ProbingUltra-Diffuse Galaxy Formation in a Group Environment
Jonah S. Gannon, ★ Bililign T. Dullo, Duncan A. Forbes, R. Michael Rich, Javier Román, Warrick J. Couch, Jean P. Brodie, , Anna Ferré-Mateu, , Adebusola Alabi, and Jeremy Mould Centre for Astrophysics and Supercomputing, Swinburne University, John Street, Hawthorn VIC 3122, Australia Departamento de Física de la Tierra y Astrofísica, Instituto de Física de Partículas y del Cosmos IPARCOS, Universidad Complutense de Madrid, E-28040 Madrid, Spain Department of Physics & Astronomy, University of California Los Angeles, 430 Portola Plaza, Los Angeles, CA 90095-1547, USA Instituto de Astrofísica de Andalucía (CSIC), Glorieta de la Astronomía, 18008 Granada, Spain University of California Observatories, 1156 High Street, Santa Cruz, CA 95064, USA Institut de Ciéncies del Cosmos (ICCUB), Universitat de Barcelona (IEEC-UB), Barcelona 08028, Spain
Accepted XXX. Received YYY; in original form ZZZ
ABSTRACT
The dominant physical formation mechanism(s) for ultra-diffuse galaxies (UDGs) is stillpoorly understood. Here, we combine new, deep imaging from the Jeanne Rich Telescopewith deep integral field spectroscopy from the Keck II telescope to investigate the formationof UDG1137+16. Our new analyses confirm both its environmental association with the lowdensity UGC 6594 group, along with its large size of 3.3 kpc and status as a UDG. The newimaging reveals two distinct stellar components for UDG1137+16, indicating that a centralstellar body is surrounded by an outer stellar envelope undergoing tidal interaction. Both thecomponents have approximately similar stellar masses. From our integral field spectroscopywe measure a stellar velocity dispersion within the half-light radius (15 ± − ) and findthat UDG1137+16 is similar to some other UDGs in that it is likely dark matter dominated.Incorporating literature measurements, we also examine the current state of UDG observationalkinematics. Placing these data on the central stellar velocity dispersion – stellar mass relation,we suggest there is little evidence for UDG1137+16 being created through a strong tidalinteraction. Finally, we investigate the constraining power current dynamical mass estimates(from stellar and globular cluster velocity dispersions) have on the total halo mass of UDGs.As most are measured within the half-light radius, they are unable to accurately constrain UDGtotal halo masses. Key words: galaxies:formation – galaxies:kinematics and dynamics – galaxies:photometry
With large half-light radii ( 𝑅 𝑒 > 𝜇 ,𝑔 >
24 mag arcsec − ) observations of ‘ultra-diffusegalaxies’ (UDGs; van Dokkum et al. 2015) have raised multiplescientific questions that are yet to be satisfactorily explained. Primeamong these, is the dark matter content, and by extension the to-tal dark matter halo mass, of UDGs. While strongly debated, evi-dence exists for UDGs exhibiting both a paucity (van Dokkum et al.2018; Danieli et al. 2019; Mancera Piña et al. 2019) and an over-abundance (Beasley et al. 2016; van Dokkum et al. 2017, 2019b;Martín-Navarro et al. 2019; Forbes et al. 2020b; Gannon et al. 2020)of dark matter. It is clear that despite both fitting the same categor- ★ E-mail: [email protected] ical definition, such objects are likely subject to different formationpathways.While some authors suggest UDGs may primarily form exter-nal to galaxy clusters (Amorisco & Loeb 2016; Román & Trujillo2017; Di Cintio et al. 2017; Chan et al. 2018; Alabi et al. 2018;Román et al. 2019) others suggest that the tidal stripping of stel-lar material in infalling cored dark matter halos can create UDGs(Carleton et al. 2019). In contrast, some suggest tidal effects aresubtler with tidal heating increasing the UDGs stellar extent whilekeeping the stellar material of the galaxy largely intact (Jiang et al.2019; Carleton et al. 2019; Sales et al. 2020). In such tidal formationscenarios the mass and density of the dark matter halo play a keyrole governing the strength of its role in UDG formation (Yozin &Bekki 2015; Carleton et al. 2019; Martin et al. 2019; Sales et al.2020). Further investigation of both the total dark matter halo mass © a r X i v : . [ a s t r o - ph . GA ] F e b J. S. Gannon et al. in UDGs and its profile (i.e., core vs cusp) is required to betterilluminate their formation pathways.In the tidally formed UDG simulations of Carleton et al. (2019)13% of simulated UDGs experienced a pericentric passage recentenough to be expected to display tidal features. Those UDGs knownto be associated with tidal features (e.g., VLSB-A Mihos et al. 2015,CenA-MM-DW3 Crnojević et al. 2016 and NGC1052-DF4 Monteset al. 2020) are in contrast to discoveries of large numbers of UDGsthat do not exhibit such features (e.g., the catalogues of Yagi et al.2016 or Alabi et al. 2020). As of yet, there is little evidence fora large population of UDGs displaying tidal features in the mostcomplete sample studied so far located in the Coma Cluster (Mowlaet al. 2017). It is difficult to reconcile how external tidal effectscan play a role in the creation of UDGs without creating a largepopulation of UDGs with tidal features. One possible explanationis that UDGs reside in massive dark matter halos whose larger massshields the stellar material from tidal stripping (Mowla et al. 2017).We note however, that the work of Muñoz et al. (2008) suggestsdeeper imaging studies may be required to detect such featuresleaving this an open line of inquiry.Sales et al. (2020) suggested that UDGs born via tidal strippingin a cluster environment should have significantly lower velocitydispersions at fixed stellar mass than those resulting from other for-mation pathways. Similar predictions, but with smaller effect, arealso made by Carleton et al. (2019). The predicted substantial di-versity of UDG velocity dispersions can also explain the substantialvariation in UDG mass to light (M/L) ratios (Carleton et al. 2019;Sales et al. 2020).If UDGs are to form external to clusters without their tidal fieldinfluencing formation, it is necessary to study their properties in lessdense group and field environments. The processing of UDGs in agalaxy group is thought to be critical in some simulations (e.g.,Martin et al. 2019). Here UDGs are expected to form most of theirstellar mass rapidly, early in the Universe coring their dark matterhalo (Martin et al. 2019). The cored halo profile makes UDGs moresusceptible to tidal heating which both suppresses further star for-mation and expands the stellar content of the galaxy (Carleton et al.2019; Martin et al. 2019). Other studies have shown similar resultsfor UDGs undergoing an expansion cycle due to stellar feedbackthat are then accreted into a dense environment, quenching star for-mation. This combination can ‘freeze’ their large size and slowlydecrease their surface brightness due to the passive evolution of thestellar population (Di Cintio et al. 2017; Chan et al. 2018; Tremmelet al. 2020). Indeed, many UDG formation scenarios rely on eitherthe pre-existence or creation of a dark matter core in the halo densityprofile (e.g., Carleton et al. 2019, albeit see Sales et al. 2020).Here we study UDG1137+16 seeking further observationalconstraints on the aforementioned UDG formation scenarios in agalaxy group environment. UDG1137+16 was first discovered us-ing image enhancing techniques to search for new dwarf galaxy can-didates from publicly available Sloan Digital Sky Survey (SDSS)data around the UGC 6594 group (Müller et al. 2018). Dubbed‘dw1137+16’, the authors reported that its large angular size andlow surface brightness are suggestive of a good UDG candidate if lo-cated in the UGC 6594 group. We note that, without a confirmationof this distance we cannot properly convert observed photometricproperties (e.g., half-light radius) into physical units. Additionally,Müller et al. (2018) suggest that: "Better photometry is needed toderive the structural parameters more accurately", motivating thenew observations presented here.In this work we obtain and analyse new, deep imaging forUDG1137+16 from the 0.7m Jeanne Rich Telescope (Section 2). We also acquire and analyse new, deep integral field spectroscopyfrom the Keck Cosmic Web Imager (KCWI) on the 10m KeckII telescope, deriving a recession velocity and a stellar velocitydispersion for UDG1137+16 (Section 3). In Section 4 we discussboth our imaging and spectroscopic results in the context of UDGformation. We then supplement our measurement with those fromthe literature to further probe the formation of UDGs. We place aparticular emphasis on the velocity dispersions of UDGs and theresulting dynamical masses inferred from them. Finally, we examinethe constraining power that current dynamical masses derived fromvelocity dispersions have on the total halo mass of UDG halos.Concluding remarks are presented in Section 5.
The photometric data used in this work were acquired using the0.7-m Jeanne Rich Telescope (JRT) and an FLI09000 CCD camera(pixel scale 1.114 (cid:48)(cid:48) /pix), at the Polaris Observatory Association sitenear Frazier Park, California. A full description of the instrumentalsetup is given in Rich et al. (2019). Observations were taken in theSDSS 𝑟 -band on the nights 2019 February 21 (16x300s), February27 (10x300s) and March 3 (22x300s) in dark, photometric skyconditions for 4 hours of total exposure time. The seeing of the finalimage is ∼ (cid:48)(cid:48) .Data reduction was carried out with the usual bias and dark sub-tractions. Images were flat-fielded using a master flat obtained viaan automatic pipeline in which all the science images observed dur-ing all the dark nights of the run were masked with NoiseChisel (Akhlaghi & Ichikawa 2015), normalised and subsequently com-bined. This produces a flat with high stability and efficiency atlow surface brightness levels. Images were astrometrised using the
SCAMP software (Bertin 2006) and photometry was referenced toSDSS to an accuracy of ∼ ± .
03 mag. The co-addition of frameswas performed using a pipeline with a resistant median-based com-bination algorithm, efficiently rejecting satellite tails, cosmic raysand other artefacts. The depth of the final coadded image is 28.4mag arcsec − (3 sigma in a 10x10 (cid:48)(cid:48) box) following the definitionprovided in appendix A of Román et al. (2020).In Figure 1 we display a central 5’ ×
5’ cutout of the resulting ∼ × The 𝑟 -band, major-axis surface brightness profile of UDG1137+16was extracted from the deep JRT image following the prescriptionin Dullo et al. (2017). To avoid light contamination from compactsources, background galaxies and bright foreground stars, we cre-ated an initial mask for the image with SExtractor (Bertin &Arnouts 1996) and then combined these with further manual masksas required. The IRAF task ellipse (Jedrzejewski 1987) was runon the final masked image to fit elliptical isophotes to the sky-subtracted 𝑟 -band, JRT image of the UDG. In extracting the lightprofiles, the isophote centre is held fixed, while the position angle(P.A.) and ellipticity were left as free parameters. Figure 2 shows thesurface brightness, P.A. and ellipticity profiles for the galaxy. Con-sistency checks were performed fitting the galaxy in both 2D using IMFIT and in 1D with non-fixed ellipse centres when extracting the
MNRAS , 1–15 (2020) n UDG1137+16 Figure 1.
A visual overview of both the Jeanne Rich Telescope imaging data and the KCWI integral field spectroscopy.
Left:
Whitelight depictions of theKCWI data cubes for: the object being shredded in the west of the frame (Shred, top), the UDG1137+16 (centre) and the positioning of sky exposures chosento include a small dwarf galaxy (bottom). Red arrows on the left correspond to the same corner of the data cube as indicated by the labelling in the image onthe right. Whitelight data cube image stretching is defined by the min/max pixel values of each individual cube.
Right : We display a 5’ ×
5’ box centred on theUGC 6594 group out of the full ∼ × surface brightness profile. Both methods were in good agreementwith the results presented herein (see further Venhola et al. 2017for a discussion of 1D vs 2D UDG fitting). The Sérsic (1968) 𝑅 / 𝑛 model is known to provide a good descrip-tion of the underlying stellar light distributions of UDGs and otherlow-luminosity ( 𝑀 𝑉 > ∼ -21.5 mag) galaxies (e.g., Graham & Driver2005; Dullo et al. 2019; Forbes et al. 2020c) . This model is writtenas: 𝐼 ( 𝑅 ) = 𝐼 e exp (cid:40) − 𝑏 𝑛 (cid:34)(cid:18) 𝑅𝑅 e (cid:19) / 𝑛 − (cid:35)(cid:41) , (1)where 𝐼 e is the intensity at the half-light radius ( 𝑅 e ). Thequantity 𝑏 𝑛 ≈ 𝑛 − / < ∼ 𝑛 < ∼
10 (Caon et al. 1993), is a We note that some irregular morphology UDGs are known to exist wherethis will not be true (Greco et al. 2018; Prole et al. 2019). function of the Sérsic index 𝑛 , and it ensures that 𝑅 e encloses halfof the total luminosity.Before fitting, we calculate the FWHMs of a Gaussian pointspread function (PSF) using several bright, unsaturated stars inthe JRT image of the UDG. We then fit a Sérsic model, that isconvolved with the PSF in 2D, to the galaxy’s 1D major-axis surfacebrightness profiles (e.g., Trujillo et al. 2001; Dullo et al. 2019;Dullo 2019; Forbes et al. 2020c) to obtain the best-fitting structuralparameters that describe the galaxy 1D light profile. We do thisby iteratively minimising the rms residual using the Levenberg-Marquardt optimisation algorithm. Figure 2 shows our fits to the 𝑟 -band light profile of UDG1137+16 along with the correspondingrms residual profiles.We fit our 1D light profile in two ways. Our first method is toperform a representative single Sérsic fitting for best comparison toliterature methods. We do this excluding the three outermost datapoints in our surface brightness profile. These points lie > (cid:48)(cid:48) withsurface brightness fainter than 27.4 mag arcsec − (compared tothe data limit of 28.4 mag arcsec − ). Although, our single Sérsic fit( 𝑛 ∼ . ± .
05 and 𝑅 e ∼ . ± . (cid:48)(cid:48) ) is in good agreement with theSDSS- 𝑟 band data modelled by Müller et al. (2018) ( 𝑛 ∼ . ± . 𝑅 e ∼ . (cid:48)(cid:48) ), we obtain a slightly larger 𝑅 𝑒 . We suggest this MNRAS , 1–15 (2020)
J. S. Gannon et al. has arisen due to the SDSS images used in Müller et al. (2018)’sanalysis having insufficient depth to properly constrain the galaxy’sstellar light distribution at large radii (
𝑅 > ∼ (cid:48)(cid:48) ).Our second method is to fit the full light profile using atwo-component, double-Sérsic model following Dullo & Graham(2013). They fitted a double Sérsic model to the light profile ofthe elliptical galaxy NGC 5576 which has a long tidal tail. Ourdouble-Sérsic decomposition of the 1D UDG1137+16 profile givesa central UDG stellar body along with an outer stellar component,Figure 2 right. While the double-Sérsic model gives a better descrip-tion for the surface brightness profile, and hence stellar componentsof UDG1137+16, this fit is not designed to produce global prop-erties of the galaxy (e.g. half-light radius) before a possible tidalinteraction with UGC6594. We discuss the physical interpretationof this second stellar component in Section 4.1.As a consistency check we also perform this fitting on publiclyavailable shallower data of UDG1137+16 from the DECaLS survey(Dey et al. 2019). There is strong agreement of the extracted 1Dsurface brightness profiles between the JRT and DECaLS imagingwithin 𝑅 ∼ . (cid:48)(cid:48) − (cid:48)(cid:48) . The higher resolution DECaLS imaging alsoreveals a possible unresolved, low mass nucleus for UDG1137+16( 𝐿 𝑟 ∼ × L (cid:12) ). Additionally, analysis of the DECaLS 𝑔 - and 𝑟 -band data for UDG1137+16 yields a global 𝑔 − 𝑟 colour of 0.65 ± < (cid:48)(cid:48) as that is where theDECaLS 𝑔 -band data provides the best constraints. The DECaLSdata have insufficient depth to properly constrain the outskirts ofUDG1137+16 and so for the remainder of this paper we focus onlyon the JRT imaging.Table 1 lists the best-fitting structural parameters for both therepresentative single and double Sérsic fits to our JRT surface bright-ness profile. 1 𝜎 uncertainties are determined after running a seriesof Monte Carlo simulations.We create more than 100 realisations of the UDG compos-ite light profile using the residual profile obtained after subtractingthe actual galaxy light profile from the fitted model (Figure 2. Weaccount for errors arising from inaccurate sky subtraction and forpossible light contamination from bright foreground objects, back-ground galaxies and faint, extended stellar halos while performingthis process. In order to determine the error associated with inaccu-rate sky subtraction and possible light contaminations, the medianbackground values were first measured from several 10 ×
10 pixelboxes away from both the UDG and other sources in the sky sub-tracted image. The error is determined by the standard deviationabout the mean of the median values in these boxes. These realisa-tions were then modelled akin to the modelling of the actual galaxylight profile to derive the 1 𝜎 errors from the best-fitting parameters. The total integrated luminosities ( 𝐿 𝑟 ) for the single- and double-Sérsic components from our decompositions of the UDG are com-puted using the best-fitting major-axis structural parameters, to-gether with the ellipticities of the individual fitted components. Weassume a distance of 21.1 Mpc for the galaxy (Tully et al. 2016)based on an assumption of association with the nearby UGC 6594group (see Section 3.2.2 for a confirmation of this assumption). Allmagnitudes are quoted in the AB system unless stated otherwise.Using our adopted distance we convert the apparent magni-tudes into absolute magnitudes ( 𝑀 𝑟 ) and then into luminosities (insolar units) assuming 𝑀 𝑟 = +4.65 for the Sun (Willmer 2018). Thesemagnitudes are not corrected for the small foreground Galactic ex-tinction ( < . Figure 2.
Major-axis surface brightness, position angle (P.A., measured indegrees from north to east) and ellipticity profiles of UDG1137+16 basedon deep imaging from the JRT (open circle).
Left:
A single Sérsic modelfitted to the 1D surface brightness profile within 50 (cid:48)(cid:48) . Right:
The best fittingdouble-Sérsic model to the full UDG light profile. We fit a central Sérsic(red dashed curve) and an outer Sérsic component (magenta dashed curve).The green, solid curve is the total fit to the profiles. Our data provides mildevidence for two stellar components in UDG1137+16. a stellar-mass to light ratio ( 𝑀 ∗ / 𝐿 𝑟 ) ≈ ± 𝑟 -band relation of Into & Portinari (2013) and our 𝑔 − 𝑟 colour of 0.65. This yields a total stellar mass for the UDG of1 . ± . × M (cid:12) . See Table 1 for a full summary of UDG1137+16luminosities and stellar masses. The integral field spectroscopy used in this work was obtained onthe night of 2020, February 17th. Our observations were made usingKCWI on the Keck II telescope (Program ID: W140, PI: Forbes). Weused the medium slicer with the BH3 grating to maximise spectralresolution ( 𝜎 inst ∼
13 km s − ), enhancing our ability to recoverlower velocity dispersions. We set a central wavelength of 5080Å in order to allow measurement of both the H 𝛽 and Mgb tripletabsorption features assuming the UDG is part of the UGC 6594group. Offset sky exposures were observed (as depicted in Figure1), to be used as inputs to a principal component analysis basedsky subtraction, optimised for faint galaxies in the UDG regime asdescribed in Gannon et al. (2020). An additional pointing of KCWIwith this configuration was taken on a nearby galaxy undergoingshredding (see Figure 1). Total exposure times were 16800s on theUDG, 8400s on sky and 1200s on the galaxy undergoing shredding(we dub this galaxy ‘Shred’ for the remainder of this work).Initial data reduction was performed using the standard KCWI We exclude the error in the stellar-mass to light ratio when calculatingerrors on the total stellar mass. MNRAS , 1–15 (2020) n UDG1137+16 Table 1.
Structural parameters of UDG1137+16Fit 𝜇 e 𝑅 e 𝑅 e 𝑛 𝜇 𝑚 𝑟 𝑀 𝑟 𝐿 𝑟 𝑀 ∗ (mag arcsec − ) (arcsec) (kpc) (mag arcsec − ) (AB mag) (AB mag) ( × 𝐿 (cid:12) ) ( × 𝑀 (cid:12) )Single Sérsic 26.5 (0.1) 32.3 (1.4) 3.3 (0.1) 1.05 (0.05) 24.62 (0.04) 16.7 (0.2) -15.0 (0.2) 0.7 (0.1) 1.4 (0.2)Inner Sérsic 25.8 (0.1) 15.3 (1.0) 1.6 (0.1) 0.55 (0.05) 24.94 (0.03) 17.8 (0.2) -13.8 (0.2) 0.2 (0.05) 0.4 (0.1)Outer Sérsic 27.5 (0.1) 62.3 (1.6) 6.4 (0.2) 0.40 (0.11) – 16.6 (0.3) -15.0 (0.3) 0.7 (0.2) 1.4 (0.4)Total I+O – – – – 24.94 (0.03) 16.5 (0.4) -15.2 (0.4) 0.9 (0.5) 1.8 (1.0)Notes.— Structural parameters from the 1D single and double Sérsic model fits to the 𝑟 -band surface brightness profiles of UDG+1137. Errors in parametersare given in (brackets) following them. I+O = Inner Sérsic + Outer Sérsic. Stellar masses were calculated assuming M ∗ /L ≈ 𝑟 -band. We assume a distanceof 21.1 Mpc (i.e., association with UGC 6594) when converting distance dependant parameters. Central surface brightnesses come from extrapolation of thefit to the centre. pipeline (Morrissey et al. 2018). Following this, we took the output ocubes which were non sky–subtracted and standard star calibrated,cropping them both spatially and spectrally to the regions of full cov-erage. We then performed an additional flat fielding step to correctfor low level gradient remaining in the data (see further - Gannonet al. 2020). Spectra of the UDG were extracted using the KCWIdata in light-bucket mode, collapsing the data cube into a singlespectrum. Sky spectra were extracted in a similar manner from theoffset sky exposures, avoiding the area of the data containing thedwarf galaxy. We took our extracted sky spectra and used themto perform a principal component analysis sky subtraction as de-scribed in Gannon et al. (2020). A model for the UDG emissionis required to perform this sky subtraction, we therefore selected12 models of varying spectral type (K-type giant, A, F, G), metal-licity (-1.3 ≤ [Fe/H] ≤ -0.1) and alpha enhancement ([ 𝛼 /Fe] = 0 or0.4) from the Coelho (2014) library of high resolution syntheticstellar populations to fit to our data. We took the best fitting tem-plate (K-type giant; [Fe/H] = −
1; [ 𝛼 /Fe] = 0) and used it as thegalaxy emission model to sky subtract our spectra as part of thesky subtraction routine. Sky subtraction for both the dwarf galaxyin the sky frame and ‘Shred’ were performed by the subtractionof on-chip sky. After applying the relevant barycentric corrections(Tollerud 2015), the sky–subtracted spectra were median stacked.For our UDG spectrum we estimate a final signal to noise ratio of20 per Å in the continuum. We display our extracted spectrum for the ‘Shred’ galaxy in Figure3. In order to measure a recession velocity from this spectrum weidentify the three clear emission lines as H 𝛽 and the [OIII] doublet,and fit Gaussian profiles to each to measure their centroid. Based onthe centroids of these Gaussian fits we measure a recession velocityof 1237 ± − and thus confirm the association of this objectwith the low density UGC 6594 group (UGC 6594, 𝑉 𝑠𝑦𝑠 = ± − ; van Driel et al. 2016).We also attempt to extract a spectrum for the dwarf galaxy inthe sky frame. Unfortunately, the spectrum we extract has a S/N toolow to perform analysis. The extraction of a velocity dispersion in the UDG regime is knownto be a particularly onerous task as UDG spectra often have a combi- nation of low S/N, exotic chemical abundances (e.g. Martín-Navarroet al. 2019; Ferré-Mateu et al. 2018), and velocity dispersions at orbelow the instrumental resolution. Based on our previous investi-gation of these issues in Gannon et al. (2020), we chose to fit ourspectrum with pPXF (Cappellari et al. 2013; Cappellari 2017) usinga wide ranging set of input parameters and fitting templates. For ourtemplates we chose to use the synthesised stellar library of Coelho(2014) along with a KCWI observation of the Milky Way GC M3.The KCWI observation of M3 was taken on 2019, April 2nd inthe same KCWI configuration as our science observation but withlower central wavelengths to account for the assumed differencein redshift between our template and the target. Being observed inthe same KCWI configuration it ideally models the instrumentalresolution, removing the mischaracterisation of instrumental reso-lution as a possible source of error in our fitting. The Coelho (2014)library has > pPXF in 241different input configurations. These configurations contain bothpure Gaussian fitting and those with the higher Gauss-Hermite mo-ments h and h also. Our pPXF configurations also include theaddition of a wide range of both additive and multiplicative polyno-mials (0-10th order for both respectively) to the templates to correctfor possible errors in recovery of the UDG spectrum. Moreover, wefit the spectrum in five distinct regions to test that our recovered ve-locity dispersion is not driven by any particular spectral region andis not adversely effected by the accidental inclusion of sky residualsin the fitting. We discard fits with reported errors in recession veloc-ity and/or velocity dispersion that are greater than 25 km s − as wedeem the fit to have ineffectively modelled our data. We display onefit, that we deem ‘good’, in Figure 3. Here we fit a pure Gaussianprofile with 4 additive and 4 multiplicative polynomials to the fullUDG spectrum using the Coelho (2014) library.We display the median kinematic values (i.e., recession veloc-ity and velocity dispersion with associated errors) resulting fromthe exhaustive fitting of our spectrum in Figure 4. We find goodagreement between fits done with both of our templates. We there-fore take the average of the M3 and Coelho (2014) library resultsas our final recession velocity = 1014 ± − and velocitydispersion = 15 ± − . As the fitting with the M3 templateis robust to a possible instrumental resolution mischaracterisationwe expect this to not be a major source of error in our result. Asfitting with the Coelho (2014) library minimises the possibility oftemplate mismatch, we also expect this to not be a major source of MNRAS , 1–15 (2020)
J. S. Gannon et al.
Figure 3.
KCWI Spectra.
Upper:
The spectrum extracted for the ‘Shred’ object. H 𝛽 and [OIII] emission lines are labelled. Based on Gaussian fits to theseemission lines we measure a recession velocity of 1237 ± − confirming its association with the low density UGC 6594 group. Lower:
The spectrumextracted for UDG1137+16 (black) and a representative pPXF fit (red). H 𝛽 , Fe I and Mgb absorption features have been labelled. Based on our fitting of theUDG1137+16 spectrum we measure a recession velocity of 1014 ± − and a velocity dispersion of 15 ± − . error in our result. Finally, fits to 5 different spectral regions of thespectrum report similar velocity dispersions which are consistentwithin errors, suggestive that a particular wavelength region doesnot drive our final reported values.In the double Sérsic decomposition described in Section 2.2.2,the half-light radius of the central stellar body in UDG1137+16 wasfound to be 15.4 (cid:48)(cid:48) . Due to the slightly off-centre pointing of the ∼ (cid:48)(cid:48) × (cid:48)(cid:48) medium slicer of KCWI, our velocity dispersion mea-surement represents a flux-weighted measurement within approx-imately one half-light radius of the central stellar body. Here thecentral regions are slightly over-represented in our measurement, asthey are fully sampled by the slicer, while the outer regions becomeincreasingly under-sampled in comparison. When using the singleSérsic half-light radius, our velocity dispersion only represents aflux-weighted measurement made within ∼ 𝑅 𝑒 ) and the luminosity–weighted average line-of-sight ve-locity dispersion within this radius ( 𝜎 𝑒 ) to determine the dynamicalmass within the 3D deprojected half-light radius ( 𝑟 / ). It takes theform: 𝑀 ( < 𝑟 / ) = ( 𝜎 𝑒 ( km s − ) )( 𝑅 𝑒 pc ) M (cid:12) ; where 𝑟 / ≈ 𝑅 𝑒 (2)We extrapolate our stellar velocity dispersion measurementout to the single Sérsic half-light radius assuming a flat velocityprofile for UDG1137+16. Using this velocity dispersion and ourcalculated half-light radius from the single Sérsic fitting we infer adynamical mass 𝑀 ( < . ) = . ± . × M (cid:12) . Combiningthis with our earlier luminosity measurement from JRT photometry MNRAS , 1–15 (2020) n UDG1137+16 Figure 4.
The recession velocity and velocity dispersion results of our pPXF fitting for our UDG1137+16 spectrum. Data are included for fitting with eachof two template sets: the Coelho (2014) synthesised stellar library (red) andan observed template from the Milky Way GC M3 (green). Fitting for fivespectral regions is included for each template set: the entire spectrum (All- circular points), the spectrum less than 5100 Å (<5100 Å- triangles), thespectrum excluding the initial 175 pixels to exclude the H 𝛽 sky subtractionresidual (No H 𝛽 - squares), the red half of the spectrum (Red - pentagons)and the blue half of the spectrum (Blue - stars). We take the average of thesefits to obtain our adopted recession velocity (1014 ± − ) and velocitydispersion ( 𝜎 = 15 ± − ) of UDG1137+16. . ± . × L (cid:12) ) we infer an 𝑟 -band M/L ratio of 17 . ± . Due to their extremely similar recession velocities we suggestUDG1137+16 is likely part of the low density UGC 6594 group(UGC 6594, 𝑉 𝑠𝑦𝑠 = ± − ; UDG1137+16, 𝑉 𝑠𝑦𝑠 = ± − ) placing it at a distance of 21.1 Mpc. Basedon the JRT data and this assumed distance we note UDG1137+16meets the standard definition for a UDG ( 𝜇 ,𝑔 >
24 mag arcsec − ; 𝑅 𝑒 > . In Section 2.2.2, the proper decomposition of the 1D surface bright-ness profile of UDG1137+16 required (at least) two distinct com-ponents. In order to investigate this further, in Figure 5 we dis-play a smoothed, contoured 5’ ×
5’ cutout of our JRT data aroundUDG1137+16. Visual inspection of the image reveals a clear distor-tion of the stellar envelope in the direction of UGC 6594 (labelledwith a red arrow), indicative of a tidal interaction between the galax-ies. Moreover, the extraction of our 1D surface brightness profilereveals a clear twisting of the elliptical isophotes to align with the di-rection of UGC 6594, a known effect of tidal interactions (Carletonet al. 2019).We argue the close projected distance between the galaxies ( ∼
20 kpc between galaxy centres at the adopted distance of 21.1 Mpc)and their extremely similar recession velocities provides further
Figure 5.
A smoothed, contoured version of the imaging data displayed inFigure 1. We smooth the 5’ ×
5’ JRT cutout using a 2D Gaussian kernel ofwidth 2 pixels and apply 10 contours to help highlight low surface bright-ness structure in the frame. The elongation in contours directly south ofUDG1137+16 is caused by a nearby bright star. A blue circle indicates theposition of the dwarf galaxy in the frame which appears faintly connectedto our UDG. A red arrow indicates an elongation in the UDG towards UGC6594 suggestive of possible tidal interaction. Qualitative evidence exists forUDG1137+16 undergoing a tidal interaction.Property Value SourceR.A. [J2000] 11 : 37 : 46 Müller et al. (2018)Dec. [J2000] +
15 : 31 : 09 Müller et al. (2018)Dist. [Mpc] 21.1 Adopted (Tully et al. 2016) 𝑚 𝑟 [mag] 16.7 ± 𝑀 𝑟 [mag] -15.0 ± 𝜇 [mag arcsec − ] 24.62 ± 𝑅 𝑒 [arcsec] 32.3 ± 𝑅 𝑒 [kpc] 3.3 ± 𝑀 ∗ [ × M (cid:12) ] 1.4 ± − ] 1014 ± 𝜎 [km s − ] 15 ± 𝑀 dyn [ × M (cid:12) ] 6.18 ± 𝑀 / 𝐿 ± Table 2.
A summary of the basic properties of UDG1137+16. For basicphotometric properties we use the single Sérsic fit in Section 2.2.2 whichallows the best comparison to the methods used in other works. For distancedependent quantities we use our adopted distance of 21.1 Mpc. evidence that they are likely an interacting pair. While we do notknow the precise physical separation between galaxies, if the averagetangential distance along line-of-sight is similar to the projecteddistance, 3D separation will be on average √ M (cid:12) wecalculate tidal radii for our measured dynamical mass using equation1 of Mowla et al. (2017). They are 1.2/2.4/3.5 kpc based on physicalseparations of 20/40/60 kpc with UGC6594. All these radii are lessthan that which our dynamical mass is measured, suggesting tidalfeatures should form. We therefore propose a scenario where theUDG outer stellar material is redistributed via tidal interaction intoa stellar envelope of approximately similar stellar mass to the centralstellar body.Our deep image does not reveal a tidal bridge between the MNRAS , 1–15 (2020)
J. S. Gannon et al. galaxies, as may be expected for interacting galaxies. This suggestseither that we have insufficient depth to detect such a feature or thatthis interaction is in its infancy and such structures are yet to form.It should be noted that UDGs embedded in tidal structures belowthe surface brightness limit of our data are known to exist (e.g.CenA-MM-Dw3; Crnojević et al. 2016). Additionally, while thereis no obvious evidence for widespread tidal disruption of UDGs inthe Coma cluster (Mowla et al. 2017), some UDGs are known toexhibit tidal features (e.g. Mihos et al. 2015).There may be cause for concern that tidal stripping could ad-versely affect our stellar velocity dispersion and hence dynamicalmass measurement (see e.g. Montes et al. 2020). However, tidalstripping operates outside–in, affecting the outskirts of the galaxyand leaving the central velocity dispersion approximately constant(Bender et al. 1992; Chilingarian et al. 2009; Blom et al. 2014;Penny et al. 2015). For example, an extreme Local Group UDGanalogue with suspected tidal origins, Andromeda XIX, exhibits alargely flat velocity dispersion profile (Collins et al. 2020). Addi-tionally, our stellar velocity dispersion measurement is made withinthe half-light radius of the central stellar component, which wouldbe subject to lessened tidal forces in comparison to the outskirts ofthe galaxy. We conclude that tidal stripping is not likely to stronglybias our centralised velocity dispersion measurement.The galaxy we label ‘Shred’ in Figure 1 is also located closein both projection and velocity space to UGC 6594. It shows thegalaxy is being disrupted and exhibits an irregular morphology. InFigure 5 this galaxy also hints at the ‘S-shaped’ features indicativeof a tidal disruption, however the coincidental association of othernearby objects makes it difficult to ascertain whether or not thesefeatures are real or merely incidental. Additionally, our spectrumfor this object displays strong emission features indicative of activestar formation which is known to be enhanced by tidal interactions(Martig & Bournaud 2008). We suggest it is likely that this galaxyis also undergoing tidal interactions in the group.
In Figure 6 we plot our stellar velocity dispersion for UDG1137+16,along with measurements for other UDGs (van Dokkum et al. 2017,2019b; Martín-Navarro et al. 2019; Danieli et al. 2019; Gannon et al.2020; Forbes et al. 2020a) and objects called UDGs where somedebate remains as to their precise classification/properties (Chilin-garian et al. 2019; Collins et al. 2020). Additional to this, we includeUDGs from Toloba et al. (2018) and van Dokkum et al. (2019a)with GC kinematics under the assumption their measured GC kine-matics are representative of the stellar velocity dispersion of theirassociated UDG (see further Forbes et al. 2020a). Our UDG sampleis broadly selected with no criteria related to galaxy environment.We study the stellar velocity dispersion – stellar mass relation using asample of non-UDGs (dwarf ellipticals; Chilingarian et al. 2009, Lo-cal Group dwarfs; McConnachie 2012, elliptical/S0/irregular/spiralgalaxies; Harris et al. 2013 and early type galaxies; Cappellari et al.2013). For more details on the UDG sample, including environ-mental associations, see Appendix A. From Figure 6 it is clear that,with the possible exception of NGC1052-DF2 and NGC1052-DF4,all UDGs appear to follow the same central velocity dispersion –stellar mass relation.Interestingly, there appears to be a lack of non-UDG objects at We exclude VLSB-B in plotting the Toloba et al. (2018) sample due to itslarge errors. a similar velocity dispersion but higher stellar mass than UDGs (Fig-ure 6). Tidal stripping has been a formation mechanism proposed forUDGs (Yozin & Bekki 2015; Carleton et al. 2019). As previouslydiscussed it is not necessarily expected to change the central veloc-ity dispersion in a galaxy but will remove stellar mass. In a strongstripping formation scenario (e.g. >
99% mass loss; Carleton et al.2019) we might expect to see progenitor galaxies for UDGs in thisregion of parameter space. The lack of UDG progenitor candidatesat a similar velocity dispersion and a stellar mass of ∼ M (cid:12) disfavours strong tidal stripping as a natural formation pathway forUDGs without a mechanism to decrease the central stellar velocitydispersion.Multiple pericentric passages my be an example of such amechanism, although it would be expected to more significantlydisrupt the morphology of the observed galaxy (Errani et al. 2015).We also note recent work demonstrating certain radial orbits canlead to both a dark matter and a stellar velocity dispersion reduction(Macciò et al. 2020) in UDGs. In this case we would not expectour galaxy to follow the stellar velocity dispersion – stellar massrelation, similar to NGC1052-DF2/NGC1052-DF4. UDG1137+16follows the relation however, and does not appear to have lost darkmatter in its central regions suggesting it is not in such a radial orbit.We suggest future work studying the resolved velocity profiles offield and group UDGs may be vital to properly understand the effectsof tidal interactions on their formation and evolution.In the case of UDG1137+16, the UDG formation simulationsof Sales et al. (2020) in denser environments would suggest thatit may not be formed through strong tidal interaction as it obeysthe stellar velocity dispersion – stellar mass relation. We also notethat, while the excess light from the stellar envelope at large radiiincreases the fitted half-light radius in the single component decom-position of the surface brightness profile, the central stellar body inthe multi-component fitting still meets the standard UDG definition.This suggests that UDG1137+16 was already of large size beforeit began its current tidal interaction. We therefore disfavour for-mation scenarios for UDG1137+16 that rely solely on strong tidalstripping/interactions to produce the large sizes observed in UDGs.In order to pose definitive constraints on the formation process ofUDG1137+16 and other UDGs, the need of high quality spectra toperform stellar population analysis becomes imperative and will besubject of future work.We take our calculated dynamical mass and M/L ratio andcompare them to the ‘U-shaped’ relation of non-UDGs in Figure 7.We plot a selection of non-UDGs to establish the relation (Zarit-sky et al. 2006; Wolf et al. 2010; Forbes et al. 2011; Cappellariet al. 2013; Toloba et al. 2014) and include those UDG objectspreviously plotted in Figure 6. As has been previously established(Toloba et al. 2018; van Dokkum et al. 2019b; Gannon et al. 2020),UDGs in Figure 7 do not necessarily follow the ‘U-shaped’ relation,with some lying above and others lying below their expected M/Lratio given their dynamical mass. We find this also to be true forUDG1137+16. It lies above the standard ‘U-shaped’ relation witha higher M/L ratio for its dynamical mass than other non-UDGs.Similar results have been found for other UDGs that are known to beassociated with large GC systems (e.g. VCC 1287). Previously, thecorrelation between GC system richness and total dark matter halomass (e.g. Burkert & Forbes 2020) has been used to infer two typesof UDG, those with dwarf galaxy-like and those with more massivedark matter halos (Forbes et al. 2020b). Unfortunately, our JRT datahave insufficient resolution to adequately probe the GC system ofthis UDG. We can therefore only suggest that UDG1137+16 dis-plays inner dark matter halo properties similar to other GC rich MNRAS , 1–15 (2020) n UDG1137+16 Figure 6.
Central stellar velocity dispersion vs galaxy stellar mass. We plot a sample of non-UDGs: early type galaxies (Cappellari et al. 2013), ellipti-cal/S0/irregular/spiral galaxies (Harris et al. 2013), dwarf ellipticals (Chilingarian et al. 2009) and Local Group dwarfs (McConnachie 2012) (all black points).We also plot our data (red star) along with other UDGs with central stellar velocity dispersions (blue points - van Dokkum et al. 2017, 2018; Danieli et al. 2019;van Dokkum et al. 2019b; Martín-Navarro et al. 2019; Gannon et al. 2020; Forbes et al. 2020a) and ‘UDGs’ from the literature where some debate remains asto their precise classification/properties (green triangles - Chilingarian et al. 2019; Collins et al. 2020). We also include measured GC velocity dispersions ofUDGs assuming they are comparable to the stellar velocity dispersion (yellow triangles - Toloba et al. 2018; van Dokkum et al. 2019a). For a full summary ofthe literature UDGs see Appendix A. Most UDGs, to date, have central stellar velocity dispersions consistent with non-UDGs for their stellar mass.
UDGs. We investigate further the total halo mass of UDG1137+16in Section 4.3.Given that only NGC1052-DF4 clearly lies below the trendin Figure 7, we focus on those UDGs lying above the ‘U-shaped’relation to higher M/L ratio given their dynamical mass. Noting thatall the UDGs plotted are of broadly similar luminosity and stellarmass, the observed deviations from the ‘U-shaped’ relation mustbe driven by their dynamical masses. Additionally, these dynamicalmasses are proportional to the square of the velocity dispersionand the half-light radius of the galaxy (e.g., Equation 2). We havealready demonstrated in Figure 6 that UDGs do not typically exhibitabnormal stellar velocity dispersions for their stellar mass. Thus, thedriver for dark matter dominated UDGs being offset towards a higherM/L ratio given their dynamical mass is, at least partially, their largersizes compared to that of non-UDGs with similar dynamical masslying on the ‘U-shaped’ relation. As noted by van Dokkum et al.(2019b): "The effective radius always contains 50 % of the light, butit does not contain a fixed fraction of the dark matter" . If we assumea similar dark matter halo structure, for both UDGs and the non-UDGs, then the larger half-light radii of UDGs encapsulates more ofthe dark matter halo. The natural effect of this is a higher dynamicalmass and hence a higher M/L ratio. However, we cannot rule outthe possibility that there is another parameter, so far unaccountedfor, causing these deviations.
The extrapolation of a single dynamical mass measurement into atotal halo mass requires the assumption of a halo profile. Here, weuse the standard cuspy Navarro et al. (1996) (NFW) halo profile,along with cored halo profiles motivated by dwarf galaxy observa-tions (Burkert 1995) and simulations (Di Cintio et al. 2014). We notethere, exists evidence favouring cored halo profiles for UDGs fromboth observations (van Dokkum et al. 2019b; Wasserman et al.2019; Gannon et al. 2020) and galaxy formation simulations (DiCintio et al. 2017; Carleton et al. 2019; Martin et al. 2019). NFWand Di Cintio et al. (2014) profiles are plotted using the methodol-ogy in the appendix of Di Cintio et al. (2014). For our over-densitycriterion we use the standard Δ where our halo mass is measuredwhen the halo density reaches 200 times the critical density of theUniverse. As the Di Cintio et al. (2014) profiles also require a stellarmass for the galaxy we use a stellar mass of 𝑀 ∗ = M (cid:12) whichis approximately the stellar mass of the currently studied UDGs.Burkert (1995) mass profiles are calculated in a similar manner.Noting that, by definition the mass at 𝑅 is 𝑀 , it is possible torecover the core radius of a Burkert (1995) halo using equations 2and 4 of Salucci & Burkert (2000). Once this parameter is knownequation 2 of Salucci & Burkert (2000) provides the mass for aBurkert (1995) halo as a function of radius. MNRAS , 1–15 (2020) J. S. Gannon et al.
Figure 7.
Dynamical mass to light ratio vs dynamical mass within the half-light radius. Non-UDGs (black points) are plotted to establish the ‘U-shaped’relation (Zaritsky et al. 2006; Wolf et al. 2010; Forbes et al. 2011; Cappellari et al. 2013; Toloba et al. 2014). For UDGs, the symbols and colours are as inFigure 6. Most UDGs tend to deviate from ‘U-shaped’ relation.
In Figure 8 we display the resulting halo mass profiles for 1000halos with masses ( 𝑀 ) equally spaced logarithmically between10 M (cid:12) and the 95% confidence upper limit of UDG halo masses(10 . M (cid:12) ; Sifón et al. 2018). We also display the minimum andmaximum halo masses in each of 400 bins (20 in radius and 20in enclosed mass). Finally, we include the logarithmic differencebetween the minimum and maximum halo mass in each bin in orderto quantify the predictive power of differing regions of parameterspace. Halos with total mass below 10 M (cid:12) are not plotted as boththe Di Cintio et al. (2014) and Burkert (1995) halo profiles quicklyrun into numerical issues in this regime. Noting this, Figure 8 isstrictly only valid for differentiating between UDGs in dwarf-likeand more massive halos. It does not display the predictive power ofobservational measurements to distinguish between different dwarf-like dark matter halos.For both the cored profiles of Di Cintio et al. (2014) and Burk-ert (1995), the presence of a dark matter core leads to a complexinterpretation of masses measured within ∼
10 kpc. In both casesthe presence of more massive dark matter halos can lead to less mass within the central regions as these halos become more effi-cient at making a core. A problematic result of this effect is thatobservational mass measurements coming within these central re-gions poorly constrain the total halo mass of the galaxy. This is nottrue for the cuspy NFW halo profile where, at all radii, an increasein mass corresponds to an increase in total halo mass.In the case of our UDG1137+16 dynamical mass measurement,the best fitting NFW profile corresponds to a ∼ × M (cid:12) halo with 1- 𝜎 range extending from approximately 6 × M (cid:12) to 4 × M (cid:12) . At face value this would suggest an under-massive darkmatter halo for UDG1137+16. The only UDG with a resolved stellarvelocity dispersion profile and total halo mass fitting, Dragonfly 44(van Dokkum et al. 2019b; Wasserman et al. 2019), requires astrong tangential anisotropy to fit a cuspy NFW profile to its data.Additionally, there is observational evidence against using NFWprofiles to fit UDG data, coming from the comparison of dynamicalmass measurements to total halo mass estimates from GC counts(Gannon et al. 2020; Forbes et al. 2020a). Noting the previousobservational evidence against NFW profiles for UDGs that aresimilarly dark matter dominated and the improbably low inferredtotal halo masses, we suggest it likely the cuspy NFW profile poorlydescribes the true halo profile of UDG1137+16.In the case of UDG1137+16 all Di Cintio et al. (2014) halosup to a total halo mass of ∼ × M (cid:12) lie within 1- 𝜎 of ourdynamical mass measurement. For the Burkert (1995) halo profiletotal halo masses up to ∼ × M (cid:12) lie within 1- 𝜎 of our dynamicalmass measurement. Once halos reach a total dark matter mass of ∼ M (cid:12) the Burkert (1995) halos again lie within 1- 𝜎 of ourmass measurement as the halos develop a sufficiently large coreto enclose masses less than other less massive halos. We note thatthe Burkert (1995) profiles were observationally motivated, beingbased on dwarf galaxies, and so may not scale well into this moremassive regime. Note that halos of this mass range are not plotted in Figure 8.MNRAS , 1–15 (2020) n UDG1137+16 Figure 8.
An investigation of the predictive power of observational mass measurements. Throughout the plot we are showing the mass enclosed within a radiusfrom halo centre, in solar masses, ( 𝑀 ( < 𝑅 ) ) vs that radius from halo centre, in kpc, (R). We plot for three different halo profiles, that of Navarro et al. (1996)( left column ), Di Cintio et al. (2014) ( middle column ) and Burkert (1995) ( right column ). Top row:
The enclosed mass as a function of radius for each haloprofile. We plot for 1000 halos with masses ( 𝑀 ) equally spaced logarithmically between 10 M (cid:12) and the 95% confidence upper limit of UDG halo masses(10 . M (cid:12) ; Sifón et al. 2018). Second row:
We bin the data in both radius and enclosed mass into 20 equal bins in log space for each respectively. In this rowwe display the minimum halo mass passing through each region.
Third row:
The same as the second row, but instead plotting the maximum halo mass passingthrough each region rather than the minimum.
Bottom row:
The logarithmic difference between the maximum and minimum halo mass in each binned region.Darker regions indicate those where measured masses poorly constrain the total halo mass of the galaxy (refer to scale bar on right). Many UDG dynamicalmasses fall in the ∼ We conclude that it is difficult for our dynamical mass mea-surement to precisely recover a total halo mass for UDG1137+16.This problem will be one common to most UDGs with singularmass measurements at small radii such as those usually calculatedusing velocity dispersions and Equation 2. Differentiation betweendifferent total halo masses requires either a mass measurement ata radius larger than what is typically available in the low surfacebrightness regime (e.g., from X-rays or lensing) or a resolved massprofile for the UDG (e.g., van Dokkum et al. 2019b; Wassermanet al. 2019). As a consequence it is entirely possible that, despitemany UDGs being measured to have dark matter halos of similar inner mass, they reside in dark matter halos of quite different totalmass.
In this work we have studied UDG1137+16 using deep 𝑟 -bandimaging from the Jeanne Rich Telescope and deep spectroscopicdata from the Keck Cosmic Web Imager. Our main conclusions areas follows: • Through analysis of the photometric properties of UDG1137+16
MNRAS , 1–15 (2020) J. S. Gannon et al. and our newly measured recession velocity we determine thisgalaxy is a bona fide UDG ( 𝑅 𝑒 = . ± . 𝜇 ,𝑟 = . ± .
04 mag arcsec − ), confirming Müller et al. (2018)’s hypothesis.Furthermore we place it, along with a nearby galaxy undergoing astrong tidal interaction, in the low density UGC 6594 group. • Our analysis also provides qualitative evidence for a dual stel-lar component in UDG1137+16. We suggest it is likely thatUDG1137+16 has a stellar envelope ( 𝑀 ∗ ≈ . ± . × M (cid:12) ), sur-rounding its central stellar component ( 𝑀 ∗ ≈ . ± . × M (cid:12) ),comprising material redistributed via tidal interaction. • We measure a stellar velocity dispersion of 15 ± − andcalculate a dynamical mass and mass to light ratio for UDG1137+16within 3.9 kpc. Our measurements are consistent with the centralstellar velocity dispersion – stellar mass relation but inconsistentwith the dynamical mass – mass to light ratio relation establishedfor non-UDGs. UDG1137+16 lies at higher mass to light ratio thannon-UDGs at similar dynamical mass. • We examine the relationship between central stellar velocity dis-persion and stellar mass for a sample of UDGs finding that mostare fully consistent with the relationship established for a sampleof non-UDG galaxies. Additionally, we note that there exist fewnon-UDG galaxies, at similar velocity dispersions but much higherstellar masses. This suggests a lack of progenitors for strong tidalstripping formation scenarios (e.g., >
99% mass loss) which woulddisfavour their ability to reproduce the known UDG population.Alternatively, a tidal stripping mechanism that decreases the cen-tral stellar velocity dispersion without significantly disrupting thegalaxy’s morphology is needed. • Finally, we investigate the ability of current dynamical mass mea-surements based on a stellar velocity dispersion to predict total halomasses for UDGs. We find that even under reasonable assumptionsfor a halo profile, current lone dynamical masses based on a stellarvelocity dispersion are unable to constrain total halo mass. In orderto properly constrain total halo mass, mass measurements made atlarger radii or a resolved mass profile are required (e.g., Dragonfly44 van Dokkum et al. 2019b; Wasserman et al. 2019. This also im-plies UDGs that display similar dark matter characteristics at smallradii may not necessarily have the same total halo masses.
ACKNOWLEDGEMENTS
We thank the anonymous referee for their careful consideration ofour work and for providing useful comments to improve upon it. Wethank R. Turner and A. Romanowsky for insightful conversationsthroughout the creation of this work. Some of the data presentedherein were obtained at the W. M. Keck Observatory, which is op-erated as a scientific partnership among the California Institute ofTechnology, the University of California and the National Aeronau-tics and Space Administration. The Observatory was made possibleby the generous financial support of the W. M. Keck Foundation.The authors wish to recognise and acknowledge the very significantcultural role and reverence that the summit of Maunakea has al-ways had within the indigenous Hawaiian community. We are mostfortunate to have the opportunity to conduct observations from thismountain. JSG acknowledges financial support received througha Swinburne University Postgraduate Research Award through-out the creation of this work. DAF thanks the ARC for financialsupport via DP160101608. AFM has received financial supportthrough the Post-doctoral Junior Leader Fellowship Programmefrom “La Caixa” Banking Foundation (LCF/BQ/LI18/11630007).JPB gratefully acknowledges support from National Science foun- dation grants AST- 1518294 and AST-1616598. B.T.D acknowl-edges supports from a Spanish postdoctoral fellowship ‘Ayudas1265 para la atracción del talento investigador. Modalidad 2: jóvenesinvestigadores.’ funded by Comunidad de Madrid under grant num-ber 2016-T2/TIC-2039 and from the grant ‘High-resolution, multi-band analysis of galaxy centers (HiMAGC)’ with reference numberPR65/19-22417 financed by Comunidad de Madrid and Univer-sidad Complutense de Madrid. JR acknowledge financial supportfrom the grants AYA2015-65973-C3-1-R and RTI2018-096228- B-C31 (MINECO/FEDER, UE), as well as from the State Agency forResearch of the Spanish MCIU through the “Center of ExcellenceSevero Ochoa” award to the Instituto de Astrofísica de Andalucía(SEV-2017-0709).
The KCWI data presented are available via the Keck Observa-tory Archive (KOA):
18 months after observations are taken. The JRTdata can be acquired by contacting JSG through the email providedfor correspondence.
REFERENCES
Akhlaghi M., Ichikawa T., 2015, ApJS, 220, 1Alabi A., et al., 2018, MNRAS, 479, 3308Alabi A. B., Romanowsky A. J., Forbes D. A., Brodie J. P., Okabe N., 2020,MNRAS, 496, 3182Amorisco N. C., Loeb A., 2016, MNRAS, 459, L51Beasley M. A., Romanowsky A. J., Pota V., Navarro I. M., Martinez DelgadoD., Neyer F., Deich A. L., 2016, ApJ, 819, L20Bender R., Burstein D., Faber S. M., 1992, ApJ, 399, 462Bertin E., 2006, in Gabriel C., Arviset C., Ponz D., Enrique S., eds, Astro-nomical Society of the Pacific Conference Series Vol. 351, AstronomicalData Analysis Software and Systems XV. p. 112Bertin E., Arnouts S., 1996, A&AS, 117, 393Blom C., Forbes D. A., Foster C., Romanowsky A. J., Brodie J. P., 2014,MNRAS, 439, 2420Burkert A., 1995, ApJ, 447, L25Burkert A., Forbes D. A., 2020, AJ, 159, 56Caon N., Capaccioli M., D’Onofrio M., 1993, MNRAS, 265, 1013Cappellari M., 2017, MNRAS, 466, 798Cappellari M., et al., 2013, MNRAS, 432, 1862Carleton T., Errani R., Cooper M., Kaplinghat M., Peñarrubia J., Guo Y.,2019, MNRAS, 485, 382Chan T. K., Kereš D., Wetzel A., Hopkins P. F., Faucher-Giguère C. A.,El-Badry K., Garrison-Kimmel S., Boylan-Kolchin M., 2018, MNRAS,478, 906Chilingarian I., Cayatte V., Revaz Y., Dodonov S., Durand D., Durret F.,Micol A., Slezak E., 2009, Science, 326, 1379Chilingarian I. V., Afanasiev A. V., Grishin K. A., Fabricant D., Moran S.,2019, ApJ, 884, 79Coelho P. R. T., 2014, MNRAS, 440, 1027Cohen Y., et al., 2018, ApJ, 868, 96Collins M. L. M., Tollerud E. J., Rich R. M., Ibata R. A., Martin N. F.,Chapman S. C., Gilbert K. M., Preston J., 2020, MNRAS, 491, 3496Crnojević D., et al., 2016, ApJ, 823, 19Danieli S., van Dokkum P., Conroy C., Abraham R., Romanowsky A. J.,2019, ApJ, 874, L12Danieli S., van Dokkum P., Abraham R., Conroy C., Dolphin A. E., Ro-manowsky A. J., 2020, ApJ, 895, L4Dey A., et al., 2019, AJ, 157, 168Di Cintio A., Brook C. B., Dutton A. A., Macciò A. V., Stinson G. S., KnebeA., 2014, MNRAS, 441, 2986 MNRAS , 1–15 (2020) n UDG1137+16 Di Cintio A., Brook C. B., Dutton A. A., Macciò A. V., Obreja A., DekelA., 2017, MNRAS, 466, L1Dullo B. T., 2019, ApJ, 886, 80Dullo B. T., Graham A. W., 2013, ApJ, 768, 36Dullo B. T., Graham A. W., Knapen J. H., 2017, MNRAS, 471, 2321Dullo B. T., et al., 2019, ApJ, 871, 9Emsellem E., et al., 2019, A&A, 625, A76Errani R., Penarrubia J., Tormen G., 2015, MNRAS, 449, L46Ferré-Mateu A., et al., 2018, MNRAS, 479, 4891Forbes D. A., Spitler L. R., Graham A. W., Foster C., Hau G. K. T., BensonA., 2011, MNRAS, 413, 2665Forbes D. A., Gannon J. S., Romanowsky A. J., Alabi A., Brodie J. P., CouchW. J., Ferre-Mateu A., 2020a, arXiv e-prints, p. arXiv:2010.07313Forbes D. A., Alabi A., Romanowsky A. J., Brodie J. P., Arimoto N., 2020b,MNRAS, 492, 4874Forbes D. A., Dullo B. T., Gannon J., Couch W. J., Iodice E., Spavone M.,Cantiello M., Schipani P., 2020c, MNRAS, 494, 5293Gannon J. S., Forbes D. A., Romanowsky A. J., Ferré-Mateu A., CouchW. J., Brodie J. P., 2020, MNRAS,Graham A. W., Driver S. P., 2005, Publ. Astron. Soc. Australia, 22, 118Greco J. P., et al., 2018, ApJ, 857, 104Harris W. E., Harris G. L. H., Alessi M., 2013, ApJ, 772, 82Into T., Portinari L., 2013, MNRAS, 430, 2715Jedrzejewski R. I., 1987, MNRAS, 226, 747Jiang F., Dekel A., Freundlich J., Romanowsky A. J., Dutton A. A., MacciòA. V., Di Cintio A., 2019, MNRAS, p. 1490Macciò A. V., Huterer Prats D., Dixon K. L., Buck T., Waterval S., AroraN., Courteau S., Kang X., 2020, arXiv e-prints, p. arXiv:2010.02245Makarov D. I., Sharina M. E., Karachentseva V. E., Karachentsev I. D.,2015, A&A, 581, A82Mancera Piña P. E., et al., 2019, ApJ, 883, L33Martig M., Bournaud F., 2008, MNRAS, 385, L38Martín-Navarro I., et al., 2019, MNRAS, 484, 3425Martin N. F., et al., 2016, ApJ, 833, 167Martin G., et al., 2019, MNRAS, 485, 796Martínez-Delgado D., et al., 2016, AJ, 151, 96McConnachie A. W., 2012, AJ, 144, 4Mihos J. C., et al., 2015, ApJ, 809, L21Monelli M., Trujillo I., 2019, ApJ, 880, L11Montes M., Infante-Sainz R., Madrigal-Aguado A., Román J., MonelliM., Borlaff A. S., Trujillo I., 2020, The galaxy "missing dark matter"NGC1052-DF4 is undergoing tidal disruption ( arXiv:2010.09719 )Morrissey P., et al., 2018, ApJ, 864, 93Mowla L., van Dokkum P., Merritt A., Abraham R., Yagi M., Koda J., 2017,ApJ, 851, 27Muñoz R. R., Majewski S. R., Johnston K. V., 2008, ApJ, 679, 346Müller O., Jerjen H., Binggeli B., 2018, A&A, 615, A105Müller O., et al., 2020, arXiv e-prints, p. arXiv:2006.04606Navarro J. F., Frenk C. S., White S. D. M., 1996, ApJ, 462, 563Pandya V., et al., 2018, ApJ, 858, 29Penny S. J., Janz J., Forbes D. A., Benson A. J., Mould J., 2015, MNRAS,453, 3635Prole D. J., van der Burg R. F. J., Hilker M., Davies J. I., 2019, MNRAS,488, 2143Rich R. M., et al., 2019, MNRAS, 490, 1539Román J., Trujillo I., 2017, MNRAS, 468, 4039Román J., Beasley M. A., Ruiz-Lara T., Valls-Gabaud D., 2019, MNRAS,486, 823Román J., Trujillo I., Montes M., 2020, A&A, 644, A42Ruiz-Lara T., et al., 2019, MNRAS, 486, 5670Sales L. V., Navarro J. F., Peñafiel L., Peng E. W., Lim S., Hernquist L.,2020, MNRAS, 494, 1848Salucci P., Burkert A., 2000, ApJ, 537, L9Schlafly E. F., Finkbeiner D. P., 2011, ApJ, 737, 103Sérsic J. L., 1968, Atlas de Galaxias AustralesSifón C., van der Burg R. F. J., Hoekstra H., Muzzin A., Herbonnet R., 2018,MNRAS, 473, 3747Tollerud E., 2015, https://gist.github.com/eteq/5000843
Toloba E., et al., 2014, ApJS, 215, 17Toloba E., et al., 2018, ApJ, 856, L31Tremmel M., Wright A. C., Brooks A. M., Munshi F., Nagai D., Quinn T. R.,2020, MNRAS, 497, 2786Trujillo I., Aguerri J. A. L., Cepa J., Gutiérrez C. M., 2001, MNRAS, 321,269Trujillo I., et al., 2019, MNRAS, 486, 1192Tully R. B., Courtois H. M., Sorce J. G., 2016, AJ, 152, 50Venhola A., et al., 2017, A&A, 608, A142Walker M. G., Mateo M., Olszewski E. W., Peñarrubia J., Evans N. W.,Gilmore G., 2009, ApJ, 704, 1274Wasserman A., et al., 2019, ApJ, 885, 155Willmer C. N. A., 2018, ApJS, 236, 47Wolf J., Martinez G. D., Bullock J. S., Kaplinghat M., Geha M., MuñozR. R., Simon J. D., Avedo F. F., 2010, MNRAS, 406, 1220Yagi M., Koda J., Komiyama Y., Yamanoi H., 2016, ApJS, 225, 11Yozin C., Bekki K., 2015, MNRAS, 452, 937Zaritsky D., Gonzalez A. H., Zabludoff A. I., 2006, ApJ, 638, 725van Dokkum P. G., Abraham R., Merritt A., Zhang J., Geha M., Conroy C.,2015, ApJ, 798, L45van Dokkum P., et al., 2016, ApJ, 828, L6van Dokkum P., et al., 2017, ApJ, 844, L11van Dokkum P., et al., 2018, Nature, 555, 629van Dokkum P., Danieli S., Abraham R., Conroy C., Romanowsky A. J.,2019a, ApJ, 874, L5van Dokkum P., et al., 2019b, ApJ, 880, 91van Driel W., et al., 2016, A&A, 595, A118
APPENDIX A: UDG PROPERTIES FROM THELITERATURE
In this appendix we provide a literature summary of UDGs witheither stellar or GC velocity dispersions and the resulting dynamicalmasses for them using Equation 2. For objects with both a GC andstellar velocity dispersion (i.e., NGC 5846_UDG1, VCC 1287 andNGC1052-DF2) we quote only the stellar velocity dispersion. Forobjects with velocity dispersions from both MUSE ( 𝜎 𝑖𝑛𝑠𝑡 = 45km s − ) and KCWI ( 𝜎 𝑖𝑛𝑠𝑡 = 13 km s − ) we give preference tothe latter as we expect them to be more accurate due to the betterinstrumental resolution. In both cases literature works that derivethe same parameters (e.g., velocity dispersion) for a UDG but areunused in the table are listed in the notes.Equation 2 was derived with the intent of using the luminosityweighted line-of-sight velocity dispersion as measured within 2Dhalf-light radius (Wolf et al. 2010). Generally, this radius differsfrom that in which the velocity dispersion is actually measured. Inorder to calculate a dynamical mass from the van Dokkum et al.(2019a) GC velocity dispersion we assume that their velocity dis-persion measurement is equivalent to the line-of-sight velocity dis-persion within half-light radius (i.e., a flat velocity profile for thegalaxy). Toloba et al. (2018) used Equation 2 along with the GCvelocity dispersion to calculate their dynamical masses. We notethat, for some GC velocity dispersions the small number statisticscan lead to both accuracy and precision loss in the measurement.There is also some ambiguity as to whether or not all of thegalaxies listed in Table A1 should be treated equally when investi-gating the formation of UDGs. As such we caution the following: • Only 2 of the 9 UDGs in the Chilingarian et al. (2019) sample(those beginning with J in Table A1) meet the standard UDG defi-nition with the other 7 being either slightly too small or slightly toobright. • While Andromeda XIX meets the formal UDG definition it has
MNRAS , 1–15 (2020) J. S. Gannon et al. orders of magnitude less stellar mass than other UDGs and is sig-nificantly fainter than what is realistically observable for non-LocalGroup UDGs. • The UDGs NGC1052-DF2 and NGC1052-DF4 have both beeninferred to have little to no dark matter and there exists debate in theliterature as to their basic properties (e.g., their distance; Trujilloet al. 2019; Monelli & Trujillo 2019 although see Danieli et al.2020). While both clearly fit the standard UDG definition, it isunclear how any single current proposed formation scenario couldcreate both these UDGs and the other known UDGs (over-massiveor otherwise). We therefore advise caution to those treating thesegalaxies as the other UDGs, seeking to find a singular formationmechanism to explain all.It should also be noted that Makarov et al. (2015) reports stellarvelocity dispersions for three UDG-like galaxies. We choose not toinclude them in this compilation as they are measured at ∼
10% ofthe instrumental resolution and are listed as being highly uncertain.
MNRAS , 1–15 (2020) n UDG + Table A1.
A summary of the UDG data plotted in Figures 6 and 7.
Bold values indicate they have been calculated in this work. We use equation 2 to calculate dynamical masses for literature data. ‘–’ denotes valuesmissing. Errors are given in (brackets).Object 𝑅 e , circ 𝜎 𝑀 ( < 𝑟 / ) 𝑀 ∗ M/L Filter References Notes[kpc] [km s − ] [ × 𝑀 (cid:12) ] [ × 𝑀 (cid:12) ] [ 𝑀 (cid:12) / 𝐿 (cid:12) ]Stellar velocity dispersionsUDG1137+16 † ( ) ( ) ( ) ( ) ( ) r This work 1VCC 1287 ∗ ‡ ( ) 56 (10) 130 (50) 27 (2) 71 (25) V Martínez-Delgado et al. (2016); Martín-Navarro et al. (2019) 3Dragonfly 44 ∗ (–) 33 (3) 39 (5) 30 (–) 26 ( + − ) 𝐼 van Dokkum et al. (2016, 2017, 2019b) 4DFX1 ∗ (–) 30 (7) ( ) (–) ( ) van Dokkum et al. (2017) 1; 5; 6NGC 5846_UDG1 † † + . − . ) 1.3 (0.8) 20 (–) ( ) V Cohen et al. (2018); van Dokkum et al. (2018); Danieli et al. (2019) 8; 9Andromeda XIX † + . − . ) 1.1 (0.5) 0.079 (–) 278 ( + − ) V Martin et al. (2016); Collins et al. (2020) 10J125846 . + . ∗ ( ) (–) 16 (4) R Chilingarian et al. (2019) 11; 12J125904 . + . ∗ ( ) (–) 22 (8) R Chilingarian et al. (2019) 11; 12J125904 . + . ∗ ( ) (–) 12 (5) R Chilingarian et al. (2019) 11; 12J125929 . + . ∗ ( ) (–) 4.5 (2.5) R Chilingarian et al. (2019) 11J125937 . + . ∗ ( ) (–) 22 (10) R Chilingarian et al. (2019) 11; 12J130005 . + . ∗ ( ) (–) 47 (17) R Chilingarian et al. (2019) 11; 12J130026 . + . ∗ ( ) (–) 2 (0.6) R Chilingarian et al. (2019) 11J130028 . + . ∗ ( ) (–) 22 (6) R Chilingarian et al. (2019) 11; 11J130038 . + . ∗ ( ) (–) 11 (4) R Chilingarian et al. (2019) 11; 13GC velocity dispersionsVLSB-B ∗ + − ) 49 ( + − ) 0.6 (0.1) 407 ( + − ) V Toloba et al. (2018) 14; 15VLSB-D ∗ + − ) 32 ( + − ) 7.9 (0.1) 21 ( + − ) V Toloba et al. (2018) 14; 15VCC 615 ∗ + − ) 25 ( + . − . ) 2.1 (0.1) 60 ( + − ) V Toloba et al. (2018) 14; 15NGC1052-DF4 † (–) 4.2 ( + . − . ) ( + . − . ) 15 (4) 0.64 ( + . − . ) 𝑉 van Dokkum et al. (2019a) 8; 15; 16 Notes.—1) A 𝑀 ∗ / 𝐿 = 𝑅 𝑒 circularised using literature b/a (0.87 - Martínez-Delgado et al. 2016).4) 𝑅 𝑒 circularised using literature b/a (0.68 - van Dokkum et al. 2017).5) 𝑅 𝑒 circularised using literature b/a (0.62 - van Dokkum et al. 2017).6) It is unclear if the velocity dispersion reported in van Dokkum et al. (2017) for DFX1 isalso effected by the same problem reported in van Dokkum et al. (2019b) for Dragonfly 44from the same work.7) See also Müller et al. (2020) who dubNGC 5846_UDG1, MATLAS-2019.8) 𝑀 ∗ / 𝐿 = 𝑅 𝑒 > . 𝜇 ,𝑔 >
24 mag arcsec − ).14) Dynamical mass is calculated using Equation 2 with 𝑅 𝑒 as the radius containing half thenumber of GCs.15) Dynamical mass calculation assumes GC velocity dispersion = stellar velocity dispersion.16) 𝑅 𝑒 circularised using literature b/a (0.89 - van Dokkum et al. 2019a). ∗ Cluster UDG † Group UDG ‡ Field UDG
This paper has been typeset from a TEX/L A TEX file prepared by the author. M N R A S , (2020