Extensive Globular Cluster Systems Associated with Ultra Diffuse Galaxies in the Coma Cluster

The discovery of large, extremely faint, spheroidal objects in galaxy clusters dates at least to Impey et al. (1988), who noticed several such objects in photographic studies of the Virgo cluster. Over the following three decades several more were found (e.g., Dalcanton et al. 1997), but it was only recognized recently how common they are. Using the Dragonfly Telephoto Array (Abraham & van Dokkum 2014), 47 galaxies with half-light radii ${R}_{{\rm{e}}}\gtrsim 1.5$ kpc and central surface brightness ${\mu }_{g,0}\gtrsim 24$ mag arcsec⁻² were found in the Coma cluster (van Dokkum et al. 2015a). The galaxies appear smooth and spheroidal, and have a much lower Sérsic (1968) index than elliptical galaxies ($n\sim 1$ versus $n\sim 4$ for ellipticals). These remarkable objects were dubbed "ultra diffuse galaxies," or UDGs. The number of known UDGs quickly expanded in the past two years, with many more examples found in Coma (Koda et al. 2015), Virgo (Mihos et al. 2015), other clusters (van der Burg et al. 2016), and in low density environments (Martínez-Delgado et al. 2016; Merritt et al. 2016; Román & Trujillo 2017).

It is still unknown how UDGs fit in the general framework of galaxy formation and evolution. One possibility is that most UDGs are closely related to smaller galaxies of the same luminosity: they may have originated as small galaxies that were puffed up by tidal interactions (see, e.g., Collins et al. 2013), or represent the high angular momentum tail of the general population of dwarf galaxies (Amorisco & Loeb 2016). Another possibility is that many UDGs are "failed" galaxies, with truncated star formation histories. Strong feedback from supernovae or active nuclei could produce underluminous galaxies, perhaps in combination with environmental effects (Agertz & Kravtsov 2015; Yozin & Bekki 2015; Di Cintio et al. 2017).

Intriguingly, an important clue to the formation of these diffuse galaxies comes from the most compact stellar systems in the universe. Beasley et al. (2016) found that the UDG VCC 1287 in Virgo has a surprisingly large number of globular clusters for its luminosity. Similar results were subsequently reported for the Coma UDGs Dragonfly 17 (Peng & Lim 2016) and Dragonfly 44 (van Dokkum et al. 2016). These early results, together with the first measurements of the kinematics of UDGs (Beasley et al. 2016; van Dokkum et al. 2016), indicated that UDGs are fundamentally different from other galaxies of the same luminosity.

However, other studies have cast doubt on this interpretation. Some large, low surface brightness objects seem to be tidally disrupted low mass galaxies (Collins et al. 2013; Merritt et al. 2016), and there is large variation in the cold gas fraction among field UDGs (Papastergis et al. 2017). Furthermore, it has been suggested that massive, globular cluster-rich systems are the exception, not the rule: Amorisco et al. (2017) report that UDGs have no statistically significant excess of globular clusters compared to normal dwarf galaxies with the same stellar mass. Amorisco et al. come to this conclusion from a comparison of the positions of compact objects in the Hammer et al. (2010) HST/ACS Coma Cluster Treasury program (CCTp; Carter et al. 2008) catalog to the positions of low surface brightness objects in the Yagi et al. (2016) catalog.

In this Letter we contribute to this discussion by measuring the globular cluster populations in two large Coma UDGs using the Hubble Space Telescope (HST). We also analyze all archival HST/ACS images of UDGs in Coma.

2.1. Kinematics

2.2. HST Imaging

HST imaging for Dragonfly 44 and DFX1 was obtained in the Cycle 24 program GO-14643. Each galaxy was observed for three orbits in V₆₀₆ and one orbit in I₈₁₄, using a standard dither pattern to eliminate hot pixels. We used ACS/WFC for DFX1 but WFC3/UVIS for Dragonfly 44 as this enabled us to simultaneously observe Dragonfly 42 in a parallel ACS observation. The CTE-corrected, drizzled images created by the STScI pipeline were used. The three V₆₀₆ images were rotated and shifted to the frame of the I₈₁₄ image and combined. In the combination step remaining deviant pixels in the individual images were replaced by the average of the other two frames. The point-source depth was measured from the rms of the counts in empty apertures with diameter d = 8 pixels, corrected to $d=\infty$ using theoretical growth curves (see Labbé et al. 2003). We find $5\sigma$ AB depths of ${V}_{606}=28.4$ and ${I}_{814}=26.8$ for Dragonfly 44 and ${V}_{606}=27.9$ and ${I}_{814}=27.0$ for DFX1. The relatively modest depth of the ACS imaging can be attributed to the now quite severe CTE effects.

The HST images of Dragonfly 44 and DFX1 are shown in Figure 1; Dragonfly 42 is discussed in Section 4. The left panels are color images created from the V₆₀₆ and I₈₁₄ exposures; the right panels show the deep V₆₀₆ data at high contrast after masking spatially extended objects (see Section 4). Both galaxies are smooth and elongated, with no obvious tidal features, spiral arms, star-forming regions, or other irregularities. The most striking aspect of Figure 1, and the central topic of this Letter, is the fact that both UDGs are associated with a large number of compact objects. For Dragonfly 44 this was already seen in ground-based imaging, although not as clearly (see van Dokkum et al. 2016). For both galaxies the distribution of compact objects has a broadly similar orientation and flattening as the smooth light.

Zoom In Zoom Out Reset image size

The compact objects were identified and characterized in the following way. First, the V₆₀₆ light of the UDGs was fit with a 2D Sérsic profile, using the GALFIT code (Peng et al. 2002). Neighboring objects, as well as the compact sources, were masked. The fit was done multiple times, improving the mask in each iteration. The best-fitting Sérsic model has effective radius ${R}_{{\rm{e}}}=4.7\,\mathrm{kpc}$, central surface brightness ${\mu }_{0,V}=24.1$, and Sérsic index n = 0.94 for Dragonfly 44 and ${R}_{{\rm{e}}}=3.5\,\mathrm{kpc}$, ${\mu }_{0,V}=24.0$, and n = 0.90 for DFX1. The results for Dragonfly 44 are in good agreement with previous ground-based measurements (van Dokkum et al. 2015b, 2016). We also fit the I₈₁₄ data, keeping all parameters except the sky value and the normalization fixed to the V₆₀₆ results. The colors of the two galaxies are the same, within the uncertainties: ${V}_{606}-{I}_{814}=0.48\pm 0.06$ for Dragonfly 44 and ${V}_{606}-{I}_{814}=0.45\pm 0.06$ for DFX1. The total magnitudes are ${V}_{606}=18.8$ and ${V}_{606}=19.3$ respectively.

After subtracting the best-fitting GALFIT models an object catalog was created using SExtractor (Bertin & Arnouts 1996), using default parameters. Globular cluster candidates were selected by the criterion $0.5\lt c\lt 1.0$, where c is the flux ratio in d = 4 pixel and d = 8 pixel apertures. Stars with a high signal-to-noise ratio (S/N) have $c\approx 0.75;$ the broad selection range ensures that unresolved objects with low S/N are included in the sample, at the expense of some contamination by compact galaxies.

The surface density of compact objects associated with the two galaxies is shown in the left panel of Figure 2. The number density was measured in elliptical annuli that are scaled to the half-light radius. In each annulus the number of compact objects with ${V}_{606}\lt 28$ was measured and divided by the area of the annulus (open circles). A contamination correction was applied by subtracting the average number density of objects with $0.5\lt c\lt 1.0$, $V\lt 28$, and distances $R\gt 3{R}_{{\rm{e}}}$. The measurements were done separately for Dragonfly 44 and DFX1 and then averaged.

Zoom In Zoom Out Reset image size

The radial distribution confirms the visual impression of a significant overdensity of compact objects. We fit a Sérsic profile to the combined, binned distribution using the emcee methodology (Foreman-Mackey et al. 2013), with a prior on the Sérsic index of $n\leqslant 4$. The best fit has $n={3.1}_{-0.9}^{+0.6}$ and ${R}_{\mathrm{gc}}={2.2}_{-0.7}^{+1.3}{R}_{{\rm{e}}}$, with ${R}_{\mathrm{gc}}$ the half-number radius of the globular cluster distribution.⁷ Forcing n = 1, i.e., a similar functional form as the stellar light, we find ${R}_{\mathrm{gc}}={1.4}_{-0.2}^{+0.2}{R}_{{\rm{e}}}$. We conclude that the distribution of globular clusters is more extended than the galaxy light, as was previously found for luminous galaxies (Hargis & Rhode 2014; Kartha et al. 2014) and the UDG Dragonfly 17 (Peng & Lim 2016), but that the precise value of ${R}_{\mathrm{gc}}$ is not well constrained by our data.

The luminosity function of compact objects within $R=1.5{R}_{{\rm{e}}}$ is shown in the right panel of Figure 2. The number rises sharply from ${V}_{606}\sim 26$ to ${V}_{606}\sim 27.5$, where it seems to plateau. The canonical luminosity function of globular clusters is a Gaussian with a width of $\sigma \approx 1$ mag and a peak at $\langle {V}_{606}\rangle \approx 27.6$ for the Coma distance (see, e.g., Miller & Lotz 2007; Lee & Jang 2016; Peng & Lim 2016). We cannot constrain the peak magnitude very well with our data, as the luminosity function does not show a clear turnover. Fitting a Gaussian with a prior $\langle {V}_{606}\rangle \lt 28$, we find $\langle {V}_{606}\rangle ={27.7}_{-0.2}^{+0.2}$ mag and $\sigma ={0.82}_{-0.15}^{+0.16}$ mag.

We conclude that the properties of the compact objects in Dragonfly 44 and DFX1 are consistent with those expected from previously studied globular cluster populations of other galaxies. To estimate the total number of clusters in each galaxy we use ${N}_{\mathrm{gc}}=4{N}_{\mathrm{gc},\mathrm{obs}}$, where ${N}_{\mathrm{gc},\mathrm{obs}}$ is the contamination-corrected number of compact objects with $R\lt 1.5{R}_{{\rm{e}}}$ and ${V}_{606}\lt 27.6$. We find ${N}_{\mathrm{gc}}=74\pm 18$ for Dragonfly 44 and ${N}_{\mathrm{gc}}=62\pm 17$ for DFX1. The number for Dragonfly 44 is consistent with our previous measurement from ground-based imaging (${N}_{\mathrm{gc}}={94}_{-20}^{+25};$ van Dokkum et al. 2016).

Finally, we note that the globular clusters are blue and that their colors are similar to that of the smooth light of the UDGs, as was previously found by Beasley & Trujillo (2016) for Dragonfly 17. Due to the limited depth of the I₈₁₄ data we can only measure reliable colors for the brightest clusters. For ${V}_{606}\lt 26.5$ and $R\lt 1.5{R}_{{\rm{e}}}$ we find $\langle {V}_{606}-{I}_{814}\rangle =0.37\pm 0.06$.

We obtained the ACS images of all 54 low surface brightness objects from the Yagi et al. (2016) catalog that fall in the Coma Cluster Treasury program area from MAST,⁸ and analyzed these galaxies in the same way as described above. Most of the CCTp data consist of a single orbit in g₄₇₅ and a single orbit in I₈₁₄. We added the images to increase the S/N, using $V^{\prime} =\sqrt{2}{g}_{475}+{I}_{814}/\sqrt{2}$. We use a zeropoint of 27.14, as for this value derived magnitudes are equivalent to V₆₀₆ for objects with the colors of UDGs and their globular clusters.

Structural parameters of the galaxies were determined using GALFIT, following the same masking procedures as described in Section 3. Only 12 of the 54 objects have ${R}_{{\rm{e}}}\gt 1.5$ kpc and are classified as UDGs. The remaining galaxies are up to a factor of two smaller than this limit. After subtracting the best-fitting GALFIT models compact objects were identified, again using the same methodology and criteria as used for Dragonfly 44 and DFX1. The number of globular clusters was then determined by measuring the number of compact objects with $V^{\prime} \lt 27.6$ in an elliptical aperture with radius $R=1.5{R}_{{\rm{e}}}$ and multiplying this by 4.

The results are listed in Table 1 and summarized in Figure 3. Dragonfly 44 and DFX1 have the most dramatic globular cluster populations of all HST-observed UDGs in Coma, but they are not the only ones with significant overdensities of point sources.⁹ We find that half of the 12 Yagi et al. (2016) UDGs have overdensities that are significant at the $\gt 2\sigma$ level. Among the Yagi et al. (2016) objects galaxy 358 has the largest number of globular clusters, with ${N}_{\mathrm{gc}}=45\pm 14$. We also include Dragonfly 42, observed in parallel with Dragonfly 44, and Dragonfly 17. The globular clusters in Dragonfly 17 were previously studied by Peng & Lim (2016) and Beasley et al. (2016). Our measurement is consistent with these studies (${N}_{\mathrm{gc}}=25\pm 11$ versus ${N}_{\mathrm{gc}}=28\pm 14$).

Zoom In Zoom Out Reset image size

Table 1.  Structural Parameters and Globular Cluster Counts

Id	MV	${R}_{{\rm{e}}}$	$\mu (0,g)$	n	b/a	${N}_{\mathrm{gc}}$
		(kpc)
DF 17	−15.3	3.3	25.0	0.61	0.71	25 ± 11
DF 42	−14.7	2.8	25.2	0.64	0.61	9 ± 7
DF 44	−16.2	4.7	24.2	0.94	0.68	76 ± 18
DF X1	−15.8	3.5	24.1	0.90	0.62	63 ± 17
Y 112	−14.2	1.8	23.8	1.43	0.81	15 ± 9
Y 121	−14.0	1.5	24.7	0.60	0.69	25 ± 11
Y 122	−13.8	2.4	25.5	0.64	0.54	0 ± 4
Y 358	−14.8	2.3	24.4	0.99	0.83	45 ± 14
Y 367	−13.7	1.7	24.9	0.84	0.73	31 ± 12
Y 370	−13.9	2.1	25.5	0.78	0.92	9 ± 10
Y 386	−14.7	2.6	25.2	0.53	0.63	3 ± 8
Y 419a	−14.6	1.7	24.4	0.62	0.78	23 ± 11
Y 424b	−11.7:	1.7:	26.8:	0.50:	0.41:	17 ± 9
Y 425	−13.3	1.8	25.2	1.33	0.99	24 ± 11
Y 436	−13.5	1.7	25.4	0.58	0.69	34 ± 13
Y 534	−13.9	1.9	25.0	1.03	0.96	28 ± 11

Notes.

^aY 419 may be a superposition of two smaller galaxies. ^bY 424 is barely detected in the HST images.

Download table as:  ASCIITypeset image

The globular counts in UDGs are compared to those in other galaxies in Figure 4. Open symbols in this Figure are taken from the literature compilation of Harris et al. (2013).¹⁰ UDGs have more globular clusters than other galaxies of the same total luminosity. For this sample of 16 UDGs the median difference¹¹ is a factor of ${6.9}_{-2.4}^{+1.0}$. The specific frequency, defined as ${S}_{{\rm{N}}}={N}_{\mathrm{gc}}\times {10}^{0.4({M}_{V}+15)}$, is $10\lesssim {S}_{{\rm{N}}}\lesssim 100$ for UDGs.

Zoom In Zoom Out Reset image size

Using newly obtained HST images of the two UDGs in Coma with measured kinematics we find that they have remarkable globular cluster populations. No other known galaxies look like the objects in Figure 1: very diffuse "blobs" as large as the Milky Way, sprinkled with many extremely compact sources. Dragonfly 44 and DFX1 are both large and relatively bright among UDGs. Although none of the smaller and fainter UDGs that were imaged serendipitously in the Coma Cluster Treasury program have quite as many globular clusters, several come close (see Figure 4). Their median globular cluster specific frequency is actually higher than that of Dragonfly 44 and DFX1 ($\langle {S}_{{\rm{N}}}\rangle =45$ versus $\langle {S}_{{\rm{N}}}\rangle =27$), because they are so faint.

Our results seem to be at odds with Amorisco et al. (2017), who report that UDGs in the Coma cluster do not have a statistically significant excesss of compact objects compared to normal dwarf galaxies (see Section 1). This tension may be partly due to the inclusion of galaxies with ${R}_{{\rm{e}}}\lt 1.5\,\mathrm{kpc}$ in that study, and partly to differences in selection techniques: as an example, Amorisco et al. (2017) identify only a single compact object in galaxy 358 (N. Amorisco 2017, private communication), whereas we find 13 with ${V}_{606}\lt 27.6$ and $R\lt 1.5{R}_{{\rm{e}}}$ (${N}_{\mathrm{gc},\mathrm{obs}}=11.2$ after correcting for contamination, and ${N}_{\mathrm{gc},\mathrm{tot}}\,=4{N}_{\mathrm{gc},\mathrm{obs}}=45$).¹²

The results presented here, and particularly Figure 4, put to rest the suggestion that most cluster UDGs are directly related to smaller galaxies of the same total luminosity. Although some UDGs may be rapidly spinning low mass galaxies (Amorisco & Loeb 2016), and quite a few are probably tidally distorted objects on the verge of complete disruption (see Collins et al. 2013; Merritt et al. 2016), the majority appear to have a different origin.

Several studies have suggested that the number of globular clusters is more closely related to the dark matter halo mass of a galaxy than to its stellar content (Blakeslee et al. 1997; Forbes et al. 2016; Harris et al. 2017). Although we cannot test this directly for UDGs, as no halo masses out to large radius have yet been measured, we can determine whether UDGs fall on the same relation between ${N}_{\mathrm{gc}}$ and the dynamical mass within the effective radius as other galaxies. This relation is shown in the left panel of Figure 5, with ${M}_{\mathrm{dyn}}(\lt {R}_{{\rm{e}}})\approx 9.3\times {10}^{5}{\sigma }^{2}{R}_{{\rm{e}}}\sqrt{b/a}$ (Wolf et al. 2010). Solid symbols are the three UDGs with measured kinematics: Dragonfly 44, DFX1, and the Virgo galaxy VCC 1287 (Beasley et al. 2016). Open symbols are from Harris et al. (2013), after applying an offset of 0.3 dex to account for the contribution of baryons inside ${R}_{{\rm{e}}}$ (Auger et al. 2010; Grillo 2010).¹³ The UDGs fall on the same relation as other galaxies.

Zoom In Zoom Out Reset image size

Encouraged by this result, and following Peng & Lim (2016) and Beasley & Trujillo (2016), we converted ${N}_{\mathrm{gc}}$ to halo mass using $\mathrm{log}{M}_{\mathrm{halo}}=9.62+1.12\mathrm{log}{N}_{\mathrm{gc}}$ (Harris et al. 2017). The median inferred halo mass is ${M}_{\mathrm{halo}}\sim 1.5\times {10}^{11}$ ${M}_{\odot }$ for the sixteen galaxies; Dragonfly 44 and DFX1 have inferred ${M}_{\mathrm{halo}}\sim 5\times {10}^{11}$ ${M}_{\odot }$. In the right panel of Figure 5 we show the relation between stellar mass and halo mass. The stellar masses were determined from the total magnitudes using Bell & de Jong (2001), with the assumption that all UDGs have the same V − I color as Dragonfly 44 and DFX1. Open symbols are derived from the Harris et al. (2013) sample of normal galaxies, using their V − K colors to transform luminosity to mass. The UDGs fall below the canonical relations between stellar mass and halo mass, suggesting they are "failed" galaxies that quenched after forming their globular clusters but before forming a disk and bulge (see also Beasley & Trujillo 2016; Peng & Lim 2016; van Dokkum et al. 2016).

This study can be extended and improved in various ways. More dynamical measurements are needed to test whether the globular cluster counts are indeed directly related to the dark matter content, and to test whether there is a simple relation between the structure and kinematics of UDGs (Zaritsky 2017). The UDGs that overlap with the CCTp program are relatively small—none were in the original Dragonfly sample—and HST imaging of more UDGs with ${R}_{{\rm{e}}}\gtrsim 3\,\mathrm{kpc}$ may turn up even more spectacular objects than Dragonfly 44 and DFX1. The best way to measure total masses is probably through weak lensing. A recent ground-based study has provided the first upper limits (Sifón et al. 2017, see Figure 5), and future HST studies of large samples could probe deeper into the relevant halo mass range of 10¹¹ ${M}_{\odot }$–10¹² ${M}_{\odot }$.

We thank the anonymous referee for a constructive report and for independently inspecting the HST images. Support from HST grant GO-14643 and NSF grants AST-1616598, AST-1616710, and AST-13123761 is gratefully acknowledged. A.J.R. is a Research Corporation for Science Advancement Cottrell Scholar.