The Densest Galaxy

INTRODUCTION

Objects with sizes and masses between those of globular clusters and compact ellipticals (r h ∼ 10-100 pc; M V ∼ −9 to −14) were first discovered in spectroscopic surveys of galaxy clusters (Hilker et al. 1999;Drinkwater et al. 2000). They were quickly dubbed "ultra-compact dwarf" galaxies (UCDs), even though their galaxian nature was unclear. Large populations of UCDs have been discovered in Fornax, Virgo, and other galaxy clusters, as well as in group and field environments-see reviews in Chilingarian et al. (2011), Norris & Kannappan (2011), and Brodie et al. (2011).

UCD formation scenarios have coalesced around two poles: star cluster or galaxy. In the former scenario, UCDs form the massive end of the normal sequence of globular clusters (Mieske et al. 2012). Further, if some star clusters form in gravitationally-bound complexes, these can merge to make objects that are larger and more massive than single clusters (Brüns et al. 2011).

Alternatively, UCDs could be galaxies that formed in individual dark matter halos-either "in situ", as unusual, extremely compact galaxies-or as the products of tidal stripping of more massive progenitor galaxies (e.g., Drinkwater et al. 2003).

A reasonable synthesis of these scenarios may be that the least-massive "UCDs", with ∼ 10 6 M , are largely star clusters, while the most massive objects ( 10 8 M ) are galaxies, or the tidally stripped remnants thereof. At intermediate masses both star clusters and galaxies may co-exist (e.g., Norris & Kannappan 2011;Brodie et al. 2011).

There is more at stake than the natural desire to understand these novel stellar systems. If a significant fraction of UCDs contain dark matter, then they form a populous class that must be included in counts of subhalos for comparisons to cosmological theory. Further, if some UCDs are formed by tidal stripping, their chemical and structural properties help trace galaxy transformation.

Here we report the discovery of an extraordinary UCD around the Virgo elliptical NGC 4649 (M60). It has a half-light radius of 24 pc but a stellar mass of 2×10 8 M , giving it the highest surface density of any galaxy in the local universe. We also present evidence that this UCD may contain a central supermassive black hole.

DATA

Imaging

We discovered M60-UCD1 in the Hubble Space Telescope/Advanced Camera for Surveys imaging of Strader et al. (2012). We have a single orbit of imaging split between F 475W and F 850LP (hereafter g and z). M60-UCD1 is located at (R.A., Dec.) = (190.8999, 11.5347) arXiv:1307.7707v2 [astro-ph.CO] 14 Aug 2013 in decimal J2000 coordinates. This is at a projected distance of only ∼ 6.6 kpc from the center of M60 ( Figure 1; assuming a distance of 16.5 Mpc; Blakeslee et al. 2009). No mention is made of M60-UCD1 in previous Virgo surveys, including the ACS Virgo Cluster Survey (Côté et al. 2004). It is present in the SDSS DR7 photometric catalog (Abazajian et al. 2009) as J124335.96+113204.6, and was classified by Simard et al. (2011) as a background galaxy.

Aperture photometry of M60-UCD1 gives a total integrated magnitude of z = 15.86 ± 0.02 and g = 17.40 ± 0.02, yielding g − z = 1.54 ± 0.03 (the g and z magnitudes in this paper are AB). The measured half-light radius (see below) is r e = 24.2 ± 0.5 pc. The inferred total luminosities are: L g = (3.26 ± 0.06) × 10 7 L ; L z = (7.88 ± 0.14) × 10 7 L ; and L V = (4.12 ± 0.08) × 10 7 L . With M V = −14.2, M60-UCD1 is the most luminous UCD known (see §4.1 for further discussion).

We fit the optical imaging with one and two-component elliptical Sérsic models. These fits, shown in Figure 2, were performed by fitting two-dimensional models convolved with an empirical point spread function (PSF) using a custom software package as described in Seth et al. (2006). The PSF (10× subsampled) was derived from point sources in the images. The fits were performed on a 5 ×5 image centered on the UCD, with the background galaxy gradient from M60 itself removed. The fits are not very sensitive to the fitting box size.

From the residual map ( Figure 2), it is clear that a single Sérsic component provides a poor fit. In particular, the ellipticity of M60-UCD1 becomes more circular at larger radii, leaving a residual along the minor axis. The radial shape of the surface brightness profile is also poorly fit.

However, using a two-component Sérsic model, a very good fit (χ 2 ν =1.07 in g) is obtained. Table 1 gives the parameters for the single and double Sérsic g band fits and the double Sérsic z fits. For the two-component fits the Sérsic parameters n and r e are very similar between the filters. Because the g band provides a much better fit (probably due to PSF modeling issues in z), all structural values cited are from the g fits.

For the best-fit two-component model, the inner component is compact (r e = 14 pc), with modest ellipticity ( = 0.25), and has about 58% of the total luminosity of M60-UCD1. The outer component is more extended (r e = 49 pc), round, and with a nearly exponential profile (n ∼ 1.2). The overall half-light radius is r e = 24.2 ± 0.5 pc, empirically measured using the deconvolved g surface brightness profile. This value is similar to the radius derived from the single component Sérsic fit; within this radius we estimate Using the Keck/DEIMOS spectrum, the integrated velocity dispersion of M60-UCD1 was measured by crosscorrelating the region around the Ca triplet with a library of templates of the same resolution and wavelength coverage, as described by Strader et al. (2011). This value is σ p = 68 ± 5 km s −1 . The radial velocity of M60-UCD1 is 1290 ± 5 km s −1 ; the systemic velocity of M60 is 1117 km s −1 (González 1993).

We estimate a dynamical mass for M60-UCD1 using the virial theorem:

e r e /G. β is a parameter that depends on the structure of the galaxy and is smaller for more concentrated systems; σ e is the integrated velocity dispersion within r e . Following the results of Cappellari et al. 2006for a range of Sérsic profiles, we assume β = 7, intermediate between the applicable values for the n = 3.3 and n = 1.2 profiles (corresponding to the inner and outer components respectively). We further estimate that σ e = 71 ± 5 km s −1 , slightly higher than the measured value of σ p , by integrating over our DEIMOS extraction window (1.2 × 1.0 ) and accounting for seeing.

The dynamical mass determined in this manner is M vir = (2.0 ± 0.3) × 10 8 M . The systematic uncertainties are significant: we have assumed isotropy, sphericity, and mass-follows-light. Dividing this dynamical mass by the total luminosity of M60-UCD1 yields a mass-to-light ratio of M/L V = 4.9±0.7. The flexible stellar population synthesis models of Conroy et al. (2009), using Padova isochrones and a Kroupa initial mass function (IMF), predict M/L V = (3.5, 4.2, 4.7) for solar metallicity and ages of (8,10,12) Gyr, respectively 8 . If M60-UCD1 has a younger age, the dynamical M/L V could imply an elevation in lowmass stars or stellar remnants over Kroupa IMF model predictions. For older ages there is an excellent match between the observed M/L V and the model predictions. As discussed in §3.3, a modest increase in the central velocity dispersion (and hence M/L) could also be caused by the presence of a supermassive black hole with a mass ∼ 10% of that of the UCD (Mieske et al. 2013). Dark matter is not expected to contribute to the M/L ( §4.2).

M60-UCD1 is marginally resolved in our DEIMOS observations, and so some spatially-resolved kinematic information is available. The 5 slitlet was aligned close to the major axis of the object. Using the sky-subtracted two-dimensional spectrum, we determined the radial ve- locity and velocity dispersion on a pixel-by-pixel basis (one pixel is ∼ 0.12 ). There is clear rotation present, with an amplitude of ∼ 30 km s −1 to a projected radius of ∼ 1.1 . A decline of comparable amplitude in the velocity dispersion is also observed. Since the radial profiles are strongly affected by seeing, we do not attempt dynamical modeling. However, these data provide motivation to obtain improved kinematic maps in the future using integral-field spectroscopy.

Spectroscopy

A spectrum of M60-UCD1 was obtained on the night of 17 January 2012 with Keck/DEIMOS (Faber et al. 2003), utilizing the 1200 l/mm grating centered at 7800Å and a 1 slit (resolution ∼ 1.5Å). We obtained three 30min exposures in 0.8 seeing. Using the spec2d pipeline (Cooper et al. 2012), the spectra were extracted, calibrated, and combined in the standard manner to produce a final one-dimensional spectrum.

To improve stellar population constraints, further spectroscopy was undertaken with MMT/Hectospec (Fabricant et al. 2005) on 16 May 2012, using the 270 l/mm grating with wavelength coverage from 3700 to 9100Å and 5Å resolution. Three 20-min exposures were taken in 0.9 seeing. These Hectospec data were pipelinereduced in a standard manner as described in Mink et al. (2007).

ANALYSIS AND RESULTS

Abundances

We constrain the stellar populations of M60-UCD1 using our MMT/Hectospec spectrum (with its wide wavelength range) and the models of Conroy & van Dokkum (2012a; with additions from Conroy & van Dokkum 2012b). These are stellar population synthesis models with variable abundance ratios for 11 elements. A Markov Chain Monte Carlo method is used to simultaneously fit the entire available optical spectrum.

The derived values are listed in Table 2. The uncertainties quoted are solely statistical, and do not include the substantial systematic uncertainties necessarily present in any integrated-light study of stellar populations. The rms residuals in the fit were < 1% over most of the spectrum. The formal age is 14.5 ± 0.5 Gyr, indicating an old stellar population.

M60-UCD1 is of solar metallicity with a mild elevation in [α/Fe] over solar. The abundances for C, O, and αelements appear very similar to the mean values for highσ local early-type galaxies determined in a similar manner (Conroy et al. 2013). However, the abundance of N is unusual: it is strongly enhanced, with [N/Fe] ∼ +0.6, comparable to the average value observed in globular clusters (e.g., Briley et al. 2004). The high abundance of N in globular clusters is generally attributed to selfenrichment by the winds of asymptotic giant branch stars (Gratton et al. 2012). Our results suggest light-element self-enrichment may also be present in UCDs, presum- The two-dimensional residuals of the bestfitting single Sérsic (left) and double Sérsic (right) fits. Contours show the g surface brightness at µg = 17 to 23 mag/arcsec 2 . Bottom panel: One-dimensional profile showing the results of our two-dimensional fits to the g surface brightness profile of M60-UCD1. The fit is shown in the top panel and the residual (data minus model, units of mag/arcsec 2 ) in the bottom panel. Open black diamonds show the data and residuals, green dashed and blue dot-dashed lines the inner and outer components, and solid red line the sum. All fits were performed in two dimensions; these profiles are for display only.

ably related to their high stellar densities.

We note [Na/Fe] varies strongly with σ in early-type galaxies, increasing from ∼ 0 to ∼ +0.4 for σ ∼ 140 to 300. The M60-UCD1 abundance ([Na/Fe] ∼ +0.4) is therefore difficult to interpret: it could be expected, or could represent a large enhancement over baseline.

As a check on these values, we performed a standard Lick index analysis using EZ Ages (Graves & Schiavon 2008). These values are also listed in Table 2, and the same caveats apply. This analysis gave a formal age of ∼ 9-11 Gyr and similar abundance values for most of the elements in common with the full-spectrum analysis (C, Ca, Mg, Fe). For N the EZ Ages analysis does not yield a reliable value, as the CN index strength is off the grids. The Lick CN 2 index for M60-UCD1 (0.24 mag) is comparable to that in most metal-rich M31 globular clusters, which are also thought to be self-enriched (Schiavon et al. 2012).

The very high abundance of N appears to be a robust conclusion of the analysis.

3.3. X-ray Data: Central Black Hole or X-ray Binary? An X-ray source at the position of M60-UCD1 is present in the Chandra/ACIS catalog of Luo et al. (2013). The central astrometric matching between the Chandra and HST data is excellent due to the large X-ray binary and globular cluster populations of M60, with an rms scatter of 0.17 . This X-ray source, termed XID 144 by Luo et al. (2013), has a position consistent with the optical center of M60-UCD1. There is evidence that it is variable, with its X-ray luminosity (from 0.3 to 8 keV) ranging from ∼ 6 × 10 37 to ∼ 1.3 × 10 38 erg/s over timescales as short as a few months. The X-ray spectrum is well-fit by a absorbed power-law with a photon index of 1.8.

There are two reasonable possibilities for this central X-ray source: it could either be an active galactic nucleus associated with a massive black hole or a low-mass X-ray binary.

The case for a central black hole is straightforward. If the black hole occupation fraction of dwarf galaxies is high, and if UCDs are the products of tidal stripping of dwarf galaxies, then a significant fraction of UCDs should have "overmassive" black holes that could be detected through dynamical or accretion signatures. If UCDs have been stripped of 99% or more of their original mass (we estimate in §4.2 that the progenitor of M60-UCD1 was ∼ 50-200 times more massive), then they could host supermassive black holes that are 10% of their presentday masses (Mieske et al. 2013). Frank et al. (2011) constrain a putative black hole to be < 5% of the total mass of one Fornax cluster UCD through integral-field spectroscopy.

The observed X-ray luminosity would be consistent with a ∼ 10 7 M black hole accreting at 10 −4 of the Eddington rate with a radiative efficiency of 10 −3 . This Eddington ratio of 10 −7 would be typical of nuclei with old stellar populations (Ho 2009).

We can also estimate the odds that M60-UCD1 contains a bright X-ray binary. Sivakoff et al. (2007) derive formulae to estimate the probability that a globular cluster contains a low-mass X-ray binary with L X > 3.2 × 10 38 erg/s. The odds are higher for metal-rich clusters and those with high encounter rates. Applying their results, but extrapolating to the fainter luminosity observed, suggests a ∼ 25% chance of having observed a low-mass X-ray binary in M60-UCD1. However, these results are of uncertain relevance for an object with a different structure and formation history than a globular cluster (since what is pertinent is the integratednot instantaneous-collision rate). Dabringhausen et al. (2012) suggest UCDs have a lower occurrence of lowmass X-ray binaries than expected on the basis of the Sivakoff et al. (2007) results. vs. log L V for dispersion-supported stellar systems (GC=globular cluster; cE=compact elliptical; E=early-type galaxy; dE=dwarf elliptical). The inner component and overall parameters for M60-UCD1 (red stars) are marked, as are the comparison objects M32 and the luminous M31 cluster G1 (blue squares). Globular clusters (the union of objects with re < 10 pc and non-dwarf galaxies with M V > −9; Brodie et al. 2011) are small points; galaxies are large points. M60-UCD1 has a higher Σ than any other galaxy. The black arrow represents the proposed evolution of the progenitor of M60-UCD1 as it was tidally stripped. The principal data source for this figure is the spectroscopically-confirmed compilation of Brodie et al. (2011) a , with updates from Forbes et al. (2013).

a See http://sages.ucolick.org/downloads/sizetable.txt

Future observations can help clarify the nature of the X-ray source. For example, if M60-UCD1 hosts a 10 6 M black hole that lies on the radio-X-ray fundamental plane for black holes (Plotkin et al. 2012), it should be detectable with the Very Large Array. Figure 3 shows a plot of log Σ vs. log L V for dispersionsupported stellar systems. Σ is the mean surface luminosity density within r e . Globular clusters are plotted with different symbols than galaxies. It is clear that M60-UCD1 is an unusual object: it is much denser than any other object classified as a galaxy. It is more massive than any UCD or star cluster of comparable size, but is much more compact than other galaxies of similar luminosity.

DISCUSSION

The Densest Galaxy?

M60-UCD1 is not the densest stellar system known. That honor goes to any of a number of nuclear star clusters, which can reach mean surface densities of > 10 5 M pc −2 within r e (Walcher et al. 2005; these are not plotted in Figure 3). Many massive globular clusters are also extremely dense. However, M60-UCD1 is arguably the densest galaxy known in the local universe. Using the M/L V from §3.2.1, its mean effective surface density is Σ = 5.4 × 10 4 M pc −2 , a factor of 2.5-3 higher than for M32. The inner component of M60-UCD1, with r e ∼ 14 pc, has a mean Σ ∼ 9 × 10 4 M pc −2 , comparable to that of many nuclear star clusters. The central volume density of M60-UCD1 is not well-constrained by the present data.

The object most similar to M60-UCD1 is HUCD1, a Hydra Cluster UCD, which has r e = 25 pc and M V = −13.4 (Misgeld et al. 2011), though M59cO (Chilingarian & Mamon 2008) and several Coma Cluster UCDs (Chiboucas et al. 2011) are also similar, if less extreme. It seems likely that ongoing surveys for UCDs will turn up additional objects with properties comparable to M60-UCD1.

Table 1 Surface