THE MOST MASSIVE ULTRA-COMPACT DWARF GALAXY IN THE VIRGO CLUSTER

This study is part of a systematic search for UCDs across the entire Virgo Cluster using the homogeneous imaging of the Next Generation Virgo Cluster Survey (NGVS; Ferrarese et al. 2012). The NGVS imaged the Virgo cluster out to the virial radii of its two main subclusters, for a total sky coverage of 104 deg², in four optical bands (${u}^{*}g^{\prime} i^{\prime} z^{\prime}$). The NGVS produced two sets of stacked images for each $1^\circ \times 1^\circ$ field: a "long exposure" stack consisting of images with exposure times from 2055 to 6400 s (depending on the band), and a "'short exposure" stack, consisting of images with exposure times from 40 to 250 s. The long exposure images saturate at $g^{\prime} \sim 18.5$ mag for point sources, so we use the short exposure images to study compact galaxies and galaxy nuclei. The methods used to generate catalogs and aperture photometry are similar to those described in Durrell et al. (2014) and Liu et al. (2015).

We selected UCD candidates based on their colors in the (${u}^{*}-{g}^{\prime }$)–(${g}^{\prime }-{z}^{\prime }$) diagram, then fit them with King (1966) models convolved with the point-spread function (PSF) using the KINGPHOT software package (Jordán et al. 2005). The King model fits allow us to determine which objects are extended in appearance.

We identified NGVS J124211.05+113841.24 (M59-UCD3) as a bright, extended object $130^{\prime\prime}$ to the east of the massive early-type galaxy M59 (9.4 kpc at the distance of M59). There is no HST imaging of M59-UCD3, so our photometric analysis is based solely on the NGVS images, which had image quality of $\mathrm{FWHM}(g^{\prime} )=0\buildrel{\prime\prime}\over{.} 58$ and $\mathrm{FWHM}(i^{\prime} )=0\buildrel{\prime\prime}\over{.} 46$ (R_e of 12.0 and 9.5 pc, respectively, at M59). Figure 1 shows M59-UCD3 relative to the massive early-type galaxies, M60 (left) and M59 (right), and the other bright UCDs, M60-UCD1 (S13) and M59cO (CM08).

Zoom In Zoom Out Reset image size

The radial velocity of M59-UCD3 was measured as 440 $\mathrm{km}\;{{\rm{s}}}^{-1}$ in a 320 s exposure obtained on 2014 November 29 with the Supernovae Integral Field Spectrograph (SNIFS) on the UH 2.2 m telescope (Lantz et al. 2004). This confirmed that M59-UCD3 is both in the Virgo Cluster, and likely to be associated with M59 (${V}_{\mathrm{hel}}=467\pm 5$ $\mathrm{km}\;{{\rm{s}}}^{-1}$, Cappellari et al. 2011). The distance to M59 is 14.9 ± 0.5 Mpc (Mei et al. 2007; Blakeslee et al. 2009), and for the rest of the paper, we assume that M59-UCD3 is at the same distance.

To confirm the SNIFS radial velocity and obtain a measure of the internal velocity dispersion of M59-UCD3, on 2015 April 14, we obtained a 300 s exposure with the Keck/DEIMOS spectrograph (Faber et al. 2003) using the 600-line grating and the $0\buildrel{\prime\prime}\over{.} 7$-wide long slit, giving a spectral range of $\lambda \lambda$ 4800–9500 $\mathring{\rm A} ,$ a spectral resolution of $2.8\;\mathring{\rm A}$ (corresponding to $\sigma =50$ $\mathrm{km}\;{{\rm{s}}}^{-1}$), and peak signal-to-noise ratio (S/N) of 107 ${\mathring{\rm A} }^{-1}.$ These data were reduced using the Simon & Geha (2007) modification of the DEIMOS spec2d pipeline (Cooper et al. 2012; Newman et al. 2013), and the spectrum was optimally extracted by fitting a Gaussian function ($\mathrm{FWHM}=10$ pixels or $1\buildrel{\prime\prime}\over{.} 85$) to the spatial intensity profile.

To provide an internally consistent comparison of the photometric properties of M59-UCD3 with M60-UCD1 and M59cO, we used NGVS data for all three objects in the same way (Table 1). In the g'-band, we find that M59-UCD3 is 0.75 mag brighter than M60-UCD1, and 1.19 mag brighter than M59cO. M59-UCD3 has the highest effective surface brightness of the three UCDs, with $\langle {\mu }_{g^{\prime} }{\rangle }_{{\rm{e}}}=16.42$ mag arcsec⁻². After correcting for foreground extinction (Schlafly & Finkbeiner 2011), the total luminosity of M59-UCD3 is ${M}_{g^{\prime} }=-14.2$ mag, or ${L}_{g^{\prime} }=5.3\times {10}^{7}{L}_{\odot }.$

Table 1.  Photometric Properties of Luminous Virgo Cluster UCDs

Name	R.A. (J2000)	Decl. (J2000)	${g}^{\prime }$	${u}^{*}-{g}^{\prime }$	${g}^{\prime }-{i}^{\prime }$	${i}^{\prime }-{z}^{\prime }$	${R}_{{\rm{e}},\mathrm{NGVS}}$	${R}_{{\rm{e}},{HST}}$	$\langle {\mu }_{{g}^{\prime }}{\rangle }_{{\rm{e}}}$
			(mag)	(mag)	(mag)	(mag)	(pc)	(pc)	(mag arcsec−2)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
M59-UCD3a	190.5460471	11.6447937	16.74	1.63	1.10	0.28	25 ± 2	⋯	16.42
M60-UCD1	190.8998699	11.5346389	17.49	1.77	1.13	0.27	36 ± 2	27	17.72
M59cO	190.4805689	11.6677212	17.93	1.70	1.11	0.27	37 ± 2	35	18.50

Notes. (4) Total ${g}^{\prime }$ magnitude from a single-Sérsic fit to M59-UCD3 and M59cO, and a double-Sérsic fit to M60-UCD1. (5)–(7) Colors measured within a $3^{\prime\prime}$ diameter aperture. (8) Effective radius and uncertainty measured from profile fits in the NGVS image, in parsecs, assuming distances of 14.9 ± 0.5 Mpc for M59 UCDs and 16.5 ± 0.6 Mpc for M60 (Blakeslee et al. 2009). Size uncertainties include the fitting uncertainty and the distance uncertainty added in quadrature, but not systematic errors (PSF, galaxy subtraction) that could be at the level of 20%. For M60-UCD1, R_e contains half the total light from a double-Sérsic fit. (9) Effective radius measured using HST imaging, from Norris et al. (2014) (10) Mean g'-band surface brightness within R_e, assuming the effective radius in Column (8).

^aSandoval et al. (2015) reported ${g}_{\mathrm{SDSS}}=16.81$ mag, ${(g-i)}_{\mathrm{SDSS}}=1.20$ mag and ${R}_{{\rm{e}},{Subaru}}=20\pm 4$ pc, all of which are consistent with our measurements.

Download table as:  ASCIITypeset image

We fit the g'-band surface brightness profile of M59-UCD3 with a PSF-convolved Sérsic profile (Figure 2). Beyond $R\approx 1\buildrel{\prime\prime}\over{.} 2,$ the data show a slight excess above a Sérsic model. The uncertainty in subtracting the light from M59 makes it difficult to ascertain whether M59-UCD3 is better fit by a double-Sérsic, so we only report results for a single-Sérsic fit. Unlike for M60-UCD1, M59-UCD3's ellipticity does not vary much with radius. The object is fairly round, except at the very center where systematic PSF uncertainties dominate. M59-UCD3 is best fit by a Sérsic profile with effective radius, ${R}_{{\rm{e}}}=0\buildrel{\prime\prime}\over{.} 345\pm 0\buildrel{\prime\prime}\over{.} 03$ (25 ± 2 pc), Sérsic index, $n=2.5,$ and surface brightness at R_e, ${\mu }_{{\rm{e}}}(g^{\prime} )=17.57$ mag arcsec⁻².

Zoom In Zoom Out Reset image size

Table 1 includes R_e from HST imaging for M60-UCD1 and M59cO from (Norris et al. 2014, hereafter N14). The size we measure for M60-UCD1 (36 ± 2 pc) is larger than that measured in HST/ACS images (27 pc), although our measurement for M59cO is consistent with (N14). Our R_e estimates may be biased toward larger sizes, given the larger PSF in the NGVS data (see, e.g., Puzia et al. 2014), or R_e may be affected by the details of subtracting light from M60. This is why we perform a relative comparison of all three UCDs using the same images and same fitting software, as relative measurements are more reliable. Even if we adopt the smaller size value for M60-UCD1, it would still have a lower effective surface brightness than M59-UCD3.

Using our Keck/DEIMOS spectrum, we measured the line of sight velocity and velocity dispersion with the penalized pixel fitting (pPXF) software of Cappellari & Emsellem (2004), which fits a broadened combination of stellar templates with different weights. We used 31 stellar templates with spectral types from B to M, observed previously with DEIMOS using the same grating. Because sources can be mis-centered on the slit, we applied a velocity correction using the atmospheric A and B absorption bands, whose centers depend on how the source illuminates the slit. The heliocentric radial velocity of M59-UCD3 is ${V}_{r}=447\pm 3$ $\mathrm{km}\;{{\rm{s}}}^{-1}$, and the integrated velocity dispersion is $\sigma =77.8\pm 1.6$ $\mathrm{km}\;{{\rm{s}}}^{-1}$. As a consistency check, we also measured the velocity dispersion of M59cO, which we observed simultaneously with M59-UCD3 through the same slit. Our measurement of $\sigma =28.1\pm 4.2$ $\mathrm{km}\;{{\rm{s}}}^{-1}$ is consistent with that of (N14), who reported a value of $\sigma =29.0\pm 2.5$ $\mathrm{km}\;{{\rm{s}}}^{-1}$.

We estimate the mass of M59-UCD3 using a spherical isotropic Jeans equation analysis. We deproject the best-fit Sérsic profile, assume mass follows light, derive the projected dispersion profile, convolving both the model dispersion and luminosity profiles with a Moffat PSF (FWHM $=\;1\buildrel{\prime\prime}\over{.} 2$). We then apply a $0\buildrel{\prime\prime}\over{.} 7$-wide slit to our model to simulate the observed integrated velocity dispersion. Matching our observed dispersion gives a total dynamical mass of $(3.7\pm 0.4)\times {10}^{8}{M}_{\odot },$ with a mass-to-light ratio (M/L) of ${M}_{\mathrm{dyn}}/{L}_{{g}^{\prime }}=7.1\pm 0.7,$ or ${M}_{\mathrm{dyn}}/{L}_{V}=4.9\pm 0.5$ assuming $g^{\prime} -V=0.40$ mag (using the transformation in Peng et al. 2006). Our model gives an average velocity dispersion within R_e of ${\sigma }_{{\rm{e}}}=94.1\pm 1.9$ $\mathrm{km}\;{{\rm{s}}}^{-1}$.

We compared the absorption line indices of M59-UCD3 with single stellar population (SSP) models to derive age, metallicity, [α/Fe], and a stellar $M/L.$ We used a Kroupa (2001) initial mass function (IMF) and the LIS-5 Å system (Vazdekis et al. 2010) where we put the models and the target spectrum at a constant resolution of 5 Å. By comparing the Hβ and ${[\mathrm{MgFe}]}^{\prime }$ indices with the SSP models of Vazdekis et al. (2015) we obtain age and metallicity estimates of 14 Gyr and [Fe/H] = −0.01 ± 0.21. The age is fixed to the oldest model available because Hβ is outside the model grid. This makes M59-UCD3 comparable in age and metallicity to the oldest and most metal-rich dE nuclei, and more metal-rich than typical dEs (e.g., Michielsen et al. 2008; Toloba et al. 2014). We compare the Mgb and $\langle \mathrm{Fe}\rangle$ indices to 14 Gyr SSP models to obtain [α/Fe] = +0.21 ± 0.07. For these SSP parameter values, the models provide a stellar ${M}_{\star }/{L}_{V}=3.4\pm 0.9$ for a Kroupa IMF or 5.7 ± 1.5 for a Salpeter IMF.

M59-UCD3 has a dynamical ${M}_{\mathrm{dyn}}/L$ between the ${M}_{\star }/L$ values derived from SSP models using Kroupa and Salpeter IMFs, and could be consistent with either, within the uncertainties.

Sandoval et al. (2015) reported the discovery of M59-UCD3 using SDSS and Subaru imaging, and spectroscopy from SOAR/Goodman. They reported $g,r,i$ photometry, a size of ${R}_{{\rm{e}}}=20\pm 4$ pc, a radial velocity ${V}_{r}=373\pm 18$ $\mathrm{km}\;{{\rm{s}}}^{-1}$, and SSP age and abundances. The photometry between our two studies for common filters (g and i) is consistent, as are the sizes, and the derived values of [Fe/H] and [α/Fe]. Our age estimate (14 Gyr) is larger than theirs (8.6 ± 2.2 Gyr), but the model grids at old ages are closely spaced and this discrepancy is not alarming, although merits further investigation. Our radial velocity is 74 $\mathrm{km}\;{{\rm{s}}}^{-1}$ higher, a significant difference that is difficult to explain given the high S/N of both of our observations. Sandoval et al. (2015) do not report a velocity dispersion.

M59-UCD3 is the most extreme object of its type discovered to date, although it is basically similar to M60-UCD1. Figure 3 shows the location of dynamically hot stellar systems, mostly in the Virgo Cluster, in the parameter space of size (R_e), effective B-band surface brightness ($\langle {\mu }_{B}{\rangle }_{{\rm{e}}}$), and B-band luminosity (M_B). The three luminous UCDs in M59 and M60 are more luminous than other UCDs in Virgo, but they are significantly more compact than objects categorized as "compact ellipticals" (cEs), which typically have ${R}_{{\rm{e}}}\gtrsim 100$ pc. Lines of constant $\langle {\mu }_{B}{\rangle }_{{\rm{e}}}$ show that M59-UCD3 is one of the densest galaxies known. These luminous UCDs appear to bridge the parameter space between massive star clusters and compact galaxies, further blurring the boundary between the two classes. The only objects in Virgo that approach or exceed the effective surface density of M59-UCD3 are nuclear star clusters (which may have brighter B-band fluxes due to rejuvenated stellar populations) and some of the most massive GCs. The total effective surface mass density (${{\rm{\Sigma }}}_{\mathrm{eff}},$ mass density within 1R_e) is $9.4\times {10}^{10}{M}_{\odot }\;{\mathrm{kpc}}^{-2},$ which is roughly the maximum ${{\rm{\Sigma }}}_{\mathrm{eff}}$ observed anywhere in the universe. As discussed in Hopkins et al. (2010), finding these extremely dense systems can illuminate the physical feedback processes that limit the density of stellar systems. Finding more UCDs in other environments (e.g., N14) will help disentangle the relationships between these families of objects.

Zoom In Zoom Out Reset image size

If UCDs are the compact remnants of relatively recent tidal stripping events, we may find stripped stellar material in their vicinities. There has been little evidence of tidal debris linked to UCDs (with a few exceptions, cf. Foster et al. 2014; Mihos 2015), but simulations show that the window of time during which tidally stripped starlight would be visible with current facilities is relatively short (Rudick et al. 2009). Given the high mass and metallicity of M59-UCD3, however, any stripped progenitor could be very massive, leaving a significant amount of debris. The average present-day stellar mass fraction of nuclear star clusters in early-type galaxies is of order 4 × 10⁻³ (e.g., Turner et al. 2012). If this was true for a system that once hosted M59-UCD3, then the progenitor galaxy would have had a stellar mass of $\sim {10}^{10}{M}_{\odot },$ or about 15% the mass of M59. This logic, however, is uncertain in a few ways: (1) the $1\sigma$ scatter in nuclear mass fraction is a factor ∼10, (2) this fraction is largely defined by lower mass nuclei, and (3) present-day UCDs may not be equivalent to today's nuclear star clusters, but may instead combine a nuclear component with a stripped bulge-like component.

We searched for fine structure and tidal streams around M59 that could be associated with either M59-UCD3, or M59cO. We used the long exposure NGVS g'-band image reduced with the Elixir-LSB pipeline (Ferrarese et al. 2012), and subtracted an elliptical isophotal model of M59. Figure 4 shows the model-subtracted image. The cross-shaped residual is typical in galaxies that show significant diskiness, such as M59. What we find notable is the broad linear feature to the south of M59, which extends for at least $\sim 200^{\prime\prime} ,$ or 14 kpc at the distance of M59. This "plume" has $\langle {\mu }_{{g}^{\prime }}\rangle \approx 28.0$ mag arcsec⁻², which is 1 mag above the detection threshold for NGVS images (Ferrarese et al. 2012, see Figure 18). If extended toward the inner regions of M59, the plume roughly lines up with M59-UCD3. Although it is possibile the plume is associated with the nearby irregular galaxy IC 3665 (VCC 1890) to the southwest, its existence in the vicinity of M59 and M59-UCD3 suggests a common origin with the larger galaxy.

Zoom In Zoom Out Reset image size

We mask bright sources and median bin the image by 9 × 9 pixels (following Mihos et al. 2005), and measure photometry of the "plume" along the length of the line shown in Figure 4, avoiding the residuals in the inner region, and using sky regions alongside the feature. The total magnitude is $g^{\prime} =16.8\pm 0.4$ mag, which at M59's distance corresponds to ${M}_{g^{\prime} }=-14.2\pm 0.4$ mag (${L}_{g^{\prime} }=5.3\times {10}^{7}{L}_{\odot }$), and is comparable to M59-UCD3 itself. The quoted uncertainty is for random errors, but we found that selecting different sky regions around the plume, or using different isophotal model subtractions of M59, can change the measured brightness by a few tenths of a magnitude. A similar measurement in the i'-band yields $i^{\prime} =15.7\pm 0.2$ mag and $g^{\prime} -i^{\prime} =1.1\pm 0.4$ mag, which is consistent with the colors of M59-UCD3 and M59, within the uncertainties.

If associated with M59-UCD3, the luminosity of this plume would be a lower-limit on the amount of material stripped from M59-UCD3's progenitor. Although significant mass could be hidden at lower surface brightnesses, hiding 1–2 orders of magnitude more mass would require it to be spread over 10–100 times more area. It seems more likely that either M59-UCD3 was a high fraction of its progenitor's total stellar mass, the stripping happened a long time ago, or M59-UCD3 was originally formed compact and dense. Measuring the mass of any central black hole in M59-UCD3 could help differentiate these scenarios.