1 INTRODUCTION
Globular clusters (GCs) have long been used as luminous tracers of the star-formation histories, chemical enrichment, and gravitational potential of their host galaxies, especially in the stellar halos where the density of stars is low (Brodie & Strader Reference Brodie and Strader2006). GCs in dwarf galaxies are of particular interest; for example, in recent years, they have been used to provide evidence that some dark matter halos have central cores rather than cusps (e.g., Goerdt et al. Reference Goerdt, Moore, Read, Stadel and Zemp2006) and to constrain the extreme mass loss from GCs predicted in some models of the formation of multiple stellar populations (Larsen, Strader, & Brodie Reference Larsen, Strader and Brodie2012). The presence of GCs also hints at active early star formation in dwarfs, as measurements in the local universe suggest that the fraction of stars formed in bound clusters and the maximum mass of star clusters correlates strongly with the star formation rate density (e.g., Goddard, Bastian, & Kennicutt Reference Goddard, Bastian and Kennicutt2010; Cook et al. Reference Cook2012; Kruijssen Reference Kruijssen2012; Johnson et al. Reference Johnson2017).
Most dwarfs in the Local Group do not host GCs, especially among the less massive galaxies. The least luminous Local Group galaxy known to have a GC is the recently discovered dwarf Eridanus II. This galaxy has M V = −7.1 and hosts a GC with M V = −3.5, which by itself makes up 4% of the luminosity of its host (Koposov et al. Reference Koposov, Belokurov, Torrealba and Evans2015; Crnojević et al. Reference Crnojević2016a). Contenta et al. (Reference Contenta2017) and Amorisco (Reference Amorisco2017) argue that the survival of this GC against dynamical friction, despite its projected location close to the galaxy centre (offset by only ~45 pc), is evidence for a cored dark matter halo in Eridanus II: If a standard cuspy halo were present, the GC would have been very unlikely to have survived to the present day. Amorisco (Reference Amorisco2017) makes a similar argument for the M31 satellite Andromeda XXV, which hosts a faint extended star cluster near its centre, though the precise offset is not well-determined as the galaxy centre itself is poorly constrained (Cusano et al. Reference Cusano2016). These conclusions are important as it is unclear whether galaxies with such low luminosities can effectively transform cuspy dark matter profiles to cored ones through central bursts of star formation (e.g., El-Badry et al. Reference El-Badry2016; Read, Agertz, & Collins Reference Read, Agertz and Collins2016), and the presence of cores could provide evidence for self-interacting dark matter rather than standard cold dark matter (e.g., Robles et al. Reference Robles2017).
Here we present a discussion of another faint Local Group galaxy with a GC: the M31 satellite Andromeda I (And I). The existence of this GC, which we term And I–GC1, was briefly noted in a conference proceeding many years ago (Grebel, Dolphin, & Guhathakurta 2000) but the basic data of the cluster have not been discussed in the subsequent literature, and it has not been included in lists of Local Group dwarfs that host GCs (e.g., Mackey Reference Mackey and Freeman2015).
2 BACKGROUND, DATA REDUCTION, AND ANALYSIS
Grebel et al. (Reference Grebel, Dolphin and Guhathakurta2000) noted the presence of a faint GC in the archival Hubble Space Telescope (HST)/WFPC2 imaging of And I published in da Costa et al. (Reference Da Costa, Armandroff, Caldwell and Seitzer1996). This GC is located at a J2000 degree position (R.A., Dec.) = (11.42881 +38.03162), calculated from our own images discussed below. This position is located ~207 pc from the centre of And I (we use a distance of 745 kpc; McConnachie Reference McConnachie2012). As the half-light radius of And I is 672 ± 69 pc (McConnachie Reference McConnachie2012), the GC sits well within the central regions of the galaxy. New, deep HST/ACS observations of And I have been obtained as part of programme GO-13739 (Skillman et al. Reference Skillman2017), which fortunately include the GC (see Figure 1). These data comprise a total exposure time of approximately 29 ks in F475W and 23 ks in F814W. We retrieved the individual CTE-corrected exposures from the HST archive then processed the data using DrizzlePac. The relative exposure offsets were determined using TweakReg and then drizzled together with AstroDrizzle to a pixel scale of 0.03 arcsec.
A log-scaled HST/ACS F814W image of And I–GC1, 10 arcsec (38 pc) on a side. An insert shows the entire And I galaxy, from a ground-based image. This has a width of 7.2 arcmin (1.6 kpc). The location of the GC within the dSph is indicated by the white spot.
Using an isophote fitting programme, we determined integrated magnitudes of the GC out to a radius of 3 arcsec (100 pixels or 11 pc), the radius at which light from the GC could not be distinguished from the background. As the cluster is faint and there are a few likely contaminating stars in the field, we masked pixels 4σ above the local isophote. The integrated Vega magnitudes of the cluster are F475W = 21.7 ± 0.1 and F814W = 20.3 ± 0.1. These photometric uncertainties represent only the shot noise in the masked profile. If the bright stars are not masked, the cluster is brighter by about 0.5 mag in F814W and 0.3 mag in F475W, which should be taken as a qualitative estimate of the systematic uncertainties in the integrated photometry. We note that even in the case of perfect measurements, there are substantial stochastic fluctuations in the absolute magnitudes and colours of low-mass clusters due to sparse sampling of the stellar mass function and the short lifetimes of bright post-main sequence stars (e.g., Fouesneau et al. Reference Fouesneau2014).
To reference these measurements to the commonly used V band, we used Padova single stellar population models (Bressan et al. Reference Bressan, Marigo, Girardi, Salasnich, Dal Cero, Rubele and Nanni2012) for a 13 Gyr population to convert F475W to V, finding V ~ 21.1, equivalent to M V ~ −3.4 (–3.7 if the bright stars are not masked). The luminosity of And I GC would rank among the lowest of confirmed M31 GCs, though most GC searches have not been complete to this depth (e.g., Huxor et al. Reference Huxor2014).
To empirically derive a metallicity for the cluster, we corrected the magnitudes for foreground reddening using the maps of Schlafly & Finkbeiner (Reference Schlafly and Finkbeiner2011) and then converted them to AB g and i, again using the Padova models. These values are i 0 = 20.7 ± 0.1 and (g − i)0 = 0.70 ± 0.11. As a check on this value, we also calculated integrated colours within the half-light radius (see below), for which we might expect the relative effect of contaminating stars to be lower. This value is (g − i)0 = 0.73, well within the uncertainty on the total colour. Peacock et al. (Reference Peacock, Zepf, Maccarone and Kundu2011) publish observed (g − r) and (r − i) colour vs. [Fe/H] plots for M31 GCs; We combine these data to estimate a rough conversion between (g − i) colour and metallicity. Using the value (g − i)0 = 0.70, we find an estimated photometric metallicity of [Fe/H] = –1.5 ± 0.4, suggesting the GC is metal poor, but not remarkably so: its metallicity is typical of the typical of the metal-poor GC populations in the Milky Way and M31. This metallicity is also consistent with the overall low metallicity of And I itself ([Fe/H] ~− 1.5; Kalirai et al. Reference Kalirai2010).
We also estimated a rough half-light radius via integrated light, finding a value of 4.2 ± 0.4 pc (the individual F475W and F814W measurements were 4.3 and 4.1 pc, respectively). This implies a half-mass relaxation time of ~360 Myr assuming a mean stellar mass of 0.6 M⊙ (Koposov et al. Reference Koposov2007). The evaporation time will be approximately 10 times longer than the relaxation time, though the actual mass loss rate depends on the current stellar mass function (including remnants), the true galactocentric distance, and the mass of the dark halo (Gieles, Heggie, & Zhao Reference Gieles, Heggie and Zhao2011). In any case, it is clear that the least massive GCs in dwarfs are susceptible to evaporation on timescales of Gyrs, suggesting that clusters like the one in And I are likely to be dissolving.
It is worth considering whether this GC could be an M31 object projected onto And I, as the dwarf sits at a projected radius of 45 kpc from M31. The surface density profile of M31 GCs from Huxor et al. (Reference Huxor2011) predicts 0.004 GCs kpc−2 at the projected distance of And I. Since the area within the half-light radius itself is ~1.4 kpc2, the expected number of contaminant GCs from M31 is about 0.006. Even if we conservatively increase this estimate by about 50% to account for incompleteness of the GC search for objects with M V > −6, the predicted number of M31 GCs within the half-light radius of And I is still <0.01. Given that the GC is much closer to the galaxy centre than the half-light radius, this calculation shows the GC is very unlikely to be an M31 interloper, though of course it would be desirable to obtain a radial velocity.
3 DISCUSSION
As noted in the Introduction, GCs are rare among Local Group galaxies with low stellar masses, and And I (with M V = –11.7; McConnachie Reference McConnachie2012) is among the most extreme examples—only Eridanus II (M V = −7.1) and Andromeda XXV (M V = −9.7) have lower luminosities and still host a GC. Here we place this galaxy and star cluster in the larger context of low-mass galaxies with GCs.
The best-studied GC system of a low-mass galaxy in the Local Group is that of the Fornax dwarf spheroidal (M V = –13.4; McConnachie Reference McConnachie2012), which has five clusters with absolute magnitudes ranging from M V = –5.1 to –8.1 (Strader et al. Reference Strader, Brodie, Forbes, Beasley and Huchra2003). One of these GCs, with M V = –7.3, is more metal rich than the others and is located near the centre of the galaxy, and hence is sometimes cited as a candidate nuclear star cluster, though its radial velocity appears offset from the field stars in the galaxy centre (Hendricks et al. Reference Hendricks, Boeche, Johnson, Frank, Koch, Mateo and Bailey2016). The dynamical friction timescale for the Fornax GCs is less than a Hubble time, so the survival of its GC system is a puzzle; one proposed solution is that the dark matter distribution in the Fornax dSph is cored rather than cusped, leading to a longer inspiral time (Goerdt et al. Reference Goerdt, Moore, Read, Stadel and Zemp2006).
Considering other M31 satellite galaxies: besides the faint And I and Andromeda XXV GCs, more luminous clusters were proposed to be affiliated with And XVII (M V = −8.7) by Irwin et al. (Reference Irwin, Ferguson, Huxor, Tanvir, Ibata and Lewis2008), but kinematic studies have shown these GCs are unlikely to be bound to the dwarf (Veljanoski et al. Reference Veljanoski2014).
The Local Group dwarf irregular WLM hosts one massive GC (Stephens, Catelan, & Contreras Reference Stephens, Catelan and Contreras2006), while WLM is more luminous than the Fornax dwarf spheroidal, its stellar mass is lower, such that this GC makes up at least 5% of the total mass of the galaxy, and an astonishing ~25% of the stars with [Fe/H] <− 2 (Larsen et al. Reference Larsen, Brodie, Forbes and Strader2014).
Beyond the Local Group, a number of galaxies in the luminosity range −10 ≲ M V ≲ −11.5 likely host at least one GC (e.g., Georgiev et al. Reference Georgiev, Puzia, Goudfrooij and Hilker2010; Da Costa et al. Reference Da Costa, Grebel, Jerjen, Rejkuba and Sharina2009). While a firm association between candidate GCs and host galaxies has not been confirmed via spectroscopy for most of these objects, in many cases, the GCs are definitively identified via HST imaging, and it is reasonable to assume that most of these proposed associations are correct. Noting a few of the extremes among this group, the galaxy KK27 (M V = −10.1) with two candidate GCs is currently the lowest luminosity galaxy known outside the Local Group with GCs (Georgiev et al. Reference Georgiev, Puzia, Goudfrooij and Hilker2010). The IKN galaxy, in the M81 group, is slightly more luminous (M V = −11.5) but has five GCs (the most luminous with M V = −8.5), which together make up at least 10% of the total stellar mass of this galaxy (Tudorica, Georgiev, & Chies-Santos Reference Tudorica, Georgiev and Chies-Santos2015; Larsen et al. Reference Larsen, Brodie, Forbes and Strader2014).
Overall, we conclude that while uncommon, GC systems in dwarf galaxies fainter than M V ~ −12 are not exceptional. Deeper observations of Local Group dwarfs, and the ongoing discovery of low-mass galaxies beyond the Local Group (e.g., Crnojević et al. Reference Crnojević2016b), are likely to reveal less massive galaxies with GCs, and the properties of these GCs may offer compelling constraints on the structure of the dark matter halos in these galaxies (Contenta et al. Reference Contenta2017; Amarisco 2017).
ACKNOWLEDGEMENTS
JS acknowledges support from NSF grant AST-1514763. DJS acknowledges support from NSF grant AST-1412504. This work was partially performed at the Aspen Center for Physics, which is supported by National Science Foundation grant PHY-1066293.
