DISCOVERY OF THE MOST ISOLATED GLOBULAR CLUSTER IN THE LOCAL UNIVERSE

Through a visual search we discovered two new globular clusters, JM81GC-1 and JM81GC-2 (called GC-1 and GC-2 hereafter), in the ACS/WFC and WFC3/UVIS fields, respectively. Figure 1(a) shows the location of these globular clusters. GC-1 is 1590 west of M82 and 2958 north of M81 (corresponding projected distances are 17.07 kpc and 31.77 kpc, respectively). GC-2 is 1334 west of M82 and 3725 north of M81 (corresponding projected distances are 14.32 kpc and 40.00 kpc, respectively). Their positions are R.A. (2000) = 09^h53^m2622, decl. (2000) = 69° 31' 175 for GC-1, and R.A. (2000) = 09^h53^m2017, decl. (2000) = 69° 39' 164 for GC-2.

Zoom In Zoom Out Reset image size

We checked the images of these clusters in the Sloan Digital Sky Survey (SDSS; York et al. 2000). GC-1 and GC-2 appear as extended sources in the SDSS images and were classified as galaxies (SDSS ID: J095326+693117 for GC-1 and J095320+693916 for GC-2). However, F814W images in Figures 1(b) and (c) show some resolved stars in the outer region of these clusters, proving that they are genuine star clusters. We checked the existence of any nearby dwarf galaxies around these two clusters using the data in Chiboucas et al. (2009) as well as the HST images, but found none.

We derived instrumental magnitudes of point sources in the images using the IRAF/DAOPHOT package that is designed for point-spread function (PSF) fitting photometry (Stetson 1994). We used 2σ as the detection threshold, and derived the PSFs using isolated bright stars in the images. We applied aperture correction derived from isolated bright stars. We calibrated the instrumental magnitudes to the Vega magnitude system using photometric zero points for ACS/WFC (http://www.stsci.edu/hst/acs/analysis/zeropoints/#tablestart) and WFC3/UVIS (http://www.stsci.edu/hst/wfc3/phot_zp_lbn). Then we converted this system to the Johnson–Cousins VI system using Sirianni et al. (2005).

Figures 2(a)−(d) display the color–magnitude diagrams for GC-1 and GC-2 as well as corresponding fields (called Field 1 and Field 2, respectively). Because the crowding is severe in the central region of the globular clusters, we plotted the stars at 08 < r < 60 from the cluster center. The aperture with a radius of 6'' covers about 96% of the total luminosity of each star cluster. In the case of fields, we plotted the stars at 15'' < r ≲ 100''. Both clusters show a relatively well defined red giant branch (RGB), indicating that they may be old globular clusters.

Zoom In Zoom Out Reset image size

We derived the distance to these clusters as well as to the corresponding halos using the TRGB method (Lee et al. 1993; Sakai et al. 1996; Méndez et al. 2002; McConnachie et al. 2004; Mouhcine et al. 2010; Conn et al. 2011). In Figure 2(e), the I-band luminosity function of the red giants in GC-1 and Field-1 shows a sudden jump at I ≈ 24.0, which corresponds to the TRGB. However, the TRGB magnitude for GC-2 is about 0.3 mag fainter than that for Field-2 in Figure 2(g), showing that GC-2 may be behind the halo stars.

Using the edge-detecting algorithm, we determined the TRGB magnitude more quantitatively. We calculated an edge-detection response function E(m) (=Φ(m + σ_m) − Φ(m − σ_m), where Φ(m) is the luminosity function of magnitude m and σ_m is the mean photometric error within a bin of ±0.05 mag about magnitude m), and we weighted it according to the Poisson noise of the luminosity function $E(m)\sqrt{\Phi (m) }$ (Méndez et al. 2002), as shown in Figures 2(f) and (h).

Thus derived TRGB magnitudes are I_TRGB = 24.039 ± 0.021 for GC-1, 24.069 ± 0.011 for Field-1, 24.295 ± 0.036 for GC-2, and 24.036 ± 0.022 for Field-2. The errors for the TRGB magnitudes were determined using a bootstrap resampling method with one million simulations. In each simulation, we randomly resampled the RGB sample with a replacement to make a new sample of the same size. We estimated the TRGB magnitude for each simulation using the same procedure and derived the standard deviation of the estimated TRGB magnitudes.

The mean color of the TRGB is derived from the colors of the bright red giants close to the TRGB and is corrected for the foreground reddening (E(B − V) = 0.09 for GC-1 and E(B − V) = 0.10 for GC-2 (Schlegel et al. 1998)): (V − I)_{0, TRGB} = 1.32 ± 0.02 for GC-1, 1.41 ± 0.03 for Field-1, 1.31 ± 0.04 for GC-2, and 1.49 ± 0.04 for Field-2. We derive the intrinsic I-band magnitude of the TRGB using the relation between the bolometric magnitude M_bol and the bolometric correction BC: $M_{\rm I_0}$ = $M_{\rm bol} - {\rm BC}_{\rm I_0}$. We calculate the bolometric magnitude using M_bol = −0.19[Fe/H] − 3.81 and the bolometric correction using ${\rm BC}_{\rm I_0} = 0.881\hbox{--}0.243(V -I)_{\rm 0,TRGB}$ (Da Costa & Armandroff 1990). [Fe/H] is derived from the mean color of the RGB stars.

The mean color of the RGB stars 0.5 mag fainter than the TRGB is derived from the colors of the bright red giants close to this magnitude: (V − I)_{0, −3.5} = 1.21  ±  0.02 for GC-1, 1.22  ±  0.03 for Field-1, 1.21  ±  0.03 for GC-2, and 1.31  ±  0.02 for Field-2. From these we derive [Fe/H]: [Fe/H] = −2.23  ±  0.11 for GC-1, −2.18  ±  0.13 for Field-1, −2.23  ±  0.12 for GC-2, and −1.84  ±  0.10 for Field-2. Using these we obtain M_{I, TRGB} = −3.95 for GC-1, −3.93 for Field-1, −3.95 for GC-2, and −3.98 for Field-2. Thus we finally calculate the distance modulus using (m − M)₀ = I_{0, TRGB} − M_{I, TRGB}. The distances derived are (m − M)₀ = 27.80  ±  0.03 (d = 3.63  ±  0.05 Mpc) for GC-1, 27.81  ±  0.03 (d = 3.65  ±  0.05 Mpc) for Field-1, 28.04  ±  0.04 (d = 4.05  ±  0.08 Mpc) for GC-2, and 27.81  ±  0.03 (d = 3.65  ±  0.05 Mpc) for Field-2. This value is similar to the previous estimates for M81, 3.63  ±  0.14 Mpc (Durrell et al. 2010; Gerke et al. 2011) and M82, 3.55  ±  0.11 Mpc (M. G. Lee & S. Lim 2012, in preparation). While GC-1 is at the same distance as M81, GC-2 is about 400 kpc behind the M81 halo along our line of sight.

GC-2 was previously observed and classified as a galaxy in the SDSS and its optical spectrum is available in the SDSS. The spectrum of GC-2 shows several absorption lines typical for globular clusters. We derived [Fe/H], [α/Fe], and the age for GC-2 from the comparison of Lick line index diagrams with the simple stellar population (SSP) models by Thomas et al. (2003).

Figure 3 displays Hβ versus [MgFe]' diagram, and 〈Fe〉 versus Mg₂ diagram for GC-2. The composite index [MgFe]' is a combination of magnesium and iron-sensitive indices, defined as [MgFe]$^\prime = \sqrt{ {\rm Mgb(0.72\cdot Fe5270+0.28\cdot Fe5335)} }$. It is an excellent metallicity indicator because it is completely independent of [α/Fe], and this behavior is almost independent of the adopted age or metallicity (Thomas et al. 2003). Hβ is an efficient indicator for age. We used the line index data for these clusters provided by the SDSS. The values derived following the technique described in Puzia et al. (2005) and Park et al. (2012) are [Fe/H] = –2.3  ±  0.12, [α/Fe] = –0.05  ±  0.40, and age = 14.9  ±  1.0 Gyr. This metallicity is consistent with the value derived from the color of the RGB. These show that GC-2 is indeed very metal-poor and old.

Zoom In Zoom Out Reset image size

We also derived age and mass from the spectral energy distribution fit using SDSS ugriz magnitudes in comparison with the SSP model given by Bruzual & Charlot (2003) (u = 20.50 ± 0.12, g = 19.13 ± 0.02, r = 18.50 ± 0.02, i = 18.21 ± 0.02, and z = 18.03 ± 0.05 for GC-1, and u = 19.51 ± 0.06, g = 18.25 ± 0.01, r = 17.60 ± 0.01, i = 17.31 ± 0.01, and z = 17.10 ± 0.03 for GC-2). Derived ages are log(age[y]) ∼10.2 for both clusters as long as we adopt Z = 0.0001, and derived cluster masses are log(M/M_☉) ∼ 6.40 for GC-1, and 6.85 for GC-2. This age for GC-2 is also consistent with the value derived from the spectrum analysis. Thus these clusters are very massive and old.

We derived the surface photometry of GC-1 and GC-2 using the IRAF/ELLIPSE task from the HST images. Figure 4 displays the radial profiles of the surface brightness for F814W images. The radial profiles of the inner region look similar to the King profiles. However, the outer parts in the radial profiles do not show any tidal cutoff, but continue to decrease smoothly. From the King model fits for the inner region at 00 < r < 40, we derived core radii (r_c), 00755 ± 00003 (1.329 ± 0.005 pc) for GC-1 and 01269 ± 00003 (2.491 ± 0.006 pc) for GC-2. Also we fit the data using the model with a power-law form in the outer region in Elson et al. (1987), ∑(r_p) = ∑₀(1 + r²_p/a²)^−γ/2, where the scale radius a is related to the King core radius as $r_c = a \sqrt{2^{2/\gamma } - 1 }$ for r_t ≫ r_c. There is a break at r = 280 (55 pc) for GC-2 so that we fit the data in two parts: a = 1.96, γ = 2.48 for GC-1, and a = 3.16, γ = 2.42, a = 3.27, γ = 3.62 for GC-2. The radial surface brightness profiles of GC-1 and GC-2 gradually decrease out to r = 15''. By integrating the radial surface brightness profiles to r = 15'', we derived integrated magnitudes of the star clusters, which we shall present in Section 4.1. From these we derived half-light radii (r_h): 0352 ± 0006 (6.13 ± 0.01 pc) for GC-1 and 0511 ± 0005 (9.81 ± 0.01 pc) for GC-2.

Zoom In Zoom Out Reset image size

Integrated magnitudes and colors of GC-1 and GC-2 are derived as, respectively, V₀ = 18.46  ±  0.01 and (V − I)₀ = 0.91  ±  0.01 and V₀ = 17.53  ±  0.01 and (V − I)₀ = 0.90  ±  0.01. Corresponding absolute magnitudes are M_V = −9.34  ±  0.01 for GC-1 and −10.51  ±  0.01 for GC-2, derived using our TRGB distances to the clusters. In Figure 5, we plot half-light radii (r_h) versus the absolute magnitude M_V for GC-1 and GC-2 in comparison with globular clusters in the MWG (Harris 1996), M81 (Nantais et al. 2011), and M82 (S. Lim et al. 2012, in preparation). We also plotted the data for globular clusters and extended star clusters in M31 (Huxor et al. 2009, 2011).

Zoom In Zoom Out Reset image size

van den Bergh & Mackey (2004) presented a boundary relation between normal globular clusters and extended globular clusters: log (r_h) = 0.2M_V + 2.6. GC-1 and GC-2 are found to be located above this boundary line, like the case of ω Cen and NGC 2419 in our Galaxy, and Mayall II in M31. Thus GC-1 and GC-2 are larger than typical globular clusters with similar luminosity, and they are much brighter than the extended star clusters in M31. The latter globular clusters (ω Cen, NGC 2419, and Mayall II) show some peculiar features so that they are often considered to be remnants of dwarf galaxies. Therefore, GC-1 and GC-2 also may be in the same vein.

At the moment the most isolated globular cluster in the Local Group is known to be MGC1, located at the projected distance from M31, 117 kpc. Considering it is ∼160 kpc closer than M31, its deprojected distance from M31 was derived to be 200  ±  20 kpc (Mackey et al. 2010). It is brighter (M_V = −9.2) than typical globular clusters, but fainter than GC-1 in M81. In the MWG, the most distant globular clusters and their galactocentric distances are NGC 2419 (91.5 kpc), Pal 3 (95.9 kpc), Eridanus (95.2 kpc), Pal 4 (111.8 kpc), and AM-1 (123.2 kpc) (Harris 1996).

On the other hand, the most distant satellite dwarf galaxies in the two main Local Group galaxies are Leo I in the MWG and And XXVIII at d > 350 kpc in M31 (Slater et al. 2011). Leo I is a dwarf spheroidal galaxy located at 270 kpc, and has long been considered the most distant satellite of the MWG (Lee et al. 1993). And XXVIII is a newly discovered dwarf galaxy about 100 kpc closer than M31. Its deprojected distance from M31 is estimated to be 365⁺¹⁷ _{− 1} kpc (Slater et al. 2011). Thus, the most remote globular clusters in our Galaxy and M31 are closer than the most distant satellite dwarf galaxies. This raises interesting questions: Can there be any globular clusters more distant than the most distant satellite galaxies in a galaxy? If so, what are they?

GC-2 lies 406  ±  97 kpc behind M81 along our line of sight, according to our TRGB distance estimates to GC-2 and Field-2. Considering this and the projected distance in the sky we derive a three-dimensional distance from M81 of 408  ±  97 kpc. Although it is closer to M82 than M81 in the sky, its three dimension distance is significantly larger than the projected distance from M82 (also from M81). Also M81 is twice as massive as M82 and is the most massive member in the M81 Group. These indicate that the main host for GC-2 may be, if any, M81 rather than M82. Therefore GC-2 is the most isolated globular cluster among the known globular clusters in the local universe. In addition, it is more distant than any known satellite dwarf galaxies around M81. The radial velocity for GC-2 given in the SDSS is 159  ±  4 km s⁻¹, which is much larger than the value for M81 (−35  ±  4 km s⁻¹) (Chynoweth et al. 2008). Thus GC-2 is receding from M81 with a velocity ≈200 km s⁻¹. If it were moving at this velocity along the radial orbit, it would have taken about 2 Gyr to reach the current position from the center of M81. Thus GC-2 can be considered an intragroup globular cluster wandering in the M81 Group.

How can GC-2 be that far from M81, receding with a velocity of ≈200 km s⁻¹? The origin of GC-2 is not clear. Possible scenarios are as follows. First, it might have been ejected during the interaction of three main galaxies about 2 Gyr ago. However, there is no evidence supporting this at the moment. Second, it may be one of the primordial globular clusters that formed early in isolation. If so, where are the others? Third, it may be a remnant of a dwarf galaxy that accreted to M81 and is receding now. If so, how could it survive during the perigalactic passage around M81? These possibilities need to be investigated with future observations or simulations.