LAMOST 1: A DISRUPTED SATELLITE IN THE CONSTELLATION DRACO

With the rise of massive spectroscopic surveys such as The Sloan Digital Sky Survey (York et al. 2000), the RAdial Velocity Experiment (Steinmetz et al. 2006), the Large Sky Area Multi-object Fibre Spectroscopic Telescope survey (LAMOST; Cui et al. 2012) and others, new phase space information is becoming available on massive scales.

However, while spectroscopic surveys provide additional tools such as radial velocity profiles, chemical compositions, and surface gravities, new hindrances such as incomplete photometry and poorly defined parameter-space coverage need to be accounted for. Structure searches traditionally rely on basic star counting techniques such as: statistical photometric density contrasting (see for example Newberg et al. 2002; Majewski et al. 2003; Belokurov et al. 2006) and photometric matched filter analyses (Odenkirchen et al. 2001). But since spectroscopic data are inherently incomplete, new methods must be used to uncover structure, usually by looking for velocity clumping; see, for example Schlaufman et al. (2009) and Starkenburg et al. (2009).

We trawl the LAMOST spectroscopic data for distinct clumpings of stars in both velocity and metallicity spaces which are incompatible with the background distributions. The technique uncovers numerous known objects and structures as well as one strong detection of an unknown object which is the subject of this letter.

We postulate that this object is a star cluster being observed near the apocenter of its eccentric orbit. It manifests as a signature distinct from the background in velocity and metallicity space. We conclude that the object is likely to be a nearby, intermediate age globular cluster which is being disrupted by the galactic potential.

In Section 2 we describe the data used in this search; in Section 3 we describe the search algorithm, rejected candidates and the Draco clump as well as its significance; in Section 4 we calculate the age and distance to the clump via isochrone fitting, the object's size, mass and luminosity, and its orbital parameters as well; we discuss and conclude in Section 5.

For our data, we use the ∼2.5 million stellar spectra available in the second data release of the LAMOST spectroscopic survey. The LAMOST telescope is 4 m Schmidt reflector located at Xinglong Station, Hebei, China. The spectroscopic survey is performed by a novel system of 4000 individually positionable fibers which collect moderate resolution spectra (R ∼ 1800) in the optical range from 370 to 900 nm. For information on the telescope and survey, see Luo et al. (2015).

Since the LAMOST survey includes multiple observations for many objects, these data are duplicate cleaned with a pairing radius of 7'' where the observations with the highest signal-to-noise ratio are preferred.

In Sections 3.1 and 4.2 we will briefly investigate the kinematic properties of the stars using proper motions from the fourth edition of the United States Naval Observatory CCD Astrograph Catalog (UCAC4; Zacharias et al. 2013).

We search for substructure in the LAMOST data set with one main constraint: the data are photometrically incomplete. With this in mind, we perform a gridwise search through the LAMOST spectroscopic data set searching for metallicity–velocity clumping that is inconsistent with the field.

The search analyzes 1° radius pencil beams in 05 steps over the entirety of the LAMOST data set. The radial velocities of the stars in the pencil beam are then fit to a mixture of Gaussians with a varying number of component Gaussians (from one Gaussian to seven). If any component Gaussian has a velocity dispersion of less than 20 km s⁻¹ and has a minimum population of three objects (to avoid delta spikes), we check if that component is also two standard deviations distinct from the total field.

This search yields a large amount of stream-like candidates (>1000). We further reduce the candidate sample by investigating only those whose velocity outliers are also quite compact in [Fe/H] space, having metallicity dispersions of less than 0.1 dex. This reduces our sample to a more manageable 21 candidates.

We next compare the sky-coordinate positioning of these candidates to: (1) the globular cluster catalog of Harris (1996), 2010 revision³ ; (2) the nearby galaxy catalog of Karachentsev et al. (2013); (3) the composite open cluster catalog of Kharchenko et al. (2012), Kharchenko et al. (2013), and Schmeja et al. (2014); and (4) the Sagittarius stream simulations of Law & Majewski (2010).

We find that, of the 21 candidates: four are coincident with the Sagittarius stream, four are consistent with M13, and one is a known open cluster. Of the 12 remaining, three are single detections and six are exact duplicate detections in adjacent fields; we ignore these. The remaining three candidates are clustered in a three degree area, do not coincide with any known structure in the catalogs listed above, have consistent properties, and are not duplicates. This overdensity is the subject of this letter.

In Figure 1 we plot the velocity–metallicity plane for spectra in a 15 (radius) circle about the mean detection coordinates (α, δ) = (274.5, 52.33). These data are fitted to a mixture of Gaussians in velocity space, as shown. Then the metallicity of the structure is estimated via an iterative 2σ rejection of the data falling into the outlying velocity profile. This algorithm selects the solid selection box shown.

Zoom In Zoom Out Reset image size

A looser, dashed selection box is constructed by eye around the region: −182 km s⁻¹ < V_radial < −150 km s⁻¹, −0.75 dex < [Fe/H] < −0.5 dex. This selection box is chosen to encompass all members in a larger sky area which appear as if they could plausibly be associated with the structure.

3.1. Significance of Overdensity

3.1.1. Significance in Metallicity Velocity Space

To calculate the significance of the overdensity in metallicity and velocity space, we compare objects within 5° of the target field at (l, b) = (80.6, 26.1) to a background of objects in a local annulus of 5°–10°. Here we assume that the target selection function is equivalent in these two fields.

To remove the disk from our data sample, the fraction of which may vary between the two fields, we consider only stars which have radial velocities less than −100 km s⁻¹ and metallicities less than −0.5 dex.⁴ Our on-target field has 204 stars in this velocity–metallicity range, of which 24 fall into the loose overdensity selection box described above. For the background annulus the numbers are 376 and 27, respectively, i.e., the probability of finding a star in the loose overdensity selection box is only 7% and our on-target field is expected to have less than 15 stars. Therefore, using binomial statistics, the chance of obtaining at least 24 such stars from a sample of 204 is only 1.2%. This constitutes our first probability.

This significance is shown visually in the inset of Figure 1. The frequency of stars falling into this selection box is plotted as a function of radius from the center of the detection and the significance compared to a background frequency is shown in shaded regions. The center of the group is almost 4σ above the background and there are three independent bins with detection over 1σ.

3.1.2. Significance in Distance and Proper Motion Space

To calculate the significance of the overdensity in distance and proper motion space, we refer to an analysis technique outlined in Schlaufman et al. (2009). The essential idea is to compare the slopes of the cumulative distribution functions of two populations in some parameter space, a steeper slope in the cumulative distribution function relative to that of a smooth background indicates an overdensity at that point in the parameter space. This test is more attuned to local overdensities than the widely used Kologmorov–Smirnov test, which is commonly used to compare global distribution functions. By sampling the background distribution many times, confidence intervals may be inferred. This test is shown in Figure 2 and yields probabilities of 0.005, 0.05, and 0.09⁵ for the distance, ${\mu }_{\alpha }$, and ${\mu }_{\delta }$ distributions (respectively) having been sampled from the background annulus described above.

Zoom In Zoom Out Reset image size

3.1.3. Total Significance

Assuming that these probabilities are independent of each other, we combine them as follows:

$$\begin{eqnarray}&&P=P([\mathrm{Fe}/{\rm{H}}],{v}_{r})\cdot P({\mu }_{\alpha })\cdot P({\mu }_{\delta })\cdot P(\mathrm{dist})\end{eqnarray} \tag{ 1 }$$

finding a total probability of 3.0e−7 that it is drawn from the local background (5.1σ).

The 5° on-target field of our overdensity contains 204 stars, which is drawn from a total LAMOST population of 22,798 such stars (after the disk cut) across the entire survey footprint. Therefore our full sample contains approximately 112 similarly sized samples. The binomial probability of finding one or more detections at this magnitude in 112 random samples is approximately zero.

4.1. Age and Distance

To determine the age of the object, we collect the sample of objects within 5° of (l, b) = (80.6, 26.1) that also fall into the dashed selection box in Figure 1. Using the spectroscopic data from the LAMOST pipeline, along with dereddened Two Micron All Sky Survey photometry (2MASS, Skrutskie et al. 2006; dereddened using the maps of Schlegel et al. 1998 and the filter coefficients of Davenport et al. 2014 which are calculated relative to the r_SDSS band extinction coefficient of Schlafly & Finkbeiner 2011), we fit these data to a series of Padova isochrones (Bressan et al. 2012, Chen et al. 2014, Tang et al. 2014) with the metallicity set to the average of the candidate objects ([Fe/H] = −0.64). Fitting is accomplished by finding the age of the isochrone which minimizes the ${\chi }^{2}$ value:

$$\begin{eqnarray}&&{\chi }^{2}={\left(\displaystyle \frac{{T}_{i}-{T}_{j}}{{\sigma }_{{T}_{i}}}\right)}^{2}+{\left(\displaystyle \frac{\mathrm{log}{(g)}_{i}-\mathrm{log}{(g)}_{j}}{{\sigma }_{\mathrm{log}{(g)}_{i}}}\right)}^{2}.\end{eqnarray} \tag{ 2 }$$

For each observed point i and each isochrone j, where T is the effective temperature and log(g) is the surface gravity.

Once an age is selected, the distance to the population is estimated by minimizing the total Euclidean distance of the candidates to the isochrone in temperature, surface gravity, and magnitude space for varying distance moduli, where all coordinates are normalized over the range of that parameter in the isochrone.

We find that the best fitting isochrone is 11 Gyr old (90% probability of being >4.4 Gyr old) at a distance of 2.6 kpc. This fit is shown in Figure 3. We note that the distances to the stars independently estimated by the spectra (Wang et al. 2016; Carlin et al. 2015) are in good agreement (see Figure 2). The normalized histogram above Figure 3 shows the distribution of all stars which fall into our dashed selection box. The overdensity of cool giants in this part of the sky, relative to the general distribution of stars in this selection box, is further evidence that our detection is real.

Zoom In Zoom Out Reset image size

4.2. Velocity and Orbit

4.3. Observational Size, Mass, and Luminosity

From the independent distance estimates of Wang et al. (2016), the clump has a standard deviation in distance of ∼1.2 kpc; when considering the reported distance errors of ∼0.9 kpc, this is an actual depth of ∼910 pc. For a rough estimate of the angular extent, we look at the frequency of stars falling into our metallicity–velocity window as a function of distance from the center (see Figure 1). We find that the frequency falls to near-background levels at ∼4° in radius, which, at the distance estimated by the isochrone fit of 2.63 kpc corresponds to a width of 367 pc. Estimating the width with certainty is difficult without uniform photometric data, owing to the patchiness of the LAMOST coverage. However, the width and depth are similar to each other and the slightly more radial extent is consistent with the orientation of the orbit (i.e., viewing the apocenter at a tangent).

We did attempt to investigate the surface density of M-Giant stars in Wide-field Infrared Survey Explorer (Wright et al. 2010) data (see Li et al. 2015) and 2MASS data as a way to characterize the spatial dimensions, but our overdensity was not found in those data.

We next make a crude estimate of the mass and luminosity of this object by integrating the functions:

$$\begin{eqnarray}&&\mathrm{mass}=K\int m(J)\cdot {\mathrm{IMF}}_{\mathrm{num}.}(J)\cdot {dJ},\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&\mathrm{luminosity}=K\int L(J)\cdot {\mathrm{IMF}}_{\mathrm{num}.}(J)\cdot {dJ},\end{eqnarray} \tag{ 4 }$$

over the total 2MASS J band ranges of the isochrones. L(J) and m(J) are estimates of the luminosity and mass of a star given a J band magnitude, interpolated from the isochrone. The IMF is the Chabrier (2001) initial mass function. K is determined to satisfy the equation:

$$\begin{eqnarray}&&N=K{\displaystyle \int }_{10}^{14}c(J)\cdot {\mathrm{IMF}}_{\mathrm{num}.}(J)\cdot {dJ},\end{eqnarray} \tag{ 5 }$$

where N = 22 $\pm \sqrt{22}$ is the number of stars we observe in the population, c(J) is the completeness of the LAMOST spectroscopy in the relevant color range, as estimated by 2MASS photometry. This rough estimate places the total mass at $2.1\pm 0.4\cdot {10}^{4}$ ${M}_{\odot }$ and the luminosity around $1.2\pm 0.3\cdot {10}^{4}$ ${L}_{\odot }$.

4.4. Progenitor

The system's orbit, passing so deep into the Galactic potential, implies that an object of mass $2.1\pm 0.4\cdot {10}^{4}$ ${M}_{\odot }$ would be tidally disrupted on relatively short timescales compared to the estimated age of 11 Gyr. However, it is known that structures remain coherent in velocity space long after being physically dissociated in the Milky Way potential (Helmi et al. 1999). The stream is also observed close to its calculated apogalacticon, where stream members are known to pile up in physical space (Dehnen et al. 2004).

The stream's progenitor could be either a dwarf galaxy or a star cluster. In Figure 4 we compare the half light radius and bolometric magnitude of the clump, a collection of globular clusters taken from Harris (1996), and a collection of dwarf galaxies taken from McConnachie (2012). The high metallicity of our stream implies that, if the progenitor were a dwarf galaxy, it should be relatively massive, on par with that of the Sagittarius Dwarf. Since we do not see any prominent physical streams in the photometric data, we look to another possible explanation. If instead the progenitor is a star cluster, the elliptical orbit and advanced age of our stream argues toward a globular cluster origin.

Zoom In Zoom Out Reset image size

We have performed a gridwise search of LAMOST data for clumps in [Fe/H] and radial velocity space. Most of the algorithmic detections belong to known structures such as the Sagittarius tidal streams and the galaxy M13. A number appear to be spurious detections. However, a significant detection has been noticed moving away from the Sun at ∼164 km s⁻¹ at a metallicity of −0.64 dex.

We collect the believed constituents of this clump and compare to the metallicity–velocity, proper motion, and distance distribution of stars in a local background field, we find that this sample is a 5.1σ signal over the background.⁶

Fitting an isochrone to the stars yields a best fit isochrone at 2.6 kpc that is most likely around 11 Gyr old and unlikely to be younger than 4.4 Gyr. Considering our observational completeness along with this best fit isochrone, we estimate the clump to have a mass of $2.1\pm 0.4\cdot {10}^{4}$ ${M}_{\odot }$ and a bolometric luminosity of $1.2\pm 0.3\cdot {10}^{4}$ ${L}_{\odot }$. The properties are summarized in Table 1. Note that since this clump is likely merely a portion of a larger stream, these are just estimates of the progenitors mass and luminosity.

Table 1.  Properties of the Clump

R.A.	2745
Decl.	523
[Fe/H]	−0.64 dex
${\mu }_{\alpha }$	−7.8 mas yr−1
${\mu }_{\delta }$	6.0 mas yr−1
Distance	2.6 kpc
Age	11.0 Gyr

Download table as:  ASCIITypeset image

The large physical size and the fact that the tangential velocity dispersion is so much higher than the radial velocity dispersion imply that this could be part of a tidal stream, although some tangential velocity dispersion is certainly an effect of proper motion uncertainty at this distance. Integrating its position and velocity through a galactic potential implies that we are observing the clump near its apogalacticon on an eccentric orbit. The high metallicity could point to a large dwarf galaxy progenitor; however, the lack of a strong physical overdensity argues against this interpretation The above properties lead to our classification of this comoving clump as a disrupted globular cluster, which we dub Lamost 1.

We thank Dr. Emma Small for useful discussion. We thank the developers and maintainers of the following software libraries which were used in this work: Topcat (Taylor 2005), Aladin (Bonnarel et al. 2000), NumPy, SciPy, AstroPy (Astropy Collaboration et al. 2013), astroML (VanderPlas et al. 2014), matplotlib, and Python. Guoshoujing Telescope (the Large Sky Area Multi-object Fiber Spectroscopic Telescope LAMOST) is a National Major Scientific Project built by the Chinese Academy of Sciences. Funding for the project has been provided by the National Development and Reform Commission. LAMOST is operated and managed by the National Astronomical Observatories, Chinese Academy of Sciences.

J.J.V. acknowledges the support of a LAMOST fellowship. M.C.S. acknowledges financial support from the CAS One Hundred Talent Fund. This work was also supported by National Key Basic Research Program of China 2014CB845700, and from NSFC grants 11173002 and 11333003 as well as the Strategic Priority Research Program The Emergence of Cosmological Structures of the Chinese Academy of Sciences, Grant No. XDB09000000.