© ESO, 2014

1. Introduction

With this paper we continue to present the results of the Milky Way Star Cluster (MWSC) survey undertaken on the basis of the two all-sky catalogues, 2MASS (Skrutskie et al. 2006) and PPMXL (Röser et al. 2010). The MWSC survey was initiated a few years ago with the aim of building a comprehensive sample of Galactic star clusters with well-determined parameters, which is complete enough to enable an unbiased study of the content and evolution of the star clusters of our Galaxy. The first paper of this series (Kharchenko et al. 2012, hereafter Paper I), gave an introduction to the survey, explained the underlying motivation, provided a short review of similar studies, described the observational basis of the survey, the data processing pipeline, and presented preliminary results obtained in the second Galactic quadrant. The second paper (Kharchenko et al. 2013, Paper II) summarises the results of the full survey carried out for a compiled input list of 3784 known objects, covering the whole sky. It presents uniform structural, kinematic, and astrophysical data for 3006 open clusters, globular clusters, and compact associations.

The first-look analysis of the MWSC data carried out in Paper II has shown that the MWSC sample is complete up to a distance of d = 1.8 kpc from the Sun for clusters of all ages except the older clusters (log (t [ yr ]) >9). Although this shortage concerns primarily the oldest clusters, the effect can be seen in the general distribution of all Galactic open clusters in Fig. 1, where we show the cluster distribution in the plane [ Z, d_XY ], with Z the vertical distance from the Galactic plane, and d_XY the distance from the Sun projected onto the Galactic plane. One can clearly see that at d_XY ≲ 2 kpc the number of high-latitude clusters diminishes with decreasing d_XY.

The general lack of old open clusters has already been noted in the 1950s (e.g. Oort 1958). Since then, old clusters have been mainly discovered at distances ≳1 kpc, resulting in a striking apparent absence of old clusters in the solar neighbourhood.

There are two main reasons that nearby old open clusters may have escaped previous searches, both based on their proximity.

• 1.

  Old open clusters show a larger scale height (van den Bergh & McClure 1980; Froebrich et al. 2010), so in combination with small distances they may be located at higher Galactic latitudes, while systematic searches for open clusters have typically been restricted to areas close to the Galactic plane (e.g. Mercer et al. 2005: | b | < 1°; Froebrich et al. 2007: | b | < 20°; Glushkova et al. 2010: | b | < 24°);

• 2.

  With a large angular extent (up to several degrees), they do not stand out prominently as overdensities from the field.

The primary goal of this paper is to get a complete list of clusters within the MWSC survey. To reach this goal we expand previous searches of star clusters in 2MASS performed typically at | b | ≲ 20° to higher Galactic latitudes. This work can be considered as an extension of the search by Froebrich et al. (2007), which used the same data and a similar approach (without filters), but was restricted to the area | b | < 20°.

In Sect. 2 we describe the data set and our method of identifying clusters. The results are presented in Sect. 3 and discussed in Sect. 4.

2. Method

2.1. Data

Cluster candidates were identified as density enhancements in the Two Micron All Sky Survey (2MASS) point-source catalogue (Skrutskie et al. 2006). The 2MASS survey provides the photometric basis of the MWSC survey with a uniformly calibrated photometry of the entire sky, complete down to K_s ≈ 14.3 mag, depending on the position on the sky. We only considered sources that were detected in all three bands (J, H, K_s) with high quality (R_flg = 1, 2 or 3). We applied our search algorithm to the entire sky at Galactic latitudes | b | > 20°.

2.2. Filtering the sample

Since nearby old clusters may not exhibit a significant overdensity in the plain 2MASS, we had to enhance the contrast between potential clusters and the field. Therefore, we used cuts in colour and magnitude according to typical colour–magnitude diagrams (CMDs) of clusters in different age, distance, and extinction bins. This is an approach comparable, albeit somewhat simpler, to what has been used to detect tidal tails of globular clusters (e.g. Grillmair et al. 1995; Odenkirchen et al. 2003). We set up 18 different filters to cover the colour–magnitude space expected for clusters with 0 ≲ A_Ks ≲ 0.3 mag and 8.8 ≲ log (t [ yr ] ) ≲ 9.4 at distances around 0.5 and 1 kpc (Fig. 2). This filtering procedure reduces the number of sources in a field to between about 10 and 40 per cent. Figure 3 illustrates the effect of the filtering: While no significant density enhancement can be detected in the unfiltered distribution, a density enhancement above the 4σ level shows up after applying one of the filters. This feature is subsequently confirmed as an open cluster (MWSC 5723).

The filters are not designed to model a specific type of cluster, but to cover the parameter range in the CMD expected for clusters in the desired age and distance range, to reduce the contamination from unrelated background objects. Since the filters are rather wide, overlap strongly and occupy a wide range, they also cover other parameter combinations, in particular for smaller and larger distances.

2.3. Finding cluster candidates

The filtered sample, together with the unfiltered catalogue, is then used as input for a cluster search algorithm based on star counts (e.g. Carpenter et al. 1995; Lada & Lada 1995; Ivanov et al. 2002; Reylé & Robin 2002). This rather simple approach is nevertheless a very efficient way of creating stellar density maps and identifying density enhancements in a field, comparable to or better than more sophisticated approaches, such as the nearest neighbour density or the separation of minimum spanning trees (Schmeja 2011). We use fields of 5° × 5° in size. Every field is subdivided into a rectilinear grid of overlapping squares that are separated by half the side length of an individual square (the Nyquist spatial sampling interval). The size of the bins is chosen such that they contain 15 stars on average. This results in bins with side lengths between about 3 and 20 arcmin. All areas showing a density ≥4σ above the average density of the field are considered potential clusters, if they contain at least ten sources. Tests showed that bins with a size that on average gives 15 stars per bin and an overdensity threshold of 4σ are best suited to detecting clusters without missing a signifiant number of clusters and picking up too many random density enhancements. Density enhancements that by visual inspection could obviously not be Galactic stellar clusters (such as fragments of M 31 or the Magellanic Clouds) were neglected. As a result we prepared a list of candidate clusters that contains the coordinates of the centres of the density enhancements and their sizes.

2.4. Veryfing the candidates and determination of the cluster parameters

To be certain that we do not re-discover already known objects, we checked if any candidate was in the MWSC input list, in the SIMBAD data base¹, and because many compact galaxies may appear as point sources in the 2MASS, in the list of galaxy clusters from the Abell et al. (1989) catalogue. The correctness of the preliminary choice of the candidate objects is supported by frequent coincidence of the candidates found with already known objects. The list of unidentified candidates, together with preliminary data on their positions and sizes, was processed with the MWSC pipeline for further checks for the construction of cluster membership and parameter determination.

The pipeline uses kinematic, photometric, and spatial information on stars in the candidate area and is described in more detail in Paper I. The main purpose of the pipeline is to clean a candidate from the fore- and background contamination using kinematic, photometric, and spatial criteria, to produce a list of probable members and to determine the basic cluster parameters in the case of success. The pipeline consists of iterative series of interactive checks of vector point diagram of proper motions, radial density profiles, magnitude-proper motion relation, and various colour–magnitude, two-colour, and Q_JHKs-colour diagrams. As a theoretical basis, we use recent Padova stellar models of Marigo et al. (2008) and Girardi et al. (2008) with isochrones computed with the CMD2.2 on-line server², whereas the pre-main sequence isochrones were computed by us from the models of Siess et al. (2000) and then transformed to the JHK_s photometric system using transformation tables provided by the Padova team with the dustyAGB07 database³. The membership probabilities of stars in the diagrams take data accuracy into account and are determined from the star location with respect to the reference sequences (represented either by isochrones or the average cluster proper motion), which themselves depend on the cluster parameters we want to find. This therefore requires an iterative approach, allowing us to successively improve both cluster membership and cluster parameters. The initial approximation was based on a visual inspection of the diagrams. As a rule, the process converges after a few iterations. Including spatial and kinematic criteria greatly helps reduce ambiguities in determining age, distance, and reddenning, which may arise if only photometric membership is considered. Details of this effect, called degeneracy, are described in detail in Paper I (Sect. 3.4.3).

The verification of the overdensities as clusters is based on the most probable members only with P_m> 61% (deviating from the reference by less than one rms-error). If their distribution in the vector point diagram of proper motions is more compact than for the rest of the stars, and if they fit the critical points of the isochrone (turn-off, red-giant branch), a candidate is considered to be confirmed, and the most probable members are used for computing the cluster parameters. Otherwise it is rejected as a random clustering of field stars (asterism). The verification by visual inspection of the diagrams is supported by objective statistical arguments. Applying a Fisher test to the identified clusters, we find that the populations of the most probable cluster members (P_m> 61%) and of “field” stars (P_m< 1%) have significantly different dispersions both in the vector point diagram (for 120, or 88% of the clusters) and in the CMD (for all clusters). Figures A.1 and A.2 show the atlas page of an exemplary cluster (MWSC 5224) with its spatial distribution, the radial density profile, the CMDs, and proper motion diagrams.

3. Results

The statistics of results of our cluster search is given in Table 1, which shows the number of candidates, divided into three groups of objects: new real clusters, asterisms, and re-identified known stellar or galaxy clusters. About half of the candidates match known objects: 338 galaxy clusters, 33 globular and 6 Milky Way open clusters, and 8 clusters in the Large Magellanic Cloud. Comparing these statistics to the data present in the catalogues, we can estimate the efficiency of the applied search algorithm. At | b | > 20° there are 49 Galactic globular clusters in the catalogue of Harris (1996, edition 2010). This means that we were able to detect 67% of the known globular clusters. The remaining globular clusters are too faint or too poorly represented in 2MASS to be detected. There were 61 open clusters at | b | > 20° in the MWSC catalogue prior to this work. Excluding associations, moving groups, embedded clusters, and cluster remnants from the sample, there are 18 clusters (called “compact” here), of which we were able to identify six (NGC 188, NGC 2682, NGC 1662, NGC 1980, NGC 2632, and Blanco 1), corresponding to a detection rate of 10% of all open clusters or 33% of the compact clusters. According to the SIMBAD database, there are 26 227 clusters of galaxies at | b | > 20°. For those our detection rate is of the order of 1% (338). The detection rate of open clusters is therefore ten times higher than the detection rate of clusters of galaxies.

Out of the 139 new clusters, 104 were detected using the CMD filters described in Sect. 2.2, 34 were only found without filters, and one was detected by both applying one of the filters and using the unfiltered field. Since we performed both a filtered and an unfiltered search, we found clusters outside the targeted age and distance limits implied by the filters.

In Fig. 4 we show the distribution of the newly discovered clusters on the sky, together with the previosuly known open clusters from the MWSC survey. The majority of the confirmed clusters are located within | b | ≲ 30°, though a few open clusters were found up to | b | ≈ 60°. However, most of the high-latitude candidates turned out to be galaxy clusters.

In Fig. 5 we compare the distributions of the parameters of newly detected clusters and of known high-latitude (| b | > 20°) clusters from the MWSC survey. We present the distributions of “structural” parameters, such as the total apparent radius r₂ of a cluster, the apparent radius r₁ of its densest central part, and the tidal radius r_t derived by fitting a King profile to the observed distribution. We also show an empirical estimator of cluster richness n₂, i.e. the number of the most probable cluster members within r₂. The lower panel of Fig. 5 shows the distributions of the so-called “photometric” parameters, derived from fitting cluster CMDs: age log t, reddening E(J − K_s), distance log d, and the height Z above the Galactic plane.

The data of the 139 new open clusters are available at the CDS and the GAVO data center as an extension to the MWSC catalogue⁴. The format is the same as that of the MWSC survey in Paper I. An overview with positions, radii, distances and ages of the new clusters is given in Table B.1.

4. Discussion

The initial goal of this search was to find unknown old star clusters at high Galactic latitudes, which might fill the local “hole” around the Sun, as we hoped. The results are illustrated in Figs. 6 and 7 where we compare the distribution of known and new clusters in the XY-plane and show the contribution of new clusters to the surface density of Galactic open clusters.

4.1. The “hole” around the Sun

Figure 6 shows that most of the newly discovered clusters occupy a ring around the Sun with inner and outer borders of d_XY ≈ 1 and 2 kpc (d_XY is the cluster distance projected on the Galactic plane), with almost no clusters at d_XY< 1 kpc. Figure 7 indicates that the new clusters slightly increased (by about 8%) the total surface density. The latter contribute mostly to the surface density of the oldest clusters (log t [ yr ] ) > 9.0), which becomes higher and flatter within the ring. At d_XY< 1 kpc, the shortage of the oldest clusters is now even more prominent. Assuming the average surface density within the ring to be typical for the whole range of the projected distances d_XY, we expect about 50 clusters still to be discovered in the solar vicinity. On the other hand, the new clusters do not significantly affect the surface density distribution of clusters with ages 8.3 < log t< 9.0, where a “hole” is only marginally visible at d_XY ≲ 0.5 kpc. Possibly about ten clusters are missing in this age and distance range. There is no convincing reason for old clusters to avoid the area around the Sun, therefore it is more likely that they escaped our search because of its limitations as discussed below.

4.2. Limitations of the search method

Most likely, the missing clusters are just too sparse and too extended to be found as overdensities, even when applying our colour–magnitude filters. For example, it was not possible to detect the cluster Ruprecht 147 (d = 175 pc, log t = 9.39, ; Kharchenko et al. 2005b) with our algorithm. In the area of Ruprecht 147 there are, even when applying our filters, more than 11 000 field stars in 2MASS, compared to about 150 members found for this cluster in the MWSC survey. When only considering the cluster core, there are about 480 field stars compared to 20 cluster members. This is much smaller than the average noise. Even using a very narrow filter specifically tailored to the CMD of Ruprecht 147 instead of our standard filters does not reduce the background to a level where the cluster becomes detectable as an overdensity. Similar to Mercer et al. (2005), who added artificial clusters to their catalogue and tried to recover them, we did additional tests by simulating the Hyades (d = 45 pc, t = 650 Myr) at different distances between 0.6 and 2 kpc at a latitude of b ≈ 30°. It turns out that only at distances ≥1 kpc is the innermost core (r ≈ 3 pc) of the cluster detected as a significant overdensity.

We also investigated the effect of our filters on the search. A comparison of the results of both filtered and unfiltered searches (see Fig. 8, showing the distances of new clusters identified with different filters) indicates that the distances do not strongly depend on a specific filter or on its absence. Another experience we gained from the results of this search is that it seems that the presence of cluster members on the giant branch facilitates their discovery with the filters, so an absence of giants may result in the clusters not being detected.

To estimate the effect of the search method and the underlying catalogue, we compare the distributions with distance of clusters detected in recent optical and NIR surveys (Fig. 9). Both surveys differ by the basic catalogues they use and by the search algorithm. The optical data are represented by the Catalogue of Open Cluster Data (COCD, Kharchenko et al. 2005a,b), based on the catalogue ASCC-2.5, which provides a higher accuracy of kinematic and photometric data and a lower level of background contamination than the combination PPMXL+2MASS does. Unlike the current detection algorithm, the new clusters in this study were searched as density enhancements in the four-dimensional space of proper motions and coordinates in the fields around bright stars (V< 9 mag). In the case of MWSC, in addition to the current set of high latitude clusters we consider data on low latitude clusters of Froebrich et al. (2007), which are included in the MWSC input list. While the total distribution of clusters in the NIR-based MWSC extends to higher distances than those of the optical survey COCD, their subsets of newly identified clusters differ with respect to the lower limit of their distances. While the bulk of new clusters found in the optical reside at distances less than 1 kpc, all the objects detected in the NIR are located outside the 1 kpc limit. This tendency is also seen in other detections of new objects based on the 2MASS catalogue (see e.g. Glushkova et al. 2010). One should note that all these results are based both on the same data source (2MASS) and use a similar approach of searching new clusters as density enhancements in the sky.

Other approaches, such as a search using proper motions (Scholz et al., in prep.), may be more successful in finding the missing nearby clusters. In the long run, the Gaia mission is expected to fill the gap.

5. Summary

From a first-look analysis of the MWSC in Paper II, we found evidence for a lack of nearby old clusters at high Galactic latitudes and projected distances d_XY ≲ 1 kpc. An additional search for star clusters was carried out on the basis of 2MASS and PPMXL at latitudes | b | > 20°. We applied colour–magnitude filters and a star count algorithm to search for these old open clusters. This resulted in the detection of 782 overdensities, regarded as cluster candidates. A comparison with lists of known objects (MWSC input list, SIMBAD data base, and the list of Abell galaxy clusters) has shown that 385 of them are already known objects. The remaining 397 cluster candidates were processed with the standard MWSC pipeline, which confirmed the cluster nature of 139 objects. All of them are open clusters with ages 8.3 < log t< 9.7, distances <3 kpc, and distances from the Galactic plane 0.3 <Z< 1 kpc. This increased the total number of known high-latitude open clusters by about 150%. Nevertheless, the “hole” with a radius of about 1 kpc around the Sun could not be filled. This dearth of old clusters is expected to be an artefact from the bias against sparse overdensities with large angular size on the sky. We estimate that still about 60 old open clusters are missing in this volume.

---

Acknowledgments

We wish to thank the referee for a detailed and helpful report. This study was supported by Sonderforschungsbereich SFB 881 “The Milky Way System” (subproject B5) of the German Research Foundation (DFG) and by DFG grant RO 528/10-1. This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. This research has made use of the SIMBAD database, operated at the CDS, Strasbourg, and of the WEBDA database, operated at the Institute for Astronomy of the University of Vienna.

References

Appendix A: Atlas page of MWSC 5244

Figures A.1 and A.2 show the pipeline output for the newly identified cluster MWSC 5224 to illustrate the star member selection procedure and the quality of the determined cluster parameters. The example cluster is selected at random and represents a typical case among the analysed objects.

The main diagram of the first page of the Atlas is a cluster map, while on the second page this role is played by the K_s,(J − H), and K_s,(J − K_s) diagrams. Stars are shown as coloured circles or dots. Symbols and their colours have the same meaning in all plots. Cyan symbols mark stars outside the cluster radius r₂ and green symbols stars within r₂. The most probable kinematic and photometric members (1σ-members) are indicated in black for members located within r₁, red for members between r₁ and r₂, and blue for stars outside r₂. Cyan bars show the uncertainty for 1σ-members (page 2).

Page 1 of the Atlas (Fig. A.1) contains five diagrams with spatial information, as well as a legend on the derived cluster parameters. The right panel is a map of the cluster surroundings. The left panels show magnitudes K_s, proper motions PM_x, PM_y, and surface density N versus distances r of stars from the cluster centre.

In the sky map, stars are shown by circles. Their size corresponds to the brightness arranged in six K_s magnitude bins. The blue cross indicates the cluster centre determined in this study. If by chance other clusters appear in this area, their centres are marked by magenta plus signs. Large blue circles (shown by dotted, solid, or dashed curves) indicate the cluster radii r₀, r₁, and r₂, respectively. In the left-hand panels, the blue vertical lines (dotted, solid, or dashed) mark r₀, r₁, or r₂. Magenta horizontal lines in the PM vs. r diagrams correspond to the derived average proper motion of the cluster. Radial density profiles in the bottom panel are shown with green for all stars, blue for 3σ-members, magenta for 2σ-members, and black for 1σ-members.

The legend gives cluster name, MWSC number, and COCD number in parentheses; equatorial RA_J2000, Dec_J2000, and Galactic l,b coordinates of the cluster centre; apparent cluster sizes r₀, r₁, r₂, and number of 1σ-members within the corresponding radius; weighted average components PM_X,Y of proper motion with their rms errors and number of stars used to compute the average; the average radial velocity, RV, rms error, and the number of stars used to compute the average; distance to the cluster, d, distance modulus, (K_s − M_Ks); NIR interstellar reddening, E(J − H),E(J − Ks), and interstellar extinction, A(K_s); cluster age, and its rms error. The number in brackets gives the number of stars used to compute the average age, or it is − 1 if an isochrone fitting was applied. The ΔH shown below the photometric diagrams indicates the empirical correction to the H-magnitude introduced in Paper I.

The parameters are shown as they were derived in the pipeline without taking their real accuracy into account, which was estimated by us from comparison with literature data after the MWSC was completed (for details see Paper II). Typically the cluster proper motions are accurate within 1 mas/yr, and the derived distances and reddenings are accurate within 11% and 7% respectively. An accuracy of the order of 10% is achieved for the ages of older open clusters (log (t [ yr ] ) > 8.2). Page 2 (Fig. A.2) contains three diagrams with kinematic information (left panels), and six diagrams with photometric information (right panels).

The three left panels with kinematic data: the two upper diagrams show PM_X,Y vs. K_s relations, i.e. “PM-magnitude equation”. Magenta vertical lines correspond to the average proper motion of the cluster. The magenta dashed line shows the apparent magnitude ${\mathit{K}}_{\mathrm{s}}^{\mathrm{mc}}$, which corresponds to the bluest colour (J − K_s) of the adopted isochrone. The bottom panel is the vector point diagram of proper motions.

The six right panels with photometric data: the two upper diagrams are CMDs (K_s,(J − H) and K_s,(J − K_s)). The magenta curve is the apparent isochrone closest to the determined cluster age. Solid blue lines outline a domain of 100% photometric members. Solid red lines (shown only in K_s,(J − K_s)) are the ZAMS (zero-age main sequence) and TAMS (terminal-age main sequence), described in more detail in Paper I. The magenta dashed line shows the apparent magnitude of minimum colour ${\mathit{K}}_{\mathrm{s}}^{\mathrm{mc}}$. The thick yellow circles mark the stars used for the age determination (see Kharchenko et al. 2005b, for details). The black arrows show the vectors of increasing extinction. The four bottom panels show the two-colour (H − K_s) / (J − H) diagram (left column) and Q_JHK-colour diagram (right column). The upper row is for stars brighter than ${\mathit{K}}_{\mathrm{s}}^{\mathrm{mc}}$, the lower row is for stars fainter than ${\mathit{K}}_{\mathrm{s}}^{\mathrm{mc}}$. Magenta curves indicate the apparent isochrone (i.e., apparent colours), whereas cyan curves show the intrinsic isochrone.

The legend is the same as on page 1.

The atlas pages for all new clusters are available in electronic form at the CDS and at the GAVO data center⁵.

Appendix B: Table of newly identified clusters

Table B.1 gives an overview of the 139 newly identified clusters. Since the total list of determined parameters is too long (37 columns), we only show the most important parameters here for quick reference (cluster names, equatorial and Galactic coordinates, their total sizes r₂, distance, and age). The full list of cluster parameters is available in electronic form at the CDS. It is in the same format as the table determined earlier in Paper II for the main body of the MWSC survey.

All Tables

All Figures