The main sequence of three red supergiant clusters

Abstract

1 INTRODUCTION

Massive stars are most commonly formed in clusters and associations (Gvaramadze et al. 2012), even if there are potential exceptions (e.g. Oey et al. 2013). These building blocks are the sole observational characteristic of star formation that is observable in more distant galaxies. It is thus of importance to investigate the details of such massive clusters locally in our own Galaxy.

The identification of local examples of clusters and associations of massive stars is complicated by a number of facts. Such objects are generally rare in the Milky Way and are thus typically at large distances. The Sun's position near the Galactic plane, hence, makes it difficult to identify and investigate these objects due to the large amounts of extinction along the line of sight. Furthermore, most massive star formation is projected towards the general direction of the Galactic Centre, which causes additional difficulties due to crowding.

Thus, the identification and characterization of massive star clusters in our own Galaxy has only recently gained considerable momentum due to advances in infrared astronomy that allow us to probe these distant and obscured objects. Many of them have been known for a considerable time, but have not been recognized as massive star clusters. For example, Westerlund 1 was discovered by Westerlund (1961) but only recognized as young, very massive cluster more than 30 yr later (Clark et al. 1998; Piatti, Bica & Claria 1998). Similarly, the cluster Stephenson 2 (Stephenson 1990) had been known for many years until its true nature as Red Supergiant Cluster 2 (RSGC 2) was uncovered (Ortolani et al. 2002; Davies et al. 2007). Further, more systematic searches to uncover the population of the most massive clusters in the Galaxy are ongoing, e.g. Rahman, Matzner & Moon (2013).

Clusters at the top end of the mass distribution in our Galaxy (above 10⁴ M_⊙) have, depending on their age, a sizeable number of evolved massive stars. These are either Wolf–Rayet stars, or blue/yellow/red supergiants. There are now a number of galactic clusters with a large population of RSG stars known, many of which are in a region confined to where the Scutum spiral arm meets the near side of the Galactic bar. These are e.g. RSGC 1–5 (Figer et al. 2006; Davies et al. 2007; Alexander et al. 2009; Clark et al. 2009; Negueruela et al. 2010, 2011). All these clusters are about 6 kpc from the Sun, highly extincted, between 7 and 20 Myr old and have masses in excess of 10 000 M_⊙. These clusters are thus an ideal testbed to study the formation and evolution of massive stars as well as their influence on the environment in great detail. However, it is important in the context of understanding the formation and evolution of massive clusters to investigate the distribution and properties of the lower mass stars as well as the bright supergiants and Wolf–Rayet stars. In particular, since the intrinsic properties of lower mass stars are much better understood, this should enable us to determine the distances, ages and reddening of these clusters more accurately.

The main-sequence (MS) stars of these clusters have so far, however, not been investigated (though the slightly less massive and less reddened cluster NGC 7419 has recently been studied by Marco & Negueruela 2013). Given that the RSG cluster members have an apparent brightness of about K = 6 mag, one would expect to detect the MS in deep near-infrared (NIR) images. In particular the Galactic Plane Survey (GPS; Lucas et al. 2008) data from the UKIRT Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007) seems to be ideal to detect the fainter members of the RSG clusters mentioned above. But even data from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) should be deep enough to detect the brightest MS stars in these clusters.

Only for the cluster RSGC 2 has the MS been detected so far (Froebrich 2013). The author uses 2MASS and GPS data to identify the top of the MS at colours of about J − K = 1.5 mag and at a brightness of slightly fainter than K = 10 mag. The author also investigates the data for RSGC 1 and RSGC 3 with the same methods but is unable to detect the MS for these clusters. It is speculated that this is caused by problems with accurate aperture photometry in the vicinity of the bright RSG stars in those objects. The MS in RSGC 2 is the easiest to detect due to the spatial extent of the cluster, which is the largest amongst the RSG clusters. Here, we hence try to identify the MS of other known and candidate RSG clusters by means of point spread function (PSF) photometry in the JHK images of the UKIDSS GPS data. This will include some of the new RSG cluster candidates (F 3 and F 4) identified by (Froebrich 2013) based on colour-selected star density maps from 2MASS.

This paper is structured as follows. In Section 2, we briefly outline the data and analysis methods employed. We then discuss our results in Section 3 with focus on the properties of the MS detected in RSGC 1–3.

2 DATA AND ANALYSIS METHODS

2.1 UKIDSS data

In Froebrich (2013), the MS of RSGC 2 has been detected in aperture photometry of UKIDSS GPS data between K = 11 and 14.5 mag. All stars brighter than this are saturated. Tentatively, this MS was also visible in 2MASS data starting from about K = 10 mag. The RSGs in this cluster are between 5th and 6th magnitude in the K band (Davies et al. 2007). Thus, the brightest MS members in this cluster are five or six magnitudes fainter than the RSGs. Given that several of the other known RSG clusters are in a similar position on the sky and potentially have similar ages, distances and reddening (Figer et al. 2006; Davies et al. 2007, 2008; Alexander et al. 2009; Clark et al. 2009), one can assume that their MS should have analogue properties, i.e. should have similarly bright members. However, for none of the other investigated RSG clusters or candidates in Froebrich (2013) has the MS been detected in the GPS aperture photometry data. This was attributed to crowding in the clusters, and thus low-quality aperture photometry of potential MS stars in the vicinity of the bright RSG cluster members. In other words, the GPS aperture photometry catalogue is highly incomplete near the bright RSG cluster stars. We hence selected RSGC 1–5 and the cluster candidates F 3 and F 4 from Froebrich (2013) for further investigations, i.e. PSF photometry, to reveal their MS.

We downloaded 10 arcmin × 10 arcmin cutouts of the JHK images from the UKIDSS GPS around the nominal position of each of these clusters via the Wide Field Camera Science Archive¹ webpage. We further obtained a complete list of all source detections (independent of the photometric quality) in each field. Typically, there are about 30–45 000 detections in each of the fields. We then manually removed all detections which were obvious image artefacts (detector cross talk, persistence, etc.) and added all visible real stars missed by the source detection software to these catalogues. In particular, near the bright RSG cluster members a number of real stars are missing. Several hundred detections have been removed/added for each field by visually inspecting the K-band images and manually deleting all obvious false detections and appending every object that has the appearance of a star but was missing in the photometric catalogue. In Table 1, we list the number of objects in these input catalogues for each cluster field.

2.2 PSF photometry

We performed PSF-fitting photometry in each image at all positions from the above generated input catalogue. For this purpose, we used the standard routines from daophot, as implemented in iraf (Stetson 1987). To define the PSF for a given frame, about 10–15 isolated reference stars distributed across the images were selected. Since the crowding was significant near the field centre, most of these reference stars are located in the outskirts of the clusters. We did not notice any sign of a spatially varying PSF. The average PSF of these reference stars is modelled in daophot with a combination of an analytical function and an empirical image, which better represents the wings of the PSF. This model PSF was then fitted to all objects in the catalogue, to create the final list of magnitudes. As measured by the χ² and the PSF-subtracted images, the PSF fitting performs well in the UKIDSS GPS images. Only in one case, the K-band image of the cluster candidate F 3 is the PSF undersampled and larger residuals are visible in the PSF-subtracted images. However, the quality of the photometry is not significantly degraded in this image (see Table 1).

We merged the photometric catalogues for each of the JHK filters and used the GPS aperture photometry of sources in each field to calibrate the instrumental magnitudes. Only objects with pstar ≥ 0.999 65 (see Lucas et al. 2008 for details on how this is defined) are used in the calibration and a nominal shift of the magnitudes and no colour terms are considered. The root-mean-square (rms) scatter of the calibrated magnitudes is listed in Table 1 together with the magnitude range in which the calibration was performed. There are some images where we detect a clear non-linearity for bright sources. This is expected, since the PSF fitting should be reliable over a wider range of the non-linear regime. Thus, the colours and magnitudes for the bright stars could be systematically off. Note that this only influences the very top of the potential cluster MS and will have no influence on our analysis.

All point sources which did lack a detection by the PSF-fitting routine in at least one of the three NIR bands were removed from the final catalogues. We also removed all stars which were fainter than the completeness limit (determined as the peak in the magnitude distribution) in at least one filter. Finally, all objects that had a photometric uncertainty which differed by more than 2σ from the average of stars with the same apparent magnitude were removed. The number of stars in this output catalogue for each cluster is listed in Table 1.

2.3 Photometric decontamination

Note that these cluster membership probabilities are strictly speaking not real probabilities (Buckner & Froebrich 2013), since fluctuations of the field star density in principle allow negative values for |$P^i_{\rm cl}$|⁠. If this occurs in our analysis the |$P^i_{\rm cl}$| value is set to zero. However, the sum of all |$P^i_{\rm cl}$| values equals the excess number of stars in the cluster area compared to the control field. High values of |$P^i_{\rm cl}$| identify the stars in the cluster field which are the most likely members and thus allow us to establish the overall population of cluster stars statistically. Typically, this method identifies a few hundred stars in the fields of the clusters RSGC 1–3 which have a membership probability above 50 per cent (see later).

3 RESULTS

Here we discuss the results obtained for all the investigated clusters. Some further details on the individual objects can be found in Appendix A.

3.1 General

Theoretically, one would expect the MS of these massive clusters to appear as a vertical accumulation of high-probability cluster members in a K versus J − K CMD, since most stars visible should be of high mass and thus have similar intrinsic NIR colours. Some scatter in colour is expected due to differential reddening along the line of sight. Furthermore, these high-mass MS stars should be situated at the bottom of the reddening band in a NIR H − K versus J − H colour–colour diagram (CCD) and not in the middle/top, which is usually occupied by giant stars.

In the left-hand panels in Fig. 1, we show as examples the decontaminated K versus J − K CMD of RSGC 1–3. All stars fainter than 8th magnitude in K are from our PSF photometry. All bright, potential RSGs (large black squares) are taken from 2MASS since they are saturated in the UKIDSS images. All small black dot symbols are stars in the cluster area with less than 50 per cent cluster membership probability |$P^i_{\rm cl}$|⁠. Green plus signs indicate stars with |$50 \le P^i_{\rm cl} < 60$| per cent, blue triangles stars with |$60 \le P^i_{\rm cl} < 70$| per cent and red squares |$P^i_{\rm cl} > 70$| per cent.

Figure 1.

Open in new tabDownload slide

Colour–magnitude (left) and Colour–colour (right) diagrams for RSGC 1 (top), RSGC 2 (middle) and RSGC 3 (bottom). Objects brighter than K = 8 mag are stars detected by 2MASS within 3 arcmin of the cluster centre and are shown as black filled squares. The remaining stars are from our PSF photometry. The colour coding indicates the photometric cluster membership probability of the stars: |$P^i_{\rm cl} > 70$| per cent red squares; |$60 \le P^i_{\rm cl} < 70$| per cent blue triangles; |$50 \le P^i_{\rm cl} < 60$| per cent green plus signs; |$P^i_{\rm cl} < 50$| per cent black dots. In the CMDs, the black solid line indicates the average position of the most likely cluster members, the blue solid line an isochrone from Lejeune & Schaerer (2001) (using the ages listed in Table 2) and the blue dashed line an isochrone from Siess, Dufour & Forestini (2000). The dotted boxes indicate the region of stars overplotted on the CCDs. The reddening band in the CCDs is based on the extinction laws from Indebetouw et al. (2005).

We determined a running weighted average of the J − K colours of the most likely cluster members along the detected MS. This is indicated by the solid black line in the CMDs in Fig. 1. As a weighting factor for each star, we used the square of the membership probability |$P^i_{\rm cl}$|⁠. To compare the cluster data with model isochrones, we utilize the Geneva isochrones by Lejeune & Schaerer (2001) which are overplotted in each panel as a blue solid line, using the parameters specified in Table 2. We also overplot as a dashed line the isochrones for low- and intermediate-mass stars from Siess et al. (2000).

In the right-hand panels in Fig. 1, we show the corresponding H − K versus J − H CCDs for the same three clusters. Symbols and colours are identical in their meaning to the CMDs. However, we plot all stars in the field as small black dots and only high-probability cluster members (⁠|$P^i_{\rm cl} > 50$| per cent) from the potential MS (as indicated by the dotted boxes in the CMDs) are shown in large coloured symbols. These boxes exclude bright, potentially saturated stars as well as faint, low signal-to-noise objects and obvious background giants. The indicated reddening band for each cluster indicated, is based on the reddening law by Indebetouw et al. (2005).

From all the clusters investigated, we can detect a MS only in the known objects RSGC 1–3 (see Fig. 1). We also investigated the fields around RSGC 4 (Negueruela et al. 2010) and RSGC 5 (Negueruela et al. 2011) but there are no apparent overdensities of stars, in particular, none that would indicate a MS (see Fig. D1 in Appendix D). There are several possibilities that could explain this. (i) These clusters have less mass, i.e. fewer members, than the other objects and thus they do not manifest themselves as overdensities in colour–magnitude space. (ii) The clusters are much more extended spatially than our search area for potential MS stars; they are more association like in appearance than cluster like. Both points seem to contribute, since both clusters have fewer confirmed members than the brighter RSGCs and they seem to be embedded in more extended regions of massive young stars (Negueruela et al. 2010, 2011).

For the new cluster candidates F 3 and F 4 from Froebrich (2013), only features that look like a tentative MS in the CMDs are found (see Fig. D2 in Appendix D). For F 3, a clump of stars at K = 13 mag and J − K = 2.4 mag can be identified, while for F 4 a more MS like feature can be seen at J − K = 2.6 mag. Both of these features contain a few hundred high-probability members. However, when utilizing the CCDs for both cluster candidates, one can identify that these features are not caused by MS stars. The high-probability members clearly are not situated near the bottom of the reddening band, indicating that they are giants. Only in the case of F 3, there are some potential MS stars. Thus, both candidates are most likely holes in the general extinction and not real clusters.

3.2 MS properties

Here, we will concentrate on determining the principal properties of the detected MS for RSGC 1–3. The isochrones overplotted to the CMDs and CCDs use age and distance estimates from the literature (see Table 2 for these parameters). We adopt a distance of 6 kpc for all clusters, which seems an appropriate average of the published values (Figer et al. 2006; Davies et al. 2007, 2008; Alexander et al. 2009; Clark et al. 2009). The ages are taken for each individual cluster from the same references. We only vary the extinction in the K band to shift the isochrone on to the detected MS. Since the upper end of the MS is almost vertical in the K versus J − K CMDs, the actual choice of age and distance will not influence the required extinction value.

All MS are ‘vertical’ in the CMDs for the top 2–4 mag in the K band. Hence, these are clearly massive MS stars and we can try to estimate the colour excess towards the cluster by shifting an isochrone until it fits the MS. This will determine the extinction towards potential cluster members, i.e. the column density of material along the line of sight that is not associated with the cluster itself and is independent of the actual age chosen for the isochrone. Utilizing an extinction law (we use Indebetouw et al. 2005), we can convert this to a foreground extinction value. These are the values we used to overplot the isochrones in Fig. 1 and which are listed as |$A_K^{{\rm MS}}$| in Table 2. In the CCDs in the right-hand panels of Fig. 1, one can see that the reddening direction of the background giants confirms the validity of this reddening law. In essence, the reddening law in these fields is in agreement with Indebetouw et al. (2005) or Stead & Hoare (2009), but using a less steep dependence of extinction on wavelength such as in Mathis (1990) or Rieke & Lebofsky (1985) can be ruled out from the CCDs. Thus, please note that the use of an extinction law in agreement with the CCDs will change the inferred K-band extinction by only about 0.1 mag. However, larger differences are expected when a less steep extinction law is applied. The cluster RSGC 1 has the largest extinction of 2.4 mag in the K band, while the other two clusters have about half this value of reddening, i.e. A_K ≈ 1.2 mag. Compared to the literature values (listed as |$A_K^{{\rm Lit}}$| in Table 2), our isochrone fit to the MS systematically finds smaller extinction values. This could be caused by the fact that: (i) the literature extinction values are estimated using a different extinction law (such as by Davies et al. 2007 for RSGC 2); (ii) the observations are taken in different filters, e.g. 2MASS K_S versus UKIDSS K; (iii) the extinction values are determined from the spectral types of the RSGs in the cluster and not by the better understood MS stars. Recently, Davies et al. (2013) have shown that there can indeed be issues with the RSG temperature scale.

We further estimate the amount of material associated with the cluster itself. This can be done by measuring the width of the MS in J − K and convert this to a value of differential reddening. We list these values in the ΔA_K column in Table 2. As for the general interstellar extinction, RSGC 1 shows the highest amount of differential reddening with about 0.75 mag in the K band. This is about three times as high as for the other two clusters, but comparable to, or even smaller than for other young embedded clusters (e.g. the Orion nebula Cluster; Scandariato et al. 2011). There are several possible explanations for the differences. (i) This cluster is younger and thus still more deeply embedded in its parental molecular cloud – however, this is unlikely given the age of the cluster. (ii) The differential reddening is not caused by intrinsic dust, but by the variations in extinction of the foreground material. The larger value for |$A^{{\rm MS}}_K$| for this cluster could support this.

In order to investigate the number of potential MS stars in each cluster, we defined a colour range in J − K for each cluster (as indicated in the CMDs in Fig. 1) that encloses all the potential MS stars. We select all stars within this colour range and determine the K-band luminosity function along the MS. These are shown as dotted blue lines in Fig. E1 in Appendix E. If we only count the membership probabilities |$P^i_{\rm cl}$| for each star, then we obtain a more realistic luminosity function which is shown as solid red line in Fig. E1. The smallest difference between the two luminosity functions is evident for RSGC 2. Thus, this is the cluster where the MS stands out most significantly from the field stars. This is in agreement with the fact that this is the only cluster where the MS has been detected previously (Froebrich 2013). The typical membership probability for the MS cluster members is about 80 per cent for RSGC 2, while it is of the order of 60 per cent for the other two clusters.

We investigate the brightest MS stars in each cluster and define |$K^{3}_{{\rm MS}}$| as the apparent K-band magnitude where there are at least three stars per 0.5 mag bin along the MS. In other words, we treat this as the top of the MS. Similarly we define |$K^{10}_{{\rm MS}}$| and list both values in Table 2. The most populated cluster MS at bright K-band magnitudes occurs in RSGC 2. There, |$K^{10}_{{\rm MS}}$| = 11 mag, while |$K^{3}_{{\rm MS}}$| is one magnitude brighter. RSGC 3 has by far the fewest bright cluster MS stars, or the MS starts only at fainter magnitudes. The absolute magnitudes for |$K^{3}_{{\rm MS}}$| and |$K^{10}_{{\rm MS}}$| are determined from our adopted distance and de-reddened with |$A_K^{{\rm MS}}$|⁠. They are listed in brackets in Table 2 and show that RSGC 1 has the brightest (M_K = −5.9 mag) end of the MS while RSGC 3 has intrinsically the faintest end of the MS (M_K = −3.7 mag). This is in good agreement with the ages for the clusters determined in the literature, which range from 8–12 Myr for RSGC 1 (Figer et al. 2006; Davies et al. 2008) to about 20 Myr for RSGC 3 (Alexander et al. 2009). Note that according to the isochrones used (Lejeune & Schaerer 2001), stars with a mass above 20 M_⊙ (or O-type stars) have M_K = −2.9 mag or brighter on the MS. Thus, in particular RSGC 1 and RSGC 2 could contain a significant number of massive MS or post-MS objects that can easily be verified spectroscopically. Note that the total number of these cluster members is likely to be larger by about 50 per cent, since we have only analysed the stars within one cluster radius. There are spectroscopically confirmed cluster members in the region between one or two cluster radii; typically only about two-thirds of the known members are within one cluster radius.

The completeness limit determined as the peak of the K-band luminosity function, for all clusters is between K = 15 and 16 mag. In all cases, this is fainter than stars of about 8 M_⊙ which have M_K = −0.9 mag (Lejeune & Schaerer 2001) or an apparent magnitude of K = 13.0 mag at our adopted distance and without considering extinction. We thus can compare the total number of stars along the MS, brighter than these stars. The numbers are weighted by the membership probabilities and are listed as |$N_{{\rm MS}}^{>8\,\mathrm{M}_{\odot }}$| in Table 2. We find that RSGC 1 has about twice as many of these OB-type stars than RSGC 2 and 3. Please note that we expect increased crowding in the cluster centres and thus the estimated OB-type cluster member numbers should be treated as lower limits. Since RSGC 1 is the most compact of the three clusters, its numbers should be most affected. The number of confirmed RSGs (see column N_RSG in Table 2) in RSGC 2 is much higher than in RSGC 1, which might be due to the lower age of the latter. Hence, out of the three clusters, RSGC 1 seems to be the most massive object as estimated in Figer et al. (2006) and Davies et al. (2008). Based on the number of stars along the MS and the age, RSGC 2 and 3 might be less massive, partly (for RSGC 3) in agreement with the predictions from Alexander et al. (2009) and at the lower end of the mass range suggested in Clark et al. (2009).

In Fig. 2, we show the spatial distribution of all stars along the MS for the three clusters. Only stars within the cluster radius are shown, since we have not determined membership probabilities outside this area. In all three cases, the most likely MS stars are concentrated towards the nominal centre of the clusters. However, the distribution seems not to be centrally condensed, but rather filamentary, especially for RSCG 2. If this is a real effect, or caused by crowding in the cluster centre and ‘missing’ objects near the bright RSGs is unclear. The spatial density of the MS cluster members (each star is weighted by its membership probability) is between two and three times higher in the cluster centre compared to the outer regions as defined by the cluster radius.

Figure 2.

Open in new tabDownload slide

Positions of the stars in RSGC 1 (top), RSGC 2 (middle) and RSGC 3 (bottom). Colour coding is the same as in Fig. 1. Only stars which are suspected MS stars (objects in the blue dotted boxes in the CMDs in Fig. 1) are plotted as coloured symbols while the remaining stars in the field are represented by black dots and confirmed RSGs are shown as black filled squares.

4 CONCLUSIONS

We have used PSF photometry on deep NIR JHK imaging data from the UKIDSS GPS to investigate the fields of known and candidate RSG clusters. We confirm the detection by Froebrich (2013) of the upper MS of the cluster RSGC 2 and for the first time detect the MS for RSGC 1 and RSGC 3.

We use the age and distance estimates from the literature to overplot isochrones on the NIR CMDs for all clusters in order to establish the reddening of the MS by utilizing the reddening law from Indebetouw et al. (2005). In all cases, the inferred K-band extinction values for the MS are smaller than the quoted values in the literature, which are determined for the RSG stars. We also infer the differential reddening towards each cluster based on the width of the detected MS. The youngest of the clusters (RSGC 1) has the highest extinction and differential reddening in accordance with its evolutionary status. It also contains the most number of stars (about 200) with masses above 8 M_⊙.

The spatial distribution of the candidate MS stars in all clusters shows a concentration towards the nominal cluster centre. However, there is no indication of a centrally condensed distribution, which could either be real or caused by increased crowding and blending effects from the bright RSG cluster members.

We also investigated fields near the clusters RSGC 4 and RSGC 5, as well as the candidate RSG clusters F 3 and F 4 from Froebrich (2013). In all cases, no MS could be detected. In the case of the already known clusters, this could be caused by them having less members or being spatially more extended, i.e. more association like. Our results indicate that the new candidates can most likely be interpreted as holes in the background extinction.

We would like to thank I. Negueruela and C. Gonzalez for fruitful discussions during the earlier stages of the project. We further acknowledge the constructive comments by the referee B. Davies which helped to improve the paper. Part of this work was funded by the Science Foundation Ireland through grant no. 10/RFP/AST2780.

REFERENCES

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article:

Appendix A: Notes on individual clusters.

Appendix B: Cluster images.

Appendix C: Cluster and control fields.

Appendix D: RSGC 4, 5 and F 3, 4.

Appendix E:K-Band luminosity functions (Supplementary Data).

Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper.