We present a VI photometric catalogue towards the open cluster Cr 121. XMM–Newton and ROSAT data are used to discover a low-mass pre-main sequence (PMS) along this sightline. de Zeeuw et al. have identified Cr 121 as a moving group, using Hipparcos data, at a distance of 592 pc; we reject the scenario that these low-mass PMS stars are associated with that association. By considering the higher mass main sequence stellar membership of the groups along this sightline, the density of low-mass PMS stars and their age spread we argue that the low-mass PMS stars are associated with a young, compact cluster at a distance of 1050 pc. This is consistent with Collinder's original description of Cr 121, and we argue that this distant compact cluster should retain its original designation. The moving group detected by de Zeeuw et al. resembles a foreground association and we agree with Eggen that this should be called CMa OB2.
This study demonstrates that although the de Zeeuw et al. census of OB associations is an invaluable resource for studying local star formation, it must be interpreted in the context of other data when considering structure over distances of the same order as the limits of the Hipparcos parallaxes.
Collinder 121 (Cr 121) is listed in the New List of OB Associations (Melnik & Efremov 1997) as an OB association at a distance of 540 pc. However, since its first identification as an open cluster of about 1° diameter by Collinder (1931), Cr 121 has had its membership re-assessed a number of times. In this paper we use the low-mass pre-main sequence (PMS) to resolve the long-standing controversy over the nature of Cr 121, and demonstrate the power of the using the low-mass PMS to pick apart structure in OB associations. We also illustrate the difficulties associated with using proper motion and parallax data of limited range in isolation.
The cluster was first discovered as a group of 18–20 main-sequence B stars found within a 40 × 60 arcmin2 box. Collinder (1931) used the cluster diameter, along with the mean separation of stars within it, to derive a distance of 1260 pc. Feinstein (1967) studied the cluster using UBV photometry and extended the membership to 40 stars brighter than V = 7 within a 10°× 10° box. This group was characterized as an OB association and the position of the zero-age main sequence (ZAMS) indicated a distance of 630 pc. Already, two very different structures had been described by different authors. Eggen (1981) suggested that the original compact group and the larger OB association may be distinct. Using intermediate-band and Hβ photometry Eggen (1981) placed the compact group, which he referred to as Cr 121, at a distance of 1.17 kpc at an age of 1.5 Myr. The more diffuse group, which Eggen referred to as CMa OB2, was found to be at a distance of 740 pc.
More recently, emphasis has been placed on the use of common motion criteria for defining cluster and association memberships. de Zeeuw et al. (1999) used Hipparcos proper motion and parallax data to select a moving group in the direction of Cr 121. They selected 103 stars in a region 13°× 16° in extent, and find a mean distance of 592 ± 28 pc. The presence of an O star, a Wolf–Rayet (WR) star and early-type B stars in the membership list led to an estimated age of ∼5 Myr. Since stars were selected from a large region of sky compared with the original boundaries of the distant interpretation of Cr 121, and the fact the Hipparcos parallaxes become unreliable at about 1 kpc, this selection method is biased towards detecting a larger, more diffuse foreground association rather than the distant compact cluster of Collinder (1931). If both groups were present, as suggested by Eggen (1981), only the foreground association would be detected as this would represent a large number of stars with common motion at a similar distance, within the reliable domain of Hipparcos parallaxes. Dias, Lépine & Alessi (2001) used data from the Tycho2 catalogue to redetermine the mean proper motion of Cr 121, amongst other clusters. This study was restricted to open clusters within 1 kpc, so again was biased towards only detecting the foreground group, and would have been insensitive to any more distant cluster along the same line of sight. Apparently working within the assumption of the de Zeeuw et al. (1999) view of Cr 121 and the Dias et al. (2001) membership list, Dias et al. (2002) determine the distance to Cr 121 as 470 pc with an age of 11 Myr.
Kaltcheva (2000) carried out Strömgren and Hβ photometry of bright B stars within a 5° radius of the classical centre ) of Cr 121. This revealed a group of stars at 660–730 pc, with characteristics similar to an OB association, and a more compact group of stars at a distance of 1085 pc, which she calls Cr 121. This would indicate that the original cluster Cr 121 is situated at a distance of over 1 kpc, whilst the Cr 121 selected by de Zeeuw et al. (1999) is a less distant OB association.
Inspection of the ROSAT Position Sensitive Proportional Counter (PSPC) catalogue revealed a number of X-ray point sources in the vicinity of WR 6, near the centre of the region studied by Kaltcheva (2000) and the original centre of Collinder's cluster. WR 6 is listed as a member of Cr 121 by de Zeeuw et al. (1999), but not by Kaltcheva (2000). The Hipparcos parallax for WR 6 of π= 1.74 ± 0.76 mas, puts it at a distance of 575 pc.
Low-mass PMS stars are expected to be X-ray bright owing to high levels of coronal activity, as such, X-ray data has been used as a diagnostic tool for identifying low-mass PMS stars (e.g. Naylor & Fabian 1999; Pozzo et al. 2000). The presence of low-mass PMS stars would allow us to fit isochrones to the PMS in a colour–magnitude diagram (CMD), constraining the distance and age of the PMS. With this in mind we have obtained deep VI imaging photometry of a region 20 arcmin in radius, covering the central region of the ROSAT PSPC field of view (FoV) for the pointings carried out towards WR 6. Since the optical data were obtained, XMM–Newton data for this region of sky have become available, allowing us to perform a more sensitive X-ray selection of PMS stars.
The layout of our optical survey is shown in Fig. 1, along with the location of the B stars identified as being members of the more distant group by Kaltcheva (2000) and WR 6. Although our survey region coincides with only two of the B stars identified by Kaltcheva (2000) as members of the more distant cluster, and lies towards the edge of the region enclosed by them, we would still expect to find a significant number of low-mass PMS stars from that group within our survey. Studies of other young clusters and OB associations have always revealed O and B stars located within a sea of low-mass PMS stars, not low-mass PMS stars only within a region enclosed by higher mass stars (e.g. Dolan & Mathieu 2001; Pozzo et al. 2003).
The VI CCD imaging of 8 13.5-arcmin FoVs centred on WR 6 was carried out using the 0.9-m telescope at the Cerro Tololo Interamerican Observatory (CTIO) in Chile on the night of 2002 February 12/13.
The FoVs were selected to cover the central region of the ROSAT PSPC pointings. Each field was observed with short (10, 6 s) and long exposures (300, 180 s) in the V and I bands in photometric conditions. The short exposures were aimed at capturing the brighter stars in each field, which would be saturated in the long exposures. To reduce the number of cosmic ray hits in any one image the long exposures were split into three parts, of 100 s in V and 60 s in I. The fields were overlapped with each other to provide a check for the internal consistency of the photometry (see Section 2.2). A number of Landolt (1992) standard star fields, containing standards with −0.6 ≤V−I≤ 5.8, were also observed at several times during the night. This ensured adequate calibration over the colour range of interest.
Data reduction and optimal photometry
The data were reduced, and photometry was performed using the optimal extraction algorithm originally described by Naylor (1998) with its application for constructing CMDs described in depth by Naylor et al. (2002). The standard star fields were reduced using the same algorithm. The instrumental magnitudes and colours for 116 standard star observations were compared with those given by Landolt (1992) to yield zero-points, colour terms and extinction coefficients. We found that an additional, magnitude independent, uncertainty of 0.02 mag was required to give a χ2 of 116 for the V-band calibration. No such adjustment was required for V−I.
Each image was bias subtracted and flat-field corrected, using files constructed from twilight sky flat-field images and bias frames taken at the beginning of the same evening as the observations. The I-band images were combined (after determining their offsets) before using the deep coadded image for object detection. Optimal photometry was then carried out on the list of stars produced by the detection software to produce an optical catalogue for each FoV. Although the I-band images were combined for object detection, photometry was carried out on the individual frames. This ensures that good signal-to-noise ratio in one frame is not swamped by poor signal-to-noise ratio in another when the images are combined. The individual measurements of each star are combined later, weighting each measurement according to its signal-to-noise ratio.
An astrometric solution was obtained through comparison of the optical catalogues with SuperCosmos catalogues (Hambly et al. 2001a) of the same FoVs. The root-mean squared (RMS) residuals of the 6-coefficient fit were approximately 0.2 arcsec. The resulting catalogues were then combined to produce an optical catalogue for the entire region covered by our survey, listing a position, a V magnitude, a V−I colour and a data quality flag (see Naylor et al. 2002) for 26 104 stars. The overlap regions between the fields were used to assess the internal consistency of the optimal photometry, which was found to be consistent at the 0.005-mag level. The final catalogue is given as Table 1. Only a sample is provided in the text; the full table is available in the online version of the journal on Synergy, and also via the CDS online data base. The reduction carried out here differs from that described by Naylor et al. (2002) in that an additional data quality flag has been introduced, and the flags are now indicated by character values rather than integers, since this gives greater flexibility for the format. There is also an entirely new flag, ‘I’, which indicates that the star's photometry is unreliable owing to ill-determined sky. The threshold for this flag to be invoked is when the fit to the sky histogram for the star has a reduced χ2 > 3. The resulting CMD is shown below in Fig. 2. A population of PMS stars is clearly visible on the CMD of the optical catalogue, lying redward of the background contamination.
Nine pointings of the ROSAT PSPC were made towards WR 6 between 1991 and 1992, totalling 28 ks. The data set for each pointing was retrieved from the public archives via the LEDAS website () hosted by Leicester University. The data were reduced using Asterix version 2.3-b1, a suite of X-ray data reduction programs available from the University of Birmingham (). The data were sorted into 20-arcmin radius image cylinders (spectral data was retained at this stage), excluding events recorded during periods of high background or poor aspect quality and selecting corrected pulse height (PHA) channels 11 to 201. The resulting images were then visually inspected and obvious sources masked in duplicate images, which were then used to create a background image cylinder for each pointing. These were then used, along with other calibration files supplied with the data, to create a background model for each pointing. These, and the unmasked image cylinders were then projected to create 2D images and background models, which were used for source detection using Cash-statistic maximization in the PSS algorithm (Saxton, Ashley & Vallance 2000). The resulting source lists were then used to further mask sources in the background image cylinders to produce refined background models. The refined background models and the images were then coadded. This resulted in an image for source detection with an effective exposure time of 25.8 ks. This image was then searched for sources using PSS returning 39 X-ray sources. The archival release of XMM–Newton observations of this region of sky means that much of the data in this source list has been superseded in sensitivity and positional accuracy. As a result we retrieved the XMM–Newton data and removed from the ROSAT PSPC source list those objects that appeared in both sets of data.
Data from two XMM–Newton observations using the European Photon Imaging Camera (EPIC) were used to make the X-ray source catalogue. One directed at WR 6 with 12 864 s total integration time during revolution 346 (Skinner et al. 2002), and another of 52 496 s directed at the north-west quadrant of the ring nebula S308 carried out during revolution 343 (unpublished). Skinner et al. (2002) investigated the possible existence of a close companion to WR 6. They make note of one other source in their FoV, 57.4 arcmin south of WR 6, which they fail to correlate with a SIMBAD counterpart. This source correlates with a star in our photometric PMS selection (field 1, star 73, see Table 5 later). Other than this they do not discuss the other sources in the FoV. The final EPIC source lists were retrieved from the ESA XMM–Newton Science Data Archive () and filtered to remove sources with detection likelihood values of less than 20. This was to allow for an error in the calculation of detection likelihoods by the XMM–SAS pipeline, in versions 5.4.1 and earlier (see XMM–Newton news no. 29). The two source lists were then cross-correlated with one another to remove duplicates. In each case of a duplicate source the observation with best signal-to-noise ratio was used. The source list was then further restricted to cover just the region covered by the optical survey. The resulting source list contained 138 sources, and is given as Table 2. This table is available in full in the online version of the journal in Synergy, and also via the CDS online data base. Listed for each source are the XMM–Newton source number, right ascension, declination, error in position, count rate and error in count rate. For those X-ray sources that correlate with an optical counterpart (see Section 4) the field number and star index are also listed.
Of the 39 ROSAT PSPC sources, 13 lay within the optical survey region and did not appear in the XMM–Newton source list. We will refer to this set of sources as the reduced ROSAT PSPC source list. This source list is given as Table 3, and is available via the CDS data base. For each source we list index number, right ascension, declination and flux. Again, for those sources that correlate with an optical counterpart we also list the field number and star index.
Identification of X-Ray Sources
Cross-correlation of optical and X-ray source ists
XMM—Newton source list and the reduced ROSAT PSPC source list were cross-correlated against the optical catalogue. XMM–Newton sources were matched using a search radius of 5 arcsec, except in cases where the error in the position given in the EPIC source list was greater than this. In these cases the radius of the position error circle given in the XMM–Newton data was used. Since the accuracy of their positions was poorer, a search radius of 10 arcsec was used to match the ROSAT sources. A boresight correction of +3 arcsec in declination was applied to the ROSAT coordinates. This was based on the offset between the coordinates of the brightest X-ray source in the original ROSAT PSPC source list and the coordinates of WR 6 in the optical catalogue. Of the 151 X-ray sources in optical survey region (138 XMM–Newton sources and 13 ROSAT PSPC sources), 103 correlate with stars in the optical catalogue (91 XMM–Newton sources and 12 ROSAT PSPC sources). The positions in colour–magnitude space of the stars in the optical catalogue with X-ray correlations are shown in Fig. 3. We have excluded stars with data quality flags other than ‘OO’, and those with signal-to-noise ratio less than 10. It is clear from this plot that the majority of the correlated stars lie the region of the diagram occupied by the PMS. In fact, of the 68 sources plotted in Fig. 3, 43 lie in the PMS region of the diagram. The inferred X-ray fluxes for these objects are consistent with their status as PMS stars (Feigelson et al. 1993).
We estimated how many chance correlations we would expect to find in different regions of the CMD by performing the cross-correlation procedure for 8 different offsets of 30 arcsec in each axis between the catalogues. We find that we would expect up to 5 of the correlations in the PMS region of the CMD to spurious, whilst essentially all of the correlations in the background region would be spurious.
Investigation of the proper motions of the X-ray selected PMS candidates
To remove objects that were not members of the group of interest, we refined our selection to remove objects that do not share common motion. We obtained SuperCosmos proper motions for 52 stars in our X-ray selected PMS (we did not exclude stars with poor photometry flags or low signal-to-noise ratio for this experiment). A mean proper motion was calculated from the data, with each star being given a weight inversely proportional to the square of its error. Stars which contributed a χ2 > 4 were removed from the selection and the weighted mean recalculated. This was repeated until no stars in the sample contributed a χ2 above this value. In total 17 stars were removed from the sample, and these are marked on Figs 5 and 6 with faint circles. A value of χ2 > 4 was chosen as the deselection criterion since, for a sample of 52 stars, a 2σ clip would be expected to remove 2 or 3 bonafide members of the group. It is thought that all of the stars removed from the sample were non-members. The distribution of the proper motions for the final selection are shown in Fig. 4. The width of the distribution is consistent with the mean errors and an internal velocity dispersion which is close to zero. Those objects that were classified as PMS members of an association along this sightline by both photometric and proper motion criteria are listed in Table 4. Those that were deselected based on their proper motions are listed in Table 5.
An interesting experiment would be to compare the proper motions of the confirmed members of this group of PMS stars with those measured by other surveys for stars along this sightline. Unfortunately the systematic error in the SuperCosmos proper motions is large at galactic latitudes |b| ≤ 30° (Hambly et al. 2001b). As such, SuperCosmos proper motions for stars within our survey region with b≈−10° will have a large systematic error. This is both magnitude and survey plate dependant.
We have assessed this issue by calculating the weighted mean of the SuperCosmos proper motions for a broad selection of background field stars in our survey. The proper motion we find is (μα cos δ, μδ) = (2.43 ± 0.26, −2.07 ± 0.27) mas yr−1. We calculate that the proper motion that would be expected due to galactic rotation and solar reflex for stars at 1 kpc distance in the direction of our survey should be approximately (μα cos δ, μδ) = (−2, 1) mas yr−1. Since, for a random selection of field stars, one would expect the mean proper motion to coincide with the proper motion of the LSR and solar reflex, it is clear that whilst the SuperCosmos proper motions are sufficiently internally consistent for crude membership selection, they are not suitable for drawing comparisons with proper motions obtained from other surveys.
Discussion: how far away and how old are The pms stars?
The fitting of theoretical isochrones to the observed PMS can yield distances and ages for stars found there. However there is a degeneracy between distance and age for low-mass stars. As such any estimates of distance and age must be made within the context of the estimates of the distances to the higher mass stars, as well as the constraints that higher mass stars place on the age of any associations that may be present.
Isochrones in V/V−I were generated from the solar metallicity models of D'Antona & Mazzitelli (1997) in the same manner as described by Jeffries, Thurston & Hambly (2001). Empirical colour–Teff conversions were obtained by fitting a 120-Myr isochrone to the Pleiades at a distance modulus of 5.6, E(B−V) = 0.04, E(V−Ic) = 0.05.
We scaled the resultant isochrones for a distance of 600 pc, in keeping with the distance derived for Cr 121 by de Zeeuw et al. (1999) of 592 ± 28 pc. In addition to shifting the isochrones for distance, they were also shifted for an extinction of AV = 0.17 (Kaltcheva 2000). This corresponds to a reddening of E(V−I) = 0.07 (if we assume a galactic reddening law), though this has little effect on the results as the reddening vector lies along the isochrones. If the low-mass PMS stars are located within the association at the distance found from the Hipparcos data, the best fitting isochrones are those corresponding to ages between 20 and 50 Myr (see Fig. 5). These imply a median age of 30 Myr, with an age spread of about 30 Myr. Such an age spread is not considered likely for an association of this age (Jeffries & Tolley 1998, and references therein). Additionally, as was noted by de Zeeuw et al. (1999), the presence of an O star and early B stars indicates that this moving group is young, about 5 Myr in age. For such a large number of PMS stars to be found within such a small survey region would also be surprising if they were associated with the group described by de Zeeuw et al. (1999). Assuming a Salpeter initial mass function (IMF), we calculate that we would expect to find 3900 0.20–0.73 M⊙ (14 ≤V≤ 20 at 592 pc) stars for the 85 B stars selected as members by de Zeeuw et al. (1999). If we assume that these would be distributed evenly over a region at least as large as the 13°× 16° region occupied by the B stars, then we would expect to find fewer than five 0.20–0.73 M⊙ stars within the region observed here. As such, the PMS found here would represent quite a concentration of the low-mass stars of the association. It should be noted that a more realistic IMF would predict even fewer stars in this mass range, since several observational studies have found that the IMF in young clusters flattens out at masses below ∼0.8 M⊙ (e.g. Preibisch et al. 2002; Briceño et al. 2002, and references therein).
The three lines of argument described above lead us to reject the scenario in which the low-mass PMS stars are associated with the moving group detected using the Hipparcos data. Instead, we use the presence of early-type main sequence B-stars in the membership lists for both interpretations of Cr 121 to constrain the age to less than 15 Myr. Such a low age is also more in keeping the scatter of stars in the PMS region of the diagram. Scaling sets of isochrones for ages of less than 15 Myr for distance until they fitted as much of the data as possible gave a distance of 1050 pc, and a modal age of between 1 and 10 Myr, applying the same extinction as before. This is consistent with the distance given by Kaltcheva (2000) for a compact group of stars at 1085 ± 41 pc. Fig. 6 shows that the age spread implied by fitting isochrones for this distance is less than 10 Myr, much more in keeping with age spreads observed in other young clusters and associations (e.g. Pozzo et al. 2003) than the 30 Myr age spread suggested by the fit for 600 pc. Isochrones for 1 and 20 Myr, scaled for the same distance as the 5 and 10 Myr ones, are included in Fig. 6 for comparison.
If we carry out a similar calculation to the one above, using a Salpeter IMF, we find that we would expect there to be around 540 stars in the 0.29–0.98 M⊙ mass range (14 ≤V≤ 20 at 1050 pc) for the 11 B-type members of the more distant cluster suggested by Kaltcheva (2000). If we assume an even distribution of stars over the ≈1° radius surveyed there, then we would expect to find, at most, 130 stars in this mass range in the region covered by our optical survey. This is consistent with the population of the PMS observed here. This would indicate that the PMS we have identified consists of stars in the ≈1 kpc distant cluster identified by Collinder (1931), Eggen (1981) and Kaltcheva (2000). To avoid confusion we shall refer to this distant cluster as Cr 121, and we shall refer to the association at approximately 600 pc as CMa OB2, following Eggen (1981).
It seems clear that the low-mass PMS stars detected here are associated with the compact group of stars found by Kaltcheva (2000), Eggen (1981) and Collinder (1931) at a distance of over 1 kpc, originally designated Cr 121. If this cluster were of the extent and at the distance found by de Zeeuw et al. (1999), the discovery of such a concentration of low-mass PMS stars which are so much older than the rest of the association, and with an age spread of 30 Myr, seems incredible. It is also clear however, that there is a young moving group that was detected using the Hipparcos data by de Zeeuw et al. (1999) at a distance of 592 ±28 pc, of which WR 6 is likely a member, in the same direction. The characteristics of the group described there, however, seem more in keeping with an OB association, than with the compact open cluster originally described as Cr 121. We argue that the distant open cluster should retain its original designation as Collinder 121, whilst the nearer OB association described by Feinstein (1967), de Zeeuw et al. (1999) and Dias et al. (2002) should be re-designated as CMa OB2, following Eggen (1981).
Whilst the Hipparcos census of OB associations within 1 kpc is an invaluable resource for studying recent local star formation, this work demonstrates the importance of interpreting proper-motion and parallax data within the context of age and distance constraints imposed by main-sequence high-mass and low-mass PMS stars. This is particularly true when investigating structures that extend beyond the range of the Hipparcos data.
We thank Nigel Hambly for useful comments regarding the use of proper motions from the SuperCosmos sky surveys, and Mike Watson for alerting us to the existence of the XMM–Newton data for WR 6 and S308. We also thank our referee Ronnie Hoogerwerf for many useful comments which improved this paper, and for alerting us to problems in our use of the SuperCosmos proper motions. The Cerro Tololo Interamerican Observatory is operated by the Association of Universities for Research in Astronomy, Inc., under contract to the US National Science Foundation. This work made use of ROSAT data obtained from the Leicester Data base Archive Service at the Department of Physics and Astronomy, Leicester University.