A&A 486, 453-466 (2008)
DOI: 10.1051/0004-6361:200809917

The outskirts of Cygnus OB2[*],[*]

F. Comerón1,[*] - A. Pasquali2 - F. Figueras3 - J. Torra3

1 - European Southern Observatory, Karl-Schwarzschild-Strasse 2, 85748 Garching, Germany
2 - Max-Planck-Institut für Astronomie, Königstuhl 17, 69117 Heidelberg, Germany
3 - Departament d'Astronomia i Meteorologia, Universitat de Barcelona, 08028 Barcelona, Spain

Received 7 April 2008 / Accepted 15 May 2008

Context. Cygnus OB2 is one of the richest OB associations in the local Galaxy, and is located in a vast complex containing several other associations, clusters, molecular clouds, and HII regions. However, the stellar content of Cygnus OB2 and its surroundings remains rather poorly known largely due to the considerable reddening in its direction at visible wavelength.
Aims. We investigate the possible existence of an extended halo of early-type stars around Cygnus OB2, which is hinted at by near-infrared color-color diagrams, and its relationship to Cygnus OB2 itself, as well as to the nearby association Cygnus OB9 and to the star forming regions in the Cygnus X North complex.
Methods. Candidate selection is made with photometry in the 2MASS all-sky point source catalog. The early-type nature of the selected candidates is confirmed or discarded through our infrared spectroscopy at low resolution. In addition, spectral classifications in the visible are presented for many lightly-reddened stars.
Results. A total of 96 early-type stars are identified in the targeted region, which amounts to nearly half of the observed sample. Most of them have featureless near-infrared spectra as expected from OB stars at the available resolution. Another 18 stars that display Brackett emission lines can be divided between evolved massive stars (most likely Be stars) and Herbig Ae/Be stars based on their infrared excesses. A component associated with Cygnus OB9/NGC 6910 is clearly identified, as well as an enhancement in the surface density of early-type stars at Cygnus X North. We also find a field population, consisting largely of early B giants and supergiants, which is probably the same as identified in recent studies of the inner $1^\circ $ circle around Cygnus OB2. The age and large extension of this population discards a direct relationship with Cygnus OB2 or any other particular association.
Conclusions. Earlier claims of the possible large extent of Cygnus OB2 beyond its central, very massive aggregate seem to be dismissed by our findings. The existence of a nearly ubiquitous population of evolved stars with massive precursors suggests a massive star formation history in Cygnus having started long before the formation of the currently observed OB associations in the region.

Key words: stars: early-type - Galaxy: open clusters and associations: individual: Cygnus OB2 - Galaxy: open clusters and associations: individual: Cygnus OB9

1 Introduction

Cygnus OB2 is one of the richest OB associations known in our Galaxy, and it is the most nearby one of its kind at less than 2 kpc from the Sun (Knödlseder 2003). It is also a unique resource for the observational study of the upper end of the stellar mass function (e.g., Herrero et al. 1999): O stars, Wolf-Rayet (WR) stars, Luminous Blue Variables (LBVs), B[e] stars... are all classes with representatives in Cygnus OB2. This association thus provides, in a single complex, the most complete and accessible showcase of the variety found among the hottest and most massive stars.

UBV photometry reveals the presence of hundreds of early-type, heavily-obscured stars in Cygnus OB2 (Massey & Thompson 1991; Kiminki et al. 2007). However, a more real measure of its richness is given by star counts in the near infrared, where the large, patchy foreground extinction in the direction of the association is greatly reduced. Using this technique, Knödlseder (2000) has inferred a content of $\sim$2600 OB stars, including over 100 O-type stars and evolved stars having massive progenitors. Support for this estimate has been provided by Comerón et al. (2002) using near-infrared low resolution spectroscopy. High signal-to-noise ratio spectra in the visible of the least obscured objects of this sample carried out by Hanson (2003) has confirmed the classification of the vast majority of the early-type star candidates listed by Comerón et al. (2002). However, it has also shown that a large fraction of them are giant and supergiant B stars rather than O stars. This suggests a picture more complex than that of a single, recent starburst less than 3 Myr ago giving rise to the association (Massey & Thompson 1991). It also casts doubts on the actual membership of these stars in Cygnus OB2, given their older ages and their extended spatial distribution when compared to the strongly-clustered, earliest-type stars at the center of the association. Since the precursors of the extended population dominated by early B-type giants and supergiants must have been O-type stars, the region must have been producing massive stars long before the formation of the major young clusters and associations currently observed there.

We investigate the nature of this population and its possible relationship with Cygnus OB2 by studying the early-type stellar content in a region lying between $1^\circ $ an $2^\circ $from the center of the association. Earlier results of this work have already revealed the existence of interesting objects possibly related to Cygnus OB2 well beyond the boundaries of the central cluster. Examples are WR 142a, a WC8 Wolf-Rayet star about $1.3^\circ$ from the center of Cygnus OB2 (Pasquali et al. 2002), and the very massive O4If runaway star BD $+43^\circ 3654$ more than two degrees away (Comerón & Pasquali 2007). The study of the outskirts of Cygnus OB2 is also motivated by its overlap with other structures tracing recent star formation, most notably the Cygnus OB9 association and several compact HII regions and embedded clusters belonging to the Cygnus X molecular cloud complex (Schneider et al. 2006).

We use the same technique based on 2MASS JHK colors used in Comerón et al. (2002) to identify OB stars in the targeted region. Infrared spectroscopy allows us to identify new young intermediate-mass stars and evolved massive stars, both characterized by their emission in Br$\gamma $, as well as other likely early-type stars without noticeable emission lines. We also provide spectral types of many of these objects whose relatively light obscuration makes them accessible to spectral classification in the blue. This paper can thus be regarded as a continuation of the previous work of Comerón et al. (2002) & Hanson (2003), now probing the largely uncharted region beyond the limits of Cygnus OB2 explored thus far. The reader is referred to the recent review by Reipurth & Schneider (2008) for a detailed description of the current state of knowledge of the vast extent of Cygnus star forming complexes, and, in particular, of the Cygnus OB2 association and the Cygnus X region.

2 Target selection

The use of the (J-H), $(H-K_{\rm S})$ diagram to distinguish distant early-type stars from late-type evolved stars is founded on the difference in intrinsic colors between both populations. The interstellar reddening law keeps the loci occupied by bright, reddened early- and late-type stars separated by a gap of $\sim$0.4 mag in (J-H) at any given $(H-K_{\rm S})$ color, in the sense of the (J-H) color being bluer; see, e.g., Comerón & Pasquali (2005) for an application of this technique to the identification of the star responsible for the ionization of the North America and Pelican nebulae. The location of a star along the band defined by the reddened colors of early-type stars obscured by different amounts is not a proof of its early spectral type, since a certain degree of contamination of this locus is caused by evolved stars with particular photometric properties. Nevertheless, the signature of a rich association such as Cygnus OB2 in the (J-H), $(H-K_{\rm S})$ diagram is clearly indicated by the presence of an enhancement of the population of this locus (see, e.g., Fig. 2 of Comerón et al. 2002). This makes the early-type population easily identifiable even in regions where its areal density is far below that of cool giants, thus extending the sensitivity threshold to the early-type population of the association well beyond the $\simeq$$1^\circ $ radius where it stands out in infrared starcounts.

\par\includegraphics[width=8.5cm]{9917fig1.eps}\end{figure} Figure 1: (J-H), (H-K) diagram of the field surrounding Cygnus OB2 between $1^\circ $ and $2^\circ $ from the approximate center of the association, at $\alpha (2000) = 20^{\rm h}32^{\rm m}26^{\rm s}$, $\delta (2000) = +40^\circ 52'$. The plotted magnitudes are from the 2MASS catalog. Only objects brighter than $K_{\rm S} = 10$ are included. Candidate early-type stars are located along a band that has its origin near (J-H) = 0, (H-K) = 0 and traces the reddening vector.
Open with DEXTER

\par\includegraphics[width=7cm,clip]{9917fig2.eps}\end{figure} Figure 2: Typical examples of suspected reddened, early-type stars. These stars appear featureless at the resolution and signal-to-noise ratio of our near-infrared spectra.
Open with DEXTER

The presence of early-type stars beyond $1^\circ $ from the center of Cygnus OB2 is clearly indicated in Fig. 1, where we plot the 2MASS (J-H), $(H-K_{\rm S})$ diagram of point sources located between $1^\circ $ and $2^\circ $ from the approximate center of Cygnus OB2 at $\alpha (2000) = 20^{\rm h}32^{\rm m}26^{\rm s}$, $\delta (2000) = +40^\circ 52'$. A band of objects with varying reddening runs along the locus of early-type stars, well separated from the dominant band defined by cool background stars, which runs above it. It is interesting to note that the number of candidate OB stars in the region between $1^\circ $ and $2^\circ $ from the center of Cygnus OB2 is comparable to that found within $1^\circ $ from the center. In addition, the spread in extinctions is larger than in the central region, indicating that many OB stars may be very obscured and thus inaccessible to survey in the visible.

To confirm the nature of individual stars lying in the early-type locus described above and their likely membership in the association, we obtained low-resolution near-infrared spectroscopy of 232 stars in the region between $1^\circ $ and $2^\circ $ from the center of the association. The selected stars have $(J-H)-1.70
(H-K_{\rm S}) < 0.20$, implying a closer position to the locus of early-type stars than to that of late-type stars; and $K_{\rm S} < 8.8$, allowing us to obtain, with a reasonable investment of observing time, a large number of K-band spectra. We also obtained classification-quality spectra in the blue of 30 of the stars with near-infrared spectra consistent with early types and membership in the association, and whose foreground obscuration was relatively low as judged from their B magnitude listed in the USNO-B catalog (Monet et al. 2003).

We adopt here the rather loose denomination ``Cygnus OB2 halo members'' to refer to early-type stars and young stellar objects located in the area under discussion, and suspected to lie at the same distance as the association. This is not meant to indicate confirmed membership in the Cygnus OB2 association. Moreover, some of the stars we discuss are actually located within the boundaries of Cygnus OB9 (including some members of the open cluster NGC 6910 belonging to that association) or are associated with star forming regions in Cygnus X like DR 17 or DR 21 that are normally not considered to be a part of Cygnus OB2.

Hanson (2003) has produced a detailed discussion on the distance to Cygnus OB2 based on current available calibrations of the intrinsic properties of upper main sequence stars. Following these conclusions, we adopt a distance of 1.45 kpc (distance modulus 10.80) to the region. We also assume that all the structures discussed here belong to a physically-coherent region lying at approximately that distance, rather than belonging to disconnected entities along the line of sight, as further discussed in Sect. 4.4.

3 Observations

3.1 Near-infrared spectroscopy

We carried out near-infrared spectroscopic observations during two observing runs at the Calar Alto observatory. The first one took place from 16 June to 1 July 2002 at the 1.23 m telescope, and the second one from 31 July to 2 August 2004 at the 2.2 m telescope. The instrument used was in both cases MAGIC, a NICMOS3-based near-infrared camera and spectrograph. We obtained the spectra using the resin-replica grism providing a resolution $\lambda / \Delta \lambda = 240$ over the 1.50-2.40 $\mu $m range with the 1'' slit used. Each star was observed at six positions along the slit, with exposure times per position ranging from 20 s (stacking 10 integrations of 2 s) for the brightest stars to 60 s (stacking 20 integrations of 3 s) for the faintest, with exposure times determined by the 2MASS $K_{\rm S}$ magnitude and the aperture of the telescope used. We carried out the extraction and calibration of the spectra with dedicated IRAF[*] scripts. The frames obtained at consecutive slit positions were subtracted from each other to cancel out the sky contribution to the spectrum, and were divided by a flat field frame. A one-dimensional object spectrum was then extracted from each sky-subtracted, flat-fielded frame. We performed the wavelength calibration of each individual spectrum with the OH airglow lines in each frame as a reference (Oliva & Origlia 1992). The wavelength-calibrated spectra extracted at each sky position were then coadded, with deviant pixels due to detector defects or cosmic ray hits automatically clipped off. Cancellation of telluric features was achieved by ratioing the object spectra by those of the nearby G5IV star HD 190771, which is expected to be featureless at the resolution employed, reduced in the same manner. Finally, we performed relative flux calibration by multiplying the reduced spectra by that of a 5700 K blackbody, which should be a good approximation to the spectral energy distribution of an unreddened G5IV star in the wavelength range covered by our spectra.

3.2 Visible spectroscopy

Spectroscopy in the visible of 26 targets suspected to be reddened early-type stars from their near-infrared photometric and spectroscopic properties was obtained at Calar Alto using CAFOS, the facility visible-light imager and low-resolution spectrograph, in an observing run between 1 and 12 July 2005. We obtained spectra of four additional stars with an identical instrumental setup at the same telescope in an observing run between 17 and 25 August 2006. The grism used covered the range shortwards of $\lambda = 6350$ Å at a resolution of $\lambda / \Delta \lambda = 1000$ and with a 1''5 slit. We based the exposure times on the B magnitude listed in the USNO-B catalog, and ranged from 10 min to 180 min. For integration times longer than 30 min, the exposures were split in blocks of 30 min, and the individual extracted and reduced spectra were then stacked together. Spectra of three lamps of HgCd, He, and Rb were taken between subsequent exposures for wavelength calibration, to minimize the effects of instrument flexure. The frames containing the raw spectra were subtracted from bias and divided by a spectroscopic flat field, and the spectra were subsequently extracted from each one of them. The individual wavelength-calibrated spectra were coadded after identification and removal of cosmic ray hits. The coadded spectra were then normalized to the interpolated continuum, to facilitate feature recognition and comparison to spectral atlases of these generally heavily-reddened objects.

4 Results

4.1 Infrared spectral classification and membership

Despite the limited resolution and signal-to-noise ratio of our infrared spectra, it is possible to broadly classify our entire sample of stars into eight classes:

\par\includegraphics[width=7cm,clip]{9917fig6.eps} \end{figure} Figure 6: Spectra of the extremely red sources #158 and #235. Star #158 is the well-studied massive protostellar object AFGL 2591 (see text). Note the similar appearance of the weak Br$\gamma $emission line at 2.166 $\mu $m.
Open with DEXTER

4.2 Infrared color-color diagram

The distribution of objects of our sample in the near-infrared color-color diagram is presented in Fig. 7. The most abundant class of stars, those displaying featureless near-infrared spectra, are well aligned along the reddening vector, displaying the colors of reddened normal early-type photospheres without indications of near-infrared excess. Based on the position of the reddest stars with the colors and near-infrared spectra of normal photospheres, we estimate a maximum reddening toward the halo of Cygnus OB2 of $A_{V_ {\rm max}} \simeq 15$ mag, or $A_{K_ {\rm max}} \simeq 1.7$ mag.

\par\includegraphics[width=9cm,clip]{9917fig7.eps} \end{figure} Figure 7: Color-color diagram of all the objects in our sample, with the exception of foreground stars. Circles: stars with featureless spectra; squares: emission-line stars; triangles: massive young stellar objects; 3-pointed asterisks: red giant branch stars; 4-pointed asterisks: carbon stars; and 8-pointed asterisks: long-period variables. The length of the reddening vector (dotted line) corresponds to 20 mag of visual extinction. The dashed line separates the loci of classical Be from Herbig Ae/Be stars.
Open with DEXTER

The bluest emission-line stars tend to appear slightly to the right of the limiting reddening vector having its origin at the position of the earliest unreddened stars, thus indicating the existence of moderate amounts of circumstellar excess. However, a break is easily identified at (H-K) >0.8 as emission-line stars with redder colors are all far more removed from the limiting reddening vector. We interpret this as a consequence of the existence of two classes of emission-line stars in our sample, as discussed in Sect. 3.1. Classical Be stars have infrared excesses produced by free-free and free-bound transitions in the ionized circumstellar gas, which is thought to be distributed in a decretion disk formed as a consequence of mass loss from the atmosphere of a star rotating at nearly the break-up speed (Sigut & Jones 2007; Ekström et al. 2008). Although the Be phenomenon may appear at all stellar ages, and also among young stars with a rapid initial rotation speed (Zorec & Briot 1997), several observational lines of evidence indicate that it most usually results from spin-up of the star via mass transfer from a companion, near the end of the main sequence lifetime (McSwain & Gies 2007). On the other hand, Herbig Ae/Be stars are pre-main sequence stars that display strong infrared excesses produced by dust in their circumstellar accretion disks, particularly in the puffed-up walls delimiting their inner rims (Dullemond et al. 2001). As shown by Hernández et al. (2005), the JHK color-color diagram is a useful tool for distinguishing between both classes of star and our own sample confirms the split of both classes according to their position in the (J-H), (H-K) diagram, or more precisely according to the reddening-free parameter Q = (J-H) - 1.70(H-K)already used in Sect. 2. Classical Be star candidates cluster around - 0.18 < Q < 0.07, whereas Herbig Ae/Be candidates cover a broad range starting at Q = -0.22 and extending to Q=-1.38. The adopted dividing line between classical Be and Herbig Ae/Be stars, defined as Q=-0.20, is indicated in Fig. 7. Since Herbig Ae/Be fulfill our selection criteria due to their infrared excess rather than to their photospheric colors, by including them in our sample of candidate Cygnus OB2 halo members we extend it to stars less massive than the OB stars on which we focus. However, including the Herbig Ae/Be in our discussion is still useful as they are additional tracers of recent star formation.

Two of the massive young stellar objects that we include in our sample stand out in Fig. 7 due to their extreme infrared excess. Star #158 (= AFGL 2591) is represented by the triangle at (J-H) = 1.24, (H-K) = 2.76, whereas the even more extreme star #235 (=IRAS 20249 + 3953) appears near the upper right corner of the diagram. The candidate massive young stellar object #221 shows a modest amount of excess by comparison, occupying a position similar to that of some Herbig Ae/Be stars, which probably indicates a more evolved status. We note the apparent existence of a third object with extreme infrared colors, star #141, which we classified among the red giant candidates on the basis of the well-defined CO absorption bands longwards of 2.29 $\mu $m characteristic of a late-type star. However, an inspection of the 2MASS images at the position of this star clearly shows the presence of a nearby star of similar brightness at $K_{\rm S}$, which dominates at shorter wavelengths; in fact, star #141 is not noticeable in the 2MASS J image. The strong apparent infrared excess is thus almost certainly an artifact due to contamination of the photometric measurement by the dominating bright companion.

The cool stars in our sample that we classify as non-members appear almost entirely removed from the loci occupied by the candidate members of the Cygnus OB2 halo, having in general much redder colors. This is partly caused by their intrinsically redder spectral energy distributions, but the dominant factor is their larger average distances due to their brighter infrared absolute magnitudes. Whereas we expect OB stars in Cygnus OB2 to have absolute magnitudes in the - 3 < MK < -5.5 range (Martins & Plez 2006), cool red giants are expected to reach MK = -7 (e.g., Ferraro et al. 2000), long period variables are in the range -6.4 > MK > -8.2 (Knapp et al. 2003), and carbon stars usually have MK < -8 (Weinberg & Nikolaev 2001; Demers & Battinelli 2007). They are thus easily detectable at the distance of the Perseus arm and beyond in a magnitude-limited sample like ours.

\par\includegraphics[width=18.4cm,clip]{9917fig8.eps}\end{figure} Figure 8: Visible spectra of the stars for which a new MK spectral classification is given in this work. Note the strong interstellar feature centered at 4428 Å, due to the strong extinction. The main criteria used for classification are outlined in Sect. 4.3.
Open with DEXTER

4.3 Spectral classification in the visible

The new spectroscopic classifications of 29 stars presented for the first time in this paper[*] are listed in Col. 7 of Tables 1 and 2. Classifications were based on a comparison between our spectra and the atlas of Walborn & Kirkpatrick (1990), which contains an abundant selection of early stars providing a very finely-grained coverage of the two-dimensional classification of the earliest spectral types, as well as an extensive discussion of the classification criteria. In the range of spectral types covered by our stars, spectral subtype classification is mainly based on the relative strengths of the HeII features with respect to HeI, most notably HeII $\lambda$4541 vs. HeI $\lambda$4471, for O-type stars in which HeII lines are visible; and on the SiIV/SiIII and SiII/SiIII ratios, as well as the HeI/H ratios, for the early B stars. In turn, the luminosity class is largely derived from the appearance of the lines produced by ionized metals (CIII, SiIV) in the proximities of H$\gamma $ and, most notably, the intensity of the CIII+OII feature at 4650 Å. Because of the large extinction in their direction, all our stars show prominent interstellar absorption features, most notably the broad band centered at 4428 Å and narrower lines at 4501 Å, 4726 Å, and 4762 Å. To estimate the uncertainty in the spectral classification, each spectrum was separately classified by each of the authors using the same reference atlas. A comparison of the individual results obtained then led us to estimate an uncertainty in each dimension of the spectral classification amounting to approximately one luminosity class and less than one spectral subtype, respectively. The spectral types listed in Tables 1 and 2 are the averages of the classifications attributed by each author.

The spectral types obtained for the new members of the region are encompassed within a very narrow range going from O7 to B2, with the vast majority of the members having subtypes between O9 and B1. In contrast, stars can be found in all the luminosity classes from I to V. Our sample includes both objects with featureless spectra in the near-infrared (Table 1) and objects observed to display lines of the Brackett series in emission (Table 2). Since we restricted our visible spectroscopy to targets with relatively bright magnitudes in the blue, we only expect classical Be stars in the latter category to be present in the sample observed in the visible. Indeed, none of the stars that we observed in the visible displays the strong infrared excess characteristic of Herbig Ae/Be stars (Hernández et al. 2005). On the other hand, of all the classical Be star candidates only one, star #128, also displays noticeable emission in the visible in the form of a faint emission core at the center of the Balmer lines.

4.4 Spatial distribution

Schematic plots with the spatial distributions of stars that we consider as likely members and non-members of the Cygnus OB2 halo are presented in Fig. 9. Candidate members are those listed in Tables 1-3, 9, as well as the young stellar object #221. Both populations have clearly different distributions: the non-members are roughly uniformly spread out, as we may expect from a population composed by old stars. Such uniform distribution implies that there are no significant large-scale differences in extinction along the line of sight at least out to the typical distances of those stars, which as noted above are greater on the average than those of early-type stars of the same apparent magnitude.

\par\includegraphics[width=16cm,clip]{9917fig9.eps} \end{figure} Figure 9: Left panel: location of all the candidate members identified through their infrared spectra (Tables 1-3), including the protostellar candidates listed in Table 9 and star #221. Symbols are as follows: open circles, stars with featureless near-infrared spectra; filled circles, stars with Brackett emission lines; filled triangles, massive young stellar objects; and open triangle, Wolf-Rayet star WR 142a. Right panel: same for three classes of non-members; open circles, red giant stars; filled circles, carbon stars; and open triangles, long-period variables.
Open with DEXTER

The stars that we have considered as likely members, on the other hand, do display structure in their spatial distribution. Although OB stars are found all over the field, two areas of surface density greater than the average appear toward the (galactic) northwest (upper right in the left panel of Fig. 9) and toward the east (left in that same figure).

\par\includegraphics[width=9cm,clip]{9917f10.eps} \end{figure} Figure 10: Location of all the 98 candidate early-type stars overlaid on a radiocontinuum image of the region at 1420 MHz, from the Canadian Galactic Plane Survey (Taylor et al. 2003). Stars are represented by the following symbols: red squares, stars with featureless near-infrared spectra; black circles, stars with Brackett emission lines; blue triangles, massive young stellar objects; and diamond, Wolf-Rayet star WR 142a. The large inner and outer blue dashed circles are respectively $1^\circ $ and $2^\circ $ in radius, and they delimit the region in the outskirts of Cygnus OB2 covered by the present study. The location of the main HII regions, as well as the W75N star forming region and the NGC 6910 open cluster, are indicated. The HII regions noted here with prefix ``DR'' were identified by Downes & Rinehart (1966). The prefix ``LDK'' refers to the list of embedded clusters of Le Duigou & Knödlseder (2002). The position of the compact HII region DR 7 is marked here, although it is probably more distant and associated to the background Perseus arm (Comerón & Torra 2001).
Open with DEXTER

Figure 10 identifies the main structures known in the region, superimposed on a 1420 MHz radio continuum map from the Canadian Galactic Plane Survey (Taylor et al. 2003) tracing the emission by ionized gas. A comparison to previously-known structures in that region shows that the northeastern concentration corresponds to the eastern boundary of the Cygnus OB9 association (e.g., Garmany & Stencel 1992), and in particular to the cluster NGC 6910 (the tight concentration of stars near l = 78.7, b = +2.0, which are recovered by our candidate member selection procedure). On the eastern side, the broad enhancement of stellar surface density contains the chain of compact HII regions DR 17, DR 21 in the W75 complex, DR 23, and DR 22, all of which are embedded in the Cygnus X North complex. The reality of Cygnus X as a single physical entity, rather than as a chance alignment of unrelated structures lying at different distances along a spiral arm seen nearly end-on, has been widely debated in the literature (see, e.g., Odenwald & Schwartz 1993). Recent detailed observations of the molecular gas in the region, mainly by Schneider et al. (2006, 2007), strongly argue for a real interconnection among these structures and present evidence for their interaction with the massive stars of Cygnus OB2. Based on those results, we consider here that DR 17, W75/DR 21, DR 22, and DR 23 are all structures belonging to Cygnus X and lying at the same distance as Cygnus OB2, although we note that a distance of 2-3 kpc to W75/DR21, significantly larger than the one adopted here, has been often adopted in the literature (Kumar et al. 2007).

The concentration of early-type stars on the eastern side is particularly obvious at the position of DR 17, composed by the earliest-type members of the embedded clusters hosted by this HII region (Le Duigou & Knödlseder 2002). However, it is more widespread and extends beyond the boundaries of the compact HII regions. A similar result has been reported recently by Kumar et al. (2007), who find that many young stars revealed by their infrared excesses in the Spitzer bands are located outside the boundaries of the main molecular complexes in the region. We note, in particular, the concentration of emission-line stars toward this general region, with a preference toward the surroundings of DR 21, although DR 21 itself is too obscured for its embedded stars to be included in our magnitude-limited sample (Nadeau et al. 1991). Almost all these stars have clear infrared excess placing them among the Herbig Ae/Be stars as discussed in Sect. 4.2, with the only exception of star #118, appearing in the direction of W75N and whose near-infrared colors place it in the classical Be star locus. Only another candidate classical Be star is possibly associated with a compact HII region, namely star #122 near DR 6, although the spatial coincidence between them is not precise.

The southwestern (lower right) part of the area covered by this study South of the galactic equator appears particularly devoid of early-type stars. This is interesting since that region coincides with Cygnus X South (Schneider et al. 2006), and recent stellar surface density estimates based on extinction-corrected starcounts (Bontemps et al., in prep; preliminary results shown in Reipurth & Schneider 2008) indicate an enhanced stellar density in that direction, which appears to be an extension of Cygnus OB2. The starcount enhancement detected by Bontemps et al. is contained within $1^\circ $ from the center of Cygnus OB2 and is in rough agreement with the distribution of early-type stars in the same region found by Comerón et al. (2002, see Fig. 14 in that paper), which preferentially extends toward the west. The same distribution is well matched by that of A0V-A5V stars recently identified by Drew et al. (2008). As discussed by those authors, the magnitudes of these stars suggest an average age somewhat older than that of the Cygnus OB2 cluster if they lie at the distance of 1.45 kpc assumed here, in agreement with the results of Hanson (2003), and it is most likely the lower mass counterpart of the sample discussed by Comerón et al. (2002). The fact that we detect no traces of it in our color-selected sample beyond the $1^\circ $ circle suggests that it either does not reach beyond that distance, or that its possible extension does not contain massive stars picked up by our selection criterion. On the other hand, 13CO maps of that region (Simon et al., in prep.; preliminary results shown in Reipurth & Schneider 2008) indicate the presence of large amounts of molecular gas in the zone where we do not detect a corresponding overdensity of early-type stars. It thus appears that any star formation currently going on in that region of Cygnus X is not producing massive stars. It should be noted that this does not apply to the entire Cygnus X South region, which extends well beyond the boundaries of the area considered here and does contain massive embedded clusters and HII regions.

4.5 Evolutionary status

It is thus possible to identify three distinct components in the area targeted by this study. The first one is the population associated with the eastern edge of Cygnus OB9, particularly the cluster NGC 6910. The second component extends over the region occupied by the DR 17, W75/DR 21, DR 22, and DR 23 complexes in Cygnus X North. The third component is a distributed population whose members are stars that do not belong to either of those two groups, which may be related to the extended population in the central $1^\circ $-radius circle around Cygnus OB2 reported by Comerón et al. (2002) and Hanson (2003), and probably also by Drew et al. (2008).

\par\includegraphics[width=18cm,clip]{9917f11.eps} \end{figure} Figure 11: Temperature-absolute magnitude diagram with the location of the stars for which spectral classification is available in each of the three regions discussed in Sect. 4.5. Circles indicate luminosity class V, crosses class IV, triangles class III, squares class II, and asterisks class I. The evolutionary tracks and isochrones from Lejeune & Schaerer (2001) are plotted for comparison, highlighting the older age of the field population. We believe that most of the stars in the Cygnus X North area actually belong to the part of the field population lying along that line of sight.
Open with DEXTER

The new availability of detailed spectral classifications for many stars in these three groups allows us to investigate in more detail their nature and evolutionary status. To this purpose we have adopted for each O-type star an effective temperature $T_{\rm eff}$based on the $T_{\rm eff}$-spectral type calibration of Martins et al. (2005), using their observational effective temperature scale. To our knowledge no similar work is available to date covering the early B spectral interval, in particular the whole range of luminosity classes covered by our spectra. We have thus used the $T_{\rm eff}$-spectral type compilation by Tokunaga (2000) for luminosity classes I and V, applying a scaling factor to the temperatures so as to force agreement with Martins et al.'s (2005) observational temperature scale at spectral types O9I and O9.5V respectively. The calibrations of Tokunaga (2000) and Martins et al. (2005) closely match each other in the overlapping late-O range, implying that our arbitrary scaling factors deviate from unity by less than 4%. Because the compilation of Tokunaga (2000) does not include luminosity class III early-type stars, we have estimated their temperatures by means of a weighted interpolation between the temperatures adopted for classes I and V. The weight factor is derived once again by forcing agreement between the temperatures derived in this way for spectral type O9III and the corresponding value given by Martins et al. (2005). Approximate temperatures of B stars of luminosity classes II and IV have then been computed by averaging the temperatures adopted for stars of luminosity class I and III, and III and V, respectively. We estimate that the systematic errors introduced in this way are below 2000 K for all the spectral types represented in our sample. A summary of $T_{\rm eff}$-spectral type calibrations for B-type stars presented by Fitzpatrick & Massa (2005) shows that existing systematic uncertainties in this domain exceed this value. We thus consider that the ad-hoc procedure that we have described to assign effective temperatures to our stars does not introduce large additional uncertainties.

The large scatter in absolute magnitudes at a given spectral type among O and early-B stars (e.g., Jaschek & Gómez 1998) cautions against the use of the spectral classification alone to derive the positions of our stars in the temperature-absolute magnitude diagram, and compare them to the predictions of evolutionary tracks to assess their evolutionary status. Instead, we have preferred to adhere to the underlying hypothesis already noted in Sect. 4.4 that all our stars have essentially the same distance modulus of 10.80 mag, and derive individual absolute magnitudes as

MV = K - 10.80 - 0.659[(J-K)-(J-K)0] + (V-K)0

where (J-K)0 and (V-K)0 are the intrinsic, unreddened color indices and the factor 0.659 is derived from the extinction law of Rieke & Lebofsky (1985). We have used the intrinsic colors derived by Martins & Plez (2006) for O stars and those compiled by Tokunaga (2000) for B stars. We estimate that interpolations needed to cover all the spectral types and luminosity classes in our sample introduce a systematic error in MV not exceeding 0.1 mag. The scatter in intrinsic colors among actual stars is likely to be a more important source of errors, as illustrated by the results of Robert et al. (2003).

The $T_{\rm eff}$-MV diagrams are shown in Fig. 11 for the eastern edge of Cygnus OB9, the region in Cygnus X North around DR 21, and the field population. In these figures, we also plot the isochrones calculated by Lejeune & Schaerer (2001) in their Case e with solar metallicity, high mass loss and no rotation, for initial stellar masses (60), 40, 25, 20, 15, 12 and 9 $M_{\odot}$ (solid lines) and stellar ages 2, (3.1), 4, 6, 8, 10, 15 and 25 Myr (dotted lines). Individual $T_{\rm eff}$ and MV derived for each star in the different regions are listed in Table 10.

Table 10: Adopted temperatures and absolute magnitudes.

There are significant differences between the distributions of stars in the temperature-luminosity diagrams among the three components considered. The Cygnus OB9 region contains abundant O-type main sequence stars, particularly in the cluster NGC 6910 whose members display other signatures of youth (Delgado & Alfaro 2000), and only four stars outside the main sequence (the B giants #181, 193, and 206 and the B1.5IV star #142) are found within this region in our sample. The class V O-type stars appear mostly above the main sequence, which may indicate that they lie at a shorter distance than that adopted by us. Garmany et al. (1992) find a very uncertain distance modulus DM = 10.0 to NGC 6910, in contrast with the better determined value for the rest of the association, DM = 11.0. However, this latter value is in closer agreement with the distance found by Delgado & Alfaro (2000) for NGC 6910 itself, $DM = 11.2 \pm 0.2$, which solves the discrepancy noted by Garmany & Stencel (1992) and is only marginally larger than that adopted by us. A value significantly shorter than DM = 10.8thus does not seem supported by other observations. The position of the class V O-type stars above the lowest isochrone may be at least partly due to binarity, as well as to rotation, whose effects can brighten O stars by as much as $\Delta M_V = 0.6$ (Meynet & Maeder 2000) for an initial velocity of 200 km s-1.

The situation is less clear in Cygnus X North, as the higher extinction in the area prevented us from obtaining visible spectra of many stars. The presence of several Brackett emission-line stars with the infrared characteristics of Herbig Ae/Be stars suggests recent star formation over a widespread area. The five stars with spectral classification in the zone are all lightly reddened and may be on the near side of the complex. The luminosity classes (Ib-III) of four of them indicates that they are not related to the ongoing star formation in the area, with the possible exception of the O9V star #100 that is very close to DR 21. It can be seen in Fig. 11 that a physical association of this star with DR 21 would support the distance adopted here, rather than the 2-3 kpc often used in the literature (Davis et al. 2007), as a greater distance would move it even further away from the main sequence.

In contrast, most of the population outside those two areas, which we consider as the field component, is lightly reddened, and visible spectroscopy is available for 20 of its members, of which 14 are found to have luminosity classes placing them above the main sequence. Of the remaining six stars, four are classified as B1V or B2V and thus main sequence lifetimes longer than 10 Myr; and only three stars, #150 (O8V; Morgan et al. 1955), #152 (O9.5V; this work) and WR 142a (WC8, for which Pasquali et al. (2002) estimate an age of at least 3 Myr) indicate more recent massive star formation. The first two are also the ones closest to the main sequence plotted in Fig. 11, in agreement with the common distance adopted for all our stars. Besides WR 142a, 7 out of 9 emission line stars in the region appear to be classical Be stars, with the only exceptions of stars #178 and #180. A comparison of the position of the stars belonging to the field component with the evolutionary tracks in the temperature-luminosity diagram indicates ages older than 6 Myr for this population.

It may be noted that several stars classified as B1V or B2V appear up to 2 mag above the main sequence in the temperature-magnitude diagrams for both Cygnus OB9 and the field population. This is considerably more than what one may expect from the effects of rotation, binarity, or the spread in intrinsic magnitudes within their spectral types. In fact such stars are not expected to be present in our sample, as normal main sequence stars later than B1V at the distance of Cygnus OB2 should have $K_{\rm S}$ magnitudes fainter than our target selection limit (see Sect. 2). In principle, these may be unrelated foreground stars at much closer distances. However, an inspection of their spectra in Fig. 8 shows interstellar features of a depth similar to those of the other stars in the region, indicating a similar level of extinction that is confirmed by their infrared photometry. A misclassification of the luminosity class thus appears as a more likely cause, which may be confirmed with better quality spectra. However, we have decided not to modify the spectral types given in this paper so as not to bias our classification with a posteriori knowledge.

5 Discussion

Our observations sample an interesting region where the outskirts of the Cygnus OB2 association merge with neighbor structures, allowing us to explore the possible existence of links among them and the history of star formation in the whole region. Such questions are rendered all the more relevant given the increasing evidence for their common distance (Schneider et al. 2007) and the possibility that the distinct structures recognized thus far are parts of a single association, particularly Cygnus 1, 8, and 9 (Mel'nik & Efremov 1995).

One of our main results is the identification of an extended distribution of OB stars around Cygnus OB2 and probably beyond, selected through their near-infrared colors and confirmed on the basis of their near-infrared spectrum. Visible spectroscopy confirms the very high reliability of infrared spectroscopy at modest resolution and signal-to-noise ratio in identifying early-type stars, as already found by Hanson (2003) when investigating the sample of candidates identified by Comerón et al. (2002). Our results also show a remarkable coincidence with those of Hanson (2003) in that the newly-identified early-type stars do not have, as a rule, spectral types as early as those at the center of Cygnus OB2. Instead, they are dominated by massive stars that have already evolved off the main sequence (i.e., giants, supergiants, and a Wolf-Rayet) or are thought to be at the end of their main sequence lifetime (the classical Be stars). Such evolved stars are dominant among the field component described in Sect. 4.5. Their presence also among the populations that we identified as belonging to the complex that includes DR 17, DR 21, DR 22, and DR 23, on one side, and Cygnus OB9, on the other, is not surprising since our allocation of stars to those groups is purely based on position. A certain level of contamination of those two components by stars actually belonging to the field population is thus to be expected.

What is the relationship between the field component and Cygnus OB2, if there is one? Hanson (2003) already pointed out that the spatial distribution of the evolved component differed markedly from that of the most massive stars of Cygnus OB2 in being much more evenly distributed over the studied area. In fact, based on both the wider spatial distribution and the gap in age Hanson suggested that the evolved component is unrelated to Cygnus OB2. Our new observations allow us to reevaluate this conclusion by extending the analysis of the spatial distribution to a larger area.

Unfortunately, the somewhat different selection of candidate early-B stars carried out in Comerón et al. (2002) and the present paper prevents a direct comparison between the stellar densities inside and outside the $1^\circ $ radius. It is however possible to perform the color selection of candidates over a much larger area that includes the central 1$^\circ$ radius around Cygnus OB2, the region covered by this study, and the surrounding regions, using the results of Sect. 4.1 as a way to remove statistically the contamination by cool background stars and by foreground stars.

\par\includegraphics[width=9cm,clip]{9917f12.eps} \end{figure} Figure 12: Surface density of OB stars selected according to the criteria K < 8.8, $-0.15 < (J-H)-1.70 (H-K_{\rm S}) < 0.20$. The shade of grey scales with the number of stars per square degree, ranging from zero (white) to 100 or more (black). The Cygnus OB2 central cluster is the black square near the center, with a density of 176 stars per square degree. A pedestal surface density of 12.5 stars per square degree has been subtracted from each cell to account for the estimated contamination by cool and foreground stars that are also selected by the color-based criteria.
Open with DEXTER

We have carried out this exercise over the area limited by $70^\circ < l < 90^\circ$, $-5^\circ < b < +5^\circ$, which has Cygnus OB2 near its center and includes the entire Cygnus OB9 region as well as parts of Cygnus OB1 to the West and Cygnus OB7 to the East. Stars were selected according to the criteria $K_{\rm S} < 8.8$, $-0.15 < (J-H)-1.70 (H-K_{\rm S}) < 0.20$. The results of the spectroscopic follow-up discussed in Sect. 4.1 indicate that these criteria select approximately 12.5 cool (red giants, long-period variables, and carbon stars) and foreground (mostly late B to F-type) stars per square degree, in addition to OB stars. We assume that these contaminants are uniformly distributed across the field, and subtract this number density from the total number density of selected objects. The result is displayed in Fig. 12, which is similar to the surface density maps presented by Knödlseder (2000) and Bontemps et al. (in prep., presented by Reipurth & Schneider 2008). The main difference is that Fig. 12 is restricted to early-type stars; the surface densities are thus much lower and the resolution much coarser than in their contour maps. As one may expect, Cygnus OB2 is the most prominent feature appearing as a very tight concentration near the center of the figure, with a peak surface density reaching 176 OB stars per square degree in the central 0.25 square degrees. The stellar density remains high toward the western half occupied by Cygnus OB9 next to Cygnus OB2, and then Cygnus OB1. The easternmost quarter of the figure is occupied by the more nearby association Cygnus OB7 (Garmany & Stencel 1992). Despite the overall similarity, there are significant differences between our map and the general starcount-based map of Bontemps et al.: Cygnus OB9 appears much more prominently in our map, whereas the DR 17/DR 21/DR 22/DR23 region, while easily recognizable as a stellar density enhancement in our Fig. 10, stands out much more clearly in Bontemps et al.'s map.

Interestingly, Fig. 12 shows that a non-zero surface density of early-type star candidates pervades the region even outside the boundaries of the already-known OB associations. By extrapolating the results obtained in the area between $1^\circ $ and $2^\circ $ from the center of Cygnus OB2, we tentatively identify this as the field population mainly composed of evolved stars, noting that this population does not appear to be associated with Cygnus OB2 nor with any other specific feature in the region. The massive precursors required by this widespread field population suggests that massive star formation has been taking place in the Cygnus region for a time exceeding 10 Myr, and long before the formation of the currently observed associations like Cygnus OB2, whose age is less than 4 Myr (Massey et al. 1995). The fact that such a population appears as an extended component with no apparent concentration is not entirely surprising, given the high density of OB associations in the region, if we assume that this density was similarly high in the past. The typical center-to-center distance among the four associations currently defined in the galactic longitude interval $74^\circ < l < 82^\circ$ (Cygnus OB1, OB2, OB8, and OB9) is $\sim$$ 2^\circ 5$. At the distance of 1.45 kpc adopted throughout this paper, an internal velocity dispersion of 3.5 km s-1 would suffice to make them overlap after 10 Myr if they are gravitationally unbound, even if they occupied a small initial volume. This velocity dispersion is comparable to the internal velocity dispersion derived for members of Cygnus OB2 by Kiminki et al. (2007), $2.44 \pm 0.07$ km s-1. Moreover it can be regarded as an upper limit, since even very young OB associations like Cygnus OB9 occupy a sizeable volume comparable to the center-to-center angular distance quoted above, and the field population contains members older than 10 Myr leaving more time for dispersal of the original associations. This population may be the equivalent to an extended component also observed in other galaxies (Pasquali & Castangia 2008), and it can thus be accounted for by assuming a previous generation of OB associations, now dispersed, with characteristic similar to those of the current associations. It appears unlikely that the field population has its origin in bound clusters having become unbound in their early evolutionary stages (Lada & Lada 2003): recent observational evidence discussed by Gieles & Bastian (2008) indicates dispersal timescales at least one order of magnitude longer than the age of the field population in a wide range of environments, including our galactic neighborhood. Our results rather suggest that those massive OB stars originate in either isolation or unbound aggregates.

The sustained massive star forming activity in Cygnus must have been essential in arranging the large-scale distribution of the interstellar gas and in driving the current generation of star formation. It may also be at the origin of the large Cygnus Supperbubble observed in X rays (Cash et al. 1980; Bochkarev & Sitnik 1985; Uyaniker et al. 2001) and possibly related phenomena such as the apparent large-scale expansion pattern noted by Comerón et al. (1998) in the proper motions of stars in the area.

6 Conclusions

We have extended previous work carried out mainly by Comerón et al. (2002) and Hanson (2003) in characterizing the population of early-type stars in the vicinity of Cygnus OB2. We have done this by identifying a magnitude-limited sample of candidate early-type stars through near-infrared imaging in the area between $1^\circ $ and $2^\circ $ from the center of Cygnus OB2, confirming them through near-infrared spectroscopy, and providing an accurate spectral classification of many of them through spectroscopy in the visible. Our main results can be summarized as follows:

We believe that the present study conclusively dismisses the case for a large extent of Cygnus OB2 much beyond the boundaries of its central concentration, which was already suggested by Garmany & Stencel (1992) and supported by Comerón et al. (2002), but then questioned by Hanson (2003). Previous studies such as those by Massey & Thompson (1991) and Knödlseder (2000) clearly show that most of the massive stellar content of the association still awaits identification and classification (see also Kiminki et al. 2007, for recent work on the confirmation of new OB members in the inner regions of Cygnus OB2). Our results indicate that future studies of Cygnus OB2 aiming to characterize its upper main sequence should concentrate on heavily-reddened members near its core, rather than on identifying new members at large distances from it.

We note that over half of the early-type stars for which we have obtained confirming near-infrared spectroscopy do not yet have a detailed spectral classification available. This is particularly important for the new candidates identified in Cygnus X North, whose proximity to compact HII regions, some of which contain embedded clusters, leads us to suspect a young age for most of them. If confirmed, the extent and composition of this group may warrant its status as an OB association in its own right, although we feel that a better characterization is still needed before taking that step. On the other hand, the as yet unclassified stars located elsewhere offer a potential for the discovery of new interesting members deserving detailed study, such as WR 142a or BD $+43^\circ 3654$. The fact that the vast majority of stars that form the basis for the present study are recognized here as early-type members of the region for the first time is a reminder of the large amount of observational work still needed to characterize the stellar population in the Cygnus region. The importance of this task is stressed by the rapid progress being made at different wavelengths in the characterization of its interstellar content.

Once again it is a pleasure to thank the staff of the Calar Alto observatory, and especially Santos Pedraz and Ana Guijarro, for their unfailingly competent and friendly support while observing on Calar Alto. We are very thankful to Bo Reipurth and Nicola Schneider for making available to us their chapter on the Cygnus region for the Handbook of Star Forming Regions prior to publication. We also thank the anonymous referee for constructive comments that helped improve both the style and content of this paper. F.C. warmly acknowledges the hospitality of the Vatican Observatory during the preparation of this paper. A.P. acknowledges support from the OPTICON Network. J.T. and F.F. acknowledge support by the Spanish Ministry of Science and Technology under contract AYA2006-15623-C02-02. This research has made use of the SIMBAD database operated at CDS, Strasbourg, France. It also 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; and of data from the Canadian Galactic Plane Survey, a Canadian project with international partners, supported by the Natural Sciences and Engineering Research Council.



Online Material

Table 1: Stars with reddened, featureless spectra.

Table 2: Emission line stars.

Table 3: Lightly reddened OB stars.

Table 4: Foreground intermediate-type stars and lightly reddened, unclassified stars.

Table 5: Late B-F stars with Brackett absorption lines.

Table 6: Red giant candidates.

Table 7: Carbon stars.

Table 8: Likely long-period variables.

Table 9: Extremely red objects.

Copyright ESO 2008