We used the Second Incremental Release of the 2MASS Point Source Catalog to extract a sample of stars within one degree of the center of Cygnus OB2, assumed to lie near the position of the multiple, Trapezium-like system VI Cyg 8 (Schulte 1958). This area contains practically the full extent of the association as determined by Knödlseder (2000). Figure 1 shows the (H-K), K diagram of the region together with the limits corresponding to stars earlier than a given spectral type having different amounts of foreground reddening. Cygnus OB2 members earlier than spectral type S are expected to fulfill

Figure 2 shows the (J-H), (H-K) diagram of the stars brighter than the limit given by Eq. (1) when using the values of M_K(S) and (H-K)₀(S) appropriate for a spectral type B0, showing the signature of the presence of the association. The extinction vector traced by the distribution of heavily reddened stars is well represented by the extinction law of Rieke & Lebofsky (1985). This was already shown to be so by Torres-Dodgen et al. (1991) in the infrared, who also confirmed early findings by Johnson & Borgman (1963) at visible wavelengths. At low-to-moderate values of the color indices there are clearly two parallel sequences each of which follows the reddening vector. The upper one has its origin at the position occupied by unreddened late-type giants, while the lower one starts at the position of the earliest stars. This alone already may be used to obtain a first estimate of the early-type contents of the association and of the average reddening in its direction, by just tracing back the position of the points in the lower sequence along the reddening vector until intersecting the locus of unreddened early-type stars. Another obvious group of objects in Fig. 2 appears at the highest values of (H-K), while having relatively blue (J-H), probably indicative of the existence of strong infrared excesses of circumstellar origin. Nearly all these objects are relatively faint (K>10) and, although they are of great interest to study the ongoing star forming activity in Cygnus OB2, their faintness in the bands that are less affected by circumstellar emission suggests that they may be intermediate mass stars, rather than members of the high end of the mass function of Cygnus OB2 that is the main purpose of our study. Their faintness also places them outside the limits accessible within moderate integration times with the telescope and instrument used for this survey, and therefore we will not discuss them in the present paper. The reader is referred to comprehensive studies on IRAS sources or molecular gas in the Cygnus region (Odenwald 1989; Odenwald & Schwartz 1989; Dobashi et al. 1996) for discussions on the youngest component of this region.

Most of the objects near the lower sequence in Fig. 2 have relatively blue colors, indicating that much of the early-type stellar contents of Cygnus OB2 is only moderately reddened. The upper sequence dominates at redder colors, showing that the vast majority of red stars seen in the direction of Cygnus OB2 are actually background, unrelated late-type stars. However, there is still a considerable number of stars in the region lying to the right of the upper sequence in the color-color diagram. These infrared-bright, very red objects are in principle candidate members of the heavily reddened massive component of Cygnus OB2 whose existence has been suggested by previous works (Reddish et al. 1966; Partharasathy & Jain 1995; Knödlseder 2000). Contrarily to what happens with the lightly obscured component, these objects do not tend to align along the extension of the lower sequence. This may be due to factors such as peculiarities in their intrinsic colors, circumstellar emission by hot dust, anomalous behavior of the extinction law in their direction, or contamination by late-type stars like long-period Mira variables, whose intrinsic H-K is redder than that of normal giants with the same intrinsic J-H (Glass et al. 1995). The spectroscopy that we present in this paper should help in clarifying the nature of these objects.

We made a first selection of targets for spectroscopic observations by using Eq. (1) in combination with the estimated sensitivity limit of K < 10 of the telescope and instrument used within a reasonable integration time. We adopted a limiting spectral type B0 in Eq. (1), corresponding to M_K = -3.1, (H-K)₀ = -0.04. The K < 10 condition implies some incompleteness affecting possible highly obscured members of the association, appearing at (H-K) = 1.1 (corresponding to A_V > 17.5) for B0 stars, and even higher extinctions for earlier types. However, as we discuss below in view of the location of the actually observed stars in the (J-H), (H-K) diagrams, we do not expect such incompleteness to affect our results in a significant way.

$\begin{figure} \par\includegraphics[width=7cm,clip]{MS1699f3.eps} \end{figure}$

Figure 3: Same as Fig. 1, but now marking with a small circle the stars for which spectra have been obtained. Most of the non-observed stars redder than H-K = 0.5 and brighter than K=10 lie very close to the reddening vector that has its origin at the locus of the late-type giants, and therefore are not expected to be members of Cygnus OB2. Some other stars in this region of the color-magnitude diagram are in the region of strong infrared excess (lower right part) of Fig. 2 and are therefore unlikely to be members of Cygnus OB2 as well for the reasons discussed in the text. Most of the stars with H-K < 0.5 and JHK colors suggestive of an early-type were also observed spectroscopically.

$\begin{figure} \par\includegraphics[width=7cm,clip]{MS1699f4.eps} \end{figure}$

Figure 4: Same as Fig. 2, but now marking with a small circle the stars for which spectra have been obtained. This figure shows that our spectroscopically observed sample with H-K > 0.5extends well into the band dominated by reddened late-type giants, thus providing a highly complete sampling of the possible obscured early-type component of Cygnus OB2. No stars from the region of high infrared excess were spectroscopically observed due to their faintness; as discussed in the text, these stars are unlikely to be a part of the massive component of the association. This might not be the case for at least some of the relatively few stars near the upper end of main reddened band, which are also too faint (K>10) to be included in our sample.

$\begin{figure} \par\includegraphics[width=7cm,clip]{MS1699f5.eps} \end{figure}$

Figure 5: A detail of Fig. 4 focusing on the region with H-K <0.5. Although our observations clearly include the majority of stars with colors consistent with them being early-type, lightly-reddened members of Cygnus OB2, the completeness of the spectroscopic sample in this color range is expected to be less than that in the range H-K > 0.5. This incompleteness is partly overcome by the adoption of spectroscopic classifications in the visible that are available in the literature for many of these stars. In any case, the obvious gap seen at H-K < 0.5 between the group of stars spectroscopically observed and the non-observed ones lying above them (many of which may be foreground stars) argues against a significant incompleteness of our spectroscopic sample over the full color range.

The selection criteria given by Eq. (1) and K < 10 reduced the sample of 2MASS stars within one degree from the center of Cygnus OB2 to 985. We further constrained our sample by calculating for each star its distance in the (J-H), (H-K)diagram to the upper reddening vector, and retaining only those with a distance greater than $\sim$ 0.1 mag. As can be seen from Fig. 2 this is a rather generous cutoff that actually includes a large fraction of stars closer to the upper than to the lower sequence. In this way we produced a list of 397 star, that we further divided into two priority groups separated by (H-K)greater or smaller than 0.5. We gave higher priority to the more reddened group (corresponding to A_V > 8 approximately), since it is here where we would expect to find the largest fraction of previously unidentified Cygnus OB2 members and the bulk of the hypothesized obscured population. We observed the 232 stars in this group. The available observing time allowed us to obtain spectra also for 83 stars in the second, bluer group, namely about half of our initially planned targets in this group. We proceeded by order of decreasing distance to the upper reddening sequence reaching a minimum distance of $\sim$ 0.2 mag to it (as compared to $\sim$ 0.1 for the redder sample), thus observing nearly all the objects clustering around the lower reddening sequence. To assess the possible incompleteness of our spectroscopically observed sample with respect to the total population of O-type members of Cygnus OB2 induced by our K < 10 cutoff, Figs. 3 and 4 plot the color-magnitude and color-color diagrams like in Figs. 1 and 2 respectively, but now highlighting the objects for which we obtained spectra. These figures suggest that indeed we should expect to have sampled the vast majority of the O-star contents of the association, especially the more obscured members.

3 Target selection