A High Angular Resolution Survey of Massive Stars in Cygnus OB2: $JHK$ Adaptive Optics Results from the
Gemini Near-InfraRed Imager

Massive stars profoundly influence the evolution of the Universe, from galactic dynamics and structure to star formation. They are often found with bound companions. Massive stars have a higher frequency of multiplicity than cooler, less massive stars (Raghavan et al., 2010; Duchêne & Kraus, 2013), especially when they are found in clusters (Mason et al., 2009). Spectroscopic studies of massive stars in the Milky Way (Sana et al., 2012) and in the Tarantula Nebula region of the Large Magellanic Cloud (Sana et al., 2013) demonstrate that perhaps $75\%$ of massive O-type stars have binary companions in orbits small enough that the stars will interact over their lifetime. Our knowledge of O-type multiple systems in larger orbits with periods in the range from years to thousands of years is incomplete due their great distances, but high angular resolution methods are beginning to fill in this period gap (Maíz Apellániz et al., 2019; Le Bouquin et al., 2017; Aldoretta et al., 2015; Sana et al., 2014).

At a distance of 1.33 – 1.7 kpc (Massey & Thompson, 1991; Torres-Dodgen et al., 1991; Hanson, 2003; Rygl et al., 2012; Kiminki et al., 2015), Cygnus OB2 = Cyg OB2 is the second closest OB association (after Ori OB1) that provides us with an example of a nearby, young stellar environment, rich in high-mass stars. Due to uneven extinction towards the region (Wright et al., 2015), the cluster begins to be unveiled in the infrared (IR). Torres-Dodgen et al. (1991) estimate the age of the association to be least 3 Myr through analysis of their Strömgren and infrared photometry, and Wright et al. (2015) argue that star formation has occurred more or less continuously over the last 1 to 7 Myr based upon the positions of the stars in the Hertzsprung - Russell (H-R) diagram. The young nature of the association is further established by the detection of X-rays from young, low mass stars in the region (Albacete Colombo et al., 2007; Wright & Drake, 2009; Wright et al., 2012). Spectroscopic surveys by Massey & Thompson (1991), Hanson (2003), and Kiminki et al. (2007) have established the early-type classifications of these stars. Massive stars are short-lived and therefore spend most of their formative years shrouded in their natal clouds, so that when they shed these clouds and a hot star is revealed, it is usually well into the main sequence stage of its life. The multiplicity of massive stars must play an important role in their formation because so many are members of binary systems (Zinnecker & Yorke, 2007). Massive stars are formed through the turbulent core collapse of a single cloud or by competitive accretion of multiple stellar seeds in a dense cloud (see the review by Rosen et al. 2020), and models of these processes predict a large incidence of binary stars with specific distributions of mass ratio and separation (Peter et al., 2012; Gravity Collaboration et al., 2018). Therefore, by studying the multiplicity properties of massive stars at an early stage we can test the role of companions in formation theories of massive stars.

The Cyg OB2 association is close enough that with modern-day adaptive optics (AO) we are able to resolve relatively close companions. The ALTAIR AO system and the Near-InfraRed Imager and Spectrograph (NIRI) at the Gemini North Observatory provides an effective tool to search for binaries, as was demonstrated by Lafrenière et al. (2014) in a multiplicity study of young stars in the Upper Sco association. With a resolution of $\sim 0\farcs 06$ and a sensitivity contrast limit of about 10 mag for differential photometry, the ALTAIR AO infrared system can delve into the depths of the association and find faint companions with periods in the range from hundreds to thousands of years. Our results complement the radial velocity survey of Kobulnicky et al. (2014) (and references therein) who searched for short period, spectroscopic systems in Cyg OB2. They determined that $30\%$ of their sample are spectroscopic binaries with periods less than 45 days.

In this paper, we provide measurements of $JHK$-band relative photometry and positions of candidate companions to our target stars. These results provide the first step in determining the true multiplicity fraction of widely separated systems. In section 2 we describe the observations of the sample in Cygnus OB2. We present the results of the survey in section 3 along with further details of the calibration in appendices for the astrometry (Appendix A) and photometry (Appendix B). We discuss the detection limits and the identification of probable physical companions in section 4. Section 5 presents the multiplicity fraction and companion frequency for the Cyg OB2 sample and compares these to similar results from studies of massive stars in other locations. We summarize the results and their significance in section 6.

We were able to observe 74 of the brightest O- and B-type stars in Cyg OB2 and one misidentified non-member using the infrared ALTAIR AO system (Richardson et al., 1998; Roberts & Singh, 1998a) at the Gemini North Observatory. We provide a list of our targets in Table 2 (given in full in the electronic version) that gives the target name, celestial coordinates (J2000), spectral classification and reference, optical and infrared (IR) magnitudes, and three measures of interstellar reddening. The majority of stars in this study were selected from the optical survey of Massey & Thompson (1991), who presented Johnson $B$ and $V$ magnitudes for the brighter stars in the sample as well as reddening towards each star. These targets are identified with prefix “MT” by their number assigned by Massey & Thompson (1991). Seventeen of our targets were selected from the infrared surveys by Comerón et al. (2002, 2008), and these are referenced by a prefix “A” or “B” from those papers. These are redder sources that are not readily detected in the optical surveys, but $V$-band and spectral information are available for some of these from Straižys & Laugalys (2008). An “S” designation is given for three stars from the compilation by Schulte (1958), and the final object is given by its Wolf-Rayet catalog number (van der Hucht, 2001). The spectral classifications are taken from a variety of sources, indicated in the notes below the table. The $V$ magnitudes reported are from Massey & Thompson (1991) for the MT # stars and from Straižys & Laugalys (2008) for others. The coordinates and infrared $JHK_{s}$ photometry are from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006). The reddening estimates are discussed in section 4.3 below. After we completed our observations we learned that the object MT 140 is in fact an intermediate mass object that is not a member of Cyg OB2 (Maíz Apellániz et al., 2016). We include our measurements here for completeness, but it is excluded from the multiplicity analysis.

Wright et al. (2015) describe the massive star content of the Cyg OB2 association, and they suggest that the association hosts 52 O-type stars and 3 Wolf-Rayet stars. Our sample includes 56 O-type stars and one Wolf-Rayet star, plus a number of luminous and/or early type B-stars. Thus, our target list should represent an almost complete sample of the most massive stars ($M/M_{\odot}>18$) in Cyg OB2 (missing the Wolf-Rayet Stars WR 144 and WR 146).

Our observations were made in three queue observing programs at the 8.1-m Gemini North Observatory during the 2005B, 2008A and 2008B observing semesters. Using the Near InfraRed Imager and Spectrograph (NIRI) with the ALTAIR adaptive optics (AO) system (Hodapp et al. 2003; Richardson et al. 1998; Roberts & Singh 1998b), we collected high resolution images ($0\farcs 022$ pixel^-1 with the $f/32$ camera) with a field of view (FOV) of approximately $22\arcsec\times 22\arcsec$. The only exception is for our $K$-band observations of MT 304 = Cyg OB2 #12. Due to its extreme IR brightness ($K=2.7$) MT 304 was observed with the shortest exposure time possible, and therefore, a smaller FOV ($11\arcsec\times 11\arcsec$) was used so that the data could be read out without over-exposing the images. The detector chip used the deep well setting for improved dynamic range, and the 2008 data were obtained with the ALTAIR field lens which improves the AO correction. The telescope sits on an altitude-azimuth mount, so that when NIRI is held fixed, the sky appears to rotate between frames. For these observations, NIRI was held fixed and the exposure times for each frame ranged between 0.02 s to 800 s in $K$ and between 0.1 s to 1869 s in $J$, depending of the brightness of the target star in each band in order to reach about half of the full well depth of the detector and achieve uniform S/N ratio measurements of the target stars.

Table 2 provides the central wavelength and the pass band for each filter. Every target was observed with the $K$ continuum filter, Kcon(209), to detect possible companions. The numbering corresponds to the central wavelength in hundreds of angstroms. We followed up on 43 stars with $J$-band observations to get additional color information on those systems with obvious companions. The 2005 data were obtained using the $J$ continuum filter, Jcon(112). The wider Jcon(121) filter was used for the 2008 observations because the companions appear fainter in the $J$-band than in the $K$-band. The seven targets observed during the 2005B semester were also imaged with the $H$ continuum filter, Hcon(157), with the exception of MT 304 which was only observed in $J$ at the time. These filters all have narrow pass bands that were needed because the stars are so bright in the infrared.

Each observation consisted of approximately 90 frames. Table 2 (given in full in the electronic version) lists the observation dates of the beginning of the first exposure and the number of frames combined to produce the final co-added image for each filter. Each target was observed at nine dither positions, set up on a $3\times 3$ grid, offset by about 50 pixels and with 10 exposures at each position. For the cases where the observations were taken over two nights, observations from each night were combined individually and also combined together. For the detection of sources, the images from each night were analyzed separately due to differences in image quality, but only data from one night were used for photometric and astrometric measurements (denoted by *). For A 25 in $K$ and A 41 in $J$, we analyzed the combined image from both nights (denoted by ^C) because they were of comparable quality. The fourth and fifth columns give the Strehl ratio and full-width at half-maximum (FWHM), respectively, of the point spread function associated with the primary target. These were determined using the IDL Strehl ratio meter code¹¹1http://www2.keck.hawaii.edu/optics/aochar/Strehl_meter2.htm written by M. van Dam.

In addition to the NIRI $K$-band observation, MT 421 was observed with the Palomar High Angular Resolution Optics (PHARO; Hayward et al. 2001) camera and the Palm-3000 AO system (Dekany et al., 2013) on the 5-m Hale telescope in 2009 July. We were able to get observations in all three IR bands, $J$, $H$, and $K_{S}$, with a field of view comparable to that of NIRI ($\sim 25\arcsec\times 25\arcsec$). The filter information for PHARO is also listed in Table 2. The PHARO images provide a pixel scale of $0\farcs 025$ pixel^-1 (Hayward et al., 2001).

The NIRI data were reduced using the tools provided as part of the Gemini reduction package in IRAF. With the images rotated, reduced, and the data quality robustly quantified through the various reduction steps, we used two different combining programs to co-add all of the frames. Most of the images were co-added using the IRAF tool IMCOADD to derive an average image taking into account the bad pixel mask. In the cases where IMCOADD failed (i.e., poor seeing, observations over multiple nights, or blended point spread functions), GEMCOMBINE was used with manual input of the central star pixel position. GEMCOMBINE produces a slightly different median image than the mean coadded IMCOADD, but the capability of allowing the user to define the pixel shifts makes the final co-added image better aligned than when IMCOADD fails. The final images from GEMCOMBINE and IMCOADD produce a slightly larger field of view than the $22\arcsec\times 22\arcsec$ FOV of a single frame, but depending on the observing conditions (e.g., exposure time and observations spanning multiple nights) some fields can be larger than others. The PHARO data were reduced by debiasing, flat fielding, bad pixel correction, and background subtraction and then shift-and-added to create a single image.

We identified possible point sources by visually inspecting each frame using SAO Image display software. This proved more successful than automated methods due to the abundance of hot pixels from the IR detector confused as point sources. The faintest companions that we detect ($\Delta K\approx 9$ mag) have signals that are just above the threshold set by the readnoise of the camera and the number of coadded frames. We identified at least one source in addition to the main target in each $K$-band frame through visual inspection. After identifying each point source and estimating the approximate pixel position of its peak, we used SExtractor (Bertin & Arnouts, 1996) to find each source and measure the centroid position and relative brightness. The positions were determined from the XWIN_IMAGE and the YWIN_IMAGE keywords in SExtractor. The relative flux returned by SExtractor is measured using the FLUX_APER parameter, which estimates the flux above the background within a circular aperture. We used nine aperture diameters on each star to create an enclosed energy curve. For close systems with blended point spread functions (PSFs) ($\rho\leq 0\farcs 1$), we used a PSF deconvolution program, FITSTARS (ten Brummelaar et al., 1996, 2000), to measure the differential magnitude and separation.

We present the astrometric and photometric results for all the stars in Table 4 (given in full in the electronic version). The relative magnitudes and positions are determined with respect to the target stars. The columns of Table 4 give the main target name, the angular separation $\rho$ and position angle $\theta$ (measured east from north) of the companion, its celestial coordinates, the magnitude difference and uncertainty in $J$, $H$, and $K$, the probability of chance alignment with a background field star $P_{ca}$ (see section 4.2), the identification number in the UKIRT Infrared Deep Sky Survey (UKIDSS) (Lawrence et al., 2007), and notes indicating other names, correspondence in another field, or measurement by FITSTARS (FS). The first row for a given target corresponds to the bright central star, and succeeding rows list data where available for each detected companion star (arranged in order of increasing separation).