The Coordinated Radio and Infrared Survey for High-Mass Star Formation (The CORNISH Survey). I. Survey Design

Advances in various widefield technologies have led to a new generation of Galactic plane surveys. Those with relevance here, and to star formation in particular, are listed in Table 1. Leading the way is the GLIMPSE I (Galactic Legacy Infrared Mid-Plane Survey Extraordinaire) survey (Benjamin et al. 2003; Churchwell et al. 2009). This Spitzer IRAC Legacy Programme covered the inner Galaxy (10° < l < 65° and -65° < l < -10°, |b| < 1°) at 3.6, 4.5, 5.8, and 8.0 μ m. GLIMPSE is 2 orders of magnitude more sensitive (limit at L -band (3.6 μm) ≈ 14–15 magnitude, although usually confusion limited) and has 10 times higher spatial resolution (1.4–1.9'') than any previous mid-IR survey. It has catalogued over 49 million sources in a wavelength regime which preferentially selects sources with hot circumstellar dust emission, such as young and evolved stars (e.g., Robitaille et al. 2008). More embedded and cooler sources are the subject of the MIPSGAL survey with Spitzer (Carey et al. 2009) and the PACS element of the Hi-GAL survey with Herschel (Molinari et al. 2010).

An even deeper probe of the general stellar population is becoming available via the deep near-IR Galactic Plane Survey (GPS) that is part of the UK IR Deep Sky Surveys (UKIDSS) program (Lucas et al. 2008). This survey is about 3 mag more sensitive and has two to three times better resolution than the 2MASS all sky near-IR survey (Skrutskie et al. 2006) and will detect upwards of a billion sources. The addition of near-IR data, where photospheric and scattered light contributions dominate, allows a much better separation and characterization of the general stellar population than from mid-IR colours alone. Near-IR surveys are also being used to map the extinction due to molecular clouds (e.g., Rowles & Froebrich 2009). The optical Hα survey IPHAS (Drew et al. 2005) is providing both stellar characterisation and three-dimensional extinction maps, and tracing nebular emission in less obscured regions of the northern plane with the southern VPHAS+survey²² scheduled to do the same in the south.

To place the stellar populations in a Galactic context, commensurate studies of the various components of the interstellar medium (ISM) are required. For the study of star formation in particular, the molecular component is provided in part by the BU-FCRAO ¹³CO 1-0 Galactic Ring Survey (Jackson et al. 2006), which covers over half of the northern GLIMPSE area with extensions over much of the northern midplane (Mottram & Brunt 2010). The mostly optically thin ¹³CO 1-0 data traces the distribution and dynamics of the cold molecular gas. The cool dust in these molecular clouds can also be mapped via its submillimeter continuum emission. This is currently being achieved with the ATLASGAL survey (Schuller et al. 2009) at 870 μ m and BOLOCAM survey at 1100 μ m (Aguirre et al. 2011), while the forthcoming JCMT Galactic Plane survey²³ will map the continuum at 450 and 850 μ m with higher sensitivity and resolution. Together with the SPIRE element of the Hi-GAL survey with the Herschel Space Observatory satellite (Molinari et al. 2010) this will provide temperature and dust emissivity information across the GLIMPSE region.

The more widely distributed atomic hydrogen component is covered by the International GPS at about 1^' resolution.²⁴ This is made up of the VLA GPS (VGPS, Stil et al. 2006), Southern GPS (SGPS; McClure-Griffiths et al. 2005) and Canadian GPS (Taylor et al. 2003). A subsection of the plane is covered by the much deeper GALFA survey of H I at 4^' resolution using the Arecibo telescope (Peek et al. 2011). The H I component traces where most of the mass of interstellar gas is located and its velocity structure. Also, 21 cm absorption lines can help to solve the near/far distance ambiguity in relation to the kinematic distances derived towards the embedded H II regions (Sewilo et al. 2004; Fish et al. 2003) and dense molecular clouds (Jackson et al. 2002; Gibson et al. 2005; Busfield et al. 2006; Roman-Duval et al. 2009). This combination of atomic and molecular kinematic data can therefore provide the three-dimensional Galactic setting of the neutral gas.

The one major component missing from the multiwavelength surveys of the Galaxy is that of the ionized gas. No existing radio continuum survey has sufficient resolution, depth, or coverage to provide the data that would complete the picture of the Galaxy. Ionized gas arises in nebulae and stellar winds around hot stars and is crucial to understanding key phases of the early and late evolution of high- and intermediate-mass stars. Dense, photoionized regions around young and evolved stars are often optically thick with a turnover frequency of around a few GHz, where the radio continuum spectral index α (S_ν ∝ ν^α) changes from the optically thick value of +2.0 to the optically thin slope of -0.1. Ionized stellar winds have a positive spectral index α ∼ +0.6, and so it is also much more efficient to carry out systematic searches for these sources at high frequencies. Furthermore, the resolution should be comparable to the arcsecond resolution of the IR surveys to allow IR counterparts to be uniquely identified. It was to fill this need for a high frequency, high resolution radio continuum survey covering the area of the Spitzer survey that the concept of the Co-Ordinated Radio 'N' Infrared Survey for High-mass star formation (CORNISH)²⁵ survey was conceived.

Many single-dish surveys of the Galactic plane have been carried out at 5 GHz with a resolution of a few arcminutes that is totally inadequate to complement the new generation of IR surveys (e.g., Altenhoff et al. 1978; Haynes et al. 1978; Gregory & Condon 1991; Griffith et al. 1994). Most higher resolution interferometric surveys have been carried out at the relatively low frequency of 1.4 GHz (Zoonematkermani et al. 1990; Becker et al. 1990; Condon et al. 1998; Giveon et al. 2005). Although the multiconfiguration MAGPIS survey (Helfand et al. 2006) and MGPS-2 survey (Murphy et al. 2007) are good for tracing the optically thin thermal emission from extended, evolved H II regions and non-thermal supernova remnants, they are not appropriate for the study of dense, thermal sources due to the unfavourable spectral index and compactness. Becker et al. (1994), with infills by Giveon et al. (2005) and White et al. (2005), have surveyed part of the inner Galaxy (-10° < l < 42°, |b| < 0.4°) at 5 GHz with the VLA in C configuration, giving a resolution of 4'' × 9''. However, these surveys cover only one-fifth of the northern GLIMPSE region, and the spatial resolution is poorer than that of Spitzer IRAC, which causes problems with source identification. At this resolution the vast majority of their sources are unresolved, and thus we have no morphological information. In the southern hemisphere, the AT20G survey at 20 GHz and ∼30'' resolution (Murphy et al. 2010), primarily designed to study the extragalactic radio source population, is also detecting many H II regions, but again not at sufficient depth or resolution to address the science opened up by current surveys of the Galactic plane at infrared wavelengths.

Our main science driver is the study of the formation of massive stars. These stars impart large amounts of UV radiation, wind energy, strong shocks, and chemically enriched material into the ISM during their lifetime. This is particularly true in starbursts, when they form in dense concentrations (Leitherer 1999). However, relatively little is known about the physics that controls the formation of stars more massive than about 8 M_⊙ (10³ L_⊙) and hence about the upper initial mass function (IMF). The fast collapse and contraction time scales and extreme conditions resulting from increased turbulence and radiation pressure makes the theoretical treatment of their formation more challenging than for solar-type stars (e.g., Behrend & Maeder 2001; Yorke & Sonnhalter 2002; McKee & Tan 2003; Bonnell & Bate 2006). Their rarity and predilection to form in dense clusters makes observational studies more challenging and high resolution is a prerequisite (e.g., Beuther et al. 2002).

Current numerical simulations indicate that accretion via a disk can continue despite strong radiation pressure (Krumholz et al. 2009; Kuiper et al. 2010), while submillimeter interferometry reveals evidence of flattened, disklike structures (e.g., Patel et al. 2005; Torrelles et al. 2007) around massive young stellar objects (MYSOs). Although these deeply embedded IR objects are already very luminous, they have yet to ionize the surrounding ISM, most likely due to ongoing accretion keeping them swollen and cool (Hoare & Franco 2007; Hosokawa & Omukai 2009; Davies et al. 2011). MYSOs drive highly collimated ionized jets (Martí et al. 1998; Curiel et al. 2006) and equatorial disk winds (Hoare et al. 1994; Hoare 2006; Gibb & Hoare 2007) which are manifest as compact thermal radio emission (a few mJy for nearby examples) with a spectral index close to +0.6. What drives these jets and winds at speeds of a few hundreds of km s^-1 (e.g., Bunn et al. 1995) is not known. A possible scenario is that the radio jets are accelerated and collimated by an MHD mechanism as in low-mass young stellar objects early on, and these give way to less-collimated radiation pressure dominated flows (Drew et al. 1998) at later stages, akin to the scheme outlined by Beuther & Shepherd (2005).

The strong radio emission from H II regions, formed once the star has contracted down to the main sequence and heated up, is a very useful probe of massive star formation (Hoare et al. 2007). It is unaffected by extinction and can thus be used to measure the formation rate of OB stars right across a galaxy (e.g., Condon, Cotton & Broderick 2002). The radio brightness of H II regions can be used to infer the flux of Lyman continuum photons from the exciting star or cluster and hence constrain the spectral type and mass of the star(s). H II regions provide one of the clearest signatures of spiral structure which can be related to the global distribution of molecular and atomic hydrogen from complementary surveys. This is particularly important in the context of the emerging picture of our Galaxy as having two stellar arms linked to the central bar while having four gaseous star-forming arms (Benjamin 2009).

High-resolution studies of Galactic objects have revealed both large numbers of ultracompact objects (UCH IIs) and a preponderance of cometary morphologies (Wood & Churchwell 1989; Kurtz et al. 1994; Walsh et al. 1998). These studies spawned a reexamination of the dynamical models of H II regions (Van Buren & Mac Low 1992; De Pree, Rodríguez & Goss 1995; Williams, Dyson & Redman 1996). The latest cometary models (Arthur & Hoare 2006) require the young OB star to be located in a density gradient to explain the cometary shape, together with a strong stellar wind to give the limb-brightened appearance and an element of proper motion up the density gradient to reconcile the ionized and molecular velocity structure. Such a picture is consistent with one where the formation of these massive stars was triggered by the passage of an external shock wave through the molecular cloud.

There are many instances where the expansion of H II regions appears to be triggering formation of new episodes of massive star formation (e.g., Deharveng et al. 2005; Urquhart, Morgan & Thompson 2009; Moore et al. 2007; Zavagno et al. 2010; Thompson et al. 2012). This is readily apparent in the 8 μ m images from the GLIMPSE survey, where the strong polycyclic aromatic hydrocarbon (PAH) emission delineates the photodissociation region (PDR) at the interface between the ionized region and the molecular cloud. Furthermore, the 24 μ m images from MIPSGAL show up the warm dust heated by Lα within the ionized zone and correlates well with the radio free-free emission (e.g., Watson et al. 2008). Younger phases such as mid-IR bright MYSOs or maser sources are often seen just ahead of the ionization front, sometimes within IR dark regions (Hoare et al. 2007). What fraction of massive star formation occurs in this way, as opposed to spontaneous collapse, is unclear. Conversely, the action of outflows, expanding H II regions, and stellar winds can also disperse the remaining molecular cloud and quench further star formation (e.g., Franco, Shore, & Tenorio-Tagle 1994). How these opposing feedback mechanisms affect the resultant star formation rate and efficiency is unknown.

Many aspects of massive star formation studies suffer from the lack of a large and unbiased sample of both UCH IIs and MYSOs. Samples based on the IRAS Point Source Catalogue (e.g., Kurtz et al. 1994; Molinari et al.1996; Sridharan et al. 2002) are heavily biased towards bright, isolated sources due to the poor spatial resolution of IRAS, which resulted in confusion in the densest regions of the plane, precisely where massive stars form. The situation for mid-IR bright MYSOs has been alleviated by the Red MSX Source (RMS) survey (Lumsden et al. 2002) that utilized the 18^'' resolution mid-IR survey by the MSX satellite (Price et al. 2001) and high resolution multiwavelength follow-up to yield a large, well-selected sample of MYSOs. This is fairly complete out to about 15 kpc for O star luminosities (Urquhart et al.2011; 2012), and the GLIMPSE survey can easily push this down to the B star range (Robitaille et al. 2008). However, the IR spectral energy distributions (SEDs) of MYSOs and UCH IIs are very similar, and therefore radio continuum data are vital to distinguish these two evolutionary phases (Fig. 1) (e.g., Urquhart et al. 2007). Hence, radio observations are crucial in distinguishing the "radio-quiet" MYSOs from the "radio-loud" UCH II regions. The large numbers of very red objects in the GLIMPSE survey with colors of embedded YSOs makes radio follow-up of each one impractical. The uniform coverage of the CORNISH survey makes this automatic.

Zoom In Zoom Out Reset image size

The RMS survey also found many UCH II regions, but such IR-selected samples are biased against very embedded objects such as those found in IR dark clouds (IRDCs) (Egan et al. 1998). Many examples of methanol maser sources (Ellingsen 2006), extended green objects (EGOs) (Cyganowski et al. 2008), and faint 24 μ m MIPSGAL sources (Rathborne et al. 2010) are showing up in IRDCs. Is there radio continuum emission associated with these very early phases of massive star formation? Cyganowski et al. (2011) performed deep radio continuum observations of a sample of EGOs and found mostly weak or nondetected sources consistent with jets or winds in the diagnostic plots of Hoare & Franco (2007). The rarity of known MYSOs with luminosities above 10⁵ L_⊙ raises the question of whether early O stars ever go through a detectable mid-IR bright MYSO phase, or whether they manifest themselves firstly as UCH IIs.

The clean selection function of the CORNISH survey combined with dynamical models and population simulation will allow us to determine the lifetime of this phase as a function of luminosity (e.g., Davies et al. 2011). Together with similar work for MYSOs from the RMS survey and earlier phases from the Hi-GAL and SCUBA2 surveys, this will yield the first comprehensive picture of the evolutionary scheme from high mass star formation. Modelling the flux, size, and shape distributions of the UCH II region population will test the dynamical models of UCH II regions and their relative locations in the dense molecular cloud material will enable statistical tests of triggering scenarios. These tests are only possible from uniform, wide-area surveys with sufficient sensitivity and resolution.

In the high mass star formation context, a spatial resolution of about 1'' is needed in a complementary radio survey for several reasons. Firstly, the IR source density in regions of massive star formation is very high in dense star forming regions. Hence, to accurately identify the IR counterpart to a compact radio source, arcsecond resolution is needed. Secondly, UCH IIs themselves often appear in complexes with separations of a few arcseconds (e.g., Sewilo et al. 2004). Finally, as previous surveys have shown, to reveal the structure of most UCH IIs, a resolution finer than an arcsecond is needed.

Zijlstra & Pottasch (1991) estimated that there should be about 23,000 planetary nebulae (PN) in the galaxy, although only just over 3000 have been identified (Frew & Parker 2010). Combined radio and IRAS searches have been successful in turning up new PN (Van de Steene & Pottasch 1995; Condon, Kaplan, Terzian 1999). A high-resolution radio survey is particularly suited to discovering young, compact PN that are heavily reddened by line-of-sight extinction in the Galactic plane, and we expect about a thousand of these in the survey area. Their radio flux densities are expected to be in the range of 5–50 mJy for typical distances with sizes of a few arcseconds. A complete sample over a large region of the Galactic plane will constrain the density of PN and hence their formation rate and distribution in the Galaxy. This has implications for models of post-AGB (asymptotic giant branch) evolution and mass loss which are not well constrained (e.g., Casassus & Roche 2001).

As in the case of UCH IIs, the immediate progenitors of PN, the proto-PN (PPN), are also radio-quiet since the central star has not yet reached 30,000 K. The radio-quiet PPN will also have very similar mid-IR colors to the radio-loud PN, and so the only way to distinguish them is through their lack of radio emission. This is another illustration of the usefulness of nondetections as well as detections in the radio survey. PPN hold the key to understanding the envelope ejection process and testing the interacting winds models for the shaping of PN (e.g., Mellema & Frank 1995). It is not known whether aspherical mass loss at the end of the AGB is due to AGB star spin-up, detached or common-envelope binaries (Mastrodemos & Morris 1999), or dynamically important magnetic fields (Gardiner & Frank 2001). If we take the birth rate of PN to be about 1 per year (Kwok 2000) and the same for PPN with a PPN lifetime of 1500 years, then we expect about 250 PPN in our survey region.

Mass loss from the later evolutionary stages of massive stars (e.g., OB supergiants, Wolf-Rayet stars, and luminous blue variables) is of great importance in the evolution of the ISM in our own and other galaxies, through both the kinetic power deposited by their stellar winds and their strong ionizing radiation fields. Thermal radio emission plays a key role in determining mass loss rates for the evolved massive stars since it arises at large distances from the underlying stars where the outflow has attained a constant velocity (e.g., Wright & Barlow, 1975). Renewed interest in this important input to the ISM has arisen due to clumpiness of the winds (Fullerton et al. 2006). Radio deduced mass loss rates may also provide evidence of the bistability jump and a drop in ratio of wind terminal velocity and the escape velocity (Lamers, Snow & Lindholm 1995), and perhaps also be important to understanding the LBV phase of stellar evolution (Vink & de Koter 2002).

Many of these massive stars exhibit nonthermal emission in addition to the thermal emission from their stellar winds (Bieging et al. 1989), which has been shown to be associated with massive binary systems, with the emission arising where the stellar winds of the two companions collides. The archetype of these colliding-wind binaries is WR140, where the nonthermal emission has led to the determination of all the orbit parameters, the distance, and the wind-momentum and mass ratios (e.g. Dougherty et al. 2005; Dougherty et al. 2011). Among WR stars that exhibit nonthermal emission, over 90% are binaries (Dougherty & Williams, 2000), and there is now evidence that a similar fraction of nonthermal emitters is also observed in O-star binaries (de Becker 2007). Challenges remain for the colliding-wind model. Not all stars with nonthermal emission are found to be binaries, and many binaries exhibit thermal emission, which among short-period systems might be attributed to absorption of the synchrotron power. Alternative mechanisms for accelerating electrons to necessary relativistic speeds have been proposed, e.g., shocks within the stellar winds, that work in single star scenarios.

Studies of both the thermal and nonthermal emitters are limited by the small sample size of the detected systems, and the samples are often biased via spectral type, optical luminosity, or known binary systems. A large, unbiased sample of radio-detected evolved OB stars has previously not been available. Since they inhabit the densest parts of the plane, there are likely many evolved OB stars hidden from optical view by high extinction. For a 5 GHz detection limit of 2 mJy, the limiting radio luminosity is 2.4 × 10¹⁸ (D/kpc)² erg s^-1 Hz^-1. Using the empirical relationship between and L_bol from Garmany & Conti (1984) and the relationship between L_6 cm and from radio flux from Bieging et al. (1989), we expect this survey to detect all stars with a bolometric luminosity L_bol≥2 × 10⁶ L_⊙ within ∼1 kpc.

Sensitive radio observations have revealed the ubiquity of high-energy processes across the radio H-R diagram (Güdel 2002). As well as dynamo effects in convective envelopes, high resolution radio observations have revealed large-scale magnetospheric structures in various stellar types, including both low and intermediate mass pre-main-sequence stars. Radio observations reveal dozens of such objects in Orion, and Chandra now finds over 1000 active stars in Orion (Garmire et al.2000). With 5 GHz flux densities reaching several tens of mJy, these sources will be detected to several kpc. Particularly intriguing is the role of magnetic fields in intermediate mass pre-main-sequence stars, which are not expected to have convective envelopes and hence large scale dynamos (e.g., Garrington et al.2002). Combined radio, IR, and X-ray studies will reveal how common these magnetic phenomena are.

Mass transfer in close binaries often leads to strong activity in radio and other wave bands. RS CVn stars, cataclysmic variables, and others are likely to be detected in the survey. However, the most important class of active binary are the so-called microquasars where the accreting object is a black hole or neutron star (Mirabel & Rodríguez 1999; Fender 2002). These produce relativistic jets analogous to AGN where ejections of plasma can be followed in real time (Fender et al. 1999). These objects are perhaps the best laboratories for studying the relativistic jet formation process and test magnetohydrodynamic models (e.g., Meier, Koide, & Uchida 2001). The key signature of these jets is strong synchrotron emission in the radio band, and since they often originate from massive stars one expects to find them close to the Galactic plane and in the locale of giant molecular clouds. In such dense environments, the X-ray and optical signatures of accretion may be heavily absorbed, preventing the detection of accretion by more conventional means. Hence, a coordinated radio and IR survey should uncover many new examples of this phenomena.

There is a strong relationship between the X-ray state and radio behavior in these objects, which has led to the concept of disk-jet coupling. Several black hole candidate objects have flat or inverted radio spectra when in the low/hard X-ray state, indicative of the presence of a self-absorbed compact jet supported by a hot X-ray corona. The jets have been resolved by VLBI in GRS1915 + 105 (Dhawan et al. 2000) and Cygnus X-1 (Stirling et al.2001). When the accretion rate increases, the thermal disk becomes dominant and the X-rays become softer. Radio outbursts often occur during the transition between hard and soft states (Fender, Homan and Belloni 2009) due to the formation of a strong shock in the outflowing jet, and radio interferometer observations are able to follow the evolution of the flow (e.g., Fender et al.1999). Our survey has the resolution to resolve such structures, and at high accretion rates the objects are detectable throughout the disk of the galaxy.

There will be a significant number of background extragalactic sources detected during the survey. Using equation A2 in Anglada et al. (1998), which is based on the source count data from Condon (1984), at our completeness level we expect to detect about 2500 extragalactic sources in the 110 square degree area. The vast majority of these radio galaxies will be at too high a redshift to determine distances via H I detections with current telescopes (Magliocchetti et al. 2000; Obreschkow et al. 2009). As well as penetrating the deepest part of the zone of avoidance, these sources will be useful in other respects. The most compact ones are potential phase calibrators for use in future interferometric studies in the Galactic plane. At 5 GHz the CORNISH survey will preferentially pick up flat spectrum sources, which are potential mm-interferometric phase calibrators. Extragalactic sources that are also visible in the optical or near-IR can be used to bring together the radio and optical coordinate reference frames, which is important for multiwavelength high resolution studies in the Galactic plane.

Radio surveys can also deliver a set of extragalactic background sources that can act as pencil beam probes of molecular clouds in follow-up absorption line studies (Greaves & Nyman 1996). Sufficiently strong sources could be used for Zeeman absorption line measurements of the magnetic field strength (Crutcher 1999). For instance, radio sources that lie behind dense, starless molecular cloud cores can be used to probe the initial conditions for star formation in these dense clouds with future sensitive instruments such as the SKA. Polarized background sources can also be used to probe the magnetic field of the Galaxy, individual H II regions, and supernova remnants via Faraday rotation (Taylor et al. 2009). High-frequency data are useful in disentangling components in sources with complex Faraday depth spectra (O'Sullivan et al. 2012).