SMA^* OBSERVATIONS OF GOODS 850−11 AND GOODS 850−13: FIRST EXAMPLES OF MULTIPLE SUBMILLIMETER SOURCES RESOLVED BY AN INTERFEROMETER

Since the first discoveries of distant submillimeter galaxies (SMGs; Smail et al. 1997; Barger et al. 1998; Hughes et al. 1998; Eales et al. 1999), tremendous progress has been made in understanding their nature and their role in galaxy evolution. With single-dish submillimeter surveys, we are now able to resolve approximately 30% of the 850 μm background into point sources with S_850 μm ≳ 2 mJy and constrain their number counts fairly well at this bright end (e.g., Coppin et al. 2006). However, the low resolution of single-dish telescopes and the associated large positional uncertainties make the identification and follow-up of these sources quite difficult. Roughly 60%–70% of bright SMGs have counterparts in deep radio interferometric images (Barger et al. 2000; Ivison et al. 2002; Chapman et al. 2003b) and their positions are known with subarcsec accuracy. Spectroscopic follow-up of the radio-identified SMGs shows a redshift distribution between z ∼ 1.5–3.5 and that the SMGs dominate the total star formation in this redshift range (Chapman et al. 2003a, 2005). The recent advent of the Submillimeter Array (SMA; Ho et al. 2004) further helps to identify the radio-faint SMGs, and the redshift distribution has been extended to z>4 (Iono et al. 2006; Wang et al. 2007; Younger et al. 2007, 2009; Cowie et al. 2009). In addition to the redshift identifications, detailed follow-up observations have been made to study the properties of the SMGs in the X-ray, near-infrared, mid-infrared, and molecular line transitions (e.g., Alexander et al. 2003a, hereafter A03; Swinbank et al. 2004; Pope et al. 2008; Yun et al. 2008; Greve et al. 2005; Tacconi et al. 2006).

The follow-up studies of SMGs have been overwhelmingly focused on the brighter sources (S_850 μm ≳ 5 mJy), which can easily be detected by single-dish telescopes. The nature of fainter SMGs that comprise the bulk of the submillimeter background has been much less explored. Surveys of lensing cluster fields yield small samples of sub-mJy sources (Blain et al. 1999; Cowie et al. 2002; Knudsen et al. 2008). The number counts indicate that the background is dominated by sources with S_850 μm ∼ 1 mJy and that the full resolution of the background requires detections of ∼0.1 mJy sources. Stacking analyses statistically detect faint SMGs at 500 μm to 1.2 mm, and various attempts have been made to study the redshift distribution of these faint SMGs (e.g., Wang et al. 2006; Serjeant et al. 2008; Marsden et al. 2009; Penner et al. 2010). However, the results from these stacking analyses have not converged to a consistent picture. A full understanding of the submillimeter background population will likely require next-generation interferometers, such as the Atacama Large Millimeter/Submillimeter Array (ALMA).

Fortunately, it is now possible to detect more typical submillimeter sources with the SMA. The recent upgrade of the SMA to a 4 GHz bandwidth not only greatly boosts its continuum sensitivity, but it also makes the calibrations with fainter quasars easier than before. This provides a first glance of what we may find in deep ALMA surveys and what issues may be present in current studies of SMGs. In this Letter, we report new SMA 340 GHz continuum observations of two SMGs, GOODS 850−11 and GOODS 850−13 (Wang et al. 2004, hereafter W04), aka GN12 and GN21 (Pope et al. 2005) in the Great Observatories Origins Deep Survey-North (GOODS-N; Giavalisco et al. 2004). Interestingly, both single-dish sources are resolved by the SMA into multiple physically unrelated galaxies. To our knowledge, these are the first examples of resolved, unrelated, multiple sources in the SMG population.

The SMA observations of GOODS 850−11 and 13 were carried out on 2009 December 30 and 31, respectively. Seven of the eight SMA antennas were available in the compact configuration. Two sidebands of 4 GHz each were centered at 334 and 346 GHz. Titan was observed to provide flux calibration. The bright radio source 3C 273 was observed to provide passband calibration. Quasars 0958+655 and 1642+689, which are 174 and 251 from GOODS-N, respectively, were observed every 15 minutes for complex gain calibrations. The 225 GHz opacity was excellent, ∼0.04–0.05 the first night and ∼0.05–0.07 the second night. The averaged single-sideband system temperature was 380 K on both nights. The effective integration was ∼4.8 and 7.2 hr per antenna on GOODS 850−11 and 13, respectively.

The calibration and data inspection were performed with the IDL-based Caltech package MIR modified for the SMA. Continuum data were generated by averaging the spectral channels after the passband phase calibration. Both gain calibrators were used to derive gain curves. However, we compared the results made with just adopting one calibrator for consistency checks, and we did not find a systematic difference. Flux calibrations were performed using data taken under conditions (time, hour angle, and elevation) similar to that of the flux calibrator. The flux calibration error is typically within ∼10% with this method.

The calibrated interferometric visibility data were exported to the package MIRIAD for subsequent imaging and analysis. The visibility data were weighted inversely proportional to the system temperature and Fourier transformed to form images. The "robust weighting" of Briggs (1995) was also applied, with a robust parameter of 1.0, to obtain a better balance between beam size and signal-to-noise ratio (S/N). The images were CLEANed around detected sources to approximately 1.5× the noise level to remove the effects of the sidelobes. The noises measured from the CLEANed images are 0.72 and 0.68 mJy for the images of GOODS 850−11 and 13, respectively. The synthesized beams are 28 × 20 and 23 × 20 in the images of GOODS 850−11 and 13, respectively. The images were corrected for the SMA primary beam response. All our fluxes and flux errors are primary beam corrected. Source positions and fluxes were measured by fitting the image with point-source models using the MIRIAD IMFIT routine.

We present the SMA and multiwavelength images of GOODS 850−11 and 13 in Figures 1 and 2, their SMA astrometry, long-waveband fluxes, infrared luminosities, and redshifts in Table 1, and their optical to near-infrared spectral energy distributions (SEDs) in Figure 3. The 1.4 GHz and 24 μm fluxes in Table 1 are adopted from Morrison et al. (2010) and the GOODS Spitzer Legacy Program DR2+ (M. Dickinson et al. 2011, in preparation), respectively, if the sources are detected. We carried out our own measurements for weakly detected sources. Below we describe the results.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.1. GOODS 850−11

GOODS 850−11 is known to be located in a region crowded with radio and X-ray sources (W04). The SMA detected two sources, GOODS 850−11a and 11b (see Figure 1), at significance levels of 5.6σ and 4.6σ, respectively. In W04 we found a SCUBA point-source flux of S_850 μm = 10.8 mJy and a total flux of 12.7 mJy in a 30'' region. The sum of the SMA fluxes (Table 1) of the two sources is consistent with our SCUBA measurements. Pope et al. (2006, hereafter P06) found an 850 μm SCUBA flux of 8.6 mJy, and they considered GOODS 850−11a as an unambiguous identification (see also Greve et al. 2008; Chapin et al. 2009). A03 also considered this source to be the X-ray counterpart of the SCUBA source. Our SMA measurements show that this identification only accounts for ∼50% of the SCUBA flux in this region. The brightest 24 μm source in this field (a blue ACS galaxy) is also detected in the radio and X-ray, but not in the submillimeter. It is at z = 2.004 and is consistent with an AGN with warm dust.

GOODS 850−11a. The source is detected at 1.4 GHz by the latest Very Large Array (VLA) imaging (Morrison et al. 2010). It is also detected in the 2 Ms Chandra image (source number 239 in Alexander et al. 2003b). Because it is severely blended with nearby sources in the MIPS 24 μm image, its 24 μm flux is highly uncertain. It does not have a spectroscopic redshift from our redshift survey (Barger et al. 2008). To estimate its redshift, we carried out photometric redshift fitting using the EAZY package (Brammer et al. 2008). We adopted the U-band flux from Capak et al. (2004), the ACS fluxes from the GOODS v2.0 catalog, the J-band flux from a deep CFHT image that we will describe elsewhere, the HST WFC3 F140W flux from A. Barger et al. (2011, in preparation), and the K_S to Infrared Array Camera (IRAC) fluxes from Wang et al. (2010). We adopted the default set of SED templates of Blanton & Roweis (2007) plus a dusty starburst model (t = 50 Myr, A_V = 2.75), all of which are provided in the EAZY package (see Brammer et al. 2008 for more details). These templates all include certain amounts of reddening, and we further reddened them by A_V = 0, 0.5, and 1.0 with the extinction law of Calzetti et al. (2000) to account for very dusty sources. We allowed for any combinations of the above templates in the fitting. The photometric redshift result for GOODS 850−11a is z = 3.11, and the best-fit SED is presented in Figure 3. This redshift agrees with the photometric redshift in P06, which is 3.1.

GOODS 850−11b. The source is detected by the VLA at 1.4 GHz but not by the 2 Ms Chandra imaging. Similar to GOODS 850−11a, it is severely blended with nearby sources in the MIPS 24 μm image, and thus its 24 μm flux is highly uncertain. It has a spectroscopic redshift of 2.095 (Reddy et al. 2006).

Zoom In Zoom Out Reset image size

3.2. GOODS 850−13

Since the commissioning of the SMA, it has been used for follow-up observations of SMGs, either for identification (e.g., Iono et al. 2006; Wang et al. 2007; Younger et al. 2007; Cowie et al. 2009; Younger et al. 2009) or for morphology (e.g., Younger et al. 2008). This is the first time that the SMA has discovered new SMGs. The multiple SMA detections illustrate the limitations of identifying SMGs in any wavelengths other than the submillimeter itself. Both sources had radio, 24 μm, and X-ray identifications in P06 and A03. All of the previously proposed identifications are only partially correct, i.e., they are all legitimate SMGs, but the submillimeter fluxes and source numbers will be misinterpreted by up to a factor of three. If such cases are common, then our understanding of the SMG population is fundamentally flawed.

The effects of multiple SMGs are commonly included in single-dish number counts simulations (e.g., Eales et al. 2000; Scott et al. 2002; Coppin et al. 2006; W04). However, its importance (relative to other effects such as Eddington bias) has not been directly demonstrated by observations. So far only sensitive millimeter/submillimeter interferometric observations can reveal the existence of multiple SMGs like GOODS 850−11 and 13, since existing single-dish telescopes are still severely confusion limited (at >850 μm) or noise limited (at 450 μm). The SMA surveys of Younger et al. (2007, 2009) imaged 15 SMGs selected at 1.1 mm. They did not find evidence of multiple sources, consistent with the argument made by Ivison et al. (2007) that multiple sources are rare. However, the SMA surveys of Younger et al. have 345 GHz rms sensitivities of 1–2 mJy. Even if there are secondary sources with S_345 GHz ∼ 3–5 mJy in their survey fields, such sources would not be easily detected. Our SMA survey is the first one that is deep enough to reveal such cases.

After the SMA upgraded to the 4 GHz bandwidth, we observed five GOODS-N SMGs at similar depths (A. Barger et al., 2011 in preparation). Two of them are resolved into multiple sources and reported here. There is a third resolved source that will be reported by Barger et al. It is unclear whether this source is a merger or a physically unrelated pair. Even if we exclude the third source, the frequency of multiple sources in our SMA sample still seems unusually high. To better understand this, we performed Monte Carlo simulations that mimic our SCUBA and SMA surveys.

We first created 1000 simulated SCUBA images using the differential counts in Cowie et al. (2002) and Coppin et al. (2006), the "true-noise" map created in our GOODS-N SCUBA survey in W04 and the SCUBA beam in W04. We then extracted the simulated SCUBA sources. For each SCUBA source detected at >5 mJy and >4σ (the selection criteria for our SMA observations), we searched the input catalog for any >3 mJy (our SMA detection limit) sources within 17'' (the SMA primary beam HWHM) of the SCUBA position. In the simulations the mean number of >5 mJy sources in a W04 SCUBA area is 11.4 ± 3.9, where the uncertainty is the dispersion in the 1000 realizations and is nearly Poissonian. This value is consistent with the W04 SCUBA observations (15 sources). On the other hand, it is significantly larger than the cumulative count, as it is affected by blending and flux boosting. The measured counts take these effects out. The mean numbers of double and triple systems are 1.29 ± 1.33 and 0.06 ± 0.27, respectively.

We had observed 10 of the 15 >5 mJy SCUBA sources with the SMA. The most recent five of the 10 SMA observations were deep enough to detect >3 mJy sources. The earlier SMA observations were shallower, so there was a selection bias against SCUBA sources with multiple counterpart candidates. Given this bias and to be conservative, we only scale the above values of 1.29 ± 1.33 and 0.06 ± 0.27 by a factor of 10/15, rather than 5/15. We thus expect to find 0.86 ± 0.89 double and 0.04 ± 0.18 triple systems. The probability of finding one triple system like GOODS 580−13 is only 4%, inconsistent with the actual observations. Among the possible explanations, the most likely one is clustering of SMGs, which is not included in the simulations. This is plausible because the photometric redshift of GOODS 850−13a has a confidence range (Table 1) covering the redshift of 13b or 13c. This can be tested with future spectroscopic observations in the near-infrared or in the millimeter.

In the same simulations, we increased the SMA detection limit to 4 mJy, and we found that the probability of multiple systems dramatically decreases to ∼6%. This is consistent with the SMA survey results of Younger et al. (2007, 2009). By altering the details of the simulations, we also found that the above results are fairly insensitive to the following SCUBA and SMA observing strategies: (1) the source extraction (various detection thresholds), (2) flux measurement in the SCUBA map (simple aperture flux versus optimally filtered flux using the beam), (3) the shape of the SCUBA sidelobes (determined by the secondary chopping), and (4) the decision on where to point the SMA (as long as it is within the SCUBA positional uncertainty).

We can use our simulations to predict the early results of ALMA identifications of SCUBA sources. We adopt the primary beam HWHM of 85 of ALMA at 340 GHz. In the ALMA early science phase we expect to have at least 10 antennae, and we can detect 0.5 mJy sources in roughly one hour. If we point ALMA at a >5 mJy SCUBA source, the probabilities of detecting double and triple SMGs within the primary beam will be 29% and 6.5%, respectively. The combined fraction of ∼35% is very high. We therefore predict that multiple detections in early ALMA observations will be quite common.

At the beginning of this section, we raised the issue about incomplete identifications of SMGs in the X-ray, 24 μm, and radio. Based on our SMA observations and the above simulations, we believe that multiple systems and therefore incomplete identifications are common. Thus, we suggest that large numbers of single-dish sources should be re-identified with sensitive interferometric observations, even if the sources were previously thought to be securely identified.

We thank Y.-W. Tang, M. Gurwell, and the SMA staff for the help in acquiring and reducing the data, and the referee for the review. We gratefully acknowledge support from the National Science Council of Taiwan grants 98-2112-M-001-003-MY2 and 99-2112-M-001-012-MY3 (W.H.W.), NSF grants AST 0709356 (L.L.C.) and AST 0708793 (A.J.B.), the University of Wisconsin Research Committee with funds granted by the Wisconsin Alumni Research Foundation (A.J.B.), and the David and Lucile Packard Foundation (A.J.B.).