3.2. Multi-waveband follow-up studies
A great deal of telescope time has been spent so far to detect and study submm-selected galaxies in other wavebands. In many cases, rich archival data predated the submm observations: most notably in HDF-N (Hughes et al., 1998). Considerable data was also available in the fields of rich clusters (Smail et al. 1997, 1998a, 2002; Cowie et al., 2002), in the region of the Eales et al. (1999, 2000) surveys in Canada-France Redshift Survey (CFRS) fields, which include the Groth Strip, and in the deep Hawaii survey fields (Barger et al., 1999a). The results of follow-up deep optical and near-IR imaging (Frayer et al., 2000) and spectroscopy (Barger et al., 1999b), mm-wave continuum imaging (Downes et al., 1999a; Bertoldi et al., (2000); Frayer et al. 2000; Gear et al. 2000; Lutz et al., 2001; Dannerbauer et al., 2002) and molecular line spectroscopy (Frayer et al., 1998, 1999; Kneib et al., 2002) have been published, and many additional studies are under way. The time spent following up the SCUBA Lens Survey (Smail et al., 2002) exceeds by almost an order of magnitude the time required to make the submm discovery observations (Smail et al., 1997). The difficulty of the task is highlighted by the identification of plausible counterparts to these galaxies being only about 60% complete over 4 y later at the start of 2002. The follow-up results from a well-studied subset of galaxies in the SCUBA Lens Survey are shown in Figs. 14 - 20, in order of decreasing 850-µm flux density. These are chosen neither to be a representative sample of submm galaxies, nor to be a sufficiently large sample for statistical studies, but rather to present a flavor of the range of galaxies that can be detected in submm-wave surveys for which high-quality multi-waveband data is available. The galaxies that are presented are chosen to have good positional information, and redshifts where possible. Other detections for which excellent multi-wavelength follow-up data are available include the brightest source in the HDF-N SCUBA image (Hughes et al., 1998; Downes et al., 1999a), an ERO detected in the CUDSS survey by Eales et al. (1999) (Gear et al., 2000), a z = 2.8 QSO in a cluster field (Kraiberg-Knudsen et al., 2001), and the substantially overlapping catalogs of galaxies detected by Cowie et al. (2002) in deeper images of 3 of the 7 clusters in the Smail et al. lens survey.
The properties of submm galaxies in the optical waveband, corresponding to the rest-frame UV waveband, appear to be very diverse (Ivison et al., 2000a). This may be due in part to their expected broad redshift distribution. However, given that two submm galaxies at the reasonably high redshifts z = 2.5 and 2.8 are known to be readily detectable at B 23 (Ivison et al., 1998, 2000, 2001) - before correcting each for the magnitude (factor of about 2.5) of amplification due to the foreground cluster lens - while most others are very much fainter (Smail et al., 2002; Dannerbauer et al., 2002), it is likely that much of the spread in their observed properties is intrinsic. As most counterparts are extremely faint, confirmation of their nature requires a large, completely identified sample of submm galaxies with known redshifts, which is likely to be some time away. It is possible that optically faint submm galaxies have similar properties, an issue that can be addressed when deep near-IR observations are available.
Ivison et al. (2000a) and Smail et al. (2002) proposed a 3-tier classification system to stress the varied nature of submm galaxies (three being an eminently sensible number of classes for 15 galaxies!). Class-0 galaxies are extremely faint in both the observed optical and near-IR wavebands. Class-1 galaxies are EROs, very faint in the optical but detectable in the near-IR, while Class-2 galaxies are relatively bright in both bands. It is unclear how closely this scheme reflects the underlying astrophysics of the submm galaxies; however, the classification separates the optically bright galaxies (Class 2s), for which the acquisition of optical redshifts and confirming CO redshifts are likely to be practical, and the fainter galaxies, for which this will be a great challenge (Class 0s). A similar approach for MAMBO sources has been discussed by Dannerbauer et al. (2002). Note that submm galaxies could change classification by having different redshifts, despite identical intrinsic SEDs.
3.2.2. Ultradeep radio images
The surface density of the faintest radio sources that can be detected using the VLA (Richards, 2000) is significantly less than that of optical galaxies, and so an incorrect radio counterpart to a submm-selected galaxy is relatively unlikely to be assigned by chance. If 1.4-GHz VLA images are available at a flux limit approximately 1000 times deeper than 850-µm images, then the radio counterparts to non-AGN submm galaxies should be detectable to any redshift (see Fig. 7). Deep radio follow-up observations of submm-selected galaxies yielding cross identifications have been discussed by Smail et al. (2000), and further information about very faint radio sources in the field of the UK 8-mJy SCUBA survey (Fox et al., 2002; Scott et al., 2002) should soon be available in Ivison et al. (2002). Despite an extremely deep radio image (Richards, 2000), the brightest submm galaxy detected in HDF-N (Hughes et al., 1998) does not have a radio detection, probably indicating a very high redshift. The survey results reported by Eales et al. (2000) and Webb et al. (2002a) were discussed in the context of radio data covering the same fields; however, the radio survey is not deep enough to detect a significant fraction of the relatively faint submm sources. As can be seen for the specific submm galaxies shown in Figs. 14 - 20, when they are available, deep high-resolution radio images are very useful for determining accurate positions and even astrophysical properties of submm galaxies: see Ivison et al. (2001).
3.2.3. CO rotation line emission and continuum mm-wave interferometry
The detection of CO line emission from submm-selected galaxies is a crucial step in the confirmation of their identification. It has been demonstrated in only three cases so far (see Figs. 14, 18 and 19), using the OVRO Millimeter Array (MMA; Frayer et al., 1998, 1999), in one case in combination with the BIMA array (Ivison et al., 2001), and the IRAM PdBI (Kneib et al., 2002). These observations are very time-consuming, typically requiring tens of hours of observing time. The 1 GHz instantaneous bandwidth of existing line-detection systems also means that a redshift accurate to at least 0.5% must be known before attempting a CO detection. In other cases, continuum emission is detected using the interferometers, confirming the reality of the initial submm detection and providing a better position (Downes et al., 1999a; Bertoldi et al., 2000; Frayer et al., 2000; Gear et al., 2000; Lutz et al., 2001; Dannerbauer et al., 2002), but no absolute confirmation of a correct optical/near-IR identification or a crucial redshift.
The ALMA interferometer array will have the collecting area and bandwidth to make rapid searches for CO line emission in the direction of known submm continuum sources from about 2010. Specialized wide-band mm-wave spectrographs to search for multiple high-redshift CO lines separated by 115 GHz / (1 + z) that are currently under development (Glenn, 2001). Wide-band cm-wave receivers at the 100-m clear-aperture Green Bank Telescope (GBT) could detect highly redshifted 115-GHz CO (1 0) line emission.
3.2.4. X-ray observations
Based on synthesis models of the X-ray background radiation intensity (Fabian and Barcons, 1992, Hasinger et al., 1996), Almaini et al. (1999) and Gunn and Shanks (2002) suggested that 10-20% of the submm galaxy population could be associated with the hard X-ray sources that contribute this background. Observations of fields with common deep Chandra and SCUBA images were discussed in Section 2.8.
The small degree of observed overlap between the submm and X-ray sources, implies that if a significant fraction of submm galaxies are powered by accretion in AGN, then the accretion must occur behind an extremely thick absorbing column of gas, and less than 1% of the X-ray emission from the AGN can be scattered into the line of sight (Fabian et al., 2000; Barger et al., 2001; Almaini et al., 2002). In order to avoid detection using SCUBA, high-redshift hard X-ray Chandra sources must either contain a very small amount of gas and dust, and thus have only a small fraction of their energy reprocessed into the far-IR waveband, which seems unlikely given their hard spectra; or they must contain dust at temperatures much higher than appears to be typical for submm-selected galaxies. The detection of Chandra sources in Abell 2390 using ISO at 15 µm, but not using SCUBA at 850 µm, argues in favor of at least some Chandra sources having very hot dust temperatures (Wilman et al., 2000). Comparison of larger, deep ISO 15-µm images with Chandra images shows that most of the faint, red AGN detected by Chandra are detected in the mid-IR (Franceschini et al., 2002).
This should be readily confirmed using wide-field sensitive mid-IR observations of Chandra fields using SIRTF, images which will also yield well-determined SEDs for the dust emission from the detected galaxies.
3.2.5. Mid- and far-IR observations
Distant submm galaxies are too faint at far- and mid-IR wavelengths to have been detected in the all-sky IRAS survey. However, the first submm-selected galaxies were detected while the next-generation ISO space observatory was still operating, and there were both late-time ISO observations of submm fields, and some serendipitous overlap of fields. In general, the small aperture and small-format detector arrays of ISO still led to relatively little overlap between SCUBA and ISO galaxies, for example in the HDF-N (Hughes et al., 1998; Elbaz et al., 1999) and Abell 2390 (Fabian et al., 2000; Wilman et al., 2000). The brightest SCUBA galaxies in the cluster Abell 370 (Ivison et al., 1998a) have fluxes measured by ISO at 15 µm (Metcalfe, 2001), providing valuable, and otherwise difficult to obtain, constraints on their short-wavelength SEDs (Fig. 2), and thus their dust temperatures. The much larger detector arrays aboard SIRTF should allow many more constraints to be imposed on the mid-IR SEDs of submm galaxies. Existing deep submm fields are included for imaging within the SIRTF guaranteed time programs, and individual submm galaxies have been targeted for mid-IR spectroscopy. 13
13 Details of SIRTF observing programs can be found at sirtf.caltech.edu. Back.