2.9. Alternative strategy for deep submm surveys
An alternative strategy for finding submm galaxies has also been tried: targeted submm observations of faint radio-selected galaxies with accurate positions and radio spectral index information detected at a 1.4-GHz flux density level brighter than 40-µJy using the VLA (for example Richards, 2000). The radio source population at these faint flux density levels is expected to include mostly high-redshift star-forming galaxies, and only a minority of sources with the more powerful synchrotron-emitting jets and lobes associated with particles accelerated by AGN. By selecting those faint radio sources with faint K-band counterparts (Barger et al., 2000; Chapman et al., 2001b), it should be possible to sift out low-redshift galaxies from the target sample, and so generate a concentrated sample of luminous high-redshift galaxies to study in the submm. From Fig. 4, it is clear that a high-redshift faint 20-100-µJy non-AGN 1.4-GHz radio source is likely to be a very luminous galaxy. Chapman et al. (2001b) claim that high-redshift submm galaxies can be detected using SCUBA at a rate of about one every hour using this method, which is significantly more rapid than the rate of about one every 10 h achieved in blank-field surveys. The efficiency of submm detection of optically-faint 1.4-GHz 20-µJy radio galaxies at an 850-µm flux density greater than about 5 mJy is of order 30-40%. Barger et al. (2000) detect 5/15 with I > 24 at 6 mJy, and Chapman et al. (2001b) detect 20/47 with I > 25 at 4.5 mJy.
The difficulty comes in the interpretation of the results. The additional selection conditions of requiring first a radio detection, and then a faint optical counterpart will inevitably lead to the omission of a certain fraction of the submm galaxy population. Galaxies missing would include both the order of 15% of distant submm-selected galaxies that have relatively bright optical counterparts - the `Class-2' SCUBA galaxies (Ivison et al., 2000a; Smail et al., 2002) - and any very high-redshift submm galaxies that cannot be detected at the VLA, despite lying on the far-IR-radio correlation (Condon, 1992; see Fig. 8). There is also likely to be a bias (at all redshifts) towards both detecting AGNs, in which radio flux densities are boosted above the level expected from the standard radio-far-IR correlation, and the exclusion of a small population of low-redshift submm galaxies, which would be too bright in the radio and K band to be included in the survey. The rare, and perhaps especially interesting, submm galaxies at the lowest and highest redshifts are thus likely to be missing from radio pre-selected surveys, as are the distant submm galaxies with brighter optical magnitudes that are likely to be easiest to follow-up and investigate.
The size of these selection effects is difficult to quantify at present. However, their existence can be inferred from the diverse multiwaveband properties displayed by submm-selected galaxies from imaging surveys, a significant fraction of which are known not to be detected in deep radio observations (Downes et al., 1999; Smail et al., 2000). In some cases, the depth of the radio images used to make the comparison could be improved by a factor of several; it is possible, but we suspect unlikely, that all submm sources have radio counterparts lurking just below existing detection thresholds.
It is instructive to compare the results of the blank-field and faint radio pre-selected submm surveys. In a true 260-arcmin2 blank-field survey, Scott et al. (2002) found surface densities of 550+100-170 and 180 ± 60 deg-2 submm galaxies brighter than 6 and 10 mJy respectively, while Borys et al. (2002) estimate 164 ± 28 deg-2 brighter than 12 mJy in a 125-arcmin2 survey, with a rather conservative error estimate from 12 detections. The surface densities resulting from the faint radio-selected investigations of Barger et al. (2000) and Chapman et al. (2001b) are 430 deg-2 brighter than 4 mJy and 135 deg-2 brighter than 10 mJy. 11 This indicates incompleteness in the faint radio-selected counts by at least about 25%. This is perfectly acceptable for gathering a list of high-redshift galaxies for further study; however, for statistical analysis of a large sample of many tens of submm galaxies, the uncertainty introduced in the derived properties of the population due to incompleteness is expected to dominate the Poisson uncertainty, and so limit the accuracy of the inferences that can be made. Once the bright counts of distant 850-µm galaxies are known from direct, hopefully unbiased, large-area SCUBA imaging surveys (Fox et al., 2002; Scott et al., 2002) and matched with very deep radio images (Ivison et al., 2002), then the value of this shortcut for compiling a large sample of submm galaxies can be assessed. The same strategy would be very valuable for wide-field MAMBO 1.2-mm surveys (Dannerbauer et al., 2002). At present, some care needs to be taken in the interpretation of faint radio pre-selected surveys.
11 Barger et al., (2000) also impose a limit of less than 840 deg-2 brighter than 6 mJy at 850-µm, based on serendipitous detections of sources in their 2.5-arcmin-wide SCUBA images around the AGN radio galaxies. Back.