8 Discussion

The systematic optical identification of radio-selected samples with flux densities in the range 1-30 mJy began nearly twenty years ago using the first deep radio images obtained by Westerbork and the VLA. In Table 5, we summarize the published work on this problem. The area surveyed covers a little more than 50 deg$^{2}$ and includes 1448 radio sources between the limits of the all-sky, single dish surveys ($\sim$30 mJy) and the FIRST survey threshold (1.0 mJy). A total of 455 sources have suggested identifications, albeit to different levels of optical sensitivity, completeness, and reliability. Only 270 counterpart candidates have been published for radio-selected objects at thresholds roughly equivalent to the POSS-I plates. The total number from this flux density range in the work reported here represents a sample of $\sim65,400$ confidently identified radio sources to this limit. Our project, then, represents an increase of more than two orders of magnitude in the number of optically identified faint radio sources. What source classes are represented in this sample, and how do the results compare with our expectations from earlier work?

The raw fraction of detections at the POSS-I limit, $\sim 18.3\%$, is similar to the mean value derived from previous work (Table 5); the different optical bands used and the fuzzy lower bound to the POSS-I magnitudes easily explains the small discrepancy introduced when the corrections for multiple-component sources are applied (§ 5.4). The distribution amongst stellar and non-stellar counterparts is also roughly similar; detailed comparisons are not warranted given the inevitable difficulty of classifying objects near the plate limits where most of the candidates lie. Further progress in quantifying the source populations represented, and in establishing their luminosity functions, requires CCD imaging and spectroscopy, work we have begun to pursue for a variety of subsamples derived from this FIRST/APM database. We discuss these briefly here.

The comprehensive catalog of stellar radio observations by Wendker (1995) contains 18 stars in the FIRST survey area detected at any radio frequency between 100 MHz and 30 GHz; of these, only a dozen have ever been seen at a 20 cm flux density in excess of 1 mJy. Our comprehensive study of various stellar proper motion catalogs has produced an additional fourteen stellar radio source identifications (Helfand et al. 1999). Proper motions remain the largest obstacle to our identification of stellar counterparts fainter than Tycho catalog limit ($\sim11^{th}$ magnitude), and the advent of the GSC-II should significantly increase the number of stellar identifications to $\sim85$. Nonetheless, stars make a trivial contribution to the total radio source population above our 20 cm flux density threshold of 1 mJy. Other Galactic counterparts are similarly rare: we have detected only 3 (of the 8) known planetary nebulae and 7 (of the 18) known radio pulsars. In total, we expect that less than 0.03% of the FIRST catalog entries represent Galactic objects.

There are, of course, a much larger number of optical counterparts with stellar images on the POSS plates. These represent a mixture of quasars, BL Lac objects, AGN with sufficiently bright nuclei that they are classified as stellar, H II galaxies with similarly bright nuclei, and radio galaxies with sufficiently small diameters that they are unresolved. The FIRST Bright Quasar Survey (White et al. 2000) has established the distribution amongst these categories for objects brighter than E=17.8: $\sim55\%$ quasars, $\sim8\%$ BL Lacs, $\sim9\%$ AGN, $\sim 4\%$ radio galaxies, and $\sim17\%$ H II galaxies, with the remaining objects being a collection of stars and chance coincidences. We have begun a smaller survey to establish this distribution at fainter magnitudes (Becker et al. 2001).

The majority of the FIRST counterparts brighter than the POSS-I limit are galaxies. Two principal populations are represented: active galactic nuclei including both relatively nearby Seyferts and luminous elliptical hosts of more powerful radio galaxies up to $z\sim0.5$, and a local population of star-forming galaxies (Condon 1992). Our work to date has consisted primarily of following up the counterparts to the several hundred bent, double-lobed radio galaxies which can act as tracers for galaxy clusters (Blanton et al. 2000, 2001). Studies of selected, magnitude-limited samples of the galaxy population are in progress.

One important use of FIRST radio sources is as tracers of the general matter density of the Universe; their broad redshift distribution and high mean $z$ samples a much larger volume than most surveys in other wavelength regimes. In applications such as determining the two-point (Cress et al. 1996), and higher order (Magliochetti et al. 1998) angular correlation functions, searching for a weak lensing signal on large angular scales (Refregier et al. 1999) and other such statistical studies, it may be useful to eliminate the low-redshift, star-forming galaxy population. The catalog presented here can be applied for such purposes.