Recent VLA and WSRT deep surveys (Mitchell and Condon 1985, Oort 1987) have extended the 1.4-GHz source count to S 100 µJy directly and S 10µJy statistically (Figure 15.5). The weighted source count flattens below S 1 mJy, suggesting the emergence of a significant low-luminosity source population - L < 1024 W Hz-1) at z 1, L < 1022 W Hz-1 at z 0.1. This luminosity range brackets the transition between elliptical and spiral galaxies in the local radio luminosity function (Figure 15.3), so the faint (S < 1 mJy) sources are likely to be quite different from the strong ones.
Figure 15.15. Most faint (S < 1 Jy at = 1.4 GHz) radio sources fall in the shaded redshift band 0.3 < z < 3, so they are at nearly the same angular-size distance D if = 1. Abscissa: angular-size distance (Mpc). Ordinate: redshift.
Subrahmanya and Kapahi (1983) associated the faint sources with nonevolving spiral galaxies at low redshifts (z < 0.1). Their conclusion depends on a very steep local luminosity function (to yield a flat visibility function for L < 1022 W Hz-1 derived from Pfleiderer's (1977) 1.4-GHz survey of bright spiral and irregular galaxies. Such a luminosity function implies much higher space densities of radio sources with L < 1022 W Hz-1 than the luminosity function based on recent VLA observations of a similar optical galaxy sample (Figure 15.3), probably because confusion by background sources in the larger beam of the NRAO 91-m telescope produces some spurious detections. The 1.4-GHz source count from nonevolving spiral galaxies described by the newer local luminosity function is an order of magnitude lower than the observed count, as indicated by the dotted line in Figure 15.17. With either luminosity function, the median redshift of nonevolving spiral galaxies in the flux-density range 0.1 to 1 mJy is only <z>0.1. Such a low median redshift does not appear to be consistent with the magnitude distribution of galaxies identified with faint sources selected at 1.4 GHz (Windhorst 1986).
If radio sources in spiral galaxies do evolve at about the same rate as those in elliptical galaxies (Condon 1984a, b), they can account for the flattening of the source counts below S 1 mJy (Figure 15.5) as well as the higher redshifts (Figure 15.10) indicated by the optical-identification magnitude distribution. It should be emphasized that in this scenario, most of the faint radio sources are not the "normal" spiral galaxies found in optically selected catalogues. Typical radio-selected spiral galaxies are those with L 1022 W Hz-1 found at the peak of the visibility function, most of which are interacting "starburst" galaxies (e.g., Condon et al. 1982a), Markarian galaxies, and Seyfert galaxies (e.g., Meurs and Wilson 1984). Since there is a close correlation between the far-infrared ( = 60 µm) and radio continuum flux densities of spiral galaxies, radio-selected spiral galaxies are more akin to galaxies found in the IRAS survey than to normal spiral galaxies.
A third explanation for the source-count flattening is that there is a "new population" of radio galaxies with L 1023 W Hz-1 at = 1.4 GHz (Windhorst 1984) that was somehow missed by the radio luminosity functions derived from optically selected samples of spiral and elliptical galaxies (e.g., Figure 15.3) or Markarian and Seyfert galaxies (Meurs and Wilson 1984). Using spectroscopy and four-band photographic photometry of galaxies identified with radio sources stronger than S = 0.6 mJy (Kron et al. 1985), Windhorst (1984) estimated that the local space density of radio sources associated with his "blue galaxy" identifications exceeds that of radio sources in known spiral and elliptical galaxies by an order of magnitude in the luminosity range L 1022 to 1023 W Hz-1. Only a moderate amount of evolution is needed for such a population to account for the source counts below S 5 mJy (Windhorst et al. 1985).
The morphological composition of the blue radio-galaxy population is still unclear, and it may vary with both flux density and apparent magnitude. Those galaxies brighter than F 16 mag are spiral galaxies; the fainter identifications "are often of peculiar compact morphology, sometimes interacting or merging" and have optical luminosities about equal to those of bright spiral galaxies (Windhorst 1984, Kron et al. 1985). Some of the faintest may be blue broad-line radio galaxies similar to those found in strong-source samples (Wall et al. 1986), photometrically misclassified elliptical galaxies (Kron et al. 1985), or misidentifications (Wall et al. 1986, Windhorst 1986, Weistrop et al. 1987). While the blue galaxies gradually displace red elliptical galaxies as the dominant population in faint radio-selected samples, it is not clear that they are a new population in the sense of surpassing the known local radio luminosity function near L 1023 W Hz-1. If a weakly evolving population is significant in radio flux-limited samples at cosmological distances, it probably should be significant even in the small samples of optically bright galaxies used to construct the local luminosity function. Also, the radio identifications of the large (N 104) UGC galaxy sample (Nilson 1973) south of = + 82° with sources stronger than S = 150 mJy at = 1.4 GHz (Broderick and Condon, manuscript in preparation) do not appear to be consistent with a factor-of-ten increase in the local luminosity function near L = 1023 W Hz-1. The UGC galaxies actually detected in this luminosity range are classified by Nilson as a mixture of active spirals, elliptical, S0, and "compact" galaxies. If the radio sources in this mixture evolve, they may account for most of the blue radio galaxies seen at cosmological distances in deep surveys. Because there is such a sharp decline in the luminosity function of spiral galaxies above L 1023 W Hz-1 the blue galaxy population identified with sources in the 1 to 10 mJy range by Kron et al. (1985) probably contains a smaller fraction of spiral galaxies than sub-mJy samples do.
A sharp distinction between radio-emitting spiral and elliptical galaxies can be made by the relative strengths of their far-infrared emission - nearly all infrared-selected galaxies with detectable radio sources are spirals, not ellipticals. Furthermore, there is a remarkably tight correlation between the far-infrared and radio continuum flux densities of spiral galaxies. For infrared-selected spirals, the quantity u log(S60 µm / S1.4GHz) has a median value <u> = + 2.15 ± 0.15 and a scatter u < 0.3 (Condon and Broderick 1986). Biermann et al. (1985) used the infrared-radio correlation and a static Euclidean extrapolation of the = 60 µm source count to predict that spiral galaxies will contribute significantly to the 5-GHz source count below S 100 µJy. Now that the = 60 µn source count has been extended to S = 50 mJy (Hacking et al. 1987), the contribution of infrared-selected spiral galaxies to the 1.4-GHz count of sources as faint as S = 50 mJy / dex(u) = 0.35 mJy can be estimated directly from the infrared data - the weighted count scales as u-3/2, so S5/2 n(S | = 1.4 GHz) 1. At least 20% of the radio sources with S 0.35 mJy at = 1.4 GHz can be identified with spiral galaxies in this way; if u > 0, the percentage rises. Finally, the = 60 µn source count is consistent with strong evolution of infrared-selected spiral galaxies (Hacking et al. 1987).