3. DEEP RADIO IMAGING OF THE HDF

In order to detect even a handful of radio sources in the HDF-N, noise levels of a few microJy must be achieved. These in turn require integration times ranging from a few days - in the case of the WSRT and VLA, to many days in the case of MERLIN.

Figure 3. The WSRT 1.4 GHz contour map of the HDF (see incomplete rotated square) and part of the HFF superimposed upon a deep CFHT optical image of the HDF region made by the Canada-France-Hawaii telescope (courtesy Amy Barger). Crosses indicate previously known VLA detections, boxes indicate new WSRT detections. The detection of a nearby, extended star forming galaxy is highlighted (upper left).

3.1. VLA 8.3 GHz, VLA-MERLIN & WSRT 1.4 GHz observations

Deep VLA observations of the HDF (including the HFF) have been conducted at both 8.3 and 1.4 GHz (Fomalont et al. 1997, Richards et al. 1998, Richards 2000). The 8.3 GHz observations reach noise levels of a few microJy per beam (several times better than the 1.4 GHz observations) but more sources are actually detected at 1.4 GHz where the source counts are steeper and the VLA field of view wider. Perhaps the most "complete" radio view of the HDF (see Fig. 3) is provided by the WSRT 1.4 GHz observations (Garrett et al. 2000a). These are sensitive to very extended radio structures, although for the WSRT (and indeed the VLA), the vast majority of the microJy radio source population remain barely or completely unresolved at arcsecond resolution. Combined VLA-MERLIN 1.4 GHz observations with a resolution of 0.2" (Muxlow et al. 1999) begin to resolve most of these sources but the detailed morphology of the microJy radio source population still remains unknown. The main results of the VLA, MERLIN and WSRT data can be summarised as follows:

• Most of the radio sources are identified with relatively bright (R < 25^m), moderate redshift, optical counterparts - often identified as interacting, irregular or peculiar morphological type (Richards et al. 1998)
• The 1.4 GHz VLA source sample is steep spectrum in nature (typically ~ -0.85 - Richards 2000). Sources selected by the VLA at 8.3 GHz are significantly flatter specutrum ( ~ - 0.35)
• There is a strong correspondence between the Mid-IR ISO detections and the radio detections (see Fig. 4 right). Indeed the majority of radio sources (after applying appropriate k-correction factors) appear to closely follow the FIR-radio correlation (Garrett 2002).

  Figure 4. Left: MERLIN-VLA contour map (super-imposed on the HST gray-scale image) of J123651+621221 - an optically faint dust-obscured starburst system in the HDF with a total flux density of 49 µJy (Muxlow et al. 1999). (b) Right: The strong correlation between FIR and radio luminosity for high-z galaxies detected by both ISO and the WSRT in the region of the HDF-N (Garrett 2002).

• Of the 91 sources detected with the combined MERLIN-VLA 1.4 GHz array (see Fig. 4 left), the majority show radio structure on sub-galactic scales (Muxlow et al. 1999). About 50% of the MERLIN-VLA 1.4 GHz detections show extended structure aligned with the optical isophotes of the galaxy.
• The WSRT detects a small but significant population of both star forming galaxies and AGN, some of which are resolved by the higher resolution VLA and MERLIN observations (Garrett et al. 2000a; Muxlow priv. comm).
• A comparison between the VLA and WSRT 1.4 GHz images (separated by several years) shows evidence for significant variability (factors of 2 or more) for a few percent of the sub-mJy radio source population (presumably low-luminosity AGN).
• Around 10-20% of the microJy source population (see Fig. 4 -left and Richards et al. 1999) are optically faint or completely unidentified (R > 25^m).