The EVN observations of the HDF suggest that deep, wide-field VLBI studies are not only possible, but in principle they can deliver important astronomical results. So far we have only scratched the surface. In a sense, we are just beginning to appreciate the fact, that VLBI has reached a sensitivity level where we can expect to detect many discrete radio sources in a single field of view (irrespective of where you point the telescopes!). This is quite a departure from the traditional role of isolated VLBI observations of very compact, and often very bright radio sources.
In the short term the use of in-beam phase-calibration techniques (Fomalont et al. 1999) should permit us to reach the expected thermal noise level of only a few microJy/beam (assuming a global VLBI array operating at 256 Mbits/sec and an on-source integration time of 24 hours). The real advance, however, will be in making full use of the raw data i.e to map out the primary beam response of individual VLBI elements in their entirety. Simultaneous multiple-field correlation, coupled with incredibly fast data output rates, is now being pursued at the EVN MkIV Data Processor at JIVE. When complete, this development will provide astronomers with the ability to image dozens of faint sub-mJy radio sources - all observed simultaneously with microJy sensitivity, full uv-coverage and milliarcsecond resolution (Garrett 2000b). Fig. 6 summarises the concept of deep, in-beam, wide-field VLBI but with current sensitivity limits employed. Large areas of the sky (such as that shown in Fig. 6) are now being routinely surveyed in great detail by optical and near-IR instruments. These deep surveys (e.g. the NOAO Deep Wide-Field Survey, Januzi & Dey 1999) have the great advantage that they cover enormous areas of sky (many square degrees) and thus there is always some region of the survey area that will include an appropriate "in-beam" VLBI calibrator.
Figure 6. A small portion of the NOAO-DWFS with a shallow WSRT 1.4 GHz contour map superimposed (image courtesy of Raffaella Morganti, de Vries et al. 2001, first radio contour at 30 µJy/beam). The dashed circle shows the FWHM of a 70-m telescope primary beam. Assuming modest r.m.s. noise levels of ~ 10 µJy/beam, 5 (6) target fields (of extent 2' × 2') can be identified. These fields are all located within the primary beam, and can be correlated simultaneously, and mapped out in their entirety. A (faint) VLBI calibrator to the North of the field permits accurate and continuous phase calibration to be applied to the target fields.
What fraction of these target sources can be detected ? The HDF-N results suggest that about 1/3 of the targets in a randomly selected field will be AGN, and that most of these will have compact structure. The remaining distant starburst systems will most likely be resolved by VLBI (even with microJy sensitivity) - detecting and imaging these systems with VLBI scale resolution, must await the construction of a much more sensitive, next generation radio telescope, such as the SKA (see Lecture by A. Kus, this volume). Nevertheless, I suspect that by the time we return to Castel St. Pietro Therme for the next NATO-ASI school, the use of VLBI as a deep, wide area survey instrument will already be well established.