7 Discussion

Our results show that combining CCD observations with conventional photographic astrometry, it is possible to obtain "absolute" (i.e. refered to a standard astrometric system) optical positions of very faint ($B\sim20-23$) CERS with a precision level as good as 60 mas.

Although the ultimate precision attainable with our method is hampered by the accuracy with which the positions of the reference stars can be determined by measuring the plates with a conventional manual measuring device, an important reduction of the total internal error results from the significant decrease of $\epsilon_{\rm r}$, which completely dominates the total internal error in a purely photographic approach (this is true even in the case of moderately bright ($B\sim19.5$) objects; see Costa & Loyola 1992).

It should be stressed furthermore that the method proposed is also capable of precise astrometry of objects that are not measurable -or simply not visible- on a wide-field telescope plate, and is very effective at exploring suspected empty fields in search of faint optical counterparts (for roughly 50% of the empty fields explored we are proposing a new optical identification).

In Table 3 we present a comparison in the ICRS with the VLBI radio positions given by Ma et al. (1998), in the sense radio minus this work (CL). To calculate the mean differences presented at the bottom of the table we considered only optical positions derived from CCD observations. The statistics are based on only 11 objects; this because source 1604-333 has a right ascension radio-CL residual three times the standard deviation of the mean of the complete sample. Keeping in mind that the errors in the measurement of the primary, secondary and tertiary reference stars used to calculate the total internal errors of our positions were average values based on previous experience (see Sect. 5) -which opens the possibility that for some objects these errors may have been underestimated- the residuals obtained are consistent with the estimated precision of our optical data.

Since our optical positions are displayed in essentially the same reference system as the radio positions, the radio-CL differences given in Table 3 provide an evaluation of these objects as possible radio/optical frame link sources (or their identification as astrophysically interesting astrometric outliers); in five cases (namely 0936-853, 1221-829, 1604-333, 1647-296 and 1706-174) the first one available. A few objects show large residuals which are not consistent with the estimated errors of their positions presented in Table 2. These cases are difficult to explain; they could be astrometric outliers, or, as explained in the above paragraph, their positional errors may have been underestimated. Also, we cannot rule out the existence of an undetected magnitude equation.

Considering that 12 of the CERS observed are either defining or candidate fiducial radio sources of the present realization of the ICRS, our optical data is potentially useful to help maintain, and possibly improve, the current link of the Hipparcos reference frame to the ICRS.

Acknowledgements

We are indebted to Cerro Tololo Interamerican Observatory for donating the plate material that made possible our survey. We are also indebted to Prof. C. Anguita for his interesting comments, and to M. Wishnjewski and L.E. González for measuring the plates. This work was financed by the Fondo Nacional de Investigación Científica y Tecnológica (proyecto N$^\circ$1970767 Fondecyt).