Essentially, the LW2 survey yields only two sources above the 4 level of confidence. They both have an LW3 couterpart. We conclude that at the 60 level, there are 2 (say at most a few) sources in the 11 arcmin2 main area of the HDF, a result at odds with the numerous sources found by Goldschmidt et al. (1997). The discrepancy has been addressed by Aussel et al. (1997). We note that almost all LW2 sources have an LW3 counterpart. This small number of sources is not in disagreement with the 15 sources found in a different region but with a similar area by Taniguchi et al. (1997) at a sensitivity level twice to 3 times better than here in the same LW2 filter. The sensitivity level is here achieved with 1.7 hours of integration per sky pixel of 3 arcseconds. In the following, we concentrate on the LW3 results which, because of the larger number of sources, gives a better statistics on number counts. In the final Table (6), 34 objects are above the 4 limit of typically 100 to 150 in an approximate area of 25 arcmin2. The best sensitivity limit quoted is obtained with an effective on-source integration time of 2.7 hours. Number counts are discussed by Aussel et al. (1998).
The overall reliability of the method can now be assessed both internally and externally:
The complete identification, as well as the detailed comparison with the source list given by Goldschmidt et al. (1997), Rowan-Robinson et al. (1997) and Aussel et al. (1998) and by radio surveys, as well as the analysis of supplementary data taken as a repeat of LW2 observations will be dealt with in a forthcoming paper. A third external level of consistency can already be made: more than a third of the HDF LW3 sources are found in the VLA radio sample of Richards et al (1998). The analysis of Fig. 9 and Fig. 10 already reveals that, in the central HDF area, one always finds one or several bright counterparts in the optical (B magnitude less than 22). An even more clear cut case is that the sources can also be always associated with a K counterpart of relatively bright 17 to 18 magnitude, except for one source HDF_ALL_LW3_20 at the lower west border with RA=12 36 46.2 and Dec= 62 11 33.5 (confirmed in LW3 observations by the 2 rasters HDF-3_LW3 and HDF-4_LW3). This could be a good example of a galaxy very dim in the optical (29th magnitude) and near-infrared domain relative to its mid-infrared luminosity, for which no redshift can be measured except in the mid, far infrared or radio domains. Another example is HDF_ALL_LW3_24 which is associated with a radio source identified with a elliptical galaxy by Richards et al. (1998). Table 6 gives the redshift (when available in the internet lists, Cowie et al. 1998) of the galaxy which is the brightest and nearest source (within 6 arcseconds) to a given ISOCAM source (in a simple eye-ball sense). Almost all redshifts are within the 0.5 to 1 range. We are probably seeing the PAH spectral features around 7.7 redshifted in the LW3 band. The fact that most of these sources are not detected in LW2 (when observed) but detected in K means that LW2 witnesses the break between stellar emission and interstellar dust emission, for the ISOCAM HDF source redshifts. Two double sources can be spotted in this table. These are HDF_ALL_LW3_13 with 15, and 27 with 29. Only one object is a star (at the border of the map) and not a galaxy: HDF_ALL_LW3_41 (see Aussel et al. 1998).
It is clear that the redundancy of the rasters plays a crucial role in assessing the reliability of sources. The reliability of our method is clearly demonstrated here on the HDF ISOCAM data, because of their large (up to 50-100) redundancy. It will be used in various other ISOCAM deep surveys, where such a redundancy cannot be afforded. It is shown here that the temporal triple beam-switch method plus a classical spatial detection on an optimally coadded map of individual measurements allow the source noise to be less than 2 parts per ten thousand of the zodiacal background (as numerically found with the present data). The method is linear and does not deal in different ways with high and low flux sources (as long as they are small compared with the zodiacal background). This is achieved with a modest-size telescope and modest-size detector array with a strong reaction to cosmic rays and with some non-linear behaviours. Finally, we note that the LW3 map is above the camera confusion limit by a factor two.
© European Southern Observatory (ESO) 1999
Online publication: February 22, 1999