3. Catalogue and identification of counterparts
To derive accurate positions of the ISO sources is limited by (a) the large pixel size () of the initial image, (b) the size of the diffraction disk (2 to ), and (c) errors in the data reduction procedure (including those due to distortion corrections). The astrometric accuracy of the CFRS galaxies in 1415+52 field is 0:0015 relative to radio positions (Hammer et al. 1995). The overall centering of the ISOCAM field was based on the positions of the 5 brightest stars and the bright CFRS14.1157 source. The resulting differences between the optical and ISO positions for all sources with counterparts (see next section) are shown in Fig. 4. All but 9 of the sources are within a radius equivalent to one pixel () in the original ISOCAM data. From comparison of the positions of optical and ISOCAM LW2 sources, the median difference is 4:002.
3.2. Optical counterparts
We have first compared the 6.75µm frame with µJy radio sources (Fomalont et al, 1991), and have calculated the probability of a pure coincidence, assuming Poisson statistics:
where d is the angular distance between the ISOCAM LW3 source and the radio source in degrees and n is the integrated density of radiosources at flux (n) = 83520 ). Six ISOCAM LW2 sources are thus identified (see Table 1) with their optical counterparts derived from Hammer et al (1995).
For all the other sources without radio counterpart we have used the band counts. All optical sources within of an ISOCAM LW2 source have been considered. Assuming Poisson statistics, the probability density of a pure coincidence between an ISOCAM and optical source is:
where d is the angular distance between the ISO object and the optical source in degrees and n() is the integrated density of galaxies at magnitude from the analysis by Lilly et al. (1995c). Fig. 5 shows the relationship between the S/N of the ISOCAM observation and the probability (P) that the optical source is projected by pure coincidence.
We are aware that possible non gaussian positional uncertainties related to extended emission, source confusion and residuals can affect our probability calculations for low S/N sources. As shown in Fig. 6, some sources present asymmetric shapes, and for few of them, only a part of the ISOCAM source overlaps the optical counterpart. Indeed the final noise structure is far from being gaussian, due to possible glitch residuals, and the ISOCAM position has not been corrected for possible image distortions. To calibrate our probabilities in an empirical way, we have applied a random match control test to the ISO field by rotating it by 90, 180 and 270 degrees relative to the optical image. Only 21 of the 54 ISOCAM sources are found randomly associated to an optical counterparts (22.5 and P0.02), which should be compared to the 22 P0.02 counterparts displayed in Table 1, this strengthens our probability calculations. Non gaussian cannot affect our probability calculations by more than a factor 2.
3.3. Source catalogs
As illustrated in Fig. 5, six catalogues were constructed which contain identifications of varying degrees of confidence (from 1 - highest, to 6 - no identification) according to their probability (P) and S/N ratio. Table 1 lists the identifications of ISO sources in four of the catalogues (in decreasing confidence level in each successive catalogue). The ISOCAM LW2 sources and their associated optical sources are listed in columns 1 & 2. Column (3) gives the redshift if available. In the redshift column , "star" indicates that the object has been identified spectroscopically as a star, " - " indicates that no redshift is available. Columns (4) (5) and (6) give the , and magnitudes (a " - " indicates that no photometry is available); column (7) lists the angular distance (in arcsec) between the ISOCAM source and optical counterparts; column (8) gives the associated probability that the identification is spurious; and columns (9) and (10) give the flux at 5-8.5 µm and the error (in µJy).
The 21 sources listed in catalogues 1 and 2 are considered to have secure identifications since they are relatively strong sources (S/N 4) and have a relatively low probability of chance coincidence. All but two of the counterparts have , the limit of the CFRS spectroscopic survey, and spectra are available for 13 of these counterparts.
The 24 sources in the supplementary catalogues 4 and 5 are identifications with lower S/N sources (4S/N 3). In five cases in the "least secure" catalogue 5, more than one optical counterpart is listed as possibly being associated with an ISOCAM source. In most cases, the more probable counterpart is optically brighter than, or as bright as, the alternate, providing additional evidence that it is the likely counterpart. In the following analysis we use only the "best" optical identifications (i.e., those with the lowest probability of chance coincidence), but all counterparts with a probability within a factor two of the smallest probability are retained in Table 1 for reference. Statistically, 1.5 out of the 18 sources in catalogue 5 are expected to be purely chance coincidences. As well as having fainter mid-IR fluxes, the optical counterparts of these 24 sources are also fainter optically on average (8 are fainter than ).
Fig. 6 shows the superposition of ISOCAM LW2 source contours and the optical counterparts of 14 extragalactic sources (from catalogues 1, 2, 4 & 5) with known redshifts. The optical image was extracted from a composite image centered at the ISO position.
Among the 54 ISOCAM LW2 sources with S/N 3, 9 were not identified to the limit of our combined image. Unfortunately, we have K band images for only a small portion of the field, so it is not possible to determine whether all sources are visible in the near-IR and what their colors are. The positions and fluxes of the 9 "non-identified" ISO sources are listed in Table 2.
In summary, 45 of the 54 ISOCAM LW2 sources (83%) were identified with galaxies and stars on composite images () of the CFRS1415+52 field. 21 of these have secure identifications with optical counterparts (including 7 stars). As shown below, some of the visible-light properties of the 24 less secure counterparts add considerable support to their identifications. In addition to the 7 spectroscopically confirmed stars, three other sources have profiles of compact sources (stars, QSO's, etc) on our I band image, so that 10 of the 45 may be stars (although the three could also be QSOs, from B,V and I band images and colors).
© European Southern Observatory (ESO) 1999
Online publication: March 1, 1999