3. The galaxy identifications
3.1. Optical identification procedure
Generally the optical identification procedure is quite straightforward: an identification is considered to be positive if the radio and optical positions are within some adopted maximum distance, and for the whole identification sample the completeness and reliability can easily be established (see e.g. De Ruiter et al. 1977).
In the case of bright galaxies the usual criteria have to be applied with somewhat more caution, because many of the galaxies are extended, and in a few cases very extended (cf. NGC 6503 in Table 1). Therefore there is the risk of losing very bright galaxies if we only use a pre-established limit in the radio-optical separation (the radio source does not necessarily fall exactly on the optical center, in particular in the case of spiral or starburst galaxies). For this reason we started with a large search area with radius 30 arcsec around the radio position, and then decided on the basis of the extension of the galaxy if it could be accepted as the optical counterpart of the radio source. Basically this was done for galaxies with magnitude brighter than 14, while fainter galaxies were treated in the usual way.
Considering the large quantity of radio sources (the more so in the future when we will use the entire WENSS), the identification procedure has to be automatized as much as possible. We proceeded in several steps: first, we cross-correlated the minisurvey sources with the Automated Plate Machine (APM) catalog of POSS I objects (McMahon & Irwin 1992), using a search radius of 30 arcsec. Of this list of optical objects thus produced, we selected a subset of those that were classified by the APM as non-stellar in at least one color. At this point in the analysis we therefore eliminated objects of stellar appearance (quasars for example), and are left with galaxies. Their nominal APM magnitudes were required to be brighter than 18.5 in the blue or brighter than 17 in the red. This resulted in a first, tentative, list of about 800 possible galaxy identifications. We then extracted small optical images for this subset, using the "jukebox" and the Digital Sky Survey (based on red POSS prints). We are developing software that can decide whether an extended object is present near the radio source position, determine its magnitude, and produce an automatic listing of proposed galaxy identifications. Thus the whole identification procedure would then be reduced to preparing a list of source positions, extracting corresponding DSS images and determine the presence of optical counterparts. However, for the moment we have used the subset of proposed galaxy identifications based on the APM, and inspected each individual field visually. Thus we checked the nature of the object (galaxy or star-like), and its distance from the radio position. Since the errors in the radio position are typically of the order of a 2-3 arcsec (see Rengelink et al. 1997), the initial search radius of 30 arcsec is usually far too big. Only very bright and therefore very extended galaxies may be associated with radio sources that are relatively far away ( arcsec) from the galaxy center. Such bright galaxies are often spirals, in which we may detect giant HII regions in the outer parts of the galaxy. Considering this we tightened our criteria at this point, requiring that the radio/optical position difference should not be more than 5 arcsec, except for the brighter galaxies, where we decided on an individual basis if the radio source was likely to be associated with the galaxy.
The magnitudes were calibrated using i) stars of selected area 57 (Grueff, private communication) and ii) the galaxies observed photometrically by Sandage (1972), Sandage (1973). From i) we could calibrate the point source magnitude and from ii) the galaxy diameter, which together then give the galaxy magnitude. It is therefore quite feasible to obtain in an automated fashion galaxy magnitudes, by determining the central photographic density, a mean source diameter (e.g. through a gaussian fit) and the background density. The rms uncertainty thus obtained for the Sandage galaxies is mag.
Our final list of identifications contains 402 galaxies. This list is by no means complete, since the selection was done using APM magnitudes. We believe optical completeness is achieved at about . This point is further discussed in Sect. 3.2.
In Fig. 1 we show the difference between radio and optical positions (in and separately). Relatively large positional differences ( arcsec) occur only for the brighter galaxies (see the discussion in the previous paragraph).
A simple estimate of the reliability of our identifications is straightforward: assuming that there are roughly 30 galaxies per square degree down to a red magnitude of 16, and assuming an average search radius of 10 arcsec, we expect of the order of 7 spurious identifications, among the minisurvey sources. The actual number of identifications with in Table 1 is 185, and therefore the contamination rate should be around 4 %. For fainter objects the search radius was actually smaller (close to 5 arcsec, see above), so that the higher density of background galaxies, down to , is almost entirely compensated by the tighter search criterion. The overall contamination of the sample given in Table 1 is therefore %, which is quite acceptable.
The completeness is much harder to estimate, since we used a first selection criterion based on APM magnitudes required to be brighter than 17 in the red. These magnitudes are accurate for stellar objects, but may be wrong by as much as magnitude for galaxies around , and more than that at the bright end. Fortunately, compared to the magnitudes as calibrated from the Digital Sky Survey, the APM tends to overestimate the brightness of objects. If we impose a limit of we should therefore be statistically complete, although individual objects may erroneously have crept in or dropped out of the sample due to the photometric uncertainty of our magnitudes (approximately 0.4 mag.). We conclude that very few objects will be missing.
3.2. The galaxy sample
The sample of galaxy identifications is given in Table 1. The lay-out of this table is: a running number based on the minisurvey radio catalogue, ordered in right ascension, in column 1, the WENSS name in column 2, peak and integrated flux densities at 325 MHz ( peak and total respectively) in columns 3 and 4, galaxy name (if any) and type (elliptical or spiral) in columns 5 and 6, the red magnitude (as discussed above) and (galactocentric) redshift, if known, in columns 7 and 8. Redshifts with an asterisk are based on the new observations described in the next section. The differences (in arcsec) of the radio and optical positions (in the sense radio minus optical) in right ascension and declination are given in columns 9 and 10, and finally the spectral index between 325 MHz (from WENSS) and 1400 MHz (from NVSS, see below) in column 11: only spectral indices of sources with a 325 MHz peak flux density above 100 mJy are shown, because total fluxes of fainter sources may be systematically understimated (see Sect. 5).
Table 1 contains all 402 galaxies that were found using the above selection criteria and it does not represent in any sense a complete sample. However it is straightforward to construct a radio and optically complete sample, once we take into account the following points:
i) The WENSS survey lists radio sources down to a flux density of 15 mJy, but is essentially complete only above mJy. There are 274 galaxy identifications in Table 1 with radio flux density above this limit.
ii) Galaxies were first selected on the basis of their blue and red APM magnitudes, which means that some galaxies have (determined afterwards) in the range 17-18 mag. Obviously we cannot claim to be complete at that magnitude level. A safe limit appears to be ; there are 119 galaxies brighter than 16 and with radio flux density above 30 mJy. This should be considered a radio and optically complete sample. We may have missed some very low surface brightness galaxies, which should have been included on the basis of their integrated magnitude: especially the spiral galaxies given in Table 1 have a wide range of mean surface brightness, and one or two are barely detected although they are so extended as to have an integrated magnitude brighter than the limit. We assume that such objects are rare and therefore only marginally affect the completeness of the sample.
In Fig. 2 we give the distribution of radio flux densities at 325 MHz of the galaxy identifications given in Table 1. The contribution of spiral galaxies in the histogram is represented by the shaded area. As is well known, spiral galaxies are mainly found below mJy, reflecting the fact that they are on average much weaker radio emitters than elliptical galaxies.
© European Southern Observatory (ESO) 1998
Online publication: September 30, 1998