Radio Source Catalogue

The Cluster sample

The sample to be studied was chosen as a complete, representative subset of the overall survey. This was achieved by selecting the 28 clusters within a 5 declination strip centred on . Details of the clusters are listed in Table 1. All data are from ACO unless otherwise noted. The columns are as follows:

Column 1: The cluster name.
Columns 2 & 3: The coordinates of the cluster centre.
Column 4: The Abell type; a colon indicates a mean type or uncertain type estimate.
Column 5: The Bautz-Morgan type; a colon means uncertain type.
Column 6: The Abell richness class, R.
Column 7: The Abell distance class, D.
Column 8: The magnitude of the tenth brightest galaxy in the cluster, .
Column 9: The best available redshift. If no redshift has been measured, the one estimated from (by ACO) is shown, indicated by an asterisk (*).
Columns 10 and 11: The pixel value rms in units of mJy/beam of the entire MOST radio image, and that of the central , using iterative 2 clipping on a histogram of image pixel values. When the inner field value was unavailable, the rms of a source-free region near the field centre was obtained, using the AIPS task IMEAN, and used instead. These values are indicated with an asterisk (*).

The image quality for one cluster (A3395) was badly affected by sidelobes from a strong source in A3391 and was not used in the current analysis. Therefore, the sample on which the identification process was carried out contains 27 cluster fields.

Source Fitting

The new source-fitting program VSAD (W. Cotton, 1995, private communication), developed for the NRAO VLA Sky Survey (NVSS), was incorporated into AIPS and used to compile preliminary source lists. VSAD is an enhancement of the standard AIPS task SAD; it fits elliptical Gaussians, as described by Condon (1996), and subtracts the fitted components from the image to form a residual image. It performed well on the majority of MOST images. In particular, for many extended and closely blended sources there was far less evidence of fitting errors than with SAD. All residual images were carefully inspected and in cases where fitting errors were evident, peak positions and flux densities were found using the AIPS verb MAXFIT and integrated flux densities were found using TVSTAT, which integrates the total flux in a polygonal area defined by the user. Whenever the peak flux density values given by VSAD and MAXFIT differed by more than 2% for strong (> 50 mJy) unresolved sources, the VSAD value was replaced by the MAXFIT value. The same was done for integrated flux densities using TVSTAT. This ensured that for strong sources any uncertainty in these quantities from the fitting was smaller than the calibration uncertainty inherent in the MOST, which is typically about 5%. The calibration error was established by comparing the strong sources in 12 pairs of repeat observations from the large overall survey dataset. A larger relative difference was allowed for weaker sources (< 20 mJy) where flux density uncertainties are becoming dominated by noise. Where the positions differed by more than about 20 the MAXFIT positions were preferred.

To ensure reliability of the source lists, a 5 cutoff was applied to the peak flux densities. The value of was chosen conservatively. In most cases, we adopted the larger of the two estimates given in Table 1. In the extreme case of A3391, the smaller full-field estimate was used because the larger inner value was boosted by radial artifacts associated with the strong central source; visual inspection was used to check the reality of catalogued sources in this inner region. All images were similarly inspected to ensure that image artifacts or sidelobe confusion were not mistaken for sources. Where confusion was evident the flux density errors were increased substantially over those derived in the next section.

Uncertainties

The 12 repeat observations mentioned above were used as a check on the uncertainties given by VSAD. It was found that the position differences could be modelled by the combination in quadrature of a systematic (calibration) offset, due to residual pointing errors, and a random scatter. Figure 1 shows that the scatter in both and , after removing the systematic offsets, depends inversely on peak flux density up to some level, above which a constant term dominates. The model can be described by the equation \

\

where is the variance of the differences in RA or Dec for a pair of observations at a given peak flux density; A represents the error due to the overall signal-to-noise, B represents the calibration error, and S is the peak flux density. In general there are approximately equal contributions to A from thermal noise and confusion. However, there are some regions where sidelobe confusion (eg. grating rings) dominates. As mentioned in section 2.2, only those sources deemed to be reliable detections were included in the error analysis. While A and B would normally be obtained directly from a fit to equation (1), in the present case B was set first by the rms scatter among different pairs of strong sources, and A was then found from fitting. For RA differences we find A = 39 arcsec mJy and B = 1.3 arcsec, and for Dec differences cosec() arcsec mJy and B = 1.2cosec() arcsec. The cosec() factor comes from the shape of the synthesised beam, assuming that errors are a fixed fraction of the beamwidth. The position uncertainties and are then given by for a source of peak flux density S. The formal position errors from VSAD were found to be similar to the flux density dependent component of errors determined empirically.

© Copyright Astronomical Society of Australia 1997