A multicolour search for quasars
Mitchell, Paul Simon
The Edinburgh Multicolour Survey is a dataset of approximately 1.3 million images covering a contiguous area of 0.1 steradians at high Galactic latitudes. These data are derived from 130 UK Schmidt photographic plates taken in passbands U, B, V, R and I in two strips at declinations -5°and 0°, comprising the standard UKST fields 789-794 and 861-867 respectively. The aim of the survey is to produce a dataset containing accurate UBVRI information and morphological classification for every image detected in any waveband in the measured area, enabling the discrimination of intrinsically rare populations of objects with non-stellar colours from the foreground of normal Galactic stars. This thesis is concerned with (a) the development of the dataset from its initiation as raw plate material through to the production of the final five-band catalogues, and (b) the selection of quasars over a wide range in redshift.The plate material was scanned with the COSMOS fast-measuring machine at Edinburgh. Care was taken to reduce to a minimum the number of incorrectly measured images, which would badly contaminate candidate lists, and to minimise photometric errors. The presence of two plates for each field/colour enables the rejection of spurious images which was performed after applying a local coordinate transformation between plates. Each set of UBVRI plates was taken close together in time to allow the correct measurement of the colours of variable stars and the rejection of images varying significantly in colour between epochs. Three different parameters were used to morphologically classify images over the full range in apparent magnitude. The data were calibrated with CCD sequences providing a total of about 30 measurements per UKST field, and zero-pointed using photoelectric measurements of bright (B = 15 —16) stars. Systematic calibration errors outside the well-calibrated magnitude range and field-effects due to variations in image structure with position were minimised by new techniques such that the final photometric accuracy of the mean magnitude in each band is 0.02-0.05 magnitudes at about two magnitudes brighter than each plate-limit. The median depths of the final catalogues are U = 20.7, B = 20.8, V — 19.6, R = 20.0 and I = 18.4. The survey, complete in all but fields 793 and 794, was used (i) to enable the selection of a new low-redshift sample of 69 quasars using a modification of the UVX technique which improves its efficiency, and (ii) as a basis for the selection of bright quasars at the highest redshifts z > 3.4.The UVX results compare very favourably with two similar samples, producing a corrected surface density of 0.40±0.05deg-2 for quasars with 15 < B < 18 and 0.3 < 2 < 2.2; examination of the distribution of objects in colour space provides further confirmation of the sample’s completeness. Power spectrum analysis and correlation function analysis produce some evidence for the presence of clustering in the3-D distribution on scale sizes up to r ~ 250 fi-1 Mpc; factors capable of producing such a signal spuriously are outlined and the implications of such a result discussed.For the highest redshift quasars, nearest-neighbour analysis was used to locate objects with non-stellar colours in multicolour space. The full range of survey colours increases sensitivity to z — 4.1-4.5 as shown by two approaches: synthesizing quasar colours for a range of spectral type and redshift in a real dataset shows that the efficiency for most quasars with 3.4 < z < 4.5 is very high, and the inclusion of genuine quasars into a survey dataset confirms this conclusion. Three such quasars were confirmed spectroscopically with redshifts 2 = 3.4, 3.5 and 3.7 from a candidate list selected with 17 < R < 18.5, which imply a corrected surface density of 0.02-0.08 deg-2 for3.4 < z < 4.1. This is lower but not significantly different from estimates based on visual emission-line surveys, and implies that the luminosity function for bright quasars has the same amplitude as that at z ~ 2. Various simple parameterisations for the evolution of the optical luminosity function are examined in the light of the new result.