The 2dF gravitational lens survey

The process of finding lensed quasars in the 2dF galaxy spectra can be split into three quite distinct phases. The first of these is the identification of candidate lenses on purely spectroscopic grounds. The vast majority of the spectra should be uninteresting, at least in the present application, but a small fraction should stand out as having either unusual continua, or, more likely, what may be broad emission lines. This set of (probably) several thousand spectra will include white dwarfs; unlensed quasars misidentified by the APM software; and, with luck, several lensed quasars. The second phase of the analysis is to examine each spectrum in detail (possibly by eye) with a view to removing all those objects which are clearly not quasars. This step could also profitably include using any other data which is already available, such as images from the Digitized Sky Survey or the Schmidt plates from which the 2dF objects were selected. Finally, the (hopefully) small set of strong lens candidates would have to be imaged at high resolution (< 1 arcsec) with a view to determining their nature morphologically. In terms of quantifiable resources this last step is by far the most expensive, as the automatic spectral analysis should take less than a second per object. However the opposite is true when intellectual effort is considered, as the manual inspection of the initial candidate list and (especially) the imaging observations are conceptually straight-forward. The complex task is determining how best to find a subset of the survey spectra that contains all the lenses and not too much else.

One approach to this problem would be to work backwards - if it is assumed that all the detectable lenses will primarily have broad quasar emission lines then the task reduces to that of identifying the spectra that have emission line-like features. Cross-correlation techniques, combined with continuum removal, could certainly detect some such objects. Unfortunately many galaxy spectra also have features that could be confused with the broad lines of a fainter, superimposed spectrum. Further, this is clearly an inefficient procedure, as much of the available information remains unused. Finally, such a method becomes increasingly difficult once any attempt is made to include non-Gaussian effects such as variation in the continuum levels and the presence of sky lines.

The method that has been adopted is based on Monte Carlo simulations of a large number of lenses. Using the formalism described in Mortlock & Webster (2000) it is possible to generate deflector-source pairs that reflect the variation in lensing likelihood with source redshift and deflector properties. A galaxy-quasar pair generated in this way is a composite object with only one degree of freedom: the relative contribution of the two components to the resultant spectrum. This is very sensitive to the properties of the lens, but such details need not be treated explicitly as the range of possible weightings can be analysed without knowing the actual image configurations. The most important aspect of this method is that, for a given galaxy-quasar pair, a very realistic composite spectrum can be generated by simply combining real 2dF galaxy and quasar spectra. The noise properties of the spectra are certainly correct provided both spectra are kept in their observed frame; the one possible important omission is reddening of the background quasar by dust in the lens. This too can be included in the simulations, but it would incur the penalty of parameterisation; this process is otherwise model-independent. A promising candidate for the analysis method is principal components analysis (PCA; Murtagh & Hecht 1987), which is already being used extensively within the 2dF collaboration, and is well suited to the task at hand. The survey galaxies cover a definite region in the multi-dimensional space of component coefficients (e.g., Folkes et al. 1999) and the lenses should inhabit a contiguous region, with the distance from the galactic locus increasing with the brightness of the quasar component. Unfortunately this process has not yet been implemented; the central ideas are quite clear, but quantitative estimates of the completeness and efficiency of the lens survey cannot yet be made.

© Copyright Astronomical Society of Australia 1997