It would appear from the previous discussion concerning the likely properties of PGs that a search covering only a limited area of sky, but reaching to faint flux limits, most closely represents the optimum search strategy. The high number density of PGs expected also ensures that one part of the sky will be pretty much as good as any other in terms of the likelihood of detection - although clearly avoiding regions of high galactic extinction and high stellar density, such as the plane of the Milky Way or towards the Large Magellanic Clouds, is advisable. PG searches have been carried out for more than 20 years and have always used the largest, and therefore most sensitive, telescopes and accompanying instruments to conduct such surveys. The observations have always been difficult and the story is one of astronomers, goal-oriented in their quest to uncover the earliest population of galaxies, only making significant observational inroads when accompanying technological advances in telescope or instrument design take place. The first serious searches for PGs, carried out in the 1970s, utilised the large areal coverage obtainable with photographic plates (Partridge 1974), although it was only when optical charge-coupled device (CCD) technology was developed, in the early 1980s, that really useful limits were first reached.
5.1 Optical continuum and emission-line searches
A number of experiemnts have been carried out designed to detect the emission from redshifted Lyman - possibly the strongest single emission feature from young galaxies. One group of these observations utilises the technique of narrow-band imaging: a CCD image taken with a filter centred on redshifted Lyman with a passband well-matched to the expected width of the emission line ( 20 Å), greatly enhances the contrast of the line against the continuum of the galaxy and, more importantly, against the terrestrial sky-emission. In this way Pritchet and Hartwick (1987) set strong limits on the space density of PGs at 4.5 < z < 6 from narrow-band imaging at wavelengths 6000 8000 Å using the Canada-France-Hawaii 3.6m telescope in Hawaii, while results on the same telescope by De Propris et al. (1993) using a system of narrow-band filters covering 4000 5000 Å found upper limits on Lyman emission that are inconsistent with the simple model predictions discussed above by a factor of about 10 at redshifts 2 z 3.
The other technique designed to detect redshifted Lyman from PGs utilises a grating spectrograph incorporating a long object slit to maximise sky coverage during the observation. While the sky coverage obtainable is significantly less than that of the narrow-band technique, the advantages are that a significantly larger wavelength range can be observed simultaneously and the sky background can be subtracted accurately. For example, Thompson and Djorgovski (1995) used spectroscopic data from the 200 inch Hale telescope at the Palomar Observatory, obtained for other purposes over the course of a 7-year period, to search for serendipitous Lyman emission over the wavelength range 5000 7500 Å and covering a total area of 15 arcmin2 of sky (sufficient to contain at least 20 PGs) - placing limits on the SFR in PGs of 100 M yr-1.
Accompanying experimental limits on the brightness of the extragalactic background light at optical wavelengths are also consistent with the lack of success in detecting individual PGs at optical wavebands, helping to rule out the existence of a substantial PG population at z < 5 (e.g. Dube et al. 1977).