Annu. Rev. Astron. Astrophys. 2000. 38: 667-715 Copyright © 2000 by Annual Reviews. All rights reserved

1. INTRODUCTION

Deep surveys have a long history in optical astronomy. Indeed, nearly every telescope ever built has at some point in its lifetime been pushed to its practical limit for source detection. An important historical motivation for such surveys has been the desire to test cosmological models via the classical number-magnitude and angular-diameter-magnitude relations [Sandage 1961, Sandage 1988]. However interpretation has always been hampered by the difficulty of disentangling the effects of galaxy selection and evolution from the effects of cosmic geometry. The modern era of deep surveys in optical astronomy began with the advent of CCD detectors, which allowed 4-m telescopes to reach to nearly the confusion limit allowed by the point spread function [Tyson 1988]. These images revealed a high surface density of faint, blue galaxies, and the combination of their number-magnitude relation, number-redshift relation, and angular correlation function was difficult to reconcile with the critical-density universe (_M = _tot = 1) that seemed at the time to be a robust expectation from inflationary models [Kron 1980, Fukugita et al. 1990, Guiderdoni & Rocca-Volmerange 1990, Efstathiou et al. 1991].

Prior to launch, simulations based on fairly conservative assumptions suggested that the Hubble space telescope (HST) would not provide an overwhelming advantage over ground-based telescopes for studies of distant galaxies [Bahcall et al. 1990]:

Our results show that the most sensitive exposures achieved so far from the ground reveal more galaxies per unit area than will be seen by planned HST observations unless galaxy sizes decrease with the maximum rate consistent with ground-based observations. In this, the most favorable case for HST, the space exposures will show almost as many galaxy images as have been observed so far in the most sensitive ground-based data.

It was apparent from the images returned immediately after the HST refurbishment in 1993 that distant galaxies had substantially higher surface brightnesses than predicted by such simulations [Dressler et al. 1994, Dickinson 1995]. Despite the pessimistic predictions, extensive field galaxy surveys were always a top priority for HST, and large amounts of observing time were awarded to deep, pointed surveys and to the medium-deep survey [Griffiths et al. 1996]. These observations provided valuable information on galaxy scale lengths and morphologies, on the evolution of clustering and the galaxy merger rate, and on the presence of unusual types of galaxies [Mutz et al. 1994, Schade et al. 1995a, Cowie et al. 1995a, Griffiths et al. 1996]. In this context, the idea of using the HST to do an ambitious deep field survey began to look quite attractive, and was eventually adopted as one of the primary uses of the HST director's discretionary time.