Annu. Rev. Astron. Astrophys. 1997. 35: 389-443 Copyright © 1997 by Annual Reviews. All rights reserved

3.4. Imaging Surveys with Hubble Space Telescope

Given the heroic efforts to resolve the faint blue population from ground-based telescopes (Giraud 1992, Colless et al 1994), one might imagine rapid progress in understanding their nature would be possible from the first refurbished HST images that became available in 1994. In practice, interpretation of the HST data has been hindered by a number of factors introduced in Section 2. The following discussion concentrates on post-refurbishment data. Discussion of the HDF is below in Section 5.

Two survey techniques have been used to image the faint population. The MDS (Griffiths et al 1994) has utilized parallel WFPC-2 data that sample random high-latitude fields in F785W or F814W and, where possible, F555W. The typical primary exposure time has seriously limited long parallel exposures and thus the bulk of the useful data comes from exposures of 1 h in F814W. The primary advantage of the MDS is the total survey field area. Morphological data and image parameters have been presented to I₈₁₄ = 22-23 (Glazebrook et al 1995b, Driver et al 1995a) for over 300 sources to I₈₁₄ = 22 within 13 WFPC-2 fields (0.02 deg²). Driver et al (1995b) also analyze 227 galaxies from a single deeper pointed exposure of 5.7 h to I₈₁₄ = 24.5. The main disadvantage of the MDS is that the small field of view of each WFPC-2 image is poorly matched to ground-based multiobject spectrographs, and thus follow-up redshift work is rather inefficient. The MDS survey is thus best viewed as a deep 2-D survey of the faint sky. Windhorst et al (1996) provides a good summary of the overall results. The alternative technique discussed by Ellis (1995), Schade et al (1995), Cowie et al (1995a, b), Koo et al (1996), LeFevre et al (1996a) involves taking HST images in primary mode of ground-based redshift survey fields. By arranging WFPC-2 exposures in a contiguous strip, an effective match is obtained with existing redshift data. A variant here is the exploitation of the "Groth strip," a GTO exposure in F606W and F814W that consists of 28 overlapping WFPC-2 fields (Koo et al 1996).

Glazebrook et al (1995b), Driver et al (1995a, b) have classified the MDS and related samples visually into spheroidal/compact, spiral, and "irregular/peculiar/merger" categories. They claimed the number of regular galaxies (spheroidals and spirals) to I = 22-24.5 is approximately as expected on the basis of the local passively evolving populations, whereas the irregular/peculiar/merger population is considerably in excess of expectations with a much steeper count slope. Abraham et al (1996b) derived the same conclusion from an automated treatment of image morphology based on the concentration of the galaxy light (which correlates closely with the bulge/disk ratio) and the asymmetry (used to locate irregulars). Odewahn et al (1996) investigate the use of artifical neural networks to classify the data and reach similar conclusions. Note that the latter techniques do not yet, as presented, include corrections for bandpass shifting biases. In each of these studies, galaxies were located after smoothing the MDS images to ground-based resolution to avoid double-counting close pairs of galaxies, and the conclusions were based on the increased normalization of the LF discussed earlier.

It is tempting to connect the rapidly evolving blue galaxies in the redshift surveys with the irregular/peculiar/merger systems seen in the MDS data (Ellis 1995), but central to the MDS results is the physical interpretation of the irregular/peculiar/merger class. Could this category of objects not simply be an increasing proportion of sources rendered unfamiliar by redshift or other effects? A convincing demonstration that this is not the case requires a detailed analysis of the inferred morphology, when viewed at likely redshifts, of a representative multicolor CCD sample of nearby galaxies of known type. Although the machinery is available to conduct such simulations (Abraham et al 1996b) (Figure 4), it is not yet clear whether the available local samples are properly represented in all classes. However, thus far it seems unlikely such a large bias could occur within the redshift range appropriate for I < 22 samples as defined from the CFRS, because the F814W images correspond typically to rest-frame B. However, the precise distinction between late-type spiral and irregular/peculiar/merger may remain uncertain (Ellis 1996a).

Within the limited sample sizes available through HST images of the redshift survey galaxies, the identification of such a high fraction of irregular galaxies is less clear. Schade et al (1995) studied 32 CFRS galaxies with z > 0.5 using WFPC-2 images in F450W and F814W and commented on the high proportion (30%) of blue nucleated galaxies. Many appear asymmetric and some shows signs of interaction. Cowie et al (1995a, b) discussed the unfamiliar nature of their distant sample and, for the faintest bluest sources with z > 1, introduced the terminology of chain galaxies - multiple systems apparently merging along one dimension (but see Dalcanton & Schectman 1996). The varied interpretation of these HST images again underlines the difficulty of connecting high redshift morphology with local ground-based data. Possibly the most robust conclusion so far is that the bulk of the faint blue population are galaxies comparable in size to the remainder of the population, a result that is apparent in earlier ground-based efforts (Colless et al 1994). Few are truly compact systems of the kind selected for detailed study by Koo et al (1995), Guzman et al (1996).